From msa@redifon.com Thu Feb 1 01:09:05 1996 From: "M. S. Armstrong" Date: Thu Feb 1 01:09:08 PST 1996 Subject: Re: NVRAM File System Implementations??? Hello Leonid Rosenboim, Concerning your request for experience with the Greenspring IP-serial. We have no direct experience with that IP but have programmed other serial IPs from GreenSpring, TEWS and Actis you may be interested in. Our requirement was to receive as many channels of synchronous RS422 at 500kbit/sec per channel as possible using one MVME162FX (the 32MHz version with DMA on the IP bus). The results of our testing are: 1) TEWS quad IP using Z85C30, 1 channel at 100% processor loading. 2) Actis quad IP using Z85C230 with 4 byte input FIFO, 8 channels at 100% loading. (interrupt rate reduced by four) Both of these with extremely customised drivers to get round constant SCC interrupts on the processor and the associated delays in responding to them. 3)GreenSpring IP MP-Serial, usies the Z16C30 with 32byte receive FIFO and 16-bit DMA IP bus access (the only commercially developed IP allowing DMA todate). This IP should permit 4 channels on DMA and 4 channels interrupt driven to be implemented with 80-90% loading. Note: this is the maximum that can fit on one board as the IP MP-serial has only a dual channel capability. We have tested 2 channels using DMA and this gives ~10% loading. We're still tring to get our driver to go faster and waiting for fixes to the DMA channel decode by Greenspring to fully realise the 8 channelconfiguration. It should be possible to push the IP MP-Serial to much higher data rates when using a few channels per board. If your interested in the driver software for the IP MP-Serial, it should be possible to come to an arrangement. The IP-Serial using the Z85C30 like the TEWS IP is unlikely to fulfill your high data-rate requirement. Best of Luck, Mark Armstrong Proessional Engineer Redifon MEL Ltd. From daemon@csg.lbl.gov Thu Feb 1 04:00:34 1996 From: daemon@csg.lbl.gov Date: Thu Feb 1 04:00:37 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Thu Feb 1 04:00:28 PST 1996 Subject: Re: Zero-copy TCP driver Subject: Re: Multiple networks and proxy arp Subject: More VME boards/systems available Subject: VxWorks 5.1.1 to VxWorks 5.2 Subject: Re: (no subject) Subject: Q: Error message Subject: PPP protocol Subject: Timer Questions Revisited ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Zero-copy TCP driver Date: 31 Jan 1996 02:11:06 GMT From: KEVIN SWEENEY Organization: MCI/News Corp. Message-ID: <4emj3q$hpg@merlin.delphi.com> References: <4e3qba$t45@lll-winken.llnl.gov> <4e90o6$sm4@bcarh8ab.bnr.ca> cjwein@bnr.ca (Christopher Wein) wrote: >In article <4e3qba$t45@lll-winken.llnl.gov>, >David P Wiltzius,b113,21551, wrote: >>I'm interested in acquiring the source of any >>working zero-copy TCP driver for VxWorks, although >>particularly for Ethernet, ATM, and HIPPI. >> >>Indeed, at this instant I would be most interested >>in one for Ethernet so I can proceed with development >>of the applications. Later I'll be using this >>as a basis for ATM and HIPPI to get the performance. >> >>I'll be glad to share my efforts, esp. to those >>that respond to the extent of our review and release >>policy. > >I was under the impression that VxWorks 5.2 tcp/ip was zero copy... > >Or am I mistaken? >-- >*********************************************************************** >Chris Wein * Phone: (613) 765-5648 >Bell-Northern Research, Ottawa * Fax: (613) 763-2286 >cjwein@bnr.ca * >*********************************************************************** Isn't Zero-copy what the Zbuf socket interface is all about? Or does that simply elimate copying between the application and TCP/IP stack. Perhaps there is more copying performed internal to the stack. Does anyone know? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Multiple networks and proxy arp Date: 31 Jan 1996 13:58:01 GMT From: jford@arcturus.gb.nrao.edu (John Ford) Organization: National Radio Astronomy Observatory Message-ID: References: <9601261720.AA11068@arcturus.gb.nrao.edu> Reply-To: jford@nrao.edu In article lawrence@world.std.com (Scott Lawrence) writes: In article <9601261720.AA11068@arcturus.gb.nrao.edu>, jford@sadira.gb.nrao.edu (John Ford) wrote: ] MV167 MV167 ] +---------+ Sm +--------+ ] Eth---+ net1.20 | Net | net1.21+--- Eth ] Net | net2.20 |--------| net2.21| Net ] 1 +---------+ +--------+ 2 ] Proxy from Proxy from ] net1.20 to net1.21 net2.21 to net2.20 You appear to contradict yourself here - you have Ethernets 1 & 2, but they both have the same IP network number? Are they connected otherwise (a bridge perhaps?). No, ethernets 1 and 2 are distinct class C networks. I would like to be able to go through either class C network to get to either board, without the knowledge of the external system (i.e., routers,gateways, etc.) The system exists now, with both boards on one ethernet. We are adding in a second ethernet, and need to keep the connection to the old network. John --------------------------- Newsgroups: comp.os.vxworks Subject: More VME boards/systems available Date: 31 Jan 1996 15:10:57 GMT From: innov8@ix.netcom.com (M.W. ) Organization: Netcom Message-ID: <4eo0q1$buu@cloner3.netcom.com> I have a couple of Motorola M4520 systems available. MVME167 68040 processor, 12-slot rack, 300 SCSI, 150 Tape. For more information e-mail or http://205.230.70.48/computer.html E-mail: innov8@ix.netcom.com - -- E-mail: innov8@ix.netcom.com --------------------------- Newsgroups: comp.os.vxworks Subject: VxWorks 5.1.1 to VxWorks 5.2 Date: 31 Jan 1996 16:46:43 GMT From: abrahamb@utica.ge.com Organization: Lockheed Martin Corp, Valley Forge PA Message-ID: <4eo6dj$c2j@morgan.vf.mmc.com> Reply-To: abrahamb@utica.ge.com We have been using VxWorks 5.1.1 for quite a while. I believe the current version of VxWorks is 5.3(Tornado). Is that correct? I was thinking of switching to 5.2 if it wasn't a hassle. I don't want to transition to Tornado just yet. I was wondering the following. 1) How hard is the transition from 5.1.1 to 5.2 ? 2) What is the difference between 5.1.1 to 5.2 ? Thanks --------------------------- Newsgroups: comp.os.vxworks Subject: Re: (no subject) Date: Wed, 31 Jan 1996 11:45:16 -0800 From: Greg Rasche Message-ID: <310FC6CC.5495@gi.com> References: <1996Jan22.231112.21922@nosc.mil> Peter, I fixed a few bugs in one written at APLabs(San Diego)...while contracting for APLabs. That is a very very harry board & probably what I'd call the driver from hell!! and I've written several drivers. The ASIC doesn't necessarily work the way it is documented to work (luckily someone else at APLabs figured it out before I got that far). You'll need lots of luck to get things working and if you're doing any high rate serial I/O you should plan on full load for 3 - 6 man months. It is NOT easy, clean, or straight forward. That's a nightmare in my past & I hope to never go through it again. Good luck, Greg peter donich wrote: > > I would like to receive a SBE VCOM24 device driver for VxWorks Unix > kernel. The driver will allow me to transmit and receive on the serial > port of the VCOM24 board. Please respond to tran_sp@nosc.mil. > Thanks. Sean Tran --------------------------- Newsgroups: comp.os.vxworks Subject: Q: Error message Date: 31 Jan 1996 20:28:01 GMT From: bpatil@bnr.ca (Basavaraj Patil) Organization: Bell-Northern Research, Richardson, TX Message-ID: <4eojch$774@crchh327.rich.bnr.ca> Hi, What does this error message mean??? 0x1184788 (tRdbRun): memPartFree: invalid block 0x1184704 in partition 0x5b5930. 0x11b794c (t2): memPartFree: invalid block 0x11ad3e4 in partition 0x5b5930. Any info appreciated. :) Raj - -- --------------------------- Newsgroups: comp.os.vxworks Subject: PPP protocol Date: Thu, 1 Feb 1996 06:32:54 GMT From: yokada@icrl.mew.co.jp (Yukio Okada) Organization: Information & Communication R&D Lab Message-ID: Sender: news@icrl.mew.co.jp Does anyone know if VxWorks supports PPP protocol? If not, is there any PPP products from third party? Any information will be appreciated. Thanks, ___Yukio Okada Matsushita Electric Works, Japan --------------------------- Newsgroups: comp.os.vxworks Subject: Timer Questions Revisited Date: Thu, 1 Feb 1996 09:09:33 GMT From: Michael Migdol Organization: Matsushita El. Works, Ltd., Central Research Lab Message-ID: <3110834D.3A9@crl.mew.co.jp> Sender: news@isc.mew.co.jp (USENET News System) About a week ago, I posted a question asking "Why does VxWorks (or any OS for that matter) need a periodic interrupt?" I got a lot of responses; thanks to everyone who took the time to help. The problem is that I didn't state the question very well, and most everyone told me what I already knew : they're used for pre-emptive multitasking and watchdog timers. However, what I really meant to ask was: "is there any benefit to having a periodic interrupt as opposed to scheduling each interrupt separately based on the current timer value?" In other words, if your context switching is occurring at 6 hertz (every 10 ticks for the default 60 hz PIT), the interrupts on the intermediate 9 ticks are useless unless you have a need for a fairly low-resolution (16 ms unit) timer. Similarly, if you do taskDelay(5), why not schedule an interrupt for 5 ticks in the future using the timer. The OS requires slightly more overhead per interrupt (since it has to now keep track of which event it needs to set up an interrupt for), but the required number of interrupts goes down significantly. This might actually be a better question for comp.os.research, but I figured I'd ask this question in a VxWorks context first. On a more applications-oriented side... My original goal was a microsecond-resolution timer, which turned out to be a fairly simple problem. I can read the PIT chip counter (on my platform, a little over 1 MHz) directly. The only problem is that I need to be able to atomically read both the system tick counter (with tickGet) and the PIT chip in order to avoid problems during the rollover. Disabling interrupts is one possibility. Rewriting tickAnnounce is another. Does anyone have better suggestions or experience? Sorry so long-winded, but I hope at least some of you made it this far and can help me out. :-) Thanks in advance, Michael - -- =================================================================== Michael A. Migdol マイケル・ミグドル Matsushita Electric Works, Ltd. Tel. (06) 906 - 8075 Central Research Laboratory Fax (06) 904 - 7104 1048 Kadoma, Osaka 571 Japan e-mail mmigdol@crl.mew.co.jp 　　松下電工株式会社中央研究所信号処理・ソフトウェア研究室 --------------------------- End of New-News digest ********************** From daemon@csg.lbl.gov Fri Feb 2 04:00:39 1996 From: daemon@csg.lbl.gov Date: Fri Feb 2 04:00:42 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Fri Feb 2 04:00:34 PST 1996 Subject: Question on filesystem performance Subject: BOOT_PARAMS Subject: Re: Booting 167 using FTP Subject: Re: Flash files system Subject: Re: Flash files system (me too!) Subject: Access Fault ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Question on filesystem performance Date: 1 Feb 1996 14:06:50 GMT From: Romano Lanzarotti Organization: Aitek Srl (Genova, Italy) Message-ID: <4eqhdq$mog@alpha.it.net> We have found that after creating a large number of files (several thousands) in one directory, the performance of filesystem operations (directory listing, file creation, file deletion, etc) drops to an extremely low level (5 - 10 seconds per operation). This holds true even after all the original files have been removed and the directory has been emptied. The speed of operations in other directories is normal, though. Our setup is: VxWorks 5.1.1 Motorola 147 board SCSI disk with dosFs filesystem Does anyone know if there is an operating system-level solution or workaround for this problem? Thanks in advance. Andrea Oddera oddera@aitek.it Romano Lanzarotti rem@aitek.it --------------------------- Newsgroups: comp.os.vxworks Subject: BOOT_PARAMS Date: 1 Feb 1996 16:42:47 GMT From: idawson@mlsma.mlm.att.com (nsi3520000-Dawson) Organization: AT&T Message-ID: <4eqqi7$6l6@ssbunews.ih.att.com> Hi Folks I want to update the boot parameters from my application software. Is there a way to do this ?. After using bootParamsShow and bootStringToStruct updating the value using bootStructToString how can I write this new boot string back to nvram ?. Is there a function which will accept the boot parameter string pre-pended with a $ and write the new boot parameters ?. Alternatively, is there a way to obtain the address of the boot parameters in nvram and use some of the above as described - finally writing them to nvram, the BOOT_PARAMS is a fixed size structure. Thanks Ian --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Booting 167 using FTP Date: 2 Feb 1996 03:15:36 GMT From: steved@199.2.117.161 Organization: Pacifier BBS, Vancouver, Wa. ((360) 693-0325) Message-ID: <4ervko$as6@news.pacifier.com> References: Reply-To: steved@pacifier.com (Steve Doiel) In , cjones@digsys1.bwi.wec.com (Chris Jones) writes: > > >We are using Tornado on a Win 95 platform. > >Do we need to be running an NFS client or can VxWorks be botted using FTP? > That depends... What version of VxWorks and what FTP server? We found that Microsoft's FTP server on Windows 3.11 would not work with VxWorks 5.1.1. WRS recommended using Chamelion, but we opted for WPFTP (SP?) which is a shareware FTP server that worked on Windows 3.11. The newer version of VxWorks (5.2) is not as fussy. Hope this helps. SteveD --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Flash files system Date: 2 Feb 1996 01:13:12 GMT From: nungeste@sdd.hp.com (Rick Nungester) Organization: Hewlett Packard, San Diego Division Message-ID: <4erof8$738@news.sdd.hp.com> References: <4eam0p$6vb@isoit109.bbn.hp.com> Phone: 1(800)628-8686 mail: flash@inside.intel.com FaxBack: 1(800)628-2283 BBS: (916)356-3600 Web site: http://www.intel.com/embedded/index.html --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Flash files system (me too!) Date: 2 Feb 1996 01:14:03 GMT From: nungeste@sdd.hp.com (Rick Nungester) Organization: Hewlett Packard, San Diego Division Message-ID: <4erogr$738@news.sdd.hp.com> References: <4eam0p$6vb@isoit109.bbn.hp.com> <4elhrs$ue@news.sanders.lockheed.com> <4elr8f$oms@news.sdd.hp.com> My mistake -- the correct information is: Phone: 1(800)628-8686 mail: flash@inside.intel.com FaxBack: 1(800)628-2283 BBS: (916)356-3600 Web site: http://www.intel.com/embedded/index.html --------------------------- Newsgroups: comp.os.vxworks Subject: Access Fault Date: 1 Feb 1996 16:48:14 +1100 From: dmr@f111.iassf.easams.com.au (Dewan RASHID) Organization: EASAMS (Australia) Pty Ltd Message-ID: <4epk6u$ej8@f111.iassf.easams.com.au> Dear VxPerts: I am running a s/w which copies data to and from a 1553 card memory. For sometime it runs quite fine but after a while I get the following fault. Access Fault Program Counter: 0x00d94e96 Status Register: 0x3010 Access Address : 0x3ff70dba Special Status : 0x0505 Task: 0xe33cdc "t14" I am not sure about the proble or the cause of the problem. Any help/suggestions will be greatly appreciated. Many thanks, Dewan GEC-Marconi Systems Sydney, Australia. - ------------------ PS: I am using VxWorks in MVME-167. --------------------------- End of New-News digest ********************** From msa@redifon.com Fri Feb 2 05:03:22 1996 From: "M. S. Armstrong" Date: Fri Feb 2 05:03:24 PST 1996 Subject: Re: Green Spring IP-Serial at high rates Dear all, (Please excuse the poor subject label last time I exploded this.) Hello Leonid Rosenboim, Concerning your request for experience with the Greenspring IP-serial. We have no direct experience with that IP but have programmed other serial IPs from GreenSpring, TEWS and Actis you may be interested in. Our requirement was to receive as many channels of synchronous RS422 at 500kbit/sec per channel as possible using one MVME162FX (the 32MHz version with DMA on the IP bus). The results of our testing are: 1) TEWS quad IP using Z85C30, 1 channel at 100% processor loading. 2) Actis quad IP using Z85C230 with 4 byte input FIFO, 8 channels at 100% loading. (interrupt rate reduced by four) Both of these with extremely customised drivers to get round constant SCC interrupts on the processor and the associated delays in responding to them. 3)GreenSpring IP MP-Serial, usies the Z16C30 with 32byte receive FIFO and 16-bit DMA IP bus access (the only commercially developed IP allowing DMA todate). This IP should permit 4 channels on DMA and 4 channels interrupt driven to be implemented with 80-90% loading. Note: this is the maximum that can fit on one board as the IP MP-serial has only a dual channel capability. We have tested 2 channels using DMA and this gives ‾10% loading. We're still tring to get our driver to go faster and waiting for fixes to the DMA channel decode by Greenspring to fully realise the 8 channelconfiguration. It should be possible to push the IP MP-Serial to much higher data rates when using a few channels per board. If your interested in the driver software for the IP MP-Serial, it should be possible to come to an arrangement. The IP-Serial using the Z85C30 like the TEWS IP is unlikely to fulfill your high data-rate requirement. Best of Luck, Mark Armstrong Proessional Engineer Redifon MEL Ltd. From ztf@one.one.com Fri Feb 2 06:09:14 1996 From: ztf@one.com (Zachary T. Frey) Date: Fri Feb 2 06:09:22 PST 1996 Subject: Re: Question on filesystem performance >We have found that after creating a large number of >files (several thousands) in one directory, the performance >of filesystem operations (directory listing, file >creation, file deletion, etc) drops to an extremely low >level (5 - 10 seconds per operation). > >This holds true even after all the original files have been >removed and the directory has been emptied. The speed of >operations in other directories is normal, though. > >Our setup is: VxWorks 5.1.1 > Motorola 147 board > SCSI disk with dosFs filesystem This is a DOS file system "feature", not specific to VxWorks. You would see the same performance problem with any other OS using the DOS file system. I don't remember all the details, but it happens because of the way DOS organizes its directory entries. Once you get over a certain number of files in a directory, the directory entries start being stored in a very inefficient way, which will lead to the access times you're seeing. >Does anyone know if there is an operating system-level solution >or workaround for this problem? The workaround I ended up having to use was to segregate my files into subdirectories. The only way I know to restore performance on a directory that has had that many files in it is to delete and recreate the directory. -- Zach Frey Open Networks Engineering, Inc. (ONE) Voice: (313) 996-9900 x 1235 777 E. Eisenhower Parkway, Suite 650 Fax: (313) 996-9908 Ann Arbor, MI 48108 Internet: ztf@one.com From Flemming_Kevin/macom_pbl_ctc@mailback.macom.com Fri Feb 2 08:33:54 1996 From: Flemming_Kevin/macom_pbl_ctc@mailback.macom.com Date: Fri Feb 2 08:33:57 PST 1996 Subject: Re: Question on filesystem performance Item Subject: Message text >We have found that after creating a large number of >files (several thousands) in one directory, the performance >of filesystem operations (directory listing, file >creation, file deletion, etc) drops to an extremely low >level (5 - 10 seconds per operation). > >This holds true even after all the original files have been >removed and the directory has been emptied. The speed of >operations in other directories is normal, though. > >Our setup is: VxWorks 5.1.1 > Motorola 147 board > SCSI disk with dosFs filesystem >Does anyone know if there is an operating system-level solution >or workaround for this problem? The DOS filesystem allocates a fixed size block for a directory. As the directory grows, it will need to allocate more space for the directory. Since you've been adding data to the disk, this usually means that the clusters closest to the original directory have been allocated for data. Therefore the directory ends up getting fragmented across the disk. For a large directory, that can mean many sections. Once allocated, directory space is not deallocated until the directory is deleted, EVEN IF all of the files are deleted from the directory. For each directory operation, the disk heads have to seek to each of these fragmented clusters. At 15-30ms per seek, it adds up. Possible solutions: If possible, use smaller directories. Delete directories when possible. It was rumored last year that WindRiver would be putting in a command to allow pre-allocation of a directory. Assuming that you know how big it can grow, this would save tremendous time because it would leave all of the directory in one area of the disk. Track-to-track seeks are much faster than random seeks. If WindRiver hasn't come up with that yet, I have another idea. Starting with a blank partition, you can write a small program to open and immediately close a number (how many do you want?) of zero length temporary file names. This should allocate the directory in one small area of the disk. The temporary files can then be deleted. As I said earlier, this leaves the directory allocated. I haven't tried it yet, but it should work. A better long-term solution is for WindRiver or a third party to provide a utility analogous to the Norton Utilities for DOS, which among other things can clean-up directories. Good luck! Kevin J. Flemming M/A-COM, Inc. M/S 115 P.O. Box 3295 Lowell, MA 01853 (Work) (508) 442-5025 flemming@ctc.macom.com From pardoe@ncp.gpt.co.uk Fri Feb 2 08:36:05 1996 From: "Matthew Pardoe" Date: Fri Feb 2 08:36:07 PST 1996 Subject: problem sith socket over slip Hello, I am having a problem running sockets over a SLIP link between two vxworks platforms. One platform runs as the server and the other as a client. On startup from power on I can set up the socket ok and send data between the two without any problems. The problem occurs when I simulate failure conditions that cause the client to reboot. When this happens the server end correctly notices that the client is no longer there and closes the socket. the server then loops back to wait at its accept statement for the client to reconnect. After the client platform has rebooted it attempts to connect to the server. At this point everything stops. The server is sat in its accept statement, and the client on its connect statement. On looking at tcpstatShow it appears that the original connections have not closed down correctly. If I delay the reboot of the client for long enough (several minutes) then the connections timeout and the client is able to successfully connect to the server again. I have tried disabling so_linger and setting up keep alive, but with no success. I am not too familiar with exactly what affect these settings should have, can anyone suggest what socket options I need to set to ensure failed socket connections are cleared down as soon as possible. =============================================================================== Matthew Pardoe _/_/ _/_/ _/ _/_/_/_/ email : pardoe@ncp.gpt.co.uk GPT Ltd. _/ _/ _/ _/ _/ _/ tel : +44 1202 782704 Sopers Lane _/ _/ _/ _/ _/ _/_/_/_/ fax : +44 1202 782571 Poole BH17 7EQ _/ _/ _/_/_/_/ _/ sneakernet : B Suite "The Pit" England =============================================================================== From zot!Schilling.com!stevec@netcom.com Fri Feb 2 09:04:46 1996 From: stevec@schilling.com (Steve Cohan) Date: Fri Feb 2 09:04:49 PST 1996 Subject: Re: Access Fault > I am running a s/w which copies data to and from > a 1553 card memory. For sometime it runs quite > fine but after a while I get the following fault. I sent (directly) Dewan some suggestions on resolving access faults that I received previously from exploder readers. If anyone else would like this info, I can pass it on. Maybe it should go into the FAQ. . . ...................................................................... . . . Steve Cohan Telerobotics for nuclear, . . Chief Engineer, Software/Controls subsea, and other extreme . . GEC Alsthom environments . . Schilling Robotic Systems, Inc. . . . . 1632 Da Vinci Court tel : 916 753 6718 . . Davis, CA 95616 fax : 916 753 8092 . . email : stevec@schilling.com . . . ...................................................................... From smith@ampersand.com Fri Feb 2 12:25:07 1996 From: smith@ampersand.com (Steve Smith) Date: Fri Feb 2 12:25:10 PST 1996 Subject: VxWorks East Coast Users Group Meeting I am pleased to announce that the 1996 East Coast VxWorks Users Group meeting will be held on Monday, April 1 (no fooling :-) ). The meeting will be held in Boston, MA just before Embedded Systems Conference which is scheduled for April 2 - 4. For new users or users who have never attended, this is a great opportunity to learn about future directions of VxWorks. It can also be very valuable to discuss problems, issues, and ideas with other users, many of whom have been working with VxWorks for 5 or more years. I have attended this meeting every year since 1989, and have found the signal to noise ratio at these meetings to be extremely high. The meeting is inexpensive, $65 at the door (cash and check only), or $50 by check or credit card (Visa or Mastercard -No AMEX) if you RSVP by March 20. Wind River Systems is graciously providing its customary level of support and has arranged for space at the Marriott Burlington. Sessions begin at 8:00 a.m. and last until approximately 4:30 p.m., with a break at noon for lunch. The day will wrap up with a cocktail reception with vendor exhibits. The morning sessions will consist of updates and presentations from Wind River Systems personnel, including: * a review of new products focusing on Tornado * a presentation on trends in real-time computing by Dave Wilner, Wind River Systems co-founder * a tutorial on Tornado APIs * a review of WRS's approach to C++ and the Wind Foundation Classes * an update and Q&A period with Dave Wilner, Wind River Systems co-founder and VP of Engineering. The afternoon session has traditionally been organized by us users. Past meetings have included: * guest speaker lectures (Mike Tieman of Cygnus a few years ago) * user presentations * birds-of-a-feather discussions on a variety of topics, including C++, ADA, networking, X & Motif and many others. Request for Suggestions !!! --------------------------- As the volunteer chairperson, I will arrange speakers and presentations and put together a final agenda for the afternoon session during the next month. In this vein, I would like to solicit suggestions >from you, the VxWorks community, regarding what you would like to see. Some of my thoughts are: 1) It was quite interesting to have Mike Tieman as a guest speaker a few years ago (good job Todd). It would be nice to have another distinguished speaker this year. Any ideas about whom people might like to see as a speaker, either on this topic or any other? 2) We have often had very interesting presentations by users with unique or exciting applications. If you are involved in such an application and would be willing to share details with the Users Group, please get in touch. (Is anyone from the New York City traffic system out there?) So please get in touch if you would like to give a presentation or have any ideas with regards to speakers, presentations, or anything else you'd like to see on the afternoon agenda. For those of you who don't know me, my name is Steve Smith. I preside over Ampersand, Inc., a 15-person software consulting/contracting firm specializing in Unix and real-time systems. I've worked with VxWorks on a variety of projects since 1988, and can remember the old days of VxWorks when people ran kernels other than wind. I also co-chaired the User's group meeting with Rudy DeBruyn in Fall of 1990, and on my own in 1993, thus making me the first person foolish enough to volunteer for this job three times :-). I can be contacted at: Stephen Smith President Ampersand, Inc. 700 Technology Park Dr., Bldg. 9 Billerica, MA 01821 s.smith@ampersand.com Phone: 508-671-0785 Fax: 508-671-0786 You can also follow up with Claire Chaplais Wind River Systems 1010 Atlantic Ave. Alameda, CA 94501 claire@wrs.com Phone: 510-814-2177 Fax: 510-814-2010 To register: Checks can be made payable to Wind River Systems. Please include your name, title, company, address, phone, fax numbers and email address. Send to Claire Chaplais at the above address. She will then fax you a confirmation letter and directions to the hotel. The address and phone for the meeting site are: Boston Marriott Burlington Route 128 and 3A Burlington, MA 01803 617-229-6565 I welcome your input, and look forward to a great meeting! Steve -------------------------------------------------------------------------- Stephen Smith Phone: (508) 671-0785 Ampersand, Inc. President Fax: (508) 671-0786 700 Technology Park Dr., Bldg 9 s.smith@ampersand.com Billerica, MA 01821 From byron@wrs.com Fri Feb 2 16:46:20 1996 From: byron@wrs.com (Byron Goulding) Date: Fri Feb 2 16:46:22 PST 1996 Subject: 360-040 companion mode I am looking for a BSP for the Motorola M68360QUADS-040 board. Has anyone out there run vxWorks on this board? Byron Goulding Wind River Systems Field Application Engineer (408) 562-5769 byron@wrs.com From jcn@atcwest.com Fri Feb 2 19:00:55 1996 From: jcn@atcwest.com (Chris Nickles) Date: Fri Feb 2 19:01:06 PST 1996 Subject: Re: Data Sheets for Serial Communication Controllers Zilog's phone number is (408) 370-8000. Signetics has been absorbed by Phillips Semiconductor. Phone (408) 991-2000. Also, Greenspring provides "engineering kits" which generally contain user manuals for the important chips on board. Chris Nickles Alliant Techsystems San Diego From daemon@csg.lbl.gov Sat Feb 3 04:00:40 1996 From: daemon@csg.lbl.gov Date: Sat Feb 3 04:00:43 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sat Feb 3 04:00:33 PST 1996 Subject: Re: Access Fault Subject: How to get The MORTIMER Story Subject: Re: VxWorks Software Architecture Subject: Re: VxWorks 5.1.1 to VxWorks 5.2 Subject: Re: How to format SCSI disk with scsiFormatUnit()? ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Access Fault Date: 2 Feb 96 08:26:06 -0500 From: "Bruce K. Gitlin" Message-ID: <1996Feb2.082606.1276@corning.com> References: <4epk6u$ej8@f111.iassf.easams.com.au> This is a multi-part message in MIME format. - ---------------------------------7524277964748 Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset=us-ascii dmr@f111.iassf.easams.com.au (Dewan RASHID) wrote: >Dear VxPerts: > >I am running a s/w which copies data to and from >a 1553 card memory. For sometime it runs quite >fine but after a while I get the following fault. > > Access Fault > Program Counter: 0x00d94e96 > Status Register: 0x3010 > Access Address : 0x3ff70dba > Special Status : 0x0505 > Task: 0xe33cdc "t14" > >I am not sure about the proble or the cause of the >problem. > >Any help/suggestions will be greatly appreciated. > > >Many thanks, > >Dewan >GEC-Marconi Systems >Sydney, Australia. >------------------ > >PS: I am using VxWorks in MVME-167. > > Use 'lkAddr' and the value of the Program Counter. It should give you the symbol that is causing the access violation, or at least point you in the right direction. - ---------------------------------7524277964748 Content-Transfer-Encoding: 7bit Content-Type: text/plain ________________________________________________________________ | ___ __ |¥ | | ¥ _ _ | / / ¥ . | | . | | | |--( |‾‾ | | / ` /_¥ |( ( __ | -+- | | |/¥ | | | |__/ | ¥_/ ¥_, ¥__ | ¥ . ¥__/ | | | | | | | | | | | | B r u c e K. G i t l i n Gitlin_BK@Corning.com | | |________________________________________________________________| | ¥________________________________________________________________¥| - ---------------------------------7524277964748-- --------------------------- Newsgroups: comp.os.vxworks Subject: How to get The MORTIMER Story Date: 2 Feb 96 11:56:29 GMT From: rabbit@tygra.Michigan.COM (Roger Rabbit) Organization: Shut Up Blair Message-ID: <14497@tygra.Michigan.COM> This posting was made automatically by machine. It will appear once every minute or so. Direct any questions to rabbit@Buster.Michigan.COM - ------- The MORTIMER story can be retreived from a special archive server (the Mort-Server) To request a part of the MORTIMER story, send mail to one of the addresses below. Put your request on the Subject: line of your message. Requests are can be the in the following form: ALL - (in upper case) will cause the entire story to be sent to you. n - (where n is an integer) will cause part one to be sent n,m... - (where n and m, etc are integers) will cause parts n, m and so forth, to be sent. You may send your request to any of the following addresses: MORTIMER@Buster.Michigan.COM MORTIMER@Babs.Michigan.COM MORTIMER@Plucky.Michigan.COM MORTIMER@Hamton.Michigan.COM MORTIMER@McLoon.Michigan.COM MORTIMER@Elmyra.Michigan.COM If you send off a request and don't get a response within 5 days, send e-mail to the sysop here (NOT ME!!). His address is: ARCHIVER@Michigan.COM - -- >>> BAN: Nuclear Power, US Intervention in The Gulf, Toxic Waste, >>> rdc, carasso, Trash Incinerators, Nuclear Weapons, Poverty, KiBoIsM >>> Racism, Kent Paul Dolan, Specieism, etc... Write: Rabbits for a Better >>> Hutch, Roscommon, MI 48653 E-MAIL: rabbit@Buster.Michigan.COM --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks Software Architecture Date: 2 Feb 1996 16:13:33 GMT From: Mike Campbell Organization: BTG, Inc. Message-ID: <4etd7d$577@server.btg.com> References: <310D60EF.13C8@digsys1.bwi.wec.com> <31118D70.21CC2C3@netcom.com> Roy Rowan wrote: >Chris Jones wrote: >> >> >> Are there any references for "good" programming techniques with VxWorks? >> >> Thanks >> Chris Jones >> Westinghouse Electronic Systems >> cjones@digsys1.bwi.wec.com > >Chris, > >All the posts that I have seen to date are very good and will do what >you wish. Chris, I agree with roy, just thought I would throw in my 2 cents. We had the problem of starting up (in a coordinated manner) mutiple task on multiple processors. Our solution was to have a starter task on the master board. All other tasks on all other boards would be started via the startup scripts, connect to our messaging task, and await a startup command. The last task started on all the boards connected to the startup task (all this via sockets). When the startup task detected that all these last tasks had connected, it would then begin sending out startup messages to the task awaiting the startup messages. The order of the startups was determined via a text file that could be edited. