Job Opportunity

I just started developing stuff with To= rnado II=20 today, and ran already into three

harmless, yet=20 extremely annoying bugs:

When you select New, you have to select the type of the= new=20 file, the file name,

whether or not to add it to the workspace, etc. You can= not=20 click 'Ok' before the

filename is entered. This is all = really=20 nice.

Bug 1:

When you've entered the file name, and select a different type (e.g. C/C++ Source

file instead of C/C++ Header file), you can't click on Ok anymore. You *have* to

change the name (add and remove a char) before you can click Ok again.

The docked debugging windows are really neat.

The available windows are :

-All local variables

-The watches

-Backtrace

-Memory dump

-Registers

The variables and watch windows are tre= e views,=20 so when the var= iable is a=20 struct,

you can click on the '+' to open a sub-tree and = see all the struct-members.

Bug 2: In the variables window I wanted to view the str= uct=20 members in hex, so

I set the variable to Hex (you ca= n't=20 change the format of a single member!).

But then the members are not view= ed in=20 Hex. It is still decimal. You have to

close and open the sub-tree first= , before=20 they are viewed in hexadecimal.

To make this even worse, the format snaps=20 back to decimal as soon as you

do a single-step. Close and open = the=20 subtree, and it's hex again.

The same thing happens in the watch window.

These two bugs are merely annoying. Here's a real=20 one:

It is in the Watch window. I have= a=20 structure defined as following:

typedef struct {

= ; int=20 Type; /* ADDRESS_TYPE_MEM or ADDRESS_TYPE_PORT */<= /DIV>

int BaseAddress;

int MemSize

} TPCIResource;

In my function:

TPCIResource=20 *ResourceList;

Then I allocated memory for six structu= res, so=20 ResourceList is an

array with six me= mbers. I=20 fill the array and add a watch:

= ;=20 ResourceList[1]

I= t seems Ok.=20 The values are right. But when I try to change it to hex

it completely fai= ls. I=20 select Hex (a radio button in the properties window),

and as soon as I click Ok, the properties window is all= empty.=20 This is

the first half of the bug. The second half: I restarted= the=20 debugger, added

ResourceList[1] again, and started single stepping. Aft= er=20 every step

weird crap is added to the watch list (e.g [0]). It is = all in=20 someway related

to Resource list. You can't remove these watches anymor= e, and=20 have

to restart the entire debugger to empty the watch=20 list.

Same stuff happens with:

ResourceList + 1

ResourceList[Index]

I now really start wondering how Tornad= o II is=20 developed. I figure they got

2 weeks. The firs= t week they=20 quickly collected some code, and the 2nd

week they merged it together and labeled it Tornado II. They needed at

least a third week for testing, but whatever...<= /DIV>

Geurt.

-- warnings are just like errors, only not=20 - --

- ------=_NextPart_000_0075_01BEB813.134AA120-- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How to Handle "Large" Data File (~50Mb)? Date: Fri, 18 Jun 1999 13:18:40 -0700 From: "David Wiggins" Message-ID: <92960669@NEWS.SAIC.COM> References: <37695B37.1433F7AA@dfrc.nasa.gov> If your 50Mb file is mounted on your host machine, we had this exact problem (although on a different board, a Radstone PPC 604e running vxWorks 5.3.1). If the file is on a target-mounted drive (e.g. Scsi), then disregard this message. The problem for us was that the default network driver is RPC, which loads the ENTIRE file across the network into target memory on the fopen call. So the size of the file you're able to open is limited by the amount of target memory available at the time. It looks like this is problem you're having, as evidenced by the memPartAlloc error. We also got this error along with the Broken pipe error you're getting. So, our solution was merely to nfsMount the host partition on our target, then all the open, read, write, close, etc. functions get routed through the nfs driver instead of the rpc driver. The nfs driver does not have this "feature" of the rpc driver, but instead works across the network as you read/write the file (with some reasonable amount of memory buffering, of course). Hope this helps, Dave Wiggins SAIC-San Diego Ken Norlin wrote in message news:37695B37.1433F7AA@dfrc.nasa.gov... > Hi, > > I'm trying to open a ~50Mb file. When I do an fopen on the file, I get the > following error: > 0x17c47f0 (tShell): memPartAlloc: block too big - 524 in partition 0x19b9f4. > ISRS.DAT: Broken pipe. > > If I cut the file down to 14 Mb, I'm able to open the file & read the data. > So it must be the file size that's causing the problem. > > I'm using an MVME2604 w/ BSP 1.1/5. Please let me know if you've run across > this problem and how you solved it. (I'm assuming there must be some way to > bump up the memory buffer size). > > Thanks a bunch, > Ken N. > > > -- > Ken Norlin --- Simulation Engineer > NASA Dryden Flight Research Center > ken.norlin@dfrc.nasa.gov > (661) 258-2046 FAX (661) 258-2792 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Putting a task at the end of the READY queue. Date: Fri, 18 Jun 1999 13:36:23 -0800 From: mlang@fst.jpl.nasa.gov (M Lang) Organization: JPL Message-ID: References: <7k809n$hbd$1@nnrp1.deja.com> <9F1EA55D2B1016B4.5BD5B7FE9D091CF6.77BE3D78F3CD18CA@lp.airnews.net> In article <9F1EA55D2B1016B4.5BD5B7FE9D091CF6.77BE3D78F3CD18CA@lp.airnews.net>, Dave Schmidt wrote: > That will put the current task at the end of the ready queue for the > current priority. It will allow any other tasks on the ready queue of > the same priority to be run (sort of round robin, if you issue > taskDelay(0) for them also when you want to go to the next task), of > course all higher priority tasks will run and intterrupt service > routines will also run. > > Please note that this method will NOT allow tasks of lower priority to > run, it just forces a rescheduling. If the task being delayed is the > only one of that priority, then it will restart immediately. > > Good Luck > > Dave Schmidt > > > vivek1796@my-deja.com wrote: > > > > hi everybody, > > > > I am looking for a way to put a running task at the end of the > > READY queue.Will a call to taskDelay(0) provide what I need?? > > > > I will be grateful to any suggestions. > > > > -regards, > > vivek. > > > > Sent via Deja.com http://www.deja.com/ > > Share what you know. Learn what you don't. However, that's not what I observed (I might be missing something here). I ran the following test: ... lowPrioTask () { for (;;) printf ("low priority task running\n); } hiPrioTask ( { for (;;) { taskDelay (0); printf ("hi priority task running\n");` } } test() { taskSpawn ("tLow", 110, 0, 10000, lowPrioTask, 0,0,0,0,0,0,0,0,0,0); taskSpawn ("tHi", 105, 0, 10000, hiPrioTask, 0,0,0,0,0,0,0,0,0,0); } [There was no other task at the priority of 105 and 110] I did see the line "low priority task running" popped up once in a while so: 1) the lower priority task did get rescheduled even though the higer priority one calls taskDelay(0) 2) The higher priority task "hiPrioTask" was the only one with priority 105 but it didn't restart **immediately**. This is with vxWorks 5.3.1 on a PPC. = Minh = --------------------------- Newsgroups: comp.os.vxworks Subject: Where can I find VxWorks Programmer's Guied or Reference Manual? Date: Sat, 19 Jun 1999 11:53:23 +0900 From: "��" Organization: Pusan National University Message-ID: <7kf0v1$sl7$1@jagalchi.cc.pusan.ac.kr> Where can I find Library functions and system calls for VxWorks? How much it is differ with linux? How much differ VxWorks data types and linux data types? --------------------------- Newsgroups: comp.os.vxworks Subject: FREE NASTY TEEN SEX PICTURES!! Date: 19 Jun 1999 07:31:09 GMT From: piofvk@hotmail.com Organization: Virgin News Service Message-ID: <7kfgvt$hsn$4805@nclient5-gui.server.virgin.net> FOR THE ULTIMATE FREE NASTY TEEN SEX PICTURES VISIT http://website.lineone.net/~charlie.appleby/free-section.htm NO CC OR ONLINE CHECK REQUIRED, ITS ALL FREE!!!! --------------------------- End of New-News digest ********************** From owner-vxwexplo-process Sun Jun 20 04:00:08 1999 From: daemon@csg.lbl.gov Date: Sun Jun 20 04:00:11 PDT 1999 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sun Jun 20 04:00:04 PDT 1999 Subject: Re: Tornado II - defining include directories Subject: Re: Putting a task at the end of the READY queue. Subject: Re: dosFs woes ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Tornado II - defining include directories Date: Sat, 19 Jun 1999 15:11:08 GMT From: "Ken Hayber" Message-ID: References: <375BE5E8.E83B5254@email.mot.com> Daniel, Try this... 1) Go to the Builds tab on the Workspace window. 2) Open the tree to 'Default' or whatever you are using. 3) Right click and select Properties from the menu 3) Select the C/C++ tab and enter your -I information there (you need to use Ctrl-Enter to add a new line) I haven't tried it yet myself, but was planning to soon. I haven't got a good kernel build yet. Good luck, Ken H. Daniel Stewart wrote in message news:375BE5E8.E83B5254@email.mot.com... > I am attempting to use Tornado II (switching over from Tornado I) and am > becoming increasing frustrated. Does anyone know how to specify in > Tornado II a directory for my own include files? I would think this is > an extremely simple and straightforward thing to do, but obviously it is > not. Since Tornado II remakes the makefile each time you build, futzing > with that is of no use. I continue to receive the > "No rule to make target XXX.h needed to make XXX.o" > Surely there is a method of specifying additional include directories to > search during the make process (similar to adding a vpath variable to > the makefile). > > Thanks, > Daniel Stewart > > Motorola Labs > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Putting a task at the end of the READY queue. Date: Sat, 19 Jun 1999 12:32:06 -0400 From: Tim Sohacki Organization: Nortel (Northern Telecom) Message-ID: <376BC606.E9E3D083@nortelnetworks.com> References: <7k809n$hbd$1@nnrp1.deja.com> <9F1EA55D2B1016B4.5BD5B7FE9D091CF6.77BE3D78F3CD18CA@lp.airnews.net> Minh, I ran this on a vxWorks 5.4 BSP, the MV1603, a 603e PowerPC. About 1 in 30,000 times, the low priority task prints. Note that this is on average, though I sometimes see tLow print twice within a couple hundred lines. I changed the "printf" to "logMsg" and was surprised at the result: ^M0x1925588 (tHi): hi priority task running 1 ^M0x1925588 (tHi): hi priority task running 2 ^M0x1925588 (tHi): hi priority task running 3 ^M0x1925588 (tHi): hi priority task running 66th ^M0x1925588 (tHi): hi priority task running 67th (about) ^M0x1927eb0 (tLow): low priority task running ^M0x1925588 (tHi): hi priority task running ^M0x1927eb0 (tLow): low priority task running ^M0x1925588 (tHi): hi priority task running ^M0x1927eb0 (tLow): low priority task running ^M0x1925588 (tHi): hi priority task running ... etc ... I added a taskPriorityGet() and verified that tLow is still running at 110 when this happens. There's an an SPR that talks about a scheduling problem; this may be related. Cheers, Tim > lowPrioTask () > { > for (;;) > printf ("low priority task running\n"); > } > > hiPrioTask ( > { > for (;;) > { > taskDelay (0); > printf ("hi priority task running\n"); > } > } > > test() > { > taskSpawn ("tLow", 110, 0, 10000, lowPrioTask, 0,0,0,0,0,0,0,0,0,0); > taskSpawn ("tHi", 105, 0, 10000, hiPrioTask, 0,0,0,0,0,0,0,0,0,0); > } --------------------------- Newsgroups: comp.os.vxworks Subject: Re: dosFs woes Date: Fri, 18 Jun 1999 23:01:54 -0700 From: "Cameron Fowler" Organization: AT&T WorldNet Services Message-ID: <7khmqk$gau$1@bgtnsc03.worldnet.att.net> References: <374E8F4B.5EEC0DDD@syncro-tech.com> <929571122.5827@www.remarq.com> In order to flush your data down to disk, you will need to do a fflush() followed by ioctly(FIOSYNC) This should push your data down to the disk Alan Yakiwchuk wrote in message <929571122.5827@www.remarq.com>... >I think the information (directories, etc.) is only updated >when you close the file. > >Have you tried flushing the output buffers to force and >update? i.e. FIOFLUSH Let me know if this works 'cause >I may be facing a similar problem soon... > > > >**** Posted from RemarQ - http://www.remarq.com - Discussions Start Here (tm) **** --------------------------- End of New-News digest ********************** From owner-vxwexplo-process Sun Jun 20 05:22:33 1999 From: "Slutsker, Rasty" Date: Sun Jun 20 05:22:37 PDT 1999 Subject: Question about SLIP connection This message is in MIME format. Since your mail reader does not understand this format, some or all of this message may not be legible. ------_=_NextPart_001_01BEBB1F.DD4238BC Content-Type: text/plain Hi, Does anybody know how to have SLIP/PPP in Windows over raw COM1/2 port (without modem)? Rasty Keywords: vxWorks, Tornado, Windows ------_=_NextPart_001_01BEBB1F.DD4238BC Content-Type: text/html Content-Transfer-Encoding: quoted-printable Question about SLIP connection

Hi,

Does anybody know how = to have SLIP/PPP in Windows = over raw COM1/2 port (without = modem)?

Rasty

Keywords: vxWorks, = Tornado, Windows

------_=_NextPart_001_01BEBB1F.DD4238BC-- From owner-vxwexplo-process Sun Jun 20 18:42:46 1999 From: Fred Roeber Date: Sun Jun 20 18:42:50 PDT 1999 Subject: Re: Putting a task at the end of the READY queue. Dave Schmidt wrote: > That will put the current task at the end of the ready queue for the > current priority. It will allow any other tasks on the ready queue of > the same priority to be run (sort of round robin, if you issue > taskDelay(0) for them also when you want to go to the next task), of > course all higher priority tasks will run and intterrupt service > routines will also run. > > Please note that this method will NOT allow tasks of lower priority to > run, it just forces a rescheduling. If the task being delayed is the > only one of that priority, then it will restart immediately. Minh Lang (mlang@fst.jpl.nasa.gov) replied: > However, that's not what I observed (I might be missing something > here). I ran the following test: > > ... > lowPrioTask () > { > for (;;) > printf ("low priority task running\n); > } > > hiPrioTask ( > { > for (;;) > { > taskDelay (0); > printf ("hi priority task running\n");` > } > } <> > > I did see the line "low priority task running" popped up once in a while > so: > > 1) the lower priority task did get rescheduled even though the higer > priority one calls taskDelay(0) > > 2) The higher priority task "hiPrioTask" was the only one with priority > 105 but it didn't restart **immediately**. > > This is with vxWorks 5.3.1 on a PPC. First, its nice to see a post from a WRS person about how their OS works. That happens far too rarely. As far as what Minh saw, its just an artifact of his test method. The general problem is that you need special methods to examine low level OS behaviour. The best method (I think) for seeing how things like tasking work in VxWorks is to use the WindView tool which shows true time based traces with low intrusion. Using printf calls to try to follow low level OS operation just doesn't work because printf interferes with operation too much. In this case what I think happens is that the high priority task runs for a number of interations printing it's message and then rescheduling itself. At some point, the ring buffer used by the serial driver to buffer output characters fills up. At this point, the high priority task blocks allowing the low priority task to run. The low priority task then does it's output request which gets mixed in. The blocking action of the printf call is what skews the results since it takes so long to output characters. To get around this, you could use WindView to look at things or toggle some external signals in place of the printf calls and look at them on a logic analyzer. In that case, you should see no execution of the low priority task. Fred | Fred J Roeber, BBN Systems & Technologies | | 4 John Clarke Road Middletown, RI 02842-5202 | | froeber@bbn.com 401-848-3548 | | TraceMaker product manager | | http://www.tracemaker.bbn.com | From owner-vxwexplo-process Mon Jun 21 04:00:10 1999 From: daemon@csg.lbl.gov Date: Mon Jun 21 04:00:14 PDT 1999 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Mon Jun 21 04:00:06 PDT 1999 Subject: Re: initialization of dosFs Subject: vxWorks and asynchronous IO Subject: Re: Games on vxworks Subject: Re: Games on vxworks Subject: Re: SDS (BDM) on MPC860 Subject: Is raise() a no-op? Subject: Re: vxWorks.st_rom booting Subject: questions on memory management Subject: Re: Games on vxworks Subject: Re: Is raise() a no-op? Subject: Re: Putting a task at the end of the READY queue. Subject: Re: Where can I find VxWorks Programmer's Guied or Reference Manual? ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: initialization of dosFs Date: 17 Jun 99 22:06:55 GMT From: georg@sgl.crestech.ca (Georg Feil) Organization: Centre for Research in Earth and Space Technology Message-ID: References: <376616CC.39AF4382@stest.ch> Rolf Kistler writes: >I have a problem with the initialization of dosFs when >booting vxWorks. >The initialization of dosFs fails with the following >error message: > >> Attaching to scsi device... 0xf73bc (tBoot): memPartAlloc: >> block too big - 65600 in partition 0x744c8. > > >There are some quite deeply nested directory trees on the dosFs >which are created without any problems by vxWorks applications. Hi... VxWorks stores the directory entry for all dosFs files and dirs in memory. If there are a lot of files there is a big flurry of small mallocs when the device is mounted, and sometimes dosFs tends to fragment memory pretty badly, especially with the "improved" dosFs memory handling introduced about half a year ago. Try disabling the private dosFs memory partition (check the documentation for dosFs patches). Or if you're using an old dosFs version without separate memory partition ability then try the patch to see if it helps. Good luck, Georg - ----------------- - -- Georg Feil | http://www.sgl.crestech.ca/ Space Geodynamics Laboratory | Email: georg@sgl.crestech.ca CRESTech | Phone: (416) 665-5458 4850 Keele St./North York/Ont/Canada/M3J 3K1 | Fax: (416) 665-1815 --------------------------- Newsgroups: comp.os.vxworks Subject: vxWorks and asynchronous IO Date: Sun, 20 Jun 1999 20:13:04 +0200 From: Ralf Gerlich Organization: BSSE Message-ID: <376D2F30.7B9BC20E@t-online.de> This is a multi-part message in MIME format. - --------------248C8348E639139D6A1C7A68 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Hi! I have a problem with vxWorks. I want to use asynchronous IO but I can't find an fcntl() function to set the FASYNC flag. I know that there's an asynchronous IO library but I'd rather use UN*X-style AIO, i.e. using FASYNC. If I try to use fcntl() I get a unresolved symbol message from the loader. Can anybody out there help me? Please, it's urgent. Ciao, Ralf - --------------248C8348E639139D6A1C7A68 Content-Type: text/x-vcard; charset=us-ascii; name="Ralf.Gerlich.vcf" Content-Transfer-Encoding: 7bit Content-Description: Card for Ralf Gerlich Content-Disposition: attachment; filename="Ralf.Gerlich.vcf" begin:vcard n:Gerlich;Ralf tel;fax:+49/7545/911240 tel;home:+49/7545/529 x-mozilla-html:TRUE url:http://www.d-design.net/rgerlich/ org:knewcleus Development Team adr:;;Auf dem Ruhbuehl 181;Immenstaaad;;88090;Germany version:2.1 email;internet:Ralf.Gerlich@t-online.de title:Chief Developer note:ICQ# 38288578 x-mozilla-cpt:;-27424 fn:Ralf Gerlich end:vcard - --------------248C8348E639139D6A1C7A68-- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Games on vxworks Date: Sun, 20 Jun 1999 18:30:56 GMT From: coder@POTTEDMEAT.texas.net (Bruce) Organization: Giganews.Com - Premium News Outsourcing Message-ID: <376d32e5.8677720@news.texas.net> References: <7k4f8g$ag0$1@nnrp1.deja.com> In comp.os.vxworks "Clifford Slocombe" wrote: >I think that you have chosen the wrong OS. If you want games, and a serious >RTOS use QNX, it has a port of Doom! I am interested to know why you would >want do this. There is a graphical slot machine written in VxWorks. My VxWorks application is graphical and uses Ease Software's port of MetaWindows. It can be done, though I agree it may not be the best for blazing fast graphics. Neither is QNX. Bruce --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Games on vxworks Date: Sun, 20 Jun 1999 21:45:05 GMT From: "Clifford Slocombe" Message-ID: References: <7k4f8g$ag0$1@nnrp1.deja.com> <376d32e5.8677720@news.texas.net> What I meant was that there was no standard graphical interface (apart from X/OpenGL), with a wide choice of graphic proprietary graphic API which horse would a games programmer back (if any)? It occurred to me that games does not necessarily mean graphics, but text adventures are a little too obscure these days to be an attractive product feature! W.R.T QNX/Photon, lets not get into that one here, I probably get flamed! Given time, games may appear for the Zinc interface. This at least has the benefit of being (now) a WindRiver product, so is likely to become widely supported. Clifford. Bruce wrote in message news:376d32e5.8677720@news.texas.net... > In comp.os.vxworks > "Clifford Slocombe" wrote: > > >I think that you have chosen the wrong OS. If you want games, and a serious > >RTOS use QNX, it has a port of Doom! I am interested to know why you would > >want do this. > > There is a graphical slot machine written in VxWorks. My VxWorks > application is graphical and uses Ease Software's port of MetaWindows. It > can be done, though I agree it may not be the best for blazing fast > graphics. Neither is QNX. > > Bruce > --------------------------- Newsgroups: comp.os.vxworks,comp.sys.powerpc.tech Subject: Re: SDS (BDM) on MPC860 Date: Mon, 21 Jun 1999 10:52:24 +1000 From: Tom Evans Organization: OzEmail Ltd, Australia Message-ID: <376D8CC8.699D@nowhere.tennyson.com.au> References: <2DDC62F5EEA63247.2DE6DAA0CB52DF9F.E2F447A8A9299D1C@lp.airnews.net> <37675C62.6271@nowhere.tennyson.com.au> <3767CF35.89F0CF67@BoschTelecomInc.com> Pete Kockritz wrote: > > I checked the ICR, it is 0. Looking at some of the other registers I > see: > DAR = ff001e00 I assume you have the MPC860 registers at 0xff000000? Then 0xff001e00 is in the middle of the "Specialized Ram - Reserved" area. If your code accesses this address it will get a bus error. Switch SDS to display mixed code. Which register has this address in it and why is it trying to access this address? What misloaded pointer is it using and why? This is a code bug. > DER = fdc77c0f (debugger handles everything but EXT and DEC) Isn't that what I said to CHANGE? >> Put Breakpoints on EVERY (I mean EVERY) Interrupt vector from >> 0x100 up and also disable SDS from trapping them by clearing >> bits in the DER register Zero bits in the DER and let the hardware handle the trap, and then see which trap it hit. > Setting breakpoints at the interrupt locations didn't show anything. Because you have the DER set to prevent the interrupts. > DSISR = 00004401 > Is the DSISR significant? Read section 7.1.2.1 of the PDF manual. It can be used to decode which instruction failed, but for what you're doing it isn't going to help. gstefkov@sprynet.com said: > Try some other JTAG tools. I have used Corelis and Applied Microsystems > without too many problems. SDS demos never thrilled me. Maybe they put more effort into making reliable tools than they put into flashy demos? We find SDS excellent. Tom Evans InitialSurname@somewhere. --------------------------- Newsgroups: comp.os.vxworks Subject: Is raise() a no-op? Date: Sun, 20 Jun 1999 19:01:46 -0700 (PDT) From: silvermang@webtv.net (Greg Silverman) Organization: WebTV Subscriber Message-ID: <21067-376D9D0A-135@newsd-221.iap.bryant.webtv.net> I was suspicious that the raise function did not work, so I wrote a simple test .. int main() { printf("start\n"); raise(SIGSEGV); printf("exit\n"); } If raise() works "start" but not "exit" should be printed. However, both are output. What gives? Greg Silverman TRW Sunnyvale,Ca --------------------------- Newsgroups: comp.os.vxworks Subject: Re: vxWorks.st_rom booting Date: Sun, 20 Jun 1999 20:38:06 -0700 From: "Steve Doiel" Message-ID: <376db163.0@news.pacifier.com> References: <7k27jk$d29$1@b5nntp2.channeli.net> >Hi everybody. >I'm Vxworks beginner. > >I coded vxworks bsp code based on ADS860 > >I succeded in testing bootrom code. > - bootrom.hex in ROM , and downloading vxWorks image > and then downloading application. > >but, I can't boot using vxworks.st_rom.hex in ROM. >neither bootrom_res_high.hex, vxworks.res_rom_res_low... > >Would you give hints for me? > I haven't worked on an ADS860 BSP but with an MVME162 you have to run "aouttobin" on the vxworks.st_rom file to create a bin file, which may be loaded into flash memory. The command is something like: aouttobin romfile.bin I hope this helps, SteveD --------------------------- Newsgroups: comp.os.vxworks Subject: questions on memory management Date: Mon, 21 Jun 1999 04:52:20 GMT From: vivek1796@my-deja.com Organization: Deja.com - Share what you know. Learn what you don't. Message-ID: <7kkge1$np$1@nnrp1.deja.com> hi everybody, I would like to know how I can free the memory allocated by memPartAlignedAlloc(). Will memPartFree() work??? Also, I would appreciate if anybody could tell me where memalign() is preferred over malloc(). Thanks for any help, - -regards, vivek. Sent via Deja.com http://www.deja.com/ Share what you know. Learn what you don't. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Games on vxworks Date: Mon, 21 Jun 1999 06:47:01 GMT From: bobarmstrong@home.nothiis.net (Robert Armstrong) Organization: @Home Network Message-ID: References: <7k4f8g$ag0$1@nnrp1.deja.com> <376d32e5.8677720@news.texas.net> Sorry but I think you significantly miss the point. I play Quake II deathmatches a lot (used to be 2 hours minimum each day in a ten way game). A game like quake is a little bit of networking, a little bit of configuration and a huge amount of rendering. I always used OpenGL. So why is vxWorks fundamentaly incapable or unsuitable? The networking is mostly fine now (do read the chapter on network customization, I think it may be chapter 4.5 of the vxWorks 5.4 programmers guide, the default configuration is intended for debug only and reconfiguration is simple). Recovering the configuration from somewhere is simple. There are plenty of fast platforms available (including PCs). There is also an OpenGL product in the partner directory:- http://www.wrs.com/html/trade/wrs00154.html So I think that addresses the "could you", but it still leaves the "why would you want to" and that I have no answer for. Mostly vxWorks has been too expensive for this sort of development, perhaps this new pricing model I've seen mentioned in tech stocks will change that. I drove my graphics card direct last time I used one, then again that was with o/s9 and I'd rather have had a bare machine than that (OK the compiler sort of helped). To be fair, one of my friends programmed some games for Phillips pagers, they went down very well. Bob In article , "Clifford Slocombe" wrote: >What I meant was that there was no standard graphical interface (apart from >X/OpenGL), with a wide choice of graphic proprietary graphic API which horse >would a games programmer back (if any)? > >It occurred to me that games does not necessarily mean graphics, but text >adventures are a little too obscure these days to be an attractive product >feature! > >W.R.T QNX/Photon, lets not get into that one here, I probably get flamed! > >Given time, games may appear for the Zinc interface. This at least has the >benefit of being (now) a WindRiver product, so is likely to become widely >supported. > >Clifford. > >Bruce wrote in message >news:376d32e5.8677720@news.texas.net... >> In comp.os.vxworks >> "Clifford Slocombe" wrote: >> >> >I think that you have chosen the wrong OS. If you want games, and a >serious >> >RTOS use QNX, it has a port of Doom! I am interested to know why you >would >> >want do this. >> >> There is a graphical slot machine written in VxWorks. My VxWorks >> application is graphical and uses Ease Software's port of MetaWindows. It >> can be done, though I agree it may not be the best for blazing fast >> graphics. Neither is QNX. >> >> Bruce >> > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Is raise() a no-op? Date: Mon, 21 Jun 1999 06:49:25 GMT From: bobarmstrong@home.nothiis.net (Robert Armstrong) Organization: @Home Network Message-ID: References: <21067-376D9D0A-135@newsd-221.iap.bryant.webtv.net> If this is the compiler with t101 then exceptions are broken in the base version proted by Wind River. If it's the compiler with t2 then the exceptions should work and I'd suggest calling Wind River (since I expect TRW has support). Bob In article <21067-376D9D0A-135@newsd-221.iap.bryant.webtv.net>, silvermang@webtv.net (Greg Silverman) wrote: >I was suspicious that the raise function did not work, so I wrote a >simple test > >... > >int main() >{ > printf("start\n"); > > raise(SIGSEGV); > > printf("exit\n"); > >} > >If raise() works "start" but not "exit" should be printed. However, both >are output. What gives? > >Greg Silverman >TRW >Sunnyvale,Ca > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Putting a task at the end of the READY queue. Date: Mon, 21 Jun 1999 06:56:46 GMT From: bobarmstrong@home.nothiis.net (Robert Armstrong) Organization: @Home Network Message-ID: References: <7k809n$hbd$1@nnrp1.deja.com> <9F1EA55D2B1016B4.5BD5B7FE9D091CF6.77BE3D78F3CD18CA@lp.airnews.net> <376BC606.E9E3D083@nortelnetworks.com> There was a reschedule problem in 5.3.1 but this was fixed and a patch was published on WindSurf long before 5.4 (Tornado 2) was released. Printf is a very poor way of trying to find out what tasks are doing, I presume someone out there can try this using WindView? That would put this discussion to bed once and for all. If you want to get information out of running tasks then you need to include a suitable method, message queues might well do the trick (logMsg() uses a message queue surprisingly enough, however the message queue is not especially long so you wills still overrun the que with fast tasks). Bob In article <376BC606.E9E3D083@nortelnetworks.com>, Tim Sohacki wrote: >Minh, > >I ran this on a vxWorks 5.4 BSP, the MV1603, a 603e PowerPC. > >About 1 in 30,000 times, the low priority task prints. >Note that this is on average, though I sometimes see >tLow print twice within a couple hundred lines. > >I changed the "printf" to "logMsg" and was surprised >at the result: > >^M0x1925588 (tHi): hi priority task running 1 >^M0x1925588 (tHi): hi priority task running 2 >^M0x1925588 (tHi): hi priority task running 3 > > > >^M0x1925588 (tHi): hi priority task running 66th >^M0x1925588 (tHi): hi priority task running 67th (about) >^M0x1927eb0 (tLow): low priority task running >^M0x1925588 (tHi): hi priority task running >^M0x1927eb0 (tLow): low priority task running >^M0x1925588 (tHi): hi priority task running >^M0x1927eb0 (tLow): low priority task running >^M0x1925588 (tHi): hi priority task running >.... etc ... > >I added a taskPriorityGet() and verified that tLow is still >running at 110 when this happens. There's an an SPR that >talks about a scheduling problem; this may be related. > >Cheers, >Tim > > >> lowPrioTask () >> { >> for (;;) >> printf ("low priority task running\n"); >> } >> >> hiPrioTask ( >> { >> for (;;) >> { >> taskDelay (0); >> printf ("hi priority task running\n"); >> } >> } >> >> test() >> { >> taskSpawn ("tLow", 110, 0, 10000, lowPrioTask, 0,0,0,0,0,0,0,0,0,0); >> taskSpawn ("tHi", 105, 0, 10000, hiPrioTask, 0,0,0,0,0,0,0,0,0,0); >> } --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Where can I find VxWorks Programmer's Guied or Reference Manual? Date: Mon, 21 Jun 1999 07:08:11 GMT From: bobarmstrong@home.nothiis.net (Robert Armstrong) Organization: @Home Network Message-ID: References: <7kf0v1$sl7$1@jagalchi.cc.pusan.ac.kr> Well, vxWorks is a commercial o/s and you have to buy a licence to develop using it. If you have payed for it then you will also have the manuals which include the programmers guide which, among other things, lists the functions. If you don't have it yet then I am given to understand that the WRS sales people can loan, or perhaps give, you the manual set as a pre-sales incentive. The web site also has the documents on-line in html form, this section is password protected though so again you would need to talk to the WRS sales people. I'm afraid I don't know where you are based but you should be able to find your nearest WRS office or representative from their web site at www.wrs.com. HTH Bob In article <7kf0v1$sl7$1@jagalchi.cc.pusan.ac.kr>, "��" wrote: >Where can I find Library functions and system calls for VxWorks? >How much it is differ with linux? >How much differ VxWorks data types and linux data types? > > --------------------------- End of New-News digest ********************** From owner-vxwexplo-process Mon Jun 21 05:00:15 1999 From: "Woolverton, Eric D." Date: Mon Jun 21 05:00:19 PDT 1999 Subject: VxWorks Tornado II I have been working with Tornado II for a day and a half now and have also seen many bugs/features. Feature 1: You can only remove a single file from a project at a time. Very annoying!! Bug 1: When looking at local variables that are arrays, a 'feature' indicates that 'n' number of entries are exactly alike. OK, but here is the problem: x[0] = 0 x[1] = 1 repeates 5 times x[2] = 0 <-------------------------------This should not be x[2] but x[6] Bug 2: Question, how long does it take to crash Tornado II? Answer, one minute. It takes that long to start up the debugger on this lousy machine of mine. Anyway, with the debugger running, all you have to do is start a compile and the entire Tornado environment crashes. Bug 3: When stopping the debugger, it triggers an exception. Why????? Eric Woolverton From owner-vxwexplo-process Mon Jun 21 06:32:59 1999 From: "Kerry-Gorden Williams" Date: Mon Jun 21 06:33:03 PDT 1999 Subject: Answer: SENS Multiple Interface Drivers for TCP/IP Protocol Hi Robert, I ran into the exactly the same issue a couple months ago. In the BSP directory (tornado/target/config/) the line: #define IP_MAX_UNITS 1 Needs to be changed to a number >= the number of IP interfaces you want to attach. In addition you need to make sure that the devices you are using as IP interfaces are properly initialized. If you need further information, please drop me a line. Cheers! Kerry /* VxWorks Tornado Wind River */ -------------------------------------------- Kerry Williams Software Design Engineer Nortel Networks Broadband Wireless Access Division 14 Fultz Boulevard Winnipeg, Manitoba R3Y 1V3 kerryw@nortelnetworks.com ESN: 771-2407 PH : (204)631-2407 From owner-vxwexplo-process Mon Jun 21 08:26:39 1999 From: "Land Systems LSA2" Date: Mon Jun 21 08:26:43 PDT 1999 Subject: -fwritable-strings VxWorks Tornado Can anyone explain what the -fwritable-strings flag for gcc does, and what its advantages and disadvantages are likely to be. I have some documentation from Mercury computer systems which say I 'MAY' like to use this flag. Pete Gardiner DERA Chertsey UK lsa2@taz.dra.hmg.gb From owner-vxwexplo-process Mon Jun 21 11:02:44 1999 From: Steve Houtchen Date: Mon Jun 21 11:02:48 PDT 1999 Subject: i960vh Tornado Vxworks This message is in MIME format. Since your mail reader does not understand this format, some or all of this message may not be legible. ------_=_NextPart_001_01BEBC0F.D7D3A910 Content-Type: text/plain; charset="iso-8859-1" Has anybody done a BSP port to the Intel i960VH development board that Cyclone built? If so would you share the BSP?? Steve Houtchen Senior Software Engineer Systran Corporation 4126 Linden Avenue, Dayton, OH 45432 (937)252-5601 ext 327 shoutchen@systran.com Steve Houtchen Senior Software Engineer Systran Corporation 4126 Linden Avenue, Dayton, OH 45432 (937)252-5601 ext 327 shoutchen@systran.com ------_=_NextPart_001_01BEBC0F.D7D3A910 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable i960vh Tornado Vxworks

