From erxatech@inrete.it Thu Aug 1 02:50:25 1996 From: erxatech@inrete.it (Paolo Magistris) Date: Thu Aug 1 02:50:28 PDT 1996 Subject: real-time clock for a PC-board Hello, There is someone that can help me in the development of a real-time clock for a PC-board? There are some template driver available ? Thank you Paolo ================================================================= _______ |E______| Engineering |__R____| Robotics |____X__| Control Systems |______A| Applications Field Application Engineer Voice: +39-11-7412749 Erxa S.r.l. Fax: +39-11-7412044 via Morghen 22, 10143 Torino (ITALY) E-Mail: erxatech@inrete.alpcom.it ================================================================= From jhillman@wrs.com Thu Aug 1 06:28:14 1996 From: Jon Hillman Date: Thu Aug 1 06:28:15 PDT 1996 Subject: Re: Serial Communication Controller At 08:47 AM 7/31/96 PDT, Pierre Wong wrote: >I am trying to set up a Zilog Z85C30 Serial Communications Controller >in an asynchronous polling mode. Does anyone know the correct >initialization sequence to put the SCC in this mode? Pierre, VxWorks 5.3 has an 8530 serial driver which is a polling-type driver (or at least can be put in that mode). It's at target/src/drv/sio/z8530Sio.c. Tornado uses it with the debugger via the serial i/f to allow for system debugging. I hope this is what you were asking for. _________________________ | Jon Hillman, FAE | | Wind River Systems | /) (407) 273-4646 (\ / ) Email:jhillman@wrs.com ( \ _( (|_________________________) ) /> (((\ \) /,) / ) / //))/ (\\\\ \_/ / \ \_/ ///// \ / \ / \ _/ \_ / ----/ /----------------------\ \---- / / \ \ From 100656.1507@compuserve.com Thu Aug 1 09:07:50 1996 From: John Shaw <100656.1507@compuserve.com> Date: Thu Aug 1 09:07:51 PDT 1996 Subject: Motorola PowerPC 1604 power up failure Greetings, We are using MVME1600-001 cards from Motorola with VxWorks 5.2 boot PROMs. We are finding that these cards get into quite a state where their row of green lights show red! They then cannot be reset from the front panel and will not power off and on cleanly unless they are allowed to cool or discharge over a period of time after power off. The red light phenomenon occurs at power up. No messages appear if a terminal is attached. Has anyone else experienced this? Is a solution available? The boot PROMS were generated using the Green Hills compiler and BSP for VxWorks 5.2-beta. An added mezzanine processor card contains the PowerPC 604 on our kit. John Shaw, Senior Software Engineer, Shade (Computer Services) Limited Calne, Wiltshire, England, UK. From jiaqp@nuhepz.phys.nwu.edu Thu Aug 1 14:07:35 1996 From: QiuPing Jia Date: Thu Aug 1 14:07:37 PDT 1996 Subject: Re: comp.os.vxworks newsdigest Hi, everybody, I have a Motorola 68k board as my target which has vxworks running on it. My host system is SGI workstation. I am using vxgdb on my SGI to debug a program on 68K, my program has 10 parameters, but it seems vxgdb does not like to take 10th parameter. I got the following message when I gave 10 parameters from vxgdb: > (vxgdb) run te_test 3 8 0 100 1 1 0 0 0 1 > Starting program: te_test 3 8 0 100 1 1 0 0 0 1 > Can't encode argumentsCan't create task on remote target machine > But if I give less than 10 parameters, vxgdb can work happily. > (vxgdb) run te_test 3 8 0 100 1 1 0 0 0 > Starting program: te_test 3 8 0 100 1 1 0 0 0 > > Breakpoint 1, te_test (slot=3, clock_delay=8, nrz_delay=0, delay_multi=100, > svx_chain=1, r_mode=1, d_mode=0, h_mode=0, p_mode=0, n_event=0, non=0) at > sar_sgit.c:1082 I thought vxgdb should take 10 parameters, but it seems not. Is there anybody has the same experience? Or somebody knows how it is built, or somebody can think something wrong with my program? Any clue will be appreciate. Thanks, Qiuping jiaqp@nuhepz.phys.nwu.edu From intrepid!pandya@uunet.uu.net Thu Aug 1 15:11:58 1996 From: intrepid!pandya@uunet.uu.net (Janak Pandya) Date: Thu Aug 1 15:12:00 PDT 1996 Subject: Exception handler task in VxWorks Hi, My target is MVME167 (68040 based Motorola board). I am using VxWorks 5.2 All the application code is in ADA. I have sereral ADA tasks running and here's what I am trying to do. I want to create an exception handler task (just like VxWorks "tExcTask") to catch hardware exceptions caused by any of the tasks running on the board. I also want to process interrupts that are not handled by other tasks by setting up interrupt handler for all possible interrupts before other tasks are activated. This way, if any of tasks installs the interrupt handlers after the exception handler task has done so then that task will handle the interrupts. My problem is for the exception processing. 1. Somehow I want to figure out the task Id of task or the Interrupt Service Routine (ISR) which caused the exception. 2. I also want to notify the offending task that there was an exception. PLEASE, let me know if you have any suggestions or questions. Thank you. Janak Pandya ------------------------------------------------------------------------------ Janak Pandya 813-347-7645 Work E-Systems 813-347-2890 Fax 1501 72nd Street N, MS-61 pandya@eci-cec.com St. Petersburg, FL 33710 ------------------------------------------------------------------------------ From kbaumann@itr.ch Thu Aug 1 23:30:00 1996 From: Kurt Baumann Date: Thu Aug 1 23:30:03 PDT 1996 Subject: Re: Exception handler task in VxWorks ---------- X-Sun-Data-Type: text X-Sun-Data-Description: text X-Sun-Data-Name: text X-Sun-Content-Lines: 11 Hi 1. To show all installed interrupt service routines, you can use veclist.c with the function veclist() 2. a demo how to handle interrupts with a message queue handler, you can see bsp 3.c (it doesn't run, because there are missing some includes, just for lookin g) I hope, it will help you Kurt Baumann and Theo Hugentobler, ITR, CH kbaumann@itr.ch ---------- X-Sun-Data-Type: c-file X-Sun-Data-Name: veclist.c X-Sun-Content-Lines: 216 /* * @(#)veclist.c 1.10 * * list fuctions attached to the interrupt vector table * * Created 28Mar89 Jeffrey O. Hill * hill@atdiv.lanl.gov * (505) 665 1831 * * .01 010393 Applied fix for zero C ISR param causes incorrect * identification as MACRO ISR problem. */ /* * * makefile * * * V5VW = /.../vx/v502b * * veclist.o: * cc68k -c -DCPU_FAMILY=MC680X0 -I$(V5VW)/h veclist.c * * */ #include "vxWorks.h" #include "stdio.h" #include "intLib.h" #include "vxLib.h" #include "iv.h" #include "ctype.h" #include "sysSymTbl.h" static char *sccsID = "@(#)veclist.c 1.10\t1/3/94 J. Hill hill@atdiv.lanl.gov"; /* * * VME bus dependent * */ #define NVEC 0x100 static char *ignore_list[] = {"_excStub","_excIntStub"}; int veclist(int); int cISRTest(FUNCPTR proutine, FUNCPTR *ppisr, void **pparam); static void *fetch_pointer(unsigned char *); /* * * veclist() * */ int veclist(int all) { int vec; int value; SYM_TYPE type; char name[MAX_SYS_SYM_LEN]; char function_type[10]; FUNCPTR proutine; FUNCPTR pCISR; int cRoutine; void *pparam; int status; int i; for(vec=0; vec My target is MVME167 (68040 based Motorola board). I am using VxWorks 5.2 > All the application code is in ADA. I have sereral ADA tasks running and > here's what I am trying to do. > > I want to create an exception handler task (just like VxWorks "tExcTask") > to catch hardware exceptions caused by any of the tasks running on the > board. I also want to process interrupts that are not handled by > other tasks by setting up interrupt handler for all possible interrupts > before other tasks are activated. This way, if any of tasks installs the > interrupt handlers after the exception handler task has done so > then that task will handle the interrupts. My problem is for > the exception processing. > > 1. Somehow I want to figure out the task Id of task or the Interrupt > Service Routine (ISR) which caused the exception. > > 2. I also want to notify the offending task that there was an exception. > I did something like this on a 167 board under VADSworks. I replaced all of the vectors that the Ada runtime used for exceptions with my own interrupt handler. These vectors were mapped to Ada exceptions like storage_error and constraint_error. The new interrupt handler saved the context of the processor in a fixed location. The context included all of the registers and the first few hundred bytes of the stack. After this was done the Ada runtime interrupt routine was called, which generated an Ada exception as usual. If the exception handler decided that the exception was serious, it could log the state information that was collected at interrupt level. This allowed us to pinpoint the cause of the exception off-line. I suppose you could send the state information to your exception task in a queue. The currently executing task ID is available from the variable taskIdCurrent, so you could send it along with the processor state. If an exception occurs at interrupt level, you could just use the PC to identify the ISR. Be careful though, many Ada runtime systems require that all exceptions be suppressed in an interrupt handler. Good luck, Jim ------ Jim Getker Sentient Networks Inc. 941-371-5454 ext. 30 getker@sentientnet.com From daemon@csg.lbl.gov Fri Aug 2 04:01:05 1996 From: daemon@csg.lbl.gov Date: Fri Aug 2 04:01:08 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Fri Aug 2 04:01:01 PDT 1996 Subject: Re: printf() broken in Tornado's version of GNU C Subject: Re: printf() broken in Tornado's version of GNU C Subject: Industrial Automation Job Opportunities Subject: Re: NTP availability? Subject: Re: --- PowerPC guru's --- Subject: Re: Booting from Windows NT Subject: SMC Driver Subject: SMC Driver Subject: Re: Minimal pcnfsd For VxWorks? Subject: Re: Debugging a task at startup Subject: Re: Why can't VxWorks handle multi-protocol sockets? Subject: VxWorks driver for SMC91C94 ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: printf() broken in Tornado's version of GNU C Date: 29 Jul 1996 19:00:59 GMT From: garyjohn@spk.hp.com (Gary Johnson) Organization: Hewlett-Packard, Spokane Division Message-ID: <4tj1pb$jec@hpcvsnz.cv.hp.com> References: <4tgn51$n34@uhura.phoenix.net> Bruce Wedding (brucew@phoenix.net) wrote: > I'm having trouble getting either of the following lines to work > properly in Tornado vxWorks 5.3. > > printf("%25s", buf); > > or > > printf("%*s", 25, buf); > > I'm passing a buffer that is not nul terminated, so if the length > specification doesn't work, I print off into hyperspace. That is > exactly what is happening now. Anyone else ever tried to use > printf() this way? Is there a fix? This is not a bug--printf() is supposed to work this way. Referring to the print() man page, the number following the % is the "minimum field width." Printf() will always attempt to print the argument string in its entirety. To print only the first 25 characters of a non-null-terminated "string", you will have to do something like this: char pbuf[26]; strncpy(pbuf, buf, 25); pbuf[25] = '\0'; printf("%25s", pbuf); I added pbuf because I assumed that you didn't want to alter the contents of buf. Gary --------------------------- Newsgroups: comp.os.vxworks Subject: Re: printf() broken in Tornado's version of GNU C Date: 29 Jul 1996 21:10:24 GMT From: litwin@chroma.jpl.nasa.gov (Todd Litwin) Organization: Jet Propulsion Laboratory, Pasadena, CA Message-ID: <4tj9c0$m2p@netline-fddi.jpl.nasa.gov> References: <4tgn51$n34@uhura.phoenix.net> <4tj1pb$jec@hpcvsnz.cv.hp.com> In article <4tj1pb$jec@hpcvsnz.cv.hp.com> garyjohn@spk.hp.com (Gary Johnson) writes: >Bruce Wedding (brucew@phoenix.net) wrote: >> I'm having trouble getting either of the following lines to work >> properly in Tornado vxWorks 5.3. >> >> printf("%25s", buf); >> >> or >> >> printf("%*s", 25, buf); >> >> I'm passing a buffer that is not nul terminated, so if the length >> specification doesn't work, I print off into hyperspace. That is >> exactly what is happening now. Anyone else ever tried to use >> printf() this way? Is there a fix? > >This is not a bug--printf() is supposed to work this way. Referring to >the print() man page, the number following the % is the "minimum field >width." Printf() will always attempt to print the argument string in >its entirety. To print only the first 25 characters of a >non-null-terminated "string", you will have to do something like this: Well, yes and no. Yes, the first printf() above will try to print the entire string. But, no, you don't need to go to great lengths to do what you want. The problem above is that the wrong format string was used. To limit the string to a specific length, say 25, you should do printf("%.25s", buf); Note the period before the 25. See page 147 in "The C Programming Language" by Kernighan and Ritchie. - -- Todd Litwin Jet Propulsion Laboratory (818) 354-5028 Todd.E.Litwin@jpl.nasa.gov --------------------------- Newsgroups: comp.os.vxworks Subject: Industrial Automation Job Opportunities Date: Wed, 31 Jul 1996 11:27:34 -0700 From: "David Loop" Organization: Newnes Machine Ltd. Message-ID: <01bb7f0d.f1a2eeb0$7d0814ac@bae-pc> NEWNES MACHINE LTD JOB OPPORTUNITIES Imagine working where others dream of having their vacation... where commuting is like a Sunday drive... where the word 'community' means more than a place your taxes go to. Salmon Arm, British Columbia, the heart of the spectacular Shuswap Lake region, is one of these places. Many of the people lucky enough to live there work for Newnes Machine Ltd. Newnes is North America's leading supplier of state-of-the-art, fully automated sawmill equipment, with over 400 employees. If you are a results oriented team player with qualifications and skills that match our job opportunities, we would like you to consider joining us and our way of life. Our Optimization team is looking for people with the following skills software design - real time industrial automation VxWorks, optimization algorithms, 3D modelling, decision logic software design - industrial man machine interface WinNT, SQL databases, object oriented / multi-threaded designs software design - embedded signal processing 2D/3D image processing, DSP, high speed comm and data acquisition system design - industrial automation VME and PC platforms, networking, distributed database, high speed data acquisition, power conditioning hardware design - profile measurement sensors laser triangulation, vision, CCD arrays, embedded image processing For immediate consideration please forward a cover letter and resume to David Loop, Optimization Business Manager. Newnes Machine Ltd. PO Box 8 Salmon Arm, B.C. V1E 4N2 email david.loop@newnes.com tel 604 832-7116 fax 604 832-7890 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: NTP availability? Date: 29 Jul 1996 20:11:05 -0700 From: mwette@mr-ed.jpl.nasa.gov (Matt Wette) Organization: Jet Propulsion Laboratory Message-ID: <4tjug9$br7@mr-ed.jpl.nasa.gov> References: <4tid19$mcm@crc-news.doc.ca> In article <4tid19$mcm@crc-news.doc.ca>, Donald McLachlan wrote: > >I work with Norm Young (who started this thread), we sort of have gettimeofday, >settimeofday, adjtime etc. Currently we have this stuff working, but without >being able to access the real time clock (even with the MMU support disabled) so >time at boot up is Jan 1 1977, but xntpd should fix this :-). >Wanna trade? > >Anyone out there able to access the real time clock on the hkv4f using the macros >supplied by vxWorks 5.2? The clock on the HKV4F is not of very high precision if I remember right. Something like 10ms comes to mind. You may want to just work with the kernel tick by using clock_gettime and clock_settime. Anyway the routines you want are in config/hkv4f/dp8750aRTC.c, more or less. I started down this track a few years back and gave up. (I couldn't get xntpd to converge). Below is what I've dug out from my efforts. In the meantime I've been trying to work bottom-up from the NTP RFC's. Good luck and let us all know if you get it working! Matt #include #include #include #include static int gettime_tm(struct tm *tm); static int settime_tm(struct tm *tm, int nsec); static long nsec_adj = 0; static long tick_adj = 0; int sysclk_tick_handler(int arg) { struct timespec ts; if (nsec_adj == 0) tickAnnounce(); clock_gettime(CLOCK_REALTIME, &ts); if (nsec_adj > 0) { ts.tv_nsec += nsec_adj; if (ts.tv_nsec >= 1000000000) { ts.tv_sec++; ts.tv_nsec -= 1000000000; } nsec_adj -= tick_adj; } else { ts.tv_nsec -= nsec_adj; if (ts.tv_nsec < 0) { ts.tv_sec--; ts.tv_nsec += 1000000000; } nsec_adj += tick_adj; } clock_settime(CLOCK_REALTIME, &ts); tickAnnounce(); return 0; } int setup_ticker(void) { tick_adj = 5000000; sysClkConnect((FUNCPTR)sysclk_tick_handler, 0); } int showtime(void) { struct timespec ts; clock_gettime(CLOCK_REALTIME, &ts); printf("%09d\n", ts.tv_nsec); return 0; } int gettimeofday(struct timeval *tp, void *tz) { struct timespec ts; clock_gettime(CLOCK_REALTIME, &ts); tp->tv_sec = ts.tv_sec; tp->tv_usec = ts.tv_nsec/1000; return 0; } int settimeofday(struct timeval *tp, void *tz) { struct timespec ts; struct tm tm; int status; ts.tv_sec = tp->tv_sec; ts.tv_nsec = 1000*(tp->tv_usec); clock_settime(CLOCK_REALTIME, &ts); localtime_r(&tp->tv_sec, &tm); settime_tm(&tm, ts.tv_nsec); return 0; } int adjtime(const struct timeval *delta, struct timeval *olddelta) { struct timeval tv; struct timespec ts; int key; #if 1 tick_adj += (delta->tv_usec)*1000; #else key = intLock(); clock_gettime(CLOCK_REALTIME, &ts); ts.tv_sec += delta->tv_sec; ts.tv_nsec += 1000*(delta->tv_usec); if (ts.tv_nsec >= 1000000000) { ts.tv_sec++; ts.tv_nsec -= 1000000000; } else if (ts.tv_nsec < 0) { ts.tv_sec++; ts.tv_nsec += 1000000000; } clock_settime(CLOCK_REALTIME, &ts); intUnlock(key); #endif return 0; } xdate() { time_t t; t = time(0); printf("%s", asctime(localtime(&t))); } time_t time(time_t *tp) { struct timespec ts; clock_gettime(CLOCK_REALTIME, &ts); if (tp) *tp = ts.tv_sec; return ts.tv_sec; } - -- matthew.r.wette@jpl.nasa.gov --------------------------- Newsgroups: comp.os.vxworks Subject: Re: --- PowerPC guru's --- Date: Mon, 29 Jul 1996 12:22:42 -0600 From: chris_estes@smtp.svl.trw.com (Chris Estes) Organization: TRW Message-ID: References: <4t8nnp$a4t@badger.heurikon.com> Ted, Your are absolutely correct. RomInit.s is failing to initialize the MPC105 Eagle chip. We discovered this back in April or so, right after the non-Beta release of vxWorks 5.2 for the PowerPC came out. With the beta version, our PCI device worked fine. In the non-beta, our PCI device was having trouble getting the PCI bus at times. In any case, Wind River has been informed of this oversight and it is on their list of things to do. It may have been somewhat fixed in 5.3, I haven't had time yet to try and port to the newer version. By the way Fred, our you looking at this code to produce a BSP for Heurikon's PowerPC board? - -Chris Estes TRW SIG Sunnyvale --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Booting from Windows NT Date: 1 Aug 1996 04:48:56 GMT From: steved@pacifier.com@199.2.117.163 (Steve Doiel) Organization: Pacifier BBS, Vancouver, Wa. ((360) 693-0325) Message-ID: <4tpcvo$7n5@news.pacifier.com> References: <4tmmgd$m47@nnrp1.news.primenet.com> Reply-To: steved@pacifier.com (Steve Doiel) Douglas Jerome writes: >Help! > >My vxworks target boots fine from my sparc, but it won't boot from >my Windows NT computer. Actually, vxworks boots but it hangs forever >trying to download the symbol table. > >I've seen a reply to a similar posting, but I did not relly understand >it. Do I need to make new boot roms? Or do I simply make some >changes to the vxworks load module? I boot VxWorks from my NT machine frequently. I initially encountered several problems relating to NT security and FTP security. On my machine I created a user for my VxWorks target and gave it a password. This password is used by VxWorks to gain access to the machine. In addition, under Control Panel their is an FTP setup. For VxWorks (or other systems) to read the drive using FTP the drive must be set up as readable. From file manager I set up the directoy containing the file I was booting with read access for the VxWorks target user. I'm not sure I hit all of the bases, but I think so. Steve Doiel --------------------------- Newsgroups: comp.os.vxworks Subject: SMC Driver Date: Thu, 25 Jul 1996 17:06:14 -0700 From: Emanoel Organization: Global Village Communication, Inc. Message-ID: <31F80BF6.7619@globalvillage.com> Does anyone have a current 68360 SMC TTY driver compatible with VxWorks? --------------------------- Newsgroups: comp.os.vxworks Subject: SMC Driver Date: Thu, 25 Jul 1996 17:07:59 -0700 From: Emanoel Organization: Global Village Communication, Inc. Message-ID: <31F80C5E.6BD0@globalvillage.com> Does anyone have a 68360 SMC TTY driver compatible with VxWorks? Please respond to emanoel@globalvillage.com thank you very much, emanoel. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Minimal pcnfsd For VxWorks? Date: 30 Jul 1996 18:19:11 -0700 From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star, Oakland, CA Message-ID: <4tmcaf$eqm@idiom.com> References: Mark Fisher (mfisher@dyn.com) wrote: : I want to be able to mount a disk on vxWorks on a PC. I can get this : to work by specifying a server that is running pcnfsd to do the user : authentication. However, I need to be able to have the vxWorks system : do the user authentication in the field. Is there any port (even a : minimal one) of pcnfsd that is available that does this? Or does : anyone have any suggestions as to how to get around this? You can get source code for pcnfsd from various FTP sites. Porting it to vxWorks should take less than several hours. I've ported pcnfsd to a custom embedded kernel (not VxWorks) based on 4.4 BSD network code and it took trivial amount of time. - -- H.J. Bae http://www.peacefulstar.oakland.ca.us --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Debugging a task at startup Date: 30 Jul 1996 18:28:59 -0700 From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star, Oakland, CA Message-ID: <4tmcsr$fb1@idiom.com> References: <4t1gm2$ece@news.pacifier.com> If you use Tornado, you can click on the little box in RUN dialog box under debug menu which says something like 'break at task start' or something like that. Other people have suggested several different solutions for vxgdb and older VxWorks. If you build a custom GDB cross-debugger for your VxWorks (using original GNU source code), you can build a version of debugger and add your own gdb_trigger() routine which can be used to invoke the debugger trap anywhere in your code (including at the start of the task). Perhaps vxgdb has this? - -- H.J. Bae http://www.peacefulstar.oakland.ca.us --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Why can't VxWorks handle multi-protocol sockets? Date: 30 Jul 1996 18:13:31 -0700 From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star, Oakland, CA Message-ID: <4tmbvr$em7@idiom.com> References: <4tkjcb$bc2@idiom.com> Blaine Kubesh (blaine@texas.net) wrote: : I don't know how to just replace the native network code without : the source code to the VxWorks kernel. Do they distribute the : kernel source? You can replace the native network kernel code with 4.4 BSD without VxWorks kernel source code. This has been done before (but the code is not available for public consumption -- done under a contract). The real issue has to do with making it general enough so that it will work with all those network devices. WRS will sell you source code for significant amount of money if you really need the source code. - -- H.J. Bae http://www.peacefulstar.oakland.ca.us --------------------------- Newsgroups: comp.os.vxworks Subject: VxWorks driver for SMC91C94 Date: 2 Aug 1996 10:07:24 GMT From: <100632.3602@compuserve.com> Organization: CompuServe Incorporated Message-ID: <4tsk0s$3dp@dub-news-svc-4.compuserve.com> Does anyone have a VxWorks driver for SMC91C94 or SMC91C92? C. Burmann --------------------------- End of New-News digest ********************** From sy@wrs.com Fri Aug 2 12:16:49 1996 From: Sayeed Choudhury Date: Fri Aug 2 12:16:51 PDT 1996 Subject: SMC 919x drivers Hello, >From: <100632.3602@compuserve.com> >Does anyone have a VxWorks driver for SMC91C94 or SMC91C92? > >C. Burmann WRS has just started a web page, in order to provide freeware drivers that have been written for VxWorks. They may or may not work, and their origin may or may not be Wind River Systems. They may have been originally written for use with Tornado 1.0, VxWorks 5.2, VxWorks 5.1, or earlier releases. Please keep in mind that they may also contain architecture-related specifics. The first driver just so happens to be the SMC91C9x, so youre in luck! The page is located at: http://www.wrs.com/frundd.html I would encourage all developers to contact me as they develop drivers, such that if appropriate, I can work with you to put the driver on our Web page to help other developers. Cheers, Sy From vasta@proton.llumc.edu Fri Aug 2 17:44:05 1996 From: Steve Vasta Date: Fri Aug 2 17:44:07 PDT 1996 Subject: ONC/TLI RPC Compatibility problems? Does anyone know of any compatibility problems using TLI RPC client software (from Solaris 2.4) against ONCRPC servers running under VxWorks 5.2? I'm able to run RPC TLI client/server tests OK if both machines are Solaris (i.e. both machines use TLI RPC . . . also known as TIRPC), and if both machines run ONCRPC (such as SunOS4.1.x and VxWorks 5.2). However, I am getting corrupted data (on the server side within the proper callback) when I run a TLI client against an ONC server and vice versa. My questions at this point are: 1) Can TLI RPC and ONC RPC be mixed like this? 2) If not, is there a version of TLI RPC that can be built for VxWorks 5.2 and/or Sun4 machines? Any help would be greatly appreciated! :-) Steve Vasta From daemon@csg.lbl.gov Sat Aug 3 04:00:44 1996 From: daemon@csg.lbl.gov Date: Sat Aug 3 04:00:46 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sat Aug 3 04:00:40 PDT 1996 Subject: Getting up to speed with VxWorks Subject: Purify like tools for VXWORKS?? Subject: Re: System Controller in slot 1 Subject: NCR/Symbios Logic 53C720 SCSI Driver Source Subject: Re: printf() broken in Tornado's version of GNU C Subject: Re: Has anyone recieved this email? Subject: ping and VxWorks 5.3 Subject: interrupt handler question Subject: VxWorks vs. OS-9 ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Getting up to speed with VxWorks Date: 1 Aug 1996 22:03:20 GMT From: Karin Larsen Organization: Nichols Research Corporation Message-ID: <4tr9j8$g41@michp1.redstone.army.mil> Can anyone tell me if you have limited real-time embedded experience, how hard is it to get up to speed with VxWorks? Are the training courses from Wind Rivers useful? Please e-mail to masseys@nichols.com --------------------------- Newsgroups: comp.os.vxworks Subject: Purify like tools for VXWORKS?? Date: Fri, 02 Aug 1996 10:47:50 -0600 From: Aseem Bakshi Organization: U S WEST - !NTERACT Services Message-ID: <32023136.8AA@phbtsus.com> Hi, I am looking for tools that perform memory checks similar to the capabilities that Purify provides on VXWORKS. Could anybody who knows of or has used such tools post back into this group or send me E-mail Thanx in advance Aseem - -- - ----------------------------------------------------------------------------- Aseem Bakshi Phone: (801)9771574 Philips Broadcast Television Systems Inc Alias: Polka Boy Salt Lake City, UT E-mail Abakshi@phbtsus.com --------------------------- Newsgroups: comp.arch.bus.vmebus;,comp.os.vxworks Subject: Re: System Controller in slot 1 Date: 2 Aug 1996 12:02:26 GMT From: tfm@sei.cmu.edu (Theodore Marz) Organization: The Software Engineering Institute Message-ID: <4tsqoi$mju@news.sei.cmu.edu> References: <4tr41n$ogd@fountain.mindlink.net> In article <4tr41n$ogd@fountain.mindlink.net>, Philip.Wiebe@newnes.com (Philip J. Wiebe) writes: |> |> |> The VMEBus specificaiton specifies that the system controller MUST be |> in slot 1. Most VME boards (at least the MVME162) specifies this as |> well. However, placing the system controller in any other slot works |> as well. We have not tried it in a production version, but in-house |> only. Has this restriction been eliminated, or will the system |> controller be unreliable in any other slot than slot 1? |> |> Philip Wiebe |> Newnes Machine Ltd. |> |> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |> Philip J. Wiebe "It has yet to be proven that intelligence |> Software Engineer has any survival value." |> Newnes Machine Ltd. - Arthur C. Clarke |> P.O.Box 8, 3550-45th Street S.E. |> Salmon Arm, BC Canada |> email: Philip.Wiebe@newnes.com |> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |> Given how the VME bus arbitrates (bus requests, bus grants, interrupts, etc) I don't think so. If you aren't using interrupts, then you PROBABLY could declair either end of the chassis to be "slot 1" and proceed. What I wouldn't do is put bus mastering cards on BOTH sides of the system controller card. You could PROBABLY get away with putting passive cards (ie. memory, non-mastering I/O cards, etc) on the "wrong" side of the system controller, as they do not try to grab the bus. Ted Marz --------------------------- Newsgroups: comp.os.vxworks Subject: NCR/Symbios Logic 53C720 SCSI Driver Source Date: Thu, 01 Aug 1996 23:03:51 +0000 From: landl@voicenet.com Organization: Voicenet - Internet Access - (215)674-9290 Message-ID: <320137D7.232E@voicenet.com> Reply-To: landl@voicenet.com Please forward any leads to an NCR/Symbios Logic 53C720 SCSI VxWorks Driver to leer@acttechnico.com Thank you, Lee J. Raguz --------------------------- Newsgroups: comp.os.vxworks Subject: Re: printf() broken in Tornado's version of GNU C Date: Tue, 30 Jul 1996 14:29:10 GMT From: brucew@phoenix.net (Bruce Wedding) Organization: NeoSoft, Inc. Message-ID: <4tl6g6$3f4@uuneo.neosoft.com> References: <4tgn51$n34@uhura.phoenix.net> <4tj1pb$jec@hpcvsnz.cv.hp.com> <4tj9c0$m2p@netline-fddi.jpl.nasa.gov> litwin@chroma.jpl.nasa.gov (Todd Litwin) wrote: >But, no, you don't need to go to great lengths to do what you want. The >problem above is that the wrong format string was used. To limit the string to >a specific length, say 25, you should do > printf("%.25s", buf); Yep, I figured it out after reading the first response. I've used the precision operator with floats, but never can seem to remember it with strings. Thanks, Bruce --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Has anyone recieved this email? Date: 26 Jul 1996 14:45:59 GMT From: tfm@sei.cmu.edu (Theodore Marz) Organization: The Software Engineering Institute Message-ID: <4taln7$hri@news.sei.cmu.edu> References: John, I sure have. Seems she is a recruiter working for a company in the Atlanta area looking for VxWorks types. Ted Marz tfm@sei.cmu.edu --------------------------- Newsgroups: comp.os.vxworks Subject: ping and VxWorks 5.3 Date: Fri, 26 Jul 1996 10:56:17 -0400 From: Michael Charnoky Organization: NASA Lewis Research Center Message-ID: <31F8DC91.72C5@fcfsrv1.lerc.nasa.gov> Hello, Our software lab recently installed VxWorks on a few remote 68040 targets. I've been writing code to test the ethernet connection between the boards. In order to troubleshoot some promlems, i tried to call the "ping" routine directly from the shell, but got an error: - -> ping "ourtarget",1,0 undefined symbol: ping This seemed strange... so i wrote a short program, including "pingLib.h" to exercise "ping." Unfortunately, I got the same error. Has anyone else had this problem?? I'm stumped... the tech support web page lends no help - plus the keyword search doesn't even work. Any help is much appreciated!! mike - ----------------------------------------------------------------------------- Michael Charnoky | phone: 216-433-3162 NASA Lewis Research Center | Cleveland, Ohio 44135 | email: mcharnok@fcfws3.lerc.nasa.gov - ----------------------------------------------------------------------------- --------------------------- Newsgroups: comp.os.vxworks Subject: interrupt handler question Date: Tue, 30 Jul 1996 14:20:19 -0500 From: Jeff Tseng Organization: Massachusetts Institute of Technology Message-ID: <31FE6073.41C6@fnal.gov> Hi, I am not completely sure that this is a VxWorks problem, but here goes: I am writing an interrupt handler for a PowerPC-based MVME1603 single-board computer which uses the VME2PCI ASIC and VMEchip2 for its VME interface. I've connected the handler to interrupt number 11 using intConnect(INUM_TO_IVEC(11),...); the handler is to signal the end of a DMA transfer over the VMEbus. It clears the interrupt and logs a message just before it returns with the call logMsg ("return from interrupt\n",0,0,0,0,0,0); return; When I run the program, what I get is the logged message and then something else: interrupt: return from interrupt interrupt: uninitialized interrupt 4144 before logging any other messages (the main program merely starts the DMA, logs another message, and returns--yes, the last message does come out after the interrupt-originated messages). So my question is: what is this uninitialized interrupt 4144? I cannot connect to it with intConnect. Otherwise the program works fine: it does the DMA transfer and logs all the messages I expect along with this one I don't. Any help would be appreciated. Jeff Tseng --------------------------- Newsgroups: comp.os.vxworks Subject: VxWorks vs. OS-9 Date: Fri, 02 Aug 1996 15:09:02 -0700 From: "Timothy A. Meier" Organization: Lawrence Livermore National Laboratory Message-ID: <32027C7E.4A6C@llnl.gov> Does anyone have a feature comparison chart/discussion of VxWorks vs. OS-9? I have used both a little, and am about to start a large project that could last several years and involve many other developers. My experience (so far) tells me that either will probably work for our application. If I can't justify a preference based on features and capabilities, the decision will boil down to cost. I beleive that OS-9 is less expensive per seat, but in the long run VxWorks may save money in development/debug time. Info and Opinions appreciated. - -- Tim Meier (x2-3341) B482 R1024 L-492 (meier3@llnl.gov) --------------------------- End of New-News digest ********************** From daemon@csg.lbl.gov Sun Aug 4 04:00:32 1996 From: daemon@csg.lbl.gov Date: Sun Aug 4 04:00:35 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sun Aug 4 04:00:29 PDT 1996 Subject: Re: NTP availability? Subject: Re: Getting up to speed with VxWorks Subject: Condition variables, reader/writer locks, barriers, etc ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: NTP availability? Date: 26 Jul 1996 19:36:44 GMT From: young@uranus.dgrc.doc.ca (Norm Young) Organization: The Communications Research Centre Message-ID: <4tb6oc$nco@crc-news.doc.ca> References: <9606268383.AA838396366@ccgate.hac.com> In article <9606268383.AA838396366@ccgate.hac.com>, sbadstuebner@ccgate.hac.com writes: > Does anyone know of an NTP port to VxWorks? > > thanks, > > Stefan Badstuebner > sBadstuebner@ccgate.hac.com > There was one several years ago but but I haven't seen one for the more recent versions of VxWorks. I'm also interested in xntp on VxWorks 5.2. If anyone knows of a source, please post here. - -- Norm Young | DoD #0770 | E-Mail: Norm.Young@crc.doc.ca Communications | Tel: 1-613-990-8779 | Fax: 1-613-990-7987 Research Centre | Ottawa, Ontario | No Disclaimer, No Quote --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Getting up to speed with VxWorks Date: 2 Aug 1996 13:59:51 GMT From: johncoby@ghgcorp.com (JOHN R. COBARRUVIAS) Organization: Personal Account Message-ID: References: <4tr9j8$g41@michp1.redstone.army.mil> Read the book "Programmers guide vxWorks 5.2". That's how I learned and I had limited realtime embedded experience. (Actually I had NONE!) Very good set of documents. I must have read it 3 or 4 times over the past 2 years. Also, the WindRiver courses are pretty good. They pretty much cover the book from end to end. It is best to read the book, know the book, live the book before going to class to get the most out of it. Also, try Sparta Inc. They have intro,intermediate, and advanced courses. they have a web page with details. In article <4tr9j8$g41@michp1.redstone.army.mil>, Karin Larsen wrote: > Can anyone tell me if you have limited real-time embedded experience, > how hard is it to get up to speed with VxWorks? Are the training > courses from Wind Rivers useful? > > Please e-mail to masseys@nichols.com - -- ,-_/ . .