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks 5.1.1 to VxWorks 5.2 Date: Fri, 02 Feb 1996 11:43:48 +0100 From: Arjon Vlasblom Organization: High Tech Automation B.V. Message-ID: <3111EAE4.41C67EA6@HTA.nl> References: <4eo6dj$c2j@morgan.vf.mmc.com> abrahamb@utica.ge.com wrote: > > We have been using VxWorks 5.1.1 for quite a while. I believe the current > version of VxWorks is 5.3(Tornado). Is that correct? I was thinking of switching > to 5.2 if it wasn't a hassle. I don't want to transition to Tornado just yet. > I was wondering the following. > > 1) How hard is the transition from 5.1.1 to 5.2 ? > 2) What is the difference between 5.1.1 to 5.2 ? > > Thanks We have recently changed from 5.1.1 to 5.2 ! 1) No problems at all. Just install according the installation guide. 2) Besides a lot of bug fixes (45 pages in the release notes) some new features are added. - WindView support - additional POSIX functionality - NFS server - 'zero-copy TCP' - online hypertext reference manual (unix) Arjon Vlasblom System Engineer - -- //// (0 0) ____________________________________o00_( )_00o______________________________ / ¥ | Arjon.Vlasblom@HTA.nl : Eindhoven : Bilthoven | | High Tech Automation BV : Voice +31 40 2447465 : Voice +31 30 2251700 | | The Netherlands : Fax +31 40 2441260 : Fax +31 30 2251717 | | http://www.hta.nl | |-----------------------------------------------------------------------------| | Disclaimer: This message does not necessarily reflect the company's opinion,| | just my own humble one. | ¥_____________________________________________________________________________/ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How to format SCSI disk with scsiFormatUnit()? Date: Fri, 2 Feb 1996 15:53:24 GMT From: dap@network.com (Dave Peterson) Organization: Network Systems Corporation Message-ID: References: <4elmvs$jgv@cisu2.jsc.nasa.gov> Reply-To: dap@network.com Sender: news@network.com In article <4elmvs$jgv@cisu2.jsc.nasa.gov>, "Tom R. Burns" writes: > I need to write code to format a SCSI disk but the vxWorks 5.1 > Reference Manual does not explain the arguments to scsiFormatUnit() > such as the defect list format. I checked the include files for > appropriate #defines and data structures but found nothing. Is this > stuff drive and/or media unique? Here is some information which may > or may not be relevant: > > The drive is a Sony Magneto Optical (SMO) model 501; 320 Mbytes. > The vxWorks target is a MVME167. > > I've scanned through the SCSI FAQ and the comp.periphs.scis but I'm > still clueless. Any information or ideas? Look at the FORMAT UNIT command in the drive's interface manual or look at the SCSI spec for the FORMAT UNIT command for an explanation of the arguments to supply. - -- =================================================================== Dave Peterson phone : 612-391-1008 Senior Engineer Network Systems Corp. email: dap@network.com =================================================================== --------------------------- End of New-News digest ********************** From micf@lannet.com Sun Feb 4 01:52:44 1996 From: micf@lannet.com (Michael Feldman) Date: Sun Feb 4 01:52:47 PST 1996 Subject: Re: vxWorks System Architecture > < ...................... stuff deleted ...................... > Don't > use taskLock/taskUnlock unless someone threatens you with bodily harm, > those two functions are as bad as 'goto' and more often than not a sign > of poorly written multi-tasking code. I guess that it isn't 100% right to separate C operators to 'good' and 'bad', they aren't film personages. Sometimes 'goto' may be used in the code, and sometimes it even makes a code more understandable. Similarly, I guess that task locking for predictable time in some situations, such as a system initialization, doesn't seem too terrible. Regards, Michael Feldman, Madge Networks ESD (Israel) From leonid@rst.co.il Sun Feb 4 05:55:57 1996 From: leonid@rst.co.il (Leonid Rosenboim) Date: Sun Feb 4 05:55:59 PST 1996 Subject: Re: Flash files system We have developed a software package that implements the Microsoft Flash File System (FFS) specification on VxWorks. It supports various embedded Flash devices, as well as PCMCIA cards on PC targets and VME. A version that supports data compression will be available very soon, and will also provide a CRC-32 data integrity check. FlashLib is hsipped with a collection of drivers for many Intel and AMD devices, as well as PC cards. For full details please refer to the product data sheet at the FTP site below. ----------------------------------------------------------------------- Leonid Rosenboim Phone: +972-3-559-8144 R S T Software Industries Ltd. Mobile: +972-50-307-142 P.O.Box 11502, AZUR 58017, Israel Fax: +972-3-559-8244 ftp://knight.trendline.co.il/pub/rst E-Mail: leonid@rst.co.il From RUSSELL@qpsx.oz.au Sun Feb 4 23:11:54 1996 From: RUSSELL@qpsx.oz.au (Jones, Russell) Date: Sun Feb 4 23:12:39 PST 1996 Subject: GNU make I'm try to use GNU make to generate .o files in a different directory from the one containing my source code. I remember a mailing about this not too long ago, but a recent disaster deleted my mailbox, so I apologize if I'm going over old ground. I'm beta testing Tornado 1.0 under Windows NT, so it's possible I've got a bug, but I'm not too familiar with GNU make so it could well be me. I have set up several parallel directories, LIB, H, DEP and OBJ. LIB contains source code, H contains common headers, OBJ is for the object files, and DEP contains dependency files, generated using the -M compiler option. For example: D:¥TORNADO¥BETA¥SRC¥OBJ¥revreg.o: revreg.c D:¥TORNADO¥BETA¥SRC¥H/qpsxstd.h ¥ D:¥Tornado¥target/h/stdio.h D:¥Tornado¥target/h/types/vxANSI.h ¥ ... In my makefile I include the dependency files (this is the method recommended in section 4.12 of WindRivers Gnu toolkit users guide). As far as I can work out, make should then use the implicit rule (.c.o) to determine how to remake the object file. However, when I try to make D:¥TORNADO¥BETA¥SRC¥OBJ¥revreg.o, make insists that there is nothing to do. I've used the -d switch to see what make is doing internally, and that has confused me even more, since it says that the dependencies are newer than the target (see the listing at the end)! It looks to me like makes implicit rules only apply if the target and the dependency are in the same directory, although section 4.3.4 says they don't (I tried using VPATH and vpath, but they didn't help). Any suggestions would be gratefully received. +--------------------------------+----------------------------------+ | Russell Jones | rjones@qpsx.oz.au | +--------------------------------+----------------------------------+ | QPSX Communications Pty. Ltd., | http://webstar.qpsx.oz.au | | Private Bag No.24, | | | West Perth 6005, | phone : (+61) 9 262 2000 | | Western Australia. | fax : (+61) 9 324 1642 | +--------------------------------+----------------------------------+ --- make debug listing (edited) --- GNU Make version 3.72-95q2, by Richard Stallman and Roland McGrath. Copyright (C) 1988, 89, 90, 91, 92, 93, 94 Free Software Foundation, Inc. This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. Default shell = D:¥WINNT35¥system32¥cmd.exe Reading makefiles... Reading makefile `makefile'... Reading makefile `D:¥TORNADO¥BETA¥SRC¥DEP¥revreg.d' (search path) (no ‾ expansion)... Updating makefiles.... ... Updating goal targets.... Considering target file `D:¥TORNADO¥BETA¥SRC¥OBJ¥revreg.o'. Looking for an implicit rule for `D:¥TORNADO¥BETA¥SRC¥OBJ¥revreg.o'. Trying pattern rule with stem `D:¥TORNADO¥BETA¥SRC¥OBJ¥revreg'. Trying implicit dependency `D:¥TORNADO¥BETA¥SRC¥OBJ¥revreg.c'. ... No implicit rule found for `D:¥TORNADO¥BETA¥SRC¥OBJ¥revreg.o'. ... Finished dependencies of target file `D:¥TORNADO¥BETA¥SRC¥OBJ¥revreg.o'. Dependency `revreg.c' is newer than dependent `D:¥TORNADO¥BETA¥SRC¥OBJ¥revreg.o'. Dependency `D:¥TORNADO¥BETA¥SRC¥H/qpsxstd.h' is newer than dependent `D:¥TORNADO¥BETA¥SRC¥OBJ¥revreg.o'. No commands for `D:¥TORNADO¥BETA¥SRC¥OBJ¥revreg.o' and no dependencies actually changed. No need to remake target `D:¥TORNADO¥BETA¥SRC¥OBJ¥revreg.o'. make: Nothing to be done for `D:¥TORNADO¥BETA¥SRC¥OBJ¥revreg.o'. From CJONES@digsys1.bwi.wec.com Mon Feb 5 07:19:32 1996 From: "CHRIS JONES" Date: Mon Feb 5 07:20:16 PST 1996 Subject: Compiling and MAKE files I am having some startup problems regarding compiling C++ code and make files. If anyone has some sample make files you could send me or some advice regarding what commands to use in the Projects menu, it would be greatly appreciated. I am currently reading the documentation regarding compiling and use of make files. As you well know, there is a lot of info to sift through. I could use a few friendly push in the right direction. Please respond to my email so as not to flood the exploder Thanks, Chris Jones cjones@digsys1.bwi.wec.com Electrical Engineer Westinghouse Electronic Systems (410) 765-7959, (410) 993-8822 (FAX) From CJONES@digsys1.bwi.wec.com Mon Feb 5 07:22:47 1996 From: "CHRIS JONES" Date: Mon Feb 5 07:23:29 PST 1996 Subject: Compiling and make files - addition Sorry for the double post, But I am using Tornado for the PC platform. Chris Jones cjones@digsys1.bwi.wec.com Electrical Engineer Westinghouse Electronic Systems (410) 765-7959, (410) 993-8822 (FAX) From ekins%sifvsj.dnet.@sifd01.solartron.com Mon Feb 5 07:46:33 1996 From: ekins%sifvsj.dnet.@sifd01.solartron.com (Phil Ekins - Solartron R&D - Farnborough UK) Date: Mon Feb 5 07:47:31 PST 1996 Subject: Result of Modicon Modbus driver request. I placed a request for a Modicon Modbus driver about a week ago: > >I am after a VxWorks driver for: > > Modicom Modbus > >I am interested in "free" and bought/supported versions. > >The driver would be required to run on a MVME-167 to drive one of the serial >ports (either 3 or 4), running under VxWorks 5.1 (currently). > >Thanks (for any information), > > Phil. > >P.S. I will report back to exploder the result(s) of this request. > Here are the result(s) as promised: I received a much appreciated response from Phil Wilshire (which I am following up): >I am working on one using a terminal server via ethernet. >Will take about another month to finish due to other work. >Email me if you are interested. >This is for the ALPHA AXPvme platform but it should port ok. > Phil Wilshire This was the only response, so if anyone has anymore information I'd like to hear from you. Cheers, Phil. ******************************************************************* * Phil Ekins, Internet: pekins@solartron.com * * Solartron, Tel: (44) 1252 376666 Ext.2287 * * Victoria Road, Fax: (44) 1252 543854 * * Farnborough, GU14 7PW, * * England. * ******************************************************************* From zot!Schilling.com!stevec@netcom.com Mon Feb 5 09:22:35 1996 From: stevec@schilling.com (Steve Cohan) Date: Mon Feb 5 09:22:38 PST 1996 Subject: Re: Access Fault Due to popular demand, here are the previous postings regarding access faults that I referred to earlier (guess in retrospect it was worth the bandwidth). Happy hunting, Steve Cohan ............ Following are excerpts from previous responses: ----------------------------------------------------------------------------- It sounds like a timeout trying to get access to the VMEbus. If you are using a card that supports fairness (such as the MVME-167), then you will want the following setup: - make sure that all of your 167 processor cards are requesting on the same VMEbus level. In our case, we have chosen level 2. And we put peripherals at level 1. - set up the bus requester to be fair and to release on request. Fairness basicly says not to request the bus again until nobody else is requesting the bus. Thus, each board takes its turn and then waits. *VMECHIP2_DMACR1 = (DMACR1_SINK_DATA | DMACR1_LVFAIR | DMACR1_LVROR | DMACR1_LVREQ_L2); (note our cache is copyback so we setup snooping to sink data) - Also if you are using a 167, note the timeouts for accessing the bus: /*********************************************************/ /*********************************************************/ /* @prb 4/26/93 WARNING */ /* */ /* The global VMEbus timeout VGTO must be set to 256Us */ /* because the ethernet chip (in un-throtled mode) can */ /* hog the local bus for ‾80Us. If another board tries */ /* to access this board while the ethernet chip is */ /* hogging the bus and the VGTO is less than 80Us */ /* the board would timeout and get a bus error */ /* Similarly, the VGTO must be greater than the LBTO */ /* or a timeout could occur if the local CPU is probing */ /* to see if the VSBchip is present (i.e. am I an MV166) */ /* So I recommend: VGTO_256US, VATO_32MS, LBTO_8US */ /*********************************************************/ /*********************************************************/ *VMECHIP2_TIMEOUTCR = (TIMEOUTCR_VGTO_256US | TIMEOUTCR_VATO_32MS | TIMEOUTCR_LBTO_8US | TIMEOUTCR_WDTO_32MS | ((256 - cpuSpeed) & 0xff)); --------------------------------------------------------------------------- Are all your processors on different bus grant lines? They should be. The daisy chain on BG3 (the default on most cards) won't work reliably for intensive bus activity (like spinning on a TAS). --------------------------------------------------------------------------- Both Stan Schneider's (use different bus grant levels) and Paul "Big Sig" :) Bade's (turn on fairness) suggestions are good things to try. I would only add to Stan's comment that you should make sure to switch the system controller to round robin if you put them on different levels. Recollection is that VxWorks tends to set things up with priority arbitration by default. --------------------------------------------------------------------------- I need more Information about your configuration. But in general: Most VMEbus interfaces are build in a way, which generates a deadlock sitiation when accessing the VMEbus in master and slave mode as well. This situation is a deadlock, since both the local and the VMEbus are granted to the requesters. The most used way is to force a local retry. (Usually done on simple Master/Slave-deadlocks. But when a readModifyWrite cycle has it's write cycle, no retry can be successful other than forcing a so called soft-bus-error. This buserror should force the software to retry the RMW-cycle. This behavior should be implemented within your BSP!!! On Opcode-prefetching CPU's like th 68040 or 68060 there has to be a 'nop' prior to the 'tas' instruction to invalidate the prefetch pipeline of the CPU! ---------------------------------------------------------------------------- ............ From jromine@btsslc.com Mon Feb 5 12:26:52 1996 From: JEFF ROMINE Date: Mon Feb 5 12:26:55 PST 1996 Subject: RPC and VxWorks I would like to hear from anyone who has used RPC under VxWorks, especially if they've overcome some difficulties. I'm using VxWorks 5.2 on the M68360. If anyone thinks they can help, I'd be glad to send more details etc. Thank You. -Jeff Romine, Broadcast Television Systems, jromine@btsslc.com From Sofin.Raskin@ecitele.com Mon Feb 5 13:00:21 1996 From: Sofin Raskin (SDH-S) Date: Mon Feb 5 13:00:25 PST 1996 Subject: unsubscribe unsubscribe From daemon@csg.lbl.gov Tue Feb 6 04:00:29 1996 From: daemon@csg.lbl.gov Date: Tue Feb 6 04:00:33 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Tue Feb 6 04:00:24 PST 1996 Subject: Driver for VMIC quad serial board (68564)? Subject: CDROM for VxWorks ??? Subject: VxWorks Applications Engineer AUSTRALIA ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Driver for VMIC quad serial board (68564)? Date: 6 Feb 1996 00:01:42 GMT From: alan@elroy.jpl.nasa.gov (Alan S. Mazer) Organization: Image Analysis Systems Group, JPL Message-ID: <4f65p6$87f@elroy.jpl.nasa.gov> I need a VxWorks driver for a VMIC 6015 quad serial I/O board based on the 68564. Anyone know where I can get one of these beasts? If I have to write it, it will be the first driver I've written and I'm not looking forward to debugging it :-). Thanks. - -- - -- Alan # Mountain Dew and doughnuts... alan@elroy.jpl.nasa.gov # because breakfast is the most important meal # of the day. --------------------------- Newsgroups: comp.os.vxworks Subject: CDROM for VxWorks ??? Date: Tue, 06 Feb 1996 09:36:12 -0800 From: Ray Peng Organization: Systems & Technonlogy Corp. Message-ID: <3117918C.41C6@systech.hinet.net> Hi, Does anyone out there know any brands of CDROM which can be used for VxWorks ? We are looking for CDROM drive for the 68K target running VxWorks. The CDROM device driver which could be executed on 5.1 or 5.2 VxWorks version is also needed. Any kinds of infomation, products resource will greatly helpful! Thanks in adviced. - -- -= Ray Peng =- Systems & Technology Corp. Taipei, Taiwan R.O.C. Voice: 886-2-6981599 Fax : 886-2-6981211 e-mail: peng@systech.hinet.net --------------------------- Newsgroups: comp.os.vxworks Subject: VxWorks Applications Engineer AUSTRALIA Date: Tue, 06 Feb 96 12:22:32 PDT From: wb@mkc.com.au Organization: AUSNet Services pty. ltd. Message-ID: Mayer Krieg & Co., the distributor Force Computers and Windriver Systems , leaders in embedded industrial computer hard- and software, seeks to appoint one Applications Engineer Operating System Software - - Adelaide based - for technical support on VxWorks and Solaris based industrial computer systems. This includes pre sales demonstrations (locally and interstate) and post sales technical backup mainly through phone, fax and email. Eperience in one of the areas VxWorks , Solaris, VME would be advantageous. Candidates should have a degree in electronic engineering or equivalent. A competitive salary package and exciting growth potential will reflect the importance of this position. Please forward your written applications to Mayer Krieg & Co., 50 Mary Street, Unley S.A. 5061, Walter Breunig : wb@mkc.com.au Tel. 08 291 3222, Fax 08 291 3200 --------------------------- End of New-News digest ********************** From daemon@csg.lbl.gov Wed Feb 7 04:00:41 1996 From: daemon@csg.lbl.gov Date: Wed Feb 7 04:00:44 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Wed Feb 7 04:00:36 PST 1996 Subject: Re: BOOT_PARAMS Subject: Re: re: Question on filesystem performance Subject: Re: RPC and VxWorks Subject: Re: VXWORKS Software contractor(s) needed Subject: GNAT target for VxWorks/68040 Subject: GNAT target for VxWorks/68040 Subject: GNAT target for VxWorks/68040 Subject: GNAT target for VxWorks/68040 Subject: Re: VERY SORRY for multiple copies Subject: Re: re: select on host system file Subject: Re: problem sith socket over slip ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: BOOT_PARAMS Date: 2 Feb 1996 19:23:54 GMT From: Don Small Organization: Sandia National Laboratories Message-ID: <4etoca$7a3@news.sandia.gov> References: <4eqqi7$6l6@ssbunews.ih.att.com> idawson@mlsma.mlm.att.com (nsi3520000-Dawson) wrote: >Hi Folks > >I want to update the boot parameters from my application >software. Is there a way to do this ?. > Yes, there is. >After using bootParamsShow and bootStringToStruct updating >the value using bootStructToString how can I write this >new boot string back to nvram ?. Is there a function which >will accept the boot parameter string pre-pended with a $ >and write the new boot parameters ?. > >Alternatively, is there a way to obtain the address of the >boot parameters in nvram and use some of the above as described - >finally writing them to nvram, the BOOT_PARAMS is a fixed >size structure. First, look at bootConfig in the vw/config/all directory. There function called bootCmdLoop is a very good example of how to Wind River modifies the boot parameters and then writes them to the NvRam. Specifically to this problem is the code for changing the boot parameters. switch() { .... case 'c': /* change boot params */ bootParamsPrompt (BOOT_LINE_ADRS); (void) sysNvRamSet (BOOT_LINE_ADRS, strlen (BOOT_LINE_ADRS) + 1, 0); break; .... } This function uses sysNvRamSet() to write the boot parameters to NvRam. I would recommend just using sysNvRamSet() for your problem. If you want to know more about how this function works, do the following. Look at sysLib.c in config/your_board. Search for a description of the chip drivers included for your board. Ours is the MV167 which includes nvRam.c. The memory drivers can be found at src/drv/mem. nvRam.c contains the code for sysNvRamSet() for the mv167 processor. In this file you find that the actual address of your boards NvRam is defined by NV_RAM_ADRS. You will find the definition of most board related addresses in config/your_board/config.h or your_board.h. Hope this helps. Don --------------------------- Newsgroups: comp.os.vxworks Subject: Re: re: Question on filesystem performance Date: 6 Feb 1996 15:16:46 GMT From: Mike Campbell Organization: BTG, Inc. Message-ID: <4f7rcu$r58@server.btg.com> References: Flemming_Kevin/macom_pbl_ctc@mailback.macom.com wrote: >Item Subject: Message text >>We have found that after creating a large number of >>files (several thousands) in one directory, the performance >>of filesystem operations (directory listing, file >>creation, file deletion, etc) drops to an extremely low >>level (5 - 10 seconds per operation). >> >>Does anyone know if there is an operating system-level solution >>or workaround for this problem? > [snip] >Possible solutions: If possible, use smaller directories. Delete >directories when possible. It was rumored last year that WindRiver >would be putting in a command to allow pre-allocation of a directory. They were supposed to release it last quarter, but I guess supporting the Tornado release go in thier way. >A better long-term solution is for WindRiver or a third party to provide >a utility analogous to the Norton Utilities for DOS, which among other >things can clean-up directories. > For clean up stuff see: Leonid Rosenboim Phone: +972-3-559-8144 R S T Software Industries Ltd. Mobile: +972-50-307-142 P.O.Box 11502, AZUR 58017, Israel Fax: +972-3-559-8244 ftp://knight.trendline.co.il/pub/rst E-Mail: leonid@rst.co.il They have some great looking stuff for vxWorks DOS disk maintenance. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: RPC and VxWorks Date: 6 Feb 1996 17:26:34 GMT From: Rendell Fong Organization: TRW-SIG Message-ID: <4f830a$lj6@gatekeeper2.svl.trw.com> References: <9602052025.AA16269@lbl.gov> I've successfully used RPC for VxWorks 5.2 on the PowerPC603 and an R3000. I didn't encounter any difficulties except for forgetting to invoke rpcTaskInit() for one of my tasks which performed the RPC. - -Rendell Fong, TRW-SIG --------------------------- Newsgroups: ba.jobs.contract,comp.os.vxworks Subject: Re: VXWORKS Software contractor(s) needed Date: Tue, 06 Feb 1996 23:42:37 GMT From: bbogart@well.com () Organization: The Whole Earth 'Lectronic Link, Sausalito, CA Message-ID: <4f8opo$kr@nkosi.well.com> References: <4er58l$jho@nkosi.well.com> We have an exciting new project integrating several VME computer systems into a semiconductor processing tool. We need one or more experienced VxWorks software engineers for 6 - 12 month efforts. If you have at least 2 years of relevant experience with VxWorks, at least 5 years of experience in software engineering, and you are immediately available, forward a resume to bbogart@well.com or bob.bogart@agai.com --------------------------- Newsgroups: comp.lang.ada,comp.os.vxworks,comp.realtime Subject: GNAT target for VxWorks/68040 Date: Tue, 06 Feb 1996 17:32:29 +0000 From: Keith Hyman Organization: Loral Space and Range Systems Message-ID: <311790AD.2700@srs.loral.com> Reply-To: hyman_keith@srs.loral.com Has anyone yet developed, or is anyone developing, a GNAT (GNU/NYU Ada) cross-compiler for VxWorks target on m68040 from a SunSolaris 2.4 system? Any information or suggestions would be appreciated. I have read various portions of the GNU documentation and have attempted some installations, but am as yet unsure as to how much actually needs to be done. As it turns out, VxWorks comes with GNU C/C++ cross-compiler for a 68k target already. I do realize that Ada tasking will not be supported... Please post or e-mail responses to: Evenings (PST): khyman@hooked.net or Days (PST): hyman@scf28.scf.loral.com Thanks, Keith Hyman --------------------------- Newsgroups: comp.lang.ada,comp.os.vxworks,comp.realtime Subject: GNAT target for VxWorks/68040 Date: Tue, 06 Feb 1996 17:32:45 +0000 From: Keith Hyman Organization: Loral Space and Range Systems Message-ID: <311790BD.1B3E@srs.loral.com> Reply-To: hyman_keith@srs.loral.com Has anyone yet developed, or is anyone developing, a GNAT (GNU/NYU Ada) cross-compiler for VxWorks target on m68040 from a SunSolaris 2.4 system? Any information or suggestions would be appreciated. I have read various portions of the GNU documentation and have attempted some installations, but am as yet unsure as to how much actually needs to be done. As it turns out, VxWorks comes with GNU C/C++ cross-compiler for a 68k target already. I do realize that Ada tasking will not be supported... Please post or e-mail responses to: Evenings (PST): khyman@hooked.net or Days (PST): hyman@scf28.scf.loral.com Thanks, Keith Hyman --------------------------- Newsgroups: comp.lang.ada,comp.os.vxworks,comp.realtime Subject: GNAT target for VxWorks/68040 Date: Tue, 06 Feb 1996 17:34:06 +0000 From: Keith Hyman Organization: Loral Space and Range Systems Message-ID: <3117910E.5D40@srs.loral.com> Reply-To: hyman_keith@srs.loral.com Has anyone yet developed, or is anyone developing, a GNAT (GNU/NYU Ada) cross-compiler for VxWorks target on m68040 from a SunSolaris 2.4 system? Any information or suggestions would be appreciated. I have read various portions of the GNU documentation and have attempted some installations, but am as yet unsure as to how much actually needs to be done. As it turns out, VxWorks comes with GNU C/C++ cross-compiler for a 68k target already. I do realize that Ada tasking will not be supported... Please post or e-mail responses to: Evenings (PST): khyman@hooked.net or Days (PST): hyman@scf28.scf.loral.com Thanks, Keith Hyman --------------------------- Newsgroups: comp.lang.ada,comp.os.vxworks,comp.realtime Subject: GNAT target for VxWorks/68040 Date: Tue, 06 Feb 1996 17:34:16 +0000 From: Keith Hyman Organization: Loral Space and Range Systems Message-ID: <31179118.5553@srs.loral.com> Reply-To: hyman_keith@srs.loral.com Has anyone yet developed, or is anyone developing, a GNAT (GNU/NYU Ada) cross-compiler for VxWorks target on m68040 from a SunSolaris 2.4 system? Any information or suggestions would be appreciated. I have read various portions of the GNU documentation and have attempted some installations, but am as yet unsure as to how much actually needs to be done. As it turns out, VxWorks comes with GNU C/C++ cross-compiler for a 68k target already. I do realize that Ada tasking will not be supported... Please post or e-mail responses to: Evenings (PST): khyman@hooked.net or Days (PST): hyman@scf28.scf.loral.com Thanks, Keith Hyman --------------------------- Newsgroups: comp.lang.ada,comp.os.vxworks,comp.realtime Subject: Re: VERY SORRY for multiple copies Date: Tue, 06 Feb 1996 17:38:51 +0000 From: Keith Hyman Organization: Loral Space and Range Systems Message-ID: <3117922B.B90@srs.loral.com> References: <31179118.5553@srs.loral.com> Reply-To: hyman_keith@srs.loral.com Sorry for the multiple copies --- KPH --------------------------- Newsgroups: comp.os.vxworks Subject: Re: re: select on host system file Date: 5 Feb 1996 15:48:56 GMT From: chaos@linux.peacefulstar.oakland.ca.us (H.J. Bae) Organization: Peaceful Star, Oakland, CA Message-ID: <4f58t8$k4o@idiom.com> References: <4e69lh$dnj@cebaf4.cebaf.gov> <4eb9g3$8ro@cebaf4.cebaf.gov> Reply-To: hjb@peacefulstar.oakland.ca.us On 26 Jan 1996 19:19:31 GMT, Carl Timmer wrote: : : However, when waiting on this file, the select statement always returns the : error, "S_netDrv_UNKNOWN_REQUEST" which I assume means that the network file : i/o driver is failing to find any information on the file of interest. VxWorks select() emulation is done via FIOSELECT ioctl to various I/O devices. The only ones I know that support FIOSELECT are sockets, ttys, and pipes. - -- H.J. Bae, Consultant hjb@peacefulstar.oakland.ca.us http://peacefulstar.oakland.ca.us/ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: problem sith socket over slip Date: 5 Feb 1996 15:53:01 GMT From: chaos@linux.peacefulstar.oakland.ca.us (H.J. Bae) Organization: Peaceful Star, Oakland, CA Message-ID: <4f594t$k4o@idiom.com> References: <9602021635.AA09314@lepy07> Reply-To: hjb@peacefulstar.oakland.ca.us On Fri, 2 Feb 96 16:35:10 GMT, "Matthew Pardoe" wrote: : : exactly what affect these settings should have, can anyone suggest what : socket options I need to set to ensure failed socket connections are cleared : down as soon as possible. Turn off SO_LINGER (set linger time to 0) and be sure to shutdown() and close() the socket. That should do it. - -- H.J. Bae, Consultant hjb@peacefulstar.oakland.ca.us http://peacefulstar.oakland.ca.us/ --------------------------- End of New-News digest ********************** From prb@aplexus.jhuapl.edu Wed Feb 7 07:46:44 1996 From: prb@aplexus.jhuapl.edu (Paul R. Bade) Date: Wed Feb 7 07:46:47 PST 1996 Subject: FAIRNESS Arbitration on PowerPC boards Hi, I am conducting an informal Poll to see how many people are planning to use a PowerPC board from vendors such as Motorola, VI Computer, Themis, Force, ... And of those people, how many people are planning to use FAIRNESS (i.e. request on no request) as the arbitration mode. For those unfamiliar with FAIRNESS, here is why one may chose to use it ... When one has several boards in a VME chassis competing for the bus at the same bus request level, the boards closer to slot 1 have a higher priority in obtaining the VMEbus than those further away. This occurs because the bus grant signals are daisy chained. What can happen in this case, is that the boards furthest from slot 1 can get starved. This problem is worse when the number of boards increases. To avoid this problem, some board implement a FAIRNESS mode. With the FAIRNESS mode, when a board requests the bus and makes a transfer, the board will not request the bus again until no other board is requesting the bus (i.e. the request line is not pulled low). This allows each board to take its turn. Thanks, +====================================================================+ | __ ____ __ __ | | /¥ ¥ / ¥ ¥ /¥_¥__ /¥ ¥ Johns Hopkins University | | / ¥_¥ | /¥ ¥ ¥ / / /¥_¥¥ ¥ ¥ Applied Physics Lab. | | / /¥ | |¥ ¥/ ¥ ¥ / / / / / ¥ ¥ ¥ | | / ¥/ |_| ¥ /¥ ¥_¥ / / / / / ¥ ¥ ¥ Paul R. Bade | | / /¥__/_/ ¥ ¥ ¥/_// / / / / / / / (301)-953-6000 x8681 | | / / / ¥ ¥_¥ ¥ ¥/ / / / / / prb@aplexus.jhuapl.edu | | ¥/_/ ¥/_/ ¥__/_/ ¥/_/ | | __ ____ __ __ | | /¥ ¥ / ¥ ¥ /¥ ¥ /¥ ¥ | | / ¥_¥ | /¥ ¥ ¥ / ¥_¥ / ¥ ¥ | | / /¥ | |¥ ¥/ ¥ ¥ / /¥ |_| / /¥ ¥ ¥ | | / ¥/ |_| ¥ /¥ ¥_¥ / / / | | ¥/ / ¥ ¥ | | / /¥ / / ¥ ¥ ¥/_// / / / / / /¥ ¥_¥ | | ¥ ¥/ / / ¥ ¥_¥ ¥ ¥/ / / ¥/ / / / | | ¥__/_/ ¥/_/ ¥__/_/ /_/_/ | | | +====================================================================+ From cjones@digsys1.bwi.wec.com Wed Feb 7 17:04:55 1996 From: Chris Jones Date: Wed Feb 7 17:04:58 PST 1996 Subject: Re: Compiling and MAKE files Wang, Tom (HT-MS) wrote: > > Hi, > > The experience that I have is on the ObjectCenter's C++ compiler and WindC++ > gateway > for the project on the vxWorks 5.11 & 5.2. > > A couple of points for the developments with the C++: > > 1. The taskSpawn() need a parameter for the function_entry_point in "C", > The C function can be implemented to create the C++ object and switch to > OOP. Does this mean that one cannot use a member function of a class to spawn a new task? Forgive my ignorance, but I currently have successfully compiled a generic task class and a specialty class derived from this generic task. What are the steps