Has anybody done a BSP port = to the Intel i960VH development board that Cyclone built?
If so would you share = the BSP??

Steve = Houtchen
Senior = Software Engineer
Systran = Corporation
4126 Linden = Avenue, Dayton, OH 45432
(937)252-5601 = ext 327
shoutchen@systran.com =

Steve = Houtchen
Senior = Software Engineer
Systran Corporation
4126 = Linden Avenue, Dayton, OH 45432
(937)252-5601 ext 327
shoutchen@systran.com

------_=_NextPart_001_01BEBC0F.D7D3A910-- From owner-vxwexplo-process Tue Jun 22 01:49:42 1999 From: David Laight Date: Tue Jun 22 01:49:47 PDT 1999 Subject: Re: -fwritable-strings > Can anyone explain what the -fwritable-strings flag for > gcc does, and what its advantages and disadvantages are > likely to be. I have some documentation from Mercury computer > systems which say I 'MAY' like to use this flag. Makes strings writable! It causes quoted strings to be placed in the read-write data section instead of the read-only text section. Really old compilers put strings into the data area and generated a separate instance of the string for each use. This allows code like the following to work: char *year = "19xx"; if (year[2] = 'x') { time_t now = time( 0 ); struct tm *tm = localtime( &now ); year[2] = tm->tm_year / 10 + '0'; year[3] = tm->tm_year % 10 + '0'; } Without affecting any other instances of the string "19xx". (Sorry: could not resist a really horrid Y2K bug!) If the strings are in read only data the above will generate a trap (if you have an mmu set to disallow writes to readonly data). This also allows the compiler to use a single instance of a string for all occurrances (in a single module). There once was a system that held the small integers at known fixed locations (something to do with the instruction set not having 'load immediate). The fortran compiler passed everything by reference and someone passed a constant 1 to a routine that incremented its argument. Thereafter EVERY constant 1 on that system (including all the O/S) has value 2. Writeable constants are not a good idea! Unless you have separate limits on instruction and data space (x86 segmented model) I would suggest that you don't want to set -fwriteable strings. David Did I creep a tornado in? ---------------------------------------------------------------- David Laight email: dsl@tadpole.co.uk Tadpole Technology plc phone: +44 1223 278 256 Cambridge, UK fax: +44 1223 278 201 From owner-vxwexplo-process Tue Jun 22 04:00:21 1999 From: daemon@csg.lbl.gov Date: Tue Jun 22 04:00:25 PDT 1999 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Tue Jun 22 04:00:17 PDT 1999 Subject: Re: Putting a task at the end of the READY queue. Subject: Re: questions on memory management Subject: Re: How to Handle "Large" Data File (~50Mb)? Subject: Using the target-based version of ld Subject: Re: Question about SLIP connection Subject: RDM/Vx White Paper Subject: Re: VXWorks start-up/boot-up timings Subject: Re: -fwritable-strings Subject: Re: Question about SLIP connection Subject: Re: Putting a task at the end of the READY queue. Subject: Looking for a cheat sheet for Vx works command line utilities Subject: Re: Is raise() a no-op? Subject: Re: Undefined symbol Subject: Re: Configuring vxWorks for bootp Subject: Re: HDLC Driver for MPC860 Subject: Re: Using the target-based version of ld Subject: ISR answering times Subject: Re: Looking for a cheat sheet for Vx works command line utilities ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Putting a task at the end of the READY queue. Date: Mon, 21 Jun 1999 09:37:10 -0400 From: Douglas Fraser Organization: Lucent Technologies Message-ID: <376E4006.B5084F48@lucent.com> References: <7kk775$7i61@overload.lbl.gov> Fred Roeber wrote: > work because printf interferes with operation too much. > > In this case what I think happens is that the high priority task runs > for a number of interations printing it's message and then rescheduling > itself. At some point, the ring buffer used by the serial driver to > buffer output characters fills up. At this point, the high priority > task blocks allowing the low priority task to run. The low priority > task then does it's output request which gets mixed in. The blocking > action of the printf call is what skews the results since it takes so > long to output characters. To get around this, you could use WindView > to look at things or toggle some external signals in place of the > printf calls and look at them on a logic analyzer. In that case, you > should see no execution of the low priority task. Fred Exactly. Another very simple thing to try would be assign global variables to each task and have the task increment its global as it runs. The low priority task counter will be at zero at the end of your test if the high priority task never blocks. Printf is dangerous in your 'real-time' app. If the ring buffer fills, your task will block, and the time to completion for that printf call will be the time it takes to dump the contents of the message on stout. If you have a slow 9600 baud link, that is - -r-e-a-l-l-y- slow. Properly used, logMsg is much more predictable. Doug --------------------------- Newsgroups: comp.os.vxworks Subject: Re: questions on memory management Date: Sun, 20 Jun 1999 17:37:07 +0300 From: Leonid Rosenboim Organization: Bezeq International Ltd. ISRAEL Message-ID: <376CFC93.FD34A511@bitband.com> References: <7kkge1$np$1@nnrp1.deja.com> vivek1796@my-deja.com wrote: > hi everybody, > > I would like to know how I can free the memory allocated by > memPartAlignedAlloc(). Will memPartFree() work??? > memPartFree() will no doubt release that chunk of memory and join it with neighbouring chunks if these are free. Essentially memalign() is the same as malloc(0 except it will potentially split a free chunk into three chunks, allocate the middle one and leave the other two as free, all for the sake of keeping the required alignment. > > Also, I would appreciate if anybody could tell me where memalign() is > preferred over malloc(). > These are rare cases where the HARDWARE requires aligned addresses for imporved performance or for any kind of success at all. One such specific case is when you need to play with the MMU's attributes or mappings, you can do it in whole PAGES so one way to get a whole PAGE(s) is to memalign() on page boundaries a chink which is multiple of page size. Hope this helps, Leonid --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How to Handle "Large" Data File (~50Mb)? Date: Mon, 21 Jun 1999 14:02:32 GMT From: jrm@mclean.sparta.com Organization: Deja.com - Share what you know. Learn what you don't. Message-ID: <7klglg$b04$1@nnrp1.deja.com> References: <37695B37.1433F7AA@dfrc.nasa.gov> <92960669@NEWS.SAIC.COM> The default network driver is actually called netDrv. It uses either rsh or ftp as its communications protocol, and it does copy the entire file into local memory when you open it (if you have entered a password in the boot configuration parameters it will use FTP, otherwise RSH. when you create your own netDrv you specify which protocol you want in netDevCreate) NFS uses RPC and UDP for its data transfer and works with small pieces of the file at a time. It does use a lot more stack, so each task that will be using NFS should add about 7K more stack space. John Moore In article <92960669@NEWS.SAIC.COM>, "David Wiggins" wrote: > If your 50Mb file is mounted on your host machine, we had this exact problem > (although on a different board, a Radstone PPC 604e running vxWorks 5.3.1). > If the file is on a target-mounted drive (e.g. Scsi), then disregard this > message. > > The problem for us was that the default network driver is RPC, which loads > the ENTIRE file across the network into target memory on the fopen call. So > the size of the file you're able to open is limited by the amount of target > memory available at the time. It looks like this is problem you're having, > as evidenced by the memPartAlloc error. We also got this error along with > the Broken pipe error you're getting. > > So, our solution was merely to nfsMount the host partition on our target, > then all the open, read, write, close, etc. functions get routed through the > nfs driver instead of the rpc driver. The nfs driver does not have this > "feature" of the rpc driver, but instead works across the network as you > read/write the file (with some reasonable amount of memory buffering, of > course). > > Hope this helps, > > Dave Wiggins > SAIC-San Diego > > Ken Norlin wrote in message > news:37695B37.1433F7AA@dfrc.nasa.gov... > > Hi, > > > > I'm trying to open a ~50Mb file. When I do an fopen on the file, I get > the > > following error: > > 0x17c47f0 (tShell): memPartAlloc: block too big - 524 in partition > 0x19b9f4. > > ISRS.DAT: Broken pipe. > > > > If I cut the file down to 14 Mb, I'm able to open the file & read the > data. > > So it must be the file size that's causing the problem. > > > > I'm using an MVME2604 w/ BSP 1.1/5. Please let me know if you've run > across > > this problem and how you solved it. (I'm assuming there must be some way > to > > bump up the memory buffer size). > > > > Thanks a bunch, > > Ken N. > > > > > > -- > > Ken Norlin --- Simulation Engineer > > NASA Dryden Flight Research Center > > ken.norlin@dfrc.nasa.gov > > (661) 258-2046 FAX (661) 258-2792 > > Sent via Deja.com http://www.deja.com/ Share what you know. Learn what you don't. --------------------------- Newsgroups: comp.os.vxworks Subject: Using the target-based version of ld Date: Mon, 21 Jun 1999 18:33:15 GMT From: rsg@nospam4me.com (Ralph Grommet) Organization: Defence Evaluation & Research Agency Message-ID: <7klntd$1vj$1@trog.dera.gov.uk> Hi I'm transitioning my project from host-based development (using tgtsvr, etc) to target-based development using rlogin. The target is a Motorola PPC MVME2307. The host is a SPARC running Solaris. I am using the standard gnu development tools. When I rlogin onto the target and download a (working) executable using ld() I get the following error message scrolling about 1 million times: "Relocation value does not fit in 24 bits." This message is generated by the loadModule() call within ld() and as a result nothing is added to the target symbol table. The executable loads properly if I do an ld() from within a tgtsvr shell. I suspect my executable is of the wrong flavour to be read by the target-based version of ld() but I have no idea what compiler switches/linking phases I need to apply/remove. I currently do an "ldppc" with the -r option as the final phase of compilation to produce the executable I download. If anyone has seen this before or can correct my gross misconceptions about what is happening I would much appreciate some advice. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Question about SLIP connection Date: Mon, 21 Jun 1999 15:58:27 GMT From: saustin_deja@my-deja.com Organization: Deja.com - Share what you know. Learn what you don't. Message-ID: <7klnes$dpd$1@nnrp1.deja.com> References: <7kipgl$3qn1@overload.lbl.gov> "Slutsker, Rasty" wrote: > Does anybody know how to have SLIP/PPP in Windows over raw COM1/2 port > (without modem)? Go to a shareware site like www.winfiles.com and download the appropriate (NT, 95, 98) "Null Modem Cable" device driver. After installation, the device will be available to your dial-up networking/RAS software. Sent via Deja.com http://www.deja.com/ Share what you know. Learn what you don't. --------------------------- Newsgroups: comp.os.vxworks Subject: RDM/Vx White Paper Date: Mon, 21 Jun 1999 10:35:09 -0700 From: "Alexander V. Voinov" Organization: Posted via RemarQ, http://www.remarQ.com - The Internet's Discussion Network Message-ID: <376E77CD.2EBBA053@raima.com> Hi All, Having seen here several times questions about databases in general and Raima Database Manager in particular, I'm brave enough to announce my white paper on databasing for VxWorks, a general overview, which (surprisingly :-) covers our C/C++ API database engine with small footprint. http://www.raima.com/wrs Regards Alexander --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VXWorks start-up/boot-up timings Date: Sat, 19 Jun 1999 15:34:25 +0300 From: Leonid Rosenboim Organization: Bezeq International Ltd. ISRAEL Message-ID: <376B8E51.915D5E81@bitband.com> References: <7k7tf8$ec1@overload.lbl.gov> "Tapia, Eduardo" wrote: > We are trying to use vxworks for an embedded system that needs to be up and > running (including recording to flash) whithin 250ms. > The processor will be a Motorola 860T running at 50 Mhz. It seems there is insufficient detail in this post to determine the cause of this delay, so here is my best 2 guesses, and a mathod to figure out the true answer: Ny best guess is that the MMu intialization routine, which builds page tables for every peice of RAM/ROM and Peripheral address space on the board. This is done weather you use MMU_BASIC or MMU_FULL. My second best guess is that some of this delay has to do with the Flash programming you mentioned above. Mind you, Flash writes are 6-14 usec per byte and erase of a sector is 10 to 900 milliseconds. My third best answer ;-)) is that the startup time is due to a few factors, not just one. So here is how you can MEASURE your start up time distribution: You can modify the usrCOnfig.c startup routine (yep, these are provided in source so you could modify them every now and then), and insert a statement which emits a timing signal between evgery two calls to any library or driver intialization function. The timing signal you emit could be anything from a pulse on a diescrete output line measured by a digital scope, to a simple printf to the console serial port. Chose your preferred method based on required prevision and available equipment. Note that printf will work only after the kernel, BSP and serial drivers have been initialized, i.e. after tyDrvCreate(), while descrete puls can be done from any code. Hope this helps, Leonid --------------------------- Newsgroups: comp.os.vxworks Subject: Re: -fwritable-strings Date: Mon, 21 Jun 1999 15:25:31 -0400 From: Joe Durusau Organization: Lockheed Martin Corporation Message-ID: <376E91AB.E2F3340D@dragon.decnet.lmco.com> References: <199906211526.IAA28729@SpamWall.lbl.gov> According to the gcc manual I have, it makes the compiler set up each string so that it is stored in a unique place in memory. Normally, the compiler would overlap strings if it could. The idea is to allow programs to write into string constants, which the compiler folk consider nasty. Speaking only for myself, Joe Durusau Land Systems LSA2 wrote: > > VxWorks Tornado > > Can anyone explain what the -fwritable-strings flag for > gcc does, and what its advantages and disadvantages are > likely to be. I have some documentation from Mercury computer > systems which say I 'MAY' like to use this flag. > > Pete Gardiner > DERA Chertsey > UK > lsa2@taz.dra.hmg.gb --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Question about SLIP connection Date: Mon, 21 Jun 1999 21:10:10 GMT From: "Clifford Slocombe" Message-ID: References: <7kipgl$3qn1@overload.lbl.gov> <7klnes$dpd$1@nnrp1.deja.com> My NT station has a null modem cable connection already listed under Standard Modems (select don't detect modem)., I installed it from the NT distribution CD. But now what do I do to connect to VxWorks? I have built in the pppLib, when I do pppInit, I get an LCP failure. If I set it to passive mode (I am guessing what this is for), and 'dial-up' from the NT end I still get nothing. I have tested the link using Hyperterminal, and a simple character echo program on the VxWorks target. Clifford. wrote in message news:7klnes$dpd$1@nnrp1.deja.com... > "Slutsker, Rasty" wrote: > > Does anybody know how to have SLIP/PPP in Windows over raw COM1/2 port > > (without modem)? > Go to a shareware site like www.winfiles.com and download the > appropriate (NT, 95, 98) "Null Modem Cable" device driver. After > installation, the device will be available to your dial-up > networking/RAS software. > > > Sent via Deja.com http://www.deja.com/ > Share what you know. Learn what you don't. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Putting a task at the end of the READY queue. Date: Mon, 21 Jun 1999 18:24:34 -0400 From: Tim Sohacki Organization: Nortel (Northern Telecom) Message-ID: <376EBBA2.6F083AE8@nortelnetworks.com> References: <7k809n$hbd$1@nnrp1.deja.com> <9F1EA55D2B1016B4.5BD5B7FE9D091CF6.77BE3D78F3CD18CA@lp.airnews.net> <376BC606.E9E3D083@nortelnetworks.com> > I changed the "printf" to "logMsg" and was surprised > at the result: > > ^M0x1925588 (tHi): hi priority task running 1 > ^M0x1925588 (tHi): hi priority task running 2 > ^M0x1925588 (tHi): hi priority task running 3 > > > > ^M0x1925588 (tHi): hi priority task running 66th > ^M0x1925588 (tHi): hi priority task running 67th (about) > ^M0x1927eb0 (tLow): low priority task running > ^M0x1925588 (tHi): hi priority task running > ^M0x1927eb0 (tLow): low priority task running > ^M0x1925588 (tHi): hi priority task running > ... etc ... > I assumed that when the logMsg() message queue was full, the caller would NOT block, and the message would just be tossed. This is not the case when a task calls logMsg() (though it sure better be for ISR's!). Instead, the msgQSend wait parameter is "WAIT_FOREVER" not "NO_WAIT." The above results are explained as follows: The message queue fills around the time the 66th message is output, blocking tHi so that tLow runs. Once this happens, tHi and tLow continue to block on every execution, because tLogTask can only unload one message from its queue every 20+ ms, much longer than it takes either task to run. --------------------------- Newsgroups: comp.os.vxworks Subject: Looking for a cheat sheet for Vx works command line utilities Date: Tue, 22 Jun 1999 02:56:53 GMT From: the_rileys@yahoo.com (Riley) Message-ID: <376efa3c.9439505@news.lvdi.net> Meets and greets.. I'm looking for a listing of VxWorks command line utilities, with syntax / switches...PLEASE! I will put you in my will... - - Riley --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Is raise() a no-op? Date: Mon, 21 Jun 1999 19:35:51 -0700 (PDT) From: silvermang@webtv.net (Greg Silverman) Organization: WebTV Subscriber Message-ID: <21039-376EF687-83@newsd-222.iap.bryant.webtv.net> References: what are t101 and t2? how i find out whati i have? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Undefined symbol Date: Tue, 22 Jun 1999 09:55:40 +0200 From: Leonid Rosenboim Organization: BitBand Technologies Ltd. Message-ID: <376F417C.BB4B44C7@bitband.com> References: <376587B7.45E6FBE1@metrolink.net> Peter wrote: > [snip] > > -> ld < util.o > Undefined symbol: _$_13Iostream_init (binding 1 type 0) > value = 24799816 = 0x17a6a48 Peter, your util.o is a result of a C++ source which uses C++ iostreams, but your VxWorks image does not contain the iostreams package, which is supposedly part of standard Tornado, but require you to define the INCLUDE_CPLUS and INCLUDE_CPLUS_IOSTREAMS options in your VxWorks configuration. HTH, Leonid --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Configuring vxWorks for bootp Date: Tue, 22 Jun 1999 09:58:23 +0200 From: Leonid Rosenboim Organization: BitBand Technologies Ltd. Message-ID: <376F421F.73C6773D@bitband.com> References: <37659104.8832417C@swbell.net> Indeed, SunOS does not come bundled with a BOOTP server (Sun had RARP as its standard diskless booting and configuration via another RPC-based protocol). Your Tornado tree has a public domain bootp server that should work on SunOS. Look in $WIND_BASE/target/unsupported/ tree. Matt Henderson wrote: > Can somebody provide an example of how you configure the SunOS 4.1.4 > /etc/bootparams file to support booting a Heurikon Baja 4700 single > board computer with a VXworks bootstrap kernel? HTH, Leonid --------------------------- Newsgroups: comp.os.vxworks Subject: Re: HDLC Driver for MPC860 Date: Tue, 22 Jun 1999 10:09:38 +0200 From: Leonid Rosenboim Organization: BitBand Technologies Ltd. Message-ID: <376F44C1.4E547A0D@bitband.com> References: The HDLC driver part should be pretty easy. However, the standard VxWorks PPP (SENS and pre-SENS) will not work over HDLC, so you will have to purchase PPP source code from Wind River to make it work over HDLC. Another option is to buy PPP from another company (Wind Partner) which already has got it working over HDLC. Raj K Meel wrote: > Hi, > > Does anyone know of a good industrial quality HDLC driver for MPC860. > We need to use it for PPP over HDLC over T1. > > Thanks > Raj Meel HTH, Leonid --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Using the target-based version of ld Date: Tue, 22 Jun 1999 09:51:50 +0200 From: Magnus L Andersson Organization: Ericsson Microwave Systems AB Message-ID: <376F4096.A3CD3291@emv.ericsson.se> References: <7klntd$1vj$1@trog.dera.gov.uk> Reply-To: magnus.l.andersson@gbg.frontec.se I can't help you with the swithces and such, but my guess is that you've got a lot of relative bracnhes in your code (e.g. OS calls) and you're placing your code more than 24 bits (16 Mb) away from the OS memory location. The relative jumps in the PPC architecture (32 bit at least) only support 24 bit relative branches according to the EABI (Embedded Application Binary Interface). For a reference on the EABI, check out http://www.chips.ibm.com/techlib/products/powerpc/appnotes.html ('Developing PowerPC Embedded Application Binary Interface (EABI) Compliant Programs') Magnus L Andersson Frontec Konsulter Goteborg AB Ralph Grommet wrote: > > Hi > > I'm transitioning my project from host-based development (using > tgtsvr, etc) to target-based development using rlogin. The target is a > Motorola PPC MVME2307. The host is a SPARC running Solaris. I am using > the standard gnu development tools. > > When I rlogin onto the target and download a (working) executable > using ld() I get the following error message scrolling about 1 million > times: > > "Relocation value does not fit in 24 bits." > > This message is generated by the loadModule() call within ld() and as > a result nothing is added to the target symbol table. The executable > loads properly if I do an ld() from within a tgtsvr shell. > > I suspect my executable is of the wrong flavour to be read by the > target-based version of ld() but I have no idea what compiler > switches/linking phases I need to apply/remove. I currently do an > "ldppc" with the -r option as the final phase of compilation to > produce the executable I download. > > If anyone has seen this before or can correct my gross misconceptions > about what is happening I would much appreciate some advice. --------------------------- Newsgroups: comp.os.vxworks Subject: ISR answering times Date: Tue, 22 Jun 1999 11:07:20 +0200 From: Christian Richter Organization: SBS-or Industrial Computers GmbH Message-ID: <376F5248.E6FAE73E@or-computers.de> Reply-To: c.richter@or-computers.de Hello, I wrote a SW that processes communication over a MIL-STD-1553 using VxWorks on a Pentium Processor Board. After every transmission an interrupt occurs. My Interrupt Service Routine will normally be executed after 2 or 3 microseconds. This time was measured using an oscilloscope. Every some hundred program cycles the answering time of the ISR is 100 microseconds or more. What may be the reason? A second interrupt which follows will be lost due to this fact. The long answering times occur irregularily but the program loops are identical. So I do not think that there is an error in the software. Does anybody know about the problem of long ISR answering times at VxWorks? C. Richter --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Looking for a cheat sheet for Vx works command line utilities Date: Tue, 22 Jun 1999 13:22:35 +0400 From: "Victor V. Vengerov" Organization: TEPKOM Message-ID: <376F55DB.FBDFDEB2@tepkom.ru> References: <376efa3c.9439505@news.lvdi.net> Hi, Riley wrote: > Meets and greets.. > > I'm looking for a listing of VxWorks command line utilities, with > syntax / switches...PLEASE! > Do you mean VxWorks target shell, isn't it? In fact VxWorks shell understand C-like syntax for usual C procedures call. You can invoke any global C procedure (defined either in VxWorks library or in your application) and pass parameters for it. If some parameters is omitted at end, value 0 will be passed. Also, you can see (or set) value of variable entering its name or assignment operator. Nothing more! Some useful "utility" procedures intended for using primary from shell defined in usrLib. Read reference manual for more information. Please, note: no any type checks etc performed during function invocation. It is your responsibility to pass correct parameters! > > I will put you in my will... > > - Riley Best regards. Victor - -- Victor V. Vengerov --------------------------- End of New-News digest ********************** From owner-vxwexplo-process Tue Jun 22 05:25:50 1999 From: tim allen Date: Tue Jun 22 05:25:54 PDT 1999 Subject: Re: Using the target-based version of ld >I'm transitioning my project from host-based development (using >tgtsvr, etc) to target-based development using rlogin. The target is a >Motorola PPC MVME2307. The host is a SPARC running Solaris. I >am using the standard gnu development tools. The difference between host-based and target based development, as far as the problem you describe is concerned, is that for host based development the dynamically loaded modules are loaded into the memory pool at the bottom of memory, but for target-based development, ld loads the modules from the top of memory. The relevance of this should become clear in a moment! >When I rlogin onto the target and download a (working) executable >using ld() I get the following error message scrolling about 1 million >times: > >"Relocation value does not fit in 24 bits." This is because the PowerPC architecture limits relative addressing to 24-bit offsets to conform to the EABI standard. Since you are now loading your modules at the top of memory, the dynamic linking is unable to resolve the offsets in the 24 bits. >This message is generated by the loadModule() call within ld() and as a >result nothing is added to the target symbol table. The executable loads >properly if I do an ld() from within a tgtsvr shell. Because it goes in low memory and the relative address fits in 24 bits. >I suspect my executable is of the wrong flavour to be read by the >target-based version of ld() but I have no idea what compiler >switches/linking phases I need to apply/remove. I currently do an >"ldppc" with the -r option as the final phase of compilation to >produce the executable I download. > >If anyone has seen this before or can correct my gross misconceptions >about what is happening I would much appreciate some advice. The solution we used is to boot VxWorks with LOCAL_MEM_SIZE set to the maximum, but with USER_RESERVED_MEM set to leave a maximum 32MB memory pool, load all the modules you want, and then if you want more memory use memAddToPool to add the reserved memory over 32MB. However I believe there are more details and some workarounds described in the harware/ppc FAQ on the WRS WindSurf site. Regards, Tim. Tim Allen Ltd Real Time Embedded Software Consultancy ______________________________________________________ Get Your Private, Free Email at http://www.hotmail.com From owner-vxwexplo-process Tue Jun 22 08:01:17 1999 From: "Tapia, Eduardo" Date: Tue Jun 22 08:01:22 PDT 1999 Subject: VXWorks start-up/boot-up timings /// Network (TCP/IP) /// pSOS VXWorks Hello again. Sorry it took a bit to reply, but we have done it now. We manage to start it in about 60 ms. The bare system. It was taking a long time in clearing the memory. Thanks a lot to everyone who helped. Now the problem is to load the TCP/IP stack during runtime, because if we load it at start-up it takes way too long. Any ideas about this? We have asked WindRiver Tech. Supp. but they don't seem to be very quick in responding. Has anyone experienced any problems with their technical support? We have got pSOS also working within that timescale and they seem more helpful. Any comments against or for Integrated Systems (pSOS)? Eduardo. Eduardo Tapia Software Engineer Engineering Dept. Penny & Giles Aerospace Ltd. From owner-vxwexplo-process Tue Jun 22 09:18:38 1999 From: Leonid Rosenboim Date: Tue Jun 22 09:18:43 PDT 1999 Subject: Re: VXWorks start-up/boot-up timings /// Network (TCP/IP) /// pSOS "Tapia, Eduardo" wrote: > Sorry it took a bit to reply, but we have done it now. We manage to start it > in about 60 ms. The bare system. It was taking a long time in clearing the > memory. > Yes, the standard bootInit code will cler all emmory when COLD boot, assuming your DRAM may have parity to be cleared. Also on many BSPs the reboot allways gets the COLD flag. so this is also often a time consuming issue. > > Thanks a lot to everyone who helped. > > Now the problem is to load the TCP/IP stack during runtime, because if we > load it at start-up it takes way too long. This can be done if you simply comment out the call to usrNetInit() or underfine the macro which contains this call in usrConfig.c, and sometime later when you got the time do the same call. We have done this many times by building with STANDALONE and not definint STANDALONE_NET_INIT, and later from Shell we do usrBootLineInit(); useNetInit(); HTH, - Leonid From owner-vxwexplo-process Tue Jun 22 11:42:00 1999 From: Reyhan ERGUN Date: Tue Jun 22 11:42:04 PDT 1999 Subject: using same vxWorks on UNIX and NT platforms This is a multi-part message in MIME format. --------------308A945FCDE6C0B45AD44642 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit I use MBX860 Board, MBX860 BSP, vxworks,.... I use Tornado I on both UNIX and NT platforms. I didn't encounter so many problems while working on NT, however working on UNIX is very annoying. Using CrossWind on UNIX crashes Tornado very often. I am not sure the problem is because of EXCEED which I use to connect UNIX machine from my PC or the tool itself. Anyway, I would like to ask if I could make vxWorks on UNIX, and use Tornado tools from my PC, do all debugging work from PC... I tried this but it doesn' work!....I encountered problems from the shell as follows: -> i NAME ENTRY TID PRI STATUS PC SP ERRNO DELAY ---------- ------------ -------- --- ---------- -------- -------- ------- ----- value = 0 = 0x0 -> devs drv name value = 0 = 0x0 -> iosFdShow fd name drv WTX Error 0x100df (AGENT_GOPHER_SYNTAX) value = -1 = 0xffffffff -> printErrno 0x0 = OK value = 0 = 0x0 ......etc. I should make vxWorks on UNIX, because I do some work integrated with another tool working on UNIX. I get vxWorks file via ftp from UNIX to PC, I tried on some other object file which are get from UNIX to PC by the same way, I can load this files from PC to the target and these applications can run on the target without any problem. Is there any utility to obtain a vxWorks made on UNIX platorm which can be run while host tool is running on NT platform? Any comments would be appreciated, Regards, Reyhan --------------308A945FCDE6C0B45AD44642 Content-Type: text/x-vcard; charset=us-ascii; name="vcard.vcf" Content-Transfer-Encoding: 7bit Content-Description: Card for Reyhan (FINDIK) ERGUN Content-Disposition: attachment; filename="vcard.vcf" begin: vcard fn: Reyhan (FINDIK) ERGUN n: ;Reyhan (FINDIK) ERGUN org: ASELSAN INC. adr: Microwave and System Technologies Division ;;Software Engineering Dept.;ANKARA;;;TURKEY email;internet: findik@mst.aselsan.com.tr tel;work: (90-312) 385 19 00 (Ext. 2668) tel;fax: (90-312) 354 52 05 note: http://www.aselsan.com.tr x-mozilla-cpt: ;0 x-mozilla-html: FALSE version: 2.1 end: vcard --------------308A945FCDE6C0B45AD44642-- From owner-vxwexplo-process Tue Jun 22 18:33:06 1999 From: Fred Roeber Date: Tue Jun 22 18:33:09 PDT 1999 Subject: Re: using same vxWorks on UNIX and NT platforms On Tue, 22 Jun 1999, Reyhan ERGUN wrote: > I should make vxWorks on UNIX, > because I do some work integrated with another tool working on UNIX. > I get vxWorks file via ftp from UNIX to PC, > I tried on some other object file which are get from UNIX to PC by the > same way, > I can load this files from PC to the target and these applications can > run on the target without any problem. Make sure you transfer the files in binary mode. > Is there any utility to obtain a vxWorks made on UNIX platorm which can > be run while host tool is running on NT platform? I might suggest running SAMBA on your UNIX machine and then storing all your VxWorks files on the UNIX machine. SAMBA is a publically available (free) program that can let a Unix machine export its disk file systems using the SMB protocol that Windows uses for remote file access. Basically, you run the SAMBA daemon on your Unix machine telling it to "export" the file system with all the VxWorks software and your application stuff. You can then use the "map network drive" commands on your PCs to get easy access to the files from the PC. Then when you work on the PC using the remote mounted disks the file transfers are handled automatically. Setting all this up isn't trivial but it isn't that hard either and is well explained in the documentation that comes with the Samba distribution. I don't have a web address handy for where to get Samba but it is readily available. Fred | Fred J Roeber, BBN Systems & Technologies | | 4 John Clarke Road Middletown, RI 02842-5202 | | froeber@bbn.com 401-848-3548 | | TraceMaker product manager | | http://www.tracemaker.bbn.com | From owner-vxwexplo-process Tue Jun 22 19:25:37 1999 From: Fred Roeber Date: Tue Jun 22 19:25:41 PDT 1999 Subject: Re: ISR answering times On Tue, 22 Jun 1999, Christian Richter wrote: > I wrote a SW that processes communication over a MIL-STD-1553 using > VxWorks on a Pentium Processor Board. After every transmission an > interrupt occurs. > My Interrupt Service Routine will normally be executed after 2 or 3 > microseconds. This time was measured using an oscilloscope. Every some > hundred program cycles the answering time of the ISR is 100 microseconds > or more. > What may be the reason? > A second interrupt which follows will be lost due to this fact. VxWorks locks out all interrupts when it has to perform critical kernel functions (actually, the intLock/intUnlock functions are used which don't have to lock out all interrupts but normally do). The time that interrupts are locked is typically small though (much less than 100 usec) since the "longer" kernel critical section synchronization is done using a different mechanism based on a "kernelState" variable which doesn't involve locking interrupts. Thus I don't think your delay is a function of this. User code can also call the intLock/intUnlock functions to lock out interrupts. I assume you don't have any code that is running that could lock out interrupts for a long time? The most likely cause of the delay is that the processor is handling some higher priority interrupt periodically which holds yours off. I don't know the X86 VxWorks timing details but 100 usec would be about right for how long it might take to handle the periodic clock interrupt on a 68K processor. If you had WindView you could figure this out easily since you would see the interrupt processing directly. I suggest checking the priority levels that are set up for the different interrupt sources. With Moto processors I have often adjusted the levels to get better response to your critical interrupts. In extreme cases, I have even made it so that VxWorks won't lock out certain interrupt levels by using the intLockLevelSet function to not mask my high priority interrupt. If you do this then be very careful what you do in the interrupt handler; you can't use any VxWorks functions that use critical sections. Fred | Fred J Roeber, BBN Systems & Technologies | | 4 John Clarke Road Middletown, RI 02842-5202 | | froeber@bbn.com 401-848-3548 | | TraceMaker product manager | | http://www.tracemaker.bbn.com | From owner-vxwexplo-process Tue Jun 22 22:36:55 1999 From: "Slutsker, Rasty" Date: Tue Jun 22 22:36:59 PDT 1999 Subject: RE: ISR answering times This message is in MIME format. Since your mail reader does not understand this format, some or all of this message may not be legible. ------_=_NextPart_001_01BEBD42.B682CB0E Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable We've seen the similar problem. When we tested our application = (originally targeted to 486) with the Pentium board, we've figured out that = interrupt latency with Pentium is significantly BIGGER than with 486. Many modern Pentium boards have built-in VGA, which shares the RAM with CPU. I = guess that VGA periodically takes local bus to access the VGA RAM, thus = halting the CPU and decreasing interrupt responsiveness. Also Pentium has = bigger cache that may affect determinism of application. Rasty ---------- From: vxwexplo@lbl.gov [SMTP:vxwexplo@lbl.gov] Sent: =E9=E5=ED =F8=E1=E9=F2=E9 23 =E9=E5=F0=E9 1999 04:26 To: vxworks_users@csg.lbl.gov Subject: Re: ISR answering times Submitted-by owner-vxwexplo-process Tue Jun 22 19:25:37 1999 Submitted-by: Fred Roeber On Tue, 22 Jun 1999, Christian Richter wrote: > I wrote a SW that processes communication over a MIL-STD-1553 using > VxWorks on a Pentium Processor Board. After every transmission an > interrupt occurs. > My Interrupt Service Routine will normally be executed after 2 or 3 > microseconds. This time was measured using an oscilloscope. Every some > hundred program cycles the answering time of the ISR is 100 microseconds > or more. > What may be the reason? > A second interrupt which follows will be lost due to this fact. VxWorks locks out all interrupts when it has to perform critical kernel =09 ------_=_NextPart_001_01BEBD42.B682CB0E Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable RE: ISR answering times

