From vxwexplo-errs@csg.lbl.gov Tue Oct 1 09:25:32 2002 From: Bruce Wilson Date: Tue Oct 1 09:25:34 PDT 2002 Subject: Re: Target shell command line limitation Hi Kevin, I searched in google ("line length group:comp.os.vxworks"), and found the following answers to your question about increasing the cmd line limit of 128 chars: (1) the limit is hardcoded; get WRS to compile a special version (2) store your long input into a string and pass a ptr to the string variable The string ptr worked well for dealing with long pathnames, but might not work for your situation. --Bruce Wilson Lawrence Livermore Nat Lab Livermore, CA bawilson@llnl.gov >Submitted-by vxwexplo-errs@csg.lbl.gov Mon Sep 30 18:37:08 2002 >Submitted-by: Kevin Tsubota > > >Hi, > >I'm running VxWorks 5.3.1 on MV230x CPUs. > >I have a generic application which I instantiate multiple times >using many name=value arguments. I've run into a target shell >command line limit of 128chars. > >Is there an easy way to increase the target shell command line length? > >Thank you. > >Kevin Tsubota >W.M. Keck Observatory > >********** From vxwexplo-errs@csg.lbl.gov Wed Oct 2 09:31:57 2002 From: Atkinson Andrew Date: Wed Oct 2 09:31:58 PDT 2002 Subject: CodeTest/Coverage vs. CoverageScope - Hi All We are looking at purchasing a coverage analysis tool for our project - has anyone had experiences (good or bad) with either of these tools? Any info appreciated. vxworks Andy Atkinson Software Engineer Bldg 31a QinetiQ, Bincleaves Technology Park Weymouth Dorset DT4 8UR Tel : 01305 764660 Fax : 01305 764661 e-mail : AATKINSON1@QinetiQ.com The Information contained in this E-Mail and any subsequent correspondence is private and is intended solely for the intended recipient(s). For those other than the recipient any disclosure, copying, distribution, or any action taken or omitted to be taken in reliance on such information is prohibited and may be unlawful. From vxwexplo-errs@csg.lbl.gov Wed Oct 2 13:39:05 2002 From: "John W Cosgrove" Date: Wed Oct 2 13:39:07 PDT 2002 Subject: tftp and interface addresses Oops, I forgot the magic words.... vxWorks 5.4, DY-4 svme-179+ board and DY-4 (Ixthos) IA7 PPC quad (5?) boards. John ---------------------- Forwarded by John W Cosgrove/RES/Raytheon/US on 10/02/2002 04:23 PM --------------------------- From: John W Cosgrove on 09/30/2002 08:29 AM To: vxwexplo@lbl.gov cc: Subject: tftp and interface addresses I have an application that runs on 4 identical boards, which uses a dummy IP address to perform a bootp request which returns the correct IP and the initial file to load. The board then setshis IP and attempts to load the file named in the bootp response. My investigation shows that the tftp server is receiving the tftp request, but from the original IP address, not the one I have just set the interface to. How do I fix this? TIA, John Cosgrove (401)842-4167 jwc@ssd.ray.com john_w_cosgrove@raytheon.com From vxwexplo-errs@csg.lbl.gov Wed Oct 2 19:40:11 2002 From: "narayananl" Date: Wed Oct 2 19:40:13 PDT 2002 Subject: Text Relocation Overflow in vxworks --=_MAILER_ATTACH_BOUNDARY1_2002103482141649760492 Content-Type: text/plain; charset=us-ascii dear all When we try to *build* our application (text segment size is abt 2MB) in vxworks5.4( Tornado2.0) we face ".text relocation overflow" problem during the partialImage.o linking. Since we use ARM processor and COFF file format, used a linker option "-split-by-reloc 65535" (or 65000). also tried with "-traditional-format" linker option but the problem remains same. when i searched through the vxworks archieve(FAQ) found a compiler option for powerPC such as -mlongcall (or -mlong-call) but i doubt any such option is available for ARMcompilers (ccarm) ? when excluded few modules from my application then the build succeeds and vxWorks image is created. But it is been very difficult(merely not possible) for me to exclude any of the modules from our application. also come across a suggestion of upgrading to Tornado2.2 which uses Elf ARM toolchain which i guess is a complete rework to be done! my questions are: 1) Is this problem solvable in Tornado2.0 ? how ? 2) Is there anyother solution ? 3) how to change the file format (COFF to ELF etc) i understand this could be a common problem posted many times into this list but any help in this regard is highly appreciated. Thanks in advance , lakshminarayanan r -------------------------------------------build capture start----------------------------------------------------- ccarm -g -mcpu=arm7tdmi -mapcs-32 -mno-thumb-interwork -ansi -nostdinc -DRW_MULTI_THREAD - D_REENTRANT -mno-sched-prolog -fno-builtin -fvolatile -IC:\Tornado\target\proj\Project1 -I C:\Tornado\target\config\sbcARM7 -IC:\Tornado\target\h -IC:\Tornado\target\config\comps\sr c -IC:\Tornado\target\src\drv -DCPU=ARM7TDMI -DPRJ_BUILD -Wcomment -g -c C:\Tornado\target \config\comps\src\version.c ccarm -nostdlib -r -Wl,-X \ -Wl,@C:\Tornado\target\proj\Project1\prjObjs.lst C:\Tornado\target\config\sbcARM7\sngk s32cEnd.obj C:\Tornado\target\proj\future\LR\FutureDSL.o version.o C:\Tornado\target\lib\ libARM7TDMIgnuvx.a C:\Tornado\host\x86-win32\lib\gcc-lib\arm-wrs-vxworks\2.7.9-970819\libg cc.a -o partialImage.o C:\Tornado\host\x86-win32\lib\gcc-lib\arm-wrs-vxworks\2.7.9-970819\..\..\..\..\arm-wrs-vxw orks\bin\\ld.exe: partialImage.o: .text: reloc overflow: 0x1012e > 0xffff partialImage.o: final close failed: File truncated collect2: ld returned 1 exit status ccarm: Internal compiler error: program ld got fatal signal 1 make: *** [partialImage.o] Error 0x1 Done. -------------------------------------------build capture end----------------------------------------------------- Get Your Private, Free E-mail from Indiatimes at http://email.indiatimes.com Buy Music, Video, CD-ROM, Audio-Books and Music Accessories from http://www.planetm.co.in Change the way you talk. Indiatimes presents "Valufon", Your PC to Phone service with clear voice at rates far less than the normal ISD rates. Go to http://www.valufon.indiatimes.com. Choose your plan. BUY NOW. --=_MAILER_ATTACH_BOUNDARY1_2002103482141649760492 Content-Type: text/html; charset=us-ascii dear all

When we try to *build* our application (text segment size is abt 2MB) in vxworks5.4(
Tornado2.0)  we face ".text relocation overflow" problem during the partialImage.o linking.

Since we use ARM processor and COFF file format, used a linker option "-split-by-reloc 65535" (or 65000). also tried with "-traditional-format" linker option but the problem remains same.

when i searched through the vxworks archieve(FAQ) found a compiler option for powerPC such as -mlongcall (or -mlong-call) but i doubt any such option is available for ARMcompilers (ccarm) ?

when excluded few modules from my application then the build succeeds and vxWorks image is created. But it is been very difficult(merely not possible) for me to exclude any of the modules from our application.

also come across a suggestion of upgrading to Tornado2.2 which uses Elf ARM toolchain which i guess is a complete rework to be done!

my questions are:

1) Is this problem solvable in Tornado2.0 ? how ?

2) Is there anyother solution ?

3) how to change the file format (COFF to ELF etc)

i understand this could be a common problem posted many times into this list

but any help in this regard is highly appreciated.

Thanks in advance ,

lakshminarayanan r

-------------------------------------------build capture start-----------------------------------------------------

ccarm -g -mcpu=arm7tdmi -mapcs-32 -mno-thumb-interwork -ansi -nostdinc -DRW_MULTI_THREAD -

D_REENTRANT -mno-sched-prolog -fno-builtin -fvolatile -IC:\Tornado\target\proj\Project1 -I

C:\Tornado\target\config\sbcARM7 -IC:\Tornado\target\h -IC:\Tornado\target\config\comps\sr

c -IC:\Tornado\target\src\drv -DCPU=ARM7TDMI -DPRJ_BUILD -Wcomment -g -c C:\Tornado\target

\config\comps\src\version.c

ccarm -nostdlib -r -Wl,-X \

-Wl,@C:\Tornado\target\proj\Project1\prjObjs.lst C:\Tornado\target\config\sbcARM7\sngk

s32cEnd.obj C:\Tornado\target\proj\future\LR\FutureDSL.o version.o C:\Tornado\target\lib\

libARM7TDMIgnuvx.a C:\Tornado\host\x86-win32\lib\gcc-lib\arm-wrs-vxworks\2.7.9-970819\libg

cc.a -o partialImage.o

C:\Tornado\host\x86-win32\lib\gcc-lib\arm-wrs-vxworks\2.7.9-970819\..\..\..\..\arm-wrs-vxw

orks\bin\\ld.exe: partialImage.o: .text: reloc overflow: 0x1012e > 0xffff

partialImage.o: final close failed: File truncated

collect2: ld returned 1 exit status

ccarm: Internal compiler error: program ld got fatal signal 1

make: *** [partialImage.o] Error 0x1

Done.

-------------------------------------------build capture end-----------------------------------------------------
Get Your Private, Free E-mail from Indiatimes at http://email.indiatimes.com
Buy Music, Video, CD-ROM, Audio-Books and Music Accessories from http://www.planetm.co.in
Change the way you talk. Indiatimes presents "Valufon", Your PC to Phone service with clear voice at rates far less than the normal ISD rates. Go to http://www.valufon.indiatimes.com. Choose your plan. BUY NOW. --=_MAILER_ATTACH_BOUNDARY1_2002103482141649760492-- From vxwexplo-errs@csg.lbl.gov Sat Oct 5 01:21:31 2002 From: youzen@263.net Date: Sat Oct 5 01:21:32 PDT 2002 Subject: Re: relocation overflow dp1234 has forwarded a link to a post entitled 'Re: relocation overflow' on 'Dsicussion Board'! The post is below. You have received a notification about this post from dp1234: dear all When we try to *build* our application (text segment size is abt 2MB) in vxworks5.4( Tornado2.0) we face ".text relocation overflow" problem during the partialImage.o linking. Since we use ARM processor and COFF file format, used a linker option "-split-by-reloc 65535" (or 65000). also tried with "-traditional-format" linker option but the problem remains same. when i searched through the vxworks archieve(FAQ) found a compiler option for powerPC such as -mlongcall (or -mlong-call) but i doubt any such option is available for ARMcompilers (ccarm) ? when excluded few modules from my application then the build succeeds and vxWorks image is created. But it is been very difficult(merely not possible) for me to exclude any of the modules from our application. also come across a suggestion of upgrading to Tornado2.2 which uses Elf ARM toolchain which i guess is a complete rework to be done! my questions are: 1) Is this problem solvable in Tornado2.0 ? how ? 2) Is there anyother solution ? 3) how to change the file format (COFF to ELF etc) 4) May the issue be solved in Tornado2.2 ? i understand this could be a common problem posted many times into this list but any help in this regard is highly appreciated. Thanks in advance , chen The actual post is below Hello, while building our application i face relocation overflow problem. So, i used linker option (-split-by-reloc 65535) and re-built our application. But the problem is not solved. when i searched the FAQ i found -mlongcall(or -mlong-call) compiler option. but when i try to compile with our ccarm the compiler is not recognising the -mlongcall option. guess the reference manual doesnot have this -mlongcall option listed for ARM processor(ccarm) seems this option is supported only for powerPc ? Firstly, this is not the best forum to post questions like this in - you would stand more chance of getting an answer by posting in either the operating system forum, or one of the tools ones. Your problem is a little strange though as the limit for ARM should be the same as for PPC, +/- 32MB from current PC. Since your text must be less than 7MB this should not be a problem. By the way, you should use sizearm to get the size of the text, rather than looking at the file size - the file size will always be larger than the actual image since it contains additional information. The error message seems to be implying that you have a 16 bit limit on the range though. One thing you might want to try would be something like this: 1) From the command line, perform a partial link of all the code apart from the VxWorks libraries: ccarm -nostdlib -r -Wl,-X -Wl,@C:\Tornado\target\proj\Project1\prjObjs.lst C:\Tornado\target\config\sbcARM7\sngks32cEnd.obj C:\Tornado\target\proj\future\LR\FutureDSL.o version.o -o partialImage.o This might also fail, in which case you must discover where in this set of objects the 16 bit relocation is coming from. Try reducing the number of objects linked until you get something that links successfully here. Be warned though, the object(s) you take out could include the cause of the problem, so examine each one for 16 bit relocations using objdumparm --reloc (see step 2). 2) Examine partialImage.o using objdumparm --reloc to see where the 16 bit relocations are (I only have VxWorks AE handy, and that uses ELF, but I hope this option is there for the COFF version of objdumparm too). When I look in my ELF images I actually don't have any 16 bit relative relocations, so I hope you won't have too many to search through. 3) Once you have the list of them, complete with the symbol to which each corresponds, you can try to work out (a) where in your code each reference is coming from, and (b) which one will be too far away to fit 16 bits. This is going to be a slow process I'm afraid, unless you don't have that many 16 bit relative relocations. HTH, John... View this thread on the web at http://www.windriver.com/windsurf/dboard/showthreaded.php?Board=windsurf&Number=803 From vxwexplo-errs@csg.lbl.gov Mon Oct 7 18:34:31 2002 From: "Matthew Peltzer" Date: Mon Oct 7 18:34:32 PDT 2002 Subject: x86 pci interrupt sharing Greetings, I am developing a device driver for custom PCI hardware under vxWorks 5.4/Tornado 2.0.2, and am experiencing difficulties in handling interrupts. I have no problem when the two devices are on their own interrupts, and I have no problem when using a shared interrupt independently. I am not able to have two threads at the same time listen to the same interrupt. According to the manual page for pciIntConnect(), I should not have to care about shared interrupts: DESCRIPTION This routine connects an interrupt handler to the PCI interrupt line(A - D). Link list is created if multiple handlers are assigned to the single PCI interrupt. To me this says that every time I call pciIntConnect(), my function and parameter will be added to a list that when the interrupt vector is triggered, each of my functions will be called with the appropriate parameter. Unfortunately, this is not what I am experiencing, although I am not 100% sure of what I am experiencing. It seems as though the second call to pciIntConnect() effectively wipes out the first parameter, for as soon as an interrupt is generated on this irq, my system deadlocks. I can only guess that this is because I am not turning off the interrupt on the correct card because my parameter is not what I expect it to be. My solution to this is to do my own isr chaining. I call pciIntConnect() and sysIntEnablePIC() only when no other connection has been made for this irq. pciIntDisconnect() and sysIntDisablePIC() are called only when no other device is listening to the irq. This setup has the same flaw as the previous: vxWorks reboots when two tasks are run that listen to a separate PCI card, yet share the same irq. I do not know if my callbacks are being called. the first operation in my callback is a logMsg(), but I am not seeing this. I also have some logMsg()s to tell me if each interrupt is being caught or ignored, which is decided by looking at the interrupt line on the hardware device as specified in the parameter passed into the isr by the system. I have yet to see an interrupt be ignored. This is the code I am using to enable interrupts: pciIntConnect(INUM_TO_IVEC((pci->irq + 0x20)), myDummyISR, list[pci->irq]); sysIntEnablePIC(pci->irq); 'pci' is a structure that among other things contains the irq number as read from the pci configuration space with: pciConfigInByte(pci->busNum, pci->devNum, pci->fncNum, 0x3c, &(pci->irq)); Is there anything that I am missing for irq sharing to work in vxWorks? Has anyone had any luck/mis-luck with similar methods? Matthew Peltzer goose23@spu.edu From vxwexplo-errs@csg.lbl.gov Tue Oct 8 09:59:31 2002 From: "Matthew Peltzer" Date: Tue Oct 8 09:59:32 PDT 2002 Subject: RE: ISR on VxWorks Thanks for you reply. My ISR is very similar to yours, with the only real difference being that my lParam is a pointer to a structure that looks like: struct isr_node { struct isr_node* next; struct my_device dev; }; ...and I have a for loop that iterates through the list, checking for the device that caused the interrupt. I have a list for each irq (0x0-0xf), and pass in the appropriate list for each irq when I connect the ISR. My real problem is when two devices are driving the same irq. I can service them separately on the same irq, and I can service them concurrently on different irqs. But when both cards are generating interrupts on the same irq, vxWorks reboots. Card A is initialized to generate an interrupt every 1/2 second, and Card B generates an interrupt on cue with a trigger function. It does not matter which card I install the interrupt for first, nor which card is 'started' first. As soon as the second card (in the sequence), vxWorks halts and reboots. So, if I start Card A generating interrupts at 2 Hz, as soon as Card B is triggered, vxWorks reboots, of, if I trigger card B ONCE, and then start Card A generating interrupts at 2 Hz, vxWorks reboots. Today I am going to hack out as simple a test program as possible to illustrate this, which I will post as soon as it either works, or demonstrates my problems succinctly. Matthew Peltzer goose23@spu.edu -----Original Message----- From: Joseph Hadzima [mailto:josef508@yahoo.com] Sent: Tuesday, October 08, 2002 5:31 AM To: goose23@spu.edu Subject: ISR on VxWorks When writting the ISR the very first thing the ISR MUST do is check it's associated device to determine if it generated the interrupt. If not it should just return. Shared interrupts are placed on a linked list, and ALL ISRs are called in turn. you may wish to try something like: void myISR( int param )/* show generic structure of a typical VxWorks ISR */ { DEVICE_STRUCT *ds = (DEVICE_STRUCT*)param; /* test to ensure that my device generated an interrupt (PPC)*/ if( !(*DEVICE_CSR & DEVICE_INTBIT_MSK) ) /* did my device really interrupt? */ return; /* service the interrupt */ moveData( DEVICE_FIFO, ds->DataBuffer ); /* make the HW happy (acknolwedge it's interrupt)*/ *DEVICE_CSR |= DEVICE_INT_ACK; /* tell processing task to process data */ semGive( ds->DataRdyFlg ); } HOpe this helps __________________________________________________ Do you Yahoo!? Faith Hill - Exclusive Performances, Videos & More http://faith.yahoo.com From vxwexplo-errs@csg.lbl.gov Tue Oct 15 09:06:04 2002 From: Sean McNeil Date: Tue Oct 15 09:06:05 PDT 2002 Subject: test I have not received any email from the VxWorks list in quite a while. Is the system down because of the upgrade or are things really this quiet? Sean From vxwexplo-errs@csg.lbl.gov Tue Oct 15 20:18:01 2002 From: Alan K Biocca Date: Tue Oct 15 20:18:02 PDT 2002 Subject: Vx Exploder Update The VxWorks Exploder here at Lawrence Berkeley Laboratory was having a problem with some postings from the Newsgroup, and so was not relaying the postings to the mailing list. This has been fixed, and the backlog of articles to email out was 2.3 megabytes. This block of messages was not emailed out due to the large size. Normal daily digests of netnews articles should resume. Thanks for your patience, Alan Biocca AKBiocca at lbl dot gov From vxwexplo-errs@csg.lbl.gov Tue Oct 15 19:57:28 2002 From: Vxworks Exploder Date: Tue Oct 15 19:57:41 PDT 2002 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Tue Oct 15 19:56:58 PDT 2002 Subject: Re: what's wrong with my cache flush function? Subject: Re: Why msgQSend can be called by ISR? Subject: Re: Anyone using Tornado 2.1.1 gcc with optimization.... Subject: Re: ipAttach ipDetach memory leak problem Subject: Re: Continue from the same point after power fail (PPC VxWorks 2)? Subject: Re: Continue from the same point after power fail (PPC VxWorks 2)? Subject: Re: Ip Checksum Subject: Re: tftpCopy disable context switching ? Subject: Re: problem connecting a flahs Block Driver to DosFsLib Subject: Re: tftpCopy disable context switching ? Subject: Re: somebody can show me how to use the function symFindByName() Subject: jungos driver development kit Subject: Re: Need to watch DTR status on a PPC Subject: localtime () Subject: Re: jungos driver development kit Subject: Wind Web Server File Uploads Subject: Ethernet Driver for the MPC8260 FEC ves CPM37 Errata Subject: Re: Design ideeas on shared memories Subject: Re: Design ideeas on shared memories Subject: Re: FTP Subject: Re: localtime () Subject: Command-line compiler problems with Tornado 2.1/NT for MIPS Subject: Re: FTP Subject: Re: Design ideeas on shared memories Subject: Re: tftpCopy disable context switching ? Subject: Re: localtime () Subject: hardware compability Subject: Re: porting vxworks to i386ex custom board Subject: Re: Design ideeas on shared memories Subject: Re: sysMemTop() doesn't quite work... Subject: Re: Strange files appearing on the hard drive Subject: Re: Debug question Subject: Re: jungos driver development kit Subject: How to make ccppc still use user-defined lables after compilation? Subject: Does redirect route in vxworks bring memory leak? Subject: Re: IP_MAX_UNITs on custom built simulator Subject: Re: Why msgQSend can be called by ISR? Subject: Does anyone have a ping program, similar to Sisco's catalyst 2900, available under vxWorks? Subject: ipAttach ipDetach memory leak problem Subject: Re: Wind Web Server - httpBlockPut Subject: Re: Debug question Subject: Re: ipAttach ipDetach memory leak problem Subject: Re: How to trace memory access in VxWorks/Tonado? Subject: how can i solve reboot(0x02)-problem Subject: "Virtual" end driver Subject: Re: tftpCopy disable context switching ? Subject: Re: Command-line compiler problems with Tornado 2.1/NT for MIPS Subject: Re: Continue from the same point after power fail (PPC VxWorks 2)? Subject: Re: How to trace memory access in VxWorks/Tonado? Subject: Re: How to make ccppc still use user-defined lables after compilation? Subject: Re: debug interrupt Subject: Re: Does anyone have a ping program, similar to Sisco's catalyst 2900, Subject: syslog in vxWorks Subject: DHCP client Subject: i have trouble connecting to target server with tornado Subject: Re: syslog in vxWorks Subject: Re: Wind Web Server File Uploads Subject: Re: Ethernet Driver for the MPC8260 FEC ves CPM37 Errata Subject: Re: Changing a password / loginUserDelete() doesn't free memory Subject: Re: Strange files appearing on the hard drive Subject: Re: debug interrupt Subject: Make problem w/Torando II, parameter is incorrect Subject: Re: file level redundancy Subject: Re: wdbSystemSuspend : linker error Subject: 8260 BSP autosizing memory? Subject: Re: file level redundancy Subject: what's wrong with my cache flush function? Subject: FTP Subject: tPortmapd out of memory Subject: Re: DHCP client Subject: Re: i have trouble connecting to target server with tornado Subject: Re: howto: locating vxworks to "alternative" flash address for PowerPC Subject: Re: Continue from the same point after power fail (PPC VxWorks 2)? Subject: Re: 8260 BSP autosizing memory? Subject: How to use a different 'target' directory with Tornado? Subject: Re: Design ideeas on shared memories Subject: Re: Vxworks Drivers for 8260 MCC and HDLC SCCs Subject: Re: Continue from the same point after power fail (PPC VxWorks 2)? Subject: Re: How to access complete snmp ObjectID from get, next, set, text Subject: Need idea for implementing watchdog Subject: new compiler win2k host/x86 target? Subject: Re: hardware compability Subject: Re: sysMemTop() doesn't quite work... Subject: Re: Design ideeas on shared memories Subject: Design ideeas on shared memories Subject: Re: Configureing PCI bus interface on IBM PPC405GP Subject: Re: Wind Web Server - httpBlockPut Subject: Re: Anyone using Tornado 2.1.1 gcc with optimization.... Subject: Re: debug interrupt Subject: Q: Does Tornado 2.1 support BSP of MV2700 or MV2604 Subject: Re: what's wrong with my cache flush function? Subject: Re: Need idea for implementing watchdog Subject: Re: context switch hooks Subject: Re: Need idea for implementing watchdog Subject: Re: ll() on /tgtsvr broken Subject: Re: DHCP client Subject: TCP connection released, but socket not closed Subject: Re: Need idea for implementing watchdog Subject: Re: i have trouble connecting to target server with tornado Subject: Re: TCP connection released, but socket not closed Subject: Re: Who have intel 82559End driver source code? How about it efficiency? Subject: Re: Need idea for implementing watchdog Subject: Re: Need idea for implementing watchdog Subject: Re: Vxworks Drivers for 8260 MCC and HDLC SCCs Subject: messages from interrupt level lost Subject: Re: Need idea for implementing watchdog Subject: Re: significance of value in target Subject: Re: TCP connection released, but socket not closed Subject: Re: significance of value in target Subject: Re: Design ideeas on shared memories Subject: Re: C++ Unresolved Symbols Subject: Unsupported FTP Server Subject: Can't boot MV2700 over ethernet Subject: Re: "Virtual" end driver Subject: Re: TCP connection released, but socket not closed Subject: Re: ipAttach ipDetach memory leak problem Subject: Re: Need idea for implementing watchdog Subject: INUM_TO_IVEC Subject: Re: Can't boot MV2700 over ethernet Subject: Re: Can't boot MV2700 over ethernet Subject: ping question!!! Subject: no carrier? Subject: float point in ARM problem Subject: binutil 2.10.1, gcc 2.95.4 & elfHex problem on Solaris. Subject: Re: Unix equivalents for VxWorks. Subject: Boot from IDE Flash Disk with x86 target Subject: Re: TCP connection released, but socket not closed Subject: Re: ata-dma on vxworks Subject: How Can I send raw data in vxworks through tty driver ? Subject: examples Subject: error massage whlie trying to download.. Subject: Re: How Can I send raw data in vxworks through tty driver ? Subject: Re: examples Subject: Re: "Virtual" end driver Subject: tt Subject: Re: Wind Web Server - httpBlockPut Subject: Re: tt Subject: Set RTS and DTR on MVME2700 Subject: Re: INUM_TO_IVEC Subject: Re: error massage whlie trying to download.. Subject: Thread safe in math.h Subject: Re: File System Problem Subject: Re: float point in ARM problem Subject: Re: Boot from IDE Flash Disk with x86 target Subject: Re: debug interrupt Subject: Re: INUM_TO_IVEC Subject: Re: Thread safe in math.h Subject: Use Parallel Port instead of the required Serial Port ? Need Null Modem line ? Subject: Re: "Virtual" end driver Subject: [Q] ibm ppc 405gp / dec21x4x / cacheDmaMalloc / malloc Subject: Re: examples Subject: Re: examples Subject: Re: examples Subject: Re: newsgroup spreading virus Subject: VxWorks 5.1 and tar Subject: Re: Use Parallel Port instead of the required Serial Port ? Need Null Modem line ? Subject: reading registers from the ixf440 using the SA on ixp1200eb Subject: Re: newsgroup spreading virus Subject: Re: newsgroup spreading virus Subject: Re: newsgroup spreading virus Subject: Re: newsgroup spreading virus Subject: Re: newsgroup spreading virus Subject: Re: newsgroup spreading virus Subject: Floating point on MBX860 Subject: Re: no carrier? Subject: How Change a TTL Value of ICMP Packet ? Subject: How to include macros inside TORNADOII for vxworks Subject: Re: trouble in ping with a larger packet on ne2000 Ethernet Subject: Re: Porting Tornado 2.0 Subject: Re: elf/ coff/ Motorola S-record/Intel Hex/Bin ? Subject: Soft Interrupts and getch() in VxWorks Subject: Re: Soft Interrupts and getch() in VxWorks Subject: Re: Why msgQSend can be called by ISR? Subject: Source of mips toolchain Subject: Re: no carrier? Subject: waitforever Subject: Re: Source of mips toolchain Subject: Re: How to include macros inside TORNADOII for vxworks Subject: MVME5100-013x versus -016x Subject: Re: waitforever Subject: Re: Porting Tornado 2.0 Subject: Re: Soft Interrupts and getch() in VxWorks Subject: flash boot for MV177 Subject: Re: Boot from IDE Flash Disk with x86 target Subject: Re: Error in boot line Subject: changing the PTE table Subject: OEMs, WE WANT TO PURCHASE OR CONSIGN YOUR VME BOARDS Subject: Re: Use Parallel Port instead of the required Serial Port ? Need Null Modem line ? Subject: Where can i get the documents about DosFs2.0? Subject: Re: elf/ coff/ Motorola S-record/Intel Hex/Bin ? Subject: driver delivery process Subject: Re: Where can i get the documents about DosFs2.0? Subject: Three questions about PPPoE Subject: Re: newsgroup spreading virus Subject: Re: VXWORKS Subject: VxWorks for different PPC. Subject: Regarding System time Subject: cpu usage Subject: Re: VxWorks for different PPC. Subject: Re: trouble in ping with a larger packet on ne2000 Ethernet Subject: Re: Soft Interrupts and getch() in VxWorks Subject: Re: Porting Tornado 2.0 Subject: Re: VxWorks bh() command question? Subject: IDE Flash disk Subject: Basic Command Line Subject: Re: Floating point on MBX860 Subject: Re: Basic Command Line Subject: Re: IDE Flash disk Subject: Re: Floating point on MBX860 Subject: How to seperate data, Subject: difference between pciautoconfiglib.c and pciconfiglib.c Subject: Re: File system shutdown Subject: Re: driver delivery process Subject: how to use the SYSTEM MODE DEBUGGER Subject: Doku for vxPpcLib Subject: Re: Doku for vxPpcLib Subject: Re: VxWorks bh() command question? Subject: Re: Source of mips toolchain Subject: Re: how to execute vxworks shell routines from other programs Subject: Re: Regarding System time Subject: Re: how to use the SYSTEM MODE DEBUGGER Subject: Interrupt vector table Subject: Re: Interrupt vector table Subject: using RngLib w/ my own memory Subject: Re: Three questions about PPPoE Subject: new Subject: mkbootAta Subject: Re: using RngLib w/ my own memory Subject: system mode vs. task mode debugging Subject: Re: newsgroup spreading virus Subject: Re: How to include macros inside TORNADOII for vxworks Subject: How to fix module link order when rebuilding in Tornado? Subject: Re: VxWorks for different PPC. Subject: how to set up WDB connection to 3C905C-TX? Subject: Re: Doku for vxPpcLib Subject: how generate an interrupt with SIMNT? Subject: Re: Doku for vxPpcLib Subject: Re: how generate an interrupt with SIMNT? Subject: Re: how generate an interrupt with SIMNT? Subject: dofference between vxWorks Image and vxWorks.st image Subject: NFS server throughput Subject: Re: dofference between vxWorks Image and vxWorks.st image Subject: where can i find source of real drivers? Subject: is anybody working on ixp1200 Subject: Re: Need idea for implementing watchdog Subject: Re: dofference between vxWorks Image and vxWorks.st image Subject: Re: Need idea for implementing watchdog Subject: absolute variable locations for GNUC Subject: Re: absolute variable locations for GNUC Subject: Re: localtime () Subject: Target shell symbol table disapperance Subject: help required for developing device driver for pcicard Subject: First time to here Subject: Re: Use Parallel Port instead of the required Serial Port ? Need Null Modem line ? Subject: Re: difference between vxWorks Image and vxWorks.st image Subject: Re: Target shell symbol table disapperance Subject: Re: booting from tffs Subject: Re: waitforever Subject: Re: booting from tffs Subject: Re: waitforever Subject: 10Base-T and 100Base-T problems!!! Subject: where can i get the BSP tool Kit? Subject: test Subject: Re: do not want vxworks shell to attach to console by default Subject: Re: where can i get the BSP tool Kit? Subject: Re: "Virtual" end driver Subject: Re: 10Base-T and 100Base-T problems!!! Subject: Help: disk cache error Subject: Re: absolute variable locations for GNUC Subject: interrupt with VxSim Subject: error with -C -redirectShell Subject: Re: disk cache error Subject: Re: Source of mips toolchain Subject: Blocking vs Non-Blocking Sockets Subject: Re: Blocking vs Non-Blocking Sockets Subject: Re: trouble in ping with a larger packet on ne2000 Ethernet Subject: Re: USB to USB Subject: Re: trouble in ping with a larger packet on ne2000 Ethernet Subject: Re: booting from tffs Subject: Re: booting from tffs Subject: Re: booting from tffs Subject: Re: Target shell symbol table disapperance Subject: Re: absolute variable locations for GNUC Subject: Re: Interrupt vector table Subject: newsgroup spreading virus Subject: Re: do not want vxworks shell to attach to console by default Subject: Re: booting from tffs Subject: elf/ coff/ Motorola S-record/Intel Hex/Bin ? Subject: Detecting is a key has been pressed Subject: Re: Source of mips toolchain Subject: Re: Detecting is a key has been pressed Subject: Re: Detecting is a key has been pressed Subject: VMIC 6001 with Vxworks Subject: 300M8260 Subject: mimicking pSOS event group call Subject: Help on a Page Fault under Intel Pentium CPU Subject: Re: Basic Command Line Subject: can load .out files off network but not TFFS Subject: booting from tffs Subject: Re: booting from tffs Subject: Re: Basic Command Line Subject: Re: Detecting is a key has been pressed Subject: Problem about TCP link under vxWorks 5.4 Subject: Re: mimicking pSOS event group call Subject: PPPoE quesion Subject: Re: DosFs2 chkdsk error message Subject: Re: disk cache error Subject: Re: DosFs2 chkdsk error message Subject: Re: VxWorks Target Scripts - Subject: Re: mimicking pSOS event group call Subject: I changed exception handler, but how to restore it. Subject: anyone out there to help me plz?very urgent help requested Subject: Does any know what "panic : soclose : NOFDREF" mean ........??? Subject: Re: Basic Command Line Subject: Re: anyone out there to help me plz?very urgent help requested Subject: Re: anyone out there to help me plz?very urgent help requested Subject: Re: Problem about TCP link under vxWorks 5.4 Subject: Re: trouble in ping with a larger packet on ne2000 Ethernet Subject: How to make a large project in the Tornado 2.0 IDE Subject: TCP/ip Subject: Re: How to make a large project in the Tornado 2.0 IDE Subject: Re: TCP/ip Subject: Re: How to make a large project in the Tornado 2.0 IDE Subject: TFFS Subject: Re: Basic Command Line Subject: Re: Basic Command Line Subject: Re: TFFS Subject: Re: TCP/ip Subject: Re: no longer with WRS Subject: Is it the hardware problem or software problem? Samsung + Intersil wireless card Subject: Re: Is it the hardware problem or software problem? Samsung + Intersil wireless card Subject: Re: Does any know what "panic : soclose : NOFDREF" mean ........??? Subject: communication with i/ocards in cpci chassis Subject: Gurus out there please help -- compact flash Subject: Re: Gurus out there please help -- compact flash Subject: Intel Web Pad - Help! Subject: [Q] ppc405, no snoop, cacheDmaMalloc Subject: Re: Gurus out there please help -- compact flash Subject: Target directory listing Subject: Re: Target directory listing Subject: Re: [Q] ppc405, no snoop, cacheDmaMalloc Subject: Re: Gurus out there please help -- compact flash Subject: Regarding Stack Size in Task Spawn Subject: Re: Regarding Stack Size in Task Spawn Subject: where can I have free downloadable Tornado II Subject: Implementation of intHandlerCreate() Subject: Re: Implementation of intHandlerCreate() Subject: Re: Gurus out there please help -- compact flash Subject: Re: VMIC 6001 with Vxworks Subject: Re: Gurus out there please help -- compact flash Subject: Re: where can I have free downloadable Tornado II Subject: Re: Gurus out there please help -- compact flash Subject: Re: Gurus out there please help -- compact flash Subject: Re: Regarding Stack Size in Task Spawn Subject: Re: TCP/ip Subject: gets() skips first char Subject: gcc cross compiler linux-sh3 Subject: netBufLib Subject: Re: [Q] ppc405, no snoop, cacheDmaMalloc Subject: WindView problem Subject: Re: Gurus out there please help -- compact flash Subject: launching the target server Subject: Re: Target directory listing Subject: vmBaseGlobalMapInit takes toooo long! Subject: Re: netBufLib Subject: ifAddrAdd problem Subject: Re: [Q] ppc405, no snoop, cacheDmaMalloc Subject: Re: VMIC 6001 with Vxworks Subject: Re: WindView problem Subject: Re: Target directory listing Subject: Re: vmBaseGlobalMapInit takes toooo long! Subject: Re: gcc cross compiler linux-sh3 Subject: Re: Target directory listing Subject: Re: launching the target server Subject: Re: [Q] ppc405, no snoop, cacheDmaMalloc Subject: Re: [Q] ppc405, no snoop, cacheDmaMalloc Subject: Re: anyone out there to help me plz?very urgent help requested Subject: shell input output Subject: Problem with HTTP Server and SLIP Subject: Re: ifAddrAdd problem Subject: Re: netBufLib Subject: Re: Problem with HTTP Server and SLIP Subject: timer use Subject: Re: timer use Subject: Subroutines and libraries Subject: CORBA/static function problem Subject: Re: Subroutines and libraries Subject: Re: CORBA/static function problem Subject: Re: ipAttach ipDetach memory leak problem Subject: Re: Subroutines and libraries Subject: Rescheduling Problem Subject: Re: VMIC 6001 with Vxworks Subject: Re: Rescheduling Problem Subject: how to implement break point Subject: Re: Problem with HTTP Server and SLIP Subject: Re: netBufLib Subject: FLASH lockup on PPC405GP eval board? Subject: Re: Problem with HTTP Server and SLIP Subject: Re: FLASH lockup on PPC405GP eval board? Subject: Re: netBufLib Subject: Training Subject: Re: VMIC 6001 with Vxworks Subject: Re: Problem with HTTP Server and SLIP Subject: Re: Training Subject: Re: Training Subject: Re: gets() skips first char Subject: Re: Help: disk cache error Subject: visionCLICK Simple vs. Complete/Full RAM tests Subject: Re: netBufLib Subject: USB Host driver Subject: How to connect an ISP to a VME interrupt Subject: Re: How to connect an ISP to a VME interrupt Subject: Re: How to connect an ISP to a VME interrupt Subject: Re: FLASH lockup on PPC405GP eval board? Subject: VxWorks Applications Engineer Subject: Exception number 14? Subject: Hierarchial subproject model and vxworks with tornado Subject: Re: how to implement break point Subject: Athlon XP Subject: TFFS Subject: vxworks clean image building Subject: Re: TFFS Subject: Re: Athlon XP Subject: Re: FLASH lockup on PPC405GP eval board? Subject: Re: Help: disk cache error Subject: Re: Rescheduling Problem Subject: Receiving from sockets Subject: Re: gets() skips first char Subject: Re: Exception number 14? Subject: Diab and visual slickEdit problem Subject: Anyone have WRS GCC sources? Subject: Re: Help: disk cache error Subject: Re: Athlon XP Subject: Re: Exception number 14? Subject: Re: Anyone have WRS GCC sources? Subject: 855 MII and 860 BSP question Subject: Re: Anyone have WRS GCC sources? Subject: Re: Receiving from sockets Subject: Low level hardware tests Subject: Re: vmBaseGlobalMapInit takes toooo long! Subject: Re: Anyone have WRS GCC sources? Subject: Re: Low level hardware tests Subject: Mulitiple END driver's question. (cannot add 3 instances) Subject: Re: "Virtual" end driver Subject: ZNYX 346Q on VxWorks 5.4 Subject: IS THE MACRO REPLACED WHEN COMPILING? Subject: Re: Receiving from sockets Subject: Re: vmBaseGlobalMapInit takes toooo long! Subject: Re: IS THE MACRO REPLACED WHEN COMPILING? Subject: Re: Athlon XP Subject: Re: Mulitiple END driver's question. (cannot add 3 instances) Subject: Re: ZNYX 346Q on VxWorks 5.4 Subject: problems after installing patch SPR68322 Subject: Re: gets() skips first char Subject: Re: gets() skips first char Subject: Multiple network interfaces in simulator Subject: Re: Multiple network interfaces in simulator Subject: Re: disk cache error Subject: Re: Anyone have WRS GCC sources? Subject: VisionCLICK BUG! Subject: resolvGetHostByName Timeout? Subject: Tornado 2 vxSim font problem Subject: Re: FLASH lockup on PPC405GP eval board? Subject: Re: VisionCLICK BUG! Subject: Re: disk cache error Subject: Re: Anyone have WRS GCC sources? Subject: linker problem Subject: Re: Tornado 2 vxSim font problem Subject: Re: 855 MII and 860 BSP question Subject: Re: visionCLICK Simple vs. Complete/Full RAM tests Subject: memory allocation Subject: VxWorks 5.4.2 on Motorola Power PC 82xx Subject: Re: ZNYX 346Q on VxWorks 5.4 Subject: Filename from FILE * or descriptor? Subject: Re: VxWorks 5.4.2 on Motorola Power PC 82xx Subject: Project management in tornado Subject: Re: Anyone have WRS GCC sources? Subject: selection of proessor Subject: Re: disk cache error Subject: Re: memory allocation Subject: Re: memory allocation Subject: Re: Project management in tornado Subject: FTP Daemon doesn't work Subject: Re: FTP Daemon doesn't work Subject: Exception: what's the meaning? Subject: CNF: Embedded Systems Club and Ada UK User Group Autumn Conference Subject: Re: NFS server throughput Subject: Re: NFS server throughput Subject: Flash Invalidate Subject: Re: Flash Invalidate Subject: Help with ZNYX 346Q for vxworks 5.4 Subject: Re: ZNYX 346Q on VxWorks 5.4 Subject: Re: memory allocation Subject: How to read the route table? Subject: Re: selection of proessor Subject: Re: memory allocation Subject: RAM Disk Subject: Re: RAM Disk Subject: writing flash driver in vxworks Subject: VxWorks Applications Engineer Subject: Re: selection of proessor Subject: [Q] netperf UDP_STREAM / 0 Mbps / WIindows XP / dec21x4x Subject: fail to transfer large file using FTP client Subject: Re: writing flash driver in vxworks Subject: Re: Receiving from sockets Subject: how to get "Physical Address" directly from PCI enumeration? Subject: vxWorks compilation-time errors Subject: Re: selection of proessor Subject: Error messages Subject: Re: How to read the route table? Subject: Re: writing flash driver in vxworks Subject: Re: Error messages Subject: Re: how to get "Physical Address" directly from PCI enumeration? Subject: tornado debugger taking lot of time to start Subject: ping problem on FEC port Vxworks 5.4 Subject: Wtx Error 0x1001e (REGISTRY_NAME_NOT_FOUND) Subject: Problems with TFFS with DiskOnChip Subject: dumping static variables? Subject: "Not enough room for program headers" linker script error help! Subject: Re: AVL tree implementation Subject: Re: Problems with TFFS with DiskOnChip Subject: Access to the Software through Tornado IDE Subject: Re: Problems with TFFS with DiskOnChip Subject: sleep for nanoseconds without using POSIX Subject: Re: sleep for nanoseconds without using POSIX Subject: Re: sleep for nanoseconds without using POSIX Subject: Re: sleep for nanoseconds without using POSIX Subject: need help with interrupts a MPC8260 Subject: Re: Access to the Software through Tornado IDE Subject: Re: dumping static variables? Subject: Re: ping problem on FEC port Vxworks 5.4 Subject: Re: how to get "Physical Address" directly from PCI enumeration? Subject: Re: tornado debugger taking lot of time to start Subject: Re: Flash Invalidate Subject: Re: sleep for nanoseconds without using POSIX Subject: Re: Flash Invalidate Subject: Curious Newbie Subject: problem on ftp client , command freeze on PWD Subject: tickGet is too slow ... anything else i can use? Subject: Re: problem on ftp client , command freeze on PWD Subject: Re: sleep for nanoseconds without using POSIX Subject: Re: sleep for nanoseconds without using POSIX Subject: Re: AVL tree implementation Subject: Coldfire 5272 BSP Serial Problem Subject: Re: Coldfire 5272 BSP Serial Problem Subject: AMD Athlon XP Subject: host shell does not resume Subject: Re: tickGet is too slow ... anything else i can use? Subject: Re: vxWorks compilation-time errors Subject: Re: AMD Athlon XP Subject: printf causing WTX Errors. Subject: Re: Wtx Error 0x1001e (REGISTRY_NAME_NOT_FOUND) Subject: Re: "Not enough room for program headers" linker script error help! Subject: Re: need help with interrupts a MPC8260 Subject: Re: sleep for nanoseconds without using POSIX Subject: Strange Interrupt/Rescheduling behavior on PowerPC Subject: Re: problem on ftp client , command freeze on PWD Subject: Re: problem on ftp client , command freeze on PWD Subject: Re: problem on ftp client , command freeze on PWD Subject: Re: problem on ftp client , command freeze on PWD Subject: Re: Coldfire 5272 BSP Serial Problem -- I am silly!! Subject: Re: sleep for nanoseconds without using POSIX Subject: Re: dumping static variables? Subject: UDP recvfrom problem Subject: Re: sleep for nanoseconds without using POSIX Subject: Re: sleep for nanoseconds without using POSIX Subject: Re: sleep for nanoseconds without using POSIX Subject: make me understand this plz. Subject: Re: tickGet is too slow ... anything else i can use? Subject: Re: need help with interrupts a MPC8260 Subject: Re: AMD Athlon XP Subject: Re: tickGet is too slow ... anything else i can use? Subject: Re: Strange Interrupt/Rescheduling behavior on PowerPC Subject: How do I set the tNetTaks stack size Subject: Re: How do I set the tNetTaks stack size Subject: Re: how to get "Physical Address" directly from PCI enumeration? Subject: Re: How do I set the tNetTaks stack size Subject: Re: AMD Athlon XP Subject: Re: make me understand this plz. Subject: Re: AMD Athlon XP Subject: Re: how to get "Physical Address" directly from PCI enumeration? Subject: ANN: PragmaDev Real Time Developer Studion V1.2 Subject: Re: host shell does not resume Subject: Support for /etc/services & set./getservent system calls.. Subject: Re: tickGet is too slow ... anything else i can use? Subject: Re: VxWorks 5.4.2 on Motorola Power PC 82xx Subject: Re: sleep for nanoseconds without using POSIX Subject: Re: sleep for nanoseconds without using POSIX Subject: VxWorks error : WTX Error 0x10140 (API_FILE_NOT_ACCESSIBLE) Subject: Re: Strange Interrupt/Rescheduling behavior on PowerPC Subject: Re: sleep for nanoseconds without using POSIX Subject: watchdog Timer Subject: Re: sleep for nanoseconds without using POSIX Subject: Re: host shell does not resume Subject: Re: watchdog Timer Subject: watchdog timer and tExcTask Queue Subject: Re: watchdog timer and tExcTask Queue Subject: Re: sleep for nanoseconds without using POSIX Subject: Re: sleep for nanoseconds without using POSIX Subject: preprocessed files Subject: Re: sleep for nanoseconds without using POSIX Subject: Re: host shell does not resume Subject: Re: sleep for nanoseconds without using POSIX Subject: why my NTP showing error: S_hostLib_UNKNOWN_HOST? Subject: BSD IP Socket buffer question Subject: Re: sleep for nanoseconds without using POSIX Subject: Re: BSD IP Socket buffer question Subject: why IFF_NOTRAILERS flag is for? Subject: linux->powerpc-vxworks cross compiler Subject: Excessive link/compile times for PPC823 target using Diab 4.4b or 5.0 Subject: A question on obtaining vxworks BSD driver Subject: Re: why IFF_NOTRAILERS flag is for? Subject: Re: sleep for nanoseconds without using POSIX Subject: Re: sleep for nanoseconds without using POSIX Subject: how to use 860 SCC1? Subject: Re: host shell does not resume Subject: Re: why my NTP showing error: S_hostLib_UNKNOWN_HOST? Subject: Re: how to get "Physical Address" directly from PCI enumeration? Subject: Help: traps in VxSim Subject: Re: sleep for nanoseconds without using POSIX Subject: Re: sleep for nanoseconds without using POSIX Subject: Re: sleep for nanoseconds without using POSIX Subject: Problem establishing communication on a PPP link. Subject: Re: how to use 860 SCC1? Subject: Re: A question on obtaining vxworks BSD driver Subject: Exception Handling in vxWorks. Subject: any website on vxworks faq? Subject: Re: File problem with either DOS or TFFS Subject: testing Subject: adding rcs2 space under powerpc Subject: Re: any website on vxworks faq? Subject: muxDevStart()? Subject: disk cache error Subject: SetupSDK Subject: Re: Exception Handling in vxWorks. Subject: Re: sleep for nanoseconds without using POSIX Subject: Queue Placement Subject: re: File problem with either DOS or TFFS Subject: Re: muxDevStart()? Subject: TLB MISS ERROR (Q) Subject: Re: muxDevStart()? Subject: Re: tftpXfer problems :Boot/Loading a VxWorks Image Subject: Re: any website on vxworks faq? Subject: i will join this newsgroup! Subject: Placing Tornado Project files into RCS Subject: Placing Tornado Project Files into RCS Subject: Re: muxDevStart()? Subject: disk cache error (new posting) Subject: Re: Placing Tornado Project files into RCS Subject: Exception when booting up a standalone image Subject: Re: How Change a TTL Value of ICMP Packet ? Subject: Re: TLB MISS ERROR (Q) Subject: Re: muxDevStart()? Subject: Re: any website on vxworks faq? Subject: symSyncLib problems Subject: muxDevStart() problems! Subject: intLock + printf Subject: Re: Exception Handling in vxWorks. Subject: Re: intLock + printf Subject: Re: Exception Handling in vxWorks. Subject: Looking for a SNMPv2 Trap example Subject: Re: any website on vxworks faq? Subject: muxDevStopAll() Subject: equivalent of unix api' getdomainname(), setdomainname() Subject: Just received: Tornado 2.2 Subject: Re: symSyncLib problems Subject: Re: Exception when booting up a standalone image Subject: Re: Just received: Tornado 2.2 Subject: windview and trigger tool doubts Subject: Can't launch target server Subject: Re: TCP read problem in VxWorks ......! Subject: Re: Can't launch target server Subject: 1 second becomes 4 seconds! Subject: Re: Just received: Tornado 2.2 Subject: Vxworks and Jungo? Subject: Limitation on number of UDP ports Subject: Re: 1 second becomes 4 seconds! Subject: Re: Just received: Tornado 2.2 Subject: Re: Limitation on number of UDP ports Subject: Re: VxWorks Applications Engineer Subject: Re: intLock + printf Subject: Re: intLock + printf Subject: anyone out there who can help me on windview and trigger tool? Subject: Re: 1 second becomes 4 seconds! Subject: Re: anyone out there who can help me on windview and trigger tool? Subject: Re: 1 second becomes 4 seconds! Subject: Re: 1 second becomes 4 seconds! Subject: SBS VG4 + bootable + flash = undefined references Subject: Re: Vxworks and Jungo? Subject: Is it necessary to program the bootloader on different chip than the application or TFFS? Subject: Target Sever and Target Agent? Subject: about bsp makefile under diab compiler Subject: RTL8029 Subject: Re: 1 second becomes 4 seconds! Subject: Re: 1 second becomes 4 seconds! Subject: Re: anyone out there who can help me on windview and trigger tool? Subject: Newbie needs help Subject: Re: Is it necessary to program the bootloader on different chip than the application or TFFS? Subject: Re: 1 second becomes 4 seconds! Subject: Re: Intel chipset 815E Subject: How to generate rom images Subject: Printf Vs. FPrintf Subject: Re: Printf Vs. FPrintf Subject: Re: network task on vxworks Subject: endPoolShow shows 0 free clusters Subject: Re: endPoolShow shows 0 free clusters Subject: Re: Limitation on number of UDP ports Subject: Re: Target Sever and Target Agent? Subject: Re: 1 second becomes 4 seconds! Subject: Loading object module from Disk on Chip Subject: Re: Is it necessary to program the bootloader on different chip than the application or TFFS? Subject: obsolete file descriptor Subject: Re: endPoolShow shows 0 free clusters Subject: Re: about bsp makefile under diab compiler Subject: Booting Subject: Multiple initialization problems with Tornado2.2/Gnu2.96 Subject: Re: TLB MISS ERROR (Q) Subject: Select() on changing fdset Subject: Re: Select() on changing fdset Subject: Re: Exception Handling in vxWorks. Subject: Re: endPoolShow shows 0 free clusters Subject: Re: Select() on changing fdset Subject: How to set a breakpoint when the content of EAX is equal to specified value?? Subject: Help for DosFs in vxworks Subject: Strang Stack Items, Need Help!!! Subject: Re: 1 second becomes 4 seconds! Subject: Re: How to generate rom images Subject: Re: Strang Stack Items, Need Help!!! Subject: bootrom_uncmp build fails on adding #defs for tsfs on pcpentium Subject: Re: Exception Handling in vxWorks. Subject: Re: Multiple initialization problems with Tornado2.2/Gnu2.96 Subject: Re: Select() on changing fdset Subject: deciphering values from windview and trigger tool usage Subject: Re: obsolete file descriptor Subject: Re: Help for DosFs in vxworks Subject: EST backend for MDP823 & Tornado Subject: Re: Select() on changing fdset Subject: Re: Strang Stack Items, Need Help!!! Subject: differences between LogMsg() and printf() functions Subject: Re: differences between LogMsg() and printf() functions Subject: linking module Subject: Re: differences between LogMsg() and printf() functions Subject: Re: Select() on changing fdset Subject: Re: Select() on changing fdset Subject: Re: linking module Subject: Re: Select() on changing fdset Subject: using "/pcConsole/0" and "/tyCo/0" simultaneously Subject: Re: linking module Subject: Re: bootrom_uncmp build fails on adding #defs for tsfs on pcpentium Subject: multiple session telnet server code Subject: Re: Multiple initialization problems with Tornado2.2/Gnu2.96 Subject: Re: multiple session telnet server code Subject: Re: endPoolShow shows 0 free clusters Subject: Convert GNU Make to vxWorks Project ?? Subject: Re: 1 second becomes 4 seconds! Subject: Re: Newbie needs help Subject: CrossWind bug viewing static variables? Subject: Maximum number for software breakpoints?? Subject: Has bash ever been ported to vxWorks? Subject: Re: anyone out there who can help me on windview and trigger tool? Subject: ASM macro in PPC860 with results in args Subject: Re: Select() on changing fdset Subject: Location of exception vectors in PPC BSPs Subject: udpstatShow question Subject: Re: Can't launch target server Subject: Re: bootrom_uncmp build fails on adding #defs for tsfs on pcpentium Subject: Re: network task on vxworks Subject: Strang Stack Items, Need Help!! Subject: semGive failed with errno of S_semLib_INVALID_OPERATION. Subject: Re: Location of exception vectors in PPC BSPs Subject: [Q] strange endDevTbl Subject: Re: 1 second becomes 4 seconds! Subject: Re: Strang Stack Items, Need Help!! Subject: sntpcTimeGet Subject: Re: Strang Stack Items, Need Help!! Subject: Re: endPoolShow shows 0 free clusters Subject: Re: Target Sever and Target Agent? Subject: Re: Booting Subject: Re: Select() on changing fdset Subject: Re: semGive failed with errno of S_semLib_INVALID_OPERATION. Subject: Re: Location of exception vectors in PPC BSPs Subject: Re: Location of exception vectors in PPC BSPs Subject: Re: Strang Stack Items, Need Help!! Subject: Using GNAT pro with VxWorks Subject: ADS8260 BSP Subject: help with vision ICE Subject: Re: Select() on changing fdset Subject: Booting from Flash Disk Subject: Newbie Question (POSIX queues) Subject: Re: Location of exception vectors in PPC BSPs Subject: dosFs2 Subject: double linked list Subject: Re: dosFs2 Subject: Re: double linked list Subject: Re: ADS8260 BSP Subject: Re: ASM macro in PPC860 with results in args Subject: Is there a limit on the number of set breakpoints?? Subject: Re: Filename from FILE * or descriptor? Subject: Re: Curious Newbie Subject: Re: Problem establishing communication on a PPP link. Subject: Re: /tyCo/2 does not work Subject: Re: Booting from Flash Disk Subject: Re: Select() on changing fdset Subject: RTI NDDS experience Subject: task trace during exception Subject: Layer 2 switch Subject: Re: semGive failed with errno of S_semLib_INVALID_OPERATION. Subject: Re: semGive failed with errno of S_semLib_INVALID_OPERATION. Subject: Re: Newbie Question (POSIX queues) Subject: Re: ADS8260 BSP Subject: Re: Booting from Flash Disk Subject: Re: semGive failed with errno of S_semLib_INVALID_OPERATION. Subject: Re: Strang Stack Items, Need Help!! Subject: Re: task trace during exception Subject: modf() function problem Subject: Re: ADS8260 BSP Subject: Re: Booting from Flash Disk Subject: protocol stack problem Subject: Re: Quad Uart Driver Subject: Re: help with vision ICE Subject: Re: Booting from Flash Disk Subject: Re: Strang Stack Items, Need Help!! Subject: semTake()/exit()/taskSpawn() Subject: Re: disk cache error Subject: Re: Newbie Question (POSIX queues) Subject: Re: Newbie Question (POSIX queues) Subject: kernel crash at symFindByValueAndType() function Subject: HELP - access s/w through the Tornado IDE Subject: Re: kernel crash at symFindByValueAndType() function Subject: Re: disk cache error Subject: Configure w/Tornado 2 Win and VxWorks Subject: Re: semTake()/exit()/taskSpawn() Subject: Re: Booting from Flash Disk Subject: Re: ASM macro in PPC860 with results in args Subject: Re: ASM macro in PPC860 with results in args Subject: symTbl.c vs. ELF .stab Subject: Re: symTbl.c vs. ELF .stab Subject: Re: Is it necessary to program the bootloader on different chip than the application or TFFS? Subject: Re: help with vision ICE Subject: bootloader and TrueFFS interaction Subject: Re: help with vision ICE Subject: Re: task trace during exception Subject: Re: Query on vxWorks.st_rom.bin Subject: Re: bootloader and TrueFFS interaction Subject: Wind Web Server and OpenSSL Subject: Re: Query on vxWorks.st_rom.bin Subject: Re: ASM macro in PPC860 with results in args Subject: Re: protocol stack problem Subject: Re: help with vision ICE Subject: Re: help with vision ICE Subject: Re: Wind Web Server and OpenSSL Subject: Fastest way to transfere data between a PC with win2000 and a PC with vxWorks? Subject: Re: help me out Subject: Newbie Starter Kit Subject: Re: Newbie Starter Kit Subject: Re: Fastest way to transfere data between a PC with win2000 and a PC with vxWorks? Subject: Re: Configure w/Tornado 2 Win and VxWorks Subject: Re: ASM macro in PPC860 with results in args Subject: New Object appear at NULL Subject: Re: ADS8260 BSP Subject: Re: ASM macro in PPC860 with results in args Subject: Re: Booting from Flash Disk Subject: Error when updating VxWorks version Subject: Re: Error when updating VxWorks version Subject: Re: Fastest way to transfere data between a PC with win2000 and a PC with vxWorks? Subject: Re: symTbl.c vs. ELF .stab Subject: Re: HELP - access s/w through the Tornado IDE Subject: how to create a bootable VxWorks CDROM Subject: Re: how to create a bootable VxWorks CDROM Subject: Re: Fastest way to transfere data between a PC with win2000 and a PC with vxWorks? Subject: Printf Vs. Fprintf Subject: netMblkToBufCopy and copy penalty Subject: Re: Printf Vs. Fprintf Subject: Re: how to create a bootable VxWorks CDROM Subject: Re: Fastest way to transfere data between a PC with win2000 and a PC with vxWorks? Subject: Tcl for Tornado Subject: Re: New Object appear at NULL Subject: Re: ADS8260 BSP Subject: Re: symTbl.c vs. ELF .stab Subject: Help with lists Subject: Re: Newbie Starter Kit Subject: Re: ADS8260 BSP Subject: Re: Booting from Flash Disk Subject: Re: semTake()/exit()/taskSpawn() Subject: Re: Help with lists Subject: Re: vxworks watch dog Subject: Re: semTake()/exit()/taskSpawn() Subject: Re: Help with lists Subject: Re: Help with lists Subject: Suspended task debugging Subject: The mib2Tbl of END Driver seems doesn't take effect Subject: resolver Subject: Re: binary redistributable Subject: Re: Suspended task debugging Subject: Suspended task debugging Subject: Ada in VxWorks Subject: Re: Ada in VxWorks Subject: Re: symTbl.c vs. ELF .stab Subject: Module loading problems Subject: Re: Ada in VxWorks Subject: Re: resolver Subject: Re: Ada in VxWorks Subject: Re: Fastest way to transfere data between a PC with win2000 and a PC with vxWorks? Subject: Re: Fastest way to transfere data between a PC with win2000 and a PC with vxWorks? Subject: Re: Fastest way to transfere data between a PC with win2000 Subject: Re: vxworks watch dog Subject: Re: semTake()/exit()/taskSpawn() Subject: Re: protocol stack problem Subject: Re: Fastest way to transfere data between a PC with win2000 and a PC with vxWorks? Subject: Re: Fastest way to transfere data between a PC with win2000 and a PC with vxWorks? Subject: Context Switch Time? Subject: Task Reschedule on mutex unlock Subject: Harddisk Copy!! Subject: Re: Harddisk Copy!! Subject: Re: Task Reschedule on mutex unlock Subject: Re: File read problem using shared memory system Subject: Experienced VxWorks Developer Available For Contract Work Subject: The FTP server access Subject: Re: The FTP server access Subject: Image or Copy from a harddisk with vxWorks OS ???!! Subject: Re: dosFs2 Subject: Breakpoints in Single Step with Vision Subject: Re: The FTP server access Subject: signal and taskCreateHookAdd Subject: VxWorks on MCF5206 Subject: Has anyone built binutils from source to be cross platform, was it Subject: Re: Tcl for Tornado Subject: Re: Newbie Starter Kit Subject: Re: Context Switch Time? Subject: Re: dosFs2 Subject: Re: Harddisk Copy!! Subject: Re: Fastest way to transfere data between a PC with win2000 and a PC with vxWorks? Subject: Re: Task Reschedule on mutex unlock Subject: Re: Task Reschedule on mutex unlock Subject: Re: Fastest way to transfere data between a PC with win2000 and a PC with vxWorks? Subject: Re: Task Reschedule on mutex unlock Subject: Re: Task Reschedule on mutex unlock Subject: signal and taskCreateHookAdd Subject: Re: symTbl.c vs. ELF .stab Subject: Re: Task Reschedule on mutex unlock Subject: Re: Task Reschedule on mutex unlock Subject: Getting crazy with VXWSymTab ... Can anyone look at this? ... please! Subject: BSP for TQM8xxL module Subject: Re: Breakpoints in Single Step with Vision Subject: HELP !!! `netJobRing' undeclared ?? Subject: Re: VxWorks on MCF5206 Subject: Re: The FTP server access Subject: Debugger stepping problem Subject: Re: Module loading problems Subject: Re: HELP !!! `netJobRing' undeclared ?? Subject: Re: Getting crazy with VXWSymTab ... Can anyone look at this? ... please! Subject: Does VxWorks offer True SMP / multiprocessor support ? Subject: Re: HELP !!! `netJobRing' undeclared ?? Subject: Re: The FTP server access Subject: Re: BSP for TQM8xxL module Subject: Re: HELP !!! `netJobRing' undeclared ?? Subject: Graphs in Tornado 2.0 Subject: Unable to access other hosts from VxSim using ULIP on win2k Subject: Searching for old MVME 68k boards Subject: pinging the target Subject: UsrNetMuxCfg Subject: easy way to store BSP config in text form? Subject: Re: Newbie Starter Kit Subject: use of bootline Subject: Re: easy way to store BSP config in text form? Subject: Re: easy way to store BSP config in text form? Subject: Re: Vxworks downloadable image size Subject: Re: use of bootline Subject: visionICEII / STK8xxL problem Subject: Re: pinging the target Subject: Re: easy way to store BSP config in text form? Subject: Re: Unable to access other hosts from VxSim using ULIP on win2k Subject: END driver can't receive the packet larger than 60 bytes Subject: Re: pinging the target Subject: Re: END driver can't receive the packet larger than 60 bytes Subject: Re: Getting crazy with VXWSymTab ... Can anyone look at this? ... please! Subject: Re: Does VxWorks offer True SMP / multiprocessor support ? Subject: Re: Graphs in Tornado 2.0 Subject: Re: END driver can't receive the packet larger than 60 bytes Subject: Re: pinging the target Subject: arpresolve: can't allocate llinfo Subject: How can I merge two Tornados into single one based on different BSP? Subject: Re: How can I merge two Tornados into single one based on different BSP? Subject: Cannot capture OSPF packets from IP stack Subject: Re: Unable to access other hosts from VxSim using ULIP on win2k Subject: Re: Vxworks downloadable image size Subject: Re: easy way to store BSP config in text form? Subject: Re: pinging the target Subject: Re: pinging the target Subject: Re: pinging the target Subject: Re: use of bootline Subject: icmpMaskSet() Subject: Re: pinging the target Subject: Re: How can I merge two Tornados into single one based on different BSP? Subject: Re: Does VxWorks offer True SMP / multiprocessor support ? Subject: Re: pinging the target Subject: Re: visionICEII / STK8xxL problem Subject: Re: pinging the target Subject: Re: pinging the target Subject: Re: Filename from FILE * or descriptor? Subject: ARP could not bind to MUX Subject: Re: visionICEII / STK8xxL problem Subject: Job tip Subject: Re: ARP could not bind to MUX Subject: htmlBook utility Subject: Re: arpresolve: can't allocate llinfo Subject: Re: arpresolve: can't allocate llinfo Subject: Re: How can I merge two Tornados into single one based on different Subject: Re: help with vision ICE Subject: Re: How can I merge two Tornados into single one based on different BSP? Subject: Re: How can I merge two Tornados into single one based on different BSP? Subject: Re: arpresolve: can't allocate llinfo Subject: Socket option SO_LINGER? Subject: Telnet Security (using LoginLib) Subject: Re: Socket option SO_LINGER? Subject: Enable cache in BSP for PPC860 Subject: Re: Enable cache in BSP for PPC860 Subject: Building projects fron another project Subject: 4 Dos partitions with tffs question? Subject: Re: Building projects fron another project Subject: req guide for making changes in pid7t bsp for evaluator7t Subject: Task status values Subject: Re: Task status values Subject: source of stackTrace (tt) Subject: Re: source of stackTrace (tt) Subject: Re-download boot strap loader to flash Subject: Re: Building projects fron another project Subject: visionICE software breakpoints in-op Subject: unresolved symbol for semInfo Subject: Problem with vxworks Subject: Re: visionICE software breakpoints in-op Subject: Re: Target shell command line limitation Subject: Stack Trace of current stack? Subject: TNetTask: arp info overwritten for 81026502 by 00:04:76:21:56:34 Subject: Re: TNetTask: arp info overwritten for 81026502 by 00:04:76:21:56:34 Subject: Re: Stack Trace of current stack? Subject: Stack Trace of current stack? Subject: Re: TNetTask: arp info overwritten for 81026502 by 00:04:76:21:56:34 Subject: Problems making boot floppy for T2.2/VxWorks 5.5 Subject: Re: Task Reschedule on mutex unlock Subject: Re: Task Reschedule on mutex unlock Subject: Interfacing NTDS board to vxworks Subject: Instantiating "extern" variables Subject: Chain DMA on GT-64260A Subject: Build specification - Link order Subject: Porting of vxWorks. Subject: Re: Instantiating "extern" variables Subject: Re: Porting of vxWorks. Subject: Tornado II on XP Subject: Re: Stack Trace of current stack? Subject: Re: Instantiating "extern" variables Subject: Re: Instantiating "extern" variables Subject: Re: TNetTask: arp info overwritten for 81026502 by 00:04:76:21:56:34 Subject: Re: Location of exception vectors in PPC BSPs Subject: bringingup vxworks on a board with two 750cxe processors Subject: Re: Tornado II on XP Subject: Re: Instantiating "extern" variables Subject: Re: Problem with vxworks Subject: Re: Instantiating "extern" variables Subject: Re: Instantiating "extern" variables Subject: Re: Stack Trace of current stack? Subject: Re: Problems making boot floppy for T2.2/VxWorks 5.5 Subject: newbie - device drivers in c++ Subject: Re: Porting of vxWorks. Subject: Re: Stack Trace of current stack? Subject: Re: Porting of vxWorks. Subject: Re: bringingup vxworks on a board with two 750cxe processors Subject: Re: Stack Trace of current stack? Subject: Re: newbie - device drivers in c++ Subject: Mixed C & Assembly listing Subject: Re: Mixed C & Assembly listing Subject: Re: bringingup vxworks on a board with two 750cxe processors Subject: Serial Debugging of vxWorks based image Subject: Target resident shell Subject: Cold boot and Warm Boot Subject: Re: Stack Trace of current stack? Subject: Standalone application Subject: Re: Cold boot and Warm Boot Subject: Re: Cold boot and Warm Boot Subject: Re: Standalone application Subject: Re: Target resident shell Subject: Re: Problems making boot floppy for T2.2/VxWorks 5.5 Subject: Adding PCI support in Vx. Subject: Re: Target resident shell Subject: Re: Instantiating "extern" variables Subject: for help Subject: Re: Task Reschedule on mutex unlock Subject: Re: bringingup vxworks on a board with two 750cxe processors Subject: Re: Stack Trace of current stack? Subject: Re: Mixed C & Assembly listing Subject: Re: Stack Trace of current stack? Subject: Re: Mixed C & Assembly listing Subject: Re: Mixed C & Assembly listing Subject: Re: bringingup vxworks on a board with two 750cxe processors Subject: Re: Mixed C & Assembly listing Subject: question: partition a disk using dosFS Subject: UDP broadcast down multiple ENDs Subject: Re: Porting of vxWorks. Subject: Re: Problems making boot floppy for T2.2/VxWorks 5.5 Subject: Strange " warning: control reaches end of non-void function" Subject: Re: bringingup vxworks on a board with two 750cxe processors Subject: Re: Porting of vxWorks. Subject: Re: question: partition a disk using dosFS Subject: Re: UDP broadcast down multiple ENDs Subject: Re: bringingup vxworks on a board with two 750cxe processors Subject: Can't find boot file Subject: Can't find boot file Subject: Re: Interfacing NTDS board to vxworks Subject: Re: Porting of vxWorks. Subject: Serial communication example code Subject: AE, user ISR, how in PD? Plus, is anybody using AE? Subject: Re: Adding PCI support in Vx. Subject: Pci interrupts in a Pentium system. Subject: Re: newbie - device drivers in c++ Subject: Re: newbie - device drivers in c++ Subject: Re: Porting of vxWorks. Subject: Re: Porting of vxWorks. Subject: can't load .out files with dosfs2 Subject: help with objcopyppc Subject: netjobadd question Subject: Re: Mixed C & Assembly listing Subject: Re: Problem with vxworks Subject: Re: question: partition a disk using dosFS Subject: Re: Porting of vxWorks. Subject: Re: question: partition a disk using dosFS Subject: Re: Pci interrupts in a Pentium system. Subject: Re: Pci interrupts in a Pentium system. Subject: Re: question: partition a disk using dosFS Subject: problems with ldppc Subject: Re: Pci interrupts in a Pentium system. Subject: Re: question: partition a disk using dosFS Subject: Missing Header files (*.h) in the "External Dependencies" in Tornado. Subject: Re: Build specification - Link order Subject: Creating Singleton objects at run time and using it in a shared library Subject: MPC860 watchdog help ... real time scheduling problem Subject: Re: Suspended task debugging Subject: Re: Pci interrupts in a Pentium system. Subject: Re: Pci interrupts in a Pentium system. Subject: Re: MPC860 watchdog help ... real time scheduling problem Subject: Re: MPC860 watchdog help ... real time scheduling problem Subject: Re: MPC860 watchdog help ... real time scheduling problem Subject: Re: problems with ldppc Subject: Re: help with objcopyppc Subject: Re: MPC860 watchdog help ... real time scheduling problem Subject: How do I intConnect an IRQ to the FCC on wrSbc8260 Subject: Msg Q issues Subject: Re: Msg Q issues Subject: Re: Creating Singleton objects at run time and using it in a shared library Subject: Re: Strange " warning: control reaches end of non-void function" Subject: Re: problems with ldppc Subject: Strange Interrupt log msg Subject: Re: Tornado II on XP Subject: Re: Problem with vxworks Subject: help with ldppc Subject: hibernation ? Subject: Re: hibernation ? Subject: Addition to the FAQ: "newbie" page Subject: MPC8260 - FCC - Busy LAN (RET_LIM, P_PER & FPSMR[STB] settings) Subject: Re: problems with ldppc Subject: Re: Pci interrupts in a Pentium system. Subject: Re: MPC8260 - FCC - Busy LAN (RET_LIM, P_PER & FPSMR[STB] settings) Subject: Re: MPC8260 - FCC - Busy LAN (RET_LIM, P_PER & FPSMR[STB] settings) Subject: Re: Intel chipset 815E Subject: Re: Intel chipset 815E Subject: Re: MPC8260 - FCC - Busy LAN (RET_LIM, P_PER & FPSMR[STB] settings) Subject: Using mHdr.mNextPkt Subject: Re: Pci interrupts in a Pentium system. Subject: What to charge for developing a BSP? Subject: taskDelay Subject: Re: taskDelay Subject: Re: Msg Q issues Subject: How ro remove IP END addresses Subject: Re: Pci interrupts in a Pentium system. Subject: Re: What to charge for developing a BSP? Subject: PPC440GP EMAC transmit problems Subject: Re: What to charge for developing a BSP? Subject: Ping semaphore problem Subject: pls needed dosFsLib.c Subject: Re: pls needed dosFsLib.c Subject: Re: How ro remove IP END addresses Subject: Re: VxWorks Applications Engineer Subject: Re: Using mHdr.mNextPkt Subject: ataResources/sysIntEnablePIC: undefined reference ...i just want to add dosFs2.0? Subject: anyone using VMIC-7755 Subject: Re: hibernation ? Subject: Re: What to charge for developing a BSP? Subject: Re: hibernation ? Subject: Re: Porting of vxWorks. Subject: Re: bringingup vxworks on a board with two 750cxe processors Subject: Re: Porting of vxWorks. Subject: Re: hibernation ? Subject: Re: can't load .out files with dosfs2 Subject: Re: hibernation ? Subject: Re: Strange Interrupt log msg Subject: Re: hibernation ? Subject: Source file dependencies in Tornado 2.0 Subject: Re: Porting of vxWorks. Subject: Re: Source file dependencies in Tornado 2.0 Subject: Re: Porting of vxWorks. Subject: Re: Porting of vxWorks. Subject: Re: Porting of vxWorks. Subject: Link Bootable with Downloadable project Subject: ANNOUNCE: Embeddable Kermit File Transfer Protocol Subject: Re: hibernation ? Subject: Slow boot on powerpc with VxWorks 5.4 Subject: tffs/d_cache Subject: Routing problem Subject: Re: x86 pci interrupt sharing Subject: Re: Routing problem Subject: finding interrupted task.. Subject: Re: tffs/d_cache Subject: Re: finding interrupted task.. Subject: Re: x86 pci interrupt sharing Subject: Re: Slow boot on powerpc with VxWorks 5.4 Subject: How to use reserved memory Subject: Re: Link Bootable with Downloadable project Subject: Re: Slow boot on powerpc with VxWorks 5.4 Subject: Re: finding interrupted task.. Subject: Re: finding interrupted task.. Subject: diskon chip probelm Subject: Help Subject: Re: Slow boot on powerpc with VxWorks 5.4 Subject: Re: Help Subject: Re: can't load .out files with dosfs2 Subject: Tornado in multi-person projects Subject: Re: Tornado in multi-person projects Subject: Re: Slow boot on powerpc with VxWorks 5.4 Subject: Re: can't load .out files with dosfs2 Subject: Re: Routing problem Subject: Re: Routing problem Subject: Re: Routing problem Subject: Re: finding interrupted task.. Subject: Re: Target resident shell Subject: Re: Tornado in multi-person projects Subject: Re: Target resident shell Subject: Re: can't load .out files with dosfs2 Subject: Re: How to use reserved memory Subject: Re: Porting of vxWorks. Subject: socket sharing Subject: Re: Porting of vxWorks. Subject: inflate Lib.... Subject: Re: Porting of vxWorks. Subject: Re: socket sharing Subject: Re: can't load .out files with dosfs2 Subject: Re: x86 pci interrupt sharing Subject: Re: x86 pci interrupt sharing Subject: Hardware Flow Control for Serial Ports Subject: Re: Porting of vxWorks. Subject: Re: Hardware Flow Control for Serial Ports Subject: Re: Routing problem Subject: Re: Source file dependencies in Tornado 2.0 Subject: Re: Porting of vxWorks. Subject: Re: Routing problem Subject: Re: Routing problem Subject: Re: Routing problem Subject: Re: inflate Lib.... Subject: How do i print unsigned long long ? Subject: IP forwarding through PPP connection Subject: Static linking i tornado for simpc Subject: Re: Porting of vxWorks. Subject: Re: How do i print unsigned long long ? Subject: Re: Hardware Flow Control for Serial Ports Subject: Re: Hardware Flow Control for Serial Ports Subject: Help: Random Reboots of target Subject: Re: x86 pci interrupt sharing Subject: Re: can't load .out files with dosfs2 Subject: sending out Ethernet packets Subject: Re: Help: Random Reboots of target Subject: Re: How do i print unsigned long long ? Subject: Re: can't load .out files with dosfs2 Subject: Re: inflate Lib.... Subject: What is needed for symFindByName? Subject: Re: can't load .out files with dosfs2 Subject: Re: Random Reboots of target Subject: Re: How do i print unsigned long long ? Subject: Re: Random Reboots of target Subject: Re: How do i print unsigned long long ? Subject: Re: Help: Random Reboots of target Subject: change the console port in vxWorks Subject: Re: change the console port in vxWorks Subject: Re: How do i print unsigned long long ? Subject: vxWorks_rom, NVRAM , and Bootline access. Subject: CORBA --> VisibrokerRT vs. ORBexpressRT Subject: Re: How do i print unsigned long long ? Subject: Redirect i/o to the shell fails Subject: Re: can't load .out files with dosfs2 Subject: Re: How do i print unsigned long long ? Subject: Re: can't load .out files with dosfs2 Subject: priority changing in the shell Subject: Re: priority changing in the shell Subject: Re: priority changing in the shell Subject: Re: What is needed for symFindByName? Subject: Starter pack ??? Subject: Re: taskDelay Subject: Re: Porting of vxWorks. Subject: Re: Porting of vxWorks. Subject: How to configure two network cards in Vxworks? Subject: Re: Porting of vxWorks. Subject: Re: Porting of vxWorks. Subject: Re: How to configure two network cards in Vxworks? Subject: taskRestart Subject: Re: Starter pack ??? Subject: Re: Porting of vxWorks. Subject: Re: Porting of vxWorks. Subject: Re: taskRestart Subject: Where can I find a sample of SCC driver for mpc8260 Subject: Re: Porting of vxWorks. Subject: about chip select question Subject: Re: Creating Singleton objects at run time and using it in a shared library Subject: How can a Shared Library access Protection Domain data Subject: Re: Porting of vxWorks. Subject: ANNOUNCE: TargetFFS is being used with VxWorks for Guaranteed Integrity Across Unexpected Resets today! Subject: Re: Where can I find a sample of SCC driver for mpc8260 Subject: Re: How can a Shared Library access Protection Domain data Subject: Re: How can a Shared Library access Protection Domain data Subject: Re: taskRestart Subject: Re: ANNOUNCE: TargetFFS is being used with VxWorks for Guaranteed Integrity Across Unexpected Resets today! Subject: PCI devices not on page boundary Subject: Re: Porting of vxWorks. Subject: Tornado PPC8xx GNU compiler conventions Subject: Re: Routing problem Subject: Re: Tornado PPC8xx GNU compiler conventions Subject: Re: taskRestart Subject: Request: ftpdlib server code Subject: Re: Request: ftpdlib server code Subject: [comp.os.vxworks] FAQ - index Subject: Re: Tornado PPC8xx GNU compiler conventions Subject: Re: Tornado in multi-person projects Subject: TMS: Web Server and MIB Subject: Re: can't load .out files with dosfs2 Subject: Writing an assmbler trap routine Subject: question about the target server. Subject: Re: question about the target server. Subject: Re: question about the target server. Subject: Re: Porting of vxWorks. Subject: Re: Porting of vxWorks. Subject: Re: Tornado PPC8xx GNU compiler conventions Subject: Re: Porting of vxWorks. Subject: Trouble debugging 8265 with VisionProbe isync and sync instructions. Subject: Re: Writing an assmbler trap routine Subject: address translation on MPC750 / Page Table Segment Descriptor Subject: Re: address translation on MPC750 / Page Table Segment Descriptor Subject: Boot Code Integrity check with CRC Subject: Re: Starter pack ??? Subject: Re: Boot Code Integrity check with CRC Subject: Re: Boot Code Integrity check with CRC Subject: Re: Boot Code Integrity check with CRC Subject: Re: Boot Code Integrity check with CRC Subject: Re: Boot Code Integrity check with CRC ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: what's wrong with my cache flush function? Date: 15 Aug 2002 01:44:44 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0208150044.2b37a733@posting.google.com> References: <21a3d7e1.0208141728.1e703881@posting.google.com> Hello, Is there a good reason why you didn't just use the VxWorks cacheLib routines? Check out the reference manual entry for cacheLib and in particular cacheFlush() and probably cacheInvalidate() since I suspect you'll need to use that at some point too. HTH, John... PS When writing a routine totally in assembler, you'd also be better off just writing it in assembler (as a .s file), and passing that to the compiler. Look in target/h/make for the flags to pass if you want to use the C pre-processor with your assembler, or from your BSP directory just try 'make myAssembler.o' having added myAssembler.s there with something in it. du.jianjun@zte.com.cn (du) wrote in message news:<21a3d7e1.0208141728.1e703881@posting.google.com>... > I am usring Intel 82559er ethernet pci controller,because it use DMA > to transport data,so cache and memory appear disconherence,the 82559 > can not work properly.So I write a cache flush function,but it can not > work properly.My cpu is MPC7410.When I disable cache of 7410,the 82559 > can work well.So I think my cache flush function is not correct,follow > is my function,who can help me correct it?I need your help,thanks. > LOCAL void DCACHE_FLUSH(int staddr,int length) > { > __asm__("lwz 3,%0"::"m"(staddr) ); /*r3 = flush address*/ > __asm__("lwz 5,%0"::"m"(length) ); /*r5 = flush length*/ > __asm__("xor 4,4,4" ); > __asm__("flush_7410: cmpw 5,4"); > __asm__("ble flush_7410_done"); > > __asm__("dcbf 3,4"); > __asm__("addi 4,4,0x20"); > __asm__("b flush_7410"); > __asm__("flush_7410_done:"); > } --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Why msgQSend can be called by ISR? Date: Thu, 15 Aug 2002 10:43:14 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <8f77d4a.0208121740.796881b1@posting.google.com> <2025ac82.0208140116.13880440@posting.google.com> <8f77d4a.0208142319.47dc32dd@posting.google.com> Since I do not have access to Wind sources anymore, I have to rely on my memory, I still beleive that netJobAdd uses rngLib, as should excJobAdd, because these only need to support a single reader, and a single writer at any given time. In any event, you can devise your won deferral mechanism with rngLib and a server task to receive the requests and execute on them. So ytou do not need to use msgQ. Regarding the Wind kernel internal calls, some of them to intLock but for only a few machine instructions, sort of software test-and-set operation, and it also has an internal execution deferral mechanism called windWorkQ. Do not try to reverse engineer or understand the microkernel API, it eint worth the effort. "Bryan" wrote in message news:8f77d4a.0208142319.47dc32dd@posting.google.com... > Claudio Potenza > Leonid Rosenboim > > The key is that how realize the same function of the msgQSend in > VxWorks in my own_designed way.Leonid Rosenboim had give some > suggestion.I think the microcall is just like the inlock/unlock which > Claudio Potenza mentioned.So I believed that it must exits > inlock/unlock in msgQSend.In fact, msgQSend is called by > excJobAdd,which I find by disassembling.So,excJobAdd is the same as > msgQSend essentially except that some additional mechanism.More > step,excJobAdd use msgQSend to notice another task.It is up the task > to receive message but not the ISR. > > Waiting for your reply! > > > Sincerely, > Bryan --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Anyone using Tornado 2.1.1 gcc with optimization.... Date: 15 Aug 2002 01:56:27 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0208150056.21f16b15@posting.google.com> References: Hello, Your posting is very unlikely to get you any useful replies... You need to tell us much more than you have. For example: 1) Which CPU? 2) What level of optimisation? 3) Exactly which compiler? [at least cc --version output] 4) What errors you are seeing... 5) What makes you think that it is the compiler that is at fault There are many experienced people reading this group, and contributing answers regularly. On the whole we are also very tolerant of new users and very basic questions, but one thing I think I can say with some certainty is that none of us are psychic. We cannot guess what tools/CPU/BSP/etc you are using, nor what you are doing with it, what is failing and how it fails. Please post your question again with more detail, and perhaps then somebody here will be able to get you some help. Rgds, John... dsa0000@hotmail.com (dsa) wrote in message news:... > Hi All, > > Anyone using Tornado 2.1.1 tool chain ? > We are facing some issues with optimization.. > > > Thanks > DSA --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ipAttach ipDetach memory leak problem Date: 20 Aug 2002 03:14:58 -0700 From: gabi@tdsoft.com (Gabi) Organization: http://groups.google.com/ Message-ID: <5607ff2c.0208200214.5468d0c9@posting.google.com> References: <5607ff2c.0208180549.7ad53054@posting.google.com> <4a7f0d21.0208181900.64784a5c@posting.google.com> lisheng_811@yahoo.com.cn (Saledom) wrote in message news:<4a7f0d21.0208181900.64784a5c@posting.google.com>... > gabi@tdsoft.com (Gabi) wrote in message news:<5607ff2c.0208180549.7ad53054@posting.google.com>... > > Hi > > When I operate the following sequence from my code, NOT from shell: > > ipDetach (1,"fec"); > > ipAttach(1,"fec"); > > I notice in memShow of a memory leak (6-12 bytes) in each sequence. Do > > I miss something? > > tanks > > Gabi > > try the following code, please. > > free(ifunit("fec")->if_name); > ipDetach(1, "fec"); > ipAttach(1,"fec"); Hi, I didn't quit follow you up, but if you are in attach mode and operate your steps you still suffer from leak problem. Gabi --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Continue from the same point after power fail (PPC VxWorks 2)? Date: Thu, 15 Aug 2002 11:35:53 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <3D5B7083.3020908@spamm.me.l8s.co.uk> David, I agree with you completely, except that I think that well designed compact Flash controller should be able to handle power loss anytime, and be able to recover from that. Because you can't write on Flash location perviously written, only on previously erased block, the internal structure of any Flash media is akin to journal, and every sector is written in a new physical location, then some metadata is updated to indicate that new version of that logical sector appeared. Old version will be subsequently destroyed and block erased to make room for new writes. I am curious as to what make of CF cards had problems with power remoival, so I will know which card not to use in my digial camera. - - Leonid "David Laight" wrote in message news:3D5B7083.3020908@spamm.me.l8s.co.uk... > Leonid Rosenboim wrote: > > Haim, > > > > What you are trying to do is in fact far more complex than seems at first, > > because saving the entire memory image and CPU registers can restore the CPU > > state, but the hard part is the assorted peripherals, which would too need > > to be restored, or re-initialized to a working state following this > > recovery. For this, every driver must have a function to safely > > re-initialize the chip to a working state without affecting any of the > > software state, and none of the VxWorks drivers I have seen can do this - > > not even the couter-timer or serial drivers in the BSP ! > > This is made even harder by the fact that the 'power fail' interrupt > might interrupt device driver code that is part way through an > (otherwise atomic) sequence (eg reading an ethernet frame from the > on-board fifo). > > I got sleep/resume going on the system I used, but I wrote drivers > for all the hardware that suported a sleep lock (cf interrupt lock). > > Worse still a compact flash card could be part way through a data > reorganisation. Pulling the power then makes a right mess of the > data! > > Davic > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Continue from the same point after power fail (PPC VxWorks 2)? Date: Thu, 15 Aug 2002 09:09:38 +0000 (UTC) From: David Laight Organization: BT Openworld Message-ID: <3D5B7083.3020908@spamm.me.l8s.co.uk> References: Leonid Rosenboim wrote: > Haim, > > What you are trying to do is in fact far more complex than seems at first, > because saving the entire memory image and CPU registers can restore the CPU > state, but the hard part is the assorted peripherals, which would too need > to be restored, or re-initialized to a working state following this > recovery. For this, every driver must have a function to safely > re-initialize the chip to a working state without affecting any of the > software state, and none of the VxWorks drivers I have seen can do this - > not even the couter-timer or serial drivers in the BSP ! This is made even harder by the fact that the 'power fail' interrupt might interrupt device driver code that is part way through an (otherwise atomic) sequence (eg reading an ethernet frame from the on-board fifo). I got sleep/resume going on the system I used, but I wrote drivers for all the hardware that suported a sleep lock (cf interrupt lock). Worse still a compact flash card could be part way through a data reorganisation. Pulling the power then makes a right mess of the data! Davic --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Ip Checksum Date: Thu, 15 Aug 2002 09:17:37 +0000 (UTC) From: David Laight Organization: BT Openworld Message-ID: <3D5B7265.7080706@spamm.me.l8s.co.uk> References: <3d5b2217$1@post.newsfeed.com> Naveen wrote: > > Hi, > I am looking for IP Checksum written in vxworks assembly language. Please > let me know somebody has already written it. WRS - most ports use an assembler version...... Clearly you don't know what you are talking about, since vxworks is an OS not a CPU. Davic --------------------------- Newsgroups: comp.os.vxworks Subject: Re: tftpCopy disable context switching ? Date: Tue, 20 Aug 2002 06:23:01 -0400 From: joe durusau Organization: WEBUSENET.com Message-ID: <3D621885.E09703AD@bellsouth.net> References: <8a5dc325.0208200105.355c76b3@posting.google.com> Shay wrote: > Hi, > I'm using VxWorks "tftpCopy". While "tftpCopy()" is downloading > I see that no other task is running. > > Anyone knows whether "tftpCopy" disabled context switching ? > How to get around ? > > Thanks, > Shay. I don't know how it does it, but I do know that tftp uses UDP, which means that it can't stand to be stopped by higher pri tasks. You need to use something else if you want other behavior. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Re: problem connecting a flahs Block Driver to DosFsLib Date: Thu, 15 Aug 2002 10:58:02 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Didn't WindRiver provide you with a DosFs 2.0 manual ? dosFs 2.0 will actually access the media not when you initialize the module, but when you first access the device. The error messages you see are an indication you have not formatted the device for DosFs. This is something you need to do once for each board (assuming your Flash driver is bug free), and dosFs 2.0 has a function similar to FORMAT on PCs. Use it. "aviv halperin" wrote in message news:d0a6f188.0208150023.34dd9ad9@posting.google.com... > The problem is as follows: > 1) I wrote a Block Driver for a 2MB Flash device. > 2) I connected the Flash Block Driver to the dos FS with the following > code: > > > BOOL DevCreate() > { > int dosFsCacheSizeDefault = 128 * 1024 ; > void * pCbio; > char bootDir [BOOT_FILE_LEN]; > > #define FLASH_MAX_NUM_OF_FILES 300 /* number of simultaneously open > files */ > > pBlkDev =flashDevCreate(8,8,0x1000000/8,0x100); > if( dosFsLibInit( 0 ) == ERROR ) > { > printErr ("dosFsLibInit failed.\n"); > return (ERROR); > } > /* Sub-module: VFAT Directory Handler */ > dosVDirLibInit(); > /* Sub-module: FAT12/FAT16/FAT32 FAT Handler */ > dosFsFatInit(); > devSplit ("BLKDEV:", bootDir); > if ( (pCbio = dcacheDevCreate( (void *) pBlkDev, NULL, > dosFsCacheSizeDefault, bootDir)) == NULL ) > { > printErr ("dcacheDevCreate failed.\n"); > return (ERROR); > } > > if (dosFsDevCreate (bootDir, pCbio, FLASH_MAX_NUM_OF_FILES,NONE) != > OK) > { > printErr ("dosFsDevCreate failed.\n"); > return (ERROR); > } > return (OK); > } > > > This worked fine(I think) cause when I checked the device list > ("devs") > I saw the new device "BLKDEV:". > > 3) now I tryed using the io's and wrote: > > open("BLKDEV:",0x200,0) and got the messages: > > 0x76da88 (tDbgTask): dosFsLib.c : Malformed boot sector. Offset 0, > value 255. > 0x76da88 (tDbgTask): dosFsLib.c : Try to accept other boot block > copy... > 0x76da88 (tDbgTask): dosFsLib.c : Malformed boot sector. Offset 0, > value 255. > 0x76da88 (tDbgTask): dosFsLib.c : Malformed boot sector. Offset 0, > value 255. > 0x76da88 (tDbgTask): dosFsLib.c : Malformed boot sector. Offset 0, > value 255. > 0x76da88 (tDbgTask): dosFsLib.c : Malformed boot sector. Offset 0, > value 255. > 0x76da88 (tDbgTask): dosFsLib.c : Malformed boot sector. Offset 0, > value 255. > 0x76da88 (tDbgTask): dosFsLib.c : Malformed boot sector. Offset 0, > value 255. > 0x76da88 (tDbgTask): dosFsLib.c : Malformed boot sector. Offset 0, > value 255. > > > > > please help me cause I'm lost here! > > Thanks very much. > > Aviv. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: tftpCopy disable context switching ? Date: Tue, 20 Aug 2002 12:56:58 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <8a5dc325.0208200105.355c76b3@posting.google.com> <3D621885.E09703AD@bellsouth.net> Joe, I disagree: Although UDP is an unreliable protocol, the socket buffer can still contain a few packets, so if higehr priority tasks take some CPU, packets won't necesarily get lost. Also, TFTP protocol has a retransmit mechanism, so as to make file transfer reliable even on top of UDP, so even lost packets (which is inevitable) wont corrupt files. Shai, I strongly doubt that tftpCopy has any significant effect on OS and task switching behaviour, moreso, tftpCopy uses more then one task (the caller of tftpCopy) to do its job (if my memory serves me well), meaning tftpCopy cant work at all wiithout multitasking in full operation ... "joe durusau" wrote in message news:3D621885.E09703AD@bellsouth.net... > > > Shay wrote: > > > Hi, > > I'm using VxWorks "tftpCopy". While "tftpCopy()" is downloading > > I see that no other task is running. > > > > Anyone knows whether "tftpCopy" disabled context switching ? > > How to get around ? > > > > Thanks, > > Shay. > > I don't know how it does it, but I do know that tftp uses UDP, which > means that it can't stand to be stopped by higher pri tasks. You need > to use something else if you want other behavior. > > Speaking only for myself, > > Joe Durusau > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: somebody can show me how to use the function symFindByName() Date: 20 Aug 2002 04:26:21 -0700 From: tyx@huawei.com (Tao Yongxiang) Organization: http://groups.google.com/ Message-ID: <8f8b64c7.0208200326.5d581c69@posting.google.com> References: <7efa47da.0208192051.596e712c@posting.google.com> liao@ztenet.com (ao.li) wrote in message news:<7efa47da.0208192051.596e712c@posting.google.com>... > I am very glad to see you can give me an example ,thanks! The following code maybe do some help. #include "symLib.h" extern SYMTAB_ID sysSymTbl; STATUS SymFindByName(char *pName, ULONG *pSymAddr) { SYM_TYPE symType; if (ZERO == sysSymTbl) { return ERROR; } if (OK == symFindByName(sysSymTbl, pName, (char **)pSymAddr, &symType)) { return OK; } else { return ERROR; } } --------------------------- Newsgroups: comp.os.vxworks Subject: jungos driver development kit Date: 20 Aug 2002 05:01:11 -0700 From: vxnerd@rediffmail.com (subbarayan) Organization: http://groups.google.com/ Message-ID: dear group, Have anyone experience in using jungos driver development kit for vxworks?For those who are unaware of whats it i give a brief intro:Jungo.com is a website which offers trial version of driver development kit for vxworks and makes driver development easier.U get the trial version free of cost from website. Incase any of u have worked with it i would like to have an interaction regarding the proceedure of using it in my application fro building device drivers for pci cards from vxworks. Anyone with experience can mail me to vxnerd@rediffmail.com or jus reply here in board. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Need to watch DTR status on a PPC Date: 15 Aug 2002 06:50:23 -0700 From: akiggins@celitesystems.com (andrew) Organization: http://groups.google.com/ Message-ID: <2eced96.0208150550.65be7e6f@posting.google.com> References: <2eced96.0208131224.42163659@posting.google.com> <3D59A77D.D839CC84@bellsouth.net> FYI, the SCC on the PPC does not capture DTR, we had to have the hw guys add a latch and pass it to an IRQ. andrew joe durusau wrote in message news:<3D59A77D.D839CC84@bellsouth.net>... > andrew wrote: > > > Ladles and Jellyspoons, > > does anyone have a clue how I can grab DTR on the serial port of a > > PPC? > > > > I need to figure out if someone is disconnecting from the serial port > > and reset the shell (which is password protected) and cause them to be > > logged out if they haven't explicitly. > > > > Any pointers to docs or code are great. > > thanks in advance > > andrew > > Consult the hardware docs for your board. It should tell you what the > port address is. Read it when you want to. I don't use PPC's, so > I can't guess at the address. > > Speaking only for myself, > > Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: localtime () Date: Tue, 20 Aug 2002 12:35:21 +0000 (UTC) From: "Peter Moseley" Organization: BT Openworld Message-ID: Reply-To: "Peter Moseley" I have a feeling that this is a bit of a basic question but I can't find the answer. Whenever my program calls "localtime ()" I always get the current time returned as GMT, not as my local time. How do I tell vxWorks to return the latter? Thanks Peter Moseley --------------------------- Newsgroups: comp.os.vxworks Subject: Re: jungos driver development kit Date: Tue, 20 Aug 2002 13:44:32 +0100 From: Paul Caswell Organization: Land & Sea Systems Message-ID: <3D6239B0.DF078AC8@baesystems.com> References: We have used it to develop a PCI driver under NT4. It's about as simple as it can get and surprisingly quick. For us the main attraction was the portability of the resultant code, NT, 2000, 95, 98, Me, VxWorks, Linux, etc.... Regards Paul --------------------------- Newsgroups: comp.os.vxworks Subject: Wind Web Server File Uploads Date: 15 Aug 2002 07:51:47 -0700 From: lchamberlain@pgaero.co.uk (Laura Chamberlain) Organization: http://groups.google.com/ Message-ID: I am trying to upgrade from the Beta version of wws 2 to the released version but I can't get file uploads to work as I want. I have a custom RPM which uploads a binary file to a temporary area, checks it's valid and then programs it to flash. What now happens is the file is uploaded to the location specified by HTTP_POST_FILE_UPLOAD_PATH retaining the name specified in the upload form. Once the whole file has been uploaded my RPM is then called. This is no good because :- 1. A file of any name can be uploaded - how does my RPM know what it is called in order to process it and remove it if invalid 2. I want to give the user feedback as the file is uploaded (it is >1Mb and takes a while). How can I stop the file being automatically uploaded so my RPM can do it (as it did before)? --------------------------- Newsgroups: comp.os.vxworks Subject: Ethernet Driver for the MPC8260 FEC ves CPM37 Errata Date: 15 Aug 2002 08:02:32 -0700 From: gabi@tdsoft.com (Gabi) Organization: http://groups.google.com/ Message-ID: <5607ff2c.0208150702.31dff97e@posting.google.com> Hi, I'm using the MPC8260 under VxWorks, and I wonder how you guys faces the need to handle the Motorola Errata CPM37 which are relevant directly to the MPC8260 FEC. tanks Gabi --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Design ideeas on shared memories Date: 20 Aug 2002 06:49:14 -0700 From: avsrk@mailcity.com (Subrahmanyam Arya) Organization: http://groups.google.com/ Message-ID: <25c1993e.0208200549.2703e980@posting.google.com> References: <25c1993e.0208160613.81842c9@posting.google.com> <754b62e0.0208191425.17e12a7f@posting.google.com> Just to make it more clear , I am going to have multiple processes accessing the shared memory , each of them could be single / multi threaded but i am going to access the shared memory from only a particular thread of each process , i.e a thread from a particular process can acquire lock and read/write data and release the lock . I am assumming the suggestions you made are for a scenario like this .... - -subra emil@lapel.com (Emil Rojas) wrote in message news:<754b62e0.0208191425.17e12a7f@posting.google.com>... > I would recommend strongly against a data structure that exposes > semaphores. You would do well to impose some abstraction between the > data and the consumers of the data. Unless there was some overriding > imperative not to do so, I would create a single thread that read and > wrote the acquired data. If a single thread is responsible for writing > data into and reading data from the shared memory then you don't have > to worry about all that contention crap. > Much of course depends on the structure of the actual data, its > arrival rate, etc. But in most case a simple thread that handles both > write and read requests simplifies the unnecessary complexity that > semaphore and the like you get you into. > > emil > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Design ideeas on shared memories Date: Tue, 20 Aug 2002 15:42:37 +0100 From: Paul Caswell Organization: Land & Sea Systems Message-ID: <3D62555D.B52D0646@baesystems.com> References: <25c1993e.0208160613.81842c9@posting.google.com> I suspect that your task would be made a lot easier if you had access to Wind Rivers shared memory objects. They are not part of the standard vxWorks but I'm sure they would simplify your job significantly. Paul --------------------------- Newsgroups: comp.os.vxworks Subject: Re: FTP Date: Thu, 15 Aug 2002 12:29:55 -0400 From: david lindauer Message-ID: <3D5BD703.3ECCC488@notifier-is.net> References: <7CG69.35670$me6.5347@sccrnsc01> there should be a setting somewhere for what directory should the FTP server default to when you log in. I haven't gone looking for it yet though, we just CD to the directory every time we log in. David weddick wrote: > We just upgrade our BSP to the pentium pro. The FTP server does not seem to > work the way it did before. Previously, we could use the windows ftp client > and see the files on /ide0/. Now the system wants to connect to host: and I > have to change the directory to /ide0/. Our Exceed FTP client won't work at > all. > > Anyone else have this problem? > > Thanks, --------------------------- Newsgroups: comp.os.vxworks Subject: Re: localtime () Date: Tue, 20 Aug 2002 14:50:47 +0000 (UTC) From: David Laight Organization: BT Openworld Message-ID: <3D625804.3000906@spamm.me.l8s.co.uk> References: > > Whenever my program calls "localtime ()" I always get the current time > returned as GMT, not as my local time. How do I tell vxWorks to return the > latter? RTFM: putenv( "TIMEZONE=XXX::min_west:start_summer:end_summer" ) where start_summer and end_summer are mmddhh and XXX is the name of the zone. unfortunately: - - you can't have a different name for winter - - you have to specify a day in the month (ie not 4th sunday etc) - - pJava has its own rules and only takes the name, it doesn't have a zone that is valid for the UK in the summer :-( David --------------------------- Newsgroups: comp.os.vxworks Subject: Command-line compiler problems with Tornado 2.1/NT for MIPS Date: Wed, 14 Aug 2002 02:45:47 GMT From: "Digital Man" Organization: Prodigy Internet http://www.prodigy.com Message-ID: <3D59CBFE.1357.UVXWORKS@vert.synchro.net> If I just try to build a simple "hello world" program with c++mips.exe (as an example), I get the following error: $ c++mips test.c /Tornado/host/x86-win32/mips-wrs-vxworks/bin/ld: cannot find -lm collect2: ld returned 1 exit status The only libm.a file in my Tornado (2.1) installation is in the Tornado/host/x86-win32/i386-pc-mingsw32/lib directory (and it's just an empty archive anyway). I've tried creating an empty libm.a with armips in the lib search path, but then I get an access violation in the linker: $ c++mips test.c (C:\Tornado\host\x86-win32\mips-wrs-vxworks\bin\ld.exe 1005) Exception: STATUS_ACCESS_VIOLATION (C:\Tornado\host\x86-win32\mips-wrs-vxworks\bin\ld.exe 1005) Dumping stack trace to ld.exe.core /Tornado/host/x86-win32/mips-wrs-vxworks/bin/nm: a.out: File truncated collect2: C:/Tornado/host/x86-win32/bin/../lib/gcc-lib/mips-wrs-vxworks/gcc- 2.96/../../../../mips-wrs-vxworks/bin/nm ret urned 1 exit status I can link with --nostdlib, but then of course I get hundreds of unresolved symbols, so I'd like to avoid specifying every library name by hand if I don't have to. Do you have any idea why the C++ compiler is requiring libm.a and why it would not be present in the Tornado MIPS library directories? Now, just as a test, I used the C compiler (ccmips) and got different results: $ ccmips test.c /Tornado/host/x86-win32/mips-wrs-vxworks/bin/ld: cannot find -lgcc collect2: ld returned 1 exit status Now, libgcc.a for MIPS I do have, so I just added that library search directory to the command-line: $ ccmips test.c - -L/tornado/host/x86-win32/lib/gcc-lib/mips-wrs-vxworks/gcc-2.96/MIPS32sfgnu /Tornado/host/x86-win32/mips-wrs-vxworks/bin/ld: warning: cannot find entry symbol _start; defaulting to 000000000040000 0 c:\DOCUME~1\swindell\LOCALS~1\Temp\ccyWaaaa.o: In function `main': c:\DOCUME~1\swindell\LOCALS~1\Temp\ccyWaaaa.o(.text+0x14): undefined reference to `printf' collect2: ld returned 1 exit status It now finds libgcc.a, but still has some unresolved symbols. Anyway, I *do* need to use the C++ compiler so that was just a test to see if would have any more success with the ccmips.exe. Thanks for any help you can offer, digital man - --- Synchronet 3.10j-Win32 NewsLink 1.29 * Vertrauen - Fullerton, California - telnet://vert.synchro.net --------------------------- Newsgroups: comp.os.vxworks Subject: Re: FTP Date: Thu, 15 Aug 2002 13:32:53 -0400 From: david lindauer Message-ID: <3D5BE5C5.FBD71676@notifier-is.net> References: <7CG69.35670$me6.5347@sccrnsc01> <3D5BD703.3ECCC488@notifier-is.net> david lindauer wrote: > there should be a setting somewhere for what directory should the FTP server > default to when you log in. I haven't gone looking for it yet though, we just > CD to the directory every time we log in. actually , I think the boot configuration lines (in the BSP's config.h) specify a default directory, and it seems like I noticed one day it is the same directory as used when you log into FTP. David > > David > > weddick wrote: > > > We just upgrade our BSP to the pentium pro. The FTP server does not seem to > > work the way it did before. Previously, we could use the windows ftp client > > and see the files on /ide0/. Now the system wants to connect to host: and I > > have to change the directory to /ide0/. Our Exceed FTP client won't work at > > all. > > > > Anyone else have this problem? > > > > Thanks, --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Design ideeas on shared memories Date: 20 Aug 2002 09:29:06 -0700 From: info@mcobject.com (Information) Organization: http://groups.google.com/ Message-ID: <7a4ba6ba.0208200829.1aa12e6d@posting.google.com> References: <25c1993e.0208160613.81842c9@posting.google.com> I'd like to invite you to take a look at McObject's eXtremeDB (McObject is a WindLink partner). eXtremeDB is an in-memory database system. It will permit you to accomplish everything you describe below without having to worry about locks and semaphores (eXtremeDB does all that for you; just begin and end your transactions). You'll also get the benefit of ACID-compliant transactions, the ability to prioritize your database transactions, ability to incorporate different access methods, and much more. A documentation and a free evaluation version of eXtremeDB for VxWorks is available at our website. Good luck. McObject LLC info@mcobject.com http://www.mcobject.com avsrk@mailcity.com (Subrahmanyam Arya) wrote in message news:<25c1993e.0208160613.81842c9@posting.google.com>... > Hi Vxworkers , > > I am trying to implement a real-time application on VxWorks which > uses shared > memories . I am going to have one task which is going to do data > acquisition and control and needs higher priority . > > There would be another 3 to 4 tasks like a datalogger , gui for > showing the data > acquired over the web and on a serial line etc . > > So the real -time data acquisition and control job writes into the > shared memory and the other processes just read from it . > > > I am familiar with the shared memories and semaphores . Is the simple > mechanism of acquiring a lock doing a read/write on to the > sharedmemory an d > releasing the lock with the real time control job running at a higher > priority be sufficient , or do we need some thing special like > read/write locks etc . > > Does any one have any specific experience with this type of issues . > > Apprecaiate your thoughts and ideas in this > > -Subra --------------------------- Newsgroups: comp.os.vxworks Subject: Re: tftpCopy disable context switching ? Date: Tue, 20 Aug 2002 10:07:31 -0700 From: "Dan Gold" Organization: Mailgate.ORG Server - http://www.Mailgate.ORG Message-ID: References: <8a5dc325.0208200105.355c76b3@posting.google.com> <3D621885.E09703AD@bellsouth.net> Sender: gold@ensemble.com "Leonid Rosenboim" wrote in message news:newscache$zi251h$0gf$1@lnews.actcom.co.il... > Shai, I strongly doubt that tftpCopy has any significant effect on OS and > task switching behaviour, moreso, tftpCopy uses more then one task (the > caller of tftpCopy) to do its job (if my memory serves me well), meaning > tftpCopy cant work at all wiithout multitasking in full operation ... I believe tNetTask would be a useful one to have running too :-) And how did the OP determine that no other task was running - by doing an 'i' at the shell ? It could be that no other tasks with priorities below the net task, and caller of tftpCopy are running. In that case, some priority adjustment might be necessary. hth, Dan Gold dgold1@san.rr.com > > Shay wrote: > > > > > Hi, > > > I'm using VxWorks "tftpCopy". While "tftpCopy()" is downloading > > > I see that no other task is running. > > > > > > Anyone knows whether "tftpCopy" disabled context switching ? > > > How to get around ? > > > > > > Thanks, > > > Shay. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: localtime () Date: Tue, 20 Aug 2002 19:45:33 +0100 From: "Dimitrios Siganos" Organization: Newport Networks Ltd. Message-ID: <1029869138.936917@azores.network-i.net> References: I got the follwoing extract from my online help. It appears to answer your question. When you read the documentation about a function don't just read the function specific text. Read the beginning of the file as well, which contains general info that applies to all the functions decsribed in that file. Haven't used any time functions yet so i don't know if this is all you need... anyway HTH. Here's the extract: If the environment variable TIMEZONE is set, the information is retrieved from this variable, otherwise from the locale information. TIMEZONE is of the form: name_of_zone:<(unused)>:time_in_minutes_from_UTC:daylight_start:daylight_end To calculate local time, the value of time_in_minutes_from_UTC is subtracted from UTC; time_in_minutes_from_UTC must be positive. Daylight information is expressed as mmddhh (month-day-hour), for example: UTC::0:040102:100102 "Peter Moseley" wrote in message news:ajtd29$mkp$1@paris.btinternet.com... > I have a feeling that this is a bit of a basic question but I can't find the > answer. > > Whenever my program calls "localtime ()" I always get the current time > returned as GMT, not as my local time. How do I tell vxWorks to return the > latter? > > Thanks > > Peter Moseley > > --------------------------- Newsgroups: comp.os.vxworks Subject: hardware compability Date: Thu, 15 Aug 2002 16:19:23 -0700 From: "Li Tan" Organization: The University of Oklahoma Message-ID: <7%U69.81$95.2335@news.ou.edu> Hi, I am new to Vxworks, and I am planning to use it in my robot project, but I caould not find the information on the hardware supported by vxworks in its web site. Could you give me a hand or hint? thanks Li --------------------------- Newsgroups: comp.os.vxworks Subject: Re: porting vxworks to i386ex custom board Date: 13 Aug 2002 21:56:40 -0700 From: jyothimshetty@yahoo.com (Jyothi M) Organization: http://groups.google.com/ Message-ID: References: <3D58108F.7AFBAAB1@notifier-is.net> hi everyone, thanks i have written the bsp,and built the following files, vxWorks.res_rom_low.bin vxWorks.res_rom.hex vxWorks.st vxWorks_rom.hex and i want to load the bsp into my custom board flash, through serial communication only. which to load, and how to load it into flash. pls help me regards. jyothi david lindauer wrote in message news:<3D58108F.7AFBAAB1@notifier-is.net>... > you have to write a BSP, which is the low-level interface between Vxworks > and the hardware. I would start with the standard 386 Bsp and rewrite/add > the parts that are specific to your board but non-standard. > > David > > Jyothi M wrote: > > > i am working on custom build i386ex board. i want to port vxworks on it. > > my board has 2mb flash, 256kb ram, and 2serial ports. > > > > how can i proced further. > > > > regards. > > Jyothi.M --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Design ideeas on shared memories Date: 20 Aug 2002 13:48:10 -0700 From: avsrk@mailcity.com (Subrahmanyam Arya) Organization: http://groups.google.com/ Message-ID: <25c1993e.0208201248.7b58bd@posting.google.com> References: <25c1993e.0208160613.81842c9@posting.google.com> <3D62555D.B52D0646@baesystems.com> Windriver Shared memory objects are for having a shared memory access in a target hardware where you have multiple processors ( not processes ) . Ordinary shared memories and semaphores are enough when you have single process and multiple processes , so the issue i was trying to discuss is having simple , lock , unlock mechnism with semget() and semrelease() verses developing read / write locks using this semphore related primitive calls . What i want to know is wether the primitive calls are sufficent or do we need to develop this type of locks . My application is kind of simple having about 5 processes accessing the shared memories , one does the right and the rest all do the read . What is the best way of doing this , any ideas please !!! thanks in advance - -subra Paul Caswell wrote in message news:<3D62555D.B52D0646@baesystems.com>... > I suspect that your task would be made a lot easier if you had access to > Wind Rivers shared memory objects. They are not part of the standard > vxWorks but I'm sure they would simplify your job significantly. > Paul --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sysMemTop() doesn't quite work... Date: Tue, 20 Aug 2002 21:10:24 +0000 (UTC) From: Johan Borkhuis Organization: Agere Systems Message-ID: References: jchen@terayon.com (john) wrote: >> You don't say but my guess is that you are using the Tornado II >> project facility to build the image. If this is the case, it is >> likely that the project facility provides is own definition of >> USER_RESERVED_MEM which overrides the one in config.h. I think you >> will find this parameter in the project manager under >> hardware->memory although I could be wrong (I'm not at work so I >> can't confirm this!) > We are compiling from command line, not from Tornado II. Hmmm. Now > I'm more suspecious of what vxWOrks is doing behind my back.... Did you create your project from the Project facility? If so, it is redefined in one of prj files in your project. Or you can just do a search for USER_RESERVED_MEM in your BSP and project directory. Or give the option "-save-temps" to the compiler and look at the sysLib.i file to see how the file was expanded. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Strange files appearing on the hard drive Date: Wed, 14 Aug 2002 17:40:14 GMT From: "John Kevin Fabiani" Organization: Prodigy Internet http://www.prodigy.com Message-ID: <2Cw69.5098$fm.2387@newssvr19.news.prodigy.com> References: <95ebbe31.0208130848.27aead6f@posting.google.com> This is fixed in Tornado 2.2 FCS. This is also fixed in the latest dosFs2.2 patch, see SPR79795 for x86. John "Aaron Graham" wrote in message news:95ebbe31.0208130848.27aead6f@posting.google.com... > Using DosFs2, I have been noticing quite often that I will get strange > files appearing on my hard drive. Here's what they look like when I > do a "ll" from the vxWorks shell: > > -rwxrwxrwx 1 0 0 2618 Aug 13 09:03 info.txt > ?--------- 0 0 0 0 Jan 1 1970 AA > > The file is usually called "AA", "A", or "Br", and cannot be read, > written, or removed, and running chkdsk does not find the problem. It > causes lots of problems while using file utilties in our application, > simply because it shows up, but can't be accessed. The only way to > fix it is to re-format the partition and re-install the software. > > I would assume that something has gone wrong with the directory > structure. Probably yanked the plug on the machine while it was doing > some sort of critical file access. RTOS...reliable > filesystem...?...nah, guess not. > > Let me know if you've seen it too. > Aaron --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Debug question Date: 20 Aug 2002 17:35:03 -0700 From: g_ch@21cn.com (GANGCHENG) Organization: http://groups.google.com/ Message-ID: References: No,it is only the test example for using netbuflib . Why? "carria" wrote in message news:... > I have encountered it also, maybe your application is time limited. > > "GANGCHENG" wrote in message > news:ce273620.0208191708.3cbac770@posting.google.com... > > Dear all, when I launch the Debugger under Tornado2 and debug the > > program step by step, the following exception is shown in the target > > shell . But when I directly run the program , there is no exception > > and error. > > > > > > Exception current instruction address: 0x00d8cf90 > > Machine Status Register: 0x00089030 > > Condition Register: 0x24000000 > > Task: 0xd8d400 "tDbgTask" > > > > > > why ? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: jungos driver development kit Date: 20 Aug 2002 21:02:00 -0700 From: vxnerd@rediffmail.com (subbarayan) Organization: http://groups.google.com/ Message-ID: References: <3D6239B0.DF078AC8@baesystems.com> Dear paul and other mentors, Paul thanks personally for your reply.Have u ported your code to vxworks using jungo?If in case u have done can u tell me the configuration proceedure in vxworks to use jungo? Can anyone tell me how to configure jungos kit and use it in vxworks and develop driver in vxworks?I am expecting something like step by step proceedure if possible. Advanced thanks to the person replying me.Anyone with a solution can mail me to vxnerd@rediffmail.com or jus reply in the board itself. --------------------------- Newsgroups: comp.os.vxworks Subject: How to make ccppc still use user-defined lables after compilation? Date: Tue, 20 Aug 2002 22:30:23 -0700 From: "Liang Yang" Organization: Arizona State University Message-ID: Hi, I insert some c labels into .c source file as follows: begin: i = 1; end: But after I use ccppc to compile this .c file into a .s file,all these labels are replaced with ccppc format lables .Lxx:(such as .L01:, .L02:). How can I make ccppc still use my own lables(begin:, end:) instead of its own lables in assembly file after compilation? Thank you. Liang --------------------------- Newsgroups: comp.os.vxworks Subject: Does redirect route in vxworks bring memory leak? Date: 20 Aug 2002 23:04:43 -0700 From: tyx@huawei.com (Tao Yongxiang) Organization: http://groups.google.com/ Message-ID: <8f8b64c7.0208202204.28b9143e@posting.google.com> Does redirect route in vxworks bring memory leak? Recently we met a problem. Our system based on vxworks could not to be telneted, but can be pinged. Here is our network topo: RAS server 10.190.46.158 / \ E1 network PSTN / \ router1 router2 remote PC 10.190.47.158 10.190.47.149 10.190.46.145 / 10.190.47.154/24 SYS on vxworks 129.9.1.9 Remote PC get an ip address from RAS server, which has an ip address pool from 10.190.46.145 to 10.190.46.155. SYS, route1 and route2 are located at the same place, all of those connected to a LAN which has a subnetwork 0xfffffff0. Route2 is the gateway to E1 network. When I set default gateway in SYS to route1, i.e. - -> mRouteShow Destination Mask TOS Gateway Flags RefCnt Use 0.0.0.0 0 0 10.190.47.158 3 0 7613 10.190.47.144 fffffff0 0 10.190.47.154 101 0 0 127.0.0.1 0 0 127.0.0.1 5 0 0 129.9.0.0 ffff0000 0 129.9.1.9 101 0 0 value = 0 = 0x0 and from remote pc I telnet to SYS, found have some redirect routes: - -> mRouteShow Destination Mask TOS Gateway Flags RefCnt Use 0.0.0.0 0 0 10.190.47.158 3 0 7613 10.190.46.145 0 0 10.190.47.149 17 1 1731045 10.190.46.146 0 0 10.190.47.149 17 0 425 10.190.46.147 0 0 10.190.47.149 17 0 269 10.190.46.148 0 0 10.190.47.149 17 0 2 10.190.46.149 0 0 10.190.47.149 17 0 238 10.190.46.150 0 0 10.190.47.149 17 0 0 10.190.47.144 fffffff0 0 10.190.47.154 101 0 0 127.0.0.1 0 0 127.0.0.1 5 0 0 129.9.0.0 ffff0000 0 129.9.1.9 101 0 0 value = 0 = 0x0 Those redirect routes never release. so when remote pc do offline and online repeatly. It's possible that remote pc get an address same as before. When remote pc get an 10.190.46.145 at the second time, SYS has a route in route table, 10.190.46.145 0 0 10.190.47.149 17 1 1731045 So I guess it will do work well without alloc route, but before connect and after connect, there is some "RTABLE" memory to be alloced. (but mRouteShow show there is no difference between the operation) BEFORE telnet: - -> netStackSysPoolShow type number - --------- ------ FREE : 458 DATA : 0 HEADER : 0 SOCKET : 7 PCB : 10 RTABLE : 29 HTABLE : 0 ATABLE : 0 SONAME : 0 ZOMBIE : 0 SOOPTS : 0 FTABLE : 0 RIGHTS : 0 IFADDR : 5 CONTROL : 0 OOBDATA : 0 IPMOPTS : 0 IPMADDR : 3 IFMADDR : 0 MRTABLE : 0 TOTAL : 512 number of mbufs: 512 number of times failed to find space: 0 number of times waited for space: 0 number of times drained protocols for space: 0 __________________ CLUSTER POOL TABLE _______________________________________________________________________________ size clusters free usage - ------------------------------------------------------------------------------- 64 64 46 40 128 64 48 37 256 64 52 39 512 64 57 28 - ------------------------------------------------------------------------------- value = 80 = 0x50 = 'P' AFTER telnet - -> netStackSysPoolShow type number - --------- ------ FREE : 453 DATA : 0 HEADER : 0 SOCKET : 8 PCB : 12 RTABLE : 31 HTABLE : 0 ATABLE : 0 SONAME : 0 ZOMBIE : 0 SOOPTS : 0 FTABLE : 0 RIGHTS : 0 IFADDR : 5 CONTROL : 0 OOBDATA : 0 IPMOPTS : 0 IPMADDR : 3 IFMADDR : 0 MRTABLE : 0 TOTAL : 512 number of mbufs: 512 number of times failed to find space: 0 number of times waited for space: 0 number of times drained protocols for space: 0 __________________ CLUSTER POOL TABLE _______________________________________________________________________________ size clusters free usage - ------------------------------------------------------------------------------- 64 64 45 41 128 64 47 38 256 64 50 41 512 64 56 29 - ------------------------------------------------------------------------------- value = 80 = 0x50 = 'P' So, do this repeatly: remote pc dial to RAS Server, telnet to SYS, and hangup. - -> netStackSysPoolShow type number - --------- ------ FREE : 256 DATA : 0 HEADER : 0 SOCKET : 13 PCB : 22 RTABLE : 213 HTABLE : 0 ATABLE : 0 SONAME : 0 ZOMBIE : 0 SOOPTS : 0 FTABLE : 0 RIGHTS : 0 IFADDR : 5 CONTROL : 0 OOBDATA : 0 IPMOPTS : 0 IPMADDR : 3 IFMADDR : 0 MRTABLE : 0 TOTAL : 512 number of mbufs: 512 number of times failed to find space: 0 number of times waited for space: 0 number of times drained protocols for space: 0 __________________ CLUSTER POOL TABLE _______________________________________________________________________________ size clusters free usage - ------------------------------------------------------------------------------- 64 64 0 88 128 64 0 131 256 64 0 115 512 64 1 153 - ------------------------------------------------------------------------------- value = 80 = 0x50 = 'P' The free memory will be soonly consumed in "RTABLE", when there is no free memory block in network stack, telnet request could not to be passed to TCP from IP, so telnet failed. Why? any advice or message is welcome! --------------------------- Newsgroups: comp.os.vxworks Subject: Re: IP_MAX_UNITs on custom built simulator Date: 15 Aug 2002 19:12:27 -0700 From: joge709@hotmail.com (joge) Organization: http://groups.google.com/ Message-ID: <1d168114.0208151812.6dd7ef7e@posting.google.com> References: <290f4d3c.0208140254.5c545a64@posting.google.com> hi ravurirr, You can set the value in the project at the site of : workspace--vxworks--networks component--network buffer initialization--params--IP_MAX_UNITS good luck, joge ravurirr@hotmail.com (Rao R R) wrote in message news:<290f4d3c.0208140254.5c545a64@posting.google.com>... > Hi > I tried to implement a template NPT style driver on Custorm Simulator > BSP.I could add the driver through END_DEV object table and could load > the device through muxDevLoad. When I used ipAttach(0,"tmp"), os asked > me to increase the IP_MAX_UNITS. I checked the configNet.h in > /target\config/simpc/configNet.h. But there is no > IP_MAX_UNITS defined. > I tried defining one like > #ifndef IP_MAX_UNITS > #define IP_MAX_UNITS 5 > #endif > > It still gives the same error saying increase IP_MAX_UNITS. > Any ideas how can I do this. > > thanks in advance > Rao --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Why msgQSend can be called by ISR? Date: Wed, 14 Aug 2002 10:27:59 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <8f77d4a.0208121740.796881b1@posting.google.com> <8f77d4a.0208132226.578b8509@posting.google.com> Assuming you do not have VxWorks source code and its confidential internal kernel architecture documentation, you should not attempt to use the microkernel calls. Instead what I recommend you do, is when your qsend function is called from an ISR, call excJobAdd() to have tExcTask to the actual qsend call for the ISR, and then it's OK to use Mutex semaphore for protection etc. inside your library. intLib.h contains a macro that determines if if is run in ISR context or not. Good luck. "Bryan" wrote in message news:8f77d4a.0208132226.578b8509@posting.google.com... > Leonid Rosenboim, > > Thank you very much! > > I would design a type of queue which increase dyanmicly when queue is > full,it also support urgent priority message.When I accomplish it ,I > find it is almost the same as queue in VxWorks.So I have the idea to > replace the queue in Vxworks.But msgQSend can be called by ISR,my > QSend can not.I believe it is the most difficult. > It semms that I should do two things If I want to replace queue in > VxWorks: > 1.Design the microcall alike VxWorks's > 2.When cause a race,the qsend call must be move to another work > queue,just like logmsg. > > Wait for your reply. > Thank you again! > Yours Bryan. --------------------------- Newsgroups: comp.os.vxworks Subject: Does anyone have a ping program, similar to Sisco's catalyst 2900, available under vxWorks? Date: 21 Aug 2002 00:04:40 -0700 From: billy_chxf@sina.com (billy chen) Organization: http://groups.google.com/ Message-ID: <8266a3fc.0208202304.458b796c@posting.google.com> Hi all, Does anyone have a ping program, similar to Sisco's catalyst 2900, available under vxWorks? Help, Thanks, - --Billy Chen --------------------------- Newsgroups: comp.os.vxworks Subject: ipAttach ipDetach memory leak problem Date: 14 Aug 2002 01:43:35 -0700 From: gabi@tdsoft.com (Gabi) Organization: http://groups.google.com/ Message-ID: <5607ff2c.0208140043.d278f90@posting.google.com> Hi When I operate the following sequence: ipDetach (1,"fec"); ipAttach(1,"fec"); And not from the shell, I notice in memShow of a memory leak (6-12 bytes) in each sequence. Do I miss something? Gabi Segev Serusi gabi@tdSoft.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Wind Web Server - httpBlockPut Date: 21 Aug 2002 00:35:52 -0700 From: lchamberlain@pgaero.co.uk (Laura Chamberlain) Organization: http://groups.google.com/ Message-ID: References: "drdiags" wrote in message news:... > Just more questions. When the system stops responding, does it ever recover? No, the whole thing locks up (the main application stops too). If the watchdog is connected the system reboots. > When it stops responding, are any tasks suspended? See above - can't get debug tools in > Have you taken a look on WindSurf for any possible SPR fixes for your version of Wind Web Server? BTW, what version is this? Yes, was running WWS 2.0 Beta but have since upgraded to the released version and installed patch 1 from WindSurf, can't see any other similar problems > Maybe someone who has used this product may have their memory jogged if you provide this. > Could you print out the errno upon the ERROR return, this may help. HttpBlockPut only returns HTTP_OK or HTTP_ERROR, when the system dies or does a short download can't see any error message on the terminal - is there something else I should be looking at? > What version of Tornado, what architecture, BSP and network interface type? Tornado 2.0.2, VxWorks Ver 5.4, ADS860 BSP ver 1.2/5, MotFec Network using source 01i,28jun00 (waiting for later ver from windriver). The data to send is compiled from DiskOnChip (TFFS and DOSFS2) and FLASH. > How many sockets are open when the error occurs? Don't know once crashed cant get debug tools in but I will find out for short downloads. During normal operation you can get several connections in the TIME_WAIT state, occasionally you get too many and the system pauses while they free up, do you think this is releated to the crashing? Here is an extract from my configuration:- #define HTTP_MAX_SOCKQ 5 #define HTTP_MAX_SOCK_BUFF 4096 #define HTTP_SOCK_TIMEOUT 60 #define HTTP_SOCK_ARRAY 4 #define HTTP_MAX_GLOBAL_MEM_PART_SIZE 15360 #define HTTP_MAX_REQ_PART_SIZE 15360 #define HTTP_MAX_REQLINE_SIZE 1024 #define HTTP_MAX_ARGUMENTS 10 #define HTTP_MIME_MAXLINES 10 #define HTTP_NR_OF_TASKS 4 #define HTTP_TASK_STACK_SIZE 6114 > Sorry I don't have an answer, but if you share the above information, > someone will join in. Also, did you contact Wind River support about > this? Yes but there response has been slow and so far no useful suggestions, they are now asking me to provide a test case but this is going to take me some time to prepare. Thanks Laura > HTH....... > "Laura Chamberlain" wrote in message > news:efdaef76.0208190544.5ebd4eac@posting.google.com... > > Hi, > > > > Thanks for your response, I have built a webpage which shows the web > > server config, memory usage and the network stats you suggest along > > with tcpstatShow as it uses TCP, but I can't see anything wrong. Any > > other ideas? > > > > "drdiags" wrote in message > news:... > > > Hello Laura, > > > > > > Not sure I can help, but is it possible that you are running out of > > > network > > > resources, i.e. #file descriptors are maxed out. Are the transfers done > > > using > > > UDP or TCP (I don't have the product installed)? I would run the various > > > network debugging tools from the target shell: > > > > > > inetstatShow > > > mbufShow > > > netStackSysPoolShow > > > netStackDataPoolShow > > > endPoolShow "interface name" <<-- Look on vxWorks FAQ for > utilities.c > > > for this > > > routine. > > > iosFdShow > > > i > > > others that I cannot think of > > > > > > HTH. > > > > > > "Laura Chamberlain" wrote in message > > > news:efdaef76.0208160356.3b1636dc@posting.google.com... > > > > I am using a custom RPM to compile data from several places and send > > > > to the user as a single file, up to 40Mb in total, this is sent using > > > > httpBlockPut in blocks of 1024 - similar to the httpRpmDownload > > > > function in Tornado\target\unsupported\http\rpm\httpRpmUpload.c. > > > > > > > > The first problem is sometimes the httpBlockPut function fails > > > > (returns ERROR), resulting in an incomplete download - using internet > > > > explorer you don't receive an error just a small file. > > > > > > > > The second problem I have found while testing the first - after an > > > > hour or so of repeatedly downloading data the target stops responding > > > > and there is no debug output to help me determine why. > > > > > > > > Can anyone help? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Debug question Date: 21 Aug 2002 00:36:43 -0700 From: eyalk@radlan.com (Eyal) Organization: http://groups.google.com/ Message-ID: References: g_ch@21cn.com (GANGCHENG) wrote in message news:... > No,it is only the test example for using netbuflib . > > Why? > > "carria" wrote in message news:... > > I have encountered it also, maybe your application is time limited. > > > > "GANGCHENG" wrote in message > > news:ce273620.0208191708.3cbac770@posting.google.com... > > > Dear all, when I launch the Debugger under Tornado2 and debug the > > > program step by step, the following exception is shown in the target > > > shell . But when I directly run the program , there is no exception > > > and error. > > > > > > > > > Exception current instruction address: 0x00d8cf90 > > > Machine Status Register: 0x00089030 > > > Condition Register: 0x24000000 > > > Task: 0xd8d400 "tDbgTask" > > > > > > > > > why ? It seems to be a timing problem. You have to enter the assembly code of the problematic line and follow the registers' values and behaviour. Second time, try to do it with cache disable and see the consequence. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ipAttach ipDetach memory leak problem Date: Wed, 14 Aug 2002 08:59:34 +0000 (UTC) From: Johan Borkhuis Organization: Agere Systems Message-ID: References: <5607ff2c.0208140043.d278f90@posting.google.com> gabi@tdsoft.com (Gabi) wrote: > When I operate the following sequence: > ipDetach (1,"fec"); > ipAttach(1,"fec"); > And not from the shell, I notice in memShow of a memory leak (6-12 > bytes) in each sequence. Do I miss something? If you call this from the shell the string "fec" is declared somewhere in memory. But it is not freed, because the shell does not know if it is still in use. If youy want to avoid this you should create a string in memory (using malloc) and pass this string to these functions and free this agter you are finished. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How to trace memory access in VxWorks/Tonado? Date: 21 Aug 2002 01:15:09 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0208210015.3b71b336@posting.google.com> References: Yuan Zheng-ZYUAN1 wrote in message news:... > we found some one byte of > prestored CRC table has been corrupted. it seems you already have located the exact address of the corrupted byte. in case you're lucky enough your target cpu supports hardware data breakpoints, take advantage of them... toni --------------------------- Newsgroups: comp.os.vxworks Subject: how can i solve reboot(0x02)-problem Date: 21 Aug 2002 01:39:19 -0700 From: ssgaga@hanmail.net (Kang sungsuk) Organization: http://groups.google.com/ Message-ID: <456bd5cf.0208210039.43200b2d@posting.google.com> The bootString Value is change.. and reboot(0) - -> rebooting is ok.. but BOOT_PARAMS is not update. one update(10%), other not update(90%). for example, bootString ip value is changed, but ifAddrGet()is old value. if i turn off(power off) my equipment.. the BOOT_PARAMS value is update clearly so i doubt the reboot() function. i expect to my equipment re-read my BOOT_PARAMS area..after reboot() i think that excute-reboot(2): [memory clear and normal reboot] will satisfy my expect. when i excute reboot(2): communication interface is all off and not start. it is same to call sysToMonitor(2). cpu: mpc 860. how can i excute reboot(2).. and which other possiblity to solve the BOOT_PARAMS value not update. sysToMonitor function: STATUS sysToMonitor ( int startType /* parameter passed to ROM to tell it how to boot */ ) { FUNCPTR pRom = (FUNCPTR) (ROM_TEXT_ADRS + 4); /* Warm reboot */ *CIMR(vxImmrGet()) = 0; /* disable all cpm interupts */ sysCpmEnetDisable (0); /* disable the ethernet device */ sysSerialReset(); /* reset the serail device */ (*pRom) (startType); /* jump to bootrom entry point */ return (OK); /* in case we ever continue from ROM monitor */ } --------------------------- Newsgroups: comp.os.vxworks Subject: "Virtual" end driver Date: 21 Aug 2002 02:49:48 -0700 From: Gunter.Ritter@tenovis.com (Gunter Ritter) Organization: http://groups.google.com/ Message-ID: Hello VxWorkers, I have written a "virtual" end driver as a secondary end driver for our system. "Virtual" means that this end driver has no "physical" ethernet hardware. The primary end driver is working with ethernet hardware. I need this secondary end driver only to exchange ip packets with the TCP/IP stack of vxworks and let vxworks do the routing between the two networks. I have a problem when I try to assign an ip address. muxDevLoad(), muxBind(), ipAttach() and ifMaskSet() work fine. ifShow() shows me the settings of the secondary driver. But calling ifAddrSet() with an unique ip address causes the network functionality of vxworks to crash. Even "ping" does no more work with the primary end driver after calling ifAddrSet(). Does anybody can give me a hint whats going wrong with my secondary end driver? Thanks. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: tftpCopy disable context switching ? Date: 21 Aug 2002 03:19:33 -0700 From: shy017@yahoo.com (Shay) Organization: http://groups.google.com/ Message-ID: <8a5dc325.0208210219.54e0f8f2@posting.google.com> References: <8a5dc325.0208200105.355c76b3@posting.google.com> <3D621885.E09703AD@bellsouth.net> Indeed, it is task priority problem. Thanks, Shay. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Command-line compiler problems with Tornado 2.1/NT for MIPS Date: Wed, 14 Aug 2002 22:34:47 GMT From: "Digital Man" Organization: Prodigy Internet http://www.prodigy.com Message-ID: <3D5AE8AD.1376.UVXWORKS@vert.synchro.net> References: <3D5A0E24.1360.UVXWORKS@vert.synchro.net> To: john_94501@yahoo.com (John) Re: Re: Command-line compiler problems with Tornado 2.1/NT for MIPS By: john_94501@yahoo.com (John) to comp.os.vxworks on Tue Aug 13 2002 11:40 pm > You need to read the VxWorks Programmer's Guide and the Tornado > documentation. You are not using the tools in the correct way at all > for VxWorks development. > > Unlike Windoze or unix, VxWorks does not use executables (which is > what c++mips was trying to generate for you since you gave it no > options). Instead, simple .o files are downloaded onto the target > system that is already running the VxWorks kernel. The loader that > performs that step will resolve the unresolved symbols with those > already in the kernel image. Actually, I already knew this and have been using Torando and VxWorks for quite some time. It's just in converting an existing Linux project to VxWorks, I simply forgot that the linking stage was unnecessary and is handled when downloaded to the target. Thanks for the reminder! digital man - --- Synchronet 3.10j-Win32 NewsLink 1.29 * Vertrauen - Fullerton, California - telnet://vert.synchro.net --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Continue from the same point after power fail (PPC VxWorks 2)? Date: Thu, 15 Aug 2002 11:47:08 +0000 (UTC) From: David Laight Organization: BT Openworld Message-ID: <3D5B956E.1040502@spamm.me.l8s.co.uk> References: <3D5B7083.3020908@spamm.me.l8s.co.uk> Leonid Rosenboim wrote: > David, I agree with you completely, except that I think that well designed > compact Flash controller should be able to handle power loss anytime, and be > able to recover from that. > > Because you can't write on Flash location perviously written, only on > previously erased block, the internal structure of any Flash media is akin > to journal, and every sector is written in a new physical location, then > some metadata is updated to indicate that new version of that logical sector > appeared. Old version will be subsequently destroyed and block erased to > make room for new writes. > > I am curious as to what make of CF cards had problems with power remoival, > so I will know which card not to use in my digial camera. I'm not sure that was the cause, but we got some CF cards (under vxworks and dosfs1) that got COMPLETELY stuffed occaisionally. Unfortunately attempting to reproduce the problem failed (as always). I dumped out a duff card and tried to put the filesystem back together by hand (ok by writing a program). However a hex dump showed why the program was having trouble - blocks of the fat were replicated over random sectors. Indeed (IIRC) even over the sectors between the start of the filesystem and the FAT. Some of these sectors are never written to, so my best guess was that the CF card did it all on its own. The cards were badged sandisk or premier (forgotten which). None of them met the CF spec - we had three different broken responses to the 'identify drive' command. David --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How to trace memory access in VxWorks/Tonado? Date: Wed, 21 Aug 2002 11:58:13 GMT From: sylvain.gagnon@mindready.com (Sylvain Gagnon) Organization: Bell Sympatico Message-ID: <%fL89.19543$_75.1013222@news20.bellglobal.com> References: Reply-To: sylvain.gagnon@mindready.com In article , Yuan Zheng-ZYUAN1 wrote: >All- > >In our application code, we have a problem like this: Our CRC check seems >not working right and after careful checking, we found some one byte of >prestored CRC table has been corrupted. > >I am wondering who has the experience to detect which pointer corrupted the >CRC table? > >Thanks a lot! > Windriver provide some tools to trace code. It is named TraceScope, ProfileScope and MemScope. Those tools could help you Regards, Sylvain G. Mindready Solutions --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How to make ccppc still use user-defined lables after compilation? Date: 21 Aug 2002 05:16:29 -0700 From: seado@21cn.com (Seado) Organization: http://groups.google.com/ Message-ID: References: "Liang Yang" wrote in message news:... > Hi, > > I insert some c labels into .c source file as follows: > > begin: > i = 1; > end: > > But after I use ccppc to compile this .c file into a .s > file,all these labels are replaced with ccppc format lables > .Lxx:(such as .L01:, .L02:). Define these labels as global symbols, a fissible method is to use inline assembly. Regards, Seado --------------------------- Newsgroups: comp.os.vxworks Subject: Re: debug interrupt Date: 21 Aug 2002 05:27:21 -0700 From: philippe.faure@techn.alstom.com (Faure) Organization: http://groups.google.com/ Message-ID: References: <7efa47da.0208192055.5558344a@posting.google.com> liao@ztenet.com (ao.li) wrote in message news:<7efa47da.0208192055.5558344a@posting.google.com>... > I want to know how to let my project be the system debug mode! Under tornado 2.0.2 (for Windows): launch the debugger: Tools/Debugger attach to system: Debug/ Attach (Alt F6) / system place a GLOBAL breakpoint in yr code: Debug /Toogle Global breakpoint (Shift F9) Note: I have some difficulties to debug in system mode, there are some bugs with that mode under Tornado 2.0.2: SPR#72141, T202: system mode debugging fails after some time / target hangs) if it is sufficient,I advice you to debug in task mode (not in system mode), it is what I choosed to do after spending too much time to try to make work without success the system mode. Speaking for myself, Philippe --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Does anyone have a ping program, similar to Sisco's catalyst 2900, available under vxWorks? Date: Wed, 21 Aug 2002 14:29:32 +0100 From: Paul Caswell Organization: Land & Sea Systems Message-ID: <3D6395BC.38A05ADC@baesystems.com> References: <8266a3fc.0208202304.458b796c@posting.google.com> If you are using Tornado 2 you can simply include the ping component in the project view and build. You should then be able to ping the outside world from your target. Regards Paul --------------------------- Newsgroups: comp.os.vxworks Subject: syslog in vxWorks Date: 21 Aug 2002 06:31:01 -0700 From: cwnam@korealink.ne.kr (changwoo) Organization: http://groups.google.com/ Message-ID: <3d64ee1a.0208210531.3a451015@posting.google.com> vxWorks.. I need the syslog for logging messages to UNIX. I tried it as follow which was mentioned before but it didn't work. " Open a UDP socket (port 514) to the syslogd running on your unix host and pass the socket's file descriptor to logInit() - no need to 'modify' logLib. " " /* revised code */ if((sFd = socket(AF_INET, SOCK_DGRAM, 0)) == ERROR) { printf("socket\n"); } sockAddrSize = sizeof(struct sockaddr_in); bzero((char*) &serverAddr, sockAddrSize); serverAddr.sin_len = (u_char)sockAddrSize; serverAddr.sin_port = htons(514); serverAddr.sin_addr.s_addr = inet_addr("200.100.12.244"); if (connect(sFd, &serverAddr, sockAddrSize) < 0) { printf("error connect syslog socket\n"); } logInit (sFd, MAX_LOG_MSGS); /* initialize logging */ " If you know how to solve this or have the source code about syslog client, please send me a message. thanks --------------------------- Newsgroups: comp.os.vxworks Subject: DHCP client Date: Wed, 21 Aug 2002 13:38:18 +0200 From: "Øyvind Holmeide" Message-ID: <3d637c51$1@news.broadpark.no> I have tried to include DHCP client functionality (DHCPv4 runtime client), but then SNMP deamon support must be added as well, which I do not have. Is SNMP required for DHCP to work? If yes, can someone propose a DHCP freeware alternative? BR Øyvind Holmeide --------------------------- Newsgroups: comp.os.vxworks Subject: i have trouble connecting to target server with tornado Date: Wed, 21 Aug 2002 16:49:43 +0200 From: "tomer" Organization: The Israel Inter-University Computation Center Message-ID: when I click the launch button at the tools\targetserver\configure I experience this massage tgtsvr (nnlabrouter@nnws-1): Wed Aug 21 16:47:04 2002 Wind River Systems Target Server: NT/Win95 version Connecting to target agent... Error: rpccore backend client Timed out failed. can anyone tell me what is the source of this error.... thanks tomer --------------------------- Newsgroups: comp.os.vxworks Subject: Re: syslog in vxWorks Date: Wed, 21 Aug 2002 17:42:21 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <3d64ee1a.0208210531.3a451015@posting.google.com> THis did not work because syslog has a rather simple protocol on top of UDP, the packet needs to begin with a few headers, including process id, message facility and priority, timestamp, and only after those comes the message text. If you insist on itegrating your syslog client with logLib, you need to pass logTask a file descriptor of a PIPE, and have your own task read the messaes from the pipe, add headers etc., and send it over to the syslog server. "changwoo" wrote in message news:3d64ee1a.0208210531.3a451015@posting.google.com... > vxWorks.. > > I need the syslog for logging messages to UNIX. > I tried it as follow which was mentioned before but it didn't work. > > " > Open a UDP socket (port 514) to the syslogd running on your unix host > and pass the socket's file > descriptor to logInit() - no need to 'modify' logLib. > " > " > /* revised code */ > if((sFd = socket(AF_INET, SOCK_DGRAM, 0)) == ERROR) > { > printf("socket\n"); > } > sockAddrSize = sizeof(struct sockaddr_in); > bzero((char*) &serverAddr, sockAddrSize); > serverAddr.sin_len = (u_char)sockAddrSize; > serverAddr.sin_port = htons(514); > serverAddr.sin_addr.s_addr = inet_addr("200.100.12.244"); > if (connect(sFd, &serverAddr, sockAddrSize) < 0) > { > printf("error connect syslog socket\n"); > } > logInit (sFd, MAX_LOG_MSGS); /* initialize logging */ > " > > If you know how to solve this or have the source code about syslog client, > please send me a message. thanks --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Wind Web Server File Uploads Date: Thu, 15 Aug 2002 12:48:31 -0400 From: david lindauer Message-ID: <3D5BDB5F.61B29529@notifier-is.net> References: I don't know internals well enough to know the details of how to upload, but it seems like you ought to be able to do *something* by changing the access RPM. I don't know what the difference in our version of wind web server but mine has an httpUploadConfInit followed by RPM configuration for it in httpconf.c. (I would assume that this actually defines the upload process rather than the end of the upload process, but maybe I am wrong). But since then they may have moved this functionality to the configlettes. ( see the directory \tornado\target\config\comps\src\net\http for example) David Laura Chamberlain wrote: > I am trying to upgrade from the Beta version of wws 2 to the released > version but I can't get file uploads to work as I want. > > I have a custom RPM which uploads a binary file to a temporary area, > checks it's valid and then programs it to flash. > > What now happens is the file is uploaded to the location specified by > HTTP_POST_FILE_UPLOAD_PATH retaining the name specified in the upload > form. Once the whole file has been uploaded my RPM is then called. > This is no good because :- > 1. A file of any name can be uploaded - how does my RPM know what it > is called in order to process it and remove it if invalid > 2. I want to give the user feedback as the file is uploaded (it is > >1Mb and takes a while). > > How can I stop the file being automatically uploaded so my RPM can do > it (as it did before)? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Ethernet Driver for the MPC8260 FEC ves CPM37 Errata Date: Thu, 15 Aug 2002 09:56:46 -0700 From: "Dan Gold" Organization: Mailgate.ORG Server - http://www.Mailgate.ORG Message-ID: References: <5607ff2c.0208150702.31dff97e@posting.google.com> Sender: gold@ensemble.com "Gabi" wrote in message news:5607ff2c.0208150702.31dff97e@posting.google.com... > Hi, > I'm using the MPC8260 under VxWorks, and I wonder how you guys faces > the need to handle the Motorola Errata CPM37 which are relevant > directly to the MPC8260 FEC. If you're using the real vxWorks 8260 BSP, it looks like they stop and then re-start the entire FCC in case of any transmit errors. If you've done your own 8260 FCC driver, then you must handle it. You only need to re-initialize the Tx side. I.e. clear GFMR[ENT], clean up the TxBD list, re-init the Tx parameters, and issue INIT_TX. If you get the errata list from the Mot web site, the errata describes what needs to be done. hth, Dan Gold dgold1@san.rr.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Changing a password / loginUserDelete() doesn't free memory Date: Wed, 14 Aug 2002 19:43:03 +0100 From: "cecil" Message-ID: <1029350596.19299.0.nnrp-01.d4e4b8d1@news.demon.co.uk> References: <1029268579.10443.0.nnrp-01.d4e4b8d1@news.demon.co.uk> <488e459a.0208132258.1870f4de@posting.google.com> Thanks, that seems to work "John" wrote in message news:488e459a.0208132258.1870f4de@posting.google.com... > Hello, > > The password string will not be freed by the library. It is actually > built on top of symLib, and deleting a user just performs a > symRemove() on the user name (the password is stored as a pointer to > the string passed in loginUserAdd(), so freeing it might not be > possible since it might not be dynamically allocated memory). > > So, to change the password you simply need to change the original > string passed to loginUserAdd(). Assuming that you can find this > string, changing the password should be simple. > > HTH, > > John... > > PS You could also try asking Wind River support or your local FAE for > the source file; then you could make whatever changes you like to the > password and username management. > > "cecil" wrote in message > news:<1029268579.10443.0.nnrp-01.d4e4b8d1@news.demon.co.uk>... > > Does anybody know how to change a password without deleting the user and > > adding another? > > I'm not keen on using loginUserDelete() the loginUserAdd() because it doen't > > seem to free up the memory it's used. > > Anybody got any thoughts? > > Nigel --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Strange files appearing on the hard drive Date: 14 Aug 2002 14:35:51 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: References: <95ebbe31.0208130848.27aead6f@posting.google.com> <2Cw69.5098$fm.2387@newssvr19.news.prodigy.com> Sender: bpringle@DeadDuck Does anyone know where to find Tornado 2.0.1 patches on WindSurf? It seems that there are many patches available for Tornado 2.0.2, but there are none for the ARM processor. Regards, Bill Pringlemeir. - -- Don't you realise that perverted people are stupid? How can you be so braindamaged? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: debug interrupt Date: Wed, 21 Aug 2002 18:26:25 +0100 From: Wade Oram Organization: None Message-ID: <2UYH+bGB18Y9EAb0@wtoram.co.uk> References: <7efa47da.0208192055.5558344a@posting.google.com> Reply-To: Wade Oram In message , Faure writes > >Note: I have some difficulties to debug in system mode, there are some >bugs with that mode under Tornado 2.0.2: >SPR#72141, T202: system mode debugging fails >after some time / target hangs) > I have successfully used system mode in T2 CP2 (Windows NT4 host, X86 target) with additional patches although I have not had to use it for extended periods. How long is 'after some time'? >if it is sufficient,I advice you to debug in task mode (not in system >mode), >it is what I choosed to do after spending too much time to try to make >work without success the system mode. The original poster wanted to set a break point in an interrupt handler. For this purpose, only system mode can be used. However, the advice still stands - find out as much information as you can about the fault in task mode and restrict system mode debugging to the occasions that really demand it. If an interrupt handler really is causing you grief and system mode debugging is not reliable enough to solve the problem, you *may* be able to change the interrupt handling so that the interrupt handler does nothing except give a semaphore to a (very) high priority task which does the interrupt handling. This will then allow you to put a break point in the original interrupt handler whilst still in task mode. Once the fault is found the original interrupt handler - hopefully now fixed - - can be restored to its rightful place. As a technique, it won't be applicable all problems with interrupt handlers but it can be used to help solve many - especially when the interrupt in question is independent of any other interrupts in the system. - -- Wade Oram --------------------------- Newsgroups: comp.os.vxworks Subject: Make problem w/Torando II, parameter is incorrect Date: 14 Aug 2002 15:50:34 -0700 From: emil@lapel.com (Emil Rojas) Organization: http://groups.google.com/ Message-ID: <754b62e0.0208141450.bee3a43@posting.google.com> I just installed Tornado II on a PC, and when I try to complie the first tutorial I get the following output. - --------------------------------------- ccsimpc -U__WINNT -UWIN32 -U__WINNT__ -UWINNT -U__MINGW32__ -U_WIN32 - -U__WIN32 -U__WIN32__ -U__WIN32 -mpentium -ansi -nostdinc -g -nostdlib -fno-builtin - -fno-defer-pop -Wall -DRW_M ULTI_THREAD -D_REENTRANT -I. -IC:\Tornado\target\h -DCPU=SIMNT -c C:\Tornado\target\src\de mo\start\cobble.c C:\Tornado\target\src\demo\start\cobble.c: In function `monitor': C:\Tornado\target\src\demo\start\cobble.c:322: warning: unused variable `isNot' vxrm C:\Tornado\target\proj\Project1\prjObjs.lst Generating C:\Tornado\target\proj\Project1\prjObjs.lst... fatal error in make: The parameter is incorrect. make: *** [C:\Tornado\target\proj\Project1\prjObjs.lst] Error 0x57 - ------------------------------------- Any clues greatly appreicated. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: file level redundancy Date: 14 Aug 2002 16:06:57 -0700 From: emil@lapel.com (Emil Rojas) Organization: http://groups.google.com/ Message-ID: <754b62e0.0208141506.464443da@posting.google.com> References: <1029252744.708216@azores.network-i.net> "Dimitrios Siganos" wrote in message news:<1029252744.708216@azores.network-i.net>... > Hi, > > does anyone know of a file level redundancy solution that works over IP? > Or has anyone worked on such project and is willing to offer some tips? > I believe that there are commercial solutions to this problem that are readily available. You might use google to search on some of your key words. I think that Storage Area Network (SAN) products would address what I think you are asking about, but may be overkill. If this doesn't make sense, then you should try restating your problem. emil --------------------------- Newsgroups: comp.os.vxworks Subject: Re: wdbSystemSuspend : linker error Date: 14 Aug 2002 16:23:51 -0700 From: rishi_bhargava@yahoo.com (Rishi Bhargava) Organization: http://groups.google.com/ Message-ID: <2474871.0208141523.4f2375@posting.google.com> References: <2474871.0208081254.16b73a2f@posting.google.com> <488e459a.0208082101.2cea1ca3@posting.google.com> Tried the following things.. but still get the same error... any suggestion as to what else should I try... thanks, Rishi john_94501@yahoo.com (John) wrote in message news:<488e459a.0208082101.2cea1ca3@posting.google.com>... > Hello, > > You don't have the WDB system mode debugging support in your kernel. > Check for INCLUDE_WDB_SYS and also that you have the WDB connection > set to WDB_COMM_END or WDB_COMM_SERIAL (the older WDB_COMM_NETWORK > excludes system mode support as it does not support polling mode). > > HTH, > > John... > > rishi_bhargava@yahoo.com (Rishi Bhargava) wrote in message news:<2474871.0208081254.16b73a2f@posting.google.com>... > > I tried to include the call to wdbSystemSuspend in my code as I want > > to suspend the system at an early stage so that I can conenct the > > target and start a debugger. When I include this I get a linker error > > : undefined reference ... > > > > I am linking the libPPC860gnuvx.a. Can anybody suggest what could be > > goinf wrong .. any include definitions I need?? > > > > thanks > > Rishi --------------------------- Newsgroups: comp.os.vxworks Subject: 8260 BSP autosizing memory? Date: 14 Aug 2002 16:35:39 -0700 From: jchen@terayon.com (john) Organization: http://groups.google.com/ Message-ID: hi everyone! Does anyone know if the Windriver supplied MPC8260 BSP supports autosizing memory? Just curious. We are currently using hard coded memory sizes. That means when the memory changes the code will need to change as well.... Also, if it does support LOCAL_MEM_AUTOSIZE, is it just a one line add in config.h to #define LOCAL_MEM_AUTOSIZE? what are the implications? thanks john --------------------------- Newsgroups: comp.os.vxworks Subject: Re: file level redundancy Date: Thu, 15 Aug 2002 02:00:45 +0100 From: Dimitrios Siganos Message-ID: <3D5AFD3D.6040301@engineer.com> References: <1029252744.708216@azores.network-i.net> <754b62e0.0208141506.464443da@posting.google.com> I guess I have not made clear that the hardware is fixed. I have two PPC7410 processors, each connected to a SCSI bus. They both act as file servers. The servers work as a redundant pair - when one goes down the other must take over without any service disruptions. So what I need is the software to run on the hardware that I already have. What you are proposing is a hardware/software solution, which is not what I was looking for. Thanks for you reply though. Dimitris Emil Rojas wrote: > "Dimitrios Siganos" wrote in message news:<1029252744.708216@azores.network-i.net>... > >>Hi, >> >>does anyone know of a file level redundancy solution that works over IP? >>Or has anyone worked on such project and is willing to offer some tips? >> > > I believe that there are commercial solutions to this problem that are > readily available. You might use google to search on some of your key > words. I think that Storage Area Network (SAN) products would address > what I think you are asking about, but may be overkill. If this > doesn't make sense, then you should try restating your problem. > > emil --------------------------- Newsgroups: comp.os.vxworks Subject: what's wrong with my cache flush function? Date: 14 Aug 2002 18:28:49 -0700 From: du.jianjun@zte.com.cn (du) Organization: http://groups.google.com/ Message-ID: <21a3d7e1.0208141728.1e703881@posting.google.com> I am usring Intel 82559er ethernet pci controller,because it use DMA to transport data,so cache and memory appear disconherence,the 82559 can not work properly.So I write a cache flush function,but it can not work properly.My cpu is MPC7410.When I disable cache of 7410,the 82559 can work well.So I think my cache flush function is not correct,follow is my function,who can help me correct it?I need your help,thanks. LOCAL void DCACHE_FLUSH(int staddr,int length) { __asm__("lwz 3,%0"::"m"(staddr) ); /*r3 = flush address*/ __asm__("lwz 5,%0"::"m"(length) ); /*r5 = flush length*/ __asm__("xor 4,4,4" ); __asm__("flush_7410: cmpw 5,4"); __asm__("ble flush_7410_done"); __asm__("dcbf 3,4"); __asm__("addi 4,4,0x20"); __asm__("b flush_7410"); __asm__("flush_7410_done:"); } --------------------------- Newsgroups: comp.os.vxworks Subject: FTP Date: Thu, 15 Aug 2002 05:02:59 GMT From: "weddick" Organization: AT&T Broadband Message-ID: <7CG69.35670$me6.5347@sccrnsc01> We just upgrade our BSP to the pentium pro. The FTP server does not seem to work the way it did before. Previously, we could use the windows ftp client and see the files on /ide0/. Now the system wants to connect to host: and I have to change the directory to /ide0/. Our Exceed FTP client won't work at all. Anyone else have this problem? Thanks, --------------------------- Newsgroups: comp.os.vxworks Subject: tPortmapd out of memory Date: Wed, 21 Aug 2002 19:13:26 GMT From: Raymond Kristiansen Organization: Kongsberg Spacetec AS Message-ID: <3D63E655.8F923A74@spacetec.no> Hello all, I am using Tornado 2.0 for development of my vxWorks system running on ppc with BSP mv2400. I have currently a bootable project that is loaded by a bootrom over a network connection. When I run this image over a period of time (typical 2 hours), it finally comes up with the error message: 0x1b8c888 (tPortMapd): panic: rpc: svc_getreq: out of memory I understand that this indicates that my system has some memory problems in some way. But is it the tPortMapd task itself that causes a memory problem? My first thought is that my application code has a memory leak in it somewhere, and that the tPortMapd task run low of memory because one of my application tasks eats up the memory over time. Comments, anyone? Regards, Ray --------------------------- Newsgroups: comp.os.vxworks Subject: Re: DHCP client Date: Wed, 21 Aug 2002 19:15:31 GMT From: David Cooper Organization: none Message-ID: References: <3d637c51$1@news.broadpark.no> If you have DHCP, then you should have everything needed to include DHCP. You might want to run your Tornado installation again and insure everything is selected. ...david - -- +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ ô¿ô Genealogy Utilities & Family Research Researching: ALLEN,BERZONER,COHEN,COOPER,KUPFERZMIDT, POSNER (POIZNER), PRITIKIN, STARKMAN Dachau Camp Entry List, Genealogy Utilities http://home.adelphia.net/~dcooper000 In article <3d637c51$1@news.broadpark.no>, oeyvind@ontimenet.com says... > I have tried to include DHCP client functionality (DHCPv4 runtime client), > but then SNMP deamon support must be added as well, which I do not have. Is > SNMP required for DHCP to work? If yes, can someone propose a DHCP freeware > alternative? > > BR > > Øyvind Holmeide > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: i have trouble connecting to target server with tornado Date: Wed, 21 Aug 2002 19:17:44 GMT From: David Cooper Organization: none Message-ID: References: Some dumb questions; but, need to ask: (1) do you have WDB included in your image ? (2) Do you have the END Driver selected for WDB in the image ? ...david - -- +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ ô¿ô Genealogy Utilities & Family Research Researching: ALLEN,BERZONER,COHEN,COOPER,KUPFERZMIDT, POSNER (POIZNER), PRITIKIN, STARKMAN Dachau Camp Entry List, Genealogy Utilities http://home.adelphia.net/~dcooper000 In article , bahart@bgumail.bgu.ac.il says... > when I click the launch button at the tools\targetserver\configure I > experience this massage > > tgtsvr (nnlabrouter@nnws-1): Wed Aug 21 16:47:04 2002 > Wind River Systems Target Server: NT/Win95 version > Connecting to target agent... > Error: rpccore backend client Timed out > failed. > > > > can anyone tell me what is the source of this error.... > > > thanks > > tomer > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: howto: locating vxworks to "alternative" flash address for PowerPC Date: Thu, 15 Aug 2002 07:08:54 GMT From: "Garym" Organization: AT&T Broadband Message-ID: References: <2aac1ba4.0208010743.1b05087a@posting.google.com> bootInit.c is where the boot loader invokes the deflashing of the main image. It uses the ROM_xxx equates to locate the base of the flash image, along with the assumption that it follows the boot loader image, but this can really be any address you like. You can then take the main image by itself (which doesn't care where it is stored in flash) and get that programmed in your preferred flash region. You then only have to worry about getting the boot loader programmed into the reset vector. In the PPC8xx world, the boot loader is prepended to the main image, and all programmed into the same flash device, which is physically mapped to the reset vector. "flingylingy" wrote in message news:2aac1ba4.0208010743.1b05087a@posting.google.com... > I want to build my vxworks image so that it is locatable (in flash) to > an address of my choice. I'll have a bootloader that does a system > check and a few speciailized things (probably DINK32 variant), and > then passes control to vxworks for loading and startup. I want to use > the uncompressed version of vxworks if possible, which should load > itself into sdram and start running. > > Is there a straightforward way to do this? I'm using T2.0.2 and the > gnu toolchain. I'm looking for ideas on how to do this. I don't care > if I have to revector romInit to point to my bootloader, but I just > don't know how to make vxworks resolve to the "alternative" address > (in another region of flash memory) so that it knows where its stuff > is, so that it can do its checksum and load process properly. > > thx for any info --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Continue from the same point after power fail (PPC VxWorks 2)? Date: Thu, 15 Aug 2002 07:23:47 GMT From: "Garym" Organization: AT&T Broadband Message-ID: <6GI69.145452$uj.210940@rwcrnsc51.ops.asp.att.net> References: In order to completely preserve the machine context at the time of a power failure, you will need some hardware assistance which will generate an external interrupt when power is going down. This circuit will guarantee you sufficient time to preserve the machine context (all registers and task data in SDRAM) which can be restored on power up. You could also save just enough of the current operating context so that you do a sufficient restore after a power failure so as to make it indistinguishable from the complete restore described above. You can also consider the "Self-refresh" option that many SDRAM devices have, in which the SDRAMs preserve their contents with very low power and no requirement for external refresh. This may be simpler but would require more battery capacity. You can certainly use the linker to relocate data segments from code segments, and write some tools to split the final binary image into code and data segments. "Haim Hillel" wrote in message news:aiqno7$bei$1@newshost.mot.com... > For environment of PPC 855/860/8260 & VxWorks 2. > I think to put the image (VxWorks app. And Data) lay compressed in flash, at > startup the code will decompress into SDRAM (cheap memory) and the data will > direct to SRAM which backup by battery, when the power fail to keep also the > Cpu reg. > For recover after power fail restore the code from the flash (with out > reinitialize the data) and load the Cpu reg. from the SRAM, then continue to > run. > > Is someone having another idea? Is someone having any code, examples or > algorithm? > or even idea how to build and put into vxWorks_romCompress a split sections > for Code and Data? > > Best Regards > Haim Hillel > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: 8260 BSP autosizing memory? Date: 15 Aug 2002 01:50:19 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0208150050.347f7d64@posting.google.com> References: Hello, You need to write the piece of code in the BSP that works out the size of the RAM and adjusts things accordingly. The following tech note explains what you need to do to implement this feature from the BSP's perspective. Obviously, detecting the size of RAM is something that you'll need to work out for your hardware. http://www.windriver.com/windsurf/appnotes/ide/tornado/t1/APN42/WTN42.html You might try some searches on WindSurf for LOCAL_MEM_AUTOSIZE and see if there are any examples of how to do this there. Normally, the autosizing reads the size information from serial PROMs connected to a I2C bus or likewise (most DRAM memory modules have these on them; most soldered down RAM systems do not). You could also try probing, checking jumpers on the board etc, etc. HTH, John... jchen@terayon.com (john) wrote in message news:... > hi everyone! > Does anyone know if the Windriver supplied MPC8260 BSP supports > autosizing memory? Just curious. We are currently using hard coded > memory sizes. That means when the memory changes the code will need > to change as well.... > > Also, if it does support LOCAL_MEM_AUTOSIZE, is it just a one line add > in config.h to #define LOCAL_MEM_AUTOSIZE? what are the implications? > > thanks > john --------------------------- Newsgroups: comp.os.vxworks Subject: How to use a different 'target' directory with Tornado? Date: Wed, 21 Aug 2002 21:27:06 +0200 From: "Werner Schiendl" Message-ID: <3d63e9ca$1@brateggebdc5.br-automation.co.at> Hello, for some reason I'd need to use a different 'target' directory for a special project (bootable VxWorks project). I'm using Tornado 2.0 on a Windows XP host, VxWorks 5.4 for I80486 and PENTIUM targets. The entire 'target' directory is together in one place, located on a local harddrive but it is _not_ located in the tornado installation directory. The TGT_DIR macro seems to be made for this exact purpose and specifying it I am able to build the current state of the project. However, the Tornado environment seems to be insist on using the files in $(WIND_BASE)/target and I have not seen any possibility to change that. I already changed the project file (.wpj) and replaced all occurences of $(WIND_BASE)/target with the absolute path. That way I'm able to open the project with Tornado and build it - but I cannot change the VxWorks configuration because it still uses the files in $(WIND_BASE)/target to check for installed features and dependencies. Does anyone know how I can make Tornado use a different 'target' directory? Thanks in advance! Werner --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Design ideeas on shared memories Date: Wed, 21 Aug 2002 14:21:04 -0500 From: Andrew Johnson Organization: APS, Argonne National Laboratory Message-ID: References: <25c1993e.0208160613.81842c9@posting.google.com> <3D62555D.B52D0646@baesystems.com> <25c1993e.0208201248.7b58bd@posting.google.com> Subrahmanyam Arya wrote: > What i want to know is wether the primitive calls are sufficent or do > we need to develop this type of locks . My application is kind of > simple having about 5 processes accessing the shared memories , one > does the right and the rest all do the read . Semaphores are probably the right tool for your task, just make sure that you use the right kind: 1. A vxWorks binary semaphore is really an event announcement - an ISR or task that wants to announce that the event just occurred calls semGive(), and the task that is waiting for the event will then return from its semTake() call. When the task waiting for the event has finished processing the data it can then re-arm the system for the next event. 2. If your events are not necessarily synchronized between the two tasks and you want to be sure that the event reading task gets exactly as many events as the sending task gave (even though the reader might temporarily get behind in processing those events), you should use a counting semaphore. If you need to pass the data between the two though, use a message queue or ring buffer instead. You can also use a counting semaphore within a resource allocator. 3. To protect a resource such as a structure in shared memory against simultaneous modification by multiple tasks, you probably want to use a mutex semaphore, which the task accessing the structure must take before doing so and release after it has finished. 4. Where you have a single writer task and multiple readers a mutex isn't quite so good, as all your readers can be working at the same time without interfering with each other. VxWorks doesn't provide a suitable semaphore for this kind of protection though, you have to make up your own out of the above primitives. In some circumstances you might not even need to use a semaphore to protect shared memory, but only if the data you're modifying is small enough to fit completely in a single machine word so the data structure can never be in an invalid state (i.e. only partially updated). If you're not sure about this, don't use the approach. - - Andrew - -- Edsger W. Dijkstra, 1930-2002: "The use of COBOL cripples the mind; its teaching should, therefore, be regarded as a criminal offence." --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Vxworks Drivers for 8260 MCC and HDLC SCCs Date: Wed, 21 Aug 2002 13:07:38 -0700 From: "Dan Gold" Organization: Mailgate.ORG Server - http://www.Mailgate.ORG Message-ID: References: Sender: gold@ensemble.com "Matchett, John" wrote in message news:ak0ld1$knl$1@overload.lbl.gov... > I am in need of MCC and SCC-HDLC Drivers for the Mot 8260. The reference board > my company bought only had a BSP that included UART and Ethernet. Does anyone > have any recommendations . Go to the Mot website and download their *examples*. Of course they won't be specific to your BSP (you'll have to port them) but it's a start. Don't look for things like performance, configurability, etc... They're just examples. hth, Dan Gold dgold1@san.rr.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Continue from the same point after power fail (PPC VxWorks 2)? Date: Thu, 15 Aug 2002 10:36:49 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Haim, What you are trying to do is in fact far more complex than seems at first, because saving the entire memory image and CPU registers can restore the CPU state, but the hard part is the assorted peripherals, which would too need to be restored, or re-initialized to a working state following this recovery. For this, every driver must have a function to safely re-initialize the chip to a working state without affecting any of the software state, and none of the VxWorks drivers I have seen can do this - not even the couter-timer or serial drivers in the BSP ! "Haim Hillel" wrote in message news:aiqno7$bei$1@newshost.mot.com... > For environment of PPC 855/860/8260 & VxWorks 2. > I think to put the image (VxWorks app. And Data) lay compressed in flash, at > startup the code will decompress into SDRAM (cheap memory) and the data will > direct to SRAM which backup by battery, when the power fail to keep also the > Cpu reg. > For recover after power fail restore the code from the flash (with out > reinitialize the data) and load the Cpu reg. from the SRAM, then continue to > run. > > Is someone having another idea? Is someone having any code, examples or > algorithm? > or even idea how to build and put into vxWorks_romCompress a split sections > for Code and Data? > > Best Regards > Haim Hillel > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How to access complete snmp ObjectID from get, next, set, text Date: 15 Aug 2002 07:32:02 -0700 From: akiggins@celitesystems.com (andrew) Organization: http://groups.google.com/ Message-ID: <2eced96.0208150632.7145696d@posting.google.com> References: Frank, I think you actually can get this from the varbind structure, I confess I haven't looked with the debugger to see if it's actually valid or not. If you have void BlahEntryGet(OIDC_T lastmatch, int tcount, OIDC_T *tlist, SNMP_PKT_T *pktp, VB_T *vbp) { ... } vb_obj_id in the VB_T * vbp looks like a suspect. andrew typedef struct VB_S { struct VB_S * vb_link; /* For the user to use, usually */ /* to build a list of related */ /* VB_Ts. */ ALENGTH_T vb_seq_size; /* Overall length of the VarBind */ /* sequence when encoded. */ OBJ_ID_T vb_obj_id; /* Object id for this VarBind */ OCTET_T vb_data_flags_n_type; /* Class form & type of data */ OCTET_T vb_flags; /* See below */ ALENGTH_T vb_data_length; /* Length of the data when encode */ MIBLOC_T vb_ml; /* The mib-leaf associated with */ /* this VarBind. */ union { INT_32_T v_number; /* Integer kinds of things */ UINT_32_T v_counter; /* Unsigned int things */ EBUFFER_T v_string; /* Octetstring things */ OBJ_ID_T v_object; /* Object id things */ unsigned char v_network_address[4]; UINT_64_T v_counter64; /* 64 bit unsigned things */ } value_u; union { ptr_t p_ptr; /* storage space for moving info */ bits32_t p_int; /* between test and set routines */ } priv; VBPRIVPROC_T * vb_free_priv; /* pointer to routine to free the */ /* private storage space */ UNDOPROC_T * undoproc; /* pointer to routine to undo sets*/ #if (INSTALL_ENVOY_EPI_SUBAGENT_SCHEME) UINT_32_T sa_flags; OCTET_T vb_access; #endif #if (INSTALL_ENVOY_AGENTX) OBJ_ID_T ax_search_end; OCTET_T ax_flags; #endif } VB_T; fgleeson@gi.com (FrankG) wrote in message news:... > I am implementing an agent that uses a c++ place to associate a single > MIB node with 1 or more data bases objects. I do not understand how > OIDs of singly and doubly indexed tables get passed in to as the > parameters passed to the 4 methods. > > Assume the following MIB node > > STATIC > MIBARC_T blahEntry_arcs[] = { > ARC(1, blahIndex1_node), > ARC(2, blahIndex2_node), > ARC(3, blahRowData1_leaf), > ARC(4, blahRowData1_leaf), > ARC(5, blahRowData2_leaf), > END_OF_ARC_LIST }; > STATIC > MIBNODE_T blahEntry_node = NODE("blahEntry", > blahEntry_arcs); > > i.e. > If the node is x.x...x.1 > Then index1 is x.x...x.1.1.0.0 > Then index2 is x.x...x.1.2.0.0 > Then rowData1 is x.x...x.1.3.0.0 > Then rowData2 is x.x...x.1.4.0.0 > Then rowData3 is x.x...x.1.5.0.0 > > In this case there would be 3 databases > > get(OIDC_T mastMatch, int unsused, OIDC_T* pUnused, ... > > I would rather deal with the complete OID as sent in the manager > request (I know how the manager works) then try to understand how it > has been mangled to make the 3 parameters above. > > In the Envoy docs it says that the SNMP_PKT_T* pckt pointer is > available so that the method can see the original packet. But when I > look at the structure for SNMP_PKT_T it is not at all obvious how to > get this information. --------------------------- Newsgroups: comp.os.vxworks Subject: Need idea for implementing watchdog Date: 21 Aug 2002 14:31:18 -0700 From: nedren@yahoo.com (Ned Ren) Organization: http://groups.google.com/ Message-ID: <98c05101.0208211331.4b320cea@posting.google.com> Hi vxworks gurus, I'm wondering if you could give me some idea for implementing a software watchdog. We have a system running a fixed number of tasks, there are bugs in the code which would cause the system running into touble over a peroid of 1 week. We want to implement a watchdog and if it sees the system is in trouble it'll reset it (of course, the best solution is find the bug, but... :) The definition of trouble is: 1. a task is suspended (due to invalid pointer access, or some other bad things) 2. a task is running wild, in a tight loop or something 3. a task is no long existing for 1 and 3 I sort of know how to do, but any idea for 2? Is there any lib call in vxworks that gives task running time? (we are using vxworks 5.4) Thanks!! --------------------------- Newsgroups: comp.os.vxworks Subject: new compiler win2k host/x86 target? Date: 15 Aug 2002 18:55:52 GMT From: Denis Perelyubskiy Message-ID: hello, we are using the following compiler version: cygnus-2.7.2-960126 egcs-971225 tornado 2.0 (includes WRS patch 27834) ... which is giving us lots of problems... there was a rumor that sometime in august WRS was going to release a new win2k-host/x86-target compiler, based on 2.96.x or something. has anyone heard anything like that? is it out? if not, when is it supposed to be out? denis - -- if you want to reply by mail, please tweak my address in a reasonable way. sorry for the inconvenience. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: hardware compability Date: Thu, 15 Aug 2002 22:52:53 GMT From: Jerry Hudgins Message-ID: <3D5C30C0.9070106@e-farm.com> References: <7%U69.81$95.2335@news.ou.edu> Reply-To: Jerry Hudgins Li Tan wrote: > Hi, I am new to Vxworks, and I am planning to use it in my robot project, > but I caould not find the information on the hardware supported by vxworks > in its web site. Could you give me a hand or hint? Basic support information can be found in the data sheets: For more specific information, talk to your WRS sales rep or your hardware vendor. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sysMemTop() doesn't quite work... Date: 21 Aug 2002 15:56:01 -0700 From: jchen@terayon.com (john) Organization: http://groups.google.com/ Message-ID: References: Got it. Thanks a alot Wade! Wade Oram wrote in message news:... > In message , john > writes > >Wade Oram wrote in message > >news:... > >> In message , john > >> writes > >> >Hello developers! > >> > > >> >I wanted to reserve some memory for personal use in vxWorks on a > >> >PPC8260. The current sysMemTop routine is as follows: > >> > > >> >char *sysMemTop( void ) > >> >{ > >> > static char * memTop = NULL; > >> > > >> > if ( memTop == NULL ) > >> > { > >> > memTop = sysPhysMemTop() - USER_RESERVED_MEM ; > >> > } > return memTop ; > >> >} > >> > > >> >However, with this routine, when sysPhysMemTop returns 0x4000000 > >> >(64MB) and I tried to define USER_RESERVED_MEM to 0x20000 (2MB) it > >> >doesn't seem to work. My total memory is still 64MB and sysMemTop > >> >still returns 0x4000000. I thought if you define USER_RESERVED_MEM > >> >then it substracts it, like the code says. but it doesn't. I took > >> >out the if loop and everything looks ok. > >> > > >> >Why is the if loop there? When I removed the if statement then > >> >everything seems to be ok. > >> > >> Its a performance optimisation: memTop is a static variable so its value > >> is calculated the first time around and then on subsequent calls to > >> sysMemTop, the pre-calculated value of memTop is returned instead of > >> repeating the calculation. This is done because sysPhysMemTop() is often > >> a non trivial function which will take a relatively long time to > >> execute. > >> > > > >I see. Our BSP doesn't allow us to do dynamic memory sizing anyways, > >and so we #define a lot of our stuff. I'll proabbly comment this out > >in our code. > > > >> > > >> >Does anyone know why it doesn't *seem* to work the say it is? my > >> >USER_RESERVED_MEM is defined in config.h and its included in the > >> >beginning of sysLib.c > >> > >> You don't say but my guess is that you are using the Tornado II project > >> facility to build the image. If this is the case, it is likely that the > >> project facility provides is own definition of USER_RESERVED_MEM which > >> overrides the one in config.h. I think you will find this parameter in > >> the project manager under hardware->memory although I could be wrong > >> (I'm not at work so I can't confirm this!) > > > >char* memTop is a local variable, though. Right? I mean, its defined > >only within the scope of the function. Yet, it seems to behave as a > >global variable. Why is that? > > > > memTop is a 'static' local variable (static char* memTop = NULL;) > meaning that it retains its value over successive calls to sysMemTop. In > fact, storage for memTop will be allocated from the BSS segment and not > the stack. In this respect, it is like a global variable. The difference > is that the compiler and linker between them conspire to make it > inaccessible from anywhere other than the body of sysMemTop. > > Thus the first time you call sysMemTop, memTop is initialised to NULL. > The if(memTop == NULL) condition then evaluates to TRUE and memTop is > then set to sysPhysMemTop() - USER_RESERVED_MEM. > > On subsequent calls to the function, memTop is not re-initialised and is > not NULL and so the if(...) condition evaluates to FALSE and > sysPhyMemTop is not called. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Design ideeas on shared memories Date: 21 Aug 2002 16:53:07 -0700 From: emil@lapel.com (Emil Rojas) Organization: http://groups.google.com/ Message-ID: <754b62e0.0208211553.dcae473@posting.google.com> References: <25c1993e.0208160613.81842c9@posting.google.com> <754b62e0.0208191425.17e12a7f@posting.google.com> <25c1993e.0208200549.2703e980@posting.google.com> avsrk@mailcity.com (Subrahmanyam Arya) wrote in message news:<25c1993e.0208200549.2703e980@posting.google.com>... > Just to make it more clear , > > I am going to have multiple processes accessing the shared memory , > each of them could be single / multi threaded but i am going to access the > shared memory from only a particular thread of each process , i.e a thread > from a particular process can acquire lock and read/write data and release the > lock . > > I am assumming the suggestions you made are for a scenario like this .... > > > -subra > Exactly, with the caveat that VxWorks does not have processes. So you mean processors then things will be more complex but with the shared memory objects you should be able to do the same thing. So to summerize and again, unless you have some high performance need you don't need to use semaphores et al. Just have a single thread manage the pointers. Allow other threads to look into data structures returned by your control thread via a message, likes request to put new data in the shared memmory are messages to the shared memory control thread. emil --------------------------- Newsgroups: comp.os.vxworks Subject: Design ideeas on shared memories Date: 16 Aug 2002 07:13:40 -0700 From: avsrk@mailcity.com (Subrahmanyam Arya) Organization: http://groups.google.com/ Message-ID: <25c1993e.0208160613.81842c9@posting.google.com> Hi Vxworkers , I am trying to implement a real-time application on VxWorks which uses shared memories . I am going to have one task which is going to do data acquisition and control and needs higher priority . There would be another 3 to 4 tasks like a datalogger , gui for showing the data acquired over the web and on a serial line etc . So the real -time data acquisition and control job writes into the shared memory and the other processes just read from it . I am familiar with the shared memories and semaphores . Is the simple mechanism of acquiring a lock doing a read/write on to the sharedmemory an d releasing the lock with the real time control job running at a higher priority be sufficient , or do we need some thing special like read/write locks etc . Does any one have any specific experience with this type of issues . Apprecaiate your thoughts and ideas in this - -Subra --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Configureing PCI bus interface on IBM PPC405GP Date: Thu, 22 Aug 2002 01:45:35 GMT From: inoue Organization: @Home Network Message-ID: <3D6442D4.BE5AA403@yahoo.co.jp> References: <3D604756.F7E4B15B@yahoo.co.jp> <3D60C947.EFB7EF3D@hotmail.com> Fortunately, I could configure PCI bus interface according to your advice. Thank you & Kind regards, Inoue Varndell Engineering wrote: > > inoue wrote: > > > > Hi, > > > > I'm configureing PCI bus interface on IBM PPC405GP walnut BSP to add > > two DP83815 (National Semiconductor) LAN devices. > > > > I configured PCI bus related definitions in config.h, made vxWorks > > image, and ran it. I failed to load DP83815 LAN driver. > > > > Then, I did pciDeviceShow and pciHeaderShow, following are the result. > > > > Please help me how config.h should be configured. > > > > I want to assign 1st DP83815 device to memory space 0x80200000, and > > 2nd to 0x80300000. Now I see only ffffffff when I dump that space. > > [snip] > > If you want to assign PCI address spaces, then you should disable > PCI auto-configuration in your BSP and call pciDevConfig(...) as > appropriate for each device. > In the PCI dumps you posted, neither device had been configured, > or both had been badly misconfigured. > > Regards, > George Varndell > -- > Varndell Engineering, LLC > Embedded Systems Expertise > http://www.varndellengineering.com > > -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- > http://www.newsfeed.com The #1 Newsgroup Service in the World! > -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Wind Web Server - httpBlockPut Date: Thu, 22 Aug 2002 02:50:17 GMT From: "drdiags" Organization: AT&T Broadband Message-ID: References: Hello Laura, Here is something I found on WindSurf: TITLE: Wind Web Server 2.0 sucks up buffers at a high rate causing crashes. SPR #: 34756 STATUS: Accepted IDE: RTOS: PRODUCT NAME: Wind Web 2.0 RELEASE STATUS: FCS PRODUCTS AFFECTED: Wind Web HOST: All HOST OS: All All ARCH FAMILY: All ARCH CPU: All BSP: DESCRIPTION: Wind Web Server 2.0 sucks up buffers at a high rate causing crashes. Wind Web Server can't handle a light load very well with crashes occuring very easily. Using RPM's to service the http requests will cause problems on simple test cases. See the test case for exact info. I don't have this product installed, but is it possible to run the Tornado Browser and look at memory consumption while your application is running. This will let you know if the above problem is the issue or whether you are leaking memory someplace. The buffers above are probably the mbufs for networking and won't be visible using the Browser. I will look around some more for other possible answers, but your depletion of mbufs and the TIME_WAIT sockets could mean you may have to increase your network buffers or NUM_FILES (WAG on my part). "Laura Chamberlain" wrote in message news:efdaef76.0208202335.75f3c35a@posting.google.com... > "drdiags" wrote in message news:... > > > Just more questions. When the system stops responding, does it ever recover? > No, the whole thing locks up (the main application stops too). If the > watchdog is connected the system reboots. > > > When it stops responding, are any tasks suspended? > See above - can't get debug tools in > > > Have you taken a look on WindSurf for any possible SPR fixes for your version of Wind Web Server? BTW, what version is this? > Yes, was running WWS 2.0 Beta but have since upgraded to the released > version and installed patch 1 from WindSurf, can't see any other > similar problems > > > Maybe someone who has used this product may have their memory jogged if you provide this. > > Could you print out the errno upon the ERROR return, this may help. > HttpBlockPut only returns HTTP_OK or HTTP_ERROR, when the system dies > or does a short download can't see any error message on the terminal - > is there something else I should be looking at? > > > What version of Tornado, what architecture, BSP and network interface type? > Tornado 2.0.2, VxWorks Ver 5.4, ADS860 BSP ver 1.2/5, MotFec Network > using source 01i,28jun00 (waiting for later ver from windriver). The > data to send is compiled from DiskOnChip (TFFS and DOSFS2) and FLASH. > > > > How many sockets are open when the error occurs? > Don't know once crashed cant get debug tools in but I will find out > for short downloads. During normal operation you can get several > connections in the TIME_WAIT state, occasionally you get too many and > the system pauses while they free up, do you think this is releated to > the crashing? > > > Here is an extract from my configuration:- > > #define HTTP_MAX_SOCKQ 5 > #define HTTP_MAX_SOCK_BUFF 4096 > #define HTTP_SOCK_TIMEOUT 60 > #define HTTP_SOCK_ARRAY 4 > #define HTTP_MAX_GLOBAL_MEM_PART_SIZE 15360 > #define HTTP_MAX_REQ_PART_SIZE 15360 > #define HTTP_MAX_REQLINE_SIZE 1024 > #define HTTP_MAX_ARGUMENTS 10 > #define HTTP_MIME_MAXLINES 10 > #define HTTP_NR_OF_TASKS 4 > #define HTTP_TASK_STACK_SIZE 6114 > > > > Sorry I don't have an answer, but if you share the above information, > > someone will join in. Also, did you contact Wind River support about > > this? > Yes but there response has been slow and so far no useful suggestions, > they are now asking me to provide a test case but this is going to > take me some time to prepare. > > > Thanks > > Laura > > > HTH....... > > "Laura Chamberlain" wrote in message > > news:efdaef76.0208190544.5ebd4eac@posting.google.com... > > > Hi, > > > > > > Thanks for your response, I have built a webpage which shows the web > > > server config, memory usage and the network stats you suggest along > > > with tcpstatShow as it uses TCP, but I can't see anything wrong. Any > > > other ideas? > > > > > > "drdiags" wrote in message > > news:... > > > > Hello Laura, > > > > > > > > Not sure I can help, but is it possible that you are running out of > > > > network > > > > resources, i.e. #file descriptors are maxed out. Are the transfers done > > > > using > > > > UDP or TCP (I don't have the product installed)? I would run the various > > > > network debugging tools from the target shell: > > > > > > > > inetstatShow > > > > mbufShow > > > > netStackSysPoolShow > > > > netStackDataPoolShow > > > > endPoolShow "interface name" <<-- Look on vxWorks FAQ for > > utilities.c > > > > for this > > > > routine. > > > > iosFdShow > > > > i > > > > others that I cannot think of > > > > > > > > HTH. > > > > > > > > "Laura Chamberlain" wrote in message > > > > news:efdaef76.0208160356.3b1636dc@posting.google.com... > > > > > I am using a custom RPM to compile data from several places and send > > > > > to the user as a single file, up to 40Mb in total, this is sent using > > > > > httpBlockPut in blocks of 1024 - similar to the httpRpmDownload > > > > > function in Tornado\target\unsupported\http\rpm\httpRpmUpload.c. > > > > > > > > > > The first problem is sometimes the httpBlockPut function fails > > > > > (returns ERROR), resulting in an incomplete download - using internet > > > > > explorer you don't receive an error just a small file. > > > > > > > > > > The second problem I have found while testing the first - after an > > > > > hour or so of repeatedly downloading data the target stops responding > > > > > and there is no debug output to help me determine why. > > > > > > > > > > Can anyone help? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Anyone using Tornado 2.1.1 gcc with optimization.... Date: 16 Aug 2002 15:59:10 -0700 From: dsa0000@hotmail.com (dsa) Organization: http://groups.google.com/ Message-ID: References: <488e459a.0208150056.21f16b15@posting.google.com> Hi John, Thank you for the response and sorry for not providing more details. Yep, as you guessed I am new to this. Here are the details : 1) Which CPU? MIPS 5000 2) What level of optimisation? -O2 3) Exactly which compiler? gcc version gcc-2.96 19990621 VxWorks 5.4 4) What errors you are seeing We are trying to upgrade from Tornado 1.0.1 to Tornado 2.1.1. With optimization turned on (-O2) we are seeing weird issues including crashes. I am attaching a testcase below (just one example). 5) What makes you think that it is the compiler that is at fault If we turn off optimization completly (without any -O flags) things looks OK. Turning off optimization is not a viable solution for us. Here is the test case - --------------------- #include "stdio.h" void xor_fn(unsigned long ipp) { int i; int value= 0; unsigned long ip; for (i=0; i < 1; i++) { ip = ipp; value |= (((unsigned short *)&ip)[0] ^ ((unsigned short *)&ip)[1]); } printf ("Value = %x\n", value); return; } Input values (for example) : 0xffff0000 Result expected : 0xffff ^ 0x0000 = 0xffff Result obtained : 0x0 (it may vary) The ccmips compile options used is as follows ccmips -Bhost/sun4-solaris2/lib/gcc-lib/ -EB -mips4 -O2 - -mno-branch-likely -G 0 -ansi -fno-builtin -funsigned-char - -funroll-loops -fno-strict-aliasing -Wall -Itarget/h -c testcode.c -o testcode.o Initially we thought it is because of the strict-aliasing. But we turned that off too. I have brought up this issue with Wind River already. I wish to know 1) Whether anybody has tried T2.1.1 gcc optimization and faced any issues. 2) Any work arounds other than compiling without optimization. (Tornodo libraries are compiled with optimization). 3) Whether it make sense to compile with other compilers ? 4) Tornado 2.1.1 Tool chain reliable ?? It was released only this year January end. Thanks in advance. Regards DSA. john_94501@yahoo.com (John) wrote in message news:<488e459a.0208150056.21f16b15@posting.google.com>... > Hello, > > Your posting is very unlikely to get you any useful replies... You > need to tell us much more than you have. > > For example: > > 1) Which CPU? > > 2) What level of optimisation? > > 3) Exactly which compiler? [at least cc --version output] > > 4) What errors you are seeing... > > 5) What makes you think that it is the compiler that is at fault > > There are many experienced people reading this group, and contributing > answers regularly. On the whole we are also very tolerant of new users > and very basic questions, but one thing I think I can say with some > certainty is that none of us are psychic. We cannot guess what > tools/CPU/BSP/etc you are using, nor what you are doing with it, what > is failing and how it fails. > > Please post your question again with more detail, and perhaps then > somebody here will be able to get you some help. > > Rgds, > > John... > > dsa0000@hotmail.com (dsa) wrote in message news:... > > Hi All, > > > > Anyone using Tornado 2.1.1 tool chain ? > > We are facing some issues with optimization.. > > > > > > Thanks > > DSA --------------------------- Newsgroups: comp.os.vxworks Subject: Re: debug interrupt Date: 21 Aug 2002 20:19:46 -0700 From: hsingyuan@yahoo.com (Hsing Yuan) Organization: http://groups.google.com/ Message-ID: <70d78653.0208211919.2ef7cf38@posting.google.com> References: <7efa47da.0208192055.5558344a@posting.google.com> <2UYH+bGB18Y9EAb0@wtoram.co.uk> Tornado 2.0.2 does have a bug in system mode. I reported the bug, and WRS reproduced the problem. However, they found that the problem is not reproducable in the new tornado to be released (this month(8/2002) ?). Using a semaphore in a interrupt handler may not be a good idea since it may cause workQ overflow. We use ICE to debug interrupt handlers. Hsing Wade Oram wrote in message news:<2UYH+bGB18Y9EAb0@wtoram.co.uk>... > In message , Faure > writes > > > >Note: I have some difficulties to debug in system mode, there are some > >bugs with that mode under Tornado 2.0.2: > >SPR#72141, T202: system mode debugging fails > >after some time / target hangs) > > > > I have successfully used system mode in T2 CP2 (Windows NT4 host, X86 > target) with additional patches although I have not had to use it for > extended periods. How long is 'after some time'? > > >if it is sufficient,I advice you to debug in task mode (not in system > >mode), > >it is what I choosed to do after spending too much time to try to make > >work without success the system mode. > > The original poster wanted to set a break point in an interrupt handler. > For this purpose, only system mode can be used. However, the advice > still stands - find out as much information as you can about the fault > in task mode and restrict system mode debugging to the occasions that > really demand it. > > If an interrupt handler really is causing you grief and system mode > debugging is not reliable enough to solve the problem, you *may* be able > to change the interrupt handling so that the interrupt handler does > nothing except give a semaphore to a (very) high priority task which > does the interrupt handling. This will then allow you to put a break > point in the original interrupt handler whilst still in task mode. Once > the fault is found the original interrupt handler - hopefully now fixed > - can be restored to its rightful place. > > As a technique, it won't be applicable all problems with interrupt > handlers but it can be used to help solve many - especially when the > interrupt in question is independent of any other interrupts in the > system. --------------------------- Newsgroups: comp.os.vxworks Subject: Q: Does Tornado 2.1 support BSP of MV2700 or MV2604 Date: Fri, 16 Aug 2002 09:50:10 -0400 From: yu zhou Organization: Nortel Message-ID: <3D5D0312.90507@nortelnetworks.com> Dear friends, Does Tornado 2.1 support BSP of MV2700 or MV2604? Thanks Yu --------------------------- Newsgroups: comp.os.vxworks Subject: Re: what's wrong with my cache flush function? Date: 16 Aug 2002 19:23:36 -0700 From: du.jianjun@zte.com.cn (du) Organization: http://groups.google.com/ Message-ID: <21a3d7e1.0208161823.28059992@posting.google.com> References: <21a3d7e1.0208141728.1e703881@posting.google.com> <488e459a.0208150044.2b37a733@posting.google.com> john_94501@yahoo.com (John) wrote in message news:<488e459a.0208150044.2b37a733@posting.google.com>... > Hello, > > Is there a good reason why you didn't just use the VxWorks cacheLib > routines? Check out the reference manual entry for cacheLib and in > particular cacheFlush() and probably cacheInvalidate() since I suspect > you'll need to use that at some point too. > > HTH, > > John... > > PS When writing a routine totally in assembler, you'd also be better > off just writing it in assembler (as a .s file), and passing that to > the compiler. Look in target/h/make for the flags to pass if you want > to use the C pre-processor with your assembler, or from your BSP > directory just try 'make myAssembler.o' having added myAssembler.s > there with something in it. > Thanks John.Because the vxworks cachelib rountine can't work,maybe the cachelib of my vxworks installed can't support mpc7410,so I should write my cachelib routines. I have write the .s file,but the 82559 also don't works well. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Need idea for implementing watchdog Date: 21 Aug 2002 22:54:34 -0700 From: tyx@huawei.com (Tao Yongxiang) Organization: http://groups.google.com/ Message-ID: <8f8b64c7.0208212154.6ded50dc@posting.google.com> References: <98c05101.0208211331.4b320cea@posting.google.com> nedren@yahoo.com (Ned Ren) wrote in message news:<98c05101.0208211331.4b320cea@posting.google.com>... > Hi vxworks gurus, > > I'm wondering if you could give me some idea for implementing a > software watchdog. We have a system running a fixed number of tasks, > there are bugs in the code which would cause the system running into > touble over a peroid of 1 week. We want to implement a watchdog and if > it sees the system is in trouble it'll reset it (of course, the best > solution is find the bug, but... :) The definition of trouble is: > > 1. a task is suspended (due to invalid pointer access, or some other > bad things) > 2. a task is running wild, in a tight loop or something > 3. a task is no long existing > > for 1 and 3 I sort of know how to do, but any idea for 2? Is there any > lib call in vxworks that gives task running time? (we are using > vxworks 5.4) > > Thanks!! you can install two hook functions: one installed by taskSwitchHookAdd(), called when tasks switch another installed by sysClkConnect(), called when ticks add and define a gobal counter in the first function, clear the counter; in the second function, increase the counter, if the counter is greater than the longest time slice which one task should take, it show the condition 2 in the above. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: context switch hooks Date: Thu, 22 Aug 2002 05:56:50 GMT From: "Krishna Balusu" Message-ID: References: Could you try adding the fn to taskDeleteHookAdd. "Pramod Srinivasan" wrote in message news:ajvlon$6d6$1@overload.lbl.gov... > Hi, > > I am having a weird problem in vxWorks, I have a > context switch hook which checks to see if the task > that is being switched out is being suspended, if so I > record this for debugging. But here is the problem. > > looking through the vxWorks source code, calling exit > in a task in vxWorks, actually suspends the current > task, and expects the exception task to clean up the > task. So when a task exits, from what I understand the > task switch hook should have been called when the next > task possibly exception task is switched in. To my > surprise I see that the task switch hook is never > called when the task exits, except in one situation > where its being called. I dont understand this > behavior. > > Does anyone else have a clue to this? > > Thanks, > --Pramod > > __________________________________________________ > Do You Yahoo!? > HotJobs - Search Thousands of New Jobs > http://www.hotjobs.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Need idea for implementing watchdog Date: Thu, 22 Aug 2002 06:32:11 +0000 (UTC) From: Johan Borkhuis Organization: Agere Systems Message-ID: References: <98c05101.0208211331.4b320cea@posting.google.com> nedren@yahoo.com (Ned Ren) wrote: > Hi vxworks gurus, > > I'm wondering if you could give me some idea for implementing a > software watchdog. We have a system running a fixed number of tasks, > there are bugs in the code which would cause the system running into > touble over a peroid of 1 week. We want to implement a watchdog and if > it sees the system is in trouble it'll reset it (of course, the best > solution is find the bug, but... :) The definition of trouble is: > > 1. a task is suspended (due to invalid pointer access, or some other > bad things) > 2. a task is running wild, in a tight loop or something > 3. a task is no long existing > > for 1 and 3 I sort of know how to do, but any idea for 2? Is there any > lib call in vxworks that gives task running time? (we are using > vxworks 5.4) You need a watchdog task that monitors the status of the other tasks. Also all monitored tasks need to signal to the watchdog task that they are still alive, for example every 100 msec. If a task misses 2 triggers it is considered "wild", and you can reset the system. Another way to detect if a task is running wild is to have one task at a low priotity siginalling the watchdog task. If this iask does not get any time other tasks must be eating all the CPU time, and are probably "wild". Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ll() on /tgtsvr broken Date: Sat, 17 Aug 2002 10:59:23 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Well, since you got source code for usrLib.c (or usrFsLib.c) that is the implementation of both ls() and ll(), you could just debug this yourself. Anyway, "ls" before actually trying to access a device, checks for a special hook, which was created for netDrv, which lists file in a special way by th enetDrv itself. If the hook is discovered to work, "ls" will not try to opendir() etc. "Pete Flugstad" wrote in message news:dje79.58892$me6.7555@sccrnsc01... > Looks like running ls on a /tgtsvr based system works OK, but > ll() (long listing) does not? It goes into an endless loop > printing out the "." directory entry. > > This is bizarre, as they both use opendir and then readdir - ll > just tries to do a stat or some such on each filename it finds > to fill out the extra info. What's ls() doing that makes it > work properly on "/tgtsvr" based directories? > > Anyone know what's going on? > > Thanks, > Pete Flugstad > Icon Labs > > PS: as a test, the code below works fine on something like "/fd0", > but as with ll, loops endlessly on "/tgtsvr" (modulo the count > check which I added to avoid locking up) even through it's not > doing any kind of stat. What does ls do that let's it work > properly on "/tgtsvr" based dirs? > > int dor( char *dirname ) > { > struct dirent *dirPtr; > int count = 0; > > DIR *aDir = opendir( dirname ); > if ( !aDir ) > { > printf( "do_readdir - unable to open '%s' directory, error: '%s'\n", > dirname, strerror(errno) ); > return -1; > } > > while ( count < 50 ) > { > dirPtr = readdir( aDir ); > if ( !dirPtr ) > break; > > printf( "dir entry: '%s'\n", dirPtr->d_name ); > count++; > } > > closedir( aDir ); > > printf( "after dir loop, count: %d\n", count ); > return 0; > } > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: DHCP client Date: Thu, 22 Aug 2002 09:47:10 +0200 From: WiMos Organization: disorganized Message-ID: <3D6496FE.1D176181@I-dont-like-spam.nl> References: <3d637c51$1@news.broadpark.no> "Øyvind Holmeide" wrote: > I have tried to include DHCP client functionality (DHCPv4 runtime client), > but then SNMP deamon support must be added as well, which I do not have. Is > SNMP required for DHCP to work? If yes, can someone propose a DHCP freeware > alternative? > > BR > > Øyvind Holmeide Some information from WindRiver which we received after asking the same question: - ------ begin quote You have encountered a known bug, which was fixed by SPR 25580 It should be fixed by installing the T2CP2, T2CP3 or T2CP4 cumulative patches, and in T2.0.2. The problem has a sting attached: when you include SNMP and DHCP without the fix, there's a duplicate symbol, so you still can't build. You should ideally have at least one of the cumulative patches installed anyway. T2CP4, which upgrades to Tornado 2.0.2 is downloadable from Windsurf. The others used to be, but I'm not sure if they still are available to customers direct. Try these links... http://web1.windriver.com/cgi-bin/windsurf/downloads/view_binary.cgi?binaryid=5 http://web1.windriver.com/cgi-bin/windsurf/downloads/view_binary.cgi?binaryid=97 http://web1.windriver.com/cgi-bin/windsurf/downloads/view_binary.cgi?binaryid=406 - ----- end quote What I find strange is that the patches are older that the CD's we recently bought. How people are supposed to know that patches are available (even directly after purchasing it) I don't understand. Wim --------------------------- Newsgroups: comp.os.vxworks Subject: TCP connection released, but socket not closed Date: Thu, 22 Aug 2002 09:48:38 +0200 From: "Jihad Haddad" Organization: TDC Internet Message-ID: <3d649756$0$53634$edfadb0f@dspool01.news.tele.dk> Hi I am a total newbe to this, and my use of terms is still that of a 6 year old. So please be patient with me. We have a "box" that contains a Strong Arm processor, an ethernet plug and that runs VxWorks. I can easily create a TCP connection from a windows 2000 PC to that box. The problem is that when I close the windows socket that has the TCP connection, the TCP connection is closed in both PC and "box". But the socket in the box (that is the VxWorks socket) is not closed. This I can see if I have a telnet connection to the box, and I run the command "iosFdShow" which shows the list of descriptors opened. I can here see that the socket that was opened with the TCP connection, is not closed. I can close the socket if I call "close fd" where fd is the number of the descriptor the socket has. But a write to the socket gives an error. If I call "errno" afterI make the "write", i get the following error code: errno = 0x179b4c: value = 1835009 = 0x1c0001 I do not know if the above error code has anything to do with the error of the "write". I have also sniffed the packet sent by the PC to close the TCP connection. Some information about the packet is the following: - ----------------------------------------------------------------------- Protocol: cadlock Summary: TCP (RST) Size: 60 The packet looks as follows: 0000: 00 02 15 00 00 3F 00 04 76 1A C7 2A 08 00 45 00 0010: 00 28 2E 34 40 00 80 06 00 00 0A 07 0A 25 0A 07 0020: 65 5D 02 C1 03 E8 85 55 55 6A 7C 69 3F B9 50 04 0030: 00 00 8E C5 00 00 - ----------------------------------------------------------------------- Does anyone have any idea why this is happening. Is it the way I initialize the TCP protocol? I do not know if this is enough information, but I can give more if needed. Thank you very much in advance Jihad Haddad --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Need idea for implementing watchdog Date: 22 Aug 2002 01:14:45 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0208220014.33fb38ef@posting.google.com> References: <98c05101.0208211331.4b320cea@posting.google.com> nedren@yahoo.com (Ned Ren) wrote in message news:<98c05101.0208211331.4b320cea@posting.google.com>... > 2. a task is running wild, in a tight loop or something well... two things: a.) typically you will know program flow for a given task, therefore it has to pass certain code parts (in a certain sequence, during a certain period of time with lower/upper boundary). you could have the system check for this. b.) a task non-blocking in a tight loop will starve all tasks with lower priority. typically, a realtime system will always report idle time (or has a dedicated task consuming the excess cpu cycles) you could monitor for one of those conditions, or for both... --------------------------- Newsgroups: comp.os.vxworks Subject: Re: i have trouble connecting to target server with tornado Date: 22 Aug 2002 01:19:32 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0208220019.5c735f25@posting.google.com> References: David Cooper wrote in message news:... > > can anyone tell me what is the source of this error.... this is a general error message... in those cases it's always a good idea to: a.) configure the image for Network connection (INCLUDE_WDB_COMM_NETWORK, check for included WDB) b.) check, can the host ping the target (check for included ICMP) c.) check target server configuration (IP address, image, wdprpc selected) if all of the above fails d.) take an ethernet traffic analyzer and see whats going on (WDB traffic is the UDP traffic for port 0x4321) --------------------------- Newsgroups: comp.os.vxworks Subject: Re: TCP connection released, but socket not closed Date: Thu, 22 Aug 2002 11:13:45 +0200 From: "Jihad Haddad" Organization: TDC Internet Message-ID: <3d64ab49$0$88404$edfadb0f@dspool01.news.tele.dk> References: <3d649756$0$53634$edfadb0f@dspool01.news.tele.dk> I found out what the error is when I "write" to not closed socket: It says: S_errno_EPIPE if that should help a bit. Thanks again Jihad Haddad "Jihad Haddad" wrote in message news:3d649756$0$53634$edfadb0f@dspool01.news.tele.dk... > Hi > > I am a total newbe to this, and my use of terms is still that of a 6 > year old. So please be patient with me. > > We have a "box" that contains a Strong Arm processor, an ethernet plug > and that runs VxWorks. I can easily create a TCP connection from a > windows 2000 PC to that box. The problem is that when I close the > windows socket that has the TCP connection, the TCP connection is closed > in both PC and "box". But the socket in the box (that is the VxWorks > socket) is not closed. This I can see if I have a telnet connection to > the box, and I run the command "iosFdShow" which shows the list of > descriptors opened. I can here see that the socket that was opened with > the TCP connection, is not closed. > I can close the socket if I call "close fd" where fd is the number of > the descriptor the socket has. But a write to the socket gives an error. > If I call "errno" afterI make the "write", i get the following error > code: > errno = 0x179b4c: value = 1835009 = 0x1c0001 > I do not know if the above error code has anything to do with the error > of the "write". > > I have also sniffed the packet sent by the PC to close the TCP > connection. Some information about the packet is the following: > ---------------------------------------------------------------------- - - > Protocol: cadlock > Summary: TCP (RST) > Size: 60 > The packet looks as follows: > 0000: 00 02 15 00 00 3F 00 04 76 1A C7 2A 08 00 45 00 > 0010: 00 28 2E 34 40 00 80 06 00 00 0A 07 0A 25 0A 07 > 0020: 65 5D 02 C1 03 E8 85 55 55 6A 7C 69 3F B9 50 04 > 0030: 00 00 8E C5 00 00 > ---------------------------------------------------------------------- - - > > Does anyone have any idea why this is happening. Is it the way I > initialize the TCP protocol? > > I do not know if this is enough information, but I can give more if > needed. > > Thank you very much in advance > Jihad Haddad > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Who have intel 82559End driver source code? How about it efficiency? Date: 17 Aug 2002 11:28:40 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0208171028.6343944e@posting.google.com> References: <21a3d7e1.0208162333.f5d4099@posting.google.com> > How to > write the driver to make it work in flexible mode? you will need INTELs driver manual therefore, however this is difficult to obtain from them (it requires signing an NDA and stuff like this). unless you have the original documentation i'd not recommend starting on this. > and how about its > performance? this depends on your architecture. flexible has no direct performance advantage compared with simplified memory structure. however it would allow you to implement mblk scatter gather in hw. du.jianjun@zte.com.cn (du) wrote in message news:<21a3d7e1.0208162333.f5d4099@posting.google.com>... > I am using intel 82559er Ethernet controller,but my driver is changed > from fei82557End.c,it does't work well,and the performance is bad.As > you know the fei82557End.c make 82559 work in simplified mode,How to > write the driver to make it work in flexible mode?and how about its > performance? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Need idea for implementing watchdog Date: Thu, 22 Aug 2002 08:40:05 -0400 From: Keith Arner Organization: Marconi Message-ID: References: <98c05101.0208211331.4b320cea@posting.google.com> Reply-To: Keith Arner On Thu, 22 Aug 2002, Johan Borkhuis wrote: > nedren@yahoo.com (Ned Ren) wrote: > > > solution is find the bug, but... :) The definition of trouble is: > > > > 1. a task is suspended (due to invalid pointer access, or some other > > bad things) > > 2. a task is running wild, in a tight loop or something > > 3. a task is no long existing > > You need a watchdog task that monitors the status of the other tasks. > Also all monitored tasks need to signal to the watchdog task that they > are still alive, for example every 100 msec. If a task misses 2 triggers > it is considered "wild", and you can reset the system. > Another way to detect if a task is running wild is to have one task at a > low priotity siginalling the watchdog task. If this iask does not get any > time other tasks must be eating all the CPU time, and are probably > "wild". Another trouble spot which you might want to monitor for is deadlock between tasks. For this the watchdog approach described above would work well, but the low priority task would not help. Additionally, I would point out that you probably want to implement the watchdog as a VxWorks watchdog timer (wdCreate()), rather than using a high priority task. This will help you catch things like: taskLock(); while (1) { ... } Keith - -- mv ~karner /loony/bin --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Need idea for implementing watchdog Date: Thu, 22 Aug 2002 13:11:23 +0000 (UTC) From: Johan Borkhuis Organization: Agere Systems Message-ID: References: <98c05101.0208211331.4b320cea@posting.google.com> Keith Arner wrote: > On Thu, 22 Aug 2002, Johan Borkhuis wrote: > >> nedren@yahoo.com (Ned Ren) wrote: >> >> > solution is find the bug, but... :) The definition of trouble is: >> > >> > 1. a task is suspended (due to invalid pointer access, or some >> > other bad things) >> > 2. a task is running wild, in a tight loop or something >> > 3. a task is no long existing >> >> You need a watchdog task that monitors the status of the other tasks. >> Also all monitored tasks need to signal to the watchdog task that >> they are still alive, for example every 100 msec. If a task misses 2 >> triggers it is considered "wild", and you can reset the system. >> Another way to detect if a task is running wild is to have one task >> at a low priotity siginalling the watchdog task. If this iask does >> not get any time other tasks must be eating all the CPU time, and are >> probably "wild". > > Another trouble spot which you might want to monitor for is deadlock > between tasks. For this the watchdog approach described above would > work well, but the low priority task would not help. > > Additionally, I would point out that you probably want to implement > the watchdog as a VxWorks watchdog timer (wdCreate()), rather than > using a high priority task. This will help you catch things like: > > taskLock(); > while (1) { ... } The problem with doing things using wdCreate is that you are running in an interrupt, and that yoiu are limited in the number of functions you can use. The only thing you could use the wd function for is to perform the real reset: create a wd function that does the reset, and start it with wdCreate. When all tasks are still alive and kicking the watchdog task restart the wd-counter, otherwise this counter will expire and the system will be reset. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Vxworks Drivers for 8260 MCC and HDLC SCCs Date: 22 Aug 2002 15:02:59 GMT From: Mike Gilbert Organization: Concentric Internet Services Message-ID: <3D64FED9.8090700@baymicrosystems.com> References: If money's not an issue check out: http://www.aisysinc.com/ We used these guys at a previous job and it really pulled in time to market. Speaking only for myself, Mike Matchett, John wrote: > I am in need of MCC and SCC-HDLC Drivers for the Mot 8260. The reference board > my company bought only had a BSP that included UART and Ethernet. Does anyone > have any recommendations . > > Thanks > > John --------------------------- Newsgroups: comp.os.vxworks Subject: messages from interrupt level lost Date: Thu, 22 Aug 2002 16:24:43 +0100 From: Adam S Organization: Lucent Technologies Message-ID: <3D65023B.5EFC2586@yahoo.se> Early in the application startup I see messages like the following: 0x7bf3110 (tExcTask): 2 messages from interrupt level lost. This is something generated by a function running in the tExcTask context. My understanding of the exception handling was that exception handler pointers were posted to a message queue which is polled by the tExcTask. This would make tExcTask a message consumer rather than producer so I wonder where tExcTask might try to send messages? One obvious answer is the logLib task, but since this message appeared at all it doesn't strike me as a likely candidate. Adam --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Need idea for implementing watchdog Date: 22 Aug 2002 09:37:04 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0208220837.e4f7e68@posting.google.com> References: <98c05101.0208211331.4b320cea@posting.google.com> Hello, > 2. a task is running wild, in a tight loop or something > 3. a task is no long existing > > for 1 and 3 I sort of know how to do, but any idea for 2? Is there any > lib call in vxworks that gives task running time? (we are using > vxworks 5.4) Not really. You can use the spyLib facilities, but there is no programmatic interface to the results in VxWorks 5.4 (there is in VxWorks AE though :). Perhaps a simpler approach is to just sample the current task ID periodically, and keep your own statistics on them. You can connect your sampling routine to the auxilliary clock's interrupt, assuming that your board has one, and set that to whatever rate you feel is appropriate. To get the ID of the task that you interrupted, simply read taskIdCurrent (the kernel's global variable for the current running task). HTH, John... --------------------------- Newsgroups: comp.os.vxworks Subject: Re: significance of value in target Date: Thu, 22 Aug 2002 17:28:51 +0100 From: Adam S Organization: Lucent Technologies Message-ID: <3D651143.47F4071@yahoo.se> References: If you have a function returning int, the value will be printed to the shell. If you have a function returning void, the memory location where the return value would normally be written to is uninitialised and then it seems to return a garbage value. Adam subbarayan wrote: > > dear mentors, > Kindly make me understand whats the value vxworks is showing in target > after exe cuting any function or command?I have always wondered whats > this value?Hows it calculated?What are the factors of dependency for > this value and what is the significance of this value with respect to > the application on target? > expecting a reply, > anyone with a reply can mail me to:vxnerd@rediffmail.com or jus reply > here in this board itself --------------------------- Newsgroups: comp.os.vxworks Subject: Re: TCP connection released, but socket not closed Date: Thu, 22 Aug 2002 17:46:39 +0100 From: Phil Ashby Organization: IDL Message-ID: <3D65156F.9090702@idl.bt.co.uk> References: <3d649756$0$53634$edfadb0f@dspool01.news.tele.dk> Jihad Haddad wrote: > Hi > > I am a total newbe to this, and my use of terms is still that of a 6 > year old. So please be patient with me. > > We have a "box" that contains a Strong Arm processor, an ethernet plug > and that runs VxWorks. I can easily create a TCP connection from a > windows 2000 PC to that box. The problem is that when I close the > windows socket that has the TCP connection, the TCP connection is closed > in both PC and "box". But the socket in the box (that is the VxWorks > socket) is not closed. This I can see if I have a telnet connection to > the box, and I run the command "iosFdShow" which shows the list of > descriptors opened. I can here see that the socket that was opened with > the TCP connection, is not closed. > I can close the socket if I call "close fd" where fd is the number of > the descriptor the socket has. But a write to the socket gives an error. The 'normal' way to detect that a TCP stream has been closed is to check the return value from a read() call, which is only 0 if the stream is closed (0 == EOF). If you are never reading from the stream in VxWorks then the EPIPE error you have noticed can also be used to indicate that the remote system has closed the stream. In either case, you MUST call close(fd) in your VxWorks application to clean up the TCP stream state correctly. HTH, - -- Phil "Phlash" Ashby (speaking only for himself) Work: http://www.internet-designers.net/ Home: http://www.ashbysoft.com/ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: significance of value in target Date: Thu, 22 Aug 2002 09:53:46 -0700 From: "Michael R. Kesti" Organization: MK Associates Message-ID: <3D65171A.313A36B4@gv.net> References: <3D651143.47F4071@yahoo.se> Reply-To: mkesti@gv.net Adam S wrote: >If you have a function returning int, the value >will be printed to the shell. If you have a function >returning void, the memory location where the return >value would normally be written to is uninitialised >and then it seems to return a garbage value. It will be the return value of the most recently called non-void function. For example, in a typical "show" function that returns void but calls printf() just before returning, the reported value is the printf() return value which is the number of characters it had output. - -- ======================================================================== Michael Kesti | "And like, one and one don't make | two, one and one make one." mkesti@gv.net | - The Who, Bargain --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Design ideeas on shared memories Date: 22 Aug 2002 10:16:31 -0700 From: avsrk@mailcity.com (Subrahmanyam Arya) Organization: http://groups.google.com/ Message-ID: <25c1993e.0208220916.264ac51b@posting.google.com> References: <25c1993e.0208160613.81842c9@posting.google.com> <754b62e0.0208191425.17e12a7f@posting.google.com> <25c1993e.0208200549.2703e980@posting.google.com> <754b62e0.0208211553.dcae473@posting.google.com> > > > > Exactly, with the caveat that VxWorks does not have processes. So you > mean processors then things will be more complex but with the shared > memory objects you should be able to do the same thing. So to > summerize and again, unless you have some high performance need you > don't need to use semaphores et al. Just have a single thread manage > the pointers. Allow other threads to look into data structures > returned by your control thread via a message, likes request to put > new data in the shared memmory are messages to the shared memory > control thread. > > emil The tasks that Vxworks has are like processes , and we can have the tasks totally independent i guess (that is not spawned for a common parent process ) , that is if the functionality is logically independent , i like to make them independent tasks , now i know all the memory is shared in VxWorks ( no special shared memory required may be ) but still i need to synchronize access to the shared memory . Using threads is like making one fat process which tries to do all the work ... any ideas / comments please - -subra --------------------------- Newsgroups: comp.os.vxworks Subject: Re: C++ Unresolved Symbols Date: Thu, 22 Aug 2002 18:38:55 +0100 From: Wade Oram Organization: None Message-ID: References: Reply-To: Wade Oram In message , Mark D Chavira writes >Hi, > >I have used C on VxWorks for a long time. I recently tried C++, and though >the program compiles fine, I get unresolved symbols when I load. The >symbols are from the string and stream standard libraries. Does anybody >know how to fix this problem. I am guessing that I just need to load a >library, but I have not been able to figure it out on my own. As far as I recall, the default C++ options that are included when you configure the OS from the project manager do no include string or iostream support. You should just be able to reconfigure the OS to include them without worrying about additional libraries. They are both sub-categories of C++ Support in the VxWorks tab of the project facility dialogue - at least on windows hosts. I have not used the project facility at all under unix hosts so I can't comment. - -- Wade Oram --------------------------- Newsgroups: comp.os.vxworks Subject: Unsupported FTP Server Date: Sun, 18 Aug 2002 03:25:54 GMT From: "weddick" Organization: AT&T Broadband Message-ID: <6tE79.75738$me6.10044@sccrnsc01> I notice in a past message a reference to an FTP server in the unsupported directory. I found the file put would like help on how to go about building this into my BSP to replace the default server. Thanks, Joel --------------------------- Newsgroups: comp.os.vxworks Subject: Can't boot MV2700 over ethernet Date: 22 Aug 2002 11:29:16 -0700 From: morganed@nswc.navy.mil (Eric Morgan) Organization: http://groups.google.com/ Message-ID: <190f55f8.0208221029.73da90c8@posting.google.com> I am trying to load vxworks into a MV2700 board over an ethernet network with a UNIX host. With PPCBug I can ping the host, but the host cannot ping the MV2700. The client can ping the host, but the host can't ping the client. Is the problem due to the ethernet address on the MV2700 using 33 bits, not the standard 32 bits? The ethernet address has 5 nibbles, not 4. Does this create an address problem with the ARP? I can replace the 2700 board with a 167 board and can ping from either direction. Of course if I try to do an NBO, (attempting to load vxworks over the ethernet) it gives me a code of $8100, which is file not found. Also, the mv2700 and the host are on the same network, so that is not the problem. Any ideas? Eric Morgan --------------------------- Newsgroups: comp.os.vxworks Subject: Re: "Virtual" end driver Date: 22 Aug 2002 14:03:10 -0700 From: happyhappy779@hotmail.com (Satish G) Organization: http://groups.google.com/ Message-ID: References: Hi, You have not given the IP addresses used for the interfaces, so I would suspect that the IPaddress/netmask combination is resulting in same network for both interfaces. Satish G Gunter.Ritter@tenovis.com (Gunter Ritter) wrote in message news:... > Hello VxWorkers, > > I have written a "virtual" end driver as a secondary end driver for > our system. > "Virtual" means that this end driver has no "physical" ethernet > hardware. > > The primary end driver is working with ethernet hardware. > > I need this secondary end driver only to exchange ip packets with the > TCP/IP stack of vxworks and let vxworks do the routing between the two > networks. > > I have a problem when I try to assign an ip address. > muxDevLoad(), muxBind(), ipAttach() and ifMaskSet() work fine. > ifShow() shows me the settings of the secondary driver. > But calling ifAddrSet() with an unique ip address causes the network > functionality of vxworks to crash. > > Even "ping" does no more work with the primary end driver after > calling ifAddrSet(). > > Does anybody can give me a hint whats going wrong with my secondary > end driver? > > Thanks. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: TCP connection released, but socket not closed Date: 22 Aug 2002 14:11:25 -0700 From: happyhappy779@hotmail.com (Satish G) Organization: http://groups.google.com/ Message-ID: References: <3d649756$0$53634$edfadb0f@dspool01.news.tele.dk> <3d64ab49$0$88404$edfadb0f@dspool01.news.tele.dk> The windows end is sending a TCP RESET to close the connection. The socket is is in a state where the TCP control block associated with the socket is removed by VxWorks, which is expected. Next time when you make a call to send something the OS returns error EPIPE. When this happens the aplication should close the socket upon which OS will remove the it. Hope this helps. Satish G "Jihad Haddad" wrote in message news:<3d64ab49$0$88404$edfadb0f@dspool01.news.tele.dk>... > I found out what the error is when I "write" to not closed socket: > It says: S_errno_EPIPE > if that should help a bit. > > Thanks again > > Jihad Haddad > > "Jihad Haddad" wrote in message > news:3d649756$0$53634$edfadb0f@dspool01.news.tele.dk... > > Hi > > > > I am a total newbe to this, and my use of terms is still that of a 6 > > year old. So please be patient with me. > > > > We have a "box" that contains a Strong Arm processor, an ethernet plug > > and that runs VxWorks. I can easily create a TCP connection from a > > windows 2000 PC to that box. The problem is that when I close the > > windows socket that has the TCP connection, the TCP connection is > closed > > in both PC and "box". But the socket in the box (that is the VxWorks > > socket) is not closed. This I can see if I have a telnet connection to > > the box, and I run the command "iosFdShow" which shows the list of > > descriptors opened. I can here see that the socket that was opened > with > > the TCP connection, is not closed. > > I can close the socket if I call "close fd" where fd is the number of > > the descriptor the socket has. But a write to the socket gives an > error. > > If I call "errno" afterI make the "write", i get the following error > > code: > > errno = 0x179b4c: value = 1835009 = 0x1c0001 > > I do not know if the above error code has anything to do with the > error > > of the "write". > > > > I have also sniffed the packet sent by the PC to close the TCP > > connection. Some information about the packet is the following: > > ---------------------------------------------------------------------- > - > > Protocol: cadlock > > Summary: TCP (RST) > > Size: 60 > > The packet looks as follows: > > 0000: 00 02 15 00 00 3F 00 04 76 1A C7 2A 08 00 45 00 > > 0010: 00 28 2E 34 40 00 80 06 00 00 0A 07 0A 25 0A 07 > > 0020: 65 5D 02 C1 03 E8 85 55 55 6A 7C 69 3F B9 50 04 > > 0030: 00 00 8E C5 00 00 > > ---------------------------------------------------------------------- > - > > > > Does anyone have any idea why this is happening. Is it the way I > > initialize the TCP protocol? > > > > I do not know if this is enough information, but I can give more if > > needed. > > > > Thank you very much in advance > > Jihad Haddad > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ipAttach ipDetach memory leak problem Date: 22 Aug 2002 14:23:14 -0700 From: happyhappy779@hotmail.com (Satish G) Organization: http://groups.google.com/ Message-ID: References: <5607ff2c.0208180549.7ad53054@posting.google.com> <4a7f0d21.0208181900.64784a5c@posting.google.com> <5607ff2c.0208200214.5468d0c9@posting.google.com> I think you should pass the interface name in form . The free call should look like: free(ifunit("fec1")->if_name); Satish G. gabi@tdsoft.com (Gabi) wrote in message news:<5607ff2c.0208200214.5468d0c9@posting.google.com>... > lisheng_811@yahoo.com.cn (Saledom) wrote in message news:<4a7f0d21.0208181900.64784a5c@posting.google.com>... > > gabi@tdsoft.com (Gabi) wrote in message news:<5607ff2c.0208180549.7ad53054@posting.google.com>... > > > Hi > > > When I operate the following sequence from my code, NOT from shell: > > > ipDetach (1,"fec"); > > > ipAttach(1,"fec"); > > > I notice in memShow of a memory leak (6-12 bytes) in each sequence. Do > > > I miss something? > > > tanks > > > Gabi > > > > try the following code, please. > > > > free(ifunit("fec")->if_name); > > ipDetach(1, "fec"); > > ipAttach(1,"fec"); > Hi, > I didn't quit follow you up, but if you are in attach mode and operate > your steps you still suffer from leak problem. > Gabi --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Need idea for implementing watchdog Date: 22 Aug 2002 15:42:06 -0700 From: emil@lapel.com (Emil Rojas) Organization: http://groups.google.com/ Message-ID: <754b62e0.0208221442.57b37cfb@posting.google.com> References: <98c05101.0208211331.4b320cea@posting.google.com> <8400a733.0208220014.33fb38ef@posting.google.com> Assuming that each thread is recieving messages you can use a low priority watchdog thread send each an every thread a message periodicly. You know the receiving thread is alive because it replies to the sender. Then you need one watch dog timer (at interrupt level) between your watchdog thread and the timer. Both the timer and the thread need to be able to reset your system. The timer can also watch that periodic system interrupts are progressing also by monitoring interrupt counts. It would be wise at that point, of course, to log as much of the system state as possible in hope of discovering the bug(s). emil --------------------------- Newsgroups: comp.os.vxworks Subject: INUM_TO_IVEC Date: 22 Aug 2002 15:43:39 -0700 From: lin_mu@hotmail.com (Forest) Organization: http://groups.google.com/ Message-ID: Dear All: I am debugging a existing embedded application developed on VxWorks 5.3.1 and motorola mvme2300-MPC 603. There is an interrupts installed using intConnect(INUM_TO_IVEC(MyIntNum), MyISR, 0) where the first argument is the "address offset of the entry in the interrupt vector table". however, I found the macros definition INUM_TO_IVEC in wind/target/h/arch/ppc/ivPpc.h is #define INUM_TO_IVEC(intNum) ((VOIDFUNCPTR *) (intNum)) Seems to me this marco definition is wrong since offset = intNum * 4; In my case intNum=0xF0; The intConnect returns ok, but when I looked at the address where MyISR supposed to resided either at 0xF0 or 0x3c0, the value didn't match with MyISR. Please comment. Regards LM --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Can't boot MV2700 over ethernet Date: Thu, 22 Aug 2002 19:24:44 -0400 From: joe durusau Organization: WEBUSENET.com Message-ID: <3D6572BC.644C93B1@bellsouth.net> References: <190f55f8.0208221029.73da90c8@posting.google.com> Eric Morgan wrote: > I am trying to load vxworks into a MV2700 board over an ethernet > network with a UNIX host. With PPCBug I can ping the host, but the > host cannot ping the MV2700. The client can ping the host, but the > host can't ping the client. Is the problem due to the ethernet > address on the MV2700 using 33 bits, not the standard 32 bits? The > ethernet address has 5 nibbles, not 4. Does this create an address > problem with the ARP? I can replace the 2700 board with a 167 board > and can ping from either direction. Of course if I try to do an NBO, > (attempting to load vxworks over the ethernet) it gives me a code of > $8100, which is file not found. Also, the mv2700 and the host are on > the same network, so that is not the problem. Any ideas? > > Eric Morgan I would suggest that you run snoop or an equiv. program on the host system. That will let you see the net traffic and figure out what's going on. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Can't boot MV2700 over ethernet Date: Thu, 22 Aug 2002 18:54:22 -0500 From: Andrew Johnson Organization: APS, Argonne National Laboratory Message-ID: References: <190f55f8.0208221029.73da90c8@posting.google.com> Eric Morgan wrote: > I am trying to load vxworks into a MV2700 board over an ethernet > network with a UNIX host. With PPCBug I can ping the host, but the > host cannot ping the MV2700. The client can ping the host, but the > host can't ping the client. That's probably because PPCbug doesn't really have a full TCP/IP stack inside it, and it isn't expecting to be pinged so it doesn't have the code to respond to a ping. > Is the problem due to the ethernet > address on the MV2700 using 33 bits, not the standard 32 bits? The > ethernet address has 5 nibbles, not 4. All ethernet addresses are 32 bits or 8 nibbles wide. PPCbug only allows you to set the bottom 20 bits of the 2700's address because the top 12 bits are fixed to Motorola's allocated range of addresses. Other manufacturers may have a different number of bits allocated to their fixed ethernet address range, so allow you to set a different number of bits. This has nothing to do with your problem, ignore it (and don't change the ethernet addresses of your boards from the factory default). > problem with the ARP? I can replace the 2700 board with a 167 board > and can ping from either direction. Is that with the 167's debug monitor running, or with vxWorks booted on it? > Of course if I try to do an NBO, > (attempting to load vxworks over the ethernet) it gives me a code of > $8100, which is file not found. Also, the mv2700 and the host are on > the same network, so that is not the problem. Any ideas? Can you get the 167 to read the 2700's boot file? It obviously can't execute it properly, but it'll give you another clue why the 2700 is failing. If it can't read it, you need to check your tftp server directory. If it can, is your tftp server somehow configured to limit which machines it will respond to? - - Andrew - -- Edsger W. Dijkstra, 1930-2002: "The use of COBOL cripples the mind; its teaching should, therefore, be regarded as a criminal offence." --------------------------- Newsgroups: comp.os.vxworks Subject: ping question!!! Date: 22 Aug 2002 17:18:39 -0700 From: essenlee@hanmail.net (LEE MYUNG SUK) Organization: http://groups.google.com/ Message-ID: 질문> NTP > I NAME ENTRY TID PRI STATUS PC SP ERRNO DELAY - ---------- ------------ -------- --- ---------- -------- -------- - ------- ----- tExcTask excTask fe7018 0 PEND c15e4 fe6f38 0 0 tLogTask logTask fe4690 0 PEND c15e4 fe45b8 0 0 tWdbTask 88194 f48d78 3 DELAY 95330 f48998 0 162 lanRx lanrx_main f46b50 10 PEND 91060 f46558 33 0 main NTPMAIN f41b18 20 DELAY 95330 f41a78 0 53 usrPort serial_main ffd948 40 READY 95d6c ffd7b8 c0002 0 gpsEngine gps_main f455b0 40 DELAY 95330 f45518 0 1 usrLcdKey lcd_key_main f44010 40 DELAY 95330 f43f60 0 1 timeCode tcd_main f41130 40 DELAY 95330 f41088 0 1 telnetRx telnetrx_mai f3fb90 40 PEND 91060 f3fa50 0 0 tNetTask netTask fa6748 50 PEND 91060 fa6690 3d 0 tSnmpd 98b54 f50860 150 PEND 91060 f4feb8 0 0 What does mean this errno(tNetTask, usrPort, lanRx task) In fact, now I have a problem. The prolem is this. This is ping prolem(ping is working and after some time ping timeout. target information is upside. lan chip is using lxt901. Do I add this chip initial? but I don't know that this data sheet routine. Do you know where ethernet driver debug? I want to using half duplex. so psmr value is setting In sysCpmEnetEnable funtion. *PSMR1(immr) &= ~(0x00000001); Is OK? and What does mean SDCR? I can't understand. please anser me.. then thanks... --------------------------- Newsgroups: comp.os.vxworks Subject: no carrier? Date: 22 Aug 2002 19:17:04 -0700 From: joge709@hotmail.com (joge) Organization: http://groups.google.com/ Message-ID: <1d168114.0208221817.842e3b2@posting.google.com> Hi all, Can anybody tell me why the "interrupt 0:dc0 no carrier!" emerages when the bootrom start up on mpcn750? I do not change any value in dec21x40end.h,DEC_USR_FD and DEC_USR_PHY_CHK do not change. Thanks advanced. Joge --------------------------- Newsgroups: comp.os.vxworks Subject: float point in ARM problem Date: 22 Aug 2002 19:21:16 -0700 From: frankzhou@peoplemail.com.cn (frankzhou) Organization: http://groups.google.com/ Message-ID: <58706286.0208221821.55e95ae8@posting.google.com> hi,all, I have a problem of float point operate in ARM architecture. My test environment: Host: win2000 OS: tornado2.0.2/vxworks5.4 Target: ARM7TDMI BSP: SBCARM7 in my projcet, the c:\tornado\host\x86-win32\lib\gcc-lib\arm-wrs-vxworks\2.7.9-970819\libgcc.a is included. content of demo file: #include "vxworks.h" #include "stdio.h" #include "math.h" #include "taskLib.h" int testMain(void); int test(void) { taskSpawn("test",100, VX_FP_TASK ,20000,(FUNCPTR)testMain,0,1,2,3,4,5,6,7,8,9); return 1; } int testMain(void) { double temp1,temp2,temp3,temp4,temp5; temp1 = sin(3.1415926/2); temp2 = cos(0); temp3 = sqrt(16); temp4 = sqrt(0); temp5 = 1/temp3; printf("\ntemp1= %f\n",temp1); printf("temp2= %f\n",temp2); printf("temp3= %f\n",temp3); printf("temp4= %f\n",temp4); printf("temp5= %f\n",temp5); return 1; } int test1(void) { float temp6 = 1.123; printf(" temp6 = %f\n",temp6); return 1; } test rusult is: Attaching interface lo0...done Adding 3053 symbols for standalone. ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]] ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]] ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]] ]]]]]]]]]]] ]]]] ]]]]]]]]]] ]] ]]]] (R) ] ]]]]]]]]] ]]]]]] ]]]]]]]] ]] ]]]] ]] ]]]]]]] ]]]]]]]] ]]]]]] ] ]] ]]]] ]]] ]]]]] ] ]]] ] ]]]] ]]] ]]]]]]]]] ]]]] ]] ]]]] ]] ]]]]] ]]]] ]]] ]] ] ]]] ]] ]]]]] ]]]]]] ]] ]]]]]]] ]]]] ]] ]]]] ]]]]] ] ]]]] ]]]]] ]]]]]]]] ]]]] ]] ]]]] ]]]]]]] ]]]] ]]]]]] ]]]]] ]]]]]] ] ]]]]] ]]]] ]] ]]]] ]]]]]]]] ]]]] ]]]]]]] ]]]]] ] ]]]]]] ] ]]] ]]]] ]] ]]]] ]]]] ]]]] ]]]] ]]]]]]]] ]]]]] ]]] ]]]]]]] ] ]]]]]]] ]]]] ]]]] ]]]] ]]]]] ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]] ]]]]]]]]]]]]]]]]]]]]]]]]]]]]] Development System ]]]]]]]]]]]]]]]]]]]]]]]]]]]] ]]]]]]]]]]]]]]]]]]]]]]]]]]] VxWorks version 5.4.1 ]]]]]]]]]]]]]]]]]]]]]]]]]] KERNEL: WIND version 2.5 ]]]]]]]]]]]]]]]]]]]]]]]]] Copyright Wind River Systems, Inc., 1984-1999 CPU: KS32C50100 for WindRiver SBC ARM7. Processo r #0. Memory Size: 0x400000. BSP version 1.2/10. WDB: Ready. - -> test value = 1 = 0x1 - -> Data abort Exception address: 0x0009ec2c Current Processor Status Register: 0x80000013 Task: 0x30fafc "test" - -> sp test1 task spawned: id = 0x30ab30, name = t1 value = 3189552 = 0x30ab30 - -> Data abort Exception address: 0x000a03e8 Current Processor Status Register: 0x00000013 Task: 0x30ab30 "t1" - -> in the 'vxworks programmer's guide',it describe: On ARM processors, double variables have a different format from IEEE double on most other processors. The bit pattern used by the ARM hardware and software floating point implementations follows the IEEE standard; however, the byte order is different from standard practice leading to a cross-endian implementation. Be careful when sharing double values in memory between ARM and other processors. it seems the problem is result of the double variables format. Does anyone has a good idea to slove this problem? any suggestion is welcome. --------------------------- Newsgroups: comp.os.vxworks Subject: binutil 2.10.1, gcc 2.95.4 & elfHex problem on Solaris. Date: 22 Aug 2002 20:19:22 -0700 From: chkim@irts.co.kr (Changhwan Kim) Organization: http://groups.google.com/ Message-ID: Hi. I compiled binutils-2.10.1 and gcc-2.95.4 for powerpc-wrs-vxwors target on solaris host. (Thanks to Dave K.) It seems working fine with application build. But when I build boot rom (bootrom_uncmp.hex) with new compiler, it cause problem. The problem is, In bootrom_uncmp.hex, .rodata section located from the start of s-record. .text should at the start of .hex for proper boot. I don't know why this happens. Even worse I don't have any information how elfHex works... Anyone have experienced this problem? or any good idea to resolve this problem? Now I'm trying to generate s-record file with objcopy. But, objcopy cannot modify start address with '--set-start' option. Need some modification with generated s-record to adjust them. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Unix equivalents for VxWorks. Date: 22 Aug 2002 23:04:14 -0700 From: kiranc@tataelxsi.co.in (Kiran.C) Organization: http://groups.google.com/ Message-ID: References: <74371f59.0208190832.6115b14d@posting.google.com> Check out these... http://www.xenomai.org/ http://legacy2linux.sourceforge.net/ Kiran tulasi@sionintl.com (Tulasidas) wrote in message news:<74371f59.0208190832.6115b14d@posting.google.com>... > Hi VxWorks Folks, > > I am trying to convert VxWorks function calls to Unix equivalent so > that > out product will run on UniX platform. I have identified the VxWorks > calls in our product. > > Any one done this type of porting before.? Do you know of any URL > where I can get the equivalent Unix calls? This will save my time in > developing wrappers for Unix functions. > > > Thanks. > > S.Tulasidas --------------------------- Newsgroups: comp.os.vxworks Subject: Boot from IDE Flash Disk with x86 target Date: 22 Aug 2002 23:43:38 -0700 From: andreas_lindell75@hotmail.com (Rikard) Organization: http://groups.google.com/ Message-ID: <5947a04.0208222243.76bbfe@posting.google.com> Hello, I am trying to boot vxworks from an IDE Flash disk IFD-350-64 but to now succes. I have done vxsys on the flash disk and then copied me bootrom.sys to the flash disk and placed a vxworks image in the root of the flash disk. To boot it I type @ and then enter. The error i get is: ""Attaching to TFFS... drive is out of range (0--1)" "usrTffsConfig failed" Error loading file:errno = 0x0 dosFsDevInit failed usrFdConfig failed Now I have some questions: 1. Does vxworks make a difference between a IDE flash disk and a IDE hard disk? According to the manufacturer the IFD-350-64 is "full ide hard disk compatible . Does that mean that the system/vxworks can not tell if its using a ide hard disk or a ide flash disk? 2. If I want to boot from a ide flash disk how do I do that? I have done the following: 2.1 Replaced the sysTffs.c in the bsp directory with the doc2k-sysTffs.c (and renamed it). After that I added MATCH EXTRA = sysTffs.o in the make file. 2.2 In config.h I included: #define INCLUDE_DOSFS /* include dosFs file system */ #define INCLUDE_FD /* include floppy disk driver */ #define INCLUDE_LPT /* include parallel port driver */ #undef INCLUDE_ATA /* include IDE/EIDE(ATA) hard disk driver */ #undef INCLUDE_TIMESTAMP /* include TIMESTAMP timer for Wind View */ #define INCLUDE_TFFS /* include TrueFFS driver for Flash */ #undef INCLUDE_PCMCIA /* include PCMCIA driver */ #ifdef INCLUDE_PCMCIA #define INCLUDE_ATA /* include ATA driver */ #define INCLUDE_SRAM /* include SRAM driver */ #undef INCLUDE_TFFS /* include TFFS driver */ #ifdef INCLUDE_NETWORK #define INCLUDE_ELT /* include 3COM EtherLink III driver */ #endif /* INCLUDE_NETWORK */ #endif /* INCLUDE_PCMCIA */ #ifdef INCLUDE_TFFS #define INCLUDE_SHOW_ROUTINES #endif /* INCLUDE_TFFS */ and added the line: #define DEFAULT_BOOT_LINE \ "tffs=0,0(0,0)host:/tffs0/vxworks h=192.168.0.247 e=192.168.0.241 u=vxworks pw=vxworks tn=target" Then I built the bootrom and copied it to the IDE flash disk under then name bootrom.sys. 2.3 Place a vxworks image in the flash root and turned my system on and then I get these errors: "Attaching to TFFS... drive is out of range (0--1) usrTffsConfig failed Error loading file:errno = 0x0 dosFsDevInit failed usrFdConfig failed" Any one have any clue to what causes these errors and how to fix them? 3. I have also tried to see if I can use a bootrom.sys that specifies an IDE/ATA hard drive but connecting the ide flash disk just to try if the system makes any difference. This bootrom.sys file works fine with a Quantum IDE hard disk and when replacing the Quantum with the flash disk the bootrom starts up and I do a @ and get the following error: "Attaching to ATA disk device... Error during dosFsDevInit: 380001 usrAtaConfig falied Error loading file: errno = 380001 dosFsDevInit failed usrFdConfig failed" Any idea what this can be? Of course if vxworks makes a big difference between ide flash disk and ide hard disk I understand why this does not work but since IFD-350-64 is compatible with hard disk, according to the manufacturer this should work as well. This question brings us back to question number 1. However any input in this question would be appreciated. 4. I have one option left that I haven't tried out and the reason is that this will take some time to do and I would like to have some input on this. I have one Readme file form M-Systems: * * * Installation Guide for DiskOnChip(R) Product Family on VxWorks(R) * * Operating System (x86 edition). * * Version 5.0, August 01 2001 * * Copyright M-Systems (c) 2001 * *********************************************************************** This file specifies on how to integrate DiskOnChip with vxworks. Is a DiskOnChip the same as an IDE flash disk? This document is fairly long and I may do it and then find out that it wont work:(:( The doucment is divided into the following sections: CONTENTS - -------- 1. Introduction 2. Installing DiskOnChip driver 3. Booting VxWorks from DiskOnChip 4. Driver's run-time configuration options 5. Formatting DiskOnChip from VxWorks application 6. Considerations related to unexpected power shutdowns 7. Alternative "garbage collection" scenario 8. Access to advanced features of DiskOnChip 9. Known problems Appendix A. Re-formatting DiskOnChip under MS-DOS Appendix B. Partitioning the disk with MS-DOS FDISK Appendix C. Using WindRiver's VXSYS and VXCOPY utilities And my questions is as following. Do I need to do step 2. Installing DiskOnChip driver before I do: 3. Booting VxWorks from DiskOnChip I have only taken a quick glance at these steps and my first impression is that step 2 is required if I intend to use the flash in my application but that is not my intention. I just want to boot the vxworks image from the flash not do any read or write from the application. So if anyone has worked with this document maybe she/he can inform me if step 2 is a "must do" before step 3? I am not including the entire document from M-Systems becasue it is fairly big but anyone who has worked on flash disk may have seen it and could answer my question. If any one knows the answer for any of the question pls assist me or maybe you can push me in the right direction because now I am running in circles. Regards Andreas --------------------------- Newsgroups: comp.os.vxworks Subject: Re: TCP connection released, but socket not closed Date: Fri, 23 Aug 2002 09:04:19 +0200 From: "Jihad Haddad" Organization: TDC Internet Message-ID: <3d65de74$0$53625$edfadb0f@dspool01.news.tele.dk> References: <3d649756$0$53634$edfadb0f@dspool01.news.tele.dk> <3d64ab49$0$88404$edfadb0f@dspool01.news.tele.dk> Thanks for all the replies. Now still keeping in mind that I am new to this I will try to describe the situation as I see it in my language, hopefully close to yours: The problem is as I see it, and which I did not mention in my first post, is that I am using ONCRPC on the PC side, and the TCP connections I create are to perform RPC calls into the "box". Now the procedure that waits for TCP packets over the net, which is a procedure in the RPC server on VxWorks ( svc_run() ), makes a select(0 , &readfds, (int *)0, (int *)0, (struct timeval *)0) ) As I understand it, the RST TCP packet should send an error message to the select, where the 4'th parameter (which is a 0 (zero)) is supposed to wait for it. And when it is received, it should call close(fd). But the 4th parameter is not supported in VxWorks as far as I understand. So either my understanding of how to perform RPC calls to VxWorks is wrong, or the implementation of the RPC server on VxWorks is erronious, because there has to be some mechanism to close the descriptors when we close the TCP connection. I can for example see in the implementation of the ftp server, that it performs reads instead of selects, and it can therefore detect that the connection is non-existent, and can therefore close the fd. But in the case of a select, something else has to be performed, which I cannot see any support for in VxWorks. Please I would be so glad for any suggestions you come with. Thank you very much Jihad Haddad "Jihad Haddad" wrote in message news:3d64ab49$0$88404$edfadb0f@dspool01.news.tele.dk... > I found out what the error is when I "write" to not closed socket: > It says: S_errno_EPIPE > if that should help a bit. > > Thanks again > > Jihad Haddad > > "Jihad Haddad" wrote in message > news:3d649756$0$53634$edfadb0f@dspool01.news.tele.dk... > > Hi > > > > I am a total newbe to this, and my use of terms is still that of a 6 > > year old. So please be patient with me. > > > > We have a "box" that contains a Strong Arm processor, an ethernet plug > > and that runs VxWorks. I can easily create a TCP connection from a > > windows 2000 PC to that box. The problem is that when I close the > > windows socket that has the TCP connection, the TCP connection is > closed > > in both PC and "box". But the socket in the box (that is the VxWorks > > socket) is not closed. This I can see if I have a telnet connection to > > the box, and I run the command "iosFdShow" which shows the list of > > descriptors opened. I can here see that the socket that was opened > with > > the TCP connection, is not closed. > > I can close the socket if I call "close fd" where fd is the number of > > the descriptor the socket has. But a write to the socket gives an > error. > > If I call "errno" afterI make the "write", i get the following error > > code: > > errno = 0x179b4c: value = 1835009 = 0x1c0001 > > I do not know if the above error code has anything to do with the > error > > of the "write". > > > > I have also sniffed the packet sent by the PC to close the TCP > > connection. Some information about the packet is the following: > > ---------------------------------------------------------------------- > - > > Protocol: cadlock > > Summary: TCP (RST) > > Size: 60 > > The packet looks as follows: > > 0000: 00 02 15 00 00 3F 00 04 76 1A C7 2A 08 00 45 00 > > 0010: 00 28 2E 34 40 00 80 06 00 00 0A 07 0A 25 0A 07 > > 0020: 65 5D 02 C1 03 E8 85 55 55 6A 7C 69 3F B9 50 04 > > 0030: 00 00 8E C5 00 00 > > ---------------------------------------------------------------------- > - > > > > Does anyone have any idea why this is happening. Is it the way I > > initialize the TCP protocol? > > > > I do not know if this is enough information, but I can give more if > > needed. > > > > Thank you very much in advance > > Jihad Haddad > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ata-dma on vxworks Date: 23 Aug 2002 00:36:56 -0700 From: zhoushan@371.net (zhoushan) Organization: http://groups.google.com/ Message-ID: References: hi,Leonid Rosenboim,can you help me? --------------------------- Newsgroups: comp.os.vxworks Subject: How Can I send raw data in vxworks through tty driver ? Date: 23 Aug 2002 00:58:31 -0700 From: robert_fan@citiz.net (Robert Fan) Organization: http://groups.google.com/ Message-ID: <4c2e8c4.0208222358.57fc1715@posting.google.com> Hi all , My board only have one serial port , it act as vxworks's console as system start up . then I want to use this port to send out some raw data , but I find it cant send out the HEX zero at all when I use standard I/O function with tty driver . How can I do this ? Any one have idea ? Thanks a lot Robert Fan --------------------------- Newsgroups: comp.os.vxworks Subject: examples Date: 23 Aug 2002 03:06:49 -0700 From: l.salvemini@infosol.it (Lorenzo Salvemini) Organization: http://groups.google.com/ Message-ID: Hi everybody, i'm a new Tornado 2 VxWorks user, does anybody know where I can find some execrises or example codes??? Thanks a lot bye --------------------------- Newsgroups: comp.os.vxworks Subject: error massage whlie trying to download.. Date: Fri, 23 Aug 2002 12:18:29 +0200 From: "tomer" Organization: The Israel Inter-University Computation Center Message-ID: hi. while trying to download an object file to the target i get this massage. - -> ld Organization: WEBUSENET.com Message-ID: <3D660D39.1044F247@bellsouth.net> References: <4c2e8c4.0208222358.57fc1715@posting.google.com> Robert Fan wrote: > Hi all , > My board only have one serial port , it act as vxworks's console as > system start up . then I want to use this port to send out some raw > data , but I find it cant send out the HEX zero at all when I use standard > I/O function with tty driver . > How can I do this ? Any one have idea ? Thanks a lot > > Robert Fan Standard IO drivers can't handle raw data on serial ports. You could use kermit or some such if you need to send a lot of data, or if just a little you could sprintf() to gen up a string. The serial driver should handle that OK. I suspect that anything else would be more trouble. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Re: examples Date: Fri, 23 Aug 2002 06:25:06 -0400 From: joe durusau Organization: WEBUSENET.com Message-ID: <3D660D82.CB1C0EA0@bellsouth.net> References: Lorenzo Salvemini wrote: > Hi everybody, > i'm a new Tornado 2 VxWorks user, does anybody know where I can find > some execrises or example codes??? > Thanks a lot > bye Not T-II specific, but you could check Embry-Riddle's website. They have a bunch of examples of vxWorks stuff there. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Re: "Virtual" end driver Date: 23 Aug 2002 03:26:59 -0700 From: Gunter.Ritter@tenovis.com (Gunter Ritter) Organization: http://groups.google.com/ Message-ID: References: Hi, the ip adresses are totaly different. One is 192.1.2.141, the other I tried to set to 143.123.19.141. The subnetmask is set to 255.255.255.0. happyhappy779@hotmail.com (Satish G) wrote in message news:... > Hi, > > You have not given the IP addresses used for the interfaces, so I > would suspect that the IPaddress/netmask combination is resulting in > same network for both interfaces. > > Satish G > > Gunter.Ritter@tenovis.com (Gunter Ritter) wrote in message news:... > > Hello VxWorkers, > > > > I have written a "virtual" end driver as a secondary end driver for > > our system. > > "Virtual" means that this end driver has no "physical" ethernet > > hardware. > > > > The primary end driver is working with ethernet hardware. > > > > I need this secondary end driver only to exchange ip packets with the > > TCP/IP stack of vxworks and let vxworks do the routing between the two > > networks. > > > > I have a problem when I try to assign an ip address. > > muxDevLoad(), muxBind(), ipAttach() and ifMaskSet() work fine. > > ifShow() shows me the settings of the secondary driver. > > But calling ifAddrSet() with an unique ip address causes the network > > functionality of vxworks to crash. > > > > Even "ping" does no more work with the primary end driver after > > calling ifAddrSet(). > > > > Does anybody can give me a hint whats going wrong with my secondary > > end driver? > > > > Thanks. --------------------------- Newsgroups: comp.os.vxworks Subject: tt Date: Fri, 23 Aug 2002 12:36:03 +0200 From: "tomer" Organization: The Israel Inter-University Computation Center Message-ID: --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Wind Web Server - httpBlockPut Date: 23 Aug 2002 04:07:32 -0700 From: lchamberlain@pgaero.co.uk (Laura Chamberlain) Organization: http://groups.google.com/ Message-ID: References: Thanks for your ideas but I am fairly sure that it is not a memory configuration problem as the mbuf's, num_files etc were all increased sometime ago. Also my web page which shows memory usage, TCP stats, INET stats and network buffer usage seems to indicate there is no problem in that area. Anyway.. WindRiver have now supplied another version of MotFecEnd.c and 2 out of 3 overnight tests ran without crashing, the 3rd had an exception on the net task (something not previously seen) so it is looking like the crashing problem resides there. Still have not seen evidence of short downloads on the released version of WWS 2.0 but testing is still in early stages. Thanks for your help, I'll let you know how it goes. Laura "drdiags" wrote in message news:... > Hello Laura, > > Here is something I found on WindSurf: > > TITLE: Wind Web Server 2.0 sucks up buffers at a high rate causing > crashes. > SPR #: 34756 > STATUS: Accepted > IDE: > RTOS: > PRODUCT NAME: Wind Web 2.0 > RELEASE STATUS: FCS > PRODUCTS AFFECTED: Wind Web > HOST: All > HOST OS: All All > ARCH FAMILY: All > ARCH CPU: All > BSP: > DESCRIPTION: Wind Web Server 2.0 sucks up buffers at a high rate > causing crashes. > > Wind Web Server can't handle a light load very well with crashes > occuring very easily. Using RPM's to service the http requests will cause > problems on simple test cases. See the test case for exact info. > > I don't have this product installed, but is it possible to run the > Tornado Browser and look > at memory consumption while your application is running. This will let > you know if the above problem > is the issue or whether you are leaking memory someplace. The buffers > above are probably > the mbufs for networking and won't be visible using the Browser. I > will look around some more > for other possible answers, but your depletion of mbufs and the > TIME_WAIT sockets could mean > you may have to increase your network buffers or NUM_FILES (WAG on my > part). > > > "Laura Chamberlain" wrote in message > news:efdaef76.0208202335.75f3c35a@posting.google.com... > > "drdiags" wrote in message > news:... > > > > > Just more questions. When the system stops responding, does it ever > recover? > > No, the whole thing locks up (the main application stops too). If the > > watchdog is connected the system reboots. > > > > > When it stops responding, are any tasks suspended? > > See above - can't get debug tools in > > > > > Have you taken a look on WindSurf for any possible SPR fixes for your > version of Wind Web Server? BTW, what version is this? > > Yes, was running WWS 2.0 Beta but have since upgraded to the released > > version and installed patch 1 from WindSurf, can't see any other > > similar problems > > > > > Maybe someone who has used this product may have their memory jogged if > you provide this. > > > Could you print out the errno upon the ERROR return, this may help. > > HttpBlockPut only returns HTTP_OK or HTTP_ERROR, when the system dies > > or does a short download can't see any error message on the terminal - > > is there something else I should be looking at? > > > > > What version of Tornado, what architecture, BSP and network interface > type? > > Tornado 2.0.2, VxWorks Ver 5.4, ADS860 BSP ver 1.2/5, MotFec Network > > using source 01i,28jun00 (waiting for later ver from windriver). The > > data to send is compiled from DiskOnChip (TFFS and DOSFS2) and FLASH. > > > > > > > How many sockets are open when the error occurs? > > Don't know once crashed cant get debug tools in but I will find out > > for short downloads. During normal operation you can get several > > connections in the TIME_WAIT state, occasionally you get too many and > > the system pauses while they free up, do you think this is releated to > > the crashing? > > > > > > Here is an extract from my configuration:- > > > > #define HTTP_MAX_SOCKQ 5 > > #define HTTP_MAX_SOCK_BUFF 4096 > > #define HTTP_SOCK_TIMEOUT 60 > > #define HTTP_SOCK_ARRAY 4 > > #define HTTP_MAX_GLOBAL_MEM_PART_SIZE 15360 > > #define HTTP_MAX_REQ_PART_SIZE 15360 > > #define HTTP_MAX_REQLINE_SIZE 1024 > > #define HTTP_MAX_ARGUMENTS 10 > > #define HTTP_MIME_MAXLINES 10 > > #define HTTP_NR_OF_TASKS 4 > > #define HTTP_TASK_STACK_SIZE 6114 > > > > > > > Sorry I don't have an answer, but if you share the above information, > > > someone will join in. Also, did you contact Wind River support about > > > this? > > Yes but there response has been slow and so far no useful suggestions, > > they are now asking me to provide a test case but this is going to > > take me some time to prepare. > > > > > > Thanks > > > > Laura > > > > > HTH....... > > > "Laura Chamberlain" wrote in message > > > news:efdaef76.0208190544.5ebd4eac@posting.google.com... > > > > Hi, > > > > > > > > Thanks for your response, I have built a webpage which shows the web > > > > server config, memory usage and the network stats you suggest along > > > > with tcpstatShow as it uses TCP, but I can't see anything wrong. Any > > > > other ideas? > > > > > > > > "drdiags" wrote in message > news:... > > > > > Hello Laura, > > > > > > > > > > Not sure I can help, but is it possible that you are running out > of > > > > > network > > > > > resources, i.e. #file descriptors are maxed out. Are the transfers > done > > > > > using > > > > > UDP or TCP (I don't have the product installed)? I would run the > various > > > > > network debugging tools from the target shell: > > > > > > > > > > inetstatShow > > > > > mbufShow > > > > > netStackSysPoolShow > > > > > netStackDataPoolShow > > > > > endPoolShow "interface name" <<-- Look on vxWorks FAQ for > utilities.c > > > > > for this > > > > > routine. > > > > > iosFdShow > > > > > i > > > > > others that I cannot think of > > > > > > > > > > HTH. > > > > > > > > > > "Laura Chamberlain" wrote in message > > > > > news:efdaef76.0208160356.3b1636dc@posting.google.com... > > > > > > I am using a custom RPM to compile data from several places and > send > > > > > > to the user as a single file, up to 40Mb in total, this is sent > using > > > > > > httpBlockPut in blocks of 1024 - similar to the httpRpmDownload > > > > > > function in Tornado\target\unsupported\http\rpm\httpRpmUpload.c. > > > > > > > > > > > > The first problem is sometimes the httpBlockPut function fails > > > > > > (returns ERROR), resulting in an incomplete download - using > internet > > > > > > explorer you don't receive an error just a small file. > > > > > > > > > > > > The second problem I have found while testing the first - after an > > > > > > hour or so of repeatedly downloading data the target stops > responding > > > > > > and there is no debug output to help me determine why. > > > > > > > > > > > > Can anyone help? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: tt Date: Fri, 23 Aug 2002 11:57:33 +0000 (UTC) From: Johan Borkhuis Organization: Agere Systems Message-ID: References: - -> tt Sorry, traces of my own stack begin at tt (). value = -1 = 0xffffffff = udbinfo + 0xffe48247 Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Set RTS and DTR on MVME2700 Date: 23 Aug 2002 05:41:57 -0700 From: f.petri@steinbichler.de (F. Petri) Organization: http://groups.google.com/ Message-ID: <839bd939.0208230441.5f772780@posting.google.com> Hello, I need to set the RTS and DTR of COM2 on an MVME2700. I guess I should use the ioctl-function, but do not find an appropriate option for it. I invoked: *fd = open("/tyCo/1", O_RDWR, 0); /* open COM 2 */ ioctl(*fd, FIOBAUDRATE, baud); /* set baud rate to 'baud' */ ioctl(*fd, FIOOPTIONS, OPT_RAW) /* set raw options */ I'm a newbie and any help would be much appreciated. Frieder. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: INUM_TO_IVEC Date: Fri, 23 Aug 2002 14:42:26 +0200 From: Klaus Zibell Message-ID: <0sacmuodrtplrmglj1td4mj1k2931jemcj@4ax.com> References: On 22 Aug 2002 15:43:39 -0700, lin_mu@hotmail.com (Forest) wrote: >Dear All: > >I am debugging a existing embedded application developed on VxWorks >5.3.1 and >motorola mvme2300-MPC 603. > >There is an interrupts installed using >intConnect(INUM_TO_IVEC(MyIntNum), MyISR, 0) where the first argument >is the "address offset of the entry in the interrupt vector table". > >however, I found the macros definition INUM_TO_IVEC in >wind/target/h/arch/ppc/ivPpc.h is >#define INUM_TO_IVEC(intNum) ((VOIDFUNCPTR *) (intNum)) > >Seems to me this marco definition is wrong since >offset = intNum * 4; >In my case intNum=0xF0; > >The intConnect returns ok, but when I looked at the address where >MyISR supposed to resided either at 0xF0 or 0x3c0, the value didn't >match with MyISR. > You are new to PowerPC? The MPC603 has only one IRQ-Pin, so there is only 1 vector at 0x500 for all external interupts. I dont't know your board but normaly there is an external IRQ controller which has to be programmed which is done by a device driver. This device driver maintains a list of which ISR belongs to which IRQ number. And this liost is manipulated with the call to intConnect (). Only my 2 cents Klaus --------------------------- Newsgroups: comp.os.vxworks Subject: Re: error massage whlie trying to download.. Date: Fri, 23 Aug 2002 12:51:29 GMT From: David Cooper Organization: none Message-ID: References: Is the working directory set to the directory where the test3.o file is found ? Do a "pwd" command to see the current working directory and then "cd" commands to set to proper directory. Or specify the directory path in the "ld" command. ...david In article , bahart@bgumail.bgu.ac.il says... > hi. > > while trying to download an object file to the target i get this massage. > > -> ld Loading test3.o WTX Error 0x10140 (API_FILE_NOT_ACCESSIBLE) > value = -1 = 0xffffffff > > what can be the reason ? > > p.s i am a beginner . > > > - -- +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ ô¿ô Genealogy Utilities & Family Research Researching: ALLEN,BERZONER,COHEN,COOPER,KUPFERZMIDT, POSNER (POIZNER), PRITIKIN, STARKMAN Dachau Camp Entry List, Genealogy Utilities http://home.adelphia.net/~dcooper000 --------------------------- Newsgroups: comp.os.vxworks Subject: Thread safe in math.h Date: 23 Aug 2002 06:43:59 -0700 From: andreas_lindell75@hotmail.com (Rikard) Organization: http://groups.google.com/ Message-ID: <5947a04.0208230543.7cde3852@posting.google.com> Hello, I have some questions about thread safety wit vxworks. 1. Are the functions in the Maths library (Math.h) thread safe ? (i.e. Will the results of math library functions return strange values if the maths functions are being called from different tasks and the maths calls overlap ?) 2. Are there any specific flags used to enable Math Library thread safety ? Regards Rikard --------------------------- Newsgroups: comp.os.vxworks Subject: Re: File System Problem Date: Fri, 23 Aug 2002 14:56:44 +0100 From: "Dimitrios Siganos" Organization: Newport Networks Ltd. Message-ID: <1030111005.389899@azores.network-i.net> References: #define INCLUDE_DISK_UTIL Dimitris wrote in message news:ak5922$rja$1@overload.lbl.gov... > > > Hi > > We are using Tornado 2.1 for MIPS for developing the VxWorks 5.4 BSP for MIPS > processor. > We have included the Dos File System using ICLUDE_DOSFS and have created > RamDisk. > But we are unable to run the commands such as dir, cd, mkdir etc. from target > shell. These are > not available in the target symbol table also. > > We have inluded the usrLib.h in our BSP and we could run the rest of the > commands mentioned > in that header file. > > Please advise if there is any configuration that need to be set-up or included > in the BSP for executing > these commands from the shell prompt. > > Thanks in advance. > > Regards > Ramakrishna > Force Computers- India Design Center, > Bangalore > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: float point in ARM problem Date: 23 Aug 2002 08:55:53 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0208230755.4cdb8bec@posting.google.com> References: <58706286.0208221821.55e95ae8@posting.google.com> Hello, Your conclusion that it is to do with the format of the double on ARM seems illogical since you are not (obviously) getting these numbers from another system. If all the floating point numbers are managed by the compiler/floating point libraries, then there should be no problem. Far more useful here would be to get the stack trace from the task once it has crashed, and try to determine exactly what is causing it to raise an exception like this. frankzhou@peoplemail.com.cn (frankzhou) wrote in message : [snip] > -> test > value = 1 = 0x1 > -> > Data abort > Exception address: 0x0009ec2c > Current Processor Status Register: 0x80000013 > Task: 0x30fafc "test" Try 'tt 0x30fafc' here so you know what function it was in when it failed, and how it got there. > -> sp test1 > task spawned: id = 0x30ab30, name = t1 > value = 3189552 = 0x30ab30 > -> > Data abort > Exception address: 0x000a03e8 > Current Processor Status Register: 0x00000013 > Task: 0x30ab30 "t1" Again, tt 0x30ab30 should provide some useful information. HTH, John... --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Boot from IDE Flash Disk with x86 target Date: 23 Aug 2002 09:26:44 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0208230826.17b12393@posting.google.com> References: <5947a04.0208222243.76bbfe@posting.google.com> andreas_lindell75@hotmail.com (Rikard) wrote in message news:<5947a04.0208222243.76bbfe@posting.google.com>... > I am trying to boot vxworks from an IDE Flash disk IFD-350-64 but to > #undef INCLUDE_ATA /* include IDE/EIDE(ATA) hard disk driver > #define INCLUDE_TFFS /* include TrueFFS driver for Flash */ > #define DEFAULT_BOOT_LINE \ > "tffs=0,0(0,0)host:/tffs0/vxworks h=192.168.0.247 e=192.168.0.241 > u=vxworks pw=vxworks tn=target" I don't know what exactly went wrong when you tried ATA, but vxWorks definitly handles ATA compatible falsh disks like fixed disks. According to your description it seems your IDE disk has an ATA interface, therefore you have to use ataDrv (neither DiskOnChip nor TFFS). INCLUDE_ATA should be the first step todo this. Maybe you'll have todo some adaptions in your BSP (e.g. pc.h, interrupt of this ATA channel, I/O space) or you simply have to adjust partition mapping (are you using DosFS2, do you have a partition table on the disk)? I'd suggest to bring up a vxWorks image with debug interface (using a floppy?) and try to mount/access the disk using ata to further debug this issue. In case this is a simply filesystem/partition issue, keep in mind there is an ATA RawIO command (for testing). Of Course using ATA will also reflect in your bootline.... toni --------------------------- Newsgroups: comp.os.vxworks Subject: Re: debug interrupt Date: 23 Aug 2002 10:08:50 -0700 From: rishi_bhargava@yahoo.com (Rishi Bhargava) Organization: http://groups.google.com/ Message-ID: <2474871.0208230908.1335e7be@posting.google.com> References: <7efa47da.0208192055.5558344a@posting.google.com> <2UYH+bGB18Y9EAb0@wtoram.co.uk> <70d78653.0208211919.2ef7cf38@posting.google.com> This is not related to interrupts but related to system mode debugging. I am trying to use wdbSystemSuspend when the system comes up so that I can attach the debugger at an early stage. I get linker error (undefined reference to wdbSystemSuspend) when trying to compile with wdbSystemSuspend. I have the WDB_COMM_TYPE defined as WDM_COM_END. END driver are enabled... and INCLUDE_WDB_SYS is also defined. Any clue what might be going wrong thanks, Rishi hsingyuan@yahoo.com (Hsing Yuan) wrote in message news:<70d78653.0208211919.2ef7cf38@posting.google.com>... > Tornado 2.0.2 does have a bug in system mode. I reported the bug, and > WRS reproduced the problem. However, they found that the problem is > not reproducable in the new tornado to be released (this month(8/2002) > ?). > > Using a semaphore in a interrupt handler may not be a good idea since > it may cause workQ overflow. > > We use ICE to debug interrupt handlers. > > Hsing > > Wade Oram wrote in message news:<2UYH+bGB18Y9EAb0@wtoram.co.uk>... > > In message , Faure > > writes > > > > > >Note: I have some difficulties to debug in system mode, there are some > > >bugs with that mode under Tornado 2.0.2: > > >SPR#72141, T202: system mode debugging fails > > >after some time / target hangs) > > > > > > > I have successfully used system mode in T2 CP2 (Windows NT4 host, X86 > > target) with additional patches although I have not had to use it for > > extended periods. How long is 'after some time'? > > > > >if it is sufficient,I advice you to debug in task mode (not in system > > >mode), > > >it is what I choosed to do after spending too much time to try to make > > >work without success the system mode. > > > > The original poster wanted to set a break point in an interrupt handler. > > For this purpose, only system mode can be used. However, the advice > > still stands - find out as much information as you can about the fault > > in task mode and restrict system mode debugging to the occasions that > > really demand it. > > > > If an interrupt handler really is causing you grief and system mode > > debugging is not reliable enough to solve the problem, you *may* be able > > to change the interrupt handling so that the interrupt handler does > > nothing except give a semaphore to a (very) high priority task which > > does the interrupt handling. This will then allow you to put a break > > point in the original interrupt handler whilst still in task mode. Once > > the fault is found the original interrupt handler - hopefully now fixed > > - can be restored to its rightful place. > > > > As a technique, it won't be applicable all problems with interrupt > > handlers but it can be used to help solve many - especially when the > > interrupt in question is independent of any other interrupts in the > > system. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: INUM_TO_IVEC Date: 23 Aug 2002 10:27:50 -0700 From: lin_mu@hotmail.com (Forest) Organization: http://groups.google.com/ Message-ID: References: > The intConnect returns ok, but when I looked at the address where > MyISR supposed to resided either at 0xF0 or 0x3c0, the value didn't > match with MyISR. The is ok , since MyISR is wrapped by intHandlerCreate. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Thread safe in math.h Date: Fri, 23 Aug 2002 12:27:36 -0500 From: Andrew Johnson Organization: APS, Argonne National Laboratory Message-ID: References: <5947a04.0208230543.7cde3852@posting.google.com> Rikard wrote: > 1. Are the functions in the Maths library (Math.h) thread safe ? (i.e. > Will the results of math library functions return strange values if the > maths functions are being called from different tasks and the maths calls > overlap ?) Yes, (no). > 2. Are there any specific flags used to enable Math Library thread safety ? There is a flag VX_FP_TASK that you have to give when you create any task that uses floating-point calculations. That tells the task switch code in the kernel to save/restore the floating poing variables when switching into and out of this task. If you don't set the flag everywhere it is really needed then you can expect wierd things to happen at random to just about any other FP task in the system. Note that some versions of ccppc (GCC for PowerPC) actually use the floating point registers to do non-FP things, and you might have to set the flag even when your code shouldn't need it (or give a -mstrict-align flag to ccppc which appears to cure that little foible). - - Andrew - -- Edsger W. Dijkstra, 1930-2002: "The use of COBOL cripples the mind; its teaching should, therefore, be regarded as a criminal offence." --------------------------- Newsgroups: comp.os.vxworks Subject: Use Parallel Port instead of the required Serial Port ? Need Null Modem line ? Date: 23 Aug 2002 13:16:34 -0700 From: patrickyu_ca@yahoo.com (patrick) Organization: http://groups.google.com/ Message-ID: Hi, all I want to use the notebook as the host. But I do not have the serial port except the parallel port. Can I use Parallel port instead of the required serial port ? If can, need any special configuration ? Need I use the Null Modem line or common line is ok for Vxworks serial connection? Thanks first. Patrick --------------------------- Newsgroups: comp.os.vxworks Subject: Re: "Virtual" end driver Date: 23 Aug 2002 14:07:49 -0700 From: happyhappy779@hotmail.com (Satish G) Organization: http://groups.google.com/ Message-ID: References: Hi, Would you mind posting the route table after ifAddrSet() call(routeShow/mrouteShow)? Gunter.Ritter@tenovis.com (Gunter Ritter) wrote in message news:... > Hi, > > the ip adresses are totaly different. One is 192.1.2.141, the other I > tried to set to 143.123.19.141. The subnetmask is set to > 255.255.255.0. > > > happyhappy779@hotmail.com (Satish G) wrote in message news:... > > Hi, > > > > You have not given the IP addresses used for the interfaces, so I > > would suspect that the IPaddress/netmask combination is resulting in > > same network for both interfaces. > > > > Satish G > > > > Gunter.Ritter@tenovis.com (Gunter Ritter) wrote in message news:... > > > Hello VxWorkers, > > > > > > I have written a "virtual" end driver as a secondary end driver for > > > our system. > > > "Virtual" means that this end driver has no "physical" ethernet > > > hardware. > > > > > > The primary end driver is working with ethernet hardware. > > > > > > I need this secondary end driver only to exchange ip packets with the > > > TCP/IP stack of vxworks and let vxworks do the routing between the two > > > networks. > > > > > > I have a problem when I try to assign an ip address. > > > muxDevLoad(), muxBind(), ipAttach() and ifMaskSet() work fine. > > > ifShow() shows me the settings of the secondary driver. > > > But calling ifAddrSet() with an unique ip address causes the network > > > functionality of vxworks to crash. > > > > > > Even "ping" does no more work with the primary end driver after > > > calling ifAddrSet(). > > > > > > Does anybody can give me a hint whats going wrong with my secondary > > > end driver? > > > > > > Thanks. --------------------------- Newsgroups: comp.os.vxworks Subject: [Q] ibm ppc 405gp / dec21x4x / cacheDmaMalloc / malloc Date: 23 Aug 2002 18:44:38 -0700 From: kotaeji@hanmail.net (Kim Tae Hyung) Organization: http://groups.google.com/ Message-ID: Hi. I have a target board based on ibm ppc 405gp, and it has dec21x4x ethernet chip. When dec21x4x driver initializes memories for descriptors and clusters, it uses cacheDmaMalloc. But for some reasons, I must replace cacheDmaMalloc with malloc(or memalign). But, it doesn't work just by replacing cacheDmaMalloc and considering memory alignment. So, I have some questions on this subject. 1. 405 manual says that "The DCU does not provide snooping. Application programs must carefully use cache-inhibited regions and cache control instructions to ensure proper operation of the cache in systems where external devices can update memory." Can't we have cache-coherent memory without cacheDmaMalloc if 405 does not provide snooping? If so, how can we get cache-coherent memory with cacheDmaMalloc? Just by software mechanism? 2. My related sysPhysMemDesc says: { (void *) RAM_LOW_ADRS, (void *) RAM_LOW_ADRS, LOCAL_MEM_SIZE - RAM_LOW_ADRS, VM_STATE_MASK_VALID | VM_STATE_MASK_WRITABLE | VM_STATE_MASK_CACHEABLE | M_STATE_MASK_MEM_COHERENCY, VM_STATE_VALID | VM_STATE_WRITABLE | VM_STATE_CACHEABLE | VM_STATE_MEM_COHERENCY }, If we set VM_STATE_MEM_COHERENCY in this memory space, does it guarantee the cache-coherency in this memory space? 3. The difference between cacheDmaMalloc and malloc. Except cache coherency. Thanks in advance. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: examples Date: Sat, 24 Aug 2002 09:26:14 -0400 From: "Cedric Le Roux" Message-ID: <3d678931_1@nopics.sjc> References: <3D660D82.CB1C0EA0@bellsouth.net> http://www.rt.db.erau.edu/ go Eagles "joe durusau" wrote in message news:3D660D82.CB1C0EA0@bellsouth.net... > > > Lorenzo Salvemini wrote: > > > Hi everybody, > > i'm a new Tornado 2 VxWorks user, does anybody know where I can find > > some execrises or example codes??? > > Thanks a lot > > bye > > Not T-II specific, but you could check Embry-Riddle's website. > They have a bunch of examples of vxWorks stuff there. > > Speaking only for myself, > > Joe Durusau > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: examples Date: Sat, 24 Aug 2002 09:26:14 -0400 From: "Cedric Le Roux" Message-ID: <3d678b2c$1_2@nopics.sjc> References: <3D660D82.CB1C0EA0@bellsouth.net> http://www.rt.db.erau.edu/ go Eagles "joe durusau" wrote in message news:3D660D82.CB1C0EA0@bellsouth.net... > > > Lorenzo Salvemini wrote: > > > Hi everybody, > > i'm a new Tornado 2 VxWorks user, does anybody know where I can find > > some execrises or example codes??? > > Thanks a lot > > bye > > Not T-II specific, but you could check Embry-Riddle's website. > They have a bunch of examples of vxWorks stuff there. > > Speaking only for myself, > > Joe Durusau > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: examples Date: Sat, 24 Aug 2002 09:26:14 -0400 From: "Cedric Le Roux" Message-ID: <3d67a56c$1_1@nopics.sjc> References: <3D660D82.CB1C0EA0@bellsouth.net> http://www.rt.db.erau.edu/ go Eagles "joe durusau" wrote in message news:3D660D82.CB1C0EA0@bellsouth.net... > > > Lorenzo Salvemini wrote: > > > Hi everybody, > > i'm a new Tornado 2 VxWorks user, does anybody know where I can find > > some execrises or example codes??? > > Thanks a lot > > bye > > Not T-II specific, but you could check Embry-Riddle's website. > They have a bunch of examples of vxWorks stuff there. > > Speaking only for myself, > > Joe Durusau > > > --------------------------- Newsgroups: alt.os.linux.suse,comp.os.linux,comp.os.linux.advocacy,comp.os.linux.embedded,comp.os.vxworks Subject: Re: newsgroup spreading virus Date: Sun, 25 Aug 2002 13:09:16 GMT From: Bill Baker Organization: EarthLink Inc. -- http://www.EarthLink.net Message-ID: References: On Sat, 24 Aug 2002 19:43:47 -0400, Russ wrote: > Also today I spent 1-1/2 hours trying to get through to a live person at > earthlink. I did eventually reach one Brian (an expert on viruses who > can't remember his employee number) who has a boss named Matt who is too > busy to talk to customers. So don't expect earthlink to do anything > soon. > Pardon me, but I have worked in a technical support call center before, and can clear some things up for you. Call centers are typically outsourced to keep costs low, so this Brian you spoke to probably doesn't even work for Earthlink directly. His supervisor, Matt, more than likely does have a lot of responsibilities, which is why he can't come to the phone, especially to answer a question that Brian is fully capable of answering himself. I suspect they told you that they would have a supervisor call you back when one was available. Besides, in these situations, Brian probably knows more than Matt on the subject, so anything that Matt says is likely to be exactly the same thing that Brian said. Also, why would they do anything about it? The From: address is probably forged, anyway, and the sender probably isn't even an Earthlink subscriber. If you want to complain to the correct ISP, run the e-mail through SpamCop (http://spamcop.net, free registration required) and find out where it really came from. - --Bill --------------------------- Newsgroups: comp.os.vxworks Subject: VxWorks 5.1 and tar Date: Sun, 25 Aug 2002 14:23:54 GMT From: "Don Peterson" Message-ID: <_L5a9.24539$WJ3.3979429@news1.news.adelphia.net> Does anyone have and updated version tar that will run with v5.1. The version I have is tarzcb.o and it does not support compression. I running on mvme-167. Any help is appreciated... V/R DP --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Use Parallel Port instead of the required Serial Port ? Need Null Modem line ? Date: 25 Aug 2002 07:30:12 -0700 From: patrickyu_ca@yahoo.com (patrick) Organization: http://groups.google.com/ Message-ID: References: <488e459a.0208240913.756150a2@posting.google.com> Hi John, Thanks for your great help first. I want to design a driver, which need the system mode(INT). And I will have a 25 pin to 9 pin changer. The USB serial port adapter is a good way if I can't use the changer. Just like you said, now it is difficult to find. If the changer can work, in this connection, do I need a NULL MODEM changer ? (I do not have the NULL MODEM changer now.) I am not sure whether the serial connection need the NULL Modem line or not? And if this can work, Is there any configuration I should take care? Patrick --------------------------- Newsgroups: comp.os.vxworks Subject: reading registers from the ixf440 using the SA on ixp1200eb Date: Sun, 25 Aug 2002 17:40:05 +0200 From: "tomer" Organization: The Israel Inter-University Computation Center Message-ID: hi i am working on the ixp1200eb supported by vxworks os. i could not find a way to read the statistics registers on the ixf 440 by the strongArm, i belive there is function that does it but i couldn't find it. any information will be helpfull. thanks tomer. --------------------------- Newsgroups: alt.os.linux.suse,comp.os.linux,comp.os.linux.advocacy,comp.os.linux.embedded,comp.os.vxworks Subject: Re: newsgroup spreading virus Date: Sun, 25 Aug 2002 16:25:26 GMT From: Russ Organization: EarthLink Inc. -- http://www.EarthLink.net Message-ID: References: Followup-To: alt.os.linux.suse Bill Baker wrote: > On Sat, 24 Aug 2002 19:43:47 -0400, Russ wrote: > >> Also today I spent 1-1/2 hours trying to get through to a live person at >> earthlink. I did eventually reach one Brian (an expert on viruses who >> can't remember his employee number) who has a boss named Matt who is too >> busy to talk to customers. So don't expect earthlink to do anything >> soon. >> > > Pardon me, but I have worked in a technical support call center before, > and can clear some things up for you. >Call centers are typically > outsourced to keep costs low, so this Brian you spoke to probably doesn't > even work for Earthlink directly. Not my problem. >His supervisor, Matt, more than likely > does have a lot of responsibilities, which is why he can't come to the > phone, especially to answer a question that Brian is fully capable of > answering himself. Brian couldn't answer the question : put me in touch with someone at earthlink who. >I suspect they told you that they would have a > supervisor call you back when one was available. Besides, in these > situations, Brian probably knows more than Matt on the subject, so > anything that Matt says is likely to be exactly the same thing that Brian > said. Brian:"what operating system are you running" Me: "Irrelevant, I'm trying to report a problem earthlink is having spreading viruses" Brian: "I can't help you if you don't tell me your operating system" Me: "Linux, but the problem is that earthlink is spreading viruses" Brian: "we don't support Linux" Me: (very tired after spending 1-1/2 hours on this) "Look, I'm getting lots of emails originating from earthlink addresses containing the Klez virus" Brian: "Have you rebooted your system?" Me: "Not for the last year or so. I'm just trying to help earthlink to stop the transmission of viruses from your domain. Can you connect me with someone who can work on this." Brian: "I know a lot about viruses. Let me explain how they work. You have to reboot your machine.." Me: "Let me talk to your supervisor or someone else who can connect me to an earthlink technical person." Brian: "I know a lot about viruses. Send a report to abuse@earthlink.net." Me: "I did two weeks ago, but I haven't gotten a response. This bounced mail is annoying and causing me problems. Will you get your supervisor?" Brian goes away for a few minutes and comes back. Brian: "he's busy. leave your number and he will call back in an hour or two." Me: "I won't be here in an hour or two. I've wasted too much of my time already." > > Also, why would they do anything about it? The From: address is probably > forged, anyway, and the sender probably isn't even an Earthlink > subscriber. If you want to complain to the correct ISP, run the e-mail > through SpamCop (http://spamcop.net, free registration required) and find > out where it really came from. > earthlink.net Let me explain this slowly so a former help desk employee can understand : I'm getting bounced email The email has my address in the "From". The email contains attachments, including one of several jpegs, and the klez virus. None of the attachments are on my system anywhere either plain or encrypted or otherwise hidden. The code attached will not run on my system. It requires IE. I never sent the email. Therefore, the From address was forged. The real origin is one of several earthlink accounts. Therefore, it is an earthlink problem. It is an especial big problem, because the attack targets US government installations. I've spent all the time I have allocated for earthlink on this. If it doesn't stop NOW, my next stop is the FBI and DIS. DIS will not be amused. But they are probably already tracking these messages anyway.Some of the bounces took a strange route back. > --Bill - -- Russ Lyttle - Not powered by ActiveX Please visit for OBD-II information and links. --------------------------- Newsgroups: alt.os.linux.suse,comp.os.linux,comp.os.linux.advocacy,comp.os.linux.embedded,comp.os.vxworks Subject: Re: newsgroup spreading virus Date: Sun, 25 Aug 2002 19:00:16 GMT From: francesca Organization: Cox Communications Message-ID: <3D692941.10700@cox.net> References: Bill Baker wrote: > On Sat, 24 Aug 2002 19:43:47 -0400, Russ wrote: > > >>Also today I spent 1-1/2 hours trying to get through to a live person at >>earthlink. I did eventually reach one Brian (an expert on viruses who >>can't remember his employee number) who has a boss named Matt who is too >>busy to talk to customers. So don't expect earthlink to do anything >>soon. >> >> > > Pardon me, but I have worked in a technical support call center before, > and can clear some things up for you. Call centers are typically > outsourced to keep costs low, so this Brian you spoke to probably doesn't > even work for Earthlink directly. His supervisor, Matt, more than likely > does have a lot of responsibilities, which is why he can't come to the > phone, especially to answer a question that Brian is fully capable of > answering himself. I suspect they told you that they would have a > supervisor call you back when one was available. Besides, in these > situations, Brian probably knows more than Matt on the subject, so > anything that Matt says is likely to be exactly the same thing that Brian > said. > Sad to say, but Bill is right, to a degree. True customer service is just about dead today. However, from my experience, if you can manage to get a supervisor on the phone, they are more likely to go the extra effort to make the customer happy. But that seems to be changing to, A few months ago I paid Sprint $5 for an account I had no access to. I did't know I had no access to it until MCI sent me a bill, after the Sprint bill was paid, for access I did have. They initially didn't want to give me credit for the service I didn't have access to because I was canceling the account and therefore would not be able to use the credit. They don't send out refund checks, they said. After going back and forth about this, the girl on the phone finally got off the phone to talk to her supervisor who told her to give the credit, though the question of getting a refund check would be sent to "reseach" to decide. --------------------------- Newsgroups: alt.os.linux.suse,comp.os.linux,comp.os.linux.advocacy,comp.os.linux.embedded,comp.os.vxworks Subject: Re: newsgroup spreading virus Date: Sun, 25 Aug 2002 20:55:52 GMT From: Russ Organization: EarthLink Inc. -- http://www.EarthLink.net Message-ID: References: <3D692941.10700@cox.net> Followup-To: alt.os.linux.suse francesca wrote: > > > Bill Baker wrote: > >> On Sat, 24 Aug 2002 19:43:47 -0400, Russ wrote: >> >> >>>Also today I spent 1-1/2 hours trying to get through to a live person at >>>earthlink. I did eventually reach one Brian (an expert on viruses who >>>can't remember his employee number) who has a boss named Matt who is too >>>busy to talk to customers. So don't expect earthlink to do anything >>>soon. >>> >>> >> >> Pardon me, but I have worked in a technical support call center before, >> and can clear some things up for you. Call centers are typically >> outsourced to keep costs low, so this Brian you spoke to probably doesn't >> even work for Earthlink directly. His supervisor, Matt, more than likely >> does have a lot of responsibilities, which is why he can't come to the >> phone, especially to answer a question that Brian is fully capable of >> answering himself. I suspect they told you that they would have a >> supervisor call you back when one was available. Besides, in these >> situations, Brian probably knows more than Matt on the subject, so >> anything that Matt says is likely to be exactly the same thing that Brian >> said. >> > > > Sad to say, but Bill is right, to a degree. True customer service is > just about dead today. However, from my experience, if you can manage > to get a supervisor on the phone, they are more likely to go the extra > effort to make the customer happy. But that seems to be changing to, A > few months ago I paid Sprint $5 for an account I had no access to. I > did't know I had no access to it until MCI sent me a bill, after the > Sprint bill was paid, for access I did have. They initially didn't want > to give me credit for the service I didn't have access to because I was > canceling the account and therefore would not be able to use the credit. > They don't send out refund checks, they said. After going back and > forth about this, the girl on the phone finally got off the phone to > talk to her supervisor who told her to give the credit, though the > question of getting a refund check would be sent to "reseach" to decide. Bill is absolutely right about how "customer service" is handled. As this is a security issue, I thought that it would be nice to try to work with earthlink to resolve it. Failing that, I posted here and elsewhere to help out those who might get hurt by earthlinks neglect. There was a time when I would argue that those who found a security problem should notify the company responsible and give them a chance to fix it before going public. After my big waste of time this weekend, I now wonder why anyone should bother. Linux people have a good rep for responding so I'll work with them. But as for MS, Earthlink, AOL, etc., I think I'll just go ahead and post. I just don't have the time and energy to fight with the idiots they put on help desks. - -- Russ Lyttle - Not powered by ActiveX Please visit for OBD-II information and links. --------------------------- Newsgroups: alt.os.linux.suse,comp.os.linux,comp.os.linux.advocacy,comp.os.linux.embedded,comp.os.vxworks Subject: Re: newsgroup spreading virus Date: Sun, 25 Aug 2002 22:19:13 GMT From: Bill Baker Organization: EarthLink Inc. -- http://www.EarthLink.net Message-ID: References: On Sun, 25 Aug 2002 12:25:26 -0400, Russ wrote: > Bill Baker wrote: > >> On Sat, 24 Aug 2002 19:43:47 -0400, Russ wrote: >> >>> Also today I spent 1-1/2 hours trying to get through to a live person >>> at earthlink. I did eventually reach one Brian (an expert on viruses >>> who can't remember his employee number) who has a boss named Matt who >>> is too busy to talk to customers. So don't expect earthlink to do >>> anything soon. >>> >>> >> Pardon me, but I have worked in a technical support call center before, >> and can clear some things up for you. > >>Call centers are typically >> outsourced to keep costs low, so this Brian you spoke to probably >> doesn't even work for Earthlink directly. > Not my problem. I'm just explaining how it works. >>His supervisor, Matt, more than likely >> does have a lot of responsibilities, which is why he can't come to the >> phone, especially to answer a question that Brian is fully capable of >> answering himself. > Brian couldn't answer the question : put me in touch with someone at > earthlink who. It's a good bet that you'll never find anyone, as these are only low-level employees who probably don't even work directly for Earthlink. >>I suspect they told you that they would have a >> supervisor call you back when one was available. Besides, in these >> situations, Brian probably knows more than Matt on the subject, so >> anything that Matt says is likely to be exactly the same thing that >> Brian said. > Brian:"what operating system are you running" Me: "Irrelevant, I'm > trying to report a problem earthlink is having spreading viruses" Brian: > "I can't help you if you don't tell me your operating system" Me: > "Linux, but the problem is that earthlink is spreading viruses" Brian: > "we don't support Linux" > Me: (very tired after spending 1-1/2 hours on this) "Look, I'm getting > lots of emails originating from earthlink addresses containing the Klez > virus" Brian: "Have you rebooted your system?" Me: "Not for the last > year or so. I'm just trying to help earthlink to stop the transmission > of viruses from your domain. Can you connect me with someone who can > work on this." > Brian: "I know a lot about viruses. Let me explain how they work. You > have to reboot your machine.." > Me: "Let me talk to your supervisor or someone else who can connect me > to an earthlink technical person." > Brian: "I know a lot about viruses. Send a report to > abuse@earthlink.net." Me: "I did two weeks ago, but I haven't gotten a > response. This bounced mail is annoying and causing me problems. Will > you get your supervisor?" Brian goes away for a few minutes and comes > back. Brian: "he's busy. leave your number and he will call back in an > hour or two." > Me: "I won't be here in an hour or two. I've wasted too much of my time > already." Sounds like Brian was reading from a script, which is common in many technical support call centers. True, his answers didn't help you, but I'm willing to bet his supervisor's answers wouldn't have helped you, either. >> Also, why would they do anything about it? The From: address is >> probably forged, anyway, and the sender probably isn't even an >> Earthlink subscriber. If you want to complain to the correct ISP, run >> the e-mail through SpamCop (http://spamcop.net, free registration >> required) and find out where it really came from. >> >> > earthlink.net > Let me explain this slowly so a former help desk employee can understand Don't patronize me, jackass. > : I'm getting bounced email > The email has my address in the "From". The email contains attachments, > including one of several jpegs, and the klez virus. None of the > attachments are on my system anywhere either plain or encrypted or > otherwise hidden. > The code attached will not run on my system. It requires IE. I never > sent the email. > Therefore, the From address was forged. The real origin is one of > several earthlink accounts. Therefore, it is an earthlink problem. It is > an especial big problem, because the attack targets US government > installations. > I've spent all the time I have allocated for earthlink on this. If it > doesn't stop NOW, my next stop is the FBI and DIS. DIS will not be > amused. But they are probably already tracking these messages > anyway.Some of the bounces took a strange route back. Sounds like that will be your best bet to getting this stopped. If it is attacking US government computers, then it was probably a mistake to try to talk to technical support, as their only course of action is to tell you to send the message to abuse@earthlink.net. And the abuse people are probably so busy, they haven't even had time to look at it yet. - --Bill --------------------------- Newsgroups: alt.os.linux.suse,comp.os.linux,comp.os.linux.advocacy,comp.os.linux.embedded,comp.os.vxworks Subject: Re: newsgroup spreading virus Date: Mon, 26 Aug 2002 00:44:46 GMT From: Russ Organization: EarthLink Inc. -- http://www.EarthLink.net Message-ID: <2Sea9.7547$ob2.665771@newsread1.prod.itd.earthlink.net> References: Followup-To: alt.os.linux.suse Bill Baker wrote: > On Sun, 25 Aug 2002 12:25:26 -0400, Russ wrote: > >> Bill Baker wrote: >> >>> On Sat, 24 Aug 2002 19:43:47 -0400, Russ wrote: >>> >>>> Also today I spent 1-1/2 hours trying to get through to a live person >>>> at earthlink. I did eventually reach one Brian (an expert on viruses >>>> who can't remember his employee number) who has a boss named Matt who >>>> is too busy to talk to customers. So don't expect earthlink to do >>>> anything soon. >>>> >>>> >>> Pardon me, but I have worked in a technical support call center before, >>> and can clear some things up for you. >> >>>Call centers are typically >>> outsourced to keep costs low, so this Brian you spoke to probably >>> doesn't even work for Earthlink directly. >> Not my problem. > > I'm just explaining how it works. > >>>His supervisor, Matt, more than likely >>> does have a lot of responsibilities, which is why he can't come to the >>> phone, especially to answer a question that Brian is fully capable of >>> answering himself. >> Brian couldn't answer the question : put me in touch with someone at >> earthlink who. > > It's a good bet that you'll never find anyone, as these are only low-level > employees who probably don't even work directly for Earthlink. > >>>I suspect they told you that they would have a >>> supervisor call you back when one was available. Besides, in these >>> situations, Brian probably knows more than Matt on the subject, so >>> anything that Matt says is likely to be exactly the same thing that >>> Brian said. >> Brian:"what operating system are you running" Me: "Irrelevant, I'm >> trying to report a problem earthlink is having spreading viruses" Brian: >> "I can't help you if you don't tell me your operating system" Me: >> "Linux, but the problem is that earthlink is spreading viruses" Brian: >> "we don't support Linux" >> Me: (very tired after spending 1-1/2 hours on this) "Look, I'm getting >> lots of emails originating from earthlink addresses containing the Klez >> virus" Brian: "Have you rebooted your system?" Me: "Not for the last >> year or so. I'm just trying to help earthlink to stop the transmission >> of viruses from your domain. Can you connect me with someone who can >> work on this." >> Brian: "I know a lot about viruses. Let me explain how they work. You >> have to reboot your machine.." >> Me: "Let me talk to your supervisor or someone else who can connect me >> to an earthlink technical person." >> Brian: "I know a lot about viruses. Send a report to >> abuse@earthlink.net." Me: "I did two weeks ago, but I haven't gotten a >> response. This bounced mail is annoying and causing me problems. Will >> you get your supervisor?" Brian goes away for a few minutes and comes >> back. Brian: "he's busy. leave your number and he will call back in an >> hour or two." >> Me: "I won't be here in an hour or two. I've wasted too much of my time >> already." > > Sounds like Brian was reading from a script, which is common in many > technical support call centers. True, his answers didn't help you, but > I'm willing to bet his supervisor's answers wouldn't have helped you, > either. > > >>> Also, why would they do anything about it? The From: address is >>> probably forged, anyway, and the sender probably isn't even an >>> Earthlink subscriber. If you want to complain to the correct ISP, run >>> the e-mail through SpamCop (http://spamcop.net, free registration >>> required) and find out where it really came from. >>> >>> >> earthlink.net >> Let me explain this slowly so a former help desk employee can understand > > Don't patronize me, jackass. Well, dumbshit, you don't seem to understand the plain english that said THE MESSAGES WERE ORIGINATING FROM EARTHLINK ADDRESSES. > >> : I'm getting bounced email >> The email has my address in the "From". The email contains attachments, >> including one of several jpegs, and the klez virus. None of the >> attachments are on my system anywhere either plain or encrypted or >> otherwise hidden. >> The code attached will not run on my system. It requires IE. I never >> sent the email. >> Therefore, the From address was forged. The real origin is one of >> several earthlink accounts. Therefore, it is an earthlink problem. It is >> an especial big problem, because the attack targets US government >> installations. >> I've spent all the time I have allocated for earthlink on this. If it >> doesn't stop NOW, my next stop is the FBI and DIS. DIS will not be >> amused. But they are probably already tracking these messages >> anyway.Some of the bounces took a strange route back. > > Sounds like that will be your best bet to getting this stopped. If it is > attacking US government computers, then it was probably a mistake to try > to talk to technical support, as their only course of action is to tell > you to send the message to abuse@earthlink.net. And the abuse people are > probably so busy, they haven't even had time to look at it yet. > > --Bill Try reading the message : the answers that Brian gave and the answers that I suspect the supervisor would give didn't/wouldn't help earthlink. I was trying to be a nice guy and help earthlink stop spreading the virus.But earthlink doesn't care. I hope all here on c.o.l.a. heed my advice and don't open any postings containing images. I think this particular script kiddie got a kick out of making a bunch of computer wizes think they were going to see some nekkid wimmen. I wonder how you managed to miss getting caught. - -- Russ Lyttle - Not powered by ActiveX Please visit for OBD-II information and links. --------------------------- Newsgroups: alt.os.linux.suse,comp.os.linux,comp.os.linux.advocacy,comp.os.linux.embedded,comp.os.vxworks Subject: Re: newsgroup spreading virus Date: Mon, 26 Aug 2002 01:22:15 GMT From: Bill Baker Organization: EarthLink Inc. -- http://www.EarthLink.net Message-ID: References: <2Sea9.7547$ob2.665771@newsread1.prod.itd.earthlink.net> On Sun, 25 Aug 2002 20:44:46 -0400, Russ wrote: > Bill Baker wrote: >>> earthlink.net >>> Let me explain this slowly so a former help desk employee can >>> understand >> >> Don't patronize me, jackass. > Well, dumbshit, you don't seem to understand the plain english that said > THE MESSAGES WERE ORIGINATING FROM EARTHLINK ADDRESSES. Well YOU didn't specify whether you were only looking at the From: line or at the Received: line. By the way, nice way to ask for help, by insulting the people you intend to ask help from. Try being civil next time. >>> : I'm getting bounced email >>> The email has my address in the "From". The email contains >>> attachments, including one of several jpegs, and the klez virus. None >>> of the attachments are on my system anywhere either plain or encrypted >>> or otherwise hidden. >>> The code attached will not run on my system. It requires IE. I never >>> sent the email. >>> Therefore, the From address was forged. The real origin is one of >>> several earthlink accounts. Therefore, it is an earthlink problem. It >>> is an especial big problem, because the attack targets US government >>> installations. >>> I've spent all the time I have allocated for earthlink on this. If it >>> doesn't stop NOW, my next stop is the FBI and DIS. DIS will not be >>> amused. But they are probably already tracking these messages >>> anyway.Some of the bounces took a strange route back. >> >> Sounds like that will be your best bet to getting this stopped. If it >> is attacking US government computers, then it was probably a mistake to >> try to talk to technical support, as their only course of action is to >> tell you to send the message to abuse@earthlink.net. And the abuse >> people are probably so busy, they haven't even had time to look at it >> yet. >> >> --Bill > > Try reading the message : the answers that Brian gave and the answers > that I suspect the supervisor would give didn't/wouldn't help earthlink. > I was trying to be a nice guy and help earthlink stop spreading the > virus.But earthlink doesn't care. I hope all here on c.o.l.a. heed my > advice and don't open any postings containing images. I think this > particular script kiddie got a kick out of making a bunch of computer > wizes think they were going to see some nekkid wimmen. > > I wonder how you managed to miss getting caught. I haven't seen any messages like the kind you describe. - --Bill --------------------------- Newsgroups: comp.os.vxworks Subject: Floating point on MBX860 Date: Sun, 25 Aug 2002 22:18:57 -0700 From: "Sandeep P" Organization: Arizona State University Message-ID: Hello, I am new to the MBX860. I am running vxworks on it. Is anybody aware of a solution to perform floating point calculations. Obviously, it would imply that I can have floating point variables. Bye, Sandeep. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: no carrier? Date: Mon, 26 Aug 2002 08:38:05 +0200 From: Marc Waber Organization: STS Switching Test Solutions Message-ID: <3D69CCCD.6F3D50CF@stest.ch> References: <1d168114.0208221817.842e3b2@posting.google.com> joge wrote: > > Hi all, > > Can anybody tell me why the "interrupt 0:dc0 no carrier!" emerages > when the bootrom start up on mpcn750? > I do not change any value in dec21x40end.h,DEC_USR_FD and > DEC_USR_PHY_CHK do not change. > > Thanks advanced. > > Joge You seem to configure your dc interface without connecting an ethernet cable. Quick workaround is to connect an ethernet cable, but you should get a new driver for your dc interface. Contact windriver, and they will send you an update, which includes also a lot of other improvements (eg 100BaseT capability). hth, Marc - -- --------------------------- Newsgroups: comp.os.vxworks Subject: How Change a TTL Value of ICMP Packet ? Date: Mon, 26 Aug 2002 16:24:11 +0900 From: "Jay" Organization: Inet Internet Services Message-ID: Hi, Everyone. I Try to make a route trace program on a vxWorks, but I can't change the TTL value of ICMP Packet. My test code is, int optval; icmpSock = socket(AF_INET, SOCK_RAW, IPPROTO_ICMP); optval = 1; setsockopt (icmpSock, IPPROTO_IP, IP_TTL, &optval, sizeof(optval)); Does anyone know how I can change a TTL Value? Thanks in advance! Jay. --------------------------- Newsgroups: comp.os.vxworks Subject: How to include macros inside TORNADOII for vxworks Date: 26 Aug 2002 00:30:12 -0700 From: vxnerd@rediffmail.com (subbarayan) Organization: http://groups.google.com/ Message-ID: dear mentors, Kindly let me know how to include macros inside vxworks using tornadoII.I am asking this becos I was adviced by the wind technical Notes to include macro INCLUDE_PCI macro inside vxworks to enable pci support.Further should i include the said macro by simply typing it in the configuration file or is there some other way from tornadoII environment to include the macros needed? I also checked with the project find object utility to find whether thers a macro called INCLUDE_PCI.But i was unable to find it and further i found only INCLUDE_PCI_PARAMS and INCLUDE_PCMCIA macros but not INCLUDE_PCI.Kindly can anyone tell me how to locate this macro or how to include it inside my appln using the tornadoII IDE? Further WTN49 states that i need to define a value for macro PCI_CFG_TYPE to enable vxworks to support pci operations.How to define a value for this said macro?should i directly go to pc.h and do direct typing the value there near the macro or is there a way in environment to define the value inside the macro? I am now assigned a job of developing device driver for pci cards from vxworks using tornadoII for x86 target with normal pc architecture.can anyone tell me where to start with and how to actually use vxworks supplied files like pciautoconfiglib.c,.h and pciconfig show files?sorry if my query looks very basic i don't know these things since i am new with vxworks. Advanced thanks for persons replying me. Any one with a reply can mail me to vxnerd@rediffmail.com or reply in this board itself. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: trouble in ping with a larger packet on ne2000 Ethernet Date: Mon, 26 Aug 2002 09:34:59 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <7f29c2ad.0208080439.62e501b7@posting.google.com> <7f29c2ad.0208091812.5eadfa30@posting.google.com> I agree that MTU has to do with it, but in no way should the MTU be a limitation. It is up to the IP protocol to perform Fragmentation at the sending host and Reassembly of parger packets at the receiving host. Failure of Ping with packets that require freagmentation is a well known symptom of buggy ethernet drivers. The fragments often arrive at very short inter-packet intervals, which are within spec, but the driver is incapable of recovering from a previous packet fast enough to receive the net packet. Similar symptom is caused by certain half-baked Ethernet switches, that do not support back pressure and are used with non-matching link speeds, e.g. of two hosts communicating, one is at 10Mbps while the otehr is at 100Mbps. Whatever is the source of this problem, it will definitely cause problems with different application protocols as well, hence it is worth to solve. "cfk" wrote in message news:thx59.1066$8_7.44448036@newssvr14.news.prodigy.com... > Let me take a wild stab at this one and suggest that the size of the ping > packet needs to be less then the smallest MTU between the host you are > pinging from and the host you are pinging to. I believe MTU is usually 1500 > on an ethernet link. > > > "rover" wrote in message > news:7f29c2ad.0208091812.5eadfa30@posting.google.com... > > rover2@163.com (rover) wrote in message > news:<7f29c2ad.0208080439.62e501b7@posting.google.com>... > > > Hi,all > > > > > > I have a 486(pc104) with a ne2000 ethernet card,use vxworks 5.4. > > > > > > when I ping it with a litter packet(<1500 bytes),it is ok,but when the > > > packet is larger(>1500),ping is failed! why? > > > > > > blow is: > > > > > > ping 192.168.144.1 -t -l 1500 > > > > > > Pinging 192.168.144.1 with 1500 bytes of data: > > > > > > Request timed out. > > > Request timed out. > > > Request timed out. > > > Request timed out. > > > > > > ping 192.168.144.1 -t -l 1472 > > > > > > Pinging 192.168.144.1 with 1472 bytes of data: > > > > > > Reply from 192.168.144.1: bytes=1472 time<10ms TTL=64 > > > Reply from 192.168.144.1: bytes=1472 time<10ms TTL=64 > > > Reply from 192.168.144.1: bytes=1472 time<10ms TTL=64 > > > > > > regards > > > > who can help me? > > > > thank very much. > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting Tornado 2.0 Date: 26 Aug 2002 00:54:07 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0208252354.34b46031@posting.google.com> References: Ahlawi Hasan wrote in message news:... > Is it possible to port Tornado 2.0 working on a > NT-station to a station with Win XP professional? rumours say tornado 2 (or tornado 2.0.2) works on windows xp. since there is no license managing with this version, you need not care about this (unless you're using diab). --------------------------- Newsgroups: comp.os.vxworks Subject: Re: elf/ coff/ Motorola S-record/Intel Hex/Bin ? Date: 26 Aug 2002 01:18:10 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0208260018.4993d07d@posting.google.com> References: xiaohua.zhang@163.com (xiaohua) wrote in message news:... > Could anybody tell me how to use these format in VxWorks envioroments? well... the executable format is toolchain specific... e.g. T2 PPC uses elf T2 x86 uses a.out the rest are derived formats, you can use tools like elfToBin or elfHex to get them... (or more sophisticated ones, like the convert utility from sds to get motorola s-record format) toni --------------------------- Newsgroups: comp.os.vxworks Subject: Soft Interrupts and getch() in VxWorks Date: 26 Aug 2002 02:25:28 -0700 From: ajay_garg@delhi.tcs.co.in (Ajay Garg) Organization: http://groups.google.com/ Message-ID: Hi, Do we have any soft interrups available in VxWorks. Actually I am looking for a service similar to int86() of DOS. If there is any other way out, please let me know. One more point, there is no support for getch() kind of thing too though getchar() is there. But getchar() does not display the character immediately on the screen, it requires a return key. So how I can implement this functionality. Any help will be greatly appreciated. Regards, Ajay Garg. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Soft Interrupts and getch() in VxWorks Date: Mon, 26 Aug 2002 06:24:38 -0400 From: joe durusau Organization: WEBUSENET.com Message-ID: <3D6A01E6.E9691603@bellsouth.net> References: Ajay Garg wrote: > Hi, > Do we have any soft interrups available in VxWorks. Actually I am > looking for a service similar to int86() of DOS. If there is any other > way out, please let me know. > One more point, there is no support for getch() kind of thing too > though getchar() is there. But getchar() does not display the > character immediately on the screen, it requires a return key. So how > I can implement this functionality. > > Any help will be greatly appreciated. > > Regards, > Ajay Garg. You can gen. interrupts if your BSP supports them. For example, if you are in a VME box, the function sysBusIntGen() can cause any int you wish. As to how to get single characters, it would depend on the source. From a serial port, you sould be able to use select() and write the chars back out. There might also be some ioctl magic required to get the thing into raw mode. If the data is coming over some other interface, it might get to be more trouble. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Why msgQSend can be called by ISR? Date: 15 Aug 2002 00:19:19 -0700 From: hotJacky@163.net (Bryan) Organization: http://groups.google.com/ Message-ID: <8f77d4a.0208142319.47dc32dd@posting.google.com> References: <8f77d4a.0208121740.796881b1@posting.google.com> <2025ac82.0208140116.13880440@posting.google.com> Claudio Potenza Leonid Rosenboim The key is that how realize the same function of the msgQSend in VxWorks in my own_designed way.Leonid Rosenboim had give some suggestion.I think the microcall is just like the inlock/unlock which Claudio Potenza mentioned.So I believed that it must exits inlock/unlock in msgQSend.In fact, msgQSend is called by excJobAdd,which I find by disassembling.So,excJobAdd is the same as msgQSend essentially except that some additional mechanism.More step,excJobAdd use msgQSend to notice another task.It is up the task to receive message but not the ISR. Waiting for your reply! Sincerely, Bryan --------------------------- Newsgroups: comp.os.vxworks Subject: Source of mips toolchain Date: 26 Aug 2002 06:39:06 -0700 From: chs@lucent.com (ChengHsin Hsu) Organization: http://groups.google.com/ Message-ID: <8b53eb96.0208260539.7ac17e6e@posting.google.com> All, Has anyone had the experience to get a copy of toolchain source code from wrs? I have sent several emails via their supports but nobody replies at all....I'm wondering whether this is possible..And interested in how can I do that... Following is my cross gcc's version and configuration. We bought this binary compiler from wrs, but wanna move to Linux as our development environment.... bear%ccmips -v gcc driver version cygnus-2.7.2-960126 egcs-971225 tornado 2.0 executing gcc version cygnus-2.7.2-960126 Thanks, Bear. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: no carrier? Date: 26 Aug 2002 07:34:23 -0700 From: richness@sina.com (richard) Organization: http://groups.google.com/ Message-ID: <465da109.0208260634.7818a56a@posting.google.com> References: <1d168114.0208221817.842e3b2@posting.google.com> joge709@hotmail.com (joge) wrote in message news:<1d168114.0208221817.842e3b2@posting.google.com>... > Hi all, > > Can anybody tell me why the "interrupt 0:dc0 no carrier!" emerages > when the bootrom start up on mpcn750? > I do not change any value in dec21x40end.h,DEC_USR_FD and > DEC_USR_PHY_CHK do not change. > > Thanks advanced. > > Joge I have ever got the similar problem. To solve this problem, I used a separate hub to connect the target and the host. Hope this can help. richard --------------------------- Newsgroups: comp.os.vxworks Subject: waitforever Date: 26 Aug 2002 07:44:35 -0700 From: senthil_rag@rediffmail.com (senthil) Organization: http://groups.google.com/ Message-ID: <39c3b976.0208260644.374f746f@posting.google.com> Hi, How are the statements below implemented by vxworks.. semTake(semid, WAITFOREVER); or semTake(semid,-1) -- both the statements waits on a semaphore forever. How is this implemented considering it in hardware point of view or kernel point of view. regards senthil --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Source of mips toolchain Date: Mon, 26 Aug 2002 14:52:08 +0000 (UTC) From: Johan Borkhuis Organization: Agere Systems Message-ID: References: <8b53eb96.0208260539.7ac17e6e@posting.google.com> chs@lucent.com (ChengHsin Hsu) wrote: > All, > > Has anyone had the experience to get a copy of toolchain source > code from wrs? I have sent several emails via their supports but > nobody replies at all....I'm wondering whether this is possible..And > interested in how can I do that... > Following is my cross gcc's version and configuration. We bought > this binary compiler from wrs, but wanna move to Linux as our > development environment.... Yes, no problem. With the 2.1 release it is included on the normal CD if I remember correctly, and for the 2.0 you can order the source cd from them for free. They have to provide it, as they are using GNU tools (see GPL). Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How to include macros inside TORNADOII for vxworks Date: Mon, 26 Aug 2002 15:12:08 GMT From: David Cooper Organization: none Message-ID: References: [This followup was posted to comp.os.vxworks and a copy was sent to the cited author.] If you are using Tornado's IDE, then you use the Project Facility, VxWorks Tab to enable PCI support. If you are configuring a BSP from command line then you edit Config.h in your BSP directory. ...david In article , vxnerd@rediffmail.com says... > dear mentors, > Kindly let me know how to include macros inside vxworks using > tornadoII.I am asking this becos I was adviced by the wind technical > Notes to include macro INCLUDE_PCI macro inside vxworks to enable pci > support.Further should i include the said macro by simply typing it in > the configuration file or is there some other way from tornadoII > environment to include the macros needed? > I also checked with the project find object utility to find whether > thers a macro called INCLUDE_PCI.But i was unable to find it and > further i found only INCLUDE_PCI_PARAMS and INCLUDE_PCMCIA macros but > not INCLUDE_PCI.Kindly can anyone tell me how to locate this macro or > how to include it inside my appln using the tornadoII IDE? > Further WTN49 states that i need to define a value for macro > PCI_CFG_TYPE to enable vxworks to support pci operations.How to > define a value for this said macro?should i directly go to pc.h and do > direct typing the value there near the macro or is there a way in > environment to define the value inside the macro? > I am now assigned a job of developing device driver for pci cards from > vxworks using tornadoII for x86 target with normal pc architecture.can > anyone tell me where to start with and how to actually use vxworks > supplied files like pciautoconfiglib.c,.h and pciconfig show > files?sorry if my query looks very basic i don't know these things > since i am new with vxworks. > Advanced thanks for persons replying me. > Any one with a reply can mail me to vxnerd@rediffmail.com or reply in > this board itself. > --------------------------- Newsgroups: comp.os.vxworks Subject: MVME5100-013x versus -016x Date: 26 Aug 2002 08:13:38 -0700 From: quicksilver@icelab.net (Andy Kaiser) Organization: http://groups.google.com/ Message-ID: I'm wondering if anyone out there is using Motorola's 5100-013x (PPC750, 64MB RAM) and 5100-016x (PPC750, 512MB RAM) with the same software. My problem is that the VxWorks BSP config I've produced for the 0163 works fine, and our application runs with no problems, but when I flash the same kernel (adjusting for the difference in memory, of course) on the 0133, VxWorks comes up fine, but the application crashes in some odd places. (For instance, on an "if (TRUE)" statement.) It's not like I can't run any applications on the 64MB board, since I've got some less intensive apps running on the board fine. I've tinkered, and the places it seems to crash are immediately following a burst of VME traffic. (This may be coincidence, since another app I've tried on the board actually programs 12MB of FPGAs over VME, and doesn't crash. Go figure.) And yes, there's enough memory on the 64MB to run the application, so that's not a problem. I've searched the Installation & Use guide, but I don't see any difference between the two boards other than on-board memory. If someone knows of a difference, other than memory complement, that needs to be adjusted within the BSP to run the 64MB (0133) board, I'd appreciate it if you could share! Thanks, Andy Kaiser L-3 Communications andy.kaiser@L-3Com.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: waitforever Date: Mon, 26 Aug 2002 17:19:20 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <39c3b976.0208260644.374f746f@posting.google.com> The "Wait forever" case is actually simplier then the case of pending on a semaphore with a specific timeout. When there is no timeout (i.e. wait forever), the task TCB will be put on the semaphore pendQ. When there is a timeout, the task will be put into semaphore's pendQ, and also on the system clock delayQ. For that reason, the TCB has a separate pointer-pair for pendQ, and a separate pointer-pair for delayQ. When the task becomes ready again, it will be kicked off both queus at the same time, and added to the readyQ, using the same pointers used for the pendQ. BTW, I used to teach this stuff, as part of the standard VxWorks workshop training. "senthil" wrote in message news:39c3b976.0208260644.374f746f@posting.google.com... > Hi, > How are the statements below implemented by vxworks.. > semTake(semid, WAITFOREVER); or semTake(semid,-1) -- both the > statements waits on a semaphore forever. How is this implemented > considering it in hardware point of view or kernel point of view. > regards > senthil --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting Tornado 2.0 Date: 26 Aug 2002 09:01:58 -0700 From: zenk78@hotmail.com (Robert Templeman) Organization: http://groups.google.com/ Message-ID: <32bae160.0208260801.1cbe073b@posting.google.com> References: straymail@itozi.com (Stray Mail for itozi.com) wrote in message news:... > Please note, this is an _automated reply_ > > The e-mail you sent was incorrectly addressed and could not be delivered. Please recheck the e-mail address. > > However, your e-mail will be reviewed and forwarded to the appropriate person where possible. > > IT @ OZI can be contacted on: > Sydney (02) 9955 5388 > International +61 2 9955 5388 >     I just got back from a Tornado Training. Tornado 2.0.x will NOT support XP, but....Tornado 2.2 was released TODAY and has full support for XP Professional as well as many other features. If you have maintainance on 2.0 you can upgrade to 2.2 for FREE. It is a pull release so you much contact Windriver for it. And it does have extensive license management so that must be resolved. Rob Templeman --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Soft Interrupts and getch() in VxWorks Date: Mon, 26 Aug 2002 10:25:16 -0700 From: "Michael R. Kesti" Organization: MK Associates Message-ID: <3D6A647C.44B6061@gv.net> References: Reply-To: mkesti@gv.net Ajay Garg wrote: >Hi, > Do we have any soft interrups available in VxWorks. Actually I am >looking for a service similar to int86() of DOS. If there is any other >way out, please let me know. Whether software interrupts are available for your target depends on whether your target processor supports them. > One more point, there is no support for getch() kind of thing too >though getchar() is there. But getchar() does not display the >character immediately on the screen, it requires a return key. So how >I can implement this functionality. You need to set the port from which you getchar() to raw mode. This is accomplished using the ioctl() call and its FIOSETOPTIONS and OPT_RAW args. >Any help will be greatly appreciated. You're welcome. - -- ======================================================================== Michael Kesti | "And like, one and one don't make | two, one and one make one." mkesti@gv.net | - The Who, Bargain --------------------------- Newsgroups: comp.os.vxworks Subject: flash boot for MV177 Date: Mon, 26 Aug 2002 14:21:31 -0400 From: Joe Durusau Organization: Lockheed Martin Corporation Message-ID: <3D6A71AB.730F634@lmco.com> I am setting up an MV177 board from Motorola. I have seen the method of enabling boot from flash previously posted by Eric Schott. However, I would like to be able to have the board run its internal diagnostics first, then boot automatically to vxWorks, and I was wondering whether anyone had gotton this to work. I need the diagnostics because this is a mission-critical system handled by personnel who will not be familiar with its internal workings, so pretesting of the processor board is strongly preferred. Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Boot from IDE Flash Disk with x86 target Date: Mon, 26 Aug 2002 22:21:23 GMT From: Andray Kaganovsky Organization: Futureway Message-ID: References: <5947a04.0208222243.76bbfe@posting.google.com> andreas_lindell75@hotmail.com (Rikard) wrote in news:5947a04.0208222243.76bbfe@posting.google.com: > Hello, > > I am trying to boot vxworks from an IDE Flash disk IFD-350-64 but to > now succes. I have done vxsys on the flash disk and then copied me > bootrom.sys to the flash disk and placed a vxworks image in the root > of the flash disk. To boot it I type @ and then enter. The error i get > is: > ""Attaching to TFFS... drive is out of range (0--1)" > "usrTffsConfig failed" > > Error loading file:errno = 0x0 > dosFsDevInit failed > usrFdConfig failed > > Now I have some questions: > 1. Does vxworks make a difference between a IDE flash disk and a IDE > hard disk? According to the manufacturer the IFD-350-64 is "full ide > hard disk compatible . Does that mean that the system/vxworks can not > tell if its using a ide hard disk or a ide flash disk? > > 2. If I want to boot from a ide flash disk how do I do that? I have > done the following: > > 2.1 Replaced the sysTffs.c in the bsp directory with the > doc2k-sysTffs.c (and renamed it). After that I added MATCH EXTRA = > sysTffs.o in the make file. > 2.2 In config.h I included: > #define INCLUDE_DOSFS /* include dosFs file system */ > #define INCLUDE_FD /* include floppy disk driver */ > #define INCLUDE_LPT /* include parallel port driver */ > #undef INCLUDE_ATA /* include IDE/EIDE(ATA) hard disk driver */ > #undef INCLUDE_TIMESTAMP /* include TIMESTAMP timer for Wind View */ > #define INCLUDE_TFFS /* include TrueFFS driver for Flash */ > #undef INCLUDE_PCMCIA /* include PCMCIA driver */ > > #ifdef INCLUDE_PCMCIA > #define INCLUDE_ATA /* include ATA driver */ > #define INCLUDE_SRAM /* include SRAM driver */ > #undef INCLUDE_TFFS /* include TFFS driver */ > #ifdef INCLUDE_NETWORK > #define INCLUDE_ELT /* include 3COM EtherLink III driver */ > #endif /* INCLUDE_NETWORK */ > #endif /* INCLUDE_PCMCIA */ > > #ifdef INCLUDE_TFFS > #define INCLUDE_SHOW_ROUTINES > #endif /* INCLUDE_TFFS */ > > and added the line: > #define DEFAULT_BOOT_LINE \ > "tffs=0,0(0,0)host:/tffs0/vxworks h=192.168.0.247 e=192.168.0.241 > u=vxworks pw=vxworks tn=target" > > Then I built the bootrom and copied it to the IDE flash disk under > then name bootrom.sys. > 2.3 Place a vxworks image in the flash root and turned my system on > and then I get these errors: > "Attaching to TFFS... drive is out of range (0--1) > usrTffsConfig failed > Error loading file:errno = 0x0 > dosFsDevInit failed > usrFdConfig failed" > > Any one have any clue to what causes these errors and how to fix them? > > 3. I have also tried to see if I can use a bootrom.sys that specifies > an IDE/ATA hard drive but connecting the ide flash disk just to try if > the system makes any difference. This bootrom.sys file works fine with > a Quantum IDE hard disk and when replacing the Quantum with the flash > disk the bootrom starts up and I do a @ and get the following error: > "Attaching to ATA disk device... Error during dosFsDevInit: 380001 > usrAtaConfig falied > > Error loading file: errno = 380001 > dosFsDevInit failed > usrFdConfig failed" > > Any idea what this can be? Of course if vxworks makes a big difference > between ide flash disk and ide hard disk I understand why this does > not work but since IFD-350-64 is compatible with hard disk, according > to the manufacturer this should work as well. This question brings us > back to question number 1. However any input in this question would be > appreciated. > > 4. I have one option left that I haven't tried out and the reason is > that this will take some time to do and I would like to have some > input on this. I have one Readme file form M-Systems: > * > * > * Installation Guide for DiskOnChip(R) Product Family on VxWorks(R) > * > * Operating System (x86 edition). > * > * Version 5.0, August 01 2001 > > * > * Copyright M-Systems (c) 2001 > * > *********************************************************************** > > > This file specifies on how to integrate DiskOnChip with vxworks. Is a > DiskOnChip the same as an IDE flash disk? This document is fairly long > and I may do it and then find out that it wont work:(:( The doucment > is divided into the following sections: > > CONTENTS > -------- > > 1. Introduction > 2. Installing DiskOnChip driver > 3. Booting VxWorks from DiskOnChip > 4. Driver's run-time configuration options > 5. Formatting DiskOnChip from VxWorks application > 6. Considerations related to unexpected power shutdowns > 7. Alternative "garbage collection" scenario > 8. Access to advanced features of DiskOnChip > 9. Known problems > > Appendix A. Re-formatting DiskOnChip under MS-DOS > Appendix B. Partitioning the disk with MS-DOS FDISK > Appendix C. Using WindRiver's VXSYS and VXCOPY utilities > > And my questions is as following. Do I need to do step > 2. Installing DiskOnChip driver > before I do: > 3. Booting VxWorks from DiskOnChip > I have only taken a quick glance at these steps and my first > impression is that step 2 is required if I intend to use the flash in > my application but that is not my intention. I just want to boot the > vxworks image from the flash not do any read or write from the > application. So if anyone has worked with this document maybe she/he > can inform me if step 2 is a "must do" before step 3? > > I am not including the entire document from M-Systems becasue it is > fairly big but anyone who has worked on flash disk may have seen it > and could answer my question. > > If any one knows the answer for any of the question pls assist me or > maybe you can push me in the right direction because now I am running > in circles. > > Regards > Andreas > [AndrayK]. Neither WindRiver's TFFS nor M-Systems's DiskOnChip driver work with IDE/ATA flash disks. The former supports NOR flash arrays only; the latter supports M-Systems' DiskOnChip parts only. You will need to use WindRiver's ATA driver with flash IDE/ATA disks. Andray --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Error in boot line Date: Mon, 26 Aug 2002 16:13:20 -0700 From: "Michael R. Kesti" Organization: MK Associates Message-ID: <3D6AB610.7321AFE@gv.net> References: <200208261952.MAA11902@baja.tuc.noao.edu> Reply-To: mkesti@gv.net "Chirag H. Shroff" wrote: >Hello All, > >We are using MVME2700 PPC. I use the "bootChange" command to change the boot >parameters. > >I add the following...... > >boot device : dc >processor number : 0 >host name : console >file name : /opt/vxworks/tornado1PPC/target/config/mv2700/vxWorks >inet on ethernet (e) : 192.168.19.3:ffffff00 >inet on backplane (b): >host inet (h) : 192.168.19.1 >gateway inet (g) : >user (u) : gong >ftp password (pw) (blank = use rsh): >flags (f) : 0x8 >target name (tn) : data >startup script (s) : /ram2/startup.cmd >other (o) : > >Everything works fine as long as power is "ON". When i turn off the power >supply and then after some time turn the power supply "on", I get the following >error messages..... > >data> Error in boot line: >dc(0,0)Condole:/opt/vxwOrks/tornado1PPC/target/config/mV700/vxWorks e=q0.168.19.3:fffFfn00 h=192.168.y.1 u=gong f=0xh tn=data s=/ram/startup.cmd >Before that it was fine. Now if i change the boot parameters again and >reboot it (dont turn the power supply "off") everything works fine. > >Pleaseeee help me. I'll really appreciate it. > >Thanks a lot. > >Chirag The boot parameter data are stored in NVRAM (Non-Volatile RAM) in order to be retained through power cycles. It would seem that the NVRAM on your target is not performing well. I did a bit of research, and it appears that the NVRAM on the MVME2700 is a battery backed SRAM (other boards use other technologies). If this is correct, it may be that the battery is not capable of holding enough charge to hold the memory or perhaps the battery charger is not doing its job well. Do you experience this problem on only one hardware set? If yes, then my suspicion of the battery/charger is supported. If the problem occurs on more than one board, however, then the problem may be due to using the wrong NVRAM support software or a flaw in the "correct" software. Good luck! - -- ======================================================================== Michael Kesti | "And like, one and one don't make | two, one and one make one." mkesti@gv.net | - The Who, Bargain --------------------------- Newsgroups: comp.os.vxworks Subject: changing the PTE table Date: 26 Aug 2002 17:35:02 -0700 From: jchen@terayon.com (john) Organization: http://groups.google.com/ Message-ID: Hello developers: Has anyone written any code to change the PTE table entries? We're working on a PPC 8260. I'm curious because I have written code and it doesn't seem to work.... My project involves "fooling" the mmu to think that data is stored somewhere else. I want to allocate a structure, and have it point to different memory regions (that have the same structure definition) by manually re-writing the PTE table... any ideas? Everytime I tried to modify the PTE table entry, vxWorks throws me an exception. I added tasklock and intlock functions around it to no avail. We're not running shared memory model so I assume we don't need to lock the PTE as described by Motorola. Any thoughts? john --------------------------- Newsgroups: comp.arch.bus.vmebus,comp.os.os9,comp.os.vxworks,comp.arch.embedded Subject: OEMs, WE WANT TO PURCHASE OR CONSIGN YOUR VME BOARDS Date: Tue, 27 Aug 2002 01:29:27 GMT From: "SHS" Organization: Prodigy Internet http://www.prodigy.com Message-ID: We can purchase your excess VME inventory or consign it in our inventory for sale to third parties. SHS is a VME dealer of primarily Morotorla VME CPU boards. Please fax your list of boards to 281-488-5183 or email wmcalpin@sunharborsales.com with prices and board condition. Include dash numbers such as MVME162-533, etc. Bill McAlpin Sun Harbor Sales Houston, Texas --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Use Parallel Port instead of the required Serial Port ? Need Null Modem line ? Date: 26 Aug 2002 19:39:00 -0700 From: patrickyu_ca@yahoo.com (patrick) Organization: http://groups.google.com/ Message-ID: References: <488e459a.0208240913.756150a2@posting.google.com> Thanks all, I had thought the changer my friend would give me sould be a parallel to serial changer. It is not, but a DB25 to DB9 changer. So No use for this condition. Anyway, thanks very much, especially John(john_94501@yahoo.com). Patrick --------------------------- Newsgroups: comp.os.vxworks Subject: Where can i get the documents about DosFs2.0? Date: 26 Aug 2002 19:55:11 -0700 From: oe_zhou@sohu.com (oe_zhou) Organization: http://groups.google.com/ Message-ID: <758a30ff.0208261855.6ec796fd@posting.google.com> thanks. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: elf/ coff/ Motorola S-record/Intel Hex/Bin ? Date: 26 Aug 2002 22:58:09 -0700 From: vijay.peshkar@wipro.com (Vijay Peshkar) Organization: http://groups.google.com/ Message-ID: References: <8400a733.0208260018.4993d07d@posting.google.com> Hi, Adding to Toni's reply, the format you will need will be decided by the processor & the format your flash device understands. regards, Vijay > > Could anybody tell me how to use these format in VxWorks envioroments? > > well... the executable format is toolchain specific... e.g. > T2 PPC uses elf > T2 x86 uses a.out > the rest are derived formats, you can use tools like > elfToBin > or > elfHex > to get them... > (or more sophisticated ones, like the convert utility from sds to get > motorola s-record format) > > toni --------------------------- Newsgroups: comp.os.vxworks Subject: driver delivery process Date: 26 Aug 2002 23:33:18 -0700 From: vxnerd@rediffmail.com (subbarayan) Organization: http://groups.google.com/ Message-ID: dear mentors, Can anyone tell me whats the type of file delivered normally for vxworks projects and applications and drivers?What sort of file is ususally the industry norm and practiced norm while delivering vxworks apps and drivers? As I am new to this sort of OS it will be helpful If any one of u could help me know these things.It would also be helpful if u could suggest me what sort of file is better for what sort of application and a comparitive study between the final application file type with respect to vxworks. Advanced thanks to the person replying. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Where can i get the documents about DosFs2.0? Date: Tue, 27 Aug 2002 06:41:11 +0000 (UTC) From: Johan Borkhuis Organization: Agere Systems Message-ID: References: <758a30ff.0208261855.6ec796fd@posting.google.com> oe_zhou@sohu.com (oe_zhou) wrote: > thanks. http://web2.windriver.com/windsurf/techpubs/ide/tornado/t2/index.html Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Three questions about PPPoE Date: 26 Aug 2002 23:44:22 -0700 From: thatbird@sina.com (thatbird) Organization: http://groups.google.com/ Message-ID: I. I now use PPPoE component to develop a pppoe host device. The version is 1.0. I set Params: PPPOE_DEV_NAME "motfec" PPPOE_DEV_UNIT "0" PPPOE_MAX_SERVICES 5 PPPOE_OPERATING_MODE PPPOE_HOST_MODE PPPOE_SERVICE_NAME "gold" After read pppoe_10draft.pdf, I wrote the following function. void PPPoE_HOST_open() { PFW_STACK_OBJ * stackObj; UINT32 PPPoE_HOST_usrHandle; if ((stackObj = pfwStackAdd(pppOeSysProfile, &applCallbacks, PPPoE_HOST_usrHandle))== NULL) { printf("PPPOE: Stack add failed\n"); return; } printf("stackObj = 0x%x\n", stackObj ); } In the function, applCallbacks is a set of example callbacks that is defined in /target/src/ppp/unsupported/pppAppl.c when PPPoE_HOST_open() finished, I got: *************************** - -> PPPoE_HOST_open pfwStackAddComplete end stackObj = 0x18f0488 value = 21 = 0x15 - -> *************************** "pfwStackAddComplete end" is printed at the end of "pfwStackAddComplete" defined in pfwStack.c But I can not get the callbacks which are defined in applCallbacks. When I excuted "pppConnectionOpen (0x18f0488,&pppUpcalls, tickGet)", I got: *************************** - -> pppConnectionOpen (0x18f0488,&pppUpcalls, tickGet) PPP: Invalid OR Incomplete stackObj 0x18f0488 value = -1 = 0xffffffff = udbinfo + 0xffed5907 - -> *************************** How can I deal with it? II. Should I config the file usrBasicPPPCfg.c and usrBasicPPPoECfg.c? III. How could I use the set of function "PPP_UPCALL_FUNCTIONS pppUpcalls" correctly? For example, I want to exchange usrname and passWord with remote peer(AC). --------------------------- Newsgroups: alt.os.linux.suse,comp.os.linux,comp.os.linux.advocacy,comp.os.linux.embedded,comp.os.vxworks Subject: Re: newsgroup spreading virus Date: Tue, 27 Aug 2002 06:59:12 GMT From: Tor Slettnes Organization: Slett.Net Message-ID: References: In article , Russ wrote: > Let me explain this slowly so a former help desk employee can understand : No need. These groups are typically read by fairly technical people. Bills (and mine, and I am sure, more people's) confusion was a result of a crucial omission in our first post: That the e-mails (or newsgroup postings?) was forged _with your address_ in the From: field: > I'm getting bounced email > The email has my address in the "From". ... > I never sent the email. > Therefore, the From address was forged. > The real origin is one of several earthlink accounts. > Therefore, it is an earthlink problem. If you have the original e-mail, take a look at the "Received:" headers. (Some of them may be forged, but you should see at least YOUR computer's idea about where it got the mails.) Once you have the transmitting computer's e-mail address, you can try to do a little bit of probing (probably illegal in the US and a handful of other very restrictive countries, but if you live in the free world, you can continue). If you have Samba installed, try for instance: nmblookup -A If you get a response, you are not only dealing with a Windows machine, but a Windows machine that is shared directly on the internet. Chances are, therefore, that that computer's owner is a victim of this virus, just like you. If so, you can also warn the machine's owner with a little pop-up message on their screen: smbclient -M -I You wouldn't neccessarily identify yourself in the message. People with the maximum IQ expected of Windows users tend to be very irrational about warnings like these, probably suspecting YOU of "hacking" their machine. Anyway, regarding your post, I can understand that you are probably frustrated and tired of dealing with this - but I think you ought to be a bit more civil all the time YOU are the one asking for help here. - -tor --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VXWORKS Date: Tue, 27 Aug 2002 09:12:15 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Ray, if you want the right candidates, you should disclose a bit more about the opportunity, e.g. contractor or payroll, location, term, relocation availability ... Good luck. "Ray Pavri" wrote in message news:akf1ui$jc$1@overload.lbl.gov... > > This is a multi-part message in MIME format. > > ------=_NextPart_000_00D2_01C24DDC.DA2B2640 > Content-Type: text/plain; > charset="iso-8859-1" > Content-Transfer-Encoding: quoted-printable > > Employment opportunity for someone with VxWorks experience developing = > device drivers. Those interested and with a VxWorks background should = > email raypavri@sustainergy.com.au in the first instance. > > regards > > Ray Pavri > > ------=_NextPart_000_00D2_01C24DDC.DA2B2640 > Content-Type: text/html; > charset="iso-8859-1" > Content-Transfer-Encoding: quoted-printable > > > > http-equiv=3DContent-Type> > > > > >
Employment opportunity for someone with VxWorks = > experience=20 > developing device drivers. Those interested and with a VxWorks = > background should=20 > email href=3D"mailto:raypavri@sustainergy.com.au">raypavri@sustainergy.com.au A> in=20 > the first instance.
>
 
