From daemon@csg.lbl.gov Wed May 1 04:00:55 1996 From: daemon@csg.lbl.gov Date: Wed May 1 04:01:00 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Wed May 1 04:00:50 PDT 1996 Subject: HELP: File System/Disk mirroring Subject: Re: VXWORKS on the HP ???? Subject: Changing default NFS perms. on create Subject: Re: TCP-IP Accelerator ??? Subject: Re: Forcing Cold Boot from Vxworks Subject: Re: TCP-IP Accelerator ??? Subject: Re: Forcing Cold Boot from Vxworks ------------------------------------------------------- Newsgroups: nortel.magellan.*,comp.os.vxworks Subject: HELP: File System/Disk mirroring Date: 18 Apr 1996 18:51:06 GMT From: Ashwin Kumar Organization: Nortel Message-ID: <4l62uq$q2n@nrchh52.rich.nt.com> Hi, Any ideas on disk mirroring software on top of vxworks ? Any info will be appreciated, Thanks. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VXWORKS on the HP ???? Date: 30 Apr 1996 13:06:55 GMT From: partin@heman.jsc.nasa.gov (Charles Partin) Organization: Lockheed Engineering & Sciences Co. Message-ID: References: <31815875.2BB7@phbtsus.com> Followup-To: comp.os.vxworks,comp.sys.hp.hpux In article <31815875.2BB7@phbtsus.com>, Aseem Bakshi wrote: > Hi, > > Under Sun Solaris it is possible to run Vxworks. Is it possible to do > the > same with HP/UX ????? > We run VxWorks 5.2 under HPUX 9.05 on a 9000/755 with no problems. > Are there any other options available to run real time under Hp/UX ??? > Thanks Aseem I am sure there are other real-time systems (pSOS, VTRX, etc) running on HP's. I have never had the need to look for them. Charlie - -- Charlie Partin email: partin@heman.jsc.nasa.gov Lockheed Engineering voice: 713-333-6803 & Sciences Co. fax: 713-333-6908 --------------------------- Newsgroups: comp.os.vxworks Subject: Changing default NFS perms. on create Date: 30 Apr 96 10:09:00 EST From: m1w@icf.hrb.com (Michael E. Warren) Organization: HRB Systems, Inc. Message-ID: <1996Apr30.100900.24586@hrbicf> Is there a convenient was to change the default permissions used when a file is created on an NFS file system? When using the copy command, new files are created on the file server using the 644 permissions. Is there anything like the csh umask? Mike Warren HRB Systems State College, PA --------------------------- Newsgroups: comp.realtime,comp.os.qnx,comp.os.vxworks,comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains,vmsnet.networks.tcp-ip.misc,vmsnet.networks.tcp-ip.multinet,de.comm.protocols.tcp-ip,comp.arch.embedded Subject: Re: TCP-IP Accelerator ??? Date: 26 Apr 1996 17:18:46 GMT From: raj@cup.hp.com (Rick Jones) Organization: the Unofficial Hewlett-Packard Message-ID: <4lr0hm$5is@hpindda.cup.hp.com> References: Followup-To: comp.realtime,comp.os.qnx,comp.os.vxworks,comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains,vmsnet.networks.tcp-ip.misc,vmsnet.networks.tcp-ip.multinet,de.comm.protocols.tcp-ip,comp.arch.embedded Reply-To: raj@cup.hp.com Russell Nelson (nelson@ns.crynwr.com) wrote: : In article peeds@netcom.com (LataAndPd) writes: : > Does anyone know of any add-on accelerator cards for : > TCP/IP which will offload the system's TCP/IP load ? If so, : Problems with this idea: : o TCP has a pretty low overhead for a clean link (Van Jacobsen's 30 : ... : o Very often the "accellerator" cards run a processor MUCH slower : than the main processor. : o The interface between the system and the card often becomes the : bottleneck. "Smart" interfaces can indeed be too smare for their own good. However, "idiot-savant" cards can be helpful - those that can provide internet checksum offload in each direction, and then perhaps also determine the split between protocol headers and data on inbound and do the DMA accordingly. At least some OSes have support for such things today - Irix and HP-UX to name two, there may be others. Now if we could just get all the third-party card developers on board :) rick jones --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Forcing Cold Boot from Vxworks Date: 26 Apr 1996 13:46:14 -0700 From: kla@leland.Stanford.EDU (Earl Mitchell) Organization: Stanford University, CA 94305, USA Message-ID: <4lrcmm$bn4@elaine25.Stanford.EDU> References: <4lpa5f$o4e@netline-fddi.jpl.nasa.gov> <4lpcmd$sq3@elaine25.Stanford.EDU> <4lqma5$8tn@hppadbk.waterloo.hp.com> In article <4lqma5$8tn@hppadbk.waterloo.hp.com>, Don Bowman wrote: >Earl Mitchell (kla@leland.Stanford.EDU) wrote: >: In article <4lpa5f$o4e@netline-fddi.jpl.nasa.gov>, >: Gerard Benenyan wrote: >: > >: >Does anyone have experience on forcing a system reboot >: >in a vxworks environment. All attempts to jump to >: >location 0xffffff0, which is the cold boot rom location have failed. >: > >: The stack is typically used to store a return instruction >: pointer (RIP) which is stuffed in the IP register when >: stack frame is popped. By modifying this you can trick the >: system into jumping to 0xffffff0 when returning from a call. >: You just need to figure out where RIP is located. > >: coldBoot() >: { >: rip_addr = (U32) 0xffffff0 ; >: return >: } > > >That's awfully error prone. >Why not just call the reboot() function? > >If that's no good for some reason, then insert an __asm__ thing there. > >coldBoot() >{ > __asm__("bx 0xfffffff0"); >} > >I don't know which processor you're using, but the opcode should be >something like that. From his description of the problem (i.e. all attempts to jump to ... have failed) I assumed he had tried this already. --------------------------- Newsgroups: comp.realtime,comp.os.qnx,comp.os.vxworks,comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains,vmsnet.networks.tcp-ip.misc,vmsnet.networks.tcp-ip.multinet,de.comm.protocols.tcp-ip,comp.arch.embedded Subject: Re: TCP-IP Accelerator ??? Date: 25 Apr 1996 22:10:55 -0400 From: mo@uunet.uu.net (Mike O'Dell) Organization: Centre for Chaotic Repeatabilty Message-ID: <4lpbbf$7nv@rodan.UU.NET> References: Reply-To: mo@uunet.uu.net usually the key to good TCP performance is not a front-end processor but getting a good Ethernet card which doesn't require much processor intervention. having built front-end processors before, I've never seen one where the protocol to talk to the front-end wasn't as complex as TCP itself, so there was no real gain at the end of the day. good TCP code + good network card = good performance. enjoy. -mo --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Forcing Cold Boot from Vxworks Date: 26 Apr 1996 14:24:05 GMT From: bowman@waterloo.hp.com (Don Bowman) Organization: H-P Panacom Div, Waterloo, ON Canada Message-ID: <4lqma5$8tn@hppadbk.waterloo.hp.com> References: <4lpa5f$o4e@netline-fddi.jpl.nasa.gov> <4lpcmd$sq3@elaine25.Stanford.EDU> Earl Mitchell (kla@leland.Stanford.EDU) wrote: : In article <4lpa5f$o4e@netline-fddi.jpl.nasa.gov>, : Gerard Benenyan wrote: : > : >Does anyone have experience on forcing a system reboot : >in a vxworks environment. All attempts to jump to : >location 0xffffff0, which is the cold boot rom location have failed. : > : >Vxworks does not allow jumping into another task space. : >Someone at Wind River suggested forcing a "triple" exception. : >Too contrived. There must be a more elegant solution. : > : >Any help will be much appreciated. : The stack is typically used to store a return instruction : pointer (RIP) which is stuffed in the IP register when : stack frame is popped. By modifying this you can trick the : system into jumping to 0xffffff0 when returning from a call. : You just need to figure out where RIP is located. : coldBoot() : { : rip_addr = (U32) 0xffffff0 ; : return : } That's awfully error prone. Why not just call the reboot() function? If that's no good for some reason, then insert an __asm__ thing there. coldBoot() { __asm__("bx 0xfffffff0"); } I don't know which processor you're using, but the opcode should be something like that. - -- bowman@waterloo.hp.com ** 519-883-3019 ** http://www.waterloo.hp.com/~bowman --------------------------- End of New-News digest ********************** From rbair@astro.ge.com Wed May 1 06:32:40 1996 From: "Robert E. Bair" Date: Wed May 1 06:32:43 PDT 1996 Subject: unsubscribe unsubscribe From edk@wrs.com Wed May 1 08:20:04 1996 From: Ed Klein Date: Wed May 1 08:20:07 PDT 1996 Subject: Re: How long to semGive()? did you get what you wanted?? I have a set of benchmarks that I could fax or email to you ?? Ed Klein 1901 N. Roselle Rd # 800 Wind River Systems Schaumburg, Il 60195 phone: (847)490-6062 fax: (847)490-6511 email: edk@wrs.com WWW: http://www.wrs.com From YF.Siu@eng.efi.com Wed May 1 10:15:45 1996 From: "Yuet Fung Siu" Date: Wed May 1 10:15:48 PDT 1996 Subject: Re: dosFs troubles... The problem you described is very similar to the one I discovered last year with vxWorks 5.2. To confirm this, instead of using the shell, try to spawn a task, i.e., 'sp copy,"srcfile","desfile"', to do the copy and observe what is happening. If you discovered that the new task never exits and is always ready, that's it. If you do a 'tt' or so, you should find that the task is doing some hashLib or sllLib stuff. Since we have the source code to 5.2, I did find the bug in the dosFsLib. The problem was due to a hashList corruption inside the dosFsLib. I've reported the bug to WRS: SPR 4418. I don't think there's a work around without the source code. You may want to contact WRS and ask for a fix or patch. I'm just surprised that it takes so long before anyone else discover this. Probably noone else is using dosFsLib with 5.2 ?? Good Luck, YF From mea@mclean.sparta.com Wed May 1 16:05:34 1996 From: Mike Anderson Date: Wed May 1 16:05:36 PDT 1996 Subject: PRI-48XP Dual T1 Greetings! We're interested in terminating a T1 (1.544 MBPs) line into a VME crate for VxWorks. We're not interested in ISDN (*I*t *S*till *D*oes *N*othing ;-). We purchased a PRI-48XP dual T1 board, but all the documentation talks about is ISDN. We've been trying for almost a week now to get someone on the phone at PRI in tech support to no avail. So, the question is, has anyone out there in netland interfaced the PRI-48XP to a T1 interface for full T1 throughput? If so, what software did you use and is it available? Regards, =============================================================================== __ Real-Time System Development, Integration, Training and Services //\\ // \\ Mike Anderson // /\ \\ Chief Engineer Voice : (703) 448-0210 ext. 235 // / \ \\ SPARTA, Inc. FAX : (703) 734-3323 // \ \\ 7926 Jones Branch Drive EMAIL : mea@mclean.sparta.com \\ \ // Suite 900 Web : http://www.mclean.sparta.com \\ \ / // McLean, VA 22102 \\ \/ // "Software development is like making \\ // a baby... You can't make a baby in one \\// month by impregnating nine women. -- "Pride in Performance" Some things just take time." =============================================================================== From daemon@csg.lbl.gov Thu May 2 04:00:34 1996 From: daemon@csg.lbl.gov Date: Thu May 2 04:00:37 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Thu May 2 04:00:28 PDT 1996 Subject: Re: Socket Reuse Problems Subject: Re: Socket Reuse Problems Subject: Q: Producing boot proms for MVME167 Subject: vxMp with Reflective Memory Subject: Re: Forcing Cold Boot from Vxworks Subject: Custom application under telnet ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Socket Reuse Problems Date: Tue, 30 Apr 1996 14:15:07 -0400 From: Kevin Rice Organization: NASA Goddard Space Flight Center -- Greenbelt, Maryland USA Message-ID: <318658AB.2A2F@vlsi.gsfc.nasa.gov> References: I don't have very many suggestions but I did notice in one of my TCP/IP books that REUSEADDR basically allows the port # to be re-used so if you are trying to bind to the same address, this wouldn't work properly. If that's in order make sure the setsockopt is being called before each bind in *EACH* app! (i should add that i've never actually tried this myself -- just playing back the 'tape.') It wasn't clear if that's what you trying to do this or not. Too bad you aren't running 4.4bsd apparently there is an option called REUSEPORT that would allow multiply open both a port & address! Kevin Rice --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Socket Reuse Problems Date: 1 May 1996 19:49:25 GMT From: j.gordon@dial.pipex.com (John Gordon) Organization: N/A Message-ID: <4m8f85$iib@soap.news.pipex.net> References: <318658AB.2A2F@vlsi.gsfc.nasa.gov> I have a piece of code written for SunOS 4.1.3 on a Sparc that makes use of SO_REUSEADDR and functions correctly, but I did notice the other day when trying to get it to run on our VxWorks 5.2/PowerPC combination that the code was getting address in use errors. At the time I ignored it as I had other problems with the code too - I will investigate a bit further tomorrow! John... --------------------------- Newsgroups: comp.os.vxworks Subject: Q: Producing boot proms for MVME167 Date: 1 May 1996 15:41:58 GMT From: jschamba@physics.utexas.edu (Jo Schambach) Organization: Physics Department, University of Texas at Austin Message-ID: <4m80o6$5cg@geraldo.cc.utexas.edu> Reply-To: JSchamba@utpapa.ph.utexas.edu Forgive my ignorance in this matter, but: how does one produce bootrom code for VxWorks from the BSP in 5.2 that can be used by a PROM burner for MVME167 proms? I found several targets relating to this subject in the Makefile. However, they all seem to produce one file only, while the manual states that there are 2 boot proms needed for VxWorks, arranged in 32 bits, one for even, and one for odd words. I don't see a target that can split the resulting file in such a manner. Am I missing something fundamental here? I would really appreciate if someone could mail me a step-by-step procedure to create boot proms, beginning with a pair of empty proms and the (non-compiled) BSP, that is understandable by a layman like myself ... Thanks in advance, Jo ***************************************************************** Joachim Schambach Research Associate Physics Department, RLM 5.208 The University of Texas at Austin Austin, TX 78712 USA email: JSchamba@physics.utexas.edu Tel: x1 (512) 471-1303 FAX: x1 (512) 471-9637 ***************************************************************** --------------------------- Newsgroups: comp.os.vxworks Subject: vxMp with Reflective Memory Date: 1 May 1996 16:17:07 GMT From: john.weir@msfc.nasa.gov (John Weir) Organization: NASA MSFC Message-ID: <4m82q3$m3t@hammer.msfc.nasa.gov> We have three VME-based vxWorks systems interconnected with a VMIC RMS system which replicates all data writes to an address range to the corresponding address in each system. We are considering using the vxMp product with this system to implement shared memory among our VME chassis. Has anyone tried this? How did it work? Thanks in advance for all replies? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Forcing Cold Boot from Vxworks Date: Fri, 26 Apr 1996 20:51:21 GMT From: mfisher@dyn.com (Mark Fisher) Organization: The World @ Software Tool & Die Message-ID: References: <4lpa5f$o4e@netline-fddi.jpl.nasa.gov> Reply-To: mfisher@dyn.com Sender: news@world.std.com (Mr Usenet Himself) gerard@uccs.jpl.nasa.gov (Gerard Benenyan) wrote: >Does anyone have experience on forcing a system reboot >in a vxworks environment. All attempts to jump to >location 0xffffff0, which is the cold boot rom location have failed. >Vxworks does not allow jumping into another task space. >Someone at Wind River suggested forcing a "triple" exception. >Too contrived. There must be a more elegant solution. >Any help will be much appreciated. > Gerard B. If, and this is a big if, your hardware has a hardware watchdog on the board then enable it and LET IT timeout, then this will usually cause the board to reset itself as if from a cold reboot. Unfortunately, one cannot rely on the vxWorks reboot() function to work in all cases -- it is really nothing more than a jump to re-start the code and as such is really a warm restart. reboot() can fail in cases where an exception is caused by a function that was registered to be called on reboot() -- see rebootHookAdd(). I think vxWorks code calls in quite a few places themselves. reboot() will nearly always work if nothing has gone wrong in the application. But when you need it most, to reset the board after some bad behavior is detected, then it less likely that it will work! I have even seen reboot() cause a double bus error -- which on a Moto CPU makes the CPU go into the halt state. The only thing that will get you out of that is manually re-setting the board or a hardware run watchdog. HTH, Mark. --------------------------- Newsgroups: comp.os.vxworks Subject: Custom application under telnet Date: Thu, 25 Apr 1996 18:02:13 -0400 From: "Jeffrey D. Prem" Organization: FORE Systems, Inc. Message-ID: <317FF665.41C67EA6@fore.com> I'd like to run a custom application, not the VxWorks shell, when a user telnets to my device. Unfortunately, there doesn't seem to be an API that provides this capability. Has anyone succeeded in doing this without having to modify the VxWorks source? Thanks, Jeff - ---------- Jeff Prem jdp@fore.com FORE Systems, Inc. --------------------------- End of New-News digest ********************** From Mike.Milde.at.SRC-Home_Office@scires.com Thu May 2 04:24:53 1996 From: Mike Milde at SRC-Home_Office Date: Thu May 2 04:24:56 PDT 1996 Subject: ENETUNREACH? I have a system where the shell occasionally spits out these messages that say "ENETUNREACH". Surprisingly, everything appears to work as it should. Any idea where these are created? I'm doing some IP socket stuff, so I assume that this means something like: Ethernet Unreachable. Any ideas? Anyone know if this is caused by an ICMP message coming back from the system I am communicating with? Any help/hints/clues are greatly appreciated! Mike Milde Scientific Research Corporation mmilde@scires.com From P.H.A.Mutsaers@cycl.phys.tue.nl Thu May 2 07:38:34 1996 From: Peter Mutsaers Date: Thu May 2 07:38:37 PDT 1996 We are using Tornado/VxWorks5.3 on a x86 target. Does anyone know how to access, in an application, a device on the PCI-bus since WRS does not provide a PCI-driver for this target. Does anyone know if there is a PCI-driver for an x86 target available from another source? Is it possible to access a PCI compatible BIOS and if yes, how? Greetings, Peter Mutsaers Eindhoven University of Technology tnndmuts@cycl.phys.tue.nl From P.H.A.Mutsaers@cycl.phys.tue.nl Thu May 2 07:39:58 1996 From: Peter Mutsaers Date: Thu May 2 07:40:01 PDT 1996 Subject: pci on x86 support We are using Tornado/VxWorks5.3 on a x86 target. Does anyone know how to access, in an application, a device on the PCI-bus since WRS does not provide a PCI-driver for this target. Does anyone know if there is a PCI-driver for an x86 target available from another source? Is it possible to access a PCI compatible BIOS and if yes, how? Greetings, Peter Mutsaers Eindhoven University of Technology tnndmuts@cycl.phys.tue.nl From bqv@eng102.eng.i-o.com Thu May 2 08:15:41 1996 From: "Bang Q. Vu" Date: Thu May 2 08:15:44 PDT 1996 Subject: Re: comp.os.vxworks newsdigest Joachim Schambach penned: > >Forgive my ignorance in this matter, but: how does one produce bootrom code for >VxWorks from the BSP in 5.2 that can be used by a PROM burner for MVME167 > proms? > >I found several targets relating to this subject in the Makefile. However, they >all seem to produce one file only, while the manual states that there are 2 > boot proms needed for VxWorks, arranged in 32 bits, one for even, and one for >odd words. I don't see a target that can split the resulting file in such a > manner. >Am I missing something fundamental here? >I would really appreciate if someone could mail me a step-by-step procedure to create boot proms, beginning with a pair of empty proms and the (non-compiled) >BSP, that is understandable by a layman like myself ... > >Thanks in advance, Jo >***************************************************************** >Joachim Schambach Research Associate >Physics Department, RLM 5.208 >The University of Texas at Austin >Austin, TX 78712 >USA > >email: JSchamba@physics.utexas.edu >Tel: x1 (512) 471-1303 FAX: x1 (512) 471-9637 >***************************************************************** > Once, i got tired of waiting for vxWorks to boot with the compressed boot ROMS so i built new boot ROMS with uncompressed codes. Go to where the BSP is, for example, vw/config/mv147, run "make bootrom_hex|bootrom_uncmp.hex". There are 2 PROMS. All of this is mentioned in the Programmer's Guide. The splitting of the boot file (i.e. bootrom_hex|bootrom_uncmp.hex) into even/odd bytes will have to be done by another piece of software not provided by WRS. The PROM burner (Unisite), that i used, is interfaced to a PC which does the splitting for me. Look at your PROM burner documentation to see if it provides the byte splitting for you. Sorry, can't help you here. -- Bang Vu | I/O, Inc. bvu@eng.i-o.com | 12300 Parc Crest Dr. Voice 713-879-2087 | Stafford, Texas 77477 Fax 713-879-2096 | From don@carts.com Thu May 2 15:31:57 1996 From: Don Bingman Date: Thu May 2 15:31:59 PDT 1996 Subject: consultants I have an imediate need for several vxworks consultants in the San diego area. Contact Don Bingman Voice: 619-587-0888 Fax: 619-587-0444 Email: don@carts.com From fchen@mpl.ucsd.edu Thu May 2 15:57:16 1996 From: fchen@mpl.ucsd.edu (Hsiufang Chen) Date: Thu May 2 15:57:20 PDT 1996 Subject: Re: Returned mail: Cannot send message for 3 days Hi: Your mail bounced because barnacle has been down or disconnected from network for several days. If you are trying to reach Bob Lawhead, his address is rml@mpl.ucsd.edu. Please resend your mail to this address. Otherwise, if you need my assistance to figure out a correct address, please let me know. Fang > From Mailer-Daemon Thu May 2 14:03:47 1996 > Date: Thu, 2 May 1996 14:03:47 -0700 > From: Mailer-Daemon (Mail Delivery Subsystem) > Subject: Returned mail: Cannot send message for 3 days > To: Postmaster > > ----- Transcript of session follows ----- > 421 barnacle: Host barnacle is down > > ----- Message header follows ----- > Return-Path: > Received: from csg.lbl.gov by chiton.ucsd.edu (5.x/SMI-SVR4-UCSD.1) > id AA16489; Mon, 29 Apr 1996 13:28:56 -0700 > Errors-To: vxwexplo-errs@lbl.gov > Received: by csg.lbl.gov (4.1/1.39) > id AA02825; Mon, 29 Apr 96 13:01:45 PDT > Date: Mon, 29 Apr 96 13:01:45 PDT > Message-Id: <9604292001.AA02825@csg.lbl.gov> > Errors-To: vxwexplo-errs@lbl.gov > To: vxworks_users@csg.lbl.gov > From: vxwexplo@lbl.gov (the vxWorks Users Group Exploder) > Subject: Re: once more plz: How do you time a piece of code? > From daemon@csg.lbl.gov Fri May 3 04:00:32 1996 From: daemon@csg.lbl.gov Date: Fri May 3 04:00:36 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Fri May 3 04:00:25 PDT 1996 Subject: New Development Tools and Embedded Database Subject: Equivalence to inittab needed for VxWorks Subject: Re: 162 flash memory Subject: Re: Rpc 4.0 and Microsoft Rpc ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: New Development Tools and Embedded Database Date: Fri, 19 Apr 96 14:14:35 From: Phillip Thoren Organization: Web Professionals' Network Message-ID: VisionSoft Corporation Welcomes You - to the http://www.synervision.com web site VisionSoft Development Products Cross-Platform Software Runtime Error Checking :& Coverage Analysis Tools - --- Automatically Detect: Variable Reference Errors, Array Boundary Condition Errors, Memory Leaks --- - --- Perform Integrated Single-Pass SoftLab(tm) Analysis for Runtime Error Checking & Coverage Analysis --- - --- Automatically Retrofit Your Application and Include Remote Software Monitoring and Runtime Diagnostics --- VisionSoft Embedded Realtime Database Products Fast, Efficient Portable Embedded C & JAVA Database API Toolkit - ------------------------------------------------------------------------------------------------------------------------------------------------ Please visit our web site http://www.synervision.com, browse our site and freely Roam!!!! Datasheets, White Papers, Manuals and Software in (HTML Format), (Acrobat Format) and (Self-Extracting Executable Trial Software) - ------------------------------------------------------------------------------------------------------------------------------------------------ - --- Cross-Platform Software Quality Tools, Internet Search, Find, Store and Replace Tools --- - --- HOST Platforms Supported Include:-- Windows 3.1, Windows 95, Windows/NT, HP-UX, VxWorks, Solaris, Macintosh, OS/2, SCO (Locus) --- - --- TARGET Platforms Supported Include:-- All Platforms Above, VMS, Ultrix, Tandem Guardian, MVS, XINU, etc. (Virtually ALL platforms) - --- Realtime TARGET Platforms Supported Include:-- VxWorks (All Processors), Will modify products as required to support other RTOS's and Non-OS type embedded environments VisionSoft Corporation Pruneyard Towers 1999 S. Bascom Ave, Suite 700 Campbell, CA 95008-9872, USA (408) 879-2672 sales@synervision.com http://www.synervision.com (c) copyright 1995-1996 VisionSoft Inc. All rights reserved. --------------------------- Newsgroups: comp.os.vxworks Subject: Equivalence to inittab needed for VxWorks Date: 2 May 1996 21:04:29 GMT From: townley@nortel.com (Jeanette Chickles) Organization: Bell-Northern Research Ltd. Message-ID: <4mb80t$3uk@crchh327.rich.bnr.ca> Hi, I was wondering if anyone out there has developed any software that's equivalent to inittab on HPUX for VxWorks? I'm interested in writing a process that monitors particular tasks running on VxWorks and restarts them if they shut down. I have some ideas on how to implement this myself, but before I do, I wanted to check to see if anyone knew if something like this is already available. Thanks in advance! Jeanette (townley@nortel.com) --------------------------- Newsgroups: comp.os.vxworks Subject: Re: 162 flash memory Date: 19 Apr 1996 17:16:57 GMT From: witts@PROBLEM_WITH_INEWS_GATEWAY_FILE (Steve Witt) Organization: Hughes Aircraft Company Message-ID: <4l8hq9$7ce@venice.sedd.trw.com> References: <199604151822.AA24425@darya.wrs.com> Reply-To: sawitt@hac2arpa.hac.com > Pete writes: > > > > I'm trying to program the MVME162's flash ram. The vxWorks documentation > > wimps out by refering the user to the Motorola's debug command 'pflash'. > > Anybody have any example code that they could share? > > > > Thanks, > > > > Pete > > > > faill@eng.insol.com An alternate method for downloading a FLASH image over Ethernet, which is much faster than the serial method - both are useful however. This may look pretty complicated, but it really isn't once you understand what is going on. 1) make your image (e.g. vxWorks.st_rom) 2) convert your image to binary format with aoutToBin % aoutToBin < vxWorks.st_rom > vxWorks.st_rom.bin 3) make sure you have the Motorola 'BUG' ROM in the PROM socket and boot from the ROM (pull jumper J22) 4) use NIOT command to specify IP addresses of target and host for image transfer 162-BUG> niot 162-BUG> ... {Read Motorola docs, only thing you have to enter are the IP address of your host and IP address of your target board} 5) use NIOP command to actually download the image over the Ethernet 162-BUG> niop 162-BUG> ... {Again, read the docs, you must enter the file name (vxWorks.st_rom.bin) and the start address in RAM that you want the load to begin at. We use 0x10008.} There is something a little tricky about this if you don't understand the TFTP protocol. This is true of SunOS and Solaris, I don't know about other OSs. TFTP is used to download the image from your host to the target. TFTP has no security mechanisms, like the password protect in FTP. So Sun (and probably others) have set up the TFTP daemon to only allow file access from a designated directory. This directory is designated by the 'inetd' config file, '/etc/inetd.conf'. Edit the line in /etc/inetd.conf for TFTP and replace the directory that is there with the one in which your image resides. Then you must restart 'inetd' (which controls the TFTP daemon) by sending it the hangup signal (kill -HUP pid, where pid is inetd's process ID retrieved using 'ps'). Read the man pages about all of these things if it isn't clear and my apologies for the seminar if this is old hat to you. 6) Ok, the image is in RAM, you must edit the first 8 bytes of the image to be programmed for some stuff needed at boot 162-BUG> mm 0x10000 00010000 0000? 0000 00010002 0000? 1000 00010004 0000? ff80 00010006 0000? 0008 00010008 46fc? . {beginning of your image, '.' stops the edit} 7) Now program flash 162-BUG> pflash 0x10000: 0xffa00000 ... prints a bunch of status, and when its done, you've got your image into FLASH, replace J22 to boot from FLASH and go for it. This image size parameter is the size of the chunk of RAM you want to program into FLASH. I use 0x80000 for bootrom images and 0x100000 for our application images, but this is of course dependent upon your image. Credits: 90% of the above was given to me by Jerry Pendleton of WRS Tech Support when we asked the same question several years ago. Hope this helps. What I've done now is to write an expect script (the TCL 'expect' utility) to do all of this almost automatically. After you've done it 5 or 10 times (not to mention hundreds of times), it loses its excitement. - --------------------------------------------------------------------- Steve Witt Hughes Aircraft Company sawitt@hac2arpa.hac.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Rpc 4.0 and Microsoft Rpc Date: Sat, 20 Apr 1996 08:09:54 GMT From: noblenet@world.std.com (NobleNet Inc.) Organization: NobleNet, Inc. Keywords: RPC, XDR, Client/Server, NobleNet, ONC, VxWorks Message-ID: References: <829821431.11023@hathsys.demon.co.uk> <829822771.12586@hathsys.demon.co.uk> Followup-To: poster In article <829822771.12586@hathsys.demon.co.uk>, Nicholas Doyle wrote: >I know there is an ONC implemention available, but it seems that good >old Microsofts MSVC 4.0 Rpc will not talk to it. Please contact NobleNet for ONC RPC tools and solutions for UNIX, Windows (3.1/95/NT), Macintosh, VMS, OS/2, NetWare, VxWorks, etc. NobleNet, Inc. 337 Turnpike Road Southboro, MA 01772 Voice +1 508 460 8222 Fax +1 508 460 3456 sales@noblenet.com --------------------------- End of New-News digest ********************** From senthil@ssi.navtech.com Fri May 3 07:39:12 1996 From: "Senthil Natesan" Date: Fri May 3 07:39:15 PDT 1996 Subject: Multibyte wide char support Hi Does any body know about multibyte or wide char support for vxworks. Have you used it? Do you know about any third party support for it? I heard vxworks does not support this. Any information is appreciated. Senthil Natesan e-mail: senthil@ssi.navtech.com From seaweed!seaweed.com!bob@netcom.com Sat May 4 19:19:43 1996 From: bob schulman Date: Sat May 4 19:19:45 PDT 1996 Subject: Re: Multibyte wide char support > Does any body know about multibyte or wide char support for vxworks. > Have you used it? Do you know about any third party support for it? > I heard vxworks does not support this. vxWorks does and doesn't support it. Some wide char functions (e.g. wcslen and wcstombs) are not supported. Some functions have entry points in the distributed vxWorks libraries (e.g. mblen and mbstowcs) but are documented us unimplemented. How a function can have been written by Wind River, distributed by Wind River, but then said to be unimplemented is bit beyond my meager mental facilities. The crowning glory of Wind River's mblen implementation (just to pick one gem of this thread) is that it returns a value of -1. The SunOS definition of mblen does not include -1 as a legal return value. I don't know, but I guess that no Unix vendor's implementation of mblen includes -1 as a legal return value. So, if you attempt to port a piece of software from, e.g. SunOS, to vxWorks, you don't have any problems resolving references to mblen, but your code may die in mysterious ways because your code did not anticipate the zen-like quality of Wind River's there-but-not-there-and-doesn't-return-legal-value mblen. By the way, did I tell you guys and gals that I've been burned by this? So what you heard is correct: Wind River does not support this. bob From ceselsa@coastalnet.com Mon May 6 06:19:18 1996 From: ceselsa@coastalnet.com (Carlos Martinez) Date: Mon May 6 06:19:21 PDT 1996 Subject: vxworks parallel driver I am looking for a VxWorks parallel driver (e.g. to use with a Laser printer). We are currently using target VMEbus Motorola PowerPC boards and host SUN SPARC Stations. Is there any merciful person on the net who could help ? Thank you. Carlos Martinez INDRA DTD Cherry Point, NC. ceselsa@coastalnet.com (919) 466-9133 (919) 466-9134 Fax From Mike.Milde.at.SRC-Home_Office@scires.com Mon May 6 07:37:01 1996 From: Mike Milde at SRC-Home_Office Date: Mon May 6 07:37:05 PDT 1996 Subject: Ruggedized VME boards? Can someone point me to a collection of information on ruggedized VME single board computers? We are currently in need of a ruggedized Motorola MVME-162 card or something as similar as possible. A WRS board support package must be available for it. There is no need for military standard components; commercial boards will be adequate. Anyone have any favorites? Thanks! Mike Milde Scientific Research Corporation (770) 989-9550 From 75104.3460@compuserve.com Mon May 6 21:38:04 1996 From: David Parkerson <75104.3460@compuserve.com> Date: Mon May 6 21:38:07 PDT 1996 Subject: Re: Ruggedized VME boards? Mike, A well supported architecture with VxWorks that is used in many Mil apps is SPARC. SPARC 5, 10 and 20's are available all with Vx. The 10/20 will run it on ea of four processors!! Also PowerPC is available with 100 Mhz, secondary cache and the Universe VME chip with Vx Support too! Call Themis at 510-252-0870. Regards David Parkerson From daemon@csg.lbl.gov Tue May 7 04:00:39 1996 From: daemon@csg.lbl.gov Date: Tue May 7 04:00:44 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Tue May 7 04:00:31 PDT 1996 Subject: Re: How to determine current memory usage Subject: Re: How to determine current memory usage Subject: Native Java VM or bytecode to 68k/i960 etc for embedded system Subject: Any HTTP servers available on VxWorks? ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: How to determine current memory usage Date: 4 May 1996 08:12:26 GMT From: romsky@x4u2.desy.de (Vladimir Romanovski) Organization: DESY Message-ID: <4mf3ha$n60@dscomsa.desy.de> References: Andy Van Pelt (avanpelt@qualcomm.com) wrote: : I would like to track memory usage from within a program, and have not : been able to find a routine to call to tell me this. memShow will display : the information I want, but I don't want the data displayed, I want to : store it and analyze it. Look this ftp://epics.aps.anl.gov:pub/epics/extensions/devAiStats.c I use long way(redirect output of memShow to socket and after _scanf_).. Vladimir --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How to determine current memory usage Date: 6 May 1996 19:03:49 GMT From: Don Small Organization: Sandia National Laboratories Message-ID: <4mliel$79b@news.sandia.gov> References: Andy Van Pelt wrote: > I would like to track memory usage from within a program, and have not >been able to find a routine to call to tell me this. memShow will display >the information I want, but I don't want the data displayed, I want to >store it and analyze it. > > How do I find out how much memory is currently allocated and how much is >free? I know how to find out the largest block that can be allocated >(memFindMax), but that's not quite the same. > > I hope the answer doesn't involve calling memShow and parsing >the output. That's gross! > The number of currently allocated "words" may be found by indexing into the system memory partition. To determine the number of bytes use: 2*memSysPartId->curWordsAllocated You must include to use this method. I would also assume that this method (since it is under the private directory) is not guaranteed by Wind River to work under any other version. I know it works with version 5.2 & 5.3. For other info about this structure view memPartLibP.h and look for the structure definition for PARTITION. Don --------------------------- Newsgroups: comp.lang.java,comp.os.vxworks,comp.realtime Subject: Native Java VM or bytecode to 68k/i960 etc for embedded system Date: Mon, 06 May 1996 14:29:32 -0700 From: Shaun Case Organization: Graham-Patten Systems, Inc. Message-ID: <318E6F3C.17F4@gpsys.com> I have been hunting around for several days for a Java development environment suitable for a realtime embedded system, preferably a 68k or i960-based platform with a well-supported realtime operating system. For performance reasons we'd like something that compiles java bytecode into native instructions. After extensive searching I have come up completely empty-handed. Can anyone provide me with pointers to something along these lines, or point me in the direction of someone who is working on something like this? Maybe I'm barking up the wrong tree, but this seems like a very reasonable thing to want to do. Thanks, Shaun --------------------------- Newsgroups: comp.os.vxworks Subject: Any HTTP servers available on VxWorks? Date: 29 Apr 1996 12:31:51 GMT From: markham@bnr.ca (Andrew Markham) Organization: Bell Northern Research Message-ID: <4m2crn$41c@nrtphba6.bnr.ca> My group has some interesting ideas for using HTTP in our RT system and are wondering if anyone has ported an HTTP server (preferably NCSA httpd) to VxWorks. Any help will be greatly appreciated! - -- Andy Markham Nortel, Inc. --------------------------- End of New-News digest ********************** From jlao@istnet.com Tue May 7 05:58:21 1996 From: jlao@istnet.com Date: Tue May 7 05:58:23 PDT 1996 Subject: Development Tools I am looking for development tool for VxWorks which the target board is MIPS. So far I have two candidates the StethoScope from Real-Time Innovations, Inc. and the Insight from the Parasoft. Is there anybody use these tools. Or is there other tools that provides better error catching capabilities. Thanks, James From aitech@rugged.com Tue May 7 10:05:19 1996 From: aitech@rugged.com (Aitech ) Date: Tue May 7 10:05:22 PDT 1996 Subject: Re: Ruggedized VME boards? > > > Can someone point me to a collection of information on ruggedized > VME single board computers? > > We are currently in need of a ruggedized Motorola MVME-162 card > or something as similar as possible. A WRS board support package > must be available for it. There is no need for military standard > components; commercial boards will be adequate. > I suggest that you take a look at the Aitech line of rugged VME boards. You can llok them up in http://www.rugged.com The C140 model is quite similar to the 162 you mentioned, and there is a VxWorks BSP available. Daniel Kord Aitech From fchen@mpl.ucsd.edu Tue May 7 19:29:08 1996 From: fchen@mpl.ucsd.edu (Hsiufang Chen) Date: Tue May 7 19:29:11 PDT 1996 Subject: Re: Returned mail: Cannot send message for 3 days Hi, Barnacle is back to the network now. Please resend your mail to him. Fang > From Mailer-Daemon Thu May 2 18:00:24 1996 > Received: by chiton.ucsd.edu (5.x/SMI-SVR4-UCSD.1) > id AA13331; Thu, 2 May 1996 18:00:24 -0700 > Date: Thu, 2 May 1996 18:00:24 -0700 > From: Mailer-Daemon (Mail Delivery Subsystem) > Subject: Returned mail: Cannot send message for 3 days > To: Postmaster > > ----- Transcript of session follows ----- > 421 barnacle: Host barnacle is down > > ----- Message header follows ----- > Return-Path: > Received: from csg.lbl.gov by chiton.ucsd.edu (5.x/SMI-SVR4-UCSD.1) > id AA21526; Mon, 29 Apr 1996 17:19:08 -0700 > Errors-To: vxwexplo-errs@lbl.gov > Received: by csg.lbl.gov (4.1/1.39) > id AA04312; Mon, 29 Apr 96 16:56:01 PDT > Date: Mon, 29 Apr 96 16:56:01 PDT > Message-Id: <9604292356.AA04312@csg.lbl.gov> > Errors-To: vxwexplo-errs@lbl.gov > To: vxworks_users@csg.lbl.gov > From: vxwexplo@lbl.gov (the vxWorks Users Group Exploder) > Subject: Re: once more plz: How do you time a piece of code? > From daemon@csg.lbl.gov Wed May 8 04:00:23 1996 From: daemon@csg.lbl.gov Date: Wed May 8 04:00:32 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Wed May 8 04:00:19 PDT 1996 Subject: VxMP Shared Memory Objects - Performance Subject: Re: Multibyte wide char support Subject: Minimum VxWorks Size? ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: VxMP Shared Memory Objects - Performance Date: 29 Apr 1996 22:33:15 GMT From: Jorge Rodriguez Organization: stp Message-ID: <4m3g3b$nb0@nrtphba6.bnr.ca> Good Day, Has anyone out there had any experience with the optional package VxMP -- Shared Memory Objects. I am interested in using it on a PPC architecture & am looking for some performance information on the package. Also any comments about the product are also appreciated. Thanks in advance, _________________________ | Jorge Rodriguez | | Bell Northern Research | /) (919) 991-4638 (\ / ) Email:Jorge@bnr.ca ( \ _( (|_________________________) ) /> (((\ \) /,) / ) / //))/ (\\\\ \_/ / \ \_/ ///// \ / \ / \ _/ \_ / - ----/ /----------------------\ \---- / / \ \ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Multibyte wide char support Date: 6 May 1996 22:15:58 -0700 From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star, Oakland, CA Message-ID: <4mmmae$r9r@idiom.com> References: <9605050206.AA29130@seaweed.com> In article <9605050206.AA29130@seaweed.com>, bob schulman wrote: > I don't know, but I guess >that no Unix vendor's implementation of mblen includes -1 as a legal return >value. The -1 is one of the correct return values for mblen(). It is an error indication when illegal character is passed. Whether WRS implements this correctly or not, I don't know. - -- H.J. Bae http://www.peacefulstar.oakland.ca.us --------------------------- Newsgroups: comp.os.vxworks Subject: Minimum VxWorks Size? Date: Tue, 07 May 1996 06:57:25 -0700 From: Todd Hoff Organization: Possibility Outpost Message-ID: <318F56C5.3FCB@possibility.com> What is the smallest size you have gotten a VxWorks image? There seems to be much confusion on this issue :-) We don't need IP, BOOTP, etc., keeping core services, and 100K seems to be the smallest it will go. This seems too large. - ------------------------------------------------------------- tmh@possibility.com | The loyalty of small men can be http://www.possibility.com | bought cheaply, for greed has no | pride. - Michael Kube-McDowell --------------------------- End of New-News digest ********************** From ztf@one.one.com Wed May 8 05:12:20 1996 From: ztf@one.com (Zachary T. Frey) Date: Wed May 8 05:12:24 PDT 1996 Subject: Re: Development Tools >I am looking for development tool for VxWorks which the target board is MIPS. >So far I have two candidates the StethoScope from Real-Time Innovations, Inc. >and the Insight from the Parasoft. Is there anybody use these tools. >Or is there other tools that provides better error catching capabilities. Does Insight now support VxWorks? I looked at them earlier this year, and the only real-time OS they said they supported was LynxOS. Could you post or email a contact for Real-Time Innovations about StethoScope? I'd be interested in learing more about them. Another test tool for VxWorks is C-Cover from Bullseye Testing Technology (info@bullseye.com, (800)-278-4268). C-Cover provides coverage analysis of code, just like the TCA (Total Coverage Analysis) module of Insight. Thanks, Zach -- Zach Frey Open Networks Engineering, Inc. (ONE) Voice: (313) 996-9900 x 1227 777 E. Eisenhower Parkway, Suite 650 Fax: (313) 996-9908 Ann Arbor, MI 48108 Internet: ztf@one.com From andyg@emass.com Wed May 8 06:49:33 1996 From: Andy Gross Date: Wed May 8 06:49:36 PDT 1996 Subject: Re: comp.os.vxworks newsdigest Todd Hoff wrote: > Subject: Minimum VxWorks Size? > > What is the smallest size you have gotten > a VxWorks image? There seems to be much confusion > on this issue :-) We don't need IP, BOOTP, etc., > keeping core services, and 100K seems to be the > smallest it will go. This seems too large. > I couldn't even get it that small. I think the smallest I got was 160K, and I tried to strip it all out. I'm sure my problem has to do with being on an i960 and having the gnu libraries linked in. If you have any call into the library it seems to cascade into a huge linking. Oh, I'm using Tornado (VxWorks 5.3). I figure when we move out of debug into production I'll be able to reduce it some: I'd be happy (and surprised) if I could get it under 100K. BTW, Wind River quotes a minimum size of 40K. I'm wondering what processor that was on and if there's any debug support in it (target agent?). Andy -- Andy Gross EMASS Senior Software Engineer 10949 E. Peakview Ave. E-Mail: andyg@emass.com Englewood, CO 80111 (303) 705-3762 From seaweed!seaweed.com!bob@netcom.com Wed May 8 07:44:01 1996 From: bob schulman Date: Wed May 8 07:44:04 PDT 1996 Subject: Re: Multibyte wide char support > The -1 is one of the correct return values for mblen(). > It is an error indication when illegal character is passed. > Whether WRS implements this correctly or not, I don't know. My SunOS man page for mblen says that -1 is returned "if they [the passed characters of the string] do not form a valid multi-byte character". A string of a simple ASCII character, especially in the "C" locale, *does* constitute a sequence of valid multi-byte characters. So I am still befuddled how simply returning -1 can be a reasonable implementation of mblen(). In any event, calling mblen() with a NULL string should never return -1 (which is what Wind River's implementation illegally does). bob From jlao@istnet.com Wed May 8 11:27:22 1996 From: jlao@istnet.com Date: Wed May 8 11:27:26 PDT 1996 Subject: Re[2]: Development Tools Zach, >Could you post or email a contact for Real-Time Innovations about >StethoScope? I'd be interested in learing more about them. The Real-Time Innovations address is 954 Aster, Sunnyvale CA 94086 tel (408)720-8312 Fax (408)720-8419. The stethoScope is a collection of utilities and tools that it easier to use and debug you VxWorks application. This is what it says on the brochure. This is similar to windview basically-I think. Thanks for the info on Bullseve. Best regards, James From stan@rti.com Wed May 8 11:45:08 1996 From: Stan Schneider Date: Wed May 8 11:45:12 PDT 1996 Subject: Re: Development Tools Hi, >> Submitted-by ztf@one.one.com Wed May 8 05:12:20 1996 >> Submitted-by: ztf@one.com (Zachary T. Frey) >> >> Could you post or email a contact for Real-Time Innovations about >> StethoScope? I'd be interested in learing more about them. >> That would be us! We market many tools for VxWorks, including: StethoScope: Real-time graphical monitoring and data acquisition Use StethoScope to watch any of your program variables evolve in real time; any value in memory can be monitored. StethoScope opens a window into your application and shows you what's really happening. It collects and saves data, exports to analysis tools (e.g. Matlab), etc. ScopeProfile: Real-time dynamic execution profiler ScopeProfile shows you exactly where you're spending your CPU cycles. It gives you an in-depth hierarchical analysis of each subroutine's CPU utilization. RTILib: A collection of focused utilities and debugging tools RTILib tools include memory analysis and leak detection, a re-entrant shell, execution tracing, fast buffer management, and more. New for tornado (available soon) is a "networking" package that works through the net or the target server, target-based remote symbol-table access and dynamic loading, etc. [ScopeProfile & RTILib are included free with StethoScope.] ControlShell: The component-based real-time programming system ControlShell is a graphical software composition environment for complex real-time applications, the result of years of joint research with Stanford University. ControlShell users build system from small reusable objects called "components." It especially targets the electro-mechanical systems market. NDDS: The Network Data Delivery Service; an advanced dissemination architecture for real-time networking. NDDS lets many diverse nodes communicate easily with no network programming. NDDS makes it easy to develop complex communication and data interactions. And NDDS implements a true real-time model, so it won't impact real-time performance of its client nodes. It delivers the needed information, on time, with minimal programmer effort. For more information, contact "info@rti.com," check out our web page at "http://www.rti.com," or call (408) 720-8312. -- Stan ============================================================================= = = = = Stan Schneider = email: stan@rti.com = = Real-Time Innovations, Inc. = Phone: (408) 720-8312 = = 155A Moffett Park Drive, Suite 111 = Fax: (408) 734-5009 = = Sunnyvale, CA 94089 = http://www.rti.com = = = = ============================================================================= From ceselsa@coastalnet.com Wed May 8 12:27:56 1996 From: ceselsa@coastalnet.com (Carlos Martinez) Date: Wed May 8 12:27:59 PDT 1996 Subject: VxWorks V5.3Beta & FORTRAN Is anybody out there using VxWorks V5.3 Beta with a Fortran compiler in a VME PowerPC Board? I would like to swap info about that. Carlos Martinez, ceselsa@coastalnet.com (919) 466-9133 Cherry Point, NC. From stan@rti.com Wed May 8 12:34:35 1996 From: Stan Schneider Date: Wed May 8 12:34:39 PDT 1996 Subject: Re: Re[2]: Development Tools >> Submitted-by jlao@istnet.com Wed May 8 11:27:22 1996 >> Submitted-by: jlao@istnet.com >> >> >> Zach, >> >> >Could you post or email a contact for Real-Time Innovations about >> >StethoScope? I'd be interested in learing more about them. >> >> The Real-Time Innovations address is 954 Aster, Sunnyvale CA 94086 >> tel (408)720-8312 Fax (408)720-8419. This is an old address, see below... >> The stethoScope is a collection of utilities and tools that it >> easier to use and debug you VxWorks application. This is what >> it says on the brochure. This is similar >> to windview basically-I think. Argggh... StethoScope is not "similar" to WindView. We get this a lot; I've appended our standard comparison... -- Stan ============================================================================= = = = = Stan Schneider = email: stan@rti.com = = Real-Time Innovations, Inc. = Phone: (408) 720-8312 = = 155A Moffett Park Drive, Suite 111 = Fax: (408) 734-5009 = = Sunnyvale, CA 94089 = http://www.rti.com = = = = ============================================================================= >> What is the difference between StethoScope and WindView? >> There is a superficial similarity---both graphically represent information gathered from an executing VxWorks target---but there's very little real functional overlap. There are three "types" of visualization/analysis tools for VxWorks: a source-level debugger (vxgdb/CrossWind), a real-time data monitor (StethoScope), and a operating system visualization tool (WindView). A debugger can be thought of as similar to a voltmeter. It measures static values. It's nearly useless for finding peak values, occasional "glitches", or time-dependent quantities like durations of events or noise characteristics. StethoScope is like a digital oscilloscope; it displays all these with ease. WindView is like a logic analyzer. It doesn't show analog values (program variables), but it excels at displaying events, such as task switches, semaphore activity, interrupts, etc. In particular, StethoScope is a window into your *application*. You can view any variable or memory location in your system. For example, you can see how much overshoot your controller has, measure how long it took a door to close, analyze noise in a sensor, or see how long a data queue is getting. StethoScope will graphically display your data, collect it for analysis, and save it to disk. StethoScope helps you understand what your program is doing. WindView, on the other hand, allows you to visualize the relationships and timing between operating system events. For example, you can measure how long it took for an interrupt to execute, how much later that interrupt was serviced, etc. It's also useful for watching and understanding task switching activity: which tasks are running, why tasks are preempted, how long the task switch took, etc. Issues like semaphore priority inversion, lock-outs, and interrupt latency are easily revealed. So, StethoScope shows you what your VxWorks application is doing, and WindView shows you how VxWorks is running your application. There's a big difference. There's almost nothing WindView can show that could be viewed with StethoScope and vice-versa. On another note, StethoScope also includes a dynamic execution profiler known as ScopeProfile. ScopeProfile displays which routines in your system are using the CPU, as a percentage of total CPU time. It's very useful for tuning your application code for maximum performance. WindView is also useful for this purpose, but it shows entirely different information. The tools are quite complementary. For example, if your problem is that your application is thrashing (spending most of its time context switching), ScopeProfile will simply report that some large percentage of the CPU is being spent in the kernel. It can show you which kernel routines are active, but that may not be of much help. WindView will show you which tasks are switching, and display the events (e.g. interrupts) that are causing the switches. On the other hand, if the problem is that you have one task with inefficient code, WindView will only tell you that one task is running most of the time. ScopeProfile will break down the individual routines within the task that are burning the CPU in great detail. It will give you a direct map of what that task is doing, what routines are being called, what routines they call, and point out exactly where the inefficiencies are. ScopeProfile provides a detailed function-by-function analysis. So, both WindView and ScopeProfile will help tune the performance of your code. However, they provide different perspectives: WindView offers precise measurements of event timing, ScopeProfile provides a detailed analysis of where your CPU time is being used. From hjb@peacefulstar.oakland.ca.us Wed May 8 13:18:01 1996 From: "H.J. Bae" Date: Wed May 8 13:18:04 PDT 1996 Subject: Re: Multibyte wide char support > > > The -1 is one of the correct return values for mblen(). > > It is an error indication when illegal character is passed. > > Whether WRS implements this correctly or not, I don't know. > My SunOS man page for mblen says that -1 is returned "if they [the > passed characters of the string] do not form a valid multi-byte character". Exactly. I'm just pointing out that -1 is a possible and valid return code for this function. Your original message implied it isn't. Maybe it's just misunderstanding. > So I am still befuddled how simply returning -1 can be a reasonable > implementation of mblen(). All the newer library stuff, like SVID->ANSI multibyte stuff, is probably from GNU or BSD libc. I don't know exactly which version of mblen() (or mbtowc() which is really what's doing the work) WRS has ported. However, WRS would not have written it from scratch, so it's not entirely their fault. If anything, they can be accused of not testing it thoroughly. Also, WRS implementation (at least in 5.3 I have here) of mblen() does not simply return -1. If you disassemble the code you will see that it actually does something, and have other return values than the -1. However, it certainly looks buggy to me. > In any event, calling mblen() with a NULL string should never return > -1 (which is what Wind River's implementation illegally does). Agreed. GNU version (1.09 libc) correctly returns 0, as per ANSI 4.10.7.2. Perhaps time to upgrade the library. -- H.J. Bae http://www.peacefulstar.oakland.ca.us From grd@telco-nac.com Wed May 8 17:51:29 1996 From: "Gautam R. Desai" Date: Wed May 8 17:51:32 PDT 1996 Subject: i960 coff to binary I am trying to convert i960 coff file to binary. Does anyone know how to. I tried objcopy but it complains about invalid bfd. ---------------------------------------------------------------------- Gautam R. Desai (510) 624-5676 TELCO Systems, Inc. email: grd@nad.telco.com ---------------------------------------------------------------------- From getker@cellstream.com Thu May 9 02:21:27 1996 From: getker@cellstream.com (Jim Getker) Date: Thu May 9 02:21:30 PDT 1996 Subject: Re: i960 coff to binary > I am trying to convert i960 coff file to binary. Does anyone know how > to. I tried objcopy but it complains about invalid bfd. > Have you tried $(VX_HSP_BASE)/bin/$(VX_HOST_TYPE)/coffToBin? I'm using vxWorks 5.2 under sun4 for the i960 and it seems to work OK. Jim Getker CellStream Networks Inc. getker@cellstream.com From daemon@csg.lbl.gov Thu May 9 04:00:41 1996 From: daemon@csg.lbl.gov Date: Thu May 9 04:00:45 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Thu May 9 04:00:34 PDT 1996 Subject: Re: TCP-IP Accelerator ??? Subject: Suggestions for R6000 dev environment Subject: Re: Shared memory driver to HP-UX card interface Subject: Problems with fopen and NFS Subject: How many tasks have you run with? Subject: Re: Custom application under telnet Subject: Re: [Q]: fopen problem Subject: Re: exception messages Subject: Re: Changing default NFS perms. on create ------------------------------------------------------- Newsgroups: comp.realtime,comp.os.qnx,comp.os.vxworks,comp.protocols.tcp-ip,comp.protocols.tcp-ip.domains,vmsnet.networks.tcp-ip.misc,vmsnet.networks.tcp-ip.multinet,de.comm.protocols.tcp-ip,comp.arch.embedded Subject: Re: TCP-IP Accelerator ??? Date: Wed, 1 May 1996 14:15:38 GMT From: hmo@sep.hamburg.com (Helge Oldach) Organization: Somebody Else's Problem Keywords: TCP-IP Accelerator Card Message-ID: References: Sender: hmo@sep.hamburg.com (Helge Oldach) peeds@netcom.com (LataAndPd) writes: | Does anyone know of any add-on accelerator cards for | TCP/IP which will offload the system's TCP/IP load ? Sure. There is a channel adapter board available from Cisco which connects a high-end router directly to the mainframe's channel. With newer software releases, this allows to offload TCP/IP protocol from NCP but runs that on the CIP board. | - what are the typical issues in using such a card in your system ? The main issue typically is that TCP/IP on the mainframe is pretty expensive (both in term of CPU cycles and software costs) compared to Cisco's solution. (To add another slightly off-topic interpretation of the question... :-)) Regards, Helge - -- Bumper sticker seen on the Data SuperHighway: Welcome to the Internet, now go home! --------------------------- Newsgroups: comp.arch.embedded,comp.realtime,comp.os.vxworks Subject: Suggestions for R6000 dev environment Date: Tue, 07 May 1996 19:19:24 GMT From: bonnett@sunland.gsfc.nasa.gov (Joe Vitale) Organization: NASA Goddard Space Flight Center -- Greenbelt, Maryland USA Message-ID: <4mo7ns$l94@post.gsfc.nasa.gov> Reply-To: bonnett@sunland.gsfc.nasa.gov Hello, In the very near future we are going to be starting a new development effort for an embedded system using an R6000 processor. We are currently in the process of setting up our development lab and I would like to try and get some advice and insight from others who have done similar work. Here we go... Our target hardware is an R6000 with a PCI bus interface. Since it will be some time before the target hardware is available, we are planning to purchase an Ultra PowerPC 603 (or 604) motherboard (with PCI connectors) to allow us to being software development while waiting for the target hardware to arrive. We are planning on using VxWorks from Wind River for our operating system. Since the new Tornado environment will not support our hardware in time to meet our schedule, we are planning on using Version 5.2 of VxWorks. We will buy VxWorks for the Ultra PowerPC board and also purchase an upgrade to the R6000 for use when our target hardware becomes available. We are planning on using the Green Hills software development tools hosted on a unix workstation (probably a sun solaris or dec alpha machine). The interface between the GH host and the Ultra PowerPC will be over an ethernet network while the interface to the target hardware will be over a serial interface. Can anybody highlight any potential gotchas with this proposed environment? It appears that at the "high level" this approach will work as intended. The problem is that at this stage it is hard to see all the details that may come up to bite us. Is anybody familiar with the RISCWatch interface supported by the R6000 and PowerPC architectures? According to the Green Hills people, their tools are a superset of the features available with just RISCWatch. Can anyone confirm or deny this? Exactly what type of capability is provided by this interface for performing hardware/software debug? I have about a hundred more questions slowly coming to the surface, but I think I'll stop here and see what kind of responses I get. :) Thanks for any help/advice/suggestions. Joe ================================================================= Joe Vitale, NASA Goddard Space Flight Center, the Hammers Company Check out "OOG, The Object Orientation Game" by MCM Productions A Windows polyform puzzle player - http://www.mcmprod.com/ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Shared memory driver to HP-UX card interface Date: 30 Apr 1996 23:14:31 -0700 From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star, Oakland, CA Message-ID: <4m6vg7$46b@idiom.com> References: <3186ABE3.3ED3@btg.com> In article <3186ABE3.3ED3@btg.com>, Mike Campbell wrote: >Got a 743i card running HP-UX 9.05 that needs to talk to a >vxWorks (5.2?) card over VME. Has anyone implemented an interface >to vxWorks backplane driver for this setup? does vxWorks even >publish any docs on how to do this? all help much appreciated.... Usually this requires bp driver source code. Long ago, at WRS, I wrote a version of bp driver for STREAMS TCP/IP in SVR3.2 Unix running on MVME147. It was a complete rewrite. For a BSD-derived networking kernel, such as HPUX, porting existing bp driver is easier. Last time I ported the "tunnel" driver to HPUX kernel for WRS, it took a couple of days work. You just have to be careful about subtle differences in mbuf and IP interface. - -- H.J. Bae http://www.peacefulstar.oakland.ca.us --------------------------- Newsgroups: comp.os.vxworks Subject: Problems with fopen and NFS Date: Wed, 8 May 1996 16:04:06 GMT From: Tom Fuda Organization: Northrop Grumman Norden Systems Message-ID: Sender: usenet@tron.bwi.wec.com (USEnet News Poster) I am developing a system which uses multiple 486 single board computers. The processors communicate via Ethernet. One board posseses an IDE drive on which the DOS filesystem has been mounted and exported via a call to nfsExport. The other board does an nfsMount to obtain access to the filesystem. I am attempting to open a file remotely from the second board using fopen(filename, "a") and write a text string to the file. I am finding that the "a" (append) access modifier does not appear to work, i.e. the existing file is overwritten, rather than appended to. I am also seeing that if an invalid path/file name is specified fopen does not return a null file pointer as it should. Has anybody else seen these problems? WRS tech support has not given me a satisfactory answer yet. Thanks in advance. Thomas Fuda Northrop Grumman Norden Systems fuda@norden.com --------------------------- Newsgroups: comp.os.vxworks Subject: How many tasks have you run with? Date: 8 May 1996 17:37:26 GMT From: otteson@b4pph10b.bnr.ca (John Otteson) Organization: Bell-Northern Research, Ottawa, Canada Message-ID: <4mqm4m$fl6@nrtphba6.bnr.ca> Has anybody experience with running over a thousand tasks on VxWorks? Have you run into a practical upper limit? 2000? 3000? 4000? Obviously you'll run out of memory and real time at some point. I am wondering if there are other more subtle problems you encounter... Or does VxWorks just keep humming right along? Thank you, - ---John otteson@nortel.ca --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Custom application under telnet Date: 30 Apr 1996 23:46:04 -0700 From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star, Oakland, CA Message-ID: <4m71bc$514@idiom.com> References: <317FF665.41C67EA6@fore.com> In article <317FF665.41C67EA6@fore.com>, Jeffrey D. Prem wrote: >I'd like to run a custom application, not the VxWorks shell, when a >user telnets to my device. Unfortunately, there doesn't seem to be an >API that provides this capability. Has anyone succeeded in doing this >without having to modify the VxWorks source? > VxWorks telnet doesn't provide API that provides such capability. It's sort of hard-coded in the daemon to start the VxWorks shell. What you can do, as an alternative, is to write a custom socket program that listens/accepts on a different port. When someone connects to you, you can then run your program of choice. Remember, most telnet clients take extra argument (port #). You can, for example, telnet to SMTP port and talk to sendmail daemon directly to send email to someone, if you don't have a Mail program. Just an idea. - -- H.J. Bae http://www.peacefulstar.oakland.ca.us --------------------------- Newsgroups: comp.os.vxworks Subject: Re: [Q]: fopen problem Date: 1 May 1996 12:20:30 -0700 From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star, Oakland, CA Message-ID: <4m8dhu$qf5@idiom.com> References: In my opinion, this is a bug. It is a matter of consistency in the interface. Lacking that, a detailed documentation that informs the users must be provided. It's not entirely fair to characterize this as "pilot error." I suspect that the right interface, consistent with how FTP netDrv works, as well as how other file I/O (dos, rt11, etc.) works is to implement rsh dependent part of netDrv to use "cat >!" rather than "cat >". There is no compelling reason not to, and it would have prevented inconsistent behavior (rsh file I/O vs. all others). Not to mention confusion among users. It is an unexpected side-effect of using rsh in netDrv that was originally not anticipated. If the original implementor was aware of this, "cat >!" would have been used. Ask Bob Cohen if he knew this when he wrote it. I doubt he did. - -- H.J. Bae http://www.peacefulstar.oakland.ca.us --------------------------- Newsgroups: comp.os.vxworks Subject: Re: exception messages Date: 1 May 1996 12:49:56 -0700 From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star, Oakland, CA Message-ID: <4m8f94$rdv@idiom.com> References: <199604261615.JAA05318@dvs18.hdvs.com> In article <199604261615.JAA05318@dvs18.hdvs.com>, Wim Dewilder wrote: > >on 68K type boards, the kernel can store exception messages starting at >address 0x800 upto address 0x900 >Like once we got a "Kernel Panic" message in this location. >It seems the kernel writes an ASCII string in these memory locations. In configAll.h (in config/all directory), is definition for this offset 0x800 called EXC_MSG_OFFSET. It is located after the bootline storage area at offset 0x700. Kernel will write critical error messages in that area when serious errors (i.e. work Q overflow) happens. > >Is there any information available on which messages could be stored there, >and how they are stored. You can just sprintf() into the memory. - -- H.J. Bae http://www.peacefulstar.oakland.ca.us --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Changing default NFS perms. on create Date: 1 May 1996 12:57:23 -0700 From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star, Oakland, CA Message-ID: <4m8fn3$rl1@idiom.com> References: <1996Apr30.100900.24586@hrbicf> In article <1996Apr30.100900.24586@hrbicf>, Michael E. Warren wrote: > >Is there a convenient was to change the default permissions used when >a file is created on an NFS file system? When using the copy command, >new files are created on the file server using the 644 permissions. >Is there anything like the csh umask? > The default mask was a hard-coded value in old VxWorks. This might have changed since 5.0 but I doubt it. No umask support as far as I know. - -- H.J. Bae http://www.peacefulstar.oakland.ca.us --------------------------- End of New-News digest ********************** From gordon@tdf01.gsfc.nasa.gov Thu May 9 05:17:06 1996 From: Gordon Miller Date: Thu May 9 05:17:09 PDT 1996 Subject: Re: Shared memory driver to HP-UX card interface Mike Campbell wrote: >Got a 743i card running HP-UX 9.05 that needs to talk to a >vxWorks (5.2?) card over VME. Has anyone implemented an interface >to vxWorks backplane driver for this setup? does vxWorks even >publish any docs on how to do this? all help much appreciated.... You might want to talk with HP on this. They have a product for the HP 700's called VME Backplane Networking. While it is meant to be used HP-UX or HP-RT (derived from Lynx OS, a real time UNIX) backplane master and HP-RT slaves, there is a possiblity that it could be used in your setup. The latest manual I have on it (10/95, software release 2.02) has (in addition to many others) a Wind River Systems copyright. I have not tried a VxWorks slave in my HP 748's. I have used multiple HP 747's running HP-UX and communicating across the backplane without too many problems. If you have any questions let me know. -- Gordon Miller Integral Systems, Inc. gordon@integ.com 1100 West Street, CSC/LR (301) 497-2416 Laurel, MD 20707 Don't be afraid to take a big step. You can't cross a chasm in two small jumps. -David Lloyd George From goodrich@noao.edu Thu May 9 10:34:55 1996 From: goodrich@noao.edu (Bret Goodrich) Date: Thu May 9 10:34:58 PDT 1996 Subject: POSIX threads Hi All... I am looking for a port of POSIX thread functions to VxWorks 5.2. Does anyone have a reference for such a library? Thanks, bret -------------------- Bret Goodrich, Gemini 8 Meter Telescopes Project, Tucson, Arizona, USA Internet: goodrich@gemini.edu SPAN/HEPNET: noao::goodrich Phonenet: 1.520.318.8522 Snailnet: PO Box 26732, Tucson, AZ 85726 From banta@cebaf.gov Thu May 9 11:50:02 1996 From: Paul Banta Date: Thu May 9 11:50:07 PDT 1996 Subject: this may be crazy, but... I have a colleague who wants to ingore bus errors. He has a couple of cards out of a bunch that get too hot and stop responding to bus accesses. He's attempted to cool the crate enough to operate, but, nevertheless, wants the software to be able to recover from a bus error should one occur. Being the resident VxWorks person he asked me to help him out. Hence, the hack below. This is a little bit of code I put together to test catching SIGSEGV and SIGBUS signals. The signal catching part works, but the task that generates the bus error still gets suspended. I've browsed through the online docs (VxWorks 5.2) looking at sigLib. I've tried using setjmp & longjmp, but I think I'm misusing them. From the sigLib docs I got the impression that I could prevent the task that generates the exception from being suspended. Perhaps I just misunderstood? Any insights? Thanks, Paul /*********************************************** * File: sigtest.c * Blame: Paul Banta banta@cebaf.gov * Date: 05-09-96 * Note: * * Modification History: * 05-09-96 -- original version * * ***********************************************/ #include #include #include #include /* external functions */ extern sysClkRateGet(); /* global variables */ static int bus_error_denial_count; static int segv_denial_count; static jmp_buf task_environment; /* function to catch signals */ mySigHandler(signo, code, pSigContext) int signo; int code; struct sigcontext *pSigContext; { /* recognizing denial is the first step toward recovery */ switch(signo) { case SIGSEGV: segv_denial_count++; if (!(segv_denial_count % 5)) /* print warning every 5 signals */ { logMsg("denying %d segmentation faults is dangerous\n",segv_denial_count); } break; case SIGBUS: bus_error_denial_count; if (!