We've seen the = similar problem. When = we tested our application = (originally targeted to = 486) with the Pentium board, we've = figured out that interrupt latency with Pentium is significantly BIGGER than with 486. Many = modern Pentium boards have built-in VGA, which shares the RAM with CPU. = I guess that VGA periodically takes local bus to access the VGA RAM, thus halting the CPU and decreasing = interrupt responsiveness. Also Pentium has = bigger cache that may = affect determinism of = application.

Rasty

----------

From: = vxwexplo@lbl.gov [SMTP:vxwexplo@lbl.gov]

Sent: =E9=E5=ED =F8=E1=E9=F2=E9 23 = =E9=E5=F0=E9 1999 04:26

To: = vxworks_users@csg.lbl.gov

Subject: Re: ISR = answering times

Submitted-by = owner-vxwexplo-process Tue = Jun 22 19:25:37 1999

Submitted-by: Fred Roeber = <froeber@bbn.com>

On Tue, 22 Jun 1999, = Christian Richter <c.richter@or-computers.de> wrote:

> I wrote a SW that = processes communication over a MIL-STD-1553 using

> VxWorks on a = Pentium Processor Board. After = every transmission an

> interrupt = occurs.

> My Interrupt Service = Routine will normally be executed after 2 or 3

> microseconds. This = time was measured using an oscilloscope. Every some

> hundred program = cycles the answering time of the = ISR is 100 microseconds

> or more.

> What may be the = reason?

> A second interrupt = which follows will be lost due to this fact.

