From harry@ellips.nl Tue Oct 1 03:55:12 1996 From: Harry Stox Date: Tue Oct 1 03:55:15 PDT 1996 Subject: ACE for VxWorks? Hello, Has anyone ported (parts of) ACE (ADAPTIVE Communication Environment) to the VxWorks real-time operating system? We are mainly interested in the C++ wrappers for BSD sockets (and associated framework parts) to implement some cross-platform networking code (VxWorks - NT). Thanks in advance for your replies to harry@ellips.nl. I'll summarize to both mailing lists (ace-users and VxWExploder). Best regards Harry Stox (harry@ellips.nl) Hard- and Software Engineer Ellips B.V. (http://www.ellips.nl) From daemon@csg.lbl.gov Tue Oct 1 04:00:49 1996 From: daemon@csg.lbl.gov Date: Tue Oct 1 04:00:51 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Tue Oct 1 04:00:46 PDT 1996 Subject: BFDs and COFF Subject: Re: multi-protocol implementation Subject: MVME162 Flash Programming ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: BFDs and COFF Date: Mon, 30 Sep 1996 15:52:01 -0700 From: Chuck Boatwright Message-ID: <32504F11.7B59@cisnc.canon.com> References: <199609110314.LAA04086@b2.nixltd.com.au> <324FC8A7.219B@mlswc.mlm.lucent.com> Reply-To: Charles_Boatwright@cisnc.canon.com Multi platform library question. Does anyone have any experience (success or otherwise) linking libraries NOT created with GNU compilers into a vxworks target? specifcally. I have a COFF (x86) library, naturally the source code is not available, and I need to link into a vxworks target, also x86. I am working on a BFD for the x86 COFF file, however, things are not working. . . TIA for the help. - -chuck --------------------------- Newsgroups: comp.os.vxworks Subject: Re: multi-protocol implementation Date: Thu, 19 Sep 1996 15:13:26 GMT From: gerlach@netcom.com (Matthew H. Gerlach) Organization: Netcom Online Communications Services (408-241-9760 login: guest) Message-ID: References: <51q4m5$rq@lex.zippo.com> Sender: gerlach@netcom23.netcom.com It's been some time that I messed with VxWorks protocol stacks, but at one time I worked on a product with IP and Link Layer SNA. In short the protocol stacks in VxWorks are not implemented as individual tasks. There is no IP task per say. There is the netjob task which actually executes the IP code. In our case the SNA Link Layer protocol stack was implemented so that the code executes in the context of the net job tasks as well. Usually in multi-protocol envirnments the different protocol don't usually talk to each other directly. Usually the device driver hands the given packet to the appropriate protocol stack and that's about it. One might have an application that interacts with multiple protocols, like a protocol translater, but in that case data is received from on protocol and then transmittedd out another protocol. I hope this helps rather than confuses. Matthew In article <51q4m5$rq@lex.zippo.com> Venu writes: >Hi all, > >I would like to know how multi-protocol stacks (for e.g., TCP/IP and SPX/IPX), >using same MAC interface (say 802.3), are implemented in VxWorks. I would >appreciate if anybody can provide me information as to how these three entities >(in the example) -- TCP/IP, SPX/IPX, and 802.3 -- are organized as tasks and >how they communicate. > >Thanks, >Venu --------------------------- Newsgroups: comp.os.vxworks Subject: MVME162 Flash Programming Date: 24 Sep 1996 02:28:06 GMT From: steved@pacifier.com@199.2.117.163 (Steve Doiel) Organization: Pacifier BBS, Vancouver, Wa. ((360) 693-0325) Message-ID: <527gvm$h9@news.pacifier.com> Reply-To: steved@pacifier.com (Steve Doiel) >From: Stephanie Lynn >Sent: Friday, September 13, 1996 10:40 AM >To: Steve Doiel > >Steve, > Can you post a note the the wind river internet help group asking if >anyone has been able to program flash memory on the mvme162 from a >vxworks application? And if it's possible, an example would be most >welcome. > What we want to do is send out software updates via modem and have our program read the new program into memory and program it into flash... Any Hints? Steve Doiel --------------------------- End of New-News digest ********************** From del@rst.com Tue Oct 1 05:47:49 1996 From: del@rst.com (Mark Del Giorno) Date: Tue Oct 1 05:47:51 PDT 1996 Subject: TCP over a serial link Heyyyyyyyyyyyyyyyyyyyyyyyyyyyyyy, We're using a wireless ethernet-based link to a robotic vehicle for command and control information from a base station. The protocal is TCP or UDP, depending on the type of data being sent. We also have a backup radio link that provides 2-way serial (RS-232) communication. We'd like to be able to switch between the ethernet and the serial ports "seamlessly". Question: Is there soffware available to open a serial port and treat it like a TCP connection - i.e. "connects" to the other size, packetizes data, auto-retries when there are errors, etc.....? I'm running an MVME-162 on both sides of the link. Thanks in advance to anyone who may have simple solution. Mark _/_/_/_/ _/_/_/_/ _/_/_/_/_/ Mark Del Giorno (del@rst.com) _/ _/ _/ _/ Robotic Systems Technology _/_/_/_/ _/_/_/_/ _/ 1110 Business Parkway South _/ _/ _/ _/ Westminster, MD 21157 _/ _/ _/_/_/_/ _/ (410) 876-9200 Fax: 876-9470 http://www.rst.com From mea@mclean.sparta.com Tue Oct 1 06:54:06 1996 From: Mike Anderson Date: Tue Oct 1 06:54:09 PDT 1996 Subject: Re: TCP over a serial link At 05:47 AM 10/1/96 PDT, you wrote: >Submitted-by del@rst.com Tue Oct 1 05:47:49 1996 >Submitted-by: del@rst.com (Mark Del Giorno) > >Heyyyyyyyyyyyyyyyyyyyyyyyyyyyyyy, > >We're using a wireless ethernet-based link to a robotic vehicle for >command and control information from a base station. The protocal >is TCP or UDP, depending on the type of data being sent. > >We also have a backup radio link that provides 2-way serial (RS-232) >communication. We'd like to be able to switch between the ethernet >and the serial ports "seamlessly". > >Question: Is there soffware available to open a serial port and treat >it like a TCP connection - i.e. "connects" to the other size, packetizes >data, auto-retries when there are errors, etc.....? > >I'm running an MVME-162 on both sides of the link. > Greetings! Yes, simply use SLIP on the serial port. (#INCLUDE_SLIP in configAll.h) When you do a slipInit, the serial port will be configured to run the IP stack (whether or not the radio link is up). You can then do connects, send, recv, sendmsg, etc. over the serial link as if it were an Ethernet and the higher level protocol (i.e., TCP) will handle the error detection/ correction. HTH, =============================================================================== __ Real-Time System Development, Integration, Training and Services //\\ // \\ Mike Anderson // /\ \\ Chief Engineer Voice : (703) 448-0210 ext. 235 // / \ \\ SPARTA, Inc. FAX : (703) 734-3323 // \ \\ 7926 Jones Branch Drive EMAIL : mea@mclean.sparta.com \\ \ // Suite 900 Web : http://www.mclean.sparta.com \\ \ / // McLean, VA 22102 \\ \/ // "Software development is like making \\ // a baby... You can't make a baby in one \\// month by impregnating nine women. -- "Pride in Performance" Some things just take time." =============================================================================== From avlasblo@best.ms.philips.com Tue Oct 1 08:08:38 1996 From: Arjon Vlasblom Date: Tue Oct 1 08:08:40 PDT 1996 Subject: Re: TCP over a serial link > > Submitted-by del@rst.com Tue Oct 1 05:47:49 1996 > Submitted-by: del@rst.com (Mark Del Giorno) > > Heyyyyyyyyyyyyyyyyyyyyyyyyyyyyyy, > > We're using a wireless ethernet-based link to a robotic vehicle for > command and control information from a base station. The protocal > is TCP or UDP, depending on the type of data being sent. > > We also have a backup radio link that provides 2-way serial (RS-232) > communication. We'd like to be able to switch between the ethernet > and the serial ports "seamlessly". > > Question: Is there soffware available to open a serial port and treat > it like a TCP connection - i.e. "connects" to the other size, packetizes > data, auto-retries when there are errors, etc.....? > > I'm running an MVME-162 on both sides of the link. > > Thanks in advance to anyone who may have simple solution. > Just go for a PPP connection on your serial link. Regards, -- ____________________________________________________________________ / \ | Arjon Vlasblom | Philips Medical Systems | | Software Project Leader URF/CAS | Building QJ2310 | | Voice : +31-40-2763126 | P.O. Box 10000 | | Fax : +31-40-2765657 | 5680 DA Best | | E-mail: avlasblo@best.ms.philips.com | The Netherlands | |--------------------------------------------------------------------| | You cannot achieve the impossible without attempting the absurd. | \____________________________________________________________________/ From GLWICKS@sandia.gov Tue Oct 1 11:11:00 1996 From: "Wickstrom, Gregory L 2671 M" Date: Tue Oct 1 11:11:02 PDT 1996 Subject: Booting VxWorks from a PC We are trying to boot VxWorks 5.1 from a PC running Windows NT 3.51 using TFTP service that comes with the ChameleonNFS product. It appears to boot VxWorks to a runnable state but fails near the end of the transfer or in the closing of the connection. The result is that the connection is refused by the PC when the symbol table is attempted to be transferred. All we get is a VxWorks prompt with no ability to do anything since there is no symbol table. Does anyone have experience in trying something similar? Thanks, Greg Wickstrom Sandia National Laboratories glwicks@sandia.gov From bwenholz@pacesetter.com Tue Oct 1 13:51:37 1996 From: bwenholz@pacesetter.com (Bruce Wenholz) Date: Tue Oct 1 13:51:40 PDT 1996 Subject: Re: Booting VxWorks from a PC ) )Submitted-by GLWICKS@sandia.gov Tue Oct 1 11:11:00 1996 )Submitted-by: "Wickstrom, Gregory L 2671 M" ) ) )We are trying to boot VxWorks 5.1 from a PC running Windows NT 3.51 using )TFTP service that comes with the ChameleonNFS product. It appears to boot )VxWorks to a runnable state but fails near the end of the transfer or in the )closing of the connection. The result is that the connection is refused by )the PC when the symbol table is attempted to be transferred. All we get is )a VxWorks prompt with no ability to do anything since there is no symbol )table. ) )Does anyone have experience in trying something similar? ) ) This sounds like a problem where vxWorks is trying to open the same socket number for the symblo table and the kernel file. The reason this happens is the bootrom.sys is a limited kernel. When the full kernel is loaded and trys to open a socket, it trys to open the same socket because the full kernel erinitializes the initial socket number for tftp. The problem is that the time for the NT system to close the socket is longer that the time for the vxWorks bootrom.sys to jump to the full vxWorks kernel and for that kernel to try to open the socket. I have heard of this problem with other hosts. ======================8<---work---------------------- e-mail: bwenholz@pacesetter.com (Bruce Wenholz) Pacesetter, Inc Sylmar, Ca 91342, USA ======================8<---personal------------------ e-mail: MMCo@DeltaNet.Com (Bruce Wenholz) WEB: http://www.deltanet.com/users/mmco/essay.html From cpires@br.homeshopping.com.br Tue Oct 1 14:15:49 1996 From: Carlos Pires Date: Tue Oct 1 14:15:51 PDT 1996 Subject: unsubscribe unsubscribe From MikeMcQuade@connect.com Tue Oct 1 16:56:56 1996 From: Mike McQuade Date: Tue Oct 1 16:56:58 PDT 1996 Subject: Questions about exceptions. Im using Tornado, and PowerPC. My application blows up, and spits this info to the Windshell, what is the best way to interperet this info ? Does anyone know where to find out what the exception numbers mean ? Exception number 768: Task: 0xdefbc8 (sendData) data access program counter: 0x104834 access address: 0x1000c90 machine status: 0xb030 14d800 vxTaskEntry +60 : xmitData ([]) 2578e8 xmitData +70 : fill_packet_buffer ([]) 257de4 fill_packet_buffer+258: bcopyWords ([]) The tt command now works on PowerPC, Im not really sure how it works, can anyone explain it in a nutshell ? Thanks. MikeMcQuade@connect.com From meijer@mms.matra-espace.fr Tue Oct 1 19:02:13 1996 From: Bruno MEIJER Date: Tue Oct 1 19:02:15 PDT 1996 Subject: VxWorks as a COTS experience ... [This message is converted from WPS-PLUS to ASCII] Dear reader, did you use the VxWorks product as a COTS (Commercial Of the Shelf) product in any project ? We would like you to be kind to help us in getting some answers or in sharing similar experiences from your projects (any project, particularly critical ones or space related ones). Within your experience, our concerns mainly deal with the following questions : 1. What was your selection and procurement process for VxWorks ? 2. Did you need additional developments and/or adaptations ? How did you manage them ? 3. What was your policy for testing, verification, qualification and acceptance of this commercial product ? 4. Did you perform an engineering assessment and/or a software quality assessment (e.g. a SW quality audit)? Could you provide us with the results / report ? 5. Did you perform a "risks assessment", taking into account for example the following classes of risks : technical risks : functionalities, reliability, performances, integration, interoperability, ... quality risks : maintainability, documentation (completness, correctnes, ...), tests coverage, configuration management, ... maintenance risks : support, source code availability, ... ... We will be very interested in reading any experience on that subject. Please feel free to foreward this message to people that may give us some informations. Thank you very much for your contributions. Sincerily Yours, Bruno MEIJER my e-mail address is : meijer@mms.matra-espace.fr@umc@CMTMRP ------------------------------------------------------------------------------ Systeme automatique de messagerie *** Attention *** *** Attention *** *** Attention *** *** Attention *** A dater du 23/09/96 Matra Marconi Space France change le nom de son domaine. matra-espace.fr est remplace par mms.fr. Desormais, veuillez utiliser les regles d'adressage suivantes: Site de Toulouse :adresse Messagerie MMS -----> prenom.nom@tls.mms.fr adresse stations -----> username@station.tls.mms.fr Site de Velizy :adresse Messagerie MMS -----> prenom.nom@vlz.mms.fr adresse stations -----> username@station.vlz.mms.fr L'ancien domaine sera definitivement inaccessible a dater du 30/11/96 ------------------------------------------------------------------------------ *** Important *** *** Important *** *** Important *** On the 23th September Matra Marconi Space France will change its domain name. matra-espace.fr is replaced by mms.fr In the future, use the following address syntax : Toulouse site :MMS mail adress -----> firstname.name@tls.mms.fr Workstation adress -----> username@workstation.tls.mms.fr Velizy site :MMS mail adress -----> firstname.name@vlz.mms.fr Workstation adress -----> username@workstation.vlz.mms.fr From the 30th November, the old domain will no longer be available ------------------------------------------------------------------------------- ------------------------------------------------------------------------------ Systeme automatique de messagerie *** Attention *** *** Attention *** *** Attention *** *** Attention *** A dater du 23/09/96 Matra Marconi Space France change le nom de son domaine. matra-espace.fr est remplace par mms.fr. Desormais, veuillez utiliser les regles d'adressage suivantes: Site de Toulouse :adresse Messagerie MMS -----> prenom.nom@tls.mms.fr adresse stations -----> username@station.tls.mms.fr Site de Velizy :adresse Messagerie MMS -----> prenom.nom@vlz.mms.fr adresse stations -----> username@station.vlz.mms.fr L'ancien domaine sera definitivement inaccessible a dater du 30/11/96 ------------------------------------------------------------------------------ *** Important *** *** Important *** *** Important *** On the 23th September Matra Marconi Space France will change its domain name. matra-espace.fr is replaced by mms.fr In the future, use the following address syntax : Toulouse site :MMS mail adress -----> firstname.name@tls.mms.fr Workstation adress -----> username@workstation.tls.mms.fr Velizy site :MMS mail adress -----> firstname.name@vlz.mms.fr Workstation adress -----> username@workstation.vlz.mms.fr From the 30th November, the old domain will no longer be available ------------------------------------------------------------------------------- From daemon@csg.lbl.gov Wed Oct 2 04:01:04 1996 From: daemon@csg.lbl.gov Date: Wed Oct 2 04:01:06 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Wed Oct 2 04:01:02 PDT 1996 Subject: Modification of bootp port number Subject: Re: MVME167 SCSI performance under Vxworks Subject: (G) UI development on VxWorks ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Modification of bootp port number Date: 20 Sep 1996 09:08:19 -0500 From: tsnider@sol.ks.symbios.com (Tim Snider) Organization: Symbios Logic Message-ID: <51u8gj$mfk@sol.ks.symbios.com> We have a need to change the normal port number for boot request and reply messages. Is there an interface, variable, hack that would allow us to do this in vxWorks? If we had the source for the vxWorks bootp we'd just modify that since we've already modified the servers bootp code to do other unique things. TIA Tim - -- Timothy Snider tim.snider@symbios.com Symbios Logic 3718 N. Rock Road phone 316-636-8736 Wichita, KS. 67226 fax 316-636-8889 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: MVME167 SCSI performance under Vxworks Date: 24 Sep 1996 10:52:52 -0400 From: eschott@psu.edu Organization: The Pennsylvania State University, Univeristy Park, PA Message-ID: References: <3247DEB8.5518@morgan.ucs.mun.ca> Sender: els@batlaf.arl.psu.edu Bill Hunt writes: > I'm wondering what kind of performance I can expect from the SCSI > interface on the MVME167. The Vxworks FAQ gives figures approaching > 4 Mbytes a second for large transfer sizes on a Herikon 040 board which > uses the same SCSI chip as the 167. A vendor of third party drivers for > SCSI boards has told me not to expect more than 600-700 Kbytes/sec. > Does anyone have any experiance with this they can share ? I have a MVME177. I have done some testing with optical drives (which are slower that conventional hard drives). With two drive, I have achieved write rates of 1.7 Mb/s and read rates of 2.1 Mb/s. During this testing, I have made the following observations - The current drivers do not support connect/disconnect. With multiple drives and a correct driver, better rates should be observable. Wind River is working on a high performance driver for this board. You can also order one from Real Time Products Ltd. Contact pak@rtp.co.uk (Paul Keir). - Your throughput is limited by the drives you select. Eric. +--------------------------------------------------------------------------+ | Eric L. Schott "And when you get lost in | | The Pennsylvania State University life, know that the road of | | Office: (814) 863-4194 return to God is always open." | | Fax (814) 863-7841 Albert Schweitzer, | | http://www.personal.psu.edu/els6 Letter to