(bus_error_denial_count % 5)) /* print warning every 5 signals */ { logMsg("denying %d bus errors is dangerous\n",bus_error_denial_count); } break; default: logMsg("this should never print\n"); break; } longjmp(task_environment, 0); } /* lets make some bus errors */ void myBusErrorGenerator() { unsigned int *ptr; int data; int status; ptr = (unsigned int *)0xffff5000; /* use this address in short i/o */ signal(SIGSEGV, (void *)mySigHandler); signal(SIGBUS, (void *)mySigHandler); FOREVER { taskDelay(sysClkRateGet()); data = *ptr; /* this is going to crash */ setjmp(task_environment); /* I don't think this is correct */ } } /* initialize this system */ int mySigTestInit() { int status; status = taskSpawn("tBusErrorGen", 100, VX_STDIO, 1024, (void *)myBusErrorGenerator, 0,0,0,0,0,0,0,0,0,0); } /* shutdown bus error generator */ mySigTestShutdown() { int status; status = taskDelete (taskNameToId("tBusErrorGen")); } From lepage@ftms.com Thu May 9 12:31:30 1996 From: Rob Lepage Date: Thu May 9 12:31:33 PDT 1996 Subject: VME device access I am having difficulty accessing a set of VME devices from an MVME1603 running VxWorks 5.2. These devices are mapped in the VME 32 bit address space, at a starting address of either 0x20000000 or 0x24000000. The actual number of addresses used can range from 2 Meg up to 64 Meg. To access these devices, I would like to map the A32 VME address range 0x20000000-0x27ffffff into the 1603 address range 0xd0000000-0xd7ffffff. From looking at information in mv1600.h, I determined that by default, 64 Meg of the A32 VME space can be accessed by the 1603, starting at an address of 0xd0000000. I tried increasing the amount of accessible A32 VME space from 64 Meg to 128 Meg by redefining VME_WINDOW_A32_SIZE, but could not get the modified kernel to boot. I also tried using two 64 Meg entries instead of one 128 Meg entry in the sysPhysMemDesc table, but still could not get the kernel to boot. What else needs to be done to increase the accessible address space? Second, how can I change the accessible VME address range from the default offset of 0x0 to the desired offset of 0x20000000? Any help, including pointers to what sections of the FM to R, will be greatly appreciated. Rob Lepage lepage@ftms.com From pata@aisg.com Thu May 9 12:58:33 1996 From: pata@aisg.com (Pat Abruzzise) Date: Thu May 9 12:59:16 PDT 1996 Subject: SW Engineering Openings I have an urgent need for SW Engineers using Vx Works for a 5 month contract on the west coast of Florida. Please contact Pat Abruzzise AISG 1 800 780-2598 or personal pager 1 800 570-0054 From YF.Siu@eng.efi.com Thu May 9 14:56:49 1996 From: "Yuet Fung Siu" Date: Thu May 9 14:56:51 PDT 1996 Subject: Re: this may be crazy, but... I think setjmp was misused, it should be something like: if (setjmp (task_environment) == 0) { /* do your stuff here */ } else { /* from longjmp() in exception handler */ } You should not give longjmp a return value of 0, however, longjmp will convert it to a 1 in this case. See the man on setjmp and longjmp. Good Luck, YF From jromine@phbtsus.com Thu May 9 16:22:27 1996 From: JEFF ROMINE Date: Thu May 9 16:22:30 PDT 1996 Subject: Re: POSIX threads > I am looking for a port of POSIX thread functions to VxWorks 5.2. > Does anyone have a reference for such a library? I am working w/ a company which has recently done a port of POSIX threads to VxWorks. They may be able to help you out. They are: Avalon Consulting Denver, CO (303) 293-9331 Talk to or send email to either: Joy Foglesong (joy@ava.com) or Carole Hogan (carole@ava.com) -Jeff Romine (Philips Broadcast Television Systems) From emwlbi@mo.emw.ericsson.se Fri May 10 01:36:01 1996 From: emwlbi@emw.ericsson.se (FY/DX Lennart Bie) Date: Fri May 10 01:36:06 PDT 1996 Subject: Re: VME device access On Thu, 9 May 96 12:31 Rob Lepage wrote: >I am having difficulty accessing a set of VME devices from an MVME1603 running >VxWorks 5.2. >>From looking at information in mv1600.h, I determined that by default, >64 Meg of the A32 VME space can be accessed by the 1603, starting at an >address of 0xd0000000. I tried increasing the amount of accessible >A32 VME space from 64 Meg to 128 Meg by redefining VME_WINDOW_A32_SIZE, >but could not get the modified kernel to boot. I also tried using two >64 Meg entries instead of one 128 Meg entry in the sysPhysMemDesc table, >but still could not get the kernel to boot. What else needs to be done >to increase the accessible address space? I myself haven't had any problems getting the kernel to boot after I changed the sysPhysMemDesc table. You get problems if you open a too big window (256M or so) that will collide with other sysPhysMemDesc table entries. >Second, how can I change the accessible VME address range from the default >offset of 0x0 to the desired offset of 0x20000000? Se below. >Any help, including pointers to what sections of the FM to R, will be >greatly appreciated. > >Rob Lepage >lepage@ftms.com > I have had severe problems with VME addressing with MVME1604 in 5.2 FCS. Note that the A32 window changes between 5.2 beta and the FCS. Some of the address space became corrupted, especially when a bigger VME window was opened in the sysPhysMemDesc table. The problem showed up on addresses 16M above where I opened my window. With the original code (FCS) the A32 space starts at 0xd8000000 (but sysBusToLocalAdrs returns 0xd0000000!). - The processor adrs 0xd8000000 yields the VME adrs of 0x08000000 and it looks as you get monotonic increasing addresses as you increase you processor address for 16M. But.. - The processor adrs 0xd9000000 yields the VME adrs of 0x091ff000 up to 0xd9f00000 where it starts to be o.k. again (I think, haven't tested all addresses). At 0xda000000 it starts to fail again and so on. The corrupt addresses have 0x001ff000 or'ed into the VME address. Such a faulty behaviour can also be found with the other addressing modes. The problem seems to lay not within the vxWorks kernel but rather inside the romInit.s routine in the bootrom! 5.3 solves the problem but only if booted with a 5.3 boot routine (bootrom or boot.bin). Note that the vxWorks kernel isn't setting up all related chips from scratch even if it looks that way (the MMU, the BATS, the MPC 105, the Vme2PCI and the VMEChip2 chips), rather it is dependent on the banks being properly set up by either romInit.s or the PPC-Bug (which probably does a better job at the time). > Second, how can I change the accessible VME address range from the default > offset of 0x0 to the desired offset of 0x20000000? This is the way addresses gets translated (I think) The processor address 0xd0000000 corresponds to the PCI address 0x10000000 which corresponds to the 040 bus address (between Vme2PCI and the VMEChip2) of 0x00000000 and the VME address of 0x00000000. Normally the Vme2PCI applies an offset (0xffffffff - 0x10000000) which decreases the address by 0x1000000. By manipulating this offset, PCI2VME_SLV2_OFFSE in mv1600.h, you could get the desired VME starting address provided you stay within the window opened by the VMEChip2 chip (0..0x7ffffff if I remember right) Lennart Bie emwlbi@emw.ericsson.se From kindl@dreisoft.de Fri May 10 05:05:12 1996 From: Wolfgang Kindl Date: Fri May 10 05:05:22 PDT 1996 Subject: CAN driver I am looking for a driver for intel 82527. Does anyone have a (basic) CAN driver? Thanks in advance. ------------------------------------------------------------ | Wolfgang Kindl | phone +49-9131-7701-0 | | 3SOFT GmbH | fax +49-9131-7701-80 | | Wetterkreuz 19a | | | D-91058 Erlangen | e-mail kindl@DreiSoft.de | | Germany | | ------------------------------------------------------------ From jhillman@wrs.com Fri May 10 08:37:37 1996 From: Jon Hillman Date: Fri May 10 08:37:40 PDT 1996 Subject: Any VxWorks module that supports RFC1305 (NTP) ? > >There is a need to synchronize the time of day for our VxWorks-based products >which are interconnected in a subnetwork. > >Is there an implementation of the Network Time Protocol (NTP) >that has been ported to VxWorks ? > >Version 2 or Version 3 NTP are OK. > _________________________ | Jon Hillman, FAE | | Wind River Systems | /) (407) 273-4646 (\ / ) Email:jhillman@wrs.com ( \ _( (|_________________________) ) /> (((\ \) /,) / ) / //))/ (\\\\ \_/ / \ \_/ ///// \ / \ / \ _/ \_ / ----/ /----------------------\ \---- / / \ \ From wbrown@csg.lbl.gov Fri May 10 09:02:50 1996 From: wbrown@csg.lbl.gov (Bill Brown) Date: Fri May 10 09:02:53 PDT 1996 Subject: Re: Any VxWorks module that supports RFC1305 (NTP) ? I think there is such a thing in the archives. Information on the vxWorks archive is available by sending e-mail to vxworks_archive@ncar.ucar.edu with the text send index as the body of the message. The VxWorks exploder archives is located at 'csg.lbl.gov' in '/pub/vxwexplo'. nonymous login is permitted. -bill "duty mailing-list-grunt of the day" From gquiroz@msmail3.hac.com Fri May 10 09:33:23 1996 From: "Quiroz, Gabriel" Date: Fri May 10 09:33:25 PDT 1996 Subject: socket buffer I am trying to perform some non-blocking writes from a vxworks board to a socket where the reader (highes priority application task) on another vxworks board is suspended for a short period of time (This only happens on occasional CPU spikes in the system). My writer task gets an EWOULDBLOCK error as expected. My question is how I can increase the socket buffer size in vxWorks 5.2? I can't afford to have the writer block or try over and over again to send the data. In ../vw/h/net/socketvar.h I see a #define SB_MAX (64*1024) which indicates the max chars in sockbuf. Is this where vxWorks gets its socket buffer sizes? Also, there is a socket structure called "socket" which I assume is used by vxWorks to maintain the socket information. Does anyone know if I can access this socket structure via some vxWorks pointer? Any insight would be appreciated. Gabriel Quiroz Hughes Aircraft Co. (619)543-4173 gquiroz@msmail3.hac.com From espin@serval.kla.com Fri May 10 09:38:51 1996 From: espin@serval.kla.com (Geoffrey Espin x2496) Date: Fri May 10 09:39:34 PDT 1996 Subject: RE: Suggestions for R6000 dev environment (really about PPC) Joe, > Subject: Suggestions for R6000 dev environment > From: bonnett@sunland.gsfc.nasa.gov (Joe Vitale) > Organization: NASA Goddard Space Flight Center -- Greenbelt, Maryland USA > > Our target hardware is an R6000 with a PCI bus interface. Since it > will be some time before the target hardware is available, we are > planning to purchase an Ultra PowerPC 603 (or 604) motherboard (with > PCI connectors) to allow us to being software development while I've been working with the VxWorks on the Ultra PPC604 for the last few months. Migrating legacy s/w from VxWorks 5.2/68k to VxWorks 5.2/PPC went well. Then upgrading to VxWorks 5.3/PPC-beta has been equally painless. The BSP and drivers are better but still lack polish. VxWorks has truly minimal support for PCIbus (mostly some header file defines). A library to write to configuration space, find devices, setup bridge(s), assign IRQs and memory, is needed (e.g. VxBIOS :-)). There may be something happening on the VxWorks/x86 front at WRS (or rather, a subcontractor of theirs) but it is not on the PPC. I've written enough to support drivers for a couple of off-the-shelf boards and a custom board -- but admittedly you end up hard-coding a certain amount. Not exactly Plug 'n' Play. :-) > system. Since the new Tornado environment will not support our > hardware in time to meet our schedule, we are planning on using > Version 5.2 of VxWorks. We will buy VxWorks for the Ultra PowerPC >... > We are planning on using the Green Hills software development tools > hosted on a unix workstation (probably a sun solaris or dec alpha > machine). The interface between the GH host and the Ultra PowerPC WRS is only supporting GNU toolset in 5.3 (beta at least) which seems to have its own set of problems for PPC. The DIAB Data worked fine in 5.2. Always heard dubious reports about GH. Moving code back and forth between 5.3 and 5.2 shouldn't be a big deal, so go with the latest and greatest. Using the PPC as an interim solution will be of only minor benefit to learn about PCIbus configuration without BIOS. I would give VxWorks/x86 a try. Geoff -- Geoffrey Espin espin@idiom.com From @lazarus.nrtc.northrop.com:wong@csd.nad.northrop.com Fri May 10 10:08:26 1996 From: "J. Pierre Wong" Date: Fri May 10 10:08:40 PDT 1996 Subject: Z85C30 driver I am looking for a driver for a Z85C30 Enhanced Serial Communications Controller. Does anyone have or know of a simple driver that they would be willing to share? Thanks in advance. Pierre Wong Northrop Grumman wong@csd.nad.northrop.com tel: 310.332.6840 fax: 310.644.3652 From uchenick@bud.tate.com Fri May 10 10:30:01 1996 From: Gordon Uchenick Date: Fri May 10 10:30:08 PDT 1996 Subject: Re: NTP support > >Is there an implementation of the Network Time Protocol (NTP) >that has been ported to VxWorks ? > >Version 2 or Version 3 NTP are OK. > Try ftp.atd.ucar.edu File is /pub/vxworks/ntp3.tar.gz HTH, Gordon uchenick@tate.com From johill@lanl.gov Fri May 10 10:32:04 1996 From: Jeff Hill Date: Fri May 10 10:32:13 PDT 1996 Subject: Re: socket buffer This is a multi-part message in MIME format. --------------446B9B3D794BDF3215FB7483 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit > > Submitted-by gquiroz@msmail3.hac.com Fri May 10 09:33:23 1996 > Submitted-by: "Quiroz, Gabriel" > > My question is how I can increase the socket buffer size in vxWorks > 5.2? try setsockopt() on parameters SO_SNDBUF and SO_RCVBUF Jeff --------------446B9B3D794BDF3215FB7483 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="signature" ______________________________________________________________________ Jeffrey O. Hill Internet johill@lanl.gov LANL MS H820 Voice 505 665 1831 Los Alamos, NM 87545 USA FAX 505 665 5107 --------------446B9B3D794BDF3215FB7483-- From uchenick@bud.tate.com Fri May 10 11:28:40 1996 From: Gordon Uchenick Date: Fri May 10 11:28:46 PDT 1996 Subject: Re: socket buffer You can expand the socket buffer "on the fly" by using setsockopt(). I've done this successfully many times. The following code chunks will give you the basic idea. This code will not block the sending task. It's very useful where one task is writing to many different sockets and can't afford to block for any one of them. int buffSize; int intSize; int true = 1; int fd; { . . . fd = socket(PF_INET, SOCK_STREAM, 0); ioctl(fd, FIONBIO, &true); intSize = sizeof(buffSize); getsockopt(fd, SOL_SOCKET, SO_SNDBUF, &buffSize, &intSize); . . . sendStuff(buff, nbytes) . . . } sendStuff(char *buff, int nbytes) { int bytesSent; if ((bytesSent = send(fd, buff, nbytes, 0)) == ERROR) { bytesSent = 0; if (errno == EWOULDBLOCK) { buffSize += nbytes; if (setsockopt(fd, SOL_SOCKET, SO_SNDBUF, &buffSize, sizeof(buffSize)) == ERROR) { close(fd); return (ERROR); } } else { close(fd); return (ERROR); } else if (bytesSent == nbytes) return (!ERROR); return (sendStuff(buff + bytesSent, nbytes - bytesSent)); } HTH, Gordon From smorgan@yurie.com Fri May 10 13:56:51 1996 From: smorgan@yurie.com Date: Fri May 10 13:56:57 PDT 1996 Subject: Re: socket buffer Take a look at setsockopt() SO_SNDBUF option. I think you will want to use this. Also look at SO_RCVBUF and the TCP receive window implication. Steve Morgan smorgan@yurie.com >I am trying to perform some non-blocking writes from a vxworks board to a >socket where the reader (highes priority application task) on another vxworks >board is suspended for a short period of time (This only happens on occasional >CPU spikes in the system). My writer task gets an EWOULDBLOCK error as >expected. My question is how I can increase the socket buffer size in vxWorks >5.2? I can't afford to have the writer block or try over and over again to >send the data. >In ../vw/h/net/socketvar.h I see a #define SB_MAX (64*1024) which indicates >the max chars in sockbuf. Is this where vxWorks gets its socket buffer sizes? >Also, there is a socket structure called "socket" which I assume is used by >vxWorks to maintain the socket information. Does anyone know if I can access >this socket structure via some vxWorks pointer? From daemon@csg.lbl.gov Sat May 11 04:00:38 1996 From: daemon@csg.lbl.gov Date: Sat May 11 04:00:43 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sat May 11 04:00:31 PDT 1996 Subject: trackball for a 68k Subject: Re: Native Java VM or bytecode to 68k/i960 etc for embedded system Subject: Re: VxWorks v5.1.1 (tNetTask) Subject: Re: dosFs troubles... Subject: Re: vxMp with Reflective Memory Subject: Re: How do you make your code "faster"? ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: trackball for a 68k Date: 1 May 1996 20:57:18 GMT From: isr@ix.netcom.com(Robert C. Smith) Organization: Netcom Message-ID: <4m8j7e$cce@dfw-ixnews3.ix.netcom.com> We're running VxWorks 5.3 on a MVME162 (68040) with a MVIP302 octal serial industry pack. Does anyone know of a source for a trackball and drivers? Thanks, Bob Smith ETEC Polyscan 520-806-3482 isr@ix.netcom.com --------------------------- Newsgroups: comp.lang.java,comp.os.vxworks,comp.realtime Subject: Re: Native Java VM or bytecode to 68k/i960 etc for embedded system Date: Fri, 10 May 1996 21:50:56 GMT From: tlewis@iadfw.net (Ted Lewis) Organization: customer of Internet America Message-ID: <4n06sj$v2f@news-f.iadfw.net> References: <318E6F3C.17F4@gpsys.com> You might want to check out www.isi.com Look for the Java press release. Integrated Systems has a Java/Embedded system solution. Ted Lewis Shaun Case wrote: >I have been hunting around for several days for >a Java development environment suitable for a >realtime embedded system, preferably a 68k or i960-based >platform with a well-supported realtime operating system. >For performance reasons we'd like something that compiles >java bytecode into native instructions. >After extensive searching I have come up completely empty-handed. >Can anyone provide me with pointers to something along these >lines, or point me in the direction of someone who is working on >something like this? >Maybe I'm barking up the wrong tree, but this seems like a very >reasonable thing to want to do. >Thanks, >Shaun --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks v5.1.1 (tNetTask) Date: 9 May 1996 22:44:37 -0700 From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star, Oakland, CA Message-ID: <4mul45$2ca@idiom.com> References: <31922B6A.160E@jpl.nasa.gov> In article <31922B6A.160E@jpl.nasa.gov>, Henry Valtier wrote: >(tNetTask): duplicate IP address!! sent from ethernet address: <..addr..> This pretty much means what it says. When a network interface is initialized (SIOCSIFADDR, setting its IP address), one of the first things that happens is an "announcement" ARP. This is to see if anyone else is using the same IP address on the same network. If it receives an ARP reply, then it is due to another host on the net having the same IP address. This is clearly not desirable. - -- H.J. Bae http://www.peacefulstar.oakland.ca.us --------------------------- Newsgroups: comp.os.vxworks Subject: Re: dosFs troubles... Date: Fri, 03 May 1996 01:30:25 GMT From: mfisher@shore.net (Mark Fisher) Organization: Shore.Net/Eco Software, Inc; (info@shore.net) Message-ID: <4mbd3v$nje@shore.shore.net> References: <199605011715.KAA10007@paris.efi.com> "Yuet Fung Siu" wrote: >The problem you described is very similar to the one I discovered last >year with vxWorks 5.2. To confirm this, instead of using the shell, >try to spawn a task, i.e., 'sp copy,"srcfile","desfile"', to do the >copy and observe what is happening. If you discovered that the new >task never exits and is always ready, that's it. If you do a 'tt' or >so, you should find that the task is doing some hashLib or sllLib >stuff. >Since we have the source code to 5.2, I did find the bug in the dosFsLib. >The problem was due to a hashList corruption inside the dosFsLib. I've >reported the bug to WRS: SPR 4418. I don't think there's a work around >without the source code. You may want to contact WRS and ask for a fix >or patch. >I'm just surprised that it takes so long before anyone else discover this. >Probably noone else is using dosFsLib with 5.2 ?? I have found, and reported a number of problems to WRS concerning the dosFsLib. As a result I have had more than a few updates sent to me. As one tech support person said to me "dosFsLib is the only code we are continually updating and pathcing in 5.2". Thus if anyone EVER even suspects trouble with the dosFsLib then don't waste a lot of time trying to debug it; instead get the latest version of dosFsLib from WRS. If the problem persists after this, then you may need to do some investigation. We use the disk a LOT in our system and I have found the version I have now is quite stable. The original poster had problems with IDE, I have only ever used SCSI, so there could also be problems in the IDE drivers. Mark Fisher --------------------------- Newsgroups: comp.os.vxworks Subject: Re: vxMp with Reflective Memory Date: Thu, 02 May 1996 10:19:40 -0400 From: Fred Roeber Organization: BBN Systems and Technologies Message-ID: <3188C47C.11524120@bbn.com> References: <4m82q3$m3t@hammer.msfc.nasa.gov> John Weir wrote: > > We have three VME-based vxWorks systems interconnected with a VMIC RMS > system which replicates all data writes to an address range to the > corresponding address in each system. We are considering using the vxMp > product with this system to implement shared memory among our VME > chassis. Has anyone tried this? How did it work? Thanks in advance for > all replies? I have used VxMP and VMIC reflective memories but never tried both together. I would, expect, however that you would run into serious problems in the area of synchronization and notification. At the lowest levels, VxMP uses spin locks to synchronize access to critical information in the shared VxMP memory areas and uses interrupts between processors to signal when actions should be taken (eg semGive from one processor to another). Further, VxMP normally places the shared memory structures in local memory on the VME backplane master (ie processor 0). To get what you want to work you would (at a minimum) need to: - - set it up so that the VxMP structures were all stored in the VMIC reflective memory. - - Since you would have multiple VME chassis with their own backplane masters sharing the same VxMP structures, one of them would need to assume overall control of the area and set it up. The other processors in the system would end up waiting for first one to complete setup of area which would introduce some bad cross chassis dependency conditions - - you would need to set up VxMP code to use VMIC cross chassis interrupts - - you would need to replace the test-and-set logic used in the VxMP spin locks with something that would work over reflective memory. Since atomic operations are not supported on VMIC card locations as far as I remember, you would have to switch to some SW sync mechanism. From what I remember when I used to be able to look at the VxMP source, there was some way of using a SW synchronization scheme but it had terrible performance. So my answer to the question is that with a lot of work and a source license one could probably get VxMP to work over reflective memory. Performance wouldn't be too good though. While this is an interesting concept, I would suggest trying to use the reflective memory with other mechanisms. The mechanism we normally use in managing our use of reflective memories is the concept of ring buffers. These work especially well if you have a consumer/producer situation with only a single writer/single reader. BTW, one of the better books I have seen on these issues is "Concurrent Programming, Fundamental Techniques for Real-Time and Parallel Software Design" by Tom Axford (Wiley publishers). Fred - -- | Fred J Roeber, BBN Systems & Technologies | | 50 Enterprise Place Middletown, RI 02842-5202 | | froeber@bbn.com 401-849-2543 (X48) | | See http://www.bbn.com/tv for TraceMaker info | --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How do you make your code "faster"? Date: Thu, 02 May 1996 10:25:14 -0400 From: Fred Roeber Organization: BBN Systems and Technologies Message-ID: <3188C5CA.6E9DAAD9@bbn.com> References: JOHN R. COBARRUVIAS wrote: > My task required a rather high data collection rate, which after some > trial and error I was able to accomodate. The analog boards were taking > about 70u seconds to complete the data collection. I found the data > transfer from the board to RAM was about 140u sec. I used the bcopy > routine instead and the data transfer is now about 70u seconds. So.....now > comes the question....... > > Is there any other generally known time savering hints, techiques, > do's/don'ts, etc I can use? There are lots of ways to get faster data transfer but they usually depend on the board and CPU you are using. One simple thing is to make sure data areas are aligned on good boundaries (ie 16 byte) and you are moving large data blocks. bcopy will run faster (on 68K CPUs at least) under these conditions. To get more speed you need to look at using DMA engines on your CPU and block transfers if you are working with VME boards. Fred - -- | Fred J Roeber, BBN Systems & Technologies | | 50 Enterprise Place Middletown, RI 02842-5202 | | froeber@bbn.com 401-849-2543 (X48) | | See http://www.bbn.com/tv for TraceMaker info | --------------------------- End of New-News digest ********************** From daemon@csg.lbl.gov Sun May 12 04:00:43 1996 From: daemon@csg.lbl.gov Date: Sun May 12 04:00:49 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sun May 12 04:00:35 PDT 1996 Subject: Re: Any HTTP servers available on VxWorks? Subject: Re: Bit-fields and byte/word access on 68K Subject: Re: Native Java VM or bytecode to 68k/i960 etc for embedded system Subject: Re: VXWORKS on the HP ???? Subject: Re: Q: Producing boot proms for MVME167 Subject: Re: i960 RP board with high speed serial Subject: Unaligned data in MIPS bootrom Subject: Looking for a Synchronous z8530 Driver? Subject: MVME167/162 DMA question. Subject: How to determine current memory usage ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Any HTTP servers available on VxWorks? Date: 2 May 1996 18:53:08 GMT From: alexm@vlsi19.gsfc.nasa.gov (Alex Measday) Organization: NASA Goddard Space Flight Center -- Greenbelt, Maryland USA Message-ID: References: <4m2crn$41c@nrtphba6.bnr.ca> <3185165F.41C2@ucar.edu> Joe Van Andel wrote: > > [Bob Van Andel] has an HTTP server specifically designed > for small, real-time systems. > Sounds like nepotism to me! :) You also might try writing your own HTTP server; the HTTP protocol is fairly simple. I run a Tcl-based HTTP server on our VxWorks system. In addition to supporting normal file transfers, the server can also access information in global memory regions and dynamically generate HTML status pages for our various subystems. You can see this server in action at: http://vlsi19.gsfc.nasa.gov:8224/software/gentle/hope.html ======================================================================== | Alex Measday Code 521 - Microelectronic Systems Branch | | Integral Systems, Inc. NASA Goddard Space Flight Center | | alexm@vlsi.gsfc.nasa.gov Greenbelt, MD | ======================================================================== ) http://vlsi22.gsfc.nasa.gov/~alexm/ ( ======================================================================== --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Bit-fields and byte/word access on 68K Date: Fri, 10 May 1996 14:41:34 -0400 From: Kevin Rice Organization: NASA Goddard Space Flight Center -- Greenbelt, Maryland USA Message-ID: <31938DDE.2DB6@vlsi.gsfc.nasa.gov> References: <31922C57.5224@cliffy.lmtas.lmco.com> <31934B07.6E6D@bnl.gov> Gonzalo Maldonado wrote: > > Shash Chatterjee wrote: > > > > Hi! > > > > I'm trying to use bit-fields to access registers on a > > VMIC VMIVME-3123 A/D board. > Don't use bit-fields. They aren't portable. Every compiler does them differently. It's fine to use them if you are doing some high-level abstract data type but when talking to hardware skip 'em. Just go with ptrs & masks. Use macros to make things look nice & structures to map to a devices memory map to keep things relatively abstract. Also you could out & grab linux/netbsd/freebsd source and look at a few drivers to see how the masters do it. Kevin Rice --------------------------- Newsgroups: comp.lang.java,comp.os.vxworks,comp.realtime Subject: Re: Native Java VM or bytecode to 68k/i960 etc for embedded system Date: Fri, 10 May 1996 12:12:31 -0700 From: gumby@cygnus.com (D.V. Henkel-Wallace) Organization: Cygnus Support Message-ID: References: <318E6F3C.17F4@gpsys.com> <318F1DE5.A8@lmtas.lmco.com> <319127A1.3F03@oro.net> > I have been hunting around for several days for > a Java development environment suitable for a > realtime embedded system, preferably a 68k or i960-based > platform with a well-supported realtime operating system. > For performance reasons we'd like something that compiles > java bytecode into native instructions. Cygnus Support has started work on a new Java front-end for Gcc. This will be a portable optimizing compiler for Java byte-code files (initially) as well as Java source files. We also intend to enhance gdb (the GNU debugger) to fully support Java. As I'm sure you know, the GNU development tools already support cross-compilation for 68k and i960 running in many embedded systems, to a large part due to the work we've put into this area, as well as the work of volunteers around the net. Of course a Java compiler needs to target a Java run-time system, even if the compiler is generating machine code. There is no portable Java run-time system yet. Sun's JDK is difficult to port, and its license is also a problem. It's alsowell-known to be inappropriate for deeply embedded systems. The Kaffe run-time system for Java looks promising. However,it is currently limited to x86-based systems (though that will change soon). > Maybe I'm barking up the wrong tree, but this seems like a very > reasonable thing to want to do. We agree, that is why we are doing it. Regards, David Henkel-Wallace Cygnus Support --------------------------- Newsgroups: comp.os.vxworks,comp.sys.hp.hpux Subject: Re: VXWORKS on the HP ???? Date: 03 May 1996 07:20:35 -0400 From: dave@sanders.com (Dave Uhrenholdt) Organization: Sanders SSD Message-ID: References: <31815875.2BB7@phbtsus.com> Sender: dave@cushing.nowhere.com In article partin@heman.jsc.nasa.gov (Charles Partin) writes: In article <31815875.2BB7@phbtsus.com>, Aseem Bakshi wrote: > Hi, > > Under Sun Solaris it is possible to run Vxworks. Is it possible to do > the > same with HP/UX ????? > We run VxWorks 5.2 under HPUX 9.05 on a 9000/755 with no problems. > Are there any other options available to run real time under Hp/UX ??? > Thanks Aseem I am sure there are other real-time systems (pSOS, VTRX, etc) running on HP's. I have never had the need to look for them. HP has a product called HP-RT which is a repackaging of LynxOS. Dave --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Q: Producing boot proms for MVME167 Date: 4 May 1996 00:01:57 -0400 From: thesharps@aol.com (THE SHARPS) Organization: America Online, Inc. (1-800-827-6364) Message-ID: <4mekrl$pvm@newsbf02.news.aol.com> References: <4m80o6$5cg@geraldo.cc.utexas.edu> Reply-To: thesharps@aol.com (THE SHARPS) Sender: root@newsbf02.news.aol.com When you recreate the bootrom image in motorola s record format, you will get a single file called bootrom.hex. This single file must then be uploaded into a prom burner memory which is properly configured to write the even addresses to one prom and the odd to the other. Hope this helps. Tim Sharp Cambridge Parallel Processing tsharp@cppus.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: i960 RP board with high speed serial Date: 4 May 1996 00:02:05 -0400 From: thesharps@aol.com (THE SHARPS) Organization: America Online, Inc. (1-800-827-6364) Message-ID: <4mekrt$pvo@newsbf02.news.aol.com> References: <9604301349.AA22785@denoon.positron.med.ge.com> Reply-To: thesharps@aol.com (THE SHARPS) Sender: root@newsbf02.news.aol.com Mike, Check with VMETRO. They are now in the board business in addition to logic analyzers, and are currently developing MIDAS boards using the i960RP chip with a at least two PMC slots. They are also working on a version to connect to the RaceWay if that interests you. Do not expect these boards until July-August time frame. Hope this helps. Tim Sharp Cambridge Parallel Processing tsharp@cppus.com --------------------------- Newsgroups: comp.os.vxworks Subject: Unaligned data in MIPS bootrom Date: Fri, 10 May 1996 17:19:59 -0400 From: Jeff Daly Organization: Sanders, A Lockheed Martin Co. Message-ID: <3193B2FF.C28@rapnet.sanders.lockheed.com> hi all, i'm attempting to compile a bootrom for an R4X00 board. i'm using vxWorks 5.2b, and the sde-mips toolset for solaris 2.x my first attempt to run the boot code resulted in an exception which i decoded as an address error (load). after jumping to romStart(), the data section is copied from _etext to DRAM. dumping the symbol table for the bootrom, i saw where my problem was: bfc40c80 r __ctype_tab bfc421c5 r strayString bfc42942 r cacheMsg bfc435b9 A _etext bfc435b9 A etext the end of text is at an unaligned address, so the first read causes an error. i couldn't find any information for the tools that said anything about alignment of read-only data sections. attempting to hack in some dummy variables didn't work, it just put them before __ctype_tab and the _etext symbol was unaligned at a different address. i'd rather not muck with the vxWorks source code itself, just fix the bootrom build. does anyone have any suggestions?? - -- ************************************************************************ *arrh, thar be disclaimers ahead! \daly@rancid.sanders.lockheed.com* *opinions expressed here are entirely | * *mine and do not represent those of the|"Me on the cover of a mag?, * *largest defense contractor on earth. / * ************************************************************************ --------------------------- Newsgroups: comp.os.vxworks Subject: Looking for a Synchronous z8530 Driver? Date: 2 May 1996 17:20:27 GMT From: mtj@space.mit.edu (Tim Jones) Organization: MIT Center for Space Research Message-ID: <4maqsr$80m@senator-bedfellow.MIT.EDU> Reply-To: mtj@space.mit.edu Has anyone updated the WRS supplied z8530 SCC driver for synchronous communication (as supplied it supports only async), and if so would they mind sharing tips (or source)? I'm running Ver5.2-beta on a PowerPC MVME1603. Thanks, M. Tim Jones AeroAstro Corp. --------------------------- Newsgroups: comp.os.vxworks Subject: MVME167/162 DMA question. Date: 3 May 1996 21:16:40 GMT From: srejto@ll.mit.edu (Stephen Rejto) Organization: MIT Lincoln Lab Message-ID: <4mdt3o$287@llnews.ll.mit.edu> We are using MVME167s and 162 to DMA local moto memory into into a VME slave memory. We are doing D64 MBLT transfers and are seeing very poor rates. ~28 MB/S. It appears that every 8 transfers (64 bits) the motorola board is getting off teh BUS and then back on. The manual says that the DMAC will get off teh bus if its FIFO goes dry. Can this really be happening and how do I prevent it. Does anyone know teh speed at which teh DMAC is proudcing data. (i.e. reading from local moto memory). It appears that our VME slave is faster then the DMA which causes the DMA to release the BUS. YUCK! Every time it releases the bus it causes a 1 usec latency. DTACK times are about 140-150 ns. ANY help appreciated. --------------------------- Newsgroups: comp.os.vxworks Subject: How to determine current memory usage Date: Fri, 3 May 1996 15:53:08 -0700 From: Andy Van Pelt Organization: QUALCOMM, Incorporated; San Diego, CA, USA Message-ID: I would like to track memory usage from within a program, and have not been able to find a routine to call to tell me this. memShow will display the information I want, but I don't want the data displayed, I want to store it and analyze it. How do I find out how much memory is currently allocated and how much is free? I know how to find out the largest block that can be allocated (memFindMax), but that's not quite the same. I hope the answer doesn't involve calling memShow and parsing the output. That's gross! - -- Andy Van Pelt Internet: avanpelt@qualcomm.com QUALCOMM Inc. Work : 619) 658-3214 --------------------------- End of New-News digest ********************** From phil@nevis.naic.edu Sun May 12 09:05:05 1996 From: phil@nevis.naic.edu Date: Sun May 12 09:05:10 PDT 1996 Subject: RE:MVME167/162 DMA question. Stephen Rejto writes:> From root@csg.lbl.gov Sun May 12 07:48:47 1996 > > We are using MVME167s and 162 to DMA local moto memory into into a > VME slave memory. > > We are doing D64 MBLT transfers and are seeing very poor rates. > ~28 MB/S. It appears that every 8 transfers (64 bits) the motorola board > is getting off teh BUS and then back on. The manual says that the DMAC > will get off teh bus if its FIFO goes dry. > > Can this really be happening and how do I prevent it. Does anyone know teh > speed at which teh DMAC is proudcing data. (i.e. reading from local moto > memory). > > It appears that our VME slave is faster then the DMA which causes > the DMA to release the BUS. YUCK! Every time it releases the bus it > causes a 1 usec latency. > > DTACK times are about 140-150 ns. When i tried D64 dma transfers between a mv167 and a mv162 I got about 28 mb/sec too. The dtack times were also around 100+ns. I talked to motorola about the dtack times and they said that is normal for a 25 Mhz vmechip2. A 33mhz mv167 should give about 80ns for dtack. I went through the mv167/mv162,mv162fx manauls that i have and found the following local bus i/o rates for burst transfers. 