VxWorks locks out all = interrupts when it has to perform critical kernel

------_=_NextPart_001_01BEBD42.B682CB0E-- From owner-vxwexplo-process Wed Jun 23 01:49:00 1999 From: David Laight Date: Wed Jun 23 01:49:04 PDT 1999 Subject: Re: ISR answering times On Tue, 22 Jun 1999, Christian Richter wrote: > My Interrupt Service Routine will normally be executed after 2 or 3 > microseconds. This time was measured using an oscilloscope. Every some > hundred program cycles the answering time of the ISR is 100 microseconds > or more. Last time I had a similar problem we located the offending code as follows: 1) Connect a logic analiser to the interrupt request line, and trigger on it being active for > 100us (in your case). 2) Connect the trigger out of the analiser (on the back somewhere) to drive a very high priority interrupt (eg the NMI) 3) use the NMI to generate a stack traceback. Saving the registers and leaping into a 'rom debugger' will probably do. After all you don't need to return from the NMI! In our case it was Unix (not VxWorks) and it was the console scroll that was locking out interupts for > 10ms and causing our X.25 to fail to receive back to back packets. But the principle is the same. David ---------------------------------------------------------------- David Laight email: dsl@tadpole.co.uk Tadpole Technology plc phone: +44 1223 278 256 Cambridge, UK fax: +44 1223 278 201 From owner-vxwexplo-process Wed Jun 23 04:00:22 1999 From: daemon@csg.lbl.gov Date: Wed Jun 23 04:00:28 PDT 1999 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Wed Jun 23 04:00:18 PDT 1999 Subject: Re: Using the target-based version of ld Subject: Re: Using the target-based version of ld Subject: Solaris Icon Problem Subject: Emacs in Tornado on NT Subject: Re: VXWorks start-up/boot-up timings /// Network (TCP/IP) /// pSOS Subject: Re: ISR answering times Subject: Where is Vxworks task scheduler? Subject: Re: Question about SLIP connection Subject: Re: PPC860 GDB hosted on linux/x86 Subject: Re: Where is Vxworks task scheduler? Subject: Re: Question about SLIP connection Subject: Re: Solaris Icon Problem Subject: Re: Is raise() a no-op? Subject: Tornado 2 and Dependencies Subject: Re: Problems in sending data over socket Subject: Contract positions in Massachusetts Subject: Re: SENS Multiple Interface Drivers for TCP/IP Protocol Subject: Re: etherInputHookAdd Subject: Re: Configuring vxWorks for bootp Subject: Re: UDP datagram size and the MTU Subject: Re: Symbol table won't download Subject: Re: SCSI Problem for MCP750 BSP Subject: Re: Tornado 2 and Dependencies Subject: static linking vs. dynamic link Subject: a problem while linking Subject: Re: Emacs in Tornado on NT Subject: Re: Is raise() a no-op? Subject: Re: Where is Vxworks task scheduler? Subject: MVME1604 64MB memory board available Subject: Re: Games on vxworks ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Using the target-based version of ld Date: Tue, 22 Jun 1999 08:45:07 -0400 From: Douglas Fraser Organization: Lucent Technologies Message-ID: <376F8553.A8AD1B93@lucent.com> References: <7klntd$1vj$1@trog.dera.gov.uk> Ralph, How much memory do you have TOTAL? We had a similar problem on a product I worked on last year. The target had 64MB of SDRAM, which is WAY outside the 24 bit address limit. loadModule() calls malloc() which always allocates from the top of the heap first, which means you get something out around 64MB in our case. The solution was to reserve 32MB at the top of memory until the dynamic loading phase was complete, then add that reserved space to the heap with memAddToPool(). The other solution is manage your own private heap space for loading and use loadModuleAt(). I found the method we used to be very easy to administer. Doug Ralph Grommet wrote: > > Hi > > I'm transitioning my project from host-based development (using > tgtsvr, etc) to target-based development using rlogin. The target is a > Motorola PPC MVME2307. The host is a SPARC running Solaris. I am using > the standard gnu development tools. > > When I rlogin onto the target and download a (working) executable > using ld() I get the following error message scrolling about 1 million > times: > > "Relocation value does not fit in 24 bits." > > This message is generated by the loadModule() call within ld() and as > a result nothing is added to the target symbol table. The executable > loads properly if I do an ld() from within a tgtsvr shell. > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Using the target-based version of ld Date: Tue, 22 Jun 1999 14:12:58 GMT From: wliu5011@my-deja.com Organization: Deja.com - Share what you know. Learn what you don't. Message-ID: <7ko5ku$9rj$1@nnrp1.deja.com> References: <7klntd$1vj$1@trog.dera.gov.uk> In article <7klntd$1vj$1@trog.dera.gov.uk>, rsg@nospam4me.com (Ralph Grommet) wrote: Add the switch -mlongcall for compiling. I have met the similar problem before. Good Luck! Wenzhong Liu Senior Software Engineer CSP Inc. Billerica, MA. > Hi > > I'm transitioning my project from host-based development (using > tgtsvr, etc) to target-based development using rlogin. The target is a > Motorola PPC MVME2307. The host is a SPARC running Solaris. I am using > the standard gnu development tools. > > When I rlogin onto the target and download a (working) executable > using ld() I get the following error message scrolling about 1 million > times: > > "Relocation value does not fit in 24 bits." > > This message is generated by the loadModule() call within ld() and as > a result nothing is added to the target symbol table. The executable > loads properly if I do an ld() from within a tgtsvr shell. > > I suspect my executable is of the wrong flavour to be read by the > target-based version of ld() but I have no idea what compiler > switches/linking phases I need to apply/remove. I currently do an > "ldppc" with the -r option as the final phase of compilation to > produce the executable I download. > > If anyone has seen this before or can correct my gross misconceptions > about what is happening I would much appreciate some advice. > > Sent via Deja.com http://www.deja.com/ Share what you know. Learn what you don't. --------------------------- Newsgroups: comp.os.vxworks Subject: Solaris Icon Problem Date: Tue, 22 Jun 1999 08:28:19 -0700 From: Jeff King Message-ID: <376FAB93.60992103@radixtek.com> Hi, More of an adminstrative question here. Running Tornado 1.0.1 on a recently upgraded solaris7 box. When I do a "launch" the 'tornado launcher' window comes up, but the icons (buttons) for WindSh, CrossWind, etc are messed up. They still work, but the buttons say "objIconButton". I ran into this awhile back with 2.5.1 when I installed the 2.5.1 recommended patch set. Solved by backing out the patches. However, I can't do that with Solaris7. Anyone have a solution/workaround. Please cc me on any reponse as i don't loiter much on this group. Thanks in advance, - -jeff - -- Jeff King | Supervisor, Communications Systems jking@radixtek.com | Radix Technologies, Inc. 650-944-1219 | http://www.radixtek.com - -The beatings will continue until morale improves- --------------------------- Newsgroups: comp.os.vxworks Subject: Emacs in Tornado on NT Date: Tue, 22 Jun 1999 15:47:01 GMT From: "Vikram Varma" Message-ID: Does anyone use emacs as their default editor in Tornado on NT? I would like to know how to configure it so that when I click the file in the workspace, I can bring up the file in emacs. Thanks in advance, Vikram --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VXWorks start-up/boot-up timings /// Network (TCP/IP) /// pSOS Date: Mon, 21 Jun 1999 19:18:08 +0300 From: Leonid Rosenboim Organization: Bezeq International Ltd. ISRAEL Message-ID: <376E65C0.5CF3305E@bitband.com> References: "Tapia, Eduardo" wrote: > Sorry it took a bit to reply, but we have done it now. We manage to start it > in about 60 ms. The bare system. It was taking a long time in clearing the > memory. > Yes, the standard bootInit code will cler all emmory when COLD boot, assuming your DRAM may have parity to be cleared. Also on many BSPs the reboot allways gets the COLD flag. so this is also often a time consuming issue. > > Thanks a lot to everyone who helped. > > Now the problem is to load the TCP/IP stack during runtime, because if we > load it at start-up it takes way too long. This can be done if you simply comment out the call to usrNetInit() or underfine the macro which contains this call in usrConfig.c, and sometime later when you got the time do the same call. We have done this many times by building with STANDALONE and not definint STANDALONE_NET_INIT, and later from Shell we do usrBootLineInit(); useNetInit(); HTH, - Leonid --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ISR answering times Date: Mon, 21 Jun 1999 19:21:30 +0300 From: Leonid Rosenboim Organization: Bezeq International Ltd. ISRAEL Message-ID: <376E668A.AA510344@bitband.com> References: <376F5248.E6FAE73E@or-computers.de> Perhaps there is a higher priority interrupt(s) in your target which cause this delay. Also, since by default on X86 interrupt nesting is not supported this delay may be caused by a lower priority interrupt too. Many times WindView has helped tremendeously to determine the source of this problems. HTH, -Leonid Christian Richter wrote: > Hello, > > I wrote a SW that processes communication over a MIL-STD-1553 using > VxWorks on a Pentium Processor Board. After every transmission an > interrupt occurs. > My Interrupt Service Routine will normally be executed after 2 or 3 > microseconds. This time was measured using an oscilloscope. Every some > hundred program cycles the answering time of the ISR is 100 microseconds > or more. > What may be the reason? > A second interrupt which follows will be lost due to this fact. > > The long answering times occur irregularily but the program loops are > identical. So I do not think that there is an error in the software. > > Does anybody know about the problem of long ISR answering times at > VxWorks? > > C. Richter --------------------------- Newsgroups: comp.os.vxworks Subject: Where is Vxworks task scheduler? Date: Tue, 22 Jun 1999 15:52:29 GMT From: marlowe_phil@my-deja.com Organization: Deja.com - Share what you know. Learn what you don't. Message-ID: <7kobfm$cct$1@nnrp1.deja.com> I have (so I'm told) the source to the VxWorks kernel somewhere in some Clearcase VOB. I need to find the scheduler, so I can put some isntrumentation in it. (I'm trying to find out how frequently we're task switching, and how much CPU time the scheduler eats up.) Can anyone tell me what to grep on? What's the name of the scheduler's routine? Or the name of the module it lives in? Thanks in advance. Sent via Deja.com http://www.deja.com/ Share what you know. Learn what you don't. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Question about SLIP connection Date: Tue, 22 Jun 1999 16:53:35 GMT From: saustin_deja@my-deja.com Organization: Deja.com - Share what you know. Learn what you don't. Message-ID: <7kof29$dul$1@nnrp1.deja.com> References: <7kipgl$3qn1@overload.lbl.gov> <7klnes$dpd$1@nnrp1.deja.com> > But now what do I do to connect to VxWorks? We turn off PAP and CHAP authentication then do our own username/password prompting, verify those then run pppInit on the vxWorks side. To connect from Windows Dial-up Networking, I say "Pop up a terminal window before dialing" that lets me respond to the username/password prompt. When pppInit starts on the vxWorks side, it will spit out a message like "My IP is a.b.c.d, your IP is e.f.g.h, MTU is 1500 bytes" _then_ begin transmitting the LCP packets. On Windows, I exit from the popup window, then Dial-up networking starts dealing with the LCP packets and establishes the link. Sent via Deja.com http://www.deja.com/ Share what you know. Learn what you don't. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: PPC860 GDB hosted on linux/x86 Date: Tue, 22 Jun 1999 11:14:15 -0700 From: Fadi Nasser Organization: ALCATEL USA Message-ID: <376FD277.CB77F37@usa.alcatel.com> References: <3742B7BE.159ED10B@omnia.com> Try putting a Motorola ADI card in your workstation and use that as a debugger for the PPC860 platform. That's what I do. Unless of course your 860 board doesn't have the wide cable connectors to connect to the ADI card. "Shaheen H. Ali" wrote: > I have a pentium linux workstation that I would like to put to use as a > debugging platform for my motorola powerpc 860 embedded computer. > > Has anyone ported gdb for this application? > > Shaheen - -- +++++++++++++++++++++++++++++++++++++ Fadi Nasser Software Engineer Alcatel USA (707)665 8049 FAX (707)792 7807 Fadi.Nasser@usa.alcatel.com ++++++++++++++++++++++++++++++++++++++ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Where is Vxworks task scheduler? Date: Tue, 22 Jun 1999 18:41:55 GMT From: "Clifford Slocombe" Message-ID: References: <7kobfm$cct$1@nnrp1.deja.com> W.R.T. task switching, WindView is the easiest way to tell this, but costs extra. W.R.T. CPU time, use build kernel with INCLUDE_SPY, then use the Spy Chart in the Tornado browser. wrote in message news:7kobfm$cct$1@nnrp1.deja.com... > I have (so I'm told) the source to the VxWorks kernel somewhere in some > Clearcase VOB. I need to find the scheduler, so I can put some > isntrumentation in it. (I'm trying to find out how frequently we're task > switching, and how much CPU time the scheduler eats up.) > > Can anyone tell me what to grep on? What's the name of the scheduler's > routine? Or the name of the module it lives in? Thanks in advance. > > > Sent via Deja.com http://www.deja.com/ > Share what you know. Learn what you don't. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Question about SLIP connection Date: Tue, 22 Jun 1999 18:41:56 GMT From: "Clifford Slocombe" Message-ID: References: <7kipgl$3qn1@overload.lbl.gov> <7klnes$dpd$1@nnrp1.deja.com> <7kof29$dul$1@nnrp1.deja.com> wrote in message news:7kof29$dul$1@nnrp1.deja.com... > > > But now what do I do to connect to VxWorks? > We turn off PAP and CHAP authentication then do our > own username/password prompting, Can this also be done woyh a pppAddHook? > verify those then run > pppInit on the vxWorks side. To connect from Windows > Dial-up Networking, I say "Pop up a terminal window before dialing" > that lets me respond to the username/password prompt. When > pppInit starts on the vxWorks side, it will spit out a message > like "My IP is a.b.c.d, your IP is e.f.g.h, MTU is 1500 bytes" > _then_ begin transmitting the LCP packets. On Windows, I exit > from the popup window, then Dial-up networking starts dealing > with the LCP packets and establishes the link. > Thank you! > > > Sent via Deja.com http://www.deja.com/ > Share what you know. Learn what you don't. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Solaris Icon Problem Date: Tue, 22 Jun 1999 17:48:11 GMT From: BruceClements@mediaone.nospam.net Organization: Take nospam out of the address Message-ID: <376fca8c.942907869@mer-news> References: <376FAB93.60992103@radixtek.com> On Tue, 22 Jun 1999 08:28:19 -0700, Jeff King wrote: Try including the "cd" lines at the following places in the launch script. These are the last lines in the script. It whould be located in ${WIND_BASE}/host/sun4-solaris/bin. This screws up your default location in crosswind so if you run any local scripts you may need to do a cd where you previously didn't have to. if [ $# -eq 0 ]; then cd ${WIND_BASE}/host/resource/bitmaps exec launch.ex else cd ${WIND_BASE}/host/resource/bitmaps exec launch.ex "$@" fi >Hi, >More of an adminstrative question here. > >Running Tornado 1.0.1 on a recently upgraded solaris7 box. >When I do a "launch" the 'tornado launcher' window comes up, >but the icons (buttons) for WindSh, CrossWind, etc are messed up. >They still work, but the buttons say "objIconButton". > >I ran into this awhile back with 2.5.1 when I installed the 2.5.1 >recommended >patch set. Solved by backing out the patches. However, I can't do that > >with Solaris7. Anyone have a solution/workaround. > >Please cc me on any reponse as i don't loiter much on this group. > >Thanks in advance, >-jeff > >-- >Jeff King | Supervisor, Communications Systems >jking@radixtek.com | Radix Technologies, Inc. >650-944-1219 | http://www.radixtek.com >-The beatings will continue until morale improves- > > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Is raise() a no-op? Date: Tue, 22 Jun 1999 15:42:30 -0500 From: Mike Beede Organization: Ciprico Inc. Message-ID: <376FF536.485E5640@ciprico.com> References: <21039-376EF687-83@newsd-222.iap.bryant.webtv.net> Greg Silverman wrote: > > what are t101 and t2? how i find out whati i have? Tornado 1.0.1 is what was distributed when we got Tornado last December. We STILL HAVEN'T GOT TORNADO 2, so you probably don't have it either. Check the CD you received--it should be marked. I think there was a Tornado 1.0, but I don't know how long ago that was distributed. As an aside, I'd sure like to know how Wind River decides what order to ship the T2 update in. I believe it's been over a week since we heard from T2 Shipping Command that it was going out that day. It **does** make for an amusing gag when my partner goes to his mail slot every day "for that package from Wind River...." Not extremely amusing, but then he isn't an extremely amusing guy. Mike Beede --------------------------- Newsgroups: comp.os.vxworks Subject: Tornado 2 and Dependencies Date: Tue, 22 Jun 1999 17:04:34 -0400 From: Matthew Donadio Organization: Image & Signal Processing, Inc. Message-ID: <376FFA62.F9A570C9@isptechinc.com> Hello all, I am having a problem porting a library to VxWorks / Tornado 2. The library uses the convention that header files are included using <> and then assumes that -I. is a command line option to the compiler. This is giving the Tornado 2 project tool heartburn, though. When I generate dependencies for the library, the project tool thinks that all of the #include files are external to the library and creates a makefile dependency like blah.o: $(PRJ_DIR)/blah.c blah.h instead of blah.o: $(PRJ_DIR)/blah.c $(PRJ_DIR)/blah.h Building the library crashes because the make gets called from a subdirectory, and it can not find the header files, and assumes that they need to be created. I could change all of the #include lines to use quotes instead of <>, but there are 150+ source files, and I don't want to do it every time the library gets released. I would also like to keep the dependencies because full recompiles would take way too long. Any ideas on how I can get around this problem? Thanks. - -- Matt Donadio (donadio@isptechinc.com) | 43 Leopard Rd, Suite 102 Sr. Software Engineer | Paoli, PA 19301-1552 Image & Signal Processing, Inc. | Phone: +1 610 407 4391 http://www.isptechinc.com | FAX: +1 610 407 4405 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Problems in sending data over socket Date: Tue, 22 Jun 1999 15:42:08 -0700 From: Stephen Macmanus Organization: Wind River Systems Message-ID: <37701140.5438@wrs.com> References: <7jt4u7$8q1@overload.lbl.gov> Venkatesh Prakash PS wrote: > > Hi All, > I am using VxWorks version 5.3.1, I have the following problems. > > 1. I have opened a UDP Socket and the "sendto" function is failing to send after > 2-3 minutes. I don't have idea how to get error number when some socket function > fails, since "perror()" function says undefined symble. > > I am using "fei" interface, with disabled "INCLUDE_END" and defined > "END_OVERRIDE". > > 2. I am getting following error messages on the screen, > > "arp info overwritten for xxxxxxxx by yyyyyyyyyy" > "arp_rtrequest: malloc failed" > "arptnew failed on xxxxx" > "arpresolve: can't allocate llinfo" > > How to correct these problems. > > Thanks in advance, > > Regards, > > vprakash The first error message ("arp info overwritten...") indicates that an ARP response is reporting a hardware address which differs from the existing entry. This message is just a warning. The situation might not be a problem if the sending host actually changed Ethernet cards recently while retaining the same IP address. It can't be prevented by any local changes. The remaining three messages indicate that the system is unable to allocate space for a new entry. The solution is to increase the size of the network pools above the default values, which are suitable for the simple Tornado host-target communication but are often too small for a production environment. - -- - ------------------ Stephen Macmanus #include Software Engineer Wind River Systems --------------------------- Newsgroups: comp.os.vxworks,comp.realtime Subject: Contract positions in Massachusetts Date: Tue, 22 Jun 1999 22:48:02 +0000 From: George Walk Organization: Road Runner Message-ID: <377012A2.37EE7784@mediaone.net> Project: SQA tester/developer for VxWorks based systems Location: North of Boston Duration: 6+ months (full-time, on site) Contact: George Walk, Responsive Software, Inc. email: walk@mediaone.net Experience: 3-7 years experience Skills: I am looking for 2 people to work on a contract for 6+ months. They should have experience with Unix development environments, shell scripts, and C programming. Also needed is experience/familiarity with VxWorks. The experience doesn't need to be with drivers or developing BSPs. You should be familiar with configuring hardware and applications in a VxWorks environment using Tornado tools. While this is an SQA position, the people I am looking for should be able to program in C to develop new tests and debug, analyze, and fix problems in existing tests. Please email resumes or questions to George Walk (walk@mediaone.net) Responsive Software, Inc. is a consulting company specializing in real-time and embedded systems development. We offer services in application development, driver development, and porting. Real-time operating systems experience includes: VxWorks, VRTX, and pSOS. Other systems include Solaris, Irix, and Linux. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: SENS Multiple Interface Drivers for TCP/IP Protocol Date: Tue, 22 Jun 1999 15:50:22 -0700 From: Stephen Macmanus Organization: Wind River Systems Message-ID: <3770132E.41B6@wrs.com> References: <3762A38E.65350DB2@eci.esys.com> rdfa wrote: > > Using SENS tornado 1.0 > > I need to define multiple network interfaces Ether net and a Proprietary > Interface Driver I can get the complete muxdevLoad on both drivers but > it seem to attach only the first device to the TCP/IP stack. > > I want to control routing by specifying the routes such as: > > ifAddrSet("lnpci0", "193.1.3.x") > ifAddrSet("Orb" ,"193.1.4..x") > > and have the SENS TCP/IP Protocol Stack route to the correct interface. > I can not get the ifShow() to show the interface but can see it if I do > a muxShow() both drivers are loaded. > > Any help would be appreciated > > Robert Flanders > rdfa@eci.esys.com > Raytheon Systems Company The muxDevLoad() routine installs the devices (which is why they are visible with muxShow) but does not attach them to a protocol. The ifShow() routine, which predates the MUX/END model, only displays devices registered with the IP protocol. There are several routines needed to complete that process, including muxDevStart() and muxBind(). In the case of the IP protocol, the bind calls are completed within the ipAttach() routine. See the manual pages for details. The usrNetwork.c startup code can also be used as an example. - -- - ------------------ Stephen Macmanus #include Software Engineer Wind River Systems --------------------------- Newsgroups: comp.os.vxworks Subject: Re: etherInputHookAdd Date: Tue, 22 Jun 1999 16:05:40 -0700 From: Stephen Macmanus Organization: Wind River Systems Message-ID: <377016C4.3706@wrs.com> References: <7k2fka$7fs$1@news.kren.nm.kr> hhs wrote: > > Hi. > > I have a question about etherInputHookAdd at PPP connection (or ethernet). > > I know that I can see the raw input packet using etherInputHookAdd. > > I want to change the raw input packet data and then return it to the IP > stack by etherInputHookAdd. > > Is it possible? > > I tried to do that, but it was impossible to grap the raw input packet and > change the packet and return it to the IP stack under my power. > > If it is possible , let me know the way. > > If it is not, is there any alternate way to do that? > > Thank you . It is theoretically possible to change packet contests using Ethernet hooks. The general idea is to make the necessary changes to the frame payload, then return FALSE so that the driver continues processing the frame instead of discarding it. While feasible, this approach is not recommended. Not only is examining every frame expensive, but altering them could cause any number of unpredictable side effects. For instance, shortening the frame could be needed, but altering the reported length could cause memory leaks when the M_BLK chain was finally freed. In A somewhat better alternative is to use the IP filter hook. It examines all IP packets only, avoiding the extra overhead of testing other frames. This hook returns TRUE instead of FALSE to continue the built-in processing. It has some of the same limitations, so changing the packet size is still not recommended. See the ipFilterLib entries in the manual for details. A final possibility is to open a raw socket which receives copies of all relevant IP packets matching the source and destination values. - -- - ------------------ Stephen Macmanus #include Software Engineer Wind River Systems --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Configuring vxWorks for bootp Date: Tue, 22 Jun 1999 16:14:33 -0700 From: Stephen Macmanus Organization: Wind River Systems Message-ID: <377018D9.3DBB@wrs.com> References: <37659104.8832417C@swbell.net> <376598B1.D2F0FAC9@swbell.net> Matt Henderson wrote: > > In what I can find online, I am not sure if the SunOS /etc/bootparams > actually is used to support bootp (it only talks about supporting > diskless clients and has support for specifying bootfiles, dumpfiles, > but does not indicate how the hardware address is specified...) If this > is the case, then is there a bootp daemon available for SunOS in the > public domain? Tornado installations include a bootp server in the $WIND_BASE/target/unsupported/bootp2.1 directory. - -- - ------------------ Stephen Macmanus #include Software Engineer Wind River Systems --------------------------- Newsgroups: comp.os.vxworks Subject: Re: UDP datagram size and the MTU Date: Tue, 22 Jun 1999 16:19:46 -0700 From: Stephen Macmanus Organization: Wind River Systems Message-ID: <37701A12.17D2@wrs.com> References: Dave Craig wrote: > > Hi, > > Does anyone have any experience of using the IP_DO_LARGE_BCAST flag (in > ipCfgParams)? It seems to allow UDP broadcasts of datagrams which are bigger > than the MTU to be fragmented and sent. Has anyone used this, and if so does > it work completely transparently? (It doesn't seem to be documented) > > Thanks > > Dave Craig > Palo Alto Yes, the flag does allow fragmentation of broadcast datagrams. It works transparently in the sense that no additional changes are required, but could have a (possibly significant) negative impact on throughput if fragments are actually lost. It will also increase the memory requirements of the receiver since the fragments must be queued until reassembly is complete. - -- - ------------------ Stephen Macmanus #include Software Engineer Wind River Systems --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Symbol table won't download Date: Sat, 19 Jun 1999 20:27:31 -0700 From: "Cameron Fowler" Organization: AT&T WorldNet Services Message-ID: <7kphk3$ap4$1@bgtnsc01.worldnet.att.net> References: <3755B72B.4C440338@lmco.com> <375BE6A3.1407EDBD@agie.ch> Also #define INCLUDE_LOADER Roberto Nunnari wrote in message <375BE6A3.1407EDBD@agie.ch>... >#define INCLUDE_NET_SYM_TBL /* load symbol table from network */ > >Best regards. >-- > Roberto Nunnari -software engineer- > mailto:roberto.nunnari@agie.ch > AGIE - http://www.agie.com > Via dei pioppi 16 tel: +41-91-8069138 > 6616 Losone """ > Switzerland (o o) >=======================oOO==(_)==OOo======================== > MY OPNIONS ARE NOT NECESSARELY THOSE OF MY EMPLOYER --------------------------- Newsgroups: comp.os.vxworks Subject: Re: SCSI Problem for MCP750 BSP Date: Sat, 19 Jun 1999 20:34:00 -0700 From: "Cameron Fowler" Organization: AT&T WorldNet Services Message-ID: <7kphk4$ap4$2@bgtnsc01.worldnet.att.net> References: <375FE3A6.78DC0F00@siemens.at> Is there anything else on the scsi bus such as a CD rom if so, make sure that you have wide transfers undefined in config.h Andreas Lugbauer wrote in message <375FE3A6.78DC0F00@siemens.at>... >Hi, > >we have a problem when using our SCSI disk in our MOTOROLA CompactPCI >system. > >Our configuration is as follows: > >MCP750-1332 CompactPCI board >MPMC101 PMC SCSI-2 Adapter >Seagate ST39173W 9GB SCSI disk > >The problem we have is that the scsiPhysDevCreate call hangs, when >creating the device for the disk. Also we have tried to use the >scsiAutoConfig call but this hangs too. > >We have verified when we load the PPCBug firmware of MOTOROLA the disk >is detected in the system by using the IOI command. So we think our HW >configuration is ok. > >It seem also that the initialisation of the SCSI controller works >because it is detected during booting vxWorks. > >What can be the cause of our problem? > >Thanks in advance for any hints > >Andreas > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Tornado 2 and Dependencies Date: Wed, 23 Jun 1999 03:00:59 GMT From: David Hamel Organization: @Home Network Message-ID: <37704D93.28A4021F@ntanet.net> References: <376FFA62.F9A570C9@isptechinc.com> Matthew Donadio wrote: > Hello all, > > I am having a problem porting a library to VxWorks / Tornado 2. > > The library uses the convention that header files are included using <> > and then assumes that -I. is a command line option to the compiler. > This is giving the Tornado 2 project tool heartburn, though. > > When I generate dependencies for the library, the project tool thinks > that all of the #include files are external to the library and > creates a makefile dependency like > > blah.o: $(PRJ_DIR)/blah.c blah.h > > instead of > > blah.o: $(PRJ_DIR)/blah.c $(PRJ_DIR)/blah.h > > Building the library crashes because the make gets called from a > subdirectory, and it can not find the header files, and assumes that > they need to be created. > > I could change all of the #include lines to use quotes instead of <>, > but there are 150+ source files, and I don't want to do it every time > the library gets released. I would also like to keep the dependencies > because full recompiles would take way too long. > > Any ideas on how I can get around this problem? Thanks. > I had the exact same problem and WRS Support sent me this summary. TITLE: How to generate external dependencies in Tornado 2.0? AUTHOR: Eric Lo SCOPE DISCOVERED ON HOST: All Windows NT 4.0 ------------------- ------------------- ARCHITECTURE: All MIPS ------------------- ------------------- BSP: All idts381 ------------------- ------------------- PRODUCT / REVISION: Tornado 2.0 Tornado 2.0 ------------------- ------------------- DESCRIPTION: How can I generate external dependencies in Tornado 2.0? I have two directories, source_code (where all the source codes are) and headers (where all the header files are). I add all the source codes into the downloadable project, then select "Dependencies" from the workspace context menu. All the header files are shown in the "external Dependencies" folder in the "Files" view of the workspace window. Then I click the "Build Project" from the tool bar, I got the following error message: make: *** No rule to make target `bee.h', needed by `bee.o'. Stop. Why can't the files build? RESEARCH: Tornado User's Guide 2.0., Trial and Error. RESOLUTION: The following procedures will generate a correct dependencies in order to build the object modules. Step 1: Modify the build specification To display the build specification, double-click on the build name in the "Build" view of the workspace to display the property sheet for the build. Then, select the "C/C++ compiler" tab of the build specification property sheet to view the compiler options. Add another option in the text box: -Idirectory_path_of_the_header_files For example: -IE:\header Step 2: Rebuild the dependencies Select "Dependencies" from the workspace context menu and click the OK button to rebuild the dependencies. Step 3: Rebuild the object modules/application You can now click "ReBuild All" button in the tool bar to rebuild the object modules/application. The new compiler option "-I" will be pulled into the dependencies and the compiler will search the specified directories for the header files. In conclusion, new dependency generation is required every time when the C/C++ compiler options are changed in the build specification. KEYWORDS: dependencies, project, makefile, compiler option --------------------------- Newsgroups: comp.os.vxworks Subject: static linking vs. dynamic link Date: Wed, 23 Jun 1999 11:03:19 +0800 From: "liuf" Organization: Alcatel/Bell Message-ID: <3770501a.0@dnews.sbell.com.cn> hi, When I try to link my modules, I wonder what's the main difference between dynamic linking and static linking? And how to make my modules be dynamic libs in vxWorks? Thanks! --------------------------- Newsgroups: comp.os.vxworks Subject: a problem while linking Date: Wed, 23 Jun 1999 11:31:28 +0800 From: "liuf" Organization: Alcatel/Bell Message-ID: <377056b3.0@dnews.sbell.com.cn> hi, I just begin to use vxworks. I use NT as host and use a PC as target. When I link my test code, as >cc386 -o testExe -nostdinc -fno-builtin -v -LC:\x86\target\lib testCode.o -lI80486gnuvx it shows an error: cannot open crt0.o: No such file or directory Can anyone tell me what ctr0.o is? why cc386 automatically link it with my code? where can I find this? Thanks! --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Emacs in Tornado on NT Date: Wed, 23 Jun 1999 04:40:31 GMT From: bobarmstrong@home.nothiis.net (Robert Armstrong) Organization: @Home Network Message-ID: <3zZb3.14893$NY6.5986@news.rdc1.sfba.home.com> References: In T2 this is trivial and in T101 impossible. If you are on T2 then look to your options, the editor configuration is prominent. I've used emacs in T2 but use PFE now (whatever you are used to). Bob In article , "Vikram Varma" wrote: >Does anyone use emacs as their default editor in Tornado on NT? >I would like to know how to configure it so that when I click the file in >the workspace, I can bring up the file in emacs. >Thanks in advance, >Vikram > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Is raise() a no-op? Date: Wed, 23 Jun 1999 04:43:22 GMT From: bobarmstrong@home.nothiis.net (Robert Armstrong) Organization: @Home Network Message-ID: References: <21039-376EF687-83@newsd-222.iap.bryant.webtv.net> <376FF536.485E5640@ciprico.com> Wind often ship via UPS, in this case you can get the UPS tracking number from internal sales in Alameda where they actually do the shipping. Alternatively you are obviously on support so just call and ask them what's happening. Bob >It **does** make for an amusing gag when my partner goes to his >mail slot every day "for that package from Wind River...." Not >extremely amusing, but then he isn't an extremely amusing guy. > > Mike Beede --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Where is Vxworks task scheduler? Date: Wed, 23 Jun 1999 04:51:10 GMT From: bobarmstrong@home.nothiis.net (Robert Armstrong) Organization: @Home Network Message-ID: <2JZb3.14898$NY6.5986@news.rdc1.sfba.home.com> References: <7kobfm$cct$1@nnrp1.deja.com> SO you have the source code? Well look in WIND_BASE/target/src/arch/ and dig around in the assembler source code, this stuff is all too time critical to work out in C. Enjoy yourself :) Bob In article <7kobfm$cct$1@nnrp1.deja.com>, marlowe_phil@my-deja.com wrote: >I have (so I'm told) the source to the VxWorks kernel somewhere in some >Clearcase VOB. I need to find the scheduler, so I can put some >isntrumentation in it. (I'm trying to find out how frequently we're task >switching, and how much CPU time the scheduler eats up.) > >Can anyone tell me what to grep on? What's the name of the scheduler's >routine? Or the name of the module it lives in? Thanks in advance. > > >Sent via Deja.com http://www.deja.com/ >Share what you know. Learn what you don't. --------------------------- Newsgroups: comp.os.vxworks Subject: MVME1604 64MB memory board available Date: Tue, 22 Jun 1999 23:49:04 -0700 From: Innovative Research Technologies Organization: Innovative Research Technologies Message-ID: <37708360.8C8B8C2A@innovative-research.com> Reply-To: sales@innovative-research.com I have quantity (1) 64MB memory add-on board available for the Motorola MVME1604 PPC VME board. I also have a few MVME1604 boards configured with 16MB and 32MB of memory. E-mail if interested. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Games on vxworks Date: Wed, 23 Jun 1999 06:31:49 GMT From: sminar@ntouch.com Organization: Deja.com - Share what you know. Learn what you don't. Message-ID: <7kpv0l$uke$1@nnrp1.deja.com> References: <7k4f8g$ag0$1@nnrp1.deja.com> <376d32e5.8677720@news.texas.net> I appreciate the input. A few more details may be helpful to the discussion... The choice of OS was not mine to make and would not be my first choice if I was in a hurry to develop some games. The OS was chosen due to timing, technical, pricing and political issues. If we were to start again today on the project we might choose something else but for now we are forging ahead. Gaming is just one small part of the product they are developing and in some ways is an afterthought. They understand that developing games on the platform will not necessarily be easy and are willing to accept games that can be developed within the limitations of the platform and timeframe. Even JAVA games might be acceptable (to them, not to me). While I have not yet studied VxWorks in great detail, I have not found anything that would make it "impossible" to develop games. I would very much like to have a code base to start with, gaming specific if possible, but at least a set of tested graphics routines. If need be we could write directly to the hardware since we are controlling the hardware platform. In terms of possible solutions, I see several directions that one could pursue (If your a true gamer, please don't hate me for bringing up a few ideas that will make you gag...): 1) Browser based games. We have a specialized version of the Netscape browser running on our platform. Unfortunately I do not have the source to the browser so I'm stuck waiting for someone else to fix it. And there are many bugs. It is operating at a level above Netscape 2.0 and below 3.0. JAVA is coming. Javascript is there but very buggy. But it still might be possible to come up with a few "games" (using the term lightly) that would run under the browser. 2) Develop or port games to VxWorks using one of the available graphics packages. I had heard about Zinc but I know little about its capabilities. I'm interested in MetaWindows as I have done projects using that package. And I did not know that an X/OpenGL package was available. In the short term I would love to find one or more simple games that run on these platforms that could be ported to VxWorks. 3) Another possibility would be to write a game platform emulator and license ROMS. This has become popular for many other OS's. There are ROMS for hundreds of arcade games available... the difficulty would be in getting reasonble licensing terms. This may be the quickest way to get something up and running that would be a more flexible solution. 4) In my dreams I would like to find someone who is developing a gaming library for VxWorks that implements all the core gaming functionality you would ever want. To go with that would be a few freeware example games that I could use immediately for demonstration purposes while developing / porting other games. (I can dream, can't I?) 5) What I really want is to find someone who is already working on this problem so that I can hire / license what they are creating. I'm buried under related work so it makes sense for me/them to have someone else do it. In article , bobarmstrong@home.nothiis.net (Robert Armstrong) wrote: > Sorry but I think you significantly miss the point. I play Quake II > deathmatches a lot (used to be 2 hours minimum each day in a ten way game). A > game like quake is a little bit of networking, a little bit of configuration > and a huge amount of rendering. I always used OpenGL. > > So why is vxWorks fundamentaly incapable or unsuitable? The networking is > mostly fine now (do read the chapter on network customization, I think it may > be chapter 4.5 of the vxWorks 5.4 programmers guide, the default > configuration is intended for debug only and reconfiguration is simple). > Recovering the configuration from somewhere is simple. There are plenty of > fast platforms available (including PCs). There is also an OpenGL product in > the partner directory:- > > http://www.wrs.com/html/trade/wrs00154.html > > So I think that addresses the "could you", but it still leaves the "why would > you want to" and that I have no answer for. > > Mostly vxWorks has been too expensive for this sort of development, perhaps > this new pricing model I've seen mentioned in tech stocks will change that. I > drove my graphics card direct last time I used one, then again that was with > o/s9 and I'd rather have had a bare machine than that (OK the compiler sort > of helped). > > To be fair, one of my friends programmed some games for Phillips pagers, they > went down very well. > > Bob > > In article , "Clifford Slocombe" > wrote: > >What I meant was that there was no standard graphical interface (apart from > >X/OpenGL), with a wide choice of graphic proprietary graphic API which horse > >would a games programmer back (if any)? > > > >It occurred to me that games does not necessarily mean graphics, but text > >adventures are a little too obscure these days to be an attractive product > >feature! > > > >W.R.T QNX/Photon, lets not get into that one here, I probably get flamed! > > > >Given time, games may appear for the Zinc interface. This at least has the > >benefit of being (now) a WindRiver product, so is likely to become widely > >supported. > > > >Clifford. > > > >Bruce wrote in message > >news:376d32e5.8677720@news.texas.net... > >> In comp.os.vxworks > >> "Clifford Slocombe" wrote: > >> > >> >I think that you have chosen the wrong OS. If you want games, and a > >serious > >> >RTOS use QNX, it has a port of Doom! I am interested to know why you > >would > >> >want do this. > >> > >> There is a graphical slot machine written in VxWorks. My VxWorks > >> application is graphical and uses Ease Software's port of MetaWindows. It > >> can be done, though I agree it may not be the best for blazing fast > >> graphics. Neither is QNX. > >> > >> Bruce > >> > > > > > Sent via Deja.com http://www.deja.com/ Share what you know. Learn what you don't. --------------------------- End of New-News digest ********************** From owner-vxwexplo-process Wed Jun 23 04:10:20 1999 From: "Tapia, Eduardo" Date: Wed Jun 23 04:10:24 PDT 1999 Subject: RE: Starting the network at run-time > > Assuming that you have built network support into the VxWorks > image, you need to call the code > > usrBootLineInit(sysStartTime); > usrNetInit(BOOT_LINE_ADRS); > > This should then start up the network for you and attach > whatever device you have specified as the boot device in > the boot line, as the WDB/Tornado Link. > > Pretty impressive timings by the way - 60ms is not long ! How > did pSOS compare ? pSOS started in 110ms, but clearing RAM. > Did you go for an uncompressed, run directly from FLASH > VxWorks image ? Yes, without the Clearing Ram bit. Regards, Eduardo Tapia. From owner-vxwexplo-process Wed Jun 23 04:10:27 1999 From: "Tapia, Eduardo" Date: Wed Jun 23 04:10:31 PDT 1999 Subject: RE: VXWorks start-up boot-up timings > > Dear Eduardo, > > I recorded the conversations to a file for posterity. > > Could you say what exactly solved your problem? > > I wish to keep this information on file, if a similar situation occurs > here. > We went for an uncompressed, run directly from FLASH VxWorks image. Then changed to code so it would not run the 'Clearing Ram' bit. We think it works OK, but we have not had time to test it properly yet. Now we want to add network capabilities at run time. Regards, Eduardo Tapia. From owner-vxwexplo-process Wed Jun 23 07:57:11 1999 From: bwaite@cspi.com Date: Wed Jun 23 07:57:15 PDT 1999 Subject: Re: using same vxWorks on UNIX and NT platforms Rehan, I am actually doing development on both Unix and NT myself and have worked though only a few minor problems before having a reletively stable system. I too use Exceed to run my UNIX based Luancher and debugger but have had no problems. What kind of hangs are you seeing with Crosswind on UNIX? There are a few Unix based Crosswind patches that make it much more stable. The first one that comes to mind is a patch that fixes a hang in Crosswind that occurs whenever you enter text at the gdb prompt. Also, maybe Crosswind itself is not hanging but the WTX agent is not responding and causing your launcher to hang waiting for a response. I have seen this behavior many times. I am not exactly sure why your output is not working, but it almost looks like you are not redirecting I/O on the Unix side, although I cannot explain why you see header when you type i(). Make sure on the Unix side of the world you have a windsh file loacted in your .wind directory. I have also been successful building on NT and debugging on UNIX and vice versa. The only difference is that I am using a PPC based Motorola board. If you are still having problems mail me and I can probably help you get through the growing pains. Good Luck, Brian Waite CSPI Customer Support Email: bwaite@cspi.com vxwexplo@lbl.gov on 06/22/99 02:42:06 PM To: vxworks_users@csg.lbl.gov cc: (bcc: Brian Waite/CSP) Subject: using same vxWorks on UNIX and NT platforms Submitted-by owner-vxwexplo-process Tue Jun 22 11:42:00 1999 Submitted-by: Reyhan ERGUN This is a multi-part message in MIME format. --------------308A945FCDE6C0B45AD44642 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit I use MBX860 Board, MBX860 BSP, vxworks,.... I use Tornado I on both UNIX and NT platforms. I didn't encounter so many problems while working on NT, however working on UNIX is very annoying. Using CrossWind on UNIX crashes Tornado very often. I am not sure the problem is because of EXCEED which I use to connect UNIX machine from my PC or the tool itself. Anyway, I would like to ask if I could make vxWorks on UNIX, and use Tornado tools from my PC, do all debugging work from PC... I tried this but it doesn' work!....I encountered problems from the shell as follows: -> i NAME ENTRY TID PRI STATUS PC SP ERRNO DELAY ---------- ------------ -------- --- ---------- -------- -------- ------- ----- value = 0 = 0x0 -> devs drv name value = 0 = 0x0 -> iosFdShow fd name drv WTX Error 0x100df (AGENT_GOPHER_SYNTAX) value = -1 = 0xffffffff -> printErrno 0x0 = OK value = 0 = 0x0 ......etc. I should make vxWorks on UNIX, because I do some work integrated with another tool working on UNIX. I get vxWorks file via ftp from UNIX to PC, I tried on some other object file which are get from UNIX to PC by the same way, I can load this files from PC to the target and these applications can run on the target without any problem. Is there any utility to obtain a vxWorks made on UNIX platorm which can be run while host tool is running on NT platform? Any comments would be appreciated, Regards, Reyhan --------------308A945FCDE6C0B45AD44642 Content-Type: text/x-vcard; charset=us-ascii; name="vcard.vcf" Content-Transfer-Encoding: 7bit Content-Description: Card for Reyhan (FINDIK) ERGUN Content-Disposition: attachment; filename="vcard.vcf" begin: vcard fn: Reyhan (FINDIK) ERGUN n: ;Reyhan (FINDIK) ERGUN org: ASELSAN INC. adr: Microwave and System Technologies Division ;;Software Engineering Dept.;ANKARA;;;TURKEY email;internet: findik@mst.aselsan.com.tr tel;work: (90-312) 385 19 00 (Ext. 2668) tel;fax: (90-312) 354 52 05 note: http://www.aselsan.com.tr x-mozilla-cpt: ;0 x-mozilla-html: FALSE version: 2.1 end: vcard --------------308A945FCDE6C0B45AD44642-- From owner-vxwexplo-process Thu Jun 24 04:00:12 1999 From: daemon@csg.lbl.gov Date: Thu Jun 24 04:00:16 PDT 1999 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Thu Jun 24 04:00:07 PDT 1999 Subject: snmpIoCommunityValidate causes manager to hangup Subject: Re: snmpIoCommunityValidate causes manager to hangup Subject: Re: snmpIoCommunityValidate causes manager to hangup Subject: FAQ ? Subject: Re: UDP performance over PPP Subject: Re: Starting the network at run-time Subject: Disney vacation 3057 Subject: VxWorks and Zip drives Subject: etherInputHookAdd and datagram handling Subject: Re: etherInputHookAdd Subject: Re: FAQ ? Subject: Tornado on NT vs. Tornado on Solaris? ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: snmpIoCommunityValidate causes manager to hangup Date: Wed, 23 Jun 1999 12:14:50 +0100 From: "Will Fookes" Organization: NDS UK Message-ID: <7kqei3$i36$1@ndsukns01.ndsuk.com> I'm developing an SNMP Agent using WindNet SNMPv1/v2c and I find that if a manager attempts a Get or Set (or whatever) using an invalid community name (which is detected by snmpIoCommunityValidate) I get no response from the Agent. This causes the manager to timeout and repeat the request... I've put debug code in the snmpIoWrite function, so I am certain that no return packet is generated. The Agent is not dead, however, and processes subsequent packets properly. My suspicion is that within snmpIoCommunityValidate I need a call (on detection of the bad community) to a function like getproc_nosuchins or setproc_error - but this fuction is called for PDUs of ALL types... Has anybody else resolved this one? Will --------------------------- Newsgroups: comp.os.vxworks Subject: Re: snmpIoCommunityValidate causes manager to hangup Date: Wed, 23 Jun 1999 12:46:05 GMT From: saustin_deja@my-deja.com Organization: Deja.com - Share what you know. Learn what you don't. Message-ID: <7kqkua$5rl$1@nnrp1.deja.com> References: <7kqei3$i36$1@ndsukns01.ndsuk.com> You can certainly do whatever you like in snmpIoCommunityValidate - for instance, in our application we generate an "authentication failed" trap, and also write to our diagnostic log, etc. How the manager reacts is probably not an issue for your agent - as far as I know the agent has no responsibility to respond to a request with an invalid community string (correct me if I'm wrong?). If your manager times out and sends the request again, I'd guess that's something in the way the manager is configured, and really doesn't have to do with the agent. Some other managers may behave differently, so you shouldn't necessarily code your agent to dovetail with only one particular manager. In article <7kqei3$i36$1@ndsukns01.ndsuk.com>, "Will Fookes" wrote: > I'm developing an SNMP Agent using WindNet SNMPv1/v2c and I find that if a > manager attempts a Get or Set (or whatever) using an invalid community name > (which is detected by snmpIoCommunityValidate) I get no response from the > Agent. This causes the manager to timeout and repeat the request... > > I've put debug code in the snmpIoWrite function, so I am certain that no > return packet is generated. The Agent is not dead, however, and processes > subsequent packets properly. > > My suspicion is that within snmpIoCommunityValidate I need a call (on > detection of the bad community) to a function like getproc_nosuchins or > setproc_error - but this fuction is called for PDUs of ALL types... > > Has anybody else resolved this one? > > Will > > Sent via Deja.com http://www.deja.com/ Share what you know. Learn what you don't. --------------------------- Newsgroups: comp.os.vxworks,comp.protocols.snmp Subject: Re: snmpIoCommunityValidate causes manager to hangup Date: Wed, 23 Jun 1999 15:52:09 +0100 From: "Will Fookes" Organization: NDS UK Message-ID: <7kqr9s$olt$1@ndsukns01.ndsuk.com> References: <7kqei3$i36$1@ndsukns01.ndsuk.com> <7kqkua$5rl$1@nnrp1.deja.com> I agree with you - up to a point. I too have added the "authentication failure" trap, but I was under the impression that the Agent should return "noSuchName" (SNMPv1) or "authorizationError" (SNMPv2c) if the object was not available in the current view. Sitting there like a lemon seems to me to be a tad unhelpful! I notice also that snmpInBadCommunityNames gets incremented when a community name is not matched, and that snmpInBadCommunityUses is never(?) incremented. My reading of the MIB suggests that snmpInBadCommunityNames should be incremented when an unrecognised community name is received, whereas snmpInBadCommunityUses would be incremented when you try to access objects that exist but are not in the view of the quoted community. BTW, I found a function ENVOY_Send_SNMP_Error_Packet() in the library, but the declaration doesn't tell me what its 2nd & 3rd parameters are! - any ideas if/how I can use this? I'll cross-post this to comp.protocols.snmp for further opinions. Will saustin_deja@my-deja.com wrote in message <7kqkua$5rl$1@nnrp1.deja.com> in newsgroup comp.os.vxworks... >You can certainly do whatever you like in snmpIoCommunityValidate - for >instance, in our application we generate an "authentication failed" >trap, and also write to our diagnostic log, etc. How the manager >reacts is probably not an issue for your agent - as far as I know the >agent has no responsibility to respond to a request with an invalid >community string (correct me if I'm wrong?). > >If your manager times out and sends the request again, I'd guess that's >something in the way the manager is configured, and really doesn't have >to do with the agent. Some other managers may behave differently, so >you shouldn't necessarily code your agent to dovetail with only one >particular manager. > >In article <7kqei3$i36$1@ndsukns01.ndsuk.com>, > "Will Fookes" wrote: >> I'm developing an SNMP Agent using WindNet SNMPv1/v2c and I find that >> if a manager attempts a Get or Set (or whatever) using an invalid >> community name (which is detected by snmpIoCommunityValidate) I get >> no response from the Agent. This causes the manager to timeout and >> repeat the request... >> >> I've put debug code in the snmpIoWrite function, so I am certain that >> no return packet is generated. The Agent is not dead, however, and >> processes subsequent packets properly. >> >> My suspicion is that within snmpIoCommunityValidate I need a call (on >> detection of the bad community) to a function like getproc_nosuchins >> or setproc_error - but this fuction is called for PDUs of ALL types... >> >> Has anybody else resolved this one? >> >> Will > >Sent via Deja.com http://www.deja.com/ >Share what you know. Learn what you don't. --------------------------- Newsgroups: comp.os.vxworks Subject: FAQ ? Date: Wed, 23 Jun 1999 20:19:09 +0200 From: "Geir R. Westermann" Organization: Telenor Online Public Access Message-ID: I'm quite new to this group, and I've been looking for a FAQ. Is there one ? Geir --------------------------- Newsgroups: comp.protocols.ppp,comp.os.vxworks Subject: Re: UDP performance over PPP Date: Wed, 23 Jun 1999 15:12:13 -0700 From: Fadi Nasser Organization: ALCATEL USA Message-ID: <37715BBD.48B89842@usa.alcatel.com> References: <7i1brc$47a$1@nnrp1.deja.com> Hi there I was running into a somewhat similar problem until I got pulled off it. Where you able to find a way to sync. buffer utilization with traffic burstiness? Could you monitor those buffers without probably looking into their location in memory or writing your own monitoring routines??? Thanks in advance, Fadi. saustin_deja@my-dejanews.com wrote: > I'm interested in finding out more detail about how the higher layers > of a protocol stack (specifically SNMP/UDP) are affected by running > over IP/PPP instead of IP/Ethernet. My application seems to have > serious problems when all of a sudden the bandwidth dips from > 10 Mbps to 9600 bps. > > I would like to know how buffers are managed > within the protocol stack and if there's any API to the buffer > management that would allow me to track utilization. Then I could > "throttle back" the generation of UDP traffic to match the buffer > availability. > > I'd also like to know how PPP arbitrates between traffic demands from > each peer. That is, if one peer is sending a ton of UDP traffic, it > seems like the other peer can't get a packet in (or perhaps it's the > case that the responses are queued way back in the first peer?). Are > there any priority schemes within the protocols that would allow me to > always have fast response, regardless of how much "bulk" UDP traffic is > queued up in one peer? > > My envrionment is VxWorks 5.3.1 on i386. Their documentation states > "PPP attaches to the TCP/IP stack at the driver (link) level." Does > this mean that all the buffer management is handled within the serial > port driver, which likely uses a fairly small character FIFO? > > Thanks for any help. > > --== Sent via Deja.com http://www.deja.com/ ==-- > ---Share what you know. Learn what you don't.--- - -- +++++++++++++++++++++++++++++++++++++ Fadi Nasser Software Engineer Alcatel USA (707)665 8049 FAX (707)792 7807 Fadi.Nasser@usa.alcatel.com ++++++++++++++++++++++++++++++++++++++ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Starting the network at run-time Date: Wed, 23 Jun 1999 22:32:55 GMT From: rjdejoie@kmrmail.kmr.ll.mit.edu (Bob DeJoie) Organization: Massachvsetts Institvte of Technology Message-ID: <37715e79.517327818@news> References: <7kqfjo$7i69@overload.lbl.gov> On Wed, 23 Jun 1999 11:54:36 +0100, "Tapia, Eduardo" wrote: > >> >> Assuming that you have built network support into the VxWorks >> image, you need to call the code >> >> usrBootLineInit(sysStartTime); >> usrNetInit(BOOT_LINE_ADRS); >> Possibly an alternate approach: I am running rom resident and want to run one code base on multiple instances of the target. I use some configuration bits to index into a table of IP addresses and then pass strings with unique addresses usrNetInit based on the board configuration. I don't bother to call usrBootInit because I don't use anything from the NVM. It seems to work , so far. ================================================ Bob DeJoie Kwajalein Atoll Republic of the Marshall Islands email: rjdejoie@kmrmail.kmr.ll.mit.edu�� phonek: 805-355-5855 fax: 805-355-3833 Time Zone: GMT +12 hours�� ================================================ --------------------------- Newsgroups: comp.os.vxworks Subject: Disney vacation 3057 Date: 23 Jun 1999 22:55:49 GMT From: kgqisg@thankyou.com Organization: Global Datalink, Inc. Message-ID: <7kroll$4dt$5802@news.gdi.net> Would you like to find out where can you get, the best priced ever, Disney vacation packages? Go To: http://www.altavista.com/cgi-bin/query?pg=q&kl=XX&q=Disney+vacation+packages and you will find. Thank you. rxkedxxjtmbrotcewzhjgodkntoekqjsbxcbvmyypvgjzbpqgqdiuf --------------------------- Newsgroups: comp.os.vxworks Subject: VxWorks and Zip drives Date: Wed, 23 Jun 1999 13:00:48 -0700 From: Kevin Havig Organization: NAWCWPNS Message-ID: <37713CF0.621D836@navair.navy.mil> I am trying to use a Zip drive with my Motorola MVME 162 CPU (SCSI-1 interface). The scsiAutoConfig command hangs when I call it. If I attach a SCSI hard drive to the CPU the scsiAutoConfig works just fine (using the scsiShow command to display the information). Do I need a special driver to get the Zip drive to work? Any help would be appreciated. Kevin Havig Code 526110D Naval Air Warfare Center China Lake, CA, 93555 havigkk@navair.navy.mil --------------------------- Newsgroups: comp.os.vxworks Subject: etherInputHookAdd and datagram handling Date: Wed, 23 Jun 1999 17:03:15 -0700 From: "JMQ" Organization: Network for Education and Research in Oregon Message-ID: <7krske$ebt$1@news.NERO.NET> I cannot find any information on what vx/works actually does when it grabs a raw packet... does it just strip the network header off and deliver the entire ip datagram (in the tcp/ip protocol instance) to the hook function as *buffer? Or does vxworks actually deterimine what protocol the packet is and strip off the appropriate header, i.e. determine that the datagram is a tcp packet and strip off it's header? Any help on how vxworks deals with this would be appreciated... Thanks. JQ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: etherInputHookAdd Date: Wed, 23 Jun 1999 17:09:12 -0700 From: "JMQ" Organization: Network for Education and Research in Oregon Message-ID: <7krsu9$eit$1@news.NERO.NET> References: <7k2fka$7fs$1@news.kren.nm.kr> <377016C4.3706@wrs.com> This IP Filter hook you speak of... is this only in Tornado II? I can't seem to find any reference to it in 1.0.1. Thanks, JQ Stephen Macmanus wrote in message news:377016C4.3706@wrs.com... It is theoretically possible to change packet contests using > Ethernet hooks. The general idea is to make the necessary changes > to the frame payload, then return FALSE so that the driver > continues processing the frame instead of discarding it. > > While feasible, this approach is not recommended. Not only is > examining every frame expensive, but altering them could cause > any number of unpredictable side effects. For instance, shortening > the frame could be needed, but altering the reported length could > cause memory leaks when the M_BLK chain was finally freed. In > > A somewhat better alternative is to use the IP filter hook. > It examines all IP packets only, avoiding the extra overhead > of testing other frames. This hook returns TRUE instead of FALSE > to continue the built-in processing. It has some of the same > limitations, so changing the packet size is still not recommended. > See the ipFilterLib entries in the manual for details. > > A final possibility is to open a raw socket which receives copies > of all relevant IP packets matching the source and destination > values. > > -- > ------------------ > Stephen Macmanus #include > Software Engineer > Wind River Systems --------------------------- Newsgroups: comp.os.vxworks Subject: Re: FAQ ? Date: Thu, 24 Jun 1999 06:54:15 GMT From: David Hamel Organization: @Home Network Message-ID: <3771D5B5.850F8C7A@ntanet.net> References: Geir R. Westermann wrote: > I'm quite new to this group, and I've been looking for a FAQ. Is there one ? > > Geir comp.os.vxworks has been in need of a FAQ maintainer for a long time. The most recent one covers up to vxworks 5.2 while the current version is vxworks 5.4 --------------------------- Newsgroups: comp.os.vxworks Subject: Tornado on NT vs. Tornado on Solaris? Date: Thu, 24 Jun 1999 09:56:51 GMT From: Ronny Organization: Deja.com - Share what you know. Learn what you don't. Message-ID: <7ksvd0$1r8$1@nnrp1.deja.com> Can someone has experiences regarding advantages/disadvantages of running Tornado on NT, as opposed to Solaris? Ronald - -- Ronald Fischer http://ourworld.compuserve.com/homepages/ronald_fischer/ Sent via Deja.com http://www.deja.com/ Share what you know. Learn what you don't. --------------------------- End of New-News digest ********************** From owner-vxwexplo-process Thu Jun 24 13:35:44 1999 From: Sammer Ostogratzk Date: Thu Jun 24 13:35:48 PDT 1999 Subject: zbuf, mbuf, mblk? What's the story? Hi all vxworks gurus: I noticed that the Tornado documentation describes a zbuf library (zbufLib), however, it looks like networking code uses mbufs (mblk?) instead. I am a bit confused on what we are really supposed to use if we want to use buffer chains. Also, I can't find a description of mbufs, mblks in the documentation. has anyone used the zbufLib? Is it ok to use it or is it an antiquated library that will disappear in newer releases of Tornado? Thanks. ______________________________________________________ Get Your Private, Free Email at http://www.hotmail.com From owner-vxwexplo-process Thu Jun 24 17:25:42 1999 From: Roger Cruz Date: Thu Jun 24 17:25:46 PDT 1999 Subject: RIP is not learning its routes... I need to understand why RIP is not learning the routes broadcasted by another RIP server. Here's the scenario. Environment: Tornado 2, BSP= Yellow Knife 750, Host = Windows NT Config.h changes: #define INCLUDE_RIP /* Routing Information Protocol RIP */ #define RIP_VERSION 2 /* Set to RIP 2 (RFC 1388) */ #define RIP_SUPPLIER 1 /* See pg 120 of Network Prg's guide */ #define RIP_GATEWAY 0 /* See pg 120 of Network Prg's guide */ #define RIP_MULTICAST 1 /* Use RIP multicast addr 224.0.0.9 */ PROBLEM: YK_3 does not learn about route to corporate Enet. Network: -------|--------------------|-------- Corporate Enet | | --------------- ------------------- |172.16.14.100| | YK_1: 172.16.4.1| |Development | | (FEI driver) | |Host PC | | YK_1: 192.0.1.1 | --------------- | (DC driver) | ----------|-------- | ------------------------------------- Experimental Enet | ------------------- | YK_3: 192.0.1.3 | | (FEI driver | ------------------- The second interface in YK_1 (the one with the decchip NIC) and the interface in YK_3 are brought up manually from the shell after the OS has started. I use ripIfSearch() to cause RIP to start looking at the new interfaces. This appears to work. I can put a sniffer on the Exp. Enet and I can see YK_1 RIP broadcasts to the RIP bcast address with the following information: net: 172.16.0.0 mask: 255.255.0.0 gw: 192.0.1.1 metric: 1 This information is what I expect. I also see YK_3 broadcasts which do not contain any entries. When I enable rip debugging (routedDebug = 3), I get the following messages in YK_1: 0xe3e800 (tRipTimerTask): Transmitting RIP message. 0xe3e800 (tRipTimerTask): Looking at IF dc0, route 192 0 1 1 0xe3e800 (tRipTimerTask): Looking at IF fei0, route 172 16 4 1 0xe3e800 (tRipTimerTask): Looking at IF lo0, route 127 0 0 1 0xe3e800 (tRipTimerTask): Transmitting RIP message. 0xe42d78 (tRipTask): Received RIP message. This is telling me that YK_1 can hear the RIP broadcasts from YK_3. In YK_3 I get the following message: 0xe3ed70 (tRipTimerTask): Looking at IF lo0, route 127 0 0 1 0xe3ed70 (tRipTimerTask): Transmitting RIP message. 0xe3ed70 (tRipTimerTask): Looking at IF fei0, route 192 0 1 3 0xe3ed70 (tRipTimerTask): Looking at IF lo0, route 127 0 0 1 0xe3ed70 (tRipTimerTask): Transmitting RIP message. Notice that it never sees the transmit from YK_1 which I know is being broadcasted because of the sniffer output. Here is the current state of the tables and interfaces in YK_3: -> routeShow ROUTE NET TABLE destination gateway flags Refcnt Use Interface ---------------------------------------------------------------------------- 192.0.1.0 192.0.1.3 101 1 0 fei0 ---------------------------------------------------------------------------- ROUTE HOST TABLE destination gateway flags Refcnt Use Interface ---------------------------------------------------------------------------- 127.0.0.1 127.0.0.1 5 0 0 lo0 ---------------------------------------------------------------------------- value = 77 = 0x4d = 'M' -> ripRouteShow 192.0.1.0 192.0.1.3 255.255.255.0 Metric: 1 Timer: 0 RTS_INTERFACE 127.0.0.0 127.0.0.1 255.0.0.0 Metric: 1 Timer: 0 RTS_EXTERNAL RTS_PASSIVE RTS_INTERFACE value = 0 = 0x0 -> ifShow lo (unit number 0): Flags: (0x8069) UP LOOPBACK MULTICAST ARP RUNNING Type: SOFTWARE_LOOPBACK Internet address: 127.0.0.1 Netmask 0xff000000 Subnetmask 0xff000000 Metric is 0 Maximum Transfer Unit size is 32768 0 packets received; 0 packets sent 0 multicast packets received 0 multicast packets sent 0 input errors; 0 output errors 0 collisions; 0 dropped fei (unit number 0): Flags: (0x8063) UP BROADCAST MULTICAST ARP RUNNING Type: ETHERNET_CSMACD Internet address: 192.0.1.3 Broadcast address: 192.0.1.255 Netmask 0xffffff00 Subnetmask 0xffffff00 Ethernet address is 00:90:27:61:48:70 Metric is 0 Maximum Transfer Unit size is 1500 6 packets received; 87 packets sent 2 multicast packets received 86 multicast packets sent 0 input errors; 0 output errors 0 collisions; 0 dropped value = 29 = 0x1d -> -> udpstatShow UDP: 86 total packets 0 input packets 86 output packets 0 incomplete header 0 bad data length field 0 bad checksum 0 broadcasts received with no ports 0 full socket 0 pcb cache lookup failed 0 pcb hash lookup failed value = 26 = 0x1a This tells me that UDP does not seem to be getting the packets, right? The vxWorks image running in both boards is identical, yet YK_1 sees the broadcasts from YK_3, why? any help is extremely appreciated. Thanks. Roger PS: Send replies to me directly to the address below. Remove "_nojunk" first. rogerc_nojunk@ignitus.com vxworks, tornado From owner-vxwexplo-process Fri Jun 25 04:00:11 1999 From: daemon@csg.lbl.gov Date: Fri Jun 25 04:00:15 PDT 1999 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Fri Jun 25 04:00:07 PDT 1999 Subject: Booting VxWorks from PCMCIA/Flashdisk, nearly there but ... Subject: Sample VxWorks driver for raw device? Subject: Re: Looking for a cheat sheet for Vx works command line utilities Subject: Re: a problem while linking Subject: Re: Question about SLIP connection Subject: Re: sendmst Subject: Re: etherInputHookAdd Subject: Re: etherInputHookAdd and datagram handling Subject: Re: Sample VxWorks driver for raw device? Subject: Re: Where is Vxworks task scheduler? Subject: Re: Tornado on NT vs. Tornado on Solaris? Subject: Re: UDP performance over PPP Subject: Re: Tornado on NT vs. Tornado on Solaris? ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Booting VxWorks from PCMCIA/Flashdisk, nearly there but ... Date: Thu, 24 Jun 1999 12:28:30 +0100 From: Simon Roberts Message-ID: <3772165E.8962794B@roke.co.uk> - --------------4EB6432B0FC00789A728C0D6 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit We are using an MVME2400 with a PMC host bus adaptor which allows us to plug in a pcmcia flashdisk (220Mb at present). BOOTP params have been changed along with some kernel files in order that we can boot directly from the flashdisk and use it as a mounted DOS filesystem. So far so good. I was having some problems with network interfaces until I set the VxWorks boot parameter 'others' to 'dc', the full configuration settings now being: boot device : pcmcia=2 unit number : 0 processor number : 0 host name : wheelie file name : /pcfsA/vxworks inet on ethernet (e) : 193.118.193.191:fffff800 host inet (h) : 193.118.193.186 flags (f) : 0x8 target name (tn) : umtscu4 other (o) : dc Whilst the boot now shows an attached TCP/IP interface ... Attached TCP/IP interface to dc unit 0 Attaching network interface lo0... done. ... network related commands still don't work as well as they do had I booted the card conventionally from a server over the ethernet network. But at least now I can use a host target shell in Tornado. Several of the more obvious failures at this point are: 1. If you type 'i' in to the host target shell you just get a banner, but no task information 2. If you type '@ls' into host target shell you get an error message resembling: couldn't squeeze "" out of the serial port on the card, and the board stops. 3. An 'ls' in the target shell fails (ie. value = -1 ...), but listing the dosFs explicitly (i.e. ls "/pcfsA") works. Having had no problems booting cards from a network server, I am suspicious of the differences in boot diagnostics, is the fact that the network version first attaches network interfaces significant? Network boot: Attaching network interface dc0... done. Attaching network interface lo0... done. Loading... 804856 Starting at 0x100000... Attached TCP/IP interface to dc unit 0 Attaching network interface lo0... done. NFS client support not included. Loading symbol table from wheelie:mv2400/umtscu5/vxworks.sym ...done Flashdisk boot: PMC230: Detected a 220200960 Byte Disk !!! done. loading /pcfsA/vxworks 845180 Starting at 0x100000... PMC230: Detected a 220200960 Byte Disk !!! User: target Attached TCP/IP interface to dc unit 0 Attaching network interface lo0... done. NFS client support not included. Loading symbol table from /pcfsA/vxworks.sym ...done Any hints or kicks in the right direction much appreciated! Simon Roberts. - --------------4EB6432B0FC00789A728C0D6 Content-Type: text/html; charset=us-ascii Content-Transfer-Encoding: 7bit We are using an MVME2400 with a PMC host bus adaptor which allows us to plug in a pcmcia flashdisk (220Mb at present). BOOTP params have been changed along with some kernel files in order that we can boot directly from the flashdisk and use it as a mounted DOS filesystem. So far so good.