god-child. | +--------------------------------------------------------------------------+ --------------------------- Newsgroups: comp.os.vxworks Subject: (G) UI development on VxWorks Date: Wed, 18 Sep 1996 20:14:00 +0500 From: Ramaswami K Viswanathan Organization: WIPROGE MEDICAL SYSTEMS Message-ID: <324011B8.4ECB@wiproge.med.ge.com> I would like to know about UI development tools available on VxWorks. -- esp. fast prototyping tools with code generating option. If I have a cross compilating environment from Solaris 2.4 to Vxworks, Can I use X/Motif Windows Code ? WindRiver Provides VxWindows - an X11R6 Server, Window Managers (twm, mwm) and a set of clients Plus a development library for Motif, Xt, and Xlib. Anybody has any pointers to its use? pros/cons? Or If I have a Windows-95 or NT platform, can the UI be cross compiled to VxWorks? Thanks in Advance! rAmA. Wipro GE Medical Systems, Bangalore, India. --------------------------- End of New-News digest ********************** From namiq@wrs.com Wed Oct 2 04:12:04 1996 From: namiq@wrs.com Date: Wed Oct 2 04:12:05 PDT 1996 Subject: Seagate Drives Hello Folks: I would appreciate some feedback for a customer from anyone who has had previous experience both good and bad in using the Seagate Barracuda 4 (4.29 GB formatted capacity) and Barracuda 9 (9.1 GB formatted capacity)Hard Disk Drives with VxWorks. Thanks in advance, Namiq Kunimoto -------------------------------------------------------------------------- Namiq Kunimoto Direct Tel#: 81-3-5467-5885 Director of Sales General Tel#: 81-3-5467-5900 Asia Pacific Operations Fax #: 81-3-5467-5877 WWW http://www.wrs.com Email: namiq@wrs.com Wind River Systems Pola Ebisu Bldg. 11F, 3-9-19 Higashi, Shibuya-ku, Tokyo 150, Japan -------------------------------------------------------------------------- From mmilde@scires.com Wed Oct 2 05:34:49 1996 From: Mike Milde Date: Wed Oct 2 05:34:52 PDT 1996 Subject: comp.os.vxworks newsdigest -Reply I'm currently writing a GUI for a VxWorks based embedded system using Java. A (really simple) HTTP server runs on our VME cards and contains all the Java class files. The Java applets open up UDP (could be TCP too if you like) connections back to a server running on the VME card. This all provides a GUI with real-time data on any platform that can run Netscape, etc. and you don't even have to go install your GUI on another system. Anyone else pursuing Java for their GUI? Mike Milde Scientific Research Corporation mmilde@scires.com --------------------------------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: (G) UI development on VxWorks Date: Wed, 18 Sep 1996 20:14:00 +0500 From: Ramaswami K Viswanathan Organization: WIPROGE MEDICAL SYSTEMS Message-ID: <324011B8.4ECB@wiproge.med.ge.com> I would like to know about UI development tools available on VxWorks. -- esp. fast prototyping tools with code generating option. If I have a cross compilating environment from Solaris 2.4 to Vxworks, Can I use X/Motif Windows Code ? WindRiver Provides VxWindows - an X11R6 Server, Window Managers (twm, mwm) and a set of clients Plus a development library for Motif, Xt, and Xlib. Anybody has any pointers to its use? pros/cons? Or If I have a Windows-95 or NT platform, can the UI be cross compiled to VxWorks? Thanks in Advance! rAmA. Wipro GE Medical Systems, Bangalore, India. From jhillman@wrs.com Wed Oct 2 06:53:55 1996 From: Jon Hillman Date: Wed Oct 2 06:53:57 PDT 1996 Subject: Re: Booting VxWorks from a PC At 01:51 PM 10/1/96 PDT, you wrote: >) >)We are trying to boot VxWorks 5.1 from a PC running Windows NT 3.51 using >)TFTP service that comes with the ChameleonNFS product. It appears to boot >)VxWorks to a runnable state but fails near the end of the transfer or in the >)closing of the connection. The result is that the connection is refused by >)the PC when the symbol table is attempted to be transferred. All we get is >)a VxWorks prompt with no ability to do anything since there is no symbol >)table. >) >)Does anyone have experience in trying something similar? >) >) > >This sounds like a problem where vxWorks is trying to open the same >socket number for the symblo table and the kernel file. The reason >this happens is the bootrom.sys is a limited kernel. When the full >kernel is loaded and trys to open a socket, it trys to open the same >socket because the full kernel erinitializes the initial socket number >for tftp. The problem is that the time for the NT system to close the >socket is longer that the time for the vxWorks bootrom.sys to jump to >the full vxWorks kernel and for that kernel to try to open the socket. >I have heard of this problem with other hosts. >======================8<---work---------------------- >e-mail: bwenholz@pacesetter.com (Bruce Wenholz) >Pacesetter, Inc Sylmar, Ca 91342, USA >======================8<---personal------------------ >e-mail: MMCo@DeltaNet.Com (Bruce Wenholz) > WEB: http://www.deltanet.com/users/mmco/essay.html Bruce is exactly right, and I can suggest at least one solution, and that would be to load the vxWorks.st file, in leiu of vxWorks and the vxWorks.sym file. Of course, you would need to INCLUDE_STANDALONE_SYM and exclude INCLUDE_NET_SYM_TBL (or whatever its named) and build the vxWorks.st file. _________________________ | Jon Hillman, FAE | | Wind River Systems | /) (407) 273-4646 (\ / ) Email:jhillman@wrs.com ( \ _( (|_________________________) ) /> (((\ \) /,) / ) / //))/ (\\\\ \_/ / \ \_/ ///// \ / \ / \ _/ \_ / ----/ /----------------------\ \---- / / \ \ From mea@mclean.sparta.com Wed Oct 2 07:00:19 1996 From: Mike Anderson Date: Wed Oct 2 07:00:23 PDT 1996 Subject: Re: Seagate Drives At 04:12 AM 10/2/96 PDT, you wrote: >Submitted-by namiq@wrs.com Wed Oct 2 04:12:04 1996 >Submitted-by: namiq@wrs.com > >Hello Folks: > >I would appreciate some feedback for a customer from anyone who has had >previous experience both good and bad in using the Seagate Barracuda 4 (4.29 >GB formatted capacity) and Barracuda 9 (9.1 GB formatted capacity)Hard Disk >Drives with VxWorks. > We use the 4GB Barracudas. The drives are awesome in terms of performance (especially if WRS will ever enable all of the SCSI-2 Fast/Wide support), but they run *very* hot. At 7800 RPM, the Barracuda will be too hot to touch within 5-10 minutes after power on. So, make sure you have plenty of cooling capacity. HTH, =============================================================================== __ Real-Time System Development, Integration, Training and Services //\\ // \\ Mike Anderson // /\ \\ Chief Engineer Voice : (703) 448-0210 ext. 235 // / \ \\ SPARTA, Inc. FAX : (703) 734-3323 // \ \\ 7926 Jones Branch Drive EMAIL : mea@mclean.sparta.com \\ \ // Suite 900 Web : http://www.mclean.sparta.com \\ \ / // McLean, VA 22102 \\ \/ // "Software development is like making \\ // a baby... You can't make a baby in one \\// month by impregnating nine women. -- "Pride in Performance" Some things just take time." =============================================================================== From mea@mclean.sparta.com Wed Oct 2 07:04:22 1996 From: Mike Anderson Date: Wed Oct 2 07:04:25 PDT 1996 Subject: Re: Questions about exceptions. At 04:57 PM 10/1/96 PDT, you wrote: >Submitted-by MikeMcQuade@connect.com Tue Oct 1 16:56:56 1996 >Submitted-by: Mike McQuade > > >Im using Tornado, and PowerPC. > >My application blows up, and spits this info to the Windshell, what is the >best way to interperet this info ? > >Does anyone know where to find out what the exception numbers mean ? > > > > >Exception number 768: Task: 0xdefbc8 (sendData) > >data access >program counter: 0x104834 >access address: 0x1000c90 >machine status: 0xb030 >14d800 vxTaskEntry +60 : xmitData ([]) >2578e8 xmitData +70 : fill_packet_buffer ([]) >257de4 fill_packet_buffer+258: bcopyWords ([]) > > >The tt command now works on PowerPC, Im not really sure how it works, can >anyone explain it in a nutshell ? > Greetings! I can't explain the exception numbers, but what this is telling you is that you had a bus error during your bcopyWords call when it hit address 0x1000c90. Now, if this address works *most* of the time and it happens to be located out on the VMEbus, then you've got a bus arbitration problem. Otherwise, you may be getting a bad pointer from a poorly debugged sysLocalToBusAdrs call or you may just have a bad pointer, period. HTH, =============================================================================== __ Real-Time System Development, Integration, Training and Services //\\ // \\ Mike Anderson // /\ \\ Chief Engineer Voice : (703) 448-0210 ext. 235 // / \ \\ SPARTA, Inc. FAX : (703) 734-3323 // \ \\ 7926 Jones Branch Drive EMAIL : mea@mclean.sparta.com \\ \ // Suite 900 Web : http://www.mclean.sparta.com \\ \ / // McLean, VA 22102 \\ \/ // "Software development is like making \\ // a baby... You can't make a baby in one \\// month by impregnating nine women. -- "Pride in Performance" Some things just take time." =============================================================================== From biocca@csg.lbl.gov Wed Oct 2 08:40:16 1996 From: biocca@csg.lbl.gov (Alan K Biocca) Date: Wed Oct 2 08:40:18 PDT 1996 Subject: VxWorks as a COTS experience ... I'm forwarding this to the exploder at their request. -- Alan K Biocca -- forwarded message -- Because we are interested to use VxWorks as a COTS (Commercial Off The Shelf) Software directly in ongoing projects, we would like you to forward these questions to the users Group so that we can share experiences. (any project, particularly critical ones or space ones). Within your experience in any project, our concerns mainly deal with the following questions : 1. What was your selection and procurement process for VxWorks ? 2. Did you need additional developments and/or adaptations ? How did you manage them ? 3. What was your policy for testing, verification, qualification and acceptance of this commercial product ? 4. Did you perform an engineering assessment and/or a software quality assessment (e.g. a SW quality audit)? Could you provide us with the results / report ? 5. Did you perform a "risks assessment", taking into account for example the following classes of risks : technical risks : functionalities, reliability, performances, integration, interoperability, ... quality risks : maintainability, documentation (completness, correctnes, ...), tests coverage, configuration management, ... maintenance risks : support, source code availability, ... ... We will be very interested in reading any experience on that subject. Please feel free to foreward this message to people that may give us some informations. Thank you very much for your contributions. Sincerily Yours, Bruno MEIJER my e-mail address is : meijer@mms.matra-espace.fr@umc@CMTMRP From kkoshimizu@scl.teainet.com Wed Oct 2 09:27:52 1996 From: "TEA Koshimizu, Kaz" Date: Wed Oct 2 09:27:54 PDT 1996 Subject: NFS vs FTP This might be repeated question but any input is appreciated. 1. NFS vs FTP We need to exchange relatively small (50-100Kb) of data with "file server" using e-net. Our choices are FTP or NFS (obviously). But I was told NFS is much efficient than FTP for vxworks's implementation. Does any one have specific comparison data ? or any information ? 2. NFS server for PC Does any one had bad experiace with specific implementation of NFS for PC ? Thank you in advance. Kazu Koshimizu Tokyo Electron America From leonid@rst.co.il Wed Oct 2 10:43:49 1996 From: leonid@rst.co.il (Leonid Rosenboim) Date: Wed Oct 2 10:43:51 PDT 1996 Subject: i960RP with VxWorks Has anobody done already a port of VxWorks 5.2 or 5.3 on to Intel i960RP processor on the Cyclone evaluation board, or on any other board. I know some of you have, and it looks that we will have to do this too on a very short notice. Hence, any tips diffs and advise I can get now may be a great help. Thanks in advance ! ----------------------------------------------------------------------- Leonid Rosenboim Phone: +972-3-559-8144 R S T Software Industries Ltd. Mobile: +972-50-307-142 P.O.Box 11502, AZUR 58017, Israel Fax: +972-3-559-8244 WWW: http://www.rst.co.il E-Mail: leonid@rst.co.il From jhillman@wrs.com Wed Oct 2 10:46:56 1996 From: Jon Hillman Date: Wed Oct 2 10:46:58 PDT 1996 Subject: Re: comp.os.vxworks newsdigest -Reply At 05:34 AM 10/2/96 PDT, Mike Milde wrote: > >I'm currently writing a GUI for a VxWorks based embedded system >using Java. A (really simple) HTTP server runs on our VME cards >and contains all the Java class files. The Java applets open up UDP >(could be TCP too if you like) connections back to a server running >on the VME card. This all provides a GUI with real-time data on any >platform that can run Netscape, etc. and you don't even have to go >install your GUI on another system. > >Anyone else pursuing Java for their GUI? > >Mike Milde >Scientific Research Corporation >mmilde@scires.com Mike, Are you aware that JavaOS is available for vxWorks? Wind River is the first embedded software company to sign a JavaOS licensing agreement. Contact me directly if this is something you're interested in and want to know more. _________________________ | Jon Hillman, FAE | | Wind River Systems | /) (407) 273-4646 (\ / ) Email:jhillman@wrs.com ( \ _( (|_________________________) ) /> (((\ \) /,) / ) / //))/ (\\\\ \_/ / \ \_/ ///// \ / \ / \ _/ \_ / ----/ /----------------------\ \---- / / \ \ From rich@immrc1.eng.mcmaster.ca Wed Oct 2 15:02:49 1996 From: rich@immrc1.eng.mcmaster.ca (Richard Teltz) Date: Wed Oct 2 15:02:51 PDT 1996 Subject: backplane TCP/IP Dear all, I posted a question a while ago regarding maximum bandwidth for the VxWorks TCP/IP implementation across the VMEbus. I recieved no response... Does anybody use this feature? What am I missing? Thanks in advance Rich *------------------------------------------------------------------* | INTELLIGENT MACHINES AND MANUFACTURING RESEARCH CENTRE (IMMRC) | *------------------------------------------------------------------* | Richard Teltz, | Dept. of Mechanical Engineering,| | principal research engineer, | McMaster University, | | tel: (905) 525-9140, x27351 | Hamilton, Ontario, | | fax: (905) 572-7944 | Canada, | | email: rich@immrc.mcmaster.ca | L8S 4L7 | *------------------------------------------------------------------* From aac!sol!applanix.com!babak@seraph.uunet.ca Wed Oct 2 15:14:31 1996 From: babak@applanix.com Date: Wed Oct 2 15:14:33 PDT 1996 Subject: unsubscribe unsubscribe From aac!sol!applanix.com!babak@seraph.uunet.ca Wed Oct 2 15:14:34 1996 From: babak@applanix.com Date: Wed Oct 2 15:14:36 PDT 1996 Subject: unsubscribe unsubscribe From seaweed!seaweed.com!bob@netcom.com Wed Oct 2 15:45:26 1996 From: bob schulman Date: Wed Oct 2 15:45:28 PDT 1996 Subject: Re: (G) UI development on VxWorks rama@wiproge.med.ge.com asks: > I would like to know about UI development tools available > on VxWorks. -- esp. fast prototyping tools with code generating > option. > > If I have a cross compilating environment from Solaris 2.4 > to Vxworks, Can I use X/Motif Windows Code ? Seaweed Systems also offers a complete X11R6 environment for vxWorks. Please contact info@seaweed.com or call us at +1 510.482-3575 for more information. bob schulman bob@seaweed.com From froeber@bbn.com Wed Oct 2 18:20:16 1996 From: Fred Roeber Date: Wed Oct 2 18:20:19 PDT 1996 Subject: Re: backplane TCP/IP On Wed, 2 Oct 1996, Richard Teltz asked: > I posted a question a while ago regarding maximum bandwidth > for the VxWorks TCP/IP implementation across the VMEbus. > > I recieved no response... > > Does anybody use this feature? What am I missing? I would guess lots of people use this feature but fewer people use it for time critical data transfer than used to now that there is VxMP. The thing is that the backplane network implementation tends to run SLOWER than an ethernet connection (at least when comparing transfers over a quite network and an idle backplane using a 68040 processor). This may seem counterintuitive until you look at the timing of the TCP/IP protocol stack. What we found was that the biggest bottleneck with TCP/IP is the CPU overhead in the protocol stack itself. When going over ethernet, the MAC layer of the transfer stack is usually handled by some dedicated hardware chip like a LANCE chip which uses a DMA engine to move the data onto the wire. With the backplane, the data is actually moved from the local CPU to a memory buffer on the backplane master using a bcopy operation on the local CPU. This is often slower than the ethernet chip can move the data (or at least not much faster since no DMA engine is involved). On top of that, there are synchronization actions required in the backplane driver and interprocessor interrupts which all add extra overhead. So the bottom line is not to expect much the backplane interface to be much faster than ethernet; say 600-700 KB/sec max. If you want high speed communication look into VxMP and actually switch to using shared message queues and semaphores. Fred | Fred J Roeber, BBN Systems & Technologies | | 50 Enterprise Place Middletown, RI 02842-5202 | | froeber@bbn.com 401-849-2543 (X48) | | See http://www.bbn.com/tv for TraceMaker info | From froeber@bbn.com Wed Oct 2 18:35:58 1996 From: Fred Roeber Date: Wed Oct 2 18:36:33 PDT 1996 Subject: Re: BOOTP port changes On Wed, 2 Oct 1996, Tim Snider asked: > We have a need to change the normal port number for boot request and reply > messages. Is there an interface, variable, hack that would allow us to do > this in vxWorks? If we had the source for the vxWorks bootp we'd just > modify that since we've already modified the servers bootp code to do other > unique things. I assume you looked at the bootpMsgSend function. I have used it in the past and it lets you switch the port you want to address the bootp request to on the server. There never used to be any way other than patching the object code to change the receive port that VxWorks expects to receive the bootp reply on, though. The code used an ethernet input hook routine that looked for a UDP input packet addressed to the standard BOOTP reply port (IPPORT_BOOTPC) to receive BOOTP replies. Fred | Fred J Roeber, BBN Systems & Technologies | | 50 Enterprise Place Middletown, RI 02842-5202 | | froeber@bbn.com 401-849-2543 (X48) | | See http://www.bbn.com/tv for TraceMaker info | From froeber@bbn.com Wed Oct 2 19:06:17 1996 From: froeber@bbn.com Date: Wed Oct 2 19:06:19 PDT 1996 Subject: Year 2000 and VxWorks I was wondering if anybody has worried about (and possibly addressed) the issue of handling dates in VxWorks applications past the year 2000? This hasn't been an issue in many of the systems we have done since most have used relative time values based on the system clock or other relative time source. One of the applications we are doing now, though, requires absolute time values. VxWorks "sort of" supports the POSIX/ANSI notion of absolute time by providing functions like "clock" that are supposed to handle time conversions. Since there is no standard BSP interface for supporting real-time clock (RTC) chips, though, there isn't really full POSIX support in this area. We are trying to add full support for time using the RTC chip on our