>
regards
>
 
>
Ray Pavri
> > ------=_NextPart_000_00D2_01C24DDC.DA2B2640-- > --------------------------- Newsgroups: comp.os.vxworks Subject: VxWorks for different PPC. Date: 27 Aug 2002 00:39:32 -0700 From: jadeok@sina.com (happylad) Organization: http://groups.google.com/ Message-ID: Can any one tell me why Vxwork for 8260 and 860 is different?for 74xx is different from 7xx. for arm 7 is different arm 9. I am confused with this .?can anyone tell me the reason?How about Tornado? Thanks in advance Happylad --------------------------- Newsgroups: comp.os.vxworks Subject: Regarding System time Date: 27 Aug 2002 00:53:11 -0700 From: vxworksonline@yahoo.com (Shiny) Organization: http://groups.google.com/ Message-ID: Hi, I have a problem in getting the system time. It gives me the time only from 0:0 instead of the exact time. When I used SysOutByte to use RTC methods as discussed in this groups earlier there was some problem .Can anybody help me in getting the system time. T, Shiny --------------------------- Newsgroups: comp.os.vxworks Subject: cpu usage Date: 27 Aug 2002 00:56:26 -0700 From: gokul@cynetanetworks.com (gnp) Organization: http://groups.google.com/ Message-ID: Hi Can anyone recommend code for detecting cpu usage by different tasks? SpyReport prints the report on the shell, but is there a way to call this from a task and get the cpu usage distribution for each task or for a particular task. Is there a way to detect bad memory access from within the code on vxworks to avoid data aborts or reboots. Thanks --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks for different PPC. Date: Tue, 27 Aug 2002 10:02:48 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: This is because the different PowerPC derivatives have a. different instruction subset or additions, and the VxWorks is compiled to use the proper insruction subset to be optimized for a particular CPU b. there are assembly modules in VxWorks that are modified for certain CPUs, to take care of their specific system functions (MMU, Cache) etc. Tornado IDE for a certain target architecture support all CPUs in that architecture, except there might be version issues, i.e. a certain CPU was added starting on a particular version, so pervious versions might not support certain derivatives. Note that the compiler also needs to be slightly modified to support new CPUs on a supported architecture, to be able to generate optimized code for the specific instruction subset. This is usually referred to as "mini architecture port". "happylad" wrote in message news:d6d79388.0208262339.63ad853b@posting.google.com... > Can any one tell me why Vxwork for 8260 and 860 is different?for 74xx > is different from 7xx. for arm 7 is different arm 9. > > I am confused with this .?can anyone tell me the reason?How about > Tornado? > > Thanks in advance > > > Happylad --------------------------- Newsgroups: comp.os.vxworks Subject: Re: trouble in ping with a larger packet on ne2000 Ethernet Date: 27 Aug 2002 01:06:16 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0208270006.726e3f09@posting.google.com> References: <7f29c2ad.0208080439.62e501b7@posting.google.com> <3D54F62E.7060304@spamm.me.l8s.co.uk> David Laight wrote in message news:<3D54F62E.7060304@spamm.me.l8s.co.uk>... > > > > when I ping it with a litter packet(<1500 bytes),it is ok,but when the > > packet is larger(>1500),ping is failed! why? This might be a driver bug? In any case it would be a good idea considering to update to a later version (to give it at least a try)... e.g. Tornado 2.0.2 + SPR68322 toni --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Soft Interrupts and getch() in VxWorks Date: 27 Aug 2002 01:08:42 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0208270008.31021f05@posting.google.com> References: ajay_garg@delhi.tcs.co.in (Ajay Garg) wrote in message news:... > One more point, there is no support for getch() kind of thing too > though getchar() is there. But getchar() does not display the > character immediately on the screen, it requires a return key. So how TTY options have already been pointed out in another posting (switch to RAW mode). In addition, be sure NOT todo this from the WindShell... the WindShell only works in line mode (as it is a TCL based interpreter which interprets some of the stuff you enter localy... and only forwards unknown commands to the target...) toni --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting Tornado 2.0 Date: 27 Aug 2002 01:11:07 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0208270011.10e515a1@posting.google.com> References: <32bae160.0208260801.1cbe073b@posting.google.com> zenk78@hotmail.com (Robert Templeman) wrote in message news:<32bae160.0208260801.1cbe073b@posting.google.com>... > I just got back from a Tornado Training. Tornado 2.0.x will NOT > support XP, but....Tornado 2.2 was released TODAY and has full support > for XP Professional as well as many other features. If you have > maintainance on 2.0 you can upgrade to 2.2 for FREE. It is a pull > release so you much contact Windriver for it. And it does have > extensive license management so that must be resolved. Rob Templeman Yes, Tornado 2.2 is the officially supported version on Windows XP. Tornado 2.0.2 should also work, although it is considered as unsupported by wrs. There are people working on the XP T202 configuration. However, wrs will only support the T22 SP platform. toni --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks bh() command question? Date: 27 Aug 2002 01:13:21 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0208270013.5350fa77@posting.google.com> References: Yuan Zheng-ZYUAN1 wrote in message news:... > Does this mean archetecture on my system (Motorola MPC8260) > doesn't support bh on memory write and can only support memory > read? Also does the bh() can only break for one memory > location instead of multiple location? I'm not enterely sure here, but from what I know by hard the 8260 does NOT support data hardware breakpoints. E.g. the 860 would support this. toni --------------------------- Newsgroups: comp.os.vxworks Subject: IDE Flash disk Date: 27 Aug 2002 01:19:34 -0700 From: andreas_lindell75@hotmail.com (Rikard) Organization: http://groups.google.com/ Message-ID: <5947a04.0208270019.718c3c0b@posting.google.com> Hello, I am working with a IFD-350-64 IDE flash disk and I am trying to get vxworks to boot from it, bootrom.sys and vxworks image on the flash disk. I am using Tornado 2.0.2 and I have upgraded to dosFs2.2 and I am using a pentium target. I have done the following in config.h: #elif (CPU_VARIANT == PENTIUM) #define DEFAULT_BOOT_LINE \ "ata=0,0(0,0)host:/ata0/vxWorks h=90.0.0.3 e=90.0.0.50 u=target" #elif (CPU_VARIANT == PENTIUMPRO) #define DEFAULT_BOOT_LINE \ "fd=0,0(0,0)host:/fd0/vxWorks.st h=90.0.0.3 e=90.0.0.50 u=target" #endif /* (CPU == I80386) */ #define INCLUDE_DOSFS /* include dosFs file system */ #define INCLUDE_FD /* include floppy disk driver */ #undef INCLUDE_LPT /* include parallel port driver */ #define INCLUDE_ATA /* include IDE/EIDE(ATA) hard disk driver */ #undef INCLUDE_TIMESTAMP /* include TIMESTAMP timer for Wind View */ #undef INCLUDE_TFFS /* include TrueFFS driver for Flash */ #undef INCLUDE_PCMCIA /* include PCMCIA driver */ #ifdef INCLUDE_PCMCIA #define INCLUDE_ATA /* include ATA driver */ #define INCLUDE_SRAM /* include SRAM driver */ #undef INCLUDE_TFFS /* include TFFS driver */ #ifdef INCLUDE_NETWORK #define INCLUDE_ELT /* include 3COM EtherLink III driver */ #endif /* INCLUDE_NETWORK */ #endif /* INCLUDE_PCMCIA */ #ifdef INCLUDE_TFFS #define INCLUDE_SHOW_ROUTINES #endif /* INCLUDE_TFFS */ The bootrom load fine and I get the [VxWorks Boot]: prompt, but when I do a "@" I get the following: Attaching to ATA disk device... done Loading /ata0/vxWorks...0+0+0 Starting at 0x530b1... [VxWorks Boot]: This means that the vxworks kernel is not up and running? If I do another "@" I get a: general protection fault and then its game over. Any idea to what this can be? I have the image on the flash disk and therefore its a little bit strange when I get the "Loading /ata0/vxWorks...0+0+0"?? If anyone have seen this error befor pls let me know. Regards Andreas --------------------------- Newsgroups: comp.os.vxworks Subject: Basic Command Line Date: Tue, 27 Aug 2002 08:40:47 GMT From: "CHANGE username to just westes" Organization: Giganews.Com - Premium News Outsourcing Message-ID: We have a computer device that uses VxWorks internally to control robotics. The device's administrator is supposed to login and configure the device using the VxWorks command line. After reading the long list of VxWorks commands available within the device I don't see any way to explore the file system and read the contents of various files. Where can I get basic command line documentation that would explain how to do this? Is there any chance that the manufacturer would disable these commands? We have a situation where two similar pieces of equipment behave slightly differently, and one takes the latest firmware and the other does not, complaining about missing files during the bootup process. We need to be able to see what is on the file system in order to figure out what is missing or different on the system that won't work with the latest firmware. Needless to say, the manufacturer isn't interested in helping us at that level of detail (and maybe doesn't have anyone on their tech support staff who could even do this, even if they were willing to support it). Any tips on how we can explore the device and figure out what is wrong with the file system structure and files is appreciated. - -- Will NOTE: To reply, CHANGE the username to westes AT uscsw.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Floating point on MBX860 Date: Tue, 27 Aug 2002 09:45:03 +0100 From: "David J Edgar" Organization: Cedaryacht Limited Message-ID: References: "Sandeep P" wrote in message news:akcdiq$3jp$1@news.asu.edu... > Hello, > I am new to the MBX860. I am running vxworks on it. Is anybody > aware of a solution to perform floating point calculations. Obviously, it > would imply that I can have floating point variables. > > Bye, > Sandeep. > > The MBX860 is MPC860 based and does not support floating point calculations. - -- David J Edgar Cedaryacht Limited, Macclesfield, Cheshire, UK E-mail: david@cd-yacht.demon.co.uk Web: http://www.cedaryacht.co.uk --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Basic Command Line Date: Tue, 27 Aug 2002 08:59:01 +0000 (UTC) From: Johan Borkhuis Organization: Agere Systems Message-ID: References: "CHANGE username to just westes" wrote: > We have a computer device that uses VxWorks internally to control > robotics. The device's administrator is supposed to login and > configure the device using the VxWorks command line. After reading > the long list of VxWorks commands available within the device I don't > see any way to explore the file system and read the contents of > various files. Where can I get basic command line documentation that > would explain how to do this? Is this a commandline specific for this device, or is it the generic VxWorks command line? (What does it show after you connect or on startup?) Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: IDE Flash disk Date: Tue, 27 Aug 2002 11:11:19 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <5947a04.0208270019.718c3c0b@posting.google.com> An IDE Flash disk behaves exactly like magnetic IDE disk drive, except it has a very small seek time and slowish write performance. Therefore, you do not need TFFS for this setup, Follow the example of setting up a regular IDE hard disk, and that should do the trick. Before trying to boot fromt he IDE disk, make a VxWorks image that supports the disk, and boot it from network, then see if the disk mounts OK, partitions decoded correctly and you can see the files on the disk. Only after the VxWorks image works well with the disk, then try to make a bootable image. "Rikard" wrote in message news:5947a04.0208270019.718c3c0b@posting.google.com... > Hello, > > I am working with a IFD-350-64 IDE flash disk and I am trying to get > vxworks to boot from it, bootrom.sys and vxworks image on the flash > disk. > > I am using Tornado 2.0.2 and I have upgraded to dosFs2.2 and I am > using a pentium target. I have done the following in config.h: > > #elif (CPU_VARIANT == PENTIUM) > #define DEFAULT_BOOT_LINE \ > "ata=0,0(0,0)host:/ata0/vxWorks h=90.0.0.3 e=90.0.0.50 u=target" > #elif (CPU_VARIANT == PENTIUMPRO) > #define DEFAULT_BOOT_LINE \ > "fd=0,0(0,0)host:/fd0/vxWorks.st h=90.0.0.3 e=90.0.0.50 u=target" > #endif /* (CPU == I80386) */ > > #define INCLUDE_DOSFS /* include dosFs file system */ > #define INCLUDE_FD /* include floppy disk driver */ > #undef INCLUDE_LPT /* include parallel port driver */ > #define INCLUDE_ATA /* include IDE/EIDE(ATA) hard disk driver */ > #undef INCLUDE_TIMESTAMP /* include TIMESTAMP timer for Wind View */ > #undef INCLUDE_TFFS /* include TrueFFS driver for Flash */ > #undef INCLUDE_PCMCIA /* include PCMCIA driver */ > > #ifdef INCLUDE_PCMCIA > #define INCLUDE_ATA /* include ATA driver */ > #define INCLUDE_SRAM /* include SRAM driver */ > #undef INCLUDE_TFFS /* include TFFS driver */ > #ifdef INCLUDE_NETWORK > #define INCLUDE_ELT /* include 3COM EtherLink III driver */ > #endif /* INCLUDE_NETWORK */ > #endif /* INCLUDE_PCMCIA */ > > #ifdef INCLUDE_TFFS > #define INCLUDE_SHOW_ROUTINES > #endif /* INCLUDE_TFFS */ > > The bootrom load fine and I get the > [VxWorks Boot]: > prompt, but when I do a "@" I get the following: > Attaching to ATA disk device... done > Loading /ata0/vxWorks...0+0+0 > Starting at 0x530b1... > > [VxWorks Boot]: > > This means that the vxworks kernel is not up and running? If I do > another "@" I get a: > general protection fault > and then its game over. > > Any idea to what this can be? I have the image on the flash disk and > therefore its a little bit strange when I get the "Loading > /ata0/vxWorks...0+0+0"?? > > If anyone have seen this error befor pls let me know. > > Regards > Andreas --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Floating point on MBX860 Date: Tue, 27 Aug 2002 11:15:32 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: Reply-To: "Michael Lawnick" Sender: mlawnick@p5088f91d.dip.t-dialin.net Hi Sandeep, David I have no MBX860 of my own, but I'm pretty sure that if the 860 H/W does not support floating point, VxWorks will support S/W-floating point anyway. Just include it in your project. Regards, Michael "David J Edgar" schrieb im Newsbeitrag news:CDYC98894D2CD@cd-yacht.demon.co.uk... > "Sandeep P" wrote in message > news:akcdiq$3jp$1@news.asu.edu... > > Hello, > > I am new to the MBX860. I am running vxworks on it. Is anybody > > aware of a solution to perform floating point calculations. Obviously, it > > would imply that I can have floating point variables. > > > > Bye, > > Sandeep. > > > > > > The MBX860 is MPC860 based and does not support floating point calculations. > > -- > David J Edgar > Cedaryacht Limited, Macclesfield, Cheshire, UK > E-mail: david@cd-yacht.demon.co.uk > Web: http://www.cedaryacht.co.uk > > --------------------------- Newsgroups: comp.os.vxworks Subject: How to seperate data, Date: Mon, 26 Aug 2002 10:41:41 +0300 From: "Martin Roth" Organization: Motorola Message-ID: For my project I need to separate the code and data into two physical memory sections: 1) Physical memory section 1 consist of static RAM memory which is mapped to address X and should include the link sections that may change during the program flow. For example: data , bss. The static RAM area is also supposed to include the dynamic memory area that is handled by Vxworks. 2) Physical memory section 2 consist of SDRAM memory which is mapped to some other memory address Y and it includes the sections that do not change during the program flow. For example : text, const Question 1: How can I perform the separation between sections that do not change (text, const) and sections that do change (data,bss, dynamic memory handled by Vxworks). My purpose is to populate the static ram memory with data,bss sections and dynamic memory handled by Vxworks. The SDRAM (other physical memory) is planned to be populated with text, const sections. The reason for all that is the relative high price of static RAM and the requirements that we have use static RAM to support power off and on survival. To lower the price of the product, I intend to use where possible SDRAM. Question 2: Are there any additional link sections I should be aware ? --------------------------- Newsgroups: comp.os.vxworks Subject: difference between pciautoconfiglib.c and pciconfiglib.c Date: 27 Aug 2002 03:19:06 -0700 From: vxnerd@rediffmail.com (subbarayan) Organization: http://groups.google.com/ Message-ID: dear mentors, Can anyone jus tell me how to go abt proceeding the development of device driver for pci card for x86 processor with standard pc architecture?I use tornadoII and windows development host and a stand alone target. Further can anyone tell me when will be this pciautoconfiglib file called?will it be called automatically as a startup routine or need it to be called explicitly to enable pci scanning? What exactly is the difference between the files pci autoconfiglib.c and pciconfiglib.c?which file should be used for which sort of appln?and can these library files directly compiled and used or should we call it from some appln to invoke and access it? And sorry for my ignorance and unknowing things.One thing i found in the manual for pci support on vxworks says we need to include the macro INCLUDE_PCI for enabling pci support.Where exactly should this macro be included and where exactly is this macro defined?How to include this macro and whats the exact location where this macro has to be included?And whats the role of bios with respect to vxworks in detecting pci card and pci bus?It would be helpful if some one could tell me what would be printed on screen if pci support is enabled and makes me confirm that my code has started working with pci support. Sorry if my questions look amature,but i am helpless as i am new to this sort of coding. Advance thanks to persons replying me. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: File system shutdown Date: Tue, 27 Aug 2002 11:23:33 +0100 From: "Dimitrios Siganos" Organization: Newport Networks Ltd. Message-ID: <1030443813.80534@azores.network-i.net> References: If I understand the problem correctly, all you have to do is to call ioctl(FIOFLUSH) for each filesystem and then shutdown. It is your responsibility to make sure that FIOFLUSH is the last disk transaction on each system. Also, I don't know if ioctl(FLUSH) is synchronous or asynchronous. If it is async then you will have to find a way to detect the completion of the command. Dimitris "Joe Welfeld" wrote in message news:ak9sn3$n1b$1@overload.lbl.gov... > > > Hello, > > We are running Tornado2.02 for PPC, with TFFS and DOSFS2, I cant seem to > find the > call to make for an orderly "disk" shutdown ? perhaps like a TFFSSync > command to > insure everything is written before I shut down ? perhaps someone knows how > I should do this ? > > Thanks, > > Joe Welfeld > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: driver delivery process Date: Tue, 27 Aug 2002 06:29:01 -0400 From: joe durusau Organization: WEBUSENET.com Message-ID: <3D6B546D.E2D1372C@bellsouth.net> References: subbarayan wrote: > dear mentors, > Can anyone tell me whats the type of file delivered normally for > vxworks projects and applications and drivers?What sort of file is > ususally the industry norm and practiced norm while delivering vxworks > apps and drivers? > As I am new to this sort of OS it will be helpful If any one of u > could help me know these things.It would also be helpful if u could > suggest me what sort of file is better for what sort of application > and a comparitive study between the final application file type with > respect to vxworks. > Advanced thanks to the person replying. I think that the basic answer to your question is that you need to attend a vxWorks training class. In general, you are going to have to deliver vxWorks itself (and a license), since you cannot expect customers to buy their own copies. That being the case, sometimes it is just as well to bundle app code into the OS. Other projects, other answers. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: how to use the SYSTEM MODE DEBUGGER Date: 27 Aug 2002 03:34:15 -0700 From: l.salvemini@infosol.it (Lorenzo Salvemini) Organization: http://groups.google.com/ Message-ID: Hi all, i've some problems with the system mode debugging. These are the steps that i follow each time: 1: launch the target server 2: launch the debug 3: attach to system (or inside the shell i use sysSuspend) 4: set the breakpoints with the buttons Now i've some problems: i'd like to follow the source code step-by-step jumping in and out from different tasks and stopping at different breakpoints. How can i do this procedure??? If i use the debugger buttons the yellow arrow (current code line) doesn't jump to other tasks, it stack only in that running task. Thanks for the help Lorenzo --------------------------- Newsgroups: comp.os.vxworks Subject: Doku for vxPpcLib Date: 27 Aug 2002 08:09:04 -0700 From: joerg.krein@ims.fhg.de (=?ISO-8859-1?Q?J=F6rg_Krein?=) Organization: http://groups.google.com/ Message-ID: Hi, where can I find the docu for the vxPpcLib? I found vxPpcLib.h, the 'header for board dependent library routines for the PowerPC'. Great, but where are the functions like vxImmrGet() documented? Couldn't find it in the html docs. Thanx! --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Doku for vxPpcLib Date: Tue, 27 Aug 2002 17:34:29 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Look aththe end of the VxWorks Programmer's Guide, there is an Appendix per each architecture supported, listing all architecture specific functions and implementation details. "Jörg Krein" wrote in message news:adcf93df.0208270709.2802e5c1@posting.google.com... > Hi, > where can I find the docu for the vxPpcLib? > I found vxPpcLib.h, the 'header for board dependent > library routines for the PowerPC'. > Great, but where are the functions like vxImmrGet() > documented? > Couldn't find it in the html docs. > Thanx! --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks bh() command question? Date: 27 Aug 2002 09:15:36 -0700 From: cpotenza@onetel.net.uk (Claudio Potenza) Organization: http://groups.google.com/ Message-ID: <2025ac82.0208270815.bb9fdc6@posting.google.com> References: > I am trying to trace the write operation for a block of memory address, > and bh() comes to my attention, however when i use it. I got the following > error -> bh 0x019A23D8, 1 > WTX Error 0x100d9 (AGENT_INVALID_HW_BP) > value = -1 = 0xffffffff > while i dont have error with > -> bh 0x019A23D8, 0 > value = 0 = 0x0 > Does this mean archetecture on my system (Motorola MPC8260) doesn't support > bh on memory write and can only support memory read? Also does the bh() can > only break for one memory location instead of multiple location? As far as I know, on the 8260, the command "bh 0x019A23D8, 0" sets an instruction breakpoint at that address (break when you fetch from here) and not a data read breakpoint. Other values for the second parameter of "bh" are not supported because the 8260 implements only a "IABR" register (only one) for instruction breakpoints. For instance on a PPC 750 you can use: bh
,0 instruction breakpoint bh
,1 data read/write breakpoint bh
,2 data read breakpoint bh
,3 data write breakpoint because this PPC implements also the DABR register. The 860 has much richer debug capabilities (several instruction and data breakpoint registers and can do less/equal/greater comparisons) but I don't know if they can be activated from bh or other vxWorks facilities. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Source of mips toolchain Date: 27 Aug 2002 09:51:10 -0700 From: chs@lucent.com (ChengHsin Hsu) Organization: http://groups.google.com/ Message-ID: <8b53eb96.0208270851.43ab2243@posting.google.com> References: <8b53eb96.0208260539.7ac17e6e@posting.google.com> Johan Borkhuis wrote in message news:... > chs@lucent.com (ChengHsin Hsu) wrote: > > > All, > > > > Has anyone had the experience to get a copy of toolchain source > > code from wrs? I have sent several emails via their supports but > > nobody replies at all....I'm wondering whether this is possible..And > > interested in how can I do that... > > Following is my cross gcc's version and configuration. We bought > > this binary compiler from wrs, but wanna move to Linux as our > > development environment.... > > Yes, no problem. With the 2.1 release it is included on the normal CD if I > remember correctly, and for the 2.0 you can order the source cd from them > for free. They have to provide it, as they are using GNU tools (see GPL). > > Kind regards, > Johan Thanks!!! I'll try it again. :p Regards, Bear. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: how to execute vxworks shell routines from other programs Date: Tue, 27 Aug 2002 13:40:03 -0400 From: "RonnoBonno" Message-ID: References: Jackie, I wrote the following function to do what you are looking for: it takes the shell command as a string, and the output file, executes the shell command and redirects the output to a file, it used in the shell by: -> shellToFile("ifShow","ifShow.out") You can easily modify it to do what you want. Good Luck, Ron Naismith - ------------------------------------ // // // This function runs a shell command and captures the output to the specified file // extern int consoleFd; typedef unsigned int (*UINTFUNCPTR) (); extern "C" int shellToFile(char * shellCmd, char * outputFile) { int rtn; int STDFd; int outFileFd; outFileFd = creat( outputFile, O_RDWR); printf("creat returned %x as a file desc\n",outFileFd); if (outFileFd != -1) { STDFd=ioGlobalStdGet(STD_OUT); ioGlobalStdSet(STD_OUT,outFileFd); rtn=execute(shellCmd); if (rtn !=0) printf("execute returned %d \n",outFileFd); ioGlobalStdSet(STD_OUT,STDFd); } close(outFileFd); return (rtn); } - ---------------------------- Jackie Bhadange wrote in message ... > >This is a multi-part message in MIME format. > >------=_NextPart_000_001C_01C24DAE.4F6BD370 >Content-Type: text/plain; > charset="iso-8859-1" >Content-Transfer-Encoding: 7bit > > >Hi All, > >I have a question regarding shell routines. Can I execute vxworks shell >routines(ex.inetstatShow) from my program? I would like to control the >output of these routines. I have a custom command line interface where I >would like to display this information. >I cannot use the vxworks shell because 1. I would like to use my own >interface and 2. vxworks shell is not reentrant(and I would like to have >this capability for more than one session). > >Any help is appreciated. >Thanks, >Jackie Bhadange > > > >------=_NextPart_000_001C_01C24DAE.4F6BD370 >Content-Type: text/html; > charset="iso-8859-1" >Content-Transfer-Encoding: quoted-printable > >xmlns:w=3D"urn:schemas-microsoft-com:office:word" = >xmlns:st1=3D"urn:schemas-microsoft-com:office:smarttags" = >xmlns=3D"http://www.w3.org/TR/REC-html40"> > > >charset=3Diso-8859-1"> > > > > > > >namespaceuri=3D"urn:schemas-microsoft-com:office:smarttags" > name=3D"PersonName"/> > > > > > >style=3D'tab-interval:.5in'> > >
> >

style=3D'font-size:10.0pt; >font-family:"Courier New"'> 

> >

style=3D'font-size:10.0pt; >font-family:"Courier New"'>Hi All,

> >

style=3D'font-size:10.0pt; >font-family:"Courier New"'> 

> >

style=3D'font-size:10.0pt; >font-family:"Courier New"'>I have a question regarding shell routines. = >Can I >execute vxworks shell class=3DGramE>routines(class=3DSpellE>ex.inetstatShow) from my program? I would like to = >control >the output of these routines. I have a custom command line interface = >where I >would like to display this information.

> >

style=3D'font-size:10.0pt; >font-family:"Courier New"'>I cannot use the class=3DSpellE>vxworks >shell because 1. I would like to use my own interface and 2. class=3DSpellE>vxworks shell is not = >reentrant(and >I would like to have this capability for more than one = >session).

> >

style=3D'font-size:10.0pt; >font-family:"Courier New"'> 

> >

style=3D'font-size:10.0pt; >font-family:"Courier New"'>Any help is = >appreciated.

> >

style=3D'font-size:10.0pt; >font-family:"Courier New"'>Thanks,

> >

New"> style=3D'font-size:10.0pt;font-family:"Courier New"'>Jackie = >Bhadangesize=3D2 face=3DArial>style=3D'font-size:10.0pt;font-family:Arial'>> > >