-,--. ,---. . ' | ,-. |-. ,-. `|__/ | -' ,-. |-. ,-. ,-. ,-. . . . , . ,-. ,-. | | | | | | | | \ ,. | -. | | | | ,-| | | | | | / | ,-| `-. | `-' ' ' ' ' `-' ` `' `---' `-' `-' `-^ ' ' `-' `' ' `-^ `-' / | John R. Cobarruvias johncoby@ghgcorp.com Houston, Tx `--' --------------------------- Newsgroups: comp.os.vxworks,comp.programming.threads Subject: Condition variables, reader/writer locks, barriers, etc Date: Thu, 01 Aug 1996 23:19:49 -0700 From: Roger McDanell Organization: NetUSA1 Inc. Message-ID: <32019E05.254E@netusa1.net> I am interested in implementing condition variables, reader/writer locks, barriers, etc. by building on top of VxWorks primitives (i.e. binary semaphores) or perhaps the classical P(), V() functions. I have done fairly simple 1st-cut implementations, but am not satisfied that they are robust. I would appreciate any information related to implementation of these items. These items are part of POSIX, so Wind River Systems may eventually supply this functionality. I need it now. --------------------------- End of New-News digest ********************** From tkb@mclean.sparta.com Sun Aug 4 21:31:59 1996 From: "Thomas K. Buchanan" Date: Sun Aug 4 21:32:01 PDT 1996 Subject: Re: System Controller in slot 1 >|> The VMEBus specificaiton specifies that the system controller MUST be >|> in slot 1. Most VME boards (at least the MVME162) specifies this as >|> well. However, placing the system controller in any other slot works >|> as well. We have not tried it in a production version, but in-house >|> only. Has this restriction been eliminated, or will the system >|> controller be unreliable in any other slot than slot 1? >Given how the VME bus arbitrates (bus requests, bus grants, interrupts, etc) >I don't think so. If you aren't using interrupts, then you PROBABLY could >declair either end of the chassis to be "slot 1" and proceed. > >What I wouldn't do is put bus mastering cards on BOTH sides of the system >controller card. You could PROBABLY get away with putting passive cards >(ie. memory, non-mastering I/O cards, etc) on the "wrong" side of the system >controller, as they do not try to grab the bus. The VME controller must be in the leftmost occupied slot, normally slot 1. Bus arbitration will not work properly if you place bus masters left of the master. It is also a design violation to put I/O cards left of the bus master. Both the bus master arbitration and interrupt cycles depend on some signals propogating left to right, it doesn't work in reverse. In summary, if you only have one card in the system put it in slot 1 for consistencies sake. If you have two or more cards that use interrupts and/or will arbitrate for bus mastership, you MUST put the controller in slot 1. Taking any other course is really asking for trouble... adios From chwe@force.de Mon Aug 5 00:35:46 1996 From: chwe@force.de (Christof Wehner) Date: Mon Aug 5 00:35:49 PDT 1996 Subject: Re: [comp.os.vxworks] 5v Solaris/167 vxWorks > I have a system with a Force 5V running solaris 2.4 and a Motorola > mvme167 running vxWorks 5.2 in a vme chassis. The Force 5v is in slot > one and is the system controller and the mvme167 is not system > controller and is in slot 2. I am using the PROMS and the vxWorks kernel > from the board support package unmodified. > > When I turn on the system the mvme167 loads the vxWorks kernel but the > Force 5V hangs loading solaris. If I stop the mvme167 from loading the > vxWorks kernel until after the Force 5V has finished loading solaris, the > system runs fine. > > I have changed the mvme167 PROMS to wait 150 seconds before loading > the vxWorks kernel. Has anybody had this problem and is there a fix or > a better work around for this. Dear Peter, yes there is a better way around this. Contact your local FORCE COMPUTERS support for the following: this is a known 'problem' (ahem feature) of Solaris! When booting, Solaris will probe the VMEbus for a device called 'mcp' in several positions. If your VxWorks board happens to be in this location on VME, Solaris will try to initialize the mcp device and hang due to a lack of cooperation of your device 8-). To suppress this, add the following to /etc/system: exclude: /drv/mcp exclude: /drv/mcpp exclude: /drv/mcpzsa Now, your system should come up without any problem! BTW: try support@fci.com!! Best regards i.A. Christof Wehner -- Christof Wehner ____ ____ ____ ____ ____ Technical Support Engineer / / / / / / / / Customer Support Department /__ / / /___/ / /__ FORCE COMPUTERS GmbH / / / / \ / / Prof.-Messerschmitt-Str. 1 / /___/ / \ /___/ /___ D-85579 Neubiberg/Muenchen email: chwe@force.de C O M P U T E R S G m b H http://www.forcecomputers.com/ - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - phone +49-89-60814-240 fax +49-89-609 77 93 From health@moneyworld.com Mon Aug 5 14:05:02 1996 From: health@moneyworld.com Date: Mon Aug 5 14:05:04 PDT 1996 Subject: Scientific Discoveries Minimize Aging (DHEA) http://dhea.natureplus.com Take advantage of the amazing benefits of DHEA. In the search for the FOUNTAIN OF YOUTH, DHEA is a must README. "I can't wait 30 More years for the National Institute on Aging to save my ass." says Dr. Regelson, a leading DHEA researcher. Click on: http://dhea.natureplus.com From stan@rti.com Mon Aug 5 14:29:57 1996 From: Stan Schneider Date: Mon Aug 5 14:29:58 PDT 1996 Subject: Re: Printing Activation Records Hi Ramandeep, >> Submitted-by rbakshi@hns.com Thu Jul 25 11:47:51 1996 >> Submitted-by: "Ramandeep Bakshi" >> >> Hi all, >> >> I am trying to debug some memory problems in one of the applications that >> is running on a Intel960. This application has been compiled for VxWorks >> using the gnu960 compiler. What I am required to do is to print out >> all the activation records in memory at a particular time. Anyone have >> any idea where I should start ? Any help will be much appreciated. >> >> Thanks a bunch, I'm not sure if this is what you (really) want, but RTI markets a dynamic execution profiler that will display the statistical history of the activation records of all running tasks. For example, a TCP server might have a profile like this: Profiling task 0x3b20e0: net4 Current sum % direct % 10.43 .12 _TcpWrite 10.31 0.01 . _write 10.27 .18 . . _iosWrite 10.06 .16 . . . _send 9.35 .64 . . . . _sosend 7.92 .27 . . . . . _tcp_usrreq 7.25 .88 . . . . . . _tcp_output 4.97 .49 . . . . . . . _ip_output 3.87 0.06 . . . . . . . . _eiOutput 3.80 .51 . . . . . . . . . _ether_output 2.32 .31 . . . . . . . . . . _eiTxStartup 1.12 1.12 . . . . . . . . . . . _bcopy .69 .69 . . . . . . . _cksum .63 .63 . . . . . _bcopy There's also the venerable "tt" command. It will show you the activation records of a single task at the time the command is typed. RTI also markets a memory trace facility that will show the call stack that was active when memory is allocated. Send email to "info@rti.com" for details. Hope this helps, -- Stan stan@rti.com From stan@rti.com Mon Aug 5 15:47:24 1996 From: Stan Schneider Date: Mon Aug 5 15:47:26 PDT 1996 Subject: Re: NTP availability? >> Date: 30 Jul 1996 18:29:27 GMT >> From: don@mars.dgrc.doc.ca (Donald McLachlan) >> ... >> Unfortunatetely when I try to make the kernel it fails with ... >> ... >> gmtime.o(.text+0): multiple definition of `_gmtime (.text)' >> >> Any idea how I can either remove localtime and gmtime from their >> libraries, or ignore this error so vxworks will build ( and hopefully >> sellect the usrTime versions of these functions) ? This is a fairly common problem when trying to fix bugs in shipped libraries. *** HACK ALERT! *** The quick fix is to patch the symbol so it doesn't clash. To do that, compile the attached program: gcc -Wall -o patrep patrep.c Then patch your library image (say to remove a symbol called "hypot"): patrep hypot Hypot < libMC68040gnuvx.so > libMC68040gnuvx2.so The new symbol can be anything but "hypot" (I used "Hypot" above), but must still be 5 characters. -- Stan /* Patrep \- Pattern replacement program Usage: patrep pattern1 pattern2 < infile > outfile This program will replace pattern1 with pattern2 anywhere it appears in infile. It works on binary as well as text files (as opposed to "sed", which only works on text files). */ #include void main(int argc, char *argv[]) { int c, len, matches; char *pat, *pt; if (argc != 3) { printf("Usage: patrep pattern1 pattern2 < infile > outfile\n"); exit (1); } pat = argv[1]; len = strlen(pat); matches = 0; while((c = getchar()) != EOF) { if (pat[matches] == c) { matches++; if (matches == len) { for (pt = argv[2]; *pt != '\0'; pt++) { putchar(*pt); } matches = 0; } } else { for (pt = argv[1]; matches > 0; pt++, matches--) { putchar(*pt); } putchar(c); } } } From harvey@wrs.com Mon Aug 5 17:35:23 1996 From: Harvey Wong Date: Mon Aug 5 17:35:25 PDT 1996 Subject: multibus drivers or bsps Dear VxWorks Users, Has anyone used VxWorks on Multibus (I, II, or even III) boards based on 80x86 processors? I'm looking for BSP's or device drivers for these boards. Thanks in advance! Harvey harvey@wrs.com From namiq@wrs.com Tue Aug 6 01:06:08 1996 From: namiq@wrs.com Date: Tue Aug 6 01:06:10 PDT 1996 Subject: LonTalk Hello Folks: I am looking for a LonTalk Adapter (ISA or PC104 card) that has a VxWorks driver. Any recommendations appreciated. Thanks in advance, Best Regards, ---------------------------------------------------------------------------- - Namiq Kunimoto Email: namiq@wrs.com Director of Sales Phone: 81-3-5467-5900 Asia Pacific Operations FAX: 81-3-5467-5877 Wind River Systems WWW: http://www.wrs.com Japan Office | POLA Ebisu Bldg. 11F +-------------------------------------------- 3-9-19, Higashi, Shibuya-ku, Tokyo, 150, Japan ---------------------------------------------------------------------------- - From john@vms.masd.northgrum.com Tue Aug 6 13:11:36 1996 From: john@vms.masd.northgrum.com (John Sicklick) Date: Tue Aug 6 13:11:59 PDT 1996 Subject: subscribe vxworks.exploder@vms.masd.northgrum.com subscribe vxworks.exploder@vms.masd.northgrum.com From john@vms.masd.northgrum.com Tue Aug 6 13:12:50 1996 From: john@vms.masd.northgrum.com (John Sicklick) Date: Tue Aug 6 13:13:11 PDT 1996 Subject: test message Pierre, Kiran, A test message to see if csd.nad.northrop.com is still working --slick From daemon@csg.lbl.gov Wed Aug 7 04:01:05 1996 From: daemon@csg.lbl.gov Date: Wed Aug 7 04:01:07 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Wed Aug 7 04:01:01 PDT 1996 Subject: Re: Booting from Windows NT Subject: Malloc Problems in 5.2 Subject: Re: VxWorks vs. OS-9 Subject: Re: NT/pSOS RPC timeout to embedded system Subject: Re: ONC/TLI RPC Compatibility problems? Subject: Re: bootp server for 95/NT ? Subject: Filesystem Shadowing Software, anyone? ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Booting from Windows NT Date: Wed, 31 Jul 1996 21:07:29 GMT From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star Message-ID: <4tohur$ad@idiom.com> References: <4tmmgd$m47@nnrp1.news.primenet.com> Reply-To: hjb@peacefulstar.oakland.ca.us jerome@primenet.com (Douglas Jerome) wrote: >Help! >My vxworks target boots fine from my sparc, but it won't boot from >my Windows NT computer. Actually, vxworks boots but it hangs forever >trying to download the symbol table. One way to figure out what is happening is to use Tornado-supplied FTP daemon (wftpd.exe) instead of NT FTD server. Using wftpd you can figure out what client is doing by turning on "logging" options. the wftpd menu option for logging can be turned on for various things like "get, put, winsock events, ..." You should then be able to figure out what is happening better. HJB --------------------------- Newsgroups: comp.os.vxworks Subject: Malloc Problems in 5.2 Date: Tue, 06 Aug 1996 11:03:27 -0700 From: Grant Haidinyak Organization: Interwave Communications Message-ID: <320788EF.41C67EA6@iwv.com> Hi, We're having some problems with the vxworks malloc in 5.2. The problem is that malloc returns a NULL pointer. The interesting part is that We call MemFindMax() to make sure that there is enough memory before the call. And when malloc() returns a NULL, we shut down the application tasks. At this point MemShow() shows that there is 10x the amount of memory available than was requested. And since the tasks were suspended, no task could allocate memory, then free it. So, does anyone have any ideas why malloc() would return a NULL when there appeared to be plenty of memory available. thanks, grant ps Here's the code fragment that's failing. It's failing at the last check. if (intContext()) { msg = NULL; goto exit; } max = memFindMax(); if (max < iwmalloc_chk_min) { msg = NULL; iwmalloc_no_mem = TRUE; goto exit; } msg = malloc(SIZE_IWMSG + Size + IWMSG_OVERAGE_SIZE); if (msg == NULL)) { /* suspend tasks */ } exit: /* the rest of the story */ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks vs. OS-9 Date: 5 Aug 1996 18:03:23 -0700 From: nigel@idiom.com (Nigel Standing) Organization: A poorly-installed InterNetNews site Message-ID: References: <32027C7E.4A6C@llnl.gov> "Timothy A. Meier" writes: > I believe that OS-9 is less expensive per seat, but in the long run > VxWorks may save money in development/debug time. Another factor which may not be important at the moment, but might be down the road, is the range of CPU architectures supported. VxWorks has a pretty wide coverage, so that in the future if you have to scale your application up/down there's likely to be suitable CPU type supported. BTW I don't think you can underestimate the usefulness of the VxWorks simulator, another VxWorks plus. My 2 cents worth, Nigel - --------------------------- Newsgroups: comp.arch.embedded,comp.os.vxworks Subject: Re: NT/pSOS RPC timeout to embedded system Date: Tue, 6 Aug 1996 08:10:13 GMT From: noblenet@world.std.com (NobleNet Inc.) Organization: NobleNet, Inc. Message-ID: References: In article , Rich Hogle wrote: >I have a high-performance data acq system, the actions of which >are coordinated by software running on a remote WindowsNT system. >The NT host starts an acquisition via a blocking RPC function call, served >by the embedded system (pSOS on PPC1604). The embedded system >performs the acquistion, then issues a return and the NT host continues. > > - We are using Chameleon/NetManage RPC-SDK environment The problem may be in your NT RPC library. Please contact NobleNet for ONC/RPC tools and solutions for UNIX, Windows (3.1/95/NT), Macintosh, VMS, OS/2, NetWare, VxWorks, etc. NobleNet, Inc. 337 Turnpike Road Southboro, MA 01772 Voice +1 508 460 8222 Fax +1 508 460 3456 sales@noblenet.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: ONC/TLI RPC Compatibility problems? Date: Tue, 6 Aug 1996 08:05:24 GMT From: noblenet@world.std.com (NobleNet Inc.) Organization: NobleNet, Inc. Message-ID: References: In article , Steve Vasta wrote: > >Does anyone know of any compatibility problems using TLI RPC client >software (from Solaris 2.4) against ONCRPC servers running under VxWorks >5.2? I'm able to run RPC TLI client/server tests OK if both machines are >Solaris (i.e. both machines use TLI RPC . . . also known as TIRPC), and if >both machines run ONCRPC (such as SunOS4.1.x and VxWorks 5.2). However, I >am getting corrupted data (on the server side within the proper >callback) when I run a TLI client against an ONC server and vice versa. >My questions at this point are: > > 1) Can TLI RPC and ONC RPC be mixed like this? > 2) If not, is there a version of TLI RPC that can be built for > VxWorks 5.2 and/or Sun4 machines? Yes, you should be able to mix TLI/RPC and ONC/RPC. Please contact NobleNet for ONC/RPC tools and solutions for UNIX, Windows (3.1/95/NT), Macintosh, VMS, OS/2, NetWare, VxWorks, etc. NobleNet, Inc. 337 Turnpike Road Southboro, MA 01772 Voice +1 508 460 8222 Fax +1 508 460 3456 sales@noblenet.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: bootp server for 95/NT ? Date: 2 Aug 1996 01:39:51 GMT From: steved@pacifier.com@199.2.117.163 (Steve Doiel) Organization: Pacifier BBS, Vancouver, Wa. ((360) 693-0325) Message-ID: <4trm97$1j7@news.pacifier.com> References: Reply-To: steved@pacifier.com (Steve Doiel) Victor K Saffrin writes: >Is anyone aware of a bootp server available for 95 or NT? I would like to not >have to customize the boot code on all targets that I ship (rather I wouldn't >want to trust manufacturing to do it), and would like to have the target get >it's IP address from the NT workstation it's hooked to. The code in >...\unsupported\bootp2.1 that WRS ships with Tornado is specifically for >Unix systems. >Thanks. >-- >Victor Saffrin >Instrumentation Laboratory, Inc. >113 Hartwell Ave. >Lexington, MA 02173 I'm not sure, but I think I saw such a critter on the NewT CDROM from Walnut Creek. I would suggest looking at ftp.cdrom.com for NT stuff. I was browsing the CDROM (incidently... a good investment if you are working on NT) this afternoon. I recall seeing something about bootp but can't remember the details. I hope this helps. Steve Doiel --------------------------- Newsgroups: comp.arch.embedded,comp.os.vxworks Subject: Filesystem Shadowing Software, anyone? Date: 5 Aug 1996 18:57:50 GMT From: bhaskar@bnr.ca (Shaji Bhaskar) Organization: Bell-Northern Research Ltd. Message-ID: <4u5g7e$34b@brtph500.bnr.ca> Hi, Here is my problem: We have two Power PC cards running the VxWorks embedded OS. Both cards have a file system that uses SCSI drives. The cards are connected by a LAN. One of the cards is intended to be a standby in case the other fails. Therefore, I need to be able to shadow the file systems to be identical (as far as practical) on both cards. I am looking for software that does some or all of the following: 1. Shadows the file system on the master card to the file system on the standby card, transparent to users of the file system. 2. If one of the cards fails, and is later brough back into service, the software will resolve any differences between the failed card's file system and the file system on the running card before resuming mirrioring activity. In other words, the software should handle network partitions and similar failures. I know I could just go out and buy a database that supports replication, but we don't have the $$$. Regards, Shaji Bhaskar --------------------------- End of New-News digest ********************** From han2326@aminet.co.kr Wed Aug 7 08:02:45 1996 From: "Han, Jeong Soo" Date: Wed Aug 7 08:02:48 PDT 1996 Subject: About bootrom Hello, As a beginer in vxWorks, I'd like to ask some question about bootrom. When do I try to re-install the bootrom using rom-writer.(reloading of bootrom.hex file using make command in the bsp of VME CPU board) Due to any modification of bsp files, Sould I re-install the bootrom? I want to know the bsp files to effect to the bootrom. (~/all/configAll.h,... & ~target/config.h, ... etc) Your detail description would be highly appreciated. J.S.Han/HDPIC research center, Seoul, Korea From snordmey@scitexdpi.com Wed Aug 7 11:39:14 1996 From: Steve Nordmeyer Date: Wed Aug 7 11:39:16 PDT 1996 Subject: stop bits, SECS-II I am using VxWorks 5.2 on an x86 platform. I need to communicate with serial devices and want to change the stop bits. I can't find any command which affects the number of stop bits. Any suggestions? Also, I may have to communicate with a device using SECS-II. Does anyone have any information about this? Thanks, Steve From stan@rti.com Wed Aug 7 12:06:55 1996 From: Stan Schneider Date: Wed Aug 7 12:06:57 PDT 1996 Subject: Sales Opportunity! At Real-Time Innovations, Inc. (RTI), we are expanding our direct sales team! We are seeking extremely talented sales people that will help us market some of the most innovative products in the area of real-time systems. This is a once-in-a-lifetime opportunity to join the core sales force at a hot growing company. You will need an exceptional track record and solid communication skills---both written and verbal. ------ RTI is a rapidly-growing Stanford spinoff in the exploding field of real-time tools and architectures. We market an object-oriented software framework, data visualization and performance analysis tools, and a real-time network communication system. We have a strong, on-going research program focused on software architectures in conjunction with Stanford University. RTI is an exciting place that will offer you the freedom and opportunity to excel. Our products are used in all sorts of applications from undersea vehicles to interplanetary robotics, on platforms from embedded microcontrollers to supercomputers! Come help shape the future of real-time... ------ * Please reply via email to "jobs@rti.com" (please include your name in the "subject" line), or fax to the number below. * Non-technical sales personnel are encouraged to apply, but you must have experience selling complex high-tech products. * No contractors or paid recruiters, please. =========================================================================== = = = = Real-Time Innovations, Inc. = Phone: (408) 720-8312 = = 155A Moffett Park Drive, Suite 111 = Fax: (408) 734-5009 = = Sunnyvale, CA 94089 = email: jobs@rti.com = = = = =========================================================================== From claire@wrs.com Wed Aug 7 16:49:07 1996 From: claire@wrs.com (Claire Chaplais) Date: Wed Aug 7 16:49:09 PDT 1996 Subject: Wind River Users Conference, Call for Speakers Wind River Systems Users Conference Call for Speakers We want you! The next Wind River Users Conference & Exhibition will be held on November 4-5, 1996 in Santa Clara, CA. To prepare for this event, we are actively soliciting VxWorks=81 / Tornado users to share ideas and tips, brainstorm tricky development problems and show how Wind River products and third party partner tools can help speed your time-to-market. You, as a user, are a perfect speaker for this event! To participate, just submit a short abstract on design, development, debug tip, trick or experience you would like to share with your peers. We are looking specifically for technical papers relating to a VxWorks or Tornado-based project. E:mail your abstract with your name, company, address and telephone number to wrug@wrs.com or fax it to me @ 510 814 2010. Deadline for submission is September 10, 1996. We'll notify you by September 20, 1996 if your abstr= act is accepted. In order to avoid blatant commercialism, your abstract must be technical in nature and without obvious sales pitches. If you have questions about the appropriateness of your topic or would like to discuss potential presentation ideas, please contact myself (claire@wrs.com) Stan Schneider, conference chair. Stan can be reached at Real-Time Innovations @ 408 720 8312 x104 or stan@rti.com. If you have any other questions about format or general conference requests, please call me @ 510 814 2177. I look forward to working with you! Regards, Claire Chaplais Marketing Coordinator P.S. As a conference speaker, you will have the undivided attention of 60-70 users and several of Wind River's senior staff - what better opportunity to become well known as a leader in the VxWorks community. (It doesn't hurt your resume either!) Wind River Systems Claire Chaplais Marketing Coordinator 1010 Atlantic Avenue Alameda, CA 94501 (510) 814-2177 From daemon@csg.lbl.gov Thu Aug 8 04:00:22 1996 From: daemon@csg.lbl.gov Date: Thu Aug 8 04:00:25 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Thu Aug 8 04:00:17 PDT 1996 Subject: Q: Memory Loss (?) Subject: Re: Purify like tools for VXWORKS?? Subject: LibVGA for VxWorks x86? Subject: Disabling Network ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Q: Memory Loss (?) Date: 6 Aug 1996 17:05:20 GMT From: scott@tcville.es.hac.com Organization: Hughes Aircraft Message-ID: <4u7u0g$sq@hacgate2.hac.com> Reply-To: scott@tcville.es.hac.com Host = AIX Target = RAD6k VME wi 4MB RAM, ethernet, NVRAM card OS = VxWorks 5.1.1 COMPILER = GreenHills Multi LANG = Ada ( I know, I know, ... ) We are seeing only 1/2 the expected memory available after loading our S/W (ELF format) as we would expect after examining the Text+Data blocks as printed via GreenHills' gdump/gnm utilities. We obtained the memory size status via VxWorks' memShow function. I *thought* we were using netDrv() but, after re-examining the config files, it looks like we are using nfsDrv() -- I am rebuilding VxWorks just to make certain. I've looked through all of the VxWorks manuals but haven't seen a nice way to display the memory map as to what objects are loaded where in order to locate where/why we have this memory loss. Is there such a utility ?? Any/all comments & suggestions would be most welcome, Thank you, E. Scott Channell scott@tcville.es.hac.com - --- Hughes Aircraft Co. voice: (310) 616-1059 Image and Signal Processing Lab addr: scott@tcville.es.hac.com PO Box 902, EO/E01/A172, El Segundo, Ca. 90245 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Purify like tools for VXWORKS?? Date: 6 Aug 1996 17:43:56 -0700 From: kla@leland.Stanford.EDU (Earl Mitchell) Organization: Stanford University, CA 94305, USA Message-ID: <4u8osc$5fn@elaine25.Stanford.EDU> References: <32023136.8AA@phbtsus.com> In article <32023136.8AA@phbtsus.com>, Aseem Bakshi wrote: >Hi, >I am looking for tools that perform memory checks similar to the >capabilities that >Purify provides on VXWORKS. Could anybody who knows of or has used such >tools >post back into this group or send me E-mail Company called Real Time Innovations sells tools like this for VxWorks. Haven't actually use their stuff but I've heard its pretty good. Don't know how to contact them but they usually monitor this group to reccomend their products. --------------------------- Newsgroups: comp.os.vxworks Subject: LibVGA for VxWorks x86? Date: Fri, 02 Aug 1996 01:05:30 -0700 From: Nick Triantos Organization: Netcom Message-ID: <3201B6CA.7F7E@ix.netcom.com> Hi everyone, Does anyone know if there are any libraries ported to or written for VxWorks which allow basic VGA graphics on VxWorks for Intel x86 machines? I've looked at libsvga for Linux, but I'm kinda new to VxWorks, so I'm a little unsure that I'll be able to port it quickly. Ideally, I'd love to find a freeware/public domain library. I know about RTGL, but it seems to only support a small number of VME boards. Many thanks, - -Nick Triantos --------------------------- Newsgroups: comp.os.vxworks Subject: Disabling Network Date: 5 Aug 1996 22:58:26 GMT From: Doug Owens Organization: cao Message-ID: <4u5uai$44a@lll-winken.llnl.gov> Greetings: I have a vxWorks 5.0 system that is getting clobbered by network activity. It's a Force CPU-30 running under 5.0. We discovered that activity on the network occasionally used up enough CPU time for it to not be able to keep up with it's real time requirements. This was fixed by disconnecting the network cable (for now). Is there a software fix for this? (i.e a way to turn off the network - and turn it back on if necessary). This particular system only needs the network to boot, but it isn't really stable enough for booting from ROM. I have another system (MV 147) that needs to occasionally access NFS file systems. I would like to be able to turn on/off the network on this system also. Any suggestions? Thanks, Doug Owens (owens2@llnl.gov) --------------------------- End of New-News digest ********************** From lwatson@arcserv.dasd.honeywell.com Thu Aug 8 08:59:08 1996 From: lwatson@arcserv.dasd.honeywell.com (Linda Watson) Date: Thu Aug 8 08:59:10 PDT 1996 Subject: Re: Subject: Q: Memory Loss (?) Scott, On memory loss... We are also using Ada with GreenHills adamulti with vxWorks 5.2 and C with the additional twist of using the 80960 and having HUGE load modules in COFF format. I don't know how much of what we noticed would apply to your situation since you are using ELF and another host. The following covers what we have observed. If we have missed something, we would appreciate the advice. Are you loading through GreenHills or through the vxShell command line? In our version, adamulti always uses a netDrv to load. Due to other problems, we always load from vxShell and then connect up the adamulti debugger. We noticed these differences for loading via netDrv versus nfsMount: netDrv - requires about three times the size of the entire COFF file during loading, but once the module is loaded, it takes up less memory than a module loaded with nfsMount nfsMount - requires about two times the size of the entire COFF file during loading, but once the module is loaded, it consistently uses up more memory than the same module loaded with netDrv. We reported this problem to Wind River and corresponded for awhile, but due to more pressing issues, haven't had time to send them more load modules, examples and statistics. We ended up dividing our system into submodules, and alternating linking them directly into vxworks (which means no debug information for the adamulti modules) or building with mainly no debug and a few modules with debug on, dyanmically loading, and using assembly level debugging just to get our modules into the system. I don't know how the ELF loader works, but the COFF loader (in addition to memory required for either nfsMount or netDrv loads), reads ENTIRE sections of the COFF load module into memory. This can cause problems if you have large load modules at half the size of total RAM or more. It would be nice if the COFF loader would work on small portions of the load module at a time. Two of the things they did find: o It appeared that the nfsMount load did not release all of the memory it had allocated for loading once it was done loading (or at least not right away). The amount of memory appeared to be negligable for the modules they tried, but I think it is actually a function of the size of the module. In any case, I don't think they found the cause. o It appeared that if a load error was encounted, not all of the memory was released back to the system that was allocated for loading. I think they are supposed to fix this in future releases but not for our version, 5.2. memShow: memShow(1) displays free segments available as well as the totals. moduleShow displays where the module segments were loaded, and siz960 or gdump can be used to determine the size of these segments. However, neither of these show the amount of memory allocated for the symbol table and other uses, and where. It would be extremely useful to have a more detailed memory map. One suggestion Wind River passed on that definitely helps reduce the amount of memory required to load a module, is to do the nfsMount load using only global symbols instead of all (global and local). adamulti puts alot more symbols into the coff load module than those produced with the gnu tools. ld 0,0,"/nfsmountdir/module.o" One more issue to watch: The symbol table generator does not include symbols from COFF sections which are not labeled "text", "data" or "bss". This is true for both makeSymTbl which generates the symbol table for vxworks and modules linked with it, as well as the dynamic linker loader. With COFF modules, once the number of relocatable entries exceeds 64K, another section is created (text0, text1, data0...). We encountered this with several of our load modules generated by adamulti, and we either need to link these in with vxWorks or load them dynamically. If later modules are loaded dynamically, they may not find all the symbols/functions they need if they were in those other sections. This also makes debugging more difficult, since you have to rely on maps and gdumps to determine addresses of routines for tasks to spawn and variables to examine. This has been reported to Wind River. and another issue: We are using a SLIP connection for all communication with the target. We don't know if this is causing this particular problem, but we have seen bad data loaded for some of the text sections in our large load modules. (Actual code in the load module, but strange values after it has been loaded into memory.) This was found by accident while examining some of our routines after receiving faults. We don't see this problem with modules which are linked with vxWorks directly and loaded using our load software since we verify every value that has been loaded. It might help if this was also done in the dynamic linker/loader. I don't think moduleCheck can help since it doesn't pass back the checksum value and it only checks what is already in memory, not against the module loaded. We haven't reported this problem yet. I hope some of this helps, and we would appreciate it if you could also request help with the loading software and the memory show facilities, or let us know if you have any suggestions. Thanks, Linda Watson lwatson@dasd.honeywell.com From anthonyl@atlantis.softeng.lamrc.com Thu Aug 8 09:45:16 1996 From: Anthony Le Date: Thu Aug 8 09:45:18 PDT 1996 Subject: subcribe subscribe Anthony Le Sr. Software Engineer LAM RESEARCH CORP. anthony.le@lamrc.com From seaweed!seaweed.com!bob@netcom.com Thu Aug 8 09:48:33 1996 From: bob schulman Date: Thu Aug 8 09:48:34 PDT 1996 Subject: Re: comp.os.vxworks newsdigest > We are seeing only 1/2 the expected memory available after loading > our S/W (ELF format) as we would expect after examining the Text+Data > blocks as printed via GreenHills' gdump/gnm utilities. We obtained the > memory size status via VxWorks' memShow function. When ELF-format object files are "ld"-d by vxWorks, a large amount of system memory is taken up by what appears to be relocation records. Supposedly, one can eliminate those records from the object file by use of "strip-sde -g", but strip-sde is buggy and doesn't work as advertised. Attempts to use ld-sde to reduce object size were also (surprisingly) unsuccessful. In sum, I think you're stuck. Perhaps a new release of the tool chain will fix these deficiencies. bob From anthony.le@lamrc.com Thu Aug 8 09:52:30 1996 From: Lam Research Date: Thu Aug 8 09:52:32 PDT 1996 Subject: subscribe subscribe From stan@rti.com Thu Aug 8 11:21:43 1996 From: Stan Schneider Date: Thu Aug 8 11:21:45 PDT 1996 Subject: Re: Q: Memory Loss (?) >> Date: 6 Aug 1996 17:05:20 GMT >> From: scott@tcville.es.hac.com >> I've looked through all of the VxWorks manuals but haven't seen a nice >> way to display the memory map as to what objects are loaded where in order >> to locate where/why we have this memory loss. Is there such a utility ?? >> Date: 6 Aug 1996 17:43:56 -0700 >> From: kla@leland.Stanford.EDU (Earl Mitchell) >> Company called Real Time Innovations sells tools like this for >> VxWorks. Haven't actually use their stuff but I've heard its >> pretty good. Don't know how to contact them but they usually monitor >> this group to reccomend their products. Interesting confluence of postings... We do indeed have a utility that will display where your memory is being used. The "memory loss" above sounds like it may be symbol-table usage, but HeapTrace would show you for sure. Contact "info@rti.com" for details. -- Stan stan@rti.com From rmoore@gic.gi.com Thu Aug 8 11:52:53 1996 From: "Moore, Rich (HT-MS)" Date: Thu Aug 8 11:52:55 PDT 1996 Subject: Prevention if Broadcast Loopback Hello VxWorkers, We are using VxWorks 5.2 for an application that requires the transmission of multicast ethernet frames carrying UDP datagrams. Unfortunately VxWorks 5.2 doesn't support multicast addressing from the IP stack. So we are using an etherOutputHook function to substitue a multicast MAC address for the broadcast address if certain criteria are met. The application is sending UDP datagrams to the broadcast IP address, which is necessary to force the receiving IP stacks to deliver the datagrams to the application. This approach works. However, the drawback to this approach is that the frame is being transmitted as a broadcast from the perspective of the protocol stack and socket layer. In typical BSD implementations, a broadcasted frame is actually copied to the receive queue of the protocol stack (internal loopback) before sending it out to the network. This is because the definition of a broadcast is to sent to "all hosts on the LAN including ourselves." This internal loopback copy has performance ramifications on our applicaion. We don't own the VxWorks source code, so we can't modify the stack. Does anybody have any ideas how to prevent the internal loopback of broadcast ethernet packets? Thanks in advance, Rich Moore General Instrument Corp. Hatboro, PA USA rmoore@gic.gic.com From leonid@rst.co.il Thu Aug 8 23:20:45 1996 From: leonid@rst.co.il (Leonid Rosenboim) Date: Thu Aug 8 23:20:47 PDT 1996 Subject: Re: Disabling Network On 5 Aug 1996 22:58:26 Doug Owens Wrote: > Greetings: > I have a vxWorks 5.0 system that is getting clobbered by network activity. > It's a Force CPU-30 running under 5.0. We discovered that activity on the > network occasionally used up enough CPU time for it to not be able to > keep up with it's real time requirements. VxWorks designers have thought about such a situation. All you need to do, is make sure that the tNetTask task is set at a priority lower (numerically hugher) then the rest of your application tasks which are on the critical path. By default its priority is 50, but it is easy to change from the Shell or a startup script or from the application with taskPrioritySet(). This will put almost all network related processing at such a priority that it will be executed only when the real-time tasks have nothing to do. Still, there are network interrupts that will steal some CPU cycles, but that is usually not significant. But if you are feeling that still too much cycles are stolen by the Interrupt Service Reoutine of the network driver, you can do sysLanIntDisable() to disable LANCE interrupts for a while. I doubt however tht this will be necesary. ----------------------------------------------------------------------- Leonid Rosenboim Phone: +972-3-559-8144 R S T Software Industries Ltd. Mobile: +972-50-307-142 P.O.Box 11502, AZUR 58017, Israel Fax: +972-3-559-8244 http://knight.trendline.co.il/~leonid/ E-Mail: leonid@rst.co.il From daemon@csg.lbl.gov Fri Aug 9 04:00:28 1996 From: daemon@csg.lbl.gov Date: Fri Aug 9 04:00:31 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Fri Aug 9 04:00:24 PDT 1996 Subject: Ada and VxWorks Subject: Re: Ada and VxWorks Subject: Pattern Matching Subject: Re: stop bits, SECS-II Subject: Re: Disabling Network Subject: Re: Malloc Problems in 5.2 Subject: Re: Ada and VxWorks ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Ada and VxWorks Date: 8 Aug 1996 11:23:25 GMT From: tk@ Organization: Lockheed Martin Federal Systems Message-ID: <4ucind$5ft@lfsserv1.fsc.ibm.com> Reply-To: tom.kodey@lmco.com I'm considering VxWorks as the RTOS for an Ada based imbedded application and am considering various development environments. The target will likely be a Pentium, PowerPC or R4xxx. I'm interested to hear what Ada development environment you were using and, briefly, what you thought of it (host, ease of interfacing with VxWorks, ease of use, maturity, support, host and target based debug, etc.)