SBC. The problem we have is that our RTC chip (like many RTC chips I am aware of) only provides a 2 digit year field. In trying to figure out how to address this problem we ran across the following comment in the VxWorks h/private/timeP.h include file: #define TM_YEAR_BASE 1900 /* Change this before the year 2000! */ This would make it appear that there is some issue with changing the VxWorks code so that it will work after the turn of the century. So, my question is whether anyone has come up with some clean way of implementing absolute time under VxWorks that will handle the century rollover issue? Thanks for any suggestions. Fred | Fred J Roeber, BBN Systems & Technologies | | 50 Enterprise Place Middletown, RI 02842-5202 | | froeber@bbn.com 401-849-2543 (X48) | | See http://www.bbn.com/tv for TraceMaker info | From jjs@mclean.sparta.com Wed Oct 2 19:21:41 1996 From: "John J. Smith" Date: Wed Oct 2 19:21:43 PDT 1996 Subject: Re: backplane TCP/IP At 06:20 PM 10/2/96 PDT, you wrote: >Submitted-by froeber@bbn.com Wed Oct 2 18:20:16 1996 >Submitted-by: Fred Roeber > >On Wed, 2 Oct 1996, Richard Teltz asked: > >> I posted a question a while ago regarding maximum bandwidth >> for the VxWorks TCP/IP implementation across the VMEbus. >> >> I recieved no response... >> >> Does anybody use this feature? What am I missing? > >I would guess lots of people use this feature but fewer people use it for >time critical data transfer than used to now that there is VxMP. The thing >is that the backplane network implementation tends to run SLOWER than an >ethernet connection (at least when comparing transfers over a quite network >and an idle backplane using a 68040 processor). Lots of stuff deleted... Sorry Fred Fred is absolutely right about the SM algorithm slowing down the transfer rate. There IS an awful lot of overhead involved, making CPU speed a major determining factor in throughput rate. In addition, bus access speed varies quite a bit from VME interface to VME interface, further clouding the issue. While testing our SM for Solaris product (sorry, couldn't avoid the plug), I found that throughput ranged from a low of about 600KB/s to well over 2.5MB/s, depending on which cards I tested. I also found that some cards were much better at hosting (exporting) the shared memory block than others, and made a HUGE difference in the throughput rate. Note that all tests were done using blaster -> blastee software, and I never tested anything slower that a 68040. There are other factors that also contribute to throughput rate such as signalling method (mailbox, interrupt, or polling) and TAS method (soft or hard). Given the number of variables associated with the original question, It's no wonder that nobody wanted to respond. I would suggest that this question is best answered by setting up a system and trying it. Any other type of performance estimate for sm is basically a WAG. Regards, _________________________________________________________ | |\ | John J Smith | Voice -> 703 448 1683 x253 | | | jjs@mclean.sparta.com | FAX -> 703 893 5494 | | | | | | Web -> http://www.mclean.sparta.com | | |_________________________________________________________| | | | | | /\\ 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 daemon@csg.lbl.gov Thu Oct 3 04:00:30 1996 From: daemon@csg.lbl.gov Date: Thu Oct 3 04:00:33 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Thu Oct 3 04:00:27 PDT 1996 Subject: question about VM62 memory size Subject: main() in VxWorks ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: question about VM62 memory size Date: Wed, 2 Oct 1996 15:10:09 +0100 From: Bartlomiej Piwek Organization: University of Mining & Metallurgy Message-ID: Could anonone tell me how to increase the amount of VM62 memory. I have VM62 with 16Mb of RAM, but I'm able to use only 4Mb. Is it possible that amount of RAM depends on settings in the kernel? Where are this settings? I tried to change LOCLAL_MEM_SIZE from 0x00400000 to 0x01000000 in config.h, but after this chage the computer was unable to boot here is what i got: Attaching network interface pep0... done. Attaching network interface lo0... done. Loading... 422968 + 90636 + 35454 Starting at 0x2000... Exception before kernel initialized: Access Fault Program Counter: 0x0001566a Status Register: 0x3704 Access Address : 0x00fffea4 Fault Status : 0x00850010 Regs at 0x3f88fc Are there any other settings I fogot about? Thanks In Advance, Bartek Piwek Dept. of Automatic Control, UMM, Krakow, Poland. --------------------------- Newsgroups: comp.os.vxworks Subject: main() in VxWorks Date: 2 Oct 1996 03:45:16 GMT From: bao@bnr.ca (Bao Dang) Organization: Bell Northern Research Message-ID: <52sogc$6ck@bmtlh10.bnr.ca> Reply-To: bdang@nortel.ca I have to port a number of Unix programs to VxWorks. These programs are ANSI-C so I don't expect a lot of modifications (I may be wrong ! I'm new with VxWorks). One thing that I have been able to figure out is how to deal with the main() function. Here are a number of approaches that I can think of: 1. Rename main to main_module_name, and write a "start" function which packages the arguments required by "module" into argc and argv pair and calls main_module_name with this argc and argv. 2. Completely change the interface of main. That is, replace main(argc, argv) with, for example, module_entry(int option_flags, int arg1, char *arg2), depending on what kind of arguments a program expects. I prefere a solution with the least of modification required, since I'd like to keep these program portable between Unix and VxWorks. Any suggestions and comments are very much appreciated. - -- - -------------------------------------------------------------------- Bao Dang Nortel Member of Scientific Staff 16 Place du Commerce Tel.: (514) 765-7722 6V44, Mail Stop 062 Fax.: (514) 761-8710 Ile des Soeurs, Verdun, QC Email: bdang@nortel.ca Canada H3E 1H6 - -------------------------------------------------------------------- --------------------------- End of New-News digest ********************** From @positron.cti-pet.com:wittner%pcmail.cti-pet.com@cti-pet.com Thu Oct 3 05:13:51 1996 From: Eli Wittner Date: Thu Oct 3 05:13:54 PDT 1996 Subject: Re: NFS vs FTP Mr. Koshimizu wrote: >Does any one have specific comparison data ? or any information ? A problem we have experienced using NFS with vxWorks is that vxWorks does not support the documented rename() function, which we were going to use to backup a file prior to recreating it. WindRiver acknowledged this as a documented problem that they did not plan to fix. Otherwise, we have had good luck with NFS/vxWorks. This was with a Sun/NFS Host. Cheers -- Eli Wittner CTI Cyclotron Systems, Inc. From evdsande@emt.ie.philips.nl Thu Oct 3 05:45:30 1996 From: "E. v.d. Sande" Date: Thu Oct 3 05:45:32 PDT 1996 Subject: running vxworks on a intel 386ex based controller board Does anybody have some experiences in running vxWorks on a intel 80386EX based system? We are using Tornado / vxWorks version 5.3 on a win95 host and the target is a 80386EX based controller board on which we use a modified ev386EX board support package. Debugging is done by connecting the wdbagent over a serial line, which turn out to be slow. Further we have trouble using the gnu C compiler shipped with Tornado on a PC-NFS network, include file are not found on nfs drives. Does anybody use a PC104 network card on a 368EX based board for connecting with the Tornado targetserver ? Thanks in advance ! Best Regards Eric van der Sanden _______________________________________________________ Philips Electronic Manufacturing Technology B.V. Ing. E. van der Sanden Electrical Development Building BAF-110 P.O.box 218 5600 MD Eindhoven The Netherlands Tel: +31 40 2722762 Fax: +31 40 2724610 E-mail: evdsande@emt.ie.philips.nl _______________________________________________________ From mea@mclean.sparta.com Thu Oct 3 07:17:33 1996 From: Mike Anderson Date: Thu Oct 3 07:17:35 PDT 1996 Subject: Re: backplane TCP/IP At 03:02 PM 10/2/96 PDT, you wrote: >Submitted-by rich@immrc1.eng.mcmaster.ca Wed Oct 2 15:02:49 1996 >Submitted-by: rich@immrc1.eng.mcmaster.ca (Richard Teltz) > > >Dear all, > >I posted a question a while ago regarding maximum bandwidth >for the VxWorks TCP/IP implementation across the VMEbus. > >I recieved no response... > >Does anybody use this feature? What am I missing? > >Thanks in advance > Greetings! Well, as is generally the case, the answer is... it depends. With WRS's out-of-the-box configuration on a 33 MHz 68040, you can expect ~600K - 1MByte/sec TCP/IP (stream socket) kind of performance (depending on *whose* 68040 you use). Here at SPARTA, we've tweeked many of the parameters in our SM for Solaris port and it's not unusual to see 2.5x that performance using VME32. Our MVP port of the Quad HyperSPARC Themis boards sees 6+ MBytes/sec on the Themis MBus. Also, if you get the source to SM, then you can tweek it even more to incorporate VME64 MBLT transfers, the DMA engines from many of the VME chip manufacturers and other games using combination VME/VSB backplanes. In fact, I've heard rumors of data rates in the teens of MBytes/sec for SM in those conditions. However, remember that if you start making those kind of mods to SM, that your compatibility with other VxWorks targets that, for instance, don't support VME64, goes out the window. So, as a point solution to get a specific hardware combination running with greater data rates, buy the SM source and hack away ;-). But, given the multitude of hardware implementations out there, WRS has taken the path of lowest common denominator (rightfully so, I believe in this case) and they deliver an SM implementation that can out run an Ethernet, but just barely. But, if you want real speed, then use WRS's VxMP product. You'll trade the socket interface for semaphores, pipes, and message queues, but if you have the need for speed... HTH, =============================================================================== __ Real-Time System Development, Integration, Training and Services //\\ // \\ Mike Anderson // /\ \\ Chief Engineer Voice : (703) 448-0210 ext. 235 // / \ \\ SPARTA, Inc. FAX : (703) 734-3323 // \ \\ 7926 Jones Branch Drive EMAIL : mea@mclean.sparta.com \\ \ // Suite 900 Web : http://www.mclean.sparta.com \\ \ / // McLean, VA 22102 \\ \/ // "Software development is like making \\ // a baby... You can't make a baby in one \\// month by impregnating nine women. -- "Pride in Performance" Some things just take time." =============================================================================== From tkb@mclean.sparta.com Thu Oct 3 08:42:31 1996 From: "Thomas K. Buchanan" Date: Thu Oct 3 08:42:34 PDT 1996 Subject: Re: backplane TCP/IP At 03:02 PM 10/2/96 PDT, you wrote: >Submitted-by rich@immrc1.eng.mcmaster.ca Wed Oct 2 15:02:49 1996 >Submitted-by: rich@immrc1.eng.mcmaster.ca (Richard Teltz) > > >Dear all, > >I posted a question a while ago regarding maximum bandwidth >for the VxWorks TCP/IP implementation across the VMEbus. > >I recieved no response... > It depends upon many things, for example: - polling, mailbox or bus interrupt notification to receivers - VME bus arbitration setup - VME bus release setup - VME bus bandwidth used by other applications - size/speed of shared memory used to store packets - CPU bandwidth of nodes being measured - VxWorks scheduling setup of application vs netTask I have seen a range of values from 600 KBytes/s to almost 7 MBytes/s. adios ----------------------------------------- /\\ Keith Buchanan ///\\ /////\\ SPARTA, Inc. ///// \\ 7926 Jones Branch Drive / \\\\\ \\ Suite 900 \ \\\\\ // McLean, VA 22102 \ ////// \////// Voice: 703 448 0210 \//// FAX: 703 893 5494 \// WWW: www.mclean.sparta.com From pean@mbari.org Thu Oct 3 08:48:50 1996 From: Andrew Pearce Date: Thu Oct 3 08:48:52 PDT 1996 Subject: Re: NFS vs FTP Hi, we have noticed other problems with NFS. On our configuration, HP hosts and Motorola 162 CPU's, vxWorks is unable to follow symbolic links between files. We have also had a problem with NFS timeouts when the client CPU is connected to the host via another vxWorks CPU acting as a gateway. This problem causes an application that writes data to a file on the host using NFS to take the amount of time needed to transfer the data plus N times the NFS timeout period. For example, with an NFS timeout period of 5 seconds, fwrite of 8K bytes takes 0.8, 5.8 10.8, 15.8 or even 20.8 seconds. This was reported as TSR#29927 and we use ftp as a workaround for this problem. -- Andrew Pearce Monterey Bay Aquarium Research Institute tel: (408) 775-1746 7700 Sandholt Rd fax: (408) 775-1646 Moss Landing, CA 95039 email: pean@mbari.org "Coping's easy. Not pureeing your loved ones, that's the hard part." Basil Fawlty From tkb@mclean.sparta.com Thu Oct 3 08:56:10 1996 From: "Thomas K. Buchanan" Date: Thu Oct 3 08:56:12 PDT 1996 Subject: Re: backplane TCP/IP At 06:20 PM 10/2/96 PDT, you wrote: >Submitted-by froeber@bbn.com Wed Oct 2 18:20:16 1996 >Submitted-by: Fred Roeber > >On Wed, 2 Oct 1996, Richard Teltz asked: > >> I posted a question a while ago regarding maximum bandwidth >> for the VxWorks TCP/IP implementation across the VMEbus. >> >> I recieved no response... >> >> Does anybody use this feature? What am I missing? > >I would guess lots of people use this feature but fewer people use it for >time critical data transfer than used to now that there is VxMP. The thing >is that the backplane network implementation tends to run SLOWER than an >ethernet connection (at least when comparing transfers over a quite network >and an idle backplane using a 68040 processor). This may seem The 68040 can push data a wee bit faster than Ethernet in tests we have conducted. An important thing to remember is that the speed will scale up with faster processors, VME64 etc. Ethernet is electrically limited to ~1 MByte/s no matter what the CPU is capable of. Ethernet is also a multiple access single medium setup. The data structures that SM uses make it similar to having a multiple access multiple media setup. Only one VME transaction can take place at a time but the maximum throughput of VME32 is 40 MBytes/s ( 80 MBytes/s MVE64 ) and is a secondary consideration unlike Ethernet. If you really need to optimize it, the source can be optimized to use VME block transfer modes. adios ----------------------------------------- /\\ Keith Buchanan ///\\ /////\\ SPARTA, Inc. ///// \\ 7926 Jones Branch Drive / \\\\\ \\ Suite 900 \ \\\\\ // McLean, VA 22102 \ ////// \////// Voice: 703 448 0210 \//// FAX: 703 893 5494 \// WWW: www.mclean.sparta.com From schen3@ccgate.hac.com Thu Oct 3 10:04:35 1996 From: schen3@ccgate.hac.com Date: Thu Oct 3 10:04:37 PDT 1996 Subject: Need SLIP and ethernet driver for Solaris 2.5 Is there anyone who has SLIP driver and ethernet driver for Solaris 2.5? Thanks, Steven schen3@ccgate.hac.com From Scott.Peters@agai.com Thu Oct 3 13:02:59 1996 From: Scott Peters Date: Thu Oct 3 13:03:33 PDT 1996 Subject: Floating Point Operations Hello Vx'rs: When I add two floats together I get strange results. Attached is the coding segment. Has anyone seen this problem before? We are using a Motorla MVME162FX. What's the secret ? Thanks Scott Peters scott.peters@agai.com ---- #include #include #include void frog() { float x = 10.0; int i; for (i=0;i<10;i++) { x = x + (float) 0.1; printf("x=%f\n",x); } } out put: ... x=10.300001 x=10.400002 x=10.500002 x=10.600002 x=10.700003 x=10.800003 x=10.900003 x=11.000004 x=11.100004 x=11.200005 ... From carl@themis.com Thu Oct 3 13:59:56 1996 From: Carl C. Chesbrough Date: Thu Oct 3 13:59:59 PDT 1996 Subject: Re: Floating Point Operations Scott, This is because of the precision of the float numbers. I get the same results running this on a SPARC10 machine. Try changing the float to a double. Unless you need more decimal places this should take care of the problem. HTH, -Carl. ******************************************************************** * * * Carl C. Chesbrough Telephone: (510) 252-0870 x 126 * * Themis Computer Facsimile: (510) 490-5529 * * 3185 Laurelview Court E-Mail : carl@themis.com * * Fremont, California 94538 * * * ******************************************************************** > Hello Vx'rs: > > When I add two floats together I get strange results. Attached is the > coding segment. Has anyone seen this problem before? We are using a > Motorla MVME162FX. What's the secret ? > > Thanks > Scott Peters > scott.peters@agai.com > ---- > #include > #include > #include > void frog() > { > float x = 10.0; > int i; > for (i=0;i<10;i++) > { > x = x + (float) 0.1; > printf("x=%f\n",x); > } > } > > out put: > ... > x=10.300001 > x=10.400002 > x=10.500002 > x=10.600002 > x=10.700003 > x=10.800003 > x=10.900003 > x=11.000004 > x=11.100004 > x=11.200005 > ... > > ********** From toprakci@venus.aselsan.com.tr Fri Oct 4 01:22:09 1996 From: Yalcin TOPRAKCI Date: Fri Oct 4 01:22:11 PDT 1996 Subject: Unsubscribe unsubscribe From daemon@csg.lbl.gov Fri Oct 4 04:00:34 1996 From: daemon@csg.lbl.gov Date: Fri Oct 4 04:00:36 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Fri Oct 4 04:00:31 PDT 1996 Subject: Re: VxWorks on DSPs? ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks on DSPs? Date: 3 Oct 1996 12:31:38 GMT From: Alex Goodyear Organization: JET Joint Undertaking Message-ID: <530bna$se3@postman.jet.uk> References: <3252614B.71E3@equator.com> As far as I am aware VxWorks is only available on CISC, RISC and SPARC based processors. Maybe you should take a look at SPOX, it is only available for DSP chips but VxWorks does have a communication BSP for use with SPOX. Alex. =============================================================================== The above article is the personal view of the poster and should not be considered as an official comment from the JET Joint Undertaking =============================================================================== --------------------------- End of New-News digest ********************** From jayr@protocol.com Fri Oct 4 07:52:32 1996 From: Jay Rasmussen Date: Fri Oct 4 07:52:34 PDT 1996 Subject: Tornado and Source Control Hello, I am curious as to which version control software packages people are successfully using in conjunction with Tornado/VxWorks. I am looking at a couple of separate projects, perhaps some shared code between them, about 20 S/W engineers and both Win/NT and Sun development platforms. Any recommendations, warnings or advice are appreciated. Thanks, Jay Rasmussen jayr@protocol.com From robin.hearn@gecm.com Fri Oct 4 08:01:03 1996 From: Robin Hearn +44 1935 442109 Date: Fri Oct 4 08:01:05 PDT 1996 Subject: Ada/VxWorks Bindings Hi there, We're trying to use VisiComms X-Windows package for VxWorks, the problem is the preferred programming language is Ada (Green Hills Ada on the PowerPC). Has anyone created bindings for Ada to either X/Motif or VxWorks? Any hints, tips or problems? Any information/help will be appreciated. Thanks in anticipation Robin ------------------------------------------------------------------------------- Robin Hearn Thomson Marconi Sonar Limited Wilkinthroop House, e-mail: robin.hearn@gecm.com Templecombe, tel: +44 (0) 1935 442109 Somerset, BA8 0DH, England fax: +44 (0) 1935 442185 ------------------------------------------------------------------------------- From pean@mbari.org Fri Oct 4 09:10:41 1996 From: Andrew Pearce Date: Fri Oct 4 09:10:43 PDT 1996 Subject: Re: Tornado and Source Control Hi, Source and revision control systems that work with Tornado are of interest to us here too. One tool that I have been looking at recently is CVS. Source code for CVS is available on the net, subject to the Free Software Foundation rules for software use and copying. Its been ported to many host environments, including Sun and HP and uses rcs to implement the underlying revision control methods. So far, we have it installed, but I haven't had time to take it for a test drive. -- Andrew Pearce Monterey Bay Aquarium Research Institute tel: (408) 775-1746 7700 Sandholt Rd fax: (408) 775-1646 Moss Landing, CA 95039 email: pean@mbari.org "Coping's easy. Not pureeing your loved ones, that's the hard part." Basil Fawlty From tkb@mclean.sparta.com Fri Oct 4 11:01:43 1996 From: "Thomas K. Buchanan" Date: Fri Oct 4 11:01:45 PDT 1996 Subject: Re: Tornado and Source Control At 09:10 AM 10/4/96 PDT, you wrote: >Submitted-by pean@mbari.org Fri Oct 4 09:10:41 1996 >Submitted-by: Andrew Pearce > >Hi, > >Source and revision control systems that work with Tornado are >of interest to us here too. One tool that I have been looking at recently We use RCS in Unix environments and PVCS in Win32 environments with no problems. We have not attempted to integrate into an IDE but both tools are integrated into our editors of choice ( Emacs & SlickEdit ) and work pretty well. adios ----------------------------------------- /\\ Keith Buchanan ///\\ /////\\ SPARTA, Inc. ///// \\ 7926 Jones Branch Drive / \\\\\ \\ Suite 900 \ \\\\\ // McLean, VA 22102 \ ////// \////// Voice: 703 448 0210 \//// FAX: 703 893 5494 \// WWW: www.mclean.sparta.com From joeco@dy4.com Fri Oct 4 11:25:06 1996 From: joeco@dy4.com (Joe Coballe) Date: Fri Oct 4 11:25:08 PDT 1996 Subject: Re: Ada/VxWorks Bindings On Fri, Oct 4, 1996, Robin Hearn asked: > We're trying to use VisiComms X-Windows package for VxWorks, the problem is the > preferred programming language is Ada (Green Hills Ada on the PowerPC). Has > anyone created bindings for Ada to either X/Motif or VxWorks? > Any hints, tips or problems? > I suggest looking into VADSworks from Rational Software (www.rational.com). It comes with a library(source included) of ADA bindings for most of the VxWorks libraries. Joe __________________________________________________________________________ | ***** **** | | Joseph Coballe **** ** *** **** **** DY 4 Systems Inc | | Software Engineer *** **** ** **** **** * 333 Palladium Dr | | Tel : 613-599-9199 x299 ** ***** * **** **** ** Kanata, ON | | Fax : 613-599-7777 * ***** **** **** *** Canada | | Email : jcoballe@dy4.com **** **** K2V 1A6 | | | |_________________________ CUSTOMER FIRST, QUALITY ALWAYS _________________| From dlin@elvis.ds.boeing.com Fri Oct 4 16:17:53 1996 From: dlin@elvis.ds.boeing.com (Dongliang Lin) Date: Fri Oct 4 16:17:54 PDT 1996 Subject: PMC(PCI Mezzanine Card) driver on AXPvme-160 SBC Hi, Does anyone have PMC drivers for AXPvme card. We designed on PMC module for the AXPvme. I need help(some example program) on the driver. Thanks. Don Lin Boeing Company From thomas@vmetro.com Sat Oct 5 00:00:15 1996 From: Thomas Nygaard Date: Sat Oct 5 00:00:18 PDT 1996 Subject: VME board has i960RP with VxWorks > Submitted-by leonid@rst.co.il (Leonid Rosenboim) Wed Oct 2 10:43:49 19= 96 > =20 > Has anobody done already a port of VxWorks 5.2 or 5.3 on to Intel > i960RP processor on the Cyclone evaluation board, or on any other board= ? VMETRO has developed a family of VMEbus boards called "MIDAS" that includ= e one or two i960RP processors running VxWorks 5.3. These boards are design= ed to operate as a PMC I/O Subsystem for VMEbus and RACEway (PMC=3DPCI Mezza= nine Card). The VxWorks port is completed. First customer shipments and formal product announcement will take place later this month.=20 VxWorks drivers for PMC modules offering the following functions are in various stages of development: - Digital Parallel I/O (FPDP) - Fiber Channel - SCI - Ethernet For more information, look at VMETRO=B4s homepage: www.vmetro.com or contact addresses below: USA/Canada:=20 Leif E. Laerum VMETRO Inc., Houston, TX Email: info@vmetro.com Tel: (713)584-0728 Fax: (713)584-9034 =20 Europe/Asia:=09 Thomas Nygaard VMETRO A/S, Oslo, Norway Email: thomas@vmetro.com Tel : + 47 2210 6090 Fax : + 47 2210 6202 Best regards, Thomas Nygaard VMETRO A/S Nedre Rommen 5 N-0988 Oslo, Norway Ph. +47 2210 6090 Fax +47 2210 6202 Email: thomas@vmetro.com From daemon@csg.lbl.gov Sat Oct 5 04:00:58 1996 From: daemon@csg.lbl.gov Date: Sat Oct 5 04:01:00 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sat Oct 5 04:00:55 PDT 1996 Subject: Re: RPC and VxWorks Subject: Re: Booting VxWorks from a PC ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: RPC and VxWorks Date: Sun, 22 Sep 1996 08:59:08 GMT From: noblenet@world.std.com (NobleNet Inc.) Organization: NobleNet, Inc. Message-ID: References: <1.5.4.32.19960921082737.0070552c@pop01.connect.com> Reposting article removed by rogue canceller. In article <1.5.4.32.19960921082737.0070552c@pop01.connect.com>, Mike McQuade wrote: >Im trying to use RPC to run functions on my VxWorks target, from a Unix >application, and XDR to send data structs to the target. > >does anyone know of, or have any sample code for RPC ? Please contact NobleNet for ONC/RPC tools and solutions for UNIX, Windows (3.1/95/NT), Macintosh, VMS, OS/2, NetWare, VxWorks, etc. NobleNet, Inc. 337 Turnpike Road Southboro, MA 01772 Voice +1 508 460 8222 Fax +1 508 460 3456 sales@noblenet.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Booting VxWorks from a PC Date: 2 Oct 1996 18:19:20 GMT From: poirot@pendragon.jsc.nasa.gov (Daniel Poirot) Organization: NASA Johnson Space Center Message-ID: References: <199610012051.NAA02029@greenbay.Pacesetter.Com> On Tue, 1 Oct 1996 13:51:33 -0700, Bruce Wenholz wrote: >) >)Submitted-by GLWICKS@sandia.gov Tue Oct 1 11:11:00 1996 >)Submitted-by: "Wickstrom, Gregory L 2671 M" >) >) >)VxWorks to a runnable state but fails near the end of the transfer or in the >)closing of the connection. The result is that the connection is refused by >)the PC when the symbol table is attempted to be transferred. All we get is >)a VxWorks prompt with no ability to do anything since there is no symbol >)table. >This sounds like a problem where vxWorks is trying to open the same >socket number for the symblo table and the kernel file. The reason Have you tried booting vxWorks.st, the vxWorks image with the symbol table compiled in? - -- Daniel Poirot dan.t.poirot1@jsc.nasa.gov Metrica/NASA-JSC http://tommy.jsc.nasa.gov/~poirot Mail Code: ER6 tel: (713)483-8793 Houston, TX 77058 fax: (713)483-3204 --------------------------- End of New-News digest ********************** From currie@ghgcorp.com Mon Oct 7 06:37:03 1996 From: Alan Currie Date: Mon Oct 7 06:37:05 PDT 1996 Subject: Interrupt time What is a reasonable time for my board to spend handling interrupts? Using spy (vxWorks 5.2), my board (DY4 68040) reports that it is using 17% of its time handling interrupts. We are using 1553 interrupts so that is some of the reason it is higher than expected. But what should the design limit be? Thanks. From intrepid!minidds@uunet.uu.net Mon Oct 7 16:30:31 1996 From: intrepid!minidds@uunet.uu.net (Lee Kappel) Date: Mon Oct 7 16:30:34 PDT 1996 Subject: Re: Ada/VxWorks Bindings >On Fri, Oct 4, 1996, Robin Hearn asked: > > We're trying to use VisiComms X-Windows package for VxWorks, the problem is the > preferred programming language is Ada (Green Hills Ada on the PowerPC). Has > anyone created bindings for Ada to either X/Motif or VxWorks? > Any hints, tips or problems? > We have created Ada bindings for the VxWorks library routines which we need using Greenhills Ada. This is easy and painless to do. A sample of such a binding follows: --****************************************************************************** -- -- Package VX_KERNEL_SERVICES -- -- File Name : VX_KERNEL_SERVICES_.ADA -- -- User Information : This package encapsulates the VxWorks Ada bindings -- for the following VxWorks Libraries: -- -- kernelLib -- tickLib -- vxLib -- --****************************************************************************** with SYSTEM; with VX_TYPES; package VX_KERNEL_SERVICES is ----------------------------------------------------------------------------- -- Subprogram Declarations ----------------------------------------------------------------------------- procedure kernelInit (rootRtn : in SYSTEM.ADDRESS; rootMemSize : in INTEGER; pMemPoolStart : in SYSTEM.ADDRESS; pMemPoolEnd : in SYSTEM.ADDRESS; intStackSize : in INTEGER; lockOutLevel : in INTEGER ); pragma INTERFACE (C,kernelInit,"kernelInit"); ----------------------------------------------------------------------------- function kernelVersion return SYSTEM.ADDRESS; pragma INTERFACE (C,kernelVersion,"kernelVersion"); ----------------------------------------------------------------------------- function kernelTimeSlice (ticks : in INTEGER) return VX_TYPES.STATUS; pragma INTERFACE (C,kernelTimeSlice,"kernelTimeSlice"); ----------------------------------------------------------------------------- procedure tickAnnounce; pragma INTERFACE (C,tickAnnounce,"tickAnnounce"); ----------------------------------------------------------------------------- procedure tickSet (ticks : in INTEGER); pragma INTERFACE (C,tickSet,"tickSet"); ----------------------------------------------------------------------------- function tickGet return INTEGER; pragma INTERFACE (C,tickGet,"tickGet"); ----------------------------------------------------------------------------- function vxTas (address : in SYSTEM.ADDRESS) return BOOLEAN; pragma INTERFACE (C,vxTas,"vxTas"); ----------------------------------------------------------------------------- function vxMemProbe (adrs : in SYSTEM.ADDRESS; mode : in INTEGER; length : in INTEGER; pVal : in SYSTEM.ADDRESS ) return VX_TYPES.STATUS; pragma INTERFACE (C,vxMemProbe,"vxMemProbe"); ----------------------------------------------------------------------------- function vxMemProbeAsi (adrs : in SYSTEM.ADDRESS; mode : in INTEGER; length : in INTEGER; pVal : in SYSTEM.ADDRESS; adrsAsi : in INTEGER ) return VX_TYPES.STATUS; pragma INTERFACE (C,vxMemProbeAsi,"vxMemProbeAsi"); end VX_KERNEL_SERVICES; ================================================================================ Lee Kappel Raytheon E-Systems Communications Division kappel@eci-cec.com From daemon@csg.lbl.gov Tue Oct 8 04:00:20 1996 From: daemon@csg.lbl.gov Date: Tue Oct 8 04:00:22 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Tue Oct 8 04:00:17 PDT 1996 Subject: Re: Wind River Web Site Subject: Re: Newbie ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Wind River Web Site Date: Mon, 07 Oct 1996 06:44:53 -0400 From: Richard Price Organization: United Space Alliance Message-ID: <3258DF25.30B4@lsoc114x.ksc.nasa.gov> References: <53371j$8il@lantana.singnet.com.sg> user wrote: > > I would like to know more about VxWork, can anyone tell me the address of Wind River Web Site. > > My Email address is : chuayh@pacific.net.sg > > Thanks in advance.wind river url is www.wri.com richard in titusville --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Newbie Date: 7 Oct 1996 15:42:05 -0700 From: kla@leland.Stanford.EDU (Earl Mitchell) Organization: Stanford University, CA 94305, USA Message-ID: <53c0vt$dgr@elaine25.Stanford.EDU> References: <324BE468.41C67EA6@mc.com> In article <324BE468.41C67EA6@mc.com>, Matt Sheppard wrote: >I'm just learning VXworks and am looking for recommended references. > >I've got the programmers guide, and it seems pretty good. > >Anyone have any others? Yeah. Now checkout the reference manual you got with the programmer's guide. ;-) --------------------------- End of New-News digest ********************** From EJaquith@mis-resources.com Tue Oct 8 06:35:23 1996 From: Eric Jaquith Date: Tue Oct 8 06:35:25 PDT 1996 Subject: V x W o r k s P R O J E C T W A N T S U ==> Read This! M I S R E S O U R C E S I N T E R N A T I O N A L Information Technology Consultants http://www.mis-resources.com ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ This is only a partial listing of positions available. Multiple positions open! pOSITION: Systems Engineer / Systems Programmer-Developer LOCATION: Charlotte, NC WAGES: $32-$45/hr. long-term contract SKILLS: Real-Time, RTOS, VxWorks, C programming, Control Systems Description: You will be involved in the design, software development, debug, and implementations of VxWorks based control systems. Some weekend and/or after hours work will be expected. This is a long-term contract project with possible contract extension. . Good communication skills, team environment desirable. Qualifications: You must have three plus years industry experience with RTOS (real-time operating systems) and current skills work with VxWorks specifically. BSEE, BSCS, MSEE, or equivalent work experience required. DATE POSTED: 10/8/96 PROJECT #: 1026 Resources International is an information technology consulting firm with corporate offices in Atlanta, Georgia. We provide clients with contract software development and technical supplemental staffing. Depending on the project, consultants may work at client sites throughout the southeast or completely off-site in our development center located in Roswell, GA. Consultants are either W-2 hourly contract employees or salaried permanent employees with benefits and paid bench time between projects. Both long and short term projects are currently available. ARE YOU INTERESTED, BUT IN A HURRY? We understand... Just cut & paste this profile and email back now. (NO RESUME REQUIRED) - - - - - - - - PROFESSIONAL PROFILE (CONFIDENTIAL) - - - - - - - - - <><><><><><><><><><><> Project #:1026 <><><><><><><><><><><> NOTE: Primary skills are languages, tools, software & platforms. FOR EXAMPLE: C/C++, UNIX, Visual C++, OOP, GUI ___ YES! I'd like to SUBSCRIBE to the CURRENT_JOBS E-MAIL LIST. Name: ______________________________________________________ Primary Skills: ___________________________________ Yrs.____ Secondary Skills: _________________________________ Yrs.____ Telephone (Day): ___________________________________________ Telephone (Night): _________________________________________ Geographical Preference: ___________________________________ Positions of Interest: _____________________________________ Total Years Experience: ____________________________________ OR send us your resume at: _/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/ _/ _/ _/ Resources International, Inc. Contact: Eric Jaquith _/ _/ 570 Colonial Park Dr., #301 Toll Free: (800) 486-1704 _/ _/ Roswell, GA 30075 Voice: (770) 587-7500 _/ _/ E-mail: ejaquith@mis-resources.com Fax: (770) 594-1932 _/ _/ _/ _/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/ Please paste your ASCII resume directly into an e-mail message. You must include the Project # and Position Title in the SUBJECT LINE. # # # From uucp@relay1.fnet.fr Tue Oct 8 08:15:45 1996 From: leverone@zeus.egt.fr (Philippe Leverone) Date: Tue Oct 8 08:15:48 PDT 1996 Subject: Re: Floating Point Operations Scott Peters wrote : ] Hello Vx'rs: ] ] When I add two floats together I get strange results. Attached is the ] coding segment. Has anyone seen this problem before? We are using a ] Motorla MVME162FX. What's the secret ? ] ] Thanks ] Scott Peters ] scott.peters@agai.com ] ---- ] #include ] #include ] #include ] void frog() ] { ] float x = 10.0; ] int i; ] for (i=0;i<10;i++) ] { ] x = x + (float) 0.1; ] printf("x=%f\n",x); ] } ] } ] ] out put: ] ... ] x=10.300001 ] x=10.400002 ] x=10.500002 ] x=10.600002 ] x=10.700003 ] x=10.800003 ] x=10.900003 ] x=11.000004 ] x=11.100004 ] x=11.200005 ] ... As Carl C. Chesbrough tell you, this is a problem of precision. On Motorola floating point implementation, 23 bits are used to code the floating point mantissa of a number, and 2^-23 makes 0.000000119. This number is the precision of C type 'float'. This means that you cannot make sure that you can use your increment of 0.1 as an exact value, but with a potential error of +/- 1E-06. Have you seen the result of your 'frog' program when you add 0.2 ? I give you : x=10.200000 x=10.400000 x=10.599999 :-) x=10.799999 x=10.999999 x=11.199999 x=11.399999 x=11.599998 ;-) x=11.799998 x=11.999998 :-( With 'double' floating point type, the precision is 2^-47, ie 7.1E-15. If you use a double typed variable, with the same printf format, the result will be better, but if you let the printf format be "%20.15f", you'll see : x= 10.100000000000000 x= 10.199999999999999 x= 10.299999999999999 x= 10.399999999999999 x= 10.499999999999998 x= 10.599999999999998 x= 10.699999999999998 x= 10.799999999999997 x= 10.899999999999997 x= 10.999999999999996 So, you can use float or double type in your programs : just choice what precision you need during calculation. Best regards. +------------------------------------------------------------------------------+ | /\ | | ///\ /\ __, Philippe Leverone, ingenieur logiciel | | /\ ///. \///\ \O ALPLOG SA, groupe ALTRAN | | // \ /. /////\_| L, 4 rue du Moulin, BP 23 - Wolfisheim | | ///. \/ ///////\ / 67032 STRASBOURG cedex | | /. / //////. \- chez EGT SA FRANCE | | / ///. \ [*] | | / \ e-mail : pleverone@egt.fr ,_"_ | | [_(O)] | +------------------------------------------------------------------------------+ From froeber@bbn.com Tue Oct 8 17:27:30 1996 From: Fred Roeber Date: Tue Oct 8 17:27:32 PDT 1996 Subject: Re: Ada/VxWorks Bindings On Fri, Oct 4, 1996, Robin Hearn asked: > We're trying to use VisiComms X-Windows package for VxWorks, the problem is the > preferred programming language is Ada (Green Hills Ada on the PowerPC). Has > anyone created bindings for Ada to either X/Motif or VxWorks? > Any hints, tips or problems? To which Joseph Coballe replied: > I suggest looking into VADSworks from Rational Software (www.rational.com). > It comes with a library(source included) of ADA bindings for most of the > VxWorks libraries. It has been a few years since I did Ada work with VxWorks but at that time we made extensive use of VadsWorks from Verdix (now Rational). Their compiler and toolset did come with a pretty extensive set of bindings to VxWorks functions. Unfortunately, unless something has changed, I wouldn't think you could get their bindings unless you bought their toolset and VADSWorks interface. We used them because they always had real good compiler technology. Seems you have chosen GreenHills though (I would guess due to front end cost issues?). As others have mentioned in other posts, creating Ada binding to VxWorks functions isn't too hard although getting things like tasking and exception handling and debug support integrated in is a bit trickier. As to X Windows, those bindings are another story. Creating Ada X windows bindings is not trivial. The 1992 Tri-Ada conference proceedings, for instance, have several articles on developing and comparing different Ada/X bindings. We have been using the Visicom X windows interface for VxWorks (with C) and have found that it allows very good portability of code from UNIX X windows environments to VxWorks (as long as memory size issues are taken into account). Thus, I would think you could probably use any of the commercial Ada X windows bindings. One of the better ones a few years ago was from Systems Engineering Research Corporation (SERC). I haven't worked with the GreenHills people but I would guess that if you were using the VADSWorks product you could certainly get advice on this issue from Rational. Good luck. Fred | Fred J Roeber, BBN Systems & Technologies | | 50 Enterprise Place Middletown, RI 02842-5202 | | froeber@bbn.com 401-849-2543 (X48) | | See http://www.bbn.com/tv for TraceMaker info | From daemon@csg.lbl.gov Wed Oct 9 04:00:26 1996 From: daemon@csg.lbl.gov Date: Wed Oct 9 04:00:29 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Wed Oct 9 04:00:23 PDT 1996 Subject: 68360 port C interrupts Subject: How To Disable CTRL-X (^X)? ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: 68360 port C interrupts Date: Tue, 08 Oct 1996 19:35:35 +0100 From: Owen Friel Organization: Nortel Message-ID: <325A9EF7.41C67EA6@nt.com> Has anyone ever tried using the port C interrupts on the 68360 under VxWorks? I'm at a dead end and would appreciate any help or suggestions. Cheers, Owen Friel Nortel Ph +353 91 733360 Fax +353 91 756050 Mail owen.friel@nt.com --------------------------- Newsgroups: comp.os.vxworks Subject: How To Disable CTRL-X (^X)? Date: Sat, 28 Sep 1996 19:22:09 GMT From: wchung@netcom.com (Wing-kuen Chung) Organization: NETCOM On-line Communication Services (408 261-4700 guest) Keywords: shell, tty, ptty, ctrl-x, ^X, ioctl, fd, OPT_MON_TRAP Message-ID: Sender: wchung@netcom9.netcom.com Hello, I wrote a simple route to enable and disable CTRL-X. The routine simply make use of ioctl. But one problem is that I don't know where to invoke the routine so that it's not too early or too late in its invocation such that by the time the shell comes up. CTRL-X is complete disabled. Here are places that I put it in but they don't work at all. 1. end of sysHwInit() 2. end of sysHwInit2() 3. shell script file invoked from bootparam. 4. end of usrRoot() (after shell startup) My function only works when I manually invoke my function, ctrlXEnable(1) in both tty shell and ptty shell (like rlogin). The purpse is to make sure when shell comes up, a naive user cannot accidentally type in ctrl-X to reboot the machine but the setup still alow "reboot" command to be issued by application codes. Hopefully, there is a lead on this. Greatly appreciated in advanced. Wing. - -- --------------------------- End of New-News digest ********************** From froeber@bbn.com Wed Oct 9 18:06:58 1996 From: Fred Roeber Date: Wed Oct 9 18:07:10 PDT 1996 Subject: Re: disabling reboot On Wed, 9 Oct 1996, Wing-kuen Chung wrote: > I wrote a simple route to enable and disable CTRL-X. The > routine simply make use of ioctl. But one problem is that > I don't know where to invoke the routine so that it's not > too early or too late in its invocation such that by the > time the shell comes up CTRL-X is complete disabled. Rather than use the OPT_MON_TRAP ioctl which has the problem of needing to be specifically configured/set on each terminal device you use, you could simply call the tyMonitorTrapSet function to change the character that the user has to type to cause a reboot. If you set it to some character that the user can't type in easily (e.g. 0x80) then you will be effectively disabling the feature. The two nice things about this approach are that it affects all terminal (and therefor also pseudo terminal) devices and that you can call this function at any point in time since it changes a global variable used in the driver. To restore the reboot function, you just set the character back to ^X (i.e. 0x18). Fred | Fred J Roeber, BBN Systems & Technologies | | 50 Enterprise Place Middletown, RI 02842-5202 | | froeber@bbn.com 401-849-2543 (X48) | | See http://www.bbn.com/tv for TraceMaker info | From 100656.1507@compuserve.com Thu Oct 10 02:29:36 1996 From: John Shaw <100656.1507@compuserve.com> Date: Thu Oct 10 02:29:38 PDT 1996 Subject: VxWorks booting failures Hello, We are running Sun Solaris 2.4 on the VxWorks host machine. We use NFS on disk partitions between this Sun and a Solaris 2.3 machine. We run VxWorks 5.2 on several 68K and PPC cards. Our Sun systems are SPARCstation LX models. Most of the 68K cards are in a VME crate and boot via the back- plane. If both Sun systems are left on over the weekend and remain idle during this time, we have significant problems when booting the cards on the following Monday. The boot sequence usually times out. The only way to resolve the problem is to reboot the Sun systems! This just adds to the usual blues felt on that particular day by most of the Western World!! We have also received an unusual message after VxWorks boots but before, or during, the running of the boot parameter script. The message is Fork (to start shell) failed on server. Please try again later. We used to boot VxWorks 5.1.1 on a Solaris 2.3 machine without any of these problems being encountered, though we were doing without NFS then. We do now have quite busy network segments in our computer room, what with X terminals, NetWare servers, and the VME crate of 68K cards competing on the Ethernet.... Has anyone else had similar problems? John Shaw, Shade (Computer Services) Limited, Calne, Wiltshire, England, UK. Email: 100656.1507@compuserve.com From daemon@csg.lbl.gov Thu Oct 10 04:00:32 1996 From: daemon@csg.lbl.gov Date: Thu Oct 10 04:00:35 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Thu Oct 10 04:00:28 PDT 1996 Subject: Re: Using Select() for a delay / System Clock Rate Subject: Re: Tornado and Source Control Subject: re: Boot failures Subject: remote diagnotics via modem Subject: Re: Details on 360 port C Subject: Re: TCP Acknowledgement Subject: TCP Acknowledgement Subject: Details on 360 port C Subject: Booting from SCSI messes up socket connections Subject: Re: TCP over a serial link Subject: Token Ring Driver for the Power PC Subject: Re: Using msqQReceive() in interrupt handlers ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Using Select() for a delay / System Clock Rate Date: Sat, 05 Oct 1996 11:10:21 -0700 From: "H.J. Bae" Organization: Peaceful Star Message-ID: <3256A48D.5047@peacefulstar.oakland.ca.us> References: <52e0oe$d49@inetserv2.fs.lmco.com> Reply-To: hjb@peacefulstar.oakland.ca.us The timeout for select() is system clock rate dependent. Making the system clock tick rate higher would me more precise timing, although it would incurr more overhead due to higher frequency of clock interrupts. - -- http://peacefulstar.oakland.ca.us mailto:hjb@peacefulstar.oakland.ca.us --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Tornado and Source Control Date: Sat, 05 Oct 1996 11:17:02 -0700 From: "H.J. Bae" Organization: Peaceful Star Message-ID: <3256A61E.4717@peacefulstar.oakland.ca.us> References: <9610041610.AA08358@lbl.gov> Reply-To: hjb@peacefulstar.oakland.ca.us I have used CVS successfully on both Solaris and WinNT for a fairly large size project using Tornado. The simplest (perhaps not the best) way to allow access to the CVS repository over the LAN to workstation clients is to put the repository on a WinNT server and add NFS server to the WinNT server. This way, WinNT clients can access the repository over SMB and Solaris clients can access the repository over NFS. CVS has its own remote access protocol based on RSH, but the free versions of CVS available for WinNT do not seem to work well over RSH (it could be just pilot error). - -- http://peacefulstar.oakland.ca.us mailto:hjb@peacefulstar.oakland.ca.us --------------------------- Newsgroups: comp.os.vxworks Subject: re: Boot failures Date: 27 Sep 1996 20:15:45 GMT From: Barbara Spinney Organization: Raytheon Electronic Systems Message-ID: <52hclh$iuu@news.res.ray.com> References: <2.2.32.19960926180517.00756754@mailhost.mclean.sparta.com> I thought I had the memory map straight but after reading some of the responses to this message I am confused again. This is what I thought the memory mapping was set up to be - If anyone has more information I certainly would appreciate hearing about it. ---------------- sysMemTop - top of the RAM | | from LOCAL_MEM_SIZE | bss segment | | for data | | - - - - | | VxWorks | | image loaded | | by bootInit | |--------------| RAM_HIGH_ADRS | Stack | | | | - - - - | | extra space | | | |--------------| RAM_ADRS + RAM_LOW_ADRS | exc vectors, | | exc msg, | | bootline | |______________| RAM_ADRS I am going to port vxWorks and a tiny little ethernet message passing application onto an old product with limited memory. Right no vxWorks image with a dummy ethernet message passing program in the place of usrDemo. I have built this and tried it on an MVME167 su my final application with the limited memory space I will need to know exactly where everything goes. I'd like to cut out a small s save a bit from the top (take say 64K off the LOCAL_MEM_SIZE) or should I take space between the Stack and the exception vectors area. --------------------------- Newsgroups: comp.os.vxworks Subject: remote diagnotics via modem Date: Wed, 09 Oct 1996 15:47:16 GMT From: isr@ix.netcom.com (Robert C. Smith) Organization: Netcom Message-ID: <53ght4$n4a@dfw-ixnews4.ix.netcom.com> We're interested in performing diagnostic routines / maintenance remotely via a modem. The target is an MVME-162 running VxWorks 5.3. Does anyone have experience doing this kind of thing? Do we need to track down a modem that can establish a link to a local modem prior to booting the target (which, I guess, should be configured to boot using CSLIP)? A commercial, off-the-shelf solution would be my preference. Thanks, Bob Smith Etec Polyscan 520-806-3482 robert_smith@fc.etec.com OR isr@ix.netcom.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Details on 360 port C Date: Wed, 09 Oct 1996 13:44:51 -0700 From: "H.J. Bae" Organization: Peaceful Star Message-ID: <325C0EC3.45EC@peacefulstar.oakland.ca.us> References: <325B67C4.41C67EA6@nt.com> Reply-To: hjb@peacefulstar.oakland.ca.us The CIPR bit (including PC0) will not be set if the error vector is issued by the CPM. Quoting the 360 manual," THe error vector is issued by the CPM if an interrupt was requested by the CPM but was masked by the user prior to being serviced by CPU32+ core and if no other pending interrupts for the CPM are present." I've only used Port C for general purpose I/O (no interrupt), so I'm not sure if this helps. - -- http://peacefulstar.oakland.ca.us mailto:hjb@peacefulstar.oakland.ca.us --------------------------- Newsgroups: comp.os.vxworks Subject: Re: TCP Acknowledgement Date: Wed, 09 Oct 1996 14:12:48 -0700 From: "H.J. Bae" Organization: Peaceful Star Message-ID: <325C1550.4BA9@peacefulstar.oakland.ca.us> References: <53ed53$52d@newsserver.rrzn.uni-hannover.de> Reply-To: hjb@peacefulstar.oakland.ca.us There is no exact way to control the delay for ACK generation in TCP code. Nor would you really want to have such mechanism. TCP code has a lot of heuristics to figure out when to delay an ACK to piggyback next time or to avoid congestion on already choking TCP stream. However, there are a couple of indirect ways to affect some of this behavior to make ACK generation go "faster". First, you can use TCP_NODELAY option to turn of Nagle's algorithm in TCP code. Second, you can inject out-of-band data into your TCP stream to force an output. However, you should not have to resort to using these mechanisms. There really is no default time delay for ACK in the TCP code. - -- http://peacefulstar.oakland.ca.us mailto:hjb@peacefulstar.oakland.ca.us --------------------------- Newsgroups: comp.os.vxworks Subject: TCP Acknowledgement Date: 8 Oct 1996 20:21:55 GMT From: degner@juno.amp.uni-hannover.de (Ralf Degner) Organization: RRZN - Newsserver Message-ID: <53ed53$52d@newsserver.rrzn.uni-hannover.de> Hi ! I want to decrease the delay time for TCP Acknowledgement. Is this possible under VxWorks and if yes, how ? What is the default time for the TCP acknowledgement ? I measured somethink about 200 ms. It is a bit long for my application. I would be very happy abou every help. Ralf Degner - -- EMail: degner@pallas.amp.uni-hannover.de WWW : http://sun1.rrzn.uni-hannover.de/~nhbidegn --------------------------- Newsgroups: comp.os.vxworks Subject: Details on 360 port C Date: Wed, 09 Oct 1996 09:52:20 +0100 From: Owen Friel Organization: Nortel Message-ID: <325B67C4.41C67EA6@nt.com> I'm running VxWorks on a 68360 and am having difficulty getting port C bit 0 interrupts to work. I've done the following: Set PCPAR to 0 Set PCDIR to 0 Set PCSO to 0 Set PCINT to 0 Set CIMR PC0 to 1 I'm using SCC1 SCC2 and SCC3 without any problems but can't get the PC0 interrupt to occur. I can check the values in all 5 registers at run time and they are correct. I can toggle the input level on the PC0 pin and read the PCDAT register to verify that the level is changing, but the interrupt never occurs. As stated in 7.15.3 of the 360 manual, the PC0 bit in the CIPR should be set if an interrupt is pending even if the interrupt is masked. This bit is not set when I toggle the input level on the PC0 pin. Has anyone any suggestions as to where I am going wrong? Cheers, Owen Friel Nortel (Northern Telecom) Galway, Ireland Phone +353 91 733360 Fax +353 91 756050 Mail owen.friel@nt.com --------------------------- Newsgroups: comp.os.vxworks Subject: Booting from SCSI messes up socket connections Date: 26 Sep 1996 13:42:55 GMT From: fred.getz@lmco.com Organization: Lockheed Martin Federal Systems Message-ID: <52e18v$c04@inetserv2.fs.lmco.com> Reply-To: fred.getz@lmco.com When our 5CE VxWorks 5.2 systems are set to boot from their scsi disks, we see strange network problems. Yes the "other" boot parameter is set to 'ln' to cause the network to be initialized at boot time. After the systems boot, we can use ftp and telnet, but attempting to establish socket connections with other workstations is very flaky. There is no problem with this when we boot the 5CE's over the network from a host system. The only difference is the following boot parameters: boot-device, file, other All other boot parameters remain the same. Does anyone have any idea what is happening here? How can this be resolved so that socket connections work correctly when booting from the scsi disk? Thanks, Fred Getz - Lockheed Martin Federal Systems --------------------------- Newsgroups: comp.os.vxworks Subject: Re: TCP over a serial link Date: 2 Oct 1996 09:07:46 GMT From: lazarus@sun-valley.Stanford.EDU (Howard Wang) Organization: Aerospace Robotics Lab, Stanford University Message-ID: <52tbd2$1up@nntp.Stanford.EDU> Well, since this is in the VxWorks newsgroup, I'm assuming you're using the VxWorks OS. If so, then you can run SLIP (part of the normal VxWorks distribution) or PPP (an optional package) which gives IP over serial links. That means your code that does TCP/UDP doesn't have to change. You only need to change routes and gateways and such at the operating system level when switching between TCP and UDP. Check your manual for slip (vxman if_sl, or slipInit, etc.). Works pretty well. Howard Wang Stanford ARL --------------------------- Newsgroups: comp.os.vxworks Subject: Token Ring Driver for the Power PC Date: Wed, 09 Oct 1996 12:56:42 -0400 From: David Pyle Organization: Teklogix Inc. Message-ID: <325BD94A.14F2@teklogix.com> Does anyone know where I can get a PCI based Token Ring driver for the Power PC? I do not care what chip set. I would like to source the Token Ring card based on driver availability - if possible. Dave --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Using msqQReceive() in interrupt handlers Date: 3 Oct 1996 12:28:09 GMT From: Alex Goodyear Organization: JET Joint Undertaking Message-ID: <530bgp$se3@postman.jet.uk> References: <52oct7$ijp@nameserver.ge.cnr.it> Hi, ISR's produce data in response to an external (to the kernel) event, they do not consume data. Also buy their very nature ISR's should be FAST, this is why potentially time consuming and pre-emptive actions are added to the workQ and deferred until after the ISR has finished. A msgQReceive requires that a message be extracted from the internal msgQ structures and all the links updated but defer any task manipulation (re-schedules) until later. All this processing combined with the fact that an ISR should never need to receive internally generated messages means that msgQReceive is not supported by VxWorks (within an ISR). If you are using the msgQLib to pass pointers to empty data buffers that the ISR can use, then you should be using rngLib to pass the data to the ISR & msgQLib to pass full buffers to the task level code. Hope this helps, Alex. =============================================================================== The above article is the personal view of the poster and should not be considered as an official comment from the JET Joint Undertaking =============================================================================== --------------------------- End of New-News digest ********************** From luettgen@nis.lanl.gov Thu Oct 10 05:50:55 1996 From: Allen Luettgen Date: Thu Oct 10 05:50:57 PDT 1996 Subject: unsubscribe unsubscribe From SJohnson@systran.com Thu Oct 10 12:49:58 1996 From: Steve Johnson Date: Thu Oct 10 12:50:01 PDT 1996 Subject: PowerPC 603 64MB We have been running the 5.2 kernel on a MVME-1603-004 PowerPC board for over a year with the kernel configured as 32MB LOCAL_MEM_SIZE=0x2000000 in the config.h file. Our PPC board actually has 64MB of on-board RAM. When I change LOCAL_MEM_SIZE to 0x400000 and rebuild, it boots up fine but when I load any of my applications I get a bunch of warning messages scrolling to the display in the shell. The message reads as follows: warning: value does not fit in 24 bits but the load is completed. When I try to run any of these applications it will crash the system. We are using the Diab C compiler 3.6f. Has anyone ever seen this problem before ? I found some switches in my inherited makefile for Diab that were suspicious. They are: -Xsmall-data=0 -Xsmall-const=0 I cannot find these switches in any of the Diab documentation. Removing these switches from the makefile made no difference in the problem described above. Can anyone tell me what the difference is between the 32 and 64 MB kernel configuration or is it something in the compiler configuration ? From cwsicki@sandia.gov Thu Oct 10 15:30:47 1996 From: cwsicki@sandia.gov (Carl Sicking) Date: Thu Oct 10 15:30:51 PDT 1996 Subject: Re: PowerPC 603 64MB >Submitted-by SJohnson@systran.com Thu Oct 10 12:49:58 1996 >Submitted-by: Steve Johnson > > > >I found some switches in my inherited makefile for Diab that were >suspicious. They are: -Xsmall-data=0 -Xsmall-const=0 > >I cannot find these switches in any of the Diab documentation. Removing >these switches from the makefile made no difference in the