style=3D'font-size:10.0pt; >font-family:Arial'> 

> >

style=3D'font-size: >12.0pt'> 

> >
> > > > > >------=_NextPart_000_001C_01C24DAE.4F6BD370-- > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Regarding System time Date: Tue, 27 Aug 2002 14:11:37 -0400 From: david lindauer Message-ID: <3D6BC0D9.D6132833@notifier-is.net> References: Shiny wrote: > Hi, > I have a problem in getting the system time. > It gives me the time only from 0:0 instead of the exact time. > When I used SysOutByte to use RTC methods as discussed in this > groups earlier there was some problem .Can anybody help me in getting > the system time. > T, > Shiny you have to do it by hand (as discussed earlier). Bear in mind that a task switch/interrupt may cause timing issues when reading/writing RTC registers. You probably want to bracket the core code to do RTC I/O with intLock() statements. David --------------------------- Newsgroups: comp.os.vxworks Subject: Re: how to use the SYSTEM MODE DEBUGGER Date: Tue, 27 Aug 2002 14:13:46 -0400 From: david lindauer Message-ID: <3D6BC15A.5871A903@notifier-is.net> References: Lorenzo Salvemini wrote: > Hi all, > i've some problems with the system mode debugging. These are the steps > that i follow each time: > 1: launch the target server > 2: launch the debug > 3: attach to system (or inside the shell i use sysSuspend) > 4: set the breakpoints with the buttons > Now i've some problems: i'd like to follow the source code > step-by-step jumping in and out from different tasks and stopping at > different breakpoints. > How can i do this procedure??? > If i use the debugger buttons the yellow arrow (current code line) > doesn't jump to other tasks, it stack only in that running task. > > Thanks for the help Lorenzo you will have to set breakpoints in your other tasks, then let it run till it hits one. David --------------------------- Newsgroups: comp.os.vxworks Subject: Interrupt vector table Date: Tue, 27 Aug 2002 21:54:41 +0200 From: "JJS" Message-ID: Our system : Tornado 2.0 VxWorks 5.4 mvme 2304 board In a 68000 or x86 system, the interrupt vector table begins in memory at address 0 So vector 10 is at address 10*4 Where is the vector table in a ppc architecture? Thanks Jack --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Interrupt vector table Date: Tue, 27 Aug 2002 13:17:53 -0700 From: "Michael R. Kesti" Organization: MK Associates Message-ID: <3D6BDE71.8B3BE02E@gv.net> References: Reply-To: mkesti@gv.net JJS wrote: >Our system : Tornado 2.0 VxWorks 5.4 mvme 2304 board >In a 68000 or x86 system, the interrupt vector table begins in memory at >address 0 >So vector 10 is at address 10*4 >Where is the vector table in a ppc architecture? There is no interrupt vector table in the PPC. Instead, interrupt handlers are executed immediately at 0x100 byte offsets from a base address. The base address is determined by the setting of a register. That register is initialized to either 0x00000000 or 0xfff00000 at reset by reading a "configuration data word" from hardware. >Thanks You're welcome. As this is not a vxWorks question, per se, you might have asked it one of the comp.sys.powerpc.* groups. - -- ======================================================================== Michael Kesti | "And like, one and one don't make | two, one and one make one." mkesti@gv.net | - The Who, Bargain --------------------------- Newsgroups: comp.os.vxworks Subject: using RngLib w/ my own memory Date: 27 Aug 2002 14:36:54 -0700 From: matthew.sample@med.ge.com (Matt) Organization: http://groups.google.com/ Message-ID: I want to specify where the my ring is located. How do I do this with the rngLib.h? Can I create the ring of size 0, and then reassign the pointers and reset the size? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Three questions about PPPoE Date: 27 Aug 2002 18:01:08 -0700 From: thatbird@sina.com (thatbird) Organization: http://groups.google.com/ Message-ID: References: thatbird@sina.com (thatbird) wrote in message news:... anyone can give me some advices. Thanks any way. --------------------------- Newsgroups: comp.os.vxworks Subject: new Date: 27 Aug 2002 18:28:42 -0700 From: thatbird@sina.com (thatbird) Organization: http://groups.google.com/ Message-ID: I. I am now useing PPPoE component to develop a pppoe host device. The version is 1.0. I set Params: PPPOE_DEV_NAME "motfec" PPPOE_DEV_UNIT "0" PPPOE_MAX_SERVICES 5 PPPOE_OPERATING_MODE PPPOE_HOST_MODE PPPOE_SERVICE_NAME "gold" After read pppoe_10draft.pdf, I wrote the following function. void PPPoE_HOST_open() { PFW_STACK_OBJ * stackObj; UINT32 PPPoE_HOST_usrHandle; if ((stackObj = pfwStackAdd(pppOeSysProfile, &applCallbacks, PPPoE_HOST_usrHandle))== NULL) { printf("PPPOE: Stack add failed\n"); return; } printf("stackObj = 0x%x\n", stackObj ); } In the function, applCallbacks is a set of example callbacks that is defined in /target/src/ppp/unsupported/pppAppl.c when PPPoE_HOST_open() finished, I got: *************************** - -> PPPoE_HOST_open pfwStackAddComplete end stackObj = 0x18f0488 value = 21 = 0x15 - -> *************************** "pfwStackAddComplete end" is printed at the end of "pfwStackAddComplete" defined in pfwStack.c But I can not get the callbacks which are defined in applCallbacks. When I excuted "pppConnectionOpen (0x18f0488,&pppUpcalls, tickGet)", I got: *************************** - -> pppConnectionOpen (0x18f0488,&pppUpcalls, tickGet) PPP: Invalid OR Incomplete stackObj 0x18f0488 value = -1 = 0xffffffff = udbinfo + 0xffed5907 - -> *************************** How can I deal with it? II. Should I config the file usrBasicPPPCfg.c and usrBasicPPPoECfg.c? III. How could I use the set of function "PPP_UPCALL_FUNCTIONS pppUpcalls" correctly? For example, I want to exchange usrname and passWord with remote peer(AC). --------------------------- Newsgroups: comp.os.vxworks Subject: mkbootAta Date: 27 Aug 2002 18:56:58 -0700 From: andreas_lindell75@hotmail.com (Rikard) Organization: http://groups.google.com/ Message-ID: <5947a04.0208271756.4ab69a83@posting.google.com> Hello, I am trying to create a bootable IDE flash disk from the target shell. I am using the following commands with the appropriate parameters for the system: - -> ataShow - -> ataDevCreate - -> dosfsMkfs - -> devs - -> mkbootAta - -> copy "vxWorks", "/flash/vxWorks" 3. Reboot the target and try to boot from the IDE flash again. However when I do a mkbootAta from the target shell I get an undefined symbol. What components must be included in the vxworks kernel to use this command. I have a built in symbol table and a target shell all ready. The funny thing is that I can not even find the mkbootAta in the "VxWorks Reference Manual". I have seen many postings on this site where people refer to this command and all of these seems to have no problem to use it but I cant even find it:) Regards Andreas --------------------------- Newsgroups: comp.os.vxworks Subject: Re: using RngLib w/ my own memory Date: 28 Aug 2002 02:09:33 GMT From: Denis Perelyubskiy Message-ID: References: * Matt [08-Tue-02 14:36 -0700]: > > I want to specify where the my ring is located. How do I > do this with the rngLib.h? Are there no any kind of ring buffer libraries available that you can use? > Can I create the ring of size 0, and then reassign the > pointers and reset the size? Perhaps avoid using the create/delete functions? For example, create merely allocates memory and sets up the structure -- you can do it yourself. After all, you have the layout of struct RING in the header file. so, you just need to initializea pToBuf and pFromBuf to 0, and adjust bufSize (and allocate memory) Judging from the source, you need to allocate 1 more byte then you'll actually set the bufSize to, as 1 byte is always empty for some reason. ... I'd still probably try to find another ring buffer library or write your own. Sounds like a hack to me to modify something like this.. I mean, the ringbuf library is hardly an intricate piece of software... denis - -- if you want to reply by mail, please tweak my address in a reasonable way. sorry for the inconvenience. --------------------------- Newsgroups: comp.os.vxworks Subject: system mode vs. task mode debugging Date: 28 Aug 2002 02:13:06 GMT From: Denis Perelyubskiy Message-ID: Hello, what are the differences / tradeoffs when using system-mode vs task mode debugging? As I understand it, system mode is going to be slower, but it allows one to do things like setting a breakpoint, and having a task which encounters it be suspended (to allow to attach to itself) however, this is the only difference I know (provided that it is correct :) ) Would watchpoints fall under this category? (For some reason, would be curious to know why, gdb for vxworks x86 targets refuses to understand watchpoints) thanks, denis - -- if you want to reply by mail, please tweak my address in a reasonable way. sorry for the inconvenience. --------------------------- Newsgroups: alt.os.linux.suse,comp.os.linux,comp.os.linux.advocacy,comp.os.linux.embedded,comp.os.vxworks Subject: Re: newsgroup spreading virus Date: Wed, 28 Aug 2002 02:31:40 GMT From: Russ Organization: EarthLink Inc. -- http://www.EarthLink.net Message-ID: References: Followup-To: alt.os.linux.suse Tor Slettnes wrote: > In article , > Russ wrote: > >> Let me explain this slowly so a former help desk employee can understand >> : > > No need. These groups are typically read by fairly technical people. > > Bills (and mine, and I am sure, more people's) confusion was a result of a > crucial omission in our first post: That the e-mails (or newsgroup > postings?) was forged _with your address_ in the From: field: > It's e-mails. My appologies if I didn't make that clear. >> I'm getting bounced email >> The email has my address in the "From". > ... >> I never sent the email. >> Therefore, the From address was forged. > > >> The real origin is one of several earthlink accounts. >> Therefore, it is an earthlink problem. > > If you have the original e-mail, take a look at the "Received:" headers. > (Some of them may be forged, but you should see at least YOUR computer's > idea about where it got the mails.) > > Once you have the transmitting computer's e-mail address, you can try to > do a little bit of probing (probably illegal in the US and a handful of > other very restrictive countries, but if you live in the free world, you > can > continue). If you have Samba installed, try for instance: > nmblookup -A > > If you get a response, you are not only dealing with a Windows machine, > but > a Windows machine that is shared directly on the internet. Chances are, > therefore, that that computer's owner is a victim of this virus, just like > you. > > If so, you can also warn the machine's owner with a little pop-up message > on their screen: > smbclient -M -I > > > > > You wouldn't neccessarily identify yourself in the message. People with > the maximum IQ expected of Windows users tend to be very irrational about > warnings like these, probably suspecting YOU of "hacking" their machine. > > Anyway, regarding your post, I can understand that you are probably > frustrated and tired of dealing with this - but I think you ought to be a > bit more civil all the time YOU are the one asking for help here. > > -tor What you say works fine. Some of the military bounces included enough information for me to do just as you said. I didn't do the smbclient thing, but I sent the owners I could identify a message. Right now its down to accounts I can't identify, but in all cases the trail dead ends at Earthlink. It would also appear that these account(s?) are active mostly on weekends and visit lots of new sites each weekend. There is a limit to what I have the time and capacity to do, which is why I tried to notify earthlink. There could very well be only one scriptkiddie playing silly games of lonely Friday nights. - -- Russ Lyttle - Not powered by ActiveX Please visit for OBD-II information and links. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How to include macros inside TORNADOII for vxworks Date: 27 Aug 2002 20:26:31 -0700 From: zuokf@263.net (KefengZuo) Organization: http://groups.google.com/ Message-ID: <980f6f4d.0208271926.35a686f2@posting.google.com> References: Hi: In Workspace-->Builds-->c/c++ compile, If the core of your target is PPC604, the definition "-DCPU=PPC604" defines macro "CPU" = "PPC604".The definition "_DMyMacro" defines macro "MyMacro". --------------------------- Newsgroups: comp.os.vxworks Subject: How to fix module link order when rebuilding in Tornado? Date: 27 Aug 2002 21:56:56 -0700 From: yangliang_mr@hotmail.com (Liang Yang) Organization: http://groups.google.com/ Message-ID: <4f57a77f.0208272056.68918b32@posting.google.com> Hi, I have joystick.c, joyio.c, joytimer.c and joylib.c in my project. And I want to make joystick.c always be the first module to be linked when I rebuild the project. But I found i have to change the module linker order manually whenever I rebuild since the module linker order is always restored to "joytimer.o joylib.o joystick.o joyio.o" when I start Tornado next time. It seems the order of module link is not listed according to alphabetical order. Anybody can give me a good solution for this? Thank you. Liang --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks for different PPC. Date: Wed, 28 Aug 2002 06:45:53 GMT From: "Gary M" Organization: AT&T Broadband Message-ID: References: Additionally, the BSPs are different between MPC86x and MPC82xx, in regards to the SIU and integrated subsystems. "Leonid Rosenboim" wrote in message news:newscache$o4th1h$seh$1@lnews.actcom.co.il... > This is because the different PowerPC derivatives have > a. different instruction subset or additions, and the VxWorks is compiled to > use the proper insruction subset to be optimized for a particular CPU > b. there are assembly modules in VxWorks that are modified for certain CPUs, > to take care of their specific system functions (MMU, Cache) etc. > > Tornado IDE for a certain target architecture support all CPUs in that > architecture, except there might be version issues, i.e. a certain CPU was > added starting on a particular version, so pervious versions might not > support certain derivatives. Note that the compiler also needs to be > slightly modified to support new CPUs on a supported architecture, to be > able to generate optimized code for the specific instruction subset. This is > usually referred to as "mini architecture port". > > "happylad" wrote in message > news:d6d79388.0208262339.63ad853b@posting.google.com... > > Can any one tell me why Vxwork for 8260 and 860 is different?for 74xx > > is different from 7xx. for arm 7 is different arm 9. > > > > I am confused with this .?can anyone tell me the reason?How about > > Tornado? > > > > Thanks in advance > > > > > > Happylad > > --------------------------- Newsgroups: comp.os.vxworks Subject: how to set up WDB connection to 3C905C-TX? Date: 27 Aug 2002 23:55:48 -0700 From: kuznetsov_gv@mail.ru (Gennady) Organization: http://groups.google.com/ Message-ID: <551abbdb.0208272255.7bd94c3f@posting.google.com> Hi all. Excuse me for a little dumb question but I really need to know... I try to set up WDB connection to PC with 3C905C-TX ethernet card. Well, I included "WDB END driver connection" in Tornado project. Then it looks I have to include ELT netif BSD driver. So there are two questions there. First one is that in manual of el3c90xEND driver this ethernet card's modification is not listed as supported. Is it critical or 3C905B or 3C905 are the same? The second one is that the parameters I have to set up for ELT are not clear to me (especially NRF_ELT). Do you know any manual references about it? Well, anyway after I set it up and I see that muxDevLoad calls in procedures usrEndLibInit (usrEndLib.c) and wdbCommDevInit (wdbEnd.c) are failed, even sysEl3c90xEndLoad had not been called. pDevTbl->endLoadString is empty for these calls, and I don't see where it should be formed. I think that some parameters are not correct but I have no idea which of them. So can anyone give me any clue, manual reference or just an idea? VxWorks 5.4, Tornado 2, Pentium BSP. Thank you. Best regards, Gennady --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Doku for vxPpcLib Date: 28 Aug 2002 01:48:05 -0700 From: joerg.krein@ims.fhg.de (=?ISO-8859-1?Q?J=F6rg_Krein?=) Organization: http://groups.google.com/ Message-ID: References: Yes, thank you. But there isnt't a doc for the vxPpcLib and the functions defined therein?! "Leonid Rosenboim" wrote in message news:... > Look aththe end of the VxWorks Programmer's Guide, there is an Appendix per > each architecture supported, listing all architecture specific functions and > implementation details. > > "Jörg Krein" wrote in message > news:adcf93df.0208270709.2802e5c1@posting.google.com... > > Hi, > > where can I find the docu for the vxPpcLib? ..... --------------------------- Newsgroups: comp.os.vxworks Subject: how generate an interrupt with SIMNT? Date: 28 Aug 2002 02:17:44 -0700 From: l.salvemini@infosol.it (Lorenzo Salvemini) Organization: http://groups.google.com/ Message-ID: Hi all, i tried to generate an interrupt with sysBusIntGen(INT_LEV,INT_GEN) but this function always returns ERROR. Which could be the error??? I set the handler in this way: intConnect(INUM_TO_IVEC(INTERRUPT_LEVEL)),(VOIDFUNCPTR)interruptHandler,i)) thanks in advance lorenzo --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Doku for vxPpcLib Date: Wed, 28 Aug 2002 11:35:44 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Yesm you are right, the PowerPC appendig does not document the vxPpcLib module, and it is neother in the Reference manual. It might be in the on-line manual, but if it is not, this is obviously a documentation bug (one of maybe thousands) and should be reported to WindRiver support. "Jörg Krein" wrote in message news:adcf93df.0208280048.6191cea2@posting.google.com... > Yes, thank you. > But there isnt't a doc for the vxPpcLib and the > functions defined therein?! > > "Leonid Rosenboim" wrote in message news:... > > Look aththe end of the VxWorks Programmer's Guide, there is an Appendix per > > each architecture supported, listing all architecture specific functions and > > implementation details. > > > > "Jörg Krein" wrote in message > > news:adcf93df.0208270709.2802e5c1@posting.google.com... > > > Hi, > > > where can I find the docu for the vxPpcLib? > ..... --------------------------- Newsgroups: comp.os.vxworks Subject: Re: how generate an interrupt with SIMNT? Date: Wed, 28 Aug 2002 11:37:12 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: The sysBusIntGen() function defined specifically for VME boards, and I would not expect it to work on anything else. "Lorenzo Salvemini" wrote in message news:e2e5ad95.0208280117.2b6dfcca@posting.google.com... > Hi all, > i tried to generate an interrupt with sysBusIntGen(INT_LEV,INT_GEN) > but this function always returns ERROR. Which could be the error??? > I set the handler in this way: > intConnect(INUM_TO_IVEC(INTERRUPT_LEVEL)),(VOIDFUNCPTR)interruptHandler,i)) > > thanks in advance lorenzo --------------------------- Newsgroups: comp.os.vxworks Subject: Re: how generate an interrupt with SIMNT? Date: Wed, 28 Aug 2002 06:34:18 -0400 From: joe durusau Organization: WEBUSENET.com Message-ID: <3D6CA72A.BF31EC99@bellsouth.net> References: Leonid Rosenboim wrote: > The sysBusIntGen() function defined specifically for VME boards, > and I would not expect it to work on anything else. > > "Lorenzo Salvemini" wrote in message > news:e2e5ad95.0208280117.2b6dfcca@posting.google.com... > > Hi all, > > i tried to generate an interrupt with sysBusIntGen(INT_LEV,INT_GEN) > > but this function always returns ERROR. Which could be the error??? > > I set the handler in this way: > > > intConnect(INUM_TO_IVEC(INTERRUPT_LEVEL)),(VOIDFUNCPTR)interruptHandler,i)) > > > > thanks in advance lorenzo True enough, if it's not a VME board, it won't work. If it is a VME board, store the global variable errno _immediately_ after the function call. That will tell you what's wrong. Read the errno documentation to find out how it works. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: dofference between vxWorks Image and vxWorks.st image Date: 28 Aug 2002 04:40:07 -0700 From: forvxworks@rediffmail.com (Anand) Organization: http://groups.google.com/ Message-ID: HI All, what are the differences between vxWorks Image and vxWorks.st , the standalone image?? Regards. Anand. --------------------------- Newsgroups: comp.os.vxworks Subject: NFS server throughput Date: Wed, 28 Aug 2002 14:12:07 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: I guess now it's my turn to ask a question for once: If anyone is running the VxWorks NFS server on top of DosFs 2.0, where the disk storage is either IDE or SCSI hard drives (not Flash disks of any kind), I would appreciate if you could share your experience with respect to the throughput and performance of this configuration. Please specify what platform is your NFS client runs on too. Flash drives are not interesting to me simply because their inherent slow throughput. Thanks, Leonid --------------------------- Newsgroups: comp.os.vxworks Subject: Re: dofference between vxWorks Image and vxWorks.st image Date: Wed, 28 Aug 2002 09:31:31 -0400 From: Joe Durusau Organization: Lockheed Martin Corporation Message-ID: <3D6CD0B3.77B3AD58@lmco.com> References: Anand wrote: > > HI All, > what are the differences between vxWorks Image and vxWorks.st , > the standalone image?? > > Regards. > Anand. For a really good discussion, see the programmer's guide, chapter 8. In general, vxWorks uses a host shell and symbol table, while vxWorks.st has a target shell and includes a symbol table internally. It is , of course, possible to configure the shell out if you don't want it. IIRC, the .st version has networking in it, but doesn't initialize the net unless you tell it to do so. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: where can i find source of real drivers? Date: 28 Aug 2002 08:29:19 -0700 From: l.salvemini@infosol.it (Lorenzo Salvemini) Organization: http://groups.google.com/ Message-ID: Does anybody know where can i find source of real embedded drivers? thanks a lot Lorenzo --------------------------- Newsgroups: comp.os.vxworks Subject: is anybody working on ixp1200 Date: Wed, 28 Aug 2002 19:32:43 +0200 From: "tomer" Organization: The Israel Inter-University Computation Center Message-ID: hi to all. i am a new user of the ixp1200 supported by vxworks os. i will be thankfull for any examples, manuals etc. tomer --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Need idea for implementing watchdog Date: 28 Aug 2002 10:41:41 -0700 From: nedren@yahoo.com (Ned Ren) Organization: http://groups.google.com/ Message-ID: <98c05101.0208280941.5468b618@posting.google.com> References: <98c05101.0208211331.4b320cea@posting.google.com> Thank you all for giving me so much ideas!! One more question, how do I make all tasks to send a message to the watchdog (running low priority) every 100 msec? Thanks!! Johan Borkhuis wrote in message news:... > Keith Arner wrote: > > > On Thu, 22 Aug 2002, Johan Borkhuis wrote: > > > >> nedren@yahoo.com (Ned Ren) wrote: > >> > >> > solution is find the bug, but... :) The definition of trouble is: > >> > > >> > 1. a task is suspended (due to invalid pointer access, or some > >> > other bad things) > >> > 2. a task is running wild, in a tight loop or something > >> > 3. a task is no long existing > >> > >> You need a watchdog task that monitors the status of the other tasks. > >> Also all monitored tasks need to signal to the watchdog task that > >> they are still alive, for example every 100 msec. If a task misses 2 > >> triggers it is considered "wild", and you can reset the system. > >> Another way to detect if a task is running wild is to have one task > >> at a low priotity siginalling the watchdog task. If this iask does > >> not get any time other tasks must be eating all the CPU time, and are > >> probably "wild". > > > > Another trouble spot which you might want to monitor for is deadlock > > between tasks. For this the watchdog approach described above would > > work well, but the low priority task would not help. > > > > Additionally, I would point out that you probably want to implement > > the watchdog as a VxWorks watchdog timer (wdCreate()), rather than > > using a high priority task. This will help you catch things like: > > > > taskLock(); > > while (1) { ... } > > The problem with doing things using wdCreate is that you are running in > an interrupt, and that yoiu are limited in the number of functions you > can use. The only thing you could use the wd function for is to perform > the real reset: create a wd function that does the reset, and start it > with wdCreate. When all tasks are still alive and kicking the watchdog > task restart the wd-counter, otherwise this counter will expire and the > system will be reset. > > Kind regards, > Johan --------------------------- Newsgroups: comp.os.vxworks Subject: Re: dofference between vxWorks Image and vxWorks.st image Date: 28 Aug 2002 12:58:05 -0700 From: arohb@cc.usu.edu (Aroh Barjatya) Organization: http://groups.google.com/ Message-ID: References: vxWorks.st image is a "standalone" image... you can configure a target shell with in the image and hence have a development system running just out of a hyperterminal(if the target shell runs on a RS232 port) or telnet(if the standalone image has networking enabled and uses it). vxWorks image basically "needs" another computer to be hooked with the CPU board... it then runs the tornado host based shell on the computer and not on the CPU board that... Anand... no hard feelings, but please RTFM before you ask questions like this.... there are pages and pages of info that describes this in the programmers guide and other documentation. Aroh forvxworks@rediffmail.com (Anand) wrote in message news:... > HI All, > what are the differences between vxWorks Image and vxWorks.st , > the standalone image?? > > Regards. > Anand. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Need idea for implementing watchdog Date: Wed, 28 Aug 2002 20:17:11 +0000 (UTC) From: Johan Borkhuis Organization: Agere Systems Message-ID: References: <98c05101.0208211331.4b320cea@posting.google.com> <98c05101.0208280941.5468b618@posting.google.com> nedren@yahoo.com (Ned Ren) wrote: > Thank you all for giving me so much ideas!! > > One more question, how do I make all tasks to send a message to the > watchdog (running low priority) every 100 msec? That is something you have to build into your tasks. If you have a task that waits for an event make sure you have a timeout of less than 100 msec and check if the timeout occurred. Send the trigger, check if the timeout occurs and handle the event. Also it is no problem if you send more often, so you can put more trigger moments in your code. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: absolute variable locations for GNUC Date: 28 Aug 2002 15:01:31 -0700 From: pwiebe@bellsouth.net (Philip Wiebe) Organization: http://groups.google.com/ Message-ID: <7dc75e7f.0208281401.656432eb@posting.google.com> I need to define a variable to an absolute location. I haven't found a way to do this using the GNU compiler. I know for the linker you can define sections; however, for downloadable apps it doesn't call the linker. Philip Wiebe --------------------------- Newsgroups: comp.os.vxworks Subject: Re: absolute variable locations for GNUC Date: Wed, 28 Aug 2002 18:07:00 -0400 From: joe durusau Organization: WEBUSENET.com Message-ID: <3D6D4984.2C265377@bellsouth.net> References: <7dc75e7f.0208281401.656432eb@posting.google.com> Philip Wiebe wrote: > I need to define a variable to an absolute location. I haven't found > a way to do this using the GNU compiler. I know for the linker you > can define sections; however, for downloadable apps it doesn't call > the linker. > > Philip Wiebe If you mean in the sense that you can define the hardware address of a device, just create a pointer of the appropriate type and set the value. Note that this is fraught with error for true variables, because the memory in question may not really exist. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Re: localtime () Date: Wed, 28 Aug 2002 17:25:49 -0500 From: Chris Richards Organization: Nortel Message-ID: <3D6D4DED.5060403@nortelNADASPAMnetworks.com> References: I had a similar issue - however, I needed both local time and UTC time + the offset to local time. I.e. local time = 11:20:30, UTC = 08:20:30 + 03:00:00 I never found a good way to get it - without doing all the calculations myself. The time source code does have a UTC and UTC offset fields in the struct typedefs, but these are never used by VxWorks. Chris. Peter Moseley wrote: > I have a feeling that this is a bit of a basic question but I can't find the > answer. > > Whenever my program calls "localtime ()" I always get the current time > returned as GMT, not as my local time. How do I tell vxWorks to return the > latter? > > Thanks > > Peter Moseley > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Target shell symbol table disapperance Date: 28 Aug 2002 15:55:20 -0700 From: arohb@cc.usu.edu (Aroh Barjatya) Organization: http://groups.google.com/ Message-ID: HI All, I am trying to make a standalone vxWorks image (vxWorks.st). My system has two options to boot: 1> TSFS(target server file system) over RS232 port 0 2> using on the CPU board flash...that should give me target shell running on port 1...and still be able to run tornado host based shell off port 0 When I make a simple vxWorks.st image...simple meaning 'with no user added modules'--just the operating system, then the image once burnt in the flash comes up with target shell on port 1 and the target shell has its symbols visible to itself. But when I add my own modules( for installing additional drivers for other components in the system) and rebuild vxWorks.st.... and boot after burning the image in flash...the target shell gives "undefined symbol" for anything typed on the terminal. The user application does get executed though, as I can hook up a tornado host based shell on port 0 of my CPU board and see all the drivers installed and devices created. The weird thing is that if I boot the CPU using the same image (with user modules linked in) using the TSFS boot the target shell comes up on port 1 with the user startup application executing and the target shell also has all symbols visible to it. Its just that when this image is turned binary (using elftobin utility) does the target shell forget the symbols. Can anybody help me on this ? What possibly could be going wrong. please send me an email along with posting here on the newsgroup as that is faster. thanks Aroh --------------------------- Newsgroups: comp.os.vxworks Subject: help required for developing device driver for pcicard Date: 28 Aug 2002 21:18:20 -0700 From: vxnerd@rediffmail.com (subbarayan) Organization: http://groups.google.com/ Message-ID: dear mentors, Can anyone jus tell me how to go abt proceeding the development of device driver for pci card for x86 processor with standard pc architecture?I use tornadoII and windows development host and a stand alone target. Further can anyone tell me when will be this pciautoconfiglib file called?will it be called automatically as a startup routine or need it to be called explicitly to enable pci scanning? What exactly is the difference between the files pci autoconfiglib.c and pciconfiglib.c?which file should be used for which sort of appln?and can these library files directly compiled and used or should we call it from some appln to invoke and access it? And sorry for my ignorance and unknowing things.One thing i found in the manual for pci support on vxworks says we need to include the macro INCLUDE_PCI for enabling pci support.Where exactly should this macro be included and where exactly is this macro defined?How to include this macro and whats the exact location where this macro has to be included?And whats the role of bios with respect to vxworks in detecting pci card and pci bus?It would be helpful if some one could tell me what would be printed on screen if pci support is enabled and makes me confirm that my code has started working with pci support. Sorry if my questions look amature,but i am helpless as i am new to this sort of coding. Advance thanks to persons replying me. --------------------------- Newsgroups: comp.os.vxworks Subject: First time to here Date: Thu, 29 Aug 2002 14:13:34 +0800 From: "Seraph Ma" Organization: Bentium Ltd. (CN99) Message-ID: It's so great that I finally get the NNTP server name for this newsgroup. Great --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Use Parallel Port instead of the required Serial Port ? Need Null Modem line ? Date: 24 Aug 2002 10:13:33 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0208240913.756150a2@posting.google.com> References: Hello, patrickyu_ca@yahoo.com (patrick) wrote in message news:... > Hi, all > > I want to use the notebook as the host. But I do not have the serial > port except the parallel port. Can I use Parallel port instead of the > required serial port ? If can, need any special configuration ? No, and the fact that you had to ask scares me. Please tell us what you are designing with VxWorks so we know to avoid buying one ;-) Serial and parallel ports are very different, and as an embedded software engineer I would have hoped that you knew that... Now, if you have a USB connection on your laptop, try getting a USB serial port adapter (there were a few of these around in the early days of USB, I suspect that they still exist). That will give you a standard serial port on your laptop that you can use for the connection to the target. A final thought: most of the time you won't need the serial connection. Once you have set the bootline, the target will boot happily without it and you can use the network connection for all communication with the target. So, you could simply configure the boot line by plugging the target's console serial port into another PC initially. Once that's done, you can forget the serial port most of the time - you can even change the bootline from the host shell (see bootChange), just don't make a mistake otherwise you'll have to go back to the serial connection to correct it. > Need I use the Null Modem line or common line is ok for Vxworks serial > connection? What is "common line" meant to mean? You can use a three wire serial connection as long as your laptop can be instructed to ignore the hardware handshaking lines (set it to software flow control). HTH, John... --------------------------- Newsgroups: comp.os.vxworks Subject: Re: difference between vxWorks Image and vxWorks.st image Date: 28 Aug 2002 23:39:13 -0700 From: forvxworks@rediffmail.com (Anand) Organization: http://groups.google.com/ Message-ID: References: HI Aroh, Thank you for the reply. yes i have read the programmers' guide, but i feel that we can learn "more" from people who have years of experience in a field in which i am a starter. i have seen that i get to learn "something more",other than the info given in the reference guides, from good people like you, all over the world, who are ready to answer even the silliest questions. Regards. Anand. arohb@cc.usu.edu (Aroh Barjatya) wrote in message news:... > vxWorks.st image is a "standalone" image... you can configure a target > shell with in the image and hence have a development system running > just out of a hyperterminal(if the target shell runs on a RS232 port) > or telnet(if the standalone image has networking enabled and uses it). > > vxWorks image basically "needs" another computer to be hooked with the > CPU board... it then runs the tornado host based shell on the computer > and not on the CPU board that... > > Anand... no hard feelings, but please RTFM before you ask questions > like this.... there are pages and pages of info that describes this in > the programmers guide and other documentation. > > Aroh > > > > forvxworks@rediffmail.com (Anand) wrote in message news:... > > HI All, > > what are the differences between vxWorks Image and vxWorks.st , > > the standalone image?? > > > > Regards. > > Anand. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Target shell symbol table disapperance Date: Thu, 29 Aug 2002 14:41:24 +0800 From: "Seraph Ma" Organization: Bentium Ltd. (CN99) Message-ID: References: hey Perhaps you should include the item of 'symbol table componets->synchorize host and target symbol table ' in you project when you make a vxWorks image. "Aroh Barjatya" wrote in message news:ecdec0b9.0208281455.725eed70@posting.google.com... > HI All, > > I am trying to make a standalone vxWorks image (vxWorks.st). My system > has two options to boot: > 1> TSFS(target server file system) over RS232 port 0 > 2> using on the CPU board flash...that should give me target shell > running on port 1...and still be able to run tornado host based shell > off port 0 > > When I make a simple vxWorks.st image...simple meaning 'with no user > added modules'--just the operating system, then the image once burnt > in the flash comes up with target shell on port 1 and the target shell > has its symbols visible to itself. > > But when I add my own modules( for installing additional drivers for > other components in the system) and rebuild vxWorks.st.... and boot > after burning the image in flash...the target shell gives "undefined > symbol" for anything typed on the terminal. The user application does > get executed though, as I can hook up a tornado host based shell on > port 0 of my CPU board and see all the drivers installed and devices > created. > > The weird thing is that if I boot the CPU using the same image (with > user modules linked in) using the TSFS boot the target shell comes up > on port 1 with the user startup application executing and the target > shell also has all symbols visible to it. Its just that when this > image is turned binary (using elftobin utility) does the target shell > forget the symbols. > > Can anybody help me on this ? What possibly could be going wrong. > > please send me an email along with posting here on the newsgroup as > that is faster. > > thanks > > Aroh --------------------------- Newsgroups: comp.os.vxworks Subject: Re: booting from tffs Date: Thu, 29 Aug 2002 08:39:13 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: Reply-To: "Michael Lawnick" Sender: mlawnick@pd9524356.dip.t-dialin.net Hi Chad, sounds that you are driving your TFFS with different versions of dosFs on bootrom/vxWorks/flash. Decide for one, dosFs or dosFs2, rebuild both, bootrom *and* vxWorks, reboot and reformat your flash, then everything should work. Regards, Michael "Chad" schrieb im Newsbeitrag news:rbdb9.225121$f05.12388023@news1.calgary.shaw.ca... > I found a post from a while back regarding a possible conflict between > dosfs2 and tffs where they disagree about where sector 0 should point > (posted by Andray Kaganovsky) and that a possible solution would be to > remove the call to dosFsDevCreate as used by dosfs2 and use the (hopefully) > backward compatible dosFsDevInit. This is promising so far, I made these > changes, reformatted the flash and put the updated copy of vxWorks on it. > Now I get this boot scenario: > > boot device : tffs=0,0 > unit number : 0 > processor number : 0 > host name : rnd > file name : /flash/vxWorks > inet on ethernet (e) : 192.0.1.90 > host inet (h) : 192.0.1.58 > flags (f) : 0x8 > other (o) : cpm > > tffsLoad: Tffs initialized... > tffsLoad: Attaching to TFFS... > BOOT: Mounting TFFS device at: /flash/vxWorks > CFISCS-PGC: entering CFISCS identification routine. > CFISCS-PGC: reading CFI for word mode. > CFISCS-PGC: identified 16-bit CFISCS. > Eraseable block size: 0x20000 > Chip size: 0x2000000 > Number of chips: 1 > Interleaving: 2 > BOOT: tffsDevCreate(0,0) successful. > BOOT: calling cbioDevVerify... > usrTffsConfig(0,0,/flash/vxWorks) succeeded. > done. > Loading /flash/vxWorks...2669040 > Starting at 0x100000... > > Failed to attach to device tffs=0,0Attaching interface lo0...done > ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ > > Adding 7413 symbols for standalone. > > > VxWorks > > Copyright 1984-1998 Wind River Systems, Inc. > > CPU: MPC823 - Thompson Technology PGC400 p0 > VxWorks: 5.4.2 > BSP version: 1.2/26 > Creation date: Aug 28 2002 > WDB: Ready. > > Mounting TFFS device at: /flash > CFISCS-PGC: entering CFISCS identification routine. > CFISCS-PGC: reading CFI for word mode. > CFISCS-PGC: identified 16-bit CFISCS. > Eraseable block size: 0x20000 > Chip size: 0x2000000 > Number of chips: 1 > Interleaving: 2 > tffsDevCreate(0,0) successful. > calling cbioDevVerify... > Flash File System Mounted > 0xfffdf8 (): task deadCan't attach unknown device tffs=0,0 (unit 0). > -> > ******************************* > It looks like it actually booted from the flash image, although had a hiccup > getting tffs started from the downloaded image. This image worked with the > file system when I obtained it from the network however. > > Can anybody shed some light on what I've done wrong? > > > > > "Chad" wrote in message > news:oscb9.222394$Ag2.11467798@news2.calgary.shaw.ca... > > I'm running a ppc860 with tornado 2.02 and can't seem to boot from my > intel > > strata flash. I am able to do the following: > > > > ************** > > > > > > -> tffsRawio 0,3,0,128,0 > > value = 0 = 0x0 > > -> tffsDevFormat 0,0 > > CFISCS-PGC: entering CFISCS identification routine. > > CFISCS-PGC: reading CFI for word mode. > > CFISCS-PGC: identified 16-bit CFISCS. > > Eraseable block size: 0x20000 > > Chip size: 0x2000000 > > Number of chips: 1 > > Interleaving: 2 > > CFISCS-PGC: entering CFISCS identification routine. > > CFISCS-PGC: reading CFI for word mode. > > CFISCS-PGC: identified 16-bit CFISCS. > > Eraseable block size: 0x20000 > > Chip size: 0x2000000 > > Number of chips: 1 > > Interleaving: 2 > > CFISCS-PGC: entering CFISCS identification routine. > > CFISCS-PGC: reading CFI for word mode. > > CFISCS-PGC: identified 16-bit CFISCS. > > Eraseable block size: 0x20000 > > Chip size: 0x2000000 > > Number of chips: 1 > > Interleaving: 2 > > value = 0 = 0x0 > > > > ************** > > > > I can ftp my vxWorks image to the flash and back to my system and verify > > that it was not corrupted, and boot from the network and mount/view the > > device with no apparent problems. When I boot from flash I get this: > > > > ************** > > > > boot device : tffs=0,0 > > unit number : 0 > > processor number : 0 > > host name : rnd > > file name : /flash/vxWorks > > inet on ethernet (e) : 192.0.1.90 > > host inet (h) : 192.0.1.58 > > flags (f) : 0x8 > > other (o) : cpm > > > > tffsLoad: Tffs initialized... > > tffsLoad: Attaching to TFFS... > > Mounting TFFS device at: /flash/vxWorks > > CFISCS-PGC: entering CFISCS identification routine. > > CFISCS-PGC: reading CFI for word mode. > > CFISCS-PGC: identified 16-bit CFISCS. > > Eraseable block size: 0x20000 > > Chip size: 0x2000000 > > Number of chips: 1 > > Interleaving: 2 > > tffsDevCreate(0,0) successful. > > src/config calling cbioDevVerify... > > dosFS device created. > > usrTffsConfig(0,0,/flash/vxWorks) succeeded. > > done. > > /flash/vxWorks/ - disk check in progress ... > > WARNING : dosChkLib : system clock is being set to FRI JAN 01 00:04:04 > 1999 > > Value obtained from file system referenced by volume descriptor pointer: > > 0xffdf7 > > 0 > > The old setting was THU JAN 01 00:00:01 1970 > > Accepted system dates are greater than FRI JAN 01 00:00:00 1999 > > /flash/vxWorks/ - Volume is OK > > > > total # of clusters: 15,806 > > # of free clusters: 9,371 > > # of bad clusters: 0 > > total free space: 18,742 Kb > > max contiguous free space: 19,191,808 bytes > > # of files: 2 > > # of folders: 0 > > total bytes in files: 12,862 Kb > > # of lost chains: 0 > > total bytes in lost chains: 0 > > Erroneous header read > > error loading file: status = 0x610001. > > Error loading file: errno = 0x610001. > > > > ************** > > > > I'm not sure what the Erroneous header read message is indicating, is it > > simply that my file is not found? /flash/vxWorks is there and visible if > I > > ftp or boot from network and afaik is a valid image to boot from flash > (the > > same image I boot from network) > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: waitforever Date: Thu, 29 Aug 2002 15:36:06 +0800 From: "Seraph Ma" Organization: Bentium Ltd. (CN99) Message-ID: References: <39c3b976.0208260644.374f746f@posting.google.com> Leonid But I do not find the pointer for delayQ in WIND_TCB data strucutrue defintion. Why????? I am using vxWorks 5.4. "Leonid Rosenboim" wrote in message news:newscache$9oig1h$qjf$1@lnews.actcom.co.il... > The "Wait forever" case is actually simplier then the case of pending > on a semaphore with a specific timeout. > > When there is no timeout (i.e. wait forever), the task TCB will be put > on the semaphore pendQ. > When there is a timeout, the task will be put into semaphore's > pendQ, and also on the system clock delayQ. For that reason, > the TCB has a separate pointer-pair for pendQ, and a separate > pointer-pair for delayQ. > > When the task becomes ready again, it will be kicked off both queus > at the same time, and added to the readyQ, using the same pointers > used for the pendQ. > > BTW, I used to teach this stuff, as part of the standard VxWorks workshop > training. > > "senthil" wrote in message > news:39c3b976.0208260644.374f746f@posting.google.com... > > Hi, > > How are the statements below implemented by vxworks.. > > semTake(semid, WAITFOREVER); or semTake(semid,-1) -- both the > > statements waits on a semaphore forever. How is this implemented > > considering it in hardware point of view or kernel point of view. > > regards > > senthil > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: booting from tffs Date: Thu, 29 Aug 2002 09:55:41 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Chad, block 0 isn't your problem, the chkdsk goes through OK, the file system is perfectly OK. Here is your problem: According to bootline, the boot FILE name is /flash/VxWorks, but your init routine mounts the whole DEVICE as /flash/vxWorks, so naturally opening "/flash/VxWorks" yields a DIRECTORY as opposed to a FILE, thus the loading error. You need to make sure that the device is mounted with the device name being "/flash/" with or without the trailing slash, and all will work. No need to fiddle with dosFsDevInit either. - - Leonid "Chad" wrote in message news:oscb9.222394$Ag2.11467798@news2.calgary.shaw.ca... > I'm running a ppc860 with tornado 2.02 and can't seem to boot from my intel > strata flash. I am able to do the following: > > ************** > > > -> tffsRawio 0,3,0,128,0 > value = 0 = 0x0 > -> tffsDevFormat 0,0 > CFISCS-PGC: entering CFISCS identification routine. > CFISCS-PGC: reading CFI for word mode. > CFISCS-PGC: identified 16-bit CFISCS. > Eraseable block size: 0x20000 > Chip size: 0x2000000 > Number of chips: 1 > Interleaving: 2 > CFISCS-PGC: entering CFISCS identification routine. > CFISCS-PGC: reading CFI for word mode. > CFISCS-PGC: identified 16-bit CFISCS. > Eraseable block size: 0x20000 > Chip size: 0x2000000 > Number of chips: 1 > Interleaving: 2 > CFISCS-PGC: entering CFISCS identification routine. > CFISCS-PGC: reading CFI for word mode. > CFISCS-PGC: identified 16-bit CFISCS. > Eraseable block size: 0x20000 > Chip size: 0x2000000 > Number of chips: 1 > Interleaving: 2 > value = 0 = 0x0 > > ************** > > I can ftp my vxWorks image to the flash and back to my system and verify > that it was not corrupted, and boot from the network and mount/view the > device with no apparent problems. When I boot from flash I get this: > > ************** > > boot device : tffs=0,0 > unit number : 0 > processor number : 0 > host name : rnd > file name : /flash/vxWorks > inet on ethernet (e) : 192.0.1.90 > host inet (h) : 192.0.1.58 > flags (f) : 0x8 > other (o) : cpm > > tffsLoad: Tffs initialized... > tffsLoad: Attaching to TFFS... > Mounting TFFS device at: /flash/vxWorks > CFISCS-PGC: entering CFISCS identification routine. > CFISCS-PGC: reading CFI for word mode. > CFISCS-PGC: identified 16-bit CFISCS. > Eraseable block size: 0x20000 > Chip size: 0x2000000 > Number of chips: 1 > Interleaving: 2 > tffsDevCreate(0,0) successful. > src/config calling cbioDevVerify... > dosFS device created. > usrTffsConfig(0,0,/flash/vxWorks) succeeded. > done. > /flash/vxWorks/ - disk check in progress ... > WARNING : dosChkLib : system clock is being set to FRI JAN 01 00:04:04 1999 > Value obtained from file system referenced by volume descriptor pointer: > 0xffdf7 > 0 > The old setting was THU JAN 01 00:00:01 1970 > Accepted system dates are greater than FRI JAN 01 00:00:00 1999 > /flash/vxWorks/ - Volume is OK > > total # of clusters: 15,806 > # of free clusters: 9,371 > # of bad clusters: 0 > total free space: 18,742 Kb > max contiguous free space: 19,191,808 bytes > # of files: 2 > # of folders: 0 > total bytes in files: 12,862 Kb > # of lost chains: 0 > total bytes in lost chains: 0 > Erroneous header read > error loading file: status = 0x610001. > Error loading file: errno = 0x610001. > > ************** > > I'm not sure what the Erroneous header read message is indicating, is it > simply that my file is not found? /flash/vxWorks is there and visible if I > ftp or boot from network and afaik is a valid image to boot from flash (the > same image I boot from network) > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: waitforever Date: Thu, 29 Aug 2002 10:08:45 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <39c3b976.0208260644.374f746f@posting.google.com> That's because I dont have access to VxWorks files at the moment, and I spelled out the field name from memory (I am human you know ;-) ) But this should be easy: The TCB begins with 3 ptr-pairs: one is for rdyQ or pendQ (these two never used at the same time so one hook does it), second is for delayQ (used for timeouts and taskDelay, and can be concurrent to PEND) and the third is for the auxiliary task list, i.e. a list of all valid TCBs. The last queue is not used by kernel, only for displaying task list, i.e. the famous "i" command. TCB, unlike other WIND object, does not begin with objCore You should be able to connect the dots from here. "Seraph Ma" wrote in message news:akkj17$c14$1@debian.bentium.com... > Leonid > But I do not find the pointer for delayQ in WIND_TCB data strucutrue > defintion. > > Why????? > I am using vxWorks 5.4. > > "Leonid Rosenboim" wrote in message > news:newscache$9oig1h$qjf$1@lnews.actcom.co.il... > > The "Wait forever" case is actually simplier then the case of pending > > on a semaphore with a specific timeout. > > > > When there is no timeout (i.e. wait forever), the task TCB will be put > > on the semaphore pendQ. > > When there is a timeout, the task will be put into semaphore's > > pendQ, and also on the system clock delayQ. For that reason, > > the TCB has a separate pointer-pair for pendQ, and a separate > > pointer-pair for delayQ. > > > > When the task becomes ready again, it will be kicked off both queus > > at the same time, and added to the readyQ, using the same pointers > > used for the pendQ. > > > > BTW, I used to teach this stuff, as part of the standard VxWorks workshop > > training. > > > > "senthil" wrote in message > > news:39c3b976.0208260644.374f746f@posting.google.com... > > > Hi, > > > How are the statements below implemented by vxworks.. > > > semTake(semid, WAITFOREVER); or semTake(semid,-1) -- both the > > > statements waits on a semaphore forever. How is this implemented > > > considering it in hardware point of view or kernel point of view. > > > regards > > > senthil > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: 10Base-T and 100Base-T problems!!! Date: 29 Aug 2002 02:16:37 -0700 From: oe_zhou@sohu.com (oe_zhou) Organization: http://groups.google.com/ Message-ID: <758a30ff.0208290116.7977c5d@posting.google.com> I am using ppc860 and the OS is VxWorks. Now i have implemented the 10Base-T through SCC1. And the board can boot up in debug mode. But when the board is linked to a 10/100Base-T self-adaptive adaptor the 10Base-T ethernet cann't be connected. What's the matter? --------------------------- Newsgroups: comp.os.vxworks Subject: where can i get the BSP tool Kit? Date: 29 Aug 2002 02:17:55 -0700 From: oe_zhou@sohu.com (oe_zhou) Organization: http://groups.google.com/ Message-ID: <758a30ff.0208290117.654e7f6c@posting.google.com> Hi. I am using VxWorks. But where can i get the BSP tool kit? Thanks! --------------------------- Newsgroups: comp.os.vxworks Subject: test Date: Thu, 29 Aug 2002 17:22:20 +0800 From: "wenyong" Organization: Bentium Ltd. (CN99) Message-ID: - -- Wenyong Xia Ext 4376 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: do not want vxworks shell to attach to console by default Date: Thu, 29 Aug 2002 09:53:18 +0000 (UTC) From: Johan Borkhuis Organization: Agere Systems Message-ID: References: "Jackie Bhadange" wrote: > Is there a way to disable the shell from attaching to the console > port? We want the shell but we want it accessible only through telnet > (and not through the console port). > Any ideas? Assign a non existing port to the console. P.S. Try to post in non-html format. It looks quite messy in a text-only browser, and this is still the most used type. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: where can i get the BSP tool Kit? Date: Thu, 29 Aug 2002 09:57:18 +0000 (UTC) From: Johan Borkhuis Organization: Agere Systems Message-ID: References: <758a30ff.0208290117.654e7f6c@posting.google.com> oe_zhou@sohu.com (oe_zhou) wrote: > Hi. I am using VxWorks. > But where can i get the BSP tool kit? From WindRiver. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: "Virtual" end driver Date: 29 Aug 2002 03:00:06 -0700 From: gunter.ritter@tenovis.com (Gunter Ritter) Organization: http://groups.google.com/ Message-ID: References: Hello, I think that I have solved the problem. After adding the functions for multicast and polling in the net functions table it is now possible to set an ip address for the interface. ifShow() looks good now. Thanks for your help. happyhappy779@hotmail.com (Satish G) wrote in message news:... > Hi, > > Would you mind posting the route table after ifAddrSet() call(routeShow/mrouteShow)? > > > Gunter.Ritter@tenovis.com (Gunter Ritter) wrote in message news:... > > Hi, > > > > the ip adresses are totaly different. One is 192.1.2.141, the other I > > tried to set to 143.123.19.141. The subnetmask is set to > > 255.255.255.0. > > > > > > happyhappy779@hotmail.com (Satish G) wrote in message news:... > > > Hi, > > > > > > You have not given the IP addresses used for the interfaces, so I > > > would suspect that the IPaddress/netmask combination is resulting in > > > same network for both interfaces. > > > > > > Satish G > > > > > > Gunter.Ritter@tenovis.com (Gunter Ritter) wrote in message news:... > > > > Hello VxWorkers, > > > > > > > > I have written a "virtual" end driver as a secondary end driver for > > > > our system. > > > > "Virtual" means that this end driver has no "physical" ethernet > > > > hardware. > > > > > > > > The primary end driver is working with ethernet hardware. > > > > > > > > I need this secondary end driver only to exchange ip packets with the > > > > TCP/IP stack of vxworks and let vxworks do the routing between the two > > > > networks. > > > > > > > > I have a problem when I try to assign an ip address. > > > > muxDevLoad(), muxBind(), ipAttach() and ifMaskSet() work fine. > > > > ifShow() shows me the settings of the secondary driver. > > > > But calling ifAddrSet() with an unique ip address causes the network > > > > functionality of vxworks to crash. > > > > > > > > Even "ping" does no more work with the primary end driver after > > > > calling ifAddrSet(). > > > > > > > > Does anybody can give me a hint whats going wrong with my secondary > > > > end driver? > > > > > > > > Thanks. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: 10Base-T and 100Base-T problems!!! Date: Thu, 29 Aug 2002 12:08:37 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <758a30ff.0208290116.7977c5d@posting.google.com> Zhou, what the hell is "self-adaptive adaptor" ?? how does it "fails to connect" ?? - does the "link" led come up or not ? - - if "link" is ok, does ping with small packets work ? "oe_zhou" wrote in message news:758a30ff.0208290116.7977c5d@posting.google.com... > I am using ppc860 and the OS is VxWorks. > Now i have implemented the 10Base-T through SCC1. > And the board can boot up in debug mode. > But when the board is linked to a 10/100Base-T self-adaptive > adaptor the 10Base-T ethernet cann't be connected. > What's the matter? --------------------------- Newsgroups: comp.os.vxworks Subject: Help: disk cache error Date: Thu, 29 Aug 2002 08:27:47 -0400 From: "R Hilderman" Organization: Nortel Message-ID: Can someone please help explain exactly what the following logMsg means? 0x1f7bd490 (tDcacheUpd): disk cache error: device 1f7bd6b8 block 252 errno c0006, disk removed while writing data, probable data loss The errno 0xc0006 means S_ioLib_DISK_NOT_PRESENT. Does the error message mean that tDcacheUpd has removed the disk while writing data or does it mean it mearly timed out while writing to the disk (or something similar). I would like to restart the system if the disk is removed, as the disk is our flash file system. Does anyone know how I would bind in for this? Any help would be appreciated. Thanks, Randy --------------------------- Newsgroups: comp.os.vxworks Subject: Re: absolute variable locations for GNUC Date: 29 Aug 2002 05:41:39 -0700 From: pwiebe@bellsouth.net (Philip Wiebe) Organization: http://groups.google.com/ Message-ID: <7dc75e7f.0208290441.ec752ad@posting.google.com> References: <7dc75e7f.0208281401.656432eb@posting.google.com> <3D6D4984.2C265377@bellsouth.net> joe durusau wrote in message news:<3D6D4984.2C265377@bellsouth.net>... > Philip Wiebe wrote: > > > I need to define a variable to an absolute location. I haven't found > > a way to do this using the GNU compiler. I know for the linker you > > can define sections; however, for downloadable apps it doesn't call > > the linker. > > > > Philip Wiebe > > If you mean in the sense that you can define the hardware address of > a device, just create a pointer of the appropriate type and set the > value. Note that this is fraught with error for true variables, because > the memory in question may not really exist. > > Speaking only for myself, > > Joe Durusau Unfortunately I don't have the luxury to change all of the existing code that uses the variable. We were using the diab compiler with nucleus before and so we could define a section using #pragma and put the section at the desired location. --------------------------- Newsgroups: comp.os.vxworks Subject: interrupt with VxSim Date: 29 Aug 2002 05:48:21 -0700 From: l.salvemini@infosol.it (Lorenzo Salvemini) Organization: http://groups.google.com/ Message-ID: Hi all, from the VxWorks programmer's guide appendice H4 i read that interrupt can be generated inside the simulator without any board attached to the host. I've a problem: how can I signal to VxWorks that the interrupt occurred?? By which system function can I send messages to the simulator??? Best regards Lorenzo --------------------------- Newsgroups: comp.os.vxworks Subject: error with -C -redirectShell Date: 29 Aug 2002 06:07:48 -0700 From: l.salvemini@infosol.it (Lorenzo Salvemini) Organization: http://groups.google.com/ Message-ID: HI all, when i try to redirect the IN/OUT to WindSh by setting -C - -redirectShell inside the command options , i got athe following message error: "The target shell task cannot be located on the target" only the task directly spawned from the shell redirect his out to the shell. What could be the problem?? thanks in advance Lorenzo --------------------------- Newsgroups: comp.os.vxworks Subject: Re: disk cache error Date: Thu, 29 Aug 2002 15:35:24 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Randy, Dcache does not remove the disk, but rather has detected that the disk has been physically removed from its slot. This is designed to support removable devices, e.g. PC cards, Compact Flash, Zip drives etc. Because many of these devices do not explicitly inform the software that a door has been opened and a card is removed, Dcache has to use a combination of techniques to detect this event. If the block 0 of the device has been unexpectedly modified (not by the file system), it will be assumed a card has been chenged. If the block driver returns certain types of error, again, that would mean the device is no longer available, hence it was removed. If your device is not removable, you should set the device "removable" flag to FALSE, thereby disabling these autodetecion mechanisms. If it is in fact a removable device, as soon as a new card has been inserted, any subsequent call to open() will mount the new device and you will be able to access it. The file descriptors that where opened before the exchange took place, will be invalid, and need to be closed, to avoid FD leakage. HTH, - - Leonid "R Hilderman" wrote in message news:akl3nv$8no$1@bcarh8ab.ca.nortel.com... > Can someone please help explain exactly what the following logMsg means? > > 0x1f7bd490 (tDcacheUpd): disk cache error: device 1f7bd6b8 block 252 errno > c0006, disk removed while writing data, probable data loss > > The errno 0xc0006 means S_ioLib_DISK_NOT_PRESENT. > Does the error message mean that tDcacheUpd has removed the disk while > writing data > or does it mean it mearly timed out while writing to the disk (or something > similar). > > I would like to restart the system if the disk is removed, as the disk is > our flash file system. > Does anyone know how I would bind in for this? > > Any help would be appreciated. > > Thanks, > Randy > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Source of mips toolchain Date: 29 Aug 2002 06:54:45 -0700 From: chs@lucent.com (ChengHsin Hsu) Organization: http://groups.google.com/ Message-ID: <8b53eb96.0208290554.2b5583fb@posting.google.com> References: <8b53eb96.0208260539.7ac17e6e@posting.google.com> <8b53eb96.0208270851.43ab2243@posting.google.com> Have submitted a TSR again. Useless....No reply at all. Any idea about how to proceed? Thanks, Bear. chs@lucent.com (ChengHsin Hsu) wrote in message news:<8b53eb96.0208270851.43ab2243@posting.google.com>... > Johan Borkhuis wrote in message news:... > > chs@lucent.com (ChengHsin Hsu) wrote: > > > > > All, > > > > > > Has anyone had the experience to get a copy of toolchain source > > > code from wrs? I have sent several emails via their supports but > > > nobody replies at all....I'm wondering whether this is possible..And > > > interested in how can I do that... > > > Following is my cross gcc's version and configuration. We bought > > > this binary compiler from wrs, but wanna move to Linux as our > > > development environment.... > > > > Yes, no problem. With the 2.1 release it is included on the normal CD if I > > remember correctly, and for the 2.0 you can order the source cd from them > > for free. They have to provide it, as they are using GNU tools (see GPL). > > > > Kind regards, > > Johan > Thanks!!! > I'll try it again. :p > > Regards, > Bear. --------------------------- Newsgroups: comp.os.vxworks Subject: Blocking vs Non-Blocking Sockets Date: 29 Aug 2002 06:59:31 -0700 From: prichtmyer@yahoo.com (Peter Richtmyer) Organization: http://groups.google.com/ Message-ID: <1b585154.0208290559.2937af1f@posting.google.com> My recent experience with sockets is limited to Unix, where we used Blocked reads/writes with a timeout. It worked very well, and was simple to code. Now I am working on a VxWorks system (PowerPC) where the application uses sockets with non-blocking I/O. Can someone tell me the advantages / disadvantages of this wrt VxWorks? It seems like alot more coding with non-blocking. There are quite a few tasks, so task blocking should not the issue. They use loops with delays until the read or write is finished. Is the "timeout" option with Blocked I/O available with VxWorks? thanks, Peter --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Blocking vs Non-Blocking Sockets Date: Thu, 29 Aug 2002 16:12:57 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <1b585154.0208290559.2937af1f@posting.google.com> Sure enough, the normal blocking I/O is available on VxWorks sockets, but then you need to either use select or dedicate a task to each connection. When there are a large number of connection. select can't be used because it is limitted to 254 file descriptors, and it is slow to perform when more then a few sockets are used. Dedicating a task per connection is also not such a great idea, increases overhead of both CPU and memory per connection. I have experienced non-blocking sockets on Solaris and VxWorks, and for many applications this is absolutely the best approach, specifically to bulk transfer and streaming. Only drawback I found is that in VxWorks there is no call similar to FIONREAD that would tell you how much free buiffer space there is in a socket, so you know how much you can write into it before actually writing. - - Leonid "Peter Richtmyer" wrote in message news:1b585154.0208290559.2937af1f@posting.google.com... > My recent experience with sockets is limited to Unix, where > we used Blocked reads/writes with a timeout. It worked > very well, and was simple to code. > > Now I am working on a VxWorks system (PowerPC) where the > application uses sockets with non-blocking I/O. Can > someone tell me the advantages / disadvantages of this > wrt VxWorks? It seems like alot more coding with non-blocking. > There are quite a few tasks, so task blocking should not > the issue. They use loops with delays until the read or write > is finished. > > Is the "timeout" option with Blocked I/O available with VxWorks? > > thanks, > Peter --------------------------- Newsgroups: comp.os.vxworks Subject: Re: trouble in ping with a larger packet on ne2000 Ethernet Date: Sat, 24 Aug 2002 19:20:20 GMT From: "cfk" Organization: Prodigy Internet http://www.prodigy.com Message-ID: References: <7f29c2ad.0208080439.62e501b7@posting.google.com> <7f29c2ad.0208091812.5eadfa30@posting.google.com> <7f29c2ad.0208111739.76239413@posting.google.com> Dear Rover: I believe that MTU is intrinsic to a particular link. In other words, an ethernet link will have a certain maximum MTU that is unchangeable. It typically is around 1500 bytes. Gigabit ethernets may have larger MTU's upon occaision, but they usually default to the same as an ethernet MTU. Some links may have smaller (or larger) MTU's. Certainly, you can change the settings in the driver if you wish. There is a field for the MTU in the END_DEVICE for that purpose. However, if you change it and the hardware doesnt support it, then you get a little deeper into your logical quandry. Perhaps it might help if you described why you are trying to ping with packets greater then the MTU and perhaps we might be able to help you work through your "bigger picture". "rover" wrote in message news:7f29c2ad.0208111739.76239413@posting.google.com... > When I ping another pc(win2000),same thing happend!! > > How can I get and modify MTU? > > thanks advance. > > "cfk" wrote in message news:... > > Let me take a wild stab at this one and suggest that the size of the ping > > packet needs to be less then the smallest MTU between the host you are > > pinging from and the host you are pinging to. I believe MTU is usually 1500 > > on an ethernet link. > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: USB to USB Date: Sat, 24 Aug 2002 19:23:43 GMT From: "cfk" Organization: Prodigy Internet http://www.prodigy.com Message-ID: <33R99.268$pM1.25068686@newssvr14.news.prodigy.com> References: <1028887367.61862@ernani.logica.co.uk> <1028897040.726175@ernani.logica.co.uk> Well, no, it isnt an ethernet device, it is a USB device. It knows about the concepts of URB's that are transfer elements from and to a bulk endpoint. It is possible to make a combination ethernet and usb network driver, but it be necessity will need to have some combinations of USB packet machination and ethernet packet machination. There is an example of such a driver in the Linux sources whose file name is pegasus.c. If you wish to create such a driver, I found last year that pegasus.c was a useful place to start from. "David Cooper" wrote in message news:MPG.17c0c16ca976c63d9896a2@news1.news.adelphia.net... > Sorry, do not know the details of the USB Developers Kit. Anyone else ? > > ...david > > In article <1028897040.726175@ernani.logica.co.uk>, peedeesee@hotmail.com > says... > > Hi David, > > > > Thanks for that, but I already have the USB Kit. What I need to know > > is how these USB to USB devices appear. It would seem that they > > pretend to be an Ethernet connection. > > > > Is this correct? If so, then does the VxWorks USB kit provide suitable > > drivers? I'm still waiting for the hardware, hence I can't test it myself. > > > > > > > > > > "David Cooper" wrote in message > > news:MPG.17bd7fbbb7561de19896a1@news1.news.adelphia.net... > > > Wind River sells a separate USB Developers Kit (both Host and Peripheral > > > stacks) -- USB drivers are NOT contained in the standard vxWorks / > > > Tornado distribution. > > > > > > ...david > > > > > > In article <1028887367.61862@ernani.logica.co.uk>, peedeesee@hotmail.com > > > says... > > > > I've been trying to find a pc/104 processor baord that can run VxWorks > > > > AND act as a USB device (not host!), but with no success. > > > > > > > > My next option is to use a pc/104 SBC with a USB host controller and > > > > then use a 'USB to USB' chip to connect to my host PC. > > > > > > > > The question is, what does a 'USB to USB' chip enumerate as under > > > > VxWorks, i.e. how do I drive it? Does it look like an ethernet comms > > > > device? > > > > > > > > Thanks in advance for any advice/experience. > > > > > > > > -- > +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ > ô¿ô Genealogy Utilities & Family Research > Researching: ALLEN,BERZONER,COHEN,COOPER,KUPFERZMIDT, > POSNER (POIZNER), PRITIKIN, STARKMAN > Dachau Camp Entry List, Genealogy Utilities > http://home.adelphia.net/~dcooper000 > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: trouble in ping with a larger packet on ne2000 Ethernet Date: Thu, 29 Aug 2002 18:37:55 +0300 From: "Martin Roth" Organization: Motorola Message-ID: References: <7f29c2ad.0208080439.62e501b7@posting.google.com> <3D54F62E.7060304@spamm.me.l8s.co.uk> <8400a733.0208270006.726e3f09@posting.google.com> I have already reported this problem to Vxworks. The problem ocured on the FEC 10/100 bps port on MPC860T family. There were actual two problems. The second one I reported to Vxworks and you can find it below. The first problem was that you have to verify that the DSHW bit in SIUMCR register is set off. This was an errata of Motorola . Martin Roth Motorola CommunicationsIsrael Mailto:bmr006@email.mot.com Hi, My name is Martin Roth and I am from Motorola Communications Israel. My license number is #201203. I am using the following tools: Tornado 2.02 Vxworks 5.4 The target board is with MPC855T (POWERPC) processor of Motorola. The BSP is for ADS860 Motorola board and it is ported to our board. We have also purchased the driver sources including the FEC (Fast Ethernet Controller) for MPC8XX family We started testing our new board based on MPC855T 48MHz processor. The FEC port was programmed to 100 Mbps. When pinging the FEC port I started experiencing "Transmit underrun" errors. I collected the statistics by activating the motFecErrorShow() from the target shell. The problem occurred when I was continuously pinging the board with large (30000 bytes) packets. The command from NT: ping 145.9.199.237 -l 30000 -t The problem did not occur on a standard ADS Motorola board. We suspected that our board has hardware problems, but after a weeks of investigation my conclusion that there is a problem with the FEC driver or with the upper software layer. The problem description: The upper layers call to motFecSend() with an M_BLK data structure that may consist of many (8-10) successive clusters with zero size. For each cluster a separate descriptor is used. Reading 8-10 empty (void) descriptors (8 bytes each) and analyzing them may take more than 5 microseconds. The Ethernet transmit buffer length is 64 bytes, which is ~ 5 microseconds transmission on 100 Mbps media. When we have successive descriptors with "zero length data", the transmit buffer is emptied before the descriptors a read and this is the reason for the transmit underrun error. The problem did not occur on ADS, because it works on 66 or 66 MHz and our board uses 48 MHz clock. Question: What are the clusters of size 0 in the M_BLK and why does the driver use a separate descriptor for this cluster? What is the sense of transmitting of a cluster which size is zero ? In order to fix the problem the following should be done: Not to have a cluster with size 0 within the M_BLK (created by upper layers) Not to try to use a transmit descriptor within motFecEnd.c when a cluster size in M_BLK is zero I have patched the motfecend.c and the problem disappeared. The reason I am reporting this for you is very simple. I would like Windriver to fix the problem properly, so I do not have to deal with it on next releases of Tornado/motfecend.c. "Toni Kurzberger" wrote in message news:8400a733.0208270006.726e3f09@posting.google.com... > David Laight wrote in message news:<3D54F62E.7060304@spamm.me.l8s.co.uk>... > > > > > > when I ping it with a litter packet(<1500 bytes),it is ok,but when the > > > packet is larger(>1500),ping is failed! why? > This might be a driver bug? In any case it would be a good idea > considering to update to a later version (to give it at least a > try)... e.g. Tornado 2.0.2 + SPR68322 > > toni --------------------------- Newsgroups: comp.os.vxworks Subject: Re: booting from tffs Date: Thu, 29 Aug 2002 16:35:42 GMT From: "Chad" Organization: Shaw Residential Internet Message-ID: References: I _think_ they are all ok.. I called devsplit on the filename after a lot of the debug, which could be misleading. *********************************************** STATUS usrTffsConfig ( int drive, /* drive number of TFFS */ int removable, /* 0 - nonremovable flash media */ char * fileName /* mount point */ ) { BLK_DEV * pBootDev; char bootDir [BOOT_FILE_LEN]; CBIO_DEV_ID cbio_dev = 0; STATUS status = -1; printErr("BOOT: Mounting TFFS device at: %s\n",fileName); if ((UINT)drive >= noOfDrives) { printErr ("BOOT: drive is out of range (0-%d).\n", noOfDrives - 1); return (ERROR); } /* create a block device spanning entire disk (non-distructive!) */ if ((pBootDev = tffsDevCreate (drive, removable)) == NULL) { printErr ("BOOT: tffsDevCreate(%d,%d) failed.\n",drive,removable); return (ERROR); } else { printErr ("BOOT: tffsDevCreate(%d,%d) successful.\n",drive,removable); } printErr("BOOT: calling cbioDevVerify...\n"); cbio_dev = cbioDevVerify(pBootDev, 0); if (cbio_dev == 0) { printf("BOOT: can't create cbio wrapper: %x\n", errno); return (ERROR); } /* split off boot device from boot file */ devSplit (fileName, bootDir); <-------------------------------------------------This should provide the correct name right? /* initialize the boot block device as a dosFs device named */ if (dosFsDevInit (bootDir, pBootDev, NULL) == NULL) { printErr ("BOOT: dosFsDevInit failed.\n"); return (ERROR); } else status = 0; ********************************************************* The error I got, "Failed to attach to device tffs=0,0Attaching interface lo0...done" seems to be generated by an ipAttach call in either usrNetInit() or usrNetEndDevStart(). It seems to me that these calls should be made with my 'other' device, cpm rather than the tffs device? I placed cpm in the 'other' field of the bootline after reading in the FAQ: Q: When my system boots from network the network link is opened , but when I use TFFS as a boot device the network is not opened. A: Define the ethernet device in the "other" boot parameter, for example if thr original boot device is cpm(0,0) the other (o) parameter should be cpm. The bootline in this case could be if you want to boot from a flash disk: $tffs=0,0(0,0)host:vxworks e= o=cpm Could this also be the problem? The calls to these functions appear to originate from library code, but I presume they are controlled from the bootline, perhaps mine is wrong? "Leonid Rosenboim" wrote in message news:newscache$v4il1h$kin$1@lnews.actcom.co.il... > Chad, > > block 0 isn't your problem, the chkdsk goes through OK, the file system is > perfectly OK. > > Here is your problem: > > According to bootline, the boot FILE name is /flash/VxWorks, but your init > routine mounts the whole DEVICE as /flash/vxWorks, so naturally opening > "/flash/VxWorks" yields a DIRECTORY as opposed to a FILE, thus the loading > error. > > You need to make sure that the device is mounted with the device name being > "/flash/" with or without the trailing slash, and all will work. > > No need to fiddle with dosFsDevInit either. > > - Leonid > > "Chad" wrote in message > news:oscb9.222394$Ag2.11467798@news2.calgary.shaw.ca... > > I'm running a ppc860 with tornado 2.02 and can't seem to boot from my > intel > > strata flash. I am able to do the following: > > > > ************** > > > > > > -> tffsRawio 0,3,0,128,0 > > value = 0 = 0x0 > > -> tffsDevFormat 0,0 > > CFISCS-PGC: entering CFISCS identification routine. > > CFISCS-PGC: reading CFI for word mode. > > CFISCS-PGC: identified 16-bit CFISCS. > > Eraseable block size: 0x20000 > > Chip size: 0x2000000 > > Number of chips: 1 > > Interleaving: 2 > > CFISCS-PGC: entering CFISCS identification routine. > > CFISCS-PGC: reading CFI for word mode. > > CFISCS-PGC: identified 16-bit CFISCS. > > Eraseable block size: 0x20000 > > Chip size: 0x2000000 > > Number of chips: 1 > > Interleaving: 2 > > CFISCS-PGC: entering CFISCS identification routine. > > CFISCS-PGC: reading CFI for word mode. > > CFISCS-PGC: identified 16-bit CFISCS. > > Eraseable block size: 0x20000 > > Chip size: 0x2000000 > > Number of chips: 1 > > Interleaving: 2 > > value = 0 = 0x0 > > > > ************** > > > > I can ftp my vxWorks image to the flash and back to my system and verify > > that it was not corrupted, and boot from the network and mount/view the > > device with no apparent problems. When I boot from flash I get this: > > > > ************** > > > > boot device : tffs=0,0 > > unit number : 0 > > processor number : 0 > > host name : rnd > > file name : /flash/vxWorks > > inet on ethernet (e) : 192.0.1.90 > > host inet (h) : 192.0.1.58 > > flags (f) : 0x8 > > other (o) : cpm > > > > tffsLoad: Tffs initialized... > > tffsLoad: Attaching to TFFS... > > Mounting TFFS device at: /flash/vxWorks > > CFISCS-PGC: entering CFISCS identification routine. > > CFISCS-PGC: reading CFI for word mode. > > CFISCS-PGC: identified 16-bit CFISCS. > > Eraseable block size: 0x20000 > > Chip size: 0x2000000 > > Number of chips: 1 > > Interleaving: 2 > > tffsDevCreate(0,0) successful. > > src/config calling cbioDevVerify... > > dosFS device created. > > usrTffsConfig(0,0,/flash/vxWorks) succeeded. > > done. > > /flash/vxWorks/ - disk check in progress ... > > WARNING : dosChkLib : system clock is being set to FRI JAN 01 00:04:04 > 1999 > > Value obtained from file system referenced by volume descriptor pointer: > > 0xffdf7 > > 0 > > The old setting was THU JAN 01 00:00:01 1970 > > Accepted system dates are greater than FRI JAN 01 00:00:00 1999 > > /flash/vxWorks/ - Volume is OK > > > > total # of clusters: 15,806 > > # of free clusters: 9,371 > > # of bad clusters: 0 > > total free space: 18,742 Kb > > max contiguous free space: 19,191,808 bytes > > # of files: 2 > > # of folders: 0 > > total bytes in files: 12,862 Kb > > # of lost chains: 0 > > total bytes in lost chains: 0 > > Erroneous header read > > error loading file: status = 0x610001. > > Error loading file: errno = 0x610001. > > > > ************** > > > > I'm not sure what the Erroneous header read message is indicating, is it > > simply that my file is not found? /flash/vxWorks is there and visible if > I > > ftp or boot from network and afaik is a valid image to boot from flash > (the > > same image I boot from network) > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: booting from tffs Date: Thu, 29 Aug 2002 16:44:34 GMT From: "Chad" Organization: Shaw Residential Internet Message-ID: References: Here is what I was trying to get at: ************************************** else { if ((strncmp (params.bootDev, "scsi", 4) == 0) || (strncmp (params.bootDev, "ide", 3) == 0) || (strncmp (params.bootDev, "ata", 3) == 0) || (strncmp (params.bootDev, "fd", 2) == 0) || (strncmp (params.bootDev, "tffs", 4) == 0)) <------------------------------------- { /* booting from disk, configure network if requested */ if (params.other [0] != EOS) <---------------- my 'other' is set to 'cpm' so pNetDev should also be set to cpm here... { /* * Protocol attachment is required and automatic * configuration with DHCP or BOOTP is supported * for all these network devices. */ attachFlag = TRUE; pNetDev = params.other; } } else { pNetDev = params.bootDev; <----------------- but if it is set to bootdev here, the error I got could easily follow /* * Booting via network: protocol attachment is required and * automatic configuration with DHCP or BOOTP is supported * for all these network devices. */ attachFlag = TRUE; } } if (attachFlag) { /* * Attempt to attach and setup the network boot device * (which is capable of automatic configuration). */ BOOL attached = FALSE; netmask = 0; bootNetmaskExtract (pBootString, &netmask); #ifdef INCLUDE_END /* Check for the END driver, if any. */ pEnd = endFindByName (pNetDev, params.unitNum); <---------------------- if (pEnd != NULL) { if (ipAttach(params.unitNum, pNetDev) != OK) <------------------- "Failed to attach to device tffs=0,0".. it should be cpm! { logMsg ("Failed to attach to device %s", (int)pNetDev, 0, 0, 0, 0, 0); return (ERROR); } ************************************* "Chad" wrote in message news:y3sb9.226743$Ag2.11718902@news2.calgary.shaw.ca... > I _think_ they are all ok.. I called devsplit on the filename after a lot of > the debug, which could be misleading. > > *********************************************** > STATUS usrTffsConfig > ( > int drive, /* drive number of TFFS */ > int removable, /* 0 - nonremovable flash media */ > char * fileName /* mount point */ > ) > { > BLK_DEV * pBootDev; > char bootDir [BOOT_FILE_LEN]; > CBIO_DEV_ID cbio_dev = 0; > STATUS status = -1; > > printErr("BOOT: Mounting TFFS device at: %s\n",fileName); > > if ((UINT)drive >= noOfDrives) > { > printErr ("BOOT: drive is out of range (0-%d).\n", noOfDrives - 1); > return (ERROR); > } > > /* create a block device spanning entire disk (non-distructive!) */ > > if ((pBootDev = tffsDevCreate (drive, removable)) == NULL) > { > printErr ("BOOT: tffsDevCreate(%d,%d) failed.\n",drive,removable); > return (ERROR); > } > else > { > printErr ("BOOT: tffsDevCreate(%d,%d) > successful.\n",drive,removable); > } > > printErr("BOOT: calling cbioDevVerify...\n"); > cbio_dev = cbioDevVerify(pBootDev, 0); > > if (cbio_dev == 0) > { > printf("BOOT: can't create cbio wrapper: %x\n", errno); > return (ERROR); > } > > /* split off boot device from boot file */ > devSplit (fileName, bootDir); > <-------------------------------------------------This should provide the > correct name right? > > /* initialize the boot block device as a dosFs device named */ > > if (dosFsDevInit (bootDir, pBootDev, NULL) == NULL) > { > printErr ("BOOT: dosFsDevInit failed.\n"); > return (ERROR); > } else status = 0; > > ********************************************************* > > The error I got, "Failed to attach to device tffs=0,0Attaching interface > lo0...done" seems to be generated by an ipAttach call in either usrNetInit() > or usrNetEndDevStart(). It seems to me that these calls should be made with > my 'other' device, cpm rather than the tffs device? I placed cpm in the > 'other' field of the bootline after reading in the FAQ: > > > Q: When my system boots from network the network link is opened , but when I > use TFFS as a boot device the network is not opened. > A: Define the ethernet device in the "other" boot parameter, for example if > thr original boot device is cpm(0,0) the other (o) parameter should be cpm. > The bootline in this case could be if you want to boot from a flash disk: > > $tffs=0,0(0,0)host:vxworks e= o=cpm > > Could this also be the problem? The calls to these functions appear to > originate from library code, but I presume they are controlled from the > bootline, perhaps mine is wrong? > > > "Leonid Rosenboim" wrote in message > news:newscache$v4il1h$kin$1@lnews.actcom.co.il... > > Chad, > > > > block 0 isn't your problem, the chkdsk goes through OK, the file system is > > perfectly OK. > > > > Here is your problem: > > > > According to bootline, the boot FILE name is /flash/VxWorks, but your init > > routine mounts the whole DEVICE as /flash/vxWorks, so naturally opening > > "/flash/VxWorks" yields a DIRECTORY as opposed to a FILE, thus the loading > > error. > > > > You need to make sure that the device is mounted with the device name > being > > "/flash/" with or without the trailing slash, and all will work. > > > > No need to fiddle with dosFsDevInit either. > > > > - Leonid > > > > "Chad" wrote in message > > news:oscb9.222394$Ag2.11467798@news2.calgary.shaw.ca... > > > I'm running a ppc860 with tornado 2.02 and can't seem to boot from my > > intel > > > strata flash. I am able to do the following: > > > > > > ************** > > > > > > > > > -> tffsRawio 0,3,0,128,0 > > > value = 0 = 0x0 > > > -> tffsDevFormat 0,0 > > > CFISCS-PGC: entering CFISCS identification routine. > > > CFISCS-PGC: reading CFI for word mode. > > > CFISCS-PGC: identified 16-bit CFISCS. > > > Eraseable block size: 0x20000 > > > Chip size: 0x2000000 > > > Number of chips: 1 > > > Interleaving: 2 > > > CFISCS-PGC: entering CFISCS identification routine. > > > CFISCS-PGC: reading CFI for word mode. > > > CFISCS-PGC: identified 16-bit CFISCS. > > > Eraseable block size: 0x20000 > > > Chip size: 0x2000000 > > > Number of chips: 1 > > > Interleaving: 2 > > > CFISCS-PGC: entering CFISCS identification routine. > > > CFISCS-PGC: reading CFI for word mode. > > > CFISCS-PGC: identified 16-bit CFISCS. > > > Eraseable block size: 0x20000 > > > Chip size: 0x2000000 > > > Number of chips: 1 > > > Interleaving: 2 > > > value = 0 = 0x0 > > > > > > ************** > > > > > > I can ftp my vxWorks image to the flash and back to my system and verify > > > that it was not corrupted, and boot from the network and mount/view the > > > device with no apparent problems. When I boot from flash I get this: > > > > > > ************** > > > > > > boot device : tffs=0,0 > > > unit number : 0 > > > processor number : 0 > > > host name : rnd > > > file name : /flash/vxWorks > > > inet on ethernet (e) : 192.0.1.90 > > > host inet (h) : 192.0.1.58 > > > flags (f) : 0x8 > > > other (o) : cpm > > > > > > tffsLoad: Tffs initialized... > > > tffsLoad: Attaching to TFFS... > > > Mounting TFFS device at: /flash/vxWorks > > > CFISCS-PGC: entering CFISCS identification routine. > > > CFISCS-PGC: reading CFI for word mode. > > > CFISCS-PGC: identified 16-bit CFISCS. > > > Eraseable block size: 0x20000 > > > Chip size: 0x2000000 > > > Number of chips: 1 > > > Interleaving: 2 > > > tffsDevCreate(0,0) successful. > > > src/config calling cbioDevVerify... > > > dosFS device created. > > > usrTffsConfig(0,0,/flash/vxWorks) succeeded. > > > done. > > > /flash/vxWorks/ - disk check in progress ... > > > WARNING : dosChkLib : system clock is being set to FRI JAN 01 00:04:04 > > 1999 > > > Value obtained from file system referenced by volume descriptor pointer: > > > 0xffdf7 > > > 0 > > > The old setting was THU JAN 01 00:00:01 1970 > > > Accepted system dates are greater than FRI JAN 01 00:00:00 1999 > > > /flash/vxWorks/ - Volume is OK > > > > > > total # of clusters: 15,806 > > > # of free clusters: 9,371 > > > # of bad clusters: 0 > > > total free space: 18,742 Kb > > > max contiguous free space: 19,191,808 bytes > > > # of files: 2 > > > # of folders: 0 > > > total bytes in files: 12,862 Kb > > > # of lost chains: 0 > > > total bytes in lost chains: 0 > > > Erroneous header read > > > error loading file: status = 0x610001. > > > Error loading file: errno = 0x610001. > > > > > > ************** > > > > > > I'm not sure what the Erroneous header read message is indicating, is it > > > simply that my file is not found? /flash/vxWorks is there and visible > if > > I > > > ftp or boot from network and afaik is a valid image to boot from flash > > (the > > > same image I boot from network) > > > > > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: booting from tffs Date: Thu, 29 Aug 2002 16:56:00 GMT From: "Chad" Organization: Shaw Residential Internet Message-ID: References: Is it just me or do the brackets appear screwed up in the code posted.. could somebody email me a clean copy of usrNetwork.c so that I can ensure it wasn't corrupted in house. > Here is what I was trying to get at: > > ************************************** > else > { > if ((strncmp (params.bootDev, "scsi", 4) == 0) || > (strncmp (params.bootDev, "ide", 3) == 0) || > (strncmp (params.bootDev, "ata", 3) == 0) || > (strncmp (params.bootDev, "fd", 2) == 0) || > (strncmp (params.bootDev, "tffs", 4) == 0)) > <------------------------------------- > { > /* booting from disk, configure network if requested */ > > if (params.other [0] != EOS) <---------------- my 'other' is set to 'cpm' so pNetDev should also be set to cpm here... > { > /* > * Protocol attachment is required and automatic > * configuration with DHCP or BOOTP is supported > * for all these network devices. > */ > > attachFlag = TRUE; > pNetDev = params.other; > } ^^^^^^^^^^^^^^^^^^^^^^ this bracket appears misplaced to me, and similarly in other statements in this code > } > else > { > pNetDev = params.bootDev; <----------------- but if it is set to > bootdev here, the error I got could easily follow > > /* > * Booting via network: protocol attachment is required and > * automatic configuration with DHCP or BOOTP is supported > * for all these network devices. > */ > > attachFlag = TRUE; > } > } > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Target shell symbol table disapperance Date: 29 Aug 2002 10:20:21 -0700 From: arohb@cc.usu.edu (Aroh Barjatya) Organization: http://groups.google.com/ Message-ID: References: Seraph, I am just using the tornado host based shell as a development shell right now and intend to use the target shell as a reprogramming shell when my application is comepletely built..... in other words when I deliever the application with the hardware, there will be no tornado host based shell, there will be only target shell running.... Hence I dont want "synchronise host and target shell" Thanks Aroh "Seraph Ma" wrote in message news:... > hey > Perhaps you should include the item of 'symbol table > componets->synchorize host and target symbol table ' in you project when you > make a vxWorks image. > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: absolute variable locations for GNUC Date: 29 Aug 2002 13:36:12 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: References: <7dc75e7f.0208281401.656432eb@posting.google.com> <3D6D4984.2C265377@bellsouth.net> <7dc75e7f.0208290441.ec752ad@posting.google.com> Sender: bpringle@DeadDuck PW> Unfortunately I don't have the luxury to change all of the PW> existing code that uses the variable. We were using the diab PW> compiler with nucleus before and so we could define a section PW> using #pragma and put the section at the desired location. Nucleus? You poor sucker. Look at the documentation for the linker. "file:///C|/Tornado/docs/gnu/toolkit/ld.html" "C:\Tornado\docs\gnu\toolkit\ld.html" For example, "-defsym symbol=expression" might be useful. Regards, Bill Pringlemeir. - -- If our behavior is strict, we do not need fun! --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Interrupt vector table Date: Thu, 29 Aug 2002 13:20:27 -0500 From: Andrew Johnson Organization: APS, Argonne National Laboratory Message-ID: References: JJS wrote: > Our system : Tornado 2.0 VxWorks 5.4 mvme 2304 board > In a 68000 or x86 system, the interrupt vector table begins in memory at > address 0 > So vector 10 is at address 10*4 > Where is the vector table in a ppc architecture? The PowerPC 603/604 family don't directly support vectored hardware interrupts - there is actually only one interrupt vector used for all hardware interrupts, and it's up to the software to decide what actually caused the interrupt (which is why PowerPC interrupt response often isn't any faster than the 68K family). In vxWorks for the PowerPC architecture, intConnect() delegates the actual vector connection to a BSP function. Look for a statement in your BSP source code that sets the variable _func_intConnectRtn and then examine the routine it is set to point to. VME board BSPs will set up their own vector table in this routine for VMEbus interrupt vectors - on the 2700 it's an array called sysIntTbl[]. - - Andrew - -- Edsger W. Dijkstra, 1930-2002: "The use of COBOL cripples the mind; its teaching should, therefore, be regarded as a criminal offence." --------------------------- Newsgroups: alt.os.linux.suse,comp.os.linux,comp.os.linux.advocacy,comp.os.linux.embedded,comp.os.vxworks Subject: newsgroup spreading virus Date: Sat, 24 Aug 2002 23:43:47 GMT From: Russ Organization: EarthLink Inc. -- http://www.EarthLink.net Message-ID: Followup-To: alt.os.linux.advocacy Pardon me if you have seen this before. A virus is being circulated with spoofed origin addresses. It comes as an e-mail with 0,0 frame hidden in an image and two attachments. The last one I got had an attachment "Hereby.exe" that appears to be a variant of the Klez virus. I first saw this as a posting in some of the news groups I've cross posted this message to. I have received several e-mails with the virus, all purporting to come from earthlink.net addresses. I have also received a number of bounced messages not originated by me containing the virus. In addition, these e-mails had destinations that were not in my address book. I notified abuse@earthlink.net, but have received no response. Also today I spent 1-1/2 hours trying to get through to a live person at earthlink. I did eventually reach one Brian (an expert on viruses who can't remember his employee number) who has a boss named Matt who is too busy to talk to customers. So don't expect earthlink to do anything soon. The simplest way to avoid the virus is to pre-filter all mail and dump any that contains HTML. I, myself, do not send or receive mail with HTML. Also, I do not send or receive files with attachments, much less exe attachments, unless the recipient and I have previously agreed to a unique key for the subject line. I would recommend you do the same. - -- Russ Lyttle - Not powered by ActiveX Please visit for OBD-II information and links. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: do not want vxworks shell to attach to console by default Date: 29 Aug 2002 11:32:02 -0700 From: akiggins@celitesystems.com (andrew) Organization: http://groups.google.com/ Message-ID: <2eced96.0208291032.74fe31ac@posting.google.com> References: Here's a lump of code that when activated will restart the shell using the filehandle of the current process. If you call this from your telnet session you will "jump" into the console. BTW it assumes that at this point you have killed the shell process. I think you have to do that or at least a restart to get it to behave nicely. andrew MIError_t spawnShell() { int n; int shelltid; /* remove the entry in the symbol table map which maps "exit" to exit() * that would be bad, we need to catch the exit so we can put the * console back into raw mode */ for( n = 0; n < standTblSize; n++ ) { if ( ! strcmp("exit", standTbl[n].name)) { standTbl[n].value = (char*)shellExit; break; } } shellDeath = semBCreate( SEM_Q_FIFO, SEM_EMPTY); /* remap the shell to local file handle */ shellOrigStdSet(STD_IN, ioTaskStdGet( taskIdSelf(), STD_IN)); shellOrigStdSet(STD_OUT, ioTaskStdGet( taskIdSelf(), STD_OUT)); shellOrigStdSet(STD_ERR, ioTaskStdGet( taskIdSelf(), STD_ERR)); shellInit(SHELL_STACK_SIZE, TRUE); semTake(shellDeath, WAIT_FOREVER); return mi_noerror; } "Jackie Bhadange" wrote in message news:... > This is a multi-part message in MIME format. > > ------=_NextPart_000_0001_01C24ED8.676CEF10 > Content-Type: text/plain; > charset="iso-8859-1" > Content-Transfer-Encoding: 7bit > > Hi, > > Is there a way to disable the shell from attaching to the console port? > We want the shell but we want it accessible only through telnet (and not > through the console port). > Any ideas? > > Thanks, > Jackie > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: booting from tffs Date: Thu, 29 Aug 2002 18:53:01 GMT From: "Chad" Organization: Shaw Residential Internet Message-ID: References: I booted from flash! It turns out I missed a modification required in usrNetBoot.c::usrNetDevNameGet if ( (strncmp (sysBootParams.bootDev, "scsi", 4) == 0) || (strncmp (sysBootParams.bootDev, "ide", 3) == 0) || (strncmp (sysBootParams.bootDev, "ata", 3) == 0) || (strncmp (sysBootParams.bootDev, "fd", 2) == 0) || (strncmp (sysBootParams.bootDev, "tffs", 4) == 0)) <------------- had to be added This is what had to be done also in usrNetwork.c::usrNetInit Apologies for dragging this on, and thanks to all who helped. Cheers, Chad --------------------------- Newsgroups: comp.os.vxworks Subject: elf/ coff/ Motorola S-record/Intel Hex/Bin ? Date: 24 Aug 2002 19:09:33 -0700 From: xiaohua.zhang@163.com (xiaohua) Organization: http://groups.google.com/ Message-ID: Could anybody tell me how to use these format in VxWorks envioroments? --------------------------- Newsgroups: comp.os.vxworks Subject: Detecting is a key has been pressed Date: 29 Aug 2002 13:39:24 -0700 From: pwiebe@bellsouth.net (Philip Wiebe) Organization: http://groups.google.com/ Message-ID: <7dc75e7f.0208291239.4712c639@posting.google.com> Is there a method of detecting if a key has been pressed on the keyboard when the console is in LINE mode? The statement ioctl(STD_IN, FIONREAD, &n) only returns a valid number once enter has been hit. I know one way is to change to RAW mode, but then you have to process each character at a time, including backspace, etc. Phil. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Source of mips toolchain Date: Thu, 29 Aug 2002 21:21:29 +0000 (UTC) From: Johan Borkhuis Organization: Agere Systems Message-ID: References: <8b53eb96.0208260539.7ac17e6e@posting.google.com> <8b53eb96.0208270851.43ab2243@posting.google.com> <8b53eb96.0208290554.2b5583fb@posting.google.com> chs@lucent.com (ChengHsin Hsu) wrote: > Have submitted a TSR again. Useless....No reply at all. > Any idea about how to proceed? Just contact you sales rep or FAE directly. They will be able to provide you with the SW. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Detecting is a key has been pressed Date: Thu, 29 Aug 2002 19:17:43 -0400 From: joe durusau Organization: WEBUSENET.com Message-ID: <3D6EAB96.41218AE8@bellsouth.net> References: <7dc75e7f.0208291239.4712c639@posting.google.com> Philip Wiebe wrote: > Is there a method of detecting if a key has been pressed on the > keyboard when the console is in LINE mode? The statement > ioctl(STD_IN, FIONREAD, &n) only returns a valid number once enter has > been hit. > > I know one way is to change to RAW mode, but then you have to process > each character at a time, including backspace, etc. > > Phil. Not that I've ever heard of. You could write your own driver for whatever hardware you are using for input, but that seems a bit extreme. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Detecting is a key has been pressed Date: 29 Aug 2002 23:36:12 GMT From: Denis Perelyubskiy Message-ID: References: <7dc75e7f.0208291239.4712c639@posting.google.com> * Philip Wiebe [08-Thu-02 13:39 -0700]: > > Is there a method of detecting if a key has been pressed > on the keyboard when the console is in LINE mode? The > statement ioctl(STD_IN, FIONREAD, &n) only returns a > valid number once enter has been hit. Is it not possible to simply select on STD_IN ? It seems that would depend on your driver though... Just a guess... denis - -- if you want to reply by mail, please tweak my address in a reasonable way. sorry for the inconvenience. --------------------------- Newsgroups: comp.os.vxworks Subject: VMIC 6001 with Vxworks Date: Thu, 29 Aug 2002 20:43:37 -0400 From: "Cedric Le Roux" Message-ID: <3d6ebfe6_3@nopics.sjc> I'm looking for some clarification regarding the programming for diagnostic mode for the VMIC 6001 MIL-STD-1553 board. The manual is kinda vague. If anyone out there has worked with this board, I'd like to hear from ya. or if anyone knows a better place for this post, I'd like to hear about that too. Thanks. --------------------------- Newsgroups: comp.os.vxworks Subject: 300M8260 Date: 29 Aug 2002 18:09:13 -0700 From: lionjackyfish@sina.com.cn (JackyLiu) Organization: http://groups.google.com/ Message-ID: <425faaf3.0208291709.7658ff5b@posting.google.com> Did anyone is used 8260 300/208/83MHZ ? I use it now, and i can succeed to use 8260 working in 133/133/66MHZ before. i change the voltage from 2.5 to 1.8 in hardware, and i change the MODCK_H and MODCK[1-3] to 0b0110110 too. I also change FREQ_66_MHZ to FREQ_83_MHZ, and change the define of the SYS_CPU_FREQ and the DEC_CLOCK_FREQ. I also change the value of BRG1 to adapt the SCC1 UART baud rate to 9600. Now , i use the program for memroy testing which copyright by MOTOROLA to test the SDRAM, the testing is passed. either, the vxworks must be fly at the following time! how can i resolve the problem? or just i must modify more in my BSP ? Thanks anyone for any discussing. THANKS. --------------------------- Newsgroups: comp.os.vxworks Subject: mimicking pSOS event group call Date: Thu, 29 Aug 2002 09:50:10 +0200 From: Pascal GULA Organization: Axlog =?ISO-8859-1?Q?Ingeni=E9rie?= Message-ID: <3D6DD232.7020104@axlog.fr> Reply-To: pascal.gula@axlog.fr Hi all, doing some porting work of a pSOS to vxWorks application, I am building an abstract library from an old one which encapsulate pSOS API. As a result, I found a AEvent class based on the pSOS event API (ev_receive(), ev_send(), ev_asend()), which is very useful. But I didn't found any direct API in vxWorks... As of now, I think I will mimic this behavior with some kind of semaphore vectoring for all the tasks of the system. Although, if anyone faced this problem in the past any comment about its solution is welcome ! Thanks in advance, Good Bye, - -- ============================================= GULA Pascal Ingénieur d'études chez AXLOG Ingénierie 19-21 rue du 8 mai 1945 94110 ARCUEIL Tel : 01.41.24.31.00 (Demander GULA Pascal) ============================================= --------------------------- Newsgroups: comp.os.vxworks Subject: Help on a Page Fault under Intel Pentium CPU Date: Fri, 30 Aug 2002 11:00:32 +0800 From: "Seraph Ma" Organization: Bentium Ltd. (CN99) Message-ID: Hey, I got such a PGF when my application runs on Intel Pentium CPU target. Page Fault Program Counter: 0x0da17668 Status Register: 0x00000003 Error Code: 0x0f2cc429 Task: 0x2ee80dc "IVR" ISR,Error,FFFD300C,F2CC78E,SAT NOV 18 22:11:17 2000,CPU:Page Fault I just wanna know what's the meaning of the PGF information.I have refered the Intel-provided IA-32 Intel CPU system programming Manual. But I can not find enough information illustrating such a PGF. Can anyone tell me the meaning of the following information? "Error Code: 0x0f2cc429 //----what's the meaning of this error code???? ISR,Error,FFFD300C,F2CC78E,SAT NOV 18 22:11:17 2000,CPU:Page Fault" // ---what's the meaning of the two // hex number FFFD300C,F2CC78E?? Thank you firstly! Best Regards, Seraph Ma £­£­£­£­£­£­£­£­£­£­£­£­£­£­£­£­£­£­£­£­£­£­£­£­£­£­ UTStarcom Shenzhen Research & Development Center Address: 3/F,Legend Research & Development Building, High-tech Industrial Park,Nanshan Shenzhen,PRC Zip Code: 518057 Tel: (86-755)2698 3555-3530 Fax: (86-755)2698 3786 Email: seraph.ma@utstar.com Web Site: http://www.utstar.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Basic Command Line Date: Fri, 30 Aug 2002 03:45:31 GMT From: "CHANGE username to just westes" Organization: Giganews.Com - Premium News Outsourcing Message-ID: References: It's the generic VxWorks command line with particular commands disabled. What commands do I need to list files on the file system? What are the critical startup configuration/script files and how do I list their contents? - -- Will NOTE: To reply, CHANGE the username to westes AT uscsw.com "Johan Borkhuis" wrote in message news:Xns92776F2DD5AD2borkhuisagerecom@192.19.193.66... > "CHANGE username to just westes" wrote: > > > We have a computer device that uses VxWorks internally to control > > robotics. The device's administrator is supposed to login and > > configure the device using the VxWorks command line. After reading > > the long list of VxWorks commands available within the device I don't > > see any way to explore the file system and read the contents of > > various files. Where can I get basic command line documentation that > > would explain how to do this? > > Is this a commandline specific for this device, or is it the generic > VxWorks command line? (What does it show after you connect or on startup?) > > Kind regards, > Johan > > -- > o o o o o o o . . . _____________________________ > o _____ || Johan Borkhuis | > .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | > >(________|__|_[_________]_|__________________________| > _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` > === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === > --------------------------- Newsgroups: comp.os.vxworks Subject: can load .out files off network but not TFFS Date: 29 Aug 2002 21:54:40 -0700 From: term@xerosoft.com (Tom) Organization: http://groups.google.com/ Message-ID: <15cdb0fb.0208292054.64cfaac9@posting.google.com> I have been working on getting a strong arm based motherboard to run vxworks. One problem I have run into is that I can't load any .out files off of my TFFS (True Flash File System). When I try to it just gives me an error saying "error reading file (errno = 0x3d0001).". I looked up the errno and it resolved to S_objLib_OBJ_ID_ERROR , but this is generated from something at a low level that I don't have the source code to. Every other operation with the TFFS seems to work just fine. I can create, delete, read, write, and rename files. As an example I tried copying an out file off my FTP server onto flash and loading it, but it failed. Then after that I copied the file off of the TFFS to a new file name on my FTP server and then loaded it from that file (over the network) and it worked fine. - -- start dump -- - -> cd "/tffs0/" value = 0 = 0x0 - -> ll - -rwxrwxrwx 1 0 0 10064 Jan 1 1980 test.out value = 0 = 0x0 - -> ld < test.out ld error: error reading file (errno = 0x3d0001). value = 0 = 0x0 - -> copy "/tffs0/test.out", "host:newtest.out" Copy OK: 10064 bytes copied value = 0 = 0x0 - -> ld < host:newtest.out value = 12531032 = 0xbf3558 - -> - -- end dump -- After these tests I thought the problem might have been size related, so I tried to create a minimal "hello world" out file, but that didn't work either. Has anyone else had any problems like this? I'm running VxWorks 5.4.2 with Tornado 2.1.1. Best Regards, Tom --------------------------- Newsgroups: comp.os.vxworks Subject: booting from tffs Date: Wed, 28 Aug 2002 22:49:56 GMT From: "Chad" Organization: Shaw Residential Internet Message-ID: I'm running a ppc860 with tornado 2.02 and can't seem to boot from my intel strata flash. I am able to do the following: ************** - -> tffsRawio 0,3,0,128,0 value = 0 = 0x0 - -> tffsDevFormat 0,0 CFISCS-PGC: entering CFISCS identification routine. CFISCS-PGC: reading CFI for word mode. CFISCS-PGC: identified 16-bit CFISCS. Eraseable block size: 0x20000 Chip size: 0x2000000 Number of chips: 1 Interleaving: 2 CFISCS-PGC: entering CFISCS identification routine. CFISCS-PGC: reading CFI for word mode. CFISCS-PGC: identified 16-bit CFISCS. Eraseable block size: 0x20000 Chip size: 0x2000000 Number of chips: 1 Interleaving: 2 CFISCS-PGC: entering CFISCS identification routine. CFISCS-PGC: reading CFI for word mode. CFISCS-PGC: identified 16-bit CFISCS. Eraseable block size: 0x20000 Chip size: 0x2000000 Number of chips: 1 Interleaving: 2 value = 0 = 0x0 ************** I can ftp my vxWorks image to the flash and back to my system and verify that it was not corrupted, and boot from the network and mount/view the device with no apparent problems. When I boot from flash I get this: ************** boot device : tffs=0,0 unit number : 0 processor number : 0 host name : rnd file name : /flash/vxWorks inet on ethernet (e) : 192.0.1.90 host inet (h) : 192.0.1.58 flags (f) : 0x8 other (o) : cpm tffsLoad: Tffs initialized... tffsLoad: Attaching to TFFS... Mounting TFFS device at: /flash/vxWorks CFISCS-PGC: entering CFISCS identification routine. CFISCS-PGC: reading CFI for word mode. CFISCS-PGC: identified 16-bit CFISCS. Eraseable block size: 0x20000 Chip size: 0x2000000 Number of chips: 1 Interleaving: 2 tffsDevCreate(0,0) successful. src/config calling cbioDevVerify... dosFS device created. usrTffsConfig(0,0,/flash/vxWorks) succeeded. done. /flash/vxWorks/ - disk check in progress ... WARNING : dosChkLib : system clock is being set to FRI JAN 01 00:04:04 1999 Value obtained from file system referenced by volume descriptor pointer: 0xffdf7 0 The old setting was THU JAN 01 00:00:01 1970 Accepted system dates are greater than FRI JAN 01 00:00:00 1999 /flash/vxWorks/ - Volume is OK total # of clusters: 15,806 # of free clusters: 9,371 # of bad clusters: 0 total free space: 18,742 Kb max contiguous free space: 19,191,808 bytes # of files: 2 # of folders: 0 total bytes in files: 12,862 Kb # of lost chains: 0 total bytes in lost chains: 0 Erroneous header read error loading file: status = 0x610001. Error loading file: errno = 0x610001. ************** I'm not sure what the Erroneous header read message is indicating, is it simply that my file is not found? /flash/vxWorks is there and visible if I ftp or boot from network and afaik is a valid image to boot from flash (the same image I boot from network) --------------------------- Newsgroups: comp.os.vxworks Subject: Re: booting from tffs Date: Wed, 28 Aug 2002 23:40:07 GMT From: "Chad" Organization: Shaw Residential Internet Message-ID: References: I found a post from a while back regarding a possible conflict between dosfs2 and tffs where they disagree about where sector 0 should point (posted by Andray Kaganovsky) and that a possible solution would be to remove the call to dosFsDevCreate as used by dosfs2 and use the (hopefully) backward compatible dosFsDevInit. This is promising so far, I made these changes, reformatted the flash and put the updated copy of vxWorks on it. Now I get this boot scenario: boot device : tffs=0,0 unit number : 0 processor number : 0 host name : rnd file name : /flash/vxWorks inet on ethernet (e) : 192.0.1.90 host inet (h) : 192.0.1.58 flags (f) : 0x8 other (o) : cpm tffsLoad: Tffs initialized... tffsLoad: Attaching to TFFS... BOOT: Mounting TFFS device at: /flash/vxWorks CFISCS-PGC: entering CFISCS identification routine. CFISCS-PGC: reading CFI for word mode. CFISCS-PGC: identified 16-bit CFISCS. Eraseable block size: 0x20000 Chip size: 0x2000000 Number of chips: 1 Interleaving: 2 BOOT: tffsDevCreate(0,0) successful. BOOT: calling cbioDevVerify... usrTffsConfig(0,0,/flash/vxWorks) succeeded. done. Loading /flash/vxWorks...2669040 Starting at 0x100000... Failed to attach to device tffs=0,0Attaching interface lo0...done ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Adding 7413 symbols for standalone. VxWorks Copyright 1984-1998 Wind River Systems, Inc. CPU: MPC823 - Thompson Technology PGC400 p0 VxWorks: 5.4.2 BSP version: 1.2/26 Creation date: Aug 28 2002 WDB: Ready. Mounting TFFS device at: /flash CFISCS-PGC: entering CFISCS identification routine. CFISCS-PGC: reading CFI for word mode. CFISCS-PGC: identified 16-bit CFISCS. Eraseable block size: 0x20000 Chip size: 0x2000000 Number of chips: 1 Interleaving: 2 tffsDevCreate(0,0) successful. calling cbioDevVerify... Flash File System Mounted 0xfffdf8 (): task deadCan't attach unknown device tffs=0,0 (unit 0). - -> ******************************* It looks like it actually booted from the flash image, although had a hiccup getting tffs started from the downloaded image. This image worked with the file system when I obtained it from the network however. Can anybody shed some light on what I've done wrong? "Chad" wrote in message news:oscb9.222394$Ag2.11467798@news2.calgary.shaw.ca... > I'm running a ppc860 with tornado 2.02 and can't seem to boot from my intel > strata flash. I am able to do the following: > > ************** > > > -> tffsRawio 0,3,0,128,0 > value = 0 = 0x0 > -> tffsDevFormat 0,0 > CFISCS-PGC: entering CFISCS identification routine. > CFISCS-PGC: reading CFI for word mode. > CFISCS-PGC: identified 16-bit CFISCS. > Eraseable block size: 0x20000 > Chip size: 0x2000000 > Number of chips: 1 > Interleaving: 2 > CFISCS-PGC: entering CFISCS identification routine. > CFISCS-PGC: reading CFI for word mode. > CFISCS-PGC: identified 16-bit CFISCS. > Eraseable block size: 0x20000 > Chip size: 0x2000000 > Number of chips: 1 > Interleaving: 2 > CFISCS-PGC: entering CFISCS identification routine. > CFISCS-PGC: reading CFI for word mode. > CFISCS-PGC: identified 16-bit CFISCS. > Eraseable block size: 0x20000 > Chip size: 0x2000000 > Number of chips: 1 > Interleaving: 2 > value = 0 = 0x0 > > ************** > > I can ftp my vxWorks image to the flash and back to my system and verify > that it was not corrupted, and boot from the network and mount/view the > device with no apparent problems. When I boot from flash I get this: > > ************** > > boot device : tffs=0,0 > unit number : 0 > processor number : 0 > host name : rnd > file name : /flash/vxWorks > inet on ethernet (e) : 192.0.1.90 > host inet (h) : 192.0.1.58 > flags (f) : 0x8 > other (o) : cpm > > tffsLoad: Tffs initialized... > tffsLoad: Attaching to TFFS... > Mounting TFFS device at: /flash/vxWorks > CFISCS-PGC: entering CFISCS identification routine. > CFISCS-PGC: reading CFI for word mode. > CFISCS-PGC: identified 16-bit CFISCS. > Eraseable block size: 0x20000 > Chip size: 0x2000000 > Number of chips: 1 > Interleaving: 2 > tffsDevCreate(0,0) successful. > src/config calling cbioDevVerify... > dosFS device created. > usrTffsConfig(0,0,/flash/vxWorks) succeeded. > done. > /flash/vxWorks/ - disk check in progress ... > WARNING : dosChkLib : system clock is being set to FRI JAN 01 00:04:04 1999 > Value obtained from file system referenced by volume descriptor pointer: > 0xffdf7 > 0 > The old setting was THU JAN 01 00:00:01 1970 > Accepted system dates are greater than FRI JAN 01 00:00:00 1999 > /flash/vxWorks/ - Volume is OK > > total # of clusters: 15,806 > # of free clusters: 9,371 > # of bad clusters: 0 > total free space: 18,742 Kb > max contiguous free space: 19,191,808 bytes > # of files: 2 > # of folders: 0 > total bytes in files: 12,862 Kb > # of lost chains: 0 > total bytes in lost chains: 0 > Erroneous header read > error loading file: status = 0x610001. > Error loading file: errno = 0x610001. > > ************** > > I'm not sure what the Erroneous header read message is indicating, is it > simply that my file is not found? /flash/vxWorks is there and visible if I > ftp or boot from network and afaik is a valid image to boot from flash (the > same image I boot from network) > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Basic Command Line Date: Fri, 30 Aug 2002 06:36:58 +0000 (UTC) From: Johan Borkhuis Organization: Agere Systems Message-ID: References: "CHANGE username to just westes" wrote: > It's the generic VxWorks command line with particular commands > disabled. > > What commands do I need to list files on the file system? With the "devs" command you can see the devices on your system. One of them will probably be your file system. You can use the "ls" command on this device to check. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Detecting is a key has been pressed Date: Fri, 30 Aug 2002 08:50:01 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <7dc75e7f.0208291239.4712c639@posting.google.com> tyLib has a input hook that gets called for every character received. The hook is called at ISR context, you could install your hook to release a binary semaphore when any key at all is pressed, and still keep your LINE mode processing done by tyLib. Look in tyLib reference manual, and tyLib.h header file. "Philip Wiebe" wrote in message news:7dc75e7f.0208291239.4712c639@posting.google.com... > Is there a method of detecting if a key has been pressed on the > keyboard when the console is in LINE mode? The statement > ioctl(STD_IN, FIONREAD, &n) only returns a valid number once enter has > been hit. > > I know one way is to change to RAW mode, but then you have to process > each character at a time, including backspace, etc. > > Phil. --------------------------- Newsgroups: comp.os.vxworks Subject: Problem about TCP link under vxWorks 5.4 Date: 30 Aug 2002 02:22:27 -0700 From: ellin_lin@263.net (ellin) Organization: http://groups.google.com/ Message-ID: <44a01eba.0208300122.316cd7b4@posting.google.com> I use TCP link in my program and it works fine. But I found a problem, that is: when the TCP Client halts(just like reboot), the TCP Server can't detect this and can't turn into the right status. When the TCP Client is colsed properly, the TCP Server surely detects this and turn into the right status. Anyone has the same experience? Any help will be appreciated! Ellin --------------------------- Newsgroups: comp.os.vxworks Subject: Re: mimicking pSOS event group call Date: Fri, 30 Aug 2002 12:41:50 GMT From: "drdiags" Organization: AT&T Broadband Message-ID: References: <3D6DD232.7020104@axlog.fr> Hello, T2.2 will support events (not sure about ev_asend() though). So maybe your workaround will not have to be long lived........ "Pascal GULA" wrote in message news:3D6DD232.7020104@axlog.fr... > Hi all, > > doing some porting work of a pSOS to vxWorks application, I am building > an abstract library from an old one which encapsulate pSOS API. > As a result, I found a AEvent class based on the pSOS event API > (ev_receive(), ev_send(), ev_asend()), which is very useful. But I > didn't found any direct API in vxWorks... > As of now, I think I will mimic this behavior with some kind of > semaphore vectoring for all the tasks of the system. Although, if anyone > faced this problem in the past any comment about its solution is welcome ! > > Thanks in advance, > Good Bye, > > -- > ============================================= > GULA Pascal > Ingénieur d'études chez AXLOG Ingénierie > > 19-21 rue du 8 mai 1945 > 94110 ARCUEIL > > Tel : 01.41.24.31.00 (Demander GULA Pascal) > ============================================= > --------------------------- Newsgroups: comp.os.vxworks Subject: PPPoE quesion Date: 30 Aug 2002 07:24:11 -0700 From: thatbird@sina.com (thatbird) Organization: http://groups.google.com/ Message-ID: Hello VxWorkers, I am now adding a PPPoE service to a HOST MODE device. I config the usrBasicPPP.c where I change the localusrName to "vxTarget", PassWord to "vxTarget", and authenticating mode from "Remote:Negotiation Not Required,Not Negotiable:c223" to "Remote:Negotiation Not Required, Negotiable:c223". Code is list below. ******************************************************************************* void usrBasicPPPInit () { ... ... ... /* say that we can deal with both CHAP and PAP */ pppParam = pfwParameterIdGet(pppSysFramework, "lcp_authProtocol"); /* initialize the secrets database */ m2pppSecuritySecretsLibInit(); /* add our name and password to the local secrets database */ m2pppSecretsEntrySet(0,PPP_SECURITY_SECRETS_DIRECTION_LOCAL_TO_REMOTE, PPP_SECURITY_CHAP_MD5_PROTOCOL,"vxTarget","vxTarget", PPP_SECURITY_SECRETS_STATUS_VALID); m2pppSecretsEntrySet(0,PPP_SECURITY_SECRETS_DIRECTION_LOCAL_TO_REMOTE, PPP_SECURITY_PAP_PROTOCOL,"vxTarget","vxTarget", PPP_SECURITY_SECRETS_STATUS_VALID ); /* say that we are open to authenticating ourselves using CHAP or PAP */ pfwProfileSet(pppSysProfile, pppParam, "Remote:Negotiation Not Required, Negotiable:c223");//"Remote:Negotiation Not Required,Not Negotiable:c223"); pfwProfileSet(pppSysProfile,pppParam, "Remote:Alternate:c023"); ... ... ... } ******************************************************************************* In my application, I wrote hostopen() using functions defined in pppAppl.c - ----------------------------------------------------------------------------- void hostopen() { void * PPPoE_usrHandle; PFW_STACK_OBJ * stackObj; if((PPPoE_usrHandle=malloc(UINT32))==NULL) { printf("PPPoE_HOST_open: cannot init usrHandle\n"); return; } if ((stackObj = pfwStackAdd(pppOeSysProfile, &applCallbacks, PPPoE_usrHandle))== NULL) { printf("PPPOE: Stack add failed\n"); return; } printf("stackObj = 0x%x\n", stackObj ); free(PPPoE_usrHandle); } - ----------------------------------------------------------------------------- When I excute hostopen() in shell, following debug infomation is printed. ... 0x18f6760 (SYS_PPP_C_1): CHAP:Received Challenge:ID 0, gotSecret vxTarget for localName vxTarget 0x18f6760 (SYS_PPP_C_1): CHAP: Sent Response, ID 0; retry = 0; time 343 0x18f6760 (SYS_PPP_C_1): CHAP: Sent Response, ID 0; retry = 1; time 643 0x18f6760 (SYS_PPP_C_1): CHAP: Sent Response, ID 0; retry = 2; time 943 0x18f6760 (SYS_PPP_C_1): CHAP: Sent Response, ID 0; retry = 3; time 1243 0x18f6760 (SYS_PPP_C_1): CHAP: Sent Response, ID 0; retry = 4; time 1543 0x18f6760 (SYS_PPP_C_1): CHAP: Sent Response, ID 0; retry = 5; time 1843 0x18f6760 (SYS_PPP_C_1): CHAP: Sent Response, ID 0; retry = 6; time 2143 Stack 1 is Deleted. 'sniffer' tell me the correct sequence should be: - ------------------------------------------------------------------- PPPoE: : Discovery Stage (PADI) PPPoE: : Discovery Stage (PADI) PPPoE: : Discovery Stage (PADI) PPPoE: : Discovery Stage (PADI) PPP: LCP Configure Request PPP: LCP Configure Request PPP: LCP Configure Ack PPP: LCP Configure Reject PPP: LCP Configure Request PPP: LCP Configure Ack PPP: LCP Identification PPP: LCP Identification CHAP: MESSAGE TYPE = Challenge Packet IDENTIFIER = 0 LENGTH = 29 CHAP: MESSAGE TYPE = Response Packet IDENTIFIER = 0 LENGTH = 62 CHAP: MESSAGE TYPE = Success Packet IDENTIFIER = 0 LENGTH = 46 ... - ------------------------------------------------------------------- While my device result: - ------------------------------------------------------------------- PPPoE: : Discovery Stage (PADI) PPPoE: : Discovery Stage (PADI) PPPoE: : Discovery Stage (PADI) PPPoE: : Discovery Stage (PADI) PPP: LCP Configure Request PPP: LCP Configure Request PPP: LCP Configure Ack PPP: LCP Configure Reject PPP: LCP Configure Request PPP: LCP Configure Ack CHAP: MESSAGE TYPE = Challenge Packet IDENTIFIER = 0 LENGTH = 29 CHAP: MESSAGE TYPE = Response Packet IDENTIFIER = 0 LENGTH = 29 ... CHAP: MESSAGE TYPE = Response Packet IDENTIFIER = 0 LENGTH = 29 PPP: LCP Echo Request PPP: LCP Echo Reply CHAP: MESSAGE TYPE = Response Packet IDENTIFIER = 0 LENGTH = 29 PPP: LCP Echo Request PPP: LCP Echo Reply ... PPPoE: : Discovery Stage - ------------------------------------------------------------------- I guess the challenge Response Packet is something error. Can you give me some advices? Thanks. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: DosFs2 chkdsk error message Date: 30 Aug 2002 08:17:07 -0700 From: dcooke@orthstar.com (Dan Cooke) Organization: http://groups.google.com/ Message-ID: <4e264d23.0208300717.1425dbdf@posting.google.com> References: <4e264d23.0207250701.1abc362a@posting.google.com> <1027613917.872841@azores.network-i.net> <4e264d23.0208021248.20b426c5@posting.google.com> > What arch are you on? If it is T2.0.2 x86 or PPC > then upgrade to T2.2 (best choice), or get the uber > patch for SPR#79569 or 79795 from Windsurf. > Using this patch, I am still getting the 0xC0006 S_ioLib_DISK_NOT_PRESENT. See error below: 0x30b4ab8 (LogDTask): logDTask.cpp line=420 error seeking to start ptr in LOG.BIN file, errno = 0xC0006, start ptr = 0x0000000065A10022. The wierd thing about this error, is that it always happens right around this file location 0x0000000065Axxxxx (the "xxxxx" address is variable). There were no other debug messages that came from the block layer or dosFs layers. Additional cbio, dcache, dosFs info to follows below... - -> dcacheShow Disk Cache, /sc0: Cached Block I/O Device, handle=0x37d42c8 Description: CBIO Disk Data Cache - LRU Disk size 67 Gbytes, RAM Size 131072 bytes Block size 512, heads 0, blocks/track 0, # of blocks 141940736 partition offset 0 blocks, type Fixed, Media changed No Current ready changed bit for this layer and all subordinates: CBIO_DEV_ID 0x037d42c8 ready changed is 0. CBIO_DEV_ID 0x037f4370 ready changed is 0. BLK_DEV ptr 0x037f43f0 ready changed is 0. Total cache Blocks 181, 0 blocks (%0) dirty Tunable Params: Bypass Threshold 24, Max Dirty 128, Read Ahead 22 blocks, Sync interval 0 sec Hit Stats: Cookie Hits 132633 Miss 1177, Hash size 269 Hits 1934 Miss 2821 LRU Hits 1929, Misses 133, Hit Ratio %93 Write Sstats: Foreground 0, Background 0, Hidden 1, Forced 27 Subordinate Device Start: Cached Block I/O Device, handle=0x37f4370 Description: CBIO to BLK_DEV Wrapper Disk size 67 Gbytes, RAM Size 0 bytes Block size 512, heads 0, blocks/track 0, # of blocks 141940736 partition offset 0 blocks, type Fixed, Media changed No Current ready changed bit for this layer and all subordinates: CBIO_DEV_ID 0x037f4370 ready changed is 0. BLK_DEV ptr 0x037f43f0 ready changed is 0. Subordinate Device End: value = 24 = 0x18 - -> dosFsShow "/sc0" volume descriptor ptr (pVolDesc): 0x37d2c58 cache block I/O descriptor ptr (cbio): 0x37d42c8 auto disk check on mount: NOT ENABLED max # of simultaneously open files: 32 file descriptors in use: 1 # of different files in use: 1 # of descriptors for deleted files: 0 # of obsolete descriptors: 1 current volume configuration: - volume label: N_DISK ; (in boot sector: N_DISK ) - volume Id: 0x1f760000 - total number of sectors: 141,940,736 - bytes per sector: 512 - # of sectors per cluster: 128 - # of reserved sectors: 32 - FAT entry size: FAT32 - # of sectors per FAT copy: 8,664 - # of FAT table copies: 1 - # of hidden sectors: 0 - first cluster is in sector # 8,696 - Update last access date for open-read-close = FALSE - names style: VxLong - root dir start cluster: 2 FAT handler information: - ------------------------ - allocation group size: 111 clusters - free space on volume: 69,271,683,016 bytes value = 0 = 0x0 - -> chkdsk "/sc0" /sc0/ - disk check in progress ... "/sc0/log/LOG.BIN" file larger then its chain of clusters Entry will be deleted. /sc0/ - Errors detected. To correct disk structure, please start check disk with a permis sion to repair. total # of clusters: 1,108,846 # of free clusters: 1,073,144 # of bad clusters: 0 total free space: 67,071 Mb max contiguous free space: 69,257,658,302 bytes # of files: 14 # of folders: 3 total bytes in files: 784,510 # of lost chains: 2 total bytes in lost chains: 2,229 Mb value = -1 = 0xffffffff - -> Any insights or info? Dan --------------------------- Newsgroups: comp.os.vxworks Subject: Re: disk cache error Date: Fri, 30 Aug 2002 11:30:27 -0400 From: "R Hilderman" Organization: Nortel Message-ID: References: Thanks for the response. It does clarify a few things. Randy --------------------------- Newsgroups: comp.os.vxworks Subject: Re: DosFs2 chkdsk error message Date: Fri, 30 Aug 2002 17:44:14 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <4e264d23.0207250701.1abc362a@posting.google.com> <1027613917.872841@azores.network-i.net> <4e264d23.0208021248.20b426c5@posting.google.com> <4e264d23.0208300717.1425dbdf@posting.google.com> Dan, this looks like a "new bug" to me, this was not the way it was done when initially released by my team. What I think is happening is that you are trying to seek past end of file to then write data. This has worked such that it is OK to seek past EOF, but subsequent read wil return an error. A write after that seek should complete ok, and actually extend the file as necesary with 0-blocks to accomodate the new size and data. FAT does not support sparse files like Unix, so either you support this and have the holes zero-filled, or not support it. We did zero-filling initially, and now it seems someone screwed this code up, or decided not to support seek-and-write past EOF. In any case, the errno you are getting is crap, and it should not update file size in directory if it does not allocate all these FAT entries and clusters approprietely. - - Leonid P.S. My conclustion is based on the offset on one hand, and disk space used as reported by chkdsk on the other hand. much smaller than this offset ! "Dan Cooke" wrote in message news:4e264d23.0208300717.1425dbdf@posting.google.com... > > What arch are you on? If it is T2.0.2 x86 or PPC > > then upgrade to T2.2 (best choice), or get the uber > > patch for SPR#79569 or 79795 from Windsurf. > > > > Using this patch, I am still getting the 0xC0006 > S_ioLib_DISK_NOT_PRESENT. > > See error below: > > 0x30b4ab8 (LogDTask): logDTask.cpp line=420 error seeking to start ptr > in LOG.BIN file, errno = 0xC0006, start ptr = 0x0000000065A10022. > > The wierd thing about this error, is that it always happens right > around > this file location 0x0000000065Axxxxx (the "xxxxx" address is > variable). > There were no other debug messages that came from the block layer or > dosFs layers. > > Additional cbio, dcache, dosFs info to follows below... > > -> dcacheShow > Disk Cache, /sc0: > Cached Block I/O Device, handle=0x37d42c8 > Description: CBIO Disk Data Cache - LRU > Disk size 67 Gbytes, RAM Size 131072 bytes > Block size 512, heads 0, blocks/track 0, # of blocks 141940736 > partition offset 0 blocks, type Fixed, Media changed No > > Current ready changed bit for this layer and all subordinates: > CBIO_DEV_ID 0x037d42c8 ready changed is 0. > CBIO_DEV_ID 0x037f4370 ready changed is 0. > BLK_DEV ptr 0x037f43f0 ready changed is 0. > Total cache Blocks 181, 0 blocks (%0) dirty > Tunable Params: > Bypass Threshold 24, Max Dirty 128, Read Ahead 22 blocks, Sync > interval 0 sec > Hit Stats: Cookie Hits 132633 Miss 1177, Hash size 269 Hits 1934 Miss > 2821 > LRU Hits 1929, Misses 133, Hit Ratio %93 > Write Sstats: Foreground 0, Background 0, Hidden 1, Forced 27 > > Subordinate Device Start: > Cached Block I/O Device, handle=0x37f4370 > Description: CBIO to BLK_DEV Wrapper > Disk size 67 Gbytes, RAM Size 0 bytes > Block size 512, heads 0, blocks/track 0, # of blocks 141940736 > partition offset 0 blocks, type Fixed, Media changed No > > Current ready changed bit for this layer and all subordinates: > CBIO_DEV_ID 0x037f4370 ready changed is 0. > BLK_DEV ptr 0x037f43f0 ready changed is 0. > Subordinate Device End: > value = 24 = 0x18 > > -> dosFsShow "/sc0" > > volume descriptor ptr (pVolDesc): 0x37d2c58 > cache block I/O descriptor ptr (cbio): 0x37d42c8 > auto disk check on mount: NOT ENABLED > max # of simultaneously open files: 32 > file descriptors in use: 1 > # of different files in use: 1 > # of descriptors for deleted files: 0 > # of obsolete descriptors: 1 > > current volume configuration: > - volume label: N_DISK ; (in boot sector: N_DISK ) > - volume Id: 0x1f760000 > - total number of sectors: 141,940,736 > - bytes per sector: 512 > - # of sectors per cluster: 128 > - # of reserved sectors: 32 > - FAT entry size: FAT32 > - # of sectors per FAT copy: 8,664 > - # of FAT table copies: 1 > - # of hidden sectors: 0 > - first cluster is in sector # 8,696 > - Update last access date for open-read-close = FALSE > - names style: VxLong > - root dir start cluster: 2 > > FAT handler information: > ------------------------ > - allocation group size: 111 clusters > - free space on volume: 69,271,683,016 bytes > value = 0 = 0x0 > > > -> chkdsk "/sc0" > /sc0/ - disk check in progress ... > "/sc0/log/LOG.BIN" file larger then its chain of clusters Entry will > be deleted. > /sc0/ - Errors detected. To correct disk structure, please start > check disk with a permis > sion to repair. > > total # of clusters: 1,108,846 > # of free clusters: 1,073,144 > # of bad clusters: 0 > total free space: 67,071 Mb > max contiguous free space: 69,257,658,302 bytes > # of files: 14 > # of folders: 3 > total bytes in files: 784,510 > # of lost chains: 2 > total bytes in lost chains: 2,229 Mb > value = -1 = 0xffffffff > -> > > > Any insights or info? > > Dan --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks Target Scripts - Date: Fri, 30 Aug 2002 17:48:30 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Andy, this is quite simple: When you shoot your loop from the tShell, the tShell code waits till your loop finishes off. When you start your code from the startup, it will run from tRoot task context, i.e. concurrently with tShell, meaning now the COMPETE FOR INPUT, and thus the tShell is stealing your input. Either you get your loop to use another comm channel, or you make sure tShell isnt reading from input while your loop does. There has been quite a lot of discussion on how that can be done on this newsgroup. HTH, - - Leonid "Atkinson Andrew" wrote in message news:ako27h$llc$1@overload.lbl.gov... > Hi > > I have a module which simply sits in a while loop getting input from the user (using scanf) and performs one of a number of options depending on the users choice - pretty straightforward. > > Now, when I download this module from either the host tools or from the command line of the target shell and run the 'main' routine, everything is OK. We enter the while loop, the user is prompted for input and the system waits until input is entered - as you would expect using scanf. However, if I use the target shell download the module and call the 'main' routine using a startup script (specified as a boot parameter) the while loop is entered, but the user is repeatedly prompted for input - the program > does not block on the call to scanf. > > Can anyone explain what is happening here - hopefully it is something very simple, but I cannot think what at the moment. > > Regards > > Andy Atkinson > Software Engineer > Bldg 31a > QinetiQ, Bincleaves Technology Park > Weymouth > Dorset > DT4 8UR > > Tel : 01305 764660 > Fax : 01305 764661 > e-mail : AATKINSON1@QinetiQ.com > > vxworks > > > The Information contained in this E-Mail and any subsequent correspondence > is private and is intended solely for the intended recipient(s). > For those other than the recipient any disclosure, copying, distribution, > or any action taken or omitted to be taken in reliance on such information > is prohibited and may be unlawful. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: mimicking pSOS event group call Date: 30 Aug 2002 10:06:52 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0208300906.2915733@posting.google.com> References: <3D6DD232.7020104@axlog.fr> Hello, Have you looked at the transition kit on the Wind River web site? You can download it for free using the form at: http://www.windriver.com/corporate/html/transkit.html HTH, John... Pascal GULA wrote in message news:<3D6DD232.7020104@axlog.fr>... > Hi all, > > doing some porting work of a pSOS to vxWorks application, I am building > an abstract library from an old one which encapsulate pSOS API. > As a result, I found a AEvent class based on the pSOS event API > (ev_receive(), ev_send(), ev_asend()), which is very useful. But I > didn't found any direct API in vxWorks... > As of now, I think I will mimic this behavior with some kind of > semaphore vectoring for all the tasks of the system. Although, if anyone > faced this problem in the past any comment about its solution is welcome ! > > Thanks in advance, > Good Bye, > > -- > ============================================= > GULA Pascal > Ingénieur d'études chez AXLOG Ingénierie > > 19-21 rue du 8 mai 1945 > 94110 ARCUEIL > > Tel : 01.41.24.31.00 (Demander GULA Pascal) > ============================================= --------------------------- Newsgroups: comp.os.vxworks Subject: I changed exception handler, but how to restore it. Date: 30 Aug 2002 11:27:18 -0700 From: yangliang_mr@hotmail.com (Liang Yang) Organization: http://groups.google.com/ Message-ID: <4f57a77f.0208301027.3897914b@posting.google.com> Hi, I inserted one instruction--"bla my_int_hook" into external interrupt handler (before "bla ppc860IntrInit + 0x304"). Program A works well after exception handler is changed. 00000500 7c7343a6 mtspr SPRG3, r3 00000504 7c6802a6 mfspr r3, LR 00000508 4801c3eb bla intEnt 0000050c 38610000 addi r3, r1, 0x0 (0) 00000510 9421fff0 stwu r1, 0xfff0(r1) 00000514 480eca63 bla my_int_hook !!! added by me 00000518 480135fb bla ppc860IntrInit + 0x304 0000051c 38210010 addi r1, r1, 0x10 (16) 00000520 4801c4bf bla intExit But after I remove "bla my_int_hook" and restore the exception handler to the exactly the same as the original one, I find program B doesn't work and system report the error as follows: - -> Exception number 4096: Task: 0x1e1018 (tExcTask) instruction translation miss program counter: 0xec9e4 machine status: 0x9032 88fb0 vxTaskEntry +5c : excTask ([]) 43de4 excTask +a4 : ec9e4 ([]) if I power down the target machine and restart, program B works well. Anybody can give me a good suggestion about this problem? Thanks. Liang --------------------------- Newsgroups: comp.os.vxworks Subject: anyone out there to help me plz?very urgent help requested Date: 30 Aug 2002 21:49:38 -0700 From: vxnerd@rediffmail.com (subbarayan) Organization: http://groups.google.com/ Message-ID: dear mentors, Can anyone jus tell me how to go abt proceeding the development of device driver for pci card for x86 processor with standard pc architecture?I use tornadoII and windows development host and a stand alone target. Further can anyone tell me when will be this pciautoconfiglib file called?will it be called automatically as a startup routine or need it to be called explicitly to enable pci scanning? What exactly is the difference between the files pci autoconfiglib.c and pciconfiglib.c?which file should be used for which sort of appln?and can these library files directly compiled and used or should we call it from some appln to invoke and access it? And sorry for my ignorance and unknowing things.One thing i found in the manual for pci support on vxworks says we need to include the macro INCLUDE_PCI for enabling pci support..i found only the macro INCLUDE_PCI_PARAMS and not INCLUDE_PCI.Are both same or different?if different where can i find INCLUDE_PCI and use it in my application? Where exactly should this macro be included and where exactly is this macro defined?How to include this macro and whats the exact location where this macro has to be included?And whats the role of bios with respect to vxworks in detecting pci card and pci bus?It would be helpful if some one could tell me what would be printed on screen if pci support is enabled and makes me confirm that my code has started working with pci support. Sorry if my questions look amature,but i am helpless as i am new to this sort of coding. Advance thanks to persons replying me. --------------------------- Newsgroups: comp.os.vxworks Subject: Does any know what "panic : soclose : NOFDREF" mean ........??? Date: 30 Aug 2002 21:51:23 -0700 From: mohasin.zaki@wipro.com (Mohasin) Organization: http://groups.google.com/ Message-ID: Hi All, I am having a client where I write data into sockets. After some amount of data is written I close the connection. Then for next set of data, I connect again to Server and send the data and close the socket adn this goes on. But sometime (rarely) I am getting an error on the shell which says "(TASKNAME) - panic : soclose" Which I think is OS panic. I am not sure if this error is returned when some error occurs while close()'ing the socket. Can anyone help me on this. Thankz a lot. Regards, MOhasin --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Basic Command Line Date: Sat, 31 Aug 2002 07:05:25 GMT From: "CHANGE username to just westes" Organization: Giganews.Com - Premium News Outsourcing Message-ID: References: What is the command to copy the contents of a file to the console? What is the fullpath expression for file A.TXT in directory B on device C? From any arbitrary location on the disk, how would you list the contents of A to the console? On Windows 2000 command line of course this would be any of: type c:\b\a.txt // not case sensitive copy c:\b\a.txt con etc - -- Will NOTE: To reply, CHANGE the username to westes AT uscsw.com "Johan Borkhuis" wrote in message news:Xns927A571AC94E9borkhuisagerecom@192.19.193.66... > "CHANGE username to just westes" wrote: > > > It's the generic VxWorks command line with particular commands > > disabled. > > > > What commands do I need to list files on the file system? > > With the "devs" command you can see the devices on your system. One of them > will probably be your file system. You can use the "ls" command on this > device to check. > > Kind regards, > Johan > > -- > o o o o o o o . . . _____________________________ > o _____ || Johan Borkhuis | > .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | > >(________|__|_[_________]_|__________________________| > _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` > === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: anyone out there to help me plz?very urgent help requested Date: Sat, 31 Aug 2002 06:08:54 -0400 From: Varndell Engineering Organization: Posted Via Newsfeeds.com = SPEED+RETENTION+COMPLETION = http://www.newsfeeds.com Message-ID: <3D7095B6.1AFAA372@hotmail.com> References: subbarayan wrote: > > dear mentors, > Can anyone jus tell me how to go abt proceeding the development of > device driver for pci card for x86 processor with standard pc > architecture?I use tornadoII and windows development host [snip] > Sorry if my questions look amature,but i am helpless as i am new to > this sort of coding. Wind River has an appnote (126) and tech note (49). These should answer most of your questions, and you can find them on Windsurf. Also, Windows will help you find all files (*.c;*.h) containing INCLUDE_PCI and other strings. Amateurs are inexperienced, not necessarily helpless. Regards, George Varndell - -- Varndell Engineering, LLC Embedded Systems Expertise http://www.varndellengineering.com - -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- http://www.newsfeed.com The #1 Newsgroup Service in the World! - -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: anyone out there to help me plz?very urgent help requested Date: 31 Aug 2002 07:35:51 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0208310635.481c6455@posting.google.com> References: vxnerd@rediffmail.com (subbarayan) wrote in message news:... Hi, you're lucky, on x86 (e.g. pcPentium) the system will typically (well, there are BIOSless BSPs... however I doubt you use one of them) let the BIOS configure the whole PCI stuff for you. You won't have to care too much. I guess you use a std. BSP from wrs, most likely pcPentium. You might want to have a look at one of the pci implementations, e.g. the INTEL 82557/558/559/559ER card driver. This one is called fei. Look for the macro INCLUDE_FEI. You'll e.g. find out that this macro automatically pulls in INCLUDE_PCI when activated (in config.h). In addition, look at the BSPs PCI initialization for this card. You'll find it in the file sysNetif.c (look for the *fei* functions) or in the file sysFei82557End.c (for T22). All thats necessary for a typical PCI card on x86 (w/ BIOS) is done there (find the PCI device, read out the config...). good luck, toni --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Problem about TCP link under vxWorks 5.4 Date: 31 Aug 2002 07:58:06 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0208310658.3f432dd3@posting.google.com> References: <44a01eba.0208300122.316cd7b4@posting.google.com> ellin_lin@263.net (ellin) wrote in message news:<44a01eba.0208300122.316cd7b4@posting.google.com>... Ellin, unfortunatelly this has todo with TCP specification. When the remote system resets (this is different than halting... since it will come up again, but w/o knowledge about the old TCP connection) it will return a RST (TCP RESET) packet when receiving a TCP packet destinated to the old connection. So the other side will be able to detect this condition (and close the connection properly). When the remote side goes down, all TCP frames will end up unaknowledged. Unfortunately it's impossible to decide whether the remote side is down, or whether the packet has been lost somewhere in between. The local stack will therefore retry (and in the default vxWorks configuration, it will retry forever). You should activate the socket option SO_KEEPALIVE to have the stack send keepalive probes on connections you want to monitor. You could use tcp_keepidle/tcp_keepintvl/tcp_maxidle to modify timing behaviour (global for the vxWorks system) hope this helps, toni --------------------------- Newsgroups: comp.os.vxworks Subject: Re: trouble in ping with a larger packet on ne2000 Ethernet Date: 31 Aug 2002 08:12:19 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0208310712.4fec2e1f@posting.google.com> References: <7f29c2ad.0208080439.62e501b7@posting.google.com> <3D54F62E.7060304@spamm.me.l8s.co.uk> <8400a733.0208270006.726e3f09@posting.google.com> "Martin Roth" wrote in message news:... > I have already reported this problem to Vxworks. The problem ocured on the > FEC 10/100 bps port on MPC860T family. The ne2000End.c device driver does "scatter gather" (aka "gather write") in software, since the old NE2000 is not capable of doing so in hardware. Although I still expect this to be a driver bug, this is something different... toni --------------------------- Newsgroups: comp.arch.embedded,comp.os.vxworks Subject: How to make a large project in the Tornado 2.0 IDE Date: 31 Aug 2002 20:46:24 -0700 From: love_pistol@sina.com (Michael) Organization: http://groups.google.com/ Message-ID: <6ad4da78.0208311946.7cda29cb@posting.google.com> Hi, I am developing a project with vxworks IDE, that's Tornado 2.0. Now I have known that tornado could help us build the project and create the makefile of each project. but now I wonder if my project is large and consists of many sourse file folds, I could create the makefile of each file fold by Tornado; how could I create the makefile of the whole project? could you please give me any information about this? Thank you very much for any suggestion. Michael --------------------------- Newsgroups: comp.os.vxworks Subject: TCP/ip Date: 1 Sep 2002 02:12:39 -0700 From: compaqbala@sify.com (Balakumar) Organization: http://groups.google.com/ Message-ID: <46992b69.0209010112.143f3b3f@posting.google.com> i want to know what is TCP/IP stack in embedded system. what is the difference between TCP/IP and TCP/IP stack. and one more question. I have demo version of vxworks.can i use it for tcp/ip implementation and networking protocol.my question is whether it will support or not? Please tell me. --------------------------- Newsgroups: comp.arch.embedded,comp.os.vxworks Subject: Re: How to make a large project in the Tornado 2.0 IDE Date: Sun, 01 Sep 2002 16:53:08 GMT From: "drdiags" Organization: AT&T Broadband Message-ID: References: <6ad4da78.0208311946.7cda29cb@posting.google.com> Hello Michael, On the vxWorks FAQ, there is a link to the concept of a Super Project. I have tried using the technique, and after a little setup time, it works okay. See the following link: http://www.xs4all.nl/~borkhuis/vxworks/project.html#1.7.2 "Michael" wrote in message news:6ad4da78.0208311946.7cda29cb@posting.google.com... > Hi, > > I am developing a project with vxworks IDE, that's Tornado 2.0. Now I > have known that tornado could help us build the project and create the > makefile of each project. but now I wonder if my project is large and > consists of many sourse file folds, I could create the makefile of > each file fold by Tornado; how could I create the makefile of the > whole project? > > could you please give me any information about this? > > Thank you very much for any suggestion. > > Michael --------------------------- Newsgroups: comp.os.vxworks Subject: Re: TCP/ip Date: 1 Sep 2002 19:24:24 GMT From: Denis Perelyubskiy Message-ID: References: <46992b69.0209010112.143f3b3f@posting.google.com> * Balakumar [09-Sun-02 02:12 -0700]: > > i want to know what is TCP/IP stack in embedded system. > what is the difference between TCP/IP and TCP/IP stack. > and one more question. tcp/ip stack in embedded system is the same thing (conceptually) as the tcp/ip stack of any other operating system I think "tcp/ip stack" may mean a couple of things. the meaning you're referrring to is something like a set of routines and data structures allowing a system to "speak" a tcp/ip protocol. this, of course, is not a "formal" definition. for a more formal/complete definition, google. many things come up > I have demo version of vxworks.can i use it for tcp/ip > implementation and networking protocol.my question is > whether it will support or not? don't know. sorry. someone else probably does though :) denis - -- if you want to reply by mail, please tweak my address in a reasonable way. sorry for the inconvenience. --------------------------- Newsgroups: comp.arch.embedded,comp.os.vxworks Subject: Re: How to make a large project in the Tornado 2.0 IDE Date: Sun, 1 Sep 2002 23:12:43 +0200 From: "Martin Raabe" Organization: 1&1 Internet AG Message-ID: References: <6ad4da78.0208311946.7cda29cb@posting.google.com> Hello Michael, I recommend the use of SNiFF+ as the Software Development Environment, especially for large sources. Try to get a look at 4.0.2 or the latest version! Ciao PS: No, I am no longer with WRS, so no illegal advertisements anymore! - -- Martin Raabe E: Martin.RaabeATNOSPAMBaSystem.de "Michael" schrieb im Newsbeitrag news:6ad4da78.0208311946.7cda29cb@posting.google.com... > Hi, > > I am developing a project with vxworks IDE, that's Tornado 2.0. Now I > have known that tornado could help us build the project and create the > makefile of each project. but now I wonder if my project is large and > consists of many sourse file folds, I could create the makefile of > each file fold by Tornado; how could I create the makefile of the > whole project? > > could you please give me any information about this? > > Thank you very much for any suggestion. > > Michael --------------------------- Newsgroups: comp.os.vxworks Subject: TFFS Date: 1 Sep 2002 22:24:42 -0700 From: pradeeppatel77@yahoo.com (pradeep) Organization: http://groups.google.com/ Message-ID: <54575a52.0209012124.61f54385@posting.google.com> I amusinf flas IFM 350 can u please tell me when to include TFFS when when not, is it nedded to have TFFS if i am using flash disk to load the code regards pradeep --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Basic Command Line Date: Mon, 2 Sep 2002 09:19:53 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: Reply-To: "Michael Lawnick" Sender: mlawnick@pd952440a.dip.t-dialin.net Hi Will, use - ->copy /[,0] devname : the name of device as listed per command 'devs' dirname : the main-/and subdirectory sequence The argument ',0' is optional for shell, but not within code. It will let copy move the contents of your file to stdout (fd=0). HTH Michael "CHANGE username to just westes" schrieb im Newsbeitrag news:VUZb9.306135$m91.12545814@bin5.nnrp.aus1.giganews.com... > What is the command to copy the contents of a file to the console? > > What is the fullpath expression for file A.TXT in directory B on device C? > From any arbitrary location on the disk, how would you list the contents of > A to the console? On Windows 2000 command line of course this would be any > of: > > type c:\b\a.txt // not case sensitive > > copy c:\b\a.txt con > > etc > > -- > Will > > NOTE: To reply, CHANGE the username to westes AT uscsw.com > > "Johan Borkhuis" wrote in message > news:Xns927A571AC94E9borkhuisagerecom@192.19.193.66... > > "CHANGE username to just westes" wrote: > > > > > It's the generic VxWorks command line with particular commands > > > disabled. > > > > > > What commands do I need to list files on the file system? > > > > With the "devs" command you can see the devices on your system. One of > them > > will probably be your file system. You can use the "ls" command on this > > device to check. > > > > Kind regards, > > Johan > > > > -- > > o o o o o o o . . . _____________________________ > > o _____ || Johan Borkhuis | > > .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | > > >(________|__|_[_________]_|__________________________| > > _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` > > === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Basic Command Line Date: Mon, 2 Sep 2002 07:42:40 +0000 (UTC) From: Johan Borkhuis Organization: Agere Systems Message-ID: References: "CHANGE username to just westes" wrote: > What is the command to copy the contents of a file to the console? copy < file.txt > What is the fullpath expression for file A.TXT in directory B on > device C? From any arbitrary location on the disk, how would you list > the contents of A to the console? On Windows 2000 command line of > course this would be any of: > > type c:\b\a.txt // not case sensitive > > copy c:\b\a.txt con The filename would be C/B/A.TXT Some examples: - -> pwd host: value = 6 = 0x6 - -> ls .... test.txt .... - -> copy < test.txt This is a test value = 0 = 0x0 - -> copy < host:test.txt This is a test value = 0 = 0x0 - -> copy < host:/Tornado/test.txt This is a test value = 0 = 0x0 - -> copy < host:D:/Tornado/test.txt This is a test value = 0 = 0x0 (host: points to the Tornado directory, this is the default directory of my FTP server) Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: TFFS Date: Mon, 2 Sep 2002 10:32:43 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <54575a52.0209012124.61f54385@posting.google.com> You can't be using that. You either use an IFD-350-xxx or an IFM-4xx, http://www.eagle.co.za/display_cat_155.htm Whereever I looked, IFM-350 does not exist. In case it is an IFD-350, which is a Flash based IDE disk, you do not neet TFFS, just the standard setup that will work with any IDE hard disk. "pradeep" wrote in message news:54575a52.0209012124.61f54385@posting.google.com... > I amusinf flas IFM 350 > can u please tell me when to include TFFS when when not, is it nedded > to have TFFS if i am using flash disk to load the code > > regards > pradeep --------------------------- Newsgroups: comp.os.vxworks Subject: Re: TCP/ip Date: Mon, 2 Sep 2002 10:34:12 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <46992b69.0209010112.143f3b3f@posting.google.com> The "free" version of Tornado/VxSim that used to be available off the WindRiver web site (not anymore) has a stripped down version of VxSim, which does not have the TCP/IP stack. "Denis Perelyubskiy" wrote in message news:slrnan4qb8.5om.denis_newsREMOVE-TO-REPLY@finch.lecs.cs.ucla.edu... > * Balakumar [09-Sun-02 02:12 -0700]: > > > > i want to know what is TCP/IP stack in embedded system. > > what is the difference between TCP/IP and TCP/IP stack. > > and one more question. > > tcp/ip stack in embedded system is the same thing > (conceptually) as the tcp/ip stack of any other operating > system > > I think "tcp/ip stack" may mean a couple of things. the > meaning you're referrring to is something like a set of > routines and data structures allowing a system to "speak" a > tcp/ip protocol. > > this, of course, is not a "formal" definition. > for a more formal/complete definition, google. many things > come up > > > I have demo version of vxworks.can i use it for tcp/ip > > implementation and networking protocol.my question is > > whether it will support or not? > > don't know. sorry. someone else probably does though :) > > denis > > -- > if you want to reply by mail, please tweak my > address in a reasonable way. sorry for the > inconvenience. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: no longer with WRS Date: Mon, 2 Sep 2002 10:37:31 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <6ad4da78.0208311946.7cda29cb@posting.google.com> Join the club ! ;-)) > > > PS: No, I am no longer with WRS, so no illegal advertisements anymore! > > -- > Martin Raabe --------------------------- Newsgroups: comp.os.vxworks Subject: Is it the hardware problem or software problem? Samsung + Intersil wireless card Date: 2 Sep 2002 04:18:48 -0700 From: thomas_yangt@sina.com (thomas yang) Organization: http://groups.google.com/ Message-ID: <2349d304.0209020318.2cfd2e50@posting.google.com> cpu: Samsumg arm os : vxworks wireless card: intersil the other side is lan card. function: bridge if the throughput is big for long time, about 5 hours. the intersil wireless card will be down. 1. After it is down, there are no packets received from wireless card. The intersil interrupt status is always 0, which indicate there are no interrupt come in. 2. The task and the memory pool are ok. Other functions are all ok. 3. When the network is down, we can send packets out from wireless land successfully. (proved by getting the packets from other machine) 4. If we disabel the intersil card and then enable it, it will work again. How can I decide if it is the hardware problem or software problem? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Is it the hardware problem or software problem? Samsung + Intersil wireless card Date: Mon, 2 Sep 2002 13:28:20 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <2349d304.0209020318.2cfd2e50@posting.google.com> Thomas, If you call the intersil driver interrupt handler manually, will it revive the WLAN card ? If this works, you have a classic case of missed interrupt, which is a hardware-software integration problem (yup, this is how I call this when hardware folks have no consideration for software limitations and/or vice versa). Contact me if this is really your problem, and I will try to help. - - Leonid "thomas yang" wrote in message news:2349d304.0209020318.2cfd2e50@posting.google.com... > cpu: Samsumg arm > os : vxworks > wireless card: intersil > the other side is lan card. > function: bridge > > > > if the throughput is big for long time, about 5 hours. > the intersil wireless card will be down. > > > 1. After it is down, there are no packets received from wireless card. > The intersil interrupt status is always 0, which indicate there are > no interrupt come in. > > 2. The task and the memory pool are ok. Other functions are all ok. > > 3. When the network is down, we can send packets out from wireless > land successfully. > (proved by getting the packets from other machine) > > 4. If we disabel the intersil card and then enable it, it will work > again. > > How can I decide if it is the hardware problem or software problem? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Does any know what "panic : soclose : NOFDREF" mean ........??? Date: 2 Sep 2002 04:48:50 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0209020348.1cf5ac8e@posting.google.com> References: mohasin.zaki@wipro.com (Mohasin) wrote in message news:... > "(TASKNAME) - panic : soclose" are you using this socket within a select statment? do you close the socket while a select is pending on it? if so, try not to close sockets with pending selects on them... toni --------------------------- Newsgroups: comp.os.vxworks Subject: communication with i/ocards in cpci chassis Date: 2 Sep 2002 06:28:02 -0700 From: be_gyus@yahoo.com (james) Organization: http://groups.google.com/ Message-ID: Hi Techis, In a typical cpci chassis can a sbc card (also a system controller) communicate with any i/o plugged on the same chassis.... or if atall it communicates how is it doing .... what interface does it use in achieving it?? Any suggestions will be appreciated. --------------------------- Newsgroups: comp.os.vxworks Subject: Gurus out there please help -- compact flash Date: 2 Sep 2002 09:52:35 -0700 From: ritalin@comcast.net (Rita) Organization: http://groups.google.com/ Message-ID: I'm having problem mounting DOS FS on my compact flash. When it was fresh from the factory, I was able to mount DOS FS, and did a "mkdir" on it. Then windRiver engineer told me that I was using the wrong usrAtaConfig.c. (The one I used was pre-dos 2.0.) After using the "correct" usrAtaConfig.c, the dosFs didn't mount, and was complaining about "No partition table found, and malformed boot sector." I then used "usrFdiskCreate()" as suggested to create a single partition table. When trying to mount the dos fs, it again complaind about "Malformed boot sector." I found that the partition table sits in sector 0, and there was no dos boot sector. Also, I found that the dosFsSecGet() was looking at sector 0x20 for the boot sector information. I think there is some serious screw up here. I read somewhere in usrFdiskLib.c (I don't remember exactly where), that I don't need a partition table if I'm using the entire disk as a whole drive. Anyway, I formatted the disk under w2k, and dumped the first sector to see what a dos boot sector looks like, and then tried to boot it up from vxWorks. It didn't work. It still prints out error message, "Malformed boot sector." I think I might have started out from the wrong side of the road. I got couple of other compact flashes, but none contains factory format. They were all formatted under w2k. (I forgot to mention that I'm using it with a mips board.) Can someone suggest how to solve this problem? My questions are, 1) do I need a partition table, if not, what should I do. 2) how does the dosFs calculate the location of the boot sector? 3) where can I get an image of a dos boot sector, or should I write my own? (I have the FAT16 document for the boot sector.) 4) If I write my own dos boot sector, do I have to create a FAT root directory? Please help me out. I've never done anything like this before, and am really stuck in mounting dos fs. Thanks a lot in advance. Rita --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Gurus out there please help -- compact flash Date: Mon, 2 Sep 2002 19:18:09 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Rita, The WRS engineer was in a way right to suggest you use a newer initalization code, but you need to look into this code and make sure it does everything the way YOU need it done. Generally, concerning CompactFlash, I recommend you first mount an duse it with its detault factiry format, or after its been formatted by a PC. So you do not need to use usrFdiskFormat(), but in the initialization code, you need to use dpartLib, which handles partition tables for DosFs2.0. Yu should find examples of using dpartLib with dosFs in the on-line and manual and DosFs component (or release) manual. Only if you are building an appliance which MUST be able to format a CF card (all my digital camreas do that), will you need to get your teeth into usrFdiskFormat, which is delivered to you in source code, so you can make it work the way you like it to. Hope this helps, Leonid Rosenboim P.S. I wrote usrFdiskFormat() as wellas dcacheLib, so please be gentle. "Rita" wrote in message news:d3a5605c.0209020852.a0a788b@posting.google.com... > I'm having problem mounting DOS FS on my compact flash. When it was > fresh from > the factory, I was able to mount DOS FS, and did a "mkdir" on it. Then > windRiver > engineer told me that I was using the wrong usrAtaConfig.c. (The one I > used was pre-dos 2.0.) After using the "correct" usrAtaConfig.c, the > dosFs didn't mount, and was complaining about "No partition table > found, and malformed boot sector." > I then used "usrFdiskCreate()" as suggested to create a single > partition table. When trying to mount the dos fs, it again complaind > about "Malformed boot sector." > > I found that the partition table sits in sector 0, and there was no > dos boot sector. Also, I found that the dosFsSecGet() was looking at > sector 0x20 for > the boot sector information. I think there is some serious screw up > here. I read somewhere in usrFdiskLib.c (I don't remember exactly > where), that > I don't need a partition table if I'm using the entire disk as a whole > drive. > > Anyway, I formatted the disk under w2k, and dumped the first sector to > see what a dos boot sector looks like, and then tried to boot it up > from vxWorks. It didn't work. It still prints out error message, > "Malformed boot sector." > > I think I might have started out from the wrong side of the road. I > got couple of other compact flashes, but none contains factory format. > They were all formatted under w2k. (I forgot to mention that I'm using > it with a mips board.) > Can someone suggest how to solve this problem? > > My questions are, > 1) do I need a partition table, if not, what should I do. > 2) how does the dosFs calculate the location of the boot sector? > 3) where can I get an image of a dos boot sector, or should I write my > own? (I have the FAT16 document for the boot sector.) > 4) If I write my own dos boot sector, do I have to create a FAT root > directory? > > Please help me out. I've never done anything like this before, and am > really stuck in mounting dos fs. > > Thanks a lot in advance. > Rita --------------------------- Newsgroups: comp.os.vxworks Subject: Intel Web Pad - Help! Date: 2 Sep 2002 12:17:11 -0700 From: cma371@aol.com (cma371) Organization: http://groups.google.com/ Message-ID: <1e487d07.0209021117.7b75c676@posting.google.com> Anybody know how to get out of Demo mode on an Intel Web Pad running Wind River OS? Thanks --------------------------- Newsgroups: comp.os.vxworks Subject: [Q] ppc405, no snoop, cacheDmaMalloc Date: 2 Sep 2002 19:17:52 -0700 From: kotaeji@hanmail.net (Kim Tae Hyung) Organization: http://groups.google.com/ Message-ID: Hi. I have a ppc405 board. And two dec21x4x pci enthernet cards are connected through ppc405 pci interface. Device Driver of dec21x4x uses cacheDmaMalloc to allocate memory for rx,tx descriptors and network clusters. But I should replace cacheDmaMalloc with malloc(or memalign) for some reasons. By the way, ppc405 has no snooping ability. I want to know about the mechanism of cacheDmaMalloc. Should I disable data cache for the memory? Thanks in advance. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Gurus out there please help -- compact flash Date: 2 Sep 2002 19:54:29 -0700 From: ritalin@comcast.net (Rita) Organization: http://groups.google.com/ Message-ID: References: Hi leonid, Thank you for your reply. The newer initialization code does use dpartLib and dcacheLib as part of the initialization before calling dosFsDevInit(). I looked into usrFdiskLib.c and did not see usrFdiskFormat(). There are only 3 routines, one is to create a partition table, one to read it, and one to parse it. I will check with WRS engineer to see if I have the correct source code to work with dosFs2.0. I assume that usrFdiskFormat() calls usrFdiskCreate() to create a partition table, as well as creating a dos boot sector. The appliance I build needs a format tool, so I will need to include usrFdiskFormat(). I just got hold of dosFs2.0 manual, and am getting documentation on dcacheLib and dpartLib. I will give you a followup if I have any breakthrough. Thanks for your suggestion. P.S. *ouch* didn't know you wrote usrFdiskFormat(), sorry, didn't mean it bad. I was getting frustrated at being stuck in mounting dosFs for a while. Rita "Leonid Rosenboim" wrote in message news:... > Rita, > > The WRS engineer was in a way right to suggest you use a newer initalization > code, but you need to look into this code and make sure it does everything > the way YOU need it done. > > Generally, concerning CompactFlash, I recommend you first mount an duse it > with its detault factiry format, or after its been formatted by a PC. > > So you do not need to use usrFdiskFormat(), but in the initialization code, > you need to use dpartLib, which handles partition tables for DosFs2.0. > Yu should find examples of using dpartLib with dosFs in the on-line and > manual > and DosFs component (or release) manual. > > Only if you are building an appliance which MUST be able to format a CF card > (all my digital camreas do that), will you need to get your teeth into > usrFdiskFormat, which is delivered to you in source code, so you can make it > work the way you like it to. > > Hope this helps, > Leonid Rosenboim > > P.S. I wrote usrFdiskFormat() as wellas dcacheLib, so please be gentle. > > "Rita" wrote in message > news:d3a5605c.0209020852.a0a788b@posting.google.com... > > I'm having problem mounting DOS FS on my compact flash. When it was > > fresh from > > the factory, I was able to mount DOS FS, and did a "mkdir" on it. Then > > windRiver > > engineer told me that I was using the wrong usrAtaConfig.c. (The one I > > used was pre-dos 2.0.) After using the "correct" usrAtaConfig.c, the > > dosFs didn't mount, and was complaining about "No partition table > > found, and malformed boot sector." > > I then used "usrFdiskCreate()" as suggested to create a single > > partition table. When trying to mount the dos fs, it again complaind > > about "Malformed boot sector." > > > > I found that the partition table sits in sector 0, and there was no > > dos boot sector. Also, I found that the dosFsSecGet() was looking at > > sector 0x20 for > > the boot sector information. I think there is some serious screw up > > here. I read somewhere in usrFdiskLib.c (I don't remember exactly > > where), that > > I don't need a partition table if I'm using the entire disk as a whole > > drive. > > > > Anyway, I formatted the disk under w2k, and dumped the first sector to > > see what a dos boot sector looks like, and then tried to boot it up > > from vxWorks. It didn't work. It still prints out error message, > > "Malformed boot sector." > > > > I think I might have started out from the wrong side of the road. I > > got couple of other compact flashes, but none contains factory format. > > They were all formatted under w2k. (I forgot to mention that I'm using > > it with a mips board.) > > Can someone suggest how to solve this problem? > > > > My questions are, > > 1) do I need a partition table, if not, what should I do. > > 2) how does the dosFs calculate the location of the boot sector? > > 3) where can I get an image of a dos boot sector, or should I write my > > own? (I have the FAT16 document for the boot sector.) > > 4) If I write my own dos boot sector, do I have to create a FAT root > > directory? > > > > Please help me out. I've never done anything like this before, and am > > really stuck in mounting dos fs. > > > > Thanks a lot in advance. > > Rita --------------------------- Newsgroups: comp.os.vxworks Subject: Target directory listing Date: Tue, 03 Sep 2002 04:44:08 GMT From: "weddick" Organization: AT&T Broadband Message-ID: I have a new system that I am unable to perform a @ls "/ide0/" command that I am able to do on other systems. What did I leave out in the new bsp? Thanks, Joel --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Target directory listing Date: Tue, 3 Sep 2002 06:38:47 +0000 (UTC) From: Johan Borkhuis Organization: Agere Systems Message-ID: References: "weddick" wrote: > I have a new system that I am unable to perform a @ls "/ide0/" command > that I am able to do on other systems. What did I leave out in the > new bsp? Can you do a ls "/ide0/" command on the target shell? Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: [Q] ppc405, no snoop, cacheDmaMalloc Date: Tue, 3 Sep 2002 08:40:46 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: Reply-To: "Michael Lawnick" Sender: mlawnick@pd9524d77.dip.t-dialin.net Hi Kim, cacheDmaMalloc is used to malloc memory and mark it as uncacheable. If you can't use the function, you'd have to call MMU-functions - which are not present if you haven't bought full MMU support :-( So I see only one way: You will have to call cacheFlush after every write function (before telling NIC-H/W that there is a new packet to send) and cacheInvalidate before every read function. This method works, I've done it for a DP83815. Take care: Buffers may not share cache lines so adjust your buffer sizes and addresses properly! Switching off data cache for complete target is a knock out, your target will slow down to snail. BTW: The same discussion occured some months ago (last year?). Do a google search. HTH Michael "Kim Tae Hyung" schrieb im Newsbeitrag news:b1b90dae.0209021817.5a8f5b46@posting.google.com... > Hi. > I have a ppc405 board. And two dec21x4x pci enthernet cards are connected > through ppc405 pci interface. > > Device Driver of dec21x4x uses cacheDmaMalloc to allocate memory for > rx,tx descriptors and network clusters. But I should replace cacheDmaMalloc with > malloc(or memalign) for some reasons. > > By the way, ppc405 has no snooping ability. > I want to know about the mechanism of cacheDmaMalloc. > Should I disable data cache for the memory? > > Thanks in advance. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Gurus out there please help -- compact flash Date: Tue, 3 Sep 2002 10:53:23 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Rita, Here is an example intialization code from teh DosFs2.0 manual: It assumes the disk is execpected to have at most three partitions, and does not include the block device driver intialization. usrPartDiskFsInit( void * blkDevId ) { const char * devNames[] = { "/sd0a", "/sd0b", "/sd0c" }; int dcacheSize = 0x30000 ; void * cbio, * cbio1 ; /* create disk cache */ if( (cbio = dcacheDevCreate(blkDevId, NULL, dcacheSize, "/sd0")) == NULL ) return ERROR ; /* create partition manager with FDISK style decoder, up to 3 parts */ cbio1 = dpartDevCreate( cbio, 3, usrFdiskPartRead ); /* create the 1st file system, 8 simult. open files, with CHKDSK */ dosFsDevCreate( devNames[0], dpartPartGet(cbio1,0), 8, 0 ); /* create the 2nd file system, 6 simult. open files, with CHKDSK */ dosFsDevCreate( devNames[1], dpartPartGet(cbio1,1), 6, 0 ); /* create the 3rd file system, 4 simultaneously open files, no CHKDSK */ dosFsDevCreate( devNames[2], dpartPartGet(cbio1,2), 4, NONE ); } Do note that old ATA intialization code had some partition handling code of its own (because DosFs 1.0 did not handle partitions) and you need to make sure that the ATA driver intialization code does not do anything like that, and that the block device "sees" the entire CF card, starting from block 0. As to formatting, you are right, useFdiskPartCreate() creates the partition table, then you need to call dosFsVolFormat() to actually initialize the dosFs. I fyou do not plan on booting from the CF (even PCs dont use CF or PCMCIA for booting normally), the code in usrFdiskLib should satisfy your needs. "Rita" wrote in message news:d3a5605c.0209021854.42796f97@posting.google.com... > Hi leonid, > > Thank you for your reply. The newer initialization code does use > dpartLib and dcacheLib as part of the initialization before calling > dosFsDevInit(). I looked into usrFdiskLib.c and did not see > usrFdiskFormat(). There are only 3 routines, one is to create a > partition table, one to read it, and one to parse it. I will check > with WRS engineer to see if I have the correct source code to work > with dosFs2.0. > > I assume that usrFdiskFormat() calls usrFdiskCreate() to create a > partition table, as well as creating a dos boot sector. The appliance > I build needs a format tool, so I will need to include > usrFdiskFormat(). I just got hold of dosFs2.0 manual, and am getting > documentation on dcacheLib and dpartLib. I will give you a followup if > I have any breakthrough. > > Thanks for your suggestion. > P.S. *ouch* didn't know you wrote usrFdiskFormat(), sorry, didn't mean > it bad. > I was getting frustrated at being stuck in mounting dosFs for a while. > Rita > > "Leonid Rosenboim" wrote in message news:... > > Rita, > > > > The WRS engineer was in a way right to suggest you use a newer initalization > > code, but you need to look into this code and make sure it does everything > > the way YOU need it done. > > > > Generally, concerning CompactFlash, I recommend you first mount an duse it > > with its detault factiry format, or after its been formatted by a PC. > > > > So you do not need to use usrFdiskFormat(), but in the initialization code, > > you need to use dpartLib, which handles partition tables for DosFs2.0. > > Yu should find examples of using dpartLib with dosFs in the on-line and > > manual > > and DosFs component (or release) manual. > > > > Only if you are building an appliance which MUST be able to format a CF card > > (all my digital camreas do that), will you need to get your teeth into > > usrFdiskFormat, which is delivered to you in source code, so you can make it > > work the way you like it to. > > > > Hope this helps, > > Leonid Rosenboim > > > > P.S. I wrote usrFdiskFormat() as wellas dcacheLib, so please be gentle. > > > > "Rita" wrote in message > > news:d3a5605c.0209020852.a0a788b@posting.google.com... > > > I'm having problem mounting DOS FS on my compact flash. When it was > > > fresh from > > > the factory, I was able to mount DOS FS, and did a "mkdir" on it. Then > > > windRiver > > > engineer told me that I was using the wrong usrAtaConfig.c. (The one I > > > used was pre-dos 2.0.) After using the "correct" usrAtaConfig.c, the > > > dosFs didn't mount, and was complaining about "No partition table > > > found, and malformed boot sector." > > > I then used "usrFdiskCreate()" as suggested to create a single > > > partition table. When trying to mount the dos fs, it again complaind > > > about "Malformed boot sector." > > > > > > I found that the partition table sits in sector 0, and there was no > > > dos boot sector. Also, I found that the dosFsSecGet() was looking at > > > sector 0x20 for > > > the boot sector information. I think there is some serious screw up > > > here. I read somewhere in usrFdiskLib.c (I don't remember exactly > > > where), that > > > I don't need a partition table if I'm using the entire disk as a whole > > > drive. > > > > > > Anyway, I formatted the disk under w2k, and dumped the first sector to > > > see what a dos boot sector looks like, and then tried to boot it up > > > from vxWorks. It didn't work. It still prints out error message, > > > "Malformed boot sector." > > > > > > I think I might have started out from the wrong side of the road. I > > > got couple of other compact flashes, but none contains factory format. > > > They were all formatted under w2k. (I forgot to mention that I'm using > > > it with a mips board.) > > > Can someone suggest how to solve this problem? > > > > > > My questions are, > > > 1) do I need a partition table, if not, what should I do. > > > 2) how does the dosFs calculate the location of the boot sector? > > > 3) where can I get an image of a dos boot sector, or should I write my > > > own? (I have the FAT16 document for the boot sector.) > > > 4) If I write my own dos boot sector, do I have to create a FAT root > > > directory? > > > > > > Please help me out. I've never done anything like this before, and am > > > really stuck in mounting dos fs. > > > > > > Thanks a lot in advance. > > > Rita --------------------------- Newsgroups: comp.os.vxworks Subject: Regarding Stack Size in Task Spawn Date: 3 Sep 2002 03:08:06 -0700 From: vxworksonline@yahoo.com (Shiny) Organization: http://groups.google.com/ Message-ID: Hi, Is there any problem in case we don't give any task size while spawning a Task(0). For a simple program it didn't give any problem.Will there be a problem in Real time. Do clear. Shiny. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Regarding Stack Size in Task Spawn Date: Tue, 03 Sep 2002 06:40:55 -0400 From: joe durusau Organization: WEBUSENET.com Message-ID: <3D7491B7.6A4BE41A@bellsouth.net> References: Shiny wrote: > Hi, > Is there any problem in case we don't give any task size while > spawning a Task(0). > For a simple program it didn't give any problem.Will there be a > problem in Real time. > Do clear. > Shiny. If you don't specify, you get a default size. It can be a problem. It all depends on what the task in question does. It is much better to make intelligent decisions, and not use defaults. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: where can I have free downloadable Tornado II Date: 3 Sep 2002 06:07:38 -0700 From: gofukdiya@yahoo.co.in (bhanu prakash) Organization: http://groups.google.com/ Message-ID: <8076f4da.0209030507.73231afc@posting.google.com> Hi, I would like to know where can I have freely downloadable Tornado II simpc. Recently I had an interactive demo class,I came to know later that its freely downloadable and can be used for some 2 months.Through this we can see the graphical pictures of semophores give and take,scheduling and we can implement all the VxWorks API in it.Its actually a simulator of BSP.Can any one give pointers where can I get this?I went through www.windriver.com but couldn't able to get it. Thanks and rgds, Bhanu Prakash --------------------------- Newsgroups: comp.os.vxworks Subject: Implementation of intHandlerCreate() Date: 3 Sep 2002 06:56:58 -0700 From: fxcrazy@yahoo.com (Tom) Organization: http://groups.google.com/ Message-ID: <633e72d2.0209030556.4e201144@posting.google.com> Hi, I am just inquisitive to know how intHandlerCreate() is implemented. Anybody has an idea ? This call accepts a 'C' routine and turns it as an interupt handler which we can register through an intVecSet(). what that 'extra' stuff would be which makes this 'C' routine an Interrupt Handler ? - - Tom --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Implementation of intHandlerCreate() Date: Tue, 3 Sep 2002 17:13:28 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <633e72d2.0209030556.4e201144@posting.google.com> Actually it is rather simple - intHandlerCreate() allocates a small chunk of memory, which is filled with machine isntructions, specific as per a given architectures that saves all processor context, which is not normally saved by the "C" compiled routines in the stack. It also calls a kernel entry point announcing the interrupt entry to the kernel "intEtner". In the middle of the generated function si a call to your C routine. Then it will restore all the processor xontext, and will call a kernet intExit routine, which is vcery important, it will execute some key kernel code as a result of system calls made by the ISR. You can simply disassemble the address returned by intHandlerCreate(). intConnect() is simply a call to intHandlerCreate() and then to intVecSet(). "Tom" wrote in message news:633e72d2.0209030556.4e201144@posting.google.com... > Hi, > > I am just inquisitive to know how intHandlerCreate() is implemented. > Anybody has an idea ? > > This call accepts a 'C' routine and turns it as an interupt handler > which we can register through an intVecSet(). > > what that 'extra' stuff would be which makes this 'C' routine an > Interrupt Handler ? > > - Tom --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Gurus out there please help -- compact flash Date: 3 Sep 2002 08:29:55 -0700 From: ritalin@comcast.net (Rita) Organization: http://groups.google.com/ Message-ID: References: Leonid, Thanks for your help. I think I miss out a step. - --> dosFsVolFormat() after dosFsDevCreate(). I used dosFsDevInit(), and passed in NULL for DOS_VOL_CONFIG *. That might have been a problem. Currently, I'm not booting up from the compact flash, and hopefully, I won't need to do that. This is what I have right now (printf results from initialization.) - ->ataDevCreate: usrAtaconfig: cbio=0x8efb6050 /* my printf */ usrFdiskPartCraete(0x8efb6050, 1, 0, 0, 0) /* my printf */ usrAtaConfig: dpartDevCreate((FUNCPTR)usrFdiskPartRead) /* my printf */ usrFdiskPartLib warning: disk partition table is not intiialized /* vxWorks */ usrFdiskPartLib warning: use usrFdiskPartCreate(0x8efb6050, 1) to initialize the disk as a single partition Retrieved old volume params with %63 confidence: Volume Prameters: FAT type: FAT16, sectors per cluster 114 116 FAT copies, 0, clusters, 18798 sectors per FAT Sectors reserved 21369, hidden 0, FAT sectors 2180568 Root dir entries 28019, sysId (null), serial numbger 74727450 Label:" "... Diskwith 503808 sectors of 512 bytes will be formatted with: Volume Parameters: FAT type: FAT16, sectors per cluster 8 2 FAT copies, 62910 clusters, 246 sectors per FAT Sectors reserved 1, hidden 0, FAT sectors 492 Root dir entries 512, sysId VXDOS16, serial number 74727450 Label:" "... This is the dump of sector 0: [0x0000] 3eeb 5690 4458 534f 3631 0000 0802 0001 *>.V.DXSO61'.'.'.* [0x0010] 0002 0002 f800 00f6 0020 0010 0000 0000 *'.'.'.'.' '.'.'.* [0x0020] b000 0007 0080 5029 7274 0074 0000 0000 *'.'.'.P)rt't'.'.* [0x0030] 0000 0000 0000 4146 3154 2036 2020 0000 *'.'.'.AF1T 6 '.* [0x0040] 0000 0000 0000 0000 0000 0000 0000 0000 *'.'.'.'.'.'.'.'.* ...... .... same as 0x0040 all the way til 0x1e0... [0x01f0] 0000 0000 0000 0000 0000 0000 0000 aa55 *'.'.'.'.'.'.'.'U* Read out reversed: [0x0000] eb3e 9056 5844 4f53 3136 0000 0208 0100 *'>'VXDOS16'.'.'.* [0x0010] 0200 0200 00f8 f600 2000 1000 0000 0000 *'.'.'.'.' '.'.'.* [0x0020] 00b0 0700 8000 2950 7472 7400 0000 0000 *'.'.'.)Ptrt.'.'.* [0x0030] 0000 0000 0000 4641 5431 3620 2020 0000 *'.'.'.FAT16 '.* [0x0040] 0000 0000 0000 0000 0000 0000 0000 0000 *'.'.'.'.'.'.'.'.* ...... .... same as 0x0040 all the way til 0x1e0... [0x01f0] 0000 0000 0000 0000 0000 0000 0000 55aa *'.'.'.'.'.'.'.'U* I think the byte order is wrong. Please let me know which one is the correct one. Since I'll be the only one writing and reading out from the compact flash, I didn't think I need to do byte swapping. I could be wrong. Thanks so much for all your help. Rita "Leonid Rosenboim" wrote in message news:... > Rita, > > Here is an example intialization code from teh DosFs2.0 manual: > > It assumes the disk is execpected to have at most three partitions, > and does not include the block device driver intialization. > > > usrPartDiskFsInit( void * blkDevId ) > { > const char * devNames[] = { "/sd0a", "/sd0b", "/sd0c" }; > int dcacheSize = 0x30000 ; > void * cbio, * cbio1 ; > /* create disk cache */ > if( (cbio = dcacheDevCreate(blkDevId, NULL, dcacheSize, "/sd0")) > == NULL ) > return ERROR ; > /* create partition manager with FDISK style decoder, up to 3 parts */ > cbio1 = dpartDevCreate( cbio, 3, usrFdiskPartRead ); > /* create the 1st file system, 8 simult. open files, with CHKDSK */ > dosFsDevCreate( devNames[0], dpartPartGet(cbio1,0), 8, 0 ); > /* create the 2nd file system, 6 simult. open files, with CHKDSK */ > dosFsDevCreate( devNames[1], dpartPartGet(cbio1,1), 6, 0 ); > /* create the 3rd file system, 4 simultaneously open files, no CHKDSK */ > dosFsDevCreate( devNames[2], dpartPartGet(cbio1,2), 4, NONE ); > } > > Do note that old ATA intialization code had some partition handling code of > its own (because DosFs 1.0 did not handle partitions) and you need to make > sure that the ATA driver intialization code does not do anything like that, > and that the block device "sees" the entire CF card, starting from block 0. > > As to formatting, you are right, useFdiskPartCreate() creates the partition > table, > then you need to call dosFsVolFormat() to actually initialize the dosFs. > I fyou do not plan on booting from the CF (even PCs dont use CF or PCMCIA > for booting normally), the code in usrFdiskLib should satisfy your needs. > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VMIC 6001 with Vxworks Date: 3 Sep 2002 08:43:20 -0700 From: terry.butman@vmic.com (Terry Butman) Organization: http://groups.google.com/ Message-ID: <6db76185.0209030743.6396707f@posting.google.com> References: <3d6ebfe6_3@nopics.sjc> "Cedric Le Roux" wrote in message news:<3d6ebfe6_3@nopics.sjc>... > I'm looking for some clarification regarding the programming for diagnostic > mode for the VMIC 6001 MIL-STD-1553 board. The manual is kinda vague. > > If anyone out there has worked with this board, I'd like to hear from ya. > or if anyone knows a better place for this post, I'd like to hear about that > too. > > Thanks. What exactly do you want to do? The diagnostics mode was designed to test the ability to read/write data through the 1553 interface. As best as I recall you do this with a series of diagnostic blocks which include: the number of memory locations to test the number of 1553 data words to test whether or not the data goes through the full data path or just through the transmitter/receiver etc. As I recall there is a reserved memory space in the diagnostic control blocks for what is called the "echo". It goes something like this. memory data written memory data read (echo) 1553 data written 1553 data read (echo) I seem to recall that the 1553 data might include the transmission status information and the interrupt status information. Any further details and I'll have to dig out the manual. regards, Terry --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Gurus out there please help -- compact flash Date: Tue, 3 Sep 2002 17:44:57 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Tell you what - put in a CF formatted on a PC, and dump its memory in a byte width: -> d address,1 But from this stuff it seems there is indeed a btre swapping issue ! Your CompactFlash socket (sometimes its actual PCMCIA) has a 16 bit bus, and if connected to the MIPS (Big Enian) bus naively, you will get every two bytes order reversed, because the standard PCMCIA (and CF) bus layout is Little Endian. run a swab() function over each sector at the driver level, and see what happens. - - Leonid "Rita" wrote in message news:d3a5605c.0209030729.36d7f843@posting.google.com... > Leonid, > > Thanks for your help. I think I miss out a step. > --> dosFsVolFormat() after dosFsDevCreate(). I used dosFsDevInit(), > and > passed in NULL for DOS_VOL_CONFIG *. That might have been a problem. > Currently, I'm not booting up from the compact flash, and hopefully, I > won't need to do that. > > This is what I have right now (printf results from initialization.) > ->ataDevCreate: > usrAtaconfig: cbio=0x8efb6050 /* my printf */ > usrFdiskPartCraete(0x8efb6050, 1, 0, 0, 0) /* my printf */ > usrAtaConfig: dpartDevCreate((FUNCPTR)usrFdiskPartRead) /* my printf > */ > usrFdiskPartLib warning: disk partition table is not intiialized /* > vxWorks */ > usrFdiskPartLib warning: use usrFdiskPartCreate(0x8efb6050, 1) to > initialize the disk as a single partition > > Retrieved old volume params with %63 confidence: > Volume Prameters: FAT type: FAT16, sectors per cluster 114 > 116 FAT copies, 0, clusters, 18798 sectors per FAT > Sectors reserved 21369, hidden 0, FAT sectors 2180568 > Root dir entries 28019, sysId (null), serial numbger 74727450 > Label:" "... > > Diskwith 503808 sectors of 512 bytes will be formatted with: > Volume Parameters: FAT type: FAT16, sectors per cluster 8 > 2 FAT copies, 62910 clusters, 246 sectors per FAT > Sectors reserved 1, hidden 0, FAT sectors 492 > Root dir entries 512, sysId VXDOS16, serial number 74727450 > Label:" "... > > This is the dump of sector 0: > [0x0000] 3eeb 5690 4458 534f 3631 0000 0802 0001 > *>.V.DXSO61'.'.'.* > [0x0010] 0002 0002 f800 00f6 0020 0010 0000 0000 *'.'.'.'.' > '.'.'.* > [0x0020] b000 0007 0080 5029 7274 0074 0000 0000 > *'.'.'.P)rt't'.'.* > [0x0030] 0000 0000 0000 4146 3154 2036 2020 0000 *'.'.'.AF1T 6 > '.* > [0x0040] 0000 0000 0000 0000 0000 0000 0000 0000 > *'.'.'.'.'.'.'.'.* > ...... .... same as 0x0040 all the way til 0x1e0... > [0x01f0] 0000 0000 0000 0000 0000 0000 0000 aa55 > *'.'.'.'.'.'.'.'U* > > Read out reversed: > [0x0000] eb3e 9056 5844 4f53 3136 0000 0208 0100 > *'>'VXDOS16'.'.'.* > [0x0010] 0200 0200 00f8 f600 2000 1000 0000 0000 *'.'.'.'.' > '.'.'.* > [0x0020] 00b0 0700 8000 2950 7472 7400 0000 0000 > *'.'.'.)Ptrt.'.'.* > [0x0030] 0000 0000 0000 4641 5431 3620 2020 0000 *'.'.'.FAT16 > '.* > [0x0040] 0000 0000 0000 0000 0000 0000 0000 0000 > *'.'.'.'.'.'.'.'.* > ...... .... same as 0x0040 all the way til 0x1e0... > [0x01f0] 0000 0000 0000 0000 0000 0000 0000 55aa > *'.'.'.'.'.'.'.'U* > > I think the byte order is wrong. Please let me know which one is the > correct one. Since I'll be the only one writing and reading out from > the compact flash, I didn't think I need to do byte swapping. I could > be wrong. > > Thanks so much for all your help. > Rita > > > "Leonid Rosenboim" wrote in message news:... > > Rita, > > > > Here is an example intialization code from teh DosFs2.0 manual: > > > > It assumes the disk is execpected to have at most three partitions, > > and does not include the block device driver intialization. > > > > > > usrPartDiskFsInit( void * blkDevId ) > > { > > const char * devNames[] = { "/sd0a", "/sd0b", "/sd0c" }; > > int dcacheSize = 0x30000 ; > > void * cbio, * cbio1 ; > > /* create disk cache */ > > if( (cbio = dcacheDevCreate(blkDevId, NULL, dcacheSize, "/sd0")) > > == NULL ) > > return ERROR ; > > /* create partition manager with FDISK style decoder, up to 3 parts */ > > cbio1 = dpartDevCreate( cbio, 3, usrFdiskPartRead ); > > /* create the 1st file system, 8 simult. open files, with CHKDSK */ > > dosFsDevCreate( devNames[0], dpartPartGet(cbio1,0), 8, 0 ); > > /* create the 2nd file system, 6 simult. open files, with CHKDSK */ > > dosFsDevCreate( devNames[1], dpartPartGet(cbio1,1), 6, 0 ); > > /* create the 3rd file system, 4 simultaneously open files, no CHKDSK */ > > dosFsDevCreate( devNames[2], dpartPartGet(cbio1,2), 4, NONE ); > > } > > > > Do note that old ATA intialization code had some partition handling code of > > its own (because DosFs 1.0 did not handle partitions) and you need to make > > sure that the ATA driver intialization code does not do anything like that, > > and that the block device "sees" the entire CF card, starting from block 0. > > > > As to formatting, you are right, useFdiskPartCreate() creates the partition > > table, > > then you need to call dosFsVolFormat() to actually initialize the dosFs. > > I fyou do not plan on booting from the CF (even PCs dont use CF or PCMCIA > > for booting normally), the code in usrFdiskLib should satisfy your needs. > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: where can I have free downloadable Tornado II Date: Tue, 03 Sep 2002 12:09:16 -0400 From: Joe Durusau Organization: Lockheed Martin Corporation Message-ID: <3D74DEAC.8ED672FF@lmco.com> References: <8076f4da.0209030507.73231afc@posting.google.com> bhanu prakash wrote: > > Hi, > I would like to know where can I have freely downloadable Tornado > II simpc. > Recently I had an interactive demo class,I came to know later that its > freely downloadable and can be used for some 2 months.Through this we > can see the graphical pictures of semophores give and take,scheduling > and we can implement all the VxWorks API in it.Its actually a > simulator of BSP.Can any one give pointers where can I get this?I went > through www.windriver.com but couldn't able to get it. > > Thanks and rgds, > Bhanu Prakash You can't. It is possible to get a limited use demo license from a local rep (that expires in so many days), but the days of the free version of vxSim are over. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Gurus out there please help -- compact flash Date: 3 Sep 2002 09:28:19 -0700 From: ritalin@comcast.net (Rita) Organization: http://groups.google.com/ Message-ID: References: Hi Leonid, Thanks! It works! I still got some questions though. If I don't call usrFdiskPartCreate(), usrFdiskPartLib prints out warning "disk partition table is not intiialized". When calling format, dosFsVolFormat() writes to sector 0, and dosFs gets mounted correctly, and I'm able to read and write files. If I call usrFdiskPartCreate() before dpartDevCreate(), the partition table is written to sector 0, and dosFsVolFormat() writes to other sectors, that it later complains cannot be found. Can you explain what I'm doing wrong? Thanks. Rita "Leonid Rosenboim" wrote in message news:... > Rita, > > Here is an example intialization code from teh DosFs2.0 manual: > > It assumes the disk is execpected to have at most three partitions, > and does not include the block device driver intialization. > > > usrPartDiskFsInit( void * blkDevId ) > { > const char * devNames[] = { "/sd0a", "/sd0b", "/sd0c" }; > int dcacheSize = 0x30000 ; > void * cbio, * cbio1 ; > /* create disk cache */ > if( (cbio = dcacheDevCreate(blkDevId, NULL, dcacheSize, "/sd0")) > == NULL ) > return ERROR ; > /* create partition manager with FDISK style decoder, up to 3 parts */ > cbio1 = dpartDevCreate( cbio, 3, usrFdiskPartRead ); > /* create the 1st file system, 8 simult. open files, with CHKDSK */ > dosFsDevCreate( devNames[0], dpartPartGet(cbio1,0), 8, 0 ); > /* create the 2nd file system, 6 simult. open files, with CHKDSK */ > dosFsDevCreate( devNames[1], dpartPartGet(cbio1,1), 6, 0 ); > /* create the 3rd file system, 4 simultaneously open files, no CHKDSK */ > dosFsDevCreate( devNames[2], dpartPartGet(cbio1,2), 4, NONE ); > } > > Do note that old ATA intialization code had some partition handling code of > its own (because DosFs 1.0 did not handle partitions) and you need to make > sure that the ATA driver intialization code does not do anything like that, > and that the block device "sees" the entire CF card, starting from block 0. > > As to formatting, you are right, useFdiskPartCreate() creates the partition > table, > then you need to call dosFsVolFormat() to actually initialize the dosFs. > I fyou do not plan on booting from the CF (even PCs dont use CF or PCMCIA > for booting normally), the code in usrFdiskLib should satisfy your needs. > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Gurus out there please help -- compact flash Date: Tue, 3 Sep 2002 18:56:28 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Let me try to sort out some confusion: WHen you initialize the Dosfs "stack", there needs not be a disk in the slot, or there could be an unformatted disk or whatever. Intialization itself does not cause disk access, and therefore disk needs not be legal. Initialization is when you call ataDevCreate, then dcacheDevCreate, dpartDevCreate, and { dosFsDevCreate with dpartPartGet ). When you attempt to access a file (or do "ll"), dosFs will send a RESET ioctl all the way down the stack to the driver, at that point the driver needs to reinitialize (this is a removable so it needs to do some init if a new card was inserted), and then dcache reinits, and dpartLib will re-read the partition table, and dosFs will re-read its first block. Now, suppose you done that initialization, and a card is inserted with an illegal partition table, there will be a warning, and you can ignore it if you intend to later create the partition table and format the partition. Before you go into formatting, I strongly recommend you get comfortable with using CF cards that where formatted on a PC, and make sure everyting works to your satisfaction. When you want to reformat the CF, do it AFTER the normail initialization procedure, and after all is formatted, the stack will be RESET again (as if a card was cahnged) so everything will be reinitialized as if a card was changed. Hope this helps. - - Leonid "Rita" wrote in message news:d3a5605c.0209030828.70de7b29@posting.google.com... > Hi Leonid, > > Thanks! It works! I still got some questions though. If I don't call > usrFdiskPartCreate(), usrFdiskPartLib prints out warning "disk > partition table is not intiialized". When calling format, > dosFsVolFormat() writes to sector 0, and dosFs gets mounted correctly, > and I'm able to read and write files. > > If I call usrFdiskPartCreate() before dpartDevCreate(), the partition > table is written to sector 0, and dosFsVolFormat() writes to other > sectors, that it later complains cannot be found. > > Can you explain what I'm doing wrong? > Thanks. > Rita > > "Leonid Rosenboim" wrote in message news:... > > Rita, > > > > Here is an example intialization code from teh DosFs2.0 manual: > > > > It assumes the disk is execpected to have at most three partitions, > > and does not include the block device driver intialization. > > > > > > usrPartDiskFsInit( void * blkDevId ) > > { > > const char * devNames[] = { "/sd0a", "/sd0b", "/sd0c" }; > > int dcacheSize = 0x30000 ; > > void * cbio, * cbio1 ; > > /* create disk cache */ > > if( (cbio = dcacheDevCreate(blkDevId, NULL, dcacheSize, "/sd0")) > > == NULL ) > > return ERROR ; > > /* create partition manager with FDISK style decoder, up to 3 parts */ > > cbio1 = dpartDevCreate( cbio, 3, usrFdiskPartRead ); > > /* create the 1st file system, 8 simult. open files, with CHKDSK */ > > dosFsDevCreate( devNames[0], dpartPartGet(cbio1,0), 8, 0 ); > > /* create the 2nd file system, 6 simult. open files, with CHKDSK */ > > dosFsDevCreate( devNames[1], dpartPartGet(cbio1,1), 6, 0 ); > > /* create the 3rd file system, 4 simultaneously open files, no CHKDSK */ > > dosFsDevCreate( devNames[2], dpartPartGet(cbio1,2), 4, NONE ); > > } > > > > Do note that old ATA intialization code had some partition handling code of > > its own (because DosFs 1.0 did not handle partitions) and you need to make > > sure that the ATA driver intialization code does not do anything like that, > > and that the block device "sees" the entire CF card, starting from block 0. > > > > As to formatting, you are right, useFdiskPartCreate() creates the partition > > table, > > then you need to call dosFsVolFormat() to actually initialize the dosFs. > > I fyou do not plan on booting from the CF (even PCs dont use CF or PCMCIA > > for booting normally), the code in usrFdiskLib should satisfy your needs. > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Regarding Stack Size in Task Spawn Date: 3 Sep 2002 10:06:53 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0209030906.533604e4@posting.google.com> References: vxworksonline@yahoo.com (Shiny) wrote in message news:... > Is there any problem in case we don't give any task size while > spawning a Task(0). > For a simple program it didn't give any problem.Will there be a > problem in Real time. Hi, give the "checkStack" tool a try. It will show you the max stack size, the actual stack size and the high water mark (it does so by pattern filling the memory), this will aid you in making your stacksize decision, regards, toni --------------------------- Newsgroups: comp.os.vxworks Subject: Re: TCP/ip Date: 3 Sep 2002 10:08:51 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0209030908.1bba6036@posting.google.com> References: <46992b69.0209010112.143f3b3f@posting.google.com> "Leonid Rosenboim" wrote in message news:... > The "free" version of Tornado/VxSim that used to be available > off the WindRiver web site (not anymore) has a stripped down > version of VxSim, which does not have the TCP/IP stack. I agree, there even is a seperate CD for the "Full" Simulator. IMHO the T22 version will include the full simulator again... toni --------------------------- Newsgroups: comp.os.vxworks Subject: gets() skips first char Date: Tue, 3 Sep 2002 19:39:04 +0200 From: "Joachim Bassendowski" Organization: HARZ.DE Internet fuer die Region (XeNET GmbH, Clausthal) Message-ID: Hi, this is a part of the unittest-code for a SBS ASF-V6-1: [..] char response[ 81 ]; [..] /* Display the "Quit Unit Test" option. */ cursor_write( line_number, 25, "q - Quit Unit Test" ); line_number = (SBS16)( line_number + 3 ); while (1) { gets( response ); printf("\"%s\"\n",response); } [..] And the output is: VxWorks Copyright 1984-1998 Wind River Systems, Inc. CPU: Motorola MVME2432-1 - MPC 750 VxWorks: 5.4 BSP version: 1.2/0 Creation date: Jun 29 2001 WDB: Ready. [..] SBS Technologies, Inc. Integrated Avionics Library Unit Test Version 6.3 Build Sep 3 2002 a - Arinc 429 Unit Test m - MIL 1553 Unit Test q - Quit Unit Test - -> m 00000000: 0011-123 "23" abc "bc" After displaying "q - Quit Unit Test" I entered return. The result was the prompt ->. Other entries here lead to an error. Then I entered 'm' and the computer showed "00000000: 0011". Then I entered 123 but the first char wasn't stored to the char-array. The same with input abc and result bc. Please, can someone explain this behaviour to me? Thanks, J. Bassendowski --------------------------- Newsgroups: comp.os.vxworks Subject: gcc cross compiler linux-sh3 Date: 3 Sep 2002 19:14:01 GMT From: Carsten Nitsch Organization: Uni Leipzig Message-ID: hi all, did anybody of you compile the gnu gcc / binutils successfully as a cross compiler for SH3 targets, running under linux (or solaris)? We own the windows version of tornado. Due to many limitation of the ms-system, we tried to build gcc/binutils for the linux platform. Unfortunately the build failed already for binutils: ./configure --program-suffix=sh --target=sh-wrs-vxworks \ --with-headers=/home/samba/Tornado/target/h/ \ --with-libs=/home/samba/Tornado/target/lib make make runs for a while an ends with different error messages such as ./config/tc-sh.c:163: warning: (near initialization for `md_relax_table[0].rlx_backward') or ./config/tc-sh.c:2409: structure has no member named `fx_r_type' We use binutils-2.10.1 Any ideas or hints? thanx Carsten --------------------------- Newsgroups: comp.os.vxworks Subject: netBufLib Date: 3 Sep 2002 12:20:25 -0700 From: tulasi@sionintl.com (Tulasidas) Organization: http://groups.google.com/ Message-ID: <74371f59.0209031120.74311c68@posting.google.com> I am porting the VxWorks code to UNIX OS. The VxWorks code uses data structures such as mBlks and clBlks. Do we have an equivalent data structures in Unix OS. Thanks Tulasi --------------------------- Newsgroups: comp.os.vxworks Subject: Re: [Q] ppc405, no snoop, cacheDmaMalloc Date: 3 Sep 2002 12:54:35 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0209031154.251a49cb@posting.google.com> References: kotaeji@hanmail.net (Kim Tae Hyung) wrote in message news:... > Should I disable data cache for the memory? This will not work for the walnut BSP (at least not for the T2 version). This version of the BSP requires Data Cache... toni --------------------------- Newsgroups: comp.os.vxworks Subject: WindView problem Date: Tue, 03 Sep 2002 21:42:18 GMT From: "Michael B." Organization: [Infostrada] Message-ID: <_0ad9.145989$lu5.3602070@twister1.libero.it> Hi, I would like to show you my strange problem: I have installed Tornado 2.0 by now and at startup it says "couldn't open file TRY=localhost/host/resource/.../WindView.win32.tcl: no such file or directory". There is a lonely button OK and Tornado IDE has no WindView buttons and functionalities. What does this mean? Why "localhost/..." instead of "Tornado/...". I can find that tcl file, it esists! I did not make any strange choice with installation, just provide license codes and specify what I wanted to install! I guess to try a re-install, what about? Thanks in advance, M. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Gurus out there please help -- compact flash Date: 3 Sep 2002 15:09:30 -0700 From: ritalin@comcast.net (Rita) Organization: http://groups.google.com/ Message-ID: References: Thanks, Leonid. That explains a lot. I checked the byte swap, and found where I made the mistake. When formatted under PC, I have to read it out in *reverse* order to get it out right. When writing using vxWorks, I assume if I don't do byte-swap while writing it, I can read it without byte-swapping later. So after couple of testing and rearranging byte order, I ended up having straight write, and byte-swap read. Thanks for pointing that out. Thanks for the suggestion. I'll get a CF and play with it a bit with PC before I dip into CF formatting. Again, many thanks to your help. I can finally see some daylight. Rita "Leonid Rosenboim" wrote in message news:... > Let me try to sort out some confusion: > > WHen you initialize the Dosfs "stack", there needs not be a disk in the > slot, or there could be an unformatted disk or whatever. Intialization > itself does not cause disk access, and therefore disk needs not be legal. > > Initialization is when you call ataDevCreate, then dcacheDevCreate, > dpartDevCreate, and { dosFsDevCreate with dpartPartGet ). > > When you attempt to access a file (or do "ll"), dosFs will send a RESET > ioctl all the way down the stack to the driver, at that point the driver > needs to reinitialize (this is a removable so it needs to do some init if a > new card was inserted), and then dcache reinits, and dpartLib will re-read > the partition table, and dosFs will re-read its first block. > > Now, suppose you done that initialization, and a card is inserted with an > illegal partition table, there will be a warning, and you can ignore it if > you intend to later create the partition table and format the partition. > > Before you go into formatting, I strongly recommend you get comfortable with > using CF cards that where formatted on a PC, and make sure everyting works > to your satisfaction. > > When you want to reformat the CF, do it AFTER the normail initialization > procedure, and after all is formatted, the stack will be RESET again (as if > a card was cahnged) so everything will be reinitialized as if a card was > changed. > > Hope this helps. > - Leonid --------------------------- Newsgroups: comp.os.vxworks Subject: launching the target server Date: Tue, 3 Sep 2002 16:30:38 -0700 From: "kdb" Message-ID: greetage - I've been unable to get the target server to connect correctly to my target. It does appear to connect but I get a checksum mismatch between the core image file on the host and the target core image in target memory. The sizes of the areas being checksummed are the same in both target memory and the core file on the host, but in different address ranges. In Tornado it looks like the target server is connected and I can start the debugger but when I set breakpoints in the code they do not halt program execution. I tried attaching the debugger to a task and it crashed the target. I've tried connecting via both the network and the serial but with similar results. I tried working this with our WindRiver field engineer but we couldn't make it work. When I described our system and the project build output he was surprised to hear that it linked with symTbl.c to get symbol data rather than ctdt.c, which is not even a part of the project. He had me do a make in the BSP directory and load up the vxWorks file this produced, which is the kernel only, no user application code. The target server should connect right up with this (that's part of their acid test for releasing a BSP) but in this case it timed out both with the serial and network options. Any idea why this is not working? thx - k --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Target directory listing Date: Wed, 04 Sep 2002 00:10:17 GMT From: "weddick" Organization: AT&T Broadband Message-ID: References: I didn't build the system with the target shell. The other systems don't have the target shell either. Joel "Johan Borkhuis" wrote in message news:Xns927E576574C5borkhuisagerecom@192.19.193.66... > "weddick" wrote: > > > I have a new system that I am unable to perform a @ls "/ide0/" command > > that I am able to do on other systems. What did I leave out in the > > new bsp? > > Can you do a ls "/ide0/" command on the target shell? > > Kind regards, > Johan > > -- > o o o o o o o . . . _____________________________ > o _____ || Johan Borkhuis | > .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | > >(________|__|_[_________]_|__________________________| > _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` > === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === > --------------------------- Newsgroups: comp.os.vxworks Subject: vmBaseGlobalMapInit takes toooo long! Date: 3 Sep 2002 17:58:29 -0700 From: jchen@terayon.com (john) Organization: http://groups.google.com/ Message-ID: Dear vxworks developers. I've read most of the posts before about how vmBaseGlobalMapInit takes a long time to map/remap memory. I have pointers and global vars that at some point in my program I would like them to have different physical addresses in memory. so in my code I redefined sysPhysMemDesc and called vmBaseGlobalMapInit to initialize it. everything works fine except it takes way too long. Does anyone have vxVMI option for their PPC 8260 BSP? can you use vmMap to just remap/add a region of virtual-physical translation (instead of the whole thing)? If you have it, would anyone mind giving me an idea if its quick (in order of ms for ~20MB)? we are using a ppc 8260 We'll buy the VxVMI option if this will work. just curious..... thanks in advance! john --------------------------- Newsgroups: comp.os.vxworks Subject: Re: netBufLib Date: 4 Sep 2002 01:52:24 GMT From: Denis Perelyubskiy Message-ID: References: <74371f59.0209031120.74311c68@posting.google.com> * Tulasidas [09-Tue-02 12:20 -0700]: > > I am porting the VxWorks code to UNIX OS. The VxWorks > code uses data structures such as mBlks and > clBlks. Do we have an equivalent data structures in Unix > OS. I haven't dealt with that, but it sounds like mbufs would be the structures you're looking for. in fact, there is a net/mbuf.h header that maps mbufs to mblocks / clusters. i could be way off on this one though. check it out. i'd be interested to hear the final word on this. denis - -- if you want to reply by mail, please tweak my address in a reasonable way. sorry for the inconvenience. --------------------------- Newsgroups: comp.os.vxworks Subject: ifAddrAdd problem Date: 3 Sep 2002 18:56:12 -0700 From: kevinatnj@yahoo.com.cn (kewin) Organization: http://groups.google.com/ Message-ID: <221caa76.0209031756.114c3674@posting.google.com> hi all There is a problem confusing me for a long time about ifAddrAdd.I have a board ,called MP,with two ethernet ports. I have set the ip,"171.1.1.1", to port 1 and "171.2.1.1" to port 2;and another board,called PP, has one ethernet port with ip address,"171.0.1.1".Two boards are in one local lan.First, i added "171.0.1.2" to MP port 1,and i can ping PP by MP port 1;Second,i deleted "171.0.1.2",from MP port1 and added it to MP port 2.I want to ping PP by this port,but it didn't work.If I was patient enough and waited for 20 minutes,I can ping PP again by MP port2.I known this is probaly the problem of arptable.After I added or deleted IP address, I have called arpFlush.It seems no use.Anybody can help we with this porblem? thanks kevin --------------------------- Newsgroups: comp.os.vxworks Subject: Re: [Q] ppc405, no snoop, cacheDmaMalloc Date: 3 Sep 2002 19:05:27 -0700 From: kotaeji@hanmail.net (Kim Tae Hyung) Organization: http://groups.google.com/ Message-ID: References: "Michael Lawnick" wrote in message news:... > Hi Kim, > > cacheDmaMalloc is used to malloc memory and mark it as uncacheable. If you > can't use the function, you'd have to call MMU-functions - which are not > present if you haven't bought full MMU support :-( > > So I see only one way: You will have to call cacheFlush after every write > function (before telling NIC-H/W that there is a new packet to send) and > cacheInvalidate before every read function. This method works, I've done it > for a DP83815. Take care: Buffers may not share cache lines so adjust your > buffer sizes and addresses properly! > > Switching off data cache for complete target is a knock out, your target > will slow down to snail. > > BTW: The same discussion occured some months ago (last year?). Do a google > search. > > HTH > Michael > > "Kim Tae Hyung" schrieb im Newsbeitrag > news:b1b90dae.0209021817.5a8f5b46@posting.google.com... > > Hi. > > I have a ppc405 board. And two dec21x4x pci enthernet cards are connected > > through ppc405 pci interface. > > > > Device Driver of dec21x4x uses cacheDmaMalloc to allocate memory for > > rx,tx descriptors and network clusters. But I should replace > cacheDmaMalloc with > > malloc(or memalign) for some reasons. > > > > By the way, ppc405 has no snooping ability. > > I want to know about the mechanism of cacheDmaMalloc. > > Should I disable data cache for the memory? > > > > Thanks in advance. Thanks a lot. I solved this problem by allocating memory multiple of page size and align it to page size. And I disabled data cache by calling vmBaseStateSet like the following code. But I still don't know why the performance becomes extremely low. :( And I tried the method which cacheFlush before write and cacheInvalidate before read. But it fails. :( Any comments appreciated. /* * Make size multiple of page size. */ sz = ((sz / VM_PAGE_SIZE) +1) * VM_PAGE_SIZE; pDrvCtrl->memBase = pShMem = (char *) valloc (sz); /* disable data cache */ if (vmBaseStateSet(NULL, pShMem, sz , VM_STATE_MASK_VALID | VM_STATE_MASK_WRITABLE | VM_STATE_MASK_CACHEABLE , VM_STATE_VALID | VM_STATE_WRITABLE | VM_STATE_CACHEABLE_NOT) != OK) printf (">> cannot vmBaseStateSet 1\n"); --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VMIC 6001 with Vxworks Date: Tue, 3 Sep 2002 23:10:52 -0400 From: "Cedric Le Roux" Message-ID: <3d757959$1_2@nopics.sjc> References: <3d6ebfe6_3@nopics.sjc> <6db76185.0209030743.6396707f@posting.google.com> I keep getting the Error bit in the Active Status Register By changing the Error Mask, I was able to determine that the error being generated is "transmission error" I've got diag mode working for the internal memory test, but neither the positive nor negative sync tests echo back anything. (actually, the board is changing one of the values I put into one of the 'data' locations?!?!) The board's BIT returns successfully, however. A search for clarification on this in the manual is useless, trust me. (I'm not going to rant here about how bad that manual is) I was just curious to see if someone else had successfully implemented some code for the diag features of this board. I want to know if I am missing something. As you described, all I do is put some test value in the 'data' locations and then start the card in Diag mode. Then the card puts the test values into the 'echo' locations for me to read. I've got two of these boards and they both do the same thing...so I doubt I've got bad hardware. Thanks for the reply, Cedric "Terry Butman" wrote in message news:6db76185.0209030743.6396707f@posting.google.com... > "Cedric Le Roux" wrote in message news:<3d6ebfe6_3@nopics.sjc>... > > I'm looking for some clarification regarding the programming for diagnostic > > mode for the VMIC 6001 MIL-STD-1553 board. The manual is kinda vague. > > > > If anyone out there has worked with this board, I'd like to hear from ya. > > or if anyone knows a better place for this post, I'd like to hear about that > > too. > > > > Thanks. > > What exactly do you want to do? > > The diagnostics mode was designed to test > the ability to read/write data through the 1553 interface. > > As best as I recall you do this with a series of diagnostic blocks which > include: > > the number of memory locations to test > the number of 1553 data words to test > whether or not the data goes through the full data path or just through the > transmitter/receiver etc. > > As I recall there is a reserved memory space in the diagnostic control blocks > for what is called the "echo". > > It goes something like this. > > memory data written > memory data read (echo) > > 1553 data written > 1553 data read (echo) > > I seem to recall that the 1553 data might include the transmission status > information and the interrupt status information. > > Any further details and I'll have to dig out the manual. > > regards, Terry --------------------------- Newsgroups: comp.os.vxworks Subject: Re: WindView problem Date: Wed, 4 Sep 2002 12:53:58 +0800 From: "Jim.chu" Organization: Bentium Ltd. (CN99) Message-ID: References: <_0ad9.145989$lu5.3602070@twister1.libero.it> Do you install tornado 2.0 on Win2000 platform ? If is, you need install spr62517(Tornado 2.0 w2k Patch). You can find it on WindSurf of WindRiver web. "Michael B." wrote in message news:_0ad9.145989$lu5.3602070@twister1.libero.it... > Hi, I would like to show you my strange problem: I have installed Tornado > 2.0 by now and at startup it says "couldn't open file > TRY=localhost/host/resource/.../WindView.win32.tcl: no such file or > directory". There is a lonely button OK and Tornado IDE has no WindView > buttons and functionalities. What does this mean? Why "localhost/..." > instead of "Tornado/...". I can find that tcl file, it esists! I did not > make any strange choice with installation, just provide license codes and > specify what I wanted to install! I guess to try a re-install, what about? > > Thanks in advance, M. > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Target directory listing Date: Wed, 4 Sep 2002 06:24:50 +0000 (UTC) From: Johan Borkhuis Organization: Agere Systems Message-ID: References: "weddick" wrote: > I didn't build the system with the target shell. The other systems > don't have the target shell either. Is it possible for you to build a system with the target shell? Also, can you check what devices are available (the devs, or @devs, command)? Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: vmBaseGlobalMapInit takes toooo long! Date: 3 Sep 2002 23:47:37 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0209032247.31f32382@posting.google.com> References: Hello, Since the 8260 is really just a 603 inside, there is no VxVMI product for it (the 60x cores have a strange MMU that doesn't work well with the VxVMI model). You probably have three options: 1) VxWorks AE which contains all the VM code to do this kind of thing, though changing the mappings for 20MB of memory is no small task so this might not be any quicker than your current solution. 2) Set up the effective address of your data in its own segment (256MB) of effective address space, and set up all the mappings for both sets in the hash table using different virtual addresses (read chapter 7 of the PPC programming environments, or green book if you don't know about PPC virtual and effective addresses). If you did this, swapping between sets would be simply changing one register value (much, much quicker!). 3) Use a BAT register to create the mapping (again using a separate block of the address space), and swap the BAT register value to swap between the two sets. If you use (2) or (3) make sure that you remove the physical memory used for both sets of data from the system memory pool (e.g. reduce LOCAL_MEM_SIZE by an appropriate amount). HTH, John... jchen@terayon.com (john) wrote in message news:... > Dear vxworks developers. > > I've read most of the posts before about how vmBaseGlobalMapInit takes > a long time to map/remap memory. I have pointers and global vars that > at some point in my program I would like them to have different > physical addresses in memory. so in my code I redefined > sysPhysMemDesc and called vmBaseGlobalMapInit to initialize it. > everything works fine except it takes way too long. > > Does anyone have vxVMI option for their PPC 8260 BSP? can you use > vmMap to just remap/add a region of virtual-physical translation > (instead of the whole thing)? If you have it, would anyone mind giving > me an idea if its quick (in order of ms for ~20MB)? we are using a > ppc 8260 > > We'll buy the VxVMI option if this will work. just curious..... > > thanks in advance! > john --------------------------- Newsgroups: comp.os.vxworks Subject: Re: gcc cross compiler linux-sh3 Date: 3 Sep 2002 23:50:33 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0209032250.5d623623@posting.google.com> References: Hello, If you haven't already done so, get the GNU sources from Wind River (speak to your local sales office or FAE). Don't know if that is the problem, but it is probably better to start with the same sources in case there were changes made to get it to work. HTH, John... Carsten Nitsch wrote in message news:... > hi all, > > > did anybody of you compile the gnu gcc / binutils successfully as a > cross compiler for SH3 targets, running under linux (or solaris)? > We own the windows version of tornado. Due to many limitation > of the ms-system, we tried to build gcc/binutils for the linux > platform. Unfortunately the build failed already for binutils: > > ./configure --program-suffix=sh --target=sh-wrs-vxworks \ > --with-headers=/home/samba/Tornado/target/h/ \ > --with-libs=/home/samba/Tornado/target/lib > > make > > make runs for a while an ends with different error messages such as > ./config/tc-sh.c:163: warning: > (near initialization for `md_relax_table[0].rlx_backward') > > or ./config/tc-sh.c:2409: structure has no member named `fx_r_type' > > We use binutils-2.10.1 > > Any ideas or hints? > > thanx > Carsten --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Target directory listing Date: 3 Sep 2002 23:55:28 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0209032255.111bf78d@posting.google.com> References: Hello, ls on the target (which is what is needed for @ls to work) is part of the usrLib.c file IIRC (under target/src/usr in your tree). You need to make a reference somehow to something in this file if you want the build to include it in the system (normally it is added when the target shell is added, but it is probably possible to include it on its own). If you want a really simply way to reference it, just add something like this in sysLib.c: static FUNCPTR dummy; then inside an existing function somewhere: dummy = (FUNCPTR) ls; You may need to add an include of usrLib.h to get rid of the warning about ls being unknown. HTH, John... "weddick" wrote in message news:... > I have a new system that I am unable to perform a @ls "/ide0/" command that > I am able to do on other systems. What did I leave out in the new bsp? > > Thanks, > Joel --------------------------- Newsgroups: comp.os.vxworks Subject: Re: launching the target server Date: 4 Sep 2002 00:10:46 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0209032310.24e85175@posting.google.com> References: Hello, Before you will get any useful answers I suspect that you will need to provide the following: 1) Tornado & VxWorks versions 2) Architecture, CPU type and BSP 3) The type of connection you are using 4) The type and version of the toolchain 5) Any special processing, or changes in the build process that you may have made I don't understand the references to symTbl.c and ctdt.c since they have nothing to do with the target server's connection to the target. FYI, ctdt.c is where the C++ static object constructors & destructors for the bootable image are listed (nothing to do with symbols). symTbl.c is, as you noted, the place where the target-side standalone symbol table's initial symbols come from. The target server obtains its symbol information from the bootable image file directly (i.e. it reads the object file's internal symbol table). Things that could cause the checksum to fail are: a) The images not being matched for some reason (e.g. booting from an older image by mistake) b) Something in the code changing read-only data (or code) dynamically (e.g. writing through a pointer to a literal string - string literals for some architectures are stored in the text section). You might be able to change this using -fwritable-strings as an option to the compiler if the string in question is in the application code, or BSP sources. c) Changes in the build process causing the object file to be without something needed by the target server (all I can think of here is stripping symbols, but I don't know if that does cause a problem, or if there is anything else that might). HTH, John... "kdb" wrote in message news:... > greetage - > > I've been unable to get the target server to connect correctly to my target. > It does appear to connect but I get a checksum mismatch between the core > image file on the host and the target core image in target memory. The > sizes of the areas being checksummed are the same in both target memory and > the core file on the host, but in different address ranges. In Tornado it > looks like the target server is connected and I can start the debugger but > when I set breakpoints in the code they do not halt program execution. I > tried attaching the debugger to a task and it crashed the target. I've > tried connecting via both the network and the serial but with similar > results. > > I tried working this with our WindRiver field engineer but we couldn't make > it work. When I described our system and the project build output he was > surprised to hear that it linked with symTbl.c to get symbol data rather > than ctdt.c, which is not even a part of the project. He had me do a make > in the BSP directory and load up the vxWorks file this produced, which is > the kernel only, no user application code. The target server should connect > right up with this (that's part of their acid test for releasing a BSP) but > in this case it timed out both with the serial and network options. > > Any idea why this is not working? > > thx - k --------------------------- Newsgroups: comp.os.vxworks Subject: Re: [Q] ppc405, no snoop, cacheDmaMalloc Date: Wed, 04 Sep 2002 07:18:22 GMT From: "Gary M" Organization: AT&T Broadband Message-ID: <2tid9.291116$983.536199@rwcrnsc53> References: "Kim Tae Hyung" wrote in message news:b1b90dae.0209031805.38207e2f@posting.google.com... > > Thanks a lot. > > I solved this problem by allocating memory multiple of page size > and align it to page size. And I disabled data cache by calling > vmBaseStateSet like the following code. > > But I still don't know why the performance becomes extremely low. :( Performance will certainly be lower. Are you sure memory speed is properly optimized in the MMU? are the NICs performing bursting DMA to and from the buffers? > And I tried the method which cacheFlush before write and > cacheInvalidate before read. But it fails. :( > You should do cacheInvalidate AFTER the read. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: [Q] ppc405, no snoop, cacheDmaMalloc Date: Wed, 4 Sep 2002 09:12:35 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: Reply-To: "Michael Lawnick" Sender: mlawnick@p5088ff37.dip.t-dialin.net Hi Kim, >I solved this problem by allocating memory multiple of page size >and align it to page size. And I disabled data cache by calling >vmBaseStateSet like the following code. That should be along the way how cacheDmaMalloc works. If you get performance problems, you seem to do something basically wrong. Haven't done it on my own, so can't tell you any caveAts. >And I tried the method which cacheFlush before write and >cacheInvalidate before read. But it fails. :( For us it works perfectly: We have switched off the MMU and are working with the 32bit descriptor (128MB pages). So if you say it fails: what are the problems ? Did you consider, that the allocated clusters will be prepended by 4 bytes of chain-pointer ? These 4 bytes must be on same cache line too! Regards, Michael "Kim Tae Hyung" schrieb im Newsbeitrag news:b1b90dae.0209031805.38207e2f@posting.google.com... > "Michael Lawnick" wrote in message news:... > > Hi Kim, > > > > cacheDmaMalloc is used to malloc memory and mark it as uncacheable. If you > > can't use the function, you'd have to call MMU-functions - which are not > > present if you haven't bought full MMU support :-( > > > > So I see only one way: You will have to call cacheFlush after every write > > function (before telling NIC-H/W that there is a new packet to send) and > > cacheInvalidate before every read function. This method works, I've done it > > for a DP83815. Take care: Buffers may not share cache lines so adjust your > > buffer sizes and addresses properly! > > > > Switching off data cache for complete target is a knock out, your target > > will slow down to snail. > > > > BTW: The same discussion occured some months ago (last year?). Do a google > > search. > > > > HTH > > Michael > > > > "Kim Tae Hyung" schrieb im Newsbeitrag > > news:b1b90dae.0209021817.5a8f5b46@posting.google.com... > > > Hi. > > > I have a ppc405 board. And two dec21x4x pci enthernet cards are connected > > > through ppc405 pci interface. > > > > > > Device Driver of dec21x4x uses cacheDmaMalloc to allocate memory for > > > rx,tx descriptors and network clusters. But I should replace > > cacheDmaMalloc with > > > malloc(or memalign) for some reasons. > > > > > > By the way, ppc405 has no snooping ability. > > > I want to know about the mechanism of cacheDmaMalloc. > > > Should I disable data cache for the memory? > > > > > > Thanks in advance. > > Thanks a lot. > > I solved this problem by allocating memory multiple of page size > and align it to page size. And I disabled data cache by calling > vmBaseStateSet like the following code. > > But I still don't know why the performance becomes extremely low. :( > And I tried the method which cacheFlush before write and > cacheInvalidate before read. But it fails. :( > > Any comments appreciated. > > /* > * Make size multiple of page size. > */ > sz = ((sz / VM_PAGE_SIZE) +1) * VM_PAGE_SIZE; > > pDrvCtrl->memBase = pShMem = (char *) valloc (sz); > > /* disable data cache */ > if (vmBaseStateSet(NULL, pShMem, sz , > VM_STATE_MASK_VALID | VM_STATE_MASK_WRITABLE | VM_STATE_MASK_CACHEABLE , > VM_STATE_VALID | VM_STATE_WRITABLE | VM_STATE_CACHEABLE_NOT) != OK) > printf (">> cannot vmBaseStateSet 1\n"); --------------------------- Newsgroups: comp.os.vxworks Subject: Re: anyone out there to help me plz?very urgent help requested Date: Wed, 04 Sep 2002 07:34:59 GMT From: "Gary M" Organization: AT&T Broadband Message-ID: References: There comes a time when you need to start investigating your "urgent" issues on your own, or to invest in hiring someone who can. Over the past few months you have demonstrated that you have issues with basic concepts of system architecture and development (not to mention basic investigative engineering). As others have pointed out, there are plenty of freely available resources available for you to solve the problem on your own. It is doubtful that anyone here is going to teach you how to write a device driver. "subbarayan" wrote in message news:f81c4277.0208302049.2e4770ad@posting.google.com... > dear mentors, > Can anyone jus tell me how to go abt proceeding the development of > device driver for pci card for x86 processor with standard pc > architecture?I use tornadoII and windows development host and a stand > alone target. > Further can anyone tell me when will be this pciautoconfiglib file > called?will it be called automatically as a startup routine or need it > to be called explicitly to enable pci scanning? > What exactly is the difference between the files pci autoconfiglib.c > and pciconfiglib.c?which file should be used for which sort of > appln?and can these library files directly compiled and used or should > we call it from some appln to invoke and access it? > And sorry for my ignorance and unknowing things.One thing i found in > the manual for pci support on vxworks says we need to include the > macro INCLUDE_PCI for enabling pci support..i found only the macro > INCLUDE_PCI_PARAMS and not INCLUDE_PCI.Are both same or different?if > different where can i find INCLUDE_PCI and use it in my application? > Where exactly should this > macro be included and where exactly is this macro defined?How to > include this macro and whats the exact location where this macro has > to be included?And whats the role of bios with respect to vxworks in > detecting pci card and pci bus?It would be helpful if some one could > tell me what would be printed on screen if pci support is enabled and > makes me confirm that my code has started working with pci support. > > Sorry if my questions look amature,but i am helpless as i am new to > this sort of coding. > Advance thanks to persons replying me. --------------------------- Newsgroups: comp.os.vxworks Subject: shell input output Date: Wed, 04 Sep 2002 09:53:23 +0200 From: Pescoller Dietrich Message-ID: the function shellOrigStdSet(STD_IN,myfd) seems to have no effect... wyis this possible what is the problem? My task is to filter the keyboard input so i do something like this... shellOrigStdSet(STD_IN, fd_shellinput) in an task then i do read(STD_IN,buf,size) if (some filter on buf) write (fd_shellinput,buf,nrread)/*iwrite to the shell input*/ else write (fd_elsewhere, buf, nread) but shellOrigStdSet doesn t make their work can someone help me understanding why or whath the solution could be? thanks diti --------------------------- Newsgroups: comp.os.vxworks Subject: Problem with HTTP Server and SLIP Date: 4 Sep 2002 01:05:29 -0700 From: Oleg.Chtchoupak@mtech-ag.com (Oleg) Organization: http://groups.google.com/ Message-ID: <16d0deb4.0209040005.2cfca43d@posting.google.com> Hallo all, we use the distributed system with two CPU under VxWorks. The CPUs communicate using SLIP. One CPU is DIMM PC Pentium under vxWorks 5.4 and here runs also GoAhead Web server. The second one is M68020 under VxWorks 5.31. When the Webserver is no in use (PEND+T) the SLIP communication between CPUs is OK. After we start a browser and get some images from the DIMM PC Target, the SLIP communication is broken. Do anyone now the explanation for this phenomenon? Can anyone help to resolve this problem? Best regards. Oleg. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ifAddrAdd problem Date: 4 Sep 2002 01:26:39 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0209040026.24abf85@posting.google.com> References: <221caa76.0209031756.114c3674@posting.google.com> kevinatnj@yahoo.com.cn (kewin) wrote in message news:<221caa76.0209031756.114c3674@posting.google.com>... > If I was patient enough and waited for 20 minutes,I can > ping PP again by MP port2.I known this is probaly the problem of > arptable.After I added or deleted IP address, I have called > arpFlush.It seems no use. well... your description and the 20 minutes definitly sound like an ARP Table problem. 20 minutes is the time dynamic ARP tables typically expire. IMHO, the system you used to ping (send ICMP ECHO REQUEST) to your vxWorks boards had the wrong ARP entry in it's cache. arpFlush on the system you used to ping the two vxWorks boards should resolve this. In case this doesn't work, maybe try taking an ethernet analyzer (ethereal?) and see whats going on on thw wire? toni --------------------------- Newsgroups: comp.os.vxworks Subject: Re: netBufLib Date: Wed, 4 Sep 2002 10:24:47 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <74371f59.0209031120.74311c68@posting.google.com> This would greatly depend on the type of code you are porting. If you are porting an appkication module, it will not be able to access the Unix mbufs structures, which are part of the kernel space. If you are porting a device driver, it would be quite surprising for you to ask this kind of questions, because you should have been already familiar with the device driver APIs on Unix by now. "Tulasidas" wrote in message news:74371f59.0209031120.74311c68@posting.google.com... > I am porting the VxWorks code to UNIX OS. The VxWorks code uses > data structures such as mBlks and clBlks. Do we have an > equivalent data structures in Unix OS. > > > Thanks > > Tulasi --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Problem with HTTP Server and SLIP Date: Wed, 4 Sep 2002 10:30:31 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <16d0deb4.0209040005.2cfca43d@posting.google.com> Oleg, can you be more specific about exactly how the SLIP links appears to be broken ? Does the SLIP link brake if you simply transfer large files over the SLIP link or not ? "Oleg" wrote in message news:16d0deb4.0209040005.2cfca43d@posting.google.com... > Hallo all, > > we use the distributed system with two CPU under VxWorks. The CPUs > communicate using SLIP. > One CPU is DIMM PC Pentium under vxWorks 5.4 and here runs also > GoAhead Web server. > > The second one is M68020 under VxWorks 5.31. > When the Webserver is no in use (PEND+T) the SLIP communication > between CPUs is OK. After we start a browser and get some images from > the DIMM PC Target, the SLIP communication is broken. > Do anyone now the explanation for this phenomenon? > Can anyone help to resolve this problem? > > Best regards. > Oleg. --------------------------- Newsgroups: comp.os.vxworks Subject: timer use Date: 4 Sep 2002 01:42:44 -0700 From: taissimon@yahoo.com.tw (simon tsai) Organization: http://groups.google.com/ Message-ID: <25a2f1f.0209040042.1d50eec9@posting.google.com> may i use a timer routine in END driver? i need it to check link status. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: timer use Date: Wed, 4 Sep 2002 10:50:30 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <25a2f1f.0209040042.1d50eec9@posting.google.com> I would recommend against POSIX timer (if that what you meant) in a driver, instead you should use the VxWorks software watchdog (see wdLib manual), and keep in mind the watchdog function will execute in system clock ISR context - keep it short, or just do netJobAdd() for the actual work. Link status check that invonves MII exchanges can take a long time on some MACs. "simon tsai" wrote in message news:25a2f1f.0209040042.1d50eec9@posting.google.com... > may i use a timer routine in END driver? > i need it to check link status. --------------------------- Newsgroups: comp.os.vxworks Subject: Subroutines and libraries Date: 4 Sep 2002 03:59:55 -0700 From: compaqbala@sify.com (Balakumar) Organization: http://groups.google.com/ Message-ID: <46992b69.0209040259.96fc2d8@posting.google.com> Hi vxworks users, this is simple question.can anybody tell me what is the difference between subroutines and libraries in vxworks. --------------------------- Newsgroups: comp.os.vxworks Subject: CORBA/static function problem Date: Wed, 4 Sep 2002 08:02:07 -0400 From: "John Sestak" Organization: Raytheon Company Message-ID: <9Dmd9.1541$c6.2317@bos-service2.ext.raytheon.com> Hi all, I'm currently running the ORBexpress CORBA ORB on two different processors in our system. One of them is an INOVA Pentium3 board and the other is an INOVA K6 board. Both of the cards have the exact same CORBA .out files loaded on them (libOEorb.out and libOEtcp.out) but when I try to run a static function called 'oneway_sync_scope' it works fine on the K6 baord but, it fails on the P3 board and I get an error saying: unresolved symbol: oneway_sync_scope( ) When I try running a 'lkup' on the oneway_sync_scope command, both processors say the following: CORBA::ORB::oneway_sync_scope(short) 0x07e510e0 text (libOEorb.out) CORBA::ORB::oneway_sync_scope(void) 0x07e51100 text (libOEorb.out) CORBA::ORB::oneway_sync_scope 0x07e789ec data (libOEorb.out) So, it looks like they both have the function loaded. The only thing I can think of that would be the problem is that there is a difference in the vxWorks image between the one running on the K6 and the one running on the P3 that is causing this. Other than this problem with this one CORBA function, the rest of the CORBA functionality seems to work fine on both processors. Anyone have any idea what the problem might be (i.e. what the difference is between the K6 and P3 kernels would be that would cause this)? Thank you, John Sestak --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Subroutines and libraries Date: Wed, 04 Sep 2002 08:01:58 -0400 From: Joe Durusau Organization: Lockheed Martin Corporation Message-ID: <3D75F636.8EBD4B44@lmco.com> References: <46992b69.0209040259.96fc2d8@posting.google.com> Balakumar wrote: > > Hi vxworks users, > this is simple question.can anybody tell me what is the difference > between subroutines and libraries in vxworks. A library is a collection of files that can be downloaded in a batch, or in many cases, selected files can be pulled out and linked into an executable. For example, libMC68040GNUvx.a is one such. WRS uses the term to describe a collection of routines that are related to a functionality group, for example Syslib. No such file exists in the system, but Syslib.h is used to advise the compiler of linkage rules to the routines in the group. A subroutine (usually called a function in C-ish) is just a thing you can call to perform some function for you. Example printf(). One reason for WRS non-standard use of the term 'library' is that it is possible to build a copy of vxWorks with only selected components included, to save space in the target. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Re: CORBA/static function problem Date: Wed, 4 Sep 2002 14:39:01 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <9Dmd9.1541$c6.2317@bos-service2.ext.raytheon.com> I think that the problem might be because the function you are calling is a C++ function, and there are two of them, depending on the argument. Its either that one of your target doe not have C++ support, or because one of them has C++ name mangling turned off. "John Sestak" wrote in message news:9Dmd9.1541$c6.2317@bos-service2.ext.raytheon.com... > Hi all, > I'm currently running the ORBexpress CORBA ORB on two different > processors in our system. One of them is an INOVA Pentium3 board and the > other is an INOVA K6 board. Both of the cards have the exact same CORBA > .out files loaded on them (libOEorb.out and libOEtcp.out) but when I try to > run a static function called 'oneway_sync_scope' it works fine on the K6 > baord but, it fails on the P3 board and I get an error saying: > > unresolved symbol: oneway_sync_scope( ) > > When I try running a 'lkup' on the oneway_sync_scope command, both > processors say the following: > > CORBA::ORB::oneway_sync_scope(short) 0x07e510e0 text (libOEorb.out) > CORBA::ORB::oneway_sync_scope(void) 0x07e51100 text (libOEorb.out) > CORBA::ORB::oneway_sync_scope 0x07e789ec data (libOEorb.out) > > So, it looks like they both have the function loaded. The only thing I can > think of that would be the problem is that there is a difference in the > vxWorks image between the one running on the K6 and the one running on the > P3 that is causing this. Other than this problem with this one CORBA > function, the rest of the CORBA functionality seems to work fine on both > processors. Anyone have any idea what the problem might be (i.e. what the > difference is between the K6 and P3 kernels would be that would cause this)? > > Thank you, > John Sestak > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ipAttach ipDetach memory leak problem Date: 4 Sep 2002 06:37:08 -0700 From: gabi@tdsoft.com (Gabi) Organization: http://groups.google.com/ Message-ID: <5607ff2c.0209040537.51ab2227@posting.google.com> References: <5607ff2c.0208180549.7ad53054@posting.google.com> <4a7f0d21.0208181900.64784a5c@posting.google.com> <5607ff2c.0208200214.5468d0c9@posting.google.com> hi All The following code seem to solve the problem: { struct ifnet *Unit; ipAttach(2,"fec"); Unit = ifunit("fec2"); ipDetach(2,"fec"); free(Unit->if_name); } tanks all Gabi --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Subroutines and libraries Date: 4 Sep 2002 07:03:34 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0209040603.43a96f07@posting.google.com> References: <46992b69.0209040259.96fc2d8@posting.google.com> compaqbala@sify.com (Balakumar) wrote in message news:<46992b69.0209040259.96fc2d8@posting.google.com>... > between subroutines and libraries in vxworks. In vxWorks 5.4.1 there is no (technical) difference. Object code is linked together w/o making a difference. This gives a nice flat binary. vxWorks AE makes this difference. (shared) libraries have deticated entry points and are to be stored at deticated locations. toni --------------------------- Newsgroups: comp.os.vxworks Subject: Rescheduling Problem Date: Wed, 4 Sep 2002 17:23:38 +0100 From: "David" Message-ID: <3d76257a$0$23626$4d4ef98e@read.news.ch.uu.net> Hi, i have a task with prio 100. Now i want to give the CPU to an other task that has the same prio. Because the second task has only very little work to do, I do not want the first task to delay for 10ms (that is one system tick). Is there a way to get the first task running just after the second task has finished it's work? we do not have the "sched_yield" function available in our kernel. Thanks a lot for all tips David --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VMIC 6001 with Vxworks Date: 4 Sep 2002 08:29:17 -0700 From: terry.butman@vmic.com (Terry Butman) Organization: http://groups.google.com/ Message-ID: <6db76185.0209040729.957d7be@posting.google.com> References: <3d6ebfe6_3@nopics.sjc> "Cedric Le Roux" wrote in message news:<3d6ebfe6_3@nopics.sjc>... > I'm looking for some clarification regarding the programming for diagnostic > mode for the VMIC 6001 MIL-STD-1553 board. The manual is kinda vague. > > If anyone out there has worked with this board, I'd like to hear from ya. > or if anyone knows a better place for this post, I'd like to hear about that > too. > > Thanks. Heres is an example. as best I remember you must have the 1553 bus properly connected and terminated to use external mode, you don't need it to run internal mode. In this case the board is addressed at VMEbus A32 address 0x4000000 This is the test setup for running diagmode on board 1 bus a - external There will be a total of 8 memory test words There will be a total of 8 1553 bus test words that should echo There will be a two extra 1553 bus test words that should not echo Step 1. set the diamode control block at offset 0x2000 Step 2. execute a clear unit command by writing to the operation command register 4000000 0201= Step 3. write the control block starting address by writing to the operation command data in register by writing to the operation command register 4000002 1000= 4000000 0100= Step 3. write the diagmode start command by writing to the operation command data in register by writing to the operation command register 4000002 0002= 4000000 0200= Step 4. examine the appropriate registers to determine pass/fail examine the output waveform for amplitude/timing examine the output waveform for data count Memory dump of the I/O memory map m 040020000 4000000 0000 : word 000 operation command register 4000002 0002 : word 001 operation command data in register 4000004 0000 : word 002 operation command data out register 4000006 0000 : word 003 active status register ( should = 0x0000 ) 4000008 0000 : word 004 interrupt pointer 400000A 0000 : word 005 off pointer 400000C 0002 : word 006 operational mode register ( should = 0x0002 ) 400000E 2000 : word 007 programming error register ( should = 0x0000 ) 4000010 1000 : word 008 current control block pointer ( should = 0x1000 ) 4000012 0000 : word 009 reserved 4000014 0000 : word 010 reserved 4000016 0000 : word 011 last status response 4000018 0000 : word 012 last command word 400001A 0000 : word 013 reserved 400001C 0000 : word 014 reserved 400001E 0000 : word 015 reserved Diagmode control block is set up at address 04002000 ? m 04002000 4002000 0000 : word 000 diagmode control word ( bus a external ) 4002002 1000 : word 001 link address ( linked to ourselves ) 4002004 0008 : word 002 memory test count set to 8 4002006 0008 : word 003 bus test count set to 8 4002008 FFFF : word 004 memory test word #1 400200A FFFF : word 005 memory test word #1 echo 400200C AAAA : word 006 memory test word #2 400200E AAAA : word 007 memory test word #2 echo 4002010 DEAD : word 008 memory test word #3 4002012 DEAD : word 009 memory test word #3 echo 4002014 0000 : word 010 memory test word #4 4002016 0000 : word 011 memory test word #4 echo 4002018 0f0f : word 012 memory test word #5 400201A 0f0f : word 013 memory test word #5 echo 400201C f0f0 : word 014 memory test word #6 400201E f0f0 : word 015 memory test word #6 echo 4002020 dead : word 016 memory test word #7 4002022 dead : word 017 memory test word #7 echo 4002024 beef : word 018 memory test word #8 4002026 beef : word 019 memory test word #8 echo 4002028 ffff : word 020 1553 bus test word #1 400202A ffff : word 021 1553 bus test word #1 echo 400202C 0000 : word 022 1553 bus test word #1 ISR 400202E 0000 : word 023 1553 bus test word #1 TSR 4002030 aaaa : word 024 1553 bus test word #2 4002032 aaaa : word 025 1553 bus test word #2 echo 4002034 0000 : word 026 1553 bus test word #2 ISR 4002036 0000 : word 027 1553 bus test word #2 TSR 4002038 5555 : word 028 1553 bus test word #3 400203A 5555 : word 029 1553 bus test word #3 echo 400203C 0000 : word 030 1553 bus test word #3 ISR 400203E 0000 : word 031 1553 bus test word #3 TSR 4002040 0000 : word 032 1553 bus test word #4 4002042 0000 : word 033 1553 bus test word #4 echo 4002044 0000 : word 034 1553 bus test word #4 ISR 4002046 0000 : word 035 1553 bus test word #4 TSR 4002048 0f0f : word 036 1553 bus test word #5 400204A 0f0f : word 037 1553 bus test word #5 echo 400204C 0000 : word 038 1553 bus test word #5 ISR 400204E 0000 : word 039 1553 bus test word #5 TSR 4002050 f0f0 : word 040 1553 bus test word #6 4002052 f0f0 : word 041 1553 bus test word #6 echo 4002054 0000 : word 042 1553 bus test word #6 ISR 4002056 0000 : word 043 1553 bus test word #6 TSR 4002058 dead : word 044 1553 bus test word #7 400205A dead : word 045 1553 bus test word #7 echo 400205C 0000 : word 046 1553 bus test word #7 ISR 400205E 0000 : word 047 1553 bus test word #7 TSR 4002060 beef : word 048 1553 bus test word #8 4002062 beef : word 049 1553 bus test word #8 echo 4002064 0000 : word 050 1553 bus test word #8 ISR 4002066 0000 : word 051 1553 bus test word #8 TSR 4002068 1234 : word 052 1553 bus test word #9 400206A 0000 : word 053 1553 bus test word #9 echo 400206C 0000 : word 054 1553 bus test word #9 ISR 400206E 0000 : word 055 1553 bus test word #9 TSR 4002070 5678 : word 056 1553 bus test word #10 4002072 0000 : word 057 1553 bus test word #10 echo 4002074 0000 : word 058 1553 bus test word #10 ISR 4002076 0000 : word 059 1553 bus test word #10 TSR 4002078 0000 : 400207A 0000 : pass/fail criterial memory test word echoes 1-8 should match their respective test words 1553 bus test word echoes 1-8 should match their respective test words 1553 bus test word echoes 9-10 should remain 0x0000 ( test count = 8 ) 1553 bus test word ISRs 1-8 should == 0x0000 ( no errors ) 1553 bus test word TSRs 1-8 should == 0x0000 ( no errors ) Active status error register should == 0x0000 ( no errors ) Programming error register should == 0x0000 ( no programming errors ) Hope this helps, by the way thanks for the praise of my manual. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Rescheduling Problem Date: Wed, 4 Sep 2002 17:38:00 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <3d76257a$0$23626$4d4ef98e@read.news.ch.uu.net> You can do one of these: 1. If the current task does taskDelay(0), it will give CPU to the next READY task with equal priority, and put itself at the end of the ReadyQ for the same priority. Hence what the second task is done (pends), the first will run again, assuming these are the only two with priority 100, and there are no higher priority tasks becoming ready at the time. 2. If the second task has so little work to do, why won't you set its priority at e.g. 99 ? "David" wrote in message news:3d76257a$0$23626$4d4ef98e@read.news.ch.uu.net... > Hi, > i have a task with prio 100. Now i want to give the CPU to an other > task that has the same prio. Because the second task has only very > little work to do, I do not want the first task to delay for 10ms > (that is one system tick). > Is there a way to get the first task running just after the second task > has finished it's work? > we do not have the "sched_yield" function available in our kernel. > > Thanks a lot for all tips > David > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: how to implement break point Date: Wed, 04 Sep 2002 16:00:55 GMT From: "Hadi" Organization: Magma Communications Ltd. Message-ID: Hello everyone, I am trying to implement some thing like "break point", which would allow me to hook my user define code at any address during runtime. After excuting my code I want to return back to the original execution path. I am using MPC 860 architecture. I sort of know how to do that in assembly. But if any of you guys already has done it in C or assembly , Please send me your source code. Thanks Hadi --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Problem with HTTP Server and SLIP Date: 4 Sep 2002 09:06:56 -0700 From: Oleg.Chtchoupak@mtech-ag.com (Oleg) Organization: http://groups.google.com/ Message-ID: <16d0deb4.0209040806.6381649b@posting.google.com> References: <16d0deb4.0209040005.2cfca43d@posting.google.com> "Leonid Rosenboim" wrote in message news:... > Oleg, can you be more specific about exactly how > the SLIP links appears to be broken ? > We control the connection timeout that is more then 2 sec from both sides (the message MUST be sent each 500 msec) and wen timeout is > 2 sec Errorhandling will activate. > Does the SLIP link brake if you simply transfer large files > over the SLIP link or not ? > Sometimes, but very seldom the SLIP brake (timeout) without large file transfer. After timeout the socket is alive ( We test that with ping). SLIP with COM1 is IRG4 Ehternet is IRG12. > "Oleg" wrote in message > news:16d0deb4.0209040005.2cfca43d@posting.google.com... > > Hallo all, > > > > we use the distributed system with two CPU under VxWorks. The CPUs > > communicate using SLIP. > > One CPU is DIMM PC Pentium under vxWorks 5.4 and here runs also > > GoAhead Web server. > > > > The second one is M68020 under VxWorks 5.31. > > When the Webserver is no in use (PEND+T) the SLIP communication > > between CPUs is OK. After we start a browser and get some images from > > the DIMM PC Target, the SLIP communication is broken. > > Do anyone now the explanation for this phenomenon? > > Can anyone help to resolve this problem? > > > > Best regards. > > Oleg. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: netBufLib Date: Wed, 04 Sep 2002 12:37:02 -0400 From: Neil Horman Organization: WEBUSENET.com Message-ID: <3D7636AE.5020104@lvl7.com> References: <74371f59.0209031120.74311c68@posting.google.com> Reply-To: nhorman@lvl7.com Tulasidas wrote: > I am porting the VxWorks code to UNIX OS. The VxWorks code uses > data structures such as mBlks and clBlks. Do we have an > equivalent data structures in Unix OS. > > > Thanks > > Tulasi > There is no "standard" network buffer management API for unix systems, so unfortunately you need to interrogate the manual or the source code for the system you are porting to. I can tell you off the top of my head that if linux is your targeted OS, the structure you are looking for is the sk_buff. Hope that helps! Neil --------------------------- Newsgroups: comp.os.vxworks Subject: FLASH lockup on PPC405GP eval board? Date: Wed, 04 Sep 2002 17:13:36 GMT From: "Charles" Organization: Prodigy Internet http://www.prodigy.com Message-ID: <4brd9.291$AO.1843367@newssvr16.news.prodigy.com> We are using WindRivers PPC405GP evaluation board. The onboard FLASH periodically locks up. The unit no longer responds to the terminal port on boot and will not download VxWorks image on ethernet connection. The only way to recover is to rewrite the FLASH with JTAG connection. Anyone else experience this? What is the cause? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Problem with HTTP Server and SLIP Date: Wed, 4 Sep 2002 19:18:56 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <16d0deb4.0209040005.2cfca43d@posting.google.com> <16d0deb4.0209040806.6381649b@posting.google.com> "Oleg" wrote in message news:16d0deb4.0209040806.6381649b@posting.google.com... > "Leonid Rosenboim" wrote in message news:... > > Oleg, can you be more specific about exactly how > > the SLIP links appears to be broken ? > > > > We control the connection timeout that is more then 2 sec from both > sides (the message MUST be sent each 500 msec) and wen timeout is > 2 > sec Errorhandling will activate. > So, you are sending hearbeat packets eachway every half a second, and if four packets have been dropped in a row, you declare the SLIP connection dead ? Consider this: when a slow link is loaded with a data transfer task that is more than the link speed, some of the packets will need to be dropped, while others will be significantly delayed while waiting in the buffer queue. Example: you have a 19200 link with 16 KByte buffers, a packet could be waiting in the buffer queue for up to 8 seconds before being transmitted. I would suggest a different approach to detecting link down condition: continue to send these heartbeat packets, but measure the "received pkt" counter of the network interface (see with ifShow), so if no packets at all where received in 2 seconds (where even if no application traffic, the HB should have xmitted 4 packets), then the link is indeed down. > > Does the SLIP link brake if you simply transfer large files > > over the SLIP link or not ? > > > > Sometimes, but very seldom the SLIP brake (timeout) without large file > transfer. > After timeout the socket is alive ( We test that with ping). > SLIP with COM1 is IRG4 > Ehternet is IRG12. > There really isn't any problem with all this, except that you are not sufficiently competent in the area of communications. Please see this a constructive criticism, and hire someone who knows communications to work with you. - - Leonid > > "Oleg" wrote in message > > news:16d0deb4.0209040005.2cfca43d@posting.google.com... > > > Hallo all, > > > > > > we use the distributed system with two CPU under VxWorks. The CPUs > > > communicate using SLIP. > > > One CPU is DIMM PC Pentium under vxWorks 5.4 and here runs also > > > GoAhead Web server. > > > > > > The second one is M68020 under VxWorks 5.31. > > > When the Webserver is no in use (PEND+T) the SLIP communication > > > between CPUs is OK. After we start a browser and get some images from > > > the DIMM PC Target, the SLIP communication is broken. > > > Do anyone now the explanation for this phenomenon? > > > Can anyone help to resolve this problem? > > > > > > Best regards. > > > Oleg. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: FLASH lockup on PPC405GP eval board? Date: Wed, 04 Sep 2002 14:39:05 -0400 From: Varndell Engineering Organization: Posted Via Newsfeeds.com = SPEED+RETENTION+COMPLETION = http://www.newsfeeds.com Message-ID: <3D765349.18493F90@hotmail.com> References: <4brd9.291$AO.1843367@newssvr16.news.prodigy.com> Charles wrote: > > We are using WindRivers PPC405GP evaluation board. > The onboard FLASH periodically locks up. The unit no > longer responds to the terminal port on boot and will > not download VxWorks image on ethernet connection. > > The only way to recover is to rewrite the FLASH with JTAG connection. > > Anyone else experience this? What is the cause? Yep, all the time. Sectors get erased and/or locked somehow. Don't know why, but you're not alone. Regards, George Varndell - -- Varndell Engineering, LLC Embedded Systems Expertise http://www.varndellengineering.com - -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- http://www.newsfeed.com The #1 Newsgroup Service in the World! - -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: netBufLib Date: 4 Sep 2002 12:55:50 -0700 From: tulasi@sionintl.com (Tulasidas) Organization: http://groups.google.com/ Message-ID: <74371f59.0209041155.71e7bc5c@posting.google.com> References: <74371f59.0209031120.74311c68@posting.google.com> Hi Leonid, "Leonid Rosenboim" wrote in message news:... > This would greatly depend on the type of code you are porting. > > If you are porting an appkication module, it will not be able to > access the Unix mbufs structures, which are part of the kernel > space. If you are porting a device driver, it would be quite surprising > for you to ask this kind of questions, because you should have been > already familiar with the device driver APIs on Unix by now. I am not porting a device driver. What I am doing is that to replace the VxWorks MUX layer, so that my application will work on a UNIX system. And more over the memory management in UNIX is different from VxWorks memory pool management. If you have done this type of work before or you have any constructive ideas please do so. Thanks. Tulasidas --------------------------- Newsgroups: comp.os.vxworks Subject: Training Date: 4 Sep 2002 13:38:09 -0700 From: pnp.williams@cox.net (PaulW) Organization: http://groups.google.com/ Message-ID: <34f85dc7.0209041238.631e8b84@posting.google.com> Dear VxWorks Experts, We are considering taking on a project to develop some communications control software for a PowerPC-based SBC and do not have VxWorks experience. We have some time before we begin and would like to hear recommendations for training activities that we should look into prior to startup. The team mostly comes from a C++ and Java background developing message handling and comm protocol software for UNIX platforms. Any pratical experience with training programs out there, or are you all self-taught? Thanks in advance. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VMIC 6001 with Vxworks Date: Wed, 4 Sep 2002 17:11:03 -0400 From: "Cedric Le Roux" Message-ID: <3d767681$1_2@nopics.sjc> References: <3d6ebfe6_3@nopics.sjc> <6db76185.0209040729.957d7be@posting.google.com> Terry, You should've mentioned that you worked on it. i would've been more tactful. But now that I've got your attention.... The cutting and pasting that took place in this manual is driving me nuts. I've got some of the same information pasted in three different places. I've got the Remote Terminal Control Word bit definition pasted from somewhere in the Bus Controller pages and is completely wrong. (Making it impossible to program RTs with this information I might add). And the most frustrating is stuff like this: "Major Time Frame The Major time frame is a 16 bit register that should be set to what the major time frame should be." VMIC 6001 manual I'm convinced the statement is factually accurate, but it's one of those things that you read and say, "Well duh.....but what is it?" It's lucky for me that I had a manual from the VMIC 6000 and an older version of the VMIC 6001 manual, since they have fewer errors due to cutting and pasting. You have my sincere respect, but please understand my position. (1553 newbie under time constraints) Thank you for the example, I'll post again tomorrow after I run the test. Cedric --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Problem with HTTP Server and SLIP Date: Wed, 4 Sep 2002 21:47:50 +0000 (UTC) From: David Laight Organization: BT Openworld Message-ID: References: <16d0deb4.0209040005.2cfca43d@posting.google.com> <16d0deb4.0209040806.6381649b@posting.google.com> > > So, you are sending hearbeat packets eachway every half a second, > and if four packets have been dropped in a row, you declare the SLIP > connection dead ? You also need to allow for the 'modem retrain' time if you might have real modems. This is several seconds long... David --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Training Date: Thu, 05 Sep 2002 01:14:37 GMT From: "drdiags" Organization: AT&T Broadband Message-ID: <1eyd9.299833$983.535861@rwcrnsc53> References: <34f85dc7.0209041238.631e8b84@posting.google.com> Hello Paul, You could review the available vxWorks training on the www.windriver.com site. If you are on the East Coast, the consultants from the Bryan Group gave a pretty good vxWorks BSP training class, which now runs 4 days instead of 3 when I went. I would also recommend that you send 1-2 guys from your team to take the device driver class. The rest of your team could try to work a deal to get a vxWorks Tornado trainer to come on-site to reduce cost, but it is a 3 day class (I think). Since your team is probably comfortable with command line builds and such, Tornado training may not be high on your list of requirements, so maybe send 1 guy to understand the two build environments (command line vs GUI). The last suggestion I would share is that once you get the product, get a feel for the on-line help. There is a lot of information available there, though if you monitor this group you will find varying opinions on how clear and concise it is in every aspect. How much time before the project starts may impact you, since the WindRiver produced classes you need may not all fall in the time cycle you want. HTH "PaulW" wrote in message news:34f85dc7.0209041238.631e8b84@posting.google.com... > Dear VxWorks Experts, > > We are considering taking on a project to develop some communications > control software for a PowerPC-based SBC and do not have VxWorks > experience. We have some time before we begin and would like to hear > recommendations for training activities that we should look into prior > to startup. The team mostly comes from a C++ and Java background > developing message handling and comm protocol software for UNIX > platforms. Any pratical experience with training programs out there, > or are you all self-taught? Thanks in advance. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Training Date: Thu, 05 Sep 2002 01:31:43 GMT From: "drdiags" Organization: AT&T Broadband Message-ID: <3uyd9.154656$_91.210324@rwcrnsc51.ops.asp.att.net> References: <34f85dc7.0209041238.631e8b84@posting.google.com> <1eyd9.299833$983.535861@rwcrnsc53> Paul, Also, that may not have been the correct name, it might be the Ptr Group (?). Also, on the East Coast you could try contacting this website: www.mccengineering.com A startup, but several former Wind River field engineers with extensive training expertise. HTH..... > You could review the available vxWorks training on the www.windriver.com > site. If you are on the East Coast, the consultants from the Bryan Group [......snip.....] > "PaulW" wrote in message > news:34f85dc7.0209041238.631e8b84@posting.google.com... > > Dear VxWorks Experts, > > > > We are considering taking on a project to develop some communications > > control software for a PowerPC-based SBC and do not have VxWorks > > experience. We have some time before we begin and would like to hear > > recommendations for training activities that we should look into prior > > to startup. The team mostly comes from a C++ and Java background > > developing message handling and comm protocol software for UNIX > > platforms. Any pratical experience with training programs out there, > > or are you all self-taught? Thanks in advance. > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: gets() skips first char Date: Thu, 05 Sep 2002 02:01:22 GMT From: Pete Kockritz Organization: AT&T Broadband Message-ID: References: In article , "Joachim Bassendowski" wrote: > > a - Arinc 429 Unit Test > m - MIL 1553 Unit Test > q - Quit Unit Test > -> m > 00000000: 0011-123 > "23" > abc > "bc" > > > After displaying "q - Quit Unit Test" I entered return. The result was the > prompt ->. Other entries here lead to an error. Then I entered 'm' and the > computer showed "00000000: 0011". Then I entered 123 but the first char > wasn't stored to the char-array. The same with input abc and result bc. > > Please, can someone explain this behaviour to me? The target shell is getting some of the input that you intended for your program. The 'm' you entered is interpreted by the shell as the command to modify memory. The '->' chars are the shell's prompt: > -> m > 00000000: 0011-123 The shell is prompting you to modify the word at address zero, the current value at that address is 0x0011. Your program is getting some of the input, but not all. Try running your program without the shell running. Is this output that you show coming out on the windShell, target console or the virtual console? - -- Pete Kockritz --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Help: disk cache error Date: 4 Sep 2002 19:37:35 -0700 From: ritalin@comcast.net (Rita) Organization: http://groups.google.com/ Message-ID: References: If you get any answer for this, please let me know as well. I've seen this problem before, and our flash was not removed. rita "R Hilderman" wrote in message news:... > Can someone please help explain exactly what the following logMsg means? > > 0x1f7bd490 (tDcacheUpd): disk cache error: device 1f7bd6b8 block 252 errno > c0006, disk removed while writing data, probable data loss > > The errno 0xc0006 means S_ioLib_DISK_NOT_PRESENT. > Does the error message mean that tDcacheUpd has removed the disk while > writing data > or does it mean it mearly timed out while writing to the disk (or something > similar). > > I would like to restart the system if the disk is removed, as the disk is > our flash file system. > Does anyone know how I would bind in for this? > > Any help would be appreciated. > > Thanks, > Randy --------------------------- Newsgroups: comp.os.vxworks Subject: visionCLICK Simple vs. Complete/Full RAM tests Date: 4 Sep 2002 19:55:32 -0700 From: ryeung@earthlink.net (Raymond Yeung) Organization: http://groups.google.com/ Message-ID: Does anyone know the main differences in test cases the "Simple" and the "complete/Full" RAM tests are? The complete tests are a lot longer than the simple ones, and I'm wondering what faults I'd be missing if I use the simple ones instead. Thanks, Raymond --------------------------- Newsgroups: comp.os.vxworks Subject: Re: netBufLib Date: Thu, 5 Sep 2002 08:40:47 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <74371f59.0209031120.74311c68@posting.google.com> <74371f59.0209041155.71e7bc5c@posting.google.com> Hi again, I might have constructive ideas, but you have provided very limitted information to go on. My guess is that your reference to MUX actually refers to the ability to install a packet filter or a custom protocol stack to work on top of END drivers in VxWorks. On UNIX these things are done differently: for packet filtering, there is a specific packet level interface, such as use dby "tcpdump" to sniff all packets off an ethernet ( ised this type of API many years ago on SunOS 4.1, and I know it has changed in Solaris 2.x, and is probably different on Unix), so your solution may depend on WHICh type of Unix you are talking about. If you are porting a protocol stack, then if it was Solaris, you would need to convert it into a set of STREAMS modules, and make them part of kernel. Linux and other BSD derivatives dont use STREAMS but have their own method of adding protocols. Lastly, you might actually get much better information by posting to the UNIX-specific newsgroups, where you will find experts with up to date info on the perticular Unix derivative you need to work with. HTH, - - Leonid "Tulasidas" wrote in message news:74371f59.0209041155.71e7bc5c@posting.google.com... > Hi Leonid, > > "Leonid Rosenboim" wrote in message news:... > > This would greatly depend on the type of code you are porting. > > > > If you are porting an appkication module, it will not be able to > > access the Unix mbufs structures, which are part of the kernel > > space. If you are porting a device driver, it would be quite surprising > > for you to ask this kind of questions, because you should have been > > already familiar with the device driver APIs on Unix by now. > > I am not porting a device driver. What I am doing is that to replace > the VxWorks MUX layer, so that my application will work on a UNIX > system. And more over the memory management in UNIX is different from > VxWorks memory pool management. If you have done this type of work > before or you have any constructive ideas please do so. > > Thanks. > > Tulasidas --------------------------- Newsgroups: comp.os.vxworks Subject: USB Host driver Date: Thu, 5 Sep 2002 12:23:09 +0530 From: "Biju MG" Organization: Texas Instruments Message-ID: Hi, Has anybody used the USB Host driver from Vxworks, How easy is it to port to some platforms, and has anybocy come across any bugs in Host driver --------------------------- Newsgroups: comp.os.vxworks Subject: How to connect an ISP to a VME interrupt Date: Thu, 05 Sep 2002 07:06:18 GMT From: Hans Berglund Message-ID: <3D770268.20509@spacetec.no> I have an MVME2400, and I am going to connect an Interrupt Service Routine to an interrupt on VMEbus level 6. I do the following calls: intConnect ((INUM_TO_IVEC(UNIV_INT_LVL0)), myISP, 0); intEnable (UNIV_INT_LVL0); sysIntEnable (Level); The three calls return OK. "UNIV_INT_LVL0" is where I think the VME interrupts are routed. "Level" is 6. I can see in the Universe registers that the interrupt status is set for level 6, but myISP is never called. This must be used a lot in other systems, so why does not mine work? Any suggestions are welcome. Hans Berglund --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How to connect an ISP to a VME interrupt Date: Thu, 5 Sep 2002 09:42:51 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <3D770268.20509@spacetec.no> Reply-To: "Michael Lawnick" Sender: mlawnick@pd9e656a9.dip.t-dialin.net Hi Hans, you are digging one step to deep in system! VME is fully supported by MVMExxxx BSPs. The interrupt vector you have to connect to is the value (function number) given on sysBusIntGen on the other side: destination target: intConnect((VOIDFUNCPTR *)funcNum,(VOIDFUNCPTR)vmeISR,0); sysIntEnable(1); source target: sysBusIntGen(intLvl,funcNum); The interrupt level on VME-bus (intLvl) will determine the priority and must be enabled for input exactly for one target in system. It is not visible for S/W on destination. The function number you transfer (funcNum) should not collide with the predefined interrupt numbers on target, you should use 0x80<=intNum<=0xFF. HTH Michael "Hans Berglund" schrieb im Newsbeitrag news:3D770268.20509@spacetec.no... > > I have an MVME2400, and I am going to connect an Interrupt > Service Routine to an interrupt on VMEbus level 6. > > I do the following calls: > > intConnect ((INUM_TO_IVEC(UNIV_INT_LVL0)), myISP, 0); > intEnable (UNIV_INT_LVL0); > sysIntEnable (Level); > > The three calls return OK. > "UNIV_INT_LVL0" is where I think the VME interrupts are routed. > "Level" is 6. > > I can see in the Universe registers that the interrupt status is set > for level 6, but myISP is never called. This must be used a lot in > other systems, so why does not mine work? > > Any suggestions are welcome. > > Hans Berglund > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How to connect an ISP to a VME interrupt Date: Thu, 5 Sep 2002 09:46:13 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <3D770268.20509@spacetec.no> Reply-To: "Michael Lawnick" Sender: mlawnick@pd9e656a9.dip.t-dialin.net Correction: must be sysIntEnable(intLvl), not sysIntEnable(1) - (copy&paste). Regards Michael --------------------------- Newsgroups: comp.os.vxworks Subject: Re: FLASH lockup on PPC405GP eval board? Date: Thu, 5 Sep 2002 10:05:09 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <4brd9.291$AO.1843367@newssvr16.news.prodigy.com> <3D765349.18493F90@hotmail.com> Reply-To: "Michael Lawnick" Sender: mlawnick@pd9e656a9.dip.t-dialin.net Can't agree. Our flashes run perfectly. BTW: PPC405GP evaluation board is from IBM, not WRS. Regards, Michael "Varndell Engineering" schrieb im Newsbeitrag news:3D765349.18493F90@hotmail.com... > Charles wrote: > > > > We are using WindRivers PPC405GP evaluation board. > > The onboard FLASH periodically locks up. The unit no > > longer responds to the terminal port on boot and will > > not download VxWorks image on ethernet connection. > > > > The only way to recover is to rewrite the FLASH with JTAG connection. > > > > Anyone else experience this? What is the cause? > > Yep, all the time. Sectors get erased and/or locked somehow. > Don't know why, but you're not alone. > > Regards, > George Varndell > -- > Varndell Engineering, LLC > Embedded Systems Expertise > http://www.varndellengineering.com > > > -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- > http://www.newsfeed.com The #1 Newsgroup Service in the World! > -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: VxWorks Applications Engineer Date: 5 Sep 2002 01:42:36 -0700 From: Bam2974@hotmail.com (Bamidelle) Organization: http://groups.google.com/ Message-ID: <66a11529.0209050042.72b7875d@posting.google.com> Hi, I am looking for an experienced Engineer with either an applications engineering or a software engineering background and strong VxWorks experience. We are a leading edge developer of: VME * CPCI * Multibus II * PMC * Real-time OS * Custom Solutions *. We are looking for someone highly technical with good communication skills. You need to be authorised to work in the UK. We will pay up to £ 35 k and offer strong package and career path. Is there anyone out there who can help ? Please reply with full CV and contact details for informal + totally confidential discussion. --------------------------- Newsgroups: comp.os.vxworks Subject: Exception number 14? Date: Thu, 05 Sep 2002 11:09:52 +0200 From: Paul Augart Organization: OCE Printing Systems GmbH Message-ID: <3D771F60.69368270@ops.de> This is a multi-part message in MIME format. - --------------3CB48FFCCCCDA0D9E466E793 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Hello All, I seem to be have "writers blocked". I am getting this error: "-> Exception number 14: Task: 0x78938cc (cePPWRK) Page Fault Program Counter: 0x0120194c Status Register: 0x00010246 Error Code : 0x00000002 11f9886 _SendPagedensityInfo__FUsUlUlUli14teCeppDevUnits+133e: 11f7e18 (7892898, 5, 2) 11f80e1 _initTraceMask +320d: 1201878 (&_Pag_Inf, 5, 7892798) " I know Exception number 14 is a Page Fault. But what is Error Code : 0x00000002. Where do I find it. when I type printErrno in shell I get this: - -> printErrno 0x00000002 0x2 = S_errno_ENOENT value = 0 = 0x0 I do not know if this is the correct error message for my error. Also what does NOENT stand for? What I would like to know is where to look for Exception number information, so I can help myself in the future Thank you in advance. - -Paul Paul.Augart@ops.de - --------------3CB48FFCCCCDA0D9E466E793 Content-Type: text/x-vcard; charset=us-ascii; name="Paul.Augart.vcf" Content-Transfer-Encoding: 7bit Content-Description: Card for Paul Augart Content-Disposition: attachment; filename="Paul.Augart.vcf" begin:vcard n:Augart;Paul tel;fax:+49-8121-72-4322 tel;work:+49-8121-72-3797 x-mozilla-html:TRUE org:Océ Printing Systems GmbH;PDE 41 adr:;;Siemensalle 2;85586 Poing;;;Germany version:2.1 email;internet:Paul.Augart@ops.de fn:Augart, Paul end:vcard - --------------3CB48FFCCCCDA0D9E466E793-- --------------------------- Newsgroups: comp.os.vxworks Subject: Hierarchial subproject model and vxworks with tornado Date: 5 Sep 2002 02:30:09 -0700 From: vxnerd@rediffmail.com (subbarayan) Organization: http://groups.google.com/ Message-ID: Dear mentors, Tornado2.0 release notes documentation pdf files gives and mentions about making hierarchial subprojects and linking them to gether.I am using tornado2.0 and target is standard pc architecture with x86 processor.I followed the same proceedure exactly whats mentioned in the manual to make subprojects and making them as single object code.After following the steps mentioned to make subprojects and link them ,when i tried to build vxworks i get the following error: make: *** No rule to make target `BuildProject4', needed by `master'. Stop. To clarify further i would like to mention that i made a bootable project called master and two downloadable projects called project4 and project5.Then in my master project in the builds tab i selected default and accessed the build properties dialogue box.Then i selected the rules pane and then clicked on new/edit button. Then i defined a new rule and typed master in target box and entered BuildProject4 BuildProject5 vxWorks in dependencies dialog box.Then i unchecked invoke this rule before building project check box. Then similarly as mentioned in manual i created a clean rule. then after that clicked apply.Then i set my default rule to be master. after all these things selected macros in build properties and defined the following path where my gnu source for file is present: D:/Tornado/target/proj/master/project4/SIMNTgnu/project4.a D:/Tornado/target/proj/master/project5/SIMNTgnu/project5.a.Then i clicked apply.then returned to files tab and clicked on build menu.I got the above said error. But one more thing to say theres not actually a .a file in the path mentioned. will the .a file be created automatically or before building i must compile both of my downloadable projects? can anyone point me where will be the source of error and help me overcome this error? all the 3 projects are in same workspace and same folder called master. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: how to implement break point Date: 5 Sep 2002 02:42:10 -0700 From: cpotenza@onetel.net.uk (Claudio Potenza) Organization: http://groups.google.com/ Message-ID: <2025ac82.0209050142.1ff33daa@posting.google.com> References: "Hadi" wrote in message news:... > Hello everyone, > I am trying to implement some thing like "break point", which would allow me > to hook my user define code at any address during runtime. After excuting > my code I want to return back to the original execution path. I am using MPC > 860 architecture. I sort of know how to do that in assembly. But if any of > you guys already has done it in C or assembly , Please send me your source > code. A no-effort way to hook your code is to use the "e" function (that sets 'eventpoints', it is in dbgLib). if you make a hook function like: int myHook(void) { return 1; } and then you do the following on the shell prompt: -> e ,1,0,myHook,0 then every time the instruction at is executed, 'myHook()' also will be executed (first). This can be quite useful but there are a few drawbacks: 1) you cannot hook interupt code 2) you cannot hook on unbreakable task (but you can make them breakable using taskOptionSet) 3) As far as I know there is no easy way to look at locals of the hooked function See the entry for the "e" function in the reference manual for more details. --------------------------- Newsgroups: comp.os.vxworks Subject: Athlon XP Date: 5 Sep 2002 03:37:27 -0700 From: egoisone@libero.it (Mic) Organization: http://groups.google.com/ Message-ID: Hi VxWorkers, does BSP-Pentium/Pro work on a Athlon XP? I made a boot disk, but the blue boot screen does not appear and everything stops. I guess it should be many many problems, but not so early! Have you any suggestions(BIOS problem, video card or what else. Everything happens before the system tries to load VxWorks image so I think the problem is the target pc...problably it does non depends on Athlon...)? --------------------------- Newsgroups: comp.os.vxworks Subject: TFFS Date: 5 Sep 2002 03:56:18 -0700 From: pradeeppatel77@yahoo.com (pradeep) Organization: http://groups.google.com/ Message-ID: <54575a52.0209050256.2ad0d9b3@posting.google.com> References: <54575a52.0209012124.61f54385@posting.google.com> I am using flash IFM 441 can u please tell me when to include TFFS when when not, is it nedded to have TFFS if i am using flash disk to load the code. IS TFFS will work with IFM 441 or not I am able to boot from IFM 441(Without TFFS ) but as frequently i have to do write operations will u suggest me to include TFFS?? regards pradeep pradeeppatel77@yahoo.com (pradeep) wrote in message news:<54575a52.0209012124.61f54385@posting.google.com>... > I amusinf flas IFM 350 > can u please tell me when to include TFFS when when not, is it nedded > to have TFFS if i am using flash disk to load the code > > regards > pradeep --------------------------- Newsgroups: comp.os.vxworks Subject: vxworks clean image building Date: 5 Sep 2002 04:06:24 -0700 From: vxnerd@rediffmail.com (subbarayan) Organization: http://groups.google.com/ Message-ID: dear mentors, Can anyone explain me the role of vxworks clean image?whats the purpose of clean image where will it be useful? One more question in tornado release notes manual i was asked to define the following rule for managing hierarchial subprojects: on the build tab of downloadable project,I was asked to define a new rule by adding Clean in target box and in dependencies box i was asked to give Cleanproject8 and Cleanproject9 where proj8 and 9 are my downloadable projects and i defined this clean rule for a bootable project by name master. I am quite unaware what does these Cleanproject8 and 9 mean? Can anyone explain what it is? After defining the rule if i compile tornado i get following error: make: *** No rule to make target `CleanProject9', needed by `clean'. Stop. Now can anyone point whats this error due to and how to overcome this error? Thanks in advance to person replying me. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: TFFS Date: Thu, 5 Sep 2002 13:10:40 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <54575a52.0209012124.61f54385@posting.google.com> <54575a52.0209050256.2ad0d9b3@posting.google.com> If it works without TFFS, then that's it, you dont need it ! Your flash disk has an internal controller that does all the write leveling for you etc., and there is no way for you to circumvent the internal controller with any software, in other words, if it works without TFFS, as a regular IDE device, TFFS will not work with it. "pradeep" wrote in message news:54575a52.0209050256.2ad0d9b3@posting.google.com... > I am using flash IFM 441 > can u please tell me when to include TFFS when when not, is it nedded > to have TFFS if i am using flash disk to load the code. > IS TFFS will work with IFM 441 or not > I am able to boot from IFM 441(Without TFFS ) but as frequently i > have to do write operations will u suggest me to include TFFS?? > > regards > pradeep > > > pradeeppatel77@yahoo.com (pradeep) wrote in message news:<54575a52.0209012124.61f54385@posting.google.com>... > > I amusinf flas IFM 350 > > can u please tell me when to include TFFS when when not, is it nedded > > to have TFFS if i am using flash disk to load the code > > > > regards > > pradeep --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Athlon XP Date: Thu, 5 Sep 2002 13:17:42 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: Reply-To: "Michael Lawnick" Sender: mlawnick@pd9e656a9.dip.t-dialin.net Hi Mic, as a first step check whether you have set INCLUDE_PC_CONSOLE. If not, your bootloader will use COM1 for boot menu. HTH Michael "Mic" schrieb im Newsbeitrag news:c6b19a0b.0209050237.5d6b6f2@posting.google.com... > Hi VxWorkers, > does BSP-Pentium/Pro work on a Athlon XP? I made a boot disk, but the > blue boot screen does not appear and everything stops. I guess it > should be many many problems, but not so early! Have you any > suggestions(BIOS problem, video card or what else. Everything happens > before the system tries to load VxWorks image so I think the problem > is the target pc...problably it does non depends on Athlon...)? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: FLASH lockup on PPC405GP eval board? Date: Thu, 5 Sep 2002 13:23:05 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <4brd9.291$AO.1843367@newssvr16.news.prodigy.com> <3D765349.18493F90@hotmail.com> <3D7722D6.4C46B446@hotmail.com> Reply-To: "Michael Lawnick" Sender: mlawnick@pd9e656a9.dip.t-dialin.net > BTW, the OP and I were referring to this board... uups, sorry did not know they sell an own one. ll*( Regards, Michael "Varndell Engineering" schrieb im Newsbeitrag news:3D7722D6.4C46B446@hotmail.com... > Michael Lawnick wrote: > > > > Can't agree. > > Our flashes run perfectly. > > > > BTW: PPC405GP evaluation board is from IBM, not WRS. > > Flash on the Embedded Planet 405GP board works perfectly, as does > the flash on several other 405GP-based boards I've worked with. > Flash on the Wind River 405GP reference design has problems. > > BTW, the OP and I were referring to this board... > http://www.windriver.com/products/html/sbc405.htm > > Regards, > George Varndell > -- > Varndell Engineering, LLC > Embedded Systems Expertise > http://www.varndellengineering.com > > > -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- > http://www.newsfeed.com The #1 Newsgroup Service in the World! > -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Help: disk cache error Date: 5 Sep 2002 04:51:52 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0209050351.67a55a78@posting.google.com> References: ritalin@comcast.net (Rita) wrote in message news:... > > The errno 0xc0006 means S_ioLib_DISK_NOT_PRESENT. > > Does the error message mean that tDcacheUpd has removed the disk while > > writing data or does it mean it mearly timed out while writing to the > > disk (or something similar). I don't know which file system you're using... but I'd recomend not to use TFFS driver with Dcache. Instead initialize TFFS and the File System (e.g. FAT) manually w/o Dcache (although this might be a performance hit). toni --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Rescheduling Problem Date: 5 Sep 2002 05:00:41 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0209050400.3f928765@posting.google.com> References: <3d76257a$0$23626$4d4ef98e@read.news.ch.uu.net> Besides taskDelay(0) you could also enable the time sliceing in the scheduler (by calling kernelTimeSlice). However you would loose the more direct control on the scheduling process... so generally solving something like this manually might be the prefered solution. toni --------------------------- Newsgroups: comp.os.vxworks Subject: Receiving from sockets Date: 5 Sep 2002 06:31:06 -0700 From: ravi368@yahoo.com (Ravi Uday) Organization: http://groups.google.com/ Message-ID: Hi, I have a little problems in moving character pointers. The input ' buf' in the below function is malloced for 2048 bytes before hand. This input variable will finally contain all the bytes read from the socket. This function works just fine if the final data to be retrieved is below 1024 bytes ( in which case ' len ' is inited to 1024 ) and anything over this will result in a memory failure ( Address load Exception ). I would appreciate if anyone can help me find a fault in this function ( particularly while loop ! ). Is there anyother way to do the same operation. void getBuffer ( char *buf, int *inBodySize ) { char *resPtr = NULL; /* Temporary pointer to collect intermediate responses */ int len = 2048, i = 0; resPtr = buf; *inBodySize = 0; while ( *inBodySize <= len ) { i = recv (sockDesc, resPtr, 400, 0); resPtr = resPtr + i; *inBodySize += i; } printf ("\nTotal buffer collected - %s\n", buf ); printf ("\nTotal size of buffer - %d\n", *inBodySize ); } Prototype of ' recv ': int recv ( int s, /* socket to receive data from */ char * buf, /* buffer to write data to */ int bufLen, /* length of buffer */ int flags /* flags to underlying protocols */ ); This routine receives data from a connection-based (stream) socket. and RETURNS The number of bytes actually received, or ERROR if the call fails. The function ' recv ' returns the data currently available in the socket. If ( say ) we are retrieving a stream of 1000 bytes, then the function will return 1000 bytes in two calls ( 1st time may be 600 then 400 ) so actually we dont know how much data is actually read using this function. Since i know the size of the total data to be retrieved i use a while loop in here. As and when i have higher data to receive i get this memory error ( Address load Exception ).. Note: In my case i am reading for 400 bytes at a time ( assumption ) All kind of help is welcome. - - Ravi --------------------------- Newsgroups: comp.os.vxworks Subject: Re: gets() skips first char Date: 5 Sep 2002 06:34:00 -0700 From: Rubio@telephonics.com (Carlton Rubio) Organization: http://groups.google.com/ Message-ID: References: Hi, Had a similar problem, then looked at where the i/o was actually being directed. Once we figured out that we weren't directing our i/o properly we used the vxworks functions "ioGlobalStdGet" and "ioGlobalStdSet" to correct the problem. You just have to know the devices you wish to communicate through. The following is a code snippet that may help you. void Op (char *s) { int c, saved_desc; /* get current io descriptor */ saved_desc = ioGlobalStdGet(0); printf("\n\r"); printf("%s\n \n >>Hit a key to continue<<\n\r",s); /* get file descriptor for device we want to read from */ /* un-comment Port you want to use */ /* Open Com Port for input */ /*input_descriptor = open("/tyCo/0", 0, 0); */ /* Open Enet Port for input */ input_descriptor = open("/pty/telnet.S/0",0, 0); /*assign input to that descriptor*/ ioGlobalStdSet(0, input_descriptor); c = getchar(); close(input_descriptor); ioGlobalStdSet(0, saved_desc); printf("c = %x\n",c); } Note that there are also task specific "io...Get" and "io...Set" functions that are also available (look through your manuals). Based on your output it looks like the Shell task is getting the input that you wanted directed toward your menu-ing function. I believe the shell uses /vio as the device name (I could be wrong here). Anyway I hope this helps. Carlton --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Exception number 14? Date: Thu, 5 Sep 2002 09:36:01 -0400 From: Keith Arner Organization: Marconi Message-ID: References: <3D771F60.69368270@ops.de> Reply-To: Keith Arner On Thu, 5 Sep 2002, Paul Augart wrote: > Exception number 14: Task: 0x78938cc (cePPWRK) > > Page Fault > Program Counter: 0x0120194c > Status Register: 0x00010246 > Error Code : 0x00000002 > 11f9886 _SendPagedensityInfo__FUsUlUlUli14teCeppDevUnits+133e: 11f7e18 > (7892898, 5, 2) > 11f80e1 _initTraceMask +320d: 1201878 (&_Pag_Inf, 5, 7892798) > " > > I know Exception number 14 is a Page Fault. > But what is Error Code : 0x00000002. Where do I find it. This is really an Pentium question, rather than a VxWorks question, so you might want to find a different newsgroup to use, if you have more such questions, but... In my copy of _Pentium Processor Family Developer's Manual_, in section 14.9.14, it shows the format for the error code of a page fault. Your code in particular means "write access to a not-present page while in supervisor mode". You probably tried to write through a NULL pointer or some such. Keith - -- "Once something has gone horribly wrong..." "...it just gets horribler and horribler." "Then, it's a matter of finding the original horriblism." -- from a conversation on debugging problems in a linked list --------------------------- Newsgroups: comp.os.vxworks Subject: Diab and visual slickEdit problem Date: Thu, 5 Sep 2002 16:42:30 +0200 From: "Ofer Goren" Organization: Verio Message-ID: - ------------------------ Win2k Sp3 Tornado 2.02 for PPC Diab 4.4b Envoy 9.2 - ------------------------ Hi. We have a very strange problem: In Tornado2, I can set an external editor, to be used from "file menu", "project" and "build output window". So, if I double click a compilation warning or error, my external editor will open the faulty file in the exact location (line number). Well, we're using visual slickedit as our external editor. When we use the GNU toolkit as our toolchain, everything works fine. However, when we use diab to compile/link our code, when double click a compilation warning/error, the default Tornado editor is opened, and not visual slickedit. What can be the cause of that, and how can it be fixed. Regardless to say, this is a common problem on all our PCs, not just one. - -- Ofer Goren Iamba Technologies --------------------------- Newsgroups: comp.os.vxworks Subject: Anyone have WRS GCC sources? Date: 05 Sep 2002 13:15:13 -0400 From: "Paul D. Smith" Organization: Nortel Networks, Inc. / Billerica, MA Message-ID: Reply-To: pausmith@nortelnetworks.com I'm trying to get the GCC (2.96) and binutils (2.9.5) sources for the VxWorks 5.4.2 tool suite. I'm working on getting the CD from WRS, but it's not exactly going quickly. I don't understand why they don't just put the tarballs up for FTP! So, I wonder if anyone here has these sources lying around somewhere they could make available for download so I could get them directly... Anyone? Thanks... - -- - ------------------------------------------------------------------------------- Paul D. Smith HASMAT--HA Software Mthds & Tools "Please remain calm...I may be mad, but I am a professional." --Mad Scientist - ------------------------------------------------------------------------------- These are my opinions---Nortel Networks takes no responsibility for them. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Help: disk cache error Date: Thu, 5 Sep 2002 19:51:37 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <8400a733.0209050351.67a55a78@posting.google.com> Toni, this is an interesting thought, I am curious as to the reason for this recommendation. Meanwhile, since the Flash disk is not mechanical, it has almost zero access time (no seek time, no rotational latency), there is really little to gain from Dcache there, This holds for TFFS or IDE Flash drives (e.g. CompactFlash), or anything else which has no moving parts. On the otehr hand I do not see how Dcache will make your flash loose performance ... unless the "disk removal detection" is causing you trouble ? "Toni Kurzberger" wrote in message news:8400a733.0209050351.67a55a78@posting.google.com... > ritalin@comcast.net (Rita) wrote in message news:... > > > The errno 0xc0006 means S_ioLib_DISK_NOT_PRESENT. > > > Does the error message mean that tDcacheUpd has removed the disk while > > > writing data or does it mean it mearly timed out while writing to the > > > disk (or something similar). > I don't know which file system you're using... but I'd recomend not to > use TFFS driver with Dcache. Instead initialize TFFS and the File > System (e.g. FAT) manually w/o Dcache (although this might be a > performance hit). > > toni --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Athlon XP Date: 5 Sep 2002 11:25:52 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0209051025.5a4da405@posting.google.com> References: egoisone@libero.it (Mic) wrote in message news:... > Hi VxWorkers, > does BSP-Pentium/Pro work on a Athlon XP? I made a boot disk, but the > blue boot screen does not appear and everything stops. I guess it > should be many many problems, but not so early! Have you any > suggestions(BIOS problem, video card or what else. Everything happens > before the system tries to load VxWorks image so I think the problem > is the target pc...problably it does non depends on Athlon...)? Depending on the version you're using the bootcode (bootsector of your disk, created with vxsys) should print out a number of "+" or ".". If not, the bootsector/bootcode/vxsys is already failing (to locate your bootrom.sys). If it does so and stops after printing out a number of "+" or "." the mini vxworks (bootloader) crashed while initializing... or does not output to pc console. is the bootsector working? toni --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Exception number 14? Date: 5 Sep 2002 11:28:42 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0209051028.3d9f99c9@posting.google.com> References: <3D771F60.69368270@ops.de> > Exception number 14: Task: 0x78938cc (cePPWRK) > Page Fault > Program Counter: 0x0120194c > Status Register: 0x00010246 > Error Code : 0x00000002 > 11f9886 _SendPagedensityInfo__FUsUlUlUli14teCeppDevUnits+133e: 11f7e18 > (7892898, 5, 2) > 11f80e1 _initTraceMask +320d: 1201878 (&_Pag_Inf, 5, 7892798) Are you on x86? > I know Exception number 14 is a Page Fault. > But what is Error Code : 0x00000002. Where do I find it. This has nothing todo with errno, this is the error code from the exception stack frame. Take the processors users manual to look this up for your architecture. This has nothing todo with the posix error ENOENT. > What I would like to know is where to look for Exception number > information, so I can help myself in the future > Thank you in advance. This is information you'll usually find in the processor Documentation, e.g. in MPC8260 PowerQUICC II user's manual (yes... for the 8260 ;-) --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Anyone have WRS GCC sources? Date: Thu, 05 Sep 2002 13:28:36 -0500 From: Andrew Johnson Organization: APS, Argonne National Laboratory Message-ID: References: Paul D. Smith wrote: > I'm trying to get the GCC (2.96) and binutils (2.9.5) sources for the > VxWorks 5.4.2 tool suite. I'm working on getting the CD from WRS, > but it's not exactly going quickly. > > I don't understand why they don't just put the tarballs up for FTP! > > So, I wonder if anyone here has these sources lying around somewhere > they could make available for download so I could get them > directly... Anyone? Try ftp://snsftp.sns.ornl.gov/UPLOAD/controls/CROSS/ - I'm not sure it their .tgz file includes binutils as well, although I suspect it does. Note that although it says gcc-2.7.2, it's actually the compiler for Tornado 2.0.2 (which is not gcc 2.96 BTW, that version of gcc only comes with Tornado 2.1, 2.2 or 3.0). - - Andrew - -- Edsger W. Dijkstra, 1930-2002: "The use of COBOL cripples the mind; its teaching should, therefore, be regarded as a criminal offence." --------------------------- Newsgroups: comp.os.vxworks Subject: 855 MII and 860 BSP question Date: Thu, 05 Sep 2002 19:40:25 GMT From: "Jim" Organization: Magma Communications Ltd. Message-ID: Hello, I need to write an MII driver to replace the SCC driver for ethernet for a custom board based on MPC855 architecture ... what work is involved in vxworks to integrate the MII driver ? Will I need the code for IPSTACK ... or can I add it as easily as adding a second uart port ... Jim --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Anyone have WRS GCC sources? Date: 05 Sep 2002 15:49:18 -0400 From: "Paul D. Smith" Organization: Nortel Networks, Inc. / Billerica, MA Message-ID: References: Reply-To: pausmith@nortelnetworks.com %% Andrew Johnson writes: aj> Paul D. Smith wrote: >> I'm trying to get the GCC (2.96) and binutils (2.9.5) sources for the >> VxWorks 5.4.2 tool suite. I'm working on getting the CD from WRS, >> but it's not exactly going quickly. >> I don't understand why they don't just put the tarballs up for FTP! >> So, I wonder if anyone here has these sources lying around somewhere >> they could make available for download so I could get them >> directly... Anyone? aj> Try ftp://snsftp.sns.ornl.gov/UPLOAD/controls/CROSS/ - I'm not sure it aj> their .tgz file includes binutils as well, although I suspect it aj> does. Note that although it says gcc-2.7.2, it's actually the compiler aj> for Tornado 2.0.2 (which is not gcc 2.96 BTW, that version of gcc only aj> comes with Tornado 2.1, 2.2 or 3.0). I dunno from Tornado, I only do VxWorks itself :). What versions of VxWorks correspond to Tornado 2.0.2, 2.1, 2.2, and 3.0? Our version says VxWorks 5.4.2, and comes with gcc 2.96 (according to ccmips --version) and binutils 2.9.5 (according to ldmips --version). I think we have an enhanced version of 5.4.2: we are using it to build for the new Broadcom BSP. Apparently this was unified in VxWorks 5.5 but we don't have a copy of that yet. Does anyone have the gcc 2.96 compiler sources? Thanks! - -- - ------------------------------------------------------------------------------- Paul D. Smith HASMAT--HA Software Mthds & Tools "Please remain calm...I may be mad, but I am a professional." --Mad Scientist - ------------------------------------------------------------------------------- These are my opinions---Nortel Networks takes no responsibility for them. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Receiving from sockets Date: Thu, 05 Sep 2002 16:34:52 -0400 From: Neil Horman Organization: WEBUSENET.com Message-ID: <3D77BFEC.1080905@lvl7.com> References: Reply-To: nhorman@lvl7.com It looks offhand like a tcp receive buffer size issue. look in the vxworks reference under setsockopt, and then look for the discussion of the SO_SNDBUF and SO_RCVBUF options. Ususally they are set pretty large, but in your case it looks like they are reduced to be around 1024 bytes, which is where you are seeing your problem. Also, I would instrument your code to check the return code of recv for -1 instead of just adding it to your overall length. If its -1 then dump out errno to see what the specific error is. That should give you more of an idea whats going wrong Neil Ravi Uday wrote: > Hi, > > I have a little problems in moving character pointers. > The input ' buf' in the below function is malloced for > 2048 bytes before hand. This input variable will finally > contain all the bytes read from the socket. > This function works just fine if the final data to be > retrieved is below 1024 bytes ( in which case ' len ' is inited to > 1024 ) and anything over this will result in a memory > failure ( Address load Exception ). > > I would appreciate if anyone can help me find a fault in this function > ( particularly while loop ! ). Is there anyother way to do the same operation. > > > void getBuffer ( char *buf, int *inBodySize ) > { > > char *resPtr = NULL; /* Temporary pointer to collect intermediate > responses */ > int len = 2048, i = 0; > > resPtr = buf; > *inBodySize = 0; > > while ( *inBodySize <= len ) { > > i = recv (sockDesc, resPtr, 400, 0); > resPtr = resPtr + i; > *inBodySize += i; > } > > printf ("\nTotal buffer collected - %s\n", buf ); > printf ("\nTotal size of buffer - %d\n", *inBodySize ); > } > > Prototype of ' recv ': > int recv > ( > int s, /* socket to receive data from */ > char * buf, /* buffer to write data to */ > int bufLen, /* length of buffer */ > int flags /* flags to underlying protocols */ > ); > > This routine receives data from a connection-based (stream) socket. > and RETURNS > The number of bytes actually received, or ERROR if the call fails. > > > The function ' recv ' returns the data currently available in the socket. > If ( say ) we are retrieving a stream of 1000 bytes, then the function > will return 1000 bytes in two calls ( 1st time may be 600 then 400 ) > so actually we dont know how much data is actually read using this function. > Since i know the size of the total data to be retrieved i use a while loop in here. > As and when i have higher data to receive i get this memory error > ( Address load Exception ).. > > Note: In my case i am reading for 400 bytes at a time ( assumption ) > > All kind of help is welcome. > > - Ravi > --------------------------- Newsgroups: comp.os.vxworks Subject: Low level hardware tests Date: Thu, 5 Sep 2002 20:38:56 +0000 (UTC) From: "John Ash" Organization: BT Openworld Message-ID: We are using VxWorks for the first time - having used inhouse 'kernels' in the past. As experienced (although not in VxWorks !) emebedded developers, we like to think we know what we want to do and when.... ....However, when it comes to porting low level boot up functionality (such as ROM checksum, RAM check and other hardware checks) there seems to be no 'obvious' place to put it and asking WRS for any off the shelf 'standard' components draws little more than a blank, but plenty of offers to sell us more stuff like the BSP developer's kit, Visionware boot loader authoring and consultancy. Doesn't anyone else incorporate low level checks in their systems *below* the kernel/application level - why do I feel as if what we're trying to do is so unusual - have I missed the point ? !!?? J --------------------------- Newsgroups: comp.os.vxworks Subject: Re: vmBaseGlobalMapInit takes toooo long! Date: 5 Sep 2002 14:38:38 -0700 From: jchen@terayon.com (john) Organization: http://groups.google.com/ Message-ID: References: <488e459a.0209032247.31f32382@posting.google.com> Thanks John We decided to use approach #3, which is the BAT register approach. Question... Why do I need to remove the memory area from the system memory pool? From the diagrams, it looks like it will search the BAT registers first when searching for a particular EA. So, if its a EA hit, it doesn't do an address translation - so it shouldn't matter whether vxWorks is managing that area or not, right? If I remove some areas of RAM from sysPhysMemDesc, then when i try to address those locations I would get an vxworks exception. I guess, what I am after, is that if I define USER_RESERVED_MEM to be some number, how do I use that region of memory? john john_94501@yahoo.com (John) wrote in message news:<488e459a.0209032247.31f32382@posting.google.com>... > Hello, > > Since the 8260 is really just a 603 inside, there is no VxVMI product > for it (the 60x cores have a strange MMU that doesn't work well with > the VxVMI model). > > You probably have three options: > > 1) VxWorks AE which contains all the VM code to do this kind of thing, > though changing the mappings for 20MB of memory is no small task so > this might not be any quicker than your current solution. > > 2) Set up the effective address of your data in its own segment > (256MB) of effective address space, and set up all the mappings for > both sets in the hash table using different virtual addresses (read > chapter 7 of the PPC programming environments, or green book if you > don't know about PPC virtual and effective addresses). If you did > this, swapping between sets would be simply changing one register > value (much, much quicker!). > > 3) Use a BAT register to create the mapping (again using a separate > block of the address space), and swap the BAT register value to swap > between the two sets. > > If you use (2) or (3) make sure that you remove the physical memory > used for both sets of data from the system memory pool (e.g. reduce > LOCAL_MEM_SIZE by an appropriate amount). > > HTH, > > John... > > jchen@terayon.com (john) wrote in message news:... > > Dear vxworks developers. > > > > I've read most of the posts before about how vmBaseGlobalMapInit takes > > a long time to map/remap memory. I have pointers and global vars that > > at some point in my program I would like them to have different > > physical addresses in memory. so in my code I redefined > > sysPhysMemDesc and called vmBaseGlobalMapInit to initialize it. > > everything works fine except it takes way too long. > > > > Does anyone have vxVMI option for their PPC 8260 BSP? can you use > > vmMap to just remap/add a region of virtual-physical translation > > (instead of the whole thing)? If you have it, would anyone mind giving > > me an idea if its quick (in order of ms for ~20MB)? we are using a > > ppc 8260 > > > > We'll buy the VxVMI option if this will work. just curious..... > > > > thanks in advance! > > john --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Anyone have WRS GCC sources? Date: Thu, 5 Sep 2002 21:48:12 +0000 (UTC) From: markm@xmission.xmission.com (Mark A. Matthews) Organization: xmission Message-ID: References: In article , Paul D. Smith wrote: >Does anyone have the gcc 2.96 compiler sources? Sounds like WRS made the same mistake as Red Hat in shipping a gcc 2.96: You'll have to get the sources from WRS. 2.96 was never an official release, name mangling is incompatible with 2.95 or 3.0, and the only way to be sure you have the same source snapshot as WRS is to get a copy from them. - -- - -Mark --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Low level hardware tests Date: Thu, 05 Sep 2002 23:36:58 GMT From: "Jim" Organization: Magma Communications Ltd. Message-ID: References: Here what I tried: romStart.c ---> before decompressing OS into RAM ... run RAM(S) tests, basic flash check ... store the results in USER_RESERVED_MEM (vxworks macro to reserve chunk of mem at end of ram ... so its not used by OS) . Boot up with jumper ... instead of starting application we test our ASICs in loop back etc ... and store the results in RAM for inspection ... Where you place your test code will really depend on what "way u start your application" from ROM, from compressed ROM, run in ROM, run in flash etc etc ... Good luck, Jim "John Ash" wrote in message news:al8fd0$8ct$1@helle.btinternet.com... > We are using VxWorks for the first time - having used inhouse 'kernels' in > the past. As experienced (although not in VxWorks !) emebedded developers, > we like to think we know what we want to do and when.... > > ....However, when it comes to porting low level boot up functionality (such > as ROM checksum, RAM check and other hardware checks) there seems to be no > 'obvious' place to put it and asking WRS for any off the shelf 'standard' > components draws little more than a blank, but plenty of offers to sell us > more stuff like the BSP developer's kit, Visionware boot loader authoring > and consultancy. > > Doesn't anyone else incorporate low level checks in their systems *below* > the kernel/application level - why do I feel as if what we're trying to do > is so unusual - have I missed the point ? > > !!?? > > J > > --------------------------- Newsgroups: comp.os.vxworks Subject: Mulitiple END driver's question. (cannot add 3 instances) Date: 5 Sep 2002 16:46:52 -0700 From: roswel_ajf@hotmail.com (roswell) Organization: http://groups.google.com/ Message-ID: <6dcae13d.0209051546.6262104d@posting.google.com> Hi All: I am having trouble adding 3 END drivers (for 3 device instances). 1. With muxShow() i can see all three instances of my device. (so muxDevLoad(), muxDevStart() and ipAttache are OK). 2. With usrNetIfConfig() {ipMaskSet(), ifAddrSet(), hostAdd() etc) works only for the instance 1 and 2. It fails for the 3rd instance. i get following errors for 3rd device (nm2). nm2 ifMaskSet: S_errno_ENXIO nm2 ifAddrSet: S_errno_ENXIO 3. ifShow() is correct and clean only for 1st device. for 2nd device, all FFFs for counters, and for 3rd device, even MTU is wrong, which set to 10. shoudl be 1500. well, clearly, with 2nd, 3rd addition has some issues..Can anyone please give me a clue. Tornado version 2.0.2 I increased the IP_MAX_UNITS to 5. what i am missing. Many thanks --------------------------- Newsgroups: comp.os.vxworks Subject: Re: "Virtual" end driver Date: 5 Sep 2002 17:43:34 -0700 From: roswel_ajf@hotmail.com (roswell) Organization: http://groups.google.com/ Message-ID: <6dcae13d.0209051643.4b62d475@posting.google.com> References: Hi , I have the same issue, i am trying to add 2 vitural interfaces. Its ok for me have them as PPP (no need for ethernet like). so ifShow() only gives correct settings for vif 1. and of course userNetIfConfig() failed for vif-3. when i do routeShow() i see the first two virtual interfaces. My problem is similar to Gunter's. Any hints?... Gunter.Ritter@tenovis.com (Gunter Ritter) wrote in message news:... > Hello VxWorkers, > > I have written a "virtual" end driver as a secondary end driver for > our system. > "Virtual" means that this end driver has no "physical" ethernet > hardware. > > The primary end driver is working with ethernet hardware. > > I need this secondary end driver only to exchange ip packets with the > TCP/IP stack of vxworks and let vxworks do the routing between the two > networks. > > I have a problem when I try to assign an ip address. > muxDevLoad(), muxBind(), ipAttach() and ifMaskSet() work fine. > ifShow() shows me the settings of the secondary driver. > But calling ifAddrSet() with an unique ip address causes the network > functionality of vxworks to crash. > > Even "ping" does no more work with the primary end driver after > calling ifAddrSet(). > > Does anybody can give me a hint whats going wrong with my secondary > end driver? > > Thanks. --------------------------- Newsgroups: comp.os.vxworks Subject: ZNYX 346Q on VxWorks 5.4 Date: 5 Sep 2002 21:13:02 -0700 From: andreas_lindell75@hotmail.com (Rikard) Organization: http://groups.google.com/ Message-ID: <5947a04.0209052013.6da7b8fd@posting.google.com> Hello, I am trying to intergrate a ZNYX 346Q network card with my vxworks image. I can build it from command line like they descibe in the manual. Now I want to move my development into the workspace and I have contacted ZNYX and they say that vxworks is a "learning experience". Has anyone intergrated ZNYX 346Q so that it can be compiled from workspace? The reason I want it in workspace is that I want to be able to call: zxeUp("znb0", "192.168.0.241", 0xffffff00, 0, 0); zxeUp("znb1", "192.168.0.240", 0xffffff00, 0, 1); from UsrAppInit plus that all my other files are in the project facilities. So if anyone have worked with ZNYX 346Q in workspace then maybe you would be kind enough to give me some pointers. Regards Andreas --------------------------- Newsgroups: comp.os.vxworks Subject: IS THE MACRO REPLACED WHEN COMPILING? Date: Fri, 6 Sep 2002 13:30:52 +0800 From: "carria" Organization: power Message-ID: Reply-To: "carria" I defined some extra macros in bspname.h, and to compile other files(.c),does the macros such as : #define PLD_REG(off) (something+(off)) be replaced when compiling? and after compiled the files(.c) to the files(.o), for I want to make the source code invisible to user, but with other source code visible, shall I put the extra macros in the bspname.h that delivered to user? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Receiving from sockets Date: 5 Sep 2002 22:35:41 -0700 From: ravi368@yahoo.com (Ravi Uday) Organization: http://groups.google.com/ Message-ID: References: <3D77BFEC.1080905@lvl7.com> Neil Horman wrote in message news:<3D77BFEC.1080905@lvl7.com>... > It looks offhand like a tcp receive buffer size issue. look in the vxworks > reference under setsockopt, and then look for the discussion of the SO_SNDBUF > and SO_RCVBUF options. Ususally they are set pretty large, but in your case it > looks like they are reduced to be around 1024 bytes, which is where you are > seeing your problem. Also, I would instrument your code to check the return > code of recv for -1 instead of just adding it to your overall length. If its -1 > then dump out errno to see what the specific error is. That should give you > more of an idea whats going wrong > Neil > Neil : Thanks for the suggestion. Well i shall look thr those socket options for improving my tcp connection behaviour. Furrther .. The 3rd parameter to 'recv' specifies number of bytes to read..but..say i specify 1000 then not necessarily it reads all 1000 bytes. We get the indication of the actual bytes read in ' i ' the return of 'recv' function. Hence i had set a while loop like that. I see something fishy in the way i move the ' resptr' variable. Do i need it to bzero it..( reset its memory contents ) everytime before calling ' recv '. Its continuing to give memory failures when the size of data on the other side is over 1024 bytes !! A printf inside while loop for actual bytes read ' i ' for a 'len' of '1872' ( i.e. len = 1872 ) was like this. [ ' buf ' variable is malloced for 2048 bytes and len = 2048) 1st iteration - i = 400 2nd iteration - i = 100 3rd iteration - i = 62 4th iteration - i = 400 Then - - MEMORY FAILURE - ( Address load Exception ) while ( *inBodySize <= len ) { i = recv ( SockDesc, resPtr, 400, 0); resPtr = resPtr + i; *inBodySize += i; printf ("\n~~~~~~~~...SizeACCUMULATED *inBodySize - %d \n", *inBodySize); } Any improvements in here Neil that you can think of.. - - Ravi --------------------------- Newsgroups: comp.os.vxworks Subject: Re: vmBaseGlobalMapInit takes toooo long! Date: 5 Sep 2002 22:47:53 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0209052147.17e5062c@posting.google.com> References: <488e459a.0209032247.31f32382@posting.google.com> Hello, > We decided to use approach #3, which is the BAT register approach. > > Question... Why do I need to remove the memory area from the system > memory pool? From the diagrams, it looks like it will search the BAT > registers first when searching for a particular EA. So, if its a EA > hit, it doesn't do an address translation - so it shouldn't matter > whether vxWorks is managing that area or not, right? But you don't want a malloc() from some piece of code in your system to allocate memory from the "reserved" portion of the physical memory, so you need to make sure that it is not in the heap partition. > If I remove some areas of RAM from sysPhysMemDesc, then when i try to > address those locations I would get an vxworks exception. Not if the BAT is set up to map the address somewhere. The BAT will match first, so the exception will not occur, even though the address is not in sysPhysMemDesc. > I guess, what I am after, is that if I define USER_RESERVED_MEM to be > some number, how do I use that region of memory? USER_RESERVED_MEM should be accessible without any problems - the mappings should still be made for that memory if the BSP is working correctly. If you do this, then you can access the memory before you set up any BAT mapping for it. Just use the address as normal. All the USER_RESERVED_MEM does is prevent the partition from getting a portion of the RAM. You can use either USER_RESERVED_MEM or LOCAL_MEM_SIZE to "hide" the portion of the physical memory that you are going to map with the BAT register(s) from VxWorks. Changing LOCAL_MEM_SIZE will stop the system even creating the mappings in the normal way. Depending on how much you plan to map using the BAT, this might save you some time, and maybe a little memory too if the system can shrink the hash table at all (the size of which depends on the amount of physical memory being mapped). HTH, John... > john > > john_94501@yahoo.com (John) wrote in message news:<488e459a.0209032247.31f32382@posting.google.com>... > > Hello, > > > > Since the 8260 is really just a 603 inside, there is no VxVMI product > > for it (the 60x cores have a strange MMU that doesn't work well with > > the VxVMI model). > > > > You probably have three options: > > > > 1) VxWorks AE which contains all the VM code to do this kind of thing, > > though changing the mappings for 20MB of memory is no small task so > > this might not be any quicker than your current solution. > > > > 2) Set up the effective address of your data in its own segment > > (256MB) of effective address space, and set up all the mappings for > > both sets in the hash table using different virtual addresses (read > > chapter 7 of the PPC programming environments, or green book if you > > don't know about PPC virtual and effective addresses). If you did > > this, swapping between sets would be simply changing one register > > value (much, much quicker!). > > > > 3) Use a BAT register to create the mapping (again using a separate > > block of the address space), and swap the BAT register value to swap > > between the two sets. > > > > If you use (2) or (3) make sure that you remove the physical memory > > used for both sets of data from the system memory pool (e.g. reduce > > LOCAL_MEM_SIZE by an appropriate amount). > > > > HTH, > > > > John... > > > > jchen@terayon.com (john) wrote in message news:... > > > Dear vxworks developers. > > > > > > I've read most of the posts before about how vmBaseGlobalMapInit takes > > > a long time to map/remap memory. I have pointers and global vars that > > > at some point in my program I would like them to have different > > > physical addresses in memory. so in my code I redefined > > > sysPhysMemDesc and called vmBaseGlobalMapInit to initialize it. > > > everything works fine except it takes way too long. > > > > > > Does anyone have vxVMI option for their PPC 8260 BSP? can you use > > > vmMap to just remap/add a region of virtual-physical translation > > > (instead of the whole thing)? If you have it, would anyone mind giving > > > me an idea if its quick (in order of ms for ~20MB)? we are using a > > > ppc 8260 > > > > > > We'll buy the VxVMI option if this will work. just curious..... > > > > > > thanks in advance! > > > john --------------------------- Newsgroups: comp.os.vxworks Subject: Re: IS THE MACRO REPLACED WHEN COMPILING? Date: Fri, 06 Sep 2002 05:48:29 GMT From: nunya@damn.business.com (Bruce) Organization: Prodigy Internet http://www.prodigy.com Message-ID: <3d7841e2.6711480@news.houston.sbcglobal.net> References: In comp.os.vxworks "carria" wrote: >I defined some extra macros in bspname.h, and to compile other >files(.c),does the macros such as : > #define PLD_REG(off) (something+(off)) >be replaced when compiling? >and after compiled the files(.c) to the files(.o), >for I want to make the source code invisible to user, but with other >source code visible, shall I put the extra macros in the bspname.h that >delivered to user? Macros are handled by the preprocessor. There are commands that will allow you can see the output of the preprocessed source files. As for what you want to do, a couple of things. I hate consultants that try to hide the code, as do most people. Anyway, look for a code obfuscator. That is about the best you can do. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Athlon XP Date: Fri, 06 Sep 2002 07:04:28 GMT From: "Michael B." Organization: [Infostrada] Message-ID: <0sYd9.7193$AW6.163299@twister2.libero.it> References: <8400a733.0209051025.5a4da405@posting.google.com> Hi, my bootsector/bootcode/vxsys works fine printing out as many "." as it does on a Pentium: the difference is what I described: with Pentium no problems, the default countdown begins and so on, with AMD on the contrary, after that point (in the right sense of the word ;-) ) everything stops, the bootloader does not appear and no blue screen! Looking for something wrong... Mic. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Mulitiple END driver's question. (cannot add 3 instances) Date: Fri, 6 Sep 2002 08:09:32 +0000 (UTC) From: Johan Borkhuis Organization: Agere Systems Message-ID: References: <6dcae13d.0209051546.6262104d@posting.google.com> roswel_ajf@hotmail.com (roswell) wrote: > Hi All: > > I am having trouble adding 3 END drivers (for 3 device instances). > > 1. With muxShow() i can see all three instances of my device. (so > muxDevLoad(), muxDevStart() and ipAttache are OK). > > 2. With usrNetIfConfig() {ipMaskSet(), ifAddrSet(), hostAdd() etc) > works only for the instance 1 and 2. It fails for the 3rd instance. > > i get following errors for 3rd device (nm2). > > nm2 ifMaskSet: S_errno_ENXIO > nm2 ifAddrSet: S_errno_ENXIO > > 3. ifShow() is correct and clean only for 1st device. for 2nd device, > all FFFs for counters, and for 3rd device, even MTU is wrong, which > set to 10. shoudl be 1500. > > well, clearly, with 2nd, 3rd addition has some issues..Can anyone > please give me a clue. Tornado version 2.0.2 > > I increased the IP_MAX_UNITS to 5. what i am missing. It looks like your driver cannot do more that 2 devices, and already has problems with the second device, looking at the counters. I would take a close look at the driver-source to see if it allows more than 1 instance. If you do not have the sourcecode ask your FAE for it. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ZNYX 346Q on VxWorks 5.4 Date: Fri, 6 Sep 2002 08:11:34 +0000 (UTC) From: Johan Borkhuis Organization: Agere Systems Message-ID: References: <5947a04.0209052013.6da7b8fd@posting.google.com> andreas_lindell75@hotmail.com (Rikard) wrote: > Hello, > > I am trying to intergrate a ZNYX 346Q network card with my vxworks > image. I can build it from command line like they descibe in the > manual. Now I want to move my development into the workspace and I > have contacted ZNYX and they say that vxworks is a "learning > experience". > > Has anyone intergrated ZNYX 346Q so that it can be compiled from > workspace? The reason I want it in workspace is that I want to be able > to call: > zxeUp("znb0", "192.168.0.241", 0xffffff00, 0, 0); > zxeUp("znb1", "192.168.0.240", 0xffffff00, 0, 1); > from UsrAppInit plus that all my other files are in the project > facilities. So if anyone have worked with ZNYX 346Q in workspace then > maybe you would be kind enough to give me some pointers. You could just add all the sourcefiles that are mentioned in the makefile to your project and recompile, that would be the easiest way. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: problems after installing patch SPR68322 Date: Fri, 06 Sep 2002 08:20:17 GMT From: bachmeier.andreas@foerstergroup.de (Andreas Bachmeier) Organization: Deutsche Telekom AG Message-ID: <3d78647c.11969761@news.dtag.de> Hi, has anybody installed the patch SPR68322? I use Tornado 2.02. Host: W2000, Target x86 Pentium To get Tornado 2.02 I already installed SPR-T2CP4_x86. Because we have still problems with the 557 network driver we installed the patch SPR68322. But now I get the following messages: +++++++++++++++++++++++++++++++ ld386 -X -N -e _sysInit -Ttext 00108000 \ dataSegPad.o partialImage.o -o vxWorks partialImage.o: In function `IntWriteLogEnable': D:\DS\DS_DCMFL\Projekt\default\D:\DS\AWS_Tools\Tools.c:250: undefined reference to `ipMaxUnits' D:\DS\DS_DCMFL\Projekt\default\D:\DS\AWS_Tools\Tools.c:250: undefined reference to `ipMaxUnits' make: *** [vxWorks] Error 0x1 Done. +++++++++++++++++++++++++++++++ The linker can't find the reference `ipMaxUnits' anymore. Is there a bug in the patch? Or do I need more work than the plain install of the patch? Andreas --------------------------- Newsgroups: comp.os.vxworks Subject: Re: gets() skips first char Date: Fri, 6 Sep 2002 10:36:02 +0200 From: "Joachim Bassendowski" Organization: HARZ.DE Internet fuer die Region (XeNET GmbH, Clausthal) Message-ID: References: Hi, thank you very much for your code snippet. I will try to use it. "Carlton Rubio" schrieb im Newsbeitrag news:cb87ce90.0209050534.3129bda3@posting.google.com... > > Had a similar problem, then looked at where the i/o was actually being > directed. Once we figured out that we weren't directing our i/o > properly we used the vxworks functions "ioGlobalStdGet" and > "ioGlobalStdSet" to correct the problem. You just have to know the > devices you wish to communicate through. The following is a code > snippet that may help you. > > [..] --------------------------- Newsgroups: comp.os.vxworks Subject: Re: gets() skips first char Date: Fri, 6 Sep 2002 10:37:28 +0200 From: "Joachim Bassendowski" Organization: HARZ.DE Internet fuer die Region (XeNET GmbH, Clausthal) Message-ID: References: Hi, thank you for your reply. I'm using windows with xterm rlogin to the host (SUN Blade 100). From there I use tip to communicate with VxWorks. "Pete Kockritz" schrieb im Newsbeitrag news:pkockritz-A9574A.21012104092002@netnews.attbi.com... > [..] > Your program is getting some of the input, but not all. > > Try running your program without the shell running. Is this output > that you show coming out on the windShell, target console or the virtual > console? > > -- > Pete Kockritz > --------------------------- Newsgroups: comp.os.vxworks Subject: Multiple network interfaces in simulator Date: 6 Sep 2002 05:34:24 -0700 From: rajendra_krsingh@yahoo.com (rajendra) Organization: http://groups.google.com/ Message-ID: <479ba395.0209060434.3734f77a@posting.google.com> Hi all, Is it possible to have more than one network interface (nt0,nt1...) in simulator. Have anyone tried that? Please let me know the procedure. thanks rajendra --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Multiple network interfaces in simulator Date: Fri, 6 Sep 2002 16:25:31 +0200 From: "Martin Raabe" Organization: 1&1 Internet AG Message-ID: References: <479ba395.0209060434.3734f77a@posting.google.com> Hello rajendra, it is possible and described in the documentaion, but it is not available in older Versions of Tornado or Tornado Prototyper. This is a licensing issue! More Information can be obtained from WindRiver. Ciao - -- - -- Martin Raabe E: Martin.RaabeATNOSPAMBaSystem.de "rajendra" schrieb im Newsbeitrag news:479ba395.0209060434.3734f77a@posting.google.com... > Hi all, > Is it possible to have more than one network interface (nt0,nt1...) in simulator. > Have anyone tried that? Please let me know the procedure. > > thanks > rajendra --------------------------- Newsgroups: comp.os.vxworks Subject: Re: disk cache error Date: Fri, 06 Sep 2002 10:39:56 -0400 From: Mary McHugh Organization: Rancho McHugh Message-ID: <3D78BE3C.2DBC1BAD@notroll.com> References: Reply-To: trollbusters@notroll.com Leonid Rosenboim wrote: > If your device is not removable, you should set the device "removable" flag > to FALSE, thereby disabling these autodetecion mechanisms. > > If it is in fact a removable device, as soon as a new card has been > inserted, any subsequent call to open() will mount the new device and you > will be able to access it. The file descriptors that where opened before the > exchange took place, will be invalid, and need to be closed, to avoid FD > leakage. > > HTH, > - Leonid We are experiencing the same problem with our flash. However, we DO have the removable flag set to false and we still get the same error as the OP. Any other ideas? Additionally, another problem we're having is regading tuning the dcache for the flash to have a 0 sync interval in order to force dcache to write immediately (dcacheDevTune). We start out OK, but after running for "awhile" (not sure what the exact interval is), something changes the sync interval back to 1 second. I know for sure that chkdsk will do it however we only run chkdsk on start/restart before we tune the cache. Problem 3 is if we have a burst of flash activity, for example we unzip a package containing many files, we occasionally crash. Sometimes it is in the task that is executing the unzip, sometimes it crashes in the tDcacheUpd task but it ALWAYS is at the same instruction in flDosFormat. I'm not really sure why we're in flDosFormat -- it sounds ominous. ;-) Here's what we get: - ---------------------- data access Exception current instruction address: 0x004aff6c Machine Status Register: 0x00009032 Data Access Register: 0xeeeeeeee Condition Register: 0x42200040 Data storage interrupt Register: 0x00009032 Task: 0x6a6ecb8 "tDcacheUpd" - -> l 0x004aff6c 0x4aff6c a81e0000 lha r0,0(r30) 0x4aff70 2c800000 cmpi crf1,0,r0,0 0x4aff74 40850078 bc 4,5,0x004affec 0x4aff78 7fe3fb78 or r3,r31,r31 0x4aff7c 83bf0030 lwz r29,48(r31) 0x4aff80 7fc4f378 or r4,r30,r30 0x4aff84 3bbdffff addi r29,r29,-1 0x4aff88 7f7de838 and r29,r27,r29 0x4aff8c 3bbd0001 addi r29,r29,1 0x4aff90 4bfff16d bl 0x004af0fc value = 4915092 = 0x4aff94 = flDosFormat + 0xf48 Any input that anyone can shed on this problem will be appreciated. I'm off to "windsurf" and google for more info. Mary mchug06@_disregard_attglobal.net --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Anyone have WRS GCC sources? Date: 06 Sep 2002 12:23:33 -0400 From: "Paul D. Smith" Organization: Nortel Networks, Inc. / Billerica, MA Message-ID: References: Reply-To: pausmith@nortelnetworks.com %% markm@xmission.xmission.com (Mark A. Matthews) writes: mam> You'll have to get the sources from WRS. 2.96 was never an mam> official release, name mangling is incompatible with 2.95 or 3.0, I'm very familiar with the whole tale of 2.96. This is the least of my worries since WRS goes in and adds their own tweaks to the code anyway (at least they used to). For my purposes the only upgrade path that matters is the one the WRS versions of the tools follow. mam> and the only way to be sure you have the same source snapshot as mam> WRS is to get a copy from them. Yep, that's why I'm posting here, not in the GCC groups; I'm wondering if anyone here has already done that and can make them available to me for download. My efforts at getting the source CD are proceeding, but veeeerrrryy slllloooooowwwwlllllyyy... - -- - ------------------------------------------------------------------------------- Paul D. Smith HASMAT--HA Software Mthds & Tools "Please remain calm...I may be mad, but I am a professional." --Mad Scientist - ------------------------------------------------------------------------------- These are my opinions---Nortel Networks takes no responsibility for them. --------------------------- Newsgroups: comp.os.vxworks Subject: VisionCLICK BUG! Date: Fri, 6 Sep 2002 21:31:59 +0400 From: "Denis Dubrovin" Organization: WEBPlus Ltd. Message-ID: I purchase your VisionCLICK\VisionICE II bundle and have troubles with flash programming with it. I have EST`s MDP823 Referense Design board with "AMD 29F016\17(2048x8) 2 Devices" flash chip as listed in result of "TF DEVICE" command (it have number 16). Using VisionCLICK Flash Programming dialog I picked up appropriate device from device list and saw result "TF CONF 16 ...." in my terminal window. BUT! When I push ANY button in this dialog (ok, cancel, erase and prog, erase or prog - ANY!) I see first command in my terminal window is "TF CONF 6 ....." and when, for example - ''ERASE' button pushed, "TF ERASE" command. And I have error message poped in terminal window because I haven`t chip number 6 (AMD 29F010(128x8) 1 Device - as listed in device list). What I must do? Or may be I do wrong something? My solution is take device number 116 and..... magician!!! I saw TF CONF 16 in my terminal! :) VisionCLICK Flash dialog simply LOST first digit of flash device type! PS : VisionCLICK 7.9a Any patches avaliable? - -- Regards! Den --------------------------- Newsgroups: comp.os.vxworks Subject: resolvGetHostByName Timeout? Date: 6 Sep 2002 10:40:10 -0700 From: dchou4u@hotmail.com (Dennis C) Organization: http://groups.google.com/ Message-ID: <65ee8e47.0209060940.413fdccb@posting.google.com> I am using resolvGetHostByName(), but if the DNS server is down, the timeout from this call is very very long. If I install multiple DNS servers using the resolvParamsSet() call, does resolvGetHostByName() return immediately if any of the DNS servers respond? If anyone of the DNS servers is down, does it cause resolvGetHostByName() to timeout with the long delay? Thanks --------------------------- Newsgroups: comp.os.vxworks Subject: Tornado 2 vxSim font problem Date: Fri, 6 Sep 2002 21:56:01 +0400 From: "Denis Dubrovin" Organization: WEBPlus Ltd. Message-ID: Hi! WinXP host. How to change font size in vxSim simulator`s window? It so tiny and unreadable..... - -- Regards! Den --------------------------- Newsgroups: comp.os.vxworks Subject: Re: FLASH lockup on PPC405GP eval board? Date: Thu, 05 Sep 2002 05:24:38 -0400 From: Varndell Engineering Organization: Posted Via Newsfeeds.com = SPEED+RETENTION+COMPLETION = http://www.newsfeeds.com Message-ID: <3D7722D6.4C46B446@hotmail.com> References: <4brd9.291$AO.1843367@newssvr16.news.prodigy.com> <3D765349.18493F90@hotmail.com> Michael Lawnick wrote: > > Can't agree. > Our flashes run perfectly. > > BTW: PPC405GP evaluation board is from IBM, not WRS. Flash on the Embedded Planet 405GP board works perfectly, as does the flash on several other 405GP-based boards I've worked with. Flash on the Wind River 405GP reference design has problems. BTW, the OP and I were referring to this board... http://www.windriver.com/products/html/sbc405.htm Regards, George Varndell - -- Varndell Engineering, LLC Embedded Systems Expertise http://www.varndellengineering.com - -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- http://www.newsfeed.com The #1 Newsgroup Service in the World! - -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VisionCLICK BUG! Date: Fri, 6 Sep 2002 20:28:51 +0200 From: "Martin Raabe" Organization: 1&1 Internet AG Message-ID: References: Hello Denis, I am sorry, but this is not a WindRiver sponsored UseGroup. To get Support, please contact: support@WindRiver.com Take Care! - -- Martin Raabe E: Martin.RaabeATNOSPAMBaSystem.de "Denis Dubrovin" schrieb im Newsbeitrag news:alaoqg$aih$1@news.wplus.spb.ru... > I purchase your VisionCLICK\VisionICE II bundle and have troubles with flash > programming with it. > > I have EST`s MDP823 Referense Design board with "AMD 29F016\17(2048x8) 2 > Devices" flash chip as listed in result of "TF DEVICE" command (it have > number 16). > > Using VisionCLICK Flash Programming dialog I picked up appropriate device > from device list and saw result "TF CONF 16 ...." in my terminal window. > > BUT! When I push ANY button in this dialog (ok, cancel, erase and prog, > erase or prog - ANY!) I see first command in my terminal window is "TF CONF > 6 ....." and when, for example - ''ERASE' button pushed, "TF ERASE" command. > And I have error message poped in terminal window because I haven`t chip > number 6 (AMD 29F010(128x8) 1 Device - as listed in device list). > > What I must do? Or may be I do wrong something? > > My solution is take device number 116 and..... magician!!! I saw TF CONF 16 > in my terminal! :) > > VisionCLICK Flash dialog simply LOST first digit of flash device type! > > PS : VisionCLICK 7.9a > > Any patches avaliable? > > -- > Regards! > Den > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: disk cache error Date: Fri, 06 Sep 2002 14:37:22 -0400 From: Mary McHugh Organization: Rancho McHugh Message-ID: <3D78F5E2.B1ECA4EB@notroll.com> References: <3D78BE3C.2DBC1BAD@notroll.com> Reply-To: trollbusters@notroll.com Mary McHugh wrote: > Additionally, another problem we're having is regading tuning the dcache > for the flash to have a 0 sync interval in order to force dcache to > write immediately (dcacheDevTune). We start out OK, but after running > for "awhile" (not sure what the exact interval is), something changes > the sync interval back to 1 second. I know for sure that chkdsk will do > it however we only run chkdsk on start/restart before we tune the > cache. Apologies for following up to my own post, but for anyone that is interested I did discover the above problem in the SPR database as SPR#30130. It's fixed with the appropriate patch or in T2.2 FCS. Mary --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Anyone have WRS GCC sources? Date: Fri, 06 Sep 2002 13:32:28 -0500 From: Andrew Johnson Organization: APS, Argonne National Laboratory Message-ID: References: Paul D. Smith wrote: > > I dunno from Tornado, I only do VxWorks itself :). What versions of > VxWorks correspond to Tornado 2.0.2, 2.1, 2.2, and 3.0? Usually: Tornado 2.0 comes with vxWorks 5.4 Tornado 2.0.2 comes with vxWorks 5.4.2 Tornado 2.2 comes with vxWorks 5.5 Tornado 3.0 comes with vxWorks AE > Our version says VxWorks 5.4.2, and comes with gcc 2.96 (according to > ccmips --version) and binutils 2.9.5 (according to ldmips --version). > > I think we have an enhanced version of 5.4.2: we are using it to build > for the new Broadcom BSP. Apparently this was unified in VxWorks 5.5 > but we don't have a copy of that yet. It sounds like you have a special version which doesn't follow the above numbering. According to the vxWorks 5.5 GNU Toolkit Release Notes the version of GCC supplied with T2.2 is called "2.96+" which may be slightly different to 2.96 (you might have a version that was used to build T2.1 or T3.0, which came out before T2.2). I guess you're just going to have to wait for the source from WRS. - - Andrew - -- "Life is what happens while you're busy making other plans." - -- John Lennon --------------------------- Newsgroups: comp.os.vxworks Subject: linker problem Date: Fri, 6 Sep 2002 20:29:25 +0100 From: "Dimitrios Siganos" Organization: Newport Networks Ltd. Message-ID: <1031340570.680577@azores.network-i.net> Hi all, my linker is throwing me a strange message that i do not understand at all. I have a BSP which works fine if I compile at the BSP directory and use it without Tornado. However, when I create Tornado project based on the BSP, and compile it, I get the following linker error. Has anyone seen this before? The message is complaining about a file in drive E: (I don't have drice E: on my computer) and the file that it is listing look like something went horribly wrong... but what? Thank you. Dimitris ldppc -X -N -e _sysInit -Ttext 00300000 \ dataSegPad.o partialImage.o ctdt.o -o vxWorks partialImage.o: In function `taskStackAllot': E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `endObjInit' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `mib2Init' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `endObjFlagSet' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `mib2ErrorAdd' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `endRcvRtnCall' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `mib2ErrorAdd' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `mib2ErrorAdd' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `endTxSemTake' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `mib2ErrorAdd' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `endTxSemGive' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `mib2ErrorAdd' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `mib2ErrorAdd' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `endTxSemGive' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `endFlagsClr' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `endFlagsSet' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `endFlagsGet' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `mib2ErrorAdd' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `mib2ErrorAdd' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `mib2ErrorAdd' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `mib2ErrorAdd' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `mib2ErrorAdd' partialImage.o:E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21p pcAltivec\PPC6 04gnu\E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec \taskLib.c:209 0: more undefined references to `mib2ErrorAdd' follow partialImage.o: In function `taskStackAllot': E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `etherMultiAdd' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `etherMultiDel' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `etherMultiGet' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `endObjectUnload' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `endFlagsGet' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `endMultiLstFirst' E:\vxWorksProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\PPC60 4gnu\E:\vxWork sProj\SPR_fixes\SelectLib_mxFdProblemT202\Tornado21ppcAltivec\taskLib.c:2090 : undefined re ference to `endMultiLstNext' partialImage.o: In function `wdbRtIfInit': C:\dimitris\proj\tornado\misc\Project2\default\C:\Tornado\target\config\comp s\src\usrWdbCo re.c:388: undefined reference to `endEtherAddressForm' C:\dimitris\proj\tornado\misc\Project2\default\C:\Tornado\target\config\comp s\src\usrWdbCo re.c:388: undefined reference to `endEtherPacketDataGet' C:\dimitris\proj\tornado\misc\Project2\default\C:\Tornado\target\config\comp s\src\usrWdbCo re.c:388: undefined reference to `endEtherPacketAddrGet' make: *** [vxWorks] Error 0x1 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Tornado 2 vxSim font problem Date: Fri, 6 Sep 2002 20:11:33 +0000 (UTC) From: Johan Borkhuis Organization: Agere Systems Message-ID: References: "Denis Dubrovin" wrote: > WinXP host. > > How to change font size in vxSim simulator`s window? It so tiny and > unreadable..... See http://www.xs4all.nl/~borkhuis/vxworks/vxw_pt5.html#5.8-D Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: 855 MII and 860 BSP question Date: 6 Sep 2002 13:55:31 -0700 From: langus@iskratel.si (Matevz Langus) Organization: http://groups.google.com/ Message-ID: References: Hi Jim, no, you don't need the code for IP stack. You only need to write your own END (Enhanced Network Driver) or maybe you can find it at Windsurf somewhere??? END drivers are attached to VxWorks IP stack thru MUXm which enables usage of different ethernet drivers (also at the same time). regards, Matevz "Jim" wrote in message news:... > Hello, > I need to write an MII driver to replace the SCC driver for ethernet for a > custom board based on MPC855 architecture ... what work is involved in > vxworks to integrate the MII driver ? Will I need the code for IPSTACK ... > or can I add it as easily as adding a second uart port ... > > Jim --------------------------- Newsgroups: comp.os.vxworks Subject: Re: visionCLICK Simple vs. Complete/Full RAM tests Date: 6 Sep 2002 13:57:49 -0700 From: langus@iskratel.si (Matevz Langus) Organization: http://groups.google.com/ Message-ID: References: vClick RAM test are worth nothing! vClick reported that complete tests are OK but software wouldn't run! regards, Matevz ryeung@earthlink.net (Raymond Yeung) wrote in message news:... > Does anyone know the main differences in test cases the > "Simple" and the "complete/Full" RAM tests are? > > The complete tests are a lot longer than the simple ones, > and I'm wondering what faults I'd be missing if I use the > simple ones instead. > > Thanks, > Raymond --------------------------- Newsgroups: comp.os.vxworks Subject: memory allocation Date: 6 Sep 2002 14:29:57 -0700 From: langus@iskratel.si (Matevz Langus) Organization: http://groups.google.com/ Message-ID: Hi! Does anyone know how it would be possible to allocate memory bellow certain address (32 MB), eventhough system partition is bigger (128 MB)? I need this to solve the 24-bit reallocation value problem. thanks, Matevz Langus --------------------------- Newsgroups: comp.os.vxworks Subject: VxWorks 5.4.2 on Motorola Power PC 82xx Date: 6 Sep 2002 14:50:48 -0700 From: vtriplic@yahoo.com (vt) Organization: http://groups.google.com/ Message-ID: <66f277de.0209061350.50ac78b0@posting.google.com> We are planning to use VxWorks on Motorola PPC 82xx. I searched this group for any user feedbacks, but I couldn't find much. Does anybody use it? Any issues that we need to be aware of up front? Any fedback is welcome. Thanks vt --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ZNYX 346Q on VxWorks 5.4 Date: 6 Sep 2002 20:19:41 -0700 From: andreas_lindell75@hotmail.com (Rikard) Organization: http://groups.google.com/ Message-ID: <5947a04.0209061919.4ee3831a@posting.google.com> References: <5947a04.0209052013.6da7b8fd@posting.google.com> Johan Borkhuis wrote in message news:... > andreas_lindell75@hotmail.com (Rikard) wrote: > > > Hello, > > > > I am trying to intergrate a ZNYX 346Q network card with my vxworks > > image. I can build it from command line like they descibe in the > > manual. Now I want to move my development into the workspace and I > > have contacted ZNYX and they say that vxworks is a "learning > > experience". > > > > Has anyone intergrated ZNYX 346Q so that it can be compiled from > > workspace? The reason I want it in workspace is that I want to be able > > to call: > > zxeUp("znb0", "192.168.0.241", 0xffffff00, 0, 0); > > zxeUp("znb1", "192.168.0.240", 0xffffff00, 0, 1); > > from UsrAppInit plus that all my other files are in the project > > facilities. So if anyone have worked with ZNYX 346Q in workspace then > > maybe you would be kind enough to give me some pointers. > > You could just add all the sourcefiles that are mentioned in the makefile > to your project and recompile, that would be the easiest way. > > Kind regards, > Johan Hello Johan, Thanks for you reply. However it made me little bit puzzled, the makefile in the project directory is regenerated each time a change is make in the project, new file is added etc. If I add the files in the makefile the addins will be deleted when an alteration is made, correct? This is what I have done so far: 1. I have clean vxworks image that compiles and runs on my target. 2. Added the 3 files if_zxeRlk.c, if_zxePci.c and configZxe.c in my project workspace and then did a dependency which resulted in that if_zxeRlk.h and if_zxePci.h was added under the external dependencies. 3. Included the INCLUDE_PCI in the vxworks image. 4. Put the 2 object files (if_zxe.obj and zxe_apps.obj) under EXTRA_MODULES and in the default directory of the project. 