162fx: page 1-29 of the installation guide... clocks memory max i/o rate local bus 3,2,2,2 write 25mhz 44mb/sec 4,2,2,2 rd non-interleaved 25mhz 40mb/sec 4,1,1,1 rd interleaved 25mhz 57mb/sec 4,2,2,2 write 32mhz 52mb/sec 5,3,3,3 rd non-interleaved 32Mhz 36mb/sec 5,2,2,2 rd interleaved 32mhz 46mb/sec mv167 programmers ref guide pg 4-1 clocks memory max i/o rate local bus 2,1,1,1 writes 25mhz 80mb/sec 4,1,1,1 rd 25mhz 57mb/sec 5,1,1,1 rd 25mhz,with parity 25mhz 50mb/sec (i'm not sure what the mv167 33mhz values are...) mv162 prog. ref guide pg 3.5 clocks memory max i/o rate local bus 3,2,2,2 write,25mhz 44mb/sec 4,2,2,2 rd,25mhz,non-intlv,no tea on pe 40mb/sec 4,1,1,1 rd,25mhz,intlv,no tea on pe. 57mb/sec 5,3,3,3 rd,25mhz,non-intlv, tea on pe 28mb/sec 5,2,2,2 read,intlv,tea on pe. 36mb/sec 4,2,2,2 write,33mhz 53mb/sec 5,3,3,3 rd,33mhz,non-intlv,no tea on pe 38mb/sec 5,2,2,2 rd,33mhz,intlv,no tea on pe 48mb/sec 6,4,4,4 rd,33mhz,non-intlv,tea on pe 29mb/sec 6,3,3,3 rd,33mhz,intlv,tea on pe 35mb/sec As you can see, they vary a lot. Some memory configurations don't allow interleaving. 28mb/sec would be a very respectable throughput for some of these memory configurations. What caught me was that the mv162fx 33mhz board reads the memory slower than the 25mhz board!!!! You are guaranteed that the i/o transfers out of the memory will not be larger than these values. These times also assume a burst xfer (i'm not sure if the vmechip2 does that for dma xfers). So when you read all the nifty advertisements about 80mb/sec with D64 over the vmebus, take it with a grain of salt.... phil perillat arecibo observatory phil@naic.edu ps. Another thing to be careful with is the local bus priorities which are hard wired. We are using a mv162fx to xfer into the local memory through the ip interface and eventually read the data out of memory over the vmebus using D64 xfers. We are also getting an interrupt from the ip device at the end of each block xfer. Things would work fine until the cpu tried to do an iack cycle to the ip device. Since the cpu has the lowest priority on the local bus, it would sometimes not get the iack cycle done in time if a D64 xfer was active. From aorad@netvision.net.il Sun May 12 12:38:22 1996 From: Amir Orad Date: Sun May 12 12:38:25 PDT 1996 Subject: Loading an object and calling its functions ... Hi. I'm trying to find an elegant way of solving the following problem : I need to load at run-time an object and call one of its functions. The problem is that I know the function's name only when I load the code. My program is already running and knows the function's name after reading a definition file (where the object name is also written) Any ideas of how can I do it in an elegant way. Right now, the only way to solve this (As I see it) is to use lkup in some way and find the function's pointer according to its name (What will happen if I have more than one function with that name ?). BTW: The function's format is pre-defined, and should be of no problem. ------------------------------------------------------------------ Amir Orad E-Mail : aorad@netvision.net.il ------------------------------------------------------------------ From stan@rti.com Sun May 12 13:51:39 1996 From: Stan Schneider Date: Sun May 12 13:51:42 PDT 1996 Subject: Re: Loading an object and calling its functions ... >> >> I'm trying to find an elegant way of solving the following problem : >> >> I need to load at run-time an object and call one of its functions. This is what symLib is for. Are you running a target-resident symbol table (the default before 5.3)? -- Stan From tja@radstone.co.uk Mon May 13 02:20:40 1996 From: Tim Allen Date: Mon May 13 02:20:55 PDT 1996 Subject: RE: Suggestions for R6000 dev environment (really about PPC) Hi, I missed the original mailing so I hope my comments won't be out of context. >VxWorks has truly minimal support for PCIbus (mostly some header file >defines). A library to write to configuration space, find devices, >setup bridge(s), assign IRQs and memory, is needed (e.g. VxBIOS :-)). >There may be something happening on the VxWorks/x86 front at WRS (or >rather, a subcontractor of theirs) but it is not on the PPC. >I've written enough to support drivers for a couple of off-the-shelf boards >and a custom board -- but admittedly you end up hard-coding a certain amount. >Not exactly Plug 'n' Play. :-) I agree the PCI bus support is minimal and, for the code I looked at, actually in contravention of the PCI spec. for some boards (SPR# 6490). I've ported PowerPC 5.2 and 5.3/Tornado to the Radstone PPC603 & PPC604 PowerPC boards and included our own PCI configuration and access routines as part of the BSP. These routines include probing and initialising all PCI devices found on any bus as well as dumping/updating configuration space etc. I have offered this source code to WRS for incorporation into VxWorks as it complies with the PCI spec. and offers extra features. >> system. Since the new Tornado environment will not support our >> hardware in time to meet our schedule, we are planning on using >> Version 5.2 of VxWorks. We will buy VxWorks for the Ultra PowerPC >>... The Radstone Tornado BSP is available now for use with the PowerPC beta test release, and as I understand it the full Tornado release from WRS is not far away (weeks?). >> We are planning on using the Green Hills software development tools >> hosted on a unix workstation (probably a sun solaris or dec alpha >> machine). The interface between the GH host and the Ultra PowerPC > >WRS is only supporting GNU toolset in 5.3 (beta at least) which seems to >have its own set of problems for PPC. The DIAB Data worked fine in 5.2. >Always heard dubious reports about GH. Moving code back and forth >between 5.3 and 5.2 shouldn't be a big deal, so go with the latest >and greatest. I used the Green Hills tools in 5.2 without much trouble and switched to GNU in 5.3 also. There were a few compiler/assembler differences to sort out but now real problems. If you're interested in the Radstone PowerPC product please contact Radstone Technology Corp. Montvale, NJ. email radstone@radstone.com or phone (800) 368 2738. Regards, Tim. +------------------------------------------------------------------+ | Tim Allen phone: +44 (0)1327 359444 | | Radstone Technology PLC, UK. fax: +44 (0)1327 358112 | +------------------------------------------------------------------+ From mslifcak@hitel.com Mon May 13 05:16:41 1996 From: mike slifcak Date: Mon May 13 05:16:45 PDT 1996 Subject: Subscribe subscribe From froeber@bbn.com Mon May 13 06:43:50 1996 From: Fred Roeber Date: Mon May 13 06:43:52 PDT 1996 Subject: Re: Z85C30 driver J. Pierre Wong wrote: > > I am looking for a driver for a Z85C30 Enhanced Serial Communications > Controller. Does anyone have or know of a simple driver that they > would be willing to share? The Z8530 chip is a standard serial chip WRS supports. You should be able to get a driver from them. The source for the driver comes with at least some BSPs. Check the src/drv/serial directory in your VxWorks distribution for the file z8530Serial.c (even object licenses to VxWorks contain some source code in the src area). Fred -- | Fred J Roeber, BBN Systems & Technologies | | 50 Enterprise Place Middletown, RI 02842-5202 | | froeber@bbn.com 401-849-2543 (X48) | | See http://www.bbn.com/tv for TraceMaker info | From aisie.aisinc.com!aisinc.com!tam@msen.com Mon May 13 07:54:22 1996 From: tam@aisinc.com (Tod More) Date: Mon May 13 07:54:25 PDT 1996 Subject: Re: SW Engineering Openings > > Submitted-by pata@aisg.com Thu May 9 12:58:33 1996 > Submitted-by: pata@aisg.com (Pat Abruzzise) > > I have an urgent need for SW Engineers using Vx Works for a 5 month contract on > the west coast of Florida. Please contact > > Pat Abruzzise > AISG > 1 800 780-2598 or personal pager 1 800 570-0054 > Do you want people on site, or can the work be done remotly ? Tod More. From andyg@emass.com Mon May 13 09:50:38 1996 From: Andy Gross Date: Mon May 13 09:50:41 PDT 1996 Subject: Re: Z85C30 driver the vxWorks Users Group Exploder wrote: > > Submitted-by froeber@bbn.com Mon May 13 06:43:50 1996 > Submitted-by: Fred Roeber > > J. Pierre Wong wrote: > > > > I am looking for a driver for a Z85C30 Enhanced Serial Communications > > Controller. Does anyone have or know of a simple driver that they > > would be willing to share? > > The Z8530 chip is a standard serial chip WRS supports. You should be > able to get a driver from them. The source for the driver comes with > at least some BSPs. Check the src/drv/serial directory in your VxWorks > distribution for the file z8530Serial.c (even object licenses to VxWorks > contain some source code in the src area). Fred > If you're using Tornado, then the driver is in src/drv/sio/z8530Sio.c. The Upgrading BSP to Tornado uses that driver as a reference. It didn't come with my BSP, but WRS sent it to me. Andy -- Andy Gross EMASS Senior Software Engineer 10949 E. Peakview Ave. E-Mail: andyg@emass.com Englewood, CO 80111 (303) 705-3762 From bpsattler@anet.rockwell.com Mon May 13 10:58:12 1996 From: "Bradley P. Sattler" Date: Mon May 13 10:58:16 PDT 1996 Subject: Problems producing IEEE695 files from GNU ld linker Using the GNU "ld" linker (version number 1.97.1) to produce "ieee" output format files, error code 134 is output. The resultant file has symbol data in it, but no code. The linker seems to produce a.out, COFF, and srec output format files with no errors. The ld68k command line, and script file used are as follows: /app/green/ld68k -Ur -T script obj/ADC_ADC.o obj/ADC_Run_BIT.o .... -o coreofp -M >./coreofp/map contents of the script file: SECTIONS { .text : { *(.text) } .data : { *(.data) } .bss : { *(.bss) } } OUTPUT_FORMAT(ieee) Any ideas/similar experiences? Thanks. bpsattler@anet.rockwell.com From ferrara@adissw.enet.dec.com Mon May 13 11:17:11 1996 From: "Bob Ferrara DTN:264-3094" Date: Mon May 13 11:17:14 PDT 1996 Subject: RAM Disk questions... Has anyone used a RAM Disk under VxWorks that supports VMS style files? I need to create a VxWorks RAM disk in my local memory space in order to copy (probably via ftp) files from a OpenVMS system (RMS file format). I understand native VxWorks supports Dos, RT-11 and a Raw file system. Thanks for any comments, Bob Ferrara Digital Equipment Corporation From ceselsa@coastalnet.com Mon May 13 14:27:26 1996 From: ceselsa@coastalnet.com (Carlos Martinez) Date: Mon May 13 14:27:29 PDT 1996 Subject: VxWorks and FORTRAN Is anybody out there using WSR VxWorks with a Fortran compiler? I would like to exchange info. about that. Carlos Martinez, ceselsa@coastalnet.com (919) 466-9133 Cherry Point, NC. From erxatech@inrete.it Tue May 14 00:52:42 1996 From: erxatech@inrete.it (Paolo Magistris) Date: Tue May 14 00:52:45 PDT 1996 Subject: 802.3 protocol Hello, I need to know is someone has information about 802.3 protocol on vxWorks. Normally vxWorks is unable to manage 802.3 frame. There is a solution for this lacks? Thank You Paolo Magistris ================================================================= _______ |E______| Engineering |__R____| Robotics |____X__| Control Systems |______A| Applications Field Application Engineer Voice: +39-11-7412749 Erxa S.r.l. Fax: +39-11-7412044 via Morghen 22, 10143 Torino (ITALY) E-Mail: erxatech@inrete.alpcom.it ================================================================= From AS_Bridge@msmail3.hac.com Tue May 14 03:31:25 1996 From: "AS Bridge" Date: Tue May 14 03:31:30 PDT 1996 Subject: Undeliverable Mail Unknown Microsoft mail form. Approximate representation follows. Message: 802.3 protocol Sent: Tue, May 14, 1996 2:28 AM To: Haberman, Robert C On Server: 631 Segment3 Date: Tue, May 14, 1996 3:31 AM Reason: Could not be delivered because the destination Microsoft Mail server could not be found. From daemon@csg.lbl.gov Tue May 14 04:02:04 1996 From: daemon@csg.lbl.gov Date: Tue May 14 04:02:09 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Tue May 14 04:00:36 PDT 1996 Subject: SCSI size configuration Subject: Re: Native Java VM or bytecode to 68k/i960 etc for embedded system Subject: DMA controller: L64853 & L64854 Subject: gnu compiler native on vxWorks 68k? Subject: Re: trackball for a 68k ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: SCSI size configuration Date: Mon, 13 May 1996 10:56:33 -0500 From: Alan Currie Organization: NASA/Johnson Space Center Message-ID: <31975BB1.4A57@ghgcorp.com> Reply-To: currie@ghgcorp.com I am switching from a 128Meg SCSI electronic recorder to a 500 Meg SCSI hard disk. I changed the number of blocks to relect change in config.h and rebuilt VxWorks. Number of blocks change shows up in scsiShow as expected. However, I can only put the same amount of stuff out on the disk that I did before. It still thinks I only have 128 Meg of space. Any suggestions? --------------------------- Newsgroups: comp.lang.java,comp.os.vxworks,comp.realtime Subject: Re: Native Java VM or bytecode to 68k/i960 etc for embedded system Date: Sun, 12 May 1996 20:12:10 GMT From: markj@netaccess.co.nz (Mark Jordan) Message-ID: <5d7cc$9d2c.363@mac.co.nz> References: <318E6F3C.17F4@gpsys.com> Shaun Case wrote: >I have been hunting around for several days for >a Java development environment suitable for a >realtime embedded system, preferably a 68k or i960-based >platform with a well-supported realtime operating system. >For performance reasons we'd like something that compiles >java bytecode into native instructions. >After extensive searching I have come up completely empty-handed. >Can anyone provide me with pointers to something along these >lines, or point me in the direction of someone who is working on >something like this? Do you realise that interpreted Java is about 20 times slower in it's execution than standard C code? Having said that, I think that SUN have developed a chip which runs Java byte code directly. Not sure about it's performance though. Cheers, Mark Jordan. +----------------------------------------------------------------------+ | Mark Jordan | Never put off anything you can do today | | Engineer | until tomorrow, because if you liked it, | | markj@netaccess.co.nz | you can do it again tomorrow! | +----------------------------------------------------------------------+ --------------------------- Newsgroups: comp.os.vxworks Subject: DMA controller: L64853 & L64854 Date: 6 May 1996 02:13:01 GMT From: ybpark@dokeby.sst.co.kr (Yun-Bo Park) Organization: Samsung Advanced Institute of Technology Keywords: DMA Message-ID: <4mjn7d$bkq@noc.sait.samsung.co.kr> I use L64854 as S-Bus DMA controller. But I found there is a library for L64853, L64853Dma.c, in VxWorks 5.1. Is there no problem to use the library ? Thanks, Yun-Bo Park ================-=-=-=--=--=---=-----=-----=---=--=--=-=-=-=-================= SAMSUNG Electronics Co.,Ltd. Fax: (02) 559-4314 Daehwa Bldg, 169, Samsung-Dong, Voice: (02) 559-4358 Kangnam-Ku Seoul, Korea 135-090 Internet: ybpark@dokeby.sst.co.kr ================-=-=-=--=--=---=-----=-----=---=--=--=-=-=-=-================= --------------------------- Newsgroups: comp.os.vxworks Subject: gnu compiler native on vxWorks 68k? Date: 13 May 1996 17:32:20 GMT From: Daniel Brodsky Organization: Wind River Systems Message-ID: <4n7rn4$ho9@darya.wrs.com> Hello All, Has anyone ported the gnu compilers and other tools to run native in a vxWorks system? Preferably on a 68k? I would appreciate any pointers to info. Thanks in advance --------------------------- Newsgroups: comp.os.vxworks Subject: Re: trackball for a 68k Date: 13 May 1996 17:36:15 GMT From: Daniel Brodsky Organization: Wind River Systems Message-ID: <4n7ruf$ho9@darya.wrs.com> References: <4m8j7e$cce@dfw-ixnews3.ix.netcom.com> I am pretty sure that Visicom laboratories, the people who make RTGL and VX-Windows, have this driver. http://www.vigra.com isr@ix.netcom.com(Robert C. Smith) wrote: >We're running VxWorks 5.3 on a MVME162 (68040) with a MVIP302 octal >serial industry pack. Does anyone know of a source for a trackball >and drivers? > >Thanks, > >Bob Smith >ETEC Polyscan >520-806-3482 >isr@ix.netcom.com > > - -- Daniel Brodsky |Phone: (303)486-0850 Field Application Engineer |Fax: (303)486-0853 Wind River Systems |Pager: (800)722-1601 4600 South Ulster Street, Suite 700 |Email: danb@wrs.com Denver, CO 80237 |WWW: http://www.wrs.com --------------------------- End of New-News digest ********************** From leonid@rst.co.il Tue May 14 05:57:15 1996 From: leonid@rst.co.il (Leonid Rosenboim) Date: Tue May 14 05:57:19 PDT 1996 Subject: Re: RAM Disk questions > From: "Bob Ferrara DTN:264-3094" > Date: Mon May 13 11:17:14 PDT 1996 > Subject: RAM Disk questions... > > Has anyone used a RAM Disk under VxWorks that supports > VMS style files? > > I need to create a VxWorks RAM disk in my local memory > space in order to copy (probably via ftp) files from a > OpenVMS system (RMS file format). > > I understand native VxWorks supports Dos, RT-11 and a Raw > file system. > > Thanks for any comments, > Bob Ferrara > Digital Equipment Corporation As you probably know, most operating systems these days do not intervene in the file's contents, i.e. there is no more fixed or variable records in a file, just plain byte-strem, and the application in turn needs to treat the file contents as needed. However, if you are transferring them files with FTP, then your VMS implementation of FTP will translate the files to the common bytestream format in accordance with file type. You need to make sure you instruct FTP to transfer the file as TEXT or BINARY, and they will be converted as follows: for TEXT files, every record will be appended with CR+LF, and VxWorks FTP will strip the CR and leave LF at the end of each record. I.e. for TEXT files every record is a Line. For BINARY files, FTP will just send the records as they come, loosing any boundaries between the records, and keeping only the bytes in their order. This is quite appropriate for object files. Once upon a time I have succesfully made VxWorks boot its image file off a VAX/VMS system. In a word: use your VMS FTP implementation for translating file contents correctly, hence it would be recommended to use the FTP client side on VMS, and server on VxWorks. Good Luck, ----------------------------------------------------------------------- Leonid Rosenboim Phone: +972-3-559-8144 R S T Software Industries Ltd. Mobile: +972-50-307-142 P.O.Box 11502, AZUR 58017, Israel Fax: +972-3-559-8244 http://knight.trendline.co.il/~leonid/ E-Mail: leonid@rst.co.il From AS_Bridge@msmail3.hac.com Tue May 14 06:09:13 1996 From: "AS Bridge" Date: Tue May 14 06:09:16 PDT 1996 Subject: Undeliverable Mail Unknown Microsoft mail form. Approximate representation follows. Message: Undeliverable Mail Sent: Tue, May 14, 1996 4:54 AM To: Haberman, Robert C On Server: 631 Segment3 Date: Tue, May 14, 1996 6:06 AM Reason: Could not be delivered because the destination Microsoft Mail server could not be found. From LEONID@vsl.mhs.compuserve.com Tue May 14 06:10:23 1996 From: Leonid Keller Date: Tue May 14 06:10:26 PDT 1996 Subject: Re: Unsubscribe leonid@vsl.mhs.comp Please, unsubscribe leonid@vsl.mhs.comp Thank you -------------------------------------------------------------------------- --- Leonid Keller Gilat Satellite Networks Ltd. Fax: (972)4-9894539 P.O.B. 205, Industrial Park Tel: (972)4-9894543 (ext.015) Yokneam 20692, E-mail: leonid@vsl.mhs.compuserve.com Israel -------------------------------------------------------------------------- --- From AS_Bridge@msmail3.hac.com Tue May 14 06:39:13 1996 From: "AS Bridge" Date: Tue May 14 06:39:16 PDT 1996 Subject: Undeliverable Mail Unknown Microsoft mail form. Approximate representation follows. Message: comp.os.vxworks newsdigest Sent: Tue, May 14, 1996 5:32 AM To: Haberman, Robert C On Server: 631 Segment3 Date: Tue, May 14, 1996 6:38 AM Reason: Could not be delivered because the destination Microsoft Mail server could not be found. From AS_Bridge@msmail3.hac.com Tue May 14 08:52:43 1996 From: "AS Bridge" Date: Tue May 14 08:52:49 PDT 1996 Subject: Undeliverable Mail Unknown Microsoft mail form. Approximate representation follows. Message: Undeliverable Mail Sent: Tue, May 14, 1996 7:47 AM To: Haberman, Robert C On Server: 631 Segment3 Date: Tue, May 14, 1996 8:51 AM Reason: Could not be delivered because the destination Microsoft Mail server could not be found. From AS_Bridge@msmail3.hac.com Tue May 14 08:52:43 1996 From: "AS Bridge" Date: Tue May 14 08:52:49 PDT 1996 Subject: Undeliverable Mail Unknown Microsoft mail form. Approximate representation follows. Message: Re: RAM Disk questions Sent: Tue, May 14, 1996 7:47 AM To: Haberman, Robert C On Server: 631 Segment3 Date: Tue, May 14, 1996 8:51 AM Reason: Could not be delivered because the destination Microsoft Mail server could not be found. From FREGEAU@digsys1.bwi.wec.com Tue May 14 09:22:55 1996 From: "JEAN FREGEAU" Date: Tue May 14 09:22:58 PDT 1996 Subject: shared memory heartbeat I am trying to set up a shared memory region on CPU0 (MVME-162, 4M) by allowing VxWorks to select the shared memory region when it boots. I have set up \target\config\mv162\config.h as follows: #define SM_OFF_BOARD FALSE #if SM_OFF_BOARD #undef SM_ANCHOR_ADRS #define SM_ANCHOR_ADRS ((char *)0xf0800000) /* off-board anchor #address */ define SM_MEM_ADRS SM_ANCHOR_ADRS /* off-board shared #memory address */ define SM_MEM_SIZE 0x00080000 /* 512K */ define #SM_OBJ_MEM_ADRS (SM_MEM_ADRS+SM_MEM_SIZE)/* sh. mem Objects pool adrs #*/ define SM_OBJ_MEM_SIZE 0x80000 /* sh. mem Objects pool #size 512K */ else /* shared memory pool address define NONE = allocate from memory */ // #define SM_MEM_ADRS (RAM_LOW_ADRS - (SM_MEM_SIZE + SM_OBJ_MEM_SIZE)) #define SM_MEM_ADRS NONE /* sh. mem Objects pool address define NONE to allocate from memory */ // #define SM_OBJ_MEM_ADRS (SM_MEM_ADRS+SM_MEM_SIZE) #define SM_OBJ_MEM_ADRS NONE #define SM_MEM_SIZE 0x000e000 /* 64K - 8k */ #define SM_OBJ_MEM_SIZE 0x10000 /* sh. mem Objects pool size #64K */ endif /* SM_OFF_BOARD */ The message I get on boot up is as follows: VxWorks System Boot Copyright 1984-1995 Wind River Systems, Inc. CPU: Motorola MVME162 Version: 5.3 BSP version: 1.1/2 Creation date: Nov 15 1995 Press any key to stop auto-boot... 0 auto-booting... boot device : ei processor number : 0 host name : digsys2 file name : c:\tornado\target\config\mv162\vxWorks inet on ethernet (e) : 100.0.0.50:ff000000 host inet (h) : 100.0.0.10 gateway inet (g) : 0 user (u) : snoopy ftp password (pw) : valid flags (f) : 0x8 target name (tn) : snoopy Attaching network interface ei0... done. Attaching network interface lo0... done. Loading... 337644 + 28820 + 34622 Starting at 0x20000... Attaching network interface ei0... done. Attaching network interface lo0... done. NFS client support not included. Attaching shared memory objects at 0x600... done I was then expecting to see the Shared Memory Anchor at address 0x600 with the offset of the shared memory packet header which would contain the shared memory heartbeat. This is what I see (i.e. offset of 0). -> d 0x600 00000600: 8765 4321 0000 0001 0000 0000 003a 9e18 *.eC!.........:..* 00000610: 0000 0000 003a 9f5c 0000 0000 0000 0000 *.....:.\........* 00000620: 0000 0000 0000 0000 0000 0000 0000 0000 *................* I have another board in the system (CPU1, mv167) that does not seen to see the sm heartbeat. I am assuming that this board is not issuing one. Any ideas or examples of how you got the on board sm to work? Thanks, Jean Fregeau Northrup Grumman 410-765-8797 FREGEAU@digsys1.bwi.wec.com From ceselsa@coastalnet.com Tue May 14 11:39:46 1996 From: ceselsa@coastalnet.com (Carlos Martinez) Date: Tue May 14 11:39:49 PDT 1996 Subject: Parallel driver I am trying to find a VxWorks parallel interface driver. We are currently using target VMEbus PowerPC boards and host SUN SPARC stations. Is there any merciful person out there who could help ? Thank you. Carlos Martinez ceselsa@coastalnet.com (919) 466-9133 (919) 466-9134 Fax From daemon@csg.lbl.gov Wed May 15 04:00:51 1996 From: daemon@csg.lbl.gov Date: Wed May 15 04:00:54 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Wed May 15 04:00:45 PDT 1996 Subject: BSP for M68340EVS Subject: Re: Ruggedized VME boards? Subject: Problem debugging PPC Tornado Beta Subject: Re: Native Java VM or bytecode to 68k/i960 etc for embedded system Subject: Anyone used vxWorks with closely coupled RT processes Subject: Software Engineers Needed ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: BSP for M68340EVS Date: Mon, 06 May 1996 09:38:51 +0000 From: Florian Bittner Organization: Kayser-Threde GmbH Message-ID: <318DC8AB.5F28@kayser-threde.de> Hi there, does anybody know if there is already a Board Support Package for the Motorola M68340 EVS Evaluation System. Please mail an answer if possible. TNX, Florian - -- Florian Bittner Software Engineer Kayser-Threde GmbH Tel +49 89 78720 83 Perchtingerstr.3 Fax +49 89 78720 98 81379 Munich Germany --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Ruggedized VME boards? Date: Sat, 11 May 1996 13:38:36 +0100 From: John Gordon Organization: UnipalmPIPEX server (post doesn't reflect views of UnipalmPIPEX) Message-ID: <31948A4C.EF2@dial.pipex.com> References: <318F8385.3BE8@rugged.com> Radstone Technology also offer a range of PowerPC boards in various levels of ruggedisation. See their home page at http://www.radstone.co.uk for more details John Gordon --------------------------- Newsgroups: comp.os.vxworks Subject: Problem debugging PPC Tornado Beta Date: Tue, 14 May 1996 22:23:47 GMT From: Philip.Wiebe@newnes.com (Philip J. Wiebe) Organization: MIND LINK! - British Columbia, Canada Message-ID: <4nb17p$j4@fountain.mindlink.net> I've been having difficulty debugging PowerPC code on a MVME1604. The problem is when attaching to a task and looking at the call stack (Backtrace option). The call stack sometimes only displays the last called routine, normally some type of VxWorks system call. In fact, the only time I've gotten the call stack to be even remotely close to what is actually the true call stack is for the demo under target/src/demo/color/vxColor.o. Even then, sometimes it does not work. I've tried both the -g and -gdwarf compiler options for generating debug information. Has anyone else had similar problems? Phil. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Philip J. Wiebe "It has yet to be proven that intelligence Software Engineer has any survival value." Newnes Machine Ltd. - Arthur C. Clarke P.O.Box 8, 3550-45th Street S.E. Salmon Arm, BC Canada email: Philip.Wiebe@newnes.com ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ --------------------------- Newsgroups: comp.lang.java,comp.os.vxworks,comp.realtime Subject: Re: Native Java VM or bytecode to 68k/i960 etc for embedded system Date: Tue, 07 May 1996 09:54:45 +0000 From: Ken Garlington Organization: Lockheed Martin Tactical Aircraft Systems Message-ID: <318F1DE5.A8@lmtas.lmco.com> References: <318E6F3C.17F4@gpsys.com> Shaun Case wrote: > > I have been hunting around for several days for > a Java development environment suitable for a > realtime embedded system, preferably a 68k or i960-based > platform with a well-supported realtime operating system. How about an alternative: a language that can used to create both Java Bytecodes (executables) _and_ executables for realtime embedded systems, such as 68k or i960? See: http://www.inmet.com/ada.html http://sw-eng.falls-church.va.us/AdaIC/ http://lglwww.epfl.ch/Ada/ --------------------------- Newsgroups: comp.os.vxworks Subject: Anyone used vxWorks with closely coupled RT processes Date: 7 May 1996 15:07:16 GMT From: abg@ast.cam.ac.uk (Bruce Gentles) Organization: Royal Greenwich Observatory, Cambridge (UK) Keywords: vxworks DSP multiprocessing Message-ID: <4mnov4$rg2@lyra.csx.cam.ac.uk> Reply-To: abg@ast.cam.ac.uk We are developing a RT system which is somewhat latency critical. we have some maths to do. The input arrives in 400 microsecs, the result to be ready 100 microsecs later. we can pipeline the process to get 500 usec to do it in. This takes 3000 usec on a MIPS R4700 or a powerPC603, running vxWorks single processor. We are looking at the trade between several (ie 12) C40 DSP's complete with nightmare spaghetti and a 6-10 processor card vme based system using vxWorks for each processor. does anyone have experience of building this kind of tightly coupled system with vxWorks as the RTOS. We have vxWorks in the layer above this sub-system and I am trying to avoid the embedded DSP array approach with another RTOS as well as vxWorks. Also, any experience or kowedge of vxWorks for DSP's or vxWorks/SPOX interaction would help a lot. Bruce - --------------------------------------------------- Bruce Gentles, Royal Greenwich Observatory Madingley Road, Cambridge, CB3 0EZ Tel +44 (0)1223 374839 Fax 374700 WWW http://www.ast.cam.ac.uk/~abg/ --------------------------- Newsgroups: comp.os.vxworks Subject: Software Engineers Needed Date: 14 May 1996 18:09:34 GMT From: Patty Madry Organization: Honeywell SSD, Clearwater Message-ID: <4nai8u$hm3@spacenns.space.honeywell.com> Honeywell Space and Strategic Systems Operation in Clearwater Florida has a wide variety of openings for Software Engineers, ranging from entry level to senior staff positions. All candidates must possess a BS or MS in Computer Engineering, Computer Science or Electrical Engineering along with expertise in several of the following: C language 1750A experience including assembly language Development of real time operating systems and I/O drivers UNIX VxWorks R3000 experience including assembly language Real time test software (Inertial Systems and electronics) Built in test software Redundancy management software We offer competitive salaries, excellent benefits and professional challenge. For consideration mail or FAX your resume to Ann Sherman, Honeywell Inc., 13350 U.S. Hwy. 19 North, Clearwater FL 34624. FAX (813) 539-5299. Or e-mail resume to mkert@space.honeywell.com. Some positions require U.S. citizenship. An equal opportunity employer. --------------------------- End of New-News digest ********************** From froeber@bbn.com Wed May 15 05:01:51 1996 From: Fred Roeber Date: Wed May 15 05:01:53 PDT 1996 Subject: Re: shared memory heartbeat JEAN FREGEAU says: > > I am trying to set up a shared memory region on CPU0 (MVME-162, 4M) > by allowing VxWorks to select the shared memory region when it boots. > I have set up \target\config\mv162\config.h as follows: > > #define SM_OFF_BOARD FALSE > > #if SM_OFF_BOARD > #undef SM_ANCHOR_ADRS > #define SM_ANCHOR_ADRS ((char *)0xf0800000) /* off-board anchor > #address */ define SM_MEM_ADRS SM_ANCHOR_ADRS /* off-board shared > #memory address */ define SM_MEM_SIZE 0x00080000 /* 512K */ define > #SM_OBJ_MEM_ADRS (SM_MEM_ADRS+SM_MEM_SIZE)/* sh. mem Objects pool adrs > #*/ define SM_OBJ_MEM_SIZE 0x80000 /* sh. mem Objects pool > #size 512K */ else > > /* shared memory pool address define NONE = allocate from memory */ // > #define SM_MEM_ADRS (RAM_LOW_ADRS - (SM_MEM_SIZE + SM_OBJ_MEM_SIZE)) > #define SM_MEM_ADRS NONE > > /* sh. mem Objects pool address define NONE to allocate from memory > */ // #define SM_OBJ_MEM_ADRS (SM_MEM_ADRS+SM_MEM_SIZE) > > #define SM_OBJ_MEM_ADRS NONE > > #define SM_MEM_SIZE 0x000e000 /* 64K - 8k */ > #define SM_OBJ_MEM_SIZE 0x10000 /* sh. mem Objects pool size > #64K */ endif /* SM_OFF_BOARD */ The code you show above has been completely munged by something but I think what you are doing is setting the SM_MEM_ADRS and SM_OBJ_MEM_ADRS to none. As you guess, this will get VxWorks to malloc the memory area from memory rather than use the fixed address that is normally in config.h for the 162. It doesn't change the address of the "backplane anchor" area which still remains at address 0x600 (hence the address you saw below). Problem is that trying to run this way won't work on the MVME162. There is a HW problem with bus snooping that is explained in the 162 target.nr file. The backplane memory area is put at a specific location and then the MMU is set to specifically not cache that area of memory. Since you moved the backplane area to normal memory, it is in an area that is set for copyback caching. This means that when the 162 code updates the backplane heartbeat, until the cache gets flushed, no other system will see the heartbeat. I suggest not doing what you are doing with a 162LX. Fred -- | Fred J Roeber, BBN Systems & Technologies | | 50 Enterprise Place Middletown, RI 02842-5202 | | froeber@bbn.com 401-849-2543 (X48) | | See http://www.bbn.com/tv for TraceMaker info | From francis@aaec.com Wed May 15 05:49:17 1996 From: francis@aaec.com (Kris Francis) Date: Wed May 15 05:49:19 PDT 1996 Subject: Request for technical assistance Can someone please tell me how to perform a burst read from a VME memory board to a Motorola MVME167 board? Thank you for your assistance. Kris Francis francis@aaec.com From ztf@one.one.com Wed May 15 10:17:19 1996 From: ztf@one.com (Zachary T. Frey) Date: Wed May 15 10:17:22 PDT 1996 Subject: File globbing in vxTcl7.4 Has anyone out there using vxTcl7.4 had trouble getting the "glob" command to work? I get the following behavior -- % pwd gnat:/store/ztf/vxtcl % exec ls tclsh tcltest tests % glob * vxtcl: not a plain file. no files matched glob pattern "*" % ls invalid command name "ls" % I'm using VxWorks 5.2 on a MVME167. Are there known problems with 5.2 with using dirLib on an NFS filesystem? That could cause this symptom ... Thanks in advance, Zach Frey -- Zach Frey Open Networks Engineering, Inc. (ONE) Voice: (313) 996-9900 x 1227 777 E. Eisenhower Parkway, Suite 650 Fax: (313) 996-9908 Ann Arbor, MI 48108 Internet: ztf@one.com From greg@eng.nmr.varian.com Wed May 15 10:20:27 1996 From: greg@eng.nmr.varian.com (Greg Brissey x6951) Date: Wed May 15 10:20:30 PDT 1996 Subject: RE:MVME167/162 DMA question. > Stephen Rejto writes:> From root@csg.lbl.gov Sun May 12 07:48:47 1996 > > > > We are using MVME167s and 162 to DMA local moto memory into into a > > VME slave memory. > > > > We are doing D64 MBLT transfers and are seeing very poor rates. > > ~28 MB/S. It appears that every 8 transfers (64 bits) the motorola board > > is getting off teh BUS and then back on. The manual says that the DMAC > > will get off teh bus if its FIFO goes dry. > > > > Can this really be happening and how do I prevent it. Does anyone know teh > > speed at which teh DMAC is proudcing data. (i.e. reading from local moto > > memory). > > > > It appears that our VME slave is faster then the DMA which causes > > the DMA to release the BUS. YUCK! Every time it releases the bus it > > causes a 1 usec latency. > > > > DTACK times are about 140-150 ns. > I ran into a problem with with our proprietary hardware when using D64 DMA with a MVME162LX that probably has nothing to do with your bandwidth problem but I thought I might mention it. It came down to the "time off the bus" and the "time on the bus" parameters on the vmechip2 I had not set. This resulted in a throttled down bandwidth and an on/off Address transaction during the D64 that our hardware wasn't expecting. I then program the vmechip2 stay on the bus till it was completed with the D64 transfer. ( E.G. *((long *)VMECHIP2_TIMEOUTCR) &= ~(TIMEOUTCR_OFF_1024US); /* TIMEOUTCR_OFF_0US */ *((long *)VMECHIP2_TIMEOUTCR) |= TIMEOUTCR_ON_DONE; ) This solve our problem with our hardware and increased our bandwidth. Just be careful not to starve everybody else trying to get the bus. greg brissey varian greg.brissey@nmr.varian.com From SECKA@hqgate.orbital.com Wed May 15 11:27:39 1996 From: SECKA@hqgate.orbital.com Date: Wed May 15 11:27:42 PDT 1996 Subject: CrossWind Task Debugging I have an application in which all my OS resources (semaphores, watchdogs, queues) are created/initialized and all my tasks are spawned by one function. I'd like to use CrossWind to debug one of my tasks while the others are running, and I'd like to be able to get control of the task right after its spawned. But I can't "attach" it until it exists. How can I do this ??? I tried having the task suspend itself upon startup, attaching to the suspended task, and then setting a breakpoint. But I could never get control in CrossWind once I resumed the task (via the shell). Any ideas ??? Thanks in advance! Suzie ================================================================= = Suzie Eckardt = email: secka@hqgate.orbital.com = = Orbital Sciences Corp. = Phone: (703) 406-5478 = = 21700 Atlantic Boulevard = Fax: (703) 406-5553 = = Dulles, VA 20166 = ================================================================= From mwbrown@sass1318.sandia.gov Wed May 15 12:27:19 1996 From: mwbrown@sass1318.sandia.gov (Matthew W. K. Brown) Date: Wed May 15 12:27:21 PDT 1996 Subject: niVXIcpu030 BSP gotcha FYI I have found a gotcha in National Instruments BSP for the niVXIcpu030 board. The problem is actually being caused by a C&H VX431c fiber optic module; but, National's BSP makes recovery *very* difficult. The situation goes like this ... We have C&H VX431c fiber optic modules in multiple crates sending digital I/O back and forth. The boards are setup to generate an interrupt when the input changes. When one crate is powered down, the VX431c sends a tiny glitch--this is not the problem. The other crate (still powered up) detects this glitch and generates an interrupt. Now, here comes the problem ... apparently, because the amplitude of the glitch is so small, the module appears to "change it's mind" and decides that it wasn't an change of state afterall and during the subsequent IACK cycle it doesn't respond. So, the niVXIcpu030 thinks that there is a problem w/ the IACK* daisy chain and disables all interrupts. The niVXIcpu030 reports that it did this--over the console port via logMsg. Naturally, we hardly every watch the output on the console so we totally missed it for a long time. All we knew was that suddenly none of our interrupts were working. Oh Yeah, this only happens some of the time (naturally)--more often in some crates and hardly at all in others ... swaping board didn't help either. Our fix was to put a hook in sysLib.c that allows the user to install a handler for an IACK BERR. If anyone using this board is interested, email me and I'll be happy to send you specific details. Just a warning! Matthew Brown mwbrown@sandia.gov mwbrown@sandia.gov From ihw@rtp.co.uk Thu May 16 01:16:31 1996 From: ihw@rtp.co.uk (Ian Willats) Date: Thu May 16 01:16:34 PDT 1996 Subject: Anyone used vxWorks with closely coupled RT processes In response to Bruce Gentles' (abg@ast.cam.ac.uk) query: > We are looking at the trade between several (ie 12) C40 DSP's complete with nightmare spaghetti and a 6-10 processor card vme based system using vxWorks for each processor. I'd say the pros and cons of using DSPs vs the other processors you mentioned will depend as much on how much data you have to move around between processors, and where it has to live in memory, as on the nature of the maths you have to do. By "nightmare spaghetti" I take it you mean C40 comm port cables all over the place! Nightmarish or not, comm ports can be a very effective way of getting lots of data down a pipeline and directly into the "private" memory of a DSP. Whereas I guess with most RISC VME boards you would have to transfer data via VMEbus into some dual-ported RAM, which might prove to be a bottle-neck. Having said that, if you need that kind of architecture, there are some VME DSP boards with pretty respectable VME/VSB interfaces and lots of dual-ported RAM. > Also, any experience or kowedge of vxWorks for DSP's or vxWorks/SPOX interaction would help a lot. To the best of my knowledge, VxWorks does not run on any "traditional" DSPs (C3x, C4x, etc.). I don't see any fundamental reason why it shouldn't, maybe it's just a question of market demand. Using VxWorks and SPOX (a real-time OS designed for DSPs) together can be quite effective. There is a thing called SPOX-LINK which can be used to allow tasks on VxWorks and SPOX to synchronise and communicate using nice, high-level facilities. It also supports services like remote file access, standard i/o, sockets, etc on the DSPs, which is great for prototyping and debugging. Feel free to call or e-mail me directly if you'd like to talk about it in more detail. Ian ---------------------------------------------------------------- Ian Willats Real-Time Products Ltd. Unit 8, Holt Court, e-mail: ihw@rtp.co.uk Aston Science Park, Jennens Road, tel: +44 121 333 6955 Birmingham. B7 4EJ. England. fax: +44 121 333 5433 ---------------------------------------------------------------- From mey@wzl-wm1.wzl.rwth-aachen.de Thu May 16 01:24:59 1996 From: Andreas Meylahn Date: Thu May 16 01:25:03 PDT 1996 Subject: Re: CrossWind Task Debugging > Date: Wed, 15 May 96 11:27:44 PDT > I have an application in which all my OS resources (semaphores, > watchdogs, queues) are created/initialized and all my tasks are > spawned by one function. I'd like to use CrossWind to debug one of my > tasks while the others are running, and I'd like to be able to get > control of the task right after its spawned. But I can't "attach" it > until it exists. How can I do this ??? I tried having the task > suspend itself upon startup, attaching to the suspended task, and then > setting a breakpoint. But I could never get control in CrossWind once > I resumed the task (via the shell). Any ideas ??? > We are still using vxWorks 5.2 with the Centerline C++ Frontend and the gdb, and we solve the problem as follows: 1.) Load your application to the target 2.) start gdb and connect it to the target 3.) list your source code in the gdb list sourcefilename.extension:linenumber 4.) set a (local) breakpoint just behind the beginning of your task entry point. Then gdb tells you the address 0xhexaddress of this breakpoint This breakpoint is in a task scope but while you are not attached to a task, it has no further meaning. it's just an easy way to get the right address. 