necessary to load these object modules to my target being that they are C++ modules. In a related matter, I envision our current system as a group of tasks each defined in a seperate class with the appropriate member functions. These tasks will all communicate via message queues and some message passing functions which are included in the generic parent class. I also plan to use a single "creator" class/task which will spawn all other tasks, create/initialize system semaphores and then delete itself after starting the rest of the system. Does this sound acceptable? Is it too simplistic, that is are there any pitfalls out there that I don't know about? I would appreciate input from those who have used C++ with VxWorks. Another question while I'm here. When using multiple processors with VxMP, does each processor have it's own code image containing only those functions relevant to that processors assigned tasks, or does each board contain a complete image of the entire code with different system 'entry' functions? That is, does board one call StartupFunc1 which sets a series of events into motion and board two calls StartupFunc2 which causes a different set of events? Thanks Again Chris Jones cjones@digsys1.bwi.wec.com From daemon@csg.lbl.gov Thu Feb 8 04:00:49 1996 From: daemon@csg.lbl.gov Date: Thu Feb 8 04:00:52 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Thu Feb 8 04:00:43 PST 1996 Subject: Re: Setting up Multiple MVME167 boards to function as VMEbus Masters/Slaves Subject: LonWorks driver Subject: Re: VxWorks Software Architecture Subject: TCP/IP/PPP 3rd Party Software Subject: Re: DejaGNU Subject: Re: RPC and VxWorks Subject: Help My Ballistic Coworker Subject: How does sysPhysMemDesc work? Subject: Re: BOOT_PARAMS ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Setting up Multiple MVME167 boards to function as VMEbus Masters/Slaves Date: Mon, 5 Feb 1996 11:15:30 From: hanson@pat.mdc.com (GeoControl Systems, Inc.) Organization: GeoControl Systems, Inc. Message-ID: References: <4ema75$kab@newsbf02.news.aol.com> In article <4ema75$kab@newsbf02.news.aol.com> thesharps@aol.com (THE SHARPS) writes: >From: thesharps@aol.com (THE SHARPS) >Subject: Setting up Multiple MVME167 boards to function as VMEbus Masters/Slaves >Date: 30 Jan 1996 18:39:17 -0500 >I would like to communicate with anyone who already has experience >modifying the VxWorks kernel to map multiple MVME167 boards into the VME >global address space as both masters and slaves. I have read the Motorola >Programmer's Reference Manual section on the VMEChip2, but would like to >establish contacts to use as sounding boards for questions that may arise >as I progress. >Tim Sharp >Member Of The Technical Staff >Hughes Aircraft Co. >tsharp@ccgate.hac.com Dear Tim, Our project team did just that about 3 years ago, then (finally) Wind River released VxMP which makes the job quite simple. Why don't you check into that product? It allows you to set up a name database for global variables (shared across the VME backplane), etc. If you want, I can tell you more about our work also. Besides setting up the global memory, we developed an automated ethernet (RPC) connection to a GUI server on a SparcStation. Send me an e-mail Dave Hanson Computer & Systems Engineer GeoControl Systems, Inc 1720 NASA Road 1, Suite 100 Houston, TX 77062 (713) 333-2561 hanson@pat.mdc.com --------------------------- Newsgroups: comp.os.vxworks Subject: LonWorks driver Date: Tue, 06 Feb 1996 02:05:05 GMT From: datalogic@bo.nettuno.it (Marco Bottazzi) Organization: Cineca Message-ID: <4f5cr5$ksu@sirio.cineca.it> Hello, Where can I get a driver (done) for the LonWorks field bud under VxWorks? Marco Bottazzi --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks Software Architecture Date: Fri, 2 Feb 1996 04:05:04 GMT From: Roy Rowan Organization: NETCOM On-line Communication Services (408 261-4700 guest) Message-ID: <31118D70.21CC2C3@netcom.com> References: <310D60EF.13C8@digsys1.bwi.wec.com> Sender: royr@netcom11.netcom.com Chris Jones wrote: > > I am relatively new to real time programming. > > I have seen a method used in the VRTX O/S to "kick off" the tasks in a > system. A high level executive task initializes every one else and waits > for them to report back via a message that they have completed > initialization and are pending on an event flag. The executive task then > posts this "kick off" event flag and goes to sleep forever. > > My question: Since VxWorks does not have event flags, is there a similar > method for starting up a system simultaneously? > > Are there any references for "good" programming techniques with VxWorks? > > Thanks > Chris Jones > Westinghouse Electronic Systems > cjones@digsys1.bwi.wec.com Chris, All the posts that I have seen to date are very good and will do what you wish. I did things slightly different. I have the task that does the initial startup create an empty semaphore. All other tasks then call a simple common subroutine that takes then gives the semaphore. The difference is that you have the control when you creat the semaphore to make priority based or fifo based, this gives you control on the ordering in which your task start up. Since the semaphore was created in our root task, this is before any static construction if you are using C++ and want to create tasks in the constructor. best of luck roy --------------------------- Newsgroups: comp.os.vxworks Subject: TCP/IP/PPP 3rd Party Software Date: 5 Feb 1996 19:01:17 GMT From: Cecil Taylor Organization: Bell Northern Research Message-ID: <4f5k5t$gpf@crchh327.rich.bnr.ca> I know of Wind River's protocol stack. Is there any freeware that implements TCP/IP/PPP under VxWorks? Where can I find it? Also, what kind of interprocess communications take place between the TCP/IP/PPP components of Wind River's stack? Finally, can you place a custom device driver at layer 1 underneath Wind River's PPP? Any help you can give to this VxWorks novice is much appreciated! --------------------------- Newsgroups: comp.os.vxworks,comp.software.testing Subject: Re: DejaGNU Date: 7 Feb 1996 11:13:55 -0800 From: rob@cygnus.com (Rob Savoye) Organization: Cygnus Support Message-ID: References: <4fagm1$ko6@superb.csc.ti.com> gbruce@ti.com (Gene Bruce) writes: >I have recently discovered DejaGNU and would like to evaluate it for >potential use. >The problem is while the distribution comes with a good reference, >there is no tutorial information nor any good hold-your-hand examples. >Does any material of this nature exist? And where? Sorry, I never got around to writing a decent tutorial. There is a short "calc" example & tutorial in the manual. The best examples are to look at the testsuite (in the sources) for Tcl, expect, and runtest. You can also look at the GDB and G++ testsuites as well. - rob - (btw - the new release of DejaGnu 1.3 should be out by next week) ‾ - -- Cygnus Support << KERNEL: Panic, core dumped >> Headquarters PO Box 1006 Darkstar crashes, 1937 Landings Drive Nederland, CO 80466 pouring its light into ashes, Mountain View, CA 94043 +1 (303) 258-0506 reason tatters, ... +1 (415) 903-1400 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: RPC and VxWorks Date: 8 Feb 1996 03:17:05 GMT From: Tom Burns Organization: GHG Corporation Message-ID: <4fbpvh$1em@pan.ghgcorp.com> References: <9602052025.AA16269@lbl.gov> JEFF ROMINE wrote: > >I would like to hear from anyone who has used RPC under VxWorks, >especially if they've overcome some difficulties. I'm using VxWorks 5.2 >on the M68360. If anyone thinks they can help, I'd be glad to send more >details etc. Thank You. > I have used RPC (and rpcgen) on a project using VxWorks 5.1 running on a MVME167 (68040). I stuck close to the RPC model (make a remote procedure call, then block until it returns). If you are trying to do something fancy like broadcasting I probably wouldn't be much help. So, I would need more details. By the way, RPC is the basis for several of the VxWorks network features including NFS. So, if NFS is working on your target then so is RPC. --------------------------- Newsgroups: comp.os.vxworks Subject: Help My Ballistic Coworker Date: 6 Feb 1996 16:30:10 GMT From: smalwood@tetons.den.mmc.com (Scott K. Smallwood) Organization: Martin Marietta Keywords: Ballistic Coworker Message-ID: <4f7vmi$a5o@tel.den.mmc.com> Help! My poor coworker is having a mild coronary over the loss of control with this, his first experience with VxWorks. He comes from the school of "I could really do every damn thing I want to using a 68000 and the debugger I had on that little microprocessor kit in college". He was suspcious when he first used C on an embedded program. He was downright rude when he first hacked around in IRIX on a real workstation in a real-time environment. He can't stand automatic transmissions in his vehicles. Now he is having a full-blown O.S. shoved down his throat for a VME project with a short fuse. In short he is a major control freak and is driving all of us nuts. Please help before we kill him. Here is our simple problem: The first step of this project is to simply get the memory maps setup. Of course our board without VxWorks just sets aside large chunks of address space that corolate to VME addresses in various flavors of address modifiers. The real rub is that in our applications we may not know upfront what VME address range we need to access. We find instructions to set up the system if addresses are fixed such as in the case of a controller base address, but what is the procedure to set this up at run time? I see a reference to building a address context and swapping it in, but has anyone dealt with this and what is the easiest way to pull it off? I would like to see my group eventually move to this architecture but I really need this guy to survive his first experience in a "high level" (low control) environment. If any of you have any job openings, please pass them onto me and I'll forward them to him, maybe that would be the best way to deal with this ugly situation. Sign me, "Desperate In Devner" - -- ____________________________________________________________ Scott Smallwood, (303) 977-6102, sks@den.mmc.com Lockheed Martin Real Time Simulation Lab, Denver, Colorado --------------------------- Newsgroups: comp.os.vxworks Subject: How does sysPhysMemDesc work? Date: Thu, 08 Feb 1996 10:32:19 +0100 From: harald.grundner@erno.de (Harald Grundner) Organization: DASA Message-ID: Dear VxGurus, What does the sysPhysMemDesc structure (located in memDesc.c) do in a system that has no Memory Management Unit? Does it do anything? I'm trying to create a BSP for an MMU-less SPARC 1 system, taking the BSP of the Force 2CE (which has an MMU) as a template. Can I throw away the whole sysPhysMemDesc thing? Or is there still some magic inside VxWorks using this information, even if there is no MMU? Unfortulately I don't have the sources ... Many thanks in advance for any insight you are able to give me. - -- Harald Grundner harald.grundner@erno.de --------------------------- Newsgroups: comp.os.vxworks Subject: Re: BOOT_PARAMS Date: 4 Feb 1996 16:43:02 GMT From: michaelv@qualcomm.com (Michael Vakulenko) Organization: Qualcomm Israel Message-ID: References: <4eqqi7$6l6@ssbunews.ih.att.com> In article <4eqqi7$6l6@ssbunews.ih.att.com>, idawson@mlsma.mlm.att.com (nsi3520000-Dawson) wrote: Ian, > Hi Folks > > I want to update the boot parameters from my application > software. Is there a way to do this ?. We are using MVME162 board and at least on this board it is possible. The boot parameters are stored in the NVRAM as a character string. On its startup VxWorks copies the boot string from NVRAM to RAM. There two constants defined in the configAll.h file, which specify the addresses of boot string. ../$VW/config/all/configAll.h: BOOT_LINE_ADRS - address of the boot string in RAM ../$VW/config/all/configAll.h: NV_BOOT_OFFSET - address of the boot line in NVRAM. Address of NVRAM is defined by the NV_RAM_ADRS in one of the BSP *.h files. For an example of how to deal with this stuff, see ../$VW/src/config/usrNetwork.c file. Just search the file for word "BOOT". > After using bootParamsShow and bootStringToStruct updating > the value using bootStructToString how can I write this > new boot string back to nvram ?. Is there a function which > will accept the boot parameter string pre-pended with a $ > and write the new boot parameters ?. Just write it to the appropriate address : /* ------------------------------ */ bootString = NV_RAM_ADRS + NV_BOOT_OFFS; params.bootDev [ 0] = 0; /* boot device code */ params.hostName [ 0] = 0; /* name of host */ params.targetName [ 0] = 0; /* name of target */ params.bad [ 0] = 0; /* backplane internet addr */ params.had [ 0] = 0; /* host internet addr */ params.bootFile [ 0] = 0; /* name of boot file */ params.startupScript [ 0] = 0; /* name of startup script file */ params.usr [ 0] = 0; /* user name */ params.passwd [ 0] = 0; /* password */ params.other [ 0] = 0; /* available for applications */ params.procNum =0; /* processor number */ params.flags =0; /* configuration flags */ bootStructToString( bootString, ¶ms); /*---------------------------------*/ > > Alternatively, is there a way to obtain the address of the > boot parameters in nvram and use some of the above as described - > finally writing them to nvram, the BOOT_PARAMS is a fixed > size structure. > See above. ___________________________________________________________ Michael Vakulenko, : Email: michaelv@qualcomm.com Software Engineer, : Voice: +972-4-8577999 QUALCOMM Israel, Ltd. : Fax: +972-4-8577998 ___________________________________________________________ --------------------------- End of New-News digest ********************** From twang@gic.gi.com Thu Feb 8 11:54:15 1996 From: "Wang, Tom (HT-MS)" Date: Thu Feb 8 11:54:18 PST 1996 Subject: Re: Compiling and MAKE files Chris Jones Wrote: > Does this mean that one cannot use a member function of a class to spawn a > new task? Forgive my ignorance, but I currently have successfully compiled a > generic task class and a specialty class derived from this generic task. What are the > steps necessary to load these object modules to my target being that they are C++ > modules. Yes, You can call the taskSpawn() in your class member function. But, The task entry_point has to be a "C" function, as far as I know. I assumed you are using the class approach, as I did, for your application. You can re-visit my response when you start implementing and loading your code. === I didn't post my response to your C++ development question, but, I think it will be better to post it here for the completeness. 1. The taskSpawn() need a parameter for the function_entry_point in "C", The C function can be implemented to create the C++ object and switch to OOP. 2. If you have any template class, then you have to do some template instantiation. ( if you are using the ObjectCenter C++ compiler and WindC++ gateway, P307, Reference Manual of the ObjectCenter on the template switch, e.g. -ptv -ptd) The sequence for make script is as followings: 1) $(CXX) $(CXXFLAGS) -c *.cpp 2) $(LD) -r -o OUT1.O *.o 3) $(CXX) -ptv -ptdtemp_list OUT1.O ;; Template instantiation 4) $(LD) -r -o OUT2.O *.o OUT1.O `cat temp_list` 5) $(NM) OUT2.O | $(MH) > OUT3.c ;; munching step 6) $(CC) $(CFLAGS) -c -o OUT4.O OUT3.c 7) $(LD) -r -o OUT5.O OUT4.O OUT2.O ;; OUT5.o is the final object Where: CXX = CC68k LD = ld68k NM = nm68k MH = munch68k CC = cc68k CXXFLAGS = flags for CXX CFLAGS = flags for CC 3. Before you load the objects into the target, You have to issue the "cplusXtorSet 1" without the ("). This will construct or destruct any static object during loading and unloading. === > In a related matter, I envision our current system as a group of tasks each > defined in a seperate class with the appropriate member functions. These > tasks will all communicate via message queues and some message passing functions > which are included in the generic parent class. I also plan to use a single "creator" > class/task which will spawn all other tasks, create/initialize system semaphores > and then delete itself after starting the rest of the system. > Does this sound acceptable? Is it too simplistic, that is are there any pitfalls > out there that I don't know about? It looks fine, from my experience on implementing similar design. We have developed a frame work for our application. This framework contains classes that encapsulates the vxWork's task and message queue with a thin wrapper, and provides a message engine to process the incoming message. The task creation design requires the "task entry_point" that you have to specify in the implementation. > Another question while I'm here. When using multiple processors with VxMP, > does each processor have it's own code image containing only those functions > relevant to that processors assigned tasks, or does each board contain a complete > image of the entire code with different system 'entry' functions? > That is, does board one call StartupFunc1 which sets a series of events into > motion and board two calls StartupFunc2 which causes a different set of > events? Every target will has its own image loaded. We don't load the other target's image. Why you load other target's image in ? Code sharing ? We partitioned the task into 3 phases, i.e. Startup, Normal Operation, and Close-down phase. Each target has a task creator that spawns all the tasks running in the target, and then use the shared_memory message_queue to exchange the message between different targets. We then use the messages through message engine to do the rest for the application. Best. -Tom Wang ===================================================================== My Email address: s.wang@ieee.org Phone: 215-773-8526 From mgarlich@me.umn.edu Thu Feb 8 13:41:46 1996 From: "Micah Garlich-Miller" Date: Thu Feb 8 13:41:49 PST 1996 Subject: unsubscribe unsubscribe mgarlich@me.umn.edu From CHCHEN@aol.com Thu Feb 8 17:40:46 1996 From: CHCHEN@aol.com Date: Thu Feb 8 17:40:48 PST 1996 Subject: Token Ring adapter driver I am looking for a Token Ring adapter driver for VxWorks. Is anybody know where I can get it from internet? Thanks, David Chen From stevem@syd.csa.com.au Thu Feb 8 19:15:35 1996 From: stevem@syd.csa.com.au (Steven McCoy) Date: Thu Feb 8 19:15:37 PST 1996 Subject: S_rpcLib_RPC_TIMEOUT We have mounted an NFS file system onto our VxWorks target. Under certain circumstances we flush the write cache to update the file immediately on the Sun server using the FIOSYNC ioctl(). Occasionally we get the following error, EVXWFAIL error raised, errno = 3080197, code = S_rpcLib_RPC_TIMEOUT EVXWFAIL exception raised in ASSIP_ERROR_LOG.IOLIB.IOCTL(IOLIB.FIOSYNC) EVXWFAIL exception raised in ASSIP_ERROR_LOG.FLUSH_CACHE The NFS driver is built on top of the Sun Microsystems' RPC facility. It appears that the RPC timeout expires presumably during times of high network activity? Can anyone tell me what the RPC timeout period is and where it is defined. Is it the nfsTimeoutSec/nfsTimeoutUSec values in VxWorks or is this something that can be tailored in SunOS? An nfsTimeoutSec of 25 seconds seems pretty long to me (there are no gateways involved although I have seen the error occur more frequently when there is one). Can the server really be that busy for that long not to service a request from the VxWorks client? ENV VME1>printf("nfsTimeoutSec = %d¥n", nfsTimeoutSec); nfsTimeoutSec = 25 ENV VME1>printf("nfsTimeoutUSec = %d¥n", nfsTimeoutUSec); nfsTimeoutUSec = 0 VADSworks version 2.0.3(b). VxWorks (for Motorola MVME167) version 5.0.2b. Kernel: WIND version 2.0. Thanks in advance, Steven McCoy CSC Australia A Unit of Computer Sciences Corporation From erxa@inrete.it Fri Feb 9 00:39:04 1996 From: erxa@inrete.it (Federico De Sario) Date: Fri Feb 9 00:39:07 PST 1996 Subject: RIP protocol under VxWorks I'm looking for a RIP protocol implementation under VxWorks. Is there anyone who has it available or has implemented it? Any help you can give at this subject will be much appreciated! Thx in advance. Federico De Sario ================================================================== _______ |E______| Engineering |__R____| Robotics & |____X__| Control Systems |______A| Applications ERXA S.r.l. E-mail: erxa@inrete.it via Morghen 22 Voice: +39-11-741.27.49 10143 Torino (ITALY) Fax: +39-11-741.20.44 ================================================================== From daemon@csg.lbl.gov Fri Feb 9 04:00:31 1996 From: daemon@csg.lbl.gov Date: Fri Feb 9 04:00:35 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Fri Feb 9 04:00:26 PST 1996 Subject: Mil 1553 boards with support under VxWorks ? Subject: Re: How does sysPhysMemDesc work? Subject: I am looking for a CPU, preferably written in Verilog Subject: Re: Setting up Multiple MVME167 boards to function as VMEbus Masters/Slaves Subject: Re: Help My Ballistic Coworker Subject: Re: TCP/IP/PPP 3rd Party Software ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Mil 1553 boards with support under VxWorks ? Date: Thu, 08 Feb 1996 18:10:44 GMT From: ctaylor@wd.estec.esa.nl Organization: European Space Agency Message-ID: <4fcemh$15ip@info.estec.esa.nl> We are looking for a dual channel 1553 VME interface board with Vxworks support. So far the boards we have found have been very expensive. Does anyone have experience in this area and can recommend a cheap but good board. //ct --------------------------- Newsgroups: comp.os.vxworks Subject: Re: How does sysPhysMemDesc work? Date: 9 Feb 1996 03:48:22 GMT From: Trung La Organization: Advanced Processing Laboratories, Inc. Message-ID: <4feg66$3c6@sulu.sd.aplabs.com> References: Harald, Can't guarantee it but you could try undefining "#define INCLUDE_MMU_BASIC" in configAll.h and config.h. Trung. --------------------------- Newsgroups: comp.os.vxworks Subject: I am looking for a CPU, preferably written in Verilog Date: 06 Feb 1996 08:46:31 GMT From: rj@eli.wariat.org (Robert J. Brown) Organization: APK - Internet Provider for Ohio. Message-ID: I am looking for a CPU, preferably written in Verilog, that is suitable to use as a timing and control sequencer component in a complex ASIC. It should operate with a minimun of a 56 MHz clock, and preferably be a 32-bit architecture. We have already looked at ARM and MIPS standard cells from silicon foundries, but these CPU standard cells have aparently only been used in foundry originated designs so far, and the documentation details are not sufficient for us to proceed with our design. Besides, with these CPUs, we are locked into a single vendor for the chip, and that may not be desirable. If you have, or know of, any such CPU design, please contact me via email or fax. Thank you. - -- - ----------- "... And the men went up and viewed Ai." [Jos 7:2] ----------- Robert Jay Brown III rj@eli.wariat.org http://eli.wariat.org 1 847 705-0370 Elijah Laboratories Inc; 759 Independence Drive; Suite 5; Palatine IL 60074 - ----- M o d e l i n g t h e M e t h o d s o f t h e M i n d ------ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Setting up Multiple MVME167 boards to function as VMEbus Masters/Slaves Date: 9 Feb 1996 04:35:15 GMT From: Trung La Organization: Advanced Processing Laboratories, Inc. Message-ID: <4feiu3$3c6@sulu.sd.aplabs.com> References: <4ema75$kab@newsbf02.news.aol.com> Dear Tim, Someone has already mentioned solutions such as RPC or using VxMP which will allow you to communicate between processors via global semaphores and global message queues. However, if you just want to know how to map each MV167's memory out to the VMEbus in an orderly manner (sounds like your original question) then the place to do it is the sysProcNumSet() routine in sysLib.c. By default, WRS only maps CPU0's memory out on to the VMEbus. However, you can modify sysProcNumSet to map out each CPU's memory based on the CPU number. If you want full blown multi-processors capability then use VxMP. Trung. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Help My Ballistic Coworker Date: 9 Feb 1996 04:03:53 GMT From: Trung La Organization: Advanced Processing Laboratories, Inc. Message-ID: <4feh39$3c6@sulu.sd.aplabs.com> References: <4f7vmi$a5o@tel.den.mmc.com> Dear Scott, If possible, I would try setting your VME boards in A16 or A24 space because you can map them in completely (64K and 16M respectively) in VxWorks. If not, I would shoot for a range of addresses in A32. You can't map in all 4GB but you don't need to be exact either. If the above doesn't do it then you might have to fork over money to buy the VxVMI optional package that will allow you to map virtual/physical address dynamically. By the way, I need to know your friend's name so we don't have to post a message like yours. Trung. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: TCP/IP/PPP 3rd Party Software Date: Fri, 9 Feb 1996 04:44:04 GMT From: dab@netcom.com (Don Brooks) Organization: NETCOM On-line Communication Services (408 261-4700 guest) Message-ID: References: <4f5k5t$gpf@crchh327.rich.bnr.ca> Sender: dab@netcom14.netcom.com Cecil Taylor (taylorc@bnr.ca) wrote: : I know of Wind River's protocol stack. Is there any freeware that implements : TCP/IP/PPP under VxWorks? Where can I find it? Also, what kind of interprocess : communications take place between the TCP/IP/PPP components of Wind River's : stack? Finally, can you place a custom device driver at layer 1 underneath : Wind River's PPP? : Any help you can give to this VxWorks novice is much appreciated! A public domain port of ppp-2.1.2 to VxWorks is available via anonymous ftp from ftp.netcom.com as /pub/da/dab/ppp-2.1.2.tar.gz. The file if_ppp.c con The source file if_ppp.c contains the interface to Wind River's IP stack. It may be possible to place your custom device driver underneath the PPP stack depending on how the driver is constructed since the PPP stack communicates with layer below it via read and write system calls. Modifications to the port and/or your driver may be necessary. Don Brooks dab@netcom.com --------------------------- End of New-News digest ********************** From bmontgomery@tsi-telsys.com Fri Feb 9 08:59:42 1996 From: jobs@tsi-telsys.com Date: Fri Feb 9 08:59:45 PST 1996 Subject: Columbia, MD. Realtime OS SW designer needed immediately ---------------------- B e y o n d L i m i t s ---------------------- Technology advances have become commonplace, but every now and then, a group of brilliant people get together and challenge yet another limit. At TSI TelSys, exceptional design engineers will find a dynamic, entrepreneurial environment and breakthrough technology. Our talented team develops and manufactures high-speed satellite & ground telecommunications equipment for scientific and commercial markets. We currently have Senior and Junior openings available in R&D and Applications Engineering for: VxWorks Real-time Software Design Engineers. All positions require a BS/MS in CS/EE/CE, and 2+ years of hands-on experience. In addition to the specifics listed, experience in packet-switched telecommunications, RF processing, or commercial SatCom products a big plus! Real-time Operating Systems Design - VxWorks BSP Design, PCMCIA drivers, pushing the frontiers in VxWorks setup and configuration! Real-time Application Software - C, VxWorks, Motorola 680x0 & PowerPC microprocessors, LAN, WAN, and Satellite communication protocols including ATM, FDDI, CCSDS. We offer talented visionaries an environment without limits and an attractive compensation package. Send or email your resume to: TSI TelSys, Inc. 7100 Columbia Gateway Drive Columbia, Maryland 21046 fax: (410) 872-3901 jobs@tsi-telsys.com http://www.tsi-telsys.com Principals only. No phone calls, please. Equal Opportunity Employer From lwatson@arcserv.dasd.honeywell.com Fri Feb 9 09:21:46 1996 From: lwatson@arcserv.dasd.honeywell.com (Linda Watson) Date: Fri Feb 9 09:21:49 PST 1996 Subject: SLIP, vxworks and a Sun workstation I'm looking for some advice on SLIP for a vxworks on an i960 and a Sun workstation. Background info: We have a custom i960 board with a custom ASIC controller and a serial interface (no VME) to our host, which is a Sun workstation running SunOS 4.1.3. Currently, we are able to bring up vxworks on the 960 with a home-hacked terminal interface on the Sun workstation which comes up with vxworks shell. We can run vxworks commands on the 960 or load custom modules into vxworks, but we need SLIP to load in user modules "on the fly". I think we're ok with SLIP on the 960 vxworks side, but I'm not completely sure about setting up SLIP on the Sun for communication with the 960, which I'm pretty sure is not the same as the configuration for dialup access. We are using the "unsupported" cslip package that came with vxworks, which seems to be geared toward users dialing in and out with modems, which we won't be doing. Questions: How would we need to setup the slip.hosts and the /etc/remote slip entries? We can either run at 9600 baud or 38K and we are using /dev/ttyb for our serial interface. slip.hosts file: from the example and a web source: login local-addr remote-addr mastk options for compress/proxy/def. "# Server installs proxy arp for client (and allows compression) Ssatan `hostname` satan.lbl.gov 0xffffff00 auto-comp proxy-arp # Client installs default route towards server (and asserts compression) Shell `hostname` hell.lbl.gov 0xffffff00 compress default" and "Shelmerg dc-slip sl-helmerg 0xfffffc00 autocomp" o Do we have to enter all the users or a specific user for the login entry, or is there a default to allow anyone to login to vxworks on the 960. It would seem to me that this part should be handled by vxworks on the 960 side, and it would act as the host with the sun user logging into it. For that matter, do we need to have a SLIP group on the Sun side? o If we have to enter user names, should they be the same as names on the Sun accounts, or different names, or does it matter? o For the options, would we need to use a subnet TCP/IP, proxy ARP, or leave it blank? o We assigned a new IP address to the 960. /etc/remote: From the example "dial19200|Telebit attributes:¥ :dv=/dev/cua0:br#19200:nt:fc:at=telebit:du: UNIX|telebit|Telebit dial-out to another Unix system:¥ :el=^U^C^R^O^D^S^Q@:ie=#%$:oe=^D:tc=dial19200: dial-foo:¥ :pn=dt5551212:tc=telebit: foo|slip-foo:¥ :st=slip:ls=/etc/login.script.unix S%h {passwd}:¥ :cc=/etc/sliplogin Sfoo:tc=dial-foo:" o From the example, I'm not sure which items in the file are required for SLIP, and which are supposed to be specific to the user configuration. o Which are specific to a dialup modem interface versus what we want to do, or does it matter? o I'm guessing we tack on the SLIP and tip definitions with the other interfaces for ttyb, but it is not clear since the cslip instructions also mention setting up the ttys to dial in/out, moving the current ttya to ttyd0, creating /dev/cua0 Would this be what we would need, or would this affect other applications using ttyb (like our load and boot program)? o Does an /etc/remote with these lines look close to what we need: cua1:dv=/dev/cua1:br#1200 dialup1|Dial-up system:¥ :pn=2015551212%:tc=UNIX-1200: hardwire:¥ :dv=/dev/ttyb:br#9600:el=^C^S^Q^U^D:ie=%$:oe=^D: tip0|tip9600:tc=UNIX-9600: cu1|cu9600:tc=UNIX-9600: slipit:st=slip:ls=/etc/login.script.unix S%h: ¥ :cc=/etc/sliplogin:tc=dialers: UNIX-9600:¥ :el=^D^U^C^S^Q^O@:du:at=ventel:ie=#$%:oe=^D:br#9600:tc=dialers: dialers:¥ :dv=/dev/cua2,/dev/cua1:¥ kernel config file: o This item may be ok, but per the per the directions, we receive errors when we enter "pseudo-device s1N init slattach" instead of just "pseudo-device s1N" Any suggestions or literature references would be appreciated. Thank you, Linda Watson lwatson@dasd.honeywell.com From MikeMcQuade@prometheus.connectsoft.net Fri Feb 9 18:07:15 1996 From: Mike McQuade Date: Fri Feb 9 18:07:19 PST 1996 Subject: PowerPC and Diab compiler Anyone using the Diab compiler with a PowerPC target ? if so, have you used floating point ? Im getting some strange results when I do floating point math on my moto MVME-1603 board. Mike McQuade MikeMcQuade@medio.net From daemon@csg.lbl.gov Sat Feb 10 04:00:47 1996 From: daemon@csg.lbl.gov Date: Sat Feb 10 04:00:50 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sat Feb 10 04:00:40 PST 1996 Subject: Re: problem with taskSpawn Subject: "UNIX" domain sockets Subject: Re: DejaGNU Subject: problem with taskSpawn Subject: Hex code conversion routine. ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: problem with taskSpawn Date: Fri, 9 Feb 1996 13:10:55 GMT From: ram@mr.picker.com (Bob Marino) Organization: Picker International, NMR Division Message-ID: <1996Feb9.131055.9583@picker.com> References: <4eqlsp$i9i@orphee.ensieta.fr> Sender: news@picker.com In article <4eqlsp$i9i@orphee.ensieta.fr> Jean-Luc Fleureau PERS writes: >Hi, > >I am a new user of vxWorks and I have a problem with the function taskSpawn >I have written : >taskSpawn("ttaskName",100,0,5000,myfunction,0,0,0,0,0,0,0,0,0,0); >and the compiler tells me : > >warning : passing arg 5 of 'taskSpawn' from incompatible pointer type. > >So I don't understand where the problem is. > >If someone can help me, thank you. >bye. > >-- > > >Jean-Luc FLEUREAU. >**************************************************************** >Dpt Electronique Informatique Automatique >Ecole Nationale Superieure des Etudes et Techniques d'Armement >2 Rue Francois VERNY 29806 BREST cedex 9 FRANCE > >Tout ce que vous avez toujours voulu savoir sur moi sans jamais >oser le demander ----> http://www.ensieta.fr/‾fleureje/ > >TEL : (33) 98 34 88 45 FAX : (33) 98 34 87 50 >Email : Jean-Luc.Fleureau@ensieta.fr >*************************************************************** > The compiler is complaining that myfunction is not declared as a function returning int. Either declare the function as such or type cast the function as FUNCPTR in the call to taskSpawn. I hope this helps. - -Roberto Marino principal engineer - ------------------------------------------------------------------------------- | Picker International, Inc. email: ram@mr.picker.com | | 5500 Avion Park Dr. phone: (216) 473-5763 | | Highland Hts., OH 44143 fax: (216) 473-5728 | - ------------------------------------------------------------------------------- --------------------------- Newsgroups: comp.os.vxworks Subject: "UNIX" domain sockets Date: Thu, 08 Feb 1996 10:07:20 -0800 From: Ricker Organization: Pacifier BBS, Vancouver, Wa. ((360) 693-0325) Message-ID: <311A3BD8.711D@pacifier.com> I know VxWorks has AF_INET and ZBUF type sockets, but is there a way to get "UNIX domain" (AF_UNIX) type sockets too ? Thanks for any advice on this. Regards, Alan Ricker --------------------------- Newsgroups: comp.os.vxworks,comp.software.testing Subject: Re: DejaGNU Date: Fri, 09 Feb 1996 10:02:58 +0000 From: Richard O'Shaughnessy Organization: Digital Equipment Corporation Message-ID: <311B1BD2.15FB@then17.ilo.dec.com> References: <4fagm1$ko6@superb.csc.ti.com> Gene Bruce wrote: > > I have recently discovered DejaGNU and would like to evaluate it for > potential use. > > The problem is while the distribution comes with a good reference, > there is no tutorial information nor any good hold-your-hand examples. > > Does any material of this nature exist? And where? > > This is also a solicitation for advice and opinions of experienced > users. > I have just recently started to evaluate DejaGnu also and the document I was using was Rob Savoye's "The DejaGnu Testing Framework", it's a good starting point and I was able to get up and running without much problem but it's not a 'user guide' in the classic sense of the word. There is a Dejagnu web page that has a section on documention that you can buy which I would presume would include user guides. The European site is http://www.maths.lancs.ac.uk:2080/‾maa036/GNU/GNUWeb/DejaGnu.html At the moment there is a thread called 'Pointers for Dejagnu' which hopefully will have more details on the latest release and on docs. Richard - -- Richard O'Shaughnessy, Email: oshaugh@then17.ilo.dec.com Digital Equipment Corporation, Phone: +353-91-754985 Ballybrit, DTN: 822-4985 Galway, Ireland. Fax: +353-91-754899 --------------------------- Newsgroups: comp.os.vxworks Subject: problem with taskSpawn Date: 1 Feb 1996 15:23:05 GMT From: Jean-Luc Fleureau PERS Organization: E.N.S.I.E.T.A, France. Message-ID: <4eqlsp$i9i@orphee.ensieta.fr> Hi, I am a new user of vxWorks and I have a problem with the function taskSpawn I have written : taskSpawn("ttaskName",100,0,5000,myfunction,0,0,0,0,0,0,0,0,0,0); and the compiler tells me : warning : passing arg 5 of 'taskSpawn' from incompatible pointer type. So I don't understand where the problem is. If someone can help me, thank you. bye. - -- Jean-Luc FLEUREAU. **************************************************************** Dpt Electronique Informatique Automatique Ecole Nationale Superieure des Etudes et Techniques d'Armement 2 Rue Francois VERNY 29806 BREST cedex 9 FRANCE Tout ce que vous avez toujours voulu savoir sur moi sans jamais oser le demander ----> http://www.ensieta.fr/‾fleureje/ TEL : (33) 98 34 88 45 FAX : (33) 98 34 87 50 Email : Jean-Luc.Fleureau@ensieta.fr *************************************************************** --------------------------- Newsgroups: comp.os.vxworks Subject: Hex code conversion routine. Date: 10 Feb 1996 02:16:23 GMT From: tomwang@ix.netcom.com(Tom Wang ) Organization: Netcom Message-ID: <4fgv5n$ldk@cloner4.netcom.com> Hi Folks, Does anyone have experience on debuging the vxWork's bootrom code ? I have done some modification on the vxWork's bootrom code. The boot code was compiled and programmed into the flash memory, but it doesn't boot. I tried to use the printf() to print the state information, but it doesn't work either. Then, I try to use the BDM and HP emulator to debug the boot code. This throw me into another domain. The BDM and HP emulator program require .abs ans .x format(IEEE 965 ?), but the tool that we have only generate Motorola S record. Where can I get the conversion routine for it ? Thanks. - -Tom Wang ======================================================================= eMail: s.wang@ieee.org Tel: 215-773-8526 --------------------------- End of New-News digest ********************** From lbecker@avtec.com Sat Feb 10 06:55:28 1996 From: lbecker@avtec.com (Lynn W. Becker II) Date: Sat Feb 10 06:55:31 PST 1996 Subject: Re: PowerPC and Diab compiler >Anyone using the Diab compiler with a PowerPC target ? >if so, have you used floating point ? >Im getting some strange results when I do floating point math on my >moto MVME-1603 board. I tried the Cetia PowerPC board with the Diab compiler about 4 months ago. I had some floating point problems. I contacted the power pc early release team and they responded "The VxWorks for PowerPC with the Diab compiler does not support floating point yet..." I later received and update saying that it was ready, but I already had chosen a MIPS R4700 based solution. You should contact Wind River systems... -- hope that helps. Lynn Becker lbecker@avtec.com From theo@ai.mit.edu Sat Feb 10 10:27:01 1996 From: theo@ai.mit.edu (Daniel) Date: Sat Feb 10 10:27:04 PST 1996 Subject: 6U frame grabber Hi, I am trying to find an inexpensive 6U color frame grabber, hopefully with VxWorks drivers or C I/F function modules available. Any suggestions? >Daniel A. Theobald >theo@mit.edu >617-253-6625 >fax 617-253-5060 >545 Technology Square Rm #831 >Cambridge, MA 02139 From daemon@csg.lbl.gov Sun Feb 11 04:00:32 1996 From: daemon@csg.lbl.gov Date: Sun Feb 11 04:00:36 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sun Feb 11 04:00:27 PST 1996 Subject: Sharing RTOS experience? Subject: Re: GNU make Subject: CDROM for VxWorks ??? ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Sharing RTOS experience? Date: Fri, 9 Feb 1996 18:13:23 GMT From: br771@torfree.net (Victor Gorin) Organization: Toronto Free-Net Message-ID: Dear VxWorks users, I am at the point where I have to select a RTOS for our new project that will target DSP boards and either 68K or PowerPC boards in a multiprocessor VME bus based system. We expect to do a cross-development on Sun Sparcstations. I am looking for information regarding VxWorks support. What is your experience? Would you advise us to switch to VxWorks or pSOS, for example. Are you satisfied with the development environment that WindRiver provides? I wonder what you do like about it and what you don't. I would appreciate if you can send me e-mail and share you thoughts, your good and bad experience. Regards, Victor Gorin. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: GNU make Date: 5 Feb 1996 13:39:27 GMT From: dyer@alx.sticomet.com (Doug Dyer) Organization: Naval Research Laboratory Message-ID: <4f51af$ptg@ra.nrl.navy.mil> References: <199602050720.PAA13067@webster.qpsx.oz.au> RUSSELL@qpsx.oz.au (Jones, Russell) writes: >I have set up several parallel directories, LIB, H, DEP and OBJ. LIB >contains source code, H contains common headers, OBJ is for the object >files, and DEP contains dependency files, generated using the -M compiler >option. For example: >D:¥TORNADO¥BETA¥SRC¥OBJ¥revreg.o: revreg.c >D:¥TORNADO¥BETA¥SRC¥H/qpsxstd.h ¥ > D:¥Tornado¥target/h/stdio.h D:¥Tornado¥target/h/types/vxANSI.h ¥ >... It gives me a headache to look at DOS paths, so Ill assume these are on the money. On thing to note is that revreg.c has no path in front of it. Make will do a few things here, one being to look for the source file with the same path as the result file. >In my makefile I include the dependency files (this is the method >recommended in section 4.12 of WindRivers Gnu toolkit users guide). As >far as I can work out, make should then use the implicit rule (.c.o) to >determine how to remake the object file. However, when I try to make >D:¥TORNADO¥BETA¥SRC¥OBJ¥revreg.o, make insists that there is nothing to >the same directory, although section 4.3.4 says they don't (I tried using >VPATH and vpath, but they didn't help). VPATH will have make scan those directories for the dependency file when it cannot find it locally. VPATH will store the resulting files in the current directory however unless you specify otherwise, but its hard to automate without helping make out too much. I just use a mv (I guess dos could use an rn?) to move the file to where it should be. Not pretty, but a solution. The easiest and less confusing method is to use absolute paths: SRC = D:¥....¥LIB¥ OBJ = D:¥....¥OBJ¥ H = D:¥....¥H¥ $(OBJ)revreg.o: $(SRC)revreg.c $(CC) $(CFLAGS) -i$(H) $? -o $@ It makes your makefile a little less flexible, so I recommend environment variables if appropriate,like: SRC = $(PROJDIR)¥SRC¥LIB¥ This can still mess up trees if you have to specify physical paths and want the same makefile cut and pasted everywhere, so this is just as good: SRC = .¥LIB¥ OBJ = .¥OBJ¥ H = .¥H¥ and just have the make file reside in D:¥...¥SRC This goes along with what I was saying earlier about make. We should have some utility that WANTS separation directores (obj, src, h, etc). Make requires a modest amount of experience to get satisfactory results, and for a tool thats too much (IMHO). Imake was designed for this but is a bit on the hard side :) cpp -M, makedepend, and a suite of other make assists don't help that much since in the end, you are dealing with make itself. Good luck, Doug "make cheese -> make: fatal error: don't know how to make cheese" Dyer - -- Doug Dyer - dyer@alx.sticomet.com | ECL: embedded scripting STI: voice (703) 329-9707 | for the 8051 family These opinions are yours --------------------------- Newsgroups: comp.os.vxworks Subject: CDROM for VxWorks ??? Date: Mon, 05 Feb 1996 12:07:04 -0800 From: Ray Peng Organization: Systems & Technonlogy Corp. Message-ID: <31166368.41C6@systech.hinet.net> Helo, Does anyone out there know any brands of CDROMs that can be used on vxworks 68K target ? The vxworks versions can be 5.1 or 5.2. The CDROM is SCSI-II interface (16-bits). Also the CDROM must have the device driver. Thanks for any help in advance. - -- -= Ray Peng =- Systems & Technology Corp. Taipei, Taiwan R.O.C. Voice: 886-2-6981599 Fax : 886-2-6981211 e-mail: peng@systech.hinet.net --------------------------- End of New-News digest ********************** From daemon@csg.lbl.gov Mon Feb 12 04:00:25 1996 From: daemon@csg.lbl.gov Date: Mon Feb 12 04:00:43 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Mon Feb 12 04:00:21 PST 1996 Subject: Motorola VME 68030 & 68040 boards and systems Subject: DejaGNU Subject: VME framegrabber/videoproc. ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Motorola VME 68030 & 68040 boards and systems Date: 12 Feb 1996 03:09:35 GMT From: innov8@ix.netcom.com (M.W. ) Organization: Netcom Message-ID: <4fmb1f$c8i@reader2.ix.netcom.com> I have several Motorola VME 68030 and 68040 cards and systems available. You can see at my web page http://205.230.70.48 or e-mail for listing. Regards - -- E-mail: innov8@ix.netcom.com --------------------------- Newsgroups: comp.os.vxworks,comp.software.testing Subject: DejaGNU Date: Wed, 07 Feb 1996 15:31:43 GMT From: gbruce@ti.com (Gene Bruce) Organization: Texas Instruments Message-ID: <4fagm1$ko6@superb.csc.ti.com> I have recently discovered DejaGNU and would like to evaluate it for potential use. The problem is while the distribution comes with a good reference, there is no tutorial information nor any good hold-your-hand examples. Does any material of this nature exist? And where? This is also a solicitation for advice and opinions of experienced users. Gene Bruce Texas Instruments History teaches that wars Systems Group begin when governments Electronic Systems Division believe the price of _/_/_/_/_/ _/_/_/ Ocean Surveillance Radar aggression is cheap. _/ _/ 2501 W. University, MS 8049 - Ronald Reagan _/ _/ McKinney, Texas 75070 _/ _/ Phone/Voice Mail: (214)952-4937 _/ _/_/_/ FAX: (214)952-2104 All expressions of opinion are strictly my own and do not reflect those of my employer. --------------------------- Newsgroups: comp.os.vxworks,comp.realtime,comp.arch.bus.vmebus Subject: VME framegrabber/videoproc. Date: Thu, 08 Feb 1996 08:41:33 +0100 From: Klaus Woerle Organization: hardly any! Message-ID: <3119A92D.45A7@fdt.de> Followup-To: comp.arch.bus.vmebus Reply-To: woerle@mpa-garching.mpg.de We are looking for VME framegrabber/videoprocessor modules. Who can give us names/addresses of manufactors/distributors? Any hint appreciated - thanks a lot! Klaus - -- Klaus Woerle Max-Planck-Institut fuer phone: (+49) 089/3299-3596 Extraterrestrische Physik fax: (+49) 089/3299-3569 Giessenbachstrasse eMail: kaw@rzg.mpg.de D-85748 Garching, Germany WWW: http://www.rzg.mpg.de/‾kaw --------------------------- End of New-News digest ********************** From twang@gic.gi.com Mon Feb 12 16:30:22 1996 From: "Wang, Tom (HT-MS)" Date: Mon Feb 12 16:30:44 PST 1996 Subject: How do I modify the Boot ROM size ? Hi Netter, Has anyone had experience on modifying the boot rom and changing the bootrom size related parameter ? I have modified the bootrom code and the code generated by the standard make script is larger than the original size. The size of the bootrom.hex file was changed from 310K to 566k. What are those MACROs or symbols related to the boot rom size on the make script and header file that I have to change ? Is it only the ROM_SIZE ? Following are the information related to our target: Our target is 68360 under vxWorks 5.2. We have 2M of flash memory and the flash is at 0x0, and we also have 4M of DRAM and it is at 0x8000000. The ROM_SIZE was modified from 0x80000 and to 0x90000(about 9 * 64K). (The size of the bootrom.hex is 567014(decimal), about 8.5 * 64K ) Is the ROM_SIZE good for the bootrom.hex ? I used nm68k on the "bootrom" after making the "bootrom.hex", and I got the following information. _romInit() is at 0x8010000. _uncompress() is at 0x8010268. When I ran the boot code under BDM, a debugger, and I set the break points in those two address, the content of the address was modified by other routine. I noticed that the content at _uncompress() was changed to 0x4E560000. Then, l let the target run after setting the breakpoints and it crashes at 0x4E560000. Thanks In advance for any help. -Tom Wang =============================================================== my Email: s.wang@ieee.org Tel: 215-773-8526 From jjs@mclean.sparta.com Mon Feb 12 18:24:46 1996 From: jjs@mclean.sparta.com (John J. Smith) Date: Mon Feb 12 18:24:49 PST 1996 Subject: Re: How do I modify the Boot ROM size ? > Submitted-by twang@gic.gi.com Mon Feb 12 16:30:22 1996 > Submitted-by: "Wang, Tom (HT-MS)" > > > > Hi Netter, > > Has anyone had experience on modifying the boot rom and > changing the bootrom size related parameter ? > > I have modified the bootrom code and the code generated by the standard > make script is larger than the original size. The size of the bootrom.hex > file was changed from 310K to 566k. What are those MACROs or symbols > related to the boot rom size on the make script and header file that I have > to > change ? Is it only the ROM_SIZE ? > Well... Yes, but you need to make sure that ALL references to ROM_SIZE are changed. Try this: 1. Modify ROM_SIZE in config.h 2. Modify ROM_SIZE in MakeSkel 3. Run Make. This should cause the makefiles to be regenerated with new values for ROM_SIZE that are taken from MakeSkel. 4. Just to be sure, grep for ROM_SIZE. All of the values (config.h, MakeSkel, Makefile, and Makefile.CPU32gnu) should have the new value. > Following are the information related to our target: > > Our target is 68360 under vxWorks 5.2. > We have 2M of flash memory and the flash is at 0x0, > and we also have 4M of DRAM and it is at 0x8000000. > Are you using the 2M flash as a boot ROM, or do you have a ROM mapped somewhere else? Assuming that you use the flash (you would have to, unles your bootrom does some fancy footwork with CS0), you could use any number up to 0x200000. > The ROM_SIZE was modified from 0x80000 and to 0x90000(about 9 * 64K). > (The size of the bootrom.hex is 567014(decimal), about 8.5 * 64K ) > Is the ROM_SIZE good for the bootrom.hex ? > bootrom.hex is a Motorola S-record file that should be at least twice the size of your binary image. You should have had no trouble at all fitting it into either rom size. Sounds like something else is wrong. If the ROM image was too large, the linker would have complained loudly. > I used nm68k on the "bootrom" after making the "bootrom.hex", > and I got the following information. > _romInit() is at 0x8010000. > _uncompress() is at 0x8010268. > This would be the locations of these functions after they were copied from ROM to RAM. > When I ran the boot code under BDM, a debugger, and I set the break points > in those two address, the content of the address was modified by other > routine. > I noticed that the content at _uncompress() was changed to 0x4E560000. > Then, l let the target run after setting the breakpoints and it crashes at > 0x4E560000. > Questions: How did you program the bootROM? Did you use a chip programmer, or a S-record filter and flash driver running on the 360? What are the first few lines of the bootrom.hex file? What are the other ROM and RAM related macros in your system? > Thanks In advance for any help. > > -Tom Wang > =============================================================== > my Email: s.wang@ieee.org Tel: 215-773-8526 > > ********** Basically, your ROM_SIZE looks OK. It looks like you may have incorrectly programmed the flash. _________________________________________________________ | |¥ |John J Smith | Voice -> 703 448 1683 x253 | | |jjs@mclean.sparta.com | FAX -> 703 734 3323 | | |_________________________________________________________| | | | | | /¥¥ SPARTA, Inc. | | | | ///¥¥ 7926 Jones Branch Drive | "Isn't Virtual | | | /////¥¥ Suite 900 | Reality an | | | ///// ¥¥ McLean, VA 22102 | Oxymoron?" | | | / ¥¥¥¥¥ ¥¥ 703-448-0120 | | | | ¥ ¥¥¥¥¥ // | | | | ¥ ////// | | | | ¥////// SPARTA | | | | ¥//// Pride In Performance | | | | ¥// -------------------- | | | |_________________________________________________________| | ¥__________________________________________________________¥| From has@info.kobelco.co.jp Mon Feb 12 22:01:27 1996 From: has@info.kobelco.co.jp (Hasegawa Shigeo) Date: Mon Feb 12 22:01:29 PST 1996 Subject: IEEE802.3 with SNAP Hi, Does anyone have ever done IEEE802.3 with SNAP on VxWorks? We'd like to know how you change Ethernet to IEEE802.3. Best Regards, ---------------------------------------- Shigeo Hasegawa *Electronics & Information Division *Kobe Steel, Ltd. TEL: 81-78-261-6461 FAX: 81-78-261-6499 E-mail: has@info.kobelco.co.jp URL: http://www.kobelco.co.jp/p069/ ---------------------------------------- From Arjon.Vlasblom@hta.nl Tue Feb 13 04:58:12 1996 From: Arjon Vlasblom Date: Tue Feb 13 04:58:14 PST 1996 Subject: VxSim crash when debugging. DESCRIPTION OF PROBLEM : When multiplying two integers (int) and stepping through the code whitin the debugger, VxSim will crash. After a short while the debugger will respond with: "RPC: Timed out" "Can't get status of target task. Detach from task? (y or n)" Invoking the function from the shell prompt will give no problems. DEMONSTRATION PROCEDURE : The following code does demonstrate the problem. #include #include void bug_test( void ) { int nr; int a = 4, b = 250; printf( "¥n - - - Bug test - - -¥n¥n" ); nr = a * b; printf( "Result: %d * %d => %d¥n", a, b, nr ); } This code is compiled with the following statement (from inside a makefile): /vol/vx/gnu/sun4.sparc/bin/ccsparc -g -ansi -pedantic -nostdinc -fno-builtin -pipe -DCPU=SIMSPARCSUNOS -DVX_IGNORE_GNU_LIBS -I/vol/vx/vxsim/h -c -o bug_tst.o bug_tst.c When this function is invoked from the shell in VxSim it will run normally, but stepping in the debugger will result in a crash at the multiply instruction. Looking at the assembly level for this source line a call to 'umul' function is done, while the used types are signed. QUESTION(S) : Does anybody know about this problem? And is their a workaround? Thanks in advance, Arjon Vlasblom System Engineer -- //// (0 0) _____________________________o00_( )_00o______________________________ / ¥ | Arjon Vlasblom High Tech Automation BV | | voice: +31 40 2447465 Piazza 408, 5611 AG Eindhoven | | fax : +31 40 2441260 The Netherlands | | email: arjon.vlasblom@hta.nl http://www.hta.nl | |----------------------------------------------------------------------| | T H E E M B E D D E D S O F T W A R E C O M P A N Y | |----------------------------------------------------------------------| | Disclaimer: This message does not necessarily reflect the company's | | opinion just my own humble one. | ¥______________________________________________________________________/ From CJONES@digsys1.bwi.wec.com Tue Feb 13 13:08:07 1996 From: "CHRIS JONES" Date: Tue Feb 13 13:08:09 PST 1996 Subject: Linker Can't Find Any Symbols Here it is. Anybody know what could be causing this one. This behavior just started to occur cc68k -X -N test.o -o test C:¥Tornado¥host¥x86-win32¥lib¥gcc-lib¥m68k-wrs-vxworks¥2.6-95q2¥ld.exe warning: cannot fi nd entry symbol start; defaulting to 00002000 test.o(.text+0x7e): undefined reference to `ftpXfer' test.o(.text+0x92): undefined reference to `printf' test.o(.text+0xa8): undefined reference to `read' test.o(.text+0xc2): undefined reference to `close' test.o(.text+0xd4): undefined reference to `ftpReplyGet' test.o(.text+0xf4): undefined reference to `ftpCommand' test.o(.text+0x106): undefined reference to `close' cc68k: unrecognized option `-X' cc68k: Internal compiler error: program ld got fatal signal 1 make.exe: *** [test] Error 0x1 Done. Chris Jones cjones@digsys1.bwi.wec.com Electrical Engineer Westinghouse Electronic Systems (410) 765-7959, (410) 993-8822 (FAX) From twang@gic.gi.com Tue Feb 13 13:10:51 1996 From: "Wang, Tom (HT-MS)" Date: Tue Feb 13 13:10:53 PST 1996 Subject: RE: How do I modify the Boot ROM size ? Hi Netter, Thanks for the help on the BOOT_ROM size. The BOOT_ROM size is fine for my modified boot Rom code. I decided to simplify the case and built the un-compressed version of the bootrom and loaded it into the flash memory of the ADS360 target board. In the built, I added the taskDelay(60) after every printf(), and this help the printf() to print the status until it reaches the cplusCtorLink() call, which was added in the usrRoot() to construct the static C++ objects. The program seems to fail at here, and the function pointer array, for the static constructors is defined after munching the C++ code. VOIDFUNCPTR _ctor[] = { __sti__xxx_cpp_func1_, // which address is 0x08308306 __sti__xxx_cpp_func2_, // which address is 0x083076ca __sti__xxx_cpp_func3_, // which address is 0x083071c8 __sti__xxx_cpp_func4_, // which address is 0x08305046 }; And the content at these address are all the same, e.g. 0x08308306 4E56 0000 4879 0834 .... After the code had been moved from the ROM to RAM. Then I setup the BDM with the target, the target fails when it try to run at the PC = 0x4E56 0000. (This happens to be the op-code for JSR (A6)). All these information point to the _ctors[] arrays, which might be setup incorrectly during memory copy from ROM to RAM. Has anyone had this experience ? Thanks in advance. -Tom Wang ======================================================== Email: Twang@gic.gi.com Tel: 215:773-8526 From matt@wrs.com Tue Feb 13 13:53:09 1996 From: Matt Rodgers Date: Tue Feb 13 13:53:12 PST 1996 Subject: Re: Linker Can't Find Any Symbols the vxWorks Users Group Exploder wrote: > Submitted-by CJONES@digsys1.bwi.wec.com Tue Feb 13 13:08:07 1996 > Submitted-by: "CHRIS JONES" > > > Here it is. Anybody know what could be causing this one. > This behavior just started to occur > You will need to create a object that is compiled, but not linked. use the -c option in your compile line, or better yet, use a makefile that resembles the makefile of your 68k targets. Example from mv177: cc68k -c -O2 -fvolatile -fno-builtin -m68040 -Wall -I/h -I. -I/net/shannon/shannon2/tornado/target/config/all -I/net/shannon/shannon2/tornado/target/h -I/net/shannon/shannon2/tornado/target/src/config -I/net/shannon/shannon2/tornado/target/src/drv -ansi -nostdinc -DCPU=MC68060 usrConfig.c > > cc68k -X -N test.o -o test > C:¥Tornado¥host¥x86-win32¥lib¥gcc-lib¥m68k-wrs-vxworks¥2.6-95q2¥ld.exe > warning: cannot fi nd entry symbol start; defaulting to 00002000 > test.o(.text+0x7e): undefined reference to `ftpXfer' > test.o(.text+0x92): undefined reference to `printf' > test.o(.text+0xa8): undefined reference to `read' test.o(.text+0xc2): > undefined reference to `close' test.o(.text+0xd4): undefined reference > to `ftpReplyGet' test.o(.text+0xf4): undefined reference to > `ftpCommand' test.o(.text+0x106): undefined reference to `close' > cc68k: unrecognized option `-X' cc68k: Internal compiler error: > program ld got fatal signal 1 make.exe: *** [test] Error 0x1 See ya. -- Thanks, Matt. |---Matt Rodgers Technical Support Engineer (510) 814-2565 matt@wrs.com---| |--Wind River Systems 1010 Atlantic Ave. 1(800) 872-4977 support@wrs.com--| |--------------------------fax: 1(510)814-2164----------------------------| From dit@bach.jhuapl.edu Tue Feb 13 14:44:35 1996 From: dit@bach.jhuapl.edu (Daryl I. Tewell) Date: Tue Feb 13 14:44:37 PST 1996 Subject: Re: problem with taskSpawn > Hi, > > I am a new user of vxWorks and I have a problem with the function taskSpawn > I have written : > taskSpawn("ttaskName",100,0,5000,myfunction,0,0,0,0,0,0,0,0,0,0); > and the compiler tells me : > > warning : passing arg 5 of 'taskSpawn' from incompatible pointer type. > > So I don't understand where the problem is. > > If someone can help me, thank you. > bye. Because it is generally useless to have a vxWorks task return something, most people type them as functions returning void. WRS, however, opted to type this parameter as FUNCPTR instead of VOIDFUNCPTR. Do the following: - in file taskLib.h, change the type of the entryPt parameter of taskSpawn from FUNCPTR to VOIDFUNCPTR. An alternative here is to simply cast myfunction to FUNCPTR in your taskSpawn call. - define __PROTOTYPE_5_0 in either your source file or your makefile. This selects an alternate taskSpawn prototype, allowing you to call taskSpawn without those confusing and superfluous 0s. -Daryl From vdbent@hea.sron.ruu.nl Wed Feb 14 01:09:17 1996 From: Kees van der Bent Date: Wed Feb 14 01:09:19 PST 1996 Subject: 80386EX Board Support Hi, Are there any processor specific considerations for creating a BSP for a 80386EX board - say in comparison with 68k? Does some of you have (prototype) BSP (fragments) for boards having a 80386 you're willing to share with me? Who sells BSPs for 80386 boards? Thanks a lot! | ### ## | #### ### ## # | Kees van der Bent | #### ### | ## ## # ### ##### | Space Research Organisation Netherlands | #### ## | #### ## ## ## ## | Sorbonnelaan 2 3584 CA Utrecht | #### | ## ## ### ## ## | The Netherlands | ## #### | | | ### #### | | +31-30-2538595 | ## ### | | vdbent@sron.ruu.nl From AS_Bridge@msmail3.hac.com Wed Feb 14 04:23:19 1996 From: "AS Bridge" Date: Wed Feb 14 04:23:21 PST 1996 Subject: Undeliverable Mail Unknown Microsoft mail form. Approximate representation follows. Message: 80386EX Board Support Sent: Wed, Feb 14, 1996 3:08 AM To: Carroll, John C; Claudio, Michael A On Server: 631 Segment5 Date: Wed, Feb 14, 1996 4:22 AM Reason: Could not be delivered because the destination Microsoft Mail server could not be found. From AS_Bridge@msmail3.hac.com Wed Feb 14 07:21:55 1996 From: "AS Bridge" Date: Wed Feb 14 07:21:57 PST 1996 Subject: Undeliverable Mail Unknown Microsoft mail form. Approximate representation follows. Message: Undeliverable Mail Sent: Wed, Feb 14, 1996 6:16 AM To: Carroll, John C; Claudio, Michael A On Server: 631 Segment5 Date: Wed, Feb 14, 1996 7:21 AM Reason: Could not be delivered because the destination Microsoft Mail server could not be found. From uchenick@bud.tate.com Wed Feb 14 07:30:33 1996 From: Gordon Uchenick Date: Wed Feb 14 07:30:35 PST 1996 Subject: Compressed Image After Boot? Currently my startup script uses ld commands to load several .o files. I'm trying to keep the size of them down. Is it possible to compress them and have them expanded into RAM as part of the loading process? This appears to be the same thing that happens when you boot a compressed ROM image. Is that facility available to me, or just at boot time? Looking at the docs for loadLib doesn't present any hints to me. TIA, Gordon Uchenick uchenick@tate.com From twang@gic.gi.com Wed Feb 14 10:05:09 1996 From: "Wang, Tom (HT-MS)" Date: Wed Feb 14 10:05:12 PST 1996 Subject: RE: How do I modify the Boot ROM size ? Hi Netter, Following is my previous post, and I have the solusion for this. > In the built, I added the taskDelay(60) after every printf(), and this help > the printf() to print the status until it reaches the cplusCtorLink() call, > which was added in the usrRoot() to construct the static C++ objects. > > The program seems to fail at here, and the function pointer array, > for the static constructors is defined after munching the C++ code. > VOIDFUNCPTR _ctor[] = > { > __sti__xxx_cpp_func1_, // which address is 0x08308306 > __sti__xxx_cpp_func2_, // which address is 0x083076ca > __sti__xxx_cpp_func3_, // which address is 0x083071c8 > __sti__xxx_cpp_func4_, // which address is 0x08305046 > }; >And the content at these address are all the same, e.g. > 0x08308306 4E56 0000 4879 0834 .... > After the code had been moved from the ROM to RAM. > Then I setup the BDM with the target, the target fails when > it try to run at the PC = 0x4E56 0000. > (This happens to be the op-code for JSR (A6)). > > All these information point to the _ctors[] arrays, which might be > setup incorrectly during memory copy from ROM to RAM. After setting up several test cases under BDM, I finally located the problem right in the cpluCtorsLink(), which I believe it has a bug. The cplusCtorLink() calls the internal function with pointer to the first entry of the cotrs[] and it fails with the PC=4E56 0000. I patched the assembly code to point to the "ctors" and it went through this static object construction. Thanks. -Tom Wang ============================================================ EMail: s.wang@ieee.org Tel: 215-773-8526 From ZINGLEBJ@ocd231.bwi.wec.com Wed Feb 14 11:20:13 1996 From: "Zingler, Bill J." Date: Wed Feb 14 11:20:23 PST 1996 Subject: Tornado Device Drivers for S2681, 88C681, 68681 serial devices Greetings to all, Does any one have or know the where-a-abouts of Tornado Device Drivers for S2681, 88C681, or 68681 serial devices? Thanks in advance, Bill _________________________________________________________________ Bill Zingler Senior Software Engineer Westinghouse Electric Corporation Oceanic Division/D.W. Complex P.O. Box1488/MS-9RDW Annapolis, MD 21404 Email: zinglebj@ocd231.bwi.wec.com Phone: 410-260-5965 Fax: 410-260-5916 _________________________________________________________________ From CJONES@digsys1.bwi.wec.com Wed Feb 14 12:24:57 1996 From: "CHRIS JONES" Date: Wed Feb 14 12:24:59 PST 1996 Subject: Displaying Filenames in Minimized Windows In the Tornado 1.0 for Windows, when a source file is minimized, how do you get only the filename to be displayed in the title block instead of the entire pathname for the file. I have 10 files open and they all look the same when minimized because only the first few characters of the pathname are displayed. Word and Excel etc display the file name only. Is there an option to change this behavior or is this just a very , VERY annoying flaw of Tornado? Chris Jones cjones@digsys1.bwi.wec.com Electrical Engineer Northrop Grumman Electronic Sensors and Systems Division (ESSD) (410) 765-7959, (410) 993-8822 (FAX) From daemon@csg.lbl.gov Thu Feb 15 04:00:39 1996 From: daemon@csg.lbl.gov Date: Thu Feb 15 04:00:42 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Thu Feb 15 04:00:31 PST 1996 Subject: type casting problem Subject: Re: Help My Ballistic Coworker Subject: Programming the Bootrom Subject: TCL 7.5b1 port anywhere? Subject: (no subject) Subject: Re: Booting VxWorks from PC via ethernet Subject: Re: Domain name services for VxWorks Subject: VGA device drives for VxWorks ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: type casting problem Date: 12 Feb 1996 12:56:12 GMT From: jose.marin@ksc.nasa.gov (Jose Marin) Organization: NASA KSC Message-ID: <4fnddc$eke@itchy.ksc.nasa.gov> Reply-To: jose.marin@ksc.nasa.gov The following works in UNIX but doesn't work in VxWorks: unsigned short raw_count; float data, coef[2]; raw_count = (unsigned short) ((data - coef[0]) / coef[1]); This is how I got it to work in VxWorks: unsigned short raw_count; float data, coef[2]; raw_count = (unsigned short) (irint(trunc((data - coef[0]) / coef[1]))); I am looking for a better alternative. Any suggestions? --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Help My Ballistic Coworker Date: 9 Feb 1996 17:41:37 GMT From: "Dennis R. Perlman" Organization: Lockheed Martin -- Denver InterNetNews site Message-ID: <4fg10h$qv8@tel.den.mmc.com> References: <4f7vmi$a5o@tel.den.mmc.com> <4feh39$3c6@sulu.sd.aplabs.com> Trung La wrote: >Dear Scott, > >If possible, I would try setting your VME boards in A16 or A24 space because >you can map them in completely (64K and 16M respectively) in VxWorks. If not, >I would shoot for a range of addresses in A32. You can't map in all 4GB but >you don't need to be exact either. > Why can't you map in all 4GB? With 32bit page entries, each mapping a 0x2000 byte page, it would only take 1MB for the page tables. Adding in the overhead for the pointer tables wouldn't add much more. One of the problems we're facing is that our SBCs are in a SGI challenge XL. The SGI hardware will assign VME addresses dynamically as a process allocates memory. This address could be anywhere in the upper 2GB space. One idea to make sure that the SBC with vxWorks knows about the allocated space is to do the above and map in the entire range of possibilities. Another concern is what is the performance hit for doing address translation? What if we just flat out disable address translation in the TC register? If anybody has answers to this, please let us know. Another co-worker of Mr. Ballistic --------------------------- Newsgroups: comp.os.vxworks Subject: Programming the Bootrom Date: 13 Feb 1996 14:44:11 GMT From: getz@lfs.loral.com Organization: Loral Federal Systems, Manassas Message-ID: <4fq83r$a6v@lfsserv1.fsc.ibm.com> Reply-To: getz@lfs.loral.com I'm trying to determine the proper way to program the bootrom for a Force CPU-5CE which runs VxWorks 5.2. The 5CE Bootrom is in flash memory. We need SCSI Boot support, thus the need to reprogram the bootrom. I've made the changes to vxWorks for SCSI support and we can use the SCSI disk now. There are BSP routines provided for writing files to flash, so I think I understand how to do that. My question is which bootrom file is loaded to flash (bootrom, bootrom.hex, bootrom.uncomp, ....) Also, is there a good way to test changes to the bootrom code prior to overwriting the current bootrom? Thanks, Fred Getz --------------------------- Newsgroups: comp.os.vxworks Subject: TCL 7.5b1 port anywhere? Date: 9 Feb 1996 04:38:55 GMT From: tavares@connix.com (Christopher C. Tavares) Organization: Connix - The Connecticut Internet Exchange Message-ID: <4fej4v$2tj@comet.connix.com> Has anyone tried porting the new version of TCL to Vxworks yet? I got 7.4 running, but 7.5 looks like it may be more of an effort that I have time to put in right now. Thanks, - -Chris Tavares --------------------------- Newsgroups: comp.os.vxworks,comp.arch.bus.vme Subject: (no subject) Date: 14 Feb 1996 20:02:06 GMT From: Lawrence Pfeffer Organization: NIST Message-ID: <4ftf3u$ke@dove.nist.gov> Greetings all, I'm looking for recommendations for VME cage vendors and also disk drives that people have had good experiences using in VxWorks-based systems (or similar VME-based systems.) For local disks, I'm more interested in reliability/ease of integration than amazing speed/capacity. U.S. based vendors somewhat preferred. Please let me know what you've used and liked (or, alternately, things to avoid.) Email responses probably best, since other newgroup readers may not be interested. However, if there is interest on either group, let me know, and I'll post a summary. Thanks in advance, Larry - -- Lawrence E. Pfeffer, Ph.D. (AKA Larry), voice: (301)975-6067, FAX: (301)869-6275 *Preferred* email: lawrence.pfeffer@nist.gov U.S. mail: NIST, Bldg. 226/Rm. B158, Gaithersburg, MD 20899 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Booting VxWorks from PC via ethernet Date: 15 Feb 1996 08:31:11 GMT From: Chris Wallett Organization: Cranfield Precision Message-ID: <4fur0f$8q1@zinc.compulink.co.uk> References: <850819176wnr@test.demon.co.uk> You need to setup the boot parameters in the MVME167 to use FTP and not rsh by putting in a password. You will also require TCP/IP and FTP server software for Windows. We use the SuperTCP package for WIndows 3.X but are moving to Windows NT as it includes TCP/IP and FTP Server as standard. (Windows 95 includes TCP/IP but not FTP Server) Regards Chris W. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Domain name services for VxWorks Date: Fri, 09 Feb 1996 15:03:31 -0500 From: Kevin Rice Organization: NASA Goddard Space Flight Center -- Greenbelt, Maryland USA Message-ID: <311BA893.B01@vlsi.gsfc.nasa.gov> References: <4ffver$21pm@news-s01.ny.us.ibm.net> brucej@ibm.net wrote: > > Does anyone know where I can get a simple stub implementation of DNS > for VxWorks either chargeable or free ? post any positive results please! Thanks, Kevin Rice --------------------------- Newsgroups: comp.os.vxworks Subject: VGA device drives for VxWorks Date: 15 Feb 1996 11:14:46 GMT From: brucej@ibm.net Message-ID: <4fv4j6$38ti@news-s01.ny.us.ibm.net> Reply-To: brucej@ibm.net Does anyone have any VGA device drivers for VxWorks ? The m6845 driver supplied is a bit crappy, doesn't have all the features and every function is LOCAL to the driver. Please mail me. Cheers Bruce Jackson (brucej@ibm.net) --------------------------- End of New-News digest ********************** From tkb@mclean.sparta.com Thu Feb 15 08:41:04 1996 From: Thomas Keith Buchanan Date: Thu Feb 15 08:41:07 PST 1996 Subject: Re: Help My Ballistic Coworker > Why can't you map in all 4GB? With 32bit page entries, each mapping a 0x2000 > byte page, it would only take 1MB for the page tables. Adding in the overhead > for the pointer tables wouldn't add much more. Because 2^32 = 4Gb... You can't map it all in unless your CPU has and address space greater than 32 bits, there would be no room left for local memory. Having said that, aren't you guys using an older SPARC target? I have worked with MicroSPARCs, SuperSPARCs and HyperSPARCs all of which have a 36-bit address space. You could map in all of A32 with any of those CPUs. You guys mentioned something about having no MMU or a static map of some kind on an older SPARC variant. You need to query Wind River about the capabilities of your MMU and cache libraries with your unique target. > Another concern is what is the performance hit for doing address translation? > What if we just flat out disable address translation in the TC register? It's not just the MMU translations you need to consider. If your architecture has them, PTE caches, instruction caches, data caches, the Sbus interface, the Mbus interface and the VMEbus interface all play a role in your overall performance. The purpose of the MMU, caches etc is to feed the RISC engine in the most efficent fashion. The output of a good compiler can execute at rates approaching 3 instructions/cycle in a SuperSPARC. Embedded systems pros often overlook the down side of such architectures, the time to execute any given instruction is not deterministic. It could 1/3 of a clock cycle or 12 clock cycles depending on the status of the MMU and the cache tables. It could take even longer if you have to arbitrate for the Mbus and/or the VMEbus. These sorts of issues are a big reason why vxWorks sets up a single simple memory context by default. Having multiple contexts and switching between them on a task basis ( ala Unix, OS/2, Win95-hahaha!! ) would make the system performance much less deterministic and make the time to do a task context switch an order of magnitude larger. If you change the TC register, it is likely that the boot code will have to be altered. In a SPARC, the goofy vxWorks boot sequence is further complicated by having 3 different memory contexts, the normal SPARC boot mode, the context setup before the kernel is copied into RAM and the context determined by the normal MMU and cache library setup and sysPhysMemDesc. I guess you can't you can't get something for nothing, the newer RISC architectures are great tools and are capable of awesome throughput but setting them up is a much more involved matter than the good ole 68k. Analyzing and tuning their performance in a deadline driven environment is an even more daunting exercise. adios ______________________________________________________ | |¥ |Thomas Keith Buchanan Voice -> 703 448 1683 x233| | |tkb@mclean.sparta.com FAX -> 703 734 3323 | | |_____________________________________________________| | | | | | /¥¥ SPARTA, Inc. | | | ///¥¥ 7926 Jones Branch Drive | | | /////¥¥ Suite 900 | | | ///// ¥¥ McLean, VA 22102 | | | / ¥¥¥¥¥ ¥¥ 703-448-0120 | | | ¥ ¥¥¥¥¥ // | | | ¥ ////// | | | ¥////// SPARTA | | | ¥//// Pride In Performance | | | ¥// -------------------- | | |_____________________________________________________| | ¥______________________________________________________¥| From tkb@mclean.sparta.com Thu Feb 15 08:49:09 1996 From: Thomas Keith Buchanan Date: Thu Feb 15 08:49:11 PST 1996 Subject: Re: Programming the Bootrom > There are BSP routines provided for writing files to flash, so I think I > understand how to do that. My question is which bootrom file is loaded to > flash (bootrom, bootrom.hex, bootrom.uncomp, ....) None of the above, you can use the compressed or uncompressed ( depending on your architecture ) binary file after passing it through the utility aoutToBin to strip the a.out header. The normal binaries files are suitable for use with a loader but not to put directly into flash. Be sure to check to make sure what type of file your flash programming functions expect to see, I've seen OEMs that write them to take a.out files and/or binary files. adios ______________________________________________________ | |¥ |Thomas Keith Buchanan Voice -> 703 448 1683 x233| | |tkb@mclean.sparta.com FAX -> 703 734 3323 | | |_____________________________________________________| | | | | | /¥¥ SPARTA, Inc. | | | ///¥¥ 7926 Jones Branch Drive | | | /////¥¥ Suite 900 | | | ///// ¥¥ McLean, VA 22102 | | | / ¥¥¥¥¥ ¥¥ 703-448-0120 | | | ¥ ¥¥¥¥¥ // | | | ¥ ////// | | | ¥////// SPARTA | | | ¥//// Pride In Performance | | | ¥// -------------------- | | |_____________________________________________________| | ¥______________________________________________________¥| From MikeMcQuade@prometheus.connectsoft.net Thu Feb 15 16:25:09 1996 From: Mike McQuade Date: Thu Feb 15 16:25:21 PST 1996 Subject: PowerPC interupt handler question Im at the point where I need to service an interupt on our PowerPC board (Motorola 1603) does anyone have a little code sample of how to connect an interupt to a C routine ? or know of any examples ? thanks. MikeMcQuade@connect.com From k_ksmz@tyl.co.jp Thu Feb 15 21:45:08 1996 From: K Koshimizu/TYL Date: Thu Feb 15 21:45:11 PST 1996 Subject: creating sub-direcotry on net device By using "mkdir" or "open" to netDevice, such as boothost: sun WS, they return ok and the named file is created. But it is a file not direcotry. If NFS is used, either "mkdir","open" will create directory. I was woundering if it is possible to create directory on WS from vxWorks without using NFS. I am using vxWorks 5.1.1 on MVME147. WS SS/5 with SunOS 4.1.3. Does any one have any information ? Thanks in advance. Kazu Koshimizu k_ksmz@tyl.co.jp From daemon@csg.lbl.gov Fri Feb 16 04:00:26 1996 From: daemon@csg.lbl.gov Date: Fri Feb 16 04:00:31 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Fri Feb 16 04:00:21 PST 1996 Subject: Re: DejaGNU Subject: frame grabber PC platform / PCI / BIOS HELP!! Subject: irq3, intConnect, dvme660 ? Subject: Re: SLIP, vxworks and a Sun workstation Subject: processor/memory PMC Subject: Re: Question on filesystem performance Subject: (no subject) Subject: xntp Subject: Re: Hex code conversion routine. Subject: Re: Seeking recommnedations for VME cage, disk drive vendors (second try) Subject: Problem: very long booting time Subject: Re: Programming the Bootrom Subject: Re: Booting VxWorks from PC via ethernet ------------------------------------------------------- Newsgroups: comp.os.vxworks,comp.software.testing Subject: Re: DejaGNU Date: 14 Feb 1996 10:18:41 GMT From: giersig@atusel23 (Roland Giersig) Organization: Alcatel Austria AG Message-ID: <4fscu1$nqk@atusks02.aut.alcatel.at> References: <4fagm1$ko6@superb.csc.ti.com> Followup-To: comp.os.vxworks,comp.software.testing Gene Bruce (gbruce@ti.com) wrote: : I have recently discovered DejaGNU and would like to evaluate it for : potential use. : The problem is while the distribution comes with a good reference, : there is no tutorial information nor any good hold-your-hand examples. : Does any material of this nature exist? And where? : This is also a solicitation for advice and opinions of experienced : users. I would suggest that you also get a good book on Expect ("Exploring Expect", O'Reilly & Ass., by Don Libes himself is excellent) and look into that first. DejaGnu is written wholly in Expect, and once you have started using Expect, the concepts and workings of DejaGnu are much more clear. We are using DejaGnu for regression-testing our own OS and I wrote the testsuite-lib that handles the whole cross-testing environment. In my opinion DejaGnu is quite heavily trimmed to compiler testing (not astonishing since this was its first application) but can be used for other straight-program-tests as well. But when you want to use it for something more different (like tests that span multiple platforms), it only makes sense to use the most basic DejaGnu functionality. The rest must be added by yourself (for which you should be fluent in Expect anyway). IMHO you cannot pull DejaGnu out of the box and have it up and running immediately. I'd suggest starting to get acquainted with Expect, then writing some Expect scripts which run some testprograms and only then start looking into DejaGnu to see what it can do to help you. Roland - -- Roland.Giersig@aut.alcatel.at (speaking only to, err, for myself) ALCATEL Austria, Scheydgasse 41, A-1210 WIEN. Phone: +43-1-27722-3755 GS/CS/IT d-(+) s++:++ a>? C++$ US*++$ P+>+++ L>++ E+>++ W+>$ N+ w--- V>? PS+ PE Y+ PGP?>++ t++>+++ 5? tv+ b>++++ D++(---) G+ e+++>* h----(*) r+++ --------------------------- Newsgroups: comp.os.vxworks Subject: frame grabber PC platform / PCI / BIOS HELP!! Date: Wed, 14 Feb 1996 12:16:12 -0800 From: peter Barrie Organization: University of Strathclyde Computer Centre Message-ID: <3122430C.4A2B@cs.strath.ac.uk> Urgent application ... I need to attach a frame grabber to a Pentium PC system running VxWorks. This is for a video compression application. I haven't purchased the grabber yet. I'm open to any ideas - an absolute dream for sales persons, but ......... Most cards come with software and setup (usually complex) for Windows , DOS OS/2 etc. Although many provide the source code, there are nearly always BIOS calls. BIOS calls are not possible under VxWorks once it has booted(I think I am correct here, can anybody advise ??). Many of the better cards are PCI . Can VxWorks be set up to support PCI peripherals on a PC platform ?? The frame grabber source I've got for a Data Translation PCI card has lots of PCI BIOS specific stuff. Has anybody solved any of these problems ? Should I program under DOS or Windows instead ! Pete Barrie Computer Science University of Strathclyde --------------------------- Newsgroups: comp.os.vxworks Subject: irq3, intConnect, dvme660 ? Date: Thu, 15 Feb 1996 09:26:17 GMT From: rla@netcom.com Organization: NETCOM On-line Communication Services (408 261-4700 guest) Message-ID: Sender: rla@netcom16.netcom.com I'm using a sparc1e VME card & VxWorks 5.0.2b I need to get interrupts working on a Datel dvme660 board. The dvme660 has a ID register which is output on D0-D7 to develop the address of the IRQ interrupt routine. I'm assuming the longword vector address offset of IRQ3 goes in the ID register. So far all I've got to work is reboot at interrupt. I'm not sure if I'm setting up the Datel ID register wrong and/or the intConnect. I'm looking for sample code. Thanks rla@netcom.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: SLIP, vxworks and a Sun workstation Date: 10 Feb 1996 08:16:10 GMT From: chaos@linux.peacefulstar.oakland.ca.us (H.J. Bae) Organization: Peaceful Star, Oakland, CA Message-ID: <4fhk8a$ij0@idiom.com> References: <9602091724.AA20607@arcserv.dasd.honeywell.com> Reply-To: hjb@peacefulstar.oakland.ca.us it's really not all that complicated. depending on the version of slip code you're running on the sunos side, you will need to run sliplogin or slattach program which takes tty device name through which slip is run. that's all it takes for a functional slip line. - -- H.J. Bae, Consultant hjb@peacefulstar.oakland.ca.us http://peacefulstar.oakland.ca.us/ --------------------------- Newsgroups: comp.os.vxworks Subject: processor/memory PMC Date: Tue, 13 Feb 1996 21:42:01 GMT From: ram@mr.picker.com (Bob Marino) Organization: Picker International, NMR Division Message-ID: <1996Feb13.214201.6214@picker.com> Sender: news@picker.com Does anyone know of, or is anyone planning on putting at least one RISC processor, some memory, and optionally some L2 cache on a PMC format card? - -Bob Marino principal engineer - ------------------------------------------------------------------------------- | Picker International, Inc. email: ram@mr.picker.com | | 5500 Avion Park Dr. phone: (216) 473-5763 | | Highland Hts., OH 44143 fax: (216) 473-5728 | - ------------------------------------------------------------------------------- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Question on filesystem performance Date: Thu, 15 Feb 1996 14:38:16 GMT From: mfisher@dyn.com (Mark Fisher) Organization: The World @ Software Tool & Die Message-ID: References: Reply-To: mfisher@dyn.com Sender: news@world.std.com (Mr Usenet Himself) Flemming_Kevin/macom_pbl_ctc@mailback.macom.com wrote: >Possible solutions: (Snip) > It was rumored last year that WindRiver >would be putting in a command to allow pre-allocation of a directory. Preallocation can be done with vxWorks 5.2, I'm not sure about earlier releases. What you have to do is when you first create the directory (or maybe even when it is empty?) you open the directory and then issue an ioctl() call with FIOCONTIG. This allows one to preallocate contiguous space for the directory. The same can also be done for files. (Note that when using the DOSFS file system be sure to get the latest patch from WRS -- there have been various problem with the dosFsLib. At the moment I believe I have found a problem where it is using mututal exclusion semaphores that are NOT inversion safe. I'm waiting for an official fix for this right now.) Hope this helps. Mark Fisher --------------------------- Newsgroups: comp.os.vxworks Subject: (no subject) Date: 10 Feb 1996 19:14:40 GMT From: Vladimir Romanovski Organization: DESY Message-ID: <4fiqr0$o04@dscomsa.desy.de> Hello, I have an interest of monitoring memory usege on VME board. I try to use memShow(0). Unfortunately, that routine send its report in stdout. How can redirect that in stdin of the same task or another? I have tryed to use for such kind redirection a pipe, but without any success.. May be someone has some an experience in that metter? Thanks in advance, Vladimir --------------------------- Newsgroups: comp.os.vxworks Subject: xntp Date: 15 Feb 1996 20:12:09 GMT From: duhrenho@vette.sanders.com (David Uhrenholdt) Organization: Lockheed Sanders Message-ID: Is there a port of xntp available for VxWorks? Thanks for any replies Dave - -- Dave Uhrenholdt duhrenho@vette.sanders.lockheed.com Phone (603) 885-7398 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Hex code conversion routine. Date: Sat, 10 Feb 1996 16:04:03 GMT From: krister@unidata.se (Krister Lagerstrom) Organization: Tripnet AB (modem +46 31 831683) Message-ID: <4fif49$5uf@minox.tripnet.se> References: <4fgv5n$ldk@cloner4.netcom.com> tomwang@ix.netcom.com(Tom Wang ) wrote: >I have done some modification on the vxWork's bootrom code. >The boot code was compiled and programmed into the flash memory, >but it doesn't boot. >I tried to use the printf() to print the state >information, but it doesn't work either. You can't use printf before the kernel is set up properly, and there are lot's of things that can go wrong before that. Instead of printf(), you can try to use digital outputs connected to LEDs, an oscilloscope or a logic analyzer to see how far you get before the crash. In your particular case however, I highly recommend using the BDM-emulator instead. >Then, I try to use the BDM and HP emulator to debug the boot code. >This throw me into another domain. The BDM and HP emulator program >require .abs ans .x format(IEEE 965 ?), but the tool that we have only >generate Motorola S record. Where can I get the conversion routine for >it ? It is possible that the GNU program objcopy can be used for this. I'm not sure if it's included in your VxWorks release, but you can find it in the binutils package. It's not really necessary to have the symbol information in order to debug the bootrom code, the listing of variable and function addresses produced by nm68k together with assembly level display in the BDM emulator is enough in many cases. It sure beats LEDs anyway. >-Tom Wang >======================================================================= >eMail: s.wang@ieee.org >Tel: 215-773-8526 / Krister - --- Krister Lagerstrvm Email: krister@unidata.se UniData HB Tel: +46 31 20 13 07 Box 240 59 Fax: +46 31 20 58 10 S-400 22 Gvteborg --------------------------- Newsgroups: comp.os.vxworks,comp.arch.bus.vme Subject: Re: Seeking recommnedations for VME cage, disk drive vendors (second try) Date: 16 Feb 1996 05:58:51 GMT From: "Phil S. Wilshire" Organization: Kaiser Aluminum Message-ID: <4g16er$3nj@express.ior.com> References: <4ftfdj$ke@dove.nist.gov> Lawrence Pfeffer wrote: >I'm looking for recommendations for VME cage vendors and also disk >drives that >people have had good experiences using in VxWorks-based systems (or >similar >VME-based systems.) HI, I'm using Electronic Solutions cages and Digital's Storage works disk drives and racks. The cages are self jumpering... nice The disks are hot swappable real nice.. You can buy cdroms and tape units to plug into the storage works bays. I can create a system disk from a tape on a system on the fly and then move that disk to the target system without touching a power switch. All those tricky SCSI cables are out of the way... Look into it.. For a high reliability system ( having also spent months tracking down a bad SCSI cable as well ) this is what I need. Phil --------------------------- Newsgroups: comp.os.vxworks Subject: Problem: very long booting time Date: 16 Feb 1996 15:46:06 +1100 From: dmr@hercules.iassf.easams.com.au (Dewan RASHID) Organization: EASAMS (Australia) Pty Ltd Message-ID: <4g126e$j45@hercules.iassf.easams.com.au> Dear VxWorkers: In the file sysLib.c, if I change the number for the the some vme are (see below) to 0x90000000, the booting requires around 15 minutes of time. Is there any reason for such a long booting time? How can I avoid this delay without reducing the number (in this case 0x90000000). Any help will be appreciated. Thanks, Dewan Rashid. - ---------------------------------- PHYS_MEM_DESC sysPhysMemDesc [] = { /* adrs and length parameters must be page-aligned (multiples of 0x2000) */ /* ram */ { (void *) LOCAL_MEM_LOCAL_ADRS, (void *) LOCAL_MEM_LOCAL_ADRS, 0x2000000, /* 32 Mbytes (adjust if necessary) */ VM_STATE_MASK_VALID | VM_STATE_MASK_WRITABLE | VM_STATE_MASK_CACHEABLE, VM_STATE_VALID | VM_STATE_WRITABLE | VM_STATE_CACHEABLE }, /* some vme */ { (void *) 0x2000000, (void *) 0x2000000, 0x90000000, /* changed from 16 Mbytes : DMR */ VM_STATE_MASK_VALID | VM_STATE_MASK_WRITABLE | VM_STATE_MASK_CACHEABLE, VM_STATE_VALID | VM_STATE_WRITABLE | VM_STATE_CACHEABLE_NOT }, --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Programming the Bootrom Date: Wed, 14 Feb 1996 20:04:09 GMT From: krister@unidata.se (Krister Lagerstrom) Organization: Tripnet AB (modem +46 31 831683) Message-ID: <4ftend$1es@minox.tripnet.se> References: <4fq83r$a6v@lfsserv1.fsc.ibm.com> getz@lfs.loral.com wrote: >I'm trying to determine the proper way to program the bootrom for a >Force CPU-5CE which runs VxWorks 5.2. The 5CE Bootrom is in flash memory. >We need SCSI Boot support, thus the need to reprogram the bootrom. I've made >the changes to vxWorks for SCSI support and we can use the SCSI disk now. >There are BSP routines provided for writing files to flash, so I think I >understand how to do that. My question is which bootrom file is loaded to >flash (bootrom, bootrom.hex, bootrom.uncomp, ....) First, there are a couple of variants of the bootrom. The differences are whether or not it copies itself to RAM before execution, and if it should be stored in a compressed form. The standard is to decompress and copy to RAM. Next, the compiled bootrom is in the a.out format which most eprom programmers can't read, so the corresponding format is .hex, which is Motorola S-Records in your case. If the flash chips are socketed I strongly suggest that you remove them and use copies instead so you have some means to restore the board to its' original condition if something happens, like a brownout during programming. >Also, is there a good way to test changes to the bootrom code prior to >overwriting the current bootrom? I'm not familiar with your particular hardware, but if you have 2 megs of RAM you could try it from there first: o Compile a version of VxWorks that only knows about the first meg of RAM. o Compile a version of the bootrom that believes that the next meg of RAM is its' ROM area. Which version you use probably doesn't matter, but bootrom.rom_res (sp?) uses the least RAM. Make sure you change all definitions for this (MakeSkel, config.h, etc, see the manuals). o Convert the a.out file to raw data using aoutToBin. o Load to RAM. You could probably write a simple TCP app to do this from your workstation, or you could use NFS. Note: It should be loaded to start of high meg + 8 bytes, since the link phase assumes that the initial SP and PC vectors are stored there. o Jump to the start of the bootrom (eg 0x100008). That should be it! I'm writing this from home, so the details might be incorrect. Feel free to contact me if you have further questions. / Krister - --- Krister Lagerstrvm Email: krister@unidata.se UniData HB Tel: +46 31 20 13 07 Box 240 59 Fax: +46 31 20 58 10 S-400 22 Gvteborg --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Booting VxWorks from PC via ethernet Date: Fri, 16 Feb 1996 09:33:47 GMT From: rla@netcom.com Organization: NETCOM On-line Communication Services (408 261-4700 guest) Message-ID: References: <850819176wnr@test.demon.co.uk> <4fur0f$8q1@zinc.compulink.co.uk> <4g0j3q$q41@xmission.xmission.com> Sender: rla@netcom16.netcom.com In article <4g0j3q$q41@xmission.xmission.com>, bgeer wrote: >Chris Wallett writes: > > >You need to setup the boot parameters in the MVME167 to use FTP > >and not rsh by putting in a password. > > >You will also require TCP/IP and FTP server software for Windows. > >We use the SuperTCP package for WIndows 3.X but are moving to > >Windows NT as it includes TCP/IP and FTP Server as standard. > >(Windows 95 includes TCP/IP but not FTP Server) > >One of our guys found a Win95 compatible ftp daemon called FTPD. It's >a piece of shareware that seemed to work just fine as an ftp daemon. >Unfortunately it didn't solve our problem as we needed a *tftp* >daemon. > >An alternative we are looking at is bringing up Linux on a PC laptop. >Both Linux & MSDOG/Windoze [my prejudice] should coexist just fine on >a large enough disk. > >-- ><> Robert Geer & Donna Tomky / * <> ><> bgeer@xmission.com | _o * o * o <> ><> Salt Lake City, Utah | -¥<, * <¥ ><> U S A | O/ O __ /__, /> <> Putting in the password, boot uses FTP. Without a password, boot uses tftp or bootp depending on the flags (0x80 or 0x40). --------------------------- End of New-News digest ********************** From tkb@mclean.sparta.com Fri Feb 16 07:37:59 1996 From: Thomas Keith Buchanan Date: Fri Feb 16 07:38:01 PST 1996 Subject: Re: Programming the Bootrom > >I'm trying to determine the proper way to program the bootrom for a > >Force CPU-5CE which runs VxWorks 5.2. The 5CE Bootrom is in flash memory. ... > First, there are a couple of variants of the bootrom. The differences > are whether or not it copies itself to RAM before execution, and if it > should be stored in a compressed form. The standard is to decompress > and copy to RAM. I believe that the compression algorithms do not work for the SPARC architecture. The man pages and/or technical documentation from Force will tell you if they expect bootrom.hex, bootrom_uncmp or bootrom_uncmp with the a.out header stripped out ( a straight binary image ). If you are trying to include the application and the full kernel, you will use the standalone files though I forget the file names. > If the flash chips are socketed I strongly suggest that you remove > them and use copies instead so you have some means to restore the > board to its' original condition if something happens, like a brownout > during programming. Flash chips are rarely socketed, they are usually surface mounted flat packs. Do not try to remove them!!!! You're SPARC card can boot using it's OBP EPROM rather than the flash, correct? You can always boot via OBP and follow Force directions to reprogram the boot flash using OBP commands. There's no way you can lose the ability to recreate the system as shipped unless you erase OBP or alter the BSP w/o saving a copy. > >Also, is there a good way to test changes to the bootrom code prior to > >overwriting the current bootrom? The standard approach to such problems is to make the changes to the full kernel first. VxWorks actually boots twice, once from EPROM/Flash and again from the full kernel. The bootConfig and usrConfig files in the config/all directory are very similar. You can modify the boot sequence in usrConfig and assorted files, make sure it works and then alter bootConfig and it's companion files and program the boot flash. In your case, I think you can reprogram the flash from VxWorks or from OBP. > o Compile a version of VxWorks that only knows about the first meg of > RAM. ... Don't go this route, Force gave you the capability to program the Flash from OBP and you can test 90% of your proposed alterations by trying new kernels loaded from your boot host. This is how Wind River suggests you approach this problem and the method SPARTA teaches in it's intermediate training courses. I have done what you are trying to do with Force, GMS and Themis SPARC targets, it'll work... adios ______________________________________________________ | |¥ |Thomas Keith Buchanan Voice -> 703 448 1683 x233| | |tkb@mclean.sparta.com FAX -> 703 734 3323 | | |_____________________________________________________| | | | | | /¥¥ SPARTA, Inc. | | | ///¥¥ 7926 Jones Branch Drive | | | /////¥¥ Suite 900 | | | ///// ¥¥ McLean, VA 22102 | | | / ¥¥¥¥¥ ¥¥ 703-448-0120 | | | ¥ ¥¥¥¥¥ // | | | ¥ ////// | | | ¥////// SPARTA | | | ¥//// Pride In Performance | | | ¥// -------------------- | | |_____________________________________________________| | ¥______________________________________________________¥| From carlos@branet.unex.es Fri Feb 16 08:42:29 1996 From: "Carlos J. Garcia Orellana" Date: Fri Feb 16 08:42:32 PST 1996 Subject: Re:( no subject) Hello, I think you should try with ioTaskStdGet/ioTaskStdSet. I use it to do something like that. I send the out of ls() into a socket, after call ls(), I restart the default stdout, and it works fine. Bye.. Carlos Javier Garcia Orellana Dpto. Electronica. Universidad de Extremadura. Badajoz - SPAIN. e-mail: carlos@nernet.unex.es WWW: http://nernet.unex.es From jfd@rtp.co.uk Fri Feb 16 09:49:11 1996 From: jfd@rtp.co.uk (Julian Day) Date: Fri Feb 16 09:49:14 PST 1996 Subject: Re: 80386EX Board Support From npr@jach.hawaii.edu Fri Feb 16 11:31:51 1996 From: npr@jach.hawaii.edu (Nick Rees) Date: Fri Feb 16 11:31:54 PST 1996 Subject: Access fault I have a task that was suspended because of an access fault. However, I had a close look at the area of the access address and couldn't find anything wrong. I changed nothing and did a task resume on the task. The task resumed fine and has been running for the past 12 hours with the piece of code that generated the fault being called at 300 Hz. Can this be a software problem? If my understanding of an access fault is correct the task shouldn't have been able to resume, so I am reluctantly suspecting the hardware. Is this more likely? The CPU is a 68060 (recently upgraded because of a cache hit/cache miss problem with earlier 68060's). The TCB of the task after the suspension was: -> ti 0x72f404 NAME ENTRY TID PRI STATUS PC SP ERRNO DELAY ---------- ------------ -------- --- ---------- -------- -------- ------- ----- callback _callbackTas 72f404 40 SUSPEND 77dcb0 72f03c 0 0 stack: base 0x72f404 end 0x72ccf4 size 9664 high 1932 margin 7732 options: 0xc VX_DEALLOC_STACK VX_FP_TASK d0 = 0 d1 = 1 d2 = 764600 d3 = 72f0a4 d4 = 3 d5 = 0 d6 = 0 d7 = 0 a0 = 780538 a1 = 762ed0 a2 = 6a3008 a3 = 75eeec a4 = 762b28 a5 = 75fbe4 a6/fp = 72f064 a7/sp = 72f03c sr = 3000 pc = 77dcb0 fpcr = 0 fpsr = 4000008 fpiar = 780558 f0 = 0.00620161 f1 = 1.4652e-07 f2 = 86400 f3 = NaN f4 = 50129.4 f5 = NaN f6 = NaN f7 = NaN Access Fault Program Counter: 0x0077dcb0 Status Register: 0x3000 Access Address : 0x0077dcb0 Fault Status : 0x01468020 value = 0 = 0x0 Poking around near the place of the error gave: -> 0x77dcb0 value = 7855280 = 0x77dcb0 = _bc635GetTimespec + 0xc -> l bc635GetTimespec _bc635GetTimespec: 77dca4 4e56 ffe8 LINK .W A6,#0xffe8help 77dcac 242e 0008 MOVE .L (0x8,A6),D2 77dcb0 262e 000c MOVE .L (0xc,A6),D3 77dcb4 246e 0010 MOVEA .L (0x10,A6),A2 77dcb8 4a82 TST .L D2 77dcba 6776 BEQ 0x0077dd32 77dcbc b4b9 0078 10f0 CMP .L _pFirst,D2 77dcc2 670c BEQ 0x0077dcd0 77dcc4 2f02 MOVE .L D2,-(A7) value = 7855302 = 0x77dcc6 = _bc635GetTimespec + 0x22 Any help would be appreciated. Nick Rees Joint Astronomy Centre Ph: +1 (808) 961-3756 660 N. Aohoku Place Fax: +1 (808) 961-6516 Hilo, HI. 96720 Internet: npr@jach.hawaii.edu From DTHOMPSON@hqgate.orbital.com Fri Feb 16 12:54:19 1996 From: DTHOMPSON@hqgate.orbital.com Date: Fri Feb 16 12:54:22 PST 1996 Subject: RAM usage on 68302 I am running Tornado 1.0 on a Motorola ADS302 board. I am limited to a 128K RAM and 128K EPROM memory configuration. I have configured VxWorks to include the agent and minimal functionality which has resulted in 80K text image and approx. 