5. When I compile I get the following errors: cc386 -g -mpentium -ansi -nostdinc -DRW_MULTI_THREAD -D_REENTRANT - -fvolatile -nostdlib -fn o-builtin -fno-defer-pop -IC:\Tornado\target\proj\Project0 - -IC:\Tornado\target\config\pcPe ntium -IC:\Tornado\target\h -IC:\Tornado\target\config\comps\src - -IC:\Tornado\target\src\d rv -DCPU=PENTIUM -DPRJ_BUILD -DCPU_VARIANT=PENTIUM -c C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c: In function `zxePciMemSet': C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:67: `PHYS_MEM_DESC' undeclared (first use this function) C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:67: (Each undeclared identifier is reported only once C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:67: for each function it appears in.) C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:67: `pMmu' undeclared (first use this function) C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:68: syntax error before `i' C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:72: `i' undeclared (first use this function) C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:72: `sysPhysMemDescNumEnt' undeclared (first use this function) C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:73: `sysPhysMemDesc' undeclared (first use this function) C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:74: `pAddr' undeclared (first use this function) C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:84: `DUMMY_PHYS_ADDR' undeclared (first use this function) C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:89: `ZXE_INIT_STATE_MASK' undeclared (first use this function) C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:90: `ZXE_INIT_STATE' undeclared(first use this function) C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c: In function `zxePciInit': C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:129: `PCI_CFG_SUB_VENDER_ID' undeclared (first use this function) C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:130: `PCI_CFG_VENDOR_ID' undeclared (first use this function) C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:131: `PCI_CFG_DEVICE_ID' undeclared (first use this function) When I examine the makefile the 3 source files are already included, so I guess that happends when I include the source files in my project. 1. When compiling from workspace do I need to include the if_zxeRlk.h and if_zxePci.h files in any of the other project files like sysLib.c? 2. When I worked in command line I did a few alterations in some files(config.h, and configNet.h and syslib.c). Do I need to make these alterations when I work in workspace. Regards Rikard --------------------------- Newsgroups: comp.os.vxworks Subject: Filename from FILE * or descriptor? Date: 06 Sep 2002 23:15:50 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: Sender: bpringle@DeadDuck Is there a vxWorks function (v5.4.x) that will get a file name from either a file descriptor or the ANSI file handle. I have something like this currently... #include ... const char *getFileName(int fd) { const char * lval = "UNKNOWN"; if(ERROR != iosValidFd(fd)) { lval = fdTable[fd-3].name; } return lval; } This is rather ugly, but it is functional. You must subtract three for the stdin, stdout and stderr that are intrinsic and not in fdTable. It would be really nice if there was a function to do this, but my search of the documentation turned up nothing. However, I may have developed a blind spot while reading the manual... I need this as I am writing a library/function that take a file handle and not the file name. The file name is only needed as a debugging aid. Thanks, Bill Pringlemeir. - -- CLICKED BY THOUSANDS OF ATOMIC BOMBS. SINKING. U-63. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks 5.4.2 on Motorola Power PC 82xx Date: 6 Sep 2002 20:54:35 -0700 From: andreas_lindell75@hotmail.com (Rikard) Organization: http://groups.google.com/ Message-ID: <5947a04.0209061954.67b8e472@posting.google.com> References: <66f277de.0209061350.50ac78b0@posting.google.com> vtriplic@yahoo.com (vt) wrote in message news:<66f277de.0209061350.50ac78b0@posting.google.com>... > We are planning to use VxWorks on Motorola PPC 82xx. I searched this > group for any user feedbacks, but I couldn't find much. Does anybody > use it? Any issues that we need to be aware of up front? Any fedback > is welcome. > Thanks > vt Hello VT, I have used a WRS SBC8240 board with no problem. WR have reference designs for various of different PPC platforms. You can check them out on WR webpage. If you buy one of those together with vxworks its a "out of the box" solution and you can start your development with no problem. If you go for a custom design with a 82xx CPU then its a different thing. Then you will need a BSP that supports your board and those can be expensive about 20000usd, not sure though and up to 2 months development depending on what peripherals you are using. Check out this link: http://www.windriver.com/products/html/reference_designs.html However I was using VxWorks 5.4, didnt even know that there was a 5.4.2 version? Regards Andreas --------------------------- Newsgroups: comp.os.vxworks Subject: Project management in tornado Date: 6 Sep 2002 21:17:52 -0700 From: vxnerd@rediffmail.com (subbarayan) Organization: http://groups.google.com/ Message-ID: Dear mentors, Tornado2.0 release notes documentation pdf files gives and mentions about making hierarchial subprojects and linking them to gether.I am using tornado2.0 and target is standard pc architecture with x86 processor.I followed the same proceedure exactly whats mentioned in the manual to make subprojects and making them as single object code.After following the steps mentioned to make subprojects and link them ,when i tried to build vxworks i get the following error: make: *** No rule to make target `BuildProject4', needed by `master'. Stop. To clarify further i would like to mention that i made a bootable project called master and two downloadable projects called project4 and project5.Then in my master project in the builds tab i selected default and accessed the build properties dialogue box.Then i selected the rules pane and then clicked on new/edit button. Then i defined a new rule and typed master in target box and entered BuildProject4 BuildProject5 vxWorks in dependencies dialog box.Then i unchecked invoke this rule before building project check box. Then similarly as mentioned in manual i created a clean rule. then after that clicked apply.Then i set my default rule to be master. after all these things selected macros in build properties and defined the following path where my gnu source for file is present: D:/Tornado/target/proj/master/project4/SIMNTgnu/project4.a D:/Tornado/target/proj/master/project5/SIMNTgnu/project5.a.Then i clicked apply.then returned to files tab and clicked on build menu.I got the above said error. But one more thing to say theres not actually a .a file in the path mentioned. will the .a file be created automatically or before building i must compile both of my downloadable projects? can anyone point me where will be the source of error and help me overcome this error? all the 3 projects are in same workspace and same folder called master. Also can anyone explain me the use of building clean type of image(vxworks clean)? advanced thanks to the person replying me. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Anyone have WRS GCC sources? Date: 6 Sep 2002 23:29:54 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0209062229.267e1f2e@posting.google.com> References: Hello all, Andrew Johnson wrote in message news:... > Paul D. Smith wrote: > > > > I dunno from Tornado, I only do VxWorks itself :). What versions of > > VxWorks correspond to Tornado 2.0.2, 2.1, 2.2, and 3.0? > > Usually: > Tornado 2.0 comes with vxWorks 5.4 > Tornado 2.0.2 comes with vxWorks 5.4.2 > Tornado 2.2 comes with vxWorks 5.5 > Tornado 3.0 comes with vxWorks AE To be 100% accurate: Tornado 3.0 came with VxWorks AE 1.0 Tornado 3.1 comes with VxWorks AE 1.1 You shouldn't be able to buy 3.0 anymore - it has been superceded by 3.1 (if you still have 3.0, contact WR to upgrade to the newer version). > > Our version says VxWorks 5.4.2, and comes with gcc 2.96 (according to > > ccmips --version) and binutils 2.9.5 (according to ldmips --version). > > > > I think we have an enhanced version of 5.4.2: we are using it to build > > for the new Broadcom BSP. Apparently this was unified in VxWorks 5.5 > > but we don't have a copy of that yet. > > It sounds like you have a special version which doesn't follow the above > numbering. According to the vxWorks 5.5 GNU Toolkit Release Notes the > version of GCC supplied with T2.2 is called "2.96+" which may be slightly > different to 2.96 (you might have a version that was used to build T2.1 or > T3.0, which came out before T2.2). I guess you're just going to have to > wait for the source from WRS. The Wind River 2.96 (and upcoming 2.96+) are not the same as the Cygnus/RedHat version 2.96. You will need to get the sources from Wind River. Currently, you should do this by calling your local sales guy or FAE - they should be able to help you get the CD. Hopefully soon they will be available for download, or at least online ordering... HTH, John... --------------------------- Newsgroups: comp.os.vxworks Subject: selection of proessor Date: 7 Sep 2002 00:43:52 -0700 From: compaqbala@sify.com (Balakumar) Organization: http://groups.google.com/ Message-ID: <46992b69.0209062343.7b31f237@posting.google.com> i am newbie for embedded system Can anybody tell me what are the major processors(micro controllers) is used for embedded system developmet. In the market , MIPS, Power PC, ARM, X86, SUN sparc like so many processors are there. If i want to build embedded application, should i know the architecture and instruction set of the above processor. Please tell me. Majorly what processor or microcontrollers are used for embedded system. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: disk cache error Date: Sat, 7 Sep 2002 09:51:20 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <3D78BE3C.2DBC1BAD@notroll.com> If you need the sync interval set to 0, then maybe your better off without dcache - you can call dosFsDevCreate with a Blk device id. Also, if dcache detects that block 0 has been unexpectedly changed, and you have not changed the device, it is a goo dreason to worry. In other words, if the driver (TFFS?) changes block 0 without an apparent reason, there could be other problems that could result in data corruption. Remember that dcache only checks on block 0 as an additional means for detecting card removal, and you can eliminate dcache, which will not eliminate the actual problem though. Regarding the exception, it seems that it happens inside TFFS code, which indeed has its own DosFs formatting function, but the actual crash could be somewhere after that function, in another function defined as "static", so its entry addr isnt in the symbol table. "Mary McHugh" wrote in message news:3D78BE3C.2DBC1BAD@notroll.com... > Leonid Rosenboim wrote: > > If your device is not removable, you should set the device "removable" flag > > to FALSE, thereby disabling these autodetecion mechanisms. > > > > If it is in fact a removable device, as soon as a new card has been > > inserted, any subsequent call to open() will mount the new device and you > > will be able to access it. The file descriptors that where opened before the > > exchange took place, will be invalid, and need to be closed, to avoid FD > > leakage. > > > > HTH, > > - Leonid > > We are experiencing the same problem with our flash. However, we DO > have the removable flag set to false and we still get the same error as > the OP. Any other ideas? > > Additionally, another problem we're having is regading tuning the dcache > for the flash to have a 0 sync interval in order to force dcache to > write immediately (dcacheDevTune). We start out OK, but after running > for "awhile" (not sure what the exact interval is), something changes > the sync interval back to 1 second. I know for sure that chkdsk will do > it however we only run chkdsk on start/restart before we tune the > cache. > > Problem 3 is if we have a burst of flash activity, for example we unzip > a package containing many files, we occasionally crash. Sometimes it is > in the task that is executing the unzip, sometimes it crashes in the > tDcacheUpd task but it ALWAYS is at the same instruction in > flDosFormat. I'm not really sure why we're in flDosFormat -- it sounds > ominous. ;-) > > Here's what we get: > ---------------------- > data access > Exception current instruction address: 0x004aff6c > Machine Status Register: 0x00009032 > Data Access Register: 0xeeeeeeee > Condition Register: 0x42200040 > Data storage interrupt Register: 0x00009032 > Task: 0x6a6ecb8 "tDcacheUpd" > > -> l 0x004aff6c > 0x4aff6c a81e0000 lha r0,0(r30) > 0x4aff70 2c800000 cmpi crf1,0,r0,0 > 0x4aff74 40850078 bc 4,5,0x004affec > 0x4aff78 7fe3fb78 or r3,r31,r31 > 0x4aff7c 83bf0030 lwz r29,48(r31) > 0x4aff80 7fc4f378 or r4,r30,r30 > 0x4aff84 3bbdffff addi r29,r29,-1 > 0x4aff88 7f7de838 and r29,r27,r29 > 0x4aff8c 3bbd0001 addi r29,r29,1 > 0x4aff90 4bfff16d bl 0x004af0fc > value = 4915092 = 0x4aff94 = flDosFormat + 0xf48 > > > Any input that anyone can shed on this problem will be appreciated. I'm > off to "windsurf" and google for more info. > > Mary > mchug06@_disregard_attglobal.net --------------------------- Newsgroups: comp.os.vxworks Subject: Re: memory allocation Date: Sat, 7 Sep 2002 10:14:45 +0000 (UTC) From: David Laight Organization: BT Openworld Message-ID: References: > Does anyone know how it would be possible to allocate memory bellow > certain address (32 MB), eventhough system partition is bigger (128 > MB)? > I need this to solve the 24-bit reallocation value problem. Try: 1) Reverse engineer and re-writing memPartAlignedAlloc() so that (with a flag) it will allocate from the low end of a free block. 2) Keep allocating memory until you get a block you like, then free all the ones you didn't use. 3) Rebuild your image with a larger RAM_LOW_ADRS and statically allocate the data below the image. David --------------------------- Newsgroups: comp.os.vxworks Subject: Re: memory allocation Date: Sat, 7 Sep 2002 13:21:05 +0100 From: Wade Oram Organization: None Message-ID: References: Reply-To: Wade Oram In message , David Laight writes >> Does anyone know how it would be possible to allocate memory bellow >> certain address (32 MB), eventhough system partition is bigger (128 >> MB)? >> I need this to solve the 24-bit reallocation value problem. > >Try: > >1) Reverse engineer and re-writing memPartAlignedAlloc() so that > (with a flag) it will allocate from the low end of a free block. > >2) Keep allocating memory until you get a block you like, then free > all the ones you didn't use. > >3) Rebuild your image with a larger RAM_LOW_ADRS and statically allocate > the data below the image. > > David > As 3 above but instead of using static allocation, add the memory that you have reserved (from the Original RAM_LOW_ADDRS to the modified RAM_LOW_ADDRS) to a new memory partition (created with memPartCreate). You can then use memPartAlignedAlloc on this new memory partition to allocate memory from the region that you have reserved. - i.e below the new value of RAM_LOW_ADRS. - -- Wade Oram --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Project management in tornado Date: Sat, 7 Sep 2002 19:42:55 +0200 From: "Martin Raabe" Organization: 1&1 Internet AG Message-ID: References: Hello subbarayan, my recommendation is to give SNiFF+ 4.x a try. (the latest version!) It has everything you want to do with the Tornado IDE. Ask WindRiver for a trial version. Ciao - -- Martin Raabe E: Martin.RaabeATNOSPAMBaSystem.de "subbarayan" schrieb im Newsbeitrag news:f81c4277.0209062017.34726315@posting.google.com... > Dear mentors, > Tornado2.0 release notes documentation pdf files gives and mentions > about making hierarchial subprojects and linking them to gether.I am > using tornado2.0 and target is standard pc architecture with x86 > processor.I followed the same proceedure exactly whats mentioned in > the manual to make subprojects and making them as single object > code.After following the steps mentioned to make subprojects and link > them ,when i tried to build vxworks i get the following error: > > make: *** No rule to make target `BuildProject4', needed by `master'. > Stop. > > To clarify further i would like to mention that i made a > bootable project called master and two downloadable projects called > project4 and project5.Then in my master project in the builds tab i > selected default and accessed the build properties dialogue box.Then i > selected the rules pane and then clicked on new/edit button. > Then i defined a new rule and typed master in target box and entered > BuildProject4 BuildProject5 vxWorks in dependencies dialog box.Then i > unchecked > invoke this rule before building project check box. > Then similarly as mentioned in manual i created a clean rule. > then after that clicked apply.Then i set my default rule to be master. > > after all these things selected macros in build properties and defined > the following path where my gnu source for file is present: > D:/Tornado/target/proj/master/project4/SIMNTgnu/project4.a > D:/Tornado/target/proj/master/project5/SIMNTgnu/project5.a.Then i > clicked apply.then returned to files tab and clicked on build menu.I > got the above said error. > > But one more thing to say theres not actually a .a file in the path > mentioned. > will the .a file be created automatically or before building i must > compile both of my downloadable projects? > > can anyone point me where will be the source of error and help me > overcome this error? > all the 3 projects are in same workspace and same folder called > master. > > Also can anyone explain me the use of building clean type of image(vxworks clean)? > > advanced thanks to the person replying me. --------------------------- Newsgroups: comp.os.vxworks Subject: FTP Daemon doesn't work Date: Sat, 7 Sep 2002 14:40:19 -0400 From: "John Sestak" Organization: Raytheon Company Message-ID: Hi all, I'm using the FTP daemon that came with the Dos Fs 2.0 software (I think) but it doesn't seem to work terribly well. I can't seem to do 'puts' without the transfer rate slowing to a crawl or dieing completely (the same thing goes for the 'gets'). Is there a patch for this thing that fixes some of the problems I'm seeing while using this thing? Is there perhaps another FTP Daemon out there that works better that someone might know about? Thank you. John Sestak --------------------------- Newsgroups: comp.os.vxworks Subject: Re: FTP Daemon doesn't work Date: Sat, 7 Sep 2002 21:47:09 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: First, the enhanced FTP server that you get with dosFs 2.0 is in the "unsupported" directory, which you need to copy somewhere and compile with your project. Are you sure you did that ? Secondly, did you measure performance locally, i.e. the file system and all the way down to the disk and its device driver prior to using FTP ? Thirdly, have you tested the network interface and driver for throughput before using FTP ? Lastly, I suggest to check out the state of network stack buffers. Will be happy to assist if you send out some more details. - - Leonid "John Sestak" wrote in message news:rKre9.1558$c6.2365@bos-service2.ext.raytheon.com... > Hi all, > I'm using the FTP daemon that came with the Dos Fs 2.0 software (I > think) but it doesn't seem to work terribly well. I can't seem to do 'puts' > without the transfer rate slowing to a crawl or dieing completely (the same > thing goes for the 'gets'). Is there a patch for this thing that fixes some > of the problems I'm seeing while using this thing? Is there perhaps another > FTP Daemon out there that works better that someone might know about? Thank > you. > > John Sestak > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Exception: what's the meaning? Date: 7 Sep 2002 21:11:57 -0700 From: roswel_ajf@hotmail.com (roswell) Organization: http://groups.google.com/ Message-ID: <6dcae13d.0209072011.547f913@posting.google.com> hi folks, i added 3 virtual interfaces ( all ethernettype, hooked up to IPstack). Afterwards, i "think" that some instruction memory has got corrupted (in taskInit). It is shown when i do a ping from target to host. Or invoking any other taskSpawn crash it too. see the dump below. target is PPC750, vxwork 5.4 i think, for sure adding the 3 vitural muxdrivers causing this. if i add only 1 vif, then, no crash. any clue how to decode this exception dump, and hint what might happen with adding 3 interfaces. thanks a million for any hint. - -> ping "host", 1 program Exception current instruction address: 0x00000000 Machine Status Register: 0x0008b030 Condition Register: 0x44400080 b4408 vxTaskEntry +60 : shell () 853f4 shell +18c: 85420 () 85644 shell +3dc: execute () 857c8 execute +d8 : yyparse () db5dc yyparse +7a8: d9524 () d969c yystart +8f8: ping () 60674 ping +2d8: taskSpawn () 9aac4 taskSpawn +64 : taskCreat () 9accc taskCreat +1c0: taskInit () 9b1b4 taskInit +474: 0 () no answer from host OR - -> taskSpawn "hi", 100, 0, 5000, ssm_vif_ready program Exception current instruction address: 0x00000000 Machine Status Register: 0x0008b030 Condition Register: 0x44400080 b4408 vxTaskEntry +60 : shell () 853f4 shell +18c: 85420 () 85644 shell +3dc: execute () 857c8 execute +d8 : yyparse () db5dc yyparse +7a8: d9524 () d969c yystart +8f8: taskSpawn () 9aac4 taskSpawn +64 : taskCreat () 9accc taskCreat +1c0: taskInit () 9b1b4 taskInit +474: 0 () --------------------------- Newsgroups: comp.os.vxworks Subject: CNF: Embedded Systems Club and Ada UK User Group Autumn Conference Date: Sun, 8 Sep 2002 10:50:10 +0100 From: "John Robinson" Organization: John Robinson And Associates Ltd Message-ID: Followup-To: comp.os.vxworks Reply-To: "John Robinson" Conference highlights include: * "Correctness by construction: Developing the MULTOS secure certification authority for smart cards" * "C++: Fit for embedded systems?" * "Test-First Design" (Tutorial) * "Principles of high availability embedded systems design" * "Automating software design/code verification" * "Scenario based testing" Abstracts for all 22 sessions are now available at http://www.EmbeddedSystemsClub.com. More details follow: =========================== Biggest and Best Event Yet! =========================== Our Autumn conference promises to be our best event yet, with 22 sessions in 3 streams. The event is to be held in Swindon, UK on 29th October 2002 and is open to all: * Member rate only £75(?120)+vat. * Associate Members & Non-Members only £170(?271)+vat (including a 12-month membership of the Embedded Systems Club or Ada UK User Group - value £95(?156)+vat The event includes: * a dedicated testing stream * a dedicated Ada stream * tutorials on "Test-First Design" and "Principles of High Availability Embedded Systems Design" * the opportunity to learn about, and contribute to, the development of new standards * a session dedicated to Ada code generation from UML models * a pre-event survey and birds-of-a-feather session on recruitment and retention of Ada staff * technical presentations on many embedded systems topics, including "C++ in Embedded Systems", "Automated software design/code verification" and "Scenario Based Testing". ============================ Not Just Another Conference! ============================ From the outset we have planned this conference to provide more than a sequence of presentations to sit and listen to. There are many opportunities to actively contribute to the event, including: * a pre-event survey and birds-of-a-feather session on staff recruitment and retention * two panel sessions, one dedicated to a proposal to develop a standardised Ada code generation profile for the UML In addition, we have a wealth of sessions providing training and education on tools and techniques including two tutorials on Test-First Design (a key technique in eXtreme Programming) and Principles of High Availability Embedded Systems Design. All of which means that this event should not be missed by anyone working in the embedded systems sector. ================================ Do not miss this opportunity to: ================================ * Acquire new skills * Contribute to new standards and industry developments: * Address key industry issues * Learn from the experience of others in your industry * Update yourself on latest market developments with vendor updates from some of the leading players in the embedded market ============ Registration ============ Registrations can be made via the websites: http://www.EmbeddedSystemsClub.com http://www.AdaUK.org.uk Or by requesting a PDF or hard copy registration form from the conference manager: mailto:Hazel@Adaxia.com We look forward to seeing you on the 29th October 2002. Best regards John Robinson Conference Director mailto:John@JohnRobinsonAndAssociates.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: NFS server throughput Date: Sun, 8 Sep 2002 13:05:46 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Folks, i have received no response on this query, Does this represent a total lack of interest in this subject ? does this mean nobody is using the VxWorks NFS server ? Or does this mean the ones who do use it only have Flash disks ? So if anyone has still got any comment on VxWorks NFS, you whish to share with this group, please do. - - Leonid "Leonid Rosenboim" wrote in message news:newscache$9czj1h$c7l$1@lnews.actcom.co.il... > I guess now it's my turn to ask a question for once: > > If anyone is running the VxWorks NFS server on top of DosFs 2.0, where the > disk storage is either IDE or SCSI hard drives (not Flash disks of any > kind), I would appreciate if you could share your experience with respect to > the throughput and performance of this configuration. Please specify what > platform is your NFS client runs on too. > > Flash drives are not interesting to me simply because their inherent slow > throughput. > > Thanks, > Leonid > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: NFS server throughput Date: Sun, 08 Sep 2002 13:11:48 GMT From: "Jeffrey Creem" Organization: AT&T Broadband Message-ID: References: "Leonid Rosenboim" wrote in message news:newscache$2m942h$98f$1@lnews.actcom.co.il... > Folks, i have received no response on this query, > Does this represent a total lack of interest in this subject ? > does this mean nobody is using the VxWorks NFS server ? > Or does this mean the ones who do use it only have Flash disks ? > > So if anyone has still got any comment on VxWorks NFS, > you whish to share with this group, please do. In my case it means that I use the NFS server but I use an Ata flash disk so I did not originally respond. In addition, I am using it with a non 2.0 DosFs package (which I had to get the source code for from my tech rep and modify in order to make the whole thing actually work in all cases). --------------------------- Newsgroups: comp.os.vxworks Subject: Flash Invalidate Date: 8 Sep 2002 06:12:09 -0700 From: yahoo_m1@hotmail.com (John) Organization: http://groups.google.com/ Message-ID: Hi All, I am using MPC750 processor on a custom board and trying to do a Flash Invalidation of the data cache, by setting the DCFI bit in the HID0 register. But, I am finding some unusual results. After the invalidation, either the board completely hangs, or it reboots with a workQPanic message or even if the board is alive, i cannot access it from the Shell. And, these results are not consistent. I have the interrupts locked and also calling my invalidation routine from the task context. Any comments on this will be highly appreciated. Thanks in Advance, John --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Flash Invalidate Date: Sun, 8 Sep 2002 17:02:52 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: By invaliadtiong data cache, you can sometimes loose some of the recent modifications, i.e. if your cache is in write-back mode, there could be some "dirty" cache lines that where not yet written to memory, which invalidation will erase. If you want to clear the entire cache, the only safe option to use is "flush & Invalidate". Cache invalidate can be used on a specific address range which has been give to a DMA (aka Bus Master) device for fetching data, and since the device will subsequently write into main memory at this range, whatever value was stored in cache need to be invalidated. "John" wrote in message news:baa5411a.0209080512.116502db@posting.google.com... > Hi All, > I am using MPC750 processor on a custom board and trying to do a Flash > Invalidation of the data cache, by setting the DCFI bit in the HID0 > register. But, I am finding some unusual results. After the > invalidation, either the board completely hangs, or it reboots with a > workQPanic message or even if the board is alive, i cannot access it > from the Shell. And, these results are not consistent. I have the > interrupts locked and also calling my invalidation routine from the > task context. > > Any comments on this will be highly appreciated. > > Thanks in Advance, > John --------------------------- Newsgroups: comp.os.vxworks Subject: Help with ZNYX 346Q for vxworks 5.4 Date: 8 Sep 2002 22:58:56 -0700 From: andreas_lindell75@hotmail.com (Rikard) Organization: http://groups.google.com/ Message-ID: <5947a04.0209082158.555302ed@posting.google.com> Hello Everyone, I am working with a znyx 346q network card that I try to get running with vxworks 5.4. I am trying to get it to compile and run using workspace not command line as specified in the manual. Everyone tells me to include the files in my project and then modify then builds extra modules by adding the object files. Sounds easy but so complicated for me. This is what I have done so far: 1. I have clean vxworks image that compiles and runs on my target. 2. Added the 3 files if_zxeRlk.c, if_zxePci.c and configZxe.c in my project workspace and then did a dependency which resulted in that if_zxeRlk.h and if_zxePci.h was added under the external dependencies. 3. Included the INCLUDE_PCI in the vxworks image. 4. Put the 2 object files (if_zxe.obj and zxe_apps.obj) under EXTRA_MODULES and in the default directory of the project. 5. When I compile I get the following errors: cc386 -g -mpentium -ansi -nostdinc -DRW_MULTI_THREAD -D_REENTRANT - -fvolatile -nostdlib -fn o-builtin -fno-defer-pop -IC:\Tornado\target\proj\Project0 - -IC:\Tornado\target\config\pcPe ntium -IC:\Tornado\target\h -IC:\Tornado\target\config\comps\src - -IC:\Tornado\target\src\d rv -DCPU=PENTIUM -DPRJ_BUILD -DCPU_VARIANT=PENTIUM -c C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c: In function `zxePciMemSet': C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:67: `PHYS_MEM_DESC' undeclared (first use this function) C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:67: (Each undeclared identifier is reported only once C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:67: for each function it appears in.) C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:67: `pMmu' undeclared (first use this function) C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:68: syntax error before `i' C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:72: `i' undeclared (first use this function) C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:72: `sysPhysMemDescNumEnt' undeclared (first use this function) C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:73: `sysPhysMemDesc' undeclared (first use this function) C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:74: `pAddr' undeclared (first use this function) C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:84: `DUMMY_PHYS_ADDR' undeclared (first use this function) C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:89: `ZXE_INIT_STATE_MASK' undeclared (first use this function) C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:90: `ZXE_INIT_STATE' undeclared(first use this function) C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c: In function `zxePciInit': C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:129: `PCI_CFG_SUB_VENDER_ID' undeclared (first use this function) C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:130: `PCI_CFG_VENDOR_ID' undeclared (first use this function) C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:131: `PCI_CFG_DEVICE_ID' undeclared (first use this function) 1. When compiling from workspace do I need to include the if_zxeRlk.h and if_zxePci.h files in any of the other project files like sysLib.c? 2. When I worked in command line I did a few alterations in some files(config.h, and configNet.h and syslib.c). Do I need to make these alterations when I work in workspace? Any other pointers in the right direction would be great. Regards Rikard --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ZNYX 346Q on VxWorks 5.4 Date: Mon, 9 Sep 2002 06:11:19 +0000 (UTC) From: Johan Borkhuis Organization: Proxim Corp. Message-ID: References: <5947a04.0209052013.6da7b8fd@posting.google.com> <5947a04.0209061919.4ee3831a@posting.google.com> Hello, > 4. Put the 2 object files (if_zxe.obj and zxe_apps.obj) under > EXTRA_MODULES and in the default directory of the project. You did receive these 2 files from your supplier? > 5. When I compile I get the following errors: > C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c: In > function `zxePciMemSet': > C:\Tornado\target\proj\pcpentium_ata_with_znyx\if_zxePci.c:67: > `PHYS_MEM_DESC' undeclared (first use this function) > ..... It looks like the problem you are having is caused by some missing included in your own source files. > When I examine the makefile the 3 source files are already included, > so I guess that happends when I include the source files in my > project. > > 1. When compiling from workspace do I need to include the if_zxeRlk.h > and if_zxePci.h files in any of the other project files like sysLib.c? No. > 2. When I worked in command line I did a few alterations in some > files(config.h, and configNet.h and syslib.c). Do I need to make these > alterations when I work in workspace. No, these files are included when you build using Tornado. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: memory allocation Date: Mon, 9 Sep 2002 06:14:37 +0000 (UTC) From: Johan Borkhuis Organization: Agere Systems Message-ID: References: langus@iskratel.si (Matevz Langus) wrote: > Hi! > > Does anyone know how it would be possible to allocate memory bellow > certain address (32 MB), eventhough system partition is bigger (128 > MB)? > I need this to solve the 24-bit reallocation value problem. Build your application with the "-mlongcall" option. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: How to read the route table? Date: Mon, 9 Sep 2002 14:14:20 +0800 From: "Xin Zhen" Organization: Bentium Ltd. (CN99) Message-ID: How to read the route table of vxworks, just like what "routeShow" has done? I'm developing a network device using vxworks, and want to provide users with these function "show route", "empty route table(set to default)", etc. So i'll want to get the whole route table. I've checked the Reference Manual but still can't get it. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: selection of proessor Date: Mon, 9 Sep 2002 06:21:16 +0000 (UTC) From: Johan Borkhuis Organization: Agere Systems Message-ID: References: <46992b69.0209062343.7b31f237@posting.google.com> compaqbala@sify.com (Balakumar) wrote: > Can anybody tell me > what are the major processors(micro controllers) is used for embedded > system developmet. > In the market , MIPS, Power PC, ARM, X86, SUN sparc like so many > processors are there. They are all used a lot, but it largely depends on your application. You should define what you need (memory, IO, processor power, power consumption, price, etc) and then look at the different processors and select the one that comes closest to your requirements. > If i want to build embedded application, should i know the > architecture and instruction set of the above processor. Architecture: yes, instuction set: not realy needed if you are using a pre-build OS (like VxWorks), but it might help. If you are using your own OS, you need to learn the ins and outs of the processor. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: memory allocation Date: Mon, 9 Sep 2002 16:06:37 +0800 From: "Seraph Ma" Organization: Bentium Ltd. (CN99) Message-ID: References: I totally agree with Johan's suggestion. The 24-bit reallocation problem can be sloved by adding the -mlongcall option in you makefile. "Johan Borkhuis" wrote in message news:Xns92845351439BAborkhuisagerecom@192.19.193.66... > langus@iskratel.si (Matevz Langus) wrote: > > > Hi! > > > > Does anyone know how it would be possible to allocate memory bellow > > certain address (32 MB), eventhough system partition is bigger (128 > > MB)? > > I need this to solve the 24-bit reallocation value problem. > > Build your application with the "-mlongcall" option. > > Kind regards, > Johan > > -- > o o o o o o o . . . _____________________________ > o _____ || Johan Borkhuis | > .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | > >(________|__|_[_________]_|__________________________| > _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` > === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === > --------------------------- Newsgroups: comp.os.vxworks Subject: RAM Disk Date: 9 Sep 2002 01:20:26 -0700 From: vasu@inorbit.com (Vasudev Mulchandani) Organization: http://groups.google.com/ Message-ID: Hello there ! Just wondering if there are any public domain code for RAM Disk kind of driver available... TIA Vasu --------------------------- Newsgroups: comp.os.vxworks Subject: Re: RAM Disk Date: Mon, 9 Sep 2002 10:50:59 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: If you have VxWorks, there is a RAM disk driver delivered in source code with the system. If you don't use VxWorks, you have contacted the wrong newsgroup "Vasudev Mulchandani" wrote in message news:f9630ce5.0209090020.6badb32f@posting.google.com... > Hello there ! > > Just wondering if there are any public domain code for RAM Disk kind of > driver available... > > TIA > Vasu --------------------------- Newsgroups: comp.os.vxworks Subject: writing flash driver in vxworks Date: 9 Sep 2002 02:05:42 -0700 From: sjith2000@yahoo.com (Sreejith) Organization: http://groups.google.com/ Message-ID: <8b17c637.0209090105.243c015@posting.google.com> hi everybody I am a newbie in vxworks.We have to develop a flash driver for the chip AM29F080B of entry level MBX860 embedded controller.We went thru the manual of the chip.Anybody can suggest further manuals or books which tells more about flash driver writing?Thanks in advance. regards sreejith --------------------------- Newsgroups: comp.os.vxworks Subject: VxWorks Applications Engineer Date: 9 Sep 2002 02:55:00 -0700 From: Bam2974@hotmail.com (Bamidelle) Organization: http://groups.google.com/ Message-ID: <66a11529.0209090154.4a72b3c9@posting.google.com> Hi, I am looking for an experienced Engineer with either an applications engineering or a software engineering background and strong VxWorks experience. We are a leading edge developer of: VME * CPCI * Multibus II * PMC * Real-time OS * Custom Solutions *. We are looking for someone highly technical with good communication skills. You need to be authorised to work in the UK, and want to take a position here. We will pay up to £ 35 k and offer strong package and career path. Is there anyone out there who can help ? Please reply with full CV and contact details for informal + totally confidential discussion. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: selection of proessor Date: 9 Sep 2002 03:11:34 -0700 From: langus@iskratel.si (Matevz Langus) Organization: http://groups.google.com/ Message-ID: References: <46992b69.0209062343.7b31f237@posting.google.com> it really depends on application you would like to build with your processor :-) compaqbala@sify.com (Balakumar) wrote in message news:<46992b69.0209062343.7b31f237@posting.google.com>... > i am newbie for embedded system > Can anybody tell me > what are the major processors(micro controllers) is used for embedded > system developmet. > In the market , MIPS, Power PC, ARM, X86, SUN sparc like so many > processors are there. > If i want to build embedded application, should i know the > architecture and instruction set of the above processor. > Please tell me. > Majorly what processor or microcontrollers are used for embedded > system. --------------------------- Newsgroups: comp.os.vxworks Subject: [Q] netperf UDP_STREAM / 0 Mbps / WIindows XP / dec21x4x Date: 9 Sep 2002 03:54:45 -0700 From: kotaeji@hanmail.net (Kim Tae Hyung) Organization: http://groups.google.com/ Message-ID: Hi, I have a question about benchmarking with netperf(UDP) on windows XP(NT) My testbed is as follows. PC1 -------100Mb Switch Hub------- PC2 | | | | | | Network Equipment I have two PCs of Windows XP, and connected to 100Mb Switch Hub. And Our Network Equipment has two ports and also connected to 100Mb Switch Hub. Okay, this is very simple forwarding test. :) netserver is on PC2, and netperf client works on PC1. [netperf -t UDP_STREAM -H 11.x.x.xxx -l 10] So, the UDP packet starts from PC1, goes to port 1 of Network Equipment, and port 2 of Network Equipment, and finally goes to PC2. If we check packets by windump, it seems that PC2 receives all packets from PC1. But the result of netperf always says 0 Mbps. But, if we do this same test by FreeBSD PCs, it works fine. netperf(Of course, freeBsd netperf) says XX Mbps. And, if we remove 100Mb Switch Hub and Network Equipment, (PC1 and PC2 are directly connected) PC1--------------------PC2 netperf works fine, and says XX Mbps even on Windows XP. Strangely to say, this strange behavior occurs only with 100Mb Switch Hub. Not with 100Mb dummy hub nor 10Mb dummy hub. And only occurs with UDP_STREAM test. Any comments appreciated. Thanks Thanks Thanks in advance. Network Equipment : VxWorks with dec21x4x driver --------------------------- Newsgroups: comp.os.vxworks Subject: fail to transfer large file using FTP client Date: 9 Sep 2002 03:57:07 -0700 From: g_ch@21cn.com (GANGCHENG) Organization: http://groups.google.com/ Message-ID: My enviroment is : VxSim win2000 ramDisk created on VxSim FTP client on VxSim/TFTP client on VxSim wftpd32 on PC. But when I transfer the file over about 1480 bytes using FTP client, it does not work again. It blocks on the sentence "read()" as follows.But similarly code as TFTP, it does work. Why ? if (ftpXfer2 (ipaddr, user,pwd, "", "RETR %s", dirname, filename, &ctrlSock, &dataSock) == ERROR) return (ERROR); /* read the remote file and write to the local file */ while ((nBytes = read (dataSock, (char*)buf, sizeof (buf))) > 0) { testBytes = write(fd,(char*)buf,nBytes); if(testBytes!=nBytes) { printf("Write to file %s, failed.\n", filename); status = ERROR; break; } } - ---------------------------------------------------------------------------- if (tftpXfer (ipaddr, 0, filename, "get", "binary", &dataFd, &errorFd) == ERROR) return (ERROR); /* read the remote file and write to the local file */ while ((nBytes = read (dataFd, (char*)buf, sizeof (buf))) > 0) { if(write(fd,(char*)buf,nBytes)!=nBytes) { printf("Write to file %s, failed.\n", filename); status = ERROR; break; } } --------------------------- Newsgroups: comp.os.vxworks Subject: Re: writing flash driver in vxworks Date: Mon, 9 Sep 2002 19:05:51 -0000 From: "christopher wang" Organization: Bentium Ltd. (CN99) Message-ID: References: <8b17c637.0209090105.243c015@posting.google.com> If you want, I can send my flash driver for vxworks to you. This driver support AM29F800B, AM29F800T, AM29F160B, AM29F160T. pls send mail to me christopher "Sreejith" wrote in message news:8b17c637.0209090105.243c015@posting.google.com... > hi everybody > I am a newbie in vxworks.We have to develop a flash driver for > the chip AM29F080B of entry level MBX860 embedded controller.We went > thru the manual of the chip.Anybody can suggest further manuals or > books which tells more about flash driver writing?Thanks in advance. > > regards > sreejith --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Receiving from sockets Date: Mon, 09 Sep 2002 08:33:54 -0400 From: Neil Horman Organization: WEBUSENET.com Message-ID: <3D7C9532.7020302@lvl7.com> References: <3D77BFEC.1080905@lvl7.com> Reply-To: nhorman@lvl7.com Can you post the snippet of code where you create the socket? It would be helpful to know in diagnosing your problem. Also, the address of the memory load exception would be helpful, as it may point to an error other than your socket configuration. Thanks Neil Ravi Uday wrote: > Neil Horman wrote in message news:<3D77BFEC.1080905@lvl7.com>... > >>It looks offhand like a tcp receive buffer size issue. look in the vxworks >>reference under setsockopt, and then look for the discussion of the SO_SNDBUF >>and SO_RCVBUF options. Ususally they are set pretty large, but in your case it >>looks like they are reduced to be around 1024 bytes, which is where you are >>seeing your problem. Also, I would instrument your code to check the return >>code of recv for -1 instead of just adding it to your overall length. If its -1 >>then dump out errno to see what the specific error is. That should give you >>more of an idea whats going wrong >>Neil >> >> > Neil : > Thanks for the suggestion. Well i shall look thr those socket options > for improving my tcp connection behaviour. Furrther .. > > > The 3rd parameter to 'recv' specifies number of bytes to > read..but..say i specify 1000 then not necessarily it reads all 1000 > bytes. We get the indication of the actual bytes read in ' i ' the > return of 'recv' function. Hence i had set a while loop like that. I > see something fishy in the way i move the ' resptr' variable. Do i > need it to bzero it..( reset its memory contents ) everytime before > calling ' recv '. Its continuing to give memory failures when the size > of data on the other side is over 1024 bytes !! > > A printf inside while loop for actual bytes read ' i ' for a 'len' of > '1872' ( i.e. len = 1872 ) was like this. > [ ' buf ' variable is malloced for 2048 bytes and len = 2048) > 1st iteration - i = 400 > 2nd iteration - i = 100 > 3rd iteration - i = 62 > 4th iteration - i = 400 > Then > - MEMORY FAILURE - ( Address load Exception ) > > while ( *inBodySize <= len ) { > > i = recv ( SockDesc, resPtr, 400, 0); > resPtr = resPtr + i; > *inBodySize += i; > printf ("\n~~~~~~~~...SizeACCUMULATED *inBodySize - %d \n", > *inBodySize); > } > Any improvements in here Neil that you can think of.. > > - Ravi > --------------------------- Newsgroups: comp.os.vxworks Subject: how to get "Physical Address" directly from PCI enumeration? Date: 09 Sep 2002 13:57:44 GMT From: charly Message-ID: <200299-145744-758413@foorum.com> Hello, I would like to read the "Physical Address" directly from PCI enumeration. Is it possible to get it from "base address 0 " or "base address n"? Thanks charly - ---------------------- example of Physical Address ------------- Ethernet adapter : Description . . . . . . . . : 3Com 3C90x Ethernet Adapter Physical Address. . . . . . : 00-C0-4F-89-DB-22 DHCP Enabled. . . . . . . . : Yes .... - ---------------------- example of enumeration ------------- - -> pciHeaderShow(0,0xb,0) vendor ID = 0x10b5 device ID = 0x9050 command register = 0x0003 status register = 0x0280 revision ID = 0x02 class code = 0x06 sub class code = 0x80 programming interface = 0x00 cache line = 0x08 latency time = 0x00 header type = 0x00 BIST = 0x00 base address 0 = 0x13711000 base address 1 = 0x0000d401 base address 2 = 0x00000000 base address 3 = 0x0000d801 base address 4 = 0x0000dc01 base address 5 = 0x13712000 cardBus CIS pointer = 0x00000000 sub system vendor ID = 0x10b5 sub system ID = 0x9050 expansion ROM base address = 0xe6000000 interrupt line = 0x0a interrupt pin = 0x01 min Grant = 0x00 max Latency = 0x00 - -- Use our news server 'news.foorum.com' from anywhere. More details at: http://nnrpinfo.go.foorum.com/ --------------------------- Newsgroups: comp.os.vxworks Subject: vxWorks compilation-time errors Date: Mon, 9 Sep 2002 17:18:52 +0400 From: "Denis Dubrovin" Organization: WEBPlus Ltd. Message-ID: Hi! WinNT\Tornado 2\Diab RTA Suite When I compile vxWorks image ( vxWorks.rom ) "as is" - in wizard configuration - all OK. But! 1. If I include "Diab RTA support" component I get error: ================================== F:\Tornado\host\x86-win32\bin\binToAsm tmp.Z > vxWorks.Z.s dcc -c -g -tPPC823FS:tornado -D__asm__=__asm -Xansi -Xstrings-in-text=0 -ew1 551 -Xintrinsi c-mask=0x3 -c -Xpreprocess-assembly -IC:\USERS\DEN\vxWorks\vxTest -IF:\Torna do\target\conf ig\wrMdpBasic8xxdiab -IF:\Tornado\target\h -IF:\Tornado\target\config\comps\ src -IF:\Torna do\target\src\drv -DCPU=PPC860 -DPRJ_BUILD -g -O vxWorks.Z.s -o vxWorks.Z.o "vxWorks.Z.s", line 35643: error: syntax error make: *** [vxWorks_rom] Error 0x1 ================================== 2. If I include "spy" component I get another error: ===================================== dld -tPPC823FS:tornado -X -e _sysInit \ dataSegPad.o partialImage.o ctdt.o -o vxWorks \ ld_low.lcf dld: warning: Undefined symbol 'sysAuxClkDisable' in file 'partialImage.o' dld: warning: Undefined symbol 'sysAuxClkRateGet' in file 'partialImage.o' dld: warning: Undefined symbol 'sysAuxClkRateSet' in file 'partialImage.o' dld: warning: Undefined symbol 'sysAuxClkEnable' in file 'partialImage.o' dld: warning: Undefined symbol 'sysAuxClkConnect' in file 'partialImage.o' dld: Undefined symbols found - no output written make: *** [vxWorks] Error 0x1 ===================================== How to make it working? - -- Regards! Den --------------------------- Newsgroups: comp.os.vxworks Subject: Re: selection of proessor Date: 9 Sep 2002 07:35:11 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0209090635.326bd2b9@posting.google.com> References: <46992b69.0209062343.7b31f237@posting.google.com> compaqbala@sify.com (Balakumar) wrote in message news:<46992b69.0209062343.7b31f237@posting.google.com>... > In the market , MIPS, Power PC, ARM, X86, SUN sparc like so many > processors are there. Another important number here is the total number of units you'll finally produce. E.g., code density will have influence on the size of the flash chips you'll have to use (we're talking about your money here). Some chips (like SH) are difficult to obtain for low quantities... but get very cheap when using for high quantities (especially when they're integrated in a design). PPC might be your choice when doing communication heavy work (with reasonable quantity). If you start with very low quantities you might as well take a Pentium... however I'd guess you won't be really happy with this choice (unless doing more than a few dozen, you don't have to forget they typically require a BIOS to come up, they're quite expensive, require extensive external hardware support (unless taking one of the embedded versions))... I'd say lot's of low quantity products typically start off with PPC... all: I guess this might give lot's of ground for discussion, but let me remind you I just fired out a few thoughts (where I would be happy to hear comments, suggestions) toni --------------------------- Newsgroups: comp.os.vxworks Subject: Error messages Date: 9 Sep 2002 07:38:56 -0700 From: yahoo_m1@hotmail.com (John) Organization: http://groups.google.com/ Message-ID: Hi All, My system generated the following error message while I was trying to run my code. - ------------------------------------------------------------------------------ Exception at interrupt level: data access Exception current instruction address: 0x0002c444 Machine Status Register: 0x00009030 Data Access Register: 0xeeeeeef2 Condition Register: 0x22020040 Data storage interrupt Register: 0x40000000 Regs at 0x7e80000 - ------------------------------------------------------------------------------ From this I understand that there was a data access exception. For some reason the system tried to access the address 0xeeeeeef2 (which is a invalid address). I also understand that the instruction which caused this exception was at address 0x0002c444. But, I am not able to decipher the other error messages like the "Data Storage interrupt Register: 0x40000000", "Regs at 0x7e80000" etc. Can you share some information related to this as to how to decipher this and use it for debugging. Basically, the information from the current instruction address and the contents of Data access registers were not enough for my debugging. I am using VxWorks 5.4, Tornado 2.0.2 for PowerPC architecture. Thanks in Advance, John. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How to read the route table? Date: 9 Sep 2002 07:43:58 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0209090643.317e9012@posting.google.com> References: You'll find the clues to what you're searching within: struct radix_node_head *rt_tables[]; struct radix_node *radixTree= rt_tables[2]->rnh_treetop; radixTree will be your start point, you'll have to follow the tree structure to the right and to the left: radixTree->rn_l radixTree->rn_r radixTree->rn_b distinguishes between leafes and nodes. radixTree->rn_dupedkey tells you whether there are duplicates available. This should be at least a start, toni --------------------------- Newsgroups: comp.os.vxworks Subject: Re: writing flash driver in vxworks Date: 9 Sep 2002 07:52:09 -0700 From: roswel_ajf@hotmail.com (roswell) Organization: http://groups.google.com/ Message-ID: <6dcae13d.0209090652.3c0d77a7@posting.google.com> References: <8b17c637.0209090105.243c015@posting.google.com> sjith2000@yahoo.com (Sreejith) wrote in message news:<8b17c637.0209090105.243c015@posting.google.com>... > hi everybody > I am a newbie in vxworks.We have to develop a flash driver for > the chip AM29F080B of entry level MBX860 embedded controller.We went > thru the manual of the chip.Anybody can suggest further manuals or > books which tells more about flash driver writing?Thanks in advance. > > regards > sreejith there is "cfiscs.c" file, a flash driver template. once u verify your write/erase routines are working, then, you can install the driver with vxworks. which will be used by tffs etc. for write, the manual will show u the flow chart. same for erase. generally, 1st put the operation cmd, then, data, then, done-like cmd. this various from op to op. the manual usuall talks about a single flash chip, but if u have few lined up, then, addressing is "interleaving". i.e 8bytes of data may live in 4 flash chips. so u have "program" all 4 chips with at the same time. and give them the data next. if u have a single chip then, no such worry. when erasing usually u erase a whole "unit". but u don't have to worry about saving data etc..tffs will handle once u give it the write/erase routines. hth --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Error messages Date: Mon, 9 Sep 2002 15:09:13 +0000 (UTC) From: Johan Borkhuis Organization: Proxim Corp. Message-ID: References: yahoo_m1@hotmail.com (John) wrote: > Hi All, > My system generated the following error message while I was trying to > run my code. > ----------------------------------------------------------------------- > ------- Exception at interrupt level: > > data access > Exception current instruction address: 0x0002c444 > Machine Status Register: 0x00009030 > Data Access Register: 0xeeeeeef2 > Condition Register: 0x22020040 > Data storage interrupt Register: 0x40000000 > Regs at 0x7e80000 > ----------------------------------------------------------------------- > ------- > > From this I understand that there was a data access exception. For > some reason the system tried to access the address 0xeeeeeef2 (which > is a invalid address). I also understand that the instruction which > caused this exception was at address 0x0002c444. But, I am not able to > decipher the other error messages like the > "Data Storage interrupt Register: 0x40000000", > "Regs at 0x7e80000" > etc. > > Can you share some information related to this as to how to decipher > this and use it for debugging. Basically, the information from the > current instruction address and the contents of Data access registers > were not enough for my debugging. Try to find out from which routine the exception was caused: look into the linker outpur, or do a "l 0x0002c444" from the shell. This can also be done the next time your target runs, as this information is static. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: how to get "Physical Address" directly from PCI enumeration? Date: Mon, 9 Sep 2002 17:04:27 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <200299-145744-758413@foorum.com> Reply-To: "Michael Lawnick" Sender: mlawnick@pd9524307.dip.t-dialin.net Hi charly, [...] > I would like to read the "Physical Address" directly from PCI enumeration. Is it > possible to get it from "base address 0 " or "base address n"? [...] > Description . . . . . . . . : 3Com 3C90x Ethernet Adapter > Physical Address. . . . . . : 00-C0-4F-89-DB-22 > DHCP Enabled. . . . . . . . : Yes [...] > base address 0 = 0x13711000 > base address 1 = 0x0000d401 [...] The Physical Address above is the MAC-Address on ethernet. It has nothing to do with the physical bus address on PCI (BAR0-BAR5). So to your question: No. Regards Michael --------------------------- Newsgroups: comp.os.vxworks Subject: tornado debugger taking lot of time to start Date: 9 Sep 2002 08:28:58 -0700 From: rajendra_krsingh@yahoo.com (rajendra) Organization: http://groups.google.com/ Message-ID: <479ba395.0209090728.18b8c04d@posting.google.com> Hi all, The tornado debugger is taking very long time to start. I am running target server with serial mode, 38400 baud rate. I observed one thing that memory taken by cgdbi86 is increasing slowly to 38MB and when it reaches that limit it gets started. What seems to me that, it is downloading symbols from the targets. But why it is downloading symbols , is there a way to ask debugger not to load symbol on start ? thanks rajendra --------------------------- Newsgroups: comp.os.vxworks Subject: ping problem on FEC port Vxworks 5.4 Date: Mon, 9 Sep 2002 19:09:19 +0300 From: "Martin Roth" Organization: Motorola Message-ID: When I ping the FEC port on ADS860 , I do not experience any problem with short fames. When the ping fame is large enough (60000 bytes) the FEC stops working. After I quit pinging, the FEC becomes operational after a period of ~20-30 seconds. Did anybody have such a problem on FEC port of MPC860T family ? Martin Roth Mailto:bmr006@email.mot.com --------------------------- Newsgroups: comp.os.vxworks Subject: Wtx Error 0x1001e (REGISTRY_NAME_NOT_FOUND) Date: 9 Sep 2002 09:29:17 -0700 From: rdevins@hotmail.com (Ronan Devins) Organization: http://groups.google.com/ Message-ID: Folks, Could anyone shed some light on what the above error is as there doesn't seem to be any documentation on the particular error (or alternatively point me in the direction of documentation that does explain the error). The error occurs after the following sequence of events: 1) Start target server 2) Start simulator (on Solaris machine) Then in a loop for multiple simulations { 3) Load executable 4) Run simulation 5) Unload executable 6) Reboot target server 7) Wtx Error 0x1001e (REGISTRY_NAME_NOT_FOUND) occurs } The error reported in step 7 does not occur every time thruogh the loop. Any help would be appreciated !!!! --------------------------- Newsgroups: comp.os.vxworks Subject: Problems with TFFS with DiskOnChip Date: 9 Sep 2002 11:33:56 -0700 From: intmain@yahoo.com (Bruce) Organization: http://groups.google.com/ Message-ID: <2b26d6bb.0209091033.58cda528@posting.google.com> We've recently discovered that occasionally, an fopen() or an fclose() with a DiskOnChip using the VxWorks TFFS is taking 600-900 mSecs instead of the normal 50-60 mSecs. Unfortunately, I have a 500mSec hardware watchdog that fires everytime this happens. Does anyone know the cause or have any suggestions? Currently, the best I can think of is to fopen() the file once and the flush() it everytime I write it and only close it on a shutdown. Any other ideas? Thanks, Bruce --------------------------- Newsgroups: comp.os.vxworks Subject: dumping static variables? Date: 9 Sep 2002 11:48:45 -0700 From: mahendrayk@hotmail.com (Mahendra) Organization: http://groups.google.com/ Message-ID: Hi, If a variable is declared as static, it is not visible from vxWorks session. I would like to know if there is a way to see its contents, like for eg. finding out its address and dumping the contents at that address? Thanks for your help, Mahendra --------------------------- Newsgroups: comp.os.vxworks Subject: "Not enough room for program headers" linker script error help! Date: 9 Sep 2002 12:50:19 -0700 From: jchen@terayon.com (john) Organization: http://groups.google.com/ Message-ID: Dear vxWorks developers: In my own linker script file (which is a version that I have taken from vxWorks) I have added a section: .data1 : { .mydatastart = .; PROVIDE (_data1Start = .); *(.data1) PROVIDE (_data1End = .); } . = . + 0x10000; in the end of my section, I wanted to add 0x10000 to it. However, when I compile this I get: vxWorks.st: Not enough room for program headers (allocated 2, need 3) ldppc: final link failed: Bad value make[1]: *** [vxWorks.st] Error 1 Funny part is, if I change my statement so that it adds only 0x1000 to the end of my section, then everything compiles fine. Any ideas? What is the size limitation? I have 128MB of RAM in our system, and that should be no problem..... thanks in advance. Please help.... john --------------------------- Newsgroups: comp.os.vxworks Subject: Re: AVL tree implementation Date: Mon, 9 Sep 2002 21:20:06 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Implementation should be pretty straight forward - copy from Knuth, you can use a scanner, but then you need to translate it from Pascal. Most pro's and con's are the same on VxWorks as everywhere else, except that it is a recursive algorythm, meaning you will need to make sure the stack is large enough for the worst case, and also would be a good idea to monitor stack space from within the code on every iteration. "Chahal, Dilbagh S (Dilbagh)" wrote in message news:alioid$4ms$1@overload.lbl.gov... > Can someone who has experimented with AVL tree implementation in vxWorks5.4, please comment on the strength and weaknesses of the implementation. > Thanks > -Dilbagh --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Problems with TFFS with DiskOnChip Date: Mon, 9 Sep 2002 21:23:06 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <2b26d6bb.0209091033.58cda528@posting.google.com> My best guess is that you are waiting for a sector Erase operation, which is about that range. Are you by chanse filling up the device to capacity ? If there is no free space left over, the TFFS code will need to make you wait until it makes space vacant by means of block reclamation, which involves an Erase. Your best solution would be to move the TFFS access to another task, so it would not block the critical task that handles that puppie... "Bruce" wrote in message news:2b26d6bb.0209091033.58cda528@posting.google.com... > We've recently discovered that occasionally, an fopen() or an fclose() > with a DiskOnChip using the VxWorks TFFS is taking 600-900 mSecs > instead of the normal 50-60 mSecs. Unfortunately, I have a 500mSec > hardware watchdog that fires everytime this happens. > > Does anyone know the cause or have any suggestions? Currently, the > best I can think of is to fopen() the file once and the flush() it > everytime I write it and only close it on a shutdown. > > Any other ideas? > > Thanks, > Bruce --------------------------- Newsgroups: comp.os.vxworks Subject: Access to the Software through Tornado IDE Date: 9 Sep 2002 13:28:10 -0700 From: robert_markey@hotmail.com (robert) Organization: http://groups.google.com/ Message-ID: <1fac8c75.0209091228.12bdaf3a@posting.google.com> Hello everybody, Our software running on the VxWorks. Maybe somebody know (some example) how I can get(create) access to our s/w through the Tornado IDE, either through a toolbar button or a menu option? Thanks a lot, Rob --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Problems with TFFS with DiskOnChip Date: Mon, 09 Sep 2002 21:18:45 GMT From: Andray Kaganovsky Organization: Futureway Message-ID: References: <2b26d6bb.0209091033.58cda528@posting.google.com> intmain@yahoo.com (Bruce) wrote in news:2b26d6bb.0209091033.58cda528 @posting.google.com: > We've recently discovered that occasionally, an fopen() or an fclose() > with a DiskOnChip using the VxWorks TFFS is taking 600-900 mSecs > instead of the normal 50-60 mSecs. Unfortunately, I have a 500mSec > hardware watchdog that fires everytime this happens. > > Does anyone know the cause or have any suggestions? Currently, the > best I can think of is to fopen() the file once and the flush() it > everytime I write it and only close it on a shutdown. > > Any other ideas? > > Thanks, > Bruce [AndrayK]. Try downloading and installing latest DiskOnChip driver from M-Systems' web site. The driver comes with README.TXT; take a look of chapter titled "Considerations related to I/O determinism". Hopefully this will solve your problem. regards, Andray --------------------------- Newsgroups: comp.os.vxworks Subject: sleep for nanoseconds without using POSIX Date: 9 Sep 2002 22:46:48 -0700 From: manjeetsc@yahoo.com (Manjeet Singh Chhabra) Organization: http://groups.google.com/ Message-ID: <8d0fbe67.0209092146.3674b6fb@posting.google.com> hi all, I have a problem i want to slepp a task for a few nanoseconds. but i can no use nanosleep because of restriction of not using POSIX methods. can anyone tell me how can i do this. Thanks Manjeet --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: 10 Sep 2002 02:09:17 -0400 From: "Paul D. Smith" Organization: Nortel Networks, Inc. / Billerica, MA Message-ID: References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> Reply-To: pausmith@nortelnetworks.com %% manjeetsc@yahoo.com (Manjeet Singh Chhabra) writes: msc> I have a problem i want to slepp a task for a few nanoseconds. msc> but i can no use nanosleep because of restriction of not using msc> POSIX methods. ??? You have a restriction that you _CAN'T_ use POSIX functions, no matter what? Bizarre. Why? - -- - ------------------------------------------------------------------------------- Paul D. Smith HASMAT--HA Software Mthds & Tools "Please remain calm...I may be mad, but I am a professional." --Mad Scientist - ------------------------------------------------------------------------------- These are my opinions---Nortel Networks takes no responsibility for them. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: Tue, 10 Sep 2002 07:24:28 GMT From: "Gary M" Organization: AT&T Broadband Message-ID: References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> How many is a "few"? Your processor type and clock speed have much to do with determining if you can do this at all. "Manjeet Singh Chhabra" wrote in message news:8d0fbe67.0209092146.3674b6fb@posting.google.com... > hi all, > I have a problem i want to slepp a task for a few nanoseconds. > but i can no use nanosleep because of restriction of not using POSIX methods. > > can anyone tell me how can i do this. > > Thanks > Manjeet --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: Tue, 10 Sep 2002 09:53:36 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> Look man, POSIX would have not helped you here a bit, because it's actual resolution still depends on the system clock interrupts, which come by default 60 times a second, and if needed could be stretched to a few thousands Hertz, but for nanosecond resolution it would require an interrupt rate no processor can sustain. Therefore, your onlu choise is a busy-loop delay, but for nanoseconds it would probably sufficie if you simply write a series of CPU instructions that have a deterministic execution time, either NOPs or accesses to an uncached memory location. "Manjeet Singh Chhabra" wrote in message news:8d0fbe67.0209092146.3674b6fb@posting.google.com... > hi all, > I have a problem i want to slepp a task for a few nanoseconds. > but i can no use nanosleep because of restriction of not using POSIX methods. > > can anyone tell me how can i do this. > > Thanks > Manjeet --------------------------- Newsgroups: comp.os.vxworks Subject: need help with interrupts a MPC8260 Date: Tue, 10 Sep 2002 10:18:41 +0200 From: "Achim Zimmer" Message-ID: Hi, I try to install in interrupt service function to a general IRQ line (IRQ5) of the MPC8260 processor. These are the code lines that should (in my opinion) do that: // IRQ5 handles all the chips interrupts // set SIEXR bit to determine which edges cause interrupt *M8260_SIEXR(regBase) |= 0x00000400; // we got an int at HIGH_TO_LOW change // connect interruptroutine status = intConnect (INUM_TO_IVEC(INUM_IRQ5), (VOIDFUNCPTR)globalIntHandler, 0); if (status == ERROR) printf("Error connect INT\n"); // set SIMR bit status = m8260IntEnable(INUM_IRQ5); if (status == ERROR) printf("Error enable INT\n"); I don't gt any errors with this functions, and at an Oszi I can see, that the interrupts physically occur, but my routine isn't raised. I seems to me, that i have missed something. Any help is welcome. Thanks Achim --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Access to the Software through Tornado IDE Date: 10 Sep 2002 01:58:31 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0209100058.7657970@posting.google.com> References: <1fac8c75.0209091228.12bdaf3a@posting.google.com> robert_markey@hotmail.com (robert) wrote in message news:<1fac8c75.0209091228.12bdaf3a@posting.google.com>... > Hello everybody, > Our software running on the VxWorks. > Maybe somebody know (some example) how I can get(create) access to our > s/w through the Tornado IDE, either through a toolbar button or a menu > option? I've never done this myself, but this is something which could be done using the Tornado API and WDB/WTX. There is a Tornado API Programers guide, which covers this (basically, this comes down to extending the Tornado TCL code by some WDB/WTX calls... and doing the same on vxWorks) toni --------------------------- Newsgroups: comp.os.vxworks Subject: Re: dumping static variables? Date: 10 Sep 2002 02:12:28 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0209100112.1cfa57a7@posting.google.com> References: mahendrayk@hotmail.com (Mahendra) wrote in message news:... > If a variable is declared as static, it is not visible from vxWorks > session. Do an nmxyz on the image (elf, a-out, ...) and extract the address, you may dump the content of it later, e.g.: #> nmppc vxWorks | grep ftpsActive 0023dc60 d ftpsActive #> toni --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ping problem on FEC port Vxworks 5.4 Date: 10 Sep 2002 02:16:15 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0209100116.2d615e40@posting.google.com> References: "Martin Roth" wrote in message news:... > When I ping the FEC port on ADS860 , I do not experience any problem with > short fames. When the ping fame is large enough (60000 bytes) the FEC stops > working. After I quit pinging, the FEC becomes operational after a period of > ~20-30 seconds. > Did anybody have such a problem on FEC port of MPC860T family ? This problem is inherited by the IP Fragment Reassembly mechanism. Basically, what you're doing comes down to being a DoS attack on a Stack with very limited ressources. You may either increase the ressources (Clusters available to the END device) or decrease the time the reassembly Q is waiting to reassemble fragments (ipfragttl in SLOW_HZ ticks... 1/2 secs). Tornado 2.2 Network Stack uses a different mechanism, this particular problem is gone completely. I'd recomend upgrading. toni --------------------------- Newsgroups: comp.os.vxworks Subject: Re: how to get "Physical Address" directly from PCI enumeration? Date: 10 Sep 2002 02:18:02 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0209100118.7e012db3@posting.google.com> References: <200299-145744-758413@foorum.com> > The Physical Address above is the MAC-Address on ethernet. It has nothing to > do with the physical bus address on PCI (BAR0-BAR5). correct, imho you're looking for something like this: #include #include #include int getmac( char *name, int nr ) { END_OBJ * pEnd; unsigned char mac[6]; pEnd= endFindByName( name, nr ); if ( !pEnd ) { printf("NIC %s%i not found!\n",name,nr); return( -1 ); } muxIoctl(pEnd, EIOCGADDR, mac); printf("MAC Address of %s%i is %02x:%02x:%02x:%02x:%02x:%02x\n", name, nr, mac[0], mac[1], mac[2], mac[3], mac[4], mac[5], mac[6] ); return( 0 ); } toni --------------------------- Newsgroups: comp.os.vxworks Subject: Re: tornado debugger taking lot of time to start Date: 10 Sep 2002 02:53:48 -0700 From: heiko_elger@arburg.com (Heiko Elger) Organization: http://groups.google.com/ Message-ID: <97e6274e.0209100153.58870c48@posting.google.com> References: <479ba395.0209090728.18b8c04d@posting.google.com> Hi rajendra, perhaps it's then known problem: SPR 25918: When loading the symbol table the memory consumption is large!! Solution: set symbol-readnow off (gdb command) Place the gdb command in a file named .vxgdbinit (!! with leading period) The file is located in the WIND_BASE directory i.e. c:\tornado rajendra_krsingh@yahoo.com (rajendra) wrote in message news:<479ba395.0209090728.18b8c04d@posting.google.com>... > Hi all, > The tornado debugger is taking very long time to start. I am running > target server with serial mode, 38400 baud rate. I observed one thing > that > memory taken by cgdbi86 is increasing slowly to 38MB and when it > reaches > that limit it gets started. > What seems to me that, it is downloading symbols from the targets. But > why it is downloading symbols , is there a way to ask debugger not to > load symbol on start ? > > thanks > rajendra --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Flash Invalidate Date: 10 Sep 2002 03:16:43 -0700 From: yahoo_m1@hotmail.com (John) Organization: http://groups.google.com/ Message-ID: References: Leo, What you are saying seems to be happening in my case. That is, by invalidating the data cache, i feel i am loosing some recent modifications and the function from where I called the invalidate does not seem to return back properly (could be because it lost track of where to go back). But, I am in a situation where in I cannot afford to do a cache flush since I want to preserve the data that is present in the memory and discard the data that is present in the cache (at least for some of the cache lines). Any pointers, or any suggestions are highly appreciated. Thanks in Advance, John. "Leonid Rosenboim" wrote in message news:... > By invaliadtiong data cache, you can sometimes loose some > of the recent modifications, i.e. if your cache is in write-back > mode, there could be some "dirty" cache lines that where not > yet written to memory, which invalidation will erase. > > If you want to clear the entire cache, the only safe option > to use is "flush & Invalidate". > > Cache invalidate can be used on a specific address range which has > been give to a DMA (aka Bus Master) device for fetching data, and since > the device will subsequently write into main memory at this range, > whatever value was stored in cache need to be invalidated. > > "John" wrote in message > news:baa5411a.0209080512.116502db@posting.google.com... > > Hi All, > > I am using MPC750 processor on a custom board and trying to do a Flash > > Invalidation of the data cache, by setting the DCFI bit in the HID0 > > register. But, I am finding some unusual results. After the > > invalidation, either the board completely hangs, or it reboots with a > > workQPanic message or even if the board is alive, i cannot access it > > from the Shell. And, these results are not consistent. I have the > > interrupts locked and also calling my invalidation routine from the > > task context. > > > > Any comments on this will be highly appreciated. > > > > Thanks in Advance, > > John --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: 10 Sep 2002 03:44:00 -0700 From: manjeetsc@yahoo.com (Manjeet Singh Chhabra) Organization: http://groups.google.com/ Message-ID: <8d0fbe67.0209100243.5f9fe38f@posting.google.com> References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> "Paul D. Smith" wrote in message news:... > %% manjeetsc@yahoo.com (Manjeet Singh Chhabra) writes: > > msc> I have a problem i want to slepp a task for a few nanoseconds. > msc> but i can no use nanosleep because of restriction of not using > msc> POSIX methods. > > ??? > > You have a restriction that you _CAN'T_ use POSIX functions, no matter > what? > > Bizarre. Why? that is because out system has such a restriction do you have any idea how to do this ? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Flash Invalidate Date: Tue, 10 Sep 2002 13:27:22 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Well, I do not understand why these changes in memory are more important then any changes in the cache, but here are two possible solutions: #1 - When you invalidate cache, you need to specify an address range, so that the invalidation is only for any cache lines that contain data within that range of virtual addresses. Note that for this to work, the region being invalidated needs to be aligned to cache line boundary (usually 16 bytes). #2 - Assuming this memory data comes from an I/O bus master device, (aka DMA), you can actually do flush_invalidate BEFORE the DMA device begins the operation, and make sure your code does not touch the DMA buffers until the I/O operation is complete. Hope this helps, - Leonid "John" wrote in message news:baa5411a.0209100216.577115b7@posting.google.com... > Leo, > What you are saying seems to be happening in my case. That is, by > invalidating the data cache, i feel i am loosing some recent > modifications and the function from where I called the invalidate does > not seem to return back properly (could be because it lost track of > where to go back). But, I am in a situation where in I cannot afford > to do a cache flush since I want to preserve the data that is present > in the memory and discard the data that is present in the cache (at > least for some of the cache lines). > > Any pointers, or any suggestions are highly appreciated. > > Thanks in Advance, > John. > > > "Leonid Rosenboim" wrote in message news:... > > By invaliadtiong data cache, you can sometimes loose some > > of the recent modifications, i.e. if your cache is in write-back > > mode, there could be some "dirty" cache lines that where not > > yet written to memory, which invalidation will erase. > > > > If you want to clear the entire cache, the only safe option > > to use is "flush & Invalidate". > > > > Cache invalidate can be used on a specific address range which has > > been give to a DMA (aka Bus Master) device for fetching data, and since > > the device will subsequently write into main memory at this range, > > whatever value was stored in cache need to be invalidated. > > > > "John" wrote in message > > news:baa5411a.0209080512.116502db@posting.google.com... > > > Hi All, > > > I am using MPC750 processor on a custom board and trying to do a Flash > > > Invalidation of the data cache, by setting the DCFI bit in the HID0 > > > register. But, I am finding some unusual results. After the > > > invalidation, either the board completely hangs, or it reboots with a > > > workQPanic message or even if the board is alive, i cannot access it > > > from the Shell. And, these results are not consistent. I have the > > > interrupts locked and also calling my invalidation routine from the > > > task context. > > > > > > Any comments on this will be highly appreciated. > > > > > > Thanks in Advance, > > > John --------------------------- Newsgroups: comp.os.vxworks Subject: Curious Newbie Date: 10 Sep 2002 05:56:13 -0700 From: gtaylor@lowebroadway.com (Garry Taylor) Organization: http://groups.google.com/ Message-ID: Hi, Please forgive my ignorance about embedded systems :-) I'm quite interested in getting a little single board computer, maybe based on the 1Ghz MIPS chips with two cores from PMC-Sierra, as this mentions VxWorks as supported. A few questions: Does anyone have any experience of this chip? A 1Ghz MIPS is bound to be pretty fast, even for Java, can anyone confirm? When creating a user interface on a VxWorks system, is simply a matter of getting an LCD panel which is supported by the SBC and it'll work? Can anyone advise on how much these boards cost? Where to buy in quantities of one? Any pointers to Dummie's guides or FAQs would be great. Thanks Garry --------------------------- Newsgroups: comp.os.vxworks Subject: problem on ftp client , command freeze on PWD Date: 10 Sep 2002 06:29:33 -0700 From: ilyasg@yahoo.com (Ilyas G) Organization: http://groups.google.com/ Message-ID: Hi all I am trying to make a ftp client over Rapid control (on vxworks) I have problem with the pwd and cd command for both the command I need to get the curent directory. so i use the following command if (ftpXfer (ftp_session.host, ftp_session.login, ftp_session.password, "", "PWD", ftp_session.directory, NULL, &ctrlSock, &dataSock) == ERROR) return (ERROR); while ((nBytes = read (dataSock, buf, sizeof (buf))) > 0) { printf("Current directory = %s",buf); } However , for the PWD command the ftp server returns the curenbtdirectory, however, it's never returned on the dataSock . Does any one used this command ? Thanks --------------------------- Newsgroups: comp.os.vxworks Subject: tickGet is too slow ... anything else i can use? Date: Tue, 10 Sep 2002 13:42:21 GMT From: "Jim" Organization: Magma Communications Ltd. Message-ID: <1Fmf9.11243$V21.263765@news> Hi, I know this is a pretty basic question ... but I need to find a function to time a control loop ... to get a ball park runtime on it ... tickLib is too slow ... taskDelay(60) gets us 1 second ... I can't change the number from 60 to anything else because to much code will need to be changed to get results ... Is there anyother library/function I can use ? Thnx, Jim --------------------------- Newsgroups: comp.os.vxworks Subject: Re: problem on ftp client , command freeze on PWD Date: Tue, 10 Sep 2002 15:45:44 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Read the ftpXfer() manual more carefully, and read a bit about FTP protocol itself. Not all FTP commands create a separate data connection, for example, PUT, GET, LIST, NLST - create a data conenction PWD, CWD, CDUP, ... - output their result on the control connection. If what you really need is the current directory on the server, read it from the control socket ! "Ilyas G" wrote in message news:afdb26d2.0209100529.15505438@posting.google.com... > Hi all > I am trying to make a ftp client over Rapid control (on vxworks) > I have problem with the pwd and cd command > > for both the command I need to get the curent directory. > so i use the following command > > if (ftpXfer (ftp_session.host, ftp_session.login, > ftp_session.password, "", > "PWD", ftp_session.directory, NULL, > &ctrlSock, &dataSock) == ERROR) > return (ERROR); > > while ((nBytes = read (dataSock, buf, sizeof (buf))) > 0) > { > printf("Current directory = %s",buf); > } > > > However , for the PWD command the ftp server returns the > curenbtdirectory, however, it's never returned on the dataSock . > Does any one used this command ? > > Thanks --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: 10 Sep 2002 09:55:34 -0400 From: "Paul D. Smith" Organization: Nortel Networks, Inc. / Billerica, MA Message-ID: References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> <8d0fbe67.0209100243.5f9fe38f@posting.google.com> Reply-To: pausmith@nortelnetworks.com %% manjeetsc@yahoo.com (Manjeet Singh Chhabra) writes: >> %% manjeetsc@yahoo.com (Manjeet Singh Chhabra) writes: msc> I have a problem i want to slepp a task for a few nanoseconds. msc> but i can no use nanosleep because of restriction of not using msc> POSIX methods. >> You have a restriction that you _CAN'T_ use POSIX functions, no matter >> what? >> >> Bizarre. Why? msc> that is because out system has such a restriction Yes, I understood that from your original post. The question is "why?" Do you also have a rule that you cannot use any ISO C standard functions, like memcpy(), etc.? Note that the POSIX standard includes (by reference) all the functions in the ISO C standard, so technically if you're not allowed to use any POSIX functions you can't use ISO C functions (malloc, free, strcpy, strcmp, memcpy, memcmp, assert, etc. etc. etc.), either. msc> do you have any idea how to do this ? As others have pointed out there is no way to sleep for less than the clock tick size of your platform (which you have neglected to mention)--the tick size is certainly going to be larger than "a few nanoseconds" unless you're working on very strange hardware indeed! :). A PPC, for example, uses a tick size of 8 ms (IIRC). If you want to pause for "a few nanoseconds" and that is a hard limit (you need it to be pretty accurate) then you have no choice but to spin in a busy loop. Luckily for you, I don't think there's any POSIX function which provides this capability. I'm interested to know what the VxWorks function that does a calibrated busy delay loop is, if someone knows it: on QNX for example there is nanospin()--very nice. - -- - ------------------------------------------------------------------------------- Paul D. Smith HASMAT--HA Software Mthds & Tools "Please remain calm...I may be mad, but I am a professional." --Mad Scientist - ------------------------------------------------------------------------------- These are my opinions---Nortel Networks takes no responsibility for them. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: 10 Sep 2002 07:16:44 -0700 From: cpotenza@onetel.net.uk (Claudio Potenza) Organization: http://groups.google.com/ Message-ID: <2025ac82.0209100616.37eaf267@posting.google.com> References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> > I have a problem i want to slepp a task for a few nanoseconds. > but i can no use nanosleep because of restriction of not using > POSIX methods. Even if you were allowed (?!) to use nanosleep (what a strange negative requirement: "don't use POSIX"!!), it would not be of any help, I believe. As the reference manual says: The suspension may be longer than requested due to the rounding up of the request to the timer's resolution... That is, "nanosleep()" has the granularity of the system clock rate. This is by "default" 60 ticks per second. You can check the exact value by calling "sysClkRateGet()". Said another way, nanosleep('1 nanosecond') is exactly like taskDelay(1) and delays your task anywhere between 0 and 1/60 of a second --------------------------- Newsgroups: comp.os.vxworks Subject: Re: AVL tree implementation Date: 10 Sep 2002 07:21:16 -0700 From: cpotenza@onetel.net.uk (Claudio Potenza) Organization: http://groups.google.com/ Message-ID: <2025ac82.0209100621.5f285c78@posting.google.com> References: "Chahal, Dilbagh S (Dilbagh)" wrote in message news:... > Can someone who has experimented with AVL tree implementation in > vxWorks5.4, please comment on the strength and weaknesses of the > implementation. I never used this library myself (yet), but it seemed intersting when I bokmarked the link: http://www.msu.edu/user/pfaffben/avl/ --------------------------- Newsgroups: comp.os.vxworks Subject: Coldfire 5272 BSP Serial Problem Date: 10 Sep 2002 07:48:17 -0700 From: pcda@lightlink.com (Paul C. Davis) Organization: http://groups.google.com/ Message-ID: <35306a13.0209100648.3a09c032@posting.google.com> Hi: Looking for someone with experience with the MCF5272C3 Coldfire BSP or helpful hints. I can't get "/tyCo/1" to function. "/tyCo/0" works fine. NUMTTY is set to 2. Rest of the BSP seems fine. The hardware scope's out back to the processor and there aren't any shunts in the way. I'm sure I'm going to feel silly! Thanks in advance --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Coldfire 5272 BSP Serial Problem Date: Tue, 10 Sep 2002 15:28:52 +0000 (UTC) From: Johan Borkhuis Organization: Proxim Corp. Message-ID: References: <35306a13.0209100648.3a09c032@posting.google.com> pcda@lightlink.com (Paul C. Davis) wrote: > Looking for someone with experience with the MCF5272C3 Coldfire BSP > or helpful hints. I can't get "/tyCo/1" to function. "/tyCo/0" works > fine. NUMTTY is set to 2. Rest of the BSP seems fine. The hardware > scope's out back to the processor and there aren't any shunts in the > way. Does your BSP support 2 TTY's? I have been looking at a similar problem, to find out that my BSP only supported 1 serial port. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: AMD Athlon XP Date: 10 Sep 2002 08:41:33 -0700 From: egoisone@libero.it (Mic) Organization: http://groups.google.com/ Message-ID: Hi! Here again...last time probably I did not explain what's my problem. Be patient. Is there anybody who can boot VxWorks on AMD Athlon XP with a current available BSP? If not, does Pentium/PentiumPro can work with some changes anyway? If possible in few words, what kind of these changes are to be done? My floppy bootloader (VxLd 1.2) prints out about three rows of dots and then stops. What does VxWorks try to do at that moment (and it fails)? Thank you again, Mic. --------------------------- Newsgroups: comp.os.vxworks Subject: host shell does not resume Date: Tue, 10 Sep 2002 15:37:05 GMT From: "Chad" Organization: Shaw Residential Internet Message-ID: My console shell does not function after a target server connects, but does up until that point. Is there a setting that will produce this behavior? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: tickGet is too slow ... anything else i can use? Date: Tue, 10 Sep 2002 11:39:25 -0400 From: Joe Durusau Organization: Lockheed Martin Corporation Message-ID: <3D7E122D.6614D5F@lmco.com> References: <1Fmf9.11243$V21.263765@news> Jim wrote: > > Hi, > > I know this is a pretty basic question ... but I need to find a function to > time a control loop ... to get a ball park runtime on it ... tickLib is too > slow ... taskDelay(60) gets us 1 second ... I can't change the number from > 60 to anything else because to much code will need to be changed to get > results ... > > Is there anyother library/function I can use ? > > Thnx, > Jim You can time a function with timex or timexN, if that would help. See the man pages for how to use them. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Re: vxWorks compilation-time errors Date: 10 Sep 2002 08:51:25 -0700 From: egoisone@libero.it (Mic) Organization: http://groups.google.com/ Message-ID: References: Hi, you have to pay attention to both Tornado and code (config.h and so on into your BPS dir) configurations. I mean that there are some inclusion that need to be done two times: the former into the workspace, the latter directly within the code. Try to get some docs or read the code. Bye, M. "Denis Dubrovin" wrote in message news:... > Hi! > > WinNT\Tornado 2\Diab RTA Suite > > When I compile vxWorks image ( vxWorks.rom ) "as is" - in wizard > configuration - all OK. > > But! > > 1. If I include "Diab RTA support" component I get error: > ================================== > F:\Tornado\host\x86-win32\bin\binToAsm tmp.Z > vxWorks.Z.s > dcc -c -g -tPPC823FS:tornado -D__asm__=__asm -Xansi -Xstrings-in-text=0 -ew1 > 551 -Xintrinsi > c-mask=0x3 -c -Xpreprocess-assembly -IC:\USERS\DEN\vxWorks\vxTest -IF:\Torna > do\target\conf > ig\wrMdpBasic8xxdiab -IF:\Tornado\target\h -IF:\Tornado\target\config\comps\ > src -IF:\Torna > do\target\src\drv -DCPU=PPC860 -DPRJ_BUILD -g -O vxWorks.Z.s -o vxWorks.Z.o > "vxWorks.Z.s", line 35643: error: syntax error > make: *** [vxWorks_rom] Error 0x1 > ================================== > > 2. If I include "spy" component I get another error: > ===================================== > dld -tPPC823FS:tornado -X -e _sysInit \ > dataSegPad.o partialImage.o ctdt.o -o vxWorks \ > ld_low.lcf > dld: warning: Undefined symbol 'sysAuxClkDisable' in file 'partialImage.o' > dld: warning: Undefined symbol 'sysAuxClkRateGet' in file 'partialImage.o' > dld: warning: Undefined symbol 'sysAuxClkRateSet' in file 'partialImage.o' > dld: warning: Undefined symbol 'sysAuxClkEnable' in file 'partialImage.o' > dld: warning: Undefined symbol 'sysAuxClkConnect' in file 'partialImage.o' > dld: Undefined symbols found - no output written > make: *** [vxWorks] Error 0x1 > ===================================== > > How to make it working? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: AMD Athlon XP Date: Tue, 10 Sep 2002 18:11:41 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: When VxLd prints these dots, it is reading the boot file from the floppy (or hard disk) using BIOS calls, with teh destination belng the low memory. The dots are being printed with BIOS too, and there is not much details about what this does because this code sits in the boot block of the boot disk, occupying only a fraction of a 512 byte sector. When the dots are all printed, the processor will simply jump to a pre-defined address where it should begin executing the initial instructions of romInit.s assembly file from the BSP. Because this is an x86, the first portion of romInit.s is a real-mode code code, which initializes a dummy GDT and IDT and jumps into "protected mode", and then there are many initializations that would take place before any other output on the screen would be emitted. Hence, it would be quite impossible to tell where does this exactly fail, without adding some additional output indicators in that initialization code, a.k.a. debugging! HTH, - - Leonid "Mic" wrote in message news:c6b19a0b.0209100741.8de6a9e@posting.google.com... > Hi! Here again...last time probably I did not explain what's my > problem. Be patient. Is there anybody who can boot VxWorks on AMD > Athlon XP with a current available BSP? If not, does > Pentium/PentiumPro can work with some changes anyway? If possible in > few words, what kind of these changes are to be done? > My floppy bootloader (VxLd 1.2) prints out about three rows of dots > and then stops. What does VxWorks try to do at that moment (and it > fails)? > > Thank you again, Mic. --------------------------- Newsgroups: comp.os.vxworks Subject: printf causing WTX Errors. Date: 10 Sep 2002 09:14:44 -0700 From: rdevins@hotmail.com (Ronan Devins) Organization: http://groups.google.com/ Message-ID: Hi, An assertion in my code (a macro, wrappping a printf and assert) is causing my target server to hang. The tests are being run from a perl script - this starts the target server and simulator, establishes the connection between them, then loads the executable and runs it (the load and run are in a loop) and finally kills the target server and simulator. I have run the test suite sucessfully. However, inserting a deliberate assert causes the target server to hang. When the assertion happens, the printf text is printed to the screen, then nothing happens. When I do a "ps" from another shell, I see all my processes running (target server, simulator, windshell). Hitting ^C give me 2 WTX errors: 0x1019c - EXCHANGE_TRANSPORT_DISCONNECT 0x1001e - REGISTRY_NAME_NOT_FOUND The output (including assert printf output) looks as follows: ASSERT failed: test.c:342 ^CWTX Error 0x1019c (EXCHANGE_TRANSPORT_DISCONNECT) WTX Error 0x1001e (REGISTRY_NAME_NOT_FOUND) Has anyone had any previous experience of WTX Errors when using printf ? Does this cause the windshell to lose communication with the target server ? Any help on this would be appreciated. Thanks, Ronan. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Wtx Error 0x1001e (REGISTRY_NAME_NOT_FOUND) Date: 10 Sep 2002 09:25:04 -0700 From: egoisone@libero.it (Mic) Organization: http://groups.google.com/ Message-ID: References: Hi, I try a tip: what about timing? Could your loop be too fast to allow WTX to resolve registry name? I have no more guesswork, good luck! rdevins@hotmail.com (Ronan Devins) wrote in message news:... > Folks, > Could anyone shed some light on what the above error is as there > doesn't seem to be any documentation on the particular error (or > alternatively point me in the direction of documentation that does > explain the error). > > The error occurs after the following sequence of events: > > 1) Start target server > 2) Start simulator (on Solaris machine) > > Then in a loop for multiple simulations > { > 3) Load executable > 4) Run simulation > 5) Unload executable > 6) Reboot target server > 7) Wtx Error 0x1001e (REGISTRY_NAME_NOT_FOUND) occurs > } > > The error reported in step 7 does not occur every time thruogh the > loop. > > Any help would be appreciated !!!! --------------------------- Newsgroups: comp.os.vxworks Subject: Re: "Not enough room for program headers" linker script error help! Date: 10 Sep 2002 09:44:57 -0700 From: jchen@terayon.com (john) Organization: http://groups.google.com/ Message-ID: References: sorry. I forgot to include that I am using Powerpc 8260 w/ est8260 BSP on a custom board. thanks jchen@terayon.com (john) wrote in message news:... > Dear vxWorks developers: > > In my own linker script file (which is a version that I have taken > from vxWorks) I have added a section: > > .data1 : > { > .mydatastart = .; > PROVIDE (_data1Start = .); > *(.data1) > PROVIDE (_data1End = .); > } > > . = . + 0x10000; > > in the end of my section, I wanted to add 0x10000 to it. However, > when I compile this I get: > > vxWorks.st: Not enough room for program headers (allocated 2, need 3) > ldppc: final link failed: Bad value > make[1]: *** [vxWorks.st] Error 1 > > Funny part is, if I change my statement so that it adds only 0x1000 to > the end of my section, then everything compiles fine. > > Any ideas? What is the size limitation? I have 128MB of RAM in our > system, and that should be no problem..... > > thanks in advance. Please help.... > > john --------------------------- Newsgroups: comp.os.vxworks Subject: Re: need help with interrupts a MPC8260 Date: Tue, 10 Sep 2002 10:19:00 -0700 From: "Dan Gold" Organization: Mailgate.ORG Server - http://www.Mailgate.ORG Message-ID: References: Sender: gold@ensemble.com "Achim Zimmer" wrote in message news:alk9t4$1qgr3n$1@ID-102608.news.dfncis.de... > I try to install in interrupt service function to a general IRQ line(IRQ5) > of the MPC8260 processor. > // IRQ5 handles all the chips interrupts > // set SIEXR bit to determine which edges cause interrupt > *M8260_SIEXR(regBase) |= 0x00000400> > status = intConnect (INUM_TO_IVEC(INUM_IRQ5), (VOIDFUNCPTR)globalIntHandler, 0); > if (status == ERROR) printf("Error connect INT\n"); > // set SIMR bit > status = m8260IntEnable(INUM_IRQ5); > if (status == ERROR) printf("Error enable INT\n"); > I don't gt any errors with this functions, and at an Oszi I can see, that > the interrupts physically occur, but my routine isn't raised. > I seems to me, that i have missed something. Do you have SIUMCR[DPPC] set correctly ? It should be 00 to allow IRQ5. If the BSP code still has it set to 10 for the ancient TBEN errata, you should be able to change it now (assuming you're running an 8260 greater than A.0 :-)) hth, Dan Gold gold@ensemble.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: Tue, 10 Sep 2002 10:55:10 -0700 From: "Brett Wilson" Organization: Motorola Message-ID: References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> <8d0fbe67.0209100243.5f9fe38f@posting.google.com> I also would like to know what the VxWorks function is to do a calibrated busy loop or how to sleep for less than the system tick time which is 1ms in our system (which is less than what WindRiver recommends and we would like to go lower!). Brett Wilson DSP Software Engineer General Dynamics "Paul D. Smith" wrote in message news:p5wupux82x.fsf@lemming.engeast.baynetworks.com... > %% manjeetsc@yahoo.com (Manjeet Singh Chhabra) writes: > > >> %% manjeetsc@yahoo.com (Manjeet Singh Chhabra) writes: > > msc> I have a problem i want to slepp a task for a few nanoseconds. > msc> but i can no use nanosleep because of restriction of not using > msc> POSIX methods. > > >> You have a restriction that you _CAN'T_ use POSIX functions, no matter > >> what? > >> > >> Bizarre. Why? > > msc> that is because out system has such a restriction > > Yes, I understood that from your original post. The question is "why?" > > Do you also have a rule that you cannot use any ISO C standard > functions, like memcpy(), etc.? Note that the POSIX standard includes > (by reference) all the functions in the ISO C standard, so technically > if you're not allowed to use any POSIX functions you can't use ISO C > functions (malloc, free, strcpy, strcmp, memcpy, memcmp, assert, > etc. etc. etc.), either. > > msc> do you have any idea how to do this ? > > As others have pointed out there is no way to sleep for less than the > clock tick size of your platform (which you have neglected to > mention)--the tick size is certainly going to be larger than "a few > nanoseconds" unless you're working on very strange hardware indeed! :). > A PPC, for example, uses a tick size of 8 ms (IIRC). > > If you want to pause for "a few nanoseconds" and that is a hard limit > (you need it to be pretty accurate) then you have no choice but to spin > in a busy loop. > > Luckily for you, I don't think there's any POSIX function which provides > this capability. > > I'm interested to know what the VxWorks function that does a calibrated > busy delay loop is, if someone knows it: on QNX for example there is > nanospin()--very nice. > > -- > -------------------------------------------------------------------------- - ----- > Paul D. Smith HASMAT--HA Software Mthds & Tools > "Please remain calm...I may be mad, but I am a professional." --Mad Scientist > -------------------------------------------------------------------------- - ----- > These are my opinions---Nortel Networks takes no responsibility for them. --------------------------- Newsgroups: comp.os.vxworks Subject: Strange Interrupt/Rescheduling behavior on PowerPC Date: 10 Sep 2002 11:21:29 -0700 From: sslartybart@hotmail.com (Sim Nanda) Organization: http://groups.google.com/ Message-ID: <31135f9a.0209101021.5cde867d@posting.google.com> Hello, I have a task (called tUartSend) that loads up a messageQ and turns on the txInterrupt on a UART. The tx ISR fires whenever there is room in the FIFO. All the ISR does is unload the messageQ (with NO_WAIT of course), load up the UART, and exit. This setup works most of the time, but I'm seeing a strange lockup condition that occurs intermittently. I've used the debugger to look at the stack trace when the tUartSend is locked up. It looks like this: Line 1: 0x19a85c in intLock() <--- This is where I'm locked up Line 2: 0x2216ac in reschedule() Line 3: 0x34f00 in XgUart::EnableTxInterrupt(this=0x6e6830) Line 4: 0x3102c in XgUart::Write(this=0x6e6830, pBuf=0x6ef324, numWords=1) Line 5: 0xfe198 in sendToUart(pBuf = 0x6ef324, numWords=1) Line X: ..... other previous routines. The stack trace looks normal to me up to line 3. But I don't understand lines 1 and 2. I didn't call reschedule(). It must've been called by vxWorks, but why is it in tUartSend's context? More importantly why is tUartSend hung in the intLock() routine? Looking at the Program Counter tUartSend is locked up on the instruction: 0x19a85c: mtmsr r4 (where r4 contains 0x1200). The rest of the tasks are running fine, it's just the tUartSend task that is frozen inside intLock()! mtmsr shouldn't block or anything. Other relevant details: - tUartSend routines don't touch the MSR or call any VxWorks routines besides msgQSend() - Always running in Privileged mode. - I've checked the task's stack and high water mark, plenty of room. - There is a network task (tNetTask) running in the background. - All other tasks including the tShell task are running fine. My hunch is that a context switch ocurred at the same time as when the interrupt was firing, or maybe there were mulitple interrupts. My Questions for anyone able to help are: Has anyone seen this before? Is there an ISR stack that's possibly being corrupted by too nested interrupts? How could I examine the ISR stack? Any suggestions on how to debug this? regards, Sim Nanda email: sslartybart hotmail com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: problem on ftp client , command freeze on PWD Date: Tue, 10 Sep 2002 20:51:07 +0200 From: ilyas G Organization: Wanadoo, l'internet avec France Telecom Message-ID: <3D7E3F1B.9020109@no-spam.guennoun.org> References: Yes you are right, the RFC say that the result is send in the control socket however, there is no way to read the control socket just after the transfer. unless i send the command on the socket. i will try write (ctrlData,"PWD", 3); read (ctrlData,buff ...); I will try it tomorrow, but does any one tried that ???, thanks ? Leonid Rosenboim wrote: > Read the ftpXfer() manual more carefully, and read a bit about FTP protocol > itself. > > Not all FTP commands create a separate data connection, for example, > PUT, GET, LIST, NLST - create a data conenction > PWD, CWD, CDUP, ... - output their result on the control connection. > > If what you really need is the current directory on the server, > read it from the control socket ! > > > "Ilyas G" wrote in message > news:afdb26d2.0209100529.15505438@posting.google.com... > >>Hi all >>I am trying to make a ftp client over Rapid control (on vxworks) >>I have problem with the pwd and cd command >> >>for both the command I need to get the curent directory. >>so i use the following command >> >> if (ftpXfer (ftp_session.host, ftp_session.login, >>ftp_session.password, "", >> "PWD", ftp_session.directory, NULL, >> &ctrlSock, &dataSock) == ERROR) >> return (ERROR); >> >> while ((nBytes = read (dataSock, buf, sizeof (buf))) > 0) >> { >> printf("Current directory = %s",buf); >> } >> >> >>However , for the PWD command the ftp server returns the >>curenbtdirectory, however, it's never returned on the dataSock . >>Does any one used this command ? >> >>Thanks > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: problem on ftp client , command freeze on PWD Date: Tue, 10 Sep 2002 20:51:21 +0200 From: ilyas G Organization: Wanadoo, l'internet avec France Telecom Message-ID: <3D7E3F29.8010508@no-spam.guennoun.org> References: Yes you are right, the RFC say that the result is send in the control socket however, there is no way to read the control socket just after the transfer. unless i send the command on the socket. i will try write (ctrlData,"PWD", 3); read (ctrlData,buff ...); I will try it tomorrow, but does any one tried that ???, thanks ? Leonid Rosenboim wrote: > Read the ftpXfer() manual more carefully, and read a bit about FTP protocol > itself. > > Not all FTP commands create a separate data connection, for example, > PUT, GET, LIST, NLST - create a data conenction > PWD, CWD, CDUP, ... - output their result on the control connection. > > If what you really need is the current directory on the server, > read it from the control socket ! > > > "Ilyas G" wrote in message > news:afdb26d2.0209100529.15505438@posting.google.com... > >>Hi all >>I am trying to make a ftp client over Rapid control (on vxworks) >>I have problem with the pwd and cd command >> >>for both the command I need to get the curent directory. >>so i use the following command >> >> if (ftpXfer (ftp_session.host, ftp_session.login, >>ftp_session.password, "", >> "PWD", ftp_session.directory, NULL, >> &ctrlSock, &dataSock) == ERROR) >> return (ERROR); >> >> while ((nBytes = read (dataSock, buf, sizeof (buf))) > 0) >> { >> printf("Current directory = %s",buf); >> } >> >> >>However , for the PWD command the ftp server returns the >>curenbtdirectory, however, it's never returned on the dataSock . >>Does any one used this command ? >> >>Thanks > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: problem on ftp client , command freeze on PWD Date: Tue, 10 Sep 2002 20:53:02 +0200 From: ilyas G Organization: Wanadoo, l'internet avec France Telecom Message-ID: <3D7E3F8E.7040707@no-spam.guennoun.org> References: Yes you are right, the RFC say that the result is send in the control socket however, there is no way to read the control socket just after the transfer. unless i send the command on the socket. i will try write (ctrlData,"PWD", 3); read (ctrlData,buff ...); I will try it tomorrow, but does any one tried that ???, thanks ? Leonid Rosenboim wrote: > Read the ftpXfer() manual more carefully, and read a bit about FTP protocol > itself. > > Not all FTP commands create a separate data connection, for example, > PUT, GET, LIST, NLST - create a data conenction > PWD, CWD, CDUP, ... - output their result on the control connection. > > If what you really need is the current directory on the server, > read it from the control socket ! > > > "Ilyas G" wrote in message > news:afdb26d2.0209100529.15505438@posting.google.com... > >>Hi all >>I am trying to make a ftp client over Rapid control (on vxworks) >>I have problem with the pwd and cd command >> >>for both the command I need to get the curent directory. >>so i use the following command >> >> if (ftpXfer (ftp_session.host, ftp_session.login, >>ftp_session.password, "", >> "PWD", ftp_session.directory, NULL, >> &ctrlSock, &dataSock) == ERROR) >> return (ERROR); >> >> while ((nBytes = read (dataSock, buf, sizeof (buf))) > 0) >> { >> printf("Current directory = %s",buf); >> } >> >> >>However , for the PWD command the ftp server returns the >>curenbtdirectory, however, it's never returned on the dataSock . >>Does any one used this command ? >> >>Thanks > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: problem on ftp client , command freeze on PWD Date: Tue, 10 Sep 2002 20:53:51 +0200 From: ily G Organization: Wanadoo, l'internet avec France Telecom Message-ID: <3D7E3FBF.9080800@netscape.net> References: Yes you are right, the RFC say that the result is send in the control socket however, there is no way to read the control socket just after the transfer. unless i send the command on the socket. i will try write (ctrlData,"PWD", 3); read (ctrlData,buff ...); I will try it tomorrow, but does any one tried that ???, thanks ? Leonid Rosenboim wrote: > Read the ftpXfer() manual more carefully, and read a bit about FTP protocol > itself. > > Not all FTP commands create a separate data connection, for example, > PUT, GET, LIST, NLST - create a data conenction > PWD, CWD, CDUP, ... - output their result on the control connection. > > If what you really need is the current directory on the server, > read it from the control socket ! > > > "Ilyas G" wrote in message > news:afdb26d2.0209100529.15505438@posting.google.com... > >>Hi all >>I am trying to make a ftp client over Rapid control (on vxworks) >>I have problem with the pwd and cd command >> >>for both the command I need to get the curent directory. >>so i use the following command >> >> if (ftpXfer (ftp_session.host, ftp_session.login, >>ftp_session.password, "", >> "PWD", ftp_session.directory, NULL, >> &ctrlSock, &dataSock) == ERROR) >> return (ERROR); >> >> while ((nBytes = read (dataSock, buf, sizeof (buf))) > 0) >> { >> printf("Current directory = %s",buf); >> } >> >> >>However , for the PWD command the ftp server returns the >>curenbtdirectory, however, it's never returned on the dataSock . >>Does any one used this command ? >> >>Thanks > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Coldfire 5272 BSP Serial Problem -- I am silly!! Date: 10 Sep 2002 14:01:22 -0700 From: pcda@lightlink.com (Paul C. Davis) Organization: http://groups.google.com/ Message-ID: <35306a13.0209101301.3ff3533@posting.google.com> References: <35306a13.0209100648.3a09c032@posting.google.com> Answer to my own question: The answer is configuration bit settings in the PDCNT register. The transmitter and receiver are not being enabled by the BSP. I got the BDM to do it and all is well. Now about that BSP...... pcda@lightlink.com (Paul C. Davis) wrote in message news:<35306a13.0209100648.3a09c032@posting.google.com>... > Hi: > > Looking for someone with experience with the MCF5272C3 Coldfire BSP > or helpful hints. I can't get "/tyCo/1" to function. "/tyCo/0" works > fine. NUMTTY is set to 2. Rest of the BSP seems fine. The hardware > scope's out back to the processor and there aren't any shunts in the > way. > > I'm sure I'm going to feel silly! > > Thanks in advance --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: Tue, 10 Sep 2002 15:33:56 -0700 From: "Behzad Abareshi" Organization: NOAO Message-ID: References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> > I have a problem i want to slepp a task for a few nanoseconds. Tell us why you want a few nanosecs delay in the first place; you may be pleasantly surprised that what you think you need is often times very different from what you really need! Talking from my own experience, of course :-) --------------------------- Newsgroups: comp.os.vxworks Subject: Re: dumping static variables? Date: Tue, 10 Sep 2002 15:45:09 -0700 From: "Behzad Abareshi" Organization: NOAO Message-ID: References: > If a variable is declared as static, it is not visible from vxWorks > session. > That's the default behavior of ld (ld 0). You can add static/local variables to the symbol table by using ld 1 (see man ld). Behzad. --------------------------- Newsgroups: comp.os.vxworks Subject: UDP recvfrom problem Date: 10 Sep 2002 16:23:27 -0700 From: pepsi_000@hotmail.com (vxJunkie) Organization: http://groups.google.com/ Message-ID: <676eb679.0209101523.bc3bf14@posting.google.com> hi folks, I have a problem with recvfrom(UDP sockets) on my PPC-8240 board with a broadcom ethernet driver. When i send a packet say 3000 bytes in size the board receives only 1472 bytes. I am not sure what happens to the rest of the packet. ipStats shows that 1 packet was reassembled. Everytime i send this packet the reassemble count in the ipStats increases. I did not see any error counts for fragment dropped/timeouts etc. Any ideas how the driver can affect this? I feel the driver cannot differentiate between multiple IP fragments. I tried the same experiment with a PPC-855t board running a motfec driver, and was able to receive 3000 bytes. Any help is appreciated. Thanks, kumar --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: Tue, 10 Sep 2002 19:44:49 -0400 From: joe durusau Organization: WEBUSENET.com Message-ID: <3D7E83F1.A7FA77E6@bellsouth.net> References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> <8d0fbe67.0209100243.5f9fe38f@posting.google.com> Well, you could write a short function, then time it with timexN(), to see how long it takes on the average. Note that this is not really a matter of the task sleeping, it's a matter of the task marking time and burning CPU time for some short minimum. You can get hardware timers that will send you an int after a given time delay, (the shortest I've ever heard of is 1 uS), but there is no way that you are going to get a repeatable and calibrated time of say 10 nS unless your task is the only one active at the time and interrupts are disabled. Note that the hardware timers I mentioned will give you a _minimum_ delay, not a calibrated _exact_ delay. If you use a busy loop, remember that it will work only for one specific platform and that specific version of the OS. I suggest that you probably really don't need this kind of thing in software. There are people out there who will be able to fairly easily give you a device that will output one or several signals in a very precise time sequence, up to seconds and down to nS long, and do it reliably. It all depends on exactly what you are trying to do, but this is efinitely a job for hardware amd not for software. Speaking only for myself, Joe Durusau Brett Wilson wrote: > I also would like to know what the VxWorks function is to do a calibrated > busy loop or how to sleep for less than the system tick time which is 1ms in > our system (which is less than what WindRiver recommends and we would like > to go lower!). > > Brett Wilson > DSP Software Engineer > General Dynamics > > "Paul D. Smith" wrote in message > news:p5wupux82x.fsf@lemming.engeast.baynetworks.com... > > %% manjeetsc@yahoo.com (Manjeet Singh Chhabra) writes: > > > > >> %% manjeetsc@yahoo.com (Manjeet Singh Chhabra) writes: > > > > msc> I have a problem i want to slepp a task for a few nanoseconds. > > msc> but i can no use nanosleep because of restriction of not using > > msc> POSIX methods. > > > > >> You have a restriction that you _CAN'T_ use POSIX functions, no > matter > > >> what? > > >> > > >> Bizarre. Why? > > > > msc> that is because out system has such a restriction > > > > Yes, I understood that from your original post. The question is "why?" > > > > Do you also have a rule that you cannot use any ISO C standard > > functions, like memcpy(), etc.? Note that the POSIX standard includes > > (by reference) all the functions in the ISO C standard, so technically > > if you're not allowed to use any POSIX functions you can't use ISO C > > functions (malloc, free, strcpy, strcmp, memcpy, memcmp, assert, > > etc. etc. etc.), either. > > > > msc> do you have any idea how to do this ? > > > > As others have pointed out there is no way to sleep for less than the > > clock tick size of your platform (which you have neglected to > > mention)--the tick size is certainly going to be larger than "a few > > nanoseconds" unless you're working on very strange hardware indeed! :). > > A PPC, for example, uses a tick size of 8 ms (IIRC). > > > > If you want to pause for "a few nanoseconds" and that is a hard limit > > (you need it to be pretty accurate) then you have no choice but to spin > > in a busy loop. > > > > Luckily for you, I don't think there's any POSIX function which provides > > this capability. > > > > I'm interested to know what the VxWorks function that does a calibrated > > busy delay loop is, if someone knows it: on QNX for example there is > > nanospin()--very nice. > > > > -- > > -------------------------------------------------------------------------- > ----- > > Paul D. Smith HASMAT--HA Software Mthds & > Tools > > "Please remain calm...I may be mad, but I am a professional." --Mad > Scientist > > -------------------------------------------------------------------------- > ----- > > These are my opinions---Nortel Networks takes no responsibility for > them. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: Wed, 11 Sep 2002 00:56:05 GMT From: "H. E. Taylor" Organization: MTS Internet Message-ID: <3D7EAF04.2E3B@despam.autobahn.mb.ca> References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> In article , Leonid Rosenboim wrote: >[...] > > Therefore, your onlu choise is a busy-loop delay, but for nanoseconds > it would probably sufficie if you simply write a series of CPU instructions > that have a deterministic execution time, either NOPs or accesses to > an uncached memory location. > Manjeet did not mention what CPU he is using, but if it is an x86 or PPC he can use an assembly routine to access the clock tick counter. You need to calibrate this counter initially, but the technique works fine. - -het - -- "progress in software has not followed Moore's law." -John Holland Energy Alternatives: http://www.autobahn.mb.ca/~het/energy.html H.E. Taylor http://www.autobahn.mb.ca/~het/ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: 10 Sep 2002 21:09:14 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> <8d0fbe67.0209100243.5f9fe38f@posting.google.com> <3D7E83F1.A7FA77E6@bellsouth.net> Sender: bpringle@DeadDuck Joe> Well, you could write a short function, then time it with Joe> timexN(), to see how long it takes on the average. Note that Joe> this is not really a matter of the task sleeping, it's a matter [snip] Alternatively, you could look at the vxWorks FAQ. I also posted some code, that was modified from the FAQ entry, in the last few months. hth, Bill Pringlemeir. - -- ANNOYED BY CORVETTE. DEPTH CHARGES. EXPLODING. U-135. --------------------------- Newsgroups: comp.os.vxworks Subject: make me understand this plz. Date: 10 Sep 2002 22:17:05 -0700 From: vxnerd@rediffmail.com (subbarayan) Organization: http://groups.google.com/ Message-ID: dear mentors, Can anyone help me understand the vxworks clean image?i mean what it does and where its exactly used and why it should be used etc?I tried to search for some info abt it in vxworks help files and manuals but of no use. Advanced thanks to persons helping me in this regards. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: tickGet is too slow ... anything else i can use? Date: Wed, 11 Sep 2002 05:19:13 GMT From: "Gary M" Organization: AT&T Broadband Message-ID: References: <1Fmf9.11243$V21.263765@news> We could probably provide more assistance if you let us know what the processor type is and the system clock speed (or the make/model of the eval board you are using). "Jim" wrote in message news:1Fmf9.11243$V21.263765@news... > Hi, > > I know this is a pretty basic question ... but I need to find a function to > time a control loop ... to get a ball park runtime on it ... tickLib is too > slow ... taskDelay(60) gets us 1 second ... I can't change the number from > 60 to anything else because to much code will need to be changed to get > results ... > > Is there anyother library/function I can use ? > > Thnx, > Jim > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: need help with interrupts a MPC8260 Date: Wed, 11 Sep 2002 07:50:55 +0200 From: "Achim Zimmer" Message-ID: References: Thank you very much, this was the right help. And one thing more: No where in the descriptions of the tornado intlib was written, that I have to acknoledge the INT with writing the SIPNR register. But after doing that first in the ISR it runs fine. Thanks Dan "Dan Gold" schrieb im Newsbeitrag news:all9l0$17k$1@newsreader.mailgate.org... > "Achim Zimmer" wrote in message > news:alk9t4$1qgr3n$1@ID-102608.news.dfncis.de... > > I try to install in interrupt service function to a general IRQ line(IRQ5) > > of the MPC8260 processor. > > // IRQ5 handles all the chips interrupts > > // set SIEXR bit to determine which edges cause interrupt > > *M8260_SIEXR(regBase) |= 0x00000400> > > status = intConnect (INUM_TO_IVEC(INUM_IRQ5), > (VOIDFUNCPTR)globalIntHandler, 0); > > if (status == ERROR) printf("Error connect INT\n"); > > // set SIMR bit > > status = m8260IntEnable(INUM_IRQ5); > > if (status == ERROR) printf("Error enable INT\n"); > > > I don't gt any errors with this functions, and at an Oszi I can see, that > > the interrupts physically occur, but my routine isn't raised. > > I seems to me, that i have missed something. > > Do you have SIUMCR[DPPC] set correctly ? It should be 00 to allow IRQ5. If > the BSP code still has it set to 10 for the ancient TBEN errata, you should > be able to change it now (assuming you're running an 8260 greater than A.0 > :-)) > > > hth, > Dan Gold gold@ensemble.com > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: AMD Athlon XP Date: Wed, 11 Sep 2002 07:44:31 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: Reply-To: "Michael Lawnick" Sender: mlawnick@p5088ef94.dip.t-dialin.net Hi Mic, last time you posted this, I asked you whether INCLUDE_PC_CONSOLE is set in your bootrom-image - you didn't answer. Michael "Mic" schrieb im Newsbeitrag news:c6b19a0b.0209100741.8de6a9e@posting.google.com... > Hi! Here again...last time probably I did not explain what's my > problem. Be patient. Is there anybody who can boot VxWorks on AMD > Athlon XP with a current available BSP? If not, does > Pentium/PentiumPro can work with some changes anyway? If possible in > few words, what kind of these changes are to be done? > My floppy bootloader (VxLd 1.2) prints out about three rows of dots > and then stops. What does VxWorks try to do at that moment (and it > fails)? > > Thank you again, Mic. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: tickGet is too slow ... anything else i can use? Date: 10 Sep 2002 23:31:37 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0209102231.72478601@posting.google.com> References: <1Fmf9.11243$V21.263765@news> Hello, Try looking at the timestamp driver routines in your timer/clock driver (look in target/src/drv/timer for the source code to these drivers, and check your BSP to see which one you are using). If your driver supports it, then something like INCLUDE_TIMESTAMP should be enough to include it in the kernel - the resolution is usually much, much better than the clock tick since it is just reading the underlying timer's count register. HTH, John... "Jim" wrote in message news:<1Fmf9.11243$V21.263765@news>... > Hi, > > I know this is a pretty basic question ... but I need to find a function to > time a control loop ... to get a ball park runtime on it ... tickLib is too > slow ... taskDelay(60) gets us 1 second ... I can't change the number from > 60 to anything else because to much code will need to be changed to get > results ... > > Is there anyother library/function I can use ? > > Thnx, > Jim --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Strange Interrupt/Rescheduling behavior on PowerPC Date: 10 Sep 2002 23:42:57 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0209102242.1be31250@posting.google.com> References: <31135f9a.0209101021.5cde867d@posting.google.com> Hello, I think, assuming I have read your message correctly, that the problem is simple: msgQReceive() is never usable from an ISR, irrespective of the timeout. The msgQSend() routine is usable with a NO_WAIT timeout, but the receive routine is not. Check out table 2-23 (page 87) in the VxWorks Programmer's Guide for the routines that are safe to call from an ISR. Only msgQSend() is listed for msgQLib. HTH, John... sslartybart@hotmail.com (Sim Nanda) wrote in message news:<31135f9a.0209101021.5cde867d@posting.google.com>... > Hello, > > I have a task (called tUartSend) that loads up a messageQ and turns on the > txInterrupt on a UART. The tx ISR fires whenever > there is room in the FIFO. All the ISR does is unload the messageQ > (with NO_WAIT of course), load up the UART, and exit. > > This setup works most of the time, but I'm seeing a strange > lockup condition that occurs intermittently. I've used the > debugger to look at the stack trace when the tUartSend is locked up. > It looks like this: > > Line 1: 0x19a85c in intLock() <--- This is where I'm locked up > Line 2: 0x2216ac in reschedule() > Line 3: 0x34f00 in XgUart::EnableTxInterrupt(this=0x6e6830) > Line 4: 0x3102c in XgUart::Write(this=0x6e6830, pBuf=0x6ef324, numWords=1) > Line 5: 0xfe198 in sendToUart(pBuf = 0x6ef324, numWords=1) > Line X: ..... other previous routines. > > The stack trace looks normal to me up to line 3. > But I don't understand lines 1 and 2. > I didn't call reschedule(). > It must've been called by vxWorks, but why is it in > tUartSend's context? More importantly why is > tUartSend hung in the intLock() routine? > > Looking at the Program Counter tUartSend is locked up on the instruction: > 0x19a85c: mtmsr r4 (where r4 contains 0x1200). > The rest of the tasks are running fine, it's just the tUartSend task that > is frozen inside intLock()! mtmsr shouldn't block or anything. > > Other relevant details: > - tUartSend routines don't touch the MSR or call any VxWorks routines > besides msgQSend() > - Always running in Privileged mode. > - I've checked the task's stack and high water mark, plenty of room. > - There is a network task (tNetTask) running in the background. > - All other tasks including the tShell task are running fine. > > My hunch is that a context switch ocurred at the same time as when > the interrupt was firing, or maybe there were mulitple interrupts. > > My Questions for anyone able to help are: > Has anyone seen this before? > Is there an ISR stack that's possibly being corrupted by too nested interrupts? > How could I examine the ISR stack? > Any suggestions on how to debug this? > > regards, > Sim Nanda > email: sslartybart hotmail com --------------------------- Newsgroups: comp.os.vxworks Subject: How do I set the tNetTaks stack size Date: Wed, 11 Sep 2002 11:07:57 +0300 From: "Barak Hermesh" Organization: Texas Instruments Message-ID: I have a stack overflow problem on vxWorks' tNetTask so I want it to be created with a larger stack. Is there an elegant way to do this ? Thanks, Barak. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How do I set the tNetTaks stack size Date: Wed, 11 Sep 2002 10:18:38 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: This is realy simple: The global variable netTaskStackSize is initialized statically to the default stack size. You can change it anywhere in the intialization code (usrNetwork.c or usrConfig.c or even sysLib.c) but make sure you do it BEOFRE the netLibInit() function is called (in usrNetwork.c). Gmar Chatima Tova! - - Leonid "Barak Hermesh" wrote in message news:almtl3$shl$1@tilde.itg.ti.com... > I have a stack overflow problem on vxWorks' tNetTask so I want it to be > created with a larger stack. > Is there an elegant way to do this ? > > Thanks, > Barak. > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: how to get "Physical Address" directly from PCI enumeration? Date: 11 Sep 2002 01:39:10 -0700 From: pdufrene.other@free.fr (Pierre D.) Organization: http://groups.google.com/ Message-ID: <8726dedf.0209110039.273c1ad2@posting.google.com> References: <200299-145744-758413@foorum.com> <8400a733.0209100118.7e012db3@posting.google.com> Thanks for your answer but this network adapter is not used by VxWorks. I want to use the MAC address to identify the computer, but I don't need any other network funtionnalities. So my question was how to get the MAC address only by PCI ennumeration and by doing some dump from memory/IO/... Charly > correct, imho you're looking for something like this: > > #include > #include > #include > > int getmac( char *name, int nr ) > { > END_OBJ * pEnd; > unsigned char mac[6]; > > pEnd= endFindByName( name, nr ); > if ( !pEnd ) > { > printf("NIC %s%i not found!\n",name,nr); > return( -1 ); > } > > muxIoctl(pEnd, EIOCGADDR, mac); > > printf("MAC Address of %s%i is %02x:%02x:%02x:%02x:%02x:%02x\n", > name, nr, mac[0], mac[1], mac[2], mac[3], mac[4], mac[5], mac[6] ); > > return( 0 ); > } > > toni --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How do I set the tNetTaks stack size Date: Wed, 11 Sep 2002 12:24:16 +0300 From: "Barak Hermesh" Organization: Texas Instruments Message-ID: References: Thanks a lot and Shana Tova. "Leonid Rosenboim" wrote in message news:newscache$cvl92h$oao$1@lnews.actcom.co.il... > This is realy simple: > > The global variable > > netTaskStackSize > > is initialized statically to the default stack size. You can change it > anywhere in the intialization code (usrNetwork.c or usrConfig.c or even > sysLib.c) but make sure you do it BEOFRE the netLibInit() function is called > (in usrNetwork.c). > > Gmar Chatima Tova! > - Leonid > > "Barak Hermesh" wrote in message > news:almtl3$shl$1@tilde.itg.ti.com... > > I have a stack overflow problem on vxWorks' tNetTask so I want it to be > > created with a larger stack. > > Is there an elegant way to do this ? > > > > Thanks, > > Barak. > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: AMD Athlon XP Date: Wed, 11 Sep 2002 09:34:31 GMT From: "Michael B." Organization: [Infostrada] Message-ID: References: Sorry Michael, once I saw INCLUDE_PCI_CONSOLE already included and the problem still I forgot your question and I tried something new...what drives me crazy is that the same bootloader (in other words the same floopy) works on a similar computer without any changes, but that PC is a Pentium IV! Thank you for your interest, Mic. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: make me understand this plz. Date: Wed, 11 Sep 2002 06:30:46 -0400 From: joe durusau Organization: WEBUSENET.com Message-ID: <3D7F1B56.5EEB8DEA@bellsouth.net> References: subbarayan wrote: > dear mentors, > Can anyone help me understand the vxworks clean image?i mean what it > does and where its exactly used and why it should be used etc?I tried > to search for some info abt it in vxworks help files and manuals but > of no use. > Advanced thanks to persons helping me in this regards. There is no such image. Traditionally in Makefiles, the target clean is used to delete everything that make itself has/may have created, in order to return the directory to its 'standard ' starting condition. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Re: AMD Athlon XP Date: Wed, 11 Sep 2002 12:30:50 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: Reply-To: "Michael Lawnick" Sender: mlawnick@p5088ef94.dip.t-dialin.net Hi Mic, so as Leonid suggested you will have to debug. A hint: reduce NUM_TTY by one and use the freed tty (aka COM-port) for polled output. Instrument your bootConfig.c with "a","b",... and see where it hangs. Good luck, Michael "Michael B." schrieb im Newsbeitrag news:H6Ef9.11653$6O4.198049@twister1.libero.it... > Sorry Michael, once I saw INCLUDE_PCI_CONSOLE already included and the > problem still I forgot your question and I tried something new...what drives > me crazy is that the same bootloader (in other words the same floopy) works > on a similar computer without any changes, but that PC is a Pentium IV! > Thank you for your interest, Mic. > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: how to get "Physical Address" directly from PCI enumeration? Date: Wed, 11 Sep 2002 12:36:06 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <200299-145744-758413@foorum.com> <8400a733.0209100118.7e012db3@posting.google.com> <8726dedf.0209110039.273c1ad2@posting.google.com> Reply-To: "Michael Lawnick" Sender: mlawnick@p5088ef94.dip.t-dialin.net Hi Pierre, this depends on your NIC but is (AFAIK) typically not possible. MAC is mostly saved in NIC's SRAM that is not memory mapped into the PCI address space. You will have to look into the documentation of your specific card. Regards, Michael "Pierre D." schrieb im Newsbeitrag news:8726dedf.0209110039.273c1ad2@posting.google.com... > Thanks for your answer but this network adapter is not used by > VxWorks. I want to use the MAC address to identify the computer, but I > don't need any other network funtionnalities. So my question was how > to get the MAC address only by PCI ennumeration and by doing some dump > from memory/IO/... > > Charly > > > correct, imho you're looking for something like this: > > > > #include > > #include > > #include > > > > int getmac( char *name, int nr ) > > { > > END_OBJ * pEnd; > > unsigned char mac[6]; > > > > pEnd= endFindByName( name, nr ); > > if ( !pEnd ) > > { > > printf("NIC %s%i not found!\n",name,nr); > > return( -1 ); > > } > > > > muxIoctl(pEnd, EIOCGADDR, mac); > > > > printf("MAC Address of %s%i is %02x:%02x:%02x:%02x:%02x:%02x\n", > > name, nr, mac[0], mac[1], mac[2], mac[3], mac[4], mac[5], mac[6] ); > > > > return( 0 ); > > } > > > > toni --------------------------- Newsgroups: comp.os.vxworks Subject: ANN: PragmaDev Real Time Developer Studion V1.2 Date: Wed, 11 Sep 2002 12:51:10 +0200 From: "Emmanuel Gaudin" Organization: PragmaDev Message-ID: Reply-To: "Emmanuel Gaudin" Dear Sir / Madame, PragmaDev announces Real Time Developer Studio V1.2, the latest release of the first commercial off the shelf tool based on SDL-RT graphical language to develop real time and embedded software. Real Time Developer Studio V1.2 is a major breakthrough in SDL-RT technology: - - Execution and graphical debug of SDL-RT systems is now possible on Linux, Solaris, and Windows hosts without any third party tool. - - Following partnership with Tasking, Real Time Developer Studio is now integrated with C166/ST10 Tasking Cross View Pro debugger and CMX RTX RTOS allowing graphical debugging at SDL-RT level of cross compiled code on host and target. More information can be found on the following web sites: http://www.pragmadev.com http://www.sdl-rt.org Regards, Emmanuel Gaudin --------------------------- Newsgroups: comp.os.vxworks Subject: Re: host shell does not resume Date: Wed, 11 Sep 2002 03:52:37 -0700 From: "Behzad Abareshi" Organization: NOAO Message-ID: References: "Chad" wrote in message news:Bkof9.287064$v53.15113569@news3.calgary.shaw.ca... > My console shell does not function after a target server connects, but does > up until that point. Is there a setting that will produce this behavior? > > You can have only one active shell at a time; you regain control over your shell after the other client disconnects. Behzad. --------------------------- Newsgroups: comp.os.vxworks Subject: Support for /etc/services & set./getservent system calls.. Date: 11 Sep 2002 04:38:20 -0700 From: vijay.peshkar@wipro.com (Vijay Peshkar) Organization: http://groups.google.com/ Message-ID: Dear VxWorkers, Am porting BIND to VxWorks. This code inherited from my linux team is the original BIND without resolver library code (was told resolver in already part of linux, so no need of resolver code). The code has lot of references to set/getservent() and getservbyname() system calls which use /etc/services. Now as the resolver of VXWorks does not address these features, am in a dilemma if I have to port the complete resolver lib too and use the same instead of VxWorks resolver. Couldn't find any satisfying answers in the archives and hence this post. Any suggestion would be helpful. Thanks and Regards, Vijay --------------------------- Newsgroups: comp.os.vxworks Subject: Re: tickGet is too slow ... anything else i can use? Date: Wed, 11 Sep 2002 11:49:09 GMT From: "Jim" Organization: Magma Communications Ltd. Message-ID: References: <1Fmf9.11243$V21.263765@news> Thanks all, I found what I needed on line here it is: #define TIMEBASE_HZ (5000000)/*This is specific to my 860 config*/ #define TIMEBASE_PERIOD (1.0 / TIMEBASE_HZ) double double_time(void) { unsigned tbu, tbl; vxTimeBaseGet(&tbu, &tbl); /* Get 64-bit value */ return (tbu * 4294967296.0 + tbl) * TIMEBASE_PERIOD; } extern "C" void my_code() { taskDelay(60); } extern "C" void benchmark(void) { double ts, te; ts = double_time(); my_code(); te = double_time(); printf("Elapsed time = %f sec\n", te - ts); printf(" = %f usec\n", (te - ts) * 1.0e6); } double BenchMarkGetStartTime(){ return double_time(); } double BenchMarkGetEndTime(){ return double_time(); } "Jim" wrote in message news:1Fmf9.11243$V21.263765@news... > Hi, > > I know this is a pretty basic question ... but I need to find a function to > time a control loop ... to get a ball park runtime on it ... tickLib is too > slow ... taskDelay(60) gets us 1 second ... I can't change the number from > 60 to anything else because to much code will need to be changed to get > results ... > > Is there anyother library/function I can use ? > > Thnx, > Jim > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks 5.4.2 on Motorola Power PC 82xx Date: 11 Sep 2002 05:26:52 -0700 From: gabi@tdsoft.com (Gabi) Organization: http://groups.google.com/ Message-ID: <5607ff2c.0209110426.1912c63@posting.google.com> References: <66f277de.0209061350.50ac78b0@posting.google.com> <5947a04.0209061954.67b8e472@posting.google.com> andreas_lindell75@hotmail.com (Rikard) wrote in message news:<5947a04.0209061954.67b8e472@posting.google.com>... > vtriplic@yahoo.com (vt) wrote in message news:<66f277de.0209061350.50ac78b0@posting.google.com>... > > We are planning to use VxWorks on Motorola PPC 82xx. I searched this > > group for any user feedbacks, but I couldn't find much. Does anybody > > use it? Any issues that we need to be aware of up front? Any fedback > > is welcome. > > Thanks > > vt > > Hello VT, > > I have used a WRS SBC8240 board with no problem. WR have reference > designs for various of different PPC platforms. You can check them out > on WR webpage. If you buy one of those together with vxworks its a > "out of the box" solution and you can start your development with no > problem. If you go for a custom design with a 82xx CPU then its a > different thing. Then you will need a BSP that supports your board and > those can be expensive about 20000usd, not sure though and up to 2 > months development depending on what peripherals you are using. > > Check out this link: > http://www.windriver.com/products/html/reference_designs.html > > However I was using VxWorks 5.4, didnt even know that there was a > 5.4.2 version? > > Regards > Andreas VT Hi, I made several BSP for MPC8260 under VxWorks on different custom boards, and I'm using Aisys (DriveWay) code generation for it. This will short your development period to one/two month for someone familiar with VxWorks and Tornado environment. Gabi --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: 11 Sep 2002 05:46:04 -0700 From: manjeetsc@yahoo.com (Manjeet Singh Chhabra) Organization: http://groups.google.com/ Message-ID: <8d0fbe67.0209110446.22a56e50@posting.google.com> References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> "Behzad Abareshi" wrote in message news:... > > I have a problem i want to slepp a task for a few nanoseconds. > > Tell us why you want a few nanosecs delay in the first place; you may be > pleasantly surprised that what you think you need is often times very > different from what you really need! Talking from my own experience, of > course :-) Actually we have a router. Running on a COM-CPU ( i don't know hardware Architeture ) and i am writing a device driver for Galnet cross bar. what driver does is it receives packets from the cross bar and sends these packets to upper layers. my driver is a task which is polling all the 8 ports of the cross bar. in eache iteration it looks for packets and send at max 8 packets to upperlayers. and sleeps for 1 tick ( which is processing time for 8 packets ). to allow other task to process those packets. but the problem is that if there are less no of packets still i have to sleep for processing time equivalent to 8 packets because i don't have any machnism to slepp for less than one tick. How can i do this ? Manjeet --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: Wed, 11 Sep 2002 13:30:00 +0000 (UTC) From: David Laight Organization: BT Openworld Message-ID: References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> <8d0fbe67.0209110446.22a56e50@posting.google.com> > > Actually we have a router. > Running on a COM-CPU ( i don't know hardware Architeture ) > and i am writing a device driver for Galnet cross bar. > > what driver does is it receives packets from the cross bar and sends > these packets to upper layers. > > my driver is a task which is polling all the 8 ports of the cross bar. > > in eache iteration it looks for packets and send at max 8 packets to > upperlayers. and sleeps for 1 tick ( which is processing time for 8 > packets ). > > to allow other task to process those packets. > > but the problem is that if there are less no of packets still i have > to sleep for processing time equivalent to 8 packets because i don't > have any machnism to slepp for less than one tick. Seems to me you should be playing with process priorities, not timers. David --------------------------- Newsgroups: comp.os.vxworks Subject: VxWorks error : WTX Error 0x10140 (API_FILE_NOT_ACCESSIBLE) Date: Wed, 11 Sep 2002 16:35:55 +0200 From: "tomer" Organization: The Israel Inter-University Computation Center Message-ID: hi, I try to load an object file and this is what I get on the shell screen. what can I do to successfully load the file? thanks rotem! --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Strange Interrupt/Rescheduling behavior on PowerPC Date: 11 Sep 2002 09:06:19 -0700 From: sslartybart@hotmail.com (Sim Nanda) Organization: http://groups.google.com/ Message-ID: <31135f9a.0209110806.6a3e18ef@posting.google.com> References: <31135f9a.0209101021.5cde867d@posting.google.com> <488e459a.0209102242.1be31250@posting.google.com> Thanks for the reply John. Yes you're right, I'm not allowed to do a msgQReceive at interrupt level. This particular piece of code was converted from pSOS. pSOS does allow loading and unloading msgQ's at interrupt level. I completely missed this difference between the OS's. Boy, is my face red! bye, sim john_94501@yahoo.com (John) wrote in message news:<488e459a.0209102242.1be31250@posting.google.com>... > Hello, > > I think, assuming I have read your message correctly, that the problem > is simple: msgQReceive() is never usable from an ISR, irrespective of > the timeout. The msgQSend() routine is usable with a NO_WAIT timeout, > but the receive routine is not. > > Check out table 2-23 (page 87) in the VxWorks Programmer's Guide for > the routines that are safe to call from an ISR. Only msgQSend() is > listed for msgQLib. > > HTH, > > John... > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: 11 Sep 2002 11:58:23 -0400 From: "Paul D. Smith" Organization: Nortel Networks, Inc. / Billerica, MA Message-ID: References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> <8d0fbe67.0209100243.5f9fe38f@posting.google.com> <3D7E83F1.A7FA77E6@bellsouth.net> Reply-To: pausmith@nortelnetworks.com > Alternatively, you could look at the vxWorks FAQ. Umm, which one? Google returns a bunch of them. I looked in the "Unofficial FAQ" and the FAQ for this newsgroup and didn't see anything. The "VxWorks/Tornado II FAQ" seemed to have some timer-related stuff, but I didn't really see what I wanted. I realize that busy loops are to be avoided in general, but sometimes you have to have them. And hardware is not always the answer. On QNX they have a very nice setup: you can run functions like: int nanospin( const struct timespec *when ); to do a busy-wait. The first time you call this function it will do a calibration to determine how long to spin on this system (you can also force calibration, say early in your program, by directly calling nanospin_calibrate()). - -- - ------------------------------------------------------------------------------- Paul D. Smith HASMAT--HA Software Mthds & Tools "Please remain calm...I may be mad, but I am a professional." --Mad Scientist - ------------------------------------------------------------------------------- These are my opinions---Nortel Networks takes no responsibility for them. --------------------------- Newsgroups: comp.os.vxworks Subject: watchdog Timer Date: 11 Sep 2002 09:30:43 -0700 From: mynamvs@nsc.network.com (Venkat Mynampati) Organization: http://groups.google.com/ Message-ID: <65b2abfc.0209110830.29485f36@posting.google.com> Hi, In my system, i have a watchdog timer that, rearms itself every 16.67msec and in the associated function, all i does is to write(increment) to a memory location over pci bus. Very rarely, what;s hapening is that, this memory location is not getting incremented and the cpu finds itself in idleLoop forever. Can you help me in understanding how this watchdog timer works... also, > What will happen if the system has locked out interrupts via intLock(), and the watchdog timer has expired? this is a real scenario in my system, where we lock interrupts, to process them! > Can the watchdog timer, after having expired a long time back be still active, once the system has found itself in idleLoop? > Is it posible that the system is not receiving any events at all? Thanks for your insights and help. Regards, Venkat --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: 11 Sep 2002 12:17:25 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> <8d0fbe67.0209100243.5f9fe38f@posting.google.com> <3D7E83F1.A7FA77E6@bellsouth.net> Sender: bpringle@DeadDuck > Alternatively, you could look at the vxWorks FAQ. Paul> Umm, which one? Google returns a bunch of them. I looked in Paul> the "Unofficial FAQ" and the FAQ for this newsgroup and didn't Paul> see anything. google -> "vxworks faq", first link. Paul> The "VxWorks/Tornado II FAQ" seemed to have some timer-related Paul> stuff, but I didn't really see what I wanted. I realize that Paul> busy loops are to be avoided in general, but sometimes you have Paul> to have them. And hardware is not always the answer. Didn't look much. There is a timer heading. There is also a "Hard delay" section. "http://www.xs4all.nl/~borkhuis/vxworks/vxw_pt6.html#6.1" A search on google of "delayLib.c" in comp.os.vxworks gives, "http://groups.google.com/groups?as_q=delayLib.c&as_ugroup=comp.os.vxworks" This includes the original post from "Geoffrey Espin". The version I posted doesn't do the calibration (ala Qnx) and must be called explicitly. I prefer this as the actual first call will delay for more than necessary and you have nothing to accommodate a system with a PLL that can change the system clock. I also added an assembler loop for the ARM, which increases the nanosleep resolution. Nothing like being spoon fed. hth, Bill Pringlemeir. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: host shell does not resume Date: Wed, 11 Sep 2002 16:54:17 GMT From: "Chad" Organization: Shaw Residential Internet Message-ID: References: My console shell never resumes, even if I kill my tornado shell and server. "Behzad Abareshi" wrote in message news:aln7j1$28jo$1@noao.edu... > > "Chad" wrote in message > news:Bkof9.287064$v53.15113569@news3.calgary.shaw.ca... > > My console shell does not function after a target server connects, but > does > > up until that point. Is there a setting that will produce this behavior? > > > > > > You can have only one active shell at a time; you regain control over your > shell after the other client disconnects. > > Behzad. > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: watchdog Timer Date: Wed, 11 Sep 2002 18:54:53 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <65b2abfc.0209110830.29485f36@posting.google.com> Post the source of your watchdog function here, it may help. "Venkat Mynampati" wrote in message news:65b2abfc.0209110830.29485f36@posting.google.com... > Hi, > In my system, i have a watchdog timer that, rearms itself > every 16.67msec and in the associated function, all i does > is to write(increment) to a memory location over pci bus. > Very rarely, what;s hapening is that, this memory location is not > getting incremented and the cpu finds itself in idleLoop forever. > > Can you help me in understanding how this watchdog timer works... > also, > > What will happen if the system has locked out interrupts > via intLock(), and the watchdog timer has expired? this is a > real scenario in my system, where we lock interrupts, to > process them! > > Can the watchdog timer, after having expired a long time back > be still active, once the system has found itself in idleLoop? > > Is it posible that the system is not receiving any events at all? > > Thanks for your insights and help. > Regards, > Venkat --------------------------- Newsgroups: comp.os.vxworks Subject: watchdog timer and tExcTask Queue Date: 11 Sep 2002 10:37:01 -0700 From: mynamvs@nsc.network.com (Venkat Mynampati) Organization: http://groups.google.com/ Message-ID: <65b2abfc.0209110937.4d015fec@posting.google.com> Hi, In the vxWorks 5.4 programmers guide, it is stated that (pg 90) WatchDog Timers .."However, if the kernel is unable to execute the function immediately for any *reason*,then the function is placedon the tExcTask work Queue." My doubt is, what if the tExcTask Queue is full, then what would happen? Is it possible for this watchdog timer function to be lost? i.e not get a chance to execute? Also, what does the document meant when it says any *reason*? does that meant, a watchdog timer expired, while my cpu is in an ISR? Thank You very much for your help and patience. Regards, Venkat --------------------------- Newsgroups: comp.os.vxworks Subject: Re: watchdog timer and tExcTask Queue Date: Wed, 11 Sep 2002 19:42:29 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <65b2abfc.0209110937.4d015fec@posting.google.com> "Venkat Mynampati" wrote in message news:65b2abfc.0209110937.4d015fec@posting.google.com... [snip] > My doubt is, what if the tExcTask Queue is full, then what would > happen? Is it possible for this watchdog timer function to be lost? > i.e not get a chance to execute? > You will get a panic and a reboot. > Also, what does the document meant when it says any *reason*? > does that meant, a watchdog timer expired, while my cpu is in > an ISR? > No, it means the ISR happened while some task was executing kernel code, so the ISR system calls will be deferred until the task is done with the system call. > Thank You very much for your help and patience. > Welcome. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: 11 Sep 2002 13:40:08 -0400 From: "Paul D. Smith" Organization: Nortel Networks, Inc. / Billerica, MA Message-ID: References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> <8d0fbe67.0209100243.5f9fe38f@posting.google.com> <3D7E83F1.A7FA77E6@bellsouth.net> Reply-To: pausmith@nortelnetworks.com %% Bill Pringlemeir writes: >> Alternatively, you could look at the vxWorks FAQ. Paul> Umm, which one? Google returns a bunch of them. I looked in Paul> the "Unofficial FAQ" and the FAQ for this newsgroup and didn't Paul> see anything. bp> google -> "vxworks faq", first link. As I said: Paul> The "VxWorks/Tornado II FAQ" seemed to have some timer-related stuff, This is the first link. I looked at it. Paul> but I didn't really see what I wanted. I realize that busy Paul> loops are to be avoided in general, but sometimes you have to Paul> have them. And hardware is not always the answer. bp> Didn't look much. There is a timer heading. I read it. The only question that seemed remotely related was the first one, and the discussion there talks about hardware timers etc. The code referenced in this section is nothing like what I'm looking for. bp> There is also a "Hard delay" section. I didn't see this one: I was searching for "time", "timers", "spin", etc., not "delay". It would be nice if this section were moved into, or at least referenced by, the time/timers section. bp> A search on google of "delayLib.c" in comp.os.vxworks gives, And why would I search for that? bp> Nothing like being spoon fed. Pfeh. It's all very easy when you already know what you're looking for and what terms to look for it under. Have a nice day. - -- - ------------------------------------------------------------------------------- Paul D. Smith HASMAT--HA Software Mthds & Tools "Please remain calm...I may be mad, but I am a professional." --Mad Scientist - ------------------------------------------------------------------------------- These are my opinions---Nortel Networks takes no responsibility for them. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: 11 Sep 2002 13:52:39 -0400 From: "Paul D. Smith" Organization: Nortel Networks, Inc. / Billerica, MA Message-ID: References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> <8d0fbe67.0209100243.5f9fe38f@posting.google.com> <3D7E83F1.A7FA77E6@bellsouth.net> Reply-To: pausmith@nortelnetworks.com %% Bill Pringlemeir writes: bp> The version I posted doesn't do the calibration (ala Qnx) and must bp> be called explicitly. I prefer this as the actual first call will bp> delay for more than necessary and you have nothing to accommodate bp> a system with a PLL that can change the system clock. As I mentioned, there's an explicit nanospin_calibrate() function you can use if you want to (re-)calibrate at a specific time rather than as a side effect of the first nanospin() call. - -- - ------------------------------------------------------------------------------- Paul D. Smith HASMAT--HA Software Mthds & Tools "Please remain calm...I may be mad, but I am a professional." --Mad Scientist - ------------------------------------------------------------------------------- These are my opinions---Nortel Networks takes no responsibility for them. --------------------------- Newsgroups: comp.os.vxworks Subject: preprocessed files Date: Wed, 11 Sep 2002 19:56:03 +0200 From: "attilio" Organization: parvis Message-ID: Reply-To: "attilio" Hi, I'm using Tornado 2.0.2 and when I create and build some projects I'd like to have beside the *.o files generated also the preprocessed files with a different estension ( say *.i). It's possible to do that without blocking the compilation? tks --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: 11 Sep 2002 14:14:31 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> <8d0fbe67.0209100243.5f9fe38f@posting.google.com> <3D7E83F1.A7FA77E6@bellsouth.net> Sender: bpringle@DeadDuck [snip] bp> Didn't look much. There is a timer heading. Paul> I read it. The only question that seemed remotely related was Paul> the first one, and the discussion there talks about hardware Paul> timers etc. The code referenced in this section is nothing Paul> like what I'm looking for. bp> There is also a "Hard delay" section. Paul> I didn't see this one: I was searching for "time", "timers", Paul> "spin", etc., not "delay". Paul> It would be nice if this section were moved into, or at least Paul> referenced by, the time/timers section. bp> A search on google of "delayLib.c" in comp.os.vxworks gives, Paul> And why would I search for that? Unfortunately, I assumed that you might scan the entire contents for more than just "time". There is other good information in the FAQ that Johan has put there. You might look at it just for reference. Finding the section entitled delay, with a reference to "delayLib.c" would give you all the information that I posted. bp> Nothing like being spoon fed. Paul> Pfeh. It's all very easy when you already know what you're Paul> looking for and what terms to look for it under. Then, I assume I helped you? Your welcome. I did my work to find this (exactly as I described). I use Johan's FAQ all the time (and even WindSurf). Then and only then do I ask questions here. I assume people on c.o.v are busy. I guess you don't. I am sorry. Manjeet (and all replies) have created over 12 posts in a thread that isn't really all that illuminating. But then again, not much here is... Regards, Bill Pringlemeir. - -- Have you ever seen large lymph nodes turn putrid 29 times in a row? Or had large lymph nodes eat your mail while driving down the road at 120mph? Or had someone you saw at a party share intimate feelings while drinking Sterno straight from the can? You will. And the company that will bring it to you: AT&T. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: host shell does not resume Date: Wed, 11 Sep 2002 11:39:26 -0700 From: "Behzad Abareshi" Organization: NOAO Message-ID: References: > My console shell never resumes, even if I kill my tornado shell and server. > Hmmm... I usually have a shell session through a serial port, and allow others to have shell sessions through the network. Whenever someone logs in I lose control of my shell, but after they disconnect I hit Enter and I am back again. I know this ain't helpful to you; sorry. Behzad. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: Wed, 11 Sep 2002 11:54:08 -0700 From: "Behzad Abareshi" Organization: NOAO Message-ID: References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> <8d0fbe67.0209110446.22a56e50@posting.google.com> "Manjeet Singh Chhabra" wrote in message news:8d0fbe67.0209110446.22a56e50@posting.google.com... > "Behzad Abareshi" wrote in message news:... > > > I have a problem i want to slepp a task for a few nanoseconds. > > > > Tell us why you want a few nanosecs delay in the first place; you may be > > pleasantly surprised that what you think you need is often times very > > different from what you really need! Talking from my own experience, of > > course :-) > > Actually we have a router. > Running on a COM-CPU ( i don't know hardware Architeture ) > and i am writing a device driver for Galnet cross bar. > > what driver does is it receives packets from the cross bar and sends > these packets to upper layers. > > my driver is a task which is polling all the 8 ports of the cross bar. > > in eache iteration it looks for packets and send at max 8 packets to > upperlayers. and sleeps for 1 tick ( which is processing time for 8 > packets ). > > to allow other task to process those packets. > > but the problem is that if there are less no of packets still i have > to sleep for processing time equivalent to 8 packets because i don't > have any machnism to slepp for less than one tick. > > How can i do this ? > > Manjeet Use select to watch the ports; read as soon as data is ready in any port; store the packets in an array if you read less than 8 packets; repeat until you have at least 8 packets; do the processing. The nice thing about select is that it doesn't hog the CPU while waiting, so you don't need to finish processing everything before reading more packets. Behzad. --------------------------- Newsgroups: comp.os.vxworks Subject: why my NTP showing error: S_hostLib_UNKNOWN_HOST? Date: 11 Sep 2002 13:31:08 -0700 From: roswel_ajf@hotmail.com (roswell) Organization: http://groups.google.com/ Message-ID: <6dcae13d.0209111231.7ffa3de8@posting.google.com> Hi, I have my NTP client which goes out every 12 hours and obtain the timestamp. anyway, got this errno 0x320001 displayed. don't know why..any hint?....thnks --------------------------- Newsgroups: comp.os.vxworks Subject: BSD IP Socket buffer question Date: Wed, 11 Sep 2002 15:51:42 -0500 From: Chris Richards Organization: Nortel Message-ID: <3D7FACDE.5080103@nortelNADASPAMnetworks.com> Hi All, This question is not just related to VxWorks, but since the VxWorks IP Stack is based on BSD code, I;ll ask away: I was wondering if anyone knew why a tail pointer to the mbuf list was not implemented. This seems a huge oversight especially for the receive sovket buffer:- Every time a packet is added to a socket buffer, the entire linked list of packets (contained in mbufs) must be traversed in order to find the tail and add the new packet (mbuf). If you have a large receive socket buffer which potentially tens or even hundreds of packets in the list (During a short packet burst), it takes significant CPU processing just to add them to the list. Is there any particular reason why a tail pointer was never added? Cheers, Chris. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: 11 Sep 2002 18:07:48 -0400 From: "Paul D. Smith" Organization: Nortel Networks, Inc. / Billerica, MA Message-ID: References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> <8d0fbe67.0209100243.5f9fe38f@posting.google.com> <3D7E83F1.A7FA77E6@bellsouth.net> Reply-To: pausmith@nortelnetworks.com %% Bill Pringlemeir writes: bp> Nothing like being spoon fed. Paul> Pfeh. It's all very easy when you already know what you're Paul> looking for and what terms to look for it under. bp> Then, I assume I helped you? Your welcome. I really do appreciate the helpful part of your response. bp> I did my work to find this (exactly as I described). I use bp> Johan's FAQ all the time (and even WindSurf). Then and only then bp> do I ask questions here. If you had said ''look in the faq; there's something there called "delay" or "delayLib"'' that would have been sufficient, and no more work for you and much more helpful to me than the original response. As I said, it's easy to find things when you know the right keywords. bp> I assume people on c.o.v are busy. I guess you don't. Ah, good. So, I guess I know what to tell someone the next time they have a question about, say, GNU make. I'll just say "It's in the manual". And if they still can't find it, I'll just say "read it all." To think I've been wasting so much time over the years... heck, I could just bind that "it's in the manual--read the whole thing" response to an autoremailer in the help-make mailing list, and I could unsubscribe!! - -- - ------------------------------------------------------------------------------- Paul D. Smith HASMAT--HA Software Mthds & Tools "Please remain calm...I may be mad, but I am a professional." --Mad Scientist - ------------------------------------------------------------------------------- These are my opinions---Nortel Networks takes no responsibility for them. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: BSD IP Socket buffer question Date: Wed, 11 Sep 2002 22:33:49 +0000 (UTC) From: David Laight Organization: BT Openworld Message-ID: References: <3D7FACDE.5080103@nortelNADASPAMnetworks.com> Chris Richards wrote: > Hi All, > > This question is not just related to VxWorks, but since the VxWorks IP > Stack is based on BSD code, I;ll ask away: > > I was wondering if anyone knew why a tail pointer to the mbuf list was > not implemented. Maybe because someone was lazy... didn't know about some list types... assumed the list would be short... Actually the code could (probably) be written to queue them in reverse order, and invert the list in order to remove items. That only needs a singly linked list. David --------------------------- Newsgroups: comp.os.vxworks Subject: why IFF_NOTRAILERS flag is for? Date: 11 Sep 2002 15:44:36 -0700 From: roswel_ajf@hotmail.com (roswell) Organization: http://groups.google.com/ Message-ID: <6dcae13d.0209111444.7dffe44f@posting.google.com> HI... whats the point of this flag? b'c my MTU is 1500, with this IFF_NOTRAILERS flag. is that mean, L3 payload will be 1500 - 14 (6.+6+2) without ethnerchecksum. pls clarify. --------------------------- Newsgroups: comp.os.vxworks Subject: linux->powerpc-vxworks cross compiler Date: Wed, 11 Sep 2002 22:53:40 GMT From: "Bruce and Jeanne" Organization: Giganews.Com - Premium News Outsourcing Message-ID: I've been trying to build a GCC cross compiler for a powerpc-vxworks that will run on linux and have run into a few problems. I'm using GCC-3.0.1 and binutils-2.1.1 from RedHat The first problem was that the fixincludes script is corrupting the included headder files. Almost all instances of the #define CPU are converted to __CPU__ and almost all instances of #define CPU_FAMILY to __CPU_FAMILY__ . If I correct all instances of this garbeling the make process continues happily until the liblibrary make process starts to run. When concat.c starts to compile I get an error "__builtin_saveregs is not is not supported on this platform". Has anyone seem this sort of thing before? Is this a problem with later versions of GCC, I've seem instructions for building the compiler from GCC-2.96? Is there an EXPAND_BUILTIN_SAVEREGS script for the powerpc that I don't have? Any help would be greatly apreciated. Bruce --------------------------- Newsgroups: comp.os.vxworks Subject: Excessive link/compile times for PPC823 target using Diab 4.4b or 5.0 Date: 11 Sep 2002 16:04:49 -0700 From: babuyama@yahoo.com (Babu Yama) Organization: http://groups.google.com/ Message-ID: We are experienceing excessive link/compile times (C++, Debug mode -g option, link=7-10 min, compile=30 min) for PPC823 target using Diab 4.4b or 5.0. Target compiler = Diab 4.4b / 5.0 Target = MPC823 Language = c/c++ Host = Windows 2000 Host compiler = MS VC6.0 RTOS = Nucleaus During a link using a 2.0 GHz Pentium system neither CPU time nor file I/O are being utilized at anywhere their limits. CPU utilization is 2-3% most of the time. There is not much disk activity either. It doesn't seem like file I/O is the gating factor here. Even with Node locked (local machine) Diab Lincense manager, we get the same poor performance. The exact code base using Microsoft compiler on the host(Visual Studio 6.0) links in 1/10 th of the time and compiles in about 1/2 the time with –g option with or WITHOUT incremental linking. Why is Diab performing so poorly? Windriver customer support was not helpful in getting to the bottom of the issue. They however, told that Diab does not have incremental linking feature (What I mean by incremental linking is discussed in the this paper. http://www.usenix.org/events/osdi2000/wiess2000/full_papers/mikulin/mikulin.pdf) Is there any thing we can do to improve our re-link times? That is after the first clean build, when we change one .cpp file and generate just one new .o file and relink, why does the linking take the same amount of time as the first clean build? This is what incremental linking solves, right? Unfortunately, Diab does not have incremental linking feature. Does gnu ccppc have the incremental linking feature for faster re-link times? Your comments are greatly appreciated. - - Babu --------------------------- Newsgroups: comp.os.vxworks Subject: A question on obtaining vxworks BSD driver Date: 11 Sep 2002 18:30:30 -0700 From: li.mingzheng@zte.com.cn (M.Z.LI) Organization: http://groups.google.com/ Message-ID: <54a46c51.0209111730.414e0b4a@posting.google.com> Hi, Has WindRiver ever written BSD style network driver for 3com 3c90x card? I could not find it in my vxworks source.( we have bought the complete target source for Vxworks). If they have this part with later release, how can I request this update? Thanks in advance! Sincerely Yours M.Z.Li --------------------------- Newsgroups: comp.os.vxworks Subject: Re: why IFF_NOTRAILERS flag is for? Date: Thu, 12 Sep 2002 07:24:25 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <6dcae13d.0209111444.7dffe44f@posting.google.com> Rosewel this is a peice of history, in fact. Many years ago, when TCP/IP was first implemented on a DEC VAX machine running BSD Unix, they devised an extention to the standard IP encapsulation over Ethernet, that would put the IP header at the end of the packet rather than at the beginning, and effectively put the payload data at the beginning of a 512 byte page, as a means to gain effeciency. The "Trailers" protocol demanded that any two parties need to agree they both support this extention as part of the ARP discovery, before using it. The "Notrailers" flag effectively disables the support for "trailers" protocol in the lower part of the IP stack (including ARP), and you wil see this flag on almost every machine supporting TCP/IP. In fact I would not be surprised if the "trailers" protocol would fail if attempted on any modern Unix system. - - Leonid "roswell" wrote in message news:6dcae13d.0209111444.7dffe44f@posting.google.com... > HI... > > whats the point of this flag? b'c my MTU is 1500, with this > IFF_NOTRAILERS flag. is that mean, L3 payload will be 1500 - 14 > (6.+6+2) without ethnerchecksum. pls clarify. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: Thu, 12 Sep 2002 05:55:16 GMT From: "Gary M" Organization: AT&T Broadband Message-ID: <7%Vf9.397677$me6.46015@sccrnsc01> References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> <8d0fbe67.0209110446.22a56e50@posting.google.com> Galnet crossbar. Ecch! I hope you have better luck than we have had thus far. Many holes in the datasheet, etc. Nanosecond sleeping should be a breeze compared to Galnet programming. "Manjeet Singh Chhabra" wrote in message news:8d0fbe67.0209110446.22a56e50@posting.google.com... > "Behzad Abareshi" wrote in message news:... > > > I have a problem i want to slepp a task for a few nanoseconds. > > > > Tell us why you want a few nanosecs delay in the first place; you may be > > pleasantly surprised that what you think you need is often times very > > different from what you really need! Talking from my own experience, of > > course :-) > > Actually we have a router. > Running on a COM-CPU ( i don't know hardware Architeture ) > and i am writing a device driver for Galnet cross bar. > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: 11 Sep 2002 23:04:29 -0700 From: manjeetsc@yahoo.com (Manjeet Singh Chhabra) Organization: http://groups.google.com/ Message-ID: <8d0fbe67.0209112204.62004622@posting.google.com> References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> <8d0fbe67.0209110446.22a56e50@posting.google.com> "Behzad Abareshi" wrote in message news:... > "Manjeet Singh Chhabra" wrote in message > news:8d0fbe67.0209110446.22a56e50@posting.google.com... > > "Behzad Abareshi" wrote in message > news:... > > > > I have a problem i want to slepp a task for a few nanoseconds. > > > > > > Tell us why you want a few nanosecs delay in the first place; you may be > > > pleasantly surprised that what you think you need is often times very > > > different from what you really need! Talking from my own experience, of > > > course :-) > > > > Actually we have a router. > > Running on a COM-CPU ( i don't know hardware Architeture ) > > and i am writing a device driver for Galnet cross bar. > > > > what driver does is it receives packets from the cross bar and sends > > these packets to upper layers. > > > > my driver is a task which is polling all the 8 ports of the cross bar. > > > > in eache iteration it looks for packets and send at max 8 packets to > > upperlayers. and sleeps for 1 tick ( which is processing time for 8 > > packets ). > > > > to allow other task to process those packets. > > > > but the problem is that if there are less no of packets still i have > > to sleep for processing time equivalent to 8 packets because i don't > > have any machnism to slepp for less than one tick. > > > > How can i do this ? > > > > Manjeet > > Use select to watch the ports; read as soon as data is ready in any port; > store the packets in an array if you read less than 8 packets; repeat until > you have at least 8 packets; do the processing. The nice thing about select > is that it doesn't hog the CPU while waiting, so you don't need to finish > processing everything before reading more packets. > > Behzad. your idea looks to me fine but if there are less than 8 packets and there are no more packets those packets will remain unserved. delaying task for those many packets is a better option i think what do you suggest ? Manjeet --------------------------- Newsgroups: comp.os.vxworks Subject: how to use 860 SCC1? Date: 11 Sep 2002 23:06:53 -0700 From: wangjf@wmc.ee.tsinghua.edu.cn (Jeff) Organization: http://groups.google.com/ Message-ID: Hi, all, I have configured the MPC860 SCC1 as an ethernet channel, as the debugging channel with Tornado2.0 on my board. However, an idea come to my mind that whether I could use this SCC1 as a real ethernet port, not just as a debugging port? Can I transfer data through it as an normal "NIC card" does? What should I do? And I think change the original BSP is indispensible, right? And then? Write new driver codes? Could anyone give me some demos or examples? Thanks in advancd! Andi Wang wangjf@wmc.ee.tsinghua.edu.cn 2002/09/12 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: host shell does not resume Date: Thu, 12 Sep 2002 08:30:24 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: Reply-To: "Michael Lawnick" Sender: mlawnick@pd9524698.dip.t-dialin.net Hi Chad, depending on your target server configuration and the script $(WIND_BASE)/.wind/windSh.tcl stdIn and stdOut are redirected to /vio (your target server). To get your target shell under control reconfigure the target server or press Ctrl-C to restart it. Host shell will display an exception - don't care, its the text from killing the target shell. Another story is that your serial shell will be suspended as long as you have a telnet connection. It will get active again after closing telnet. HTH Michael "Chad" schrieb im Newsbeitrag news:Bkof9.287064$v53.15113569@news3.calgary.shaw.ca... > My console shell does not function after a target server connects, but does > up until that point. Is there a setting that will produce this behavior? > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: why my NTP showing error: S_hostLib_UNKNOWN_HOST? Date: 11 Sep 2002 23:41:16 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0209112241.5e91b939@posting.google.com> References: <6dcae13d.0209111231.7ffa3de8@posting.google.com> Hello, roswel_ajf@hotmail.com (roswell) wrote in message news:<6dcae13d.0209111231.7ffa3de8@posting.google.com>... > Hi, > > I have my NTP client which goes out every 12 hours and obtain the > timestamp. anyway, got this errno 0x320001 displayed. don't know > why..any hint?....thnks Assuming that you are looking up the IP address of the time server using a name and DNS, the error is probably generated by the name resolver's search order: first it checks hostLib (the equivalent of /etc/hosts if you like), then it goes to DNS if it didn't get an answer. As ever in VxWorks, if you didn't get an error back from any system function you called, you should ignore the errno value. In this case I'm guesing that the resolver code used it to determine that it should send the request out to your DNS server as hostLib was unable to match the name to an IP address. HTH, John... --------------------------- Newsgroups: comp.os.vxworks Subject: Re: how to get "Physical Address" directly from PCI enumeration? Date: 12 Sep 2002 01:02:31 -0700 From: pdufrene.other@free.fr (Pierre D.) Organization: http://groups.google.com/ Message-ID: <8726dedf.0209120002.17a2bc@posting.google.com> References: <200299-145744-758413@foorum.com> <8400a733.0209100118.7e012db3@posting.google.com> <8726dedf.0209110039.273c1ad2@posting.google.com> Hi Michael, Thank you for your answer, I will look for something else to identify the computer. Pierre > Hi Pierre, > > this depends on your NIC but is (AFAIK) typically not possible. MAC is > mostly saved in NIC's SRAM that is not memory mapped into the PCI address > space. You will have to look into the documentation of your specific card. > > Regards, > Michael > > "Pierre D." schrieb im Newsbeitrag > news:8726dedf.0209110039.273c1ad2@posting.google.com... > > Thanks for your answer but this network adapter is not used by > > VxWorks. I want to use the MAC address to identify the computer, but I > > don't need any other network funtionnalities. So my question was how > > to get the MAC address only by PCI ennumeration and by doing some dump > > from memory/IO/... > > > > Charly > > > > > correct, imho you're looking for something like this: > > > > > > #include > > > #include > > > #include > > > > > > int getmac( char *name, int nr ) > > > { > > > END_OBJ * pEnd; > > > unsigned char mac[6]; > > > > > > pEnd= endFindByName( name, nr ); > > > if ( !pEnd ) > > > { > > > printf("NIC %s%i not found!\n",name,nr); > > > return( -1 ); > > > } > > > > > > muxIoctl(pEnd, EIOCGADDR, mac); > > > > > > printf("MAC Address of %s%i is %02x:%02x:%02x:%02x:%02x:%02x\n", > > > name, nr, mac[0], mac[1], mac[2], mac[3], mac[4], mac[5], > mac[6] ); > > > > > > return( 0 ); > > > } > > > > > > toni --------------------------- Newsgroups: comp.os.vxworks Subject: Help: traps in VxSim Date: Thu, 12 Sep 2002 09:34:10 -0400 From: "neuralking" Organization: neural Message-ID: Reply-To: "neuralking" Got following traps when my application were running. I can't debug because these traps happened in low-level OS area (tExcTask). Any suggention is greatly appreciated! - ---------------------------------------------------------------------------- - ---------------------------------------------------------- Exception number 13: Task: 0x117e0740 (tExcTask) General Protection Fault Program Counter: 0x0041808a Status Register: 0x00010287 5490a51 _writeTrapLog +55 : _generateLog_skipback_for_task ([1000b, 117e0740, 254fbfb, 70, 0]) 5463995 _generateLog_skipback_for_task+25 : _generateLog_skipback_for_task_t ([1000b, 117e0740, 254fbfb, 70, 0]) 54639e0 _generateLog_skipback_for_task_t+3c : _genericGenerateLog ([1000b, 117e0740, 254fbfb, 70, 0]) 5463606 _genericGenerateLog+2ee: _addStack (117e0740, 53bef60, 534, 8000, 32, 0, 5) 545324f _addStack +127: _writeLogData ([53bef60, 534, 8000, 69, 254fc87]) 5452133 _writeLogData +3b : _memcpy ([53bf4a0, 254fc87, 7ac0, 2557747, 2557747]) 458f1a _memcpy +e : 53bf1b9 ([254fc87, 53bf4a0, 7ac0, 540, 254faeb]) 0 : _simCallHandler@4 ([7ac0, 254facb, 458f1f, 254fc87, 53bf4a0]) 410646 _simCallHandler@4+2a : _excExcHandle (0, &_simVxExcInfo, &_simExcRegs) 438f3c _excExcHandle +108: _traplog_ExcepHook (117e0740, 0, &_simVxExcInfo) 5490d7f _traplog_ExcepHook+13 : _writeTrapLog (5490d5f, 117e0740, 0, &_simVxExcInfo, 1) 5490a51 _writeTrapLog +55 : _generateLog_skipback_for_task ([1000b, 117e0740, 254f9e7, 70, 0]) 5463995 _generateLog_skipback_for_task+25 : _generateLog_skipback_for_task_t ([1000b, 117e0740, 254f9e7, 70, 0]) 54639e0 _generateLog_skipback_for_task_t+3c : _genericGenerateLog ([1000b, 117e0740, 254f9e7, 70, 0]) 5463606 _genericGenerateLog+2ee: _addStack (117e0740, 53cef80, 534, 8000, 32, 0, 5) 545324f _addStack +127: _writeLogData ([53cef80, 534, 8000, 69, 254fa73]) 5452133 _writeLogData +3b : _memcpy ([53cf4c0, 254fa73, 7ac0, 2557533, 2557533]) 458f1a _memcpy +e : 53cf1d9 ([254fa73, 53cf4c0, 7ac0, 540, 254f8d7]) 0 : _simCallHandler@4 ([7ac0, 254f8b7, 458f1f, 254fa73, 53cf4c0]) 410646 _simCallHandler@4+2a : _excExcHandle (0, &_simVxExcInfo, &_simExcRegs) 438f3c _excExcHandle +108: _traplog_ExcepHook (117e0740, 0, &_simVxExcInfo) 5490d7f _traplog_ExcepHook+13 : _writeTrapLog (5490d5f, 117e0740, 0, &_simVxExcInfo, 1) 5490a51 _writeTrapLog +55 : _generateLog_skipback_for_task ([1000b, 117e0740, 254f7d3, 70, 0]) 5463995 _generateLog_skipback_for_task+25 : _generateLog_skipback_for_task_t ([1000b, 117e0740, 254f7d3, 70, 0]) 54639e0 _generateLog_skipback_for_task_t+3c : _genericGenerateLog ([1000b, 117e0740, 254f7d3, 70, 0]) 5463606 _genericGenerateLog+2ee: _addStack (117e0740, 5386ef0, 534, 8000, 32, 0, 5) 545324f _addStack +127: _writeLogData ([5386ef0, 534, 8000, 69, 254f85f]) 5452133 _writeLogData +3b : _memcpy ([5387430, 254f85f, 7ac0, 255731f, 255731f]) 458f1a _memcpy +e : 5387149 ([254f85f, 5387430, 7ac0, 540, 254f6c3]) 0 : _simCallHandler@4 ([7ac0, 254f6a3, 458f1f, 254f85f, 5387430]) 410646 _simCallHandler@4+2a : _excExcHandle (0, &_simVxExcInfo, &_simExcRegs) 438f3c _excExcHandle +108: _traplog_ExcepHook (117e0740, 0, &_simVxExcInfo) 5490d7f _traplog_ExcepHook+13 : _writeTrapLog (5490d5f, 117e0740, 0, &_simVxExcInfo, 1) 5490a51 _writeTrapLog +55 : _generateLog_skipback_for_task ([1000b, 117e0740, 254f5bf, 70, 0]) 5463995 _generateLog_skipback_for_task+25 : _generateLog_skipback_for_task_t ([1000b, 117e0740, 254f5bf, 70, 0]) 54639e0 _generateLog_skipback_for_task_t+3c : _genericGenerateLog ([1000b, 117e0740, 254f5bf, 70, 0]) 5463606 _genericGenerateLog+2ee: _addStack (117e0740, 539ef20, 534, 8000, 32, 0, 5) 545324f _addStack +127: _writeLogData ([539ef20, 534, 8000, 69, 254f64b]) 5452133 _writeLogData +3b : _memcpy ([539f460, 254f64b, 7ac0, 255710b, 255710b]) 458f1a _memcpy +e : _bcopy ([254f64b, 539f460, 7ac0, 540, 254f4af]) --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: Thu, 12 Sep 2002 07:02:10 -0700 From: "Behzad Abareshi" Organization: NOAO Message-ID: References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> <8d0fbe67.0209110446.22a56e50@posting.google.com> <8d0fbe67.0209112204.62004622@posting.google.com> > your idea looks to me fine but if there are less than 8 packets and > there are no more packets those packets will remain unserved. > > delaying task for those many packets is a better option i think > what do you suggest ? > Select has a timeout that you can set to suit your app; you process whatever packages you have if you don't get anything more and select times out. Behzad. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: Thu, 12 Sep 2002 16:28:02 +0000 (UTC) From: David Laight Organization: BT Openworld Message-ID: References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> <8d0fbe67.0209110446.22a56e50@posting.google.com> <8d0fbe67.0209112204.62004622@posting.google.com> > > Select has a timeout that you can set to suit your app; you process whatever > packages you have if you don't get anything more and select times out. It is still limited to the system clock tick. And if you increase the clock frequency above 2147Hz attempts to sleep for sub-second times are likely to block the task forever (and probably stuff the wall time). David --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: Thu, 12 Sep 2002 10:30:20 -0700 From: "Behzad Abareshi" Organization: NOAO Message-ID: References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> <8d0fbe67.0209110446.22a56e50@posting.google.com> <8d0fbe67.0209112204.62004622@posting.google.com> "David Laight" wrote in message news:alqfah$qtl$1@venus.btinternet.com... > > > > Select has a timeout that you can set to suit your app; you process whatever > > packages you have if you don't get anything more and select times out. > > It is still limited to the system clock tick. > And if you increase the clock frequency above 2147Hz attempts to > sleep for sub-second times are likely to block the task forever > (and probably stuff the wall time). > > David > Note that the response was to the concern that the initial scheme I suggested failed to explicitly address the case where less than eight packages arrive and no package comes after those for a long time. So the select timeout is to handle this noncritical detail, and the timeout value would be several times the clock tick. I think this is adequate: the scheme takes care of heavy traffic, and is relaxed when there is little traffic. I think I have said enough on this thread :-) Peace. --------------------------- Newsgroups: comp.os.vxworks Subject: Problem establishing communication on a PPP link. Date: 12 Sep 2002 13:29:17 -0700 From: dmasse@ele.etsmtl.ca (David Masse) Organization: http://groups.google.com/ Message-ID: <14fc63a8.0209121229.54d4ea7a@posting.google.com> Hi, I'm trying to establish a PPP link between two VxWorks targets. I create a serial TTY driver on a HDLC port of a MPC860 CPU for each target. I want to install PPP over this serial channel to allow the two targets to communicate. I initializes the passive target with no option like this: PPP_OPTIONS pppOpt; pppOpt.flags = OPT_PASSIVE_MODE | OPT_NO_PAP | OPT_NO_CHAP | OPT_NO_VJ; pppOpt.lcp_echo_interval = "30"; pppOpt.lcp_echo_failure = "10"; if( pppInit( 0, "/tyCo/1", localAddr, remoteAddr, 0, &pppOpt, NULL ) == OK ){ logMsg("pppTest: ppp link established\n",0,0,0,0,0,0); } else { logMsg("pppTest: ERROR! Cannot establish ppp link\n",0,0,0,0,0,0); } Almost same thing for the active target: PPP_OPTIONS pppOpt; pppOpt.flags = OPT_NO_PAP | OPT_NO_CHAP | OPT_NO_VJ; pppOpt.lcp_echo_interval = "30"; pppOpt.lcp_echo_failure = "10"; if( pppInit( 0, "/tyCo/1", localAddr, remoteAddr, 0, &pppOpt, NULL ) == OK ){ logMsg("pppTest: ppp link established\n",0,0,0,0,0,0); } else { logMsg("pppTest: ERROR! Cannot establish ppp link\n",0,0,0,0,0,0); } The problem is that for both, the pppInit() call return OK but then, I received the following error: [vxKernel] -> ppp0: pppd: vj-max-slots value must be between 2 and 16 ppp0: vj_max_slots error 0x241db0 (tPPP0): memPartFree: invalid block 0xce7ec in partition 0x235ed0 Someone know what is the problem? Thanks in advance. David --------------------------- Newsgroups: comp.os.vxworks Subject: Re: how to use 860 SCC1? Date: Thu, 12 Sep 2002 14:28:23 -0700 From: "Michael R. Kesti" Organization: MK Associates Message-ID: <3D8106F7.17F3BFF3@gv.net> References: Reply-To: mkesti@gv.net Jeff wrote: >Hi, all, > >I have configured the MPC860 SCC1 as an ethernet channel, as the >debugging channel with Tornado2.0 on my board. However, an idea come >to my mind that whether I could use this SCC1 as a real ethernet port, >not just as a debugging port? Can I transfer data through it as an >normal "NIC card" does? Yes. >What should I do? And I think change the original BSP is >indispensible, right? And then? Write new driver codes? There is, to my knowledge, nothing that you need to do. The network interface is available for debugging and for application purposes. >Could anyone give me some demos or examples? Section 7.2.2 of the VxWorks 5.4 Network Programmer's Guide. >Thanks in advancd! You're welcome. - -- ======================================================================== Michael Kesti | "And like, one and one don't make | two, one and one make one." mkesti@gv.net | - The Who, Bargain --------------------------- Newsgroups: comp.os.vxworks Subject: Re: A question on obtaining vxworks BSD driver Date: 12 Sep 2002 21:52:32 -0700 From: wolfhe@sina.com (wolfhe) Organization: http://groups.google.com/ Message-ID: <70ed4eed.0209122052.431df5fd@posting.google.com> References: <54a46c51.0209111730.414e0b4a@posting.google.com> Hi, If you use Tornado II, you should have it in your system library. The name of the 3c90x series drivers is "el3c90xEnd.o". If you provide 2 functions 90xEndIntEnale/Disable, it should work. Good luck He Jin li.mingzheng@zte.com.cn (M.Z.LI) wrote in message news:<54a46c51.0209111730.414e0b4a@posting.google.com>... > Hi, > > Has WindRiver ever written BSD style network driver for 3com 3c90x > card? I could not find it in my vxworks source.( we have bought the > complete target source for Vxworks). If they have this part with later > release, how can I request this update? > > Thanks in advance! > > Sincerely Yours > > M.Z.Li --------------------------- Newsgroups: comp.os.vxworks Subject: Exception Handling in vxWorks. Date: 12 Sep 2002 22:44:43 -0700 From: tgisfree@yahoo.com (tzeghuan) Organization: http://groups.google.com/ Message-ID: <290ade1c.0209122144.5566acf4@posting.google.com> Hi all, Recently i tried out some exception handling in both windows and vxworks. The same piece of code works in win environment but not in vxworks. Can anyone know about this give some clue? Is there a need to enable exception handling or any thing like that? Thanks in advance. regards tzeghuan --------------------------- Newsgroups: comp.os.vxworks Subject: any website on vxworks faq? Date: 13 Sep 2002 01:01:36 -0700 From: tgisfree@yahoo.com (tzeghuan) Organization: http://groups.google.com/ Message-ID: <290ade1c.0209130001.6751ea62@posting.google.com> any website on vxworks faq? thks --------------------------- Newsgroups: comp.os.vxworks Subject: Re: File problem with either DOS or TFFS Date: Fri, 13 Sep 2002 08:20:03 +0000 (UTC) From: David Laight Organization: BT Openworld Message-ID: References: > The corrupted data in the file is always all zero bytes. Maybe there is always a 2k random overwrite - it just doesn't always hit anywhere important. It might be possible to scan for a 2k block of zeros... David --------------------------- Newsgroups: comp.os.vxworks Subject: testing Date: 13 Sep 2002 01:21:48 -0700 From: tgisfree@yahoo.com (tzeghuan) Organization: http://groups.google.com/ Message-ID: <290ade1c.0209130021.55ce982f@posting.google.com> sorry can't c my msg posted!? --------------------------- Newsgroups: comp.os.vxworks Subject: adding rcs2 space under powerpc Date: Fri, 13 Sep 2002 09:45:30 GMT From: "Kelly Hornsby" Organization: Cox Communications Message-ID: <_sig9.17711$8F4.612598@news2.east.cox.net> Hello All I have a Sandpoint (PowerPC 745) development system that runs VxWorks 5.4.1 very nicely. There is a serial device in the 64 bit space of rcs2 on this system. I want to access this serial device. The access for the device is not in the BSP, and rcs2 is apparently not activated under this bsp, so I'm trying to add access to the rcs2 area by modifying my bootrom and vxworks images to open up this space. Question: What structure should I modify in order to add rcs2 access to vxworks? Should I modify the BATs, or should I add PTEs, or am I off totally? I realize this space is a 64-bit space, and that I must access the area via a double float pointer that points to the byte registers in the serial device. I'm not sure what to do here. I'm looking for some direction here, because I'm not sure where to go with this. Any inputs are welcome. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: any website on vxworks faq? Date: Fri, 13 Sep 2002 06:26:34 -0400 From: joe durusau Organization: WEBUSENET.com Message-ID: <3D81BD5A.B7312DBC@bellsouth.net> References: <290ade1c.0209130001.6751ea62@posting.google.com> tzeghuan wrote: > any website on vxworks faq? > thks I can't spell the guy's name, but the pointer to the FAQ is posted here periodically. A google search will surely show it. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: muxDevStart()? Date: 13 Sep 2002 04:22:35 -0700 From: oe_zhou@sohu.com (oe_zhou) Organization: http://groups.google.com/ Message-ID: <758a30ff.0209130322.6a686e10@posting.google.com> here is one segment of my bsp:(ppc860+vxworks) /*=================================================*/ pEnd = endFindByName (bootDev, params.unitNum); printf("\n here: endFindByName().\n"); /* added by zhou */ if (!attached && (pEnd != NULL)) { printf("\n here: muxDevStart. \n"); if (muxDevStart (pEnd) != OK) { printf("Failed to start device %s\n", bootDev); return (ERROR); } printf("\n here: muxIoctl.\n"); .................... It can display the " here:muxDevStart." ,but it can't output the "here: muxIoctl" or "Failed to start device.". And it seems stopping and does nothing. Help me! --------------------------- Newsgroups: comp.os.vxworks Subject: disk cache error Date: 13 Sep 2002 07:16:52 -0700 From: ritalin@comcast.net (Rita) Organization: http://groups.google.com/ Message-ID: Hello, When we write to our disk, after a long while, we see the following error message: copy "0310FEF6.DRV","abc" 0x8697fcb0 (tShell): disk cache error: device 87f936f0 block -1 errno 0, disk read failed copy: error reading file after copying 327680 bytes. value = -1 = 0xffffffff - -> copy "0310FEF6.DRV","abc" 0x8697fcb0 (tShell): disk cache error: device 87f936f0 block -1 errno c0003, disk read failed copy: error reading file after copying 327680 bytes. value = -1 = 0xffffffff It seems to me that the block number is out of range. Does anyone know what might have cause this problem? It's not an easy reproducible problem, and we are using scsi flash drive. We've used the same thing for over 2 years, and never saw the problem before. Thanks. Rita --------------------------- Newsgroups: comp.os.vxworks Subject: SetupSDK Date: 13 Sep 2002 08:23:11 -0700 From: robert_markey@hotmail.com (robert) Organization: http://groups.google.com/ Message-ID: <1fac8c75.0209130723.5c9f89f0@posting.google.com> Hi, Does somebody have chance to use or using SetupSDK utility produced by VxWorks. Rob --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Exception Handling in vxWorks. Date: 13 Sep 2002 10:32:14 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0209130932.1c877e18@posting.google.com> References: <290ade1c.0209122144.5566acf4@posting.google.com> Hello, Your question is so vague that I doubt you will get a useful answer. Try posting some more information... here are a few questions for you to think about: 1) Did you mean hardware exception handling or C++ exception handling (very different things)? 2) What doesn't work exactly? Does it not compile? Are there errors when you download it? Does it cause errors when you run it? 3) Have you got a small code fragment that you could post here? 4) Which architecture/BSP are you using? 5) What compiler flags did you use when you compiled it? 6) Have you read the manual? 7) Have you searched the archives of this group for possible answers? 8) Have you read the FAQ for this group? Please remember that people in this forum are not paid to answer your questions, and are usually very busy with their own work. If you want to get a good answer you will need to think up front about the kinds of information that are relevant; most of the people here do not have the time, nor the desire to extract the information from you over a series of postings. That said, if you post enough information about your problem, often you will get an answer. HTH, John... tgisfree@yahoo.com (tzeghuan) wrote in message news:<290ade1c.0209122144.5566acf4@posting.google.com>... > Hi all, > > Recently i tried out some exception handling in both windows and vxworks. > The same piece of code works in win environment but not in vxworks. > Can anyone know about this give some clue? > Is there a need to enable exception handling or any thing like that? > > Thanks in advance. > > regards > tzeghuan --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sleep for nanoseconds without using POSIX Date: Fri, 13 Sep 2002 15:54:01 -0400 From: Neil Horman Organization: WEBUSENET.com Message-ID: <3D824259.3000905@lvl7.com> References: <8d0fbe67.0209092146.3674b6fb@posting.google.com> <8d0fbe67.0209110446.22a56e50@posting.google.com> Reply-To: nhorman@lvl7.com Manjeet Singh Chhabra wrote: > "Behzad Abareshi" wrote in message news:... > >>>I have a problem i want to slepp a task for a few nanoseconds. >>> >>Tell us why you want a few nanosecs delay in the first place; you may be >>pleasantly surprised that what you think you need is often times very >>different from what you really need! Talking from my own experience, of >>course :-) >> > > Actually we have a router. > Running on a COM-CPU ( i don't know hardware Architeture ) > and i am writing a device driver for Galnet cross bar. > > what driver does is it receives packets from the cross bar and sends > these packets to upper layers. > > my driver is a task which is polling all the 8 ports of the cross bar. > > in eache iteration it looks for packets and send at max 8 packets to > upperlayers. and sleeps for 1 tick ( which is processing time for 8 > packets ). > > to allow other task to process those packets. > > but the problem is that if there are less no of packets still i have > to sleep for processing time equivalent to 8 packets because i don't > have any machnism to slepp for less than one tick. > > How can i do this ? > > Manjeet > Does the cross bar not have an interrupt line that you can register a handler against? Then you wouldn't have to worry about timer loops at all. Neil --------------------------- Newsgroups: comp.os.vxworks Subject: Queue Placement Date: Fri, 13 Sep 2002 17:35:29 -0400 From: Steven Offenbacher Organization: Johns Hopkins University Applied Physics Lab, Laurel, MD, USA Message-ID: <3D825A21.F854FD3D@jhuapl.edu> Is there someway to get VxWorks to force a queue to be created at a specific memory address ? Is there some efficient way (i.e. without forcing interrupts off or blocked) for a program to get the amount of available queue space ? Thanks --------------------------- Newsgroups: comp.os.vxworks Subject: re: File problem with either DOS or TFFS Date: Fri, 13 Sep 2002 17:59:34 -0400 From: Andray Kaganovsky Organization: WorldCom Canada Ltd. News Reader Service Message-ID: References: On Fri, 13 Sep 2002 04:17:34 -0400, Cory Dixon wrote: [AndrayK]. Are you using DOS_OPT_LONGNAMES with dosFs ? If you do, that will confuse TFFS. Try calling routine tffsDevOptionsSet(), that could help. Andray > Please excuse me all, for my problem is a problem that seems to have > already been beaten to death. I have answered my own question, upgrade > to dosFs 2.0! This was answered in October of 2001 ( search for _TFFS > and dosFS problems_ at google groups). > > Though, I am still curious if anyone else has implemented TFFS on AMD > Flash Am29DL323C. > > I am bothered by the fact that we didn't detect this until late in the > game (as of days ago), and now we have to request dosFs 2.0 from > WindRiver and upgrade all of our systems. It seems to me that upgrading > to dosFs 2.0 is something WindRiver should push, not hide. > > cory > > Cory Dixon wrote: >> Hello all, >> >> I am at the University of Colorado, and we are coming down to crunch >> time on our Three Corner Sat software deadline. And, of course, we have >> run into a major problem. My questions, for now, are if any one can >> tell me if the problem described below could be manifested by DOS, the >> TFFS layer, or possibly memory paging, or as I fear a bad combination >> of the above? Or for that matter, has any one had similar problems, and >> did you find the cause? Finally, is there anyone who has successfully >> implemented TFFS on AMD Flash (Am29DL323C)? >> >> I will try to be as brief and concise as I can. If I have skipped >> something please let me know. >> >> We are running VxWorks 5.4.2 for EmbeddedPlanet's RPX Light board, >> which runs a PowerPC MPC 823 with 16 MB of ram and 16 MB of AMD Flash >> (Am29DL323C) on four chips. I have modified the socket and MTD level >> code to get TFFS up and running on the targets. We have Tornado 2.0.2 >> but use our own makefiles as to avoid Tornado's hidden fields. For >> various reasons, we have divided the 16MB of flash into 5 different >> flash mounts consisting of 8, 4, 2, 1, 1 MB each. Each of them is >> mounted as a separate TFFS drive, with a DOS file system. We use the >> last 1 MB partition, which doesn't get used or have DOS mounted on it, >> for our kernel and is the only thing that has saved us at this point. >> >> We are using dosFs 1.0, and this late in the game changing to 2.0 is a >> drastic move. But, at this point I simply don't know. I am mounting DOS >> with the AUTOSYNC option, but this did not seem to have any affect our >> problem. >> >> The following is a description of the problem as given by one of our >> testers. The problem was first noticed when FTP'ing many files in a row >> to a target, and exhibits the same problem. >> >> ------------------------------------------------------- >> >> The failure is manifested as bad data in a file even though no read or >> write errors have been reported by the file I/O. This is consistent >> with at least some of the operational errors we have seen during >> testing. >> >> A test program has ben run that randomly creates and deletes files of >> random sizes and containing random data patterns. Before a file is >> deleted, the data is read back and compared to the data pattern that >> was written. >> >> 1) The bad data in a corrupted file is always a multiple >> of 2048 bytes even though the file lengths are random. The >> corrupted data in the file is always all zero bytes. >> 2) Within a partition, corruption of an existing file >> always occurs during the creation of a new file in that same >> partition. >> 3) Corruption doesn't occur every time a file is created, >> but the corruption is not rare. It almost always occurs by the >> time 7 files have been created in a partition. Quite commonly it >> occurs on the creation of the 5th file. I have seen it happen on >> the 4th file a couple of times. >> 4) Usually, but not always, the file that is corrupted is the >> last previous file created in the partition before the file >> creation that causes the corruption. (ie corruption doesn't happen >> across partitions) >> >> >> Please, any insight into our problem would be greatly appreciated. Cory >> Dixon >> --------------------------- Newsgroups: comp.os.vxworks Subject: Re: muxDevStart()? Date: 13 Sep 2002 16:18:48 -0700 From: czhao@hns.com (Charlie Zhao) Organization: http://groups.google.com/ Message-ID: <75b047bf.0209131518.3daa96dc@posting.google.com> References: <758a30ff.0209130322.6a686e10@posting.google.com> What is your bootDev? It seems the MUX driver for this device is not working. At least the muxstart function of this driver got some problems. oe_zhou@sohu.com (oe_zhou) wrote in message news:<758a30ff.0209130322.6a686e10@posting.google.com>... > here is one segment of my bsp:(ppc860+vxworks) > /*=================================================*/ > pEnd = endFindByName (bootDev, params.unitNum); > > printf("\n here: endFindByName().\n"); /* added by zhou */ > > if (!attached && (pEnd != NULL)) > { > printf("\n here: muxDevStart. \n"); > if (muxDevStart (pEnd) != OK) > { > printf("Failed to start device %s\n", bootDev); > return (ERROR); > } > printf("\n here: muxIoctl.\n"); > .................... > It can display the " here:muxDevStart." ,but it can't output the > "here: muxIoctl" or "Failed to start device.". And it seems stopping > and does nothing. > Help me! --------------------------- Newsgroups: comp.os.vxworks Subject: TLB MISS ERROR (Q) Date: 13 Sep 2002 18:08:56 -0700 From: essenlee@hanmail.net (LEE MYUNG SUK) Organization: http://groups.google.com/ Message-ID: USING : VxWorks 5.4 mpc860 board question> application program is working, after some time, exception error is occured and rebooted. what's the metter? what is the mean of TLB Miss? please answer me vxworker. < error message > Exception at interrupt level: Implementation Dependent Data TLB Miss Exception current instruction address: 0x00001208 Machine Status Register: 0x08209030 Condition Register: 0x42000040 Regs at 0x1fedc0 Press any key to stop auto-boot... 4 3 2 1 0 auto-booting... NTP > I NAME ENTRY TID PRI STATUS PC SP ERRNO DELAY - ---------- ------------ -------- --- ---------- -------- -------- - ------- ----- tExcTask excTask fe7018 0 PEND c3110 fe6f38 0 0 tLogTask logTask fe4690 0 PEND c3110 fe45b8 0 0 tWdbTask 9462c f48d78 3 PEND 9d4f8 f48bd0 0 0 lanRx lanrx_main f46b50 10 PEND 9d4f8 f46568 33 0 main NTPMAIN f41b18 20 DELAY a17c8 f41a80 0 20 usrPort serial_main ffd948 40 READY a2204 ffd7b8 c0002 0 gpsEngine gps_main f41130 40 READY a17c8 f41098 0 0 usrLcdKey lcd_key_main f3fb90 40 DELAY a17c8 f3fae0 0 1 timeCode tcd_main f3e5f0 40 DELAY a17c8 f3e548 0 1 telnetRx telnetrx_mai f3d050 40 PEND 9d4f8 f3cf10 0 0 tNetTask netTask fa6748 50 PEND 9d4f8 fa6690 3d 0 tSnmpd a4fec f50860 150 PEND 9d4f8 f4feb8 0 0 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: muxDevStart()? Date: Sat, 14 Sep 2002 02:00:19 GMT From: "drdiags" Organization: AT&T Broadband Message-ID: References: <758a30ff.0209130322.6a686e10@posting.google.com> Hello oe_zhou, Charlie Zhao is correct about your failure. I was just going through this with a wrMdpPro8xx BSP with a PPC860 daughter card. In my case, I was trying to add an motfec to the bootrom and the real problem was that in my case I had put a PPC850 daughter card on the eval board. Thus the motFecStart() was hanging trying to read back from the hardware. In my case, it was the fact that the PPC850 didn't support the MII Fast Ethernet controller. After I RTFM and put in the right ICM (PPC860), the interface was initialized and all was good. Doubt this is your problem, but I would take Charlie's advice and if you have the driver source, isolate what call in the xxxStart() routine is hanging. HTH "oe_zhou" wrote in message news:758a30ff.0209130322.6a686e10@posting.google.com... > here is one segment of my bsp:(ppc860+vxworks) > /*=================================================*/ > pEnd = endFindByName (bootDev, params.unitNum); > > printf("\n here: endFindByName().\n"); /* added by zhou */ > > if (!attached && (pEnd != NULL)) > { > printf("\n here: muxDevStart. \n"); > if (muxDevStart (pEnd) != OK) > { > printf("Failed to start device %s\n", bootDev); > return (ERROR); > } > printf("\n here: muxIoctl.\n"); > .................... > It can display the " here:muxDevStart." ,but it can't output the > "here: muxIoctl" or "Failed to start device.". And it seems stopping > and does nothing. > Help me! --------------------------- Newsgroups: comp.os.vxworks Subject: Re: tftpXfer problems :Boot/Loading a VxWorks Image Date: Sat, 14 Sep 2002 02:09:03 GMT From: "drdiags" Organization: AT&T Broadband Message-ID: <2Twg9.311734$aA.54523@sccrnsc02> References: Hello Shiv, Don't you need to have the TFTP Daemon running on your Windows host and not the WFTPD, since they use different ports. If you have the www.ethereal.com network analyzer, you should see the request going out from the target, though I am surprised you didn't get an error pack somewhere indicating a time-out. There is a TFTP daemon in the %WIND_BASE%/host/x86-win32/bin directory called Tftpd32.exe? HTH. (If I misread your post, sorry.....) "Shiv" wrote in message news:altkii$1a5$1@overload.lbl.gov... > > Hi there ! > > I'm in the process of downloading a VxWorks Image from the network with my > bootrom code. (Win2000 host, mpc855T PowerPC target) > > I've verified the boot parameters and I've made sure the WFTPD program is > configured correctly on the host-side. > > TftpXfer() (the vxworks BSP module netload()'s routine that initiates t'fer) > doesn't give me an error message, yet i see no indication of an FTP transfer > (via the logging screen on the WFTPD program) ? Has anyone experienced a > similar problem in getting their TFTP to work in their bootcode? > > Would appreciate any advice I could gather !! > > Thanks, > Shiv --------------------------- Newsgroups: comp.os.vxworks Subject: Re: any website on vxworks faq? Date: 13 Sep 2002 20:26:08 -0700 From: vijay.peshkar@wipro.com (Vijay Peshkar) Organization: http://groups.google.com/ Message-ID: References: <290ade1c.0209130001.6751ea62@posting.google.com> <3D81BD5A.B7312DBC@bellsouth.net> joe durusau wrote in message news:<3D81BD5A.B7312DBC@bellsouth.net>... > tzeghuan wrote: > > > any website on vxworks faq? > > thks > > I can't spell the guy's name, but the pointer to the FAQ is posted > here periodically. A google search will surely show it. > > Speaking only for myself, > > Joe Durusau http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html --------------------------- Newsgroups: comp.os.vxworks Subject: i will join this newsgroup! Date: 13 Sep 2002 20:40:27 -0700 From: my2008post@hotmail.com (chris) Organization: http://groups.google.com/ Message-ID: <3cc05ce9.0209131940.1eb246b0@posting.google.com> i will join this newsgroup! Sincerely, Chris come from china. --------------------------- Newsgroups: comp.os.vxworks Subject: Placing Tornado Project files into RCS Date: Fri, 13 Sep 2002 23:45:36 -0500 From: "fred" Organization: Posted via Supernews, http://www.supernews.com Message-ID: Reply-To: "fred" I currently have all my application code, drivers, and BSP all under version control (we're using PVCS Version Manager). I would also like to put the Tornado project files under PVCS but I've found that they contain hard coded paths to the files included in the project (the Tornado generated files and the ones picked up from the BSP all use the $WIND_BASE environment variable instead of hardcoded paths). The hard coded paths will cause problems for the person checking out the project files if they didn't follow the exact directory structure (using another drive for example). I'm told that the dependencies file will also contain hard coded paths. Right now we will enforce a rule that the application software (all the files that would be included in the Tornado project) reside in a directory tree anchored alongside the Tornado tree. We could force each developer to define another environment variable (that would identical to WIND_BASE accept the trailing "Tornado" would be removed). The project files could be hand edited to use this variable -- not very appealing. The environment variable could be dynamically generated using a tcl script and the current value of WIND_BASE and still have the hand edited project files use this variable to point to the included files in the project. All this sounds like allot of work and I'm not sure if the project files are rewritten when you save them (blowing away the hand edits). I'm also not sure how to get a tcl script to run at startup or how to handle the dependencies file. Maybe it's not worth the effort. Thanks, David --------------------------- Newsgroups: comp.os.vxworks Subject: Placing Tornado Project Files into RCS Date: Sat, 14 Sep 2002 00:00:44 -0500 From: "fred" Organization: Posted via Supernews, http://www.supernews.com Message-ID: Reply-To: "fred" I have my application, drivers, and BSP all under revision control (we use PVCS Version Manager) and would like to place my Tornado project files there also. The problem is that the project files and dependencies files both use hard coded paths for the application files that I've included in the Tornado project. We have a directory tree for our project such that each user will have the application ode in a fixed location relative to the Tornado tree, but they may have Tornado on different drives (drive D: rather than C:, for example). With hardcoded paths this will cause the project files to be potentially unusable for people checking them out form PVCS. I've never done this before but I'm told that you can setup tcl startup scripts that run each time Tornado starts. I can that a tcl script could set an environment variable (based on WIND_BASE) to point to the application software tree and the project files edited to remove hard coded paths and use the environment variable. I'm not sure how to handle the dependencies file since it is dynamically built. All this sounds like allot of work and something that would be very hard to maintain over new releases of Tornado. Does anyone have a slick solution to the problem? Thanks David --------------------------- Newsgroups: comp.os.vxworks Subject: Re: muxDevStart()? Date: 14 Sep 2002 04:02:07 -0700 From: oe_zhou@sohu.com (oe_zhou) Organization: http://groups.google.com/ Message-ID: <758a30ff.0209140302.7594aa6a@posting.google.com> References: <758a30ff.0209130322.6a686e10@posting.google.com> My bootDev is "motfec". And my cpu is 860T. But it seems still don't work! --------------------------- Newsgroups: comp.os.vxworks Subject: disk cache error (new posting) Date: 14 Sep 2002 04:34:06 -0700 From: ritalin@comcast.net (Rita) Organization: http://groups.google.com/ Message-ID: Hello, When we write to our disk, after a long while, we see the following error message: copy "0310FEF6.DRV","abc" 0x8697fcb0 (tShell): disk cache error: device 87f936f0 block -1 errno 0, disk read failed copy: error reading file after copying 327680 bytes. value = -1 = 0xffffffff - -> copy "0310FEF6.DRV","abc" 0x8697fcb0 (tShell): disk cache error: device 87f936f0 block -1 errno c0003, disk read failed copy: error reading file after copying 327680 bytes. value = -1 = 0xffffffff It seems to me that the block number is out of range. Does anyone know what might have cause this problem? It's not an easy reproducible problem, and we are using scsi flash drive. We've used the same thing for over 2 years, and never saw the problem before. Thanks. Rita --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Placing Tornado Project files into RCS Date: Sat, 14 Sep 2002 08:29:33 -0700 From: "Michael R. Kesti" Organization: MK Associates Message-ID: <3D8355DD.7D0B08CE@gv.net> References: Reply-To: mkesti@gv.net fred wrote: >I currently have all my application code, drivers, and BSP all under version >control (we're using PVCS Version Manager). I would also like to put the >Tornado project files under PVCS but I've found that they contain hard coded >paths to the files included in the project (the Tornado generated files and >the ones picked up from the BSP all use the $WIND_BASE environment variable >instead of hardcoded paths). The hard coded paths will cause problems for >the person checking out the project files if they didn't follow the exact >directory structure (using another drive for example). I'm told that the >dependencies file will also contain hard coded paths. > >Right now we will enforce a rule that the application software (all the >files that would be included in the Tornado project) reside in a directory >tree anchored alongside the Tornado tree. We could force each developer to >define another environment variable (that would identical to WIND_BASE >accept the trailing "Tornado" would be removed). The project files could be >hand edited to use this variable -- not very appealing. The environment >variable could be dynamically generated using a tcl script and the current >value of WIND_BASE and still have the hand edited project files use this >variable to point to the included files in the project. All this sounds >like allot of work and I'm not sure if the project files are rewritten when >you save them (blowing away the hand edits). I'm also not sure how to get a >tcl script to run at startup or how to handle the dependencies file. Maybe >it's not worth the effort. I have successfully dealt with this under Windows-NT by specifying a mapped drive or two that the project files must reference. This allow individual users to install their Tornado and project directories on any physical drive/directory they wish, while retaining the commonality required for the RCS based project files. - -- ======================================================================== Michael Kesti | "And like, one and one don't make | two, one and one make one." mkesti@gv.net | - The Who, Bargain --------------------------- Newsgroups: comp.os.vxworks Subject: Exception when booting up a standalone image Date: 14 Sep 2002 11:07:26 -0700 From: vividh@hotmail.com (rasayani) Organization: http://groups.google.com/ Message-ID: <28809481.0209141007.63d22c24@posting.google.com> I am having a strange problem. I have two Motorola CPCI MCP750 boards. I run standalone vxWorks image. On one of the boards after tftping the image from a server, it gives a exception. The same image on the other board works fine. Is there a way to debug where the exception occurs? I don't get the target shell even, so I presume it fails in some OS initialization part. thanks, vividh --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How Change a TTL Value of ICMP Packet ? Date: 14 Sep 2002 11:20:47 -0700 From: vividh@hotmail.com (rasayani) Organization: http://groups.google.com/ Message-ID: <28809481.0209141020.3ad0a393@posting.google.com> References: try char optval; icmpSock = socket(AF_INET, SOCK_RAW, IPPROTO_ICMP); optval = 1; setsockopt (icmpSock, IPPROTO_IP, IP_TTL, &optval, sizeof(char)); vividh "Jay" wrote in message news:... > Hi, Everyone. > I Try to make a route trace program on a vxWorks, but I can't change the TTL > value of ICMP Packet. > My test code is, > > int optval; > > icmpSock = socket(AF_INET, SOCK_RAW, IPPROTO_ICMP); > optval = 1; > setsockopt (icmpSock, IPPROTO_IP, IP_TTL, &optval, sizeof(optval)); > > Does anyone know how I can change a TTL Value? > > Thanks in advance! > Jay. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: TLB MISS ERROR (Q) Date: 14 Sep 2002 11:48:19 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0209141048.4038caea@posting.google.com> References: Hello, essenlee@hanmail.net (LEE MYUNG SUK) wrote in message news:... > USING : VxWorks 5.4 mpc860 board > > question> > application program is working, after some time, exception error is > occured and rebooted. > what's the metter? what is the mean of TLB Miss? A TLB miss exception is meant to cause the operating system to read the page tables and load a translation from there into the TLB. The TLB is a cache for page table entries, necessary to improve performance. The exception is normally handled by VxWorks, but judging from the address in the exception message something has failed in the handler itself. Are you running with the cache and MMU enabled, or not? Have you changed these settings in your BSP? What revision of the 860 silicon are you using? Have you checked the BSP's target.nr file to see if there are any known problems with your particular silicon revision - the 8xx family is perhaps the worst CPU family out there in terms of serious silicon-related errata. > please answer me vxworker. > > < error message > > > Exception at interrupt level: This is perhaps a more important statement than the exact exception that occurred. One of your interrupt service routines is causing an exception for some reason. You should try to find out which ISR it is by testing them thoroughly one at a time. Of course, another possibility is that some part of your application is corrupting system memory somewhere. You could try to examine the stack using the stack pointer in the regs (see the address in the last line of the exception message for where the regs were saved, and the REG_SET definition in target/h/arch/ppc/regsPpc.h for the layout of the reg set - the stack pointer for PPC is GPR 1). HTH, John... > Implementation Dependent Data TLB Miss > Exception current instruction address: 0x00001208 > Machine Status Register: 0x08209030 > Condition Register: 0x42000040 > Regs at 0x1fedc0 > > > Press any key to stop auto-boot... > 4 3 2 1 0 > auto-booting... > > > > NTP > I > > NAME ENTRY TID PRI STATUS PC SP > ERRNO DELAY > ---------- ------------ -------- --- ---------- -------- -------- > ------- ----- > tExcTask excTask fe7018 0 PEND c3110 fe6f38 > 0 0 > tLogTask logTask fe4690 0 PEND c3110 fe45b8 > 0 0 > tWdbTask 9462c f48d78 3 PEND 9d4f8 f48bd0 > 0 0 > lanRx lanrx_main f46b50 10 PEND 9d4f8 f46568 > 33 0 > main NTPMAIN f41b18 20 DELAY a17c8 f41a80 > 0 20 > usrPort serial_main ffd948 40 READY a2204 ffd7b8 > c0002 0 > gpsEngine gps_main f41130 40 READY a17c8 f41098 > 0 0 > usrLcdKey lcd_key_main f3fb90 40 DELAY a17c8 f3fae0 > 0 1 > timeCode tcd_main f3e5f0 40 DELAY a17c8 f3e548 > 0 1 > telnetRx telnetrx_mai f3d050 40 PEND 9d4f8 f3cf10 > 0 0 > tNetTask netTask fa6748 50 PEND 9d4f8 fa6690 > 3d 0 > tSnmpd a4fec f50860 150 PEND 9d4f8 f4feb8 > 0 0 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: muxDevStart()? Date: Sat, 14 Sep 2002 18:50:11 GMT From: "drdiags" Organization: AT&T Broadband Message-ID: References: <758a30ff.0209130322.6a686e10@posting.google.com> <758a30ff.0209140302.7594aa6a@posting.google.com> oe_zhou, "oe_zhou" wrote in message news:758a30ff.0209140302.7594aa6a@posting.google.com... > My bootDev is "motfec". And my cpu is 860T. > But it seems still don't work! Maybe listing more pertinent information will help you resolve your problem. 1 - What version of vxWorks/Tornado? 2 - Is this your own BSP or one of WindRiver's? 3 - Is your board a custom or reference board? 4 - Does this BSP allow you to enable another END? 5 - Have you contacted either Wind River or the BSP developer for help? 6 - Have others at your company using the same setup successfully installed the motfec in their bootrom? In other words is this just a single case where you are having problems and others don't or you have others who are facing the same problem. 7 - Have you checked Wind Surf for any possible motFecEnd object updates? If all else fails, you should contact your local FAE to see if you can get the source to debug your problem or you should contact support. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: any website on vxworks faq? Date: 14 Sep 2002 12:00:33 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0209141100.23f8533e@posting.google.com> References: <290ade1c.0209130001.6751ea62@posting.google.com> Hello, The FAQ link has already been posted, but there are some other resources that can be useful when searching for answers: 1) Google archives this group and allows searching: http://groups.google.com/groups?hl=en&lr=&safe=off&group=comp.os.vxworks 2) WindSurf has a search tool that will search this group as well as tech tips, application notes, published SPRs and product manuals. http://www.windriver.com/windsurf/ 3) WindSurf also has a community forum, which, though less active than this one could have the answer you need. HTH, John... tgisfree@yahoo.com (tzeghuan) wrote in message news:<290ade1c.0209130001.6751ea62@posting.google.com>... > any website on vxworks faq? > thks --------------------------- Newsgroups: comp.os.vxworks Subject: symSyncLib problems Date: 14 Sep 2002 17:33:48 -0700 From: ands1234321@yahoo.com (Anthony) Organization: http://groups.google.com/ Message-ID: <811802c7.0209141633.283a1369@posting.google.com> I am using Tornado 2.0/VxWorks 5.4 with a Motorola PPMC750 processor card and BSP. On bootup, I load an object module using a startup script. The loading works fine and I can see the symbols in the module from the target shell. However I can not see the symbols from the Tornado shell, even though I have symbol synchronization turned on, and I load the module with the LOAD_ALL_SYMBOLS flag. There are no unresolved symbols in the module I am loading. Directly from the symSyncLib documentation: "When the task starts, it synchronizes target and host symbol tables so that every module loaded on the target before the target server was started can be seen by the host tools. This feature is particularly useful if VxWorks is started with a target-based startup script before the target server has been launched." This is EXACTLY the behavior I am NOT seeing. When the task starts, it reports "Added target_modules to target-server.....done", but I still can not see the symbols with the Tornado shell. However if I load the very same module after the task starts, everything gets properly synchronized. The purpose of all of this is to automatically load an object module on bootup, and have all of the module symbols available to the host tools. (Tornado shell, Crosswind, etc.) Perhaps I am going about this incorrectly? Thanks in advance for any help you can provide. --------------------------- Newsgroups: comp.os.vxworks Subject: muxDevStart() problems! Date: 14 Sep 2002 19:39:13 -0700 From: oe_zhou@sohu.com (oe_zhou) Organization: http://groups.google.com/ Message-ID: <758a30ff.0209141839.ba9172@posting.google.com> In my bsp(VxWorks for ppc860T) the function muxDevStart hangs and doesn't return. I boot from "motfec(0,0)". Who can tell me what's happened in muxDevStart() and which function it invokes? Adanced thanks! --------------------------- Newsgroups: comp.os.vxworks Subject: intLock + printf Date: 14 Sep 2002 23:09:52 -0700 From: yahoo_m1@hotmail.com (John) Organization: http://groups.google.com/ Message-ID: Hi All, What happens if I call intLock () and then try to do a printf? Will it release the lock? Or does it raise an exception and what kind of exception? Something like this... lockKey = intLock (); . . do some processing; . . intUnlock (lockKey); I was suspecting a part of my code which has this sequence and would like to get some suggestions / comments on this. Your replies/ideas are highly appreciated. John. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Exception Handling in vxWorks. Date: 15 Sep 2002 02:12:55 -0700 From: tgisfree@yahoo.com (tzeghuan) Organization: http://groups.google.com/ Message-ID: <290ade1c.0209150112.485fe964@posting.google.com> References: <290ade1c.0209122144.5566acf4@posting.google.com> <488e459a.0209130932.1c877e18@posting.google.com> Hello, Thanks for the reply. It just a simple C++ codes for testing: try() { //create a divide by zero error here. } catch(...) { //print a error statement here. } with f-exception option enabled. I saw a similar question posted here a year ago, the reply was rather vague. Hope I didn't miss out some other settings. Thanks in advance. tzeghuan john_94501@yahoo.com (John) wrote in message news:<488e459a.0209130932.1c877e18@posting.google.com>... > Hello, > > Your question is so vague that I doubt you will get a useful answer. > Try posting some more information... here are a few questions for you > to think about: > > 1) Did you mean hardware exception handling or C++ exception handling > (very different things)? > > 2) What doesn't work exactly? Does it not compile? Are there errors > when you download it? Does it cause errors when you run it? > > 3) Have you got a small code fragment that you could post here? > > 4) Which architecture/BSP are you using? > > 5) What compiler flags did you use when you compiled it? > > 6) Have you read the manual? > > 7) Have you searched the archives of this group for possible answers? > > 8) Have you read the FAQ for this group? > > Please remember that people in this forum are not paid to answer your > questions, and are usually very busy with their own work. If you want > to get a good answer you will need to think up front about the kinds > of information that are relevant; most of the people here do not have > the time, nor the desire to extract the information from you over a > series of postings. > > That said, if you post enough information about your problem, often > you will get an answer. > > HTH, > > John... > > tgisfree@yahoo.com (tzeghuan) wrote in message news:<290ade1c.0209122144.5566acf4@posting.google.com>... > > Hi all, > > > > Recently i tried out some exception handling in both windows and vxworks. > > The same piece of code works in win environment but not in vxworks. > > Can anyone know about this give some clue? > > Is there a need to enable exception handling or any thing like that? > > > > Thanks in advance. > > > > regards > > tzeghuan --------------------------- Newsgroups: comp.os.vxworks Subject: Re: intLock + printf Date: Sun, 15 Sep 2002 12:10:55 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: John, It seems to me you are putting too much stuff in between intLock/intUnlock - this is bad practice because it takes away the flexibility to control and prioritize resources later. Specifically if you put a printf(), it will prolong the time of lockout, as printf() could take several milliseconds to execute (depending on CPU and format string). Also, printf() could return immediately, or might need to WAIT if the output buffer (e.g. of the serial console) is full. In order to avoid losing data, the write() call (called by printf() ), will PEND until there is sufficient space available in the output buffer. When a task in interrupt lockout pends, interrupts are re-enabled, until that task gets to run again. By the way, replacing printf() with logMsg() won't make a difference, because logMsg() will too have to wait if the logTask msgQ is full (unless it was called from an ISR), and will unlock interrupts when waiting. What you need to do is to all your critical stuff before unlocking interrupts, and move the printf() and other non-crticical calls to follow intUnlock. Hope this helps, - - Leonid "John" wrote in message news:baa5411a.0209142209.43e9106a@posting.google.com... > Hi All, > What happens if I call intLock () and then try to do a printf? Will it > release the lock? Or does it raise an exception and what kind of > exception? > > Something like this... > > lockKey = intLock (); > . > . > do some processing; > . > . > intUnlock (lockKey); > > I was suspecting a part of my code which has this sequence and would > like to get some suggestions / comments on this. > > Your replies/ideas are highly appreciated. > > John. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Exception Handling in vxWorks. Date: 15 Sep 2002 12:06:26 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0209151106.3c45b4a@posting.google.com> References: <290ade1c.0209122144.5566acf4@posting.google.com> <488e459a.0209130932.1c877e18@posting.google.com> <290ade1c.0209150112.485fe964@posting.google.com> Hello, You are confusing C++ exceptions with hardware exceptions. The C++ exception handling mechanism does not catch hardware exceptions (and it is not meant to according to the specification - the fact that MS C++ does this is because they have implemented a non-standard language extension to do this). You did not specify the CPU or BSP that you are using (see my previous comments about the minimum information you should supply when posting questions here), so I cannot say if the architecture you have even generates an exception for divide by zero - most software floating point implementations do not, they simply return one of the IEEE defined NaN values. Assuming that the architecture you are using does generate hardware exceptions for divide by zero, you will need to catch the error using signals. If you know enough about the internals of the compiler, then you can create a signal handler that will throw you a C++ exception, but it is not as simple as just throwing an C++ exception from the signal handler. HTH, John... tgisfree@yahoo.com (tzeghuan) wrote in message news:<290ade1c.0209150112.485fe964@posting.google.com>... > Hello, > > Thanks for the reply. > > It just a simple C++ codes for testing: > > try() > { > //create a divide by zero error here. > } > catch(...) > { > //print a error statement here. > } > > with f-exception option enabled. > > I saw a similar question posted here a year ago, the reply was rather vague. > > Hope I didn't miss out some other settings. > > Thanks in advance. > > tzeghuan > > > john_94501@yahoo.com (John) wrote in message news:<488e459a.0209130932.1c877e18@posting.google.com>... > > Hello, > > > > Your question is so vague that I doubt you will get a useful answer. > > Try posting some more information... here are a few questions for you > > to think about: > > > > 1) Did you mean hardware exception handling or C++ exception handling > > (very different things)? > > > > 2) What doesn't work exactly? Does it not compile? Are there errors > > when you download it? Does it cause errors when you run it? > > > > 3) Have you got a small code fragment that you could post here? > > > > 4) Which architecture/BSP are you using? > > > > 5) What compiler flags did you use when you compiled it? > > > > 6) Have you read the manual? > > > > 7) Have you searched the archives of this group for possible answers? > > > > 8) Have you read the FAQ for this group? > > > > Please remember that people in this forum are not paid to answer your > > questions, and are usually very busy with their own work. If you want > > to get a good answer you will need to think up front about the kinds > > of information that are relevant; most of the people here do not have > > the time, nor the desire to extract the information from you over a > > series of postings. > > > > That said, if you post enough information about your problem, often > > you will get an answer. > > > > HTH, > > > > John... > > > > tgisfree@yahoo.com (tzeghuan) wrote in message news:<290ade1c.0209122144.5566acf4@posting.google.com>... > > > Hi all, > > > > > > Recently i tried out some exception handling in both windows and vxworks. > > > The same piece of code works in win environment but not in vxworks. > > > Can anyone know about this give some clue? > > > Is there a need to enable exception handling or any thing like that? > > > > > > Thanks in advance. > > > > > > regards > > > tzeghuan --------------------------- Newsgroups: comp.os.vxworks Subject: Looking for a SNMPv2 Trap example Date: 15 Sep 2002 13:12:29 -0700 From: akiggins@celitesystems.com (andrew) Organization: http://groups.google.com/ Message-ID: <2eced96.0209151212.15361035@posting.google.com> 1 and all, Plowing through the ( as normal spotty ) SNMP docs for Epilogue I can't find an example of how to transmit a SNMPv2 NOTIFICATION from the agent. Can anyone point me at some (re)sources ? thanks in advance andrew --------------------------- Newsgroups: comp.os.vxworks Subject: Re: any website on vxworks faq? Date: Sun, 15 Sep 2002 11:00:44 +0800 From: "Jim.chu" Organization: Bentium Ltd. (CN99) Message-ID: References: <290ade1c.0209130001.6751ea62@posting.google.com> Here : http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html "tzeghuan" wrote in message news:290ade1c.0209130001.6751ea62@posting.google.com... > any website on vxworks faq? > thks --------------------------- Newsgroups: comp.os.vxworks Subject: muxDevStopAll() Date: 15 Sep 2002 22:10:55 -0700 From: oe_zhou@sohu.com (oe_zhou) Organization: http://groups.google.com/ Message-ID: <758a30ff.0209152110.3374b181@posting.google.com> My system consits of ppc860T and VxWorks. The function, muxDevStopAll, in the bootConfig.c file hangs. Why? I use cpm(0,0) to boot and i set up the motfec. --------------------------- Newsgroups: comp.os.vxworks Subject: equivalent of unix api' getdomainname(), setdomainname() Date: 15 Sep 2002 23:28:58 -0700 From: vijay.peshkar@wipro.com (Vijay Peshkar) Organization: http://groups.google.com/ Message-ID: Hi All, Reference Manual does not list these api's. Any ideas on how to implement them?. Thanks and Regards, Vijay --------------------------- Newsgroups: comp.os.vxworks Subject: Just received: Tornado 2.2 Date: Mon, 16 Sep 2002 07:37:56 +0000 (UTC) From: Johan Borkhuis Organization: Proxim Corp. Message-ID: I just received my upgrade to Tornado 2.2 for PPC (Win2k). Some first (incomplete) impressions. - - VxWorks version 5.5. - - License manager: Tornado now includes a license manager, with a fixed number of floating or fixed licenses. If you have support you can choose to convert your licenses to floating or node-locked, but support for floating will be more expensive. - - Installation: One big improvement: you can select which filetypes will be associated with Tornado! So no more resetting of the filetypes to point to a real editor after a Tornado install. - - Link scripts: link scripts are used instead of the buildin link scripts - - GCC version 2.96 - - Codetest: not yet available for PPC, but an upgrade has to be purchased, because of new features (around $500) - - Documentation includes a list of error codes and a migration guide - - Paper version of documentation has to be ordered seperately, but it is still free - - GNU sources are not included in the standard distribution. I don't have any performance numbers yet, but I will try to start working with this release as soon as possible. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: symSyncLib problems Date: Mon, 16 Sep 2002 09:29:01 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <811802c7.0209141633.283a1369@posting.google.com> Reply-To: "Michael Lawnick" Sender: mlawnick@p5088ebb4.dip.t-dialin.net Hi Anthony, this is a known 'feature': The first module loaded by startup-script won't be synchronised. Loading a dummy object should fix it. HTH Michael "Anthony" schrieb im Newsbeitrag news:811802c7.0209141633.283a1369@posting.google.com... > I am using Tornado 2.0/VxWorks 5.4 with a Motorola PPMC750 processor > card and BSP. > > On bootup, I load an object module using a startup script. The > loading works fine and I can see the symbols in the module from the > target shell. However I can not see the symbols from the Tornado > shell, even though I have symbol synchronization turned on, and I load > the module with the LOAD_ALL_SYMBOLS flag. There are no unresolved > symbols in the module I am loading. > > Directly from the symSyncLib documentation: > > "When the task starts, it synchronizes target and host symbol tables > so that every module loaded on the target before the target server was > started can be seen by the host tools. This feature is particularly > useful if VxWorks is started with a target-based startup script before > the target server has been launched." > > This is EXACTLY the behavior I am NOT seeing. When the task starts, > it reports "Added target_modules to target-server.....done", > but I still can not see the symbols with the Tornado shell. However > if I load the very same module after the task starts, everything gets > properly synchronized. > > The purpose of all of this is to automatically load an object module > on bootup, and have all of the module symbols available to the host > tools. (Tornado shell, Crosswind, etc.) Perhaps I am going about this > incorrectly? > > Thanks in advance for any help you can provide. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Exception when booting up a standalone image Date: Mon, 16 Sep 2002 09:32:21 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <28809481.0209141007.63d22c24@posting.google.com> Reply-To: "Michael Lawnick" Sender: mlawnick@p5088ebb4.dip.t-dialin.net Hi rasayani, check that processor number' in your bootline differ. The cause of your crash could be the initialisation of shared memory on PCI-bus. If processor numbers don't differ, invalid addresses with undifined behaviour might occur. HTH Michael "rasayani" schrieb im Newsbeitrag news:28809481.0209141007.63d22c24@posting.google.com... > I am having a strange problem. I have two Motorola CPCI MCP750 boards. > I run standalone vxWorks image. On one of the boards after tftping the > image from a server, it gives a exception. The same image on the other > board works fine. Is there a way to debug where the exception occurs? > I don't get the target shell even, so I presume it fails in some OS > initialization part. > > thanks, > vividh --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Just received: Tornado 2.2 Date: Mon, 16 Sep 2002 10:36:36 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: Reply-To: "Michael Lawnick" Sender: mlawnick@p5088ebb4.dip.t-dialin.net In addition: - - contains DosFs2, - - optional TFFS should work up to 2GB (not tested yet) - - still old unfixed bugs (e.g. in WindView) Regards Michael "Johan Borkhuis" schrieb im Newsbeitrag news:Xns928B61683FC54jborkhuisproximcom@192.19.193.66... > I just received my upgrade to Tornado 2.2 for PPC (Win2k). Some first > (incomplete) impressions. > > - VxWorks version 5.5. > - License manager: Tornado now includes a license manager, with a fixed > number of floating or fixed licenses. If you have support you can choose > to convert your licenses to floating or node-locked, but support for > floating will be more expensive. > - Installation: One big improvement: you can select which filetypes will > be associated with Tornado! So no more resetting of the filetypes to point > to a real editor after a Tornado install. > - Link scripts: link scripts are used instead of the buildin link scripts > - GCC version 2.96 > - Codetest: not yet available for PPC, but an upgrade has to be purchased, > because of new features (around $500) > - Documentation includes a list of error codes and a migration guide > - Paper version of documentation has to be ordered seperately, but it is > still free > - GNU sources are not included in the standard distribution. > > I don't have any performance numbers yet, but I will try to start working > with this release as soon as possible. > > Kind regards, > Johan > > -- > o o o o o o o . . . _____________________________ > o _____ || Johan Borkhuis | > .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | > >(________|__|_[_________]_|__________________________| > _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` > === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === > --------------------------- Newsgroups: comp.os.vxworks Subject: windview and trigger tool doubts Date: 16 Sep 2002 03:37:01 -0700 From: vxnerd@rediffmail.com (subbarayan) Organization: http://groups.google.com/ Message-ID: dear mentors, 1)Can anyone tell me how to configure windview to view a graph of user defined tasks?My typical appln requires that the graph should show the name of tasks used in my programme.I have been unable to find my task names in graph where as i find generic tasknames as t14,t15,t16 etc and some other tasks like tnettask,idle task,texectask etc.Can anyone point me to info abt these various tasks displayed in graph of windview?I created a root task and spawned a task called taskone in my programme.I am unable to find taskone in my graph of windview.Is it appearing in some other name or what settings should i make to view info abt taskone in windview? the same code shown below was the one i tried.I expected taskone to be viewed in graph after calling root and executing it.I went through the manual but was unable to find the info i was looking for. 2)I was going through the triggering tool manual.After going through it ,i tried to use it to trigger windview for a sample application to start logging.I did the following: Opened trigger tool. Then defined context matches as tid1 which is my taskid of my task called taskone.defined actions as startwv. Then the following is my code: #include #include"vxWorks.h" #include"taskLib.h" void taskone(); void tasktwo(); int tid1,tid2; void root() { tid1=taskSpawn("t1",90,0x100,20000,FUNCPTR(taskone),0,0,0,0,0,0,0,0,0,0); tid2=taskSpawn("t2",91,0x100,20000,FUNCPTR(tasktwo),0,0,0,0,0,0,0,0,0,0); } void taskone() { printf("%d\n",taskIdSelf()); printf("i am first task in this pc \n"); printf("tasktwo created"); } void tasktwo() { printf("i am second task in this pc \n"); printf("releasing cpu to taskone \n"); } I am unable to find the trigger working inspite of me calling either root from shell or by calling taskone directly from shell.Is there some configuration error in it?I went through the manual but was unable to find the info i was looking for. I will be thankful to persons clarifying my doubt.I use tornado2.0 and i tried in simulator version. --------------------------- Newsgroups: comp.os.vxworks Subject: Can't launch target server Date: 16 Sep 2002 03:38:07 -0700 From: emsindhwani@yahoo.com.sg (Mohit Sindhwani) Organization: http://groups.google.com/ Message-ID: Hi. I'm using Tornado II/ VxWorks for the Tricore target. The board in question does not have a network controller. I'm trying to connect to it over the serial connection, but nothing seems to work. When I try to launch the target server from the GUI, nothing happens. When I execute 'tgtsvr -V target -d com1 -B wdbserial -bps 9600' on a command prompt, it dies with the following error: 'The procedure entry point symAddDa could not be located in the dynamic link library tgtmgt.dll' I'm more than a bit confused about how to proceed. Any pointers on where to look? Thanks, Mohit. p.s. I'm using Windows XP (if that helps) and have installed the patch that is supposed to be required to run Tornado under Windows 2000. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: TCP read problem in VxWorks ......! Date: Mon, 16 Sep 2002 08:25:24 -0400 From: Neil Horman Organization: WEBUSENET.com Message-ID: <3D85CDB4.1030606@lvl7.com> References: Reply-To: nhorman@lvl7.com Could also be a read call issue. When you call read on a socket for X bytes of data, you may actaully read anything from 0 to X bytes of data, dependent on your socket options (and default socket options are very platform specific). If you have a buffer of 100 bytes allocated and do a read(sockfd, buf, 100) it may be that you only read 80 bytes, and the remaining 20 were unchanged. You're just interpreting them as "junk" from the packet. Check the return code from the read call, that will tell you how many bytes you actually read, and you will know for sure if your garbage data is comming from the packet, or from an uninitalized buffer in your code. Neil vvkoduri@hss.hns.com wrote: > Hi Experts ! > Here we are facing a very basic problem in VxWorks, related to TCP Socket > read. > > Here we wrote a small appl by using TCP sockete, will be running on Solaris 2.6, > remote application will be > running on Motorola PPC 7400, with VxWorks. > > Our Application always expects 16bytes header first and then User Payload, > length will be specified in the Header 16 bytes. > > Here When the remote appl is running on the Solaris machine, Our appl is getting > the message perfectly from remote application. > > Lets say .. header consists of buflen = 100, > > read(sockfd, buf, 100) working well and no trailing junk. > > If we run the remote Appl on VxWorks, after reading the bytes specified in the > header, still some bytes are coming. > > Some garbage bytes are trailing the message. > > PLease suggest anything wrong with sokcet operations on Vxworks .. let say send > or write functions. > > With thanx and regards, > Ramakrishna kvv. > > DISCLAIMER: This message is proprietary to Hughes Software Systems > Limited (HSS) and is intended solely for the use of the individual > to whom it is addressed. It may contain privileged or confidential > information and should not be circulated or used for any purpose other > than for what it is intended. If you have received this message in error, > please notify the originator immediately. If you are not the intended > recipient, you are notified that you are strictly prohibited from using, > copying, altering, or disclosing the contents of this message. HSS accepts > no responsibility for loss or damage arising from the use of the information > transmitted by this email including damage from virus. > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Can't launch target server Date: Mon, 16 Sep 2002 13:16:07 +0000 (UTC) From: Johan Borkhuis Organization: Proxim Corp. Message-ID: References: emsindhwani@yahoo.com.sg (Mohit Sindhwani) wrote: > p.s. I'm using Windows XP (if that helps) and have installed the patch > that is supposed to be required to run Tornado under Windows 2000. This could be the problem. Tornado 2.2 is available and this should support XP, so you might consider upgrading. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: 1 second becomes 4 seconds! Date: 16 Sep 2002 07:13:53 -0700 From: welson_sun@yahoo.com (Welson Sun) Organization: http://groups.google.com/ Message-ID: <9d9eae54.0209160613.3043746e@posting.google.com> I have got a problem with our 860 board: When the vxWorks is starting up and counting down, the interval is not 1 second, actually it is 4 seconds! I wrote a POSIX timer program and set the timer expiration duration to be 1 second, but it actually expires after 4 seconds! What could be the cause, I would like to know the root cause of this problem. I have tried sysClkRateSet(sysClkRateGet()*4) but still the case. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Just received: Tornado 2.2 Date: Mon, 16 Sep 2002 10:22:15 -0400 From: "Stephane Ross" Message-ID: References: I heard that they made modifications in the network stack to improve it's performance/reliability. Will you be looking into that? Stephane Ross "Johan Borkhuis" wrote in message news:Xns928B61683FC54jborkhuisproximcom@192.19.193.66... > I just received my upgrade to Tornado 2.2 for PPC (Win2k). Some first > (incomplete) impressions. > > - VxWorks version 5.5. > - License manager: Tornado now includes a license manager, with a fixed > number of floating or fixed licenses. If you have support you can choose > to convert your licenses to floating or node-locked, but support for > floating will be more expensive. > - Installation: One big improvement: you can select which filetypes will > be associated with Tornado! So no more resetting of the filetypes to point > to a real editor after a Tornado install. > - Link scripts: link scripts are used instead of the buildin link scripts > - GCC version 2.96 > - Codetest: not yet available for PPC, but an upgrade has to be purchased, > because of new features (around $500) > - Documentation includes a list of error codes and a migration guide > - Paper version of documentation has to be ordered seperately, but it is > still free > - GNU sources are not included in the standard distribution. > > I don't have any performance numbers yet, but I will try to start working > with this release as soon as possible. > > Kind regards, > Johan > > -- > o o o o o o o . . . _____________________________ > o _____ || Johan Borkhuis | > .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | > >(________|__|_[_________]_|__________________________| > _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` > === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === > --------------------------- Newsgroups: comp.os.vxworks Subject: Vxworks and Jungo? Date: 16 Sep 2002 08:26:14 -0700 From: dj_thorrin71@yahoo.com (Johnathan Ross) Organization: http://groups.google.com/ Message-ID: <40cb3a14.0209160726.297f8afa@posting.google.com> Has anyone successfully used this combination to send data across the PCI bus? We are on an x86 target with a TI C64 DSP on a PCI card. We are trying to send data to/from the host processor. Simple enough. The jungo tools work, and the data actually gets from place to place, but it is INCREDIBLY slow. Our messages can be 32K max, but Im seeing times in the 15ms range instead of the micro second range. Someone here mentioned that we may need to be in 'burst mode' for the PCI data. The TI chip looks like it defaults to this mode, but Im not sure about the pentium. Also, from looking at the Jungo page they say that if your PCI transfers are slow, look into using DMA to transfer data. The concept looks easy... Copy some data, give destination memory location, and go. What is the 'go' button? A register on the pentium? Does Vxworks handle this? So my questions are: 1) Why are my PCI transfers so slow? 2) How do I set/verify the pentium for burst mode? 3) Does Vxworks handle DMA or is this something I need to figure out on my own? Im sure there is a sample somewhere on how to do this on an x86. Where can I find this sample? 4) Do I even NEED jungo to handle bus transfers? Is PCI transfer easy enough to do without this tool? Thanks in advance! - -Johnathan Ross --------------------------- Newsgroups: comp.os.vxworks Subject: Limitation on number of UDP ports Date: Mon, 16 Sep 2002 11:26:13 -0500 From: Jankesh Piparia Organization: iDEN Systems Division Message-ID: <3D860625.4F4EC369@email.mot.com> Hi, I am trying to create 100 UDP listener(server) ports in vxWorks. The OS does not allow me to create more than 37 of them. On trying to create the 38th port, the program crashes. Can anybody tell me what configurable parameter I need to change in the vxWorks kernel to get this working. Also what is the system limitation on the number of ports. I know the theoretical limitation is 64k but how can I calculate the practical limitation. Thanks in advance, Jankesh --------------------------- Newsgroups: comp.os.vxworks Subject: Re: 1 second becomes 4 seconds! Date: Mon, 16 Sep 2002 18:58:01 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <9d9eae54.0209160613.3043746e@posting.google.com> You have a bug in the sysClk timer driver in your BSP, it incorrectly calculates the timer prescaler. Perhaps all you nede to to define the BSP bacro representing CPU clock to reflect actual hardware. "Welson Sun" wrote in message news:9d9eae54.0209160613.3043746e@posting.google.com... > I have got a problem with our 860 board: When the vxWorks is starting > up and counting down, the interval is not 1 second, actually it is 4 > seconds! > > I wrote a POSIX timer program and set the timer expiration duration to > be 1 second, but it actually expires after 4 seconds! > > > What could be the cause, I would like to know the root cause of this > problem. I have tried sysClkRateSet(sysClkRateGet()*4) but still the > case. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Just received: Tornado 2.2 Date: 16 Sep 2002 18:28:27 GMT From: Denis Perelyubskiy Message-ID: References: Hello, while we're using x86, still awaiting anxiously for comments :) Does anyone, by any chance, know if there are any bugfixes in t2.2 that are not available as standalone downloadable fixes or cumulative patches? going from 2.0 -> 2.2 is a little unpleasant due to NPT and recompilation of bootloaders and such (probably other things as well?) so trying to figure out what is it that it buys us... * Johan Borkhuis [09-Mon-02 00:37 -0700]: > > I just received my upgrade to Tornado 2.2 for PPC (Win2k). Some first > (incomplete) impressions. > - VxWorks version 5.5. > - License manager: Tornado now includes a license manager, with a fixed > number of floating or fixed licenses. If you have support you can choose > to convert your licenses to floating or node-locked, but support for > floating will be more expensive. > - Installation: One big improvement: you can select which filetypes will > be associated with Tornado! So no more resetting of the filetypes to point > to a real editor after a Tornado install. > - Link scripts: link scripts are used instead of the buildin link scripts > - GCC version 2.96 > - Codetest: not yet available for PPC, but an upgrade has to be purchased, > because of new features (around $500) > - Documentation includes a list of error codes and a migration guide > - Paper version of documentation has to be ordered seperately, but it is > still free > - GNU sources are not included in the standard distribution. > I don't have any performance numbers yet, but I will try to start working > with this release as soon as possible. > Kind regards, > Johan denis - -- if you want to reply by mail, please tweak my address in a reasonable way. sorry for the inconvenience. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Limitation on number of UDP ports Date: 16 Sep 2002 15:02:58 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0209161402.41123ffd@posting.google.com> References: <3D860625.4F4EC369@email.mot.com> Hello, This is a very frequently asked question - please learn to search the archives for the group *before* posting (you can search at groups.google.com as well as at a number of other places). To answer your question, it is not the UDP ports that is limiting you per-se. There are two things that you will need to change to have so many open listenerss though: 1) The number of file descriptors allowed (default is 50). Change this using NUM_FILES. If you plan to use select(), and you are using VxWorks 5.4 or earlier, then you are limited to a maximum of 256 for this value. 2) You will probably need to allocate more resources to the network stack itself. This is changed using the NUM_xxx and NUM_SYS_xxx (where xxx is replaced by a series of numbers reflecting the size of each block). There is information on WindSurf about this. This app note contains some information about this process: http://www.windriver.com/windsurf/appnotes/osExtensions/pjworks/APN51/WTN51.pdf HTH, John... Jankesh Piparia wrote in message news:<3D860625.4F4EC369@email.mot.com>... > Hi, > I am trying to create 100 UDP listener(server) ports in vxWorks. > The OS does not allow me to create more than 37 of them. On trying > to create the 38th port, the program crashes. > Can anybody tell me what configurable parameter I need to change in > the vxWorks kernel to get this working. Also what is the system > limitation > on the number of ports. I know the theoretical limitation is 64k but how > can I > calculate the practical limitation. > > Thanks in advance, > Jankesh --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks Applications Engineer Date: Mon, 16 Sep 2002 20:58:35 +0100 From: molochai Organization: Griffin Internet - http://www.griffin.net.uk Message-ID: References: <66a11529.0209090154.4a72b3c9@posting.google.com> Bamidelle wrote: Any chancve of providing a few details such as location please? M > Hi, I am looking for an experienced Engineer with either an > applications engineering or a software engineering background and > strong VxWorks experience. We are a leading edge developer of: VME * > CPCI * Multibus II * PMC * Real-time OS * Custom Solutions *. We are > looking for someone highly technical with good communication skills. > You need to be authorised to work in the UK, and want to take a > position here. We will pay up to £ 35 k and offer strong package and > career path. > > Is there anyone out there who can help ? Please reply with full CV and > contact details for informal + totally confidential discussion. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: intLock + printf Date: Tue, 17 Sep 2002 02:54:29 GMT From: "Gary M" Organization: AT&T Broadband Message-ID: References: The results of calling printf() in the intLock() context are unpredictable, meaning it may work, but it is not guaranteed to work. The documentation for intLock() states: "Do not call VxWorks system routines with interrupts locked. Violating this rule may re-enable interrupts unpredictably." By inference, calling printf will cause multiple system calls to be invoked. "John" wrote in message news:baa5411a.0209142209.43e9106a@posting.google.com... > Hi All, > What happens if I call intLock () and then try to do a printf? Will it > release the lock? Or does it raise an exception and what kind of > exception? > > Something like this... > > lockKey = intLock (); > . > . > do some processing; > . > . > intUnlock (lockKey); > > I was suspecting a part of my code which has this sequence and would > like to get some suggestions / comments on this. > > Your replies/ideas are highly appreciated. > > John. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: intLock + printf Date: 16 Sep 2002 21:38:55 -0700 From: yahoo_m1@hotmail.com (John) Organization: http://groups.google.com/ Message-ID: References: Hi Leo, I appreciate your replies. Your suggestions are helpful. John. "Leonid Rosenboim" wrote in message news:... > John, > It seems to me you are putting too much stuff in between intLock/intUnlock - > this is bad practice because it takes away the flexibility to control and > prioritize resources later. > Specifically if you put a printf(), it will prolong the time of lockout, as > printf() could take several milliseconds to execute (depending on CPU and > format string). Also, printf() could return immediately, or might need to > WAIT if the output buffer (e.g. of the serial console) is full. In order to > avoid losing data, the write() call (called by printf() ), will PEND until > there is sufficient space available in the output buffer. > When a task in interrupt lockout pends, interrupts are re-enabled, until > that task gets to run again. > > By the way, replacing printf() with logMsg() won't make a difference, > because logMsg() will too have to wait if the logTask msgQ is full (unless > it was called from an ISR), and will unlock interrupts when waiting. > What you need to do is to all your critical stuff before unlocking > interrupts, and move the printf() and other non-crticical calls to follow > intUnlock. > > Hope this helps, > - Leonid > "John" wrote in message > news:baa5411a.0209142209.43e9106a@posting.google.com... > > Hi All, > > What happens if I call intLock () and then try to do a printf? Will it > > release the lock? Or does it raise an exception and what kind of > > exception? > > > > Something like this... > > > > lockKey = intLock (); > > . > > . > > do some processing; > > . > > . > > intUnlock (lockKey); > > > > I was suspecting a part of my code which has this sequence and would > > like to get some suggestions / comments on this. > > > > Your replies/ideas are highly appreciated. > > > > John. --------------------------- Newsgroups: comp.os.vxworks Subject: anyone out there who can help me on windview and trigger tool? Date: 16 Sep 2002 21:44:45 -0700 From: vxnerd@rediffmail.com (subbarayan) Organization: http://groups.google.com/ Message-ID: dear mentors, 1)Can anyone tell me how to configure windview to view a graph of user defined tasks?My typical appln requires that the graph should show the name of tasks used in my programme.I have been unable to find my task names in graph where as i find generic tasknames as t14,t15,t16 etc and some other tasks like tnettask,idle task,texectask etc.Can anyone point me to info abt these various tasks displayed in graph of windview?I created a root task and spawned a task called taskone in my programme.I am unable to find taskone in my graph of windview.Is it appearing in some other name or what settings should i make to view info abt taskone in windview? the same code shown below was the one i tried.I expected taskone to be viewed in graph after calling root and executing it.I went through the manual but was unable to find the info i was looking for. 2)I was going through the triggering tool manual.After going through it ,i tried to use it to trigger windview for a sample application to start logging.I did the following: Opened trigger tool. Then defined context matches as tid1 which is my taskid of my task called taskone.defined actions as startwv. Then the following is my code: #include #include"vxWorks.h" #include"taskLib.h" void taskone(); void tasktwo(); int tid1,tid2; void root() { tid1=taskSpawn("t1",90,0x100,20000,FUNCPTR(taskone),0,0,0,0,0,0,0,0,0,0); tid2=taskSpawn("t2",91,0x100,20000,FUNCPTR(tasktwo),0,0,0,0,0,0,0,0,0,0); } void taskone() { printf("%d\n",taskIdSelf()); printf("i am first task in this pc \n"); printf("tasktwo created"); } void tasktwo() { printf("i am second task in this pc \n"); printf("releasing cpu to taskone \n"); } I am unable to find the trigger working inspite of me calling either root from shell or by calling taskone directly from shell.Is there some configuration error in it?I went through the manual but was unable to find the info i was looking for. I will be thankful to persons clarifying my doubt.I use tornado2.0 and i tried in simulator version. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: 1 second becomes 4 seconds! Date: 16 Sep 2002 21:52:01 -0700 From: welson_sun@yahoo.com (Welson Sun) Organization: http://groups.google.com/ Message-ID: <9d9eae54.0209162052.646ec3ec@posting.google.com> References: <9d9eae54.0209160613.3043746e@posting.google.com> Thanks. I am searching the BSP source code. The question should be: what is the clock source for the countdown interval and the POSIX timer? "Leonid Rosenboim" wrote in message news:... > You have a bug in the sysClk timer driver in your BSP, it incorrectly > calculates the timer prescaler. Perhaps all you nede to to define the > BSP bacro representing CPU clock to reflect actual hardware. > > "Welson Sun" wrote in message > news:9d9eae54.0209160613.3043746e@posting.google.com... > > I have got a problem with our 860 board: When the vxWorks is starting > > up and counting down, the interval is not 1 second, actually it is 4 > > seconds! > > > > I wrote a POSIX timer program and set the timer expiration duration to > > be 1 second, but it actually expires after 4 seconds! > > > > > > What could be the cause, I would like to know the root cause of this > > problem. I have tried sysClkRateSet(sysClkRateGet()*4) but still the > > case. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: anyone out there who can help me on windview and trigger tool? Date: Tue, 17 Sep 2002 07:51:49 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: Reply-To: "Michael Lawnick" Sender: mlawnick@pd95240a5.dip.t-dialin.net Hi subbarayan, I can't believe your current question! This is one more RTFM and look at your own code. Do you ever think before posting? Your code calls taskSpawn("t1",90,0x100,20000,FUNCPTR(taskone),0,0,0,0,0,0,0,0,0,0); So you create a task with name "t1" who's entry point is taskone. Thats what you see in WindView. Try 'i' on shell and you'll get just this name. Michael "subbarayan" schrieb im Newsbeitrag news:f81c4277.0209162044.16c389c0@posting.google.com... > dear mentors, > 1)Can anyone tell me how to configure windview to view a graph of user > defined tasks?My typical appln requires that the graph should show the > name of tasks used in my programme.I have been unable to find my task > names in graph where as i find generic tasknames as t14,t15,t16 etc > and some other tasks like tnettask,idle task,texectask etc.Can anyone > point me to info abt these various tasks displayed in graph of > windview?I created a root task and spawned a task called taskone in my > programme.I am unable to find taskone in my graph of windview.Is it > appearing in some other name or what settings should i make to view > info abt taskone in windview? the same code shown below was the one i > tried.I expected taskone to be viewed in graph after calling root and > executing it.I went through the manual but was unable to find the info > i was looking for. > > 2)I was going through the triggering tool manual.After going through > it ,i tried to use it to trigger windview for a sample application to > start logging.I did the following: > Opened trigger tool. > Then defined context matches as tid1 which is my taskid of my task > called taskone.defined actions as startwv. > Then the following is my code: > #include > #include"vxWorks.h" > #include"taskLib.h" > void taskone(); > void tasktwo(); > int tid1,tid2; > void root() > { > tid1=taskSpawn("t1",90,0x100,20000,FUNCPTR(taskone),0,0,0,0,0,0,0,0,0,0); > tid2=taskSpawn("t2",91,0x100,20000,FUNCPTR(tasktwo),0,0,0,0,0,0,0,0,0,0); > } > void taskone() > { > printf("%d\n",taskIdSelf()); > printf("i am first task in this pc \n"); > printf("tasktwo created"); > } > void tasktwo() > { > printf("i am second task in this pc \n"); > printf("releasing cpu to taskone \n"); > } > I am unable to find the trigger working inspite of me calling either > root from shell or by calling taskone directly from shell.Is there > some configuration error in it?I went through the manual but was > unable to find the info i was looking for. > > I will be thankful to persons clarifying my doubt.I use tornado2.0 and > i tried in simulator version. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: 1 second becomes 4 seconds! Date: 16 Sep 2002 23:11:59 -0700 From: welson_sun@yahoo.com (Welson Sun) Organization: http://groups.google.com/ Message-ID: <9d9eae54.0209162211.78346c5@posting.google.com> References: <9d9eae54.0209160613.3043746e@posting.google.com> I have just discovered that the outcome of taskDelay(sysClkRateGet()) is to make the task delay 4 seconds! "Leonid Rosenboim" wrote in message news:... > You have a bug in the sysClk timer driver in your BSP, it incorrectly > calculates the timer prescaler. Perhaps all you nede to to define the > BSP bacro representing CPU clock to reflect actual hardware. > > "Welson Sun" wrote in message > news:9d9eae54.0209160613.3043746e@posting.google.com... > > I have got a problem with our 860 board: When the vxWorks is starting > > up and counting down, the interval is not 1 second, actually it is 4 > > seconds! > > > > I wrote a POSIX timer program and set the timer expiration duration to > > be 1 second, but it actually expires after 4 seconds! > > > > > > What could be the cause, I would like to know the root cause of this > > problem. I have tried sysClkRateSet(sysClkRateGet()*4) but still the > > case. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: 1 second becomes 4 seconds! Date: Tue, 17 Sep 2002 09:25:30 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <9d9eae54.0209160613.3043746e@posting.google.com> <9d9eae54.0209162052.646ec3ec@posting.google.com> I thought I made it clear the first time around - the timing source for everything, including autonoot countdown and POSIX timers is the system clock, which is driven by the BSP's sysClk timer driver (usually included from sysLib.c) The timer used, and its source clock is by large a hardware issue, and I clealry remember that ppc86- BSP has a macro which defines the hardware clock that drives the CPU, which is the source for the timer. "Welson Sun" wrote in message news:9d9eae54.0209162052.646ec3ec@posting.google.com... > Thanks. I am searching the BSP source code. The question should be: > what is the clock source for the countdown interval and the POSIX > timer? > > "Leonid Rosenboim" wrote in message news:... > > You have a bug in the sysClk timer driver in your BSP, it incorrectly > > calculates the timer prescaler. Perhaps all you nede to to define the > > BSP bacro representing CPU clock to reflect actual hardware. > > > > "Welson Sun" wrote in message > > news:9d9eae54.0209160613.3043746e@posting.google.com... > > > I have got a problem with our 860 board: When the vxWorks is starting > > > up and counting down, the interval is not 1 second, actually it is 4 > > > seconds! > > > > > > I wrote a POSIX timer program and set the timer expiration duration to > > > be 1 second, but it actually expires after 4 seconds! > > > > > > > > > What could be the cause, I would like to know the root cause of this > > > problem. I have tried sysClkRateSet(sysClkRateGet()*4) but still the > > > case. --------------------------- Newsgroups: comp.os.vxworks Subject: SBS VG4 + bootable + flash = undefined references Date: 17 Sep 2002 00:28:16 -0700 From: adrian.baumgarth@elac-nautik.com (Adrian Baumgarth) Organization: http://groups.google.com/ Message-ID: Hi, I am converting a downloadable project to a bootable one. The application uses the flash built in the PowerPC board VG4 by SBS. Everything works fine in the downloadable projekt, but in the bootable there are some linker errors, that i just don't get rid off. The text below is the complete build output for a small example project which I made for solving the problem. As you can see everything compiles fine, but the linker is very unhappy. If anybody has a suggestion don't hesitate to answer. Adrian P.S.: I edited the build output a bit to shorten it. vxrm *.o *.rpo ctdt.c symTbl.c vxWorks* vxrm D:\Programme\WindRiver\Tornado\target\proj\include_test\prjComps.h D:\Programme\WindR iver\Tornado\target\proj\include_test\prjParams.h D:\Programme\WindRiver\Tornado\target\pr oj\include_test\prjConfig.c D:\Programme\WindRiver\Tornado\target\proj\include_test\linkSy ms.c vxrm D:\Programme\WindRiver\Tornado\target\proj\include_test\libs.nm D:\Programme\WindRive r\Tornado\target\proj\include_test\libs.size wtxtcl D:\Programme\WindRiver\Tornado\host\src\hutils\configGen.tcl D:\Programme\WindRiver \Tornado\target\proj\include_test\Include_Test.wpj ccppc -g -mstrict-align -ansi -nostdinc -fvolatile -fno-builtin - -fno-for-scope -P -x assem bler-with-cpp -ID:\Programme\WindRiver\Tornado\target\proj\include_test - -ID:\Programme\Win dRiver\Tornado\target\config\sbs_vg4 - -ID:\Programme\WindRiver\Tornado\target\h -ID:\Progra mme\WindRiver\Tornado\target\config\comps\src - -ID:\Programme\WindRiver\Tornado\target\src\ drv -DCPU=PPC604 -DPRJ_BUILD -I. - -ID:\Programme\WindRiver\Tornado\target\config\sbs_vg4\in clude -ID:\Programme\WindRiver\Tornado\target\h\tffs -DVG4 - -DTARGET_DIR="\"sbs_vg4\"" -g - mlongcall -c D:\Programme\WindRiver\Tornado\target\config\sbs_vg4\sysALib.s - -o sysALib.o ccppc [...] sysLib.c ccppc [...] usrAppInit.c ccppc [...] prjConfig.c ccppc [...] linkSyms.c ccppc [...] mainprog.cpp ccppc [...] version.c ccppc -BD:\Programme\WindRiver\Tornado\host\x86-win32\lib\gcc-lib\ - -nostdlib -r -Wl,-X \ -Wl,@D:\Programme\WindRiver\Tornado\target\proj\include_test\prjObjs.lst version.o D :\Programme\WindRiver\Tornado\target\config\sbs_vg4\Vg4Lib.a D:\Programme\WindRiver\Tornad o\target\lib\libPPC604gnuvx.a -o partialImage.o nmppc partialImage.o @D:\Programme\WindRiver\Tornado\target\proj\include_test\prjObjs.lst | wtxtcl D:\Programme\WindRiver\Tornado\host\src\hutils\munch.tcl -asm ppc > ctdt.c ccppc -c -fdollars-in-identifiers -g -mstrict-align -ansi - -DRW_MULTI_THREAD -D_REENTRANT - fvolatile -fno-builtin -fno-for-scope - -ID:\Programme\WindRiver\Tornado\target\proj\include _test -ID:\Programme\WindRiver\Tornado\target\config\sbs_vg4 - -ID:\Programme\WindRiver\Torn ado\target\h -ID:\Programme\WindRiver\Tornado\target\config\comps\src - -ID:\Programme\WindR iver\Tornado\target\src\drv -DCPU=PPC604 -DPRJ_BUILD -I. - -ID:\Programme\WindRiver\Tornado\ target\config\sbs_vg4\include - -ID:\Programme\WindRiver\Tornado\target\h\tffs -DVG4 -DTARGE T_DIR="\"sbs_vg4\"" -g -mlongcall ctdt.c -o ctdt.o copy partialImage.o tmp.o 1 file(s) copied. D:\Programme\WindRiver\Tornado\host\x86-win32\bin\makeSymTbl PPC tmp.o > symTbl.c ccppc -c -fdollars-in-identifiers -g -mstrict-align -ansi - -DRW_MULTI_THREAD -D_REENTRANT - fvolatile -fno-builtin -fno-for-scope - -ID:\Programme\WindRiver\Tornado\target\proj\include _test -ID:\Programme\WindRiver\Tornado\target\config\sbs_vg4 - -ID:\Programme\WindRiver\Torn ado\target\h -ID:\Programme\WindRiver\Tornado\target\config\comps\src - -ID:\Programme\WindR iver\Tornado\target\src\drv -DCPU=PPC604 -DPRJ_BUILD -I. - -ID:\Programme\WindRiver\Tornado\ target\config\sbs_vg4\include - -ID:\Programme\WindRiver\Tornado\target\h\tffs -DVG4 -DTARGE T_DIR="\"sbs_vg4\"" -g -mlongcall symTbl.c -o symTbl.o ccppc -g -mstrict-align -ansi -DRW_MULTI_THREAD -D_REENTRANT - -fvolatile -fno-builtin -fno- for-scope -ID:\Programme\WindRiver\Tornado\target\proj\include_test - -ID:\Programme\WindRiv er\Tornado\target\config\sbs_vg4 - -ID:\Programme\WindRiver\Tornado\target\h -ID:\Programme\ WindRiver\Tornado\target\config\comps\src - -ID:\Programme\WindRiver\Tornado\target\src\drv - -DCPU=PPC604 -DPRJ_BUILD -I. - -ID:\Programme\WindRiver\Tornado\target\config\sbs_vg4\includ e -ID:\Programme\WindRiver\Tornado\target\h\tffs -DVG4 - -DTARGET_DIR="\"sbs_vg4\"" -g -mlon gcall -c D:\Programme\WindRiver\Tornado\target\config\comps\src\dataSegPad.c ldppc -X -N -e _sysInit -Ttext 00100000 \ dataSegPad.o partialImage.o ctdt.o symTbl.o -o vxWorks partialImage.o: In function `runItAll': D:\Programme\WindRiver\Tornado\target\proj\include_test\default\D:\Programme\WindRiver\Tor nado\target\proj\include_test\mainprog.cpp:10: undefined reference to `intelFlashGetManufC ode(unsigned char)' D:\Programme\WindRiver\Tornado\target\proj\include_test\default\D:\Programme\WindRiver\Tor nado\target\proj\include_test\mainprog.cpp:10: undefined reference to `intelFlashGetManufC ode(unsigned char)' make: *** [vxWorks] Error 0x1 Done. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Vxworks and Jungo? Date: 17 Sep 2002 00:37:17 -0700 From: egoisone@libero.it (Mic) Organization: http://groups.google.com/ Message-ID: References: <40cb3a14.0209160726.297f8afa@posting.google.com> Hi Ross, probably I haven't all the answers you need but I found myself in the same situation months ago. At that time I tried Jungo but, probably due to my PCI card, nothing went well. Anyway, I saw that Jungo code was built to cover a wide set of devices so it cannot be optimized for everyone. My solution was writing by myself a driver and I think you can do the same once you have installed the card and got addresses of its registers. Bye. Mic. dj_thorrin71@yahoo.com (Johnathan Ross) wrote in message news:<40cb3a14.0209160726.297f8afa@posting.google.com>... > Has anyone successfully used this combination to send data across the > PCI bus? > > We are on an x86 target with a TI C64 DSP on a PCI card. We are > trying to send data to/from the host processor. Simple enough. The > jungo tools work, and the data actually gets from place to place, but > it is INCREDIBLY slow. Our messages can be 32K max, but Im seeing > times in the 15ms range instead of the micro second range. > > Someone here mentioned that we may need to be in 'burst mode' for the > PCI data. The TI chip looks like it defaults to this mode, but Im not > sure about the pentium. > > Also, from looking at the Jungo page they say that if your PCI > transfers are slow, look into using DMA to transfer data. The concept > looks easy... Copy some data, give destination memory location, and > go. What is the 'go' button? A register on the pentium? Does > Vxworks handle this? > > So my questions are: > > 1) Why are my PCI transfers so slow? > 2) How do I set/verify the pentium for burst mode? > 3) Does Vxworks handle DMA or is this something I need to figure out > on my own? Im sure there is a sample somewhere on how to do this on an > x86. Where can I find this sample? > 4) Do I even NEED jungo to handle bus transfers? Is PCI transfer > easy enough to do without this tool? > > Thanks in advance! > -Johnathan Ross --------------------------- Newsgroups: comp.os.vxworks Subject: Is it necessary to program the bootloader on different chip than the application or TFFS? Date: 17 Sep 2002 02:18:54 -0700 From: dannyc@harmonicinc.com (Danny Cohen) Organization: http://groups.google.com/ Message-ID: <41c1fd2b.0209170118.76648c6c@posting.google.com> From you're experience: Do you see any SW or HW or System drawbacks in the solution of programming both boot loader and application and File System on the same Flash Chip? --------------------------- Newsgroups: comp.os.vxworks Subject: Target Sever and Target Agent? Date: 17 Sep 2002 02:25:42 -0700 From: oe_zhou@sohu.com (oe_zhou) Organization: http://groups.google.com/ Message-ID: <758a30ff.0209170125.16150b2@posting.google.com> I use FEC of the 860T to boot, and it works well. But when the system image has been downloaded , the 100BASE-T network does'n work. And the link to target agent is losed.But the serial port and the system both works well. (860T + VxWorks). Help me!!! --------------------------- Newsgroups: comp.os.vxworks Subject: about bsp makefile under diab compiler Date: 17 Sep 2002 03:27:14 -0700 From: lijun1@datangmobile.com (lijun) Organization: http://groups.google.com/ Message-ID: <50cc31ed.0209170227.ac4d35c@posting.google.com> HI: i should complie the bsp,but i have not tornado IDE in unix enviroment.so i have to write makefile to realize my purpose.so who can tell me how to write the makefile? --------------------------- Newsgroups: comp.os.vxworks Subject: RTL8029 Date: 17 Sep 2002 03:40:44 -0700 From: egoisone@libero.it (Mic) Organization: http://groups.google.com/ Message-ID: Hi VxWorkers, I read many interesting things about RTL8029 and VxWorks but I cannot find a simple answer: does NE2000 driver works "as is" with RTL8029(AS) or not (does not matter if communication goes slowly...)? I mean, is enough to put "#define INCLUDE_ENE" and correct I/O address and IRQ line in config.h? Thanks, Mic. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: 1 second becomes 4 seconds! Date: 17 Sep 2002 03:45:40 -0700 From: welson_sun@yahoo.com (Welson Sun) Organization: http://groups.google.com/ Message-ID: <9d9eae54.0209170245.9596556@posting.google.com> References: <9d9eae54.0209160613.3043746e@posting.google.com> <9d9eae54.0209162211.78346c5@posting.google.com> welson_sun@yahoo.com (Welson Sun) wrote in message news:<9d9eae54.0209162211.78346c5@posting.google.com>... > I have just discovered that the outcome of taskDelay(sysClkRateGet()) > is to make the task delay 4 seconds! > > "Leonid Rosenboim" wrote in message news:... > > You have a bug in the sysClk timer driver in your BSP, it incorrectly > > calculates the timer prescaler. Perhaps all you nede to to define the > > BSP bacro representing CPU clock to reflect actual hardware. > > > > "Welson Sun" wrote in message > > news:9d9eae54.0209160613.3043746e@posting.google.com... > > > I have got a problem with our 860 board: When the vxWorks is starting > > > up and counting down, the interval is not 1 second, actually it is 4 > > > seconds! > > > > > > I wrote a POSIX timer program and set the timer expiration duration to > > > be 1 second, but it actually expires after 4 seconds! > > > > > > > > > What could be the cause, I would like to know the root cause of this > > > problem. I have tried sysClkRateSet(sysClkRateGet()*4) but still the > > > case. Latest finding: If I set the TBS bit of SCCR register, the taskDelay(sysClkRateGet()) will delay for 2 seconds! And if I clear the TBS bit, it will just delay 4 seconds. Does this mean that this problem is related with Timebase clock? And I don't understand the following definations in the BSP: #define TMBCLK_FREQ CRISTAL_FREQ #define DEC_CLOCK_FREQ TMBCLK_FREQ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: 1 second becomes 4 seconds! Date: Tue, 17 Sep 2002 13:01:46 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <9d9eae54.0209160613.3043746e@posting.google.com> <9d9eae54.0209162211.78346c5@posting.google.com> <9d9eae54.0209170245.9596556@posting.google.com> "Welson Sun" wrote in message news:9d9eae54.0209170245.9596556@posting.google.com... [snip] > Latest finding: > > If I set the TBS bit of SCCR register, the taskDelay(sysClkRateGet()) > will delay for 2 seconds! And if I clear the TBS bit, it will just > delay 4 seconds. > Does this mean that this problem is related with Timebase clock? Yes, the timebase register is probably involved, but since I dont have this particular hardware manual handy, you will need to figure it out yourself. > > And I don't understand the following definations in the BSP: > #define TMBCLK_FREQ CRISTAL_FREQ > #define DEC_CLOCK_FREQ TMBCLK_FREQ CRYSTAL_FREQ is the frequency of the clock fed to the CPU on your board, find and make sure it's definition reflects your actual crystal! DEC_CLOCK is the frequency of the "decrementer" which I beleive is used for the WindView timestamp. You should have the hardware manual, and you have source code of the sysClk driver, I do not see what should keep you from resolving this. - - Leonid --------------------------- Newsgroups: comp.os.vxworks Subject: Re: anyone out there who can help me on windview and trigger tool? Date: 17 Sep 2002 05:26:17 -0700 From: vxnerd@rediffmail.com (subbarayan) Organization: http://groups.google.com/ Message-ID: References: Dear Mr.Lawnik, Sorry I made a mistake when i posted my code.actually i gave task name as "tas1" and not t1 as mentioned in my code.But still I could not find a tas1 in my windview.I am finding only generic names like t4,t5,t6 etc.Where is the mistake? Thanks for ur reply. --------------------------- Newsgroups: comp.os.vxworks Subject: Newbie needs help Date: Tue, 17 Sep 2002 13:18:36 GMT From: "Yan Meunier" Organization: Aliant Internet Message-ID: Hi everyone. I am new to VxWorks, and real time programming. I am trying to find some links or any articles which could help me out with the programing of timers and signals. I've already taken a look at the man-pages, but they were not much help. Any suggestions would be greatfull. Thank you, Yan Meunier --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Is it necessary to program the bootloader on different chip than the application or TFFS? Date: Tue, 17 Sep 2002 15:17:31 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <41c1fd2b.0209170118.76648c6c@posting.google.com> Reply-To: "Michael Lawnick" Sender: mlawnick@pd95240a5.dip.t-dialin.net Hi Danny, as always: that depends ... ;-) Generally no problem. In this case you may not use a vxWorks_res of course or have to move your flash update routines to RAM. The only remarkable drawback : If an update of your flash fails it might damage your bootimage. So you either need to be able to lock the boot sectors or should have JTAG or back-upflash for worst case. Otherwise you'll have a piece of H/W you can't use anymore. Regards, Michael "Danny Cohen" schrieb im Newsbeitrag news:41c1fd2b.0209170118.76648c6c@posting.google.com... > From you're experience: Do you see any SW or HW or System drawbacks in > the solution of programming both boot loader and application and File > System on the same Flash Chip? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: 1 second becomes 4 seconds! Date: 17 Sep 2002 06:39:23 -0700 From: welson_sun@yahoo.com (Welson Sun) Organization: http://groups.google.com/ Message-ID: <9d9eae54.0209170539.28250303@posting.google.com> References: <9d9eae54.0209160613.3043746e@posting.google.com> <9d9eae54.0209162211.78346c5@posting.google.com> Problem update: 1. taskDelay( sysClkRateGet() ) delays 4 clock instead of 1 clock. sysClkRateGet() = 60 2. The BRG clock on my 860 board is working fine because my SMC1 (debug port) is at the exact baud rate as I have set 3. My board is using a 25Mhz oscilliator and the CPU clock is 50Mhz, I have checked the MF field in PLCR, it is correct ( = 1 ). 4. If I set the TBS bit in SCCR register, taskDelay( sysClkRateGet() ) will delay 2 seconds and if I clear this bit, taskDelay( sysClkRateGet() ) will just delay 4 seconds! I think the problem should be related with the timebase clock rate or the decrement clock rate. I found the two defines in the BSP: #define TMBCLK_FREQ CRISTAL_FREQ /* = 25Mhz */ #define DEC_CLOCK_FREQ TMBCLK_FREQ Because this BSP is derived from ADS, I don't know what is the use of these two defines. Is vxWorks using these two defines? If yes, where? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Intel chipset 815E Date: Tue, 17 Sep 2002 15:40:50 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Yes it works. And you only need to modify the PCI device ID constant for the standard FEI driver to work on it. It also has a good Bus Master IDE controller, but you need to modfy the ataDrv driver to make use of it and get excellend disk performance. wrote in message news:am79uj$kn7$1@overload.lbl.gov... > Hello VXWORKS and TORNADO Gurus, > > does anyone already works with the intel chipset 815E ? > I am looking for a driver for the ethernet component integrated inside ... > > Thanks in advance > > > > François PERTIN > > > STÄUBLI SCA > Robotics Division > Place Robert Stäubli - BP 70 > F - 74210 Faverges (France) > Tel.: +33 (0)4.50.65.62.59 > Fax: +33 (0)4.50.65.61.30 > mailto:f.pertin@staubli.com > http://www.staubli.com > --------------------------- Newsgroups: comp.os.vxworks Subject: How to generate rom images Date: Tue, 17 Sep 2002 13:59:42 GMT From: Hans Berglund Message-ID: <3D873554.3090909@spacetec.no> Hi I have a problem with generating ROM images for an mvve2400 board. I have tried vxWorks_rom and vxWorks_romResident images. I know there are lots of people out there who manage to get these things working. If you have no spare time to read on, please just post the "objdumpppc -h" and "elfToBin" from a working image. Both rom and romResident are of interest. First case first: I "niop" the image to address 4000, and flash it with pflash 4000:100000 ff000100. The latter becomes ff8000100 and fff00100 after the flash select jumper is moved. After a reset (and jumper is moved) it is completely silent. The objdumpppc -h says: vxWorks_rom: file format elf32-powerpc Sections: Idx Name Size VMA LMA File off Algn 0 .text 00001a3c fff00100 fff00100 00110100 2**2 CONTENTS, ALLOC, LOAD, READONLY, CODE 1 .data 000f6a68 00800000 00800000 00010000 2**2 CONTENTS, ALLOC, LOAD, DATA 2 .rodata 0000001c fff01b3c fff01b3c 00111b3c 2**2 CONTENTS, ALLOC, LOAD, READONLY, DATA 3 .sdata2 00000000 fff01b58 fff01b58 00111b58 2**2 CONTENTS, ALLOC, LOAD, READONLY, DATA 4 .got 00000010 008f6a68 008f6a68 00106a68 2**2 CONTENTS, ALLOC, LOAD, DATA 5 .sdata 00000000 008f6a78 008f6a78 00106a78 2**2 CONTENTS, ALLOC, LOAD, DATA 6 .sbss 00000000 00000000 00000000 00111b58 2**0 CONTENTS 7 .bss 00000000 008f6a78 008f6a78 00106a78 2**0 ALLOC 8 .comment 00000100 00000000 00000000 00111b58 2**0 CONTENTS, READONLY The elfToBin says: Entry address is fff00100 seg0: Flags: 00000005 Offset: 00110100 VADDR: fff00100 Size: 00001a58 seg1: Flags: 00000006 Offset: 00010000 VADDR: 00800000 Size: 000f6a78 I thought it would be easier to make an rom resident image, and so I did. objdumpppc -h says: vxWorks_romResident: file format elf32-powerpc Sections: Idx Name Size VMA LMA File off Algn 0 .text 000d8d54 fff00100 fff00100 00020100 2**3 CONTENTS, ALLOC, LOAD, READONLY, CODE 1 .data 0000b890 00100000 00100000 00010000 2**4 CONTENTS, ALLOC, LOAD, DATA 2 .rodata 0000ce18 fffd8e58 fffd8e58 000f8e58 2**3 CONTENTS, ALLOC, LOAD, READONLY, DATA 3 .sdata2 00000000 fffe5c70 fffe5c70 00105c70 2**2 CONTENTS, ALLOC, LOAD, READONLY, DATA 4 .ctors 0000000c 0010b890 0010b890 0001b890 2**0 CONTENTS, ALLOC, LOAD, DATA 5 .dtors 0000000c 0010b89c 0010b89c 0001b89c 2**0 CONTENTS, ALLOC, LOAD, DATA 6 .got 00000010 0010b8a8 0010b8a8 0001b8a8 2**2 CONTENTS, ALLOC, LOAD, DATA 7 .sdata 00000000 0010b8b8 0010b8b8 0001b8b8 2**2 CONTENTS, ALLOC, LOAD, DATA 8 .sbss 00000334 0010b8b8 0010b8b8 0001b8b8 2**2 ALLOC 9 .bss 0000581c 0010bbf0 0010bbf0 0001b8b8 2**3 ALLOC 10 .comment 00004008 00000000 00000000 00105c70 2**0 CONTENTS, READONLY The elfToBin refuses to convert it. I just says: Entry address is fff006e8 seg0 is NULL! exiting All help is appreciated. --------------------------- Newsgroups: comp.os.vxworks Subject: Printf Vs. FPrintf Date: 17 Sep 2002 08:07:53 -0700 From: nirn@rit.co.il (Nir N) Organization: http://groups.google.com/ Message-ID: <4913ea0b.0209170707.7f670f40@posting.google.com> Hi, Printf in VxWorks is Not buffered IO command and FPrintf is Buffered (they are the same with stdout). Does anyone know if the fprintf is pending like the printf ? Best Regards Nir --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Printf Vs. FPrintf Date: Tue, 17 Sep 2002 17:08:18 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <4913ea0b.0209170707.7f670f40@posting.google.com> Both would pend when they call the write() system call, if the output device buffer is full - the write(0 will PEND at the device driver. With a buffered output with fprintf() calls to write will be less frequent, but each write request will have more data, and more probably to pend. "Nir N" wrote in message news:4913ea0b.0209170707.7f670f40@posting.google.com... > Hi, > > Printf in VxWorks is Not buffered IO command and FPrintf is Buffered > (they are the same with stdout). > > Does anyone know if the fprintf is pending like the printf ? > > > > Best Regards > Nir --------------------------- Newsgroups: comp.os.vxworks Subject: Re: network task on vxworks Date: 17 Sep 2002 08:36:43 -0700 From: mchandnani@empirix.com (Monica Chandnani) Organization: http://groups.google.com/ Message-ID: <6fe3f9e6.0209170736.79fb118a@posting.google.com> References: I am seeing the same problem. I don't know how to solve it, but, I did get utilities.c and compiled it and downloaded it. I ran endPoolShow command from it. The number of free clusters was 0. But, I don't know how to increase the number of clusters. Do you know how to do that? Shlomo Katz wrote in message news:... > hi all > i have a strange problem, i am using vxworks on a ppc860 . > after a while that the system is running the unit does not respond to ping > request, the network task seems to be running . > does somebody know of this problem > > any tips are welcome > > thanks > > katz shlomi > www.rit.co.il > > ************************************************************************************************** > Privacy Notice: The information contained in this transmittal, including any > attachments hereto, are confidential and privileged, and intended solely for the > specified addressee(s). If you are not the intended addressee, or if you receive this > message by error, please notify the sender and delete this information from your > computer. > > Moreover, the recipient(s) may not disclose, forward, or copy this e-mail or > attachments, or any portion thereof, or permit the use of this information, by anyone > not entitled to it, or in a way that may be damaging to the sender. > > ************************************************************************************************** --------------------------- Newsgroups: comp.os.vxworks Subject: endPoolShow shows 0 free clusters Date: 17 Sep 2002 08:40:56 -0700 From: mchandnani@empirix.com (Monica Chandnani) Organization: http://groups.google.com/ Message-ID: <6fe3f9e6.0209170740.7cf60648@posting.google.com> Summary: The number of free clusters shown by endPoolShow go down to 0; and all network connections to the PowerPC die, i.e ping, ftp, telnet all fail. Description: I open 2 sockets to the PowerPC and send lots of requests at a very high speed. Eventually, the PowerPC stops responding to ftp, telnet, ping etc. While, when I see things through the serial connection, I see all tasks (including the network task) look normal and checkStack also shows no errors. However, when I run the endPoolShow command, I see the number of free clusters to 0. Is there a way of increasing the number of end driver clusters? Thanks --------------------------- Newsgroups: comp.os.vxworks Subject: Re: endPoolShow shows 0 free clusters Date: Tue, 17 Sep 2002 18:06:36 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <6fe3f9e6.0209170740.7cf60648@posting.google.com> Open your Network Programmers Guide (it's available for download from www.wrs.com if you lost your paper copy, and open it in the section: 4.6.3 Network Memory Pool Configuration It's all written up there. RTFM ! "Monica Chandnani" wrote in message news:6fe3f9e6.0209170740.7cf60648@posting.google.com... > Summary: > > The number of free clusters shown by endPoolShow go down to 0; and all > network connections to the PowerPC die, i.e ping, ftp, telnet all > fail. > > Description: > I open 2 sockets to the PowerPC and send lots of requests at a very > high speed. Eventually, the PowerPC stops responding to ftp, telnet, > ping etc. While, when I see things through the serial connection, I > see all tasks (including the network task) look normal and checkStack > also shows no errors. However, when I run the endPoolShow command, I > see the number of free clusters to 0. Is there a way of increasing the > number of end driver clusters? > > Thanks --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Limitation on number of UDP ports Date: Tue, 17 Sep 2002 10:59:53 -0500 From: Jankesh Piparia Organization: iDEN Systems Division Message-ID: <3D875179.3437E994@email.mot.com> References: <3D860625.4F4EC369@email.mot.com> <488e459a.0209161402.41123ffd@posting.google.com> John, Thanks for your response. I had figured out step 1 on my own by looking at the error code of the socket() call. I was in the process of figuring out step 2 also. Meanwhile, you resolved my problem by giving me the windriver link. Thanks again, Jankesh John wrote: > Hello, > > This is a very frequently asked question - please learn to search the > archives for the group *before* posting (you can search at > groups.google.com as well as at a number of other places). > > To answer your question, it is not the UDP ports that is limiting you > per-se. There are two things that you will need to change to have so > many open listenerss though: > > 1) The number of file descriptors allowed (default is 50). Change this > using NUM_FILES. If you plan to use select(), and you are using > VxWorks 5.4 or earlier, then you are limited to a maximum of 256 for > this value. > > 2) You will probably need to allocate more resources to the network > stack itself. This is changed using the NUM_xxx and NUM_SYS_xxx (where > xxx is replaced by a series of numbers reflecting the size of each > block). There is information on WindSurf about this. This app note > contains some information about this process: > > http://www.windriver.com/windsurf/appnotes/osExtensions/pjworks/APN51/WTN51.pdf > > HTH, > > John... > > Jankesh Piparia wrote in message news:<3D860625.4F4EC369@email.mot.com>... > > Hi, > > I am trying to create 100 UDP listener(server) ports in vxWorks. > > The OS does not allow me to create more than 37 of them. On trying > > to create the 38th port, the program crashes. > > Can anybody tell me what configurable parameter I need to change in > > the vxWorks kernel to get this working. Also what is the system > > limitation > > on the number of ports. I know the theoretical limitation is 64k but how > > can I > > calculate the practical limitation. > > > > Thanks in advance, > > Jankesh --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Target Sever and Target Agent? Date: Tue, 17 Sep 2002 11:05:35 -0500 From: Jankesh Piparia Organization: iDEN Systems Division Message-ID: <3D8752CF.C95834AC@email.mot.com> References: <758a30ff.0209170125.16150b2@posting.google.com> Hi, Did you try the routeAdd("0", ""). You will need to add this line of code in your usrAppInit.c before building the vxWorks image. - -Jankesh oe_zhou wrote: > I use FEC of the 860T to boot, and it works well. > But when the system image has been downloaded , > the 100BASE-T network does'n work. And the link > to target agent is losed.But the serial port > and the system both works well. > (860T + VxWorks). > Help me!!! --------------------------- Newsgroups: comp.os.vxworks Subject: Re: 1 second becomes 4 seconds! Date: Tue, 17 Sep 2002 09:22:49 -0700 From: "Dan Gold" Organization: Mailgate.ORG Server - http://www.Mailgate.ORG Message-ID: References: <9d9eae54.0209160613.3043746e@posting.google.com> <9d9eae54.0209162211.78346c5@posting.google.com> <9d9eae54.0209170539.28250303@posting.google.com> Sender: gold@ensemble.com "Welson Sun" wrote in message news:9d9eae54.0209170539.28250303@posting.google.com... > Problem update: > > 3. My board is using a 25Mhz oscilliator and the CPU clock is 50Mhz, > I have checked the MF field in PLCR, it is correct ( = 1 ). > > I think the problem should be related with the timebase clock rate or > the decrement clock rate. I found the two defines in the BSP: > > #define TMBCLK_FREQ CRISTAL_FREQ /* = 25Mhz */ > #define DEC_CLOCK_FREQ TMBCLK_FREQ By default, all PowerPC BSPs use the decrementer as the system clock. Look (as Leonid said) at the sysClk driver. You should find it in tornado/target/src/drv/timer/ppcDecTimer.c. It should use the DEC_CLOCK_FREQ define in sysClkInit() and sysClkRateSet(). You need to sit down with the 860 manual, in particular the chapter on dealing with the clocks, and run the numbers from your .h file to convince yourself you know how everything is working, i.e. the MF bit, TBS, etc... The decrementer counts at 1/4 the bus clock speed. So if your bus clock is 50Mhz, your decrementer should be counting at 12.5Mhz. That divide by 4 is either handled in the bsp specific .h file in the target directory, or as part of the equation in ppcDecTimer.c - I've seen it done in both places, depending on the BSP. Since the original problem was delaying for 1 second really delays for 4 seconds, I would leave the TBS bit so it's 4 seconds, then look for the missing divide by 4 in one of the defines for the decrementer, or in ppcDecTimer.c hth, Dan Gold gold@ensemble.com --------------------------- Newsgroups: comp.os.vxworks Subject: Loading object module from Disk on Chip Date: 17 Sep 2002 11:33:25 -0700 From: gibbsw@battelle.org (Bill Gibbs) Organization: http://groups.google.com/ Message-ID: We have a system we are trying to implement which consist of 2MB ROM, 1MB RAM and an 8MB disk-on-chip. We are running into difficulties with the amount of RAM and I am trying to find a solution. The kernel is ROM resident and we are using the target shell to load an application from the disk-on-chip into the RAM for execution with a startup script. The symbol table is currently loaded on the target. The one requirement that we have is that we must be able to download new application code onto the board via ethernet and the board must run this code on startup in a stand-alone configuration. Currently we download the .o file via FTP onto the disk-on-chip, and then the object file is loaded into RAM at startup via the startup script. How can I decrease the size of the kernel and the amount of RAM that is used. One idea that we have been addressing is to pre-link the module that resides on the disk-on-chip before it is loaded into the RAM and do away with the symbol table. Is this a good idea and if so how can this be accomplished? Is there a better way to do this? Would this configuration prevent us from using debugging tools during execution? Thanks, Bill Gibbs --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Is it necessary to program the bootloader on different chip than the application or TFFS? Date: Tue, 17 Sep 2002 18:54:40 GMT From: "Brian St. Pierre" Organization: Worldpath Internet Services (worldpath.net) Message-ID: References: <41c1fd2b.0209170118.76648c6c@posting.google.com> "Danny Cohen" wrote: > From you're experience: Do you see any SW or HW or System drawbacks in > the solution of programming both boot loader and application and File > System on the same Flash Chip? Not at all. This is a piece of cake. I've done this in a couple of different designs. Just make sure you reserve plenty of room for your boot loader. (The first time I did this we ran into problems because the boot loader grew and it was a mess to fix all the boards that needed to have more space reserved.) For extra security, we lock the sectors of the device that hold the boot loader. - -- Brian St. Pierre brian (at) bstpierre.org --------------------------- Newsgroups: comp.os.vxworks Subject: obsolete file descriptor Date: Tue, 17 Sep 2002 19:28:46 GMT From: "Brian St. Pierre" Organization: Worldpath Internet Services (worldpath.net) Message-ID: I'm using vxworks 5.4, dosFs2 and TFFS on an i28f128. I have a task running that keeps a file open and occasionally writes a log message to a file on the flash. After each write the file is flushed with ioctl(FIOFLUSH). Every once in a while, we'll get an error: 0x381400 == S_dosFsLib_FD_OBSOLETE. I've been able to force this in testing by doing a chkdsk "/flash" and then writing a log message. But we're not normally using chkdsk. Does anyone know what other operations will obsolete the fd's on the device? Also, is there any way of testing whether the fd is obsolete? If not, I guess I just have to go back to open/write/close-ing the file every time a new message comes in (which is quite a bit slower than keeping it open and flushing). - -- - -Brian St. Pierre brian (at) bstpierre.org --------------------------- Newsgroups: comp.os.vxworks Subject: Re: endPoolShow shows 0 free clusters Date: 17 Sep 2002 16:14:06 -0700 From: jchen@terayon.com (john) Organization: http://groups.google.com/ Message-ID: References: <6fe3f9e6.0209170740.7cf60648@posting.google.com> mchandnani@empirix.com (Monica Chandnani) wrote in message news:<6fe3f9e6.0209170740.7cf60648@posting.google.com>... > Summary: > > The number of free clusters shown by endPoolShow go down to 0; and all > network connections to the PowerPC die, i.e ping, ftp, telnet all > fail. > > Description: > I open 2 sockets to the PowerPC and send lots of requests at a very > high speed. Eventually, the PowerPC stops responding to ftp, telnet, > ping etc. While, when I see things through the serial connection, I > see all tasks (including the network task) look normal and checkStack > also shows no errors. However, when I run the endPoolShow command, I > see the number of free clusters to 0. Is there a way of increasing the > number of end driver clusters? > > Thanks Hi Monica If you have spy, try to run spy and see if your tNetTask is so overwhelmed that its not servicing the requests fast enough. That was our problem before, where tNetTask is so slow and we are sending it so many requests that it just got stuck. We eventually took out all calls to tNetTask and wrote our own driver. j --------------------------- Newsgroups: comp.os.vxworks Subject: Re: about bsp makefile under diab compiler Date: Tue, 17 Sep 2002 19:51:57 -0400 From: joe durusau Organization: WEBUSENET.com Message-ID: <3D87C01D.9FFE9EAF@bellsouth.net> References: <50cc31ed.0209170227.ac4d35c@posting.google.com> lijun wrote: > HI: > i should complie the bsp,but i have not tornado IDE in > unix enviroment.so i have to write makefile to realize my > purpose.so who can tell me how to write the makefile? If you have the bsp, the makefile is in $WIND_BASE/target/config/ as late as T II. I don't know about anything later. That one works with the gnu toolset, maybe you could make it work with the diab compiler. Never tried it, myself. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Booting Date: Wed, 18 Sep 2002 00:15:20 GMT From: "Thomas Øderud" Message-ID: Hi, I got an webphone with an StrongArm SA-1100 running VxWorks. I was hoping someone could tell me how to halt or stop the automated startup the device makes. Because it goes into some non shell GUI. Thomas Øderud Norway --------------------------- Newsgroups: comp.os.vxworks Subject: Multiple initialization problems with Tornado2.2/Gnu2.96 Date: 17 Sep 2002 17:46:11 -0700 From: sslartybart@hotmail.com (Sim Nanda) Organization: http://groups.google.com/ Message-ID: <31135f9a.0209171646.5da9010e@posting.google.com> Hi Y'all, I'm porting our code to Tornado2.2/Gnu2.96. We were using Tornado2.0/Gnu2.73. It went pretty smoothly, except that I see some strange behavior at runtime...so maybe not as smoothly as I think. It seems that some (if not all) of my static constructors are being called twice. The only reason I've noticed this is because one of my static constructors spawns a task. With Gnu2.96 compiled code I get not one but TWO of these tasks. Each one looks identical to it's twin but with different TaskId's. So it seems that the static constructor got called twice. I've looked through the code to find any pesky duplicate calls, but I can't find any. That and the fact that this double initialization doesn't happen when the same code is compiled with Gnu2.73 leads me to think there's something I don't know about the new compiler. Perhaps I've missed a compiler option or two. None of the new options *seem* relevant to static init. I've gone through the Release Notes as well, and can't find anything relevant to this behavior. Anybody else seen this behavior? bye, sim sslartybart AT hotmail --------------------------- Newsgroups: comp.os.vxworks Subject: Re: TLB MISS ERROR (Q) Date: 17 Sep 2002 18:07:40 -0700 From: essenlee@hanmail.net (LEE MYUNG SUK) Organization: http://groups.google.com/ Message-ID: References: <488e459a.0209141048.4038caea@posting.google.com> thank you for your answer. but, I don't settle this problem. when cache enabled, lcd task is broken. and after some time, exception error iss occured like below. what's the matter? - -> Implementation Dependent Data TLB Miss Exception current instruction address: 0x00001208 Machine Status Register: 0x08209032 Condition Register: 0x48000080 Task: 0xf6ede8 "tNetTask" 80154 vxTaskEntry +5c : shell () 74934 shell +180: 74960 () 74aa8 shell +2f4: ledRead () 8ef24 ledRead +164: read () 3128c read +10 : iosRead () 32718 iosRead +cc : tyRead () 7c310 tyRead +44 : semTake () 52ddc semTake +134: semBTake () tShell restarted. ohn_94501@yahoo.com (John) wrote in message news:<488e459a.0209141048.4038caea@posting.google.com>... > Hello, > > essenlee@hanmail.net (LEE MYUNG SUK) wrote in message news:... > > USING : VxWorks 5.4 mpc860 board > > > > question> > > application program is working, after some time, exception error is > > occured and rebooted. > > what's the metter? what is the mean of TLB Miss? > > A TLB miss exception is meant to cause the operating system to read > the page tables and load a translation from there into the TLB. The > TLB is a cache for page table entries, necessary to improve > performance. > > The exception is normally handled by VxWorks, but judging from the > address in the exception message something has failed in the handler > itself. Are you running with the cache and MMU enabled, or not? Have > you changed these settings in your BSP? What revision of the 860 > silicon are you using? Have you checked the BSP's target.nr file to > see if there are any known problems with your particular silicon > revision - the 8xx family is perhaps the worst CPU family out there in > terms of serious silicon-related errata. > > > please answer me vxworker. > > > > < error message > > > > > Exception at interrupt level: > > This is perhaps a more important statement than the exact exception > that occurred. One of your interrupt service routines is causing an > exception for some reason. You should try to find out which ISR it is > by testing them thoroughly one at a time. > > Of course, another possibility is that some part of your application > is corrupting system memory somewhere. You could try to examine the > stack using the stack pointer in the regs (see the address in the last > line of the exception message for where the regs were saved, and the > REG_SET definition in target/h/arch/ppc/regsPpc.h for the layout of > the reg set - the stack pointer for PPC is GPR 1). > > HTH, > > John... > > > Implementation Dependent Data TLB Miss > > Exception current instruction address: 0x00001208 > > Machine Status Register: 0x08209030 > > Condition Register: 0x42000040 > > Regs at 0x1fedc0 > > > > > > Press any key to stop auto-boot... > > 4 3 2 1 0 > > auto-booting... > > > > > > > > NTP > I > > > > NAME ENTRY TID PRI STATUS PC SP > > ERRNO DELAY > > ---------- ------------ -------- --- ---------- -------- -------- > > ------- ----- > > tExcTask excTask fe7018 0 PEND c3110 fe6f38 > > 0 0 > > tLogTask logTask fe4690 0 PEND c3110 fe45b8 > > 0 0 > > tWdbTask 9462c f48d78 3 PEND 9d4f8 f48bd0 > > 0 0 > > lanRx lanrx_main f46b50 10 PEND 9d4f8 f46568 > > 33 0 > > main NTPMAIN f41b18 20 DELAY a17c8 f41a80 > > 0 20 > > usrPort serial_main ffd948 40 READY a2204 ffd7b8 > > c0002 0 > > gpsEngine gps_main f41130 40 READY a17c8 f41098 > > 0 0 > > usrLcdKey lcd_key_main f3fb90 40 DELAY a17c8 f3fae0 > > 0 1 > > timeCode tcd_main f3e5f0 40 DELAY a17c8 f3e548 > > 0 1 > > telnetRx telnetrx_mai f3d050 40 PEND 9d4f8 f3cf10 > > 0 0 > > tNetTask netTask fa6748 50 PEND 9d4f8 fa6690 > > 3d 0 > > tSnmpd a4fec f50860 150 PEND 9d4f8 f4feb8 > > 0 0 --------------------------- Newsgroups: comp.os.vxworks Subject: Select() on changing fdset Date: Wed, 18 Sep 2002 01:02:34 GMT From: yang_christopher@hotmail.com (chris) Message-ID: Hi, In VxWorks, I know you can assign multiple IP addresses to a single interface (i.e. motfcc). If I want to allow 'x' number of IP addresses available through telnet, I know that I can create the 'x' number of file descriptors and then using select() wait for a signal. The question that I have is: If I do a select() on my fd and then afterward add one more file descriptor to the fd_set can I do this without using signals or select() timeouts? Or can I mess around with exception fds or even with the referenced read fd_set? Thanks for your help chris --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Select() on changing fdset Date: 18 Sep 2002 01:16:07 GMT From: Denis Perelyubskiy Message-ID: References: * chris [09-Tue-02 18:02 -0700]: [...] > The question that I have is: If I do a select() on my fd > and then afterward add one more file descriptor to the > fd_set can I do this without using signals or select() > timeouts? Or can I mess around with exception fds or > even with the referenced read fd_set? can you add some sort of "control" fd to your referenced read fd_set? and whenever that fires, you do something special... when you need to call select() on a new set, you just write to your "control" fd, and select() returns? hth, denis - -- 'From' email address is used as a sink. Not read. Ever. Instead, send to [t-h-e-u-k-r-a-i-n-i-a-n|h-o-t-m-a-i-l|c-o-m] (remove dashes, replace the first vertical bar with @, and second with a dot) --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Exception Handling in vxWorks. Date: 17 Sep 2002 18:46:50 -0700 From: tgisfree@yahoo.com (tzeghuan) Organization: http://groups.google.com/ Message-ID: <290ade1c.0209171746.37338a64@posting.google.com> References: <290ade1c.0209122144.5566acf4@posting.google.com> <488e459a.0209130932.1c877e18@posting.google.com> <290ade1c.0209150112.485fe964@posting.google.com> <488e459a.0209151106.3c45b4a@posting.google.com> Hi, I not too sure whether its a C++ or harware, but I think it should be a general kind of software exception. Btw, I am using a PowerPC architecture. vxWorks 5.4. According to the manual, a bit (FPSCR[zx]) will be set if there is a divide by zero exception. The problem now is i am not sure how to catch that exception. I know you guys are busy. Appreciate all kinds of helps. rdgs john_94501@yahoo.com (John) wrote in message news:<488e459a.0209151106.3c45b4a@posting.google.com>... > Hello, > > You are confusing C++ exceptions with hardware exceptions. The C++ > exception handling mechanism does not catch hardware exceptions (and > it is not meant to according to the specification - the fact that MS > C++ does this is because they have implemented a non-standard language > extension to do this). > > You did not specify the CPU or BSP that you are using (see my previous > comments about the minimum information you should supply when posting > questions here), so I cannot say if the architecture you have even > generates an exception for divide by zero - most software floating > point implementations do not, they simply return one of the IEEE > defined NaN values. > > Assuming that the architecture you are using does generate hardware > exceptions for divide by zero, you will need to catch the error using > signals. If you know enough about the internals of the compiler, then > you can create a signal handler that will throw you a C++ exception, > but it is not as simple as just throwing an C++ exception from the > signal handler. > > HTH, > > John... > > tgisfree@yahoo.com (tzeghuan) wrote in message news:<290ade1c.0209150112.485fe964@posting.google.com>... > > Hello, > > > > Thanks for the reply. > > > > It just a simple C++ codes for testing: > > > > try() > > { > > //create a divide by zero error here. > > } > > catch(...) > > { > > //print a error statement here. > > } > > > > with f-exception option enabled. > > > > I saw a similar question posted here a year ago, the reply was rather vague. > > > > Hope I didn't miss out some other settings. > > > > Thanks in advance. > > > > tzeghuan > > > > > > john_94501@yahoo.com (John) wrote in message news:<488e459a.0209130932.1c877e18@posting.google.com>... > > > Hello, > > > > > > Your question is so vague that I doubt you will get a useful answer. > > > Try posting some more information... here are a few questions for you > > > to think about: > > > > > > 1) Did you mean hardware exception handling or C++ exception handling > > > (very different things)? > > > > > > 2) What doesn't work exactly? Does it not compile? Are there errors > > > when you download it? Does it cause errors when you run it? > > > > > > 3) Have you got a small code fragment that you could post here? > > > > > > 4) Which architecture/BSP are you using? > > > > > > 5) What compiler flags did you use when you compiled it? > > > > > > 6) Have you read the manual? > > > > > > 7) Have you searched the archives of this group for possible answers? > > > > > > 8) Have you read the FAQ for this group? > > > > > > Please remember that people in this forum are not paid to answer your > > > questions, and are usually very busy with their own work. If you want > > > to get a good answer you will need to think up front about the kinds > > > of information that are relevant; most of the people here do not have > > > the time, nor the desire to extract the information from you over a > > > series of postings. > > > > > > That said, if you post enough information about your problem, often > > > you will get an answer. > > > > > > HTH, > > > > > > John... > > > > > > tgisfree@yahoo.com (tzeghuan) wrote in message news:<290ade1c.0209122144.5566acf4@posting.google.com>... > > > > Hi all, > > > > > > > > Recently i tried out some exception handling in both windows and vxworks. > > > > The same piece of code works in win environment but not in vxworks. > > > > Can anyone know about this give some clue? > > > > Is there a need to enable exception handling or any thing like that? > > > > > > > > Thanks in advance. > > > > > > > > regards > > > > tzeghuan --------------------------- Newsgroups: comp.os.vxworks Subject: Re: endPoolShow shows 0 free clusters Date: Wed, 18 Sep 2002 01:52:56 GMT From: "drdiags" Organization: AT&T Broadband Message-ID: References: <6fe3f9e6.0209170740.7cf60648@posting.google.com> Hello Monica, For the END Drivers private cluster pool, you should look in the sysXXXEnd.c file to see if the driver supports increasing the pool size through the parse string passed to the muxDevLoad() call to your END driver. If not, you may have to obtain the source from your local Wind River folks. I ran across this issue recently when I was running 3k packets between two sbc8260's at a very high rate. The motscc had < 100 clusters available before I started my test and quickly went to 0 (due to fragmentation and re-assembly requirements) The motfcc had > 500 and was able to maintain the rate quite nicely. I got the impression increasing the motscc to a higher value would require a little more research and I haven't been able to get back to it. Long story short, changing the END private cluster pool is driver dependent. See if yours supports this. Look at the parameters allowed in the parse string, it is sometimes documented in the BSP sysXXXEnd.c file or else look at the header file for your driver as a start. HTH... "Monica Chandnani" wrote in message news:6fe3f9e6.0209170740.7cf60648@posting.google.com... > Summary: > > The number of free clusters shown by endPoolShow go down to 0; and all > network connections to the PowerPC die, i.e ping, ftp, telnet all > fail. > > Description: > I open 2 sockets to the PowerPC and send lots of requests at a very > high speed. Eventually, the PowerPC stops responding to ftp, telnet, > ping etc. While, when I see things through the serial connection, I > see all tasks (including the network task) look normal and checkStack > also shows no errors. However, when I run the endPoolShow command, I > see the number of free clusters to 0. Is there a way of increasing the > number of end driver clusters? > > Thanks --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Select() on changing fdset Date: Wed, 18 Sep 2002 01:54:33 GMT From: yang_christopher@hotmail.com (chris) Message-ID: References: Hey Denis, Thanks for the reply. If I add a "control" fd to the fd_set then it would mean that I would have to know what the fd (something I would have to bind to I take) parameters are? Can I add a fd that I have absolutely no clue what the IP is (I've never tried it)? I'm taking a wild guess, but if I want to add a new fd to the list, could I make a "dummy" file descriptor, add it to the fd_set. When select() is in its blocking state (no timeout or exceptions in this call), could I bind to the dummy fd and except these new provisioning to take immediate effect? I guess its almost like a "control", except that select() doesn't return. Denis, I think you're right, it might just work, but it would be a bit ugly, since it would take another task to mess around with the fd in a different context. I'm just worry that select() is really blocking itself... err... waits in no cyclic pattern from incoming messages. chris Denis Perelyubskiy wrote: >* chris [09-Tue-02 18:02 -0700]: > [...] > > The question that I have is: If I do a select() on my fd > > and then afterward add one more file descriptor to the > > fd_set can I do this without using signals or select() > > timeouts? Or can I mess around with exception fds or > > even with the referenced read fd_set? > >can you add some sort of "control" fd to your referenced >read fd_set? and whenever that fires, you do something >special... when you need to call select() on a new set, you >just write to your "control" fd, and select() returns? > >hth, > >denis > >-- >'From' email address is used as a sink. Not read. Ever. >Instead, send to [t-h-e-u-k-r-a-i-n-i-a-n|h-o-t-m-a-i-l|c-o-m] >(remove dashes, replace the first vertical bar with @, >and second with a dot) --------------------------- Newsgroups: comp.os.vxworks Subject: How to set a breakpoint when the content of EAX is equal to specified value?? Date: Wed, 18 Sep 2002 10:06:32 +0800 From: "Seraph Ma" Organization: Bentium Ltd. (CN99) Message-ID: Hey all I just wanna set a breakpoint when the content of EAX ( Intel 32 architechture CPU like Pentium )is equal to specified value. Can anyone give me some help? Thanx a lot!!:) Seraph --------------------------- Newsgroups: comp.os.vxworks Subject: Help for DosFs in vxworks Date: 17 Sep 2002 19:32:40 -0700 From: robert_fan@citiz.net (Robert Fan) Organization: http://groups.google.com/ Message-ID: <4c2e8c4.0209171832.250308d8@posting.google.com> Hi all , I am using a big disk (ATA 80G) in vxworks with dosFS 2.0 , after I have created some files ( total > 4G Bytes) , I find that when we create new file on the disk for first time after disk is mounted , vxworks spend much time ( 10 ~ 30s ) to search FAT for a space . If we increase the size of sectors per cluster when formating , it will deduce the search time . But I find the max value for sectors per cluster is 254 , it still results vxworks spend much time to find a empty FAT when disk is big . Now I want to change the cluster size to 512k or more , Can any one tell me how to do ? when I try to change the sector size and blocks in ataDevCreate to simulate large sector with ata disk ,but dosFsVolFormat returns error . Can get big cluster size by this way ? I need your helps very much , any suggestion is welcomed ! Robert Fan --------------------------- Newsgroups: comp.os.vxworks Subject: Strang Stack Items, Need Help!!! Date: Wed, 18 Sep 2002 11:33:08 +0800 From: "Seraph Ma" Organization: Bentium Ltd. (CN99) Message-ID: hey all When I trace my application task's stack, I got this: - -> tt tNCMInp 13bfd2 vxTaskEntry +a : f2eca43 (0, 0, 0, 0, 0, 0, 0, 0, 0, 0) f2eca6d miNCMInputTask +441: WACOS_shmAllocMem (8) f2e1f25 WACOS_shmAllocMem+241: 0 (c94203c) 0 : 1bb5d4 ([f2def2d, 8, 10202, ffff, 0]) 1bb5fd excStub +35 : excExcHandle (e, d4fa0c0, d4fa098) 1643df excExcHandle +143: taskSuspend ([0, d4fa0c0, 0, 0, e0003]) value = 0 = 0x0 As what you have seen, there is a line "f2e1f25 WACOS_shmAllocMem+241: 0 (c94203c)" I just wonder what is the meaning of the function of '0'. I have seen this for many times. Do this mean my application task's stack had already been damanged by some tasks else? Or it's OK. If that,how can we explain the meaning of the '0'. Thank you all in advance.Any help informaition is appreciative! --------------------------- Newsgroups: comp.os.vxworks Subject: Re: 1 second becomes 4 seconds! Date: 17 Sep 2002 20:54:01 -0700 From: welson_sun@yahoo.com (Welson Sun) Organization: http://groups.google.com/ Message-ID: <9d9eae54.0209171954.66dcde8a@posting.google.com> References: <9d9eae54.0209160613.3043746e@posting.google.com> <9d9eae54.0209162211.78346c5@posting.google.com> <9d9eae54.0209170539.28250303@posting.google.com> "Dan Gold" wrote in message news:... > "Welson Sun" wrote in message > news:9d9eae54.0209170539.28250303@posting.google.com... > > Problem update: > > > > 3. My board is using a 25Mhz oscilliator and the CPU clock is 50Mhz, > > I have checked the MF field in PLCR, it is correct ( = 1 ). > > > > I think the problem should be related with the timebase clock rate or > > the decrement clock rate. I found the two defines in the BSP: > > > > #define TMBCLK_FREQ CRISTAL_FREQ /* = 25Mhz */ > > #define DEC_CLOCK_FREQ TMBCLK_FREQ > > By default, all PowerPC BSPs use the decrementer as the system clock. Look > (as Leonid said) at the sysClk driver. You should find it in > tornado/target/src/drv/timer/ppcDecTimer.c. It should use the > DEC_CLOCK_FREQ define in sysClkInit() and sysClkRateSet(). You need to sit > down with the 860 manual, in particular the chapter on dealing with the > clocks, and run the numbers from your .h file to convince yourself you know > how everything is working, i.e. the MF bit, TBS, etc... > > The decrementer counts at 1/4 the bus clock speed. So if your bus clock is > 50Mhz, your decrementer should be counting at 12.5Mhz. That divide by 4 is > either handled in the bsp specific .h file in the target directory, or as > part of the equation in ppcDecTimer.c - I've seen it done in both places, > depending on the BSP. > > Since the original problem was delaying for 1 second really delays for 4 > seconds, I would leave the TBS bit so it's 4 seconds, then look for the > missing divide by 4 in one of the defines for the decrementer, or in > ppcDecTimer.c > > hth, > Dan Gold gold@ensemble.com Thank you Dan, really good answer. The related configuration of my board is: External oscilliator : 25Mhz SPLL freq: 50Mhz (so the MF should be 1, I have checked that, it's right) External bus freq: 50Mhz (EBDF of SCCR should be 00b, it's right) According to Dan's answer, my DEC_CLOCK_FREQ should be SPLL_FREQ/4, so I changed my BSP, re-generated the vxWorks image, but not burned the flash( I have some problem burning the flash now), restarted target, and now taskDelay(sysClkRateGet()) now delays 2 seconds. Based on this, I set the TBS bit of SCCR, and now taskDelay(sysClkRateGet()) delays 1 second! Although it seems that this will solve my problem, but I still have two questions: 1. What the relation between TBS bit and DEC_CLOCK_FREQ? 2. What is the use of TMBCLK_FREQ defination? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How to generate rom images Date: 17 Sep 2002 21:42:55 -0700 From: maxg@korea.com (MaxG) Organization: http://groups.google.com/ Message-ID: <64440dc0.0209172042.24a4ef1e@posting.google.com> References: <3D873554.3090909@spacetec.no> When you build vxworks_rom, vxworks_rom.hex creat alse. So you can use objcopyppc tool in DOS Prompt as follow objcopyppc -I srec -O binary vxWorks_Rom.hex vxWorks_Rom.bin Download vxWorks_Rom.bin to mvme2400 board using "niop" and burn to flash memory as you did. Hans Berglund wrote in message news:<3D873554.3090909@spacetec.no>... > Hi > > I have a problem with generating ROM images for an mvve2400 board. > > I have tried vxWorks_rom and vxWorks_romResident images. I know > there are lots of people out there who manage to get these things > working. > > If you have no spare time to read on, please just post the > "objdumpppc -h" and "elfToBin" from a working image. Both rom > and romResident are of interest. > > > First case first: > > I "niop" the image to address 4000, and flash it with > pflash 4000:100000 ff000100. The latter becomes ff8000100 > and fff00100 after the flash select jumper is moved. > After a reset (and jumper is moved) it is completely silent. > > The objdumpppc -h says: > > vxWorks_rom: file format elf32-powerpc > > Sections: > Idx Name Size VMA LMA File off Algn > 0 .text 00001a3c fff00100 fff00100 00110100 2**2 > CONTENTS, ALLOC, LOAD, READONLY, CODE > 1 .data 000f6a68 00800000 00800000 00010000 2**2 > CONTENTS, ALLOC, LOAD, DATA > 2 .rodata 0000001c fff01b3c fff01b3c 00111b3c 2**2 > CONTENTS, ALLOC, LOAD, READONLY, DATA > 3 .sdata2 00000000 fff01b58 fff01b58 00111b58 2**2 > CONTENTS, ALLOC, LOAD, READONLY, DATA > 4 .got 00000010 008f6a68 008f6a68 00106a68 2**2 > CONTENTS, ALLOC, LOAD, DATA > 5 .sdata 00000000 008f6a78 008f6a78 00106a78 2**2 > CONTENTS, ALLOC, LOAD, DATA > 6 .sbss 00000000 00000000 00000000 00111b58 2**0 > CONTENTS > 7 .bss 00000000 008f6a78 008f6a78 00106a78 2**0 > ALLOC > 8 .comment 00000100 00000000 00000000 00111b58 2**0 > CONTENTS, READONLY > > The elfToBin says: > Entry address is fff00100 > seg0: Flags: 00000005 Offset: 00110100 VADDR: fff00100 Size: 00001a58 > seg1: Flags: 00000006 Offset: 00010000 VADDR: 00800000 Size: 000f6a78 > > > I thought it would be easier to make an rom resident image, and so I > did. objdumpppc -h says: > > vxWorks_romResident: file format elf32-powerpc > > Sections: > Idx Name Size VMA LMA File off Algn > 0 .text 000d8d54 fff00100 fff00100 00020100 2**3 > CONTENTS, ALLOC, LOAD, READONLY, CODE > 1 .data 0000b890 00100000 00100000 00010000 2**4 > CONTENTS, ALLOC, LOAD, DATA > 2 .rodata 0000ce18 fffd8e58 fffd8e58 000f8e58 2**3 > CONTENTS, ALLOC, LOAD, READONLY, DATA > 3 .sdata2 00000000 fffe5c70 fffe5c70 00105c70 2**2 > CONTENTS, ALLOC, LOAD, READONLY, DATA > 4 .ctors 0000000c 0010b890 0010b890 0001b890 2**0 > CONTENTS, ALLOC, LOAD, DATA > 5 .dtors 0000000c 0010b89c 0010b89c 0001b89c 2**0 > CONTENTS, ALLOC, LOAD, DATA > 6 .got 00000010 0010b8a8 0010b8a8 0001b8a8 2**2 > CONTENTS, ALLOC, LOAD, DATA > 7 .sdata 00000000 0010b8b8 0010b8b8 0001b8b8 2**2 > CONTENTS, ALLOC, LOAD, DATA > 8 .sbss 00000334 0010b8b8 0010b8b8 0001b8b8 2**2 > ALLOC > 9 .bss 0000581c 0010bbf0 0010bbf0 0001b8b8 2**3 > ALLOC > 10 .comment 00004008 00000000 00000000 00105c70 2**0 > CONTENTS, READONLY > > The elfToBin refuses to convert it. I just says: > Entry address is fff006e8 > seg0 is NULL! exiting > > > All help is appreciated. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Strang Stack Items, Need Help!!! Date: Wed, 18 Sep 2002 07:59:24 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: Reply-To: "Michael Lawnick" Sender: mlawnick@pd9524efa.dip.t-dialin.net Hi Seraph, >Do this mean my application task's stack had already been damanged by some tasks else? My tip: WACOS_shmAllocMem() calls another function via a FUNCPTR that is not initialised or on download you've got a 'unresolved symbol' message HTH Michael "Seraph Ma" schrieb im Newsbeitrag news:am8s5o$vdu$1@debian.bentium.com... > hey all > When I trace my application task's stack, I got this: > > -> tt tNCMInp > 13bfd2 vxTaskEntry +a : f2eca43 (0, 0, 0, 0, 0, 0, 0, 0, 0, 0) > f2eca6d miNCMInputTask +441: WACOS_shmAllocMem (8) > f2e1f25 WACOS_shmAllocMem+241: 0 (c94203c) > 0 : 1bb5d4 ([f2def2d, 8, 10202, ffff, 0]) > 1bb5fd excStub +35 : excExcHandle (e, d4fa0c0, d4fa098) > 1643df excExcHandle +143: taskSuspend ([0, d4fa0c0, 0, 0, e0003]) > value = 0 = 0x0 > > > As what you have seen, there is a line > > "f2e1f25 WACOS_shmAllocMem+241: 0 (c94203c)" > > I just wonder what is the meaning of the function of '0'. > I have seen this for many times. Do this mean my application task's > stack > had already been damanged by some tasks else? Or it's OK. If that,how can we > explain the meaning of the '0'. > > Thank you all in advance.Any help informaition is appreciative! > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: bootrom_uncmp build fails on adding #defs for tsfs on pcpentium Date: 17 Sep 2002 23:22:23 -0700 From: vijay.peshkar@wipro.com (Vijay Peshkar) Organization: http://groups.google.com/ Message-ID: Hi All, Am including the lines given below in config.h for booting with tsfs #undef CONSOLE_TTY #define CONSOLE_TTY NONE #undef WDB_TTY_CHANNEL #define WDB_TTY_CHANNEL 0 #undef WDB_COMM_TYPE #define WDB_COMM_TYPE WDB_COMM_SERIAL #define INCLUDE_TSFS_BOOT On compilation am getting following errors...Any suggestions on what's I am doing wrong?. C:\Tornado\target\lib\libPENTIUMgnuvx.a(uipc_mbuf.o)(.text+0x11d): undefined reference to `clDescTblNumEnt' C:\Tornado\target\lib\libPENTIUMgnuvx.a(uipc_mbuf.o)(.text+0x123): undefined reference to `clDescTbl' C:\Tornado\target\lib\libPENTIUMgnuvx.a(uipc_mbuf.o)(.text+0x128): undefined reference to `mClBlkConfig' C:\Tornado\target\lib\libPENTIUMgnuvx.a(uipc_mbuf.o)(.text+0x13f): undefined reference to `sysClDescTblNumEnt' C:\Tornado\target\lib\libPENTIUMgnuvx.a(uipc_mbuf.o)(.text+0x145): undefined reference to `sysClDescTbl' C:\Tornado\target\lib\libPENTIUMgnuvx.a(uipc_mbuf.o)(.text+0x14a): undefined reference to `sysMclBlkConfig' make.exe: *** [bootrom_uncmp] Error 0x1 Thanks and Regards, Vijay --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Exception Handling in vxWorks. Date: 17 Sep 2002 23:28:30 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0209172228.5f843cd0@posting.google.com> References: <290ade1c.0209122144.5566acf4@posting.google.com> <488e459a.0209130932.1c877e18@posting.google.com> <290ade1c.0209150112.485fe964@posting.google.com> <488e459a.0209151106.3c45b4a@posting.google.com> <290ade1c.0209171746.37338a64@posting.google.com> Hello, > I not too sure whether its a C++ or harware, but I think it should be > a general kind of software exception. No, C++ exceptions are totally different from hardware exceptions. > Btw, I am using a PowerPC architecture. vxWorks 5.4. > According to the manual, a bit (FPSCR[zx]) will be set if there is a > divide by zero exception. That will (a) only be true for floating point divide by zero, and (b) only be true on CPUs that have a h/w floating point unit (for example, the popular PPC860 family do not have h/w floating point, and will not generate an exception for any divide by zero). > The problem now is i am not sure how to catch that exception. If the hardware generates the exception, then you can catch it in your application using a signal handler. You should register to receive SIGFPE (floating point exception). Search on the web, or read the manuals for examples of registering signal handlers. HTH, John... > I know you guys are busy. > Appreciate all kinds of helps. > > rdgs > > > john_94501@yahoo.com (John) wrote in message news:<488e459a.0209151106.3c45b4a@posting.google.com>... > > Hello, > > > > You are confusing C++ exceptions with hardware exceptions. The C++ > > exception handling mechanism does not catch hardware exceptions (and > > it is not meant to according to the specification - the fact that MS > > C++ does this is because they have implemented a non-standard language > > extension to do this). > > > > You did not specify the CPU or BSP that you are using (see my previous > > comments about the minimum information you should supply when posting > > questions here), so I cannot say if the architecture you have even > > generates an exception for divide by zero - most software floating > > point implementations do not, they simply return one of the IEEE > > defined NaN values. > > > > Assuming that the architecture you are using does generate hardware > > exceptions for divide by zero, you will need to catch the error using > > signals. If you know enough about the internals of the compiler, then > > you can create a signal handler that will throw you a C++ exception, > > but it is not as simple as just throwing an C++ exception from the > > signal handler. > > > > HTH, > > > > John... > > > > tgisfree@yahoo.com (tzeghuan) wrote in message news:<290ade1c.0209150112.485fe964@posting.google.com>... > > > Hello, > > > > > > Thanks for the reply. > > > > > > It just a simple C++ codes for testing: > > > > > > try() > > > { > > > //create a divide by zero error here. > > > } > > > catch(...) > > > { > > > //print a error statement here. > > > } > > > > > > with f-exception option enabled. > > > > > > I saw a similar question posted here a year ago, the reply was rather vague. > > > > > > Hope I didn't miss out some other settings. > > > > > > Thanks in advance. > > > > > > tzeghuan > > > > > > > > > john_94501@yahoo.com (John) wrote in message news:<488e459a.0209130932.1c877e18@posting.google.com>... > > > > Hello, > > > > > > > > Your question is so vague that I doubt you will get a useful answer. > > > > Try posting some more information... here are a few questions for you > > > > to think about: > > > > > > > > 1) Did you mean hardware exception handling or C++ exception handling > > > > (very different things)? > > > > > > > > 2) What doesn't work exactly? Does it not compile? Are there errors > > > > when you download it? Does it cause errors when you run it? > > > > > > > > 3) Have you got a small code fragment that you could post here? > > > > > > > > 4) Which architecture/BSP are you using? > > > > > > > > 5) What compiler flags did you use when you compiled it? > > > > > > > > 6) Have you read the manual? > > > > > > > > 7) Have you searched the archives of this group for possible answers? > > > > > > > > 8) Have you read the FAQ for this group? > > > > > > > > Please remember that people in this forum are not paid to answer your > > > > questions, and are usually very busy with their own work. If you want > > > > to get a good answer you will need to think up front about the kinds > > > > of information that are relevant; most of the people here do not have > > > > the time, nor the desire to extract the information from you over a > > > > series of postings. > > > > > > > > That said, if you post enough information about your problem, often > > > > you will get an answer. > > > > > > > > HTH, > > > > > > > > John... > > > > > > > > tgisfree@yahoo.com (tzeghuan) wrote in message news:<290ade1c.0209122144.5566acf4@posting.google.com>... > > > > > Hi all, > > > > > > > > > > Recently i tried out some exception handling in both windows and vxworks. > > > > > The same piece of code works in win environment but not in vxworks. > > > > > Can anyone know about this give some clue? > > > > > Is there a need to enable exception handling or any thing like that? > > > > > > > > > > Thanks in advance. > > > > > > > > > > regards > > > > > tzeghuan --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Multiple initialization problems with Tornado2.2/Gnu2.96 Date: 17 Sep 2002 23:33:27 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0209172233.1c6a324f@posting.google.com> References: <31135f9a.0209171646.5da9010e@posting.google.com> Hello, As ever, what CPU are you using? Also, how are yuo loading the code (host based loader, target based loader, linked to the bootable image?). John... sslartybart@hotmail.com (Sim Nanda) wrote in message news:<31135f9a.0209171646.5da9010e@posting.google.com>... > Hi Y'all, > I'm porting our code to Tornado2.2/Gnu2.96. > We were using Tornado2.0/Gnu2.73. > It went pretty smoothly, except that I see some strange > behavior at runtime...so maybe not as smoothly as I think. > > It seems that some (if not all) of my static constructors > are being called twice. The only reason I've noticed this is > because one of my static constructors spawns a task. With Gnu2.96 > compiled code I get not one but TWO of these tasks. Each one looks > identical > to it's twin but with different TaskId's. So it seems that the static > constructor got called twice. I've looked through the code to find > any pesky duplicate calls, but I can't find any. That and the fact > that this double > initialization doesn't happen when the same code is compiled with > Gnu2.73 leads me to think there's something I don't know about the new > compiler. Perhaps I've missed a compiler option or two. None of the > new options *seem* relevant to static init. I've gone through the > Release Notes > as well, and can't find anything relevant to this behavior. > > Anybody else seen this behavior? > > bye, > sim > sslartybart AT hotmail --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Select() on changing fdset Date: 17 Sep 2002 23:40:34 -0700 From: huzy@sina.com (huzy) Organization: http://groups.google.com/ Message-ID: <3f27d50.0209172240.56517a45@posting.google.com> References: what you want to know just in <>. here, I give a small program fragment: - -------------- int listen_fd, client_fds[16+1], i; FD_SET rfds, wfds; listen_fd = socket(...); bind(listen_fd ,...); listen(listen_fd); while(1) { FD_ZERO(rfds); FD_SET(&rfds, listen_fd); for(i=0; i<16; i++) if (client_fds[i] != -1) FD_SET(&rfds, client_fds[i]); select(&rfds, &wfds, NULL, NULL); if (FD_ISSET(&rfds, listen_fd) { for(i=0; i<=16; i++) { if (client_fds[i] == -1) client_fds[i] = accept(listen_fd); } if (i == 16) { close(client_fds[i]); client_fds[i] = -1; } } for(int i=0; i<16; i++) { if (FD_ISSET(&rfds, client_fds[i]) ......... } - -------------- --------------------------- Newsgroups: comp.os.vxworks Subject: deciphering values from windview and trigger tool usage Date: 18 Sep 2002 00:27:54 -0700 From: vxnerd@rediffmail.com (subbarayan) Organization: http://groups.google.com/ Message-ID: dear mentors, 1)Can anyone tell me how to configure windview to view a graph of user defined tasks?My typical appln requires that the graph should show the name of tasks used in my programme.I have been unable to find my task names in graph where as i find generic tasknames as t14,t15,t16 etc and some other tasks like tnettask,idle task,texectask etc.Can anyone point me to info abt these various tasks displayed in graph of windview?I created a root task and spawned a task called taskone in my programme.I am unable to find taskone in my graph of windview.Is it appearing in some other name or what settings should i make to view info abt taskone in windview? the same code shown below was the one i tried.I expected taskone to be viewed in graph after calling root and executing it.I went through the manual but was unable to find the info i was looking for. 2)I was going through the triggering tool manual.After going through it ,i tried to use it to trigger windview for a sample application to start logging.I did the following: Opened trigger tool. Then defined context matches as tid1 which is my taskid of my task called taskone.defined actions as startwv. Then the following is my code: #include #include"vxWorks.h" #include"taskLib.h" void taskone(); void tasktwo(); int tid1,tid2; void root() { tid1=taskSpawn("tas1",90,0x100,20000,FUNCPTR(taskone),0,0,0,0,0,0,0,0,0,0); tid2=taskSpawn("tas2",91,0x100,20000,FUNCPTR(tasktwo),0,0,0,0,0,0,0,0,0,0); } void taskone() { printf("%d\n",taskIdSelf()); printf("i am first task in this pc \n"); printf("tasktwo created"); } void tasktwo() { printf("i am second task in this pc \n"); printf("releasing cpu to taskone \n"); } I am unable to find the trigger working inspite of me calling either root from shell or by calling taskone directly from shell.Is there some configuration error in it?I went through the manual but was unable to find the info i was looking for. I will be thankful to persons clarifying my doubt.I use tornado2.0 and i tried in simulator version. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: obsolete file descriptor Date: Wed, 18 Sep 2002 10:02:03 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: This issue has been mentioned here recently: It turns out that some people who use TFFS find out that sector 0 changes from time to time (presumably by TFFS), which in turn is detected by Dcache as "disk change", and that in turns renders all currently open FD's as obsolete, and remounts the device. Nobody has yet come up with a clear description about WHY does TFFS change values of the sector 0 (boot sector), or what sepcific values does it change, and there is no means as of yet to turn this activity off. You can however do away with Dcache - TFFS has no moving parts and thus the performance benefit of using disk cache is minimal. Without disk cache, performance will not signigicantly change, and there will not be any detection of sector 0 changes, so you will not get the "FD Obsolete" error. HTH, - - Leonid "Brian St. Pierre" wrote in message news:OnLh9.2499$Xj4.2543828@newshog.newsread.com... > I'm using vxworks 5.4, dosFs2 and TFFS on an i28f128. I have a task running > that keeps a file open and occasionally writes a log message to a file on > the flash. After each write the file is flushed with ioctl(FIOFLUSH). Every > once in a while, we'll get an error: 0x381400 == S_dosFsLib_FD_OBSOLETE. > > I've been able to force this in testing by doing a chkdsk "/flash" and then > writing a log message. But we're not normally using chkdsk. Does anyone know > what other operations will obsolete the fd's on the device? Also, is there > any way of testing whether the fd is obsolete? If not, I guess I just have > to go back to open/write/close-ing the file every time a new message comes > in (which is quite a bit slower than keeping it open and flushing). > > -- > -Brian St. Pierre > brian (at) > bstpierre.org > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Help for DosFs in vxworks Date: Wed, 18 Sep 2002 10:09:22 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <4c2e8c4.0209171832.250308d8@posting.google.com> Bob, you have misinterpretted your problem. When you make the first disk access, that is the time when dosFs 2.0 will actually mount the file system, and will check it for consistency, if it is configured to automatically do so. That is the reason for the 10 seconds delay. There are two ways to go about this: 1. (preferred) shortly after you initialize dosFs etc. in usrConfig.c, spawn a background task that does some sort of trivial access to the disk, e.g. open it's root directory and close, then exit the task. Now if you do this early enogh, the disk mount and ccanning will happen in parallel with your other initialization (network, application), and will not effectivly consume any of your critical time, and by the time your application needs disk access, the disk will be mounted and checked. 2. You can change the dosFs initialization flags and instruct it to skip checking, so the mount will happen immeidately. By the way, the dosFs data structure allows only one byte to hold the sectors-per-cluster value, thus the 254 maximum value can not be exceeded (thank MSFT). Changing size of disk sectors to something larger than 512 is possible, but requires low-level formatting of the disk, this can be done on SCSI disks, never seen anyone do it with ATA drives though, could be impossible. Hope this helps, - - Leonid "Robert Fan" wrote in message news:4c2e8c4.0209171832.250308d8@posting.google.com... > Hi all , > I am using a big disk (ATA 80G) in vxworks with dosFS 2.0 , > after I have created some files ( total > 4G Bytes) , I find > that when we create new file on the disk for first time after > disk is mounted , vxworks spend much time ( 10 ~ 30s ) to search > FAT for a space . If we increase the size of sectors per cluster > when formating , it will deduce the search time . But I find > the max value for sectors per cluster is 254 , it still results > vxworks spend much time to find a empty FAT when disk is big . > > Now I want to change the cluster size to 512k or more , Can > any one tell me how to do ? > > when I try to change the sector size and blocks in ataDevCreate > to simulate large sector with ata disk ,but dosFsVolFormat returns > error . Can get big cluster size by this way ? > I need your helps very much , any suggestion is welcomed ! > > Robert Fan --------------------------- Newsgroups: comp.os.vxworks Subject: EST backend for MDP823 & Tornado Date: Wed, 18 Sep 2002 12:36:37 +0400 From: "Denis Dubrovin" Organization: WEBPlus Ltd. Message-ID: Hi! I have some problem with use EST backends 2.5 for tornado on WinXP host. I compile vxworks as mentioned in documentation with wdbisnowexternal() function in it. And download BDX to the visionICE. When I start tornado and configuring target server for IP address of ICE and estppc as backend and ELF file as core file. When I start target server it says many warnings like "Target core CRC is 0 (computed from 0x01000000 to .....) Host core CRC .... CRCs are different! Can`t compute host core file CRC - length is invalid..." etc. As result debugger can`t connect to target and I fails. I try to start vxWorks on target previously starting target server by doing GO 100000. Same result. That I do wrong???? - -- Regards! Den --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Select() on changing fdset Date: Wed, 18 Sep 2002 10:43:16 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Chris, I think you are going the wrong way, why do you need to use select ? If all you need is to have multiple IP addresses (one or multiple interfaces) on a target, and have a service (e.g. Telnet) receive requests from any of thos interfaces, directed to any of the target's IP addresses, then all you need to do, is bind the socket to the woildcard IP address INADDR_ANY, and the specific port of your service. Thence you do the normal bind, listen and loop of accept routine. When a socket local IP address is wildcard, it will receive packets destined for any of the target's IP addresses. Then when the accept() call returns with a received new connection, it will report to you the source IP, and you will be able to get the local IP address the connection was destined to with getsockname(). My personal suggestion is to avoid using select() where possible - it is complex, slow and does not scale well for large number of connections. - - Leonid "chris" wrote in message news:KgQh9.76620$561.42473@news02.bloor.is.net.cable.rogers.com... > Hi, > > In VxWorks, I know you can assign multiple IP addresses to a single interface > (i.e. motfcc). > > If I want to allow 'x' number of IP addresses available through telnet, I know > that I can create the 'x' number of file descriptors and then using select() > wait for a signal. > > The question that I have is: If I do a select() on my fd and then afterward > add one more file descriptor to the fd_set can I do this without using signals > or select() timeouts? Or can I mess around with exception fds or even with > the referenced read fd_set? > > Thanks for your help > > chris > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Strang Stack Items, Need Help!!! Date: Wed, 18 Sep 2002 17:00:16 +0800 From: "Seraph Ma" Organization: Bentium Ltd. (CN99) Message-ID: References: Mike But it was not like that. WACOS_shmAllocMem() is written by me, and I know exactly it never calls any uninitialized or unresolved symbol. But Still thank you for you warm-hearted help! : - ) "Michael Lawnick" wrote in message news:am94ub$65n$1@snoopy.bndlg.de... > Hi Seraph, > > >Do this mean my application task's stack had already been damanged by some > tasks else? > > My tip: > WACOS_shmAllocMem() calls another function via a FUNCPTR that is not > initialised > or > on download you've got a 'unresolved symbol' message > > HTH > Michael > > "Seraph Ma" schrieb im Newsbeitrag > news:am8s5o$vdu$1@debian.bentium.com... > > hey all > > When I trace my application task's stack, I got this: > > > > -> tt tNCMInp > > 13bfd2 vxTaskEntry +a : f2eca43 (0, 0, 0, 0, 0, 0, 0, 0, 0, 0) > > f2eca6d miNCMInputTask +441: WACOS_shmAllocMem (8) > > f2e1f25 WACOS_shmAllocMem+241: 0 (c94203c) > > 0 : 1bb5d4 ([f2def2d, 8, 10202, ffff, 0]) > > 1bb5fd excStub +35 : excExcHandle (e, d4fa0c0, d4fa098) > > 1643df excExcHandle +143: taskSuspend ([0, d4fa0c0, 0, 0, e0003]) > > value = 0 = 0x0 > > > > > > As what you have seen, there is a line > > > > "f2e1f25 WACOS_shmAllocMem+241: 0 (c94203c)" > > > > I just wonder what is the meaning of the function of '0'. > > I have seen this for many times. Do this mean my application task's > > stack > > had already been damanged by some tasks else? Or it's OK. If that,how can > we > > explain the meaning of the '0'. > > > > Thank you all in advance.Any help informaition is appreciative! > > > > > > > > > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: differences between LogMsg() and printf() functions Date: 18 Sep 2002 05:01:22 -0700 From: vxnerd@rediffmail.com (subbarayan) Organization: http://groups.google.com/ Message-ID: dear mentors, Can anyone explain me the differences between LogMsg function and the printf function ?What are the advantages and disadvantages of using both?Under which circumstances should u use each of these functions? Advanced thanks to person rpelying me. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: differences between LogMsg() and printf() functions Date: Wed, 18 Sep 2002 08:45:49 -0400 From: "Stephane Ross" Message-ID: <1A_h9.29596$RO5.1188023@wagner.videotron.net> References: Hi, printf is non buffered IO (see previous discussion on printf vs fprintf). logMsg is buffered and outputs the traces using a high priority task (logTask). If you are in a multi-threaded application, you can be sure that the logMsg traces will be written in order of execution and won't be scrambled. But since every trace will produce a context switch, if you are doing a lot of tracing, the cost will be enormous. Hope this helps, Stephane "subbarayan" wrote in message news:f81c4277.0209180401.176b5eee@posting.google.com... > dear mentors, > Can anyone explain me the differences between LogMsg function and the > printf function ?What are the advantages and disadvantages of using > both?Under which circumstances should u use each of these functions? > > Advanced thanks to person rpelying me. --------------------------- Newsgroups: comp.os.vxworks Subject: linking module Date: 18 Sep 2002 05:47:30 -0700 From: gibbsw@battelle.org (Bill Gibbs) Organization: http://groups.google.com/ Message-ID: Can I statically link an object module to the kernel before download and then specify the address to load it? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: differences between LogMsg() and printf() functions Date: Wed, 18 Sep 2002 12:49:58 GMT From: David Cooper Organization: none Message-ID: References: printf() will try to display immediatley while logMsg() sends a message to logTask and logTask does the actual display. Since logMsg() does not do the display itself, it can be used in interrupt and driver code. ...david In article , vxnerd@rediffmail.com says... > dear mentors, > Can anyone explain me the differences between LogMsg function and the > printf function ?What are the advantages and disadvantages of using > both?Under which circumstances should u use each of these functions? > > Advanced thanks to person rpelying me. > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Select() on changing fdset Date: Wed, 18 Sep 2002 08:54:09 -0400 From: Keith Arner Organization: Marconi Message-ID: <313680C9A886D511A06000204840E1CF67A1A0-100000@whq-msgusr-02.pit.comms.marconi.com> References: Reply-To: Keith Arner On Wed, 18 Sep 2002, chris wrote: > In VxWorks, I know you can assign multiple IP addresses to a single interface > (i.e. motfcc). > > If I want to allow 'x' number of IP addresses available through telnet, I know > that I can create the 'x' number of file descriptors and then using select() > wait for a signal. I agree with Leonid in that you can probably get away with one socket to accept() on. > The question that I have is: If I do a select() on my fd and then afterward > add one more file descriptor to the fd_set can I do this without using signals > or select() timeouts? Or can I mess around with exception fds or even with > the referenced read fd_set? Assuming that you really do need to select() on multiple fds, the question about changing the fd set begs the question: from where do you expect to modify the fd set? If you are modifying it from within the task that is calling select(), then the answer is simple: change it however you like because once the task returns from select(), there is no "memory" of how you called select() in the first place. However, I assume that you intend to modify the fd set from a second task, while the first task is still somewhere within select(). This would be a bad idea: the second task has no insight into how select() is using the fd set that was passed in. If you need to modify it, based on a decision made in a second task, set up a pipe to wake the first task out of select(). Leonid goes on to say: > My personal suggestion is to avoid using select() where possible - it is > complex, slow and does not scale well for large number of connections. This is just misplaced advice. - I have not seen any compelling evidence that select() is any more complex than necessary. - select() may have a fair amount of overhead if you're trying to move large amounts of data around quickly. But if you're just working with telnet, the data throughput is almost negligible; after all, how fast can people reasonably type? - I agree that select() may not scale well for thousands of connections, but that is no reason to avoid it when you're only going to have a few dozen connetions. The VxWorks implementation of select() may not solve all problems, but it certainly is a useful tool for solving some problems. Keith - -- "Once something has gone horribly wrong..." "...it just gets horribler and horribler." "Then, it's a matter of finding the original horriblism." -- from a conversation on debugging problems in a linked list --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Select() on changing fdset Date: Wed, 18 Sep 2002 15:10:42 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <313680C9A886D511A06000204840E1CF67A1A0-100000@whq-msgusr-02.pit.comms.marconi.com> Keith, I have used select extensively on various types of Unix and on VxWorks, and I must insist that in the particular case of VxWorks there is no compelling reason to use select() because anything you would like to do can be done easier without select(). If you have only a small number of connection, then a dedicated task per connection is infinitely more simple. If you have more than 100 connections, the dedicated task solution becomes impractical, and that is where select() would become useless as well. It's hard limit is at 256 fd's, but even at 100 fd's its overhead is too much. Nobody has convinced me otherwise so far, and my VxWorks experience goes back to version 5.0.2 (can you remember when that was?). "Keith Arner" wrote in message news:313680C9A886D511A06000204840E1CF67A1A0-100000@whq-msgusr-02.pit.comms.m arconi.com... [snip] > Leonid goes on to say: > > > My personal suggestion is to avoid using select() where possible - it is > > complex, slow and does not scale well for large number of connections. > > This is just misplaced advice. > > - I have not seen any compelling evidence that select() is any more > complex than necessary. > > - select() may have a fair amount of overhead if you're trying to move > large amounts of data around quickly. But if you're just working with > telnet, the data throughput is almost negligible; after all, how fast can > people reasonably type? > > - I agree that select() may not scale well for thousands of connections, > but that is no reason to avoid it when you're only going to have a few > dozen connetions. > > The VxWorks implementation of select() may not solve all problems, but it > certainly is a useful tool for solving some problems. > > Keith > > -- > "Once something has gone horribly wrong..." > "...it just gets horribler and horribler." > "Then, it's a matter of finding the original horriblism." > -- from a conversation on debugging problems in a linked list > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: linking module Date: Wed, 18 Sep 2002 15:13:26 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Billy, if you have got a module statically linked to VxWorks, it's already got an address, and is loaded and ready to run as soon as VxWorks boots up. Why whould you want it to have any different an address ?! Your question looks prerry bizare, would you go a stap backwards and explain what is it exactly that you are trying to achive ? - - Leonid "Bill Gibbs" wrote in message news:de879fcf.0209180447.1c6a791@posting.google.com... > Can I statically link an object module to the kernel before download > and then specify the address to load it? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Select() on changing fdset Date: Wed, 18 Sep 2002 14:27:33 GMT From: "Brian St. Pierre" Organization: Worldpath Internet Services (worldpath.net) Message-ID: References: "chris" wrote: > Denis Perelyubskiy wrote: > >* chris [09-Tue-02 18:02 -0700]: > > > The question that I have is: If I do a select() on my fd > > > and then afterward add one more file descriptor to the > > > fd_set can I do this without using signals or select() > > > timeouts? Or can I mess around with exception fds or > > > even with the referenced read fd_set? > > > >can you add some sort of "control" fd to your referenced > >read fd_set? and whenever that fires, you do something > >special... when you need to call select() on a new set, you > >just write to your "control" fd, and select() returns? > > > If I add a "control" fd to the fd_set then it would mean that I would have to > know what the fd (something I would have to bind to I take) parameters are? > Can I add a fd that I have absolutely no clue what the IP is (I've never tried > it)? Your control fd doesn't need to be a socket. It could be a pipe. Then the other task (the one that's trying to add an fd to your set) could write to other side of the pipe to wake up the select() task. You'd know which fd was the pipe and you could test if that was the fd that woke you up. Then you could check a flag to see what you need to do. E.g. add an fd to the set, remove an fd from the set, etc. - -Brian --------------------------- Newsgroups: comp.os.vxworks Subject: using "/pcConsole/0" and "/tyCo/0" simultaneously Date: 18 Sep 2002 07:52:20 -0700 From: schanzf@thmulti.com (Friedhelm Schanz) Organization: http://groups.google.com/ Message-ID: <474839c9.0209180652.12a29513@posting.google.com> Hello, we are using X68 based hardware with standard VGA/PS2 and COM connectors. Problem: every output to STDOUT should appear on both devices (VGA and COM1) simultaneously. The system should accept inputs from both devices (PS2 and COM1) at any time. Is it possible to use the devices "/pcConsole/0" and "/tyCo/0" simultaneously for STDIN and STDOUT? I've tried to write a simple device driver, that opens "/pcConsole/0" and "/tyCo/0" in its xxxDrvOpen routine. If writing to the device, the written buffer is forwarded to "/pcConsole/0" and "/tyCo/0". This approach is working so far. But I've got problems concerning STDIN. Any tips are welcome. Thanks, Friedhelm --------------------------- Newsgroups: comp.os.vxworks Subject: Re: linking module Date: 18 Sep 2002 10:38:17 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: References: Sender: bpringle@DeadDuck "Bill Gibbs" wrote in message news:de879fcf.0209180447.1c6a791@posting.google.com... > Can I statically link an object module to the kernel before download > and then specify the address to load it? Leonid> Billy, if you have got a module statically linked to VxWorks, Leonid> it's already got an address, and is loaded and ready to run Leonid> as soon as VxWorks boots up. Why whould you want it to have Leonid> any different an address ?! I suspect that Bill wishes the code to run at an address with faster RAM, etc. He wants the code to be rommable, or "chunked" with the rest of the vxWorks image. You have two choices. 1) Compile the code as completely PC relative with no external references, add start and end markers, copy the code to the destination and use a function pointer to call the routine. a.c void doSeperateMemory(void) { /* work... */ } void endSeperateMemory(void) { /* nothing, just a marker. */ } b.c void relocater(void) { extern char doSeperateMemory; extern char endSeperateMemory; void * fastMem = (void*)FAST_MEM; /* or vector table, etc */ int size = (int) (&endSeperateMemory - &doSeperateMemory); if(size > FAST_MEM_SIZE) { printf(" BIG ERROR!\n"); while(1); } memcpy(fastMem,&doSeperateMemory,size); } void doFast(void) { void (*fastFnc)(void) = (void (*)(void))FAST_MEM; fastFnc(); } 2) Use a custom linker file, with the "LMA" load address different from the "VMA" virtual address. Copy the code to the destination and all references should be fixed appropriately by the linker. You can obtain the default linker file with "ldarm -verbose", "ldppc - -verbose", "ld68k -verbose", etc. The second is tougher to implement and often requires you just use Makefiles and not Tornado projects. However, there is less restrictions on the relocated code. hth, Bill Pringlemeir. - -- Your computer is great, right? You make me sick. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: bootrom_uncmp build fails on adding #defs for tsfs on pcpentium Date: 18 Sep 2002 08:24:53 -0700 From: egoisone@libero.it (Mic) Organization: http://groups.google.com/ Message-ID: References: What kind of BSP are you using? My config.h does not have those rows... Anyway, remember that you have to include this feature into Tornado project too. Mic. vijay.peshkar@wipro.com (Vijay Peshkar) wrote in message news:... > Hi All, > > Am including the lines given below in config.h for booting with tsfs > #undef CONSOLE_TTY > #define CONSOLE_TTY NONE > #undef WDB_TTY_CHANNEL > #define WDB_TTY_CHANNEL 0 > #undef WDB_COMM_TYPE > #define WDB_COMM_TYPE WDB_COMM_SERIAL > #define INCLUDE_TSFS_BOOT > > On compilation am getting following errors...Any suggestions on what's > I am doing wrong?. > > C:\Tornado\target\lib\libPENTIUMgnuvx.a(uipc_mbuf.o)(.text+0x11d): > undefined reference to > `clDescTblNumEnt' > C:\Tornado\target\lib\libPENTIUMgnuvx.a(uipc_mbuf.o)(.text+0x123): > undefined reference to > `clDescTbl' > C:\Tornado\target\lib\libPENTIUMgnuvx.a(uipc_mbuf.o)(.text+0x128): > undefined reference to > `mClBlkConfig' > C:\Tornado\target\lib\libPENTIUMgnuvx.a(uipc_mbuf.o)(.text+0x13f): > undefined reference to > `sysClDescTblNumEnt' > C:\Tornado\target\lib\libPENTIUMgnuvx.a(uipc_mbuf.o)(.text+0x145): > undefined reference to > `sysClDescTbl' > C:\Tornado\target\lib\libPENTIUMgnuvx.a(uipc_mbuf.o)(.text+0x14a): > undefined reference to > `sysMclBlkConfig' > make.exe: *** [bootrom_uncmp] Error 0x1 > > Thanks and Regards, > Vijay --------------------------- Newsgroups: comp.os.vxworks Subject: multiple session telnet server code Date: Wed, 18 Sep 2002 15:25:15 GMT From: inoue Organization: @Home Network Message-ID: <3D889BB6.412736F2@yahoo.co.jp> Hi, Any one knows where I can get multiple session telnet server code? I don't have VxWorks 5.5. but 5.4. So I need unsupported code. Thanks, Inoue --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Multiple initialization problems with Tornado2.2/Gnu2.96 Date: 18 Sep 2002 08:32:55 -0700 From: sslartybart@hotmail.com (Sim Nanda) Organization: http://groups.google.com/ Message-ID: <31135f9a.0209180732.6a3bbb2e@posting.google.com> References: <31135f9a.0209171646.5da9010e@posting.google.com> <488e459a.0209172233.1c6a324f@posting.google.com> Ah yes. Left out a few crucial details didn't I. Processor: PPC405 The image is downloaded using a target based loader (bootRom). thanks for your time, sim sslartybart AT hotmail john_94501@yahoo.com (John) wrote in message news:<488e459a.0209172233.1c6a324f@posting.google.com>... > Hello, > > As ever, what CPU are you using? Also, how are yuo loading the code > (host based loader, target based loader, linked to the bootable > image?). > > John... > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: multiple session telnet server code Date: Wed, 18 Sep 2002 18:10:05 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <3D889BB6.412736F2@yahoo.co.jp> You can get it from your local friendly WindRiver FAE (i.e. Field Application Engineer). In case he does not know what you are taling about, ask him for the reentrant telnet code developed by WindRiver Israel in 1998, it's supposedly in the FAE FTP area. "inoue" wrote in message news:3D889BB6.412736F2@yahoo.co.jp... > Hi, > > Any one knows where I can get multiple session telnet server code? > I don't have VxWorks 5.5. but 5.4. So I need unsupported code. > > Thanks, > Inoue --------------------------- Newsgroups: comp.os.vxworks Subject: Re: endPoolShow shows 0 free clusters Date: Wed, 18 Sep 2002 09:16:56 -0700 From: "Dan Gold" Organization: Mailgate.ORG Server - http://www.Mailgate.ORG Message-ID: References: <6fe3f9e6.0209170740.7cf60648@posting.google.com> Sender: gold@ensemble.com "drdiags" wrote in message news:Y%Qh9.504114$UU1.83800@sccrnsc03... > The motscc had < 100 clusters available before I started > my test and quickly went to 0 (due to fragmentation and re-assembly > requirements). The motfcc had > 500 and was able to maintain the rate quite nicely. I got > the impression increasing the motscc to a higher value would require a little > more research and I haven't been able to get back to it. Why "a little more research" ? Of course if you have the source, changing the number of buffers is trivial :-) Otherwise, usually the number of buffers is dependent on the number of Rx and Tx BDs, plus a small "fudge factor" for the loaned ones. So if you allocate more buffer descriptors, you get more buffers. If you just want more buffers, you may need to get the source from WRS and modify the arithmetic to get more buffers without more buffer descriptors. hth, Dan Gold gold@ensemble.com --------------------------- Newsgroups: comp.os.vxworks Subject: Convert GNU Make to vxWorks Project ?? Date: Wed, 18 Sep 2002 16:56:26 GMT From: "AKA" Organization: Compaq, part of the new hp Message-ID: <_e2i9.32$v75.633925@news.cpqcorp.net> Is there a procedure for converting a gnu makefile into a vxworks project workspace? I'm running Tornado 2 for Windows. What about "configure"? How do I replace this functionality under Tornado 2 for Windows? I would like to stay "pure" windows without using Cygnus or anything like that. Thanks, Russell --------------------------- Newsgroups: comp.os.vxworks Subject: Re: 1 second becomes 4 seconds! Date: Wed, 18 Sep 2002 10:14:12 -0700 From: "Dan Gold" Organization: Mailgate.ORG Server - http://www.Mailgate.ORG Message-ID: References: <9d9eae54.0209160613.3043746e@posting.google.com> <9d9eae54.0209162211.78346c5@posting.google.com> <9d9eae54.0209170539.28250303@posting.google.com> <9d9eae54.0209171954.66dcde8a@posting.google.com> Sender: gold@ensemble.com "Welson Sun" wrote in message news:9d9eae54.0209171954.66dcde8a@posting.google.com... > "Dan Gold" wrote in message news:... > > "Welson Sun" wrote in message > > news:9d9eae54.0209170539.28250303@posting.google.com... > The related configuration of my board is: > External oscilliator : 25Mhz > SPLL freq: 50Mhz (so the MF should be 1, I have checked that, it's > right) External bus freq: 50Mhz (EBDF of SCCR should be 00b, it's right) > > According to Dan's answer, my DEC_CLOCK_FREQ should be SPLL_FREQ/4, so > I changed my BSP and now taskDelay(sysClkRateGet()) now delays 2 seconds. > Based on this, I set the TBS bit of SCCR, and now taskDelay(sysClkRateGet()) delays 1 second! > > Although it seems that this will solve my problem, but I still have > two questions: > 1. What the relation between TBS bit and DEC_CLOCK_FREQ? > 2. What is the use of TMBCLK_FREQ defination? I don't think setting TBS is the correct thing to do. Of course I don't have your board, my 860 has a 50Mhz external oscillator and 50Mhz bus, so my settings are a little different. The ads860.h file has the #defines #define TMBCLK_FREQ CRISTAL_FREQ /* define the decrementer input clock frequency */ #define DEC_CLOCK_FREQ TMBCLK_FREQ I don't recall the clocking details on the ads board but I'm sure it's right for them :-) I have on my board TBS as 0, and setting #define TMBCLK_FREQ CRISTAL_FREQ /* define the decrementer input clock frequency */ #define DEC_CLOCK_FREQ TMBCLK_FREQ/4 This is because my CRISTAL_FREQ is 50Mhz, and my bus clock is 50Mhz. > 1. What the relation between TBS bit and DEC_CLOCK_FREQ? TBS selects the clock source for the timebase and decrementer. If you're not sure which clock source you're using, talk to the hardware guys about what they expect it to be. > 2. What is the use of TMBCLK_FREQ defination? That can be the actual frequency of the clock used for the time base/decrementer input, which is also the bus clock. I.e. if you have a 50Mhz oscillator and you divide by 1 to get the bus clock. Your clocking is a little different, because you're multiplying your 25Mhz external oscillator by 2, to get the 50Mhz bus clock, which also happens to be the time base source. So I would set TBS to 0, and use the following: #define TMBCLK_FREQ CRISTAL_FREQ /* 25Mhz for your board */ /* define the decrementer input clock frequency */ #define DEC_CLOCK_FREQ TMBCLK_FREQ/2 This is because your CRISTAL_FREQ is 25Mhz, and your bus clock is 50Mhz. The bottom line is that you know your bus clock is 50Mhz, so your decrementer will count every 12.5Mhz. Somehow you need to set the decrementer freq to get to 12.5Mhz. If you define TMBCLK to be 25Mhz (i.e. your crystal), then the DEC clock freq must be TMBCLK/2. If you define TMBCLK to be 50 (the actual bus frequency), then the DEC_CLOCK must be TMBCLK/4. hth, Dan Gold gold@ensemble.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Newbie needs help Date: 18 Sep 2002 10:33:41 -0700 From: rick@cyclone.com (Richard Nardone) Organization: http://groups.google.com/ Message-ID: <29b153d9.0209180933.5d639360@posting.google.com> References: "Yan Meunier" wrote in message news: Organization: General Dynamics, Electronics Systems Message-ID: Sender: news@wlbr.to.gd-es.com (System Administrator) I am running CrossWind/VxSim (Tornado 2.0) on a SUN platform. When I examine the static variables that are defined at the top of a single C source file (i.e., these variables are private to the C file), CrossWind (gdb) shows all of these variables as having the same memory location. Is this a bug? --------------------------- Newsgroups: comp.os.vxworks Subject: Maximum number for software breakpoints?? Date: 18 Sep 2002 14:02:23 -0700 From: quadmail@yahoo.com (Alex) Organization: http://groups.google.com/ Message-ID: <7b99c091.0209181302.315e6477@posting.google.com> The WDB_BP_MAX field is modifiable by the user, but is there an actual limit/restriction on how many BPs one can set up on a PPC target? Thanks, Alex McQuade --------------------------- Newsgroups: comp.os.vxworks Subject: Has bash ever been ported to vxWorks? Date: Wed, 18 Sep 2002 21:49:28 GMT From: "Ed Kaulakis" Organization: Giganews.Com - Premium News Outsourcing Message-ID: Reply-To: "Ed Kaulakis" Even a somewhat crippled version could be quite useful. - -- Ed Kaulakis #include --------------------------- Newsgroups: comp.os.vxworks Subject: Re: anyone out there who can help me on windview and trigger tool? Date: Wed, 18 Sep 2002 15:53:23 -0700 From: M Lang Organization: Jet Propulsion Laboratory - Pasadena CA Message-ID: <3D8903E3.4EFCB7EB@jpl.nasa.gov> References: Reply-To: Minh.Lang@jpl.nasa.gov subbarayan wrote: > > Dear Mr.Lawnik, > Sorry I made a mistake when i posted my code.actually i gave task name > as "tas1" and not t1 as mentioned in my code.But still I could not > find a tas1 in my windview.I am finding only generic names like > t4,t5,t6 etc.Where is the mistake? > Thanks for ur reply. Because as soon as you spawned your tasks, they are done and have exited and therefore no longer existed. -Minh- --------------------------- Newsgroups: comp.os.vxworks Subject: ASM macro in PPC860 with results in args Date: 18 Sep 2002 16:01:01 -0700 From: bwoodhouse@quantumdata.com (Bob W) Organization: http://groups.google.com/ Message-ID: <705be589.0209181501.171053ca@posting.google.com> Hello, I am converting a function from a Diab environment to the vxWorks-with-gnu environment. The processor remains a PPC860 throughout. I have compilable code that runs without crashes, but does not do what I desire. I am passing in 4 args. The goal is to have this same function prototype for the application code to see, but just change what is below. The arg called Result should be equal to arg p1 minus arg p2. The temp is a scratchpad area. These parameter are already declared in advance as "register INT32 p1, p2, result, temp". I have tried a dozen variations of this code, but here is one. I tried to constrain the result to appear in the same reg as the arg 2, etc. I have included asm.h and included "#define _ASMLANGUAGE". The value of result is always zero for whatever p1 and p2 may be (I use p1 > p2) for these tests. The subtraction is trying to take a POINT (x,y) stored as 2 16-bit values in an INT32 and subtract one POINT from the other, FWIW. Thanks, BobW Preview shows 1 line wrap that is not really in the real code /******************************************************* * SUBXY * Result = p1 - p2 * ********************************************************/ #define SUBXY( p1, p2, result, temp ) \ { register INT32 __p1 = (p1), __p2 = (p2), __result = (result), __temp = (temp); \ __asm__(" extsh 3,%0;\ extsh %3,%1;\ sub %3,3,%3;\ insrwi %2,%3,16,16;\ srawi 3,%0,16;\ srawi %3,%1,16;\ sub %3,3,%3;\ insrwi %2,%3,16,0"\ : "=&r" (__result)\ : "r" (__p1), "r" (__p2), "2" (__result), "r" (__temp)\ : "3", "cc"\ );\ } --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Select() on changing fdset Date: Thu, 19 Sep 2002 00:19:49 GMT From: yang_christopher@hotmail.com (chris) Message-ID: References: Keith, Leonid, I've completed implementing my code to inspect the file descriptor once accept has received a connection. I've been able to extract the local address (in this case an address part of the inaddr_any) and all that is left to do is compare this information to a table that I would construct myself to verify whether or not this is an address I would like keep a connection with. But you're both right, select() was a completely bad idea. I guess I should had mentioned that the number of connections that I would support would be very small (less than 20). But the number of IP address that I would filter would be 2 or 3 out of 7. Thank-you both for input, it has been proven extremely useful to me. chris "Leonid Rosenboim" wrote: >Keith, > >I have used select extensively on various types of Unix and on VxWorks, >and I must insist that in the particular case of VxWorks there is no >compelling >reason to use select() because anything you would like to do can be done >easier >without select(). > >If you have only a small number of connection, then a dedicated task >per connection is infinitely more simple. > >If you have more than 100 connections, the dedicated task solution >becomes impractical, and that is where select() would become useless >as well. It's hard limit is at 256 fd's, but even at 100 fd's its overhead >is too much. > >Nobody has convinced me otherwise so far, and my VxWorks experience goes >back to version 5.0.2 (can you remember when that was?). > >"Keith Arner" wrote in message >news:313680C9A886D511A06000204840E1CF67A1A0-100000@whq-msgusr-02.pit.comms.m >arconi.com... >[snip] >> Leonid goes on to say: >> >> > My personal suggestion is to avoid using select() where possible - it is >> > complex, slow and does not scale well for large number of connections. >> >> This is just misplaced advice. >> >> - I have not seen any compelling evidence that select() is any more >> complex than necessary. >> >> - select() may have a fair amount of overhead if you're trying to move >> large amounts of data around quickly. But if you're just working with >> telnet, the data throughput is almost negligible; after all, how fast can >> people reasonably type? >> >> - I agree that select() may not scale well for thousands of connections, >> but that is no reason to avoid it when you're only going to have a few >> dozen connetions. >> >> The VxWorks implementation of select() may not solve all problems, but it >> certainly is a useful tool for solving some problems. >> >> Keith >> >> -- >> "Once something has gone horribly wrong..." >> "...it just gets horribler and horribler." >> "Then, it's a matter of finding the original horriblism." >> -- from a conversation on debugging problems in a linked list >> --------------------------- Newsgroups: comp.os.vxworks Subject: Location of exception vectors in PPC BSPs Date: 18 Sep 2002 17:26:03 -0700 From: fgleeson@gi.com (FrankG) Organization: http://groups.google.com/ Message-ID: How does vxWorks locate the PPC exception vectors? Specifically I am trying to understand the wrSbc8260 BSP, but the question applies to all PPC BSPs. I think I understand the PPC exception vector scheme. The base is set according to 0xFFF00000 or 0x00000000. This depends on the IP bit in the MSR which originates at reset in the CIP bit in the HRCW. Whatever the base is you then have the following offsets: Quoting from 6-4 of MPCFPE32B.pdf. System Reset 0x0100 Machine Check 0x0200 DSI 0x0300 ISI 0x0400 External Interrupt 0x0500 Alignment 0x0600 Program 0x0700 FP unavailable 0x0900 ... Reserved 0x1000 - 0x2FFF There are several things I am confused about. It seems as though the romImage starts at address 0xFFF00100. I guess this causes the assembler routine romInit in romInit.s to execute on reset by virtue of the fact that it is linked first. It seems like there is no allowance for any other exception except the External Interrupt which is then connected from m8260IntrInit() via this API call. excIntConnect ((VOIDFUNCPTR *) _EXC_OFF_INTR, m8260IntrDeMux); I geuss this just overwrites whatever happens to be at address 0xFFF00500. Which is a predefined physical address location in the MMU? I geuss the writing of the TLBs in romInit.s set this up to whatever physical memory location. Clearly I have more questions than answers I am slowly integrating all this PPC/BSP knowledge. However I was hoping someone who had already figured this stuff out could give me a hint, or at least point me to some good documentation. --------------------------- Newsgroups: comp.os.vxworks Subject: udpstatShow question Date: 18 Sep 2002 18:01:03 -0700 From: essenlee@hanmail.net (LEE MYUNG SUK) Organization: http://groups.google.com/ Message-ID: - -> udpstatShow UDP: 8285 total packets 4830 input packets 3455 output packets 0 incomplete header 0 bad data length field 0 bad checksum 1378 broadcasts received with no ports 0 full socket 3452 pcb cache lookups failed 0 pcb hash lookup failed what is the mean of 3455 output packets and 3452 pcb cache lookups failed. Is the metter? -> 3452 pcb cache lookups failed using : vxworks 5.4 mpc860board wating for your answer. thank you. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Can't launch target server Date: 18 Sep 2002 19:48:11 -0700 From: emsindhwani@yahoo.com.sg (Mohit Sindhwani) Organization: http://groups.google.com/ Message-ID: References: Johan Borkhuis wrote in message news:... > emsindhwani@yahoo.com.sg (Mohit Sindhwani) wrote: > > p.s. I'm using Windows XP (if that helps) and have installed the patch > > that is supposed to be required to run Tornado under Windows 2000. > > This could be the problem. Tornado 2.2 is available and this should support > XP, so you might consider upgrading. > > Kind regards, > Johan Thanks Johan, I suspected that this might be the problem. Anyway, I'm following this up with WindRiver Singapore. Cheers Mohit. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: bootrom_uncmp build fails on adding #defs for tsfs on pcpentium Date: 18 Sep 2002 20:54:01 -0700 From: vijay.peshkar@wipro.com (Vijay Peshkar) Organization: http://groups.google.com/ Message-ID: References: Mic, Thanks for answering. Am using pcpentium target and those lines are not present by default and should be added in case you want tsfs support. The complete set of changes (both in config.h as well as in project) to be made to use these features are given in Tornado User's guide Pg125. Am getting the below errors even after following all the steps suggested in Manual. Regards, Vijay egoisone@libero.it (Mic) wrote in message news:... > What kind of BSP are you using? My config.h does not have those rows... > Anyway, remember that you have to include this feature into Tornado project too. > > Mic. > > vijay.peshkar@wipro.com (Vijay Peshkar) wrote in message news:... > > Hi All, > > > > Am including the lines given below in config.h for booting with tsfs > > #undef CONSOLE_TTY > > #define CONSOLE_TTY NONE > > #undef WDB_TTY_CHANNEL > > #define WDB_TTY_CHANNEL 0 > > #undef WDB_COMM_TYPE > > #define WDB_COMM_TYPE WDB_COMM_SERIAL > > #define INCLUDE_TSFS_BOOT > > > > On compilation am getting following errors...Any suggestions on what's > > I am doing wrong?. > > > > C:\Tornado\target\lib\libPENTIUMgnuvx.a(uipc_mbuf.o)(.text+0x11d): > > undefined reference to > > `clDescTblNumEnt' > > C:\Tornado\target\lib\libPENTIUMgnuvx.a(uipc_mbuf.o)(.text+0x123): > > undefined reference to > > `clDescTbl' > > C:\Tornado\target\lib\libPENTIUMgnuvx.a(uipc_mbuf.o)(.text+0x128): > > undefined reference to > > `mClBlkConfig' > > C:\Tornado\target\lib\libPENTIUMgnuvx.a(uipc_mbuf.o)(.text+0x13f): > > undefined reference to > > `sysClDescTblNumEnt' > > C:\Tornado\target\lib\libPENTIUMgnuvx.a(uipc_mbuf.o)(.text+0x145): > > undefined reference to > > `sysClDescTbl' > > C:\Tornado\target\lib\libPENTIUMgnuvx.a(uipc_mbuf.o)(.text+0x14a): > > undefined reference to > > `sysMclBlkConfig' > > make.exe: *** [bootrom_uncmp] Error 0x1 > > > > Thanks and Regards, > > Vijay --------------------------- Newsgroups: comp.os.vxworks Subject: Re: network task on vxworks Date: 18 Sep 2002 21:13:48 -0700 From: vijay.peshkar@wipro.com (Vijay Peshkar) Organization: http://groups.google.com/ Message-ID: References: Katz, Few debugging steps and posible causes of the problem.. Application/Driver buffer loaning: If sockets buffers are not being read, driver buffers will be depleted. As long as the driver is out of buffers, the driver will not be able to receive or send packets. Pings will be unanswered. Target server will be unable to contact the target. endPoolShow displays END driver's buffer pool status. If there are 0 free buffers, use inetstatShow to determine which socket has data in the receive queue. This information can be used to identify which task is not reading its sockets. See netPerformance for an explanation of network driver buffer loaning scheme. Driver: If ifShow, ipstatShow shows no data in the receive queue, the problem may be due to a faulty driver: If there are no other driver versions you can use, use the loopback driver. Run both the server and the client code in the target. Send to IP address 127.0.0.1. Regards, Vijay Ilia Bosis wrote in message news:... > Can You provide more info on the problem ? > > For example, When the target ceases to respond to ping: > If the Ethernet statistics counters, displayed by the ifShow, are > incremented ? > What counters exactly ? > > If the ping FROM Your target is dead as well ? > You can invoke the ping from the either Tornado or target shell by typing > "ping ". > > If the ping from Your target to that target itself is succeeded ? > That test involves the IP stack and not the network driver, > that will help You to isolate the driver problems from the stack' ones > > What's Your tasks' status (shown by the "i" command from the shell) ? > If some of them are suspended ? > > The tNET task must be pending most of the time. > If it's ready or running all the time, it's flooded by events - most chances > are, from an ill-behaved driver. > > Generally speaking, I'd never heard of such a catastrophic death of the > networking stack - > it's the very well-debugged BSD networking code, after all, > but there is a lot of problems with the poorly-written drivers. > Unfortunately, the WRS-supplied drivers aren't free from that plague, as > well as the home-written ones. > > So look at Your driver first. > > > Hag Sameach > Ilia. > -------------------------------------------- > Ilia Bosis > Software Engineer > PacketLight Networks > Tel: +972-9-7644228 > Fax: +972-9-7645444 > E-mail: Ilia_Bosis@packetlight.com > Web: www.packetlight.com > > > > > -----Original Message----- > > From: vxwexplo@lbl.gov [mailto:vxwexplo@lbl.gov] > > Sent: Tuesday, September 17, 2002 7:12 AM > > To: vxworks_users@csg.lbl.gov > > Subject: network task on vxworks > > > > > > Submitted-by vxwexplo-errs@csg.lbl.gov Mon Sep 16 21:53:38 2002 > > Submitted-by: Shlomo Katz > > > > hi all > > i have a strange problem, i am using vxworks on a ppc860 . > > after a while that the system is running the unit does not > > respond to ping > > request, the network task seems to be running . > > does somebody know of this problem > > > > any tips are welcome > > > > thanks > > > > katz shlomi > > www.rit.co.il --------------------------- Newsgroups: comp.os.vxworks Subject: Strang Stack Items, Need Help!! Date: Thu, 19 Sep 2002 14:17:41 +0800 From: "Seraph Ma" Organization: Bentium Ltd. (CN99) Message-ID: hey all When I trace my application task's stack, I got this: - -> tt tNCMInp 13bfd2 vxTaskEntry +a : f2eca43 (0, 0, 0, 0, 0, 0, 0, 0, 0, 0) f2eca6d miNCMInputTask +441: WACOS_shmAllocMem (8) f2e1f25 WACOS_shmAllocMem+241: 0 (c94203c) 0 : 1bb5d4 ([f2def2d, 8, 10202, ffff, 0]) 1bb5fd excStub +35 : excExcHandle (e, d4fa0c0, d4fa098) 1643df excExcHandle +143: taskSuspend ([0, d4fa0c0, 0, 0, e0003]) value = 0 = 0x0 As what you have seen, there is a line "f2e1f25 WACOS_shmAllocMem+241: 0 (c94203c)" I just wonder what is the meaning of the function of '0'. I have seen this for many times. Do this mean my application task's stack had already been damanged by some tasks else? Or it's OK. If that,how can we explain the meaning of the '0'. WACOS_shmAllocMem() never calls any uninitialized or unresolved symbol. Thank you all in advance.Any help informaition is appreciative! --------------------------- Newsgroups: comp.os.vxworks Subject: semGive failed with errno of S_semLib_INVALID_OPERATION. Date: Thu, 19 Sep 2002 15:02:12 +0800 From: "Seraph Ma" Organization: Bentium Ltd. (CN99) Message-ID: Hey all Does anyone know when will semGive fail with an errno of S_semLib_INVALID_OPERATION. The windriver's vxworks muanual say nothing about this. Also have I asked technical support from them,but they just asked me to refer their manual.God~~ Faint~~~~ If anybody ever see this errno when you call semGive,please reply. Thank you very much!!! Best Regard Seraph Ma --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Location of exception vectors in PPC BSPs Date: Thu, 19 Sep 2002 00:16:02 -0700 From: "Dan Gold" Organization: Mailgate.ORG Server - http://www.Mailgate.ORG Message-ID: References: Sender: gold@ensemble.com FrankG wrote in message ... >How does vxWorks locate the PPC exception vectors? It doesn't. They are set by the Green Book, which you've quoted below. They are either at 0x100 or 0xfff00100, as you correctly noted by MSR[IP]. >I think I understand the PPC exception vector scheme. The base is set >according to 0xFFF00000 or 0x00000000. This depends on the IP bit in >the MSR which originates at reset in the CIP bit in the HRCW. Whatever >the base is you then >have the following offsets: Quoting from 6-4 of MPCFPE32B.pdf. >System Reset 0x0100 >Machine Check 0x0200 >DSI 0x0300 >ISI 0x0400 >External Interrupt 0x0500 >Alignment 0x0600 >Program 0x0700 >FP unavailable 0x0900 >... >Reserved 0x1000 - 0x2FFF Correct. The reserved area actually contains vectors, some are only present on a 750, some only on a 74xx, etc... You need to see your processor specific HW manual. For the 8260, the 603e manual does just fine. >There are several things I am confused about. It seems as though the >romImage starts at address 0xFFF00100. I guess this causes the >assembler routine romInit in romInit.s to execute on reset by virtue >of the fact that it is linked first. Correct, and vxWorks links it at 0xfff00100. >It seems like there is no allowance for any other exception except the >External Interrupt which is then connected from m8260IntrInit() via >this API call. > > excIntConnect ((VOIDFUNCPTR *) _EXC_OFF_INTR, m8260IntrDeMux); > >I geuss this just overwrites whatever happens to be at address >0xFFF00500. There are other exceptions that are generated to the PPC, such as SMI and NMI. And in the case of the 8260 IRQ0-7. And also for the 8260 (and 860) you can have general purpose input lines generating interrupts. You are correct that there is no allowance for multiple general purpose interrupts to be hardwired to the external interrupt pin. Usually all external interrupts come through some sort of interrupt controller. The 8260 and 860 are special cases of the external interrupts. Look in tornado/target/src/intrCtl/xxx8260.c, I can't remember the exact name of the file. Basically vxWorks does the excIntConnect, because it needs a few things are the beginning of the exception handler (prologue and epilogue for the ISR at 0x500). So when you call intConnect(), vxWorks maintains a table of exception vectors. When the 8260 generates an interrupt, the code looks at the SIVEC register to get the particular vector, uses that as an index into it's table, and jumps to the function you've registered with intConnect(). That way you don't have to worry about the code at 0x500. >Which is a predefined physical address location in the MMU? I geuss >the writing of the TLBs in romInit.s set this up to whatever physical >memory location. Huh ? In romInit.s I think it only invalidates the TLBs. >Clearly I have more questions than answers I am slowly integrating all >this PPC/BSP knowledge. However I was hoping someone who had already >figured this stuff out could give me a hint, or at least point me to >some good documentation. The BSP code is the best documentation for some of this stuff. Of course the HW manuals too, and you can always call your WRS FAE and ask for source if you don't find what you're looking for. hth, Dan Gold gold@ensemble.com --------------------------- Newsgroups: comp.os.vxworks Subject: [Q] strange endDevTbl Date: 19 Sep 2002 00:25:36 -0700 From: kotaeji@hanmail.net (Kim Tae Hyung) Organization: http://groups.google.com/ Message-ID: Hi~ I cannot understand this strange behavior of my board. My target board is 405GP and has 3 pci dec21x4x mac. The internel ibmEmac is not used. My endDevTbl is as follows: END_TBL_ENTRY endDevTbl [] = /* SUCCESS */ { {0, END_LOAD_FUNC_PRI, END_LOAD_STRING, END_BUFF_LOAN_PRI, NULL, FALSE}, {1, END_LOAD_FUNC_PRI, END_LOAD_STRING, END_BUFF_LOAN_PRI, NULL, FALSE}, {2, END_LOAD_FUNC_PRI, END_LOAD_STRING, END_BUFF_LOAN_PRI, NULL, FALSE}, {3, END_LOAD_FUNC_PRI, END_LOAD_STRING, END_BUFF_LOAN_PRI, NULL, TRUE}, {0, END_TBL_END, NULL, 0, NULL, FALSE}, }; I think unit 3 is not necessary. But if I remove that sentence( like below), my board fails to initiate 3ports. END_TBL_ENTRY endDevTbl [] = /* FAIL */ { {0, END_LOAD_FUNC_PRI, END_LOAD_STRING, END_BUFF_LOAN_PRI, NULL, FALSE}, {1, END_LOAD_FUNC_PRI, END_LOAD_STRING, END_BUFF_LOAN_PRI, NULL, FALSE}, {2, END_LOAD_FUNC_PRI, END_LOAD_STRING, END_BUFF_LOAN_PRI, NULL, FALSE}, {0, END_TBL_END, NULL, 0, NULL, FALSE}, }; Any comments appreciated. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: 1 second becomes 4 seconds! Date: Thu, 19 Sep 2002 05:07:34 -0400 From: Varndell Engineering Organization: Posted Via Newsfeeds.com = SPEED+RETENTION+COMPLETION = http://www.newsfeeds.com Message-ID: <3D8993D6.213F3462@hotmail.com> References: <9d9eae54.0209160613.3043746e@posting.google.com> <9d9eae54.0209162211.78346c5@posting.google.com> <9d9eae54.0209170539.28250303@posting.google.com> <9d9eae54.0209171954.66dcde8a@posting.google.com> Welson Sun wrote: [snip] > External oscilliator : 25Mhz > SPLL freq: 50Mhz (so the MF should be 1, I have checked that, it's > right) > External bus freq: 50Mhz (EBDF of SCCR should be 00b, it's right) > > According to Dan's answer, my DEC_CLOCK_FREQ should be SPLL_FREQ/4, so > I changed my BSP, re-generated the vxWorks image, but not burned the > flash( I have some problem burning the flash now), restarted target, > and now taskDelay(sysClkRateGet()) now delays 2 seconds. > > Based on this, I set the TBS bit of SCCR, and now > taskDelay(sysClkRateGet()) delays 1 second! > > Although it seems that this will solve my problem, but I still have > two questions: > 1. What the relation between TBS bit and DEC_CLOCK_FREQ? There is no simple relationship between TBS and DEC_CLOCK_FREQ. TBS determines the source of the clock used for TMBCLK, but it is just one of several variables that determine what value *should* be assigned to DEC_CLOCK_FREQ. > 2. What is the use of TMBCLK_FREQ defination? TMBCLK can be determined at runtime knowing only the frequency of OSCCLK. Trying to define TMBCLK_FREQ at compile-time, with no corresponding runtime code to check or enforce it, is stupid. In ppc860timer.c, you'll see there are two important macros. They are DEC_CLOCK_FREQ and DEC_CLK_TO_INC. Both are assigned default values that should be over-ridden by BSP specific values. I yanked this from the 860 manual to show how simple all this could be were it not for all the unnecessary #defines. Table 15-5. TMBCLK Configuration SCCR[TBS] MODCK[1-2] MF+1 Clock Source TMBCLK at PORESET Prescaler 1 XX X GCLK2 16 0 0X X OSCCLK 4 0 1X 1,2 OSCCLK 16 0 1X >2 OSCCLK 4 From this table, you can that TBS determines the source for TMBCLK and, in combination with MODCK[1-2] and PLPRCR[MF], determines the prescaler. For 8xx processors, the simplest approach is to use TBS = 1, define DEC_CLOCK_FREQ such that it is equal to the SPLL frequency, and define DEC_CLK_TO_INC = 16. Note that this assumes that GCLK2 equals the SPLL frequency, an assumption that is only true if SCCR[DFNH] = 0. A better solution would be a rewrite of ppc860timer.c to compute this stuff at runtime. Regards, George Varndell - -- Varndell Engineering, LLC Embedded Systems Expertise http://www.varndellengineering.com - -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- http://www.newsfeed.com The #1 Newsgroup Service in the World! - -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Strang Stack Items, Need Help!! Date: Thu, 19 Sep 2002 11:13:57 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: Reply-To: "Michael Lawnick" Sender: mlawnick@pd9e65a79.dip.t-dialin.net Hi Seraph, > WACOS_shmAllocMem() never > calls any uninitialized or unresolved symbol. this does not look like a stack corruption, as excTask is able to back trace your calls. The back trace shows the call of an unknown function at address 0 in function WACOS_shmAllocMem(): > f2e1f25 WACOS_shmAllocMem+241: 0 (c94203c) As normaly there is never a function, the call is corrupt. The processor throws an exception at address 0 as there is no valid instruction or the MMU has marked this area as non-executable: > 0 : 1bb5d4 ([f2def2d, 8, 10202, ffff, 0]) > 1bb5fd excStub +35 : excExcHandle (e, d4fa0c0, d4fa098) So put a breakpoint at WACOS_shmAllocMem(), select 'Mixed Source and Disassembly' in 'View' menu and step through. Regards, Michael "Seraph Ma" schrieb im Newsbeitrag news:ambq67$q75$1@debian.bentium.com... > hey all > When I trace my application task's stack, I got this: > > -> tt tNCMInp > 13bfd2 vxTaskEntry +a : f2eca43 (0, 0, 0, 0, 0, 0, 0, 0, 0, 0) > f2eca6d miNCMInputTask +441: WACOS_shmAllocMem (8) > f2e1f25 WACOS_shmAllocMem+241: 0 (c94203c) > 0 : 1bb5d4 ([f2def2d, 8, 10202, ffff, 0]) > 1bb5fd excStub +35 : excExcHandle (e, d4fa0c0, d4fa098) > 1643df excExcHandle +143: taskSuspend ([0, d4fa0c0, 0, 0, e0003]) > value = 0 = 0x0 > > > As what you have seen, there is a line > > "f2e1f25 WACOS_shmAllocMem+241: 0 (c94203c)" > > I just wonder what is the meaning of the function of '0'. > I have seen this for many times. Do this mean my application task's > stack > had already been damanged by some tasks else? Or it's OK. If that,how can we > explain the meaning of the '0'. > WACOS_shmAllocMem() never > calls any uninitialized or unresolved symbol. > > Thank you all in advance.Any help informaition is appreciative! > > > --------------------------- Newsgroups: comp.os.vxworks Subject: sntpcTimeGet Date: 19 Sep 2002 02:40:55 -0700 From: asmita.joshi@wipro.com (Asmita Joshi) Organization: http://groups.google.com/ Message-ID: Hi We are using the SNTP Client in vxworks 5.4. We are facing a problem with the API sometimes. Our configuration is such that the client is on a LAN connected to the gateway. Across the gateway in another LAN, our SNTP server is present. It so happens that sometimes the API succeeds in retrieving time from the server whereas sometimes it times out. We have narrowed down the problem to the following: The API sntpcTimeGet succeeds when the ARP packet that is transmitted contains the gateways IP as its destination address. Now, when we connect the SNTP client in a subnet within the previous LAN, a wrong route entry gets added to the routing table. This entry contains the address of the NTP server. Because of this entry, the sntp client fails. When the API fails, the ARP packet transmitted before the NTP packet contains the SNTP servers IP as its destination address. Once this condition occurs (i.e. the sntpcTimeGet API fails), a reboot is the only solution to get the SNTP client to retrieve time from the SNTP server. Does anyone know how to solve this problem? How do we prevent incorrect route entries from being added. Also, is there any particular router setting that needs to be given to avoid this problem? thanks regards Asmita --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Strang Stack Items, Need Help!! Date: Thu, 19 Sep 2002 18:19:42 +0800 From: "Seraph Ma" Organization: Bentium Ltd. (CN99) Message-ID: References: Michael Thank you again for your warm-hearted help. These code runs in our company's gateway--which disable me from runtime debugging in Tornado IDE.It's really troublesome to debug these codes. OK, what I have got from the PGF is like this: Page Fault Program Counter: 0x00000000 Status Register: 0x00000003 Error Code: 0x0f26e429 Task: 0xd506ba8 "tNCMInp" Task not found. - -> - -> ISR,Error,FFFD300C,F26E78E,THU SEP 19 02:47:38 2002,CPU:Page Fault The task fails like this: -> taskShow tNCMInp,1 > > NAME ENTRY TID PRI STATUS PC SP ERRNO DELAY > ---------- ------------ -------- --- ---------- -------- -------- ------- - ----- > tNCMInp _miNCMInputT d7fb654 80 SUSPEND 12bd6e d7fb4f8 160068 0 > > stack: base 0xd7fb654 end 0xd7f1654 size 40952 high 1740 margin 39212 > > options: 0x4 > VX_DEALLOC_STACK > > edi = d7fb540 esi = d7fb654 ebp = d7fb500 esp = > d7fb4f8 > ebx = d7fb654 edx = fa0 ecx = f9f eax = > 0 > eflags = 246 pc = 12bd6e > -> printErrno 0x160068 > S_semLib_INVALID_OPERATION > value = 27 = 0x1b The PC counter is 0x00000000 as you have seen. BTW, have you ever got an errno of S_semLib_INVALID_OPERATION when you call semGive() ? Do you know when does this errno come up ? Thank you very very much for you,for all who pay attention to this post! Best Regard Seraph Ma "Michael Lawnick" wrote in message news:amc4gm$vsf$1@snoopy.bndlg.de... > Hi Seraph, > > WACOS_shmAllocMem() never > > calls any uninitialized or unresolved symbol. > > this does not look like a stack corruption, as excTask is able to back trace > your calls. The back trace shows the call of an unknown function at address > 0 in function WACOS_shmAllocMem(): > > f2e1f25 WACOS_shmAllocMem+241: 0 (c94203c) > As normaly there is never a function, the call is corrupt. The processor > throws an exception at address 0 as there is no valid instruction or the MMU > has marked this area as non-executable: > > 0 : 1bb5d4 ([f2def2d, 8, 10202, ffff, 0]) > > 1bb5fd excStub +35 : excExcHandle (e, d4fa0c0, d4fa098) > > So put a breakpoint at WACOS_shmAllocMem(), select 'Mixed Source and > Disassembly' in 'View' menu and step through. > > Regards, > Michael > > "Seraph Ma" schrieb im Newsbeitrag > news:ambq67$q75$1@debian.bentium.com... > > hey all > > When I trace my application task's stack, I got this: > > > > -> tt tNCMInp > > 13bfd2 vxTaskEntry +a : f2eca43 (0, 0, 0, 0, 0, 0, 0, 0, 0, 0) > > f2eca6d miNCMInputTask +441: WACOS_shmAllocMem (8) > > f2e1f25 WACOS_shmAllocMem+241: 0 (c94203c) > > 0 : 1bb5d4 ([f2def2d, 8, 10202, ffff, 0]) > > 1bb5fd excStub +35 : excExcHandle (e, d4fa0c0, d4fa098) > > 1643df excExcHandle +143: taskSuspend ([0, d4fa0c0, 0, 0, e0003]) > > value = 0 = 0x0 > > > > > > As what you have seen, there is a line > > > > "f2e1f25 WACOS_shmAllocMem+241: 0 (c94203c)" > > > > I just wonder what is the meaning of the function of '0'. > > I have seen this for many times. Do this mean my application task's > > stack > > had already been damanged by some tasks else? Or it's OK. If that,how can > we > > explain the meaning of the '0'. > > WACOS_shmAllocMem() never > > calls any uninitialized or unresolved symbol. > > > > Thank you all in advance.Any help informaition is appreciative! > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: endPoolShow shows 0 free clusters Date: Thu, 19 Sep 2002 06:47:03 -0400 From: Varndell Engineering Organization: Posted Via Newsfeeds.com = SPEED+RETENTION+COMPLETION = http://www.newsfeeds.com Message-ID: <3D89AB26.B5D985D2@hotmail.com> References: <6fe3f9e6.0209170740.7cf60648@posting.google.com> Monica Chandnani wrote: > > Summary: > > The number of free clusters shown by endPoolShow go down to 0; and all > network connections to the PowerPC die, i.e ping, ftp, telnet all > fail. > > Description: > I open 2 sockets to the PowerPC and send lots of requests at a very > high speed. Eventually, the PowerPC stops responding to ftp, telnet, > ping etc. While, when I see things through the serial connection, I > see all tasks (including the network task) look normal and checkStack > also shows no errors. However, when I run the endPoolShow command, I > see the number of free clusters to 0. Is there a way of increasing the > number of end driver clusters? You've been given some good tips and advice by other people. I wanted to add one point. If the driver's internal buffer pool is being depleted temporarily, due to unread sockets and/or buffers being help at the IP layer for extended periods to facilitate reassembly, those buffers will eventually make it back into the driver pool as various time-outs take effect. The possibility exists that the buffers are gone for good, leaked away due to mishandling by the driver. In this case, adding buffers will only delay the inevitable and you'll end up depleted anyway. Watch the driver pool for a while after it goes empty, if the buffers come back, then adding more buffers will help. In cases where a driver pool is not being leaked away at the driver level, frequent temporary depletion of the driver buffer pool *may* indicate excessive packet loss by the driver. This excessive loss will cause the stack to hold onto driver buffers while awaiting retransmission of lost packets for reassembly. The cure can be as simple as adding more buffer descriptors (as opposed to more buffers), or it might require more extensive driver repair and/or optimization. Regards, George Varndell - -- Varndell Engineering, LLC Embedded Systems Expertise Need an END driver for SiS900 devices? Contact us. http://www.varndellengineering.com - -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- http://www.newsfeed.com The #1 Newsgroup Service in the World! - -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Target Sever and Target Agent? Date: 19 Sep 2002 04:31:14 -0700 From: oe_zhou@sohu.com (oe_zhou) Organization: http://groups.google.com/ Message-ID: <758a30ff.0209190331.14f3dd80@posting.google.com> References: <758a30ff.0209170125.16150b2@posting.google.com> <3D8752CF.C95834AC@email.mot.com> But i only use a 100BASE-T hub,and no gateway. Then how can i do? \\\\\\\\\\\\\\\\\\\\\\\\ Jankesh Piparia wrote in message news:<3D8752CF.C95834AC@email.mot.com>... > Hi, > Did you try the routeAdd("0", " subnet>"). > > You will need to add this line of code in your usrAppInit.c before > building > the vxWorks image. > > -Jankesh > > oe_zhou wrote: > > > I use FEC of the 860T to boot, and it works well. > > But when the system image has been downloaded , > > the 100BASE-T network does'n work. And the link > > to target agent is losed.But the serial port > > and the system both works well. > > (860T + VxWorks). > > Help me!!! --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Booting Date: Thu, 19 Sep 2002 07:48:43 -0400 From: Joe Durusau Organization: Lockheed Martin Corporation Message-ID: <3D89B99B.7AC7127C@lmco.com> References: "Thomas Øderud" wrote: > > Hi, > > I got an webphone with an StrongArm SA-1100 running > VxWorks. > > I was hoping someone could tell me how to halt or stop the automated startup > the device makes. Because it goes into some non shell GUI. > > Thomas Øderud > Norway If it's running vxWorks.st (the standalone, non-development version of vxWorks), you just need to take out the memory and put in your own OS. Unless the developer left a hook in the system for future changes, there is no way to change the bootup sequence. If it is still running the development version, connecting a terminal to the serial port will suffice. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Select() on changing fdset Date: Thu, 19 Sep 2002 08:17:37 -0400 From: Keith Arner Organization: Marconi Message-ID: <313680C9A886D511A06000204840E1CF67A1A2-100000@whq-msgusr-02.pit.comms.marconi.com> References: <313680C9A886D511A06000204840E1CF67A1A0-100000@whq-msgusr-02.pit.comms.marconi.com> Reply-To: Keith Arner On Wed, 18 Sep 2002, Leonid Rosenboim wrote: > I have used select extensively on various types of Unix and on VxWorks, > and I must insist that in the particular case of VxWorks there is no > compelling reason to use select() because anything you would like to do > can be done easier without select(). > > If you have only a small number of connection, then a dedicated task > per connection is infinitely more simple. I disagree that it is necessarily more simple. The individual task might be less complex, but the system as a whole is not necessarily so. In a task-per-fd model, you've got to make the desicion about whether the tasks stick around or exit when the fd is closed. If they stick around, you need to build a task-pool-manager to find an available task to service requests. If you can tell a priori how many fds you will need, this could be as simple as a message queue, but becomes more complex if the pool needs to expand and contract dynamically. If tasks do exit, then you could be up against potentially ugly debugging issues like semTake()/exit()/taskSpawn() or logMsg()/exit()/taskSpawn(), wherein the identity of the offending task is ambiguous. Debugging a system with a static set of debuggable entities is generally easier than debugging a system with a dynamically changing set of debuggable entities. With multiple tasks, you need to be more concerned with dead locks and race conditions (traditionally nasty problems to deal with). With one task, these issues simply evaporate. Granted, the trade off is that in a single task model, blocking in one service function necessarily blocks all other service, but this may be acceptable (your service functions might never block, anyway). > If you have more than 100 connections, the dedicated task solution > becomes impractical, and that is where select() would become useless > as well. It's hard limit is at 256 fd's, but even at 100 fd's its overhead > is too much. But you are pre-supposing how much overhead is too much. In an application that has tight, real time deadlines, select()'ing on 100's of fds in the critical path is probably a bad idea. But if you're only using select() in a non-real time manner (such as the OP's example of user-interactive telnet), you'll probably never notice the millisecond or so that you might spend running through fds. > Nobody has convinced me otherwise so far, and my VxWorks experience goes > back to version 5.0.2 (can you remember when that was?). I will admit that I am a relative newcommer to VxWorks, and to this newsgroup. Perhaps in several years, I will learn to fear select(). In the mean time, can you provide concrete examples of why select() is so damaging as to be avoided whenever possible? Keith - -- "Once something has gone horribly wrong..." "...it just gets horribler and horribler." "Then, it's a matter of finding the original horriblism." -- from a conversation on debugging problems in a linked list --------------------------- Newsgroups: comp.os.vxworks Subject: Re: semGive failed with errno of S_semLib_INVALID_OPERATION. Date: Thu, 19 Sep 2002 08:23:03 -0400 From: Keith Arner Organization: Marconi Message-ID: References: Reply-To: Keith Arner On Thu, 19 Sep 2002, Seraph Ma wrote: > Hey all > Does anyone know when will semGive fail with an errno of > S_semLib_INVALID_OPERATION. > The windriver's vxworks muanual say nothing about this. > Also have I asked technical support from them,but they just asked me to > refer their manual.God~~ Faint~~~~ > If anybody ever see this errno when you call semGive,please reply. > Thank you very much!!! You can get S_semLib_INVALID_OPERATION if you try to do a semGive() on a mutex that you do not own. You can also get S_semLib_INVALID_OPERATION if you do a semFlush() on a mutex semaphore. Keith - -- "Once something has gone horribly wrong..." "...it just gets horribler and horribler." "Then, it's a matter of finding the original horriblism." -- from a conversation on debugging problems in a linked list --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Location of exception vectors in PPC BSPs Date: Thu, 19 Sep 2002 08:47:14 -0400 From: Varndell Engineering Organization: Posted Via Newsfeeds.com = SPEED+RETENTION+COMPLETION = http://www.newsfeeds.com Message-ID: <3D89C752.1C0309F8@hotmail.com> References: FrankG wrote: > > How does vxWorks locate the PPC exception vectors? > Specifically I am trying to understand the wrSbc8260 BSP, but the > question applies to all PPC BSPs. > > I think I understand the PPC exception vector scheme. The base is set > according to 0xFFF00000 or 0x00000000. This depends on the IP bit in > the MSR which originates at reset in the CIP bit in the HRCW. Whatever > the base is you then > have the following offsets: Quoting from 6-4 of MPCFPE32B.pdf. > System Reset 0x0100 > Machine Check 0x0200 > DSI 0x0300 > ISI 0x0400 > External Interrupt 0x0500 > Alignment 0x0600 > Program 0x0700 > FP unavailable 0x0900 > ... > Reserved 0x1000 - 0x2FFF > > There are several things I am confused about. It seems as though the > romImage starts at address 0xFFF00100. I guess this causes the > assembler routine romInit in romInit.s to execute on reset by virtue > of the fact that it is linked first. Sort of, but not all PPC systems boot from that address. > It seems like there is no allowance for any other exception except the > External Interrupt which is then connected from m8260IntrInit() via > this API call. > > excIntConnect ((VOIDFUNCPTR *) _EXC_OFF_INTR, m8260IntrDeMux); > > I geuss this just overwrites whatever happens to be at address > 0xFFF00500. By the time vx calls excIntConnect, the vectors are in RAM. > Which is a predefined physical address location in the MMU? I geuss > the writing of the TLBs in romInit.s set this up to whatever physical > memory location. No, rominit.s just invalidates the TLBs. They are initialized later. > > Clearly I have more questions than answers I am slowly integrating all > this PPC/BSP knowledge. However I was hoping someone who had already > figured this stuff out could give me a hint, or at least point me to > some good documentation. The following link is for vx 5.5, but answers questions like these. http://web2.windriver.com/windsurf/techpubs/ide/tornado/t22/pdf/vxworks/Guide1.pdf - -- Varndell Engineering, LLC Embedded Systems Expertise Need an END driver DP8382x GigE devices? Contact us. http://www.varndellengineering.com - -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- http://www.newsfeed.com The #1 Newsgroup Service in the World! - -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Location of exception vectors in PPC BSPs Date: Thu, 19 Sep 2002 08:47:30 -0400 From: Varndell Engineering Organization: Posted Via Newsfeeds.com = SPEED+RETENTION+COMPLETION = http://www.newsfeeds.com Message-ID: <3D89C762.4AA0F5DB@hotmail.com> References: FrankG wrote: > > How does vxWorks locate the PPC exception vectors? > Specifically I am trying to understand the wrSbc8260 BSP, but the > question applies to all PPC BSPs. > > I think I understand the PPC exception vector scheme. The base is set > according to 0xFFF00000 or 0x00000000. This depends on the IP bit in > the MSR which originates at reset in the CIP bit in the HRCW. Whatever > the base is you then > have the following offsets: Quoting from 6-4 of MPCFPE32B.pdf. > System Reset 0x0100 > Machine Check 0x0200 > DSI 0x0300 > ISI 0x0400 > External Interrupt 0x0500 > Alignment 0x0600 > Program 0x0700 > FP unavailable 0x0900 > ... > Reserved 0x1000 - 0x2FFF > > There are several things I am confused about. It seems as though the > romImage starts at address 0xFFF00100. I guess this causes the > assembler routine romInit in romInit.s to execute on reset by virtue > of the fact that it is linked first. Sort of, but not all PPC systems boot from that address. > It seems like there is no allowance for any other exception except the > External Interrupt which is then connected from m8260IntrInit() via > this API call. > > excIntConnect ((VOIDFUNCPTR *) _EXC_OFF_INTR, m8260IntrDeMux); > > I geuss this just overwrites whatever happens to be at address > 0xFFF00500. By the time vx calls excIntConnect, the vectors are in RAM. > Which is a predefined physical address location in the MMU? I geuss > the writing of the TLBs in romInit.s set this up to whatever physical > memory location. No, rominit.s just invalidates the TLBs. They are initialized later. > > Clearly I have more questions than answers I am slowly integrating all > this PPC/BSP knowledge. However I was hoping someone who had already > figured this stuff out could give me a hint, or at least point me to > some good documentation. The following link is for vx 5.5, but answers questions like these. http://web2.windriver.com/windsurf/techpubs/ide/tornado/t22/pdf/vxworks/Guide1.pdf - -- Varndell Engineering, LLC Embedded Systems Expertise Need an END driver DP8382x GigE devices? Contact us. http://www.varndellengineering.com - -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- http://www.newsfeed.com The #1 Newsgroup Service in the World! - -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Strang Stack Items, Need Help!! Date: Thu, 19 Sep 2002 14:43:05 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: Reply-To: "Michael Lawnick" Sender: mlawnick@pd9e65a79.dip.t-dialin.net Hi Seraph, this all shows me, that you *have* a jump to 0. As you ask for S_semLib_INVALID_OPERATION: A corrupted semId could cause this problem too. You will have to find out which line in source corresponds to WACOS_shmAllocMem+241. Show us the code of WACOS_shmAllocMem() and tell us Tornado version and H/W type, perhaps we can help. Regards, Michael "Seraph Ma" schrieb im Newsbeitrag news:amc8c1$ud0$1@debian.bentium.com... > Michael > Thank you again for your warm-hearted help. > These code runs in our company's gateway--which disable me from runtime > debugging in Tornado IDE.It's really troublesome to debug these codes. > OK, what I have got from the PGF is like this: > > Page Fault > Program Counter: 0x00000000 > Status Register: 0x00000003 > Error Code: 0x0f26e429 > Task: 0xd506ba8 "tNCMInp" > Task not found. > > -> > -> ISR,Error,FFFD300C,F26E78E,THU SEP 19 02:47:38 2002,CPU:Page Fault > > > The task fails like this: > -> taskShow tNCMInp,1 > > > > NAME ENTRY TID PRI STATUS PC SP ERRNO > DELAY > > ---------- ------------ -------- --- ---------- -------- -------- ------- > ----- > > tNCMInp _miNCMInputT d7fb654 80 SUSPEND 12bd6e d7fb4f8 160068 > 0 > > > > stack: base 0xd7fb654 end 0xd7f1654 size 40952 high 1740 margin 39212 > > > > options: 0x4 > > VX_DEALLOC_STACK > > > > edi = d7fb540 esi = d7fb654 ebp = d7fb500 esp = > > d7fb4f8 > > ebx = d7fb654 edx = fa0 ecx = f9f eax = > > 0 > > eflags = 246 pc = 12bd6e > > -> printErrno 0x160068 > > S_semLib_INVALID_OPERATION > > value = 27 = 0x1b > > > > The PC counter is 0x00000000 as you have seen. BTW, have you ever got an > errno of S_semLib_INVALID_OPERATION when you call semGive() ? Do you know > when does this errno come > up ? > Thank you very very much for you,for all who pay attention to this post! > > > Best Regard > Seraph Ma > > > > > > "Michael Lawnick" wrote in message > news:amc4gm$vsf$1@snoopy.bndlg.de... > > Hi Seraph, > > > WACOS_shmAllocMem() never > > > calls any uninitialized or unresolved symbol. > > > > this does not look like a stack corruption, as excTask is able to back > trace > > your calls. The back trace shows the call of an unknown function at > address > > 0 in function WACOS_shmAllocMem(): > > > f2e1f25 WACOS_shmAllocMem+241: 0 (c94203c) > > As normaly there is never a function, the call is corrupt. The processor > > throws an exception at address 0 as there is no valid instruction or the > MMU > > has marked this area as non-executable: > > > 0 : 1bb5d4 ([f2def2d, 8, 10202, ffff, 0]) > > > 1bb5fd excStub +35 : excExcHandle (e, d4fa0c0, d4fa098) > > > > So put a breakpoint at WACOS_shmAllocMem(), select 'Mixed Source and > > Disassembly' in 'View' menu and step through. > > > > Regards, > > Michael > > > > "Seraph Ma" schrieb im Newsbeitrag > > news:ambq67$q75$1@debian.bentium.com... > > > hey all > > > When I trace my application task's stack, I got this: > > > > > > -> tt tNCMInp > > > 13bfd2 vxTaskEntry +a : f2eca43 (0, 0, 0, 0, 0, 0, 0, 0, 0, 0) > > > f2eca6d miNCMInputTask +441: WACOS_shmAllocMem (8) > > > f2e1f25 WACOS_shmAllocMem+241: 0 (c94203c) > > > 0 : 1bb5d4 ([f2def2d, 8, 10202, ffff, 0]) > > > 1bb5fd excStub +35 : excExcHandle (e, d4fa0c0, d4fa098) > > > 1643df excExcHandle +143: taskSuspend ([0, d4fa0c0, 0, 0, e0003]) > > > value = 0 = 0x0 > > > > > > > > > As what you have seen, there is a line > > > > > > "f2e1f25 WACOS_shmAllocMem+241: 0 (c94203c)" > > > > > > I just wonder what is the meaning of the function of '0'. > > > I have seen this for many times. Do this mean my application task's > > > stack > > > had already been damanged by some tasks else? Or it's OK. If that,how > can > > we > > > explain the meaning of the '0'. > > > WACOS_shmAllocMem() never > > > calls any uninitialized or unresolved symbol. > > > > > > Thank you all in advance.Any help informaition is appreciative! > > > > > > > > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Using GNAT pro with VxWorks Date: Thu, 19 Sep 2002 14:55:08 +0200 From: "Marcus Hedlund" Organization: ericsson Message-ID: Hi! I'm trying to use GNAT pro to develop an Ada application using VxWorks. What I'm trying to figure out is how to merge the GNAT compiler into the Tornado environment? I have seen that the GNAT feature can become integrated in Tornado, but how? It has been difficult trying to get some information about this topic in the manuals. Are there anybody who have a clue how to do this? Any help would be greatly appreciated, /M.Hedlund --------------------------- Newsgroups: comp.os.vxworks Subject: ADS8260 BSP Date: Thu, 19 Sep 2002 14:59:38 +0200 From: "Achim Zimmer" Message-ID: we are using the ADS8260 BSP from Windriver. The FCC2 is used there as Ethernet interface. Now we try to use the MCC1, but the channel specific area is definied to be at 0x00 of DPRAM(DualPortRam). The problem I have, I can not find any information in the BSP were the BD's and the buffers (and also how many) for the FCC2 are located in DPRAM. Also I don't know where I have to change the init of FCC2 to the new values we need. The only thing that is described in the BSP is the UART SCC1 and SCC2. Did anybody know anything about the DPRAM layout and how to change it ?? Any help is welcome. Thanks very much. Achim --------------------------- Newsgroups: comp.os.vxworks Subject: help with vision ICE Date: Thu, 19 Sep 2002 13:04:29 GMT From: "Jim" Organization: Magma Communications Ltd. Message-ID: Hi i have a question, in my bootloader on the mpc860 i disable the busmonitor ... it seems the BDM (vision ICE) may be turning this back on (but the register sypcr is still ffffff08 as i set it) ... is it possible the bus will time out on BDM when accessing a slow device ... The problem in general is this ... i can boot my board without the ice ... but with the ice i die accessing a slow device 500 khz ... Help please, Jim --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Select() on changing fdset Date: Thu, 19 Sep 2002 15:04:03 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <313680C9A886D511A06000204840E1CF67A1A0-100000@whq-msgusr-02.pit.comms.marconi.com> <313680C9A886D511A06000204840E1CF67A1A2-100000@whq-msgusr-02.pit.comms.marconi.com> "Keith Arner" wrote in message news:313680C9A886D511A06000204840E1CF67A1A2-100000@whq-msgusr-02.pit.comms.m arconi.com... [snip] > I will admit that I am a relative newcommer to VxWorks, and to this > newsgroup. Perhaps in several years, I will learn to fear select(). In > the mean time, can you provide concrete examples of why select() is so > damaging as to be avoided whenever possible? > Because I have seen too many cases when people tried to use it and got themselves mixed up hopelessly. Several years of VxWorks support can teach you one thing or two about what is "too complex" for the average programmer, and what's not. By the way, clearly the "verage" has been going down continually for many years. Cheers! - - Leonid --------------------------- Newsgroups: comp.os.vxworks Subject: Booting from Flash Disk Date: Thu, 19 Sep 2002 14:23:14 +0100 From: "Paul" Organization: Logica UK Ltd Message-ID: <1032441796.250754@ernani.logica.co.uk> Sorry if this is an old question but... Is it possible to load a 'full' VxWorks kernel from a Flash Disk without using the intermediate 'bootrom' loader? Am I right in thinking that it should be possible provided I can get my kernel smaller than 0.5 MB? Has anyone out there managed to do this? My intention is to try to reduce boot up time by skipping this intermediate stage. If I have to strip my kernel to the bones to do this, then that's fine. I can always load in some additional modules (e.g. SIO drivers etc) after boot. I'm using a PIII based board, running VxWorks 5.4.2., with a Flash Disk that looks like a standard ATA Hard Drive. Thanks to all. --------------------------- Newsgroups: comp.os.vxworks Subject: Newbie Question (POSIX queues) Date: Thu, 19 Sep 2002 13:57:05 GMT From: "Yan Meunier" Organization: Aliant Internet Message-ID: Hi guys, I now moved one to playing with POSIX queues. I was able to send characters (which is a #define) from one task to another without any problems. Now I'm trying to send a pointer to a class I wrote onto the queue. The problem comes when I try to receive it. It return me an ERROR with a EMSGSIZE ERRNO. I am able to the same type of message queueing with the use of Wind Message Queues. Is there something I should know about the POSIX message Queue, that's diffrent then the Wind Message Queue as far as sending and receiving goes? If this Question should be asked in another group, could anyone point me to one? Thank you, Yan Meunier --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Location of exception vectors in PPC BSPs Date: 19 Sep 2002 07:16:56 -0700 From: rick@cyclone.com (Richard Nardone) Organization: http://groups.google.com/ Message-ID: <29b153d9.0209190616.1a897465@posting.google.com> References: fgleeson@gi.com (FrankG) wrote in message news:... > How does vxWorks locate the PPC exception vectors? > Specifically I am trying to understand the wrSbc8260 BSP, but the > question applies to all PPC BSPs. > > I think I understand the PPC exception vector scheme. The base is set > according to 0xFFF00000 or 0x00000000. This depends on the IP bit in > the MSR which originates at reset in the CIP bit in the HRCW. Whatever > the base is you then > have the following offsets: Quoting from 6-4 of MPCFPE32B.pdf. > System Reset 0x0100 > Machine Check 0x0200 > DSI 0x0300 > ISI 0x0400 > External Interrupt 0x0500 > Alignment 0x0600 > Program 0x0700 > FP unavailable 0x0900 > ... > Reserved 0x1000 - 0x2FFF This is true, except the Decrementer exception is connected at offset 0x0900. > > There are several things I am confused about. It seems as though the > romImage starts at address 0xFFF00100. I guess this causes the > assembler routine romInit in romInit.s to execute on reset by virtue > of the fact that it is linked first. After the HRCW is read, the first instruction fetch is from the system reset exception vector located at offset 0x0100 from the exception base. > It seems like there is no allowance for any other exception except the > External Interrupt which is then connected from m8260IntrInit() via > this API call. > > excIntConnect ((VOIDFUNCPTR *) _EXC_OFF_INTR, m8260IntrDeMux); > > I geuss this just overwrites whatever happens to be at address > 0xFFF00500. The PPC architecture has one main external interrupt. The ISR placed at this location has lots of work to do to figure out which interrupt(s) is present and then dispatch the appropriate processing (another ISR). The Decrementer exception is also used as the system cadence for many BSPS. This determines when the kernel is activated, the VxWork Tick, which is usually set to 60Hz or 100Hz. > Which is a predefined physical address location in the MMU? I geuss > the writing of the TLBs in romInit.s set this up to whatever physical > memory location. The BSP I worked with on the 8260 did a one to one mapping using the four IBATS and four DBATS. Look at the PowerPC Programming Environments for 32-Bit Microprocessors manual, chapter 7 MMU. > > Clearly I have more questions than answers I am slowly integrating all > this PPC/BSP knowledge. However I was hoping someone who had already > figured this stuff out could give me a hint, or at least point me to > some good documentation. --------------------------- Newsgroups: comp.os.vxworks Subject: dosFs2 Date: 19 Sep 2002 07:42:16 -0700 From: molochai@cybermuppet.co.uk (Molochai) Organization: http://groups.google.com/ Message-ID: We're seeing some strange behaviour with a file on a dosFs2 device becoming suddenly acquiring an extend of either 0 or 0xffffffff. Sometimes the directory entry on the device indicates the same, other times it doesn't. I'm aware of http://www.xs4all.nl/~borkhuis/vxworks/vxw_pt3.html#3.1 (With DosFS2 some of our files started disappearing). Can anybody elaborate on precisely what 'files disappearing' is here please? We're certainly running on PPC hardware and using lots of floating point. Additionally, in trying see see how the directory entry might have been corrupted on the file, I've traced all write requests through the block layer and have observed a single block write request when files are opened for reading. I wasn't aware that FAT16 had a last-accessed timestamp. Our system does not have an RTC. Is it that dosFs2 updating the actual filestamp on each file as it's opened? Thanks M --------------------------- Newsgroups: comp.os.vxworks Subject: double linked list Date: 19 Sep 2002 07:51:04 -0700 From: derfan2001@altavista.de (Mario Schattschneider) Organization: http://groups.google.com/ Message-ID: <13dfb019.0209190651.36ca7dbf@posting.google.com> Hi everyone, I want to write a linked list in C++ but it dosen't work I have a class Node which looks like this: class Node { Node *next; Node *previous; int lineNumber; int taskId; char filename[100]; Noder(int line, int taskId, char filename) { .... } } Node; How dose I use this to edit or build a linked list? Thanks for your help. Mario --------------------------- Newsgroups: comp.os.vxworks Subject: Re: dosFs2 Date: Thu, 19 Sep 2002 16:59:01 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Yes, there is a last-accessed timestamp in DosFs on all systems, and yes, DosFs2.0 does update it every time it's accessed, some people have complained about that, and it seems that WindRiver should add an option to disable this when not required. The time is taken from the current VxWorks "wall clock" (see clock_gettime() ) As to the rest of your issues, I am not sure I understand what is it exactly that happens to you, and what kind of disk ar eyou using. "Molochai" wrote in message news:e030048c.0209190642.5672023c@posting.google.com... > We're seeing some strange behaviour with a file on a dosFs2 device > becoming suddenly acquiring an extend of either 0 or 0xffffffff. > Sometimes the directory entry on the device indicates the same, other > times it doesn't. > > I'm aware of http://www.xs4all.nl/~borkhuis/vxworks/vxw_pt3.html#3.1 > (With DosFS2 some of our files started disappearing). Can anybody > elaborate on precisely what 'files disappearing' is here please? We're > certainly running on PPC hardware and using lots of floating point. > > > Additionally, in trying see see how the directory entry might have > been corrupted on the file, I've traced all write requests through the > block layer and have observed a single block write request when files > are opened for reading. I wasn't aware that FAT16 had a last-accessed > timestamp. > Our system does not have an RTC. Is it that dosFs2 updating the actual > filestamp on each file as it's opened? > > Thanks > > M --------------------------- Newsgroups: comp.os.vxworks Subject: Re: double linked list Date: Thu, 19 Sep 2002 17:01:27 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <13dfb019.0209190651.36ca7dbf@posting.google.com> Mario - go an buy yourself a book on "Data Structures and Algorythms" - this is a mandatory course in every university computer science program, and I am sure you can find a book on this subject in practically every language you can imagine. Study the book content, and I assure you, you will be rewarded in the future !! "Mario Schattschneider" wrote in message news:13dfb019.0209190651.36ca7dbf@posting.google.com... > Hi everyone, > > I want to write a linked list in C++ but it dosen't work > > I have a class Node which looks like this: > > class Node > { > Node *next; > Node *previous; > int lineNumber; > int taskId; > char filename[100]; > > Noder(int line, int taskId, char filename) > { > .... > } > > } Node; > > How dose I use this to edit or build a linked list? > > Thanks for your help. > > Mario --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ADS8260 BSP Date: Thu, 19 Sep 2002 08:57:57 -0700 From: "Dan Gold" Organization: Mailgate.ORG Server - http://www.Mailgate.ORG Message-ID: References: Sender: gold@ensemble.com "Achim Zimmer" wrote in message news:amchnt$4ovf3$1@ID-102608.news.dfncis.de... > we are using the ADS8260 BSP from Windriver. The FCC2 is used there as > Ethernet interface. > Now we try to use the MCC1, but the channel specific area is definied to be > at 0x00 of DPRAM(DualPortRam). > > The problem I have, I can not find any information in the BSP were the BD's > and the buffers (and also how many) for the FCC2 are located in DPRAM. Look at the FCC2 DPRAM structure. It's located at IMMR+0x8500. Dump it, then look at the FCC parameter ram table in chapter 28 of the 8260 UM. That will tell you the base of the BDs, and you can look for the WRAP bit to find out how many there are. That's one way. I thought WRS gave you the ability to specify in the load string (in EndDevTbl) where the BDs were located. Remember there is an errata that says BDs for Ethernet FCCs should be placed in extneral memory, not in DPRAM. So don't be surprised if you look and find them in external memory. The actual buffers themselves will certainly not be located in DPRAM. At 1500 bytes each, you'd run out of DPRAM pretty quick. This stuff should be documented in either the BSP docs, or configNet.h, or sysMotXXX.c, for the BSP. Or just ask WRS for the source to motFcc.c. > Also I don't know where I have to change the init of FCC2 to the new values we need. Again, you may not need source, just the ability to understand configNet.h or sysMotXXX.c. hth, Dan Gold gold@ensemble.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ASM macro in PPC860 with results in args Date: Thu, 19 Sep 2002 11:54:04 -0500 From: "Marty Schrader" Organization: Parsec Systems, Inc. Message-ID: <3cni9.519$b_1.203@fe01> References: <705be589.0209181501.171053ca@posting.google.com> Why do you have to do this in assembly, Bob? Call me. Oh, I'm working on the other thing we discussed, too. - -- Marty Schrader Embedded Technology Consulting from Parsec Systems, Inc. 0N107 Cottonwood Drive Wheaton, IL 60187 (630) 588-0241 www.parsecsystemsinc.com marty@parsecsystemsinc.com --------------------------- Newsgroups: comp.os.vxworks Subject: Is there a limit on the number of set breakpoints?? Date: 19 Sep 2002 13:38:48 -0700 From: quadmail@yahoo.com (Alex) Organization: http://groups.google.com/ Message-ID: <7b99c091.0209191238.7f22164d@posting.google.com> The WDB_BP_MAX field is modifiable by the user, but is there an actual limit/restriction on how many BPs one can set up on a PPC target? Thanks, Alex McQuade --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Filename from FILE * or descriptor? Date: Thu, 19 Sep 2002 16:58:37 -0400 From: david lindauer Message-ID: <3D8A3A7D.8397923D@notifier-is.net> References: if the headers declare the structure referenced by FILE *, it may have a name in it. I believe reference said elements is non-portable behavior though. David Bill Pringlemeir wrote: > Is there a vxWorks function (v5.4.x) that will get a file name from > either a file descriptor or the ANSI file handle. I have something > like this currently... > > #include > ... > const char *getFileName(int fd) > { > const char * lval = "UNKNOWN"; > > if(ERROR != iosValidFd(fd)) > { > lval = fdTable[fd-3].name; > } > > return lval; > } > > This is rather ugly, but it is functional. You must subtract three > for the stdin, stdout and stderr that are intrinsic and not in > fdTable. It would be really nice if there was a function to do this, > but my search of the documentation turned up nothing. However, I may > have developed a blind spot while reading the manual... > > I need this as I am writing a library/function that take a file handle > and not the file name. The file name is only needed as a debugging > aid. > > Thanks, > Bill Pringlemeir. > > -- > CLICKED BY THOUSANDS OF ATOMIC BOMBS. SINKING. U-63. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Curious Newbie Date: Thu, 19 Sep 2002 17:01:10 -0400 From: david lindauer Message-ID: <3D8A3B16.AF9663BE@notifier-is.net> References: it is possible to get little embedded PCs; usually they are PC compatibles on a small board. VxWorks will pretty much support those as-is (except you may have to obtain a network driver since that isn't part of the PC standard). David Garry Taylor wrote: > Hi, > Please forgive my ignorance about embedded systems :-) > > I'm quite interested in getting a little single board computer, maybe > based on the 1Ghz MIPS chips with two cores from PMC-Sierra, as this > mentions VxWorks as supported. > > A few questions: > > Does anyone have any experience of this chip? A 1Ghz MIPS is bound to > be pretty fast, even for Java, can anyone confirm? > > When creating a user interface on a VxWorks system, is simply a matter > of getting an LCD panel which is supported by the SBC and it'll work? > > Can anyone advise on how much these boards cost? Where to buy in > quantities of one? > > Any pointers to Dummie's guides or FAQs would be great. > > Thanks > > Garry --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Problem establishing communication on a PPP link. Date: Thu, 19 Sep 2002 17:03:38 -0400 From: david lindauer Message-ID: <3D8A3BAA.94DE3FF7@notifier-is.net> References: <14fc63a8.0209121229.54d4ea7a@posting.google.com> go into I think it is configall.h; basic PPP parameters are set up there. You don't necessarily have to use pppOpt... but the error may be indicative that you haven't completely initialized all fields of pppOpt. David David Masse wrote: > Hi, > > I'm trying to establish a PPP link between two VxWorks targets. I > create a serial TTY driver on a HDLC port of a MPC860 CPU for each > target. I want to install PPP over this serial channel to allow the > two targets to communicate. > > I initializes the passive target with no option like this: > > PPP_OPTIONS pppOpt; > pppOpt.flags = OPT_PASSIVE_MODE | OPT_NO_PAP | OPT_NO_CHAP | > OPT_NO_VJ; > pppOpt.lcp_echo_interval = "30"; > pppOpt.lcp_echo_failure = "10"; > > if( pppInit( 0, "/tyCo/1", localAddr, remoteAddr, 0, &pppOpt, NULL ) > == OK ){ > logMsg("pppTest: ppp link established\n",0,0,0,0,0,0); > } else { > logMsg("pppTest: ERROR! Cannot establish ppp link\n",0,0,0,0,0,0); > } > > Almost same thing for the active target: > > PPP_OPTIONS pppOpt; > pppOpt.flags = OPT_NO_PAP | OPT_NO_CHAP | OPT_NO_VJ; > pppOpt.lcp_echo_interval = "30"; > pppOpt.lcp_echo_failure = "10"; > > if( pppInit( 0, "/tyCo/1", localAddr, remoteAddr, 0, &pppOpt, NULL ) > == OK ){ > logMsg("pppTest: ppp link established\n",0,0,0,0,0,0); > } else { > logMsg("pppTest: ERROR! Cannot establish ppp link\n",0,0,0,0,0,0); > } > > The problem is that for both, the pppInit() call return OK but then, I > received the following error: > > [vxKernel] -> ppp0: pppd: vj-max-slots value must be between 2 and 16 > ppp0: vj_max_slots error > 0x241db0 (tPPP0): memPartFree: invalid block 0xce7ec in partition > 0x235ed0 > > Someone know what is the problem? > > Thanks in advance. > > David --------------------------- Newsgroups: comp.os.vxworks Subject: Re: /tyCo/2 does not work Date: Thu, 19 Sep 2002 17:05:02 -0400 From: david lindauer Message-ID: <3D8A3BFE.38092845@notifier-is.net> References: Is the BSP from the manufacturer of the board? If not you may have to tweak it. David Anand Richard wrote: > I am working on a custom vxworks BSP for an IBM 750 board which has a > SouthBridge VIATECH 82c686 chip with two ns16550 style UARTs. The board has > one other UART apart from these which is configured as the default console. > I am trying to use the 16550 uarts in the southbridge as /tyCo/1 and > /tyCo/2. The devices are set up correctly since my raw output routines work > fine. However when I open the /tyCo/1 device and write to it I am unable to > get anything at the ports. The open and write return successfully. "devs" > in the shell gives an output as follows: > > -> devs > drv name > 0 /null > 1 /tyCo/0 > 1 /tyCo/1 > 1 /tyCo/2 > 5 asterix: > 6 /pty/rlogin.S > 7 /pty/rlogin.M > > Thanks. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Booting from Flash Disk Date: Thu, 19 Sep 2002 17:08:53 -0400 From: david lindauer Message-ID: <3D8A3CE5.55903E98@notifier-is.net> References: <1032441796.250754@ernani.logica.co.uk> at a guess, you would have to incorporate the code from the bootrom that does the mode switch into protected mode. David Paul wrote: > Sorry if this is an old question but... > > Is it possible to load a 'full' VxWorks kernel from a Flash Disk > without using the intermediate 'bootrom' loader? > > Am I right in thinking that it should be possible provided I can > get my kernel smaller than 0.5 MB? Has anyone out there > managed to do this? > > My intention is to try to reduce boot up time by skipping this > intermediate stage. If I have to strip my kernel to the bones > to do this, then that's fine. I can always load in some additional > modules (e.g. SIO drivers etc) after boot. > > I'm using a PIII based board, running VxWorks 5.4.2., with a > Flash Disk that looks like a standard ATA Hard Drive. > > Thanks to all. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Select() on changing fdset Date: Thu, 19 Sep 2002 16:19:12 -0700 From: "MC" Organization: alphanet Message-ID: References: <313680C9A886D511A06000204840E1CF67A1A0-100000@whq-msgusr-02.pit.comms.marconi.com> <313680C9A886D511A06000204840E1CF67A1A2-100000@whq-msgusr-02.pit.comms.marconi.com> Reply-To: "MC" Keith I am new to VxWorks and I would like to know a little bit more about this issue I know this is an off-track question. But can you spare some time to explain what the two ugly debugging issues semTake()/exit()/taskSpawn() and logMsg()/exit()/taskSpawn() are ? I mean what problems could they cause ? Thanks Mahesh "Keith Arner" wrote in message news:313680C9A886D511A06000204840E1CF67A1A2-100000@whq-msgusr-02.pit.comms.m arconi.com... > If tasks do exit, then you could be up against potentially ugly debugging > issues like semTake()/exit()/taskSpawn() or logMsg()/exit()/taskSpawn(), > wherein the identity of the offending task is ambiguous. Debugging a > system with a static set of debuggable entities is generally easier than > debugging a system with a dynamically changing set of debuggable entities. > --------------------------- Newsgroups: comp.os.vxworks Subject: RTI NDDS experience Date: 19 Sep 2002 18:30:24 -0700 From: biao@webbasin.com (Bill) Organization: http://groups.google.com/ Message-ID: <73187923.0209191730.7199f47e@posting.google.com> Has anybody used NDDS to build a distributed close system with redundancy and hot pluggable modules? Bill --------------------------- Newsgroups: comp.os.vxworks Subject: task trace during exception Date: Thu, 19 Sep 2002 22:39:30 -0700 From: "s s" Organization: Bell Sympatico Message-ID: Hello, I have a signal handler hooked up to one of my task. This signal handler is for "BUS ERROR". Inside the handler I print the stack of the current task. So the signal handler would get called if the task gets a exception due to a bad memory access. Now when I use "trcStack" to trace the stack , I only see the signal handler. I would like to see the instructions that caused this exception. How do I dig it out of the stack. Does anyone has any suggestion Hadi --------------------------- Newsgroups: comp.os.vxworks Subject: Layer 2 switch Date: 19 Sep 2002 19:58:51 -0700 From: snower_tiger@hotmail.com (snower) Organization: http://groups.google.com/ Message-ID: <3f60ed07.0209191858.25de212a@posting.google.com> Our corporation want to develop Layer 2 switch, and we have developed our hardware board. We want to use MPC8240 cpu and switch chip BCM5615, we just use vxworks, but we dont want to use TMS. We want develop all layer 2 protocols and funtions. If anybody have software technology, or want to cooperate the project, please attach me, my email: snower_tiger@hotmail.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: semGive failed with errno of S_semLib_INVALID_OPERATION. Date: Fri, 20 Sep 2002 12:59:07 +0800 From: "Seraph Ma" Organization: Bentium Ltd. (CN99) Message-ID: References: Thank you very much,Keith. I will try to find out this. Thank you and those fellows who pay attention to this post again. "Keith Arner" wrote in message news:Pine.GSO.4.42.0209190820170.1626-100000@zimbra... > On Thu, 19 Sep 2002, Seraph Ma wrote: > > > Hey all > > Does anyone know when will semGive fail with an errno of > > S_semLib_INVALID_OPERATION. > > The windriver's vxworks muanual say nothing about this. > > Also have I asked technical support from them,but they just asked me to > > refer their manual.God~~ Faint~~~~ > > If anybody ever see this errno when you call semGive,please reply. > > Thank you very much!!! > > You can get S_semLib_INVALID_OPERATION if you try to do a semGive() on a > mutex that you do not own. You can also get S_semLib_INVALID_OPERATION if > you do a semFlush() on a mutex semaphore. > > Keith > > -- > "Once something has gone horribly wrong..." > "...it just gets horribler and horribler." > "Then, it's a matter of finding the original horriblism." > -- from a conversation on debugging problems in a linked list > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: semGive failed with errno of S_semLib_INVALID_OPERATION. Date: Fri, 20 Sep 2002 06:28:12 GMT From: "Gary M" Organization: AT&T Broadband Message-ID: <0ezi9.334686$kp.1062242@rwcrnsc52.ops.asp.att.net> References: I suppose you could also get the same result if the memory location which holds the semaphore ID gets overwritten with another object ID, like a message queue. "Keith Arner" wrote in message news:Pine.GSO.4.42.0209190820170.1626-100000@zimbra... > On Thu, 19 Sep 2002, Seraph Ma wrote: > > > Hey all > > Does anyone know when will semGive fail with an errno of > > S_semLib_INVALID_OPERATION. > > The windriver's vxworks muanual say nothing about this. > > Also have I asked technical support from them,but they just asked me to > > refer their manual.God~~ Faint~~~~ > > If anybody ever see this errno when you call semGive,please reply. > > Thank you very much!!! > > You can get S_semLib_INVALID_OPERATION if you try to do a semGive() on a > mutex that you do not own. You can also get S_semLib_INVALID_OPERATION if > you do a semFlush() on a mutex semaphore. > > Keith > > -- > "Once something has gone horribly wrong..." > "...it just gets horribler and horribler." > "Then, it's a matter of finding the original horriblism." > -- from a conversation on debugging problems in a linked list > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Newbie Question (POSIX queues) Date: Fri, 20 Sep 2002 08:44:19 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: Reply-To: "Michael Lawnick" Sender: mlawnick@p5088ea2d.dip.t-dialin.net Hi Yan, posting some of your code snippets might help. Regards, Michael "Yan Meunier" schrieb im Newsbeitrag news:RIki9.5650$0i4.751476@ursa-nb00s0.nbnet.nb.ca... > Hi guys, I now moved one to playing with POSIX queues. I was able to send > characters (which is a #define) from one task to another without any > problems. Now I'm trying to send a pointer to a class I wrote onto the > queue. The problem comes when I try to receive it. It return me an ERROR > with a EMSGSIZE ERRNO. > > I am able to the same type of message queueing with the use of Wind Message > Queues. > > Is there something I should know about the POSIX message Queue, that's > diffrent then the Wind Message Queue as far as sending and receiving goes? > > If this Question should be asked in another group, could anyone point me to > one? > Thank you, > Yan Meunier > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ADS8260 BSP Date: Fri, 20 Sep 2002 08:57:04 +0200 From: "Achim Zimmer" Message-ID: References: Yes, thats is right. This way works to figure out where the tables and buffers are. There more complicated problem is to move it, because WRS uses a very tricky init string at ADS8260. I have posted this initfunction from WRS, because my english isn't so good to explaine it. I changed this motFccEndParamTemplate to "0x%x:0x%x:0x2100:0x300:-1:-1:-1:-1:0x%x:0x%x:0x%x:0x%x:0x%x:0x%x"; but the result was that the MUX failed to load the device. The problem is, that I dont understand exactly the usage of the template fields. Thanks Achim /*************************************************************************** *** * * sysMotFccEndLoad - load an istance of the motFccEnd driver * * This routine loads the motFccEnd driver with proper parameters. It also * reads the BCSR3 to find out which type of processor is being used, and * sets up the load string accordingly. * * The END device load string formed by this routine is in the following * format. * :::::: * ::::::: * * * .IP * Internal memory address * .IP * FCC number being used * .IP * buffer descriptors base address * .IP * buffer descriptors space size * .IP * data buffers base address * .IP * data buffers space size * .IP * tx buffer in internal memory * .IP * rx buffer in internal memory * .IP * number of TBDs or NONE * .IP * number of RBDs or NONE * .IP * address of a MII-compliant PHY device * .IP * default operating mode of a MII-compliant PHY device * .IP * auto-negotiation order table for a MII-compliant PHY device or NONE * * This routine only loads and initializes one instance of the device. * If the user wishes to use more than one motFccEnd devices, this routine * should be changed. * * RETURNS: pointer to END object or ERROR. * * SEE ALSO: motFccEndLoad () */ END_OBJ * sysMotFccEndLoad ( char * pParamStr, /* ptr to initialization parameter string */ void * unused /* unused optional argument */ ) { /* * The motFccEnd driver END_LOAD_STRING should be: * :::::: * ::::::: * * Note that unit string is prepended by the mux, so we * don't put it here. */ char * pStr = NULL; char paramStr [300]; LOCAL char motFccEndParamTemplate [] = "0x%x:0x%x:-1:-1:-1:-1:-1:-1:0x%x:0x%x:0x%x:0x%x:0x%x:0x%x"; END_OBJ * pEnd; if (strlen (pParamStr) == 0) { /* * muxDevLoad() calls us twice. If the string is * zero length, then this is the first time through * this routine. */ pEnd = (END_OBJ *) motFccEndLoad (pParamStr); } else { /* * On the second pass through here, we actually create * the initialization parameter string on the fly. * Note that we will be handed our unit number on the * second pass and we need to preserve that information. * So we use the unit number handed from the input string. */ pStr = strcpy (paramStr, pParamStr); /* Now, we get to the end of the string */ pStr += strlen (paramStr); /* finish off the initialization parameter string */ sprintf (pStr, motFccEndParamTemplate, (UINT) vxImmrGet (), MOT_FCC_NUM, MOT_FCC_TBD_NUM, MOT_FCC_RBD_NUM, MOT_FCC_PHY_ADDR, MOT_FCC_DEF_PHY_MODE, &motFccAnOrderTbl, MOT_FCC_FLAGS ); if ((pEnd = (END_OBJ *) motFccEndLoad (paramStr)) == (END_OBJ *)ERROR) { logMsg ("Error: motFccEndLoad failed to load driver\n", 0, 0, 0, 0, 0, 0); } } return (pEnd); } #endif /* INCLUDE_MOT_FCC */ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Booting from Flash Disk Date: Fri, 20 Sep 2002 08:50:28 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <1032441796.250754@ernani.logica.co.uk> Reply-To: "Michael Lawnick" Sender: mlawnick@p5088ea2d.dip.t-dialin.net Hi Paul, vxWorks_rom or vxWorks_romCompress are the images you are looking for. You problem might be, that it wants to load to lower 512kB. So you'll have to find some way to get around ... HTH Michael "Paul" schrieb im Newsbeitrag news:1032441796.250754@ernani.logica.co.uk... > Sorry if this is an old question but... > > Is it possible to load a 'full' VxWorks kernel from a Flash Disk > without using the intermediate 'bootrom' loader? > > Am I right in thinking that it should be possible provided I can > get my kernel smaller than 0.5 MB? Has anyone out there > managed to do this? > > My intention is to try to reduce boot up time by skipping this > intermediate stage. If I have to strip my kernel to the bones > to do this, then that's fine. I can always load in some additional > modules (e.g. SIO drivers etc) after boot. > > I'm using a PIII based board, running VxWorks 5.4.2., with a > Flash Disk that looks like a standard ATA Hard Drive. > > Thanks to all. > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: semGive failed with errno of S_semLib_INVALID_OPERATION. Date: Fri, 20 Sep 2002 15:09:09 +0800 From: "Seraph Ma" Organization: Bentium Ltd. (CN99) Message-ID: References: <0ezi9.334686$kp.1062242@rwcrnsc52.ops.asp.att.net> Gary I have check the assemble code of WINDRIVER's API semGive(). What I got out is that the errno of S_semLib_INVALID_OPERATION can only be rose up by semGive(),never other APIs. You can 'l semGive'to see what it is. :-) But still thank you very much! "Gary M" wrote in message news:0ezi9.334686$kp.1062242@rwcrnsc52.ops.asp.att.net... > I suppose you could also get the same result if the memory location which > holds the semaphore ID gets overwritten with another object ID, like a > message queue. > > "Keith Arner" wrote in message > news:Pine.GSO.4.42.0209190820170.1626-100000@zimbra... > > On Thu, 19 Sep 2002, Seraph Ma wrote: > > > > > Hey all > > > Does anyone know when will semGive fail with an errno of > > > S_semLib_INVALID_OPERATION. > > > The windriver's vxworks muanual say nothing about this. > > > Also have I asked technical support from them,but they just asked me > to > > > refer their manual.God~~ Faint~~~~ > > > If anybody ever see this errno when you call semGive,please reply. > > > Thank you very much!!! > > > > You can get S_semLib_INVALID_OPERATION if you try to do a semGive() on a > > mutex that you do not own. You can also get S_semLib_INVALID_OPERATION if > > you do a semFlush() on a mutex semaphore. > > > > Keith > > > > -- > > "Once something has gone horribly wrong..." > > "...it just gets horribler and horribler." > > "Then, it's a matter of finding the original horriblism." > > -- from a conversation on debugging problems in a linked list > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Strang Stack Items, Need Help!! Date: Fri, 20 Sep 2002 15:21:33 +0800 From: "Seraph Ma" Organization: Bentium Ltd. (CN99) Message-ID: References: Michael WACOS_shmAllocMem+241 is a function of xlstGet() implement by our team. It just got a node from a double-list. It puzzles me why this symbol have been reallocated to the value of 0!! When the task corrupts,the symbol of 'xlstGet()' is still OK in the text segment. My vxworks version 5.4.2. Tornado 2.0.2. Host PC Pentium ,Traget CPU is Pentium too. SSI_SHM_HANDLE WACOS_shmAllocMem(unsigned int size) { unsigned short indexOfType; unsigned short indexOfNode; unsigned short i; unsigned int basesize; unsigned int offset; SHM_POOL *shmid; if((size>MAX_BLK_SIZE)||(size==0)) { gTRACE("Request block size error !\n"); return(0); } /*decide what's type of block of requested*/ basesize = MIN_BLK_SIZE; indexOfType = BLK_TYPE_NUM; for(i=0; iaccess) ) /* semaphore taking here!!!!!!!!!*/ { gTRACE("Failed to lock the share memory LOCK!\n"); return(0); } if(0 == xlstCount(&(shmid->freeList[indexOfType]))) { gTRACE("There is no free block for alloc!\n"); shmid->usedUpTimes[indexOfType]+=1; if( SSI_ERROR == UNLOCK(shmid->access) ) { gTRACE("Failed to UNlock the share memory LOCK!\n"); } return(0); } indexOfNode = xlstGet(&(shmid->freeList[indexOfType])); /*<-- just corrupts here!!!! */ /* xlstGet returned value check ,Added by Seraph Ma 2002/05/13 */ if( SSI_XLST_NONODE == indexOfNode ) { gTRACE( "WACOS_shmAllocMem:xlstGet got invalid node value\n" ); return 0; } /*count the offset address of the node*/ offset = shmid->freeList[indexOfType].param.memOffset; offset += indexOfNode * (shmid->freeList[indexOfType].param.bytesPerNode); offset += sizeof(XNODE); /*mod stat info in the freeList*/ shmid->blkAllocated[indexOfType] += 1; shmid->bytesAllocated += (basesize + sizeof(XNODE)); /* if the Max allocated number old recored value is smaller than * current allocated blocks number,renew it */ shmid->maxAllocatedNum[indexOfType] = ( shmid->blkAllocated[indexOfType] > shmid->maxAllocatedNum[indexOfType])? shmid->blkAllocated[indexOfType]:shmid->maxAllocatedNum[indexOfType]; if( SSI_ERROR == semGive(shmid->access) ) /*shmid->access is a created counting semaphore*/ { gTRACE("Failed to UNlock the share memory LOCK!\n"); return(0); } return(offset); } - -> l WACOS_shmAllocMem WACOS_shmAllocMem: 03f3adfc 55 PUSH EBP 03f3adfd 89 e5 MOV EBP, ESP 03f3adff 83 ec 1c SUB ESP, 28 03f3ae02 57 PUSH EDI 03f3ae03 56 PUSH ESI 03f3ae04 53 PUSH EBX 03f3ae05 c7 45 ec 00 00 00 00 MOV [EBP-20], 0x0 03f3ae0c 81 7d 08 00 10 00 00 CMP [EBP+8], 0x1000 03f3ae13 77 17 JNBE 0x03f3ae2c 03f3ae15 83 7d 08 00 CMP [EBP+8], 0 value = 66301465 = 0x3f3ae19 = WACOS_shmAllocMem + 0x1d - -> l WACOS_shmAllocMem+0x241 03f3b03d e8 96 ce ff ff CALL xlstGet 03f3b042 83 c4 04 ADD ESP, 4 03f3b045 89 c0 MOV EAX, EAX 03f3b047 66 OSIZE 03f3b048 89 45 fc MOV [EBP-4], AX 03f3b04b 68 a9 ad f3 03 PUSH 0x3f3ada9 03f3b050 e8 33 02 1e fc CALL printf 03f3b055 83 c4 04 ADD ESP, 4 03f3b058 66 OSIZE 03f3b059 81 7d fc ff ff CMP [EBP-4], 0xffff value = 66302046 = 0x3f3b05e = WACOS_shmAllocMem + 0x262 Faint~~ So strange it seems. Thank you ,Michael,you are a so warm-hearted guy. I will tell you the final result of this problem as soon as I get it out. Thanx again. "Michael Lawnick" wrote in message news:amcgor$o89$1@snoopy.bndlg.de... > Hi Seraph, > > this all shows me, that you *have* a jump to 0. As you ask for > S_semLib_INVALID_OPERATION: > A corrupted semId could cause this problem too. > You will have to find out which line in source corresponds to > WACOS_shmAllocMem+241. > Show us the code of WACOS_shmAllocMem() and tell us Tornado version and H/W > type, perhaps we can help. > > Regards, > Michael > > "Seraph Ma" schrieb im Newsbeitrag > news:amc8c1$ud0$1@debian.bentium.com... > > Michael > > Thank you again for your warm-hearted help. > > These code runs in our company's gateway--which disable me from > runtime > > debugging in Tornado IDE.It's really troublesome to debug these codes. > > OK, what I have got from the PGF is like this: > > > > Page Fault > > Program Counter: 0x00000000 > > Status Register: 0x00000003 > > Error Code: 0x0f26e429 > > Task: 0xd506ba8 "tNCMInp" > > Task not found. > > > > -> > > -> ISR,Error,FFFD300C,F26E78E,THU SEP 19 02:47:38 2002,CPU:Page Fault > > > > > > The task fails like this: > > -> taskShow tNCMInp,1 > > > > > > NAME ENTRY TID PRI STATUS PC SP ERRNO > > DELAY > > > > ---------- ------------ -------- --- ---------- -------- -------- ------- > > ----- > > > tNCMInp _miNCMInputT d7fb654 80 SUSPEND 12bd6e d7fb4f8 > 160068 > > 0 > > > > > > stack: base 0xd7fb654 end 0xd7f1654 size 40952 high 1740 margin > 39212 > > > > > > options: 0x4 > > > VX_DEALLOC_STACK > > > > > > edi = d7fb540 esi = d7fb654 ebp = d7fb500 esp = > > > d7fb4f8 > > > ebx = d7fb654 edx = fa0 ecx = f9f eax = > > > 0 > > > eflags = 246 pc = 12bd6e > > > -> printErrno 0x160068 > > > S_semLib_INVALID_OPERATION > > > value = 27 = 0x1b > > > > > > > > The PC counter is 0x00000000 as you have seen. BTW, have you ever got > an > > errno of S_semLib_INVALID_OPERATION when you call semGive() ? Do you know > > when does this errno come > > up ? > > Thank you very very much for you,for all who pay attention to this > post! > > > > > > Best Regard > > Seraph Ma > > > > > > > > > > > > "Michael Lawnick" wrote in message > > news:amc4gm$vsf$1@snoopy.bndlg.de... > > > Hi Seraph, > > > > WACOS_shmAllocMem() never > > > > calls any uninitialized or unresolved symbol. > > > > > > this does not look like a stack corruption, as excTask is able to back > > trace > > > your calls. The back trace shows the call of an unknown function at > > address > > > 0 in function WACOS_shmAllocMem(): > > > > f2e1f25 WACOS_shmAllocMem+241: 0 (c94203c) > > > As normaly there is never a function, the call is corrupt. The processor > > > throws an exception at address 0 as there is no valid instruction or the > > MMU > > > has marked this area as non-executable: > > > > 0 : 1bb5d4 ([f2def2d, 8, 10202, ffff, 0]) > > > > 1bb5fd excStub +35 : excExcHandle (e, d4fa0c0, d4fa098) > > > > > > So put a breakpoint at WACOS_shmAllocMem(), select 'Mixed Source and > > > Disassembly' in 'View' menu and step through. > > > > > > Regards, > > > Michael > > > > > > "Seraph Ma" schrieb im Newsbeitrag > > > news:ambq67$q75$1@debian.bentium.com... > > > > hey all > > > > When I trace my application task's stack, I got this: > > > > > > > > -> tt tNCMInp > > > > 13bfd2 vxTaskEntry +a : f2eca43 (0, 0, 0, 0, 0, 0, 0, 0, 0, 0) > > > > f2eca6d miNCMInputTask +441: WACOS_shmAllocMem (8) > > > > f2e1f25 WACOS_shmAllocMem+241: 0 (c94203c) > > > > 0 : 1bb5d4 ([f2def2d, 8, 10202, ffff, 0]) > > > > 1bb5fd excStub +35 : excExcHandle (e, d4fa0c0, d4fa098) > > > > 1643df excExcHandle +143: taskSuspend ([0, d4fa0c0, 0, 0, e0003]) > > > > value = 0 = 0x0 > > > > > > > > > > > > As what you have seen, there is a line > > > > > > > > "f2e1f25 WACOS_shmAllocMem+241: 0 (c94203c)" > > > > > > > > I just wonder what is the meaning of the function of '0'. > > > > I have seen this for many times. Do this mean my application > task's > > > > stack > > > > had already been damanged by some tasks else? Or it's OK. If that,how > > can > > > we > > > > explain the meaning of the '0'. > > > > WACOS_shmAllocMem() never > > > > calls any uninitialized or unresolved symbol. > > > > > > > > Thank you all in advance.Any help informaition is appreciative! > > > > > > > > > > > > > > > > > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: task trace during exception Date: Fri, 20 Sep 2002 07:28:13 +0000 (UTC) From: Johan Borkhuis Organization: Proxim Corp. Message-ID: References: "s s" wrote: > Hello, > I have a signal handler hooked up to one of my task. This signal > handler is for "BUS ERROR". Inside the handler I print the stack of > the current task. > > So the signal handler would get called if the task gets a exception > due to a bad memory access. Now when I use "trcStack" to trace the > stack , I only see the signal handler. I would like to see the > instructions that caused this exception. How do I dig it out of the > stack. Does anyone has any suggestion What you have to do is to find out which task caused the exception. As you did not give your architecture it is not possible to indicate how. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: modf() function problem Date: 20 Sep 2002 00:51:43 -0700 From: vividh@hotmail.com (rasayani) Organization: http://groups.google.com/ Message-ID: <28809481.0209192351.7ebffcd0@posting.google.com> Has anyone had a problem with the modf() function under Vxworks? I have include INCLUDE_ANSI_MATH in config.h. But for eg: double i = 10; double j = 10; double k = modf (i ,&j); should give k = 0 but gives a strange large number. vividh --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ADS8260 BSP Date: Fri, 20 Sep 2002 01:18:04 -0700 From: "Dan Gold" Organization: Mailgate.ORG Server - http://www.Mailgate.ORG Message-ID: References: Sender: gold@ensemble.com Achim Zimmer wrote in message ... >Yes, thats is right. This way works to figure out where the tables and >buffers are. Did you examine it ? where are the buffer descriptor tables, in DPRAM or external memory ? If they are in external memory there is nothing to move. If they are in DPRAM then you need to move them. Your main problem is you need the DPRAM starting at the base of bank 1. First, are you sure something is using that memory ? If so, how do you know ? What happens if you zero out that memory while the system is running ? Or can you dump it and see if values are changing ? Other candidates that may use that space are the buffer descriptors for the serial port (i.e. the SCC, since they cannot reside in external memory). You can look at the serial driver where it initializes it's BDs and see where it is putting them. It sounds like you need to do 1 and/or 2 things: 1. Find some documentation on the BSP that describes where everything is placed in DPRAM. This may or may not exist. And you may have to look in a header file for it. 2. Look through all the code in the BSP for the various peripherals you are using (FCC, SCC, anything else like I2c or SPI, etc...) and see where they place their stuff in DPRAM. >There more complicated problem is to move it, because WRS uses a very tricky >init string at ADS8260. Again, you don't have to move it unless you find out the BDs are in DPRAM. Make sure of that first before dealing with the init string. >I have posted this initfunction from WRS, because my english isn't so good >to explaine it. > >I changed this motFccEndParamTemplate to > "0x%x:0x%x:0x2100:0x300:-1:-1:-1:-1:0x%x:0x%x:0x%x:0x%x:0x%x:0x%x"; >but the result was that the MUX failed to load the device. what did you change it from ? >The problem is, that I dont understand exactly the usage of the template >fields. The END device load string formed by this routine is in the following format. >* :::::: >* ::::::: >* >* >* .IP >* buffer descriptors base address this is the parameter you want to change. If it's 0, it probably defaults to something inside the driver, but I don't know what. Maybe call WRS and ask for motFcc.c if you need to. >* .IP >* tx buffer in internal memory >* .IP >* rx buffer in internal memory If the above 2 fields are in an area you need, you'll need to move them too. The bottom line is find out if anything is interfering with the DPRAM you need, and what it is first. hth, Dan Gold gold@ensemble.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Booting from Flash Disk Date: Fri, 20 Sep 2002 10:35:12 +0100 From: "Paul" Organization: Logica UK Ltd Message-ID: <1032514513.184293@ernani.logica.co.uk> References: <1032441796.250754@ernani.logica.co.uk> Thanks, Yes, those are the images I've been looking at. I guess the question is, has anyone managed to get a small enough kernel for this to work? Also, if I use the compressed version, does the 512kB limit apply to the compressed image, or the image after it is uncompressed? Cheers. "Michael Lawnick" wrote in message news:amegfl$hbv$1@snoopy.bndlg.de... > Hi Paul, > > vxWorks_rom or vxWorks_romCompress are the images you are looking for. > You problem might be, that it wants to load to lower 512kB. So you'll have > to find some way to get around ... > > HTH > Michael > > "Paul" schrieb im Newsbeitrag > news:1032441796.250754@ernani.logica.co.uk... > > Sorry if this is an old question but... > > > > Is it possible to load a 'full' VxWorks kernel from a Flash Disk > > without using the intermediate 'bootrom' loader? > > > > Am I right in thinking that it should be possible provided I can > > get my kernel smaller than 0.5 MB? Has anyone out there > > managed to do this? > > > > My intention is to try to reduce boot up time by skipping this > > intermediate stage. If I have to strip my kernel to the bones > > to do this, then that's fine. I can always load in some additional > > modules (e.g. SIO drivers etc) after boot. > > > > I'm using a PIII based board, running VxWorks 5.4.2., with a > > Flash Disk that looks like a standard ATA Hard Drive. > > > > Thanks to all. > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: protocol stack problem Date: Fri, 20 Sep 2002 10:12:21 +0000 (UTC) From: He Jin Organization: Bentium Ltd. (CN99) Message-ID: Hi, Johan, Recently we meet a strange problem while porting vxWorks to a MIPS CPU design board. It seems that we made a mistake in dealing with tcp/ip protocol stack. We use an intel 82559 NIC and vxWorks' driver for 82557, 82559 NICs. At the same time, we disabled the DHCP option in the project file. When system startups, it can send a normal ARP packet to the host machine, however the ARP echo packets from the host have been dropped by vxWorks' network protocol handler, that's, use "ifShow fei0" shell command, it will report received serveral unknown protocol packets without IP checksum error. The following is the snapshot: - -> ifShow "fei" fei (unit number 0): Flags: (0x8063) UP BROADCAST MULTICAST ARP RUNNING Type: ETHERNET_CSMACD Internet address: 192.168.222.50 Broadcast address: 192.168.222.255 Netmask 0xffffff00 Subnetmask 0xffffff00 Ethernet address is 00:03:47:77:ba:f6 Metric is 0 Maximum Transfer Unit size is 1500 0 octets received 0 octets sent 83 packets received 6 packets sent 83 unicast packets received 6 unicast packets sent 0 non-unicast packets received 0 non-unicast packets sent 0 input discards 83 input unknown protocols 0 input errors 0 output errors 0 collisions; 0 dropped value = -6 = 0xfffffffa - -> The design board has a MIPS cpu and disabled all cache, so the DMA between target main-memory and NIC should not be wrong. Thanks He Jin --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Quad Uart Driver Date: Fri, 20 Sep 2002 10:29:25 GMT From: "Kelly Hornsby" Organization: Cox Communications Message-ID: <9MCi9.4262$IL6.195730@news2.east.cox.net> References: The SIO driver is what you want to use if you need to use the tyCo interfaces. The control and callback setup is specified in the SIO routines, but the control of the abstracted sio itself is found elsewhere. If you wish to use the serial port for a specialized control seperate from the SIO, then you can use the "serial" driver. I had to create a driver for a dual channel serial device on a removable TechnoBox PMC, and I did this same thing. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: help with vision ICE Date: Fri, 20 Sep 2002 06:34:36 -0400 From: Varndell Engineering Organization: Posted Via Newsfeeds.com = SPEED+RETENTION+COMPLETION = http://www.newsfeeds.com Message-ID: <3D8AF9BC.CA270123@hotmail.com> References: Jim wrote: > > in my bootloader on the mpc860 i disable the busmonitor ... it seems the BDM > (vision ICE) may be turning this back on (but the register sypcr is still > ffffff08 as i set it) ... is it possible the bus will time out on BDM when > accessing a slow device ... The BM is enabled whenever FRZ is asserted or when a debug mode request is pending; independent of SYPCR[BME]. So, yes, the bus can timeout. > The problem in general is this ... i can boot my board without the ice ... > but with the ice i die accessing a slow device 500 khz ... Perhaps a little more information would help. Can you describe what happens when you die? Are you using Click, SingleStep, what? Is your slow device accessed via GPCM or UPM? Can you slow down the system clock to a point where the max BM timeout is greater than the access time of the device? Just to see if this mysterious dying ceases? Regards, George Varndell - -- Varndell Engineering, LLC Embedded Systems Expertise Need an END driver for DP8382x GigE? Contact us. http://www.varndellengineering.com - -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- http://www.newsfeed.com The #1 Newsgroup Service in the World! - -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Booting from Flash Disk Date: Fri, 20 Sep 2002 13:58:59 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <1032441796.250754@ernani.logica.co.uk> <1032514513.184293@ernani.logica.co.uk> Reply-To: "Michael Lawnick" Sender: mlawnick@p5088ea2d.dip.t-dialin.net Hi Paul, I've not tested yet but expect that the limit only applies to the compressed. It will decompress to RAM_LOW_ADDR (which on PC is the higher one !) and should start as expected. Whether your compressed kernel fits is just a question of the features you include. Try and you'll find out. Have just made a test image from a standard I most times use (incl. FTP, targetshell, telnet etc.) Build told: vxWorks_romCompress: 13520(t) + 478016(d) = 491536 dir tells: 20.09.2002 13:55 493.534 vxWorks_romCompress Regards, Michael "Paul" schrieb im Newsbeitrag news:1032514513.184293@ernani.logica.co.uk... > Thanks, > > Yes, those are the images I've been looking at. I guess the > question is, has anyone managed to get a small enough > kernel for this to work? > > Also, if I use the compressed version, does the 512kB limit > apply to the compressed image, or the image after it is > uncompressed? > > Cheers. > > > "Michael Lawnick" wrote in message > news:amegfl$hbv$1@snoopy.bndlg.de... > > Hi Paul, > > > > vxWorks_rom or vxWorks_romCompress are the images you are looking for. > > You problem might be, that it wants to load to lower 512kB. So you'll have > > to find some way to get around ... > > > > HTH > > Michael > > > > "Paul" schrieb im Newsbeitrag > > news:1032441796.250754@ernani.logica.co.uk... > > > Sorry if this is an old question but... > > > > > > Is it possible to load a 'full' VxWorks kernel from a Flash Disk > > > without using the intermediate 'bootrom' loader? > > > > > > Am I right in thinking that it should be possible provided I can > > > get my kernel smaller than 0.5 MB? Has anyone out there > > > managed to do this? > > > > > > My intention is to try to reduce boot up time by skipping this > > > intermediate stage. If I have to strip my kernel to the bones > > > to do this, then that's fine. I can always load in some additional > > > modules (e.g. SIO drivers etc) after boot. > > > > > > I'm using a PIII based board, running VxWorks 5.4.2., with a > > > Flash Disk that looks like a standard ATA Hard Drive. > > > > > > Thanks to all. > > > > > > > > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Strang Stack Items, Need Help!! Date: Fri, 20 Sep 2002 14:18:58 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: Reply-To: "Michael Lawnick" Sender: mlawnick@p5088ea2d.dip.t-dialin.net Hi Seraph, I'm no PC-specialist, PPC is my world, but following idea: Your function xlstGet is called with the agument 0xc94203c. You should check whether this might corrupt your stack if off limit and missing check for parameter boundaries. Your rts instruction could then load a 0 to PC and the tracing of excTask would show what we see. Regards, Michael "Seraph Ma" schrieb im Newsbeitrag news:ameia6$j44$1@debian.bentium.com... > Michael > WACOS_shmAllocMem+241 is a function of xlstGet() implement by our team. > It just got a node from a double-list. > It puzzles me why this symbol have been reallocated to the value of 0!! > When the task corrupts,the symbol of 'xlstGet()' is still OK in the text > segment. > My vxworks version 5.4.2. Tornado 2.0.2. Host PC Pentium ,Traget CPU is > Pentium too. > > SSI_SHM_HANDLE WACOS_shmAllocMem(unsigned int size) > { > unsigned short indexOfType; > unsigned short indexOfNode; > unsigned short i; > unsigned int basesize; > unsigned int offset; > SHM_POOL *shmid; > > if((size>MAX_BLK_SIZE)||(size==0)) > { > gTRACE("Request block size error !\n"); > return(0); > } > > /*decide what's type of block of requested*/ > basesize = MIN_BLK_SIZE; > indexOfType = BLK_TYPE_NUM; > for(i=0; i { > if(size <= basesize) > { > indexOfType = i; > break; > } > basesize *= 2; > } > if(indexOfType == BLK_TYPE_NUM) > { > gTRACE("request block more than max block size!\n"); > return(0); > } > > shmid = (SHM_POOL *)get_shm_addr(); > if(shmid == NULL) > { > gTRACE("Get share memory pool ID failed!\n"); > return(0); > } > > if( SSI_ERROR == semTake(shmid->access) ) /* semaphore taking > here!!!!!!!!!*/ > { > gTRACE("Failed to lock the share memory LOCK!\n"); > return(0); > } > if(0 == xlstCount(&(shmid->freeList[indexOfType]))) > { > gTRACE("There is no free block for alloc!\n"); > shmid->usedUpTimes[indexOfType]+=1; > if( SSI_ERROR == UNLOCK(shmid->access) ) > { > gTRACE("Failed to UNlock the share memory LOCK!\n"); > } > return(0); > } > > indexOfNode = xlstGet(&(shmid->freeList[indexOfType])); /*<-- just > corrupts here!!!! */ > /* xlstGet returned value check ,Added by Seraph Ma 2002/05/13 */ > if( SSI_XLST_NONODE == indexOfNode ) > { > gTRACE( "WACOS_shmAllocMem:xlstGet got invalid node value\n" ); > return 0; > } > > /*count the offset address of the node*/ > offset = shmid->freeList[indexOfType].param.memOffset; > offset += indexOfNode * > (shmid->freeList[indexOfType].param.bytesPerNode); > offset += sizeof(XNODE); > > /*mod stat info in the freeList*/ > shmid->blkAllocated[indexOfType] += 1; > shmid->bytesAllocated += (basesize + sizeof(XNODE)); > /* if the Max allocated number old recored value is smaller than > * current allocated blocks number,renew it */ > shmid->maxAllocatedNum[indexOfType] = > ( shmid->blkAllocated[indexOfType] > > shmid->maxAllocatedNum[indexOfType])? > shmid->blkAllocated[indexOfType]:shmid->maxAllocatedNum[indexOfType]; > > if( SSI_ERROR == semGive(shmid->access) ) /*shmid->access is a created > counting semaphore*/ > { > gTRACE("Failed to UNlock the share memory LOCK!\n"); > return(0); > } > > return(offset); > } > > -> l WACOS_shmAllocMem > WACOS_shmAllocMem: > 03f3adfc 55 PUSH EBP > 03f3adfd 89 e5 MOV EBP, ESP > 03f3adff 83 ec 1c SUB ESP, 28 > 03f3ae02 57 PUSH EDI > 03f3ae03 56 PUSH ESI > 03f3ae04 53 PUSH EBX > 03f3ae05 c7 45 ec 00 00 00 00 MOV [EBP-20], 0x0 > 03f3ae0c 81 7d 08 00 10 00 00 CMP [EBP+8], 0x1000 > 03f3ae13 77 17 JNBE 0x03f3ae2c > 03f3ae15 83 7d 08 00 CMP [EBP+8], 0 > value = 66301465 = 0x3f3ae19 = WACOS_shmAllocMem + 0x1d > -> l WACOS_shmAllocMem+0x241 > 03f3b03d e8 96 ce ff ff CALL xlstGet > 03f3b042 83 c4 04 ADD ESP, 4 > 03f3b045 89 c0 MOV EAX, EAX > 03f3b047 66 OSIZE > 03f3b048 89 45 fc MOV [EBP-4], AX > 03f3b04b 68 a9 ad f3 03 PUSH 0x3f3ada9 > 03f3b050 e8 33 02 1e fc CALL printf > 03f3b055 83 c4 04 ADD ESP, 4 > 03f3b058 66 OSIZE > 03f3b059 81 7d fc ff ff CMP [EBP-4], 0xffff > value = 66302046 = 0x3f3b05e = WACOS_shmAllocMem + 0x262 > > > > > Faint~~ So strange it seems. > Thank you ,Michael,you are a so warm-hearted guy. I will tell you the > final result of this problem as soon as I get it out. > Thanx again. > > > "Michael Lawnick" wrote in message > news:amcgor$o89$1@snoopy.bndlg.de... > > Hi Seraph, > > > > this all shows me, that you *have* a jump to 0. As you ask for > > S_semLib_INVALID_OPERATION: > > A corrupted semId could cause this problem too. > > You will have to find out which line in source corresponds to > > WACOS_shmAllocMem+241. > > Show us the code of WACOS_shmAllocMem() and tell us Tornado version and > H/W > > type, perhaps we can help. > > > > Regards, > > Michael > > > > "Seraph Ma" schrieb im Newsbeitrag > > news:amc8c1$ud0$1@debian.bentium.com... > > > Michael > > > Thank you again for your warm-hearted help. > > > These code runs in our company's gateway--which disable me from > > runtime > > > debugging in Tornado IDE.It's really troublesome to debug these codes. > > > OK, what I have got from the PGF is like this: > > > > > > Page Fault > > > Program Counter: 0x00000000 > > > Status Register: 0x00000003 > > > Error Code: 0x0f26e429 > > > Task: 0xd506ba8 "tNCMInp" > > > Task not found. > > > > > > -> > > > -> ISR,Error,FFFD300C,F26E78E,THU SEP 19 02:47:38 2002,CPU:Page Fault > > > > > > > > > The task fails like this: > > > -> taskShow tNCMInp,1 > > > > > > > > NAME ENTRY TID PRI STATUS PC SP > ERRNO > > > DELAY > > > > > > > ---------- ------------ -------- --- ---------- -------- -------- ------- > > > ----- > > > > tNCMInp _miNCMInputT d7fb654 80 SUSPEND 12bd6e d7fb4f8 > > 160068 > > > 0 > > > > > > > > stack: base 0xd7fb654 end 0xd7f1654 size 40952 high 1740 margin > > 39212 > > > > > > > > options: 0x4 > > > > VX_DEALLOC_STACK > > > > > > > > edi = d7fb540 esi = d7fb654 ebp = d7fb500 esp = > > > > d7fb4f8 > > > > ebx = d7fb654 edx = fa0 ecx = f9f eax = > > > > 0 > > > > eflags = 246 pc = 12bd6e > > > > -> printErrno 0x160068 > > > > S_semLib_INVALID_OPERATION > > > > value = 27 = 0x1b > > > > > > > > > > > > The PC counter is 0x00000000 as you have seen. BTW, have you ever > got > > an > > > errno of S_semLib_INVALID_OPERATION when you call semGive() ? Do you > know > > > when does this errno come > > > up ? > > > Thank you very very much for you,for all who pay attention to this > > post! > > > > > > > > > Best Regard > > > Seraph > Ma > > > > > > > > > > > > > > > > > > "Michael Lawnick" wrote in message > > > news:amc4gm$vsf$1@snoopy.bndlg.de... > > > > Hi Seraph, > > > > > WACOS_shmAllocMem() never > > > > > calls any uninitialized or unresolved symbol. > > > > > > > > this does not look like a stack corruption, as excTask is able to back > > > trace > > > > your calls. The back trace shows the call of an unknown function at > > > address > > > > 0 in function WACOS_shmAllocMem(): > > > > > f2e1f25 WACOS_shmAllocMem+241: 0 (c94203c) > > > > As normaly there is never a function, the call is corrupt. The > processor > > > > throws an exception at address 0 as there is no valid instruction or > the > > > MMU > > > > has marked this area as non-executable: > > > > > 0 : 1bb5d4 ([f2def2d, 8, 10202, ffff, 0]) > > > > > 1bb5fd excStub +35 : excExcHandle (e, d4fa0c0, d4fa098) > > > > > > > > So put a breakpoint at WACOS_shmAllocMem(), select 'Mixed Source and > > > > Disassembly' in 'View' menu and step through. > > > > > > > > Regards, > > > > Michael > > > > > > > > "Seraph Ma" schrieb im Newsbeitrag > > > > news:ambq67$q75$1@debian.bentium.com... > > > > > hey all > > > > > When I trace my application task's stack, I got this: > > > > > > > > > > -> tt tNCMInp > > > > > 13bfd2 vxTaskEntry +a : f2eca43 (0, 0, 0, 0, 0, 0, 0, 0, 0, 0) > > > > > f2eca6d miNCMInputTask +441: WACOS_shmAllocMem (8) > > > > > f2e1f25 WACOS_shmAllocMem+241: 0 (c94203c) > > > > > 0 : 1bb5d4 ([f2def2d, 8, 10202, ffff, 0]) > > > > > 1bb5fd excStub +35 : excExcHandle (e, d4fa0c0, d4fa098) > > > > > 1643df excExcHandle +143: taskSuspend ([0, d4fa0c0, 0, 0, e0003]) > > > > > value = 0 = 0x0 > > > > > > > > > > > > > > > As what you have seen, there is a line > > > > > > > > > > "f2e1f25 WACOS_shmAllocMem+241: 0 (c94203c)" > > > > > > > > > > I just wonder what is the meaning of the function of '0'. > > > > > I have seen this for many times. Do this mean my application > > task's > > > > > stack > > > > > had already been damanged by some tasks else? Or it's OK. If > that,how > > > can > > > > we > > > > > explain the meaning of the '0'. > > > > > WACOS_shmAllocMem() never > > > > > calls any uninitialized or unresolved symbol. > > > > > > > > > > Thank you all in advance.Any help informaition is appreciative! > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: semTake()/exit()/taskSpawn() Date: Fri, 20 Sep 2002 08:37:33 -0400 From: Keith Arner Organization: Marconi Message-ID: References: <313680C9A886D511A06000204840E1CF67A1A0-100000@whq-msgusr-02.pit.comms.marconi.com> <313680C9A886D511A06000204840E1CF67A1A2-100000@whq-msgusr-02.pit.comms.marconi.com> Reply-To: Keith Arner On Thu, 19 Sep 2002, MC wrote: > Keith > I am new to VxWorks and I would like to know a little bit more about this > issue > I know this is an off-track question. > But can you spare some time to explain what the two ugly debugging issues > semTake()/exit()/taskSpawn() and logMsg()/exit()/taskSpawn() are ? > I mean what problems could they cause ? > > Thanks > Mahesh > > "Keith Arner" wrote in message > news:313680C9A886D511A06000204840E1CF67A1A2-100000@whq-msgusr-02.pit.comms.m > arconi.com... > > If tasks do exit, then you could be up against potentially ugly debugging > > issues like semTake()/exit()/taskSpawn() or logMsg()/exit()/taskSpawn(), > > wherein the identity of the offending task is ambiguous. Debugging a > > system with a static set of debuggable entities is generally easier than > > debugging a system with a dynamically changing set of debuggable entities. In VxWorks, tasks are identified by a task id. When one task exits, its task id becomes available for reuse. In VxWorks 5.4, if you spawn a task immediately after some task has exited, the new task is fairly likely to get the task id of the task that just exited. The problem arises when something gets a task id, then attempts to use it after the corresponding task has exited: at this point did you mean to talk about the old task, or the new task which accidentally got the same task id? At any given time, a mutex semaphore is "owned" by at most one task. What happens if a task takes the semaphore, forgets to give it up, then exits? The next task to come along and accidentally get the same task id will also accidentally own the mutex semaphore. The root problem in your code would be the task that exited while owning the semaphore, but the only symptom you see is a completely unrelated task owning the semaphore. When presented with that symptom, how does one divine what the root problem was? Another such problem arises with logMsg(). When a task logs a message, the message does not get printed out immediately, but rather it gets queued up, to be printed out by tLogTask. This usually get printed to a slow serial device, so there could be a noticably long time between when the message gets queued and when it gets printed. When the message is printed, it includes the name of the task that logged the message, but the name of the task is found by following a task id that was stored when the message was queued. If a task queues a message, then exits and another task is spawned before the message gets printed out, the message is printed out with the task name of the new task, rather than the old task. When debugging the problem indicated by your error message, you are debugging based on misleading evidence. When people are designing a new piece of functionality, and are considering spawning off some new tasks, I advise them to avoid creating and deleting tasks at run time. If you create all your tasks at initialization time and they never exit, you are debugging a much more predictable system. This is not always possible, but is a good idea, to the extent that it can be done. Keith - -- "Once something has gone horribly wrong..." "...it just gets horribler and horribler." "Then, it's a matter of finding the original horriblism." -- from a conversation on debugging problems in a linked list --------------------------- Newsgroups: comp.os.vxworks Subject: Re: disk cache error Date: Fri, 20 Sep 2002 13:15:17 GMT From: "Brian St. Pierre" Organization: Worldpath Internet Services (worldpath.net) Message-ID: References: <3D78BE3C.2DBC1BAD@notroll.com> I asked a related question in a later thread ("obsolete file descriptor"). Leonid, thanks for the tip on what obsoletes an fd. I have taken out dcache, calling dosFsDevInit with the device returned by tffsDevCreate. However, it appears that dosFsDevInit is creating a dcache automatically. When I do a dosFsShow, it prints a value for pCbio, and cbioShow says that there is a 512 byte cache. One problem that I see with this arrangement is that I haven't set the sync interval on this dcache to 0. Would it be better to create a dcache with size of zero? - -> dosFsShow "/flash" volume descriptor ptr (pVolDesc): 0xf3c3ab8 cache block I/O descriptor ptr (pCbio): 0xf3c3b78 auto disk check on mount: DOS_CHK_REPAIR | DOS_CHK_VERB_2 [ ... ] - -> cbioShow 0xf3c3b78 Cached Block I/O Device, handle=0xf3c3b78 Description: Block Device Disk size 29 Mbytes, RAM Size 512 bytes ^^^^^^^^^^^^^^^^^^ Block size 512, heads 15, blocks/track 4, # of blocks 61196 partition offset 0 blocks, type Fixed, Media changed No The code that creates this: STATUS usrTffsConfig(int drive, int removable, char* fileName) { char devName [BOOT_FILE_LEN]; BLK_DEV * pBlkDev; void* cbio; STATUS status; if ((UINT)drive >= noOfDrives) { printErr ("drive is out of range (0-%d).\n", noOfDrives - 1); return (ERROR); } /* create a block device spanning entire disk (non-destructive!) */ pBlkDev = tffsDevCreate (drive, removable); if (pBlkDev == NULL) { printErr ("usrTffsConfig: tffsDevCreate failed.\n"); return (ERROR); } /* Extract device name from file. */ devSplit (fileName, devName); #undef USE_DCACHE #if USE_DCACHE cbio = dcacheDevCreate((void*)pBlkDev, NULL, 1024, devName); if(cbio == NULL) { printErr("usrTffsConfig: dcacheDevCreate failed.\n"); return ERROR; } /* Set the syncInterval to 0 (last parameter). */ status = dcacheDevTune(cbio, 0, 0, 0, 0); if(status != OK) { printErr("usrTffsConfig: dcacheDevTune failed.\n"); return ERROR; } #else cbio = pBlkDev; #endif /* Initialize the boot block device as a dosFs device named * . Set up for 16 simultaneously open files, with * automatic CHKDSK and no disk cache. */ status = dosFsDevCreate (devName, cbio, TFFS_MAX_OPEN_FILES, 0); if (status != OK) { printErr ("usrTffsConfig: dosFsDevCreate failed.\n"); return (ERROR); } return (OK); } - -- - -Brian St. Pierre brian (at) bstpierre.org "Leonid Rosenboim" wrote: > If you need the sync interval set to 0, then maybe your better off > without dcache - you can call dosFsDevCreate with a Blk device id. > > Also, if dcache detects that block 0 has been unexpectedly changed, > and you have not changed the device, it is a goo dreason to worry. > In other words, if the driver (TFFS?) changes block 0 without an > apparent reason, there could be other problems that could result > in data corruption. > Remember that dcache only checks on block 0 as an additional means > for detecting card removal, and you can eliminate dcache, which > will not eliminate the actual problem though. > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Newbie Question (POSIX queues) Date: Fri, 20 Sep 2002 14:18:09 GMT From: "Yan Meunier" Organization: Aliant Internet Message-ID: References: for example class base{ public: virtual char* getMessage()const = 0; private: char* messagePtr; }; class A:public base{ public: A(); char* getMessage()const{return messagePtr}; private: ... }; A::A() { messagePtr = new char("Hello World"); } - ---------------------------------------------- //DEFINES mqd_t QID; char QName[] = "TestQ"; //FORWARD DECLARATIONS void receiveTask(void); void sendTask(void); //========================================================================== ========= int entryPoint(void) { if (RcvId = taskSpawn ("tRcvTask", 100, 0, 4000, (FUNCPTR)receiveTask, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0) == ERROR) { printf ("taskSpawn of tRcvTask failed\n"); return (ERROR); } else printf ("taskSpawn of tRcvTask created\n"); if (SndId = taskSpawn ("tSndTask", 95, 0, 4000, (FUNCPTR)sendTask, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0) == ERROR) { printf ("taskSpawn of tSendTask failed\n"); return (ERROR); } else printf("taskSpawn of tSendTask created\n"); return 0; } //========================================================================== ========= void sendTask(void) { base* basePtr = new A(); if((QID = mq_open(QName, O_CREAT | O_RDWR)) == (mqd_t) -1) { printf("\nsendTask: Create() failed:\n"); printError(errnoOfTaskGet(RcvId)); return; } else printf("\nsendTask: Queue created\n"); if ( mq_send(QID, basePtr , sizeof(base*), HI_PRIO) == ERROR) { printf ("sendTask: mq_send failed\n"); printError(errnoOfTaskGet(SndId)); return; } else { printf ( "sendTask: Msg (\"%d\") sent \n", basePtr); if(mq_close(QID) == ERROR) { printf("sendTask: Close() failed:\n"); printError(errnoOfTaskGet(SndId)); } else printf("sendTask: Queue Closed\n" ); } } //========================================================================== ========= void receiveTask(void) { int prio; base* basePtr; if (mq_open(QName, O_RDONLY) == (mqd_t) -1) { printf ("\nreceiveTask: Open() failed:\n"); printError(errnoOfTaskGet(SndId)); return; } if (mq_receive(QID, basePtr, sizeof(base*), &prio) == ERROR) { printf ("\nreceiveTask: Read() failed:\n"); printError(errnoOfTaskGet(RcvId)); return; } else { printf("\nreceiveTask: Msg of priority %d received: %s\n", prio, basePtr->getMessage()); if(mq_close(QID) == ERROR) { printf ("receiveTask: Close() failed:\n"); printError(errnoOfTaskGet(RcvId)); } else { printf("receiveTask: Queue Closed\n" ); if(mq_unlink(QName) == ERROR) { printf ("receiveTask: Unlink() failed:\n"); printError(errnoOfTaskGet(RcvId)); } else printf("receiveTask: Queue name Unlinked\n" ); } } delete basePtr; } class base and class A are only examples, actual code to long to write out. printError() is a function I wrote to print a string version of the ERRNO. It **very** posible I have my implementation with pointers cokmpletly wrong. But here is what I am trying to do in a nut shell... *basePtr allocation of class A ************ *************************** * 12345 * ---------------------> * * ************ * * 12300 * * *************************** 12345 I actually want to pass the address of the allocated class A (12345) to the queue. The reall class' which would inherit from "base" can very in size at run time. Hence the size of any message going into the Queue would 4 (right?). When receiving the message ("12345") I place it into the basePtr of the receiveTask and access it's contents. If there is a better way to go about it, please share your thought. I am just a newbie and playing around with some code. Any help would be gratefull. thanks in advance, Yan "Michael Lawnick" wrote in message news:ameg44$gg7$1@snoopy.bndlg.de... > Hi Yan, > > posting some of your code snippets might help. > > Regards, > Michael > > "Yan Meunier" schrieb im Newsbeitrag > news:RIki9.5650$0i4.751476@ursa-nb00s0.nbnet.nb.ca... > > Hi guys, I now moved one to playing with POSIX queues. I was able to send > > characters (which is a #define) from one task to another without any > > problems. Now I'm trying to send a pointer to a class I wrote onto the > > queue. The problem comes when I try to receive it. It return me an ERROR > > with a EMSGSIZE ERRNO. > > > > I am able to the same type of message queueing with the use of Wind > Message > > Queues. > > > > Is there something I should know about the POSIX message Queue, that's > > diffrent then the Wind Message Queue as far as sending and receiving goes? > > > > If this Question should be asked in another group, could anyone point me > to > > one? > > Thank you, > > Yan Meunier > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Newbie Question (POSIX queues) Date: Fri, 20 Sep 2002 14:22:32 GMT From: "Yan Meunier" Organization: Aliant Internet Message-ID: References: sorry, looks like the message did'nt come out the way I had hoped "Yan Meunier" wrote in message news:B6Gi9.6167$0i4.808689@ursa-nb00s0.nbnet.nb.ca... > for example > class base{ > public: > virtual char* getMessage()const = 0; > private: > char* messagePtr; > }; > > class A:public base{ > public: > A(); > char* getMessage()const{return messagePtr}; > private: > ... > }; > > A::A() > { > messagePtr = new char("Hello World"); > } > > ---------------------------------------------- > //DEFINES > mqd_t QID; > char QName[] = "TestQ"; > > //FORWARD DECLARATIONS > void receiveTask(void); > void sendTask(void); > //========================================================================== > ========= > int entryPoint(void) > { > if (RcvId = taskSpawn ("tRcvTask", 100, 0, 4000, > (FUNCPTR)receiveTask, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0) == ERROR) > { > printf ("taskSpawn of tRcvTask failed\n"); > return (ERROR); > } > else > printf ("taskSpawn of tRcvTask created\n"); > > if (SndId = taskSpawn ("tSndTask", 95, 0, 4000, (FUNCPTR)sendTask, 0, 0, > 0, 0, 0, 0, 0, 0, 0, 0) == ERROR) > { > printf ("taskSpawn of tSendTask failed\n"); > return (ERROR); > } > else > printf("taskSpawn of tSendTask created\n"); > > return 0; > } > //========================================================================== > ========= > void sendTask(void) > { > base* basePtr = new A(); > > if((QID = mq_open(QName, O_CREAT | O_RDWR)) == (mqd_t) -1) > { > printf("\nsendTask: Create() failed:\n"); > printError(errnoOfTaskGet(RcvId)); > return; > } > else > printf("\nsendTask: Queue created\n"); > > if ( mq_send(QID, basePtr , sizeof(base*), HI_PRIO) == ERROR) > { > printf ("sendTask: mq_send failed\n"); > printError(errnoOfTaskGet(SndId)); > return; > } > else > { > printf ( "sendTask: Msg (\"%d\") sent \n", basePtr); > > if(mq_close(QID) == ERROR) > { > printf("sendTask: Close() failed:\n"); > printError(errnoOfTaskGet(SndId)); > } > else > printf("sendTask: Queue Closed\n" ); > } > } > //========================================================================== > ========= > void receiveTask(void) > { > int prio; > base* basePtr; > > if (mq_open(QName, O_RDONLY) == (mqd_t) -1) > { > printf ("\nreceiveTask: Open() failed:\n"); > printError(errnoOfTaskGet(SndId)); > return; > } > > if (mq_receive(QID, basePtr, sizeof(base*), &prio) == ERROR) > { > printf ("\nreceiveTask: Read() failed:\n"); > printError(errnoOfTaskGet(RcvId)); > return; > } > else > { > printf("\nreceiveTask: Msg of priority %d received: %s\n", prio, > basePtr->getMessage()); > > if(mq_close(QID) == ERROR) > { > printf ("receiveTask: Close() failed:\n"); > printError(errnoOfTaskGet(RcvId)); > } > else > { > printf("receiveTask: Queue Closed\n" ); > > if(mq_unlink(QName) == ERROR) > { > printf ("receiveTask: Unlink() failed:\n"); > printError(errnoOfTaskGet(RcvId)); > } > else > printf("receiveTask: Queue name Unlinked\n" ); > } > } > > delete basePtr; > } > > class base and class A are only examples, actual code to long to write out. > printError() is a function I wrote to print a string version of the ERRNO. > It **very** posible I have my implementation with pointers cokmpletly wrong. > But here is what I am trying to do in a nut shell... > > *basePtr > allocation of class A > ************ > *************************** > * 12345 * ---------------------> * > * > ************ * > * > 12300 * > * > > *************************** > > 12345 > > I actually want to pass the address of the allocated class A (12345) to the > queue. The reall class' which would inherit from "base" can very in size at > run time. > Hence the size of any message going into the Queue would 4 (right?). When > receiving the message ("12345") I place it into the basePtr of the > receiveTask and access it's contents. > If there is a better way to go about it, please share your thought. I am > just a newbie and playing around with some code. Any help would be > gratefull. > > thanks in advance, > Yan > "Michael Lawnick" wrote in message > news:ameg44$gg7$1@snoopy.bndlg.de... > > Hi Yan, > > > > posting some of your code snippets might help. > > > > Regards, > > Michael > > > > "Yan Meunier" schrieb im Newsbeitrag > > news:RIki9.5650$0i4.751476@ursa-nb00s0.nbnet.nb.ca... > > > Hi guys, I now moved one to playing with POSIX queues. I was able to > send > > > characters (which is a #define) from one task to another without any > > > problems. Now I'm trying to send a pointer to a class I wrote onto the > > > queue. The problem comes when I try to receive it. It return me an ERROR > > > with a EMSGSIZE ERRNO. > > > > > > I am able to the same type of message queueing with the use of Wind > > Message > > > Queues. > > > > > > Is there something I should know about the POSIX message Queue, that's > > > diffrent then the Wind Message Queue as far as sending and receiving > goes? > > > > > > If this Question should be asked in another group, could anyone point me > > to > > > one? > > > Thank you, > > > Yan Meunier > > > > > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: kernel crash at symFindByValueAndType() function Date: 20 Sep 2002 07:41:31 -0700 From: cj_rhee@3com.com (CJ) Organization: http://groups.google.com/ Message-ID: After our application run sometime, kernel crashes if we type any command (eg. "i") Here is function trace at the moment of crash right after typing command "i". 0x2ce8fc address belongs to symFindByValueAndType() @0x2ce84c 0x2ce83c address belongs to symFindByValue() @0x2ce824 0x2d2330 address belongs to dbgPrintCall() @0x2d22b8 0x3127b0 address belongs to trcStackLvl() @0x3126a4 0x312778 address belongs to trcStackLvl() @0x3126a4 0x312778 address belongs to trcStackLvl() @0x3126a4 I think some variables in symFindByValueAndType() got corrupted. Anyone had same problem...or know what's going on...? Thanks. --------------------------- Newsgroups: comp.os.vxworks Subject: HELP - access s/w through the Tornado IDE Date: 20 Sep 2002 08:04:44 -0700 From: valrub@hotmail.com (val) Organization: http://groups.google.com/ Message-ID: <64ddf295.0209200704.77303399@posting.google.com> Hello! Please help me with: I'm trying to create access to my s/w (drl.exe) through the Tornado IDE !! I didn't find much help in the Tornado API Programmer's Guide 2.0 (chapter6), "Adding Tools to Tornado" (I also little bit new to Tcl). Maybe somebody has some example of that ... Thank you for any suggestions, examples, any help ! Val --------------------------- Newsgroups: comp.os.vxworks Subject: Re: kernel crash at symFindByValueAndType() function Date: 20 Sep 2002 10:40:49 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: References: Sender: bpringle@DeadDuck CJ> After our application run sometime, kernel crashes if we type any CJ> command (eg. "i") CJ> Here is function trace at the moment of crash right after typing CJ> command "i". [snip] CJ> I think some variables in symFindByValueAndType() got corrupted. CJ> Anyone had same problem...or know what's going on...? It is possible that the symbol table is corrupted. It is more likely that the stack is corrupted. I believe that vxWorks does not do any checking on the stack frame addresses. If the stack frame is addressing an address that is not mapped by the MMU, etc. then the code will cause a data abort or other hardware exception. Can you make a fully mapped memory space? If so, does the problem persists? I believe that the "sysSymTbl", etc can be protected by an MMU with R/O attributes after booting. However, you may have to write you own MMU protection routines to do this [also useful for protecting the vector table]. hth, Bill Pringlemeir. - -- TORPEDOED BY TWO CRASHING WEB BROWSERS. LEAVE BOAT. U-204. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: disk cache error Date: Fri, 20 Sep 2002 17:19:18 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <3D78BE3C.2DBC1BAD@notroll.com> When you use a blkIo device with dosFsDevCreate(), it will automatically create a think CBIO layer on top of the blkDev, which is necesary to support the byte-wise operations needed by the file system in addition to sector-wise operations which are supported by blkIo. This thin layer does have only one sector (512 bytes) worth of memory, and it will be synced to disk almost immediately anyway, so there is no need to worry about sync time, it is not a real disk cache. Hope this helps. - - Leonid "Brian St. Pierre" wrote in message news:FbFi9.2695$Xj4.2781520@newshog.newsread.com... > I asked a related question in a later thread ("obsolete file descriptor"). > Leonid, thanks for the tip on what obsoletes an fd. I have taken out dcache, > calling dosFsDevInit with the device returned by tffsDevCreate. However, it > appears that dosFsDevInit is creating a dcache automatically. When I do a > dosFsShow, it prints a value for pCbio, and cbioShow says that there is a > 512 byte cache. One problem that I see with this arrangement is that I > haven't set the sync interval on this dcache to 0. Would it be better to > create a dcache with size of zero? > > -> dosFsShow "/flash" > > volume descriptor ptr (pVolDesc): 0xf3c3ab8 > cache block I/O descriptor ptr (pCbio): 0xf3c3b78 > auto disk check on mount: DOS_CHK_REPAIR | DOS_CHK_VERB_2 > > [ ... ] > > -> cbioShow 0xf3c3b78 > Cached Block I/O Device, handle=0xf3c3b78 > Description: Block Device > Disk size 29 Mbytes, RAM Size 512 bytes > ^^^^^^^^^^^^^^^^^^ > Block size 512, heads 15, blocks/track 4, # of blocks 61196 > partition offset 0 blocks, type Fixed, Media changed No > > The code that creates this: > > STATUS usrTffsConfig(int drive, int removable, char* fileName) > { > char devName [BOOT_FILE_LEN]; > BLK_DEV * pBlkDev; > void* cbio; > STATUS status; > > if ((UINT)drive >= noOfDrives) > { > printErr ("drive is out of range (0-%d).\n", noOfDrives - 1); > return (ERROR); > } > > /* create a block device spanning entire disk (non-destructive!) */ > pBlkDev = tffsDevCreate (drive, removable); > if (pBlkDev == NULL) > { > printErr ("usrTffsConfig: tffsDevCreate failed.\n"); > return (ERROR); > } > > /* Extract device name from file. */ > devSplit (fileName, devName); > > #undef USE_DCACHE > #if USE_DCACHE > cbio = dcacheDevCreate((void*)pBlkDev, NULL, 1024, devName); > if(cbio == NULL) > { > printErr("usrTffsConfig: dcacheDevCreate failed.\n"); > return ERROR; > } > > /* Set the syncInterval to 0 (last parameter). */ > status = dcacheDevTune(cbio, 0, 0, 0, 0); > if(status != OK) > { > printErr("usrTffsConfig: dcacheDevTune failed.\n"); > return ERROR; > } > #else > cbio = pBlkDev; > #endif > > /* Initialize the boot block device as a dosFs device named > * . Set up for 16 simultaneously open files, with > * automatic CHKDSK and no disk cache. */ > status = dosFsDevCreate (devName, cbio, TFFS_MAX_OPEN_FILES, 0); > if (status != OK) > { > printErr ("usrTffsConfig: dosFsDevCreate failed.\n"); > return (ERROR); > } > > return (OK); > } > > -- > -Brian St. Pierre > brian (at) > bstpierre.org > > "Leonid Rosenboim" wrote: > > If you need the sync interval set to 0, then maybe your better off > > without dcache - you can call dosFsDevCreate with a Blk device id. > > > > Also, if dcache detects that block 0 has been unexpectedly changed, > > and you have not changed the device, it is a goo dreason to worry. > > In other words, if the driver (TFFS?) changes block 0 without an > > apparent reason, there could be other problems that could result > > in data corruption. > > Remember that dcache only checks on block 0 as an additional means > > for detecting card removal, and you can eliminate dcache, which > > will not eliminate the actual problem though. > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Configure w/Tornado 2 Win and VxWorks Date: Fri, 20 Sep 2002 15:39:51 GMT From: "AKA" Organization: Compaq, part of the new hp Message-ID: How do you run ./configure under the Tornado IDE for Windows w/VxWorks 5.5 or in the command window? All I get is"configure is not recognized command ..." Am I missing something? Thanks, Russell --------------------------- Newsgroups: comp.os.vxworks Subject: Re: semTake()/exit()/taskSpawn() Date: Fri, 20 Sep 2002 17:32:39 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <313680C9A886D511A06000204840E1CF67A1A0-100000@whq-msgusr-02.pit.comms.marconi.com> <313680C9A886D511A06000204840E1CF67A1A2-100000@whq-msgusr-02.pit.comms.marconi.com> "Keith Arner" wrote in message news:Pine.GSO.4.42.0209200813330.1626-100000@zimbra... > On Thu, 19 Sep 2002, MC wrote: [snip] > In VxWorks, tasks are identified by a task id. When one task exits, its > task id becomes available for reuse. In VxWorks 5.4, if you spawn a task > immediately after some task has exited, the new task is fairly likely to > get the task id of the task that just exited. The taskId is simply the address of the TCB (Task Control Block) in memory, and this behaviour, correctly described here, is due to the memPartLib which looks for new blocks to allocate beginning with the atest free'd one. The only reason this has not been solved yet, is lazyness. > > The problem arises when something gets a task id, then attempts to use it > after the corresponding task has exited: at this point did you mean to > talk about the old task, or the new task which accidentally got the same > task id? > > At any given time, a mutex semaphore is "owned" by at most one task. What > happens if a task takes the semaphore, forgets to give it up, then exits? > The next task to come along and accidentally get the same task id will > also accidentally own the mutex semaphore. The root problem in your code > would be the task that exited while owning the semaphore, but the only > symptom you see is a completely unrelated task owning the semaphore. When > presented with that symptom, how does one divine what the root problem > was? This is completely wrong. See semMLib reference manual, here is a quote: SEMAPHORE DELETION The semDelete( ) call terminates a semaphore and deallocates any associated memory. The deletion of a semaphore unblocks tasks pended on that semaphore; the routines which were pended return ERROR. Take special care when deleting mutual-exclusion semaphores to avoid deleting a semaphore out from under a task that already owns (has taken) that semaphore. Applications should adopt the protocol of only deleting semaphores that the deleting task owns. > > Another such problem arises with logMsg(). When a task logs a message, > the message does not get printed out immediately, but rather it gets > queued up, to be printed out by tLogTask. This usually get printed to a > slow serial device, so there could be a noticably long time between when > the message gets queued and when it gets printed. When the message is > printed, it includes the name of the task that logged the message, but the > name of the task is found by following a task id that was stored when the > message was queued. If a task queues a message, then exits and another > task is spawned before the message gets printed out, the message is > printed out with the task name of the new task, rather than the old task. > When debugging the problem indicated by your error message, you are > debugging based on misleading evidence. > Quote true, although not very critical. > > When people are designing a new piece of functionality, and are > considering spawning off some new tasks, I advise them to avoid creating > and deleting tasks at run time. If you create all your tasks at > initialization time and they never exit, you are debugging a much more > predictable system. This is not always possible, but is a good idea, to > the extent that it can be done. Or, one could simply work around the taskId reuse issue. Yes, there are several ways to work around this, some actually implemented by customers I have advised. - - Leonid --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Booting from Flash Disk Date: Fri, 20 Sep 2002 17:36:20 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <1032441796.250754@ernani.logica.co.uk> Yes Paul, You can do that, but I am afraid that loading the other stuff after the kernel comes up does too comsume time, and you might consider a possible outcome that the overall process will consume more time than the current scheme, rendering your efforts useless. Unless you are doing this for educational purposes, which is an excellent exercise, I must admit. Stripping down the VxWorks kernel will make you learn a lot about VxWorks and its components, a truly worthwhile experience indeed ! - - Leonid "Paul" wrote in message news:1032441796.250754@ernani.logica.co.uk... > Sorry if this is an old question but... > > Is it possible to load a 'full' VxWorks kernel from a Flash Disk > without using the intermediate 'bootrom' loader? > > Am I right in thinking that it should be possible provided I can > get my kernel smaller than 0.5 MB? Has anyone out there > managed to do this? > > My intention is to try to reduce boot up time by skipping this > intermediate stage. If I have to strip my kernel to the bones > to do this, then that's fine. I can always load in some additional > modules (e.g. SIO drivers etc) after boot. > > I'm using a PIII based board, running VxWorks 5.4.2., with a > Flash Disk that looks like a standard ATA Hard Drive. > > Thanks to all. > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ASM macro in PPC860 with results in args Date: 20 Sep 2002 10:01:01 -0700 From: bwoodhouse@quantumdata.com (Bob W) Organization: http://groups.google.com/ Message-ID: <705be589.0209200901.7aefc612@posting.google.com> References: <705be589.0209181501.171053ca@posting.google.com> <3cni9.519$b_1.203@fe01> I prefer to do as little as possible in asm. However, I am dealing with a body of work that is said to have been optimized in the past for a very time-critical portion of this device, such that timings were taken and the movement from C to asm was deemed important. There is likely not a willingness to pay me to put it all back into C to see if it is fast enough now. Since it is asm for the same PPC860 chip, there is hope that a formula exists to move everything over to the new compiler with good results. Another part to the problem is that after I compile the macro and try to step to, through, and past it in the VisionIceII emulator, in the low-level mode stepping as asm, Ifind that my asm macro appeared not to be there. It is there as a green comment, but there is no asm for it. I put other code before and after it, but not a single hint that my macro was taken seriously. I looked for optimizations that may have been made by the compiler to cause the code to appear where I wasn't looking. On that topic, I have posted a question to the WRS WindSurf support desk in hopes of getting a clue. Thanks, Bob --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ASM macro in PPC860 with results in args Date: Fri, 20 Sep 2002 13:07:29 -0500 From: "Marty Schrader" Organization: Parsec Systems, Inc. Message-ID: References: <705be589.0209181501.171053ca@posting.google.com> <3cni9.519$b_1.203@fe01> <705be589.0209200901.7aefc612@posting.google.com> Umm, is the compiler for some reason not expanding the macro at all? You can probably test that by putting a macro invocation in its own source file and compiling that. See what assembly code the compiler produces. I know you can look at the asm produced for the real source file, but it's a lot easier for test purposes to just look at the code in question by itself. You'll probably have to set up the externals, etc. Is this the only asm macro that you are having trouble with? Is it wrapped with some pragma or other that is screwing things up? How much more assembly is there that will be causing problems? One other thing: for this particular example you can try reproducing the macro's functionality in C and compile it. See what assembly the compiler produces and compare that to the "optimized" version in assembly source. It is possible that the compiler does a fine job without having to optimize by hand. I have found that, these days, a lot of the stuff I used to have to optimize is plenty fast just letting the high level compiler do all the work. - -- Marty Schrader Embedded Technology Consulting from Parsec Systems, Inc. 0N107 Cottonwood Drive Wheaton, IL 60187 (630) 588-0241 www.parsecsystemsinc.com marty@parsecsystemsinc.com "Bob W" wrote in message news:705be589.0209200901.7aefc612@posting.google.com... > I prefer to do as little as possible in asm. However, I am dealing > with a body of work that is said to have been optimized in the past > for a very time-critical portion of this device, such that timings > were taken and the movement from C to asm was deemed important. There > is likely not a willingness to pay me to put it all back into C to see > if it is fast enough now. Since it is asm for the same PPC860 chip, > there is hope that a formula exists to move everything over to the new > compiler with good results. > > Another part to the problem is that after I compile the macro and try > to step to, through, and past it in the VisionIceII emulator, in the > low-level mode stepping as asm, Ifind that my asm macro appeared not > to be there. It is there as a green comment, but there is no asm for > it. I put other code before and after it, but not a single hint that > my macro was taken seriously. I looked for optimizations that may > have been made by the compiler to cause the code to appear where I > wasn't looking. On that topic, I have posted a question to the WRS > WindSurf support desk in hopes of getting a clue. > > Thanks, > Bob --------------------------- Newsgroups: comp.os.vxworks Subject: symTbl.c vs. ELF .stab Date: 20 Sep 2002 14:22:50 -0400 From: "Paul D. Smith" Organization: Nortel Networks, Inc. / Billerica, MA Message-ID: Reply-To: pausmith@nortelnetworks.com Here's something that I'm very confused about: our link process first links all the code into a temporary .o, then it uses nm (makeSymTbl script) to extract all the symbols out and create a symTbl.c file, which is compiled and re-linked with the temporary .o. I understand why this is done, so that the CLI can invoke functions directly. But I don't understand why it's done _this_ way. This all appears to be very standard VxWorks behavior so I'm hoping someone out there can explain why this method was chosen. I mean, all the symbols and their locations are already available through the ELF .stabstr and .stab sections, so why can't the CLI simply use that existing data to call the function, instead of having to generate, compile, and link this whole big data table which effectively duplicates the ELF sections we already have. Is there some quirk here that I'm not seeing, or is it just that no one wanted to figure out (or it was too hard to figure out) how to use the ELF .stabstr / .stab sections instead? Thoughts on this? - -- - ------------------------------------------------------------------------------- Paul D. Smith HASMAT--HA Software Mthds & Tools "Please remain calm...I may be mad, but I am a professional." --Mad Scientist - ------------------------------------------------------------------------------- These are my opinions---Nortel Networks takes no responsibility for them. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: symTbl.c vs. ELF .stab Date: 20 Sep 2002 15:40:53 -0400 From: "Paul D. Smith" Organization: Nortel Networks, Inc. / Billerica, MA Message-ID: References: Reply-To: pausmith@nortelnetworks.com Hm. I get it, I think. elfToBin doesn't emit the .stab so it's not there in the image. But I still don't really understand why they decided to do that extra extract to C etc. step instead of including the .stab section in the image and walking that to create the symbol table. - -- - ------------------------------------------------------------------------------- Paul D. Smith HASMAT--HA Software Mthds & Tools "Please remain calm...I may be mad, but I am a professional." --Mad Scientist - ------------------------------------------------------------------------------- These are my opinions---Nortel Networks takes no responsibility for them. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Is it necessary to program the bootloader on different chip than the application or TFFS? Date: Fri, 20 Sep 2002 13:13:49 -0700 From: "Dave Francheski" Organization: Advanced Telcom Group Inc. Internet Services Message-ID: References: <41c1fd2b.0209170118.76648c6c@posting.google.com> I had a couple of questions regarding this - 1) I need to also store other kinds of information in the reserved area beneath TFFS (e.g., the boot-line, manufacturing parameters, etc.) along with the bootROM (bootloader) itself. Traditionally, the boot-line is stored in the very last sector of flash, so is it acceptable to relocate it to this reserved area as well? 2) How will the bootloader ("bootROM") write/read the boot-line, etc.? VxWorks provides the generic flashMem.c module with sysFlashSet() / sysFlashGet() APIs. These are indirectly called through the nvRAMtoFlash.c module sysNvRamSet() / sysNvRamGet() APIs from the bootloader (bootConfig.c). Is there any way to share this lower-level flashMem.c driver with TFFS (and its MTD module). It's always good to promote code re-use, but I don't see an easy way of sharing the two flash drivers. Am I taking the right approach here? regards, Dave Francheski davidf@caymas.com "Brian St. Pierre" wrote in message news:QTKh9.2498$Xj4.2542159@newshog.newsread.com... > "Danny Cohen" wrote: > > From you're experience: Do you see any SW or HW or System drawbacks in > > the solution of programming both boot loader and application and File > > System on the same Flash Chip? > > Not at all. This is a piece of cake. I've done this in a couple of different > designs. Just make sure you reserve plenty of room for your boot loader. > (The first time I did this we ran into problems because the boot loader grew > and it was a mess to fix all the boards that needed to have more space > reserved.) For extra security, we lock the sectors of the device that hold > the boot loader. > > -- > Brian St. Pierre > brian (at) > bstpierre.org > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: help with vision ICE Date: Fri, 20 Sep 2002 20:27:47 GMT From: "Philippe Venne" Organization: Magma Communications Ltd. Message-ID: <7xLi9.12833$V21.303352@news> References: <3D8AF9BC.CA270123@hotmail.com> Hi Goerge Here is a little more information on that bug. First a capture of my registers after the chash /* here we are loading the application in SDRAM */ BDI>load 0x10000 application.bin BIN Loading application.bin , please wait .... Loading program file passed BDI> /* Here we are seeting the PC counter and starting the core */ BDI>go 0x10000 /* We get this right after the crash */ - - TARGET: target has entered debug mode BDI>info Target state : debug mode Debug entry cause : machine check interrupt Current PC : 0x0027ab6c Data address : 0x05260024 DSISR value : 0x00000069 BDI>rd icr icr: 0x12203400 304100352 BDI> This is with the other BDM tool that we are using, the BDI2000 from Abatron. And this device is accessed using a GPCM, here ara the settings: WM32 0x0220012C 0xFF0009FC ;OR5 option register bank 5 WM32 0x02200128 0x05000001 ;BR5 base register bank 5 Finally to this point, I have not tried slowing down the system clock, but I am working on it. Do you think it has something to do with the DER register? At this point it is set to 0x71C7400F Thanks. Philippe Venne "Varndell Engineering" wrote in message news:3D8AF9BC.CA270123@hotmail.com... > Jim wrote: > > > > in my bootloader on the mpc860 i disable the busmonitor ... it seems the BDM > > (vision ICE) may be turning this back on (but the register sypcr is still > > ffffff08 as i set it) ... is it possible the bus will time out on BDM when > > accessing a slow device ... > > The BM is enabled whenever FRZ is asserted or when a debug mode request > is pending; independent of SYPCR[BME]. So, yes, the bus can timeout. > > > The problem in general is this ... i can boot my board without the ice ... > > but with the ice i die accessing a slow device 500 khz ... > > Perhaps a little more information would help. Can you describe what > happens when you die? Are you using Click, SingleStep, what? > Is your slow device accessed via GPCM or UPM? > Can you slow down the system clock to a point where the max BM timeout > is greater than the access time of the device? Just to see if this > mysterious dying ceases? > > Regards, > George Varndell > -- > Varndell Engineering, LLC Embedded Systems Expertise > Need an END driver for DP8382x GigE? Contact us. > http://www.varndellengineering.com > > > -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- > http://www.newsfeed.com The #1 Newsgroup Service in the World! > -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: bootloader and TrueFFS interaction Date: Fri, 20 Sep 2002 14:47:15 -0700 From: "Dave Francheski" Organization: Advanced Telcom Group Inc. Internet Services Message-ID: TrueFFS allows the user to reserve the lower portion of the flash Device for bootimages, etc. I have the following questions concerning this: 1) I need to also store other kinds of information in the reserved area beneath TFFS (e.g., the boot-line, manufacturing parameters, etc.) along with the bootROM (bootloader) itself. Traditionally, the boot-line is stored in the very last sector of flash, so is it acceptable to relocate it to this reserved area as well? 2) How will the bootloader ("bootROM") write/read the boot-line, etc.? VxWorks provides the generic flashMem.c module with sysFlashSet() & sysFlashGet() APIs. These are indirectly called through the nvRAMtoFlash.c module sysNvRamSet() & sysNvRamGet() APIs from the bootloader (bootConfig.c). Is there any way to share this lower-level flashMem.c driver with TFFS (ie., its MTD module). It's always good to promote code re-use, but I don't see an easy way of sharing the two flash drivers (aside from adding a special API into the MTD to allow the bootloader to access the flash). Am I taking the right approach here? regards, Dave Francheski davidf@caymas.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: help with vision ICE Date: Fri, 20 Sep 2002 21:52:37 GMT From: "Philippe Venne" Organization: Magma Communications Ltd. Message-ID: References: <3D8AF9BC.CA270123@hotmail.com> Hi Goerge It looks loke I have a "debug mode request pending". How can I get rid of this? I do not need debug capabilities... Thanks. Philippe "Varndell Engineering" wrote in message news:3D8AF9BC.CA270123@hotmail.com... > Jim wrote: > > > > in my bootloader on the mpc860 i disable the busmonitor ... it seems the BDM > > (vision ICE) may be turning this back on (but the register sypcr is still > > ffffff08 as i set it) ... is it possible the bus will time out on BDM when > > accessing a slow device ... > > The BM is enabled whenever FRZ is asserted or when a debug mode request > is pending; independent of SYPCR[BME]. So, yes, the bus can timeout. > > > The problem in general is this ... i can boot my board without the ice ... > > but with the ice i die accessing a slow device 500 khz ... > > Perhaps a little more information would help. Can you describe what > happens when you die? Are you using Click, SingleStep, what? > Is your slow device accessed via GPCM or UPM? > Can you slow down the system clock to a point where the max BM timeout > is greater than the access time of the device? Just to see if this > mysterious dying ceases? > > Regards, > George Varndell > -- > Varndell Engineering, LLC Embedded Systems Expertise > Need an END driver for DP8382x GigE? Contact us. > http://www.varndellengineering.com > > > -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- > http://www.newsfeed.com The #1 Newsgroup Service in the World! > -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: task trace during exception Date: 20 Sep 2002 15:29:33 -0700 From: alex.mo@attbi.com (alex mo) Organization: http://groups.google.com/ Message-ID: References: Hi Hadi, You can only trace the stack of a task other than itself. Therefore, try to store the task id and run the trace from another task. - --alex "s s" wrote in message news:... > Hello, > I have a signal handler hooked up to one of my task. This signal handler is > for "BUS ERROR". Inside the handler I print the stack of the current task. > > So the signal handler would get called if the task gets a exception due to a > bad memory access. Now when I use "trcStack" to trace the stack , I only > see the signal handler. I would like to see the instructions that caused > this exception. How do I dig it out of the stack. Does anyone has any > suggestion > Hadi --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Query on vxWorks.st_rom.bin Date: Fri, 20 Sep 2002 22:26:42 GMT From: "Chad" Organization: Shaw Residential Internet Message-ID: References: Make sure that your network device is set in the 'other' field of you bootline. In usrNetwork.c it is set: else { if ((strncmp (params.bootDev, "scsi", 4) == 0) || (strncmp (params.bootDev, "ide", 3) == 0) || (strncmp (params.bootDev, "ata", 3) == 0) || (strncmp (params.bootDev, "fd", 2) == 0) || (strncmp (params.bootDev, "tffs", 4) == 0)) { /* booting from disk, configure network if requested */ if (params.other [0] != EOS) { /* * Protocol attachment is required and automatic * configuration with DHCP or BOOTP is supported * for all these network devices. */ attachFlag = TRUE; pNetDev = params.other; ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ AFAIK, both problems could be related, as tornado can be set to communicate across the network. Wish I could be more help, Chad "Somisetty, Somisetty" wrote in message news:amfvm0$b6h$1@overload.lbl.gov... > Hi > I am using WindRiver's SBC8260 board. I have made a vxWorks.st_rom.bin > using standard VxWorks Makefiles. I have burnt the image on flash. I am able > to see the VxWorks Shell from the serial hyperterminal. My questions are: > 1. I am not able to connect to target using Tornado. Why? > 2. From hyperterminal windShell, I found that tNetTask is not spawned. Why > is tNetTask not spawned? How do I make stand-alone ROM-based VxWorks with > internet capabilities? > > Thanks in advance. > > Regards > Somi --------------------------- Newsgroups: comp.os.vxworks Subject: Re: bootloader and TrueFFS interaction Date: Fri, 20 Sep 2002 21:19:30 -0400 From: Andray Kaganovsky Organization: Futureway Message-ID: References: Reply-To: andrayk@futurewaynospam.com Dave Francheski wrote: > TrueFFS allows the user to reserve the lower portion of the flash > Device for bootimages, etc. I have the following questions concerning > this: > > 1) I need to also store other kinds of information in the reserved area > beneath TFFS (e.g., the boot-line, manufacturing parameters, etc.) > along with the bootROM (bootloader) itself. Traditionally, the > boot-line > is stored in the very last sector of flash, so is it acceptable to > relocate it to > this reserved area as well? > > 2) How will the bootloader ("bootROM") write/read the boot-line, etc.? > VxWorks provides the generic flashMem.c module with sysFlashSet() & > sysFlashGet() APIs. These are indirectly called through the > nvRAMtoFlash.c module sysNvRamSet() & sysNvRamGet() APIs > from the bootloader (bootConfig.c). > > Is there any way to share this lower-level flashMem.c driver with TFFS > (ie., its MTD module). It's always good to promote code re-use, but I > don't see an easy way of sharing the two flash drivers (aside from > adding a special API into the MTD to allow the bootloader to access the > flash). > > Am I taking the right approach here? > > regards, > Dave Francheski > davidf@caymas.com [AndrayK]. You can use TFFS's routine tffsRawio() to access physical flash locations. See description of FL_PHYSICAL_READ and FL_PHYSICAL_WRITE calls in TFFS manual. Andray --------------------------- Newsgroups: comp.os.vxworks Subject: Wind Web Server and OpenSSL Date: 20 Sep 2002 19:32:14 -0700 From: tjaimin@yahoo.com (Jai) Organization: http://groups.google.com/ Message-ID: <17902e07.0209201832.403345bf@posting.google.com> We use Tornado 2.0 the Wind Web Server 1.0. I am looking to port OpenSSL to Wind Web Server. Has anybody done it before? Does anyone have any ideas whether it can be done? and if yes what are the complexities involved? Has anybody used any other implementation of SSL on Wind Web Server. Thanks, Jai. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Query on vxWorks.st_rom.bin Date: 21 Sep 2002 02:20:41 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0209210120.1e0ccd52@posting.google.com> References: "Somisetty, Somisetty" wrote in message news:... > 1. I am not able to connect to target using Tornado. Why? You have to include the WDB agent and select and configure the proper connection type. Check whether WDB is included. w/o Network, you'll only be able to connet through e.g. serial interface... > 2. From hyperterminal windShell, I found that tNetTask is not spawned. Why > is tNetTask not spawned? How do I make stand-alone ROM-based VxWorks with > internet capabilities? Most likeley you're missing: #define STANDALONE_NET A complete set for standalone + symbol table would be (plus some additional usefull stuff): #define INCLUDE_STANDALONE_SYM_TBL #define INCLUDE_STAT_SYM_TBL #define INCLUDE_SYM_TBL #define INCLUDE_SYM_TBL_INIT #define INCLUDE_SHOW_ROUTINES #define INCLUDE_NET_SHOW #define INCLUDE_PING #define INCLUDE_TCP_DEBUG #define INCLUDE_DEBUG #define STANDALONE_NET + make vxWorks.st toni --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ASM macro in PPC860 with results in args Date: 21 Sep 2002 02:25:10 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0209210125.25a92845@posting.google.com> References: <705be589.0209181501.171053ca@posting.google.com> bwoodhouse@quantumdata.com (Bob W) wrote in message news:<705be589.0209181501.171053ca@posting.google.com>... Hi Bob, I don't know the answer to this by hard as the code provided looks good at first glance, however this should be easy to catch. The gnu complier provides detailed output with the "--save-temps" flag given. This will give you: * the preprocessed file * the assembler file By examining the two it should be possible to catch this (so, which stage is failing?) toni --------------------------- Newsgroups: comp.os.vxworks Subject: Re: protocol stack problem Date: 21 Sep 2002 02:28:31 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0209210128.499227c3@posting.google.com> References: He Jin wrote in message news:... Maybe try to add an input hook as a first try (either binding a SNARF Protocol or adding an ether input hook). This will allow you to have a look at the packet. There you should be able to see what's actually going wrong... mabye only an alignement issue? tnoi --------------------------- Newsgroups: comp.os.vxworks Subject: Re: help with vision ICE Date: Sat, 21 Sep 2002 07:17:35 -0400 From: Varndell Engineering Organization: Posted Via Newsfeeds.com = SPEED+RETENTION+COMPLETION = http://www.newsfeeds.com Message-ID: <3D8C554F.E2093351@hotmail.com> References: <3D8AF9BC.CA270123@hotmail.com> Philippe Venne wrote: > > It looks loke I have a "debug mode request pending". How can I get rid of > this? I do not need debug capabilities... Then why are you using a debugger? Are you just using it as an image loader? I don't think the bus monitor is the problem. You've got the timeout set to the highest value, which gives you 2040 system clocks before the BM terminates with internal TEA. Even if you're running your 860 at very high rates, the BM allows your 500Khz device plenty of time to terminate accesses before it does. So, even if the BM is being enabled behind your back, it shouldn't trigger. You should enable the BM. My hunch is that you are getting a real, externally generated, TEA. Can you use a logic analyzer to watch the bus control signals during accesses to this device? Regards, George Varndell - -- Varndell Engineering, LLC Embedded Systems Expertise Need an END driver for SiS900? Contact us. http://www.varndellengineering.com - -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- http://www.newsfeed.com The #1 Newsgroup Service in the World! - -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: help with vision ICE Date: Sat, 21 Sep 2002 16:43:52 GMT From: "Philippe Venne" Organization: Magma Communications Ltd. Message-ID: References: <3D8AF9BC.CA270123@hotmail.com> <3D8C554F.E2093351@hotmail.com> Hi Goerge, First, some more details. Let me explain my intention in this procedure I am trying to run. I am a production automation designer, so I try to get board production faster and automated. Therefore here is my plan: 1. Copy the software image to FEEPROM (Flash) with the BDM tool 2. Copy the OS image in SDRAM with the BDM tool ( to save expanding time of the image, 2:30 minutes in this case) 3. Set PC and SP and start the application from SDRAM to do configuration and testing in production I took your advice and hooked a logic analyser to the system trigerring on my suspiscious address. Because I can get the thing to work when the BDM is not connected, I looked at it going, and beleive me or not, it took to my slow device 103.6 us to send TA. With my bus running at 40 MHz, if the BM is on, it will trigger after 51 us when set to maximum timeout. I then ran the board with loading the image in SDRAM with the BDM tool, and I received a TEA after, as expected, 51.52 us. Software designers knew about this problem, so they turned of the BM. So back to my last question, how do I get rid of this "debug mode request pending" ?!? Thanks. Philippe. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Wind Web Server and OpenSSL Date: 21 Sep 2002 10:05:48 -0700 From: vdalal@hotmail.com (Vasu Dalal) Organization: http://groups.google.com/ Message-ID: <363d071b.0209210905.38dddf11@posting.google.com> References: <17902e07.0209201832.403345bf@posting.google.com> Jai: > We use Tornado 2.0 the Wind Web Server 1.0. I am looking to port > OpenSSL to Wind Web Server. Has anybody done it before? Start here : http://www.xs4all.nl/~borkhuis/vxworks/vxw_pt6.html#6.9 hth, vasu --------------------------- Newsgroups: comp.os.vxworks Subject: Fastest way to transfere data between a PC with win2000 and a PC with vxWorks? Date: 21 Sep 2002 14:46:54 -0700 From: ajn@bredband.net (Ake) Organization: http://groups.google.com/ Message-ID: <7777af4e.0209211346.1b507f81@posting.google.com> Hello! We may start a new project where we need to move a lots of between two PC one with win2000 and one with VxWorks. My ide is to set up a 100Mbits/s ethernet link and transfer data with FTP. Is there a faster way to do this. (instead of using Ehernet) The PC with win2000 will be responsible to move the files. Note no other computers will use this cable. I hope there is support for 100 Mbits/s Ethernet in vxWorks, Is it? On the same Ethernet connection I like to send commands between the two PC, what protocol should I use then? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: help me out Date: Sat, 21 Sep 2002 19:05:27 -0400 From: joe durusau Organization: WEBUSENET.com Message-ID: <3D8CFB36.CC2605C9@bellsouth.net> References: suvaranjan das wrote: > Hello room, > > Plz help me on loading VxWorks eval version. > > suva_tops@hotmail.com > > _________________________________________________________________ > Join the world’s largest e-mail service with MSN Hotmail. > http://www.hotmail.com You can get that from your local WRS representative. You can't get it off the web or any such. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Newbie Starter Kit Date: Sun, 22 Sep 2002 00:58:00 GMT From: "Blackie Beard" Message-ID: Reply-To: "Blackie Beard" ============================== I want to learn vxWorks in my spare time. How do I get started for very cheap? Is there a good kit, and what books are recommended. My background is H/W, S/W, F/W and SoC, but never studied any OS. Thanks in advance! BB ============================== --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Newbie Starter Kit Date: Sun, 22 Sep 2002 04:36:08 GMT From: nunya@damn.business.com (Bruce) Organization: Prodigy Internet http://www.prodigy.com Message-ID: <3d8d4905.43812929@news.houston.sbcglobal.net> References: In comp.os.vxworks "Blackie Beard" wrote: >============================== >I want to learn vxWorks in my spare time. >How do I get started for very cheap? >Is there a good kit, and what books are >recommended. My background is H/W, >S/W, F/W and SoC, but never studied >any OS. Thanks in advance! The best way to get started cheap is to get something free like MicroC/OS-II. You can learn how to write multi-tasking programs and all you learn will help you with VxWorks. TTBMK, there are no books out on VxWorks other than the manuals. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Fastest way to transfere data between a PC with win2000 and a PC with vxWorks? Date: Sun, 22 Sep 2002 08:22:59 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <7777af4e.0209211346.1b507f81@posting.google.com> You will be able to do even better then 100Mbps with FTP, (if for example you had a GE interface), but doing this isn't going to be easy. You wull need to spend some time enhancing (or applying patches) to the FE driver, so it will perform at 100Mbps without freezing, then you will have to do some work on the disk attachment driver (IDE or SCSI) and make sure it tops 10MBytes/sec. Lastly, the only modification for the VxWorks FTP server is to increase the buffer it uses move data between network and disk, and tue the disk cache. If this souds like a lot of work, it is, but unfortunately I can not suggest any alternative way to quickly move this data. - - Leonid "Ake" wrote in message news:7777af4e.0209211346.1b507f81@posting.google.com... > Hello! > We may start a new project where we need to move a lots of between two > PC one with win2000 and one with VxWorks. My ide is to set up a > 100Mbits/s ethernet link and transfer data with FTP. Is there a faster > way to do this. (instead of using Ehernet) > > The PC with win2000 will be responsible to move the files. Note no > other computers will use this cable. > > I hope there is support for 100 Mbits/s Ethernet in vxWorks, Is it? > > On the same Ethernet connection I like to send commands between the > two PC, what protocol should I use then? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Configure w/Tornado 2 Win and VxWorks Date: Sun, 22 Sep 2002 12:53:13 +0000 (UTC) From: Johan Borkhuis Organization: Proxim Corp. Message-ID: References: "AKA" wrote: > How do you run ./configure under the Tornado IDE for Windows w/VxWorks > 5.5 or in the command window? All I get is"configure is not > recognized command ..." Am I missing something? I don't think this will work under a standard Windows environment. You will need to install Cygwin for that. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ASM macro in PPC860 with results in args Date: 22 Sep 2002 12:27:19 -0700 From: bwoodhouse@quantumdata.com (Bob W) Organization: http://groups.google.com/ Message-ID: <705be589.0209221127.171454fe@posting.google.com> References: <705be589.0209181501.171053ca@posting.google.com> <8400a733.0209210125.25a92845@posting.google.com> In response to Toni: Toni, Thanks. I will look at the -save-temps once I get back to work on Monday. I found that if I just took the contents of the macro and just placed it into the spots where I was trying to use it, the very same code would then start appearing in the vxWorks executable -- I could step through it in visionIce. So now it is should be a piece of cake to test and debug it in that form. It is still a mystery why it vanishes when in the #define macro guise. Best, Bob W --------------------------- Newsgroups: comp.os.vxworks Subject: New Object appear at NULL Date: 22 Sep 2002 13:16:23 -0700 From: bwoodhouse@quantumdata.com (Bob W) Organization: http://groups.google.com/ Message-ID: <705be589.0209221216.73ce6f81@posting.google.com> Hello. A co-worker of mine discovered a problem on the product that I am transferring to vxWorks. I am confused about the nature of the problem and further surprised that I found no similar experiences on this group or WindSurf. So here are the facts. Any time I do a "new" the pointer returned is pointing to zero. I made a simple example where useAppInit calls a routine, appLaunch(), in my initial C++ file, appLaunch.cpp. I have distilled it down to something simple: int *ip; ip = new int[50]; printf("ip created at %x.\n",ip); So I see ip is zero, normally suggesting out-of-memory, you would think. The other failed lines of code were more like product = new TProduct; // class TProduct is allocated, returns pointer product. I did not notice this at first, as I created several objects that appeared to run fine (before adding the check for NULL pointer on them). They must have been using memory spaced at "lucky" locations for the most part. Now the memShow has no trouble telling me that I have plenty of free space. I have RAM_LO at 0x100000 where the vxWorks FTPs in at about 2-3 MB, while RAM_HI is at 0xE00000, with no user reserved portions involved. I had some code simply in C in the Tornado/vxWorks environment without this problem, as storage is just reserved during the build and found itself around the 0x350000 area mostly. The C++ stuff seems to "run" fine in many ways but for the mislocated objects. I do notice warnings about __builtin_new and the other __builtin's being redeclared from size 32 to something bigger, but I have seen other posters conclude that they could ignore the warning of that type. Any thoughts? Did I forget to turn on a flag? Should I go back to C++ school, etc? Best, BobW --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ADS8260 BSP Date: 22 Sep 2002 17:31:06 -0700 From: demorz@yahoo.com (Dennis) Organization: http://groups.google.com/ Message-ID: <4d0e703e.0209221631.73de27fa@posting.google.com> References: "Dan Gold" wrote in message news:... > Achim Zimmer wrote in message ... > >Yes, thats is right. This way works to figure out where the tables and > >buffers are. > > > Did you examine it ? where are the buffer descriptor tables, in DPRAM or > external memory ? > If they are in external memory there is nothing to move. If they are in > DPRAM then you need to move them. > > Your main problem is you need the DPRAM starting at the base of bank 1. > First, are you sure something is using that memory ? If so, how do you know > ? What happens if you zero out that memory while the system is running ? > Or can you dump it and see if values are changing ? Other candidates that > may use that space are the buffer descriptors for the serial port (i.e. the > SCC, since they cannot reside in external memory). You can look at the > serial driver where it initializes it's BDs and see where it is putting > them. > > It sounds like you need to do 1 and/or 2 things: > 1. Find some documentation on the BSP that describes where everything is > placed in DPRAM. This may or may not exist. And you may have to look in a > header file for it. > 2. Look through all the code in the BSP for the various peripherals you are > using (FCC, SCC, anything else like I2c or SPI, etc...) and see where they > place their stuff in DPRAM. > > >There more complicated problem is to move it, because WRS uses a very > tricky > >init string at ADS8260. > > Again, you don't have to move it unless you find out the BDs are in DPRAM. > Make sure of that first before dealing with the init string. > > >I have posted this initfunction from WRS, because my english isn't so good > >to explaine it. > > > >I changed this motFccEndParamTemplate to > > "0x%x:0x%x:0x2100:0x300:-1:-1:-1:-1:0x%x:0x%x:0x%x:0x%x:0x%x:0x%x"; > >but the result was that the MUX failed to load the device. > > > what did you change it from ? > > >The problem is, that I dont understand exactly the usage of the template > >fields. > The END device load string formed by this routine is in the following > format. > >* :::::: > >* ::::::: > >* > >* > >* .IP > >* buffer descriptors base address > > this is the parameter you want to change. If it's 0, it probably defaults > to something inside the driver, but I don't know what. Maybe call WRS and > ask for motFcc.c if you need to. > > >* .IP > >* tx buffer in internal memory > >* .IP > >* rx buffer in internal memory > > If the above 2 fields are in an area you need, you'll need to move them too. > > The bottom line is find out if anything is interfering with the DPRAM you > need, and what it is first. > > hth, > Dan Gold gold@ensemble.com Moving to a different FCC on 8260 requires much more than figuring out where the BDs are and changing the parameter string. I did that long ago and I don't remember the details, but I remember I had to make many changes, not just in one or two places. Make sure you physically can use another FCC. If some of its pins are used for something else, you are unlikely to be able to change FCCs. If you are using the Tornado project environment, check the build paramenters and included component properties. Look for references to FCC2 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ASM macro in PPC860 with results in args Date: 22 Sep 2002 17:43:44 -0700 From: demorz@yahoo.com (Dennis) Organization: http://groups.google.com/ Message-ID: <4d0e703e.0209221643.785064d3@posting.google.com> References: <705be589.0209181501.171053ca@posting.google.com> <8400a733.0209210125.25a92845@posting.google.com> gonozal@gmx.at (Toni Kurzberger) wrote in message news:<8400a733.0209210125.25a92845@posting.google.com>... > bwoodhouse@quantumdata.com (Bob W) wrote in message news:<705be589.0209181501.171053ca@posting.google.com>... > > Hi Bob, > > I don't know the answer to this by hard as the code provided looks > good at first glance, however this should be easy to catch. > The gnu complier provides detailed output with the "--save-temps" flag > given. This will give you: > * the preprocessed file > * the assembler file > By examining the two it should be possible to catch this (so, which > stage is failing?) > > toni I am not absolutely sure, but I believe the Diab compiler and GNU for MPC860 have different calling conventions. I believe, they even use different registers for implementing the parameter stack. After WRS bought Diab, I tried to switch from GNU to Diab. To do that, I would have had to recompile the VxWorks libraries with the Diab compiler, but I didn't have the source code... And I couldn't get any useful help from WRS, either. So, check that the calling conventions are the same in your ASM code as the ones that the compiler generates. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Booting from Flash Disk Date: 22 Sep 2002 17:54:46 -0700 From: demorz@yahoo.com (Dennis) Organization: http://groups.google.com/ Message-ID: <4d0e703e.0209221654.3e0919ae@posting.google.com> References: <1032441796.250754@ernani.logica.co.uk> "Leonid Rosenboim" wrote in message news:... > Yes Paul, > > You can do that, but I am afraid that loading the other stuff after > the kernel comes up does too comsume time, and you might > consider a possible outcome that the overall process will > consume more time than the current scheme, rendering > your efforts useless. > > Unless you are doing this for educational purposes, > which is an excellent exercise, I must admit. > Stripping down the VxWorks kernel will make you > learn a lot about VxWorks and its components, > a truly worthwhile experience indeed ! > > - Leonid > > "Paul" wrote in message > news:1032441796.250754@ernani.logica.co.uk... > > Sorry if this is an old question but... > > > > Is it possible to load a 'full' VxWorks kernel from a Flash Disk > > without using the intermediate 'bootrom' loader? > > > > Am I right in thinking that it should be possible provided I can > > get my kernel smaller than 0.5 MB? Has anyone out there > > managed to do this? > > > > My intention is to try to reduce boot up time by skipping this > > intermediate stage. If I have to strip my kernel to the bones > > to do this, then that's fine. I can always load in some additional > > modules (e.g. SIO drivers etc) after boot. > > > > I'm using a PIII based board, running VxWorks 5.4.2., with a > > Flash Disk that looks like a standard ATA Hard Drive. > > > > Thanks to all. > > > > > > I think the actual problem is that the file system support is not available until VxWorks is up and running. So, to do what you want to do, your boot code should be smart enough to figure out where the VxWorks image is in the flash, perhaps by analysing the FAT and directory. If you use TFFS, the things are even more complicated. --------------------------- Newsgroups: comp.os.vxworks Subject: Error when updating VxWorks version Date: 22 Sep 2002 20:29:36 -0700 From: ktrenor2@csc.com.au (KevinT) Organization: http://groups.google.com/ Message-ID: Hi I've looked at various FAQ's and searched this group for an answer but haven't found one so here goes. Note that I'm this is the first time I've had to get involved in the setup of vxWorks at this level so I'm a bit of a novice at this stuff. I'm in the process of updating from vxWorks 5.3.1 to 5.4.2 and am consistently getting a message indicating that the vxWorks.sym file cannot be read - the error code is 0xd001 which I think means DEVICE_NOT_FOUND (from iosLib.h). This causes the symbol table to not be loaded however vxWorks is itself loaded. Also one of the executables attempts to read a text file from the host machine and this also fails. Other than that the executable appears to behave as it did for the old vxWorks. This is for 2 different Power PC boards and the same error occurs in both cases. At load time the network device appears to load ok (no error is shown anyway) and its using ftp to do the load (to an SGI machine if that matters). The exact same setup works correctly for the old version. I've looked thru all the header files etc and done a comparison between those we've got for 5.3.1 and those for 5.4.2 and haven't noticed anything that would give this error or even what device it is complaining about. Any tips/clues would be greatly appreciated. Thanks Kevin --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Error when updating VxWorks version Date: Mon, 23 Sep 2002 04:24:05 GMT From: "drdiags" Organization: AT&T Broadband Message-ID: References: Hello Kevin, Sorry, no clues, but could you try a built-in symbol table w/network show routines until you resolve this problem? Did you have SENS in your 5.3.1 installation(END driver) or were you using the if_xxx BSD network drivers? Have you looked on Wind Surf for the T1.0.1 to T2.0 migration guide? This may help with any underlying BSP migration tips. I would try a sniffer to see what the request for the symbol table looked like, if it was sent at all. Did you see any request for the symbol file in your FTP server's logging? Other than these questions, I am only guessing. Usually when I see this scenario, the END in the kernel DEFAULT_BOOTLINE didn't match what I had in the bootrom or things along that line. That's why I believe in the target shell w/a built-in symbol table until I have stabilized things in my upgrade. HTH. "KevinT" wrote in message news:de02855f.0209221929.777403da@posting.google.com... > Hi > > I've looked at various FAQ's and searched this group for an answer but > haven't found one so here goes. Note that I'm this is the first time > I've had to get involved in the setup of vxWorks at this level so I'm > a bit of a novice at this stuff. > > I'm in the process of updating from vxWorks 5.3.1 to 5.4.2 and am > consistently getting a message indicating that the vxWorks.sym file > cannot be read - the error code is 0xd001 which I think means > DEVICE_NOT_FOUND (from iosLib.h). This causes the symbol table to not > be loaded however vxWorks is itself loaded. Also one of the > executables attempts to read a text file from the host machine and > this also fails. Other than that the executable appears to behave as > it did for the old vxWorks. This is for 2 different Power PC boards > and the same error occurs in both cases. At load time the network > device appears to load ok (no error is shown anyway) and its using ftp > to do the load (to an SGI machine if that matters). The exact same > setup works correctly for the old version. > > I've looked thru all the header files etc and done a comparison > between those we've got for 5.3.1 and those for 5.4.2 and haven't > noticed anything that would give this error or even what device it is > complaining about. > > Any tips/clues would be greatly appreciated. > > Thanks > Kevin --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Fastest way to transfere data between a PC with win2000 and a PC with vxWorks? Date: Mon, 23 Sep 2002 08:45:26 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <7777af4e.0209211346.1b507f81@posting.google.com> Reply-To: "Michael Lawnick" Sender: mlawnick@pd9524384.dip.t-dialin.net Hi, as FTP in my experience is an easy to handle but slow protocoll, I'd suggest to write your own transfer protocoll based on UDP packets. The benefit might improve your performance from about 50-60% upto 90-95% (i.e. ~100-120 kByte/s) :-)) The down part is that you are out of any standards :-( Regards, Michael "Ake" schrieb im Newsbeitrag news:7777af4e.0209211346.1b507f81@posting.google.com... > Hello! > We may start a new project where we need to move a lots of between two > PC one with win2000 and one with VxWorks. My ide is to set up a > 100Mbits/s ethernet link and transfer data with FTP. Is there a faster > way to do this. (instead of using Ehernet) > > The PC with win2000 will be responsible to move the files. Note no > other computers will use this cable. > > I hope there is support for 100 Mbits/s Ethernet in vxWorks, Is it? > > On the same Ethernet connection I like to send commands between the > two PC, what protocol should I use then? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: symTbl.c vs. ELF .stab Date: Mon, 23 Sep 2002 08:53:08 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: Reply-To: "Michael Lawnick" Sender: mlawnick@pd9524384.dip.t-dialin.net Hi Paul, two ideas: - - to make it easily possible to remove symbols that shouldn't be available on CLI - - there are more formats than ELF might be there are more ... finally: it works well, so where is the beef? Regards Michael "Paul D. Smith" schrieb im Newsbeitrag news:p51y7oh2ka.fsf@lemming.engeast.baynetworks.com... > Hm. I get it, I think. elfToBin doesn't emit the .stab so it's not > there in the image. > > But I still don't really understand why they decided to do that extra > extract to C etc. step instead of including the .stab section in the > image and walking that to create the symbol table. > > -- > -------------------------------------------------------------------------- - ----- > Paul D. Smith HASMAT--HA Software Mthds & Tools > "Please remain calm...I may be mad, but I am a professional." --Mad Scientist > -------------------------------------------------------------------------- - ----- > These are my opinions---Nortel Networks takes no responsibility for them. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: HELP - access s/w through the Tornado IDE Date: Mon, 23 Sep 2002 08:55:58 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <64ddf295.0209200704.77303399@posting.google.com> Reply-To: "Michael Lawnick" Sender: mlawnick@pd9524384.dip.t-dialin.net Hi Val, There are lots of examples under $(WINDBASE)/host/resource/tcl so dig a bit ! Regard, Michael "val" schrieb im Newsbeitrag news:64ddf295.0209200704.77303399@posting.google.com... > Hello! > Please help me with: > I'm trying to create access to my s/w (drl.exe) through the Tornado > IDE !! > I didn't find much help in the Tornado API Programmer's Guide 2.0 > (chapter6), "Adding Tools to Tornado" (I also little bit new to Tcl). > Maybe somebody has some example of that ... > Thank you for any suggestions, examples, any help ! > Val --------------------------- Newsgroups: comp.os.vxworks Subject: how to create a bootable VxWorks CDROM Date: Mon, 23 Sep 2002 12:57:17 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1032778666.293601@newsreader1.wirehub.nl> Hi, Two questions.. Is there support for ATA/IDE CDROM-drives in VxWorks? It lookslike there is only support for SCSI CDROMs, it this correct or should ATA/IDE be supported as well? And is it possible to put the bootloader on a CDROM and how should I do that?? Thanx, Renee. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: how to create a bootable VxWorks CDROM Date: Mon, 23 Sep 2002 13:16:51 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <1032778666.293601@newsreader1.wirehub.nl> Reply-To: "Michael Lawnick" Sender: mlawnick@pd9524384.dip.t-dialin.net Hi Renee, ATAPI CD-Rom support is mentioned in T2.2, don't know whether it will work under T2.0.x I assume you have a PC as a target, as you want to load the bootloader from CD. Haven't tried, but it should work as if you are building a DOS-bootable CD. IMO you will have to exchange the boot block by the block of vxworks (-->mkboot.c in BSP dir) and have to burn a contigues bootrom.sys on the CD. - Again: I haven't tried yet and assume everyone would be pleased to hear from your results ;-) Regards, Michael "Renee Teunissen" schrieb im Newsbeitrag news:1032778666.293601@newsreader1.wirehub.nl... > > > Hi, > > Two questions.. > > Is there support for ATA/IDE CDROM-drives in VxWorks? It lookslike > there is only support for SCSI CDROMs, it this correct or should > ATA/IDE be supported as well? > > And is it possible to put the bootloader on a CDROM and how should I > do that?? > > Thanx, > Renee. > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Fastest way to transfere data between a PC with win2000 and a PC with vxWorks? Date: Mon, 23 Sep 2002 13:50:34 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <7777af4e.0209211346.1b507f81@posting.google.com> Mike, with some tuning and optimized drivers, I was able to get in excess 70Mbps transfer rate with FTP. There is nothing in the FTP protocol to make it inherently slow, it is only a matter of implementation. "Michael Lawnick" wrote in message news:ammda8$drs$1@snoopy.bndlg.de... > Hi, > > as FTP in my experience is an easy to handle but slow protocoll, I'd suggest > to write your own transfer protocoll based on UDP packets. > The benefit might improve your performance from about 50-60% upto 90-95% > (i.e. ~100-120 kByte/s) :-)) > The down part is that you are out of any standards :-( > > Regards, > Michael > > "Ake" schrieb im Newsbeitrag > news:7777af4e.0209211346.1b507f81@posting.google.com... > > Hello! > > We may start a new project where we need to move a lots of between two > > PC one with win2000 and one with VxWorks. My ide is to set up a > > 100Mbits/s ethernet link and transfer data with FTP. Is there a faster > > way to do this. (instead of using Ehernet) > > > > The PC with win2000 will be responsible to move the files. Note no > > other computers will use this cable. > > > > I hope there is support for 100 Mbits/s Ethernet in vxWorks, Is it? > > > > On the same Ethernet connection I like to send commands between the > > two PC, what protocol should I use then? > > --------------------------- Newsgroups: comp.os.vxworks Subject: Printf Vs. Fprintf Date: 23 Sep 2002 05:00:34 -0700 From: nirn@rit.co.il (Nir N) Organization: http://groups.google.com/ Message-ID: <4913ea0b.0209230400.1d99cff0@posting.google.com> Hi, Printf is an UnBuffered IO command and it's a know fact that the VxWorks machine is Pednig when trying to Print. The Fprintf is Buffered IO command does it mean that the VxWorks is not pending on it when printing ? Any Help will Do.... Best Regards Nir N --------------------------- Newsgroups: comp.os.vxworks Subject: netMblkToBufCopy and copy penalty Date: Mon, 23 Sep 2002 15:12:32 +0200 From: "Ofer Goren" Organization: Verio Message-ID: - ------------------------ Win2k Sp3 Tornado 2.02 for PPC Diab 4.4b Envoy 9.2 - ------------------------ Hi. I have a question regarding netMblkToBufCopy(). I understand that since a packet can be spawned on multiple MBlks, this is one way to get the packet data into my own buffer. The problem is, that this function copies the packet from the Mblks to the buffer, a process I can't efford (time considerations). Is there away to extract the whole packet from the mblks be reference? meaning, a pointer to the data, instead of a buffer contining the data? Is there a way to FORCE vxWorks to put the entire packet in a single MBlk? Thanks, - -- Ofer Goren Iamba Technologies --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Printf Vs. Fprintf Date: Mon, 23 Sep 2002 14:22:36 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <4913ea0b.0209230400.1d99cff0@posting.google.com> Nir, fprintf() will PEND as well, but it will do so less frequently for longer periods of time, because the buffered fprintf() will call write() with typically larger buffer size (e.g. 1024 bytes), it will take it longer to fill up the buffer and call write() - thius less frequently, but will PEND for longer because it will need to wait longer untill 1024 bytes become available in the output buffer of the serial driver. "Nir N" wrote in message news:4913ea0b.0209230400.1d99cff0@posting.google.com... > Hi, > > Printf is an UnBuffered IO command and it's a know fact that the > VxWorks machine is Pednig when trying to Print. > > The Fprintf is Buffered IO command does it mean that the VxWorks is > not pending on it when printing ? > > > Any Help will Do.... > > Best Regards > Nir N --------------------------- Newsgroups: comp.os.vxworks Subject: Re: how to create a bootable VxWorks CDROM Date: Mon, 23 Sep 2002 14:31:22 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1032784312.169224@newsreader1.wirehub.nl> References: <1032778666.293601@newsreader1.wirehub.nl> Hi Michel, Well I will look into that. In the mean while I have created a bootfloppy, and made a bootable CDROM with that disk as base. Nero has an option to build a bootable CDROM of an bootable floppy. This works on older targets only, not ona newer systemsboards.. Cheers, Renee - - - - - - - - - Renee A. Teunissen PTS Software bv, Meerweg 7, 1405BA Bussum, NL. T.+31-(0)35-6926969, http://www.pts.nl "Michael Lawnick" wrote in message news:ammt72$c8g$1@snoopy.bndlg.de... > Hi Renee, > > ATAPI CD-Rom support is mentioned in T2.2, don't know whether it will work > under T2.0.x > > I assume you have a PC as a target, as you want to load the bootloader from > CD. > Haven't tried, but it should work as if you are building a DOS-bootable CD. > IMO you will have to exchange the boot block by the block of vxworks > (-->mkboot.c in BSP dir) and have to burn a contigues bootrom.sys on the > CD. - Again: I haven't tried yet and assume everyone would be pleased to > hear from your results ;-) > > Regards, > Michael > > "Renee Teunissen" > agree.with.these.terms@mail.zevenaar.org> schrieb im Newsbeitrag > news:1032778666.293601@newsreader1.wirehub.nl... > > > > > > Hi, > > > > Two questions.. > > > > Is there support for ATA/IDE CDROM-drives in VxWorks? It lookslike > > there is only support for SCSI CDROMs, it this correct or should > > ATA/IDE be supported as well? > > > > And is it possible to put the bootloader on a CDROM and how should I > > do that?? > > > > Thanx, > > Renee. > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Fastest way to transfere data between a PC with win2000 and a PC with vxWorks? Date: Mon, 23 Sep 2002 14:38:31 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <7777af4e.0209211346.1b507f81@posting.google.com> Reply-To: "Michael Lawnick" Sender: mlawnick@pd9524384.dip.t-dialin.net Hi Leonid, > > (i.e. ~100-120 kByte/s) :-)) it was too early in the morning .... cancel 100-120 kByte/s, I meant 10-12 MByte/s or 80-96 Mbps. > There is nothing in the FTP protocol to make it inherently slow, > it is only a matter of implementation. In the first step the implementation is given by object and RFC ... Hm... I go offline, as the ice is getting too thin for me, its below register level ;-) Regards, Michael --------------------------- Newsgroups: comp.os.vxworks Subject: Tcl for Tornado Date: 23 Sep 2002 06:33:21 -0700 From: valrub@hotmail.com (val) Organization: http://groups.google.com/ Message-ID: <64ddf295.0209230533.cfd71@posting.google.com> Hello everybody, I'm trying to create access to our program executable (drl.exe) through the Tornado IDE. I can create button on the Tornado IDE with the next line of Tcl code: controlCreate launch [list toolbarbutton \ -callback launchWtxapp \ -tooltip "Launch Tilcon Graphic Editor" \ -bitmap "[wtxPath host resource bitmaps launch controls]TGE.bmp"] but I don't know how to connect that button to the executable (drl.exe). Maybe some ideas ... Thank you! Val --------------------------- Newsgroups: comp.os.vxworks Subject: Re: New Object appear at NULL Date: 23 Sep 2002 06:40:58 -0700 From: bwoodhouse@quantumdata.com (Bob W) Organization: http://groups.google.com/ Message-ID: <705be589.0209230540.491ff352@posting.google.com> References: <705be589.0209221216.73ce6f81@posting.google.com> Whoa -- let's not think too hard about this one now. It turned out that there was another little memory manager that I had ported in that was overriding the new operator with code that did not quite work. That was the benefit of getting ahold of the emulator today. Sorry for any heavy mental investments that just came to a screeching halt. Best, BobW --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ADS8260 BSP Date: Mon, 23 Sep 2002 16:46:08 +0200 From: "Achim Zimmer" Message-ID: References: Thank you Dan, you are right, all Ethernet buffers and descriptors are in RAM, not DPRAM. But there was an INIT failure in m8260sio.c for the UART (served at an SCC) that has crashed the CP. Thanks. "Dan Gold" schrieb im Newsbeitrag news:amelo6$nm3$1@newsreader.mailgate.org... > Achim Zimmer wrote in message ... > >Yes, thats is right. This way works to figure out where the tables and > >buffers are. > > > Did you examine it ? where are the buffer descriptor tables, in DPRAM or > external memory ? > If they are in external memory there is nothing to move. If they are in > DPRAM then you need to move them. > > Your main problem is you need the DPRAM starting at the base of bank 1. > First, are you sure something is using that memory ? If so, how do you know > ? What happens if you zero out that memory while the system is running ? > Or can you dump it and see if values are changing ? Other candidates that > may use that space are the buffer descriptors for the serial port (i.e. the > SCC, since they cannot reside in external memory). You can look at the > serial driver where it initializes it's BDs and see where it is putting > them. > > It sounds like you need to do 1 and/or 2 things: > 1. Find some documentation on the BSP that describes where everything is > placed in DPRAM. This may or may not exist. And you may have to look in a > header file for it. > 2. Look through all the code in the BSP for the various peripherals you are > using (FCC, SCC, anything else like I2c or SPI, etc...) and see where they > place their stuff in DPRAM. > > >There more complicated problem is to move it, because WRS uses a very > tricky > >init string at ADS8260. > > Again, you don't have to move it unless you find out the BDs are in DPRAM. > Make sure of that first before dealing with the init string. > > >I have posted this initfunction from WRS, because my english isn't so good > >to explaine it. > > > >I changed this motFccEndParamTemplate to > > "0x%x:0x%x:0x2100:0x300:-1:-1:-1:-1:0x%x:0x%x:0x%x:0x%x:0x%x:0x%x"; > >but the result was that the MUX failed to load the device. > > > what did you change it from ? > > >The problem is, that I dont understand exactly the usage of the template > >fields. > The END device load string formed by this routine is in the following > format. > >* :::::: > >* ::::::: > >* > >* > >* .IP > >* buffer descriptors base address > > this is the parameter you want to change. If it's 0, it probably defaults > to something inside the driver, but I don't know what. Maybe call WRS and > ask for motFcc.c if you need to. > > >* .IP > >* tx buffer in internal memory > >* .IP > >* rx buffer in internal memory > > If the above 2 fields are in an area you need, you'll need to move them too. > > The bottom line is find out if anything is interfering with the DPRAM you > need, and what it is first. > > hth, > Dan Gold gold@ensemble.com > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: symTbl.c vs. ELF .stab Date: 23 Sep 2002 10:54:05 -0400 From: "Paul D. Smith" Organization: Nortel Networks, Inc. / Billerica, MA Message-ID: References: Reply-To: pausmith@nortelnetworks.com %% "Michael Lawnick" writes: ml> finally: it works well, so where is the beef? I guess if your link only takes a few seconds, there's no beef. But if your link takes minutes (like ours), and you double that time in order to build this C symbol table every single time anyone builds and you have lots of people (like we do), that's hundreds of links a day times 2 minutes or whatever that could be saved--that adds up to significant developer time pretty fast... If you could get the same behavior but halve the link time, that seems like it would be a win. - -- - ------------------------------------------------------------------------------- Paul D. Smith HASMAT--HA Software Mthds & Tools "Please remain calm...I may be mad, but I am a professional." --Mad Scientist - ------------------------------------------------------------------------------- These are my opinions---Nortel Networks takes no responsibility for them. --------------------------- Newsgroups: comp.os.vxworks Subject: Help with lists Date: 23 Sep 2002 08:24:08 -0700 From: derfan2001@altavista.de (Mario Schattschneider) Organization: http://groups.google.com/ Message-ID: <13dfb019.0209230724.7cf539be@posting.google.com> Hello I need some help with linked lists under vxWorks. I want to create a linked list and put some data into it. I wrote some code in C++ - it works but I don't know if my implementation is right and how I get the data from the listelements. Here is my code At first I created a class #ifndef _MYNODE_H #define _MYNODE_H #include class Mynode { public: NODE node; int line; int taskId; char* fi; /*Constructor*/ Mynode(char *filename, int linenumber ,int tID); /* Destructor */ ~Mynode(); }; #endif The code for the constructor and the destructor are in another cpp file Here is the code from the file: ... LIST myList; Mynode::Mynode(char* filename, int linenumber ,int tID) { fi = new char[100]; fi = filename; line = linenumber; taskId = tID; } Mynode::~Mynode() { delete fi; } void addDataToList() { Mynode *node = new Mynode(file, line, taskid); lstAdd(&myList, (NODE*)&node); test = lstCount(&myList); } Where is my mistake and how can a I get the data from the elements - or is that a wrong implementaion to create and fill up a list? Thanks for your help Best regards Mario Schattschneider --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Newbie Starter Kit Date: Mon, 23 Sep 2002 15:31:00 GMT From: "harry mudd" Organization: AT&T Broadband Message-ID: References: here's a good link http://www.onesmartclick.com/rtos/rtos.html "Blackie Beard" wrote in message news:sA8j9.6967$7i2.195@nwrddc02.gnilink.net... > ============================== > I want to learn vxWorks in my spare time. > How do I get started for very cheap? > Is there a good kit, and what books are > recommended. My background is H/W, > S/W, F/W and SoC, but never studied > any OS. Thanks in advance! > > BB > ============================== > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ADS8260 BSP Date: Mon, 23 Sep 2002 08:54:13 -0700 From: "Dan Gold" Organization: Mailgate.ORG Server - http://www.Mailgate.ORG Message-ID: References: Sender: gold@ensemble.com "Achim Zimmer" wrote in message news:amn9ia$7du2i$1@ID-102608.news.dfncis.de... > Thank you Dan, > > you are right, all Ethernet buffers and descriptors are in RAM, not DPRAM. > > But there was an INIT failure in m8260sio.c for the UART (served at an SCC) > that has crashed the CP. So is this the main problem, or just something else you've discovered ? I haven't looked but I would guess that the SCC puts it's BDs (and possibly buffers) in the DPRAM in bank 1. You'll need to look at m8260sio.c for the details. Another thing to consider, since this entire exercise is so you can use the MCC. What version of the 8260 are you using ? Will you need to run the RAM-based microcode patches ? If so, you can't use the first 32 (I believe) channels of MCC1 because the microcode patch needs to go there. So you still need to make sure the first part of DPRAM is free (potentially for the microcode patch), but then you can start with MCC channel 64. Maybe the DPRAM isn't used that far into the bank. hth, Dan Gold gold@ensemble.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Booting from Flash Disk Date: Mon, 23 Sep 2002 17:10:09 +0100 From: "Paul" Organization: Logica UK Ltd Message-ID: <1032797410.457536@ernani.logica.co.uk> References: <1032441796.250754@ernani.logica.co.uk> <4d0e703e.0209221654.3e0919ae@posting.google.com> "Dennis" wrote in message news:4d0e703e.0209221654.3e0919ae@posting.google.com... > "Leonid Rosenboim" wrote in message news:... > > Yes Paul, > > > > You can do that, but I am afraid that loading the other stuff after > > the kernel comes up does too comsume time, and you might > > consider a possible outcome that the overall process will > > consume more time than the current scheme, rendering > > your efforts useless. > > > > Unless you are doing this for educational purposes, > > which is an excellent exercise, I must admit. > > Stripping down the VxWorks kernel will make you > > learn a lot about VxWorks and its components, > > a truly worthwhile experience indeed ! > > > > - Leonid > > Well it sort of works. I've built a vxWorks_rom image with a built in symbol table (and very little else!). I can boot straight into the kernel from floppy (for now) without the intermediate bootrom stage. The only problem is that the symbol table seems to have vanished. It tries to initialise it but reports 'Adding 0 symbols for standalone'. I've checked the symTbl.c file and everything seems ok there. I can't see where standTblSize is being reset back to zero. Any suggestions as there is little I can do on the target without a symbol table, and, of course, I had to strip out WDB to get the image to fit. Cheers. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: semTake()/exit()/taskSpawn() Date: Mon, 23 Sep 2002 13:14:35 -0400 From: Keith Arner Organization: Marconi Message-ID: References: <313680C9A886D511A06000204840E1CF67A1A0-100000@whq-msgusr-02.pit.comms.marconi.com> <313680C9A886D511A06000204840E1CF67A1A2-100000@whq-msgusr-02.pit.comms.marconi.com> Reply-To: Keith Arner On Fri, 20 Sep 2002, Leonid Rosenboim wrote: > "Keith Arner" wrote in message > news:Pine.GSO.4.42.0209200813330.1626-100000@zimbra... > > [snip] > > In VxWorks, tasks are identified by a task id. When one task exits, its > > task id becomes available for reuse. In VxWorks 5.4, if you spawn a task > > immediately after some task has exited, the new task is fairly likely to > > get the task id of the task that just exited. > > The taskId is simply the address of the TCB (Task Control Block) in memory, > and this behaviour, correctly described here, is due to the memPartLib > which looks for new blocks to allocate beginning with the atest free'd one. > The only reason this has not been solved yet, is lazyness. True. But even if the implementation of taskSpawn() changes, you'll still need to worry about the potential of re-used task ids. [ snip ] > > At any given time, a mutex semaphore is "owned" by at most one task. What > > happens if a task takes the semaphore, forgets to give it up, then exits? > > The next task to come along and accidentally get the same task id will > > also accidentally own the mutex semaphore. The root problem in your code > > would be the task that exited while owning the semaphore, but the only > > symptom you see is a completely unrelated task owning the semaphore. When > > presented with that symptom, how does one divine what the root problem > > was? > > This is completely wrong. See semMLib reference manual, here is a quote: > > SEMAPHORE DELETION Who's talking about deleting semaphores? I'm talking about deleting tasks. [ snip ] > > message was queued. If a task queues a message, then exits and another > > task is spawned before the message gets printed out, the message is > > printed out with the task name of the new task, rather than the old task. > > When debugging the problem indicated by your error message, you are > > debugging based on misleading evidence. > > > Quote true, although not very critical. I've been stung by this enough that I worry about it. Your milage may vary. [ snip ] > Or, one could simply work around the taskId reuse issue. Yes, there are > several ways to work around this, some actually implemented by customers I > have advised. Working around the current task id reuse is going to involve working around the current implementation (which is not necessarily portable to future releases of VxWorks), or re-inventing the task id wheel. Both of these are things which you'd probably like to avoid. But task id reuse is really only the tip of the iceberg. The more your system changes at run time, the harder it is to debug. Because tasks are such a core component in the system design, it's advisable to keep the set of tasks as simple and understandable as possible. Having a dynamically changing task list often makes it much more difficult to get your head around the state of the system when things go wrong. Keith - -- "Once something has gone horribly wrong..." "...it just gets horribler and horribler." "Then, it's a matter of finding the original horriblism." -- from a conversation on debugging problems in a linked list --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Help with lists Date: 23 Sep 2002 13:03:58 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: References: <13dfb019.0209230724.7cf539be@posting.google.com> Sender: bpringle@DeadDuck class Mynode { public: NODE node; int line; int taskId; [snip] void addDataToList() { Mynode *node = new Mynode(file, line, taskid); lstAdd(&myList, (NODE*)&node); test = lstCount(&myList); } Mario> Where is my mistake and how can a I get the data from the Mario> elements - or is that a wrong implementaion to create and fill Mario> up a list? Your mistake is "lstAdd(&myList, (NODE*)&node);". Do you know how a compiler implements things in C++? It will shove a vtable, rtti, and other information at the start of the "class Mynode" structure. If you have the Tornado documents, look for the "WindFoundation classes". The URL might be, "file:///C|/Tornado2/docs/vxworks/ref/VXWList.html" Note the explicit warnings about doing what you have done. vxWorks is expecting a structure with specific memory layout. This is abstracted away by C++ and some techniques that are available to users of "struct" are not available to users of "class". Unfortunately there is no language that allows "classes" with fixed memory mappings. See the book, _Inside the C++ object model_ by "Stanley B. Lippman" for more information, or the Internal g++ documentation. btw, the include should be in your vxWorks header directory. hth, Bill Pringlemeir. - -- Many people will ask you for advice about asking for a raise. Listen to your dinoflagellate tomorrow. Your cat will be fluctuating in the next 1 or 7 years. Your efforts at cleaning the fridge turn out to be quite worthwhile. You'll often be happy. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: vxworks watch dog Date: 23 Sep 2002 10:57:59 -0700 From: alex.mo@attbi.com (alex mo) Organization: http://groups.google.com/ Message-ID: References: I don't think there is any problem with your WD, can you call the function directly from the shell to see if your task gets the message? Also, can we see the code for trec_CheckUnitTimeOut func. thanks, - --alex Shlomo Katz wrote in message news:... > hi > i am using vxworks watch utility and found some difficulties > > i have the follwing code > wd = wdCreate(); > status = wdStart(wd,(5_SEC), (FUNCPTR)trec_CheckUnitTimeOut,NULL); > when > trec_CheckUnitTimeOut - send a message to a task and have no > parameters but i do not get the message > > does someone have any idea why not > the task that should receive the message is ok and work fine with other > messages sent to it > > thanks > katz shlomi > www.rit.co.il > > ************************************************************************************************** > Privacy Notice: The information contained in this transmittal, including any > attachments hereto, are confidential and privileged, and intended solely for the > specified addressee(s). If you are not the intended addressee, or if you receive this > message by error, please notify the sender and delete this information from your > computer. > > Moreover, the recipient(s) may not disclose, forward, or copy this e-mail or > attachments, or any portion thereof, or permit the use of this information, by anyone > not entitled to it, or in a way that may be damaging to the sender. > > ************************************************************************************************** --------------------------- Newsgroups: comp.os.vxworks Subject: Re: semTake()/exit()/taskSpawn() Date: Mon, 23 Sep 2002 20:36:24 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <313680C9A886D511A06000204840E1CF67A1A0-100000@whq-msgusr-02.pit.comms.marconi.com> <313680C9A886D511A06000204840E1CF67A1A2-100000@whq-msgusr-02.pit.comms.marconi.com> "Keith Arner" wrote in message news:Pine.GSO.4.42.0209231300020.1626-100000@zimbra... > On Fri, 20 Sep 2002, Leonid Rosenboim wrote: [snip] > > > > This is completely wrong. See semMLib reference manual, here is a quote: > > > > SEMAPHORE DELETION > > Who's talking about deleting semaphores? I'm talking about deleting > tasks. > Unfortunate heading, but try reading the section underneeth, it specifically refers to protection against deletion of a task while it owns a Mutex semaphore, in order to avoid the exact probloem you where referring to. [snip] > Working around the current task id reuse is going to involve working > around the current implementation (which is not necessarily portable to > future releases of VxWorks), or re-inventing the task id wheel. Both of > these are things which you'd probably like to avoid. > This is where I draw the line, not going to discuss any specific workaround on this NG, but no workaround is really a replacement for a solution. > But task id reuse is really only the tip of the iceberg. The more your > system changes at run time, the harder it is to debug. Because tasks are > such a core component in the system design, it's advisable to keep the set > of tasks as simple and understandable as possible. Having a dynamically > changing task list often makes it much more difficult to get your head > around the state of the system when things go wrong. Yes, I agree. I too prefer to design a system where the list of tasks is more or less static, but I would not rule out dynamic tasks in certain situation, similarly I would not rule out the use of "goto" statements entirely. Every tool has its specific application. Cheers, - Leonid --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Help with lists Date: Mon, 23 Sep 2002 22:27:46 +0200 From: Mario Schattschneider Message-ID: <43uuou8ftlsikhh2c3esosttovhjq6fi9c@4ax.com> References: <13dfb019.0209230724.7cf539be@posting.google.com> Bill Pringlemeir wrote: >Your mistake is "lstAdd(&myList, (NODE*)&node);". Do you know how a >compiler implements things in C++? It will shove a vtable, rtti, and >other information at the start of the "class Mynode" structure. If >you have the Tornado documents, look for the "WindFoundation classes". >The URL might be, > > "file:///C|/Tornado2/docs/vxworks/ref/VXWList.html" > >Note the explicit warnings about doing what you have done. vxWorks is >expecting a structure with specific memory layout. This is abstracted >away by C++ and some techniques that are available to users of >"struct" are not available to users of "class". Unfortunately there >is no language that allows "classes" with fixed memory mappings. > Hi Bill, is it better to implement the list as a struct? I'll look to the VXWorks documentation. Best regards Mario --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Help with lists Date: 23 Sep 2002 17:23:20 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: References: <13dfb019.0209230724.7cf539be@posting.google.com> <43uuou8ftlsikhh2c3esosttovhjq6fi9c@4ax.com> Sender: bpringle@DeadDuck >Your mistake is "lstAdd(&myList, (NODE*)&node);". Do you know how a >compiler implements things in C++? It will shove a vtable, rtti, and >other information at the start of the "class Mynode" structure. If >you have the Tornado documents, look for the "WindFoundation classes". >The URL might be, > > "file:///C|/Tornado2/docs/vxworks/ref/VXWList.html" > >Note the explicit warnings about doing what you have done. vxWorks is >expecting a structure with specific memory layout. This is abstracted >away by C++ and some techniques that are available to users of >"struct" are not available to users of "class". Unfortunately there >is no language that allows "classes" with fixed memory mappings. Mario> is it better to implement the list as a struct? I'll look to Mario> the VXWorks documentation. /* -*- mode: c++ -*- */ #ifndef _LIST_H_ #define _LIST_H_ #include #include /** @class CListItem * */ class CListItem { public: CListItem(); virtual ~CListItem(); virtual bool IsA(int _classID) const = 0; protected: struct SNode { NODE node; class CListItem *m_Data; } * m_Node; static CListItem *getItem(NODE *p); friend class CList; private: static const int baseID = 0; // A local implementation of RTTI. }; /** @class CList * */ class CList { public: CList(); virtual ~CList() = 0; protected: /** * List manipulation routines. */ CListItem * first () const; CListItem * last () const; CListItem * next (CListItem *pNode) const; CListItem * previous (CListItem *pNode) const; void add (CListItem *pNode); void remove (CListItem *pNode); int count (void) const; /** * Class data. */ LIST list_; // vxWorks doubly linked list structure. }; #endif Here is an example of add, ////////////////////////////////////////////////// // CList::add - add a node to the end of list // void CList::add (CListItem *pNode) { if(pNode && pNode->m_Node) { lstAdd (&list_, &pNode->m_Node->node); } } CListItem::CListItem() : m_Node(0) { m_Node = static_cast(malloc(sizeof(SNode))); if(m_Node) { m_Node->m_Data = this; } } CListItem::~CListItem() { if(m_Node) { free(m_Node); m_Node = 0; } } You can derive from these classes. They don't use templates, so they are quite a small implementation of a container. I am sure that most C++ zealots will find them repugnant and several News group members will cast shame on me for not providing you with the full source and flying to your locale to help you compile the code. Compile with "-fno-exceptions -fno-rtti" if you are actually concerned with code size (and speed). The only trick is to keep a link between the "structure and the class". The line "m_Node->m_Data = this" is part of this. GetItem() is something like "return p->m_Data;", although you might like to use RTTI or you own typing mechanism. I think I have helped a lot, but you have probably learned little. Regards, Bill Pringlemeir. - -- Do not drive this week. You will meet God this week. Your endothermic therapsid will need to be petted this week. You'll soon be happy. --------------------------- Newsgroups: comp.os.vxworks Subject: Suspended task debugging Date: Mon, 23 Sep 2002 16:06:44 -0700 From: "MK" Organization: mike_texx@mail.ru Message-ID: Reply-To: "MK" I run into a problem, where my system would break down, and when I connect a debugger to it I notice that one of the tasks is suspended. The place at which it is suspended seems totally harmless though. Looking through the posts on this group I found out that apparently VxWorks will suspend an active task if there is a bus error in the interrupt handler, and there isn't anything wrong being done by the task... I am in need of suggestions, how do I debug this problem? If we assume that the bug is in the driver code, is there a way to find out the address of the command which actually caused the bus error? --------------------------- Newsgroups: comp.os.vxworks Subject: The mib2Tbl of END Driver seems doesn't take effect Date: 23 Sep 2002 17:42:37 -0700 From: g_ch@21cn.com (GANGCHENG) Organization: http://groups.google.com/ Message-ID: Dear all, I've finished an virtual END driver. When it receives packets or sends packet, it calls mib2ErrorAdd to count errors in the mib2Tbl of END Driver if error occors. But when I calls ifShow() or snmp agent gets the information of IF group, the count in the mib2Tbl of END driver can not be gotten. I am maybe sure that the END_MIB_INIT is successful on the xxEndLoad(). Why doesn't it work? B.T.W when the snmp agent gets the information of IF group, I am also sure that xxEndIoCtl has been called to get MIB2 information from my virtual END driver. Best Regards --------------------------- Newsgroups: comp.os.vxworks Subject: resolver Date: 23 Sep 2002 19:36:57 -0700 From: vijay.peshkar@wipro.com (Vijay Peshkar) Organization: http://groups.google.com/ Message-ID: Hi, Am trying to use VxWorks provided resolver on a PC based target. The DNS resolver component is selected and the domain & server IP are configured as in example. First am running resolvInit and resolvParamsSet as in code below. if (ERROR == resolvInit( pResolvParams->nameServersAddr[0], pResolvParams->domainName, 0 )) { printf("DNS: ERROR: resolvInit() failed\n"); return ERROR; } if (ERROR == resolvParamsSet (pResolvParams) { printf("DNS: ERROR: resolvParamsSet() failed\n"); return ERROR; } Have added my target's IP as nameserver in a linux machine's resolv.conf. Then running resolvGetHostByName() from shell. The resulting debug message on shell is as below. Now if the resolvSend is going thru, what's failing?. - -> resolvGetHostByName "VxWorksTarget.mynet.wipro.com",0x76ff828,512 resolvSend() ;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 1 ;; flags: rd; Ques: 1, Ans: 0, Auth: 0, Addit: 0 ;; QUESTIONS: ;; VxWorksTarget.mynet.wipro.com, type = A, class = IN resolvSend() ;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 2 ;; flags: rd; Ques: 1, Ans: 0, Auth: 0, Addit: 0 ;; QUESTIONS: ;; VxWorksTarget.mynet.wipro.com.mynet.wipro.com, type = A, class = IN value = 0 = 0x0 - -> Thanks and Regards, Vijay --------------------------- Newsgroups: comp.os.vxworks Subject: Re: binary redistributable Date: 23 Sep 2002 22:27:53 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: References: Sender: bpringle@DeadDuck Matt> So far my only success is to have made a vxWorks image with my Matt> code inserted (via ld) in the image. Yet vxWorks has no Matt> knowledge of where my code resides, or what the symbols are. I assume that you are trying to make a 3rd party library? 1) You can require that the user insert a function call to your code. 2) Another way would be to compile the code as "C++" and use a static object to get the build process to call the static constructor of your object "auto-magically". From this routine, it should be possible to spawn other tasks, etc. I am sure that there are other imaginative ways to do this as well. hth, Bill Pringlemeir. - -- Baranyi Belknap Firewalls keyhole BRLO CBNRC SAFE Europol DES bootleg White Water UOP security analyzer Becker --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Suspended task debugging Date: Tue, 24 Sep 2002 09:33:57 +0200 From: "alexv" Organization: Verio Message-ID: References: It is not nesessary bus error in interrupt handler ( in most cases such error would halt your system at all ). I think of some possible reasons: 1) stack overflow of some other task ( could be checked by Browser ) 2) memory corruption. You can check this assumption by changing your memory arrangement by changing task stack size , or adding some harmless commands ( for example printf ). This may cause your problem to appear in other place or disappear temporarilty ) Check also if your problem appears always at the same time and same instruction. H/W errors may appear randomly. Alex - ----- Original Message ----- From: "MK" Newsgroups: comp.os.vxworks Sent: Tuesday, September 24, 2002 1:06 AM Subject: Suspended task debugging > I run into a problem, where my system would break down, and when I connect a > debugger to it I notice that one of the tasks is suspended. The place at > which it is suspended seems totally harmless though. Looking through the > posts on this group I found out that apparently VxWorks will suspend an > active task if there is a bus error in the interrupt handler, and there > isn't anything wrong being done by the task... I am in need of suggestions, > how do I debug this problem? > > If we assume that the bug is in the driver code, is there a way to find out > the address of the command which actually caused the bus error? > > --------------------------- Newsgroups: comp.os.vxworks Subject: Ada in VxWorks Date: Tue, 24 Sep 2002 10:03:30 +0200 From: "Marcus Hedlund" Organization: ericsson Message-ID: Hi! I have managed to integrate GNAT pro with VxWorks and am trying to run a few ada programs. However I can only run the same Ada program once. After trying to run it twice, an exception is raised and the program hangs. Anyone had this kind of problem? Any help would appreciated, /M. Hedlund --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Ada in VxWorks Date: Tue, 24 Sep 2002 10:05:51 +0200 From: "Martin Raabe" Organization: 1&1 Internet AG Message-ID: References: Hello Marcus, there is a GNAT pro integration to Tornado and there is also one for SNiFF+. Maybe you want to take a look into that? HIH - -- Martin Raabe E: Martin.RaabeATNOSPAMBaSystem.de "Marcus Hedlund" schrieb im Newsbeitrag news:amp5s2$mki$1@newstoo.ericsson.se... > Hi! > > I have managed to integrate GNAT pro with VxWorks and am trying to run a few > ada programs. > However I can only run the same Ada program once. After trying to run it > twice, an exception is raised and the program hangs. > Anyone had this kind of problem? > > Any help would appreciated, > > /M. Hedlund > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: symTbl.c vs. ELF .stab Date: Tue, 24 Sep 2002 08:24:47 +0000 (UTC) From: David Laight Organization: BT Openworld Message-ID: References: Paul D. Smith wrote: > %% "Michael Lawnick" writes: > > ml> finally: it works well, so where is the beef? > > I guess if your link only takes a few seconds, there's no beef. But if > your link takes minutes (like ours), and you double that time in order > to build this C symbol table every single time anyone builds and you > have lots of people (like we do), that's hundreds of links a day times 2 > minutes or whatever that could be saved--that adds up to significant > developer time pretty fast... > > If you could get the same behavior but halve the link time, that seems > like it would be a win. > I'd look at the script used to extract the names. It is probably TCL - about the slowest way possible.... Certainly the one used to get the C++ constructors is hideously slow - and can be sped up significantly by a little twiddleing. I actually used a fixed (hand build) C symbol table that only contained the names that were likely to be useful. That makes the target shell slightly safer. If you fiddle with the code that creates the symbol table it is possible to a function that will download the conplete table over the network - when you need all of it. Building an image on a 40MHz sparc system definitely showed up the slow spots :-) David --------------------------- Newsgroups: comp.os.vxworks Subject: Module loading problems Date: 24 Sep 2002 01:57:44 -0700 From: san_blabla@hotmail.com (San Blabla) Organization: http://groups.google.com/ Message-ID: <6046c736.0209240057.2a686e88@posting.google.com> Hi, everyone: I have writen an application on vxWorks, when I link it into the system image, everything is O.K. But when I don't link it into my system image and try to load it as a module using ld() function, the error message "Undefined symbol: get_myaddress" occured. I found that this function is in a vxWorks library, so I guess it is because that I didn't link my application object to the system image, the linker could not know anybody would call get_myaddress() so it didn't link in the corresponding object in the library. I thought two ways to solve this problem, one is to write some dummy functions to call get_myaddress() in the system image, the other one is to link the object from the library with my application object by hand. Both way seem to be kind of ugly. Can anybody tell me how to solve this problem in a better way? Regards, Blabla. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Ada in VxWorks Date: Tue, 24 Sep 2002 11:06:53 +0200 From: "Marcus Hedlund" Organization: ericsson Message-ID: References: Hi Martin, I have already managed to integrate the adacompiler. It is just that when I run Ada program in the VxWorks windshell, I can only run it once. If I then try to run it again without rebooting, it will not work. I hope this clarifies my question /Marcus "Martin Raabe" wrote in message news:amp6cq$gtm$1@news.online.de... > Hello Marcus, > there is a GNAT pro integration to Tornado and there is also one for SNiFF+. > Maybe you want to take a look into that? > > HIH > -- > Martin Raabe > E: Martin.RaabeATNOSPAMBaSystem.de > > "Marcus Hedlund" schrieb im Newsbeitrag > news:amp5s2$mki$1@newstoo.ericsson.se... > > Hi! > > > > I have managed to integrate GNAT pro with VxWorks and am trying to run a > few > > ada programs. > > However I can only run the same Ada program once. After trying to run it > > twice, an exception is raised and the program hangs. > > Anyone had this kind of problem? > > > > Any help would appreciated, > > > > /M. Hedlund > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: resolver Date: 24 Sep 2002 04:26:36 -0700 From: vijay.peshkar@wipro.com (Vijay Peshkar) Organization: http://groups.google.com/ Message-ID: References: Few more points am seeing.. hostAdd was used to add 'VxWorksTarget'. The debug message of api resolvGetHostByAddr displays wrong ip 86.120.87.111 instead of 10.114.5.152. Any explanations?. Thanks and Regards, Vijay - -> _resolvGetHostByAddr "VxWorksTarget.mynet.wipro.com",0x7701cd0,512 resolvSend() ;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 20 ;; flags: rd; Ques: 1, Ans: 0, Auth: 0, Addit: 0 ;; QUESTIONS: ;; 111.87.120.86.in-addr.arpa, type = PTR, class = IN vijay.peshkar@wipro.com (Vijay Peshkar) wrote in message news:... > Hi, > > Am trying to use VxWorks provided resolver on a PC based target. > The DNS resolver component is selected and the domain & server IP > are configured as in example. > > First am running resolvInit and resolvParamsSet as in code below. > if (ERROR == resolvInit( pResolvParams->nameServersAddr[0], > pResolvParams->domainName, 0 )) > { > printf("DNS: ERROR: resolvInit() failed\n"); > return ERROR; > } > > if (ERROR == resolvParamsSet (pResolvParams) > { > printf("DNS: ERROR: resolvParamsSet() failed\n"); > return ERROR; > } > > Have added my target's IP as nameserver in a linux machine's resolv.conf. > > Then running resolvGetHostByName() from shell. The resulting debug > message on shell is as below. Now if the resolvSend is going thru, > what's failing?. > > -> resolvGetHostByName "VxWorksTarget.mynet.wipro.com",0x76ff828,512 > resolvSend() > ;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 1 > ;; flags: rd; Ques: 1, Ans: 0, Auth: 0, Addit: 0 > ;; QUESTIONS: > ;; VxWorksTarget.mynet.wipro.com, type = A, class = IN > > resolvSend() > ;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 2 > ;; flags: rd; Ques: 1, Ans: 0, Auth: 0, Addit: 0 > ;; QUESTIONS: > ;; VxWorksTarget.mynet.wipro.com.mynet.wipro.com, type = A, class = IN > > value = 0 = 0x0 > -> > > Thanks and Regards, > Vijay --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Ada in VxWorks Date: Tue, 24 Sep 2002 13:54:38 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: Reply-To: "Michael Lawnick" Sender: mlawnick@pd9e65697.dip.t-dialin.net Hi Marcus, don't know about Ada, but generally: May be there is a ada task (e.g garbage collection) still running? When now loading the object again you most certainly get a crash. HTH Michael "Marcus Hedlund" schrieb im Newsbeitrag news:amp9it$1dj$1@newstoo.ericsson.se... > Hi Martin, > I have already managed to integrate the adacompiler. It is just that when I > run Ada program in the VxWorks windshell, I can only run it once. If I then > try to run it again without rebooting, it will not work. > I hope this clarifies my question > > /Marcus > > "Martin Raabe" wrote in message > news:amp6cq$gtm$1@news.online.de... > > Hello Marcus, > > there is a GNAT pro integration to Tornado and there is also one for > SNiFF+. > > Maybe you want to take a look into that? > > > > HIH > > -- > > Martin Raabe > > E: Martin.RaabeATNOSPAMBaSystem.de > > > > "Marcus Hedlund" schrieb im Newsbeitrag > > news:amp5s2$mki$1@newstoo.ericsson.se... > > > Hi! > > > > > > I have managed to integrate GNAT pro with VxWorks and am trying to run a > > few > > > ada programs. > > > However I can only run the same Ada program once. After trying to run it > > > twice, an exception is raised and the program hangs. > > > Anyone had this kind of problem? > > > > > > Any help would appreciated, > > > > > > /M. Hedlund > > > > > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Fastest way to transfere data between a PC with win2000 and a PC with vxWorks? Date: 24 Sep 2002 05:36:26 -0700 From: ajn@bredband.net (Ake) Organization: http://groups.google.com/ Message-ID: <7777af4e.0209240436.186cd5c1@posting.google.com> References: <7777af4e.0209211346.1b507f81@posting.google.com> Thanks for your replies! When I asked this question I thought that some one would have suggested another cable inteface, a parallel inteface or something even beter. Åke --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Fastest way to transfere data between a PC with win2000 and a PC with vxWorks? Date: Tue, 24 Sep 2002 15:15:21 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1032880564.202126@newsreader2.wirehub.nl> References: <7777af4e.0209211346.1b507f81@posting.google.com> If you do not apply these suggested changes, and send a lot of data, can this be the reason why a system has troubles with datatransfers of about 1 or 2 MB using FTP of the target server? Even is I transfer about 700K things go alright the first time, but for second time... We are using a PIII based board of VMIC with an intel 82559 chip on it, we are using the FEI END driver. A suggestion that hubs or switches in the network might cause problems - as a support engineer at VMIC said - sounds strange.. Cheers, Renee - ----- Original Message ----- From: "Leonid Rosenboim" Newsgroups: comp.os.vxworks Sent: Sunday, September 22, 2002 8:22 AM Subject: Re: Fastest way to transfere data between a PC with win2000 and a PC with vxWorks? > You will be able to do even better then 100Mbps with FTP, > (if for example you had a GE interface), but doing this > isn't going to be easy. You wull need to spend some > time enhancing (or applying patches) to the FE driver, > so it will perform at 100Mbps without freezing, then > you will have to do some work on the disk attachment > driver (IDE or SCSI) and make sure it tops 10MBytes/sec. > Lastly, the only modification for the VxWorks FTP server > is to increase the buffer it uses move data between network > and disk, and tue the disk cache. > > If this souds like a lot of work, it is, but unfortunately I can not > suggest any alternative way to quickly move this data. > > - Leonid > > > "Ake" wrote in message > news:7777af4e.0209211346.1b507f81@posting.google.com... > > Hello! > > We may start a new project where we need to move a lots of between two > > PC one with win2000 and one with VxWorks. My ide is to set up a > > 100Mbits/s ethernet link and transfer data with FTP. Is there a faster > > way to do this. (instead of using Ehernet) > > > > The PC with win2000 will be responsible to move the files. Note no > > other computers will use this cable. > > > > I hope there is support for 100 Mbits/s Ethernet in vxWorks, Is it? > > > > On the same Ethernet connection I like to send commands between the > > two PC, what protocol should I use then? > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: vxworks watch dog Date: Tue, 24 Sep 2002 13:44:46 GMT From: Pete Kockritz Organization: AT&T Broadband Message-ID: References: In article , Shlomo Katz wrote: > hi > i am using vxworks watch utility and found some difficulties > > i have the follwing code > wd = wdCreate(); > status = wdStart(wd,(5_SEC), (FUNCPTR)trec_CheckUnitTimeOut,NULL); > when > trec_CheckUnitTimeOut - send a message to a task and have no > parameters but i do not get the message > > does someone have any idea why not > the task that should receive the message is ok and work fine with other > messages sent to it What does your watchdog callback function look like? Recall that the watchdog callback function is executed in ISR context. Does your callback do anything that should not be done in an ISR, like printf? I'm assuming that your wdStart function returns OK; are you checking the return value? - -- Pete Kockritz --------------------------- Newsgroups: comp.os.vxworks Subject: Re: semTake()/exit()/taskSpawn() Date: Tue, 24 Sep 2002 09:53:26 -0400 From: Keith Arner Organization: Marconi Message-ID: References: <313680C9A886D511A06000204840E1CF67A1A0-100000@whq-msgusr-02.pit.comms.marconi.com> <313680C9A886D511A06000204840E1CF67A1A2-100000@whq-msgusr-02.pit.comms.marconi.com> Reply-To: Keith Arner On Mon, 23 Sep 2002, Leonid Rosenboim wrote: > "Keith Arner" wrote in message > news:Pine.GSO.4.42.0209231300020.1626-100000@zimbra... > > On Fri, 20 Sep 2002, Leonid Rosenboim wrote: > > > Who's talking about deleting semaphores? I'm talking about deleting > > tasks. > > > Unfortunate heading, but try reading the section underneeth, it specifically > refers to protection against deletion of a task while it owns a Mutex > semaphore, in order to avoid the exact probloem you where referring > to. But making a mutex deletion safe only protects a task from a *different* task trying to delete it. If a task takes a deletion safe mutex then exits all on its own, the task will go away, and the mutex will remain owned by a non-existant task. And tasks exiting on their own is the only case that we should really be concerned with because tasks deleting other tasks is a recipe for disaster (are you *sure* that all the mutexes are deletion safe? are you *sure* that task is not about to give a binary semaphore that a third task is waiting for? are you *sure* that task is not using a binary semaphore as a mutex semaphore?). "taskDelete() considered harmful." > [snip] > > Working around the current task id reuse is going to involve working > > around the current implementation (which is not necessarily portable to > > future releases of VxWorks), or re-inventing the task id wheel. Both of > > these are things which you'd probably like to avoid. > > This is where I draw the line, not going to discuss any specific workaround > on this NG, but no workaround is really a replacement for a solution. Fair enough. Keith - -- "Once something has gone horribly wrong..." "...it just gets horribler and horribler." "Then, it's a matter of finding the original horriblism." -- from a conversation on debugging problems in a linked list --------------------------- Newsgroups: comp.os.vxworks Subject: Re: protocol stack problem Date: 24 Sep 2002 07:01:56 -0700 From: google@mail.zevenaar.org (Renee Teunissen) Organization: http://groups.google.com/ Message-ID: References: <8400a733.0209210128.499227c3@posting.google.com> gonozal@gmx.at (Toni Kurzberger) wrote in message news:<8400a733.0209210128.499227c3@posting.google.com>... > He Jin wrote in message news:... > > Maybe try to add an input hook as a first try (either binding a SNARF > Protocol or adding an ether input hook). This will allow you to have a > look at the packet. > There you should be able to see what's actually going wrong... mabye > only an alignement issue? > > tnoi I have about the same problem. Could a lot of Novell ipx/spx traffic be a problem - or the source of this? I'm using a VMIC PIII compact PCI board (VMICPCI-7755) with an onboard 82559ER chip. While sending data to the boards , things go wrong, target server-filesystem stops functioning, ftp stop working on files larger than 1809 bytes.. Any idears? - ---------------------- - -> ifShow "fei" ifShow "fei" fei (unit number 0): Flags: (0x8063) UP BROADCAST MULTICAST ARP RUNNING Type: ETHERNET_CSMACD Internet address: 172.25.96.183 Broadcast address: 172.25.255.255 Netmask 0xffff0000 Subnetmask 0xffff0000 Ethernet address is 00:20:38:00:58:27 Metric is 0 Maximum Transfer Unit size is 1500 0 octets received 0 octets sent 216860 packets received 8843 packets sent 216860 unicast packets received 8835 unicast packets sent 0 non-unicast packets received 8 non-unicast packets sent 0 input discards 157258 input unknown protocols 0 input errors 0 output errors 0 collisions; 0 dropped value = 29 = 0x1d - ---------------------- Cheers, Renee. - ---------------------- Renee Teunissen PTS Software bv Meerweg 7, 1405BA Netherlands --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Fastest way to transfere data between a PC with win2000 and a PC with vxWorks? Date: Tue, 24 Sep 2002 17:03:21 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <7777af4e.0209211346.1b507f81@posting.google.com> <1032880564.202126@newsreader2.wirehub.nl> "Renee Teunissen" wrote in message news:1032880564.202126@newsreader2.wirehub.nl... > If you do not apply these suggested changes, and send a lot of data, > can this be the reason why a system has troubles with datatransfers of > about 1 or 2 MB using FTP of the target server? Even is I transfer > about 700K things go alright the first time, but for second time... > I am sorry, I lost you here. What do you mean "FTP of the target server" ?? Please be specific, what is the role of the VxWorks target (FTP server or client?) Where are the files (i.e. filesystem and disk type). > We are using a PIII based board of VMIC with an intel 82559 chip on > it, we are using the FEI END driver. A suggestion that hubs or > switches in the network might cause problems - as a support engineer > at VMIC said - sounds strange.. > This i snot strange - there is one scenario I have describe on this NG serveral times that can get you in trouble - here is goes again (briefly): If there are two nodes on the network, one is connected with 10Mbps and the other with 100Mbps, and if the switch/hub does not apply flow control (aka back pressure), file transfers between the two can become unacceptably slow. The two nodes could be of any hardwar architeture or OS. Good luck, Leonid --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Fastest way to transfere data between a PC with win2000 and a PC with vxWorks? Date: Tue, 24 Sep 2002 17:08:24 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <7777af4e.0209211346.1b507f81@posting.google.com> <7777af4e.0209240436.186cd5c1@posting.google.com> Well, I thought about this a bit before replying to you, and in the past I used all sorts of interfaces for data transfer between a Real-time target and a Unix host (SCSI, DR11W, bus-to-bus adapters) but it where the days when Ethernet was only 10Mbps. Now if you add a Gigabit Ethernet to your target, it is not the Ethernet which is the bottle neck anymore, so anything I can think of, except maybe of reflective memory (proprietary or InfiniBand) there is really no good alternative, and reflective memory is awfully expensive. "Ake" wrote in message news:7777af4e.0209240436.186cd5c1@posting.google.com... > Thanks for your replies! > > When I asked this question I thought that some one would have > suggested another cable inteface, a parallel inteface or something > even beter. > > Åke --------------------------- Newsgroups: comp.os.vxworks Subject: Context Switch Time? Date: Tue, 24 Sep 2002 12:14:08 -0400 From: Alex Bruce Organization: Alcatel SDN Message-ID: <3D908F50.771434A0@alcatel.com> SSdtIGN1cmlvdXMgaWYgYW55b25lIGhhcyBtZWFzdXJlZCB0aGUgY29udGV4dCBzd2l0Y2gg dGltZSBvbiB0aGVpcg0KdGFyZ2V0IHBsYXRmb3JtPw0KDQpXZSdyZSBjdXJpb3VzIGhvdyBt dWNoIHRpbWUgaXQgdGFrZXMgZm9yIG9uZSB0YXNrIHRvIGJlIHJlcGxhY2VkIGJ5DQphbm90 aGVyLiBEZXRhaWxzIG9uICphbnkqIHBsYXRmb3JtIHdvdWxkIGJlIGhlbHBmdWwsIGhvd2V2 ZXIgd2UgYXJlDQp0eXBpY2FsbHkgcnVubmluZyA2NiBNSHogUG93ZXJQQ3MuDQoNCkNoZWVy cywNCkFsZXgNCg0KDQo= --------------------------- Newsgroups: comp.os.vxworks Subject: Task Reschedule on mutex unlock Date: Tue, 24 Sep 2002 12:16:15 -0400 From: Alex Bruce Organization: Alcatel SDN Message-ID: <3D908FCF.F75EE09C@alcatel.com> SXMgYW55b25lIGZhbWlsaWFyIHdpdGggdGhlIGlzc3VlIGluIFZ4V29ya3Mgd2hlbiB1bmxv Y2tpbmcgYSBtdXRleCB0aGF0DQpoYXMgcmFpc2VkIHRoZSBwcmlvcml0eSBvZiB0YXNrPyBX ZSd2ZSBkaXNjb3ZlcmVkIHRoYXQgVnhXb3JrcyBkb2VzIG5vdA0KcmVzY2hlZHVsZSB0aGUg dGFzayBiYWNrIHRvIGl0J3MgaW5pdGlhbCBwcmlvcml0eSB3aGVuIHVubG9jayB0aGUgbXV0 ZXgNCmFuZCB0aGF0IHRoZSB0YXNrIHdpbGwgY29udGludWUgYXQgdGhlIGhpZ2hlciBwcmlv cml0eSB1bnRpbCBhbm90aGVyDQpyZXNjaGVkdWxlIGV2ZW50IChmcm9tIHRhc2tEZWxheSwg b3IgYW5vdGhlciBzZW1hcGhvcmUpLg0KDQpXZSBoYXZlIGEgc2lsbHkgd29ya2Fyb3VuZCwg anVzdCB3b25kZXJpbmcgaWYgZm9sa3MgYXJlIGZhbWlsaWFyIHdpdGgNCnRoaXMgaXNzdWU/ DQoNCkNoZWVycywNCkFsZXgNCg== --------------------------- Newsgroups: comp.os.vxworks Subject: Harddisk Copy!! Date: 24 Sep 2002 09:55:23 -0700 From: erdal@web.de (erdal) Organization: http://groups.google.com/ Message-ID: Hallo, my question is can i make harddisk immage from a Vxworks OS with 2 partitions. i find out that th normal Image software (Drive Image and so on) not work thans for your help. PS: It is a SCSI harddisk. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Harddisk Copy!! Date: Tue, 24 Sep 2002 19:00:57 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: There is very few things easier to do than this ! Just write a simple loop, that uses the blkIo API to directly interface the SCSI driver (could be an ATA driver too), read a bunch of sectors from one drive inbto a buffer and write them to the other disk. The blkIo API is as easy as they come. - - Leonid "erdal" wrote in message news:cee71399.0209240855.596549f4@posting.google.com... > Hallo, > my question is can i make harddisk immage from a Vxworks OS with 2 > partitions. > > i find out that th normal Image software (Drive Image and so on) not > work thans for your help. > PS: It is a SCSI harddisk. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Task Reschedule on mutex unlock Date: Tue, 24 Sep 2002 13:29:26 -0400 From: Keith Arner Organization: Marconi Message-ID: References: <3D908FCF.F75EE09C@alcatel.com> Reply-To: Keith Arner On Tue, 24 Sep 2002, Alex Bruce wrote: > Is anyone familiar with the issue in VxWorks when unlocking a mutex that > has raised the priority of task? We've discovered that VxWorks does not > reschedule the task back to it's initial priority when unlock the mutex > and that the task will continue at the higher priority until another > reschedule event (from taskDelay, or another semaphore). > > We have a silly workaround, just wondering if folks are familiar with > this issue? The problem stems from the fact that there is no way to determine which mutex semaphores are owned by a task, given a task id (you can get from the semaphore to the task, but not from the task to the semaphore). Each task maintains a count of the inversion safe semaphores it holds. As it acquires more semaphores, its priority ratchets up. But the priority will not return to "normal" untill *all* of the inversion safe semaphores have been released, and the count goes back to zero. The algorithm works in a lot of common, special cases. But it only takes on non-special case to ruin your day. Last time I checked, WindRiver didn't seem interested in fixing this problem in 5.4. From what I've seen of VxWorks AE, the kernel might actually have enough information to be able to calculate priority (de)inheritance correctly. The Tornado tools allow you to navigate between all sorts of objects. It's concievable to me that they *might* have fixed it in AE. But I cannot say for certain. Keith - -- "Once something has gone horribly wrong..." "...it just gets horribler and horribler." "Then, it's a matter of finding the original horriblism." -- from a conversation on debugging problems in a linked list --------------------------- Newsgroups: comp.os.vxworks Subject: Re: File read problem using shared memory system Date: Tue, 24 Sep 2002 13:47:43 -0500 From: Andrew Johnson Organization: APS, Argonne National Laboratory Message-ID: References: Hi Alastair, Alastair Borrowman wrote: > > Does anybody have any experience of similar problems when moving between > 5.3 and 5.4 or have any ideas as to what may be causing the above > behaviour? I'm not sure why a 5.3 => 5.4 upgrade would have changed things, but this sounds like it might be explainable by the CPU caches. After doing the read(), the master should call cacheFlush() over the file's memory image to ensure everything gets pushed out to RAM, and the slaves should similarly call cacheInvalidate() over that same shared region before reading any data to ensure they don't see stale cached date. I suspect VxMP sets up its own special shared region to use cache writethrough, but your file isn't going into that area and outside of the shared region the cache behaviour is up the BSP. If the hkbaja47 doesn't fully support bus snooping then you could get symptoms like you're seeing. HTH, - - Andrew - -- "Life is what happens while you're busy making other plans." - John Lennon --------------------------- Newsgroups: comp.os.vxworks Subject: Experienced VxWorks Developer Available For Contract Work Date: Tue, 24 Sep 2002 15:14:15 -0700 From: realtime4u Organization: Bell Sympatico Message-ID: <3D90E3B7.9A86D6D5@yahoo.com> Hello, I am an experienced embedded software/firmware developer located in Ottawa Canada. I am seeking VxWorks software development work on a contact basis. SUMMARY OF QUALIFICATIONS - - 12+ years developing embedded real-time critical software. - - Extensive knowledge of the TCP/IP protocol suite, and driver implementation. - - Development experience in the UNIX (SunOS, Solaris), QNX, pSOS, VMS, and Windows operating systems. - - Extensive in-depth knowledge of VxWorks and Tornado development environment. - - Extensive experience in developing VxWorks Board Support Package (BSP) for custom targets. - - Proficient in programming in C/C++, Java, and ASSEMBLY (MIPS R4000, Motorola 680X0/PPC860, Intel 80X86). - - Extensive knowledge of low-level software debugging using logic analyzers, BDM, and ICE. - - Extensive knowledge of initial board bring-up. - - Experienced in delivering a product from design to implementation in both a small startup environment, and in a large telecommunication research and development environment. If you would like more information, or a copy of my resume please reply to realtime4u@yahoo.com --------------------------- Newsgroups: comp.os.vxworks Subject: The FTP server access Date: 24 Sep 2002 12:38:37 -0700 From: robert_markey@hotmail.com (robert) Organization: http://groups.google.com/ Message-ID: <1fac8c75.0209241138.5a6c09bb@posting.google.com> Hi guys, The FTP server provided by Wind River is accessible from the Window Start Menu. I'm tryig to create additional button to the Tornado to accomplish the same, but this file doesn't create nohing, I also try to comment proc ftpdInitUI and use "controlCreate" and "menuItemInsert" out of proc - same thing, maybe some ideas what I should do? thank you, Robert I copy the file 01Ftp.win32.tcl, which is provided below, into the following directory: $WIND_BASE\host\resource\tcl\app-config\Tornado Copy the icon image, ftp.bmp, into: $WIND_BASE\host\resource\bitmaps\Launch\controls # 01Ftp.win32.tcl - wftpd32.exe Tcl initialization file proc ftpdInitUI {} { controlCreate launch [list toolbarbutton \ -callback "LaunchFtpd" \ -name LaunchFtpd \ -tooltip "Launch FTP Server" \ -bitmap "[wtxPath host resource bitmaps launch controls]ftp.bmp"] menuItemInsert -statusmessage "Launch FTP Server" \ -callback {LaunchFtpd} \ -after -bypath \ {&Tools &Debugger...} "&FTP Server" } proc LaunchFtpd {} { toolLaunch "LaunchFtpd" "[wtxPath host x86-win32 bin]wftpd32.exe" "[wtxPath host x86-win32 bin]" 0 0 0 0 0 0 } end --------------------------- Newsgroups: comp.os.vxworks Subject: Re: The FTP server access Date: Tue, 24 Sep 2002 15:48:00 -0400 From: david lindauer Message-ID: <3D90C170.1C5E1FBB@notifier-is.net> References: <1fac8c75.0209241138.5a6c09bb@posting.google.com> what is wtxpath set to? David robert wrote: > Hi guys, > The FTP server provided by Wind River is accessible from the Window > Start Menu. > I'm tryig to create additional button to the Tornado to accomplish the > same, but this file doesn't create nohing, I also try to comment proc > ftpdInitUI and use "controlCreate" and "menuItemInsert" out of proc - > same thing, maybe some ideas what I should do? > thank you, > Robert > > I copy the file 01Ftp.win32.tcl, which is provided > below, into the following directory: > $WIND_BASE\host\resource\tcl\app-config\Tornado > Copy the icon image, ftp.bmp, into: > $WIND_BASE\host\resource\bitmaps\Launch\controls > > # 01Ftp.win32.tcl - wftpd32.exe Tcl initialization file > > proc ftpdInitUI {} { > > controlCreate launch [list toolbarbutton \ > -callback "LaunchFtpd" \ > -name LaunchFtpd \ > -tooltip "Launch FTP Server" \ > -bitmap "[wtxPath host resource bitmaps launch > controls]ftp.bmp"] > > menuItemInsert -statusmessage "Launch FTP Server" \ > -callback {LaunchFtpd} \ > -after -bypath \ > {&Tools &Debugger...} "&FTP Server" > > } > > proc LaunchFtpd {} { > toolLaunch "LaunchFtpd" "[wtxPath host x86-win32 > bin]wftpd32.exe" > "[wtxPath host x86-win32 bin]" 0 0 0 0 0 0 > } > end --------------------------- Newsgroups: comp.os.vxworks Subject: Image or Copy from a harddisk with vxWorks OS ???!! Date: 24 Sep 2002 12:55:29 -0700 From: erdal@web.de (erdal) Organization: http://groups.google.com/ Message-ID: Hallo, I have a problem we are working with printer with has an OS vxWorks with 2 partisions, and i wan´t to make a image or a complet Copy of this hardisk is it Possibel an if how?? The Normal Tools like Drive image are not Work!!! thank in avance. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: dosFs2 Date: Mon, 23 Sep 2002 21:18:11 +0100 From: molochai Organization: Griffin Internet - http://www.griffin.net.uk Message-ID: <3usnma.vmn.ln@gonzo.cybermuppet.co.uk> References: Leonid Rosenboim wrote: > Yes, there is a last-accessed timestamp in DosFs on all systems, > and yes, DosFs2.0 does update it every time it's accessed, > some people have complained about that, and it seems that WindRiver > should add an option to disable this when not required. > The time is taken from the current VxWorks "wall clock" (see > clock_gettime() ) That's what I eventually assumed was going on. This behaviour exacebated the other problem we saw. > > As to the rest of your issues, I am not sure I understand what is it > exactly that happens to you What we saw was a single file which is accessed frequently suddenly gaining an improbably file extent (either 0 or 0xffffffff). On some occasions this new extent was written back to the disc, and on others only ever affected cached directory entries presumably held by dosFs2. The problem goes away with the application of the dosFs2 patch (can't remember the number off the top of my head) from Windsurf. The documentation with the patch does't say much other than describing it as a compiler problem (which in fact it isn't at all). What I suspect is happening is that dosfs2 uses the GCC long long type (64 bit int) to hold the file extend, and that on GCC2.7.2 for PPC, this is implemented using floating point registers. As many of the contexts dosFs2 may get run in don't have the floating point context saved, this is clearly dangerous. I can only assume that the file extent is at some point held in a floating point register at a point at which the updated directory entry (with last accessed time) is written back to disc (almost guaranteed context switch) and then gets clobbered. We could reproduce it quite reliably. M --------------------------- Newsgroups: comp.os.vxworks,comp.sys.powerpc.misc,comp.sys.powerpc.tech Subject: Breakpoints in Single Step with Vision Date: 24 Sep 2002 15:14:54 -0700 From: jul123x@yahoo.com (Julie) Organization: http://groups.google.com/ Message-ID: <4f5f1a5a.0209241414.46c45ea5@posting.google.com> Hi All, I'm using Single Step with Vision for the Power PC and I want to set breakpoints in my code. However, when I do and then hit "Go" to begin execution of the program, single step will not automatically stop when the breakpoint is hit, I have to press the Stop button. No big deal I hear you say ... well I don't want to stop in the wrong place i.e. some places where I might corrupt the SRR register when I'm not in a "recoverable" state or if I'm executing code at a logical address rather than a physical address and if I don't have the logical to physical mapping set up correctly then Single Step will crash. Anyway, if you know of a way around this(an alias I need to setup or some thing else I need to set etc) you might let me know. Cheers ! --------------------------- Newsgroups: comp.os.vxworks Subject: Re: The FTP server access Date: Wed, 25 Sep 2002 01:27:35 GMT From: "drdiags" Organization: AT&T Broadband Message-ID: References: <1fac8c75.0209241138.5a6c09bb@posting.google.com> Robert, When is the ftpdInitUI() proc called? The file will be source after the 00xxx.tcl files, but all that does is make the proc available. You still have to call the InitUI routine. Could you try making the last line in your tcl script the following: ftpdInitUI This should now put the button where you want it.... "robert" wrote in message news:1fac8c75.0209241138.5a6c09bb@posting.google.com... > Hi guys, > The FTP server provided by Wind River is accessible from the Window > Start Menu. > I'm tryig to create additional button to the Tornado to accomplish the > same, but this file doesn't create nohing, I also try to comment proc > ftpdInitUI and use "controlCreate" and "menuItemInsert" out of proc - > same thing, maybe some ideas what I should do? > thank you, > Robert > > > > I copy the file 01Ftp.win32.tcl, which is provided > below, into the following directory: > $WIND_BASE\host\resource\tcl\app-config\Tornado > Copy the icon image, ftp.bmp, into: > $WIND_BASE\host\resource\bitmaps\Launch\controls > > > # 01Ftp.win32.tcl - wftpd32.exe Tcl initialization file > > proc ftpdInitUI {} { > > controlCreate launch [list toolbarbutton \ > -callback "LaunchFtpd" \ > -name LaunchFtpd \ > -tooltip "Launch FTP Server" \ > -bitmap "[wtxPath host resource bitmaps launch > controls]ftp.bmp"] > > menuItemInsert -statusmessage "Launch FTP Server" \ > -callback {LaunchFtpd} \ > -after -bypath \ > {&Tools &Debugger...} "&FTP Server" > > } > > proc LaunchFtpd {} { > toolLaunch "LaunchFtpd" "[wtxPath host x86-win32 > bin]wftpd32.exe" > "[wtxPath host x86-win32 bin]" 0 0 0 0 0 0 > } > end --------------------------- Newsgroups: comp.os.vxworks Subject: signal and taskCreateHookAdd Date: Tue, 24 Sep 2002 21:52:32 -0700 From: "s s" Organization: Bell Sympatico Message-ID: I am attempting to do is set up a signal handler to process a segmentation violation in a task. However, what I am really trying to do is for each new task that starts, this signal handler is to be set up. So I have been trying to use taskCreateHookAdd() to setup a function that is called when a new task starts. However, when the function gets called, the task is not the new task, but it does have a pointer to the new task's WIND_TCB. I do not know what fields I could update in the new task's WIND_TCB to set up a signal handler for the new task. Has anyone ever tried something like this? My signal handler is: LOCAL jmp_buf env; void sigHandler ( int sig, int code, SIGCONTEXT *sigContext ) { switch (sig) { case SIGSEGV: printErr ("\nSignal SIGSEGV received\n"); break; } longjmp (env, sig); } >My added Hook routine: >void myCreateHook (pTcb) > WIND_TCB *pTcb; { /* Register the signal handler */ } >My code to set up the hook: >taskCreateHookAdd((FUNCPTR) myCreateHook); > --------------------------- Newsgroups: comp.os.vxworks Subject: VxWorks on MCF5206 Date: 24 Sep 2002 20:25:28 -0700 From: andreas_lindell75@hotmail.com (Rikard) Organization: http://groups.google.com/ Message-ID: <5947a04.0209241925.2eb266c0@posting.google.com> Hello, I just want to know if MCF5206 is supported by WindRiver, i.e do they have a BSP. If anyone knows a good distrabutor/manufacturer where I can buy a simple board with a MCF5206 its even better, just the CPU and a COM port nothing more. I dont want to spend 3K USD on a reference desing by WR. Thanks, Andreas --------------------------- Newsgroups: comp.os.vxworks Subject: Has anyone built binutils from source to be cross platform, was it difficult? Date: Wed, 25 Sep 2002 03:39:09 GMT From: FrankG Organization: Giganews.Com - Premium News Outsourcing Message-ID: <3D913105.801@bogus.com> I am porting the wrSbc8260 to the 8265, which is based on a PPC 603e core. I was very annoyed with the 20 or so warnings that romInit.s generates when assembled. I was able to get rid of almost all of them by fixing the broken macros in the c header files. However there was 1 warning I was getting that made no sense. It was complaining about, my providing too many significant bits to a PPC instruction that expected a 16 bit offset. The wierd thing was that I was using @ha directive to obtain a 16 bit result. I searched on google and found some references to an old bug in gas that was repsonsible for my problems. The google thread went on to say that the bug has been fixed. The thread was dated August 1999. So I invoked gasppc -v to find out my version is 1.12. The latest gas is 2.91. Why does vxWorks age the free opensource software before shipping it? Do they ever make updates available on a windriver ftp site? I wasted a half hour explaining this in an email to windriver tech support. No response. But it is opensource software so I should be able to build it myself right? Has anyone successfully built the binutils package to cross target say a PPC from a solaris platform? Was it difficult or easy. I am on a tight schedule so if it takes longer than a half day I won't be able to do it. Thanks in advance. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Tcl for Tornado Date: 24 Sep 2002 21:49:42 -0700 From: senthil_rag@rediffmail.com (senthil) Organization: http://groups.google.com/ Message-ID: <39c3b976.0209242049.21bf450c@posting.google.com> References: <64ddf295.0209230533.cfd71@posting.google.com> Val, Iam assuming you need to execute an module (.o) downloaded in your target through a button event. For doing this, you need to first register you tool to the target server,then use WTX TCL API routines to invoke executables in memory of target. Refer WTX TCL API (4.4) of Tornado API Guide.Hope this helps. regards senthil valrub@hotmail.com (val) wrote in message news:<64ddf295.0209230533.cfd71@posting.google.com>... > Hello everybody, > I'm trying to create access to our program executable (drl.exe) > through the Tornado IDE. > I can create button on the Tornado IDE with the next line of Tcl code: > > controlCreate launch [list toolbarbutton \ > -callback launchWtxapp \ > -tooltip "Launch Tilcon Graphic Editor" \ > -bitmap "[wtxPath host resource bitmaps launch controls]TGE.bmp"] > > but I don't know how to connect that button to the executable > (drl.exe). > > Maybe some ideas ... > Thank you! > Val --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Newbie Starter Kit Date: Wed, 25 Sep 2002 06:00:16 GMT From: "Blackie Beard" Message-ID: References: Reply-To: "Blackie Beard" Wow, that looks really great. I'll read the lectures, too. Thanks! BB "harry mudd" wrote in message news:UsGj9.413508$_91.557821@rwcrnsc51.ops.asp.att.net... > here's a good link > http://www.onesmartclick.com/rtos/rtos.html > "Blackie Beard" wrote in message > news:sA8j9.6967$7i2.195@nwrddc02.gnilink.net... > > ============================== > > I want to learn vxWorks in my spare time. > > How do I get started for very cheap? > > Is there a good kit, and what books are > > recommended. My background is H/W, > > S/W, F/W and SoC, but never studied > > any OS. Thanks in advance! > > > > BB > > ============================== > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Context Switch Time? Date: Wed, 25 Sep 2002 08:14:56 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <3D908F50.771434A0@alcatel.com> Reply-To: "Michael Lawnick" Sender: mlawnick@pd9524d36.dip.t-dialin.net Hi Alex, > Details on *any* platform would be helpful That is not possible, it strongly depends on the particular H/W. Even the setup in your Workspace (cache mode, MMU) might influence the result. Contact your FAE and ask him for the values of your platform. If they are not available, there is a ready to compile benchmark suite he might give to you. Regards, Michael "Alex Bruce" schrieb im Newsbeitrag news:3D908F50.771434A0@alcatel.com... > I'm curious if anyone has measured the context switch time on their > target platform? > > We're curious how much time it takes for one task to be replaced by > another. Details on *any* platform would be helpful, however we are > typically running 66 MHz PowerPCs. > > Cheers, > Alex > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: dosFs2 Date: Wed, 25 Sep 2002 08:17:31 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <3usnma.vmn.ln@gonzo.cybermuppet.co.uk> Indeed, dosFs2 uses "long long" for many data types, as it supports disk size over 4G, and with the VX_LOGN directory structure it will support files over 4G in size. And indeed I know of the PPC compiler issue that it uses FTP regs for 64-bit integers. Hope all is OK now for you. "molochai" wrote in message news:3usnma.vmn.ln@gonzo.cybermuppet.co.uk... > Leonid Rosenboim wrote: > > > Yes, there is a last-accessed timestamp in DosFs on all systems, > > and yes, DosFs2.0 does update it every time it's accessed, > > some people have complained about that, and it seems that WindRiver > > should add an option to disable this when not required. > > The time is taken from the current VxWorks "wall clock" (see > > clock_gettime() ) > > That's what I eventually assumed was going on. This behaviour exacebated the > other problem we saw. > > > > > As to the rest of your issues, I am not sure I understand what is it > > exactly that happens to you > > > What we saw was a single file which is accessed frequently suddenly gaining > an improbably file extent (either 0 or 0xffffffff). On some occasions this > new extent was written back to the disc, and on others only ever affected > cached directory entries presumably held by dosFs2. > > The problem goes away with the application of the dosFs2 patch (can't > remember the number off the top of my head) from Windsurf. > The documentation with the patch does't say much other than describing it as > a compiler problem (which in fact it isn't at all). What I suspect is > happening is that dosfs2 uses the GCC long long type (64 bit int) to hold > the file extend, and that on GCC2.7.2 for PPC, this is implemented using > floating point registers. As many of the contexts dosFs2 may get run in > don't have the floating point context saved, this is clearly dangerous. I > can only assume that the file extent is at some point held in a floating > point register at a point at which the updated directory entry (with last > accessed time) is written back to disc (almost guaranteed context switch) > and then gets clobbered. We could reproduce it quite reliably. > > M > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Harddisk Copy!! Date: Wed, 25 Sep 2002 09:47:28 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1032947293.974657@newsreader2.wirehub.nl> References: Or boot linux of CD and do as root: dd if=/dev/ of=/dev/ bs=1024 If you want it as image file, replace /dev/ with the filename you want to store the image Cheers, Renee - --- Renee Teunissen PTS Software bv "Leonid Rosenboim" wrote in message news:newscache$6qcy2h$8r$1@lnews.actcom.co.il... > There is very few things easier to do than this ! > > Just write a simple loop, that uses the blkIo API > to directly interface the SCSI driver (could be an ATA driver too), > read a bunch of sectors from one drive inbto a buffer and > write them to the other disk. The blkIo API is as easy as they come. > > - Leonid > > "erdal" wrote in message > news:cee71399.0209240855.596549f4@posting.google.com... > > Hallo, > > my question is can i make harddisk immage from a Vxworks OS with 2 > > partitions. > > > > i find out that th normal Image software (Drive Image and so on) not > > work thans for your help. > > PS: It is a SCSI harddisk. > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Fastest way to transfere data between a PC with win2000 and a PC with vxWorks? Date: Wed, 25 Sep 2002 10:22:22 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1032949388.378346@newsreader2.wirehub.nl> References: <7777af4e.0209211346.1b507f81@posting.google.com> <1032880564.202126@newsreader2.wirehub.nl> Sorry about the fuzzyness.. I try to copy a large file, eg a new VxWorks kernel, to my targhet using the FTPserver on the board. I starty my ftp-session from a NT4 cmd-box. After a while I get: "netout:Connection reset" as message by my FTP-clients. Zee also a message of Gijs: - ----- Hi All, Recently we switched from a 3C905C-TX-M Etherlink 10/100 XL PCI ethernet card to a Intel PRO100+ because the 3COM card leaded occassionally to a page fault (with the EL_3C90X_END driver). We managed to get the system running with the Intel PRO100+ (based on the Intel 82559) with the FEI_END driver but are running into another problem now. When we try to FTP a larger file to the VxWorks target system the FTP session hangs. After downloading some 700+ Kbytes the download stops and after a while the FTP client displays the message "netout:Connection reset" (this could be from our FTP client (standard NT4.0 ftp client program), I'm not sure). Also, the connection state in the inetstatShow overview shows "1536". We tried disabling buffer loaning (netConfig.h) and increasing the number of transmit/receive buffer descriptors (from 32 to 256), to no avail. Has anyone discovered the same problems and/or solutions for them? Gijs Withagen Technolution BV - ----- Our disk is an ATA type 20GB harddisk, with an 1GB msdos FAT16 partition we are using. We use the ATA driver.. The same thing happens with the target server filesystem. I dont see these problems if I'm booted with linux on the same hardware. One thing, if using ifAddrSet, the boardcast address and netmask are not set. If I do this, thinks start looking better. I was in assumptionm that if the subnetmask was set, the broadcast address will be set automatically. Regarding our network topology, we do have a switched/hubbed network so there is some Novell traffic on it (boardcast packets) (ipx/spx). Cheers, Renee - ------- Renee Teunissen PTS Software bv "Leonid Rosenboim" wrote in message news:newscache$7a7y2h$0h$1@lnews.actcom.co.il... > > "Renee Teunissen" > agree.with.these.terms@mail.zevenaar.org> wrote in message > news:1032880564.202126@newsreader2.wirehub.nl... > > If you do not apply these suggested changes, and send a lot of data, > > can this be the reason why a system has troubles with datatransfers of > > about 1 or 2 MB using FTP of the target server? Even is I transfer > > about 700K things go alright the first time, but for second time... > > > I am sorry, I lost you here. What do you mean "FTP of the target server" ?? > > Please be specific, what is the role of the VxWorks target (FTP server or > client?) > Where are the files (i.e. filesystem and disk type). > > > > > We are using a PIII based board of VMIC with an intel 82559 chip on > > it, we are using the FEI END driver. A suggestion that hubs or > > switches in the network might cause problems - as a support engineer > > at VMIC said - sounds strange.. > > > > This i snot strange - there is one scenario I have describe on this NG > serveral > times that can get you in trouble - here is goes again (briefly): > If there are two nodes on the network, one is connected with 10Mbps > and the other with 100Mbps, and if the switch/hub does not apply flow > control > (aka back pressure), file transfers between the two can become unacceptably > slow. The two nodes could be of any hardwar architeture or OS. > > Good luck, > > Leonid > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Task Reschedule on mutex unlock Date: Wed, 25 Sep 2002 09:20:57 +0000 (UTC) From: David Laight Organization: BT Openworld Message-ID: References: <3D908FCF.F75EE09C@alcatel.com> > > Last time I checked, WindRiver didn't seem interested in fixing this > problem in 5.4. :-) Actually it is relatively easy to do a much better job. The answer to the question "are any tasks waiting for any mutex with 'priority invertion' enabled" is easy to calculate. If there are no possible waiting tasks it must be safe to reduce the priority. David --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Task Reschedule on mutex unlock Date: Wed, 25 Sep 2002 14:01:17 +0300 From: "Martin Roth" Organization: Motorola Message-ID: References: <3D908FCF.F75EE09C@alcatel.com> I think that I understand why is the priority is being changed only on the release of last mutex. The priority is not being updated because the operation of calculating the next priority is a complicated one. The next priority (after mutex release) should be the priority of the highest priority task that is blocked by any mutex owned by the task. The time complexity of this operation depends on number of tasks blocked by one of semaphores owned by the task. Martin Roth "David Laight" wrote in message news:amrv5o$p60$1@paris.btinternet.com... > > > > Last time I checked, WindRiver didn't seem interested in fixing this > > problem in 5.4. > > :-) > > Actually it is relatively easy to do a much better job. > The answer to the question "are any tasks waiting for > any mutex with 'priority invertion' enabled" is easy > to calculate. If there are no possible waiting tasks > it must be safe to reduce the priority. > > David > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Fastest way to transfere data between a PC with win2000 and a PC with vxWorks? Date: Wed, 25 Sep 2002 14:11:21 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <7777af4e.0209211346.1b507f81@posting.google.com> <1032880564.202126@newsreader2.wirehub.nl> <1032949388.378346@newsreader2.wirehub.nl> Thanks Renee for the long version of the story. This si indeed a complicated story, and here are some test ideas out of the top of my head: Test 1- isolate nbetwork I would recommend putting the NT host and VxWorks target running FTP server on the separate Lab network, with just one switch or hub inbetween, and making sure that both boxes are connected with the same speed to the switch (i.e. avoid speed mismatch). Test 2 - test FTP using the supplied ftpXfer() API, you can do some FTP'ing locally on the VxWorks box, write some test code to write a large file by putting large bogus data into the data socket returned by ftpXfer(). This will test the FTP server, file system and ATA driver in one punch. Let me know what you come up with. - - Leonid Rosenboim Consultant --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Task Reschedule on mutex unlock Date: Wed, 25 Sep 2002 14:17:21 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <3D908FCF.F75EE09C@alcatel.com> No Martin, this operation is a constant time - when you enable priority inheritance on a mutex sem, it will also enable Priority_Q option which means that the tasks blocked are ordered by their priority, so to look at the topmost task is simply to look at the head of the queue. "Martin Roth" wrote in message news:ams571$qds$1@newshost.mot.com... > I think that I understand why is the priority is being changed only on the > release of last mutex. The priority is not being updated because the > operation of calculating the next priority is a complicated one. The next > priority (after mutex release) should be the priority of the highest > priority task that is blocked by any mutex owned by the task. The time > complexity of this operation depends on number of tasks blocked by one of > semaphores owned by the task. > > Martin Roth > > > "David Laight" wrote in message > news:amrv5o$p60$1@paris.btinternet.com... > > > > > > Last time I checked, WindRiver didn't seem interested in fixing this > > > problem in 5.4. > > > > :-) > > > > Actually it is relatively easy to do a much better job. > > The answer to the question "are any tasks waiting for > > any mutex with 'priority invertion' enabled" is easy > > to calculate. If there are no possible waiting tasks > > it must be safe to reduce the priority. > > > > David > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Task Reschedule on mutex unlock Date: Wed, 25 Sep 2002 08:57:02 -0400 From: Keith Arner Organization: Marconi Message-ID: References: <3D908FCF.F75EE09C@alcatel.com> Reply-To: Keith Arner On Wed, 25 Sep 2002, David Laight wrote: > > Last time I checked, WindRiver didn't seem interested in fixing this > > problem in 5.4. > > :-) > > Actually it is relatively easy to do a much better job. > The answer to the question "are any tasks waiting for > any mutex with 'priority invertion' enabled" is easy > to calculate. If there are no possible waiting tasks > it must be safe to reduce the priority. If that were done, then priority inheritance for any given task would depend on the state of all other (read: unrelated) tasks. So introducing a new task could "break" the priorities of existing tasks. To fix the problem, you really need to be able to calculate the set of tasks blocked by a given task (i.e. the set of tasks blocking on semaphores owned by the given task), and choose the highest priority of all the tasks in this set. If you maintain the list in priority order, choosing the highest priority is constant time, though inserting a node is O(n). Keith - -- "Once something has gone horribly wrong..." "...it just gets horribler and horribler." "Then, it's a matter of finding the original horriblism." -- from a conversation on debugging problems in a linked list --------------------------- Newsgroups: comp.os.vxworks Subject: signal and taskCreateHookAdd Date: Wed, 25 Sep 2002 12:58:48 GMT From: "Hadi" Organization: Magma Communications Ltd. Message-ID: Hello, I am trying to hook signal to all the task so that I could catch bus Error. I thought I could use taskCreateHookAdd() function, to add my signal handler, so that every time a new task get created my signal handler will get register. But the problem is when the registration function is called then its not running under the context of the newly created task. So I could register signal handler for that new task. I looked at the WIND_TCB structure. It cantains "struct sigtcb * pSignalInfo" . Is there any function that would allow me to connect my signal handler to a task from another task. If I know the WIND_TCB of a task how do I register the signal handler to it. Hadi --------------------------- Newsgroups: comp.os.vxworks Subject: Re: symTbl.c vs. ELF .stab Date: 25 Sep 2002 09:44:23 -0400 From: "Paul D. Smith" Organization: Nortel Networks, Inc. / Billerica, MA Message-ID: References: Reply-To: pausmith@nortelnetworks.com %% David Laight writes: dl> I'd look at the script used to extract the names. dl> It is probably TCL - about the slowest way possible.... Yeah--but actually it's not the speed of generating the .c file that seems to take very long, it's the extra link required afterwards. Thanks... - -- - ------------------------------------------------------------------------------- Paul D. Smith HASMAT--HA Software Mthds & Tools "Please remain calm...I may be mad, but I am a professional." --Mad Scientist - ------------------------------------------------------------------------------- These are my opinions---Nortel Networks takes no responsibility for them. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Task Reschedule on mutex unlock Date: Wed, 25 Sep 2002 16:29:41 +0300 From: "Martin Roth" Organization: Motorola Message-ID: References: <3D908FCF.F75EE09C@alcatel.com> Leonid, I guess that the Priority_Q is for one mutex sem. So calculating the priority of the next task is not a problem for one mutex. If we have several mutex semaphores owned by the task, then I believe there should be different priority queues per each mutex sem. In this case the complexity should be as the number of mutex semaphores owned by the task. Thanks for the remark. Martin "Leonid Rosenboim" wrote in message news:newscache$j9uz2h$223$1@lnews.actcom.co.il... > No Martin, this operation is a constant time - > > when you enable priority inheritance on a mutex sem, > it will also enable Priority_Q option which means that > the tasks blocked are ordered by their priority, so to look > at the topmost task is simply to look at the head of the queue. > > "Martin Roth" wrote in message > news:ams571$qds$1@newshost.mot.com... > > I think that I understand why is the priority is being changed only on the > > release of last mutex. The priority is not being updated because the > > operation of calculating the next priority is a complicated one. The next > > priority (after mutex release) should be the priority of the highest > > priority task that is blocked by any mutex owned by the task. The time > > complexity of this operation depends on number of tasks blocked by one of > > semaphores owned by the task. > > > > Martin Roth > > > > > > "David Laight" wrote in message > > news:amrv5o$p60$1@paris.btinternet.com... > > > > > > > > Last time I checked, WindRiver didn't seem interested in fixing this > > > > problem in 5.4. > > > > > > :-) > > > > > > Actually it is relatively easy to do a much better job. > > > The answer to the question "are any tasks waiting for > > > any mutex with 'priority invertion' enabled" is easy > > > to calculate. If there are no possible waiting tasks > > > it must be safe to reduce the priority. > > > > > > David > > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Task Reschedule on mutex unlock Date: Wed, 25 Sep 2002 16:29:12 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <3D908FCF.F75EE09C@alcatel.com> Here is the deal: every mutex sem has a pri_q, problem is the TCB does not have a list of owned mutex sems, only a could of them, so the task does not really know which mutex it owns at an istant, only knows how many. Be assuredm there is a solution to this problem, if WindRiver really wants to fix this. If there is a will there is a way. Apparently Will does not work at WindRiver anymore either. - - Leonid "Martin Roth" wrote in message news:amsdtb$cu$1@newshost.mot.com... > Leonid, > > I guess that the Priority_Q is for one mutex sem. So calculating the > priority of the next task is not a problem for one mutex. > > If we have several mutex semaphores owned by the task, then I believe there > should be different priority queues per each mutex sem. > In this case the complexity should be as the number of mutex semaphores > owned by the task. > > Thanks for the remark. > > Martin > > > "Leonid Rosenboim" wrote in message > news:newscache$j9uz2h$223$1@lnews.actcom.co.il... > > No Martin, this operation is a constant time - > > > > when you enable priority inheritance on a mutex sem, > > it will also enable Priority_Q option which means that > > the tasks blocked are ordered by their priority, so to look > > at the topmost task is simply to look at the head of the queue. > > > > "Martin Roth" wrote in message > > news:ams571$qds$1@newshost.mot.com... > > > I think that I understand why is the priority is being changed only on > the > > > release of last mutex. The priority is not being updated because the > > > operation of calculating the next priority is a complicated one. The > next > > > priority (after mutex release) should be the priority of the highest > > > priority task that is blocked by any mutex owned by the task. The time > > > complexity of this operation depends on number of tasks blocked by one > of > > > semaphores owned by the task. > > > > > > Martin Roth > > > > > > > > > "David Laight" wrote in message > > > news:amrv5o$p60$1@paris.btinternet.com... > > > > > > > > > > Last time I checked, WindRiver didn't seem interested in fixing this > > > > > problem in 5.4. > > > > > > > > :-) > > > > > > > > Actually it is relatively easy to do a much better job. > > > > The answer to the question "are any tasks waiting for > > > > any mutex with 'priority invertion' enabled" is easy > > > > to calculate. If there are no possible waiting tasks > > > > it must be safe to reduce the priority. > > > > > > > > David > > > > > > > > > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Getting crazy with VXWSymTab ... Can anyone look at this? ... please! Date: 25 Sep 2002 07:45:10 -0700 From: maba@combra.se (Marco) Organization: http://groups.google.com/ Message-ID: <8ee874e3.0209250645.136ad5a7@posting.google.com> Hi! I'm trying to use the class VXWSymTab as an hash table. But even this simple code doesn't want to run. (memSysPartId is a global variable that stores the ID of the system memory. Such a variable is declared in memLib.h.) ========================================= #include "vxWorks.h" #include "memLib.h" #include "vxwSymLib.h" int run(void){ VXWSymTab* test = new VXWSymTab(6, FALSE, memSysPartId); return 0; } ========================================= The instantiation fails because symTblCreate returns 0! Am I missing something? Thanks in advance! Marco Basile ========================================= VXWSymTab (int hashSizeLog2, BOOL sameNameOk, PART_ID symPartId) : stid_ (symTblCreate (hashSizeLog2, sameNameOk, symPartId)) { if (stid_ == 0) vxwThrowErrno (); } ========================================= --------------------------- Newsgroups: comp.os.vxworks Subject: BSP for TQM8xxL module Date: Wed, 25 Sep 2002 19:01:50 +0400 From: "Denis Dubrovin" Organization: WEBPlus Ltd. Message-ID: Hi! Where are I can get BSP for vxWorks 5.4 for my TQM8xxL module? - -- Regards! Den --------------------------- Newsgroups: comp.os.vxworks,comp.sys.powerpc.misc,comp.sys.powerpc.tech Subject: Re: Breakpoints in Single Step with Vision Date: 25 Sep 2002 10:10:01 -0700 From: ralph@depping.com (ralph) Organization: http://groups.google.com/ Message-ID: <3dcfdb29.0209250910.312f8330@posting.google.com> References: <4f5f1a5a.0209241414.46c45ea5@posting.google.com> Hi, we've had the same problem. Are you using Windows 2000? If so this is an app note from WindRiver that might help! Ralph Enabling Win2K Parallel Port Interrupts How To Enable Interrupts on Windows 2000 for visionPROBE Copyright © 1984-2001 Wind River Systems Inc. ALL RIGHTS RESERVED. Introduction The visionPROBE is a tool used to connect a debugger running on a host PC (visionCLICK or SingleStep with vision), to a target development board. It connects to the parallel port on the PC and the JTAG or BDM port of the Target. In order for the visionPROBE to operate correctly, it requires an interrupt be assigned to the parallel port device driver. Unlike all previous versions of MS Windows, Windows 2000 does not assign an interrupt to its parallel port device driver. This Application Note describes how to assign a suitable interrupt to the driver in order to use a visionPROBE correctly. Symptoms The main symptom of an incorrectly configured interrupt is a failure to stop at breakpoints. Typically the problem can be described as: “I can download and run code OK on my target, but when I set a breakpoint the target fails to stop at that point. Strangely, if I hit stop in the debugger the source window shows that I am at the place when I set the breakpoint.” What is actually happening is that the target stops at the breakpoint, but that the debugger is unaware of this. If you look at the visionPROBE you will notice that the “Run” LED has gone out, and the “Debug” LED is illuminated. The probe tries to signal this to the PC using an interrupt, but the application does not see it. This also affects flash programming as the algorithm uses breakpoints to signal to the PC when it has finished erasing or programming a block of data. Note that not all versions of the visionPROBE firmware use interrupts in the same way. So you may have a situation where you are using a single PC and visionPROBE on two different architectures, one of which works and the other does not. Enabling Interrupts The Windows 2000 System Resource Editor is vastly improved over that shipped with Windows NT. Most of the information is displayed in a simple, easy to follow manner. The following steps describe how to enable interrupts and assign a suitable IRQ to the parallel port driver using the editor. The example is done using the first parallel port (LPT1) but is applicable to any parallel port. i) Boot the PC, then log in to Windows as the Administrator or with an account with administrative privileges. Normal users are prevented from changing system resources. ii) Once logged in, right-click on the “My Computer” icon and select the “Properties” option. iii) In the “System Properties” window, select the “Hardware” tab, then click the “Device Manager…” button. iv) In the “Device Manager” window, expand the “Ports (COM & LPT)” section and double-click on the “ECP Printer Port (LPT1)” entry. NOTE: If the entry does not say “ECP” then you do not have the correct driver, or your parallel port is not configured for ECP mode in the BIOS. Since there are so many BIOS types, changing this setting is beyond the scope of this Application Note. v) In the Properties dialogue, select the “Port Settings” tab and make sure “Use any interrupt assigned to the port” is selected as shown. vi) Next select the “Resources” tab, you should see the current I/O, DMA and Interrupt resources used as follows: The example above shows the normal settings for LPT1, i.e. I/O at 378 and 778 (the end addresses may vary) and IRQ 7 in use. In this case, no DMA channel is allocated to the driver. Note that the “Use automatic settings” check-box has been cleared. By clearing this box, it is possible to use the “Settings based on” drop-down menu. This menu offers a number of configurations. If your setting does not match the above configuration then you should try each of the configurations available and select the one that matches. On some systems, there will be no option that matches the above parameters exactly, usually because DMA is required. Enabling DMA will not affect the visionPROBE, so using such a setting will be fine. It is vital that there are no conflicts listed in the “Conflicting device list”. If there are then you have 2 options: change the settings for the port, or change the settings on the conflicting device(s). To change a parameter simply double-click the title of the parameter in the window (e.g. “Interrupt Request”) and you will be given a new dialogue looking like this: If you have to select an IRQ other than 7, or alternative I/O addresses in order to resolve the conflict, then the visionPROBE will need to be configured appropriately. See the COMDLL Configuration section (next). vii) Once configuration is complete, click OK. You may need to re-boot your PC for changes to take effect. COMDLL Configuration If you were unable to use the default LPT1 (or LPT2) settings when enabling interrupts, you can change the settings used by the visionPROBE by editing the “comdll.cfg” file. This file is located in the product directory (C:\ESTII\ for visionCLICK, C:\SDS772 for SingleStep). The file is a simple ASCII text file, which has the following contents by default: PAR1 0x378 0x7 PAR2 0x278 0x5 PAR1 identifies the settings for Parallel Port 1 (i.e. LPT1), and PAR2 for the second port. The entries are for I/O address and Interrupt Request number. Most often it is the IRQ number that needs to be changed. Note that you should exit all visionPROBE applications before editing this file. Editing can be done with any text editor such as Notepad. Once saved, you can re-start the application and it will use the new parameters. Testing The best method to test communications with the visionPROBE are working is to run the VPTest application. This is installed with both visionCLICK and SingleStep. When SingleStep is installed, there is an icon created to run VPTest in the Start menu, visionCLICK users will need to browse to the ESTII subdirectory and locate “vptest.exe”. Use of VPTest is documented elsewhere. References There are a number of application notes and other documents available from Wind River that may be of use in diagnosing problems with a visionPROBE. All the application notes can be found on the visionCLICK CD and on WindSurf. jul123x@yahoo.com (Julie) wrote in message news:<4f5f1a5a.0209241414.46c45ea5@posting.google.com>... > Hi All, > > I'm using Single Step with Vision for the Power PC and I want to set > breakpoints in my code. > > However, when I do and then hit "Go" to begin execution of the > program, single step will not automatically stop when the breakpoint > is hit, I have to press the Stop button. > > No big deal I hear you say ... well I don't want to stop in the wrong > place i.e. some places where I might corrupt the SRR register when I'm > not in a "recoverable" state or if I'm executing code at a logical > address rather than a physical address and if I don't have the logical > to physical mapping set up correctly then Single Step will crash. > > Anyway, if you know of a way around this(an alias I need to setup or > some thing else I need to set etc) you might let me know. > > Cheers ! --------------------------- Newsgroups: comp.os.vxworks Subject: HELP !!! `netJobRing' undeclared ?? Date: 25 Sep 2002 11:19:42 -0700 From: zhouzi@yahoo.com (Zi Zhou) Organization: http://groups.google.com/ Message-ID: <7254c696.0209251019.633ed7bd@posting.google.com> Hi, I am trying to use rngIsEmpty(netJobRing) to check if my netjob queue still have empty entries. I included rngLib.h, netLib.h, the compilation complains 'netJobRing' undeclared. Since I can't find any detailed info about netJobRing, I am assuming it's a global variable that's defined in one of the Lib.h, Am I right? What's the proper way to check if netjob queue still has empty entry and how should I use netJobRing? I am using Tornado 2.0.2 thanks Zi --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks on MCF5206 Date: Wed, 25 Sep 2002 19:12:54 +0000 (UTC) From: "Graham Baxter" Organization: Graham Baxter (Software) Limited Message-ID: References: <5947a04.0209241925.2eb266c0@posting.google.com> Andreas, Wind has a pSOS BSP for the Motorola: ColdFire M5206eLITE Evaluation Board Regards, Graham Baxter Freelance Software Engineer VxWorks and pSOS BSPs (available shortly for new assignment) gbaxter@NOSPAMbcs.org.uk "Rikard" wrote in message news:5947a04.0209241925.2eb266c0@posting.google.com... > Hello, > > I just want to know if MCF5206 is supported by WindRiver, i.e do they > have a BSP. If anyone knows a good distrabutor/manufacturer where I > can buy a simple board with a MCF5206 its even better, just the CPU > and a COM port nothing more. I dont want to spend 3K USD on a > reference desing by WR. > > Thanks, > Andreas --------------------------- Newsgroups: comp.os.vxworks Subject: Re: The FTP server access Date: 25 Sep 2002 13:37:54 -0700 From: robert_markey@hotmail.com (robert) Organization: http://groups.google.com/ Message-ID: <1fac8c75.0209251237.236a790a@posting.google.com> References: <1fac8c75.0209241138.5a6c09bb@posting.google.com> Could you try making the last line in your tcl script the following: ftpdInitUI This should now put the button where you want it.... Yes! if I do that it ceate a button and placed it into the Tornado IDE, but what i see it doesn't invoke "menuItemInsert" and doesn't run application - ftp. Maybe some ideas... "drdiags" wrote in message news:... > Robert, > > When is the ftpdInitUI() proc called? The file will be source after the > 00xxx.tcl > files, but all that does is make the proc available. You still have to call > the > InitUI routine. Could you try making the last line in your tcl script the > following: > > ftpdInitUI > > This should now put the button where you want it.... > > "robert" wrote in message > news:1fac8c75.0209241138.5a6c09bb@posting.google.com... > > Hi guys, > > The FTP server provided by Wind River is accessible from the Window > > Start Menu. > > I'm tryig to create additional button to the Tornado to accomplish the > > same, but this file doesn't create nohing, I also try to comment proc > > ftpdInitUI and use "controlCreate" and "menuItemInsert" out of proc - > > same thing, maybe some ideas what I should do? > > thank you, > > Robert > > > > > > > > I copy the file 01Ftp.win32.tcl, which is provided > > below, into the following directory: > > $WIND_BASE\host\resource\tcl\app-config\Tornado > > Copy the icon image, ftp.bmp, into: > > $WIND_BASE\host\resource\bitmaps\Launch\controls > > > > > > # 01Ftp.win32.tcl - wftpd32.exe Tcl initialization file > > > > proc ftpdInitUI {} { > > > > controlCreate launch [list toolbarbutton \ > > -callback "LaunchFtpd" \ > > -name LaunchFtpd \ > > -tooltip "Launch FTP Server" \ > > -bitmap "[wtxPath host resource bitmaps launch > > controls]ftp.bmp"] > > > > menuItemInsert -statusmessage "Launch FTP Server" \ > > -callback {LaunchFtpd} \ > > -after -bypath \ > > {&Tools &Debugger...} "&FTP Server" > > > > } > > > > proc LaunchFtpd {} { > > toolLaunch "LaunchFtpd" "[wtxPath host x86-win32 > > bin]wftpd32.exe" > > "[wtxPath host x86-win32 bin]" 0 0 0 0 0 0 > > } > > end --------------------------- Newsgroups: comp.os.vxworks Subject: Debugger stepping problem Date: Wed, 25 Sep 2002 16:40:37 -0400 From: "Daniel G Waddington" Organization: Bell Laboratories, Lucent Technologies Message-ID: Hi, I am debugging Vxwork 5.1 on a PPC604 MTX target over networked WDB. When I first load the module it debugs and steps fine. However, when I reload the module (because I've made a change), and re-debug it, it will stop at break points but when you try step/next it runs away. The only way to solve it is through a reboot. Any ideas? DAN --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Module loading problems Date: 25 Sep 2002 16:22:42 -0700 From: alex.mo@attbi.com (alex mo) Organization: http://groups.google.com/ Message-ID: References: <6046c736.0209240057.2a686e88@posting.google.com> san_blabla@hotmail.com (San Blabla) wrote in message news:<6046c736.0209240057.2a686e88@posting.google.com>... > Hi, everyone: > > I have writen an application on vxWorks, when I link it into the > system image, everything is O.K. But when I don't link it into my > system image and try to load it as a module using ld() function, the > error message "Undefined symbol: get_myaddress" occured. I found that > this function is in a vxWorks library, so I guess it is because that I > didn't link my application object to the system image, the linker > could not know anybody would call get_myaddress() so it didn't link in > the corresponding object in the library. I thought two ways to solve > this problem, one is to write some dummy functions to call > get_myaddress() in the system image, the other one is to link the > object from the library with my application object by hand. Both way > seem to be kind of ugly. Can anybody tell me how to solve this problem > in a better way? > > Regards, > Blabla. HI Blabla, The reason for the error is that the dynamic linker can not resolve the "get_myaddress" symbol because somehow your application is calling this function. The function is belong to the RPC library. My suggestion is , in your Tornado project under \ include and rebuild your system image then try to load your application. Good luck. - --alex --------------------------- Newsgroups: comp.os.vxworks Subject: Re: HELP !!! `netJobRing' undeclared ?? Date: 25 Sep 2002 16:44:00 -0700 From: alex.mo@attbi.com (alex mo) Organization: http://groups.google.com/ Message-ID: References: <7254c696.0209251019.633ed7bd@posting.google.com> zhouzi@yahoo.com (Zi Zhou) wrote in message news:<7254c696.0209251019.633ed7bd@posting.google.com>... > Hi, > > I am trying to use rngIsEmpty(netJobRing) to check if my netjob queue > still have empty entries. I included rngLib.h, netLib.h, the > compilation complains 'netJobRing' undeclared. Since I can't find any > detailed info about netJobRing, I am assuming it's a global variable > that's defined in one of the Lib.h, Am I right? What's the proper way > to check if netjob queue still has empty entry and how should I use > netJobRing? I am using Tornado 2.0.2 > > thanks > > Zi Hi Zi, No, you are NOT right, sorry, the is not a golbal variable since I couldn't find the symbol from the shell. This is the return value from function which you need to check if netjob queue is Empty. is using that ring ID to check if it has something to do. I don't know of any way to get id now, if I find any other info. I let you know. - --alex --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Getting crazy with VXWSymTab ... Can anyone look at this? ... please! Date: 25 Sep 2002 17:01:10 -0700 From: alex.mo@attbi.com (alex mo) Organization: http://groups.google.com/ Message-ID: References: <8ee874e3.0209250645.136ad5a7@posting.google.com> maba@combra.se (Marco) wrote in message news:<8ee874e3.0209250645.136ad5a7@posting.google.com>... > Hi! > I'm trying to use the class VXWSymTab as an hash table. > But even this simple code doesn't want to run. > (memSysPartId is a global variable that stores the ID of the system > memory. Such a variable is declared in memLib.h.) > > ========================================= > #include "vxWorks.h" > #include "memLib.h" > #include "vxwSymLib.h" > > > int run(void){ > > VXWSymTab* test = new VXWSymTab(6, FALSE, memSysPartId); > > return 0; > } > ========================================= > > The instantiation fails because symTblCreate returns 0! > Am I missing something? > > Thanks in advance! > > Marco Basile > > > ========================================= > VXWSymTab (int hashSizeLog2, BOOL sameNameOk, PART_ID symPartId) > : stid_ (symTblCreate (hashSizeLog2, sameNameOk, symPartId)) > { > if (stid_ == 0) > vxwThrowErrno (); > } > ========================================= Hi Macro, As far as I remebered the reason that function return NULL if memory for your hash table (in your case for 64 entries> is NOT sufficient. Now do you have enough memory? simple example that I found, SYMTAB_ID sysSymTbl; /* system symbol table id */ #define SYM_TBL_HASH_SIZE_LOG2 8 /* 256 entry hash table symbol table */ sysSymTbl = symTblCreate (SYM_TBL_HASH_SIZE_LOG2, TRUE, memSysPartId); - --alex --------------------------- Newsgroups: comp.os.vxworks Subject: Does VxWorks offer True SMP / multiprocessor support ? Date: Wed, 25 Sep 2002 17:32:17 -0700 From: "Dave Francheski" Organization: Advanced Telcom Group Inc. Internet Services Message-ID: Can anybody tell me if VxWorks offers true SMP (symmetric multi-processing) support? I know the VxMP option offers shared-memory objects (like semaphores, sockets, etc.) between tasks running on separate processors, but this is not at all SMP. An SMP kernel can schedule tasks/threads to execute on any available processor in the system. Regards, Dave Francheski --------------------------- Newsgroups: comp.os.vxworks Subject: Re: HELP !!! `netJobRing' undeclared ?? Date: Thu, 26 Sep 2002 09:21:20 +0800 From: "Nison.liu" Organization: Bentium Ltd. (CN99) Message-ID: References: <7254c696.0209251019.633ed7bd@posting.google.com> Hi Zhou zi: netJobRing is defined in netLib.c and it is a local variable and only be used in netLib.c, following is corresponding code in netLib.c: LOCAL RING_ID netJobRing; ..... if ((netJobRing = rngCreate (JOB_RING_SIZE)) == (RING_ID) NULL) panic ("netLibInit: couldn't create job ring\n"); typedef struct { FUNCPTR routine; /* routine to be called */ int param1; /* arg to routine */ int param2; int param3; int param4; int param5; } TODO_NODE; #define JOB_RING_SIZE (85 * sizeof (TODO_NODE)) if you want to has the ability access netJobRing besides netLib.c, so the only way is modify TMS source code and re-compile the kernel:-) Nison.liu Utstarcom Shenzhen R&D center "alex mo" ??????:b684bded.0209251544.3a62a37b@posting.google.com... > zhouzi@yahoo.com (Zi Zhou) wrote in message news:<7254c696.0209251019.633ed7bd@posting.google.com>... > > Hi, > > > > I am trying to use rngIsEmpty(netJobRing) to check if my netjob queue > > still have empty entries. I included rngLib.h, netLib.h, the > > compilation complains 'netJobRing' undeclared. Since I can't find any > > detailed info about netJobRing, I am assuming it's a global variable > > that's defined in one of the Lib.h, Am I right? What's the proper way > > to check if netjob queue still has empty entry and how should I use > > netJobRing? I am using Tornado 2.0.2 > > > > thanks > > > > Zi > > > Hi Zi, > > No, you are NOT right, sorry, the is not a golbal > variable since I couldn't find the symbol from the shell. This is the > return value from function which you need to check if > netjob queue is Empty. is using that ring ID to check if it > has something to do. I don't know of any way to get id > now, if I find any other info. I let you know. > > --alex --------------------------- Newsgroups: comp.os.vxworks Subject: Re: The FTP server access Date: Thu, 26 Sep 2002 01:49:48 GMT From: "drdiags" Organization: AT&T Broadband Message-ID: <%Itk9.76718$gA4.27187@sccrnsc02> References: <1fac8c75.0209241138.5a6c09bb@posting.google.com> <1fac8c75.0209251237.236a790a@posting.google.com> Robert, I tried this today and I couldn't even get the button, so you are way ahead of me. I happen to have the RTI tools installed with my Tornado and I looked at how they added buttons and added to the menu. I noticed that they used menuItemAppend instead of menuItemInsert, but in my case it still didn't matter. In your LaunchFtpd proc, you could put the following to help troubleshoot your problem: proc LaunchFtpd {} { messageBox -ok "Hello World" } This way clicking the button should let you know your procedure is linked to the button. Another option is to start the Tornado IDE from the DOS command line after running the torVars.bat file, do this: tornado -C /* This will put an option on the menu to show a hidden console */ Under the Window option you should now see the Tornado Console as a clickable option. Now whenever you put a 'puts "Hello World" ' , it will go out to this DOS Shell type window. There is a tech tip on this if you are interested. So, you now have two ways to troubleshoot your Tcl proc. You could also see if the toolLaunch returns an error or not (if it does, on-line documentation should help). Good luck! BTW, you are using Windows and not UNIX, right? "robert" wrote in message news:1fac8c75.0209251237.236a790a@posting.google.com... > Could you try making the last line in your tcl script the > following: > > ftpdInitUI > > This should now put the button where you want it.... > > Yes! if I do that it ceate a button and placed it into the Tornado > IDE, but what i see it doesn't invoke "menuItemInsert" and doesn't run > application - ftp. > Maybe some ideas... > > > > "drdiags" wrote in message news:... > > Robert, > > > > When is the ftpdInitUI() proc called? The file will be source after the > > 00xxx.tcl > > files, but all that does is make the proc available. You still have to call > > the > > InitUI routine. Could you try making the last line in your tcl script the > > following: > > > > ftpdInitUI > > > > This should now put the button where you want it.... > > > > "robert" wrote in message > > news:1fac8c75.0209241138.5a6c09bb@posting.google.com... > > > Hi guys, > > > The FTP server provided by Wind River is accessible from the Window > > > Start Menu. > > > I'm tryig to create additional button to the Tornado to accomplish the > > > same, but this file doesn't create nohing, I also try to comment proc > > > ftpdInitUI and use "controlCreate" and "menuItemInsert" out of proc - > > > same thing, maybe some ideas what I should do? > > > thank you, > > > Robert > > > > > > > > > > > > I copy the file 01Ftp.win32.tcl, which is provided > > > below, into the following directory: > > > $WIND_BASE\host\resource\tcl\app-config\Tornado > > > Copy the icon image, ftp.bmp, into: > > > $WIND_BASE\host\resource\bitmaps\Launch\controls > > > > > > > > > # 01Ftp.win32.tcl - wftpd32.exe Tcl initialization file > > > > > > proc ftpdInitUI {} { > > > > > > controlCreate launch [list toolbarbutton \ > > > -callback "LaunchFtpd" \ > > > -name LaunchFtpd \ > > > -tooltip "Launch FTP Server" \ > > > -bitmap "[wtxPath host resource bitmaps launch > > > controls]ftp.bmp"] > > > > > > menuItemInsert -statusmessage "Launch FTP Server" \ > > > -callback {LaunchFtpd} \ > > > -after -bypath \ > > > {&Tools &Debugger...} "&FTP Server" > > > > > > } > > > > > > proc LaunchFtpd {} { > > > toolLaunch "LaunchFtpd" "[wtxPath host x86-win32 > > > bin]wftpd32.exe" > > > "[wtxPath host x86-win32 bin]" 0 0 0 0 0 0 > > > } > > > end --------------------------- Newsgroups: comp.os.vxworks Subject: Re: BSP for TQM8xxL module Date: Thu, 26 Sep 2002 06:47:23 +0000 (UTC) From: Johan Borkhuis Organization: Proxim Corp. Message-ID: References: "Denis Dubrovin" wrote: > Hi! > > Where are I can get BSP for vxWorks 5.4 for my TQM8xxL module? Ask your FAE to do a search on their internal database and/or mailinglist. Or contact the manufacturer for support. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: HELP !!! `netJobRing' undeclared ?? Date: Thu, 26 Sep 2002 06:49:56 +0000 (UTC) From: Johan Borkhuis Organization: Proxim Corp. Message-ID: References: <7254c696.0209251019.633ed7bd@posting.google.com> "Nison.liu" wrote: > if you want to has the ability access netJobRing besides netLib.c, so > the only way is modify TMS source code and re-compile the kernel:-) Or, if you don't have the source code, ask your FAE to recompile this module for you with this symbol global. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Graphs in Tornado 2.0 Date: Mon, 23 Sep 2002 10:37:19 +0200 From: "pmanente" Organization: Cisco Systems Inc. Message-ID: <1033029941.20269@sj-nntpcache-5> This is a multi-part message in MIME format. - ------=_NextPart_000_000A_01C262ED.317AAE50 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Hi folks, I need to introduce Directed Acyclic Graphs as data structures but = Tornado 2.0 seems not to have such facility nor a useful library. I've found example code that's based on C++ STL, but Tornado doesn't = have STL library. Is there anybody who had the same problem or can suggest a link to = source code (C/C++) that can be ported to PPC target architecture? Thanks in advance, P.M> - ------=_NextPart_000_000A_01C262ED.317AAE50 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
Hi folks,
    I need to introduce = Directed=20 Acyclic Graphs as data structures but Tornado 2.0 seems not to have = such facility nor a useful library.
I've found example code that's based on = C++ STL,=20 but Tornado doesn't have STL library.
Is there anybody who had the same = problem or=20 can suggest a link to source code (C/C++) that can be ported to PPC = target=20 architecture?
Thanks in advance,
   =20 P.M>