5.) use this address to set a (global) breakpoint in the target shell b 0xhexaddress 6.) run your application from the shell until it hits the (global) breakpoint. Then the shell tells you a task ID 0xhextaskID 7.) use this task ID in gdb to attach attach 0xhextaskID and you should then have control under gdb 8.) delete the (global) breakpoint in the targets shell. 9.) go on with gdb I think there could be a nicer way in Tornado, I've heard there you can set a global breakpoint in the debugger and therefore you don't need a shell for debugging but the procedure is just the same. Andreas ========================================================================= RWTH-AACHEN WZL Werkzeugmaschinenlabor Dipl.-Ing. Andreas Meylahn Steinbachstr. 53b D-52074 Aachen Tel: +49-241-80 7451 FAX: +49-241-8888 293 e-mail: mey@wzl-wm1.wzl.rwth-aachen.de ========================================================================= From daemon@csg.lbl.gov Thu May 16 04:00:41 1996 From: daemon@csg.lbl.gov Date: Thu May 16 04:00:44 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Thu May 16 04:00:34 PDT 1996 Subject: Re: ENETUNREACH? Subject: Re: Development Tools ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: ENETUNREACH? Date: 3 May 1996 23:07:56 -0700 From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star, Oakland, CA Message-ID: <4mes7s$8np@idiom.com> References: In article , Mike Milde at SRC-Home_Office wrote: > > I have a system where the shell occasionally spits out these >messages that say "ENETUNREACH". Surprisingly, everything appears to >work as it should. > Any idea where these are created? I'm doing some IP socket >stuff, so I assume that this means something like: Ethernet >Unreachable. Any ideas? Anyone know if this is caused by an ICMP >message coming back from the system I am communicating with? ENETUNREACH is an indication that you are trying to send IP packets but routing table does not have corresponding entry for the destination. IP layer is unable to resolve the route, therefore unable to send the packet out. It is possible that a gateway is involved in returning ICMP network unreachable message back to the sender. However, VxWorks won't report that back as an error message. It's more likely that your routing table is incorrect. - -- H.J. Bae http://www.peacefulstar.oakland.ca.us --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Development Tools Date: 8 May 1996 15:42:10 GMT From: bowman@waterloo.hp.com (Don Bowman) Organization: H-P Panacom Div, Waterloo, ON Canada Message-ID: <4mqfci$g82@hppadbk.waterloo.hp.com> References: <9604078314.AA831484111@ist1.istnet.com> jlao@istnet.com wrote: : I am looking for development tool for VxWorks which the target board is MIPS. : So far I have two candidates the StethoScope from Real-Time Innovations, Inc. : and the Insight from the Parasoft. Is there anybody use these tools. : Or is there other tools that provides better error catching capabilities. You can also use (for those on a cheap budget :) gcc with the gct (coverage testing) and bounds checking extensions. Another one we've used is dbmalloc (which since became the commercial product sentinel). We also purchased Hindsight (now owned by integrisoft). info@integrisoft.com It does coverage analysis and some CASE stuff for visualisation. A company called synervision makes some products that do memory checking and stuff for vxworks: http://www.synervision.com - -- bowman@waterloo.hp.com ** 519-883-3019 ** http://www.waterloo.hp.com/~bowman --------------------------- End of New-News digest ********************** From daemon@csg.lbl.gov Fri May 17 04:00:54 1996 From: daemon@csg.lbl.gov Date: Fri May 17 04:00:59 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Fri May 17 04:00:44 PDT 1996 Subject: Re: Native Java VM or bytecode to 68k/i960 etc for embedded system Subject: VxWorks 5.2 and C++ Subject: Re: VxWorks v5.1.1 (tNetTask) Subject: 100 Mbit/s Ethernet with Motorola VME 167 or 177 card? Subject: VxWorks v5.1.1 (tNetTask) Subject: Shared memory layout problem Subject: Re: VxWorks v5.1.1 (tNetTask) ------------------------------------------------------- Newsgroups: comp.lang.java,comp.os.vxworks,comp.realtime Subject: Re: Native Java VM or bytecode to 68k/i960 etc for embedded system Date: 9 May 1996 03:53:41 GMT From: rlk@rational.com (Bob Kitzberger) Organization: Rational Software Corporation Message-ID: <4mrq85$m7j@rational.rational.com> References: <318E6F3C.17F4@gpsys.com> <318F1DE5.A8@lmtas.lmco.com> <319127A1.3F03@oro.net> Followup-To: comp.lang.java,comp.os.vxworks,comp.realtime gozer (gozer@oro.net) wrote: : Java is our language of choice (for religious reasons). ^^^^^^^^^^^^^^^^^^^^^ :-) - -- Bob Kitzberger Rational Software Corporation rlk@rational.com http://www.rational.com http://www.rational.com/pst/products/testmate.html --------------------------- Newsgroups: comp.os.vxworks Subject: VxWorks 5.2 and C++ Date: 16 May 1996 18:16:15 GMT From: Michael Roesler Organization: Lam Research Corporation - (510) 659-0200 Message-ID: <4nfrdf$7v@internal.lamrc.com> Using VxWorks 5.2, I would like to write a driver in C++ for a XYCOM240 digital I/O board. Can one use C++ with VxWorks 5.2? I do not see any references regarding C++ in the VxWorks programmers guide. Where can I get more information regarding using C++ and VxWorks? Thanks, Mike Roesler Lam Research miker@lamrc.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks v5.1.1 (tNetTask) Date: 16 May 1996 22:41:22 GMT From: Michael Levasseur Organization: GDE Systems Inc. Message-ID: <4ngaui$rhb@gde.GDEsystems.COM> References: <31922B6A.160E@jpl.nasa.gov> We saw this problem too. This might not be a dupticate IP address though. We saw this when we had two (or more) boards that were in a VME rack that were both attempting to download the VxWorks Kernel at the same time. Both of the Boards had different IP addresses but were located on the same Subnet. Both of the boards were booting from the front panel Ethernet connector and had a Desta Repeater connected to each. The error wasn't seen if only one of the boards was booted and allowed to complete the VxWorks kernel download. For our current download structure we are downloading the VME System Controller First and then downloading each of the other processors in the VME Rack. Each of the Processors has an IP address and the VME Backplane has a separate Subnet Address. Then again this could just be that you do have two devices with the same IP addresses.... --------------------------- Newsgroups: comp.os.vxworks Subject: 100 Mbit/s Ethernet with Motorola VME 167 or 177 card? Date: Fri, 17 May 1996 00:56:48 GMT From: georg@sgl.ists.ca (Georg Feil) Organization: Institute for Space and Terrestrial Science Message-ID: Sender: news@newshub.ists.ca Hi, Has anybody out there used 100 Mbit/s Ethernet on VME? We have a system that currently uses Motorola MVME 167 CPU cards and regular 10 Mbit/s ethernet. We would like to find a 100 Mbit/s Ethernet interface with VxWorks support that could be added to this system. I understand that we can't expect miraculous data rates from this, we just need the ability to connect to a fast Ethernet network. Wind River has suggested that we switch to a Power PC card with PMC slots such as MVME1600 so that we can use a PMC 100 Mbit network card, but I would like to do it without the hassle of changing architectures, maybe just move to the MVME177 for a little extra crunch. Any suggestions are appreciated, Thanks... Georg Feil Space Geodynamics Laboratory | Email: georg@sgl.ists.ca Institute for Space and Terrestrial Science | Phone: (416) 665-5458 4850 Keele St./North York/Ont/Canada/M3J 3K1 | Fax: (416) 665-1815 - -- Georg Feil Space Geodynamics Laboratory | Internet: georg@sgl.ists.ca Institute for Space and Terrestrial Science | Phone: (416) 665-5458 4850 Keele St./North York/Ont/Canada/M3J 3K1 | Fax: (416) 665-1815 --------------------------- Newsgroups: comp.os.vxworks Subject: VxWorks v5.1.1 (tNetTask) Date: Thu, 09 May 1996 10:29:14 -0700 From: Henry Valtier Organization: Jet Propulsion Laboratory - Pasadena CA Message-ID: <31922B6A.160E@jpl.nasa.gov> I am currently using VxWorks 5.1.1, KERNEL: WIND v2.4, running on a combination of 1 master and several slave MVME167-02A,-02B boards in a common backplane in a VME chassis. At most, 2-3 user application tasks run on each board, besides the VxWorks tasks. The master MVME167 serves as the gateway on a local LAN, while the slave boards communicate over a shared memory network. A SUN Sparc20 functions as the UNIX Host and the VME chassis is the Target H/W. Has anyone ever come across the following message (viewed in a VxWorks shell window): (tNetTask): duplicate IP address!! sent from ethernet address: <..addr..> What does it mean? How critical is this condition and what problem(s) does it cause? Thanks in advance. Henry Valtier Jet Propulsion Laboratory Pasadena, California 91109 Henry.Valtier@jpl.nasa.gov --------------------------- Newsgroups: comp.os.vxworks Subject: Shared memory layout problem Date: 9 May 1996 17:55:48 GMT From: David Perkinson Organization: TRW, Inc. Message-ID: <31923235.41C6@boris.msfc.nasa.gov> I have a problem with shared memory semaphores. When I try an example containing just a few shared queue's and semaphores, everything works fine. When I have large blocks of shared memory as well as about a dozen shared queue's and a dozen shared semaphores, I get problems. SemTake calls on the second CPU board cause the first board to go crazy. It looks like the second CPU is causing the first CPU to jump to some strange place in its code which causes access faults. Has anyone seen this kind of behavior before? The second CPU continues to run but runs out of memory after enough attempts to get the semaphore. =============================================================================== David Perkinson, TRW perkinsond@boris.msfc.nasa.gov =============================================================================== --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks v5.1.1 (tNetTask) Date: Thu, 16 May 1996 11:26:16 GMT From: gadi@hynet.co.il (Gadi Friedman) Organization: TeleOp Communication Industries, Ltd. Message-ID: References: <31922B6A.160E@jpl.nasa.gov> In article <31922B6A.160E@jpl.nasa.gov> Henry Valtier writes: > >(tNetTask): duplicate IP address!! sent from ethernet address: <..addr..> > > >What does it mean? How critical is this condition and what problem(s) >does it cause? > Two ethernet ports have the same IP addresses. Check out your arp tables. Gadi --------------------------- End of New-News digest ********************** From tnh@terma.dk Fri May 17 04:29:38 1996 From: tnh@terma.dk (Thorsten Nordmark Hansen) Date: Fri May 17 04:29:41 PDT 1996 Subject: POSIX named semaphores Have anyone of you tried to use the POSIX named semaphores in VxWorks 5.3? When I try to use them I get a -1 returned from sem_open() and errno is set to EINVAL. I tried to copy example 2.7 from 'Programmers Guide: VxWorks 5.3' directly but with the same result. I have no problems using POSIX unnamed semaphores and POSIX message queues. Thanks for any help, Thorsten Hansen TERMA Elektronik A/S tnh@terma.dk From greg@eng.nmr.varian.com Fri May 17 09:31:49 1996 From: greg@eng.nmr.varian.com (Greg Brissey x6951) Date: Fri May 17 09:32:17 PDT 1996 Subject: Re: 100 Mbit/s Ethernet Georg Feil writes: > > Hi, > > Has anybody out there used 100 Mbit/s Ethernet on VME? > > We have a system that currently uses Motorola MVME 167 CPU cards > and regular 10 Mbit/s ethernet. We would like to find a 100 Mbit/s > Ethernet interface with VxWorks support that could be added to this > system. > > I understand that we can't expect miraculous data rates from this, > we just need the ability to connect to a fast Ethernet network. > Wind River has suggested that we switch to a Power PC card with > PMC slots such as MVME1600 so that we can use a PMC 100 Mbit network > card, but I would like to do it without the hassle of changing > architectures, maybe just move to the MVME177 for a little extra > crunch. > > Any suggestions are appreciated, We use a Motorola MVME162LX, and would also like to try 100 Mbit/s Ethernet without changing architectures. I would also be interested in any suggestions. Greg Brissey Varian, Palo Alto, CA greg.brissey@nmr.varian.com From sniedze%homer.dnet.boeing.com@atc.boeing.com Fri May 17 13:24:06 1996 From: sniedze%homer.dnet.boeing.com@atc.boeing.com Date: Fri May 17 13:24:08 PDT 1996 Subject: 3d0002/3d0001 errno's Question on errno 3d0002/3d0001 ? Anyone know why errno 3d0002 appears for tShell on doing an "i" command on first starting WindShell under Tornado? I have a Sun Ultra running Solaris connected to an VME chassis that has a M68030 single board computer. The monitor crt shows no such error. Also, doing a semTake of a binary semaphore created by a different task results in errno= 3d0001. Thanks. js From vzait@xxcal.com Fri May 17 13:24:24 1996 From: vzait@xxcal.com (Veronika Zaitseva) Date: Fri May 17 13:24:28 PDT 1996 Subject: positions avail XXCAL, INC. XXCAL, INC. is a $25MM Technical Contract Recruiting company, providing contract professionals for the past 20 years. We have offices in Los Angeles, Silicon Valley, and Austin, TX. We recruit and provide contract computer professionals to the DP organizations, and high-technology companies. Currently we are looking for the following assignment: Job Description: Need Software Developers with strong 'C' programming experience in a systems, or embedded environment. Must have some sort of real-time OS experience, ie. VxWorks, or VRTX, or 486 Kernal. Development environment is Sun Sparc, and will cross compile to embedded 486 custom controller. Using Wind River C based on GNU C. Debugging with VxGDB. Duration: 12-18 months Location: Los Angeles, CA Contact: Veronika Zaitseva @XXCAL Los Angeles at 800 879-9225, ext.34, or fax to 310 478-6226, or e-mail to vzait@xxcal.com From daemon@csg.lbl.gov Sat May 18 04:02:51 1996 From: daemon@csg.lbl.gov Date: Sat May 18 04:05:02 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sat May 18 04:01:02 PDT 1996 Subject: Re: Bit-fields and byte/word access on 68K Subject: Re: Native Java VM or bytecode to 68k/i960 etc for embedded system ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Bit-fields and byte/word access on 68K Date: Fri, 10 May 1996 09:56:23 -0400 From: Gonzalo Maldonado Organization: Brookhaven National Laboratory Message-ID: <31934B07.6E6D@bnl.gov> References: <31922C57.5224@cliffy.lmtas.lmco.com> Shash Chatterjee wrote: > > Hi! > > I'm trying to use bit-fields to access registers on a > VMIC VMIVME-3123 A/D board. This board has what it calls "firmware > registers" (essentially, pseudo-registers in the DSP's SRAM), that > are 16-bit registers. The registers can be accessed as 8-bit (ie, > no bus errors) but the data is invalid. > > The problem is that the cc68k compiler uses the 68030 bit-field > instructions (BFINS, BFCLR, etc), but these seem to work with > byte accesses to the registers. So is there a way to force the compiler > to use 16-bit accesses (still using C bit-fields) and do logical > ANDs/ORs as necessary (I understand that this will be less efficient > than the bit-field instructions, but WORKING is more efficient than > NOT-WORKING ;-) > > Thanks for any help, > Shash > -- Well, why not try declaring a pointer of type short to access these registers and do your logical operations right on the read values. { INT16 *pWhassup; INT16 value; *pWhassup = (INT16 *)0xFFFF0000; value = *pWhassup; value = value & SOME_MASK; *pWhassup = value; } If you compare the assembly instructions generated you will see that this is definitely more efficient than using bit fields. __________ | _ ____\ * * * * * | |\ \\\\\ * * * * * | |\\ \ * * * * * | |__\ \ | ____\ \ zal@bnl.gov Gonzalo Maldonado | |\\\\\ \ Brookhaven National Laboratory |_________\\ Bldg. 911-C Upton, NY 11973-5000 \\\\\\\\\\\ http://www.rhichome.bnl.gov/People/zal/ --------------------------- Newsgroups: comp.lang.java,comp.os.vxworks,comp.realtime Subject: Re: Native Java VM or bytecode to 68k/i960 etc for embedded system Date: Thu, 09 May 1996 17:44:55 -0600 From: "Adam T. McClure" Organization: Colorado Space Grant College Message-ID: <31928377.24A5@colorado.edu> References: <318E6F3C.17F4@gpsys.com> <4mqi0v$s15@venice.dh.trw.com> Steve Witt wrote: > > > I have been hunting around for several days for > > a Java development environment suitable for a > > realtime embedded system, preferably a 68k or i960-based > > platform with a well-supported realtime operating system. > > For performance reasons we'd like something that compiles > > java bytecode into native instructions. > > > After extensive searching I have come up completely empty-handed. > > Can anyone provide me with pointers to something along these > > lines, or point me in the direction of someone who is working on > > something like this? > > > Maybe I'm barking up the wrong tree, but this seems like a very > > reasonable thing to want to do. > You're right! It's extremely reasonable. I may even know the right people to get it done. I could make a better argument if folks who are interested in this could give solid reasons why the Java language is so great for embedded systems. Email please. Adam McClure mcclurea@colorado.edu --------------------------- End of New-News digest ********************** From daemon@csg.lbl.gov Sun May 19 04:01:15 1996 From: daemon@csg.lbl.gov Date: Sun May 19 04:01:20 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sun May 19 04:01:10 PDT 1996 Subject: A couple of questions about Windview Subject: Re: Literature on preemptive, multitasking design? Subject: Re: Problem debugging PPC Tornado Beta Subject: Re: Error message from tExcTask ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: A couple of questions about Windview Date: Fri, 10 May 1996 08:20:46 GMT From: Michael Migdol Organization: Matsushita El. Works, Ltd., Central Research Lab Message-ID: <3192FC5E.793A@crl.mew.co.jp> Sender: news@isc.mew.co.jp (USENET News System) Hi all, We recently started using WindView, and are finding it fairly useful. Does anyone know the answer to the following questions? 1) Is there any way to remove particular tasks from the graph display? How about removing them from the graph display altogether? 2) Is the data format of the WindView files open? I would like to be able to create spreadsheet-type files containing portions of the event data. Or is this a way to force us to buy Stethoscope? Thanks in advance, Michael - -- =================================================================== Michael A. Migdol $B%^%$%1%k!&%_%0%I%k(J Matsushita Electric Works, Ltd. Tel. (06) 906 - 8075 Central Research Laboratory Fax (06) 904 - 7104 1048 Kadoma, Osaka 571 Japan e-mail mmigdol@crl.mew.co.jp $B!!!!>>2 References: <4mdrcm$2s5@maverick.tad.eds.com> In <4mdrcm$2s5@maverick.tad.eds.com> Dave Mack writes: > >Hi everyone, > >I have 2+ years experience in C programming, but I am >a new VxWorks user and have not done any real-time, >multitasking, priority based programming before. I have >read and understand the concepts presented in the VxWorks >Programmers Guide, but was hoping someone out there could >suggest some material that does more than just define >functionality. I am looking for good sources that talk about >program design concepts, structure charts, etc., when dealing >with multitasking programs. > >Your suggestion is greatly appreciated! > >David Mack >lnustoc.jzgmwk@eds.com >810-524-8807 >810-528-4264(FAX) > > You may look at this book, I think it gives a good overview, it is not an in depth tutorial on the subject though. It addresses real-time issues: Phillip A. Laplante, "Real-Time Systems Design And Analysis", IEEE Press, ISBN 0-7803-0402-0. Istvan http://www.mcs.com/~agg/ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Problem debugging PPC Tornado Beta Date: 17 May 1996 21:57:10 GMT From: nigels@synoptics.com (Nigel Standing) Organization: Bay Networks, Inc. Message-ID: <4nisnm$5fg@inn.corpwest.baynetworks.com> References: <4nb17p$j4@fountain.mindlink.net> Philip.Wiebe@newnes.com (Philip J. Wiebe) writes: > I've been having difficulty debugging PowerPC code on a MVME1604 > I've tried both the -g and -gdwarf compiler options for generating > debug information. A couple of points. Firstly, I believe that although the WRS documentation does say to use "-gdwarf" it is broken in the beta release (it produces *very* large images). Secondly, if I recall correctly, one major difference between a PPC stack dump and (say) the traditional 68k one is that the PPC stack frame may not show any arguments. This is because they are being passed in volatile registers (r3 thru r12). Once a function starts executing it is free to trash any of the regs. In addition, the same regs are used to pass arguments to subsequent function calls. All this means that at each stack level the arguments might very well not exist anymore. So you might be getting some weirdness with your stack trace because of this. Nigels - - -- **************************************************************************** Nigel Standing Vox: +1 408 764 7542 Platforms Group Fax: +1 408 988 6653 Bay Networks (SynOptics) Inc. E-m: nigels@BayNetworks.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Error message from tExcTask Date: 10 May 96 17:43:46 GMT From: bss@wrs.com (Benjamin Skrainka) Organization: Wind River Systems, Inc. Message-ID: References: Greetings! agil@shita.news.qualcomm.com (Amit Gil) writes: >We are getting the following error message: >-> 0xfde244 (tExcTask): 1 messages from interrupt level lost. >0xfde244 (tExcTask): 1 messages from interrupt level lost. [snip] This error message means that excJobAdd() is failing because the message queue, read by tExcTask, is full. excJobAdd() is used by ISRs to perform task level work by sending a request to tExcTask. excJobAdd() is implemented as a wrapper around msgQSend(). If msgQSend() fails, then excJobAdd() bumps the count of messages lost. When tExcTask next runs it will print out the error messages you see. Consequently, you are either queuing up too much work from interrupt time or tExcTask is not running often enough to drain the message queue. >We don't know what is the cause to the error or if it has any influence on the stsem. We don't see any lost messages or any strange behavior exepct this message. >What we do is use send a message from an ISR through a pipe to a task that sends a messages thruogh a UDP socket. The ISR is called from a watch-dog. The pipe is not full when the ISR writes to it. The pipe write method is probably responsible for the calls to excJobAdd(): the pipeWrite() function calls selWakeupAll() to provide support for select(), and selWakeupAll() performs an excJobAdd() if it is at interrupt time. Try making your watch dog slower: if the problem goes away, you know that pipeWrite() is the culprit. A longer term solution would be to use message queues instead of pipes: message queues will always be faster, because pipes are built on message queues. Good luck, Ben ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Benjamin S. Skrainka Wind River Systems Development Engineer 1010 Atlantic Avenue bss@wrs.com Alameda, CA 94501-1147 voice 510.814.2147 http://www.wrs.com/ facsimile 510.814.2164 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ --------------------------- End of New-News digest ********************** From daemon@csg.lbl.gov Mon May 20 04:00:27 1996 From: daemon@csg.lbl.gov Date: Mon May 20 04:00:31 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Mon May 20 04:00:21 PDT 1996 Subject: FORCE10 DOS FILE SYSTEM PROBLEMS Subject: Problems migrating from VxWorks 5.1 -> 5.2 Subject: 3U CPU suggestions Subject: SCSI device size ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: FORCE10 DOS FILE SYSTEM PROBLEMS Date: Sat, 11 May 1996 16:01:42 GMT From: DuaneAustin@worldnet.att.net (Duane C. Austin) Organization: AT&T WorldNet Services Message-ID: <4n3o4v$l58@mtinsc01-mgt.ops.worldnet.att.net> I am developing an application that runs on a FORCE10 using VxWorks 5.2 using a DOS file system. Occasionally I'm experiencing occassional disk file corruptions. If you have experienced and solved such a problem I would appreicate hearing from you. THANKS!!! DuaneAustin@worldnet.att.net --------------------------- Newsgroups: comp.os.vxworks Subject: Problems migrating from VxWorks 5.1 -> 5.2 Date: Fri, 17 May 1996 00:08:35 GMT From: pgonza@atl.com (Pablo Gonzalez-Gandolfi) Organization: Advanced Technology Laboratories, Bothell, WA Message-ID: Sender: pgonza@ironmt.atl.com I am having trouble migrating one of our applications from vxworks 5.1 to vxworks 5.2. The application makes heavy use of the vxworks directory structure and FAT table. Under vxworks 5.1, the application runs fine (at least does not display any abnormal behavior). Under 5.2, it dies, sometimes currupting the FAT table. I can get it to die every time. Without getting into the guts of the application, I was wondering if anybody out there is aware of any known porting issues between 5.1 and 5.2 that are not in the release notes. In particular, changes in the FAT table, file manipulation (copy, rename, etc). If you've had any interesting problems porting 5.1->5.2, I would be interested in that as well. Any help will be greatly appreciated. - -pgg - -- ************************************************************** Pablo Gonzalez-Gandolfi Email: pgonza@atl.com Principal Software Engineer Advanced Technology Lab. Voice: (206)487-8139 22100 Bothell-Everett Hwy. Fax: (206)486-5220 Bothell, WA 98041-3003 ************************************************************* --------------------------- Newsgroups: comp.os.vxworks Subject: 3U CPU suggestions Date: 16 May 1996 18:15:45 GMT From: msinger@griffon.oe.fau.edu (Matthew Singer) Organization: Florida Atlantic University Message-ID: <4nfrch$t75@cybernet.cse.fau.edu> Reply-To: msinger@griffon.oe.fau.edu (Matthew Singer) Sender: msinger@griffon (Matthew Singer) Anyone know of a 3U 68030 or better card that has on card or daughter card fast scsi with dma? - -- - ------------------------------------------------------------------------------- Matthew R. Singer msinger@oe.fau.edu Systems Engineer Voice: 407-367-2947 Florida Atlantic University Fax : 407-367-3885 Dept of Ocean Engineering - ------------------------------------------------------------------------------- --------------------------- Newsgroups: comp.os.vxworks Subject: SCSI device size Date: Thu, 16 May 1996 13:21:28 -0500 From: Alan Currie Organization: NASA/Johnson Space Center Message-ID: <319B7228.77CB@ghgcorp.com> Reply-To: currie@ghgcorp.com I had a problem taking a larger SCSI drive (500meg) and getting the OS to recognize that it was bigger than the original (128meg) SCSI drive. Turns out the problem was that I was writing all my files to the root directory. The root directory has a limitation of 112 files that I was not aware of until I found the dosFsConfigShow command. Solution create subdirectories and write files to these subdirectories. My open was creating an error return of -1, not because disk was full, but because number of files was exceeded. Does anyone know how to get a more verbose indication of what the problem is when the only return is a -1? Thanks. --------------------------- End of New-News digest ********************** From SMTP_Bridge_Server@msmail3.hac.com Mon May 20 05:42:48 1996 From: "SMTP Bridge Server" Date: Mon May 20 05:42:50 PDT 1996 Subject: Undeliverable Mail Unknown Microsoft mail form. Approximate representation follows. Message: comp.os.vxworks newsdigest Sent: Sun, May 19, 1996 5:40 AM To: Quiroz, Gabriel On Server: 631 Segment2 Date: Mon, May 20, 1996 5:42 AM Reason: Could not be delivered because the destination Microsoft Mail server could not be found. From YF.Siu@eng.efi.com Mon May 20 15:01:26 1996 From: "Yuet Fung Siu" Date: Mon May 20 15:01:29 PDT 1996 Subject: Re: comp.os.vxworks newsdigest >5.2 using a DOS file system. Occasionally I'm experiencing >occassional disk file corruptions. If you have experienced and solved >and FAT table. Under vxworks 5.1, the application runs fine (at least does not >display any abnormal behavior). Under 5.2, it dies, sometimes currupting the >FAT table. I can get it to die every time. I think this is not my first time replying to the same problem. 5.2 introduced a new bug to the dosFsLib that will corrupt the file system and hang the task occasionally. I've reported this WRS will probably be able to give you a patch for this since this is no workaround. Please talk to your sale person or a more technical WRS person. YF From haifang@dtc.hp.com Mon May 20 17:36:55 1996 From: Hai-Fang Yun Date: Mon May 20 17:36:58 PDT 1996 Subject: running PSIM in Tornado environment Has anyone tried PSIM (a Power PC simulator) or other Power PC simulator in Tornado environment? Need some pointers on how to set up the simulator and run it. Thanks in advance, hyun@dtc.hp.com From daemon@csg.lbl.gov Tue May 21 04:00:26 1996 From: daemon@csg.lbl.gov Date: Tue May 21 04:00:33 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Tue May 21 04:00:24 PDT 1996 Subject: 68360 and Tornado Subject: Re: Native Java VM or bytecode to 68k/i960 etc for embedded system ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: 68360 and Tornado Date: 21 May 1996 03:03:45 GMT From: martyii@ix.netcom.com(Martin Sigona ) Organization: Netcom Message-ID: <4nrbqh$9b1@dfw-ixnews9.ix.netcom.com> Has any one used the 68360 in companian mode with the 68360 ? If you have-have you ever tried to use the serial driver that comes with the BSP and use it for SCC3 and 4. In the device driver I am having problems when m68360ResetChannel trys to enable the transmit and receive for SCC3 at the end of the fucntion. It dies. Know this is the same driver I am use on SCC2. I have modified the m68360_CHAN for initialazition. Please email me back at martin_sigona@globalvillage. --------------------------- Newsgroups: comp.lang.java,comp.os.vxworks,comp.realtime Subject: Re: Native Java VM or bytecode to 68k/i960 etc for embedded system Date: Mon, 20 May 1996 18:18:03 -0500 From: Steve Lee Organization: Iowa State University Message-ID: <31A0FDAB.205D@iastate.edu> References: <318E6F3C.17F4@gpsys.com> <4mqi0v$s15@venice.dh.trw.com> <31928377.24A5@colorado.edu> <4nqn73$9b9@sjx-ixn2.ix.netcom.com> Reply-To: sjlee@iastate.edu G. David Atkins II wrote: > > I know people that want to do a similar thing for > embedded PC based real-time control. So if you could > please add me to your discussions. > You can subscribe to the mailing list real-time-java@iastate.edu Also, check out http://www.newmonics.com - -- Steve Lee Computer Engineering/Computer Science Undergraduate Student Iowa State University email -> sjlee@iastate.edu WWW -> http://www.cs.iastate.edu/~sjlee/homepage.html --------------------------- End of New-News digest ********************** From daemon@csg.lbl.gov Wed May 22 04:00:20 1996 From: daemon@csg.lbl.gov Date: Wed May 22 04:00:23 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Wed May 22 04:00:17 PDT 1996 Subject: httpd for vxWorks? Subject: VXworks works, really? Subject: Sun 1E boards Subject: Debugger Problems with Tornado 1.0 ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: httpd for vxWorks? Date: Thu, 16 May 1996 10:39:28 -0700 From: Shaun Case Organization: Graham-Patten Systems, Inc. Message-ID: <319B6850.56F1@gpsys.com> Has anyone ported or implemented an httpd to vxWorks? --------------------------- Newsgroups: comp.os.vxworks Subject: VXworks works, really? Date: 21 May 1996 16:51:23 GMT From: consign@ix.netcom.com(Consign, Inc. ) Organization: Netcom Message-ID: <4nssab$dpb@sjx-ixn5.ix.netcom.com> Why choose VXworks if a 486 running the QNX OS or kernel equals or outperforms VXworks on a 266Mhx Dec Alpha? Inquiring minds want to know. --------------------------- Newsgroups: comp.os.vxworks Subject: Sun 1E boards Date: Thu, 16 May 96 22:06:19 PDT From: hwest@csc.com Organization: Computer Sciences Corporation Message-ID: <4ngnej$c8e@explorer.csc.com> Greetings We are using sun 1e boards from Force running VxWorks 5.2 . We would like to expand the board memory using an S bus expansion board. Has anyone had any success doing this? Has anyone had any experience using this board in this way? Thanks in advance for any help you can give. Regards Harold West Computer Sciences Corporation Hanover MD --------------------------- Newsgroups: comp.os.vxworks Subject: Debugger Problems with Tornado 1.0 Date: 21 May 1996 14:03:59 GMT From: Rick Evans Organization: NeoSoft Internet Services +1 713 968 5800 Message-ID: <4nsigf$7e8@uuneo.neosoft.com> I am using Tornado version 1.0 on Windows 95. We are having following problems with the debugger. 