? Tom Kodey tom.kodey@lmco.com Lockheed Martin Federal Systems --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Ada and VxWorks Date: 8 Aug 1996 12:46:28 GMT From: mfb@mbunix.mitre.org (Michael F Brenner) Organization: The MITRE Corporation, Bedford Mass. Message-ID: <4ucnj4$g7m@linus.mitre.org> References: <4ucind$5ft@lfsserv1.fsc.ibm.com> I too am interested in a Pentium Ada Vxworks solution for a product that used to be on a 486 using PharLap. Does anyone have experience converting from Pharlap to Vxworks? Does anyone have information on multiple serial interfaces (say, a dozen or so Rs422's?). mikeb@mitre.org --------------------------- Newsgroups: comp.os.vxworks Subject: Pattern Matching Date: Thu, 8 Aug 1996 13:30:36 GMT From: mfisher@dyn.com (Mark Fisher) Organization: The World @ Software Tool & Die Message-ID: Reply-To: mfisher@dyn.com Sender: news@world.std.com (Mr Usenet Himself) Hi, Does anyone know of a text pattern matching utility, like awk, that can be included in an application running under vxWorks. Any suggestions welcomed! Thanks, Mark Fisher --------------------------- Newsgroups: comp.os.vxworks Subject: Re: stop bits, SECS-II Date: Thu, 08 Aug 1996 20:18:17 +0200 From: John Neumoege Organization: Humboldt Universitaet zu Berlin Message-ID: <320A2F69.4AB2@ipk.fhg.de> References: Steve Nordmeyer wrote: > > I am using VxWorks 5.2 on an x86 platform. I need to communicate with > serial devices and want to change the stop bits. I can't find any command > which affects the number of stop bits. Any suggestions? > As far as I know there are not any commands to change the number of stop bits. However, you can directly write to the appropoate register of the SIO chip. The serial driver (i8250Serial.c) uses the following code to change the appropiate register: sysOutByte (pTyCoDv->lcr, 0x03); The lcr (Line Control Register) controls the following values: Bits 0 and 1: specify the number of data bits as follows: 0 0 5 data bits 0 1 6 data bits 1 0 7 data bits 1 1 8 data bits Bit 2: specifies the number of STOPBITS! 0 for 1 stop bit 1 for 1 1/2 stop bits, only with 5 data bits 1 for 2 stop bits for 6, 7, or 8 data bits Bit 3: Parity enable bit. 0 no parity 1 parity on Bit 4: Even parity enable bit (only when bit 3 is set) 0 odd parity 1 even parity The other bits are probably not important for your application, so just set them to zero. So the code from the serial driver with 0x03 means: 8 data bits, 1 stop bit and no parity. pTyCoDv->lcr is a pointer to the lcr port address which is equal to the base address of the serial channel plus the offset of 3. For example, COM1 generally has the base port address 0x3f8, so the lcr of COM1 can be addressed with 0x3fb thus sysOutByte (0x3fb, yourLcrValue); By the way, the prototype for sysOutByte can be found in sysLib.h. Regards, - --John --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Disabling Network Date: 6 Aug 1996 13:31:56 GMT From: johncoby@ghgcorp.com (JOHN R. COBARRUVIAS) Organization: Personal Account Message-ID: References: <4u5uai$44a@lll-winken.llnl.gov> In article <4u5uai$44a@lll-winken.llnl.gov>, Doug Owens wrote: Hum.....we had the same problems. Turned out the boards were generating interrupts that had to be serviced. We would isolate the network in the lab and got better, but not reliable results. Then we isolated our 2 boxes. Better, but not reliable. We still had a task that would jump in and take over the CPU while we missed schedules. So...... (We were trying to use Ethernet to send data to a display device (SUN) while a real time application was running in the target. When we told the vendor of the problem, they said "Don't use Ethernet in a realtime system." After a few months, I agreed.) You can try suspending the netTask task using taskSuspend(). See if that works. Or deleting the task with taskDelete. Then restart it using netLibInit. See Also netLib, usrConfig, netTask() > Greetings: > I have a vxWorks 5.0 system that is getting clobbered by network activity. > It's a Force CPU-30 running under 5.0. We discovered that activity on the > network occasionally used up enough CPU time for it to not be able to > keep up with it's > real time requirements. This was fixed by disconnecting the network > cable > (for now). Is there a software fix for this? (i.e a way to turn off the > network - and turn it back on if necessary). This particular system only > needs the network to boot, but it isn't really stable enough for booting > from ROM. I have another system (MV 147) that needs to occasionally > access > NFS file systems. I would like to be able to turn on/off the network on > this system also. > > Any suggestions? > Thanks, > Doug Owens (owens2@llnl.gov) - -- ,-_/ . .-,--. ,---. . ' | ,-. |-. ,-. `|__/ | -' ,-. |-. ,-. ,-. ,-. . . . , . ,-. ,-. | | | | | | | | \ ,. | -. | | | | ,-| | | | | | / | ,-| `-. | `-' ' ' ' ' `-' ` `' `---' `-' `-' `-^ ' ' `-' `' ' `-^ `-' / | John R. Cobarruvias johncoby@ghgcorp.com Houston, Tx `--' --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Malloc Problems in 5.2 Date: Thu, 08 Aug 1996 14:50:26 -0700 From: Pankaj Malhotra Organization: Intel Corp Message-ID: <320A6122.3DC0@ccm.jf.intel.com> References: <320788EF.41C67EA6@iwv.com> Grant Haidinyak wrote: > > Hi, > > We're having some problems with the vxworks malloc in 5.2. > > The problem is that malloc returns a NULL pointer. The interesting part > is that We call MemFindMax() to make sure that there is enough memory > before the call. And when malloc() returns a NULL, we shut down the > application tasks. At this point MemShow() shows that there is 10x the > amount of memory available than was requested. And since the tasks were > suspended, no task could allocate memory, then free it. > > So, does anyone have any ideas why malloc() would return a NULL when > there appeared to be plenty of memory available. > > thanks, > > grant > > ps Here's the code fragment that's failing. It's failing at the last > check. > > if (intContext()) { > msg = NULL; > goto exit; > } > > max = memFindMax(); > if (max < iwmalloc_chk_min) { > msg = NULL; > iwmalloc_no_mem = TRUE; > goto exit; > } > > msg = malloc(SIZE_IWMSG + Size + IWMSG_OVERAGE_SIZE); > if (msg == NULL)) { > /* suspend tasks */ > } > > exit: > /* the rest of the story */ Grant: What memShow() shows you is the total memory prsent in the system. Unfortunately in your case the memory must be really fragmented in a number of blocks so that when memShow() shows you enough memory available but there is no single block with this much space. Try using memShow 1 to see the fragmented blocks. Pankaj - ------- Pankaj Malhotra (Work) : 503-264-7746 Network Programming Division (Home) : 503-690-9589 Intel Corp (Mail) : PankajX_Malhotra@ccm.jf.intel.com HillsBoro OR --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Ada and VxWorks Date: Thu, 08 Aug 1996 11:37:08 -0700 From: Michael Levasseur Organization: GDE Systems Inc. Message-ID: <320A33D4.41C67EA6@gdesystems.com> References: <4ucind$5ft@lfsserv1.fsc.ibm.com> We are using VxWorks on the Heurikon Baja R4700 platform with Rational VADSWorks. This is a good mix for our needs. We are currently porting and upgrading an application that was about 1 to 2 MSLOCs. We are porting 68K code to MIPS which presents a specific set of problems. These are mostly related to chapter 13;i.e., rep specs. This environment works for us. There are some questions that should be asked - Am I upgrading code? If so using the planned upgrade path is probably better. Upgrading from 68030 to 68060 propably will less painful. Do I need mixed languages? With Rational there is only Ada and you can use gcc modules, but you can't interractively debug the C. Will your hardware be supported later? Is the platform being widely used or are there only a few people using it. Is the company that builds your hardware reputable and is likely to be around ten years down the road. If anyone has any other questions that need to be considered please post them. --------------------------- End of New-News digest ********************** From prb@aplexus.jhuapl.edu Fri Aug 9 06:44:30 1996 From: prb@aplexus.jhuapl.edu (Paul R. Bade) Date: Fri Aug 9 06:44:32 PDT 1996 Subject: Network Activity > Greetings: > I have a vxWorks 5.0 system that is getting clobbered by network activity. > It's a Force CPU-30 running under 5.0. We discovered that activity on the > network occasionally used up enough CPU time for it to not be able to > keep up with it's > real time requirements. This was fixed by disconnecting the network > cable > (for now). Is there a software fix for this? (i.e a way to turn off the > network - and turn it back on if necessary). This particular system only > needs the network to boot, but it isn't really stable enough for booting > from ROM. I have another system (MV 147) that needs to occasionally > access > NFS file systems. I would like to be able to turn on/off the network on > this system also. > > Any suggestions? > Thanks, > Doug Owens (owens2@llnl.gov) > You may want to try to turn off the interface with: ifFlagChange (ifName, IFF_UP, FALSE); /* turn off the interface */ +====================================================================+ | __ ____ __ __ | | /\ \ / \ \ /\_\__ /\ \ Johns Hopkins University | | / \_\ | /\ \ \ / / /\_\\ \ \ Applied Physics Lab. | | / /\ | |\ \/ \ \ / / / / / \ \ \ | | / \/ |_| \ /\ \_\ / / / / / \ \ \ Paul R. Bade | | / /\__/_/ \ \ \/_// / / / / / / / (301)-953-6000 x8681 | | / / / \ \_\ \ \/ / / / / / prb@aplexus.jhuapl.edu | | \/_/ \/_/ \__/_/ \/_/ | | __ ____ __ __ | | /\ \ / \ \ /\ \ /\ \ | | / \_\ | /\ \ \ / \_\ / \ \ | | / /\ | |\ \/ \ \ / /\ |_| / /\ \ \ | | / \/ |_| \ /\ \_\ / / / | | \/ / \ \ | | / /\ / / \ \ \/_// / / / / / /\ \_\ | | \ \/ / / \ \_\ \ \/ / / \/ / / / | | \__/_/ \/_/ \__/_/ /_/_/ | | | +====================================================================+ Acct # 255 Site ID: 159 We now have a 3 year site license # 7513 PC# 707115 Previous Support IDs: 1726, 1944, 2031 Host: Sun Sparc Targets: MV167, MV197, PowerPC From sridhar@prc.allied.com Fri Aug 9 11:12:41 1996 From: sridhar@prc.allied.com (Sridhar Adumusilli) Date: Fri Aug 9 11:12:43 PDT 1996 Subject: Initialized data variables available to power-up module? Hi, In the project (Vxworks on CPU32 target processor, and sloaris host) I am working on, I am using C++ objects to encapsulate device data. For the power-up module to use these objects, I have specifed some public functions as static member functions. These public functions also use static const data members that contain device register addresses. My problem is, I am not sure if the initialized data segment has the static data members (of the device class) already initialized by the time power-up tries to intialize the device. If the compiler generates the code to initialze static data, then power-up can't use these static data members to read/write the device registers. I am using gnu c++ compiler for the target. I don't have target prototype yet. Any suggestions on how can I inspect the executable on host or how can I make sure that the data segment contains values for power-up use? Thanks Sridhar sridhar@prc.allied.com From mail@ppgsoft.com Fri Aug 9 11:44:21 1996 From: mail@ppgsoft.com Date: Fri Aug 9 11:44:22 PDT 1996 Subject: Software Announcement Dear the: Parallel Performance Group, of Sedona, AZ, announces a Series of FREE E-Mail Newsletters on high-tech software topics. Below is a list of the topics discussed in the Newsletter Series: Business Process Reengineering Real-Time Multiprocessing/Uniprocessing DSP Multiprocessing Neural Networks/Fuzzy Logic/Fuzzy Control Parallel Processing Animation/Solid Modelling Distributed Processing Mechanical Design Automation/FEA Object-Oriented CASE Tools CAD for Industrial and Architectural Design Project Management Discrete and Continuous Simulation Image Processing Medical Imaging Graphical Application Interfaces Virtual Reality Statistical Analysis Volume Visualization Engineering Data Management Data Warehousing/Data Mining Electronic Publishing Geographic Information Systems To receive more information, send any e-mail to info@ppgsoft.com; this will return a detailed listing of the Newsletters, and instructions on how to re- ceive sample copies and subscriptions. Visit Our Web site at http://www.ppgsoft.com/ on high-tech software products, presented with pleasing professional graphics. From froeber@bbn.com Fri Aug 9 13:17:22 1996 From: Fred Roeber Date: Fri Aug 9 13:17:24 PDT 1996 Subject: Re: Disabling Network On 5 Aug 1996 22:58:26 Doug Owens Wrote: > > I have a vxWorks 5.0 system that is getting clobbered by network activity. > > It's a Force CPU-30 running under 5.0. We discovered that activity on the > > network occasionally used up enough CPU time for it to not be able to > > keep up with it's real time requirements. Leonid Rosenboim replied: > VxWorks designers have thought about such a situation. All you need to do, > is make sure that the tNetTask task is set at a priority lower > (numerically hugher) then the rest of your application tasks which are > on the critical path. By default its priority is 50, but it is > easy to change from the Shell or a startup script or from the > application with taskPrioritySet(). > > This will put almost all network related processing at such a priority > that it will be executed only when the real-time tasks have nothing > to do. I agree with Leonid that setting netTask to be lower priority than any of your real time applications usually does the trick. The other thing to consider, though, is putting your VxWorks targets on a separate network than the rest of your development systems. We always do this to isolate the targets from random network traffic including things like routing information broadcasts, ARP requests and random other broadcast messages. Set up one of your workstations to have two network interfaces (one to the target network and one to your main network). The workstation software usually can handle the routing required to get from one network to the other so that, for instance, you can still get to the targets from other systems besides the one that has the direct ethernet connection. Since you would end up running multiple network segments, there are some addressing issues you would need to resolve with respect to using either different subnets or different networks. Still, putting your targets on their own network is usually not too hard and quite useful to minimize network overhead. Fred -- | Fred J Roeber, BBN Systems & Technologies | | 50 Enterprise Place Middletown, RI 02842-5202 | | froeber@bbn.com 401-849-2543 (X48) | | See http://www.bbn.com/tv for TraceMaker info | From froeber@bbn.com Fri Aug 9 13:37:17 1996 From: Fred Roeber Date: Fri Aug 9 13:37:19 PDT 1996 Subject: Re: Prevention if Broadcast Loopback Rich Moore wrote: > We are using VxWorks 5.2 for an application that requires the transmission > of multicast ethernet frames carrying UDP datagrams. Unfortunately VxWorks > 5.2 doesn't support multicast addressing from the IP stack. So we are using > an etherOutputHook function to substitue a multicast MAC address for the > broadcast address if certain criteria are met. The application is sending > UDP datagrams to the broadcast IP address, which is necessary to force the > receiving IP stacks to deliver the datagrams to the application. This > approach works. > > However, the drawback to this approach is that the frame is being > transmitted as a broadcast from the perspective of the protocol stack and > socket layer. > > In typical BSD implementations, a broadcasted frame is actually copied to > the receive queue of the protocol stack (internal loopback) before sending > it out to the network. This is because the definition of a broadcast is to > sent to "all hosts on the LAN including ourselves." > > This internal loopback copy has performance ramifications on our applicaion. > We don't own the VxWorks source code, so we can't modify the stack. Does > anybody have any ideas how to prevent the internal loopback of broadcast > ethernet packets? The way you are doing the multicasts, I don't know of any way to prevent the loopback message. You could, however, build your own ethernet packet with the data in it and call etherOutput to send the message directly. To do this, you need to build your own ethernet, IP and UDP headers to wrap around the data. This isn't trivial but it isn't too hard since all the header structures are defined in VxWorks include files and there are good descriptions of the packet layout in books like UNIX Network Programming by Stevens. This method not only works but can end up being much faster since you avoid the overhead of the protocol stack (which can be significant). Fred -- | Fred J Roeber, BBN Systems & Technologies | | 50 Enterprise Place Middletown, RI 02842-5202 | | froeber@bbn.com 401-849-2543 (X48) | | See http://www.bbn.com/tv for TraceMaker info | From jjoly@glenayre.com Fri Aug 9 13:48:58 1996 From: jjoly@glenayre.com (J Joly []) Date: Fri Aug 9 13:49:00 PDT 1996 Subject: Re: Software Announcement From daemon@csg.lbl.gov Sat Aug 10 04:00:40 1996 From: daemon@csg.lbl.gov Date: Sat Aug 10 04:00:44 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sat Aug 10 04:00:37 PDT 1996 Subject: Finding amount of time left on a watchdoc Subject: A32 space VME access Subject: Digital VME Products Feedback Subject: Re: Prevention if Broadcast Loopback Subject: Re: LonTalk Subject: Re: Purify like tools for VXWORKS?? Subject: UNIX - VxWorks POSITION -- SEATTLE AREA Subject: Re: Disabling Network ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Finding amount of time left on a watchdoc Date: Fri, 9 Aug 1996 11:43:26 -0700 From: Andy Van Pelt Organization: QUALCOMM, Incorporated; San Diego, CA, USA Message-ID: How do I programmatically find out how much time is left on a watchdog? There doesn't seem to be such a routine, but wdShow displays it, so it must be available *somehow*. - -- Andy Van Pelt Internet: avanpelt@qualcomm.com QUALCOMM Inc. Work : 619) 658-3214 --------------------------- Newsgroups: comp.os.vxworks Subject: A32 space VME access Date: Fri, 09 Aug 1996 10:23:51 -0400 From: "Michael J. Dempsey" Organization: Lockheed Martin Message-ID: <320B49F7.2781E494@motown.lmco.com> Hello We have a PowerPC (MV1603) running with a beta version of VxWorks 5.3 embedded in VME chassis that contains two other VME boards with memory registers that we would like our VxWorks application to read and write to. One of the boards uses A16 address space and has a base address of 0x8000. The other uses A32 address space and has a base address of 0x00300000. For A16 space board we used sysBusToLocalAdrs to map its memory to something our PowerPC VxWorks application could use. It returned (if memory serves me correctly) 0xefff8xxx (where xxx is an offset from the base address). This worked fine, we read from that board's memory and also wrote to it. We tried using the same call to access memory on the board with A32 address space (changing the first parameter for sysBusLocalAdrs from VM_SUP_SHORT_IO to VM_EXT_SUP_DATA to indicate the different address space). For an input of 0x00300000 to sysBusLocalAdrs we get returned and address of 0xd0300000, but when the program tried to access that memory location it blew up. We have a bus analyzer in the VME rack set to trigger on anything but it never did, so the application seems to have blown up before even trying to access the VME memory.Is there something I need to set up or include in the kernel to access this off-board memory? Any help is appreciated. Thanx, Mike D. dempsey@motown.lmco.com --------------------------- Newsgroups: comp.os.vxworks Subject: Digital VME Products Feedback Date: Fri, 09 Aug 1996 17:16:57 -0700 From: jeff Organization: InfiNet Message-ID: <320BD4F9.386B@infi.net> I have a client that is considering migrating to the Digital VME 4/288 and I wanted to see if anyone has had experience using it in any applications. Mainly, what I am looking for is: 1 - have you have seen improvement in your performance from previous processors (either DEC or Non-DEC) 2- how is the software support? 3 -how is the technical support? Also, any comments on the AXP-VME 230? Thanks! Jeff Haigler Pen-Tech jhaigler@infi.net --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Prevention if Broadcast Loopback Date: Fri, 09 Aug 1996 03:11:55 GMT From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star Message-ID: <4uea9q$g9g@idiom.com> References: <320A6253@gicpo.gic.gi.com> Reply-To: hjb@peacefulstar.oakland.ca.us "Moore, Rich (HT-MS)" wrote: >This internal loopback copy has performance ramifications on our applicaion. > We don't own the VxWorks source code, so we can't modify the stack. Does >anybody have any ideas how to prevent the internal loopback of broadcast >ethernet packets? try setting global variable useloopback to 0 before sending the packet. HJB --------------------------- Newsgroups: comp.os.vxworks Subject: Re: LonTalk Date: Fri, 09 Aug 1996 03:15:57 GMT From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star Message-ID: <4ueahc$g9g@idiom.com> References: <199608060805.AA01984@mail.wrs.com> <320A08DA.213@sandiegoca.ncr.com> Reply-To: hjb@peacefulstar.oakland.ca.us greg bruno wrote: >namiq@wrs.com wrote: >> >> I am looking for a LonTalk Adapter (ISA or PC104 card) that has a VxWorks >> driver. >> >> Any recommendations appreciated. >give the folks at engenuity systems a call: I don't know about ISA or PC104, but I've adapted Engenuity driver code for their VME LonWorks adapter. Their driver is strictly a device driver -- at least the one I used was. A lot of work had to be done to port LonWorks protocol stack on top of this driver in order to actually do anything useful over LonWorks protocol. There are some consulting companies like qLon that may be able to help. HJB --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Purify like tools for VXWORKS?? Date: Mon, 05 Aug 1996 15:29:04 -0500 From: Debbie Buscher Organization: NASA/JSC Message-ID: <32065990.2381@gp902.jsc.nasa.gov> References: <32023136.8AA@phbtsus.com> Aseem Bakshi wrote: > > Hi, > I am looking for tools that perform memory checks similar to the > capabilities that > Purify provides on VXWORKS. Could anybody who knows of or has used such > tools > post back into this group or send me E-mail > Thanx in advance > Aseem > -- > ----------------------------------------------------------------------------- > Aseem Bakshi Phone: (801)9771574 > Philips Broadcast Television Systems Inc Alias: Polka > Boy > Salt Lake City, UT > E-mail Abakshi@phbtsus.com The only tool that I have found to do memory checks for VxWorks is RTILib, which is made by Real-Time Innovations, Inc. out of Sunnyvale, CA. Debbie Buscher --------------------------- Newsgroups: comp.os.vxworks,comp.realtime,comp.arch.embedded,comp.sys.powerpc.tech,seattle.jobs.offered,ba.jobs.offered,comp.unix.programmer,comp.sys.hp.hpux Subject: UNIX - VxWorks POSITION -- SEATTLE AREA Date: Sat, 10 Aug 1996 01:23:39 GMT From: MikeMcQuade@connect.com (Mike McQuade) Message-ID: <4ugnpo$7fl@wisdom.connectsoft.net> SENIOR PROGRAMMER A 27 year old international electronic manufacturing firm in the Pacific Northwest is seeking a Senior Programmer to perform software design under a UNIX based operating system using HP 700 series computers. Duties will include prototyping new software products, supporting existing software products, interfacing w/technical writer in creation of operating guides and interfacing with customers in troubleshooting system problems. Potential for providing field support in software design and maintenance. Desired experience includes 3 years programming experience, programming in C under a HPUX environment, user interface design implementing X-Windows and Motif, knowledge of HPUX system administration concepts, experience writing embedded software for the PowerPC and experience using Wind Rivers VxWorks for embedded system design. Company environment supports individual responsibility of projects from beginning to end. Employees share in growth opportunities and are encouraged to pursue fullest potential. Competitive salary, DOE. Exceptional benefits package including medical, profit sharing, life insurance, retirement plan, educational stipends, etc. Equal Opportunity Employer. If you meet the above requirements, send resume with cover letter to: ATTN: Human Resources Zetec, Inc. PO Box 140 Issaquah WA 98027-0140 Or apply directly at 1370 NW Mall Street, Issaquah, WA. EMAIL to pelwell@zetec.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Disabling Network Date: Tue, 06 Aug 1996 02:40:42 GMT From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star Message-ID: <4u6bc5$d1l@idiom.com> References: <4u5uai$44a@lll-winken.llnl.gov> Reply-To: hjb@peacefulstar.oakland.ca.us Doug Owens wrote: >Greetings: >I have a vxWorks 5.0 system that is getting clobbered by network activity. >It's a Force CPU-30 running under 5.0. We discovered that activity on the >network occasionally used up enough CPU time for it to not be able to >keep up with it's >real time requirements. This was fixed by disconnecting the network >cable >(for now). Is there a software fix for this? (i.e a way to turn off the >network - and turn it back on if necessary). This particular system only >needs the network to boot, but it isn't really stable enough for booting >from ROM. I have another system (MV 147) that needs to occasionally >access >NFS file systems. I would like to be able to turn on/off the network on >this system also. You can try - -> xxx = ifunit("xx0") - -> if_down(xxx) HJB --------------------------- End of New-News digest ********************** From 100656.1507@compuserve.com Sat Aug 10 05:09:08 1996 From: John Shaw <100656.1507@compuserve.com> Date: Sat Aug 10 05:09:10 PDT 1996 Subject: Re: Pattern Matching Mark Fisher wrote >Does anyone know of a text pattern matching utility, like awk, that >can be included in an application running under vxWorks. Any >suggestions welcomed! I can't think of any VxWorks utility that can be run. But, if the host is running Unix and VxWorks networking has been enabled, you could use rcmd(2) on port 514 in order to run awk remotely on the host from VxWorks. A socket descriptor is returned that can be used as both stdin and stdout. John Shaw, Shade (Computer Services) Limited, Calne, Wiltshire, England. 100656.1507@compuserve.com From daemon@csg.lbl.gov Mon Aug 12 04:00:31 1996 From: daemon@csg.lbl.gov Date: Mon Aug 12 04:00:36 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Mon Aug 12 04:00:27 PDT 1996 Subject: Re: interrupt handler question Subject: MC68060 Bug and Fix Subject: Trapping VME bus errors. Subject: Re: Ada and VxWorks ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: interrupt handler question Date: 11 Aug 1996 18:02:15 -0400 From: thesharps@aol.com (THE SHARPS) Organization: America Online, Inc. (1-800-827-6364) Message-ID: <4ull97$m7d@newsbf02.news.aol.com> References: <31FE6073.41C6@fnal.gov> Reply-To: thesharps@aol.com (THE SHARPS) Sender: root@newsbf02.news.aol.com Jeff, I have recently performed the same tasks using the NewBridge Unviverse chip on Radstone 604s and 603es. In doing so, I have performed the following interrupt initialization steps which worked fine for me: intDisable(11); intConnect(INUMTOIVEC(11),....); intEnable(11): Obviously, you must also have the VMEChip2 setup to correctly map VME and DMA interrupts to a PCI interrupt that the ISA-PCI bridge maps to local int 11. Hope this helps. Tim Sharp Software Engineer tsharp@cppus.com --------------------------- Newsgroups: comp.os.vxworks Subject: MC68060 Bug and Fix Date: Thu, 8 Aug 1996 13:28:36 GMT From: mfisher@dyn.com (Mark Fisher) Organization: The World @ Software Tool & Die Message-ID: Reply-To: mfisher@dyn.com Sender: news@world.std.com (Mr Usenet Himself) Hi, This messages describes a flaw in the MC68060 that I have encountered and for which I have produced a test case. This problem exists in the latest mask of the 68060, namely 01F43G. It may or may not exist in earlier masks. I believe a new mask of the 060 will not be released until mid-97 -- I have no indication if a fix for this problem will be included. Motorola has been informed of this flaw. I have been told, by the vendor of my board, that the problem is in the Superscalar pipeline of the MC68060. This is enabled by default under vxWorks as stated on page 540 of the reference manual. Running the test case under these conditions the error occurs. I have verified that after calling the function vxSSDisable(), which disables the Superscalar pipeline, and re-running the test case no error is produced. So I believe that the assertion that the problem is in the Superscalar pipeline is correct. Okay so here is the test case The code below is able to duplicate this problem. The code below has two parts 1) A function, fp_test(), that is to be spawned as a task. This calls the second function repeatedly with arguments such that zero should never be produced. 2) A second function, fp_DoCalc(), that is called by the task that performs the calculation that fails when the Superscalar pipeline is enabled. The task spawned stops as soon as a zero result (the error condition) is produced and prints outs the values used when the calculation was performed. Usually a failure occurs within the first three calls to fp_DoCalc(). In order for the problem to appear the code must be compiled with the following: cc68k -g -fstrength-reduce -fno-builtin -m68040 -m68040-only -Wall \ - -Wno-trigraphs -I$VX_BSP_BASE/h -ansi -pedantic -pipe -nostdinc \ - -DCPU=MC68040 -c -DVXWORKS (The use of -DCPU=MC68040 or -DCPU=MC68060 makes no difference in this code) Once the code is compiled load it into a vxWorks system and then execute it as follows - ->sp fptest Running the produced code on a 68040 system no error is produced, on a 68060 system an error occurs. Notes: It is possible to "fix" this test case by: a) Breaking up the C expression that causes the problem into more than one C statement. b) Specifying optimization, which causes the generated code to be different. and probably a few other ways. However, none of these are really a fix. This is because there is no guarantee that every way in which the problem may occur is fixed by using optimization and it is not possible to know what the compiler will generate for all C code. I reported this problem to Wind River System as TSR 33106 a few weeks ago. But since I see no mention of it on their WEB page I have decided I should now post it to this group. The opportunity to run this test on another vendor's board presented itself lately and the results were the same -- this confirms that the problem is not in my vendor's board. To fix the problem do the following: In the function sysHwInit(), in file sysLib.c, replace the call vxSSEnable() with a call to vxSSDisable(). Note even though the MC68060 disables the Superscalar pipeline by default, the bootrom code will have turned it on. You may also need to replace your bootroms in all 68060 boards -- I guess you will only need to do this if for some reason you have added code to the bootroms that does floating point calcs. (I don't think the bootroms nornally do any floating point stuff.) Here is the actual test code: #include #include typedef struct { double xCoord; double yCoord; double zCoord; }IMGPOS; typedef struct { float sliceLocation; IMGPOS slicePos; }IMGATTR; typedef struct { IMGATTR attr; int imgNum; char sopInstanceUid[64 + 2]; }PER_IMG; typedef struct { double x; double y; double z; }DIRCOS; int fp_DoCalc(PER_IMG *iattrP, DIRCOS *depthP); #define MAX_INNER (40) int fp_test() { DIRCOS depth; PER_IMG imgInfo; int iCnt; int jCnt = 0; int err = 0; depth.x = 0.0; depth.y = 0.0; depth.z = 1.0; imgInfo.attr.slicePos.xCoord = 0.0; imgInfo.attr.slicePos.yCoord = 0.0; for(jCnt = 0; !err; jCnt++) { imgInfo.attr.slicePos.zCoord = 2.0; for (iCnt = 0; !err && iCnt < MAX_INNER; iCnt++) { err = fp_DoCalc(&imgInfo, &depth); if (err) { printf("Error on calculation #%d\n", iCnt + 1 + (jCnt * MAX_INNER)); } imgInfo.attr.slicePos.zCoord += 2.0; } } return(0); } int fp_DoCalc(PER_IMG *iattrP, DIRCOS *depthP) { int err = 0; iattrP->attr.sliceLocation = (depthP->x * iattrP->attr.slicePos.xCoord) + (depthP->y * iattrP->attr.slicePos.yCoord) + (depthP->z * iattrP->attr.slicePos.zCoord); if (iattrP->attr.sliceLocation == 0.0) { printf("Zero computed\n"); printf("\tsliceLocation = %g\n", iattrP->attr.sliceLocation); printf("\tdepth(x,y,z) = (%g, %g, %g)\n", depthP->x, depthP->y, depthP->z); printf("\tposition(x,y,z) = (%g, %g, %g)\n", iattrP->attr.slicePos.xCoord, iattrP->attr.slicePos.yCoord, iattrP->attr.slicePos.zCoord); err = 1; } return(err); } Here is the assembly code, listed on a board. ******************** This is the function fp_test() ************************ - -> l fp_test _fp_test: 1f33c3a 4e56 ff78 LINK .W A6,#0xff78 1f33c3e 2f02 MOVE .L D2,-(A7) 1f33c40 42ae ff7e CLR .L (0xff7e,A6) 1f33c44 42ae ff7a CLR .L (0xff7a,A6) 1f33c48 42ae ffe8 CLR .L (0xffe8,A6) 1f33c4c 42ae ffec CLR .L (0xffec,A6) 1f33c50 42ae fff0 CLR .L (0xfff0,A6) 1f33c54 42ae fff4 CLR .L (0xfff4,A6) 1f33c58 2d7c 3ff0 0000 fff8 MOVE .L #0x3ff00000,(0xfff8,A6) 1f33c60 42ae fffc CLR .L (0xfffc,A6) value = 32717924 = 0x1f33c64 = _fp_test + 0x2a - -> l 1f33c64 42ae ff8a CLR .L (0xff8a,A6) 1f33c68 42ae ff8e CLR .L (0xff8e,A6) 1f33c6c 42ae ff92 CLR .L (0xff92,A6) 1f33c70 42ae ff96 CLR .L (0xff96,A6) 1f33c74 42ae ff7e CLR .L (0xff7e,A6) 1f33c78 4aae ff7a TST .L (0xff7a,A6) 1f33c7c 6600 0086 BNE 0x01f33d04 1f33c80 2d7c 4000 0000 ff9a MOVE .L #0x40000000,(0xff9a,A6) 1f33c88 42ae ff9e CLR .L (0xff9e,A6) 1f33c8c 42ae ff82 CLR .L (0xff82,A6) value = 32717968 = 0x1f33c90 = _fp_test + 0x56 - -> l 1f33c90 4aae ff7a TST .L (0xff7a,A6) 1f33c94 6666 BNE 0x01f33cfc 1f33c96 7427 MOVEQ #0x27,D2 1f33c98 b4ae ff82 CMP .L (0xff82,A6),D2 1f33c9c 6d5e BLT 0x01f33cfc 1f33c9e 70e8 MOVEQ #0xe8,D0 1f33ca0 d08e ADD .L A6,D0 1f33ca2 2f00 MOVE .L D0,-(A7) 1f33ca4 7086 MOVEQ #0x86,D0 1f33ca6 d08e ADD .L A6,D0 value = 32717992 = 0x1f33ca8 = _fp_test + 0x6e - -> l 1f33ca8 2f00 MOVE .L D0,-(A7) 1f33caa 6100 00ca BSR _fp_DoCalc 1f33cae 504f ADDQ .W #0x8,A7 1f33cb0 2d40 ff7a MOVE .L D0,(0xff7a,A6) 1f33cb4 4aae ff7a TST .L (0xff7a,A6) 1f33cb8 6724 BEQ 0x01f33cde 1f33cba 222e ff7e MOVE .L (0xff7e,A6),D1 1f33cbe 2001 MOVE .L D1,D0 1f33cc0 e580 ASL .L #0x2,D0 1f33cc2 d081 ADD .L D1,D0 value = 32718020 = 0x1f33cc4 = _fp_test + 0x8a - -> l 1f33cc4 e780 ASL .L #0x3,D0 1f33cc6 2200 MOVE .L D0,D1 1f33cc8 5281 ADDQ .L #0x1,D1 1f33cca 2001 MOVE .L D1,D0 1f33ccc d0ae ff82 ADD .L (0xff82,A6),D0 1f33cd0 2f00 MOVE .L D0,-(A7) 1f33cd2 487a ff4c PEA (0xff4c,PC) 1f33cd6 61ff fe10 741e BSR _printf 1f33cdc 504f ADDQ .W #0x8,A7 1f33cde f22e 5444 ff9a FDMOVE.D (0xff9a,A6),F0 value = 32718052 = 0x1f33ce4 = _fp_test + 0xaa - -> l 1f33ce4 f23c 5466 4000 0000 0000 0000 FDADD .D #0x4000000000000000,F0 1f33cf0 f22e 7400 ff9a FMOVE .D F0,(0xff9a,A6) 1f33cf6 52ae ff82 ADDQ .L #0x1,(0xff82,A6) 1f33cfa 6094 BRA 0x01f33c90 1f33cfc 52ae ff7e ADDQ .L #0x1,(0xff7e,A6) 1f33d00 6000 ff76 BRA 0x01f33c78 1f33d04 4280 CLR .L D0 1f33d06 6000 0002 BRA 0x01f33d0a 1f33d0a 242e ff74 MOVE .L (0xff74,A6),D2 1f33d0e 4e5e UNLK A6 value = 32718096 = 0x1f33d10 = _fp_test + 0xd6 - -> - -> l 1f33d10 4e75 RTS ******************** This is the function fp_DoCalc() ************************ - -> l fp_DoCalc _fp_DoCalc: 1f33d76 4e56 fffc LINK .W A6,#0xfffc 1f33d7a f227 e004 FMOVEM.X F2,-(A7) 1f33d7e 2f0a MOVE .L A2,-(A7) 1f33d80 2f02 MOVE .L D2,-(A7) 1f33d82 42ae fffc CLR .L (0xfffc,A6) 1f33d86 206e 0008 MOVEA .L (0x8,A6),A0 1f33d8a 226e 000c MOVEA .L (0xc,A6),A1 1f33d8e 246e 0008 MOVEA .L (0x8,A6),A2 1f33d92 f211 5444 FDMOVE.D (A1),F0 1f33d96 f22a 5467 0004 FDMUL .D (0x4,A2),F0 value = 32718236 = 0x1f33d9c = _fp_DoCalc + 0x26 - -> l 1f33d9c 226e 000c MOVEA .L (0xc,A6),A1 1f33da0 246e 0008 MOVEA .L (0x8,A6),A2 1f33da4 f229 54c4 0008 FDMOVE.D (0x8,A1),F1 1f33daa f22a 54e7 000c FDMUL .D (0xc,A2),F1 1f33db0 f200 0466 FDADD .X F1,F0 1f33db4 226e 000c MOVEA .L (0xc,A6),A1 1f33db8 246e 0008 MOVEA .L (0x8,A6),A2 1f33dbc f229 54c4 0010 FDMOVE.D (0x10,A1),F1 1f33dc2 f22a 54e7 0014 FDMUL .D (0x14,A2),F1 1f33dc8 f200 0466 FDADD .X F1,F0 value = 32718284 = 0x1f33dcc = _fp_DoCalc + 0x56 - -> l 1f33dcc f200 0140 FSMOVE.X F0,F2 1f33dd0 f210 6500 FMOVE .S F2,(A0) 1f33dd4 206e 0008 MOVEA .L (0x8,A6),A0 1f33dd8 f210 4444 FDMOVE.S (A0),F0 1f33ddc f200 003a FS .X D0 1f33de0 f28e 0098 FB .W #0x98487a 1f33de4 487a ff2c PEA (0xff2c,PC) 1f33de8 61ff fe10 730c BSR _printf 1f33dee 584f ADDQ .W #0x4,A7 1f33df0 206e 0008 MOVEA .L (0x8,A6),A0 value = 32718324 = 0x1f33df4 = _fp_DoCalc + 0x7e - -> l 1f33df4 f210 4544 FDMOVE.S (A0),F2 1f33df8 f227 7500 FMOVE .D F2,-(A7) 1f33dfc 201f MOVE .L (A7)+,D0 1f33dfe 221f MOVE .L (A7)+,D1 1f33e00 2f01 MOVE .L D1,-(A7) 1f33e02 2f00 MOVE .L D0,-(A7) 1f33e04 487a ff1b PEA (0xff1b,PC) 1f33e08 61ff fe10 72ec BSR _printf 1f33e0e 504f ADDQ .W #0x8,A7 1f33e10 584f ADDQ .W #0x4,A7 value = 32718354 = 0x1f33e12 = _fp_DoCalc + 0x9c - -> l 1f33e12 206e 000c MOVEA .L (0xc,A6),A0 1f33e16 2f28 0014 MOVE .L (0x14,A0),-(A7) 1f33e1a 2f28 0010 MOVE .L (0x10,A0),-(A7) 1f33e1e 206e 000c MOVEA .L (0xc,A6),A0 1f33e22 2f28 000c MOVE .L (0xc,A0),-(A7) 1f33e26 2f28 0008 MOVE .L (0x8,A0),-(A7) 1f33e2a 206e 000c MOVEA .L (0xc,A6),A0 1f33e2e 2f28 0004 MOVE .L (0x4,A0),-(A7) 1f33e32 2f10 MOVE .L (A0),-(A7) 1f33e34 487a ff00 PEA (0xff00,PC) value = 32718392 = 0x1f33e38 = _fp_DoCalc + 0xc2 - -> l 1f33e38 61ff fe10 72bc BSR _printf 1f33e3e defc 001c ADDA .W #0x1c,A7 1f33e42 206e 0008 MOVEA .L (0x8,A6),A0 1f33e46 2f28 0018 MOVE .L (0x18,A0),-(A7) 1f33e4a 2f28 0014 MOVE .L (0x14,A0),-(A7) 1f33e4e 206e 0008 MOVEA .L (0x8,A6),A0 1f33e52 2f28 0010 MOVE .L (0x10,A0),-(A7) 1f33e56 2f28 000c MOVE .L (0xc,A0),-(A7) 1f33e5a 206e 0008 MOVEA .L (0x8,A6),A0 1f33e5e 2f28 0008 MOVE .L (0x8,A0),-(A7) value = 32718434 = 0x1f33e62 = _fp_DoCalc + 0xec - -> l 1f33e62 2f28 0004 MOVE .L (0x4,A0),-(A7) 1f33e66 487a feec PEA (0xfeec,PC) 1f33e6a 61ff fe10 728a BSR _printf 1f33e70 defc 001c ADDA .W #0x1c,A7 1f33e74 7401 MOVEQ #0x1,D2 1f33e76 2d42 fffc MOVE .L D2,(0xfffc,A6) 1f33e7a 202e fffc MOVE .L (0xfffc,A6),D0 1f33e7e 6000 0002 BRA 0x01f33e82 1f33e82 242e ffe8 MOVE .L (0xffe8,A6),D2 1f33e86 246e ffec MOVEA .L (0xffec,A6),A2 value = 32718474 = 0x1f33e8a = _fp_DoCalc + 0x114 - -> l 1f33e8a f22e d020 fff0 FMOVEM.X (0xfff0,A6),F2 1f33e90 4e5e UNLK A6 1f33e92 4e75 RTS Mark Fisher --------------------------- Newsgroups: comp.os.vxworks Subject: Trapping VME bus errors. Date: 09 Aug 1996 21:37:16 -0700 From: mcs@goblin.caltech.edu (Martin Shepherd) Organization: California Institute of Technology. Message-ID: I am trying to write some diagnostic code to trap VMEbus SIGBUS and SIGSEGV exceptions on an MVME177 MC68060 CPU running VxWorks 5.3. I have written a signal handler to trap these exceptions but now I need to know how to determine the local address that caused the exception. My understanding is that the MC68060 places the address that caused the exception on the stack, but I don't know where to look on the stack with respect to code in the signal handler. I would be grateful if somebody could give me some constructive suggestions. Martin Shepherd (mcs@astro.caltech.edu) --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Ada and VxWorks Date: 8 Aug 1996 12:48:20 GMT From: johncoby@ghgcorp.com (JOHN R. COBARRUVIAS) Organization: Personal Account Message-ID: References: <4ucind$5ft@lfsserv1.fsc.ibm.com> In article <4ucind$5ft@lfsserv1.fsc.ibm.com>, tom.kodey@lmco.com wrote: > I'm considering VxWorks as the RTOS for an Ada based imbedded application and > am considering various development environments. The target will likely be a > Pentium, PowerPC or R4xxx. I'm interested to hear what Ada development > environment you were using and, briefly, what you thought of it (host, ease of > interfacing with VxWorks, ease of use, maturity, support, host and target > based debug, etc.)