- ------=_NextPart_000_000A_01C262ED.317AAE50-- --------------------------- Newsgroups: comp.os.vxworks Subject: Unable to access other hosts from VxSim using ULIP on win2k Date: 26 Sep 2002 02:30:50 -0700 From: eyal.rozenblat@seabridgenetworks.com (Eyal Rozenblat) Organization: http://groups.google.com/ Message-ID: <12f96c64.0209260130.7835d5b8@posting.google.com> I have installed ULIP and patches on my Win2k machine. Indeed it pings this machine, the host, and any VxSim instance I run on this host. Both directions. But attempting to ping another host (of the same subnet as my host) fails. I have examined it with Ethereal and saw that the ping packet have both source and destination MAC addresses incorrect. They are 00-00-00-00-00-fe and 00-00-00-00-00-01 . Any tip? Thanks, Eyal --------------------------- Newsgroups: comp.os.vxworks Subject: Searching for old MVME 68k boards Date: Thu, 26 Sep 2002 11:40:23 +0200 From: Alexis PEUCH Organization: IFREMER Message-ID: <3D92D607.410F78EA@ifremer.fr> Hello We are searching for these Motorola VME boards: MV162-263 and MV162-01 (with IP modules or not), manufactured in 1994-1996. If you have such boards (and don't use them anymore), we will be very happy to purchase them. Alexis Peuch http://www.ifremer.fr/fleet/ --------------------------- Newsgroups: comp.os.vxworks Subject: pinging the target Date: Thu, 26 Sep 2002 17:57:43 +0800 From: "IMS" Organization: IMS Netvigator Message-ID: Hi experts, i have a runtime vxWorks image (5.4.2) burnt onto a MIPS platform. On the target shell i find that i can't ping to any host on the same network nor can i successfully ping to it from another host. ifShow() gives the following: ene (unit number 1): Flags: (0x8022) DOWN BROADCAST MULTICAST ARP Type: ETHERNET_CSMACD Internet address: 10.128.26.47 Broadcast address: 10.128.26.255 Netmask 0xff000000 Subnetmask 0xffffff00 Ethernet address is 00:e0:4c:12:34:22 Metric is 0 Maximum Transfer Unit size is 1500 0 octets received 0 octets sent 4110 packets received 0 packets sent 4110 unicast packets received 0 unicast packets sent 0 non-unicast packets received 0 non-unicast packets sent 0 input discards 0 input unknown protocols 0 input errors 0 output errors 0 collisions; 0 dropped lo (unit number 0): Flags: (0x8069) UP LOOPBACK MULTICAST ARP RUNNING Type: SOFTWARE_LOOPBACK Internet address: 127.0.0.1 Netmask 0xff000000 Subnetmask 0xff000000 Metric is 0 Maximum Transfer Unit size is 32768 0 packets received; 0 packets sent 0 multicast packets received 0 multicast packets sent 0 input errors; 0 output errors 0 collisions; 0 dropped value = 29 = 0x1d - -> version It leads me to think that packets can be properly received. Just that nothing can be sent out. Here i have a question. Do i need to suspect the interrupt setting of the NE2000 end device? Can you suggest any areas that i should look into? Thanks you all! --------------------------- Newsgroups: comp.os.vxworks Subject: UsrNetMuxCfg Date: Thu, 26 Sep 2002 08:49:58 -0400 From: david lindauer Message-ID: <3D930276.D5F40EEB@notifier-is.net> hi, the above file appears in some installations of tornado but not others. When we upgraded to windows 2000, the symbol maxMuxBinds became undefined at link time. We redid our projects from scratch which would seemingly get rid of the dependence if it really wasn't installed, but still had the same problem. Anyone know what is up, how this will affect us if we just comment out that line, or how to get maxMuxBinds symbol installed properly? Thanks, David --------------------------- Newsgroups: comp.os.vxworks Subject: easy way to store BSP config in text form? Date: Thu, 26 Sep 2002 14:53:11 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1033044796.811659@newsreader1.wirehub.nl> Hi, For documentation perposes, I'd like to "dump" my BSP config (which options are on or off) into somekind of text form. Please help. btw, in whch files in this info stored? Gr Renee. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Newbie Starter Kit Date: Thu, 26 Sep 2002 09:12:03 -0400 From: Tasvo Organization: Bell Sympatico Message-ID: <3D9307A3.B8F71755@sympatico.ca> References: Thanks, Harry! I have the same interest as Blackie; thus the thanky note :) TV harry mudd wrote: > here's a good link > http://www.onesmartclick.com/rtos/rtos.html > "Blackie Beard" wrote in message > news:sA8j9.6967$7i2.195@nwrddc02.gnilink.net... > > ============================== > > I want to learn vxWorks in my spare time. > > How do I get started for very cheap? > > Is there a good kit, and what books are > > recommended. My background is H/W, > > S/W, F/W and SoC, but never studied > > any OS. Thanks in advance! > > > > BB > > ============================== > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: use of bootline Date: Thu, 26 Sep 2002 21:38:20 +0800 From: "Yung" Organization: IMS Netvigator Message-ID: Hi all, recently I am using Tornado 2.0 to develop an application of vxWorks. I have this query. While a bootline is necessary for the bootROM to retrieve a runtime image from a network device (or maybe other local devices), why is it also needed for the runtime image itself? I have found out that the runtime image actually derives network parameters from the bootline (done by routines called in prjConfig()). But it doesn't actually boot. Is my finding correct? I am very new to this platform development. Any advice very appreciated. Thanks. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: easy way to store BSP config in text form? Date: Thu, 26 Sep 2002 21:49:17 +0800 From: "Yung" Organization: IMS Netvigator Message-ID: References: <1033044796.811659@newsreader1.wirehub.nl> afaik, BSP\config.h and target\config\all\configAll.h have all the options that govern the build of a BSP. "Renee Teunissen" wrote in message news:1033044796.811659@newsreader1.wirehub.nl... > > Hi, > > For documentation perposes, I'd like to "dump" my BSP config (which > options are on or off) into somekind of text form. Please help. > > btw, in whch files in this info stored? > > Gr > Renee. > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: easy way to store BSP config in text form? Date: Thu, 26 Sep 2002 16:06:08 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1033049178.8016@newsreader1.wirehub.nl> References: <1033044796.811659@newsreader1.wirehub.nl> Thanx, I know. I'm looking howto "convert" those project files of Tornado into something readable for documentation purposes. Puttin' those headers in a word doc, is not what I'm looking for. Thanx anyway for youre answer.. Renee. - ------ Renee Teunissen PTS Software bv Meerweg 7, 1405BA Bussum, NL. email: firstnamelastnamepts.nl "Yung" wrote in message news:amv392$cu35@imsp212.netvigator.com... > afaik, BSP\config.h and target\config\all\configAll.h have all the options > that govern the build of a BSP. > > "Renee Teunissen" > agree.with.these.terms@mail.zevenaar.org> wrote in message > news:1033044796.811659@newsreader1.wirehub.nl... > > > > Hi, > > > > For documentation perposes, I'd like to "dump" my BSP config (which > > options are on or off) into somekind of text form. Please help. > > > > btw, in whch files in this info stored? > > > > Gr > > Renee. > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Vxworks downloadable image size Date: Thu, 26 Sep 2002 16:08:06 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1033049294.454544@newsreader1.wirehub.nl> References: have you tried removing all debug symbols? cheers, Renee. - -------- Renee Teunissen PTS Software bv "Manjunath Bhat" wrote in message news:amv1aj$npd$1@overload.lbl.gov... > Hi, > I'm currently using tornado II evaluation copy. I created a downloadable > image file (.out file) which is 14MB approx. But when I do the actual > download on the target server (simulator) it is just 4MB. Why is such a huge > difference in the downloadable image and the actual downloaded image ? Is it > normal? > > MRBhat. > > > > _________________________________________________________________ > Send and receive Hotmail on your mobile device: http://mobile.msn.com > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: use of bootline Date: Thu, 26 Sep 2002 10:18:09 -0400 From: david lindauer Message-ID: <3D931721.4C2D33BA@notifier-is.net> References: Yung wrote: > Hi all, > > recently I am using Tornado 2.0 to develop an application of vxWorks. I have > this query. > While a bootline is necessary for the bootROM to retrieve a runtime image > from a network device (or maybe other local devices), why is it also needed > for the runtime image itself? I have found out that the runtime image > actually derives network parameters from the bootline (done by routines > called in prjConfig()). But it doesn't actually boot. Is my finding correct? > I am very new to this platform development. Any advice very appreciated. > Thanks. yes some of the boot line settings are inherited by the application image. This allows you to dynamically change your network settings either by distributing different boot code to different customers (without changing the application) or by changing the settings by hand in the boot code dialogs. In our case the production distributions we use have boot code set to not go on the network at all; after the app comes up it uses appropriate kernel calls to set network properties (which we configure through the serial port on our board). Of course for debug we have boot code that will go out and read the application image from one of our PCs... at least when we can, some of our projects use WindWeb and that is not very conducive to the normal debug process of downloading application modules on the fly. For those we initially bootstrap the code without using tornado at all, then use FTP to download successive versions of the kernel/application for testing. Did they ever upgrade DOSFS to work with current versions of pre-formatted disks? AFAIK when we were developing it didn't work because MS had switched to a new partition enumeration in the MBR. We fixed it by using MSDOS 6.22 format utility. But it is getting really hard to come by MSDOS 6.22 disks to do the formatting during production, much easier if we could just live with the pre-formatted disk. David --------------------------- Newsgroups: comp.os.vxworks Subject: visionICEII / STK8xxL problem Date: Thu, 26 Sep 2002 18:51:54 +0400 From: "Denis Dubrovin" Organization: WEBPlus Ltd. Message-ID: I use visionICEII with TQ-component STK8xxL starter kit board and have a problem with BDM connection. When I connect BDM connector to board and power it up , visionCLICK says "Can`t enter BKM mode! Check cable and settings!" Anyone has same problems? What I must check for take BDM connection working? - -- Regards! Den --------------------------- Newsgroups: comp.os.vxworks Subject: Re: pinging the target Date: Thu, 26 Sep 2002 17:20:58 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Hey Miller, are you really a General ? read again your own post, and search for the word "DOWN", there will you find the wisdom, or at least the answer your're after is there, as to wisdom, that'll have to wait till your next incarnation. Leonid "IMS" wrote in message news:amulmu$ctt1@imsp212.netvigator.com... > Hi experts, i have a runtime vxWorks image (5.4.2) burnt onto a MIPS > platform. On the target shell i find that i can't ping to any host on the > same network nor can i successfully ping to it from another host. ifShow() > gives the following: > > ene (unit number 1): > Flags: (0x8022) DOWN BROADCAST MULTICAST ARP > Type: ETHERNET_CSMACD > Internet address: 10.128.26.47 > Broadcast address: 10.128.26.255 > Netmask 0xff000000 Subnetmask 0xffffff00 > Ethernet address is 00:e0:4c:12:34:22 > Metric is 0 > Maximum Transfer Unit size is 1500 > 0 octets received > 0 octets sent > 4110 packets received > 0 packets sent > 4110 unicast packets received > 0 unicast packets sent > 0 non-unicast packets received > 0 non-unicast packets sent > 0 input discards > 0 input unknown protocols > 0 input errors > 0 output errors > 0 collisions; 0 dropped > lo (unit number 0): > Flags: (0x8069) UP LOOPBACK MULTICAST ARP RUNNING > Type: SOFTWARE_LOOPBACK > Internet address: 127.0.0.1 > Netmask 0xff000000 Subnetmask 0xff000000 > Metric is 0 > Maximum Transfer Unit size is 32768 > 0 packets received; 0 packets sent > 0 multicast packets received > 0 multicast packets sent > 0 input errors; 0 output errors > 0 collisions; 0 dropped > value = 29 = 0x1d > -> version > > It leads me to think that packets can be properly received. Just that > nothing can be sent out. Here i have a question. Do i need to suspect the > interrupt setting of the NE2000 end device? Can you suggest any areas that i > should look into? Thanks you all! > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: easy way to store BSP config in text form? Date: Thu, 26 Sep 2002 11:31:20 -0400 From: Joe Durusau Organization: Lockheed Martin Corporation Message-ID: <3D932848.94237608@lmco.com> References: <1033044796.811659@newsreader1.wirehub.nl> <1033049178.8016@newsreader1.wirehub.nl> Renee Teunissen wrote: > > Thanx, I know. > > I'm looking howto "convert" those project files of Tornado into > something readable for documentation purposes. Puttin' those headers > in a word doc, is not what I'm looking for. > > Thanx anyway for youre answer.. > Renee. > ------ > Renee Teunissen > PTS Software bv > Meerweg 7, 1405BA Bussum, NL. > email: firstnamelastnamepts.nl > > "Yung" wrote in message > news:amv392$cu35@imsp212.netvigator.com... > > afaik, BSP\config.h and target\config\all\configAll.h have all the > options > > that govern the build of a BSP. > > > > "Renee Teunissen" > > > agree.with.these.terms@mail.zevenaar.org> wrote in message > > news:1033044796.811659@newsreader1.wirehub.nl... > > > > > > Hi, > > > > > > For documentation perposes, I'd like to "dump" my BSP config > (which > > > options are on or off) into somekind of text form. Please help. > > > > > > btw, in whch files in this info stored? > > > > > > Gr > > > Renee. > > > > > > > > > > > > > Which files??? All the project files that tornado gens up are just ascii text, at least on my system. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Unable to access other hosts from VxSim using ULIP on win2k Date: Thu, 26 Sep 2002 18:51:49 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <12f96c64.0209260130.7835d5b8@posting.google.com> Eyal, the problem is with routing, the IP addresses used by VxSim are not know outside of your specific Win2k machine, so the ping requests may actually go out, but never come back .. hahaha. A few "route add" moves should do the trick. - - Leonid "Eyal Rozenblat" wrote in message news:12f96c64.0209260130.7835d5b8@posting.google.com... > I have installed ULIP and patches on my Win2k machine. Indeed it pings > this machine, the host, and any VxSim instance I run on this host. > Both directions. > But attempting to ping another host (of the same subnet as my host) > fails. > I have examined it with Ethereal and saw that the ping packet have > both source and destination MAC addresses incorrect. They are > 00-00-00-00-00-fe and 00-00-00-00-00-01 . > Any tip? > Thanks, > Eyal --------------------------- Newsgroups: comp.os.vxworks Subject: END driver can't receive the packet larger than 60 bytes Date: Fri, 27 Sep 2002 01:02:35 +0800 From: "minskey73" Organization: Bentium Ltd. (CN99) Message-ID: Hi,all I write a END driver for AM79c976(for MPC860),but the driver receive only 64-bytes packet. The following is some piece of code ,who can tell why,please. netClBlkJoin (pClBlk, pCluster, len, NULL, 0, 0, 0); //here ,len is right netMblkClJoin (pMblk, pClBlk); /* make the packet data coherent */ LN_CACHE_INVALIDATE (pMblk->mBlkHdr.mData, len); pMblk->mBlkHdr.mData += pDrvCtrl->offset; pMblk->mBlkHdr.mLen = len; pMblk->mBlkHdr.mFlags |= M_PKTHDR; pMblk->mBlkPktHdr.len = len; DRV_LOG (DRV_DEBUG_RX, "Calling upper layer!\n", 1, 2, 3, 4, 5, 6); /* Deal with memory alignment. */ pNewCluster += pDrvCtrl->offset; /* Give receiver a new buffer */ LN_RMD_BUF_TO_ADDR (pRmd, pTemp, pNewCluster); /* Call the upper layer's receive routine. */ END_RCV_RTN_CALL(&pDrvCtrl->endObj, pMblk); --------------------------- Newsgroups: comp.os.vxworks Subject: Re: pinging the target Date: Fri, 27 Sep 2002 01:31:54 +0800 From: "Yung" Organization: IMS Netvigator Message-ID: References: oops, right. Thanks so much Leonid. As a freshman to the platform i wasn't even alarmed that the interface is not up and running yet. I was just trying to make sure every network initialization line in prjConfig.c was executed without returning error. What have i done or not done to make the interface go DOWN? I wanted to search for some documentation about the flags of ifShow() but didn't get any clue yet. How can i bring it back UP? Thanks again. (not really a General... sorry...:-p "Leonid Rosenboim" wrote in message news:newscache$mfx13h$yn6$1@lnews.actcom.co.il... > Hey Miller, are you really a General ? > > read again your own post, and search for the word "DOWN", there will you > find the wisdom, or at least the answer your're after is there, as to > wisdom, that'll have to wait till your next incarnation. > > Leonid > > "IMS" wrote in message > news:amulmu$ctt1@imsp212.netvigator.com... > > Hi experts, i have a runtime vxWorks image (5.4.2) burnt onto a MIPS > > platform. On the target shell i find that i can't ping to any host on the > > same network nor can i successfully ping to it from another host. ifShow() > > gives the following: > > > > ene (unit number 1): > > Flags: (0x8022) DOWN BROADCAST MULTICAST ARP > > Type: ETHERNET_CSMACD > > Internet address: 10.128.26.47 > > Broadcast address: 10.128.26.255 > > Netmask 0xff000000 Subnetmask 0xffffff00 > > Ethernet address is 00:e0:4c:12:34:22 > > Metric is 0 > > Maximum Transfer Unit size is 1500 > > 0 octets received > > 0 octets sent > > 4110 packets received > > 0 packets sent > > 4110 unicast packets received > > 0 unicast packets sent > > 0 non-unicast packets received > > 0 non-unicast packets sent > > 0 input discards > > 0 input unknown protocols > > 0 input errors > > 0 output errors > > 0 collisions; 0 dropped > > lo (unit number 0): > > Flags: (0x8069) UP LOOPBACK MULTICAST ARP RUNNING > > Type: SOFTWARE_LOOPBACK > > Internet address: 127.0.0.1 > > Netmask 0xff000000 Subnetmask 0xff000000 > > Metric is 0 > > Maximum Transfer Unit size is 32768 > > 0 packets received; 0 packets sent > > 0 multicast packets received > > 0 multicast packets sent > > 0 input errors; 0 output errors > > 0 collisions; 0 dropped > > value = 29 = 0x1d > > -> version > > > > It leads me to think that packets can be properly received. Just that > > nothing can be sent out. Here i have a question. Do i need to suspect the > > interrupt setting of the NE2000 end device? Can you suggest any areas that > i > > should look into? Thanks you all! > > > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: END driver can't receive the packet larger than 60 bytes Date: Thu, 26 Sep 2002 13:58:37 -0400 From: Varndell Engineering Organization: Posted Via Newsfeeds.com = SPEED+RETENTION+COMPLETION = http://www.newsfeeds.com Message-ID: <3D934ACD.286D0C2D@hotmail.com> References: minskey73 wrote: > > Hi,all > > I write a END driver for AM79c976(for MPC860),but the driver receive > only 64-bytes packet. The following is some piece of code ,who can tell > why,please. > > netClBlkJoin (pClBlk, pCluster, len, NULL, 0, 0, 0); //here ,len is > right > netMblkClJoin (pMblk, pClBlk); > > /* make the packet data coherent */ > LN_CACHE_INVALIDATE (pMblk->mBlkHdr.mData, len); > > pMblk->mBlkHdr.mData += pDrvCtrl->offset; Did you forget to set mData to pCluster before invalidating it and adding in offset, or did you just forget to post that code? Also, if you're running this on an 860, you needn't worry yourself about offset. Of course, if you plan to deploy this driver on other, possibly alignment challenged processors, then carry on. > pMblk->mBlkHdr.mLen = len; > pMblk->mBlkHdr.mFlags |= M_PKTHDR; > pMblk->mBlkPktHdr.len = len; > > DRV_LOG (DRV_DEBUG_RX, "Calling upper layer!\n", 1, 2, 3, 4, 5, 6); > > /* Deal with memory alignment. */ > pNewCluster += pDrvCtrl->offset; > > /* Give receiver a new buffer */ > LN_RMD_BUF_TO_ADDR (pRmd, pTemp, pNewCluster); > > /* Call the upper layer's receive routine. */ > END_RCV_RTN_CALL(&pDrvCtrl->endObj, pMblk); If you are not setting pMblk->mBlkHdr.mData to point to pCluster, then I don't see how this driver receives packets of any size at all. Regards, George Varndell - -- Varndell Engineering, LLC Embedded Systems Expertise Need a vxWorks END driver for AMD79C97x devices? Contact us. http://www.varndellengineering.com - -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- http://www.newsfeed.com The #1 Newsgroup Service in the World! - -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Getting crazy with VXWSymTab ... Can anyone look at this? ... please! Date: 26 Sep 2002 11:06:42 -0700 From: maba@combra.se (Marco) Organization: http://groups.google.com/ Message-ID: <8ee874e3.0209261006.c37a73@posting.google.com> References: <8ee874e3.0209250645.136ad5a7@posting.google.com> Hi Alex! I forgot to tell that I'm running with SIMNTgnu as target ... Do you think it has something to do with that? /Marco alex.mo@attbi.com (alex mo) wrote in message news:... > Hi Macro, > > As far as I remebered the reason that function return > NULL if memory for your hash table (in your case for 64 entries> is > NOT sufficient. Now do you have enough memory? > > simple example that I found, > > SYMTAB_ID sysSymTbl; /* system symbol table id */ > #define SYM_TBL_HASH_SIZE_LOG2 8 /* 256 entry hash table symbol table > */ > > > sysSymTbl = symTblCreate (SYM_TBL_HASH_SIZE_LOG2, TRUE, > memSysPartId); > > --alex --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Does VxWorks offer True SMP / multiprocessor support ? Date: Thu, 26 Sep 2002 13:57:57 -0500 From: Andrew Johnson Organization: APS, Argonne National Laboratory Message-ID: References: Dave Francheski wrote: > Can anybody tell me if VxWorks offers true SMP (symmetric multi-processing) > support? Yes, and the answer is No it doesn't. VxWorks is definitely aimed at single processor systems, the kernel is not designed for SMP use. Given their target market that's not terribly surprising, it really doesn't seem that there would be much point to using SMP in embedded systems (although no doubt there are quite a few exceptions out there to prove the rule). - - Andrew - -- "Life is what happens while you're busy making other plans." - John Lennon --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Graphs in Tornado 2.0 Date: Thu, 26 Sep 2002 14:01:41 -0500 From: Andrew Johnson Organization: APS, Argonne National Laboratory Message-ID: References: <1033029941.20269@sj-nntpcache-5> pmanente wrote: > I need to introduce Directed Acyclic Graphs as data structures but > Tornado 2.0 seems not to have such facility nor a useful library. I've > found example code that's based on C++ STL, but Tornado doesn't have > STL library. Actually the implementation of the STL that comes with Tornado 2.2 (not 2.0.2) is supposed to be significantly better than the older stuff, although I'm not speaking from experience here. You might want to ask your FAE when you're going to receive your upgrade to 2.2 and try that... - - Andrew - -- "Life is what happens while you're busy making other plans." - John Lennon --------------------------- Newsgroups: comp.os.vxworks Subject: Re: END driver can't receive the packet larger than 60 bytes Date: Thu, 26 Sep 2002 15:30:44 -0400 From: Varndell Engineering Organization: Posted Via Newsfeeds.com = SPEED+RETENTION+COMPLETION = http://www.newsfeeds.com Message-ID: <3D936064.E3D8F75A@hotmail.com> References: Duh, let me try this again... minskey73 wrote: > > Hi,all > > I write a END driver for AM79c976(for MPC860),but the driver receive > only 64-bytes packet. The following is some piece of code ,who can tell > why,please. > > netClBlkJoin (pClBlk, pCluster, len, NULL, 0, 0, 0); //here ,len is > right Using len for the 3rd argument to netClBlkJoin is wrong -- pass the actual size of the buffer instead. > netMblkClJoin (pMblk, pClBlk); > > /* make the packet data coherent */ > LN_CACHE_INVALIDATE (pMblk->mBlkHdr.mData, len); > As many END drivers as I've written, and I never noticed that the join functions actually set mData for you. I've never seen a driver that *didn't* set it explicitly, but it seems it is indeed unnecessary. > pMblk->mBlkHdr.mData += pDrvCtrl->offset; > pMblk->mBlkHdr.mLen = len; > pMblk->mBlkHdr.mFlags |= M_PKTHDR; > pMblk->mBlkPktHdr.len = len; > > DRV_LOG (DRV_DEBUG_RX, "Calling upper layer!\n", 1, 2, 3, 4, 5, 6); > > /* Deal with memory alignment. */ > pNewCluster += pDrvCtrl->offset; Why are you adding offset here? pNewCluster here will become the pCluster above -- after it's had a packet put in it. Do you account for this offset addition when you process this descriptor later, ie subtract offset before joining to clBlk? > > /* Give receiver a new buffer */ > LN_RMD_BUF_TO_ADDR (pRmd, pTemp, pNewCluster); > > /* Call the upper layer's receive routine. */ > END_RCV_RTN_CALL(&pDrvCtrl->endObj, pMblk); There's no smoking gun here that explains receiving only 64-byte packets. BTW, how do you know you're only receiving 64 bytes? Regards, George Varndell - -- Varndell Engineering, LLC Embedded Systems Expertise Need a vxWorks END driver for AMD79C97x devices? Contact us. http://www.varndellengineering.com - -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- http://www.newsfeed.com The #1 Newsgroup Service in the World! - -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: pinging the target Date: 26 Sep 2002 12:45:17 -0700 From: demorz@yahoo.com (Dennis) Organization: http://groups.google.com/ Message-ID: <4d0e703e.0209261145.6edb6d56@posting.google.com> References: Check the routing information. If it is not set up correctly, ping requests won't go through "IMS" wrote in message news:... > Hi experts, i have a runtime vxWorks image (5.4.2) burnt onto a MIPS > platform. On the target shell i find that i can't ping to any host on the > same network nor can i successfully ping to it from another host. ifShow() > gives the following: > > ene (unit number 1): > Flags: (0x8022) DOWN BROADCAST MULTICAST ARP > Type: ETHERNET_CSMACD > Internet address: 10.128.26.47 > Broadcast address: 10.128.26.255 > Netmask 0xff000000 Subnetmask 0xffffff00 > Ethernet address is 00:e0:4c:12:34:22 > Metric is 0 > Maximum Transfer Unit size is 1500 > 0 octets received > 0 octets sent > 4110 packets received > 0 packets sent > 4110 unicast packets received > 0 unicast packets sent > 0 non-unicast packets received > 0 non-unicast packets sent > 0 input discards > 0 input unknown protocols > 0 input errors > 0 output errors > 0 collisions; 0 dropped > lo (unit number 0): > Flags: (0x8069) UP LOOPBACK MULTICAST ARP RUNNING > Type: SOFTWARE_LOOPBACK > Internet address: 127.0.0.1 > Netmask 0xff000000 Subnetmask 0xff000000 > Metric is 0 > Maximum Transfer Unit size is 32768 > 0 packets received; 0 packets sent > 0 multicast packets received > 0 multicast packets sent > 0 input errors; 0 output errors > 0 collisions; 0 dropped > value = 29 = 0x1d > -> version > > It leads me to think that packets can be properly received. Just that > nothing can be sent out. Here i have a question. Do i need to suspect the > interrupt setting of the NE2000 end device? Can you suggest any areas that i > should look into? Thanks you all! --------------------------- Newsgroups: comp.os.vxworks Subject: arpresolve: can't allocate llinfo Date: 26 Sep 2002 20:26:54 GMT From: brewe@pnc.aps.anl.gov (Dale Brewe) Organization: PNC-CAT Message-ID: I'm trying to get vxWorks 5.4.2 working on a MVME2700 cpu (PowerPC), and have a problem apparently related to arp. I can get the cpu to download vxworks and my startup script, but I get the "arpresolve: can't allocate llinfo" msg, any time I try any further network access (e.g. when trying nfsMount). I don't see anything wrong w/ the routing table, which has only a couple of entries. The arp table, however, is empty. Trying arpAdd fails. The cpu doesn't respond to pings, but other hosts can get its ethernet address. I'd appreciate any suggestions. dale brewe --------------------------- Newsgroups: comp.os.vxworks Subject: How can I merge two Tornados into single one based on different BSP? Date: Thu, 26 Sep 2002 15:49:43 -0700 From: "Liang Yang" Organization: Arizona State University Message-ID: Hi, I have a Tornado 2.0 for MBX, I also have a Tornado 2.1 for IXP1200. But I don't want to install two tornado in my computer. How can I enbale my Tornado 2.0 for MBX to support IXP without installing it? I want to try to copy BSP from IXP tornado\target directory to the MBX tornado\target directory. When I create Projects, I can find IXP listed in the available BSP? But when I try to build projects, Tornado reports a lot of error? How can I merge this two different Tornado seamlessly? Thanks. Liang --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How can I merge two Tornados into single one based on different BSP? Date: Thu, 26 Sep 2002 19:24:17 -0400 From: joe durusau Organization: WEBUSENET.com Message-ID: <3D939721.E312031C@bellsouth.net> References: Liang Yang wrote: > Hi, > > I have a Tornado 2.0 for MBX, I also have a Tornado 2.1 for IXP1200. > But I don't want to install two tornado in my computer. How can I enbale > my Tornado 2.0 for MBX to support IXP without installing it? > > I want to try to copy BSP from IXP tornado\target directory to the MBX > tornado\target directory. When I create Projects, I can find IXP listed in > the available BSP? But when I try to build projects, Tornado reports a lot > of error? > > How can I merge this two different Tornado seamlessly? > > Thanks. > > Liang I'm not familiar with those boards, but if they are different arch, you can't do that. You could update the 2.0 job to 2.1, and manage to have just one tornado directory, poviding that you are willing to use the makefiles, rather than the GUI interface. If the boards and not (for example) both X86 or some other family, it will be much harder and impossible if you try to mix 2.0 and 2.1 Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Cannot capture OSPF packets from IP stack Date: 26 Sep 2002 18:17:25 -0700 From: lihuapeng@21cn.com (DragonSpring) Organization: http://groups.google.com/ Message-ID: <1d411918.0209261717.777d8f0a@posting.google.com> Environment: target: PPC860 rev. D4, Host: WinNT 4.0, Tornado 2.0/VxWorks 5.4 for PPC, Tornado 2.0 Cumulative Patch 2 (SPR T2CP2), NPT 5.4(TDK-13274), NPT patch(SPR 30894) We are developing our OSPF protocol (rfc2328). We want to get all OSPF packet from IP layer. So we use the code as following: int ospf_init() { FAST struct protosw *p_ospf_protocol_switch; if (_protoSwIndex >= ((sizeof(inetsw))/(sizeof(inetsw[0])))) return (-1) ; p_ospf_protocol_switch = &inetsw [_protoSwIndex]; if (p_ospf_protocol_switch == NULL) return (-1); if (p_ospf_protocol_switch->pr_domain != NULL) return (1); /* already initialized */ p_ospf_protocol_switch->pr_type = SOCK_RAW; p_ospf_protocol_switch->pr_domain = &inetdomain; p_ospf_protocol_switch->pr_protocol = IPPROTO_OSPF; p_ospf_protocol_switch->pr_flags = PR_ATOMIC|PR_ADDR; p_ospf_protocol_switch->pr_input = ospf_receive; p_ospf_protocol_switch->pr_output = ospf_send; p_ospf_protocol_switch->pr_ctlinput = raw_ctlinput; p_ospf_protocol_switch->pr_ctloutput = rip_ctloutput; p_ospf_protocol_switch->pr_usrreq = rip_usrreq; p_ospf_protocol_switch->pr_init = 0 ; p_ospf_protocol_switch->pr_fasttimo = 0; p_ospf_protocol_switch->pr_slowtimo = 0; p_ospf_protocol_switch->pr_drain = 0; p_ospf_protocol_switch->pr_sysctl = 0; _protoSwIndex++; ip_protox[IPPROTO_OSPF] = p_ospf_protocol_switch - inetsw; ... } void ospf_receive(struct mbuf *m) { /* determine which interface this OSPF packet is received from */ ... } void ospf_send(struct mbuf *m) { ... } We invoke ospf_init() in usrAppInit() (which is in usrAppInit.c) to start the OSPF process. Our target board has 2 interfaces: an 10/100 Ethernet interface and a WAN interface. The 10/100 Ethernet interface driver is motFecEnd.c, the WAN interface's device driver and END driver are written by ourselves. The Ethernet interface is a BROADCAST inteface, and the WAN interface is a POINT-TO-POINT interface. According to the TCP/IP and NPT architecture, the END driver delivers OSPF packets to TCP/IP stack. And here we use inetsw[] to get all OSPF packets received from all interfaces. The problem is that we can get those OSPF packets from the Ethernet interface, but we can not get OSPF packets from the WAN interface. But we have observed that the WAN interface's END driver has delivers the OSPF packets to MUX layer using netJobAdd(). But we cann't get them from TCP/IP stack. The OSPF Hello packet's destination IP is 224.0.0.5, and we have set the IFF_PROMISC | IFF_POINTOPOINT | IFF_MULTICAST | IFF_NOARP | IFF_RUNNING flags in END_OBJ_READY(). BTW, during initialization, We call in_addmulti() to add the WAN interface into the 224.0.0.5 multicast group, but in_addmulti() fails. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Unable to access other hosts from VxSim using ULIP on win2k Date: Fri, 27 Sep 2002 01:24:50 GMT From: "drdiags" Organization: AT&T Broadband Message-ID: References: <12f96c64.0209260130.7835d5b8@posting.google.com> Eyal, In a case I tried, I had two Windows Host with say IP addresses like so: Host # 1 192.168.101.10 Host # 2 192.168.101.9 vxSim # 1 on Host # 1 10.100.1.1 vxSim # 2 on Host # 2 10.100.1.2 on Host # 1 route add 10.100.1.2 192.168.101.9 on Host # 2 route add 10.100.1.1 192.168.101.10 I could then ping vxSim#1 to vxSim#2. To ping Host #2 from vxSim #1, I had to setup a route on Host #1: route add 10.100.1.1 10.100.1.253 (ULIP adapter IP address on Host #1) and something similiar on Host #2 for vxSim #2: route add 10.100.1.2. 10.100.1.252 (ULIP adapter IP address on Host #2) And as Lenoid mentions, you need to make sure the route on the targets knows to use the 10.100.1.x gateway to reach the 192.168.101.x network. This is all from memory, so YMMV. "Eyal Rozenblat" wrote in message news:12f96c64.0209260130.7835d5b8@posting.google.com... > I have installed ULIP and patches on my Win2k machine. Indeed it pings > this machine, the host, and any VxSim instance I run on this host. > Both directions. > But attempting to ping another host (of the same subnet as my host) > fails. > I have examined it with Ethereal and saw that the ping packet have > both source and destination MAC addresses incorrect. They are > 00-00-00-00-00-fe and 00-00-00-00-00-01 . > Any tip? > Thanks, > Eyal --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Vxworks downloadable image size Date: Fri, 27 Sep 2002 01:29:49 GMT From: "drdiags" Organization: AT&T Broadband Message-ID: References: If you use the objdump --section-headers image.out file, you will see that the actual part of the image that requires allocation on the target is listed as ALLOC or you can use the size which will show the part of the image that will actually require on target memory. The rest is debug information. "Manjunath Bhat" wrote in message news:amv1aj$npd$1@overload.lbl.gov... > Hi, > I'm currently using tornado II evaluation copy. I created a downloadable > image file (.out file) which is 14MB approx. But when I do the actual > download on the target server (simulator) it is just 4MB. Why is such a huge > difference in the downloadable image and the actual downloaded image ? Is it > normal? > > MRBhat. > > > > _________________________________________________________________ > Send and receive Hotmail on your mobile device: http://mobile.msn.com > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: easy way to store BSP config in text form? Date: Fri, 27 Sep 2002 01:41:59 GMT From: "drdiags" Organization: AT&T Broadband Message-ID: References: <1033044796.811659@newsreader1.wirehub.nl> <1033049178.8016@newsreader1.wirehub.nl> Renee, There are some command line Project Tcl scripts you can use to try to get what you want. Look up the cmpXXX routines in your on-line help or the %WIND_BASE%\host\resource\tcl\app-config\Project\cmpScriptLib.tcl file. It may not help, but you can decide what information can be obtained using these routines. "Renee Teunissen" wrote in message news:1033049178.8016@newsreader1.wirehub.nl... > Thanx, I know. > > I'm looking howto "convert" those project files of Tornado into > something readable for documentation purposes. Puttin' those headers > in a word doc, is not what I'm looking for. > > Thanx anyway for youre answer.. > Renee. > ------ > Renee Teunissen > PTS Software bv > Meerweg 7, 1405BA Bussum, NL. > email: firstnamelastnamepts.nl > > "Yung" wrote in message > news:amv392$cu35@imsp212.netvigator.com... > > afaik, BSP\config.h and target\config\all\configAll.h have all the > options > > that govern the build of a BSP. > > > > "Renee Teunissen" > > > agree.with.these.terms@mail.zevenaar.org> wrote in message > > news:1033044796.811659@newsreader1.wirehub.nl... > > > > > > Hi, > > > > > > For documentation perposes, I'd like to "dump" my BSP config > (which > > > options are on or off) into somekind of text form. Please help. > > > > > > btw, in whch files in this info stored? > > > > > > Gr > > > Renee. > > > > > > > > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: pinging the target Date: Fri, 27 Sep 2002 09:43:08 +0800 From: "Nison.liu" Organization: Bentium Ltd. (CN99) Message-ID: References: <4d0e703e.0209261145.6edb6d56@posting.google.com> Do you have include Ping client software and ICMP component in your vxWorks image? Nison.liu "Dennis" ??????:4d0e703e.0209261145.6edb6d56@posting.google.com... > Check the routing information. If it is not set up correctly, > ping requests won't go through > > "IMS" wrote in message news:... > > Hi experts, i have a runtime vxWorks image (5.4.2) burnt onto a MIPS > > platform. On the target shell i find that i can't ping to any host on the > > same network nor can i successfully ping to it from another host. ifShow() > > gives the following: > > > > ene (unit number 1): > > Flags: (0x8022) DOWN BROADCAST MULTICAST ARP > > Type: ETHERNET_CSMACD > > Internet address: 10.128.26.47 > > Broadcast address: 10.128.26.255 > > Netmask 0xff000000 Subnetmask 0xffffff00 > > Ethernet address is 00:e0:4c:12:34:22 > > Metric is 0 > > Maximum Transfer Unit size is 1500 > > 0 octets received > > 0 octets sent > > 4110 packets received > > 0 packets sent > > 4110 unicast packets received > > 0 unicast packets sent > > 0 non-unicast packets received > > 0 non-unicast packets sent > > 0 input discards > > 0 input unknown protocols > > 0 input errors > > 0 output errors > > 0 collisions; 0 dropped > > lo (unit number 0): > > Flags: (0x8069) UP LOOPBACK MULTICAST ARP RUNNING > > Type: SOFTWARE_LOOPBACK > > Internet address: 127.0.0.1 > > Netmask 0xff000000 Subnetmask 0xff000000 > > Metric is 0 > > Maximum Transfer Unit size is 32768 > > 0 packets received; 0 packets sent > > 0 multicast packets received > > 0 multicast packets sent > > 0 input errors; 0 output errors > > 0 collisions; 0 dropped > > value = 29 = 0x1d > > -> version > > > > It leads me to think that packets can be properly received. Just that > > nothing can be sent out. Here i have a question. Do i need to suspect the > > interrupt setting of the NE2000 end device? Can you suggest any areas that i > > should look into? Thanks you all! --------------------------- Newsgroups: comp.os.vxworks Subject: Re: pinging the target Date: Fri, 27 Sep 2002 11:56:41 +0800 From: "Yung" Organization: IMS Netvigator Message-ID: References: <4d0e703e.0209261145.6edb6d56@posting.google.com> Hi Dennis, I have checked the routing information and they are correct. An earlier reply from Leonid reminded me that the ene1 interface was actually DOWN and not running. I am still trying to figure out the reason. I have checked that the interrupt vector setting is correct. Any suggestion? "Dennis" wrote in message news:4d0e703e.0209261145.6edb6d56@posting.google.com... > Check the routing information. If it is not set up correctly, > ping requests won't go through > > "IMS" wrote in message news:... > > Hi experts, i have a runtime vxWorks image (5.4.2) burnt onto a MIPS > > platform. On the target shell i find that i can't ping to any host on the > > same network nor can i successfully ping to it from another host. ifShow() > > gives the following: > > > > ene (unit number 1): > > Flags: (0x8022) DOWN BROADCAST MULTICAST ARP > > Type: ETHERNET_CSMACD > > Internet address: 10.128.26.47 > > Broadcast address: 10.128.26.255 > > Netmask 0xff000000 Subnetmask 0xffffff00 > > Ethernet address is 00:e0:4c:12:34:22 > > Metric is 0 > > Maximum Transfer Unit size is 1500 > > 0 octets received > > 0 octets sent > > 4110 packets received > > 0 packets sent > > 4110 unicast packets received > > 0 unicast packets sent > > 0 non-unicast packets received > > 0 non-unicast packets sent > > 0 input discards > > 0 input unknown protocols > > 0 input errors > > 0 output errors > > 0 collisions; 0 dropped > > lo (unit number 0): > > Flags: (0x8069) UP LOOPBACK MULTICAST ARP RUNNING > > Type: SOFTWARE_LOOPBACK > > Internet address: 127.0.0.1 > > Netmask 0xff000000 Subnetmask 0xff000000 > > Metric is 0 > > Maximum Transfer Unit size is 32768 > > 0 packets received; 0 packets sent > > 0 multicast packets received > > 0 multicast packets sent > > 0 input errors; 0 output errors > > 0 collisions; 0 dropped > > value = 29 = 0x1d > > -> version > > > > It leads me to think that packets can be properly received. Just that > > nothing can be sent out. Here i have a question. Do i need to suspect the > > interrupt setting of the NE2000 end device? Can you suggest any areas that i > > should look into? Thanks you all! --------------------------- Newsgroups: comp.os.vxworks Subject: Re: pinging the target Date: Fri, 27 Sep 2002 11:58:27 +0800 From: "Yung" Organization: IMS Netvigator Message-ID: References: <4d0e703e.0209261145.6edb6d56@posting.google.com> Yes, Nison, they are included, done at the Tornado Project. I just wonder why packets seem to be coming in, even when the interface is down. Thanks for your advice :-) "Nison.liu" wrote in message news:an0d4n$k5v$1@debian.bentium.com... > Do you have include Ping client software and ICMP component in your vxWorks > image? > > Nison.liu > > "Dennis" > ??????:4d0e703e.0209261145.6edb6d56@posting.google.com... > > Check the routing information. If it is not set up correctly, > > ping requests won't go through > > > > "IMS" wrote in message > news:... > > > Hi experts, i have a runtime vxWorks image (5.4.2) burnt onto a MIPS > > > platform. On the target shell i find that i can't ping to any host on > the > > > same network nor can i successfully ping to it from another host. > ifShow() > > > gives the following: > > > > > > ene (unit number 1): > > > Flags: (0x8022) DOWN BROADCAST MULTICAST ARP > > > Type: ETHERNET_CSMACD > > > Internet address: 10.128.26.47 > > > Broadcast address: 10.128.26.255 > > > Netmask 0xff000000 Subnetmask 0xffffff00 > > > Ethernet address is 00:e0:4c:12:34:22 > > > Metric is 0 > > > Maximum Transfer Unit size is 1500 > > > 0 octets received > > > 0 octets sent > > > 4110 packets received > > > 0 packets sent > > > 4110 unicast packets received > > > 0 unicast packets sent > > > 0 non-unicast packets received > > > 0 non-unicast packets sent > > > 0 input discards > > > 0 input unknown protocols > > > 0 input errors > > > 0 output errors > > > 0 collisions; 0 dropped > > > lo (unit number 0): > > > Flags: (0x8069) UP LOOPBACK MULTICAST ARP RUNNING > > > Type: SOFTWARE_LOOPBACK > > > Internet address: 127.0.0.1 > > > Netmask 0xff000000 Subnetmask 0xff000000 > > > Metric is 0 > > > Maximum Transfer Unit size is 32768 > > > 0 packets received; 0 packets sent > > > 0 multicast packets received > > > 0 multicast packets sent > > > 0 input errors; 0 output errors > > > 0 collisions; 0 dropped > > > value = 29 = 0x1d > > > -> version > > > > > > It leads me to think that packets can be properly received. Just that > > > nothing can be sent out. Here i have a question. Do i need to suspect > the > > > interrupt setting of the NE2000 end device? Can you suggest any areas > that i > > > should look into? Thanks you all! > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: use of bootline Date: Fri, 27 Sep 2002 15:10:52 +0800 From: "Yung" Organization: IMS Netvigator Message-ID: References: <3D931721.4C2D33BA@notifier-is.net> "david lindauer" wrote in message news:3D931721.4C2D33BA@notifier-is.net... > > > Yung wrote: > > > Hi all, > > > > recently I am using Tornado 2.0 to develop an application of vxWorks. I have > > this query. > > While a bootline is necessary for the bootROM to retrieve a runtime image > > from a network device (or maybe other local devices), why is it also needed > > for the runtime image itself? I have found out that the runtime image > > actually derives network parameters from the bootline (done by routines > > called in prjConfig()). But it doesn't actually boot. Is my finding correct? > > I am very new to this platform development. Any advice very appreciated. > > Thanks. > > yes some of the boot line settings are inherited by the application image. This > allows you to dynamically change your network settings either by distributing > different boot code to different customers (without changing the application) or > by changing the settings by hand in the boot code dialogs. > > In our case the production distributions we use have boot code set to not go on > the network at all; after the app comes up it uses appropriate kernel calls to > set network properties (which we configure through the serial port on our > board). Of course for debug we have boot code that will go out and read the > application image from one of our PCs... at least when we can, some of our > projects use WindWeb and that is not very conducive to the normal debug process > of downloading application modules on the fly. For those we initially bootstrap > the code without using tornado at all, then use FTP to download successive > versions of the kernel/application for testing. > > Did they ever upgrade DOSFS to work with current versions of pre-formatted > disks? AFAIK when we were developing it didn't work because MS had switched to > a new partition enumeration in the MBR. We fixed it by using MSDOS 6.22 format > utility. But it is getting really hard to come by MSDOS 6.22 disks to do the > formatting during production, much easier if we could just live with the > pre-formatted disk. > > David Hi David, thanks for your reply. It confirms my understanding of the use of bootline. Still much to learn on this... --------------------------- Newsgroups: comp.os.vxworks Subject: icmpMaskSet() Date: Fri, 27 Sep 2002 15:18:22 +0800 From: "Yung" Organization: IMS Netvigator Message-ID: can't locate any detailed description about this API call. Does anybody know? Thanks a lot. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: pinging the target Date: Fri, 27 Sep 2002 09:31:42 +0200 From: "Friedrich Ensslin" Message-ID: References: Hi, "Yung" schrieb im Newsbeitrag news:amvgag$5o81@imsp212.netvigator.com... > oops, right. Thanks so much Leonid. As a freshman to the platform i wasn't > even alarmed that the interface is not up and running yet. I was just trying > to make sure every network initialization line in prjConfig.c was executed > without returning error. What have i done or not done to make the interface > go DOWN? I wanted to search for some documentation about the flags of > ifShow() but didn't get any clue yet. How can i bring it back UP? one way to get your ethernet interface started automatically at startup is to add the name of the interface in the "other" section of the bootparameters. bye, Fritz --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How can I merge two Tornados into single one based on different BSP? Date: Fri, 27 Sep 2002 07:59:15 +0000 (UTC) From: Johan Borkhuis Organization: Proxim Corp. Message-ID: References: "Liang Yang" wrote: > Hi, > > I have a Tornado 2.0 for MBX, I also have a Tornado 2.1 for IXP1200. > But I don't want to install two tornado in my computer. How can I > enbale my Tornado 2.0 for MBX to support IXP without installing it? > > I want to try to copy BSP from IXP tornado\target directory to the MBX > tornado\target directory. When I create Projects, I can find IXP > listed in the available BSP? But when I try to build projects, Tornado > reports a lot of error? > > How can I merge this two different Tornado seamlessly? The only way to do combine 2 Tornado trees is when you use the same version. So it is possible to combine a PPC and MIPS version of Tornado 2.1, but not a 2.0 and a 2.1. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Does VxWorks offer True SMP / multiprocessor support ? Date: Fri, 27 Sep 2002 09:53:17 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: ALthough WindRiver itself has never done anything about SMP a few years back, there was a project developed by SPARTA, that had modified the kernel to support SPARC based SMP hardware. ISPARTA was a WindRiver VAR and a solution provider, don't know what's become of them. "Dave Francheski" wrote in message news:amtkii$bc5$1@nnrp.atgi.net... > Can anybody tell me if VxWorks offers true SMP (symmetric multi-processing) > support? > > I know the VxMP option offers shared-memory objects (like semaphores, > sockets, etc.) > between tasks running on separate processors, but this is not at all SMP. > An SMP kernel can schedule tasks/threads to execute on any available > processor in the system. > > Regards, > Dave Francheski > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: pinging the target Date: Fri, 27 Sep 2002 09:56:16 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: It is rather unusual that the flag is DOWN, you might tray ifFlagsSet() to set the UP flag.' "Leonid Rosenboim" wrote in message news:newscache$mfx13h$yn6$1@lnews.actcom.co.il... > Hey Miller, are you really a General ? > > read again your own post, and search for the word "DOWN", there will you > find the wisdom, or at least the answer your're after is there, as to > wisdom, that'll have to wait till your next incarnation. > > Leonid > > "IMS" wrote in message > news:amulmu$ctt1@imsp212.netvigator.com... > > Hi experts, i have a runtime vxWorks image (5.4.2) burnt onto a MIPS > > platform. On the target shell i find that i can't ping to any host on the > > same network nor can i successfully ping to it from another host. ifShow() > > gives the following: > > > > ene (unit number 1): > > Flags: (0x8022) DOWN BROADCAST MULTICAST ARP > > Type: ETHERNET_CSMACD > > Internet address: 10.128.26.47 > > Broadcast address: 10.128.26.255 > > Netmask 0xff000000 Subnetmask 0xffffff00 > > Ethernet address is 00:e0:4c:12:34:22 > > Metric is 0 > > Maximum Transfer Unit size is 1500 > > 0 octets received > > 0 octets sent > > 4110 packets received > > 0 packets sent > > 4110 unicast packets received > > 0 unicast packets sent > > 0 non-unicast packets received > > 0 non-unicast packets sent > > 0 input discards > > 0 input unknown protocols > > 0 input errors > > 0 output errors > > 0 collisions; 0 dropped > > lo (unit number 0): > > Flags: (0x8069) UP LOOPBACK MULTICAST ARP RUNNING > > Type: SOFTWARE_LOOPBACK > > Internet address: 127.0.0.1 > > Netmask 0xff000000 Subnetmask 0xff000000 > > Metric is 0 > > Maximum Transfer Unit size is 32768 > > 0 packets received; 0 packets sent > > 0 multicast packets received > > 0 multicast packets sent > > 0 input errors; 0 output errors > > 0 collisions; 0 dropped > > value = 29 = 0x1d > > -> version > > > > It leads me to think that packets can be properly received. Just that > > nothing can be sent out. Here i have a question. Do i need to suspect the > > interrupt setting of the NE2000 end device? Can you suggest any areas that > i > > should look into? Thanks you all! > > > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: visionICEII / STK8xxL problem Date: 27 Sep 2002 01:48:30 -0700 From: andreas_lindell75@hotmail.com (Rikard) Organization: http://groups.google.com/ Message-ID: <5947a04.0209270048.26f07e5f@posting.google.com> References: "Denis Dubrovin" wrote in message news:... > I use visionICEII with TQ-component STK8xxL starter kit board and have a > problem with BDM connection. > When I connect BDM connector to board and power it up , visionCLICK says > "Can`t enter BKM mode! Check cable and settings!" > > Anyone has same problems? > What I must check for take BDM connection working? Hello Dennis, My first question is an easy on: Have you turned on the power on both ICE and the target? The power for target must be on otherwise you will get that error. If it is check the following: 1. Make sure you can ping your visionICE from command line in host. 2. Make sure you have connected to personality module correctly to the target i.e the pin 1 form ICE must be on pin 1 on target. 3. In the VisionClicks terminal window you should have an >>ERR prompt when unable to enter BKM. Try to type "in" a few times. regards Andreas --------------------------- Newsgroups: comp.os.vxworks Subject: Re: pinging the target Date: Fri, 27 Sep 2002 17:17:44 +0800 From: "Yung" Organization: IMS Netvigator Message-ID: References: i see, thanks. Will have a go with ifFlagSet(). By the way, after another day's effort, i have finally managed to get the ping working (thanks again to the DOWN wisdom that gave the direction). Just that the cure is far too strange. After usrNetIfConfig(), I have added a line icmpMaskGet( ) with some dummy hardcoded parameters. And right after this line ping began to work like magic. ifShow also indicates that the interface is up and running. Peculiar? Still don't know how to rationalize this in the software.... "Leonid Rosenboim" wrote in message news:newscache$gi733h$759$1@lnews.actcom.co.il... > It is rather unusual that the flag is DOWN, you might tray ifFlagsSet() to > set the UP flag.' > > "Leonid Rosenboim" wrote in message > news:newscache$mfx13h$yn6$1@lnews.actcom.co.il... > > Hey Miller, are you really a General ? > > > > read again your own post, and search for the word "DOWN", there will you > > find the wisdom, or at least the answer your're after is there, as to > > wisdom, that'll have to wait till your next incarnation. > > > > Leonid > > > > "IMS" wrote in message > > news:amulmu$ctt1@imsp212.netvigator.com... > > > Hi experts, i have a runtime vxWorks image (5.4.2) burnt onto a MIPS > > > platform. On the target shell i find that i can't ping to any host on > the > > > same network nor can i successfully ping to it from another host. > ifShow() > > > gives the following: > > > > > > ene (unit number 1): > > > Flags: (0x8022) DOWN BROADCAST MULTICAST ARP > > > Type: ETHERNET_CSMACD > > > Internet address: 10.128.26.47 > > > Broadcast address: 10.128.26.255 > > > Netmask 0xff000000 Subnetmask 0xffffff00 > > > Ethernet address is 00:e0:4c:12:34:22 > > > Metric is 0 > > > Maximum Transfer Unit size is 1500 > > > 0 octets received > > > 0 octets sent > > > 4110 packets received > > > 0 packets sent > > > 4110 unicast packets received > > > 0 unicast packets sent > > > 0 non-unicast packets received > > > 0 non-unicast packets sent > > > 0 input discards > > > 0 input unknown protocols > > > 0 input errors > > > 0 output errors > > > 0 collisions; 0 dropped > > > lo (unit number 0): > > > Flags: (0x8069) UP LOOPBACK MULTICAST ARP RUNNING > > > Type: SOFTWARE_LOOPBACK > > > Internet address: 127.0.0.1 > > > Netmask 0xff000000 Subnetmask 0xff000000 > > > Metric is 0 > > > Maximum Transfer Unit size is 32768 > > > 0 packets received; 0 packets sent > > > 0 multicast packets received > > > 0 multicast packets sent > > > 0 input errors; 0 output errors > > > 0 collisions; 0 dropped > > > value = 29 = 0x1d > > > -> version > > > > > > It leads me to think that packets can be properly received. Just that > > > nothing can be sent out. Here i have a question. Do i need to suspect > the > > > interrupt setting of the NE2000 end device? Can you suggest any areas > that > > i > > > should look into? Thanks you all! > > > > > > > > > > > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: pinging the target Date: Fri, 27 Sep 2002 17:22:12 +0800 From: "Yung" Organization: IMS Netvigator Message-ID: References: Arch so. Let me have a check if i can do that too. Danke. "Friedrich Ensslin" wrote in message news:YPTk9.2$la.1804@news.ecrc.de... > Hi, > > "Yung" schrieb im Newsbeitrag > news:amvgag$5o81@imsp212.netvigator.com... > > oops, right. Thanks so much Leonid. As a freshman to the platform i > wasn't > > even alarmed that the interface is not up and running yet. I was just > trying > > to make sure every network initialization line in prjConfig.c was executed > > without returning error. What have i done or not done to make the > interface > > go DOWN? I wanted to search for some documentation about the flags of > > ifShow() but didn't get any clue yet. How can i bring it back UP? > > one way to get your ethernet interface started automatically at startup is > to add the name of the interface in the "other" section of the > bootparameters. > > bye, > > Fritz > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Filename from FILE * or descriptor? Date: 27 Sep 2002 03:26:56 -0700 From: Matthias.Guelck@de.heidelberg.com (Matthias) Organization: http://groups.google.com/ Message-ID: <22419d9b.0209270226.3bbddb45@posting.google.com> References: Hi Bill, sorry for this VERY late answer: Within the description of ioctl() you will find the following: FIOGETNAME Gets the file name of the file descriptor and copies it to the buffer nameBuf: status = ioctl (fd, FIOGETNAME, &nameBuf ); Matthias Bill Pringlemeir wrote in message news:... > Is there a vxWorks function (v5.4.x) that will get a file name from > either a file descriptor or the ANSI file handle. I have something > like this currently... > > #include > ... > const char *getFileName(int fd) > { > const char * lval = "UNKNOWN"; > > if(ERROR != iosValidFd(fd)) > { > lval = fdTable[fd-3].name; > } > > return lval; > } > > This is rather ugly, but it is functional. You must subtract three > for the stdin, stdout and stderr that are intrinsic and not in > fdTable. It would be really nice if there was a function to do this, > but my search of the documentation turned up nothing. However, I may > have developed a blind spot while reading the manual... > > I need this as I am writing a library/function that take a file handle > and not the file name. The file name is only needed as a debugging > aid. > > Thanks, > Bill Pringlemeir. --------------------------- Newsgroups: comp.os.vxworks Subject: ARP could not bind to MUX Date: 27 Sep 2002 04:01:46 -0700 From: bukovsek@iskratel.si (bukovsek) Organization: http://groups.google.com/ Message-ID: When calling pfwStackAdd() it returns error and I receive "ARP could not bind to the MUX! ppppfwStackAddComplete: INTERFACE_LAYER couldn't process add request: stack 0x662f6a0". I was unable to find a description of this message on the web. What would be possible reasons and corresponding solutions? Thanks in advance for your answers. Andrej --------------------------- Newsgroups: comp.os.vxworks Subject: Re: visionICEII / STK8xxL problem Date: Fri, 27 Sep 2002 15:53:16 +0400 From: "Denis Dubrovin" Organization: WEBPlus Ltd. Message-ID: References: <5947a04.0209270048.26f07e5f@posting.google.com> Hi Andreas! Yes, sure I saw >>ERR message at vision click terminal window. INs in next some times not help to me. BTW with MDP823 Reference design board all worked OK. And one more question: Why I can`t enter to >>SIM mode on my ICEII? Is that mode obsoletely in ICEII ? - -- Regards! Den "Rikard" wrote in message news:5947a04.0209270048.26f07e5f@posting.google.com... > "Denis Dubrovin" wrote in message news:... > > I use visionICEII with TQ-component STK8xxL starter kit board and have a > > problem with BDM connection. > > When I connect BDM connector to board and power it up , visionCLICK says > > "Can`t enter BKM mode! Check cable and settings!" > > > > Anyone has same problems? > > What I must check for take BDM connection working? > > Hello Dennis, > > My first question is an easy on: Have you turned on the power on both > ICE and the target? The power for target must be on otherwise you will > get that error. > > If it is check the following: > 1. Make sure you can ping your visionICE from command line in host. > 2. Make sure you have connected to personality module correctly to the > target i.e the pin 1 form ICE must be on pin 1 on target. > 3. In the VisionClicks terminal window you should have an >>ERR prompt > when unable to enter BKM. Try to type "in" a few times. > > regards > Andreas --------------------------- Newsgroups: comp.os.vxworks Subject: Job tip Date: Fri, 27 Sep 2002 14:14:19 GMT From: "Blackie Beard" Message-ID: <%IZk9.12310$va2.918@nwrddc01.gnilink.net> Reply-To: "Blackie Beard" I'm not a recruiter, and I'm just doing this out of charity... I live in Santa Barbara area of California, and have noted two job opportunities for vxWorks. I myself am having problems getting an engineering job, but there are a few niche talents which are still hot (none of which I have), one of which is vxWorks. To make a long story short, Santa Barbara is very beautiful w/ nice beaches, and the jobs are at: Computer Motion (The guy actually called me to ask if I knew vxWorks, I should have lied, I guess) The other is compusensor which has a _big_ ad in the Santa Barbara New Press. Good luck. If you get a job there, remember me for FPGA jobs! Thanks, Mark no.spam.ma.no.spam@no.spam.verizon.net --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ARP could not bind to MUX Date: 27 Sep 2002 07:26:20 -0700 From: gonozal@gmx.at (Toni Kurzberger) Organization: http://groups.google.com/ Message-ID: <8400a733.0209270626.6d399951@posting.google.com> References: bukovsek@iskratel.si (bukovsek) wrote in message news:... > When calling pfwStackAdd() it returns error and I receive "ARP could > not bind to the MUX! ppppfwStackAddComplete: INTERFACE_LAYER couldn't > process add request: stack 0x662f6a0". I was unable to find a > description of this message on the web. What would be possible reasons > and corresponding solutions? Thanks in advance for your answers. > Andrej Hi, check MUX_MAX_BINDS macro (2 bindings per interface, one for IP, one for ARP... plus additional bindings for all other protocols zou might use). In addition, it is alwazs a good idea to check the System Cluster Pool (netStackSysPoolShow...) PFW (WindNet PPP in this case) also uses one cluster pool. In the default configuration this is the Network Data Cluster Poll (netStackDataPoolShow...) Both Pools are configurable with the NUM_* and NUM_SYS_* macros... have fun, toni --------------------------- Newsgroups: comp.os.vxworks Subject: htmlBook utility Date: 27 Sep 2002 07:48:38 -0700 From: valrub@hotmail.com (val) Organization: http://groups.google.com/ Message-ID: <64ddf295.0209270648.314ff151@posting.google.com> Hi there, Tornado IDE provide us with utility htmlBook, which generates the hierarchy of HTML indexes for the HTML manuals within a directory tree. Api has very good explanation about that utility such as parameters and so on, but i found myself in trouble how actually that utility should be start or from where; tornado provide us with htmlBook.tcl and htmlBook.bat file... Maybe somebody has some experience or used that feature or know how to use it? Thank you, Val --------------------------- Newsgroups: comp.os.vxworks Subject: Re: arpresolve: can't allocate llinfo Date: 27 Sep 2002 09:03:04 -0700 From: kannan@comneti.com (kannan) Organization: http://groups.google.com/ Message-ID: <2ede7703.0209270803.1d93efc6@posting.google.com> References: Hi, I had a problem similiar to your problem earlier. The solution for this problem is as follows: In your "target\lib\libXXXXXXgnuvx.a. I have extracted the ".o" files using the "ar" tool and manually load the arpLib.o into my target board through the ld command.Now I was able to do the arpAdd in my target board. where XXXXXX refers to your CPU type. Also the problem llinfo get Solved. Regds, Kanna --------------------------- Newsgroups: comp.os.vxworks Subject: Re: arpresolve: can't allocate llinfo Date: Fri, 27 Sep 2002 16:20:22 +0000 (UTC) From: "Graham Baxter" Organization: Graham Baxter (Software) Limited Message-ID: References: Dale, Is'nt this the message that goes away when you give the add a mask to your target IP address in your boot line? E.g. target address: 192.168.2.1:FFFFFF00 where FFFFFF00 is the mask. There is a colon ':' separating the 2 fields. - -- Regards, Graham Baxter Freelance Software Engineer VxWorks and pSOS BSPs (available shortly for new assignment) gbaxter@NOSPAMbcs.org.uk "Dale Brewe" wrote in message news:amvqie$pn$1@milo.mcs.anl.gov... > I'm trying to get vxWorks 5.4.2 working on a MVME2700 cpu (PowerPC), and have > a problem apparently related to arp. I can get the cpu to download vxworks and > my startup script, but I get the "arpresolve: can't allocate llinfo" msg, any > time I try any further network access (e.g. when trying nfsMount). I don't see > anything wrong w/ the routing table, which has only a couple of entries. The > arp table, however, is empty. Trying arpAdd fails. The cpu doesn't respond to > pings, but other hosts can get its ethernet address. I'd appreciate any > suggestions. > > dale brewe > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How can I merge two Tornados into single one based on different BSP? Date: Fri, 27 Sep 2002 12:44:42 -0500 From: Andrew Johnson Organization: APS, Argonne National Laboratory Message-ID: References: Johan Borkhuis wrote: > > The only way to do combine 2 Tornado trees is when you use the same > version. So it is possible to combine a PPC and MIPS version of Tornado > 2.1, but not a 2.0 and a 2.1. The release notes for Tornado 2.2 say that you cannot install two architectures of that release at the same location, which is rather disapointing. We haven't received 2.2 yet so I can't experiment to find out why, has anyone else found out why they say this? - - Andrew - -- "Life is what happens while you're busy making other plans." - John Lennon --------------------------- Newsgroups: comp.os.vxworks Subject: Re: help with vision ICE Date: Sun, 22 Sep 2002 01:35:11 -0400 From: Varndell Engineering Organization: Posted Via Newsfeeds.com = SPEED+RETENTION+COMPLETION = http://www.newsfeeds.com Message-ID: <3D8D568F.485B552F@hotmail.com> References: <3D8AF9BC.CA270123@hotmail.com> <3D8C554F.E2093351@hotmail.com> Philippe Venne wrote: [snip] > > I took your advice and hooked a logic analyser to the system trigerring on > my suspiscious address. Because I can get the thing to work when the BDM is > not connected, I looked at it going, and beleive me or not, it took to my > slow device 103.6 us to send TA. With my bus running at 40 MHz, if the BM is > on, it will trigger after 51 us when set to maximum timeout. I then ran the > board with loading the image in SDRAM with the BDM tool, and I received a > TEA after, as expected, 51.52 us. That's a really slow device. > Software designers knew about this problem, so they turned of the BM. They clearly had no choice. > So back to my last question, how do I get rid of this "debug mode request > pending" ?!? With visionIce, configure trap exceptions to 'no'. CF TRPEXP NO in the terminal window. With an Abatron, I don't know. Regards, George Varndell - -- Varndell Engineering, LLC Embedded Systems Expertise Need an vxWorks END driver for DP8382x GigE? Contact us. http://www.varndellengineering.com - -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- http://www.newsfeed.com The #1 Newsgroup Service in the World! - -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How can I merge two Tornados into single one based on different BSP? Date: Fri, 27 Sep 2002 18:29:37 +0000 (UTC) From: Johan Borkhuis Organization: Proxim Corp. Message-ID: References: Andrew Johnson wrote: > Johan Borkhuis wrote: >> >> The only way to do combine 2 Tornado trees is when you use the same >> version. So it is possible to combine a PPC and MIPS version of >> Tornado 2.1, but not a 2.0 and a 2.1. > > The release notes for Tornado 2.2 say that you cannot install two > architectures of that release at the same location, which is rather > disapointing. We haven't received 2.2 yet so I can't experiment to > find out why, has anyone else found out why they say this? I am waiting for my second version to test this. I got information from an FAE that it is possible, but not advised, as it is not tested. Also it is possible that they will release patches for specific releases. But ask your FAE for more specific information, as there are some possible problems. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How can I merge two Tornados into single one based on different BSP? Date: 27 Sep 2002 12:50:55 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0209271150.7601890e@posting.google.com> References: Johan Borkhuis wrote in message > The only way to do combine 2 Tornado trees is when you use the same > version. So it is possible to combine a PPC and MIPS version of Tornado > 2.1, but not a 2.0 and a 2.1. Actually, it is also unlikely to work for 2.1 versions since they were all released on a per-architecture essentially as upgrades, and therefore could contain differences in common files (like the header files). The new Tornado 2.2 release should work with multiple architectures in the same tree. Upgrade to that for both architectures and you should be fine. HTH, John... --------------------------- Newsgroups: comp.os.vxworks Subject: Re: arpresolve: can't allocate llinfo Date: Sat, 28 Sep 2002 08:06:18 +0000 (UTC) From: "Graham Baxter" Organization: Graham Baxter (Software) Limited Message-ID: References: Dale, Is'nt this the message that goes away when you give the add a mask to your target IP address in your boot line? E.g. target address: 192.168.2.1:FFFFFF00 where FFFFFF00 is the mask. There is a colon ':' separating the 2 fields. - -- Regards, Graham Baxter Freelance Software Engineer VxWorks and pSOS BSPs (available shortly for new assignment) gbaxter@NOSPAMbcs.org.uk "Dale Brewe" wrote in message news:amvqie$pn$1@milo.mcs.anl.gov... > I'm trying to get vxWorks 5.4.2 working on a MVME2700 cpu (PowerPC), and have > a problem apparently related to arp. I can get the cpu to download vxworks and > my startup script, but I get the "arpresolve: can't allocate llinfo" msg, any > time I try any further network access (e.g. when trying nfsMount). I don't see > anything wrong w/ the routing table, which has only a couple of entries. The > arp table, however, is empty. Trying arpAdd fails. The cpu doesn't respond to > pings, but other hosts can get its ethernet address. I'd appreciate any > suggestions. > > dale brewe > --------------------------- Newsgroups: comp.os.vxworks,comp.protocols.tcp-ip Subject: Socket option SO_LINGER? Date: Sat, 28 Sep 2002 16:25:43 GMT From: yang_christopher@hotmail.com (chris) Message-ID: Hi, I've been investigating the consumption of mBufs when client sessions are active closed from the server. A good example is when a telnet session is established and closed by the server. I've noticed when "inetstatShow" is invoked that the client session doesn't immediately close. It enters the TIME_WAIT state which I know is part of the 2MSL requirement. I've been reading Stevens and I thought that if I set the SO_LINGER option, I would be able to close the TCP connection (the entire TCP connection) by activating SO_LINGER and setting a smaller time (not 0, because VxWorks defaults 0 to 120secs). The two questions I have is: 1. Correct me if I'm wrong, but does SO_LINGER really close the entire TCP connection within the defined time? I've tried setting the linger time to 1 (qLinger = 1) (don't worry I've made sure that the onoff is 1), but when I timed the TCP session after the close, it remained in the mBuf pool for 1 minute. Has anybody experienced such a problem? The socket that I'm setting the options to is the socket used to accept the client connection in Telnet. Maybe I've got the wrong socket to set options to? 2. If SO_LINGER doesn't do what it's supposed to do.... how can I minimize the 2MSL state, without going into the TCP source code and changing them? chris --------------------------- Newsgroups: comp.os.vxworks Subject: Telnet Security (using LoginLib) Date: Sat, 28 Sep 2002 16:32:21 GMT From: yang_christopher@hotmail.com (chris) Message-ID: Hello, Just a quick one.... (bet you've heard this already) I've initialized the libraries using loginLib functions (successfully), can when I run a session with my telnetServer, low and behold, I get the appropriate login sequence, even though, I didn't change any particular code in telnetServer. I'm not too sure how loginLib operates (I think it may have to do with my task shell, but this is irrelevant). I noticed that when I press CTRL-C extremely quickly at the login prompt, I'm able to "bust through" the login and reach the shell. Has anybody experienced this same problem? Is there a known fix for this? I guess, what I could always do is use the loginLib to store my login and passwords, but create my own login prompt from my own shell... but I'm afraid that I'll run into the same predicament. Thanks for your input chris --------------------------- Newsgroups: comp.os.vxworks,comp.protocols.tcp-ip Subject: Re: Socket option SO_LINGER? Date: Sat, 28 Sep 2002 19:49:06 GMT From: steve@nospam.Watt.COM (Steve Watt) Organization: Watt Consultants, San Jose, CA, USA Message-ID: References: Sender: usenet@Watt.COM (Wattres Root) In article , chris wrote: >1. Correct me if I'm wrong, but does SO_LINGER really close the entire TCP >connection within the defined time? I've tried setting the linger time to 1 >(qLinger = 1) (don't worry I've made sure that the onoff is 1), but when I >timed the TCP session after the close, it remained in the mBuf pool for 1 >minute. Has anybody experienced such a problem? The socket that I'm setting >the options to is the socket used to accept the client connection in Telnet. >Maybe I've got the wrong socket to set options to? No, SO_LINGER is used when you close the socket and there is still data to be sent to the other end, but the other end has a zero window. The implementation will wait until SO_LINGER has expired before giving up on sending that data. >2. If SO_LINGER doesn't do what it's supposed to do.... how can I minimize >the 2MSL state, without going into the TCP source code and changing them? Why do you believe this is a problem? Are you servicing tens of thousands of telnet connections from your vxWorks box? The resources will clean themselves up in 2MSL, and that's the way it should be. - -- Steve Watt KD6GGD PP-ASEL-IA ICBM: 121W 56' 57.8" / 37N 20' 14.9" Internet: steve @ Watt.COM Whois: SW32 Free time? There's no such thing. It just comes in varying prices... --------------------------- Newsgroups: comp.os.vxworks Subject: Enable cache in BSP for PPC860 Date: Sun, 29 Sep 2002 01:24:50 GMT From: "somebody" Organization: Magma Communications Ltd. Message-ID: Hi, I need help Enabling cache in romInt.s. Following is my code. I call "EnablePPCCache" from the romStart() function. But when I put this bootloader on my EVB board the board never boots up. Any help would be greatly appreciated. Without calling "EnablePPCCache" the board always boots up. I am not sure what I doing wrong. Currently the EVB board is using MPC 860 I am not sure about the version number. It says XPC860PZP80D4. Thanks sil. .globl EnablePPCCache EnablePPCCache: stw r31,-4(sp) /* save value of r31 to stack (SP not updated) */ mr r31,r3 /* more r3 is return, asave param in reg31 ???*/ li r3,0 /* clear return value ???*/ /* unlock the instruction/data cache */ lis r4, HIADJ ( CACHE_CMD_UNLOCK_ALL)/* load unlock cmd */ addi r4, r4, LO (CACHE_CMD_UNLOCK_ALL) mtspr IC_CST, r4 /* unlock all I cache lines */ mtspr DC_CST, r4 /* unlock all D cache lines */ /* invalidate the instruction/data cache */ lis r4, HIADJ (CACHE_CMD_INVALIDATE) addi r4, r4, LO (CACHE_CMD_INVALIDATE) mtspr IC_CST, r4 /* invalidate all I cache lines */ mtspr DC_CST, r4 /* invalidate all D cache lines */ /* Set forced write - through bit */ lis r4, HIADJ ( 0x01000000 ) addi r4, r4, LO ( 0x01000000 ) mtspr DC_CST, r4 /* enable the instruction/data cache */ lis r4, HIADJ ( CACHE_CMD_ENABLE) /* load enable cmd */ addi r4, r4, LO (CACHE_CMD_ENABLE) mtspr DC_CST, r4 /***** enable D cache ******/ mtspr IC_CST, r4 /***** enable I cache ******/ lwz r31,-4(sp) /* restore caller reg */ blr /* go to caller */ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Enable cache in BSP for PPC860 Date: Sun, 29 Sep 2002 03:27:55 -0400 From: Varndell Engineering Organization: Posted Via Newsfeeds.com = SPEED+RETENTION+COMPLETION = http://www.newsfeeds.com Message-ID: <3D96AB7B.BDCCA470@hotmail.com> References: somebody wrote: > > Hi, > I need help Enabling cache in romInt.s. Following is my code. I call > "EnablePPCCache" from the romStart() function. But when I put this > bootloader on my EVB board the board never boots up. Any help would be > greatly appreciated. Without calling "EnablePPCCache" the board always > boots up. > I am not sure what I doing wrong. Currently the EVB board is using MPC > 860 I am not sure about the version number. It says XPC860PZP80D4. Unfortunately, what you're doing wrong is enabling the dcache in rominit.s. The data cache should not be enabled before vxWorks has initialized and enabled the MMU. Think about the implications of making the entire address cacheable so early in the boot process. Regards, George Varndell - -- Varndell Engineering, LLC Embedded Systems Expertise Need a vxWorks END driver for Sis900? Contact us. http://www.varndellengineering.com - -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- http://www.newsfeed.com The #1 Newsgroup Service in the World! - -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Building projects fron another project Date: 29 Sep 2002 04:59:44 -0700 From: oded.weissman@motorola.com (Oded Weissman) Organization: http://groups.google.com/ Message-ID: Hello Would like to know how can I force Tornado 2.0.2 project to build other project in the workspace (running their makefiles) before building itself (i.e. to create libs needed to final project). Thanks Oded --------------------------- Newsgroups: comp.os.vxworks Subject: 4 Dos partitions with tffs question? Date: 29 Sep 2002 12:00:09 -0700 From: roswel_ajf@hotmail.com (roswell) Organization: http://groups.google.com/ Message-ID: <6dcae13d.0209291100.2daea303@posting.google.com> hi All. My "/tffs0" partition (32MB) flash has been working just fine for a long time. i have dosFs1.0, with tornado 2.0.2. I have total of 128MB of flash. now i want to create 3 more flash/dos partitions (32MB each). And that setup works fine too. But i cannot place files in them, can't copy etc. but, copy returns 0, when i do 'll' there is nothing. dosFsConfigShow "/tffs0", and "/tffs1" are same. everything looks good, but can't copy. i am using the same MTD for all. i can 'cd' to each directory too. and 'pwd' returns correctly. What Am I missing?..pls help. Thanks - -> devs drv name 0 /null 1 /tyCo/0 1 /tyCo/1 3 /tffs0 3 /tffs1 3 /tffs2 3 /tffs3 - -> dosFsConfigShow "/tffs1" device name: /tffs1 total number of sectors: 62780 bytes per sector: 512 media byte: 0xf8 # of sectors per cluster: 4 # of reserved sectors: 1 # of FAT tables: 2 # of sectors per FAT: 62 max # of root dir entries: 240 # of hidden sectors: 4 removable medium: false disk change w/out warning: not enabled auto-sync mode: not enabled long file names: not enabled exportable file system: not enabled lowercase-only filenames: not enabled volume mode: O_RDWR (read/write) available space: 32012288 bytes max avail. contig space: 32012288 bytes value = 0 = 0x0 - -> dosFsConfigShow "/tffs0" device name: /tffs0 total number of sectors: 62780 bytes per sector: 512 media byte: 0xf8 # of sectors per cluster: 4 # of reserved sectors: 1 # of FAT tables: 2 # of sectors per FAT: 62 max # of root dir entries: 240 # of hidden sectors: 4 removable medium: false disk change w/out warning: not enabled auto-sync mode: not enabled long file names: not enabled exportable file system: not enabled lowercase-only filenames: not enabled volume mode: O_RDWR (read/write) available space: 32012288 bytes max avail. contig space: 32012288 bytes value = 0 = 0x0 - -> devs drv name 0 /null 1 /tyCo/0 1 /tyCo/1 3 /tffs0 3 /tffs1 3 /tffs2 3 /tffs3 3 /tmp 5 host: 6 /vio --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Building projects fron another project Date: Sun, 29 Sep 2002 21:43:34 GMT From: "drdiags" Organization: AT&T Broadband Message-ID: References: Oded, "Oded Weissman" wrote in message news:acaaa989.0209290359.25d7b288@posting.google.com... > Hello > > Would like to know how can I force Tornado 2.0.2 project to > build other project in the workspace (running their makefiles) > before building itself > (i.e. to create libs needed to final project). The vxWorks FAQ has a section on how to create a super project. Maybe some of the tips in there can be applied to your case. HTH. http://www.xs4all.nl/~borkhuis/vxworks/project.html#1.7.2 Or maybe the Hierarchical project http://www.xs4all.nl/~borkhuis/vxworks/project.html#1.7.1 > > Thanks > > Oded --------------------------- Newsgroups: comp.os.vxworks Subject: req guide for making changes in pid7t bsp for evaluator7t Date: 29 Sep 2002 22:19:39 -0700 From: jagadtushar@yahoo.com (Tushar) Organization: http://groups.google.com/ Message-ID: <5865894d.0209292119.799935ff@posting.google.com> we are using Samsung KS32C50100... what are the changes required in the bsp of pid7t so as to make vxworks running on evaluator .... we made changes in the file pid7t.h so that the memory locations for all the register matches with the one in evaluator7t but much of the hardware provided with pid7t is not present on evaluator7t ... is it wise to remove those declarations from the file pid7t.h??? evaluator7t comes with bsl loaded on to it ... we would like to preserve the bsl and then try to run vxworks on top it. the moment the bsl starts running from 0x00 the memory is remapped by bsl in the following fashion 0x00000000 to 0x0003FFFF 256KB 32 bit SRAM bank, using ROMCON1 0x00040000 to 0x0007FFFF 256KB 32 bit SRAM bank, using ROMCON2 0x01800000 to 0x0187FFFF 512KB 16 bit flash bank, using ROMCON0 0x03FE0000 to 0x03FE1FFF 8KB 32 bit internal SRAM 0x03FF0000 to 0x03FFFFFF 64KB Microcontroller register space this is the memory usage of bsl in flash 0x01800000 to 0x01806FFF Bootstrap loader 0x01807000 to 0x01807FFF Production test 0x01808000 to 0x0180FFFF Reserved 0x01810000 to 0x0181FFFF Angel 0x01820000 to 0x0187FFFF Available for your programs and data SRAM usage under bsl 0x00000000 to 0x0000003F Exception vector table and address constants 0x00000040 to 0x00000FFF Unused 0x00001000 to 0x00007FFF Read-write data space for BSL 0x00008000 to 0x00077FFF Available as download area for user code and data 0x00078000 to 0x0007FFFF System and user stacks is it possible to run vxworks so that it doesnot corrupt the memory used by bsl? thanx tushar --------------------------- Newsgroups: comp.os.vxworks Subject: Task status values Date: 29 Sep 2002 23:15:02 -0700 From: vanabb@uvic.ca (M. vA) Organization: http://groups.google.com/ Message-ID: <70027fcf.0209292215.74181a56@posting.google.com> Where can I find a table listing the integer values for task status (READY, PEND, etc). These values are returned as integers to td_status of the TASK_DESC structure, but I cannot find a table of what value refers to what state. Instead of parsing taskStatusString each time, it would be simpler to just look at integers. Thank you M. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Task status values Date: Mon, 30 Sep 2002 08:41:44 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <70027fcf.0209292215.74181a56@posting.google.com> see taskLib.h "M. vA" wrote in message news:70027fcf.0209292215.74181a56@posting.google.com... > Where can I find a table listing the integer values for task status > (READY, PEND, etc). These values are returned as integers to td_status > of the TASK_DESC structure, but I cannot find a table of what value > refers to what state. > > Instead of parsing taskStatusString each time, it would be simpler to > just look at integers. > > Thank you > > M. --------------------------- Newsgroups: comp.os.vxworks Subject: source of stackTrace (tt) Date: Mon, 30 Sep 2002 09:43:40 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1033379047.324735@newsreader2.wirehub.nl> Hi, Just curious, has anyone ever implemented some debugging tools, like for example a stacktrace? I'm looking for something like this. For debuggin' reasons I like to write the stacktrace of a task to nvram just before it's scheduled in or out. tt() does not work/help, sinds I want the output in a structure in nvram and not on screen/console... Btw, is there any "public" VxWorks-source archive somewhere? Thanx, Renee - -------------- Renee Teunissen PTS Software bv --------------------------- Newsgroups: comp.os.vxworks Subject: Re: source of stackTrace (tt) Date: Mon, 30 Sep 2002 11:02:22 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1033383768.784746@newsreader2.wirehub.nl> References: <1033379047.324735@newsreader2.wirehub.nl> ehh.. Johan Borkhuis' FAQ has the answer.. Renee. "Renee Teunissen" wrote in message news:1033379047.324735@newsreader2.wirehub.nl... > Hi, > > Just curious, has anyone ever implemented some debugging tools, like > for example a stacktrace? I'm looking for something like this. > > For debuggin' reasons I like to write the stacktrace of a task to > nvram just before it's scheduled in or out. > tt() does not work/help, sinds I want the output in a structure in > nvram and not on screen/console... > > Btw, is there any "public" VxWorks-source archive somewhere? > > Thanx, > Renee > -------------- > Renee Teunissen > PTS Software bv > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re-download boot strap loader to flash Date: 30 Sep 2002 07:39:51 -0700 From: jagadtushar@yahoo.com (Tushar) Organization: http://groups.google.com/ Message-ID: <5865894d.0209300639.6832fcea@posting.google.com> Evaluator7T (Samsung KS32C50100 microcontroller) has a boot strap loader loaded initially which supports a particular command set to interact with the board like download, flashwrite, etc ...... the binary image is also provided with the board .... if the bsl gets corrupted how do i reload the bsl back on to the flash so that i get the basic functionality of the bsl back on to the board .... thanx tushar --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Building projects fron another project Date: Mon, 30 Sep 2002 14:40:09 GMT From: mbgardipee@raytheon.com Organization: Raytheon Company Message-ID: References: Oded, Sub-projects are also covered in the hard-copy Tornado Release Notes 2.0 Edition 1 on pages 19-23. Melvin Gardipee. "Oded Weissman" wrote: > Hello > > Would like to know how can I force Tornado 2.0.2 project to > build other project in the workspace (running their makefiles) > before building itself > (i.e. to create libs needed to final project). --------------------------- Newsgroups: comp.os.vxworks Subject: visionICE software breakpoints in-op Date: Mon, 30 Sep 2002 15:11:57 -0400 From: "Doug Brown" Message-ID: <3d98a1fe$0$15628$4c41069e@reader0.ash.ops.us.uu.net> Anyone had problems getting their visionICE to halt when hitting a SW breakpoint? There must be some config setting that gets wacked somehow. Setting a HW breakpoint works, but sometimes it gets into a state where SW breaks never work. Target platform is PPC7410. Thanks, Doug --------------------------- Newsgroups: comp.os.vxworks Subject: unresolved symbol for semInfo Date: 30 Sep 2002 22:34:39 -0700 From: vxworksonline@yahoo.com (Shiny) Organization: http://groups.google.com/ Message-ID: Hi, I am using semInfo to find out the number of tasks waiting on a semaphore. I am able to compile, but when I download, it gives unresolved error for semInfo. What should i include from VxWorks project facility to be able to use semInfo?? I am using simpc as my BSP. --------------------------- Newsgroups: comp.realtime,comp.os.vxworks Subject: Problem with vxworks Date: 30 Sep 2002 22:47:42 -0700 From: yathish_g@yahoo.com Organization: http://groups.google.com/ Message-ID: <9f000cfd.0209302147.1ff4c00a@posting.google.com> Hi all, I was using the compact Flash II type ethernet card (supporting only 10 Mbps) on my target board. We have been using the NE2000 compatible drivers on vxworks for this card and its working perfectly fine. Recently I had purchased a new CF card from socketcom which they claim to be 10/100 Mbps compatible. I am facing the following problem. Plz help me out in this regard. I have a FTP server running on my PC. The Host PC and the target are connected via cross cable. I am able to ping to the target and ping never fails but Boot is stuck while loading the image from the host. I have attached my FTP log and the console info. console: [VxWorks Boot]: @ boot device : ene unit number : 0 processor number : 0 host name : host file name : vxWorks inet on ethernet (e) : 10.1.12.32 host inet (h) : 10.1.12.30 user (u) : vxworks ftp password (pw) : vxworks flags (f) : 0x0 Attached TCP/IP interface to ene0. Attaching network interface lo0... done. Loading... Ftp log: [L 0017] 08/01/01 14:38:43 Connection accepted from 10.1.12.32 [C 0017] 08/01/01 14:38:43 Command "USER vxworks" received [C 0017] 08/01/01 14:38:43 PASSword accepted [L 0017] 08/01/01 14:38:43 User gayathri logged in. [C 0017] 08/01/01 14:38:43 Command "TYPE I" received [C 0017] 08/01/01 14:38:43 TYPE set to I N [C 0017] 08/01/01 14:38:43 Command "PORT 10.1.12.32,4,1" received [C 0017] 08/01/01 14:38:43 PORT set to 10.1.12.32 - 1025 (4,1) [C 0017] 08/01/01 14:38:43 Command "RETR vxWorks" received [C 0017] 08/01/01 14:38:43 RETRieve started on file vxWorks [C 0017] 08/01/01 14:38:43 Command "PORT 10.1.12.32,4,1" received [C 0017] 08/01/01 14:38:43 PORT set to 10.1.12.32 - 1025 (4,1) [C 0017] 08/01/01 14:38:43 Command "RETR vxWorks" received [C 0017] 08/01/01 14:38:43 RETRieve started on file vxWorks Line "Command "RETR ...." repeats forever and the boot never completes. These are my Queries... 1. When I use the old CF card (10 Mbps), the file is retrieved without any problem. But when I insert the New CF card, which socketcom claims to support 10/100 Mbps...the file retrievel fails. 2. Is there any problem with the MAC settings of the new card....which I suspect.? Hoping for the earlier reply, regds, Yathish. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: visionICE software breakpoints in-op Date: Tue, 01 Oct 2002 02:22:26 -0400 From: Varndell Engineering Organization: Posted Via Newsfeeds.com = SPEED+RETENTION+COMPLETION = http://www.newsfeeds.com Message-ID: <3D993F22.B93F325C@hotmail.com> References: <3d98a1fe$0$15628$4c41069e@reader0.ash.ops.us.uu.net> Doug Brown wrote: > > Anyone had problems getting their visionICE to halt when hitting a SW > breakpoint? > > There must be some config setting that gets wacked somehow. Setting a HW > breakpoint works, but sometimes it gets into a state where SW breaks never > work. > > Target platform is PPC7410. This is almost certainly caused by forgetting to do CF VECTOR LOW. Regards, George Varndell - -- Varndell Engineering, LLC Embedded Systems Expertise Need a vxWorks END driver for DP8382x GigE? Contact us. http://www.varndellengineering.com - -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- http://www.newsfeed.com The #1 Newsgroup Service in the World! - -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Target shell command line limitation Date: Tue, 01 Oct 2002 03:21:08 -0700 From: Bob Bradley Organization: Uh huh huh...It says Organ Message-ID: References: In article , Kevin Tsubota wrote: > I'm running VxWorks 5.3.1 on MV230x CPUs. > > I have a generic application which I instantiate multiple times > using many name=value arguments. I've run into a target shell > command line limit of 128chars. > > Is there an easy way to increase the target shell command line length? The following might be useful from the VxWorks FAQ : - -- Q: How do I increase the maximim linelength of the target shell from 128 characters? A: I'm not sure what you need to do exactly but I've had a similar problem. I put some of the longer paths into a string and then passed the string as the variable on the command line. This allowed us to get by the 128 character limit. Also consider setting environment variables. The first is easier. (From: Don Small, dmsmall@sandia.gov) --------------------------- Newsgroups: comp.os.vxworks Subject: Stack Trace of current stack? Date: Tue, 01 Oct 2002 03:31:48 -0700 From: Bob Bradley Organization: Uh huh huh...It says Organ Message-ID: Is there an easy way to get a stack trace of the current stack? tt() and the code from the VxWorks FAQ work except when called on the current task. I'd like to add an option to my debugging asserts to provide a stack trace (e.g. when a bad parameter is detected, I can more easily see who passed it). Rather than manually parse the stack frame, I figured I could send the current task ID to a message queue and wait on a reply then have a debug task receive the message, do a tt() or trcStack() on that task ID, and reply, but I was hoping for a little cleaner solution. --------------------------- Newsgroups: comp.os.vxworks Subject: TNetTask: arp info overwritten for 81026502 by 00:04:76:21:56:34 Date: 1 Oct 2002 04:09:02 -0700 From: YuriR@elop.co.il (Yuri Rafailsky) Organization: http://groups.google.com/ Message-ID: <6ecd7e0.0210010309.7ba6cc6@posting.google.com> Hi all, This is logMsg from tNetTask: "arp info overwritten for 81026502 by 00:04:76:21:56:34" Do you have any idea what is it and why it happens? Thanks a lot in advanced. Yuri --------------------------- Newsgroups: comp.os.vxworks Subject: Re: TNetTask: arp info overwritten for 81026502 by 00:04:76:21:56:34 Date: Tue, 1 Oct 2002 13:44:34 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1033479903.114467@newsreader2.wirehub.nl> References: <6ecd7e0.0210010309.7ba6cc6@posting.google.com> Are you sure 81026502 is correct? 81026502 = 0x04:d4:5d:c6, does 4.212.93.198 make any sence ? or the part 0x04:d4:5d:c6 as a mac-address? probably you have a duplicate IP on the network.. what you could do is: just after bootup do an arpShow and write down the ipnumber/mac combinations. If you get theis error again, compare the results of arpShow. Another thing, if you own a linux/freebsd/whatever-unixbox, try to install arpwatch, this applications tracks duplicatited ip's on the mac, ip/mac address flip/flops, etc and will send you an email if it detects something. Cheers, Renee - - - - - - - - - Renee A. Teunissen, PTS Software bv, Meerweg 7, 1405BA Bussum, NL. T.+31-(0)35-6926969, M.+31-(0)35-6926969, http://www.pts.nl "Yuri Rafailsky" wrote in message news:6ecd7e0.0210010309.7ba6cc6@posting.google.com... > Hi all, > > This is logMsg from tNetTask: > "arp info overwritten for 81026502 by 00:04:76:21:56:34" > Do you have any idea what is it and why it happens? > Thanks a lot in advanced. > > Yuri --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Stack Trace of current stack? Date: Tue, 1 Oct 2002 13:45:50 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1033479979.93726@newsreader2.wirehub.nl> References: Take a look at this wonderfull FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxw_pt5.html search for "trcStack" Greetz, Renee. - ------------------------ Renee Teunissen PTS Software bv - ------------------------ Q: How can I build a stack-trace? A: You can do this using an undocumented function called "trcStack". Here is all you need. You can even call xxxTrace(tid) from interrupt level... though logMsg() is far from ideal. /********************************************************************* ********** * * xxxTracePrint - stack trace print function * * RETURNS: OK or ERROR */ static void xxxTracePrint ( INSTR *caller, int func, int nargs, int *args ) { char buf [250]; int ix; int len = 0; len += sprintf (&buf [len], "%#10x: %#10x (", (int)caller, func); for (ix = 0; ix < nargs; ix++) { if (ix != 0) len += sprintf (&buf [len], ", "); len += sprintf (&buf [len], "%#x", args [ix]); } len += sprintf (&buf [len], ")\n"); logMsg (buf); } /********************************************************************* ********** * * xxxTrace - stack trace * * RETURNS: OK or ERROR */ int xxxTrace ( WIND_TCB *pTcb ) { REG_SET regs; if (pTcb == NULL) return (ERROR); taskRegsGet ((int)pTcb, ®s); trcStack (®s, (FUNCPTR) xxxTracePrint, (int)pTcb); return (OK); } (From: Geoffrey Espin, espin@idiom.com) - ----- Original Message ----- From: "Bob Bradley" Newsgroups: comp.os.vxworks Sent: Tuesday, October 01, 2002 12:31 PM Subject: Stack Trace of current stack? > Is there an easy way to get a stack trace of the current stack? tt() and > the code from the VxWorks FAQ work except when called on the current > task. > > I'd like to add an option to my debugging asserts to provide a stack > trace (e.g. when a bad parameter is detected, I can more easily see who > passed it). > > Rather than manually parse the stack frame, I figured I could send the > current task ID to a message queue and wait on a reply then have a debug > task receive the message, do a tt() or trcStack() on that task ID, and > reply, but I was hoping for a little cleaner solution. for a little cleaner solution. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: TNetTask: arp info overwritten for 81026502 by 00:04:76:21:56:34 Date: Tue, 01 Oct 2002 07:48:36 -0400 From: Joe Durusau Organization: Lockheed Martin Corporation Message-ID: <3D998B94.89048A6B@lmco.com> References: <6ecd7e0.0210010309.7ba6cc6@posting.google.com> Yuri Rafailsky wrote: > > Hi all, > > This is logMsg from tNetTask: > "arp info overwritten for 81026502 by 00:04:76:21:56:34" > Do you have any idea what is it and why it happens? > Thanks a lot in advanced. > > Yuri It usually means that two devices on your net are using the same IP address, or that perhaps in a network using DHCP that an IP address has been changed. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Problems making boot floppy for T2.2/VxWorks 5.5 Date: 1 Oct 2002 10:01:55 -0700 From: suenews@schooner.com (Susan LoVerso) Organization: http://groups.google.com/ Message-ID: <27127820.0210010901.375ca50b@posting.google.com> I recently received T2.2/VxWorks 5.5 and I am trying to create boot floppies to no avail. I am using a Solaris host and x86 target (p2). Here is what I have attempted. 1. Run T2.2, config/build my VxWorks image, and build my boot ROM images, bootrom_uncmp and bootrom_uncmp.bin. Copy the bootroms to a target-accessible location on my Solaris box. Exit T2.2. 2. Run T2.0 on host/VxWorks 5.4 on my target. Run mkbootFd on 5.4 and create floppy with no errors. (I tried both images, both succeed in mkbootFd but neither appears to boot.) 3. When I attempt to use this floppy, it reads the image, as shown by the V1.6 +++ ... but it apparently has a problem when it jumps to start executing it. (I assume I should see something similar to what I see when I boot VxWorks 5.4, blue screen, countdown to auto-boot, etc. and that it is "quick" - I wait about 30 seconds before deciding it is dead/hung.) I also have a tip line to a serial console that doesn't show anything (in case 5.5 changed to that default from the monitor, like it did for AE). The mkboot.c is identical in my 5.4 and 5.5 tree. I can use the same command and am successful in making boot floppies that work (using the same, brand-new floppy) using a 5.4 bootrom_uncmp. Any ideas on how to debug this would be welcome. Is there some 5.4/5.5 interaction I am not aware of preventing this from working? If so, how do I make a boot floppy if I cannot boot 5.5 and I am on Solaris? Does anyone have a generic 5.5 boot image for a p2 they could 'dd' off the floppy and send me? I would appreciate it. Thanks! --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Task Reschedule on mutex unlock Date: Tue, 1 Oct 2002 17:11:59 +0000 (UTC) From: David Laight Organization: BT Openworld Message-ID: References: <3D908FCF.F75EE09C@alcatel.com> Keith Arner wrote: > On Wed, 25 Sep 2002, David Laight wrote: > > >>>Last time I checked, WindRiver didn't seem interested in fixing this >>>problem in 5.4. >> >>:-) >> >>Actually it is relatively easy to do a much better job. >>The answer to the question "are any tasks waiting for >>any mutex with 'priority invertion' enabled" is easy >>to calculate. If there are no possible waiting tasks >>it must be safe to reduce the priority. > > > If that were done, then priority inheritance for any given task would > depend on the state of all other (read: unrelated) tasks. So introducing > a new task could "break" the priorities of existing tasks. Change, not break... > > To fix the problem, you really need to be able to calculate the set of > tasks blocked by a given task (i.e. the set of tasks blocking on > semaphores owned by the given task), and choose the highest priority of > all the tasks in this set. If you maintain the list in priority order, > choosing the highest priority is constant time, though inserting a node is > O(n). Or you could traverse the list of tasks looking for one waiting for a mutex owned by the current task. I suspect there is a spare list pointer that could be used to maintain a list of tasks waiting on semaphores with priority inversion enabled. David --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Task Reschedule on mutex unlock Date: Tue, 1 Oct 2002 19:43:54 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <3D908FCF.F75EE09C@alcatel.com> David, there already is a list of task waiting on a mutex - it's head is in the mutex semaphore itself ! Also, the inversion-safe mutex queues tasks in priority-prder, so the topmost priority task waiting for mutex is the first one off the list head off the semId. What is currently missing, is a list of mutexes owned by a given task, only a coult is available. One of the advantages of WIND kernel, is that it NEVER goes through th elist of all tasks, except when you ask it to display the list of all tasks (shell, Tornado etc.), so the system call perf does not depend on total number of tasks in the system. Experience shows however that overly large number of tasks gets the human engineer rather confused, and consumes too much memory so this advantage does not really make a big difference most of the time. - - Leonid P.S. there is no spare list, I think. Don't need to memorize source (like me), just look at header files. "David Laight" wrote in message [snip] > Or you could traverse the list of tasks looking for one waiting for > a mutex owned by the current task. > I suspect there is a spare list pointer that could be used to > maintain a list of tasks waiting on semaphores with priority > inversion enabled. > > David > --------------------------- Newsgroups: comp.os.vxworks Subject: Interfacing NTDS board to vxworks Date: 1 Oct 2002 10:53:27 -0700 From: morganed@nswc.navy.mil (Eric Morgan) Organization: http://groups.google.com/ Message-ID: <190f55f8.0210010953.1e1a56f3@posting.google.com> I am trying to interface an ANTARES NTDS board to a Motorola 167 board. I have a compiled test program from the company that I load and run, but the program spits out that there is no board at that address. I have the board address set up correctly, and the software knows that address. What do I have to do in vxworks to get it to allow the two boards to talk to each other? Is there an application note or tech note that explains everything you have to do in vxworks to allow boards to talk to each other? Does the MMU have anything to do with it? Eric Morgan --------------------------- Newsgroups: comp.os.vxworks Subject: Instantiating "extern" variables Date: 01 Oct 2002 14:27:58 -0400 From: "Paul D. Smith" Organization: Nortel Networks, Inc. / Billerica, MA Message-ID: Reply-To: pausmith@nortelnetworks.com Hi all; I'm working on getting the VxWorks 5.3.1 compiler (gcc 2.7.2) from WRS compiled hosted on Linux. I have a prebuilt compiler hosted on Solaris already. I got the sources off a CD directly from WRS, so this should be the same code as was used to build our Solaris-hosted compiler. In fact, all the .o's appear to be identical, so that's good. However, at final link time on Solaris it all works, but on Linux it fails with a whole slew of undefined symbol errors. Investigating, all these symbols are global variables, and they are all referenced everywhere with the extern keyword; that is, every time they're used I see: extern int foo; There is no place in the code where they are actually defined: int foo; Apparently in the Solaris link, even if all references to global variables are extern a global of that name/type is still created. In the Linux link, it requires one of them to be non-extern. I think there's some kind of link-time option, maybe in the ldscripts (except I've checked and they're identical) or someplace else that controls this (obviously the ld command-line options are identical--note I'm running ld directly, not through gcc). Does anyone know where I should look for this? I'd prefer not to have to go through all the code and change it in this project, even though that is the right thing to do in general. Thanks! - -- - ------------------------------------------------------------------------------- Paul D. Smith HASMAT--HA Software Mthds & Tools "Please remain calm...I may be mad, but I am a professional." --Mad Scientist - ------------------------------------------------------------------------------- These are my opinions---Nortel Networks takes no responsibility for them. --------------------------- Newsgroups: comp.os.vxworks Subject: Chain DMA on GT-64260A Date: 1 Oct 2002 12:00:55 -0700 From: akarpov@my-deja.com (Alex) Organization: http://groups.google.com/ Message-ID: <4f09bb68.0210011100.7e295708@posting.google.com> Dear vxworkers, I was wondering if somebody could share his/her experience running a GT-64260A DMA engine in chain mode? I did spend a lot of time trying to get it running properly. So far no luck. An example taken from the Galileo User Manual shipped with the VxWorks BSP DB-64260A (Version 2.2/2.1-I) did not work either. I would appreciate any suggestions. Thanks, Alex --------------------------- Newsgroups: comp.os.vxworks Subject: Build specification - Link order Date: 1 Oct 2002 12:07:33 -0700 From: lawrielk@yahoo.co.in (Lawrie) Organization: http://groups.google.com/ Message-ID: <4ac248a4.0210011107.19484809@posting.google.com> Hi..... If someone could help me in knowing the significance of changing the link order in the build specification it would be very greatful. looking forward for the reply.... thanks. Lawrie. --------------------------- Newsgroups: comp.os.vxworks Subject: Porting of vxWorks. Date: 1 Oct 2002 12:08:16 -0700 From: brian_doe@hotmail.com (brian) Organization: http://groups.google.com/ Message-ID: Hi All, Im presently using vxWorks for a product and would like to port it to another royalty free OS as a cost cutting measure. The chip that Im using has very few other supporting kernels, all of which cost a lot, either in royalties or up fron costs. The application is very simple as far as the OS is concerned.. Theres no TCPIP stack etc. Basicall just the kernel. What I need is a scheduler, message passing, semaphores etc. Basic kernel functionality. Any ideas on how to approach this? One idea I have is to buy/get the source of a generic kernel and imtegrate it into my processor. Is that as large a job as I think it is? Anyone suggest a good source-free kernel out there I can buy (dont want Linux..too big) or download for free? Any other ideas? Thanx in advance brian --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Instantiating "extern" variables Date: 01 Oct 2002 14:54:34 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: References: Sender: bpringle@DeadDuck Paul> Investigating, all these symbols are global variables, and they Paul> are all referenced everywhere with the extern keyword; that is, Paul> every time they're used I see: Paul> extern int foo; Paul> There is no place in the code where they are actually defined: Paul> int foo; Paul> Apparently in the Solaris link, even if all references to Paul> global variables are extern a global of that name/type is still Paul> created. In the Linux link, it requires one of them to be Paul> non-extern. Paul> I think there's some kind of link-time option, maybe in the [snip] This is commonly referred to as a "common" section. The problem is that with `C', there is not type-safe linking. So the "extern int foo" can be an "extern char foo[128]", etc. This is the risk of using this construct. Seeing as you are a very busy person, here is a relevant portion of the LD document. -d, -dc, -dp These three options are equivalent; multiple forms are supported for compatibility with other linkers. They assign space to common symbols even if a relocatable output file is specified (with `-r'). The script command FORCE_COMMON_ALLOCATION has the same effect. See section Option Commands. Of course, this might be different with the version of the GNU tools that you use on your Linux system (or Solaris if you are Canadian crossing). However, the main point is that this is called a "COMMON" section. I think that you were missing that idea. hth, Bill Pringlemeir. - -- While you're chewing, think of STEVEN SPIELBERG'S bank account.. This will have the same effect as two ``STARCH BLOCKERS''! --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: 01 Oct 2002 15:01:50 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: References: Sender: bpringle@DeadDuck Brian> Any ideas on how to approach this? One idea I have is to Brian> buy/get the source of a generic kernel and imtegrate it into Brian> my processor. Is that as large a job as I think it is? Anyone Brian> suggest a good source-free kernel out there I can buy (dont Brian> want Linux..too big) or download for free? You would get far better advice if you gave the processor. There are two RTOS's that I would use. eCos and uCos, they are both royalty free. (uCos-II has a one-time license fee). "http://sources.redhat.com/ecos/" ARM, IA32, Matsushita AM3x, MIPS, NEC V8xx, PowerPC, SPARC, SuperH "http://www.ucos-ii.com/" "http://www.ucos-ii.com/contents/products/ucos-ii/ports.html" Avr, TriCore, Z80, V8, TMSxxx, 8051, etc. Is the processor 4,8,16,32,64 bit? Even that would help a lot. hth, Bill Pringlemeir. - -- global enforcers Saddam Hussein investigation CESID e-cash keyhole argus Kh-11 interception quiche Cohiba Ceridian Treasury bootleg --------------------------- Newsgroups: comp.os.vxworks Subject: Tornado II on XP Date: Tue, 01 Oct 2002 21:47:38 +0000 From: molochai Organization: Griffin Internet - http://www.griffin.net.uk Message-ID: Anyone know whether Tornado II works correctly on XP? I know there was a patch for Win2k - is this sufficient? M - -- The muppetry is strong in this one. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Stack Trace of current stack? Date: Tue, 01 Oct 2002 13:52:48 -0700 From: Bob Bradley Organization: Uh huh huh...It says Organ Message-ID: References: <1033479979.93726@newsreader2.wirehub.nl> In article <1033479979.93726@newsreader2.wirehub.nl>, "Renee Teunissen" wrote: > Take a look at this wonderfull FAQ: > http://www.xs4all.nl/~borkhuis/vxworks/vxw_pt5.html > search for "trcStack" I am trying to get a stack trace of the current stack, which is not supported by the code from the FAQ (or tt()). trcStack() in general doesn't appear to support getting a stack trace of the current stack. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Instantiating "extern" variables Date: 01 Oct 2002 17:37:25 -0400 From: "Paul D. Smith" Organization: Nortel Networks, Inc. / Billerica, MA Message-ID: References: Reply-To: pausmith@nortelnetworks.com %% Bill Pringlemeir writes: bp> This is commonly referred to as a "common" section. The problem is bp> that with `C', there is not type-safe linking. So the "extern int bp> foo" can be an "extern char foo[128]", etc. This is the risk of using bp> this construct. Indeed. However, I didn't write the code and there's too much of it to allow fixing it to fall into the scope of this project. Next one, though... It's difficult because our default build environment, obviously, doesn't complain. So no one notices these until we try to build in the non-default environment--even if I fix these the problem will probably reappear shortly. If I ever figure out why the Linux ldppc is not instantiating the common section, maybe that will show me how to force the Solaris ldppc to behave the same way so we can avoid this in the future :). bp> Seeing as you are a very busy person, here is a relevant portion of bp> the LD document. Thanks, but as I pointed out in the original post the ld invocations between the Solaris and Linux versions are identical (no extra/missing flags on either one, and I'm invoking ld directly, not through gcc), so it's not -d. This did lead me to the ld --verbose option to display the linker scripts though--thanks; I was using -v and that contained very little info :)--however, adding --verbose to the two ld commands shows that the linker scripts are identical as well :(. Perhaps there's something more bizarre going on here than it first appeared. - -- - ------------------------------------------------------------------------------- Paul D. Smith HASMAT--HA Software Mthds & Tools "Please remain calm...I may be mad, but I am a professional." --Mad Scientist - ------------------------------------------------------------------------------- These are my opinions---Nortel Networks takes no responsibility for them. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Instantiating "extern" variables Date: 01 Oct 2002 18:09:04 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: References: Sender: bpringle@DeadDuck Bill> Seeing as you are a very busy person, here is a relevant Bill> portion of the LD document. Paul> Thanks, but as I pointed out in the original post the ld Paul> invocations between the Solaris and Linux versions are Paul> identical (no extra/missing flags on either one, and I'm Paul> invoking ld directly, not through gcc), so it's not -d. Hmm, apparently I am not clairvoyant. As a matter of fact, I don't even know if the link problem was when you were building GCC or when you were building a target project. I am now guessing that you have built a Linux cross compiler and the the problem is that the externals are not found when building code for the target. Ie, the x86-linux cross compiler with a PowerPC-vxworks target is not linking as with the solaris gcc from WRS with a PowerPC-vxWorks target? Paul> Perhaps there's something more bizarre going on here than it Paul> first appeared. It is usually bad to assume this to begin with. Can you link the Linux objects on the Solaris box without problems? I guess so, as you said they are identical. You could use the default linker file printed by the --verbose to add "*(COMMON)" to the bss section. It sounds like the linux hosted "ldppc" has built differently. Another possibility is that some host environment variable is affecting things. I have seen this several times before. The Linux installation may use some environment variable to change the default behavior of the linker and this maybe propagating to the cross compiler. Using a custom linker file will help to figure out whether the environment is responsible. hth, Bill Pringlemeir. - -- Do not meddle in the affairs of dragons, for you are crunchy and taste good with ketchup. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: TNetTask: arp info overwritten for 81026502 by 00:04:76:21:56:34 Date: 1 Oct 2002 15:43:56 -0700 From: akarpov@my-deja.com (Alex) Organization: http://groups.google.com/ Message-ID: <4f09bb68.0210011443.670f7040@posting.google.com> References: <6ecd7e0.0210010309.7ba6cc6@posting.google.com> Yuri, Make sure you use correct IP address mask in the boot parameters, for instance: ... inet on ethernet (e) : 190.18.64.104:ffffff00 ... YuriR@elop.co.il (Yuri Rafailsky) wrote in message news:<6ecd7e0.0210010309.7ba6cc6@posting.google.com>... > Hi all, > > This is logMsg from tNetTask: > "arp info overwritten for 81026502 by 00:04:76:21:56:34" > Do you have any idea what is it and why it happens? > Thanks a lot in advanced. > > Yuri vxworks --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Location of exception vectors in PPC BSPs Date: 1 Oct 2002 17:10:13 -0700 From: kiran.mandava@skystream.com (kiran Mandava) Organization: http://groups.google.com/ Message-ID: References: <29b153d9.0209190616.1a897465@posting.google.com> hello All, I am trying to port vxworks onto one of our board. this board has dual processors 750Cxe, CPC710 is the bridging device. they are using the same SDRAM i have problems while registering the interrupt handlers. First CPU is using the first half of the SDRAM and the second CPU the other half. Since i can't relocate the Interrupt Vector base how can i handle the interrupts on both the CPUs. if any of the processor gets the external interrupt it will jump to 0x500 and vxworks tries to save the interrupt context to the interrupt stack location and then call the ISR. In the ISR i could jump to a different location depending on the CPU on which the interrupt occured, but the interrupt context is always stored on the first CPU SDRAM and how to handle the case if the first CPU gets the interrupt at the same time as when the second one is in the middle of the ISR. thanks kiran rick@cyclone.com (Richard Nardone) wrote in message news:<29b153d9.0209190616.1a897465@posting.google.com>... > fgleeson@gi.com (FrankG) wrote in message news:... > > How does vxWorks locate the PPC exception vectors? > > Specifically I am trying to understand the wrSbc8260 BSP, but the > > question applies to all PPC BSPs. > > > > I think I understand the PPC exception vector scheme. The base is set > > according to 0xFFF00000 or 0x00000000. This depends on the IP bit in > > the MSR which originates at reset in the CIP bit in the HRCW. Whatever > > the base is you then > > have the following offsets: Quoting from 6-4 of MPCFPE32B.pdf. > > System Reset 0x0100 > > Machine Check 0x0200 > > DSI 0x0300 > > ISI 0x0400 > > External Interrupt 0x0500 > > Alignment 0x0600 > > Program 0x0700 > > FP unavailable 0x0900 > > ... > > Reserved 0x1000 - 0x2FFF > > This is true, except the Decrementer exception is connected at > offset 0x0900. > > > > > > There are several things I am confused about. It seems as though the > > romImage starts at address 0xFFF00100. I guess this causes the > > assembler routine romInit in romInit.s to execute on reset by virtue > > of the fact that it is linked first. > > After the HRCW is read, the first instruction fetch is from the > system reset exception vector located at offset 0x0100 from the > exception base. > > > It seems like there is no allowance for any other exception except the > > External Interrupt which is then connected from m8260IntrInit() via > > this API call. > > > > excIntConnect ((VOIDFUNCPTR *) _EXC_OFF_INTR, m8260IntrDeMux); > > > > I geuss this just overwrites whatever happens to be at address > > 0xFFF00500. > > The PPC architecture has one main external interrupt. The ISR > placed at this location has lots of work to do to figure out which > interrupt(s) is present and then dispatch the appropriate processing > (another ISR). > The Decrementer exception is also used as the system cadence for many > BSPS. > This determines when the kernel is activated, the VxWork Tick, which > is usually set to 60Hz or 100Hz. > > Which is a predefined physical address location in the MMU? I geuss > > the writing of the TLBs in romInit.s set this up to whatever physical > > memory location. > The BSP I worked with on the 8260 did a one to one mapping using the > four IBATS and four DBATS. Look at the PowerPC Programming > Environments for 32-Bit Microprocessors manual, chapter 7 MMU. > > > > > Clearly I have more questions than answers I am slowly integrating all > > this PPC/BSP knowledge. However I was hoping someone who had already > > figured this stuff out could give me a hint, or at least point me to > > some good documentation. --------------------------- Newsgroups: comp.os.vxworks Subject: bringingup vxworks on a board with two 750cxe processors Date: 1 Oct 2002 17:20:02 -0700 From: kiran.mandava@skystream.com (kiran Mandava) Organization: http://groups.google.com/ Message-ID: Hello all, Since the interrupt vector base is not relocatable on powrePc 750 how should i handle the interrupts on both the processors. if there is an external interrupt on any of the processors it would jump to location 0x500 where the ISR corresponding to the First CPU is located, so there is a problem handling the interrupts from both the processors as both of them try to save the interrupt conetext to the same location..did anybody came across such problem and any sort of feedback would be really helpful CPC 710 is the bridging device and both the processors use the same SDRAM first half the first CPU and the otherhalf the other CPU thanks, kiran --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Tornado II on XP Date: Tue, 01 Oct 2002 17:33:37 -0700 From: Bob Bradley Organization: Uh huh huh...It says Organ Message-ID: References: In article , molochai wrote: > Anyone know whether Tornado II works correctly on XP? I know there was a > patch for Win2k - is this sufficient? I use Tornado 2.0.2 on Windows XP and it seems to work correctly. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Instantiating "extern" variables Date: 01 Oct 2002 22:34:55 -0400 From: "Paul D. Smith" Organization: Nortel Networks, Inc. / Billerica, MA Message-ID: References: Reply-To: pausmith@nortelnetworks.com %% Bill Pringlemeir writes: Bill> Seeing as you are a very busy person, here is a relevant Bill> portion of the LD document. Paul> Thanks, but as I pointed out in the original post the ld Paul> invocations between the Solaris and Linux versions are Paul> identical (no extra/missing flags on either one, and I'm Paul> invoking ld directly, not through gcc), so it's not -d. bp> Hmm, apparently I am not clairvoyant. From my original post: > I think there's some kind of link-time option, maybe in the ldscripts > (except I've checked and they're identical) or someplace else that > controls this (obviously the ld command-line options are identical--note > I'm running ld directly, not through gcc). What I meant this to say was that I run ld on Linux and Solaris with identical command lines and that the linker scripts are identical. bp> As a matter of fact, I don't even know if the link problem was bp> when you were building GCC or when you were building a target bp> project. Yes, sorry, this wasn't clear in my post. bp> I am now guessing that you have built a Linux cross compiler and bp> the the problem is that the externals are not found when building bp> code for the target. Ie, the x86-linux cross compiler with a bp> PowerPC-vxworks target is not linking as with the solaris gcc from bp> WRS with a PowerPC-vxWorks target? Yes. Note that the x86-linux cross compiler was built from the GCC source CD provided by WRS and ostensibly is the same code that the Solaris compiler was built from. However, either that's not quite true or there were changes made to the Solaris-specific areas of GCC by WRS which still need to be made in the Linux-specific areas, because I've noted a few situations where the output .o files are not identical (obviously leaving aside any timestamping, etc.) given the same input on Solaris and Linux. For one example, building a C++ file with -g gives very different output (I mean like it's 12k bigger on Solaris), although without -g the output is identical. Paul> Perhaps there's something more bizarre going on here than it Paul> first appeared. bp> It is usually bad to assume this to begin with. Can you link the bp> Linux objects on the Solaris box without problems? I guess so, as bp> you said they are identical. You could use the default linker bp> file printed by the --verbose to add "*(COMMON)" to the bss bp> section. It sounds like the linux hosted "ldppc" has built bp> differently. Another possibility is that some host environment bp> variable is affecting things. I have seen this several times bp> before. The Linux installation may use some environment variable bp> to change the default behavior of the linker and this maybe bp> propagating to the cross compiler. Since they are the same code that shouldn't be an issue. Anyway, I too have seen far too many cases where individual users' environments caused differenced in the build, and I have long since introduced a shell script wrapper to invoke the build, the primary purpose of which is to sanitize the environment. After further investigation I have found the problem. It's _very_ bizarre. Someone wrote code like this: int some_size = SOME_SIZE_MAX; char some_array[some_size]; Obviously this is not valid C89 (although variable sizes arrays are supported in the C99 standard), but there is a GNU extension that allows it. This code actually compiles and works (I suppose) with the Solaris WRS compiler, but it fails to compile with the Linux version--in fact, the compiler segfaults! Unfortunately the build did not fail, as it should have, due to poorly written make rules. I don't fully grok the connection, although those problem variables _are_ used in that .o file. Maybe the compiler left a corrupted .o around after the segfault. Changing that code to the correct: int some_size = SOME_SIZE_MAX; char some_array[SOME_SIZE_MAX]; allows the code to compile and the link now proceeds without incident. - -- - ------------------------------------------------------------------------------- Paul D. Smith HASMAT--HA Software Mthds & Tools "Please remain calm...I may be mad, but I am a professional." --Mad Scientist - ------------------------------------------------------------------------------- These are my opinions---Nortel Networks takes no responsibility for them. --------------------------- Newsgroups: comp.realtime,comp.os.vxworks Subject: Re: Problem with vxworks Date: 1 Oct 2002 20:46:28 -0700 From: wv9557@yahoo.com (Will) Organization: http://groups.google.com/ Message-ID: <4a885870.0210011946.4271985d@posting.google.com> References: <9f000cfd.0209302147.1ff4c00a@posting.google.com> The log only shows one side of the story. Not seeing the whole packets exchange, it's hard to tell. You might want to get something like tcpdump for Windows (yah - tcpdump does run on Windows :) and look at every packet that is exchanged between the 2 boxes. Just a shoot in the dark: after you swapped the cards, wait long enough for the PC's ARP cache to timeout. Just a shoot in the dark number 2: Force both the vxWorks box and the PC to do 10mbits, by using a good old 10mbit-only switch. yathish_g@yahoo.com wrote in message news:<9f000cfd.0209302147.1ff4c00a@posting.google.com>... > Hi all, > > I was using the compact Flash II type ethernet card (supporting only > 10 Mbps) on my target board. We have been using the NE2000 compatible > drivers on vxworks for this card and its working perfectly fine. > Recently I had purchased a new CF card from socketcom which they claim > to be 10/100 Mbps compatible. I am facing the following problem. Plz > help me out in this regard. > > I have a FTP server running on my PC. The Host PC and the target are > connected via cross cable. I am able to ping to the target and ping > never fails but Boot is stuck while loading the image from the host. I > have attached my FTP log and the console info. > > console: > [VxWorks Boot]: @ > > boot device : ene > unit number : 0 > processor number : 0 > host name : host > file name : vxWorks > inet on ethernet (e) : 10.1.12.32 > host inet (h) : 10.1.12.30 > user (u) : vxworks > ftp password (pw) : vxworks > flags (f) : 0x0 > > Attached TCP/IP interface to ene0. > Attaching network interface lo0... done. > Loading... > > Ftp log: > [L 0017] 08/01/01 14:38:43 Connection accepted from 10.1.12.32 > [C 0017] 08/01/01 14:38:43 Command "USER vxworks" received > [C 0017] 08/01/01 14:38:43 PASSword accepted > [L 0017] 08/01/01 14:38:43 User gayathri logged in. > [C 0017] 08/01/01 14:38:43 Command "TYPE I" received > [C 0017] 08/01/01 14:38:43 TYPE set to I N > [C 0017] 08/01/01 14:38:43 Command "PORT 10.1.12.32,4,1" received > [C 0017] 08/01/01 14:38:43 PORT set to 10.1.12.32 - 1025 (4,1) > [C 0017] 08/01/01 14:38:43 Command "RETR vxWorks" received > [C 0017] 08/01/01 14:38:43 RETRieve started on file vxWorks > [C 0017] 08/01/01 14:38:43 Command "PORT 10.1.12.32,4,1" received > [C 0017] 08/01/01 14:38:43 PORT set to 10.1.12.32 - 1025 (4,1) > [C 0017] 08/01/01 14:38:43 Command "RETR vxWorks" received > [C 0017] 08/01/01 14:38:43 RETRieve started on file vxWorks > > Line "Command "RETR ...." repeats forever and the boot never > completes. > > These are my Queries... > > 1. When I use the old CF card (10 Mbps), the file is retrieved > without any problem. But when I insert the New CF card, which > socketcom > claims to support 10/100 Mbps...the file retrievel fails. > > 2. Is there any problem with the MAC settings of the new > card....which I suspect.? > > Hoping for the earlier reply, > > regds, > Yathish. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Instantiating "extern" variables Date: 1 Oct 2002 21:38:09 -0700 From: wv9557@yahoo.com (Will) Organization: http://groups.google.com/ Message-ID: <4a885870.0210012038.6fc01a79@posting.google.com> References: Are you sure there is not an "extern int foo = 2;" somewhere? Note the weird combo of extern and =. Wild guess: maybe someone hides it in assembly file? "Paul D. Smith" wrote in message news:... > Hi all; > > I'm working on getting the VxWorks 5.3.1 compiler (gcc 2.7.2) from WRS > compiled hosted on Linux. I have a prebuilt compiler hosted on Solaris > already. > > I got the sources off a CD directly from WRS, so this should be the same > code as was used to build our Solaris-hosted compiler. > > In fact, all the .o's appear to be identical, so that's good. However, > at final link time on Solaris it all works, but on Linux it fails with a > whole slew of undefined symbol errors. > > Investigating, all these symbols are global variables, and they are all > referenced everywhere with the extern keyword; that is, every time > they're used I see: > > extern int foo; > > There is no place in the code where they are actually defined: > > int foo; > > Apparently in the Solaris link, even if all references to global > variables are extern a global of that name/type is still created. In > the Linux link, it requires one of them to be non-extern. > > I think there's some kind of link-time option, maybe in the ldscripts > (except I've checked and they're identical) or someplace else that > controls this (obviously the ld command-line options are identical--note > I'm running ld directly, not through gcc). > > Does anyone know where I should look for this? I'd prefer not to have > to go through all the code and change it in this project, even though > that is the right thing to do in general. > > Thanks! --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Instantiating "extern" variables Date: 1 Oct 2002 21:51:46 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0210012051.19403d6c@posting.google.com> References: Hello, You have discovered something about the file formats that is not so obvious... From one of the later posts I assume you are using PowerPC. In that case, you should be generating ELF output, and when generating ELF 'extern int foo;' means generate an undefined symbol. If you want a common symbol you must write 'int foo;' (i.e. commons are bss symbols). The VxWorks code itself relies on these for many of its scalability features. In older file formats (e.g. a.out) though, the extern keyword would have generated a common (as Bill suggested). Unfortunately, this is doesn't seem to be true when generating ELF (nor COFF by the way, so Bill, you should have seen this on your ARM builds as well). The puzzle for me is why it works using your Solaris toolchain; it certainly should not consider an extern to be common. If you perform a partial link on Solaris (using the -r option), what is the type of one of the "undefined" symbols in the nm output? Are you certain that you are not missing an object file anywhere in the Linux link that would define all the missing symbols? Have you compared the nm output of all the files? If not, add them all to an archive, and generate the nm output for that on both hosts - then diff it. Or, as Bill suggested I believe, try linking the Linux generated objects on the Solaris box. Unfortunately, I am at home at the moment so I cannot try the Solaris tools. I can tell you that the Linux-hosted VxWorks AE toolchain requires a non-extern instance somewhere to link correctly though, and I believe that the Solaris hosted AE toolchain does too. Of course, that is 2.96 not 2.7.2. Rgds, John... "Paul D. Smith" wrote in message news:... > Hi all; > > I'm working on getting the VxWorks 5.3.1 compiler (gcc 2.7.2) from WRS > compiled hosted on Linux. I have a prebuilt compiler hosted on Solaris > already. > > I got the sources off a CD directly from WRS, so this should be the same > code as was used to build our Solaris-hosted compiler. > > In fact, all the .o's appear to be identical, so that's good. However, > at final link time on Solaris it all works, but on Linux it fails with a > whole slew of undefined symbol errors. > > Investigating, all these symbols are global variables, and they are all > referenced everywhere with the extern keyword; that is, every time > they're used I see: > > extern int foo; > > There is no place in the code where they are actually defined: > > int foo; > > Apparently in the Solaris link, even if all references to global > variables are extern a global of that name/type is still created. In > the Linux link, it requires one of them to be non-extern. > > I think there's some kind of link-time option, maybe in the ldscripts > (except I've checked and they're identical) or someplace else that > controls this (obviously the ld command-line options are identical--note > I'm running ld directly, not through gcc). > > Does anyone know where I should look for this? I'd prefer not to have > to go through all the code and change it in this project, even though > that is the right thing to do in general. > > Thanks! --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Stack Trace of current stack? Date: 1 Oct 2002 22:05:00 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0210012105.4e27864c@posting.google.com> References: <1033479979.93726@newsreader2.wirehub.nl> Hello, This may, or may not work... worth a try though: void traceMe (void) { REG_SET regs; taskDelay (1); /* Force context save */ taskRegsGet (0, ®s); trcStack (®s, NULL, /* NULL for default print routine */ 0); /* 0 for myself */ } The context save should get an approximately correct set of registers into the TCB. The only thing I can see causing a problem here is the heuristic scanner code that tries to determine where in the life of a function the PC is (i.e. is the PC in the prologue, the main function code or the epilogue). Try it and post the results, either way, so that others can benefit from your findings in the future. Make sure you tell us the architecture you tried too... HTH, John... Bob Bradley wrote in message news:... > In article <1033479979.93726@newsreader2.wirehub.nl>, > "Renee Teunissen" > agree.with.these.terms@mail.zevenaar.org> wrote: > > > Take a look at this wonderfull FAQ: > > http://www.xs4all.nl/~borkhuis/vxworks/vxw_pt5.html > > search for "trcStack" > > I am trying to get a stack trace of the current stack, which is not > supported by the code from the FAQ (or tt()). trcStack() in general > doesn't appear to support getting a stack trace of the current stack. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Problems making boot floppy for T2.2/VxWorks 5.5 Date: 1 Oct 2002 22:29:34 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0210012129.1d3e9ad3@posting.google.com> References: <27127820.0210010901.375ca50b@posting.google.com> Hello Susan, Do you still have the VxWorks AE boot floppies? That should be able to boot a 5.5 kernel image (just point the filename at your vxWorks image file). From there, you can use the mkboot tool for VxWorks to create yourself a 5.5 boot floppy. If you don't have the VxWorks AE images, they can be downloaded from WindSurf still. I ask for the 5.5 images to be put up there too... HTH, John... suenews@schooner.com (Susan LoVerso) wrote in message news:<27127820.0210010901.375ca50b@posting.google.com>... > I recently received T2.2/VxWorks 5.5 and I am trying to create boot floppies > to no avail. I am using a Solaris host and x86 target (p2). Here is what > I have attempted. > > 1. Run T2.2, config/build my VxWorks image, and build my boot ROM images, > bootrom_uncmp and bootrom_uncmp.bin. Copy the bootroms to a target-accessible > location on my Solaris box. Exit T2.2. > 2. Run T2.0 on host/VxWorks 5.4 on my target. Run mkbootFd on 5.4 and > create floppy with no errors. (I tried both images, both succeed in mkbootFd > but neither appears to boot.) > 3. When I attempt to use this floppy, it reads the image, as shown by the > V1.6 +++ ... but it apparently has a problem when it jumps to start executing it. > (I assume I should see something similar to what I see when I boot > VxWorks 5.4, blue screen, countdown to auto-boot, etc. and that it > is "quick" - I wait about 30 seconds before deciding it is dead/hung.) > > I also have a tip line to a serial console that doesn't show anything (in > case 5.5 changed to that default from the monitor, like it did for AE). > > The mkboot.c is identical in my 5.4 and 5.5 tree. I can use the same command > and am successful in making boot floppies that work (using the same, > brand-new floppy) using a 5.4 bootrom_uncmp. > > Any ideas on how to debug this would be welcome. Is there some 5.4/5.5 > interaction I am not aware of preventing this from working? If so, how do > I make a boot floppy if I cannot boot 5.5 and I am on Solaris? Does anyone > have a generic 5.5 boot image for a p2 they could 'dd' off the floppy > and send me? I would appreciate it. Thanks! --------------------------- Newsgroups: comp.os.vxworks Subject: newbie - device drivers in c++ Date: 1 Oct 2002 23:03:46 -0700 From: ravurirr@hotmail.com (Rao R R) Organization: http://groups.google.com/ Message-ID: <290f4d3c.0210012203.38c6685a@posting.google.com> Hi Does vxworks supports c++ object handling at kernel level? Do some one tried writing device drivers in vxworks in c++. Can someone give insight into this ? What would be the adv and disadvanteges of writing device drivers in c++. thanks in advance Ranga Rao R --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Wed, 2 Oct 2002 08:17:53 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1033539506.392359@newsreader1.wirehub.nl> References: You could try the special 8086/microcontroller version of the linux kernel (http://www.uclinux.org/) or somethng like that. It all depend what kind of process your porting for. other interesring links: http://www.lineo.com/products/embedix/index.html http://www.rtlinux.org/ or http://embedded.pagina.nl Cheers, Renee - - - - - - - - - ing. R. A. Teunissen (Renee) PTS Software bv, Meerweg 7, 1405BA Bussum, NL. T.+31-(0)35-6926969, M.+31-(0)6-22778313, http://www.pts.nl, @pts.nl personal link page: http://wittenburg10c.nl/db/dest/links.html "brian" wrote in message news:d2f67bd8.0210011108.5e6d9858@posting.google.com... > Hi All, > > Im presently using vxWorks for a product and would like to port it to > another royalty free OS as a cost cutting measure. The chip that Im > using has very few other supporting kernels, all of which cost a lot, > either in royalties or up fron costs. > > The application is very simple as far as the OS is concerned.. Theres > no TCPIP stack etc. Basicall just the kernel. What I need is a > scheduler, message passing, semaphores etc. Basic kernel > functionality. > > Any ideas on how to approach this? One idea I have is to buy/get the > source of a generic kernel and imtegrate it into my processor. Is that > as large a job as I think it is? Anyone suggest a good source-free > kernel out there I can buy (dont want Linux..too big) or download for > free? > > Any other ideas? > > Thanx in advance > > brian --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Stack Trace of current stack? Date: Wed, 2 Oct 2002 09:15:00 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1033542934.364066@newsreader1.wirehub.nl> References: <1033479979.93726@newsreader2.wirehub.nl> It does work, btw - I could trace the stack of the shell I was working on :) But stacktracing of the currect task is an illusion, think about it. The stack should - if correct - always end in trcStack or one of it's sibblings or... not ;) Please rethink your problem.. Otherwise, you could just grep the registers and task info, and trace the framepointer until you are at the end, and start with the current framepointer (see your regs). This should not be so difficult. The framepoint can easily be found on a fixed offset of the stackpointer. Every local stack frame has a pointer to the previous stack frame and so on. This is - so far I can see from here - the samething trcStack does. Therefore it needs the regs. So if oyu manually fill-in the framepointer of a regs-struct you could nice things with it ;) All function arguments are on a standard offset of the framepointer/stackpointer. Btw, you didnt compile the thing with the omit-framepointer (or something like that) option - did you? "Bob Bradley" wrote in message news:bob-841AE1.13524801102002@news.apple.com... > In article <1033479979.93726@newsreader2.wirehub.nl>, > "Renee Teunissen" > agree.with.these.terms@mail.zevenaar.org> wrote: > > > Take a look at this wonderfull FAQ: > > http://www.xs4all.nl/~borkhuis/vxworks/vxw_pt5.html > > search for "trcStack" > > I am trying to get a stack trace of the current stack, which is not > supported by the code from the FAQ (or tt()). trcStack() in general > doesn't appear to support getting a stack trace of the current stack. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Wed, 2 Oct 2002 09:18:32 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Are you serious ? To replace the OS as a cost cutting measure is one peice of BS I have heard too many times, and allways proved wrong. When you calculate the amount of effort it would take to port, and then to support the new OS, and compare it to the total amount of money you pay WindRiver, it never turns out cheaper to replace the OS. And if it does look cheaper, just call your sales guy and renegotiate the royalty price by showing him your proof that it would save you money to change the OS. "brian" wrote in message news:d2f67bd8.0210011108.5e6d9858@posting.google.com... > Hi All, > > Im presently using vxWorks for a product and would like to port it to > another royalty free OS as a cost cutting measure. The chip that Im > using has very few other supporting kernels, all of which cost a lot, > either in royalties or up fron costs. > > The application is very simple as far as the OS is concerned.. Theres > no TCPIP stack etc. Basicall just the kernel. What I need is a > scheduler, message passing, semaphores etc. Basic kernel > functionality. > > Any ideas on how to approach this? One idea I have is to buy/get the > source of a generic kernel and imtegrate it into my processor. Is that > as large a job as I think it is? Anyone suggest a good source-free > kernel out there I can buy (dont want Linux..too big) or download for > free? > > Any other ideas? > > Thanx in advance > > brian --------------------------- Newsgroups: comp.os.vxworks Subject: Re: bringingup vxworks on a board with two 750cxe processors Date: Wed, 2 Oct 2002 09:16:52 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: Reply-To: "Michael Lawnick" Sender: mlawnick@p5088fe38.dip.t-dialin.net Hi kiran, there should be no direct access of your memory (I'm looking into my PPC405 code). All references that might aim memory should be indirect via r1 or r31. So your processors should only reference their own domain. If not, you have a general problem, as system calls will get mixed up. Remember: you won't be able to use the same kernel-code with multiple processors. Every processor will need it's own vxWorks code located to a different address to avoid mixup of global variables. Regards, Michael "kiran Mandava" schrieb im Newsbeitrag news:ea368e49.0210011620.2e3bc5d1@posting.google.com... > Hello all, > > Since the interrupt vector base is not relocatable on powrePc 750 > how should i handle the interrupts on both the processors. if there is > an external interrupt on any of the processors it would jump to > location 0x500 where the ISR corresponding to the First CPU is > located, so there is a problem handling the interrupts from both the > processors as both of them try to save the interrupt conetext to the > same location..did anybody came across such problem and any sort of > feedback would be really helpful > > CPC 710 is the bridging device and both the processors use the same > SDRAM first half the first CPU and the otherhalf the other CPU > > > > thanks, > kiran --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Stack Trace of current stack? Date: Wed, 02 Oct 2002 00:36:55 -0700 From: Bob Bradley Organization: Uh huh huh...It says Organ Message-ID: References: <1033479979.93726@newsreader2.wirehub.nl> <1033542934.364066@newsreader1.wirehub.nl> In article <1033542934.364066@newsreader1.wirehub.nl>, "Renee Teunissen" wrote: > It does work, btw - I could trace the stack of the shell I was working > on :) When I try to use either tt() or trcStack() on the current task, it just prints an error that it cannot run trace the curret stack. The documentation for tt() describes this as a limitation as well. > But stacktracing of the currect task is an illusion, think about it. > The stack should - if correct - always end in trcStack or one of it's > sibblings or... not ;) Please rethink your problem.. It's okay if it ends in trcStack because I would still be able to see how it got there. For example, if a() calls b(), b() calls c(), and c() calls trcStack, I would be able to see: trcStack c b a --------------------------- Newsgroups: comp.os.vxworks Subject: Re: newbie - device drivers in c++ Date: Wed, 2 Oct 2002 10:32:43 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <290f4d3c.0210012203.38c6685a@posting.google.com> On one hand, there is no "kernel mode" in VxWorks, and thus there is very little difference between an application and a device driver. With the exception of a few standard devices (e.g. Network, storage) that need to comply with specific APIs, the hardware can be driven by application, and the notion of device driver in many cases is merely for convenience. On the other hand, C++ is a double-edged sword, and I would recommend against using this language for any code that accesses hardware, or is otherwise critical wrt performance, and that would naturally include any sort of device drivers. Except when it's in the hands of a very experienced C++ programmer who knows exactly when he/she is doing. IMHO "Rao R R" wrote in message news:290f4d3c.0210012203.38c6685a@posting.google.com... > Hi > Does vxworks supports c++ object handling at kernel level? Do some > one tried writing device drivers in vxworks in c++. Can someone give > insight into this ? What would be the adv and disadvanteges of writing > device drivers in c++. > > thanks in advance > Ranga Rao R --------------------------- Newsgroups: comp.os.vxworks Subject: Mixed C & Assembly listing Date: Wed, 2 Oct 2002 12:29:57 +0300 From: "Martin Roth" Organization: Motorola Message-ID: I am using Tornado 2.02 for MPC8xx with GNU. What i the easiest way to change the compiler switches in order to have a listing of mixed C and assembly code ? Thanks --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Mixed C & Assembly listing Date: Wed, 2 Oct 2002 11:08:33 +0100 From: Danny Chapman Organization: Lucent Technologies, Indian Hill Message-ID: References: The following (or something like it) might be useful: objdumpppc -l --source --disassemble file.o On Wed, 2 Oct 2002, Martin Roth wrote: > I am using Tornado 2.02 for MPC8xx with GNU. > What i the easiest way to change the compiler switches in order to have a > listing of mixed C and assembly code ? > > Thanks > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: bringingup vxworks on a board with two 750cxe processors Date: Wed, 02 Oct 2002 06:06:46 -0400 From: Varndell Engineering Organization: Posted Via Newsfeeds.com = SPEED+RETENTION+COMPLETION = http://www.newsfeeds.com Message-ID: <3D9AC536.8335091C@hotmail.com> References: kiran Mandava wrote: > > Hello all, > > Since the interrupt vector base is not relocatable on powrePc 750 > how should i handle the interrupts on both the processors. if there is > an external interrupt on any of the processors it would jump to > location 0x500 where the ISR corresponding to the First CPU is > located, so there is a problem handling the interrupts from both the > processors as both of them try to save the interrupt conetext to the > same location..did anybody came across such problem and any sort of > feedback would be really helpful > > CPC 710 is the bridging device and both the processors use the same > SDRAM first half the first CPU and the otherhalf the other CPU Assuming that each cpu is executing an image that has been linked to a different address, or each image executes within its own virtual space, I think you can do this. When the 750 takes an interrupt, it will be executing in real address mode for a short time. It will call intEnt() to restore MMU translation, if translation was enabled when the interrupt occurred, save context on the active stack, and return from intEnt to call whatever C function was installed by excConnect(). Right now, you have 2 images, each installing a C handler (different addresses) into the same physical address. Without working out all the details for you, I think you can install a common handler into the vector. This handler would need to decide which processor was executing it and call the interrupt handler for that processor. Obviously, this would require some hacking of the interrupt controller code -- the code that currently takes over the external interrupt vector. Given that, when the user handler is called (after return from intEnt()) virtual address translation is active, this might even be easy. The predictable disclaimer; I've never tried this before and I really haven't thought it out thoroughly. Regards, George Varndell - -- Varndell Engineering, LLC Embedded Systems Expertise Need an vxWorks END driver for Sis900 or DP83815? Contact us. http://www.varndellengineering.com - -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- http://www.newsfeed.com The #1 Newsgroup Service in the World! - -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Serial Debugging of vxWorks based image Date: 2 Oct 2002 03:38:55 -0700 From: amit.gil@seabridge.co.il (Amit G.) Organization: http://groups.google.com/ Message-ID: <9574ee82.0210020238.a5806c@posting.google.com> Hi. I'm trying to connect a serial debugger to a vxWorks based application using a serial debug jig. I built the version with WDB SERIAL communication compilation flag. When connecting to the card with "tip" I get the WDB Ready... message. When disconnecting the tip and trying to load the target server - I get error messages: Connecting to target agent... Error: rpccore backend client RPC: Timed out failed. Connecting to target agent... Error: rpccore backend client RPC: Timed out failed. Connecting to target agent... Error: clnttty_rcv: read timed out Error: clnttty_rcv: read timed out Error: clnttty_rcv: no end of frame in input Error: clnttty_rcv: no end of frame in input Error: clnttty_rcv: no end of frame in input Error: clnttty_rcv: no end of frame in input ) Any thoughts? I increased the timeout to 5 minutes - just in case. --------------------------- Newsgroups: comp.os.vxworks Subject: Target resident shell Date: 2 Oct 2002 03:43:09 -0700 From: amit.gil@seabridge.co.il (Amit G.) Organization: http://groups.google.com/ Message-ID: <9574ee82.0210020243.2f681441@posting.google.com> I'm trying to use the target resident shell with no success. I build the version with the USE_TARGET_SHELL I get the shell welcome screen on my serial console - yet no prompt. This is what I see: Attached TCP/IP interface to motscc unit 0 0xfbffdf8 (tRootTask): Attached network interface lo0 successfully NFS client support not included. Adding 15961 symbols for standalone. 0xfbffdf8 (tRootTask): rtrequest:: DEL don't free block e395fa0 ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]] ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]] ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]] ]]]]]]]]]]] ]]]] ]]]]]]]]]] ]] ]]]] (R) ] ]]]]]]]]] ]]]]]] ]]]]]]]] ]] ]]]] ]] ]]]]]]] ]]]]]]]] ]]]]]] ] ]] ]]]] ]]] ]]]]] ] ]]] ] ]]]] ]]] ]]]]]]]]] ]]]] ]] ]]]] ]] ]]]]] ]]]] ]]] ]] ] ]]] ]] ]]]]] ]]]]]] ]] ]]]]]]] ]]]] ]] ]]]] ]]]]] ] ]]]] ]]]]] ]]]]]]]] ]]]] ]] ]]]] ]]]]]]] ]]]] ]]]]]] ]]]]] ]]]]]] ] ]]]]] ]]]] ]] ]]]] ]]]]]]]] ]]]] ]]]]]]] ]]]]] ] ]]]]]] ] ]]] ]]]] ]] ]]]] ]]]] ]]]] ]]]] ]]]]]]]] ]]]]] ]]] ]]]]]]] ] ]]]]]]] ]]]] ]]]] ]]]] ]]]]] ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]] ]]]]]]]]]]]]]]]]]]]]]]]]]]]]] Development System ]]]]]]]]]]]]]]]]]]]]]]]]]]]] ]]]]]]]]]]]]]]]]]]]]]]]]]]] VxWorks version 5.4 ]]]]]]]]]]]]]]]]]]]]]]]]]] KERNEL: WIND version 2.5 ]]]]]]]]]]]]]]]]]]]]]]]]] Copyright Wind River Systems, Inc., 1984-1999 CPU: SIEMENS BIAS 2.1 - PPC7400. Processor #0. Memory Size: 0xfc00000. BSP version 1.0. WDB_COMM_TYPE: WDB_COMM_NETWORK WDB: Ready. After that I get some printouts generated by my application and that's about it, now visible communication with the shell. Any suggestions? --------------------------- Newsgroups: comp.os.vxworks Subject: Cold boot and Warm Boot Date: 2 Oct 2002 04:37:04 -0700 From: lawrielk@yahoo.co.in (Lawrie) Organization: http://groups.google.com/ Message-ID: <4ac248a4.0210020337.3eb6a4ca@posting.google.com> Hi VxWorks, I've got few doubts regarding the images that we make in VxWorks....seems to be easy for few brains who 've done lot of work in this field.... 1). What is Cold Boot and Warm Boot? 2). Why is the system startType....i.e. whether cold boot or warm boot..... taken as an argument in routines... a). romStart( int startType) b). usrInit ( int startType)....its also saved in an integer variable sysStartType, why? 3). Has anyone ever tried to port vxworks into Evalutor 7T board using ARM7TDMII architecture......its bcos we need serious help outhere ? Hoping to get some reply from the good souls..... Thanks. Lawrie. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Stack Trace of current stack? Date: Wed, 2 Oct 2002 13:40:27 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1033558864.591792@newsreader1.wirehub.nl> References: <1033479979.93726@newsreader2.wirehub.nl> <1033542934.364066@newsreader1.wirehub.nl> I see... What I would do then... get your currect framepointer. look: (pentium based target) - -> ti tPortmapd NAME ENTRY TID PRI STATUS PC SP ERRNO DELAY - ---------- ------------ -------- --- ---------- -------- -------- ---- - --- ----- tPortmapd _portmapd 3dc47dc 100 PEND 11a5b4 3dc46ac 16 0 stack: base 0x3dc47dc end 0x3dc3454 size 4988 high 528 margin 4460 options: 0x7 VX_SUPERVISOR_MODE VX_UNBREAKABLE VX_DEALLOC_STACK edi = 3dc47dc esi = 0 ebp = 3dc4744 esp = 3dc46ac ebx = 3dc4714 edx = 0 ecx = 0 eax = 0 eflags = 246 pc = 11a5b4 value = 0 = 0x0 The framepointer (ebp) = 0x3dc4744 - -> d 0x3dc4744 03dc4740: 478c 03dc 56e9 001a 0100 0000 * G...V......* 03dc4750: 476c 03dc 0000 0000 0000 0000 0000 0000 *lG..............* 03dc4760: ec74 03e7 479c 03dc 0000 0000 0000 0000 *t....G..........* 03dc4770: 0000 0000 0000 0000 0000 0000 0000 0000 *................* 03dc4780: 0000 0000 0000 0000 0000 0000 47ac 03dc *.............G..* 03dc4790: d963 0018 0000 0000 0000 0000 02ee 6f00 *c..............o* 03dc47a0: 0000 0000 eeee eeee 0530 0013 0000 0000 *........0.......* 03dc47b0: 053c 0013 0000 0000 0000 0000 0000 0000 *<...............* 03dc47c0: 0000 0000 0000 0000 0000 0000 0000 0000 *................* 03dc47d0: 0000 0000 0000 0000 0000 0000 0000 0000 *................* 03dc47e0: 0000 0000 0064 0000 47e8 c3dc 47e8 c3dc *....d....G...G..* 03dc47f0: 47f0 c3dc 47f0 c3dc 47f8 c3dc 32bc 03dc *.G...G...G...2..* 03dc4800: 1bf0 03e1 4800 c3dc 4808 c3dc a12c 001b *.....H...H..,...* 03dc4810: 3454 03dc 0007 0000 0002 0000 0064 0000 *T4..........d...* 03dc4820: 0064 0000 0000 0000 0000 0000 0000 0000 *d...............* 03dc4830: 0000 0000 0000 0000 4874 03dc 0000 0000 *........tH......* 03dc4840: 0000 0000 *................* value = 21 = 0x15 at that address you'll find the address of the next framepointer and on that address the next one until you'll find 0 here you find: ebp = 03dc.4744 -> 03dc.478c -> 03dc.47ac -> 0000.0000 the integer after that you'll find the return adres, which is the code space of the calling program. thus begin of function + offset. ebp = 03dc.4744, on 03dc.474c you'll find: 001a.56e9 so the call three is: this func is called from 001a.56e9 is called from 0018.d963 called from 0013053c if you apply: tCh_symName on 001a.56e9 you will get the symbol name + offset (diff contans this offset) put this is a function and... Please post your findings and code - if possible. also see: http://packetstormsecurity.nl/docs/infosec/buffer-overflows/random-sta ck.text which describes those nasty stackoverun exploids. Those docs are a prefect illustration how the stack works and how to use it for you own bennefits, which is in our case, stack tracing. You'll find more interesting docs on packetstorm with respect to stack and heap memory issues. - ------------------- LOCAL STATUS tCh_symName(void* addr, char* symName, int* diff) { SYM_TYPE t; int val; char buf[512]; if( symFindByValue(sysSymTbl, (int)addr, buf, &val, &t) == OK) { if( diff != NULL ) { *diff = (int) addr - val; if( *diff == 0) { sprintf(symName, "%#10x %16.16s", val, buf); } else { sprintf(symName, "%#10x %16.16s + 0x%x", val, buf, *diff); } return OK; } else { if( (int)addr != val) { sprintf(symName, "_0x%x", addr); return ERROR; } else { sprintf(symName, "%#10x %16.16s", val, buf); return OK; } } } sprintf(symName, "_0x%x", addr); return ERROR; } - ------------------- "Bob Bradley" wrote in message news:bob-0EE16C.00365502102002@news.apple.com... > In article <1033542934.364066@newsreader1.wirehub.nl>, > "Renee Teunissen" > agree.with.these.terms@mail.zevenaar.org> wrote: > > > It does work, btw - I could trace the stack of the shell I was working > > on :) > > When I try to use either tt() or trcStack() on the current task, it just > prints an error that it cannot run trace the curret stack. The > documentation for tt() describes this as a limitation as well. > > > But stacktracing of the currect task is an illusion, think about it. > > The stack should - if correct - always end in trcStack or one of it's > > sibblings or... not ;) Please rethink your problem.. > > It's okay if it ends in trcStack because I would still be able to see > how it got there. For example, if a() calls b(), b() calls c(), and c() > calls trcStack, I would be able to see: > > trcStack > c > b > a --------------------------- Newsgroups: comp.os.vxworks Subject: Standalone application Date: 2 Oct 2002 05:16:31 -0700 From: andreas_lindell75@hotmail.com (Rikard) Organization: http://groups.google.com/ Message-ID: <5947a04.0210020416.1528382d@posting.google.com> Hello, I have written an application in Tornado 2.0.2 for vxWorks 5.4.2 and now I want to compile it into a standalone application. Does anyone know how that is done, if so please provide me with some information. Regards Andreas --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Cold boot and Warm Boot Date: Wed, 2 Oct 2002 14:18:37 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <4ac248a4.0210020337.3eb6a4ca@posting.google.com> "Lawrie" wrote in message news:4ac248a4.0210020337.3eb6a4ca@posting.google.com... > Hi VxWorks, > > I've got few doubts regarding the images that we make in VxWorks....seems to > be easy for few brains who 've done lot of work in this field.... > A lot of work would be an understatement in my case. > 1). What is Cold Boot and Warm Boot? On Cold boot, the VxWorks bootRom will clear out the entire memory, assuming it starts after power-on, and assuming memory has either parity ECC, all memory bits need to be written first, so writing a 0 into all bytes makes sure that a perity error will not result from a read-before-write event. Worm boot circumvents this zeroeing phase, in order to save time. The above does not really apply for the x86/Bios (aka PC) architecture. Application may also use the startType as needed. > > 2). Why is the system startType....i.e. whether cold boot or warm boot..... > taken as an argument in routines... > a). romStart( int startType) > b). usrInit ( int startType)....its also saved in an integer variable > sysStartType, why? > romStart is actually using the argument (passed from fixed-position entry points) to decide to do the zeroing or not. usrInit does not use startType, except for propagating it towards any drivers or application that might make use of it. > 3). Has anyone ever tried to port vxworks into Evalutor 7T board using > ARM7TDMII architecture......its bcos we need serious help outhere ? > Nope, never did ARM, but did at least 20 BSPs on different architectures, i960 being the one I used most. > Hoping to get some reply from the good souls..... > Thanks. > Hope it helps. Leonid --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Cold boot and Warm Boot Date: Wed, 02 Oct 2002 08:36:30 -0400 From: Joe Durusau Organization: Lockheed Martin Corporation Message-ID: <3D9AE84E.F900EB37@lmco.com> References: <4ac248a4.0210020337.3eb6a4ca@posting.google.com> Lawrie wrote: > > Hi VxWorks, > > I've got few doubts regarding the images that we make in VxWorks....seems to > be easy for few brains who 've done lot of work in this field.... > > 1). What is Cold Boot and Warm Boot? > Cold boot is power up or hardware reset, warm boot is control-X to the shell or the reboot command. > 2). Why is the system startType....i.e. whether cold boot or warm boot..... > taken as an argument in routines... > a). romStart( int startType) It has to do with whether memory is cleared, among other things. > b). usrInit ( int startType)....its also saved in an integer variable > sysStartType, why? > usrInt could use it, never looked at exactly what is does. > 3). Has anyone ever tried to port vxworks into Evalutor 7T board using > ARM7TDMII architecture......its bcos we need serious help outhere ? Not I. You can get a copy of the BSP porting kit from WRS to start with. Call them and have you checkboot ready. > > Hoping to get some reply from the good souls..... > Thanks. > > Lawrie. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Standalone application Date: Wed, 02 Oct 2002 08:40:42 -0400 From: Joe Durusau Organization: Lockheed Martin Corporation Message-ID: <3D9AE94A.10DE39F2@lmco.com> References: <5947a04.0210020416.1528382d@posting.google.com> Rikard wrote: > > Hello, > > I have written an application in Tornado 2.0.2 for vxWorks 5.4.2 and > now I want to compile it into a standalone application. Does anyone > know how that is done, if so please provide me with some information. > > Regards > Andreas If you are using the Makefiles, set MACH_EXTRA to the name of your application object, add a call to your startup code in usrConfig.c (search for DEMO for an example), make vxWorks.st_rom or some similar target, download to PROM, and you are there. If you are using the GUI, see the FAQ, there is probably an example there. Never tried it that way. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Target resident shell Date: Wed, 02 Oct 2002 08:42:27 -0400 From: Joe Durusau Organization: Lockheed Martin Corporation Message-ID: <3D9AE9B3.50A87B4A@lmco.com> References: <9574ee82.0210020243.2f681441@posting.google.com> "Amit G." wrote: > > I'm trying to use the target resident shell with no success. > > I build the version with the USE_TARGET_SHELL > > I get the shell welcome screen on my serial console - yet no prompt. > > This is what I see: > Attached TCP/IP interface to motscc unit 0 > 0xfbffdf8 (tRootTask): Attached network interface lo0 successfully > NFS client support not included. > > Adding 15961 symbols for standalone. > 0xfbffdf8 (tRootTask): rtrequest:: DEL don't free block e395fa0 > > > ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]] > ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]] > ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]] > ]]]]]]]]]]] ]]]] ]]]]]]]]]] ]] ]]]] > (R) > ] ]]]]]]]]] ]]]]]] ]]]]]]]] ]] ]]]] > ]] ]]]]]]] ]]]]]]]] ]]]]]] ] ]] ]]]] > ]]] ]]]]] ] ]]] ] ]]]] ]]] ]]]]]]]]] ]]]] ]] ]]]] ]] > ]]]]] > ]]]] ]]] ]] ] ]]] ]] ]]]]] ]]]]]] ]] ]]]]]]] ]]]] ]] > ]]]] > ]]]]] ] ]]]] ]]]]] ]]]]]]]] ]]]] ]] ]]]] ]]]]]]] > ]]]] > ]]]]]] ]]]]] ]]]]]] ] ]]]]] ]]]] ]] ]]]] ]]]]]]]] > ]]]] > ]]]]]]] ]]]]] ] ]]]]]] ] ]]] ]]]] ]] ]]]] ]]]] ]]]] > ]]]] > ]]]]]]]] ]]]]] ]]] ]]]]]]] ] ]]]]]]] ]]]] ]]]] > ]]]] ]]]]] > ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]] > ]]]]]]]]]]]]]]]]]]]]]]]]]]]]] Development System > ]]]]]]]]]]]]]]]]]]]]]]]]]]]] > ]]]]]]]]]]]]]]]]]]]]]]]]]]] VxWorks version 5.4 > ]]]]]]]]]]]]]]]]]]]]]]]]]] KERNEL: WIND version 2.5 > ]]]]]]]]]]]]]]]]]]]]]]]]] Copyright Wind River Systems, Inc., > 1984-1999 > > CPU: SIEMENS BIAS 2.1 - PPC7400. > Processor #0. > Memory Size: 0xfc00000. BSP version > 1.0. > WDB_COMM_TYPE: WDB_COMM_NETWORK > WDB: Ready. > > After that I get some printouts generated by my application and that's > about it, now visible communication with the shell. > > Any suggestions? For one thing, you don't want WDB. take a look at config.h. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Problems making boot floppy for T2.2/VxWorks 5.5 Date: 2 Oct 2002 05:47:14 -0700 From: suenews@schooner.com (Susan LoVerso) Organization: http://groups.google.com/ Message-ID: <27127820.0210020447.6a624277@posting.google.com> References: <27127820.0210010901.375ca50b@posting.google.com> <488e459a.0210012129.1d3e9ad3@posting.google.com> john_94501@yahoo.com (John) wrote in message news:<488e459a.0210012129.1d3e9ad3@posting.google.com>... > Hello Susan, > > Do you still have the VxWorks AE boot floppies? That should be able to > boot a 5.5 kernel image (just point the filename at your vxWorks image > file). From there, you can use the mkboot tool for VxWorks to create > yourself a 5.5 boot floppy. Yes I do. That was one idea I had not thought of. Thanks a lot John! [long pause] I can now boot 5.5! It does go to the serial line, like AE, not the monitor like 5.4. I am up and running. Thanks again. --------------------------- Newsgroups: comp.os.vxworks Subject: Adding PCI support in Vx. Date: Wed, 02 Oct 2002 09:17:37 -0400 From: Jeremiah Frick Organization: Improv Systems, Inc. Message-ID: <3D9AF1F1.4020508@improvsys.com> Hello Everyone, I'm attempting to add PCI support in VxWorks on the ARM Integrator 940t for a card I have. When I initially set the system up, I could detect the PCI card and read the 9050 configuration registers (vendor ID, device ID, etc) without any problem. However, the BAR addresses weren't being set up at all. So, I did some looking around and added some code to sysLib.c in the BSP: { UINT32 pciBus; /* PCI Bus number */ UINT32 pciDevice; /* PCI Device number */ UINT32 pciFunc; /* PCI Function number */ if( pciFindDevice( VENDOR_ID, DEVICE_ID, 0, &pciBus, &pciDevice, &pciFunc ) == OK ) { pciAssignResources( pciBus, pciDevice, pciFunc ); } } This was added inside the SysHwInit function. Using this software I now get at least a BAR2 address (this is the only one I checked initially). However, the address given isn't consistant with PCI memory space. In my system the BAR2 address I'm getting is 0x01000000, and the PCI space should be 0x40000000 at the least. I was wondering if the value returned is simply an offset into the PCI base memory space or if I'm still experiencing problems and need to add more software. Any assistance would be greatly appreciated. Sincerely, Jeremy --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Target resident shell Date: Wed, 2 Oct 2002 22:19:25 +0900 From: "Girish Gulawani" Message-ID: <3d9af25b$0$22475$44c9b20d@news3.asahi-net.or.jp> References: <9574ee82.0210020243.2f681441@posting.google.com> 1> what is the priority of your application task? 2> by any chance did you press ENTER key after application messages...;-) "Amit G." wrote in message news:9574ee82.0210020243.2f681441@posting.google.com... > I'm trying to use the target resident shell with no success. > > I build the version with the USE_TARGET_SHELL > > I get the shell welcome screen on my serial console - yet no prompt. > > This is what I see: > Attached TCP/IP interface to motscc unit 0 > 0xfbffdf8 (tRootTask): Attached network interface lo0 successfully > NFS client support not included. > > Adding 15961 symbols for standalone. > 0xfbffdf8 (tRootTask): rtrequest:: DEL don't free block e395fa0 > > > ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]] > ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]] > ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]] > ]]]]]]]]]]] ]]]] ]]]]]]]]]] ]] ]]]] > (R) > ] ]]]]]]]]] ]]]]]] ]]]]]]]] ]] ]]]] > ]] ]]]]]]] ]]]]]]]] ]]]]]] ] ]] ]]]] > ]]] ]]]]] ] ]]] ] ]]]] ]]] ]]]]]]]]] ]]]] ]] ]]]] ]] > ]]]]] > ]]]] ]]] ]] ] ]]] ]] ]]]]] ]]]]]] ]] ]]]]]]] ]]]] ]] > ]]]] > ]]]]] ] ]]]] ]]]]] ]]]]]]]] ]]]] ]] ]]]] ]]]]]]] > ]]]] > ]]]]]] ]]]]] ]]]]]] ] ]]]]] ]]]] ]] ]]]] ]]]]]]]] > ]]]] > ]]]]]]] ]]]]] ] ]]]]]] ] ]]] ]]]] ]] ]]]] ]]]] ]]]] > ]]]] > ]]]]]]]] ]]]]] ]]] ]]]]]]] ] ]]]]]]] ]]]] ]]]] > ]]]] ]]]]] > ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]] > ]]]]]]]]]]]]]]]]]]]]]]]]]]]]] Development System > ]]]]]]]]]]]]]]]]]]]]]]]]]]]] > ]]]]]]]]]]]]]]]]]]]]]]]]]]] VxWorks version 5.4 > ]]]]]]]]]]]]]]]]]]]]]]]]]] KERNEL: WIND version 2.5 > ]]]]]]]]]]]]]]]]]]]]]]]]] Copyright Wind River Systems, Inc., > 1984-1999 > > CPU: SIEMENS BIAS 2.1 - PPC7400. > Processor #0. > Memory Size: 0xfc00000. BSP version > 1.0. > WDB_COMM_TYPE: WDB_COMM_NETWORK > WDB: Ready. > > > After that I get some printouts generated by my application and that's > about it, now visible communication with the shell. > > Any suggestions? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Instantiating "extern" variables Date: 02 Oct 2002 09:32:45 -0400 From: "Paul D. Smith" Organization: Nortel Networks, Inc. / Billerica, MA Message-ID: References: <488e459a.0210012051.19403d6c@posting.google.com> Reply-To: pausmith@nortelnetworks.com %% john_94501@yahoo.com (John) writes: j> I can tell you that the Linux-hosted VxWorks AE toolchain requires j> a non-extern instance somewhere to link correctly though, and I j> believe that the Solaris hosted AE toolchain does too. Of course, j> that is 2.96 not 2.7.2. Yes, we did also run into this issue if we tried to build with a newer GCC, so I think the default handling of "common" must have changed between WRS 2.7.2 and 2.96. But for this particular platform the decision was made to stay with the existing compiler... - -- - ------------------------------------------------------------------------------- Paul D. Smith HASMAT--HA Software Mthds & Tools "Please remain calm...I may be mad, but I am a professional." --Mad Scientist - ------------------------------------------------------------------------------- These are my opinions---Nortel Networks takes no responsibility for them. --------------------------- Newsgroups: comp.os.vxworks Subject: for help Date: 2 Oct 2002 06:53:54 -0700 From: buaanupt@sina.com (ZHANG Yan) Organization: http://groups.google.com/ Message-ID: <9e847e5e.0210020553.4a51473d@posting.google.com> Hi,I am a beginner in VxWorks development, can you plz give me some links on VxWorks learning and embedded development,especially wireless or mobile programming. Thanks and Regards. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Task Reschedule on mutex unlock Date: Wed, 2 Oct 2002 14:05:29 +0000 (UTC) From: David Laight Organization: BT Openworld Message-ID: References: <3D908FCF.F75EE09C@alcatel.com> Leonid Rosenboim wrote: > David, there already is a list of task waiting on a mutex - > it's head is in the mutex semaphore itself ! That is a list for a specific mutex, I was thinking of a global list for all mutex. David > P.S. there is no spare list, I think. Don't need to memorize source > (like me), just look at header files. I don't have access to the header files at the moment, I've never had the source - just an enourmous disassembly of the entire image :-) --------------------------- Newsgroups: comp.os.vxworks Subject: Re: bringingup vxworks on a board with two 750cxe processors Date: 2 Oct 2002 07:09:00 -0700 From: kiran.mandava@skystream.com (kiran Mandava) Organization: http://groups.google.com/ Message-ID: References: <3D9AC536.8335091C@hotmail.com> Yes each of My CPU have their own images and they are located at different places in the memory. lets sayfirst CPU 0x00000000 to 0x04000000 and the sencond one from 0x04000000 to 0x08000000.(image loading starts at an offset 0x20000) ok when any of the processors get an external interrupt they jump to 0x500. assume that second CPU didn't put any code at 0x500, but in the first CPU's iSR i would jump to different location depending on the CPU number. the first thing it does in the code at 0x500 is it tries to save the context at a location interrupt_stack_base(this is in reference to a variable vxIntStackBase or something), as the code is put in by the first CPU the address belongs to the first CPU, (it doesn't use the r1 to store the context on stack, so that the interrupt stack would be totally different from the current r1), so if there is an interrupt at the same time there is a problem here.! intEnt does it enable the Real mode. i wasn't sure, if that is the first thing it does probably that might help me solving this probem.! thanks, kiran >Varndell Engineering wrote in message >news:<3D9AC536.8335091C@hotmail.com>... > kiran Mandava wrote: > > > > Hello all, > > > > Since the interrupt vector base is not relocatable on powrePc 750 > > how should i handle the interrupts on both the processors. if there is > > an external interrupt on any of the processors it would jump to > > location 0x500 where the ISR corresponding to the First CPU is > > located, so there is a problem handling the interrupts from both the > > processors as both of them try to save the interrupt conetext to the > > same location..did anybody came across such problem and any sort of > > feedback would be really helpful > > > > CPC 710 is the bridging device and both the processors use the same > > SDRAM first half the first CPU and the otherhalf the other CPU > > Assuming that each cpu is executing an image that has been linked to > a different address, or each image executes within its own virtual > space, I think you can do this. > When the 750 takes an interrupt, it will be executing in real address > mode for a short time. It will call intEnt() to restore MMU translation, > if translation was enabled when the interrupt occurred, save context > on the active stack, and return from intEnt to call whatever C function > was installed by excConnect(). Right now, you have 2 images, each > installing a C handler (different addresses) into the same physical > address. > Without working out all the details for you, I think you can install > a common handler into the vector. This handler would need to decide > which processor was executing it and call the interrupt handler for > that processor. Obviously, this would require some hacking of the > interrupt controller code -- the code that currently takes over > the external interrupt vector. Given that, when the user handler > is called (after return from intEnt()) virtual address translation > is active, this might even be easy. > The predictable disclaimer; I've never tried this before and I > really haven't thought it out thoroughly. > > Regards, > George Varndell > -- > Varndell Engineering, LLC Embedded Systems Expertise > Need an vxWorks END driver for Sis900 or DP83815? Contact us. > http://www.varndellengineering.com > > > -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- > http://www.newsfeed.com The #1 Newsgroup Service in the World! > -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Stack Trace of current stack? Date: Wed, 2 Oct 2002 14:14:53 +0000 (UTC) From: David Laight Organization: BT Openworld Message-ID: References: <1033479979.93726@newsreader2.wirehub.nl> <1033542934.364066@newsreader1.wirehub.nl> <1033558864.591792@newsreader1.wirehub.nl> Renee Teunissen wrote: > I see... > What I would do then... > > get your currect framepointer. Yes - for some reason the tt function doesn't use the frame pointer, instead it (tries to) parse the instructions to find an exit path from each routine and hence the return address. Worse still, it does this recursively and only prints anything once it has found the stack base. This is done solely to reverse the order of the printout - which is completely unnecessary. If (ie when) it fails to find the stack base (eg anything called as a native method from pjava on a strongarm) you get no info at all..... If you have stack frames then they can be used to trivially generate a backtrace. Actually I wonder if it possible to patch out the check that stops tt() working on the current process? I didn't ever try. David --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Mixed C & Assembly listing Date: Wed, 2 Oct 2002 16:49:19 +0300 From: "Martin Roth" Organization: Motorola Message-ID: References: Danny, I have tried objdumpppc -l --source --debugging file.o I am getting the assembly listing, but I do not see the C source code. What I see is the line number of the C source code, but without the line itself. Martin "Danny Chapman" wrote in message news:Pine.GSO.4.43.0210021101250.22678-100000@mlgw069.uk.lucent.com... > > The following (or something like it) might be useful: > > objdumpppc -l --source --disassemble file.o > > > On Wed, 2 Oct 2002, Martin Roth wrote: > > > I am using Tornado 2.02 for MPC8xx with GNU. > > What i the easiest way to change the compiler switches in order to have a > > listing of mixed C and assembly code ? > > > > Thanks > > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Stack Trace of current stack? Date: Wed, 2 Oct 2002 16:45:37 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1033569972.232848@newsreader1.wirehub.nl> References: <1033479979.93726@newsreader2.wirehub.nl> <1033542934.364066@newsreader1.wirehub.nl> <1033558864.591792@newsreader1.wirehub.nl> "David Laight" wrote in message news:anev0s$dak$1@helle.btinternet.com... > Renee Teunissen wrote: > > I see... > Yes - for some reason the tt function doesn't use the frame pointer, > instead it (tries to) parse the instructions to find an exit path > from each routine and hence the return address. I never analysed what tt() does, just tried to find an answer :) this should work on INTEL: - --------------------------------------------- int tCh_Trace(unsigned int ebp) { unsigned int *fbptr; unsigned int *retad; fbptr = (unsigned int *)ebp; retad = fbptr+1; do { printf("fbptr : %p: %8.8x\n", fbptr, *fbptr); printf("retaddress : %8.8x\n", *retad); printf("---------------\n"); fbptr = (unsigned int *) *fbptr; retad = fbptr+1; } while( fbptr ); return 0; } - --------------------------------------------- just call it with the ebp you get from taskRegsGet or using another method :) > > Worse still, it does this recursively and only prints anything > once it has found the stack base. This is done solely to > reverse the order of the printout - which is completely unnecessary. I agree. Thou tt() displays info which cant be found otherwise, i guess.. like the callee's address. I thought I should find a 'call
' intruction at (*retad)-1 or something like that and then I found out there is an CALL EAX :( (I'm not that deep in asembler) > If (ie when) it fails to find the stack base (eg anything called > as a native method from pjava on a strongarm) you get no info at > all..... I know.. > If you have stack frames then they can be used to trivially > generate a backtrace. > > Actually I wonder if it possible to patch out the check that > stops tt() working on the current process? > I didn't ever try. call trcStack you current process and see what happens ;) put the code in johan borkhuis FAQ in to a file and compile it, you'll see what happens... (my expiriences differ from a endless loop until a reboot) The routine above can be extented to find much more. dont have time to do that, this should give a smart guy/girl enought idears... as we all known from our college textbooks "... is left as an exercise to the reader" ;) And read those hacking dox at packetstormsecurity.org, they'll provide at lot of info to geet some real debugging expirience :) Cheers, Renee - - - - - - - - - Renee A. Teunissen PTS Software bv, Meerweg 7, 1405BA Bussum, NL. T.+31-(0)35-6926969, M.+31-(0)6-22778313, http://www.pts.nl, @pts.nl personal link page: http://wittenburg10c.nl/db/dest/links.html --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Mixed C & Assembly listing Date: Wed, 2 Oct 2002 16:40:41 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: Reply-To: "Michael Lawnick" Sender: mlawnick@p5088fe38.dip.t-dialin.net Hi Martin, I had the same problem. After upgrade to T2.2 objdumpppc now works as expected. Michael "Martin Roth" schrieb im Newsbeitrag news:anetm5$ptf$1@newshost.mot.com... > Danny, > I have tried objdumpppc -l --source --debugging file.o > > I am getting the assembly listing, but I do not see the C source code. > What I see is the line number of the C source code, but without the line > itself. > Martin > > > "Danny Chapman" wrote in message > news:Pine.GSO.4.43.0210021101250.22678-100000@mlgw069.uk.lucent.com... > > > > The following (or something like it) might be useful: > > > > objdumpppc -l --source --disassemble file.o > > > > > > On Wed, 2 Oct 2002, Martin Roth wrote: > > > > > I am using Tornado 2.02 for MPC8xx with GNU. > > > What i the easiest way to change the compiler switches in order to have > a > > > listing of mixed C and assembly code ? > > > > > > Thanks > > > > > > > > > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Mixed C & Assembly listing Date: 02 Oct 2002 10:26:09 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: References: Sender: bpringle@DeadDuck Martin> I have tried objdumpppc -l --source --debugging file.o Martin> I am getting the assembly listing, but I do not see the C Martin> source code. What I see is the line number of the C source Martin> code, but without the line itself. If you are doing this on a Windows platform and compiling with the Tornado projects, it appears that the tools put weird paths in the object files. With the ARM and COFF files, I have used a hex editor to zap the file name string in the binary. The "--source" option only works if the source file can be found. If you were to recompile the file in the same directory with the same options, it should find it. Ie, change "ccppc {compile options} -c path/to/source/file.c -o path/to/obj/file.o" to "ccppc {compile options} -c file.c -o file.o" After copying the source "file.c" to the directory where the Makefile is run. There still may be other problems, but this might help. hth, Bill Pringlemeir. - -- ..I must be a VETERINARIAN.. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: bringingup vxworks on a board with two 750cxe processors Date: Wed, 2 Oct 2002 16:48:08 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <3D9AC536.8335091C@hotmail.com> Reply-To: "Michael Lawnick" Sender: mlawnick@p5088fe38.dip.t-dialin.net Hi kiran, I'm pretty sure you will have to replace the code at 0x500 by a piece of your own, where you detect the CPU and branch either to lower or higher CPU- image. Another approach: For the higher CPU redefine the MMU-entry of the lowest page in exception mode and put an address-translation into it . Regards, Michael "kiran Mandava" schrieb im Newsbeitrag news:ea368e49.0210020608.14976189@posting.google.com... > Yes each of My CPU have their own images and they are located at > different places in the memory. > > lets sayfirst CPU 0x00000000 to 0x04000000 and the sencond one from > 0x04000000 to 0x08000000.(image loading starts at an offset 0x20000) > > ok when any of the processors get an external interrupt they jump to > 0x500. assume that second CPU didn't put any code at 0x500, but in the > first CPU's iSR i would jump to different location depending on the > CPU number. the first thing it does in the code at 0x500 is it tries > to save the context at a location interrupt_stack_base(this is in > reference to a variable vxIntStackBase or something), as the code is > put in by the first CPU the address belongs to the first CPU, (it > doesn't use the r1 to store the context on stack, so that the > interrupt stack would be totally different from the current r1), so if > there is an interrupt at the same time there is a problem here.! > > > intEnt does it enable the Real mode. i wasn't sure, if that is the > first thing it does probably that might help me solving this probem.! > > > thanks, > kiran > > >Varndell Engineering wrote in message >news:<3D9AC536.8335091C@hotmail.com>... > > kiran Mandava wrote: > > > > > > Hello all, > > > > > > Since the interrupt vector base is not relocatable on powrePc 750 > > > how should i handle the interrupts on both the processors. if there is > > > an external interrupt on any of the processors it would jump to > > > location 0x500 where the ISR corresponding to the First CPU is > > > located, so there is a problem handling the interrupts from both the > > > processors as both of them try to save the interrupt conetext to the > > > same location..did anybody came across such problem and any sort of > > > feedback would be really helpful > > > > > > CPC 710 is the bridging device and both the processors use the same > > > SDRAM first half the first CPU and the otherhalf the other CPU > > > > Assuming that each cpu is executing an image that has been linked to > > a different address, or each image executes within its own virtual > > space, I think you can do this. > > When the 750 takes an interrupt, it will be executing in real address > > mode for a short time. It will call intEnt() to restore MMU translation, > > if translation was enabled when the interrupt occurred, save context > > on the active stack, and return from intEnt to call whatever C function > > was installed by excConnect(). Right now, you have 2 images, each > > installing a C handler (different addresses) into the same physical > > address. > > Without working out all the details for you, I think you can install > > a common handler into the vector. This handler would need to decide > > which processor was executing it and call the interrupt handler for > > that processor. Obviously, this would require some hacking of the > > interrupt controller code -- the code that currently takes over > > the external interrupt vector. Given that, when the user handler > > is called (after return from intEnt()) virtual address translation > > is active, this might even be easy. > > The predictable disclaimer; I've never tried this before and I > > really haven't thought it out thoroughly. > > > > Regards, > > George Varndell > > -- > > Varndell Engineering, LLC Embedded Systems Expertise > > Need an vxWorks END driver for Sis900 or DP83815? Contact us. > > http://www.varndellengineering.com > > > > > > -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- > > http://www.newsfeed.com The #1 Newsgroup Service in the World! > > -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Mixed C & Assembly listing Date: Wed, 2 Oct 2002 15:46:01 +0100 From: Danny Chapman Organization: Lucent Technologies, Indian Hill Message-ID: References: On Wed, 2 Oct 2002, Martin Roth wrote: > Danny, > I have tried objdumpppc -l --source --debugging file.o > > I am getting the assembly listing, but I do not see the C source code. > What I see is the line number of the C source code, but without the line > itself. You might have needed to compile with -g in the first place. Two other things I've got written in my notebook... gcc -g -Wa,-ahl=file.s -c file.c (this certainly works under solaris) or... there's a free perl program called showasm that takes an assembly file (compile with -S) and intermingles it with the source. I haven't had cause to do this stuff for a while, so I can't remember the details! Also, I use MPC755, and the options may differ slightly... - - Danny --------------------------- Newsgroups: comp.os.vxworks Subject: question: partition a disk using dosFS Date: Wed, 2 Oct 2002 16:56:26 +0200 From: "Khac Tri VU" Organization: Planet Internet Message-ID: Hi folks, Does anyone know which dosFS function can be used to partition a disk? Thank you. Tri. --------------------------- Newsgroups: comp.os.vxworks Subject: UDP broadcast down multiple ENDs Date: Wed, 2 Oct 2002 16:28:26 +0100 From: "Kevin Garvey" Organization: Esat Net Customer Message-ID: Hi, Does anybody know how to configure the vxWorks IP stack such that a UDP broadcast (i.e. sendto() - destination 255.255.255.255) will be sent on all configured ENDs? Currently, I have two END attached to the IP stack and I can only get the IP stack to forward the broadcast datagram down the interface which has the default gateway (i.e. 0.0.0.0) configured against it. If I delete the route to the default gateway, the sendto() call fails with error EHOSTUNREACH. I've been through most of the archive on this news group and can't find any answers. All suggestion will be gratefully received. regards, Kevin Garvey. kgarvey@com21.ie Com21, Cork Airport Business Park, Cork, Ireland. Home Page: http://www.com21.com/ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: 2 Oct 2002 08:48:57 -0700 From: bmcgonig@yahoo.com (Brian) Organization: http://groups.google.com/ Message-ID: <3a5ff53f.0210020748.3643d45a@posting.google.com> References: "Leonid Rosenboim" wrote in message news:... > Are you serious ? > > To replace the OS as a cost cutting measure is one peice of BS > I have heard too many times, and allways proved wrong. > HP just did it with their some of their printers. Have you proved them wrong yet? It's very possible that WRS just wont negotiate down since the technical risk is deemed high. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Problems making boot floppy for T2.2/VxWorks 5.5 Date: 2 Oct 2002 09:23:58 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0210020823.324e8351@posting.google.com> References: <27127820.0210010901.375ca50b@posting.google.com> <488e459a.0210012129.1d3e9ad3@posting.google.com> Hello Susan, It seems that in 5.5 you should have the relevant instructions for creating your own boot floppy on the host. Check in your BSP's target.nr or target.txt file for instructions. The Pentium team lead sent me this: Creating bootable diskettes from Solaris Use /usr/bin/fdformat that comes with Solaris. It requires a bootstrap loader file called vxld.bin located in your Tornado directory tree at $WIND_BASE/host/sun4-solaris2/bin/vxld.bin Insert a 1.44MB diskette into the Sun diskette drive, and issue the fdformat command to format the diskette and install the boot block. % fdformat -U -d -B $WIND_BASE/host/sun4-solaris2/bin/vxld.bin Formatting 1.44 MB in /vol/dev/rdiskette0/no_name#0 Press return to start formatting floppy. .................................................................... fdformat: using "vxld.bin" for MS-DOS boot loader Eject and re-insert the diskette. On many systems the eject command is required. % eject /vol/dev/rdiskette0/no_name can now be manually ejected % volcheck Copy the bootrom image to the diskette. Use cp for bin images or objcopypentium for all other images. % cp bootrom.bin /floppy/floppy0/bootrom.sys ... % objcopypentium -O binary bootrom /floppy/floppy0/bootrom.sys HTH, John... john_94501@yahoo.com (John) wrote in message news:<488e459a.0210012129.1d3e9ad3@posting.google.com>... > Hello Susan, > > Do you still have the VxWorks AE boot floppies? That should be able to > boot a 5.5 kernel image (just point the filename at your vxWorks image > file). From there, you can use the mkboot tool for VxWorks to create > yourself a 5.5 boot floppy. > > If you don't have the VxWorks AE images, they can be downloaded from > WindSurf still. I ask for the 5.5 images to be put up there too... > > HTH, > > John... > > suenews@schooner.com (Susan LoVerso) wrote in message news:<27127820.0210010901.375ca50b@posting.google.com>... > > I recently received T2.2/VxWorks 5.5 and I am trying to create boot floppies > > to no avail. I am using a Solaris host and x86 target (p2). Here is what > > I have attempted. > > > > 1. Run T2.2, config/build my VxWorks image, and build my boot ROM images, > > bootrom_uncmp and bootrom_uncmp.bin. Copy the bootroms to a target-accessible > > location on my Solaris box. Exit T2.2. > > 2. Run T2.0 on host/VxWorks 5.4 on my target. Run mkbootFd on 5.4 and > > create floppy with no errors. (I tried both images, both succeed in mkbootFd > > but neither appears to boot.) > > 3. When I attempt to use this floppy, it reads the image, as shown by the > > V1.6 +++ ... but it apparently has a problem when it jumps to start executing it. > > (I assume I should see something similar to what I see when I boot > > VxWorks 5.4, blue screen, countdown to auto-boot, etc. and that it > > is "quick" - I wait about 30 seconds before deciding it is dead/hung.) > > > > I also have a tip line to a serial console that doesn't show anything (in > > case 5.5 changed to that default from the monitor, like it did for AE). > > > > The mkboot.c is identical in my 5.4 and 5.5 tree. I can use the same command > > and am successful in making boot floppies that work (using the same, > > brand-new floppy) using a 5.4 bootrom_uncmp. > > > > Any ideas on how to debug this would be welcome. Is there some 5.4/5.5 > > interaction I am not aware of preventing this from working? If so, how do > > I make a boot floppy if I cannot boot 5.5 and I am on Solaris? Does anyone > > have a generic 5.5 boot image for a p2 they could 'dd' off the floppy > > and send me? I would appreciate it. Thanks! --------------------------- Newsgroups: comp.os.vxworks Subject: Strange " warning: control reaches end of non-void function" Date: 2 Oct 2002 09:27:55 -0700 From: ofer3443@yahoo.com (Ofer Rotschield) Organization: http://groups.google.com/ Message-ID: Please advise, >>> /*** Compile using: ccppc -mcpu=405 -mstrict-align -ffor-scope -pipe -ansi -fno-builtin - -Wall -Wno-switch -nostdinc -c -DCPU=PPC405 -g x.C results: x.C: In function `foo': x.C:29: warning: unused variable `v1' x.C:33: warning: control reaches end of non-void function `foo()' Compiler: cygnus-2.7.2-960126 egcs-971225 tornado 2.0 Host: SunOS vardia 5.8 Generic_108528-15 sun4u sparc SUNW,UltraSPARC-IIi-Engine ***/ class cc { public: ~cc() {}; }; int foo () { cc v1[1]; cc v2[1]; return 1; } <<< --------------------------- Newsgroups: comp.os.vxworks Subject: Re: bringingup vxworks on a board with two 750cxe processors Date: Wed, 02 Oct 2002 10:39:00 -0400 From: Varndell Engineering Organization: Posted Via Newsfeeds.com = SPEED+RETENTION+COMPLETION = http://www.newsfeeds.com Message-ID: <3D9B0503.7A00423F@hotmail.com> References: <3D9AC536.8335091C@hotmail.com> kiran Mandava wrote: > > Yes each of My CPU have their own images and they are located at > different places in the memory. > > lets sayfirst CPU 0x00000000 to 0x04000000 and the sencond one from > 0x04000000 to 0x08000000.(image loading starts at an offset 0x20000) > > ok when any of the processors get an external interrupt they jump to > 0x500. assume that second CPU didn't put any code at 0x500, but in the > first CPU's iSR i would jump to different location depending on the > CPU number. the first thing it does in the code at 0x500 is it tries > to save the context at a location interrupt_stack_base(this is in > reference to a variable vxIntStackBase or something), as the code is > put in by the first CPU the address belongs to the first CPU, (it > doesn't use the r1 to store the context on stack, so that the > interrupt stack would be totally different from the current r1), so if > there is an interrupt at the same time there is a problem here.! > > intEnt does it enable the Real mode. i wasn't sure, if that is the > first thing it does probably that might help me solving this probem.! Well, the restoration of the MSR translation bits is not the first thing that gets done, but it's close. And it is definitely done before R1 is loaded with the address of the interrupt stack. Also, just to be sure about terminology, intEnt doesn't enable real addressing mode. That happens automatically when the interrupt exception takes place. IntEnt enables virtual addressing mode, if virtual addressing mode was in effect before the interrupt. BTW, does this board have 2 interrupt controllers? Regards, George Varndell - -- Varndell Engineering, LLC Embedded Systems Expertise Need a vxWorks END driver for SiS900 or DP83815? Contact us. http://www.varndellengineering.com - -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- http://www.newsfeed.com The #1 Newsgroup Service in the World! - -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Wed, 2 Oct 2002 18:57:38 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <3a5ff53f.0210020748.3643d45a@posting.google.com> Now, don't take me wrong here, I never said there eint a reason to throw out WindRiver, but cost eint that reason, son! I know that HP has replaced VxWorks with something else in their printers, but can you really prove that cost was the reason ? You know there are a dozen of other reasons that could be the cause: like lack of support, too long to fix bugs, no source code availability, ... I mean, if you're pissed at WRS - you're not alone! Go ahead, throw them out, join the masses, just don't tell me it's for cost! "Brian" wrote in message news:3a5ff53f.0210020748.3643d45a@posting.google.com... > "Leonid Rosenboim" wrote in message news:... > > Are you serious ? > > > > To replace the OS as a cost cutting measure is one peice of BS > > I have heard too many times, and allways proved wrong. > > > > HP just did it with their some of their printers. Have you proved > them wrong yet? It's very possible that WRS just wont negotiate down > since the technical risk is deemed high. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: question: partition a disk using dosFS Date: Wed, 2 Oct 2002 18:58:22 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Which dosFs version ? "Khac Tri VU" wrote in message news:anf1k5$fbp$1@reader13.wxs.nl... > Hi folks, > Does anyone know which dosFS function can be used to partition a disk? > Thank you. > Tri. > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: UDP broadcast down multiple ENDs Date: Wed, 2 Oct 2002 19:01:09 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: "Kevin Garvey" wrote in message news:anf31h$n91$1@kermit.esat.net... > Hi, > > Does anybody know how to configure the vxWorks IP stack such that a UDP > broadcast (i.e. sendto() - destination 255.255.255.255) will be sent on all > configured ENDs? > You can't, not without modifying the stack. Besides, most other stacks I know will ever duplicate a packet in this way. It can go over this interface or the other one, but never ever on both. This is rooted deep in the way that IP defines a broadcast. Now Multicast is a completely different ball game. HTH --------------------------- Newsgroups: comp.os.vxworks Subject: Re: bringingup vxworks on a board with two 750cxe processors Date: 2 Oct 2002 10:39:51 -0700 From: kiran.mandava@skystream.com (kiran Mandava) Organization: http://groups.google.com/ Message-ID: References: <3D9AC536.8335091C@hotmail.com> Hi michael, First approch: i currently think that is the only way..but looking for inputs second approch: in the ISR MMU disabled. 750 interrupt bot the address and data translation is disabled.! kiran >"Michael Lawnick" wrote in message news:$1@snoopy.bndlg.de>... > Hi kiran, > > I'm pretty sure you will have to replace the code at 0x500 by a piece of > your own, where you detect the CPU and branch either to lower or higher CPU- > image. > Another approach: > For the higher CPU redefine the MMU-entry of the lowest page in exception > mode and put an address-translation into it . > > Regards, > Michael > > "kiran Mandava" schrieb im Newsbeitrag > news:ea368e49.0210020608.14976189@posting.google.com... > > Yes each of My CPU have their own images and they are located at > > different places in the memory. > > > > lets sayfirst CPU 0x00000000 to 0x04000000 and the sencond one from > > 0x04000000 to 0x08000000.(image loading starts at an offset 0x20000) > > > > ok when any of the processors get an external interrupt they jump to > > 0x500. assume that second CPU didn't put any code at 0x500, but in the > > first CPU's iSR i would jump to different location depending on the > > CPU number. the first thing it does in the code at 0x500 is it tries > > to save the context at a location interrupt_stack_base(this is in > > reference to a variable vxIntStackBase or something), as the code is > > put in by the first CPU the address belongs to the first CPU, (it > > doesn't use the r1 to store the context on stack, so that the > > interrupt stack would be totally different from the current r1), so if > > there is an interrupt at the same time there is a problem here.! > > > > > > intEnt does it enable the Real mode. i wasn't sure, if that is the > > first thing it does probably that might help me solving this probem.! > > > > > > thanks, > > kiran > > > > >Varndell Engineering wrote in message > news:<3D9AC536.8335091C@hotmail.com>... > > > kiran Mandava wrote: > > > > > > > > Hello all, > > > > > > > > Since the interrupt vector base is not relocatable on powrePc 750 > > > > how should i handle the interrupts on both the processors. if there is > > > > an external interrupt on any of the processors it would jump to > > > > location 0x500 where the ISR corresponding to the First CPU is > > > > located, so there is a problem handling the interrupts from both the > > > > processors as both of them try to save the interrupt conetext to the > > > > same location..did anybody came across such problem and any sort of > > > > feedback would be really helpful > > > > > > > > CPC 710 is the bridging device and both the processors use the same > > > > SDRAM first half the first CPU and the otherhalf the other CPU > > > > > > Assuming that each cpu is executing an image that has been linked to > > > a different address, or each image executes within its own virtual > > > space, I think you can do this. > > > When the 750 takes an interrupt, it will be executing in real address > > > mode for a short time. It will call intEnt() to restore MMU translation, > > > if translation was enabled when the interrupt occurred, save context > > > on the active stack, and return from intEnt to call whatever C function > > > was installed by excConnect(). Right now, you have 2 images, each > > > installing a C handler (different addresses) into the same physical > > > address. > > > Without working out all the details for you, I think you can install > > > a common handler into the vector. This handler would need to decide > > > which processor was executing it and call the interrupt handler for > > > that processor. Obviously, this would require some hacking of the > > > interrupt controller code -- the code that currently takes over > > > the external interrupt vector. Given that, when the user handler > > > is called (after return from intEnt()) virtual address translation > > > is active, this might even be easy. > > > The predictable disclaimer; I've never tried this before and I > > > really haven't thought it out thoroughly. > > > > > > Regards, > > > George Varndell > > > -- > > > Varndell Engineering, LLC Embedded Systems Expertise > > > Need an vxWorks END driver for Sis900 or DP83815? Contact us. > > > http://www.varndellengineering.com > > > > > > > > > -----------== Posted via Newsfeed.Com - Uncensored Usenet News > ==---------- > > > http://www.newsfeed.com The #1 Newsgroup Service in the World! > > > -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers > =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Can't find boot file Date: 2 Oct 2002 12:43:44 -0500 From: Steve Hills Organization: DMS Bliss Corporation Message-ID: **** Post for FREE via your newsreader at post.usenet.com **** Some background: I have 2 Motorola MVME172-P644SE SBCs running VxWorks. There is a transition module installed to provide serial, printer and Ethernet ports. Processor 1 boots from a file on a disk drive on a nearby PC via FTP over the backplane Ethernet. Processor 2 uses processor 1 as a gateway to load the same file over the network. Both processors have different entry points in the code. The following are from the boot traces from both processors. Processor 1 Settings: boot device : ei unit number : 0 processor number : 0 file name : \boot1\vxWorks inet on ethernet (e) : 10.1.50.23:ffff0000 Address of this SBC on the "outside" network inet on backplane (b): 10.0.0.1:ffff0000 Address of this SBC on the "backplane" network host inet (h) : 10.1.50.24 Address of the PC on the "outside" network user (u) : motorola ftp password (pw) : flags (f) : 0x8 startup script (s) : \boot1\hgc1 Processor 2 Settings: boot device : sm=0x02000600 processor number : 1 file name : \boot2\vxWorks inet on backplane (b): 10.0.0.2:ffff0000 Address of this SBC on the "backplane" network host inet (h) : 10.1.50.24 Address of the PC on the "outside" network gateway inet (g) : 10.0.0.1 Address of the first SBC on the "backplane" network user (u) : motorola ftp password (pw) : flags (f) : 0x8 startup script (s) : \boot2\hgc2 My problem is the processor 2 "Can't find the boot file". I have done a ROUTE ADD 126.0.0.0 MASK 255.255.255.0 10.1.50.23 to add the first processor to the PC routing table. I had this problem once before and adding the route fixed it, but not this time. HELP! Thanks, Steve Hills - -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= *** Usenet.com - The #1 Usenet Newsgroup Service on The Planet! *** http://www.usenet.com Unlimited Download - 19 Seperate Servers - 90,000 groups - Uncensored - -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= --------------------------- Newsgroups: comp.os.vxworks Subject: Can't find boot file Date: 2 Oct 2002 12:46:45 -0500 From: Steve Hills Organization: DMS Bliss Corporation Message-ID: **** Post for FREE via your newsreader at post.usenet.com **** Some background: I have 2 Motorola MVME172-P644SE SBCs running VxWorks. There is a transition module installed to provide serial, printer and Ethernet ports. Processor 1 boots from a file on a disk drive on a nearby PC via FTP over the backplane Ethernet. Processor 2 uses processor 1 as a gateway to load the same file over the network. Both processors have different entry points in the code. The following are from the boot traces from both processors. Processor 1 Settings: boot device : ei unit number : 0 processor number : 0 file name : \boot1\vxWorks inet on ethernet (e) : 10.1.50.23:ffff0000 Address of this SBC on the "outside" network inet on backplane (b): 10.0.0.1:ffff0000 Address of this SBC on the "backplane" network host inet (h) : 10.1.50.24 Address of the PC on the "outside" network user (u) : motorola ftp password (pw) : flags (f) : 0x8 startup script (s) : \boot1\hgc1 Processor 2 Settings: boot device : sm=0x02000600 processor number : 1 file name : \boot2\vxWorks inet on backplane (b): 10.0.0.2:ffff0000 Address of this SBC on the "backplane" network host inet (h) : 10.1.50.24 Address of the PC on the "outside" network gateway inet (g) : 10.0.0.1 Address of the first SBC on the "backplane" network user (u) : motorola ftp password (pw) : flags (f) : 0x8 startup script (s) : \boot2\hgc2 My problem is the processor 2 "Can't find the boot file". I have done a ROUTE ADD 126.0.0.0 MASK 255.255.255.0 10.1.50.23 to add the first processor to the PC routing table. I had this problem once before and adding the route fixed it, but not this time. HELP! Thanks, Steve Hills - -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= *** Usenet.com - The #1 Usenet Newsgroup Service on The Planet! *** http://www.usenet.com Unlimited Download - 19 Seperate Servers - 90,000 groups - Uncensored - -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Interfacing NTDS board to vxworks Date: Wed, 02 Oct 2002 13:15:10 -0500 From: Andrew Johnson Organization: APS, Argonne National Laboratory Message-ID: References: <190f55f8.0210010953.1e1a56f3@posting.google.com> Eric Morgan wrote: > I am trying to interface an ANTARES NTDS board to a Motorola 167 > board. I have a compiled test program from the company that I load > and run, but the program spits out that there is no board at that > address. I have the board address set up correctly, and the software > knows that address. What do I have to do in vxworks to get it to > allow the two boards to talk to each other? Is there an application > note or tech note that explains everything you have to do in vxworks > to allow boards to talk to each other? Does the MMU have anything to > do with it? To answer your last question first, that depends whereabouts the board is actually placed. If it's in A32 space then this is almost certainly an MMU issue. If you are building an application-specific version of vxWorks then the simple way to fix this is to add an entry to the sysPhysMemDesc[] table for the board's area of memory, and make sure that the VMEchip2 A32 master window encloses that address range. However be careful about the size of the sysPhysMemDesc[] entry - unless it has a small memory footprint don't make the entry significantly larger than it needs to be because the page table entries use up memory and take time to be created at boot time. If you need access to a large section of the A32 space you can set up the TTRegs to give you that without setting sysPhysMemDesc[]. Our mv167 BSP has the following code in sysHwInit2(): /* Set DTT1 to map virtual memory at 0x80000000 - 0xffffffff direct * to physical memory, marking it noncachable & serialized. */ __asm__("movec %0,dtt1" : /* no outputs */ : "d" (0x807fe040)); - - Andrew - -- "Life is what happens while you're busy making other plans." - John Lennon --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Wed, 02 Oct 2002 13:27:31 -0500 From: Andrew Johnson Organization: APS, Argonne National Laboratory Message-ID: References: brian wrote: > > Im presently using vxWorks for a product and would like to port it to > another royalty free OS as a cost cutting measure. The chip that Im > using has very few other supporting kernels, all of which cost a lot, > either in royalties or up fron costs. You haven't told us what kind of CPU you're using, so it's kind of hard to give specific suggestions. > Any ideas on how to approach this? One idea I have is to buy/get the > source of a generic kernel and imtegrate it into my processor. Is that > as large a job as I think it is? Anyone suggest a good source-free > kernel out there I can buy (dont want Linux..too big) or download for > free? To add to other people's suggestions, have a look at RTEMS http://www.rtems.com/ for another good Open Source OS. If your CPU doesn't have an MMU, then ucLinux would be worth looking at too, it's much smaller than standard Linux. - - Andrew - -- "Life is what happens while you're busy making other plans." - John Lennon --------------------------- Newsgroups: comp.os.vxworks Subject: Serial communication example code Date: Wed, 2 Oct 2002 14:42:13 -0400 From: "Jin-Woo Lee" Organization: Cornell University Message-ID: Sender: jl206@cornell.invalid (on 128.253.249.182) Hello, I'm a beginner of VxWorks. Could anyone send me a sample code of the serial communication (transmit / receive) example code or the direction for it? The VxWorks manuals are too big to read in a short time. My current system is Motorola MV2400, PowerPC and the development environment is Torndao 2.0. Thanks in advance! Jinwoo --------------------------- Newsgroups: comp.os.vxworks Subject: AE, user ISR, how in PD? Plus, is anybody using AE? Date: 2 Oct 2002 12:02:53 -0700 From: danderson@kns.com (Don) Organization: http://groups.google.com/ Message-ID: I've gotten some very useful information from John Gordon [Thanks, John], but always can use more, especially since it appears nobody is using AE! :) Info: We have a user routine attached to an ISR in our 5.4 system, specifically the AuxClock to do precision timing and sampling. It updates "global" variables and, if condition met, gives a semaphore to a "timer" task for additional processing. I believe under AE, ISRs can only be in kernel PD, so does anybody have information on what they did for issues like this? Issues include handling the global data -- using a shared data PD, but what about performance, wanting to avoid adding much user code to kernel to avoid "crashing" it, etc. [some suggestions from John deleted] Thanks. Although it's pretty time-critical, I think your suggestion to use access functions from another PD might be OK. Like everybody, we are trying to do things that probably are mutually exclusive: minimize additions to the kernel, not affect application timing, decouple as much code and data as possible, etc., and get it done quickly! ;) More info: we plan to break up our application into two PDs, one for the core code and one for the material handling system code. Both will need access to the services of the previously mentioned user timer task [and others like it] that is the one given the semaphore out of the ISR. I see two general approaches: do this/these kind of tasks [e.g., spooler task, event tasks] in the kernel PD [easier, but increases risk of crashing kernel] or have a separate PD of all of these tasks so that multiple other PDs can share [either some form of IPC or by shared data PD since it won't be kernel PD], which is more complex but reduces risk to kernel. Any suggestions again appreciated. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Adding PCI support in Vx. Date: Wed, 2 Oct 2002 19:38:46 +0000 (UTC) From: "Graham Baxter" Organization: Graham Baxter (Software) Limited Message-ID: References: <3D9AF1F1.4020508@improvsys.com> On my system (CompactPCI, MPC7410, Galileo), the system controller (VxWorks) is responsible for assigning addresses to the BAR registers for all found devices. It scans the bus (via configuration cycles), locates all connected devices, and probes each found device's functions for memory requirements. The memory requirements are then sorted to descending size sequence. The BAR registers are then programmed to be within the assigned memory space and taking care of any required alignment. There may be an example in your Wind River BSP. I hope this helps. - -- Regards, Graham Baxter Freelance Software Engineer VxWorks and pSOS BSPs (available shortly for new assignment) gbaxter@NOSPAMbcs.org.uk "Jeremiah Frick" wrote in message news:3D9AF1F1.4020508@improvsys.com... > Hello Everyone, > > I'm attempting to add PCI support in VxWorks on the ARM Integrator 940t > for a card I have. When I initially set the system up, I could detect > the PCI card and read the 9050 configuration registers (vendor ID, > device ID, etc) without any problem. However, the BAR addresses weren't > being set up at all. So, I did some looking around and added some code > to sysLib.c in the BSP: > > { > UINT32 pciBus; /* PCI Bus number */ > UINT32 pciDevice; /* PCI Device number */ > UINT32 pciFunc; /* PCI Function number */ > if( pciFindDevice( VENDOR_ID, DEVICE_ID, 0, &pciBus, &pciDevice, > &pciFunc ) == OK ) { > pciAssignResources( pciBus, pciDevice, pciFunc ); > } > } > > This was added inside the SysHwInit function. Using this software I now > get at least a BAR2 address (this is the only one I checked initially). > However, the address given isn't consistant with PCI memory space. In > my system the BAR2 address I'm getting is 0x01000000, and the PCI space > should be 0x40000000 at the least. I was wondering if the value > returned is simply an offset into the PCI base memory space or if I'm > still experiencing problems and need to add more software. > > Any assistance would be greatly appreciated. > > Sincerely, > Jeremy > --------------------------- Newsgroups: comp.os.vxworks Subject: Pci interrupts in a Pentium system. Date: 2 Oct 2002 13:55:21 -0700 From: sfinley@xes-inc.com (Scott Finley) Organization: http://groups.google.com/ Message-ID: I did a search of this newsgroup and found several posts on this subject, but non seemed to fix my issue. I am trying to service interrupts on INT A (IRQ 8) in my pentium system. I can generate interrupts (I have a card with an LED so I can see INT A being driven) and connect my handler (calling pciInt(8) successfully enters my isr) but the interrupt does not seem to be properly enabled. Here is the code I use to set things up: pciConfigOutByte(0,3,0,0x3c,8); pciIntConnect(INUM_TO_IVEC(8+0x20), ctHandler, ctPtr); sysIntEnablePIC(8); You probably noticed that I wrote the int Line configuration register on my card first. The reason I need to do this may be at the root of my problem. After I boot, the int Line pci config register has not been written (it is 0x00). I'm not sure how big a problem this is. Of the 5 PCI devices I can see, only the ethernet controller has its int line register set (it is set to 0x0B). Is my BIOS screwing me and preventing this card from generation interrupts? Clearly writing this register manually and then enabling like normal isn't doing the trick. Anyone have any ideas? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: newbie - device drivers in c++ Date: Wed, 02 Oct 2002 21:14:51 +0000 From: molochai Organization: Griffin Internet - http://www.griffin.net.uk Message-ID: References: <290f4d3c.0210012203.38c6685a@posting.google.com> Rao R R wrote: > Hi > Does vxworks supports c++ object handling at kernel level? > I can't forsee any particular problems provided that C++ runtime support is included in the vxworks iamge. Clearly static wrapper methods will be needed inbetween methods and places where a function with C linkage can be bound (interrupt vectors, device entry points etc.) Extra care will be needed to avoid generating exceptions in inapproriate places (for instance throwing an exception in a character-driver entry point such as open() would be bad form as it may be called from C). ISRs shouldn't present any extra problems to be honest, but be careful of compiler code synthesis which might well generate enexpected constructions for instance. Avoid IOSTREAMS completely - if you start your driver in hwInit2() there might be some nasty surprises with initialisation order and static initialisers. > some one tried writing device drivers in vxworks in c++. Can someone give > insight into this ? What would be the adv and disadvanteges of writing > device drivers in c++. Thye obvious advantage being a great deal of the OS and driver layer is inherrently OO. However, this is implemented the hard way in C, and the way OO features are implemented won't be simplified by using C++. - -- The muppetry is strong in this one. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: newbie - device drivers in c++ Date: Wed, 2 Oct 2002 22:41:15 +0100 From: Wade Oram Organization: None Message-ID: References: <290f4d3c.0210012203.38c6685a@posting.google.com> Reply-To: Wade Oram In message , molochai writes >Rao R R wrote: > >> Hi >> Does vxworks supports c++ object handling at kernel level? >> > >I can't forsee any particular problems provided that C++ runtime support is >included in the vxworks iamge. > >Clearly static wrapper methods will be needed inbetween methods and places >where a function with C linkage can be bound (interrupt vectors, device >entry points etc.) > >Extra care will be needed to avoid generating exceptions in inapproriate >places (for instance throwing an exception in a character-driver entry >point such as open() would be bad form as it may be called from C). ISRs >shouldn't present any extra problems to be honest, but be careful of >compiler code synthesis which might well generate enexpected constructions >for instance. I am involved in a project on vxWorks which is nearly all C++ code. The application involves custom device handling ( not device drivers as such as they don't hook into the vxWorks IO system) including interrupts. All done with C++. If you are compiling with C++ exceptions enabled, then I believe that you will have to avoid creating and destroying objects on the stack from within any interrupt service routines. I thought that I could use a simple class like: class InterruptLock { public: InterruptLock() {lock = ::intLock();} ~InterruptLock() {::intUnlock(lock);} private: int lock; // check that this is the right type to hold // the return value from ::intLock() }; // example usage: // enter critical section { InterruptLock locker; // constructor locks interrupts ... do critical stuff here! // interrupt lock gets released here by InterruptLock destructor } to do my lock handling from within the interrupt routine as I do in places outside of the interrupt context. The result was a quick and very final protection fault! We live and learn! Calling the ::intLock and ::intUnlock routines directly gave no problems. There are no problems using previously constructed objects from within the interrupt context. I can't see any problems creating POD structures on the stack from within the interrupt context although I have never tried it. If you have disabled C++ exceptions, then I can't see any reason why you should not be able to use stack objects like 'InterruptLock' from within the interrupt context but, again, I haven't tried it. - -- Wade Oram --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: 2 Oct 2002 14:56:13 -0700 From: brian_doe@hotmail.com (brian) Organization: http://groups.google.com/ Message-ID: References: > > You haven't told us what kind of CPU you're using, so it's kind of hard to > give specific suggestions. Yes. Thanx to all for their suggestions.. even Leonid above who seems pretty obstinate..but I guess he must be pretty grumpy in his old age, as he calls me "son" :) The processor is the Tensilica Xtensa processor. www.tensilica.com. Its a 32bit, no mmu, no fp RISC processor. Again, Thanx for the many suggestions. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: 2 Oct 2002 15:12:58 -0700 From: brian_doe@hotmail.com (brian) Organization: http://groups.google.com/ Message-ID: References: <3a5ff53f.0210020748.3643d45a@posting.google.com> "Leonid Rosenboim" wrote in message news:... > Now, don't take me wrong here, I never said there eint > a reason to throw out WindRiver, but cost eint that reason, son! > > I know that HP has replaced VxWorks with something else > in their printers, but can you really prove that cost was the > reason ? Actually I dont know for sure. But I can find out. Do you know for sure that it wasnt the reason? > > You know there are a dozen of other reasons that could > be the cause: like lack of support, too long to fix bugs, > no source code availability, ... > > I mean, if you're pissed at WRS - you're not alone! > Go ahead, throw them out, join the masses, > just don't tell me it's for cost! Well dad, Actually, Im not a bit pissed at WRS. In fact normally I take the same side of the equation that you do. Cost savings due to time to market etc etc. And in fact that argument is usually winnable. And in fact I would recommend vxWorks for new development any day. And do it all the time. However in this case, the product is built and working. The question is whether I can run a parallel development of swapping out the kernel. The cost? Suppose I wanna ship 4 million units. Suppose that the best that I can get is 25c /unit from WRS. Suppose that the total royalty cost in therefore $1Mill. Now. If I can do this in 3-6mos, it costs me 100k. Are there maintenance costs? In fact very little if the code is not buggy. If I can use a kernel that is already debugged (just not ported) then I'm way ahead of the game on the bugs. Personally I think 1 month of serious testing would eliminate nearly every bug. Am I going to port to another processor after this if I need to get more mips etc etc later? Probably not. But if I do I can always go back to WRS and begin again. Aside from calling me 'son' and disparaging the whole idea .. do you have any suggestions or can you point out something specific that I have missed? Please do as I am looking for holes in this argument. Brian > > "Brian" wrote in message > news:3a5ff53f.0210020748.3643d45a@posting.google.com... > > "Leonid Rosenboim" wrote in message > news:... > > > Are you serious ? > > > > > > To replace the OS as a cost cutting measure is one peice of BS > > > I have heard too many times, and allways proved wrong. > > > > > > > HP just did it with their some of their printers. Have you proved > > them wrong yet? It's very possible that WRS just wont negotiate down > > since the technical risk is deemed high. --------------------------- Newsgroups: comp.os.vxworks Subject: can't load .out files with dosfs2 Date: 2 Oct 2002 19:22:00 -0700 From: term@xerosoft.com (Tom) Organization: http://groups.google.com/ Message-ID: <15cdb0fb.0210021822.572bd28c@posting.google.com> This is a repost of a previous message I had posted with new information included. I have been working on getting a strong arm based motherboard to run vxworks. One problem I have run into is that I can't load any .out files off of my TFFS (True Flash File System). When I try to it just gives me an error saying "error reading file (errno = 0x3d0001).". I looked up the errno and it resolved to S_objLib_OBJ_ID_ERROR , but this is generated from something at a low level that I don't have the source code to. I recently tried stepping through the assembler code to see where this error was coming from and it turns out errno is not set until the cleanup code runs after fioRead fails, so this is errno is irrelevant. Every other operation with the TFFS seems to work just fine. I can create, delete, read, write, and rename files. As an example I tried copying an out file off my FTP server onto flash and loading it, but it failed. Then after that I copied the file off of the TFFS to a new file name on my FTP server and then loaded it from that file (over the network) and it worked fine. - -- start dump -- - -> cd "/tffs0/" value = 0 = 0x0 - -> ll - -rwxrwxrwx 1 0 0 10064 Jan 1 1980 test.out value = 0 = 0x0 - -> ld < test.out ld error: error reading file (errno = 0x3d0001). value = 0 = 0x0 - -> copy "/tffs0/test.out", "host:newtest.out" Copy OK: 10064 bytes copied value = 0 = 0x0 - -> ld < host:newtest.out value = 12531032 = 0xbf3558 - -> - -- end dump -- After these tests I thought the problem might have been size related, so I tried to create a minimal "hello world" out file, but that didn't work either. As a work around I thought I would try setting up a runscript that creates a ram drive and then copies each file to the ramdrive, loads from there, deletes the file, and then continues on to the next one until it is done. Unfortunatelly this failed with the same error message that I got when I loaded off of a TFFS. The only thing I can think of that the ramdrive and TFFS have in common that a ftp file system doesn't is that they use dosFS (version 2). Has anyone else had any problems like this? I'm running VxWorks 5.4.2 with Tornado 2.1.1. Best Regards, Tom --------------------------- Newsgroups: comp.os.vxworks Subject: help with objcopyppc Date: 2 Oct 2002 20:42:37 -0700 From: andreas_lindell75@hotmail.com (Rikard) Organization: http://groups.google.com/ Message-ID: <5947a04.0210021942.5f8cbc94@posting.google.com> Hello, I have a file(test.out) that I want to convert into a .s file using the objcopyppc utility provided by WR. However, I am not very good at these things so I have problems with all the switches and flags that needs to be set. I have tried the following: objcopyppc -S test.out test.s But then the size of file test.s is only 1KB (test.out is 12 kb) so that is not the correct settings. Does anyone know what switches I need to set to convert a .out file to a .s file using the objcopyppc? Regards Andreas --------------------------- Newsgroups: comp.os.vxworks Subject: netjobadd question Date: 2 Oct 2002 20:55:54 -0700 From: taissimon@yahoo.com.tw (simon tsai) Organization: http://groups.google.com/ Message-ID: <25a2f1f.0210021955.50c20f2f@posting.google.com> i am porting END driver, when the netjobadd() called in my ISR routine system crash and no any error message , how do i solve this problem --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Mixed C & Assembly listing Date: Thu, 3 Oct 2002 08:16:46 +0300 From: "Martin Roth" Organization: Motorola Message-ID: References: Bill, Thanks, you suggestion works on Tornado 2.02. Martin "Bill Pringlemeir" wrote in message news:m2smzohq7y.fsf@sympatico.ca... > > Martin> I have tried objdumpppc -l --source --debugging file.o > > Martin> I am getting the assembly listing, but I do not see the C > Martin> source code. What I see is the line number of the C source > Martin> code, but without the line itself. > > If you are doing this on a Windows platform and compiling with the > Tornado projects, it appears that the tools put weird paths in the > object files. With the ARM and COFF files, I have used a hex editor > to zap the file name string in the binary. The "--source" option only > works if the source file can be found. If you were to recompile the > file in the same directory with the same options, it should find it. > > Ie, change > > "ccppc {compile options} -c path/to/source/file.c -o path/to/obj/file.o" > > to > > "ccppc {compile options} -c file.c -o file.o" > > After copying the source "file.c" to the directory where the Makefile > is run. > > There still may be other problems, but this might help. > > hth, > Bill Pringlemeir. > > -- > ..I must be a VETERINARIAN.. --------------------------- Newsgroups: comp.realtime,comp.os.vxworks Subject: Re: Problem with vxworks Date: 2 Oct 2002 23:38:20 -0700 From: yathish_g@yahoo.com Organization: http://groups.google.com/ Message-ID: <9f000cfd.0210022238.6e71211a@posting.google.com> References: <9f000cfd.0209302147.1ff4c00a@posting.google.com> <4a885870.0210011946.4271985d@posting.google.com> Hi will, Thanks for the reply. In the first mail, I have already mentioned that I am able to ping from the host to the vxworks box. I am able to ping successfully without any packet loss. Both the host and the target are able to sniff each other as far as the packet exchange is concerned. Both the target and the host are forced to do at 10 Mbps. regds, Yathish wv9557@yahoo.com (Will) wrote in message news:<4a885870.0210011946.4271985d@posting.google.com>... > The log only shows one side of the story. > Not seeing the whole packets exchange, it's hard to tell. You might > want to get something like tcpdump for Windows (yah - > tcpdump does run on Windows :) > and look at every packet that is exchanged between the 2 boxes. > > Just a shoot in the dark: after you swapped the cards, wait long > enough for the PC's ARP cache to timeout. > > > Just a shoot in the dark number 2: > Force both the vxWorks box and the PC to do 10mbits, by using > a good old 10mbit-only switch. > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: question: partition a disk using dosFS Date: Thu, 3 Oct 2002 09:19:06 +0200 From: "Khac Tri VU" Organization: Planet Internet Message-ID: References: dosFS 2.0 "Leonid Rosenboim" wrote in message news:newscache$6y5d3h$u84$1@lnews.actcom.co.il... > Which dosFs version ? > > "Khac Tri VU" wrote in message > news:anf1k5$fbp$1@reader13.wxs.nl... > > Hi folks, > > Does anyone know which dosFS function can be used to partition a disk? > > Thank you. > > Tri. > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Thu, 3 Oct 2002 09:43:46 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <3a5ff53f.0210020748.3643d45a@posting.google.com> Brian, You're doing the math, which is good. Are you sure WRS wont go below 25 cents per unit if you COMMIT to a 4 Million volume ? Also, look at it from a cash-flow point of view, look at how many unit you will sell in one year, and compare that to the cost of development. If in the end the math tells you that you save money, go ahead, but before you do, show that math to your WRS sales guy, explain him this math slowly, and make sure they understand, thereby giving them their last chanse. If they dont miss that chanse they are indeed dumb. Back at the time when I sold VxWorks, I used a different argument, comparing the royalty price to some of the other system components (connectors, packaging etc.) In other words, saving 25 cents of your COGS - is this really the way you would like to spend your resources ? Anyway, good luck in what you;re doing. Over and out. "brian" wrote in message news:d2f67bd8.0210021412.92afe08@posting.google.com... > "Leonid Rosenboim" wrote in message news:... > > Now, don't take me wrong here, I never said there eint > > a reason to throw out WindRiver, but cost eint that reason, son! > > > > I know that HP has replaced VxWorks with something else > > in their printers, but can you really prove that cost was the > > reason ? > > Actually I dont know for sure. But I can find out. Do you know for > sure that it wasnt the reason? > > > > > You know there are a dozen of other reasons that could > > be the cause: like lack of support, too long to fix bugs, > > no source code availability, ... > > > > I mean, if you're pissed at WRS - you're not alone! > > Go ahead, throw them out, join the masses, > > just don't tell me it's for cost! > > > Well dad, > > Actually, Im not a bit pissed at WRS. In fact normally I take the same > side of the equation that you do. Cost savings due to time to market > etc etc. And in fact that argument is usually winnable. And in fact > I would recommend vxWorks for new development any day. And do it all > the time. > > However in this case, the product is built and working. The question > is whether I can run a parallel development of swapping out the > kernel. > > The cost? Suppose I wanna ship 4 million units. Suppose that the best > that I can get is 25c /unit from WRS. Suppose that the total royalty > cost in therefore $1Mill. > > Now. If I can do this in 3-6mos, it costs me 100k. Are there > maintenance costs? In fact very little if the code is not buggy. If I > can use a kernel that is already debugged (just not ported) then I'm > way ahead of the game on the bugs. Personally I think 1 month of > serious testing would eliminate nearly every bug. Am I going to > port to another processor after this if I need to get more mips etc > etc later? Probably not. But if I do I can always go back to WRS and > begin again. > > Aside from calling me 'son' and disparaging the whole idea .. do you > have any suggestions or can you point out something specific that I > have missed? Please do as I am looking for holes in this argument. > > Brian > > > > > > > > > "Brian" wrote in message > > news:3a5ff53f.0210020748.3643d45a@posting.google.com... > > > "Leonid Rosenboim" wrote in message > > news:... > > > > Are you serious ? > > > > > > > > To replace the OS as a cost cutting measure is one peice of BS > > > > I have heard too many times, and allways proved wrong. > > > > > > > > > > HP just did it with their some of their printers. Have you proved > > > them wrong yet? It's very possible that WRS just wont negotiate down > > > since the technical risk is deemed high. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: question: partition a disk using dosFS Date: Thu, 3 Oct 2002 09:47:31 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: To directly answer yoru question, partitions are created using usrFdiskPartCreate() then every partition must be initialize with dosFsVolFormat(). Also, I would strongly recommend you read the dosFs 2.0 documentation. "Khac Tri VU" wrote in message news:angr57$1lo$1@reader13.wxs.nl... > dosFS 2.0 > > > "Leonid Rosenboim" wrote in message > news:newscache$6y5d3h$u84$1@lnews.actcom.co.il... > > Which dosFs version ? > > > > "Khac Tri VU" wrote in message > > news:anf1k5$fbp$1@reader13.wxs.nl... > > > Hi folks, > > > Does anyone know which dosFS function can be used to partition a disk? > > > Thank you. > > > Tri. > > > > > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Pci interrupts in a Pentium system. Date: Thu, 3 Oct 2002 10:09:44 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1033632622.694125@newsreader1.wirehub.nl> References: if you use pciHeaderShow in the shell on you device, does it say INT8? Cheers, Renee - - - - - - - - - Renee A. Teunissen PTS Software bv, Meerweg 7, 1405BA Bussum, NL. T.+31-(0)35-6926969, M.+31-(0)6-22778313, http://www.pts.nl, @pts.nl personal link page: http://wittenburg10c.nl/db/dest/links.html "Scott Finley" wrote in message news:e2993ff2.0210021255.10c7e3eb@posting.google.com... > I did a search of this newsgroup and found several posts on this > subject, but non seemed to fix my issue. > > I am trying to service interrupts on INT A (IRQ 8) in my pentium > system. I can generate interrupts (I have a card with an LED so I can > see INT A being driven) and connect my handler (calling pciInt(8) > successfully enters my isr) but the interrupt does not seem to be > properly enabled. > > Here is the code I use to set things up: > pciConfigOutByte(0,3,0,0x3c,8); > pciIntConnect(INUM_TO_IVEC(8+0x20), ctHandler, ctPtr); > sysIntEnablePIC(8); > > You probably noticed that I wrote the int Line configuration register > on my card first. The reason I need to do this may be at the root of > my problem. After I boot, the int Line pci config register has not > been written (it is 0x00). I'm not sure how big a problem this is. Of > the 5 PCI devices I can see, only the ethernet controller has its int > line register set (it is set to 0x0B). Is my BIOS screwing me and > preventing this card from generation interrupts? Clearly writing this > register manually and then enabling like normal isn't doing the trick. > Anyone have any ideas? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Pci interrupts in a Pentium system. Date: Thu, 3 Oct 2002 10:09:44 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1033633843.758370@newsreader1.wirehub.nl> References: if you use pciHeaderShow in the shell on you device, does it say INT8? Cheers, Renee - - - - - - - - - Renee A. Teunissen PTS Software bv, Meerweg 7, 1405BA Bussum, NL. T.+31-(0)35-6926969, M.+31-(0)6-22778313, http://www.pts.nl, @pts.nl personal link page: http://wittenburg10c.nl/db/dest/links.html "Scott Finley" wrote in message news:e2993ff2.0210021255.10c7e3eb@posting.google.com... > I did a search of this newsgroup and found several posts on this > subject, but non seemed to fix my issue. > > I am trying to service interrupts on INT A (IRQ 8) in my pentium > system. I can generate interrupts (I have a card with an LED so I can > see INT A being driven) and connect my handler (calling pciInt(8) > successfully enters my isr) but the interrupt does not seem to be > properly enabled. > > Here is the code I use to set things up: > pciConfigOutByte(0,3,0,0x3c,8); > pciIntConnect(INUM_TO_IVEC(8+0x20), ctHandler, ctPtr); > sysIntEnablePIC(8); > > You probably noticed that I wrote the int Line configuration register > on my card first. The reason I need to do this may be at the root of > my problem. After I boot, the int Line pci config register has not > been written (it is 0x00). I'm not sure how big a problem this is. Of > the 5 PCI devices I can see, only the ethernet controller has its int > line register set (it is set to 0x0B). Is my BIOS screwing me and > preventing this card from generation interrupts? Clearly writing this > register manually and then enabling like normal isn't doing the trick. > Anyone have any ideas? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: question: partition a disk using dosFS Date: Thu, 3 Oct 2002 10:45:49 +0200 From: "Khac Tri VU" Organization: Planet Internet Message-ID: References: Thank you, Leonid. Can I fdisk from winshell like I used to run diskFormat()? I think that in the document of DosFS, usrFdiskPartLib documents only 2 functions: usrFdiskPartRead() and usrFdiskPartShow(). Do I miss something? Regards, Tri. "Leonid Rosenboim" wrote in message news:newscache$24be3h$z96$1@lnews.actcom.co.il... > > To directly answer yoru question, partitions are created using > usrFdiskPartCreate() > then every partition must be initialize with dosFsVolFormat(). > > Also, I would strongly recommend you read the dosFs 2.0 > documentation. > > "Khac Tri VU" wrote in message > news:angr57$1lo$1@reader13.wxs.nl... > > dosFS 2.0 > > > > > > "Leonid Rosenboim" wrote in message > > news:newscache$6y5d3h$u84$1@lnews.actcom.co.il... > > > Which dosFs version ? > > > > > > "Khac Tri VU" wrote in message > > > news:anf1k5$fbp$1@reader13.wxs.nl... > > > > Hi folks, > > > > Does anyone know which dosFS function can be used to partition a disk? > > > > Thank you. > > > > Tri. > > > > > > > > > > > > > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: problems with ldppc Date: 3 Oct 2002 02:59:13 -0700 From: andreas_lindell75@hotmail.com (Rikard) Organization: http://groups.google.com/ Message-ID: <5947a04.0210030159.7eec97fd@posting.google.com> Hello, I am working on a CHAMP AVII board from DY4 and will run VxWorks on proc A and application using the BSL on B, C and D. It are these applications that are causing my problems and I have difficulty linking them. I have created a small application in Tornado that uses a function in the BSL and a printf line. I compile it and get a file test.o that I need to link to the bsl and standard C libraries. I use the command ldppc: ldppc -X -N -o test.out test.o -L$(AVII)/lib I get errors saying: ldppc: warning: cannot find entry symbol _start; defaulting to 00040074 lab1.o: In function `vx_main': lab1.o(.text+0x10): undefined reference to `_bslBoardInit' lab1.o(.text+0x24): undefined reference to `printf' I am new to linking from command line, Tornado usualy does that for me:(, and now I am paying the price. So my questions is as follows. I need to point the linker to the correct directories, so if I have Tornado installed in d:\tornado where and how do I point the linker to solve undef ref to printf. The `_bslBoardInit' is due to me not pointing to the correct directory where the includes for AV2 is stored but I hope that problem will go away once I can point correctly for the printf, then I will know how to point. So if anyone knows how to point the linker please provide me with some help. Regards Andreas --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Pci interrupts in a Pentium system. Date: 3 Oct 2002 03:17:49 -0700 From: google@fareham.org (Roger) Organization: http://groups.google.com/ Message-ID: <63f42115.0210030217.655d7acc@posting.google.com> References: > You probably noticed that I wrote the int Line configuration register > on my card first. The reason I need to do this may be at the root of > my problem. After I boot, the int Line pci config register has not > been written (it is 0x00). I'm not sure how big a problem this is. Of > the 5 PCI devices I can see, only the ethernet controller has its int > line register set (it is set to 0x0B). Is my BIOS screwing me and > preventing this card from generation interrupts? Clearly writing this > register manually and then enabling like normal isn't doing the trick. > Anyone have any ideas? The interrupt register is just a place holder. Your card should generate interrupts on line A, (or A & B if it's got two devices, etc). The BIOS should enumerate the slots, and work out where the interrupt line A,B,C,D are actually routed and fill in the interrupt register value. If the interrupt register is 0 then the BIOS didn't enumerate the interrupt lines. I've not played with this under VxWorks, but under IxWorks it used to mess this up a bit. Try setting your BIOS to non-plug and play OS, then rebooting to DOS. Take a look at the register value, and use that. Roger. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: question: partition a disk using dosFS Date: Thu, 3 Oct 2002 12:19:03 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: usrFdiskPartCreate is equivalent to FDISK, and you can run it fron anywhere, windsh, target shell, program, ... its just a function you know. Regarding documentation, you are right, usrFdiskPartCreate is not documented on the offline supplement manual, - my fault. (should submit this to WRS support for future correction). "Khac Tri VU" wrote in message news:anh086$6lo$1@reader13.wxs.nl... > Thank you, Leonid. Can I fdisk from winshell like I used to run > diskFormat()? > I think that in the document of DosFS, usrFdiskPartLib documents only 2 > functions: usrFdiskPartRead() and usrFdiskPartShow(). Do I miss something? > Regards, > Tri. > > > > > > "Leonid Rosenboim" wrote in message > news:newscache$24be3h$z96$1@lnews.actcom.co.il... > > > > To directly answer yoru question, partitions are created using > > usrFdiskPartCreate() > > then every partition must be initialize with dosFsVolFormat(). > > > > Also, I would strongly recommend you read the dosFs 2.0 > > documentation. > > > > "Khac Tri VU" wrote in message > > news:angr57$1lo$1@reader13.wxs.nl... > > > dosFS 2.0 > > > > > > > > > "Leonid Rosenboim" wrote in message > > > news:newscache$6y5d3h$u84$1@lnews.actcom.co.il... > > > > Which dosFs version ? > > > > > > > > "Khac Tri VU" wrote in message > > > > news:anf1k5$fbp$1@reader13.wxs.nl... > > > > > Hi folks, > > > > > Does anyone know which dosFS function can be used to partition a > disk? > > > > > Thank you. > > > > > Tri. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Missing Header files (*.h) in the "External Dependencies" in Tornado. Date: 3 Oct 2002 05:00:43 -0700 From: shy017@yahoo.com (Shay) Organization: http://groups.google.com/ Message-ID: <8a5dc325.0210030400.19449365@posting.google.com> Hi all, I have a (minor) problem. I work with Tornado 2.0.2 I have several Project in the same workspace. For all the project Tornado got right the "External Dependencies" decision and gave all the .h file. BUT In one project he refuse to do so :-( No matter how may time I click the Dependencies, open/close the project ... My solution is to "Add Files..." to the project. Any better solution ? Thanks, Shay. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Build specification - Link order Date: Thu, 03 Oct 2002 07:32:04 -0400 From: david lindauer Message-ID: <3D9C2AB4.E36323D@notifier-is.net> References: <4ac248a4.0210011107.19484809@posting.google.com> it generally doesn't make a difference. But as one example, in a recent (non-vxworks) project I had a boot code segment that was 32K. We later wished to upgrade it to a boot code segment of 64K by moving some other portions of memory around. However, only the first 32K could be downloaded to versions running the old software. What we did was changed the link order so that the most important code for the functionality we needed during the upgrade was in the first 32K of the boot code; less important code such as self tests and fancy displays was placed in the second 32K. So when we first downloaded the new boot code to equipment running the old version of the program, we were able to cope with the 32K limit and still have enough functionality to upgrade the rest of the system - although certain relatively unimportant functionality wasn't invoked because it couldn't be sent down at that time. As part of the upgrade we eventually downloaded the entire boot image and got our new functionality added in. David Lawrie wrote: > Hi..... > > If someone could help me in knowing the significance of > changing the link order in the build specification it > would be very greatful. > > looking forward for the reply.... > > thanks. > Lawrie. --------------------------- Newsgroups: comp.os.vxworks Subject: Creating Singleton objects at run time and using it in a shared library Date: 3 Oct 2002 06:54:57 -0700 From: steven4443@yahoo.com (stephen hallacy) Organization: http://groups.google.com/ Message-ID: Hello, I want to create an singleton object at bootup time or some time after bootup time, and I want to be able to access that object's methods in a shared library. Is this possible, and if so, how is it done? Any help is appreciated, Thanks, stephen --------------------------- Newsgroups: comp.os.vxworks Subject: MPC860 watchdog help ... real time scheduling problem Date: Thu, 03 Oct 2002 14:16:41 GMT From: "Jim" Organization: Magma Communications Ltd. Message-ID: Hello, I am working on the a routine to service the MPC860 watchdog. I am having problems serviceing the watchdog despite the fact my watchdog service routine is a priority 0. My watch dog routine is essentailly for(;;) { giveTheDogABone(); taskDelay(60); } I am having problems servicing the watch during startup only ... I am now curious if the way I am servicing the watchdog is incorrect. Mostly I am wondering if: 1.) should i use taskDelay(60) ... if i use this i think a low priority task can still hog the processor ... unless there is an ISR or someone blocks. 2.) or should i use a scheduled interrupt to do this servicing ... Cheers, Jim --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Suspended task debugging Date: Thu, 03 Oct 2002 10:20:30 -0400 From: Mary McHugh Organization: Rancho McHugh Message-ID: <3D9C522E.C301D84@notroll.com> References: Reply-To: trollbusters@notroll.com MK wrote: > > I run into a problem, where my system would break down, and when I connect a > debugger to it I notice that one of the tasks is suspended. The place at > which it is suspended seems totally harmless though. Looking through the > posts on this group I found out that apparently VxWorks will suspend an > active task if there is a bus error in the interrupt handler, and there > isn't anything wrong being done by the task... I am in need of suggestions, > how do I debug this problem? > > If we assume that the bug is in the driver code, is there a way to find out > the address of the command which actually caused the bus error? Is the task that's suspended the one you're trying to attach the debugger to? If so, when it attaches to a task there is an option to halt the task on attach. (Tools->Options->Debug Tab, deselect "Always halt after attaching to task"). Mary --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Pci interrupts in a Pentium system. Date: 3 Oct 2002 07:44:01 -0700 From: sfinley@xes-inc.com (Scott Finley) Organization: http://groups.google.com/ Message-ID: References: <1033632622.694125@newsreader1.wirehub.nl> I don't have pciHeaderShow() compiled into my BSP (what do I include in the Tornado 2 cockpit to get this?) but as I said before, manually reading the int line register gives 0x00. "Renee Teunissen" wrote in message news:<1033632622.694125@newsreader1.wirehub.nl>... > if you use pciHeaderShow in the shell on you device, does it say INT8? > > Cheers, > Renee > - - - - - - - - > Renee A. Teunissen > PTS Software bv, Meerweg 7, 1405BA Bussum, NL. > T.+31-(0)35-6926969, M.+31-(0)6-22778313, > http://www.pts.nl, @pts.nl > personal link page: http://wittenburg10c.nl/db/dest/links.html > > "Scott Finley" wrote in message > news:e2993ff2.0210021255.10c7e3eb@posting.google.com... > > I did a search of this newsgroup and found several posts on this > > subject, but non seemed to fix my issue. > > > > I am trying to service interrupts on INT A (IRQ 8) in my pentium > > system. I can generate interrupts (I have a card with an LED so I > can > > see INT A being driven) and connect my handler (calling pciInt(8) > > successfully enters my isr) but the interrupt does not seem to be > > properly enabled. > > > > Here is the code I use to set things up: > > pciConfigOutByte(0,3,0,0x3c,8); > > pciIntConnect(INUM_TO_IVEC(8+0x20), ctHandler, ctPtr); > > sysIntEnablePIC(8); > > > > You probably noticed that I wrote the int Line configuration > register > > on my card first. The reason I need to do this may be at the root of > > my problem. After I boot, the int Line pci config register has not > > been written (it is 0x00). I'm not sure how big a problem this is. > Of > > the 5 PCI devices I can see, only the ethernet controller has its > int > > line register set (it is set to 0x0B). Is my BIOS screwing me and > > preventing this card from generation interrupts? Clearly writing > this > > register manually and then enabling like normal isn't doing the > trick. > > Anyone have any ideas? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Pci interrupts in a Pentium system. Date: 3 Oct 2002 07:48:24 -0700 From: sfinley@xes-inc.com (Scott Finley) Organization: http://groups.google.com/ Message-ID: References: <63f42115.0210030217.655d7acc@posting.google.com> I have an ethernet card in the system that has been enumerated and it's int line register is set to 0x0B. Is it possible that the vxWorks ethernet driver did this, not the BIOS? Why would the BIOS do just one card and not the others? I'm fairly sure my card is on IRQ8 because according to an LED on a card in one of the PCI slots, it is the INT after the ethernet card on IRQ12, so it should wrap back to IRQ8 right? Anyway, I tried the code above on all 4 IRQs (8-12) and it didn't work on any of them... google@fareham.org (Roger) wrote in message news:<63f42115.0210030217.655d7acc@posting.google.com>... > > You probably noticed that I wrote the int Line configuration register > > on my card first. The reason I need to do this may be at the root of > > my problem. After I boot, the int Line pci config register has not > > been written (it is 0x00). I'm not sure how big a problem this is. Of > > the 5 PCI devices I can see, only the ethernet controller has its int > > line register set (it is set to 0x0B). Is my BIOS screwing me and > > preventing this card from generation interrupts? Clearly writing this > > register manually and then enabling like normal isn't doing the trick. > > Anyone have any ideas? > > The interrupt register is just a place holder. Your card should > generate interrupts on line A, (or A & B if it's got two devices, > etc). > The BIOS should enumerate the slots, and work out where the interrupt > line A,B,C,D are actually routed and fill in the interrupt register > value. > > If the interrupt register is 0 then the BIOS didn't enumerate the > interrupt lines. I've not played with this under VxWorks, but under > IxWorks it used to mess this up a bit. > > Try setting your BIOS to non-plug and play OS, then rebooting to DOS. > Take a look at the register value, and use that. > > Roger. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: MPC860 watchdog help ... real time scheduling problem Date: Thu, 3 Oct 2002 16:49:12 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Jim, I assume this is a doggie that will reset the board if it isnt fed for 1 second. If you do this in a task with taskDelay, then you need to leave some margin, i.e. if the dog bytes in 1 sec, do taskdelay( sysClkRateGet() /2 ), give it a longer leash. During intialization, it could take a few seconds until your dog feeder task can run, you must howver enable the dog as early as possible, so even if init freezes, the dog will give it another try. During initalizations, I sugegst you add single calls (no loops) to giveTheDogABone(), e.g. after every "major" initialziation is done, sysHwInit2, MMU, network , storage, ... Then when the task is running, you need to decide to give it a "middle" priority, i.e. if the dog feeder task is running at priority 200, if any task between priority 0 to 199 runs amok, the dog will reset. If there are tasks that SHOULD hog the CPU (compute intensive tasks), assign thempriority 200 to 255, and they wont bite the dog. Of course, this is an "off the top of my head" advise, let me know if you need some real consultation. Hope this helps, Leonid "Jim" wrote in message news:djYm9.14291$V21.349154@news... > Hello, > > I am working on the a routine to service the MPC860 watchdog. I am having > problems serviceing the watchdog despite the fact my watchdog service > routine is a priority 0. > > My watch dog routine is essentailly > > for(;;) { > giveTheDogABone(); > taskDelay(60); > } > > I am having problems servicing the watch during startup only ... I am now > curious if the way I am servicing the watchdog is incorrect. Mostly I am > wondering if: > > 1.) should i use taskDelay(60) ... if i use this i think a low priority task > can still hog the processor ... unless there is an ISR or someone blocks. > > 2.) or should i use a scheduled interrupt to do this servicing ... > > > Cheers, > Jim > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: MPC860 watchdog help ... real time scheduling problem Date: Thu, 03 Oct 2002 15:03:56 GMT From: "Jim" Organization: Magma Communications Ltd. Message-ID: References: Thanks Leonid, what i think is happening to me is this ... My watchdog task is at priority 0 ... this is the first task i start ... Next I start up a series of tasks at lower priorities ... Here's my theory given I have 1.5 seconds until my watchdog starts ... 1.) I start my watchdog at priority 0 which wakes up every 0.5 seconds ... 2.) I start a task ... at priority 25 3.) The task takes say 1.6 seconds to initialize ... and doesn't block on anything ... no ISR's etc ... Is it possible for the task at 1.6 seconds (unblocking) to prevent the watchdog from being serviced? I assume it is ... I believe in vxworks a scheduling decision is made only on an ISR or a running task Blocking ... Or is there another way that the priority zero task would be made running ... Thanks again, Jim "Leonid Rosenboim" wrote in message news:newscache$vmue3h$427$1@lnews.actcom.co.il... > Jim, I assume this is a doggie that will reset the board if it isnt fed > for 1 second. > > If you do this in a task with taskDelay, then you need to leave some > margin, i.e. if the dog bytes in 1 sec, do taskdelay( sysClkRateGet() /2 ), > give it a longer leash. > > During intialization, it could take a few seconds until your dog feeder task > can run, you must howver enable the dog as early as possible, > so even if init freezes, the dog will give it another try. > > During initalizations, I sugegst you add single calls (no loops) to > giveTheDogABone(), e.g. after every "major" initialziation is done, > sysHwInit2, MMU, network , storage, ... > > Then when the task is running, you need to decide to give it a "middle" > priority, i.e. if the dog feeder task is running at priority 200, if any > task between priority 0 to 199 runs amok, the dog will reset. If there are > tasks that SHOULD > hog the CPU (compute intensive tasks), assign thempriority 200 to 255, > and they wont bite the dog. > > Of course, this is an "off the top of my head" advise, let me know if you > need some real consultation. > > Hope this helps, > Leonid > > "Jim" wrote in message > news:djYm9.14291$V21.349154@news... > > Hello, > > > > I am working on the a routine to service the MPC860 watchdog. I am having > > problems serviceing the watchdog despite the fact my watchdog service > > routine is a priority 0. > > > > My watch dog routine is essentailly > > > > for(;;) { > > giveTheDogABone(); > > taskDelay(60); > > } > > > > I am having problems servicing the watch during startup only ... I am now > > curious if the way I am servicing the watchdog is incorrect. Mostly I am > > wondering if: > > > > 1.) should i use taskDelay(60) ... if i use this i think a low priority > task > > can still hog the processor ... unless there is an ISR or someone blocks. > > > > 2.) or should i use a scheduled interrupt to do this servicing ... > > > > > > Cheers, > > Jim > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: MPC860 watchdog help ... real time scheduling problem Date: Thu, 03 Oct 2002 15:11:34 GMT From: "Jim" Organization: Magma Communications Ltd. Message-ID: References: When I say no ISR's ... I mean my pheripherals are not generating ISR's ... the 860 is runnning fully ISR's are enabled etc "Jim" wrote in message news:w%Ym9.14305$V21.349606@news... > Thanks Leonid, > > what i think is happening to me is this ... > > My watchdog task is at priority 0 ... this is the first task i start ... > Next I start up a series of tasks at lower priorities ... > > Here's my theory given I have 1.5 seconds until my watchdog starts ... > > 1.) I start my watchdog at priority 0 which wakes up every 0.5 seconds ... > 2.) I start a task ... at priority 25 > 3.) The task takes say 1.6 seconds to initialize ... and doesn't block on > anything ... no ISR's etc ... > > Is it possible for the task at 1.6 seconds (unblocking) to prevent the > watchdog from being serviced? > > I assume it is ... I believe in vxworks a scheduling decision is made only > on an ISR or a running task Blocking ... > > Or is there another way that the priority zero task would be made running > ... > > Thanks again, > Jim > > > > "Leonid Rosenboim" wrote in message > news:newscache$vmue3h$427$1@lnews.actcom.co.il... > > Jim, I assume this is a doggie that will reset the board if it isnt fed > > for 1 second. > > > > If you do this in a task with taskDelay, then you need to leave some > > margin, i.e. if the dog bytes in 1 sec, do taskdelay( sysClkRateGet() > /2 ), > > give it a longer leash. > > > > During intialization, it could take a few seconds until your dog feeder > task > > can run, you must howver enable the dog as early as possible, > > so even if init freezes, the dog will give it another try. > > > > During initalizations, I sugegst you add single calls (no loops) to > > giveTheDogABone(), e.g. after every "major" initialziation is done, > > sysHwInit2, MMU, network , storage, ... > > > > Then when the task is running, you need to decide to give it a "middle" > > priority, i.e. if the dog feeder task is running at priority 200, if any > > task between priority 0 to 199 runs amok, the dog will reset. If there are > > tasks that SHOULD > > hog the CPU (compute intensive tasks), assign thempriority 200 to 255, > > and they wont bite the dog. > > > > Of course, this is an "off the top of my head" advise, let me know if you > > need some real consultation. > > > > Hope this helps, > > Leonid > > > > "Jim" wrote in message > > news:djYm9.14291$V21.349154@news... > > > Hello, > > > > > > I am working on the a routine to service the MPC860 watchdog. I am > having > > > problems serviceing the watchdog despite the fact my watchdog service > > > routine is a priority 0. > > > > > > My watch dog routine is essentailly > > > > > > for(;;) { > > > giveTheDogABone(); > > > taskDelay(60); > > > } > > > > > > I am having problems servicing the watch during startup only ... I am > now > > > curious if the way I am servicing the watchdog is incorrect. Mostly I am > > > wondering if: > > > > > > 1.) should i use taskDelay(60) ... if i use this i think a low priority > > task > > > can still hog the processor ... unless there is an ISR or someone > blocks. > > > > > > 2.) or should i use a scheduled interrupt to do this servicing ... > > > > > > > > > Cheers, > > > Jim > > > > > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: problems with ldppc Date: 3 Oct 2002 09:10:08 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0210030810.2417be6b@posting.google.com> References: <5947a04.0210030159.7eec97fd@posting.google.com> Hello, It is not clear what you are trying to do (especially since you talk about a 'bsl' and not a BSP - did you mean BSP?). Either way, what you are doing is just not going to work for VxWorks... here's some better ideas: 1) If you are trying to link your application to the bootable VxWorks image: - Copy test.o to the BSP directory - Type: % make ADDED_MODULES=test.o If you have more than one .o file to add (in the future) use: % make ADDED_MODULES="test.o test2.o thirdModule.o" (notice the quotes). Reboot your system (making sure that the bootrom's filename parameter is pointing at the newly built image. 2) If you are trying to use your application as a downloadable file, then you don't need to do anything else: the .o file is perfect for VxWorks to download in all probability (as long as your kernel contains all the services that the application uses - if not, you'll need to change the config.h file to define the "INCLUDE_" macros for the missing feature(s)). HTH, John... andreas_lindell75@hotmail.com (Rikard) wrote in message news:<5947a04.0210030159.7eec97fd@posting.google.com>... > Hello, > > I am working on a CHAMP AVII board from DY4 and will run VxWorks on > proc A and application using the BSL on B, C and D. It are these > applications that are causing my problems and I have difficulty > linking them. > > I have created a small application in Tornado that uses a function in > the BSL and a printf line. I compile it and get a file test.o that I > need to link to the bsl and standard C libraries. I use the command > ldppc: > ldppc -X -N -o test.out test.o -L$(AVII)/lib > I get errors saying: > ldppc: warning: cannot find entry symbol _start; defaulting to > 00040074 > lab1.o: In function `vx_main': > lab1.o(.text+0x10): undefined reference to `_bslBoardInit' > lab1.o(.text+0x24): undefined reference to `printf' > I am new to linking from command line, Tornado usualy does that for > me:(, and now I am paying the price. So my questions is as follows. > > I need to point the linker to the correct directories, so if I have > Tornado installed in d:\tornado where and how do I point the linker to > solve undef ref to printf. The `_bslBoardInit' is due to me not > pointing to the correct directory where the includes for AV2 is stored > but I hope that problem will go away once I can point correctly for > the printf, then I will know how to point. > So if anyone knows how to point the linker please provide me with some > help. > > Regards > Andreas --------------------------- Newsgroups: comp.os.vxworks Subject: Re: help with objcopyppc Date: 03 Oct 2002 10:12:33 -0600 From: taashlo@sandia.gov Organization: Sandia National Laboratories, Albuquerque, NM USA Message-ID: References: <5947a04.0210021942.5f8cbc94@posting.google.com> Sender: taashlo@TYPHOON andreas_lindell75@hotmail.com (Rikard) writes: > Hello, > > I have a file(test.out) that I want to convert into a .s file using > the objcopyppc utility provided by WR. However, I am not very good at > these things so I have problems with all the switches and flags that > needs to be set. I have tried the following: > objcopyppc -S test.out test.s > But then the size of file test.s is only 1KB (test.out is 12 kb) so > that is not the correct settings. > > Does anyone know what switches I need to set to convert a .out file to > a .s file using the objcopyppc? > > Regards > Andreas Andreas, Try this: objcopyppc -O srec test.out test.s Tad --------------------------- Newsgroups: comp.os.vxworks Subject: Re: MPC860 watchdog help ... real time scheduling problem Date: Thu, 3 Oct 2002 18:35:34 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Jim, I have to assume that your BSP is funcioning OK. If it does, there is an interrupt every 1/60th of a second, that is the system clock, which calls a system function which checks which of the delayed tasks need to be awaken, and if one (or more) of the delayed tasks have awaken, and at higher priority than the currently running task, a rescheduling takes place. So if all is as you described it, the watchdog task should be awaken and hit the dog. But if that priority 25 task is doing taskLock() or even worse - intLock(), it could prevent the watchdog task from running, until it yelds the CPU, or unlocks. "Jim" wrote in message news:w%Ym9.14305$V21.349606@news... > Thanks Leonid, > > what i think is happening to me is this ... > > My watchdog task is at priority 0 ... this is the first task i start ... > Next I start up a series of tasks at lower priorities ... > > Here's my theory given I have 1.5 seconds until my watchdog starts ... > > 1.) I start my watchdog at priority 0 which wakes up every 0.5 seconds ... > 2.) I start a task ... at priority 25 > 3.) The task takes say 1.6 seconds to initialize ... and doesn't block on > anything ... no ISR's etc ... > > Is it possible for the task at 1.6 seconds (unblocking) to prevent the > watchdog from being serviced? > > I assume it is ... I believe in vxworks a scheduling decision is made only > on an ISR or a running task Blocking ... > > Or is there another way that the priority zero task would be made running > ... > > Thanks again, > Jim > > > > "Leonid Rosenboim" wrote in message > news:newscache$vmue3h$427$1@lnews.actcom.co.il... > > Jim, I assume this is a doggie that will reset the board if it isnt fed > > for 1 second. > > > > If you do this in a task with taskDelay, then you need to leave some > > margin, i.e. if the dog bytes in 1 sec, do taskdelay( sysClkRateGet() > /2 ), > > give it a longer leash. > > > > During intialization, it could take a few seconds until your dog feeder > task > > can run, you must howver enable the dog as early as possible, > > so even if init freezes, the dog will give it another try. > > > > During initalizations, I sugegst you add single calls (no loops) to > > giveTheDogABone(), e.g. after every "major" initialziation is done, > > sysHwInit2, MMU, network , storage, ... > > > > Then when the task is running, you need to decide to give it a "middle" > > priority, i.e. if the dog feeder task is running at priority 200, if any > > task between priority 0 to 199 runs amok, the dog will reset. If there are > > tasks that SHOULD > > hog the CPU (compute intensive tasks), assign thempriority 200 to 255, > > and they wont bite the dog. > > > > Of course, this is an "off the top of my head" advise, let me know if you > > need some real consultation. > > > > Hope this helps, > > Leonid > > > > "Jim" wrote in message > > news:djYm9.14291$V21.349154@news... > > > Hello, > > > > > > I am working on the a routine to service the MPC860 watchdog. I am > having > > > problems serviceing the watchdog despite the fact my watchdog service > > > routine is a priority 0. > > > > > > My watch dog routine is essentailly > > > > > > for(;;) { > > > giveTheDogABone(); > > > taskDelay(60); > > > } > > > > > > I am having problems servicing the watch during startup only ... I am > now > > > curious if the way I am servicing the watchdog is incorrect. Mostly I am > > > wondering if: > > > > > > 1.) should i use taskDelay(60) ... if i use this i think a low priority > > task > > > can still hog the processor ... unless there is an ISR or someone > blocks. > > > > > > 2.) or should i use a scheduled interrupt to do this servicing ... > > > > > > > > > Cheers, > > > Jim > > > > > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: How do I intConnect an IRQ to the FCC on wrSbc8260 Date: 3 Oct 2002 11:27:14 -0700 From: FrankGleeson@motorola.com (frankg) Organization: http://groups.google.com/ Message-ID: I am porting the wrSbc8260 BSP to run on our hardware. I have the source for sysMotFccEnd.c. How do I go about configuring this code to work on my hardware? Specifically: How do I configure it to use external interrupt pin that the PHY chip is wired to? Our design shares the same interrupt with FCC1 and FCC2. Is it possible to get source for the interrupt handler? I need to add logic that queries the 2 PHY devices and figure out which 1 is interrupting (possibly both). What is the best way to do this? --------------------------- Newsgroups: comp.os.vxworks Subject: Msg Q issues Date: 3 Oct 2002 11:37:26 -0700 From: brian_doe@hotmail.com (brian) Organization: http://groups.google.com/ Message-ID: Hi All, I have come across a piece of code that does the following: Both a task and an ISR are writing to the same message queue. It looks like the message queue is being corrupted. Im assuming that is because the ISR is coming in when the task is in the process of writing to the queue therefore causing the corruption. Does that assumption make sense? (i.e., WRS doesnt disable ints when it writes to a q?..no reason it should) One work around is that the ISR can post to a diff Q (one that it alone posts to) and that message is forwarded to the real Q. Anyone else see a more elegant solution/design? thanx :) brian --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Msg Q issues Date: Thu, 3 Oct 2002 21:03:03 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Brian, there could be two reason for this corruption that I can think of: 1. if you send your data in more than one msgQSend call, the ISR call could come between the calls. 2. If the ISR happens in the midst of the task executing msgQSend, then the ISR's msgQSend request will be DEFERRED (i.e. put into a special kernel workQ) and executed rught after the critical part of the task's sys call is done. Now this can cause trouble if the arguments for the msgQSend of the ISR are volatile - i.e. they are in stack, and the stack could dissapear before the deferred execution actually takes place. The solution for (2) is to make sure that msgQSend does not use a pointer to a stack variable for the actual message. Although I did not indend to explain exactly how the VxWorks kernel operates, you can probably figure out the principle. "brian" wrote in message news:d2f67bd8.0210031037.12839cfd@posting.google.com... > Hi All, > > I have come across a piece of code that does the following: > > Both a task and an ISR are writing to the same message queue. It looks > like the message queue is being corrupted. Im assuming that is because > the ISR is coming in when the task is in the process of writing to the > queue therefore causing the corruption. Does that assumption make > sense? (i.e., WRS doesnt disable ints when it writes to a q?..no > reason it should) One work around is that the ISR can post to a diff > Q (one that it alone posts to) and that message is forwarded to the > real Q. Anyone else see a more elegant solution/design? > > thanx :) > > brian --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Creating Singleton objects at run time and using it in a shared library Date: Thu, 03 Oct 2002 20:35:02 GMT From: "H. E. Taylor" Organization: MTS Internet Message-ID: <3D9CC452.4EF3@despam.autobahn.mb.ca> References: In article , stephen hallacy wrote: > > I want to create an singleton object at bootup time or some time after > bootup time, and I want to be able to access that object's methods in > a shared library. Is this possible, > Yes. >and if so, how is it done? > The usual way. ;-) Really. What difficulty are you having? Setup a global function GetObjectX() which uses a static pointer to your instance. The first time through you instantiate, save & return the pointer. Later times you just return the pointer. You should probably consider tear down at this stage of design as well. You may think your object will live forever, but as soon as you implement it, some manager will change the spec on you. - -het PS. Possibly useful links. http://hillside.net/patterns/ http://www.autobahn.mb.ca/~het/clinks.html#CPlusPlus - -- "progress in software has not followed Moore's law." -John Holland Energy Alternatives: http://www.autobahn.mb.ca/~het/energy.html H.E. Taylor http://www.autobahn.mb.ca/~het/ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Strange " warning: control reaches end of non-void function" Date: Thu, 03 Oct 2002 21:31:17 +0000 From: molochai Organization: Griffin Internet - http://www.griffin.net.uk Message-ID: <6vcina.d73.ln@gonzo.cybermuppet.co.uk> References: Ofer Rotschield wrote: > Please advise, > >>>> Almost certainly a compiler bug. Try comiling the fragment using a newer version of GCC to be sure. That version of GCC is quite old and has quite a few bugs in it. You ought Tornodo 2.2 is just about out now for PowerPC (certainly for Win32 hosts, not sure about Solaris though). Failing that, you ought to be able to get Dave Korn's version of GCC to compile under solaris. This is the same GCC version - or very close - to the one WRS used. > ***/ > > class cc > { > public: > ~cc() {}; > }; > > int foo () > { > cc v1[1]; > cc v2[1]; > > return 1; > } > > <<< --------------------------- Newsgroups: comp.os.vxworks Subject: Re: problems with ldppc Date: 3 Oct 2002 18:06:40 -0700 From: andreas_lindell75@hotmail.com (Rikard) Organization: http://groups.google.com/ Message-ID: <5947a04.0210031706.28913643@posting.google.com> References: <5947a04.0210030159.7eec97fd@posting.google.com> <488e459a.0210030810.2417be6b@posting.google.com> Hello Everyone/John, John thanks for the feedback but it does not solve my problem maybe due to me not expressing my self clearly. BSL stands for "Board Support Library" and is DY4 proprietary library for supporting applications not using an OS to run on CHAMP AV2(4 CPU's). This BSL allows you to run OS(vxworks) on processor 1 and then use the OS to boot normal applications that does not need a OS on proc 2, 3 and 4. This is a popular configuration to save money on run time license from WR. So what I need is just some old fashion linking:) between C libraries and the BSL library not to VxWorks. This linking will produce a .out file that will be converted into a motorola S file and then booted by VxWorks to proc 2, 3 and 4. Here is my question again: I have created a small application in Tornado that uses a function in the BSL and a printf line. I compile it and get a file test.o that I need to link to the bsl and standard C libraries. I use the command ldppc: ldppc -X -N -o test.out test.o -L$(AVII)/lib I get errors saying: dppc: warning: cannot find entry symbol _start; defaulting to 00040074 lab1.o: In function `vx_main': lab1.o(.text+0x10): undefined reference to `_bslBoardInit' lab1.o(.text+0x24): undefined reference to `printf' I am new to linking from command line, Tornado usualy does that for me:(, and now I am paying the price. So my questions is as follows. I need to point the linker to the correct directories, so if I have Tornado installed in d:\tornado where and how do I point the linker to solve undef ref to printf. The `_bslBoardInit' is due to me not pointing to the correct directory where the includes for AV2 is stored but I hope that problem will go away once I can point correctly for the printf, then I will know how to point. So if anyone knows how to point the linker please provide me with some help. Regards Andreas john_94501@yahoo.com (John) wrote in message news:<488e459a.0210030810.2417be6b@posting.google.com>... > Hello, > > It is not clear what you are trying to do (especially since you talk > about a 'bsl' and not a BSP - did you mean BSP?). Either way, what you > are doing is just not going to work for VxWorks... here's some better > ideas: > > 1) If you are trying to link your application to the bootable VxWorks > image: > > - Copy test.o to the BSP directory > > - Type: > > % make ADDED_MODULES=test.o > > If you have more than one .o file to add (in the future) use: > > % make ADDED_MODULES="test.o test2.o thirdModule.o" > > (notice the quotes). > > Reboot your system (making sure that the bootrom's filename > parameter is > pointing at the newly built image. > > > 2) If you are trying to use your application as a downloadable file, > then you don't need to do anything else: the .o file is perfect for > VxWorks to download in all probability (as long as your kernel > contains all the services that the application uses - if not, you'll > need to change the config.h file to define the "INCLUDE_" macros for > the missing feature(s)). > > HTH, > > John... > > andreas_lindell75@hotmail.com (Rikard) wrote in message news:<5947a04.0210030159.7eec97fd@posting.google.com>... > > Hello, > > > > I am working on a CHAMP AVII board from DY4 and will run VxWorks on > > proc A and application using the BSL on B, C and D. It are these > > applications that are causing my problems and I have difficulty > > linking them. > > > > I have created a small application in Tornado that uses a function in > > the BSL and a printf line. I compile it and get a file test.o that I > > need to link to the bsl and standard C libraries. I use the command > > ldppc: > > ldppc -X -N -o test.out test.o -L$(AVII)/lib > > I get errors saying: > > ldppc: warning: cannot find entry symbol _start; defaulting to > > 00040074 > > lab1.o: In function `vx_main': > > lab1.o(.text+0x10): undefined reference to `_bslBoardInit' > > lab1.o(.text+0x24): undefined reference to `printf' > > I am new to linking from command line, Tornado usualy does that for > > me:(, and now I am paying the price. So my questions is as follows. > > > > I need to point the linker to the correct directories, so if I have > > Tornado installed in d:\tornado where and how do I point the linker to > > solve undef ref to printf. The `_bslBoardInit' is due to me not > > pointing to the correct directory where the includes for AV2 is stored > > but I hope that problem will go away once I can point correctly for > > the printf, then I will know how to point. > > So if anyone knows how to point the linker please provide me with some > > help. > > > > Regards > > Andreas --------------------------- Newsgroups: comp.os.vxworks Subject: Strange Interrupt log msg Date: 3 Oct 2002 18:25:25 -0700 From: tgisfree@yahoo.com (tzeghuan) Organization: http://groups.google.com/ Message-ID: <290ade1c.0210031725.249f87a6@posting.google.com> Hi, I have this strange interrupt log msg : "interrupt: Bu Read Ram Busy!" whenever I compiled using Tornado with option -O2. But when compiled using -g, I don't get the msg. I am using Tornado 2 and vxWorks 5.4 What could be the problems? Any kind of helps are appreciated. Thanks in advance. rdgs tzeghuan --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Tornado II on XP Date: Fri, 4 Oct 2002 11:31:30 +0800 From: "Yung" Organization: IMS Netvigator Message-ID: References: me too. WinXp (Home) with Tornado 2.0. No problem so far. "Bob Bradley" wrote in message news:bob-321911.17333701102002@news.apple.com... > In article , > molochai wrote: > > > Anyone know whether Tornado II works correctly on XP? I know there was a > > patch for Win2k - is this sufficient? > > I use Tornado 2.0.2 on Windows XP and it seems to work correctly. --------------------------- Newsgroups: comp.realtime,comp.os.vxworks Subject: Re: Problem with vxworks Date: 3 Oct 2002 22:29:00 -0700 From: wv9557@yahoo.com (Will) Organization: http://groups.google.com/ Message-ID: <4a885870.0210032129.6a11a9e0@posting.google.com> References: <9f000cfd.0209302147.1ff4c00a@posting.google.com> <4a885870.0210011946.4271985d@posting.google.com> <9f000cfd.0210022238.6e71211a@posting.google.com> well, ICMP succeeds while FTP fails. All the more of a reason why you should examine FTP packets. yathish_g@yahoo.com wrote in message news:<9f000cfd.0210022238.6e71211a@posting.google.com>... > Hi will, > > Thanks for the reply. In the first mail, I have already mentioned that > I am able to ping from the host to the vxworks box. I am able to ping > successfully without any packet loss. Both the host and the target are > able to sniff each other as far as the packet exchange is concerned. > > Both the target and the host are forced to do at 10 Mbps. > > regds, > Yathish > > > wv9557@yahoo.com (Will) wrote in message news:<4a885870.0210011946.4271985d@posting.google.com>... > > The log only shows one side of the story. > > Not seeing the whole packets exchange, it's hard to tell. You might > > want to get something like tcpdump for Windows (yah - > > tcpdump does run on Windows :) > > and look at every packet that is exchanged between the 2 boxes. > > > > Just a shoot in the dark: after you swapped the cards, wait long > > enough for the PC's ARP cache to timeout. > > > > > > Just a shoot in the dark number 2: > > Force both the vxWorks box and the PC to do 10mbits, by using > > a good old 10mbit-only switch. > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: help with ldppc Date: 3 Oct 2002 23:38:55 -0700 From: andreas_lindell75@hotmail.com (Rikard) Organization: http://groups.google.com/ Message-ID: <5947a04.0210032238.2c8760e5@posting.google.com> Hello, I am new to using the ldppc from command line and have encounterd the following problem. I have a small application: #include "stdio.h" void vx_main(void) { printf("Hello world\n"); } This application has been built from Tornado workspace and is a downloadble application, I compile it and get a test.o file. This file I want to link using ldppc into a test.out file that I later want to convert into a Motorola S file. I use the following ldppc command from dos: ldppc -X -N -e vx_main -Ttext 00200000 -o test.out test.o - -Ld:/tornado/target/lib/libPPC604gnuvx.a And I get the following error: lab1.o: In function `vx_main': lab1.o(.text+0x20): undefined reference to `printf' Does anyone know how to solve the undefined reference to 'printf'? I assume that I am not including some library or I am including the libPPC604gnuvx.a in the wrong way. Any suggestion to what I am doing wrong would be helpfull. The reason I want to link from command line is becasue in a later stage I will use a Board Support Libary from DY4 and then I will need to link my application from command line into this library. Regards Andreas --------------------------- Newsgroups: comp.os.vxworks Subject: hibernation ? Date: 4 Oct 2002 01:00:18 -0700 From: s_popov@prosyst.bg (Alexander Popov) Organization: http://groups.google.com/ Message-ID: <6d1b7093.0210040000.1df7fd48@posting.google.com> Hi all, Does VxWorks support system hibernation ( suspend-to-ram, suspend-to-disk )? If yes is part of the standard os/kernel or is it an extension? How about suspend-to-disk for a single process? Regards, Alexander Popov --------------------------- Newsgroups: comp.os.vxworks Subject: Re: hibernation ? Date: Fri, 4 Oct 2002 10:25:04 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <6d1b7093.0210040000.1df7fd48@posting.google.com> Sasha, VxWorks has no support for the feature you mention. It is far from enough to support such features in the OS, this must be also supported in all device drivers to work properly. Besides, VxWorks does not allways need to have a disk to operate. If you need to conserve battery power, your first step must be to select a processor which is low-power, i.e. ARM. Also, for VxWorks you will need a lot less memory, and that will save on power too, or might even be small emough to afford to use SRAM (instead of DRAM) which is very power-effecient. Next, I would suggest to identify peripheral components that drain power significantly, and build a "suspension" mode into these devices (separately) so the main software can turn these devices off when not in use (e.g. Flash VPP, wireless transceiver). Lastly, some processors support different operating modes, allowing to compromise power for speed, and you can switch these modes transparently to the OS. Hope this helps, Leonid "Alexander Popov" wrote in message news:6d1b7093.0210040000.1df7fd48@posting.google.com... > Hi all, > > Does VxWorks support system hibernation ( suspend-to-ram, suspend-to-disk )? > If yes is part of the standard os/kernel or is it an extension? > How about suspend-to-disk for a single process? > > Regards, > Alexander Popov --------------------------- Newsgroups: comp.os.vxworks Subject: Addition to the FAQ: "newbie" page Date: Fri, 4 Oct 2002 09:20:27 +0000 (UTC) From: Johan Borkhuis Organization: Proxim Corp. Message-ID: One thing that comes up quite often lately is "newbies" asking for information and starting points. I added a section to the FAQ for this, with some basic info. Please have a look and comment on it. Also if you have links that are usefull for "newbies" or if you have book reviews or tips (usefull for "newbies" and "oldies") please let me know, so that I can add these to this page. The page can be found at: http://www.xs4all.nl/~borkhuis/vxworks/vxw_pt0.html Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.sys.powerpc.tech,comp.os.vxworks Subject: MPC8260 - FCC - Busy LAN (RET_LIM, P_PER & FPSMR[STB] settings) Date: Fri, 4 Oct 2002 11:12:07 +0100 From: "David J Edgar" Organization: Cedaryacht Limited Message-ID: I'm running an MPC8260A with Ethernet using FCC2. When connected to a 'busy LAN' I keep getting the 'RL' bit set in the TX Buffer Descriptor. Handling this requires a TX Restart which would seem overkill. Note: Originally RET_LIM = 15, P_PER = 0 & FPSMR[STB] = 0 After reading the User Manual I found that by ether increasing RET_LIM or setting FPSMR[STB] = 1 the problem went away (setting P_PER did not help). So, while I have a solution I'd like to find out what other people have done when faced with this situation? - -- David J Edgar Cedaryacht Limited, Macclesfield, Cheshire, UK E-mail: david@cd-yacht.demon.co.uk Web: http://www.cedaryacht.co.uk --------------------------- Newsgroups: comp.os.vxworks Subject: Re: problems with ldppc Date: Fri, 04 Oct 2002 06:46:49 -0400 From: joe durusau Organization: WEBUSENET.com Message-ID: <3D9D7199.73FC9B49@bellsouth.net> References: <5947a04.0210030159.7eec97fd@posting.google.com> <488e459a.0210030810.2417be6b@posting.google.com> <5947a04.0210031706.28913643@posting.google.com> In normal operation, there is _no_ C library linked in with vxWorks. Everything is in libvx.a, which is part of vxWorks itself. You need to find out what the BSL contains and restrict yourself to those things, assuming that it has a library file that comes with it. You will _not_ be able to use the standard C library, (unless you port it to the processor you are using) and I doubt that the BSL will have very much in it. HTH. Speaking only for myself, Joe Durusau Rikard wrote: > Hello Everyone/John, > > John thanks for the feedback but it does not solve my problem maybe > due to me not expressing my self clearly. > > BSL stands for "Board Support Library" and is DY4 proprietary library > for supporting applications not using an OS to run on CHAMP AV2(4 > CPU's). This BSL allows you to run OS(vxworks) on processor 1 and then > use the OS to boot normal applications that does not need a OS on proc > 2, 3 and 4. This is a popular configuration to save money on run time > license from WR. > > So what I need is just some old fashion linking:) between C libraries > and the BSL library not to VxWorks. This linking will produce a .out > file that will be converted into a motorola S file and then booted by > VxWorks to proc 2, 3 and 4. Here is my question again: > > I have created a small application in Tornado that uses a function in > the BSL and a printf line. I compile it and get a file test.o that I > need to link to the bsl and standard C libraries. I use the command > ldppc: > ldppc -X -N -o test.out test.o -L$(AVII)/lib > I get errors saying: > dppc: warning: cannot find entry symbol _start; defaulting to 00040074 > lab1.o: In function `vx_main': > lab1.o(.text+0x10): undefined reference to `_bslBoardInit' > lab1.o(.text+0x24): undefined reference to `printf' > I am new to linking from command line, Tornado usualy does that for > me:(, and now I am paying the price. So my questions is as follows. > > I need to point the linker to the correct directories, so if I have > Tornado installed in d:\tornado where and how do I point the linker to > solve undef ref to printf. The `_bslBoardInit' is due to me not > pointing to the correct directory where the includes for AV2 is stored > but I hope that problem will go away once I can point correctly for > the printf, then I will know how to point. > So if anyone knows how to point the linker please provide me with some > help. > > Regards > Andreas > > john_94501@yahoo.com (John) wrote in message news:<488e459a.0210030810.2417be6b@posting.google.com>... > > Hello, > > > > It is not clear what you are trying to do (especially since you talk > > about a 'bsl' and not a BSP - did you mean BSP?). Either way, what you > > are doing is just not going to work for VxWorks... here's some better > > ideas: > > > > 1) If you are trying to link your application to the bootable VxWorks > > image: > > > > - Copy test.o to the BSP directory > > > > - Type: > > > > % make ADDED_MODULES=test.o > > > > If you have more than one .o file to add (in the future) use: > > > > % make ADDED_MODULES="test.o test2.o thirdModule.o" > > > > (notice the quotes). > > > > Reboot your system (making sure that the bootrom's filename > > parameter is > > pointing at the newly built image. > > > > > > 2) If you are trying to use your application as a downloadable file, > > then you don't need to do anything else: the .o file is perfect for > > VxWorks to download in all probability (as long as your kernel > > contains all the services that the application uses - if not, you'll > > need to change the config.h file to define the "INCLUDE_" macros for > > the missing feature(s)). > > > > HTH, > > > > John... > > > > andreas_lindell75@hotmail.com (Rikard) wrote in message news:<5947a04.0210030159.7eec97fd@posting.google.com>... > > > Hello, > > > > > > I am working on a CHAMP AVII board from DY4 and will run VxWorks on > > > proc A and application using the BSL on B, C and D. It are these > > > applications that are causing my problems and I have difficulty > > > linking them. > > > > > > I have created a small application in Tornado that uses a function in > > > the BSL and a printf line. I compile it and get a file test.o that I > > > need to link to the bsl and standard C libraries. I use the command > > > ldppc: > > > ldppc -X -N -o test.out test.o -L$(AVII)/lib > > > I get errors saying: > > > ldppc: warning: cannot find entry symbol _start; defaulting to > > > 00040074 > > > lab1.o: In function `vx_main': > > > lab1.o(.text+0x10): undefined reference to `_bslBoardInit' > > > lab1.o(.text+0x24): undefined reference to `printf' > > > I am new to linking from command line, Tornado usualy does that for > > > me:(, and now I am paying the price. So my questions is as follows. > > > > > > I need to point the linker to the correct directories, so if I have > > > Tornado installed in d:\tornado where and how do I point the linker to > > > solve undef ref to printf. The `_bslBoardInit' is due to me not > > > pointing to the correct directory where the includes for AV2 is stored > > > but I hope that problem will go away once I can point correctly for > > > the printf, then I will know how to point. > > > So if anyone knows how to point the linker please provide me with some > > > help. > > > > > > Regards > > > Andreas --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Pci interrupts in a Pentium system. Date: 4 Oct 2002 03:58:37 -0700 From: google@fareham.org (Roger) Organization: http://groups.google.com/ Message-ID: <63f42115.0210040258.61c46bd4@posting.google.com> References: <63f42115.0210030217.655d7acc@posting.google.com> > I have an ethernet card in the system that has been enumerated and > it's int line register is set to 0x0B. Is it possible that the vxWorks > ethernet driver did this, not the BIOS? Why would the BIOS do just one > card and not the others? I'm fairly sure my card is on IRQ8 because > according to an LED on a card in one of the PCI slots, it is the INT > after the ethernet card on IRQ12, so it should wrap back to IRQ8 > right? Anyway, I tried the code above on all 4 IRQs (8-12) and it > didn't work on any of them... If the BIOS is told you have a Plag-N-play OS, then it will only assign resourses for possible boot devices, if the ethernet has a boot ROM then that would explain it. It's possible that vxWorks assigns the resourses, but I'd try changing the BIOS setting to non-Plug-and-play, to see what happens. Roger. --------------------------- Newsgroups: comp.sys.powerpc.tech,comp.os.vxworks Subject: Re: MPC8260 - FCC - Busy LAN (RET_LIM, P_PER & FPSMR[STB] settings) Date: Fri, 04 Oct 2002 06:59:24 -0400 From: Varndell Engineering Organization: Posted Via Newsfeeds.com = SPEED+RETENTION+COMPLETION = http://www.newsfeeds.com Message-ID: <3D9D748C.C2E613AF@hotmail.com> References: David J Edgar wrote: > > I'm running an MPC8260A with Ethernet using FCC2. > > When connected to a 'busy LAN' I keep getting the 'RL' bit set in the TX > Buffer Descriptor. Handling this requires a TX Restart which would seem > overkill. > > Note: Originally RET_LIM = 15, P_PER = 0 & FPSMR[STB] = 0 > > After reading the User Manual I found that by ether increasing RET_LIM or > setting FPSMR[STB] = 1 the problem went away (setting P_PER did not help). > > So, while I have a solution I'd like to find out what other people have done > when faced with this situation? I've handled this the same way as you. Given the cost of allowing RL errors, it makes sense to give the FCC a fighting chance at succeeding on every transmit attempt. Late collisions and underruns are still problematic, but they rarely occur. Regards, George Varndell - -- Varndell Engineering, LLC Embedded Systems Expertise http://www.varndellengineering.com Need a vxWorks END driver for NatSemi DP83815, DP83821, DP83820? Contact us. - -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- http://www.newsfeed.com The #1 Newsgroup Service in the World! - -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.sys.powerpc.tech,comp.os.vxworks Subject: Re: MPC8260 - FCC - Busy LAN (RET_LIM, P_PER & FPSMR[STB] settings) Date: Fri, 4 Oct 2002 14:06:48 +0100 From: "David J Edgar" Organization: Cedaryacht Limited Message-ID: References: <3D9D748C.C2E613AF@hotmail.com> "Varndell Engineering" wrote in message news:3D9D748C.C2E613AF@hotmail.com... > David J Edgar wrote: > > > > I'm running an MPC8260A with Ethernet using FCC2. > > > > When connected to a 'busy LAN' I keep getting the 'RL' bit set in the TX > > Buffer Descriptor. Handling this requires a TX Restart which would seem > > overkill. > > > > Note: Originally RET_LIM = 15, P_PER = 0 & FPSMR[STB] = 0 > > > > After reading the User Manual I found that by ether increasing RET_LIM or > > setting FPSMR[STB] = 1 the problem went away (setting P_PER did not help). > > > > So, while I have a solution I'd like to find out what other people have done > > when faced with this situation? > > I've handled this the same way as you. Given the cost of allowing > RL errors, it makes sense to give the FCC a fighting chance at > succeeding on every transmit attempt. Late collisions and underruns > are still problematic, but they rarely occur. > > Regards, > George Varndell > -- > Varndell Engineering, LLC Embedded Systems Expertise > http://www.varndellengineering.com > Need a vxWorks END driver for NatSemi DP83815, DP83821, DP83820? > Contact us. > > > -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- > http://www.newsfeed.com The #1 Newsgroup Service in the World! > -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- OK - what number do you use in RET_LIM - I know that 100 works but it would seem a bit excessive... - -- David J Edgar Cedaryacht Limited, Macclesfield, Cheshire, UK E-mail: david@cd-yacht.demon.co.uk Web: http://www.cedaryacht.co.uk --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Intel chipset 815E Date: 4 Oct 2002 06:21:28 -0700 From: phil.tatterton@micromass.co.uk (Philip Tatterton) Organization: http://groups.google.com/ Message-ID: References: Hi, Could you clarify how the 'PCI device ID constant' is modified & what to. Thanks Phil "Leonid Rosenboim" wrote in message news:... > Yes it works. And you only need to modify the PCI device ID constant for the > standard FEI driver to work on it. > > It also has a good Bus Master IDE controller, but you need to modfy the > ataDrv driver to make use of it and get excellend disk performance. > > wrote in message > news:am79uj$kn7$1@overload.lbl.gov... > > Hello VXWORKS and TORNADO Gurus, > > > > does anyone already works with the intel chipset 815E ? > > I am looking for a driver for the ethernet component integrated inside ... > > > > Thanks in advance > > > > > > > > François PERTIN > > > > > > STÄUBLI SCA > > Robotics Division > > Place Robert Stäubli - BP 70 > > F - 74210 Faverges (France) > > Tel.: +33 (0)4.50.65.62.59 > > Fax: +33 (0)4.50.65.61.30 > > mailto:f.pertin@staubli.com > > http://www.staubli.com > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Intel chipset 815E Date: Fri, 4 Oct 2002 15:25:08 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: The actual IDs of your actual device can be viewed with pciShow(), and the constants are in a BSP file, maybe sysNetIf.c (from top o fmy head). "Philip Tatterton" wrote in message news:fd1d4c48.0210040521.3b6ea820@posting.google.com... > Hi, > Could you clarify how the 'PCI device ID constant' is modified & what to. > Thanks > Phil > > > "Leonid Rosenboim" wrote in message news:... > > Yes it works. And you only need to modify the PCI device ID constant for the > > standard FEI driver to work on it. > > > > It also has a good Bus Master IDE controller, but you need to modfy the > > ataDrv driver to make use of it and get excellend disk performance. > > > > wrote in message > > news:am79uj$kn7$1@overload.lbl.gov... > > > Hello VXWORKS and TORNADO Gurus, > > > > > > does anyone already works with the intel chipset 815E ? > > > I am looking for a driver for the ethernet component integrated inside ... > > > > > > Thanks in advance > > > > > > > > > > > > François PERTIN > > > > > > > > > STÄUBLI SCA > > > Robotics Division > > > Place Robert Stäubli - BP 70 > > > F - 74210 Faverges (France) > > > Tel.: +33 (0)4.50.65.62.59 > > > Fax: +33 (0)4.50.65.61.30 > > > mailto:f.pertin@staubli.com > > > http://www.staubli.com > > > --------------------------- Newsgroups: comp.sys.powerpc.tech,comp.os.vxworks Subject: Re: MPC8260 - FCC - Busy LAN (RET_LIM, P_PER & FPSMR[STB] settings) Date: Fri, 04 Oct 2002 09:45:17 -0400 From: Varndell Engineering Organization: Posted Via Newsfeeds.com = SPEED+RETENTION+COMPLETION = http://www.newsfeeds.com Message-ID: <3D9D9B6D.2B47D6D6@hotmail.com> References: <3D9D748C.C2E613AF@hotmail.com> David J Edgar wrote: > > OK - what number do you use in RET_LIM - I know that 100 works but it would > seem a bit excessive... If you think 100 is excessive, try 1536. In some very early 8260s, the FCC seemed to interpret the value of ret_lim as mrblr instead. The device wouldn't work with ret_lim = 15. Now I just leave it 1536. Given that this is irrelevant in full duplex mode, and given that restarting the transmitter doesn't enhance the transmitter's chances of success when it does restart, I just want the damn thing to keep on trying. Eventually, if they haven't already, Mot will fix the transmitter so it doesn't croak on minor failures and none of this will be necessary. Regards, George Varndell - -- Varndell Engineering, LLC Embedded Systems Expertise http://www.varndellengineering.com Need a vxWorks END driver for NatSemi DP83815, DP83821, DP83820? Contact us. - -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- http://www.newsfeed.com The #1 Newsgroup Service in the World! - -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Using mHdr.mNextPkt Date: Fri, 4 Oct 2002 10:00:05 -0400 From: "Harjinder Chayra" Organization: Storm Internet Services Message-ID: Sender: terayon@gw-terayon.storm.ca When a packet is passed to the END from the stack is it possible that the mHdr.mNextPkt will not be set to NULL? i.e. should my code check this field and send the attached packet? Thanks Harjinder --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Pci interrupts in a Pentium system. Date: 4 Oct 2002 07:01:50 -0700 From: sfinley@xes-inc.com (Scott Finley) Organization: http://groups.google.com/ Message-ID: References: <63f42115.0210030217.655d7acc@posting.google.com> <63f42115.0210040258.61c46bd4@posting.google.com> It looks like the motherboard is just an old piece of crap, and the BIOS (written by Intel) has 3 options I can change, none of them dealing with PnP. Basically I can enable network booting, and change the order of booting to try network first or not. No matter what I do, the network card is the only thing with the Int Line set. Any other ideas? Thanks for your help. google@fareham.org (Roger) wrote in message news:<63f42115.0210040258.61c46bd4@posting.google.com>... > > I have an ethernet card in the system that has been enumerated and > > it's int line register is set to 0x0B. Is it possible that the vxWorks > > ethernet driver did this, not the BIOS? Why would the BIOS do just one > > card and not the others? I'm fairly sure my card is on IRQ8 because > > according to an LED on a card in one of the PCI slots, it is the INT > > after the ethernet card on IRQ12, so it should wrap back to IRQ8 > > right? Anyway, I tried the code above on all 4 IRQs (8-12) and it > > didn't work on any of them... > > If the BIOS is told you have a Plag-N-play OS, then it will only > assign resourses for possible boot devices, if the ethernet has a boot > ROM then that would explain it. > > It's possible that vxWorks assigns the resourses, but I'd try changing > the BIOS setting to non-Plug-and-play, to see what happens. > > Roger. --------------------------- Newsgroups: comp.os.vxworks Subject: What to charge for developing a BSP? Date: 4 Oct 2002 07:15:44 -0700 From: TedGoldblatt@yahoo.com (Ted Goldblatt) Organization: http://groups.google.com/ Message-ID: I have been asked to develop a (semi)custom BSP. While I have done this before, it was always as an employee, not as a consultant. So, I've never had to worry about pricing before. For those that do this, is there a reasonably standard price (or range) for developing BSPs? How about for a high-end comms driver? Thanks for any info! ted (PS: if you post with a no archive header, please email me a copy, as my only current usenet access is though Google... Thanks) --------------------------- Newsgroups: comp.os.vxworks Subject: taskDelay Date: 4 Oct 2002 08:40:33 -0700 From: egoisone@libero.it (Mic) Organization: http://groups.google.com/ Message-ID: Hy guys, just a question: how can I delay a task for 10^-6 seconds? I know usrRoot usually sets tick rate to 60, with a max value equal to 1000 (right?), but this means that to pass an integer to taskDelay (as its prototype says), the shortest delay I can have is 1 ms! What about? Regards, Mic. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: taskDelay Date: Fri, 4 Oct 2002 17:10:23 +0100 From: Wade Oram Organization: None Message-ID: <9QsF$2Kv1bn9EATQ@wtoram.co.uk> References: Reply-To: Wade Oram In message , Mic writes >Hy guys, just a question: how can I delay a task for 10^-6 seconds? I >know usrRoot usually sets tick rate to 60, with a max value equal to >1000 (right?), but this means that to pass an integer to taskDelay (as >its prototype says), the shortest delay I can have is 1 ms! What >about? On my hardware its impossible without using a simple loop to consume the required amount of processor time. Using (Aux/custom?) timers to interrupt after the specified amount of time and give a binary semaphore to the task waiting would appear to be a solution except that the overhead of the interrupt/semTake() is often more than 1us and sometimes much more. (You don't specify what hardware platform your running on so I don't know if this is the case on your platform but unless your right up to date, I suspect that it may well be). A "for( volatile int tDelay = 0; (tDelay < (numMicrosecs * DelayScale) - DelayOffset); ++tDelay);" loop is often the best that you can do. Of course for cross platform portability, DelayOffset and DelayScale would have to be determined by a calibration routine. I don't know of any way to get delays of this magnitude without using processor time. - -- Wade Oram --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Msg Q issues Date: 4 Oct 2002 09:37:29 -0700 From: brian_doe@hotmail.com (brian) Organization: http://groups.google.com/ Message-ID: References: Thanx Leonid. That helps a lot. Brian "Leonid Rosenboim" wrote in message news:... > Brian, > > there could be two reason for this corruption that I can think of: > > 1. if you send your data in more than one msgQSend call, the ISR call > could come between the calls. > 2. If the ISR happens in the midst of the task > executing msgQSend, then the ISR's msgQSend request > will be DEFERRED (i.e. put into a special kernel workQ) > and executed rught after the critical part of the task's sys call > is done. Now this can cause trouble if the arguments > for the msgQSend of the ISR are volatile - i.e. they are in > stack, and the stack could dissapear before the deferred > execution actually takes place. > > The solution for (2) is to make sure that msgQSend does not > use a pointer to a stack variable for the actual message. > > Although I did not indend to explain exactly how the VxWorks > kernel operates, you can probably figure out the principle. > > "brian" wrote in message > news:d2f67bd8.0210031037.12839cfd@posting.google.com... > > Hi All, > > > > I have come across a piece of code that does the following: > > > > Both a task and an ISR are writing to the same message queue. It looks > > like the message queue is being corrupted. Im assuming that is because > > the ISR is coming in when the task is in the process of writing to the > > queue therefore causing the corruption. Does that assumption make > > sense? (i.e., WRS doesnt disable ints when it writes to a q?..no > > reason it should) One work around is that the ISR can post to a diff > > Q (one that it alone posts to) and that message is forwarded to the > > real Q. Anyone else see a more elegant solution/design? > > > > thanx :) > > > > brian --------------------------- Newsgroups: comp.os.vxworks Subject: How ro remove IP END addresses Date: Fri, 4 Oct 2002 18:03:54 +0100 From: "Kevin Garvey" Organization: Esat Net Customer Message-ID: Hi, Once multiple IP addresses have been added to an END, using ifAddrAdd(), does anyboby know how to subsequently remove them, one at a time, without detaching the END from the stack. The routeDelete() utility removes the associated route OK but the address remains against the interface. regards, Kevin Garvey. kgarvey@com21.ie Com21, Cork Airport Business Park, Cork, Ireland. Home Page: http://www.com21.com/ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Pci interrupts in a Pentium system. Date: Fri, 04 Oct 2002 10:26:21 -0400 From: Varndell Engineering Organization: Posted Via Newsfeeds.com = SPEED+RETENTION+COMPLETION = http://www.newsfeeds.com Message-ID: <3D9DA50D.558AED99@hotmail.com> References: Scott Finley wrote: > > I did a search of this newsgroup and found several posts on this > subject, but non seemed to fix my issue. > > I am trying to service interrupts on INT A (IRQ 8) in my pentium > system. I can generate interrupts (I have a card with an LED so I can > see INT A being driven) and connect my handler (calling pciInt(8) > successfully enters my isr) but the interrupt does not seem to be > properly enabled. > > Here is the code I use to set things up: > pciConfigOutByte(0,3,0,0x3c,8); > pciIntConnect(INUM_TO_IVEC(8+0x20), ctHandler, ctPtr); > sysIntEnablePIC(8); > > You probably noticed that I wrote the int Line configuration register > on my card first. The reason I need to do this may be at the root of > my problem. After I boot, the int Line pci config register has not > been written (it is 0x00). I'm not sure how big a problem this is. Of > the 5 PCI devices I can see, only the ethernet controller has its int > line register set (it is set to 0x0B). Is my BIOS screwing me and > preventing this card from generation interrupts? Clearly writing this > register manually and then enabling like normal isn't doing the trick. > Anyone have any ideas? This may not be your problem at all but shouldn't the call to pciIntConnect look like this...? pciIntConnect(INUM_TO_IVEC(8 + INT_NUM_IRQ0), ctHandler, ctPtr); You may have had good reason to choose 0x20, but if INT_NUM_IRQ0 is defined to be 0x20, then why not use it? Regards, George Varndell - -- Varndell Engineering, LLC Embedded Systems Expertise http://www.varndellengineering.com Need a vxWorks END driver for NatSemi DP83815, DP83821, DP83820? Contact us. - -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- http://www.newsfeed.com The #1 Newsgroup Service in the World! - -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: What to charge for developing a BSP? Date: Fri, 04 Oct 2002 12:05:22 -0400 From: Varndell Engineering Organization: Posted Via Newsfeeds.com = SPEED+RETENTION+COMPLETION = http://www.newsfeeds.com Message-ID: <3D9DBC42.B3126493@hotmail.com> References: Ted Goldblatt wrote: > > I have been asked to develop a (semi)custom BSP. While I have done > this before, it was always as an employee, not as a consultant. So, > I've never had to worry about pricing before. For those that do this, > is there a reasonably standard price (or range) for developing BSPs? Well, this is refreshing. I get many bogus requests for quotes from people who don't know what to charge for a job and want to use my quote as a benchmark; so it is nice to see some honesty. Anyway, it's simple. You've done BSPs before so you know about long it will take you and you know what value you place on your time, making it a simple arithmetic problem. > How about for a high-end comms driver? If you don't have a clue how long this will take you, it's tough. Is it custom hardware? Does the hardware incorporate custom FPGAs (particularly messy sometimes)? You have to try to ballpark the hours based on reasonable assumptions and get the customer to buy into a T&M arrangement. If you think the hardware represents a big risk and the customer wants fixed or near-fixed pricing, don't worry about not getting the job. You don't want it and you can't afford it. > > Thanks for any info! > > ted > > (PS: if you post with a no archive header, please email me a copy, as > my only current usenet access is though Google... Thanks) Oh, well you won't see this until tomorrow then. ;) Good Luck. Regards, George Varndell - -- Varndell Engineering, LLC Embedded Systems Expertise http://www.varndellengineering.com Need a vxWorks END driver for NatSemi DP83815, DP83821, DP83820? Contact us. - -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- http://www.newsfeed.com The #1 Newsgroup Service in the World! - -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: PPC440GP EMAC transmit problems Date: Fri, 04 Oct 2002 19:29:04 +0100 From: Philipez Organization: Blaq Industries Message-ID: Hi folks, Does anybody have any experience with using the PPC440GP's inbuilt EMACs? Using the standard driver from the vxWorks 440 BSP, we can get good(ish) ttcp TCP receive performance. However, when we try to transmit (UDP or TCP), we can get only two packets out onto the wire, then transmit stalls. The last packet is always truncated to around 900-1000 bytes, with a bad TCP checksum (when using TCP). After that, we cannot even ping the board. Receiving ttcp TTCP causes the 440 to transmit ACKs, and somehow that seems OK. When the TX is dead: Using tcpdump and a vxWorks 'snoop', we can see that the 440 is still receiving packets, but does not transmit. ifShow indicates the 440 receives packets, and is trying to transmit. (The "packets sent" counter increments.) Eventually, "packets sent" stops incrementing. We can kill transmit in the same way by doing the following from Linux: ping -f -s 1100 <440GP> ping -s 4000 <440GP> In either case, TX dies and does not recover. This is the standard BSP with none of our application code loaded. This only occurs when the MMU and caches are enabled. When disabled, everything works OK (but damn slow). Any insights would be much appreciated. Thanks, Philipez. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: What to charge for developing a BSP? Date: Fri, 4 Oct 2002 21:11:07 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <3D9DBC42.B3126493@hotmail.com> pricing BSP work simply by multiplying estimated time by some pretty standard hourly fee is one way to do this, but not the best way by far. If the quote is firm, and you expect a fixed amount on completion, you must have some margins priced-in, to account for potential surprises (remember Murphy!). Most of all, if this hardware is fresh, no software has run on it (a simple rom monitor doesnt count), you need to expect hardware bugs. Also, expect that even a year after completion you will get calls from the custmer saying he found a bug, and asking you to spend time to correct it, although probably it's not one of yours, you still end-up wasting your own time. So it is nothing like being an employee our there in the wilderness, noone to protect you but yourself. I have made my share of mistakes, so I speak purely from experience. - - Leonid "Varndell Engineering" wrote in message news:3D9DBC42.B3126493@hotmail.com... > Ted Goldblatt wrote: > > > > I have been asked to develop a (semi)custom BSP. While I have done > > this before, it was always as an employee, not as a consultant. So, > > I've never had to worry about pricing before. For those that do this, > > is there a reasonably standard price (or range) for developing BSPs? > > Well, this is refreshing. I get many bogus requests for quotes from > people who don't know what to charge for a job and want to use my quote > as a benchmark; so it is nice to see some honesty. > Anyway, it's simple. You've done BSPs before so you know about long it > will take you and you know what value you place on your time, making it > a simple arithmetic problem. > > > How about for a high-end comms driver? > > If you don't have a clue how long this will take you, it's tough. > Is it custom hardware? Does the hardware incorporate custom FPGAs > (particularly messy sometimes)? You have to try to ballpark the > hours based on reasonable assumptions and get the customer to buy > into a T&M arrangement. If you think the hardware represents a big > risk and the customer wants fixed or near-fixed pricing, don't worry > about not getting the job. You don't want it and you can't afford it. > > > > > Thanks for any info! > > > > ted > > > > (PS: if you post with a no archive header, please email me a copy, as > > my only current usenet access is though Google... Thanks) > > Oh, well you won't see this until tomorrow then. ;) Good Luck. > > Regards, > George Varndell > -- > Varndell Engineering, LLC Embedded Systems Expertise > http://www.varndellengineering.com > Need a vxWorks END driver for NatSemi DP83815, DP83821, DP83820? > Contact us. > > > -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- > http://www.newsfeed.com The #1 Newsgroup Service in the World! > -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Ping semaphore problem Date: Fri, 4 Oct 2002 15:40:38 -0400 From: "Harjinder Chayra" Organization: Storm Internet Services Message-ID: Sender: terayon@gw-terayon.storm.ca On a target shell, when I issue a Ping it just returns without sending out any packets. When I debugged it (assembler only no source code) I found that it checked the variable pingSem, which is zero each time and explains why no packets are sent out. The question is why pingSem is not non-zero. Any ideas appreciated. Thanks Harjinder --------------------------- Newsgroups: comp.os.vxworks Subject: pls needed dosFsLib.c Date: 4 Oct 2002 13:19:06 -0700 From: mvi76@yahoo.com (Martin Ivanov) Organization: http://groups.google.com/ Message-ID: <5f86e92.0210041219.46db0d76@posting.google.com> Hi, I'm porting new library for vxworks but it needs dosFsLib.c file wich seems not to be in my source tree. If you have ideas how to get it please tell me. Thanx. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: pls needed dosFsLib.c Date: Sat, 5 Oct 2002 09:45:30 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <5f86e92.0210041219.46db0d76@posting.google.com> If you really need the source code for dosFs, you need to ask your WindRiver field engineer. This is not an open source product you know. But if you provide a bit more details on what you are doing, perhaps I could suggest a way to do without this source code. - ----------------------------------------------------------------------- Leonid Rosenboim Visit: http://www.masada2000.org/ Consultant Email: my first name at consultant dot com "Martin Ivanov" wrote in message news:5f86e92.0210041219.46db0d76@posting.google.com... > Hi, > I'm porting new library for vxworks but it needs dosFsLib.c file > wich seems not to be in my source tree. If you have ideas how to get > it please tell me. > Thanx. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How ro remove IP END addresses Date: Sat, 05 Oct 2002 05:18:57 -0400 From: Varndell Engineering Organization: Posted Via Newsfeeds.com = SPEED+RETENTION+COMPLETION = http://www.newsfeeds.com Message-ID: <3D9EAE80.43BDDDB1@hotmail.com> References: Kevin Garvey wrote: > > Hi, > > Once multiple IP addresses have been added to an END, using ifAddrAdd(), > does anyboby know how to subsequently remove them, one at a time, without > detaching the END from the stack. > > The routeDelete() utility removes the associated route OK but the address > remains against the interface. This is a FAQ... http://www.xs4all.nl/~borkhuis/vxworks/vxw_pt4.html#4.1-F vx5.5 now has ifAddrDelete(). Regards, George Varndell - -- Varndell Engineering, LLC Embedded Systems Expertise Need a vxWorks END driver for SiS900? Contact us. http://www.varndellengineering.com - -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- http://www.newsfeed.com The #1 Newsgroup Service in the World! - -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks Applications Engineer Date: Sat, 05 Oct 2002 17:55:45 GMT From: Russ Organization: EarthLink Inc. -- http://www.EarthLink.net Message-ID: References: <66a11529.0209090154.4a72b3c9@posting.google.com> Bamidelle wrote: > Hi, I am looking for an experienced Engineer with either an > applications engineering or a software engineering background and > strong VxWorks experience. We are a leading edge developer of: VME * > CPCI * Multibus II * PMC * Real-time OS * Custom Solutions *. We are > looking for someone highly technical with good communication skills. > You need to be authorised to work in the UK, and want to take a > position here. We will pay up to £ 35 k and offer strong package and > career path. > > Is there anyone out there who can help ? Please reply with full CV and > contact details for informal + totally confidential discussion. Perhaps in Scotland, Highlands or Islands? - -- Russ Lyttle - Not powered by ActiveX Please visit for OBD-II information and links. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Using mHdr.mNextPkt Date: Sun, 06 Oct 2002 18:27:10 GMT From: ggg+news@best.com (GGG) Organization: rest Message-ID: <2g%n9.19084$T_.441587@iad-read.news.verio.net> References: pMblk may be a chain and not have all the data contained in it. So, I think you'd better check the mNextPkt, and if available, follow the chain through. In general this is done using netMblkToBufCopy to take care of following the chain. In article , Harjinder Chayra wrote: >When a packet is passed to the END from the stack is it possible that the >mHdr.mNextPkt will not be set to NULL? > >i.e. should my code check this field and send the attached packet? > >Thanks >Harjinder > > --------------------------- Newsgroups: comp.os.vxworks Subject: ataResources/sysIntEnablePIC: undefined reference ...i just want to add dosFs2.0? Date: 6 Oct 2002 12:30:37 -0700 From: roswel_ajf@hotmail.com (roswell) Organization: http://groups.google.com/ Message-ID: <6dcae13d.0210061130.5d49f4a2@posting.google.com> Hi Folks: I just want to add dosFs2.0, no need for ATA (dont' even know what is ATA driver), so how can i diseble ATA stuff, just have the dosFs2.0?.... The dosFs2.0 intall did put the *.o into the libPPC604gnuvx.a, i had to manually copy all *.o to BSP dir, and place them in EXTRA_MODULE list. am i doing something strange?... pls help. roswel. make[2]: Leaving directory `S:\ssmc\bsp' dppc -X -N -e _sysInit -Ttext 00010000 \ -o vxWorks dataSegPad.o vxWorks.tmp ctdt.o xWorks.tmp: In function `ataDrv': xWorks.tmp(.text+0x11b26): undefined reference to `ataResources' xWorks.tmp(.text+0x11b2a): undefined reference to `ataResources' xWorks.tmp(.text+0x11d80): undefined reference to `sysIntEnablePIC' xWorks.tmp(.text+0x11e26): undefined reference to `ataTypes' xWorks.tmp(.text+0x11e2a): undefined reference to `ataTypes' xWorks.tmp: In function `ataDevCreate': xWorks.tmp(.text+0x12722): undefined reference to `ataTypes' xWorks.tmp(.text+0x12726): undefined reference to `ataTypes' xWorks.tmp: In function `ataRawio': xWorks.tmp(.text+0x12aca): undefined reference to `ataTypes' xWorks.tmp(.text+0x12ace): more undefined references to `ataTypes' follow xWorks.tmp: In function `ataPread': xWorks.tmp(.text+0x13f9c): undefined reference to `sysInWordStringRev' xWorks.tmp: In function `ataRW': xWorks.tmp(.text+0x14082): undefined reference to `ataTypes' xWorks.tmp(.text+0x14086): undefined reference to `ataTypes' xWorks.tmp(.text+0x142b8): undefined reference to `sysOutWordString' xWorks.tmp(.text+0x142d8): undefined reference to `sysOutLongString' xWorks.tmp(.text+0x14490): undefined reference to `sysInWordString' xWorks.tmp(.text+0x144b0): undefined reference to `sysInLongString' ake[1]: *** [vxWorks] Error 0x1 --------------------------- Newsgroups: comp.os.vxworks Subject: anyone using VMIC-7755 Date: Mon, 7 Oct 2002 06:55:16 +0200 From: "rat" Organization: Wirehub! Internet, The Netherlands Message-ID: <1033966227.886889@newsreader1.wirehub.nl> Just wondinrin... Is anybody using VMIC compact PCI boards, like 7755 or 7696? I'm just wondering about your expiriences regarding the stability of the BSP and others. Thanx, Renee. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: hibernation ? Date: 6 Oct 2002 23:34:07 -0700 From: s_popov@prosyst.bg (Alexander Popov) Organization: http://groups.google.com/ Message-ID: <6d1b7093.0210062234.4cd025f8@posting.google.com> References: <6d1b7093.0210040000.1df7fd48@posting.google.com> "Leonid Rosenboim" wrote in message news:... > Sasha, > > VxWorks has no support for the feature you mention. > > "Alexander Popov" wrote in message > news:6d1b7093.0210040000.1df7fd48@posting.google.com... > > Hi all, > > > > Does VxWorks support system hibernation ( suspend-to-ram, > suspend-to-disk )? > > If yes is part of the standard os/kernel or is it an extension? > > How about suspend-to-disk for a single process? Thanks Leonid, My problem is not exactly the power drain. We need to implement a sollution that will work in conditions presuming random power losses and we need to preserve the state of the running application. Maybe we should implemnt it ourselves... Can you point me up to some vxworks kernel programing guides? I'm a complete newbie with vxworks... Regards, Alexander Popov --------------------------- Newsgroups: comp.os.vxworks Subject: Re: What to charge for developing a BSP? Date: Mon, 7 Oct 2002 06:41:07 +0000 (UTC) From: Johan Borkhuis Organization: Proxim Corp. Message-ID: References: TedGoldblatt@yahoo.com (Ted Goldblatt) wrote: > I have been asked to develop a (semi)custom BSP. While I have done > this before, it was always as an employee, not as a consultant. So, > I've never had to worry about pricing before. For those that do this, > is there a reasonably standard price (or range) for developing BSPs? If possible do not give them a fixed price, but an estimate, and send the actual hours later when you are finished. Quite often you walk into HW problems or problems with certain chips, or even with VxWorks libraries. This is impossible to calculate. I never had to do this as a consultant, but several times within a project, and it is very difficult to convince a project manager that you have an uncertainty factor of 200%..... Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: hibernation ? Date: Mon, 07 Oct 2002 06:57:32 GMT From: "Gary M" Organization: AT&T Broadband Message-ID: References: <6d1b7093.0210040000.1df7fd48@posting.google.com> <6d1b7093.0210062234.4cd025f8@posting.google.com> What you describe is not a standard implementation of system hibernation or suspension, which assumes it has enough time and power to preserve the system state. Random power loss requires hardware support to detect the onset of power loss, provide a notification interrupt to the processor, and supply enough power long enough for the system state to be preserved (easier now than it was 20 years ago). Then, depending on your requirements, the system must restore operation from the saved state within a prescribed period of time after power is restored. This is straightforward if your system has some sort of battery system, more complicated if it does not, but in any event, you will probably have to do the software on your own. For information on the VxWorks kernel, check the Wind River web site and the FAQ for information. "Alexander Popov" wrote in message news:6d1b7093.0210062234.4cd025f8@posting.google.com... > "Leonid Rosenboim" wrote in message news:... > > Sasha, > > > > VxWorks has no support for the feature you mention. > > > > "Alexander Popov" wrote in message > > news:6d1b7093.0210040000.1df7fd48@posting.google.com... > > > Hi all, > > > > > > Does VxWorks support system hibernation ( suspend-to-ram, > > suspend-to-disk )? > > > If yes is part of the standard os/kernel or is it an extension? > > > How about suspend-to-disk for a single process? > > > Thanks Leonid, > > My problem is not exactly the power drain. > We need to implement a sollution that will work in conditions > presuming random power losses and we need to preserve the state of the > running application. > > Maybe we should implemnt it ourselves... > Can you point me up to some vxworks kernel programing guides? I'm a > complete newbie with vxworks... > > Regards, > Alexander Popov --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Mon, 07 Oct 2002 11:10:45 +0200 From: Tim Sinkins Organization: Alcatel Kommunikations Elektronik Message-ID: <3DA14F95.A74993DC@alcatel-ke.de> References: <3a5ff53f.0210020748.3643d45a@posting.google.com> Reply-To: timothy.sinkins@alcatel-ke.de brian wrote: > > "Leonid Rosenboim" wrote in message news:... > > Now, don't take me wrong here, I never said there eint > > a reason to throw out WindRiver, but cost eint that reason, son! > > > > I know that HP has replaced VxWorks with something else > > in their printers, but can you really prove that cost was the > > reason ? > > Actually I dont know for sure. But I can find out. Do you know for > sure that it wasnt the reason? > > > > > You know there are a dozen of other reasons that could > > be the cause: like lack of support, too long to fix bugs, > > no source code availability, ... > > > > I mean, if you're pissed at WRS - you're not alone! > > Go ahead, throw them out, join the masses, > > just don't tell me it's for cost! > > Well dad, > > Actually, Im not a bit pissed at WRS. In fact normally I take the same > side of the equation that you do. Cost savings due to time to market > etc etc. And in fact that argument is usually winnable. And in fact > I would recommend vxWorks for new development any day. And do it all > the time. > > However in this case, the product is built and working. The question > is whether I can run a parallel development of swapping out the > kernel. > > The cost? Suppose I wanna ship 4 million units. Suppose that the best > that I can get is 25c /unit from WRS. Suppose that the total royalty > cost in therefore $1Mill. Jeez, do WRS really not have a royalty free option, IME much of the competition does, why would anyone even think of putting it in a million units product with that sort of pricing, and no sorce code either. What is so great about this OS that makes anyone even consider using it on those terms? > Now. If I can do this in 3-6mos, Take it from me, you can't. Are you talking about rolling your own or just switching to another RTOS in this time? Either way you still can't do it. The latter took us about 12 man months, not including the item below (similar reason as you, old OS not ported to new processor, with hindsight, it would have been easier to write a port, but I think some of the reason the engineers made the decision that they did was a "career development" one) > it costs me 100k. Are there maintenance costs? There are bound to be little differences in the interface that you will be debugging for months. > In fact very little if the code is not buggy. If I > can use a kernel that is already debugged (just not ported) then I'm > way ahead of the game on the bugs. Personally I think 1 month of > serious testing would eliminate nearly every bug. Am I going to > port to another processor after this if I need to get more mips etc > etc later? Probably not. But if I do I can always go back to WRS and > begin again. > > Aside from calling me 'son' and disparaging the whole idea .. do you > have any suggestions or can you point out something specific that I > have missed? Please do as I am looking for holes in this argument. When you say the alternatives are 'expensive' in up-front costs what do you mean? What is expensive for you? Tim > > Brian > > > > > "Brian" wrote in message > > news:3a5ff53f.0210020748.3643d45a@posting.google.com... > > > "Leonid Rosenboim" wrote in message > > news:... > > > > Are you serious ? > > > > > > > > To replace the OS as a cost cutting measure is one peice of BS > > > > I have heard too many times, and allways proved wrong. > > > > > > > > > > HP just did it with their some of their printers. Have you proved > > > them wrong yet? It's very possible that WRS just wont negotiate down > > > since the technical risk is deemed high. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: bringingup vxworks on a board with two 750cxe processors Date: Mon, 7 Oct 2002 11:14:20 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <3D9AC536.8335091C@hotmail.com> Reply-To: "Michael Lawnick" Sender: mlawnick@p5088f9f9.dip.t-dialin.net Hi kiran, > second approch: in the ISR MMU disabled. 750 interrupt bot the > address and data translation is disabled.! But re-enabling it should bring you back to your CPU's context!? Regards, Michael "kiran Mandava" schrieb im Newsbeitrag news:ea368e49.0210020939.7581dd9f@posting.google.com... > Hi michael, > First approch: i currently think that is the only way..but looking > for inputs > second approch: in the ISR MMU disabled. 750 interrupt bot the > address and data translation is disabled.! > > > kiran > > > >"Michael Lawnick" wrote in message news:$1@snoopy.bndlg.de>... > > Hi kiran, > > > > I'm pretty sure you will have to replace the code at 0x500 by a piece of > > your own, where you detect the CPU and branch either to lower or higher CPU- > > image. > > Another approach: > > For the higher CPU redefine the MMU-entry of the lowest page in exception > > mode and put an address-translation into it . > > > > Regards, > > Michael > > > > "kiran Mandava" schrieb im Newsbeitrag > > news:ea368e49.0210020608.14976189@posting.google.com... > > > Yes each of My CPU have their own images and they are located at > > > different places in the memory. > > > > > > lets sayfirst CPU 0x00000000 to 0x04000000 and the sencond one from > > > 0x04000000 to 0x08000000.(image loading starts at an offset 0x20000) > > > > > > ok when any of the processors get an external interrupt they jump to > > > 0x500. assume that second CPU didn't put any code at 0x500, but in the > > > first CPU's iSR i would jump to different location depending on the > > > CPU number. the first thing it does in the code at 0x500 is it tries > > > to save the context at a location interrupt_stack_base(this is in > > > reference to a variable vxIntStackBase or something), as the code is > > > put in by the first CPU the address belongs to the first CPU, (it > > > doesn't use the r1 to store the context on stack, so that the > > > interrupt stack would be totally different from the current r1), so if > > > there is an interrupt at the same time there is a problem here.! > > > > > > > > > intEnt does it enable the Real mode. i wasn't sure, if that is the > > > first thing it does probably that might help me solving this probem.! > > > > > > > > > thanks, > > > kiran > > > > > > >Varndell Engineering wrote in message > > news:<3D9AC536.8335091C@hotmail.com>... > > > > kiran Mandava wrote: > > > > > > > > > > Hello all, > > > > > > > > > > Since the interrupt vector base is not relocatable on powrePc 750 > > > > > how should i handle the interrupts on both the processors. if there is > > > > > an external interrupt on any of the processors it would jump to > > > > > location 0x500 where the ISR corresponding to the First CPU is > > > > > located, so there is a problem handling the interrupts from both the > > > > > processors as both of them try to save the interrupt conetext to the > > > > > same location..did anybody came across such problem and any sort of > > > > > feedback would be really helpful > > > > > > > > > > CPC 710 is the bridging device and both the processors use the same > > > > > SDRAM first half the first CPU and the otherhalf the other CPU > > > > > > > > Assuming that each cpu is executing an image that has been linked to > > > > a different address, or each image executes within its own virtual > > > > space, I think you can do this. > > > > When the 750 takes an interrupt, it will be executing in real address > > > > mode for a short time. It will call intEnt() to restore MMU translation, > > > > if translation was enabled when the interrupt occurred, save context > > > > on the active stack, and return from intEnt to call whatever C function > > > > was installed by excConnect(). Right now, you have 2 images, each > > > > installing a C handler (different addresses) into the same physical > > > > address. > > > > Without working out all the details for you, I think you can install > > > > a common handler into the vector. This handler would need to decide > > > > which processor was executing it and call the interrupt handler for > > > > that processor. Obviously, this would require some hacking of the > > > > interrupt controller code -- the code that currently takes over > > > > the external interrupt vector. Given that, when the user handler > > > > is called (after return from intEnt()) virtual address translation > > > > is active, this might even be easy. > > > > The predictable disclaimer; I've never tried this before and I > > > > really haven't thought it out thoroughly. > > > > > > > > Regards, > > > > George Varndell > > > > -- > > > > Varndell Engineering, LLC Embedded Systems Expertise > > > > Need an vxWorks END driver for Sis900 or DP83815? Contact us. > > > > http://www.varndellengineering.com > > > > > > > > > > > > -----------== Posted via Newsfeed.Com - Uncensored Usenet News > > ==---------- > > > > http://www.newsfeed.com The #1 Newsgroup Service in the World! > > > > -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers > > =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Mon, 7 Oct 2002 11:34:19 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <3a5ff53f.0210020748.3643d45a@posting.google.com> Reply-To: "Leonid Rosenboim" Brian, could not resist, must follow on this once more (last time, promise): You will probably spend in the era of $300k to $600k on the switch of VxWorks for a semi-custom OS - using something unknown will be plenty of surprises. And that will save you 25 cents off the COGS. Alternatively, you can hire a consultant for $20k-$50k to review your design, and recommend alternative ways to reduce COGS, and realistically you can expect over $2.5 per unit savings, ten times what you would save on changing the OS. There are people who's middl ename is "cost reduction", but that won't give you the sense of job security that you will get by switching to a semi-custom OS. - ----------------------------------------------------------------------- Leonid Rosenboim Visit: http://www.masada2000.org/ Consultant Email: my first name at consultant dot com "brian" wrote in message news:d2f67bd8.0210021412.92afe08@posting.google.com... > "Leonid Rosenboim" wrote in message news:... > > Now, don't take me wrong here, I never said there eint > > a reason to throw out WindRiver, but cost eint that reason, son! > > > > I know that HP has replaced VxWorks with something else > > in their printers, but can you really prove that cost was the > > reason ? > > Actually I dont know for sure. But I can find out. Do you know for > sure that it wasnt the reason? > > > > > You know there are a dozen of other reasons that could > > be the cause: like lack of support, too long to fix bugs, > > no source code availability, ... > > > > I mean, if you're pissed at WRS - you're not alone! > > Go ahead, throw them out, join the masses, > > just don't tell me it's for cost! > > > Well dad, > > Actually, Im not a bit pissed at WRS. In fact normally I take the same > side of the equation that you do. Cost savings due to time to market > etc etc. And in fact that argument is usually winnable. And in fact > I would recommend vxWorks for new development any day. And do it all > the time. > > However in this case, the product is built and working. The question > is whether I can run a parallel development of swapping out the > kernel. > > The cost? Suppose I wanna ship 4 million units. Suppose that the best > that I can get is 25c /unit from WRS. Suppose that the total royalty > cost in therefore $1Mill. > > Now. If I can do this in 3-6mos, it costs me 100k. Are there > maintenance costs? In fact very little if the code is not buggy. If I > can use a kernel that is already debugged (just not ported) then I'm > way ahead of the game on the bugs. Personally I think 1 month of > serious testing would eliminate nearly every bug. Am I going to > port to another processor after this if I need to get more mips etc > etc later? Probably not. But if I do I can always go back to WRS and > begin again. > > Aside from calling me 'son' and disparaging the whole idea .. do you > have any suggestions or can you point out something specific that I > have missed? Please do as I am looking for holes in this argument. > > Brian > > > > > > > > > "Brian" wrote in message > > news:3a5ff53f.0210020748.3643d45a@posting.google.com... > > > "Leonid Rosenboim" wrote in message > > news:... > > > > Are you serious ? > > > > > > > > To replace the OS as a cost cutting measure is one peice of BS > > > > I have heard too many times, and allways proved wrong. > > > > > > > > > > HP just did it with their some of their printers. Have you proved > > > them wrong yet? It's very possible that WRS just wont negotiate down > > > since the technical risk is deemed high. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: hibernation ? Date: Mon, 7 Oct 2002 11:43:26 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <6d1b7093.0210040000.1df7fd48@posting.google.com> <6d1b7093.0210062234.4cd025f8@posting.google.com> Reply-To: "Leonid Rosenboim" Sasha, kernel is really not your real concern for the power-loss issue, device drivers and peripherals are. It all depends on what is your application, what kind of non-volatile memory is available, and how do you exactly define "state information" that you want to keep accross power fails. Supporting Hibernation for the entire system is difficult, and might prove to be too expensive solution for your needs. - ----------------------------------------------------------------------- Leonid Rosenboim Visit: http://www.masada2000.org/ Consultant Email: my first name at consultant dot com "Alexander Popov" wrote in message news:6d1b7093.0210062234.4cd025f8@posting.google.com... > "Leonid Rosenboim" wrote in message news:... > > Sasha, > > > > VxWorks has no support for the feature you mention. > > > > "Alexander Popov" wrote in message > > news:6d1b7093.0210040000.1df7fd48@posting.google.com... > > > Hi all, > > > > > > Does VxWorks support system hibernation ( suspend-to-ram, > > suspend-to-disk )? > > > If yes is part of the standard os/kernel or is it an extension? > > > How about suspend-to-disk for a single process? > > > Thanks Leonid, > > My problem is not exactly the power drain. > We need to implement a sollution that will work in conditions > presuming random power losses and we need to preserve the state of the > running application. > > Maybe we should implemnt it ourselves... > Can you point me up to some vxworks kernel programing guides? I'm a > complete newbie with vxworks... > > Regards, > Alexander Popov --------------------------- Newsgroups: comp.os.vxworks Subject: Re: can't load .out files with dosfs2 Date: Mon, 7 Oct 2002 11:46:29 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <15cdb0fb.0210021822.572bd28c@posting.google.com> Reply-To: "Michael Lawnick" Sender: mlawnick@p5088f9f9.dip.t-dialin.net Hi Tom, just to be sure ... You are working under Target-Shell (i.e. serial or TELNET) not WindShell (aka Host-Shell, in conjunction with target server) ? If yes (Target-Shell): ??? If no: try - ->@ld schrieb im Newsbeitrag news:15cdb0fb.0210021822.572bd28c@posting.google.com... > This is a repost of a previous message I had posted with new > information included. > > I have been working on getting a strong arm based motherboard to run > vxworks. One problem I have run into is that I can't load any .out > files off of my TFFS (True Flash File System). When I try to it just > gives me an error saying "error reading file (errno = 0x3d0001).". I > looked up the errno and it resolved to S_objLib_OBJ_ID_ERROR , but > this is generated from something at a low level that I don't have the > source code to. I recently tried stepping through the assembler code > to see where this error was coming from and it turns out errno is not > set until the cleanup code runs after fioRead fails, so this is errno > is irrelevant. > > Every other operation with the TFFS seems to work just fine. I can > create, delete, read, write, and rename files. As an example I tried > copying an out file off my FTP server onto flash and loading it, but > it failed. Then after that I copied the file off of the TFFS to a new > file name on my FTP server and then loaded it from that file (over the > network) and it worked fine. > > -- start dump -- > > -> cd "/tffs0/" > value = 0 = 0x0 > -> ll > -rwxrwxrwx 1 0 0 10064 Jan 1 1980 test.out > value = 0 = 0x0 > -> ld < test.out > ld error: error reading file (errno = 0x3d0001). > value = 0 = 0x0 > -> copy "/tffs0/test.out", "host:newtest.out" > Copy OK: 10064 bytes copied > value = 0 = 0x0 > -> ld < host:newtest.out > value = 12531032 = 0xbf3558 > -> > > -- end dump -- > > > After these tests I thought the problem might have been size related, > so I tried to create a minimal "hello world" out file, but that didn't > work either. > > As a work around I thought I would try setting up a runscript that > creates a ram drive and then copies each file to the ramdrive, loads > from there, deletes the file, and then continues on to the next one > until it is done. Unfortunatelly this failed with the same error > message that I got when I loaded off of a TFFS. The only thing I can > think of that the ramdrive and TFFS have in common that a ftp file > system doesn't is that they use dosFS (version 2). > > Has anyone else had any problems like this? I'm running > VxWorks 5.4.2 with Tornado 2.1.1. > > > Best Regards, > > Tom --------------------------- Newsgroups: comp.os.vxworks Subject: Re: hibernation ? Date: 7 Oct 2002 03:02:58 -0700 From: s_popov@prosyst.bg (Alexander Popov) Organization: http://groups.google.com/ Message-ID: <6d1b7093.0210070202.24255bda@posting.google.com> References: <6d1b7093.0210040000.1df7fd48@posting.google.com> >Sasha, > >kernel is really not your real concern for the power-loss issue, >device drivers and peripherals are. As I said I know nothing about vxWorks, but on linux systems (mainly done for the clustering technologies) a process state (it's memory + the hardware state ) is saved in order to be tranfered to another machine/CPU - I'll use that to implement what I need for our application (and this requires support in the kernel for checkpointing the process "status")... I know that vxWorks uses gdb - I was thinking of using something like gdb's core files to suspend-to-disk/flash the process... >It all depends on what is your application, what kind of >non-volatile memory is available, and how do you >exactly define "state information" that you want to >keep accross power fails. > >Supporting Hibernation for the entire system >is difficult, and might prove to be too expensive >solution for your needs. I agree completely - now I'm focused on hibernating a single process (this should be much cheaper concerning the non-volatile memory)... I'll check how the checkpointing is done on linux (I think there are projects that don't need kernel support and recompiling of the binaries) and then I'll try to implement the the same technology on vxWorks and QNX - But I'm not sure that it's posiible - that's why I wanted to learn more about the kernel of vxWorks... --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Strange Interrupt log msg Date: Mon, 7 Oct 2002 12:02:57 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <290ade1c.0210031725.249f87a6@posting.google.com> Reply-To: "Michael Lawnick" Sender: mlawnick@p5088f9f9.dip.t-dialin.net Proof your concept of task- and interrupt-synchronisation: you've got a nasty case, where you rely on a given order of execution that is not true on different optimization settings. "tzeghuan" schrieb im Newsbeitrag news:290ade1c.0210031725.249f87a6@posting.google.com... > Hi, > > I have this strange interrupt log msg : "interrupt: Bu Read Ram Busy!" > whenever I compiled using Tornado with option -O2. > But when compiled using -g, I don't get the msg. > > I am using Tornado 2 and vxWorks 5.4 > > What could be the problems? > Any kind of helps are appreciated. > > Thanks in advance. > > rdgs > tzeghuan --------------------------- Newsgroups: comp.os.vxworks Subject: Re: hibernation ? Date: Mon, 7 Oct 2002 12:17:41 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <6d1b7093.0210040000.1df7fd48@posting.google.com> <6d1b7093.0210070202.24255bda@posting.google.com> Reply-To: "Leonid Rosenboim" VxWorks is nothing like Unix - There is no such thing as a process, there are tasks, and tasks do not own all of their resources, therefore restoring the task state means nil, because the task could be using anything at all that could be anywhere in the system's memory (heap chunks, routines, ...) You could in fact draw a parallel: VxWorks task <-> Unix thread, while the entire VxWorks system is somewhat akin to a process. Perhaps you should start by simply reading the VxWorks programmers guide (available for download from windriver.com). - -- - ----------------------------------------------------------------------- Leonid Rosenboim Visit: http://www.masada2000.org/ Consultant Email: my first name at consultant dot com "Alexander Popov" wrote in message news:6d1b7093.0210070202.24255bda@posting.google.com... > >Sasha, > > > >kernel is really not your real concern for the power-loss issue, > >device drivers and peripherals are. > > As I said I know nothing about vxWorks, but on linux systems (mainly > done for the clustering technologies) a process state (it's memory + > the hardware state ) is saved in order to be tranfered to another > machine/CPU - I'll use that to implement what I need for our > application (and this requires support in the kernel for checkpointing > the process "status")... > I know that vxWorks uses gdb - I was thinking of using something like > gdb's core files to suspend-to-disk/flash the process... > > >It all depends on what is your application, what kind of > >non-volatile memory is available, and how do you > >exactly define "state information" that you want to > >keep accross power fails. > > > >Supporting Hibernation for the entire system > >is difficult, and might prove to be too expensive > >solution for your needs. > > I agree completely - now I'm focused on hibernating a single process > (this should be much cheaper concerning the non-volatile memory)... > I'll check how the checkpointing is done on linux (I think there are > projects that don't need kernel support and recompiling of the > binaries) and then I'll try to implement the the same technology on > vxWorks and QNX - But I'm not sure that it's posiible - that's why I > wanted to learn more about the kernel of vxWorks... --------------------------- Newsgroups: comp.os.vxworks Subject: Source file dependencies in Tornado 2.0 Date: Mon, 7 Oct 2002 12:56:06 +0200 From: "Carsten Olsen" Organization: TDC Internet Message-ID: <3da16847$0$91466$edfadb0f@dspool01.news.tele.dk> Hi, I have a problem with file-dependencies among different directories in Tornado 2.0. When source-files are placed in different directories, the dependencies fail and the project cannot be compiled. Example directory-file structure: Classes/ class1.cpp class1.h Project/ project.wsp project.wpj source/ main.cpp In main.cpp I include "class1.h" with '#include "class1.h"' When I compile, I get the error-mesage: "make: *** No rule to make target `class1.h', needed by `main.o'. Stop" The problem is, that the dependencies at the end of the auto-generated makefile are wrong. How do I tell Tornado, that header-files are placed in another directory? I do not want to make includes like '#include "../../classes/class1.h"' The information should belong to the project-files, not the individual source-files. Does anyone know how to handle this situation? Thanks. Kind regards, Carsten Olsen --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Mon, 7 Oct 2002 10:59:54 +0000 (UTC) From: Johan Borkhuis Organization: Proxim Corp. Message-ID: References: <3a5ff53f.0210020748.3643d45a@posting.google.com> <3DA14F95.A74993DC@alcatel-ke.de> Tim Sinkins wrote: > Jeez, do WRS really not have a royalty free option, IME much of the > competition does, why would anyone even think of putting it in a > million units product with that sort of pricing, and no sorce code > either. What is so great about this OS that makes anyone even > consider using it on those terms? Yes, they have a royalty free option. Just talk to your sales contact, and negotiate a deal with him (end of a quarter or end of year might be a good time to do this :-). But it will be expensive. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Source file dependencies in Tornado 2.0 Date: Mon, 7 Oct 2002 13:25:04 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <3da16847$0$91466$edfadb0f@dspool01.news.tele.dk> Reply-To: "Michael Lawnick" Sender: mlawnick@p5088f9f9.dip.t-dialin.net Hi Carsten, this is a FAQ, so search in e.g. google or look into http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html Regards, Michael "Carsten Olsen" schrieb im Newsbeitrag news:3da16847$0$91466$edfadb0f@dspool01.news.tele.dk... > Hi, > > I have a problem with file-dependencies among different directories in > Tornado 2.0. When source-files are placed in different directories, > the dependencies fail and the project cannot be compiled. > > Example directory-file structure: > Classes/ > class1.cpp > class1.h > Project/ > project.wsp > project.wpj > source/ > main.cpp > > In main.cpp I include "class1.h" with '#include "class1.h"' > When I compile, I get the error-mesage: > "make: *** No rule to make target `class1.h', needed by `main.o'. Stop" > > The problem is, that the dependencies at the end of the auto-generated > makefile are wrong. > How do I tell Tornado, that header-files are placed in another directory? > I do not want to make includes like '#include "../../classes/class1.h"' > The information should belong to the project-files, not the individual > source-files. > > Does anyone know how to handle this situation? > > Thanks. > > Kind regards, > Carsten Olsen > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Mon, 07 Oct 2002 14:02:38 +0200 From: Tim Sinkins Organization: Alcatel Kommunikations Elektronik Message-ID: <3DA177DE.F76FF84B@alcatel-ke.de> References: <3a5ff53f.0210020748.3643d45a@posting.google.com> <3DA14F95.A74993DC@alcatel-ke.de> Reply-To: timothy.sinkins@alcatel-ke.de Johan Borkhuis wrote: > > Tim Sinkins wrote: > > > Jeez, do WRS really not have a royalty free option, IME much of the > > competition does, why would anyone even think of putting it in a > > million units product with that sort of pricing, and no sorce code > > either. What is so great about this OS that makes anyone even > > consider using it on those terms? > > Yes, they have a royalty free option. Just talk to your sales contact, and > negotiate a deal with him (end of a quarter or end of year might be a good > time to do this :-). But it will be expensive. Thank's, but what does expensive mean? It's got to be a lot less than 1M? Tim > > Kind regards, > Johan > > -- > o o o o o o o . . . _____________________________ > o _____ || Johan Borkhuis | > .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | > >(________|__|_[_________]_|__________________________| > _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` > === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Mon, 7 Oct 2002 12:55:42 +0000 (UTC) From: Johan Borkhuis Organization: Proxim Corp. Message-ID: References: <3a5ff53f.0210020748.3643d45a@posting.google.com> <3DA14F95.A74993DC@alcatel-ke.de> <3DA177DE.F76FF84B@alcatel-ke.de> Tim Sinkins wrote: >> Yes, they have a royalty free option. Just talk to your sales >> contact, and negotiate a deal with him (end of a quarter or end of >> year might be a good time to do this :-). But it will be expensive. > > Thank's, but what does expensive mean? It's got to be a lot less than > 1M? That largely depends on your purchaser :-) But as far as I know this should be below 1M. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Mon, 07 Oct 2002 15:08:48 +0200 From: Tim Sinkins Organization: Alcatel Kommunikations Elektronik Message-ID: <3DA18760.DF629292@alcatel-ke.de> References: <3a5ff53f.0210020748.3643d45a@posting.google.com> <3DA14F95.A74993DC@alcatel-ke.de> <3DA177DE.F76FF84B@alcatel-ke.de> Reply-To: timothy.sinkins@alcatel-ke.de Johan Borkhuis wrote: > > Tim Sinkins wrote: > >> Yes, they have a royalty free option. Just talk to your sales > >> contact, and negotiate a deal with him (end of a quarter or end of > >> year might be a good time to do this :-). But it will be expensive. > > > > Thank's, but what does expensive mean? It's got to be a lot less than > > 1M? > > That largely depends on your purchaser :-) But as far as I know this should > be below 1M. OK, I'm just wondering what people think of as 'expensive' in this context. The only figures I have seen for this (for other OS) have been in the 20-50K range, which whilst not affordable by the one-man-band is not something that I would call exspensive in a corporate situation. Tim > > Kind regards, > Johan > > -- > o o o o o o o . . . _____________________________ > o _____ || Johan Borkhuis | > .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | > >(________|__|_[_________]_|__________________________| > _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` > === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Link Bootable with Downloadable project Date: 7 Oct 2002 06:40:00 -0700 From: yaronk@rafael.co.il (Yaron) Organization: http://groups.google.com/ Message-ID: Hi All I Have a Bootable projects and Downloadable projects which works quite well. Now I'm trying to link them together to one Bootable project. I read in the "Tornado Release Notes 2.0" how to link a Bootable project with Downloadable one, but I don't understand how to create a Partial Link file of the Downloadable project. Thanks for any info! Yaron --------------------------- Newsgroups: comp.os.vxworks Subject: ANNOUNCE: Embeddable Kermit File Transfer Protocol Date: 7 Oct 2002 10:53:09 -0400 From: fdc@columbia.edu (Frank da Cruz) Organization: Columbia University Message-ID: In response to increasingly frequent requests, I'm happy to announce the availability of an embeddable Kermit protocol. This is a tiny but modern implementation of Kermit protocol (send and receive only; no server) written in ANSI C suitable for embedding and realtime applications as well as for creation of DLLs. It is the real thing, written at the Kermit Project at Columbia University, the original developers of the Kermit protocol. More about Embedded Kermit (EK) here: http://www.columbia.edu/kermit/ek.html This is not free software and it is not online for public viewing. As you may know, the Kermit Project is funded entirely by revenue it generates. The purpose of EK is to generate badly needed revenue during the economic slump, and to furnish people who need a compact and embeddable version of Kermit protocol with -- at last -- an implementation "from the source" that works with other correct Kermit implementations such as C-Kermit, G-Kermit, MS-DOS Kermit, and Kermit 95, and that is fully supported by the developers. If you're interested in looking at it, please send an inquiry by email to: kermit@columbia.edu Thanks. Frank da Cruz The Kermit Project Columbia University New York NY USA http://www.columbia.edu/kermit/ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: hibernation ? Date: 07 Oct 2002 10:36:44 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: References: <6d1b7093.0210040000.1df7fd48@posting.google.com> <6d1b7093.0210062234.4cd025f8@posting.google.com> Sender: bpringle@DeadDuck "Alexander Popov" wrote... >> Does VxWorks support system hibernation ( suspend-to-ram, >> suspend-to-disk )? If yes is part of the standard os/kernel or is >> it an extension? How about suspend-to-disk for a single process? "Leonid Rosenboim" wrote... > VxWorks has no support for the feature you mention. "AP" == Alexander Popov write, AP> My problem is not exactly the power drain. We need to implement AP> a sollution that will work in conditions presuming random power AP> losses and we need to preserve the state of the running AP> application. You aren't listening. Your problem is power drain. The problem is all about power management. The first thing that must be done is to remove heavy power drain peripherals such as radios, printers, bar code scanners, LCDs, backlights, network transformers, etc. This can be done by hardware which isolates the CPU and memory supply from the hardware. Leonid (or someone) mentioned using battery backed SRAM. You can also use "Super-Caps" with .5 to 4 Farad values that store quite a bit of power to run a low power draw CPU like an ARM (SuperH or MIPS is also good. I don't remember which). You do not require any support from the OS to do this. It is next to impossible to store RAM to Non-volatile in a short period of time. You might do this with a wall-supplied unit and a UPS type battery backup. If you have a method of preserving the memory, such as SRAM with a super-CAP supplying "stay alive" current, you just need to save the processor registers and provide driver PowerOff(), and PowerOn() functions. Strange things can also happen to the vxWorks kernel if your timer tick jumps by the hibernating amount. It does sound like you have a sever misconception. "Suspend to disk", would not be a concept that is too familiar to embedded devices. If you are making a low volume VME product for some control system, this might be feasible. You will need a large battery. Even storing memory to a solid state device like flash takes quite a bit of power. What ever the case, this is a power drain problem. Unless of course, you have software that can defy physics? hth, Bill Pringlemeir. - -- Have you ever seen an infant make annoying phone calls to your neighbor five times in a row? Or wondered if it was possible to have a subway throw sparks at you? You will. And the company that will bring it to you: AT&T. --------------------------- Newsgroups: comp.os.vxworks Subject: Slow boot on powerpc with VxWorks 5.4 Date: Mon, 07 Oct 2002 19:48:29 +0200 From: =?iso-8859-1?Q?St=E9phane?= GERVAIS Organization: Dassault Data Services Message-ID: <3DA1C8ED.ED764FF1@free.fr> Reply-To: s-gervais@dassault-data-services.fr Il s'agit d'un message multivolet au format MIME. - --------------E8B80CA9FB54CA417663346B Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Hello, I've got a VME system which works very well with VxWorks 5.3.1. Since I put VxWorks 5.4 on cards, I've got problems while my system is booting. All ppc card need more than 3 minutes to load the boot image ! (No problem with 68k cards). I've tried to boot only one ppc card: the problem is always here. But if I connect the card in another switch (all card are connected in one switch), the problem disappears !!! I suppose it is a network problem (TCP/IP conflicts ?). What are the differences between VxWorks 5.3.1 and VxWorks 5.4 BSP for mv2604 BSP to make a such behavior ? How can I solve this problem ? Thanks for your solutions... Stephane. - --------------E8B80CA9FB54CA417663346B Content-Type: text/x-vcard; charset=us-ascii; name="sgervais.dds.vcf" Content-Transfer-Encoding: 7bit Content-Description: Carte pour Stéphane GERVAIS Content-Disposition: attachment; filename="sgervais.dds.vcf" begin:vcard n:GERVAIS;Stéphane tel;fax:04 92 38 82 49 tel;work:04 92 38 82 41 x-mozilla-html:FALSE org:Dassault Data Services;Marseille / Sophia version:2.1 email;internet:s-gervais@dassault-data-services.fr adr;quoted-printable:;;2791 Route de St Bernard=0D=0ALes Moulins 2;VALLAURIS;;06220; fn:Stéphane GERVAIS end:vcard - --------------E8B80CA9FB54CA417663346B-- --------------------------- Newsgroups: comp.os.vxworks Subject: tffs/d_cache Date: Tue, 08 Oct 2002 01:43:12 GMT From: "verizon" Message-ID: I read *somewhere* that if you are using TFFS that the you must undef the enable_d_cache in the project ? Running t2 with all the latest tffs/dosfs upgrads from WRS, but still have corruption of data files. Any answer? Thanks ! --------------------------- Newsgroups: comp.os.vxworks Subject: Routing problem Date: 7 Oct 2002 23:18:29 -0700 From: Gunter.Ritter@tenovis.com (Gunter Ritter) Organization: http://groups.google.com/ Message-ID: Hi vxworkers, after solving my problem with the second end driver, the next problem occures with the routing between the networks. The environment is as follows: The first network interface has the IP address 192.1.1.141 and the subnet mask is set to 255.255.255.0. This interface is connected to a physical ethernet LAN. The second network interface has the IP address 192.1.2.1 and the subnet mask is set to 255.255.255.0. This is only a virtual network interface, not connected to a physial ethernet. This interface shall provide access to the pyhsical network 192.1.1.0 for clients. These clients are dialing in to the network 192.1.2.0 using PPP and IPCP. The IP address of a client is for example 192.1.2.55. The routing between the networks shall be done by vxworks. PPP and IPCP is not done by vxworks. For this I use a different stack. It looks like this: --------------- ------------ 192.1.1.141 | |192.1.2.1 192.1.2.55 | | - -------------| vx target |-------------------------| W2K client | | | | | ---------------- ------------ Now the problem: The W2K client is dialed in. Ping from the client 192.1.2.55 to 192.1.2.1 and 192.1.1.141 does work. I get a reply. Ping to a PC in the network 192.1.1.0 (PC has IP 192.1.1.1) does not work. I do not get a reply. Ping from the PC with IP 192.1.1.1 to the client 192.1.2.55 does work when I add the target 192.1.2.141 as the default gateway in the LAN configuration of the PC. So, what am I doing wrong? What do I have to do for a working routing with this configuration? Gunter --------------------------- Newsgroups: comp.os.vxworks Subject: Re: x86 pci interrupt sharing Date: Tue, 8 Oct 2002 10:06:53 +0200 From: "Khac Tri VU" Organization: Planet Internet Message-ID: References: Hi, Matthew I think that you need to save the address of the current ISR linked to the IRQ somewhere before registering the second ISR(check intVecGet). In the second ISR, you will have to call the first ISR if needed. logMsg() in a ISR is not a good practise since ISR may have been executed, but nothing is printed out. Cheers, Tri. "Matthew Peltzer" wrote in message news:antf51$pr2$1@overload.lbl.gov... > Greetings, > > I am developing a device driver for custom PCI hardware under vxWorks > 5.4/Tornado 2.0.2, and am experiencing difficulties in handling interrupts. > > I have no problem when the two devices are on their own interrupts, and I > have no problem when using a shared interrupt independently. I am not able > to have two threads at the same time listen to the same interrupt. > > According to the manual page for pciIntConnect(), I should not have to care > about shared interrupts: > > DESCRIPTION > This routine connects an interrupt handler to the PCI > interrupt line(A - D). Link list is created if multiple > handlers are assigned to the single PCI interrupt. > > To me this says that every time I call pciIntConnect(), my function and > parameter will be added to a list that when the interrupt vector is > triggered, each of my functions will be called with the appropriate > parameter. Unfortunately, this is not what I am experiencing, although I am > not 100% sure of what I am experiencing. It seems as though the second call > to pciIntConnect() effectively wipes out the first parameter, for as soon as > an interrupt is generated on this irq, my system deadlocks. I can only > guess that this is because I am not turning off the interrupt on the correct > card because my parameter is not what I expect it to be. > > My solution to this is to do my own isr chaining. I call pciIntConnect() > and sysIntEnablePIC() only when no other connection has been made for this > irq. pciIntDisconnect() and sysIntDisablePIC() are called only when no > other device is listening to the irq. This setup has the same flaw as the > previous: vxWorks reboots when two tasks are run that listen to a separate > PCI card, yet share the same irq. > > I do not know if my callbacks are being called. the first operation in my > callback is a logMsg(), but I am not seeing this. I also have some > logMsg()s to tell me if each interrupt is being caught or ignored, which is > decided by looking at the interrupt line on the hardware device as specified > in the parameter passed into the isr by the system. I have yet to see an > interrupt be ignored. > > > This is the code I am using to enable interrupts: > > pciIntConnect(INUM_TO_IVEC((pci->irq + 0x20)), myDummyISR, list[pci->irq]); > sysIntEnablePIC(pci->irq); > > > 'pci' is a structure that among other things contains the irq number as read > from the pci configuration space with: > > pciConfigInByte(pci->busNum, pci->devNum, pci->fncNum, 0x3c, &(pci->irq)); > > > Is there anything that I am missing for irq sharing to work in vxWorks? > Has anyone had any luck/mis-luck with similar methods? > > > Matthew Peltzer > goose23@spu.edu > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Routing problem Date: 08 Oct 2002 11:26:53 +0100 From: Simon Wright Organization: Alenia Marconi Systems, ISD, Farlington Message-ID: References: Sender: sjw@galadriel.frlngtn.gecm.com Gunter.Ritter@tenovis.com (Gunter Ritter) writes: > The W2K client is dialed in. Ping from the client 192.1.2.55 to > 192.1.2.1 and 192.1.1.141 does work. I get a reply. Ping to a PC in > the network 192.1.1.0 (PC has IP 192.1.1.1) does not work. I do not > get a reply. > > Ping from the PC with IP 192.1.1.1 to the client 192.1.2.55 does > work when I add the target 192.1.2.141 as the default gateway in the > LAN configuration of the PC. Do you still not get a reply 192.1.2.55 -> 192.1.1.1 -> 192.1.2.55 after you have set this default gateway on 192.1.1.1? because 192.1.1.1 needs to know how to send the ICMP packets back. --------------------------- Newsgroups: comp.os.vxworks Subject: finding interrupted task.. Date: Tue, 8 Oct 2002 14:40:47 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1034088054.120126@newsreader2.wirehub.nl> Hi, For debugging purposes, I like to know if I'm able to find out which task got interrupted if I an in an interrupt. I've got an execption/sysbus error interrupt and like to get to the registers of the interrupted task. Thanx, Renee - - - - - - - - - Renee A. Teunissen PTS Software bv, Meerweg 7, 1405BA Bussum, NL. T.+31-(0)35-6926969, M.+31-(0)6-22778313, http://www.pts.nl, @pts.nl personal embedded / security link page: http://wittenburg10c.nl/db/dest/links.html --------------------------- Newsgroups: comp.os.vxworks Subject: Re: tffs/d_cache Date: Tue, 8 Oct 2002 15:14:17 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: Reply-To: "Michael Lawnick" Sender: mlawnick@p5088eaa5.dip.t-dialin.net You'll have to disable d-cache for the flash-area! -->sysPhysMemDesc[] or BAT in sysLib.c You are welcome "verizon" schrieb im Newsbeitrag news:QKqo9.27975$Mw4.3153@nwrddc01.gnilink.net... > I read *somewhere* that if you are using TFFS that the you must undef the > enable_d_cache in the project ? Running t2 with all the latest tffs/dosfs > upgrads from WRS, but still have corruption of data files. Any answer? > > Thanks ! > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: finding interrupted task.. Date: Tue, 8 Oct 2002 15:21:34 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <1034088054.120126@newsreader2.wirehub.nl> Reply-To: "Leonid Rosenboim" in taskLib.h: extern WIND_TCB *taskIdCurrent; /* always current executing task */ this global variable does not change when one enters an interrupt or exception. ther egisters are in the WIND_TCB structure pointed byu this task ID. - ----------------------------------------------------------------------- Leonid Rosenboim Visit: http://www.masada2000.org/ Consultant Email: my first name at consultant dot com "Renee Teunissen" wrote in message news:1034088054.120126@newsreader2.wirehub.nl... > Hi, > > For debugging purposes, I like to know if I'm able to find out which > task got interrupted if I an in an interrupt. I've got an > execption/sysbus error interrupt and like to get to the registers of > the interrupted task. > > Thanx, > Renee > - - - - - - - - > Renee A. Teunissen > PTS Software bv, Meerweg 7, 1405BA Bussum, NL. > T.+31-(0)35-6926969, M.+31-(0)6-22778313, > http://www.pts.nl, @pts.nl > personal embedded / security link page: > http://wittenburg10c.nl/db/dest/links.html > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: x86 pci interrupt sharing Date: Tue, 8 Oct 2002 15:19:29 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: Reply-To: "Michael Lawnick" Sender: mlawnick@p5088eaa5.dip.t-dialin.net Hi Matthew, sounds you are using PC arch. Did you call pciIntLibInit with the correct number of interrupts that should be shared? If not, your symptoms are as expected. Check your handler tables, you've got the sources in src/drv. It's just some debugging work. HTH Michael "Matthew Peltzer" schrieb im Newsbeitrag news:antf51$pr2$1@overload.lbl.gov... > Greetings, > > I am developing a device driver for custom PCI hardware under vxWorks > 5.4/Tornado 2.0.2, and am experiencing difficulties in handling interrupts. > > I have no problem when the two devices are on their own interrupts, and I > have no problem when using a shared interrupt independently. I am not able > to have two threads at the same time listen to the same interrupt. > > According to the manual page for pciIntConnect(), I should not have to care > about shared interrupts: > > DESCRIPTION > This routine connects an interrupt handler to the PCI > interrupt line(A - D). Link list is created if multiple > handlers are assigned to the single PCI interrupt. > > To me this says that every time I call pciIntConnect(), my function and > parameter will be added to a list that when the interrupt vector is > triggered, each of my functions will be called with the appropriate > parameter. Unfortunately, this is not what I am experiencing, although I am > not 100% sure of what I am experiencing. It seems as though the second call > to pciIntConnect() effectively wipes out the first parameter, for as soon as > an interrupt is generated on this irq, my system deadlocks. I can only > guess that this is because I am not turning off the interrupt on the correct > card because my parameter is not what I expect it to be. > > My solution to this is to do my own isr chaining. I call pciIntConnect() > and sysIntEnablePIC() only when no other connection has been made for this > irq. pciIntDisconnect() and sysIntDisablePIC() are called only when no > other device is listening to the irq. This setup has the same flaw as the > previous: vxWorks reboots when two tasks are run that listen to a separate > PCI card, yet share the same irq. > > I do not know if my callbacks are being called. the first operation in my > callback is a logMsg(), but I am not seeing this. I also have some > logMsg()s to tell me if each interrupt is being caught or ignored, which is > decided by looking at the interrupt line on the hardware device as specified > in the parameter passed into the isr by the system. I have yet to see an > interrupt be ignored. > > > This is the code I am using to enable interrupts: > > pciIntConnect(INUM_TO_IVEC((pci->irq + 0x20)), myDummyISR, list[pci->irq]); > sysIntEnablePIC(pci->irq); > > > 'pci' is a structure that among other things contains the irq number as read > from the pci configuration space with: > > pciConfigInByte(pci->busNum, pci->devNum, pci->fncNum, 0x3c, &(pci->irq)); > > > Is there anything that I am missing for irq sharing to work in vxWorks? > Has anyone had any luck/mis-luck with similar methods? > > > Matthew Peltzer > goose23@spu.edu > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Slow boot on powerpc with VxWorks 5.4 Date: Tue, 8 Oct 2002 15:24:37 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <3DA1C8ED.ED764FF1@free.fr> Reply-To: "Michael Lawnick" Sender: mlawnick@p5088eaa5.dip.t-dialin.net Hi Stéphane, Check your MAC- and IP-addresses, they MUST be unique. Check on serial connection whether a message like 'dc-no carrier' is coming up. If yes: try to get the newest dec driver. Are you sure the time is spent in network? HTH Michael "Stéphane GERVAIS" schrieb im Newsbeitrag news:3DA1C8ED.ED764FF1@free.fr... > Hello, > > I've got a VME system which works very well with VxWorks 5.3.1. > Since I put VxWorks 5.4 on cards, I've got problems while my system is > booting. > All ppc card need more than 3 minutes to load the boot image ! (No > problem with 68k cards). > > I've tried to boot only one ppc card: the problem is always here. > But if I connect the card in another switch (all card are connected in > one switch), the problem disappears !!! > > I suppose it is a network problem (TCP/IP conflicts ?). > What are the differences between VxWorks 5.3.1 and VxWorks 5.4 BSP for > mv2604 BSP to make a such behavior ? > How can I solve this problem ? > > Thanks for your solutions... > > Stephane. > --------------------------- Newsgroups: comp.os.vxworks Subject: How to use reserved memory Date: 8 Oct 2002 06:40:04 -0700 From: derfan2001@altavista.de (Mario Schattschneider) Organization: http://groups.google.com/ Message-ID: <13dfb019.0210080540.2322fd5a@posting.google.com> Hello, I would like to use USER_RESERVED_MEMORY in my program. How do I access or write into this memory section. Must I add the memory to the system pool. Thanks for your help Best regards Mario --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Link Bootable with Downloadable project Date: Tue, 08 Oct 2002 07:53:54 -0400 From: Joe Durusau Organization: Lockheed Martin Corporation Message-ID: <3DA2C752.34E09FDD@lmco.com> References: Yaron wrote: > > Hi All > I Have a Bootable projects and Downloadable projects which works > quite well. > Now I'm trying to link them together to one Bootable project. > I read in the "Tornado Release Notes 2.0" how to link a Bootable > project with Downloadable one, but I don't understand how to create a > Partial Link file of the Downloadable project. > Thanks for any info! > Yaron I'm not sure if I understand you correctly, but if you have a file that you can load with the target server command ld Organization: Dassault Data Services Message-ID: <3DA2EC86.15D074E2@free.fr> References: <3DA1C8ED.ED764FF1@free.fr> Reply-To: s-gervais@dassault-data-services.fr Il s'agit d'un message multivolet au format MIME. - --------------733CD9B9AC056CA566F72BB5 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit Hi Michael, My MAC and IP addresses are unique (they works well with VxWorks 5.3.1). I have no message like 'dc no-carrier' (the boot is completed even if it's slow). How can I be sure about time spent in network ? All I know is activity lights of my switch are off for my powerpc cards during a long time, between two activities of their processor (just after boot load message). Do you think to a bad behavior of the BSP flashed on powerpc cards ? So, why does my card work correctly on the other switch ? I know it's strange and not easy. Thanks for your ideas Stephane. Michael Lawnick a écrit : > Hi Stéphane, > > Check your MAC- and IP-addresses, they MUST be unique. > Check on serial connection whether a message like 'dc-no carrier' is coming > up. > If yes: try to get the newest dec driver. > > Are you sure the time is spent in network? > > HTH > Michael > > "Stéphane GERVAIS" schrieb im Newsbeitrag > news:3DA1C8ED.ED764FF1@free.fr... > > Hello, > > > > I've got a VME system which works very well with VxWorks 5.3.1. > > Since I put VxWorks 5.4 on cards, I've got problems while my system is > > booting. > > All ppc card need more than 3 minutes to load the boot image ! (No > > problem with 68k cards). > > > > I've tried to boot only one ppc card: the problem is always here. > > But if I connect the card in another switch (all card are connected in > > one switch), the problem disappears !!! > > > > I suppose it is a network problem (TCP/IP conflicts ?). > > What are the differences between VxWorks 5.3.1 and VxWorks 5.4 BSP for > > mv2604 BSP to make a such behavior ? > > How can I solve this problem ? > > > > Thanks for your solutions... > > > > Stephane. > > - --------------733CD9B9AC056CA566F72BB5 Content-Type: text/x-vcard; charset=us-ascii; name="sgervais.dds.vcf" Content-Transfer-Encoding: 7bit Content-Description: Carte pour Stéphane GERVAIS Content-Disposition: attachment; filename="sgervais.dds.vcf" begin:vcard n:GERVAIS;Stéphane tel;fax:04 92 38 82 49 tel;work:04 92 38 82 41 x-mozilla-html:FALSE org:Dassault Data Services;Marseille / Sophia version:2.1 email;internet:s-gervais@dassault-data-services.fr adr;quoted-printable:;;2791 Route de St Bernard=0D=0ALes Moulins 2;VALLAURIS;;06220; fn:Stéphane GERVAIS end:vcard - --------------733CD9B9AC056CA566F72BB5-- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: finding interrupted task.. Date: Tue, 8 Oct 2002 16:55:00 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1034096107.212560@newsreader2.wirehub.nl> References: <1034088054.120126@newsreader2.wirehub.nl> Hi Leonid, So you're sayin that intEnd saves all the registers in the TCB? I'm not that sure about this. The only thing I need is the program counter ;) Thanx for you answer! Renee. "Leonid Rosenboim" wrote in message news:newscache$ywzn3h$kf1$1@lnews.actcom.co.il... > > in taskLib.h: > extern WIND_TCB *taskIdCurrent; /* always current executing task */ > > this global variable does not change when one enters an interrupt > or exception. > ther egisters are in the WIND_TCB structure pointed byu this task ID. > > -------------------------------------------------------------------- - --- > Leonid Rosenboim Visit: http://www.masada2000.org/ > Consultant Email: my first name at consultant dot com > > > "Renee Teunissen" wrote in message > news:1034088054.120126@newsreader2.wirehub.nl... > > Hi, > > > > For debugging purposes, I like to know if I'm able to find out which > > task got interrupted if I an in an interrupt. I've got an > > execption/sysbus error interrupt and like to get to the registers of > > the interrupted task. > > > > Thanx, > > Renee > > - - - - - - - - > > Renee A. Teunissen > > PTS Software bv, Meerweg 7, 1405BA Bussum, NL. > > T.