15k of data and bss. When I create the image as vxworks_rom.hex (copies uncompressed image to RAM) the browser and the memShow utility indicate that VxWorks is using 155K of RAM. Where is the extra 60K of RAM coming from? I am running the exception task (8k of stack), the logTask (5K of stack) and the wdBtask (4k of stack). I am also using default values for NUM_DRIVERS (50) and NUM_FILES (50). This all may account for about 20K of RAM but I still can't account for the additional 40K. Is there a vxWorks configuration expert out there? Thanks for any help Dean Thompson dthompson@hqgate.orbital.com From ken@esd.ray.com Fri Feb 16 13:36:32 1996 From: ken@esd.ray.com (Ken Balmy) Date: Fri Feb 16 13:36:35 PST 1996 Subject: NTP on FDDI hello out there - has anybody experienced using NTP (or any similar time-synch'ing protocol) on a FDDI-based network? we need to sync up time stamps in our boxes to within 1 msec my concern is the potentially variable latency on the FDDI TIA ------------------------------------------------------------------ Ken Balmy Too fast to live E-Systems, Goleta Division Too young to die ken@esd.ray.com Bye, bye ^D B{= ------------------------------------------------------------------ From yfsiu@efi.com Fri Feb 16 14:02:35 1996 From: Yuet Fung Siu Date: Fri Feb 16 14:02:38 PST 1996 Subject: File systems for VxWorks We are currently running VxWorks 5.2 with the included DOS file system. We are now seeking methods to improve the performance of this file system or alternatives. We appreciate any informative reply. Thanks, YF From daemon@csg.lbl.gov Sat Feb 17 04:00:45 1996 From: daemon@csg.lbl.gov Date: Sat Feb 17 04:00:48 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sat Feb 17 04:00:41 PST 1996 Subject: Re: xntp found Subject: Re: Help : Task spawning given a string containing the entry function Subject: Re: Help : Task spawning given a string containing the entry function Subject: DMA with a FORCE CPU-44 Subject: Re: RPC and VxWorks Subject: Re: Booting VxWorks from PC via ethernet ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: xntp found Date: 16 Feb 1996 12:43:28 GMT From: duhrenho@vette.sanders.com (David Uhrenholdt) Organization: Lockheed Sanders Message-ID: References: In article duhrenho@vette.sanders.com (David Uhrenholdt) writes: > > Is there a port of xntp available for VxWorks? > Thanks to geger@dauntless.den.mmc.com I found that there is a reasonability recent port of ntp at ftp.atd.ucar.edu/pub/vxworks/vx. - -- Dave Uhrenholdt duhrenho@vette.sanders.lockheed.com Phone (603) 885-7398 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Help : Task spawning given a string containing the entry function Date: 13 Feb 1996 12:52:44 -0500 From: chaplin@dopey.keithley.com (Roger Chaplin) Organization: Keithley Instruments, Cleveland, Ohio Message-ID: <4fqj5c$fj@dopey.keithley.com> References: <4foi66$a54@f111.iassf.easams.com.au> Anthony MEREDITH (arm@f111.iassf.easams.com.au) wrote: : Could some please help me. I am writing a program to : load and spawn an excutable. The problem I need to : solve is how to convert the entry point given as : a string to the type required by taskSpawn(), : ie, FUNCPTR. Use symFindByName(). - -- Roger Chaplin / Test Instruments Group / chaplin_roger@keithley.com | "For the lines are long and the fighting is strong and they're breakin' | | down the distance between right and wrong." | | -- Bob Dylan, "Ring Them Bells" | --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Help : Task spawning given a string containing the entry function Date: 13 Feb 1996 12:56:31 -0500 From: chaplin@dopey.keithley.com (Roger Chaplin) Organization: Keithley Instruments, Cleveland, Ohio Message-ID: <4fqjcf$g0@dopey.keithley.com> References: <4foi66$a54@f111.iassf.easams.com.au> Anthony MEREDITH (arm@f111.iassf.easams.com.au) wrote: : Could some please help me. I am writing a program to : load and spawn an excutable. The problem I need to : solve is how to convert the entry point given as : a string to the type required by taskSpawn(), : ie, FUNCPTR. : arm@iassf.easams.com.au : Could someone please help me. I am writing a program to run on : vxworks that loads and spawns a requested excutable. The : problem I need to solve is how to convert the entry point : name which is a string to the type used by taskSpawn(), : ie, FUNCPTR. : arm@iassf.easams.com.au : Subject: Task spawning given a string containing the entry function : Summary: : Expires: : Sender: : Followup-To: : Distribution: : Organization: EASAMS (Australia) Pty Ltd : Keywords: : Cc: : Newsgroups: comp.os.vxworks : Subject: Help : Task spawning given a string as the entry function : Summary: : Expires: : Sender: : Followup-To: : Distribution: : Organization: EASAMS (Australia) Pty Ltd : Keywords: : Cc: - -- Roger Chaplin / Test Instruments Group / chaplin_roger@keithley.com | "For the lines are long and the fighting is strong and they're breakin' | | down the distance between right and wrong." | | -- Bob Dylan, "Ring Them Bells" | --------------------------- Newsgroups: comp.os.vxworks Subject: DMA with a FORCE CPU-44 Date: Wed, 14 Feb 1996 16:15:11 GMT From: wlynde@nmsu.edu (William Lynde) Organization: New Mexico State University Message-ID: <4fsuei$7nu@bubba.NMSU.Edu> I am a newcomer to VME and have a FORCE CPU-44 processor which I am trying to use to do a DMA transfer across the VMEbus. Does anyone out there have some C code to do this that they could send to me so that I could compare and see what I am doing wrong? Thanks in advance. Bill --------------------------- Newsgroups: comp.os.vxworks Subject: Re: RPC and VxWorks Date: 16 Feb 1996 20:50:51 -0800 From: pwatson@eworld.com (PWatson) Organization: eWorld Message-ID: <4g3mrb$p3h@hp5.online.apple.com> References: <9602052025.AA16269@lbl.gov> Reply-To: pwatson@eworld.com (PWatson) Sender: root@news-relay.eworld.com I have used RPC in VxWorks extensively. We don't use any third party tools to make it work, besides rpcgen supplied with our UNIX boxes. We have also integrated VxWorks RPC with DOS and Windows RPC (using FTP Software's development package). I am a big proponent of RPC. I just wish Wind River would make some enhancements to their implementation of it, like (mainly) supporting RPC broadcasts. We have also had some problems with TCP clients-- some of these are already documented problems. But anyhow, not quite sure what your asking. But just know that it works fine! Philip Watson Matsushita Avionics Systems Corp., Irvine, CA --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Booting VxWorks from PC via ethernet Date: 15 Feb 1996 17:28:42 -0700 From: bgeer@xmission.xmission.com (bgeer) Organization: XMission Internet (801 539 0900) Message-ID: <4g0j3q$q41@xmission.xmission.com> References: <850819176wnr@test.demon.co.uk> <4fur0f$8q1@zinc.compulink.co.uk> Chris Wallett writes: >You need to setup the boot parameters in the MVME167 to use FTP >and not rsh by putting in a password. >You will also require TCP/IP and FTP server software for Windows. >We use the SuperTCP package for WIndows 3.X but are moving to >Windows NT as it includes TCP/IP and FTP Server as standard. >(Windows 95 includes TCP/IP but not FTP Server) One of our guys found a Win95 compatible ftp daemon called FTPD. It's a piece of shareware that seemed to work just fine as an ftp daemon. Unfortunately it didn't solve our problem as we needed a *tftp* daemon. An alternative we are looking at is bringing up Linux on a PC laptop. Both Linux & MSDOG/Windoze [my prejudice] should coexist just fine on a large enough disk. - -- <> 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 tkb@mclean.sparta.com Sat Feb 17 06:56:42 1996 From: Thomas Keith Buchanan Date: Sat Feb 17 06:56:45 PST 1996 Subject: Re: File systems for VxWorks On Fri, 16 Feb 1996, the vxWorks Users Group Exploder wrote: > We are currently running VxWorks 5.2 with the included DOS file system. > We are now seeking methods to improve the performance of this file > system or alternatives. We appreciate any informative reply. You can improve your throughput by buffering your data in the application and doing reads/writes for larger chunks of data. If you really want to max out the SCSI capability, you will need to use the raw file system. It is a fairly common practice to have dosFs and rawFs partitions on the same physical device. You can write your data as fast as possible and store indexing information in the dosFs partition. adios ______________________________________________________ | |¥ |Thomas Keith Buchanan Voice -> 703 448 1683 x233| | |tkb@mclean.sparta.com FAX -> 703 734 3323 | | |_____________________________________________________| | | | | | /¥¥ SPARTA, Inc. | | | ///¥¥ 7926 Jones Branch Drive | | | /////¥¥ Suite 900 | | | ///// ¥¥ McLean, VA 22102 | | | / ¥¥¥¥¥ ¥¥ 703-448-0120 | | | ¥ ¥¥¥¥¥ // | | | ¥ ////// | | | ¥////// SPARTA | | | ¥//// Pride In Performance | | | ¥// -------------------- | | |_____________________________________________________| | ¥______________________________________________________¥| From daemon@csg.lbl.gov Sun Feb 18 04:00:32 1996 From: daemon@csg.lbl.gov Date: Sun Feb 18 04:00:36 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sun Feb 18 04:00:19 PST 1996 Subject: Re: RIP protocol under VxWorks Subject: Re: re: Help My Ballistic Coworker Subject: Re: re: Help My Ballistic Coworker Subject: Solaris 2.5 as a Tornado host Subject: Sorry about double post. Lost finger/mouse coord. Subject: Help : Task spawning given a string containing the entry function Subject: Task spawning given a string containing the entry function Subject: Help : Task spawning given a string as the entry function Subject: Re: Help : Task spawning given a string containing the entry function ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: RIP protocol under VxWorks Date: 12 Feb 1996 22:00:21 GMT From: witts@PROBLEM_WITH_INEWS_GATEWAY_FILE (Steve Witt) Organization: Hughes Aircraft Company Message-ID: <4fod9l$q7v@venice.sedd.trw.com> References: <9602090840.AA08927@inrete.it> Reply-To: sawitt@hac2arpa.hac.com Federico De Sario (erxa@inrete.it) wrote: > I'm looking for a RIP protocol implementation under VxWorks. > Is there anyone who has it available or has implemented it? > Any help you can give at this subject will be much appreciated! > Thx in advance. Yes, there is one at the VxWorks Archive that I've used. It implements RIP version 1 I believe. You can find it at: ftp://ftp.atd.ucar.edu/pub/vxworks/vx and the files are: vxrouted01 - vxrouted08 - --------------------------------------------------------------------- Steve Witt Hughes Aircraft Company sawitt@hac2arpa.hac.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: re: Help My Ballistic Coworker Date: 17 Feb 1996 18:39:03 GMT From: "Dennis R. Perlman" Organization: Lockheed Martin -- Denver InterNetNews site Message-ID: <4g57c8$mv9@tel.den.mmc.com> References: Thomas Keith Buchanan wrote: >> Why can't you map in all 4GB? With 32bit page entries, each mapping a 0x2000 >> byte page, it would only take 1MB for the page tables. Adding in the overhead >> for the pointer tables wouldn't add much more. > >Because 2^32 = 4Gb... You can't map it all in unless your CPU has and >address space greater than 32 bits, there would be no room left for local Actually, " Why can't you map in all 4GB? " was sorta a rhetorical question. I was merely wondering what was, if anything, preventing me from mapping in as much memory as I wanted. I'm aware that the entire 4G space is shared by local and external devices. The biggest contraints (as far as I know) to mapping in a lot of memory is, one, it takes up about a meg for the PDE's (assuming 32bit PDE maps 4k page and 4G mapped) and two, the time to initialize the PDEs. >memory. Having said that, aren't you guys using an older SPARC target? I don't know where this came from. What we are using is M680x0s. We have some targets that are M68040 based and some that are M68060. The 060 takes about 20 seconds to build PDEs for 256M. >You need to query Wind River >about the capabilities of your MMU and cache libraries with your unique >target. > Have, but haven't received any response back yet? What we've asked is, what is the affect on vxWorks if you turn off address translation? It seems to me that since, for the targets we are using, the physical address and the virtual address are the same, about the only thing we getting out of the address translation mechanism is the ability to mark, via the PE's, areas of memory as cacheable or non-cacheable. Since we are mostly concerned with memory that must be non-cached, we could control caching through the CACR. (cache control register) and blow-off the overhead for address translation. > ********************************************************** * Dennis R. Perlman * * Lockheed Martin Astronautics * * (303) 977-8891 * * Email: perlman@den.mmc.com * ********************************************************** --------------------------- Newsgroups: comp.os.vxworks Subject: Re: re: Help My Ballistic Coworker Date: 17 Feb 1996 18:39:34 GMT From: "Dennis R. Perlman" Organization: Lockheed Martin -- Denver InterNetNews site Message-ID: <4g57d6$mv9@tel.den.mmc.com> References: Thomas Keith Buchanan wrote: >> Why can't you map in all 4GB? With 32bit page entries, each mapping a 0x2000 >> byte page, it would only take 1MB for the page tables. Adding in the overhead >> for the pointer tables wouldn't add much more. > >Because 2^32 = 4Gb... You can't map it all in unless your CPU has and >address space greater than 32 bits, there would be no room left for local Actually, " Why can't you map in all 4GB? " was sorta a rhetorical question. I was merely wondering what was, if anything, preventing me from mapping in as much memory as I wanted. I'm aware that the entire 4G space is shared by local and external devices. The biggest contraints (as far as I know) to mapping in a lot of memory is, one, it takes up about a meg for the PDE's (assuming 32bit PDE maps 4k page and 4G mapped) and two, the time to initialize the PDEs. >memory. Having said that, aren't you guys using an older SPARC target? I don't know where this came from. What we are using is M680x0s. We have some targets that are M68040 based and some that are M68060. The 060 takes about 20 seconds to build PDEs for 256M. >You need to query Wind River >about the capabilities of your MMU and cache libraries with your unique >target. > Have, but haven't received any response back yet? What we've asked is, what is the affect on vxWorks if you turn off address translation? It seems to me that since, for the targets we are using, the physical address and the virtual address are the same, about the only thing we getting out of the address translation mechanism is the ability to mark, via the PE's, areas of memory as cacheable or non-cacheable. Since we are mostly concerned with memory that must be non-cached, we could control caching through the CACR. (cache control register) and blow-off the overhead for address translation. > ********************************************************** * Dennis R. Perlman * * Lockheed Martin Astronautics * * (303) 977-8891 * * Email: perlman@den.mmc.com * ********************************************************** --------------------------- Newsgroups: comp.os.vxworks Subject: Solaris 2.5 as a Tornado host Date: Sat, 17 Feb 1996 19:25:01 GMT From: ggerard@onramp.net (Greg Gerard) Organization: Salve, Inc. Message-ID: <4g5a43$fip@news.onramp.net> Does the host environment for Tornado work under Solaris 2.5? thanks, greg --------------------------- Newsgroups: comp.os.vxworks Subject: Sorry about double post. Lost finger/mouse coord. Date: 17 Feb 1996 18:55:21 GMT From: "Dennis R. Perlman" Organization: Lockheed Martin -- Denver InterNetNews site Message-ID: <4g58ap$mv9@tel.den.mmc.com> References: <4g57d6$mv9@tel.den.mmc.com> Sorry about double post. Lost finger/mouse coord. --------------------------- Newsgroups: comp.os.vxworks Subject: Help : Task spawning given a string containing the entry function Date: 13 Feb 1996 10:23:50 +1100 From: arm@f111.iassf.easams.com.au (Anthony MEREDITH) Organization: EASAMS (Australia) Pty Ltd Message-ID: <4foi66$a54@f111.iassf.easams.com.au> Could some please help me. I am writing a program to load and spawn an excutable. The problem I need to solve is how to convert the entry point given as a string to the type required by taskSpawn(), ie, FUNCPTR. arm@iassf.easams.com.au Could someone please help me. I am writing a program to run on vxworks that loads and spawns a requested excutable. The problem I need to solve is how to convert the entry point name which is a string to the type used by taskSpawn(), ie, FUNCPTR. arm@iassf.easams.com.au Subject: Task spawning given a string containing the entry function Summary: Expires: Sender: Followup-To: Distribution: Organization: EASAMS (Australia) Pty Ltd Keywords: Cc: Newsgroups: comp.os.vxworks Subject: Help : Task spawning given a string as the entry function Summary: Expires: Sender: Followup-To: Distribution: Organization: EASAMS (Australia) Pty Ltd Keywords: Cc: --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Help : Task spawning given a string containing the entry function Date: 13 Feb 1996 11:56:20 -0800 From: kla@leland.Stanford.EDU (Earl Mitchell) Organization: Stanford University, CA 94305, USA Message-ID: <4fqqd4$9rc@elaine25.Stanford.EDU> References: <4foi66$a54@f111.iassf.easams.com.au> In article <4foi66$a54@f111.iassf.easams.com.au>, Anthony MEREDITH wrote: >Could some please help me. I am writing a program to >load and spawn an excutable. The problem I need to >solve is how to convert the entry point given as >a string to the type required by taskSpawn(), >ie, FUNCPTR. Use the symbol table facilities to map the function name (string) to the function's entry point (address) which you can then pass as an arg to taskSpawn. For example you could use "lkup" command or something like "symFindByName". Checkout symLib. --------------------------- End of New-News digest ********************** From tkb@mclean.sparta.com Sun Feb 18 16:11:06 1996 From: Thomas Keith Buchanan Date: Sun Feb 18 16:11:09 PST 1996 Subject: re: Help My Ballistic Coworker On Sat, 17 Feb 1996, Dennis R. Perlman wrote: > Actually, " Why can't you map in all 4GB? " was sorta a rhetorical > question. I was merely wondering what was, if anything, preventing > me from mapping in as much memory as I wanted. I'm aware that the > entire 4G space is shared by local and external devices. The biggest > contraints (as far as I know) to mapping in a lot of memory is, one, it takes > up about a meg for the PDE's (assuming 32bit PDE maps 4k page and 4G mapped) I have heard rumors that the VxWorks MMU libraries have some limits on the amount of virtual memory that can be mapped. At least in the SPARC world, I know they don't try and optimize the tables. They just create level 3 tables for all areas. > I don't know where this came from. What we are using is M680x0s. We have > some targets that are M68040 based and some that are M68060. The 060 takes > about 20 seconds to build PDEs for 256M. sorry, I thought I read something in this same thread about a SPARC target. > It seems to me that since, for the targets we are using, the physical address > and the virtual address are the same, about the only thing we getting out of the > address translation mechanism is the ability to mark, via the PE's, areas > of memory as cacheable or non-cacheable. Since we are mostly concerned with > memory that must be non-cached, we could control caching through the CACR. > (cache control register) and blow-off the overhead for address translation. Address translation doesn't usually introduce "overhead", it's normally an integrated piece of the hardware that keeps the instruction and data pipelines full. You trade much higher average throughput for the loss of determinism due to cache hits/misses. In most cases, it's a good trade... If you don't need to use the caches or are concerned about determinism, just turn off the MMU and cache stuff. You can do that by commenting out or undefining INCLUDE_MMU_BASIC and INCLUDE_CACHE_SUPPORT. You had better check and see why your project chose the 68060 in the first place because your average performance will be much much lower w/o the caches enabled. Perhaps you should just turn off the data caches or tweak the sysPhysMemDesc tables. adios ______________________________________________________ | |¥ |Thomas Keith Buchanan Voice -> 703 448 1683 x233| | |tkb@mclean.sparta.com FAX -> 703 734 3323 | | |_____________________________________________________| | | | | | /¥¥ SPARTA, Inc. | | | ///¥¥ 7926 Jones Branch Drive | | | /////¥¥ Suite 900 | | | ///// ¥¥ McLean, VA 22102 | | | / ¥¥¥¥¥ ¥¥ 703-448-0120 | | | ¥ ¥¥¥¥¥ // | | | ¥ ////// | | | ¥////// SPARTA | | | ¥//// Pride In Performance | | | ¥// -------------------- | | |_____________________________________________________| | ¥______________________________________________________¥| From daemon@csg.lbl.gov Mon Feb 19 04:01:27 1996 From: daemon@csg.lbl.gov Date: Mon Feb 19 04:01:30 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Mon Feb 19 04:01:24 PST 1996 Subject: Re: Access fault Subject: Re: Booting VxWorks from PC via ethernet Subject: Seeking recommnedations for VME cage, disk drive vendors (second try) Subject: PCI bus/ PC ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Access fault Date: 18 Feb 1996 17:45:53 GMT From: Fabrice Cancre Organization: Sympatico Message-ID: <4g7okh$hbs@news1.sympatico.ca> References: <9602161932.AA10154@pau.jach.hawaii.edu.jach.hawaii.edu> Re: Access fault I cannot help you about this subject but I also have problems with a MVME177 board (including a 68060) with cache. I recently received a patch from WindRiver which is supposed to fix most of the problems. Actually, it does not in my case but it could help you. By the way, if you were a former user of a 68040 board, have you noticed a significant performance increase ? I personnaly do not Reply to Fabrice Cancre R/D Tech 1200 saint Jean Baptiste Quebec, G2E 5E8 Canada Tel : +1 (418)872-1155 Fax : +1 (418)872-5431 E-Mail : 73323.1612@compuserve.com Home E-Mail : famille.cancre@sympatico.ca --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Booting VxWorks from PC via ethernet Date: 18 Feb 1996 17:38:41 GMT From: Fabrice Cancre Organization: Sympatico Message-ID: <4g7o71$hbs@news1.sympatico.ca> References: <850819176wnr@test.demon.co.uk> <4fur0f$8q1@zinc.compulink.co.uk> Re: Booting VxWorks from PC via ethernet I use either a UNIX workstation or a Windows/NT workstation to boot my system (MVME 167 + other boards on a VME chassis booting using shared memory). Everything goes perfect. I had to add a user to my Windows/NT workstation and I load kernel using FTP and code using NFS. Unfortunatly, I cannot reboot the system faster than every 10 minutes. The FTP server will not reply nor return any error code. Has anybody experienced the same kind of problem ? In Windows NT 3.51, there is a parameter in the FTP server which is set to 10 minutes but even if I set it to 1 minute it will not change anaything Thanks Reply to Fabrice Cancre R/D Tech 1200 saint Jean Baptiste Quebec, G2E 5E8 Canada Tel : +1 (418)872-1155 Fax : +1 (418)872-5431 E-Mail : 73323.1612@compuserve.com Home E-Mail : famille.cancre@sympatico.ca --------------------------- Newsgroups: comp.os.vxworks,comp.arch.bus.vme Subject: Seeking recommnedations for VME cage, disk drive vendors (second try) Date: 14 Feb 1996 20:07:15 GMT From: Lawrence Pfeffer Organization: NIST Message-ID: <4ftfdj$ke@dove.nist.gov> Greetings all, Apologies for the prev msg without subject line. It's one of those days. I'm looking for recommendations for VME cage vendors and also disk drives that people have had good experiences using in VxWorks-based systems (or similar VME-based systems.) For local disks, I'm more interested in reliability/ease of integration than amazing speed/capacity. U.S. based vendors somewhat preferred. Please let me know what you've used and liked (or, alternately, things to avoid.) Email responses probably best, since other newgroup readers may not be interested. However, if there is interest on either group, let me know, and I'll post a summary. Thanks in advance, - -- Lawrence E. Pfeffer, Ph.D. (AKA Larry), voice: (301)975-6067, FAX: (301)869-6275 *Preferred* email: lawrence.pfeffer@nist.gov U.S. mail: NIST, Bldg. 226/Rm. B158, Gaithersburg, MD 20899 --------------------------- Newsgroups: comp.os.vxworks Subject: PCI bus/ PC Date: Wed, 14 Feb 1996 14:43:40 -0800 From: peter Barrie Organization: University of Strathclyde Computer Centre Message-ID: <3122659C.4339@cs.strath.ac.uk> Has anybody built any drivers for PCI peripherals on a VxWorks PC platform ? Is this possible , difficult or downright foolish ? Pete Barrie --------------------------- End of New-News digest ********************** From leonid@rst.co.il Mon Feb 19 09:45:27 1996 From: leonid@rst.co.il (Leonid Rosenboim) Date: Mon Feb 19 09:45:29 PST 1996 Subject: 100baseT Does anyone know of a working solution for 100BaseT fast ethernet for VxWorks with either SBus or VMEbus interface ? i.e. a card with a VxWorks driver for any of the above buses ? I have asked this question before with no results so far. ----------------------------------------------------------------------- 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 ftp://knight.trendline.co.il/pub/rst E-Mail: leonid@rst.co.il From @lazarus.nrtc.northrop.com:wong@csd.nad.northrop.com Mon Feb 19 18:35:10 1996 From: "J. Pierre Wong" Date: Mon Feb 19 18:35:13 PST 1996 Subject: checksum usage How do you implement a checksum on a vxWorks program that is install on an eeprom? Is the computation of the checksum just on the text section of the object code, or does it include the bss and the data sections? Also, where do you normally store the checksum value for comparision during run-time. Pierre Wong Northrop Grumman Corporation tel: 310.332.6840 fax: 310.644.3652 email: wong@csd.nad.northrop.com From daemon@csg.lbl.gov Tue Feb 20 04:00:28 1996 From: daemon@csg.lbl.gov Date: Tue Feb 20 04:13:52 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Tue Feb 20 04:00:24 PST 1996 Subject: Re: Token Ring adapter driver Subject: Booting VxWorks from PC via ethernet Subject: IP memory access Subject: network benchmark apps for Vxworks ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Token Ring adapter driver Date: 16 Feb 1996 20:54:10 -0800 From: pwatson@eworld.com (PWatson) Organization: eWorld Message-ID: <4g3n1i$p3p@hp5.online.apple.com> References: <960208204042_217598869@emout05.mail.aol.com> Reply-To: pwatson@eworld.com (PWatson) Sender: root@news-relay.eworld.com Well, we did our own. We're not normally in the business of licensing software, but maybe you could convince us. I believe the chip we used what the TI 380. Philip Watson Matsushita Avionics Systems Corp., Irvine, CA --------------------------- Newsgroups: comp.os.vxworks Subject: Booting VxWorks from PC via ethernet Date: Wed, 14 Feb 1996 23:17:45 GMT From: Derron John Roberts Organization: Test Technologies Limited Message-ID: <850819176wnr@test.demon.co.uk> Reply-To: derron@test.demon.co.uk I am using a MVME167 single board computer which boots up vxworks from a Sun station via the ethernet. Any ideas how I can perform this boot up using a PC running DOS & Windows for Workgroups. Studying the ethernet traffic seems to indicate the boot sequence is using rsh commands to retrive the boot file. Any help greatly appreciated. - ------------------- Derron Roberts Email derron@test.demon.co.uk --------------------------- Newsgroups: comp.os.vxworks Subject: IP memory access Date: 19 Feb 1996 13:51:03 GMT From: terjeh@fof.nft.no (Terje Hansen) Organization: NFT Kongsberg, Norway Message-ID: <4g9v87$hg7@kongsberg.fof.nft.no> Have anyone experience using vxworks and IP modules on the MVME162 board (MC68LC040) or MVME162FX? There seems to be a problem (bus error or corrupted data read) accessing the IP memory area. We are using vxworks 5.2, mv162lc BSP and IP-SUI (Interray) with no DMA involved. From sysLib.c: /* ip module : SUI */ { (void *) 0x04000000, (void *) 0x04000000, 0x10000, /* 64 Kbytes */ VM_STATE_MASK_VALID | VM_STATE_MASK_WRITABLE | VM_STATE_MASK_CACHEABLE, VM_STATE_VALID | VM_STATE_WRITABLE | VM_STATE_CACHEABLE_NOT }, - ---- /* set up ip2-chip */ *IP2CHIP_IP_RESET = 0x01; delay(200) /* 200ms */ *IP2CHIP_IP_RESET = 0x00; /* Set base adr. for the SUI IP module = 0x04000000 * with size = 64KB (memory range= 0x0400000-0x0401ffff */ *IP2CHIP_IP_bBASE_U = 0x04; *IP2CHIP_IP_bBASE_L = 0x00; *IP2CHIP_IP_bSIZE = SIZE_64KB; /* Set up general control reg. */ *IP2CHIP_IP_bGEN_CTRL = 0x00; *IP2CHIP_IP_bGEN_CTRL = ( GEN_CTRL_MEN | GEN_CTRL_WIDTH_16 | GEN_CTRL_RT_2US ); Thanks, Terje Hansen --------------------------- Newsgroups: comp.os.vxworks Subject: network benchmark apps for Vxworks Date: 19 Feb 1996 18:57:54 GMT From: cjs@bailey.com (Charles J. Stack) Organization: Bailey Controls Company Message-ID: <4gah7i$c5a@spider.bailey.com> Reply-To: cjs@bailey.com Looking for Vxworks implementations of network/protocol benchmarking applications (such as TTCP, Netperf). Any recommendations? --------------------------- End of New-News digest ********************** From wynia@newkla.kla.com Tue Feb 20 07:46:32 1996 From: wynia@newkla.kla.com Date: Tue Feb 20 07:46:35 PST 1996 Subject: Console Monitor Does anyone have any information about how much of a performance hit a FORCE board takes when a serial Console line is attached to the board? And is this performance more or less than when an ethernet "xterm -e rsh name.domain" type console session is used? jw. -------------------------- Jim Wynia -------------------------- KLA Instruments [wynia@newkla.kla.com] address:160 Rio Robles P.O.Box 49055 direct dial:408-468-5364 San Jose, CA 95161-9055 fax:408-434-2843 From leonid@rst.co.il Tue Feb 20 10:30:50 1996 From: leonid@rst.co.il (Leonid Rosenboim) Date: Tue Feb 20 10:30:53 PST 1996 Subject: Re: checksum usage > From: "J. Pierre Wong" > Date: Mon Feb 19 18:35:13 PST 1996 > Subject: checksum usage > How do you implement a checksum on a vxWorks program that is install > on an eeprom? Is the computation of the checksum just on the text > section of the object code, or does it include the bss and the data > sections? Also, where do you normally store the checksum value for > omparision during run-time. We done a similar thing for one of our customer here: We compute the checksum for the text only, at any time during system operation, and we recover the string containing date and time this image was created from the variable char *creationDate ; We keep a copy of the creation date and the checksum in NVRAM. If during checking, the creation date differs from the one in NVRAM, we silently replace both the cheksum and date in NVRAM. If however the creation date match but the checksum does not, this means that either the image on the file was corrupted, an I/O error occured during loading or code memory was corrupted, and the system is rebooted with the appropriate message. During the same consulting effort we discovered that VxWorks already contains a similar functionality for dynamically loadable modules, just call moduleCheck() and it checks the text checksums against the checksum recorded when the module was first loaded. You should not attempt to do a checksum on the data during system operation, cause data contains variables that change during normal operation. ----------------------------------------------------------------------- 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 