1. When we attach to a task that is loaded with symbols we can only get an assembly listing. 2. We can not control the source search path for the debugger through the IDE. We have to set it with the gcc commands. Is their someone else out their using Tornado on Windows 95? If so, are you having these types of problems? --------------------------- End of New-News digest ********************** From Mike.Milde.at.SRC-Home_Office@scires.com Wed May 22 06:07:19 1996 From: Mike Milde at SRC-Home_Office Date: Wed May 22 06:07:21 PDT 1996 Subject: Re: VXworks works, really? Because your application may very well require something other than a 80486 or a 266 Mhz Dec Alpha. Mike Milde Scientific Research Corporation mike@pegasus.scires.com -------------------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: VXworks works, really? Date: 21 May 1996 16:51:23 GMT From: consign@ix.netcom.com(Consign, Inc. ) Organization: Netcom Message-ID: <4nssab$dpb@sjx-ixn5.ix.netcom.com> Why choose VXworks if a 486 running the QNX OS or kernel equals or outperforms VXworks on a 266Mhx Dec Alpha? Inquiring minds want to know. From sblachma@aoc.nrao.edu Wed May 22 09:14:48 1996 From: Steve Blachman Date: Wed May 22 09:14:50 PDT 1996 Subject: Re: VXworks works, really? > Why choose VXworks if a 486 running the QNX OS or kernel equals or > outperforms VXworks on a 266Mhx Dec Alpha? > Inquiring minds want to know. At best, this is an poor question - it begs itself. There is no information here with which to make a decision and the answer is contained therin. If these are the only two conteders and "ourperforms" covers all considerations the answer is obvious. At worst, it is rude. Steve Blachman From luke@wrs.com Wed May 22 11:48:59 1996 From: Luke Teyssier Date: Wed May 22 11:49:01 PDT 1996 Subject: i960 and Loader Format Classes? Hello, Does anyone know of good clases in the San Francisco Bay Area for the i960 Architecture family? I'm looking for a fairly intensive class on assembly language details and idioms, comparison of the Cx, Jx, Hx, Sx, Kx, Details of interrupt and Fault handling, register allocation and similar topics, all from a low-level, intensive software perspective. Also, does anyone have a pointer to good classes on loader formats such as elf, coff, ecoff, and related formats? Thanks for your help, Luke -- ------------------------------------------------------------------------------ | Luke Teyssier email: luke@wrs.com | | Wind River Systems http://www.wrs.com | | 1010 Atlantic Avenue | | Alameda, CA 94501 | |----------------------------------------------------------------------------- From MikeMcQuade@prometheus.connectsoft.net Wed May 22 16:39:18 1996 From: Mike McQuade Date: Wed May 22 16:39:24 PDT 1996 Subject: how do you allow a TCP socket to re-connect ? In my application, my VxWorks box is the TCP socket server, Everything works good the first time the Client connects. But if my client crashes, then Im at a loss as to how I can re-establish the socket without rebooting the VxWorks target. Id like to be able to re-establish the socket fairly quick, ie < one minute. Does anyone know how to make TCP re-connect ? or have any sample code ? Thanks. MikeMcQuade@connect.com From ng@netstal.ch Wed May 22 23:26:15 1996 From: ng@netstal.ch (Niklaus Giger) Date: Wed May 22 23:26:17 PDT 1996 Subject: Re: how do you allow a TCP socket to re-connect ? >Submitted-by: Mike McQuade > > >In my application, my VxWorks box is the TCP socket server, Everything works >good the first time the Client connects. But if my client crashes, then Im >at a loss as to how I can re-establish the socket without rebooting the >VxWorks target. > >Id like to be able to re-establish the socket fairly quick, ie < one minute. > >Does anyone know how to make TCP re-connect ? >or have any sample code ? > Here is our sample code: acceptSockFd = socket (AF_INET, SOCK_STREAM, 0); if (acceptSockFd == ERROR) { printf("\ncouldn't create socket for accepting"); } /* allow bind to socket, which is associated with a "zombie" protocol control block */ optval = 1; rc = setsockopt (acceptSockFd, SOL_SOCKET, SO_REUSEADDR, (char *)&optval, sizeof (optval)); You find the answer also in the FAQ of the vxWorks Users Group. Regards Niklaus Giger Abteilung TEI Netstal Maschinen AG CH-8752 Naefels (Switzerland) Tel: +41 (0) 55/ 618 64 68 Fax: +41 (0) 55/ 612 42 41 Email: ng@netstal.ch From daemon@csg.lbl.gov Thu May 23 04:00:24 1996 From: daemon@csg.lbl.gov Date: Thu May 23 04:00:27 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Thu May 23 04:00:20 PDT 1996 Subject: FDDI for custom 386 board ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: FDDI for custom 386 board Date: Wed, 22 May 1996 15:00:04 GMT From: pete@imc4.eej.ulst.ac.uk (Pete Brett) Message-ID: <832777280.23132.0@hathsys.demon.co.uk> I am developing a distributed data aquisition system using a 386EX running vxWorks in each node. The system will be expected to operate in a noisy environment and so several custommers will demand fibre for LAN communications. The LAN will have two distinct types of traffic:- 1) File transfer: file sizes up to a few MB soft real time constraints. 2) Short command packets with firm real time constraints (8 - 16 ms for a message to be sent, processed and a reply returned) and timing information. Any system using CSMA / CD is by nature not deterministic, but is very available and fairly cheap. Has anyone used a chipset for a token passing LAN (preferably FDDI)? If so could you reply by posting or E-mail (pete@imc4.eej.ulst.ac.uk) I need to know manufacturers and distributers of chipsets, ball-park prices and availability of vxWorks drivers. Many thanks in advance, Pete --------------------------- End of New-News digest ********************** From leonid@rst.co.il Thu May 23 05:04:11 1996 From: leonid@rst.co.il (Leonid Rosenboim) Date: Thu May 23 05:04:15 PDT 1996 Subject: Re: VXworks works, really? > Why choose VXworks if a 486 running the QNX OS or kernel equals > or outperforms VXworks on a 266Mhx Dec Alpha? I do not know how well VxWorks on Alpha performs, but VxWorks surely outperforms QNX on a i486. When comaring two OS's, compare them on the same hardware. From another prespective, it is probably true that VxWorks on Alpha is a lot slower then on i486 in terms of interrupt response and context switch. This is true for all RISC vs. CISC processors. RISC provide much more computing power, with slower R-T response times. Also, according to a study by Venture Development Corp. which covers OS shipped on VMEbus based systems in 1994: VxWorks was the most popular with 15.8%, second came UNIX System V with 14.9%, third OS9 with 13.9%, ... and QNX appears last in the chart with only 1.1%. Of course, the VMEbus is not the entire market, and QNX is a lot more popular on embedded PCs, but this may after all change the mind of some skeptics. ----------------------------------------------------------------------- Leonid Rosenboim Phone: +972-3-559-8144 R S T Software Industries Ltd. Fax: +972-3-559-8244 3 Hashikma st. Industrial Zone, E-Mail: leonid@rst.co.il P.O.Box 11502, AZUR 58017, Israel WindRiver Distributor WWW: http://knight.trendline.co.il/~leonid/ From naivar@telomere.lanl.gov Thu May 23 07:14:29 1996 From: Mark Naivar Date: Thu May 23 07:14:31 PDT 1996 Subject: Re: how do you allow a TCP socket to re-connect ? Mike, Here's some code I use to reconnect automatically when the socket connection goes down. Hope this is what you're after. It's a code fragment I ripped out of some code we're using so hopefully all the important details are there. Let me know. -Mark /*************************************************************************** * * */ #include #include #include #include #include #include /********************************************************************* * NAME * sysSockCreate * * DESCRIPTION * This routine creates a socket, waits for it to be * established, then returns the server and client * socket handles. * * RETURNS * OK or ERROR */ STATUS sysSockCreate(u_short Port, int *serverSocket, int *clientSocket) { struct sockaddr serverSockAddr; struct sockaddr clientSockAddr; int addrLen=sizeof(serverSockAddr); u_short port; port = Port; /******************************************** * Create the socket. */ if( (*serverSocket=socket(AF_INET, SOCK_STREAM, 0)) == -1) { logMsg("ERROR: socket unsuccessful (0x%x)\n", errnoGet()); return(ERROR); } /******************************************** * Clear out the structure. */ bzero(&serverSockAddr, addrLen); /******************************************** * Specify the address family and the * port number. */ serverSockAddr.sa_family=AF_INET; bcopy(&port, serverSockAddr.sa_data, sizeof(port)); /******************************************** * Associate the socket with the address. */ if(bind(*serverSocket, &serverSockAddr, addrLen) == -1) { logMsg( "ERROR: bind unsuccessful (0x%x)\n", errnoGet()); close(*serverSocket); return(ERROR); } /******************************************** * Allow the socket to take connections. */ if(listen(*serverSocket, 1) == -1) { logMsg( "ERROR: listen unsuccessful (0x%x)\n", errnoGet()); close(*serverSocket); return(ERROR); } logMsg("Port %d(SYS) available.\n", port); /******************************************** * Accept the connection request when * one is received. */ return(*clientSocket=accept(*serverSocket, &clientSockAddr, &addrLen)); } /**************************************************************** * * NAME * sysTask * * SYNOPSIS * status = sysTask(); * * DESCRIPTION * This program opens a socket for communication to a host * computer. It listens for commands to arrive over the * socket and processes the commands. * * Commands are defined in the ntbcyproto.h file. * * */ STATUS sysTask(void) { int serverSocket; int clientSocket; int numsend, numrcv, totrcv; int SysPort = 4000; sysCmd *cp; char cBuf[sizeof(sysCmd)]; sysStat *sp; char sBuf[sizeof(sysStat)]; /************************************************************* * Create the socket, wait for a connection to be made. */ if(sysSockCreate(SysPort, &serverSocket, &clientSocket) != ERROR) logMsg("SYS connection established.\n"); else return(ERROR); cp = (sysCmd *)cBuf; sp = (sysStat *)sBuf; /******************************************************* * MAIN LOOP ******************************************************* * Wait for a command, then process it. ******************************************************* * MAIN LOOP ******************************************************* */ while(1) { totrcv = 0; do { numrcv=recv(clientSocket, &cBuf[totrcv], sizeof(sysCmd)-totrcv, 0); if(numrcv <= 0) { /****************************************** * Connection temporarily disconnected, * attempt to re-establish. */ logMsg("CMD connection terminated(%d).\n", numrcv); if(shutdown(clientSocket, 2) != OK) logMsg("Shutdown client: 0x%x\n", errnoGet()); if(close(clientSocket) != OK) logMsg("**ERROR closing client! (0x%x)\n", errnoGet()); if(shutdown(serverSocket, 2) != OK) logMsg("Shutdown server: 0x%x\n", errnoGet()); if(close(serverSocket) != OK) logMsg("**ERROR closing server! (0x%x)\n", errnoGet()); do taskDelay(50); while(sysSockCreate(SysPort, &serverSocket, &clientSocket) == ERROR); logMsg("SYS connection established.\n"); totrcv = 0; } else totrcv += numrcv; } while(totrcv < sizeof(sysCmd)); /************************************************** * NOTE: We should use the network order * conversion routines, but VxWorks * does not have any, and 68K CPUs * are in network byte order. ************************************************** */ /********************************* * Process the command. */ switch(cp->opcode) { default: sp->result = ERROR; numsend=send(clientSocket, sBuf, sizeof(sysStat), 0); break; } } /* while(1) */ } ----------------------------------------------------------------- | Mark Naivar | | | Los Alamos National Lab | (505) 665-2034 (office) | | LS-5, MS M888 | (505) 665-4254 (lab1) | | P.O. Box 1663 | (505) 665-6738 (lab2) | | Los Alamos, NM 87545 | (505) 667-1321 (FAX) | ----------------------------------------------------------------- From del@rst.com Thu May 23 11:22:51 1996 From: del@rst.com (Mark Del Giorno) Date: Thu May 23 11:22:54 PDT 1996 Subject: Redirecting printf()'s Heyyyyyyyyyyyyyyyyyyyy, I'm using MVME162's in an embedded application. I'm looking for a simple way of redirecting all printf()'s so I can save all output text in a battery-backed buffer until I can retrieve the robotic vehicle and get to the information - rather than having the printf()'s go to the serial port and have all those bits spilling out the 25-pin D-sub. Thanks in advance for any advice. _/_/_/_/ _/_/_/_/ _/_/_/_/_/ Mark Del Giorno (del@rst.com) _/ _/ _/ _/ (410) 876-9200 Fax: 876-9470 _/_/_/_/ _/_/_/_/ _/ Robotic Systems Technology _/ _/ _/ _/ 1110 Business Parkway South _/ _/ _/_/_/_/ _/ Westminster, MD 21157 From stan@rti.com Thu May 23 12:20:18 1996 From: Stan Schneider Date: Thu May 23 12:20:20 PDT 1996 Subject: Re: Redirecting printf()'s >> Heyyyyyyyyyyyyyyyyyyyy, >> >> I'm using MVME162's in an embedded application. I'm looking for >> a simple way of redirecting all printf()'s so I can save all >> output text in a battery-backed buffer until I can retrieve the >> robotic vehicle and get to the information - rather than having the >> printf()'s go to the serial port and have all those bits spilling >> out the 25-pin D-sub. >> I think you can just use open a file, and then use ioTaskStdSet on fd 1 (&2)... -- Stan stan@rti.com From gverdun@fibertek.com Thu May 23 14:09:54 1996 From: "Gustavo H. Verdun" Date: Thu May 23 14:09:56 PDT 1996 Subject: Has anyone use Radstone's VME PPC603? If you have any advice on this CPU I would appreciate it. Software will be written in C and Ada (Green Hills) but not together... Are there any gotcha's or other quirks that anyone may have encountered? Thanks, Gus From uchenick@bud.tate.com Thu May 23 14:11:18 1996 From: Gordon Uchenick Date: Thu May 23 14:11:20 PDT 1996 Subject: Re: Redirecting printf()'s > I'm using MVME162's in an embedded application. I'm looking for > a simple way of redirecting all printf()'s so I can save all > output text in a battery-backed buffer until I can retrieve the > robotic vehicle and get to the information - rather than having the > printf()'s go to the serial port and have all those bits spilling > out the 25-pin D-sub. I've designed something like this. The battery backed ram is generated as a DOS file system. If it's just one task doing the prints, you can use ioTaskStdSet(). A dirtier trick might be to close fd 1 for stdout or 2 for stderr and then reopen them to files in your battery backed buffer. HTH, Gordon uchenick@tate.com From 75104.3460@compuserve.com Thu May 23 20:21:01 1996 From: David Parkerson <75104.3460@compuserve.com> Date: Thu May 23 20:21:03 PDT 1996 Subject: Re: 100 Mbit/s Ethernet Hi Guys, The natural replacment for a 162 is to go to PowerPC!. The VI Computer now a part of THEMIS COMPUTER has a 604 with 100B-T standard. Call 510-252-0870 for more info. David Parkerson Themis Regional Manager From Scott.Peters@agai.com Fri May 24 11:56:28 1996 From: Scott Peters Date: Fri May 24 11:56:30 PDT 1996 Subject: subscribe subscribe scott.peters@agai.com begin 600 WINMAIL.DAT M>)\^(@$2`0:0" `$```````!``$``0>0!@`(````Y 0```````#H``$(@ <` M& ```$E032Y-:6-R;W-O9G0@36%I;"Y.;W1E`#$(`0F `0`)````.#9$,$4P M.#,`P@$!(( #``X```#,!P4`& `+`#H`.@`%`'0!`06 `P`.````S <%`!@` M"P`Z`#H`!0!T`0$-@ 0``@````(``@`!!( !``H```!S=6)S8W)I8F4`P@,! M`Y &`(P!```2````0 `Y`/ )=!"C2;L!'@`($ $````=````4U5"4T-224)% M4T-/5%10151%4E- 04=!24-/30`````#`"8```````L`(P``````"P`I```` M```#`#8``````!X`< `!````"@```'-U8G-CL"@P!0-P+S!K0"@S(#Q0(`8VC!"L!S970P( <3`H#^?0J ",\) MV0* "H0+-Q(R(0'0('-U8@3R8F4O"J,*A0J !/!O`D N<$,3( 20 Date: Fri May 24 13:37:29 PDT 1996 Subject: X Windows for VxWorks Greetings, Does anyone know of any third party products that support X Windows on VxWorks? Does anyone have any experiences they would like to share on these products? Thanks in advance, Bill _________________________________________________________________ Bill Zingler Senior Software Engineer Northrup Grumman Corp. Oceanic Division/D.W. Complex P.O. Box1488/MS-9RDW Annapolis, MD 21404 Email: zinglebj@ocd231.bwi.wec.com Phone: 410-260-5965 Fax: 410-260-5916 "The road of excess leads to the palace of wisdom." - William Blake _________________________________________________________________ From craig@hicom.hitachi.com Fri May 24 16:25:58 1996 From: craig@spd.hitachi.com (Craig Humfrey) Date: Fri May 24 16:26:00 PDT 1996 Subject: DOS File System public-ware? I'm seeking public domain versions of a DOS file system to integrate into a Flash memory application under VxWorks version 5.2 The ucar site depends on the dosFsLib.[ch] files, and the dosFsLib.c file is not part of the standard distribution. Does anyone know where such software might be available? Craig Humphreys Pacific Design Engineering craig@pacengr.com From guthrie@power.amasd.anatcp.rockwell.com Fri May 24 16:59:02 1996 From: guthrie@power.amasd.anatcp.rockwell.com (Bob Guthrie) Date: Fri May 24 16:59:04 PDT 1996 Subject: VxWorks boot EPROMS MVME167 One of my co-workers is having troubles with the VxWorks boot EPROMs he received from Wind River on his MVME167. EPROMs 120ns MVME167 33Mhz Using an emulator he can see that there are bit errors when the CPU reads the EPROM. It does not appear to be a simple big endian/little endian byte order type problem. The Motorola debugger program proms work fine. Anybody seen this? Robert Guthrie e-mail: guthrie@amasd.anatcp.rockwell.com From seaweed!seaweed.com!bob@netcom.com Fri May 24 17:20:27 1996 From: bob schulman Date: Fri May 24 17:20:30 PDT 1996 Subject: Re: X Windows for VxWorks > Does anyone know of any third party products that support X Windows on > VxWorks? Does anyone have any experiences they would like to share on > these products? Seaweed Systems has been shipping X11R6 and Motif 2.0 products for the vxWorks environment for over 2 years now. We supply both client-side and server-side solutions in either object and source form. Please contact me for more information. Bob Schulman bob@seaweed.com +1 510.482-3575 From jhillman@wrs.com Fri May 24 19:07:38 1996 From: Jon Hillman Date: Fri May 24 19:07:40 PDT 1996 Subject: Re: X Windows for VxWorks At 01:37 PM 5/24/96 PDT, you wrote: >Submitted-by ZINGLEBJ@ocd231.bwi.wec.com Fri May 24 13:37:28 1996 >Submitted-by: "Zingler, Bill J." > > >Greetings, > >Does anyone know of any third party products that support X Windows on >VxWorks? Does anyone have any experiences they would like to share on >these products? > >Thanks in advance, > >Bill >_________________________________________________________________ > >Bill Zingler >Senior Software Engineer > >Northrup Grumman Corp. >Oceanic Division/D.W. Complex >P.O. Box1488/MS-9RDW >Annapolis, MD 21404 > >Email: zinglebj@ocd231.bwi.wec.com >Phone: 410-260-5965 >Fax: 410-260-5916 > Bill, contact your local Wind River Systems sales rep for information on Visicom VX-Windows products. WRS resells these products. _________________________ | Jon Hillman, FAE | | Wind River Systems | /) (407) 273-4646 (\ / ) Email:jhillman@wrs.com ( \ _( (|_________________________) ) /> (((\ \) /,) / ) / //))/ (\\\\ \_/ / \ \_/ ///// \ / \ / \ _/ \_ / ----/ /----------------------\ \---- / / \ \ From daemon@csg.lbl.gov Sat May 25 04:01:11 1996 From: daemon@csg.lbl.gov Date: Sat May 25 04:01:14 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sat May 25 04:01:06 PDT 1996 Subject: Re: Debugger Problems with Tornado 1.0 Subject: Re: how do you allow a TCP socket to re-connect ? Subject: How to get ifIndex from WindNet SNMP Subject: 64MB and 128MB board availabe for MVME167 Subject: Re: VXworks works, really? Subject: Re: VXworks works, really? Subject: Re: VXworks works, really? Subject: Re: Redirecting printf()'s Subject: Motorola MVME147 available Subject: ATM cards with VxWorks support? Subject: JOB: VA Real-time Embedded Engineer with VxWorks exp. Subject: Re: Has anyone use Radstone's VME PPC603? Subject: Re: Debugger Problems with Tornado 1.0 Subject: Re: VXworks works, really? Subject: Re: VXworks works, really? Subject: Re: VXworks works, really? Subject: Re: Has anyone use Radstone's VME PPC603? ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Debugger Problems with Tornado 1.0 Date: 23 May 1996 11:44:11 GMT From: leif@tsl.uu.se (Leif Thuresson) Organization: The Svedberg Lab Message-ID: <4o1j2b$37cc@columba.udac.uu.se> References: <4nsigf$7e8@uuneo.neosoft.com> In article <4nsigf$7e8@uuneo.neosoft.com>, rick@scimisys.com says... > >I am using Tornado version 1.0 on Windows 95. We are having >following problems with the debugger. > >1. When we attach to a task that is loaded with symbols we >can only get an assembly listing. > >2. We can not control the source search path for the debugger >through the IDE. We have to set it with the gcc commands. > >Is their someone else out their using Tornado on Windows 95? >If so, are you having these types of problems? > > > I'm using Tornado with Sun Solaris and I have a similar problem. I have the kernel compiled with the target resident shell. If I load a module from the target shell (in a startup script) I can't make CrossWind see the symbols from that module. I thought the CrossWind command "add-symbol-file" (Tornado users guide p 236) was suppose to fix this but I only get an error message. Cannot find object module on target: LOADER_OBJ_MODULE_NOT_FOUND I have been in contact with WRS support about this but so far they have only suggested that I should use windsh -s to load the startup script. This isn't a very attractive solution in my case, since it means I then have to have the windsh running all the time for the vxWorks system to boot up and initialise correctly. I don't know if it's possible to run CrossWind over RDB the way it was done before Tornado ? I've noticed the INCLUDE_RDB option in the configuration tool but I haven't tried it yet. Any one have a solution to this ? /Leif - ------------------------------------------------------------------------ Leif Thuresson, The Svedberg Laboratory, Uppsala University, Sweden leif.thuresson@tsl.uu.se --------------------------- Newsgroups: comp.os.vxworks Subject: Re: how do you allow a TCP socket to re-connect ? Date: 23 May 96 16:11:03 -0500 From: "Stephen Tether" Organization: FERMILAB, Batavia, IL Message-ID: References: <199605222339.QAA18985@prometheus.connectsoft.net> - --Cyberdog-AltBoundary-00EE8A8C Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable >In my application, my VxWorks box is the TCP socket server, Everything works >good the first time the Client connects. But if my client crashes, then Im >at a loss as to how I can re-establish the socket without rebooting the >VxWorks target. > >Id like to be able to re-establish the socket fairly quick, ie < one minute. > > >Does anyone know how to make TCP re-connect ? > >or have any sample code ? > > >Thanks. > > >MikeMcQuade@connect.com > In brief, if you've organized the server code properly you don't need to do a thing. At the start, the server task should have been blocked in accept(), waiting for a connection. The first argument to accept() is the file descriptor of the original, or parent, socket. When a client makes a connection, accept() creates a brand new socket, transfers the connection to it, and returns its file descriptor. You can then spawn a new task to handle the connection (or allocate a pre-spawned task from a pool for real speed) and pass it the new fd. The original server task can now go back to accept(), still using the parent socket, which is now back in the same state as it was before the client connected. If the client crashes, the task handling the connection should close the the socket, using the fd returned by accept(), NOT THE ORIGINAL. Even if this task bombs and gets suspended, or is unable to detect the client crash, the original server task and its socket are unaffected. If graceful termination is impossible, you will have to reboot the server once in a while due to the accumulation of dead tasks and/or sockets. - - Steve Tether - --------------------------------------------------- This message was created and sent using the Cyberdog Mail System - --------------------------------------------------- - --Cyberdog-AltBoundary-00EE8A8C Content-Type: multipart/mixed; boundary="Cyberdog-MixedBoundary-00EE8A8C" Content-Transfer-Encoding: 7bit - --Cyberdog-MixedBoundary-00EE8A8C Content-Type: text/enriched; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable 9Monaco>In my application, my VxWorks box is the TCP socket server, Everything works >good the first time the Client connects. But if my client crashes, then Im >at a loss as to how I can re-establish the socket without rebooting the >VxWorks target. > >Id like to be able to re-establish the socket fairly quick, ie << one minute. > > >Does anyone know how to make TCP re-connect ? > >or have any sample code ? > > >Thanks. > > 10Geneva>9MonacoMikeMcQuade@connect.com 10Geneva> In brief, if you've organized the server code properly you don't need to do a thing. At the start, the server task should have been blocked in accept(), waiting for a connection. The first argument to accept() is the file descriptor of the original, or parent, socket. When a client makes a connection, accept() creates a brand new socket, transfers the connection to it, and returns its file descriptor. You can then spawn a new task to handle the connection (or allocate a pre-spawned task from a pool for real speed) and pass it the new fd. The original server task can now go back to accept(), still using the parent socket, which is now back in the same state as it was before the client connected. If the client crashes, the task handling the connection should close the the socket, using the fd returned by accept(), NOT THE ORIGINAL. Even if this task bombs and gets suspended, or is unable to detect the client crash, the original server task and its socket are unaffected. If graceful termination is impossible, you will have to reboot the server once in a while due to the accumulation of dead tasks and/or sockets. - - Steve Tether - --------------------------------------------------- This message was created and sent using the Cyberdog Mail System - --------------------------------------------------- - --Cyberdog-MixedBoundary-00EE8A8C-- - --Cyberdog-AltBoundary-00EE8A8C-- --------------------------- Newsgroups: comp.os.vxworks Subject: How to get ifIndex from WindNet SNMP Date: Thu, 23 May 1996 16:36:50 -0700 From: Nick Gandin Organization: Genral Instrument Message-ID: <31A4F692.5C5F@gi.com> I am constructing my private WindNet SNMP MIB table that I would like to have a key as ifIndex from the standard SNMP MIBII. The question is how do I get ifIndex values in my code? Do they have any API function for this? Any hints are appreciated. Nick Gandin General Instrument (619)455 1500 x.3511 ngandin@gi.com --------------------------- Newsgroups: comp.os.vxworks Subject: 64MB and 128MB board availabe for MVME167 Date: 24 May 1996 01:16:23 GMT From: innov8@ix.netcom.com(M.W.) Organization: Netcom Message-ID: <4o32l7$359@sjx-ixn2.ix.netcom.com> I have a 64MB and 128MB board available for the Motorola MVME167 card. I also have a set of VXworks proms available. E-mail if interested. - -- E-mail: innov8@ix.netcom.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VXworks works, really? Date: Thu, 23 May 96 14:36:10 GMT From: danh@qnx.com (Dan Hildebrand) Organization: QNX Software Systems Message-ID: <4jtr23g6@qnx.com> References: <4nssab$dpb@sjx-ixn5.ix.netcom.com> In article , Eric Rabinowitz wrote: >Consign, Inc. (consign@ix.netcom.com) wrote: >: Why choose VXworks if a 486 running the QNX OS or kernel equals or >: outperforms VXworks on a 266Mhx Dec Alpha? > >: Inquiring minds want to know. > >Interesting that you didn't sign you message. >Apparently QNX is starting to feel some pressure from Tornado. To the best of our knowledge, this message did not originate from QNX Software Systems. We would agree that this sort of posting isn't appropriate. >Well there are a few reasons. >I'm not sure where you got your information but according to the numbers >published by QNX, VxWorks performs significantly better then QNX on that same >66MHz 486 DX2. > >Other differences are many but include: >VxWorks is a cross-development environment, QNX is native. >There are certainly arguments for each so I won't go into it. Actually, you can develop for QNX from Windows, OS/2 or NT as well. >However this among other things makes it much easier to port VxWorks to an >entirely different processor architecture since most of the development tools >don't need to be ported to the real-time target. This is why VxWorks is >available on the most popular processor for embedded designs (68k being the >most popular, then PowerPC, Intel 80960, and 80x86 not nec. in that order) >and other more specialized chips like CPU32, Mips, SPARC, Hitachi, Siemens, >and others. Our emphasis on x86 runtime targets is deliberate, and allows us to focus on supporting virtually every device associated with x86 machines (embedded or otherwise), rather than putting our efforts into supporting more CPU families. Cross vs native development tool suites are not the reason for our x86 emphasis. >VxWorks minimum executable size is much smaller than QNX. >QNX claims to be able to scale down to as little as 256K, >VxWorks minimum is on the order of 20K (CISC). Our next generation product is in beta. :-) >Anyway for your reference: > >QNX 486 DX2 66MHz published results (QNX 4.21 Specifications) > >interrupt latency: 7 microseconds >average context switching: 6 microseconds (user-level only) > >VxWorks 486 DX2 66MHz measured results (VxWorks 5.2) > >maximum interrupt latency: < 1.0 microseconds >average context-switch: 1.9 microseconds (kernel and user-level) > >In any case the two environments are very different. True - a noteworthy difference is that QNX is a POSIX-compliant OS and runs user processes in separate, MMU-protected address spaces. - -- Dan Hildebrand (danh@qnx.com) QNX Software Systems, Ltd. http://www.qnx.com/~danh 175 Terence Matthews phone: (613) 591-0931 (voice) Kanata, Ontario, Canada (613) 591-3579 (fax) K2M 1W8 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VXworks works, really? Date: Fri, 24 May 1996 03:38:14 GMT From: elr@netcom.com (Eric Rabinowitz) Organization: NETCOM On-line Communication Services (408 261-4700 guest) Message-ID: Sender: elr@netcom.netcom.com you wrote: > Why choose VXworks if a 486 running the QNX OS or kernel equals or > outperforms VXworks on a 266Mhx Dec Alpha? > > Inquiring minds want to know. Hmmmmmmmmmm. Well there are a few reasons. I'm not sure where you got your information but according to the numbers published by QNX, VxWorks performs significantly better then QNX on that same 66MHz 486 DX2. Other differences are many but include: VxWorks is a cross-development environment, QNX is native. There are certainly arguments for each so I won't go into it. However this among other things makes it much easier to port VxWorks to an entirely different architecture since most of the development tools don't need to be ported to the real-time target. This is why VxWorks is available on the most popular processor for embedded designs (68k being the most popular, then PowerPC, Intel 80960, and 80x86 not nec. in that order) and other more specialized chips like CPU32, Mips, SPARC, Hitachi, Siemens, and some others that I can't recall. VxWorks' minimum executable size is much smaller than QNX. QNX claims to be able to scale down to as little as 256K, VxWorks minimum is on the order of 20K (CISC). QNX has no visualization tools equivalent to WindView or Stethoscope. QNX seems to have more PC-typical devices drivers, whereas VxWorks has more embedded-type device drivers. If fact VxWorks and QNX don't seem to compete too much. Although I know of one VxWorks developer who liked both VxWorks and QNX, but chose VxWorks simply because they wanted to be able to use the Intel 80960 on future designs, although at the time they were leaning toward using the 80386/80486. QNX has many happy customers who are doing native development on Intel 80x86 and don't ever intend to change to another processor architecture or O/S. VxWorks has die-hard fans as well. Anyway for your reference: QNX 486 DX2 66MHz published results (QNX 4.21 Specifications) interrupt latency: 7 microseconds average context switching: 6 microseconds (user-level only) average task response time: not published maximum task response time: not published VxWorks 486 DX2 66MHz measured results (VxWorks 5.2) maximum interrupt latency: < 1.0 microseconds average context-switch: 1.9 microseconds (kernel and user-level) average task response time: 15 microseconds maximum task response time: 25 usec In any case the two environments are very different. - -e P.S. It's curious that you didn't sign your post, I wonder if that means that QNX is feeling some pressure from Tornado?? - -- Eric Linn Rabinowitz Panoramic Systems CA 408-289-7757 elr@netcom.com WA 206-689-1707 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VXworks works, really? Date: 24 May 1996 01:23:52 GMT From: marc@wrs.com (Marc Shepard) Organization: Wind River Systems, Inc. Message-ID: <4o3338$ft5@darya.wrs.com> References: <4nssab$dpb@sjx-ixn5.ix.netcom.com> Dude, > Why choose VXworks if a 486 running the QNX OS or kernel equals or > outperforms VXworks on a 266Mhx Dec Alpha? This statement makes no sense at all. The real question is the performance of VxWorks vs QNX when both run on the 486. If you switch OS's and platforms, then the result is likely to depend only on the bechmark code and not on the OS's. For example, context switch time on the 486 is probably faster because the register set is smaller, regardless of which OS is running on the 486. VxWorks has several huge advantages over QNX: 1) The development tools are much better, so you finish writing and debugging your application faster. 2) It runs on a lot of different CPU's, so if your next project is better suited to a different processor, you can reuse your previous VxWorks code. - -Marc - -- ``` marc@wrs.com (o o) (510)814-2142 - -----------------oOO--(_)--OOo-------------------- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Redirecting printf()'s Date: 24 May 1996 12:24:00 GMT From: johncoby@ghgcorp.com (JOHN R. COBARRUVIAS) Organization: Personal Account Message-ID: References: <199605232112.RAA05066@bud.tate.com> In article <199605232112.RAA05066@bud.tate.com>, Gordon Uchenick wrote: > > I'm using MVME162's in an embedded application. I'm looking for > > a simple way of redirecting all printf()'s so I can save all > > output text in a battery-backed buffer until I can retrieve the > > robotic vehicle and get to the information - rather than having the > > printf()'s go to the serial port and have all those bits spilling > > out the 25-pin D-sub. > > I've designed something like this. The battery backed ram is generated > as a DOS file system. If it's just one task doing the prints, you can > use ioTaskStdSet(). A dirtier trick might be to close fd 1 for stdout > or 2 for stderr and then reopen them to files in your battery backed buffer. > Or try using ioGlobalStdSet(). this will reset your standard in/out to a different device. Which could be your dos file system in nvram. > HTH, > > Gordon > uchenick@tate.com - -- ,-_/ . .