I was having some problems with network interfaces until I set the VxWorks boot parameter 'others' to 'dc', the full configuration settings now being:

boot device          : pcmcia=2
unit number          : 0
processor number     : 0
host name            : wheelie
file name            : /pcfsA/vxworks
inet on ethernet (e) : 193.118.193.191:fffff800
host inet (h)        : 193.118.193.186
flags (f)            : 0x8
target name (tn)     : umtscu4
other (o)            : dc

Whilst the boot now shows an attached TCP/IP interface ...

Attached TCP/IP interface to dc unit 0
Attaching network interface lo0... done.

... network related commands still don't work as well as they do had I booted the card conventionally from a server over the ethernet network. But at least now I can use a host target shell in Tornado.

Several of the more obvious failures at this point are:

1. If you type 'i' in to the host target shell you just get a banner, but no task information
2. If you type '@ls' into host target shell you get an error message resembling: couldn't squeeze "" out of the serial port on the card, and the board stops.
3. An 'ls' in the target shell fails (ie. value = -1 ...), but listing the dosFs explicitly (i.e. ls "/pcfsA") works.

Having had no problems booting cards from a network server, I am suspicious of the differences in boot diagnostics, is the fact that the network version first attaches network interfaces significant?

Network boot:

Attaching network interface dc0... done.
Attaching network interface lo0... done.
Loading... 804856
Starting at 0x100000...

Attached TCP/IP interface to dc unit 0
Attaching network interface lo0... done.
NFS client support not included.
Loading symbol table from wheelie:mv2400/umtscu5/vxworks.sym ...done

Flashdisk boot:

PMC230: Detected a 220200960 Byte Disk !!!
done.
loading /pcfsA/vxworks
845180
Starting at 0x100000...

PMC230: Detected a 220200960 Byte Disk !!!
User: target
Attached TCP/IP interface to dc unit 0
Attaching network interface lo0... done.
NFS client support not included.
Loading symbol table from /pcfsA/vxworks.sym ...done

Any hints or kicks in the right direction much appreciated!