problem described >above. > >Can anyone tell me what the difference is between the 32 and 64 MB kernel >configuration or is it something in the compiler configuration ? > The Diab Data Tornado Application Note specifies that these switches must be used. If you don't have the App note, I recommend giving them a call. They faxed me a copy last week. They comments in the App note say that these switches are necessary because the "VxWorks loader cannot handle small data area." Good Luck, Carl From banta@alioth.cebaf.gov Thu Oct 10 15:48:40 1996 From: Paul Banta Date: Thu Oct 10 15:48:42 PDT 1996 Subject: netperf & ttcp Has anyone ported these utilities to VxWorks ? Thanks in advance. -- Paul Banta voice: 757-269-5829 TJNAF, MS 12H fax: 757-269-5800 12000 Jefferson Avenue email: banta@cebaf.gov Newport News, VA 23606 Thomas Jefferson National Accelerator Facility From banta@alioth.cebaf.gov Thu Oct 10 16:17:53 1996 From: Paul Banta Date: Thu Oct 10 16:17:55 PDT 1996 Subject: tcp/ip performance survey I'd like to take an quick survey of experience with tcp/ip in high bandwidth applications. Particularly where VxWorks targets are sending as much data as they can muster through a network interface. What is the average output bandwidth? What is the peak output bandwidth? Does the system perform well over a long period of time? Are there any special tcp parameter modifications (e.g. adjusting the send and receive sliding window sizes)? I'll collect the results and distribute them by request. My curiosity about this stems from my current problem: I have a Radstone PPC604 (100MHz PowerPC) with an ATM adapter on it. The driver is home grown (with some help from various people around the net). The other end of the fiber is plugged into a sun ultra 1. The unix system has a process that eats data from a socket as fast as it can. The vxworks target crams data through a socket as fast as it can. The system runs steady at around 4.9 MBytes/s. It will occasionally peak at just over 5MBytes/s (the driver has yet to be tweaked so I'm hoping that this number will go up. I still have CPU left at this rate). But, it doesn't run steady for long. After 30-60 seconds it begins to slowly degrade and the data rate ultimately drops dramatically. I'm guessing that I'm stuck in some funky tcp state which, of course, could be happening due to a driver bug, but I'd like to hear about other experiences. Thanks, -- Paul Banta voice: 757-269-5829 TJNAF, MS 12H fax: 757-269-5800 12000 Jefferson Avenue email: banta@cebaf.gov Newport News, VA 23606 Thomas Jefferson National Accelerator Facility From froeber@bbn.com Thu Oct 10 18:24:40 1996 From: Fred Roeber Date: Thu Oct 10 18:24:42 PDT 1996 Subject: Re: tcp/ip performance survey Paul Banta wrote: > The system runs steady at around 4.9 MBytes/s. It will occasionally peak > at just over 5MBytes/s (the driver has yet to be tweaked so I'm hoping that > this number will go up. I still have CPU left at this rate). But, it > doesn't run steady for long. After 30-60 seconds it begins to slowly > degrade and the data rate ultimately drops dramatically. I'm guessing that > I'm stuck in some funky tcp state which, of course, could be happening due > to a driver bug, but I'd like to hear about other experiences. Sounds like an interesting problem. Due to the symptoms, I would tend to think you might be running into memory fragmentation issues. Since VxWorks uses a fairly simple "first fit" linked list memory allocator, the allocator's performance can degrade over time if you are creating lots of blocks and only releasing some of them. Try to use memShow to see if you have a memory leak. As far as fragmentation, you can use "memShow 1" to see the free list but that might get old in a hurry. I think there are some vendors out there with tools to show the heap state better if you think that is the problem (I would check the third party product list on WRS home page). -- | Fred J Roeber, BBN Systems & Technologies | | 50 Enterprise Place Middletown, RI 02842-5202 | | froeber@bbn.com 401-849-2543 (X48) | | See http://www.bbn.com/tv for TraceMaker info | From cwcssk@queen.cwc.nus.sg Thu Oct 10 23:38:37 1996 From: Suresh Santhana Krishnan Date: Thu Oct 10 23:38:40 PDT 1996 Subject: Any one using VxWorks with 68302 Hi, I am interested to know anyone working on VxWorks with 68302 and 68PM302 chipsets. I would like also to know the size restriction. I am trying to use 68302 for Wireless Protocol implementation. What I want to know is the code + data size (uncompressed) that can be used. And the important layers of code that need to be there for the run-time system. And the method of loading ROM + Flash memory ? Thanks in advance, Suresh Santhana Krishnan Centre For Wireless Communications Singapore From robin.hearn@gecm.com Fri Oct 11 01:59:18 1996 From: Robin Hearn +44 1935 442109 Date: Fri Oct 11 01:59:20 PDT 1996 Subject: Re: PowerPC 603 64MB >Submitted-by SJohnson@systran.com Thu Oct 10 12:49:58 1996 >Submitted-by: Steve Johnson > > >We have been running the 5.2 kernel on a MVME-1603-004 PowerPC board for >over a year with the kernel configured as 32MB LOCAL_MEM_SIZE=0x2000000 in >the config.h file. Our PPC board actually has 64MB of on-board RAM. When I >change LOCAL_MEM_SIZE to 0x400000 and rebuild, it boots up fine but when I >load any of my applications I get a bunch of warning messages scrolling to >the display in the shell. The message reads as follows: > >warning: value does not fit in 24 bits > >but the load is completed. When I try to run any of these applications it >will crash the system. We are using the Diab C compiler 3.6f. > >Has anyone ever seen this problem before ? > >I found some switches in my inherited makefile for Diab that were >suspicious. They are: -Xsmall-data=0 -Xsmall-const=0 > >I cannot find these switches in any of the Diab documentation. Removing >these switches from the makefile made no difference in the problem described >above. > >Can anyone tell me what the difference is between the 32 and 64 MB kernel >configuration or is it something in the compiler configuration ? We had a similar problem with out Radstone PPC604 board, when I contacted our supplier they came up with the following problem report. Summary #28843 AUTHOR: jerald ------------------------------------------------------- DATE: Feb 23 1996 10:54AM Product: 5.2-PowerPC Beta-1 Host: Sun4 Board: mv1604 Board: TITLE: Value does not fit in 24 bits. AUTHOR: J. Pendleton SCOPE DISCOVERED ON HOST: N/A ------------------- ------------------- ARCHITECTURE: PowerPC ------------------- ------------------- BSP: N/A ------------------- ------------------- ------------------- ------------------- DESCRIPTION OF PROBLEM: When attempting to load an object module with the shell loader the message "Warning: Value does not fit in 24 bits appears. RESOLUTION: This is due to an architectural limitations of the powerPC. Certain instuctions in the powerPC (eg. bl) only allow a 24 bit offset. When vxWorks mallocs memory for a object being loaded, it mallocs from the top of memory. If the addressed malloced lies outside the 24 bit offset, relocation cannot occur and the load fails. Suggested workaround: 1. build vxWorks with LOCAL_MEM_SIZE set to 32m. 2. Load in the objects using the shell loader. 3. MemAddToPool the remaining memory. SPR / PATCH: 6040 / No patch planned KEYWORDS: SIGN-OFF: Hope this helps Robin ------------------------------------------------------------- Robin Hearn Thomson Marconi Sonar Limited Wilkinthroop House, e-mail: robin.hearn@gecm.com Templecombe, tel: +44 (0) 1935 442109 Somerset, BA8 0DH, England fax: +44 (0) 1935 442185 ------------------------------------------------------------- From daemon@csg.lbl.gov Fri Oct 11 04:01:31 1996 From: daemon@csg.lbl.gov Date: Fri Oct 11 04:01:34 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Fri Oct 11 04:01:28 PDT 1996 Subject: Re: backplane TCP/IP Subject: MVME1300 + VxWorks Subject: MVME1300 + VxWorks Subject: Re: remote diagnotics via modem Subject: Solaris/PPC compiler for VxWorks 5.2 - Status? Subject: VxWorks network routing capabilities Subject: Re: Solaris/PPC compiler for VxWorks 5.2 - Sta Subject: Re: VxWorks network routing capabilities Subject: US-FL-ST. PETERSBURG ADA, VxWorks, 68040, UNIX Subject: switch to output source code to assem mapping Subject: Re: Flash Disk Driver for VxWorks ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: backplane TCP/IP Date: 4 Oct 1996 11:44:40 -0600 From: bgeer@xmission.xmission.com (bgeer) Organization: XMission Internet (801 539 0900) Message-ID: <533ie8$ctr@xmission.xmission.com> References: <2.2.32.19961003142604.006b3078@mailhost.mclean.sparta.com> Mike Anderson writes: >At 03:02 PM 10/2/96 PDT, you wrote: >>Submitted-by rich@immrc1.eng.mcmaster.ca Wed Oct 2 15:02:49 1996 >>Submitted-by: rich@immrc1.eng.mcmaster.ca (Richard Teltz) >> >>I posted a question a while ago regarding maximum bandwidth >>for the VxWorks TCP/IP implementation across the VMEbus. >> >>I recieved no response... Well, I managed to miss the original post...sorry. I did some backplane network testing a couple of years ago between 2 MVME162LX 68040 cards. Based on our requirements, my testing was with large packets (32k bytes). The "sm" device achieved 3-5 megabits, depending on transfer start & end points -- slave/master was faster than slave/slave as would be expected. My "dm" device achieved 15 megabits regardless of master/slave relationship. This device used DMA to transfer packets & I built a "direct memory" manager which allowed every packet to be transferred direct to the recipients memory from the source buffer, rather than routing all packets thru the master's buffer system as "sm" does. I ported my code to a Performance Technologies 68030 board & a Force 68040 board. The downside is having to write a robust DMA driver & experiment with each different VME chip used; some of those chips don't seem to perform as well as I expected. Turns out the vmechip2 on the MVME162LX was the nicest to work with -- using DMA its 68040-25MHz was faster than a Radstone 68040-33MHz using bcopy() (the latter's DMA capability was too limiting to be useful). We noted bcopy() transfer was faster up to about 0x140 to 0x200 bytes in size; for bigger packets DMA was faster. Wish I could post my "dm" stuff to the archive...never got my employer's permission. Cheers, Bob - -- <> Robert Geer & Donna Tomky / * <> <> bgeer@xmission.com | _o * o * o <> <> Salt Lake City, Utah | -\<, * <\ <> U S A | O/ O __ /__, /> <> --------------------------- Newsgroups: comp.os.vxworks Subject: MVME1300 + VxWorks Date: Thu, 10 Oct 1996 12:50:50 +0100 From: Hepworth & Hatcher Organization: UUNet PIPEX server (post doesn't reflect views of UUNet PIPEX) Message-ID: <325CE31A.30B9@dial.pipex.com> Are there any VxWorks users out there using the an MVME1300 series board and VxWorks? If there are then are you using a network adapter and if so which one? I have ordered an MVME1302 board and an Interphase 4524 network card. Unfortunately there I have not found a driver for this board or its chipset the TI ThunderLan ! Thanks very much P.Hatcher hephat@dial.pipex.com --------------------------- Newsgroups: comp.os.vxworks Subject: MVME1300 + VxWorks Date: Thu, 10 Oct 1996 12:51:20 +0100 From: Hepworth & Hatcher Organization: UUNet PIPEX server (post doesn't reflect views of UUNet PIPEX) Message-ID: <325CE338.40A5@dial.pipex.com> Are there any VxWorks users out there using the an MVME1300 series board and VxWorks? If there are then are you using a network adapter and if so which one? I have ordered an MVME1302 board and an Interphase 4524 network card. Unfortunately there I have not found a driver for this network board or its chipset the TI ThunderLan ! Thanks very much P.Hatcher hephat@dial.pipex.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: remote diagnotics via modem Date: Thu, 10 Oct 1996 18:18:38 GMT From: 100567.1652@compuserve.com (Rolf Grzibek) Organization: Philips BTS Message-ID: <53jeit$7q4@news.Dortmund.Germany.EU.net> References: <53ght4$n4a@dfw-ixnews4.ix.netcom.com> Hi, Sorry, no answer, but the same problem... In my current project (custom hardware with i960 running VxWorks 5.3) we also want to use a modem for remote diagnostics. We like to have the following possibilities: - - target shell redirected from the standard console - - SLIP/CSLIP (or PPP) connection so that the Tornado target server can use WinNT/95 Dial-up Networking - - serial connection to the wdb-agent for system mode debugging. (nice to have via modem, but not necessary) Can anybody give some tips, how this can be achieved ? Thanks in advance. Rolf - -- Rolf Grzibek 100567.1652@compuserve.com --------------------------- Newsgroups: comp.os.vxworks Subject: Solaris/PPC compiler for VxWorks 5.2 - Status? Date: 10 Oct 1996 14:53:55 -0400 From: Henning Pangels Organization: National Robotics Engineering Consortium Message-ID: Sender: hmp@falcon Hello all, about a year ago, we were told that there really wasn't a usable compiler for the PowerPC available for Solaris host platforms -- has this improved? I'd love to hear from anyone actively developing VxWorks code under these conditions, as we're considering making that move ourselves. Thanks! - -Henning *---------------------------------------------------------------------* | Henning M. Pangels Assistant Director for Software Engineering | | hmp+@cmu.edu National Robotics Engineering Consortium | | | | http://rec.ri.cmu.edu/~hmp #10 40th Street | | Pittsburgh, PA 15201 | | Robotics Institute Phone: (412)681-6935 | | Carnegie Mellon University FAX: (412)681-6961 | *---------------------------------------------------------------------* --------------------------- Newsgroups: comp.os.vxworks Subject: VxWorks network routing capabilities Date: 10 Oct 1996 18:58:09 GMT From: "Marc Tillinghast" Organization: Storm Integration Message-ID: <01bbb6dd$9fbe5820$90411ecc@daysbetween> I'm adding two IP-Ethernet modules to a MV162 running VxWorks (currently version 5.1.1). I'm writing SW to move specific data between the three network interface via TCP sockets. I've got a customer who lives deep in the classified world who has security concerns about this configuration. I'm looking for proof that VxWorks doesn't have a native routing capability such as that contained in UNIX. The concern is that this configuration could route all packets between the Ethernet interfaces. Please tell me this isn't possible, and better yet point me to a document that states this - ideally one generated by Wind River. Thanks in advance. Marc Tillinghast (tilly@ix.netcom.com) --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Solaris/PPC compiler for VxWorks 5.2 - Sta Date: 10 Oct 1996 23:56:24 GMT From: scott@tcville.es.hac.com Organization: Hughes Aircraft Message-ID: <53k2f8$600@hacgate2.hac.com> References: Reply-To: scott@tcville.es.hac.com In article fsf@falcon.rec.ri.cmu.edu, Henning Pangels () writes: >about a year ago, we were told that there really wasn't a usable >compiler for the PowerPC available for Solaris host platforms -- >has this improved? >I'd love to hear from anyone actively developing VxWorks code under >these conditions, as we're considering making that move ourselves. GreenHills has a cross-compiler (Multi) for a Solaris host and PPC target. The compiler can be used either interactively (ie. GUI) or in batch (ie. command-line). They can be reached at (813) 781-4909 Scott - --- Hughes Aircraft Co. voice: (310) 616-1059 Image and Signal Processing Lab addr: scott@tcville.es.hac.com PO Box 902, EO/E01/A172, El Segundo, Ca. 90245 --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks network routing capabilities Date: 10 Oct 1996 17:22:38 -0700 From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star, Oakland, CA Message-ID: <53k40e$98q@idiom.com> References: <01bbb6dd$9fbe5820$90411ecc@daysbetween> Marc Tillinghast (tilly@ix.netcom.com) wrote: : I'm adding two IP-Ethernet modules to a MV162 running VxWorks (currently : version 5.1.1). : I'm writing SW to move specific data between the three network interface : via TCP sockets. : I've got a customer who lives deep in the classified world who has security : concerns about : this configuration. I'm looking for proof that VxWorks doesn't have a : native routing capability : such as that contained in UNIX. The concern is that this configuration : could route all packets : between the Ethernet interfaces. : Please tell me this isn't possible, and better yet point me to a document : that states this - : ideally one generated by Wind River. VxWorks networking code is based on 4.3-tahoe BSD code. By default, the network code turns on the global variable ipforwarding. If you reset this value (ipforwarding = 0) then IP routing code in VxWorks kernel will not forward packets. - -- http://peacefulstar.oakland.ca.us mailto:hjb@peacefulstar.oakland.ca.us --------------------------- Newsgroups: comp.os.vxworks Subject: US-FL-ST. PETERSBURG ADA, VxWorks, 68040, UNIX Date: Fri, 11 Oct 1996 01:16:25 GMT From: warnold@emi.net (W.J. Arnold) Organization: Arnold-Hanafin Corporation Keywords: Software Engineer,UNIX,ADA,68000,MVME 165,VXWorks,68040 Message-ID: <53k79r$l2@cambridge.emi.net> Reply-To: warnold@emi.net SOFTWARE ENGINEERS; 8 Positions JOB SUMMARY: Experience with ADA development in a UNIX environment required. Motorola 68040 processor on MVME 165 monoboard computer. VXWorks experience is preferred. C and C++ experience a plus REQUIRED EXPERIENCE: 3-8 years of software development experience EMPLOYMENT ARRANGEMENT: CONTRACT START DATE: IMMEDIATE LOCATION: St. Petersburg, Florida DURATION:6 months+ OTHER REQUIREMENTS: Active Secret Clearance Preferred but will renew clearance if it expired within the last 6 months. Arnold-Hanafin Corporation 9070 Kimberly Boulevard Office #27 Suite 106 Boca Raton, Florida 33434-2861 Voice: 407-488-3946/800-891-1086(toll free) Fax: 407-488-0568 Email: warnold@emi.net --------------------------- Newsgroups: comp.os.vxworks Subject: switch to output source code to assem mapping Date: 7 Oct 1996 20:15:22 GMT From: barbaras@xcalibur.cse.tek.com (Barbara Ann Siefken) Organization: Tektronix, Inc., Beaverton, OR. Message-ID: <53bocq$rh9@tekadm1.cse.tek.com> IDT R4650 base target VxWorks OS GNU Compiler My customer wants to use LA-Browser. The disassembly works great, but we haven't been able to generate the appropriate information on his compiled objects for LA-Browser to work. He links on the target and generates an a .o. file. What is missing is a companion file that correlates the source code line numbers to the assembly instructionton addresses (in this case relative addresses since the code is linked on the target). The question, are there switches for the GNU compiler to generate such a file, or does anyone know of a utility to extract the information from the .o file? Please respond by e-mail if you have help. Barbara Siefken --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Flash Disk Driver for VxWorks Date: Mon, 7 Oct 1996 10:12:38 GMT From: Leonid Rosenboim Organization: RST Software Industries Ltd. Message-ID: <3258D796.41C67EA6@rst.co.il> References: <53214o$fdv@sf18.dseg.ti.com> Sender: news@actcom.co.il (News) Ron Anderson wrote: > > Does anyone know of a source for a Flash memory driver for > VxWorks that will make Flash look like a disk drive? > In fact there is a number of solutions. If your hardware emplys PCMCIA, IDE or even SCSI interfaces, you can buy a flash-based solid state disk with ATA interface for IDE and PCMCIA hardware, or with a SCSI protocol where appropriate.For these drives usually VxWorks already has drivers, as they mimic a magnetic disk with the same interface. The trouble with these solutions is that they are very expensive, and usually made by relatively obscure companies, which may make you feel unsafe to depend upon. These companies are small because they sell few units, because they're expensive ... etc. Interestingly enough, the FLash memory components (which are used inside some of the above mentioned products) are not as expensive now as they used to be. AMD quotes in a press release that they will see their Am29F016 2 MByte component for just $20. in large quantities. My sources claim that SAMSUNG can beat even that price with their Km29N16000TR/TRS components based on NAND-Flash technology. Putting aside the question of where the price difference goes, only a software solution which is royalty-free and is adapteble to virtually any Flash vendor and model, can enable usage of the most cost-effective components available. As software solutions go, there are basically two approaches: FTL - is a software layer that makes the Flash behave like a 512-bytes sectored magnetic disk, on top of which you run the familiar FAT (i.e. dosFs) file system code. FFS - is a technology pioneered by Microsoft Embedded Products Division couple of year back, and defines a completely new, linked- list variable-record based data structure for a File System specifically designed for Flash. (In our Web page there is a reference to an article describing these two technologies in more details). We have choosen the FFS technology to implement specifically for VxWorks, and the embedded market. Our product - FlashLib, is shipping in full source code for over 6 months now, running on 68k, SPARC, PowerPC, i960 and x86 processors, and supporting Flash memories from Intel and AMD, with some customers already shipping products to customers with FlashLib embedded in its software. In terms of hardware, the basic requirement of FlashLib is that the device is composed of at least 4 individually erasable sectors, all of which are of equal size (symmetrical). In practice we can use 28F020, 28F008SA, 28F016xx (in x8 data bus), Am29F040, Am29F080, Am29F016, and recently we added KM29N16000xx in the form of a newly developed "Flash Driver". Please refer to our Web site (below) for more details on this and other products we offer. - -- - ----------------------------------------------------------------------- Leonid Rosenboim Phone: +972-3-559-8144 R S T Software Industries Ltd. Mobile: +972-50-307-142 P.O.Box 11502, AZUR 58017, Israel Fax: +972-3-559-8244 WWW: http://www.rst.co.il E-Mail: leonid@rst.co.il --------------------------- End of New-News digest ********************** From johill@lanl.gov Fri Oct 11 07:56:55 1996 From: Jeff Hill Date: Fri Oct 11 07:56:57 PDT 1996 Subject: Re: tcp/ip performance survey the vxWorks Users Group Exploder wrote: > > Submitted-by froeber@bbn.com Thu Oct 10 18:24:40 1996 > Submitted-by: Fred Roeber > > Paul Banta wrote: > > > The system runs steady at around 4.9 MBytes/s. It will occasionally peak > > at just over 5MBytes/s (the driver has yet to be tweaked so I'm hoping that > > this number will go up. I still have CPU left at this rate). But, it > > doesn't run steady for long. After 30-60 seconds it begins to slowly > > degrade and the data rate ultimately drops dramatically. I'm guessing that > > I'm stuck in some funky tcp state which, of course, could be happening due > > to a driver bug, but I'd like to hear about other experiences. > > Sounds like an interesting problem. Due to the symptoms, I would tend > to think you might be running into memory fragmentation issues. Since > VxWorks uses a fairly simple "first fit" linked list memory allocator, > the allocator's performance can degrade over time if you are creating > lots of blocks and only releasing some of them. Try to use memShow > to see if you have a memory leak. As far as fragmentation, you can > use "memShow 1" to see the free list but that might get old in a hurry. > I think there are some vendors out there with tools to show the heap > state better if you think that is the problem (I would check the > third party product list on WRS home page). Or perhaps lost packets are causing the TCP/IP congestion avoidance algorithm to kick in. Congestion avoidance is discussed in detail in "TCP/IP illustrated Vol 1" by Richard Stevens. Jeff -- ______________________________________________________________________ Jeffrey O. Hill Internet johill@lanl.gov LANL MS H820 Voice 505 665 1831 Los Alamos, NM 87545 USA FAX 505 665 5107 From pthoren@synervision.com Fri Oct 11 11:05:02 1996 From: Phil Thoren Date: Fri Oct 11 11:05:05 PDT 1996 Subject: Small, Fast C, JAVA Database product If interested please respond to return email! or see http://www.vsonline.com regards, Phil VisionSoft ------------------------------------- Name: Phil Thoren (PH: 408-879-2672, FAX: 408-879-2635 E-mail: pthoren@synervision.com (Phil Thoren) WWW: http://www.synervision.com Postal Address: VisionSoft, 1999 S Bascom Ave, Suite 700, Campbell, CA 95008 From BARONK@hub01.tds-gn.lmco.com Fri Oct 11 11:16:40 1996 From: "Baron, Kenneth @ GRNK (LI-JUPITER DIRSYNC)" Date: Fri Oct 11 11:16:42 PDT 1996 Subject: RE: tcp/ip performance survey >I'd like to take an quick survey of experience with tcp/ip in high bandwidth >applications. Particularly where VxWorks targets are sending as much data >as they can muster through a network interface. > >What is the average output bandwidth? >What is the peak output bandwidth? >Does the system perform well over a long period of time? >Are there any special tcp parameter modifications (e.g. adjusting the send and >receive sliding window sizes)? >The system runs steady at around 4.9 MBytes/s. It will occasionally peak >at just over 5MBytes/s (the driver has yet to be tweaked so I'm hoping that >this number will go up. I still have CPU left at this rate). But, it >doesn't run steady for long. After 30-60 seconds it begins to slowly I have a sonar application that needs to read about 350 Mbytes of data which I presently have on a fast SCSI hard disk on a Solaris 2.5 SPARC-20 workstation. I am using VxWorks 5.2 on a Motorola PowerPC MVME1604-014 100 MHz board. In my VxWorks application I use fread() to read the data as needed (virtually continuously). When I mount the Solaris file system and use NFS I get a transfer rate of 5.1 Mbits of my data per second (not counting the packet overhead). When I use netDrv (RSH) I get a transfer rate of 3.8 Mbits per second. Ethernet is supposed to be able to handle 10 Mbits/sec. I have only three nodes on the ethernet: SPARC-20; MVME1604 and an HP printer. I can do this all day. I am also interested in tcp and udp transfer speeds over all the types of media. If people mail to me I will compile a meddia/protocol/architecture xfer speed table. Ken Baron Lockheed Martin (516) 574-2549 kenneth.baron@lmco.com From pardo@rti.com Fri Oct 11 11:28:32 1996 From: Gerardo Pardo-Castellote Date: Fri Oct 11 11:28:34 PDT 1996 Subject: Re: Solaris/PPC compiler for VxWorks 5.2 - Status? >> about a year ago, we were told that there really wasn't a usable >> compiler for the PowerPC available for Solaris host platforms -- >> has this improved? >> I'd love to hear from anyone actively developing VxWorks code under >> these conditions, as we're considering making that move ourselves. There is one available from WRS now. Tornado (VxWorks 5.3) comes with a version of gcc that cross compiles form solaris2 to ppc. It should also work for VxWorks 5.2 -Gerardo =========================================================================== = = = = Gerardo Pardo-Castellote = email: pardo@rti.com = = Real-Time Innovations, Inc. = Phone: (408) 720-8312 = = 155A Moffett Park Drive, Suite 111 = Fax: (408) 734-5009 = = Sunnyvale, CA 94089 = http://www.rti.com = = = = =========================================================================== From daemon@csg.lbl.gov Sat Oct 12 04:01:51 1996 From: daemon@csg.lbl.gov Date: Sat Oct 12 04:01:54 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sat Oct 12 04:01:48 PDT 1996 Subject: Re: TCP Acknowledgement Subject: Re: Console Tasks in VxWorks not Functioning Subject: Re: Bare-Bone configuration? Subject: VxWorks interrupt handling Subject: Test message Subject: [Q] Vxworks SMP version?? Subject: Re: TCP over a serial link Subject: g++ for Intel 960 with exceptions ??? Subject: Re: Wind River Web Site Subject: C, JAVA Small, Fast Database Subject: Re: Looking for pcmcia ata flashdisk driver Subject: Re: VxWorks interrupt handling Subject: VxWorks on DSPs? Subject: Looking for pcmcia ata flashdisk driver Subject: Re: tcp/ip performance survey Subject: Re: netperf & ttcp ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: TCP Acknowledgement Date: Thu, 10 Oct 1996 19:10:46 +0200 From: Ralf Degner Organization: RRZN - Newsserver Message-ID: References: <53ed53$52d@newsserver.rrzn.uni-hannover.de> <325C1550.4BA9@peacefulstar.oakland.ca.us> Hi ! > There is no exact way to control the delay for ACK generation > in TCP code. Nor would you really want to have such mechanism. I realy do not want to do this, but it would be a nice think for my special application. > However, there are a couple of indirect > ways to affect some of this behavior to make ACK generation > go "faster". First, you can use TCP_NODELAY option to turn of Nagle's > algorithm in TCP code. For example at SunOS TCP_NODELAY is a kernel flag +> recompile all (if you own it). I think at VxWorks it an option you can set with setscokopt. I tried this but saw ne effect. > Second, you can inject out-of-band > data into your TCP stream to force an output. Not very nice and hard to realize with my client software (LabView). > However, you should > not have to resort to using these mechanisms. There really > is no default time delay for ACK in the TCP code. I think, there is. If the TCP code is not able to add the ACK to a normal datapackage, it will send an ACK only packed after a certain time. At SunOS this time is 50 ms. I measured 200 ms with VxWorks. I would be very very happy if I can set this to about 50 ms. A good solution would be to use UDP. But with my client this is a hard job to generate my private ACK. Thanks for your comment. Anythink to say about this ? Ralf Degner EMail: degner@pallas.amp.uni-hannover.de WWW : http://sun1.rrzn.uni-hannover.de/~nhbidegn --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Console Tasks in VxWorks not Functioning Date: 2 Oct 1996 10:22:22 -0700 From: jeffc@ee.ubc.ca (Jeffrey Chow) Organization: University of BC, Dept. of Electrical Engineering Message-ID: <52u8ce$g14@hawk.ee.ubc.ca> References: <9609251446.AA01027@softeng.lamrc.com> In article <9609251446.AA01027@softeng.lamrc.com>, Anthony Le wrote: >Hi Vxperts, > >I have a problem related to console window that needs to be solved. >I use a "tip" program (run in a Xwindow UNIX) to interact with the >console connected over a RS-232 line (9600 baud) of the SUN . >Sometime it badly hung up specially when I display a lot of messages >on the UNIX screen. I've noticed the same problem on my system. I don't know the underlying reason for the problem except that tip must use a different handshaking protocol than a VT100 terminal. I do have a fix that worked on my system. Try the following: In file usrConfigTarg.c, function usrRoot(): find the line that initializes the tty driver (tyCoDevCreate()) and increase the buffer size for both reads and write. (You probably only need to increase the write buffer but I changed both). I increased the buffer from 512 bytes to 4K. What I think happened was that the tip session responds to slow to the tty device driver and the buffers overflow causing the card to hang. Let me know if it works and also if you found another solution. Jeff - -------- Jeffrey Chow, jeffc@ee.ubc.ca, Department of Electrical Engineering, UBC, 2356 Main Mall,Vancouver, BC, V6T 1Z4, --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Bare-Bone configuration? Date: Mon, 30 Sep 1996 08:38:37 -0500 From: Steve Woloschek Organization: GE Medical Systems Message-ID: <324FCD5D.372C@med.ge.com> References: <9608268437.AA843795265@CCGATE.HAC.COM> <52h2a6$l0c@uuneo.neosoft.com> Bruce Wedding wrote: > > schen3@ccgate.hac.com wrote: > > > We try to cut vxWorks OS to bare-bone, I need to know what option need > > to be included or excluded. Can someone tell me where I can find all > > the descriptions for all the facilities (INCLUDE_*) ? > > I have the descriptions on page 374 of the VxWorks Programmer's Guide > for Version 5.3 > > Bruce You will also notice that windConfig utility does not provide facilities to include/exclude all of the options. You will notice in the BSP's that WRS assumes you want a number of features that you actually might not want. For instance in configAll.h ... #define INCLUDE_ANSI_ALL This one adds about 50k of code to your target. In a bare bones system you most likely do not need all the ansi stuff. Steve Woloschek GE Medical Systems --------------------------- Newsgroups: comp.os.vxworks Subject: VxWorks interrupt handling Date: 11 Oct 1996 15:02:09 GMT From: elw@lvld.hp.com (Eddie Williamson) Organization: Hewlett Packard Loveland (Mountain IT) Site Message-ID: <53lnhh$mds@hplvejl.lvld.hp.com> Does someone know the details of how VxWorks handles interrupts/exceptions? I'm trying to get my first BSP up and running. My cpu's are Motorolla 68000/68030/68060. I'm trying to get VxWorks 5.3 up and running on an existing card that had been shipping for a while. We periodically ship software updates on tape and I'm trying to switch to VxWorks for both a new card and the existing ones in the field. The roms currently expect to download an image and jump to its starting point. That means that the roms are already done and I have to boot within the context of our existing system. My current roms start at address 0 and go up. That means the interrupt vector table is fixed and can't be changed. Each vector points to a fixed spot in ram where I initialize a vector jump table to go to the particular exception handler just downloaded into ram. How do I integrate VxWorks' exception handling such as: intConnect()? Can I simply point all of my exception handling into a kernel entry point? Or do I have to do all of my own exception handling still. Bummer since I want to use WindView and I think this method would not allow WindView to "see" the interrupts. ______________________________________________________________________ Hewlett-Packard Eddie L. Williamson Jr. Manufacturing Test Division elw@lvld.hp.com 815 SW 14th Street, MS AU100 (970) 679-3674 Loveland, CO --------------------------- Newsgroups: comp.os.vxworks Subject: Test message Date: Fri, 11 Oct 1996 05:39:42 -0700 From: Chet Kackman Organization: GDE Systems Inc. Message-ID: <325E400E.41C67EA6@GDEsystems.com> Test message. --------------------------- Newsgroups: comp.os.vxworks Subject: [Q] Vxworks SMP version?? Date: 1 Oct 1996 14:25:41 -0400 From: jsun@fore.com (Jianyong Sun) Organization: FORE Systems, Inc. Message-ID: <52rnn5$20b@august.fore.com> Does anyone know that if Wind River System has any plan for a SMP version of VxWorks for PENTIUM PRO? If yes, what is the time frame? Prolog started shipping their dual PENTIUM PRO CompactPCI product recently. Due to SMP capability of PENTIUM PRO, it makes lot of sense for a SMP version of VxWorks. Any comments on this issue will be appreciated. Thanks. James Sun Hardware Design Engineer FORE Systems, Inc. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: TCP over a serial link Date: Tue, 1 Oct 1996 22:03:04 GMT From: dab@netcom.com (Don Brooks) Organization: NETCOM On-line Communication Services (408 261-4700 guest) Message-ID: References: <2.2.32.19961001140258.0070e98c@mailhost.mclean.sparta.com> Sender: dab@netcom12.netcom.com Mike Anderson (mea@mclean.sparta.com) wrote: : At 05:47 AM 10/1/96 PDT, you wrote: : >Submitted-by del@rst.com Tue Oct 1 05:47:49 1996 : >Submitted-by: del@rst.com (Mark Del Giorno) : > : >Heyyyyyyyyyyyyyyyyyyyyyyyyyyyyyy, : > : >We're using a wireless ethernet-based link to a robotic vehicle for : >command and control information from a base station. The protocal : >is TCP or UDP, depending on the type of data being sent. : > : >We also have a backup radio link that provides 2-way serial (RS-232) : >communication. We'd like to be able to switch between the ethernet : >and the serial ports "seamlessly". : > : >Question: Is there soffware available to open a serial port and treat : >it like a TCP connection - i.e. "connects" to the other size, packetizes : >data, auto-retries when there are errors, etc.....? : > : >I'm running an MVME-162 on both sides of the link. : > : Greetings! : Yes, simply use SLIP on the serial port. (#INCLUDE_SLIP in configAll.h) : When you do a slipInit, the serial port will be configured to run the : IP stack (whether or not the radio link is up). You can then do connects, : send, recv, sendmsg, etc. over the serial link as if it were an Ethernet : and the higher level protocol (i.e., TCP) will handle the error detection/ : correction. There is also a public domain port of the PPP (Point-to-Point) serial protocol available via anonymous ftp from ftp.netcom.com as /pub/da/dab/ppp-2.1.2.tar.gz (ftp://ftp.netcom.com/da/dab/ppp-2.1.2.tar.gz via the web). PPP differs from SLIP in being more configurable and having more error checking but its also more complex in use and implementation and is larger in size codewise if you have memory constraints. Don Brooks dab@netcom.com --------------------------- Newsgroups: gnu.g++.help,comp.os.vxworks Subject: g++ for Intel 960 with exceptions ??? Date: Fri, 11 Oct 1996 15:05:22 GMT From: 100567.1652@compuserve.com (Rolf Grzibek) Organization: Philips BTS Message-ID: <53lnkl$5cm@news.Dortmund.Germany.EU.net> Hi all, my current compiler GNU C++ version 2.6-95q2 (intel 80960) compiled by CC. running on Windows95 gave me the following warning: >warning: `catch', `throw', and `try' are all C++ reserved words >sorry, not implemented: exception handling not supported This was the first time I tried try { ... } catch () {} Does somebody know about a newer version of GCC that supports exceptions? Is this an Intel960 specific restriction, or are exceptions not supported on other target CPUs, too? Help or URLs will be appreciated. Thanks in advance Rolf - -- Rolf Grzibek 100567.1652@compuserve.