ftp://knight.trendline.co.il/pub/rst E-Mail: leonid@rst.co.il From phil@nevis.naic.edu Tue Feb 20 11:08:09 1996 From: phil@nevis.naic.edu Date: Tue Feb 20 11:08:12 PST 1996 Subject: RE: IP memory access Terje Hansen writes: -------------------- Have anyone experience using vxworks and IP modules on the MVME162 board (MC68LC040) or MVME162FX? There seems to be a problem (bus error or corrupted data read) accessing the IP memory area. We are using vxworks 5.2, mv162lc BSP and IP-SUI (Interray) with no DMA involved. >From sysLib.c: /* ip module : SUI */ { (void *) 0x04000000, (void *) 0x04000000, 0x10000, /* 64 Kbytes */ VM_STATE_MASK_VALID | VM_STATE_MASK_WRITABLE | VM_STATE_MASK_CACHEABLE, VM_STATE_VALID | VM_STATE_WRITABLE | VM_STATE_CACHEABLE_NOT }, - ---- /* set up ip2-chip */ *IP2CHIP_IP_RESET = 0x01; delay(200) /* 200ms */ *IP2CHIP_IP_RESET = 0x00; /* Set base adr. for the SUI IP module = 0x04000000 * with size = 64KB (memory range= 0x0400000-0x0401ffff */ *IP2CHIP_IP_bBASE_U = 0x04; *IP2CHIP_IP_bBASE_L = 0x00; *IP2CHIP_IP_bSIZE = SIZE_64KB; /* Set up general control reg. */ *IP2CHIP_IP_bGEN_CTRL = 0x00; *IP2CHIP_IP_bGEN_CTRL = ( GEN_CTRL_MEN | GEN_CTRL_WIDTH_16 | GEN_CTRL_RT_2US ); ----------------------------- I don't know whether this is all the initialization you've done to the ip board. On pg 4-26 of my mv162fx programmers ref. guide there is a note that says: "when programmming b_width1-bwidth0 for either 8 bits or 16 bits, a_width1-awidth0 must be programmed for one of 8-bits or 16-bits." If you don't program the width of ip_a, the 0,0 width defaults to 32 bits and conflicts with ip_b. I'm not sure this is your problem, but it's worth checking. phil perillat arecibo observatory phil@naic.edu From jromine@btsslc.com Tue Feb 20 13:14:14 1996 From: JEFF ROMINE Date: Tue Feb 20 13:14:16 PST 1996 Subject: Dynamic Linking - replacing symbols Thank you to everyone who responded to my question on RPC and VxWorks. My next question has to do with dynamic linking. What happens when a module is loaded which has symbols identical to symbols which are already in the symbol table? The reason I ask is that I want to replace certain functions which are part of the VxWorks system image for all modules which are subsequently loaded. I looked at the man page for the "ld" function. It has the option of loading all,global only, or no symbols into the symbol table, but it doesn't have an option which causes the symbols in the module being loaded to replace those already there. I know that I could do this by explicitly modifying the symbol table, but that would be awkward. -Jeff Romine, Broadcast Television Systems, jromine@btsslc.com From daemon@csg.lbl.gov Wed Feb 21 04:00:39 1996 From: daemon@csg.lbl.gov Date: Wed Feb 21 04:13:52 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Wed Feb 21 04:00:35 PST 1996 Subject: Re: Tcl7.4 for vxWorks Subject: Re: Help : Task spawning given a string containing the entry function Subject: Help! Rescheduling with Time limits Subject: Request: Makefile for 5.3 sio archives Subject: Re: RPC and VxWorks ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Tcl7.4 for vxWorks Date: 20 Feb 1996 13:28:33 GMT From: Carl Timmer Organization: CEBAF Message-ID: <4gcia1$a0l@cebaf4.cebaf.gov> For those interested, I've completed a port of Tcl7.4 to vxWorks (based on the work of others). You can find it at ftp://ftp.atd.ucar.edu/pub/vxworks/vx/vxTcl7.4.tar.Z Have fun, Carl --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Help : Task spawning given a string containing the entry function Date: 19 Feb 1996 19:04:47 GMT From: msinger@griffon.oe.fau.edu (Matthew Singer) Organization: Florida Atlantic University Message-ID: <4gahkf$pke@cybernet.cse.fau.edu> References: <4foi66$a54@f111.iassf.easams.com.au> Reply-To: msinger@griffon.oe.fau.edu (Matthew Singer) Sender: msinger@griffon (Matthew Singer) In article <4foi66$a54@f111.iassf.easams.com.au>, arm@f111.iassf.easams.com.au (Anthony MEREDITH) writes: |>Could some please help me. I am writing a program to |>load and spawn an excutable. The problem I need to |>solve is how to convert the entry point given as |>a string to the type required by taskSpawn(), |>ie, FUNCPTR. |> |>arm@iassf.easams.com.au |> |> |>Could someone please help me. I am writing a program to run on |>vxworks that loads and spawns a requested excutable. The |>problem I need to solve is how to convert the entry point |>name which is a string to the type used by taskSpawn(), |>ie, FUNCPTR. |> extern SYMTAB_ID sysSymTbl; int (*function)(); SYM_TYPE type; symFindByName(sysSymTbl, "_function_name", (char **)&function, &type); function(); - ------------------------------------------------------------------------------- Matthew R. Singer msinger@oe.fau.edu Systems Engineer Voice: 407-367-2947 Florida Atlantic University Fax : 407-367-3885 Dept of Ocean Engineering - ------------------------------------------------------------------------------- --------------------------- Newsgroups: comp.os.vxworks Subject: Help! Rescheduling with Time limits Date: 20 Feb 1996 15:34:46 GMT From: John Neumoege Organization: Fraunhofer IPK, Berlin Message-ID: <4gcpmm$cln@charly.ipk.fhg.de> I need to programm a controller with two clocks of 8ms and 64ms. Can anyone give me an idea as to how I can assure that the controller actually cycles every 8ms. To this point I have used vrtx real time os. On this system I set the clock for 8ms and if a task was finished before the 8ms, I let it "sleep" until the next cycle, where the task gives up its processor time. Is there something like this in vxWorks? I have found a routine sysDelay which seems to do this, but for our BSP, I80486, the parameters are given as'void'. Are there any other simple methods? Thanks --------------------------- Newsgroups: comp.os.vxworks Subject: Request: Makefile for 5.3 sio archives Date: 20 Feb 1996 22:29:28 GMT From: friedl@cher.heurikon.com (Ted Friedl) Organization: Heurikon Corporation Message-ID: <4gdi08$3t9@badger.heurikon.com> 5.3 developers, I would like to create a Makefile for a 5.3-style serial driver I just wrote. Wind River provides no examples in their Tornado distribution or documentation. My guess is that there are a few examples floating around on Wind River-distributed 5.3 compatible BSPs. Would anyone with a "target/src/drv/sio/Makefile*" file please email me a copy? Much thanks, Ted Friedl Senior Software Engineer Heurikon Corporation Madison, WI, USA --------------------------- Newsgroups: comp.os.vxworks Subject: Re: RPC and VxWorks Date: Thu, 8 Feb 1996 04:44:37 GMT From: noblenet@world.std.com (NobleNet Inc.) Organization: NobleNet, Inc. Message-ID: References: <9602052025.AA16269@lbl.gov> Followup-To: poster In article <9602052025.AA16269@lbl.gov>, JEFF ROMINE wrote: > >I would like to hear from anyone who has used RPC under VxWorks, especially >if they've overcome some difficulties. I'm using VxWorks 5.2 on the >M68360. If anyone thinks they can help, I'd be glad to send more details >etc. Thank You. > 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 http://www.noblenet.com/ --------------------------- End of New-News digest ********************** From prb@aplexus.jhuapl.edu Wed Feb 21 12:48:31 1996 From: prb@aplexus.jhuapl.edu (Paul R. Bade) Date: Wed Feb 21 12:48:34 PST 1996 Subject: Setting Up a proxy Client as a gateway to another subnet Hi, Has anyone successfully implemented the multi-tier proxy network shown in Figure 6-18 of the vxWorks 5.2 reference manual? If so, have you been able to send datagrams from the proxy server to the secondary network. In our setup it appears that the datagrams from the proxy server get sent to the main network instead of the proxy client that is serving as the gateway to the secondary network. It seems that routeAdd() can't figure out how to correctly update the routing tables. In our system, the proxy server on the main network is 128.244.69.200. The proxy sever on the sm network is 128.244.69.201. The proxy client on the sm network is 128.244.69.206. And the proxy client on the secondary network for which it is a gateway is 128.244.79.209. We have INCLUDE_SM_SEQ_ADDR included. On the proxy server, we find: ROUTE NET TABLE destination gateway flags Refcnt Use Interface ------------------------------------------------------------------------ 128.244.69.0 128.244.69.200 1 6 8385 ei0 128.244.79.0 128.244.69.206 3 2 1014 ei0 ------------------------------------------------------------------------ **** Note above that the 206 interface is ei0, but should be sm0 *** thus the packets get sent incorrectly to ei0 ROUTE HOST TABLE destination gateway flags Refcnt Use Interface ------------------------------------------------------------------------ 127.0.0.1 127.0.0.1 5 0 0 lo0 128.244.69.202 128.244.69.201 5 0 989 sm0 128.244.69.203 128.244.69.201 5 0 1210 sm0 128.244.69.204 128.244.69.201 5 0 1411 sm0 128.244.69.205 128.244.69.201 5 0 1432 sm0 128.244.69.206 128.244.69.201 5 0 2096 sm0 128.244.69.207 128.244.69.201 5 0 920 sm0 ------------------------------------------------------------------------ Thanks, +====================================================================+ | __ ____ __ __ | | /¥ ¥ / ¥ ¥ /¥_¥__ /¥ ¥ Johns Hopkins University | | / ¥_¥ | /¥ ¥ ¥ / / /¥_¥¥ ¥ ¥ Applied Physics Lab. | | / /¥ | |¥ ¥/ ¥ ¥ / / / / / ¥ ¥ ¥ | | / ¥/ |_| ¥ /¥ ¥_¥ / / / / / ¥ ¥ ¥ Paul R. Bade | | / /¥__/_/ ¥ ¥ ¥/_// / / / / / / / (301)-953-6000 x8681 | | / / / ¥ ¥_¥ ¥ ¥/ / / / / / prb@aplexus.jhuapl.edu | | ¥/_/ ¥/_/ ¥__/_/ ¥/_/ | | __ ____ __ __ | | /¥ ¥ / ¥ ¥ /¥ ¥ /¥ ¥ | | / ¥_¥ | /¥ ¥ ¥ / ¥_¥ / ¥ ¥ | | / /¥ | |¥ ¥/ ¥ ¥ / /¥ |_| / /¥ ¥ ¥ | | / ¥/ |_| ¥ /¥ ¥_¥ / / / | | ¥/ / ¥ ¥ | | / /¥ / / ¥ ¥ ¥/_// / / / / / /¥ ¥_¥ | | ¥ ¥/ / / ¥ ¥_¥ ¥ ¥/ / / ¥/ / / / | | ¥__/_/ ¥/_/ ¥__/_/ /_/_/ | | | +====================================================================+ From punamia@newkla.kla.com Wed Feb 21 16:09:33 1996 From: punamia@newkla.kla.com Date: Wed Feb 21 16:13:54 PST 1996 Subject: port from 5.0.2 to 5.2 Hi VxWorkers, Has anybody ported an application from VxWorks version 5.0.2 to 5.2 or 5.3 ? How much work is needed ? Does WRS provide BSP for M68020 Force Board for VxWorks 5.2/5.3 ? Thanks, Manoj Punamia e-mail: m_punami@kla.com phone: (408)456-6537 KLA Instruments Corp. 160 Rio Robles San Jose CA 94566 From daemon@csg.lbl.gov Thu Feb 22 04:00:31 1996 From: daemon@csg.lbl.gov Date: Thu Feb 22 04:14:11 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Thu Feb 22 04:00:27 PST 1996 Subject: Why Tornado? Subject: Serial port H/W handshaking? Subject: Test -- Please Answer! Subject: Maintaining multiple makefiles Subject: VxWindows and Tornado Subject: NTP on VxWorks (how?) Subject: employment opportunity Subject: competitors of VxWorks Subject: TORNADO for win95 Subject: Re: TORNADO for win95 ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Why Tornado? Date: 20 Feb 1996 15:08:10 GMT From: duhrenho@vette.sanders.com (David Uhrenholdt) Organization: Lockheed Sanders Message-ID: Has anyone (besides Wind River) comeup with a list of reasons for using Tornado? From reading the WRS Web page on it, it appears that are a couple of tools added and a bunch of tools/capabilities that are already available. The new tools I thought looked good where CrossWind and the Browser. Why I have to buy a new product to have access to a decent debugger, I don't know, but I would be happy to have something better than xxgdb. The other things in the web page don't seem new to me. C and C++ compilers, The shell, Target tools - these are already part of the standard VxWorks package. Any comments? - -- Dave Uhrenholdt duhrenho@vette.sanders.lockheed.com Phone (603) 885-7398 --------------------------- Newsgroups: comp.os.vxworks Subject: Serial port H/W handshaking? Date: Wed, 21 Feb 1996 07:55:27 -0500 From: sroth@shore.net (Scott D. Roth) Organization: Shore.Net/Eco Software, Inc; (info@shore.net) Message-ID: Has anyone succeeded in using hardware handshaking for flow control on a serial port in VxWorks? --------------------------- Newsgroups: comp.os.vxworks Subject: Test -- Please Answer! Date: 21 Feb 1996 14:53:16 GMT From: John Neumoege Organization: Fraunhofer IPK, Berlin Message-ID: <4gfbks$5p1@charly.ipk.fhg.de> I am new to the newsgroup and am not sure if the news that I am posting is actually making it around. If anyone reads this please send me a quick reply or a message to my e-mail address below. Thanks - --John - ------------------John Neumoege-------------------------- Fraunhofer-IPK e-mail: John.Neumoege@ipk.fhg.de Pascalstr. 8-9 Voice: ++49/(0)30/390 06-127 D-10587 Berlin FAX: ++49/(0)30/391 10 37 --------------------------- Newsgroups: comp.os.vxworks Subject: Maintaining multiple makefiles Date: 21 Feb 1996 17:21:12 GMT From: johncoby@ghgcorp.com (JOHN R. COBARRUVIAS) Organization: Personal Account Message-ID: I am not a guru on makeskel or makefiles in general. Looking at the makefile for vxworks can make you go blind! So, is there a swift way to maintain multiple Makefiles, one for your development environment, and one for your final target configuration? I know I can create a new configAll.h for my target configuration and then swap the two, and rebuild vxworks, but I would think there is a more "intelligent" way of doing it. ideas? - -- ,-_/ . .-,--. ,---. . ' | ,-. |-. ,-. `|__/ | -' ,-. |-. ,-. ,-. ,-. . . . , . ,-. ,-. | | | | | | | | ¥ ,. | -. | | | | ,-| | | | | | / | ,-| `-. | `-' ' ' ' ' `-' ` `' `---' `-' `-' `-^ ' ' `-' `' ' `-^ `-' / | John R. Cobarruvias johncoby@ghgcorp.com Houston, Tx `--' --------------------------- Newsgroups: comp.os.vxworks Subject: VxWindows and Tornado Date: 21 Feb 1996 16:55:45 GMT From: brucej@ibm.net Message-ID: <4gfiqh$bhk@news-s01.ny.us.ibm.net> Reply-To: brucej@ibm.net I am trying to use VxWindows 2.1 (for x86) with Tornado 1.0, and have the following problem. I have configured all the VxWorks files that need changing correctly, but when I try to run the xserver via the support.o command vxstart("main -tpri 75") I get the message "Could not find main". If I try to manually run the server from the shell by typing "main" the server runs correctly but of course I can't start any x applications. Is this a problem with the VxWindows vxstart() command not recognising symbols in Tornado ? Has anyone else had this problem & been able to sort it out ? Cheers Bruce Jackson (brucej@ibm.net) Hugh Symons Group PLC --------------------------- Newsgroups: comp.os.vxworks Subject: NTP on VxWorks (how?) Date: 21 Feb 1996 16:53:45 -0500 From: richyarr@aol.com (RichYarr) Organization: America Online, Inc. (1-800-827-6364) Message-ID: <4gg499$fhq@newsbf02.news.aol.com> Reply-To: richyarr@aol.com (RichYarr) Sender: root@newsbf02.news.aol.com I need to implement NTP on some boxes running VxWorks. I thought I heard about someone at the Monterey Bay Aquarium writing the code and making source available. Any help would be greatly appreciated. Please e-mail me at yarrington.rich@ssd.loral.com Also post for others if you wish. Thanks, Rich Yarrington --------------------------- Newsgroups: comp.os.vxworks Subject: employment opportunity Date: Wed, 21 Feb 1996 17:35:45 -0800 From: Cheryl Hoffman Organization: Heurikon Corporation Message-ID: <312BC871.5995@heurikon.com> Heurikon Corporation is a market leader in the design and manufacture of real-time single board computers. ISO-9001 certified. We are seeking a highly motivated individual for the following intermediate level position: SOFTWARE SUPPORT ENGINEER Responsibilities include: efficiently resolving a wide variety of techinical software issues, assisting in product customization, performing customer application testing, and providing the technical interface between customers and other departments. Qualified candidates will have a BS in EE or CS, effective communication skills, 2 years technical experience including customer interface, and "C" programming. Excellent problem solving skills, and familiarity with UNIX. Real-time software, networking, driver support, and microporcessor application experience desireable. We provide a great work atmosphere, health, dental, 401(k) life, tuition, time off and success sharing plan. Send your resume with salary requirements to: Human Resources Heurikon Corporation 8310 Excelsior Drive Madison, WI 53717 e-mail: hr@heurikon.com Check out our home page at: http.//www.heurikon.com Madison is recognized as one of the "most livable" cities in the entire United States by Money Magazine. Among their criteria: access to health care, quality education and employment opportunities. Madison is happily sandwiched between Lake Mendota and Lake Monona, featuring some rather unique geographical characteristics. With the lakefronts, a Big Ten University, the state capitol, recreation, shopping, night life, a farmers' market, clean air and a healthy economy, Madison is a progressive sort of place. Come check us out. --------------------------- Newsgroups: comp.os.vxworks Subject: competitors of VxWorks Date: 22 Feb 1996 00:08:55 GMT From: ywang@hera.EECS.Berkeley.EDU (Yongdong Wang) Organization: University of California, Berkeley Message-ID: <4ggc6n$qj5@agate.berkeley.edu> I am trying to collect information about real time operating systems. Which companies have products that compete with VxWorks? I have heard of VRTX from MicroTek and something called MADD (Multi Access Data Devices). But I don't have any detailed info on these two products. Any information is appreciated. Yongdong Wang ywang@cs.berkeley.edu --------------------------- Newsgroups: comp.os.vxworks Subject: TORNADO for win95 Date: Wed, 21 Feb 1996 19:02:05 -0800 From: Stephen Rejto Organization: MIT Lincoln Lab Message-ID: <312BDCAD.77A@ll.mit.edu> I'm a long time vXworks avocate and have always hosted VxWorks under UNIX (sun,sgi). We are starting up a new program and am faced with the win95 tornado vs UNIX tornado decision. Normally I would just go to the UNIX platform. However with the current steep discount on teh Win95 TORNADO platform we are very motivated to go with the win95 version. For background we do use win95 for office stuff (word,excel etc.) but we are pretty die hard UNIX types when it comes to doing "real work". Also we're developing for 68K mvme162. My question is: does it work ??? Are people happy with it? Can I still use the VX-Windows software Any related experience with win95 TORNADO would be appreciated. Thank you Stephen Rejto MIT Lincoln Lab srejto@ll.mit.edu --------------------------- Newsgroups: comp.os.vxworks Subject: Re: TORNADO for win95 Date: 22 Feb 1996 11:04:28 GMT From: John Neumoege Organization: Fraunhofer IPK, Berlin Message-ID: <4ghijs$nlq@charly.ipk.fhg.de> References: <312BDCAD.77A@ll.mit.edu> Stephen Rejto wrote: .. deleted stuff ... >My question is: > does it work ??? Hi Stephen! We bought the vxWorks 5.1b last summer and are new users to the vxworks environment, so I can't give you any comparisons to the UNIX platform. We had terrible problems (bugs) with the 5.1 version and felt it had to do with the fact that the vxWorks port to the PC world is relatively new. Since then we have updated to 5.3b and Tornado which is now running stable with only a few minor problems which we can live with until a workaround is found (or a new version is released!). On the same PC we have vxWorks 5.3b, WfW 3.11 and linux operating sysemts so we can easily switch around (by rebooting) between systems depending on our needs. I must say that I miss the project management facility from the MULTI development environment, but the other benifits (multiple shells, faster compilation times, relocation of resources to the host, etc.) from the Tornado environment more than compensate for this. Seeing as you come from the UNIX world, you will already be accustomed to the makefiles. > Are people happy with it? To this point I have worked with VRTX developed on VAX workstations using VMS and downloaded to parallel M68040 cards on VME rack. Compared to this, the vxWorks with Tornado is heavenly. The fact that I can interactively debug my software makes my job EXTREMELY earsier. And the tech support I have received from wrs has been outstanding. So I would say, "yes, I am happy with it." > Can I still use the VX-Windows software Sorry, I don't know what the VX-Windows software is. > >Any related experience with win95 TORNADO would be >appreciated. For the price, I would say buy it! Have fun... - --John - ------------------John Neumoege-------------------------- Fraunhofer-IPK e-mail: John.Neumoege@ipk.fhg.de Pascalstr. 8-9 Voice: ++49/(0)30/390 06-127 D-10587 Berlin FAX: ++49/(0)30/391 10 37 --------------------------- End of New-News digest ********************** From rmandler@bbn.com Thu Feb 22 05:24:04 1996 From: Roland Mandler Date: Thu Feb 22 05:24:06 PST 1996 Subject: VX-Windows I am interested on getting some performance information on VX-Windows 2.1 from any current users. Our embedded system must accept controls from and output data (mostly text, some graphics) to 2 remote X servers on the same network. We will be using X library calls only (no Motif), and the displays will be updated approximately once per second. Our VxWorks hosts are 33MHz 68040 boards with associated PCMCIA solid-state disks in each card cage. If possible, I would like to run both X clients on the same 68040. I would like to know: 1) Has anyone tried to run multiple clients concurrently in a similar environment? Can the 040 even handle this and still provide adequate response to the users? 2) How much disk space would be adequate for all the setup/configuration files? 3) How much memory does an active X client typically use? %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Roland Mandler % % BBN Systems & Technologies % % 50 Enterprise Center % % Suite 211A % % Middletown, RI 02842 % % % % rmandler@bbn.com % % % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% ‾ From leonid@rst.co.il Thu Feb 22 08:26:18 1996 From: leonid@rst.co.il (Leonid Rosenboim) Date: Thu Feb 22 08:26:21 PST 1996 Subject: Re: Request: Makefile for 5.3 sio archives > Date: 20 Feb 1996 22:29:28 GMT > From: friedl@cher.heurikon.com (Ted Friedl) > 5.3 developers, > I would like to create a Makefile for a 5.3-style serial driver > I just wrote. Wind River provides no examples in their Tornado > distribution or documentation. > My guess is that there are a few examples floating around on > Wind River-distributed 5.3 compatible BSPs. Would anyone with a > "target/src/drv/sio/Makefile*" file please email me a copy? Dear Ted, an Sio driver does not need a Makefile in $WIND_BASE/target/src/drv/sio directory. Instead, you're supposed to #include the sio driver source from he tyCoDrv.c file in your BSP, and build the vxWorks/bootrom images there. Leonid From MikeMcQuade@prometheus.connectsoft.net Thu Feb 22 19:41:16 1996 From: Mike McQuade Date: Thu Feb 22 19:41:23 PST 1996 Subject: PowerPC 603e, anyone have an ENV file for.. Is anybody running a Motorola PowerPC 603e with VxWorks ? If so, how did you bring it up ? with the Moto ppc-bug ? Ive heard that it can be done in this way, and Im trying to bring up a MVME-1603e board with 5.2 beta. Im not sure of the settings in the ENV on the moto board, the ENV Ive got on this 603e is different from the 603 Im currently running. Would sure like to see a copy of a "working" 603e ENV file. thanks Mike McQuade MikeMcQuade@connect.com From dugue@satie.obs-nice.fr Fri Feb 23 01:03:11 1996 From: dugue@satie.obs-nice.fr (Michel Dugue) Date: Fri Feb 23 01:03:14 PST 1996 Subject: unsubscribe unsubscribe From daemon@csg.lbl.gov Fri Feb 23 04:00:38 1996 From: daemon@csg.lbl.gov Date: Fri Feb 23 04:13:50 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Fri Feb 23 04:00:33 PST 1996 Subject: Re: Dynamic Linking - replacing symbols Subject: Re: "UNIX" domain sockets Subject: Re: S_rpcLib_RPC_TIMEOUT Subject: splitting hex files Subject: Re: creating sub-direcotry on net device Subject: Re: VxWindows and Tornado Subject: Re: RPC and VxWorks ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Dynamic Linking - replacing symbols Date: 21 Feb 1996 09:06:23 -0500 From: chaplin@dopey.keithley.com (Roger Chaplin) Organization: Keithley Instruments, Cleveland, Ohio Message-ID: <4gf8sv$gfl@dopey.keithley.com> References: <9602202114.AA25936@lbl.gov> JEFF ROMINE (jromine@btsslc.com) wrote: : Thank you to everyone who responded to my question on RPC and VxWorks. My : next question has to do with dynamic linking. What happens when a module : is loaded which has symbols identical to symbols which are already in the : symbol table? The reason I ask is that I want to replace certain functions : which are part of the VxWorks system image for all modules which are : subsequently loaded. I looked at the man page for the "ld" function. It : has the option of loading all,global only, or no symbols into the symbol : table, but it doesn't have an option which causes the symbols in the module : being loaded to replace those already there. I know that I could do this : by explicitly modifying the symbol table, but that would be awkward. It is possible for a VxWorks symbol table to contain multiple entries having the same name. The symbol table lookup returns the latest one. You can figure out for yourself how to test this (or send email, and I'll give you some hints). Therefore, simply loading your module which redefines some VxWorks-supplied routines will work, but only for modules you load afterward. Any modules already loaded and linked will still be linked to the VxWorks-supplied routines. - -- Roger Chaplin / Test Instruments Group / chaplin_roger@keithley.com | "For the lines are long and the fighting is strong and they're breakin' | | down the distance between right and wrong." | | -- Bob Dylan, "Ring Them Bells" | --------------------------- Newsgroups: comp.os.vxworks Subject: Re: "UNIX" domain sockets Date: 22 Feb 1996 20:42:51 GMT From: chaos@linux.peacefulstar.oakland.ca.us (H.J. Bae) Organization: Peaceful Star, Oakland, CA Message-ID: <4gikgb$l8i@idiom.com> References: <311A3BD8.711D@pacifier.com> Reply-To: hjb@peacefulstar.oakland.ca.us On Thu, 08 Feb 1996 10:07:20 -0800, Ricker wrote: : I know VxWorks has AF_INET and ZBUF type sockets, but is there : a way to get "UNIX domain" (AF_UNIX) type sockets too ? There is no AF_UNIX socket support in VxWorks as of version 5.1. Unless WRS has recently added support for it. - -- H.J. Bae, Consultant hjb@peacefulstar.oakland.ca.us http://www.peacefulstar.oakland.ca.us/ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: S_rpcLib_RPC_TIMEOUT Date: 22 Feb 1996 20:55:22 GMT From: chaos@linux.peacefulstar.oakland.ca.us (H.J. Bae) Organization: Peaceful Star, Oakland, CA Message-ID: <4gil7q$l8i@idiom.com> References: <9602082316.AA19398@condor.anzac> Reply-To: hjb@peacefulstar.oakland.ca.us On Fri, 9 Feb 96 10:16:19 EST, Steven McCoy wrote: : : Can anyone tell me what the RPC timeout period is and where it is : defined. Is it the nfsTimeoutSec/nfsTimeoutUSec values in VxWorks or : is this something that can be tailored in SunOS? the RPC timeout is used as parameter to RPC clnt_call macro which either becomes udp or tcp version of clnt_call routines. they are then used to timeout the RPC call being made. in UNIX implementations of NFS, the timeout value for NFS can be specified in a configuration file as mount time option. in VxWorks, you need to change the global variables nfsTimeout*. : An nfsTimeoutSec of 25 : seconds seems pretty long to me (there are no gateways involved although : I have seen the error occur more frequently when there is one). Can the : server really be that busy for that long not to service a request from : the VxWorks client? : it is possible that server may be down or slow. it is also possible that there is a problem on client side. you can trace your network for NFS packets to debug this. you can also try longer retry time. - -- H.J. Bae, Consultant hjb@peacefulstar.oakland.ca.us http://www.peacefulstar.oakland.ca.us/ --------------------------- Newsgroups: comp.os.vxworks Subject: splitting hex files Date: 22 Feb 1996 23:26:17 GMT From: jcbhrb@nic.cerf.net (Jacob Hirbawi) Organization: CERFnet Message-ID: <4giu2p$qee@news.cerf.net> This should be a fairly common utility, but I'm having trouble finding it. We have Motorola MVME167 cards with "odd" and "even" eprom sockets. How would I "split" my hex file so that the right code goes to each socket. Some of the people I talked to say their programer does that for them; mine doesn't have this capability. Any pointers about where I can find such a utility are greatly appreciated. Jacob. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: creating sub-direcotry on net device Date: 23 Feb 1996 00:37:56 GMT From: chaos@linux.peacefulstar.oakland.ca.us (H.J. Bae) Organization: Peaceful Star, Oakland, CA Message-ID: <4gj294$s4h@idiom.com> References: <9602160533.AA14203@tyl.TYL.CO.JP> Reply-To: hjb@peacefulstar.oakland.ca.us On Fri, 16 Feb 96 14:33:13 JST, K Koshimizu/TYL wrote: : : By using "mkdir" or "open" to netDevice, such as boothost: sun WS, they : return ok and the named file is created. But it is a file not direcotry. : If NFS is used, either "mkdir","open" will create directory. : I was woundering if it is possible to create directory on WS from vxWorks : without using NFS. : I am using vxWorks 5.1.1 on MVME147. WS SS/5 with SunOS 4.1.3. : you cannot do "mkdir" on netDrv. "mkdir" only works on NFS and DosFS in VxWorks. (there really is no reason why it cannot support mkdir by the way. VxWorks just does not implement it) This is only one of the problems that arise from the fact that the filesystems in VxWorks are not really filesystems, but just bunch of I/O driver routines that provide services to emulate filesystems. They are not entirely consistent from API point of view. - -- H.J. Bae, Consultant hjb@peacefulstar.oakland.ca.us http://www.peacefulstar.oakland.ca.us/ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWindows and Tornado Date: Wed, 21 Feb 1996 18:27:35 -0800 From: Gene Kennon Organization: VisiCom Laboratories, Inc. Message-ID: <312BD497.64FE@visicom.com> References: <4gfiqh$bhk@news-s01.ny.us.ibm.net> brucej@ibm.net wrote: > > I am trying to use VxWindows 2.1 (for x86) with Tornado 1.0, and have the > following problem. > > I have configured all the VxWorks files that need changing correctly, but > when I try to run the xserver via the support.o command vxstart("main -tpri 75") > I get the message "Could not find main". > > If I try to manually run the server from the shell by typing "main" the server > runs correctly but of course I can't start any x applications. > > Is this a problem with the VxWindows vxstart() command not recognising > symbols in Tornado ? > > Has anyone else had this problem & been able to sort it out ? > > Cheers > > Bruce Jackson (brucej@ibm.net) > > Hugh Symons Group PLC VX-Windows 2.1 was not designed to run under the Tornado environment. It was designed for VxWorks 5.1 and 5.2. VX-Windows 3.0, currently in beta, will support VxWorks 5.3 (Tornado) that will appropriately deal with the change in loading requirements of Tornado. Please contact our support group for more details and release dates at vx_windows@visicom.com Regards, Gene Kennon Vice President, Corporate Marketing VisiCom Laboratories, Inc. 619/457-2111 http://www.visicom.