Simon Roberts.
- --------------4EB6432B0FC00789A728C0D6-- --------------------------- Newsgroups: comp.os.vxworks Subject: Sample VxWorks driver for raw device? Date: Thu, 24 Jun 1999 14:08:06 +0200 From: psav Organization: Centre Universiaire d'Archamps Message-ID: <37721FA6.493D1172@cur-archamps.fr> I'm looking for a sample about creating and installing a raw device and its driver with VxWorks 5.2. Thank you. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Looking for a cheat sheet for Vx works command line utilities Date: Thu, 24 Jun 1999 16:12:35 GMT From: Charlie Grames Organization: Boeing Message-ID: <377258F3.FD5AA5E0@nowhere.com> References: <376efa3c.9439505@news.lvdi.net> Sender: nntp@news.boeing.com (Boeing NNTP News Access) Huh? I'm not sure what you're looking for, but if you want the list of VxWorks routines, you need to look in the VxWorks Reference Manual. Keep in mind that, with the exception of Tornado host-specific commands (which are documented in the User's Guide), VxWorks "command line utilities" are the same as VxWorks routines. If you want information on binaries packaged with Tornado (e.g., tgtsvr), it can be found in the User's Guide. Charlie Grames The Boeing Company (314) 233-1956 Charles.R.Grames @ boeing.com Riley wrote: > > Meets and greets.. > > I'm looking for a listing of VxWorks command line utilities, with > syntax / switches...PLEASE! > > I will put you in my will... > > - Riley --------------------------- Newsgroups: comp.os.vxworks Subject: Re: a problem while linking Date: Thu, 24 Jun 1999 17:11:35 GMT From: "Clifford Slocombe" Message-ID: References: <377056b3.0@dnews.sbell.com.cn> <37711288.4C78F671@triumf.ca> If you need to link multiple modules, which is likely if you have a large number of modules, and necessary if you have two or more modules that reference each other, you must incrementally link using the -r option I think but my manuals and my Internet connection are not in the same location! Graham Waters wrote in message news:37711288.4C78F671@triumf.ca... > liuf wrote: > > > > hi, > > > > I just begin to use vxworks. I use NT as host and use a PC as target. > > When I link my test code, as > > >cc386 -o testExe -nostdinc -fno-builtin -v -LC:\x86\target\lib > > testCode.o -lI80486gnuvx > > it shows an error: > > cannot open crt0.o: No such file or directory > > > > Can anyone tell me what ctr0.o is? why cc386 automatically link it with my > > code? where can I find this? > > > > Thanks! > > The vxWorks kernel include all run-time libraries and uses > dynamic loading and unloading of object modules with > run-time relocation and linking. It is only necessary to compile a > program and load it. Also do not use the statement "main()", use > some function name, and call that function by name from the the vxWorks > shell to run it. > > Your compiler line should look something like this: > > cc386 (your compile flags) -c testExec.c > > The -c option means compile but do not link. > > Regards > Graham. > > +---------------------------------------------------------------------+ > |Graham Waters, Control System Eng Internet: waters@triumf.ca| > |TRIUMF --- University of British Columbia, phone: (604)-222-1047 | > |Vancouver, BC, Canada Fax: (604)-222-7307 | > |http://edevel9pc.triumf.ca/homepage/graham01.html | > +---------------------------------------------------------------------+ > > For every expert there is an equal and opposite expert > - Arthur C Clark --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Question about SLIP connection Date: Thu, 24 Jun 1999 17:11:37 GMT From: "Clifford Slocombe" Message-ID: References: <7kipgl$3qn1@overload.lbl.gov> <7klnes$dpd$1@nnrp1.deja.com> <7kof29$dul$1@nnrp1.deja.com> In the end, I looked in some old newsgroup articles using Alta Vista to search them. What nobody was saying was that when NT dials up a server, it sends the string CLIENT, and expects the response CLIENTSERVER. This is neither PPP or obvious. I could see the CLIENT string, and initially guesssed a response of SERVER...close, but no cigar! After this exchange, it is straight PPP, and PAP, CHAP, or any other password protocol are irrellevent unless you want to enforce that kind of security. I did not, and I simply had access, I pinged, FTP'd, and Telnet'd, all OK. I assume that the same exchange in reverse must take place for dialing-in to NT. To dial into the VxWorks target I started PPP in passive mode, and then executed the code below, which waits for the CLIENT string, and responds as required. I then invoked dial-up networking from NT using the null modem device, and got straight in. I guess the code below could be added as a PPP initialisation hook, I intend to implement a dial/answer routine here for when I add a modem. /* NT originated PPP connection handshake */ #include "stdio.h" #include "string.h" #include "vxWorks.h" #include "ioLib.h" #define CLIENT_POLL "CLIENT" #define SERVER_RESPONSE "CLIENTSERVER" void pppNTDialInStarter( char *device, int rate ) { int handle ; int status ; char c ; int index ; char poll[] = CLIENT_POLL ; handle = open( device, O_RDWR, 0 ) ; status = handle ; if( status != ERROR ) { status = ioctl( handle, FIOSETOPTIONS, OPT_RAW ) ; } if( status != ERROR ) { status = ioctl( handle, FIOBAUDRATE, rate ) ; } index = 0 ; while( status != ERROR && poll[index] != '\0' ) { status = read( handle, &c, 1 ) ; if( status != ERROR && c == poll[index] ) { index++ ; } else { index = 0 ; } } if( status != ERROR ) { status = write( handle, SERVER_RESPONSE, strlen(SERVER_RESPONSE) ) ; } } wrote in message news:7kof29$dul$1@nnrp1.deja.com... > > > But now what do I do to connect to VxWorks? > We turn off PAP and CHAP authentication then do our > own username/password prompting, verify those then run > pppInit on the vxWorks side. To connect from Windows > Dial-up Networking, I say "Pop up a terminal window before dialing" > that lets me respond to the username/password prompt. When > pppInit starts on the vxWorks side, it will spit out a message > like "My IP is a.b.c.d, your IP is e.f.g.h, MTU is 1500 bytes" > _then_ begin transmitting the LCP packets. On Windows, I exit > from the popup window, then Dial-up networking starts dealing > with the LCP packets and establishes the link. > > > > Sent via Deja.com http://www.deja.com/ > Share what you know. Learn what you don't. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: sendmst Date: Tue, 22 Jun 1999 16:28:05 -0700 From: Stephen Macmanus Organization: Wind River Systems Message-ID: <37701C05.1206@wrs.com> References: <376A4991.21E48482@tekelec.com> Kevin Sapp wrote: > > VxWorks Users, > > The VxWorks documentation states that sendmsg can only be used > with a udp (datagram) socket. BSD versions (solaris) can use it > for TCP or UDP. > > We have some vendor code that is using this in a "portable" > implementation. Is the VxWorks documentation correct (only > UDP) for VxWorks or can it be used for both. > > Any insight would be appreciated > > Thanks > Kevin You can consider this a documentation screwup. I guess the author either considered the vectored send more "natural" for UDP or generically referred to all sockets as datagram sockets. The sendmsg() routine will work with both TCP and UDP sockets. - -- - ------------------ Stephen Macmanus #include Software Engineer Wind River Systems --------------------------- Newsgroups: comp.os.vxworks Subject: Re: etherInputHookAdd Date: Thu, 24 Jun 1999 14:59:45 -0700 From: Stephen Macmanus Organization: Wind River Systems Message-ID: <3772AA51.EED@wrs.com> References: <7k2fka$7fs$1@news.kren.nm.kr> <377016C4.3706@wrs.com> <7krsu9$eit$1@news.NERO.NET> JMQ wrote: > > This IP Filter hook you speak of... is this only in Tornado II? I can't > seem to find any reference to it in 1.0.1. > > Thanks, > > JQ The IP filter hook is not formally documented until Tornado 2 but is also included with the SENS version of the network stack. It is not available with the original network stack included in the Tornado 1.0.1 release. Stephen - -- - ------------------ Stephen Macmanus #include Software Engineer Wind River Systems --------------------------- Newsgroups: comp.os.vxworks Subject: Re: etherInputHookAdd and datagram handling Date: Thu, 24 Jun 1999 15:05:04 -0700 From: Stephen Macmanus Organization: Wind River Systems Message-ID: <3772AB90.5521@wrs.com> References: <7krske$ebt$1@news.NERO.NET> JMQ wrote: > > I cannot find any information on what vx/works actually does when it grabs a > raw packet... does it just strip the network header off and deliver the > entire ip datagram (in the tcp/ip protocol instance) to the hook function as > *buffer? Or does vxworks actually deterimine what protocol the packet is > and strip off the appropriate header, i.e. determine that the datagram is a > tcp packet and strip off it's header? > > Any help on how vxworks deals with this would be appreciated... > The etherInputHook mechanism passes entire frames to the registered functions, including the Ethernet header. Testing the frame for IP information and for other matches (such as desired port numbers) is the responsibility of the called routine. > Thanks. > > JQ - -- - ------------------ Stephen Macmanus #include Software Engineer Wind River Systems --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Sample VxWorks driver for raw device? Date: Thu, 24 Jun 1999 15:05:00 -0700 From: David Holloway Organization: Wind River Systems Message-ID: <3772AB8C.DBAC3A0@wrs.com> References: <37721FA6.493D1172@cur-archamps.fr> <3772A9E4.9DBD739C@systemsinterface.com> Reply-To: david.holloway@wrs.com Bob Irwin wrote: > > psav wrote: > > > > I'm looking for a sample about creating and installing a raw device and > > its driver with VxWorks 5.2. Thank you. > > What kind of raw device? If its a disk, you can use the Raw File System > (p 244 5.2 programmer's guide). If it's scsi based, but not disk, you > can use scsiLib. I think he is referring to block devices as opposed to character devices that are common in unix kernels. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Where is Vxworks task scheduler? Date: Thu, 24 Jun 1999 22:49:53 GMT From: Yiming Organization: Share what you unknow. Learn what you have known, oops Message-ID: <7kucme$jj7$1@nnrp1.deja.com> References: <7kobfm$cct$1@nnrp1.deja.com> <7krfbo$nsh$1@shell7.ba.best.com> In article <7krfbo$nsh$1@shell7.ba.best.com>, canuck@best.com (Chris Wein) wrote: > Instead of modifying the source, why don't you just use the task > switch hook (taskSwitchHookAdd) to measure your # of switches? > It may also help with measuring cpu time for the scheduler but you > run into the problem of the intrumentation affecting the measurement. > Still, the hook is a good place to start. > Yes, but beware if scheduler is activated but reschedule decide it has to pick the same one (as before rescheduling), no taskSwitchHook will be called. - -- Zhang #include Sent via Deja.com http://www.deja.com/ Share what you know. Learn what you don't. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Tornado on NT vs. Tornado on Solaris? Date: Fri, 25 Jun 1999 05:06:48 GMT From: bobarmstrong@home.nothiis.net (Robert Armstrong) Organization: @Home Network Message-ID: References: <7ksvd0$1r8$1@nnrp1.deja.com> Well, being a recidivist, I prefer the command line on Solaris for building. I'm a vxWorks user and all this Tornado stuff is very pretty but nothing to do with what I'm trying to deliver. To be fair I don't know how I'd get by without WindView sometimes... just not where I do most of my work. Other than that I actually end up using NT most of the time and now that we can dump the god awful editor I can actually write some code without a continuous feeling of irritation. So far as I can tell the compiler works the same for both platforms and likewise the Makefiles seem to survive the experience of living on a Gates platform. On the bright side Tornado seems to have its act together on NT, on the dark side 64Mb RAM works fine on a three year old Solaris box and 128Mb of RAM seems to be barely enough on a year old NT box. So if it were my money? Windows platforms are so cheap and there is so little difference in product performance that I have to go for the PC... then again, if they ever decide to port to Linux that's the end of NT for me :) Yes after 15 years of diabolical PC pain there is actually light at the end of the tunnel, first we can ditch MS and then it's just a short sidestep out of the x86 architecture. I have a dream... Bob In article <7ksvd0$1r8$1@nnrp1.deja.com>, Ronny wrote: >Can someone has experiences regarding advantages/disadvantages of >running Tornado on NT, as opposed to Solaris? > >Ronald --------------------------- Newsgroups: comp.os.vxworks Subject: Re: UDP performance over PPP Date: Fri, 25 Jun 1999 05:12:30 GMT From: bobarmstrong@home.nothiis.net (Robert Armstrong) Organization: @Home Network Message-ID: <2dEc3.22178$NY6.7012@news.rdc1.sfba.home.com> References: <7i1brc$47a$1@nnrp1.deja.com> <37715BBD.48B89842@usa.alcatel.com> In article <37715BBD.48B89842@usa.alcatel.com>, Fadi Nasser wrote: >> I'm interested in finding out more detail about how the higher layers >> of a protocol stack (specifically SNMP/UDP) are affected by running >> over IP/PPP instead of IP/Ethernet. My application seems to have >> serious problems when all of a sudden the bandwidth dips from >> 10 Mbps to 9600 bps. >> >> I would like to know how buffers are managed >> within the protocol stack and if there's any API to the buffer >> management that would allow me to track utilization. Then I could >> "throttle back" the generation of UDP traffic to match the buffer >> availability. >> >> I'd also like to know how PPP arbitrates between traffic demands from >> each peer. That is, if one peer is sending a ton of UDP traffic, it >> seems like the other peer can't get a packet in (or perhaps it's the >> case that the responses are queued way back in the first peer?). Are >> there any priority schemes within the protocols that would allow me to >> always have fast response, regardless of how much "bulk" UDP traffic is >> queued up in one peer? >> >> My envrionment is VxWorks 5.3.1 on i386. Their documentation states >> "PPP attaches to the TCP/IP stack at the driver (link) level." Does >> this mean that all the buffer management is handled within the serial >> port driver, which likely uses a fairly small character FIFO? >> >> Thanks for any help. Though it never seems like a satisfactory response, if you read the Stevens books you will find that there is a type field in the IP header and that it can be used to flag some traffic as immediate. Although this is pretty much falling out of use the one place I have still seen it used is in PPP for queue jumping (it said). Good luck, Bob --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Tornado on NT vs. Tornado on Solaris? Date: Fri, 25 Jun 1999 07:30:58 GMT From: Ronny Organization: Deja.com - Share what you know. Learn what you don't. Message-ID: <7kvb7j$tt5$1@nnrp1.deja.com> References: <7ksvd0$1r8$1@nnrp1.deja.com> (this message mailed and posted) In article , bobarmstrong@home.nothiis.net (Robert Armstrong) wrote: > Well, being a recidivist, I prefer the command line on Solaris for building. > I'm a vxWorks user and all this Tornado stuff is very pretty but nothing to do > with what I'm trying to deliver. Does this mean you don't have a command line access to all the tools when running under NT? WindRiver told us that development under Solaris and NT looks the same. We like to automate a lot, and good scripting possibilities are a necessity. Also, we want to use our own Make utility, not that one which is included with Tornado. Therefore, I would like to be able to write a Perl or Tcl Script where I could drive every aspect of the development cycle. Given that Tornado includes Tcl, and that Perl is available for Solaris and NT alike, is there anything I can not do under NT that I could do under Solaris? Ronald - -- Ronald Fischer http://ourworld.compuserve.com/homepages/ronald_fischer/ Sent via Deja.com http://www.deja.com/ Share what you know. Learn what you don't. --------------------------- End of New-News digest ********************** From owner-vxwexplo-process Fri Jun 25 11:07:45 1999 From: "Eckardt, Suzie" Date: Fri Jun 25 11:07:49 PDT 1999 Subject: VxWorks MIPS Interrupt Vectors This message is in MIME format. Since your mail reader does not understand this format, some or all of this message may not be legible. ------_=_NextPart_001_01BEBF35.7AD0CA50 Content-Type: text/plain; charset="iso-8859-1" I am trying to get a custom BSP running on a board based on a MIPS R3000 variant. My only debug tool is a Logic Analyzer, so the information I can collect is limited. I found that I am failing on all the intConnect() calls in the BSP. The Reference Manual states that intConnect() calls intHandlerCreate() and intVecSet(); I never see the call to intVecSet() so I assume I'm failing in intHandlerCreate(). Anyway, intConnect() returns ERROR and errno is set to 0x003d0001 which is S_objLib_OBJ_ID_ERROR. What does it mean to "connect" to an interrupt vector in the VxWorks/MIPS world? Where are the vectors: the Programmer's Guide says there are 256 of them in the 512 bytes starting at 0x80000000, the Reference Manual says they are in excBsrTbl? What is stored in the vector: an instruction or an address? Where does intHandlerCreate() create the ISR wrapper: in malloc'd memory or in pre-allocated static memory? Thanks, Suzie ************************************************ ** Suzie Eckardt ** Innovative Concepts, Incorporated ** McLean, VA ** eckardt@innocon.com ** 703-893-2007 x258 ************************************************ ------_=_NextPart_001_01BEBF35.7AD0CA50 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable VxWorks MIPS Interrupt Vectors

I am trying to get a custom BSP running on a board = based on a MIPS R3000 variant. My only debug tool is a Logic = Analyzer, so the information I can collect is limited.

I found that I am failing on all the intConnect() = calls in the BSP. The Reference Manual states that intConnect() = calls intHandlerCreate() and intVecSet(); I never see the call to = intVecSet() so I assume I'm failing in intHandlerCreate(). = Anyway, intConnect() returns ERROR and errno is set to 0x003d0001 which = is S_objLib_OBJ_ID_ERROR.

What does it mean to "connect" to an = interrupt vector in the VxWorks/MIPS world? Where are the = vectors: the Programmer's Guide says there are 256 of them in the 512 = bytes starting at 0x80000000, the Reference Manual says they are in = excBsrTbl? What is stored in the vector: an instruction or an = address? Where does intHandlerCreate() create the ISR wrapper: in = malloc'd memory or in pre-allocated static memory?

Thanks,
Suzie

************************************************
** Suzie Eckardt
** Innovative Concepts, Incorporated
** McLean, VA
** eckardt@innocon.com
** 703-893-2007 x258
************************************************