+31-(0)35-6926969, M.+31-(0)6-22778313, > > http://www.pts.nl, @pts.nl > > personal embedded / security link page: > > http://wittenburg10c.nl/db/dest/links.html > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: finding interrupted task.. Date: Tue, 8 Oct 2002 16:58:36 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <1034088054.120126@newsreader2.wirehub.nl> <1034096107.212560@newsreader2.wirehub.nl> Reply-To: "Leonid Rosenboim" "Renee Teunissen" wrote in message news:1034096107.212560@newsreader2.wirehub.nl... > > Hi Leonid, > > So you're sayin that intEnd saves all the registers in the TCB? I'm > not that sure about this. The only thing I need is the program counter > ;) > Depending on the architecture - if the arch you have is using a separate stack for interrupts (I think VxWorks has a separate interrupt stack for most if not all architectures), intEnt will save registers in TCB, although the only reg you're interested in is SP, and the rest are in the next frame. > Thanx for you answer! > > Renee. > > > "Leonid Rosenboim" wrote in message > news:newscache$ywzn3h$kf1$1@lnews.actcom.co.il... > > > > in taskLib.h: > > extern WIND_TCB *taskIdCurrent; /* always current executing task */ > > > > this global variable does not change when one enters an interrupt > > or exception. > > ther egisters are in the WIND_TCB structure pointed byu this task > ID. > > > > -------------------------------------------------------------------- > --- > > Leonid Rosenboim Visit: > http://www.masada2000.org/ > > Consultant Email: my first name at consultant > dot com > > > > > > "Renee Teunissen" wrote in > message > > news:1034088054.120126@newsreader2.wirehub.nl... > > > Hi, > > > > > > For debugging purposes, I like to know if I'm able to find out > which > > > task got interrupted if I an in an interrupt. I've got an > > > execption/sysbus error interrupt and like to get to the registers > of > > > the interrupted task. > > > > > > Thanx, > > > Renee > > > - - - - - - - - > > > Renee A. Teunissen > > > PTS Software bv, Meerweg 7, 1405BA Bussum, NL. > > > T.+31-(0)35-6926969, M.+31-(0)6-22778313, > > > http://www.pts.nl, @pts.nl > > > personal embedded / security link page: > > > http://wittenburg10c.nl/db/dest/links.html > > > > > > > > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks,comp.os.qnx Subject: diskon chip probelm Date: Tue, 8 Oct 2002 06:03:09 +0000 (UTC) From: "Ganesh Anand" Organization: Mailgate.ORG Server - http://www.Mailgate.ORG Message-ID: <3d47da57997f9a14d13396faba7f4e3c.98538@mygate.mailgate.org> hi all, i am facing a probelm due to diskon chip. i had done networking between two or more boards embedded with MPC823. everything works properly if i download the image to the boards. But samething is not happening if i boot the boards through diskon. i am using UDP for broadcasting & tcp for communication between two boards. it is showing some error in getting the broadcast address. if anyody cud show some insight regarding the probelm it wud be of great help to me. ganesh - -- Posted via Mailgate.ORG Server - http://www.Mailgate.ORG --------------------------- Newsgroups: comp.os.vxworks Subject: Help Date: Tue, 8 Oct 2002 17:50:24 +0000 (UTC) From: "Steffi Graph" Organization: Mailgate.ORG Server - http://www.Mailgate.ORG Message-ID: <4a0d1f9e8a13dbdff0942e086a8c2d1a.103688@mygate.mailgate.org> Hi, Could anyone help me by answering the following doubt. I want to get my IP from my interface. From where should I refer my interface name?. (i.e Network Interface Name) Can I give a printf statement and know the interface name? Thanks in Advance Steffi.G - -- Posted via Mailgate.ORG Server - http://www.Mailgate.ORG --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Slow boot on powerpc with VxWorks 5.4 Date: Tue, 8 Oct 2002 16:40:34 -0400 From: "William Dennen" Organization: Posted via Supernews, http://www.supernews.com Message-ID: References: <3DA1C8ED.ED764FF1@free.fr> <3DA2EC86.15D074E2@free.fr> Stephane Since the characteristics of the problem change when you use a different switch some questions would be: (1) what are the differences between these switchs -- in particular is one autonegiotiating and the other not? (2) if you move all 2604 boards to the other switch does the problem return? if not, is it still gone when you move everything to that switch? - -- >@< Bill Dennen MCG/GTAC.topaz "Stéphane GERVAIS" wrote in message news:3DA2EC86.15D074E2@free.fr... > Hi Michael, > > My MAC and IP addresses are unique (they works well with VxWorks 5.3.1). > I have no message like 'dc no-carrier' (the boot is completed even if it's > slow). > > How can I be sure about time spent in network ? > All I know is activity lights of my switch are off for my powerpc cards during > a long time, between two activities of their processor (just after boot load > message). > Do you think to a bad behavior of the BSP flashed on powerpc cards ? > So, why does my card work correctly on the other switch ? > > I know it's strange and not easy. > > Thanks for your ideas > > Stephane. > > > Michael Lawnick a écrit : > > > Hi Stéphane, > > > > Check your MAC- and IP-addresses, they MUST be unique. > > Check on serial connection whether a message like 'dc-no carrier' is coming > > up. > > If yes: try to get the newest dec driver. > > > > Are you sure the time is spent in network? > > > > HTH > > Michael > > > > "Stéphane GERVAIS" schrieb im Newsbeitrag > > news:3DA1C8ED.ED764FF1@free.fr... > > > Hello, > > > > > > I've got a VME system which works very well with VxWorks 5.3.1. > > > Since I put VxWorks 5.4 on cards, I've got problems while my system is > > > booting. > > > All ppc card need more than 3 minutes to load the boot image ! (No > > > problem with 68k cards). > > > > > > I've tried to boot only one ppc card: the problem is always here. > > > But if I connect the card in another switch (all card are connected in > > > one switch), the problem disappears !!! > > > > > > I suppose it is a network problem (TCP/IP conflicts ?). > > > What are the differences between VxWorks 5.3.1 and VxWorks 5.4 BSP for > > > mv2604 BSP to make a such behavior ? > > > How can I solve this problem ? > > > > > > Thanks for your solutions... > > > > > > Stephane. > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Help Date: Tue, 08 Oct 2002 15:50:23 -0700 From: "Michael R. Kesti" Organization: MK Associates Message-ID: <3DA3612F.18C745AC@gv.net> References: <4a0d1f9e8a13dbdff0942e086a8c2d1a.103688@mygate.mailgate.org> Reply-To: mkesti@gv.net Steffi Graph wrote: >Hi, > >Could anyone help me by answering the following doubt. > >I want to get my IP from my interface. >From where should I refer my interface name?. (i.e Network Interface >Name) Can I give a printf statement and know the interface name? > >Thanks in Advance See ifAddrGet() in the reference manual's ifLib section. - -- ======================================================================== Michael Kesti | "And like, one and one don't make | two, one and one make one." mkesti@gv.net | - The Who, Bargain --------------------------- Newsgroups: comp.os.vxworks Subject: Re: can't load .out files with dosfs2 Date: 8 Oct 2002 16:00:36 -0700 From: term@xerosoft.com (Tom) Organization: http://groups.google.com/ Message-ID: <15cdb0fb.0210081500.786ccc0b@posting.google.com> References: <15cdb0fb.0210021822.572bd28c@posting.google.com> Hey Michael, Thanks for the reply. I actually have been using the target shell (via serial cable) for all this testing. So far my best guess is that there is some bug in this version of VxWorks (Tornado 2.1.1/VxWorks 5.4.2). If I go back to Tornado 2.0.1/VxWorks 5.4.1 then I don't get this problem. Unfortunatelly I have no support contract with WindRiver so I may have to just use 5.4.1 :( Best regards, Tom "Michael Lawnick" wrote in message news:... > Hi Tom, > > just to be sure ... > You are working under Target-Shell (i.e. serial or TELNET) not WindShell > (aka Host-Shell, in conjunction with target server) ? > If yes (Target-Shell): ??? > If no: try > ->@ld and query google for host-/target-shell issues. > > HTH > Michael > > "Tom" schrieb im Newsbeitrag > news:15cdb0fb.0210021822.572bd28c@posting.google.com... > > This is a repost of a previous message I had posted with new > > information included. > > > > I have been working on getting a strong arm based motherboard to run > > vxworks. One problem I have run into is that I can't load any .out > > files off of my TFFS (True Flash File System). When I try to it just > > gives me an error saying "error reading file (errno = 0x3d0001).". I > > looked up the errno and it resolved to S_objLib_OBJ_ID_ERROR , but > > this is generated from something at a low level that I don't have the > > source code to. I recently tried stepping through the assembler code > > to see where this error was coming from and it turns out errno is not > > set until the cleanup code runs after fioRead fails, so this is errno > > is irrelevant. > > > > Every other operation with the TFFS seems to work just fine. I can > > create, delete, read, write, and rename files. As an example I tried > > copying an out file off my FTP server onto flash and loading it, but > > it failed. Then after that I copied the file off of the TFFS to a new > > file name on my FTP server and then loaded it from that file (over the > > network) and it worked fine. > > > > -- start dump -- > > > > -> cd "/tffs0/" > > value = 0 = 0x0 > > -> ll > > -rwxrwxrwx 1 0 0 10064 Jan 1 1980 test.out > > value = 0 = 0x0 > > -> ld < test.out > > ld error: error reading file (errno = 0x3d0001). > > value = 0 = 0x0 > > -> copy "/tffs0/test.out", "host:newtest.out" > > Copy OK: 10064 bytes copied > > value = 0 = 0x0 > > -> ld < host:newtest.out > > value = 12531032 = 0xbf3558 > > -> > > > > -- end dump -- > > > > > > After these tests I thought the problem might have been size related, > > so I tried to create a minimal "hello world" out file, but that didn't > > work either. > > > > As a work around I thought I would try setting up a runscript that > > creates a ram drive and then copies each file to the ramdrive, loads > > from there, deletes the file, and then continues on to the next one > > until it is done. Unfortunatelly this failed with the same error > > message that I got when I loaded off of a TFFS. The only thing I can > > think of that the ramdrive and TFFS have in common that a ftp file > > system doesn't is that they use dosFS (version 2). > > > > Has anyone else had any problems like this? I'm running > > VxWorks 5.4.2 with Tornado 2.1.1. > > > > > > Best Regards, > > > > Tom --------------------------- Newsgroups: comp.os.vxworks Subject: Tornado in multi-person projects Date: Tue, 8 Oct 2002 20:52:02 -0400 From: "BruceL" Message-ID: Hi, I'm currently the only person building our application code and I don't share .wpj files, but that is soon to change. I am seriously considering ditching Tornado and using a standard build script (originally generated by Tornado) from the command line and/or my editor. My 2 complaints with Tornado is that to share .wpj files, directory names have to match exactly. Also, there is no "GET" rule; i.e., no way to browse the most recent revision controlled files from the network. I would like to know if other people are managing this and what the best way to go is. Thanks, Bruce --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Tornado in multi-person projects Date: Tue, 08 Oct 2002 20:00:06 -0700 From: "Michael R. Kesti" Organization: MK Associates Message-ID: <3DA39BB6.57B20668@gv.net> References: Reply-To: mkesti@gv.net BruceL wrote: >Hi, > >I'm currently the only person building our application code and I don't >share .wpj files, but that is soon to change. I am seriously considering >ditching Tornado and using a standard build script (originally generated by >Tornado) from the command line and/or my editor. > >My 2 complaints with Tornado is that to share .wpj files, directory names >have to match exactly. Also, there is no "GET" rule; i.e., no way to browse >the most recent revision controlled files from the network. > >I would like to know if other people are managing this and what the best way >to go is. This sort of thing is somewhat FAQ so you might try searching the archives. In any event, what's your host OS? - -- ======================================================================== Michael Kesti | "And like, one and one don't make | two, one and one make one." mkesti@gv.net | - The Who, Bargain --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Slow boot on powerpc with VxWorks 5.4 Date: 8 Oct 2002 20:23:00 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0210081923.7392217d@posting.google.com> References: <3DA1C8ED.ED764FF1@free.fr> Hello, Contact WR support and get the newest driver for your card. Some versions of the dc driver have problems with certain combinations of DEC chipset and PHY... HTH, John... Stéphane GERVAIS wrote in message news:<3DA1C8ED.ED764FF1@free.fr>... > Hello, > > I've got a VME system which works very well with VxWorks 5.3.1. > Since I put VxWorks 5.4 on cards, I've got problems while my system is > booting. > All ppc card need more than 3 minutes to load the boot image ! (No > problem with 68k cards). > > I've tried to boot only one ppc card: the problem is always here. > But if I connect the card in another switch (all card are connected in > one switch), the problem disappears !!! > > I suppose it is a network problem (TCP/IP conflicts ?). > What are the differences between VxWorks 5.3.1 and VxWorks 5.4 BSP for > mv2604 BSP to make a such behavior ? > How can I solve this problem ? > > Thanks for your solutions... > > Stephane. > > -- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: can't load .out files with dosfs2 Date: Wed, 9 Oct 2002 10:04:59 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <15cdb0fb.0210021822.572bd28c@posting.google.com> Reply-To: "Leonid Rosenboim" Could you try to replace this command: > -> ld < test.out > ld error: error reading file (errno = 0x3d0001). > value = 0 = 0x0 with this command ? -> ld "/tffs0/test.out" or this command ? -> ld < /tffs0/test.out - -- HTH - ----------------------------------------------------------------------- Leonid Rosenboim Visit: http://www.masada2000.org/ Consultant Email: my first name at consultant dot com "Tom" wrote in message news:15cdb0fb.0210021822.572bd28c@posting.google.com... > This is a repost of a previous message I had posted with new > information included. > > I have been working on getting a strong arm based motherboard to run > vxworks. One problem I have run into is that I can't load any .out > files off of my TFFS (True Flash File System). When I try to it just > gives me an error saying "error reading file (errno = 0x3d0001).". I > looked up the errno and it resolved to S_objLib_OBJ_ID_ERROR , but > this is generated from something at a low level that I don't have the > source code to. I recently tried stepping through the assembler code > to see where this error was coming from and it turns out errno is not > set until the cleanup code runs after fioRead fails, so this is errno > is irrelevant. > > Every other operation with the TFFS seems to work just fine. I can > create, delete, read, write, and rename files. As an example I tried > copying an out file off my FTP server onto flash and loading it, but > it failed. Then after that I copied the file off of the TFFS to a new > file name on my FTP server and then loaded it from that file (over the > network) and it worked fine. > > -- start dump -- > > -> cd "/tffs0/" > value = 0 = 0x0 > -> ll > -rwxrwxrwx 1 0 0 10064 Jan 1 1980 test.out > value = 0 = 0x0 > -> ld < test.out > ld error: error reading file (errno = 0x3d0001). > value = 0 = 0x0 > -> copy "/tffs0/test.out", "host:newtest.out" > Copy OK: 10064 bytes copied > value = 0 = 0x0 > -> ld < host:newtest.out > value = 12531032 = 0xbf3558 > -> > > -- end dump -- > > > After these tests I thought the problem might have been size related, > so I tried to create a minimal "hello world" out file, but that didn't > work either. > > As a work around I thought I would try setting up a runscript that > creates a ram drive and then copies each file to the ramdrive, loads > from there, deletes the file, and then continues on to the next one > until it is done. Unfortunatelly this failed with the same error > message that I got when I loaded off of a TFFS. The only thing I can > think of that the ramdrive and TFFS have in common that a ftp file > system doesn't is that they use dosFS (version 2). > > Has anyone else had any problems like this? I'm running > VxWorks 5.4.2 with Tornado 2.1.1. > > > Best Regards, > > Tom --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Routing problem Date: 9 Oct 2002 01:09:21 -0700 From: Gunter.Ritter@tenovis.com (Gunter Ritter) Organization: http://groups.google.com/ Message-ID: References: Simon Wright wrote in message news:... > Gunter.Ritter@tenovis.com (Gunter Ritter) writes: > > > The W2K client is dialed in. Ping from the client 192.1.2.55 to > > 192.1.2.1 and 192.1.1.141 does work. I get a reply. Ping to a PC in > > the network 192.1.1.0 (PC has IP 192.1.1.1) does not work. I do not > > get a reply. > > > > Ping from the PC with IP 192.1.1.1 to the client 192.1.2.55 does > > work when I add the target 192.1.2.141 as the default gateway in the > > LAN configuration of the PC. > > Do you still not get a reply 192.1.2.55 -> 192.1.1.1 -> 192.1.2.55 > after you have set this default gateway on 192.1.1.1? because > 192.1.1.1 needs to know how to send the ICMP packets back. Hi Simon, thanks for that hint. Yes, I get replies when I ping the machine having set 192.1.1.141 as a default gateway. The problem seems to be that the messages from 192.1.2.55 appearing in the network 192.1.2.0 still have the original source address 192.1.2.55 in the IP header and not 192.1.1.141. This explains why it works with a default gateway entry on 192.1.1.1. But, if I have more than one of these constellations in the same network, it will not work. Is it possible to have a configuration that does not need to set a default gateway? I think about NAT. How can this be done under vxworks? Gunter --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Routing problem Date: Wed, 9 Oct 2002 13:33:49 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1034170439.162264@newsreader2.wirehub.nl> References: Hi, Have you ever thought about proxy-arp? basically with proxy arp, you ethernet interface replies to packets address to your dial-in lines, and therefore the address you assign have to be in the same subnet as your ethernet interface. This how I did this years ago with linux dial-in systems. So to all clients on the net think your dial-in-machine has more than one IP, your dial-in-machine will gate it to your serial device(s). Arp -a on your windows boxes will show the same mac-address for all dial-in lines of that dial-in-machine with the MAC-address of the dial-in-machine. Btw, you dont have to mess with default gates or static routes to make this work. Proxy arp-ing is invented for this - in my humble opinion. :) Cheers, Renee - - - - - - - - - Renee A. Teunissen PTS Software bv, Meerweg 7, 1405BA Bussum, NL. T.+31-(0)35-6926969, M.+31-(0)6-22778313, http://www.pts.nl, @pts.nl personal link page: http://wittenburg10c.nl/db/dest/links.html "Gunter Ritter" wrote in message news:ac95d252.0210090009.5b2e5adf@posting.google.com... > Simon Wright wrote in message news:... > > Gunter.Ritter@tenovis.com (Gunter Ritter) writes: > > > > > The W2K client is dialed in. Ping from the client 192.1.2.55 to > > > 192.1.2.1 and 192.1.1.141 does work. I get a reply. Ping to a PC in > > > the network 192.1.1.0 (PC has IP 192.1.1.1) does not work. I do not > > > get a reply. > > > > > > Ping from the PC with IP 192.1.1.1 to the client 192.1.2.55 does > > > work when I add the target 192.1.2.141 as the default gateway in the > > > LAN configuration of the PC. > > > > Do you still not get a reply 192.1.2.55 -> 192.1.1.1 -> 192.1.2.55 > > after you have set this default gateway on 192.1.1.1? because > > 192.1.1.1 needs to know how to send the ICMP packets back. > > Hi Simon, > > thanks for that hint. > Yes, I get replies when I ping the machine having set 192.1.1.141 as a > default gateway. > > The problem seems to be that the messages from 192.1.2.55 appearing in > the network 192.1.2.0 still have the original source address > 192.1.2.55 in the IP header and not 192.1.1.141. This explains why it > works with a default gateway entry on 192.1.1.1. > > But, if I have more than one of these constellations in the same > network, it will not work. > Is it possible to have a configuration that does not need to set a > default gateway? I think about NAT. How can this be done under > vxworks? > > > Gunter --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Routing problem Date: Wed, 9 Oct 2002 14:17:29 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Reply-To: "Leonid Rosenboim" Since I guess you dont have a solution yet, I am going to tro to help too: 1. There is absolutely no reason to have the "virtual" interface with the address 192.1.2.1 - you will only get in trouble with it. It would be enough if all the PPP interfaces (point-to-point) have the local IP address set to 192.1.2.1. Any number of p2p interfaces can share an IP address on one end (this is known as "unnumbered interface in Cisco nomenclature). 2. One possible reason to your problems is that VxWorks may be trying to send out packets on that interface which is not connected (this is a bug, but #1 above will probably mittigate the issue). 3. For the PCs connecting via PPP, make sure that IPCP is handing out addresses such that 192.1.2.1 is set as default gateway for the dialin clients (use: route print from command line to verify). 4. For the PCs on the 192.1.1.x LAN, they all have to be set up with some default route address. If you have some other router on that network, it would typically serve as the default route. If there is a DHCP server on that LAN, it will also hand out a default route to its PCs. But if all you got on 192.1.1.x is this VxWorks box and any number of PCs, you need to set the default route as you have, to the VxWorks box on all the PCs, and all should work just fine. If you elaborate a bit on what else is there on 192.1.1.x, I might be able to come up with additional ideas. - -- HTH - ----------------------------------------------------------------------- Leonid Rosenboim Visit: http://www.masada2000.org/ Consultant Email: my first name at consultant dot com "Gunter Ritter" wrote in message news:ac95d252.0210072218.4c6bd461@posting.google.com... > Hi vxworkers, > > after solving my problem with the second end driver, the next problem > occures with the routing between the networks. > > The environment is as follows: > The first network interface has the IP address 192.1.1.141 and the > subnet mask is set to 255.255.255.0. This interface is connected to a > physical ethernet LAN. > > The second network interface has the IP address 192.1.2.1 and the > subnet mask is set to 255.255.255.0. This is only a virtual network > interface, not connected to a physial ethernet. This interface shall > provide access to the pyhsical network 192.1.1.0 for clients. These > clients are dialing in to the network 192.1.2.0 using PPP and IPCP. > The IP address of a client is for example 192.1.2.55. The routing > between the networks shall be done by vxworks. > PPP and IPCP is not done by vxworks. For this I use a different stack. > > It looks like this: > > > --------------- ------------ > 192.1.1.141 | |192.1.2.1 192.1.2.55 | | > -------------| vx target |-------------------------| W2K client | > | | | | > ---------------- ------------ > > > Now the problem: > > The W2K client is dialed in. Ping from the client 192.1.2.55 to > 192.1.2.1 and 192.1.1.141 does work. I get a reply. > Ping to a PC in the network 192.1.1.0 (PC has IP 192.1.1.1) does not > work. I do not get a reply. > > Ping from the PC with IP 192.1.1.1 to the client 192.1.2.55 does work > when I add the target 192.1.2.141 as the default gateway in the LAN > configuration of the PC. > > So, what am I doing wrong? What do I have to do for a working routing > with this configuration? > > > Gunter --------------------------- Newsgroups: comp.os.vxworks Subject: Re: finding interrupted task.. Date: Wed, 9 Oct 2002 08:46:54 -0400 From: Keith Arner Organization: Marconi Message-ID: References: <1034088054.120126@newsreader2.wirehub.nl> <1034096107.212560@newsreader2.wirehub.nl> Reply-To: Keith Arner On Tue, 8 Oct 2002, Leonid Rosenboim wrote: > Depending on the architecture - if the arch you have is using a separate > stack for interrupts (I think VxWorks has a separate interrupt stack for > most > if not all architectures), intEnt will save registers in TCB, although the > only > reg you're interested in is SP, and the rest are in the next frame. In case it matters, VxWorks on the x86 does not use a seperate interrupt stack; interrupts are handled on the stack of taskIdCurrent. Keith - -- "Once something has gone horribly wrong..." "...it just gets horribler and horribler." "Then, it's a matter of finding the original horriblism." -- from a conversation on debugging problems in a linked list --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Target resident shell Date: 9 Oct 2002 07:03:56 -0700 From: amit.gil@seabridge.co.il (Amit G.) Organization: http://groups.google.com/ Message-ID: <9574ee82.0210090603.3a09fd62@posting.google.com> References: <9574ee82.0210020243.2f681441@posting.google.com> <3d9af25b$0$22475$44c9b20d@news3.asahi-net.or.jp> 1> well - I have several tasks - starting with the tNetTask (50) and other application tasks (100). 2> I did - after printouts have ended and I wanted to enter shell commands. "Girish Gulawani" wrote in message news:<3d9af25b$0$22475$44c9b20d@news3.asahi-net.or.jp>... > 1> what is the priority of your application task? > 2> by any chance did you press ENTER key after application messages...;-) > > "Amit G." wrote in message > news:9574ee82.0210020243.2f681441@posting.google.com... > > I'm trying to use the target resident shell with no success. > > > > I build the version with the USE_TARGET_SHELL > > > > I get the shell welcome screen on my serial console - yet no prompt. > > > > This is what I see: > > Attached TCP/IP interface to motscc unit 0 > > 0xfbffdf8 (tRootTask): Attached network interface lo0 successfully > > NFS client support not included. > > > > Adding 15961 symbols for standalone. > > 0xfbffdf8 (tRootTask): rtrequest:: DEL don't free block e395fa0 > > > > > > ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]] > > ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]] > > ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]] > > ]]]]]]]]]]] ]]]] ]]]]]]]]]] ]] ]]]] > > (R) > > ] ]]]]]]]]] ]]]]]] ]]]]]]]] ]] ]]]] > > ]] ]]]]]]] ]]]]]]]] ]]]]]] ] ]] ]]]] > > ]]] ]]]]] ] ]]] ] ]]]] ]]] ]]]]]]]]] ]]]] ]] ]]]] ]] > > ]]]]] > > ]]]] ]]] ]] ] ]]] ]] ]]]]] ]]]]]] ]] ]]]]]]] ]]]] ]] > > ]]]] > > ]]]]] ] ]]]] ]]]]] ]]]]]]]] ]]]] ]] ]]]] ]]]]]]] > > ]]]] > > ]]]]]] ]]]]] ]]]]]] ] ]]]]] ]]]] ]] ]]]] ]]]]]]]] > > ]]]] > > ]]]]]]] ]]]]] ] ]]]]]] ] ]]] ]]]] ]] ]]]] ]]]] ]]]] > > ]]]] > > ]]]]]]]] ]]]]] ]]] ]]]]]]] ] ]]]]]]] ]]]] ]]]] > > ]]]] ]]]]] > > ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]] > > ]]]]]]]]]]]]]]]]]]]]]]]]]]]]] Development System > > ]]]]]]]]]]]]]]]]]]]]]]]]]]]] > > ]]]]]]]]]]]]]]]]]]]]]]]]]]] VxWorks version 5.4 > > ]]]]]]]]]]]]]]]]]]]]]]]]]] KERNEL: WIND version 2.5 > > ]]]]]]]]]]]]]]]]]]]]]]]]] Copyright Wind River Systems, Inc., > > 1984-1999 > > > > CPU: SIEMENS BIAS 2.1 - PPC7400. > > Processor #0. > > Memory Size: 0xfc00000. BSP version > > 1.0. > > WDB_COMM_TYPE: WDB_COMM_NETWORK > > WDB: Ready. > > > > > > After that I get some printouts generated by my application and that's > > about it, now visible communication with the shell. > > > > Any suggestions? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Tornado in multi-person projects Date: Wed, 9 Oct 2002 16:50:37 +0200 From: "Martin Raabe" Organization: 1&1 Internet AG Message-ID: References: Hello Bruce, I would take a look at SNiFF+ for Tornado, which manages exactly what you want to achieve. Ciao - -- Martin Raabe E: Martin.RaabeATNOSPAMBaSystem.de "BruceL" schrieb im Newsbeitrag news:anvuls$bg7$1@bob.news.rcn.net... > Hi, > > I'm currently the only person building our application code and I don't > share .wpj files, but that is soon to change. I am seriously considering > ditching Tornado and using a standard build script (originally generated by > Tornado) from the command line and/or my editor. > > My 2 complaints with Tornado is that to share .wpj files, directory names > have to match exactly. Also, there is no "GET" rule; i.e., no way to browse > the most recent revision controlled files from the network. > > I would like to know if other people are managing this and what the best way > to go is. > > Thanks, > Bruce > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Target resident shell Date: Wed, 09 Oct 2002 12:08:24 -0400 From: Joe Durusau Organization: Lockheed Martin Corporation Message-ID: <3DA45478.6CCF1948@lmco.com> References: <9574ee82.0210020243.2f681441@posting.google.com> <3d9af25b$0$22475$44c9b20d@news3.asahi-net.or.jp> <9574ee82.0210090603.3a09fd62@posting.google.com> This is just a guess, but the strange number in the message about "DEL didn't release..." tends to suggest a stack overflow or perhaps an error in the way something is being stored in an array. Speaking only for myself, Joe Durusau "Amit G." wrote: > > 1> well - I have several tasks - starting with the tNetTask (50) and > other application tasks (100). > 2> I did - after printouts have ended and I wanted to enter shell > commands. > > "Girish Gulawani" wrote in message news:<3d9af25b$0$22475$44c9b20d@news3.asahi-net.or.jp>... > > 1> what is the priority of your application task? > > 2> by any chance did you press ENTER key after application messages...;-) > > > > "Amit G." wrote in message > > news:9574ee82.0210020243.2f681441@posting.google.com... > > > I'm trying to use the target resident shell with no success. > > > > > > I build the version with the USE_TARGET_SHELL > > > > > > I get the shell welcome screen on my serial console - yet no prompt. > > > > > > This is what I see: > > > Attached TCP/IP interface to motscc unit 0 > > > 0xfbffdf8 (tRootTask): Attached network interface lo0 successfully > > > NFS client support not included. > > > > > > Adding 15961 symbols for standalone. > > > 0xfbffdf8 (tRootTask): rtrequest:: DEL don't free block e395fa0 > > > > > > > > > ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]] > > > ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]] > > > ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]] > > > ]]]]]]]]]]] ]]]] ]]]]]]]]]] ]] ]]]] > > > (R) > > > ] ]]]]]]]]] ]]]]]] ]]]]]]]] ]] ]]]] > > > ]] ]]]]]]] ]]]]]]]] ]]]]]] ] ]] ]]]] > > > ]]] ]]]]] ] ]]] ] ]]]] ]]] ]]]]]]]]] ]]]] ]] ]]]] ]] > > > ]]]]] > > > ]]]] ]]] ]] ] ]]] ]] ]]]]] ]]]]]] ]] ]]]]]]] ]]]] ]] > > > ]]]] > > > ]]]]] ] ]]]] ]]]]] ]]]]]]]] ]]]] ]] ]]]] ]]]]]]] > > > ]]]] > > > ]]]]]] ]]]]] ]]]]]] ] ]]]]] ]]]] ]] ]]]] ]]]]]]]] > > > ]]]] > > > ]]]]]]] ]]]]] ] ]]]]]] ] ]]] ]]]] ]] ]]]] ]]]] ]]]] > > > ]]]] > > > ]]]]]]]] ]]]]] ]]] ]]]]]]] ] ]]]]]]] ]]]] ]]]] > > > ]]]] ]]]]] > > > ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]] > > > ]]]]]]]]]]]]]]]]]]]]]]]]]]]]] Development System > > > ]]]]]]]]]]]]]]]]]]]]]]]]]]]] > > > ]]]]]]]]]]]]]]]]]]]]]]]]]]] VxWorks version 5.4 > > > ]]]]]]]]]]]]]]]]]]]]]]]]]] KERNEL: WIND version 2.5 > > > ]]]]]]]]]]]]]]]]]]]]]]]]] Copyright Wind River Systems, Inc., > > > 1984-1999 > > > > > > CPU: SIEMENS BIAS 2.1 - PPC7400. > > > Processor #0. > > > Memory Size: 0xfc00000. BSP version > > > 1.0. > > > WDB_COMM_TYPE: WDB_COMM_NETWORK > > > WDB: Ready. > > > > > > > > > After that I get some printouts generated by my application and that's > > > about it, now visible communication with the shell. > > > > > > Any suggestions? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: can't load .out files with dosfs2 Date: 9 Oct 2002 09:22:00 -0700 From: term@xerosoft.com (Tom) Organization: http://groups.google.com/ Message-ID: <15cdb0fb.0210090822.52841a20@posting.google.com> References: <15cdb0fb.0210021822.572bd28c@posting.google.com> Hey Leonid, I have tried a number of different command lines to load the file, but I only showed one in my first e-mail. I also tried: ld 1,1,"test.out" ld 1,1,"/tffs0/test.out" ld < test.out ld < /tffs0/test.out and loadModule() with a file descriptor that I manually opened. unfortunatelly they all failed. Any other ideas? Best regards, Tom "Leonid Rosenboim" wrote in message news:... > Could you try to replace this command: > > > -> ld < test.out > > ld error: error reading file (errno = 0x3d0001). > > value = 0 = 0x0 > > with this command ? > > -> ld "/tffs0/test.out" > > or this command ? > > -> ld < /tffs0/test.out > > -- HTH > ----------------------------------------------------------------------- > Leonid Rosenboim Visit: http://www.masada2000.org/ > Consultant Email: my first name at consultant dot com > > > "Tom" wrote in message > news:15cdb0fb.0210021822.572bd28c@posting.google.com... > > This is a repost of a previous message I had posted with new > > information included. > > > > I have been working on getting a strong arm based motherboard to run > > vxworks. One problem I have run into is that I can't load any .out > > files off of my TFFS (True Flash File System). When I try to it just > > gives me an error saying "error reading file (errno = 0x3d0001).". I > > looked up the errno and it resolved to S_objLib_OBJ_ID_ERROR , but > > this is generated from something at a low level that I don't have the > > source code to. I recently tried stepping through the assembler code > > to see where this error was coming from and it turns out errno is not > > set until the cleanup code runs after fioRead fails, so this is errno > > is irrelevant. > > > > Every other operation with the TFFS seems to work just fine. I can > > create, delete, read, write, and rename files. As an example I tried > > copying an out file off my FTP server onto flash and loading it, but > > it failed. Then after that I copied the file off of the TFFS to a new > > file name on my FTP server and then loaded it from that file (over the > > network) and it worked fine. > > > > -- start dump -- > > > > -> cd "/tffs0/" > > value = 0 = 0x0 > > -> ll > > -rwxrwxrwx 1 0 0 10064 Jan 1 1980 test.out > > value = 0 = 0x0 > > -> ld < test.out > > ld error: error reading file (errno = 0x3d0001). > > value = 0 = 0x0 > > -> copy "/tffs0/test.out", "host:newtest.out" > > Copy OK: 10064 bytes copied > > value = 0 = 0x0 > > -> ld < host:newtest.out > > value = 12531032 = 0xbf3558 > > -> > > > > -- end dump -- > > > > > > After these tests I thought the problem might have been size related, > > so I tried to create a minimal "hello world" out file, but that didn't > > work either. > > > > As a work around I thought I would try setting up a runscript that > > creates a ram drive and then copies each file to the ramdrive, loads > > from there, deletes the file, and then continues on to the next one > > until it is done. Unfortunatelly this failed with the same error > > message that I got when I loaded off of a TFFS. The only thing I can > > think of that the ramdrive and TFFS have in common that a ftp file > > system doesn't is that they use dosFS (version 2). > > > > Has anyone else had any problems like this? I'm running > > VxWorks 5.4.2 with Tornado 2.1.1. > > > > > > Best Regards, > > > > Tom --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How to use reserved memory Date: 9 Oct 2002 10:07:51 -0700 From: eyalk@radlan.com (Eyal) Organization: http://groups.google.com/ Message-ID: References: <13dfb019.0210080540.2322fd5a@posting.google.com> derfan2001@altavista.de (Mario Schattschneider) wrote in message news:<13dfb019.0210080540.2322fd5a@posting.google.com>... > Hello, > > I would like to use USER_RESERVED_MEMORY in my program. > How do I access or write into this memory section. > Must I add the memory to the system pool. > > Thanks for your help > > Best regards > > Mario vxWorks using memory from LOCAL_MEM_LOCAL_ADRS to LOCAL_MEM_LOCAL_ADRS + LOCAL_MEM_SIZE - USER_RESERVED_MEMORY so, for example, if: LOCAL_MEM_LOCAL_ADRS = 0 LOCAL_MEM_SIZE = 0x4000000 USER_RESERVED_MEMORY = 0x3000000 you can use your memory between 0x1000000 to 0x4000000 for your separate application, without interfering between VxWorks and your appl. Eyal --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: 9 Oct 2002 10:29:21 -0700 From: brian_doe@hotmail.com (brian) Organization: http://groups.google.com/ Message-ID: References: <3a5ff53f.0210020748.3643d45a@posting.google.com> <3DA14F95.A74993DC@alcatel-ke.de> Tim Sinkins wrote in message news:<3DA14F95.A74993DC@alcatel-ke.de>... > > > Now. If I can do this in 3-6mos, > > Take it from me, you can't. Are you talking about rolling your own > or just switching to another RTOS in this time? Either way you still > can't do it. The latter took us about 12 man months, not including > the item below (similar reason as you, old OS not ported to new > processor, with hindsight, it would have been easier to write a port, > but I think some of the reason the engineers made the decision that > they did was a "career development" one) > > > it costs me 100k. Are there maintenance costs? > > There are bound to be little differences in the interface that you > will be debugging for months. > Tim. It really took 12 man months? Lets say that I am using Ucos or something like that. The "port" is to attach the scheduler etc to the interrupts/timers etc etc. Im not trying to write a new kernel. Is that a 12month job? (Could be...but Ive heard of guys that did it in 3mos :) Thanx --------------------------- Newsgroups: comp.os.vxworks Subject: socket sharing Date: Wed, 9 Oct 2002 13:26:44 -0400 From: "Linux Prj" Organization: Lucent Technologies, Indian Hill Message-ID: Hi, In vxWorks 5.4, if two tasks have to share a socket descriptor, do anything special have to be done. i.e., a task has opened a socket, and another task wants to access it for read/write etc. Thanks --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: 9 Oct 2002 10:33:22 -0700 From: brian_doe@hotmail.com (brian) Organization: http://groups.google.com/ Message-ID: References: <3a5ff53f.0210020748.3643d45a@posting.google.com> "Leonid Rosenboim" wrote in message news:... > Brian, could not resist, must follow on this once more > (last time, promise): Hey. Thanx for the responses :) > > You will probably spend in the era of $300k to $600k > on the switch of VxWorks for a semi-custom OS - > using something unknown will be plenty of surprises. > And that will save you 25 cents off the COGS. That's the questions I have. 300-600k is 2-4 man years. Can it really take that long? How? As in, say , porting one of the Linux baby kernels (no stack..just kernel etc) Again thanx for the input. --------------------------- Newsgroups: comp.os.vxworks Subject: inflate Lib.... Date: 9 Oct 2002 10:55:54 -0700 From: chs@lucent.com (ChengHsin Hsu) Organization: http://groups.google.com/ Message-ID: <8b53eb96.0210090955.3ff816a5@posting.google.com> Hello, Is there anyway that we can know how much memory we need to malloc as dest buffer before the inflate call like this? inflate (src, dest, nBytes); Thanks, Bear. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Wed, 9 Oct 2002 19:53:15 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <3a5ff53f.0210020748.3643d45a@posting.google.com> Reply-To: "Leonid Rosenboim" "brian" wrote in message news:d2f67bd8.0210090933.19572732@posting.google.com... > "Leonid Rosenboim" wrote in message news:... > > Brian, could not resist, must follow on this once more > > (last time, promise): > > Hey. Thanx for the responses :) > > > > > You will probably spend in the era of $300k to $600k > > on the switch of VxWorks for a semi-custom OS - > > using something unknown will be plenty of surprises. > > And that will save you 25 cents off the COGS. > > That's the questions I have. 300-600k is 2-4 man years. Can it really > take that long? How? As in, say , porting one of the Linux baby > kernels (no stack..just kernel etc) > > Again thanx for the input. It really depends on how experienced the people on this project are. Also, you need to appreciate the tools that you would need to replace, although it's OK to use the GNU toolchain you got with VxWorks, you will need to come up with a low-level debugger such as an emulator. Then, there are drivers, if the new kernel has different driver APIs, and you can't just port the current ones as they probably include a WindRiver copyright. Last but not least is the application, it is critical to make an "inventory" of OS calls used by the application, and decide on a case-by-case basis how these will be either (1) mapped to new OS existing calls; (2) emulated on top of the new OS or added to the new OS or (3) replaced at the application with a call that exists in the new OS. The risk factor here is quite large, much more substantial than any hardware modification that would save the same amount on the COGS, or even 10 times more ! HTH, Leonid --------------------------- Newsgroups: comp.os.vxworks Subject: Re: socket sharing Date: Wed, 9 Oct 2002 19:56:11 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Reply-To: "Leonid Rosenboim" "Linux Prj" wrote in message news:ao1osl$l21@netnews.proxy.lucent.com... > Hi, > In vxWorks 5.4, if two tasks have to share > a socket descriptor, do anything special have to be done. > i.e., a task has opened a socket, and another task > wants to access it for read/write etc. > > Thanks > > Depends. For a UDP socket, it is just as simple as that. For a TCP socket, being stream oriented, you need to consider how that will effect the byte stream, and will probably need to have a Mutex on the sock to make sure that any stream component (or sytnax element) is written to the sock entirely before the other socket has a chanse to write. - -- - ----------------------------------------------------------------------- Leonid Rosenboim Visit: http://www.masada2000.org/ Consultant Email: my first name at consultant dot com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: can't load .out files with dosfs2 Date: Wed, 9 Oct 2002 18:47:50 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <15cdb0fb.0210021822.572bd28c@posting.google.com> <15cdb0fb.0210090822.52841a20@posting.google.com> Reply-To: "Leonid Rosenboim" This all seems so strange that it is probably something very simple we all have ignored. If you have the exact same file effectively loading from a network, there should be no reason at all this file won't be loading from dosFs, well except that different file systems handle the lseek() call differently, and most architectures dynamic loader will use lseek(). How about if you create a pipe, copy the file into the pipe, while another task runs loadModule with the pipe being its stdin ? (I suppose you iknow how to do this). Keep me posted. - ----------------------------------------------------------------------- Leonid Rosenboim Visit: http://www.masada2000.org/ Consultant Email: my first name at consultant dot com "Tom" wrote in message news:15cdb0fb.0210090822.52841a20@posting.google.com... > Hey Leonid, > > I have tried a number of different command lines to load the file, but > I only showed one in my first e-mail. > > I also tried: > > ld 1,1,"test.out" > ld 1,1,"/tffs0/test.out" > ld < test.out > ld < /tffs0/test.out > and loadModule() with a file descriptor that I manually opened. > > unfortunatelly they all failed. > > Any other ideas? > > > Best regards, > > Tom > > "Leonid Rosenboim" wrote in message news:... > > Could you try to replace this command: > > > > > -> ld < test.out > > > ld error: error reading file (errno = 0x3d0001). > > > value = 0 = 0x0 > > > > with this command ? > > > > -> ld "/tffs0/test.out" > > > > or this command ? > > > > -> ld < /tffs0/test.out > > > > -- HTH > > ----------------------------------------------------------------------- > > Leonid Rosenboim Visit: http://www.masada2000.org/ > > Consultant Email: my first name at consultant dot com > > > > > > "Tom" wrote in message > > news:15cdb0fb.0210021822.572bd28c@posting.google.com... > > > This is a repost of a previous message I had posted with new > > > information included. > > > > > > I have been working on getting a strong arm based motherboard to run > > > vxworks. One problem I have run into is that I can't load any .out > > > files off of my TFFS (True Flash File System). When I try to it just > > > gives me an error saying "error reading file (errno = 0x3d0001).". I > > > looked up the errno and it resolved to S_objLib_OBJ_ID_ERROR , but > > > this is generated from something at a low level that I don't have the > > > source code to. I recently tried stepping through the assembler code > > > to see where this error was coming from and it turns out errno is not > > > set until the cleanup code runs after fioRead fails, so this is errno > > > is irrelevant. > > > > > > Every other operation with the TFFS seems to work just fine. I can > > > create, delete, read, write, and rename files. As an example I tried > > > copying an out file off my FTP server onto flash and loading it, but > > > it failed. Then after that I copied the file off of the TFFS to a new > > > file name on my FTP server and then loaded it from that file (over the > > > network) and it worked fine. > > > > > > -- start dump -- > > > > > > -> cd "/tffs0/" > > > value = 0 = 0x0 > > > -> ll > > > -rwxrwxrwx 1 0 0 10064 Jan 1 1980 test.out > > > value = 0 = 0x0 > > > -> ld < test.out > > > ld error: error reading file (errno = 0x3d0001). > > > value = 0 = 0x0 > > > -> copy "/tffs0/test.out", "host:newtest.out" > > > Copy OK: 10064 bytes copied > > > value = 0 = 0x0 > > > -> ld < host:newtest.out > > > value = 12531032 = 0xbf3558 > > > -> > > > > > > -- end dump -- > > > > > > > > > After these tests I thought the problem might have been size related, > > > so I tried to create a minimal "hello world" out file, but that didn't > > > work either. > > > > > > As a work around I thought I would try setting up a runscript that > > > creates a ram drive and then copies each file to the ramdrive, loads > > > from there, deletes the file, and then continues on to the next one > > > until it is done. Unfortunatelly this failed with the same error > > > message that I got when I loaded off of a TFFS. The only thing I can > > > think of that the ramdrive and TFFS have in common that a ftp file > > > system doesn't is that they use dosFS (version 2). > > > > > > Has anyone else had any problems like this? I'm running > > > VxWorks 5.4.2 with Tornado 2.1.1. > > > > > > > > > Best Regards, > > > > > > Tom --------------------------- Newsgroups: comp.os.vxworks Subject: Re: x86 pci interrupt sharing Date: 9 Oct 2002 13:18:59 -0700 From: goose23@spu.edu (Matthew Peltzer) Organization: http://groups.google.com/ Message-ID: <75a1cd5a.0210091218.38d18669@posting.google.com> References: "Michael Lawnick" wrote in message news:... > Hi Matthew, > > sounds you are using PC arch. Indeed. I am using a PentiumPro, but my BSP is based off of 'PENTIUM'. > Did you call pciIntLibInit with the correct number of interrupts that should > be shared? > If not, your symptoms are as expected. > Check your handler tables, you've got the sources in src/drv. > It's just some debugging work. > > HTH > Michael > handler tables? you mean pciIntList[]? As far as I can tell, pciIntLibInit is prototyped as: STATUS pciIntLibInit (void); And there does not seem to be anything useful in the near vincinity, with the exception of this comment: "Shared PCI interrupt are supported by three functions: pciInt(), pciIntConnect(), pciIntDisconnect(). pciIntConnect() adds the specified interrupt handler to the link list and pciIntDisconnect() removes it from the link list. Master interrupt handler pciInt() executes these interrupt handlers in the link list for a PCI interrupt. Each interrupt handler must check the device dependent interrupt status bit to determine the source of the interrupt, since it simply execute all interrupt handlers in the link list. pciInt() should be attached by intConnect() function in the BSP initialization with its parameter. The parameter is an IRQ associated to the PCI interrupt." After reading this, I have devised this algorithm: 1) upon driver/vxWorks initialization, connect & enable the hardware's irq: intConnect(INUM_TO_IVEC(pci->irq), pciInt, pci->irq); intEnable(pci->irq); 2) when the user wants to listen to interrupts generated by the hardware, add my ISR to the list: pciIntConnect(INUM_TO_IVEC(pci->irq), myISR, pci); The only thing stopping me from doing this is that I have not seen anything like this in all the user manuals or newsgroups that I have searched. But I see no reason, per the comment, that nothing other than this is the correct code. Your thoughts? Matthew Peltzer --------------------------- Newsgroups: comp.os.vxworks Subject: Re: x86 pci interrupt sharing Date: Wed, 09 Oct 2002 17:12:01 -0400 From: Varndell Engineering Organization: Posted Via Newsfeeds.com = SPEED+RETENTION+COMPLETION = http://www.newsfeeds.com Message-ID: <3DA49BA1.5E7F7242@hotmail.com> References: <75a1cd5a.0210091218.38d18669@posting.google.com> Matthew Peltzer wrote: > > "Michael Lawnick" wrote in message news:... > > Hi Matthew, > > > > sounds you are using PC arch. > Indeed. I am using a PentiumPro, but my BSP is based off of > 'PENTIUM'. > > > Did you call pciIntLibInit with the correct number of interrupts that should > > be shared? > > If not, your symptoms are as expected. > > Check your handler tables, you've got the sources in src/drv. > > It's just some debugging work. > > > > HTH > > Michael > > > > handler tables? you mean pciIntList[]? > > As far as I can tell, pciIntLibInit is prototyped as: > > STATUS pciIntLibInit (void); > > And there does not seem to be anything useful in the near vincinity, > with the exception of this comment: > > "Shared PCI interrupt are supported by three functions: pciInt(), > pciIntConnect(), pciIntDisconnect(). pciIntConnect() adds the > specified interrupt handler to the link list and pciIntDisconnect() > removes it from the link list. Master interrupt handler pciInt() > executes these interrupt handlers in the link list for a PCI > interrupt. Each interrupt handler must check the device dependent > interrupt status bit to determine the source of the interrupt, since > it simply execute all interrupt handlers in the link list. pciInt() > should be attached by intConnect() function in the BSP initialization > with its parameter. The parameter is an IRQ associated to the PCI > interrupt." > > After reading this, I have devised this algorithm: > > 1) upon driver/vxWorks initialization, connect & enable the hardware's > irq: > intConnect(INUM_TO_IVEC(pci->irq), pciInt, pci->irq); > intEnable(pci->irq); > > 2) when the user wants to listen to interrupts generated by the > hardware, add my ISR to the list: > pciIntConnect(INUM_TO_IVEC(pci->irq), myISR, pci); > > The only thing stopping me from doing this is that I have not seen > anything like this in all the user manuals or newsgroups that I have > searched. But I see no reason, per the comment, that nothing other > than this is the correct code. > > Your thoughts? You didn't answer the question as to whether or not you've called pciIntLibInit(). If your BSP is not calling it, in sysHwInit(), then you need to add a call to it. You are correct that it takes no parameters, but that doesn't mean you needn't call it. After that, simply call pciIntConnect just as you would intConnect. And don't forget to add INT_NUM_IRQ0 to the irq number when you do. You don't need to do the intConnect for pciIntLib. It will do that itself. You do need to explicitly enable the IRQ, and you also need to be aware that the IRQ will stay enabled even after you've pciIntDisconnected all ISRs from it. You also need to be aware that when/if you temporarily mask a given IRQ, you are possibly disabling an IRQ that more than device is relying on. Finally, despite it's name, there is nothing PCI about pciIntLib. It's just a generic IRQ sharing utility. Regards, George Varndell - -- Varndell Engineering, LLC Embedded Systems Expertise Need a vxWorks END driver for SiS900? Contact us. http://www.varndellengineering.com - -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- http://www.newsfeed.com The #1 Newsgroup Service in the World! - -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Hardware Flow Control for Serial Ports Date: Wed, 9 Oct 2002 16:57:37 -0700 From: "Dan Pinson" Message-ID: I'm receiving raw data on a serial port using VxWorks standard driver. I suspect that the receive FIFO is being overrun. Here is the situation: PentiumPC BSP 16550 UART After opening the port, I enable hardware flow control via ioctl. After receiving at 115K for awhile, I get unexpected data. The same device works OK on another system under a different OS. This is just an indicator; not proof. I checked the highwater mark in my RX ring buffer. It was well below the free space. In looking at the 16550 serial drivers target/src/drv/sio/ns16550Serial.c and ns16550Sio.c, I see how transmitting is throttled by CTS but not how RTS is manipulated. I notice that the UART driver does nothing with an rx overrun interrupt. I thought tyLib might clear RTS when the buffer gets full, but I don't see an API in the UART driver that could be used. Bottom-line is that I wonder if RTS/CTS flow control works on the x86 BSP and/or with the 16550 drivers. Thanks in advance, Dan --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: 09 Oct 2002 21:16:30 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: References: <3a5ff53f.0210020748.3643d45a@posting.google.com> <3DA14F95.A74993DC@alcatel-ke.de> Sender: bpringle@DeadDuck Tim> Tim. It really took 12 man months? Lets say that I am using Tim> Ucos or something like that. The "port" is to attach the Tim> scheduler etc to the interrupts/timers etc etc. Im not trying to Tim> write a new kernel. Is that a 12month job? (Could be...but Ive Tim> heard of guys that did it in 3mos :) I ported ucLinux to a custom ColdFire board in about three month. This board has bluetooth hardware and used the ucLinux network stack. I was not involved in all of the bluetooth development. Just the support for ucLinux including a boot loader. If your hardware is off the shelf, then a non-vx driver is probably available for it. If the hardware is custom, then you probably have all the rights to it. The one down side is the tools. This is really the only good thing about vxWorks. I have seen some good GUI front-ends to GDB and some of them even run on Windows. Consider that Tornado is built using GDB and Tcl and it is clear that we aren't too far away. I have also used RTEM and uCos-II in hobby mode. They aren't that hard to get working... but then again, I didn't support them. The main problem with ucLinux was the expectation that any Linux package could compile and work in a 1/2 hour. Anything that uses fork ergo pthreads, etc. will not be straight forward to port. I would recommend the full Linux instead (especially if you have the MMU). fwiw, Bill Pringlemeir. - -- Only a cretin like you would say that trees are perverted. Let me tell you something, you DAF driver, if you've seen one dog, you've seen them all. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Hardware Flow Control for Serial Ports Date: Thu, 10 Oct 2002 07:49:17 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: Reply-To: "Michael Lawnick" Sender: mlawnick@pd9524a6d.dip.t-dialin.net Hi Dan, didn't really get through your post, just the standard injection when UART is concerned: Did you enable its FIFO? Remember that the standard vxWorks driver doesn't do that. You'll have to patch the code in src/drv/sio along your needs. HTH Michael "Dan Pinson" schrieb im Newsbeitrag news:ao2fpu$ijho7$1@ID-48686.news.dfncis.de... > I'm receiving raw data on a serial port using VxWorks standard driver. I > suspect that the receive FIFO is being overrun. Here is the situation: > > PentiumPC BSP > 16550 UART > > After opening the port, I enable hardware flow control via ioctl. After > receiving at 115K for awhile, I get unexpected data. The same device works > OK on another system under a different OS. This is just an indicator; not > proof. > > I checked the highwater mark in my RX ring buffer. It was well below the > free space. > > In looking at the 16550 serial drivers target/src/drv/sio/ns16550Serial.c > and ns16550Sio.c, I see how transmitting is throttled by CTS but not how RTS > is manipulated. I notice that the UART driver does nothing with an rx > overrun interrupt. I thought tyLib might clear RTS when the buffer gets > full, but I don't see an API in the UART driver that could be used. > > Bottom-line is that I wonder if RTS/CTS flow control works on the x86 BSP > and/or with the 16550 drivers. > > Thanks in advance, > Dan > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Routing problem Date: 10 Oct 2002 00:17:51 -0700 From: Gunter.Ritter@tenovis.com (Gunter Ritter) Organization: http://groups.google.com/ Message-ID: References: Thanks for your answer. I think the following is my main problem: > 4. For the PCs on the 192.1.1.x LAN, they all have to be set up > with some default route address. If you have some other router > on that network, it would typically serve as the default route. > If there is a DHCP server on that LAN, it will also hand out > a default route to its PCs. > But if all you got on 192.1.1.x is this VxWorks box and > any number of PCs, you need to set the default route > as you have, to the VxWorks box on all the PCs, and > all should work just fine. One problem is, that I do not have access to the PC configuration in the corporate network of a customer. This can be solved by a routing entry that for my vxworks box in the existing default gateway. This might be done static or dynamic via RIP. But, this works only if I have not to many vxworks boxes and the customer grants me access to this router. But if I have many of these vxworks boxes in the main network (it can be thousands), this constellation will not work. So, I think I need NAT to access the corporate LAN from the subnet behind the vxworks box. Gunter. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Source file dependencies in Tornado 2.0 Date: Thu, 10 Oct 2002 09:22:28 +0200 From: "Ofer Goren" Organization: Verio Message-ID: References: <3da16847$0$91466$edfadb0f@dspool01.news.tele.dk> "Carsten Olsen" wrote in message news:3da16847$0$91466$edfadb0f@dspool01.news.tele.dk... > Hi, > > I have a problem with file-dependencies among different directories in > Tornado 2.0. When source-files are placed in different directories, > the dependencies fail and the project cannot be compiled. > > Example directory-file structure: > Classes/ > class1.cpp > class1.h > Project/ > project.wsp > project.wpj > source/ > main.cpp > > In main.cpp I include "class1.h" with '#include "class1.h"' > When I compile, I get the error-mesage: > "make: *** No rule to make target `class1.h', needed by `main.o'. Stop" > > The problem is, that the dependencies at the end of the auto-generated > makefile are wrong. > How do I tell Tornado, that header-files are placed in another directory? > I do not want to make includes like '#include "../../classes/class1.h"' > The information should belong to the project-files, not the individual > source-files. > > Does anyone know how to handle this situation? > > Thanks. > > Kind regards, > Carsten Olsen > > > Hi. The dependencies issue is a major flow in Tornado2. It only works sometimes. Just use the -I flags to the correct locations (path). In addition, double check that you dont have the same header file (the same name) in different locations. That should solve the problem. If not, re-submit your question, and I will try to help you further. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Thu, 10 Oct 2002 09:32:16 +0200 From: Tim Sinkins Organization: Alcatel Kommunikations Elektronik Message-ID: <3DA52D00.5F5070E@alcatel-ke.de> References: <3a5ff53f.0210020748.3643d45a@posting.google.com> <3DA14F95.A74993DC@alcatel-ke.de> Reply-To: timothy.sinkins@alcatel-ke.de brian wrote: > > Tim Sinkins wrote in message news:<3DA14F95.A74993DC@alcatel-ke.de>... > > > > > Now. If I can do this in 3-6mos, > > > > Take it from me, you can't. Are you talking about rolling your own > > or just switching to another RTOS in this time? Either way you still > > can't do it. The latter took us about 12 man months, not including > > the item below (similar reason as you, old OS not ported to new > > processor, with hindsight, it would have been easier to write a port, > > but I think some of the reason the engineers made the decision that > > they did was a "career development" one) > > > > > it costs me 100k. Are there maintenance costs? > > > > There are bound to be little differences in the interface that you > > will be debugging for months. > > > > Tim. It really took 12 man months? About that yes. > Lets say that I am using Ucos or > something like that. The "port" is to attach the scheduler etc to the > interrupts/timers etc etc. OK, We had a 20+ task multi-tasking system. All communications between tasks was (originally) done using calls to the RTOS which were inline. We had to replace all of these with the new OS calls. This was sensibly done by first changing the calls to use a generic interface layer. Then we had to rewrite the generic interface for the new RTOS. This was non trivial and some of the problems took much more investigation to get right than expected. One or two functions were completely impossible to resolve cleanly in the interface and we had to re-structure the task to get it to work. Then we had to debug it! We found the last one (I hope) more than six months after we had started and there were 4 people working on this part of the project. How hard the task is really depends on how similar the functionality of the two OSs is. If, for example, the new OS places a much larger restriction on what you can do from inside an interrupt than the old one did, then you have a biger re-write than you first thought. I suppose my estimate could be a bit excessive. We were moving to a new processor so some of the work that I now remember being OS support could have just been the new processor requirements. > Im not trying to write a new kernel. Is > that a 12month job? (Could be...but Ive heard of guys that did it in > 3mos :) But what type of kernel? Tim > > Thanx --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Routing problem Date: 10 Oct 2002 00:38:59 -0700 From: Gunter.Ritter@tenovis.com (Gunter Ritter) Organization: http://groups.google.com/ Message-ID: References: <1034170439.162264@newsreader2.wirehub.nl> "Renee Teunissen" wrote in message news:<1034170439.162264@newsreader2.wirehub.nl>... > Hi, > > Have you ever thought about proxy-arp? > > basically with proxy arp, you ethernet interface replies to packets > address to your dial-in lines, and therefore the address you assign > have to be in the same subnet as your ethernet interface. This how I > did this years ago with linux dial-in systems. > > So to all clients on the net think your dial-in-machine has more than > one IP, your dial-in-machine will gate it to your serial device(s). > Arp -a on your windows boxes will show the same mac-address for all > dial-in lines of that dial-in-machine with the MAC-address of the > dial-in-machine. > > Btw, you dont have to mess with default gates or static routes to make > this work. > > Proxy arp-ing is invented for this - in my humble opinion. :) > > Cheers, > Renee > - - - - - - - - > Renee A. Teunissen > PTS Software bv, Meerweg 7, 1405BA Bussum, NL. > T.+31-(0)35-6926969, M.+31-(0)6-22778313, > http://www.pts.nl, @pts.nl > personal link page: http://wittenburg10c.nl/db/dest/links.html > > > Hi, I thought about proxy arp, but in my understanding of the WRS documentation does this only work with shared memory networks. If this should work in my configuration ( I donßt know how to do that), there is a different problem with proxy arp. My system is connected to a corporate LAN and for each IP address in this LAN the customer has to pay some fees. In the case of proxy arp, I need "official" IP addresses for the subnet behind the vxworks system, so each of these systems needs two IP addresses plus one for each client (up to 8) in the subnet out of the corporate LAN. And if many of these vxworks systems are attached to the network, the IP addresses might run out in a class C network. Otherwise, using proxy arp, administration is more difficult, because each subnet behind the vxworks system must have individual IP adresses. I think the solution will be NAT. Gunter --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Routing problem Date: Thu, 10 Oct 2002 09:37:09 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Reply-To: "Leonid Rosenboim" "Gunter Ritter" wrote in message news:ac95d252.0210092317.40d9169b@posting.google.com... [snip] > One problem is, that I do not have access to the PC configuration in > the corporate network of a customer. > > This can be solved by a routing entry that for my vxworks box in the > existing default gateway. This might be done static or dynamic via > RIP. > But, this works only if I have not to many vxworks boxes and the > customer grants me access to this router. > I think that if you add RIP to the VxWorks box, it will be quote a reasonable requirement for the customer network admin to either have to enable RIP on the main router (if not already enabled) or add static routes. I can see no particular reason for the net admin to refuse RIP cooperation, except for security considerations. > But if I have many of these vxworks boxes in the main network (it can > be thousands), this constellation will not work. Assuming you have RIP, this can be done even if thousands of boxes are involved, although they won't be on the main network then, there would need to be a core backbone router, and the Boxes will be separated into segments. > > So, I think I need NAT to access the corporate LAN from the subnet > behind the vxworks box. Wel,, NAT is a viable option to consider, but keep in mind that with NAT, the dialup clients will be able to access the corporate network servers, but not the other way arround, unless an elaborate static map is implemented, and that would very much depend on the application. HTH - Leonid --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Routing problem Date: Thu, 10 Oct 2002 10:38:14 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1034246307.279480@newsreader2.wirehub.nl> References: <1034170439.162264@newsreader2.wirehub.nl> You could be right about this. I have no expiriences with proxy arp on VxWorks only with linux/unix - and after checking the manuals I guess you are right about this. NAT can be your solution, but remember NAT does not work for all protocols. Or you could build yor own "proxyarp" packet forwarder - but this gonna take some time - I guess. I would - i guess - try to hook into the ethernet packets, there should be a hook for it answer ARP requests for the IP address you have registered at your serial lines grep those packets matching the registered serial IP numbers, strip the ethernet header and forward it into your serial line. Cheers, Renee. "Gunter Ritter" wrote in message news:ac95d252.0210092338.5cd37ef9@posting.google.com... > "Renee Teunissen" wrote in message news:<1034170439.162264@newsreader2.wirehub.nl>... > > Hi, > > > > Have you ever thought about proxy-arp? > > > > basically with proxy arp, you ethernet interface replies to packets > > address to your dial-in lines, and therefore the address you assign > > have to be in the same subnet as your ethernet interface. This how I > > did this years ago with linux dial-in systems. > > > > So to all clients on the net think your dial-in-machine has more than > > one IP, your dial-in-machine will gate it to your serial device(s). > > Arp -a on your windows boxes will show the same mac-address for all > > dial-in lines of that dial-in-machine with the MAC-address of the > > dial-in-machine. > > > > Btw, you dont have to mess with default gates or static routes to make > > this work. > > > > Proxy arp-ing is invented for this - in my humble opinion. :) > > > > Cheers, > > Renee > > - - - - - - - - > > Renee A. Teunissen > > PTS Software bv, Meerweg 7, 1405BA Bussum, NL. > > T.+31-(0)35-6926969, M.+31-(0)6-22778313, > > http://www.pts.nl, @pts.nl > > personal link page: http://wittenburg10c.nl/db/dest/links.html > > > > > > > > Hi, > > I thought about proxy arp, but in my understanding of the WRS > documentation does this only work with shared memory networks. > > If this should work in my configuration ( I donßt know how to do > that), there is a different problem with proxy arp. > > My system is connected to a corporate LAN and for each IP address in > this LAN the customer has to pay some fees. In the case of proxy arp, > I need "official" IP addresses for the subnet behind the vxworks > system, so each of these systems needs two IP addresses plus one for > each client (up to 8) in the subnet out of the corporate LAN. And if > many of these vxworks systems are attached to the network, the IP > addresses might run out in a class C network. > > Otherwise, using proxy arp, administration is more difficult, because > each subnet behind the vxworks system must have individual IP > adresses. > > I think the solution will be NAT. > > > Gunter --------------------------- Newsgroups: comp.os.vxworks Subject: Re: inflate Lib.... Date: Thu, 10 Oct 2002 07:54:32 -0400 From: Joe Durusau Organization: Lockheed Martin Corporation Message-ID: <3DA56A78.310DA855@lmco.com> References: <8b53eb96.0210090955.3ff816a5@posting.google.com> ChengHsin Hsu wrote: > > Hello, > > Is there anyway that we can know how much memory we need to malloc > as dest buffer before the inflate call like this? > inflate (src, dest, nBytes); > > Thanks, > Bear. Try looking at quest.jpl.nasa.gov/zlib/ , which according to WRS is the source of inflateLib. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: How do i print unsigned long long ? Date: 10 Oct 2002 05:31:11 -0700 From: deepaks@sasken.com (Deepak Shetty) Organization: http://groups.google.com/ Message-ID: <421ece65.0210100431.41f75af9@posting.google.com> Hi, In VxWorks, i have a variable of type unsigned long long. This basically means that the size of this var is 8 bytes ( 64 bits ) Any idea how do i print the value contained in this variable using printf ? There is no supporting type ( as in %type ) that i can use in printf to print vale of a 64 bit variable ? We are using this var to store timestamp, which the std says shud be 64 bits long, hence the prob. Actually in VC++ u have a microsoft specific extension type called I64, which can be used to print 64 bit value, but such a thing is not present in VxWorks. any pointers from people who have already experienced such a prob on VxWorks ? Thanx, deepak --------------------------- Newsgroups: comp.os.vxworks Subject: IP forwarding through PPP connection Date: 10 Oct 2002 06:43:44 -0700 From: dmasse@ele.etsmtl.ca (David Masse) Organization: http://groups.google.com/ Message-ID: <14fc63a8.0210100543.7f5c95ac@posting.google.com> Hi, I've got two PPC860 targets with VxWorks AE. Each have two network connections: one standard ethernet connection and one PPP connection. The two targets are connected together through their PPP connection. Each are connected to its own network through their Ethernet connection. One PC is connect to each network. Here is a diagram of this configuration: ---------- | | | PC 1 | | | ---------- | 169.254.198.120 | NET 169.254.0.0 ----------------------------------------------- | | 169.254.198.121 ---------- |Target 1| | | | PPC860 | ---------- | 10.0.0.1 PPP Conection | | 10.0.0.2 ---------- |Target 2| | | | PPC860 | ---------- | 142.137.20.46 | NET 142.137.0.0 ----------------------------------------------- | | 142.137.20.176 ---------- | | | PC 2 | | | ---------- What am I trying to do is to access PC2 from PC1. This means that both target must forward IP packet. But for the moment, I'm not able to ping target 2 from PC1. I add the following entry in PC1 routing table: Destination: 10.0.0.0, Mask = 255.255.255.0, Gateway = 169.254.198.121 In target 1, I had the following entry in routing table with mRouteAdd(): Destination: 10.0.0.0 Mask: 255.255.255.0 Gateway: 10.0.0.1 With this configuration, I can ping 169.254.198.121 and 10.0.0.1 from PC 1 but not 10.0.0.2. I can ping 10.0.0.2 from target 1 but not from PC 1. It seems that the VxWorks target does not perform any IP forwarding on the PPP connection. Someone have an idea on what I am doing wrong here? Thanks in advance. David --------------------------- Newsgroups: comp.os.vxworks Subject: Static linking i tornado for simpc Date: 10 Oct 2002 07:14:47 -0700 From: vxworksonline@yahoo.com (Shiny) Organization: http://groups.google.com/ Message-ID: Hi, Does anybody know how to statically link application modules in Tornado for a simpc? I need to statically link the application modules b'cos of cross-reference across modules. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: 10 Oct 2002 10:49:02 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: References: <3a5ff53f.0210020748.3643d45a@posting.google.com> <3DA14F95.A74993DC@alcatel-ke.de> <3DA52D00.5F5070E@alcatel-ke.de> Sender: bpringle@DeadDuck Brian> Tim. It really took 12 man months? Tim> About that yes. Brian> Lets say that I am using Ucos or something like that. The Brian> "port" is to attach the scheduler etc to the interrupts/timers Brian> etc etc. Tim> OK, We had a 20+ task multi-tasking system. All communications Tim> between tasks was (originally) done using calls to the RTOS Tim> which were inline. We had to replace all of these with the new People are talking about many different things here. One is porting an OS to new hardware. Ie, writing a BSP for an OS, another is to port an OS to a new CPU/Architechure and another is to port applications to the new OS. I wouldn't consider a port of the OS, unless a significant peice of software was available for the OS that you didn't have on the original. The equation of simple man-month versus OS pricing does not always equate. There are issued of lost product revenue, market timing, resource availability, etc. That influence the decision on a per company basis. Porting an OS is rarely related to a companies main business interests. fwiw, Bill Pringlemeir. - -- ANNOYED BY CORVETTE. WEB PAGES. SINKING. U-449. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How do i print unsigned long long ? Date: Thu, 10 Oct 2002 11:52:34 -0400 From: Joe Durusau Organization: Lockheed Martin Corporation Message-ID: <3DA5A242.20F05B79@lmco.com> References: <421ece65.0210100431.41f75af9@posting.google.com> Deepak Shetty wrote: > > Hi, > In VxWorks, i have a variable of type unsigned long long. > This basically means that the size of this var is 8 bytes ( 64 bits ) > > Any idea how do i print the value contained in this variable using printf ? > > There is no supporting type ( as in %type ) that i can use in printf to > print vale of a 64 bit variable ? > > We are using this var to store timestamp, which the std says shud be 64 bits > long, hence the prob. > > Actually in VC++ u have a microsoft specific extension type called I64, > which can be used to print 64 bit value, but such a thing is not present in > VxWorks. > > any pointers from people who have already experienced such a prob on VxWorks > ? > > Thanx, > deepak The only way I've seen it doen is by splitting it into two 32-bit vars, and putting them side-by-side, at least with the gnu compiler and assoc. libs. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Hardware Flow Control for Serial Ports Date: Thu, 10 Oct 2002 11:30:44 -0700 From: "Dan Pinson" Message-ID: References: Michael, Thanks for the warning but it looks to me as if the FIFOs are enabled in the ns16550 driver. From ns16550Open(): REG(FCR, pChan) = (RxCLEAR | TxCLEAR | FIFO_ENABLE); Regards, Dan "Michael Lawnick" wrote in message news:ao34ct$fdr$1@snoopy.bndlg.de... > Hi Dan, > > didn't really get through your post, just the standard injection when UART > is concerned: > Did you enable its FIFO? Remember that the standard vxWorks driver doesn't > do that. You'll have to patch the code in src/drv/sio along your needs. > > HTH > Michael > > "Dan Pinson" schrieb im Newsbeitrag > news:ao2fpu$ijho7$1@ID-48686.news.dfncis.de... > > I'm receiving raw data on a serial port using VxWorks standard driver. I > > suspect that the receive FIFO is being overrun. Here is the situation: > > > > PentiumPC BSP > > 16550 UART > > > > After opening the port, I enable hardware flow control via ioctl. After > > receiving at 115K for awhile, I get unexpected data. The same device works > > OK on another system under a different OS. This is just an indicator; not > > proof. > > > > I checked the highwater mark in my RX ring buffer. It was well below the > > free space. > > > > In looking at the 16550 serial drivers target/src/drv/sio/ns16550Serial.c > > and ns16550Sio.c, I see how transmitting is throttled by CTS but not how > RTS > > is manipulated. I notice that the UART driver does nothing with an rx > > overrun interrupt. I thought tyLib might clear RTS when the buffer gets > > full, but I don't see an API in the UART driver that could be used. > > > > Bottom-line is that I wonder if RTS/CTS flow control works on the x86 BSP > > and/or with the 16550 drivers. > > > > Thanks in advance, > > Dan > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Hardware Flow Control for Serial Ports Date: Thu, 10 Oct 2002 11:40:16 -0700 From: "Dan Pinson" Message-ID: References: Woops! The x86 BSPs are using the 8250 drivers. Criminy, when are they going to move up to the modern world? Anyway, thanks. I'm probably being overrun because the FIFO is not being used. ..Dan "Dan Pinson" wrote in message news:ao4h14$j6hnj$1@ID-48686.news.dfncis.de... > Michael, > > Thanks for the warning but it looks to me as if the FIFOs are enabled in the > ns16550 driver. From ns16550Open(): > > REG(FCR, pChan) = (RxCLEAR | TxCLEAR | FIFO_ENABLE); > > Regards, > Dan > > > "Michael Lawnick" wrote in message > news:ao34ct$fdr$1@snoopy.bndlg.de... > > Hi Dan, > > > > didn't really get through your post, just the standard injection when UART > > is concerned: > > Did you enable its FIFO? Remember that the standard vxWorks driver doesn't > > do that. You'll have to patch the code in src/drv/sio along your needs. > > > > HTH > > Michael > > > > "Dan Pinson" schrieb im Newsbeitrag > > news:ao2fpu$ijho7$1@ID-48686.news.dfncis.de... > > > I'm receiving raw data on a serial port using VxWorks standard driver. I > > > suspect that the receive FIFO is being overrun. Here is the situation: > > > > > > PentiumPC BSP > > > 16550 UART > > > > > > After opening the port, I enable hardware flow control via ioctl. After > > > receiving at 115K for awhile, I get unexpected data. The same device > works > > > OK on another system under a different OS. This is just an indicator; > not > > > proof. > > > > > > I checked the highwater mark in my RX ring buffer. It was well below the > > > free space. > > > > > > In looking at the 16550 serial drivers > target/src/drv/sio/ns16550Serial.c > > > and ns16550Sio.c, I see how transmitting is throttled by CTS but not how > > RTS > > > is manipulated. I notice that the UART driver does nothing with an rx > > > overrun interrupt. I thought tyLib might clear RTS when the buffer gets > > > full, but I don't see an API in the UART driver that could be used. > > > > > > Bottom-line is that I wonder if RTS/CTS flow control works on the x86 > BSP > > > and/or with the 16550 drivers. > > > > > > Thanks in advance, > > > Dan > > > > > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Help: Random Reboots of target Date: 10 Oct 2002 16:25:57 -0700 From: rkintada@yahoo.com (Rama) Organization: http://groups.google.com/ Message-ID: <496ea65b.0210101525.94a11bc@posting.google.com> Hi I have an application running on PPC850 processor. The target reboots itselt randomly. I ran a few tests and ran memShow periodically. This did not show any memory leaks. I was running the spy command also, and i did not see any thing abnormal. There is no watch dog thread running. I would like get some suggestions/comments on how to debug this. Is there a way to capture the reboot condition via debugger? I have to run my application for a while as the reboots happen at random times. Thanks in advance - - Ramakrishna --------------------------- Newsgroups: comp.os.vxworks Subject: Re: x86 pci interrupt sharing Date: 10 Oct 2002 16:34:05 -0700 From: goose23@spu.edu (Matthew Peltzer) Organization: http://groups.google.com/ Message-ID: <75a1cd5a.0210101534.110d9ba4@posting.google.com> References: <75a1cd5a.0210091218.38d18669@posting.google.com> <3DA49BA1.5E7F7242@hotmail.com> George, Thanks for your help. It took a lot of mucking around in the vxWorks sources, but I have managed to solve my problems. To answer your question (finally, sorry), pciIntLibInit() is being called from within sysHwInit(). My problems turned out to be on the uninstall side of things, not the install. I had unconsciously made the assumption that the ISR chaining was done on a device basis, rather than a driver basis. So, I could install my ISR routine twice, but when I tried to uninstall either routine, only the first would be removed. I can only attribute vxWorks crashing & rebooting to the stale entry that was never getting called. My guess is that when the second card raised its interrupt line, vxWorks ended up in something that is not a function. My solution to all this is to keep a table of all the devices listening to interrups, and connect/enable on the first of each irq, and disable/disconnect on the last of each irq. My parameter to the ISR is the irq for the isr. Since the device table is global, I can iterate through each slot and see which device raised the interrupt for that irq. Again, thanks. Matthew Peltzer --------------------------- Newsgroups: comp.os.vxworks Subject: Re: can't load .out files with dosfs2 Date: 10 Oct 2002 16:59:48 -0700 From: term@xerosoft.com (Tom) Organization: http://groups.google.com/ Message-ID: <15cdb0fb.0210101559.2fa61919@posting.google.com> References: <15cdb0fb.0210021822.572bd28c@posting.google.com> <15cdb0fb.0210090822.52841a20@posting.google.com> Leonid, I tried the test you proposed and loaded the .out file into a pipe, set stdin to be the pipes FD, and then did an ld with name == NULL so that it would try to load from stdin. This was all done in one task, I just made sure the pipe was very large; this isn't a problem is it? Unfortunately the call to ld failed. I hadn't really thought about the lseek implementation being an issue, but I think you are on to something. I ran this test again but set a breakpoint on lseek (right before ld is called). What I saw was that the first time lseek gets called it is trying to do a seek on STDIN (fd = 0) to offset 300 from the begining of the file. This seems valid but I don't believe pipes support seeking, so the call returns a - -1. Maybe the version of dosFs that I have has a bug in its lseek implementation? What do you think? I guess I could write my own pipes style driver that supports lseek and use that. Can you think of any other things to try? Thanks for the help, - -Tom "Leonid Rosenboim" wrote in message news:... > This all seems so strange that it is probably something > very simple we all have ignored. > > If you have the exact same file effectively loading > from a network, there should be no reason at all > this file won't be loading from dosFs, well except > that different file systems handle the lseek() call > differently, and most architectures dynamic > loader will use lseek(). > > How about if you create a pipe, copy the file > into the pipe, while another task runs loadModule > with the pipe being its stdin ? (I suppose you iknow > how to do this). > > Keep me posted. > ----------------------------------------------------------------------- > Leonid Rosenboim Visit: http://www.masada2000.org/ > Consultant Email: my first name at consultant dot com > > > "Tom" wrote in message > news:15cdb0fb.0210090822.52841a20@posting.google.com... > > Hey Leonid, > > > > I have tried a number of different command lines to load the file, but > > I only showed one in my first e-mail. > > > > I also tried: > > > > ld 1,1,"test.out" > > ld 1,1,"/tffs0/test.out" > > ld < test.out > > ld < /tffs0/test.out > > and loadModule() with a file descriptor that I manually opened. > > > > unfortunatelly they all failed. > > > > Any other ideas? > > > > > > Best regards, > > > > Tom > > > > "Leonid Rosenboim" wrote in message > news:... > > > Could you try to replace this command: > > > > > > > -> ld < test.out > > > > ld error: error reading file (errno = 0x3d0001). > > > > value = 0 = 0x0 > > > > > > with this command ? > > > > > > -> ld "/tffs0/test.out" > > > > > > or this command ? > > > > > > -> ld < /tffs0/test.out > > > > > > -- HTH > > > ----------------------------------------------------------------------- > > > Leonid Rosenboim Visit: http://www.masada2000.org/ > > > Consultant Email: my first name at consultant dot > com > > > > > > > > > "Tom" wrote in message > > > news:15cdb0fb.0210021822.572bd28c@posting.google.com... > > > > This is a repost of a previous message I had posted with new > > > > information included. > > > > > > > > I have been working on getting a strong arm based motherboard to run > > > > vxworks. One problem I have run into is that I can't load any .out > > > > files off of my TFFS (True Flash File System). When I try to it just > > > > gives me an error saying "error reading file (errno = 0x3d0001).". I > > > > looked up the errno and it resolved to S_objLib_OBJ_ID_ERROR , but > > > > this is generated from something at a low level that I don't have the > > > > source code to. I recently tried stepping through the assembler code > > > > to see where this error was coming from and it turns out errno is not > > > > set until the cleanup code runs after fioRead fails, so this is errno > > > > is irrelevant. > > > > > > > > Every other operation with the TFFS seems to work just fine. I can > > > > create, delete, read, write, and rename files. As an example I tried > > > > copying an out file off my FTP server onto flash and loading it, but > > > > it failed. Then after that I copied the file off of the TFFS to a new > > > > file name on my FTP server and then loaded it from that file (over the > > > > network) and it worked fine. > > > > > > > > -- start dump -- > > > > > > > > -> cd "/tffs0/" > > > > value = 0 = 0x0 > > > > -> ll > > > > -rwxrwxrwx 1 0 0 10064 Jan 1 1980 test.out > > > > value = 0 = 0x0 > > > > -> ld < test.out > > > > ld error: error reading file (errno = 0x3d0001). > > > > value = 0 = 0x0 > > > > -> copy "/tffs0/test.out", "host:newtest.out" > > > > Copy OK: 10064 bytes copied > > > > value = 0 = 0x0 > > > > -> ld < host:newtest.out > > > > value = 12531032 = 0xbf3558 > > > > -> > > > > > > > > -- end dump -- > > > > > > > > > > > > After these tests I thought the problem might have been size related, > > > > so I tried to create a minimal "hello world" out file, but that didn't > > > > work either. > > > > > > > > As a work around I thought I would try setting up a runscript that > > > > creates a ram drive and then copies each file to the ramdrive, loads > > > > from there, deletes the file, and then continues on to the next one > > > > until it is done. Unfortunatelly this failed with the same error > > > > message that I got when I loaded off of a TFFS. The only thing I can > > > > think of that the ramdrive and TFFS have in common that a ftp file > > > > system doesn't is that they use dosFS (version 2). > > > > > > > > Has anyone else had any problems like this? I'm running > > > > VxWorks 5.4.2 with Tornado 2.1.1. > > > > > > > > > > > > Best Regards, > > > > > > > > Tom --------------------------- Newsgroups: comp.os.vxworks Subject: sending out Ethernet packets Date: Fri, 11 Oct 2002 10:25:59 +0800 From: "Yung" Organization: IMS Netvigator Message-ID: for the purpose of debugging the network driver, i'd like to form an Ethernet packet myself and push it out to see if it gets through. Suppose there is a MUX-related API that does the job? Can anybody suggest a method? Thanks so much! --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Help: Random Reboots of target Date: Fri, 11 Oct 2002 03:55:47 GMT From: "H. E. Taylor" Organization: MTS Internet Message-ID: <3DA6661D.30C4@despam.autobahn.mb.ca> References: <496ea65b.0210101525.94a11bc@posting.google.com> In article <496ea65b.0210101525.94a11bc@posting.google.com>, Rama wrote: > > I have an application running on PPC850 processor. The target reboots > itselt randomly. I ran a few tests and ran memShow periodically. This > did not show any memory leaks. I was running the spy command also, > and i did not see any thing abnormal. There is no watch dog thread > running. > > I would like get some suggestions/comments on how to debug this. Is > there a way to capture the reboot condition via debugger? I have to > run my application for a while as the reboots happen at random times. > This can be a most trying 'learning experience'. Do you trust your hardware? ie. If you don't run your app, is it stable? I have seen a couple of hardware techniques you might use. Scoping the bus is one possibility, or possibly a logic analyzer. Weird stuff like PS problems need to be eliminated,as well. If you're down to the code, the technique to follow is divide and conquer. What can you pull out of your app code and still get the ittermittent crash? What sections of code can you have your app not exercise & still get the crash? While your testers are exploring this, another group can be starting code analyses. Appoint one person to keep track of all hypotheses and modify and/or verify code to elimate the possibility. Getting to the point of being able to replicate the problem is nice, but not sufficient. - -het - -- "progress in software has not followed Moore's law." -John Holland Computer Links: http://www.autobahn.mb.ca/~het/clinks.html H.E. Taylor http://www.autobahn.mb.ca/~het/ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How do i print unsigned long long ? Date: 10 Oct 2002 21:42:37 -0700 From: deepaks@sasken.com (Deepak Shetty) Organization: http://groups.google.com/ Message-ID: <421ece65.0210102042.426497eb@posting.google.com> References: <421ece65.0210100431.41f75af9@posting.google.com> <3DA5A242.20F05B79@lmco.com> > > The only way I've seen it doen is by splitting it into two 32-bit > vars, and putting them side-by-side, at least with the gnu > compiler and assoc. libs. > > Speaking only for myself, > > Joe Durusau I have one more basic doubt here, hoep u can clarify me on this. In C std, the std says that operands of the bitwise operations are promoted to the integer data type, and the result ( intermediate ) is also stored in an interger data type. SO when i use bitwise shift operators with unsigned long long, which is of bigger size than the integer data type, how does the result ( intermediate ) get stored ? WHat actually happens inside ? deepak --------------------------- Newsgroups: comp.os.vxworks Subject: Re: can't load .out files with dosfs2 Date: Fri, 11 Oct 2002 08:50:38 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <15cdb0fb.0210021822.572bd28c@posting.google.com> <15cdb0fb.0210090822.52841a20@posting.google.com> <15cdb0fb.0210101559.2fa61919@posting.google.com> Reply-To: "Michael Lawnick" Sender: mlawnick@p5088f598.dip.t-dialin.net Hi Tom, (Leonid: sorry, don't want to break you) Tom: just to get you forward: whats about memDrv instead of pipe? ramDrv bases on dosFs AFAIK, but memDrv should give you a plain memory driver that should do the same trick as pipe. Just my 2 cents. Michael "Tom" schrieb im Newsbeitrag news:15cdb0fb.0210101559.2fa61919@posting.google.com... > Leonid, > > I tried the test you proposed and loaded the .out file into a pipe, > set stdin to be the pipes FD, and then did an ld with name == NULL so > that it would try to load from stdin. This was all done in one task, > I just made sure the pipe was very large; this isn't a problem is it? > Unfortunately the call to ld failed. > > I hadn't really thought about the lseek implementation being an issue, > but I think you are on to something. I ran this test again but set a > breakpoint on lseek (right before ld is called). What I saw was that > the first time lseek gets called it is trying to do a seek on STDIN > (fd = 0) to offset 300 from the begining of the file. This seems > valid but I don't believe pipes support seeking, so the call returns a > -1. > > Maybe the version of dosFs that I have has a bug in its lseek > implementation? What do you think? I guess I could write my own > pipes style driver that supports lseek and use that. Can you think of > any other things to try? > > Thanks for the help, > > -Tom > > "Leonid Rosenboim" wrote in message news:... > > This all seems so strange that it is probably something > > very simple we all have ignored. > > > > If you have the exact same file effectively loading > > from a network, there should be no reason at all > > this file won't be loading from dosFs, well except > > that different file systems handle the lseek() call > > differently, and most architectures dynamic > > loader will use lseek(). > > > > How about if you create a pipe, copy the file > > into the pipe, while another task runs loadModule > > with the pipe being its stdin ? (I suppose you iknow > > how to do this). > > > > Keep me posted. > > ----------------------------------------------------------------------- > > Leonid Rosenboim Visit: http://www.masada2000.org/ > > Consultant Email: my first name at consultant dot com > > > > > > "Tom" wrote in message > > news:15cdb0fb.0210090822.52841a20@posting.google.com... > > > Hey Leonid, > > > > > > I have tried a number of different command lines to load the file, but > > > I only showed one in my first e-mail. > > > > > > I also tried: > > > > > > ld 1,1,"test.out" > > > ld 1,1,"/tffs0/test.out" > > > ld < test.out > > > ld < /tffs0/test.out > > > and loadModule() with a file descriptor that I manually opened. > > > > > > unfortunatelly they all failed. > > > > > > Any other ideas? > > > > > > > > > Best regards, > > > > > > Tom > > > > > > "Leonid Rosenboim" wrote in message > > news:... > > > > Could you try to replace this command: > > > > > > > > > -> ld < test.out > > > > > ld error: error reading file (errno = 0x3d0001). > > > > > value = 0 = 0x0 > > > > > > > > with this command ? > > > > > > > > -> ld "/tffs0/test.out" > > > > > > > > or this command ? > > > > > > > > -> ld < /tffs0/test.out > > > > > > > > -- HTH > > > > ----------------------------------------------------------------------- > > > > Leonid Rosenboim Visit: http://www.masada2000.org/ > > > > Consultant Email: my first name at consultant dot > > com > > > > > > > > > > > > "Tom" wrote in message > > > > news:15cdb0fb.0210021822.572bd28c@posting.google.com... > > > > > This is a repost of a previous message I had posted with new > > > > > information included. > > > > > > > > > > I have been working on getting a strong arm based motherboard to run > > > > > vxworks. One problem I have run into is that I can't load any .out > > > > > files off of my TFFS (True Flash File System). When I try to it just > > > > > gives me an error saying "error reading file (errno = 0x3d0001).". I > > > > > looked up the errno and it resolved to S_objLib_OBJ_ID_ERROR , but > > > > > this is generated from something at a low level that I don't have the > > > > > source code to. I recently tried stepping through the assembler code > > > > > to see where this error was coming from and it turns out errno is not > > > > > set until the cleanup code runs after fioRead fails, so this is errno > > > > > is irrelevant. > > > > > > > > > > Every other operation with the TFFS seems to work just fine. I can > > > > > create, delete, read, write, and rename files. As an example I tried > > > > > copying an out file off my FTP server onto flash and loading it, but > > > > > it failed. Then after that I copied the file off of the TFFS to a new > > > > > file name on my FTP server and then loaded it from that file (over the > > > > > network) and it worked fine. > > > > > > > > > > -- start dump -- > > > > > > > > > > -> cd "/tffs0/" > > > > > value = 0 = 0x0 > > > > > -> ll > > > > > -rwxrwxrwx 1 0 0 10064 Jan 1 1980 test.out > > > > > value = 0 = 0x0 > > > > > -> ld < test.out > > > > > ld error: error reading file (errno = 0x3d0001). > > > > > value = 0 = 0x0 > > > > > -> copy "/tffs0/test.out", "host:newtest.out" > > > > > Copy OK: 10064 bytes copied > > > > > value = 0 = 0x0 > > > > > -> ld < host:newtest.out > > > > > value = 12531032 = 0xbf3558 > > > > > -> > > > > > > > > > > -- end dump -- > > > > > > > > > > > > > > > After these tests I thought the problem might have been size related, > > > > > so I tried to create a minimal "hello world" out file, but that didn't > > > > > work either. > > > > > > > > > > As a work around I thought I would try setting up a runscript that > > > > > creates a ram drive and then copies each file to the ramdrive, loads > > > > > from there, deletes the file, and then continues on to the next one > > > > > until it is done. Unfortunatelly this failed with the same error > > > > > message that I got when I loaded off of a TFFS. The only thing I can > > > > > think of that the ramdrive and TFFS have in common that a ftp file > > > > > system doesn't is that they use dosFS (version 2). > > > > > > > > > > Has anyone else had any problems like this? I'm running > > > > > VxWorks 5.4.2 with Tornado 2.1.1. > > > > > > > > > > > > > > > Best Regards, > > > > > > > > > > Tom --------------------------- Newsgroups: comp.os.vxworks Subject: Re: inflate Lib.... Date: Fri, 11 Oct 2002 08:59:30 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <8b53eb96.0210090955.3ff816a5@posting.google.com> Reply-To: "Michael Lawnick" Sender: mlawnick@p5088f598.dip.t-dialin.net Hi Bear, definite answer: no, you'll have to go through the algorithm :-( The output length depends on the amount of information within your data. It's just the representation that is changed on compression. To get the length of output, you'd have to go through input and calculate - just what compression already does. What you could do: inflate to NULL, get the amount of output data and then do allocation :-( Regards, Michael "ChengHsin Hsu" schrieb im Newsbeitrag news:8b53eb96.0210090955.3ff816a5@posting.google.com... > Hello, > > Is there anyway that we can know how much memory we need to malloc > as dest buffer before the inflate call like this? > inflate (src, dest, nBytes); > > Thanks, > Bear. --------------------------- Newsgroups: comp.os.vxworks Subject: What is needed for symFindByName? Date: 11 Oct 2002 00:45:09 -0700 From: glenn_is_cool@hotmail.com (Andreas F) Organization: http://groups.google.com/ Message-ID: Hi.. I'm having a problem with 'symFindByName'. It just doesn't work! I know that you need to synchronize host and target symbol tables, so that is taken care of. My question is really what components I need to include to my VxWorks image. And is there any other pitfalls with this function? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: can't load .out files with dosfs2 Date: Fri, 11 Oct 2002 09:50:01 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <15cdb0fb.0210021822.572bd28c@posting.google.com> <15cdb0fb.0210090822.52841a20@posting.google.com> <15cdb0fb.0210101559.2fa61919@posting.google.com> Reply-To: "Leonid Rosenboim" Good work, Tom. I do recall something about a certain patch version of dosFs2 that has a problem in lseek(), and the reason I suggested you try pipes was to doube check that this ld indeed uses lseek. It does. And it clearly shows what i exactly different beween simply copying the file and loading it as a module. What you need to do now is two things: 1. With the help of WRS support (sorry you have to do this), starting with on-line, see if there is an SPR about dosFs2 and lseek, and if you find it, get the patch. 2. If #1 does not work, write s simple test program that first creates a sizable file, and initializes it with very simple, expectable data, e.g. every 32-bit word in the file would simple contain its offset relative stanrt of file. Then the program begins a series of lseek() calls using random offsets modulus the file size, reads a word from that offset and verifies it is what was expected. I am pretty sure this will demonstrate the bug if there is indeed an lseek bug. You might get away with a simple script that repeats the calls made by ld, e.g. read 48 bytes from beginning, lseek 300, read 128 bytes, ... I had a test like that in the fstest suite we developed in parallel with dosFs2 (actually this suite begun with cdromFs, later was extended for FlashLib and then extended more for patching dosFs1 and dosFs2 development, but I figure nobody at WRS relelbers this suite ever existed, bammer!) 3. If you are bored in the mean time, you could also creatre a ramDisk and try to copy the file over and load it from ramDisk, but most chanses are it will fail too becase it would use the same dosFs2. Just let me know when you need my address to mail me a check ;-) - ----------------------------------------------------------------------- Leonid Rosenboim Visit: http://www.masada2000.org/ Consultant Email: my first name at consultant dot com "Tom" wrote in message news:15cdb0fb.0210101559.2fa61919@posting.google.com... > Leonid, > > I tried the test you proposed and loaded the .out file into a pipe, > set stdin to be the pipes FD, and then did an ld with name == NULL so > that it would try to load from stdin. This was all done in one task, > I just made sure the pipe was very large; this isn't a problem is it? > Unfortunately the call to ld failed. > > I hadn't really thought about the lseek implementation being an issue, > but I think you are on to something. I ran this test again but set a > breakpoint on lseek (right before ld is called). What I saw was that > the first time lseek gets called it is trying to do a seek on STDIN > (fd = 0) to offset 300 from the begining of the file. This seems > valid but I don't believe pipes support seeking, so the call returns a > -1. > > Maybe the version of dosFs that I have has a bug in its lseek > implementation? What do you think? I guess I could write my own > pipes style driver that supports lseek and use that. Can you think of > any other things to try? > > Thanks for the help, > > -Tom > > "Leonid Rosenboim" wrote in message news:... > > This all seems so strange that it is probably something > > very simple we all have ignored. > > > > If you have the exact same file effectively loading > > from a network, there should be no reason at all > > this file won't be loading from dosFs, well except > > that different file systems handle the lseek() call > > differently, and most architectures dynamic > > loader will use lseek(). > > > > How about if you create a pipe, copy the file > > into the pipe, while another task runs loadModule > > with the pipe being its stdin ? (I suppose you iknow > > how to do this). > > > > Keep me posted. > > ----------------------------------------------------------------------- > > Leonid Rosenboim Visit: http://www.masada2000.org/ > > Consultant Email: my first name at consultant dot com > > > > > > "Tom" wrote in message > > news:15cdb0fb.0210090822.52841a20@posting.google.com... > > > Hey Leonid, > > > > > > I have tried a number of different command lines to load the file, but > > > I only showed one in my first e-mail. > > > > > > I also tried: > > > > > > ld 1,1,"test.out" > > > ld 1,1,"/tffs0/test.out" > > > ld < test.out > > > ld < /tffs0/test.out > > > and loadModule() with a file descriptor that I manually opened. > > > > > > unfortunatelly they all failed. > > > > > > Any other ideas? > > > > > > > > > Best regards, > > > > > > Tom > > > > > > "Leonid Rosenboim" wrote in message > > news:... > > > > Could you try to replace this command: > > > > > > > > > -> ld < test.out > > > > > ld error: error reading file (errno = 0x3d0001). > > > > > value = 0 = 0x0 > > > > > > > > with this command ? > > > > > > > > -> ld "/tffs0/test.out" > > > > > > > > or this command ? > > > > > > > > -> ld < /tffs0/test.out > > > > > > > > -- HTH > > > > ----------------------------------------------------------------------- > > > > Leonid Rosenboim Visit: http://www.masada2000.org/ > > > > Consultant Email: my first name at consultant dot > > com > > > > > > > > > > > > "Tom" wrote in message > > > > news:15cdb0fb.0210021822.572bd28c@posting.google.com... > > > > > This is a repost of a previous message I had posted with new > > > > > information included. > > > > > > > > > > I have been working on getting a strong arm based motherboard to run > > > > > vxworks. One problem I have run into is that I can't load any .out > > > > > files off of my TFFS (True Flash File System). When I try to it just > > > > > gives me an error saying "error reading file (errno = 0x3d0001).". I > > > > > looked up the errno and it resolved to S_objLib_OBJ_ID_ERROR , but > > > > > this is generated from something at a low level that I don't have the > > > > > source code to. I recently tried stepping through the assembler code > > > > > to see where this error was coming from and it turns out errno is not > > > > > set until the cleanup code runs after fioRead fails, so this is errno > > > > > is irrelevant. > > > > > > > > > > Every other operation with the TFFS seems to work just fine. I can > > > > > create, delete, read, write, and rename files. As an example I tried > > > > > copying an out file off my FTP server onto flash and loading it, but > > > > > it failed. Then after that I copied the file off of the TFFS to a new > > > > > file name on my FTP server and then loaded it from that file (over the > > > > > network) and it worked fine. > > > > > > > > > > -- start dump -- > > > > > > > > > > -> cd "/tffs0/" > > > > > value = 0 = 0x0 > > > > > -> ll > > > > > -rwxrwxrwx 1 0 0 10064 Jan 1 1980 test.out > > > > > value = 0 = 0x0 > > > > > -> ld < test.out > > > > > ld error: error reading file (errno = 0x3d0001). > > > > > value = 0 = 0x0 > > > > > -> copy "/tffs0/test.out", "host:newtest.out" > > > > > Copy OK: 10064 bytes copied > > > > > value = 0 = 0x0 > > > > > -> ld < host:newtest.out > > > > > value = 12531032 = 0xbf3558 > > > > > -> > > > > > > > > > > -- end dump -- > > > > > > > > > > > > > > > After these tests I thought the problem might have been size related, > > > > > so I tried to create a minimal "hello world" out file, but that didn't > > > > > work either. > > > > > > > > > > As a work around I thought I would try setting up a runscript that > > > > > creates a ram drive and then copies each file to the ramdrive, loads > > > > > from there, deletes the file, and then continues on to the next one > > > > > until it is done. Unfortunatelly this failed with the same error > > > > > message that I got when I loaded off of a TFFS. The only thing I can > > > > > think of that the ramdrive and TFFS have in common that a ftp file > > > > > system doesn't is that they use dosFS (version 2). > > > > > > > > > > Has anyone else had any problems like this? I'm running > > > > > VxWorks 5.4.2 with Tornado 2.1.1. > > > > > > > > > > > > > > > Best Regards, > > > > > > > > > > Tom --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Random Reboots of target Date: Fri, 11 Oct 2002 09:56:40 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <496ea65b.0210101525.94a11bc@posting.google.com> Reply-To: "Leonid Rosenboim" "Rama" wrote in message news:496ea65b.0210101525.94a11bc@posting.google.com... > Hi > > I have an application running on PPC850 processor. The target reboots > itselt randomly. I ran a few tests and ran memShow periodically. This > did not show any memory leaks. I was running the spy command also, > and i did not see any thing abnormal. There is no watch dog thread > running. > > I would like get some suggestions/comments on how to debug this. Is > there a way to capture the reboot condition via debugger? I have to > run my application for a while as the reboots happen at random times. > > Thanks in advance > > - Ramakrishna This is trhe price you pay for hasty design, sloppy programming and dirt cheap inexperienced engineers lacking motivation. Now you need to get someone with a lot of experience to work with you on this, which can easily take several months, because of the need to manually review all of your application code, to find ALL those memory corruptions and stray pointers lurking in there. - ----------------------------------------------------------------------- Leonid Rosenboim Visit: http://www.masada2000.org/ Consultant Email: my first name at consultant dot com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How do i print unsigned long long ? Date: Fri, 11 Oct 2002 09:53:09 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <421ece65.0210100431.41f75af9@posting.google.com> <3DA5A242.20F05B79@lmco.com> <421ece65.0210102042.426497eb@posting.google.com> Reply-To: "Leonid Rosenboim" "Deepak Shetty" wrote in message news:421ece65.0210102042.426497eb@posting.google.com... > > > > The only way I've seen it doen is by splitting it into two 32-bit > > vars, and putting them side-by-side, at least with the gnu > > compiler and assoc. libs. > > > > Speaking only for myself, > > > > Joe Durusau > > I have one more basic doubt here, hoep u can clarify me on this. > In C std, the std says that operands of the bitwise operations are > promoted to the integer data type, and the result ( intermediate ) is > also stored in an interger data type. SO when i use bitwise shift > operators with unsigned long long, which is of bigger size than the > integer data type, how does the result ( intermediate ) get stored ? > WHat actually happens inside ? > > deepak If I recall correctly, the standard also says that intermediate results are promoted to the largest integer present in the expression, so whenever a long long is present, the intermediate will be too 64-bit. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Random Reboots of target Date: Fri, 11 Oct 2002 11:35:58 +0100 From: "David" Message-ID: <3da69b7f$0$1382$4d4ef98e@read.news.ch.uu.net> References: <496ea65b.0210101525.94a11bc@posting.google.com> Hi probably you get an unexpected interrupt that makes the target reboot. make a test function capturing all interrupts that you do not expect Example #define EXC_OFF_ALIGN 0x00600 /* alignment */ excIntConnect((VOIDFUNCPTR *) EXC_OFF_ALIGN, reinterpret_cast(intCatcher)); This example is for power PC This will call the function intCatcher if an alignement error occours. See your VxWorks manual for the other ints "Rama" schrieb im Newsbeitrag news:496ea65b.0210101525.94a11bc@posting.google.com... > Hi > > I have an application running on PPC850 processor. The target reboots > itselt randomly. I ran a few tests and ran memShow periodically. This > did not show any memory leaks. I was running the spy command also, > and i did not see any thing abnormal. There is no watch dog thread > running. > > I would like get some suggestions/comments on how to debug this. Is > there a way to capture the reboot condition via debugger? I have to > run my application for a while as the reboots happen at random times. > > Thanks in advance > > - Ramakrishna --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How do i print unsigned long long ? Date: Fri, 11 Oct 2002 06:39:56 -0400 From: joe durusau Organization: WEBUSENET.com Message-ID: <3DA6AA7C.A3E5CF3@bellsouth.net> References: <421ece65.0210100431.41f75af9@posting.google.com> <3DA5A242.20F05B79@lmco.com> <421ece65.0210102042.426497eb@posting.google.com> Leonid Rosenboim wrote: > "Deepak Shetty" wrote in message > news:421ece65.0210102042.426497eb@posting.google.com... > > > > > > The only way I've seen it doen is by splitting it into two 32-bit > > > vars, and putting them side-by-side, at least with the gnu > > > compiler and assoc. libs. > > > > > > Speaking only for myself, > > > > > > Joe Durusau > > > > I have one more basic doubt here, hoep u can clarify me on this. > > In C std, the std says that operands of the bitwise operations are > > promoted to the integer data type, and the result ( intermediate ) is > > also stored in an interger data type. SO when i use bitwise shift > > operators with unsigned long long, which is of bigger size than the > > integer data type, how does the result ( intermediate ) get stored ? > > WHat actually happens inside ? > > > > deepak > > If I recall correctly, the standard also says that intermediate results are > promoted to the largest integer present in the expression, so whenever > a long long is present, the intermediate will be too 64-bit. If I were you, I wouldn't put too much faith in the standard, esp. if you are using the gnu compiler that wrs ships. It is very old, and may not support long long at all. You can find out what it's doing by looking at the assembler output. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Help: Random Reboots of target Date: 11 Oct 2002 04:20:41 -0700 From: demralarood@hotmail.com (Demrala Rood) Organization: http://groups.google.com/ Message-ID: <3baeb99.0210110320.7106bbd1@posting.google.com> References: <496ea65b.0210101525.94a11bc@posting.google.com> rkintada@yahoo.com (Rama) wrote in message news:<496ea65b.0210101525.94a11bc@posting.google.com>... > Hi > > I have an application running on PPC850 processor. The target reboots > itselt randomly. I ran a few tests and ran memShow periodically. This > did not show any memory leaks. I was running the spy command also, > and i did not see any thing abnormal. There is no watch dog thread > running. > > I would like get some suggestions/comments on how to debug this. Is > there a way to capture the reboot condition via debugger? I have to > run my application for a while as the reboots happen at random times. Buy a visionProbe or visionIce and visionClick debugger. It costs less to buy proper tools than it does to flounder in the dark for a few weeks. DR --------------------------- Newsgroups: comp.os.vxworks Subject: change the console port in vxWorks Date: Fri, 11 Oct 2002 16:13:55 +0400 From: "Denis Dubrovin" Organization: WEBPlus Ltd. Message-ID: Hi! I need to change the console port from /tyCo/0 to /tyCo/1 prior to starting the shell. I edited the configAll.h file and changed: #define CONSOLE_TTY 0 /* console channel */ to: #define CONSOLE_TTY 1 /* console channel */ I rebuild, restart, and I get zip. No activity what so ever on the console. Any idea what the heck I am missing? - -- Regards! Den --------------------------- Newsgroups: comp.os.vxworks Subject: Re: change the console port in vxWorks Date: Fri, 11 Oct 2002 12:57:53 +0000 (UTC) From: Johan Borkhuis Organization: Proxim Corp. Message-ID: References: "Denis Dubrovin" wrote: > I need to change the console port from /tyCo/0 to /tyCo/1 prior to > starting the shell. > > I edited the configAll.h file and changed: > #define CONSOLE_TTY 0 /* console channel */ > > to: > > #define CONSOLE_TTY 1 /* console channel */ > > I rebuild, restart, and I get zip. No activity what > so ever on the console. Any idea what the heck I am missing? You should change this in the Tornado project tool as well. And you should check if WDB is not claiming this port. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How do i print unsigned long long ? Date: Fri, 11 Oct 2002 09:37:57 -0400 From: Keith Arner Organization: Marconi Message-ID: References: <421ece65.0210100431.41f75af9@posting.google.com> Reply-To: Keith Arner On 10 Oct 2002, Deepak Shetty wrote: > In VxWorks, i have a variable of type unsigned long long. > This basically means that the size of this var is 8 bytes ( 64 bits ) > > Any idea how do i print the value contained in this variable using printf ? > > There is no supporting type ( as in %type ) that i can use in printf to > print vale of a 64 bit variable ? The format that you're supposed to use is "%lld" for a 'signed long long' and "%llu" for an 'unsigned long long'. However, if I remember correctly, the libraries that WindRiver ships with the egcs compiler in 5.4 do not recognize this. Consequently, you have to write your own function to do it. Also, the topic of bit shifts on a long long came up in this thread. The documentation that ships with 5.4 says that bit shitfs on a long long may or may not be supported on any particular platform. This means that a bit shift on a 64 bit long long may well discard the most significant 32 bits. Use at your own risk. Keith - -- "Once something has gone horribly wrong..." "...it just gets horribler and horribler." "Then, it's a matter of finding the original horriblism." -- from a conversation on debugging problems in a linked list --------------------------- Newsgroups: comp.os.vxworks Subject: vxWorks_rom, NVRAM , and Bootline access. Date: 11 Oct 2002 07:15:07 -0700 From: ipnet@boursorama.com (Ipnet) Organization: http://groups.google.com/ Message-ID: Hi everybody, I have read all posts about vxWorks_rom, NVRAM and Bootline, but haven't found anyone that matches the problem I have. Here are problem descriptions : The processor is MPC8260. 1) I flash a VxWorks_rom kernel at the begining of the Flash Board. 2) The NVRAM is simulated at the end of the Flash board and is empty (no Boot Line yet). 3) I build the VxWorks_rom under Tornado 2, and the prjConfig.c file is generated. Here is a copy of my usrInit and usrRoot extracted from prjConfig.c : void usrInit (int startType) { sysStart (startType); /* clear BSS and set up the vector table base address. */ cacheLibInit (USER_I_CACHE_MODE, USER_D_CACHE_MODE); /* include cache support */ excVecInit (); /* exception handling */ sysHwInit (); /* call the BSPs sysHwInit routine during system startup */ usrCacheEnable (); /* optionally enable caches */ usrKernelInit (); /* context switch and interrupt handling. DO NOT REMOVE. */ } void usrRoot (char *pMemPoolStart, unsigned memPoolSize) { usrKernelCoreInit (); /* core kernel facilities */ memPartLibInit (pMemPoolStart, memPoolSize); /* core memory partition manager */ memInit (pMemPoolStart, memPoolSize); /* full featured memory allocator */ usrMmuInit (); /* basic MMU component */ sysClkInit (); /* System clock component */ usrIosCoreInit (); /* core I/O system */ usrKernelExtraInit (); /* extended kernel facilities */ usrIosExtraInit (); /* extended I/O system */ usrNetworkInit (); /* Initialize the network subsystem */ selectInit (); /* select */ usrToolsInit (); /* software development tools */ cplusLibInit (); /* C++ runtime support */ cplusDemanglerInit (); /* support library for target shell and loader */ usrAppInit (); /* call usrAppInit() (in your usrAppInit.c project file) after startup. */ } 4) The sysClkInit() will call sysClkConnect ((FUNCPTR) usrClock, 0), sysClkRateSet (60) and sysClkEnable (). 5) My sysClkConnect function calls to store the DEFAULT_BOOT_LINE in NVRAM : /* store default boot line if needed */ if ((sysNvRamGet(bootLine, BOOT_LINE_SIZE, 0) == ERROR) || (bootLine[0] == EOS) || (bootLine[0] == 0xFF)) { sysNvRamSet(DEFAULT_BOOT_LINE, strlen(DEFAULT_BOOT_LINE)+1, 0); } 6) When board jumps to the entry point of the vxWorks_rom and after executing all code the NVRAM is never updated and the VxWorks cannot find the boot device , and automatically attaches the lo0 (loopback). DIFFERENCES between a BootRom and a vxWorks_Rom : BootRom : _ In the usrInit() function, the BootRom only enables the Instructions Caches. _ In the usrRoot() funtion, it does a memInit() and then a sysClockConnect. vxWorks_rom : _ In the usrInit() function, the VxWorks_rom enables both Instruction and Data Caches. _ In the usrRoot() funtion, it does a memInit(), an Instruction and Data MMU init and then a sysClockConnect. Some additionnal testings : _ If I deactivate the Data MMU init in the vxWorks_rom, the kernel is executed, the NVRAM is updated and the boot device is attached. _ If I keep Intruction and Data MMU init the NVRAM is never updated and we found what is explained on above 6). a) Do you find anything similar? b) Is it a known bug ? b) How did you solved this problem (workaround or fix)? Thanks in advance. Ipnet. --------------------------- Newsgroups: comp.os.vxworks,comp.object.corba Subject: CORBA --> VisibrokerRT vs. ORBexpressRT Date: Fri, 11 Oct 2002 10:24:12 -0400 From: David Roberge Organization: Universite du Quebec Message-ID: <3DA6DF0C.4050809@ele.etsmtl.ca> Hi all, We are currently evaluating these two CORBA products: - - VisibrokerRT for VxWorks from Borland (formely Highlander) - - ORBexpressRT for VxWorks from OIS Does anyone had good or bad experiences with any of these product/company ? Thanks in advance. Your comments will be of great help to us. David Roberge --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How do i print unsigned long long ? Date: Fri, 11 Oct 2002 16:57:06 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <421ece65.0210100431.41f75af9@posting.google.com> <3DA5A242.20F05B79@lmco.com> <421ece65.0210102042.426497eb@posting.google.com> <3DA6AA7C.A3E5CF3@bellsouth.net> Reply-To: "Leonid Rosenboim" "joe durusau" wrote in message news:3DA6AA7C.A3E5CF3@bellsouth.net... [snip] > If I were you, I wouldn't put too much faith in the standard, esp. > if you are using the gnu compiler that wrs ships. It is very old, > and may not support long long at all. You can find out what it's > doing by looking at the assembler output. > > Speaking only for myself, > > Joe Durusau Joe, I have used the WRS GNU compiler with long long quite extensively during the development of DosFs2, and the core filesystem in DosFs2 still uses 64-bit integer math fairly succesfully. There is a know issue with long long on some PPC processors that use FP regs (there is a gcc patch for this), and there is this well know issue that WRS stdio lacks long long support so people have to write their own routines to convert log long to strings for printing. By the way, there is a 64-bit int to string function inside dosFs2, but it is not exposed. - -- hth - ----------------------------------------------------------------------- Leonid Rosenboim Visit: http://gamla.org.il/english/index.htm Consultant Email: my first name at consultant dot com --------------------------- Newsgroups: comp.os.vxworks Subject: Redirect i/o to the shell fails Date: Fri, 11 Oct 2002 15:44:02 GMT From: Eric McDaniel Organization: Vertical Networks Message-ID: This used to work for me, but now it doesn't and I don't know why... When I start up the target server (Tornado 2.0.2), I get the following message: "Warning: Cannot redirect the target shell to the target server Console. This VIO might be already open and the redirection done. If the redirection does not work, restart the target and the target server." When I execute 'vf0=open("/vio/0",2,0)' from the shell, it returns - -1 (failure). Below is my windsh.tcl script. I know it is running every time I open a shell, as I see the two printf messages. However, all of my i/o is still going to the serial port. shParse { printf("Executing windsh.tcl...\n"); } shParse { printf("Redirecting i/o to the shell...\n"); } shParse { vf0=open("/vio/0",2,0); } shParse { ioGlobalStdSet(0,vf0); } shParse { ioGlobalStdSet(1,vf0); } shParse { ioGlobalStdSet(2,vf0); } shParse { logFdAdd(vf0); } --------------------------- Newsgroups: comp.os.vxworks Subject: Re: can't load .out files with dosfs2 Date: 11 Oct 2002 09:27:36 -0700 From: term@xerosoft.com (Tom) Organization: http://groups.google.com/ Message-ID: <15cdb0fb.0210110827.127da64a@posting.google.com> References: <15cdb0fb.0210021822.572bd28c@posting.google.com> <15cdb0fb.0210090822.52841a20@posting.google.com> <15cdb0fb.0210101559.2fa61919@posting.google.com> Thanks Michael, this is exactly what I needed. I am now able to load off of TFFS if I read the file first and then use a memDrv to provide the fd that I pass to ld. "Michael Lawnick" wrote in message news:... > Hi Tom, > > (Leonid: sorry, don't want to break you) > > Tom: just to get you forward: whats about memDrv instead of pipe? ramDrv > bases on dosFs AFAIK, but memDrv should give you a plain memory driver that > should do the same trick as pipe. > > Just my 2 cents. > Michael > > "Tom" schrieb im Newsbeitrag > news:15cdb0fb.0210101559.2fa61919@posting.google.com... > > Leonid, > > > > I tried the test you proposed and loaded the .out file into a pipe, > > set stdin to be the pipes FD, and then did an ld with name == NULL so > > that it would try to load from stdin. This was all done in one task, > > I just made sure the pipe was very large; this isn't a problem is it? > > Unfortunately the call to ld failed. > > > > I hadn't really thought about the lseek implementation being an issue, > > but I think you are on to something. I ran this test again but set a > > breakpoint on lseek (right before ld is called). What I saw was that > > the first time lseek gets called it is trying to do a seek on STDIN > > (fd = 0) to offset 300 from the begining of the file. This seems > > valid but I don't believe pipes support seeking, so the call returns a > > -1. > > > > Maybe the version of dosFs that I have has a bug in its lseek > > implementation? What do you think? I guess I could write my own > > pipes style driver that supports lseek and use that. Can you think of > > any other things to try? > > > > Thanks for the help, > > > > -Tom > > > > "Leonid Rosenboim" wrote in message > news:... > > > This all seems so strange that it is probably something > > > very simple we all have ignored. > > > > > > If you have the exact same file effectively loading > > > from a network, there should be no reason at all > > > this file won't be loading from dosFs, well except > > > that different file systems handle the lseek() call > > > differently, and most architectures dynamic > > > loader will use lseek(). > > > > > > How about if you create a pipe, copy the file > > > into the pipe, while another task runs loadModule > > > with the pipe being its stdin ? (I suppose you iknow > > > how to do this). > > > > > > Keep me posted. > > > ----------------------------------------------------------------------- > > > Leonid Rosenboim Visit: http://www.masada2000.org/ > > > Consultant Email: my first name at consultant dot > com > > > > > > > > > "Tom" wrote in message > > > news:15cdb0fb.0210090822.52841a20@posting.google.com... > > > > Hey Leonid, > > > > > > > > I have tried a number of different command lines to load the file, but > > > > I only showed one in my first e-mail. > > > > > > > > I also tried: > > > > > > > > ld 1,1,"test.out" > > > > ld 1,1,"/tffs0/test.out" > > > > ld < test.out > > > > ld < /tffs0/test.out > > > > and loadModule() with a file descriptor that I manually opened. > > > > > > > > unfortunatelly they all failed. > > > > > > > > Any other ideas? > > > > > > > > > > > > Best regards, > > > > > > > > Tom > > > > > > > > "Leonid Rosenboim" wrote in message > news:... > > > > > Could you try to replace this command: > > > > > > > > > > > -> ld < test.out > > > > > > ld error: error reading file (errno = 0x3d0001). > > > > > > value = 0 = 0x0 > > > > > > > > > > with this command ? > > > > > > > > > > -> ld "/tffs0/test.out" > > > > > > > > > > or this command ? > > > > > > > > > > -> ld < /tffs0/test.out > > > > > > > > > > -- HTH > > > > > ----------------------------------------------------------------------- > > > > > Leonid Rosenboim Visit: > http://www.masada2000.org/ > > > > > Consultant Email: my first name at consultant > dot > com > > > > > > > > > > > > > > > "Tom" wrote in message > > > > > news:15cdb0fb.0210021822.572bd28c@posting.google.com... > > > > > > This is a repost of a previous message I had posted with new > > > > > > information included. > > > > > > > > > > > > I have been working on getting a strong arm based motherboard to > run > > > > > > vxworks. One problem I have run into is that I can't load any > .out > > > > > > files off of my TFFS (True Flash File System). When I try to it > just > > > > > > gives me an error saying "error reading file (errno = 0x3d0001).". > I > > > > > > looked up the errno and it resolved to S_objLib_OBJ_ID_ERROR , but > > > > > > this is generated from something at a low level that I don't have > the > > > > > > source code to. I recently tried stepping through the assembler > code > > > > > > to see where this error was coming from and it turns out errno is > not > > > > > > set until the cleanup code runs after fioRead fails, so this is > errno > > > > > > is irrelevant. > > > > > > > > > > > > Every other operation with the TFFS seems to work just fine. I > can > > > > > > create, delete, read, write, and rename files. As an example I > tried > > > > > > copying an out file off my FTP server onto flash and loading it, > but > > > > > > it failed. Then after that I copied the file off of the TFFS to a > new > > > > > > file name on my FTP server and then loaded it from that file (over > the > > > > > > network) and it worked fine. > > > > > > > > > > > > -- start dump -- > > > > > > > > > > > > -> cd "/tffs0/" > > > > > > value = 0 = 0x0 > > > > > > -> ll > > > > > > -rwxrwxrwx 1 0 0 10064 Jan 1 1980 test.out > > > > > > value = 0 = 0x0 > > > > > > -> ld < test.out > > > > > > ld error: error reading file (errno = 0x3d0001). > > > > > > value = 0 = 0x0 > > > > > > -> copy "/tffs0/test.out", "host:newtest.out" > > > > > > Copy OK: 10064 bytes copied > > > > > > value = 0 = 0x0 > > > > > > -> ld < host:newtest.out > > > > > > value = 12531032 = 0xbf3558 > > > > > > -> > > > > > > > > > > > > -- end dump -- > > > > > > > > > > > > > > > > > > After these tests I thought the problem might have been size > related, > > > > > > so I tried to create a minimal "hello world" out file, but that > didn't > > > > > > work either. > > > > > > > > > > > > As a work around I thought I would try setting up a runscript that > > > > > > creates a ram drive and then copies each file to the ramdrive, > loads > > > > > > from there, deletes the file, and then continues on to the next > one > > > > > > until it is done. Unfortunatelly this failed with the same error > > > > > > message that I got when I loaded off of a TFFS. The only thing I > can > > > > > > think of that the ramdrive and TFFS have in common that a ftp file > > > > > > system doesn't is that they use dosFS (version 2). > > > > > > > > > > > > Has anyone else had any problems like this? I'm running > > > > > > VxWorks 5.4.2 with Tornado 2.1.1. > > > > > > > > > > > > > > > > > > Best Regards, > > > > > > > > > > > > Tom --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How do i print unsigned long long ? Date: 11 Oct 2002 11:10:41 -0700 From: dcooke@orthstar.com (Dan Cooke) Organization: http://groups.google.com/ Message-ID: <4e264d23.0210111010.5425dce1@posting.google.com> References: <421ece65.0210100431.41f75af9@posting.google.com> > In VxWorks, i have a variable of type unsigned long long. > This basically means that the size of this var is 8 bytes ( 64 bits ) > > Any idea how do i print the value contained in this variable using printf ? > Quick and dirty hex output can be done as follows... logMsg("64 Bit Value = 0x%08X%08X \n", (uint32)(LongLongValue>>32), (uint32)(LongLongValue&0x0ffffffff)); Dan --------------------------- Newsgroups: comp.os.vxworks Subject: Re: can't load .out files with dosfs2 Date: 11 Oct 2002 09:47:57 -0700 From: term@xerosoft.com (Tom) Organization: http://groups.google.com/ Message-ID: <15cdb0fb.0210110847.79d170a7@posting.google.com> References: <15cdb0fb.0210021822.572bd28c@posting.google.com> <15cdb0fb.0210090822.52841a20@posting.google.com> <15cdb0fb.0210101559.2fa61919@posting.google.com> Hey Leonid, See comments below... - -Tom "Leonid Rosenboim" wrote in message news:... > Good work, Tom. > > I do recall something about a certain patch version of dosFs2 that > has a problem in lseek(), and the reason I suggested you try > pipes was to doube check that this ld indeed uses lseek. It does. > And it clearly shows what i exactly different beween simply copying > the file and loading it as a module. > > What you need to do now is two things: > 1. With the help of WRS support (sorry you have to do this), starting with > on-line, see if there is an SPR about dosFs2 and lseek, and if you find it, > get the patch. > My companies support contract has expired and we are tight on cash, so renewel probably isn't going to happen for a while. I sent them an e-mail just incase they felt like helping anyway, but I got no reply (big surprise ;) ). > 2. If #1 does not work, write s simple test program that first creates a > sizable file, and initializes it with very simple, expectable data, e.g. > every 32-bit word in the file would simple contain its offset relative > stanrt of file. Then the program begins a series of lseek() calls using > random offsets modulus the file size, reads a word from that offset and > verifies it is what was expected. I am pretty sure this will demonstrate the > bug if there is indeed an lseek bug. You might get away with a simple script > that repeats the calls made by ld, e.g. read 48 bytes from beginning, lseek > 300, read 128 bytes, ... > I had a test like that in the fstest suite we developed in parallel with > dosFs2 (actually this suite begun with cdromFs, later was extended for > FlashLib and then extended more for patching dosFs1 and dosFs2 development, > but I figure nobody at WRS relelbers this suite ever existed, bammer!) > I would like to make sure the problem is in fact with lseek, so this should be a good test. I'll try it sometime today and post the results. > 3. If you are bored in the mean time, you could also creatre a ramDisk and > try to copy the file over and load it from ramDisk, but most chanses are it > will fail too becase it would use the same dosFs2. > I tried this before and it failed (that was the reason I started to suspect the problem was related to dosFs rather than TFFS). I did however try going through memDrv and that worked fine, so now I have a workaround that I can use for loading out of flash (yay!). Of course it would be better if I could use ld directly, but that may not be an option for the time being. > Just let me know when you need my address to mail me a check ;-) I'll see what I can do, but most likely my company isn't going to want to cut a check. I do apreciate your help though. > ----------------------------------------------------------------------- > Leonid Rosenboim Visit: http://www.masada2000.org/ > Consultant Email: my first name at consultant dot com > > > "Tom" wrote in message > news:15cdb0fb.0210101559.2fa61919@posting.google.com... > > Leonid, > > > > I tried the test you proposed and loaded the .out file into a pipe, > > set stdin to be the pipes FD, and then did an ld with name == NULL so > > that it would try to load from stdin. This was all done in one task, > > I just made sure the pipe was very large; this isn't a problem is it? > > Unfortunately the call to ld failed. > > > > I hadn't really thought about the lseek implementation being an issue, > > but I think you are on to something. I ran this test again but set a > > breakpoint on lseek (right before ld is called). What I saw was that > > the first time lseek gets called it is trying to do a seek on STDIN > > (fd = 0) to offset 300 from the begining of the file. This seems > > valid but I don't believe pipes support seeking, so the call returns a > > -1. > > > > Maybe the version of dosFs that I have has a bug in its lseek > > implementation? What do you think? I guess I could write my own > > pipes style driver that supports lseek and use that. Can you think of > > any other things to try? > > > > Thanks for the help, > > > > -Tom > > > > "Leonid Rosenboim" wrote in message > news:... > > > This all seems so strange that it is probably something > > > very simple we all have ignored. > > > > > > If you have the exact same file effectively loading > > > from a network, there should be no reason at all > > > this file won't be loading from dosFs, well except > > > that different file systems handle the lseek() call > > > differently, and most architectures dynamic > > > loader will use lseek(). > > > > > > How about if you create a pipe, copy the file > > > into the pipe, while another task runs loadModule > > > with the pipe being its stdin ? (I suppose you iknow > > > how to do this). > > > > > > Keep me posted. > > > ----------------------------------------------------------------------- > > > Leonid Rosenboim Visit: http://www.masada2000.org/ > > > Consultant Email: my first name at consultant dot > com > > > > > > > > > "Tom" wrote in message > > > news:15cdb0fb.0210090822.52841a20@posting.google.com... > > > > Hey Leonid, > > > > > > > > I have tried a number of different command lines to load the file, but > > > > I only showed one in my first e-mail. > > > > > > > > I also tried: > > > > > > > > ld 1,1,"test.out" > > > > ld 1,1,"/tffs0/test.out" > > > > ld < test.out > > > > ld < /tffs0/test.out > > > > and loadModule() with a file descriptor that I manually opened. > > > > > > > > unfortunatelly they all failed. > > > > > > > > Any other ideas? > > > > > > > > > > > > Best regards, > > > > > > > > Tom > > > > > > > > "Leonid Rosenboim" wrote in message > news:... > > > > > Could you try to replace this command: > > > > > > > > > > > -> ld < test.out > > > > > > ld error: error reading file (errno = 0x3d0001). > > > > > > value = 0 = 0x0 > > > > > > > > > > with this command ? > > > > > > > > > > -> ld "/tffs0/test.out" > > > > > > > > > > or this command ? > > > > > > > > > > -> ld < /tffs0/test.out > > > > > > > > > > -- HTH > > > > > ----------------------------------------------------------------------- > > > > > Leonid Rosenboim Visit: > http://www.masada2000.org/ > > > > > Consultant Email: my first name at consultant > dot > com > > > > > > > > > > > > > > > "Tom" wrote in message > > > > > news:15cdb0fb.0210021822.572bd28c@posting.google.com... > > > > > > This is a repost of a previous message I had posted with new > > > > > > information included. > > > > > > > > > > > > I have been working on getting a strong arm based motherboard to > run > > > > > > vxworks. One problem I have run into is that I can't load any > .out > > > > > > files off of my TFFS (True Flash File System). When I try to it > just > > > > > > gives me an error saying "error reading file (errno = 0x3d0001).". > I > > > > > > looked up the errno and it resolved to S_objLib_OBJ_ID_ERROR , but > > > > > > this is generated from something at a low level that I don't have > the > > > > > > source code to. I recently tried stepping through the assembler > code > > > > > > to see where this error was coming from and it turns out errno is > not > > > > > > set until the cleanup code runs after fioRead fails, so this is > errno > > > > > > is irrelevant. > > > > > > > > > > > > Every other operation with the TFFS seems to work just fine. I > can > > > > > > create, delete, read, write, and rename files. As an example I > tried > > > > > > copying an out file off my FTP server onto flash and loading it, > but > > > > > > it failed. Then after that I copied the file off of the TFFS to a > new > > > > > > file name on my FTP server and then loaded it from that file (over > the > > > > > > network) and it worked fine. > > > > > > > > > > > > -- start dump -- > > > > > > > > > > > > -> cd "/tffs0/" > > > > > > value = 0 = 0x0 > > > > > > -> ll > > > > > > -rwxrwxrwx 1 0 0 10064 Jan 1 1980 test.out > > > > > > value = 0 = 0x0 > > > > > > -> ld < test.out > > > > > > ld error: error reading file (errno = 0x3d0001). > > > > > > value = 0 = 0x0 > > > > > > -> copy "/tffs0/test.out", "host:newtest.out" > > > > > > Copy OK: 10064 bytes copied > > > > > > value = 0 = 0x0 > > > > > > -> ld < host:newtest.out > > > > > > value = 12531032 = 0xbf3558 > > > > > > -> > > > > > > > > > > > > -- end dump -- > > > > > > > > > > > > > > > > > > After these tests I thought the problem might have been size > related, > > > > > > so I tried to create a minimal "hello world" out file, but that > didn't > > > > > > work either. > > > > > > > > > > > > As a work around I thought I would try setting up a runscript that > > > > > > creates a ram drive and then copies each file to the ramdrive, > loads > > > > > > from there, deletes the file, and then continues on to the next > one > > > > > > until it is done. Unfortunatelly this failed with the same error > > > > > > message that I got when I loaded off of a TFFS. The only thing I > can > > > > > > think of that the ramdrive and TFFS have in common that a ftp file > > > > > > system doesn't is that they use dosFS (version 2). > > > > > > > > > > > > Has anyone else had any problems like this? I'm running > > > > > > VxWorks 5.4.2 with Tornado 2.1.1. > > > > > > > > > > > > > > > > > > Best Regards, > > > > > > > > > > > > Tom --------------------------- Newsgroups: comp.os.vxworks Subject: priority changing in the shell Date: 11 Oct 2002 11:34:23 -0700 From: smolensj@yahoo.com (smonice) Organization: http://groups.google.com/ Message-ID: i have a forever repeating task (called using repeat) and need to change the priority now in the shell, how can i do this without calling the taskSpawn and making a new task, is there a way to do this. Please help asap. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: priority changing in the shell Date: Fri, 11 Oct 2002 12:03:08 -0700 From: "Michael R. Kesti" Organization: MK Associates Message-ID: <3DA7206C.69F1E973@gv.net> References: Reply-To: mkesti@gv.net smonice wrote: >i have a forever repeating task (called using repeat) and need to >change the priority now in the shell, how can i do this without >calling the taskSpawn and making a new task, is there a way to do >this. Please help asap. See taskPrioritySet() and other functions in the reference manual's taskLib section. - -- ======================================================================== Michael Kesti | "And like, one and one don't make | two, one and one make one." mkesti@gv.net | - The Who, Bargain --------------------------- Newsgroups: comp.os.vxworks Subject: Re: priority changing in the shell Date: 11 Oct 2002 16:12:55 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0210111512.6ecc56d@posting.google.com> References: Hello, You could start by reading the VxWorks Programmer's Guide that came with your purchase, but to answer your question, the routine you are looking for is taskPrioritySet(). You really do need to read the manuals though and find out more about the design of the OS and what it can do. Once you have read the programmer's guide, you might also want to take a look at the index in the reference manuals... this is a special kind of index called a permuted index. You can look in here for 'priority' and it will help you find all the routines that have the word 'priority' in their synopsis. HTH, John... smolensj@yahoo.com (smonice) wrote in message news:... > i have a forever repeating task (called using repeat) and need to > change the priority now in the shell, how can i do this without > calling the taskSpawn and making a new task, is there a way to do > this. Please help asap. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: What is needed for symFindByName? Date: 11 Oct 2002 16:15:41 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0210111515.5af86692@posting.google.com> References: Hello, You will need the symbol table and something to initialise it. The latter could be handled using INCLUDE_NET_SYM_TBL (in which case you will need the target-based loader too), or using INCLUDE_STANDALONE_SYM_TBL in which case you don't need anything else. For VxWorks 5.x builds with the standalone table you need to use a different build rule. For VxWorks AE, simply adding the component to your kernel domain will cause the build to do the right things. HTH, John... glenn_is_cool@hotmail.com (Andreas F) wrote in message news:... > Hi.. > > I'm having a problem with 'symFindByName'. It just doesn't work! I > know that you need to synchronize host and target symbol tables, so > that is taken care of. My question is really what components I need to > include to my VxWorks image. And is there any other pitfalls with this > function? --------------------------- Newsgroups: comp.os.vxworks Subject: Starter pack ??? Date: Fri, 11 Oct 2002 20:07:43 -0700 From: "A. Kaul" Organization: National Semiconductor, Santa Clara Message-ID: <3DA791FF.8080509@antispam.com> Are there any good refrences for starting driver development on VxWorks? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: taskDelay Date: Sat, 12 Oct 2002 00:08:31 -0700 From: Ian Montgomerie Message-ID: References: <9QsF$2Kv1bn9EATQ@wtoram.co.uk> On Fri, 4 Oct 2002 17:10:23 +0100, Wade Oram wrote: >In message , Mic > writes >>Hy guys, just a question: how can I delay a task for 10^-6 seconds? I >>know usrRoot usually sets tick rate to 60, with a max value equal to >>1000 (right?), but this means that to pass an integer to taskDelay (as >>its prototype says), the shortest delay I can have is 1 ms! What >>about? > >On my hardware its impossible without using a simple loop to consume the >required amount of processor time. > >Using (Aux/custom?) timers to interrupt after the specified amount of >time and give a binary semaphore to the task waiting would appear to be >a solution except that the overhead of the interrupt/semTake() is often >more than 1us and sometimes much more. (You don't specify what hardware >platform your running on so I don't know if this is the case on your >platform but unless your right up to date, I suspect that it may well >be). > >A >"for( volatile int tDelay = 0; (tDelay < (numMicrosecs * DelayScale) - >DelayOffset); ++tDelay);" > >loop is often the best that you can do. > >Of course for cross platform portability, DelayOffset and DelayScale >would have to be determined by a calibration routine. Here are some hints for writing a delay loop: 1. Write a "delay" function with useful units, such as microseconds. Have a delay loop inside it that runs N iterations per microsecond. Measure to find the value of N. This lets you change just one function to port, etc. 2. Reading from a "volatile" integer is basic to make the loop predictable, as it won't read from a register. But if possible, try reading from non-cached memory, reading from flash memory, etc. This makes the loop fairly insensitive to CPU speed and what is going on in the system, although it does depend on memory speed. I had to do short delays for a system where the flash memory usually has the same performance, but the CPU will be clocked at different levels (pain in the ass to measure this at runtime). 3. Your "delay" function will probably be very accurate to your measurements for decently long delays. But if your delay is so short that the internal loop is only executing 10 times or so, the delay will be longer due to setup time, function call overhead, etc. Be cautious of this if you want to, say, wait exactly 0.5 microseconds, but not 0.8 microseconds. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Sat, 12 Oct 2002 00:29:42 -0700 From: Ian Montgomerie Message-ID: References: <3a5ff53f.0210020748.3643d45a@posting.google.com> On Mon, 7 Oct 2002 11:34:19 +0200, "Leonid Rosenboim" wrote: >Brian, could not resist, must follow on this once more >(last time, promise): > >You will probably spend in the era of $300k to $600k >on the switch of VxWorks for a semi-custom OS - >using something unknown will be plenty of surprises. >And that will save you 25 cents off the COGS. > >Alternatively, you can hire a consultant for $20k-$50k >to review your design, and recommend alternative >ways to reduce COGS, and realistically you can >expect over $2.5 per unit savings, ten times what >you would save on changing the OS. Did he say what the nature of his product was? On the product I work on, which is a SoC + software, there are exactly two ways to reduce our own COGS. First, spin a new cost-reduced chip. Second, reduce software royalties. (We can also make the product more attractive by spinning chips with new HW features, or writing new software, so that we can inter-operate with cheaper external hardware, thus reducing our customers' COGS). And unfortunately, there are no miraculous consultants who know more about our jobs than we do, to jump in and save costs. Other people may be in similar boats. For example, one of our products has a dual CPU design. But to save the royalties of multiprocessor VxWorks, we run a custom scheduler on one of the CPUs, and we have custom mechanisms to communicate between the two. VxWorks runs on one CPU, and that is where we put externally developed code, control code that benefits highly from being able to use the Tornado debugger, etc. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Sat, 12 Oct 2002 00:33:02 -0700 From: Ian Montgomerie Message-ID: References: <3a5ff53f.0210020748.3643d45a@posting.google.com> On Wed, 9 Oct 2002 19:53:15 +0200, "Leonid Rosenboim" wrote: >"brian" wrote in message >news:d2f67bd8.0210090933.19572732@posting.google.com... >> "Leonid Rosenboim" wrote in message >news:... >> > Brian, could not resist, must follow on this once more >> > (last time, promise): >> >> Hey. Thanx for the responses :) >> >> > >> > You will probably spend in the era of $300k to $600k >> > on the switch of VxWorks for a semi-custom OS - >> > using something unknown will be plenty of surprises. >> > And that will save you 25 cents off the COGS. >> >> That's the questions I have. 300-600k is 2-4 man years. Can it really >> take that long? How? As in, say , porting one of the Linux baby >> kernels (no stack..just kernel etc) >> >> Again thanx for the input. > >It really depends on how experienced the people on this project are. >Also, you need to appreciate the tools that you would need to replace, >although it's OK to use the GNU toolchain you got with VxWorks, >you will need to come up with a low-level debugger such as an emulator. >Then, there are drivers, if the new kernel has different driver APIs, and >you can't just port the current ones as they probably include a WindRiver >copyright. >Last but not least is the application, it is critical to make an "inventory" >of OS calls used by the application, and decide on a case-by-case basis >how these will be either (1) mapped to new OS existing calls; (2) emulated >on top of the new OS or added to the new OS or (3) replaced at the >application >with a call that exists in the new OS. > >The risk factor here is quite large, much more substantial than any >hardware modification that would save the same amount on the COGS, >or even 10 times more ! You seem quite confident in the detailed costs of someone else's project. If my project paid royalties for our RTOS, which we don't, switching could be a very good idea. It would in fact be less risky than trying to save another 25 cents on our design (over 5% of overall unit cost), since our only non-software means of doing so is the risky, multimillion dollar option of cost-reducing the silicon. Then for some other projects, saving 25 cents per unit just means cutting out some unecessary screws. --------------------------- Newsgroups: comp.os.vxworks Subject: How to configure two network cards in Vxworks? Date: 12 Oct 2002 01:07:25 -0700 From: li.mingzheng@zte.com.cn (M.Z.LI) Organization: http://groups.google.com/ Message-ID: <54a46c51.0210120007.22fb189b@posting.google.com> Hi, all, I want to use two network cards on one target, I have configured two cards with END style driver support, they worked very well, however, it seems vxworks does not support multiple network cards with BSD style driver support. Can anyone give me a little hints regarding this issue? thanks a lot! --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Sat, 12 Oct 2002 11:16:45 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <3a5ff53f.0210020748.3643d45a@posting.google.com> Reply-To: "Leonid Rosenboim" "Ian Montgomerie" wrote in message news:hsjfqusv57hqols3ft8fr1anpf4psgho2j@4ax.com... [snip] > You seem quite confident in the detailed costs of someone else's > project. If my project paid royalties for our RTOS, which we don't, > switching could be a very good idea. It would in fact be less risky > than trying to save another 25 cents on our design (over 5% of overall > unit cost), since our only non-software means of doing so is the > risky, multimillion dollar option of cost-reducing the silicon. Then > for some other projects, saving 25 cents per unit just means cutting > out some unecessary screws. > My confidence is based on experience. Your project does seem to be something I have not seen before - a total COGS bill of $5 per unit (calculated from 25 cents and 5% ratio). Are you sure of that ?! Maybe your best choise to save cost is is in fact changing an RTOS, but if you also consider production of an ASIC, I think that translating any working FPGA based design into ASIC is less risky than substantial software rewrite, and the ASIC can be outsourced. Maybe you're right to go for RTOS swap in this quite unusual project. What puzzles me is that people ask for advise, and then argue with the advise they get. If you have a strong opinion, why bother ask anyone ? Strange. Leonid --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Sat, 12 Oct 2002 11:20:43 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <3a5ff53f.0210020748.3643d45a@posting.google.com> Reply-To: "Leonid Rosenboim" "Ian Montgomerie" wrote in message news:ecjfqu4dqdde3qcnpl29f3v53brn882leo@4ax.com... [snip] > And unfortunately, there are no miraculous consultants who know more > about our jobs than we do, to jump in and save costs. Other people > may be in similar boats. For example, one of our products has a dual > CPU design. But to save the royalties of multiprocessor VxWorks, we > run a custom scheduler on one of the CPUs, and we have custom > mechanisms to communicate between the two. VxWorks runs on one CPU, > and that is where we put externally developed code, control code that > benefits highly from being able to use the Tornado debugger, etc. > Just a comment: I know about several projects that use multiple processors with VxWorks that pay per number of units (i.e. systems) shipped and not by the number of processors. Definitely, you could have negotiated similar terms instead of using a custom scheduler on the second CPU. - - Leonid --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How to configure two network cards in Vxworks? Date: Sat, 12 Oct 2002 11:22:57 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <54a46c51.0210120007.22fb189b@posting.google.com> Reply-To: "Leonid Rosenboim" There is no problem with using multiple BSD network drivers or multiple devices of the same type in VxWorks, except that a particular network driver can have a limitation (reentrancy problem) on the number of its instaciations. - -- - ----------------------------------------------------------------------- Leonid Rosenboim Visit: http://gamla.org.il/english/index.htm Consultant Email: my first name at consultant dot com "M.Z.LI" wrote in message news:54a46c51.0210120007.22fb189b@posting.google.com... > Hi, all, > > I want to use two network cards on one target, I have configured two > cards with END style driver support, they worked very well, however, > it seems vxworks does not support multiple network cards with BSD > style driver support. > > Can anyone give me a little hints regarding this issue? thanks a lot! --------------------------- Newsgroups: comp.os.vxworks Subject: taskRestart Date: 12 Oct 2002 04:42:36 -0700 From: vittala@hotmail.com (Vittal) Organization: http://groups.google.com/ Message-ID: <6847c39a.0210120342.91911fc@posting.google.com> Hi All, I am facing a problem with taskRestart. I am having two tasks task1 and task2 with the same priority of 0. I am trying to restart task1 from task2. The task1 restarts properly. However, there is a problem. I am trying to check the return value of taskRestart and print appropritate message. But the message never appears even though the task is restarted. I replaced the printf with logMsg and the results were same. I also replaced the printf/logMsg with a global variable and tried to check the value of the global variable from the shell. But the global variable's value was unchanged. Something like this.... restartCount = 0; /* global count */ code snippet from task2.... if (OK == taskRestart (taskNameToId ("task1")) restartCount = 1; else restartCount = 2; After the execution of this, task1 is restarted. I have confirmed this with some prints and global variables. However, the restartCount remains 0. I expected it to be 1 since the task1 was restarted. If there was some problem, at least it should have got changed to 2. But, the restarCount remains 0 and this is something which I am not able to understand. I am not able to connect the debugger due to some problems with the network and hence not able to debug it properly. Any pointers, ideas are welcome and will be greatly appreciated. with regards, Vittal. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Starter pack ??? Date: 12 Oct 2002 12:01:07 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0210121101.377ecb05@posting.google.com> References: <3DA791FF.8080509@antispam.com> Hello, Try the VxWorks Programmer's Guide for starters, then under target/src/drv/ you should find some source for example character drivers. What kind of driver are you trying to develop? HTH, John... "A. Kaul" wrote in message news:<3DA791FF.8080509@antispam.com>... > Are there any good refrences for starting driver development on VxWorks? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Sat, 12 Oct 2002 20:18:01 -0700 From: Ian Montgomerie Message-ID: References: <3a5ff53f.0210020748.3643d45a@posting.google.com> On Sat, 12 Oct 2002 11:16:45 +0200, "Leonid Rosenboim" wrote: > >"Ian Montgomerie" wrote in message >news:hsjfqusv57hqols3ft8fr1anpf4psgho2j@4ax.com... > >[snip] > >> You seem quite confident in the detailed costs of someone else's >> project. If my project paid royalties for our RTOS, which we don't, >> switching could be a very good idea. It would in fact be less risky >> than trying to save another 25 cents on our design (over 5% of overall >> unit cost), since our only non-software means of doing so is the >> risky, multimillion dollar option of cost-reducing the silicon. Then >> for some other projects, saving 25 cents per unit just means cutting >> out some unecessary screws. >> > >My confidence is based on experience. >Your project does seem to be something I have not seen before - >a total COGS bill of $5 per unit (calculated from 25 cents and 5% ratio). >Are you sure of that ?! That's only a rough figure, and I am not going to release the actual one, but yes, our SoC sells in the realm of 10 dollars in quantity. Well, the one I work on sells for about that, other SoCs also using VxWorks sell from less than 5 dollars, on up. Obviously our unit cost is less than 10 dollars, given that we like to have positive gross margins. >Maybe your best choise to save cost is is in fact changing an RTOS, Well as I said, IIRC we don't actually pay royalties for our RTOS. We definitely got fixed-price deals for some major silicon IP. And for some other software IP (I don't know how the heck Wind River makes money licensing DosFs2 for around a dollar a unit, when ERTFS works just as well for under 10 grand fixed license). I was saying that, if we could somehow save 25 cents by changing the RTOS, it would be an option looked at very hard indeed. >but if you also consider production of an ASIC, I think that translating any >working FPGA based design into ASIC is less risky than substantial software >rewrite, and the ASIC can be outsourced. There with the assumptions again. FPGA? ASIC? My dear fellow, the product I work on is an ASSP product developed by internal VLSI. >Maybe you're right to go for RTOS swap in this quite unusual project. I'm not sure how "unusual" it is to have ASSP or ASIC SoCs, running an RTOS and selling for 20 dollars a unit or less. Last I checked that was a rather giant and mainstream component of many markets such as consumer electronics and telecommunications. >What puzzles me is that people ask for advise, and then argue with the >advise they get. If you have a strong opinion, why bother ask anyone ? When did I ask anyone for advice? I was pointing out that the advice you were giving (blanket statements that changing RTOS doesn't make sense because it could cost several hundred grand and up, there are much easier ways to save 25 cents, etc), really only applies to products with unit costs such that 25 cents is not a big deal. A fairly large number of engineers work on low-cost products that nevertheless use an RTOS. Although I will grant that the people making RTOS decisions for these products aren't necessarily likely to be posting to this newsgroup asking for advice. The point being, it's difficult to give advice without any details of someone's project. But when you do try and give advice "blind", it needs to be with caveats that strategy X may not apply to products with property Y. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Sat, 12 Oct 2002 20:30:21 -0700 From: Ian Montgomerie Message-ID: <1gphqu4qdg4db8tmpqqbath5dl5nf119dv@4ax.com> References: <3a5ff53f.0210020748.3643d45a@posting.google.com> On Sat, 12 Oct 2002 11:20:43 +0200, "Leonid Rosenboim" wrote: > >"Ian Montgomerie" wrote in message >news:ecjfqu4dqdde3qcnpl29f3v53brn882leo@4ax.com... > >[snip] > >> And unfortunately, there are no miraculous consultants who know more >> about our jobs than we do, to jump in and save costs. Other people >> may be in similar boats. For example, one of our products has a dual >> CPU design. But to save the royalties of multiprocessor VxWorks, we >> run a custom scheduler on one of the CPUs, and we have custom >> mechanisms to communicate between the two. VxWorks runs on one CPU, >> and that is where we put externally developed code, control code that >> benefits highly from being able to use the Tornado debugger, etc. >> > >Just a comment: I know about several projects that use multiple processors >with VxWorks that pay per number of units (i.e. systems) shipped and not >by the number of processors. Definitely, you could have negotiated similar >terms instead of using a custom scheduler on the second CPU. WindRiver would give us normal VxWorks for price X, and multiprocessor VxWorks for price X + Y, where Y was real money. No matter how good our purchasing guys were, WindRiver was for obvious reasons not going to throw in multiprocessor support as a freebie. Y happened to be less than the expense of developing a custom scheduler and custom interprocessor communications tools. Of course, it helped that before deciding to adopt VxWorks, we'd already developed the scheduler as a concept, and we needed many of our own communication primitives anyway because VxWorks message queues are really quite inefficient, for our purposes, compared to intelligent use of semaphore-bounded buffers. If you have the right people working for you, this stuff needn't be all that hard; I went to a university where it was expected that students taking the real time systems undergrad course would, individually or with a partner, write an x86 microkernel RTOS from scatch in four weeks, given only an ethernet bootloader and some design hints, leaving it in good enough state to control a significant robotics project. Every semester about two dozen of the better students would attempt it, and pretty much all succeed. It is not uncommon for large projects to have someone with the talent and RTOS experience to attempt such feats. A basic RTOS kernel and support features do not require a lot of code, or a lot of original thought for that matter. It can go quite quickly if you know what you're doing. (But it can go far, far less quickly if you haven't fully understood the concept of an RTOS, and don't know how to debug and test a kernel. Or if you haven't a good idea what thread synchronization and communication mechanisms, etc., are efficient for your application). --------------------------- Newsgroups: comp.os.vxworks Subject: Re: taskRestart Date: Sat, 12 Oct 2002 20:34:24 -0700 From: Ian Montgomerie Message-ID: References: <6847c39a.0210120342.91911fc@posting.google.com> On 12 Oct 2002 04:42:36 -0700, vittala@hotmail.com (Vittal) wrote: >Hi All, >I am facing a problem with taskRestart. > >I am having two tasks task1 and task2 with the same priority of 0. I >am trying to restart task1 from task2. The task1 restarts properly. >However, there is a problem. I am trying to check the return value of >taskRestart and print appropritate message. But the message never >appears even though the task is restarted. I replaced the printf with >logMsg and the results were same. I also replaced the printf/logMsg >with a global variable and tried to check the value of the global >variable from the shell. But the global variable's value was >unchanged. > >Something like this.... > >restartCount = 0; /* global count */ > >code snippet from task2.... > >if (OK == taskRestart (taskNameToId ("task1")) > restartCount = 1; >else > restartCount = 2; > >After the execution of this, task1 is restarted. I have confirmed this >with some prints and global variables. However, the restartCount >remains 0. I expected it to be 1 since the task1 was restarted. If >there was some problem, at least it should have got changed to 2. But, >the restarCount remains 0 and this is something which I am not able to >understand. > >I am not able to connect the debugger due to some problems with the >network and hence not able to debug it properly. > >Any pointers, ideas are welcome and will be greatly appreciated. Well, what does task 2 do? You said it is at the same priority as task 1. Is it possible that it is simply never relinquishing control of the CPU? You also said you can't connect the debugger and have problems with the network... at priority 0, your tasks both have higher priority than the Tornado shell task, VxWorks net task, etc. So if one of them is taking all of the CPU, those system tasks won't run and you won't be able to connect across the network. (Sorry if these things are obvious to you and the problem is more complex...) --------------------------- Newsgroups: comp.os.vxworks Subject: Where can I find a sample of SCC driver for mpc8260 Date: Sun, 13 Oct 2002 06:46:03 GMT From: "Ming Tai" Message-ID: Does anyone know Where can I find a sample of SCC driver for mpc8260 Ken --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Sun, 13 Oct 2002 08:53:42 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <3a5ff53f.0210020748.3643d45a@posting.google.com> Reply-To: "Leonid Rosenboim" Wel,, I find it both interesting and encouraging that DosFs2 - the project I headed and wrote some of its code myself, is selling for $1 per pop in large quantities, and beating competition who sells a similar FS buyout. I am surprised that they charge it a separate fee, not bundled with VxWorks. The unusual part about your situation from my perspective is that you are apparently selling chips, and pay VxWorks royalties. Usually, the people who make systems will pay for VxWorks. I know a few exceptions when VxWorks (or actually IoWorks) is bundled with i960 I2O chips, and know there are a few more who switched to the bundled model, but still the majority of VxWorks customer are system builders. Naturally, I had to assume you belong to that majority. Still, I beleive you should be able to negotiate a buyout. - -- - ----------------------------------------------------------------------- Leonid Rosenboim Visit: http://gamla.org.il/english/index.htm Consultant Email: my first name at consultant dot com --------------------------- Newsgroups: comp.os.vxworks Subject: about chip select question Date: Sun, 13 Oct 2002 15:25:04 +0800 From: "eric" Organization: DCI HiNet Message-ID: sir, I am use arm 4510b chip. I am want to use ecs0 .. ecs1 to select or produce single. How to do this? Please give me some advice. Thank you! 2002.10.13 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Creating Singleton objects at run time and using it in a shared library Date: 13 Oct 2002 08:19:56 -0700 From: steven4443@yahoo.com (stephen hallacy) Organization: http://groups.google.com/ Message-ID: References: <3D9CC452.4EF3@despam.autobahn.mb.ca> Thanks for the reply. "H. E. Taylor" wrote in message news:<3D9CC452.4EF3@despam.autobahn.mb.ca>... > In article , > stephen hallacy wrote: > > > > I want to create an singleton object at bootup time or some time after > > bootup time, and I want to be able to access that object's methods in > > a shared library. Is this possible, > > > > Yes. > > >and if so, how is it done? > > > > The usual way. ;-) > > Really. What difficulty are you having? > > Setup a global function GetObjectX() which uses a > static pointer to your instance. The first time through > you instantiate, save & return the pointer. Later times > you just return the pointer. You should probably consider > tear down at this stage of design as well. You may think > your object will live forever, but as soon as you > implement it, some manager will change the spec on you. > > > -het > > > PS. > Possibly useful links. > http://hillside.net/patterns/ > http://www.autobahn.mb.ca/~het/clinks.html#CPlusPlus --------------------------- Newsgroups: comp.os.vxworks Subject: How can a Shared Library access Protection Domain data Date: 13 Oct 2002 08:41:04 -0700 From: steven4443@yahoo.com (stephen hallacy) Organization: http://groups.google.com/ Message-ID: Hello, Background Information: I have two application domains (say domain1 and domain2) that want to access API functions from a shared library. Also, I have another application domain (domain3) that has one task that gathers serial data and performs computation on the data. I have no control over the applications in domain1 and domain2. These applications are created by another company. Question: How can my shared library have access to computation result from domain3's task? Stephen --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Sun, 13 Oct 2002 12:02:08 -0700 From: Ian Montgomerie Message-ID: References: <3a5ff53f.0210020748.3643d45a@posting.google.com> On Sun, 13 Oct 2002 08:53:42 +0200, "Leonid Rosenboim" wrote: >Wel,, I find it both interesting and encouraging that DosFs2 - >the project I headed and wrote some of its code myself, >is selling for $1 per pop in large quantities, and beating competition >who sells a similar FS buyout. I didn't say anything about how many copies of DosFs2 were being sold, just that I wondered how WindRiver could make money by inflexibly charging that much for it. In fact, when it comes to consumer electronics products using flash cards such as SmartMedia, Compact Flash, and Memory Stick (FAT16 based devices) - they all use ERTFS, nobody in consumer electronics than I know of uses DosFs2, precisely because of the ridiculous licensing fees. If you plan to ship more than 10,000 units, DosFs2 is a very uneconomical choice, and I wonder why WindRiver shot down its potential for consumer devices by being seemingly unwilling to negotiate much lower prices for volume customers. >I am surprised that they charge it a separate fee, not bundled with VxWorks. > >The unusual part about your situation from my perspective is that you >are apparently selling chips, and pay VxWorks royalties. Usually, the >people who make systems will pay for VxWorks. Why the heck would we make all of our customers individually negotiate deals with WindRiver? Or for any other license whose price isn't set in stone for everyone? Not only would that be annoying to them, but we can get a better deal than any one customer because we can negotiate based on the full volume. It's an embedded chip and the software running on it, and requiring VxWorks, is mainly ours not that of our customer. >I know a few exceptions >when VxWorks (or actually IoWorks) is bundled with i960 I2O chips, >and know there are a few more who switched to the bundled model, >but still the majority of VxWorks customer are system builders. Majority in terms of number of customers, certainly. But certain large chip builders sell a LOT of units with VxWorks. My employer, LSI Logic, sells tens of millions. In a few years it will be hundreds of millions. >Naturally, I had to assume you belong to that majority. No, you didn't. A better thing to do in terms of presenting advice would for example be to say "IF your chip is an FPGA, you could save lots of money by going ASIC, if it's an ASIC you could probably save money by outsourcing to get a cost-reduced design". Rather than the very unclear phrasing resulting from you just assuming the person in question uses an FPGA, and running from there without clearly stating your assumption. And a better thing to do that that would be, to ask for more details of the product. >Still, I beleive you should be able to negotiate a buyout. ??? --------------------------- Newsgroups: comp.os.vxworks Subject: ANNOUNCE: TargetFFS is being used with VxWorks for Guaranteed Integrity Across Unexpected Resets today! Date: Sun, 13 Oct 2002 19:55:41 GMT From: "Louis Meadows" Organization: Prodigy Internet http://www.prodigy.com Message-ID: <1dkq9.3309$a82.233150663@newssvr21.news.prodigy.com> Let's make it happen and take it to the next 9 ! Blunk Microsystems' embedded flash file system, TargetFFST works well with VxWorks, offers a rugged alternative to mechanical storage systems. Using the NOR flash memory commonly installed on COTS CPU boards, TargetFFST implements a reliable, re-entrant file system with a POSIX and ANSI C compliant application program interface. Benefits: a.. Royalty Free b.. Reliable, Re-entrant Embedded File System c.. Provides POSIX and Standard C API d.. Use of Flash Memory is Invisible to Applications e.. Supports Dynamic Creation and Deletion of Files, Directories, and Links f.. Complete Wear Leveling g.. Guaranteed Integrity Across Unexpected Resets h.. Supports Multiple Volumes i.. Includes Three Sample Applications TargetFFST supports multiple volumes and allows dynamic creation and deletion of files, directories, and links with full read and write capability. TargetFFST is not a static ROM-image file system. The use of flash media for the backing store is invisible to the application layer program. Wear leveling is implemented to prevent premature failure of the flash media. Erase cycles are spread evenly across all blocks. For end-of-life monitoring, the vstat() call provides the current wear count. The wear count is maintained starting with the first time a flash volume is formatted. TargetFFST provides guaranteed integrity across unexpected power losses. New information is written to flash memory without overwriting old information, neither control information nor user data. If an unexpected power loss occurs, directory structures, closed files, and files opened for reading are never at risk. Only data written since the last synchronizing operation (fclose(), fflush(), etc.) can be lost. When the TargetFFST reboots, it restores the file system to its state at the time of the last synchronizing operation. TargetFFST supports the large erasable block sizes typical of flash memory, usually 64KB or larger. It has a "thin" hardware dependent layer for maximum ease in porting to new platforms and supports 8, 16, or 32-bit interfaces to the CPU. A single bank of Flash memory can be shared between a boot program and a TargetFFST volume. The User's Manual includes several example drivers. TargetFFST is a source code product. The source code is 100% ANSI C and has been tested using compilers from Diab Data, Gnu, Metrowerks, and Software Development Systems. The only RTOS dependence is the use of two semaphores. Porting consists of writing wrapper functions for the operations of creating, posting tokens to, and acquiring tokens from semaphores. Prototypes and descriptions of the required wrapper functions are provided in the User's Manual. Three sample applications are supplied: a command line shell, a binary search application, and an upgrade installer. When the upgrade installer runs, it selects the most recent executable file that passes a checksum test and commences downloading an upgrade. The previous executable is deleted after the upgrade completes successfully. If an unexpected power loss occurs, the system resorts to the previously installed upgrade. Louis Meadows ( lmeadows@blunkmicro.com ) Blunkmicro Systems - Major Account Sales 44 - 46 Front Street - Santa Cruz, CA 95060 (831) 459 7673 direct (831) 459 9316 fax We develop, market and support the Target line of real-time development solutions for embedded platforms. see www.embeddedtools.net for more information. ============================================================================ ========== "Speed is fine but accuracy is final" - Wyatt Earp - --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Where can I find a sample of SCC driver for mpc8260 Date: Sun, 13 Oct 2002 14:53:02 -0700 From: "Dan Gold" Organization: Mailgate.ORG Server - http://www.Mailgate.ORG Message-ID: References: Sender: gold@ensemble.com Ming Tai wrote in message ... >Does anyone know >Where can I find a sample of SCC driver for mpc8260 I thought one came with the 8260 BSP from WRS. If not, call your local WRS FAE - they should be able to send it to you. hth, Dan Gold gold@ensemble.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How can a Shared Library access Protection Domain data Date: 13 Oct 2002 15:40:58 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0210131440.4d25ae3e@posting.google.com> References: Hello, The simplest way would be to place the data in a shared data domain (see sdLib). Then, your shared library can cause that shared data domain to be attached to domain1 and domain2 the first time they call. After the SD is attached, your shared library functions can access it without any problems. Likewise, domain3 can attach to it and use the SD space for the data it is generating. HTH, John... steven4443@yahoo.com (stephen hallacy) wrote in message news:... > Hello, > > Background Information: > I have two application domains (say domain1 and domain2) that want to > access API functions from a shared library. Also, I have another > application domain (domain3) that has one task that gathers serial > data and performs computation on the data. I have no control over the > applications in domain1 and domain2. These applications are created > by another company. > > Question: > How can my shared library have access to computation result from > domain3's task? > > Stephen --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How can a Shared Library access Protection Domain data Date: 13 Oct 2002 22:49:13 -0700 From: steven4443@yahoo.com (stephen hallacy) Organization: http://groups.google.com/ Message-ID: References: <488e459a.0210131440.4d25ae3e@posting.google.com> It helps alot! Thanks! :) john_94501@yahoo.com (John) wrote in message news:<488e459a.0210131440.4d25ae3e@posting.google.com>... > Hello, > > The simplest way would be to place the data in a shared data domain > (see sdLib). Then, your shared library can cause that shared data > domain to be attached to domain1 and domain2 the first time they call. > > After the SD is attached, your shared library functions can access it > without any problems. Likewise, domain3 can attach to it and use the > SD space for the data it is generating. > > HTH, > > John... > > steven4443@yahoo.com (stephen hallacy) wrote in message news:... > > Hello, > > > > Background Information: > > I have two application domains (say domain1 and domain2) that want to > > access API functions from a shared library. Also, I have another > > application domain (domain3) that has one task that gathers serial > > data and performs computation on the data. I have no control over the > > applications in domain1 and domain2. These applications are created > > by another company. > > > > Question: > > How can my shared library have access to computation result from > > domain3's task? > > > > Stephen --------------------------- Newsgroups: comp.os.vxworks Subject: Re: taskRestart Date: Mon, 14 Oct 2002 05:55:23 +0000 (UTC) From: Johan Borkhuis Organization: Proxim Corp. Message-ID: References: <6847c39a.0210120342.91911fc@posting.google.com> vittala@hotmail.com (Vittal) wrote: > Hi All, > I am facing a problem with taskRestart. > > I am having two tasks task1 and task2 with the same priority of 0. First, try to run these tasks at a priority of 100 and see what's happening. Then you could try increasing the priority again. But why would you need a priority of 0? Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ANNOUNCE: TargetFFS is being used with VxWorks for Guaranteed Integrity Across Unexpected Resets today! Date: Mon, 14 Oct 2002 10:35:41 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <1dkq9.3309$a82.233150663@newssvr21.news.prodigy.com> Reply-To: "Leonid Rosenboim" Hey, this looks great - I too had a royalte-free source code Flash file system for VxWorks - "FlashLib" between 1996-1998. One key feature FlashLib had is on-the-fly compression, which doubled the storage capacity of flash, and doubled writeing speed at the cost of somewhat reduced read speed (depending on CPU speed). My company was acquired by WIndRiver on Jan 1997, with the intention to discontinue FlashLib when TFFS becomes available, which took place about one year later. Good luck, Leonid "Louis Meadows" wrote in message news:1dkq9.3309$a82.233150663@newssvr21.news.prodigy.com... > Let's make it happen and take it to the next 9 ! > Blunk Microsystems' embedded flash file system, TargetFFST works well with > VxWorks, offers a rugged alternative to mechanical storage systems. Using > the NOR flash memory commonly installed on COTS CPU boards, TargetFFST > implements a reliable, re-entrant file system with a POSIX and ANSI C > compliant application program interface. > > > Benefits: > > > a.. Royalty Free > b.. Reliable, Re-entrant Embedded File System > c.. Provides POSIX and Standard C API > d.. Use of Flash Memory is Invisible to Applications > e.. Supports Dynamic Creation and Deletion of Files, Directories, and > Links > f.. Complete Wear Leveling > g.. Guaranteed Integrity Across Unexpected Resets > h.. Supports Multiple Volumes > i.. Includes Three Sample Applications > TargetFFST supports multiple volumes and allows dynamic creation and > deletion of files, directories, and links with full read and write > capability. TargetFFST is not a static ROM-image file system. The use of > flash media for the backing store is invisible to the application layer > program. > > Wear leveling is implemented to prevent premature failure of the flash > media. Erase cycles are spread evenly across all blocks. For end-of-life > monitoring, the vstat() call provides the current wear count. The wear count > is maintained starting with the first time a flash volume is formatted. > > TargetFFST provides guaranteed integrity across unexpected power losses. New > information is written to flash memory without overwriting old information, > neither control information nor user data. If an unexpected power loss > occurs, directory structures, closed files, and files opened for reading are > never at risk. Only data written since the last synchronizing operation > (fclose(), fflush(), etc.) can be lost. When the TargetFFST reboots, it > restores the file system to its state at the time of the last synchronizing > operation. > > TargetFFST supports the large erasable block sizes typical of flash memory, > usually 64KB or larger. It has a "thin" hardware dependent layer for maximum > ease in porting to new platforms and supports 8, 16, or 32-bit interfaces to > the CPU. A single bank of Flash memory can be shared between a boot program > and a TargetFFST volume. The User's Manual includes several example drivers. > > TargetFFST is a source code product. The source code is 100% ANSI C and has > been tested using compilers from Diab Data, Gnu, Metrowerks, and Software > Development Systems. The only RTOS dependence is the use of two semaphores. > Porting consists of writing wrapper functions for the operations of > creating, posting tokens to, and acquiring tokens from semaphores. > Prototypes and descriptions of the required wrapper functions are provided > in the User's Manual. > > Three sample applications are supplied: a command line shell, a binary > search application, and an upgrade installer. When the upgrade installer > runs, it selects the most recent executable file that passes a checksum test > and commences downloading an upgrade. The previous executable is deleted > after the upgrade completes successfully. If an unexpected power loss > occurs, the system resorts to the previously installed upgrade. > > Louis Meadows ( lmeadows@blunkmicro.com ) > Blunkmicro Systems - Major Account Sales > 44 - 46 Front Street - Santa Cruz, CA 95060 > (831) 459 7673 direct (831) 459 9316 fax > > We develop, market and support the Target line of real-time development > solutions > for embedded platforms. see www.embeddedtools.net for more information. > ============================================================================ > ========== > "Speed is fine but accuracy is final" - Wyatt Earp - > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: PCI devices not on page boundary Date: Mon, 14 Oct 2002 11:33:41 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1034588068.18418@newsreader1.wirehub.nl> Hi, While probing for PCI devices on Intel based (compact) PCI system, I got BAR1 or BAR2 addresses which are not on a page boundary, so usrMmuInit fails on these entries. For example I get address 0xe820:0400 for BAR0, which needs 256 bytes mem space, and this address is not on a page boundary. Any idears how to resolv this? Thanx, Renee - - - - - - - board1: SBS/Greenspring Carrierboard BAR0: 0xE820:0000 (256 bytes) BAR1: IO-space, not used BAR2: 0xEC00:0000 (64MB) board2: SBS/Greenspring Carrierboard BAR0: 0xE820:0400 (256 bytes) BAR1: IO-space, not used BAR2: 0xF000:0000 (64MB) --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Mon, 14 Oct 2002 12:59:09 +0200 From: Tim Sinkins Organization: Alcatel Kommunikations Elektronik Message-ID: <3DAAA37D.EAE13FD0@alcatel-ke.de> References: <3a5ff53f.0210020748.3643d45a@posting.google.com> Reply-To: timothy.sinkins@alcatel-ke.de Ian Montgomerie wrote: > > On Wed, 9 Oct 2002 19:53:15 +0200, "Leonid Rosenboim" > wrote: > > >"brian" wrote in message > >news:d2f67bd8.0210090933.19572732@posting.google.com... > > > >The risk factor here is quite large, much more substantial than any > >hardware modification that would save the same amount on the COGS, > >or even 10 times more ! > > You seem quite confident in the detailed costs of someone else's > project. If my project paid royalties for our RTOS, which we don't, > switching could be a very good idea. I think the basic assumption made about the worth of switching is that the user has the right deal in the first place. Anyone who is producing a product by the millions (or even 100,000s) who pays a royalty, has struck a dumb deal and they would perhaps be better off renegotiating with WRS on the threat of moving OS. OTOH if you already have a single-site/processor/product license which you want to reproduce on another site/processor/product and you simply don't like the WRS fee for doing this, it is IMHO very unlikely to be possible to switch for less than the cost of this fee unless that fee is very excessive. The new OS will have its own license fee and then there is the unknown cost of implementing it in your product and for re-training all you staff, the total cost of this is going to be into six figures and I would hope that a license from WRS costs less than this (I know that the comparable competition does). Finally, already having a single-site/processor/product license and changing OS just to save on support fees (or whatever) is so economically stupid as to not be worthy of consideration. Tim > It would in fact be less risky > than trying to save another 25 cents on our design (over 5% of overall > unit cost), since our only non-software means of doing so is the > risky, multimillion dollar option of cost-reducing the silicon. Then > for some other projects, saving 25 cents per unit just means cutting > out some unecessary screws. --------------------------- Newsgroups: comp.os.vxworks Subject: Tornado PPC8xx GNU compiler conventions Date: Mon, 14 Oct 2002 14:05:01 +0200 From: "Martin Roth" Organization: Motorola Message-ID: I have impemented some time critical code in assembly of PPC860. Where can I find the cross compiler's technical documentation to get the exact register usage conventions of the compiler? ( calling conventions, such as which registers are used to pass arguments and return values, which registers are preserved (saved) across function calls, etc... ) Martin --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Routing problem Date: 14 Oct 2002 12:22:34 +0100 From: Simon Wright Organization: Alenia Marconi Systems, ISD, Farlington Message-ID: References: Sender: sjw@galadriel.frlngtn.gecm.com Gunter.Ritter@tenovis.com (Gunter Ritter) writes: > The problem seems to be that the messages from 192.1.2.55 appearing > in the network 192.1.2.0 still have the original source address > 192.1.2.55 in the IP header and not 192.1.1.141. This explains why > it works with a default gateway entry on 192.1.1.1. That is how it is meant to work. You are thinking of NAT, I suspect. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Tornado PPC8xx GNU compiler conventions Date: Mon, 14 Oct 2002 14:52:41 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Reply-To: "Leonid Rosenboim" "Martin Roth" wrote in message news:aoebuc$gkb$1@newshost.mot.com... > I have impemented some time critical code in assembly of PPC860. > > Where can I find the cross compiler's technical documentation to get the > exact register usage conventions of the compiler? ( calling conventions, > such as which registers are used to pass arguments and return values, which > registers are preserved (saved) across function calls, etc... ) > > Martin > > Hello Martin, how are you doing these days ? There is no such documentation that you require. The exact description of register usage in GCC exist as part of the GCC source code, although it's not stricktly a C source file, it is part of the program. Alternatively, many architectures have an agreed-upon ABI (Application binary interface) standard that allows interoperability of different compilers, and I beleive the is an ABI for PowerPC designed originally by Apple+Motorola+IBM, and most chanses are the GCC on PowerPC complies with the ABI. - -- HTH - ----------------------------------------------------------------------- Leonid Rosenboim Visit: http://gamla.org.il/english/index.htm Consultant Email: my first name at consultant dot com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: taskRestart Date: Mon, 14 Oct 2002 09:52:25 -0400 From: Keith Arner Organization: Marconi Message-ID: References: <6847c39a.0210120342.91911fc@posting.google.com> Reply-To: Keith Arner On 12 Oct 2002, Vittal wrote: > restartCount = 0; /* global count */ > > code snippet from task2.... > > if (OK == taskRestart (taskNameToId ("task1")) > restartCount = 1; > else > restartCount = 2; > > After the execution of this, task1 is restarted. I have confirmed this > with some prints and global variables. However, the restartCount > remains 0. I expected it to be 1 since the task1 was restarted. If > there was some problem, at least it should have got changed to 2. But, > the restarCount remains 0 and this is something which I am not able to > understand. > > I am not able to connect the debugger due to some problems with the > network and hence not able to debug it properly. That seems particularly odd. If task1 and task2 are both the same priority, then task2 should be the active task when task1 becomes ready. Thus, task2 should run to completion before task1 gets to execute its first instruction. Try putting a taskDelay(1) at the beginning of task1, and see if that allows task2 to execute. I would normally suggest using WindView to debug this sort of thing, but if your network stack is gone, that won't work. But, as other people have suggested, if you change the priorities of your tasks to something below that of tNetTask, you might be able to get your network stack back online. Keith - -- "Once something has gone horribly wrong..." "...it just gets horribler and horribler." "Then, it's a matter of finding the original horriblism." -- from a conversation on debugging problems in a linked list --------------------------- Newsgroups: comp.os.vxworks Subject: Request: ftpdlib server code Date: 14 Oct 2002 08:20:03 -0700 From: qingdan@yahoo.com (QingDan) Organization: http://groups.google.com/ Message-ID: <704122b3.0210140720.3899aebd@posting.google.com> Hi, there, We are working on a specific build of FTP mechanism using vxWorks' ftplib and ftpdLib. However, we need to remove ftpdLib's SECURITY and incorporate some of our own features into vxworks' ftpdLib. I was told there is a unsupported source code ftpdlib.c under $WIND_BASE/target/unsupported/src/netwrs/. Unfortunately it is not found in our installation. I'm wondering if anybody can do me a favor by sending me a copy of ftpdLib.c and related files if you have access to above directory? Thanks so much in advance. QingDan --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Request: ftpdlib server code Date: Mon, 14 Oct 2002 17:45:51 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <704122b3.0210140720.3899aebd@posting.google.com> Reply-To: "Leonid Rosenboim" "QingDan" wrote in message news:704122b3.0210140720.3899aebd@posting.google.com... > Hi, there, > > We are working on a specific build of FTP mechanism using vxWorks' > ftplib and ftpdLib. However, we need to remove ftpdLib's SECURITY and > incorporate some of our own features into vxworks' ftpdLib. I was told > there is a unsupported source code ftpdlib.c under > $WIND_BASE/target/unsupported/src/netwrs/. Unfortunately it is not > found in our installation. I'm wondering if anybody can do me a favor > by sending me a copy of ftpdLib.c and related files if you have access > to above directory? Thanks so much in advance. > > QingDan This source is available only with installation of DosFs2.0. It is "unsupported" but is copyrighted nevertheless, therefore it would be an infringement on WindRiver copyrights, and therefore illegal for anyone to send you this file, except by WindRiver themselves. Contact your friendly FAE or support at WindRiver, if you want the FTP server code but not the entire DosFs2 package (which is a free upgrade for Tornado 2.0 and 2.1, and comes bundled with Tornado 2.2 and 3). Also, (since I did those ftpdLib enhancements), I'd like to inform you that the security "login" mechanism is customizable, without source modification, i.e. you can write your own function with API as in loginLib, and plug them into the FTP server. - -- HTH - ----------------------------------------------------------------------- Leonid Rosenboim Visit: http://gamla.org.il/english/index.htm Consultant Email: my first name at consultant dot com --------------------------- Newsgroups: comp.os.vxworks,comp.realtime,comp.arch.embedded,comp.answers,news.answers Subject: [comp.os.vxworks] FAQ - index Date: 13 Oct 2002 15:35:21 GMT From: johan@borksoft.xs4all.nl (Johan Borkhuis) Organization: Borksoft Message-ID: Followup-To: comp.os.vxworks Archive-name: vxworks-faq/index Posting-Frequency: monthly Last-modified: 2002/08/23 Version: 1.46 URL: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html Copyright: (c) 1999 - 2002 Johan Borkhuis Maintainer: Johan Borkhuis This is the index of the FAQ for the newsgroup comp.os.vxworks. The FAQ itself can be found at http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html Original charter of comp.os.vxworks: comp.os.vxworks will be a forum for discussion of topics related to the VxWorks real-time operating system and development environment. Current and prospective users of VxWorks will be able to discuss technical problems, report experiences with various hardware components, share code, report bugs and fixes, and exchange any information of interest to the VxWorks user community. Relevant new product announcements may be made by commercial participants, subject to a limit - e.g. 1/month - on the frequency of such posts. Contents of the FAQ: 1. Tool problems 1.1 Compiler and Linker 1.2 Debugger 1.3 FTP 1.4 Host tools 1.5 Installation 1.6 Make 1.7 Project facility 1.7.1 Hierarchical projects 1.7.2 Super projects 1.8 Target Server 1.9 Target Shell 1.10 Telnet 1.11 Tornado 1.11.1 Tornado (General) 1.11.2 Tornado (Windows) 1.12 Version Control 1.13 Visual Studio integration 1.14 Windsh 1.15 WindView 2. Changes/additions to be made to BSP's 2.1 Different BSP's 2.2 Processor specific issues 2.2.1 Power PC 2.2.2 I960 2.2.3 MIPS 2.2.4 ARM 2.3 Bootstrap 2.4 Adding component to your VxWorks configuration 2.5 VME problems 2.6 VxWorks environment 2.6.1 usrSerial.c 2.6.2 prj_vxworks.tcl 2.7 PCI problems 3. File System problems 3.1 Dos file system 3.2 Flash File system 3.3 Floppy-disk File system 3.4 RAM-disk File system 3.5 General file system questions 4. Network 4.1 Configuration problems 4.2 ARP 4.3 DHCP 4.4 FTP and TFTP 4.5 PPP 4.5.1 PPP on Windows 95 4.5.2 PPP on Windows NT 4.5.3 PPP on Solaris 4.6 Sockets 4.7 Telnet 4.8 Other network related questions 5. VxWorks Questions 5.1 C++ issues 5.2 Communication problems 5.3 Interrupts 5.4 LoadModule problems 5.5 pthreads 5.6 Reboot 5.7 Semaphores 5.8 Simulator (VxSim) 5.9 Task related items 5.10 Time/timer related items 5.11 Wind Web Server related items 5.12 Zinc/windML related items 5.13 Other items 5.14 VxWorks AE issues 6. Extra information 6.1 Hard delay 6.2 Memory leaks 6.3 Corba engines 6.4 Web servers 6.5 NTP usage 6.6 Performace / Benchmarks 6.7 SNMP 6.8 Lint 6.9 Encryption Other Links The News Department Overall question index Overall Sources index Change history - -- o o o o o o o . . . _____J_o_h_a_n___B_o_r_k_h_u_i_s___ o _____ || http://www.xs4all.nl/~borkhuis | .][__n_n_|DD[ ====_____ | johan@borksoft.xs4all.nl | >(________|__|_[_________]_|________________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` ==== VxWorks-FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html ==== - ----- End forwarded message ----- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Tornado PPC8xx GNU compiler conventions Date: Mon, 14 Oct 2002 18:19:28 +0200 From: "Martin Roth" Organization: Motorola Message-ID: References: Leonid , I have found the answers in DIAB C compiler manual. I hope that the calling conventions of GNU and DIAB compilers for PPC860 are the same. Correct me if I am wrong. Thanks, Martin "Leonid Rosenboim" wrote in message news:newscache$yk2z3h$1l8$1@lnews.actcom.co.il... > > "Martin Roth" wrote in message > news:aoebuc$gkb$1@newshost.mot.com... > > I have impemented some time critical code in assembly of PPC860. > > > > Where can I find the cross compiler's technical documentation to get the > > exact register usage conventions of the compiler? ( calling conventions, > > such as which registers are used to pass arguments and return values, > which > > registers are preserved (saved) across function calls, etc... ) > > > > Martin > > > > > Hello Martin, how are you doing these days ? > > There is no such documentation that you require. The exact description of > register usage in GCC exist as part of the GCC source code, although it's > not stricktly a C source file, it is part of the program. > > Alternatively, many architectures have an agreed-upon ABI (Application > binary interface) standard that allows interoperability of different > compilers, and I beleive the is an ABI for PowerPC designed originally by > Apple+Motorola+IBM, and most chanses are the GCC on PowerPC complies with > the ABI. > > -- HTH > ----------------------------------------------------------------------- > Leonid Rosenboim Visit: > http://gamla.org.il/english/index.htm > Consultant Email: my first name at consultant dot com > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Tornado in multi-person projects Date: Mon, 14 Oct 2002 20:27:50 +0000 From: molochai Organization: Griffin Internet - http://www.griffin.net.uk Message-ID: <7c9foa.272.ln@gonzo.cybermuppet.co.uk> References: BruceL wrote: > Hi, > > I'm currently the only person building our application code and I don't > share .wpj files, but that is soon to change. I am seriously considering > ditching Tornado and using a standard build script (originally generated > by Tornado) from the command line and/or my editor. > > My 2 complaints with Tornado is that to share .wpj files, directory names > have to match exactly. Also, there is no "GET" rule; i.e., no way to > browse the most recent revision controlled files from the network. > > I would like to know if other people are managing this and what the best > way to go is. What manage this by having to types of build: * those initiated from within Tornado II of individual libraries and applications * those initiated from a top-level makefile which uses a small fragment of TCL which ISTR is available on one of the FAQs to generate makefiles from .wpj files, and then calls the relevent build rules on them. We use both approaches as when projects start getting big, running a large set of nested makefiles can take quite a while even if few targets are to be built. Often when developing and continously building one project at a time, the former approach is much faster. We use the latter approach to synchronise a developer's sandbox and for release builds. Adding additional make rules to get source codes files from VC is not going to be too hard to do. If using Windoze you might also find forcing your .wpj files to use protject relative paths to be a good idea, and WindRiver conveniently provide a macro ($PRJ_DIR) to do this with. Unfortunately, Tornado by default places absolute paths in the .wpj file, so some handeditting is required. --------------------------- Newsgroups: comp.os.vxworks Subject: TMS: Web Server and MIB Date: Mon, 14 Oct 2002 16:22:23 -0500 From: "AW" Organization: NTT/Verio Message-ID: Hi there, Is there anybody who is familiar with Tornado for Managed Switches (TMS)? I have built custom MIB and managed to insert it into IDB+. TMS is using Wind Web Server to access the MIB tree. My problem is that I can't access my MIB from web pages. The steps I have done are: 1. Create a custom HTML page 2. Create an association .c file to define the symbols (mibgroupName_mibnodeName) 3. Recompile webif directory 4. Recompile vxWorks BSD And I still get "Symbol not found" error. Thank you very much in advance Adrian --------------------------- Newsgroups: comp.os.vxworks Subject: Re: can't load .out files with dosfs2 Date: 14 Oct 2002 16:02:53 -0700 From: term@xerosoft.com (Tom) Organization: http://groups.google.com/ Message-ID: <15cdb0fb.0210141502.18a17700@posting.google.com> References: <15cdb0fb.0210021822.572bd28c@posting.google.com> <15cdb0fb.0210090822.52841a20@posting.google.com> <15cdb0fb.0210101559.2fa61919@posting.google.com> <15cdb0fb.0210110847.79d170a7@posting.google.com> Leonid, Ok, you were right. I wrote some simple code to test lseek/read calls on a deterministic binary file (the file just counts up by 1 every 32 bit word, like you sugested) and the test fails horribly. If I run my test code and have it do 20 lseeks/reads, all 20 reads will contain the wrong data. If I do the same test using a test binary file thats located on my ftp server, then I am able to do over 1,000,000 lseeks/reads without any errors. If anybody knows of a patch for this, please let me know. Thanks again for the help, Tom term@xerosoft.com (Tom) wrote in message news:<15cdb0fb.0210110847.79d170a7@posting.google.com>... > Hey Leonid, > > See comments below... > > -Tom > > "Leonid Rosenboim" wrote in message news:... > > Good work, Tom. > > > > I do recall something about a certain patch version of dosFs2 that > > has a problem in lseek(), and the reason I suggested you try > > pipes was to doube check that this ld indeed uses lseek. It does. > > And it clearly shows what i exactly different beween simply copying > > the file and loading it as a module. > > > > What you need to do now is two things: > > 1. With the help of WRS support (sorry you have to do this), starting with > > on-line, see if there is an SPR about dosFs2 and lseek, and if you find it, > > get the patch. > > > > My companies support contract has expired and we are tight on cash, so > renewel probably isn't going to happen for a while. I sent them an > e-mail just incase they felt like helping anyway, but I got no reply > (big surprise ;) ). > > > 2. If #1 does not work, write s simple test program that first creates a > > sizable file, and initializes it with very simple, expectable data, e.g. > > every 32-bit word in the file would simple contain its offset relative > > stanrt of file. Then the program begins a series of lseek() calls using > > random offsets modulus the file size, reads a word from that offset and > > verifies it is what was expected. I am pretty sure this will demonstrate the > > bug if there is indeed an lseek bug. You might get away with a simple script > > that repeats the calls made by ld, e.g. read 48 bytes from beginning, lseek > > 300, read 128 bytes, ... > > I had a test like that in the fstest suite we developed in parallel with > > dosFs2 (actually this suite begun with cdromFs, later was extended for > > FlashLib and then extended more for patching dosFs1 and dosFs2 development, > > but I figure nobody at WRS relelbers this suite ever existed, bammer!) > > > > I would like to make sure the problem is in fact with lseek, so this > should be a good test. I'll try it sometime today and post the > results. > > > 3. If you are bored in the mean time, you could also creatre a ramDisk and > > try to copy the file over and load it from ramDisk, but most chanses are it > > will fail too becase it would use the same dosFs2. > > > > I tried this before and it failed (that was the reason I started to > suspect the problem was related to dosFs rather than TFFS). I did > however try going through memDrv and that worked fine, so now I have a > workaround that I can use for loading out of flash (yay!). Of course > it would be better if I could use ld directly, but that may not be an > option for the time being. > > > Just let me know when you need my address to mail me a check ;-) > > I'll see what I can do, but most likely my company isn't going to want > to cut a check. I do apreciate your help though. > > > ----------------------------------------------------------------------- > > Leonid Rosenboim Visit: http://www.masada2000.org/ > > Consultant Email: my first name at consultant dot com > > > > > > "Tom" wrote in message > > news:15cdb0fb.0210101559.2fa61919@posting.google.com... > > > Leonid, > > > > > > I tried the test you proposed and loaded the .out file into a pipe, > > > set stdin to be the pipes FD, and then did an ld with name == NULL so > > > that it would try to load from stdin. This was all done in one task, > > > I just made sure the pipe was very large; this isn't a problem is it? > > > Unfortunately the call to ld failed. > > > > > > I hadn't really thought about the lseek implementation being an issue, > > > but I think you are on to something. I ran this test again but set a > > > breakpoint on lseek (right before ld is called). What I saw was that > > > the first time lseek gets called it is trying to do a seek on STDIN > > > (fd = 0) to offset 300 from the begining of the file. This seems > > > valid but I don't believe pipes support seeking, so the call returns a > > > -1. > > > > > > Maybe the version of dosFs that I have has a bug in its lseek > > > implementation? What do you think? I guess I could write my own > > > pipes style driver that supports lseek and use that. Can you think of > > > any other things to try? > > > > > > Thanks for the help, > > > > > > -Tom > > > > > > "Leonid Rosenboim" wrote in message > news:... > > > > This all seems so strange that it is probably something > > > > very simple we all have ignored. > > > > > > > > If you have the exact same file effectively loading > > > > from a network, there should be no reason at all > > > > this file won't be loading from dosFs, well except > > > > that different file systems handle the lseek() call > > > > differently, and most architectures dynamic > > > > loader will use lseek(). > > > > > > > > How about if you create a pipe, copy the file > > > > into the pipe, while another task runs loadModule > > > > with the pipe being its stdin ? (I suppose you iknow > > > > how to do this). > > > > > > > > Keep me posted. > > > > ----------------------------------------------------------------------- > > > > Leonid Rosenboim Visit: http://www.masada2000.org/ > > > > Consultant Email: my first name at consultant dot > com > > > > > > > > > > > > "Tom" wrote in message > > > > news:15cdb0fb.0210090822.52841a20@posting.google.com... > > > > > Hey Leonid, > > > > > > > > > > I have tried a number of different command lines to load the file, but > > > > > I only showed one in my first e-mail. > > > > > > > > > > I also tried: > > > > > > > > > > ld 1,1,"test.out" > > > > > ld 1,1,"/tffs0/test.out" > > > > > ld < test.out > > > > > ld < /tffs0/test.out > > > > > and loadModule() with a file descriptor that I manually opened. > > > > > > > > > > unfortunatelly they all failed. > > > > > > > > > > Any other ideas? > > > > > > > > > > > > > > > Best regards, > > > > > > > > > > Tom > > > > > > > > > > "Leonid Rosenboim" wrote in message > news:... > > > > > > Could you try to replace this command: > > > > > > > > > > > > > -> ld < test.out > > > > > > > ld error: error reading file (errno = 0x3d0001). > > > > > > > value = 0 = 0x0 > > > > > > > > > > > > with this command ? > > > > > > > > > > > > -> ld "/tffs0/test.out" > > > > > > > > > > > > or this command ? > > > > > > > > > > > > -> ld < /tffs0/test.out > > > > > > > > > > > > -- HTH > > > > > > ----------------------------------------------------------------------- > > > > > > Leonid Rosenboim Visit: > http://www.masada2000.org/ > > > > > > Consultant Email: my first name at consultant > > dot > > com > > > > > > > > > > > > > > > > > > "Tom" wrote in message > > > > > > news:15cdb0fb.0210021822.572bd28c@posting.google.com... > > > > > > > This is a repost of a previous message I had posted with new > > > > > > > information included. > > > > > > > > > > > > > > I have been working on getting a strong arm based motherboard to > run > > > > > > > vxworks. One problem I have run into is that I can't load any > .out > > > > > > > files off of my TFFS (True Flash File System). When I try to it > just > > > > > > > gives me an error saying "error reading file (errno = 0x3d0001).". > I > > > > > > > looked up the errno and it resolved to S_objLib_OBJ_ID_ERROR , but > > > > > > > this is generated from something at a low level that I don't have > the > > > > > > > source code to. I recently tried stepping through the assembler > code > > > > > > > to see where this error was coming from and it turns out errno is > not > > > > > > > set until the cleanup code runs after fioRead fails, so this is > errno > > > > > > > is irrelevant. > > > > > > > > > > > > > > Every other operation with the TFFS seems to work just fine. I > can > > > > > > > create, delete, read, write, and rename files. As an example I > tried > > > > > > > copying an out file off my FTP server onto flash and loading it, > but > > > > > > > it failed. Then after that I copied the file off of the TFFS to a > new > > > > > > > file name on my FTP server and then loaded it from that file (over > the > > > > > > > network) and it worked fine. > > > > > > > > > > > > > > -- start dump -- > > > > > > > > > > > > > > -> cd "/tffs0/" > > > > > > > value = 0 = 0x0 > > > > > > > -> ll > > > > > > > -rwxrwxrwx 1 0 0 10064 Jan 1 1980 test.out > > > > > > > value = 0 = 0x0 > > > > > > > -> ld < test.out > > > > > > > ld error: error reading file (errno = 0x3d0001). > > > > > > > value = 0 = 0x0 > > > > > > > -> copy "/tffs0/test.out", "host:newtest.out" > > > > > > > Copy OK: 10064 bytes copied > > > > > > > value = 0 = 0x0 > > > > > > > -> ld < host:newtest.out > > > > > > > value = 12531032 = 0xbf3558 > > > > > > > -> > > > > > > > > > > > > > > -- end dump -- > > > > > > > > > > > > > > > > > > > > > After these tests I thought the problem might have been size > related, > > > > > > > so I tried to create a minimal "hello world" out file, but that > didn't > > > > > > > work either. > > > > > > > > > > > > > > As a work around I thought I would try setting up a runscript that > > > > > > > creates a ram drive and then copies each file to the ramdrive, > loads > > > > > > > from there, deletes the file, and then continues on to the next > one > > > > > > > until it is done. Unfortunatelly this failed with the same error > > > > > > > message that I got when I loaded off of a TFFS. The only thing I > can > > > > > > > think of that the ramdrive and TFFS have in common that a ftp file > > > > > > > system doesn't is that they use dosFS (version 2). > > > > > > > > > > > > > > Has anyone else had any problems like this? I'm running > > > > > > > VxWorks 5.4.2 with Tornado 2.1.1. > > > > > > > > > > > > > > > > > > > > > Best Regards, > > > > > > > > > > > > > > Tom --------------------------- Newsgroups: comp.os.vxworks Subject: Writing an assmbler trap routine Date: Tue, 15 Oct 2002 08:24:02 +0000 (UTC) From: "Steve Comfort" Organization: Mailgate.ORG Server - http://www.Mailgate.ORG Message-ID: I am attempting to write a small test program to generate an interrupt. After much scratching of my head over the GNU reference on asm, I have tried #define DO_INT_3 __asm__ ("int 3") Which at least does not give me syntax errors. The compiler however decides that Error: operands given don't match any known 386 instruction Which has me stumped. Any help would be appreciated.. - -- Posted via Mailgate.ORG Server - http://www.Mailgate.ORG --------------------------- Newsgroups: comp.os.vxworks Subject: question about the target server. Date: Tue, 15 Oct 2002 12:15:02 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1034677212.142808@newsreader1.wirehub.nl> For a project we need to share a filesystem between a server and a few targets (max 8), so far we where thinking about using the target server for this, sinds we only need to exchange a few settings files. But I we reboot a target with a connection to the target-server, this connection does not come back, we have to restart the target server to make things work again. Is this normal behaviour? Or should we switch to another solution, like NFS. I rather not, sinds we are in a windows based compagny without NFS servers or people to support it. Any idears? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: question about the target server. Date: Tue, 15 Oct 2002 12:51:36 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <1034677212.142808@newsreader1.wirehub.nl> Reply-To: "Michael Lawnick" Sender: mlawnick@pd9e650d5.dip.t-dialin.net Hi Renee, "Renee Teunissen" schrieb im Newsbeitrag news:1034677212.142808@newsreader1.wirehub.nl... > > > > For a project we need to share a filesystem between a server and a few > targets (max 8), so far we where thinking about using the target > server for this, sinds we only need to exchange a few settings files. > > But I we reboot a target with a connection to the target-server, this > connection does not come back, we have to restart the target server to > make things work again. Is this normal behaviour? Yes > Or should we switch > to another solution, like NFS. Yes > I rather not, sinds we are in a windows > based compagny without NFS servers or people to support it. > > Any idears? Else: FTP. Use the Windows FTP server, not the one part of Tornado, to avoid security holes. HTH Michael --------------------------- Newsgroups: comp.os.vxworks Subject: Re: question about the target server. Date: 15 Oct 2002 09:36:52 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0210150836.3163682b@posting.google.com> References: <1034677212.142808@newsreader1.wirehub.nl> Hello, "Renee Teunissen" wrote in message news:<1034677212.142808@newsreader1.wirehub.nl>... > For a project we need to share a filesystem between a server and a few > targets (max 8), so far we where thinking about using the target > server for this, sinds we only need to exchange a few settings files. > > But I we reboot a target with a connection to the target-server, this > connection does not come back, we have to restart the target server to > make things work again. Is this normal behaviour? No, normally the target server will reattach to the target once it reboots automatically. Are you using any special options to the target server? Which version fo Tornado are you using? How are you rebooting the target? > Or should we switch > to another solution, like NFS. I rather not, sinds we are in a windows > based compagny without NFS servers or people to support it. You can easily set up an NFS server on an old PC using something like Linux or FreeBSD; just install Samba as well so that your PCs can share files with it. This is a very effective way of getting a file server. Of course, you'd probably need to support it yourselves, but look on that as a learning experience ;-) HTH, John... --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: 15 Oct 2002 09:42:33 -0700 From: brian_doe@hotmail.com (brian) Organization: http://groups.google.com/ Message-ID: References: <3a5ff53f.0210020748.3643d45a@posting.google.com> "Leonid Rosenboim" wrote in message news:... > "Ian Montgomerie" wrote in message > news:hsjfqusv57hqols3ft8fr1anpf4psgho2j@4ax.com... > > [snip] > > > You seem quite confident in the detailed costs of someone else's > > project. If my project paid royalties for our RTOS, which we don't, > > switching could be a very good idea. It would in fact be less risky > > than trying to save another 25 cents on our design (over 5% of overall > > unit cost), since our only non-software means of doing so is the > > risky, multimillion dollar option of cost-reducing the silicon. Then > > for some other projects, saving 25 cents per unit just means cutting > > out some unecessary screws. > > > > > What puzzles me is that people ask for advise, and then argue with the > advise they get. If you have a strong opinion, why bother ask anyone ? > > Strange. > Leonid I dont know if this was addressed to me (the original poster)..but it wasnt me that posted the above comment/arguments. Anyway..thanx for all the input Brian --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: 15 Oct 2002 09:46:02 -0700 From: brian_doe@hotmail.com (brian) Organization: http://groups.google.com/ Message-ID: References: <3a5ff53f.0210020748.3643d45a@posting.google.com> <3DAAA37D.EAE13FD0@alcatel-ke.de> > I think the basic assumption made about the worth of switching is > that the user has the right deal in the first place. > Anyone who is producing a product by the millions (or even 100,000s) > who pays a royalty, has struck a dumb deal and they would perhaps be > better off renegotiating with WRS on the threat of moving OS. Possibly a dumb deal. But I can recount many companies such as Philips, Sony (many products). to name two, that pay royalties on products that ship in 6-7 figures. Maybe Sony is dumb as well. Funny, I never thought they were. Thanx for the input. Brian --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Tornado PPC8xx GNU compiler conventions Date: 15 Oct 2002 09:46:21 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0210150846.69a4143b@posting.google.com> References: Hello, Try a web search for two documents: PowerPC System V ABI PowerPC EABI Or you can read this document that IBM have on their web site about writing EABI compliant applications: http://www-3.ibm.com/chips/techlib/techlib.nsf/techdocs/852569B20050FF77852569970071B0D6/$file/eabi_app.pdf For the most part, VxWorks is EABI compliant, so this should get you most of the stuff you need. HTH, John... "Martin Roth" wrote in message news:... > I have impemented some time critical code in assembly of PPC860. > > Where can I find the cross compiler's technical documentation to get the > exact register usage conventions of the compiler? ( calling conventions, > such as which registers are used to pass arguments and return values, which > registers are preserved (saved) across function calls, etc... ) > > Martin --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: 15 Oct 2002 13:00:09 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: References: <3a5ff53f.0210020748.3643d45a@posting.google.com> Sender: bpringle@DeadDuck [snip] Ian> inflexibly charging that much for it. In fact, when it comes to Ian> consumer electronics products using flash cards such as Ian> SmartMedia, Compact Flash, and Memory Stick (FAT16 based Ian> devices) - they all use ERTFS, nobody in consumer electronics [snip] People do use DosFs in commercial products. I know that one ADSL modem uses DOSFS. However, I am interested in your experience with ERTFS. I am generally interested in anyones experience with COMSOL ("http://www.computer-solutions.co.uk/"). How does this software port to vxWorks? Are there seperate 8-bit friendly products or is the software for the 8051 the same that would run on a MIPS, etc 32-bit devices? Regards, Bill Pringlemeir. - -- Look for your teddy bear with one eye. Your finances will become important in the next several months. A wallflower will be a sick spouse. Loving your vehicle now will turn out to be futile after a few days. You'll soon be tall. 12 is your lucky number. --------------------------- Newsgroups: comp.os.vxworks Subject: Trouble debugging 8265 with VisionProbe isync and sync instructions. Date: 15 Oct 2002 10:16:25 -0700 From: FrankGleeson@motorola.com (frankg) Organization: http://groups.google.com/ Message-ID: I am using EST/WindRivers VisionClick software patched to v7.9A. We connect to our target using JTAG and VisionProbe II licensed for 8260. I am are using the wrSbc8260 BSP as a starting point to develope an 8265 specific BSP. While single-stepping through the initializtion code in romInit.s. I found that the VisionClick would loose control of the target whenever I executed either an isyn or sync instruction. A dialog pops up saying something like Error (30045) could not step target PC=address. If I hit the single step button again the PC is advanced to some "random" value. I do not know if this was a random jump or if the VisionProbe simply can not halt the target at the next instruction so the processor keeps executing more instructions. My first thought was an unhandled exception so I put in exception vectors for all the exceptions in the green book. All of my exception vectors spin in a tight loop. I also put a .org statement in to push my code up to FFF03000. Putting my init code out of the range reserved for exceptions. None of these measures had an effect the processor/debugger would still flake out whenever it encountered a sync instruction. Finally I commented out all the sync and isync instructions. Then I was able to step through my code and it worked correctly. I realize that at least some of these syncs are neccessary in the production code. Has anyone had a similar experience with the VisionProbe II? Or can anyone thing of a reason why JTAG single stepping would be effected by the isync/sync instructions? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Writing an assmbler trap routine Date: 15 Oct 2002 12:38:26 -0700 From: hsingyuan@yahoo.com (Hsing Yuan) Organization: http://groups.google.com/ Message-ID: <70d78653.0210151138.2bd845e@posting.google.com> References: I do not know much about intel assembly. But if this is PowerPC assembly, you should put a space before "int 3". Otherwise, "int" will be treated as a label while "3" an instruction. Hsing Yuan "Steve Comfort" wrote in message news:... > I am attempting to write a small test program to generate an interrupt. > After much scratching of my head over the GNU reference on asm, I have > tried > > #define DO_INT_3 __asm__ ("int 3") > > Which at least does not give me syntax errors. The compiler however > decides that > > Error: operands given don't match any known 386 instruction > > Which has me stumped. Any help would be appreciated.. --------------------------- Newsgroups: comp.os.vxworks Subject: address translation on MPC750 / Page Table Segment Descriptor Date: Tue, 15 Oct 2002 16:31:40 -0400 From: Michael Reed Message-ID: <3DAC7B2C.88ABB74B@ll.mit.edu> Hello, I'm having trouble debugging the processor bus (between a MPC750 and a Galileo memory controller). I've capatured data accesses on the bus w/ a logic analyzer, but the addresses being accessed don't look like the address I'm trying to access. My code was pretty simple: unsigned int *myPointer; myPointer = 0x700000; *myPointer = 0xFF; So when I single-step this code, I expected to see address 0x700000 flash across the address bus, but I didn't. So I've dived into logical to physical address translation, which took me to Page Tables and segment descriptors. Question is then, if my address is 0x700000, how do I follow it thru the page table and figure out what physical address will come out on the processor bus? THanks, Mike R. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: address translation on MPC750 / Page Table Segment Descriptor Date: Tue, 15 Oct 2002 16:58:10 -0400 From: Varndell Engineering Organization: Posted Via Newsfeeds.com = SPEED+RETENTION+COMPLETION = http://www.newsfeeds.com Message-ID: <3DAC8162.E0043F51@hotmail.com> References: <3DAC7B2C.88ABB74B@ll.mit.edu> Michael Reed wrote: > > Hello, > I'm having trouble debugging the processor bus (between a MPC750 and a > Galileo memory controller). I've capatured data accesses on the bus w/ > a logic analyzer, but the addresses being accessed don't look like the > address I'm trying to access. > > My code was pretty simple: > > unsigned int *myPointer; > > myPointer = 0x700000; > > *myPointer = 0xFF; > > So when I single-step this code, I expected to see address 0x700000 > flash across the address bus, but I didn't. So I've dived into logical > to physical address translation, which took me to Page Tables and > segment descriptors. > > Question is then, if my address is 0x700000, how do I follow it thru the > page table and figure out what physical address will come out on the > processor bus? Is data translation turned on in the msr register? If not, the MMU is not your problem. You need to provide more information about the state of your h/w and software. If you're debugging the interface between the 750 and the system controller, what does work? What part does vxWorks play in this problem? What are you single-stepping with? Where is the code you're stepping through, flash? Do you know that almost none of the MMU related registers are initialized following a POR? Regards, George Varndell - -- Varndell Engineering, LLC Embedded Systems Expertise Need a vxWorks END driver that works? Contact us. http://www.varndellengineering.com - -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- http://www.newsfeed.com The #1 Newsgroup Service in the World! - -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks,comp.arch.embedded Subject: Boot Code Integrity check with CRC Date: 15 Oct 2002 14:06:44 -0700 From: ryeung@earthlink.net (Raymond Yeung) Organization: http://groups.google.com/ Message-ID: We're developing diagnostics for our 8260 based processor board. Boot code, persistent data and VxWorks (i.e. application) are all on flash. One self-test I'm considering for the flash is CRC check. When it comes to the boot code, I've the following questions: + Boot code resides in a erasable sector (it's an AMD flash part with equal sector size). We normally burn boot code with emulator. But in future, we may implement S/W upgrade of the boot code by FTP transfer from host. Given this background, is it common for people to do CRC on boot code? + If I implement my own S-record based loader to flash, and since S-record specifies address, actual data, and its length, boot code content can be scattered over distiguous flash memory regions. How does CRC program figure out where to pick up the "pieces" to compute CRC? + How do we compute CRC on host? I presume we won't compute over an S-record file? Would appreciate any comments or references to articles. Thanks, Raymond --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Starter pack ??? Date: Tue, 15 Oct 2002 17:18:35 -0700 From: "A. Kaul" Organization: National Semiconductor, Santa Clara Message-ID: <3DACB05B.3060800@antispam.com> References: <3DA791FF.8080509@antispam.com> <488e459a.0210121101.377ecb05@posting.google.com> I am waiting for the CDs to be shipped. Any online resources mean time? Looking at an Ethernet driver! Seems that one needs to support both interrupt & polled modes? Can the switch between these modes be requested by the OS at any time? Thanks. - -ak John wrote: > Hello, > > Try the VxWorks Programmer's Guide for starters, then under > target/src/drv/ you should find some source for example character > drivers. What kind of driver are you trying to develop? > > HTH, > > John... --------------------------- Newsgroups: comp.os.vxworks,comp.arch.embedded Subject: Re: Boot Code Integrity check with CRC Date: Tue, 15 Oct 2002 21:53:56 +0000 (UTC) From: "Graham Baxter" Organization: BT Openworld Message-ID: References: Raymond, Consider also using a "magic" number just before the checksum. If it is not in the correct place there is no point in validating the CRC. Any gaps in S-records would be filled with 0xFFFFFFFF as with uninitialized flash. I always leave space for the magic number and checksum at the very end of the image. I build a binary image which just has a checksum set to 0xFFFFFFFF right at the end. As soon as the binary image is built, I use a utility to perform the checksum calculation and write a new file with the correct checksum. You should find a some checksum code free on the web to enable you to write this utility. The same code that you use on your host to compute the checksum can be compiled into your target to validate a checksum. S-records are very wasteful. A binary image with a 32-bit checksum is very safe and does not waste any space. A final word of warning, be careful if your host and target have different Endianesses. - -- Regards, Graham Baxter Freelance Software Engineer (VxWorks and pSOS BSPs) - available shortly for a new assignment Graham Baxter (Software) Limited gbaxter@NOSPAMbcs.org.uk "Raymond Yeung" wrote in message news:d83bb8e3.0210151306.55f6f8a8@posting.google.com... > We're developing diagnostics for our 8260 based processor board. > Boot code, persistent data and VxWorks (i.e. application) are all > on flash. > > One self-test I'm considering for the flash is CRC check. When it > comes to the boot code, I've the following questions: > > + Boot code resides in a erasable sector (it's an AMD flash part with > equal sector size). We normally burn boot code with emulator. But > in future, we may implement S/W upgrade of the boot code by FTP transfer > from host. > Given this background, is it common for people to do CRC on boot code? > > + If I implement my own S-record based loader to flash, and since S-record > specifies address, actual data, and its length, boot code content can > be scattered over distiguous flash memory regions. How does CRC program > figure out where to pick up the "pieces" to compute CRC? > > + How do we compute CRC on host? I presume we won't compute over an S-record > file? > > Would appreciate any comments or references to articles. > > Thanks, > Raymond --------------------------- Newsgroups: comp.os.vxworks,comp.arch.embedded Subject: Re: Boot Code Integrity check with CRC Date: Tue, 15 Oct 2002 17:57:17 -0400 From: joe durusau Organization: WEBUSENET.com Message-ID: <3DAC8F3D.584E2440@bellsouth.net> References: Raymond Yeung wrote: > We're developing diagnostics for our 8260 based processor board. > Boot code, persistent data and VxWorks (i.e. application) are all > on flash. > > One self-test I'm considering for the flash is CRC check. When it > comes to the boot code, I've the following questions: > > + Boot code resides in a erasable sector (it's an AMD flash part with > equal sector size). We normally burn boot code with emulator. But > in future, we may implement S/W upgrade of the boot code by FTP transfer > from host. > Given this background, is it common for people to do CRC on boot code? > Never heard of anyone bothering, since there are checks on the net transfer already. Most checks are some sort of a addition. Nothing wrong with a CRC, I just haven't seen anybody doing it. > > + If I implement my own S-record based loader to flash, and since S-record > specifies address, actual data, and its length, boot code content can > be scattered over distiguous flash memory regions. How does CRC program > figure out where to pick up the "pieces" to compute CRC? > If you are flashing vxWorks_rom or something similar, it doesn't have any "holes" - that is, it fill up from wherever it starts to wherever it ends. You would have to tell the CRC program the starting and ending addresses. > > + How do we compute CRC on host? I presume we won't compute over an S-record > file? > You would do it easily if the two machines have the same wordsize and endian-ness. Otherwise, it can be a mess. I've seen it done by flashing, then computing the CRC on the target, then putting the result somewhere it can be found. There are programs that can verify the flash against the S-records, bit for bit if you wish. > > Would appreciate any comments or references to articles. > > Thanks, > Raymond Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks,comp.arch.embedded Subject: Re: Boot Code Integrity check with CRC Date: Tue, 15 Oct 2002 23:08:25 +0100 From: Stephen Message-ID: References: In article , Raymond Yeung writes >We're developing diagnostics for our 8260 based processor board. >Boot code, persistent data and VxWorks (i.e. application) are all >on flash. > >One self-test I'm considering for the flash is CRC check. When it >comes to the boot code, I've the following questions: > >+ Boot code resides in a erasable sector (it's an AMD flash part with > equal sector size). We normally burn boot code with emulator. But > in future, we may implement S/W upgrade of the boot code by FTP transfer > from host. > Given this background, is it common for people to do CRC on boot code? Sure. I'd certainly consider doing it if there were no other reasons not to, but it might depend on how important CRC integrity was - such as what sort of damage or data loss might occur if the boot code was bogus. >+ If I implement my own S-record based loader to flash, and since S-record > specifies address, actual data, and its length, boot code content can > be scattered over distiguous flash memory regions. How does CRC program > figure out where to pick up the "pieces" to compute CRC? One suggestion would be to make a policy of having the unused areas of flash contain a known pattern, such as 0x00 or 0xFF, and then compute the CRC over the whole lot. >+ How do we compute CRC on host? I presume we won't compute over an S-record > file? Either write a program to analyse the S-record file and compute the CRC, or if you have the capability, allow a special diagnostics mode on the target to *report* the CRC it was expecting when the CRC fails. With a bit of appropriate number handling, you should be able to use this to go back to the source and enter the correct CRC (without modifying anything else). --------------------------- Newsgroups: comp.os.vxworks,comp.arch.embedded Subject: Re: Boot Code Integrity check with CRC Date: Tue, 15 Oct 2002 23:23:21 +0000 (UTC) From: David Laight Organization: BT Openworld Message-ID: References: > > + Boot code resides in a erasable sector (it's an AMD flash part with > equal sector size). We normally burn boot code with emulator. But > in future, we may implement S/W upgrade of the boot code by FTP transfer > from host. > Given this background, is it common for people to do CRC on boot code? I did the following (remember the image you want is binary, not S records...) 1) Add a header to the vxworks image (the ARM one has one we hacked) 2) Put a checksum and size in the header 3) Load the entire image onto the system, validate the checksum before writing to flash (makes sure the file hasn't gine through an ascii ftp) 4) On boot (in rominit.s) validate the checksum - will explode if the update was only partial, instead on going really badly wrong during the uncompress. David --------------------------- Newsgroups: comp.os.vxworks,comp.arch.embedded Subject: Re: Boot Code Integrity check with CRC Date: Tue, 15 Oct 2002 23:22:33 GMT From: Wolfgang Denk Organization: DENX Software Engineering, Germany Message-ID: References: Sender: wd@denx.muc.de (wd) ryeung@earthlink.net (Raymond Yeung) writes: >We're developing diagnostics for our 8260 based processor board. I guess you've looked around for existing solutions to that problem? > Given this background, is it common for people to do CRC on boot code? Yes. >+ If I implement my own S-record based loader to flash, and since S-record If you implement your own S-record based loader to flash, you're wasting your time. This has been done many times before. Did you - for example - have a look at the PPCBoot code? > specifies address, actual data, and its length, boot code content can > be scattered over distiguous flash memory regions. How does CRC program > figure out where to pick up the "pieces" to compute CRC? It doesn't figure it out, you tell it. >+ How do we compute CRC on host? I presume we won't compute over an S-record > file? Why not run the CRC code over a binary image? >Would appreciate any comments or references to articles. use PPCBoot Best regards, Wolfgang Denk - -- Software Engineering: Embedded and Realtime Systems, Embedded Linux Phone: (+49)-8142-4596-87 Fax: (+49)-8142-4596-88 Web: www.denx.de How does a project get to be a year late? ... One day at a time. --------------------------- End of New-News digest ********************** From vxwexplo-errs@csg.lbl.gov Tue Oct 15 20:39:39 2002 From: Alan K Biocca Date: Tue Oct 15 20:39:41 PDT 2002 Subject: Vx Exploder Issues II For those who want to catch up on those articles missing from the mailing list, they are on the website in the archive file for October 2002. We are having a temporary problem with the web interface to accounts. It updates the database, but then gives errors when transferring the data to the files that actually run the list. This means that when the mount is restored the next change will load all the database changes to the files, so the changes are not lost, but may be delayed. Hopefully this will be rectified tomorrow. It is interesting to note that the software that reads the netnews that I modified and recompiled tonite was last compiled in 1992. Amazing we can still run and compile that code trivially. That is a low-cost life-cycle maintenance. Thanks for your patience, Alan Biocca AKBiocca at lbl dot gov VxWorks Exploder Software Author From vxwexplo-errs@csg.lbl.gov Wed Oct 16 04:03:14 2002 From: Vxworks Exploder Date: Wed Oct 16 04:03:15 PDT 2002 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Wed Oct 16 04:03:12 PDT 2002 Subject: Re: Porting of vxWorks. Subject: Re: Boot Code Integrity check with CRC Subject: Re: test Subject: Re: Starter pack ??? Subject: snmp Subject: Re: Starter pack ??? Subject: Re: Porting of vxWorks. Subject: sntp client Subject: Undefined symbols: ___divdi3 with JWorks 4 Subject: Re: Undefined symbols: ___divdi3 with JWorks 4 ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Tue, 15 Oct 2002 20:14:18 -0700 From: Ian Montgomerie Message-ID: <9dkpqugkkodu9p3tdani5rgue5rm2h6g92@4ax.com> References: On 15 Oct 2002 13:00:09 -0400, Bill Pringlemeir wrote: >However, I am interested in your experience with ERTFS. I am >generally interested in anyones experience with COMSOL >("http://www.computer-solutions.co.uk/"). How does this software port >to vxWorks? Are there seperate 8-bit friendly products or is the >software for the 8051 the same that would run on a MIPS, etc 32-bit >devices? ERTFS is robust, fairly well structured, and was easy to port. Comes with support for your typical PC, mac, and Unix processors. Out of the box it is optimized for flash cards - if you want to use a hard disk with optimal performance, you have to play with some defines controlling how much caching is done (especially of the FAT itself), etc. Seems to do a good job of contiguous allocation, etc. Like any FAT file system I've seen, it's not power safe - you may need to run chkdsk after a power failure during certain file system operations, and chkdsk takes a while. Doesn't have quite the suite of "command line" usable DOS-style file management functions that DosFs2 does. If you really look closely, you'll notice a few of the FAT structs defines by ERTFS and DosFs2 don't actually match in some obscure places. I can't recall exactly which ones but I had to mess around a bit to make a common low-level regression that would work on both. I think one or the other doesn't precisely match the Microsoft FAT implementation in all respects, but I didn't bother figuring out which (I don't have much of a direct reference for the MS originals). Since our only application for reading other FAT filesystems is reading flash cards, where ERTFS dominates, I didn't worry too much. --------------------------- Newsgroups: comp.os.vxworks,comp.arch.embedded Subject: Re: Boot Code Integrity check with CRC Date: Wed, 16 Oct 2002 04:39:31 GMT From: Grant Edwards Message-ID: References: In article , Raymond Yeung wrote: > Given this background, is it common for people to do CRC on boot code? Some sort of checksum. It doesn't have to be CRC. I've never seen any failure mode data on flash ROM to indicate that CRC provides better protection than faster methods. CRC has proven advantages for use on communications lines, but I'm not convinced the failure patterns observed on communications lines are the same as those observed in ROMs. > + If I implement my own S-record based loader to flash, and since S-record > specifies address, actual data, and its length, boot code content can > be scattered over distiguous flash memory regions. How does CRC program > figure out where to pick up the "pieces" to compute CRC? I require that the "file" be contiguous. The checksum is added to the image before it is downloaded. > + How do we compute CRC on host? I presume we won't compute over an S-record > file? I wrote a C program to do it on a binary file. In the past, I've used and/or written programs that would operate on hex files (they would also fill in holes in the file). - -- Grant Edwards grante@visi.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: test Date: Wed, 16 Oct 2002 08:36:21 +0200 From: "Martin Raabe" Organization: 1&1 Internet AG Message-ID: References: <1034697969.14176.7.camel@blue.mcneil.com> Hello Sean, I read your message in the news group comp.os.vxworks. Here are about 20 postings a day, day by day! Good to read from you. Ciao - -- Martin Raabe E: Martin.RaabeATNOSPAMBaSystem.de "Sean McNeil" schrieb im Newsbeitrag news:1034697969.14176.7.camel@blue.mcneil.com... > I have not received any email from the VxWorks list in quite a while. > Is the system down because of the upgrade or are things really this > quiet? > > Sean > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Starter pack ??? Date: 16 Oct 2002 00:11:29 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0210152311.481b2b34@posting.google.com> References: <3DA791FF.8080509@antispam.com> <488e459a.0210121101.377ecb05@posting.google.com> <3DACB05B.3060800@antispam.com> Hello, "A. Kaul" wrote in message news:<3DACB05B.3060800@antispam.com>... > I am waiting for the CDs to be shipped. Any online resources mean time? WindSurf, but you'll need an account from Wind River for that - call your sales person/FAE. > Looking at an Ethernet driver! Seems that one needs to support both > interrupt & polled modes? Can the switch between these modes be > requested by the OS at any time? No, polled mode is only used when you switch to external model with the system mode debugging facility. If system mode debugging over ethernet is not important to you (say because you have a JTAG debugger, or are happy with serial line connection for system mode), then you need not bother with the polled mode. That said, it is not normally that difficult to add this mode, and it might even be simpler to debug while you get the (no ISR to deal with ;-) HTH, John... > Thanks. > -ak > > John wrote: > > > Hello, > > > > Try the VxWorks Programmer's Guide for starters, then under > > target/src/drv/ you should find some source for example character > > drivers. What kind of driver are you trying to develop? > > > > HTH, > > > > John... --------------------------- Newsgroups: comp.os.vxworks Subject: snmp Date: Wed, 16 Oct 2002 15:22:11 +0800 From: "eric" Organization: DCI HiNet Message-ID: Sir, I want pass some variable value from NMS to snmp agent. Is snmp protocal can support this?? Please give me advice. Thank you!! Lihao 2002.10.16 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Starter pack ??? Date: Wed, 16 Oct 2002 07:28:25 +0000 (UTC) From: Johan Borkhuis Organization: Proxim Corp. Message-ID: References: <3DA791FF.8080509@antispam.com> <488e459a.0210121101.377ecb05@posting.google.com> <3DACB05B.3060800@antispam.com> "A. Kaul" wrote: > I am waiting for the CDs to be shipped. Any online resources mean time? Yes, go to the Online Support on the WindRiver website. If you don't have a login create one. Or look at this link (you still need a login to view): http://www.windriver.com/windsurf/products/ide/tornado/documentation/index. html Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Wed, 16 Oct 2002 10:38:00 +0200 From: Tim Sinkins Organization: Alcatel Kommunikations Elektronik Message-ID: <3DAD2568.16C701E8@alcatel-ke.de> References: <3a5ff53f.0210020748.3643d45a@posting.google.com> <3DAAA37D.EAE13FD0@alcatel-ke.de> Reply-To: timothy.sinkins@alcatel-ke.de brian wrote: > > > I think the basic assumption made about the worth of switching is > > that the user has the right deal in the first place. > > Anyone who is producing a product by the millions (or even 100,000s) > > who pays a royalty, has struck a dumb deal and they would perhaps be > > better off renegotiating with WRS on the threat of moving OS. > > Possibly a dumb deal. In the current market definately a dumb deal. (current as in "with the number of alternatives available" rather than "because the economy's shit at the moment"). I recall working on a product 10 years ago where they had decided to use pSoS(+). The offer was a royalty and the company said, "nope offer us a royalty free option or we'll go somewhere else". A 5 figure license fee was negotiated. > But I can recount many companies such as > Philips, Sony (many products). to name two, that pay royalties on > products that ship in 6-7 figures. New products or legacy stuff. I can believe that 15 years ago it might have been difficult to get a royalty free OS and if the code is fully working the company might take the view that sticking with paying the royalty is better than a re-write, but for stuff that's being re-written anyway (to a new processor perhaps) it's really dumb. OK, your current supplier will have you over a barrel when it comes to getting a good deal from them, but not so big a barrel that they completely refuse any deal (unless *they* are dumb). Back to your example, I know of one Philips company that wrote their own RTOS to avoid the license fees and quite a comprehensive beast it is too. I guess this decision must have also been about 12 years ago (by knowing when the first product in the range was made). At the time this looked an OK decision, but long term this was also dumb, as all the people that did the work have moved on and the division that wrote it makes no improvement and gives very little support to the division that uses it (which due to the internal politics of large companies is now different, so different it's in a different country!). > Maybe Sony is dumb as well. Funny, I never thought they were. You make your own mind up! We've bought a license fee for our OS for the equivilent cost of 80,000 units and the guy's here think that they've been cheated because the future product options that the fee covers isn't broad enough to cover all the possible upgrade options that we might decide to do! Not even bothering to negotiate is really dumb. Tim > Thanx for the input. > > Brian --------------------------- Newsgroups: comp.os.vxworks Subject: sntp client Date: Wed, 16 Oct 2002 17:15:35 +0800 From: "news.bentium.com" Organization: Bentium Ltd. (CN99) Message-ID: hi , all How to run up the sntp client to sync time from another machine ? Can I decide the precision with the sntp client ? pls help me. --------------------------- Newsgroups: comp.os.vxworks Subject: Undefined symbols: ___divdi3 with JWorks 4 Date: 16 Oct 2002 03:27:46 -0700 From: andreas_lindell75@hotmail.com (Rikard) Organization: http://groups.google.com/ Message-ID: <5947a04.0210160227.50cfdcb@posting.google.com> Hello, I have a problem that seems to be common here but so far I have been unable to solve it. I am using T2.0.2, WindML2.0.3 and Jworks4.0. When I download the EVM to the VxWorks I get the following error: - -> ld < evm Loading D:/Tornado/target/proj/evm / Undefined symbols: ___divdi3 ___moddi3 ___udivdi3 I have found out that this have something with variables of the type long long, these are defined in libgcc.a and I need to link this into my VxWorks image. So far I have been unable to do it. I have tried the following from Tornado IDE: In workspace->builds->default(vxworks image)->linker. -X -N -Le:/tornado/host/x86-win32/lib/gcc-lib/i386-wrs-vxworks/cygnus-2.7.2-960126/libgcc.a When I compile the vxworks image goes through and the kernel is ok and it starts up but I still have the error when I download the EVM. So how do I link my image with the libgcc.a to take away the undefined symbols. From workspace as I did or by editing a makefile in BSP or project? Regards Andreas --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Undefined symbols: ___divdi3 with JWorks 4 Date: Wed, 16 Oct 2002 12:40:19 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: <5947a04.0210160227.50cfdcb@posting.google.com> Reply-To: "Michael Lawnick" Sender: mlawnick@pd9e65198.dip.t-dialin.net Hi Rikard, you don't need to link your vxWorks, but your application with libgcc.a. vxWorks doesn't have any unresolved symbols, otherwise you'd get an error message at link time. So your link with libgcc.a is a noop. If you want to provide the code within vxWorks, you'll have to code a dummy function that is never called but performs the operations that need to be included. HTH Michael "Rikard" schrieb im Newsbeitrag news:5947a04.0210160227.50cfdcb@posting.google.com... > Hello, > > I have a problem that seems to be common here but so far I have been > unable to solve it. I am using T2.0.2, WindML2.0.3 and Jworks4.0. When > I download the EVM to the VxWorks I get the following error: > -> ld < evm > Loading D:/Tornado/target/proj/evm / > Undefined symbols: > ___divdi3 > ___moddi3 > ___udivdi3 > > I have found out that this have something with variables of the type > long long, these are defined in libgcc.a and I need to link this into > my VxWorks image. So far I have been unable to do it. I have tried the > following from Tornado IDE: > In workspace->builds->default(vxworks image)->linker. > -X -N -Le:/tornado/host/x86-win32/lib/gcc-lib/i386-wrs-vxworks/cygnus-2.7 .2-960126/libgcc.a > When I compile the vxworks image goes through and the kernel is ok and > it starts up but I still have the error when I download the EVM. > > So how do I link my image with the libgcc.a to take away the undefined > symbols. From workspace as I did or by editing a makefile in BSP or > project? > > Regards > Andreas --------------------------- End of New-News digest ********************** From vxwexplo-errs@csg.lbl.gov Wed Oct 16 13:39:09 2002 From: Sean McNeil Date: Wed Oct 16 13:39:11 PDT 2002 Subject: Looking for Contacts I am interested in making new contacts within the vxWorks community. I have been working exclusively with Wind River in a diverse capacity for many years now. Prior to that, I worked on many successful projects in both the government and commercial sector. I would like to expand my associations to any and all companies that may find themselves in need of people with embedded solution expertise. If your company is involved in developing an embedded application running under vxWorks (or any other OS), I would like to hear back from you. Thanks, Sean McNeil From vxwexplo-errs@csg.lbl.gov Thu Oct 17 04:03:31 2002 From: Vxworks Exploder Date: Thu Oct 17 04:03:33 PDT 2002 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Thu Oct 17 04:03:28 PDT 2002 Subject: Re: Undefined symbols: ___divdi3 with JWorks 4 Subject: Re: Trouble debugging 8265 with VisionProbe isync and sync instructions. Subject: Is there a 64-bit RawFS (ala DosFS 2.0)? Subject: Re: Is there a 64-bit RawFS (ala DosFS 2.0)? Subject: Re: address translation on MPC750 / Page Table Segment Descriptor Subject: Re: Is there a 64-bit RawFS (ala DosFS 2.0)? Subject: Re: Is there a 64-bit RawFS (ala DosFS 2.0)? Subject: Re: address translation on MPC750 / Page Table Segment Descriptor Subject: Re: Undefined symbols: ___divdi3 with JWorks 4 Subject: target shell Subject: Re: CORBA --> VisibrokerRT vs. ORBexpressRT Subject: Re: Is there a 64-bit RawFS (ala DosFS 2.0)? Subject: Telnet and CLI stuck Subject: Re: Boot Code Integrity check with CRC Subject: Re: pciProbing question.. Subject: dlmalloc.2.7.2 corruption?? Subject: Porting VxWorks Subject: Re: PCI devices not on page boundary Subject: Re: dlmalloc.2.7.2 corruption?? Subject: spr #23225 Subject: Re: Boot Code Integrity check with CRC Subject: Re: Trouble debugging 8265 with VisionProbe isync and sync instructions. Subject: Re: Starter pack ??? Subject: Re: Boot Code Integrity check with CRC Subject: Re: Porting of vxWorks. Subject: Bootrom question Subject: vxworks VLAN implementation Subject: Re: spr #23225 Subject: USB Subject: strange timer bahaviour Subject: Question about aqrchive of objects Subject: Re: dlmalloc.2.7.2 corruption?? Subject: Re: Porting VxWorks Subject: Re: Question about aqrchive of objects Subject: Re: Bootrom question Subject: Re: Porting of vxWorks. ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Undefined symbols: ___divdi3 with JWorks 4 Date: Wed, 16 Oct 2002 15:56:32 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <5947a04.0210160227.50cfdcb@posting.google.com> Reply-To: "Leonid Rosenboim" If you by any chanse install DosFs2.0, you won't have these errors occuring. Or you can extract the relevant object files from libgcc.a, and add them to your vxWorks link command. Here is another alternative method: When you add libgcc.a into the link command, the linker will not use any of the objects from the .a library unless they are required by one of the .o objects mentioned previously, so your other alternative may be to explicitly define these function names in one of the init .c files, for example, you can add this into usrConfig.c #define INCLUDE_INT64_GCC_SUPPORT #ifdef INCLUDE_INT64_GCC_SUPPORT extern void __divdi3(); /* note one _ omitted vs. "nm" symbol */ extern void __moddi3(); extern void __udivdi3(); statuc VOIDFUNCPTR __gcc_i64_funcs[] = { __divdi3, __moddi3, __udivdi3 /* add as needed */ }; #endif this code will create an explicit reference to these functions, so they will be yanked from libgcc.a during the link as per your current makefile. "Rikard" wrote in message news:5947a04.0210160227.50cfdcb@posting.google.com... > Hello, > > I have a problem that seems to be common here but so far I have been > unable to solve it. I am using T2.0.2, WindML2.0.3 and Jworks4.0. When > I download the EVM to the VxWorks I get the following error: > -> ld < evm > Loading D:/Tornado/target/proj/evm / > Undefined symbols: > ___divdi3 > ___moddi3 > ___udivdi3 > > I have found out that this have something with variables of the type > long long, these are defined in libgcc.a and I need to link this into > my VxWorks image. So far I have been unable to do it. I have tried the > following from Tornado IDE: > In workspace->builds->default(vxworks image)->linker. > -X -N -Le:/tornado/host/x86-win32/lib/gcc-lib/i386-wrs-vxworks/cygnus-2.7 .2-960126/libgcc.a > When I compile the vxworks image goes through and the kernel is ok and > it starts up but I still have the error when I download the EVM. > > So how do I link my image with the libgcc.a to take away the undefined > symbols. From workspace as I did or by editing a makefile in BSP or > project? > > Regards > Andreas --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Trouble debugging 8265 with VisionProbe isync and sync instructions. Date: 16 Oct 2002 07:08:53 -0700 From: rick@cyclone.com (Richard Nardone) Organization: http://groups.google.com/ Message-ID: <29b153d9.0210160608.a49ecde@posting.google.com> References: > I am using EST/WindRivers VisionClick software patched to v7.9A. > We connect to our target using JTAG and VisionProbe II licensed > for 8260. > Frank, If you are running from Flash, make sure youre breakpoints are configured to be Internal Hardware (IHBC) and not Software. Check the configuration window setting for set breakpoint. --------------------------- Newsgroups: comp.os.vxworks Subject: Is there a 64-bit RawFS (ala DosFS 2.0)? Date: 16 Oct 2002 08:31:04 -0600 From: taashlo@sandia.gov Organization: Sandia National Laboratories, Albuquerque, NM USA Message-ID: <7kgipiaf.fsf@sandia.gov> Sender: taashlo@TYPHOON Hi All, For certain applications, the RawFS has been quite adequate. But the partition/file size limits have become a hinderance with some of the very large hard disks that I'm encountering now. I was wondering if there is now a version that supports large partitions/files the same way that DosFS 2.0 does? If so, could you please point me to some documentation so that I may RTFM? Thanks, Tad Ashlock --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Is there a 64-bit RawFS (ala DosFS 2.0)? Date: Wed, 16 Oct 2002 16:50:24 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <7kgipiaf.fsf@sandia.gov> Reply-To: "Leonid Rosenboim" wrote in message news:7kgipiaf.fsf@sandia.gov... > Hi All, > > For certain applications, the RawFS has been quite adequate. But > the partition/file size limits have become a hinderance with some > of the very large hard disks that I'm encountering now. I was > wondering if there is now a version that supports large > partitions/files the same way that DosFS 2.0 does? If so, could > you please point me to some documentation so that I may RTFM? > > Thanks, > > Tad Ashlock Yes, Tad, there is. During DosFs2.0 project development, rawFs.c has too been revamped, adding support of 64-bit integer arithmetics, the new 64-bit ioctl code support (mainly FIO_SEEK64) and support for CBIO so that it can work on top of dpartLib partition manager and/or dcacheLib - disk cache manager. At my time, this new rawFs was released as part of DosFs2.0 early access release. Unfirtunately, since I am not with WindRiver for a while, I can not reliably testify if the current dosFs2.0 product CD (or the version delivered with new versions of Tornado 2.2 and 3.0) include this new version of rawFs. If you - or someone else on this list - have a current dosFs2.0 installed, yo0u could probably identify the newer rawFs version by looking at rawFsLib.h include file or on-line manual. - -- HTH - ----------------------------------------------------------------------- Leonid Rosenboim Visit: http://gamla.org.il/english/index.htm Consultant Email: my first name at consultant dot com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: address translation on MPC750 / Page Table Segment Descriptor Date: Wed, 16 Oct 2002 11:16:14 -0400 From: Michael Reed Message-ID: <3DAD82BE.2AA971DE@ll.mit.edu> References: <3DAC7B2C.88ABB74B@ll.mit.edu> Thanks George, I was using VisionICE2 to try to single step the code, but I'm not completely comfortable w/ how (if) it's working properly. So I want to answer all of your questions using SingleStep debugger, but when I tried to attach in system mode (w/ wdbrpc backend), I got an error saying target did not support system mode debugging. Any tips on this? It's an MPC750 processor on a DY4 vme board. Thanks, Mike R. >Is data translation turned on in the msr register? If not, the MMU is not your problem. You need to provide more information about the state of your h/w and software. If you're debugging the interface between the 750 and the system controller, what does work? What part does vxWorks play in this problem? What are you single-stepping with? Where is the code you're stepping through, flash? Do you know that almost none of the MMU related registers are initialized following a POR? Michael Reed wrote: > > Hello, > I'm having trouble debugging the processor bus (between a MPC750 and a > Galileo memory controller). I've capatured data accesses on the bus w/ > a logic analyzer, but the addresses being accessed don't look like the > address I'm trying to access. > > My code was pretty simple: > > unsigned int *myPointer; > > myPointer = 0x700000; > > *myPointer = 0xFF; > > So when I single-step this code, I expected to see address 0x700000 > flash across the address bus, but I didn't. So I've dived into logical > to physical address translation, which took me to Page Tables and > segment descriptors. > > Question is then, if my address is 0x700000, how do I follow it thru the > page table and figure out what physical address will come out on the > processor bus? > > THanks, > Mike R. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Is there a 64-bit RawFS (ala DosFS 2.0)? Date: 16 Oct 2002 09:41:37 -0600 From: taashlo@sandia.gov Organization: Sandia National Laboratories, Albuquerque, NM USA Message-ID: References: <7kgipiaf.fsf@sandia.gov> Sender: taashlo@TYPHOON "Leonid Rosenboim" writes: > wrote in message news:7kgipiaf.fsf@sandia.gov... > > Hi All, > > > > For certain applications, the RawFS has been quite adequate. But > > the partition/file size limits have become a hinderance with some > > of the very large hard disks that I'm encountering now. I was > > wondering if there is now a version that supports large > > partitions/files the same way that DosFS 2.0 does? If so, could > > you please point me to some documentation so that I may RTFM? > > > > Thanks, > > > > Tad Ashlock > > Yes, Tad, there is. During DosFs2.0 project development, rawFs.c has too > been revamped, adding support of 64-bit integer arithmetics, the new 64-bit > ioctl code support (mainly FIO_SEEK64) and support for CBIO so that it can > work > on top of dpartLib partition manager and/or dcacheLib - disk cache manager. > > At my time, this new rawFs was released as part of DosFs2.0 early access > release. Unfirtunately, since I am not with WindRiver for a while, I can not > reliably testify if the current dosFs2.0 product CD (or the version > delivered > with new versions of Tornado 2.2 and 3.0) include this new version of rawFs. > > If you - or someone else on this list - have a current dosFs2.0 installed, > yo0u could probably identify the newer rawFs version by looking at > rawFsLib.h include file or on-line manual. > > -- HTH > ----------------------------------------------------------------------- > Leonid Rosenboim Visit: > http://gamla.org.il/english/index.htm > Consultant Email: my first name at consultant dot com Leonid, I have Tornado 2.1 for the PPC installed. This version comes with DosFS 2.0. I've looked at the on-line manual under "rawFsLib [DOSFS2]", but it didn't list FIOSEEK64 (nor FIO_SEEK64) among the IOCTL functions. However, the page for "dosFsLib [DOSFS2]" *did* include FIOSEEK64 among the IOCTL functions (along with several other 64-bitters). In the rawFsLib.h file, a #include for cbioLib.h has been added since Tornado 2.0. Plus it now has the following entry in the modification history: 02c,21jun00,rsh upgrade to dosFs 2.0 So based on this evidence, I'm going to assume that the documentation for rawFsLib is incomplete and 64-bit offsets are supported. Now I get to go read up on the details of all the problems associated with DosFS 2.0 on the PPC platform that I remember hearing about. (Something about the compiler using the fp processor registers to store 64-bit integers. Now these registers must be saved during a context switch?) Details, details. :-) :-) Anyway, thanks for your help and I'll report to this group how things went. Tad --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Is there a 64-bit RawFS (ala DosFS 2.0)? Date: Wed, 16 Oct 2002 18:01:03 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <7kgipiaf.fsf@sandia.gov> Reply-To: "Leonid Rosenboim" Ok, cool investigation. Don't know who "rsh" is, but I am curious, do you see any mention of "lrn" in the modification history below ? (that would be me ;-) ) You are absolutely brilliant to remember the 64-bit math issue of PPC which was discovered with connection to DsFs2.0 - since rawFs if it is really the nore code, will too use 64 bit math (albeit at a much primitive level) so it too may be subject to the same PPC compiler issue causing use of FP registers. One way to look into this would be to check if tha patch for 64-bit PPC issue of DosFs2 also includes a newer rawFsLib.o file !! Good luck ! Leonid --------------------------- Newsgroups: comp.os.vxworks Subject: Re: address translation on MPC750 / Page Table Segment Descriptor Date: Wed, 16 Oct 2002 12:05:13 -0400 From: Varndell Engineering Organization: Posted Via Newsfeeds.com = SPEED+RETENTION+COMPLETION = http://www.newsfeeds.com Message-ID: <3DAD8E39.A6604964@hotmail.com> References: <3DAC7B2C.88ABB74B@ll.mit.edu> <3DAD82BE.2AA971DE@ll.mit.edu> Michael Reed wrote: > > Thanks George, > I was using VisionICE2 to try to single step the code, but I'm not > completely comfortable w/ how (if) it's working properly. So I want to > answer all of your questions using SingleStep debugger, but when I tried > to attach in system mode (w/ wdbrpc backend), I got an error saying > target did not support system mode debugging. Any tips on this? It's Okay, I get it. Got a lot to learn, huh? If your network driver is not END, or it does not support polled mode operation, you can't attach for system mode debugging. If that's the case, you might be able to debug in system mode over the serial port. Again, the SIO driver must support polled mode, but most do. On the logic analyzer and bus debugging thing, maybe the store just went into cache. Good luck, and RTFM. Regards, George Varndell - -- Varndell Engineering, LLC Embedded Systems Expertise Need a vxWorks END driver that works? Contact us. http://www.varndellengineering.com - -----------== Posted via Newsfeed.Com - Uncensored Usenet News ==---------- http://www.newsfeed.com The #1 Newsgroup Service in the World! - -----= Over 100,000 Newsgroups - Unlimited Fast Downloads - 19 Servers =----- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Undefined symbols: ___divdi3 with JWorks 4 Date: 16 Oct 2002 12:25:18 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: References: <5947a04.0210160227.50cfdcb@posting.google.com> Sender: bpringle@DeadDuck > I have a problem that seems to be common here but so far I have been > unable to solve it. I am using T2.0.2, WindML2.0.3 and Jworks4.0. When > I download the EVM to the VxWorks I get the following error: > -> ld < evm > Loading D:/Tornado/target/proj/evm / > Undefined symbols: > ___divdi3 > ___moddi3 > ___udivdi3 Leonid> Here is another alternative method: When you add libgcc.a Leonid> into the link command, the linker will not use any of the Leonid> objects from the .a library unless they are required by one Leonid> of the .o objects mentioned previously, so your other Leonid> alternative may be to explicitly define these function names Leonid> in one of the init .c files, for example, you can add this Leonid> into usrConfig.c Leonid> #define INCLUDE_INT64_GCC_SUPPORT [snip] This is the usual solution. You can also play with linker options if you wish to save a little space (gcc_i64_funcs in Leonid's example). For example, "-u __divdi3 -u __moddi3 -u __udivdi3". Alternatively, a GNU assembler file can be code to use ".extern" and the same effect will be achieved. WindML and Zinc use a similar mechanism to provide the full tool suite in the vxWorks image. You may need a similar technique if you wish to include the full JWorks code in your kernel image (or it already has some EVMAll symbol). Another solution is to extract the module with "ar$(TOOL)" and add it to MODULE_EXTRA, etc to place the code where you wish. The array Leonid mentioned can be useful if it is placed at a know location for all vxWorks kernels you produce. Then modules/code can be produced that have no relocations, but call the vxWorks function through the table. This is advantageous if you do not wish to put all of the symbols in the image to provide target based "loadModule" functionality. fwiw, Bill Pringlemeir. - -- Do not become pregnant this month. Loving your cat now will turn out to be unhealthy after a few hours. Be sure to sleep before you find out that your long-lost brother is lost. You will meet a warmonger this week. You'll once in a while be fulfilled. Don't forget, it's only 366 days at most until Christmas. --------------------------- Newsgroups: comp.os.vxworks Subject: target shell Date: Wed, 16 Oct 2002 18:33:08 +0200 From: "francesco torre" Organization: Marconi, UK Message-ID: Hello, I am working with vxWorks on ppc860 (GNU compiler), using the target shell both via serial line and networked (telnet). I noted that calling a function by the shell with few parameters than the definition has, always forces the missing parameters to 0. This may be useful. My question: is this a documented feature? If yes, where? Thank in advance! --------------------------- Newsgroups: comp.os.vxworks,comp.object.corba Subject: Re: CORBA --> VisibrokerRT vs. ORBexpressRT Date: 16 Oct 2002 09:38:05 -0700 From: raycr@yahoo.com (Raymond Rawa) Organization: http://groups.google.com/ Message-ID: <6062b16c.0210160838.48df2b4e@posting.google.com> References: <3DA6DF0C.4050809@ele.etsmtl.ca> David Roberge wrote in message news:<3DA6DF0C.4050809@ele.etsmtl.ca>... > Does anyone had good or bad experiences with any of these product/company ? I've been using ORBexpressRT Ada on VxWorks/PPC for two years now and I love it. I only ever had one problem with it. The installer for my specific platform had a bug. Tech support has been great, they found and fixed the installer problem right away. The product itself has been rock solid. I've never tried their C++ version, but I've heard equally good things about that product too. - -Ray Rawa --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Is there a 64-bit RawFS (ala DosFS 2.0)? Date: 16 Oct 2002 11:49:03 -0700 From: danderson@kns.com (Don) Organization: http://groups.google.com/ Message-ID: References: <7kgipiaf.fsf@sandia.gov> "Leonid Rosenboim" wrote in message news:... > > Yes, Tad, there is. During DosFs2.0 project development, rawFs.c has too > been revamped, adding support of 64-bit integer arithmetics, the new 64-bit > ioctl code support (mainly FIO_SEEK64) and support for CBIO so that it can > work > on top of dpartLib partition manager and/or dcacheLib - disk cache manager. > > At my time, this new rawFs was released as part of DosFs2.0 early access > release. Unfirtunately, since I am not with WindRiver for a while, I can not > reliably testify if the current dosFs2.0 product CD (or the version > delivered > with new versions of Tornado 2.2 and 3.0) include this new version of rawFs. > > If you - or someone else on this list - have a current dosFs2.0 installed, > yo0u could probably identify the newer rawFs version by looking at > rawFsLib.h include file or on-line manual. Yep, it's there. Comments indicate the 64-bit ioctls and CBIO additions. This is for 5.4 on the x8 with dosFs2.0 > > -- HTH > ----------------------------------------------------------------------- > Leonid Rosenboim Visit: > http://gamla.org.il/english/index.htm > Consultant Email: my first name at consultant dot com HTH, Don --------------------------- Newsgroups: comp.os.vxworks Subject: Telnet and CLI stuck Date: 16 Oct 2002 11:53:46 -0700 From: gabi@tdsoft.com (Gabi) Organization: http://groups.google.com/ Message-ID: <5607ff2c.0210161053.ed2f946@posting.google.com> Hi, I'm working on the MPC8260 under vxworks, with CLI(tShell) telent and FTP server. Some of the product in the field return with the following error: The CLI or Telnet not replay but the target still response to ping. any suggestion? tanks Gabi --------------------------- Newsgroups: comp.os.vxworks,comp.arch.embedded Subject: Re: Boot Code Integrity check with CRC Date: 16 Oct 2002 11:58:04 -0700 From: ryeung@earthlink.net (Raymond Yeung) Organization: http://groups.google.com/ Message-ID: References: Thanks for all the replies. Having read all of them, I've the following thoughts: 1. S-record is a wrong form to use. Initially, I explored what format to use; VxWorks generates elf, hex (S-record). But I'm not sure I found the binary version that I can load directly to flash and have the processor boot. That's why I thought about using S-record, and write a loader on target to interpret all S-records. 2. Incidentally, has anyone done the following below: Use the bin file from visionCLICK Convert utility. Strip off the 32-bytes header, and load the remaining data section directly to flash for boot up purpose, starting at reset vector location. 3. Checksum instead of CRC may be used for boot code integrity. Assuming I'm dealing with binary, there'd be no "holes" as in S-record. Thus, I can compute the checksum from start to end of the entire binary. 4. On the host, I can put checksum at end of the boot code binary file. But once I "load" it onto flash, would I put this checksum at a fixed known location (e.g. last 32-bit of an erase sector)? Or would I create a global variable in boot code to store that? I suppose I can't just stick it at the "end" as the code doesn't know where the location of this end is, does it? In the case of using global, how do I commuicate with the host tool to initialize it to the proper checksum when I build the file? Or am I off-track here? 5. What is PPCBoot? I try searching for it on Googles, and got too many returns. Would appreciate it if one can provide a link to its tutorial? (Note: I'm currently booting using VxWorks boot rom. Is PPCBoot another boot rom?) 6. To evaluate usefulness of this checksum thing (to tell people why I even bother to go through all these), I've the following points: + flash could be defective: certain cells may have faults (e.g. stuck-at). + Boot image may first have to be transferred to RAM before writing to flash (buffer for FTP). Similar memory related problem could happen here in RAM. Furthermore, a strayed thread could overwrite its content before it's written to flash. + Read back could have problems (either address/data/flash-device) issue. + Verifying checksum right after file transfer ensures the file is not screwed up due to ASCII transfer (as David pointed out). But this would not be applicable if I program boot code only with Emulator. Did I miss anything? Thanks, Raymond --------------------------- Newsgroups: comp.os.vxworks Subject: Re: pciProbing question.. Date: 16 Oct 2002 12:44:58 -0700 From: rick@cyclone.com (Richard Nardone) Organization: http://groups.google.com/ Message-ID: <29b153d9.0210161144.e9ead4b@posting.google.com> References: <1034157627.783214@newsreader2.wirehub.nl> "Renee Teunissen" wrote in message news:<1034157627.783214@newsreader2.wirehub.nl>... > sometime I see (in sysLib.c, intel/cPCI based system) > ---------------------- > pciConfigInLong (pciBus, pciDevice, pciFunc, PCI_CFG_BASE_ADDRESS_1, > &basePtr); > > /* get size for MMU entry */ > /* Write all f's and read back value */ > pciConfigOutLong (pciBus, pciDevice, pciFunc, PCI_CFG_BASE_ADDRESS_1, > 0xffffffff); > pciConfigInLong (pciBus, pciDevice, pciFunc, PCI_CFG_BASE_ADDRESS_1, > ®Size); > > regSize= ( ~( regSize & 0xFFFFFFF0 ) ) + 1; > > /* restore specified base address */ > pciConfigOutLong (pciBus, pciDevice, pciFunc, PCI_CFG_BASE_ADDRESS_1, > basePtr); > ---------------------- > > Looks nice, but.. I f do this and compare it to he specs of the > device's vendor, I see a complete different regSize. Is the above > sample apllicable for all (compact)PCI devices or is this "just" for > the device this code was meant for? > Renee, This is pretty standard PCI bus configuation cycle activity. The value returned when the base address register is read the second time is an indication of the memory block size requested by the device. It is not the direct value of the memory size requested. The binary weighted value of the first one bit found indicates the required amount of space. If regSize = 0xFFF00000, the device is requesting 1MB of space. HTH Rick Nardone Cyclone Microsystems New Haven, CT 06511 --------------------------- Newsgroups: comp.os.vxworks Subject: dlmalloc.2.7.2 corruption?? Date: 16 Oct 2002 13:19:11 -0700 From: ndemario1@excite.com (Neil) Organization: http://groups.google.com/ Message-ID: <8c117101.0210161219.700e1cc6@posting.google.com> Hi Group, I have recently added dlmalloc to vxWorks 5.4.2 PPC with what seems to be terrific results BUT. The BUT comes when I run tcpstatShow from the target shell, part of the printouts are corrupted as follows (it varies, this is the worst I've seen, allows at the beginning of the printout): 2 control packetÿÿÿÿÿÿÿÿ 42282 packetÿÿÿÿÿÿÿÿ received 39359 ackÿÿÿÿÿÿÿÿ (for 2043266 byteÿÿÿÿÿÿÿÿ) Dlmalloc has done wonders for my memory fragmentation when running the ACE ORB (Tao). Everything else, ftp, applications, tcp sockets, etc, seem to run fine except this one item. If I return to my original vxWorks image then the tcpstatShow corruption goes away. My implementation of dlmalloc is somewhat of a hybrid. I removed memLib.o from my lib, changed the symbols "malloc and free" to "malloX and freX" in memPartLib.o. I've written a sbrk() routine that uses the USER_RESERVED_MEM area. I've also have a version that memPartAlignedAlloc's pages of system heap (8 byte address aligned) which limits vxWork's memory allocator to issuing 4096 or larger size blocks. It does this very well, without fragmentation, by the way. Both cases use no trim, non-contiguous, and semaphore locks/wrappers. Suggestions on how to debug this problem and or comments about what to do would be a great help! Neil DeMario Embedded Software Engineer Axcelis Technologies, Inc. Beverly, MA, 01915 Neil.Demario@Axcelis.com --------------------------- Newsgroups: comp.os.vxworks Subject: Porting VxWorks Date: Wed, 16 Oct 2002 20:42:01 GMT From: "JM" Organization: Road Runner - Texas Message-ID: Hello, I am new to VxWorks but need to know if the MPC8560 is supported or if it is possible to port VxWorks V5.X to that platform. If it is possible to port VxWorks, what has to be done? Jeff Mokhtar --------------------------- Newsgroups: comp.os.vxworks Subject: Re: PCI devices not on page boundary Date: 16 Oct 2002 13:47:30 -0700 From: rick@cyclone.com (Richard Nardone) Organization: http://groups.google.com/ Message-ID: <29b153d9.0210161247.29be9cb4@posting.google.com> References: <1034588068.18418@newsreader1.wirehub.nl> > While probing for PCI devices on Intel based (compact) PCI system, I > got BAR1 or BAR2 addresses which are not on a page boundary, so > usrMmuInit fails on these entries. For example I get address > 0xe820:0400 for BAR0, which needs 256 bytes mem space, and this > address is not on a page boundary. I'm not sure of the Host architecture, but I usually statically set up regions which cover PCI address space in my host CPU address space and make adjustments to the MMU before probing the bus. Conversely, you could scan the bus, group the requested space in to regions with appropriate attributes, and then set up the MMU dynamically to handle the requests. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: dlmalloc.2.7.2 corruption?? Date: 16 Oct 2002 18:49:50 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: References: <8c117101.0210161219.700e1cc6@posting.google.com> Sender: bpringle@DeadDuck Neil> I have recently added dlmalloc to vxWorks 5.4.2 PPC with what Neil> seems to be terrific results BUT. The BUT comes when I run Neil> tcpstatShow from the target shell, part of the printouts are Neil> corrupted as follows (it varies, this is the worst I've seen, Neil> allows at the beginning of the printout): [snip] Neil> Suggestions on how to debug this problem and or comments about Neil> what to do would be a great help! It is very nice to have a sane allocator isn't it? The problem might actually be in the tcpstatShow() code. I have had reboot() fail due to the network shutdown sequence. Basically, there is code like this, p = net_devs[0]; while(p) { shutdown(p); free(p); p = p->next; } The `p' pointer is freed and looking at next is not allowed. It might be that the tpcstatShow() code is dependant on a poor allocator that doesn't free memory quite so fast. If you have WindSurf access, file a TSR. You can use objdumpppc to look at the tcpstatShow assembler and try to deduce a problem. When I gave extended details of the above problem, the support engineer gave a patch to me. However, the problem might not be with dlmalloc. fwiw, Bill Pringlemeir. - -- You will be in conflict with Sagitarius since Mercury has moved into Pisces. Don't let your teddy bear take a bath with your rhinocerous. Someone you know will be a valuable enemy. A lout will think of you as your mother. You'll never be a Borg. --------------------------- Newsgroups: comp.os.vxworks Subject: spr #23225 Date: 16 Oct 2002 16:04:30 -0700 From: tjaimin@yahoo.com (Jai) Organization: http://groups.google.com/ Message-ID: <17902e07.0210161504.421c9613@posting.google.com> Hi, Wind River had a patch (23225) for OpenSSL on Wind Web Server a while back. They've stopped distributing this path since they do not support WWS1.0 anymore. Has anyone got this patch? Is it possible to get this patch from any other place? Thanks, Jai. --------------------------- Newsgroups: comp.os.vxworks,comp.arch.embedded Subject: Re: Boot Code Integrity check with CRC Date: Wed, 16 Oct 2002 23:52:52 +0100 From: nospam Organization: http://extra.newsguy.com Message-ID: References: ryeung@earthlink.net (Raymond Yeung) wrote: >4. On the host, I can put checksum at end of the boot code binary > file. But once I "load" it onto flash, would I put this > checksum at a fixed known location (e.g. last 32-bit of an > erase sector)? Or would I create a global variable in boot code > to store that? If you calculate a checksum for the entire boot block (assuming it begins in an erased state) and insert somewhere into that block the inverse of the checksum. The system then checksums the entire boot block and will always get the same value. You don't need to know what or where the checksum is, you arrange for the block to always have the same checksum. To verify the contents in RAM prior to FLASH programming you can still pretend to check the whole block or why not use a block sized amount of RAM, one day the code may occupy the whole block so you probably have to be able to allocate that amount of RAM anyway. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Trouble debugging 8265 with VisionProbe isync and sync instructions. Date: 16 Oct 2002 16:46:26 -0700 From: FrankGleeson@motorola.com (frankg) Organization: http://groups.google.com/ Message-ID: References: <29b153d9.0210160608.a49ecde@posting.google.com> Yes, I know that the jtag can not set a software breakpoint in flash. I can only use the single hardware breakpoint. I have since started debugging the c rountines in bootInit.c. I have not seen the problem outside of romInit.s. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Starter pack ??? Date: Wed, 16 Oct 2002 20:17:33 -0700 From: "A. Kaul" Organization: National Semiconductor, Santa Clara Message-ID: <3DAE2BCD.4060100@antispam.com> References: <3DA791FF.8080509@antispam.com> <488e459a.0210121101.377ecb05@posting.google.com> <3DACB05B.3060800@antispam.com> Thanks for the info! Couple of more specific queries - Is there a sytem routine that a driver could call to launch a timer? How much of a delay is sysDelay() supposed to provide? Regards. - -ak --------------------------- Newsgroups: comp.os.vxworks,comp.arch.embedded Subject: Re: Boot Code Integrity check with CRC Date: Wed, 16 Oct 2002 17:28:06 -0700 From: Rajesh Tripathi Organization: Globix Corp. Message-ID: <3DAE0416.926F5812@com21.com> References: Hi Raymond, Raymond Yeung wrote: > Thanks for all the replies. Having read all of them, I've the > following thoughts: > > 1. S-record is a wrong form to use. Initially, I explored what > format to use; VxWorks generates elf, hex (S-record). But I'm > not sure I found the binary version that I can load directly to > flash and have the processor boot. That's why I thought about > using S-record, and write a loader on target to interpret all > S-records. I used elfToBin utility for generating binary file form elf file. That was two years ago on Tornado 2.0 for PPC and I don't find that file on my T2.1 for MIPS installation now. I believe the recommended way now is to use "objcopyppc -O binary --binary-without-bss xyz.elf xyz.bin" Which version of Tornado you are using? > > 2. Incidentally, has anyone done the following below: Use the bin > file from visionCLICK Convert utility. Strip off the 32-bytes > header, and load the remaining data section directly to flash > for boot up purpose, starting at reset vector location. It Should work although I never tried. Now I'm using our own utility to extract the text and data sections from elf file and prepend with a header. A bootrom update utility on target loads the file via tftp and burns the flash. You can even embed the bootrom image it in application image to be automatically or manually updated in field. > > 3. Checksum instead of CRC may be used for boot code integrity. > Assuming I'm dealing with binary, there'd be no "holes" as in > S-record. Thus, I can compute the checksum from start to end > of the entire binary. Checksum is fine. > > 4. On the host, I can put checksum at end of the boot code binary > file. But once I "load" it onto flash, would I put this > checksum at a fixed known location (e.g. last 32-bit of an > erase sector)? Or would I create a global variable in boot code > to store that? I suppose I can't just stick it at the "end" as > the code doesn't know where the location of this end is, does it? > In the case of using global, how do I commuicate with the > host tool to initialize it to the proper checksum when I build > the file? Or am I off-track here? It works fine both ways. On PPC declare a global and initialize it with a Macro say CKSUM defined as zero in Makefile. After linking is done, extract the bin file, calculate checksum and save the compliment of checksum in a temporary file tchksum. Now do a second link and pass it at link time as $(LD) $(LDFLAGS) -defsym CKSUM=`cat tchksum` -e $(ROM_ENTRY) The global variable now has compliment of checksum and if you calculate the checksum during POST, it should always be zero. You've to play with Makefile a little but it works fine. The other way is to have a utility calculate checksum and append it to binary file at the end. The bootcode calculates checksum from start upto _edata and compares it to the stored checksum at _edata. The data section is normally linked at ram address so calculate it's size by _edata - _fdata and add it to _etext to find the last address of image in flash. > > 5. What is PPCBoot? I try searching for it on Googles, and got > too many returns. Would appreciate it if one can provide a > link to its tutorial? (Note: I'm currently booting using VxWorks > boot rom. Is PPCBoot another boot rom?) This is a great bootloder for PowerPc . Documentation should be available at http://ppcboot.sourceforge.net > > 6. To evaluate usefulness of this checksum thing (to tell people why > I even bother to go through all these), I've the following points: > + flash could be defective: certain cells may have faults (e.g. > stuck-at). > + Boot image may first have to be transferred to RAM before > writing to flash (buffer for FTP). Similar memory related > problem could happen here in RAM. Furthermore, a strayed thread > could overwrite its content before it's written to flash. > + Read back could have problems (either address/data/flash-device) > issue. > + Verifying checksum right after file transfer ensures the file > is not screwed up due to ASCII transfer (as David pointed out). > But this would not be applicable if I program boot code only > with Emulator. > Did I miss anything? > > Thanks, > Raymond Regards, Rajesh --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: 16 Oct 2002 18:22:06 -0700 From: brian_doe@hotmail.com (brian) Organization: http://groups.google.com/ Message-ID: References: <3DAAA37D.EAE13FD0@alcatel-ke.de> <3DAD2568.16C701E8@alcatel-ke.de> Tim Sinkins wrote in message news:<3DAD2568.16C701E8@alcatel-ke.de>... > brian wrote: > > > > > I think the basic assumption made about the worth of switching is > > > that the user has the right deal in the first place. > > > Anyone who is producing a product by the millions (or even 100,000s) > > > who pays a royalty, has struck a dumb deal and they would perhaps be > > > better off renegotiating with WRS on the threat of moving OS. > > > > Possibly a dumb deal. > > In the current market definately a dumb deal. > > (current as in "with the number of alternatives available" > rather than "because the economy's shit at the moment"). > > I recall working on a product 10 years ago where they had decided to > use pSoS(+). The offer was a royalty and the company said, "nope > offer us a royalty free option or we'll go somewhere else". A 5 > figure license fee was negotiated. I think your knowledge is limited. ISI was notorious for doing buyouts. Part of the reason that WRS bought them. You wont get the same buyout from WRS. > > > But I can recount many companies such as > > Philips, Sony (many products). to name two, that pay royalties on > > products that ship in 6-7 figures. > > New products or legacy stuff. Do a little research. The Sony Walkman NWE3 for one. > > > Maybe Sony is dumb as well. Funny, I never thought they were. > > You make your own mind up! I have. > that we might decide to do! Not even bothering to negotiate > is really dumb. ya think? Brian --------------------------- Newsgroups: comp.os.vxworks Subject: Bootrom question Date: 16 Oct 2002 17:59:48 -0700 From: snower_tiger@hotmail.com (snower) Organization: http://groups.google.com/ Message-ID: <3f60ed07.0210161659.10efdd12@posting.google.com> I used tornado to develop the project. but I have a question, if I build bootrom with uncompressed hex, it work well, but if I build bootrom with compressed hex, it cant work, it will call excintconnect funtion, and did not come back. Is there anyone know why, anybody can suggest any idel? my email: snower_tiger@hotmail.com --------------------------- Newsgroups: comp.os.vxworks Subject: vxworks VLAN implementation Date: Thu, 17 Oct 2002 12:24:59 +0800 From: "Nison.liu" Organization: Bentium Ltd. (CN99) Message-ID: All: Where can I find VLAN implementation information under vxWorks? Thanks! Nison --------------------------- Newsgroups: comp.os.vxworks Subject: Re: spr #23225 Date: 17 Oct 2002 00:55:09 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: References: <17902e07.0210161504.421c9613@posting.google.com> Sender: bpringle@DeadDuck Jai> Wind River had a patch (23225) for OpenSSL on Wind Web Server Jai> a while back. They've stopped distributing this path since they Jai> do not support WWS1.0 anymore. Has anyone got this patch? Is it Jai> possible to get this patch from any other place? Bzzzt. The patch is for SSLeay, not OpenSSL. The current code in CVS for OpenSSL (v 0.9.8 I think) has been set up to configure to build OpenSSL on the Solaris platform for a vxWorks target. You can also look in the vxWorks FAQ and search the OpenSSL mailing list for some more info on vxWorks and OpenSSL. It is probably less effort to get 0.9.6g working that it is chasing WRS patches. Here's an idea, use google. hth, Bill Pringlemeir. - -- BOMBED BY IRC SERVER. UNABLE TO SURFACE. SINKING. U-181. --------------------------- Newsgroups: comp.os.vxworks Subject: USB Date: 16 Oct 2002 22:28:36 -0700 From: ravi368@yahoo.com (Ravi Uday) Organization: http://groups.google.com/ Message-ID: Hi, Can anybody tell me which is the best discussion site for discussing issues regarding writing driver for USB on VxWorks O.S. I already have the VxWorks USB DDK with me and have few questions of using / begin with it. My device is DiskOnKey. Thanks for all kind of helps/experience - - Ravi --------------------------- Newsgroups: comp.os.vxworks Subject: strange timer bahaviour Date: Thu, 17 Oct 2002 09:17:42 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1034839098.567517@newsreader1.wirehub.nl> Hi, I'm trying to understand serveral timer_gettime and timer_settime related issues we have. We have an application which has to send-out triggers at specified intervals. To do this we arm a timer and hook it to our services routine, set the intervals and things go OK. But if we change the period-time, strange things happen. When we change the period time, we do not only change the it_interval parameter but also - if possible - the it_value parameter, to make the current period fit into the new period-time. So we modify the remaining time as set with it_value of the itimerspec struct to make it fit. If I do that, it seems that the new periode time (it_interval) gets the same value as the "time remaining" (it_value) and thus the timer is running much to fast. does this make a bell ring by someone? is it allow - on a running timer - to modify the it_value ? do I forget something? Another strange thing. If I call timer_gettime, I do get (sometimes) it_value.tv_sec with negative values. Is this normal bahaviour? Thanx for far. Renee. --------------------------- Newsgroups: comp.os.vxworks Subject: Question about aqrchive of objects Date: Thu, 17 Oct 2002 15:35:24 +0800 From: "Yung" Organization: IMS Netvigator Message-ID: Hi experts, under the "Builds" specifications of the Tornado workspace, if I add an archive files (e.g."xxx.a") under the macro "LIBS", does the system automatically link up all the objects contained in the archive into the final image? Or does it only pick those that are called by others? Thanks. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: dlmalloc.2.7.2 corruption?? Date: Thu, 17 Oct 2002 09:57:03 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <8c117101.0210161219.700e1cc6@posting.google.com> Reply-To: "Leonid Rosenboim" I probably can't really help you fix this right away, but I will chip in with what I know, and maybe that will contribute to the progress: The garbage you see comes from the rather silly peice of code that VxWorks has borrowed from BSD's nettstat program. The place where garbage comes, is supposed to contain either an "s" character if the couter is larger then 1, or an empty string if the counter is exactly 1. You see why this is silly, it is meant to print "1 packet" and "500 packets". Anyway, where there is supposed to be an empty string, there is a printf using something like "%d packet%s" and then the first argument is the counter, and the second is a macro which checks the counter and resolves to a string which is either "s" or "". This as you can see has NOTHING TO DO WITH heap routines, and you can verify this rather easily by disassembling the code, or by looking at the source of "netstat" untility on any Unix open source (they all borrowed from early BSD code). Now, if you really want me to make a guess, then I'd bet 10 bucks that your current issue is stack corruption, printf() requires a rather big chunk of stack space for formatting a string, so your stack could overflow because the fancy heap code requires more stack than the regular one, and would bet yet another 10 bucks that the fancy heap code somehow messes with the printf arguments, thereby doubling my chanses. - -- HTH - ----------------------------------------------------------------------- Leonid Rosenboim Visit: http://gamla.org.il/english/index.htm Consultant Email: my first name at consultant dot com "Neil" wrote in message news:8c117101.0210161219.700e1cc6@posting.google.com... > Hi Group, > I have recently added dlmalloc to vxWorks 5.4.2 PPC with what seems to > be terrific results BUT. The BUT comes when I run tcpstatShow from the > target shell, part of the printouts are corrupted as follows (it > varies, this is the worst I've seen, allows at the beginning of the > printout): > > 2 control packetÿÿÿÿÿÿÿÿ > 42282 packetÿÿÿÿÿÿÿÿ received > 39359 ackÿÿÿÿÿÿÿÿ (for 2043266 byteÿÿÿÿÿÿÿÿ) > > Dlmalloc has done wonders for my memory fragmentation when running the > ACE ORB (Tao). Everything else, ftp, applications, tcp sockets, etc, > seem to run fine except this one item. If I return to my original > vxWorks image then the tcpstatShow corruption goes away. > > My implementation of dlmalloc is somewhat of a hybrid. I removed > memLib.o from my lib, changed the symbols "malloc and free" to "malloX > and freX" in memPartLib.o. I've written a sbrk() routine that uses the > USER_RESERVED_MEM area. I've also have a version that > memPartAlignedAlloc's pages of system heap (8 byte address aligned) > which limits vxWork's memory allocator to issuing 4096 or larger size > blocks. It does this very well, without fragmentation, by the way. > Both cases use no trim, non-contiguous, and semaphore locks/wrappers. > Suggestions on how to debug this problem and or comments about what to > do would be a great help! > > Neil DeMario > Embedded Software Engineer > Axcelis Technologies, Inc. > Beverly, MA, 01915 > Neil.Demario@Axcelis.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting VxWorks Date: Thu, 17 Oct 2002 08:18:11 +0000 (UTC) From: Johan Borkhuis Organization: Proxim Corp. Message-ID: References: "JM" wrote: > I am new to VxWorks but need to know if the MPC8560 > is supported or if it is possible to port VxWorks V5.X to that > platform. If it is possible to port VxWorks, what has to be done? If you need this kind of information it is better to contact WindRiver directly. They can tell you which platform are supported and also if they have a BSP for a development platform. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Question about aqrchive of objects Date: Thu, 17 Oct 2002 08:22:13 +0000 (UTC) From: Johan Borkhuis Organization: Proxim Corp. Message-ID: References: "Yung" wrote: > Hi experts, > > under the "Builds" specifications of the Tornado workspace, if I add an > archive files (e.g."xxx.a") under the macro "LIBS", does the system > automatically link up all the objects contained in the archive into the > final image? Or does it only pick those that are called by others? This is a basic linker question, so you better check the linker manual or try to find some information on linking in general. But to answer your question: only the modules that are referenced are included in you image. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Bootrom question Date: 17 Oct 2002 02:43:55 -0700 From: be_gyus@yahoo.com (james) Organization: http://groups.google.com/ Message-ID: References: <3f60ed07.0210161659.10efdd12@posting.google.com> snower_tiger@hotmail.com (snower) wrote in message news:<3f60ed07.0210161659.10efdd12@posting.google.com>... > I used tornado to develop the project. but I have a question, if I > build bootrom with uncompressed hex, it work well, but if I build > bootrom with compressed hex, it cant work, it will call excintconnect > funtion, and did not come back. Is there anyone know why, anybody can > suggest any idel? > my email: snower_tiger@hotmail.com Which is the Tornado version u r using, the bsp and whats the error u r gettting while building? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Thu, 17 Oct 2002 12:41:32 +0200 From: Tim Sinkins Organization: Alcatel Kommunikations Elektronik Message-ID: <3DAE93DC.21E25B1D@alcatel-ke.de> References: <3DAAA37D.EAE13FD0@alcatel-ke.de> <3DAD2568.16C701E8@alcatel-ke.de> Reply-To: timothy.sinkins@alcatel-ke.de brian wrote: > > Tim Sinkins wrote in message news:<3DAD2568.16C701E8@alcatel-ke.de>... > > brian wrote: > > > > > I think your knowledge is limited. My knowledge is sufficient. In order to get a product license from an OS vendor you have to find some other supplier that will sell their product this way as leverage for those that won't. The only knowldge I need, is that there are companies with comparable RTOS to WRS that will offer product licenses. I know that this is the case. In the last 10 years none of the companies I have worked for (I freelance, so that's about 10 companies) has taken an RTOS on a royalty fee basis. WRS cannot sail against this tide and must offer a similar deal even if they only offer a gold plated deal. > ISI was notorious for doing > buyouts. Part of the reason that WRS bought them. You wont get the > same buyout from WRS. I don't recall saying you would. I recall saying that they would do a product license (given enough leverage). > > > But I can recount many companies such as > > > Philips, Sony (many products). to name two, that pay royalties on > > > products that ship in 6-7 figures. > > > > New products or legacy stuff. > > Do a little research. The Sony Walkman NWE3 for one. How can I do this. Which companies are going to tell me *honestly* that (users) they (users) pay huge roylties to an RTOS vendor or that they (suppliers) receive large reyalties from specific customer(s). This is comercially sensitive information that ought not to be given to anybody, it is potentially stock price sensitive info. > > > Maybe Sony is dumb as well. Funny, I never thought they were. > > > > You make your own mind up! > > I have. Um, IMO WRS offer nothing that cannot be bought elsewhere. Choosing to use VXWorks for a multi-million unit product range at 25c a pop when you can get a comparable roylty free option from elsewhere for 50K is stupid. OK, WRS may offer a complete solution: OS, TCP/IP, IDE etc that you will have to buy in separately and it may be worth paying 100K for an integrated package but to agree a deal that costs you 1M is stupid IMO. > > that we might decide to do! Not even bothering to negotiate > > is really dumb. > > ya think? Yep. Tim > > Brian --------------------------- End of New-News digest ********************** From vxwexplo-errs@csg.lbl.gov Fri Oct 18 04:03:35 2002 From: Vxworks Exploder Date: Fri Oct 18 04:03:37 PDT 2002 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Fri Oct 18 04:03:32 PDT 2002 Subject: Re: dlmalloc.2.7.2 corruption?? Subject: Re: USB Subject: Calling an object method in a protection domain from a shared library Subject: Boot device Subject: Anyone already ported from Tornado 2.0 to Tornado 2.2 ? Subject: sendto in vxworks Subject: Re: sendto in vxworks Subject: Re: target shell Subject: Re: strange timer bahaviour Subject: Driving Three Monitors from a single PC Subject: Tornado Registry problem Subject: Re: Driving Three Monitors from a single PC Subject: Re: Boot Code Integrity check with CRC Subject: Telnet Time-Out Subject: Target Shell Subject: Re: sendto in vxworks Subject: VxWorks dosFs 1.0 chkdsk Utility Subject: Re: VxWorks dosFs 1.0 chkdsk Utility Subject: build problem Subject: Re: vxworks VLAN implementation Subject: Re: Porting of vxWorks. Subject: dosFs 1.0 Integrity Information Needed Subject: How to force asynchronous read() to finish in time? Subject: Re: build problem Subject: Re: Boot Code Integrity check with CRC Subject: Re: spr #23225 Subject: Re: Boot Code Integrity check with CRC Subject: Re: Bootrom question Subject: how to re-size of TFFS? Subject: Re: spr #23225 Subject: Re: VxWorks dosFs 1.0 chkdsk Utility Subject: Re: dosFs 1.0 Integrity Information Needed Subject: Re: Calling an object method in a protection domain from a shared library Subject: Re: Boot device Subject: a question about vxVMI and 3C905B Subject: Re: Anyone already ported from Tornado 2.0 to Tornado 2.2 ? Subject: Re: VxWorks dosFs 1.0 chkdsk Utility Subject: Re: Porting of vxWorks. Subject: Re: How to force asynchronous read() to finish in time? Subject: Where are the shared library created C++ objects located Subject: Re: Boot device Subject: Re: VxWorks dosFs 1.0 chkdsk Utility Subject: Re: VxWorks dosFs 1.0 chkdsk Utility ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: dlmalloc.2.7.2 corruption?? Date: 17 Oct 2002 05:38:55 -0700 From: ndemario1@excite.com (Neil) Organization: http://groups.google.com/ Message-ID: <8c117101.0210170438.42244bbb@posting.google.com> References: <8c117101.0210161219.700e1cc6@posting.google.com> Others here at work agree with you that we maybe seeing a pre-existing problem that is covered up by the vxworks memory allocator. Thanks for the suggestion, I'll take a look. My testing has shown that vxWorks can take up to 6 seconds to find memory when the free list is 8000 entries long. Dlmalloc handles this same test in a few milli seconds!!! > > It is very nice to have a sane allocator isn't it? > > The problem might actually be in the tcpstatShow() code. I have had > reboot() fail due to the network shutdown sequence. Basically, there > is code like this, > > p = net_devs[0]; > while(p) > { > shutdown(p); > free(p); > p = p->next; > } > > The `p' pointer is freed and looking at next is not allowed. It might > be that the tpcstatShow() code is dependant on a poor allocator that > doesn't free memory quite so fast. > > If you have WindSurf access, file a TSR. You can use objdumpppc to > look at the tcpstatShow assembler and try to deduce a problem. When I > gave extended details of the above problem, the support engineer gave > a patch to me. > > However, the problem might not be with dlmalloc. > > fwiw, > Bill Pringlemeir. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: USB Date: Thu, 17 Oct 2002 14:58:07 +0200 From: "Khac Tri VU" Organization: Planet Internet Message-ID: References: You should try the official site of USB at www.usb.org. It will lead you to diferent discussion mailing list. Tri "Ravi Uday" wrote in message news:ecca3ca6.0210162128.c36f98c@posting.google.com... > Hi, > > Can anybody tell me which is the best discussion site for discussing > issues regarding writing driver for USB on VxWorks O.S. I already have > the VxWorks USB DDK with me and have few questions of using / begin > with it. > My device is DiskOnKey. > Thanks for all kind of helps/experience > > - Ravi --------------------------- Newsgroups: comp.os.vxworks Subject: Calling an object method in a protection domain from a shared library Date: 17 Oct 2002 06:45:42 -0700 From: steven4443@yahoo.com (stephen hallacy) Organization: http://groups.google.com/ Message-ID: Hello, I'm waiting for my company to get VXWorks and Tornado. In the meantime I'm trying to work out the architecture of a system. My questions: Is it possible to call an object member in a protection domain from a shared library? I can only think of creating a function in the shared library that accepts a pointer to the object member. This pointer could be made global to the shared library. Is this the best way to do it? Also, I have shared library functions that create objects. Will these objects reside in the protection domain (that attached to the shared library) or the shared library domain. I'm thinking it's the protection domain. Am I correct? Thanks Stephen --------------------------- Newsgroups: comp.os.vxworks Subject: Boot device Date: 17 Oct 2002 15:00:35 +0100 From: Ole Asbjørn Fadum Message-ID: I'm using a x86 board for my project. The pentium boards doesn't have any NVRAM, but some nice guys has provided the files for store the bootline on disk instead (I.e. nvRamToAta.c). Sometimes my system is booted from IDE disk, and sometimes from floppy. I want the bootparameters to reside in the bootdevice, and not as it is today as the floppy boot takes the IDE bootparameters. How could I find out which device was cause in the current boot??? --------------------------- Newsgroups: comp.os.vxworks Subject: Anyone already ported from Tornado 2.0 to Tornado 2.2 ? Date: Thu, 17 Oct 2002 10:05:28 -0400 From: "steeve gagne" Organization: Newshosting.com - Highest quality at a great price! www.newshosting.com Message-ID: <3daec4b6@news.mhogaming.com> Is There anyone using Tornado 2.2 ? Does WindRiver give you a new BSP ? becaus I am stuck with Tornado 2.2 Wihout any BSPS So, I have started to port my project, I successfully compiled my run-time image System memory seems to not be configured correctly I have tried to boot it but here is the results : Starting run-time image... SysMemTop :: Total Memory 16777216 1000000 Memory status bytes blocks avg block max block ------ ---------- --------- ---------- ---------- current free 9289344 14 663524 9216264 alloc 977096 109 8964 - cumulative alloc 977208 110 8883 - dosFsFmtLib: FAT16 indicated by BPB FSTYPE string, cluster calculation was FAT12. Returning string. WARNING: FAT16 indicated by BPB FSTYPE string, cluster calculation was FAT12. Honoring string. memPartAlloc: block too big - 27256 in partition 0x630a30. memPartAlloc: block too big - 98824 in partition 0x630a30. Steeve Gagne steeve.gagne@videotron.ca Thanks For Your Help --------------------------- Newsgroups: comp.os.vxworks Subject: sendto in vxworks Date: Thu, 17 Oct 2002 22:05:08 +0800 From: "chenxt" Organization: Bentium Ltd. (CN99) Message-ID: Hi all, Two hosts comminute with tcp protocol in vxworks, but when the peer host dose not recevie the socket, I will block the task on the 'sendto' routine for a long time. why? thanks! --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sendto in vxworks Date: Thu, 17 Oct 2002 16:16:03 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Reply-To: "Leonid Rosenboim" "chenxt" wrote in message news:aomg2l$k5c$1@mail.cn99.com... > Hi all, > > Two hosts comminute with tcp protocol in vxworks, but when the peer host > dose not recevie the socket, I will block the task on the 'sendto' routine > for a long time. why? > > thanks! Very simply, TCP is a reliable protocol, and also handles flow control, and what you see is a classic example of flow control - if the receiver is busy and not collecting any data off the socket, the sender must stop sending after the buffers at the sender and receiver sockets are at the high limit. You can switch the sender's socket into non-blocking mode and that will cause the sender to return the number of bytes actually written to sender socket buffer instead of blocking, but functionally, if the receiver does not retreive its data you can either wait or abort the connection. - -- - ----------------------------------------------------------------------- Leonid Rosenboim Visit: http://gamla.org.il/english/index.htm Consultant Email: my first name at consultant dot com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: target shell Date: 17 Oct 2002 08:02:48 -0700 From: cpotenza@onetel.net.uk (Claudio Potenza) Organization: http://groups.google.com/ Message-ID: <2025ac82.0210170702.82291e9@posting.google.com> References: "francesco torre" wrote > I noted that calling a function by the shell with few parameters > than the definition has, always forces the missing parameters to 0. > My question: is this a documented feature? If yes, where? Looking at the documentation for Tornado 2 (for Windows) The *Target Shell* is documented in: vxWorks Programmer's guide - 9.2 Target-Resident Shell there, at: 9.2.6 Summary of Target and Host Shell Differences refers you back to the documentation for the *Tornado Shell* at: Tornado User's Guide - 6 Shell there at: 6.3.3 Expressions - Function Calls you'll find your answer By the way, reading all section 6 and specifically "6.2.2 Shell Features" might reserve some interesting surprise for those that never read the manuals... --------------------------- Newsgroups: comp.os.vxworks Subject: Re: strange timer bahaviour Date: Thu, 17 Oct 2002 17:10:53 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1034867481.955098@newsreader2.wirehub.nl> References: <1034839098.567517@newsreader1.wirehub.nl> Hi, If I run the sample-code below... I do call (out of the shell): timDemoInit timDemoSet 20000 /* 20 seconds interval timer */ timDemoMove 15000 /* move time-day to 15 seconds */ /* wait a few seconds */ timDemoMove 15000 /* move time-day to 15 seconds */ /* wait a few seconds */ timDemoMove 15000 /* move time-day to 15 seconds */ And I get negative values for it_value.tv_sec, can anyone tell me why or is this happling with your taget as wel? I use a intel based PIII board. Thanx, Renee sample code: - --------- #include #include "tickLib.h" #include "sigLib.h" #include "logLib.h" #include "stdio.h" #include "timers.h" #include "stdlib.h" #define TIMER_RELTIME 0 #define Milli_Seconds_In_Second (1000) #define Nano_Seconds_In_Milli_Second (1000*1000) timer_t TimerId = 0; struct itimerspec iSpecNow; struct itimerspec iSpecDelay; struct itimerspec iSpecPrev; STATUS timDemoInit(void); STATUS timDemoSet(int seconds); STATUS timDemoMove(int seconds); STATUS timDemoStop(void); STATUS Period2Delay (struct timespec * Delay, int Period); STATUS Delay2Period (int * Period, struct timespec * Delay); void timDemoGetTest(void); void my_handler(timer_t timerid, int arg); void my_handler(timer_t timerid, int arg) { static int prevTick = 0; int now = tickGet(); timer_gettime(timerid, &iSpecNow); logMsg((char*)"TimID %x: diff=%d, interval=%d,%d, value=%d,%d\n", (int) timerid, (now - prevTick), iSpecNow.it_interval.tv_sec, iSpecNow.it_interval.tv_nsec / Nano_Seconds_In_Milli_Second, iSpecNow.it_value.tv_sec, iSpecNow.it_value.tv_nsec / Nano_Seconds_In_Milli_Second ); prevTick = now; } STATUS timDemoStop(void) { return timer_delete( TimerId); } STATUS timDemoInit(void) { if(timer_create(CLOCK_REALTIME, NULL, &TimerId) == ERROR) { logMsg((char*)"timDemoInit: timer_create FAILED\n", 0,0,0,0,0,0); return ERROR; } if(timer_connect(TimerId, my_handler, 0) == ERROR) { logMsg((char*)"timDemoInit: timer_connect FAILED\n", 0,0,0,0,0,0); return ERROR; } printf("TimerId=0x%x\n", (int)TimerId); return OK; } STATUS timDemoSet(int Period) { Period2Delay( &iSpecDelay.it_interval, Period); Period2Delay( &iSpecDelay.it_value, Period); return timer_settime(TimerId, TIMER_RELTIME, &iSpecDelay, &iSpecPrev); } void timDemoGetTest(void) { static int prevTick = 0; int now; while(1) { taskDelay(1); timer_gettime(TimerId, &iSpecNow); if(iSpecNow.it_value.tv_sec < 0) { now = tickGet(); logMsg((char*)"timDemoGetTest: now=%d, diff=%d, interval=%d,%d, value=%d,%d\n", now, (now - prevTick), iSpecNow.it_interval.tv_sec, iSpecNow.it_interval.tv_nsec / Nano_Seconds_In_Milli_Second, iSpecNow.it_value.tv_sec, iSpecNow.it_value.tv_nsec / Nano_Seconds_In_Milli_Second ); prevTick = now; } } } STATUS timDemoMove(int Period) { int time_to_go; int it_value; int it_interval; timer_gettime(TimerId, &iSpecNow); Delay2Period(&it_value, &iSpecNow.it_value); Delay2Period(&it_interval, &iSpecNow.it_interval); time_to_go = it_interval - it_value; if(it_value > Period) { time_to_go = Period - time_to_go; } else { time_to_go = 1; } Period2Delay( &iSpecDelay.it_value, time_to_go); Period2Delay( &iSpecDelay.it_interval, Period); printf("timDemoMove: Period=%d, gettime:it_value=%d, gettime:it_interval=%d, time_to_go=%d\n", Period, it_value, it_interval, time_to_go ); printf("timDemoMove: new_interval=%ld,%ld, new_value=%ld,%ld\n", iSpecDelay.it_interval.tv_sec, iSpecDelay.it_interval.tv_nsec / Nano_Seconds_In_Milli_Second, iSpecDelay.it_value.tv_sec, iSpecDelay.it_value.tv_nsec / Nano_Seconds_In_Milli_Second ); timer_settime(TimerId, TIMER_RELTIME, &iSpecDelay, &iSpecPrev); printf("timDemoMove: settime_o_interval=%ld,%ld, settime_o_value=%ld,%ld\n", iSpecPrev.it_interval.tv_sec, iSpecPrev.it_interval.tv_nsec / Nano_Seconds_In_Milli_Second, iSpecPrev.it_value.tv_sec, iSpecPrev.it_value.tv_nsec / Nano_Seconds_In_Milli_Second); return OK; } STATUS Period2Delay (struct timespec * Delay, int Period) { ldiv_t Ldiv; /* Delay time in milli-seconds (max. delay is about 49 days) */ ldiv_r (Period, Milli_Seconds_In_Second, &Ldiv); Delay->tv_sec = Ldiv.quot; /* Seconds */ Delay->tv_nsec = Ldiv.rem * Nano_Seconds_In_Milli_Second; /* Nano seconds */ return OK; } STATUS Delay2Period (int * Period, struct timespec * Delay) { /* Period time in milli-seconds (max. period is about 49 days) */ *Period = Delay->tv_sec * Milli_Seconds_In_Second + (int)(Delay->tv_nsec / Nano_Seconds_In_Milli_Second); /* Milli seconds */ if(*Period < 0) { printf("Delay2Period: Period= %ld, timespec=%ld,%ld\n", *Period, Delay->tv_sec, Delay->tv_nsec / Nano_Seconds_In_Milli_Second ); } return OK; } - --------- "Renee Teunissen" wrote in message news:1034839098.567517@newsreader1.wirehub.nl... > > > Hi, > > I'm trying to understand serveral timer_gettime and timer_settime > related issues we have. > > We have an application which has to send-out triggers at specified > intervals. To do this we arm a timer and hook it to our services > routine, set the intervals and things go OK. > > But if we change the period-time, strange things happen. > > When we change the period time, we do not only change the it_interval > parameter but also - if possible - the it_value parameter, to make the > current period fit into the new period-time. So we modify the > remaining time as set with it_value of the itimerspec struct to make > it fit. > If I do that, it seems that the new periode time (it_interval) gets > the same value as the "time remaining" (it_value) and thus the timer > is running much to fast. > > does this make a bell ring by someone? > is it allow - on a running timer - to modify the it_value ? > do I forget something? > > Another strange thing. If I call timer_gettime, I do get (sometimes) > it_value.tv_sec with negative values. Is this normal bahaviour? > > > Thanx for far. > Renee. > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Driving Three Monitors from a single PC Date: 17 Oct 2002 08:18:17 -0700 From: technokiran@yahoo.com (Kiran Kumar) Organization: http://groups.google.com/ Message-ID: <552e79d.0210170718.269fdb4d@posting.google.com> I am very new to vxworks. Its just been 1 month since I started working on it. In fact my exposure to real time OS itself is limited. I have developed only 2 simple drivers on QNX before this. I am working on a project to simulate some aircraft displays. This requires that a Single PCI based CPU module (Pentium 4 1.5 Ghz) must drive four monitors. I have absolutely no idea as to how to handle more than 1 monitor. The GUI package to be used is ZINC. Please help me out on this. There is one more very basic problem that I have - I am currently working only on the Simulator. But I am unable to create any files on the hard disk. I was told to use the usrAtaConfig() API. I tried but it dint work. I feel that it is possible only with bootable applications. Am i right or is it possible to create files on hard disk even with the simulator or a downlodable application? If it is possible, Please let me know how. --------------------------- Newsgroups: comp.os.vxworks Subject: Tornado Registry problem Date: 17 Oct 2002 08:38:05 -0700 From: compaqbala@sify.com (Balakumar) Organization: http://groups.google.com/ Message-ID: <46992b69.0210170738.7d3f0751@posting.google.com> Dear sirs, I installed tornado 2.0. When i launch my target server,it shows "Svctcp_.c -tcp SCOKET Creation problem".What should i do? please tell me. with regards balakumar.D --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Driving Three Monitors from a single PC Date: 17 Oct 2002 12:14:20 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: References: <552e79d.0210170718.269fdb4d@posting.google.com> Sender: bpringle@DeadDuck [snip] Kiran> I have absolutely no idea as to how to handle more than 1 Kiran> monitor. The GUI package to be used is ZINC. Please help me Kiran> out on this. You can not do this with standard Zinc. WindML supports multiple display contexts. You will have to sub-class the ZafDisplay class and use the multiple WindML displays. You may need to run separate Zinc applications for each display (Ie, vxWorks tasks). You may have to sub-class the ZafApplication as well to set the ZafDisplay properly. There may also be some global variables that might have to change to task specific variables, or be renamed for each application. Once that is done, the Zinc output from each task should go to the appropriate monitor. I have never ran multi-task Zinc systems. I hope it goes well for you. Regards, Bill Pringlemeir. - -- Have you ever been given eight telephones made out of an infant? Or used the toilet to travel while living on a planet where you have two left feet? Or used a conference call to sing with characters from an AT&T commercial? You will. And the company that will bring it to you: AT&T. --------------------------- Newsgroups: comp.os.vxworks,comp.arch.embedded Subject: Re: Boot Code Integrity check with CRC Date: 17 Oct 2002 10:36:50 -0700 From: ryeung@earthlink.net (Raymond Yeung) Organization: http://groups.google.com/ Message-ID: References: Thanks guys. I think I got all the details I need, based on all the good insights provided so far. --------------------------- Newsgroups: comp.os.vxworks Subject: Telnet Time-Out Date: 17 Oct 2002 11:17:48 -0700 From: gabi@tdsoft.com (Gabi) Organization: http://groups.google.com/ Message-ID: <5607ff2c.0210171017.4bf13d91@posting.google.com> Hi, How can I set the Telnet timeout or there's any timeout in the Telnet session? Further more incase of link fail does the telnet socket perform keep_alive and if not, how can I set this option to the telnet socket (can I use the ioctl option). Tanks Gabi --------------------------- Newsgroups: comp.os.vxworks Subject: Target Shell Date: Thu, 17 Oct 2002 17:08:43 +0200 From: Paolo Accinelli Organization: Marconi, UK Message-ID: <3DAED27B.4020106@marconi.com> Hello, I use vxWorks on PPC860 (GNU compiler), with the target shell both via serial line and networked (telnet). I noted that calling a "c-language" function with few parameters than its definition, give the result that lossing parameters are forced at 0 (zero). This may be an useful behaviour. Is this documented anywhere? Thanks --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sendto in vxworks Date: Thu, 17 Oct 2002 15:00:30 GMT From: mbgardipee@raytheon.com Organization: Raytheon Company Message-ID: References: > Hi all, > > Two hosts comminute with tcp protocol in vxworks, but when the peer host > dose not recevie the socket, I will block the task on the 'sendto' routine > for a long time. why? Your receiver may also be sending an ICMP source quench to the sender. If your connection is point-to-point (ethernet crossover), you can afford to not play nice on the network. You can remove ICMP support from the VxWorks image and just have the TCP flow control meter the data to the receiver. Just remember that without ICMP you will not be able to ping the VxWorks sender. Regards, Melvin Gardipee --------------------------- Newsgroups: comp.os.vxworks Subject: VxWorks dosFs 1.0 chkdsk Utility Date: 17 Oct 2002 12:55:48 -0700 From: ryeung@earthlink.net (Raymond Yeung) Organization: http://groups.google.com/ Message-ID: I got chkdsk source code from WindRiver. Before I use it on my system, I'd like to know if anybody here used it extensively. What do you think of its useability/robustness (crashes, bugs etc)? Any feedback would be appreciated. Thanks, Raymond --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks dosFs 1.0 chkdsk Utility Date: Thu, 17 Oct 2002 22:17:42 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Reply-To: "Leonid Rosenboim" "Raymond Yeung" wrote in message news:d83bb8e3.0210171155.19c0fe71@posting.google.com... > I got chkdsk source code from WindRiver. Before I use it on my > system, I'd like to know if anybody here used it extensively. > What do you think of its useability/robustness (crashes, bugs etc)? > > Any feedback would be appreciated. > Thanks, > Raymond Yeah I used it a lot, but I would not state an opition, becasue I can not be impartial - I consider chkdsk to be my creation, together with Vladimir Levertov, who wrote all of chkdsk's code, with my direction and supervision. - -- - ----------------------------------------------------------------------- Leonid Rosenboim Consultant Email: my first name at consultant dot com --------------------------- Newsgroups: comp.os.vxworks Subject: build problem Date: 17 Oct 2002 08:48:46 -0700 From: compaqbala@sify.com (Balakumar) Organization: http://groups.google.com/ Message-ID: <46992b69.0210170748.4eb707a1@posting.google.com> dear sirs, Does anyone know how to resolve this problem? When I attempt to rebuild a project in Tornado 2.0 for SPARC, the source modules compile to build the .o file, OK. Next, the "vxrm" command appears, I hear a *bonk* beep, and then the build stops. The build window stays open with no further messages and no "Done" message and no *ding* beep. The files aren't removed, --------------------------- Newsgroups: comp.os.vxworks Subject: Re: vxworks VLAN implementation Date: Thu, 17 Oct 2002 17:04:21 -0400 From: "Linux Prj" Organization: Lucent Technologies, Indian Hill Message-ID: References: there is a separate product called TMS 2.0 for vlan, garp etc protocols "Nison.liu" wrote in message news:aole3d$nog$1@debian.bentium.com... > All: > > Where can I find VLAN implementation information under vxWorks? > > Thanks! > Nison > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: 17 Oct 2002 14:38:45 -0700 From: brian_doe@hotmail.com (brian) Organization: http://groups.google.com/ Message-ID: References: <3DAAA37D.EAE13FD0@alcatel-ke.de> <3DAD2568.16C701E8@alcatel-ke.de> <3DAE93DC.21E25B1D@alcatel-ke.de> > Choosing to use VXWorks for a multi-million unit product > range at 25c a pop when you can get a comparable roylty free > option from elsewhere for 50K is stupid. OK, WRS may offer > a complete solution: OS, TCP/IP, IDE etc that you will have to buy > in separately and it may be worth paying 100K for an integrated > package but to agree a deal that costs you 1M is stupid IMO. Well in some cases you can GET a full OS/Stack etc that is ported to your processor. I guess we will agree to disagree. Im stupid. You are naive. --------------------------- Newsgroups: comp.os.vxworks Subject: dosFs 1.0 Integrity Information Needed Date: 17 Oct 2002 15:49:47 -0700 From: ryeung@earthlink.net (Raymond Yeung) Organization: http://groups.google.com/ Message-ID: Hello, Our system install dosFs 1.0 (on Tornado 2.0) on Intoto's Flash Media Manager. The flash device is a removable flash SIMM (but I'm not expecting people to remove it while the system is up, as this would amount to hot-swapping, that we're not supporting currently). I'm trying to determine what error detection/recovery mechanism available on this version of dosFs. I tried to contact WindRiver, but in no vail, as this is an old version, and there's little documentation. I suppose if there's any H/W problem (e.g. flash cells stuck-at faults), it's the responsibility of the Flash Media layer to do the detection? To guarantee file content correctness, seems like I need to implement some sort of error detection code (e.g. checksum) on a file by file basis at application layer. Any information in this arena, particularly dosFs 1.0, would be much appreciated. Thanks, Raymond --------------------------- Newsgroups: comp.os.vxworks Subject: How to force asynchronous read() to finish in time? Date: 17 Oct 2002 16:23:19 -0700 From: yangliang_mr@hotmail.com (Liang Yang) Organization: http://groups.google.com/ Message-ID: <4f57a77f.0210171523.208d1ffd@posting.google.com> Hi, I need to call TCP socket read() in the target side(MBX 860) to read the data from the server(Windows 2000) side. But since read() is asynchronous operation, the task which calls read() will yield the control to other tasks after it calls read(). So the code to check the data which is put after read() shows the data is still not ready. How can I ensure data is ready after I call read()? Thanks. Liang --------------------------- Newsgroups: comp.os.vxworks Subject: Re: build problem Date: Fri, 18 Oct 2002 08:23:56 +0900 From: "Girish Gulawani" Message-ID: <3daf4699$0$19939$44c9b20d@news3.asahi-net.or.jp> References: <46992b69.0210170748.4eb707a1@posting.google.com> i think there should be one more sound - ssshhh. "Balakumar" wrote in message news:46992b69.0210170748.4eb707a1@posting.google.com... > dear sirs, > Does anyone know how to resolve this problem? > > When I attempt to rebuild a project in Tornado 2.0 for SPARC, the source > modules compile to build the .o file, OK. Next, the "vxrm" command > appears, I hear a *bonk* beep, and then the build stops. The build window > stays open with no further messages and no "Done" message and no *ding* > beep. The files aren't removed, --------------------------- Newsgroups: comp.os.vxworks,comp.arch.embedded Subject: Re: Boot Code Integrity check with CRC Date: 17 Oct 2002 16:47:20 -0700 From: ryeung@earthlink.net (Raymond Yeung) Organization: http://groups.google.com/ Message-ID: References: <3DAE0416.926F5812@com21.com> Rajesh Tripathi wrote in message news:<3DAE0416.926F5812@com21.com>... > Hi Raymond, > > Raymond Yeung wrote: > > > Thanks for all the replies. Having read all of them, I've the > > following thoughts: > > > > 1. S-record is a wrong form to use. Initially, I explored what > > format to use; VxWorks generates elf, hex (S-record). But I'm > > not sure I found the binary version that I can load directly to > > flash and have the processor boot. That's why I thought about > > using S-record, and write a loader on target to interpret all > > S-records. > > I used elfToBin utility for generating binary file form elf file. That was > two years ago on Tornado 2.0 for PPC and I don't find that file on my T2.1 > for MIPS installation now. I believe the recommended way now is to use > "objcopyppc -O binary --binary-without-bss xyz.elf xyz.bin" > > Which version of Tornado you are using? > I use Tornado 2.0. I tried out elfToBin. When I diff its output with Emulator's convert utility output (with header stripped), I find that elfToBin's output contains lots of extra 0's at the end of the file (hence 2-3 times larger than the other file). Otherwise, it looks like the two binaries are the same. Are all those 0's the BSS? I also tried your "objcopyppc -O binary --binary-without-bss xyz.elf xyz.bin" command. Apparently, my version of objcopyppc don't seem to support "--binary-without-bss" option. Any other way to strip all the extra zeros? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: spr #23225 Date: 17 Oct 2002 14:56:28 -0700 From: tjaimin@yahoo.com (Jai) Organization: http://groups.google.com/ Message-ID: <17902e07.0210171356.308e82ef@posting.google.com> References: <17902e07.0210161504.421c9613@posting.google.com> Bill Pringlemeir wrote in message news:... > Jai> Wind River had a patch (23225) for OpenSSL on Wind Web Server > Jai> a while back. They've stopped distributing this path since they > Jai> do not support WWS1.0 anymore. Has anyone got this patch? Is it > Jai> possible to get this patch from any other place? > > Bzzzt. The patch is for SSLeay, not OpenSSL. The current code in CVS > for OpenSSL (v 0.9.8 I think) has been set up to configure to build > OpenSSL on the Solaris platform for a vxWorks target. You can also > look in the vxWorks FAQ and search the OpenSSL mailing list for some > more info on vxWorks and OpenSSL. > > It is probably less effort to get 0.9.6g working that it is chasing > WRS patches. Here's an idea, use google. > > hth, > Bill Pringlemeir. I am not looking to port OpenSSL on to vxWorks, I've taken care of that. The part I am looking for is the shim layer for Wind Web Server 1.0 to use OpenSSL sockets, I suppose thats what they have the patch for. And, OpenSSL was formerly known as SSLeay. --------------------------- Newsgroups: comp.os.vxworks,comp.arch.embedded Subject: Re: Boot Code Integrity check with CRC Date: Fri, 18 Oct 2002 00:35:35 GMT From: Wolfgang Denk Organization: DENX Software Engineering, Germany Message-ID: References: Sender: wd@denx.muc.de (wd) ryeung@earthlink.net (Raymond Yeung) writes: >5. What is PPCBoot? I try searching for it on Googles, and got > too many returns. Would appreciate it if one can provide a > link to its tutorial? (Note: I'm currently booting using VxWorks > boot rom. Is PPCBoot another boot rom?) Urrrrrghhh... If you google just for "PPCBoot", the very first link that turns up is the PPCBoot home page at Sourceforge ("Embedded PowerPC Linux Boot Project"), and the second link is the "PPCBoot-users Info Page". That was too many links for you to follow? Wolfgang Denk - -- Software Engineering: Embedded and Realtime Systems, Embedded Linux Phone: (+49)-8142-4596-87 Fax: (+49)-8142-4596-88 Web: www.denx.de As long as we're going to reinvent the wheel again, we might as well try making it round this time. - Mike Dennison --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Bootrom question Date: 17 Oct 2002 18:19:40 -0700 From: snower_tiger@hotmail.com (snower) Organization: http://groups.google.com/ Message-ID: <3f60ed07.0210171719.78185924@posting.google.com> References: <3f60ed07.0210161659.10efdd12@posting.google.com> be_gyus@yahoo.com (james) wrote in message news:... > snower_tiger@hotmail.com (snower) wrote in message news:<3f60ed07.0210161659.10efdd12@posting.google.com>... > > I used tornado to develop the project. but I have a question, if I > > build bootrom with uncompressed hex, it work well, but if I build > > bootrom with compressed hex, it cant work, it will call excintconnect > > funtion, and did not come back. Is there anyone know why, anybody can > > suggest any idel? > > my email: snower_tiger@hotmail.com > > > > > Which is the Tornado version u r using, the bsp and whats the error u > r gettting while building? I used Tornado II, vxWorks 5.4, the bsp i modified from mpc8240, and there is no error while building, but it can not work on the target board. --------------------------- Newsgroups: comp.os.vxworks Subject: how to re-size of TFFS? Date: 17 Oct 2002 19:27:02 -0700 From: wangwz@chinatopglobal.com (wangwz) Organization: http://groups.google.com/ Message-ID: <72c0062e.0210171827.7dbe013e@posting.google.com> Hi, everyone, How can I change the size of TFFS? Now I use MPC8245 with 8Mflash, the TFFS is from 0xff800000 to 0xfff00000, now I want change it from 0xff800000 to 0xffe00000, how can I do for it? Thanks --------------------------- Newsgroups: comp.os.vxworks Subject: Re: spr #23225 Date: 17 Oct 2002 23:17:43 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: References: <17902e07.0210161504.421c9613@posting.google.com> <17902e07.0210171356.308e82ef@posting.google.com> Sender: bpringle@DeadDuck Jai> Wind River had a patch (23225) for OpenSSL on Wind Web Server Jai> a while back. They've stopped distributing this path since they Jai> do not support WWS1.0 anymore. Has anyone got this patch? Is it Jai> possible to get this patch from any other place? Bill Pringlemeir wrote... > > Bzzzt. The patch is for SSLeay, not OpenSSL. The current code in CVS > for OpenSSL (v 0.9.8 I think) has been set up to configure to build > OpenSSL on the Solaris platform for a vxWorks target. You can also > look in the vxWorks FAQ and search the OpenSSL mailing list for some > more info on vxWorks and OpenSSL. > > It is probably less effort to get 0.9.6g working that it is chasing > WRS patches. Here's an idea, use google. Jai> I am not looking to port OpenSSL on to vxWorks, I've taken care Jai> of that. The part I am looking for is the shim layer for Wind Jai> Web Server 1.0 to use OpenSSL sockets, I suppose thats what they Jai> have the patch for. And, OpenSSL was formerly known as SSLeay. Sorry, my mind reading abilities are off. The OpenSSL interface has changed since SSLeay. There are quite a few good tutorial on the web on how to utilize OpenSSL [a PDF from Linux Journal?]. You might also look at the Curl project to see how a HTTP retrieval software program uses secure and plain HTTP sockets. I really think that the patch would be a disappointment if you found it... but I could be wrong. fwiw, Bill Pringlemeir. - -- Kibo, the great teacher was Kibotic. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks dosFs 1.0 chkdsk Utility Date: 17 Oct 2002 22:03:12 -0700 From: ryeung@earthlink.net (Raymond Yeung) Organization: http://groups.google.com/ Message-ID: References: Hi Leonid, I saw some comments in the source code that says crashes could happen under some conditions. I wasn't quite sure what the situations are. That's why I asked. BTW, I probably was one of the early user of FlashLib (back in 1997/98). I vaguely remembered you also created FlashLib? I might have talked with you or Vladimir. Raymond "Leonid Rosenboim" wrote in message news:... > "Raymond Yeung" wrote in message > news:d83bb8e3.0210171155.19c0fe71@posting.google.com... > > I got chkdsk source code from WindRiver. Before I use it on my > > system, I'd like to know if anybody here used it extensively. > > What do you think of its useability/robustness (crashes, bugs etc)? > > > > Any feedback would be appreciated. > > Thanks, > > Raymond > > Yeah I used it a lot, but I would not state an opition, becasue I can not > be impartial - I consider chkdsk to be my creation, together with > Vladimir Levertov, who wrote all of chkdsk's code, with my direction > and supervision. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: dosFs 1.0 Integrity Information Needed Date: 17 Oct 2002 22:05:05 -0700 From: ryeung@earthlink.net (Raymond Yeung) Organization: http://groups.google.com/ Message-ID: References: FYI. I'm looking for this information from the perspective of writing system diatnostics (POST). Before I invent any wheel, I thought I'd check out what're available to me first. ryeung@earthlink.net (Raymond Yeung) wrote in message news:... > Hello, > > Our system install dosFs 1.0 (on Tornado 2.0) on Intoto's Flash Media Manager. > The flash device is a removable flash SIMM (but I'm not expecting people to > remove it while the system is up, as this would amount to hot-swapping, that > we're not supporting currently). > > I'm trying to determine what error detection/recovery mechanism available > on this version of dosFs. I tried to contact WindRiver, but in no vail, > as this is an old version, and there's little documentation. > > I suppose if there's any H/W problem (e.g. flash cells stuck-at faults), > it's the responsibility of the Flash Media layer to do the detection? > To guarantee file content correctness, seems like I need to implement some > sort of error detection code (e.g. checksum) on a file by file basis at > application layer. > > Any information in this arena, particularly dosFs 1.0, would be much > appreciated. > > Thanks, > Raymond --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Calling an object method in a protection domain from a shared library Date: 17 Oct 2002 22:11:52 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0210172111.4d9f3672@posting.google.com> References: Hello again, > My questions: > Is it possible to call an object member in a protection domain from a > shared library? I can only think of creating a function in the shared > library that accepts a pointer to the object member. This pointer > could be made global to the shared library. Is this the best way to > do it? You cannot do it symbolically, but, as you describe, it is possible for shared library code to call functions in an application via a function pointer. In the case of C++ though, calling a member function via a function pointer is not so simple... Of course, there's nothing to stop you simply passing an object pointer to the library. That way the library code can call the member function through the object instance in the normal way. The only thing you will not be able to do is have an undefined symbol in the shared library resolved by one in the application domain. Once the application and shared library domain are linked though, they are essentially a single address space, so as long as you have the address, you can make the call. > Also, I have shared library functions that create objects. Will these > objects reside in the protection domain (that attached to the shared > library) or the shared library domain. I'm thinking it's the > protection domain. Am I correct? Yes, they will be created in the application domain's heap space. HTH, John... --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Boot device Date: Fri, 18 Oct 2002 08:19:05 +0200 From: "Michael Lawnick" Organization: Buergernetz Dillingen Message-ID: References: Reply-To: "Michael Lawnick" Sender: mlawnick@p5088fb8c.dip.t-dialin.net Hi Ole, I thought about the same some time before too, but found no good solution. As AFAIK BIOS doesn't store anywhere its current boot source, you'd have to provide two different bootimages, one for fd with nvRamToFd and one for hd with nvRamToAta, or at least with a merker for the source drive. As this will give more project overhead than it's worth, I had let it be. If you find out how to get boot source from H/W, tell it here, I'm pretty sure others would like to know too. Regards, Michael "Ole Asbjørn Fadum" schrieb im Newsbeitrag news:MPG.1818e0f24120b392989681@news.wineasy.se... > I'm using a x86 board for my project. The pentium boards doesn't have > any NVRAM, but some nice guys has provided the files for store the > bootline on disk instead (I.e. nvRamToAta.c). Sometimes my system is > booted from IDE disk, and sometimes from floppy. I want the > bootparameters to reside in the bootdevice, and not as it is today as > the floppy boot takes the IDE bootparameters. How could I find out which > device was cause in the current boot??? --------------------------- Newsgroups: comp.os.vxworks Subject: a question about vxVMI and 3C905B Date: Fri, 18 Oct 2002 14:44:09 +0800 From: "Jim Chu" Organization: utsz Message-ID: Reply-To: "Jim Chu" Hi all, I have a question about vxVMI and 3C905B. I list the hardware and software configuration first: Hardware: A PC based on x86 CPU(PIII) NIC: 3C905B Software : Tornado 2.0.2 for x86 vxVMI is installed BSP : pcPentium 1. When I only including "basic MMU support" in vxWorks kernel, everything is OK. 2. When I including " full MMU support" in vxWorks kernel, building vxWorks kernel successfully. When the kernel starting at 0x108000 , errors present as follows: Page Fault Program Counter:0x00167197 Status Register:0x00010246 Error Code:0x00000003 Task:0x7ffee7c "tRootTask" My question is what happens on 3C905B when I including " full MMU support" in vxWorks kernel? Do someone have similar experience like this? Any help is appreciate! Thanks , Cjg --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Anyone already ported from Tornado 2.0 to Tornado 2.2 ? Date: Fri, 18 Oct 2002 07:20:39 +0000 (UTC) From: Johan Borkhuis Organization: Proxim Corp. Message-ID: References: <3daec4b6@news.mhogaming.com> "steeve gagne" wrote: > Is There anyone using Tornado 2.2 ? > > Does WindRiver give you a new BSP ? > becaus I am stuck with Tornado 2.2 Wihout any BSPS As far as I know you should always reeceive a BSP with your distribution, as you did originally. So ask them again for the BSP. The BSPs are distributed on a seeperate CD. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks dosFs 1.0 chkdsk Utility Date: Fri, 18 Oct 2002 09:49:13 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Reply-To: "Leonid Rosenboim" "Raymond Yeung" wrote in message news:d83bb8e3.0210172103.74156751@posting.google.com... > Hi Leonid, > > I saw some comments in the source code that says crashes could happen > under some conditions. I wasn't quite sure what the situations are. > That's why I asked. > If I recall correctly, chkdsk might crash if there is insufficient memory to hold the number of files/dirs found on the disk. In DosFs2.0, the chkdsk stuff was reqritten, so that if memory is insufficient, and you know large FAT32 volumes can contain hundreds of thousands of files, then the 2nd FAT area on teh disk is used as scratch area instead of memory, so there is guarantee the scan will work, albeit more slowly. > BTW, I probably was one of the early user of FlashLib (back in 1997/98). > I vaguely remembered you also created FlashLib? I might have talked > with you or Vladimir. Yes, I and Vlad have created FlashLib, and cdromFs before that. FlashLib began selling in 1996, so you where not one of the first customers, but rather one of the last ones for FlashLib. I have read your other post, regarding Intoto FTL: It is the FTL responsability to make sure even disk block is consistent, like a magnetic disk that writes a CRC with every sector. chkdsk only verifies that the information in FAT is consistent with itself and with the information in all the directory entries, and it does not dive into the file data itself, and does not read-write any data blocks except to read directory content. Let me know if you need further help. Leonid --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Fri, 18 Oct 2002 09:57:23 +0200 From: Tim Sinkins Organization: Alcatel Kommunikations Elektronik Message-ID: <3DAFBEE3.1EE916A5@alcatel-ke.de> References: <3DAAA37D.EAE13FD0@alcatel-ke.de> <3DAD2568.16C701E8@alcatel-ke.de> <3DAE93DC.21E25B1D@alcatel-ke.de> Reply-To: timothy.sinkins@alcatel-ke.de brian wrote: > > > Choosing to use VXWorks for a multi-million unit product > > range at 25c a pop when you can get a comparable roylty free > > option from elsewhere for 50K is stupid. OK, WRS may offer > > a complete solution: OS, TCP/IP, IDE etc that you will have to buy > > in separately and it may be worth paying 100K for an integrated > > package but to agree a deal that costs you 1M is stupid IMO. > > Well in some cases you can GET a full OS/Stack etc that is ported to > your processor. As you can from other vendors. In any case you used the Walkman as an example. Why does a walkman need a TCP/IP or any other stack? Surely, it just needs a basic mulit-tasking RTOS and little else? > I guess we will agree to disagree. Obviously. > Im stupid. If you like, I didn't say that. > You are naive. Actually I think it is you that is naive. You seem to think that the deal that the company offer at the start of the process is the only one that is available. It isn't. Tim --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How to force asynchronous read() to finish in time? Date: Fri, 18 Oct 2002 10:31:56 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <4f57a77f.0210171523.208d1ffd@posting.google.com> Reply-To: "Leonid Rosenboim" "Liang Yang" wrote in message news:4f57a77f.0210171523.208d1ffd@posting.google.com... > Hi, > > I need to call TCP socket read() in the target side(MBX 860) > to read the data from the server(Windows 2000) side. But > since read() is asynchronous operation, the task which > calls read() will yield the control to other tasks after > it calls read(). So the code to check the data which is > put after read() shows the data is still not ready. > > How can I ensure data is ready after I call read()? You are very confused, but since you are a student (posting from a University IP address), I guess that is to be expected. If you have set the socket into non-blocking mode, read() will return immediately, weather data is available or not. If the socket is in its default blocking mode, read will WAIT (thereby allowing other tasks to use the CPU) until SOME data is available, and upon return will report the number of bytes retreived. So the code that is executed after read() returns should simply check how many butes read() returned, and that amount reported by read() has been indeed copied to your buffer and is perfectly ready for further use. HTH Leonid --------------------------- Newsgroups: comp.os.vxworks Subject: Where are the shared library created C++ objects located Date: 18 Oct 2002 01:36:41 -0700 From: steven4443@yahoo.com (stephen hallacy) Organization: http://groups.google.com/ Message-ID: Hello, With regards to c++ objects created in a shared library function, is the c++ object located in the domain where the shared library was called from or is the c++ object in the located in the shared library? Thanks, Stephen --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Boot device Date: Fri, 18 Oct 2002 08:50:52 +0000 (UTC) From: David Laight Organization: BT Openworld Message-ID: References: > As AFAIK BIOS doesn't store anywhere its current boot source, If it is 'standard' PC hardware, then the bios identifier for the boot disk should be in 'dl' when the booted code is entered. dl = 0 => floppy 0 dl = 1 => floppy 1 dl = 0x80 => hard disk 0 dl = 0x81 => hard disk 1 etc. Only trouble is working out which order the BIOS has ordered the hard disk in. What the boot code isn't (usually) told is which partition was booted. It has to rescan the MBR looking for a partition of the correct type. David --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks dosFs 1.0 chkdsk Utility Date: Fri, 18 Oct 2002 08:54:21 +0000 (UTC) From: David Laight Organization: BT Openworld Message-ID: References: > then the 2nd FAT area on teh disk is used as scratch area instead of memory, What does it do if the volume doesn't have a second copy of the FAT? For flash the extra FAT updates become tiresome, and are of little real benefit for fixing corruptions, so I only created a single copy of the FAT. David --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks dosFs 1.0 chkdsk Utility Date: Fri, 18 Oct 2002 11:13:05 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Reply-To: "Leonid Rosenboim" "David Laight" wrote in message news:aooi7s$906$2@knossos.btinternet.com... > > > then the 2nd FAT area on teh disk is used as scratch area instead of memory, > > What does it do if the volume doesn't have a second copy of the FAT? > For flash the extra FAT updates become tiresome, and are of little > real benefit for fixing corruptions, so I only created a single copy > of the FAT. > > David > If RAM has sufficient room for the scratch table, the 2nd FAT wont even be considered, that should take care of your Flash case. If neither sufficient RAM nor 2nd FAT are available, DosFs2.0 wont be able to consistency check the volume. You can reproduce this by (1) having a volume formatted with a single FAT copy, and then allocate most of the heap until free memory is under 100 bytes ro so, then running chkdsk(). Oh by the way, the 2nd FAT does have a small benefit on DosFs 2.0 - the 2nd FAT copy is updated at a different time, and that helps identify if the volume has got any inconsistencies at all. Also if there is a read error on the 1st FAT, the 2nd is used. DosFs2.0 also make backup copies of the boot bloc, so any single bad sector can be recovered from with 2 FAT copies. - - Leonid --------------------------- End of New-News digest ********************** From vxwexplo-errs@csg.lbl.gov Sat Oct 19 02:52:03 2002 From: "Sudhakar Tadi" Date: Sat Oct 19 02:52:05 PDT 2002 Subject: SCSIMgr question This is a multi-part message in MIME format. ------=_NextPart_000_0060_01C27783.3E35F5A0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Just trying for one more time ----- Original Message -----=20 From: Sudhakar Tadi=20 To: the vxWorks Users Group Exploder ; vxworks_users@csg.lbl.gov=20 Cc: Sudhakar Tadi=20 Sent: Friday, October 18, 2002 4:44 PM Subject: SCSIMgr question. Hi All, I need some clarification regarding SCSI drivers. As per the specs = there is one SCSIMgr taks for scsi controller, that's fine. But what = happens If I have the two similar cards in the same machine, do I need = to create seperate SCSIMgr task or one SCSImgr task enough for managing = the two cards. My question is can single SCSImgr task handles the two = similar type of HBA cards. Can somebody clarify. Thanks in Advance -Tadi Sudhakar Tadi HCL Technologies Ltd 184-188, NSK Road,Vadapalani Chennai-600 026,INDIA Ph:91-44-372 8366 Ext:1134 Fax:91-44-480 6640 http://san.hcltech.com ------=_NextPart_000_0060_01C27783.3E35F5A0 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
Just trying for one more time
 