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: RPC and VxWorks Date: 23 Feb 1996 00:41:38 GMT From: chaos@linux.peacefulstar.oakland.ca.us (H.J. Bae) Organization: Peaceful Star, Oakland, CA Message-ID: <4gj2g2$s4h@idiom.com> References: <9602052025.AA16269@lbl.gov> <4g3mrb$p3h@hp5.online.apple.com> Reply-To: hjb@peacefulstar.oakland.ca.us On 16 Feb 1996 20:50:51 -0800, PWatson wrote: : I am a big proponent of RPC. I just wish Wind River would make some : enhancements to their implementation of it, like (mainly) supporting RPC : broadcasts. We have also had some problems with TCP clients-- some of : these are already documented problems. : clnt_broadcast() should work in VxWorks. What sort of problem are you having? It seemed to work last time I tried. - -- H.J. Bae, Consultant hjb@peacefulstar.oakland.ca.us http://www.peacefulstar.oakland.ca.us/ --------------------------- End of New-News digest ********************** From tkb@mclean.sparta.com Fri Feb 23 13:51:10 1996 From: Thomas Keith Buchanan Date: Fri Feb 23 13:51:13 PST 1996 Subject: Re: Console Monitor On Tue, 20 Feb 1996, the vxWorks Users Group Exploder wrote: > Does anyone have any information about how much of a performance > hit a FORCE board takes when a serial Console line is attached to > the board? The impact on performance is directly proportional to how much you use the console. If you don't print anything to the console, there is no penalty at all. If you print a lot of information to the console, you will impact performance in two ways. At a higher level, the formatted I/O functions in C are very slow in all environments and generally to be avoided in vxWorks environments. At a low level, the serial console is usually interrupt-driven. It can interact with your other interrupts in interesting ways. You can use the logLib utilities to control the task level performance impact but the interrupt level stuff will still have to be considered. My advice is to NOT use the console unless it fulfills an explicit requuuirement in your design. If you do use it, manage it like any other I/O resource. Just remember that it's use is not free and when you use it for application purposes, you limit it's usefulness as a debugging aid. > And is this performance more or less than when an ethernet > "xterm -e rsh name.domain" type console session is used? The comments above still apply but the I/O is routed thorugh the protocol stack and the Ethernet interface. The throughput is much higher but Ethernet devices also use interrupts and usually DMA resources. No I/O comes w/o strings attached... ______________________________________________________ | |¥ |Thomas Keith Buchanan Voice -> 703 448 1683 x233| | |tkb@mclean.sparta.com FAX -> 703 734 3323 | | |_____________________________________________________| | | | | | /¥¥ SPARTA, Inc. | | | ///¥¥ 7926 Jones Branch Drive | | | /////¥¥ Suite 900 | | | ///// ¥¥ McLean, VA 22102 | | | / ¥¥¥¥¥ ¥¥ 703-448-0120 | | | ¥ ¥¥¥¥¥ // | | | ¥ ////// | | | ¥////// SPARTA | | | ¥//// Pride In Performance | | | ¥// -------------------- | | |_____________________________________________________| | ¥______________________________________________________¥| From daemon@csg.lbl.gov Sat Feb 24 05:05:24 1996 From: daemon@csg.lbl.gov Date: Sat Feb 24 05:05:42 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sat Feb 24 04:03:44 PST 1996 Subject: Java for VxWorks? Subject: Re: Maintaining multiple makefiles Subject: Re: Serial port H/W handshaking? ------------------------------------------------------- Newsgroups: comp.os.vxworks,comp.lang.java Subject: Java for VxWorks? Date: Wed, 21 Feb 1996 14:27:01 -0600 From: Greg Graham Organization: ACNS Research/Dev Message-ID: <312B8015.5790@bnr.ca> Any word on an effort to port Java to VxWorks? - -- Greg Graham ggraham@bnr.ca --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Maintaining multiple makefiles Date: 22 Feb 1996 09:51:40 -0800 From: espin@idiom.com (Geoffrey Espin) Organization: A poorly-installed InterNetNews site Message-ID: References: johncoby@ghgcorp.com (JOHN R. COBARRUVIAS) writes: >I am not a guru on makeskel or makefiles in general. >Looking at the makefile for vxworks can make you go blind! >So, is there a swift way to maintain multiple Makefiles, one for your >development environment, and one for your final target configuration? >I know I can create a new configAll.h for my target configuration and then >swap the two, and rebuild vxworks, but I would think there is a more For pre-5.3 VxWorks, think of the BSP MakeSkel/Makefile scheme as a chopped down imake. Don't look at the Makefile or Makefile.$(CPU)$(TOOL) and you won't go blind. After setting your environment correctly and adjusting your path go ahead and change your MakeSkel, say add your own driver: ADDED_CFLAGS = -DCOMPANY_NAME MACH_EXTRA = myDrv.o And then % mpp >Makefile The mpp script is in vw/bin/ and should be in your path. Now just do a make, and myDrv.o will be in your new vxWorks image. The COMPANY_NAME or project name identifier should be used in any WRS code you modify. Don't clone configAll.h... just modify the BSP's config.h, that way you can have different targets some say with INCLUDE_NFS and INCLUDE_RDB and some without. Even within your config.h you might have a "#if LOADED" which turns on all the INCLUDE's in the "#if FALSE" part of configAll.h. Modifying configAll.h (after keeping a copy, say, configAll.h.wrs) could be justified if you have a lot of different targets all running the identical options but leave it in all/configAll.h. And still, mark your changes with "#ifdef COMPANY_NAME", that way you can lift your code and put it into the next VxWorks release. Though, most likely it will be useful when you're moving from a 68K BSP to a PowerPC/SPARC/... within the same release version. By keeping changes localized, upgrading your BSP or upgrading the OS to say 5.3 will be easier. Tornado/5.3 eliminates MakeSkel but still uses and include() mechanism to pull in the ugly but invaluable definitions for all the CFLAGS, tools, -Includes etc.. Geoff - -- Geoffrey Espin espin@idiom.com espin@serval.kla.com KLA (408)468-2496 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Serial port H/W handshaking? Date: Thu, 22 Feb 1996 22:16:13 -0500 From: rypma@waterloo.hp.com (Ted Rypma) Organization: Hewlett-Packard Panacom Division Message-ID: References: In article , sroth@shore.net (Scott D. Roth) wrote: >Has anyone succeeded in using hardware handshaking for flow control >on a serial port in VxWorks? Sure. But I had to write the serial driver myself. In Wind River's defence, this was for 4.0.3 and 5.0.1. I'm not sure if the current serial drivers properly implement hardware handshake. The BSP included source for the serial driver, so making the additions was relatively easy. - -- - ----- Ted Rypma, Waterloo, Ontario, Canada === rypma@waterloo.hp.com --------------------------- End of New-News digest ********************** From daemon@csg.lbl.gov Sun Feb 25 04:00:47 1996 From: daemon@csg.lbl.gov Date: Sun Feb 25 04:14:18 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sun Feb 25 04:00:43 PST 1996 Subject: Re: competitors of VxWorks Subject: Re: competitors of VxWorks Subject: Re: competitors of VxWorks Subject: m2Lib Interface Question ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: competitors of VxWorks Date: Sat, 24 Feb 1996 11:54:35 -0500 From: sroth@shore.net (Scott D. Roth) Organization: Shore.Net/Eco Software, Inc; (info@shore.net) Message-ID: References: <4ggc6n$qj5@agate.berkeley.edu> In article <4ggc6n$qj5@agate.berkeley.edu>, ywang@hera.EECS.Berkeley.EDU (Yongdong Wang) wrote: > I am trying to collect information about real time operating systems. > Which companies have products that compete with VxWorks? > I have heard of VRTX from MicroTek and something called MADD > (Multi Access Data Devices). But I don't have any detailed info > on these two products. > > Any information is appreciated. Another competitor is Lynx. The newsgroup is comp.os.lynx. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: competitors of VxWorks Date: Sat, 24 Feb 1996 16:11:45 -0600 From: Dean Cheney Organization: Heurikon Corporation Message-ID: <312F8D21.37C8@heurikon.com> References: <4ggc6n$qj5@agate.berkeley.edu> Scott D. Roth wrote: > > In article <4ggc6n$qj5@agate.berkeley.edu>, ywang@hera.EECS.Berkeley.EDU > (Yongdong Wang) wrote: > > > I am trying to collect information about real time operating systems. > > Which companies have products that compete with VxWorks? > > I have heard of VRTX from MicroTek and something called MADD > > (Multi Access Data Devices). But I don't have any detailed info > > on these two products. > > > > Any information is appreciated. > > Another competitor is Lynx. The newsgroup is comp.os.lynx. Yet another is OS-9 from Microware, discussed in comp.os.os9 - -- yours in surreal-time, dc - ---------------------------------------------------------------------- ¥¥ you only // see that "Where the hell are ¥¥ what is the singing cats?!" -P.N. // not --------------------------- Newsgroups: comp.os.vxworks Subject: Re: competitors of VxWorks Date: Sat, 24 Feb 1996 19:17:39 GMT From: bowtie@cradac.mclean.mcd.mot.com (Bill Dennen) Organization: Motorola Computer Group; Eastern Marketing Area Message-ID: References: <4ggc6n$qj5@agate.berkeley.edu> ywang@hera.EECS.Berkeley.EDU (Yongdong Wang) writes: >I am trying to collect information about real time operating systems. >Which companies have products that compete with VxWorks? Integrated Systems, Inc. pSOS+ http://www.isi.com Lynx LynxOS http://www.lynx.com (Lynx isn't as hard realtime as VxWorks but it is popular) Rational Systems Rational Microware OS/9 Regards - -- {>@<} Bill Dennen [bowtie@mclean.mcd.mot.com] | we has met the enemy, SE, Virginia Sales Territory CIS: 75126.2272 | and they is us ... Motorola Computer Group (703) 714-0707 | Pogo --------------------------- Newsgroups: comp.os.vxworks Subject: m2Lib Interface Question Date: 22 Feb 1996 15:04:00 -0500 From: bmilburn@fore.com (Barton Milburn) Organization: FORE Systems, Inc. Message-ID: <4gii7g$1pb@snoopy.fore.com> I have a question about the "m2Lib" package of vxWorks. I want to set entries in the ipRouteTable. The interface to the function m2IpRouteTblEntrySet() only allows the objects M2_IP_ROUTE_DEST, M2_IP_ROUTE_NEXT_HOP, and M2_IP_ROUTE_TYPE to be set. How does one change other MIB objects (ipRouteIfIndex, ipRouteMetric1,...) that are read-write? Is it an ioctl() interface? ************************************************************************** * Barton Milburn email : bmilburn@fore.com * * FORE Systems phone : 412.635.3487 * * 174 Thorn Hill Rd URL : http://www.fore.com * * Warrendale PA 15086 * ************************************************************************** - -- ************************************************************************** * Barton Milburn email : bmilburn@fore.com * * FORE Systems phone : 412.635.3487 * * 174 Thorn Hill Rd URL : http://www.fore.com * * Warrendale PA 15086 * ************************************************************************** --------------------------- End of New-News digest ********************** From froeber@bbn.com Sun Feb 25 17:47:56 1996 From: Fred Roeber Date: Sun Feb 25 17:47:59 PST 1996 Subject: Re: Help My Ballistic Coworker On Sat, 17 Feb 1996, Dennis R. Perlman wrote: >> It seems to me that since, for the targets we are using, the physical address >> and the virtual address are the same, about the only thing we getting out of the >> address translation mechanism is the ability to mark, via the PE's, areas >> of memory as cacheable or non-cacheable. Since we are mostly concerned with >> memory that must be non-cached, we could control caching through the CACR. >> (cache control register) and blow-off the overhead for address translation. Thomas Keith Buchanan responded: >Address translation doesn't usually introduce "overhead", it's normally >an integrated piece of the hardware that keeps the instruction and data >pipelines full. You trade much higher average throughput for the loss of >determinism due to cache hits/misses. In most cases, it's a good trade... > >If you don't need to use the caches or are concerned about determinism, >just turn off the MMU and cache stuff. You can do that by commenting out >or undefining INCLUDE_MMU_BASIC and INCLUDE_CACHE_SUPPORT. You had better >check and see why your project chose the 68060 in the first place because >your average performance will be much much lower w/o the caches enabled. >Perhaps you should just turn off the data caches or tweak the >sysPhysMemDesc tables. With the 680x0 targets that Dennis is using, there are some useful alternatives to using the MMU. We use 68040s and avoid the MMU when possible. The main issue is that you want to enable caching on memory that isn't shared with other processors or hardware and disable it for areas that are shared with other processors or with hardware that isn't handled by bus snooping hardware. The 680x0 has separate instruction and data caches. There is almost never a reason to turn off the instruction cache since the instruction space isn't normally modified dynamically or shared between processors. Turning on the instruction cache results in execution speeds that can easily be twice as fast as with the cache turned off (all timing comparisions cited here are highly variable based on exact CPU board design, memory speed, and code layout). The main issue is usually how to handle the data cache. Turning on the data cache can also result in an average 2 to 1 performance gain as compared to running with the data cache off. If the data cache is used by itself, there can be problems with data coherency when data areas are shared between multiple processors or between a processor and other IO devices that write data into memory. As Dennis stated, the main use of the MMU in VxWorks is to selectively make certain areas of memory uncached even when data caching is enabled. The problems with using the MMU include the memory space for MMU page tables and the time to set up those tables as Dennis mentioned as well as the extra nondeterminism and overhead introduced by MMU operation and things like page table accesses. In some cases, the transparent translation registers that the 68040 has can be used in place of the MMU to achieve faster operation and quicker initialization. The 68040 provides two transparent translation registers that allow one to set up two different types of special caching regions. With most CPU boards we use, we set things up one way for the VxWorks backplane master processor and another way for any slave processors. The boards we use (eg Moto MV162) provide bus snooping such that local IO devices can write to on card memory in a cache coherent fashion (ie. the cache is automatically updated if an IO device writes to physical memory that is also contained in cache). This means that we only have to worry about cases where local memory is accessed from another processor over the VME bus. Such access is used by the VxWorks shared memory backplane driver with memory on the backplane master processor (usually processor ID 0). With slave boards, there normally isn't any multiprocessor access to local memory. Since there isn't access to local memory on slave boards, they are easier to set up. You can use transparent translation register 1 to mark the full address range as uncached. You then use transparent translation register 0 to mark the on card memory as cached. Since register 0 overrides register 1 for areas that both cover, the end result is that local memory is cached and all the rest is uncached. You don't use the MMU in this case at all yet you end up with the type of memory access you normally want. Things aren't as straight forward with processor 0. The problem is that the transparent translation registers only allow you to set caching in a very coarse fashion and always change caching on a minimum of 16MB blocks. With the backplane master, some local memory is usually used on processor 0 for shared communication between all processors. The memory that is shared must be marked as uncached even though you want to leave the rest of local memory as cached to get maximum performance. To achieve this fine level of caching control, you need to use the MMU. Even here, you can use the transparent translation registers to provide easy access to large parts of the VME address range. We set up one register to allow uncached access to the upper half of the address range (ie the msb of the address is set) and the other to allow uncached access to all memory areas that have the second msb set (ie 0x20000000 - 0x3fffffff and 0x40000000 - 0x5fffffff). This leaves the area from 0 - 0x1fffffff uncontrolled. We use the MMU to address any memory we need access to in this area including the on card memory which normally starts at address 0. The VxWorks function cacheDmaMalloc can then be used to allocate a piece of local memory and set it up to be uncached even though most of memory is left as cached. Setting up cache control and the MMU in VxWorks or any other OS is a complex issue. How you do it can have a large effect on run-time performance though. I figured I would try to provide some information on alternatives to using the MMU to get the type of cache control you desire since these alternatives can yield better performance in a number of ways. As with most things, getting the better performance comes at the cost of somewhat higher code complexity. But isn't that usually the case? Fred | Fred J Roeber, BBN Systems & Technologies | | 50 Enterprise Place Middletown, RI 02842-5202 | | froeber@bbn.com 401-849-2543 (X48) | From daemon@csg.lbl.gov Mon Feb 26 04:01:40 1996 From: daemon@csg.lbl.gov Date: Mon Feb 26 04:13:56 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Mon Feb 26 04:01:37 PST 1996 Subject: Is WRS.COM really down? ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Is WRS.COM really down? Date: 26 Feb 1996 01:04:56 -0500 From: hmp@frc2.frc.ri.cmu.edu (Henning Pangels) Organization: Field Robotics Center Message-ID: Reply-To: hmp@frc2.frc.ri.cmu.edu (Henning Pangels) Sender: hmp@squash.frc.ri.cmu.edu Sorry for the noise injection, but did anyone else experience trouble sending mail to wrs.com in the past week? My mailer is bouncing it right back at me claiming wrs.com is down, and I cannot reach it via other net means, either. Heck, 'ping wrs.com' segv's on me (but that may be a local problem:-) I'm not aware of any other net-related problems at our site, but I just find it hard to believe that the main email gateway of such a fine company would be *down* for a whole week! Is it me/us or is it time to adjust my expectations? Thanks, - -Henning - -- Henning Pangels | hmp@cs.cmu.edu | Field Robotics Center Senior Research Programmer| (412) 268-8911 |Carnegie-Mellon University - ---------------------------------------------------------------------------- Experience is what you get when you don't get what you want. --------------------------- End of New-News digest ********************** From tkb@mclean.sparta.com Mon Feb 26 06:08:44 1996 From: Thomas Keith Buchanan Date: Mon Feb 26 06:08:55 PST 1996 Subject: Re: comp.os.vxworks newsdigest > Sorry for the noise injection, but did anyone else experience trouble > sending mail to wrs.com in the past week? My mailer is bouncing it My mail to wrs.com is bouncing so often that they are calling me complaining about the lack of beta-test reports. I had the same problems late last year. ______________________________________________________ | |¥ |Thomas Keith Buchanan Voice -> 703 448 1683 x233| | |tkb@mclean.sparta.com FAX -> 703 734 3323 | | |_____________________________________________________| | | | | | /¥¥ SPARTA, Inc. | | | ///¥¥ 7926 Jones Branch Drive | | | /////¥¥ Suite 900 | | | ///// ¥¥ McLean, VA 22102 | | | / ¥¥¥¥¥ ¥¥ 703-448-0120 | | | ¥ ¥¥¥¥¥ // | | | ¥ ////// | | | ¥////// SPARTA | | | ¥//// Pride In Performance | | | ¥// -------------------- | | |_____________________________________________________| | ¥______________________________________________________¥| From CJONES@digsys1.bwi.wec.com Mon Feb 26 13:10:26 1996 From: "CHRIS JONES" Date: Mon Feb 26 13:10:28 PST 1996 Subject: semInfo function / browser spy window I have two questions (1) the section of code below is from a creation tasks which spawns other system tasks and then wait for them all to pend on a 'Startup' semaphore before Flushing this semaphore in order to start all tasks at the same time. In the while loop below, I added a taskDelay statement to make this task sleep brieflly before checking again how many child tasks are pending. If I do not include this delay, the board crashes hard. Any ideas why? I have round robin scheduling enabled during this startup sequence. Does it have something to do with rapid successive calls to semInfo()? ******* Code ******* const NUMBER_OF_TASKS = 2 ; int NumTasksInitialized = 0 ; int TaskIdList[NUMBER_OF_TASKS] ; // wait until all tasks pend on the startup semaphore while ((NumTasksInitialized < NUMBER_OF_TASKS)) { // get the number of tasks currently pending on the Startup semaphore NumTasksInitialized = semInfo(semIdStartup, TaskIdList, NUMBER_OF_TASKS) ; // sleep for a while to give tasks time to initialize taskDelay(10); } // all tasks have reported in, start the system !! semFlush(semIdStartup) ; ******* End ******* Next question: I have one task which just sends messages to a second task with a message Q. In the spy browser window of Tornado, the INTERRUPT line indicates that 44% of the CPU is being used for interrupts while only 27% is being used by each task. Is this a mistake? I have no interrupts in my system. What could cause this indicator with a simple system such as mine? Thanks Chris Jones cjones@digsys1.bwi.wec.com Electrical Engineer Northrop Grumman Electronic Sensors and Systems Division (ESSD) (410) 765-7959, (410) 993-8822 (FAX) From tkb@mclean.sparta.com Mon Feb 26 17:50:52 1996 From: Thomas Keith Buchanan Date: Mon Feb 26 17:50:55 PST 1996 Subject: Re: semInfo function / browser spy window On Mon, 26 Feb 1996, the vxWorks Users Group Exploder wrote: > // get the number of tasks currently pending on the Startup > semaphore NumTasksInitialized = semInfo(semIdStartup, > TaskIdList, NUMBER_OF_TASKS) ; What is the token 'semaphore' used for? It looks out of place, semInfo returns a simple integer. > > // sleep for a while to give tasks time to initialize > taskDelay(10); > > } > > Next question: > > I have one task which just sends messages to a second task with a > message Q. In the spy browser window of Tornado, the INTERRUPT line > indicates that 44% of the CPU is being used for interrupts while only > 27% is being used by each task. Is this a mistake? I have no > interrupts in my system. What could cause this indicator with a > simple system such as mine? Is the receiving task blocking while waiting for the message? Is the sending task running w/o any blocking or timing mecahanism? Are you printing any messages to the console? What is your spy sample rate? Every target I have ever seen has at least 2 interrupt sources, one for the system timer and one for the serial port used for the console. If you have SCSI, Ethernet, parallel port, extra timer, extra serial port etc. you can usually add another interrupt source for each. If your target has a bus controller on it, there will probably be more interrupt sources though they should not be active by default. adios ______________________________________________________ | |¥ |Thomas Keith Buchanan Voice -> 703 448 1683 x233| | |tkb@mclean.sparta.com FAX -> 703 734 3323 | | |_____________________________________________________| | | | | | /¥¥ SPARTA, Inc. | | | ///¥¥ 7926 Jones Branch Drive | | | /////¥¥ Suite 900 | | | ///// ¥¥ McLean, VA 22102 | | | / ¥¥¥¥¥ ¥¥ 703-448-0120 | | | ¥ ¥¥¥¥¥ // | | | ¥ ////// | | | ¥////// SPARTA | | | ¥//// Pride In Performance | | | ¥// -------------------- | | |_____________________________________________________| | ¥______________________________________________________¥| From daemon@csg.lbl.gov Tue Feb 27 04:00:39 1996 From: daemon@csg.lbl.gov Date: Tue Feb 27 04:00:43 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Tue Feb 27 04:00:34 PST 1996 Subject: zmodem (rz/sz) for vxworks? Subject: Re: Is WRS.COM really down? Subject: Re: Java for VxWorks? Subject: GNU g++ cross compiler for VxWorks Subject: mv167/hklaguna shared memory with mailbox interrupts ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: zmodem (rz/sz) for vxworks? Date: Sat, 24 Feb 1996 21:14:26 -0800 From: "Tunc N. Hoscan" Message-ID: <312FF032.63F@ijetlink.net> Hi everyone, I'm looking for a port of Chuck Forsberg's rz/sz or any other zmodem ports for vxworks 5.1.1 or better. Anyone ever heard of something like that? Any clue is appreciated. Thanks, - -- Tunc N. Hoscan e-mail: tunch@jetlink.net www : http://www.jetlink.net/‾tunch --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Is WRS.COM really down? Date: 26 Feb 1996 16:51:12 GMT From: johncoby@ghgcorp.com (JOHN R. COBARRUVIAS) Organization: Personal Account Message-ID: References: In article , hmp@frc2.frc.ri.cmu.edu (Henning Pangels) wrote: > Sorry for the noise injection, but did anyone else experience trouble > sending mail to wrs.com in the past week? My mailer is bouncing it > right back at me claiming wrs.com is down, and I cannot reach it via > other net means, either. Heck, 'ping wrs.com' segv's on me (but that > may be a local problem:-) I'm not aware of any other net-related problems at > our site, but I just find it hard to believe that the main email > gateway of such a fine company would be *down* for a whole week! > Is it me/us or is it time to adjust my expectations? Hum.....seems fine to me. 2/26/96 > Thanks, > > -Henning > -- > Henning Pangels | hmp@cs.cmu.edu | Field Robotics Center > Senior Research Programmer| (412) 268-8911 |Carnegie-Mellon University > ---------------------------------------------------------------------------- > Experience is what you get when you don't get what you want. - -- ,-_/ . .-,--. ,---. . ' | ,-. |-. ,-. `|__/ | -' ,-. |-. ,-. ,-. ,-. . . . , . ,-. ,-. | | | | | | | | ¥ ,. | -. | | | | ,-| | | | | | / | ,-| `-. | `-' ' ' ' ' `-' ` `' `---' `-' `-' `-^ ' ' `-' `' ' `-^ `-' / | John R. Cobarruvias johncoby@ghgcorp.com Houston, Tx `--' --------------------------- Newsgroups: comp.os.vxworks,comp.lang.java Subject: Re: Java for VxWorks? Date: Tue, 27 Feb 1996 04:44:37 GMT From: Kenneth Green Organization: Hewlett-Packard Company Message-ID: <31328C35.7965@hpato.aus.hp.com> References: <312B8015.5790@bnr.ca> <4gr7n0$b0n@asp.erinet.com> Sender: news@hparc12.aus.hp.com (News Daemon ID) I know that the developers of pSos have announced a Java port to that OS. This should provide some encouragement to WRS to do the same for VxWorks. We live in hope ;-) Rgds, Kenneth Steve Gaeke wrote: > > This would be great. Especially if the appletviewer was a small > kernel. This would have great marketing potential against QNX's > photon. Let me know if you hear anything!!! > > steve. > sgaeke@trinova.com > > Greg Graham wrote: > > >Any word on an effort to port Java to VxWorks? > >-- > > >Greg Graham > >ggraham@bnr.ca - -- _______________________________________________________________________________ Kenneth Green Hewlett Packard Australia Ltd Design Engineer Australian Telecom Operation (ATO) 347 Burwood Highway Phone: +613 9210 5541 Burwood East 3130 Fax : +613 9210 5550 AUSTRALIA Postal address: Email: kenneth_green@aus.hp.com P.O. Box 221, Blackburn 3130, Australia _______________________________________________________________________________ --------------------------- Newsgroups: comp.os.vxworks Subject: GNU g++ cross compiler for VxWorks Date: 26 Feb 1996 20:43:10 GMT From: Irene Barg Organization: University of Arizona, MMT Observatory Message-ID: <4gt60u$tmc@news.ccit.arizona.edu> Hello, I've installed a native g++ compiler for Sun SPARC Solaris environment, and want to build a cross-compiler for m68k-wrs-vxworks. Has anyone done this? Please respond to email address below. Thanks, - --irene Irene Barg MMT Observatory University of Arizona Tucson, AZ 85721 Ph: 520-621-7647 Email: ibarg@as.arizona.edu --------------------------- Newsgroups: comp.os.vxworks Subject: mv167/hklaguna shared memory with mailbox interrupts Date: Tue, 27 Feb 1996 05:23:05 UTC From: etang@quack.kfu.com (Eugene Tang) Organization: The Duck Pond public unix: +1 408 249 9630, log in as 'guest'. Message-ID: Sender: news@quack.kfu.com (0000-News(0000)) Hi, I'm working on configuring a VME chassis with a Motorola MV167 as a controller and a Heurikon Laguna (MIPS R4600-based) as a slave over a shared memory network using mailbox interrupts. I was able to get it working using polling. However, when using mailbox interrupts I get either a bus error, an access exception or a hang after it says "Loading..." (probably within netLoad). I've tried a16, a24 and a32 address spaces for the mailbox location without any luck. Just wondering if anyone has been able to get a configuration like this one working before, or has any suggestions. Thanks. - -Eugene --------------------------- End of New-News digest ********************** From del@rst.com Wed Feb 28 07:15:07 1996 From: del@rst.com (Mark Del Giorno) Date: Wed Feb 28 07:15:10 PST 1996 Subject: 2 Processors & Data Cache - do they mix? Heyyyyyyyyyyyyyyyyyyyyyyyyyyyyyy, We're using an MVME162 and want to add a second 162 to the VME backplane. They need to share a small amount of data, and we'd like to keep it simple by using a block of dual-ported memory on the "master" and have the "slave" just write data across the VMEbus. A simple software semaphore can (hopefully) guarantee mutual exclusion in reading/writing the data. The question is: Does the on-chip data cache of the 68040 screw up a simple arrangement like this? e.g. If the slave writes data to the master's RAM, then the master goes to read the data, is it possible that the master's cache will "hit" and master won't get the actual data that was written to the dual port RAM? (We have disabled the cache via sysLib.c, but suffer a factor of 2-3 times in speed! Something we simply can't live with....or we wouldn't need to add another processor!) Thanks in advance for any insight. _/_/_/_/ _/_/_/_/ _/_/_/_/_/ Mark Del Giorno (del@rst.com) _/ _/ _/ _/ (410) 876-9200 Fax: 876-9470 _/_/_/_/ _/_/_/_/ _/ Robotic Systems Technology _/ _/ _/ _/ 1110 Business Parkway South _/ _/ _/_/_/_/ _/ Westminster, MD 21157 From stan@rti.com Wed Feb 28 21:56:29 1996 From: Stan Schneider Date: Wed Feb 28 21:56:32 PST 1996 Subject: Re: 2 Processors & Data Cache - do they mix? >> We're using an MVME162 and want to add a second 162 to the >> VME backplane. They need to share a small amount of data, and we'd >> like to keep it simple by using a block of dual-ported memory on the >> "master" and have the "slave" just write data across the VMEbus. >> >> A simple software semaphore can (hopefully) guarantee mutual exclusion >> in reading/writing the data. >> >> The question is: Does the on-chip data cache of the 68040 screw up a >> simple arrangement like this? >> A simple way to do this is to allocate the memory with cacheDmaMalloc(). -- Stan stan@rti.com From smith.ian@traf.com Thu Feb 29 03:34:18 1996 From: smith.ian@traf.com Date: Thu Feb 29 03:34:21 PST 1996 Subject: Boot Roms What is the problem with trying to buy blank bootroms for MVME167 processors (PLCC type 27C202JK) ? Our suppliers in UK want ten weeks for delivery. I would appreciate if any of you 167 users could enlighten me with a supplier who holds stock. Email as below, thanks for any assistance Ian =================================== Ian D. Smith (smith.ian@traf.com) Phone: +44- 114 -244 9971 Davy International =================================== From del@rst.com Thu Feb 29 10:36:31 1996 From: del@rst.com (Mark Del Giorno) Date: Thu Feb 29 10:36:33 PST 1996 Subject: Re: 2 Processors & Data Cache - do they mix? Heyyyyyyyyyyyyyyyyy, >> We're using an MVME162 and want to add a second 162 to the >> VME backplane. They need to share a small amount of data, and we'd >> like to keep it simple by using a block of dual-ported memory on the >> "master" and have the "slave" just write data across the VMEbus. >> >> A simple software semaphore can (hopefully) guarantee mutual exclusion >> in reading/writing the data. >> >> The question is: Does the on-chip data cache of the 68040 screw up a >> simple arrangement like this? >> > >A simple way to do this is to allocate the memory with cacheDmaMalloc(). > > -- Stan > stan@rti.com I've used cacheDmaMalloc(), and it doesn't seem to work. I've been talking to WRS about it, and they can't explain the behavior I'm seeing either. Basically, the MMU can only deal with 8k boundaries for assigning areas to be cacheable or non-cacheable. So if you ask for a cacheDmaMalloc(10), it should allocate at least 8k, and return a pointer to 10 bytes in the 8k. This is definitely not what happens. My sample code is: #include "vxWorks.h" #include "stdio.h" #include "stdlib.h" #include "vmLib.h" #include "cacheLib.h" #include "private/vmLibP.h" unsigned char *p; void ttt(n) int n; { if (cacheEnable(DATA_CACHE) == ERROR) printf("Error enabling cache¥n"); printf("VM_PAGE_SIZE_GET() = %d¥n", VM_PAGE_SIZE_GET()); p = cacheDmaMalloc(n); printf("p = %p¥n", p); p = malloc(n); printf("p = %p¥n", p); p = cacheDmaMalloc(n); printf("p = %p¥n", p); p = malloc(n); printf("p = %p¥n", p); } When run, this code does the following: -> ttt 10 VM_PAGE_SIZE_GET() = 8192 p = 0x3fde7c p = 0x3fde68 p = 0x3fde54 p = 0x3fde40 value = 13 = 0xd The problem is, according to WRS, each cacheDmaMalloc should allocate a minimum of 8192 bytes. According to the result above, it looks like it's just doing a regular old malloc() and allocating 20 bytes (which is also a little puzzle, because I figured it would be 10 + 8 = 18....maybe it's a 4-byte boundary thing...) If anyone can try this code on a MVME162 and gets different results, or if there's anyone out there using cacheDmaMalloc() successfully and can see the error of my ways, please let me know. Thanks again. _/_/_/_/ _/_/_/_/ _/_/_/_/_/ Mark Del Giorno (del@rst.com) _/ _/ _/ _/ (410) 876-9200 Fax: 876-9470 _/_/_/_/ _/_/_/_/ _/ Robotic Systems Technology _/ _/ _/ _/ 1110 Business Parkway South _/ _/ _/_/_/_/ _/ Westminster, MD 21157 From stan@rti.com Thu Feb 29 15:19:41 1996 From: Stan Schneider Date: Thu Feb 29 15:19:43 PST 1996 Subject: Re: 2 Processors & Data Cache - do they mix? >> I've used cacheDmaMalloc(), and it doesn't seem to work. I've been >> talking to WRS about it, and they can't explain the behavior I'm seeing >> either. Basically, the MMU can only deal with 8k boundaries for >> assigning areas to be cacheable or non-cacheable. So if you ask for >> a cacheDmaMalloc(10), it should allocate at least 8k, and return a pointer >> to 10 bytes in the 8k. This is definitely not what happens. >> Hmmmm. I thought it worked for us. Do you have MMU_BASIC enabled? Are you using VxVM? I don't know that we've ever tried it with a non-page-size multiple block; that could be a bug. Have you tried that? -- Stan stan@rti.com