------_=_NextPart_001_01BEBF35.7AD0CA50-- From owner-vxwexplo-process Sun Jun 27 09:58:15 1999 From: "Jimmie T Marks" Date: Sun Jun 27 09:58:18 PDT 1999 Subject: unld I am having problems with the unld after loading and executing a file. I am using Tornado / vxWorks 5.3.1. Below is some code and shell print outputs. Could you look at this and provide some suggestions? Would appreciate any help. Thankyou. Jimmie Marks Raytheon Systems Co. jtmark@ftw.rsc.raytheon.com A file is loaded and executed successfully with the loadModule and symFindByName with the taskSpawn functions. The executed file is then unloaded unsuccessfully using the unld function. The memory error message is: 0x3e18d08 (u1): memPartFree: invalid block 0x3e190d4 in partition 0x2811a0. The following is the code used for unld: //**************************************************************************************** //**************************************************************************************** // This function uses the method unld() to unload ** // a specific file name that represents a module that is loaded. ** //**************************************************************************************** //**************************************************************************************** taskDelay(60); char *tempName = new char[strlen(fileName) + 1]; strcpy(tempName, fileName); // copy the const fileName to a temp var moduleShow(tempName, 1); MODULE_ID mod = moduleFindByName(tempName); // get the module id for the module name if(mod == NULL) { cerr << "ERROR: cannot find the module \"" << tempName << "\"" << endl; delete []tempName; return false; } else { // Unload the module if(unld(mod, 1) == ERROR) { cerr << "ERROR: cannot load the module with id: " << mod << endl; delete [] tempName; return false; } else { cerr << "File \"" << tempName << "\" unloaded." << endl; moduleShow(tempName, 1); delete [] tempName; return true; } } Here is the shell output that comes with it: TESTING LOAD()......................... Load called with: "c/lala.o" File Descriptor: 33 FILE: c/lala.o Load Successful for "lala.o". Activating _foo__Fv....... file enty point found! _foo__Fv task being activated (spawned) ********************************************** ********************************************** This is the file lala.o that was loaded by foo ********************************************** ********************************************** LOAD() DONE. --------------------------------------------------------------------- TESETING OF UNLOAD().................. Trying to unload lala.o....... MODULE NAME MODULE ID GROUP # TEXT START DATA START BSS START --------------- ---------- ---------- ---------- ---------- ---------- lala.o 0x3e12d18 6 0x3e18f94 0x3e190e8 0x3e190e8 Size of text segment: 340 Size of data segment: 0 Size of bss segment: 4 Total size : 344 0x3e18d08 (u1): memPartFree: invalid block 0x3e190d4 in partition 0x2811a0. Any Suggestions? From owner-vxwexplo-process Mon Jun 28 01:17:56 1999 From: tim allen Date: Mon Jun 28 01:18:00 PDT 1999 Subject: Re: Booting VxWorks from PCMCIA/Flashdisk, nearly there but ... >... network related commands still don't work as well as they do had I >booted the card conventionally from a server over the ethernet network. >But at least now I can use a host target shell in Tornado. It may be that your host can't find the VxWorks boot image and symbol table. When the Tornado host shell attaches to the target, by default it tries to locate the boot image by using the boot file pathname from the boot parameters on the target. The host has a different view of the local file system and so the boot image and symbols are not found when you boot, for example, from pcmcia. You can work around this by forcing the path in the Tornado Tools->Target Server->Configure option. Select Target Server Properties->Core File and Symbols, where you can give the path to the boot image stored on your host. When you boot from the host, over the network, this issue does not arise. --------------------- Tim Allen Ltd Real Time Embedded Software Consultancy ______________________________________________________ Get Your Private, Free Email at http://www.hotmail.com From owner-vxwexplo-process Mon Jun 28 08:53:04 1999 From: "Jim Hawthorne" Date: Mon Jun 28 08:53:07 PDT 1999 Subject: Announcing RTADK 4.0 Shareware for VxWorks and Tornado VxWorkers- Announcing the Realtime Application Development Kit RTADK 4.0 for Applications Running under VxWorks and Tornado Applied Technology Associates, an Albuquerque, NM based R&D company has announced the shareware release of RTADK 4.0 for VxWorks and Tornado. Over the past 23 years, ATA has earned praise for our support of R&D projects and experiments from Florida to Hawaii. As a New Mexico corporation, we maintain an emphasis on support of defense-related research sites in our home state, including the Air Force Research Laboratory at Kirtland AFB, the High Energy Laser Test Facility (HELSTF) at White Sands Missle Range, and Sandia National Laboratories. ATA has been successful in fielding realtime systems that operate on the ground, in the air, in space, on telescopes, balloons, research aircraft, and imaging satellites. From microelectronics to control room operations to laser-related pointing and tracking systems, we have been involved in the development and implementation of realtime systems. As the result of years of realtime experience we have developed a "toolkit" of reusable realtime application development components that we have called the Realtime Application Development Kit or RTADK. The intent of RTADK, is to make realtime application development faster and easier by providing a basic framework on which to build your application. Important features of the Realtime Application Development Kit include: - Resource Discovery Service allowing the target system to identify itself by name, IP address, and other application- specific information. - State-of-Health Service allowing the target to be polled for current state and condition. - Interprocess communications services providing support for message passing between the target system�s tasks and event synchronization. - Interprocess communications monitoring and logging for development and/or system debugging and testing. - Interprocess communications recording and playback to assist developer in subsystem simulation and troubleshooting difficult intermittent behavior. - Support for commanding over a LAN or the Internet using TCP and UDP. - Support for commanding via RS232. - Standard Command Set providing immediate support for a few simple tasks and providing examples of how to code application- specific command handlers. Unneeded commands can be deactivated by application. - Configurable Command Table that can be modified to allow commands to be accessed only through a specific media to prevent unauthorized access. - Application-specific commands are added simply by writing a command handler and adding the command to the Command Table at runtime. - Developer can insert application-specific Error Codes and Messages into the Error Message Table for use in error handling. - Fully configurable. Socket port numbers, serial port device names, which services to include� Use only what your application needs. Best of all, RTADK 4.0 is being distributed as shareware!!! Visit our web page at http://www.aptec.com/ to learn more about how you can develop realtime systems faster, easier, and become more productive with RTADK 4.0. From owner-vxwexplo-process Mon Jun 28 08:54:15 1999 From: Mauro Martins Date: Mon Jun 28 08:54:20 PDT 1999 Subject: Access SDRAM and DRAM in FADS8xx Hi all, This mail concerns the following tools: Tornado 1.0.1 on Sun Solaris for Power PC target Motorola's MPC860FADS board. I would like to use both SDRAM banks and DRAM banks at the same time in the board. Has anybody made this modification in the bsp and could help me? tanks in advance Mauro Martins -- _______________________________________________________ Mauro Jose Galarda Martins Siemens Tel TR-DS Rua Pedro Gusso, 2635 CEP 81310-900 Curitiba - Parana' - Brasil Tel: +55 41 3416457 Fax: +55 41 3415620 e-mail: mauro.martins@siemens.com.br _______________________________________________________ From owner-vxwexplo-process Tue Jun 29 04:00:29 1999 From: daemon@csg.lbl.gov Date: Tue Jun 29 04:00:34 PDT 1999 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Tue Jun 29 04:00:25 PDT 1999 Subject: MV2604 Universe DMA Subject: Re: MV2604 Universe DMA Subject: Re: Question about SLIP connection Subject: esmc driver (91c94) losing pages Subject: Re: Where is Vxworks task scheduler? Subject: Windview Help Subject: Re: using multiple routing protocols in SENS Subject: Network Address Translation and Application Level Gateways Subject: Re: MV2604 Universe DMA Subject: Anyone use Aisys's DriveWay 3DE Subject: R: Tornado on NT vs. Tornado on Solaris? Subject: Re: vxWorks UDP question Subject: C_MASK16 for 860 Subject: Re: Help: PPP link established, but can't do anything? Subject: HELP : Tornado1.0.1 + SENS1.1 Subject: Unaligned access fault Subject: Re: MV2604 Universe DMA Subject: Re: Network Address Translation and Application Level Gateways Subject: Re: HELP : Tornado1.0.1 + SENS1.1 Subject: Re: Unaligned access fault ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: MV2604 Universe DMA Date: Mon, 28 Jun 1999 10:51:08 -0400 From: Andrew Reddig Organization: Shore.Net/Eco Software, Inc; (info@shore.net) Message-ID: <37778BDC.74BB75DD@tekmicro.com> I'm trying to use the Universe DMA controller under VxWorks, and the DMA throughput is about 1/2 to 1/3 of what I would expect (7-10 MB/s, even using D64 MBLT). The card that we are talking to supports 30 MB/s D32 BLT to another VME host, so I'm pretty sure that the Universe is the issue. I suspect an issue with the Universe's priority on the PCI bus or the PCI bus priority into DRAM, but haven't been able to nail it down yet. Can anyone provide any hints, pointers, advice, experience? VxWorks is 5.3.1, mv2604 BSP 1.1/4. Thanks! - -- Andrew Reddig andyr@tekmicro.com TEK Microsystems, Incorporated 781.270.0808 One North Avenue Fax 781.270.0813 Burlington, MA 01803 Web http://www.tekmicro.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: MV2604 Universe DMA Date: Mon, 28 Jun 1999 08:24:54 -0700 From: Scott Kamara Organization: AP Labs Message-ID: <377793C6.81F8762D@pop.sttl.uswest.net> References: <37778BDC.74BB75DD@tekmicro.com> Reply-To: spk@uswest.net The rates we see with the mv2604/UnivII are much better. You might want to verify via a VME bus analyzer the slave response time for a given d64 or d32 blt word-cycle is sufficient to support 30 MB/s. Are you using your own dmac driver or Motorola's? If Motorola's (non-WRS-delivered), it is documented that theirs is a polled interface which could degrade the rate you would get if it was interrupt-on-complete driven. The source and destination must be 8-byte aligned. That is, the lower three bits of the source and dest address (vme addr and local mem addr) must be the same. If not aligned right, I dont know what kind of side-effects this would have. Andrew Reddig wrote: > > I'm trying to use the Universe DMA controller under VxWorks, and the DMA > throughput is about 1/2 to 1/3 of what I would expect (7-10 MB/s, even > using D64 MBLT). The card that we are talking to supports 30 MB/s D32 > BLT to another VME host, so I'm pretty sure that the Universe is the > issue. > > I suspect an issue with the Universe's priority on the PCI bus or the > PCI bus > priority into DRAM, but haven't been able to nail it down yet. > > Can anyone provide any hints, pointers, advice, experience? > > VxWorks is 5.3.1, mv2604 BSP 1.1/4. > > Thanks! > > -- > Andrew Reddig andyr@tekmicro.com > TEK Microsystems, Incorporated 781.270.0808 > One North Avenue Fax 781.270.0813 > Burlington, MA 01803 Web http://www.tekmicro.com - -- - ----------------------- Scott Kamara AP Data Systems Seattle Engr. Office (206) 525-8592 scott@aplabs.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Question about SLIP connection Date: Fri, 25 Jun 1999 08:53:32 GMT From: "Ian Rudge" Message-ID: References: <7kipgl$3qn1@overload.lbl.gov> <7klnes$dpd$1@nnrp1.deja.com> <7kof29$dul$1@nnrp1.deja.com> Ok, further to my previous message, I am now trying to get the VxWorks end to originate the connection. I am sending CLIENT from the target, and NT responds with CLIENTSERVER. But when I run pppInit, according to the modem log file, the NULL modem hangs up as soon as it starts to receive data. BTW I cannot see how to explicitly turn CHAP or PAP in NT off, so far I have just ignored it. Is this the cause of my problem? Clifford. --------------------------- Newsgroups: comp.os.vxworks Subject: esmc driver (91c94) losing pages Date: Wed, 23 Jun 1999 19:41:52 GMT From: john_dwinell@my-deja.com Organization: Deja.com - Share what you know. Learn what you don't. Message-ID: <7krd9t$gdd$1@nnrp1.deja.com> I am using the esmc driver with VxWorks. My target is a Hithachi SH-3 (7729) with the SMC91C94 Ethernet controller. The driver seems to work okay, but under heavy traffic, it loses pages. We have found that the remove and release command appears to release the wrong number of pages (possibly the number of pages for the next packet in the fifo). Once this happens, the pages are lost forever. We setup a test where the problem would occur after sending 13 large pings.If I reset the MMU after the 12th packet, the problem does not occur. Of course, I can only do this during testing. Can anyone help? Sent via Deja.com http://www.deja.com/ Share what you know. Learn what you don't. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Where is Vxworks task scheduler? Date: 23 Jun 1999 13:16:56 -0700 From: canuck@best.com (Chris Wein) Organization: a user of Best Internet Communications, Inc. www.best.com Message-ID: <7krfbo$nsh$1@shell7.ba.best.com> References: <7kobfm$cct$1@nnrp1.deja.com> In article <7kobfm$cct$1@nnrp1.deja.com>, wrote: >I have (so I'm told) the source to the VxWorks kernel somewhere in some >Clearcase VOB. I need to find the scheduler, so I can put some >isntrumentation in it. (I'm trying to find out how frequently we're task >switching, and how much CPU time the scheduler eats up.) > >Can anyone tell me what to grep on? What's the name of the scheduler's >routine? Or the name of the module it lives in? Thanks in advance. Instead of modifying the source, why don't you just use the task switch hook (taskSwitchHookAdd) to measure your # of switches? It may also help with measuring cpu time for the scheduler but you run into the problem of the intrumentation affecting the measurement. Still, the hook is a good place to start. --------------------------- Newsgroups: comp.os.vxworks Subject: Windview Help Date: Fri, 25 Jun 1999 07:44:03 -0400 From: "David R. Kuechenmeister" Organization: Georgia Institute of Technology, Atlanta GA, USA Message-ID: <250619990744033113%David.Kuechenmeister@gtri.gatech.edu> References: <376efa3c.9439505@news.lvdi.net> <377258F3.FD5AA5E0@nowhere.com> Reply-To: David.Kuechenmeister@gtri.gatech.edu I am having trouble with a new installation of Windview for our MV2604 BSP. I hoped that someone else had resolved similar problems and could offer some advice. The lab recently bought the new Windview 2.0; this came with a new SENS installation, as well. After I installed the stuff on the CD, that is a tornado patch, SENS and Windview, I couldn't load the symbol table from the host. The bootrom, compiled before the new installation, loaded the image file, started it as usual, then reattached the network interfaces. After that -- nothing. No symbol table, no ethernet activity from the target, only an error after timing out while trying to load the vxWorks.sym file. I checked my usual problem spots, permissions were set correctly, paths and file locations matched. I suspect that there is some sort of network problem between the bsp and the new Windview, but I haven't managed to convince WRS of that. Any help would be appreciated. Thanks, Dave Kuechenmeister - -- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: using multiple routing protocols in SENS Date: Mon, 28 Jun 1999 11:29:59 -0700 From: Stephen Macmanus Organization: Wind River Systems Message-ID: <3777BF27.3488@wrs.com> References: <7l84bs$jmr$1@nnrp1.deja.com> rksripada@hotmail.com wrote: > > The SENS manual says that when multiple routing protocols are used, the > SENS can dynamically specify which route has the precedence. Can someone > elaborate on how SENS can do this dynamically??? See the manual page entry for the routeProtoPrioritySet routine. > > For example, there is a Router using SENS configured to run RIP and > OSPF. RIP learns a route (let's say RIP-ROUTE) for a subnet S1. Also, > the OSPF learns a route (let's say OSPF-ROUTE) for the same subnet S1. > If the RIP-ROUTE and OSPF-ROUTE are different, which route will be given > precedence for writing into the kernel's IP Forwarding Table ??? The routing protocol with the highest priority will add its entries to the IP forwarding table. In the default setting, all routes have an equal priority, so any route to the same destination will always replace an existing route. The priority set routine exists to prevent this thrashing. Stephen - -- - ------------------ Stephen Macmanus #include Software Engineer Wind River Systems --------------------------- Newsgroups: comp.os.vxworks Subject: Network Address Translation and Application Level Gateways Date: Mon, 28 Jun 1999 14:14:18 -0700 From: "Dave Berry" Organization: Share Gate Message-ID: <3777e507.0@news.intercomm.com> I am working on creating NAT and ALG's for a VxWork ethernet (SENS). I have seen this done in a number of different ways (linux at IP, NT in the driver others as applications) and am interested in any opinions about how to perform this effort. I believe the correct placement to be in the IP stack but that does not seem to be possible under VxWorks without getting the source code from VxWorks. Dave. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: MV2604 Universe DMA Date: Mon, 28 Jun 1999 18:09:25 -0400 From: Andrew Reddig Organization: Shore.Net/Eco Software, Inc; (info@shore.net) Message-ID: <3777F295.86613CD8@tekmicro.com> References: <37778BDC.74BB75DD@tekmicro.com> <377793C6.81F8762D@pop.sttl.uswest.net> Followup for any interested readers: Scott Kamara of AP Labs gave me a couple of pointers. After poking around a bit, with nothing working, he suggested that the PWEN (Posted Write Enable) bit might not be set in the Universe VSI1_CTL register. That turned out to be the problem. With PWEN turned on, the target-driven D32 BLT is 29 MB/s and D64 MBLT is 58 MB/s. Many thanks to Scott for his very kind assistance with this problem. - --- Andy R. Scott Kamara wrote: > > The rates we see with the mv2604/UnivII are much better. > You might want to verify via a VME bus analyzer the slave > response time for a given d64 or d32 blt word-cycle > is sufficient to support 30 MB/s. > > Are you using your own dmac driver or Motorola's? > If Motorola's (non-WRS-delivered), it is documented > that theirs is a polled > interface which could degrade the rate you would get > if it was interrupt-on-complete driven. > > The source and destination must be 8-byte aligned. > That is, the lower three bits of the source and dest > address (vme addr and local mem addr) must be the same. > If not aligned right, I dont know what kind of > side-effects this would have. > > Andrew Reddig wrote: > > > > I'm trying to use the Universe DMA controller under VxWorks, and the DMA > > throughput is about 1/2 to 1/3 of what I would expect (7-10 MB/s, even > > using D64 MBLT). The card that we are talking to supports 30 MB/s D32 > > BLT to another VME host, so I'm pretty sure that the Universe is the > > issue. > > > > I suspect an issue with the Universe's priority on the PCI bus or the > > PCI bus > > priority into DRAM, but haven't been able to nail it down yet. > > > > Can anyone provide any hints, pointers, advice, experience? > > > > VxWorks is 5.3.1, mv2604 BSP 1.1/4. > > > > Thanks! > > > > -- > > Andrew Reddig andyr@tekmicro.com > > TEK Microsystems, Incorporated 781.270.0808 > > One North Avenue Fax 781.270.0813 > > Burlington, MA 01803 Web http://www.tekmicro.com > > -- > ----------------------- > Scott Kamara > AP Data Systems > Seattle Engr. Office > (206) 525-8592 > scott@aplabs.com - -- Andrew Reddig andyr@tekmicro.com TEK Microsystems, Incorporated 781.270.0808 One North Avenue Fax 781.270.0813 Burlington, MA 01803 Web http://www.tekmicro.com --------------------------- Newsgroups: comp.os.vxworks Subject: Anyone use Aisys's DriveWay 3DE Date: 28 Jun 1999 22:26:44 GMT From: jgay@drmail.dr.lucent.com (John M. Gay) Organization: Lucent Technologies, Columbus, Ohio Message-ID: <7l8sr4$imr@nntpa.cb.lucent.com> Reply-To: DELETEjgay@lucent.com Well, like the subject asks - I am looking for any opinions or experiences with DriveWay 3DE. My targe is an MPC850. Thanks for the time. - -- john gay. DELETEjgay@lucent.com Life is a near-death experience. -- George Carlin --------------------------- Newsgroups: comp.os.vxworks Subject: R: Tornado on NT vs. Tornado on Solaris? Date: Mon, 28 Jun 1999 23:05:32 +0200 From: "Yellow" Organization: TIN Message-ID: <7l8o4r$av6$1@nslave1.tin.it> References: <7ksvd0$1r8$1@nnrp1.deja.com> <7kvb7j$tt5$1@nnrp1.deja.com> Wind River usually tell you a lot of thing ... beleve 1% Ronny wrote in message 7kvb7j$tt5$1@nnrp1.deja.com... > (this message mailed and posted) > > In article , > bobarmstrong@home.nothiis.net (Robert Armstrong) wrote: > > Well, being a recidivist, I prefer the command line on Solaris for > building. > > I'm a vxWorks user and all this Tornado stuff is very pretty but > nothing to do > > with what I'm trying to deliver. > > Does this mean you don't have a command line access to all the tools > when running under NT? WindRiver told us that development under Solaris > and NT looks the same. > > We like to automate a lot, and good scripting possibilities are a > necessity. Also, we want to use our own Make utility, not that one which > is included with Tornado. Therefore, I would like to be able to write a > Perl or Tcl Script where I could drive every aspect of the development > cycle. Given that Tornado includes Tcl, and that Perl is available for > Solaris and NT alike, is there anything I can not do under NT that I > could do under Solaris? > > Ronald > -- > Ronald Fischer > http://ourworld.compuserve.com/homepages/ronald_fischer/ > > > Sent via Deja.com http://www.deja.com/ > Share what you know. Learn what you don't. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: vxWorks UDP question Date: 25 Jun 1999 19:16:22 GMT From: toni@devsoft.devsoft.com (Anton Hartl) Organization: Customer of SpaceNet GmbH Message-ID: References: <930329254.7537@www.remarq.com> In article <930329254.7537@www.remarq.com>, Robb wrote: >If I am using the sendto function, does it block until the >packet goes out on either the TCP/IP layer (I am using >ethernet), or until there is enough room in vxWorks's >transmit buffer for the data? > >I am asking because I want to write a program that sends >UDP packets out at a very rapid rate. I want to keep >calling the sendto function one after the other, and I want >to make sure that I don't lose data. > >Does vxWorks keep an internal buffer for sending UDP >packets? If it does, and I perform a sendTo call, will the >sendTo call block until there is room in the output buffer >for the next packet? If you want to make sure that you don't lose data, you must not use UDP. UDP does not guarantee delivery of packets by definition; not under VxWorks or any other OS. If you send UDP packets back-to-back you can be almost certain to have packets losses. The OS is allowed to just drop the packets you hand over in case it runs out of buffers for the data or any intermediate network component may decide to drop packets that come in a a very high rate or the receiver cannot keep up with the rate and does not pass all received data to the application ... A 5.2 VxWorks problem (perhaps also later versions) we ran into using high-packet rate UDP was that it made the network stack to stop working completely (both UDP and TCP) for about 30 seconds after which it recovered until the next attempt to send a bunch of packets was made. The only workaround was to add a small delay after each packet send, not so funny ... - -Toni - -- /dev Software Gmbh | Anton Hartl | Dumbo: You're an unusual cop. Gabelsbergerstr. 51 | Phone: +49 (0)89-28659843 | Mickey Mouse: You're an unusal 80333 Muenchen | Fax: +49 (0)89-28659845 | killer, too. Germany | Email: toni@devsoft.com | John Woo, The Killer --------------------------- Newsgroups: comp.os.vxworks Subject: C_MASK16 for 860 Date: Fri, 25 Jun 1999 20:23:41 GMT From: telfer@my-deja.com Organization: Deja.com - Share what you know. Learn what you don't. Message-ID: <7l0og0$em5$1@nnrp1.deja.com> Hello, I am trying to move a fully functional 32 channel QMC driver from SCC2 to SCC3. I have run into a problem that I am beginning to think is a bug on the 860 chip. When I write to the Global Multichannel Parameters, specifically the C_MASK16 parameter I loose functionality on the ethernet setup on SCC1 and the raw data channel SMC1. I am setting the C_MASK16 to 0xF0B8 as stated in the Multichannel Controller manual so that I can have 16bit CRC on the 32 logical channels. Initializing C_MASK16 to anything but zero will cause problems even if the 16bit CRC is not implemented on any of the channels. Has anyone else had similar problems? Thanks, Colin.Telfer@premisys.com Sent via Deja.com http://www.deja.com/ Share what you know. Learn what you don't. --------------------------- Newsgroups: comp.protocols.ppp,comp.os.vxworks Subject: Re: Help: PPP link established, but can't do anything? Date: 27 Jun 1999 08:56:47 -0500 From: kite@NoSpam.%inetport.com (Clifford Kite) Organization: |Chaos| Message-ID: <7l5aiv$5es@inetport.com> References: <37754438.54C5@webguy-prod.com> Followup-To: comp.protocols.ppp,comp.os.vxworks Melinda Stimpson (melinda@webguy-prod.com) wrote: : I am trying to connect Windows NT to a vxWorks system using PPP. : Once I got the NT non-PPP CLIENT/CLIENTSERVER nonsense out of the way, : I was able to make a connection. However, I can't *do* anything with : the connection - I can't telnet, rlogin, or even ping. .. : problem). Note specifically the "Unknown protocol 80fd" - I presume : that is telling me of a problem, but what, exactly? The 80fd refers to Compression Control Protocol (CCP) which "ppp0" either doesn't support or is configured not to support. Neither this or any of the ConfRej's for various LCP options prevented the link negotiations from reaching an apparently normal completion. A silly question: Do you have a host route on each side of the link to the other side through the PPP interface? - -- Clifford Kite Not a guru. (tm) /* The wealth of a nation is created by the productive labor of its * citizens. */ --------------------------- Newsgroups: comp.os.vxworks Subject: HELP : Tornado1.0.1 + SENS1.1 Date: Tue, 29 Jun 1999 16:30:12 +0900 From: "�躹��" Organization: LG InterNet Inc. Channeli Internet Service Message-ID: <7l9sor$dve$1@b5nntp2.channeli.net> Sender: andro@156.147.159.85 Hi~. everybody.. I hace problems using SENS1.1 My target board is Motorola's MPC860. I installed the Tornado1.0.1 and SENS1.1 on my Windows Host first, and developed my BSP. If I don't install SENS and Windview , I can compile the bootrom.hex. It works well on the target board. But, If I install SENS and recompile the bootrom.hex it dosn't work. I developed my BSP Code based on ADS860 BSP Rev '/2', and I think the bootrom.hex should run whether I install SENS or not. Following file is the setup.log file when I installed SENS. Please give me the correct installation guide or something else. Thank you in advance. === setup.log file === 29-Jun-99.10:05 TDK-12221-ZC-09 SETUP-1.5 29-Jun-99.10:05 100-8587-30 Tornado Cumulative Patch Rev 1: ppc 29-Jun-99.10:06 100-6901-30 Tornado Object: x86-win32 x ppc 29-Jun-99.10:11 TDK-12221-ZC-09 SETUP-1.5 29-Jun-99.10:11 100-6902-30 VxWorks BSP: ads860 29-Jun-99.10:12 29-Jun-99.10:12 SETUP has overwriten the following file(s). Which seems to be 29-Jun-99.10:12 newer than the one on the CDROM. Original file(s) is zipped 29-Jun-99.10:12 into C:\Tornado/.wind/uninst/data.002. 29-Jun-99.10:12 Use any unzip utility in case you need to get back the original 29-Jun-99.10:12 version of the file(s). 29-Jun-99.10:12 29-Jun-99.10:12 target/lib/objPPC403gnuvx/if_sl.o 29-Jun-99.10:12 target/lib/objPPC603gnuvx/if_sl.o 29-Jun-99.10:12 target/lib/objPPC604gnuvx/if_sl.o 29-Jun-99.10:12 target/lib/objPPC860gnuvx/if_sl.o 29-Jun-99.10:17 TDK-12221-ZC-09 SETUP-1.5 29-Jun-99.10:17 100-10532-30 WindNet SENS: ppc 29-Jun-99.10:20 29-Jun-99.10:20 SETUP has overwriten the following file(s). Which seems to be 29-Jun-99.10:20 newer than the one on the CDROM. Original file(s) is zipped 29-Jun-99.10:20 into C:\Tornado/.wind/uninst/data.003. 29-Jun-99.10:20 Use any unzip utility in case you need to get back the original 29-Jun-99.10:20 version of the file(s). 29-Jun-99.10:20 29-Jun-99.10:20 target/lib/objPPC403gnuvx/if_eihk.o 29-Jun-99.10:20 target/lib/objPPC403gnuvx/if_eitp.o 29-Jun-99.10:20 target/lib/objPPC603gnuvx/if_eihk.o 29-Jun-99.10:20 target/lib/objPPC603gnuvx/if_eitp.o 29-Jun-99.10:20 target/lib/objPPC604gnuvx/if_eihk.o 29-Jun-99.10:20 target/lib/objPPC604gnuvx/if_eitp.o 29-Jun-99.10:20 target/lib/objPPC860gnuvx/if_eihk.o 29-Jun-99.10:20 target/lib/objPPC860gnuvx/if_eitp.o 29-Jun-99.10:20 target/unsupported/bootp2.1/Makefile 29-Jun-99.10:20 target/unsupported/bootp2.1/bootpd.c --------------------------- Newsgroups: comp.os.vxworks Subject: Unaligned access fault Date: Tue, 29 Jun 1999 10:13:37 +0200 From: Robertas Simkevicius Organization: Omnitel, Lithuania Message-ID: <37788031.A5877402@fmst.vtu.lt> hi, I'm not very familar with VxWorks, but would like to know about error handling in it. I'm using VxWorks v3.2 on Alpha CPU, for some communication tasks. And sometimes I'm getting such error: unaligned access fault Fault VA = 7fffffff7fffffff Opcode = 29 Register = 10 Exception PC = fffffc0001bf92f8 Exception Stack Frame Pointer = fffffc0001c6e978 Exception Stack Pointer = fffffc0001c6ea80 Task: 0x1c6f740 Maybe there is some way to find out in which place it occurs, or why it occurs. Ther must be some line debug possibilities, or maybe someone can explain what VA, Opcode, PC and etc. means. Thank you in advance Babas --------------------------- Newsgroups: comp.os.vxworks Subject: Re: MV2604 Universe DMA Date: Tue, 29 Jun 1999 09:37:27 +0100 From: Dave Machin Organization: Marconi Electronic Systems Message-ID: <377885C7.A04A331A@REMOVE-THIS.gecm.com> References: <37778BDC.74BB75DD@tekmicro.com> One thing to look at would be the disconnect size and bus ownership timeouts - also check that the FIFOs are enabled as this can have a huge impact on performance. You might find you need to tweak various block size and ownership timer parameters on both the Universe and any bridge chips in the path, in order to get the best performance - it's possible that you're having to re-negotiate for one (or all) of the busses for every few words transferred, and this can do very bad things to your performance! Finally, check that the area mapped to the VME master image is set as uncached. In practise, I've found it quite difficult to get close to the transfer rates mentioned by the sales brochures - but there are many things you can do to improve on the situation. Hope some of this helps. Cheers Dave Machin Andrew Reddig wrote: > > I'm trying to use the Universe DMA controller under VxWorks, and the DMA > throughput is about 1/2 to 1/3 of what I would expect (7-10 MB/s, even > using D64 MBLT). The card that we are talking to supports 30 MB/s D32 > BLT to another VME host, so I'm pretty sure that the Universe is the > issue. > > I suspect an issue with the Universe's priority on the PCI bus or the > PCI bus > priority into DRAM, but haven't been able to nail it down yet. > > Can anyone provide any hints, pointers, advice, experience? > > VxWorks is 5.3.1, mv2604 BSP 1.1/4. > > Thanks! > > -- > Andrew Reddig andyr@tekmicro.com > TEK Microsystems, Incorporated 781.270.0808 > One North Avenue Fax 781.270.0813 > Burlington, MA 01803 Web http://www.tekmicro.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Network Address Translation and Application Level Gateways Date: Tue, 29 Jun 1999 10:19:01 +0100 From: "Mark Hambleton" Organization: (Posted via) Netcom Internet Ltd. Message-ID: <7la2rt$si4$1@taliesin.netcom.net.uk> References: <3777e507.0@news.intercomm.com> I have done a similar think for SENS and unfortunately I had to do it in the driver. You may be able to do is as an output protocol, I had already used the single output protocol slot though which is why it ended up in the driver. - -- Mark Hambleton http://www.xstreamis.com/ Dave Berry wrote in message <3777e507.0@news.intercomm.com>... >I am working on creating NAT and ALG's for a VxWork ethernet (SENS). I have >seen this done in a number of different ways (linux at IP, NT in the driver >others as applications) and am interested in any opinions about how to >perform this effort. I believe the correct placement to be in the IP stack >but that does not seem to be possible under VxWorks without getting the >source code from VxWorks. > >Dave. > > --------------------------- Newsgroups: comp.os.vxworks Subject: Re: HELP : Tornado1.0.1 + SENS1.1 Date: Tue, 29 Jun 1999 10:20:38 +0100 From: Simon Roberts Message-ID: <37788FE6.F6FCA6A3@roke.co.uk> References: <7l9sor$dve$1@b5nntp2.channeli.net> Andro, We had, what sounds like, similar problems installing SENS (and SNMP as it happens) with our BSP. We discovered that it was essential to install the components in the following order: 1. Install Tornado *without* SENS and SNMP and Windview (your components may be slightly different, so ignore if n/a) 2. Install your BSP by dragging and dropping the BSP release on the ./target directory (at least this is what we do with out pre release MV2400 BSP) 3. Install the remaining components of Tornado missed out in step 1 This should ensure that you have the right copies of files included for SENS (rather than BSP copies which would otherwise have overwritten them). Regards, Simon. "�躹��" wrote: > Hi~. everybody.. > > I hace problems using SENS1.1 > > My target board is Motorola's MPC860. > > I installed the Tornado1.0.1 and SENS1.1 on my Windows Host first, > and developed my BSP. > > If I don't install SENS and Windview , > I can compile the bootrom.hex. It works well on the target board. > But, If I install SENS and recompile the bootrom.hex > it dosn't work. > > I developed my BSP Code based on ADS860 BSP Rev '/2', > and I think the bootrom.hex should run whether I install > SENS or not. > > Following file is the setup.log file when I installed SENS. > > Please give me the correct installation guide or something else. > > Thank you in advance. > > === setup.log file === > > 29-Jun-99.10:05 TDK-12221-ZC-09 SETUP-1.5 > 29-Jun-99.10:05 100-8587-30 Tornado Cumulative Patch Rev 1: ppc > 29-Jun-99.10:06 100-6901-30 Tornado Object: x86-win32 x ppc > > 29-Jun-99.10:11 TDK-12221-ZC-09 SETUP-1.5 > 29-Jun-99.10:11 100-6902-30 VxWorks BSP: ads860 > 29-Jun-99.10:12 > 29-Jun-99.10:12 SETUP has overwriten the following file(s). Which seems to > be > 29-Jun-99.10:12 newer than the one on the CDROM. Original file(s) is zipped > 29-Jun-99.10:12 into C:\Tornado/.wind/uninst/data.002. > 29-Jun-99.10:12 Use any unzip utility in case you need to get back the > original > 29-Jun-99.10:12 version of the file(s). > 29-Jun-99.10:12 > 29-Jun-99.10:12 target/lib/objPPC403gnuvx/if_sl.o > 29-Jun-99.10:12 target/lib/objPPC603gnuvx/if_sl.o > 29-Jun-99.10:12 target/lib/objPPC604gnuvx/if_sl.o > 29-Jun-99.10:12 target/lib/objPPC860gnuvx/if_sl.o > > 29-Jun-99.10:17 TDK-12221-ZC-09 SETUP-1.5 > 29-Jun-99.10:17 100-10532-30 WindNet SENS: ppc > 29-Jun-99.10:20 > 29-Jun-99.10:20 SETUP has overwriten the following file(s). Which seems to > be > 29-Jun-99.10:20 newer than the one on the CDROM. Original file(s) is zipped > 29-Jun-99.10:20 into C:\Tornado/.wind/uninst/data.003. > 29-Jun-99.10:20 Use any unzip utility in case you need to get back the > original > 29-Jun-99.10:20 version of the file(s). > 29-Jun-99.10:20 > 29-Jun-99.10:20 target/lib/objPPC403gnuvx/if_eihk.o > 29-Jun-99.10:20 target/lib/objPPC403gnuvx/if_eitp.o > 29-Jun-99.10:20 target/lib/objPPC603gnuvx/if_eihk.o > 29-Jun-99.10:20 target/lib/objPPC603gnuvx/if_eitp.o > 29-Jun-99.10:20 target/lib/objPPC604gnuvx/if_eihk.o > 29-Jun-99.10:20 target/lib/objPPC604gnuvx/if_eitp.o > 29-Jun-99.10:20 target/lib/objPPC860gnuvx/if_eihk.o > 29-Jun-99.10:20 target/lib/objPPC860gnuvx/if_eitp.o > 29-Jun-99.10:20 target/unsupported/bootp2.1/Makefile > 29-Jun-99.10:20 target/unsupported/bootp2.1/bootpd.c --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Unaligned access fault Date: Tue, 29 Jun 1999 12:13:14 +0200 From: "Fred Zwarts" Organization: Kernfysisch Versneller Instituut Message-ID: <7la67q$ame$1@info.service.rug.nl> References: <37788031.A5877402@fmst.vtu.lt> Robertas Simkevicius wrote in message = news:37788031.A5877402@fmst.vtu.lt... > hi, >=20 >=20 > I'm not very familar with VxWorks, but would like to know about error > handling in it. I'm using VxWorks v3.2 on Alpha CPU, for some > communication tasks. And sometimes I'm getting such error: >=20 > unaligned access fault > Fault VA =3D 7fffffff7fffffff > Opcode =3D 29 > Register =3D 10 > Exception PC =3D fffffc0001bf92f8 > Exception Stack Frame Pointer =3D fffffc0001c6e978 > Exception Stack Pointer =3D fffffc0001c6ea80 > Task: 0x1c6f740 >=20 > Maybe there is some way to find out in which place it occurs, or why = it > occurs. Ther must be some line debug possibilities,=20 VxWorks has Debugging facilities (see the manuals). > or maybe someone can > explain what VA, Opcode, PC and etc. means. VA: Virtual Address Opcode: Operation Code (Instruction being executed) PC: Program Counter=20 (Virtual Address of current/next instruction) Stack Frame Pointer: Virtual Address of Frame=20 (Each subroutine call starts a new frame on the = stack) Stack Pointer: Virtual Address of the top of the stack. >=20 > Thank you in advance >=20 > Babas --------------------------- End of New-News digest ********************** From owner-vxwexplo-process Tue Jun 29 08:17:25 1999 From: Sammer Ostogratzk Date: Tue Jun 29 08:17:28 PDT 1999 I posted this before but did not get an answer, so I'll try again. Hi all vxworks gurus: I noticed that the Tornado documentation describes a zbuf library (zbufLib), however, it looks like networking code uses mbufs (mblk?) instead. I am a bit confused on what we are really supposed to use if we want to use buffer chains. Also, I can't find a description of mbufs, mblks in the documentation. has anyone used the zbufLib? Is it ok to use it or is it an antiquated library that will disappear in newer releases of Tornado? Thanks. ______________________________________________________ Get Your Private, Free Email at http://www.hotmail.com From owner-vxwexplo-process Tue Jun 29 08:46:04 1999 From: Smith Date: Tue Jun 29 08:46:07 PDT 1999 Subject: inter-processor comm We are considering using a 2-processor solution for the project on which I am working. The main reason is to off-load the comm functions (including MAC and TCP/IP) to a dedicated processor, freeing the other processor for functions. I am looking for possible solutions for the inter-processor comm. and have some general. Any insight would be greatly appreciated. -What type of Multi-Processor support does vxWorks have? -What type of functionality does smlib give? -Is there such a thing as a CORBA ORB for shared memory? Thanks, Brett Smith brett.smith@bktech.com vxWorks From owner-vxwexplo-process Tue Jun 29 09:02:10 1999 From: "Paul Travers" Date: Tue Jun 29 09:02:13 PDT 1999 Subject: Tornado 2 C++ Compiler Problem Hi, I have just installed Tornado (2) and have had problems compiling C++. The install is for Windows NT and building for a PowerPC target. The application c1plus.exe causes an exception followed by an application error in as.exe (Dr Watson claims a stack overflow for the exception). This module did not present problems with the previous release of Tornado. Interestingly the module instantiates template objects. Has anyone had similar problems or have any information relating to this. Regards, Paul Travers pault@epid.eurotherm.co.uk From owner-vxwexplo-process Tue Jun 29 10:35:51 1999 From: mdelgior@TIGER.rst.com (Mark Del Giorno) Date: Tue Jun 29 10:35:54 PDT 1999 Subject: -mlongcall doesn't work... Heyyyyyyyyyyyyyyyyy, We cannot get the following simple code (ttt.cc) to load on a 64MB 2700 board. We get the infamous: Undefined symbol: __builtin_new (binding 1 type 0) Undefined symbol: __builtin_delete (binding 1 type 0) (I thought this was fixed with the latest version of the compiler......?) We're using Tornado 2. Does anyone know a recommended work-around? Thanks in advance. - Mark ========= start of ttt.cc #include "vxWorks.h" class MARK_CLASS { public: MARK_CLASS() {}; }; void ttt(int n) { MARK_CLASS *m; m = new MARK_CLASS(); } ============== end of ttt.cc ********* Compilation method: /usr/wind2/host/sun4-solaris2/bin/ccppc -I/usr/wind2/target/h \ -Wall -DCPU=PPC604 -O -pipe -nostdinc \ -ansi -fvolatile -fno-builtin -mcpu=604 -mlongcall -c ttt.cc ******** Compiler version: # ccppc -v Reading specs from /usr/wind2/host/sun4-solaris2/lib/gcc-lib/powerpc-wrs-vxworks /cygnus-2.7.2-960126/specs gcc driver version cygnus-2.7.2-960126 egcs-971225 tornado 2.0 executing gcc ver sion cygnus-2.7.2-960126 From owner-vxwexplo-process Tue Jun 29 14:48:51 1999 From: Keith Potter Date: Tue Jun 29 14:48:54 PDT 1999 Subject: European Wind River Newsletter This may be of interest to those in Europe: Currents Europe, is a monthly e-mail newsletter covering the latest Tornado news, Wind River partner news, technical seminars, training and other events in Europe. To subscribe to CURRENTS Europe put SUBSCRIBE in the BODY of an e-mail to: currents-europe@oaknetpub.com Or see web page Thanks, Keith Potter European Marketing Manager Wind River Systems From owner-vxwexplo-process Tue Jun 29 21:06:26 1999 From: Randy Riggins Date: Tue Jun 29 21:06:30 PDT 1999 Subject: RE: inter-processor comm This message is in MIME format. Since your mail reader does not understand this format, some or all of this message may not be legible. ------ =_NextPart_001_01BEC2AD.64CCADBE Content-Type: text/plain Brett, VxWorks supports a backplane network which allows you to run TCP/IP over the backplane between the two processors. This allows one processor to be the gateway for the other. We currently do this. We use an RPC package from NobleNet to perform inter-processor communications. VxWorks also has an optional package called Shared Memory Objects, this give you semaphores, message queues and other facilities between multiple processors. randy > -----Original Message----- > From: vxwexplo@lbl.gov [SMTP:vxwexplo@lbl.gov] On Behalf Of Smith > Posted At: Tuesday, June 29, 1999 8:48 AM > Posted To: vxworks > Conversation: inter-processor comm > Subject: inter-processor comm > > We are considering using a 2-processor solution for the project on > which > I am working. The main reason is to off-load the comm functions > (including MAC and TCP/IP) to a dedicated processor, freeing the other > processor for functions. > > I am looking for possible solutions for the inter-processor comm. and > have some general. Any insight would be greatly appreciated. > > -What type of Multi-Processor support does vxWorks have? > > -What type of functionality does smlib give? > > -Is there such a thing as a CORBA ORB for shared memory? > > Thanks, > Brett Smith > brett.smith@bktech.com > > vxWorks ------ =_NextPart_001_01BEC2AD.64CCADBE Content-Type: text/html Content-Transfer-Encoding: quoted-printable RE: inter-processor comm

Brett,

VxWorks supports a = backplane network which allows you to run TCP/IP over the backplane = between the two processors. This allows one processor to be the gateway = for the other. We currently do this. We use an RPC package from = NobleNet to perform inter-processor communications.

VxWorks also has an = optional package called Shared Memory Objects, this give you = semaphores, message queues and other facilities between multiple = processors.

randy

-----Original Message-----
From:   vxwexplo@lbl.gov = [SMTP:vxwexplo@lbl.gov] On = Behalf Of Smith
Posted = At:      Tuesday, June 29, 1999 8:48 AM
Posted = To:      vxworks
Conversation:   inter-processor comm
Subject:       = inter-processor comm

We are considering = using a 2-processor solution for the project on which
I am working. = The main reason is to off-load the comm functions
(including MAC and = TCP/IP) to a dedicated processor, freeing the other
processor for = functions.

I am looking for = possible solutions for the inter-processor comm. and
have some = general. Any insight would be greatly appreciated.

-What type of = Multi-Processor support does vxWorks have?

-What type of = functionality does smlib give?

-Is there such a = thing as a CORBA ORB for shared memory?