 
----- Original Message -----
From:=20 Sudhakar=20 Tadi
To: the vxWorks Users Group Exploder = ; vxworks_users@csg.lbl.gov =
Sent: Friday, October 18, 2002 = 4:44=20 PM
Subject: SCSIMgr = question.

Hi All,
I need some clarification regarding = SCSI=20 drivers.  As per the specs there is one SCSIMgr taks for scsi = controller,=20 that's fine.  But what happens If I have the two similar cards in = the=20 same machine, do I need to create seperate SCSIMgr task or one SCSImgr = task=20 enough for managing the two cards. My question is can single SCSImgr = task=20 handles the two similar type of HBA cards. Can somebody = clarify.
 
Thanks in Advance
-Tadi
Sudhakar Tadi
HCL Technologies = Ltd
184-188,=20 NSK Road,Vadapalani
Chennai-600 026,INDIA
Ph:91-44-372 8366=20 Ext:1134
Fax:91-44-480 6640
http://san.hcltech.com
------=_NextPart_000_0060_01C27783.3E35F5A0-- From vxwexplo-errs@csg.lbl.gov Sat Oct 19 04:03:13 2002 From: Vxworks Exploder Date: Sat Oct 19 04:03:14 PDT 2002 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sat Oct 19 04:03:11 PDT 2002 Subject: Re: Bootrom question Subject: Re: strange timer bahaviour Subject: Re: strange timer bahaviour Subject: vxWorks BSP manual needed Subject: arptnew failed Subject: Re: VxWorks dosFs 1.0 chkdsk Utility Subject: Alignment problems in PPC Subject: Re: Where are the shared library created C++ objects located Subject: Re: diskon chip probelm Subject: Good way to detect when a PPP link is fully operational? Subject: Re: vxWorks BSP manual needed Subject: Re: Where are the shared library created C++ objects located Subject: suspend functions using semTake() Subject: Re: Porting of vxWorks. Subject: Re: Boot Code Integrity check with CRC Subject: post-mortem in WindView Subject: Re: Alignment problems in PPC Subject: Re: vxWorks BSP manual needed Subject: Re: Telnet Time-Out Subject: Re: Where are the shared library created C++ objects located Subject: Re: suspend functions using semTake() ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Bootrom question Date: Fri, 18 Oct 2002 09:41:00 -0400 From: joe durusau Organization: WEBUSENET.com Message-ID: <3DB00F6C.757641C8@bellsouth.net> References: <3f60ed07.0210161659.10efdd12@posting.google.com> <3f60ed07.0210171719.78185924@posting.google.com> snower wrote: > be_gyus@yahoo.com (james) wrote in message news:... > > snower_tiger@hotmail.com (snower) wrote in message news:<3f60ed07.0210161659.10efdd12@posting.google.com>... > > > I used tornado to develop the project. but I have a question, if I > > > build bootrom with uncompressed hex, it work well, but if I build > > > bootrom with compressed hex, it cant work, it will call excintconnect > > > funtion, and did not come back. Is there anyone know why, anybody can > > > suggest any idel? > > > my email: snower_tiger@hotmail.com > > > > > > > > > > Which is the Tornado version u r using, the bsp and whats the error u > > r gettting while building? > > I used Tornado II, vxWorks 5.4, the bsp i modified from mpc8240, and > there is no error while building, but it can not work on the target > board. For some boards, there is a problem in which the decompression doesn't work. Take a look at the files in the BSP and see if any of them mention your board being one such. If not, call your local rep so submit an spr via the website. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Re: strange timer bahaviour Date: 18 Oct 2002 07:51:09 -0700 From: rick@cyclone.com (Richard Nardone) Organization: http://groups.google.com/ Message-ID: <29b153d9.0210180651.6b2c0728@posting.google.com> References: <1034839098.567517@newsreader1.wirehub.nl> <1034867481.955098@newsreader2.wirehub.nl> > > And I get negative values for it_value.tv_sec, can anyone tell me why > or is this happling with your taget as wel? I use a intel based > PIII board. > Renee struct timespec { /* interval = tv_sec*10**9 + tv_nsec */ time_t tv_sec; /* seconds */ long tv_nsec; /* nanoseconds (0 - 1,000,000,000) */ }; struct itimerspec { struct timespec it_interval; /* timer period (reload value) */ struct timespec it_value; /* timer expiration */ }; Note that tv_sec is of type time_t, which is unsigned long. You are treating this as an int, which is why you are seeing negative values sometimes. - Richard Nardone Cyclone Microsystems --------------------------- Newsgroups: comp.os.vxworks Subject: Re: strange timer bahaviour Date: 18 Oct 2002 08:25:42 -0700 From: rick@cyclone.com (Richard Nardone) Organization: http://groups.google.com/ Message-ID: <29b153d9.0210180725.2c8f96b@posting.google.com> References: <1034839098.567517@newsreader1.wirehub.nl> <1034867481.955098@newsreader2.wirehub.nl> "> > void > timDemoGetTest(void) > { > static int prevTick = 0; > int now; > > while(1) > { > taskDelay(1); > > timer_gettime(TimerId, &iSpecNow); > > if(iSpecNow.it_value.tv_sec < 0) > { > now = tickGet(); > Renee, tv_sec is of type unsigned long. You may want to change this comparison to see if it is equal to zero. You may want to read the timerLib manual a bit more closely. Also, you need to understand that the fundamental resolution of your timer is tied to the system clock interval (i.e. tick) which is usually around 60Hz or 16.66 ms. Nanosecond resolution is not possible with this timer facility. Also, you seem to be setting both an interval timer, and a one shot timer at the same time. Is this intentional? Richard Nardone Cyclone Microsystems --------------------------- Newsgroups: comp.os.vxworks Subject: vxWorks BSP manual needed Date: Fri, 18 Oct 2002 18:36:18 GMT From: "vb" Organization: AT&T Broadband Message-ID: Hello, I wonder if anyone can share a document describing vxWorks BSP internals, API, etc. Thanks in advance! /vb --------------------------- Newsgroups: comp.os.vxworks Subject: arptnew failed Date: Fri, 18 Oct 2002 18:34:23 GMT From: "somebody" Organization: Magma Communications Ltd. Message-ID: Hi Does any one know why I get this msg 0xf7bf88 (tNetTask): arptnew failed on 50f22423 0xf7bf88 (tNetTask): arptnew failed on 50f22423 0xf7bf88 (tNetTask): arptnew failed on 50f22423 Does it mean that my network stack has run out of memory. If so, I am presumming I will get more error message and my ip connection will not work properly. Why is my ip connection is still fine but I still get arptNew problem. Hadi --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks dosFs 1.0 chkdsk Utility Date: 18 Oct 2002 11:48:22 -0700 From: ryeung@earthlink.net (Raymond Yeung) Organization: http://groups.google.com/ Message-ID: References: Hi Leonid, Thanks for your kind offer and sharing of insights. I'll keep that in mind. It seems to me the underlining file system (dosFs 1.0) and/or flash media implementation do not provide enough error detection mechanism. chkdsk would provide some coverage in terms of sectors/blocks/links management (I've yet to read in details the comments/documentation). Intoto Flash Media Manager doesn't seem to offer much in this regard either. I'm thinking about putting at application layer checksum/CRC per file for error detection. This should be an umbrela check that covers all possible problems all the way down to flash device cell level. The # of files we need to deal with really isn't that many (probably within 10 or so). This approach wouldn't help me to isolate the problem, but would at least let me know there's a problem somewhere on the flash (or perhaps the address/data lines). Raymond "Leonid Rosenboim" wrote in message news:... > "Raymond Yeung" wrote in message > news:d83bb8e3.0210172103.74156751@posting.google.com... > > Hi Leonid, > > > > I saw some comments in the source code that says crashes could happen > > under some conditions. I wasn't quite sure what the situations are. > > That's why I asked. > > > > If I recall correctly, chkdsk might crash if there is insufficient memory to > hold the number of files/dirs found on the disk. > > In DosFs2.0, the chkdsk stuff was reqritten, so that if memory is > insufficient, > and you know large FAT32 volumes can contain hundreds of thousands of files, > then the 2nd FAT area on teh disk is used as scratch area instead of memory, > so there is guarantee the scan will work, albeit more slowly. > > > BTW, I probably was one of the early user of FlashLib (back in 1997/98). > > I vaguely remembered you also created FlashLib? I might have talked > > with you or Vladimir. > > Yes, I and Vlad have created FlashLib, and cdromFs before that. > FlashLib began selling in 1996, so you where not one of the first > customers, but rather one of the last ones for FlashLib. > > I have read your other post, regarding Intoto FTL: > It is the FTL responsability to make sure even disk block is consistent, > like a magnetic disk that writes a CRC with every sector. > chkdsk only verifies that the information in FAT is consistent with itself > and with the information in all the directory entries, and it does not dive > into the file data itself, and does not read-write any data blocks > except to read directory content. > > Let me know if you need further help. > > Leonid --------------------------- Newsgroups: comp.sys.powerpc.misc,comp.os.vxworks,gnu.gcc.help Subject: Alignment problems in PPC Date: 18 Oct 2002 11:58:21 -0700 From: FrankGleeson@motorola.com (frankg) Organization: http://groups.google.com/ Message-ID: I am debugging a problem in our vxworks BSP whish is based on the wrSbc8260 BSP. The following code fails to copy the string from tyName to consoleName: In File bootInit.c char consoleName [20]; /* line 578 global char array */ In Function usrRoot strcpy (&consoleName[0], tyName); Using our visionclick debugger I was able to determine that the address that was passed in as the dest pointer was 006a3cf0 but the linker says that the address is 006a3cf4. This is consistent with the state of memory after the strcpy. The string is copied intact to 006a3cf0. ccppc -B/home/copeng/project/semboot1.0.0/wind/host/sun4-solaris2/lib/gcc-lib/ - -mstrict-align -ansi -nostdinc - -I/home/copeng/project/semboot1.0.0/wind/target/config/sembsp_proj - -I/home/copeng/project/semboot1.0.0/wind/target/config/all - -I/home/copeng/project/semboot1.0.0/wind/target/h - -I/home/copeng/project/semboot1.0.0/wind/target/src/config - -I/home/copeng/project/semboot1.0.0/wind/target/src/drv - -I/home/copeng/project/semboot1.0.0/boot - -I/home/copeng/project/semboot1.0.0/com_inc - -I/home/copeng/project/semboot1.0.0/inc - -I/home/copeng/project/semboot1.0.0 - -I/home/copeng/project/semboot1.0.0/wind/target/h/arch/ppc - -DCPU=PPC603 -DSEM -g -O0 -g2 -gstabs -gxcoff -gdwarf -P -x assembler-with-cpp -Wa,-L -c -o sysALib.o sysALib.s We are using vxWorks 5.4.2. We are using the gnu compiler. Do you have any idea what could cause this? Here is the actual compiler command for this file. Is there a compile flag that could influence the alignment? do we need to put pragmas around the consoleName variable declaration to force 8 byte alignment? If it's an alignment issue then it seems odd that the 8265 which is 603e based (32 bit implementation) would require 64 bit alignment. I noticed we are not using the -mcpu=603e or any of the -mpower switches. Do you think this would change gcc's allignment policy? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Where are the shared library created C++ objects located Date: Fri, 18 Oct 2002 19:53:32 GMT From: "H. E. Taylor" Organization: MTS Internet Message-ID: <3DB08114.EDC@despam.autobahn.mb.ca> References: In article , stephen hallacy wrote: > > With regards to c++ objects created in a shared library function, is > the c++ object located in the domain where the shared library was > called from or is the c++ object in the located in the shared library? > You appear to be confused about the runtime object model. At any rate, you can answer your question for yourself using 'lkup' and suitable test module & library. There is a book by Stan Lippman "Inside the C++ Object Model", which may be of some use to you. - -het - -- "progress in software has not followed Moore's law." -John Holland Computer Links: http://www.autobahn.mb.ca/~het/clinks.html H.E. Taylor http://www.autobahn.mb.ca/~het/ --------------------------- Newsgroups: comp.os.vxworks,comp.os.qnx Subject: Re: diskon chip probelm Date: Fri, 18 Oct 2002 15:59:19 -0400 From: david lindauer Message-ID: <3DB06817.6A3E2BDB@notifier-is.net> References: <3d47da57997f9a14d13396faba7f4e3c.98538@mygate.mailgate.org> Ganesh Anand wrote: > hi all, > > i am facing a probelm due to diskon chip. i had done networking > between two or more boards embedded with MPC823. everything works > properly if i download the image to the boards. But samething is not > happening if i boot the boards through diskon. i am using UDP for > broadcasting & tcp for communication between two boards. it is showing > some error in getting the broadcast address. if anyody cud show some > insight regarding the probelm it wud be of great help to me. > you will probably have to set the subnet mask and the broadcast address by hand... David > > ganesh > > -- > Posted via Mailgate.ORG Server - http://www.Mailgate.ORG --------------------------- Newsgroups: comp.os.vxworks Subject: Good way to detect when a PPP link is fully operational? Date: 18 Oct 2002 13:27:46 -0700 From: doug_dyer@yahoo.com (Doug Dyer) Organization: http://groups.google.com/ Message-ID: <9b386c37.0210181227.49660edd@posting.google.com> My vxworks system has a passive PPP link, immediately "connected" and awaiting a client on our serial port. Id like a way to detect when a client connection comes and goes (authenticated and link ready for data packets vs. no client to send to). Is there a good way to do this? Thanks in advance --------------------------- Newsgroups: comp.os.vxworks Subject: Re: vxWorks BSP manual needed Date: Fri, 18 Oct 2002 16:34:32 -0400 From: joe durusau Organization: WEBUSENET.com Message-ID: <3DB07058.2183509B@bellsouth.net> References: vb wrote: > Hello, > > I wonder if anyone can share a document describing vxWorks BSP internals, > API, etc. > > Thanks in advance! > /vb I don't imagine that there is a doc about internals out there, but the reference manual describes all the public functions and the API. It should have come when you ordered your copy. If not, you should be able to get it off windsurf, but calling your FAE should get a printed copy, and they are much easier to use. Speaking only for myself, Joe Durusau --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Where are the shared library created C++ objects located Date: 18 Oct 2002 14:06:34 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0210181306.7b3e65f5@posting.google.com> References: Hello, > With regards to c++ objects created in a shared library function, is > the c++ object located in the domain where the shared library was > called from or is the c++ object in the located in the shared library? I answered this on your other thread, but to reiterate, the memory used to store the instance of each object created by the shared library code will come from the heap in the home domain of the task that allocated the memory (i.e. the application domain's heap). So, the objects could be considered to be created in the application domain. Something to note here is that this applies to the kernel too. If you add a routine to the kernel that calls malloc() to allocate some memory dynamically, when you call that routine from a user domain task, the memory will be allocated from the user domain's heap, and not from the kernel's heap. HTH, John... --------------------------- Newsgroups: comp.os.vxworks Subject: suspend functions using semTake() Date: 18 Oct 2002 14:38:14 -0700 From: jsanchor@cs5.dasd.honeywell.com (Jay) Organization: http://groups.google.com/ Message-ID: <9e60ca47.0210181338.79bd7307@posting.google.com> Is there a nicer way to do this? typedef struct _buffer_type { int cmd; int status; } buffer_type; buffer_type* ptrToBuffer = (buffer_type*)some_hardware_address; /*simplified: no one calls sendCommand() till it has returned status */ int sendCommand (int cmd) { ptrToBuffer->cmd = cmd; /* writes to hardware, generates interrupt*/ /* hardware processing command, so wait */ semTake(); return ptrToBuffer->status; } /* after hardware is done processing command, it puts results at status location and generates interrupt */ void intFromHardware(void) { semGive(); } basically, I want the command to block till it has status from hardware. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: 18 Oct 2002 15:48:07 -0700 From: brian_doe@hotmail.com (brian) Organization: http://groups.google.com/ Message-ID: References: <3DAAA37D.EAE13FD0@alcatel-ke.de> <3DAD2568.16C701E8@alcatel-ke.de> <3DAE93DC.21E25B1D@alcatel-ke.de> <3DAFBEE3.1EE916A5@alcatel-ke.de> > > > > Well in some cases you can GET a full OS/Stack etc that is ported to > > your processor. Sorry that was a mistype. What I meant to state was that in some cases you CANT GET anything that is ported to your processor. The processor we used had two choices : ATI and Wind River. And ATIs deal was much more restrictive. > > As you can from other vendors. In any case you used the Walkman > as an example. Why does a walkman need a TCP/IP or any other stack? > Surely, it just needs a basic mulit-tasking RTOS and little else? Well I suggest you think about that for a little longer. Think about how the music files get from the PC to the walkman. Yes. A USB stack. regards, Brian --------------------------- Newsgroups: comp.os.vxworks,comp.arch.embedded Subject: Re: Boot Code Integrity check with CRC Date: 18 Oct 2002 09:58:37 -0700 From: ryeung@earthlink.net (Raymond Yeung) Organization: http://groups.google.com/ Message-ID: References: You're quite right, Wolfgang. Google does return the links you mentioned as the 1st and 2nd links, now that I tried again. Previously, I tried my search on the advanced_group_search, which I suppose returns things from newsgroups only. My regular search engine is alta-vista. :) Perhaps I should start using Google in other aspect more.... Wolfgang Denk wrote in message news:... > ryeung@earthlink.net (Raymond Yeung) writes: > > >5. What is PPCBoot? I try searching for it on Googles, and got > > too many returns. Would appreciate it if one can provide a > > link to its tutorial? (Note: I'm currently booting using VxWorks > > boot rom. Is PPCBoot another boot rom?) > > Urrrrrghhh... > > If you google just for "PPCBoot", the very first link that turns up > is the PPCBoot home page at Sourceforge ("Embedded PowerPC Linux Boot > Project"), and the second link is the "PPCBoot-users Info Page". > > That was too many links for you to follow? > > Wolfgang Denk --------------------------- Newsgroups: comp.os.vxworks Subject: post-mortem in WindView Date: 18 Oct 2002 18:35:36 -0700 From: dennis_r_thomas@mail.northgrum.com (Denis Thomas) Organization: http://groups.google.com/ Message-ID: Does anyone have a good example of embedded code to collect post mortem events and get this data to a file. My host is DY4 PPC using vxworks 5.4, have windview capability. The example in WindView Users Guide is a little fuzzy. How do you setup the memory area to use, I don't see anything to pass in how big then memory area is to use for capture purposes. Does anyone have some real working code for setting up this capability. Anything is greatly appreciated. --------------------------- Newsgroups: comp.sys.powerpc.misc,comp.os.vxworks,gnu.gcc.help Subject: Re: Alignment problems in PPC Date: 18 Oct 2002 19:01:06 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0210181801.7feee7a5@posting.google.com> References: Hello, It is odd that you are asking to change the alignment in the compiler, but then state that the linker is showing you an address that is different to that reported by your debugger... Whatever the linker reported was presumably derived from the compiler's instructions on alignment! Have you tried disassembling the vxWorks image and looking at the instructions that load the parameters just before this call to strcpy()? I can't believe that the instructions are loading a value that is different to the reported address of the symbol. As to compiler flags, have you read the architecture supplement for PPC - the flags you need to use should be specified in there? That said, since you're building a kernel source file the build flags should be set by the project or BSP make file. You shouldn't change these without a very good reason ;-) HTH, John... FrankGleeson@motorola.com (frankg) wrote in message news:... > I am debugging a problem in our vxworks BSP whish is based on the > wrSbc8260 BSP. > > The following code fails to copy the string from tyName to > consoleName: > In File bootInit.c > > char consoleName [20]; /* line 578 global char array */ > > In Function usrRoot > strcpy (&consoleName[0], tyName); > > Using our visionclick debugger I was able to determine that the > address that was passed in as the dest > pointer was 006a3cf0 but the linker says that the address is 006a3cf4. > This is consistent with the state > of memory after the strcpy. The string is copied intact to 006a3cf0. > > > ccppc -B/home/copeng/project/semboot1.0.0/wind/host/sun4-solaris2/lib/gcc-lib/ > -mstrict-align -ansi -nostdinc > -I/home/copeng/project/semboot1.0.0/wind/target/config/sembsp_proj > -I/home/copeng/project/semboot1.0.0/wind/target/config/all > -I/home/copeng/project/semboot1.0.0/wind/target/h > -I/home/copeng/project/semboot1.0.0/wind/target/src/config > -I/home/copeng/project/semboot1.0.0/wind/target/src/drv > -I/home/copeng/project/semboot1.0.0/boot > -I/home/copeng/project/semboot1.0.0/com_inc > -I/home/copeng/project/semboot1.0.0/inc > -I/home/copeng/project/semboot1.0.0 > -I/home/copeng/project/semboot1.0.0/wind/target/h/arch/ppc > -DCPU=PPC603 -DSEM -g -O0 -g2 -gstabs -gxcoff -gdwarf -P -x > assembler-with-cpp -Wa,-L -c -o sysALib.o sysALib.s > > > We are using vxWorks 5.4.2. We are using the gnu compiler. > Do you have any idea what could cause this? Here is the actual > compiler command for this file. > Is there a compile flag that could influence the alignment? do we need > to put pragmas around > the consoleName variable declaration to force 8 byte alignment? If > it's an alignment issue > then it seems odd that the 8265 which is 603e based (32 bit > implementation) would require > 64 bit alignment. > > I noticed we are not using the -mcpu=603e or any of the -mpower > switches. Do you think this > would change gcc's allignment policy? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: vxWorks BSP manual needed Date: Sat, 19 Oct 2002 02:56:31 GMT From: "vb" Organization: AT&T Broadband Message-ID: References: <3DB07058.2183509B@bellsouth.net> Joe, thank you for your reply. In have to admit though that I do not have any relations with VxWorks - I used to do pSOS BSPs in the past and am pretty familiar with LynxOS BSP as of recent, just wanted to get some VxWorks BSP background as well. What is windsurf - is this a `customer only' service? Thank you again, /vb "joe durusau" wrote in message news:3DB07058.2183509B@bellsouth.net... > > > vb wrote: > > > Hello, > > > > I wonder if anyone can share a document describing vxWorks BSP internals, > > API, etc. > > > > Thanks in advance! > > /vb > > I don't imagine that there is a doc about internals out there, > but the reference manual describes all the public functions > and the API. It should have come when you ordered your copy. > If not, you should be able to get it off windsurf, but calling your FAE > should get a printed copy, and they are much easier to use. > > Speaking only for myself, > > Joe Durusau > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Telnet Time-Out Date: 18 Oct 2002 20:04:49 -0700 From: vdalal@hotmail.com (Vasu Dalal) Organization: http://groups.google.com/ Message-ID: <363d071b.0210181904.40197abb@posting.google.com> References: <5607ff2c.0210171017.4bf13d91@posting.google.com> gabi: > How can I set the Telnet timeout or there's any timeout in the Telnet > session? Try setting the tcp_keepidle variable to a value u like. i.e. 600 => 10 minutes hth, vasu --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Where are the shared library created C++ objects located Date: 18 Oct 2002 22:31:35 -0700 From: steven4443@yahoo.com (stephen hallacy) Organization: http://groups.google.com/ Message-ID: References: Het, the problem is that my company hasn't purchased tornado or vxworks yet, and I just have the help manuals. I plan to do a significant amount of tinkering when we get tornado and vxworks, and I'll try 'lkup' at that time too. Thanks John, sorry for posting the same question twice. I posted this one before you answered my first message. I'll try not to be so impatient next time. You've help me a quite a bit. Thanks Stephen steven4443@yahoo.com (stephen hallacy) wrote in message news:... > Hello, > > With regards to c++ objects created in a shared library function, is > the c++ object located in the domain where the shared library was > called from or is the c++ object in the located in the shared library? > > Thanks, > Stephen --------------------------- Newsgroups: comp.os.vxworks Subject: Re: suspend functions using semTake() Date: 19 Oct 2002 01:27:22 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0210190027.f52d3d4@posting.google.com> References: <9e60ca47.0210181338.79bd7307@posting.google.com> Hello, I'd say that assuming you were after synchronous behaviour (i.e. the task that requests the operation waits for its completion), then the solution you have is good. Be careful not to have multiple threads (tasks) calling sendCommand() though. If you think that is likely, you'd probably be advised to put a mutex semaphore around it as well: int sendCommand (int cmd) { int status; semTake (mutex, WAIT_FOREVER); /* Mutex semaphore */ ptrToBuffer->cmd = cmd; /* writes to hardware, generates interrupt*/ /* hardware processing command, so wait */ semTake(intRcvd, WAIT_FOREVER); /* Binary semaphore */ status = ptrToBuffer->status; semGive (mutex); return status; } Notice that the status is taken from the device inside the mutex locked region. HTH, John... jsanchor@cs5.dasd.honeywell.com (Jay) wrote in message news:<9e60ca47.0210181338.79bd7307@posting.google.com>... > Is there a nicer way to do this? > > typedef struct _buffer_type > { > int cmd; > int status; > } buffer_type; > > buffer_type* ptrToBuffer = (buffer_type*)some_hardware_address; > > /*simplified: no one calls sendCommand() till it has returned status */ > int sendCommand (int cmd) > { > ptrToBuffer->cmd = cmd; /* writes to hardware, generates interrupt*/ > /* hardware processing command, so wait */ > semTake(); > return ptrToBuffer->status; > } > > /* after hardware is done processing command, it puts results at > status location and generates interrupt */ > void intFromHardware(void) > { > semGive(); > } > > basically, I want the command to block till it has status from hardware. --------------------------- End of New-News digest ********************** From vxwexplo-errs@csg.lbl.gov Mon Oct 21 00:47:33 2002 From: "Sudhakar Tadi" Date: Mon Oct 21 00:47:35 PDT 2002 Subject: SCSIMgr question This is a multi-part message in MIME format. ------=_NextPart_000_00AF_01C27904.2DB657B0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Hi All, Please point me some other alias so that I can get suggestions for my = question Thanks in Advance -Tadi ----- Original Message -----=20 From: Sudhakar Tadi=20 To: Sudhakar Tadi ; the vxWorks Users Group Exploder ; = vxworks_users@csg.lbl.gov=20 Sent: Saturday, October 19, 2002 3:21 PM Subject: SCSIMgr question Just trying for one more time ----- Original Message -----=20 From: Sudhakar Tadi=20 To: the vxWorks Users Group Exploder ; vxworks_users@csg.lbl.gov=20 Cc: Sudhakar Tadi=20 Sent: Friday, October 18, 2002 4:44 PM Subject: SCSIMgr question. Hi All, I need some clarification regarding SCSI drivers. As per the specs = there is one SCSIMgr taks for scsi controller, that's fine. But what = happens If I have the two similar cards in the same machine, do I need = to create seperate SCSIMgr task or one SCSImgr task enough for managing = the two cards. My question is can single SCSImgr task handles the two = similar type of HBA cards. Can somebody clarify. Thanks in Advance -Tadi Sudhakar Tadi HCL Technologies Ltd 184-188, NSK Road,Vadapalani Chennai-600 026,INDIA Ph:91-44-372 8366 Ext:1134 Fax:91-44-480 6640 http://san.hcltech.com ------=_NextPart_000_00AF_01C27904.2DB657B0 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
Hi All,
 
  Please point me some other alias = so that I=20 can get suggestions for my question
 
Thanks in Advance
-Tadi
----- Original Message -----
From:=20 Sudhakar=20 Tadi
To: Sudhakar Tadi ; the vxWorks = Users Group=20 Exploder ; vxworks_users@csg.lbl.gov =
Sent: Saturday, October 19, = 2002 3:21=20 PM
Subject: SCSIMgr question

Just trying for one more time
 
 
----- Original Message -----
From:=20 Sudhakar=20 Tadi
To: the vxWorks Users Group = Exploder ; vxworks_users@csg.lbl.gov =
Sent: Friday, October 18, = 2002 4:44=20 PM
Subject: SCSIMgr = question.

Hi All,
I need some clarification regarding = SCSI=20 drivers.  As per the specs there is one SCSIMgr taks for scsi=20 controller, that's fine.  But what happens If I have the two = similar=20 cards in the same machine, do I need to create seperate SCSIMgr task = or one=20 SCSImgr task enough for managing the two cards. My question is can = single=20 SCSImgr task handles the two similar type of HBA cards. Can somebody = clarify.
 
Thanks in Advance
-Tadi
Sudhakar Tadi
HCL Technologies=20 Ltd
184-188, NSK Road,Vadapalani
Chennai-600 = 026,INDIA
Ph:91-44-372=20 8366 Ext:1134
Fax:91-44-480 6640
http://san.hcltech.com
------=_NextPart_000_00AF_01C27904.2DB657B0-- From vxwexplo-errs@csg.lbl.gov Mon Oct 21 04:03:16 2002 From: Vxworks Exploder Date: Mon Oct 21 04:03:18 PDT 2002 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Mon Oct 21 04:03:14 PDT 2002 Subject: Obects / ringbuffer in userreserved memory Subject: Re: suspend functions using semTake() Subject: Re: Alignment problems in PPC Subject: Re: vxWorks BSP manual needed Subject: Re: post-mortem in WindView Subject: baby Question Subject: Re: baby Question Subject: CP7000 and DMA Subject: Re: taskRestart Subject: Re: Porting of vxWorks. Subject: USB camera in vxworks ? Re: USB Subject: Re: strange timer bahaviour Subject: Re: strange timer bahaviour Subject: File System in VxWorks ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Obects / ringbuffer in userreserved memory Date: 20 Oct 2002 13:29:38 -0700 From: derfan2001@altavista.de (Mario Schattschneider) Organization: http://groups.google.com/ Message-ID: <13dfb019.0210201229.5cb26603@posting.google.com> Hello, I would like to create a ringbuffer in userreserved memory. The ringbuffer should countain Objects - my question is how do I create a variable or an object in userreserved memory? I think I must write my own ringbuffer because I can't create a vxWorks ringbuffer in userreserved memory. The buffer must be in the userreserved memory because the data in the buffer should be available after a reboot of vxWorks. I hope you could help me with my problem to create a variable or a Object in userreserved memory. Best regards Mario Schattschneider --------------------------- Newsgroups: comp.os.vxworks Subject: Re: suspend functions using semTake() Date: Sat, 19 Oct 2002 09:54:09 -0700 From: "Michael R. Kesti" Organization: MK Associates Message-ID: <3DB18E31.E84A544C@gv.net> References: <9e60ca47.0210181338.79bd7307@posting.google.com> Reply-To: mkesti@gv.net Jay wrote: >Is there a nicer way to do this? > >typedef struct _buffer_type >{ >int cmd; >int status; >} buffer_type; > >buffer_type* ptrToBuffer = (buffer_type*)some_hardware_address; > >/*simplified: no one calls sendCommand() till it has returned status */ >int sendCommand (int cmd) >{ > ptrToBuffer->cmd = cmd; /* writes to hardware, generates interrupt*/ > /* hardware processing command, so wait */ > semTake(); > return ptrToBuffer->status; >} > >/* after hardware is done processing command, it puts results at >status location and generates interrupt */ >void intFromHardware(void) >{ > semGive(); >} > >basically, I want the command to block till it has status from hardware. If I understand what you're seeking, this is called "process syncronization" and semaphores are intended for exactly this, as well as other things. You're on the correct path, Jay. - -- ======================================================================== Michael Kesti | "And like, one and one don't make | two, one and one make one." mkesti@gv.net | - The Who, Bargain --------------------------- Newsgroups: comp.sys.powerpc.misc,comp.os.vxworks,gnu.gcc.help Subject: Re: Alignment problems in PPC Date: 20 Oct 2002 14:46:37 -0700 From: FrankGleeson@motorola.com (frankg) Organization: http://groups.google.com/ Message-ID: References: <488e459a.0210181801.7feee7a5@posting.google.com> Thanks for your response. The problem I saw in VisionClick was a problem with the way I created my symbol file. The problem is I am trying to debug a compressed vxworks bootrom image. In our case the bootrom is linked high with means that the -Ttext flag is passed to the linker giving it the address of destination on the uncompressed image in ram. When I created my symbol file I found that all of the symbols were offset by the size of the initialization code romInit.s that is never compressed. I was able to generate working symbol files by creating a seperate target that passed a different offset so that compressedEntry was at the destination address in ram. The function compressedEntry() is the first thing in the inflate-ed image and must reside at the target address. The problem with these "corrected" symbol files was that no correction was made for the data or bss segments. As a result global variables were offset in the symbol file by the amount of data and bss space required by the noncompressed initialization code. I got confused by this and though gcc was generating incorrect machine code. When actually it was entirely a symbol-file/tools issue. After reflecting on this I think I would get better results by creating a build that ommits the uncompressed initialization code with - -Ttext=DEST_RAM and feed that to convert to get my symbol file. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: vxWorks BSP manual needed Date: Sat, 19 Oct 2002 15:09:09 -0400 From: joe durusau Organization: WEBUSENET.com Message-ID: <3DB1ADD4.7E6E3B3B@bellsouth.net> References: <3DB07058.2183509B@bellsouth.net> It's on the WRS website, and most if not all of it is customers only. There are some guides about porting BSPs available from WRS without having a development license, but I don;t think anything is free. You could call the local rep and see what they have to offer and what the prices are. Speaking only for myself, Joe Durusau vb wrote: > Joe, thank you for your reply. > > In have to admit though that I do not have any relations with VxWorks - I > used to do pSOS BSPs in the past and am pretty familiar with LynxOS BSP as > of recent, just wanted to get some VxWorks BSP background as well. > > What is windsurf - is this a `customer only' service? > > Thank you again, > /vb > "joe durusau" wrote in message > news:3DB07058.2183509B@bellsouth.net... > > > > > > vb wrote: > > > > > Hello, > > > > > > I wonder if anyone can share a document describing vxWorks BSP > internals, > > > API, etc. > > > > > > Thanks in advance! > > > /vb > > > > I don't imagine that there is a doc about internals out there, > > but the reference manual describes all the public functions > > and the API. It should have come when you ordered your copy. > > If not, you should be able to get it off windsurf, but calling your FAE > > should get a printed copy, and they are much easier to use. > > > > Speaking only for myself, > > > > Joe Durusau > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: post-mortem in WindView Date: Sat, 19 Oct 2002 19:35:22 GMT From: ggg+news@best.com (GGG) Organization: rest Message-ID: <_tis9.19757$T_.468540@iad-read.news.verio.net> References: In post-mortem WindView configuration you specify sysMemTop() and sysPhysMemTop(). The difference between the two is the size, USER_RESERVED_MEM. A correctly made BSP should allow you to specify USER_RESERVED_MEM as an amount of memory to keep at the top of available RAM that doesn't go into the system memory pool. Your BSP's sysMemTop should return the bottom of this area and sysPhysMemTop the top. See sections 4.4, 8.1 of the WindView UG. As far as initializing the upload without using the GUI, see the tech tip on the support site, "Using Windview for post-mortem" to enhance what is in the WV UG, 4.4. It shows how to use rBuffCreate and specify the start and size of the area of memory for initializing the buffers. The result of rBuffCreate is what is used as an argument to wvEvtLogInit. GGG In article , Denis Thomas wrote: >Does anyone have a good example of embedded code to collect post >mortem events and get this data to a file. My host is DY4 PPC using >vxworks 5.4, have windview capability. The example in WindView Users >Guide is a little fuzzy. How do you setup the memory area to use, I >don't see anything to pass in how big then memory area is to use for >capture purposes. > >Does anyone have some real working code for setting up this >capability. > >Anything is greatly appreciated. --------------------------- Newsgroups: comp.os.vxworks Subject: baby Question Date: Mon, 21 Oct 2002 04:26:54 +0000 (UTC) From: kwjlee@lgcns.com (charley brown) Organization: Korea Research Environment Open Network Message-ID: Hi everyone ! What is Vx-Works? - -concepts or definitions - -Use case - -Sample Applications Please answer the baby Questions.... I am a beginner.... Thanks a lot --------------------------- Newsgroups: comp.os.vxworks Subject: Re: baby Question Date: Mon, 21 Oct 2002 04:31:26 GMT From: nunya@damn.business.com (Bruce) Organization: Prodigy Internet http://www.prodigy.com Message-ID: <3db38300.12761429@news.houston.sbcglobal.net> References: In comp.os.vxworks kwjlee@lgcns.com (charley brown) wrote: >What is Vx-Works? >-concepts or definitions >-Use case >-Sample Applications > >Please answer the baby Questions.... > >I am a beginner.... google.com vxworks 133,000 hits --------------------------- Newsgroups: comp.os.vxworks Subject: CP7000 and DMA Date: 20 Oct 2002 23:43:24 -0700 From: shenoy_sudhakar@hotmail.com (Sudhakar Shenoy) Organization: http://groups.google.com/ Message-ID: <4d53234f.0210202243.2c15e9b0@posting.google.com> Hi, we have a Ocelot board with R4000 Mips processor. the board supports 2 PMC slots on it along with the Galileo system controller.we have a custom made PMC card. if i want to do DMA some data using the gtDma() (galileo dma routine provided by the BSP of ocelot ,cp7000), if the Data is written in memory allocated through "cacheDmaMalloc", it works fine. but if i try writing the same data just by using the "malloc" and DMAing it, it doenst respond. is it necessary to do some other setting for the same to work?? please let me know, thanks n regds, Sudhakar Shenoy. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: taskRestart Date: 21 Oct 2002 08:22:13 GMT From: Denis Perelyubskiy Message-ID: References: <6847c39a.0210120342.91911fc@posting.google.com> * Vittal [10-Sat-02 04:42 -0700]: > > Hi All, I am facing a problem with taskRestart. > I am having two tasks task1 and task2 with the same > priority of 0. I am trying to restart task1 from task2. > The task1 restarts properly. However, there is a > problem. I am trying to check the return value of > taskRestart and print appropritate message. But the > message never appears even though the task is restarted. > I replaced the printf with logMsg and the results were > same. I also replaced the printf/logMsg with a global > variable and tried to check the value of the global > variable from the shell. But the global variable's value > was unchanged. > Something like this.... > restartCount = 0; /* global count */ > code snippet from task2.... > if (OK == taskRestart (taskNameToId ("task1")) > restartCount = 1; > else > restartCount = 2; > After the execution of this, task1 is restarted. I have > confirmed this with some prints and global variables. > However, the restartCount remains 0. I expected it to be > 1 since the task1 was restarted. If there was some > problem, at least it should have got changed to 2. But, > the restarCount remains 0 and this is something which I > am not able to understand. > I am not able to connect the debugger due to some > problems with the network and hence not able to debug it > properly. If you ask wrs, they'll tell you that running anything with priorities less (in a numerical sense) then tnettask is not supported by them. It may have something to do with the splnet() semaphore, taken by tNetTask for processing any incoming packet. You may not be aware of it, but deep down below calls like write() to a socket take the same semaphore. If you restart the task while holding on to the same semaphore..... btw, print() deep down below does a write to a socket. takes splnet() semaphore. Not sure about the logmsg either try writing to vga or try setting something in user-reserved memory and checking with debugger or something. But if your tasks are doing something weird, your debugger may not work due to priority problems, I think. I am not sure about this last one though. denis - -- 'From' email address is used as a sink. Not read. Ever. Instead, send to [t-h-e-u-k-r-a-i-n-i-a-n|h-o-t-m-a-i-l|c-o-m] (remove dashes, replace the first vertical bar with @, and second with a dot) --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Porting of vxWorks. Date: Mon, 21 Oct 2002 10:42:32 +0200 From: Tim Sinkins Organization: Alcatel Kommunikations Elektronik Message-ID: <3DB3BDF8.B4C35992@alcatel-ke.de> References: <3DAAA37D.EAE13FD0@alcatel-ke.de> <3DAD2568.16C701E8@alcatel-ke.de> <3DAE93DC.21E25B1D@alcatel-ke.de> <3DAFBEE3.1EE916A5@alcatel-ke.de> Reply-To: timothy.sinkins@alcatel-ke.de brian wrote: > > > > > > > Well in some cases you can GET a full OS/Stack etc that is ported to > > > your processor. > > Sorry that was a mistype. What I meant to state was that in some cases > you CANT GET anything that is ported to your processor. The processor > we used had two choices : ATI and Wind River. And ATIs deal was much > more restrictive. So choose another processor. No this is not as stupid as it sounds. It is quite common for a project hardware team to choose specific hardware just to get a few cents saving without reference to the software costs. When you tell them that the required software isn't available they say, "well you can write it, it doesn't cost anything, it's only software!". Most companies I have worked for think this way. > > As you can from other vendors. In any case you used the Walkman > > as an example. Why does a walkman need a TCP/IP or any other stack? > > Surely, it just needs a basic mulit-tasking RTOS and little else? > > Well I suggest you think about that for a little longer. Think about > how the music files get from the PC to the walkman. We'll I was unaware that there was a 'walkman' that did this . I was thinking in terms of playback of CDs etc. > Yes. A USB stack. So write it yourself. It can't be that difficult (I know nothing about the USB spec but I I do do this sort of software professionally and it really can't be that difficult to write just the User end of the USB protocol). In any case I consider this to be part of the application. Tim > > regards, > > Brian --------------------------- Newsgroups: comp.os.vxworks Subject: USB camera in vxworks ? Re: USB Date: Mon, 21 Oct 2002 09:04:40 GMT From: "Mike Kim" Organization: Dacom Message-ID: References: Hi , I'd like to use a usb camera for image capturing. Because Someone asked me if it is possible to capture image through usb camera. I'm using vxworks 5.4 and zinc 5.x. But using USB ddk is too much expensive. So I want to know if is there any another way to capture image in vxworks environment. Thank you in-advance. "Ravi Uday" wrote in message news:ecca3ca6.0210162128.c36f98c@posting.google.com... > Hi, > > Can anybody tell me which is the best discussion site for discussing > issues regarding writing driver for USB on VxWorks O.S. I already have > the VxWorks USB DDK with me and have few questions of using / begin > with it. > My device is DiskOnKey. > Thanks for all kind of helps/experience > > - Ravi --------------------------- Newsgroups: comp.os.vxworks Subject: Re: strange timer bahaviour Date: Mon, 21 Oct 2002 11:21:13 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1035192099.254012@newsreader2.wirehub.nl> References: <1034839098.567517@newsreader1.wirehub.nl> <1034867481.955098@newsreader2.wirehub.nl> <29b153d9.0210180725.2c8f96b@posting.google.com> Hi Richard, Thanx about the remark, but: sys/types.h: (vxworks) #ifndef _TIME_T_ #define _TIME_T_ typedef long time_t; #endif linux (bits/types.h) typedef long int __time_t; So I dont agree with you. time_t is signed! But to the other question, yes this is intended. what we try to do is: say one has a periodic timer with period times of 4 seconds, and our system requires it to be changed (which is a part of the simulation application). Let say we want it to become 5 seconds. now one can do two things. 1) set the period time to the new value "wait" for the current period to end then start with the new periode time of 5 seconds. 2) another solution, if you are in a period, one could try to "modify" the current period to make it fit in the new period time. I know this might not aways possible, especially if one reduces the period time. We thought might be a nice thing to do because we want to set the new period values as soon as possible. In our simulation project we like to have the new period time as soon as possible. But is this a legal thing to do? to "modify" the time a timer has left, using th it_value parameter? Cheers, Renee - - - - - - - - - Renee A. Teunissen PTS Software bv, Meerweg 7, 1405BA Bussum, NL. T.+31-(0)35-6926969, M.+31-(0)6-22778313, http://www.pts.nl, @pts.nl personal link page: http://wittenburg10c.nl/db/dest/links.html - ----- Original Message ----- From: "Richard Nardone" Newsgroups: comp.os.vxworks Sent: Friday, October 18, 2002 5:25 PM Subject: Re: strange timer bahaviour > "> > > void > > timDemoGetTest(void) > > { > > static int prevTick = 0; > > int now; > > > > while(1) > > { > > taskDelay(1); > > > > timer_gettime(TimerId, &iSpecNow); > > > > if(iSpecNow.it_value.tv_sec < 0) > > { > > now = tickGet(); > > > > > Renee, > > tv_sec is of type unsigned long. You may want to change this > comparison to see if it is equal to zero. You may want to read the > timerLib manual a bit more closely. > > Also, you need to understand that the fundamental resolution of > your timer is tied to the system clock interval (i.e. tick) which is > usually around 60Hz or 16.66 ms. Nanosecond resolution is not > possible with this timer facility. > > Also, you seem to be setting both an interval timer, and a one shot > timer at the same time. Is this intentional? > > > Richard Nardone > Cyclone Microsystems --------------------------- Newsgroups: comp.os.vxworks Subject: File System in VxWorks Date: 21 Oct 2002 03:51:20 -0700 From: technokiran@yahoo.com (Kiran Kumar) Organization: http://groups.google.com/ Message-ID: <552e79d.0210210251.a1ac354@posting.google.com> Hi, I am new to Vxworks. I am facing problems when I was trying to create a file in the VxWorks application. Can anyone suggest, how to handle files in Vxworks. Thanks in advance kiran. --------------------------- End of New-News digest ********************** From vxwexplo-errs@csg.lbl.gov Mon Oct 21 07:10:24 2002 From: Mike McCullough Date: Mon Oct 21 07:10:25 PDT 2002 Subject: Re: Obects / ringbuffer in userreserved memory (VxWorks) Mario, AFAIK you can still do a memPartAddToPool/memPartCreate using an address in user-reserved memory from your application code. Only difference is exactly what you want which is VxWorks won't clear the memory on a reset... Guys, any comments?? Seems straightforward to me... Newsgroups: comp.os.vxworks Subject: Obects / ringbuffer in userreserved memory Date: 20 Oct 2002 13:29:38 -0700 From: derfan2001@altavista.de (Mario Schattschneider) Organization: http://groups.google.com/ Message-ID: <13dfb019.0210201229.5cb26603@posting.google.com> Hello, I would like to create a ringbuffer in userreserved memory. The ringbuffer should countain Objects - my question is how do I create a variable or an object in userreserved memory? I think I must write my own ringbuffer because I can't create a vxWorks ringbuffer in userreserved memory. The buffer must be in the userreserved memory because the data in the buffer should be available after a reboot of vxWorks. I hope you could help me with my problem to create a variable or a Object in userreserved memory. Best regards Mario Schattschneider --------------------------- ********************************************************** Mike McCullough MCC Engineering President and CEO 9 Cheshire Court Londonderry, NH 03053 mike@mccengineering.com Tel: 603-537-9593 Fax: 603-434-9890 www.mccengineering.com Cell: 603-247-1219 ********************************************************** 24x7 VxWorks Support Custom Support Programs VxWorks Training Embedded Consulting Services A WindLink Partner ********************************************************** From vxwexplo-errs@csg.lbl.gov Mon Oct 21 22:37:34 2002 From: "Srinivasa Murthy A" Date: Mon Oct 21 22:37:35 PDT 2002 srinivasa.murthy@wipro.com Subject: I would like to join the group X-Mailer: Netscape Messenger Express 3.5.2 [Mozilla/4.51 [en] (X11; I; SunOS 5.8 sun4u)] Hi, I would like to join the VxWorks Users Group. Here is my email id: srinivasa.murthy@wipro.com Thank you. regards, Srinivasa Murthy begin 666 Wipro_Disclaimer.txt M*BHJ*BHJ*BHJ*BHJ*BHJ*BHJ*BHJ*BHJ*BI$:7-C;&%I;65R*BHJ*BHJ*BHJ M*BHJ*BHJ*BHJ*BHJ*BHJ*BHJ*BHJ*BHJ*BHJ*BHJ*BHJ*BHJ*BHJ*BH@(" @ M#0H@#0H@26YF;W)M871I;VX@8V]N=&%I;F5D(&EN('1H:7,@12U-04E,(&)E M:6YG('!R;W!R:65T87)Y('1O(%=I<')O($QI;6ET960@:7,@)W!R:79I;&5G M960G( T*86YD("=C;VYF:61E;G1I86PG(&%N9"!I;G1E;F1E9"!F;W(@=7-E M(&]N;'D@8GD@=&AE(&EN9&EV:61U86P@;W(@96YT:71Y('1O('=H:6-H(&ET M(&ES( T*861D6EN9R!O2!M86YN97(@=VAA M='-O979E2!P Date: Mon Oct 21 22:38:26 PDT 2002 Subject: I would like to join the group Hi, I would like to join the VxWorks Users Group. Here is my email id: srinivasa.murthy@wipro.com Thank you. regards, Srinivasa Murthy begin 666 Wipro_Disclaimer.txt M*BHJ*BHJ*BHJ*BHJ*BHJ*BHJ*BHJ*BHJ*BI$:7-C;&%I;65R*BHJ*BHJ*BHJ M*BHJ*BHJ*BHJ*BHJ*BHJ*BHJ*BHJ*BHJ*BHJ*BHJ*BHJ*BHJ*BHJ*BH@(" @ M#0H@#0H@26YF;W)M871I;VX@8V]N=&%I;F5D(&EN('1H:7,@12U-04E,(&)E M:6YG('!R;W!R:65T87)Y('1O(%=I<')O($QI;6ET960@:7,@)W!R:79I;&5G M960G( T*86YD("=C;VYF:61E;G1I86PG(&%N9"!I;G1E;F1E9"!F;W(@=7-E M(&]N;'D@8GD@=&AE(&EN9&EV:61U86P@;W(@96YT:71Y('1O('=H:6-H(&ET M(&ES( T*861D6EN9R!O2!M86YN97(@=VAA M='-O979E2!P Date: Tue Oct 22 04:03:16 PDT 2002 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Tue Oct 22 04:03:12 PDT 2002 Subject: Loading program slow using ld Subject: Re: Loading program slow using ld Subject: Re: How to force asynchronous read() to finish in time? Subject: Re: Is Thread ID unique through the session? Subject: Re: File System in VxWorks Subject: Re: arptnew failed Subject: Re: arptnew failed Subject: duplicate/truncated filenames in Flash Subject: vision ice question Subject: Re: Telnet Time-Out Subject: Re: Obects / ringbuffer in userreserved memory Subject: dosFs2.2 is printing junk during init? Subject: Re: baby Question Subject: Is Thread ID unique through the session? Subject: USB camera in vxworks ? + cost if needed + method outline Subject: Re: USB camera in vxworks ? Re: USB Subject: BSP problem running vx Bootloader without ICE Subject: Re: strange timer bahaviour Subject: profilescope doesn't initialize... Subject: Re: Multicast driver sample Subject: remote object file loading ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Loading program slow using ld Date: 21 Oct 2002 04:41:42 -0700 From: damiandixon@netscape.net (Damain) Organization: http://groups.google.com/ Message-ID: Hi, Is there any way of downloading a program which is quicker then using ld. It's just taking a very long time to download a 20MB program, via WindSh. I am using: VxWorks 5.4 Sun Solaris 8 (host) Thanks Damian --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Loading program slow using ld Date: Mon, 21 Oct 2002 14:55:45 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Reply-To: "Leonid Rosenboim" "Damain" wrote in message news:b6801b8c.0210210341.66c76141@posting.google.com... > Hi, > > Is there any way of downloading a program which is quicker then using ld. > > It's just taking a very long time to download a 20MB program, via WindSh. > > I am using: > VxWorks 5.4 > Sun Solaris 8 (host) > > Thanks > Damian Perhaps the quickest way to load a large application is by statically linking it with VxWorks, so when VxWorks boots, the application is already in the target's memory. The boot time of a combined application and VxWorks run-time will be many times faster, at the cost of somewhat longer build time. There is also a mid-way solution, where those parts of the application which do not change frequently are statically linked with VxWorks, while thos parts you are currently developing are dynamically loaded, so the longer build time has lesser significance because it is infrequently performed, and the dynamic loading tiem is much shorted due to smaller size of the loaded object. - -- - ----------------------------------------------------------------------- Leonid Rosenboim Visit: http://gamla.org.il/english/index.htm Consultant Email: my first name at consultant dot com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How to force asynchronous read() to finish in time? Date: Mon, 21 Oct 2002 16:00:37 +0200 From: natklof@hotmail.com Message-ID: References: <4f57a77f.0210171523.208d1ffd@posting.google.com> "Leonid Rosenboim" wrote: >"Liang Yang" wrote in message >news:4f57a77f.0210171523.208d1ffd@posting.google.com... >> Hi, >> >> I need to call TCP socket read() in the target side(MBX 860) >> to read the data from the server(Windows 2000) side. But >> since read() is asynchronous operation, the task which >> calls read() will yield the control to other tasks after >> it calls read(). So the code to check the data which is >> put after read() shows the data is still not ready. >> >> How can I ensure data is ready after I call read()? > >You are very confused, but since you are a student (posting >from a University IP address), I guess that is to be expected. > >If you have set the socket into non-blocking mode, read() >will return immediately, weather data is available or not. >If the socket is in its default blocking mode, read will >WAIT (thereby allowing other tasks to use the CPU) >until SOME data is available, and upon return will >report the number of bytes retreived. > >So the code that is executed after read() returns should >simply check how many butes read() returned, and that >amount reported by read() has been indeed copied to your >buffer and is perfectly ready for further use. > >HTH > Leonid > Happy Birthday, Leonid! AR --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Is Thread ID unique through the session? Date: Mon, 21 Oct 2002 16:32:01 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <6839b7a5.0210210459.5455a8f4@posting.google.com> Reply-To: "Leonid Rosenboim" "Andrei Sedelnikov" wrote in message news:6839b7a5.0210210459.5455a8f4@posting.google.com... > Hallo everybody, > > I have a rather simple but probably very specific question: > > Can I rely upon that, that ThreadIDs (TIDs) are unique through > the application's session, or they are being always reused, > after the thread was killed? > > Thank you! You mean tid = TaskId : Yeah, taskId may be reused after a task is terminated, due to a possible reuse of the same memory allocated for the Task Control Block structure. It is possible to avoid TID reuse if it is REALLY IMPORTANT. - -- Contact me directly if you need more details. - ----------------------------------------------------------------------- Leonid Rosenboim Visit: http://gamla.org.il/english/index.htm Consultant Email: my first name at consultant dot com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: File System in VxWorks Date: Mon, 21 Oct 2002 09:32:44 -0700 From: "Michael R. Kesti" Organization: MK Associates Message-ID: <3DB42C2C.EC9F16A4@gv.net> References: <552e79d.0210210251.a1ac354@posting.google.com> Reply-To: mkesti@gv.net Kiran Kumar wrote: >Hi, > I am new to Vxworks. I am facing problems when I was trying to >create a file in the VxWorks application. Can anyone suggest, how to >handle files in Vxworks. > >Thanks in advance >kiran. See creat() and other functions in the reference manual's ioLib section. - -- ======================================================================== Michael Kesti | "And like, one and one don't make | two, one and one make one." mkesti@gv.net | - The Who, Bargain --------------------------- Newsgroups: comp.os.vxworks Subject: Re: arptnew failed Date: Mon, 21 Oct 2002 12:55:01 -0500 From: Andrew Johnson Organization: APS, Argonne National Laboratory Message-ID: References: somebody wrote: > Hi Does any one know why I get this msg > 0xf7bf88 (tNetTask): arptnew failed on 50f22423 > 0xf7bf88 (tNetTask): arptnew failed on 50f22423 > 0xf7bf88 (tNetTask): arptnew failed on 50f22423 > Does it mean that my network stack has run out of memory. No. The 50f22423 number translates to the IP address 80.242.36.35 which is relevent to the problem and which you may recognize. It may be that you're getting packets from that address which is outside of the subnet that the vxWorks system is in, or that you don't have the right routing set up to get to that address, or that your subnet mask is incorrectly set on one of your interfaces. It's not a terribly helpful error message, but those are the kinds of things that have caused it in my experience. Does anybody else know anything more? - - Andrew - -- "Life is what happens while you're busy making other plans." - John Lennon --------------------------- Newsgroups: comp.os.vxworks Subject: Re: arptnew failed Date: Mon, 21 Oct 2002 21:15:50 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Reply-To: "Leonid Rosenboim" "Andrew Johnson" wrote in message news:ap1f13$uc4$1@milo.mcs.anl.gov... > somebody wrote: > > > Hi Does any one know why I get this msg > > 0xf7bf88 (tNetTask): arptnew failed on 50f22423 > > 0xf7bf88 (tNetTask): arptnew failed on 50f22423 > > 0xf7bf88 (tNetTask): arptnew failed on 50f22423 > > Does it mean that my network stack has run out of memory. > > No. The 50f22423 number translates to the IP address 80.242.36.35 which > is relevent to the problem and which you may recognize. It may be that > you're getting packets from that address which is outside of the subnet > that the vxWorks system is in, or that you don't have the right routing > set up to get to that address, or that your subnet mask is incorrectly set > on one of your interfaces. It's not a terribly helpful error message, but > those are the kinds of things that have caused it in my experience. > > Does anybody else know anything more? > > - Andrew Andrew, I will try to explain this from another perspective: An IP packet is being transmitted, and based upon how it matches network interfaces' IP addresses and mask, an outgoing interface is selected, at which point the interface driver calls arpresolve to get a MAC address for that IP address, either by sending an ARP request, or by using a cached entry. At that point, if the ARP code, looking for an ARP cache entry, finds that this particular IP address has a specific routing entry which is not an ARP entry, it will return that particular error. Since ARP uses has a common emcnahism with the IP routing table, there could not be a routing etnry and an ARP entry for the same destination IP address at the same time. One possible reason for this is if an ICMP Redirect is received for a particuar IP address, which would be considered "local" if mathcing its address to the network interface's IP address and mask. mRouteShow() at the time of this error should pretty much paint the picture in bright colour. - - Leonid --------------------------- Newsgroups: comp.os.vxworks Subject: duplicate/truncated filenames in Flash Date: Mon, 21 Oct 2002 15:34:41 -0700 From: "kdb" Message-ID: <6i%s9.260$B11.66292@news.uswest.net> Every once in a while my Flash filesystem will show duplicate and/or truncated filenames. E.g.: what was once "Logs" will now be simply "L", what was "myfile" is now "m". Or I may have multiple files all named "myfile". Any idea what's going on here and how to fix it? This happens with directories or normal files. In both duplicates and truncations it does preserve the upper/lower case of the name. The file or directory contents remain intact, it's just the names that are hosed. System env: PowerPC 8245 (603-based, not using any floating point to my knowledge but I could be wrong), VxWorks 5.4.2, Tornado 2.0.2, dosFs 2.0, TrueFFS. thanks in advance - K --------------------------- Newsgroups: comp.os.vxworks Subject: vision ice question Date: Mon, 21 Oct 2002 23:05:47 GMT From: "Jim" Organization: Magma Communications Ltd. Message-ID: hello i am using vision ice 2 ... i need a way such that the vision ice doesn't assert control over the registers SRR0 and SRR1 so I can debug some asm code calling rfi ... Can any one help? Jim --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Telnet Time-Out Date: 20 Oct 2002 05:31:55 -0700 From: gabi@tdsoft.com (Gabi) Organization: http://groups.google.com/ Message-ID: <5607ff2c.0210200431.6c90933f@posting.google.com> References: <5607ff2c.0210171017.4bf13d91@posting.google.com> <363d071b.0210181904.40197abb@posting.google.com> vdalal@hotmail.com (Vasu Dalal) wrote in message news:<363d071b.0210181904.40197abb@posting.google.com>... > gabi: > > > How can I set the Telnet timeout or there's any timeout in the Telnet > > session? > > Try setting the tcp_keepidle variable to a value u like. > i.e. 600 => 10 minutes > > hth, > vasu Hi Vasu, The TCP_IDLE_TIMEO_DFLT set to 60 (approx. 2min), but this is not what I'm looking for. I want to limit the Telnet session, incase of stop typing, for a constant period. Is this possible. tanks Gabi --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Obects / ringbuffer in userreserved memory Date: 21 Oct 2002 08:04:11 GMT From: Denis Perelyubskiy Message-ID: References: <13dfb019.0210201229.5cb26603@posting.google.com> Mario, * Mario Schattschneider [10-Sun-02 13:29 -0700]: > > I would like to create a ringbuffer in userreserved > memory. The ringbuffer should countain Objects - my > question is how do I create a variable or an object in > userreserved memory? I think I must write my own > ringbuffer because I can't create a vxWorks ringbuffer in > userreserved memory. perhaps the following will help? http://groups.google.com/groups?hl=en&lr=&ie=UTF-8&oe=UTF-8&th=173c969039271f77&seekm=slrnamoc6t.55m.denis_newsREMOVE-TO-REPLY%40concorde.cs.ucla.edu&frame=off denis - -- 'From' email address is used as a sink. Not read. Ever. Instead, send to [t-h-e-u-k-r-a-i-n-i-a-n|h-o-t-m-a-i-l|c-o-m] (remove dashes, replace the first vertical bar with @, and second with a dot) --------------------------- Newsgroups: comp.os.vxworks Subject: dosFs2.2 is printing junk during init? Date: 21 Oct 2002 18:11:59 -0700 From: roswel_ajf@hotmail.com (roswell) Organization: http://groups.google.com/ Message-ID: <6dcae13d.0210211711.6804a30d@posting.google.com> hi.. how can it get rid off these junk printed by dosFs2.x. Roswell - ------------------------- Retrieved old volume params with %33 confidence: Volume Parameters: FAT type: FAT32, sectors per cluster 0 0 FAT copies, 0 clusters, 0 sectors per FAT Sectors reserved 0, hidden 0, FAT sectors 0 Root dir entries 0, sysId (null) , serial number 0 Label:" " ... Disk with 16384 sectors of 512 bytes will be formatted with: Volume Parameters: FAT type: FAT16, sectors per cluster 2 2 FAT copies, 8143 clusters, 32 sectors per FAT Sectors reserved 1, hidden 0, FAT sectors 64 Root dir entries 512, sysId VXDOS16 , serial number 0 Label:" --------------------------- Newsgroups: comp.os.vxworks Subject: Re: baby Question Date: Mon, 21 Oct 2002 14:50:44 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Reply-To: "Leonid Rosenboim" Charley, VxWorks is a real-time operating system and run-time enviroment, which sells as part of the Tornado development environment. VxWorks has initially grown out of the scientific, industrial and military/aerospace real-time application, and in recent time has been used in many deeply embedded applications including consumer electronics, medical and telecommunications equipment. Large scale applications I am familiar woth that use VxWorks include large capital equipment for the semiconductor industry such as those of Applied Materials Inc., and large digital printers such as those made by Indigo (now part of H-P digital imaging division). Telecom applications include large optical switches, such as those from Ciena Corp, and cellular infrastructure equipment from Lucent and Qualcomm, to small devices like enterprise managed Layer2/3 switches, and digital CableTV set-top boxes. Consumer devices include personal audio devices from Sony, and digital cameras from various manufacturers. I have over 10 years of working experience with VxWorks, and I will be glad to address your specific needs, if you can elaborate on your requirements. Otherwise, WindRiver's web site www.windriver.com has all the information you may require. - -- - ----------------------------------------------------------------------- Leonid Rosenboim Consultant Email: my first name at consultant dot com "charley brown" wrote in message news:aovvme$ggs$1@news.kreonet.re.kr... > Hi everyone ! > > What is Vx-Works? > -concepts or definitions > -Use case > -Sample Applications > > Please answer the baby Questions.... > > I am a beginner.... > > Thanks a lot > --------------------------- Newsgroups: comp.os.vxworks Subject: Is Thread ID unique through the session? Date: 21 Oct 2002 05:59:23 -0700 From: toader@mail.ru (Andrei Sedelnikov) Organization: http://groups.google.com/ Message-ID: <6839b7a5.0210210459.5455a8f4@posting.google.com> Hallo everybody, I have a rather simple but probably very specific question: Can I rely upon that, that ThreadIDs (TIDs) are unique through the application's session, or they are being always reused, after the thread was killed? Thank you! --------------------------- Newsgroups: comp.os.vxworks Subject: USB camera in vxworks ? + cost if needed + method outline Date: Mon, 21 Oct 2002 23:41:11 GMT From: "Mike Kim" Organization: Dacom Message-ID: References: Thank you for mail from Leonid. I think I should know first the cost using usb stack(or lib) and outline how to use usb device. thanks... --------------------------- Newsgroups: comp.os.vxworks Subject: Re: USB camera in vxworks ? Re: USB Date: 21 Oct 2002 22:15:43 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0210212115.752d889c@posting.google.com> References: Hello, If the camera is a USB one, I'm not sure how you plan to avoid using a USB protocol stack - USB is not like a simple I/O port, it is a bus, and it has a fairly complex protocol running over it that must be used. You might be able to get other forms of camera - some older ones were parallel port based; you might be able to still get some of those, or you could look for a serial camera - some of the early Kodak digital cameras had serial port connections. You might also be able to write your own USB stack, or buy one from somewhere else... HTH, John... "Mike Kim" wrote in message news:... > Hi , > > I'd like to use a usb camera for image capturing. > > Because Someone asked me if it is possible to capture image through usb > camera. > > I'm using vxworks 5.4 and zinc 5.x. > > But using USB ddk is too much expensive. > > So I want to know if is there any another way to capture image in vxworks > environment. > > Thank you in-advance. > > "Ravi Uday" wrote in message > news:ecca3ca6.0210162128.c36f98c@posting.google.com... > > Hi, > > > > Can anybody tell me which is the best discussion site for discussing > > issues regarding writing driver for USB on VxWorks O.S. I already have > > the VxWorks USB DDK with me and have few questions of using / begin > > with it. > > My device is DiskOnKey. > > Thanks for all kind of helps/experience > > > > - Ravi --------------------------- Newsgroups: comp.os.vxworks Subject: BSP problem running vx Bootloader without ICE Date: Tue, 22 Oct 2002 05:27:40 GMT From: Icis Machine Message-ID: I'm new here and still very new to vxWorks. I'm developing a BSP for a MIPS based processor. I've got everything finished but when I try to run the vxworks bootloader in ROM so that it will load my vx image into RAM via TFTP without my Vision ICE, nothing appears to come out of the serial port (i.e. doesn't give the appearance that the bootrom is working). However when I do hook the ICE up and run the debugger (Vision Click), I can easily step through the code and even just reset and run it without problems. Also when I reset the board via the Vision Click software, everything seems fine. When I either soft reset via the reset button or power cycle, nothing comes out of the serial interface. What am I doing wrong here? Anyone have any ideas? jen +-------------------------------------------------------------------------+ If ignorance is bliss, you must be orgasmic. -forum2k --------------------------- Newsgroups: comp.os.vxworks Subject: Re: strange timer bahaviour Date: 21 Oct 2002 07:55:26 -0700 From: rick@cyclone.com (Richard Nardone) Organization: http://groups.google.com/ Message-ID: <29b153d9.0210210655.54f38bf4@posting.google.com> References: <1034839098.567517@newsreader1.wirehub.nl> <1034867481.955098@newsreader2.wirehub.nl> <29b153d9.0210180725.2c8f96b@posting.google.com> <1035192099.254012@newsreader2.wirehub.nl> "Renee Teunissen" > Hi Richard, > > Thanx about the remark, but: > > sys/types.h: (vxworks) > #ifndef _TIME_T_ > #define _TIME_T_ > typedef long time_t; > #endif > > linux (bits/types.h) > typedef long int __time_t; > > So I dont agree with you. time_t is signed! > Hi Renee, I have Tornado 2.2 with VxWorks 5.5 In the file \h\types\vxtypesbase.h the following definition is found: #ifndef _TYPE_time_t #define _TYPE_time_t typedef unsigned long time_t #endi In the file \h\time.h the following defintion is found: #ifdef _TYPE_time_t _TYPE_time_t; #undef _TYPE_time_t #endif This is where I derived my answer from. Didn't mean to offend. I can't imagine why it would be a signed value. Regards, Richard Nardone Cyclone Microsystems --------------------------- Newsgroups: comp.os.vxworks Subject: profilescope doesn't initialize... Date: 21 Oct 2002 19:39:10 -0700 From: djckdgus@dreamwiz.com (Changhyun Oe) Organization: http://groups.google.com/ Message-ID: <5d59bae9.0210211839.4efaa8e@posting.google.com> Hi~ I want to use ProfileScope, but it always shows exception as follows.. - ----------------------------------------------------------------- Exception number 4: Task: 0x4860e58 (t2) Floating-Point Disabled pc : 0x4864424 npc : 0x4864428 pcw : 0x509000c1 4257c88 _vxTaskEntry +c : 4239980 (4fffa10, 0, 10, 0, &_sysInit, &_sysBootLine) 4239a14 _wdbFuncCallLibInit+104: _ProfileScopeInit (1f4, 0, 10, 0, &_sysInit, &_sysBootLi ne) 486ad04 _ProfileScopeInit+58 : _scopeProfileInit (4976d50, 0, 2, 10, 1, 42327c0) 486a590 _scopeProfileInit+4c : _ProfileScopeConnectClk (0, 4976d50, 10, 1, 4295d60, 42976 c0) 486a534 _ProfileScopeConnectClk+a4 : _ProfileSetSampleRate (6b, 6b, 0, 7, ffffffff, 4976d 50) - ----------------------------------------------------------------- As you can see, it says 'floating point disabled' and I guess my sparc architecture has no FPU(Floating point unit). And what I want to know is ... 1. If no FPU exists, is it impossible to use Profilescope? 2. If not, how will it be possible? Thanks in advance. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Multicast driver sample Date: Tue, 22 Oct 2002 10:52:37 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Reply-To: "Leonid Rosenboim" "Stefano Pettazzoni" wrote in message news:ap2uv6$14sh$1@newsreader1.mclink.it... > Hi all, > where can I find a sample source code of a BSD 4.4 ethernet network driver > that support multicast > > Thanks, > Stefano. > -- > Stefano, when you see one such source of a driver you might be dissapointed, because some of the BSD drivers do not really have any real support for multicast - all they do is enter PROMISCOUS mode when multicasts are enabled, and let the IP stack do the rest. This is a quick a dirty solution, which sometimes may be quote sufficient, and in other times can be a source of serious problem. - -- - ----------------------------------------------------------------------- Leonid Rosenboim Consultant Email: my first name at consultant dot com --------------------------- Newsgroups: comp.os.vxworks Subject: remote object file loading Date: 22 Oct 2002 02:11:59 -0700 From: dalves@cfn.ist.utl.pt (lega) Organization: http://groups.google.com/ Message-ID: <4c7ddf17.0210220111.67da9de0@posting.google.com> Hi, I'm a recent VxWorks user and I'm trying to do a remote object file loading into VxWorks. I'm working on a Solaris (SunOS) machine and logging on to a Motorola board, running VxWorks, on a VME rack. I've been trying to use the command: vxRsh -t ld < file.o but it doesn't work and the operation is timed-out. By the way, it's funny that: vxRsh -t pwd and vxRsh -t i both work, and: vxRsh -t ls doesn't ... howcome ? Thanks for your help, lega --------------------------- End of New-News digest ********************** From vxwexplo-errs@csg.lbl.gov Tue Oct 22 07:10:31 2002 From: Mike McCullough Date: Tue Oct 22 07:10:33 PDT 2002 Subject: Re: Loading program slow using ld --=====================_10419562==_ Content-Type: text/plain; charset="us-ascii"; format=flowed Some additions to Leonid's responses for completeness... There are at least 2 other ways to increase the speed of the ld command, one simple and one not so simple. 1) If you are on a subnet try a direct connect between the UNIX box and the target board ie take it off the subnet. Many delays are due to local network congestion although with a 20Mb image this is unlikely to be your problem. You also may have problem booting your box if it isn't on the net, depends on your IT setup. 2) There is a switch for gdb that makes it only download symbols that it needs immediately rather than all the symbols at once. I've attached a tip on that as well as some basic target server fixes that usually help as well. Not sure if this will help with the ld command as I haven't tried it but you will see faster downloads from Crosswind. Newsgroups: comp.os.vxworks Subject: Re: Loading program slow using ld Date: Mon, 21 Oct 2002 14:55:45 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Reply-To: "Leonid Rosenboim" "Damain" wrote in message news:b6801b8c.0210210341.66c76141@posting.google.com... > Hi, > > Is there any way of downloading a program which is quicker then using ld. > > It's just taking a very long time to download a 20MB program, via WindSh. > > I am using: > VxWorks 5.4 > Sun Solaris 8 (host) > > Thanks > Damian Perhaps the quickest way to load a large application is by statically linking it with VxWorks, so when VxWorks boots, the application is already in the target's memory. The boot time of a combined application and VxWorks run-time will be many times faster, at the cost of somewhat longer build time. There is also a mid-way solution, where those parts of the application which do not change frequently are statically linked with VxWorks, while thos parts you are currently developing are dynamically loaded, so the longer build time has lesser significance because it is infrequently performed, and the dynamic loading tiem is much shorted due to smaller size of the loaded object. - -- - ----------------------------------------------------------------------- Leonid Rosenboim Visit: http://gamla.org.il/english/index.htm Consultant Email: my first name at consultant dot com ********************************************************** Mike McCullough MCC Engineering President and CEO 9 Cheshire Court Londonderry, NH 03053 mike@mccengineering.com Tel: 603-537-9593 Fax: 603-434-9890 www.mccengineering.com Cell: 603-247-1219 ********************************************************** 24x7 VxWorks Support Custom Support Programs VxWorks Training Embedded Consulting Services A WindLink Partner ********************************************************** --=====================_10419562==_ Content-Type: text/plain; charset="us-ascii" Content-Disposition: attachment; filename="WDB_POOL_SIZE.txt" Mike, This sounds suspiciously like the problem you showed me today. > Could you try redefining the WDB_POOL_SIZE in config.h: > > #undef WDB_POOL_SIZE > #define WDB_POOL_SIZE (3*1024*1024) /* 3 Mb for WDB agent pool */ > > Try this and get back to me? Thanx. > > > > >> Try increasing the tgtsvr cache more. > >> > >> Because the entire size of the object plus the size of the area needed > >> to > >> locate the object will ideally all live in cache during load-time, you > >> > >> will want to have a value significantly larger than your largest > >> object. > >> > >> Further, ensure that the target hasn't crashed for some reason.. like > >> include the telnet server and the shell or something.. or just ping it > >> > >> steadily during the load. > > > >Thanks.....this thought about the target crashing led me to the solution > >to the problem. The customer has a 16 Meg board. About 13 meg, after > >loading of vxWorks, is available for tools and apps. Of this 13 Meg, > >1/16 get allocated, by default, to tools (WDB_POOL_SIZE in configAll.h) > >thus only leaving around 850k for tool usage. Dynamic downloading > >allocates memory from this tool pool. Therefore, if the image I'm > >loading is 1.5M, and I only have 800K of space........... > > > > >> > Hello World, > >> > > >> > A customer in Argentina is trying to load (via WindShell) a 1.5 meg > >> > image to a target in Chicago. She is getting the above error. > >> > > >> > I snooped windsurf and found 3 suggestions: > >> > > >> > 1) Increase the target server timeout to a huge number ( set it to a > >> > >> > million - no luck) > >> > 2) Increase the memory cashe for the host tools ( set it to two meg > >> - > >> > didn't help) > >> > 3) run the following at the shell: -> ? set > >> > wtxCmdTimeout(wtxObjModuleLoad) 30000 > >> > > >> > Strike three...... > >> > > >> --=====================_10419562==_ Content-Type: text/plain; charset="us-ascii" Content-Disposition: attachment; filename="Lazy Loading Switch.txt" Not sure if you've ever seen the lazy loading switch. Believe what you are seeing is a result of compiling your entire application image with the -g switch which will typically bulk up the amount of memory (NT tmp space). Raising this often helps but does not remove the problem. There are 2 possible solutions to these problems: 1) Stop compiling the entire image for debugging, just compile the code you are debugging and remove the switch for modules that have already been tested. 2) Enable lazy-loading in the debugger which only loads the debug symbols for the current stack frame rather than the entire image at once. To do this requires you to: set symbol-readnow off You are supposed to put this in your .gdbinit file which you create. My usage of this was to add "gdb set symbol-readnow off" to the tcl file itself for Crosswind (gdb.tcl) in the target/host/resource/tcl directory. At 04:49 PM 1/29/01 +0000, you wrote: Hi All, I have a customer who is linking in a large application in to their VxWorks image (Tornado 2.0 on Solaris 2.6). The total size of the image is 32Mb, and is built with debugging symbols. Everything is fine, until they launch Crosswind ... this only occurs with really big applications linked into their image. Anyone have any clues or suggestions as to why this might be happening? --=====================_10419562==_ Content-Type: text/plain; charset="us-ascii" Content-Disposition: attachment; filename="TgtSvrCache_NTSWap.txt" A couple questions on this Roger: 1) Are you using a large debug image say over a couple Megabytes?? If this is the case you might need to start using the following: FYI The line to add to make GDB do lazy reading of debug info is: set symbol-readnow off You are supposed to put this in your .gdbinit file which you create. My usage of this was to add "gdb set symbol-readnow off" to the tcl file itself for Crosswind (gdb.tcl) in the target/host/resource/tcl directory. 2) Check that the target server cache size is more than twice the size of your debug image. You can tweak this on the target server configuration window. 3) There is a way to increase the NT swap space if you haven't figured out how to do this already. I had a customer who started with this but ended up needing to use the lazy loading switch as well. If none of these are any help, please contact me so we can look into it more. X-Sender: ticse@mail.wrs.com X-Mailer: QUALCOMM Windows Eudora Pro Version 4.2.0.58 Date: Thu, 23 Mar 2000 16:58:59 -0800 To: Mike McCullough From: Michael Ticse Subject: Ignitus: FW: Swap space running out... Cc: "Al Maillet (E-mail)" , Roger Cruz , Peter Fetterolf Mike, Can you please help Roger out on this problem. Thanks, Michael At 11:39 AM 3/23/00 -0500, Peter Fetterolf wrote: Michael, This bug is a major problem for us: (From Rogers email below...) Ever since I've used Tornado (back to T1.0), I've experienced a problem > with CrossWind crashing and leaving a lot of swap space used up, ie, never > released. This becomes a problem after a while and requires us to reboot > our server. Not a nice thing to do when you have 70 engineers depending > on this machine. We need to get a patch for this ASAP. Peter Fetterolf VP, Software Engineering Ignitus Communications LLC Phone: 978-929-9133 X302 Fax: 978-929-9141 Mobile: 617-803-4202 http://www.ignitus.com -----Original Message----- From: Lancz Griswold [mailto:lancz.griswold@windriver.com] Sent: Thursday, March 02, 2000 6:11 PM To: Roger Cruz Subject: RE: Swap space running out... Roger, I don't have any suggestions on this one. I know there are files open associated with the debugger session and I suppose they are whats eating up resources. I'll check to see if others have seen these problems and what they may do to overcome them or at least deal with them in some way. Lancz -----Original Message----- From: Roger Cruz [mailto:RogerC@Ignitus.com] Sent: Thursday, March 02, 2000 5:08 PM To: 'lancz@wrs.com' Cc: 'al@wrs.com'; 'johnb@wrs.com'; 'cruz@ignitus.com'; Peter Fetterolf Subject: RE: Swap space running out... Lancz, do you have any suggestions here? are there any fixes that I'm not aware of? R. > -----Original Message----- > From: Roger Cruz > Sent: Thursday, February 17, 2000 5:35 PM > To: 'support@wrs.com' > Cc: 'al@wrs.com'; 'johnb@wrs.com'; 'lancz@wrs.com'; 'cruz@ignitus.com' > Subject: Swap space running out... > > > License: 30463 > Host: Solaris 2.6 > BSP: YK750, MBX860 > Tornado II > > Ever since I've used Tornado (back to T1.0), I've experienced a problem > with CrossWind crashing and leaving a lot of swap space used up, ie, never > released. This becomes a problem after a while and requires us to reboot > our server. Not a nice thing to do when you have 70 engineers depending > on this machine. Has a fix or patch ever been released for this problem? > CrossWind dies when the target is rebooted and the CrossWind menus are > used to reconnect to the target server. > > Thanks. > Roger <> **************************************************************************** **** Michael Ticse Tel: 732-933-2622 Lucent Account Manager Fax: 732-933-2610 Wind River Systems Cell: 732-492-5786 125 Half Mile Rd., Suite 200 WWW: http://www.windriver.com Red Bank, NJ 07701 Email: ticse@windriver.com ******************************************************************************** ********************************************************************* Mike McCullough Onsite Service Products Program Manager 9 Cheshire Court Londonderry, NH 03053 Americas Sales Wind River Systems, Inc. Phone: (603) 537-9593 http://www.windriver.com Fax: (603) 434-8925 "Monkey Management without Going Bananas" ********************************************************************* --=====================_10419562==_-- From vxwexplo-errs@csg.lbl.gov Tue Oct 22 08:17:42 2002 From: "Brown, Matthew" Date: Tue Oct 22 08:17:43 PDT 2002 Subject: FW: Loading program slow using ld We had a similiar problem using VxWorks 5.4 (Took a few minutes to download a 10 meg object). There is a way to download an object directly to the target without the object going *through* the windshell. See if you have a windsh.tcl file in your .wind directory and add the following line : shConfig LD_SEND_MODULES off Or you can execute that line directly from the windsh by putting a '?' in front of it. That should force objects directly down to the target when you execute a ld from the windshell. There's a few other tcl environment variables you can set. Check out section 5.2.2 of the Tornado User's Guide. Matt > ---------- > From: vxwexplo@lbl.gov[SMTP:vxwexplo@lbl.gov] > Sent: Tuesday, October 22, 2002 10:12 AM > To: vxworks_users@csg.lbl.gov > Subject: Re: Loading program slow using ld > > Submitted-by vxwexplo-errs@csg.lbl.gov Tue Oct 22 07:10:31 2002 > Submitted-by: Mike McCullough > > > --=====================_10419562==_ > Content-Type: text/plain; charset="us-ascii"; format=flowed > > Some additions to Leonid's responses for completeness... > > There are at least 2 other ways to increase the speed of the ld command, > one simple and one not so simple. > > 1) If you are on a subnet try a direct connect between the UNIX box and > the > target board ie take it off the subnet. Many delays are due to local > network congestion although with a 20Mb image this is unlikely to be your > problem. You also may have problem booting your box if it isn't on the > net, > depends on your IT setup. > > 2) There is a switch for gdb that makes it only download symbols that it > needs immediately rather than all the symbols at once. I've attached a tip > > on that as well as some basic target server fixes that usually help as > well. Not sure if this will help with the ld command as I haven't tried it > > but you will see faster downloads from Crosswind. > > > Newsgroups: comp.os.vxworks > Subject: Re: Loading program slow using ld > Date: Mon, 21 Oct 2002 14:55:45 +0200 > From: "Leonid Rosenboim" > Organization: Verio > Message-ID: > References: > Reply-To: "Leonid Rosenboim" > "Damain" wrote in message > news:b6801b8c.0210210341.66c76141@posting.google.com... > > Hi, > > > > Is there any way of downloading a program which is quicker then using > ld. > > > > It's just taking a very long time to download a 20MB program, via > WindSh. > > > > I am using: > > VxWorks 5.4 > > Sun Solaris 8 (host) > > > > Thanks > > Damian > From vxwexplo-errs@csg.lbl.gov Tue Oct 22 08:22:29 2002 From: Bruce Wilson Date: Tue Oct 22 08:22:30 PDT 2002 Subject: Re: Loading program slow using ld Try increasing the MAX_CACHE_MEMORY parameter within the launch window. The default is 1000000 (1MB), and when we increased it to 2000000 (2MB), "ld > Hi, > > > > Is there any way of downloading a program which is quicker then using ld. > > > > It's just taking a very long time to download a 20MB program, via WindSh. > > > > I am using: > > VxWorks 5.4 > > Sun Solaris 8 (host) > > > > Thanks > > Damian -- Bruce A. Wilson U.C. Lawrence Livermore National Lab P.O. Box 808 (L-278) Livermore, CA 94550 voice: (925) 422-4145 fax: (925) 423-1243 From vxwexplo-errs@csg.lbl.gov Wed Oct 23 04:03:22 2002 From: Vxworks Exploder Date: Wed Oct 23 04:03:23 PDT 2002 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Wed Oct 23 04:03:19 PDT 2002 Subject: Re: mcp750 bsp probs with Tornado II Subject: RTI Utilities PatchSet Subject: Re: Is there a 64-bit RawFS (ala DosFS 2.0)? Subject: Creating a vxWorks.st system image Subject: mcp750 bsp probs with Tornado II Subject: Re: mcp750 bsp probs with Tornado II Subject: The infamous "No rule to make target" problem Subject: Newbie Interrupt Service Routine Question Subject: Re: The infamous "No rule to make target" problem Subject: vxdiskadm Subject: Q: host-target connection lost Subject: Re: dosFs2.2 is printing junk during init? Subject: Re: The infamous "No rule to make target" problem Subject: Re: arptnew failed Subject: Re: Is there a 64-bit RawFS (ala DosFS 2.0)? Subject: Re: vxdiskadm Subject: Multicast driver sample Subject: Re: USB camera in vxworks ? Re: USB Subject: Re: The infamous "No rule to make target" problem Subject: Re: The infamous "No rule to make target" problem Subject: Beginner needs help Subject: Re: strange timer bahaviour Subject: Re: mcp750 bsp probs with Tornado II Subject: Filter packets at i82559 Subject: Re: The infamous "No rule to make target" problem Subject: Sending a block of Data at MAC level Subject: Re: USB camera in vxworks ? Re: USB Subject: Source Browser in Tornado IDE Subject: Re: The infamous "No rule to make target" problem Subject: Re: Stack size Subject: Re: Stack size Subject: How to detect the failure of a hardisk or filesystem in vxWorks? Subject: Stack size Subject: How to replace VxWorks TCP/IP Stack source Subject: Re: strange timer bahaviour Subject: Re: Stack size Subject: Re: How to detect the failure of a hardisk or filesystem in vxWorks? ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: mcp750 bsp probs with Tornado II Date: Tue, 22 Oct 2002 14:37:49 +0100 From: Dave Moore Organization: Computer Sciences Corporation Message-ID: <3DB554AD.DD2E21F9@baesystem.com> References: <3DB54821.53EBE158@baesystem.com> <3DB54AEC.21B72C02@stest.ch> Marc Waber wrote: > Hi, > > afaik, vxWorks images under T2 are created in the directory > ".../target/proj//default". The one in the > config-direcrory is just a minimal image pre-configured by WindRiver and > has nothing to do with the image you create using T2. > Yes, I'm aware that the Tornado generated image will be in .target/proj/mcp750/default' > > I guess that you have a misconfiguration in your project. > Since the vxWorks image is loaded, the bootrom seems to execute ok, and > the vxWorks image has a problem. > Well the Tornado users manual suggests that all Tornado II BSP come with an already prepared project file for this purpose (ie. mcp750.wpj). However, when I install the bsp (straight from the Tornado II Disk) there isn't one there, so I guess I need to create one and thought I could by basing the new project on the exiting working BSP. > > Make sure your settings in the project tool of T2 are compatible to your > hardware and to the "config.h" settings. > If settings differ from "config.h" to the project tool, this might cause > the application to crash!! I've looked in the wpj file with a text editor and there is a list of INCLUDE_XXX whatever. So I guess I could use the Find Object.. function to find all the additional INCLUDES that Tornado insists on putting in, so the the list matches that in the configAll.h file?. --------------------------- Newsgroups: comp.os.vxworks Subject: RTI Utilities PatchSet Date: Tue, 22 Oct 2002 13:42:27 +0000 (UTC) From: Patrick Organization: Siemens Business Services Message-ID: Reply-To: pat102345@netscape.NOJUNK.net Hi Folks, I have some third party code that makes use of an RTI library function called PatchSet. PatchSet can be used to insert code into a running system. The version used is quite old, circa 2000, and is broken somewhere in the patch.so library, I have no source code. But I have a newer version of patch.so which I would like to use instead. The original RTILib seems to have been a product in its own right, and I can find info. for the old PatchSet function. But the current version appears to be part of RTI ScopeTools product. The interface to PatchSet has changed quite a bit and I can find no info. about this new interface. So, to the question, has anybody used the current RtiTools_PatchAddressSet function, and have you any information on it. I particularly need info. about the PatchInstance struct. Thanks for your help. - -- Patrick - ---------------------------------------------------------------------- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Is there a 64-bit RawFS (ala DosFS 2.0)? Date: 22 Oct 2002 07:25:58 -0700 From: taashlo@sandia.gov (Tad Ashlock) Organization: http://groups.google.com/ Message-ID: <5917f4d0.0210220625.4354a841@posting.google.com> References: <7kgipiaf.fsf@sandia.gov> "Leonid Rosenboim" wrote in message news:... > Ok, cool investigation. Don't know who "rsh" is, but I am curious, do you > see > any mention of "lrn" in the modification history below ? (that would be me > ;-) ) Yes, actually. Here's the modification history for rawFsLib.h: modification history - -------------------- 02c,21jun00,rsh upgrade to dosFs 2.0 03f,31jul99,jkf changes for CBIO API. 03e,31jul99,jkf T2 merge, tidiness & spelling. 03d,15oct98,lrn moved 64-bit extended ioctl codes to ioLib.h 03c,08oct98,vld added definition of RAWFS_DEF_MAX_FILES 03b,08oct98,vld driver interface changed to CBIO_DEV. replaced rawvd_pBlkDev with CBIO_DEV * rawvd_pCbio changed prototype of rawFsDevInit(). 03a,23jul98,vld added ioctl codes for 64-bit ioctl requests; new error codes. 02b,22sep92,rrr added support for c++ 02a,04jul92,jcf cleaned up. 01d,26may92,rrr the tree shuffle 01c,04oct91,rrr passed through the ansification filter -changed VOID to void -changed copyright notice 01b,05oct90,shl added ANSI function prototypes. added copyright notice. 01a,02oct90,kdl written > You are absolutely brilliant to remember the 64-bit math issue of PPC > which was discovered with connection to DsFs2.0 - since rawFs > if it is really the nore code, will too use 64 bit math (albeit at a much > primitive level) so it too may be subject to the same PPC compiler > issue causing use of FP registers. > > One way to look into this would be to check if tha patch for 64-bit > PPC issue of DosFs2 also includes a newer rawFsLib.o file !! > > Good luck ! > Leonid It doesn't look like rawFsLib.o was changed (in the libPPC604gnuvx.a archive). It's dated May 27, 2001 in the archive, but the patch (70400) is dated September 26, 2001. In addition, ioctl(fd, FIOSEEK64, (int)&TestAddress) is returning S_ioLib_UNKNOWN_REQUEST. Does this mean that RawFS *doesn't* support 64-bit operations? Or am I doing something wrong? Thanks, Tad --------------------------- Newsgroups: comp.os.vxworks Subject: Creating a vxWorks.st system image Date: 22 Oct 2002 07:34:53 -0700 From: vanessabenfield98@hotmail.com (Vanessa Benfield) Organization: http://groups.google.com/ Message-ID: Hi all, If you answer this message please can you be as simple as possible as I'm very new to vxWorks and can be confused easily! I'm trying to create a vxWorks.st image using tornado 2, there was no rule already in my makefile (I've created a bootable project based on an exisiting bsp - pcPentium bsp) so I've added on from the old rules.bsp file. However this doesn't seem to work! I get a couple of errors: *** no rule to make target 'depend.pcPentium' needed by vxWorks.st and Circular vxWorks.st <- vxWorks.st dependency dropped I think there is something wrong with my dependencies, though I haven't really changed anything apart from adding the xvxWorks.st rule as there wasn't one!! Could anyone please advise me how I can create an vxWorks.st image as I didn't think there would be so many problems!!! Thanks in advance. Vanessa Benfield --------------------------- Newsgroups: comp.os.vxworks Subject: mcp750 bsp probs with Tornado II Date: Tue, 22 Oct 2002 13:44:18 +0100 From: Dave Moore Organization: Computer Sciences Corporation Message-ID: <3DB54821.53EBE158@baesystem.com> Hi All, I'm basically trying to create a bootable vxWorks image using the Tornado II tool. I've got Tornado II and the 750 bsp installed. If I just use a dos command shell, cd to the bsp directory (ie. c:\Tornado\target\config\mcp750) and type 'make clean' and then 'make', this creates vxWorks and vxWorks.sym. My mcp750 will boot up from this fine. However, if I try to use the Tornado tool to create a bootable image, use the working mcp750 bsp as a base, then the card refuses to boot from it. It gets as far as 'Starting as 0x100000...' and then hangs. In fact the size of the image created by 'make'ing the bsp directy via a dos shell and building it using the Tornado tool are significantly different. Why is this?. Surely a project created bootable image createdf from Tornado II should create a working image provided that it is 'based' upon a currently working bsp?. So what am I doing wrong?. And why are the images different sizes?. Thanks in advance, Dave --------------------------- Newsgroups: comp.os.vxworks Subject: Re: mcp750 bsp probs with Tornado II Date: Tue, 22 Oct 2002 14:56:12 +0200 From: Marc Waber Organization: STS Switching Test Solutions Message-ID: <3DB54AEC.21B72C02@stest.ch> References: <3DB54821.53EBE158@baesystem.com> Hi, afaik, vxWorks images under T2 are created in the directory ".../target/proj//default". The one in the config-direcrory is just a minimal image pre-configured by WindRiver and has nothing to do with the image you create using T2. I guess that you have a misconfiguration in your project. Since the vxWorks image is loaded, the bootrom seems to execute ok, and the vxWorks image has a problem. Make sure your settings in the project tool of T2 are compatible to your hardware and to the "config.h" settings. If settings differ from "config.h" to the project tool, this might cause the application to crash!! hth, marc --------------------------- Newsgroups: comp.os.vxworks Subject: The infamous "No rule to make target" problem Date: Tue, 22 Oct 2002 10:30:30 -0600 From: "Matt Osminer" Organization: Newshosting.com - Highest quality at a great price! www.newshosting.com Message-ID: <3db57e4b$2@news.mhogaming.com> I seem to have come to an impasse with the infamous "No rule to make target" problem. As any Tornado sufferer will tell you the solution should be 1. Add -I directives to your C/C++ settings 2. Add a VPATH directive to your makefile Macros from the Project Macros Tab 3. Rebuild all dependencies, then Build-All Wondering if anybody has tried all of the above, as I have, and still been unable to get past the "No rule" error? My project was working just fine and then I did the unthinkable act of removing a few obsolete files and adding some new files. I didn't touch the -I or VPATH configurations and my new files all fell within the bounds of the original settings so there shouldn't have been a problem. The weird thing is Tornado just seems to be having a hard time locating this one directory. All my other -I directories appear just fine. I've tried: 1. Closing Tornado, rebuilding dependencies, saving the workspace, and "building-all" in about every combination you can think of. 2. Removing all the -I options and re-entering them 3. Removing the VPATH and re-entering it 4. Checked for directory existance etc. Any further ideas out there??? --------------------------- Newsgroups: comp.os.vxworks Subject: Newbie Interrupt Service Routine Question Date: Tue, 22 Oct 2002 13:59:00 GMT From: robotz@robotz.rob (Robotz) Organization: Randori News - http://www.randori.com - Try our FREE Usenet Scanner! Message-ID: <3db557bb.3671715708@news.airband.net> Reply-To: robotz@robotz.rob I'm new to Tornado and vxWorks and I'm trying to get an interrupt service routine working. I've got an SBC8260 board and I have a function generator hooked up to IRQ 4 to trigger it. Right now I'm running it at a very low speed (about 1Hz). So far I'm not getting anything. Can you guys just make sure I'm not totally missing something... /* intRecv just prints a message */ /* 0x0b000100 should be IRQ4 */ intConnect( 0x0b000100, intRecv, 0 ); intEnable(); /* wait for interrupts, press ENTER to continue */ getc( stdin ); intDisable(); thanks in advance, - -paul --------------------------- Newsgroups: comp.os.vxworks Subject: Re: The infamous "No rule to make target" problem Date: 22 Oct 2002 10:58:41 -0600 From: taashlo@sandia.gov Organization: Sandia National Laboratories, Albuquerque, NM USA Message-ID: <8z0q8l6m.fsf@sandia.gov> References: <3db57e4b$2@news.mhogaming.com> Sender: taashlo@TYPHOON "Matt Osminer" writes: > I seem to have come to an impasse with the infamous "No rule to make target" > problem. As any Tornado sufferer will tell you the solution should be > > 1. Add -I directives to your C/C++ settings > 2. Add a VPATH directive to your makefile Macros from the Project Macros Tab > 3. Rebuild all dependencies, then Build-All > > Wondering if anybody has tried all of the above, as I have, and still been > unable to get past the "No rule" error? My project was working just fine > and then I did the unthinkable act of removing a few obsolete files and > adding some new files. I didn't touch the -I or VPATH configurations and my > new files all fell within the bounds of the original settings so there > shouldn't have been a problem. > > The weird thing is Tornado just seems to be having a hard time locating this > one directory. All my other -I directories appear just fine. > > I've tried: > 1. Closing Tornado, rebuilding dependencies, saving the workspace, and > "building-all" in about every combination you can think of. > 2. Removing all the -I options and re-entering them > 3. Removing the VPATH and re-entering it > 4. Checked for directory existance etc. > > Any further ideas out there??? I'm sorry to say that I too have encountered this problem. In fact I reported it to Wind River's tech support. They were unable to reproduce the problem and demanded all of my files. Due to various reasons, this is simply out of the question and the support stalled right there. The only solution I found was to delete the entire project and recreate it from scratch. I've now had to do this three times now, and I'm getting rather sick of it. I've only seen this problem in Tornado version 2.1 for the PPC. The problem appears to be that dependencies are not being regenerated after deleting old files and then adding new files. For instance, I can delete a file from the project named pmcsio4.cpp, then add a file named serial.cpp. I then hit the "Update Dependencies" button. Then I exit Tornado and hit "Yes" when asked whether or not to save the workspace. At this point when I take a look at the makefile and the .prj file, they both still have references to the old file (pmcsio4.cpp) and no mention of the new file (serial.cpp). I didn't have this problem with earlier versions of Tornado (2.0, 1.x). If you find out any way of working around this problem, please let us know. Tad Ashlock --------------------------- Newsgroups: comp.os.vxworks Subject: vxdiskadm Date: 22 Oct 2002 10:13:39 -0700 From: sunill@hotmail.com (Sunu) Organization: http://groups.google.com/ Message-ID: <5f2cce34.0210220913.13ae27e7@posting.google.com> Vxdiskadm --------------------------- Newsgroups: comp.os.vxworks Subject: Q: host-target connection lost Date: Tue, 22 Oct 2002 15:58:33 +0200 From: "Khac Tri VU" Organization: Planet Internet Message-ID: Hi folks, We run our applications and sometimes we got host-target connection lost. Anyhow, the applications still run and no crash has been detected. It is likely that after a long while being preempted, the tNetTask cannot recover the connection. Does anyone experience this and know how to resolve the issue? Thank you in advance. Tri. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: dosFs2.2 is printing junk during init? Date: Tue, 22 Oct 2002 10:46:36 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <6dcae13d.0210211711.6804a30d@posting.google.com> Reply-To: "Leonid Rosenboim" Simply add this DOS_OPT_QUIET option when you call dosFsVolFormat(). - -- - ----------------------------------------------------------------------- Leonid Rosenboim Visit: http://gamla.org.il/english/index.htm Consultant Email: my first name at consultant dot com "roswell" wrote in message news:6dcae13d.0210211711.6804a30d@posting.google.com... > hi.. > > how can it get rid off these junk printed by dosFs2.x. > Roswell > > ------------------------- > > Retrieved old volume params with %33 confidence: > Volume Parameters: FAT type: FAT32, sectors per cluster 0 > 0 FAT copies, 0 clusters, 0 sectors per FAT > Sectors reserved 0, hidden 0, FAT sectors 0 > Root dir entries 0, sysId (null) , serial number 0 > Label:" " ... > Disk with 16384 sectors of 512 bytes will be formatted with: > Volume Parameters: FAT type: FAT16, sectors per cluster 2 > 2 FAT copies, 8143 clusters, 32 sectors per FAT > Sectors reserved 1, hidden 0, FAT sectors 64 > Root dir entries 512, sysId VXDOS16 , serial number 0 > Label:" --------------------------- Newsgroups: comp.os.vxworks Subject: Re: The infamous "No rule to make target" problem Date: 22 Oct 2002 13:26:44 -0400 From: Bill Pringlemeir Organization: Bell Sympatico Message-ID: References: <3db57e4b$2@news.mhogaming.com> <8z0q8l6m.fsf@sandia.gov> Sender: bpringle@DeadDuck [snip] Tad> At this point when I take a look at the makefile and the .prj Tad> file, they both still have references to the old file Tad> (pmcsio4.cpp) and no mention of the new file (serial.cpp). Tad> I didn't have this problem with earlier versions of Tornado Tad> (2.0, 1.x). Tad> If you find out any way of working around this problem, please Tad> let us know. I am not sure if I know about this problem. "http://www.xs4all.nl/~borkhuis/vxworks/vxw_pt1.html#1.6-B" "http://www.xs4all.nl/~borkhuis/vxworks/vxw_pt1.html#1.7-C" In our projects we have more than one target. Often I get what is being discussed, or so I believe. Changing the build target and updating dependacies seems to solve this. The "wpj" is just a list of information that a TCL script reads and uses to produce the make file. The "makeGen.tcl" in Johan's FAQ uses these TCL files. There are also some other Tornado files for doing command line builds. I filed a TSR about this and it is fairly complex to set up, but it does work. The file comments have some statements that aren't quite right. For example, "/Tornado/host/resource/tcl/app-config/Project/cmpScriptLib.tcl" Not sure if this is the actual files I was looking at... I just ended up using a modified version of "makeGen.tcl" to produce lint rules and emacs ebrowse files. hth, Bill Pringlemeir. - -- CESID CIDA 2600 Magazine secure Skipjack basement halcon PLO Rule Psix USDOJ UOP Sears Tower RSA supercomputer INSCOM AES crack --------------------------- Newsgroups: comp.os.vxworks Subject: Re: arptnew failed Date: Tue, 22 Oct 2002 12:37:16 -0500 From: Andrew Johnson Organization: APS, Argonne National Laboratory Message-ID: References: Leonid Rosenboim wrote: > Andrew, I will try to explain this from another perspective: Thanks Leonid, that helps. (That answer was one for the FAQ I think...) - - Andrew - -- "Life is what happens while you're busy making other plans." - John Lennon --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Is there a 64-bit RawFS (ala DosFS 2.0)? Date: Tue, 22 Oct 2002 19:52:00 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <7kgipiaf.fsf@sandia.gov> <5917f4d0.0210220625.4354a841@posting.google.com> Reply-To: "Leonid Rosenboim" "Tad Ashlock" wrote in message news:5917f4d0.0210220625.4354a841@posting.google.com... > "Leonid Rosenboim" wrote in message news:... > > Ok, cool investigation. Don't know who "rsh" is, but I am curious, do you > > see > > any mention of "lrn" in the modification history below ? (that would be me > > ;-) ) > > Yes, actually. Here's the modification history for rawFsLib.h: > > modification history > -------------------- > 02c,21jun00,rsh upgrade to dosFs 2.0 > 03f,31jul99,jkf changes for CBIO API. > 03e,31jul99,jkf T2 merge, tidiness & spelling. > 03d,15oct98,lrn moved 64-bit extended ioctl codes to ioLib.h Yup, that's me :-) > 03c,08oct98,vld added definition of RAWFS_DEF_MAX_FILES > 03b,08oct98,vld driver interface changed to CBIO_DEV. > replaced rawvd_pBlkDev with CBIO_DEV * rawvd_pCbio > changed prototype of rawFsDevInit(). > 03a,23jul98,vld added ioctl codes for 64-bit ioctl requests; new > error codes. And this is Vladimir, my partner who did most of the coding :-) > 02b,22sep92,rrr added support for c++ > 02a,04jul92,jcf cleaned up. > 01d,26may92,rrr the tree shuffle > 01c,04oct91,rrr passed through the ansification filter > -changed VOID to void > -changed copyright notice > 01b,05oct90,shl added ANSI function prototypes. > added copyright notice. > 01a,02oct90,kdl written > > > > You are absolutely brilliant to remember the 64-bit math issue of PPC > > which was discovered with connection to DsFs2.0 - since rawFs > > if it is really the nore code, will too use 64 bit math (albeit at a much > > primitive level) so it too may be subject to the same PPC compiler > > issue causing use of FP registers. > > > > One way to look into this would be to check if tha patch for 64-bit > > PPC issue of DosFs2 also includes a newer rawFsLib.o file !! > > > > Good luck ! > > Leonid > > It doesn't look like rawFsLib.o was changed (in the libPPC604gnuvx.a > archive). It's dated May 27, 2001 in the archive, but the patch > (70400) is dated September 26, 2001. Shoot! > > In addition, ioctl(fd, FIOSEEK64, (int)&TestAddress) is returning > S_ioLib_UNKNOWN_REQUEST. Does this mean that RawFS *doesn't* support > 64-bit operations? Or am I doing something wrong? Let me find the words ... beep... beep...beep... did they put the new include file and keep the old source file ? How ...beep... careless of them! Sorry, for having now to send you back to the support folks, but is there a choise ? Just ask them to send you the revision history of the rawFsLib.c file from the ClearCase view that is appropriate for your version and patch level, and see if you have the same history lines as you have in the header file. Hope this helps, Leonid --------------------------- Newsgroups: comp.os.vxworks Subject: Re: vxdiskadm Date: Tue, 22 Oct 2002 19:59:05 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <5f2cce34.0210220913.13ae27e7@posting.google.com> Reply-To: "Leonid Rosenboim" Sunu, I doubt anyone outside of Hitachi Data Systems (that is the company you work for) has heard of ++ Vxdiskadm ++ - -- What exactly did you want to ask ? - ----------------------------------------------------------------------- Leonid Rosenboim Consultant Email: my first name at consultant dot com "Sunu" wrote in message news:5f2cce34.0210220913.13ae27e7@posting.google.com... > Vxdiskadm --------------------------- Newsgroups: comp.os.vxworks Subject: Multicast driver sample Date: Tue, 22 Oct 2002 09:32:09 +0200 From: "Stefano Pettazzoni" Organization: MC-link S.p.A. Message-ID: Hi all, where can I find a sample source code of a BSD 4.4 ethernet network driver that support multicast Thanks, Stefano. - -- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: USB camera in vxworks ? Re: USB Date: Tue, 22 Oct 2002 07:51:15 GMT From: "Mike Kim" Organization: Dacom Message-ID: References: <488e459a.0210212115.752d889c@posting.google.com> OK, I want to know alternative way using usb device. But You said USB is a bus which has a complex bus. I know buying USB stack is a fast, convient way for me. But only I want to know if there is any other way about this. Thanks anyway for your respose. - -- ####################################### ChanSeong Kim OFFICE : 02-829-3391 FAX : 02-829-3400 PCS : 016-660-6457 KOREA COMPUTER BLDG. F9 395-65, SHNDAEBANG-DONG, DONGJAK-KU SEOUL, 156-010 KOREA ####################################### "John" wrote in message news:488e459a.0210212115.752d889c@posting.google.com... > Hello, > > If the camera is a USB one, I'm not sure how you plan to avoid using a > USB protocol stack - USB is not like a simple I/O port, it is a bus, > and it has a fairly complex protocol running over it that must be > used. > > You might be able to get other forms of camera - some older ones were > parallel port based; you might be able to still get some of those, or > you could look for a serial camera - some of the early Kodak digital > cameras had serial port connections. You might also be able to write > your own USB stack, or buy one from somewhere else... > > HTH, > > John... > > "Mike Kim" wrote in message news:... > > Hi , > > > > I'd like to use a usb camera for image capturing. > > > > Because Someone asked me if it is possible to capture image through usb > > camera. > > > > I'm using vxworks 5.4 and zinc 5.x. > > > > But using USB ddk is too much expensive. > > > > So I want to know if is there any another way to capture image in vxworks > > environment. > > > > Thank you in-advance. > > > > "Ravi Uday" wrote in message > > news:ecca3ca6.0210162128.c36f98c@posting.google.com... > > > Hi, > > > > > > Can anybody tell me which is the best discussion site for discussing > > > issues regarding writing driver for USB on VxWorks O.S. I already have > > > the VxWorks USB DDK with me and have few questions of using / begin > > > with it. > > > My device is DiskOnKey. > > > Thanks for all kind of helps/experience > > > > > > - Ravi --------------------------- Newsgroups: comp.os.vxworks Subject: Re: The infamous "No rule to make target" problem Date: Tue, 22 Oct 2002 12:15:36 -0700 From: "Michael R. Kesti" Organization: MK Associates Message-ID: <3DB5A3D8.AF34497E@gv.net> References: <3db57e4b$2@news.mhogaming.com> Reply-To: mkesti@gv.net Matt Osminer wrote: >I seem to have come to an impasse with the infamous "No rule to make target" >problem. As any Tornado sufferer will tell you the solution should be > >1. Add -I directives to your C/C++ settings >2. Add a VPATH directive to your makefile Macros from the Project Macros Tab >3. Rebuild all dependencies, then Build-All > >Wondering if anybody has tried all of the above, as I have, and still been >unable to get past the "No rule" error? My project was working just fine >and then I did the unthinkable act of removing a few obsolete files and >adding some new files. I didn't touch the -I or VPATH configurations and my >new files all fell within the bounds of the original settings so there >shouldn't have been a problem. > >The weird thing is Tornado just seems to be having a hard time locating this >one directory. All my other -I directories appear just fine. > >I've tried: >1. Closing Tornado, rebuilding dependencies, saving the workspace, and >"building-all" in about every combination you can think of. >2. Removing all the -I options and re-entering them >3. Removing the VPATH and re-entering it >4. Checked for directory existance etc. > >Any further ideas out there??? I've had luck removing the offending file from the project and again adding it. This sometimes turns into "gopher smashing", as fixing one file seems to cause another to fail. Perseverence usually pays off, however, and I eventually get the system back to the state where it will again build. - -- ======================================================================== Michael Kesti | "And like, one and one don't make | two, one and one make one." mkesti@gv.net | - The Who, Bargain --------------------------- Newsgroups: comp.os.vxworks Subject: Re: The infamous "No rule to make target" problem Date: Tue, 22 Oct 2002 20:24:13 +0000 (UTC) From: "Graham Baxter" Organization: BT Openworld Message-ID: References: <3db57e4b$2@news.mhogaming.com> Are the slashes '/' the right way around in your include paths. - -- Regards, Graham Baxter Freelance Software Engineer (VxWorks and pSOS BSPs) Graham Baxter (Software) Limited gbaxter@NOSPAMbcs.org.uk "Matt Osminer" wrote in message news:3db57e4b$2@news.mhogaming.com... > I seem to have come to an impasse with the infamous "No rule to make target" > problem. As any Tornado sufferer will tell you the solution should be > > 1. Add -I directives to your C/C++ settings > 2. Add a VPATH directive to your makefile Macros from the Project Macros Tab > 3. Rebuild all dependencies, then Build-All > > Wondering if anybody has tried all of the above, as I have, and still been > unable to get past the "No rule" error? My project was working just fine > and then I did the unthinkable act of removing a few obsolete files and > adding some new files. I didn't touch the -I or VPATH configurations and my > new files all fell within the bounds of the original settings so there > shouldn't have been a problem. > > The weird thing is Tornado just seems to be having a hard time locating this > one directory. All my other -I directories appear just fine. > > I've tried: > 1. Closing Tornado, rebuilding dependencies, saving the workspace, and > "building-all" in about every combination you can think of. > 2. Removing all the -I options and re-entering them > 3. Removing the VPATH and re-entering it > 4. Checked for directory existance etc. > > Any further ideas out there??? > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Beginner needs help Date: 22 Oct 2002 14:06:29 -0700 From: NxMorpheus@netscape.net Organization: http://groups.google.com/ Message-ID: Hi, I am fairly new to VxWorks and I am trying to run the listed examples in the VxWorks programmers guide and I am facing the following problem: I have successfully compiled the priorityTest.c sample application but when I try to load it on the board, I get unresolved symbols error with the following message: Errors while downloading d:\\priorityTest.o: sched_setparam sched_getparam I have modified my configAll.h file to define the INCLUDE_POSIX_SCHED and include this facility in the project but it doesn't seem to work. I would appreciate it if you would guide me through with resolving this problem. Thanks, N --------------------------- Newsgroups: comp.os.vxworks Subject: Re: strange timer bahaviour Date: 22 Oct 2002 16:17:37 +0100 From: Simon Wright Organization: Alenia Marconi Systems, ISD, Farlington Message-ID: References: <1034839098.567517@newsreader1.wirehub.nl> <1034867481.955098@newsreader2.wirehub.nl> <29b153d9.0210180725.2c8f96b@posting.google.com> <1035192099.254012@newsreader2.wirehub.nl> Sender: sjw@galadriel.frlngtn.gecm.com "Renee Teunissen" writes: > what we try to do is: > say one has a periodic timer with period times of 4 seconds, and our > system requires it to be changed (which is a part of the simulation > application). Let say we want it to become 5 seconds. > > now one can do two things. > 1) set the period time to the new value > "wait" for the current period to end > then start with the new periode time of 5 seconds. > > 2) another solution, if you are in a period, one could try to "modify" > the current period to make it fit in the new period time. I know this > might not aways possible, especially if one reduces the period time. > We thought might be a nice thing to do because we want to set the new > period values as soon as possible. In our simulation project we like > to have the new period time as soon as possible. > > But is this a legal thing to do? to "modify" the time a timer has > left, using th it_value parameter? Isn't there a third possibility? cancel the existing timer and set up a new one? this sounds much more likely to work to me! --------------------------- Newsgroups: comp.os.vxworks Subject: Re: mcp750 bsp probs with Tornado II Date: Tue, 22 Oct 2002 15:46:14 +0100 From: Dave Moore Organization: Computer Sciences Corporation Message-ID: <3DB564B6.326EA7B1@baesystem.com> References: <3DB54821.53EBE158@baesystem.com> <3DB54AEC.21B72C02@stest.ch> <3DB554AD.DD2E21F9@baesystem.com> Ok Guys. Apparently it's a bug - SPR 27046. Ta All, Dave Dave Moore wrote: > Marc Waber wrote: > > > Hi, > > > > afaik, vxWorks images under T2 are created in the directory > > ".../target/proj//default". The one in the > > config-direcrory is just a minimal image pre-configured by WindRiver and > > has nothing to do with the image you create using T2. > > > > Yes, I'm aware that the Tornado generated image will be in > .target/proj/mcp750/default' > > > > > I guess that you have a misconfiguration in your project. > > Since the vxWorks image is loaded, the bootrom seems to execute ok, and > > the vxWorks image has a problem. > > > > Well the Tornado users manual suggests that all Tornado II BSP come with an > already prepared project file for this purpose (ie. mcp750.wpj). However, > when I install the bsp (straight from the Tornado II Disk) there isn't one > there, so I guess I need to create one and thought I could by basing the new > project on the exiting working BSP. > > > > > Make sure your settings in the project tool of T2 are compatible to your > > hardware and to the "config.h" settings. > > If settings differ from "config.h" to the project tool, this might cause > > the application to crash!! > > I've looked in the wpj file with a text editor and there is a list of > INCLUDE_XXX whatever. So I guess I could use the Find Object.. function to > find all the additional INCLUDES that Tornado insists on putting in, so the > the list matches that in the configAll.h file?. --------------------------- Newsgroups: comp.os.vxworks Subject: Filter packets at i82559 Date: 22 Oct 2002 12:34:35 -0700 From: dannyc@harmonicinc.com (Danny Cohen) Organization: http://groups.google.com/ Message-ID: <41c1fd2b.0210221134.56ef78d7@posting.google.com> I'm using driver for i82559 and i'm tring to configure the ethernet controller to filter packets from reaching the s/w driver. I'm using the programmable filter+ flexible filter commands opcode=1000. Somehow it doesn't work. Did i miss anything? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: The infamous "No rule to make target" problem Date: Tue, 22 Oct 2002 20:32:38 -0400 From: joe durusau Organization: WEBUSENET.com Message-ID: <3DB5EE26.498E6F2@bellsouth.net> References: <3db57e4b$2@news.mhogaming.com> To the O.P: One way the works is to abandon the idea of using the GUI for building things, and use the makefile that comes in the BSP directory. You copy all the files you need into a single directory, gen up MACH_EXTRA (IIRC, not sure of the exact spelling), and away you go! You have to wean everybody away from the idea of having to click on something, but this method generally always works. I generally use sccs, and stick the makefile in sccs along with all the sources and suchlike. Speaking only for myself, Joe Durusau Graham Baxter wrote: > Are the slashes '/' the right way around in your include paths. > > -- > Regards, > > Graham Baxter > Freelance Software Engineer (VxWorks and pSOS BSPs) > Graham Baxter (Software) Limited > gbaxter@NOSPAMbcs.org.uk > > "Matt Osminer" wrote in message > news:3db57e4b$2@news.mhogaming.com... > > I seem to have come to an impasse with the infamous "No rule to make > target" > > problem. As any Tornado sufferer will tell you the solution should be > > > > 1. Add -I directives to your C/C++ settings > > 2. Add a VPATH directive to your makefile Macros from the Project Macros > Tab > > 3. Rebuild all dependencies, then Build-All > > > > Wondering if anybody has tried all of the above, as I have, and still been > > unable to get past the "No rule" error? My project was working just fine > > and then I did the unthinkable act of removing a few obsolete files and > > adding some new files. I didn't touch the -I or VPATH configurations and > my > > new files all fell within the bounds of the original settings so there > > shouldn't have been a problem. > > > > The weird thing is Tornado just seems to be having a hard time locating > this > > one directory. All my other -I directories appear just fine. > > > > I've tried: > > 1. Closing Tornado, rebuilding dependencies, saving the workspace, and > > "building-all" in about every combination you can think of. > > 2. Removing all the -I options and re-entering them > > 3. Removing the VPATH and re-entering it > > 4. Checked for directory existance etc. > > > > Any further ideas out there??? > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Sending a block of Data at MAC level Date: 22 Oct 2002 11:59:29 -0700 From: prashantwaknis@teledex.com (Prashant Waknis) Organization: http://groups.google.com/ Message-ID: Hi Have a question about Vxworks... Would appreciate help. How can I use VxWorks to send out a block of data on the network using VxWorks? I know the source MAC address and the destination MAC address. Thanks a lot. Prashant --------------------------- Newsgroups: comp.os.vxworks Subject: Re: USB camera in vxworks ? Re: USB Date: Wed, 23 Oct 2002 03:42:53 +0000 (UTC) From: Jeff Zheng Organization: Your Company Message-ID: References: <488e459a.0210212115.752d889c@posting.google.com> Usb bus is designed to be much software related. The usb controller is much like an ethernet controller. Usb device driver is much like socket users. And there must be some kind of protocol stack between them. I have tried to migrate linux usb stack to vxWorks. but havn't succeed yet. Jeff "Mike Kim" wrote in news:Tr7t9.77017$v_3.1624862@news.bora.net: > OK, I want to know alternative way using usb device. > But You said USB is a bus which has a complex bus. > > I know buying USB stack is a fast, convient way for me. > > But only I want to know if there is any other way about this. > > Thanks anyway for your respose. > > > -- > ####################################### > ChanSeong Kim > OFFICE : 02-829-3391 > FAX : 02-829-3400 > PCS : 016-660-6457 > KOREA COMPUTER BLDG. F9 > 395-65, SHNDAEBANG-DONG, > DONGJAK-KU > SEOUL, 156-010 KOREA > ####################################### > > "John" wrote in message > news:488e459a.0210212115.752d889c@posting.google.com... >> Hello, >> >> If the camera is a USB one, I'm not sure how you plan to avoid using >> a USB protocol stack - USB is not like a simple I/O port, it is a >> bus, and it has a fairly complex protocol running over it that must >> be used. >> >> You might be able to get other forms of camera - some older ones were >> parallel port based; you might be able to still get some of those, or >> you could look for a serial camera - some of the early Kodak digital >> cameras had serial port connections. You might also be able to write >> your own USB stack, or buy one from somewhere else... >> >> HTH, >> >> John... >> --------------------------- Newsgroups: comp.os.vxworks Subject: Source Browser in Tornado IDE Date: Wed, 23 Oct 2002 04:48:19 GMT From: "np" Organization: AT&T Broadband Message-ID: IS there a Source Browser (_Not the Target Browser_) in the Tornado IDE ( like in Visual Studio Tools->Source Browser). If yes how to build the project/workspace with Source Browser option? Thanks in Advance --------------------------- Newsgroups: comp.os.vxworks Subject: Re: The infamous "No rule to make target" problem Date: Wed, 23 Oct 2002 08:30:57 +0200 From: Emmanuel Herbreteau Organization: SEPRO Robotique (BEP) Message-ID: <3DB64221.764F83DA@sepro-robotique.com> References: <3db57e4b$2@news.mhogaming.com> Hi, Same problem : our Tornado project began to crash with the "no rule to make target" message. A close look at the generated makefile show the following lines : .../... $(PRJ_DIR)/programmation/instructionTimer.hpp \ $(PRJ_DIR)/programmation/ModifValeur.hpp \ $(PRJ_DIR)/programmation/clavierPLC.hpp \ {Tornado \ argetconfigsp823define_can.h} \ $(PRJ_DIR)/ihm2/FenetreManuel.hpp \ $(PRJ_DIR)/ihm2/FenetreRangement.hpp \ .../... This kind of error appears several times in the makefile. Obviously, there is a nasty bug in the makefile generator... (Tornado 2.0.2 CP4 PPC). It appears sometime when we try to add/remove a file in our project or when we try to check the dependencies. (80 files in our project) Then, to restore our project, we have to get the project file from our backup. WRS support just told us : "you have to buy the new Sniff project manager, which is, of course, much better than Tornado"... (Your car broke down ? Just buy a new car !) Funny, isn't it ? >8-( Regards Emmanuel Matt Osminer wrote: > I seem to have come to an impasse with the infamous "No rule to make target" > problem. As any Tornado sufferer will tell you the solution should be .../... > Any further ideas out there??? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Stack size Date: Wed, 23 Oct 2002 10:41:30 +0200 From: "Jakob Olsen" Organization: TDC Internet Message-ID: <3db660c8$0$97618$edfadb0f@dspool01.news.tele.dk> References: Hi Mohan Well, the hard way of doing it is to actually calculate a worst case stack size of your program. Nobody I know actually does this. The easy way is to set the stack size very high, to make sure that there is enough. Then let the task run for some time, stressing it a bit, to make sure it has been through all possible situations where a large stack is neccessary. Now, open a shell and use the command "ti NameOfYourTask" to see the task info of your task. This will reveal how big the tasks stack is, the maximum use, and what margin this yields. Now go back to your taskSpawn and adjust the stack size using the new information you just got. (I usually run with a stack size twice as big as needed, just to be sure) Jakob "Ram Mohan" wrote in message news:ap5ehc$s04$1@news.mch.sbs.de... > Hello, > > I am new to VxWorks. > > One of the taskSpawn() argument is stackSize. > > My questions are, > 1. How to guess the stack size ? Is there any procedure ? > 2. If the given stackSize is not sufficient for the task then what happens ? > > > Please help me to clarify about doubts, > > Thanks & Regards, > Mohan > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Stack size Date: Wed, 23 Oct 2002 10:41:44 +0200 From: "Jakob Olsen" Organization: TDC Internet Message-ID: <3db660c9$0$97618$edfadb0f@dspool01.news.tele.dk> References: Hi Mohan Well, the hard way of doing it is to actually calculate a worst case stack size of your program. Nobody I know actually does this. The easy way is to set the stack size very high, to make sure that there is enough. Then let the task run for some time, stressing it a bit, to make sure it has been through all possible situations where a large stack is neccessary. Now, open a shell and use the command "ti NameOfYourTask" to see the task info of your task. This will reveal how big the tasks stack is, the maximum use, and what margin this yields. Now go back to your taskSpawn and adjust the stack size using the new information you just got. (I usually run with a stack size twice as big as needed, just to be sure) Jakob "Ram Mohan" wrote in message news:ap5ehc$s04$1@news.mch.sbs.de... > Hello, > > I am new to VxWorks. > > One of the taskSpawn() argument is stackSize. > > My questions are, > 1. How to guess the stack size ? Is there any procedure ? > 2. If the given stackSize is not sufficient for the task then what happens ? > > > Please help me to clarify about doubts, > > Thanks & Regards, > Mohan > > > --------------------------- Newsgroups: comp.os.vxworks Subject: How to detect the failure of a hardisk or filesystem in vxWorks? Date: Wed, 23 Oct 2002 09:33:01 +0000 (UTC) From: ZhengZhe Organization: Your Company Message-ID: Hi everybody: We are developing a HA system and we are using harddisk. DosFs2 is much stable than its predecessor, but we are still experiencing hardisk failure (sometimes hardware failure). We need to detect runtime harddisk failures so we can handle them. Is there any kind of hook of something? Or how to detect the failure initiatively? Anybody has any idea? Jeff Zheng --------------------------- Newsgroups: comp.os.vxworks Subject: Stack size Date: Wed, 23 Oct 2002 11:44:11 +0530 From: "Ram Mohan" Organization: SISL Message-ID: Reply-To: "Ram Mohan" Hello, I am new to VxWorks. One of the taskSpawn() argument is stackSize. My questions are, 1. How to guess the stack size ? Is there any procedure ? 2. If the given stackSize is not sufficient for the task then what happens ? Please help me to clarify about doubts, Thanks & Regards, Mohan --------------------------- Newsgroups: comp.os.vxworks Subject: How to replace VxWorks TCP/IP Stack source Date: Wed, 23 Oct 2002 16:12:49 +0800 From: "Mike Chen" Organization: DCI HiNet Message-ID: Hi all, We have a project to replace VxWorks TCP/IP Stack with 3th Party code. But we have met the trouble about select() function, becasue it related to tty driver, anyone has any documents to avoid this trouble or ever replace VxWorks TCP/IP Stack. Please help us. Regards, Mike Chen --------------------------- Newsgroups: comp.os.vxworks Subject: Re: strange timer bahaviour Date: Wed, 23 Oct 2002 12:17:33 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1035368287.321157@newsreader2.wirehub.nl> References: <1034839098.567517@newsreader1.wirehub.nl> <1034867481.955098@newsreader2.wirehub.nl> <29b153d9.0210180725.2c8f96b@posting.google.com> <1035192099.254012@newsreader2.wirehub.nl> Well this _is_ possible - ofcourse. I tried it and it seem to work a bit better than without it. But stell get negative it_value.tv_sec results - dont know why. > > But is this a legal thing to do? to "modify" the time a timer has > > left, using th it_value parameter? > > Isn't there a third possibility? cancel the existing timer and set up > a new one? this sounds much more likely to work to me! --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Stack size Date: Wed, 23 Oct 2002 12:20:43 +0200 From: "Renee Teunissen" Organization: Wirehub! Internet, The Netherlands Message-ID: <1035368476.528369@newsreader2.wirehub.nl> References: <3db660c9$0$97618$edfadb0f@dspool01.news.tele.dk> What happens if you overrun the stack, is that your taskname will be destroyed, sinds the taskName is at the end of the stack.. or take a look at the vxworks faq, there is some sample code to detect stack overruns. Cheers, Renee - - - - - - - - - Renee A. Teunissen PTS Software bv, Meerweg 7, 1405BA Bussum, NL. T.+31-(0)35-6926969, M.+31-(0)6-22778313, http://www.pts.nl, @pts.nl personal link page: http://wittenburg10c.nl/db/dest/links.html "Jakob Olsen" wrote in message news:3db660c9$0$97618$edfadb0f@dspool01.news.tele.dk... > Hi Mohan > > Well, the hard way of doing it is to actually calculate a worst case stack > size of your program. Nobody I know actually does this. > The easy way is to set the stack size very high, to make sure that there is > enough. Then let the task run for some time, stressing it a bit, to make > sure it has been through all possible situations where a large stack is > neccessary. > Now, open a shell and use the command "ti NameOfYourTask" to see the task > info of your task. This will reveal how big the tasks stack is, the maximum > use, and what margin this yields. > Now go back to your taskSpawn and adjust the stack size using the new > information you just got. (I usually run with a stack size twice as big as > needed, just to be sure) > > Jakob > > "Ram Mohan" wrote in message > news:ap5ehc$s04$1@news.mch.sbs.de... > > Hello, > > > > I am new to VxWorks. > > > > One of the taskSpawn() argument is stackSize. > > > > My questions are, > > 1. How to guess the stack size ? Is there any procedure ? > > 2. If the given stackSize is not sufficient for the task then what happens > ? > > > > > > Please help me to clarify about doubts, > > > > Thanks & Regards, > > Mohan > > > > > > > > > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How to detect the failure of a hardisk or filesystem in vxWorks? Date: Wed, 23 Oct 2002 06:42:03 -0400 From: joe durusau Organization: WEBUSENET.com Message-ID: <3DB67CFB.73F7AF95@bellsouth.net> References: ZhengZhe wrote: > Hi everybody: > > We are developing a HA system and we are using harddisk. > DosFs2 is much stable than its predecessor, but we are still > experiencing hardisk failure (sometimes hardware failure). > We need to detect runtime harddisk failures so we can handle them. > Is there any kind of hook of something? Or how to detect the failure > initiatively? Anybody has any idea? > > Jeff Zheng How about when an IO request returns an error lookng at errno, analyzing the contents, then doing whatever you need to do. If you write your own drivers, you could make the driver send an int to kick off a special recovery routine, but that seems extreme. Speaking only for myself, Joe Durusau --------------------------- End of New-News digest ********************** From vxwexplo-errs@csg.lbl.gov Wed Oct 23 10:44:34 2002 From: "Small, Gene" Date: Wed Oct 23 10:44:35 PDT 2002 Subject: Relocation error when loading vxWorks modules I've recently added memory to one of my PPC 604 targets (from 32M to 40M) and I am now getting: ld (1,0,"fstream.o") Relocation value does not fit in 24 bits. Relocation value does not fit in 24 bits. Relocation value does not fit in 24 bits. Relocation value does not fit in 24 bits. when trying to load objects. Has anyone come across this or know how to fix it? Thanks, Gene Small From vxwexplo-errs@csg.lbl.gov Wed Oct 23 11:32:24 2002 From: Mike Anderson Date: Wed Oct 23 11:32:25 PDT 2002 Subject: Re: Relocation error when loading vxWorks modules Greetings! On Wed, 2002-10-23 at 13:52, the vxWorks Users Group Exploder wrote: > Submitted-by vxwexplo-errs@csg.lbl.gov Wed Oct 23 10:44:34 2002 > Submitted-by: "Small, Gene" > > I've recently added memory to one of my PPC 604 targets (from 32M to 40M) and I > am now getting: > > ld (1,0,"fstream.o") > Relocation value does not fit in 24 bits. Ahh, the infamous 24-bit relocation problem. This results from the EABI spec and the way jumps are called on the PPC. Typically, they use a branch local call which can only deal with a 24-bit address. However, if you compile the code with the -mlong_calls option, then you'll use the indirect jump via register call that allows for the full 32-bits. Also, you'll find that the code that you have will probably load just fine using the Tornado tools. That's because the T2 shell uses a memory cache in low RAM to load objects like the executable code. That's below the 24-bit boundary and hence, works just fine. HTH, Mike -- ======================================================= T h e P T R G r o u p, I n c. ======================================================= Embedded, Real-Time Solutions and Training Mike Anderson mailto:mike@ThePTRGroup.com Chief Scientist http://www.ThePTRGroup.com C: (703) 585-9384 V: (703) 430-3748 ======================================================= "Software development is like making a baby... You can't make a baby in one month by impregnating nine women. Some things just take time." From vxwexplo-errs@csg.lbl.gov Mon Oct 7 18:34:31 2002 From: "Matthew Peltzer" Date: Wed Oct 23 13:57:56 PDT 2002 Subject: x86 pci interrupt sharing Greetings, I am developing a device driver for custom PCI hardware under vxWorks 5.4/Tornado 2.0.2, and am experiencing difficulties in handling interrupts. I have no problem when the two devices are on their own interrupts, and I have no problem when using a shared interrupt independently. I am not able to have two threads at the same time listen to the same interrupt. According to the manual page for pciIntConnect(), I should not have to care about shared interrupts: DESCRIPTION This routine connects an interrupt handler to the PCI interrupt line(A - D). Link list is created if multiple handlers are assigned to the single PCI interrupt. To me this says that every time I call pciIntConnect(), my function and parameter will be added to a list that when the interrupt vector is triggered, each of my functions will be called with the appropriate parameter. Unfortunately, this is not what I am experiencing, although I am not 100% sure of what I am experiencing. It seems as though the second call to pciIntConnect() effectively wipes out the first parameter, for as soon as an interrupt is generated on this irq, my system deadlocks. I can only guess that this is because I am not turning off the interrupt on the correct card because my parameter is not what I expect it to be. My solution to this is to do my own isr chaining. I call pciIntConnect() and sysIntEnablePIC() only when no other connection has been made for this irq. pciIntDisconnect() and sysIntDisablePIC() are called only when no other device is listening to the irq. This setup has the same flaw as the previous: vxWorks reboots when two tasks are run that listen to a separate PCI card, yet share the same irq. I do not know if my callbacks are being called. the first operation in my callback is a logMsg(), but I am not seeing this. I also have some logMsg()s to tell me if each interrupt is being caught or ignored, which is decided by looking at the interrupt line on the hardware device as specified in the parameter passed into the isr by the system. I have yet to see an interrupt be ignored. This is the code I am using to enable interrupts: pciIntConnect(INUM_TO_IVEC((pci->irq + 0x20)), myDummyISR, list[pci->irq]); sysIntEnablePIC(pci->irq); 'pci' is a structure that among other things contains the irq number as read from the pci configuration space with: pciConfigInByte(pci->busNum, pci->devNum, pci->fncNum, 0x3c, &(pci->irq)); Is there anything that I am missing for irq sharing to work in vxWorks? Has anyone had any luck/mis-luck with similar methods? Matthew Peltzer goose23@spu.edu From vxwexplo-errs@csg.lbl.gov Wed Oct 23 13:57:55 2002 Received: from postal1.lbl.gov (postal1.lbl.gov [128.3.7.82]) by csg.lbl.gov (8.9.3/8.9.3) with ESMTP id NAA04384 for ; Wed, 23 Oct 2002 13:57:55 -0700 (PDT) Received: from postal1.lbl.gov (localhost [127.0.0.1]) by postal1.lbl.gov (8.11.2/8.11.2) with ESMTP id g9NKvsN17823 for ; Wed, 23 Oct 2002 13:57:54 -0700 (PDT) Received: from mailgw2a.lmco.com (mailgw2a.lmco.com [192.91.147.7]) by postal1.lbl.gov (8.11.2/8.11.2) with ESMTP id g9NKvqq17803 for ; Wed, 23 Oct 2002 13:57:53 -0700 (PDT) Received: from emss01g01.ems.lmco.com ([129.197.181.54]) by mailgw2a.lmco.com (8.11.6/8.11.6) with ESMTP id g9NKvph17091 for ; Wed, 23 Oct 2002 16:57:51 -0400 (EDT) Received: from CONVERSION-DAEMON.lmco.com by lmco.com (PMDF V6.1-1 #40643) id <0H4G00B018ONKC@lmco.com> for vxwexplo@lbl.gov; Wed, 23 Oct 2002 13:57:43 -0700 (PDT) Received: from EMSS09I00.us.lmco.com ([158.183.26.31]) by lmco.com (PMDF V6.1-1 #40643) with ESMTP id <0H4G00OZ4B7GX4@lmco.com> for vxwexplo@lbl.gov; Wed, 23 Oct 2002 13:21:20 -0700 (PDT) Received: by EMSS09I00.us.lmco.com with Internet Mail Service (5.5.2653.19) id ; Wed, 23 Oct 2002 16:21:16 -0400 Content-return: allowed Date: Wed, 23 Oct 2002 16:21:06 -0400 From: "Garner, David (N-Progeny)" Subject: 5.4 allocates significantly more memory on the heap compared to 5 .3 To: "'vxwexplo@lbl.gov'" Message-id: MIME-version: 1.0 X-Mailer: Internet Mail Service (5.5.2653.19) Content-type: text/plain Content-transfer-encoding: 7BIT We are migrating our apps to Tornado II with vxworks v5.4. Our system previously used Tornado I with vxworks v5.3. The problem is that when we issue memShow from the target shell with no startup script and nothing extra running that the allocated memory is 8 meg higher on T2 versus T1. This is causing a problem because we no longer have enough memory to load our objects. The two kernels are configured identically. Does anyone know why T2 uses 8 more meg from the heap? Is there a tool that will analyze blocks allocated on the heap and list the associated tasks? vxworks tornado Thanks, Dave Garner From vxwexplo-errs@csg.lbl.gov Thu Oct 24 04:03:21 2002 From: Vxworks Exploder Date: Thu Oct 24 04:03:22 PDT 2002 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Thu Oct 24 04:03:18 PDT 2002 Subject: Re: WDB Configuration error Subject: Re: Sending a block of Data at MAC level Subject: Linux Subject: How to solve "rpccore backend client timed out"? Subject: vision can't step thru rfi when i use mmu ... why not? Subject: Re: Linux Subject: Re: Linux Subject: compiling with -frepo and linking Subject: Re: Relocation error when loading vxWorks modules Subject: etherOutput and VLAN tagged frames Subject: Re: Linux Subject: considering vxWorks for ASIC/Networking task. Advice? Subject: Re: Linux Subject: Re: Relocation error when loading vxWorks modules Subject: Grrr, still having PCI problems Subject: SOS - how to put a short delay in a function Subject: Re: Grrr, still having PCI problems Subject: Re: vision can't step thru rfi when i use mmu ... why not? Subject: Re: SOS - how to put a short delay in a function Subject: CNF: Embedded Systems Club Autumn conference - last call for delegates Subject: Re: Linux Subject: Re: Source Browser in Tornado IDE Subject: Re: The infamous "No rule to make target" problem Subject: EtherOutput() Subject: building mips toolchain on cygwin Subject: Re: vision can't step thru rfi when i use mmu ... why not? Subject: Interrupt Handling and Scheduler Subject: Re: Linux Subject: Re: Source Browser in Tornado IDE Subject: Re: The infamous "No rule to make target" problem Subject: WDB Configuration error Subject: Re: Interrupt Handling and Scheduler Subject: Re: Grrr, still having PCI problems ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: WDB Configuration error Date: Wed, 23 Oct 2002 16:58:26 +0100 From: Wade Oram Organization: None Message-ID: References: Reply-To: Wade Oram In message , Vanessa Benfield writes >Hi all, > >I was wondering if anyone out there can help me!! > >I've created a vxWorks image by including a symbol table to a normal >vxWorks image and then building it to create a standalone system >image. > >I'm booting this standalone image up using a floppy disk which loads >vxWorks.st, however I get the following error: > >wdbConfig: error configuring WDB communication interface > >WDB: Agent configuration failed > >It also says that the muxLoad failed. > >Does anyone out there know how to boot a standalone image on a pc, >using a floppy disk?? - I'm new to vxWorks so any help would be very >much appreciated, as I can't seem to find any info about this anywhere >else! > >Thanks, Vanessa Benfield I can't be sure, but this sounds as if you need to add an "o=????" entry to your default boot line. The o= entry is used to specify a network interface to initialise when the image is being booted from a non-network device (in this case the floppy disk). - -- Wade Oram --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Sending a block of Data at MAC level Date: Wed, 23 Oct 2002 16:42:06 GMT From: "Steve Mowbray" Organization: blueyonder (post doesn't reflect views of blueyonder) Message-ID: References: You will not be able to create the entire ethernet frame in software as the max size MAC header and trailer are mostly fixed at the hardware level in the controller chip. As for sending raw network packets, I suggest you investigate the MUX interface. This will give you the full logical link layer functionality provided by the (presumably) commercial grade network driver that you have, this would be my tip for the easiest approach. If you want to get into the business of writing your own NIC device driver, this will be an order of magnitude harder than using the MUX spec for very little extra flexibility (if any). However if you are undeterred the Enhanced Network Driver (END) and Network Protocol Toolkit (NPT) are WindRiver standards that may serve as a starting point Hope this helps Steve "Prashant Waknis" wrote in message news:ea70aa8a.0210221059.7fa5cd84@posting.google.com... > Hi > > Have a question about Vxworks... Would appreciate help. > > How can I use VxWorks to send out a block of data on the network using VxWorks? > I know the source MAC address and the destination MAC address. > > Thanks a lot. > > Prashant --------------------------- Newsgroups: comp.os.vxworks Subject: Linux Date: 23 Oct 2002 10:27:40 -0700 From: brian_doe@hotmail.com (brian) Organization: http://groups.google.com/ Message-ID: Hi Folx, Has anyone out there been told by their company, or heard of others being told, that they can't use Linux, due to potential GPL issues? I realise that the debate goes on, however Im wondering what the large companies with lawyers think? Thanx in advance brian --------------------------- Newsgroups: comp.os.vxworks Subject: How to solve "rpccore backend client timed out"? Date: 23 Oct 2002 10:28:27 -0700 From: yangliang_mr@hotmail.com (Liang Yang) Organization: http://groups.google.com/ Message-ID: <4f57a77f.0210230928.67f140d1@posting.google.com> Hi, I added a device driver loader to vxWorks image, so when vxWorks image begin to run, it will load this device driver from usrAppInit.c. Though I found device driver is loaded successfully (from the message displayed in console window), I found target server can't connect to the target agent any more, it always repeat reporting error message as follows: "rpccore backend client timed out" Whar does this mean? Thanks. Liang --------------------------- Newsgroups: comp.os.vxworks Subject: vision can't step thru rfi when i use mmu ... why not? Date: Wed, 23 Oct 2002 17:38:20 GMT From: "Jim" Organization: Magma Communications Ltd. Message-ID: vision can't step thru rfi when i use mmu ... why not? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Linux Date: Wed, 23 Oct 2002 19:52:41 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Reply-To: "Leonid Rosenboim" "brian" wrote in message news:d2f67bd8.0210230927.164179ce@posting.google.com... > Hi Folx, > > Has anyone out there been told by their company, or heard of others > being told, that they can't use Linux, due to potential GPL issues? I > realise that the debate goes on, however Im wondering what the large > companies with lawyers think? > > Thanx in advance > > brian Brian, I am not working for any large company right now, but I find this sibject rather interesting too (BTW, why didn't you cross-post it to comp.os.linux ?). I think that it is somewhat risky to makie a commercial product on Linux code because the GPL language states that all modifications you make must return to the public domain. There are certain ways used by different companies (including WIndRiver who use GPL'ed code in the tool chain) that think the have figured it out (i.e. separate their proprietary stuff from GPL stuff), but since this has never been tested at court, every lawyer will look at it with a certain amount of distrust. This GPL for a lawyer is like a program which compiles nicely but has never been tested, because it has bever been to court, and because it is in fact very different from any other license in the industry. Most licensing disputes (as well as patent disputes) end up in a negotiated compromise, but with this GPL, who are you going to negotiate with ?? The code has been written by hundreds (if not thousands) of poeple, how on earth can anyone track them all and have them sign a compromise agreement ? It is interesting what otehr people have to say on this, but even more interesting is the opinion of their (and your) company lawyer ! Cheers! Leonid --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Linux Date: Wed, 23 Oct 2002 20:43:11 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Reply-To: "Leonid Rosenboim" Here are some useful links I digged up on Google, interesting stuff. http://www.eweek.com/print_article/0,3668,a=28000,00.asp http://www.cs.helsinki.fi/u/campa/teaching/oss/papers/jarvinen.pdf http://www.uwsg.iu.edu/hypermail/linux/kernel/9807.1/0722.html http://www.econ.umn.edu/~mboldrin/Issues/OpenText/gpl_michele.html But this is possibly the best (Jeff Espin who write this, and Curt Chacker who is mentioned both used to work for WindRiver) http://www.linuxdevices.com/articles/AT9161119242.html --------------------------- Newsgroups: comp.os.vxworks Subject: compiling with -frepo and linking Date: 23 Oct 2002 11:44:42 -0700 From: vprashant@illinoisalumni.org (V Prashant) Organization: http://groups.google.com/ Message-ID: <25050bb6.0210231044.5bda6b7@posting.google.com> I am using Tornado2.2 (Host->Windows, Target->MIPS) and compiling my project with the -frepo option. Though it generates .rpo files, it does not seem to link them using "collect2" as mentioned in the gcc documentation. However, when I change the linker to ccmips instead of ldmips, collect2 is invoked, though it faitfully crashes... ccmips -EB -r -Lc:\tornado\host\x86-win32\lib\gcc-lib\mips-wrs-vxworks\gcc-2.96\Mips32Sfgnu @c:\proj\ilon\ilon100\v1.0\rni\proj\prjObjs.lst -o partialImage.o (C:\Tornado\host\x86-win32\lib\gcc-lib\mips-wrs-vxworks\gcc-2.96\collect2.exe 1001) Exception: STATUS_ACCESS_VIOLATION ccmips: Internal compiler error: program collect2 got fatal signal 11 make: *** [RNI.out] Error 0x1 Done. I am trying to avoid duplicate template instantiation using the -frepo option but just cannot get it to work. Any help/pointers will be appreicated. Thanks Prashant --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Relocation error when loading vxWorks modules Date: Wed, 23 Oct 2002 11:51:50 -0700 From: "Michael R. Kesti" Organization: MK Associates Message-ID: <3DB6EFC6.C7BF6364@gv.net> References: Reply-To: mkesti@gv.net "Small, Gene" wrote: >I've recently added memory to one of my PPC 604 targets (from 32M to 40M) and I >am now getting: > >ld (1,0,"fstream.o") >Relocation value does not fit in 24 bits. >Relocation value does not fit in 24 bits. >Relocation value does not fit in 24 bits. >Relocation value does not fit in 24 bits. > >when trying to load objects. Has anyone come across this or know how to fix it? The following is quoted from the WRSBC8240 BSP's target.nr file: - --- The 64Meg is divided into two halves. The first 32Meg is used for code and data, and the second 32Meg is specified as USER_RESERVED_MEM memory. This is done to accomodate the 24 bit address EABI compilation option. This results in the compiler generating 'bl' branch instructions which increases performance but places a 32Meg address limit on the code. There are two ways to 'work around' this limitation. 1.) recompile all the source with the -mlongcall compilation flag, which results in poorer performance, or, 2.) add remaining 32Meg to the memory pool at execution time via the memAddToPool() function. Option #1 would require all the libraries, driver, and the BSP to be recompiled with the -mlongcall compiler option. This results in a number of code changes; the most obvious code change is branches are implemented via brlr instruction verses the bl instruction. Option #2, used by the BSP, restores the upper 32Meg of RAM to the heap, making it available via malloc() and such. See WindTech Note WTN41 for details. - --- I suspect that this or something very much like it applies to your target and BSP, as well. - -- ======================================================================== Michael Kesti | "And like, one and one don't make | two, one and one make one." mkesti@gv.net | - The Who, Bargain --------------------------- Newsgroups: comp.os.vxworks Subject: etherOutput and VLAN tagged frames Date: Wed, 23 Oct 2002 21:18:23 +0200 From: "Ofer Goren" Message-ID: <3db6f69d$1@news.012.net.il> Hi. Can I use etherOutput() to send VLAN-tagged frames? (Ethertype 0x8100)? Thanks, Ofer. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Linux Date: Wed, 23 Oct 2002 21:25:33 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: Reply-To: "Leonid Rosenboim" Also, feel free to join in to the discussion on the same matter on Yahoo Message boards-> WIND http://messages.yahoo.com/bbs?.mm=FN&action=l&board=4688317&tid=wind&sid=468 8317&mid=22111 --------------------------- Newsgroups: comp.os.vxworks Subject: considering vxWorks for ASIC/Networking task. Advice? Date: Wed, 23 Oct 2002 15:41:43 -0400 From: "Warren Postma" Organization: Bell Sympatico Message-ID: What we are looking for is an RTOS that can be used for both (a) traditional development board work on something like an Intel ARM based CPU Reference/Development Board, and also (b) run inside a single chip ASIC "System on Chip" (SOC) designed chipset product. Unfortunately, most ASIC-based RTOS solutions don't have proper TCP/IP stacks, and most full featured RTOSs don't scale down for ASIC/SOC designs. The choice of RTOS is for two separate projects which should both use the same RTOS to permit some code-sharing between them. One is on a dev board, one is on an ASIC. The tough thing is vxWorks might be too big for an ASIC design. We can afford maybe 128K for a kernel and a full TCP/IP stack with Ethernet, PPPoE, and PPP, leaving 128K in our ASIC for our application code. What would a kernel size (ballpark in 100s of kbs) for a vxWorks kernel with TCP/IP stack, and without TCP/IP, on say an ARM9 or StrongARM/XScale CPU. Is anyone using vxWorks in ASIC/SOC or Soft-core designs (ARM9 IP block inside FPGA, for instance)? Does anyone have any other RTOS suggestions? Something with ASIC ultra-tiny potential, but a decent TCP/IP stack (full TCP/IP source code provided with SDK preferred) and good hardwdare debugging. I know vxWorks has great toolsets, and is used in network processing (NP) applications, but I haven't heard of it being used in SOC single-chip embedded systems. Thanks for any help or pointers. Warren Postma Toronto Canada --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Linux Date: Wed, 23 Oct 2002 14:42:01 -0500 From: Andrew Johnson Organization: APS, Argonne National Laboratory Message-ID: References: Leonid Rosenboim wrote: > "brian" wrote: > >>Has anyone out there been told by their company, or heard of others >>being told, that they can't use Linux, due to potential GPL issues? I >>realise that the debate goes on, however Im wondering what the large >>companies with lawyers think? I don't work in a large company, but I can point you to lots of such companies that are selling products on Linux - Oracle would be one extremely prominent example. > I think that it is somewhat risky to makie a commercial product on Linux > code > because the GPL language states that all modifications you make must return > to the public domain. To be pedantic about this, you are _not_ forced to release your changes into the public domain, but if you're selling (or even giving away) a product that makes use of GPL code you have to follow the rules of the GPL, which means providing source code to your customers and licensing your changes under the GPL. However it is legal to mix GPL and proprietary code in the same product (although not in the same program), so you're not forced to give away your secrets provided you're careful about your design. > There are certain ways used by different companies > (including WIndRiver who use GPL'ed code in the tool chain) that think > the have figured it out (i.e. separate their proprietary stuff from GPL > stuff), The Linux license includes a specific waiver which permits non-GPL drivers to be loaded into the kernel, thus you can include specialist hardware in a Linux-based system without too much difficulty and still retain all rights to your specialist driver code. Application programs running on Linux have never been forced to use the GPL (unless they also use some other GPL code or library) - the system call interface between User and Kernel code is a GPL barrier. You can guarantee that some very well-paid lawyers have accepted the latter or there's no way you'd ever have been able to buy Oracle running on Linux. > This GPL for a lawyer is like a program which compiles nicely but has never > been tested, because it has bever been to court, and because it is in fact > very different from any other license in the industry. The GPL has never been tested in court because it's seen to be legally pretty water-tight - I believe the Free Software Foundation's legal adviser Eben Moglen would love to test it, but nobody will take him on! Every case of violation of GNU's copyright has been resolved to the satisfaction of the FSF by negotiation, either by the violator removing the GPL code from their proprietary product or by licensing their product under the GPL. Of course not all GPL software is copyrighted (i.e. owned) by the FSF. In the case of the Linux kernel, the copyright subsists with the individual authors or their employers, some of whom are big companies with Legal Departments (e.g. IBM, HP, SGI): > Most licensing disputes (as well as patent disputes) end up in a negotiated > compromise, but with this GPL, who are you going to negotiate with ?? The > code > has been written by hundreds (if not thousands) of poeple, how on earth can > anyone track them all and have them sign a compromise agreement ? Hence when dealing with a GPL violation, the GPL copyright owners never compromise; the violator always fixes the problem one way or the other, and _usually_ they get forgiven. If your product can't [be made to] work without the GPL code then you'd better fix the violation and get forgiven or else you've just thrown all your work on the scrap-heap (because while you're in violation the GPL explicitly denies you the right to resditribute the GPL code). However it's not actually that hard to work out how to stay within the law, and lots of companies are doing so, you just have to be careful about what libraries you use, and ensure your programmers understand the issues properly so they don't go stealing chunks of code from GPL applications. In some cases a company's main area of interest isn't in software anyway but in selling the hardware it runs on or interfaces with, so releasing the code under the GPL isn't actually much of a hardship, and can often result in a better product than with closed-source software. - - Andrew - -- "Life is what happens while you're busy making other plans." - John Lennon --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Relocation error when loading vxWorks modules Date: Wed, 23 Oct 2002 14:45:40 -0500 From: Andrew Johnson Organization: APS, Argonne National Laboratory Message-ID: References: Small, Gene wrote: > I've recently added memory to one of my PPC 604 targets (from 32M to 40M) and I > am now getting: > > ld (1,0,"fstream.o") > Relocation value does not fit in 24 bits. > Relocation value does not fit in 24 bits. > Relocation value does not fit in 24 bits. > Relocation value does not fit in 24 bits. > > when trying to load objects. Has anyone come across this or know how to fix it? http://www.aps.anl.gov/epics/modules/base/ppc.php#Relocation - - Andrew - -- "Life is what happens while you're busy making other plans." - John Lennon --------------------------- Newsgroups: comp.os.vxworks Subject: Grrr, still having PCI problems Date: 23 Oct 2002 13:49:58 -0700 From: dj_thorrin71@yahoo.com (Johnathan Ross) Organization: http://groups.google.com/ Message-ID: <40cb3a14.0210231249.733d9228@posting.google.com> Hey guys, I am still having problems sending data across the PCI bus. We are on a Pentium target and are trying to send up to 32k of data across the PCI bus to another card. The data gets to the card, but it is very slow. I have looked at some of the PCI registers on the Pentium (Bus Master) and I see a register called latency time which determines how long the master can hold the PCI bus. I have tried raising and lowering this value, but the times are still the same. Any other registers that I can try? I would think that there aren't any registers to controll the speed of data going across the bus, you just tell it to move a block of data. I was hoping that by raising the latency timer, the host would grab control of the bus and not let go, but I am seeing no change in the amount of time to send the data. Basic flow of the code is: memDevCreate (READMEM, 0, sysMemTop()); fd_read = open ("/mem/read/0", O_RDONLY, 0); ... ioctl (fd_read, FIOSEEK, address); read (fd_read, ptr_message, length); basically trying to copy data from one memory address to another (address to ptr_message). Address is on the PCI card and ptr_message is on the host. Any ideas will be greatly appreciated. - -Johnathan Ross --------------------------- Newsgroups: comp.os.vxworks Subject: SOS - how to put a short delay in a function Date: Wed, 23 Oct 2002 15:56:47 -0500 From: "L" Message-ID: Hi, I'm new in VxWorks, and would appreciate it if someone could tell me how I can easily put a short delay (5 seconds) inside a function. It seems VxWorks doesn't have sleep( ) function, is it right? Thanks, L. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Grrr, still having PCI problems Date: Wed, 23 Oct 2002 21:31:29 GMT From: "H. E. Taylor" Organization: MTS Internet Message-ID: <3DB72F84.2D03@despam.autobahn.mb.ca> References: <40cb3a14.0210231249.733d9228@posting.google.com> In article <40cb3a14.0210231249.733d9228@posting.google.com>, Johnathan Ross wrote: > > Hey guys, > > I am still having problems sending data across the PCI bus. We are on > a Pentium target and are trying to send up to 32k of data across the > PCI bus to another card. The data gets to the card, but it is very > slow. > I have run into this in the past, as well. > I have looked at some of the PCI registers on the Pentium (Bus Master) > and I see a register called latency time which determines how long the > master can hold the PCI bus. I have tried raising and lowering this > value, but the times are still the same. > Same here. > Any other registers that I can try? I would think that there aren't > any registers to controll the speed of data going across the bus, you > just tell it to move a block of data. I was hoping that by raising > the latency timer, the host would grab control of the bus and not let > go, but I am seeing no change in the amount of time to send the data. > > Basic flow of the code is: > > memDevCreate (READMEM, 0, sysMemTop()); > fd_read = open ("/mem/read/0", O_RDONLY, 0); > ... > ioctl (fd_read, FIOSEEK, address); > read (fd_read, ptr_message, length); > > basically trying to copy data from one memory address to another > (address to ptr_message). Address is on the PCI card and ptr_message > is on the host. > > Any ideas will be greatly appreciated. > I'm not sure why you are messing with a memDev, but no matter. I used to use base registers to find a buffer and set up a pointer and just execute a read/write increment loop (which is probably what the memDev does). Is there a CPU on the PCI card? If so, ask your hardware guys what address arbitration has to take place on every read/write. The only way I saw this operation running fast was using DMA, and that I think was because the above arbitration only took place once per block. - -het - -- "progress in software has not followed Moore's law." -John Holland Computer Links: http://www.autobahn.mb.ca/~het/clinks.html H.E. Taylor http://www.autobahn.mb.ca/~het/ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: vision can't step thru rfi when i use mmu ... why not? Date: Wed, 23 Oct 2002 22:29:05 GMT From: "Jim" Organization: Magma Communications Ltd. Message-ID: References: Seems like the visionICE is no good for debugging OS's with memory protection ... "Your visionICE should work just as well as the Abatron for debugging your 2K bootloader and PPCBoot. Where it falls apart is when the MMU gets turned on ..." Jim "Jim" wrote in message news:g8Bt9.16396$V21.452582@news... > vision can't step thru rfi when i use mmu ... why not? > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: SOS - how to put a short delay in a function Date: Wed, 23 Oct 2002 15:56:16 -0700 From: "Michael R. Kesti" Organization: MK Associates Message-ID: <3DB72910.DF77AF45@gv.net> References: Reply-To: mkesti@gv.net L wrote: > > Hi, > > I'm new in VxWorks, and would appreciate it if someone could tell me how I > can easily put a short delay (5 seconds) inside a function. It seems > VxWorks doesn't have sleep( ) function, is it right? > > Thanks, > > L. #include "vxworks.h" #include "taskLib.h" #include "sysLib.h" void someFunction ( void ) { taskDelay(sysClkRateGet() * 5); } - -- ======================================================================== Michael Kesti | "And like, one and one don't make | two, one and one make one." mkesti@gv.net | - The Who, Bargain --------------------------- Newsgroups: comp.os.vxworks Subject: CNF: Embedded Systems Club Autumn conference - last call for delegates Date: Thu, 24 Oct 2002 00:15:18 +0100 From: "John Robinson" Organization: John Robinson And Associates Ltd Message-ID: Followup-To: comp.os.vxworks Reply-To: "John Robinson" The Embedded Systems Club Autumn conference is being held in Swindon, UK on Tuesday 29th October 2002. Late bookings are still being taken at http://www.EmbeddedSystemsClub.com for this low cost conference and exhibition, which has already attracted over 110 attendees. The conference features 2 streams of 22 sessions, including several sessions designed to give delegates the opportunity to play an active role in the conference. Two tutorials on "Test-First Design" (Kevlin Henney, Curbralan) and "Principles of High Availability Embedded Systems Design" (David Kalinsky, OSE) provide a training element to the event, while technical presentations by Silver Software, Praxis, ADI, IPL, ACT Europe, Programming Research and I-Logix help to make this a conference with significant technical content. The exhibition has attracted key players in the embedded systems markets including Wind River, Vsystems, Telelogic, Silver Software, Programming Research, Praxis, PolySpace, OSE, Objektum, LDRA, Kennedy Carter, John Robinson & Associates, IPL, I-Logix, Headway Software, Green Hills Software, First Matrix, ARTiSAN and Aonix. At the exhibition a number of product vendors will announce and demonstrate new products and/or significant product upgrades, including IPL, Telelogic, Rational Software, Headway Software, PolySpace Technologies, First Matrix and OSE. The mix of exhibitors, sponsors, speakers and delegates attracted to this event represent a wide spread of interests across the embedded systems industry. You can still join us, by registering via our website: http://www.EmbeddedSystemsClub.com To receive information on our 2003 events, register as an associate member via the same websites. Associate membership is free of charge. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Linux Date: Thu, 24 Oct 2002 08:27:19 +0900 From: "Girish Gulawani" Message-ID: <3db73049$0$19955$44c9b20d@news3.asahi-net.or.jp> References: > Has anyone out there been told by their company, or heard of others > being told, that they can't use Linux, due to potential GPL issues? I > realise that the debate goes on, however Im wondering what the large > companies with lawyers think? yes i have been told that Linux has GPL problem & that can not be used in products. i do not have any experiences of lawyers explaining this, but the boss says that is end of our idea to put linux. we continue to drag on with poor vxworks features. i think wrs should atleast consider re-writing kernel if not open source. processors are giving newer newer features but vxworks continues to be same old kernel-mode-only & no-mmu. IMHO, this forum deserves this discussion more. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Source Browser in Tornado IDE Date: Wed, 23 Oct 2002 09:50:11 +0200 From: "Martin Raabe" Organization: 1&1 Internet AG Message-ID: References: Hello np, yes there is a Source Browser available as a Wind Power Tool: SNiFF+ To create a SNiFF+ project, press the Tools>SNiFF+ button and you get it done automatically. You may also start your project in SNiFF+. For more details ask WIndRiver for SNoFF+ Version 4.1. Hope it helps. Ciao - -- Martin Raabe E: Martin.RaabeATNOSPAMBaSystem.de "np" schrieb im Newsbeitrag news:nSpt9.64490$La5.202635@rwcrnsc52.ops.asp.att.net... > IS there a Source Browser (_Not the Target Browser_) in the Tornado IDE ( > like in Visual Studio Tools->Source Browser). If yes how to build the > project/workspace with Source Browser option? > > Thanks in Advance > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: The infamous "No rule to make target" problem Date: Wed, 23 Oct 2002 09:51:44 +0200 From: "Jakob Olsen" Organization: TDC Internet Message-ID: <3db65511$0$97648$edfadb0f@dspool01.news.tele.dk> References: <3db57e4b$2@news.mhogaming.com> Hi Matt. We have run extensive research into this problem in the company I am working in. Here's a few tips: 1. Always use forward slashes in -I directives. use /target/h not \target\h. 2. Remove the "-I." that Tornado adds to projects automatically, and which does searches in the "current directory" that can be pretty undefined. 3. Make sure that the file extension of all included files is lowercase ( foo.h not foo.H ). TOrnado has a bad habit of just ignoring files with uppercase extensions. Following these rules very strictly has eliminated the problem in all our projects. Try it out. Jakob "Matt Osminer" wrote in message news:3db57e4b$2@news.mhogaming.com... > I seem to have come to an impasse with the infamous "No rule to make target" > problem. As any Tornado sufferer will tell you the solution should be > > 1. Add -I directives to your C/C++ settings > 2. Add a VPATH directive to your makefile Macros from the Project Macros Tab > 3. Rebuild all dependencies, then Build-All > > Wondering if anybody has tried all of the above, as I have, and still been > unable to get past the "No rule" error? My project was working just fine > and then I did the unthinkable act of removing a few obsolete files and > adding some new files. I didn't touch the -I or VPATH configurations and my > new files all fell within the bounds of the original settings so there > shouldn't have been a problem. > > The weird thing is Tornado just seems to be having a hard time locating this > one directory. All my other -I directories appear just fine. > > I've tried: > 1. Closing Tornado, rebuilding dependencies, saving the workspace, and > "building-all" in about every combination you can think of. > 2. Removing all the -I options and re-entering them > 3. Removing the VPATH and re-entering it > 4. Checked for directory existance etc. > > Any further ideas out there??? > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: EtherOutput() Date: 23 Oct 2002 17:32:01 -0700 From: prashantwaknis@teledex.com (Prashant Waknis) Organization: http://groups.google.com/ Message-ID: Hi I have a block of data that I would like to multicast to the network. I tried to use EtherOut() in the Ethernet Library (EtherLib)... Using EtherLib, it seems that I can send a packet to a specific MAC address, or to a broadcast address (FFFFFFFFFFFF). But I can't send it to a multicast MAC address (such as 01000ccccccc). Can someone tell me what in VxWorks I should be using to achieve this? Thanks. Prashant --------------------------- Newsgroups: comp.os.vxworks Subject: building mips toolchain on cygwin Date: 23 Oct 2002 20:05:58 -0700 From: tckchuang@yahoo.com (Ken) Organization: http://groups.google.com/ Message-ID: <3db17119.0210231905.30abff29@posting.google.com> I am building a mips toolchain on cygwin environmet using the toolchain source codes whcih has been successfully built under linux. However, I got stuck when first compiling the gcc. The error message is like the following: mipsel-linux-ar: not found mv tmplibgcc2.a libgcc2.a mv: cannot stat 'tmplibgcc2.a': No such file or directory make[1]: ***[libgcc2.a] Error 1 make[1]: Leaving dircetory '/cygdrive/c/....' make: *** [all-gcc] Error 2 This kind of problem was happened when I built an ARM toolchain on cygwin. Would someone please help me to go over the problem? Thanks a lot. Ken --------------------------- Newsgroups: comp.os.vxworks Subject: Re: vision can't step thru rfi when i use mmu ... why not? Date: 23 Oct 2002 22:24:32 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0210232124.649c9b07@posting.google.com> References: Hello, For those that are interested, this is due to the mechanism that the JTAG debug engine in the CPU uses sharing the critical save registers (SRR0 and SRR1) with the CPU itself. These registers are used during exception/interrupt handling to hold the PC and MSR values; rfi then restores the PC and MSR in one operation from the SRR0 and SRR1 save registers. If you try to step through this, the ICE will 'corrupt' the values that the OS was trying to restore to the PC and MSR. Normally this just results in the rfi becoming a block in the stepping; each step takes you back to the rfi instruction. You should be able to configure the VisionICE (and probably other versions of these debug attachments) to ignore the exception handlers when stepping, and not to interrupt the CPU when it is in the non-recoverable state (i.e. before the OS has saved the SRR0 and SRR1 values). Turning the MMU on will certainly make this more likely to happen on the cores that have software miss handling (860, 603 and 82xx). That is because the software miss handlers never save SRR0 and SRR1 to keep their performance as high as possible, hence they remain in non-recoverable state for a longer time than most handlers, increasing the probability of stopping the CPU while it is in a miss handler. It could also happen in interrupt handlers, or any other exception handlers though, even with the MMU disabled. HTH, John... "Jim" wrote in message news:... > Seems like the visionICE is no good for debugging OS's with memory > protection ... > > "Your visionICE should work just as > well as the Abatron for debugging your 2K bootloader and > PPCBoot. Where it falls apart is when the MMU gets turned on ..." > > Jim > > "Jim" wrote in message > news:g8Bt9.16396$V21.452582@news... > > vision can't step thru rfi when i use mmu ... why not? > > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Interrupt Handling and Scheduler Date: 23 Oct 2002 22:48:48 -0700 From: anjaneyulu@mistralsoftware.com (vemanj) Organization: http://groups.google.com/ Message-ID: <53c6857a.0210232148.5268569a@posting.google.com> Hi, I have a doubt about the interrupt handling and the scheduling. When an ISR is being executed, that particular interrupt is disable. My question is :if the interrupts are coming at very high frequency, the other tasks will never get the chance to execute. Is it right??? Please tell me the scheduler reaction to an interrupt. What happens to the current execution???? Thanks in advance. Regards, Anj --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Linux Date: 23 Oct 2002 23:37:21 -0700 From: john_94501@yahoo.com (John) Organization: http://groups.google.com/ Message-ID: <488e459a.0210232237.165e1a7e@posting.google.com> References: Hello, There's a new white paper on Wind River's home page about Linux too written by Jerry Fiddler. HTH, John... "Leonid Rosenboim" wrote in message news:... > Here are some useful links I digged up on Google, interesting stuff. > > http://www.eweek.com/print_article/0,3668,a=28000,00.asp > http://www.cs.helsinki.fi/u/campa/teaching/oss/papers/jarvinen.pdf > http://www.uwsg.iu.edu/hypermail/linux/kernel/9807.1/0722.html > http://www.econ.umn.edu/~mboldrin/Issues/OpenText/gpl_michele.html > > But this is possibly the best (Jeff Espin who write this, and Curt Chacker > who > is mentioned both used to work for WindRiver) > http://www.linuxdevices.com/articles/AT9161119242.html --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Source Browser in Tornado IDE Date: Thu, 24 Oct 2002 11:45:28 +0200 From: Red Fenris Organization: No way - anarchy makes me feel better ;) Message-ID: References: Reply-To: fenris@netcourrier.com We use Source insight from Source Dynamics (www.sourceinsight.com) and it's really fine, even if it's a commercial program np wrote: > IS there a Source Browser (_Not the Target Browser_) in the Tornado IDE ( > like in Visual Studio Tools->Source Browser). If yes how to build the > project/workspace with Source Browser option? > > Thanks in Advance > > - -- - ------------------------------------- Support third world debt cancellation and reduction of military spendings - ------------------------------------- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: The infamous "No rule to make target" problem Date: Thu, 24 Oct 2002 11:47:12 +0200 From: Red Fenris Organization: No way - anarchy makes me feel better ;) Message-ID: References: <3db57e4b$2@news.mhogaming.com> Reply-To: fenris@netcourrier.com > Any further ideas out there??? > Yes - forget about the infamous Tornado and make your own makefiles from those producted by the tcl environment Speaking for myself only - ------------------------------------- Support third world debt cancellation and reduction of military spendings - ------------------------------------- --------------------------- Newsgroups: comp.os.vxworks Subject: WDB Configuration error Date: 23 Oct 2002 03:58:02 -0700 From: vanessabenfield98@hotmail.com (Vanessa Benfield) Organization: http://groups.google.com/ Message-ID: Hi all, I was wondering if anyone out there can help me!! I've created a vxWorks image by including a symbol table to a normal vxWorks image and then building it to create a standalone system image. I'm booting this standalone image up using a floppy disk which loads vxWorks.st, however I get the following error: wdbConfig: error configuring WDB communication interface WDB: Agent configuration failed It also says that the muxLoad failed. Does anyone out there know how to boot a standalone image on a pc, using a floppy disk?? - I'm new to vxWorks so any help would be very much appreciated, as I can't seem to find any info about this anywhere else! Thanks, Vanessa Benfield --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Interrupt Handling and Scheduler Date: Thu, 24 Oct 2002 12:37:14 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: References: <53c6857a.0210232148.5268569a@posting.google.com> Reply-To: "Leonid Rosenboim" "vemanj" wrote in message news:53c6857a.0210232148.5268569a@posting.google.com... > Hi, > I have a doubt about the interrupt handling and the scheduling. > When an ISR is being executed, that particular interrupt is disable. > > My question is :if the interrupts are coming at very high frequency, > the other tasks will never get the chance to execute. Is it right??? > > Please tell me the scheduler reaction to an interrupt. What happens to > the current execution???? > > Thanks in advance. > > Regards, > Anj I dont want to be rude, but this is a very beginner-type question. Handling of interrupts is not an issue of any particular scheduler or operating system - it is how the processor (and associated interrupt controller) is handling interrupts, this is why it is sometimes called a "hardware interrupt" You are basically right, that when an ISR take precedence over all tasks in a system, and over interrupts that have lesser hardware priority. High priority interrupts can still occur if your hardware supports nested interrupts. You are also right that if interrupts happen at an extremely high frequency (depends on processor, but surely over 100kHz), there could be very little or no processor time at all left to execute any tasks. - -- - ----------------------------------------------------------------------- Leonid Rosenboim Visit: http://gamla.org.il/english/index.htm Consultant Email: my first name at consultant dot com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Grrr, still having PCI problems Date: 24 Oct 2002 03:39:37 -0700 From: dj_thorrin71@yahoo.com (Johnathan Ross) Organization: http://groups.google.com/ Message-ID: <40cb3a14.0210240239.22471532@posting.google.com> References: <40cb3a14.0210231249.733d9228@posting.google.com> <3DB72F84.2D03@despam.autobahn.mb.ca> Thanks for the response. Yes, there is a processor on the PCI card. We actually had a hardware guy plug in some sort of PCI bus card connected to a laptop. This bus card was in a vacant slot on the target in question. He was able to burst data to the processor on the PCI card from his laptop so I know there is no problem with the settings on the PCI card. I just cant get things to work effeciently from that target locally. Where can I find an example of a DMA transfer? - -JR "H. E. Taylor" wrote in message news:<3DB72F84.2D03@despam.autobahn.mb.ca>... > In article <40cb3a14.0210231249.733d9228@posting.google.com>, > Johnathan Ross wrote: > > > > Hey guys, > > > > I am still having problems sending data across the PCI bus. We are on > > a Pentium target and are trying to send up to 32k of data across the > > PCI bus to another card. The data gets to the card, but it is very > > slow. > > > > I have run into this in the past, as well. > > > I have looked at some of the PCI registers on the Pentium (Bus Master) > > and I see a register called latency time which determines how long the > > master can hold the PCI bus. I have tried raising and lowering this > > value, but the times are still the same. > > > > Same here. > > > Any other registers that I can try? I would think that there aren't > > any registers to controll the speed of data going across the bus, you > > just tell it to move a block of data. I was hoping that by raising > > the latency timer, the host would grab control of the bus and not let > > go, but I am seeing no change in the amount of time to send the data. > > > > Basic flow of the code is: > > > > memDevCreate (READMEM, 0, sysMemTop()); > > fd_read = open ("/mem/read/0", O_RDONLY, 0); > > ... > > ioctl (fd_read, FIOSEEK, address); > > read (fd_read, ptr_message, length); > > > > basically trying to copy data from one memory address to another > > (address to ptr_message). Address is on the PCI card and ptr_message > > is on the host. > > > > Any ideas will be greatly appreciated. > > > > I'm not sure why you are messing with a memDev, but no matter. > I used to use base registers to find a buffer and set up a > pointer and just execute a read/write increment loop (which > is probably what the memDev does). > > Is there a CPU on the PCI card? > If so, ask your hardware guys what address arbitration has to take > place on every read/write. > > The only way I saw this operation running fast was using DMA, and > that I think was because the above arbitration only took place once > per block. > > -het --------------------------- End of New-News digest ********************** From vxwexplo-errs@csg.lbl.gov Fri Oct 25 04:05:48 2002 From: Vxworks Exploder Date: Fri Oct 25 04:05:49 PDT 2002 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Fri Oct 25 04:05:45 PDT 2002 Subject: Re: Grrr, still having PCI problems Subject: Re: Interrupt Handling and Scheduler Subject: Former WR employee club Subject: Re: Relocation error when loading vxWorks modules Subject: Contigous Memory Allocation Subject: Re: Contigous Memory Allocation Subject: Seeking PPC target simulator for Solaris 7/8 Subject: ip route path Subject: Higher heap usage with Tornado II kernel Subject: Re: suspend functions using semTake() Subject: Difficulity in buying USB stack for vxworks Subject: Good luck to your USB migration... Subject: WindRiver Tornado problem: ftp server with wu-ftpd in Linux Subject: FAT32 and harddisks connected to fiber channel Subject: Configuring network devices - Please Help! Subject: Load vxWorks modules twice Subject: Re: Load vxWorks modules twice Subject: Re: Contigous Memory Allocation Subject: Re: Linking up ".o" files ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Grrr, still having PCI problems Date: 24 Oct 2002 04:31:36 -0700 From: dj_thorrin71@yahoo.com (Johnathan Ross) Organization: http://groups.google.com/ Message-ID: <40cb3a14.0210240331.7cceac0b@posting.google.com> References: <40cb3a14.0210231249.733d9228@posting.google.com> <3DB72F84.2D03@despam.autobahn.mb.ca> Heh, following up my own post... I forgot to mention that we are on a pentium target. I was looking at the 'rep movsw' assembly command. I saw on a webpage somewhere that this command was actually more efficient than DMA (?!) This doesnt seem too hard to implement either. I will give it a try and let you know of the results. (sample pulled from comp.lang.asm.x86) lds si, ; put src address in DS:SI les di, ; put dest address in ES:DI mov cx, ; how many words to move cld; increment src and dest addresses rep movsw; move the data - -JR "H. E. Taylor" wrote in message news:<3DB72F84.2D03@despam.autobahn.mb.ca>... > In article <40cb3a14.0210231249.733d9228@posting.google.com>, > Johnathan Ross wrote: > > > > Hey guys, > > > > I am still having problems sending data across the PCI bus. We are on > > a Pentium target and are trying to send up to 32k of data across the > > PCI bus to another card. The data gets to the card, but it is very > > slow. > > > > I have run into this in the past, as well. > > > I have looked at some of the PCI registers on the Pentium (Bus Master) > > and I see a register called latency time which determines how long the > > master can hold the PCI bus. I have tried raising and lowering this > > value, but the times are still the same. > > > > Same here. > > > Any other registers that I can try? I would think that there aren't > > any registers to controll the speed of data going across the bus, you > > just tell it to move a block of data. I was hoping that by raising > > the latency timer, the host would grab control of the bus and not let > > go, but I am seeing no change in the amount of time to send the data. > > > > Basic flow of the code is: > > > > memDevCreate (READMEM, 0, sysMemTop()); > > fd_read = open ("/mem/read/0", O_RDONLY, 0); > > ... > > ioctl (fd_read, FIOSEEK, address); > > read (fd_read, ptr_message, length); > > > > basically trying to copy data from one memory address to another > > (address to ptr_message). Address is on the PCI card and ptr_message > > is on the host. > > > > Any ideas will be greatly appreciated. > > > > I'm not sure why you are messing with a memDev, but no matter. > I used to use base registers to find a buffer and set up a > pointer and just execute a read/write increment loop (which > is probably what the memDev does). > > Is there a CPU on the PCI card? > If so, ask your hardware guys what address arbitration has to take > place on every read/write. > > The only way I saw this operation running fast was using DMA, and > that I think was because the above arbitration only took place once > per block. > > -het --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Interrupt Handling and Scheduler Date: 24 Oct 2002 06:16:19 -0700 From: krishgroups@yahoo.com (Krishna) Organization: http://groups.google.com/ Message-ID: <927eeb1e.0210240516.6cc14b8e@posting.google.com> References: <53c6857a.0210232148.5268569a@posting.google.com> Hi, If interrupts are coming at a very high frequency (as frequent as a task can't get scheduled), then it is as good as the same interrupt ISR not clearing the IRQ status. Then Kernel will raise an exception: workQPanic: Kernel work queue overflow For more details regarding this, you can also refer to the link. http://www.xs4all.nl/~borkhuis/vxworks/vxw_pt5.html ~Krish anjaneyulu@mistralsoftware.com (vemanj) wrote in message news:<53c6857a.0210232148.5268569a@posting.google.com>... > Hi, > I have a doubt about the interrupt handling and the scheduling. > When an ISR is being executed, that particular interrupt is disable. > > My question is :if the interrupts are coming at very high frequency, > the other tasks will never get the chance to execute. Is it right??? > > Please tell me the scheduler reaction to an interrupt. What happens to > the current execution???? > > Thanks in advance. > > Regards, > Anj --------------------------- Newsgroups: comp.os.vxworks Subject: Former WR employee club Date: Thu, 24 Oct 2002 15:29:03 +0200 From: "Leonid Rosenboim" Organization: Verio Message-ID: Reply-To: "Leonid Rosenboim" Just found out this forum already exists, lost of familiar names: http://groups.yahoo.com/group/windx/ - -- - ----------------------------------------------------------------------- Leonid Rosenboim Consultant Email: my first name at consultant dot com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Relocation error when loading vxWorks modules Date: Thu, 24 Oct 2002 13:39:05 +0000 (UTC) From: Johan Borkhuis Organization: Proxim Corp. Message-ID: References: "Small, Gene" wrote: > I've recently added memory to one of my PPC 604 targets (from 32M to > 40M) and I am now getting: > > ld (1,0,"fstream.o") > Relocation value does not fit in 24 bits. > Relocation value does not fit in 24 bits. > Relocation value does not fit in 24 bits. > Relocation value does not fit in 24 bits. > > when trying to load objects. Has anyone come across this or know how > to fix it? See http://nl0030-proxim-www/Vxworks/vxw_pt2_ppc.html#2.2.1-A Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Contigous Memory Allocation Date: Thu, 24 Oct 2002 17:32:50 +0200 From: "- Ludo -" Organization: Guest of ProXad - France Message-ID: <3db812a5$0$6102$626a54ce@news.free.fr> Hi, Just a little question : does malloc function under VxWorks, allocates a contiguous area of memory (just as linux kmalloc does) ? If not, is there a way to do it ? Thanks - - Ludo - --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Contigous Memory Allocation Date: Thu, 24 Oct 2002 09:09:53 -0700 From: "Michael R. Kesti" Organization: MK Associates Message-ID: <3DB81B51.2F617441@gv.net> References: <3db812a5$0$6102$626a54ce@news.free.fr> Reply-To: mkesti@gv.net - - Ludo - wrote: > Hi, > >Just a little question : does malloc function under VxWorks, allocates a >contiguous area of memory (just as linux kmalloc does) ? > >If not, is there a way to do it ? I'm not familiar with Linux' kmalloc and I don't have actual knowledge of VxWorks' malloc internals, but I can't see how the latter could do anything but allocate a contiguous memory area. malloc() returns only a pointer to the memory (or a null pointer if unable to allocate the requested block size) and applications are free to use the allocated memory as they see fit. Without knowledge of any discontinuities in the allocation, such applications would eventually clobber any multiple allocations, therefore allocated blocks must be contiguous. - -- ======================================================================== Michael Kesti | "And like, one and one don't make | two, one and one make one." mkesti@gv.net | - The Who, Bargain --------------------------- Newsgroups: comp.os.vxworks Subject: Seeking PPC target simulator for Solaris 7/8 Date: 24 Oct 2002 11:15:33 -0700 From: bruce@scherzinger.net (Bruce Scherzinger) Organization: http://groups.google.com/ Message-ID: I am looking for a place to get source or binary files for a VxSim PPC target that will run on Solaris 7/8. It would be wonderful if this target could be preconfigured with the OS configured as vxWorks_romResident, but if not we can certainly build it with our code. Does anyone know if this is available either as open source or commercially and, if so, where can I get it or information about it? Thanks in advance, Bruce P.S. email address is defunct. please post reply here. thanks --------------------------- Newsgroups: comp.os.vxworks Subject: ip route path Date: Thu, 24 Oct 2002 21:55:34 GMT From: "somebody" Organization: Magma Communications Ltd. Message-ID: Hi, I have a general question. In our system we only have two interfaces. When a packet comes in I just want to route the packet to the other interface without going through the routing table. Is there any document somewhere, where it talks about how packet are handler between different module of IP layer. Thanks jon --------------------------- Newsgroups: comp.os.vxworks Subject: Higher heap usage with Tornado II kernel Date: 24 Oct 2002 14:31:08 -0700 From: daveg@progeny.net (David Garner) Organization: http://groups.google.com/ Message-ID: <8b62c40.0210241331.2db5665a@posting.google.com> We are migrating our apps to Tornado II with vxworks v5.4. Our system previously used Tornado I with vxworks v5.3. The problem is that when we issue memShow from the target shell with no startup script and nothing extra running that the allocated memory is 8 meg higher on T2 versus T1. This is causing a problem because we no longer have enough memory to load our objects. The two kernels are configured identically. Does anyone know why T2 uses 8 more meg from the heap? Is there a tool that will analyze blocks allocated on the heap and list the associated tasks? VxWorks Tornado Thanks, Dave Garner --------------------------- Newsgroups: comp.os.vxworks Subject: Re: suspend functions using semTake() Date: 23 Oct 2002 08:02:40 -0700 From: jsanchor@cs5.dasd.honeywell.com (Jay) Organization: http://groups.google.com/ Message-ID: <9e60ca47.0210230702.32d434dc@posting.google.com> References: <9e60ca47.0210181338.79bd7307@posting.google.com> <3DB18E31.E84A544C@gv.net> "Michael R. Kesti" wrote in message news:<3DB18E31.E84A544C@gv.net>... > Jay wrote: > > >Is there a nicer way to do this? > > > >typedef struct _buffer_type > >{ > >int cmd; > >int status; > >} buffer_type; > > > >buffer_type* ptrToBuffer = (buffer_type*)some_hardware_address; > > > >/*simplified: no one calls sendCommand() till it has returned status */ > >int sendCommand (int cmd) > >{ > > ptrToBuffer->cmd = cmd; /* writes to hardware, generates interrupt*/ > > /* hardware processing command, so wait */ > > semTake(); > > return ptrToBuffer->status; > >} > > > >/* after hardware is done processing command, it puts results at > >status location and generates interrupt */ > >void intFromHardware(void) > >{ > > semGive(); > >} > > > >basically, I want the command to block till it has status from hardware. > > If I understand what you're seeking, this is called "process syncronization" > and semaphores are intended for exactly this, as well as other things. > > You're on the correct path, Jay. Thank you. Sorry it took me so long to get back. --------------------------- Newsgroups: comp.os.vxworks Subject: Difficulity in buying USB stack for vxworks Date: Fri, 25 Oct 2002 00:55:22 GMT From: "Mike Kim" Organization: Dacom Message-ID: <_D0u9.89766$v_3.1855177@news.bora.net> References: <488e459a.0210212115.752d889c@posting.google.com> I think buying a usb stack is a expensive one. I was told this usb stack would be 38,000~40,000 USD. So we couldn't decide buying this usb stack even if we have possiblity about ataching a usb camera requested by our customer. Anyway I hope there would be a good alternative using usb device with good reliablity. bye~~ --------------------------- Newsgroups: comp.os.vxworks Subject: Good luck to your USB migration... Date: Fri, 25 Oct 2002 00:50:06 GMT From: "Mike Kim" Organization: Dacom Message-ID: <2z0u9.89765$v_3.1854724@news.bora.net> References: <488e459a.0210212115.752d889c@posting.google.com> Hi~ Jeff.. I wish you would have a good result. "Jeff Zheng" wrote in message news:ap55rt$1968$1@mail.cn99.com... > > Usb bus is designed to be much software related. > The usb controller is much like an ethernet controller. > Usb device driver is much like socket users. > And there must be some kind of protocol stack between them. > I have tried to migrate linux usb stack to vxWorks. but havn't succeed yet. > > Jeff > > > "Mike Kim" wrote in > news:Tr7t9.77017$v_3.1624862@news.bora.net: > > > OK, I want to know alternative way using usb device. > > But You said USB is a bus which has a complex bus. > > > > I know buying USB stack is a fast, convient way for me. > > > > But only I want to know if there is any other way about this. > > > > Thanks anyway for your respose. > > > > > > -- > > ####################################### > > ChanSeong Kim > > OFFICE : 02-829-3391 > > FAX : 02-829-3400 > > PCS : 016-660-6457 > > KOREA COMPUTER BLDG. F9 > > 395-65, SHNDAEBANG-DONG, > > DONGJAK-KU > > SEOUL, 156-010 KOREA > > ####################################### > > > > "John" wrote in message > > news:488e459a.0210212115.752d889c@posting.google.com... > >> Hello, > >> > >> If the camera is a USB one, I'm not sure how you plan to avoid using > >> a USB protocol stack - USB is not like a simple I/O port, it is a > >> bus, and it has a fairly complex protocol running over it that must > >> be used. > >> > >> You might be able to get other forms of camera - some older ones were > >> parallel port based; you might be able to still get some of those, or > >> you could look for a serial camera - some of the early Kodak digital > >> cameras had serial port connections. You might also be able to write > >> your own USB stack, or buy one from somewhere else... > >> > >> HTH, > >> > >> John... > >> --------------------------- Newsgroups: comp.os.vxworks Subject: WindRiver Tornado problem: ftp server with wu-ftpd in Linux Date: Fri, 25 Oct 2002 13:36:10 +0800 From: Richard Wang Organization: DCI HiNet Message-ID: <3DB8D84A.70601@ms3.senao.com.tw> Hi, To boot a board, I need to download the image through ftp. There are at least two ways to do this. I can build the server with Tornado's ftp server. I can also build a ftp server on another computer. When I tried to put downloaded image from a remote computer running Linux system, I can not boot the image successfully even though I use the same image that just successfully booted the board through a ftp server on Windows, that comes with Tornado package. Both images seem to be downloaded successfully. However, the one downloaded through wu-ftpd (lINUX) can not be started successfully. Have you have this problem before? Why? Is there other way to solve this? Thanx, Richard --------------------------- Newsgroups: comp.os.vxworks Subject: FAT32 and harddisks connected to fiber channel Date: Fri, 25 Oct 2002 08:45:34 +0200 From: Christer Lund Organization: Ericsson Microwave Systems AB Message-ID: <3DB8E88E.1C6F3C75@emw.ericsson.se> Hi, I'm involved i a project where we plan to use a couple of VME based Power PCs (PPC) connected to disks (HD) via fiber channel (FC) arbitrating loop (AL). _________VME_________ | | | PPC1 PPC2 PPC3 | |___AL___| |_______| | | | | _________ _______ | | | | | | | HD1 HD2 HD3 1 The operating system on the PPC's is vXWorks 2 All PPC's shall have access to all HD's, directly via AL, i.e. the filesystem resides on all PPC's. 3 To be compatible whith another system the we would prefer FAT32 as filesystem. 4 The filesystem should be resistant/repairable in runtime if an unplanned stop, e.g. power failure should occur. I dont know if the requirements can be fulfilled. FAT32 is to my knowledge not a "safe" filesystem. My ide is to use a journaling sytem to record actions to the filesystem. Is there better idees? Is it possible to use FAT32 according to (2) above? My ide is that only PPC1 is permitted to create files. When the files are created the max filelength is allocated. The FAT is then updated only by PPC1. PPC2 and PPC3 are only permitted to open, write and close already created files. A file is accessed only from one PPC. This should work I think. However, I'm not sure what happens when PPC2 and PPC3 flushes or syncing the filesystem. Does they affect the whole FAT or only their own files? In other words, is there a risc that e.g. PPC2 holds an old version of FAT that destroys the FAT (updated by PPC1) when it flushes the filesystem? Is NFS a better solution for vXWorks? Should I develop my own filesystem? I'm grateful for all suggestions, thoughts and ide's you come up with. Thanks, Christer --------------------------- Newsgroups: comp.os.vxworks Subject: Configuring network devices - Please Help! Date: 25 Oct 2002 00:43:23 -0700 From: vanessabenfield98@hotmail.com (Vanessa Benfield) Organization: http://groups.google.com/ Message-ID: Hi all, I hope someone out there can help me, as I've hit a dead end and don't know what to do!!! I have created a standalone image and booted it using a floppy disk on my pc. However when I boot I get it attaching to the local (looping) interface - I don't want this as I've been told I need to attach it to some other interface (ie. I don't know what one, just not a local one?!) I have a 3com Ethernet Link PCI TPC NIC (3C900B-TPB) network card and have tried different settings using the GUI to change The BSD ethernet drivers (selected from network devices), none of these have changed the interface when booting though. I also have a WDB network connection set. If anyone can help that would be great, as I am new to vxWorks, and seem to have hit one problem after another!! Thanks in advance, Vanessa --------------------------- Newsgroups: comp.os.vxworks Subject: Load vxWorks modules twice Date: Fri, 25 Oct 2002 16:42:46 +0800 From: "Hans" Organization: Lucent Technologies, Indian Hill Message-ID: Hello When I load an object file from winshell twice, I found there are two module with same name. Does anybody know why? I want the later module overwrite the previous one. Thanks Hans --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Load vxWorks modules twice Date: Fri, 25 Oct 2002 09:27:12 +0000 (UTC) From: Johan Borkhuis Organization: Proxim Corp. Message-ID: References: "Hans" wrote: > Hello > > When I load an object file from winshell twice, I found there are two > module with same name. Does anybody know why? I want the later module > overwrite the previous one. Then you have to unld it first. BTW: it is no problem to have the same object file twice in memory: the new symbols will overwrite the old ones, so when you call something it will be called in the new module. Kind regards, Johan - -- o o o o o o o . . . _____________________________ o _____ || Johan Borkhuis | .][__n_n_|DD[ ====_____ | JBorkhuis@Proxim.com | >(________|__|_[_________]_|__________________________| _/oo OOOOO oo` ooo ooo 'o!o!o o!o!o` === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html === --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Contigous Memory Allocation Date: Fri, 25 Oct 2002 12:33:22 +0200 From: "- Ludo -" Organization: Guest of ProXad - France Message-ID: <3db91eb1$0$6124$626a54ce@news.free.fr> References: <3db812a5$0$6102$626a54ce@news.free.fr> > I presume you mean physically contiguous! Yes, exactly. > In which case the 'usual' answer is 'yes' because vxworks uses a fixed > map for memory. OTOH some systems (eg strongarm) have gaps in physical > memory which will be joined up using the mmu, in that case malloc() > could return memory that crosses the boundary. Ok, thanks for this. > 1) Hope! > 2) Allocate a buffer and check it is ok > 3) Allocate it from a pool that is known to be in contiguous memory I'll check this. Thanks for the answer --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Linking up ".o" files Date: Fri, 25 Oct 2002 06:34:52 -0400 From: joe durusau Organization: WEBUSENET.com Message-ID: <3DB91E4C.EC70DA22@bellsouth.net> References: Yung wrote: > Hi experts, > > I am trying to link up a number of ".o" and ".a" files to form a single > image. It seems that the order of the linking pays an important role because > the final image is functional only if objects are linked in specific > sequence. What is the trick inside? > The strange thing is that the image can usually be built successfully. Just > that incorrectly linked images may suffer from, say, some network connection > problems. > Anybody has experience like this? Thanks. In general, for any unix-like linker any .a files must follow all .o files, because the linker selectes the modules to load out of the .a files, based on unsatisfied externals it has inits internal tables at the time it encounters the .a file on the linker command line. The order of the .a files is also significant. If a particular .a files uses objects in another .a file, the _using_ .a file must show up on the link line before the one that is being used. See the man pages for whatever version of ld you are using for more details. The situation can get especially tricky of you make the mistake of having a situation in which say, foo.a uses things in bar.a and bar.a also uses things in foo.a. For a better answer, post a specific example, preferably a short one. Speaking only for myself, Joe Durusau --------------------------- End of New-News digest ********************** From vxwexplo-errs@csg.lbl.gov Fri Oct 25 17:06:11 2002 From: "Dr. Jifang Tian" Date: Fri Oct 25 17:06:13 PDT 2002 Subject: Need Help on extending the vxworks PPC memory --------------9F7C8B3E1BE2D878940C34B5 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: base64 Hello, All, We are running vxWorks 5.3.1 (Tornado 1.0.1) (cpu: MVME 2300-mpc603) on the PPC with 64M memory. But the maximum memory we can use is 32M. Since the applications are getting larger, we need more memory. How to solve this problem ? it is possible ? You reply regarding this message are highly appreciated. Thank a million ! Jeff Tian Spectral Dynamics Tel: 408-918-2550 --------------9F7C8B3E1BE2D878940C34B5 Content-Type: text/html; charset=us-ascii Content-Transfer-Encoding: 7bit 
Hello, All,
We are running vxWorks 5.3.1 (Tornado 1.0.1) (cpu: MVME 2300-mpc603) on the PPC with
64M memory. But the maximum memory we can use is 32M. Since the applications are
getting larger, we need more memory. How to solve this problem ?  it is possible ?
You reply regarding this message are highly
appreciated. Thank a million !



Jeff Tian



Spectral Dynamics
Tel: 408-918-2550

 

--------------9F7C8B3E1BE2D878940C34B5--



From vxwexplo-errs@csg.lbl.gov  Fri Oct 25 23:32:49 2002
From: dht 
Date: Fri Oct 25 23:32:51 PDT 2002
Subject: RE: Need Help on extending the vxworks PPC memory

Just add -mlongcall to CFLAGS and rebuild everything.  Makes the code a bit 
larger since the jumps and calls now have to load two long words to execute 
instead of one but it the only way to get around the 32M memory limit since 
vxWorks loads low and apps load high making them >32M apart in a 64+Meg board.
 Otherwise the compiler uses a relative jump/call that has only 24 bits of 
offset address.



From vxwexplo-errs@csg.lbl.gov  Sat Oct 26 04:03:14 2002
From: Vxworks Exploder 
Date: Sat Oct 26 04:03:16 PDT 2002
Subject: comp.os.vxworks newsdigest

Comp.Os.Vxworks Daily Digest    Sat Oct 26 04:03:12 PDT 2002

        Subject: Re: Load vxWorks modules twice
        Subject: Diab vs. GNU for ARM
        Subject: Linking up ".o" files
        Subject: Re: Load vxWorks modules twice
        Subject: Re: File System in VxWorks
        Subject: Re: Driving Three Monitors from a single PC
        Subject: Re: Diab vs. GNU for ARM
        Subject: TCP/IP stacks performance
        Subject: When Changing IP address, old IP address doesn't quite go away..
        Subject: Re: ip route path 
        Subject: Re: ip route path 
        Subject: Re: ip route path 
        Subject: Re: Regarding Semaphore
        Subject: Re: Need Help on extending the vxworks PPC memory

-------------------------------------------------------

Newsgroups: comp.os.vxworks
Subject: Re: Load vxWorks modules twice
Date: Fri, 25 Oct 2002 17:53:40 +0800
From: "Ren Hantao" 
Organization: Lucent Technologies, Indian Hill
Message-ID: 
References:  

But for some reason, I don't want to unload it. Are there any other ways? I
remember when I uses vxworks some times ago, if the module are same, the
later one will overwrite the previous one, but I don't know why on our
current system, things are different.

Hans

"Johan Borkhuis"  wrote in message
news:Xns92B273E287D77jborkhuisproximcom@192.19.193.66...
> "Hans"  wrote:
>
> > Hello
> >
> >   When I load an object file from winshell twice, I found there are two
> > module with same name. Does anybody know why? I want the later module
> > overwrite the previous one.
>
> Then you have to unld it first. BTW: it is no problem to have the same
> object file twice in memory: the new symbols will overwrite the old ones,
> so when you call something it will be called in the new module.
>
> Kind regards,
>         Johan
>
> --
>         o o o o o o o . . .   _____________________________
>        o      _____           ||    Johan Borkhuis        |
>      .][__n_n_|DD[  ====_____  |    JBorkhuis@Proxim.com  |
>     >(________|__|_[_________]_|__________________________|
>     _/oo OOOOO oo`  ooo   ooo  'o!o!o                o!o!o`
> === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html ===
>



---------------------------

Newsgroups: comp.os.vxworks
Subject: Diab vs. GNU for ARM
Date: Fri, 25 Oct 2002 05:13:24 -0600
From: "Kyle Unice" 
Message-ID: 

Rumor is that Diab does 20% better (speed wise) than GNU.  I was wondering
if anyone had any experience to confirm or deny this.   20% performance
improvement may be worth the $2800 WR is asking for Diab.

Thanks,
kyle



---------------------------

Newsgroups: comp.os.vxworks
Subject: Linking up ".o" files
Date: Fri, 25 Oct 2002 18:00:47 +0800
From: "Yung" 
Organization: IMS Netvigator
Message-ID: 

Hi experts,

I am trying to link up a number of ".o" and ".a" files to form a single
image. It seems that the order of the linking pays an important role because
the final image is functional only if objects are linked in specific
sequence. What is the trick inside?
The strange thing is that the image can usually be built successfully. Just
that incorrectly linked images may suffer from, say, some network connection
problems.
Anybody has experience like this? Thanks.




---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: Load vxWorks modules twice
Date: Fri, 25 Oct 2002 13:48:10 +0200
From: "Michael Lawnick" 
Organization: Buergernetz Dillingen
Message-ID: 
References:   
Reply-To: "Michael Lawnick" 
Sender: mlawnick@145.253.156.162

Unload means 'free the memory and unoad the symbols', so if you want to
reload and the second has to cover the primary, this *is* just a prior
unload. All other may lead to bad symbols etc.
As far as I experienced it, the WindShell (Tornado) does an automatic
unload, target shell (serial/TELNET) just loads a second instance.
If you load a module via one shell, the other one doesn't see it, as long as
you haven't started symbol synchronisation.

Regards,
Michael

"Ren Hantao"  schrieb im Newsbeitrag
news:apb4hu$k6k@netnews.proxy.lucent.com...
> But for some reason, I don't want to unload it. Are there any other ways?
I
> remember when I uses vxworks some times ago, if the module are same, the
> later one will overwrite the previous one, but I don't know why on our
> current system, things are different.
>
> Hans
>
> "Johan Borkhuis"  wrote in message
> news:Xns92B273E287D77jborkhuisproximcom@192.19.193.66...
> > "Hans"  wrote:
> >
> > > Hello
> > >
> > >   When I load an object file from winshell twice, I found there are
two
> > > module with same name. Does anybody know why? I want the later module
> > > overwrite the previous one.
> >
> > Then you have to unld it first. BTW: it is no problem to have the same
> > object file twice in memory: the new symbols will overwrite the old
ones,
> > so when you call something it will be called in the new module.
> >
> > Kind regards,
> >         Johan
> >
> > --
> >         o o o o o o o . . .   _____________________________
> >        o      _____           ||    Johan Borkhuis        |
> >      .][__n_n_|DD[  ====_____  |    JBorkhuis@Proxim.com  |
> >     >(________|__|_[_________]_|__________________________|
> >     _/oo OOOOO oo`  ooo   ooo  'o!o!o                o!o!o`
> > === VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html ===
> >
>
>



---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: File System in VxWorks
Date: 25 Oct 2002 06:33:15 -0700
From: technokiran@yahoo.com (Kiran Kumar)
Organization: http://groups.google.com/
Message-ID: <552e79d.0210250533.3ea4f449@posting.google.com>
References: <552e79d.0210210251.a1ac354@posting.google.com> <3DB42C2C.EC9F16A4@gv.net> <3d425998.0210232307.74c6a2b1@posting.google.com>

r1y2g3@sina.com (Eagle) wrote in message news:<3d425998.0210232307.74c6a2b1@posting.google.com>...
> "Michael R. Kesti"  wrote in message news:<3DB42C2C.EC9F16A4@gv.net>...
> > Kiran Kumar wrote:
> > 
> > >Hi,
> > >    I am new to Vxworks. I am facing problems when I was trying to
> > >create a file in the VxWorks application. Can anyone suggest, how to
> > >handle files in Vxworks.
> > >
> > >Thanks in advance
> > >kiran.
> > 
> > See creat() and other functions in the reference manual's ioLib section.
> 
> yes

Look, I am currently working using the simulator. Calling creat()
didn't work.
It gave me an Error which i decoded (in the iosLib module) - "Device
not found".
Someone suggested the use of usrAtaConfig(). But When I use it, I get
an error when dowloading the application (Unresolved Symbols). I have
tried to include many header files. But they only result in more
unresolved symbols. Can anyone give the Exact proceedure of creating a
file?

---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: Driving Three Monitors from a single PC
Date: 25 Oct 2002 06:39:43 -0700
From: technokiran@yahoo.com (Kiran Kumar)
Organization: http://groups.google.com/
Message-ID: <552e79d.0210250539.312650f5@posting.google.com>
References: <552e79d.0210170718.269fdb4d@posting.google.com> 

Bill Pringlemeir  wrote in message news:...
> [snip]
> 
>  Kiran> I have absolutely no idea as to how to handle more than 1
>  Kiran> monitor. The GUI package to be used is ZINC. Please help me
>  Kiran> out on this.
> 
> You can not do this with standard Zinc.  WindML supports multiple
> display contexts.  You will have to sub-class the ZafDisplay class and
> use the multiple WindML displays.  You may need to run separate Zinc
> applications for each display (Ie, vxWorks tasks).  You may have to
> sub-class the ZafApplication as well to set the ZafDisplay properly.
> There may also be some global variables that might have to change to
> task specific variables, or be renamed for each application.
> 
> Once that is done, the Zinc output from each task should go to the
> appropriate monitor.  I have never ran multi-task Zinc systems.  I
> hope it goes well for you.
> 

Thanks a Lot. I dint understand much as I am very new to this
environment. So I need some time to get the hang of it. Hope you could
eloborate on it a little bit. Please oblige.
> Regards,
> Bill Pringlemeir.

---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: Diab vs. GNU for ARM
Date: Fri, 25 Oct 2002 17:18:43 +0000 (UTC)
From: David Laight 
Organization: BT Openworld
Message-ID: 
References: 

Kyle Unice wrote:
> Rumor is that Diab does 20% better (speed wise) than GNU.  I was wondering
> if anyone had any experience to confirm or deny this.   20% performance
> improvement may be worth the $2800 WR is asking for Diab.

Well the version of gcc WRS were shipping for arm was fairly
archaic and didn't generate very good code at all.
The newer gcc ought to be a lot better - but I don't know if WRS
are shipping it yet.

Of course if you do any floating point maths you need to get some
decent FP routines from somewhere.  The ones compiled from the C
source (from gnu) are particularly dire.

	David


---------------------------

Newsgroups: comp.os.vxworks
Subject: TCP/IP stacks performance
Date: Fri, 25 Oct 2002 19:23:23 +0100
From: Frederic 
Organization: SchlumbergerSema - Grenoble Technology and Projets Center
Message-ID: <3DB98C1B.84785269@NOSPAM.slb.com>
Reply-To: frederic.torras@NOSPAM.slb.com

hi all,

does anyone know where to find informations about TCP/IP stack
performance under VxWorks ?

benchmarks should be appreciate.

Thanks.


---------------------------

Newsgroups: comp.os.vxworks
Subject: When Changing IP address, old IP address doesn't quite go away..
Date: 25 Oct 2002 13:23:39 -0700
From: softrealtime@attbi.com (Bruce Clements)
Organization: http://groups.google.com/
Message-ID: <45bfa250.0210251223.2840a84d@posting.google.com>

I have noticed that when I change an ip address on the fly, VxWorks
still replies to icmp echo requests for the previous IP address The
new address  works fine. On the echo request for the old IP address I
get two packets:

The first is an arp from my device with the senders protocol address
being the new IP address and the target protocol address being the old
IP address.

The second is what sniffer pro describes as and ICMP Redirect for Host
which is
the echo reply addressed to the old IP address, from the new IP
address.

I have tried using the ifAddrDel call from Johan Borkius's website but
this does not resolve the problem. I have also gotten some unsupported
code from WRS that does not help either. The provided routine
basically does the following. I have tried several tweaks to no avail:

//out with the old:
arpflush
ifFlagChange UP | Running FALSE
ifMaskGet 
ifBroadcastGet (this fails as we don't explicitly set it)
ifAddrGet
// both the following AddrDel calls appear to ioctl then close a
socket slightly differently
ifAddrDel (Johan's patch, only takes ipAddress and interface name -
plugged
  it in to see if it would make a difference, it did not.)
ifAddrDelete (code from WRS)
ifRouteDeleteAll
ipDetach

// In with the new:
ipAttach
Then this:
ifAddrAdd (Broadcast address NULL)
Or this: (neither seem to behave differently)
ifMaskSet
ifAddrSet
ifFlagChange UP | RUNNING TRUE

Any help would be... well helpful

Thanks,
 Bruce Clements

---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: ip route path 
Date: Fri, 25 Oct 2002 20:37:18 GMT
From: "somebody" 
Organization: Magma Communications Ltd.
Message-ID: <2Yhu9.16691$V21.462284@news>
References:  

Thanks Leon,
Yes you are right I am bit concern about IP routing performance. Our box is
very similiar to a  DSL modem. We  have only  two interfaces. Atm interface
and  an ethernet interface. When a packet comes from from the ATM interface
I just wants to forward the packet  to the ethernet interface without
making any routing decision. Is there any way of doing it.
                          Hadi
"Leonid Rosenboim"  wrote in message
news:newscache$y59j4h$ni3$1@lnews.actcom.co.il...
> "somebody"  wrote in message
> news:q%Zt9.16579$V21.458656@news...
> > Hi,
> > I have a general question. In our system we only have two interfaces.
When
> a
> > packet comes in I just want to route the packet to the other  interface
> > without going through the routing table.
> >
> > Is there any document somewhere, where it talks about how packet are
> handler
> > between different module of IP layer.
> >             Thanks
> >                 jon
>
> Dear Jon at Spacebridge.com,
>
> your question is too general to be answered logically.
> When you say you want to "route" the packet, you basically mean you need
to
> "forward" the packet to teh otehr interface, but do you really want to
> forward ALL packets ? - if so, you are talking about a Layer-2 forwarding
> mechanism, which is a bit more complicated then that.
>
> The role fo the routing table in the IP stack is used for Layer-3
> orwarding  - also know as "routing", and the information in the routing
> table is used for making decisions about which packets to forward, and
which
> not to.
>
> Are you dissatisfied with the standard IP Layer-3 forwarding because of
its
> functionality or because of performance ?
>
> -- Please be more specific, and I will be happy to assist.
> -----------------------------------------------------------------------
>    Leonid Rosenboim                Visit:
> http://gamla.org.il/english/index.htm
>         Consultant                Email: my first name at consultant dot
com
>
>
>
>



---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: ip route path 
Date: Fri, 25 Oct 2002 12:29:24 +0200
From: "Leonid Rosenboim" 
Organization: Verio
Message-ID: 
References: 
Reply-To: "Leonid Rosenboim" 

"somebody"  wrote in message
news:q%Zt9.16579$V21.458656@news...
> Hi,
> I have a general question. In our system we only have two interfaces. When
a
> packet comes in I just want to route the packet to the other  interface
> without going through the routing table.
>
> Is there any document somewhere, where it talks about how packet are
handler
> between different module of IP layer.
>             Thanks
>                 jon

Dear Jon at Spacebridge.com,

your question is too general to be answered logically.
When you say you want to "route" the packet, you basically mean you need to
"forward" the packet to teh otehr interface, but do you really want to
forward ALL packets ? - if so, you are talking about a Layer-2 forwarding
mechanism, which is a bit more complicated then that.

The role fo the routing table in the IP stack is used for Layer-3
orwarding  - also know as "routing", and the information in the routing
table is used for making decisions about which packets to forward, and which
not to.

Are you dissatisfied with the standard IP Layer-3 forwarding because of its
functionality or because of performance ?

- -- Please be more specific, and I will be happy to assist.
- -----------------------------------------------------------------------
   Leonid Rosenboim                Visit:
http://gamla.org.il/english/index.htm
        Consultant                Email: my first name at consultant dot com





---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: ip route path 
Date: Sat, 26 Oct 2002 11:30:15 +0200
From: "Leonid Rosenboim" 
Organization: Verio
Message-ID: 
References:   <2Yhu9.16691$V21.462284@news>
Reply-To: "Leonid Rosenboim" 

Are you sire tis is like a DSL modem ? Because if it is, a DSL modem
does not forward just any packet, it usually recovers a PPPoA
packet and sends it to one particular MAC address as PPPoE,
all this is done at Layer 2, without ever recovering an IP frame.

When you get an ATM packet, what encapsulation does it use ?
Is it Classical IP over ATM or Lan Emulation ?

In any event, you can construct yoru own Ethernet MAC frame
and simply shoot it out with etherOutput() - only the source MAC
and FCS fields will be overridden by the MAC, all the rest of the
MAC frame will be transmitted exactly like you have assembled it.

HTH,
 Leonid

P.S. You have signed two different names in your postings,
which is rather confusing.

"somebody"  wrote in message
news:2Yhu9.16691$V21.462284@news...
> Thanks Leon,
> Yes you are right I am bit concern about IP routing performance. Our box
is
> very similiar to a  DSL modem. We  have only  two interfaces. Atm
interface
> and  an ethernet interface. When a packet comes from from the ATM
interface
> I just wants to forward the packet  to the ethernet interface without
> making any routing decision. Is there any way of doing it.
>                           Hadi
> "Leonid Rosenboim"  wrote in message
> news:newscache$y59j4h$ni3$1@lnews.actcom.co.il...
> > "somebody"  wrote in message
> > news:q%Zt9.16579$V21.458656@news...
> > > Hi,
> > > I have a general question. In our system we only have two interfaces.
> When
> > a
> > > packet comes in I just want to route the packet to the other
interface
> > > without going through the routing table.
> > >
> > > Is there any document somewhere, where it talks about how packet are
> > handler
> > > between different module of IP layer.
> > >             Thanks
> > >                 jon
> >
> > Dear Jon at Spacebridge.com,
> >
> > your question is too general to be answered logically.
> > When you say you want to "route" the packet, you basically mean you need
> to
> > "forward" the packet to teh otehr interface, but do you really want to
> > forward ALL packets ? - if so, you are talking about a Layer-2
forwarding
> > mechanism, which is a bit more complicated then that.
> >
> > The role fo the routing table in the IP stack is used for Layer-3
> > orwarding  - also know as "routing", and the information in the routing
> > table is used for making decisions about which packets to forward, and
> which
> > not to.
> >
> > Are you dissatisfied with the standard IP Layer-3 forwarding because of
> its
> > functionality or because of performance ?
> >
> > -- Please be more specific, and I will be happy to assist.
> > -----------------------------------------------------------------------
> >    Leonid Rosenboim                Visit:
> > http://gamla.org.il/english/index.htm
> >         Consultant                Email: my first name at consultant dot
> com
> >
> >
> >
> >
>
>



---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: Regarding Semaphore
Date: Sat, 26 Oct 2002 11:38:03 +0200
From: "Leonid Rosenboim" 
Organization: Verio
Message-ID: 
References: 
Reply-To: "Leonid Rosenboim" 

"Shiny"  wrote in message
news:df8a7b26.0210252119.7c2d739d@posting.google.com...
> Hi,
> Will there be any problem if Semaphore is taken twice for the same
> data structure one after the other and given later.
> I am using Mutex semaphore.

The Mutex semaphore was designed with this scenario in mind -
if a mutex is owned by a given task, that sae task may take the
sem again and again succesfully, because orderly execution
of the protected code is maintained.
The mutex must be given back exactly the same number of times
it was taken.

> Also I want to know will there be any problem if we use Binary
> Semaphore in the similar manner.
>

No, if you use a Binrary semaphore, it only ahs two satets,
so it can not be taken more then once in a row.

Hope this helps,
 Leonid



---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: Need Help on extending the vxworks PPC memory
Date: Sat, 26 Oct 2002 10:57:07 +0000 (UTC)
From: "Graham Baxter" 
Organization: BT Openworld
Message-ID: 
References: 

Jeff,

If your application does fit in 32MB another possibility is to give VxWorks
the other 32MB with a call to:
    memAddToPool()
in usrAppInit() in usrAppInit.c

There is an application note somewhere on Wind River's website on how to do
this.

- --
Regards,


Graham Baxter
Freelance Software Engineer (VxWorks and pSOS BSPs)
Graham Baxter (Software) Limited
gbaxter@NOSPAMbcs.org.uk

"dht"  wrote in message news:apdffk$lhp$1@overload.lbl.gov...
> Just add -mlongcall to CFLAGS and rebuild everything.  Makes the code a
bit
> larger since the jumps and calls now have to load two long words to
execute
> instead of one but it the only way to get around the 32M memory limit
since
> vxWorks loads low and apps load high making them >32M apart in a 64+Meg
board.
>  Otherwise the compiler uses a relative jump/call that has only 24 bits of
> offset address.
>



---------------------------

End of New-News digest
**********************


From vxwexplo-errs@csg.lbl.gov  Sun Oct 27 04:03:09 2002
From: Vxworks Exploder 
Date: Sun Oct 27 04:03:10 PST 2002
Subject: comp.os.vxworks newsdigest

Comp.Os.Vxworks Daily Digest    Sun Oct 27 04:03:07 PST 2002

        Subject: Re: Source Browser in Tornado IDE
        Subject: Re: Diab vs. GNU for ARM
        Subject: Regarding Semaphore

-------------------------------------------------------

Newsgroups: comp.os.vxworks
Subject: Re: Source Browser in Tornado IDE
Date: 23 Oct 2002 20:02:11 GMT
From: Denis Perelyubskiy 
Message-ID: 
References: 

 * np  [10-Tue-02 21:48 -0700]:
 >
 > IS there a Source Browser (_Not the Target Browser_) in
 > the Tornado IDE ( like in Visual Studio Tools->Source
 > Browser). If yes how to build the project/workspace with
 > Source Browser option?

I am not sure what it is you refer to when you say "source
browser". If you're looking for things that do x-ref, etc,
then there is nothing vxworks-specific.

Tronado IDE is quite useless in anything
code-editing-related. Use standalone tools for that.

Martin had mentioned Sniff+, which is very nice, and I think
(but check) very expensive. If your company pays for it, its
quite nice. 

Somewhat buggy, but quite usable alternative includes
   http://sourcenav.sourceforge.net/
Source navigator used to be produced by Cygnus folks, but
now I think its open-sourced. Its in active development,
which progresses at a _very_ slow rate

Another tool, is not as expensive as Sniff+, less
"pretty", and probably less feature-rich. However, it is
fast, gets the job done well, and is commercially-supported
with frequent updates. The tool is called "Understand for
C++"
   http://www.scitools.com/

If you're not looking for x-ref type browsing, then take a
look at your favorite editor (emacs and vim and a host of
other editors may be integrated with the following). The
utils are mostly useful if they can be integrated with your
favorite editors. I know of scripts for vim. I assume there
are scripts for emacs. Probably, there are scripts for other
editors
   - ctags (http://ctags.sourceforge.net/)
   - idutils (http://www.gnu.org/directory/text/editors/idutils.html)
   - grep
   
Personally, I find source navigator usable enough. I have to
use it _very_ rarely, and mostly get around with
vim-ctags-idutils combo. 

If someone knows other options, I'd like to know about them
too. 

good luck,

denis

- -- 
'From' email address is used as a sink. Not read. Ever.
Instead, send to [t-h-e-u-k-r-a-i-n-i-a-n|h-o-t-m-a-i-l|c-o-m]
(remove dashes, replace the first vertical bar with @, 
and second with a dot)

---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: Diab vs. GNU for ARM
Date: 25 Oct 2002 22:54:23 -0400
From: Bill Pringlemeir 
Organization: Bell Sympatico
Message-ID: 
References: 
Sender: bpringle@DeadDuck


 Kyle> Rumor is that Diab does 20% better (speed wise) than GNU.  I
 Kyle> was wondering if anyone had any experience to confirm or deny
 Kyle> this.  20% performance improvement may be worth the $2800 WR is
 Kyle> asking for Diab.

Are you sure that is for the ARM?  I know that the PPC version is
about 20% better for plain vanilla 'C' code on a PowerPC.  I heard
that it wasn't as good for the ARM.  The tools from ARM themselves
seem to generate the best code.

Personally, I wouldn't find the 20% was worth it as the GNU tools work
with more OSs and CPUs.  Also, the license manager with DIAB is a
pain.

fwiw,
Bill Pringlemeir.

- -- 
I didn't order any WOO-WOO...  Maybe a YUBBA..  But no WOO-WOO!

---------------------------

Newsgroups: comp.os.vxworks
Subject: Regarding Semaphore
Date: 25 Oct 2002 22:19:00 -0700
From: vxworksonline@yahoo.com (Shiny)
Organization: http://groups.google.com/
Message-ID: 

Hi,
Will there be any problem if Semaphore is taken twice for the same
data structure one after the other and given later.
I am using Mutex semaphore.
Also I want to know will there be any problem if we use Binary
Semaphore in the similar manner.

TIA,
Shiny

---------------------------

End of New-News digest
**********************


From vxwexplo-errs@csg.lbl.gov  Mon Oct 28 04:03:20 2002
From: Vxworks Exploder 
Date: Mon Oct 28 04:03:22 PST 2002
Subject: comp.os.vxworks newsdigest

Comp.Os.Vxworks Daily Digest    Mon Oct 28 04:03:17 PST 2002

        Subject: Interrupting blocked I/O recvfrom calls
        Subject: Re: MPC860 training suggestions
        Subject: Re: Interrupting blocked I/O recvfrom calls
        Subject: Re: When Changing IP address, old IP address doesn't quite go away..
        Subject: JTAG debugger for MPC82xx
        Subject: Re: Newbie Interrupt Service Routine Question
        Subject: Re: JTAG debugger for MPC82xx
        Subject: Re: Diab vs. GNU for ARM
        Subject: Re: TCP/IP stacks performance
        Subject: Re: WindRiver Tornado problem: ftp server with wu-ftpd in Linux
        Subject: ospf setup
        Subject: Stop the extinction...  3549

-------------------------------------------------------

Newsgroups: comp.os.vxworks
Subject: Interrupting blocked I/O recvfrom calls
Date: 27 Oct 2002 05:23:22 -0800
From: mbenari@ecitele.com (mbenari@outthere.com)
Organization: http://groups.google.com/
Message-ID: 

I am trying to interrupt a task that is currently pending on a recvfrom()
call to a UDP Socket --- I am sort of rehashing an OLD thread 
"re: ping command available?" 1993-01-22; Chuck Meade in that I
am trying to utilize his techniques to interrupt the recvfrom call by
using signals.

There are various techniques to interrupt a socket that I have tried:
NOTE: I am testing this without actual packets being sent from remote
host.

1.  Setting SOL_SOCKET / SO_RCVTIMEO with the wanted timeout.
      I don't believe this is supported in vxWorks 5.4
      If it is, this is the preferred method.
         code would look like this:

      struct timeval tv;
      tv.tv_sec	= 5;
      tv.tv_usec	= 0;
      setsockopt(sockfd, SOL_SOCKET, SO_RCVTIMEO, &tv, sizeof(tv));

      while (1)
         {
         n= recvfrom(sockfd, recvbuf, RECVBUFSIZE, 0, NULL, NULL);
         if (n < 0)
            {
            if (errno == EWOULDBLOCK)
               {
               ... do something
               }
            else
               {  	
               ... recverror
               }
            }
            ... handle packet
         }



      THIS should work, but unless I am actually receiving packets on
      the socket, I NEVER get to EWOULDBLOCK I am assuming that SO_RCVTIMEO
      is unsupported in vxWorks.

2.  Preceeding calls to recvfrom() with calls to select() with timeout on 
      timeval.  This works and I will NOT include code HERE.  However, for
      both this case and case #1 (above), I am in essense changing the
      behavior of my task from one that is truely blocking on the socket
      itself (in _sbwait() on an IO semaphore), to one that is in essense
      NON-BLOCKING and polling with a resolution of timeval in select().
      This solution works for the purposes that I require, however it does 
      influence the behavior of the waiting task - LESS desirable.

3.  Interrupting the socket call by sending a signal to the task.
      This solution is the more desirable.  However, I am having problems 
      getting this to work.

      Code attached.

   void dummy_alarm_handler(int);
   SIGVEC  myHandlerInfo = {dummy_alarm_handler,0,0} ;
   SIGVEC  mySavedHandlerInfo ;
   int     myTaskId;
   WDOG_ID myWDId;


   /* forward declarations */

   LOCAL void myTaskBody (void);

   void dummy_alarm_handler(int i)
      {
      ;
      }

   STATUS myTaskInit (void)
      {
      .
      .   assume that the socket is setup as needed
      .
      sigInit();
      sigvec(SIGALRM, &myHandlerInfo, &mySavedHandlerInfo
   
      myTaskId = taskIdSelf() ;
      myWDId= wdCreate();
      if (myWDId == NULL)
         {
         printf("\r\nWDCreate failure!!\r\n");
         }

      return (OK);
      }

   LOCAL void myTaskBody (void)
      {
      int                       pktSize;           /* size of recvd packet */
      struct sockaddr_in        remoteAddr;    /* remote transport address */
      struct sockaddr_in        localAddr;     /* local transport address */
      char                      rcvBuf [MY_BUF_SIZE];  /* buffer for input */
      int                       sockSize;

      sockSize = sizeof (struct sockaddr_in);

      FOREVER
         {
         if (mySuspendFlag == MYTASK_SUSPEND_SELF_ID)
            {
            /* turn OFF suspend Flag for next time */
            mySuspendFlag= 0;
            
            printf("actually suspended HERE !!");
            taskSuspend(taskIdSelf());
            printf("Resumed from HERE !!\r\n");
            }

         if ( (pktSize = recvfrom (mySocket, rcvBuf, MY_BUF_SIZE, 0,
                                  (struct sockaddr *) &remoteAddr, &sockSize))  
                                  <= 0  )
            {
            
            /* If pktSize is error (<0) or other printerr data */
            if (pktSize < 0 )
               {
               switch (errno)
                  {
                  case EWOULDBLOCK:
                     printf("recvfrom TIMEOUT !!");
                     break;
                  case EINTR:
                     printf("recvfrom INTR via alarm !!");
                     break;
                  default:
                     printf("recvfrom OTHER errno !!");
                     break;
                  }
               }
            else
               {
               printf("recvfrom == 0 (No Packet Data to Process)");
               }
            
            continue;
            }
         sockSize = sizeof (struct sockaddr_in);
         getsockname (mySocket, (struct sockaddr *) &localAddr, &sockSize);

         printf("Received packet...");
  
         processPacket (pktSize, rcvBuf, &remoteAddr, &localAddr, &mySocket);
         }
    }


   int myWDRoutine(void)
      {
      kill(myTaskId, SIGALRM);
      return (0);
      }

   void myTaskSelfKill(void)
      {
      wdStart(myWDId, 1300, myWDRoutine, 0) ;
      }



Calling myTaskSelfKill starts the wdTimer correctly, However for some
reason the call to Kill in myWdRoutine() has no affect, and therefore,
I never get to interrupt my task.  I checked with wdShow() to ensure
that the WD was setup with the correct function.   It seams as if sigvec
has NO EFFECT !!  If I call sigvec(SIGALRM, 0, &tempSIGVEC) from shell
there doesn't seem to be anything set to tempSIGVEC !!  Contents are
all 0.

Signals don't seem to be setup appropraitely!

myTask remains locked on the semaphore in recvfrom() ...

Am I missing anything here?

Any/all help would be appreciated.

Thanks.

- -Mike

---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: MPC860 training suggestions
Date: Sun, 27 Oct 2002 07:45:07 -0800
From: "Michael R. Kesti" 
Organization: MK Associates
Message-ID: <3DBC0A03.2F12A1DF@gv.net>
References: <72e486b0.0210261955.3b660b3e@posting.google.com>
Reply-To: mkesti@gv.net

kishuri wrote:

>Would like to find out your comments/opinions regarding Mpc860/mpc8260
>training courses from Arnewsh Inc(www.arnewsh.com) and
>PhoenixMicro(www.phxmicro.com). Wish to master architecture plus
>programming the  Mpc860/mpc8260 PowerQuicc under Vxworks and other
>RTOS.
>Please comment.
>
>If anyone attended these, please tell me is it worth for dollar, or is
>it the same as reading user manuals.  They are $1000 plus here in US.

I took the Arnewsh course in San Jose about 2.5 years ago in preparation
for my first PPC work.  The presentation used assembler and C, but no
RTOS of any kind was even mentioned.  As the project for which I was
preparing was also my first vxWorks experience, I also took Wind River's
Tornado and BSP classes.

Although a bit expensive, I can recommend these classes as a way to
quickly get up to speed on these topics.  Don't expect to come to them
with no embedded experience, however.  They are geared toward evelopers
with at least moderate experience.

- -- 
========================================================================
          Michael Kesti            |  "And like, one and one don't make
                                   |   two, one and one make one."
          mkesti@gv.net            |          - The Who, Bargain

---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: Interrupting blocked I/O recvfrom calls
Date: Sun, 27 Oct 2002 18:05:56 +0200
From: "Leonid Rosenboim" 
Organization: Verio
Message-ID: 
References: 
Reply-To: "Leonid Rosenboim" 

Now, I didn't dive into the methods you used and why they
have failed, but here is another method you haven't tried yet:

 ioctl (sockFd, FIOCANCEL, 0);

this should make the read() or recvfrom() return with an error immediately,
and it is permitted to make this call from an ISR, if I recall correctly.

- -- HTH
P.S. How's life at ECI Telecom these days ?
- -----------------------------------------------------------------------
   Leonid Rosenboim                Visit:
http://gamla.org.il/english/index.htm
        Consultant                Email: my first name at consultant dot com


"mbenari@outthere.com"  wrote in message
news:a7a5bf22.0210270523.7123df38@posting.google.com...
> I am trying to interrupt a task that is currently pending on a recvfrom()
> call to a UDP Socket --- I am sort of rehashing an OLD thread
> "re: ping command available?" 1993-01-22; Chuck Meade in that I
> am trying to utilize his techniques to interrupt the recvfrom call by
> using signals.
>
> There are various techniques to interrupt a socket that I have tried:
> NOTE: I am testing this without actual packets being sent from remote
> host.
>
> 1.  Setting SOL_SOCKET / SO_RCVTIMEO with the wanted timeout.
>       I don't believe this is supported in vxWorks 5.4
>       If it is, this is the preferred method.
>          code would look like this:
>
>       struct timeval tv;
>       tv.tv_sec = 5;
>       tv.tv_usec = 0;
>       setsockopt(sockfd, SOL_SOCKET, SO_RCVTIMEO, &tv, sizeof(tv));
>
>       while (1)
>          {
>          n= recvfrom(sockfd, recvbuf, RECVBUFSIZE, 0, NULL, NULL);
>          if (n < 0)
>             {
>             if (errno == EWOULDBLOCK)
>                {
>                ... do something
>                }
>             else
>                {
>                ... recverror
>                }
>             }
>             ... handle packet
>          }
>
>
>
>       THIS should work, but unless I am actually receiving packets on
>       the socket, I NEVER get to EWOULDBLOCK I am assuming that
SO_RCVTIMEO
>       is unsupported in vxWorks.
>
> 2.  Preceeding calls to recvfrom() with calls to select() with timeout on
>       timeval.  This works and I will NOT include code HERE.  However, for
>       both this case and case #1 (above), I am in essense changing the
>       behavior of my task from one that is truely blocking on the socket
>       itself (in _sbwait() on an IO semaphore), to one that is in essense
>       NON-BLOCKING and polling with a resolution of timeval in select().
>       This solution works for the purposes that I require, however it does
>       influence the behavior of the waiting task - LESS desirable.
>
> 3.  Interrupting the socket call by sending a signal to the task.
>       This solution is the more desirable.  However, I am having problems
>       getting this to work.
>
>       Code attached.
>
>    void dummy_alarm_handler(int);
>    SIGVEC  myHandlerInfo = {dummy_alarm_handler,0,0} ;
>    SIGVEC  mySavedHandlerInfo ;
>    int     myTaskId;
>    WDOG_ID myWDId;
>
>
>    /* forward declarations */
>
>    LOCAL void myTaskBody (void);
>
>    void dummy_alarm_handler(int i)
>       {
>       ;
>       }
>
>    STATUS myTaskInit (void)
>       {
>       .
>       .   assume that the socket is setup as needed
>       .
>       sigInit();
>       sigvec(SIGALRM, &myHandlerInfo, &mySavedHandlerInfo
>
>       myTaskId = taskIdSelf() ;
>       myWDId= wdCreate();
>       if (myWDId == NULL)
>          {
>          printf("\r\nWDCreate failure!!\r\n");
>          }
>
>       return (OK);
>       }
>
>    LOCAL void myTaskBody (void)
>       {
>       int                       pktSize;           /* size of recvd packet
*/
>       struct sockaddr_in        remoteAddr;    /* remote transport address
*/
>       struct sockaddr_in        localAddr;     /* local transport address
*/
>       char                      rcvBuf [MY_BUF_SIZE];  /* buffer for input
*/
>       int                       sockSize;
>
>       sockSize = sizeof (struct sockaddr_in);
>
>       FOREVER
>          {
>          if (mySuspendFlag == MYTASK_SUSPEND_SELF_ID)
>             {
>             /* turn OFF suspend Flag for next time */
>             mySuspendFlag= 0;
>
>             printf("actually suspended HERE !!");
>             taskSuspend(taskIdSelf());
>             printf("Resumed from HERE !!\r\n");
>             }
>
>          if ( (pktSize = recvfrom (mySocket, rcvBuf, MY_BUF_SIZE, 0,
>                                   (struct sockaddr *) &remoteAddr,
&sockSize))
>                                   <= 0  )
>             {
>
>             /* If pktSize is error (<0) or other printerr data */
>             if (pktSize < 0 )
>                {
>                switch (errno)
>                   {
>                   case EWOULDBLOCK:
>                      printf("recvfrom TIMEOUT !!");
>                      break;
>                   case EINTR:
>                      printf("recvfrom INTR via alarm !!");
>                      break;
>                   default:
>                      printf("recvfrom OTHER errno !!");
>                      break;
>                   }
>                }
>             else
>                {
>                printf("recvfrom == 0 (No Packet Data to Process)");
>                }
>
>             continue;
>             }
>          sockSize = sizeof (struct sockaddr_in);
>          getsockname (mySocket, (struct sockaddr *) &localAddr,
&sockSize);
>
>          printf("Received packet...");
>
>          processPacket (pktSize, rcvBuf, &remoteAddr, &localAddr,
&mySocket);
>          }
>     }
>
>
>    int myWDRoutine(void)
>       {
>       kill(myTaskId, SIGALRM);
>       return (0);
>       }
>
>    void myTaskSelfKill(void)
>       {
>       wdStart(myWDId, 1300, myWDRoutine, 0) ;
>       }
>
>
>
> Calling myTaskSelfKill starts the wdTimer correctly, However for some
> reason the call to Kill in myWdRoutine() has no affect, and therefore,
> I never get to interrupt my task.  I checked with wdShow() to ensure
> that the WD was setup with the correct function.   It seams as if sigvec
> has NO EFFECT !!  If I call sigvec(SIGALRM, 0, &tempSIGVEC) from shell
> there doesn't seem to be anything set to tempSIGVEC !!  Contents are
> all 0.
>
> Signals don't seem to be setup appropraitely!
>
> myTask remains locked on the semaphore in recvfrom() ...
>
> Am I missing anything here?
>
> Any/all help would be appreciated.
>
> Thanks.
>
> -Mike



---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: When Changing IP address, old IP address doesn't quite go away..
Date: 27 Oct 2002 08:32:53 -0800
From: gabi@tdsoft.com (Gabi)
Organization: http://groups.google.com/
Message-ID: <5607ff2c.0210270832.da37805@posting.google.com>
References: <45bfa250.0210251223.2840a84d@posting.google.com>

softrealtime@attbi.com (Bruce Clements) wrote in message news:<45bfa250.0210251223.2840a84d@posting.google.com>...
> I have noticed that when I change an ip address on the fly, VxWorks
> still replies to icmp echo requests for the previous IP address The
> new address  works fine. On the echo request for the old IP address I
> get two packets:
> 
> The first is an arp from my device with the senders protocol address
> being the new IP address and the target protocol address being the old
> IP address.
> 
> The second is what sniffer pro describes as and ICMP Redirect for Host
> which is
> the echo reply addressed to the old IP address, from the new IP
> address.
> 
> I have tried using the ifAddrDel call from Johan Borkius's website but
> this does not resolve the problem. I have also gotten some unsupported
> code from WRS that does not help either. The provided routine
> basically does the following. I have tried several tweaks to no avail:
> 
> //out with the old:
> arpflush
> ifFlagChange UP | Running FALSE
> ifMaskGet 
> ifBroadcastGet (this fails as we don't explicitly set it)
> ifAddrGet
> // both the following AddrDel calls appear to ioctl then close a
> socket slightly differently
> ifAddrDel (Johan's patch, only takes ipAddress and interface name -
> plugged
>   it in to see if it would make a difference, it did not.)
> ifAddrDelete (code from WRS)
> ifRouteDeleteAll
> ipDetach
> 
> // In with the new:
> ipAttach
> Then this:
> ifAddrAdd (Broadcast address NULL)
> Or this: (neither seem to behave differently)
> ifMaskSet
> ifAddrSet
> ifFlagChange UP | RUNNING TRUE
> 
> Any help would be... well helpful
> 
> Thanks,
>  Bruce Clements

see "change IP on the fly" under gabi

---------------------------

Newsgroups: comp.os.vxworks
Subject: JTAG debugger for MPC82xx
Date: Sun, 27 Oct 2002 18:36:48 +0200
From: "Martin Roth" 
Organization: Motorola
Message-ID: 

 I consider purchasing a JTAG debugger for BSP development. The CPU board is
based on MPC82xx processor of Motorola and will run Vxworks 5.4 OS.
I have two options for JTAG debuggers:
1) Visionice II of Windriver
2) PowerTap of Applied Microsystems
I would appreciate to have any good/bad opinion regarding these JTAG
debuggers!


Regards,

  Martin Roth



---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: Newbie Interrupt Service Routine Question
Date: Sun, 27 Oct 2002 09:26:16 -0800
From: "Dan Gold" 
Organization: Mailgate.ORG Server - http://www.Mailgate.ORG
Message-ID: 
References: <3db557bb.3671715708@news.airband.net> <3db9b93d.3958838949@news.airband.net>
Sender: gold@ensemble.com

Robotz wrote in message <3db9b93d.3958838949@news.airband.net>...
>After a lot of googling and searching the source code, I found the
>file m8260IntrCtl.c which appears to have the functions to use the
>8260 interrupts.  This is what I've come up with based on an example I
>found in a source file, but I'm still not getting any interrupt
>signal.  If anyone has any tips of can give some example code it would
>be greatly appreciated.


The 8260 multiplexes various pin functions, and the interrupts (IRQ1-7) are
one of them.  Check SIUMCR[DPPC] and make sure it's set for 00b, or
SIUMCR[L2CPC] and amke sure it's 01b, depending on which IRQ4 line you're
using.

  hth,
   Dan Gold  gold@ensemble.com



---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: JTAG debugger for MPC82xx
Date: Mon, 28 Oct 2002 07:54:12 +0000 (UTC)
From: Johan Borkhuis 
Organization: Proxim Corp.
Message-ID: 
References: 

"Martin Roth"  wrote:

>  I consider purchasing a JTAG debugger for BSP development. The CPU
>  board is 
> based on MPC82xx processor of Motorola and will run Vxworks 5.4 OS.
> I have two options for JTAG debuggers:
> 1) Visionice II of Windriver
> 2) PowerTap of Applied Microsystems
> I would appreciate to have any good/bad opinion regarding these JTAG
> debuggers!

I use visionICE quite a lot, for both MIPS and PPC (824x), and it works 
very well. The SW is stable (btw: visionICE runs under pSOS :-), 
visionCLICK (the PC SW) knows how to interface with VxWorks, and you can 
use the visionICE as a WDB interface.

Kind regards,
        Johan

- -- 
        o o o o o o o . . .   _____________________________
       o      _____           ||    Johan Borkhuis        |
     .][__n_n_|DD[  ====_____  |    JBorkhuis@Proxim.com  |
    >(________|__|_[_________]_|__________________________|
    _/oo OOOOO oo`  ooo   ooo  'o!o!o                o!o!o`
=== VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html ===


---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: Diab vs. GNU for ARM
Date: Mon, 28 Oct 2002 07:58:54 +0000 (UTC)
From: Johan Borkhuis 
Organization: Proxim Corp.
Message-ID: 
References:  

David Laight  wrote:

> Well the version of gcc WRS were shipping for arm was fairly
> archaic and didn't generate very good code at all.
> The newer gcc ought to be a lot better - but I don't know if WRS
> are shipping it yet.

Yes, they are shipping T22 for ARM also, and this includes GCC 2.9 (the 
version number is 2.9-010413), so it is not the latest version compared to 
the ather architectures (PPC for example is shipping with 2.96), but still 
quite an improvement over the 2.7 version they were shipping until now.

Kind regards,
        Johan

- -- 
        o o o o o o o . . .   _____________________________
       o      _____           ||    Johan Borkhuis        |
     .][__n_n_|DD[  ====_____  |    JBorkhuis@Proxim.com  |
    >(________|__|_[_________]_|__________________________|
    _/oo OOOOO oo`  ooo   ooo  'o!o!o                o!o!o`
=== VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html ===


---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: TCP/IP stacks performance
Date: Mon, 28 Oct 2002 08:00:39 +0000 (UTC)
From: Johan Borkhuis 
Organization: Proxim Corp.
Message-ID: 
References: <3DB98C1B.84785269@NOSPAM.slb.com>

Frederic  wrote:
> does anyone know where to find informations about TCP/IP stack
> performance under VxWorks ?
> benchmarks should be appreciate.

Contact your WindRiver sales person or your FAE, they will be able to 
provide you with benchmark results for different architectures.

Kind regards,
        Johan

- -- 
        o o o o o o o . . .   _____________________________
       o      _____           ||    Johan Borkhuis        |
     .][__n_n_|DD[  ====_____  |    JBorkhuis@Proxim.com  |
    >(________|__|_[_________]_|__________________________|
    _/oo OOOOO oo`  ooo   ooo  'o!o!o                o!o!o`
=== VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html ===


---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: WindRiver Tornado problem: ftp server with wu-ftpd in Linux
Date: Mon, 28 Oct 2002 10:28:07 +0000 (UTC)
From: Johan Borkhuis 
Organization: Proxim Corp.
Message-ID: 
References: <3DB8D84A.70601@ms3.senao.com.tw>

Richard Wang  wrote:
....
> Both images seem to be downloaded successfully. However, the one
> downloaded through wu-ftpd (lINUX) can not be started successfully.
> Have you have this problem before? Why? Is there other way to solve
> this? 

See http://www.xs4all.nl/~borkhuis/vxworks/vxw_pt2.html#2.3-L. If this not 
helps check if the file is sent as binary or text.

Kind regards,
        Johan

- -- 
        o o o o o o o . . .   _____________________________
       o      _____           ||    Johan Borkhuis        |
     .][__n_n_|DD[  ====_____  |    JBorkhuis@Proxim.com  |
    >(________|__|_[_________]_|__________________________|
    _/oo OOOOO oo`  ooo   ooo  'o!o!o                o!o!o`
=== VxWorks FAQ: http://www.xs4all.nl/~borkhuis/vxworks/vxworks.html ===


---------------------------

Newsgroups: comp.os.vxworks
Subject: ospf setup
Date: 28 Oct 2002 03:46:42 -0800
From: g_ch@21cn.com (GANGCHENG)
Organization: http://groups.google.com/
Message-ID: 

I want to ask a question about the setting of ospf.
According to the help of vxworks, "ospfLib - OSPF version 2 (RFC 1583)
routing facilities (OSPF Option) " , I first define the INCLUDE_OSPF
in configAll.h. Once the system is up and running m need to invoke the
ospfInit( ) call.
Then I write :

    /* First we set up Interface A */
    /* set the interface address */
    intf.ospfIfIpAddress = 0xa00a0a05; /* 160.10.10.5 */
    
    /* address less interface is false */ 
    intf.ospfAddressLessIf = 0;
    
    /* interface area id set to 2 */ 
    intf.ospfIfAreaId =  2;
    
    /* router priority */
    intf.ospfIfRtrPriority = 5;
    
    /* various  time   intervals */
    intf.ospfIfTransitDelay   =   1;
    intf.ospfIfRetransInterval  =  3;
    intf.ospfIfHelloInterval = 10;
    intf.ospfIfRtrDeadInterval = 40;
    intf.ospfIfPollInterval = 30;
    
    /* enable OSPF on interface */
    intf.ospfIfAdminStat = M2_ospfAdminStat_enabled;
    
    /* set the parameters for this interface */
    if(m2OspfIfEntrySet (M2_OSPF_IF_AREA_ID |
        M2_OSPF_IF_RTR_PRIORITY |
        M2_OSPF_IF_RETRANS_INTERVAL |
        M2_OSPF_IF_HELLO_INTERVAL |
        M2_OSPF_IF_RTR_DEAD_INTERVAL |
        M2_OSPF_IF_POLL_INTERVAL |
        M2_OSPF_IF_ADMIN_STAT,
        &intf) != OK)
        {
        return (ERROR);
        }

But the above routine returns error. The error id is
S_m2Lib_ENTRY_NOT_FOUND.
I can be sure that the interface ip address is correct.
Why ?

Thanks a lot

---------------------------

Newsgroups: comp.os.rsts,comp.os.v,comp.os.vms,comp.os.vxworks
Subject: Stop the extinction...  3549
Date: Mon, 28 Oct 2002 11:46:26 GMT
From: pawe@hotmail.com
Organization: blueyonder (post doesn't reflect views of blueyonder)
Message-ID: 

Please read this itll only take you a minute - it could be your local area next...!
The corporations are at it again, this time focussing their attentions on a small part of the SW of England. A large international mining company claims to have found platinum in the UK and in their quest for money are going to extinquish at least 1 species from our planet. Butterflies, birds, small mammals, fish and even some plants are at risk from extinction if this corporate monster is not stopped.

Please, please, please send as many emails as you can to register your views at info@mcleaninternationalmining.com - like i said, it could be your area next...

Thanks
Harry
P.A.W.E
mcetusstbowmgytwlmzuwymuoytkxoxecgykubvzgdmnqmkl


---------------------------

End of New-News digest
**********************


From vxwexplo-errs@csg.lbl.gov  Mon Oct 28 09:10:08 2002
From: "Chuck Meade" 
Date: Mon Oct 28 09:10:10 PST 2002
Subject: Re: Interrupting blocked I/O recvfrom calls

Mike,

The original code that I posted uses the watchdog to send 
SIGALRM to the ping sender, not the receiver.  The ping sender
task is the one spawned at entrypoint start_pinging, and it is
waiting at a call to pause() for a signal to be received.  The
receiver of those pings is the task with entrypoint receive_pings.

The check that is in my example for EINTR after recvfrom is just
standard error-checking practice for error returns from a recvfrom
call.

> If I call sigvec(SIGALRM, 0, &tempSIGVEC) from shell
> there doesn't seem to be anything set to tempSIGVEC !!  Contents are
> all 0.

When you call sigvec from the shell, aren't you operating on the
signal handlers for the shell task?  I don't know if what you are
seeing there is not expected.

I would try a staged approach.  Put a logmsg call in your watchdog
routine.  Is it being called as expected?  I know that you say it
is being set up correctly, but is it ever actually called?  

Then try several of the VxWorks calls to get a signal handler 
registered by the task doing the recvfrom.  Try to get a signal 
to a task that is not in recvfrom, but rather doing a pause().  

Break your problem down into subproblems.  Determine what you can
and can't get to work.  Then you can ask a more specific question
once you have determined exactly where your implementation breaks
down.

Regards,

Chuck Meade
The PTR Group, Inc.
www.ThePTRGroup.com


> Newsgroups: comp.os.vxworks
> Subject: Interrupting blocked I/O recvfrom calls
> Date: 27 Oct 2002 05:23:22 -0800
> From: mbenari@ecitele.com (mbenari@outthere.com)
> Organization: http://groups.google.com/
> Message-ID: 
> 
> I am trying to interrupt a task that is currently pending on a recvfrom()
> call to a UDP Socket --- I am sort of rehashing an OLD thread 
> "re: ping command available?" 1993-01-22; Chuck Meade in that I
> am trying to utilize his techniques to interrupt the recvfrom call by
> using signals.
> 
> There are various techniques to interrupt a socket that I have tried:
> NOTE: I am testing this without actual packets being sent from remote
> host.
> 
> 1.  Setting SOL_SOCKET / SO_RCVTIMEO with the wanted timeout.
>       I don't believe this is supported in vxWorks 5.4
>       If it is, this is the preferred method.
>          code would look like this:
> 
>       struct timeval tv;
>       tv.tv_sec	= 5;
>       tv.tv_usec	= 0;
>       setsockopt(sockfd, SOL_SOCKET, SO_RCVTIMEO, &tv, sizeof(tv));
> 
>       while (1)
>          {
>          n= recvfrom(sockfd, recvbuf, RECVBUFSIZE, 0, NULL, NULL);
>          if (n < 0)
>             {
>             if (errno == EWOULDBLOCK)
>                {
>                ... do something
>                }
>             else
>                {  	
>                ... recverror
>                }
>             }
>             ... handle packet
>          }
> 
> 
> 
>       THIS should work, but unless I am actually receiving packets on
>       the socket, I NEVER get to EWOULDBLOCK I am assuming that SO_RCVTIMEO
>       is unsupported in vxWorks.
> 
> 2.  Preceeding calls to recvfrom() with calls to select() with timeout on 
>       timeval.  This works and I will NOT include code HERE.  However, for
>       both this case and case #1 (above), I am in essense changing the
>       behavior of my task from one that is truely blocking on the socket
>       itself (in _sbwait() on an IO semaphore), to one that is in essense
>       NON-BLOCKING and polling with a resolution of timeval in select().
>       This solution works for the purposes that I require, however it does 
>       influence the behavior of the waiting task - LESS desirable.
> 
> 3.  Interrupting the socket call by sending a signal to the task.
>       This solution is the more desirable.  However, I am having problems 
>       getting this to work.
> 
>       Code attached.
> 
>    void dummy_alarm_handler(int);
>    SIGVEC  myHandlerInfo = {dummy_alarm_handler,0,0} ;
>    SIGVEC  mySavedHandlerInfo ;
>    int     myTaskId;
>    WDOG_ID myWDId;
> 
> 
>    /* forward declarations */
> 
>    LOCAL void myTaskBody (void);
> 
>    void dummy_alarm_handler(int i)
>       {
>       ;
>       }
> 
>    STATUS myTaskInit (void)
>       {
>       .
>       .   assume that the socket is setup as needed
>       .
>       sigInit();
>       sigvec(SIGALRM, &myHandlerInfo, &mySavedHandlerInfo
>    
>       myTaskId = taskIdSelf() ;
>       myWDId= wdCreate();
>       if (myWDId == NULL)
>          {
>          printf("\r\nWDCreate failure!!\r\n");
>          }
> 
>       return (OK);
>       }
> 
>    LOCAL void myTaskBody (void)
>       {
>       int                       pktSize;           /* size of recvd packet */
>       struct sockaddr_in        remoteAddr;    /* remote transport address */
>       struct sockaddr_in        localAddr;     /* local transport address */
>       char                      rcvBuf [MY_BUF_SIZE];  /* buffer for input */
>       int                       sockSize;
> 
>       sockSize = sizeof (struct sockaddr_in);
> 
>       FOREVER
>          {
>          if (mySuspendFlag == MYTASK_SUSPEND_SELF_ID)
>             {
>             /* turn OFF suspend Flag for next time */
>             mySuspendFlag= 0;
>             
>             printf("actually suspended HERE !!");
>             taskSuspend(taskIdSelf());
>             printf("Resumed from HERE !!\r\n");
>             }
> 
>          if ( (pktSize = recvfrom (mySocket, rcvBuf, MY_BUF_SIZE, 0,
>                                   (struct sockaddr *) &remoteAddr, &sockSize))  
>                                   <= 0  )
>             {
>             
>             /* If pktSize is error (<0) or other printerr data */
>             if (pktSize < 0 )
>                {
>                switch (errno)
>                   {
>                   case EWOULDBLOCK:
>                      printf("recvfrom TIMEOUT !!");
>                      break;
>                   case EINTR:
>                      printf("recvfrom INTR via alarm !!");
>                      break;
>                   default:
>                      printf("recvfrom OTHER errno !!");
>                      break;
>                   }
>                }
>             else
>                {
>                printf("recvfrom == 0 (No Packet Data to Process)");
>                }
>             
>             continue;
>             }
>          sockSize = sizeof (struct sockaddr_in);
>          getsockname (mySocket, (struct sockaddr *) &localAddr, &sockSize);
> 
>          printf("Received packet...");
>   
>          processPacket (pktSize, rcvBuf, &remoteAddr, &localAddr, &mySocket);
>          }
>     }
> 
> 
>    int myWDRoutine(void)
>       {
>       kill(myTaskId, SIGALRM);
>       return (0);
>       }
> 
>    void myTaskSelfKill(void)
>       {
>       wdStart(myWDId, 1300, myWDRoutine, 0) ;
>       }
> 
> 
> 
> Calling myTaskSelfKill starts the wdTimer correctly, However for some
> reason the call to Kill in myWdRoutine() has no affect, and therefore,
> I never get to interrupt my task.  I checked with wdShow() to ensure
> that the WD was setup with the correct function.   It seams as if sigvec
> has NO EFFECT !!  If I call sigvec(SIGALRM, 0, &tempSIGVEC) from shell
> there doesn't seem to be anything set to tempSIGVEC !!  Contents are
> all 0.
> 
> Signals don't seem to be setup appropraitely!
> 
> myTask remains locked on the semaphore in recvfrom() ...
> 
> Am I missing anything here?
> 
> Any/all help would be appreciated.
> 
> Thanks.
> 
> - -Mike



From vxwexplo-errs@csg.lbl.gov  Tue Oct 29 14:42:43 2002
From: Jim Whyte 
Date: Tue Oct 29 14:42:44 PST 2002
Subject: Using fast RAM for .data section?



This is a multi-part message in MIME format

---------5311559753115597
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: quoted-printable


Regarding vxWorks:
Does anyone know how to put a .data section into some
fast (on chip) RAM at a given address?
It seems like the linker has some support for this with
the "MEMORY" command but how can I use this to put
one file's .data secion at a specific address?

Thanks,
Jim
email: jwhyte@brecis.com


---------5311559753115597
Content-Type: text/html; charset=iso-8859-1
Content-Transfer-Encoding: quoted-printable






one file's .data secion at a specific address?



Thanks,


Jim


email: jwhyte@brecis.com



 



---------5311559753115597--



From vxwexplo-errs@csg.lbl.gov  Wed Oct 30 04:03:18 2002
From: Vxworks Exploder 
Date: Wed Oct 30 04:03:20 PST 2002
Subject: comp.os.vxworks newsdigest

Comp.Os.Vxworks Daily Digest    Wed Oct 30 04:03:16 PST 2002

        Subject: Re: Linux
        Subject: Interrupt latency variation
        Subject: Runaway stdout?
        Subject: Re: ip route path 
        Subject: Re: USB camera in vxworks ? Re: USB
        Subject: Re: Newbie Interrupt Service Routine Question
        Subject: Re: Load vxWorks modules twice
        Subject: Re: Semaphores for WatchDog timers.
        Subject: delete file
        Subject: Re: vxWorks BSP manual needed
        Subject: Re: sysClkRate reasonable values..
        Subject: Re: TCP/IP stacks performance
        Subject: Re: Newbie Interrupt Service Routine Question
        Subject: Re: Interrupt Handling and Scheduler
        Subject: Re: Linux
        Subject: Re: CP7000 and DMA
        Subject: Bootrom size question
        Subject: Re: delete file
        Subject: Re: Using fast RAM for .data section?
        Subject: Re: Bootrom size question

-------------------------------------------------------

Newsgroups: comp.os.vxworks
Subject: Re: Linux
Date: Mon, 28 Oct 2002 13:39:54 +0100
From: "David" 
Message-ID: <3dbd3019$0$1388$4d4ef98e@read.news.ch.uu.net>
References: 

Hi,
why not using FreeBSD or OpenBSD?
Dave

"brian"  schrieb im Newsbeitrag
news:d2f67bd8.0210230927.164179ce@posting.google.com...
> Hi Folx,
>
> Has anyone out there been told by their company, or heard of others
> being told, that they can't use Linux, due to potential GPL issues?  I
> realise that the debate goes on, however Im wondering what the large
> companies with lawyers think?
>
> Thanx in advance
>
> brian



---------------------------

Newsgroups: comp.os.vxworks
Subject: Interrupt latency variation
Date: Mon, 28 Oct 2002 18:32:11 +0100
From: Christophe Couturier 
Organization: Guest of France Telecom Oleane's newsreading service
Message-ID: <3DBD749B.725966A8@st2e.com>

    Hello all,

    By looking the TimeBase Register of our MPC860 in a timer interrupt,
we mesured a +/10 µs variation of the interrupt lattency time.
    Is this normal?
    Does anoyone know the cause of this problem and if there is a way to
reduce this variation?

    Christophe


---------------------------

Newsgroups: comp.os.vxworks
Subject: Runaway stdout?
Date: Mon, 28 Oct 2002 19:22:54 GMT
From: "Chad" 
Organization: Shaw Residential Internet
Message-ID: 

I've been having some issues with maintaining control of my I/O between
target shell/host shell for some time and thought I had it sorted, but now
I'm seeing printf() spit to the target shell in one task and the tornado
shell in another.  Currently I have no target shell components at all
included, so all the IO I see there is the boot process, and IO from the
initial usrAppInit launch calls, then the sporatic output from stdout.

Has anybody encountered similar behavior?



---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: ip route path 
Date: Mon, 28 Oct 2002 23:48:47 GMT
From: "somebody" 
Organization: Magma Communications Ltd.
Message-ID: 
References:   <2Yhu9.16691$V21.462284@news> 

Leon,
Thanks for the info. I will look into that. Sorry for the confusion about
the name. I was a misshap.
       -       Hadi
"Leonid Rosenboim"  wrote in message
news:newscache$e31l4h$nw7$1@lnews.actcom.co.il...
> Are you sire tis is like a DSL modem ? Because if it is, a DSL modem
> does not forward just any packet, it usually recovers a PPPoA
> packet and sends it to one particular MAC address as PPPoE,
> all this is done at Layer 2, without ever recovering an IP frame.
>
> When you get an ATM packet, what encapsulation does it use ?
> Is it Classical IP over ATM or Lan Emulation ?
>
> In any event, you can construct yoru own Ethernet MAC frame
> and simply shoot it out with etherOutput() - only the source MAC
> and FCS fields will be overridden by the MAC, all the rest of the
> MAC frame will be transmitted exactly like you have assembled it.
>
> HTH,
>  Leonid
>
> P.S. You have signed two different names in your postings,
> which is rather confusing.
>
> "somebody"  wrote in message
> news:2Yhu9.16691$V21.462284@news...
> > Thanks Leon,
> > Yes you are right I am bit concern about IP routing performance. Our box
> is
> > very similiar to a  DSL modem. We  have only  two interfaces. Atm
> interface
> > and  an ethernet interface. When a packet comes from from the ATM
> interface
> > I just wants to forward the packet  to the ethernet interface without
> > making any routing decision. Is there any way of doing it.
> >                           Hadi
> > "Leonid Rosenboim"  wrote in message
> > news:newscache$y59j4h$ni3$1@lnews.actcom.co.il...
> > > "somebody"  wrote in message
> > > news:q%Zt9.16579$V21.458656@news...
> > > > Hi,
> > > > I have a general question. In our system we only have two
interfaces.
> > When
> > > a
> > > > packet comes in I just want to route the packet to the other
> interface
> > > > without going through the routing table.
> > > >
> > > > Is there any document somewhere, where it talks about how packet are
> > > handler
> > > > between different module of IP layer.
> > > >             Thanks
> > > >                 jon
> > >
> > > Dear Jon at Spacebridge.com,
> > >
> > > your question is too general to be answered logically.
> > > When you say you want to "route" the packet, you basically mean you
need
> > to
> > > "forward" the packet to teh otehr interface, but do you really want to
> > > forward ALL packets ? - if so, you are talking about a Layer-2
> forwarding
> > > mechanism, which is a bit more complicated then that.
> > >
> > > The role fo the routing table in the IP stack is used for Layer-3
> > > orwarding  - also know as "routing", and the information in the
routing
> > > table is used for making decisions about which packets to forward, and
> > which
> > > not to.
> > >
> > > Are you dissatisfied with the standard IP Layer-3 forwarding because
of
> > its
> > > functionality or because of performance ?
> > >
> > > -- Please be more specific, and I will be happy to assist.
> >
> -----------------------------------------------------------------------
> > >    Leonid Rosenboim                Visit:
> > > http://gamla.org.il/english/index.htm
> > >         Consultant                Email: my first name at consultant
dot
> > com
> > >
> > >
> > >
> > >
> >
> >
>
>



---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: USB camera in vxworks ? Re: USB
Date: Tue, 29 Oct 2002 01:12:55 +0000 (UTC)
From: ZhengZhe 
Organization: Your Company
Message-ID: 
References:   <488e459a.0210212115.752d889c@posting.google.com>   

Hi Ravi:
    	We are using UHCI, the chipset is intel's chipset for PentiumIII. 
OHCI is compaq's spec. The chipset is not a matter, whatever it is from 
it must follow one standard either UHCI or OHCI. For vxWorks USB u can 
buy one, or u can migrate from linux. Linux implement both UHCI and
OHCI, u only need to migrate one of them. But it is only the first step
, after that u need to use the linux usb architecture to write your own
driver.
    	We spent three weeks on this, and we met some problem at DMA, my
boss aborted the plan.

Jeff


ravi368@yahoo.com (Ravi Uday) wrote in
news:ecca3ca6.0210280101.dd850d0@posting.google.com: 

> 
> Jeff :
> 
> Whats the OHCI u r using ?? Which chipset ( Host Controller ) are u
> using.
> I am now trying my hands on ScanLogics SL11H Host Controller. But its
> on windows I need one for VxWorks. Any suggestions. ?? My transfer is
> Bulk as well
> Your views are appreciated before hand
> 
> - Ravi
> 


---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: Newbie Interrupt Service Routine Question
Date: Tue, 29 Oct 2002 16:51:01 GMT
From: robotz@robotz.rob (Robotz)
Organization: Randori News - http://www.randori.com - Try our FREE Usenet Scanner!
Message-ID: <3dbeb995.4286646621@news.airband.net>
References: <3db557bb.3671715708@news.airband.net> <3db9b93d.3958838949@news.airband.net> 
Reply-To: robotz@robotz.rob

>The 8260 multiplexes various pin functions, and the interrupts (IRQ1-7) are
>one of them.  Check SIUMCR[DPPC] and make sure it's set for 00b, or
>SIUMCR[L2CPC] and amke sure it's 01b, depending on which IRQ4 line you're
>using.

That solved the main problem and got me going in the right direction
to solve the rest of the problems.  Thanks Dan!  I needed to be
looking at the PPC processor manual (particularly the SIU section)
instead of the vxWorks manuals.


/* ISR on a SBC8260 board */
/* (using IRQ4 via the JP22 connector) */

int volatile mycount;

int immr;

LOCAL void intRecv( UINT32 temp )
{
	/* turn off interrrupt */
	temp = 0x00000800;
	*M8260_SIPNR_H( immr ) = temp;

	mycount++;
}


int sensor()
{
	volatile UINT32 temp;

	/* get base address of processor's
	   memory mapped registers */
	immr = vxImmrGet();

	/* Set up IRQ4 on the DP(4)/IRQ4 pin 
	    courtesy of Dan Gold  :o)  */
	temp = *M8260_SIUMCR( immr );
	temp &= 0xf3ffffff;    /* clear DPPC ...   */
	*M8260_SIUMCR( immr ) = temp;

	/* Set IRQ4 to only trigger
	    on high->low change */
	temp = *M8260_SIEXR( immr );
	temp |= 0x00000800;
	*M8260_SIEXR( immr ) = temp;

	mycount=0;

	intConnect(
		m8260InumToIvec(INUM_IRQ4), 
		(VOIDFUNCPTR)intRecv, 
		0 );

	intEnable( INUM_IRQ4 );

	/* delay 10 seconds */
	taskDelay(sysClkRateGet() * 10);

	intDisable( INUM_IRQ4 );

	printf( "\nmycount = %d\n", mycount );

	return( 0 );
}




---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: Load vxWorks modules twice
Date: Tue, 29 Oct 2002 09:16:52 +0800
From: "Ren Hantao" 
Organization: Lucent Technologies, Indian Hill
Message-ID: 
References:    

But what I saw was the WindShell didn't unload the old one automaticly. Are
there any way to enable it?

Thanks

Hans


"Michael Lawnick"  wrote in message
> As far as I experienced it, the WindShell (Tornado) does an automatic
> unload, target shell (serial/TELNET) just loads a second instance.



---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: Semaphores for WatchDog timers.
Date: Tue, 29 Oct 2002 09:24:05 -0800
From: "Michael R. Kesti" 
Organization: MK Associates
Message-ID: <3DBEC435.D0A0E7CE@gv.net>
References: 
Reply-To: mkesti@gv.net

Shiny wrote:

>Hi,
>This is my flow of the program. I have a task named Task, which
>creates a semaphore and takes it. The semaphore is taken before the
>timer is created. Inside the Timer CB also the semaphore is taken. The
>semaphoere is given after 10 seconds in the main task. The expected
>result is that the TimerCB should wait in the semaphore until it is
>given from the main task(for 10 seconds). But that doesnot happen. I
>think watchdog timers donot wait on semaphores. If I want to achieve
>my expected result, what kind of semaphore should I use?
>
>int Task(void)
>{
>
>    semBinary = semBCreate(SEM_Q_FIFO, SEM_FULL);
>    printf("\nTaking Semaphore in the task before creating the timer\n");
>    semTake(semBinary,WAIT_FOREVER);
>
>    printf("Creating the timer\n");
>    wdogId = wdCreate();
>    if ( wdogId == NULL )
>    {
>        printf( "timedemo: couldn't create watchdog\n" );
>        return ERROR;
>    }
>    else
>    {
>        if ( wdStart( wdogId, sysClkRateGet() * 5, (FUNCPTR)TimerCB,
>                        0) == ERROR )
>        {
>            printf( "timedemo: couldn't create watchdog\n" );
>            return ERROR;
>        }
>        else
>        {
>            printf("In 10 seconds timer\n");
>        }
>    }
>
>    printf("Waiting for 10 seconds\n");
>    taskDelay (sysClkRateGet ( ) * 10);
>
>    printf("Giving semaphore inside Main task\n");
>    return 0;
>}
>
>void TimerCB(void)
>{
>    time(&timet);
>    printf("time1 : %u\n",timet);
>
>    printf("Timer call back Fired\n");
>    printf("Taking semaphore inside timer call back\n");
>    semTake(semBinary,WAIT_FOREVER);
>
>    if (wdStart( wdogId, sysClkRateGet() * 10, (FUNCPTR)TimerCB,
>                        0) == ERROR )
>    {
>        printf( "timedemo: couldn't create watchdog\n" );
>        return;
>    }
>
>    printf("Giving semaphore inside timer call back\n");
>    semGive(semBinary);
>}
>
>TIA, Shiney.

It appears that it is your program that is faulty rather than the
semaphore and/or watchdog timer implementations.

When you create semBinary using semBCreate() with the SEM_FULL option
the semaphore is immediately available.  The semTake() in the third
line of Task() therefore returns immediately and semBinary is now
empty/unavailable and, because there is no semGive(semBinary) call,
semBinary will remain empty indefinately.  Later, when TimerCB() calls
semTake(semBinary,WAIT_FOREVER), semTake() will therefore never return.

It gets worse.  TimerCB() is called from an interrupt service routine
context, which places constraints on what it may safely do.  Two things
that it cannot safely do is call semTake() and printf().

If you deal with these issues and write a valid program you should
find that vxWorks semaphores and watchdog timers work just fine.

- -- 
========================================================================
          Michael Kesti            |  "And like, one and one don't make
                                   |   two, one and one make one."
          mkesti@gv.net            |          - The Who, Bargain

---------------------------

Newsgroups: comp.os.vxworks
Subject: delete file
Date: Tue, 29 Oct 2002 11:39:58 -0800
From: "Roger C." 
Organization: Jet Propulsion Laboratory - Pasadena CA
Message-ID: 

Is there any function to delete file under VxWorks? Thanks for your help.

- --
Regards,

Roger



---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: vxWorks BSP manual needed
Date: Tue, 29 Oct 2002 20:40:41 GMT
From: "H. E. Taylor" 
Organization: MTS Internet
Message-ID: <3DBF0B89.692D@despam.autobahn.mb.ca>
References:  

In article ,
 Chad wrote:
> "vb"  wrote in message
> news:CwYr9.216$zE6.1018@rwcrnsc51.ops.asp.att.net..
>>
>> I wonder if anyone can share a document describing vxWorks BSP internals,
>> API, etc.
> 
> I found a document entitled "VxWorks - Device drivers in a nutshell" which I
> believe was gnu public licensed several months ago, I found it quite
> informative and if I can dig it up I'll email it to you.  It gives a general
> device driver framework and example, though I believe (I've yet to look into
> it in too much detail) things have changed a bit in vxWorks 5.5.
> 

	Could you put it on the web, or alternatively 
	could someone else [Johann?] in the group?  
	This type of doc is frequently requested, but
	unfortunately I don't have online room for it myself. 
	[I'm near my 10 meg limit.]


- -het
	


- --
"The earth is the cradle of mankind, but we cannot stay 
in the cradle forever." Konstantin Tsiolkovsky	 

Computer Links: http://www.autobahn.mb.ca/~het/clinks.html
H.E. Taylor  http://www.autobahn.mb.ca/~het/

---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: sysClkRate reasonable values..
Date: 29 Oct 2002 13:02:28 -0800
From: rick@cyclone.com (Richard Nardone)
Organization: http://groups.google.com/
Message-ID: <29b153d9.0210291302.5b6b717c@posting.google.com>
References: <1035882002.428469@newsreader2.wirehub.nl>

"Renee Teunissen" > 
> For our target we would like to the sysClkRate to something about 1000
> ticks a second, to get our timers and delays more accurate.

"Renee Teunissen" 
> 
> For our target we would like to the sysClkRate to something about 1000
> ticks a second, to get our timers and delays more accurate.
> 
  You want to use a different timer/counter for higher resolution. 
Look into the BSP for a separate device with some access routines
(i.e. a driver) for it.  It is common to have a second timer/counter
which can generate an interupt.
 
  Increasing the system clock to 1000Hz is generally not advisable. 
Each tick causes the kernel to wake up and see what needs to be done. 
This activity consumes CPU cycles, and is the overhead associated with
an RTOS.

  Not sure about the network problem.
 

  -Regards,

   Richard Nardone
   Cyclone Microsystems

---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: TCP/IP stacks performance
Date: Tue, 29 Oct 2002 01:32:15 +0000 (UTC)
From: ZhengZhe 
Organization: Your Company
Message-ID: 
References: <3DB98C1B.84785269@NOSPAM.slb.com>

Frederic  wrote in news:3DB98C1B.84785269
@NOSPAM.slb.com:

> hi all,
> 
> does anyone know where to find informations about TCP/IP stack
> performance under VxWorks ?
> 
> benchmarks should be appreciate.
> 
> Thanks.
> 

Hi Frederic:
    	To improve the performance of vxWorks bundled tcp/ip, the first
thing should be increase all the mBuffer size, this would greately 
improve the performance of the stack. The default buffer size only 
allow u to operate on a small network with a few sockets, download 
somefile. But if u want to use it as the primary communication channel
on use it on a very large network, the stack would be inefficient.
Just increase it. In our case we made all the buffer grows 32X.

Jeff

---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: Newbie Interrupt Service Routine Question
Date: Tue, 29 Oct 2002 14:17:39 -0800
From: "Dan Gold" 
Organization: Mailgate.ORG Server - http://www.Mailgate.ORG
Message-ID: 
References: <3db557bb.3671715708@news.airband.net> <3db9b93d.3958838949@news.airband.net>  <3dbeb995.4286646621@news.airband.net>
Sender: gold@ensemble.com

"Robotz"  wrote in message
news:3dbeb995.4286646621@news.airband.net...
> That solved the main problem and got me going in the right direction
> to solve the rest of the problems.  Thanks Dan!  I needed to be
> looking at the PPC processor manual (particularly the SIU section)
> instead of the vxWorks manuals.

You're welcome.  One other thing worth noting - if you look at the user
manual errata for section 5.4.1 (reset), it lists about 7 fields in the
SIUMCR (including DPPC, L2CPC, etc...) and says "These bits cannot be
changed after reset".  Of course they can (because everyone's doing it).
For future reference (in case Mot suddently changes them to R/O), you might
want to use a new hardware config word (HRCW) that has the correct values so
you don't have to change the software.

  hth,
   Dan Gold  gold@ensemble.com



---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: Interrupt Handling and Scheduler
Date: Thu, 24 Oct 2002 14:01:50 +0200
From: Tim Sinkins 
Organization: Alcatel Kommunikations Elektronik
Message-ID: <3DB7E12E.51138C0F@alcatel-ke.de>
References: <53c6857a.0210232148.5268569a@posting.google.com>
Reply-To: timothy.sinkins@alcatel-ke.de



vemanj wrote:
> 
> Hi,
> I have a doubt about the interrupt handling and the scheduling.
> When an ISR is being executed, that particular interrupt is disable.
> 
> My question is :if the interrupts are coming at very high frequency,
> the other tasks will never get the chance to execute. Is it right???

Yep, If this happens you've either under-specified your processor or 
you have the wrong interrupt strategy for a specific source.

> Please tell me the scheduler reaction to an interrupt.

Nothing.  It will be interrupted and will have no knowledge of
that interruption.

> What happens to the current execution????

It stops whilst the ISR runs.  It is the ISR's job to save and 
restore the context of the previous thread.

Tim

> 
> Thanks in advance.
> 
> Regards,
> Anj

---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: Linux
Date: Mon, 28 Oct 2002 14:25:04 -0600
From: Andrew Johnson 
Organization: APS, Argonne National Laboratory
Message-ID: 
References:    

Leonid Rosenboim wrote:
> "Andrew Johnson"  wrote:
> 
>>To be pedantic about this, you are _not_ forced to release your changes
>>into the public domain, but if you're selling (or even giving away) a
>>product that makes use of GPL code you have to follow the rules of the
>>GPL, which means providing source code to your customers and licensing
>>your changes under the GPL.  However it is legal to mix GPL and
>>proprietary code in the same product (although not in the same program),
>>so you're not forced to give away your secrets provided you're careful
>>about your design.
> 
> [snip]
> 
> Legal ? Yes, in the sense that by this you do not violate any federal or
> state
> law by doing this. But being legal does not necesarily means you are
> protected
> from civil littigation by the owners of other parts of the GPL code.

IANAL, but my meaning of the term "legal" in the above paragraph was 
intended to be "you are not violating the terms of the GPL".

In the specific example of an embedded system that uses a stock Linux 
kernel but runs a completely proprietary user-mode program, I'd be mixing 
GPL and non-GPL code in the same *product* (but not in the same *program*) 
without violating the GPL.  I can run GNU programs from a proprietary 
non-GPL shell script without worrying about having legal problems with 
RMS, providing I follow the GPL terms for any modifications I make to the 
GNU code.

The FSF have never stopped a GNU/Linux distributor from including GNU 
products along with commercial non-GPL packages, because that is 
explicitly permitted by the GPL (aggregation clause).  There is no 
difference as far as this clause is concerned between selling a GNU/Linux 
distribution on CD and selling a product that has a Disk-On-Chip with the 
same binaries installed on it.

The GPL may have enforced at least one significant difference between 
traditional embedded systems and those based on Linux: Modularity, and 
partitioning of the system into multiple programs, rather than having all 
of the code linked together into a single monolithic binary.  This usually 
opens up such embedded systems more to modification by the end user, and 
as an end user of such systems I thoroughly appreciate this.

- - Andrew
- -- 
"Life is what happens while you're busy making other plans." - John Lennon


---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: CP7000 and DMA
Date: 29 Oct 2002 16:17:09 -0800
From: kiranmandava@hotmail.com (kmandava)
Organization: http://groups.google.com/
Message-ID: 
References: <4d53234f.0210202243.2c15e9b0@posting.google.com>

shenoy_sudhakar@hotmail.com (Sudhakar Shenoy) wrote in message news:<4d53234f.0210202243.2c15e9b0@posting.google.com>...
> Hi,
>     we have a Ocelot board with R4000 Mips processor. the board
> supports 2 PMC slots on it along with the Galileo system controller.we
> have a custom made PMC card.
>    if i want to do DMA some data using the gtDma() (galileo dma
> routine provided by the BSP of ocelot ,cp7000), if the Data is written
> in memory allocated through "cacheDmaMalloc", it works fine. but if i
> try writing the same data just by using the "malloc" and DMAing it, it
> doenst respond.
>    is it necessary to do some other setting for the same to work??
> please let me know,
> thanks n regds,
> Sudhakar Shenoy.

cacheDmaMalloc : it allocates the memory which is cache safe which
could be used for the DMA stuff..!

malloc doesn't care anything about the caches..

ofcourse it is not true on all the processors...
just to try this make the whole of your memory map to be not cacheable
and try it..!

kiran

---------------------------

Newsgroups: comp.os.vxworks
Subject: Bootrom size question
Date: Wed, 30 Oct 2002 00:28:35 +0000 (UTC)
From: owen@manta.nosc.mil (Wallace E. Owen)
Organization: SPAWAR Systems Center, San Diego, CA 92152
Message-ID: 

I'm trying to build a moderately large (~1.2MB text segment) app to be
booted and run from a floppy on an embedded PC.  A helloworld app builds
fine, but when I try to make my full app's boot image, romsize complains
that the imagre is ~600K too large.  I adjusted the RAM_LOW_ADRS from
0x00108000 to 0x00208000, and changed the ROM_SIZE from 0x00090000 to
0x00190000 thinking that this would make enough room for the image and
the boot loader (I'm trying to make my app boot from floppy).  The app
builds without complaint, but it just hangs after the loader loads it.

Any clues appreciated.


  // Wally

---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: delete file
Date: Tue, 29 Oct 2002 18:25:23 -0800
From: "Michael R. Kesti" 
Organization: MK Associates
Message-ID: <3DBF4313.7D4FE9C9@gv.net>
References: 
Reply-To: mkesti@gv.net

"Roger C." wrote:

>Is there any function to delete file under VxWorks? Thanks for your help.

See remove() in the reference manual's ioLib section.

- -- 
========================================================================
          Michael Kesti            |  "And like, one and one don't make
                                   |   two, one and one make one."
          mkesti@gv.net            |          - The Who, Bargain

---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: Using fast RAM for .data section?
Date: Wed, 30 Oct 2002 08:11:00 +0100
From: "Michael Lawnick" 
Organization: Buergernetz Dillingen
Message-ID: 
References: <200210292242.g9TMgi924215@solar.brecis.com>
Reply-To: "Michael Lawnick" 
Sender: mlawnick@145.253.156.162

Hi Jim,

this depends on the Tornado/VxWorks version you use.
Up to T2..0.2/VxWorks5.4.2 : IMO forget it, I tried something similiar and
failed as its much too complicated.
T2.1: don't know
T2.2/VxWorks5.5: just edit the linker script.

FYI: Up to T2..0.2/VxWorks5.4.2 the linker uses intrinsic rules for linking,
so changing one section leads to redesign of all rules - you have to be
experienced in GNU tools to accomplish this. Since T2.2 explicit linker
rules are used, so it should be not too complicated to change them along
your needs.

Regards,
Michael

"Jim Whyte"  schrieb im Newsbeitrag
news:200210292242.g9TMgi924215@solar.brecis.com...
>
>
> This is a multi-part message in MIME format
>
> ---------5311559753115597
> Content-Type: text/plain; charset=iso-8859-1
> Content-Transfer-Encoding: quoted-printable
>
>
> Regarding vxWorks:
> Does anyone know how to put a .data section into some
> fast (on chip) RAM at a given address?
> It seems like the linker has some support for this with
> the "MEMORY" command but how can I use this to put
> one file's .data secion at a specific address?
>
> Thanks,
> Jim
> email: jwhyte@brecis.com
>
>
> ---------5311559753115597
> Content-Type: text/html; charset=iso-8859-1
> Content-Transfer-Encoding: quoted-printable
>
> 
> 
>  252">
>  ">
> 
> 
>
> 
> 
Regarding vxWorks:

> 
Does anyone know how to put a .data section into some

> 
fast (on chip) RAM at a given address?

> 
It seems like the linker has some support for this with

> 
the "MEMORY" command but how can I use this to put

> 
one file's .data secion at a specific address?

>
> 
Thanks,

> 
Jim

> 
email: jwhyte@brecis.com

>
> 
 

> 
>
> ---------5311559753115597--
>



---------------------------

Newsgroups: comp.os.vxworks
Subject: Re: Bootrom size question
Date: Wed, 30 Oct 2002 11:17:51 +0200
From: "Leonid Rosenboim" 
Organization: Verio
Message-ID: 
References: 
Reply-To: "Leonid Rosenboim" 

On the PC architecture, there is the limit of low memory size
at 640KB, that is where you're stuck.

When you boot off a floppy, the boot block uses BIOS calls
to load a file from floppy into the memory, and this BIOS
works in Real Mode, therefore limitted to the lower 640K
region.

You may want to leave the Bootrom as it is, and have the
defaul boot line declare the floppy as boot device, and have
a second file, vxWorks.st on the same floppy which is loaded
into high memory (1MB+) by the bootrom, in protected mode.

Problem is, the bootrom and application won't fit a floppy, unless
you use the 2.8MByte high density floppy, os some other storage
solution.

- --
- -----------------------------------------------------------------------
   Leonid Rosenboim                Visit:
http://gamla.org.il/english/index.htm
        Consultant                Email: my first name at consultant dot com


"Wallace E. Owen"  wrote in message
news:apn93j$sei$1@newpoisson.nosc.mil...
> I'm trying to build a moderately large (~1.2MB text segment) app to be
> booted and run from a floppy on an embedded PC.  A helloworld app builds
> fine, but when I try to make my full app's boot image, romsize complains
> that the imagre is ~600K too large.  I adjusted the RAM_LOW_ADRS from
> 0x00108000 to 0x00208000, and changed the ROM_SIZE from 0x00090000 to
> 0x00190000 thinking that this would make enough room for the image and
> the boot loader (I'm trying to make my app boot from floppy).  The app
> builds without complaint, but it just hangs after the loader loads it.
>
> Any clues appreciated.
>
>
>   // Wally



---------------------------

End of New-News digest
**********************


From vxwexplo-errs@csg.lbl.gov  Wed Oct 30 08:15:04 2002
From: Mike McCullough 
Date: Wed Oct 30 08:15:05 PST 2002
Subject: Re: delete file

Roger, I believe you can just use the usrLib.c (usrFsLib.c) source code 
that Wind provides to do this. This code provides the typical dos-like 
commands such as copy, del, etc.

As far as I know you have to manually insert this into your VxWorks image 
either using the Project tool or MACH_EXTRA in the Makefile. I "think" that 
the new dosFs 2.2 patch fixes this problem but I am not 100% sure. I would 
also highly recommend adding dosFs 2.2 to a Tornado 2.0.2 tree. It solves a 
few quite nasty problems and doesn't force you to upgrade to T2.2. This 
patch is downloadable off the support page...


Newsgroups: comp.os.vxworks
Subject: delete file
Date: Tue, 29 Oct 2002 11:39:58 -0800
From: "Roger C." 
Organization: Jet Propulsion Laboratory - Pasadena CA
Message-ID: 
Is there any function to delete file under VxWorks? Thanks for your help.
- --
Regards,
Roger



**********************************************************
Mike McCullough                    MCC Engineering
President and CEO                  9 Cheshire Court
                                    Londonderry, NH 03053

mike@mccengineering.com            Tel:  603-537-9593
                                    Fax:  603-434-9890
www.mccengineering.com            Cell: 603-247-1219

**********************************************************
        24x7 VxWorks Support  Custom Support Programs
          VxWorks Training   Embedded Consulting Services
                     A WindLink Partner
**********************************************************



From vxwexplo-errs@csg.lbl.gov  Wed Oct 30 10:30:42 2002
From: Shiv 
Date: Wed Oct 30 10:30:44 PST 2002
Subject: Symbol Table problems on VisionClick , VxWorks


I am pretty sure I'm missing something a bit rudimentary , but why is it
that when I add a bunch of function calls to the UsrAppInit() module in the
WRS BSP for the 855T , that the VisionClick Debugger turns up a bunch
unresolved symbols , not for the entire code segment , but ONLY for a
selective few local variables that come up as "unresolved" when I probe them
via break-points?

Any clues or suggestions would be appreciated !!
Shiv


From vxwexplo-errs@csg.lbl.gov  Thu Oct 31 07:21:16 2002
From: Linux G Shang 
Date: Thu Oct 31 07:21:18 PST 2002
Subject: Re: Symbol Table problems on VisionClick , VxWorks

Dears:
    If I use ifAddrSet("xxxx0","1.2.3.4") change my vxWorks's ip address,it can work correctly.But I want to know what END Driver function was called for saving the new address into NVRAM? vxWorks use ipIoctl(.....)??????Thanks for your response.


From vxwexplo-errs@csg.lbl.gov  Thu Oct 31 14:36:55 2002
From: John Sambrook 
Date: Thu Oct 31 14:36:57 PST 2002
Subject: How to check dosFs/TFFS data integrity before attempting a mount?

Hello,

The system I am working on has a remote flash device located at
the end of a proprietary bus.  The customer can attach different
flash devices from time to time. 

I have written software that is able to read, write and erase the 
flash, and I have this working as well with Tffs, where I wrote my 
own MTD and socket layer code.

I can format the flash and mount dosFs, and it seems to work.  I
can run chkdsk on the mounted volume, etc.

My question is this.

In an uncontrolled environment, users will connect flash devices
that have not been properly formatted.  I don't want this to result
in a system crash.

Is there an established way of testing for the presence of a tffs-
based dosFs filesystem on the flash device, before attempting to
mount the filesystem?  

I'd be grateful for any pointers or suggestions.

Thanks,

John Sambrook

vxworks 

__________________________________________________
Do you Yahoo!?
HotJobs - Search new jobs daily now
http://hotjobs.yahoo.com/


From vxwexplo-errs@csg.lbl.gov  Thu Oct 31 20:47:31 2002
From: radha@telecruz.com
Date: Thu Oct 31 20:47:34 PST 2002
Subject: IrDA stack on vxWorks source code downloads path

Hi All,

           If you have IrDA stack on VxWorks  source code (or) down 
loads URL, please forward to me. I will be very thankfull to you.

regards,
radhakrishna