? > Hum....do you have to use Ada? (Ok, enough Ada bashing from an ex-Ada supporter) I have used VADSworks from Verdix, now Rational. It was a very adequate set of tools except... VADSworks was built upon vxWorks version 5.0.1. Verdix was not going to upgrade VADSworks to vxWorks 5.1 because it required inputs from Windriver and they were not very interested. So there was a "lag" in support. And we sometimes got the old "its not a VADSworks problem, its a vxWorks problem". Other than that, it was a pretty good system. Now, I understand Verdix is now Rational. So you may have Rational built upon VADSworks, built upon vxWorks. E-Gads! But I would check on that. My experience with Rational has been TRY BEFORE YOU BUY! Hum....do you have to use Ada? > Tom Kodey > tom.kodey@lmco.com > Lockheed Martin Federal Systems - -- ,-_/ . .-,--. ,---. . ' | ,-. |-. ,-. `|__/ | -' ,-. |-. ,-. ,-. ,-. . . . , . ,-. ,-. | | | | | | | | \ ,. | -. | | | | ,-| | | | | | / | ,-| `-. | `-' ' ' ' ' `-' ` `' `---' `-' `-' `-^ ' ' `-' `' ' `-^ `-' / | John R. Cobarruvias johncoby@ghgcorp.com Houston, Tx `--' --------------------------- End of New-News digest ********************** From rterry@cacihsv.com Mon Aug 12 05:52:24 1996 From: "Russell Terry" Date: Mon Aug 12 05:52:26 PDT 1996 Subject: Re: Software Announcement send newsletter info ---------------------------------------------------- Russell M. Terry Jr. Senior Systems Engineer CACI-ASG 1555 The Boardwalk InfraRed Sensor Systems Huntsville, AL 35816 Embedded Systems 205.830.4782, voice Custom Hardware/Software 205.837.2086, fax VME, VxWorks, SCSI From mea@mclean.sparta.com Mon Aug 12 07:53:35 1996 From: Mike Anderson Date: Mon Aug 12 07:53:37 PDT 1996 Subject: Re: A32 space VME access At 04:00 AM 8/10/96 PDT, you wrote: >From: "Michael J. Dempsey" > >Hello > > We have a PowerPC (MV1603) running with a beta version of VxWorks 5.3 > embedded in VME chassis that contains two other VME boards with memory > registers that we would like our VxWorks application to read and write > to. One of the boards uses A16 address space and has a base address of > 0x8000. The other uses A32 address space and has a base address of > 0x00300000. > For A16 space board we used sysBusToLocalAdrs to map its memory to > something our PowerPC VxWorks application could use. It returned (if > memory serves me correctly) 0xefff8xxx (where xxx is an offset from > the base address). This worked fine, we read from that board's > memory and also wrote to it. > We tried using the same call to access memory on the board with > A32 address space (changing the first parameter for sysBusLocalAdrs > from VM_SUP_SHORT_IO to VM_EXT_SUP_DATA to indicate the different > address space). For an input of 0x00300000 to sysBusLocalAdrs we get > returned and address of 0xd0300000, but when the program tried to > access that memory location it blew up. We have a bus analyzer in > the VME rack set to trigger on anything but it never did, so the > application seems to have blown up before even trying to access > the VME memory.Is there something I need to set up or include in > the kernel to access this off-board memory? Any help is appreciated. > Greetings! Not being a PPC user, I can only suggest that you take a close look at both your sysPhysMemDesc structure (make sure there's an entry for the A32 space you wish to access) and your sysLocalToBusAddr & sysBusToLocalAddr functions in your BSP (check these for accuracy, I've found many errors in these two functions over the last 9 years working with VxWorks). Make sure your compare them with the memory map figures in your PPC's documentation (yikes! You mean I have to look in the manual?!?! ;-). BTW, 0x300000 seems like a very low address to be located in A32. That sounds more like an A24 address. HTH, =============================================================================== __ Real-Time System Development, Integration, Training and Services //\\ // \\ Mike Anderson // /\ \\ Chief Engineer Voice : (703) 448-0210 ext. 235 // / \ \\ SPARTA, Inc. FAX : (703) 734-3323 // \ \\ 7926 Jones Branch Drive EMAIL : mea@mclean.sparta.com \\ \ // Suite 900 Web : http://www.mclean.sparta.com \\ \ / // McLean, VA 22102 \\ \/ // "Software development is like making \\ // a baby... You can't make a baby in one \\// month by impregnating nine women. -- "Pride in Performance" Some things just take time." =============================================================================== From tmorris@me.umn.edu Mon Aug 12 14:25:16 1996 From: "Ted Morris" Date: Mon Aug 12 14:25:18 PDT 1996 Subject: What about LINUX?? Any word on if WindRiver is going to port VxWorks or Tornado to LINUX OS?? We would be EXTREMELY interested in this. thanks, Ted _/_/_/ _/_/ _/_/ _/_/ _/_/_/ _/_/ _/_/_/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/_/ _/ _/ _/ _/ _/ _/ _/ _/_/_/_/ _/ _/ _/ _/ _/ _/_/_/_/ _/ _/_/_/ _/ _/ _/ _/ _/_/_/ _/ _/ _/_/_/ ------------------------------------------------------------------------------ Ted Morris tmorris@me.umn.edu 612-625-3520 Center For Advanced Manufacturing Design And Control 612-625-9881 Institute of Technology FAX: 612-625-8884 University of Minnesota 111 Church St. S.E. (Rm 125) Mpls MN 55455 "A lie can travel halfway around the world before the truth can get its shoes on." [Actually, NOT Mark Twain] ------------------------------------------------------------------------------ From chrisf@jna.com.au Mon Aug 12 17:13:48 1996 From: "Chris Ford" Date: Mon Aug 12 17:13:50 PDT 1996 Subject: Handling statics with C++ We are currently using VxWorks in a C environment, and are considering a move to C++. One concern we have, however, is the treatment of statics. Under VxWorks, if we wish to start several instances of the same task, and we wish each instance to have its own copy of a static, we need to use taskVarAdd(). This has been fine so far as we can combine the necessary statics into a single structure and only tasVarAdd() a single pointer to that structure. This minimises the hit on context switch time. In C++, however, the whole idea is to hide your data, meaning that any static member variables should remain part of a class. Two questions then arise: - How to taskVarAdd() these statics. One solution is to have a static member function used to initialise the class (a bit yucky, but workable). - What do you do about the performance problem? Because the data should remain hidden, it's a bad idea to pull the statics out into a combined structure. This means that every class which has a static must do its own taskVarAdd(), multiplying the number of task variables, and increasing the context switch time. The question is: what do people do about this? Many thanks in advance for any ideas. -- Chris Ford Senior Project Engineer JNA Telecommunications voice: +61 2 9935 5753 fax: +61 2 417 3862 e-mail: chrisf@jna.com.au post: 16 Smith St, Chatswood, NSW 2067, Australia. ----------------------------------------------------------------------------JNA From casey@gateway.ccii.co.za Tue Aug 13 05:18:28 1996 From: Casey Crellin Date: Tue Aug 13 05:18:30 PDT 1996 Subject: Re: What about LINUX?? > Any word on if WindRiver is going to port VxWorks or Tornado to LINUX OS?? Now that is a really good suggestion. Linux would make a very good host station. Any word from the WRS people? Casey -- Casey Crellin casey@ccii.co.za CCII Systems (Pty) Ltd ----------------------------------------------------------------------------- From SMTP_Bridge_Server@msmail3.hac.com Tue Aug 13 05:41:23 1996 From: "SMTP Bridge Server" Date: Tue Aug 13 05:41:25 PDT 1996 Subject: Undeliverable Mail Unknown Microsoft mail form. Approximate representation follows. Message: comp.os.vxworks newsdigest Sent: Mon, Aug 12, 1996 5:35 AM To: Quiroz, Gabriel On Server: 631 Segment2 Date: Tue, Aug 13, 1996 5:40 AM Reason: Could not be delivered because the destination Microsoft Mail server could not be found. From SMTP_Bridge_Server@msmail3.hac.com Tue Aug 13 07:13:30 1996 From: "SMTP Bridge Server" Date: Tue Aug 13 07:13:32 PDT 1996 Subject: Undeliverable Mail Unknown Microsoft mail form. Approximate representation follows. Message: Re: Software Announcement Sent: Mon, Aug 12, 1996 7:08 AM To: Quiroz, Gabriel On Server: 631 Segment2 Date: Tue, Aug 13, 1996 7:11 AM Reason: Could not be delivered because the destination Microsoft Mail server could not be found. From SMTP_Bridge_Server@msmail3.hac.com Tue Aug 13 09:55:45 1996 From: "SMTP Bridge Server" Date: Tue Aug 13 09:55:47 PDT 1996 Subject: Undeliverable Mail Unknown Microsoft mail form. Approximate representation follows. Message: Re: A32 space VME access Sent: Mon, Aug 12, 1996 9:54 AM To: Quiroz, Gabriel On Server: 631 Segment2 Date: Tue, Aug 13, 1996 9:55 AM Reason: Could not be delivered because the destination Microsoft Mail server could not be found. From anthonyl@atlantis.softeng.lamrc.com Tue Aug 13 15:56:42 1996 From: Anthony Le Date: Tue Aug 13 15:56:44 PDT 1996 Subject: read available number of characters of a serial port. Hello, Do anybody know how to read available number of unread bytes of a serial port? Anthony Le Lam Research Corp. From daemon@csg.lbl.gov Wed Aug 14 04:06:15 1996 From: daemon@csg.lbl.gov Date: Wed Aug 14 04:07:06 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Wed Aug 14 04:00:48 PDT 1996 Subject: Re: Standard I/O read with a timeout Subject: Re: What about LINUX?? Subject: Re: Handling statics with C++ Subject: Time synchronisation ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Standard I/O read with a timeout Date: 14 Aug 1996 00:05:28 GMT From: David Meeks Organization: Siemens Medical Systems, Ultrasound Group Message-ID: <4ur588$c18@hood.sqi.com> References: <3210D651.41C67EA6@adc.com> I had essentially the same problem when I wrote some code to initialize a modem. In my systems modems are optional and I do not know if one is attached or not. When I call read() and no modem is attached to my serial port, read() would not return so I spawned a separate task to make the read() call. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: What about LINUX?? Date: 14 Aug 1996 03:24:22 GMT From: david@ods.com (David Engel) Organization: Optical Data Systems, Inc. Message-ID: <4urgt6$l9g@news.ods.com> References: <199608122125.VAA01187@ziv.me.umn.edu> Ted Morris wrote: > Any word on if WindRiver is going to port VxWorks or Tornado to LINUX OS?? > We would be EXTREMELY interested in this. So would we! David - -- David Engel Optical Data Systems, Inc. david@ods.com 1101 E. Arapaho Road (214) 234-6400 Richardson, TX 75081 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Handling statics with C++ Date: Tue, 13 Aug 1996 08:09:46 -0700 From: Todd Hoff Organization: Possibility Outpost Message-ID: <32109ABA.28AB@possibility.com> References: <9608130013.AA29767@clix.jna.com.au> <32108499.21EF@ucar.edu> Joe Van Andel wrote: > // return pointer to single instance of this object/task. > A1Record *A1Record::Instance() > { > if (!instance_) { > instance_ = new A1Record(); > } A mutex is needed to protect against multiple accesses to Instance() during construction. We've also had a problem when the instance started ISA thread as the object will not be ready to use even though a pointer is returned. We need another level to wait on the state of the object to be ready. - ------------------------------------------------------------- tmh@possibility.com | The loyalty of small men can be http://www.possibility.com | bought cheaply, for greed has no | pride. - Michael Kube-McDowell --------------------------- Newsgroups: comp.os.vxworks Subject: Time synchronisation Date: Wed, 14 Aug 1996 08:41:05 GMT From: hiske@ict.nl (Hiske Huiting) Organization: ICT Automatisering Deventer BV Message-ID: Reply-To: hiske@ict.nl Sender: usenet@tweety.ict.nl (news) I have Sparc board with VxWorks 5.2 connected to a Silicon Graphics Challenge through VME and TCP/IP. What is the easiest way to synchronise the VxWorks time with the SGI? Thanx, Hiske ____________________________________________________________________________ name: H.Huiting / ___ \ ICT Automatisering B.V. / Aerospace internet: hiske@ict.nl / / \ \ P.O.Box 701 fax: 31-570-621362 ( ( ICT ) 7400 AS Deventer tel: 31-570-633888 \ \___/ / The Netherlands ____________________________\_______/______________________________:-)______ --------------------------- End of New-News digest ********************** From mea@mclean.sparta.com Wed Aug 14 07:20:29 1996 From: Mike Anderson Date: Wed Aug 14 07:21:52 PDT 1996 Subject: Re: read available number of characters of a serial port. At 03:56 PM 8/13/96 PDT, you wrote: >Submitted-by anthonyl@atlantis.softeng.lamrc.com Tue Aug 13 15:56:42 1996 >Submitted-by: Anthony Le > >Hello, > >Do anybody know how to read available number of unread bytes of a >serial port? Greetings! It's really quite straightforward. Open a file descriptor that corresponds to the serial port in question and use the FIONREAD ioctl call: #include "vxWorks.h" #include "ioctl.h" #include "ioLib.h" short snippet() { int fd, numBytesInBuffer; /* this open should probably only be done once in your program */ fd = open("/tyCo/1", O_RDWR); /* open for "update" */ numBytesInBuffer = ERROR; /* assume ERROR case */ /* call ioctl & check ioctl status to detect ERROR encountered in call */ if (ioctl(fd, FIONREAD, &numBytesInBuffer) == ERROR) { /* add any error correction stuff here */ close(fd); /* clean ups should be done once in your program */ return(ERROR); } else { close(fd); /* clean ups should be done once in your program */ return(numBytesInBuffer); } } Provided your vxWorks doesn't have a broken FIONREAD ioctl, this should work. BTW, if your using file pointers (fopen, fread, etc.), then replace the ioctl(fd,...) with ioctl(fileno(filePointer),...). HTH, =============================================================================== __ Real-Time System Development, Integration, Training and Services //\\ // \\ Mike Anderson // /\ \\ Chief Engineer Voice : (703) 448-0210 ext. 235 // / \ \\ SPARTA, Inc. FAX : (703) 734-3323 // \ \\ 7926 Jones Branch Drive EMAIL : mea@mclean.sparta.com \\ \ // Suite 900 Web : http://www.mclean.sparta.com \\ \ / // McLean, VA 22102 \\ \/ // "Software development is like making \\ // a baby... You can't make a baby in one \\// month by impregnating nine women. -- "Pride in Performance" Some things just take time." =============================================================================== From punamia@falcon.kla.com Wed Aug 14 10:03:44 1996 From: punamia@falcon.kla.com (Manoj Punamia) Date: Wed Aug 14 10:03:47 PDT 1996 Subject: Re: read available number of characters of a serial port. >Hello, >Do anybody know how to read available number of unread bytes of a >serial port? >Anthony Le >Lam Research Corp. Use the following: ioctl(port_name, FIONREAD, &nbytes); where: port_name = file descriptor of the port you opened. FIONREAD = 1, defined in ioLib.h to read num chars available to read. nbytes = # of bytes available to read (int). -- Manoj Punamia anthonyl@atlantis.softeng.lamrc.com KLA Instruments From lbecker@avtec.com Wed Aug 14 14:48:59 1996 From: "Lynn W. Becker, II" Date: Wed Aug 14 14:49:01 PDT 1996 Subject: 5.3 booting from 5.2 Shared memory? Hi, I am having trouble booting my mv1300. As you may know, there is no onboard ethernet on the mv1300. I am attempting to a backplane ethernet to boot the board. I have 3 Heurikon Baja boards running VxWorks 5.2. The three boards have the following bootlines Boot line: ei(0,0)skybox.avtec.com:/export/home/vw/config/hkbaja47/vxWorks e=206.161.148.30 b=90.0.0.1 h=206.161.148.29 u=vxuser tn=uno.avtec.com value = 146 = 0x92 Boot line: sm=040000600(0,2)skybox.avtec.com:/export/home/vw/config/hkbaja47/vxWorks b=90.0 .0.3:ff000000 h=206.161.148.29 g=90.0.0.1 u=vxuser tn=tres.avtec.com value = 160 = 0xa0 Boot line: sm=0x40000600(0,3)skybox.avtec.com:/export/home/vw/config/hkbaja47/vxWorks b=90. 0.0.4:ff000000 h=206.161.148.29 g=90.0.0.1 u=vxuser tn=quatro.avtec.com value = 163 = 0xa3 Note that first board boots from the host over the ethernet, and the other two are set up to boot from the backplane ethernet. They are successful in doing this. The fourth board is the mv1300. Running 5.3 beta, trying to boot from the shared memory network, and I get the following... VxWorks System Boot Copyright 1984-1995 Wind River Systems, Inc. CPU: Motorola MVME1305 - PowerPC 603 Version: 5.3-PowerPC-beta BSP version: 1.1/2 Creation date: Aug 13 1996, 14:01:15 Press any key to stop auto-boot... 0 auto-booting... boot device : sm=0x40000600 processor number : 4 host name : skybox.avtec.com file name : /export/home/wind/target/config/mv1305/vxWorks inet on backplane (b): 90.0.0.5 host inet (h) : 206.161.148.29 gateway inet (g) : 90.0.0.1 user (u) : vxuser flags (f) : 0x0 target name (tn) : cinqo.avtec.com Backplane anchor at 0x40000600... Attaching network interface sm0... done. smIsAlive: readyValue = 0x7c7343a6, heartbeat = 0x0 smIsAlive: readyValue = 0x7c7343a6, heartbeat = 0x0 [snip..] smIsAlive: readyValue = 0x7c7343a6, heartbeat = 0x0 initialization failed for device "sm0" Can't load boot file!! Can a 5.3 machine boot from the 5.2 shared memory network? Would any more information be helpful in figuring this out? ______________________________________________________________________ Mr. Lynn W. Becker lbecker@avtec.com Engineer (703) 273-2211 Avtec Systems Inc (703) 273-1313 FAX 10530 Rosehaven Street, Suite 300 http://www.avtec.com Fairfax, VA 22030 From mea@mclean.sparta.com Wed Aug 14 15:26:07 1996 From: Mike Anderson Date: Wed Aug 14 15:26:09 PDT 1996 Subject: Re: 5.3 booting from 5.2 Shared memory? At 02:49 PM 8/14/96 PDT, you wrote: >Submitted-by lbecker@avtec.com Wed Aug 14 14:48:59 1996 >Submitted-by: "Lynn W. Becker, II" > >Hi, I am having trouble booting my mv1300. As you may know, there is >no onboard ethernet on the mv1300. I am attempting to a backplane >ethernet to boot the board. > [snip] >Boot line: > >sm=0x40000600(0,3)skybox.avtec.com:/export/home/vw/config/hkbaja47/vxWorks >b=90. > >0.0.4:ff000000 h=206.161.148.29 g=90.0.0.1 u=vxuser >tn=quatro.avtec.com > >value = 163 = 0xa3 > > >Note that first board boots from the host over the ethernet, and the >other two are set up to boot from the backplane ethernet. They are >successful in doing this. > > >The fourth board is the mv1300. Running 5.3 beta, trying to boot from >the shared memory network, and I get the following... > [snip] >processor number : 4 >host name : skybox.avtec.com >file name : /export/home/wind/target/config/mv1305/vxWorks >inet on backplane (b): 90.0.0.5 >host inet (h) : 206.161.148.29 >gateway inet (g) : 90.0.0.1 >user (u) : vxuser >flags (f) : 0x0 >target name (tn) : cinqo.avtec.com > >Backplane anchor at 0x40000600... Attaching network interface sm0... >done. > >smIsAlive: readyValue = 0x7c7343a6, heartbeat = 0x0 > [snip] > >initialization failed for device "sm0" > > This looks like the mvme1300 is not "seeing" the same memory as the the Bajas. In fact, I would be somewhat surprised if the 1300 had the same VME memory map as the Baja. Therefore, my first guess would be that 0x40000600 may not be the same physical location that the Bajas see at that address. A good test would be from the 1300's [VxWorks Boot]: prompt, do a [VxWorks Boot]: d 0x40000600 You should see the "sm" signature value of 0x87654321. If you don't, then the address maps don't line up between the two processor types and you'll have to go to your 1300 documentation to find how to properly map that address. HTH, =============================================================================== __ Real-Time System Development, Integration, Training and Services //\\ // \\ Mike Anderson // /\ \\ Chief Engineer Voice : (703) 448-0210 ext. 235 // / \ \\ SPARTA, Inc. FAX : (703) 734-3323 // \ \\ 7926 Jones Branch Drive EMAIL : mea@mclean.sparta.com \\ \ // Suite 900 Web : http://www.mclean.sparta.com \\ \ / // McLean, VA 22102 \\ \/ // "Software development is like making \\ // a baby... You can't make a baby in one \\// month by impregnating nine women. -- "Pride in Performance" Some things just take time." =============================================================================== From leonid@rst.co.il Wed Aug 14 23:01:32 1996 From: leonid@rst.co.il (Leonid Rosenboim) Date: Wed Aug 14 23:01:34 PDT 1996 Subject: Re: What about LINUX?? > Any word on if WindRiver is going to port VxWorks or Tornado to LINUX OS?? Who do you call then if something to do with Linux does not work ? Ghost-Busters ! Seriously, unsupported host operating systems can cause a big support headache for Wind River, cause all problems will end up bouncing back to them. God knows how many problem reports related to e.g. "booting from Solaris" have been handeled, even the exploder had its share. And this is for Solaris 2.x - one of the best supported OSes. At least our local bunch are a zillion times better then the Microsoft people that handle support calls. Just imageine what it would be like, and imageine the whole support group will get even more work load then it has now, and you know who will be effected ... Linux - vote NO. Leonid From daemon@csg.lbl.gov Thu Aug 15 04:00:28 1996 From: daemon@csg.lbl.gov Date: Thu Aug 15 04:00:30 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Thu Aug 15 04:00:25 PDT 1996 Subject: NT/pSOS RPC timeout to embedded system ------------------------------------------------------- Newsgroups: comp.arch.embedded,comp.os.vxworks Subject: NT/pSOS RPC timeout to embedded system Date: Sun, 04 Aug 1996 08:44:36 -0500 From: hogle@crd.ge.com (Rich Hogle) Organization: GE Corp R&D Center, Schenectady NY Message-ID: I have a high-performance data acq system, the actions of which are coordinated by software running on a remote WindowsNT system. The NT host starts an acquisition via a blocking RPC function call, served by the embedded system (pSOS on PPC1604). The embedded system performs the acquistion, then issues a return and the NT host continues. All is fine until acquistions times reach/exceed 188 seconds, at which time the NT RPC returns with connection-reset status code, even though the acquisition has not completed and no return/reset has been issued. Setting RPC timeout value to high number (10000 seconds) does not help. Is there any place else under NT where this magic number of 188 seconds is configured?? The acquisition task continues on its merry way collecting data for 188+nn seconds, and when done, completes normally and tries to return status to NT client which has since departed. - If RPC client is a SunOS application, then system runs fine. Timeouts occur at the specified time that I set, exactly as expected. Acquisitions for many minutes run routinely. - during a acquisition, the embedded pSOS system suspends all networking tasks to avoid any delay in capturing data. Does NT somehow detected this and think that a connection reset has occured? SunOS doesn't seem to know/care. - We are using Chameleon/NetManage RPC-SDK environment - pSOS 2.1.2 on MVME1604 Any hints will be much appreciated. Thanks Rich Hogle GE-CRD hogle@crd.ge.com --------------------------- End of New-News digest ********************** From lbecker@avtec.com Thu Aug 15 06:38:16 1996 From: "Lynn W. Becker, II" Date: Thu Aug 15 06:38:18 PDT 1996 Subject: Re: 5.3 booting from 5.2 Shared memory? > A good test would be from the 1300's [VxWorks Boot]: >prompt, do a > >[VxWorks Boot]: d 0x40000600 > >You should see the "sm" signature value of 0x87654321. If you don't, then >the address maps don't line up between the two processor types and you'll >have to go to your 1300 documentation to find how to properly map that >address. [VxWorks Boot]: d 0x40000600 40000600: 7c73 43a6 7c68 02a6 4822 d4af 3861 0000 *|sC.|h..H"..8a..* 40000610: 9421 fff0 4822 8d1b 3821 0010 4822 d5cf *.!..H"..8!..H"..* 40000620: 0000 0000 0000 0000 0000 0000 0000 0000 *................* 40000630: 0000 0000 0000 0000 0000 0000 0000 0000 *................* 40000640: 0000 0000 0000 0000 0000 0000 0000 0000 *................* 40000650: 0000 0000 0000 0000 0000 0000 0000 0000 *................* 40000660: 0000 0000 0000 0000 0000 0000 0000 0000 *................* 40000670: 0000 0000 0000 0000 0000 0000 0000 0000 *................* [VxWorks Boot]: Thanks for the tip Mike. I will go RTFM. ______________________________________________________________________ Mr. Lynn W. Becker lbecker@avtec.com Engineer (703) 273-2211 Avtec Systems Inc (703) 273-1313 FAX 10530 Rosehaven Street, Suite 300 http://www.avtec.com Fairfax, VA 22030 From punamia@falcon.kla.com Thu Aug 15 10:40:36 1996 From: punamia@falcon.kla.com (Manoj Punamia) Date: Thu Aug 15 10:40:38 PDT 1996 Subject: Re: NT/pSOS RPC timeout to embedded system I had a similar problem with the older version of SunOS. The timeout value we set in RPC func "clnt_call" did not work. We got around this by changing the default RPC timeout by calling func "clnt_control" and resetting to default after the call. This is somewhat inefficient but it works. Other obvious suggestion is to run your acquisition task without suspending network task just for debugging purpose to check if NT is acting too smart. --Manoj Punamia KLA INSTRUMENTS From punamia@falcon.kla.com Thu Aug 15 10:48:03 1996 From: punamia@falcon.kla.com (Manoj Punamia) Date: Thu Aug 15 10:48:05 PDT 1996 Subject: Re: What about LINUX?? Why would it be any different that say using gnu tools ? From daemon@csg.lbl.gov Fri Aug 16 04:01:24 1996 From: daemon@csg.lbl.gov Date: Fri Aug 16 04:01:26 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Fri Aug 16 04:01:19 PDT 1996 Subject: Re: What about LINUX?? Subject: NVRAM logging library now in the FTP archives... Subject: Re: Source control with Tornado? Subject: S: Backup programm for vxworks Subject: Re: Standard I/O read with a timeout Subject: Re: YEAR 2000 Subject: Re: What about LINUX?? Subject: Re: HTTP server for VxWorks Subject: Re: Source control with Tornado? Subject: STL with Tornado/VxWorks Subject: Re: YEAR 2000 Subject: Re: Downloading vxWorks to PPC ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: What about LINUX?? Date: 15 Aug 1996 09:21:07 -0600 From: bgeer@xmission.xmission.com (bgeer) Organization: XMission Internet (801 539 0900) Message-ID: <4uvf93$aog@xmission.xmission.com> References: <9608150525.AA04342@rst.co.il> leonid@rst.co.il (Leonid Rosenboim) writes: >> Any word on if WindRiver is going to port VxWorks or Tornado to LINUX OS?? >Who do you call then if something to do with Linux does not work ? > Ghost-Busters ! No, you "call" the appropriate one of a large number of Linux newsgroups. I dare say the Linux developers & users are a more responsive group than *any* single-vendor OS. Of course there's no tech support phone number that I know of for Linux, but then my experience using phone support has been much worse than support I've received using email & newsgroup solicitations. >Seriously, unsupported host operating systems can cause a big >support headache for Wind River, cause all problems will end up >bouncing back to them. God knows how many problem reports related >to e.g. "booting from Solaris" have been handeled, even the exploder >had its share. Do you have the sources, *ALL* the sources, for Solaris? I have all the sources for Linux...if something doesn't work, I can track it down myself. I haven't needed to...Linux just works! I installed & configured Linux on a ProStar Pentium laptop for supporting VxWorks on a MVME162LX on a field delivery. It took about 2 hours using cdrom. I enabled tftp, fired up the Ethernet PCMCIA card, & actually booted from it, as well as wrote flash ram & was able to do everything else I needed to except compile. All I needed was a cc68k that executed on x86 & I would have had a complete portable development system. This laptop cost less than $4000 US. Say that about a Solaris machine...:-) I have a copy of the FUZZ report from UWisc (unfortunately at home & I'm posting this from work). The statistics in this report indicate Linux & its utilities are more stable than other Unix's based on the results from their test suite. Well worth reading... The only downside to Linux I know of is support for certain pieces of new hardware. A prime example is XWindow support for the Matrox Millenium(?) display adapter. Matrox would not release their specs to public domain so the XFree folks couldn't create a driver using its features. Note that either Xinside or MetroX *does* have a <$200 Linux X driver for it. So far, tho, this kind of problem has been very rare, & that Linux drivers are available for the vast majority of PC hardware products. >... At least our local bunch are a zillion times better then the Microsoft >people that handle support calls. Yeah, well, sure...I'd expect that. >Just imageine what it would be like, and imageine the whole support >group will get even more work load then it has now, and you know >who will be effected ... No doubt any new platform/OS combination would increase support demand. I'll bet the support required by Linux users would not be significantly more than that required by other OS's; actually, I'll bet it would be significantly *less*. >Linux - vote NO. Try Linux...you'll vote YES. Bob - -- <> Robert Geer & Donna Tomky / * <> <> bgeer@xmission.com | _o * o * o <> <> Salt Lake City, Utah | -\<, * <\ <> U S A | O/ O __ /__, /> <> --------------------------- Newsgroups: comp.os.vxworks Subject: NVRAM logging library now in the FTP archives... Date: Tue, 13 Aug 1996 20:01:12 GMT From: gnn@wrs.com (George Neville-Neil) Organization: Wind River Systems, Inc. Message-ID: Hi Folks, Well, I've written something those of you with NVRAM in your targets might find useful. It's something I call nvLogLib, which allows you to log messages into your NVRAM and then retreive them. This faciliy uses sysNvRamSet and sysNvRamGet so you'd better have those symbols before linking this into your code. There is a complete README file in the tar file placed on the archive. I'm not in control of the archive but I'm told the package should be there soon. Please read the disclaimer in the source code for legal/warranty information. Later, George - -- George V. Neville-Neil work: gnn@wrs.com home:gnn@netcom.com NIC: GN82 The path of my life is strewn with cowpats from the devil's own satanic herd! - Edmund Blackadder --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Source control with Tornado? Date: 15 Aug 1996 17:17:53 GMT From: friedl@cher.heurikon.com (Ted Friedl) Organization: Heurikon Corporation Message-ID: <4uvm41$a4k@badger.heurikon.com> References: <4uniph$fu4$1@mhade.production.compuserve.com> <4uuo9h$lgs@idiom.com> In article <4uuo9h$lgs@idiom.com>, H.J. Bae wrote: >Jim Wilson <102105.3240@CompuServe.COM> wrote: > >>Can anyone recommend a Source Code management system that works >>well with the Tornado for Windows? > >I use GNU RCS/CVS for windows NT. Available freely at >ftp://ftp.cyclic.com > >HJB We use CVS as well for VxWorks BSP, driver and tool development. Set up and used properly, it is a powerful tool. I was here coding VxWorks before we implemented CVS, so I am qualified to say that it is a huge time saver and helps us create top-quality VxWorks products. In an environment with many developers, I would not want to work without it. Here is an overview of how we use CVS with VxWorks 5.3 (your situation may require something different): o All Tornado packages we receive from Wind River are loaded onto our file server and made "read-only". o For each Tornado tree, we have a ksh script that sets up the environment variables for using that set of Tornado tools and code in a personal directory. o A script is employed to soft-link the "host" and "share" directories of a read-only Tornado tree into the personal directory. The script also creates a copy of every directory under "target" and soft-links all of the files (this allows the user to modify things like usrConfig.c without disrupting others). o Only our BSP/driver source code is saved in CVS and is on our file server. In this environment, working on a personal and modifiable version of a BSP with Tornado involves these steps... 1. Make a clean directory and "cd" there. 1. Choose a Tornado tree and run its ksh script. 2. Check-out the BSP/driver code from CVS. 3. Run the script to "overlay" the Tornado tree. 4. Develop away! 5. Check-in improved BSP/driver code to CVS. 6. Delete the directory. ...and to believe I thought source code "control" was going to restrict my coding freedom. It has really done the opposite and has permitted me to work at a higher level. Good luck in your choice! Ted Friedl Senior Software Engineer Heurikon Corporation Madison, WI --------------------------- Newsgroups: comp.os.vxworks Subject: S: Backup programm for vxworks Date: 15 Aug 1996 19:59:49 +0200 From: alpha@jungle.toppoint.de (Sable) Organization: Jungle Industries said Balu... Message-ID: <4uvoil$13a@jungle.toppoint.de> Hello, i search for some tool tp backup my vxwork-disks on an remote computer. are there any simple tools on the internet or any helps? thank you Hendrik S. Roepcke --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Standard I/O read with a timeout Date: 15 Aug 1996 13:11:56 -0400 From: duhrenho@sanders.com (David Uhrenholdt) Organization: Sanders SSD Message-ID: References: <3210D651.41C67EA6@adc.com> Sender: duhrenho@hickory In article <3210D651.41C67EA6@adc.com> Sifang Chu writes: We are designing a HDLC driver to work with the VxWorks seven basic I/O functions. The VxWorks read() function is defined as: nBytes = read (fd, &buffer, maxBytes). We want to make our HDLC read to timeout if data is not available. What is the best way to pass a timeout counter to the HDLC driver? -- It would seem that using "select" would be an easy way to handle this. - -- Dave Uhrenholdt duhrenho@vette.sanders.lockheed.com Phone (603) 885-7398 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: YEAR 2000 Date: 15 Aug 1996 20:46:41 GMT From: vbv@giskard.cwru.edu ("Virgilio 'Dean' B. Velasco Jr.") Organization: Case Western Reserve Univ. Cleveland, Ohio (USA) Message-ID: <4v02bh$j9m@madeline.INS.CWRU.Edu> References: <4unkoi$5rr@niven.ksc.nasa.gov> In article <4unkoi$5rr@niven.ksc.nasa.gov>, Richard Price wrote: >What effect will the transition from 1999 to 2000 have on files in VxWorks? They'll be a year older. - -- Virgilio "Dean" Velasco Jr, Department of Electrical Eng'g and Applied Physics graduate student slave, roboticist-in-training and Q wannabe http://dora.eeap.cwru.edu/vbv/vbv.html | Remember: Wise men still seek Christ. "And I shall shed my light over dark evil, for darkness cannot stand the light." --------------------------- Newsgroups: comp.os.vxworks Subject: Re: What about LINUX?? Date: Thu, 15 Aug 1996 23:22:11 GMT From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star Message-ID: <4v0bf8$1lr@idiom.com> References: <9608150525.AA04342@rst.co.il> Reply-To: hjb@peacefulstar.oakland.ca.us leonid@rst.co.il (Leonid Rosenboim) wrote: >Linux - vote NO. Linux is the best UNIX I have ever used. I've ported enough commerical UNIX kernels around in my career to know that Linux is just as good, if not better. Of course, that doesn't mean that a "real company" would even consider porting their tools to Linux. Too bad. HJB --------------------------- Newsgroups: comp.os.vxworks Subject: Re: HTTP server for VxWorks Date: Thu, 15 Aug 1996 08:41:04 GMT From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star Message-ID: <4uunr4$lgs@idiom.com> References: <4t6di0$10b@mistral.physio-control.com> <4tja59$duv@idiom.com> Reply-To: hjb@peacefulstar.oakland.ca.us A lot of people asked for the toy HTTP server for VxWorks. It is in: http://peacefulstar.oakland.ca.us/vxhacks.html It's a hack, so don't expect much. It's useful for letting your embedded target respond to HTTP clients (netscape) for management functions, etc. HJB --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Source control with Tornado? Date: Thu, 15 Aug 1996 08:48:45 GMT From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star Message-ID: <4uuo9h$lgs@idiom.com> References: <4uniph$fu4$1@mhade.production.compuserve.com> Reply-To: hjb@peacefulstar.oakland.ca.us Jim Wilson <102105.3240@CompuServe.COM> wrote: >Can anyone recommend a Source Code management system that works >well with the Tornado for Windows? I use GNU RCS/CVS for windows NT. Available freely at ftp://ftp.cyclic.com HJB --------------------------- Newsgroups: comp.os.vxworks Subject: STL with Tornado/VxWorks Date: Tue, 13 Aug 1996 12:25:08 -0700 From: Jeff Eastwood Organization: Teleport - Portland's Public Access (503) 220-1016 Message-ID: <3210D694.2EF7@teleport.com> I'd like to hear from anyone that has used the Standard Template Library with Tornado or any other version of VxWorks. Either a port of the GNU STL, or use of a commercial product such as Object Space's version, or even a direct port of the HP reference implementation would be of interest. Jeff Eastwood ESI --------------------------- Newsgroups: comp.os.vxworks Subject: Re: YEAR 2000 Date: Wed, 14 Aug 1996 11:13:04 GMT From: Tom Fuda Organization: Northrop Grumman Norden Systems Message-ID: References: <4unkoi$5rr@niven.ksc.nasa.gov> Sender: usenet@tron.bwi.wec.com (USEnet News Poster) I don't suspect that the year 2000 will cause problems with VxWorks. The dosFsLib function, dosFsDateSet accepts dates through the year 2099. This function sets the date which the DOS filesystem uses for file timestamping purposes. Another option for file timestamping is to use the dosFsDateTimeInstall function to install a user-supplied date/time handler. Since you will most likely have to write your own user-supplied date/time handler you can take care to be 2000 compliant. I believe this is the only area where VxWorks cares about a date. Tom Fuda Senior Software Engineer Northrop Grumman Norden Systems fuda@norden.