-,--. ,---. . ' | ,-. |-. ,-. `|__/ | -' ,-. |-. ,-. ,-. ,-. . . . , . ,-. ,-. | | | | | | | | \ ,. | -. | | | | ,-| | | | | | / | ,-| `-. | `-' ' ' ' ' `-' ` `' `---' `-' `-' `-^ ' ' `-' `' ' `-^ `-' / | John R. Cobarruvias johncoby@ghgcorp.com Houston, Tx `--' --------------------------- Newsgroups: comp.os.vxworks Subject: Motorola MVME147 available Date: 21 May 1996 16:33:59 GMT From: innov8@ix.netcom.com(M.W.) Organization: Netcom Message-ID: <4nsr9n$9ko@dfw-ixnews9.ix.netcom.com> I have some Motorola MVME147 VME cards available. E-mail if interested. - -- E-mail: innov8@ix.netcom.com --------------------------- Newsgroups: comp.os.vxworks Subject: ATM cards with VxWorks support? Date: 24 May 1996 14:08:42 GMT From: newell@rd.bbc.co.uk (Chris Newell) Organization: British Broadcasting Corporation, UK Message-ID: <4o4fta$t5@bbcnews.rd.bbc.co.uk> Does anyone know of an ATM Network Interface Card with VwWorks support? I would be very interested to hear from anyone who has experience of ATM (at any speed and with any host running VxWorks) Chris Newell chris.newell@rd.bbc.co.uk - --------------------------------------------------------------------------- The opinions expressed here are my own and not necessarily those of the BBC --------------------------- Newsgroups: comp.os.vxworks Subject: JOB: VA Real-time Embedded Engineer with VxWorks exp. Date: 24 May 1996 18:09:08 GMT From: The Consortium Organization: InternetMCI Message-ID: <4o4u04$lq4@news.internetmci.com> We are looking for several BSEE's with a real-time software engineering background for both permanent and consulting opportunities in Northern VA. Ideal candidate will have experience with Motorolla's 68040 chip, C, and the Vxworks operating system. If you would like to make a career change or want more information about this position contact Mike Thron at: THE CONSORTIUM FAX (301) 652-2330 PH (301) 986-1412 x3018 E-Mail Consortium@internetmci.com http://www.consortiuminc.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Has anyone use Radstone's VME PPC603? Date: 24 May 96 14:09:21 -0500 From: "Stephen Tether" Organization: FERMILAB, Batavia, IL Message-ID: References: One thing I forgot: you should apply the Grande patch to your VxWorks system library (ignore the parts for other BSPs). This fixes a lot of annoying problems common to all boards, such as a tendency to hang during loading of the symbol table and unreliable interrupt handling. Send a message to support@wrs.com to find out how to get and apply the patch. - - Steve - --------------------------------------------------- This message was created and sent using the Cyberdog Mail System - --------------------------------------------------- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Debugger Problems with Tornado 1.0 Date: Fri, 24 May 1996 12:26:44 +0200 From: John Neumoege Organization: IPK-Berlin, Dep. of Control Engineering Message-ID: <31A58EE4.7A46@ipk.fhg.de> References: <4nsigf$7e8@uuneo.neosoft.com> Rick Evans wrote: > > I am using Tornado version 1.0 on Windows 95. We are having > following problems with the debugger. > > 1. When we attach to a task that is loaded with symbols we > can only get an assembly listing. > > 2. We can not control the source search path for the debugger > through the IDE. We have to set it with the gcc commands. > > Is their someone else out their using Tornado on Windows 95? > If so, are you having these types of problems? > > Hi Rick! We are using Tornado 1.0 with vxWorks 5.3 and we are also having similiar problems. Greetings. - --John - -- - -------------------John Neumoege------------------------ Fraunhofer-IPK e-mail: John.Neumoege@ipk.fhg.de Pascalstr. 8-9 Voice: ++49/(0)30/390 06-127 D-10587 Berlin FAX: ++49/(0)30/391 10 37 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VXworks works, really? Date: 24 May 1996 09:05:15 -0700 From: mwette@mr-ed.jpl.nasa.gov (Matt Wette) Organization: Jet Propulsion Laboratory Message-ID: <4o4mnr$4ei@mr-ed.jpl.nasa.gov> References: <4nssab$dpb@sjx-ixn5.ix.netcom.com> <4jtr23g6@qnx.com> In article <4jtr23g6@qnx.com>, Dan Hildebrand wrote: >In article , Eric Rabinowitz wrote: >>Anyway for your reference: >> >>QNX 486 DX2 66MHz published results (QNX 4.21 Specifications) >> >>interrupt latency: 7 microseconds >>average context switching: 6 microseconds (user-level only) >> >>VxWorks 486 DX2 66MHz measured results (VxWorks 5.2) >> >>maximum interrupt latency: < 1.0 microseconds >>average context-switch: 1.9 microseconds (kernel and user-level) >> >>In any case the two environments are very different. > >True - a noteworthy difference is that QNX is a POSIX-compliant OS and runs >user processes in separate, MMU-protected address spaces. The QNX web site (www.qnx.com) now advertises that QNX 4.23 Proc32 does a context switch in, can you believe it, 2.6 milliseconds! Matt - -- matthew.r.wette@jpl.nasa.gov --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VXworks works, really? Date: Fri, 24 May 96 20:42:46 GMT From: danh@qnx.com (Dan Hildebrand) Organization: QNX Software Systems Message-ID: References: <9605231151.AA04387@rst.co.il> In article <9605231151.AA04387@rst.co.il>, Leonid Rosenboim wrote: >> Why choose VXworks if a 486 running the QNX OS or kernel equals >> or outperforms VXworks on a 266Mhx Dec Alpha? > >I do not know how well VxWorks on Alpha performs, but VxWorks surely >outperforms QNX on a i486. When comaring two OS's, compare them >on the same hardware. Since VxWorks is not a protected-mode OS, it does not need to manipulate the MMU on a context switch, so it has less work to do. The question is whether or not MMU protection is valuable to the realtime system, and if it is, how much overhead is worth incurring to have it. >Also, according to a study by Venture Development Corp. which covers >OS shipped on VMEbus based systems in 1994: > VxWorks was the most popular with 15.8%, > second came UNIX System V with 14.9%, third OS9 with 13.9%, ... and > QNX appears last in the chart with only 1.1%. Since QNX is an x86-only OS, and VME is a predominately non-x86 bus architecture, the fact that QNX has a 1.1% market share on VME is quite surprising. On x86 platforms (embedded or otherwise) Emerging Technologies Group ranked QNX as the leading realtime OS. - -- Dan Hildebrand (danh@qnx.com) QNX Software Systems, Ltd. http://www.qnx.com/~danh 175 Terence Matthews phone: (613) 591-0931 (voice) Kanata, Ontario, Canada (613) 591-3579 (fax) K2M 1W8 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VXworks works, really? Date: 24 May 1996 12:26:05 GMT From: johncoby@ghgcorp.com (JOHN R. COBARRUVIAS) Organization: Personal Account Message-ID: References: In article , elr@netcom.com (Eric Rabinowitz) wrote: > you wrote: > > Why choose VXworks if a 486 running the QNX OS or kernel equals or > > outperforms VXworks on a 266Mhx Dec Alpha? > > > > Inquiring minds want to know. Why buy an car when a bike will get you anywhere you need to go? Hmmmmmm. This is not as easy to answer as it is to ask. > > Hmmmmmmmmmm. > > Well there are a few reasons. > I'm not sure where you got your information but according to the numbers > published by QNX, VxWorks performs significantly better then QNX on that same > 66MHz 486 DX2. > > Other differences are many but include: > VxWorks is a cross-development environment, QNX is native. > There are certainly arguments for each so I won't go into it. > > However this among other things makes it much easier to port VxWorks to an > entirely different architecture since most of the development tools don't > need to be ported to the real-time target. This is why VxWorks is available > on the most popular processor for embedded designs (68k being the most popular, > then PowerPC, Intel 80960, and 80x86 not nec. in that order) and other more > specialized chips like CPU32, Mips, SPARC, Hitachi, Siemens, and some others > that I can't recall. > > VxWorks' minimum executable size is much smaller than QNX. > QNX claims to be able to scale down to as little as 256K, > VxWorks minimum is on the order of 20K (CISC). > > QNX has no visualization tools equivalent to WindView or Stethoscope. > > QNX seems to have more PC-typical devices drivers, whereas VxWorks > has more embedded-type device drivers. > > If fact VxWorks and QNX don't seem to compete too much. > Although I know of one VxWorks developer who liked both VxWorks and QNX, but > chose VxWorks simply because they wanted to be able to use the Intel 80960 > on future designs, although at the time they were leaning toward using > the 80386/80486. > > QNX has many happy customers who are doing native development on Intel 80x86 > and don't ever intend to change to another processor architecture or O/S. > > VxWorks has die-hard fans as well. > > > Anyway for your reference: > > QNX 486 DX2 66MHz published results (QNX 4.21 Specifications) > > interrupt latency: 7 microseconds > average context switching: 6 microseconds (user-level only) > average task response time: not published > maximum task response time: not published > > VxWorks 486 DX2 66MHz measured results (VxWorks 5.2) > > maximum interrupt latency: < 1.0 microseconds > average context-switch: 1.9 microseconds (kernel and user-level) > average task response time: 15 microseconds > maximum task response time: 25 usec > > In any case the two environments are very different. > > -e > > P.S. It's curious that you didn't sign your post, I wonder if that means that > QNX is feeling some pressure from Tornado?? > -- > > Eric Linn Rabinowitz > Panoramic Systems > CA 408-289-7757 elr@netcom.com > WA 206-689-1707 - -- ,-_/ . .-,--. ,---. . ' | ,-. |-. ,-. `|__/ | -' ,-. |-. ,-. ,-. ,-. . . . , . ,-. ,-. | | | | | | | | \ ,. | -. | | | | ,-| | | | | | / | ,-| `-. | `-' ' ' ' ' `-' ` `' `---' `-' `-' `-^ ' ' `-' `' ' `-^ `-' / | John R. Cobarruvias johncoby@ghgcorp.com Houston, Tx `--' --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Has anyone use Radstone's VME PPC603? Date: 24 May 96 13:35:29 -0500 From: "Stephen Tether" Organization: FERMILAB, Batavia, IL Message-ID: References: - --Cyberdog-AltBoundary-000A89CB Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable >If you have any advice on this CPU I would appreciate it. > >Software will be written in C and Ada (Green Hills) but not together... > >Are there any gotcha's or other quirks that anyone may have encountered? > My group installed five of those about six weeks ago. The first board support package from Radstone was missing all the man pages, so we had to discover by trial and error that the correct network interface for the boot parameters was lnPci; they came preset for ln. Radstone immediately shipped us a correct BSP when we told them about this. Our PPC603's didn't know their own Ethernet addresses. These were supposed to be contained in the non-volatile RAM but there was garbage there instead. We had to make a new boot EEPROM image (bootrom_uncmp) after making sure the preprocessor symbol ETHERNET_ADR_SET was defined in config.h. This enabled a new command "N" (not "n") taking no arguments; it walks you through the process of setting the Ethernet addresses. The lower order bits of the address are on a teeny-tiny sticker on one of the VME connectors. If you're luckier than we were, your boards will have the Ethernet addresses in EEPROM on the network interface; try "n lnPci" at the VxWorks boot prompt and see if it returns a valid address. In vmeUniverse.c, sysVmeIntHandler() calls sysMailboxInt() and sysVmeIntDispatch() before they have been defined or declared.This gives both functions the default return type of int, and our compiler (Diab) objects when it discovers that they return void. I moved sysMailboxInt() in front of sysVmeIntHandler() and added a prototype declaration for sysVmeIntDispatch(). The Green Hills compiler might not detect this error. WRS SPR #6319 states that register p0 is not set properly in sysALib.s before calling usrInit(); it should be set to BOOT_NORMAL (=3D BOOT_WARM_AUTOBOOT). The rs603 BSP also has this problem. The maximum auxiliary clock rate is 2 KHz, as opposed to 5 KHz for the system clock. The BSP doesn't support the time-of-day clock. Make sure that HALF_BUS_SPEED is either defined or undefined in config.h. Check with Radstone to see which you need to do. - --- Steve Tether Research Scientist, MIT Lab. for Nuclear Science tether@lns.mit.edu - --------------------------------------------------- This message was created and sent using the Cyberdog Mail System - --------------------------------------------------- - --Cyberdog-AltBoundary-000A89CB Content-Type: multipart/mixed; boundary="Cyberdog-MixedBoundary-000A89CB" Content-Transfer-Encoding: 7bit - --Cyberdog-MixedBoundary-000A89CB Content-Type: text/enriched; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable 9Monaco>If you have any advice on this CPU I would appreciate it. > >Software will be written in C and Ada (Green Hills) but not together... > 10Geneva>9MonacoAre there any gotcha's or other quirks that anyone may have encountered? 10Geneva> My group installed five of those about six weeks ago. The first board support package from Radstone was missing all the man pages, so we had to discover by trial and error that the correct network interface for the boot parameters was lnPci; they came preset for ln. Radstone immediately shipped us a correct BSP when we told them about this. Our PPC603's didn't know their own Ethernet addresses. These were supposed to be contained in the non-volatile RAM but there was garbage there instead. We had to make a new boot EEPROM image (bootrom_uncmp) after making sure the preprocessor symbol ETHERNET_ADR_SET was defined in config.h. This enabled a new command "N" (not "n") taking no arguments; it walks you through the process of setting the Ethernet addresses. The lower order bits of the address are on a teeny-tiny sticker on one of the VME connectors. If you're luckier than we were, your boards will have the Ethernet addresses in EEPROM on the network interface; try "n lnPci" at the VxWorks boot prompt and see if it returns a valid address. In vmeUniverse.c, sysVmeIntHandler() calls sysMailboxInt() and sysVmeIntDispatch() before they have been defined or declared.This gives both functions the default return type of int, and our compiler (Diab) objects when it discovers that they return void. I moved sysMailboxInt() in front of sysVmeIntHandler() and added a prototype declaration for sysVmeIntDispatch(). The Green Hills compiler might not detect this error. WRS SPR #6319 states that register p0 is not set properly in sysALib.s before calling usrInit(); it should be set to BOOT_NORMAL (=3D BOOT_WARM_AUTOBOOT). The rs603 BSP also has this problem. The maximum auxiliary clock rate is 2 KHz, as opposed to 5 KHz for the system clock. The BSP doesn't support the time-of-day clock. Make sure that HALF_BUS_SPEED is either defined or undefined in config.h. Check with Radstone to see which you need to do. - --- Steve Tether Research Scientist, MIT Lab. for Nuclear Science tether@lns.mit.edu - --------------------------------------------------- This message was created and sent using the Cyberdog Mail System - --------------------------------------------------- - --Cyberdog-MixedBoundary-000A89CB-- - --Cyberdog-AltBoundary-000A89CB-- --------------------------- End of New-News digest ********************** From leonid@rst.co.il Sat May 25 23:57:01 1996 From: leonid@rst.co.il (Leonid Rosenboim) Date: Sat May 25 23:58:00 PDT 1996 Subject: Flash File System > From: craig@spd.hitachi.com (Craig Humfrey) > Date: Fri May 24 16:26:00 PDT 1996 > I'm seeking public domain versions of a DOS file system to integrate > into a Flash memory application under VxWorks version 5.2 Implementation of DOS file systems on top of Flash memories is a common method, and is called FTL - Flash Translation Layer. It is a relatively simple thing to do, but the result is rather poor in terms of Flash edurance and wear leveling. In other words, if a lot of updates arte expected, the Flash memory will be worn out very quickly. Also, the public sources on UCAR require modifications to dosFsLib, which in turn require you to purxhase dosFsLib source code from Wind River. On the other hand, you may consider purchasing FlashLib(tm) - our implementation of the Microsoft Flash File System, which is a LFS (Linear File System), and is a lot more effecient in its use of the Flash memory life span, and it can save a lot of space with it's included compression mechanism. The cost of our Flash File System is roughly equal to one man month worth of engineering, and provides full source code. Also, it costs less then dosFsLib source from Wind River, it's a fully supported product, and usually requires just a day's work to integrate it with your hardware. For more detaisl about Flash Memory technologies see: http://www.systemsoft.com/wpapers/flash.htm For details about our Flash File System, see the URL below. ----------------------------------------------------------------------- Leonid Rosenboim Phone: +972-3-559-8144 R S T Software Industries Ltd. Mobile: +972-50-307-142 P.O.Box 11502, AZUR 58017, Israel Fax: +972-3-559-8244 http://knight.trendline.co.il/~leonid/ E-Mail: leonid@rst.co.il From daemon@csg.lbl.gov Sun May 26 04:00:39 1996 From: daemon@csg.lbl.gov Date: Sun May 26 04:00:41 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sun May 26 04:00:35 PDT 1996 Subject: Re: VXworks works, really? Subject: Re: VXworks works, really? Subject: Re: VXworks works, really? Subject: Re: Problems migrating from VxWorks 5.1 -> 5.2 Subject: Re: VXworks works, really? Subject: Re: How to get ifIndex from WindNet SNMP ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: VXworks works, really? Date: Fri, 24 May 96 17:34:53 GMT From: danh@qnx.com (Dan Hildebrand) Organization: QNX Software Systems Message-ID: References: <4nssab$dpb@sjx-ixn5.ix.netcom.com> <4jtr23g6@qnx.com> <4o4mdh$4cs@mr-ed.jpl.nasa.gov> In article <4o4mdh$4cs@mr-ed.jpl.nasa.gov>, Matt Wette wrote: >In article <4jtr23g6@qnx.com>, Dan Hildebrand wrote: >>True - a noteworthy difference is that QNX is a POSIX-compliant OS and runs >>user processes in separate, MMU-protected address spaces. > >OK, but is QNX multithreaded, providing users the ability to run >multiple threads in the same address space? I didn't get the impression >that this capability exists from gleening over the QNX net literature. >Where's your 1003.1c stuff? The existing product support non-POSIX threads, POSIX threads are in beta. - -- Dan Hildebrand (danh@qnx.com) QNX Software Systems, Ltd. http://www.qnx.com/~danh 175 Terence Matthews phone: (613) 591-0931 (voice) Kanata, Ontario, Canada (613) 591-3579 (fax) K2M 1W8 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VXworks works, really? Date: Tue, 21 May 1996 23:01:54 GMT From: brucew@phoenix.net (Bruce Wedding) Organization: BranPaul Systems Message-ID: <4nti67$1qj@uhura.phoenix.net> References: <4nssab$dpb@sjx-ixn5.ix.netcom.com> In comp.os.vxworks consign@ix.netcom.com(Consign, Inc. ) wrote: >Why choose VXworks if a 486 running the QNX OS or kernel equals or >outperforms VXworks on a 266Mhx Dec Alpha? > >Inquiring minds want to know. Did you come here for a reason? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VXworks works, really? Date: Fri, 24 May 96 18:49:27 GMT From: danh@qnx.com (Dan Hildebrand) Organization: QNX Software Systems Message-ID: <16wr2fan@qnx.com> References: <4nssab$dpb@sjx-ixn5.ix.netcom.com> <4jtr23g6@qnx.com> <4o4mnr$4ei@mr-ed.jpl.nasa.gov> In article <4o4mnr$4ei@mr-ed.jpl.nasa.gov>, Matt Wette wrote: > >The QNX web site (www.qnx.com) now advertises that QNX 4.23 Proc32 >does a context switch in, can you believe it, 2.6 milliseconds! Hey now - you found that on the beta section of our web site! Apparently that web page itself hadn't quite finished beta testing. :-) It now more correctly reads 2.6 microseconds. - -- Dan Hildebrand (danh@qnx.com) QNX Software Systems, Ltd. http://www.qnx.com/~danh 175 Terence Matthews phone: (613) 591-0931 (voice) Kanata, Ontario, Canada (613) 591-3579 (fax) K2M 1W8 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Problems migrating from VxWorks 5.1 -> 5.2 Date: 25 May 1996 02:35:11 GMT From: dog@sunspot.noao.edu (Fritz Stauffer) Organization: National Solar Observatory/SP Message-ID: <4o5rkv$n5o@noao.edu> References: In article pgonza@atl.com (Pablo Gonzalez-Gandolfi) writes: > >I am having trouble migrating one of our applications from vxworks 5.1 to >vxworks 5.2. The application makes heavy use of the vxworks directory structure > ... >Without getting into the guts of the application, I was wondering if anybody >out there is aware of any known porting issues between 5.1 and 5.2 that >are not in the release notes. In particular, changes in the FAT table, file > ... I have intermittant trouble with writing to files using the SCSI library. I can hang in sbicProgBytesIn(). This was found by task tracing the hung task. This started after 5.1.0. fritz stauffer --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VXworks works, really? Date: 24 May 1996 08:59:45 -0700 From: mwette@mr-ed.jpl.nasa.gov (Matt Wette) Organization: Jet Propulsion Laboratory Message-ID: <4o4mdh$4cs@mr-ed.jpl.nasa.gov> References: <4nssab$dpb@sjx-ixn5.ix.netcom.com> <4jtr23g6@qnx.com> In article <4jtr23g6@qnx.com>, Dan Hildebrand wrote: >True - a noteworthy difference is that QNX is a POSIX-compliant OS and runs >user processes in separate, MMU-protected address spaces. OK, but is QNX multithreaded, providing users the ability to run multiple threads in the same address space? I didn't get the impression that this capability exists from gleening over the QNX net literature. Where's your 1003.1c stuff? Matt - -- matthew.r.wette@jpl.nasa.gov --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How to get ifIndex from WindNet SNMP Date: Sun, 26 May 1996 11:29:12 +0300 From: Ron Keidar Organization: Qualcomm Message-ID: <31A81658.6EA@qualcomm.com> References: <31A4F692.5C5F@gi.com> Nick Gandin wrote: > > I am constructing my private WindNet SNMP MIB table > that I would like to have a key as ifIndex from > the standard SNMP MIBII. > > The question is how do I get ifIndex values in my > code? Do they have any API function for this? > > Any hints are appreciated. > > Nick Gandin > General Instrument > (619)455 1500 x.3511 > ngandin@gi.com You may access the 'ifTable' from your code using 'm2IfTblEntryGet()'. Look in WindNet man pages for this function as well as for m2SysLib, m2IfLib, m2IpLib, etc.... Ron Keidar --------------------------- End of New-News digest ********************** From daemon@csg.lbl.gov Mon May 27 04:00:17 1996 From: daemon@csg.lbl.gov Date: Mon May 27 04:00:19 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Mon May 27 04:00:14 PDT 1996 Subject: Measuring idle & interrupt time using Spy? ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Measuring idle & interrupt time using Spy? Date: Mon, 27 May 1996 07:20:54 GMT From: Michael Migdol Organization: Matsushita El. Works, Ltd., Central Research Lab Message-ID: <31A957D6.37DD@crl.mew.co.jp> Sender: news@isc.mew.co.jp (USENET News System) Hi all, I know that Spy basically measures 3 things: # of ticks spent in each task, # of ticks spent in ISR's, and # of ticks spent in idle time. The time spent in each task is actually stored in the tasks' TCB in the taskTicks field and can thus be retrieved externally from Spy. Where are the idle time and ISR time information stored? Has anyone written functions to retrieve that information? I've gone just about crazy looking for it. :-) Thanks in advance, Michael - -- Michael A. Migdol Research Engineer Matsushita Electric Works, Ltd. Tel. 81-6-906-8075 Central Research Laboratory Fax 81-6-904-7104 1048 Kadoma, Osaka 571 Japan e-mail mmigdol@crl.mew.co.jp http://nesc.me.utexas.edu/~mmigdol/ --------------------------- End of New-News digest ********************** From JAMES@lss-chq.mhs.compuserve.com Mon May 27 07:28:07 1996 From: "Hayes, James" Date: Mon May 27 07:28:09 PDT 1996 Subject: Re: VxWorks boot EPROMS MVME167 >One of my co-workers is having troubles with the VxWorks boot EPROMs >he received from Wind River on his MVME167. > > EPROMs 120ns > MVME167 33Mhz > >Using an emulator he can see that there are bit errors when the CPU >reads the EPROM. It does not appear to be a simple big endian/little endian >byte order type problem. The Motorola debugger program proms work >fine. > >Anybody seen this? > Robert Guthrie > e-mail: guthrie@amasd.anatcp.rockwell.com We experienced the same trouble with an Atmel part. 120ns versions did not work... 100ns DO WORK !!! James Legacy Storage Systems Inc. From MikeMcQuade@prometheus.connectsoft.net Mon May 27 10:28:51 1996 From: Mike McQuade Date: Mon May 27 10:28:52 PDT 1996 Subject: Passing Structs from Unix to PPC target. Im having some difficulty passing a struct from my HP-UX host to my PowerPC target. The struct gets down to the target, but in order to read everything out on the PPC side, I had to put in some padding variables (on the PPC). Then once I pad the struct to get everything out, when I try to send it back to the HP, elements are out of place. Can anyone explain whats going on here ? There must be a better way, ? Thanks. Mike McQuade MikeMcQuade@connect.com From daemon@csg.lbl.gov Tue May 28 04:00:22 1996 From: daemon@csg.lbl.gov Date: Tue May 28 04:00:24 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Tue May 28 04:00:18 PDT 1996 Subject: Re: Debugger Problems with Tornado 1.0 Subject: Re: VXworks works, really? Subject: Sweden Subject: Compatibility problem with dos and vxworks file system Subject: Re: routed for VxWorks ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Debugger Problems with Tornado 1.0 Date: 27 May 1996 19:24:13 GMT From: steved@pacifier.com@199.2.117.163 (Steve Doiel) Organization: Pacifier BBS, Vancouver, Wa. ((360) 693-0325) Message-ID: <4ocvgu$oo2@news.pacifier.com> References: <4nsigf$7e8@uuneo.neosoft.com> Reply-To: steved@pacifier.com (Steve Doiel) In <4nsigf$7e8@uuneo.neosoft.com>, Rick Evans writes: >I am using Tornado version 1.0 on Windows 95. We are having >following problems with the debugger. > >1. When we attach to a task that is loaded with symbols we >can only get an assembly listing. > >2. We can not control the source search path for the debugger >through the IDE. We have to set it with the gcc commands. > >Is their someone else out their using Tornado on Windows 95? >If so, are you having these types of problems? > > Yes, we did encounter these problems. After a visit from a WRS FAE we learned that most of our problems were due to case sensitivity. Tornado and the GNU tools are rooted in the UNIX world which is case sensitive. It was a co-worker that encoutered this problem on W95 (I'm using NT.) She learned that apparently all of her problems were due to the inconsistant use of case in file naming. The folks at support@wrs.com should be up on this. steved@pacififer.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VXworks works, really? Date: Wed, 22 May 1996 21:20:22 GMT From: elr@netcom.com (Eric Rabinowitz) Organization: NETCOM On-line Communication Services (408 261-4700 guest) Message-ID: References: <4nssab$dpb@sjx-ixn5.ix.netcom.com> Sender: elr@netcom.netcom.com Consign, Inc. (consign@ix.netcom.com) wrote: : Why choose VXworks if a 486 running the QNX OS or kernel equals or : outperforms VXworks on a 266Mhx Dec Alpha? : Inquiring minds want to know. Interesting that you didn't sign you message. Apparently QNX is starting to feel some pressure from Tornado. Well there are a few reasons. I'm not sure where you got your information but according to the numbers published by QNX, VxWorks performs significantly better then QNX on that same 66MHz 486 DX2. Other differences are many but include: VxWorks is a cross-development environment, QNX is native. There are certainly arguments for each so I won't go into it. However this among other things makes it much easier to port VxWorks to an entirely different processor architecture since most of the development tools don't need to be ported to the real-time target. This is why VxWorks is available on the most popular processor for embedded designs (68k being the most popular, then PowerPC, Intel 80960, and 80x86 not nec. in that order) and other more specialized chips like CPU32, Mips, SPARC, Hitachi, Siemens, and others. VxWorks minimum executable size is much smaller than QNX. QNX claims to be able to scale down to as little as 256K, VxWorks minimum is on the order of 20K (CISC). QNX has no visualization tools equivalent to Windview or Stethoscope. QNX seems to have more PC-typical devices drivers, whereas VxWorks has more embedded-type device drivers. VxWorks and QNX don't seem to compete too much. Although I know of one VxWorks developer who liked both VxWorks and QNX, but chose VxWorks simply because they wanted to be able to use the Intel 80960 on future designs, although at the time they were leaning toward using the 80386/80486. QNX has many happy customers who are doing native development on Intel 80x86 and don't ever intend to change to another processor architecture or O/S. VxWorks has die-hard fans as well. Anyway for your reference: QNX 486 DX2 66MHz published results (QNX 4.21 Specifications) interrupt latency: 7 microseconds average context switching: 6 microseconds (user-level only) average task response time: not published maximum task response time: not published VxWorks 486 DX2 66MHz measured results (VxWorks 5.2) maximum interrupt latency: < 1.0 microseconds average context-switch: 1.9 microseconds (kernel and user-level) average task response time: 15 microseconds maximum task response time: 25 usec In any case the two environments are very different. - -e - -- Eric Linn Rabinowitz Panoramic Systems CA 408-289-7757 elr@netcom.com WA 206-689-1707 --------------------------- Newsgroups: comp.os.vxworks Subject: Sweden Date: Mon, 27 May 1996 16:25:35 -0700 From: Brampton Library Organization: Spar Aerospace Ltd. Message-ID: <31AA39EF.1327@spar.ca> Hello! I am a software engineer with 12 years experience mostly in Real Time Embedded and GUI software development. In my last project I was the project engineer for the development of the software for an 8 Degrees of freedom robot controller. This was developed in C using vxWorks as the real time OS. The Motif Based GUI was developed on a SUN SPARC using XDesigner. I am looking for a software development position in southern Sweden and would appreciate your help greatly (like company names and possible point of contact). I am currently employed but available within a reasonably short period of time. I would also incur all moving expenses. Thank you very much.. Be sure any help is greatly appreciated. yfillion@wchat.on.ca --------------------------- Newsgroups: comp.os.vxworks Subject: Compatibility problem with dos and vxworks file system Date: Tue, 28 May 1996 00:06:15 GMT From: Tony Wu Organization: Hewlett-Packard Co. Message-ID: <31AA4377.13D1@hpato.aus.hp.com> Sender: news@hparc12.aus.hp.com (News Daemon ID) I am trying to format a scsi disk drive (1.2Gbyte) such that a dos platform (DOS6.2) can understand the file format and VxWork's dosFs can also understand it. I've experiment with 1) formating a 200Mbyte disk and installing the file system as the VxWork's dosFs. The PC does not want to recoginize the disk ( the pc does not want to recognize the disk as d: drive.) 2) fdisk the scsi drive under the pc to install the dos partition. The PC recognise the scsi drive as a pc logical drive. VxWorks does not recognise the disk and the dosFsInit failed. Would it work if I offset it by 1 block with scsiBlkDevCreate? Is there anyway to get around this problem? What is my best bet to have a 1.2Gbyte disk that can be understood by vxworks, dos and unix machine? Thanks in advance Tony --------------------------- Newsgroups: comp.os.vxworks Subject: Re: routed for VxWorks Date: 17 May 1996 19:02:09 GMT From: witts@PROBLEM_WITH_INEWS_GATEWAY_FILE (Steve Witt) Organization: Hughes Aircraft Company Message-ID: <4niifh$aa8@venice.dh.trw.com> References: <4ndt2g$2b8@neemit.synopsys.com> Reply-To: sawitt@hac2arpa.hac.com > Is there a routed-equivalent program available for VxWorks? > I am using VxWorks 5.1.1 on a Force 5V, in case it matters. Yes, in the VxWorks archive: ftp://ftp.atd.ucar.edu/pub/vxworks/vx with several files called vxrouted... - -- Steve Witt | mailto:sawitt@hac2arpa.hac.com Hughes Aircraft Company El Segundo, California --------------------------- End of New-News digest ********************** From froeber@bbn.com Tue May 28 05:06:33 1996 From: Fred Roeber Date: Tue May 28 05:06:35 PDT 1996 Subject: Re: comp.os.vxworks newsdigest Michael Migdol asked about spyLib: > Where > are the idle time and ISR time information stored? Has anyone written > functions to retrieve that information? I've gone just about crazy > looking for it. :-) The counters are kept in local variables inside spyLib so aren't easily accessible. The spyLib works by adding a special aux clock interrupt handler that bumps the local counts for idle and interrupt cycles as well as task hooks to bump task counts kept in the TCB. Long ago I tried to do the same thing you are but I had access to the source so could make the variables global. Without source you are sort of out of luck. Good luck anyway. Fred -- | Fred J Roeber, BBN Systems & Technologies | | 50 Enterprise Place Middletown, RI 02842-5202 | | froeber@bbn.com 401-849-2543 (X48) | | See http://www.bbn.com/tv for TraceMaker info | From stan@rti.com Tue May 28 10:58:34 1996 From: Stan Schneider Date: Tue May 28 10:58:36 PDT 1996 Subject: Re: Passing Structs from Unix to PPC target. >> >> Im having some difficulty passing a struct from my HP-UX host to my PowerPC >> target. >> >> The struct gets down to the target, but in order to read everything out on >> the PPC side, I had to put in some padding variables (on the PPC). >> >> Then once I pad the struct to get everything out, when I try to send it back >> to the HP, elements are out of place. >> >> >> Can anyone explain whats going on here ? >> Compilers are free to pack structures as they see fit. >> >> There must be a better way, ? Try XDR. It's easy & relatively cheap between big-endian machines. The "rpcgen" compiler will do the hard work too. See "man rpcgen" on your HP. -- Stan ============================================================================= = = = = Stan Schneider = email: stan@rti.com = = Real-Time Innovations, Inc. = Phone: (408) 720-8312 = = 155A Moffett Park Drive, Suite 111 = Fax: (408) 734-5009 = = Sunnyvale, CA 94089 = http://www.rti.com = = = = ============================================================================= From pardo@rti.com Tue May 28 11:03:45 1996 From: Gerardo Pardo-Castellote Date: Tue May 28 11:03:47 PDT 1996 Subject: Re: Passing Structs from Unix to PPC target. >> >> Im having some difficulty passing a struct from my HP-UX host to my PowerPC >> target. >> >> The struct gets down to the target, but in order to read everything out on >> the PPC side, I had to put in some padding variables (on the PPC). >> >> Then once I pad the struct to get everything out, when I try to send it back >> to the HP, elements are out of place. >> >> Can anyone explain whats going on here ? >> There must be a better way, ? Each processor architecture uses possible different storage sizes (e.g. 32 vs 64 bit pointers and integers), alignment, and even byte ordering (big endian vs. little endian). Both PPC and hppa are big endian so byte ordering isn't the problem in this case. But the others still are. You can either pad your data manually on each processor so the memory representation is the same, or alternatively, use a machine independent data-representation format such as XDR. -Gerardo =========================================================================== = = = = Gerardo Pardo-Castellote = email: pardo@rti.com = = Real-Time Innovations, Inc. = Phone: (408) 720-8312 = = 155A Moffett Park Drive, Suite 111 = Fax: (408) 734-5009 = = Sunnyvale, CA 94089 = http://www.rti.com = = = = =========================================================================== From brett@rti.com Tue May 28 15:33:41 1996 From: Brett Murphy Date: Tue May 28 15:33:43 PDT 1996 Subject: Re: X Windows for VxWorks Greetings, VisiCom has an X Windows product for VxWorks called VX-Windows. We used it on two different x86 targets just a couple of months ago. We were quite pleased with it. It was fast and has been working reliably for us. Check the TradeWinds directory for more info. Their phone number is 1-800-621-8474 -- Brett =============================================================================== = = = = Brett Murphy = email: brett@rti.com = = Real-Time Innovations, Inc. = Phone: (408) 720-8312 = = 155A Moffet Park Drive, Suite 111 = Fax: (408) 734-5009 = = Sunnyvale, CA 94089 = = = = = =============================================================================== > Submitted-by ZINGLEBJ@ocd231.bwi.wec.com Fri May 24 13:37:28 1996 > Submitted-by: "Zingler, Bill J." > > > Greetings, > > Does anyone know of any third party products that support X Windows on > VxWorks? Does anyone have any experiences they would like to share on > these products? > > Thanks in advance, > > Bill > _________________________________________________________________ > > Bill Zingler > Senior Software Engineer > > Northrup Grumman Corp. > Oceanic Division/D.W. Complex > P.O. Box1488/MS-9RDW > Annapolis, MD 21404 > > Email: zinglebj@ocd231.bwi.wec.com > Phone: 410-260-5965 > Fax: 410-260-5916 > From daemon@csg.lbl.gov Wed May 29 04:00:22 1996 From: daemon@csg.lbl.gov Date: Wed May 29 04:00:25 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Wed May 29 04:00:17 PDT 1996 Subject: VxWorks FAQ where? Subject: shared memory configuration Subject: Re: Debugger Problems with Tornado 1.0 Subject: programming the MC68302? Subject: !!! H E L P !!! MC 68060 FP Math Performance Subject: VxWorks -> PCNFS performance Subject: Software Engineer Wanted ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: VxWorks FAQ where? Date: Tue, 28 May 1996 12:05:28 GMT From: wzhang@novice.uwaterloo.ca (W.Z.) Organization: University of Waterloo Message-ID: Sender: news@novice.uwaterloo.ca (Mr. News) Where can I find the VxWorks FAQ ? --------------------------- Newsgroups: comp.os.vxworks Subject: shared memory configuration Date: 28 May 1996 08:24:04 GMT From: mabbih@gi.univ-lr.fr (ABBIH Moustapha) Organization: Universite de La Rochelle Message-ID: <4oed74$otv@hpuniv.univ-lr.fr> helo I have to configure vxworks to support the shared memory objects. The master and the slave boards are mvme162 types. The backplane is a vme bus. The problem is how to determinate the anchor adresse for the slave board?. Or how to configure the vme adresse base for a mvme162 ????. If you have some ideas please help me.. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Debugger Problems with Tornado 1.0 Date: Tue, 28 May 1996 12:42:51 GMT From: nobody@nowhere Organization: Northrop Grumman Corporation Message-ID: References: <4nsigf$7e8@uuneo.neosoft.com> <4ocvgu$oo2@news.pacifier.com> Sender: usenet@tron.bwi.wec.com (USEnet News Poster) In article <4ocvgu$oo2@news.pacifier.com>, steved@pacifier.com (Steve Doiel) wrote: >In <4nsigf$7e8@uuneo.neosoft.com>, Rick Evans writes: >>I am using Tornado version 1.0 on Windows 95. We are having >>following problems with the debugger. >> >>1. When we attach to a task that is loaded with symbols we >>can only get an assembly listing. >> >>2. We can not control the source search path for the debugger >>through the IDE. We have to set it with the gcc commands. >> >>Is their someone else out their using Tornado on Windows 95? >>If so, are you having these types of problems? >> >> > >Yes, we did encounter these problems. After a visit from a WRS FAE we >learned that most of our problems were due to case sensitivity. Tornado and >the GNU tools are rooted in the UNIX world which is case sensitive. > >It was a co-worker that encoutered this problem on W95 (I'm using NT.) She >learned that apparently all of her problems were due to the inconsistant use >of case in file naming. > >The folks at support@wrs.com should be up on this. >steved@pacififer.com > I have also had problems with the debugger. Originally, all of my source and object files were in a single directory and everything worked well. Problems started when we distributed source files in multiple directories and included them in the "SourceSearchPath" dialog box of the debugger. WRS has yet to respond to my detailed description of the problem sent last week. Chris Jones cjones@digsys1.bwi.wec.com Northrop Grumman - ESSD --------------------------- Newsgroups: comp.os.vxworks Subject: programming the MC68302? Date: 23 May 1996 13:43:56 GMT From: johncoby@ghgcorp.com (JOHN R. COBARRUVIAS) Organization: Personal Account Message-ID: I have to program the MC68302 Integrated Multiprotocol Processor, and I don't know what the hell I am doing! Can anyone shed some light, suggest a book to read? Anything? I am floundering! I am using a PEP Modular Computer PVM30 with vxWorks 5.2. - -- ,-_/ . .-,--. ,---. . ' | ,-. |-. ,-. `|__/ | -' ,-. |-. ,-. ,-. ,-. . . . , . ,-. ,-. | | | | | | | | \ ,. | -. | | | | ,-| | | | | | / | ,-| `-. | `-' ' ' ' ' `-' ` `' `---' `-' `-' `-^ ' ' `-' `' ' `-^ `-' / | John R. Cobarruvias johncoby@ghgcorp.com Houston, Tx `--' --------------------------- Newsgroups: comp.os.vxworks Subject: !!! H E L P !!! MC 68060 FP Math Performance Date: 28 May 1996 21:20:21 GMT From: ipss@roanoke.infi.net (ipss) Organization: Industrial Process Support Services Message-ID: <4ofqml$mlq@nw101.infi.net> We are using an MVE177 with 68060 CPU and seeing greatly degraded FP performance when dealing with numbers whose magnitude are less than FLT_MIN. This can cause a fabs() call to go from 1 microsec to 53 microsec! Similar degradation occurs for all FP Math operations. Has anyone else seen this and can suggest a workaround. This is of great concern to us in a high speed (1KHz) application which is math intensive. --------------------------- Newsgroups: comp.os.vxworks Subject: VxWorks -> PCNFS performance Date: 28 May 1996 23:54:22 GMT From: ashrafis@nt.com (Siamak Razzaghe-Ashrafi) Organization: Northern Telecom Inc., Mission Park Message-ID: <4og3ne$2a3@nrchh52.rich.nt.com> Hi I am having some preformace problems with my vxWorks NFS server. I have vxWorks 5.2 on a 68360 with a 170MB IDE Drive. I can transfer 32kbyte/sec from my Sparc station to my vxWorks box by NFS mounting the file system over 10Mb ethernet. Is this about the norm. From a PC client running NFS software I can only get 19Kbyte/sec. Is anyone doing vxWorks server to PC client NFS file transfer? If so what kinds of numbers are you getting? I know my lan is going at lest 5Mb and I was able to copy info from the ram disk to the hard drive at 225kbyte/sec so I know my hard drive is not the bottle neck. Any Ideas. Thankx in advance. ashrafis@nt.com - -- Hi --------------------------- Newsgroups: comp.dsp,comp.software-eng,comp.realtime,comp.unix.programmer,nj.jobs,misc.jobs.offered,comp.os.vxworks Subject: Software Engineer Wanted Date: Tue, 28 May 1996 21:10:35 GMT From: "Terry K. Lim" Organization: Pentek, Inc. Message-ID: <31AB6BCB.41C67EA6@pentek.com> Position Available: Software Engineer Company: Pentek, Inc. COMPANY DESCRIPTION Manufacturer of digital signal processing and data acquisition products for VME and VXI. QUALIFICATIONS Strong C/UNIX programming and experience interfacing with hardware devices is required. Multiple UNIX or real-time OS platform experience a plus. JOB DESCRIPTION Responsibilities include developing and maintaining UNIX and real-time OS drivers for Pentek hardware as well as developing and maintaining support packages that link 3rd party software products to Pentek hardware and software. Contact: Terry Lim Director of Software Engineering Pentek, Inc. 55 Walnut St. Norwood, NJ 07648 e-mail: tkl@pentek.com FAX: (201) 767-7179 - -- Terry Lim | Voice: (201) 767-7100 Director of Software Engineering | FAX: (201) 767-7179 Pentek, Inc. | BBS: (201) 767-8733 --------------------------- End of New-News digest ********************** From FREGEAU@digsys1.bwi.wec.com Wed May 29 08:03:50 1996 From: "JEAN FREGEAU" Date: Wed May 29 08:03:52 PDT 1996 Subject: logMsg shared memory message Q?? I would like to route output from all of the boards in my system to a single output source. Does anyone know of a way to use the logMsg feature with a shared memory message Q? Thanks! Jean Fregeau fregeau@digsys1.bwi.wec.com Northrop Grumman - ESSD From ANDERSDG@f3groups1.fsd.jhuapl.edu Wed May 29 12:53:56 1996 From: "Anderson, Derek G." Date: Wed May 29 12:53:58 PDT 1996 Subject: MVME147 Synchronous serial port driver My version of vxworks only supports asynchronous mode of operation of the on-board Z8530 SCC's. Vxworks says they do not have a driver to support synchronous mode on the MVME147, so I was wondering if anyone has written one. that I could use. I am using port 1,2,and 3 in asynchronous, port 4 in synchronous. Also, has anyone who used both modes of operation on the 147 run into major problems; possibly concerning only one interrupt vector per chip. Thanks, Derek derek.anderson@jhuapl.edu (410)792-6000 x3746 From vince@lassen.rti.com Wed May 29 15:05:03 1996 From: vince@rti.com (Vince Chen) Date: Wed May 29 15:05:46 PDT 1996 Subject: RE: !!! H E L P !!! MC 68060 FP Math Performance > We are using an MVE177 with 68060 CPU and seeing greatly degraded FP > performance when dealing with numbers whose magnitude are less than > FLT_MIN. This can cause a fabs() call to go from 1 microsec to 53 > microsec! Similar degradation occurs for all FP Math operations. > > Has anyone else seen this and can suggest a workaround. This is of > great concern to us in a high speed (1KHz) application which is math > intensive. Yes, you are getting underflow exception errors. You'll have to figure out how to turn that off. This is a problem on 167 boards as well. We had first noticed this when comparing performance of applications running single-precision vs. double-precision calculations. We were surprised to see that the single-precision version ran slower. The reason was that it (a simulated controls system) was doing such a good job that the errors in the system were below FLT_MIN, thereby generating underflow exceptions. It took the double-precision application much longer to hit its underflow limit, so it ran faster before that point. We "fixed" the problem by injection noise on the order of FLT_MIN into the system, since we couldn't disable exception-handling for underflow. -vince P.S. (plug) We figured all this out using our ScopeProfile profiling tool. ========================================================================== = = = = Vincent Chen = email: vince@rti.com = = Real-Time Innovations, Inc. = Phone: (408) 720-8312 = = 155A Moffett Park Drive, Suite 111 = Fax: (408) 734-5009 = = Sunnyvale, CA 94089 = = ========================================================================== From dbruce@targettech.com Wed May 29 15:07:38 1996 From: dbruce@targettech.com (Dave Bruce) Date: Wed May 29 15:08:21 PDT 1996 Subject: http server for VxWorks ---------- X-Sun-Data-Type: text X-Sun-Data-Description: text X-Sun-Data-Name: text X-Sun-Content-Lines: 20 My company, Target Technologies, is currently in the process of developing a http server for VxWorks for some of our internal products. The core functionality is working and we are currently deciding on the paradigm for forms support. I don't know if we have decided if we will be marketing it as a separate product yet, but you could try sending email to info@targettech.com if you are interested. Since I have seen several posts asking about the availibility of a VxWorks based http server, I thought I would post a broadcast response rather than to the individuals. ================================================================ David R. Bruce dbruce@targettech.com Project Engineer http://www.targettech.com Target Technologies, Inc. Real-Time System Integration, (508)657-9940 Products, and Consulting Services =============================================================== ---------- X-Sun-Data-Type: default X-Sun-Data-Description: default X-Sun-Data-Name: httppost X-Sun-Content-Lines: 20 My company, Target Technologies, is currently in the process of developing a http server for VxWorks for some of our internal products. The core functionality is working and we are currently deciding on the paradigm for forms support. I don't know if we have decided if we will be marketing it as a separate product yet, but you could try sending email to info@targettech.com if you are interested. Since I have seen several posts asking about the availibility of a VxWorks based http server, I thought I would post a broadcast response rather than to the individuals. ================================================================ David R. Bruce dbruce@targettech.com Project Engineer http://www.targettech.com Target Technologies, Inc. Real-Time System Integration, (508)657-9940 Products, and Consulting Services =============================================================== From "heiner@hic334.decnet.bosch.de"@hic334.decnet.bosch.de Thu May 30 03:37:45 1996 From: Andreas Heiner <"heiner@hic334.decnet.bosch.de"@hic334.decnet.bosch.de> Date: Thu May 30 03:37:47 PDT 1996 Subject: ISDN for VME and CAPI2.0 Hello guys I am looking for ISDN adapter cards to plug into a VME rack. I want to use it with a CAPI2.0 driver. Is there anybody who had the same problem and can give me some hints from which company/distributor I can buy such VME-ISDN boards and where I can get an ISDN-CAPI2.0 (maybe CAPI1.2 is enough to start) driver for VxWorks? Regards, Andreas Heiner Internet: Andreas.Heiner@fr.bosch.de c/o Robert Bosch GmbH Dept. FV/SLH (Corporate Research and Development Systems 2 / Advanced Development Communications) Robert-Bosch-Str. 200 D-31139 Hildesheim Germany Phone (+49) 5121 493946 FAX (+49) 5121 492520 From daemon@csg.lbl.gov Thu May 30 04:00:30 1996 From: daemon@csg.lbl.gov Date: Thu May 30 04:00:32 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Thu May 30 04:00:27 PDT 1996 Subject: Re: i960 and Loader Format Classes? ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: i960 and Loader Format Classes? Date: 28 May 1996 17:49:12 -0700 From: kla@leland.Stanford.EDU (Earl Mitchell) Organization: Stanford University, CA 94305, USA Message-ID: <4og6u8$3oj@elaine25.Stanford.EDU> References: <199605221848.AA21830@niagra.wrs.com> In article <199605221848.AA21830@niagra.wrs.com>, Luke Teyssier wrote: >Hello, > >Does anyone know of good clases in the San Francisco Bay Area for >the i960 Architecture family? I'm looking for a fairly intensive >class on assembly language details and idioms, comparison of the Cx, >Jx, Hx, Sx, Kx, Details of interrupt and Fault handling, register allocation >and similar topics, all from a low-level, intensive software perspective. Don't know about any classes. But Intel has lots of FREE manuals. Call their Literature Hotline 1-800-548-4725 and order the following: i960 Processor Assembler's Guide i960 Processor Software Utilities User Guide gcc960 User's manual gdb960 User's manual These will be shipped to you for FREE!!! You can download the source code etc from ftp.intel.com. Look for GNUTOOLS960 or something like that. Also purchase a copy of ... i960 Cx Microprocessor User's Manual Which will tell you 95% of what you need to know about i960? The 5% it doesn't tell you? Things like ... Can I my code use region 0 for anything (e.g. stack space)? If anybody knows please post an answer. - -earl --------------------------- End of New-News digest ********************** From andyg@emass.com Thu May 30 07:01:36 1996 From: Andy Gross Date: Thu May 30 07:01:38 PDT 1996 Subject: Re: i960 and Loader Format Classes? > Newsgroups: comp.os.vxworks > Subject: Re: i960 and Loader Format Classes? > Date: 28 May 1996 17:49:12 -0700 > From: kla@leland.Stanford.EDU (Earl Mitchell) > Organization: Stanford University, CA 94305, USA > Message-ID: <4og6u8$3oj@elaine25.Stanford.EDU> > References: <199605221848.AA21830@niagra.wrs.com> > > In article <199605221848.AA21830@niagra.wrs.com>, > Luke Teyssier wrote: > >Hello, > > > >Does anyone know of good clases in the San Francisco Bay Area for > >the i960 Architecture family? I'm looking for a fairly intensive > >class on assembly language details and idioms, comparison of the Cx, > >Jx, Hx, Sx, Kx, Details of interrupt and Fault handling, register allocation > >and similar topics, all from a low-level, intensive software perspective. > > Don't know about any classes. But Intel has lots of FREE manuals. > Call their Literature Hotline 1-800-548-4725 and order the following: > > i960 Processor Assembler's Guide > i960 Processor Software Utilities User Guide > gcc960 User's manual > gdb960 User's manual > > These will be shipped to you for FREE!!! You can download the source code etc > from ftp.intel.com. Look for GNUTOOLS960 or something like that. > > Also purchase a copy of ... > > i960 Cx Microprocessor User's Manual > > Which will tell you 95% of what you need to know about i960? > > The 5% it doesn't tell you? > > Things like ... > > Can I my code use region 0 for anything (e.g. stack space)? > If anybody knows please post an answer. > > - -earl It's been a while since I did anything with the Cx (I'm using the Jx now), but I believe region 0 can be used for anything. However, all accesses to the first 1K (this size may change based on which chip you are using - I'm not sure) will go onchip. The rest of region 0 may be configured any way you wish. BTW, I believe that in the "Solutions 960" book they describe the different families. There is also an "i960 Processors and Related Products" manual that would be of some help. Earl is correct - talk to Intel's documentation department. They have plenty of material for you. There's probably stuff on their fax-back service also... Please note; either decide on your processor first or get the User's Manual to each one you are considering. The major differences will be in the older lines (S, K, C) vs. the newer lines (J, H). While the newer ones support the i960 core instructions, they've added some. If you're planning on using vxWorks with the i960, you should be aware that the Cygnus GNU compiler that they're shipping with Tornado (at least on the PC) does not support the new i960 instructions. This is actually a problem with the assembler - it hasn't been updated for the new lines. Additionally, any new optimizations specific to the J or H are not in there - optimization is done for the C. Andy -- Andy Gross EMASS Senior Software Engineer 10949 E. Peakview Ave. E-Mail: andyg@emass.com Englewood, CO 80111 (303) 705-3762 From MikeMcQuade@prometheus.connectsoft.net Thu May 30 12:19:23 1996 From: Mike McQuade Date: Thu May 30 12:19:25 PDT 1996 Subject: whats the proper use of sysBusToLocalAdrs() Whats the correct use of sysBusToLocalAdrs() ? Ive got several adresses in my header file that Id like to use this call on to get the correct adress dynamicaly. The way I used it in the past was to run it one time, note the base address for STD I/O space, then go back and edit my header file, to put in the correct addresses. Any suggestions. ? Thanks MikeMcQuade@connect.com From aisie.aisinc.com!aisinc.com!pew@msen.com Thu May 30 12:30:15 1996 From: pew@aisinc.com (Paul E. Wilt) Date: Thu May 30 12:30:18 PDT 1996 Subject: Re: i960 and Loader Format Classes? > > Submitted-by andyg@emass.com Thu May 30 07:01:36 1996 > Submitted-by: Andy Gross > > > Newsgroups: comp.os.vxworks > > Subject: Re: i960 and Loader Format Classes? > > Date: 28 May 1996 17:49:12 -0700 > > From: kla@leland.Stanford.EDU (Earl Mitchell) > > Organization: Stanford University, CA 94305, USA > > Message-ID: <4og6u8$3oj@elaine25.Stanford.EDU> > > References: <199605221848.AA21830@niagra.wrs.com> > > > > In article <199605221848.AA21830@niagra.wrs.com>, > > Luke Teyssier wrote: > > >Hello, > > > > > >Does anyone know of good clases in the San Francisco Bay Area for > > >the i960 Architecture family? I'm looking for a fairly intensive > > >class on assembly language details and idioms, comparison of the Cx, > > >Jx, Hx, Sx, Kx, Details of interrupt and Fault handling, register allocation > > >and similar topics, all from a low-level, intensive software perspective. > > > > Don't know about any classes. But Intel has lots of FREE manuals. > > Call their Literature Hotline 1-800-548-4725 and order the following: > > > > i960 Processor Assembler's Guide > > i960 Processor Software Utilities User Guide > > gcc960 User's manual > > gdb960 User's manual > > > > These will be shipped to you for FREE!!! You can download the source code etc > > from ftp.intel.com. Look for GNUTOOLS960 or something like that. > > > > Also purchase a copy of ... > > > > i960 Cx Microprocessor User's Manual > > > > Which will tell you 95% of what you need to know about i960? > > > > The 5% it doesn't tell you? > > > > Things like ... > > > > Can I my code use region 0 for anything (e.g. stack space)? > > If anybody knows please post an answer. > > > > - -earl > > It's been a while since I did anything with the Cx (I'm using the Jx now), > but I believe region 0 can be used for anything. However, all accesses > to the first 1K (this size may change based on which chip you are using - > I'm not sure) will go onchip. The rest of region 0 may be configured > any way you wish. Be careful of the setting you use for the number of register cache sets used. The default number of register caches (5, I believe) does not use much of the internal RAM. However, if you set the cache size to a larger value, the area from 0x3ff and downward will be used. We currently use 9 register cache sets. This ends up eating internal RAM from 0x300-0x3ff for register caches. Another thing to realize is that if you use the cached interrupt vector capability of the i960, up to the first 64 bytes of internal RAM are used to cache these vectors (0x000-0x040). Another thing to watch: The "i960 Cx Microprocessor User's Manual" has an error in it when describing register caches. Table 5-1 in the March 1994 version of the manual (p 5-9) indicates that each register cache takes 16 bytes. Actually each register cache is 64 bytes (16 registers * 4 bytes/register). We ran afoul of this last year when trying to increase the number of register sets we used in our system. Just thought you might find this information useful. > > BTW, I believe that in the "Solutions 960" book they describe the different > families. There is also an "i960 Processors and Related Products" manual > that would be of some help. Earl is correct - talk to Intel's documentation > department. They have plenty of material for you. There's probably stuff > on their fax-back service also... > > Please note; either decide on your processor first or get the User's Manual > to each one you are considering. The major differences will be in the > older lines (S, K, C) vs. the newer lines (J, H). While the newer ones > support the i960 core instructions, they've added some. > > If you're planning on using vxWorks with the i960, you should be aware > that the Cygnus GNU compiler that they're shipping with Tornado (at > least on the PC) does not support the new i960 instructions. This > is actually a problem with the assembler - it hasn't been updated > for the new lines. Additionally, any new optimizations specific > to the J or H are not in there - optimization is done for the C. > > Andy > > > -- > Andy Gross EMASS > Senior Software Engineer 10949 E. Peakview Ave. > E-Mail: andyg@emass.com Englewood, CO 80111 > (303) 705-3762 -- ============================================================================= Paul Wilt, Sr Software Engineer (313) 332-7044 FAX : (313) 332-7077 Applied Intelligent Systems, Inc. pew@aisinc.com 110 Parkland Plaza Ann Arbor, MI 48103 ============================================================================= From daemon@csg.lbl.gov Fri May 31 04:00:30 1996 From: daemon@csg.lbl.gov Date: Fri May 31 04:00:32 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Fri May 31 04:00:27 PDT 1996 Subject: Re: VXworks works, really? Subject: Benchmarking VxWorks Subject: 3C595 100Mb/sec ethernet driver Subject: B E W A R E 68060 & 68040 Math Performance is Terrible ! ! ! ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: VXworks works, really? Date: 28 May 1996 17:25:35 -0700 From: kla@leland.Stanford.EDU (Earl Mitchell) Organization: Stanford University, CA 94305, USA Message-ID: <4og5hv$38f@elaine25.Stanford.EDU> References: <9605231151.AA04387@rst.co.il> In article , Dan Hildebrand wrote: >In article <9605231151.AA04387@rst.co.il>, >Leonid Rosenboim wrote: >>> Why choose VXworks if a 486 running the QNX OS or kernel equals >>> or outperforms VXworks on a 266Mhx Dec Alpha? >> >>I do not know how well VxWorks on Alpha performs, but VxWorks surely >>outperforms QNX on a i486. When comaring two OS's, compare them >>on the same hardware. > >Since VxWorks is not a protected-mode OS, it does not need to manipulate >the MMU on a context switch, so it has less work to do. The question is >whether or not MMU protection is valuable to the realtime system, and if it >is, how much overhead is worth incurring to have it. They probably made the right choice since this is an application specific issue. But this doesn't make you feel any better when you are desperately trying to find a bug in code causing memory to be overwritten. ;-) --------------------------- Newsgroups: comp.os.vxworks Subject: Benchmarking VxWorks Date: 29 May 1996 04:47:21 GMT From: gms4vme@ix.netcom.com (Benjamin Sharfi) Organization: Netcom Message-ID: <4ogksp$siu@sjx-ixn3.ix.netcom.com> Hi, I am looking for a benchmark suite that will allow me to benchmark various single board computers with various CPU's using VxWorks. Can any one out there help !! Please E-Mail reponse to gms4vme@ix.netcom.com Thank You Don Palmer - Project Manager General Micro Systems, Inc --------------------------- Newsgroups: comp.os.vxworks Subject: 3C595 100Mb/sec ethernet driver Date: Thu, 30 May 1996 12:24:44 -0400 From: Mark Nelson Organization: Eastman Kodak Company Message-ID: <31ADCBCC.35A6@Kodak.COM> Anyone got any info on the availability of a 100Mb/sec ethernet driver for the 3-Com 3C595 etherlink controller? My target system is a Pentium with multiple PCI buses, I'm using the pc486 BSP from WRS. - -- Mark J. Nelson Development Engineer Imaging Science Career Development Program (716)588-6226 Color Scanning Systems Mark_Nelson@Kodak.COM Eastman Kodak Company --------------------------- Newsgroups: comp.realtime,comp.os.vxworks Subject: B E W A R E 68060 & 68040 Math Performance is Terrible ! ! ! Date: 30 May 1996 17:24:57 GMT From: ipss@roanoke.infi.net (ipss) Organization: Industrial Process Support Services Message-ID: <4okll9$uc1@nw101.infi.net> While doing a project in VxWorks 5.2 with MVE177 (68060) we found out that the floating point performance for the 68060 & 68040 is terrible whenever numbers whose magnitude is less than FLT_MIN or greater than FLT_MAX are encountered. From correspondence with others this is apparently due to very poor software exception handling in Motorola math libraries. This exception handling was done in hardware on 68030 and earlier. Motorola is aware of the problem and is utterly unconcerned that the execution time for a simple fabs() function call can go from 1 microsec to 53 microsec. They don't seem to get it! All should be careful when selecting either the 68060 or 68040 architecture for time critical, math intensive applications. Your only recourse is to rewrite Motorola math libraries or add under/overflow checks all over your code. Steve Fournier ipss@roanoke.infi.net --------------------------- End of New-News digest ********************** From leonid@rst.co.il Fri May 31 05:12:24 1996 From: leonid@rst.co.il (Leonid Rosenboim) Date: Fri May 31 05:12:29 PDT 1996 Subject: Re: logMsg shared memory message Q?? This one is a challenging question I just could not resist: > From: "JEAN FREGEAU" > Date: Wed May 29 08:03:52 PDT 1996 > Subject: logMsg shared memory message Q?? > I would like to route output from all of the boards in my system to a > single output source. Does anyone know of a way to use the logMsg > feature with a shared memory message Q? logTask listens on amessage queue, and logMsg sends its stuff to a message queue, and shared msgQ's are treated the same way as local ones except when initialized. So, what you need is to tell logLib to send and listen with a global msgQ that you would create or bind with. This would normally call for modifications of logLib.c source code, if you have that. That is however a bad approach, the shared msgQ's can't be sent to from an interrupt level, and logMsg use frequently used exactly for that. Also, logMsg passes arguments which are typically local pointers, and another CPU will get plain garbage referring through them. This means you will need to use logFdAdd() to add a file descriptor to the logTask to send its messages to. This fd can be related to a custom driver you must write, similar to the pipeDrv stuff, which would optionally add some identification of the CPU number, and send the FORMATTED data over to the master CPU. It can be more easily done with sockets using if_sm network. Just set up a TCP conenction to the master board, and use the socket fd with logAdd() to direct all messages over there. I hope this information will help. ----------------------------------------------------------------------- Leonid Rosenboim Phone: +972-3-559-8144 R S T Software Industries Ltd. Mobile: +972-50-307-142 P.O.Box 11502, AZUR 58017, Israel Fax: +972-3-559-8244 http://knight.trendline.co.il/~leonid/ E-Mail: leonid@rst.co.il From jhillman@wrs.com Fri May 31 06:37:51 1996 From: Jon Hillman Date: Fri May 31 06:37:53 PDT 1996 Subject: Re: logMsg shared memory message Q?? At 05:12 AM 5/31/96 PDT, you wrote: >Submitted-by leonid@rst.co.il Fri May 31 05:12:24 1996 >Submitted-by: leonid@rst.co.il (Leonid Rosenboim) > >This one is a challenging question I just could not resist: > >> From: "JEAN FREGEAU" >> Date: Wed May 29 08:03:52 PDT 1996 >> Subject: logMsg shared memory message Q?? >> I would like to route output from all of the boards in my system to a >> single output source. Does anyone know of a way to use the logMsg >> feature with a shared memory message Q? > >logTask listens on amessage queue, and logMsg sends its stuff to a message >queue, and shared msgQ's are treated the same way as local ones except >when initialized. So, what you need is to tell logLib to send and listen >with a global msgQ that you would create or bind with. > >This would normally call for modifications of logLib.c source code, >if you have that. > >That is however a bad approach, the shared msgQ's can't be sent to from >an interrupt level, and logMsg use frequently used exactly for that. >Also, logMsg passes arguments which are typically local pointers, >and another CPU will get plain garbage referring through them. > >This means you will need to use logFdAdd() to add a file descriptor >to the logTask to send its messages to. This fd can be related >to a custom driver you must write, similar to the pipeDrv stuff, >which would optionally add some identification of the CPU number, >and send the FORMATTED data over to the master CPU. > >It can be more easily done with sockets using if_sm network. >Just set up a TCP conenction to the master board, and use the socket >fd with logAdd() to direct all messages over there. I thought I'd add a suggestion here Jean. Establish a generic LOCAL message queue to which you write all appropriate logMsg-type stuff. Write a low priority task which will run on each processor and will receive each message written to that Q. That task in turn will send the messages via shared message queue or sockets to the processor at which you want the message directed, and it will do the logMsg. Hey, it works. :-> _________________________ | Jon Hillman, FAE | | Wind River Systems | /) (407) 273-4646 (\ / ) Email:jhillman@wrs.com ( \ _( (|_________________________) ) /> (((\ \) /,) / ) / //))/ (\\\\ \_/ / \ \_/ ///// \ / \ / \ _/ \_ / ----/ /----------------------\ \---- / / \ \ From tony@puma.phx.mcd.mot.com Fri May 31 10:48:31 1996 From: Tony Brefeld Date: Fri May 31 10:48:39 PDT 1996 Subject: Re: comp.os.vxworks newsdigest -- Begin original message -- > From: vxwexplo@lbl.gov (the vxWorks Users Group Exploder) > ------------------------------------------------------- > > Newsgroups: comp.os.vxworks > Subject: Benchmarking VxWorks > Date: 29 May 1996 04:47:21 GMT > From: gms4vme@ix.netcom.com (Benjamin Sharfi) > Organization: Netcom > Message-ID: <4ogksp$siu@sjx-ixn3.ix.netcom.com> > > Hi, > > I am looking for a benchmark suite that will allow me to benchmark > various single board computers with various CPU's using VxWorks. > > Can any one out there help !! > > Please E-Mail reponse to gms4vme@ix.netcom.com > > > Thank You > > Don Palmer - Project Manager > General Micro Systems, Inc > -- End original message -- I too would be interested in any benchmark suites available. ------ Thanks Tony Brefeld Motorola MCG-Tempe WK: 1-602-438-3748 MS: DW220 FAX: 1-602-438-6154 2900 S. Diablo Way INTERNET: tony@phx.mcd.mot.com Tempe, Az. 85282 From kozubal@javelin.lanl.gov Fri May 31 14:15:15 1996 From: kozubal@javelin.lanl.gov (Andy Kozubal) Date: Fri May 31 14:15:21 PDT 1996 Subject: RE: whats the proper use of sysBusToLocalAdrs() > > Whats the correct use of sysBusToLocalAdrs() ? > > Ive got several adresses in my header file that Id like to use this call on > to get the correct adress dynamicaly. > > The way I used it in the past was to run it one time, note the base address > for STD I/O space, then go back and edit my header file, to put in the > correct addresses. > > > Any suggestions. ? > > > Thanks > MikeMcQuade@connect.com > The mapping of bus address to RAM address is determined by the BSP, and this mapping might change if you go to a different board or even change BSP versions for the same board. Most of the code I've seen uses a call to sysBusToLocalAdrs() for each device during initialiation and saves the returned value. Example: LOCAL struct myDevReg { .... .... } *pReg; /* Initialize ptr to register set */ status = sysBusToLocalAdrs (VME_AM_STD_SUP_DATA, MY_DEV_REG, (char **)&pReg); ...... pReg->csr = 0x40; |=======================================================================| | Andy Kozubal (DX-7) | Digital: akozubal@lanl.gov | | Mail Stop P947 | | | Los Alamos National Laboratory | Analog: (505) 667-6508 | | Los Alamos, New Mexico 87545 | | |=======================================================================| From mea@mclean.sparta.com Fri May 31 15:25:06 1996 From: Mike Anderson Date: Fri May 31 15:25:26 PDT 1996 Subject: RE: whats the proper use of sysBusToLocalAdrs() At 02:15 PM 5/31/96 PDT, you wrote: >Submitted-by kozubal@javelin.lanl.gov Fri May 31 14:15:15 1996 >Submitted-by: kozubal@javelin.lanl.gov (Andy Kozubal) > >> >> Whats the correct use of sysBusToLocalAdrs() ? >> > Here's a code snippett that I use for doing address translation: #include "vxWorks.h" #include "vme.h" VOID busAddrXlate (char *busAddr, int AM) { char *xlatedAddr; int status; status = sysBusToLocalAdrs(AM, busAddr, &xlatedAddr); if (status == OK) printf("This board sees address 0x%x at 0x%x\n", (UINT32) busAddr, (UINT32) xlatedAddr); else { printf("sysBusToLocalAdrs error!\n"); printf("Unable to convert address 0x%x with address modifer 0x%x!\n\n", (UINT32) busAddr, (UINT32) AM); } } VOID localAddrXlate (char *busAddr, int AM) { char *xlatedAddr; int status; status = sysLocalToBusAdrs(AM, busAddr, &xlatedAddr); if (status == OK) printf("This board sees address 0x%x at 0x%x\n", (UINT32) busAddr, (UINT32) xlatedAddr); else { printf("sysBusToLocalAdrs error!\n"); printf("Unable to convert address 0x%x with address modifer 0x%x!\n\n", (UINT32) busAddr, (UINT32) AM); } } -> busAddrXlate(0xfe00, 0x2d) This board sees address 0xfe00 at 0xff04fe00 To then use this techique in code: #include "vxWorks.h" #include "vme.h" STATUS accessHw (char *busAddr, int AM, UINT32 value, char charWidth) { char *xlatedAddr; int status; status = sysBusToLocalAdrs(AM, busAddr, &xlatedAddr); if (status == OK) { switch (charWidth) { case 1: *(xlatedAddr) := value & 0xff; break; case 2: *(xlatedAddr) := value & 0xffff; break; case 4: *(xlatedAddr) := value & 0xffffffff; break; default: *(xlatedAddr) := value & 0xff; break; } } else { printf("sysBusToLocalAdrs error!\n"); printf("Unable to convert address 0x%x with address modifer 0x%x!\n\n", (UINT32) busAddr, (UINT32) AM); return(ERROR); } return(OK); } Then, to write the 8-bit value 0xb5 to short I/O at 0xfe00 you would call: -> accessHw (0xfe00, 0x2d, 0xb5, 1) HTH, =============================================================================== __ Real-Time System Development, Integration, Training and Services //\\ // \\ Mike Anderson // /\ \\ Chief Engineer Voice : (703) 448-0210 ext. 235 // / \ \\ SPARTA, Inc. FAX : (703) 734-3323 // \ \\ 7926 Jones Branch Drive EMAIL : mea@mclean.sparta.com \\ \ // Suite 900 Web : http://www.mclean.sparta.com \\ \ / // McLean, VA 22102 \\ \/ // "Software development is like making \\ // a baby... You can't make a baby in one \\// month by impregnating nine women. -- "Pride in Performance" Some things just take time." ===============================================================================