Thanks,
Brett Smith
brett.smith@bktech.com

vxWorks

------ =_NextPart_001_01BEC2AD.64CCADBE-- From owner-vxwexplo-process Wed Jun 30 03:50:34 1999 From: "Tyler, Paul A (DBY)" Date: Wed Jun 30 03:50:38 PDT 1999 Subject: Network interface stops transmitting. Hello, I have a VMIC-7591 (VME PC) running VxWorks 5.3.1. I have two DEC 21143 network interfaces, one on board and one on the PMC site. The problem I see is that the on board interface (dc0), which is used as the boot device, stops transmitting. This happens during the startup of the applications by the start script, about 50% of the time. The application sees NFS errors when trying to open a configuration file. The ifShow() command shows an increasing number of transmit errors. The interface does not recover, and a reboot is required. The failure appears to be related to a health check application which runs on the host PC. If I stop that application the reboot works fine. I am not running SENS. Has anyone seen this type of failure before? Is there a way of monitoring the transmit errors and then re-launching the network interface? Any help would be very much appreciated. Paul Tyler Measurement Systems Specialist Electronics & Measurement Technology Dept. Rolls-Royce plc. Tel. +44 (1332) 247808 Fax. +44 (1332) 247928 From owner-vxwexplo-process Wed Jun 30 04:00:16 1999 From: daemon@csg.lbl.gov Date: Wed Jun 30 04:00:20 PDT 1999 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Wed Jun 30 04:00:12 PDT 1999 Subject: Unfamiliar vxWorks Error Subject: Distribution in Swiss Subject: Re: Booting VxWorks from PCMCIA/Flashdisk, nearly there but ... Subject: fcntl() missing Subject: Re: Tornado 2 and Dependencies Subject: Re: [Q] which realtime DBs are there? Subject: Re: a problem while linking Subject: Re: T2 POSIX timer broken for VxSim? Subject: Re: dosFs woes Subject: Save 40-60% on Cellular & Digital Phone Accessories Subject: viewing dynamically loaded modules Subject: Re: UDP performance over PPP Subject: Re: inter-processor comm Subject: Re: Unaligned access fault Subject: Re: Sample VxWorks driver for raw device? Subject: PPP NT to vxWorks, can't detect disconnect Subject: Re: Help: PPP link established, but can't do anything? Subject: using multiple routing protocols in SENS ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Unfamiliar vxWorks Error Date: Tue, 29 Jun 1999 10:11:15 -0400 From: "David R. Kuechenmeister" Organization: Georgia Institute of Technology, Atlanta GA, USA Message-ID: <290619991011154328%David.Kuechenmeister@gtri.gatech.edu> We have been having some uncommanded reboots of our MV2604 processor, preceeded by the following errors IP CLK RLS: BAD LOCK READ! IP CLK GET: BAD LOCK READ! These are repeated several times before the reboot. I've checked through our wrs directories and can't find the strings, plus I don't print any such message in the application code. Could anyone give me the meaning of the message and maybe a way to track down these errors in the future? Thanks Dave Kuechenmeister - -- David R. Kuechenmeister Georgia Tech Research Institute Atlanta,GA 30332-0853 mailto:David.Kuechenmeister@gtri.gatech.edu Voice: (770)528-7738 --------------------------- Newsgroups: comp.os.vxworks Subject: Distribution in Swiss Date: Tue, 29 Jun 1999 16:41:24 +0200 From: "Patrick FAFIN" Organization: i-Logix France Message-ID: <3778cc83@news.uk.ibm.net> Reply-To: "Patrick FAFIN" I-Logix is looking for an agent for the Swiss territory. I-Logix is developping Rhapsody, UML development tool for embedded code generation, and Statemate, specification validation tool for complex systems. Please contact me at patrick@ilogix.com For more information : www.ilogix.com Thanks - -- Patrick FAFIN Sales Director Southern Europe I-Logix 4 allee de Londres 91969 - Les Ulis Cedex - France Tel : + 33 1 60 92 17 17 Fax: + 33 1 60 92 17 18 Email: patrick@ilogix.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Booting VxWorks from PCMCIA/Flashdisk, nearly there but ... Date: Tue, 29 Jun 1999 15:01:34 +0100 From: Simon Roberts Message-ID: <3778D1BE.62BF4ACE@roke.co.uk> References: <7l8015$fdb29@overload.lbl.gov> Thanks Tim, your suggestion has helped alot. It didn't help that host and target accessed two different symbol files! Simon. tim allen wrote: > >... network related commands still don't work as well as they do had I > >booted the card conventionally from a server over the ethernet network. > >But at least now I can use a host target shell in Tornado. > > It may be that your host can't find the VxWorks boot image and > symbol table. > > When the Tornado host shell attaches to the target, by default it tries to > locate the boot image by using the boot file pathname from the boot > parameters on the target. The host has a different view of the local file > system and so the boot image and symbols are not found when you boot, for > example, from pcmcia. You can work around this by forcing the path in the > Tornado Tools->Target Server->Configure option. Select Target Server > Properties->Core File and Symbols, where you can give the path to the boot > image stored on your host. > > When you boot from the host, over the network, this issue does not arise. > > --------------------- > Tim Allen Ltd > Real Time Embedded Software Consultancy > > ______________________________________________________ > Get Your Private, Free Email at http://www.hotmail.com --------------------------- Newsgroups: comp.os.vxworks Subject: fcntl() missing Date: Tue, 29 Jun 1999 18:04:54 +0200 From: Ralf Gerlich Organization: BSSE Message-ID: <3778EEA6.FD2DF856@t-online.de> Hi! I'm missing the fcntl() syscall. I need to change some flags for a file using F_SETFL. Is there an fcntl() call or any other possibility to do such a thing? I posted a similar question to this newsgroup some time before but I got no answer. Thanx in advance, Ralf --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Tornado 2 and Dependencies Date: 29 Jun 1999 16:44:08 GMT From: Johan Borkhuis Organization: XS4ALL Internet BV Message-ID: <7lat4o$8s8$1@news1.xs4all.nl> References: <376FFA62.F9A570C9@isptechinc.com> Matthew Donadio wrote: : Hello all, : I am having a problem porting a library to VxWorks / Tornado 2. : The library uses the convention that header files are included using <> : and then assumes that -I. is a command line option to the compiler. : This is giving the Tornado 2 project tool heartburn, though. : When I generate dependencies for the library, the project tool thinks : that all of the #include files are external to the library and : creates a makefile dependency like : blah.o: $(PRJ_DIR)/blah.c blah.h : instead of : blah.o: $(PRJ_DIR)/blah.c $(PRJ_DIR)/blah.h : Building the library crashes because the make gets called from a : subdirectory, and it can not find the header files, and assumes that : they need to be created. : I could change all of the #include lines to use quotes instead of <>, : but there are 150+ source files, and I don't want to do it every time : the library gets released. I would also like to keep the dependencies : because full recompiles would take way too long. : Any ideas on how I can get around this problem? Thanks. Try the following: Add the following lines to youw WPJ file: BUILD_SIMNTgnu_MACRO_VPATH $(PRJ_DIR)/include This will add the macro VPATH to your project (in this case the simulator). This macro is also exported to your makefile. See the make manual for more information on the VPATH variable (it is the search path for dependencies). See also the other message in this thread with the message from WRS. Johan Borkhuis. - -- J.A. Borkhuis, Product Manager Public Transport Systems Peek Traffic B.V, Basicweg 16 P.O. Box 2542, 3800 GB Amersfoort NL Tel.: (+31)-33-454 1970, Fax: (+31)-33-454 1973 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: [Q] which realtime DBs are there? Date: Mon, 28 Jun 1999 05:34:43 GMT From: yuval_lorian Organization: Deja.com - Share what you know. Learn what you don't. Message-ID: <7l71hj$7t3$1@nnrp1.deja.com> References: <7l6mqp$k7h$1@news.etri.re.kr> You may try the "Raima Data Manager" (from Raima corporation). It works with VxWorks. Yuval. In article <7l6mqp$k7h$1@news.etri.re.kr>, "��(Jerry)" wrote: > Which realtime DBs are there based on VxWorks? > Could you recommend one? > Bye. > > -- > *************************************************** > Jehun Rhee (Jerry) > > Mobile Traffic Management Team, > Radio & Broadcasting Technology Lab., ETRI > 161 Kajong Dong, Yusong Gu, Taejon, 305-350, Korea > > Voice Phone @ +82-42-860-3820 > Fax @ +82-42-860-6403 > Home Phone @ +82-42-485-8809 > Mobile Phone @ +82-17-248-0116 > Email @ jhrhee@etri.re.kr > ICQ @ 30447192 > SoftMessenger @ jhrhee (only Korean) > > Sent via Deja.com http://www.deja.com/ Share what you know. Learn what you don't. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: a problem while linking Date: Wed, 23 Jun 1999 09:59:52 -0700 From: Graham Waters Organization: TRIUMF Message-ID: <37711288.4C78F671@triumf.ca> References: <377056b3.0@dnews.sbell.com.cn> Reply-To: waters@triumf.ca liuf wrote: > > hi, > > I just begin to use vxworks. I use NT as host and use a PC as target. > When I link my test code, as > >cc386 -o testExe -nostdinc -fno-builtin -v -LC:\x86\target\lib > testCode.o -lI80486gnuvx > it shows an error: > cannot open crt0.o: No such file or directory > > Can anyone tell me what ctr0.o is? why cc386 automatically link it with my > code? where can I find this? > > Thanks! The vxWorks kernel include all run-time libraries and uses dynamic loading and unloading of object modules with run-time relocation and linking. It is only necessary to compile a program and load it. Also do not use the statement "main()", use some function name, and call that function by name from the the vxWorks shell to run it. Your compiler line should look something like this: cc386 (your compile flags) -c testExec.c The -c option means compile but do not link. Regards Graham. +---------------------------------------------------------------------+ |Graham Waters, Control System Eng Internet: waters@triumf.ca| |TRIUMF --- University of British Columbia, phone: (604)-222-1047 | |Vancouver, BC, Canada Fax: (604)-222-7307 | |http://edevel9pc.triumf.ca/homepage/graham01.html | +---------------------------------------------------------------------+ For every expert there is an equal and opposite expert - Arthur C Clark --------------------------- Newsgroups: comp.os.vxworks Subject: Re: T2 POSIX timer broken for VxSim? Date: Tue, 29 Jun 1999 20:07:42 GMT From: Hai Vu Message-ID: <37792792.B2F42C6A@netscape.net> References: <3779102E.2BD8E19A@netscape.net> To answer my own posting, here is an excerpt of a response form WRS: | SPR# 26584 | ............................................................................ | .... | | DATE: Apr 12 1999 | TITLE: posix timer tests cause general protection faults on SIMNT | | | Problem_Description | | Running the posix timer tests (target/src/test/vxwRegresssion/posix/ | pTimerTest.c, | | run testTimers function to execute) causes general protection faults | on the NT simulator similar to below: | | Exception number 0: Task: 0x4bccd10 (t4) | | General Protection Fault | Program Counter: 0x0042ae6d | Status Register: 0x00010202 | | value = 0 = 0x0 | | test3, test4, test6, test7, test8 (see pTimerTest.c) seem to cause the | general protection fault. | | 10-Apr-99.18:03 CD manufacturing time: Tue Apr 06 04:52:38 PDT 1999 | 10-Apr-99.18:03 TDK-12840-ZC-00 SETUP-2.0 f:\tornado2.ct.fullsim.qe.apr5 | 10-Apr-99.18:03 Microsoft Windows NT version 4.0 (Build 1381) Service Pack 3 | 10-Apr-99.18:03 100-11103-30 CodeTEST/Coverage: x86-winNT x simnt | 10-Apr-99.18:04 100-11129-30 CodeTEST/Memory: x86-winNT x simnt | 10-Apr-99.18:04 100-9878-30 Driver Objects/Headers: simnt | 10-Apr-99.18:04 100-9884-30 Tornado Simulator: x86-winNT | 10-Apr-99.18:06 100-9881-30 Wind Foundation Class: simnt | | | Please include the TSR number in replies that relate to this issue. | Thank you. | - ---------------------------------------------------------------------------- | --- | Hai Vu wrote: > I have been experimenting with the new Tornado 2 (PPC target)'s VxSim > for NT and encountered some problems with the POSIX timer > implementation. VxWorks reported GENERAL PROTECTION FAULT eveytime a > POSIX timer fires. When I ran the debugger, it pointed to sigqueue() as > the last command being executed. Included below is the test code that > generates the problem. If anyone out there has the same problem or know > of a solution please let me know. > > #include > #include > #include > #include > > static void _SigHandler (int nSigno, siginfo_t* pSiginfo, void* pToken); > > int TestTimer () > { > //## begin Tester::TestTimer%929655315.body preserve=yes > struct sigevent sigEvent; > timer_t tid; > struct sigaction sa; > itimerspec timerSpec; > timespec currtime = { 0, 0 }; > > // The manual recommended this prior to calling timer functions > clock_settime( CLOCK_REALTIME, &currtime ); > > // Setup timer signal (SIGALRM) > sigEvent. sigev_signo = SIGALRM; > sigEvent. sigev_value. sival_int = 666; > sigEvent. sigev_notify = SIGEV_NONE; > > // Create the timer > timer_create( CLOCK_REALTIME, &sigEvent, &tid ); > > // Setup the signal handler to handle SIGALRM > sa. sa_sigaction = &_SigHandler; > sigemptyset( &( sa. sa_mask )); > sigaddset( &( sa. sa_mask ), SIGALRM ); > sa. sa_flags = SA_SIGINFO; > > sigaction( SIGALRM, &sa, 0 ); > > // Setup timeout pertiod for the timer > timespec timeout = { 2, 0 }; > timespec ZERO_TIME = { 0, 0 }; > > timerSpec. it_interval = ZERO_TIME; // No repeat > timerSpec. it_value = timeout; > > // Arm the timer > timer_settime( tid, 0, &timerSpec, 0 ); > > // Wait for timer to fire > timespec sleeptime = { 4, 0 }; > nanosleep( &sleeptime, 0 ); > > return 0; > } > > void _SigHandler (int nSigno, siginfo_t* pSiginfo, void* pToken) > { > > cout << "_SigHandlerr() - signo = " << nSigno << ", token = " << pToken > << endl; > > } --------------------------- Newsgroups: comp.os.vxworks Subject: Re: dosFs woes Date: Tue, 29 Jun 1999 20:31:50 GMT From: ckulach@my-deja.com Organization: Deja.com - Share what you know. Learn what you don't. Message-ID: <7lbafm$qnk$1@nnrp1.deja.com> References: <374E8F4B.5EEC0DDD@syncro-tech.com> <929571122.5827@www.remarq.com> <7khmqk$gau$1@bgtnsc03.worldnet.att.net> <37735619.53776F72@roke.co.uk> ioctl(FIOFLUSH) seems to work fine here. Our system uses PCMCIA ATA cards, and is often subject to unexpected power loss and card removal - and the bit of code below seems to do the trick. //flush the file system iFd = open("a:",O_RDWR,0); ioctl(iFd,FIOFLUSH,0); close(iFd); Chris In article <37735619.53776F72@roke.co.uk>, Simon Roberts wrote: > According to the manual fflush() is not implemented in VxWorks (see below). > Closing the file seems to be the only way of forcing a flush, then if necessary > reopening for append. > > Simon. > > NAME > > fflush( ) - flush a stream (ANSI) > > SYNOPSIS > > int fflush > ( > FILE * fp /* stream to flush */ > ) > > DESCRIPTION > > This routine writes to the file any unwritten data for a specified output or > update stream for which the most recent operation was not input; for an input > stream the behavior is undefined. > > CAVEAT > > ANSI specifies that if fp is a null pointer, fflush( ) performs the flushing > action on all streams for which the behavior is defined; however, this is not > implemented in VxWorks. > > Cameron Fowler wrote: > > > In order to flush your data down to disk, you will need to do > > a > > fflush() > > followed by > > ioctly(FIOSYNC) > > This should push your data down to the disk > > Alan Yakiwchuk wrote in message <929571122.5827@www.remarq.com>... > > >I think the information (directories, etc.) is only updated > > >when you close the file. > > > > > >Have you tried flushing the output buffers to force and > > >update? i.e. FIOFLUSH Let me know if this works 'cause > > >I may be facing a similar problem soon... > > > > > > > > > > > >**** Posted from RemarQ - http://www.remarq.com - Discussions Start Here > > (tm) **** > > Sent via Deja.com http://www.deja.com/ Share what you know. Learn what you don't. --------------------------- Newsgroups: comp.os.vxworks Subject: Save 40-60% on Cellular & Digital Phone Accessories Date: Tue, 29 Jun 1999 16:00:50 PDT From: icwpcs@aol.com Organization: ICW Message-ID: <7lbc55$96c$10388@mercury.hiline.net> Need a new battery, car charger or leather protective carrying case for your cellular or digital phone? We offer package prices on acessories for the most popular phones of today as well as older models at 40-60% savings over retail stores. Don't pay $30 for a car charger, $70 for a battery or $20 for a leather case. Our package prices for battery, car charger and leather case start at $29.99. EXAMPLE OF COMPLETE STARTER KIT W/BATTERY, CAR CHARGER AND CASE: Qualcomm 1920/2700 , Sale Price $29.99, Regularly $79.99 Nokai 5000/6000 series, Sale Price $54.99, Regularly $99.99 with vib battery Motorola 550/660 Flip phone, Sale Price $29.99, Regularly $89.99 Just need one item, not the entire package? Most leather case are just $9.99 and vehicle chargers are only $14.99 on most models. Battery's start at $24.99. Hard to find item? We can special order your accessories for older models at no additional cost. PARTIAL LIST OF PHONE WE HAVE ACCESSORIES FOR: Sprint QUALCOMM 1920/2700 SprintSAMSUNG 1000/1900/2000 SONY 1201,1301,CMZ-100,CM-777 MOTOROLA MICRO TAC 550-650 ,TELETAC, STAR TAC, MICRO TAC ELITE, BAG PHONES, FLIP STYLE NOKAI 5000-6000 SERIES, 100, 232-252 , 1000 , 638-636, 918, 2160-2170-2190 NEC 700/800 ERICCISON - MOST MODELS AUDIOVOX -MOST MODELS SANYO 3000 CREDIT CARD ORDERS ONLY, NO COD'S, MUST HAVE EXACT MODEL NUMBER 1-800-686-8988 Monday-Friday 9am-6pm, Saturday 12n-5pm Shipping and handling charge of $7.99 for starter kits and $3.99 for individual items. Most items shipped in 5 business days. This offer is brought to you by ICW, a large wireless dealer in the Southwestern US. If you no longer wish to receive e-mail please notify us as we make every effort to clean our marketing list on a regular basis. --------------------------- Newsgroups: comp.os.vxworks Subject: viewing dynamically loaded modules Date: Tue, 29 Jun 1999 18:00:58 -0400 From: Roger Steines Organization: Georgia Tech Research Institute Message-ID: <37794219.F0363754@gtri.gatech.edu> How do I configure the debugger so that it knows what source code to display for the dynamicly loaded modules? I am using the Tornado debugger for UNIX. I have a handful of tasks I wish to run in parallel, with a main task loading and spawing the other tasks. To do so, from within the main task I have used the moduleLoad command to load the various modules, the symFindByName command to get the symbol values, and VXWTask class to create and spawn the various tasks. All of this seems to work perfectly, but I cannot view the source code from the Tornado Debugger. I can view the source of the main task, but the source code for the dynamically loaded modules come up in assembly. I have already verified that the debug flag was set during compilation. I have tried the add-symbol-file to add the object files to the hosts symbol file after the modules are up and running, but I keep getting an error claiming that the symbol files were not found on the target. I'm not sure whats wrong, since I can use a wind shell to verify that the symbols are in memory and the task are running. --------------------------- Newsgroups: comp.protocols.ppp,comp.os.vxworks Subject: Re: UDP performance over PPP Date: Thu, 24 Jun 1999 14:28:31 GMT From: saustin_deja@my-deja.com Organization: Deja.com - Share what you know. Learn what you don't. Message-ID: <7ktfa0$79d$1@nnrp1.deja.com> References: <7i1brc$47a$1@nnrp1.deja.com> <37715BBD.48B89842@usa.alcatel.com> >Where you > able to find a way to sync. buffer utilization with traffic >burstiness? What I ended up doing was "throttling" the delivery of the bursty traffic (snmp traps), based on overall mbuf usage. Each time a trap was to be generated, I look through all the mbuf statistics and see if any are below a certain threshold. If they are, I delay and try again, up to some maximum number of times. When the free mbufs are above the threshold again, or I went through the loop the max # times, I generate the trap. Ideally I'd only need to throttle on the mbuf queue for the size buffer I need, but unfortunately I have no way of telling what that is because the SNMP agent actually builds the buffer with all the variable bindings, etc. Here's what the code looks like: #undef SEND_DELAY_DEBUG void trapThrottle ( void ) { /* SRA 052499 Logic to avoid flooding PPP interface with traps. Look at the percentage in use of the network mbuf clusters If below the threshold, delay here while the interface clears out a bit and frees some mbufs. In an ideal world we'd know how big the trap is and would just have to look at the mbuf cluster that would be used. If the traps are sent on an Ethernet interface, the threshold should never get hit - buffers are processed too fast to ever see a drop. However, on a slow PPP interface sometimes the trap generation will exceed the link's ability to deliver the traps to the NMS. This is the case the throttling should handle */ /* NET_POOL *mySysPtr = ( NET_POOL *)_pNetSysPool ; */ /* FYI */ NET_POOL *myDataPtr = ( NET_POOL *) _pNetDpool ; CL_POOL *myPoolPtr ; /* point at each pool cluster */ unsigned int percentFree ; unsigned int i , j , delayCount ; #define MAX_TRAP_DELAY_TRIES 20 #define MBUF_DELAY_THRESHOLD 90 /* percent */ /* find the mbuf pool "statistics" for each mbuf size */ for ( i = 0 ; i < CL_TBL_SIZE ; i++ ) { myPoolPtr = (CL_POOL *) myDataPtr->clTbl[i] ; delayCount = 0 ; /* only interested in non-zero entries */ if ( (NULL == myPoolPtr) || (NULL == myPoolPtr->clSize) ) continue ; for ( j = 0 ; j < MAX_TRAP_DELAY_TRIES ; j++ ) { /* for this mbuf size, how free is it? */ percentFree = (myPoolPtr->clNumFree * 100) / (myPoolPtr->clNum) ; if ( percentFree < MBUF_DELAY_THRESHOLD ) { /* hold off a while */ delayCount++ ; taskDelay ( 20 ) ; /* want to be able to delay up to*/ /* an empirically derived # */ } } /* end for j */ #ifdef SEND_DELAY_DEBUG if ( delayCount ) { printf("Delayed %d times mbuf %d\n", delayCount , myPoolPtr->clSize); } #endif } /* end for i */ } Sent via Deja.com http://www.deja.com/ Share what you know. Learn what you don't. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: inter-processor comm Date: Tue, 29 Jun 1999 19:16:19 -0400 From: Elliot Lipeles Organization: LNS, Cornell U., Ithaca, NY 14853 Message-ID: <377953C3.710CFD27@mail.lns.cornell.edu> References: <7lare7$fd529@overload.lbl.gov> Smith wrote: > We are considering using a 2-processor solution for the project on which > I am working. The main reason is to off-load the comm functions > (including MAC and TCP/IP) to a dedicated processor, freeing the other > processor for functions. > > I am looking for possible solutions for the inter-processor comm. and > have some general. Any insight would be greatly appreciated. > > -What type of Multi-Processor support does vxWorks have? > > -What type of functionality does smlib give? > > -Is there such a thing as a CORBA ORB for shared memory? > > Thanks, > Brett Smith > brett.smith@bktech.com > > vxWorks The visegenics ORB (sold by Inprise) says that it will dynamically decide it the client and server are on the same host and in the case use shared memory. Elliot Lipeles --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Unaligned access fault Date: Tue, 29 Jun 1999 18:51:12 -0700 From: "Steve Doiel" Message-ID: <377975a9.0@news.pacifier.com> References: <37788031.A5877402@fmst.vtu.lt> >I'm not very familar with VxWorks, but would like to know about error >handling in it. I'm using VxWorks v3.2 on Alpha CPU, for some >communication tasks. And sometimes I'm getting such error: > >unaligned access fault > Fault VA = 7fffffff7fffffff > Opcode = 29 > Register = 10 > Exception PC = fffffc0001bf92f8 > Exception Stack Frame Pointer = fffffc0001c6e978 > Exception Stack Pointer = fffffc0001c6ea80 >Task: 0x1c6f740 > It's a bit ugly, but what I have found helps to isolate the problem is to start with ti 0x1c6f70 (in your example) (BTW: I'm a little rusty since I haven't done this in a while) VxWorks then displays task control block information which includes the program counter. Find the first symbol that precedes the program counter to determine which function you're in, and then you can disassemble code and match it up with the source. There should be an easier way (and probably is) but what I just described does work. SteveD --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Sample VxWorks driver for raw device? Date: Thu, 24 Jun 1999 14:57:56 -0700 From: Bob Irwin Organization: Systems Interface Software Message-ID: <3772A9E4.9DBD739C@systemsinterface.com> References: <37721FA6.493D1172@cur-archamps.fr> psav wrote: > > I'm looking for a sample about creating and installing a raw device and > its driver with VxWorks 5.2. Thank you. What kind of raw device? If its a disk, you can use the Raw File System (p 244 5.2 programmer's guide). If it's scsi based, but not disk, you can use scsiLib. - -- Bob Irwin bobi@systemsinterface.com Phone:281-484-6661 --------------------------- Newsgroups: comp.protocols.ppp,comp.os.vxworks Subject: PPP NT to vxWorks, can't detect disconnect Date: Tue, 29 Jun 1999 22:10:23 -0700 From: Melinda Stimpson Organization: Web Guy Productions Message-ID: <3779A6BF.1769@webguy-prod.com> Reply-To: melinda@webguy-prod.com I have a PPP connection from Windows NT to a system running vxWorks. (The vxWorks system is in silent mode.) NT calls in (using dial-up networking), the connection is established, all is well - until I try to disconnect from the NT side. vxWorks never sees the disconnect (specifically, the PPP disconnect hook function is never called), no matter what I do. I have tried hanging up the connection on the NT box, or exiting dial-up networking (disconnecting as I go), or even shutting down the PPP link on the vxWorks side via pppDelete (or whatever it's called; I'm at home now). Nothing results in the disconnect hook function being called. I really need to call it, because I need to know when to go back to listening for the NT dial-up networking "CLIENT" boloney. Anyone have any ideas? E-mail replies can/should go to mstimp@artel.com (my wife does not need a mailbox full of PPP technical details). Thanks! - -- There would be fewer problems with children if they had to chop wood to keep the television set going. --------------------------- Newsgroups: comp.protocols.ppp,comp.os.vxworks Subject: Re: Help: PPP link established, but can't do anything? Date: Tue, 29 Jun 1999 22:25:29 -0700 From: Melinda Stimpson Organization: Web Guy Productions Message-ID: <3779AA49.B1F@webguy-prod.com> References: <37754438.54C5@webguy-prod.com> <7l5aiv$5es@inetport.com> Reply-To: melinda@webguy-prod.com Clifford Kite wrote: > > Melinda Stimpson (melinda@webguy-prod.com) wrote: > > : I am trying to connect Windows NT to a vxWorks system using PPP. > : Once I got the NT non-PPP CLIENT/CLIENTSERVER nonsense out of the way, > : I was able to make a connection. However, I can't *do* anything with > : the connection - I can't telnet, rlogin, or even ping. > .. > > : problem). Note specifically the "Unknown protocol 80fd" - I presume > : that is telling me of a problem, but what, exactly? > > The 80fd refers to Compression Control Protocol (CCP) which "ppp0" > either doesn't support or is configured not to support. Neither this > or any of the ConfRej's for various LCP options prevented the link > negotiations from reaching an apparently normal completion. > > A silly question: Do you have a host route on each side of the link to > the other side through the PPP interface? Apparently, the problem was something to do with subnets. When I moved the PPP IP addresses to something that wasn't on the Ethernet subnet address of the NT box, everything worked. - -- There would be fewer problems with children if they had to chop wood to keep the television set going. --------------------------- Newsgroups: comp.os.vxworks Subject: using multiple routing protocols in SENS Date: Mon, 28 Jun 1999 15:29:16 GMT From: rksripada@hotmail.com Organization: Deja.com - Share what you know. Learn what you don't. Message-ID: <7l84bs$jmr$1@nnrp1.deja.com> The SENS manual says that when multiple routing protocols are used, the SENS can dynamically specify which route has the precedence. Can someone elaborate on how SENS can do this dynamically??? For example, there is a Router using SENS configured to run RIP and OSPF. RIP learns a route (let's say RIP-ROUTE) for a subnet S1. Also, the OSPF learns a route (let's say OSPF-ROUTE) for the same subnet S1. If the RIP-ROUTE and OSPF-ROUTE are different, which route will be given precedence for writing into the kernel's IP Forwarding Table ??? thanks, radhakrishna Sent via Deja.com http://www.deja.com/ Share what you know. Learn what you don't. --------------------------- End of New-News digest ********************** From owner-vxwexplo-process Wed Jun 30 05:47:31 1999 From: Matthieu Herrb Date: Wed Jun 30 05:47:34 PDT 1999 Subject: saving bootline in NVRAM on mvme2400 Hi, I'm using the VxWorks pre-release BSP from Motorola version 1.1/0.2 on a mvme 2400 board, and I can't get the bootline to be saved in NVRAM. I've put the the boot program (with the default boot line unchanged in config.h) into flash memory, and the board doesn't keep the parameters I'm typing with the boot progrwam 'c' command between reboots. Has someone seen this problem ? is there a solution (other then flashing a boot program with the good parameters or booting directly from PPC-Bug) ? ADVthanksANCE From owner-vxwexplo-process Wed Jun 30 07:39:16 1999 From: "Eckardt, Suzie" Date: Wed Jun 30 07:39:20 PDT 1999 Subject: Tornado Module Loader & ROM-Resident MIPS Symbols This message is in MIME format. Since your mail reader does not understand this format, some or all of this message may not be legible. ------_=_NextPart_001_01BEC306.2B250C80 Content-Type: text/plain; charset="iso-8859-1" I've got a ROM-resident VxWorks image executing on a custom MIPS/R3000 board; the image file is: vxworks.res_rom_nosym_res_low. I can attach a target server to this image and access the Tornado tools with no problem. My next step is to download application code to the target; I use the shell "ld" command to do this. When I inspect the downloaded code, I find that references to ROM-resident symbols are not resolved correctly. For example, "lkup" tells me that msgQCreate resides at 0xbfc1919c, yet the downloaded code tries to access msgQCreate at 0x8fc1919c. It seems that the MIPS physical address is properly resolved, but the virtual address is not. The module loader is generating RAM-based addresses for ROM-resident symbols! Has anyone experienced this before? Am I missing a configuration option? Could this be a problem with the target server? ************************************************ ** Suzie Eckardt ** Innovative Concepts, Incorporated ** McLean, VA ** eckardt@innocon.com ** 703-893-2007 x258 ************************************************ ------_=_NextPart_001_01BEC306.2B250C80 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Tornado Module Loader & ROM-Resident MIPS Symbols

I've got a ROM-resident VxWorks image executing on a = custom MIPS/R3000 board; the image file is: = vxworks.res_rom_nosym_res_low. I can attach a target server to = this image and access the Tornado tools with no problem. My next = step is to download application code to the target; I use the shell = "ld" command to do this. When I inspect the downloaded = code, I find that references to ROM-resident symbols are not resolved = correctly. For example, "lkup" tells me that msgQCreate = resides at 0xbfc1919c, yet the downloaded code tries to access = msgQCreate at 0x8fc1919c. It seems that the MIPS physical address = is properly resolved, but the virtual address is not. The module = loader is generating RAM-based addresses for ROM-resident = symbols!

Has anyone experienced this before? Am I = missing a configuration option? Could this be a problem with the = target server?

************************************************
** Suzie = Eckardt = ; = ;
** Innovative Concepts, = Incorporated
** McLean, VA
** eckardt@innocon.com
** 703-893-2007 x258
************************************************

------_=_NextPart_001_01BEC306.2B250C80-- From owner-vxwexplo-process Wed Jun 30 13:11:12 1999 From: Erin Swieringa Date: Wed Jun 30 13:11:15 PDT 1999 Subject: lptDrv with bi-directional parallel port config I am developing on an Intel x86 target (Tornado 1.0.1) and am attempting to use the line printer driver. I cannot print if the target PC's CMOS setting for Parallel Port Type is set as "Bi-directional" - it works if Parallel Port Type is set as "Compatible". We have a large installed base of systems with the Parallel Port Type configured as "Bi-directional" so I'm hoping to avoid modifying the CMOS setting. Our current application (running under DOS) prints successfully with the "Bi-directional" setting. I would think the lpt driver under VxWorks should also work. Among other things I have tried forcing the parallel port C_INPUT (data direction) control register value (using LPT_SETCONTROL as well as sysOutByte(LPT_CTRL(ioBase), ...) prior to calling the driver write routine with no success. Has anybody resolved the above issue? Erin Swieringa Electronic Theatre Controls From owner-vxwexplo-process Wed Jun 30 13:22:37 1999 From: Alan K Biocca Date: Wed Jun 30 13:22:41 PDT 1999 Subject: Re: Expoder working? There was a problem backing up the outgoing email. It is now catching up with several days of output. -- Alan K Biocca vxworks exploder author At 10:52 AM 6/29/99 , you wrote: >Alan, > >I have not received exploder messages for several days, including the >newsgroup archives. Is the system down or did I get bumped for some reason? > >Ron Wagner >rcw@dynres.com From owner-vxwexplo-process Wed Jun 30 18:47:01 1999 From: Fred Roeber Date: Wed Jun 30 18:47:06 PDT 1999 Subject: Re: inter-processor comm On Tue, 29 Jun 1999, Brett Smith wrote on vxWorks: > We are considering using a 2-processor solution for the project on which > I am working. The main reason is to off-load the comm functions > (including MAC and TCP/IP) to a dedicated processor, freeing the other > processor for functions. > > I am looking for possible solutions for the inter-processor comm. and > have some general. Any insight would be greatly appreciated. > > -What type of Multi-Processor support does vxWorks have? I suggest getting VxMP add on package from WRS. Extends semaphores and message queues to multiple processors. Much faster communication than smlib. > -What type of functionality does smlib give? smlib provides a backplane based, Shared Memory, IP communication path so that the two processors could look like two separate network nodes. Thus, this is a loosely coupled approach. Drawback is that the communication is slow and fairly CPU intensive since you are running whole network stack and doing some functions normally offloaded to the networking chip. > -Is there such a thing as a CORBA ORB for shared memory? I don't know. Fred | Fred J Roeber, BBN Systems & Technologies | | 4 John Clarke Road Middletown, RI 02842-5202 | | froeber@bbn.com 401-848-3548 | | TraceMaker product manager | | http://www.tracemaker.bbn.com |