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Downloading vxWorks to PPC Date: 16 Aug 1996 02:38:42 GMT From: steved@pacifier.com@199.2.117.163 (Steve Doiel) Organization: Pacifier BBS, Vancouver, Wa. ((360) 693-0325) Message-ID: <4v0mvi$eit@news.pacifier.com> References: <4uvtus$bka$1@mhafc.production.compuserve.com> Reply-To: steved@pacifier.com (Steve Doiel) Keith Kingsley writes: >Has anyone had problems downloading the vxWorks image file from >NT to a PPC board? I'm using NT 3.51 as the host and a Motorola >1604 board as my target, and I can't seem to get the target board >to FTP properly. When I boot up the PPC board, it attaches the >network interfaces dc0 and lc0 fine, but when it needs to >download vxWorks, it just displays "Loading..." and hangs. I can >monitor FTP sessions on my NT machine, and I can see that the >target is successfully logging on. The PPC board seems to find >the file as well, for if rename or delete >D:\Tornado\target\config\mv1604\vxWorks, I get a "file not found" >message. Any suggestions? > We boot our MVME162 targets from NT using FTP. Thinks to look for when it doesn't work: Privileges of the "user" the target is loggin in as. If you're using NT's standard FTP server, you must set up what drives may be accessed using FTP. Also, from a different machine (if you have one available) try to FTP the same file logging in with the same user and password as the target system is using. Hope this helps, Steve Doiel steved@pacifier.com --------------------------- End of New-News digest ********************** From harvey@wrs.com Fri Aug 16 11:35:24 1996 From: Harvey Wong Date: Fri Aug 16 11:35:26 PDT 1996 Subject: hardware diagnostics Can anyone point me to to some general hardware diagnostic routines which could be executed upon booting of VxWorks? Thanks in advance. Harvey harvey@wrs.com From cpires@br.homeshopping.com.br Fri Aug 16 11:53:35 1996 From: Carlos Pires Date: Fri Aug 16 11:53:37 PDT 1996 Subject: DWord / Word alignment Hello VxWorkers, We am using VxWorks 5.2 for both SPARC and Motorola 68040 CPUs in different VME chassis. All our applications are developed on SPARC/Solaris 2.4 workstations using C++, compiled by Wind River C++ compiler. Everything is working fine except for the two-byte structure alignment for the Motorola 68040 CPU and four-byte alignment for SPARC. Both CPUs need to communicate and exchange structures. This word misalignment is a great nuisance for the communication between both CPUs. We have looked for a compiler option to enable four-byte alignment for 68040 CPU to no avail. Any ideas on how we could get the four-byte structure alignment from Wind C++ for Motorola 68040 ? Thank you in advance. Carlos Pires IES Systems From CJONES@digsys1.bwi.wec.com Fri Aug 16 12:36:11 1996 From: "CHRIS JONES" Date: Fri Aug 16 12:36:13 PDT 1996 Subject: Code Startup On Multiple Processors We have a multi board system using VxMP. Each processor has the same code image and starts up tasks depending on which board it resides on. What I would like to find is a better way to start up our main program on each board during the development cycle. Currently, I have to start a shell for each board and run a script file in each shell which downloads the code and starts the main program. This method is fine for two or three processors, however, we will be adding several more boards and this method is becoming laborious. Is there some way to download code and start it running on several boards at once? Thanks, Chris Jones cjones@digsys1.bwi.wec.com (410) 765-7959 (410) 993-8822 FAX Northrop Grumman Electronic Sensors & Systems Division From royg@semantic.com Fri Aug 16 14:44:21 1996 From: Roy Gordon Date: Fri Aug 16 14:44:23 PDT 1996 Subject: Re: DWord / Word alignment Another approach, since common understanding of differently aligned but semantically equivalent structures seems to be what's really involved... What about using the xdr compiler? For example, using rcpgen(1) to just generate the needed C source, if that's all that you would need. -- roy > Submitted-by cpires@br.homeshopping.com.br Fri Aug 16 11:53:35 1996 > Submitted-by: Carlos Pires > > Hello VxWorkers, > > We am using VxWorks 5.2 for both SPARC and Motorola 68040 CPUs in different > VME chassis. > All our applications are developed on SPARC/Solaris 2.4 workstations using > C++, compiled by Wind River C++ compiler. Everything is working fine except > for the two-byte structure alignment for the Motorola 68040 CPU and > four-byte alignment for SPARC. Both CPUs need to communicate and exchange > structures. This word misalignment is a great nuisance for the communication > between both CPUs. > We have looked for a compiler option to enable four-byte alignment for 68040 > CPU to no avail. Any ideas on how we could get the four-byte structure > alignment from Wind C++ for Motorola 68040 ? > > Thank you in advance. > Carlos Pires > IES Systems From royg@semantic.com Fri Aug 16 14:46:27 1996 From: Roy Gordon Date: Fri Aug 16 14:46:29 PDT 1996 Subject: Re: DWord / Word alignment Another approach, since common understanding of differently aligned but semantically equivalent structures seems to be what's really involved... What about using the xdr compiler? For example, using rcpgen(1) to just generate the needed C source, if that's all that you would need. -- roy > Submitted-by cpires@br.homeshopping.com.br Fri Aug 16 11:53:35 1996 > Submitted-by: Carlos Pires > > Hello VxWorkers, > > We am using VxWorks 5.2 for both SPARC and Motorola 68040 CPUs in different > VME chassis. > All our applications are developed on SPARC/Solaris 2.4 workstations using > C++, compiled by Wind River C++ compiler. Everything is working fine except > for the two-byte structure alignment for the Motorola 68040 CPU and > four-byte alignment for SPARC. Both CPUs need to communicate and exchange > structures. This word misalignment is a great nuisance for the communication > between both CPUs. > We have looked for a compiler option to enable four-byte alignment for 68040 > CPU to no avail. Any ideas on how we could get the four-byte structure > alignment from Wind C++ for Motorola 68040 ? > > Thank you in advance. > Carlos Pires > IES Systems From daemon@csg.lbl.gov Sat Aug 17 04:00:45 1996 From: daemon@csg.lbl.gov Date: Sat Aug 17 04:00:47 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sat Aug 17 04:00:42 PDT 1996 Subject: I want GCC wchar_t to be 16-bit not 32-bit Subject: Re: What about LINUX?? Subject: Re: What about LINUX?? Subject: Re: Ada and VxWorks Subject: Re: LonTalk Subject: Re: Help on building gcc/ld/as cross-comp Subject: Re: What about LINUX?? ------------------------------------------------------- Newsgroups: comp.lang.c,comp.os.vxworks,gnu.gcc.help Subject: I want GCC wchar_t to be 16-bit not 32-bit Date: Fri, 16 Aug 1996 12:37:53 +0100 From: Steve James Organization: Wandel & Goltermann Ltd Message-ID: <32145D91.A60@wago.de> If I compile the following under GNU CC (v2.6-95q2): wchar_t string[]=L"Hello world\n"; then I get an initialised array where each element of the array is, in this particular compiler, a 32-bit quantity. Clearly this falls within the ANSI standard for implementing wchar_t. But because I am working on an embedded system where memory may be short I am unhappy about filling the memory with large quantities of zeros. I would therefore like the implementation of wchar_t to be as a 16-bit quantity. Does anyone know if this is possible? One solution would be if it was possible to specify the size of the wchar_t type on the command line? An alternative approach to get what I want is to create my own type but then I cannot initialise an array as in my first example because there is no syntax which will work correctly. I would ideally like to be able to use the following: typedef unsigned short WCHAR_T; WCHAR_T string[]=x"Hello world\n"; where x represents a suitable character which tells the compiler to output an array of unsigned short elements. The only solution I have found to get what I want is: typedef unsigned short WCHAR_T; WCHAR_T string[]={'H','e','l','l','o',' ','w','o','r','l','d','\n','\0'}; which is really tortuous to type and fraught with difficulties including having to remember to manually specify the null terminator on the end. What I actually want is a primitive type which is understood by the compiler in a way which allows arrays of this type to be initialised but a type which I know to be implemented as a 16-bit object. Has anyone encountered this problem before? Does anybody have a better solution? Paul Lumb Replies by e-mail please: paul.lumb@wago.de --------------------------- Newsgroups: comp.os.vxworks Subject: Re: What about LINUX?? Date: Fri, 16 Aug 1996 08:09:38 GMT From: Mark Zusman Organization: Pointer Software Systems Message-ID: <32142CC2.109F732E@pointer.israel.net> References: <199608122125.VAA01187@ziv.me.umn.edu> <4usp9v$2ct@ncar.ucar.edu> Sender: news@actcom.co.il (News) Rich Neitzel wrote: > > |> > |> > |> Any word on if WindRiver is going to port VxWorks or Tornado to LINUX OS?? > |> > |> We would be EXTREMELY interested in this. > |> > As would we. > > -- > Richard Neitzel thor@atd.ucar.edu Torren med sitt skjegg > National Center For Atmospheric Research lokkar borni under sole-vegg > Box 3000 Boulder, CO 80307-3000 Gjx'i med sitt shinn > 303-497-2057 jagar borni inn. While I worked as contractor in Kulicke & Soffa Israel I've compiled gcc tools suite on Linux (and HP-UX 9.x for that matter) and it works perfectly till now - in fact KSL engineers bring Linux notebook to field so they can fix bugs/debug on customer's sites. The latest releases of binutils and gcc has wrs-68k-vxworks as legitimate target. So I built cross binutils (ld/as/nm) from linux to vxworks, then I took libgcc1.a from Sun set of vxworks tools and compiled gcc 2.7.x as crosscompiler. Then I've built gdb-4.15 as crossdebugger and tuned different frontends (DDD and xxgdb) to work with it - the whole setup work flawlessly (I worked with VxWorks 5.1.1) BTW, g++ worked too, just without using streams/STL - I had some problems cross-compiling libg++ to VxWorks. I'm out of KSL for 10 monthes now but I know that these tools are alive and well. I think the person who is in charge of them is Gil Katz - kgill@ksl.co.il; he should be able to answer your questions. I'm rather experienced sysadm/build manager so it took about a day to setup the whole thing - it went really well. Mark. - --- Mark Zusman marklz@pointer.israel.net Senior Software Engineer Pointer Software Systems Inc. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: What about LINUX?? Date: 16 Aug 1996 15:53:50 +0930 From: wes@spock.atlantek.com.au (Wesley Hosking) Organization: Atlantek Microsystem Pty. Ltd, Adelaide, Australia. Message-ID: References: <9608150525.AA04342@rst.co.il> <4uvf93$aog@xmission.xmission.com> Another vote here (for doing debugging in the field) for having VxWorks supported under Linux.... wes - -- - ------------------------------------------------------------------------------ Wesley Hosking Atlantek Microsystems Endeavour House, Technology Park, South Australia wes@atlantek.com.au +61-8-260-8990 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Ada and VxWorks Date: 13 Aug 1996 23:59:21 GMT From: scott@tcville.es.hac.com Organization: Hughes Aircraft Message-ID: <4ur4sp$8k7@hacgate2.hac.com> References: <4uni5s$g66@news.res.ray.com> Reply-To: scott@tcville.es.hac.com In article g66@news.res.ray.com, milesm@tif531.ed.ray.com (Michael Miles) writes: >I have had a great deal of success with the Green Hills builder tool >set with Ada, C and C++. They are also helpful with VxWorks problems. >They can be reached at (813)781-4909. .. >johncoby@ghgcorp.com says... >>In article <4ucind$5ft@lfsserv1.fsc.ibm.com>, tom.kodey@lmco.com wrote: >>> I'm considering VxWorks as the RTOS for an Ada based imbedded application and >>> am considering various development environments. The target will likely be a >>> Pentium, PowerPC or R4xxx. I'm interested to hear what Ada development >>> environment you were using and, briefly, what you thought of it (host, ease of >>> interfacing with VxWorks, ease of use, maturity, support, host and target >>> based debug, etc.)? .. >>Hum....do you have to use Ada? (Ok, enough Ada bashing from an ex-Ada >supporter) .. >>I have used VADSworks from Verdix, now Rational. It was a very adequate >>set of tools except... .. >>My experience with Rational has been TRY BEFORE YOU BUY! With past experience with Verdix and current experience with GreenHills+VxWorks I'd give the nod to GH for your effort. Their environment would more easily allow you to define target platforms, manage files, etc.. The integrated debugger allows for a host-side symbol table and works quite well save for its variable display feature (update is quirky and register values never seem to be updated [tested on an AIX host, RAD6k target]). Maturity is a major plus for Verdix, GH I'd rate as late-beta. I have a difficult time grading any compiler as post-beta so you should definitely decide for yourself and, as said above, "try before you buy." Scott - --- Hughes Aircraft Co. voice: (310) 616-1059 Image and Signal Processing Lab addr: scott@tcville.es.hac.com PO Box 902, EO/E01/A172, El Segundo, Ca. 90245 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: LonTalk Date: Thu, 08 Aug 1996 08:33:46 -0700 From: greg bruno Organization: ncr classic Message-ID: <320A08DA.213@sandiegoca.ncr.com> References: <199608060805.AA01984@mail.wrs.com> namiq@wrs.com wrote: > > I am looking for a LonTalk Adapter (ISA or PC104 card) that has a VxWorks > driver. > > Any recommendations appreciated. give the folks at engenuity systems a call: 602.275.3363 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Help on building gcc/ld/as cross-comp Date: 15 Aug 1996 12:18:35 -0400 From: duhrenho@sanders.com (David Uhrenholdt) Organization: Sanders SSD Message-ID: References: <4u5ac0$nhm@hacgate2.hac.com> Sender: duhrenho@hickory I use a line similar to ./configure --target=sparc-vxworks as, ld, nm, ar, etc, and gdb all compile without any extra intervention. I haven't done it in a while so I might have left something out. I did it without access to anything from Wind River. Also, the only Linux machine I have access to is my home machine, and I just don't do that much work from home :), so I haven't tested them out thoroughly. And as pointed out else where there are problems using gdb 4.16 with VxWorks. Quote from bowman@waterloo.hp.com (Don Bowman) I had to make 1 change to get it to work. If you configure everything, and run make, you should be able to use it, but you may find you have to hit '^c' on attach. If you don't, forget this. If you do, then add this line to gdb/gdb/config/sparc/tm-vxsparc.h /* Don't do wait_for_inferior on attach. */ #define ATTACH_NO_WAIT Building ccsparc is a bit of a problem. It appears that for the objective C compiler it wants to link-in the C library or some such thing. Again I don't recall exactly. To fix this you remove the objC target from the make file. What you are missing is some of the tools specifically for vxworks such as xsym. Fortunately, some other tools such as WindView (which I have not actually tried) are written in tcl and somewhat more portable. Of course you should be sure you are not violating any part of your licensing agreement when you use it. - -- Dave Uhrenholdt duhrenho@vette.sanders.lockheed.com Phone (603) 885-7398 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: What about LINUX?? Date: 16 Aug 1996 09:43:41 -0400 From: Dave Uhrenholdt Organization: Lockheed Sanders Message-ID: References: <9608150525.AA04342@rst.co.il> <4uvf93$aog@xmission.xmission.com> bgeer@xmission.xmission.com (bgeer) writes: > > leonid@rst.co.il (Leonid Rosenboim) writes: > > >> Any word on if WindRiver is going to port VxWorks or Tornado to LINUX OS?? > > >Who do you call then if something to do with Linux does not work ? > > Ghost-Busters ! > > No, you "call" the appropriate one of a large number of Linux > newsgroups. I dare say the Linux developers & users are a more > responsive group than *any* single-vendor OS. > > Of course there's no tech support phone number that I know of for > Linux, but then my experience using phone support has been much worse > than support I've received using email & newsgroup solicitations. You can buy tech support from Redhat, Cygnus, etc. > > >Seriously, unsupported host operating systems can cause a big > >support headache for Wind River, cause all problems will end up > >bouncing back to them. God knows how many problem reports related > >to e.g. "booting from Solaris" have been handeled, even the exploder > >had its share. [SNIP]> > I installed & configured Linux on a ProStar Pentium laptop for > supporting VxWorks on a MVME162LX on a field delivery. It took about > 2 hours using cdrom. I enabled tftp, fired up the Ethernet PCMCIA > card, & actually booted from it, as well as wrote flash ram & was able > to do everything else I needed to except compile. All I needed was a > cc68k that executed on x86 & I would have had a complete portable > development system. This laptop cost less than $4000 US. Say that > about a Solaris machine...:-) I have come close to getting a full development system, untested, but compiled for HP-UX. I just haven't tried with Linux. Part of the problem is that Wind River doesn't support the Freely available development tools that it ships with the VxWorks. [SNIP] > >Just imageine what it would be like, and imageine the whole support > >group will get even more work load then it has now, and you know > >who will be effected ... > > No doubt any new platform/OS combination would increase support > demand. I'll bet the support required by Linux users would not be > significantly more than that required by other OS's; actually, I'll > bet it would be significantly *less*. As I said, there support for the freely available tools is week to begin with. I don't see how added Linux will significantly increase the load. > > Bob > -- > <> Robert Geer & Donna Tomky / * <> > <> bgeer@xmission.com | _o * o * o <> > <> Salt Lake City, Utah | -\<, * <\ > <> U S A | O/ O __ /__, /> <> --------------------------- End of New-News digest ********************** From lyorg@scorpio.com Sat Aug 17 23:45:56 1996 From: lyorg@scorpio.com (Lyor Goldstein) Date: Sat Aug 17 23:45:58 PDT 1996 Subject: Re: What about Linux I have alos ported my VxWorks environment to Linux (no thanks to WRS...). It is rather simple, though there are some minor issues to be resolved. I fully agree with what has been said here about Linux, and I suggest that WRS take it under advisement to listen to the needs of their users/customers, instead of trying to force their own views on us.... Lyor G. From daemon@csg.lbl.gov Sun Aug 18 04:00:35 1996 From: daemon@csg.lbl.gov Date: Sun Aug 18 04:00:37 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sun Aug 18 04:00:32 PDT 1996 Subject: Re: DWord / Word alignment Subject: Re: What about LINUX?? Subject: Vxworks and imake ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: DWord / Word alignment Date: 17 Aug 1996 03:06:39 GMT From: steved@pacifier.com@199.2.117.163 (Steve Doiel) Organization: Pacifier BBS, Vancouver, Wa. ((360) 693-0325) Message-ID: <4v3cvv$27n@news.pacifier.com> References: <199608161850.PAA10348@piraque.homeshopping.com.br> Reply-To: steved@pacifier.com (Steve Doiel) Carlos Pires writes: >Hello VxWorkers, > >We am using VxWorks 5.2 for both SPARC and Motorola 68040 CPUs in different >VME chassis. >All our applications are developed on SPARC/Solaris 2.4 workstations using >C++, compiled by Wind River C++ compiler. Everything is working fine except >for the two-byte structure alignment for the Motorola 68040 CPU and >four-byte alignment for SPARC. Both CPUs need to communicate and exchange >structures. This word misalignment is a great nuisance for the communication >between both CPUs. >We have looked for a compiler option to enable four-byte alignment for 68040 >CPU to no avail. Any ideas on how we could get the four-byte structure >alignment from Wind C++ for Motorola 68040 ? > >Thank you in advance. >Carlos Pires >IES Systems > I have difficulty with the common practice of defining a program language structure and sharing the source code for the structure between machines. IMHO in the long run you are better off to define the data structures that will be communicated in an architecture independent fashion, and then making both sides of the interface adhere to the architecture. If you follow this practice you never encounter the types of problems you are describing. You may incur a small amount of overhead in the translation, but few applications I have seen couldn't stand that overhead. I am currently working on an application where we share data between different machines with different architectures via a TCP/IP network. Everything is defined in an interface specification, and structures are defined indepently on each machine. It has been working well for us. Steve Doiel steved@pacifier.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: What about LINUX?? Date: Thu, 15 Aug 1996 11:03:37 -0600 From: Aseem Bakshi Organization: U S WEST - !NTERACT Services Message-ID: <32135869.54F1@phbtsus.com> References: <199608122125.VAA01187@ziv.me.umn.edu> <4usp9v$2ct@ncar.ucar.edu> Rich Neitzel wrote: > > In article <199608122125.VAA01187@ziv.me.umn.edu>, "Ted Morris" writes: > |> > |> > |> Any word on if WindRiver is going to port VxWorks or Tornado to LINUX OS?? > |> > |> We would be EXTREMELY interested in this. > |> > As would we. > > -- > Richard Neitzel thor@atd.ucar.edu Torren med sitt skjegg > National Center For Atmospheric Research lokkar borni under sole-vegg > Box 3000 Boulder, CO 80307-3000 Gjx'i med sitt shinn > 303-497-2057 jagar borni inn. We would be intereted. Hope WindRiver has somebody reading this grou[ - -- - ----------------------------------------------------------------------------- Aseem Bakshi Phone: (801)9771574 Philips Broadcast Television Systems Inc Alias: Polka Boy Salt Lake City, UT E-mail Abakshi@phbtsus.com --------------------------- Newsgroups: comp.os.vxworks Subject: Vxworks and imake Date: Thu, 15 Aug 1996 10:59:08 -0500 From: Shreya Bhat Organization: BNR, richardson, tx Message-ID: <3213494C.5BBA@bnr.ca> Hi, Can I use imake to compile and build the libraries for the VxWorks target? I wanted to know if the imake supports the munching --- I think the munching is done on the object files. It looks at the constructors for the static objects and compiles and links the *.c files for the static object constructors. This may not be very clear explanation of munching! Please cc me to the reply. Regards, Shreya --------------------------- End of New-News digest ********************** From MikeMcQuade@prometheus.connectsoft.net Sun Aug 18 18:20:09 1996 From: Mike McQuade Date: Sun Aug 18 18:20:12 PDT 1996 Subject: Re: What about Linux At 11:46 PM 8/17/96 PDT, you wrote: >I suggest that WRS take it under >advisement to listen to the needs of their users/customers, instead of trying > to force their own views on us.... I think they would be wise to be the pioneer in this area, Ive recently seen 200Mhz Pentium Pro machines that would run circles around any Sun or Hp box. The latest Linux 2.0 is damn fast, supports about any hardware out there, has the GNU tools, has TCL/TK, turns Pentium machines into credible workstations, costs about $6.00.... I can see it now, WRS puts Linux 2.0 on a CD with the Tornado tools, becoming the first Real Time OS vendor to ship their tools with a host operating system. I think they should do the port as a matter of innovation. Don't let Lynx or pSOS be first, and you know they (ISI, LYNX) read this newsgroup. If WRS wont give thir own customers what they want, their competition, looking for any opportunity to catch up, will be glad to. there Ive done my part to keep this stirred up :) Mike McQuade MikeMcQuade@connect.com From daemon@csg.lbl.gov Mon Aug 19 04:00:24 1996 From: daemon@csg.lbl.gov Date: Mon Aug 19 04:00:26 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Mon Aug 19 04:00:20 PDT 1996 Subject: Re: Time synchronisation Subject: Re: Handling statics with C++ Subject: Re: A32 space VME access ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Time synchronisation Date: 16 Aug 1996 15:56:50 -0700 From: mwette@mr-ed.jpl.nasa.gov (Matt Wette) Organization: Jet Propulsion Laboratory Message-ID: <4v2ubi$gu2@mr-ed.jpl.nasa.gov> References: In article , Hiske Huiting wrote: >I have Sparc board with VxWorks 5.2 connected to a Silicon Graphics Challenge >through VME and TCP/IP. > >What is the easiest way to synchronise the VxWorks time with the SGI? /* rdate.c - set time remotely * * $Id: rdate.c,v 1.1 1995/09/27 04:03:35 mwette Exp $ */ #include #include #include #include #include #include #include #include #include #include static char const rcsid[] = "$Id: rdate.c,v 1.1 1995/09/27 04:03:35 mwette Exp $"; #define IPPORT_TIMESERVER 37 extern char sysBootParams; int rdate(void) { struct sockaddr_in sockaddr; struct timespec tp; int sock, n, size, status; unsigned long t; char *bufr; static char *BOOT_HOST = &sysBootParams + 20; memset(&sockaddr, '\0', sizeof(sockaddr)); sockaddr.sin_family = AF_INET; sockaddr.sin_port = htons(IPPORT_TIMESERVER); sockaddr.sin_addr.s_addr = hostGetByName(BOOT_HOST); sock = socket(AF_INET, SOCK_STREAM, 0); if (sock == -1) { perror("rdate: socket() failed"); return -1; } status = connect(sock, (struct sockaddr*)&sockaddr, sizeof(sockaddr)); if (status == -1) { close(sock); perror("rdate: connect() failed"); return -1; } bufr = (char *)&t; size = sizeof(t); while (size) { n = read(sock, bufr, size); if (n < 0) { close(sock); perror("rdate: read() failed"); return -1; } size -= n; bufr += n; } close (sock); /* Adjust for 70 difference (1900->1970) and 1 sec delay. */ tp.tv_sec = (ntohl(t) - 2208988800L) + 1; tp.tv_nsec = 0; clock_settime(CLOCK_REALTIME, &tp); return 0; } /* --- last line of rdate.c --- */ - -- matthew.r.wette@jpl.nasa.gov --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Handling statics with C++ Date: Mon, 19 Aug 1996 15:28:56 +1000 From: Earl Chew Organization: Hewlett-Packard Co. Message-ID: <3217FB98.49B3@hpato.aus.hp.com> References: <9608130013.AA29767@clix.jna.com.au> <32108499.21EF@ucar.edu> Joe Van Andel wrote: > // return pointer to single instance of this object/task. > A1Record *A1Record::Instance() > { > if (!instance_) { > instance_ = new A1Record(); > } > return instance_; > } > ------ > The advantage of this strategy is that rather than having to maintain a > global pointer to some key service object, or having to explicitly pass > an object pointer to every task/class that uses it, any code that needs > the service calls: > A1Record *ptr = A1Record::Instance(); > > I hope this advice helps - feel free to email me if you need more > information. The only problem here is that there is a potential for two or more tasks to race. There is a small probability that the race is a draw and multiple instances of A1Record get created. We've found that there are in fact two kinds of singletons: 1. Thread singletons. Each thread gets a singleton. This is most easily achieved by mixing the above pattern with task variables (thread keys in Posix). When the number of task variables grows, this simple strategy becomes cumbersome, and some way is required to amalgamate all the task variables into a single `clump' so that the task switch time becomes independent of the number of task variables. 2. Process singletons. This is more akin to the code snippet provided above. All threads share the same singleton. This needs some kind of lock to prevent multiple tasks racing to create the singleton, or can use the C++ global ctor code to ensure that the singleton is created *before* main() starts. Earl - -- +----------------------------------------------------------------------+ | Earl Chew Email: earl_chew@aus.hp.com | | Hewlett-Packard Australia Tel: +61 3 9210 5528 | | Australian Telecom Operation Fax: +61 3 9210 5550 | | 347 Burwood Highway GPS: 37 51'46"S 145 10'03"E | | Burwood East, Victoria 3151, Australia | +----------------------------------------------------------------------+ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: A32 space VME access Date: Fri, 09 Aug 1996 17:01:33 -0500 From: Debbie Buscher Organization: NASA/JSC Message-ID: <320BB53D.1AE6@gp902.jsc.nasa.gov> References: <320B49F7.2781E494@motown.lmco.com> Michael J. Dempsey wrote: > > Hello > > We have a PowerPC (MV1603) running with a beta version of VxWorks 5.3 > embedded in VME chassis that contains two other VME boards with memory > registers that we would like our VxWorks application to read and write > to. One of the boards uses A16 address space and has a base address of > 0x8000. The other uses A32 address space and has a base address of > 0x00300000. > For A16 space board we used sysBusToLocalAdrs to map its memory to > something our PowerPC VxWorks application could use. It returned (if > memory serves me correctly) 0xefff8xxx (where xxx is an offset from > the base address). This worked fine, we read from that board's > memory and also wrote to it. > We tried using the same call to access memory on the board with > A32 address space (changing the first parameter for sysBusLocalAdrs > from VM_SUP_SHORT_IO to VM_EXT_SUP_DATA to indicate the different > address space). For an input of 0x00300000 to sysBusLocalAdrs we get > returned and address of 0xd0300000, but when the program tried to > access that memory location it blew up. We have a bus analyzer in > the VME rack set to trigger on anything but it never did, so the > application seems to have blown up before even trying to access > the VME memory.Is there something I need to set up or include in > the kernel to access this off-board memory? Any help is appreciated. > > Thanx, Mike D. > dempsey@motown.lmco.com You might try setting SM_OFF_BOARD to TRUE in config.h - WRS support told me to do this when I was attempting something similar. Are you using vxMemProbe to do your reads and writes? If so, are you protecting yourself against the call failing due to bus errors? Debbie Buscher --------------------------- End of New-News digest ********************** From jjjurek@fly.aeromet.com Mon Aug 19 15:42:13 1996 From: jon jurek Date: Mon Aug 19 15:42:15 PDT 1996 Subject: integrating removable media drives I am interested in integrating removable media drives such as Iomega Bernoulli drives, Zip drives, or Jaz drives into a vxWorks system with MVME177 CPUs. Does anyone have any tips or ideas. Thanks a bunch. -- ______________________________________ Jon Jurek Aeromet, Inc. Software Group Leader 112 Beechcraft Drive jjjurek@fly.aeromet.com Jones Riverside Airport (918) 299-2621 Tulsa, OK 74132 http://www.aeromet.com/ From MikeMcQuade@connect.com Mon Aug 19 21:26:38 1996 From: MikeMcQuade@connect.com Date: Mon Aug 19 21:26:40 PDT 1996 Subject: Re: integrating removable media drives At 03:42 PM 8/19/96 PDT, you wrote: >Submitted-by jjjurek@fly.aeromet.com Mon Aug 19 15:42:13 1996 >Submitted-by: jon jurek > >I am interested in integrating removable media drives such as Iomega >Bernoulli drives, Zip drives, or Jaz drives into a vxWorks system with >MVME177 CPUs. Does anyone have any tips or ideas. > We are also interested in Zip / Jazz drives (SCSI) for use with a MVME-1603/4 system. Mike McQuade MikeMcQuade@connect.com From daemon@csg.lbl.gov Tue Aug 20 04:00:16 1996 From: daemon@csg.lbl.gov Date: Tue Aug 20 04:00:18 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Tue Aug 20 04:00:13 PDT 1996 Subject: Flash File System? ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Flash File System? Date: Mon, 19 Aug 1996 15:12:10 +0200 From: Stefan Eifert Organization: Siemens Message-ID: <3218682A.755F@mch.scn.de> Hi, we are currently discussing the use of a flash file system for booting up vxworks on our target. Does anybody know if there is a ready-to-use system out there on the market or how difficult the coding would be? Thanks in advance! --------------------------- End of New-News digest ********************** From cpires@br.homeshopping.com.br Tue Aug 20 06:57:45 1996 From: Carlos Pires Date: Tue Aug 20 06:57:46 PDT 1996 Subject: Re: DWord / Word alignment Bruce Wedding, At 21:51 17/08/96 GMT, you wrote: >In comp.os.vxworks >Carlos Pires wrote: > > >>We have looked for a compiler option to enable four-byte alignment for 68040 >>CPU to no avail. Any ideas on how we could get the four-byte structure >>alignment from Wind C++ for Motorola 68040 ? > >You just have to pad all your structs out with chars, no big >deal really. > >Assuming 4 byte ints: > >struct mystruct { > char array[50]; /* need 2 bytes to be multiple of 4 */ > char padding[2]; > unsigned short x; /* only 2 bytes */ > char unused [2]; /* only here to pad */ >} test; > > >See? > >Bruce Wedding > > > Thanks for your answer, but in fact we were looking for another solution because to do what you kindly suggested, we would have to modify almost all of our header files and structures definitions. We have already temporarily modified our system to "translate" the messages exchanged between both CPUs, SPARC and Motorola 68040. Perhaps you have not had the opportunity to read my last e-mail with a better short definition of our problem, which I trancribe below: >>Hello VxWorkers, >> >>We am using VxWorks 5.2 for both SPARC and Motorola 68040 CPUs in different VME >>chassis. >>All our applications are developed on SPARC/Solaris 2.4 workstations using C++, >>compiled by Wind River C++ compiler. Everything is working fine except for the >>two-byte structure alignment for the Motorola 68040 CPU and four-byte alignment for >>SPARC. Both CPUs need to communicate and exchange structures. This word misalignment >>is a great nuisance for the communication between both CPUs. >>We have looked for a compiler option to enable four-byte alignment for 68040 CPU to >>no avail. Any ideas on how we could get the four-byte structure alignment from Wind >>C++ for Motorola 68040 ? >> >>Thank you in advance. >>Carlos Pires >>IES Systems We have the majority of our application header files (especially those related to the sw lower layers of both systems), and some of the code as well, commom to both SPARC (5VT) and Motorola 68040 targets, enhancing portability and easier troubleshooting. We have encountered this problem only since we went past the VxWorks development phase host(Sun Station) / target(SPARC 5VT xor Motorola). Now both types of targets need to exchange lots of real-time data with each other. What we do really need is a code directive or a compiler option to instruct the Wind River C++ compiler to extend 2 bytes ints to 4 bytes on the Motorola side, since the contrary does not seem to be worthy trying. Regards to all, Carlos Pires IES Systems From Robert.Crispen@hsv.boeing.com Tue Aug 20 12:01:31 1996 From: "Crispen, Bob" Date: Tue Aug 20 12:01:35 PDT 1996 Subject: RE: DWord / Word alignment I could be a meanie and suggest you switch to Ada: type uchar is range 0..255; for uchar'size use 8; -- usually not needed, but what the heck type ushort is range 0..65535; for ushort'size use 16; -- ditto -- A teensy bit trickier to specify integers, so let's just -- assume type "integer" is 32 bits long rather than having -- to explain some non-straightforward looking stuff. -- Actually in Ada95 we'd probably use modular types all -- around. type uchar_array is array (1..50) of uchar; -- or 0..49 if you like that better type foo is record bar : uchar_array; baz : integer; bap : ushort; end record; for foo use record at mod 4 bar at 0 range 0..399; -- a couple of bytes pad here that you don't need to declare baz at 52 range 0..31; bap at 56 range 0..15; end record; for foo'size use 60*8; -- not really needed, either -- And if you declare an array of these puppies, since you've -- bounded the type at mod 4, each one will be automatically -- padded 2 bytes at the end. (from memory, but you get the gist even if I've messed up the grammar) Horses for courses: if you need to exactly specify the layout of data structures, you *might* consider using a language that was designed to do that instead of a language that wasn't. Note that you could even specify a monstrosity like: type foo is record bar : uchar; baz : integer; end record; for foo use record at mod 4 bar at 0 range 0..7; baz at 1 range 0..31; end record; That would align foo_object.baz at 1, 5, 9, and d -- and Ada would generate the code to legally manipulate that number on your processor. For example, on a SPARC, it generates the ugliest combinations of loads and shifts you ever saw. Slooooow! But it would generate legal code, in case the folks who designed your interface are idiots. In C and C++ I don't think this problem could be coped with at all. Alternately, with a weaker language, xdr makes a good deal of sense, and also solves the endian problems you'll have if you need to reuse the code. >Bob Crispen >bob.crispen@boeing.com >Speaking for myself, not my company > From leonid@rst.co.il Wed Aug 21 01:15:45 1996 From: leonid@rst.co.il (Leonid Rosenboim) Date: Wed Aug 21 01:15:47 PDT 1996 Subject: Re: Flash File System > Subject: Flash File System? > Date: Mon, 19 Aug 1996 15:12:10 +0200 > From: Stefan Eifert > Organization: Siemens > Message-ID: <3218682A.755F@mch.scn.de> > > Hi, > > we are currently discussing the use of a flash file system for > booting up vxworks on our target. > ... Dear Mr. Eifert, We sell such a product, FlashLib, it includes a full sclae dynamic Flash File System based on Microsoft FFS specifications. This product us being used already at a numer of projects with VxWorks. On the subject of booting, the solution depends on the architecture. In the case of 68k, SPARC, i960 etc embedded designs, we offer a scaled down read-only configuration of the file system, that can be built into the boot ROM, and initialized in a similar way to the SCSI and IDE booting support in VxWorks. Instead of initializing an IDE driver, you initialize a FLash driver (that we supply in source code), and instead calling dosFsInit(), you call flashFsDevCreate(). Also, the boot ROM can be a separate device, or it can reside on the same Flash as the file system, but the boot ROM space will be excluded from File System usage. We also have recently developed a spcific solution for PC 386/486 targets. It is ina form of an ISA or PC104 card. It's first 64K sector is "reserved" and programmed with a BIOS extention code, which contains a 4KByte TinyFFS implementation. It will switch to protected mode and load a specific file from the File System into RAM and jump there. Essentially, you plug in a ISA card in the PC target, switch the power on, and viola! vxWorks boots up. This way you can have a REAL EMBEDDED PC, without any magnetic storage, but still be able to store files. Visit our Web site for complete product information. > Does anybody know if there is a ready-to-use system out there > on the market or how difficult the coding would be? Usually it takes about one man-month to program and test a simple Flash implementation. e.g. a driver and a utility that can burn it with executable code. FlashLib price is roughly the same as one man-month of engineering, and offers a complete solution, with support for assorted flash chips and cards, and constantly updated and enhanced. ----------------------------------------------------------------------- Leonid Rosenboim Phone: +972-3-559-8144 R S T Software Industries Ltd. Mobile: +972-50-307-142 P.O.Box 11502, AZUR 58017, Israel Fax: +972-3-559-8244 WWW: http://www.rst.co.il E-Mail: leonid@rst.co.il From uucp@relay1.fnet.fr Wed Aug 21 01:48:18 1996 From: leverone@zeus.egt.fr (Philippe Leverone) Date: Wed Aug 21 01:48:20 PDT 1996 Subject: Re: DWord / Word alignment Carlos Pires wrote: ] ] ..., but in fact we were looking for another solution ] because to do what you kindly suggested, we would have to modify almost all ] of our header files and structures definitions. We have already temporarily ] modified our system to "translate" the messages exchanged between both CPUs, ] SPARC and Motorola 68040. Perhaps you have not had the opportunity to read ] my last e-mail with a better short definition of our problem, which I ] trancribe below: ] ] >>Hello VxWorkers, ] >> ] >>We am using VxWorks 5.2 for both SPARC and Motorola 68040 CPUs in ] different VME >>chassis. ] >>All our applications are developed on SPARC/Solaris 2.4 workstations using ] C++, >>compiled by Wind River C++ compiler. Everything is working fine ] except for the >>two-byte structure alignment for the Motorola 68040 CPU and ] four-byte alignment for >>SPARC. Both CPUs need to communicate and exchange ] structures. This word misalignment >>is a great nuisance for the ] communication between both CPUs. ] >>We have looked for a compiler option to enable four-byte alignment for ] 68040 CPU to >>no avail. Any ideas on how we could get the four-byte ] structure alignment from Wind >>C++ for Motorola 68040 ? ] >> ] >>Thank you in advance. ] >>Carlos Pires ] >>IES Systems ] ] We have the majority of our application header files (especially those ] related to the sw lower layers of both systems), and some of the code as ] well, commom to both SPARC (5VT) and Motorola 68040 targets, enhancing ] portability and easier troubleshooting. We have encountered this problem ] only since we went past the VxWorks development phase host(Sun Station) / ] target(SPARC 5VT xor Motorola). Now both types of targets need to exchange ] lots of real-time data with each other. What we do really need is a code ] directive or a compiler option to instruct the Wind River C++ compiler to ] extend 2 bytes ints to 4 bytes on the Motorola side, since the contrary does ] not seem to be worthy trying. Alignment compiler option exists for GNU C compiler. This is __attribute__ ((aligned(n))). Example : #include struct { char c1; long l1 __attribute__ ((aligned(4))); char c2; long l2; } s1; toto() { printf("&c1 = %#04x\n", &s1.c1); printf("&l1 = %#04x\n", &s1.l1); printf("&c2 = %#04x\n", &s1.c2); printf("&l2 = %#04x\n", &s1.l2); } Outputs on a 68030 target : -> ld toto &c1 = 0x3c6484 &l1 = 0x3c6488 &c2 = 0x3c648c &l2 = 0x3c648e value = 15 = 0xf Check up your C++ manual for such an option, ... or write your communication tasks in C. Good luck (we also have such problems between Motorola and Intel processors :-( +------------------------------------------------------------------------------+ | /\ | | / \ /\ Philippe Leverone, ingenieur logiciel | | /\ / \/ \ ;_o ALPLOG SA, groupe ALTRAN | | / \ / / \OL/ 4 rue du Moulin, BP 23 - Wolfisheim | | / \/ / \O 67032 STRASBOURG cedex | | / / / \ chez EGT SA FRANCE | | / / \ [*] | | \ e-mail : pleverone@egt.fr ,_"_ | | [_(O)] | +------------------------------------------------------------------------------+ From daemon@csg.lbl.gov Wed Aug 21 04:00:36 1996 From: daemon@csg.lbl.gov Date: Wed Aug 21 04:00:38 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Wed Aug 21 04:00:32 PDT 1996 Subject: Re: Handling statics with C++ Subject: PowerPC patches to gcc? Subject: Memory management Subject: Motorola 68030 VME card available ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Handling statics with C++ Date: 21 Aug 1996 01:46:52 GMT From: lijx@nt.com (Jin Li) Organization: Northern Telecom Inc., Mission Park Message-ID: <4vdpqc$i4t@nrchh52.rich.nt.com> References: <9608130013.AA29767@clix.jna.com.au> <4ups46$o2v@news.sei.cmu.edu> In article <4ups46$o2v@news.sei.cmu.edu>, Theodore Marz wrote: > >In article <9608130013.AA29767@clix.jna.com.au>, "Chris Ford" writes: >|> We are currently using VxWorks in a C environment, and are considering a move >|> to C++. One concern we have, however, is the treatment of statics. Under >|> VxWorks, if we wish to start several instances of the same task, and we wish >|> each instance to have its own copy of a static, we need to use taskVarAdd(). >|> This has been fine so far as we can combine the necessary statics into a single >|> structure and only tasVarAdd() a single pointer to that structure. This >|> minimises the hit on context switch time. >|> >[snip] >|> >|> The question is: what do people do about this? >|> >|> Many thanks in advance for any ideas. >|> >|> -- >|> Chris Ford >|> Senior Project Engineer >|> JNA Telecommunications >|> voice: +61 2 9935 5753 fax: +61 2 417 3862 e-mail: chrisf@jna.com.au >|> post: 16 Smith St, Chatswood, NSW 2067, Australia. >|> ----------------------------------------------------------------------------JNA > >I am fairly new to VxWorks, but I can tell you how I got the same tasks to not >share "global" variables with the same name (I can think of a couple of ways) >all involve loading the task multiple times in memory. > >1) If you don't use VxWorks to load-time-link anything (except the VxWorks >kernel support), then you probably could strip out all global symbols (or tell >VxWorks not to load the symbol table) and load the task multiple times in >memory. > >2) You could use VxVMI to establish private virtual memory contexts for each >task. This has the added benefit of having the text, data, and bss segments >of the task "protected" from other tasks running on the system (ie. their >physical addresses could be made read-only or invalid in the global virtual >context). This may have some interesting side-effects in how you do dynamic >memory allocation and how you have to build your class constructors. We were >doing this for the task protection, not multi-instancing tasks. > >3) finally, you could use a varient of the above theme, which may involve >writing your own loader. Using a share code segment. Load one complete copy >of the task in a virtual address space. Keep track of where it's data and bss >segments are located in virtual space. When you want to instance several >copies of the task, create new data and bss segments, a new virtual memory >context which refers to the old text segment, and the new data and bss, and >off you go. > >Let me know what you end up doing. > >Ted Marz >Software Engineering Institute >tfm@sei.cmu.edu >(412)268-5891 Hi everyone: I think there are two different things here. One is "task variable" (a variable which has different value with different task) which is not a C or C++ concept, another is static member of C++ class. When using C++, you'll be able to implement "task variable". One of implementation is following. Let us say you want to use a global object foo of type Foo as a "task variable", you define another class Foobar so that for each task that will use foo, put a member of type Foo in Foobar(of course they should be private members), when you define the access functions for these member, you should use task Id to choose the right one. You may think this will slow down your application. But when you use inline function it really create little overhead. If you consider that there are also overhead when using taskVar, you may end up with better performance. Tim Harrison and Douglas Schmidt have a paper related to this. They discuss the problem in general, not target to VxWorks. You can find it at Dr. Schmidt's home page: http://www.cs.wustl.edu/~schmidt I think it is under design pattern When you have a static member in your class, you always need to treat it as global, so when you define the members functions that deal with the static member, you always need to safe-guard it by some form of mutual exclusion. The problem of how to deal with reentry is more difficult. For one thing, when you code a member function you do not know if the function is working on a static object or a automatic object -- it dependents on how the object is used. Here are several things that I can think of: 1) Make it absurdly safe. i.e think of you are working on a static object. Of course, you'll need to use a lot of mutual exclusion and some of them are not necessary. Seem to me not a good ideal. 2) Provide two set of functions, one set for using with static object and another set for automatic object. In this case you put something extra on the user of your class. You have risk of having a unhappy user -:( And the design of the class is not clean. 3) Make the class intelligent so that it can tell what kind of storage it has(static or automatic). This solution is ideal, but it is more difficult and not portable. We are using Wind++. I think it keep the track of static objects in the system. 4) Do things as in case of C. The user of a class is responsible for thread-safe. This means if I am calling a member function of a global object, I will use semaphore or whatever to safe guard it. By saying all this, I don't mean that we are in a worse situation than in C.You can always use C++ as a better C. And there are C++ libraries that can be made thread-safe (we are using tools.h++ from RogueWave with Wind++, to make it multi-thread safe, we complied the library with a option) I do not know how much overheads we pay for when using the multi-thread safe version of the library. We moved from a C environment to C++ environment about two years ago. I think it was a good move. We have a lot of legacy C code, there are little problems to put the legacy C code and new C++ code together. If you are doing things right, I don't think you'll lose much performance when using C++. Jin Li Northern Telecom 408-565-7880 lijx@nt.com --------------------------- Newsgroups: comp.os.vxworks Subject: PowerPC patches to gcc? Date: 21 Aug 1996 03:26:49 GMT From: mcneils@app006221 (Sean McNeil) Organization: TRW Data Technologies Division, Carson CA Message-ID: I just started working on a project where VxWorks is the target OS running on a PowerPC board. I want to make a cross-compiler targeting powerpc-wrs-vxworks, but even the latest gcc snapshot doesn't support vxworks (even though it supports many other configurations, like powerpc-rtems). Does anyone know of patches I can acquire to add vxworks support for the powerpc target? --------------------------- Newsgroups: comp.os.vxworks Subject: Memory management Date: Wed, 21 Aug 1996 10:57:16 +0200 From: Harald Bjares Organization: VingMed Sound Message-ID: <321ACF6C.2572@vingmed.no> We are using C++ to implement our application. Unfortunately this involves lots of dynamic memory allocation/deallocation. After some time the heap is fragmented with many very small free memory blocks (and one large) resulting in a slow responce time. The problem is not memory leakage, but the fact that it taks a long time to search the heap for a large enough block of free memory. From the manual I see that the current algoritm uses a first-fit strategy and also combines adjacent blocks. Then to my question. Does anyone know of a memory manager for VxWorks which improves search time for free memory blocks? The alternative might be to cache our objects and reuse them, which have been done to some extent, but is not a practical solution. Thanks in advance. Harald Bjares hbj@vingmed.no --------------------------- Newsgroups: comp.os.vxworks Subject: Motorola 68030 VME card available Date: 21 Aug 1996 05:53:03 GMT From: innov8@ix.netcom.com (MaRtY) Organization: Netcom Message-ID: <4ve87v$654@dfw-ixnews3.ix.netcom.com> References: I have some Motorola MVME147 68030 25Mhz 8MB and 16MB VME cards available. All guaranteed. E-mail if interested. E-mail: marty@innovative-research.com - -- E-mail: innov8@ix.netcom.com --------------------------- End of New-News digest ********************** From CJONES@digsys1.bwi.wec.com Wed Aug 21 06:39:40 1996 From: "CHRIS JONES" Date: Wed Aug 21 06:39:43 PDT 1996 Subject: Programming Boot ROMS Are there any good sources available with instructions for programming boot roms? Our target is an MVME177. How does one seperate ROM image into the even and odd prom? Thanks, Chris Jones cjones@digsys1.bwi.wec.com (410) 765-7959 (410) 993-8822 FAX Northrop Grumman Electronic Sensors & Systems Division From han2326@aminet.co.kr Wed Aug 21 07:08:38 1996 From: "Han, Jeong Soo" Date: Wed Aug 21 07:08:40 PDT 1996 Subject: Setting of task priorities & system clock Hello, In my developing system using VME boards(cpu board:MC68030), I have designed about 10 tasks(local communication tasks, remote communication task, digital input task, digital output task, analog input task, intenal process task,...)and priority-based scheduling. My questions are 1) What is the optimal value of system clock to be set in my developing system? I know the default is 60 tick/1sec and I think this value is not appropriate, especially for the communication tasks. 2) What is the optimal value of task priority for each tasks. Are these decided only according to experience-base? Thank you in advance. J.S.Han/HDPIC Reserch Center, Seoul, Korea From seaweed!seaweed.com!bob@netcom.com Wed Aug 21 07:30:23 1996 From: bob schulman Date: Wed Aug 21 07:30:30 PDT 1996 Subject: Re: Memory management hbj@vingmed.no asks: > Does anyone know of a memory manager for VxWorks which improves > search time for free memory blocks? Seaweed Systems has a written-from-scratch memLib replacement. It has all the functionality which Wind River's memLib has...and more! It has memory leak detection, sanity checks at malloc/realloc/free time, user-callable memory arena sanity checker, constant-time allocation, better fragmentation behavior, callbacks for various low-memory conditions, and more! A real ginzu knife of an app. Please call or email for information: info@seaweed.com +1 510.482-3575 bob schulman From caot@applique.sedd.trw.com Wed Aug 21 11:21:04 1996 From: caot@applique.sedd.trw.com Date: Wed Aug 21 11:21:06 PDT 1996 Subject: HELP!!! Hi vxwexplo@lbl.gov Does anyone know the address of the GUI news group. I need some information on the GUI design tool that interfaces with Ada application on the PC platform (486 processor). Any suggestion would be greatly appreciated. Thanks, Tai Cao (310) 764-6944 From jhillman@wrs.com Wed Aug 21 12:04:28 1996 From: Jon Hillman Date: Wed Aug 21 12:04:30 PDT 1996 Subject: Re: HELP!!! At 11:21 AM 8/21/96 PDT, Tai Cao wrote: > Does anyone know the address of the GUI news group. >I need some information on the GUI design tool that interfaces with >Ada application on the PC platform (486 processor). Any suggestion >would be greatly appreciated. > >Thanks, >Tai Cao >(310) 764-6944 Tai Cao, I understand that Green Hills ADA-MULTI will be available soon on PC/Windows95. I have forwarded your request to Pat Rodenbeck of Green Hills (pat@ghs.com) to get in touch with you. Jon Hillman, FAE Wind River Systems 3504 Lake Lynda Dr., Suite 155 Orlando FL 32817 Phone: (407) 273-4646 Fax: (407) 273-5075 E-mail: jhillman@wrs.com Web site: http://www.wrs.com From froeber@bbn.com Wed Aug 21 12:16:59 1996 From: Fred Roeber Date: Wed Aug 21 12:17:01 PDT 1996 Subject: Re: Setting of task priorities & system clock Han, Jeong Soo" writes: > In my developing system using VME boards(cpu board:MC68030), I have > designed about 10 tasks(local communication tasks, remote communication > task, digital input task, digital output task, analog input task, intenal > process task,...)and priority-based scheduling. > My questions are > > 1) What is the optimal value of system clock to be set in my developing > system? I know the default is 60 tick/1sec and I think this value is not > appropriate, especially for the communication tasks. It isn't clear why you would want to change the clock rate for your system. VxWorks normally uses a priority based preemptive scheduling algorithm, not a time slice based round robin approach (although there is an option for enabling round robin type scheduling). Since the scheduling is preemptive, the clock rate doesn't have much control over scheduling. What it does affect are the granularity of time delays that can be achieved. If you want finer time delay granularity, you may want to boost the clock rate but realize that this will result in more CPU overhead to service the more frequent timer interrupts. A 68030 isn't too powerful so I wouldn't boost the timer interrupts much over 200/sec. > 2) What is the optimal value of task priority for each tasks. > Are these decided only according to experience-base? Setting task priorities in a preemption based system isn't an issue that is unique to VxWorks. There are various books on real time programming that cover these issues. A good place to start a search for such information is: http://cs-www.bu.edu/pub/ieee-rts/Home.html which is the Home Page of the IEEE Computer Society, Technical Committee on Real-Time Systems. Many people set task priorities based on heuristic methods (by guesswork and trial and error). There is a fair base of research on how to set priorities more sceintifically. One of the popular areas related to scheduling theory is "Rate Monotonic Scheduling". Under this theory (which works best for periodic systems) the tasks with the highest periodic rate get the highest priority. There is a lot more to it and all sorts of special extensions to try to handle aperiodic events but (in my opinion) a lot of the extensions are too complex and make assumptions that often aren't valid. The basic rule above is a good starting point for priority assignment though. Hope this helps. Fred -- | Fred J Roeber, BBN Systems & Technologies | | 50 Enterprise Place Middletown, RI 02842-5202 | | froeber@bbn.com 401-849-2543 (X48) | | See http://www.bbn.com/tv for TraceMaker info | From daemon@csg.lbl.gov Thu Aug 22 04:00:21 1996 From: daemon@csg.lbl.gov Date: Thu Aug 22 04:00:24 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Thu Aug 22 04:00:16 PDT 1996 Subject: Re: What about LINUX?? Subject: PC486 and PCI Subject: RPCs and multiple tasks Subject: RAM Disk Subject: dosFsLib DOS_OPT_AUTOSYNC option... ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: What about LINUX?? Date: Wed, 21 Aug 1996 14:13:49 -0400 From: Tom Flanigan Organization: Cybernet, Boca Raton, Florida. Message-ID: <321B51DD.41C6@oe.fau.edu> References: <9608150525.AA04342@rst.co.il> <4v0bf8$1lr@idiom.com> H.J. Bae wrote: > > leonid@rst.co.il (Leonid Rosenboim) wrote: > > >Linux - vote NO. > > Linux is the best UNIX I have ever used. > I've ported enough commerical UNIX kernels around in my career > to know that Linux is just as good, if not better. > > Of course, that doesn't mean that a "real company" > would even consider porting their tools to Linux. Too bad. > > HJB I personally have not used Linux yet but I'm planning on trying it out and I know of someone who has been able to port VxWorks to it. As far as I can see Linux is well test and although it may not be supported by many companies it at least has many followers who are constantly working on it. Besides something that you don't pay for can't be that bad... at least you can't say you didn't get what you payed for. - -Tom - -- - ----------------------------------------------------------------------------- Thomas C. Flanigan |Internet:tflaniga@oe.fau.edu Department of Ocean Engineering |Phone : 407-367-2947 Florida Atlantic University |Fax : 407-367-3885 - ----------------------------------------------------------------------------- "Well, when I was younger, my parents moved around a lot. But I always found them." --------------------------- Newsgroups: comp.os.vxworks Subject: PC486 and PCI Date: Wed, 21 Aug 1996 08:33:17 +0100 From: Chris Wallett Organization: Cranfield Precision Message-ID: <321ABBBD.300B@mail.compulink.co.uk> Does anybody know what is happoning with Tornado and PCI for PC targets ? We would like to use the new high speed Ethernet cards which use PCI. I know most of the new PowerPC VME cards use PCI and PCI peripherals (Ethernet/SCSI/Video) but there seems to be no support for the PC platform. Any ideas ?? --------------------------- Newsgroups: comp.os.vxworks Subject: RPCs and multiple tasks Date: Wed, 21 Aug 96 21:59:24 From: traymond@pcix.com Organization: Plymouth Commercial Internet eXchange Message-ID: <4vgett$2mn@alpha.pcix.com> Hi I have implemented an rpc server running under VxWorks that uses multiple tasks. There is one task that receives the rpc request, it is also the task that called rpcTaskInit, it then passes the request to a worker task. The worker task performs the request and then returns the rpc response. This seems to work but I am a little concerned because the VxWorks user guide warns against multiple tasks sharing the rpc structures. Does anyone know if what I have done will cause a problem? Thanks. Terry Raymond --------------------------- Newsgroups: comp.os.vxworks Subject: RAM Disk Date: 20 Aug 1996 15:25:49 GMT From: robinson@metronet.com (William Robinson) Organization: Abalone Engineering Associates Message-ID: <4vcldt$hq4@newnews.metronet.com> Hi, I'm working on a VME-based system running VxWorks in an environment with limited available power and weight. I have a datastream that requires a sustained 10 Mbytes/s storage rate for periods of up to 30 seconds. The device used to archive the data assumes there is a SCSI device available to store data at that sustained rate. Options that involve a RAID architecture aren't being considered because of the power/weight issue. Would anyone know of a "RAM Disk" device (or for that matter, some kind of SCSI device that doesn't cost a ton of money) that would integrate in my VxWorks/VME environment? I have plenty of time to transfer the data to a disk after the data is collected, but I need the burst rate capability. Any ideas? Thanks in advance, Will Robinson --------------------------- Newsgroups: comp.os.vxworks Subject: dosFsLib DOS_OPT_AUTOSYNC option... Date: Thu, 22 Aug 1996 02:39:04 GMT From: Ron Kellam Organization: Hewlett-Packard Co. Message-ID: <321BC848.219E@aus.hp.com> Sender: news@hparc12.aus.hp.com (News Daemon ID) Hi VxWorkers, For the time I have been making use of the VxWorks DOS files system, I have always had the DOS_OPT_AUTOSYNC option set. As might be expected, this has a significant impact on the file systems 'write' performance, making it much slower than with the AUTOSYNC option disabled. I am now considering disabling the AUTOSYNC option, but am concerned about how this may adversly affect the stability/integrity of the file system. Does anyone have any information or experience to share regarding this issue? Do most VxWorks users enable the AUTOSYNC option? Anything? Regards, Ron Kellam (ron_kellam@aus.hp.com) Australian Telecom Operation Hewlett Packard Australia --------------------------- End of New-News digest ********************** From franz@eng.kns.com Thu Aug 22 08:10:39 1996 From: franz@eng.kns.com (Keith Franz) Date: Thu Aug 22 08:10:41 PDT 1996 Subject: Re: memLib Just reciently, I think that I saw an email saying that someone had developed a replacement for memLib. If so, could someone please forward me that information? Thanks, -Keith Franz franz@eng.kns.com From chwe@force.de Thu Aug 22 08:37:02 1996 From: chwe@force.de (Christof Wehner) Date: Thu Aug 22 08:37:04 PDT 1996 Subject: RS-485 driver for FORCE CPU-40 MC68652? Hello There! Did anyone out there create a RS-485 driver for the above named board he/she'd be willing to share? A customer of ours is looking for it and would be very happy about this. Thanks in advance, anything is appreciated. i.A. Christof Wehner -- Christof Wehner ____ ____ ____ ____ ____ Technical Support Engineer / / / / / / / / Customer Support Department /__ / / /___/ / /__ FORCE COMPUTERS GmbH / / / / \ / / Prof.-Messerschmitt-Str. 1 / /___/ / \ /___/ /___ D-85579 Neubiberg/Muenchen email: chwe@force.de C O M P U T E R S G m b H http://www.forcecomputers.com/ - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - phone +49-89-60814-240 fax +49-89-609 77 93 From jhillman@wrs.com Thu Aug 22 08:40:38 1996 From: Jon Hillman Date: Thu Aug 22 08:40:40 PDT 1996 Subject: Re: comp.os.vxworks newsdigest At 04:00 AM 8/22/96 PDT, Tom wrote: > >H.J. Bae wrote: >> >> leonid@rst.co.il (Leonid Rosenboim) wrote: >> >> >Linux - vote NO. >> >> Linux is the best UNIX I have ever used. >> I've ported enough commerical UNIX kernels around in my career >> to know that Linux is just as good, if not better. >> >> Of course, that doesn't mean that a "real company" >> would even consider porting their tools to Linux. Too bad. >> >> HJB > >I personally have not used Linux yet but I'm planning on trying it out >and I know of someone who has been able to port VxWorks to it. As far >as I can see Linux is well test and although it may not be supported by >many companies it at least has many followers who are constantly working >on it. Besides something that you don't pay for can't be that bad... at >least you can't say you didn't get what you payed for. > I thought I would put my personal 2 cents worth in (probably not worth that). WRS has considered LINUX, but regardless of its technical merits, there is not enough volume (read that demand) to do it for Tornado. Now please do not confuse the statistics with facts here, but from what I understand, the customer base for LINUX doing embedded development is less than the customer base for MACINTOSH. WRS goes with the flow and when LINUX reaches a "critical mass" then WRS will most certainly be listening. Other than this forum, I have not had a single potential customer ask me about LINUX. I do not in this case speak officially for Wind River Systems, merely forwarding my opinion. Jon Hillman Orlando FL USA From sri1@gvl.esys.com Thu Aug 22 09:13:01 1996 From: Shawn Ingram Date: Thu Aug 22 09:13:07 PDT 1996 Subject: System Image and Application Code in FLASH I was looking for advice on how to (or if its possible at all) to burn a minimal kernel image along with my application code into FLASH on an AXPvme 160. First of all I'm not really sure how much FLASH is on the 160. But, I need to have everything boot without Ethernet or disk. All this leaves of course is FLASH. I will only need a very minimal kernel since I'm not using any network functions or file systems. I would like to retain a minimal shell and symbol table if at all possible but if not .... Also the loader and unloader stuff I can ditch since there will be nothing to load from. Let me get to my questions 1. I've examined the Makefile stuff and I'm not sure which "flavor" of a system image to use. I was thinking "system_rom_alpha" or "system.st_rom_alpha" sounded correct but I've not done this before nor has anyone where I work. Which type should I use? 2. What file(s) will I have to modify to pass control from the (minimal) shell to my application so that I may spawn my tasks, etc? 3. How much FLASH is on the AXPvme 160 and how much of this is reserved for the console? 4. When I've built this system/application image do I just burn it into flash using the "update" command in the normal way? 5. Has anyone done this? Any other advice or answers to questions that I didn't ask but should have are greatly appreciated. Thanks in advance for any responses (even flames :) /********************************************* /* * /* S.R.INGRAM sri1@esygvl.com * /* Phone: (903) 457-5255 * /* * /*********************************************/ From YF.Siu@eng.efi.com Thu Aug 22 10:35:48 1996 From: "Yuet Fung Siu" Date: Thu Aug 22 10:35:50 PDT 1996 Subject: Re: comp.os.vxworks newsdigest >Do most VxWorks users enable the AUTOSYNC option? We never enable the AUTOSYNC option in our dosFs. But we have a fixed hard drive, no removable media. In VxWorks 5.1, we discovered that if you don't enable the AUTOSYNC option, the FAT table will not be synced until you close the file. This may create large number of lost clusters if the system was interrupted before a file close. In VxWorks 5.2, WRS fixed this problem by always sync'ing the FAT whenever there is a modification, disregarding the setting of the AUTOSYNC option. Only the sync'ing of file data is governed by this option. Good Luck, YF From lbecker@avtec.com Thu Aug 22 10:56:50 1996 From: "Lynn W. Becker, II" Date: Thu Aug 22 10:56:53 PDT 1996 Subject: flexlm behavior I tried starting flexlm (license manager) from the command line to see if it would work before creating a script in /etc/rc2.d. I am running Tornado 1.0 with the version 5.3-PowerPC-beta. I am using Solaris 2.4. I typed in lmgrd -c ~lbecker/license.dat > ~lbecker/license.log & and when I dumped license.log I got... skybox# more ~lbecker/license.log 8/22 11:35:46 (lmgrd) FLEXlm - Copyright 1988, 1992, Highland Software, Inc. 8/22 11:36:11 (lmgrd) Retrying socket bind (address in use: port 7689) 8/22 11:36:41 (lmgrd) Retrying socket bind (address in use: port 7689) 8/22 11:37:11 (lmgrd) Retrying socket bind (address in use: port 7689) 8/22 11:37:41 (lmgrd) Retrying socket bind (address in use: port 7689) 8/22 11:38:11 (lmgrd) Retrying socket bind (address in use: port 7689) 8/22 11:38:41 (lmgrd) Retrying socket bind (address in use: port 7689) 8/22 11:39:11 (lmgrd) FLEXlm (v2.40) started on skybox (Sun) (8/22/96) 8/22 11:39:11 (lmgrd) License file: "/export/home/lbecker/license.dat" 8/22 11:39:11 (lmgrd) Started wrs 8/22 11:39:11 (wrs) Server started on skybox for: tgtsvr I am not sure why I got this (unless a previous failed attempt to run the license server had hung up that port). A previous attempt had the same messages. but it was never able to connect. Has anyone else out there who uses Solaris 2.4 and PPC Beta seen this behavior? Thanks! ______________________________________________________________________ Mr. Lynn W. Becker lbecker@avtec.com Engineer (703) 273-2211 Avtec Systems Inc (703) 273-1313 FAX 10530 Rosehaven Street, Suite 300 http://www.avtec.com Fairfax, VA 22030 From lbecker@avtec.com Thu Aug 22 11:04:03 1996 From: "Lynn W. Becker, II" Date: Thu Aug 22 11:04:05 PDT 1996 Subject: Does 5.3 use interrupts to communicate? I am assuming that 5.3 (Tornado) uses interrupts to communicate. I have a Heurikon Baja board running 5.2 and a Motorola MV1300 running 5.3 Beta. When I do a windsh to connect to the MV1300 the Baja terminal fills with these messages... -> interrupt: Uninitialized Interrupt! Vector number 60 (0-255). non-cpu bus error interrupt Program Counter: 0x80052f20 Status Register: 0x30006703 interrupt: Error address: 0xff04c240, Error ID: 0x0003 interrupt: Uninitialized Interrupt! Vector number 60 (0-255). non-cpu bus error interrupt Program Counter: 0x80052f20 Status Register: 0x30006703 interrupt: Error address: 0xff04c240, Error ID: 0x0003 interrupt: Uninitialized Interrupt! Vector number 60 (0-255). non-cpu bus error interrupt Program Counter: 0x80052f20 Status Register: 0x30006703 interrupt: Error address: 0xff04c240, Error ID: 0x0003 interrupt: Uninitialized Interrupt! Is this a problem or do I just need to rebuild the BSP to ignore this interrupt vector on the Baja? Thanks for any info... ______________________________________________________________________ Mr. Lynn W. Becker lbecker@avtec.com Engineer (703) 273-2211 Avtec Systems Inc (703) 273-1313 FAX 10530 Rosehaven Street, Suite 300 http://www.avtec.com Fairfax, VA 22030 From robert@orincon.com Thu Aug 22 11:13:45 1996 From: "Robert Redfield" Date: Thu Aug 22 11:13:47 PDT 1996 Subject: RAID interfaces Subject: Time:11:10 AM OFFICE MEMO RAID interfaces Date:8/22/96 Anybody driving RAID devices (ideally from Andataco) from VxWorks boards? From jhillman@wrs.com Thu Aug 22 12:40:30 1996 From: Jon Hillman Date: Thu Aug 22 12:40:32 PDT 1996 Subject: Re: Does 5.3 use interrupts to communicate? At 11:04 AM 8/22/96 PDT, Lynn Becker wrote: > >-> interrupt: >Uninitialized Interrupt! >Vector number 60 (0-255). non-cpu bus error interrupt >Program Counter: 0x80052f20 >Status Register: 0x30006703 >interrupt: Error address: 0xff04c240, Error ID: 0x0003 > ... > >Is this a problem or do I just need to rebuild the BSP to ignore this >interrupt vector on the Baja? Looks like a vme board is generating some interrupts and you don't have any application code to respond to it. You need to determine what is doing this (bus analyzer will help but typically); you probably know what else is in your vme chassis! Use intConnect(INUM_TO_IVEC(vector_number), (FUNCPTR) intFunc, (int) arg) to install your subroutine to respond to the interrupt vector. (Think of the vector number here as the index to the vector table, thus INUM_TO_IVEC(vector_number) converts this to a real address. The above stuff is typically application code. I hope this is what you're asking for. Jon Hillman, FAE Wind River Systems Orlando FL USA _________________________ | Jon Hillman, FAE | | Wind River Systems | /) (407) 273-4646 (\ / ) Email:jhillman@wrs.com ( \ _( (|_________________________) ) /> (((\ \) /,) / ) / //))/ (\\\\ \_/ / \ \_/ ///// \ / \ / \ _/ \_ / ----/ /----------------------\ \---- / / \ \ From Scott.Peters@agai.com Thu Aug 22 15:04:51 1996 From: Scott Peters Date: Thu Aug 22 15:04:53 PDT 1996 Subject: RE: Handling statics with C++ I would suggest that you make the object complete self-contained. Incorporate any task ID and semaphore ID's in the class. In the constructor of the class add all the cremation stuff like taskCreate and semCreate and store these ID's in the instantiated objects. Of course this only works if you are not concerned on how many task run on your system. scott.peters@agai.com ---------- Subject: Handling statics with C++ Date: Monday, August 12, 1996 5:13PM Submitted-by chrisf@jna.com.au Mon Aug 12 17:13:48 1996 Submitted-by: "Chris Ford" We are currently using VxWorks in a C environment, and are considering a move to C++. One concern we have, however, is the treatment of statics. Under VxWorks, if we wish to start several instances of the same task, and we wish each instance to have its own copy of a static, we need to use taskVarAdd(). This has been fine so far as we can combine the necessary statics into a single structure and only tasVarAdd() a single pointer to that structure. This minimises the hit on context switch time. In C++, however, the whole idea is to hide your data, meaning that any static member variables should remain part of a class. Two questions then arise: - How to taskVarAdd() these statics. One solution is to have a static member function used to initialise the class (a bit yucky, but workable). - What do you do about the performance problem? Because the data should remain hidden, it's a bad idea to pull the statics out into a combined structure. This means that every class which has a static must do its own taskVarAdd(), multiplying the number of task variables, and increasing the context switch time. The question is: what do people do about this? Many thanks in advance for any ideas. -- Chris Ford Senior Project Engineer JNA Telecommunications voice: +61 2 9935 5753 fax: +61 2 417 3862 e-mail: chrisf@jna.com.au post: 16 Smith St, Chatswood, NSW 2067, Australia. ------------------------------------------------------------------------ --- -JNA From P.H.A.Mutsaers@cycl.phys.tue.nl Fri Aug 23 07:44:46 1996 From: Peter Mutsaers Date: Fri Aug 23 07:44:48 PDT 1996 Subject: zbufLib Date: Fri, 16 Aug 1996 17:00:14 +0100 From: Peter Mutsaers Subject: zbufLib To: vxexplo@lbl.gov Cc: TNNDMUTS@cycl.phys.tue.nl Content-type: TEXT/PLAIN; CHARSET=US-ASCII Hello (Vx)Workers, Does anybody know if the maximum length of a zbuf or zbufsegment is limited? A call to most of the zbufLib routines fail when the original buffer I inserted into the zbuf is larger then 32767 bytes (i.e. the maximum positive value of a signed short int). Greetings, Jos Aarts. From schen3@ccgate.hac.com Fri Aug 23 11:21:15 1996 From: schen3@ccgate.hac.com Date: Fri Aug 23 11:21:17 PDT 1996 Subject: Map out the memory map ?? I am looking for way to map out the memory map on the RAD6000 VME board I am using. I would like to know where is VxWorks loaded to, end of the system image, RAM_LOW_ADRS, FREE_RAM_ADRS, RAM_HIGH_ADRS, initial stack, interrupt stack and system memory pool begin address and end address. Some address are define in config.h configAll.h and usrConfig.c but when I used d() to check the memory, the VxWorks image was not at the address define in the header file. Any advice are welcome. Thanks in advance. Steven Chen schen3@ccgate.hac.com From robert@orincon.com Fri Aug 23 13:44:24 1996 From: "Robert Redfield" Date: Fri Aug 23 13:44:26 PDT 1996 Subject: Force SPARC target Subject: Time:1:42 PM OFFICE MEMO Force SPARC target Date:8/23/96 Any opinions, comments or experiences running VxWorks on Force 5V, or 20? From han2326@aminet.co.kr Fri Aug 23 23:17:59 1996 From: "Han, Jeong Soo" Date: Fri Aug 23 23:18:01 PDT 1996 Subject: work queue overflow Does anybody experienced this error : "workQpanic : work queue overflow" I have got this error when I run my serial communication program.(i.e. 2~5 second after spawning) The hardware configuration is that piggyback communication board(MC68881 chip), connected on MC68030 cpu board. Any hints would be appreciated. Thank you. J.S. Han ----------- HDPIC From daemon@csg.lbl.gov Sat Aug 24 04:01:31 1996 From: daemon@csg.lbl.gov Date: Sat Aug 24 04:01:33 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sat Aug 24 04:01:28 PDT 1996 Subject: Re: Memory management Subject: Re: 5.3 booting from 5.2 Shared memory? ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Memory management Date: 22 Aug 1996 13:02:31 GMT From: dixon@ra.crd.ge.com (walt dixon) Organization: GE Corp. Research & Development, Schenectady, NY Message-ID: <4vhlp7$eum@crdnns.crd.ge.com> References: <321ACF6C.2572@vingmed.no> In article <321ACF6C.2572@vingmed.no> Harald Bjares writes: We are using C++ to implement our application. Unfortunately this involves lots of dynamic memory allocation/deallocation. After some time the heap is fragmented with many very small free memory blocks (and one large) resulting in a slow responce time. The problem is not memory leakage, but the fact that it taks a long time to search the heap for a large enough block of free memory. From the manual I see that the >current algoritm uses a first-fit strategy and also combines adjacent blocks. Then to my question. Does anyone know of a memory manager for VxWorks which improves search time for free memory blocks? The alternative might be to cache our objects and reuse them, which have been done >to some extent, but is not a practical solution. The classic way of solving this problem is to work with multiple heaps. One heap for small objects, another for larger objects, etc. If there are only one or two large object classes, perhaps you could implement an operator new for them which allocates memory from a different pool. ALthough you apparently have not run into the problem yet, the small objects could fragment your heap to the extent that you cannot allocate a large object. - -- Walt Dixon dixon@crd.ge.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: 5.3 booting from 5.2 Shared memory? Date: 15 Aug 1996 15:55:03 GMT From: friedl@cher.heurikon.com (Ted Friedl) Organization: Heurikon Corporation Message-ID: <4uvh8n$8u4@badger.heurikon.com> References: In article , Lynn W. Becker, II wrote: >> A good test would be from the 1300's [VxWorks Boot]: >>prompt, do a >> >>[VxWorks Boot]: d 0x40000600 >> >>You should see the "sm" signature value of 0x87654321. If you don't, then >>the address maps don't line up between the two processor types and you'll >>have to go to your 1300 documentation to find how to properly map that >>address. [snip] Lynn, This is sound advice. Your display of local 0x40000600 on the mv1300 does not appear to be the SM anchor of the Baja4700. By default, a Heurikon Baja4700 with a sysProcNum of 0 and a valid backplane address will place the SM anchor at local 0x600 and map the base of its memory to 0x40000000 of VME A32 space. It just so happens that the Baja's local to A32 space address translation is "one-to-one" (no translation) so the boot device on Bajas 1 and 2 ends up being sm=40000600. On the mv1300, local 0x40000600 may not translate to 0x40000600 of VME A32 space. Do check the manual to see if this is the case. If the mv1300 handles tranlation like the Baja, the next thing to investigate is the mv1300 cache. Good luck. Let us know how things turn out. Ted Friedl Senior Software Engineer Heurikon Corporation Madison, WI --------------------------- End of New-News digest ********************** From daemon@csg.lbl.gov Sun Aug 25 04:00:28 1996 From: daemon@csg.lbl.gov Date: Sun Aug 25 04:00:30 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sun Aug 25 04:00:25 PDT 1996 Subject: Re: Standard I/O read with a timeout Subject: VxSIM + CoStreams Subject: [ANNOUNCE] Formal design methods conference FORTE/PSTV'96 Subject: [ANNOUNCE] SDL Forum 1997 - info ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Standard I/O read with a timeout Date: 24 Aug 1996 18:24:44 -0700 From: mcs@goblin.caltech.edu (Martin Shepherd) Organization: California Institute of Technology. Message-ID: References: <3210D651.41C67EA6@adc.com> In article tadlock@bosch.dseg.ti.com (edward tadlock) writes: > >...[Original question omitted] > > Setup a timer for the task, set its value and arm it before >the call to read. If it goes off, the read will be exited and you can >check the return code (I think it is EINTR) and do whatever error >handling you want. Be sure to reset the timer when you are done. >Below is a code fragement to show the method..it may or may not be >syntactically and logically correct..its just a illustration.. > >/* setup timer */ > > timer_create(CLOCK_REALTIME, &eventhandler, &timer_id); > timer_settime(timer_id,flags, ×pec, NULL); > > i= read(fd, &buffer, maxBytes); >/* check read results */ > if (i != nbytes ){ > if (errno = EINTR) { > /* error handling */ > printf(...) > } > } > timer_cancel(..) Isn't there a race condition in the above example? If a higher priority task pre-empts this code between the timer_settime() statement and the read() statement then the timer might go off before the read() is started and the resulting signal would then not be seen by the read() statement. How about the following alternatives: 1. Use select with a timeout to wait for the target file-descriptor to become readable (or writable). This will require a loop if more than one byte is to be transfered. In particular, if the driver doesn't support non-blocking I/O then data will have to be transfered one byte at a time. This method works if just one task is performing a particular flavor of I/O on the file-descriptor. Otherwise semaphores have to be used to make sure that just one task responds when the fd becomes ready for I/O. 2. A messier way to timeout I/O is to use a watchdog timer with an auxiliary task and a semaphore. The watchdog timer signals the semaphore when the timeout is reached, and the auxiliary task then cancels the I/O either by sending the blocked task a signal or by using a driver-provided ioctl() similar to FIOCANCEL in tyLib. If the auxiliary task runs at lower priority than the task that is performing the I/O then it will only run when the primary task is blocked. This prevents it cancelling the I/O between the watchdog timer being started and the read() or write() blocking, and thus side-steps the race condition that I pointed out above. Be sure to cancel the watchdog when the read is done. Then apply semTake(sem, NO_WAIT) to the timeout semaphore to prevent spurious I/O cancelation at a later time, just in case the timer caused the semaphore to be signaled after the read() but before the wdCancel(). If anybody has any better suggestions for timing out I/O (or disagrees with the above) then I would be interested to hear what they have to suggest. Martin Shepherd (mcs@astro.caltech.edu) --------------------------- Newsgroups: comp.os.vxworks Subject: VxSIM + CoStreams Date: Sun, 25 Aug 1996 02:05:32 GMT From: sheih@netcom.com (SHOUJONG SHEIH) Organization: NETCOM On-line Communication Services (408 261-4700 guest) Message-ID: Sender: sheih@netcom.netcom.com Greetings: Does anyone know what's the right procedure to integrate CoStreams package into VxSIM and rebuild the VxSIM? Or, is it even possible. I followed the instructions in the manual and was able to rebuild VxSIM by using the ADDED_MODULE option in Makefile.SIMSPARCgnu to include object files generated during compiling CoStream files. However, after I started the new rebuilt VxSIM and spawn a function - StreamScheduler(), I got an error message - Segementation Violation. Can anyone help? Thanks a lot. S. John Sheih - -- sheih@netcom.com --------------------------- Newsgroups: comp.sys.m68k,comp.protocols.iso,comp.protocols.iso.dev-environ,comp.protocols.misc,comp.realtime,comp.robotics.research,comp.software-eng,comp.software.testing,comp.specification.misc,comp.os.qnx,comp.os.lynx,comp.os.chorus,comp.os.vxworks Subject: [ANNOUNCE] Formal design methods conference FORTE/PSTV'96 Date: 24 Aug 1996 23:52:15 GMT From: johnd@reflex-tech.co.uk (John Dalton) Organization: Reflex Technology Message-ID: <4vo4jf$nfk@raven.cc.ukans.edu> Followup-To: comp.robotics.research Reply-To: johnd@reflex-tech.co.uk FORTE/PSTV'96 will address Formal Description Techniques (FDTs) applicable to communication protocols and distributed systems (such as Lotos, SDL, Estelle, ASN.1, TTCN, Z, automata, process algebras, logics). This is relevant to communications, transport, robotics and safety-critical systems engineering as well as the more general interest in automated simulation and verification/validation. The conference will be a forum for presentation of the state of the art in theory, application, tools and industrialization of FDTs, and will provide an excellent orientation for newcomers. For the first time, the hitherto separate conferences FORTE and PSTV will be combined into a joint edition. URL: http://www.informatik.uni-kl.de/aggotz/forte.pstv96 ******* Crossposted: comp.robotics.research (moderated) ******* Summary: Academic, government & industry research in robotics. Archives and information: http://www.robot.ireq.ca/CRR Charter: ftp://ftp.robot.ireq.ca/pub/crr/Charter Meta-discussions/information: crr-request@robot.ireq.ca --------------------------- Newsgroups: comp.sys.m68k,comp.protocols.iso,comp.protocols.iso.dev-environ,comp.protocols.misc,comp.realtime,comp.robotics.research,comp.software-eng,comp.software.testing,comp.specification.misc,comp.os.qnx,comp.os.lynx,comp.os.chorus,comp.os.vxworks Subject: [ANNOUNCE] SDL Forum 1997 - info Date: 24 Aug 1996 23:52:56 GMT From: johnd@reflex-tech.co.uk (John Dalton) Organization: Reflex Technology Message-ID: <4vo4ko$nfk@raven.cc.ukans.edu> Followup-To: comp.robotics.research The eighth SDL Forum takes place at INT in Evry, France, September 22-26, 1997, organised by INT and CNET-FRANCE TELECOM. World Wide Web: http://alix.int-evry.fr/lor/SDL97/ The SDL Forum is the meeting place for established and potential users of the language, tool vendors and ITU Study Group members maintaining the language. It is held every 2 years and has become the most important event in the calendar for anyone involved in SDL and related languages and technology. The proceedings are widely referenced, and the Forum is the basis of further improvements. In addition, the SDL Forum has also become a Forum for Message Sequence Charts (MSC), the formal event trace notation which is another cornerstone of ITU-T languages. SDL is the relevant formal description technique in communications and other fields of distributed or safety-critical software where simulation, verification and validation against specifications are essential (eg robotics control, transport, aeronautics, medical, space applications). The ITU (former CCITT) recommended Specification and Description Language, SDL, has a growing user community, not only within telecommunications industry, but also in these other application areas. The SDL-92 version of the language is an object based language, with support for concurrent objects (processes) with behaviour in terms of Extended Finite State Machines and interaction by asynchronous signal exchange and remote procedure calls. SDL-92 is now available in industrial and commercial tools. A limited addendum to SDL-92 has been agreed by ITU SG10, and updated tools supporting this last version (sometimes called SDL-96) should be available by the time of the conference. ******* Crossposted: comp.robotics.research (moderated) ******* Summary: Academic, government & industry research in robotics. Archives and information: http://www.robot.ireq.ca/CRR Charter: ftp://ftp.robot.ireq.ca/pub/crr/Charter Meta-discussions/information: crr-request@robot.ireq.ca --------------------------- End of New-News digest ********************** From craig@pacengr.com Mon Aug 26 12:26:22 1996 From: Pacific Design Engineering Date: Mon Aug 26 12:26:24 PDT 1996 Subject: VxWorks Job Opening, SF Bay Area Subject: Job Opening, VxWorks Development Pacific Design Engineering, an engineering consulting business, has an immediate opening for a VxWorks systems and application development engineer. This is a full-time position. The work may be performed from home, or in our customers' site, as required. We are located in Danville, California, in the San Francisco Bay Area. The successful applicant will have a background in real-time software development - experience with VxWorks is preferred. A minimum of two years of experience is necessary for this position. A BSEE/BSCS is required, an MSCS/MSEE is a plus. The individual must be resourceful, self-motivated, and work well under pressure, and have a demonstrated history of successful completion of software development projects. Substantial training will be provided by VxWorks development experts during the initial projects. Pacific Design Engineering is a small, rapidly growing business that offers services in the real-time systems market, Internet security, and client-server application development. We offer attractive compensation and a chance to work with highly-skilled, experienced engineering consultants. We are an equal opportunity employer. For more information, send email, with your resume attached in either MIME, UUENCODE, or BINHEX format to: jobs@pacengr.com ___________________________________________________________________ Craig A. Humphreys Pacific Design Engineering System Engineering Consulting Danville, CA Real-time kernels, Internetworking pacengr@pacengr.com Digital Image and Signal Processing http://www.pacengr.com/pde.htm Internet Security; WWW services PGP Public Key: http://www.pacengr.com/pde/pgpkey.htm From p72950@gmcl.co.uk Tue Aug 27 00:43:49 1996 From: Phil Quiney Date: Tue Aug 27 00:43:51 PDT 1996 Subject: Re: work queue overflow Hi there, Yes I have seen this error when in the process of writing a synchronous device driver for the z8530 SCC. It was caused by an interrupt that was not being handled correctly by the driver. I contacted WRS & recieved the following e-mail more generally covering this subject. Mike Tarlton European Technical Support Wind River Systems 19, avenue de la Norvege ZA de Courtaboeuf 1 91953 Les Ulis Cedex Tel : +33 1 60 92 63 00 Fax : +33 1 60 92 63 15 Email: miket@wrsec.fr / support@wrsec.fr -------------------------------------------------------------------- Summary #11538 AUTHOR: davidj ------------------------------------------------------- DATE: May 27 1993 11:21AM Product: 5.0.2 Host: Sun3 Board: mv167 Board: TITLE: What's causing the workQPanic: Kernel work queue overflow, error? SYSTEM: DESCRIPTION: SYMPTOM(S): RESOLUTION: Regarding your work queue overflow question, it can be caused by several different things. In general, the work queue is used to queue kernel work that must be deferred because the kernel is already engaged by another request. Kernel work deferred to the work queue must originate from an interrupt service routine. When the kernel is in "kernel" state, certain operations can not be per- formed, they must be deferred. Some of these operations are: incrementing the timer tick, taskSuspend, taskResume, semGive, etc. The work to be deferred must be added/originated from an ISR. If a function added to the netTask queue never returns, then the netTask will never get a chance to run the next job on its queue, and the queue will, therefore, fill up. The size of the work queue is fixed and there is no way to change the size of the work queue under the current setup. It could be that the jumpers are set wrong. More typically, the work queue overflow is because the interrupt ack- nowledgement is incorrect or that the ISR frequency rate is too high. Check for other high frequency interrupts which might be interrupting the kernel calls. If there are high frequency interrupts then lower the inter- rupt frequency. What's happening is that there are more netJobAdds happening at interrupt level than the task level code running in the netTask context can process. Check to see if there are lot of network activity. The network interfaces usually call netJobAdd to add incoming packets to be processed by netTask. NetJobAdd calls semGive. If this is the case check to see if netTask is being given enough time to run. Is there an application task at a higher priority? If so try raising the priority of the netTask such that it is higher than the application. A common problem is that the interrupt code (or even the code spawned as a task) does a netJobAdd every time that the interface gets an interrupt. The correct way to do this is to have the task level code that gets run by netJobAdd, process all of the work that needs to be done (namely: dequeue incoming packets and free the buffers of transmitted packets), and have the interrupt level code only do a netJobAdd if the task level code is not currently running (specifically: this means that the task level code sets a flag that says that it's running that the interrupt level code can exam- ine.) You may, perhaps, also fix this by changing the priority of netTask (as stated above) to be higher than that of the shell. netTask must be at a higher priority to service these, or else the ring buffer overflows, which is a different issue. SEE ALSO: END----------------------------------------------------------------- Regards P.J Quiney >Does anybody experienced this error : > "workQpanic : work queue overflow" >I have got this error when I run my serial communication program.(i.e. >2~5 second after spawning) >The hardware configuration is that piggyback communication board(MC68881 >chip), connected on MC68030 cpu board. >Any hints would be appreciated. >Thank you. >J.S. Han >----------- >HDPIC =============================================================================== Phil Quiney _/_/_/_/ _/ _/_/_/_/ email : quiney@gmcl.co.uk GEC Marconi _/ _/ _/ _/ _/ tel : +44 1245 353221x3914 Communications Ltd _/_/_/_/ _/ _/ _/ _/_/ fax : +44 1245 275448 Chelmsford _/ _/_/_/_/ _/_/_/_/ England CM1 1PL _/ =============================================================================== From daemon@csg.lbl.gov Tue Aug 27 04:00:24 1996 From: daemon@csg.lbl.gov Date: Tue Aug 27 04:00:27 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Tue Aug 27 04:00:20 PDT 1996 Subject: Re: RAID interfaces Subject: Re: No UDP packet Subject: No UDP packet Subject: Re: VxWorks boot from scsi-pcmcia adapters Subject: Re: VxWorks boot from scsi-pcmcia adapters Subject: Re: Does 5.3 use interrupts to communicate? Subject: Req: Driver for Cirrus Logic CD1865/1864/180 octal UART Subject: IP-488 driver questions!!! ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: RAID interfaces Date: 26 Aug 1996 16:15:48 GMT From: friedl@cher.heurikon.com (Ted Friedl) Organization: Heurikon Corporation Message-ID: <4vsijk$jcl@badger.heurikon.com> References: In article , Robert Redfield wrote: > Subject: Time:11:10 AM > OFFICE MEMO RAID interfaces Date:8/22/96 > >Anybody driving RAID devices (ideally from Andataco) from VxWorks boards? > Robert, I have written some demanding multi-channel video applications using VxWorks and RAIDs from Ciprico and Conner (the Conners used a Mylex controller). The VxWorks boards used were the Heurikon Nitro40 and the Baja4700. I can share some RAID benchmark results for these boards with VxWorks if you are interested. Ted Friedl Senior Software Engineer Heurikon Corporation Madison, WI --------------------------- Newsgroups: comp.os.vxworks Subject: Re: No UDP packet Date: 26 Aug 1996 18:39:47 GMT From: David Perkinson Organization: TRW, Inc. Message-ID: <3221EFD7.FF6@boris.msfc.nasa.gov> References: <3221CCFD.794B@boris.msfc.nasa.gov> <4vslrr$i1d@news.sei.cmu.edu> Theodore, Thanks for the info. I have run this test with and without active TCP socket connections to the same VxWorks target. The system resides on a 3000 node bridged WAN (beyond my control) and there is at least one SGI NTP daemon talking out there. I have observed the traffic from my TCP sockets using HP NetMetrix running on a Sun SparcServer20. I have watched the test code run correctly on Unix machines. It compiles with no changes on VxWorks. Theodore Marz wrote: > > Hello, > What else is on your net, and how do you know the packets aren't going out? > I have seen (and heard of) some trouble in VxWorks 5.2 with UDP/IP. It seems > that if there is broadcast traffic out there, to a port which is not bound on > the VxWorks side to some service, that sometimes the VxWorks UDP/IP stack > shuts down and no packets propagate up/down the IP stack. > > To check this out, Why don't you try isolating your network to just your > VxWorks boot server (which will be one end of the connectionless comm link) > and the VxWorks target. > > If your boot server is SGI, then you can use netstat to see what UDP/IP > packets are happening (at least counts). > > Just for a little more background on the above, we were using sockets, > UDP/IP broadcast, and had an SG on the net as a monitor. It seemed that > timed (the SGI NTP server daemon) was putting out NTP packts that the VxWorks > system did not like. Got a bunch of "Destination Unreachable" ICMP packets > and then the VxWorks stack just kind of shut down (still said packets were > going out, but they weren't). Putting some directed traffic to the VxWorks > systems (such as an active telnet, or Ping) seems to shake it loose. > > There are a couple of open TSRs on this problem since early spring. Haven't > heard of any resolution. > > Ted Marz > > In article <3221CCFD.794B@boris.msfc.nasa.gov>, David Perkinson writes: > |> This is a multi-part message in MIME format. > |> > |> --------------15FB59E21CFB > |> Content-Type: text/plain; charset=us-ascii > |> Content-Transfer-Encoding: 7bit > |> > |> I am having trouble sending a UDP packet using VxWorks 5.2. > |> My TCP sockets work fine, and the UDP calls do not return errors, I just > |> don't get a packet. I have sniffed the network to watch the traffic. > |> The same code works correctly on our SGI Challenge. Is there something > |> I must do to enable UDP? (I didn't find UDP mentioned in UserConfig.h) - -- =============================================================================== David Perkinson, TRW Opinions that may be expressed in this email perkinsond@boris.msfc.nasa.gov are my own and not neccessarily that of TRW. =============================================================================== --------------------------- Newsgroups: comp.os.vxworks Subject: No UDP packet Date: 26 Aug 1996 16:11:05 GMT From: David Perkinson Organization: TRW, Inc. Message-ID: <3221CCFD.794B@boris.msfc.nasa.gov> This is a multi-part message in MIME format. - --------------15FB59E21CFB Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit I am having trouble sending a UDP packet using VxWorks 5.2. My TCP sockets work fine, and the UDP calls do not return errors, I just don't get a packet. I have sniffed the network to watch the traffic. The same code works correctly on our SGI Challenge. Is there something I must do to enable UDP? (I didn't find UDP mentioned in UserConfig.h) - -- =============================================================================== David Perkinson, TRW Opinions that may be expressed in this email perkinsond@boris.msfc.nasa.gov are my own and not neccessarily that of TRW. =============================================================================== - --------------15FB59E21CFB Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="udptest.c" #include #include #include #define CONNECT_TO_REMOTE 200 #define DISCONNECT_FROM_REMOTE 201 #define REMOTE_COMMAND 202 #define ENABLE_INCOMMING 203 #define DISABLE_INCOMMING 204 #define TIMELINER_EXECUTOR 205 #define TL_STEP_REQ 206 #define NUM_OF_REMOTES 20 #define CMD_HDR_SIZE 12 #define REMOTE 1 /*-------------------------------------------------------------------*/ /* TYPEDEFS */ /*-------------------------------------------------------------------*/ typedef struct socketEntryType { char address[20]; short int port; int socketFD; int protocol; struct sockaddr_in addr; int incomming; int TimeLiner; } SOCKETENTRYTYPE; typedef struct tlMessageType { short int messageType; short int tlVersion; short int objectId; short int objectSize; char data[4096]; } TLMESSAGETYPE; typedef struct tlStepRequestType { short int Request; short int Param1; short int Param2; short int Param3; } TLSTEPREQUESTTYPE; /*-------------------------------------------------------------------*/ /* FILE GLOBALS */ /*-------------------------------------------------------------------*/ char message[100]; SOCKETENTRYTYPE Remote[NUM_OF_REMOTES]; /*-------------------------------------------------------------------*/ void UserMessage(int id, char *msg) { printf("%i: %s\n",id,msg); } int connectToRemote(char *ipAddrStr, short int rHandle, short int remPort, short int rProtocol) { printf("making a remote connection\n"); if (Remote[rHandle].address[0] != '\0') { sprintf(message,"remote handle %i already in use\n",rHandle); UserMessage(REMOTE,message); return(-1); } strcpy(Remote[rHandle].address,ipAddrStr); Remote[rHandle].port = remPort; /* rProtocol must be 1 for SOCK_STREAM or anything else for SOCK_DGRAM */ Remote[rHandle].protocol = (rProtocol==SOCK_STREAM)?SOCK_STREAM:SOCK_DGRAM; Remote[rHandle].incomming = ENABLE_INCOMMING; Remote[rHandle].TimeLiner = 0; if ((Remote[rHandle].socketFD = socket(PF_INET, Remote[rHandle].protocol,0)) == ERROR) { perror("unable to open socket"); return(ERROR); } bzero((char *)&Remote[rHandle].addr,sizeof(struct sockaddr_in)); Remote[rHandle].addr.sin_family = PF_INET; Remote[rHandle].addr.sin_port = htons(Remote[rHandle].port); Remote[rHandle].addr.sin_addr.s_addr = inet_addr(Remote[rHandle].address); if(Remote[rHandle].protocol == SOCK_STREAM) { if(connect(Remote[rHandle].socketFD,&Remote[rHandle].addr, sizeof(struct sockaddr_in)) == ERROR) { perror("unable to connect"); close(Remote[rHandle].socketFD); sprintf(message,"unable to connect to %s port %i\n", Remote[rHandle].address,Remote[rHandle].port); UserMessage(REMOTE,message); Remote[rHandle].address[0] = '\0'; return(ERROR); } } sprintf(message,"REMOTE %i connect to %s port %i using protocol %i\n", rHandle,Remote[rHandle].address,Remote[rHandle].port, Remote[rHandle].protocol); UserMessage(REMOTE,message); return(0); } int disconnectFromRemote(short int rHandle) { if (Remote[rHandle].address[0] == '\0') return(-1); close(Remote[rHandle].socketFD); Remote[rHandle].address[0] = '\0'; return(0); } startTLseq(short int rHandle, short int seqNumber, short int p1, short int p2, short int p3) { TLMESSAGETYPE tlMessage; TLSTEPREQUESTTYPE tlMsgContents; /* STEP task status packets */ tlMessage.messageType = 3; tlMessage.tlVersion = 1; tlMessage.objectSize = sizeof(TLSTEPREQUESTTYPE); tlMessage.objectId = 19; /* whatever the proper id is... */ tlMsgContents.Request = seqNumber; tlMsgContents.Param1 = p1; tlMsgContents.Param2 = p2; tlMsgContents.Param3 = p3; printf("setting msg contents\n"); /* memcpy(&tlMessage.data,&tlMsgContents,sizeof(TLSTEPREQUESTTYPE)); */ sprintf(message,"sending a STEP request [%x] [%x] to TIMELINER\n", (short int)tlMessage.data[0],(short int)tlMessage.data[2]); UserMessage(REMOTE,message); sendto(Remote[rHandle].socketFD,&tlMessage,sizeof(TLMESSAGETYPE),0, (struct sockaddr *)&(Remote[rHandle].addr),sizeof(struct sockaddr_in)); /* MDM Health and Status */ /* MDM telemetry */ } udptest(char *addr, short int rPort) { connectToRemote(addr,1,rPort,2); startTLseq(1,1,1,1,1); disconnectFromRemote(1); } - --------------15FB59E21CFB-- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks boot from scsi-pcmcia adapters Date: 27 Aug 1996 02:41:17 GMT From: michaeli@stout.atd.ucar.edu (Matt Michaelis) Organization: NCAR Message-ID: <4vtn8d$n6n@ncar.ucar.edu> References: <321B20D0.7BD1@kayser-threde.de> Sender: michaeli@ncar.ucar.edu To: bf@kayser-threde.de Subject: Re: VxWorks boot from scsi-pcmcia adapters Newsgroups: comp.os.vxworks In-Reply-To: <321B20D0.7BD1@kayser-threde.de> Organization: NCAR Cc: Bcc: michaeli Hello, In article <321B20D0.7BD1@kayser-threde.de> you write: >Hi, > >does anybody out there in netland know if it is possible to boot vxworks >from a scsi device on Motorola 68k or PPC CPUs, which reads pcmcia flash >cards. These are 3 1/2 " devices, which emulate a kind of harddisk and >allow standard pcmcia ram cards to be inserted and treated like a >removable disk. > >I tried this unsuccesfully a few months ago with VMEexec and the >standard Motorola Bug, which did not recognize the swiss made scsi >"drive" and now the question arises again with vxworks as the OS. >If somebody tried this, I'ld like to hear about it. As far as booting from any device, all you need to do is install the appropriate driver in the bootrom, and then modify the boot to load from your device. The "vxWorks User Manual" has some fairly good information on booting. With that said, I'm a bit confused by what you are trying to explain. One of the areas of confusion, is that there are at least three types of PCMCIA cards which can be used as disks, not including real disks. First, there are straight FLASH cards. These are rather complex to use as a true disk, because they require the Microsoft Flash File system (I don't know if Wind River is planning on supporting this at some point in time, but I've seen it mentioned in this newsgroup before). Second, there are SRAM cards with battery backup. These are relatively easy to use, since all you really need to to do is use the standard vxWorks RAMdrive device, and layer the vxWorks MSDOS file system on top of it. Finally, there are the FLASH cards which contain a chip to emulate the standard ATA type hard drive interface. If I understand you right, you have a SCSI PCMCIA card device, and you want to use the ATA interface FLASH cards in it. If this is correct, then the manufacturer of the SCSI PCMCIA device should be responsible for providing some sort of driver for it. However, it's possible, or perhaps likely that the manufacturer has a DOS or UNIX driver for it but no vxWorks driver. In this case you will either have to spend a lot of time or money to get one. In addition, you really need two drivers. First, you need something to run the PCMCIA interface and configure it for your cards. Second, you need an ATA disk driver for your card. I have done some PCMCIA work, and I may have part of what you need. Specifically, I have written what is known as a single point enabler which configures a Dynatem DMEM20 card to work with a SanDisk ATA FLASH card. I have also written an ATA driver which works with SanDisk cards (at least it works with 5MB and 10MB ones). The PCMCIA enabler just sets up addresses so that the ATA driver can work. I believe O+R in Germany sells the DMEM20 card under its own name, so if you were interested in using that instead of your SCSI device, you might save some effort. I have been considering putting my driver in the vxWorks archive, but just haven't done it yet. If you are interested in it, let me know. Matt Michaelis michaeli@ncar.ucar.edu --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Does 5.3 use interrupts to communicate? Date: 26 Aug 1996 15:37:11 GMT From: friedl@cher.heurikon.com (Ted Friedl) Organization: Heurikon Corporation Message-ID: <4vsgb7$iqb@badger.heurikon.com> References: In article , Lynn W. Becker, II wrote: >I am assuming that 5.3 (Tornado) uses interrupts to communicate. I have a >Heurikon Baja board running 5.2 and a Motorola MV1300 running 5.3 Beta. >When I do a windsh to connect to the MV1300 the Baja terminal fills with >these messages... > > >-> interrupt: >Uninitialized Interrupt! >Vector number 60 (0-255). non-cpu bus error interrupt >Program Counter: 0x80052f20 >Status Register: 0x30006703 >interrupt: Error address: 0xff04c240, Error ID: 0x0003 [snip] Lynn, With the Baja family, the base of VME A16 space lives at local address 0xff040000. The Baja, therefore, is getting a bus error accessing 0xc240 of A16 space. I assume you are using the sm (or bp) driver - otherwise a default Baja BSP does not access A16 space. The MV1300 probably puts its mailbox at 0xc240 of A16 (check out SM_INT_TYPE, SM_INT_ARG1, SM_INT_ARG2 and SM_INT_ARG3 on the MV1300). In my mind, the two most likely scenarios are: 1. The MV1300's mailbox is not at 0xc240 of A16 like it says it is. or 2. The Baja's mailbox is colliding with the MV1300's. The second case can only happen if the Baja has a processor number of 2. [The thing that is not clear in my mind is: Why is this not seen until starting windsh? Please give us "netters" more information about your system.] As always, make sure you have your VME system controller "on" and in the left most slot. Ted Friedl Senior Software Engineer Heurikon Corporation Madison, WI --------------------------- Newsgroups: comp.os.vxworks Subject: Req: Driver for Cirrus Logic CD1865/1864/180 octal UART Date: Tue, 27 Aug 1996 07:51:42 GMT From: 100567.1652@compuserve.com (Rolf Grzibek) Organization: Philips BTS Message-ID: <4vu9co$1rj@news.Dortmund.Germany.EU.net> Hi all, does somebody know about a driver for the Cirrus Logic octal serial communication controllers CD-180, CD-1864, CD-1865? Written by WRS or others. They should all be compatiible. TIA, Rolf - -- Rolf Grzibek 100567.1652@compuserve.com --------------------------- Newsgroups: comp.os.vxworks Subject: IP-488 driver questions!!! Date: Mon, 26 Aug 1996 15:33:46 +0930 From: Peter Hoffensetz Organization: Defence Science & Technology Organisation Message-ID: <32213E42.538B@dsto.defence.gov.au> Hi, I have just purchased a Greenspring IP-488 (IEEE-488 Controller) industry pack which I will be using on either a MVME162 or a VIPC616. I've got a few quotes so far for the VxWorks driver for this IP ... all were about TEN times the cost of the IP!!! Does anyone know of any alternative cheaper sources for this software? OR Has anyone written a driver for this IP? Is this a simple task? Do you get enough documentation with the IP (I haven't recieved it yet)? Any help is greatly appreciated. Regards, Peter Hoffensetz (Peter.Hoffensetz@dsto.defence.gov.au) --------------------------- End of New-News digest ********************** From rlin@sisa.samsung.com Tue Aug 27 09:33:03 1996 From: Robert Lin - SISA/Telecom Date: Tue Aug 27 09:33:05 PDT 1996 subscribe From lbecker@avtec.com Tue Aug 27 10:56:23 1996 From: "Lynn W. Becker, II" Date: Tue Aug 27 10:56:25 PDT 1996 Subject: Re: Does 5.3 use interrupts to communicate? Ted Freidl wrote: >I assume you are using the sm (or bp) driver - otherwise >a default Baja BSP does not access A16 space. This is true. > The MV1300 >probably puts its mailbox at 0xc240 of A16 (check out >SM_INT_TYPE, SM_INT_ARG1, SM_INT_ARG2 and SM_INT_ARG3 on >the MV1300). > >In my mind, the two most likely scenarios are: > > 1. The MV1300's mailbox is not at 0xc240 of A16 like > it says it is. > > or > > 2. The Baja's mailbox is colliding with the MV1300's. > >The second case can only happen if the Baja has a processor >number of 2. > >[The thing that is not clear in my mind is: Why is this not > seen until starting windsh? Please give us "netters" more > information about your system.] I changed proc=2 proc=5 (90.0.0.6) and the problem went away. The config. Slot 0: Heurikon Baja Processor 0 90.0.0.1 206.161.148.30 Slot 1: Heurikon Baja Processor 1 90.0.0.2 Slot 2: Heurikon Baja Processor 2 90.0.0.3 Slot 3: Heurikon Baja Processor 3 90.0.0.4 Slot 4: MV1301 Processor 4 90.0.0.5 Here are the Mailbox setup values for the MV1301. #define GCSR_GROUP_ADDR (0xc2) /* A15-A8 for brd mbox addr */ #define SM_INT_TYPE SM_INT_MAILBOX_1 /* 1-byte write mbox int */ #define SM_INT_ARG1 VME_AM_SUP_SHORT_IO /* bus address space */ /* bus address */ #define SM_INT_ARG2 ((GCSR_GROUP_ADDR << 8) + \ (sysProcNumGet() << 4) + 2) #define SM_INT_ARG3 0x08 /* value (SIG3 bit) */ The processor number is set to 4. so the address would be c242 for the mailbox address if I did the arithmetic correct.. The Baja mailboxes are setup using the following #define SM_INT_TYPE SM_INT_MAILBOX_1 #endif /* SM_POLL */ #define SM_INT_ARG1 VME_AM_SUP_SHORT_IO #define SM_INT_ARG2 ((int)((HKBAJA_DEFAULT_BP_MBOX+HKBAJA_MBOX_ADDR_BIAS) | ((USHORT)(sysProcNumGet())<<8))) #define SM_INT_ARG3 0 #define HKBAJA_DEFAULT_BP_MBOX 0x0021 #define HKBAJA_MBOX_ADDR_BIAS 0xc000 so the Processors A16 Mailboxes would be at Proc 0: 0xc021 Proc 1: 0xc121 Proc 2: 0xc221 Proc 3: 0xc321 Proc 4: 0xc242 So there appears to be a conflict.... Is this allowed to happen under VxWorks? It seems like the mailboxes are aligned in A16 space as a function of ProcNum but are not standard across different boards and BSPs? ______________________________________________________________________ Mr. Lynn W. Becker lbecker@avtec.com Engineer (703) 273-2211 Avtec Systems Inc (703) 273-1313 FAX 10530 Rosehaven Street, Suite 300 http://www.avtec.com Fairfax, VA 22030 From atp@ssd.ray.com Tue Aug 27 16:50:08 1996 From: Pickering Date: Tue Aug 27 16:50:11 PDT 1996 Subject: Routine daemon Is there anyone out there in VxWorks land that has source code for a route-d routing daemon (broadcasts subnet routes) that they would be willing to share ? Thanks in advance!! Allan Pickering, Raytheon Electronic Systems Division 1847 West Main Road, Portsmouth, RI 02871-1087 Phone: (401) 842-4071 Email: atp@ssd.ray.com From daemon@csg.lbl.gov Wed Aug 28 04:00:47 1996 From: daemon@csg.lbl.gov Date: Wed Aug 28 04:00:49 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Wed Aug 28 04:00:43 PDT 1996 Subject: Re: IP-488 driver questions!!! Subject: Re: SCRAMNet by Systran, anyone? ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: IP-488 driver questions!!! Date: 27 Aug 1996 14:31:19 GMT From: leif@tsl.uu.se (Leif Thuresson) Organization: The Svedberg Lab Message-ID: <4vv0rn$i2g@columba.udac.uu.se> References: <32213E42.538B@dsto.defence.gov.au> In article <32213E42.538B@dsto.defence.gov.au>, Peter.Hoffensetz@dsto.defence.gov.au says... > >Hi, > >I have just purchased a Greenspring IP-488 (IEEE-488 >Controller) industry pack which I will be using on >either a MVME162 or a VIPC616. > >I've got a few quotes so far for the VxWorks driver >for this IP ... all were about TEN times the cost of >the IP!!! > >Does anyone know of any alternative cheaper sources >for this software? > >OR > >Has anyone written a driver for this IP? Is this a >simple task? Do you get enough documentation with the >IP (I haven't recieved it yet)? > >Any help is greatly appreciated. > > > >Regards, > >Peter Hoffensetz >(Peter.Hoffensetz@dsto.defence.gov.au) I don't know of any free vxWorks driver, but there is a driver for IP-488 in the hideos package ( http://www.aps.anl.gov/asd/controls/hideos/ ). It's written in C++ and it's supposed to run under the hideos system on a standalone CPU communicating over the backplane with a another CPU running vxWorks as an intelligent I/O module. This isn't exactly what you asked for, but if you are considering writing your own vxWorks driver this driver might be of some use to understand how the IP-488 module works. If you do write your own driver and are willing to share it I'm very interested ! /Leif leif@tsl.uu.se --------------------------- Newsgroups: comp.os.vxworks Subject: Re: SCRAMNet by Systran, anyone? Date: Tue, 27 Aug 1996 12:15:27 -0500 From: Shash Chatterjee Organization: Lockheed Martin Tactical Aircraft Systems Message-ID: <32232D2F.6034@cliffy.lmtas.lmco.com> References: <4vahnq$2q8@dopey.keithley.com> <4vf0s4$kfg@news.sei.cmu.edu> Theodore Marz wrote: > > I believe so. I think that Evans & Sutherland put out a line of image > generators that are VME based (motorola MVME-147/167?). One of the > interfacing options for these systems (ESIG-1000/2000, etc) is SCRAMNET. > > Ted Marz Although none of our VxWorks/VME systems use SCRAMNet, other people here have succesfully used them to link PC's and PDP-11s. From what I've seen, there's not much to integrate other than to initialize the cards one time at bootup. From then on, it looks like memory (ie, make sure your processor can access the VMEbus space that the SCRAMNet card resides on and that the pointers in your code point to SCRAMNet memory). Shash - -- +-----------------------------------------------------------------+ + @WORK: ChatterjeeS@lmtas.lmco.com @HOME: chatterj@flash.net + + Sasvata Chatterjee | + + "SHASH-WATA" | Lockheed Martin Tactical Aircraft Systems+ + (817) 763-1495 | Lockheed Boulevard, MZ2118 + + (817) 777-2115 (FAX) | Fort Worth, TX 76108 + +-----------------------------------------------------------------+ --------------------------- End of New-News digest ********************** From leonid@rst.co.il Wed Aug 28 04:58:27 1996 From: leonid@rst.co.il (Leonid Rosenboim) Date: Wed Aug 28 04:58:29 PDT 1996 Subject: Web site, news feed Dear VxWorkers, I appologize if you get this announcement twice, I have posted it couple of days ago on the newsgroup, and did not get it back via the exploder mailing list (yet ?!), so I am giving up, resending the announcement, this time via the exploder. For the occasion of moving our Web site to a new server, we are offering a number of free software modules for VxWorks on the Web, and of course some product information :-) Also, we are interested to add more VxWorks-related links to out page, so if you are aware of any such URL's, please let me know. Enjoy! ----------------------------------------------------------------------- Leonid Rosenboim Phone: +972-3-559-8144 R S T Software Industries Ltd. Mobile: +972-50-307-142 P.O.Box 11502, AZUR 58017, Israel Fax: +972-3-559-8244 WWW: http://www.rst.co.il E-Mail: leonid@rst.co.il From ConradBj@aol.com Wed Aug 28 09:29:26 1996 From: ConradBj@aol.com Date: Wed Aug 28 09:29:28 PDT 1996 Subject: 1553 Interface to Motorola MVME167, vxworkd 5.2 How do I set up VX works to enable VME vectored interrupts from 1553 operation. 1553 board is a DDC65522. I can be reached by telephone at 415-941-1769 or fax at 415-941-5896 or B. Conrads e-mail. Thanks Bill Bjorkman. From goodrich@noao.edu Wed Aug 28 17:53:38 1996 From: goodrich@noao.edu (Bret Goodrich) Date: Wed Aug 28 17:53:55 PDT 1996 Subject: Fast Ethernet Can anyone suggest or recommend 100BaseTX (Fast Ethernet) VME boards that have VxWorks support? Thanks, bret -------------------- Bret Goodrich, Gemini 8 Meter Telescopes Project, Tucson, Arizona, USA Internet: goodrich@gemini.edu SPAN/HEPNET: noao::goodrich Phonenet: 1.520.318.8522 Snailnet: PO Box 26732, Tucson, AZ 85726 From ozkose@venus.aselsan.com.tr Thu Aug 29 00:42:50 1996 From: Ozgu OZKOSE ERDOGAN Date: Thu Aug 29 00:42:52 PDT 1996 Subject: force cpu5v - pentek 4269 interface Hello, We have developed a project using force cpu3ce sparc board and pentek 4269 dsp board. Now, we are using cpu5v instead of cpu3ce. Using cpu5v we have some problems with the interface to pentek 4269. These two boards are using the 4269s shared memory simultaneously. But while they are accessing this memory area, SPARC reads some garbage data even though we have not encountered any bus error. We can make SPARC to read the correct data by repeating the bus cycle. We have checked the cycles with bus analyser but there is no significant reason for the error. If you any idea on the reason of problem, please inform us. Any idea will be appriciated. Hoping to hear from you. Best Regards, Ozgu ERDOGAN, ozkose@venus.aselsan.com.tr From daemon@csg.lbl.gov Thu Aug 29 04:00:20 1996 From: daemon@csg.lbl.gov Date: Thu Aug 29 04:00:23 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Thu Aug 29 04:00:15 PDT 1996 Subject: soreceive crashes in 5.1 on a Force 5CE Subject: MVME162 Questions : IP & SRAM Subject: vxWorks kernel size Subject: Re: work queue overflow Subject: spy doesn't do anything Subject: Re: MVME162 Questions : IP & SRAM Subject: sysAuxClk: help using > 1 Subject: pthreads port to vxworks ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: soreceive crashes in 5.1 on a Force 5CE Date: 28 Aug 1996 12:07:59 GMT From: Joerg Albert Organization: Ericsson Radio Systems AB Message-ID: <501cqv$7n2@erinews.ericsson.se> Hi all, we are using VxWorks 5.1 on a Force 5CE Sparc board and we get a crash ("Memory address not aligned") in soreceive() when we call read() on a socket with a high data throughput. It seems that some control data got corrupted. OK, it might be our application running wild. I don't know the data structures used by soreceive, so I cannot say if the complete struct is overwritten or just certain values. - - Is the network code of VxWorks 5.1 based on some public domain sources or elsehow available in source ? Or could someone explain the parameters of soreceive to me ? - - Does anyone knows about similar crashes ( during read/close operations on the same socket by different tasks) ? Thanks for your help ! /Joerg. - -- #include "standard.disclaimer" - -----------------+--------------------------------------+ | T/BT Joerg Albert | | | Ericsson Radio Systems AB | tel: +46-8-4045577 | | Torshamnsgatan 23, Kista | fax: +46-8-7575720 | | S-16480 Stockholm, Sweden | email: Joerg.Albert@eedn.ericsson.se | +------------------------------+--------------------------------------+ --------------------------- Newsgroups: comp.os.vxworks Subject: MVME162 Questions : IP & SRAM Date: 28 Aug 1996 02:51:59 GMT From: steved@pacifier.com@199.2.117.163 (Steve Doiel) Organization: Pacifier BBS, Vancouver, Wa. ((360) 693-0325) Message-ID: <500c8f$3j2@news.pacifier.com> Reply-To: steved@pacifier.com (Steve Doiel) Is it possible to have my application set the IP number of the VxWorks system during boot time? How? I would like to use the user defined jumpers on the 162 to select from one of several IP numbers we use on a small intranet. I plan to have the software set the IP number based on these jumpers at boot time. Also, the MVME 162 we are using has 512K of battery backed SRAM. How do I go about making use of this memory? The BSP manual has a couple of routines for accessing non-volatile memory, but gives no clue as to what VxWorks uses and what is available to the application program. Any assistance would be most appreciated. steved@woodland.usnr.com --------------------------- Newsgroups: comp.os.vxworks Subject: vxWorks kernel size Date: Wed, 28 Aug 1996 09:00:32 -0500 From: Steve Woloschek Organization: GE Medical Systems Message-ID: <32245100.17C9@med.ge.com> I am trying to configure a vxWorks 5.3 kernel with only the following services: - multitasking - message queues - binary semaphores i.e. no networking, no tty, no formatted io, etc. I can only get the image size down to ~90K. There are still alot of other librarires being brought in for some unknown reason. I think this functionality should only consume ~32K at the most. Does anyone know how to get an extremely small kernel? Steve Woloschek woloschekS@med.ge.com GE Medical Systems --------------------------- Newsgroups: comp.os.vxworks Subject: Re: work queue overflow Date: 27 Aug 1996 19:01:54 GMT From: David Meeks Organization: Siemens Medical Systems, Ultrasound Group Message-ID: <4vvgn2$7jf@hood.sqi.com> References: <321F7F76.D0B@aminet.co.kr> I had this same error a couple of years ago on a MC68040. It turned out that I was not disabling interrupts from my serial port controller in my ISR so eventually the workQ filled up. Hope this helps, Dave Meeks Siemens Medical Systems. --------------------------- Newsgroups: comp.os.vxworks Subject: spy doesn't do anything Date: 28 Aug 1996 17:04:55 -0400 From: smarple@aol.com (Smarple) Organization: America Online, Inc. (1-800-827-6364) Message-ID: <502c9n$g09@newsbf02.news.aol.com> Reply-To: smarple@aol.com (Smarple) Sender: root@newsbf02.news.aol.com Working with VxWorks 5.1.1 and VADS, and a large number of tasks (>150), spy just returns without generating any output (at least within what seems to be a reasonable amount of time). Please don't ask why there are so many tasks -- it's not my project. :) spyReport doesn't produce any output either. Is there a known (spy) problem with large numbers of tasks? I cannot say that there was no modification to the kernel which precludes spy from working. Running spy with a frequency of 30 seconds didn't help. Thanks --------------------------- Newsgroups: comp.os.vxworks Subject: Re: MVME162 Questions : IP & SRAM Date: 28 Aug 1996 15:45:00 -0600 From: bgeer@xmission.xmission.com (bgeer) Organization: XMission Internet (801 539 0900) Message-ID: <502eks$158@xmission.xmission.com> References: <500c8f$3j2@news.pacifier.com> steved@pacifier.com@199.2.117.163 (Steve Doiel) writes: >Is it possible to have my application set the IP number of the VxWorks system >during boot time? How? >I would like to use the user defined jumpers on the 162 to select from one of >several IP numbers we use on a small intranet. I plan to have the software >set the IP number based on these jumpers at boot time. The network interfaces are initialized in: /src/config/usrNetwork.c look for usrNetIfAttach() & usrNetIfConfig(). I've made flash rom'able kernels with networking included & with the bootup network initialization disabled, then included network initialization function calls in my startup script. >Also, the MVME 162 we are using has 512K of battery backed SRAM. How do >I go about making use of this memory? The BSP manual has a couple of >routines for accessing non-volatile memory, but gives no clue as to what >VxWorks uses and what is available to the application program. How do you want to use it? One way is to use it with: /src/usr/ramDrv.c as a ram disk... - -- <> Robert Geer & Donna Tomky / * <> <> bgeer@xmission.com | _o * o * o <> <> Salt Lake City, Utah | -\<, * <\ <> U S A | O/ O __ /__, /> <> --------------------------- Newsgroups: comp.os.vxworks Subject: sysAuxClk: help using > 1 Date: Mon, 26 Aug 1996 09:07:48 -0700 From: Dave Mack Organization: EDS/Controls group Message-ID: <3221CBD4.7DB@eds.com> Hello, I would greatly appreciate tips on how to use more than one Aux.clock. I am using a MVME68060. My code uses one of the board's aux. clocks for polling I/O using the sysAuxClk functions. I am trying to also use StethoScope but am not able to because it also requires the use of an aux. clock. It seems I might have to write drivers to use the other aux. clocks on by board? Thanks for any help...if I do need to write a driver any code/tips are also greatly appreciated. David Mack EDS 300 E.Big Beaver Rd. Troy, MI 48083 - ------------ lnustoc.jzgmwk@eds.com --------------- --------------------------- Newsgroups: comp.os.vxworks Subject: pthreads port to vxworks Date: 27 Aug 1996 15:35:35 -0600 From: kscdenv@stout.entertain.com (Greg Greene) Organization: ABWAM, Inc., full service ISP, voice: 1+ 303 730-6050 Message-ID: <4vvpn7$mla@stout.entertain.com> Is there a public domain port of pthreads to vxWorks available? Thanks Greg Greene KineticSystems Corp. 303-220-7455 x13 --------------------------- End of New-News digest ********************** From prb@aplexus.jhuapl.edu Thu Aug 29 04:57:53 1996 From: prb@aplexus.jhuapl.edu (Paul R. Bade) Date: Thu Aug 29 04:57:56 PDT 1996 Subject: Re: Fast Ethernet Bret Goodrich writes: > > Can anyone suggest or recommend 100BaseTX (Fast Ethernet) VME boards > that have VxWorks support? > You should check with VI Computer (800) VME-CPUS We are using their Power4B (PowerPC 604e) board which has all I/O out the P2 connector including 100BaseT4. It also has a AUI interface, SCSI, four RS-232 ports, etc. I believe that they also make a version that has 100BaseTX out of the front panel. +====================================================================+ | __ ____ __ __ | | /\ \ / \ \ /\_\__ /\ \ Johns Hopkins University | | / \_\ | /\ \ \ / / /\_\\ \ \ Applied Physics Lab. | | / /\ | |\ \/ \ \ / / / / / \ \ \ | | / \/ |_| \ /\ \_\ / / / / / \ \ \ Paul R. Bade | | / /\__/_/ \ \ \/_// / / / / / / / (301)-953-6000 x8681 | | / / / \ \_\ \ \/ / / / / / prb@aplexus.jhuapl.edu | | \/_/ \/_/ \__/_/ \/_/ | | __ ____ __ __ | | /\ \ / \ \ /\ \ /\ \ | | / \_\ | /\ \ \ / \_\ / \ \ | | / /\ | |\ \/ \ \ / /\ |_| / /\ \ \ | | / \/ |_| \ /\ \_\ / / / | | \/ / \ \ | | / /\ / / \ \ \/_// / / / / / /\ \_\ | | \ \/ / / \ \_\ \ \/ / / \/ / / / | | \__/_/ \/_/ \__/_/ /_/_/ | | | +====================================================================+ From froeber@bbn.com Thu Aug 29 07:23:44 1996 From: Fred Roeber Date: Thu Aug 29 07:23:47 PDT 1996 Subject: Re: MV162 SRAM Steve Doiel asks: > Also, the MVME 162 we are using has 512K of battery backed SRAM. How do > I go about making use of this memory? The BSP manual has a couple of > routines for accessing non-volatile memory, but gives no clue as to what > VxWorks uses and what is available to the application program. The battery backed SRAM on the 162 boards is somewhat unusual in terms of board designs. It is distinct from the "NVRAM" on the board which uses an 8KB memory block inside the real time clock chip. That NVRAM area has very specific uses in VxWorks. The board also has DRAM memory which is what VxWorks uses as the "RAM" when it runs the system. The SRAM, on the other hand, isn't used explicitely by VxWorks at all and is available for any type of application use. To use it you simply set a pointer to its local address and store whatever you want there. Hope this helps. Fred -- | Fred J Roeber, BBN Systems & Technologies | | 50 Enterprise Place Middletown, RI 02842-5202 | | froeber@bbn.com 401-849-2543 (X48) | | See http://www.bbn.com/tv for TraceMaker info | From stan@rti.com Thu Aug 29 11:51:40 1996 From: Stan Schneider Date: Thu Aug 29 11:51:42 PDT 1996 Subject: Re: sysAuxClk: help using > 1 >> Newsgroups: comp.os.vxworks >> Date: Mon, 26 Aug 1996 09:07:48 -0700 >> From: Dave Mack >> I would greatly appreciate tips on how to use more than one >> Aux.clock. >> I am using a MVME68060. My code uses one of the board's aux. clocks >> for polling I/O using the sysAuxClk functions. I am trying to also >> use StethoScope but am not able to because it also requires the use >> of an aux. clock. It seems I might have to write drivers to use the >> other aux. clocks on by board? >> Two comments: First, StethoScope's demo program uses the aux clock by default, but there are many other ways to collect data. Contact "support@rti.com" if you have questions. Second, we have drivers for the two extra clocks on the vme2 chip; I'm sure that's on the mv147 boards, it's probably on yours too. The drivers are free for the asking; contact me or "support@rti.com". HTH, -- Stan ============================================================================= = = = = Stan Schneider = email: stan@rti.com = = Real-Time Innovations, Inc. = Phone: (408) 720-8312 x104 = = 155A Moffett Park Drive, Suite 111 = Fax: (408) 734-5009 = = Sunnyvale, CA 94089 = http://www.rti.com = = = = ============================================================================= From budh@venus.gvg.tek.com Thu Aug 29 13:10:14 1996 From: budh@venus.gvg.tek.com (Bud Hensley) Date: Thu Aug 29 13:10:16 PDT 1996 Subject: Two Ethernets Has anyone run the Motorola MC68EN360 processor at 25mHZ with two ethernets. There is a controversy on performance, if two can be active the same time. Does the CPM risc have enough horsepower to run 2 copies of the ethernet protocol microcode at the same time. Thanks, Sheridan email: Sheridan.Hensley@tek.com From tkb@mclean.sparta.com Thu Aug 29 14:22:27 1996 From: "Thomas K. Buchanan" Date: Thu Aug 29 14:22:29 PDT 1996 Subject: Re: Two Ethernets At 01:10 PM 8/29/96 PDT, you wrote: >Submitted-by budh@venus.gvg.tek.com Thu Aug 29 13:10:14 1996 >Submitted-by: budh@venus.gvg.tek.com (Bud Hensley) > > >Has anyone run the Motorola MC68EN360 processor at 25mHZ >with two ethernets. There is a controversy on performance, >if two can be active the same time. Does the CPM risc have >enough horsepower to run 2 copies of the ethernet protocol >microcode at the same time. It is my understanding that the '360 can only physically and electrically support one 10Mbit/s interface at a time. At least that is what my technical manual says... adios ----------------------------------------- /\\ Keith Buchanan ///\\ /////\\ SPARTA, Inc. ///// \\ 7926 Jones Branch Drive / \\\\\ \\ Suite 900 \ \\\\\ // McLean, VA 22102 \ ////// \////// Voice: 703 448 0210 \//// FAX: 703 893 5494 \// WWW: www.mclean.sparta.com From daemon@csg.lbl.gov Fri Aug 30 04:00:54 1996 From: daemon@csg.lbl.gov Date: Fri Aug 30 04:00:57 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Fri Aug 30 04:00:51 PDT 1996 Subject: Re: Fast Ethernet Subject: Re: Force SPARC target Subject: RPC 4.0 for VxWorks.. Subject: vxWorks benchmarks? ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Fast Ethernet Date: 29 Aug 1996 15:04:47 GMT From: friedl@cher.heurikon.com (Ted Friedl) Organization: Heurikon Corporation Message-ID: <504bif$klb@badger.heurikon.com> References: <199608290053.RAA27489@icarus.gemini.edu> In article <199608290053.RAA27489@icarus.gemini.edu>, Bret Goodrich wrote: >Can anyone suggest or recommend 100BaseTX (Fast Ethernet) VME boards >that have VxWorks support? Bret, The Heurikon BajaNet has VxWorks support and can support one or two 100BaseTX ports. http://www.heurikon.com/products/BajaNet.html tells the story. The BajaNet uses the Baja4700 (MIPS 4700 processor) as a platform for Fast Ethernet. What the web page does not say is a similar product will be available using the Baja68k (Mot 68LC040, 040 or 060 processor) and VxWorks 5.3, if it is not available already. Ted Friedl Senior Software Engineer Heurikon Corporation Madison, WI --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Force SPARC target Date: 29 Aug 1996 20:54:18 GMT From: pww@cray.com (Paul Wells) Organization: Cray Research, Inc. Message-ID: <50501q$h2f@walter.cray.com> References: In article , Robert Redfield wrote: > Subject: Time:1:42 PM > OFFICE MEMO Force SPARC target Date:8/23/96 > >Any opinions, comments or experiences running VxWorks on Force 5V, or 20? We've started using the Force 5V on a project here, and it seems like an excellent board. It's a fairly complete clone of the Sun SPARCstation 5 (with the addition of an SBus/VME bridge), which we've been running VxWorks on for several years now. - - Paul Wells Silicon Graphics, Inc. --------------------------- Newsgroups: comp.os.vxworks Subject: RPC 4.0 for VxWorks.. Date: Thu, 29 Aug 1996 10:05:50 +0100 From: Pankaj Shroff Organization: Alta Technology Corporation Message-ID: <32255D6E.64CE@altatech.com> i am porting code running on IRIX/SGI to VxWorks/DECAlpha i am using version 3.2 of VxWorks from DEC. i have run into problems related to RPC/XDR library of VxWorks. is there a more recent implementation of rpc (like rpc4.0) for DEC/VxWorks. ? also, i was trying to port the sun rpc4.0/xdr routines to VxW myself but had problems particularly with xdr_float() and xdr_double() routines. they work at times and then suddenly crash! are these bugs on VxW only ? are there any fixes ? thank you much - -pps, altatech. --------------------------- Newsgroups: comp.os.vxworks Subject: vxWorks benchmarks? Date: Thu, 29 Aug 1996 11:15:39 -0800 From: mlang@tsunami.jpl.nasa.gov (M Lang) Organization: JPL Message-ID: Hi, Are there any benchmark codes that I can run on different vxWorks targets (68k, ppc, mips etc..) to get some performance numbers? I would appreciate any pointers. Thanks a lot. - Minh - --------------------------- End of New-News digest ********************** From al@columbia.mcd.mot.com Fri Aug 30 07:32:32 1996 From: Al Currano Date: Fri Aug 30 07:32:35 PDT 1996 Subject: Fast Ethernet; PowerPC VME Boards & VX/Works | Bret Goodrich writes: | | > | > Can anyone suggest or recommend 100BaseTX (Fast Ethernet) VME boards | > that have VxWorks support? | > Motorola Computer Group has announced a new family of PowerPC-based VME Single Board Computers (SBCs) based on the new PowerPlus architecture which I strongly recommend you take a look at for your application. There are versions which support both 100BaseTx and 100BaseT4. A VX/Works board support package is also available. The PowerPlus architecture is a radical improvement over existing SBC architectures. It incorporates a state-of-the-art memory controller (Falcon) which takes full advantage of PowerPC address pipelining and provides interleaved access to a 144-bit main memory bus to deliver a processor memory bandwidth up to 400+ Mbytes/sec -- three times the speed of existing implementations. PowerPlus also incorporates the Raven PCI bridge chip which supports 64-bit PCI for a maximum throughput of 64 MBytes/sec on the PCI local bus. All other VME implementations I am aware of support only 32 bit PCI transfers. PowerPlus includes two series of VME boards: the MVME2600 series is a single-slot, full-featured SBC series with either the MPC603e (200 MHz) or MPC604e (166MHz & 200 MHz) PowerPC processor, up to 256 MB ECC memory, up to 9 MB Flash, Ethernet (10 Mbit or 100 MBit), fast/wide SCSI, 4 serial ports (2 async, 2 synchronous) parallel port, floppy, keyboard & mouse interfaces; the MVME3600 is a 2 slot board series which adds SVGA graphics w/2MB Video RAM, up to 1 GByte ECC memory and optional dual ethernet and SCSI interfaces. Both series can be ordered optionally with 256K L2 cache. Both series of boards support one PCI Mezzanine Card (PMC) onboard, and two additional PMC cards via an attached carrier board. You can order these boards from Motorola now. List prices start at $4495 (quantity 1). Production shipments of some models will start in September; production shipments of other models will follow in October. Call you local Motorola Computer Group sales office for more info, or look us up on the Web at http://www.mot.com/computer/. Or feel free to call or E-mail me if you have any questions. -- Al Currano _/ _/ _/_/_/_/_/ _/_/_/_/_/ al_currano@mcg.mot.com _/_/ _/_/ _/ _/ _/ _/ Motorola Computer Group _/ _/_/ _/ _/ _/ 8945 Guilford Rd., Ste 145 _/ _/ _/ _/ _/ Columbia, MD 21046 _/ _/ _/ _/ _/_/ Phone: (410) 312-5930 _/ _/ _/ _/ _/ _/ FAX: (410) 381-4496 _/ _/ _/_/_/_/_/ _/_/_/_/_/ From schen3@ccgate.hac.com Fri Aug 30 09:18:26 1996 From: schen3@ccgate.hac.com Date: Fri Aug 30 09:18:28 PDT 1996 Subject: How to verifying a BSP How do I test or verify a BSP from my supplier? Is there a general procedure on document available out there? Regard, Steven Chen schen3@ccgate.hac.com From MikeMcQuade@connect.com Fri Aug 30 09:50:21 1996 From: Mike McQuade Date: Fri Aug 30 09:50:23 PDT 1996 Subject: Re: Fast Ethernet; PowerPC VME Boards & VX/Works At 07:32 AM 8/30/96 PDT, you wrote: >Submitted-by al@columbia.mcd.mot.com Fri Aug 30 07:32:32 1996 >The PowerPlus architecture is a radical improvement over existing SBC >architectures. It incorporates a state-of-the-art memory controller >(Falcon) which takes full advantage of PowerPC address pipelining and >provides interleaved access to a 144-bit main memory bus to deliver a >processor memory bandwidth up to 400+ Mbytes/sec -- three times the >speed of existing implementations. PowerPlus also incorporates the Raven >PCI bridge chip which supports 64-bit PCI for a maximum throughput of 64 >MBytes/sec on the PCI local bus. All other VME implementations I am >aware of support only 32 bit PCI transfers. > I think thats 264 Mbytes / Sec on the PCI bus. A couple of other things I like about the new Moto card are: the RAVEN ASIC which allows interrupts to be prioritized, this card unlike most of the other PowerPC VME cards will allow a VME interrupt to be higher priority than a Keyboard int. The Newbridge PCI-VME chip should be a major improvement over the MVME-160x boards, VMECHIP2 VME2PCI combo. Mike. MikeMcQuade@connect.com From jlao@yurie.com Fri Aug 30 11:21:08 1996 From: jlao@yurie.com Date: Fri Aug 30 11:21:11 PDT 1996 Subject: vxgdb Hi, Is there anybody exprienced the cursor of the VxGDB/MIPS 2.0 disappear when you setp into a function. We have the source of the function, and we remove the optimization option. The VxWork version we use is 5.2. We posted this question to windriver and their response is that the source should be match with the object file. And after that we get run around of answer. Is there anybody could send us a hint to fix this problem. Thanks, James From al@columbia.mcd.mot.com Fri Aug 30 13:29:24 1996 From: Al Currano Date: Fri Aug 30 13:29:26 PDT 1996 Subject: Motorola PowerPlus VME Boards - Correction & Addition My apologies for a typographical error in my earlier message. The 64 bit PCI local bus used in Motorola's PowerPlus VME boards (MVME26 and MVME360 series) is capable of achieving a maximum throughput of 264 MBytes/sec, not 64 MBytes/sec. -- Al Currano _/ _/ _/_/_/_/_/ _/_/_/_/_/ al_currano@mcg.mot.com _/_/ _/_/ _/ _/ _/ _/ Motorola Computer Group _/ _/_/ _/ _/ _/ 8945 Guilford Rd., Ste 145 _/ _/ _/ _/ _/ Columbia, MD 21046 _/ _/ _/ _/ _/_/ Phone: (410) 312-5930 _/ _/ _/ _/ _/ _/ FAX: (410) 381-4496 _/ _/ _/_/_/_/_/ _/_/_/_/_/ From stan@rti.com Fri Aug 30 14:35:51 1996 From: Stan Schneider Date: Fri Aug 30 14:35:53 PDT 1996 Subject: Re: RPC 4.0 for VxWorks.. >> Date: Thu, 29 Aug 1996 10:05:50 +0100 >> From: Pankaj Shroff >> i am porting code running on IRIX/SGI to VxWorks/DECAlpha >> i am using version 3.2 of VxWorks from DEC. >> i have run into problems related to RPC/XDR library of VxWorks. >> is there a more recent implementation of rpc (like rpc4.0) for >> DEC/VxWorks. ? >> also, i was trying to port the sun rpc4.0/xdr routines to VxW myself >> but had problems particularly with xdr_float() and xdr_double() >> routines. they work at times and then suddenly crash! >> are these bugs on VxW only ? are there any fixes ? >> There's a bug in xdr_double() for little-endian architectures (like Alpha) in versions of VxWorks prior to 5.3 (including all DEC versions). I've appended the fixed code; compile it & load it before your app. HTH, -- Stan stan@rti.com ======================================================================== #include #define RTI_ENDIAN_LITTLE bool_t xdr_double( XDR *xdrs, double *dp) { register long *lp; switch (xdrs->x_op) { case XDR_ENCODE: lp = (long *)dp; #if defined(RTI_ENDIAN_LITTLE) return (XDR_PUTLONG(xdrs, lp+1) && XDR_PUTLONG(xdrs, lp)); #else return (XDR_PUTLONG(xdrs, lp) && XDR_PUTLONG(xdrs, lp+1)); #endif break; case XDR_DECODE: lp = (long *)dp; #if defined(RTI_ENDIAN_LITTLE) return (XDR_GETLONG(xdrs, lp+1) && XDR_GETLONG(xdrs, lp)); #else return (XDR_GETLONG(xdrs, lp) && XDR_GETLONG(xdrs, lp+1)); #endif break; case XDR_FREE: return (TRUE); break; } return (FALSE); } From isys!isys.com!kareem@uunet.uu.net Fri Aug 30 18:12:56 1996 From: kareem@isys.com (Kareem Fawell) Date: Fri Aug 30 18:12:58 PDT 1996 Subject: Subject: Sockets with a timeout Hello: I am using SOCK_STREAM type sockets and I would like to have a connect and read with a timeout. If I were using UNIX I could do something with signals, but this is not supported on vxWorks. Any ideas would be appreciated. Kareem Fawell Isys Controls, Inc Alameda, CA 94501 From daemon@csg.lbl.gov Sat Aug 31 04:02:02 1996 From: daemon@csg.lbl.gov Date: Sat Aug 31 04:02:04 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sat Aug 31 04:01:59 PDT 1996 Subject: How to detect end of socket connection Subject: Re: No UDP packet Subject: Re: soreceive crashes in 5.1 on a Force 5CE Subject: Re: MVME162 Questions : IP & SRAM Subject: Re: Routine daemon Subject: Re: MVME162 Questions : IP & SRAM Subject: Scheme , SCM , Guile , port ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: How to detect end of socket connection Date: 30 Aug 1996 19:08:32 GMT From: junkes@scrutinizer.fhi-berlin.mpg.de (Heinz Junkes) Message-ID: <507e7g$j6v@fu-berlin.de> I have open a socket connection with socket, bind, listen, accept then i read endless commands form the connection..... now the other computer crached ..... the read request gets now always data ... without end How can i check that a connection is still alive ??? Heinz - ----------------------------------------------------------------------------- Fritz-Haber-Institut | Phone: (+49 30) 8413-4270 Heinz Junkes | Fax (G3+G4): (+49 30) 8413-5900 Faradayweg 4-6 | E-Mail: junkes@fhi-berlin.mpg.de D - 14195 Berlin | NeXT-Mail: junkes@mynext.rz-berlin.mpg.de - ----------------------------------------------------------------------------- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: No UDP packet Date: Wed, 28 Aug 1996 22:13:25 GMT From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star Message-ID: <502g9o$m7k@idiom.com> References: <3221CCFD.794B@boris.msfc.nasa.gov> Reply-To: hjb@peacefulstar.oakland.ca.us David Perkinson wrote: >typedef struct tlMessageType >{ > short int messageType; > short int tlVersion; > short int objectId; > short int objectSize; > char data[4096]; >} TLMESSAGETYPE; The problem is here. char data[4096] makes TLMESSAGETYPE size to be larger than maximum default UDP message size allowed to be sent on VxWorks. You can print the current limit for max UDP datagram size you can send from the shell: - -> udp_sendspace By default, it is set to 2048. This is a BSD Tahoe legacy. Most modern TCP/UDP/IP code has increased this number. You can easily make this limit higher either from the shell or in your code using the global udp_sendspace: - -> udp_sendspace=5*1024 for example. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: soreceive crashes in 5.1 on a Force 5CE Date: Wed, 28 Aug 1996 21:16:38 GMT From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star Message-ID: <502cv8$jh9@idiom.com> References: <501cqv$7n2@erinews.ericsson.se> Reply-To: hjb@peacefulstar.oakland.ca.us Joerg Albert wrote: >- Is the network code of VxWorks 5.1 based on some public domain > sources or elsehow available in source ? > Or could someone explain the parameters of soreceive to me ? ftp://gatekeeper.dec.com/pub/BSD/net1/src/sys/sys/uipc_socket.c.Z --------------------------- Newsgroups: comp.os.vxworks Subject: Re: MVME162 Questions : IP & SRAM Date: Wed, 28 Aug 1996 21:18:53 GMT From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star Message-ID: <502d3f$jh9@idiom.com> References: <500c8f$3j2@news.pacifier.com> Reply-To: hjb@peacefulstar.oakland.ca.us steved@pacifier.com@199.2.117.163 (Steve Doiel) wrote: >Is it possible to have my application set the IP number of the VxWorks system >during boot time? How? in usrConfig.c, after usrNetInit(), call ifAddrSet("ei0","xxx.xxx.xxx.xxx"); --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Routine daemon Date: Wed, 28 Aug 1996 21:22:03 GMT From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star Message-ID: <502d9d$jh9@idiom.com> References: <199608272348.TAA29078@grouper.ssd.ray.com> Reply-To: hjb@peacefulstar.oakland.ca.us Pickering wrote: >Is there anyone out there in VxWorks land that has source code for a >route-d routing daemon (broadcasts subnet routes) that they would be >willing to share ? ftp://ftp.atd.ucar.edu/pub/vxworks/vx/vxrouted* (in 8 parts) --------------------------- Newsgroups: comp.os.vxworks Subject: Re: MVME162 Questions : IP & SRAM Date: 31 Aug 1996 02:45:04 GMT From: steved@pacifier.com@199.2.117.163 (Steve Doiel) Organization: Pacifier BBS, Vancouver, Wa. ((360) 693-0325) Message-ID: <5088vg$2gg@news.pacifier.com> References: <500c8f$3j2@news.pacifier.com> <502d3f$jh9@idiom.com> Reply-To: steved@pacifier.com (Steve Doiel) In <502d3f$jh9@idiom.com>, chaos@idiom.com (H.J. Bae) writes: >steved@pacifier.com@199.2.117.163 (Steve Doiel) wrote: > >>Is it possible to have my application set the IP number of the VxWorks system >>during boot time? How? > >in usrConfig.c, after usrNetInit(), call >ifAddrSet("ei0","xxx.xxx.xxx.xxx"); > Thanks! I was kind of guessing this might be the case, but didn't know if VxWorks would be happy if you change the IP address after it is initially set based on the default command line string. SteveD --------------------------- Newsgroups: comp.os.vxworks Subject: Scheme , SCM , Guile , port Date: 30 Aug 1996 15:34:23 GMT From: constant@alph04.triumf.ca (Constantin Vrinceanu) Organization: The University of British Columbia Message-ID: Has anyone tried/succeded/failed doing a port of some Scheme implementation (on VxWorks) ? Siod should be easy (i think), but i would want more. Thank you in advance. - -- Constantin 7306 --------------------------- End of New-News digest **********************