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Wind River Web Site Date: Mon, 07 Oct 1996 20:50:00 GMT From: dhansen@btree.com (Dave Hansen) Organization: B-Tree Verification Systems, Inc. Message-ID: <53bq8m$9ct@cobra.Minn.Net> References: <53371j$8il@lantana.singnet.com.sg> <3258DF25.30B4@lsoc114x.ksc.nasa.gov> Reply-To: dhansen@btree.com Richard Price wrote: >user wrote: >> >> I would like to know more about VxWork, can anyone tell me the address of Wind River Web Site. >> >> My Email address is : chuayh@pacific.net.sg >> >> Thanks in advance.wind river url is www.wri.com >richard in titusville That would be http://www.wrs.com/ -=Dave dhansen@btree.com I can barely speak for myself, so I certainly can't speak for B-Tree --------------------------- Newsgroups: comp.os.vxworks Subject: C, JAVA Small, Fast Database Date: Fri, 11 Oct 96 11:54:38 From: Phil Thoren Organization: Web Professionals' Network Message-ID: visit http://www.vsonline.com or email phil@vsonline.com regards, VisionSoft --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Looking for pcmcia ata flashdisk driver Date: Fri, 11 Oct 1996 12:22:24 -0400 From: James Frey Organization: Anthem Electronics Inc - CT Message-ID: <325E7440.747E@pcnet.com> References: Reply-To: freyjw@pcnet.com Paul B. Lamere wrote: > > I am looking for a VxWorks driver for a pcmcia ata flashdisk > (SanDisk). Any pointers would be greatly appreciated. > > Paul Lamere Have you found one yet? Our customer is also trying to boot from a SanDisk FlashIDE drive without a floppy... - -- ================================================================== James Frey Voice: 203-575-1575 Sr. Field Applications Engineer Toll Free: 800-359-3516 Anthem Electronics Fax: 203-596-3232 61 Mattatuck Heights Road Voicemail: 800-726-8436 Box 2587 Waterbury, CT 06705 Email: freyjw@anthem-ct.com ================================================================== --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VxWorks interrupt handling Date: 11 Oct 1996 18:52:29 GMT From: elw@lvld.hp.com (Eddie Williamson) Organization: Hewlett Packard Loveland (Mountain IT) Site Message-ID: <53m51d$igr@hplvejl.lvld.hp.com> References: <53lnhh$mds@hplvejl.lvld.hp.com> I think I have an idea. Does this sound like it would work?? I can call intHandlerCreate() for my current ISR's. This returns a function pointer to the new interrupt handler which I can then plug into my interrupt vector table. Can anyone confirm/negate this? ______________________________________________________________________ Hewlett-Packard Eddie L. Williamson Jr. Manufacturing Test Division elw@lvld.hp.com 815 SW 14th Street, MS AU100 (970) 679-3674 Loveland, CO --------------------------- Newsgroups: comp.os.vxworks Subject: VxWorks on DSPs? Date: Wed, 02 Oct 1996 07:34:19 -0500 From: Eric Rehm Organization: Equator Technologies, Inc. Message-ID: <3252614B.71E3@equator.com> Reply-To: eric@equator.com Has anyone used VxWorks as a real-time O/S a DSP? If so, has it been sufficient for both soft- and hard-real-time tasks? /eric --------------------------- Newsgroups: comp.os.vxworks Subject: Looking for pcmcia ata flashdisk driver Date: Mon, 30 Sep 1996 17:25:10 GMT From: lamere@mv.mv.com (Paul B. Lamere) Organization: MV Communications, Inc. Message-ID: I am looking for a VxWorks driver for a pcmcia ata flashdisk (SanDisk). Any pointers would be greatly appreciated. Paul Lamere --------------------------- Newsgroups: comp.os.vxworks Subject: Re: tcp/ip performance survey Date: 11 Oct 1996 22:16:43 -0700 From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star, Oakland, CA Message-ID: <53n9jr$435@idiom.com> References: <199610102316.TAA26967@alioth.CCCHP> Paul Banta (banta@alioth.cebaf.gov) wrote: : The system runs steady at around 4.9 MBytes/s. It will occasionally peak : at just over 5MBytes/s (the driver has yet to be tweaked so I'm hoping that : this number will go up. I still have CPU left at this rate). But, it : doesn't run steady for long. After 30-60 seconds it begins to slowly : degrade and the data rate ultimately drops dramatically. I'm guessing that : I'm stuck in some funky tcp state which, of course, could be happening due : to a driver bug, but I'd like to hear about other experiences. This sounds like you are running out of cluster mbufs after a while and starts threshing because of lack of cluster mbufs. If you're doing the ususal mbuf loan tricks in the driver, instrument your loaned mbuf and cluster mbuf code in the driver to keep track of their behavior (i.e. how many gets used, how close you get to threshing due to lack of the number of available cluster & loan mbufs). - -- http://peacefulstar.oakland.ca.us mailto:hjb@peacefulstar.oakland.ca.us --------------------------- Newsgroups: comp.os.vxworks Subject: Re: netperf & ttcp Date: 11 Oct 1996 22:17:55 -0700 From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star, Oakland, CA Message-ID: <53n9m3$458@idiom.com> References: <199610102247.SAA26945@alioth.CCCHP> Paul Banta (banta@alioth.cebaf.gov) wrote: : Has anyone ported these utilities to VxWorks ? There is a very quick-and-dirty 'ttcp' port in http://peacefulstar.oakland.ca.us/vxhacks.html Use it at your own risk! :-) - -- http://peacefulstar.oakland.ca.us mailto:hjb@peacefulstar.oakland.ca.us --------------------------- End of New-News digest ********************** From 100656.1507@compuserve.com Sat Oct 12 10:30:53 1996 From: John Shaw <100656.1507@compuserve.com> Date: Sat Oct 12 10:30:55 PDT 1996 Subject: Re: VxWorks interrupt handling Regarding the question from Eddie L. Williamson Jr., >I can call intHandlerCreate() for my current ISR's. This returns >a function pointer to the new interrupt handler which I can then >plug into my interrupt vector table. >Can anyone confirm/negate this? Yup, intConnect () calls intHandlerCreate () and follows thru with this by calling intVecSet (). This function is described below (from the VxWorks manual pages). This routine attaches an exception/interrupt/trap handler to a specified vector. The vector is specified as an offset into the CPU's vector table. This vector table starts, by default, at: MC680x0: 0 SPARC: 0x1000 i960: `sysIntTable' in sysLib R3000/R4000: `excBsrTbl' in excArchLib 29030/29200: `LOCAL_MEM_LOCAL_ADRS' in target/config.h i386/i486: 0 However, the vector table may be set to start at any address with intVecBaseSet() (on CPUs for which it is available). The vector table is set up in usrInit(). --------------------------------------------------------------------------------- --------------------------- John Shaw, Shade (Computer Services) Limited, Calne, Wiltshire, England, UK. 100656.1507@compuserve.com, alas.... From daemon@csg.lbl.gov Sun Oct 13 04:01:25 1996 From: daemon@csg.lbl.gov Date: Sun Oct 13 04:03:21 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sun Oct 13 04:01:22 PDT 1996 Subject: UK jobs, Digital Video and Audio, HW and vxWorks Subject: zmodem Subject: [Q] Where is time zone information? Subject: Re: TCP Acknowledgement Subject: Re: Flash Disk Driver for VxWorks Subject: termios/stty VxWorks equivalent? Subject: varargs on Tornado PPC ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: UK jobs, Digital Video and Audio, HW and vxWorks Date: Tue, 1 Oct 1996 22:02:09 +0100 From: rms Organization: X Message-ID: <895jCBARbYUyEw1f@carapace.demon.co.uk> Reply-To: richard@carapace.demon.co.uk Oxtel is a fast expanding company operating in the Professional Broadcast market with products designed for digital video/audio presentation and monitoring. Oxtel presently employs 16 people at it's premises in Didcot, England and is looking to expand it's design team from 4 to 7 people. Oxtel is looking for bright self motivated designers seeking the challenge that a small company environment brings. Prospective employees will have good communication skills and will need to be able to take projects from the proposal stage right through to manufacture. Candidates need to have the relevant residence qualifications for working in the UK. We have three immediate vacancies: Software Engineer Package Salary, with profit share, by negotiation Experience 3+ years experience Degree qualified C/C++ experience essential Experience of real time operating systems such as VxWorks essential MC683xx/Intel x86 family experience useful Windows NT experience useful Ethernet experience useful Knowledge of processor architecture and hardware useful Responsibilities Design of real time software in C/C++ using MC683xx Design of real time software in C/C++ using Intel x86 and VxWorks Debugging new hardware at the hardware/software interface - ------------------------------------------------------------------------ - ---- Oxtel are also looking for two hardware engineers, one experienced and one junior. These people will be required to work on digital projects employing DSP, FPGA and Processor technology. The projects will be in the area of digital audio/digital video processing for the professional broadcast market. We would like the experienced engineer to be knowledgeable in our industry and should therefore have many of the skills listed as essential. The junior engineer needs to bring in similar skills at a less developed level but need not have all the skills listed as essential. Design Engineers Package Salary, with profit share, by negotiation Experience Degree qualified 3 years + (broadcast experience preferred) Experience of FPGA (Xilinx), EPLD useful/essential Broadcast digital video design experience useful/essential Broadcast digital audio design experience useful 270Mb/s serial video experience useful Experience of DSP (M56000) useful Experience of Intel x86 architectures with ISA/PCI useful MPEG experience useful Schematic capture using OrCAD or similar Responsibilities Specification, design and transfer to manufacturing of products involving: Embedded processors systems with Ethernet etc. DSP (56000 family) hardware and software for audio processing FPGA (Xilinx etc.) hardware for complex high speed video processing Analogue interfaces for video/audio + related analogue processing CVs By Email please to richard.simmons@oxtel.com - -- rms --------------------------- Newsgroups: comp.os.vxworks Subject: zmodem Date: Wed, 02 Oct 1996 17:32:56 -0400 From: Robert DeMartino Organization: Gerber Systems Corporation Message-ID: <3252DF88.19C9@postoffice.worldnet.att.net> Does anybody know where to find a zmodem protocol file transfer program for vxworks? --------------------------- Newsgroups: comp.os.vxworks Subject: [Q] Where is time zone information? Date: 2 Oct 1996 22:03:53 GMT From: alexm@vlsi.gsfc.nasa.gov (Alex Measday) Organization: NASA Goddard Space Flight Center -- Greenbelt, Maryland USA Message-ID: <52uos9$lf1@post.gsfc.nasa.gov> One of my programs calls strftime(2) to format the current time and "%Z" is specified in the format string for the time zone. On some of the CPUs in our rack, the strftime() call succeeds and returns "UTC" for the time zone. On one of the cards, however, the strftime() call blows up (access fault). The ansiTime(1) man page says that the time zone information is retrieved from an environment variable, TIMEZONE, or, if TIMEZONE is not defined, from the locale. If I explicitly define TIMEZONE with a putenv(2) on the offending CPU, my call to strftime() works. On the other CPUs, however, I can find no evidence of TIMEZONE being set. (Some slight delving into ANSI C locales was not promising.) Why do most of the cards seem to know the time zone (or do their random time zone pointers just happen to point to null strings) and one card doesn't? What's the recommended way of setting the time zone on a system? Thanks, Alex ======================================================================== | Alex Measday Code 521 - Microelectronic Systems Branch | | Integral Systems, Inc. NASA Goddard Space Flight Center | | c.a.measday@ieee.org Greenbelt, MD | ======================================================================== ) http://purl.oclc.org/net/alex/ ( ======================================================================== --------------------------- Newsgroups: comp.os.vxworks Subject: Re: TCP Acknowledgement Date: 11 Oct 1996 22:02:16 -0700 From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star, Oakland, CA Message-ID: <53n8oo$3jj@idiom.com> References: <53ed53$52d@newsserver.rrzn.uni-hannover.de> <325C1550.4BA9@peacefulstar.oakland.ca.us> Ralf Degner (degner@pallas.amp.uni-hannover.de) wrote: : For example at SunOS TCP_NODELAY is a kernel flag +> recompile all (if : you own it). You don't need source code, nor do you need to recompile SunOS to change this. You can use setsockopt TCP_NODELAY for this on SunOS as well. Now, if you want to globally control this (for all sockets), there is a SunOS kernel global variable tcpnodelack (or something like that) can be changed using adb on the kernel image (no need to recompile anything). Alternatively you can change this in the tunable parameters listed in one of the kernel reconfigure "/sys" files (no need to have full SunOS source code). : I think, there is. If the TCP code is not able to add the ACK to a normal : datapackage, it will send an ACK only packed after a certain time. At : SunOS this time is 50 ms. I measured 200 ms with VxWorks. I would be very : very happy if I can set this to about 50 ms. Well, we could argue on the term "default ACK delay" but if you are concerned with this particular case of ACK behavior ( it is just one of the behaviors you will see on the TCP connection ). This is obviously due to a sequence of TCP segments arriving on VxWorks side in perfect order, which makes TCP code to invoke simple delayed ACK mechanism. In both SunOS and VxWorks this type of simple delayed ACK processing is done by a TCP fast-timeout routine. In older releases of SunOS this used to be set to 200ms (5 ticks per second) which is what the original BSD UNIX code uses. The last SunOS kernel I hacked on was 4.1.1, so I could be out of touch. However, it could very well be because SunOS is receiving out-of-order packets and not delaying ACK's. In any case, VxWorks by default (unless they changed it since the last time I worked on their TCP code) uses 200ms TCP fast timeouts. If you want to crank this up to 50ms, you will need to ask WRS to send you source code for protocol fast timeout initialization code. Or, you can see how this is done in the original 4.3 Tahoe code and disassemble the pffasttimo() routine and replace the timeout arguments being passed to the watchdog init routines. However, this is NOT the "default ACK delay for TCP". There is no such thing! :-) Good luck! - -- http://peacefulstar.oakland.ca.us mailto:hjb@peacefulstar.oakland.ca.us --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Flash Disk Driver for VxWorks Date: 12 Oct 1996 13:58:23 -0700 From: nigel@idiom.com (Nigel Standing) Organization: A poorly-installed InterNetNews site Message-ID: References: <53214o$fdv@sf18.dseg.ti.com> <533mni$1pt@uuneo.neosoft.com> <3255EFB0.4DB@peacefulstar.oakland.ca.us> Actually, I'm surprised that Leonid Rosenboim hasn't already contributed to this thread. His company has a VxWorks flash file system, which he sells (in source) for about $12k. His e-mail address is "leonid@rst.co.il". Nigel - --------------------------- Newsgroups: comp.os.vxworks Subject: termios/stty VxWorks equivalent? Date: Wed, 02 Oct 1996 18:17:47 -0700 From: Eric Knudstrup Organization: Netcom Message-ID: <3253143B.2AC30AC0@mammoth-tech.com> I was wondering if there is anything similar in VxWorks. Thanks, Eric --------------------------- Newsgroups: comp.os.vxworks Subject: varargs on Tornado PPC Date: Thu, 10 Oct 1996 08:42:28 -0700 From: Norm Lamarra Organization: JPL Message-ID: <325D1964.2968@normnt.jpl.nasa.gov> Reply-To: Norm.Lamarra@jpl.nasa.gov Hi, We're experiencing trouble with porting code from Tornado/68K to Tornado/PPC which appears to be related to differences in varargs handling (we can't change everything to stdargs). Any references, suggestions, similar experiences, known problems. Thanks - -- Norman Lamarra Network Engineering Section 394 Distributed Computing and System Engineering Group Jet Propulsion Lab 4800 Oak Grove, M/S 525-3682 Pasadena, CA 91109 (voice) (818)306-6148 (fax) (818)306-6969 email: Norm.Lamarra@jpl.nasa.gov --------------------------- End of New-News digest ********************** From js@eroch.mc.xerox.com Sun Oct 13 04:31:04 1996 From: js@eroch.mc.xerox.com (Jeffrey R. Szczepanski) Date: Sun Oct 13 04:31:06 PDT 1996 Subject: Re: Solaris/PPC compiler for VxWorks 5.2 - Status? > >> about a year ago, we were told that there really wasn't a usable > >> compiler for the PowerPC available for Solaris host platforms -- > >> has this improved? > >> I'd love to hear from anyone actively developing VxWorks code under > >> these conditions, as we're considering making that move ourselves. > > There is one available from WRS now. Tornado (VxWorks 5.3) comes with a > version of gcc that cross compiles form solaris2 to ppc. It should also > work for VxWorks 5.2 > Also, We run the Diab Data Compiler for Power PC on Sun OS 4.1.4 and on Solaris 2.5 with the binary compatibility installed without problems. -Jeff ------------------------------------------------------------------------------ Jeffrey R. Szczepanski Phone: (716) 264-6613 Xerox Engineering Systems FAX: (716) 264-6612 300 Main Street East Rochester, NY 14445 Email: js@eroch.mc.xerox.com ------------------------------------------------------------------------------ From mea@mclean.sparta.com Mon Oct 14 11:13:29 1996 From: Mike Anderson Date: Mon Oct 14 11:13:31 PDT 1996 Subject: Re: tcp/ip performance survey At 04:17 PM 10/10/96 PDT, you wrote: >Submitted-by banta@alioth.cebaf.gov Thu Oct 10 16:17:53 1996 >Submitted-by: Paul Banta > >I'd like to take an quick survey of experience with tcp/ip in high bandwidth >applications. Particularly where VxWorks targets are sending as much data >as they can muster through a network interface. > >What is the average output bandwidth? >What is the peak output bandwidth? >Does the system perform well over a long period of time? >Are there any special tcp parameter modifications (e.g. adjusting the send and >receive sliding window sizes)? > For what it's worth, one of my Interphase 5211 FDDI customers has reported a steady 5.84 MBytes/sec using SPARTA's driver and the Motorola 1603 board. This was using our driver in VME64 unlimited burst mode. The numbers were repeatable and did not seem to degrade over time. This was measured using the ubiquitous blaster/blastee programs using TCP stream sockets and their send/recv buffers set to the 48Kbyte max via setsockopt. HTH, =============================================================================== __ Real-Time System Development, Integration, Training and Services //\\ // \\ Mike Anderson // /\ \\ Chief Engineer Voice : (703) 448-0210 ext. 235 // / \ \\ SPARTA, Inc. FAX : (703) 734-3323 // \ \\ 7926 Jones Branch Drive EMAIL : mea@mclean.sparta.com \\ \ // Suite 900 Web : http://www.mclean.sparta.com \\ \ / // McLean, VA 22102 \\ \/ // "Software development is like making \\ // a baby... You can't make a baby in one \\// month by impregnating nine women. -- "Pride in Performance" Some things just take time." =============================================================================== From aguilar_ronald@msmail.lansce.lanl.gov Tue Oct 15 08:28:48 1996 From: "Aguilar, Ronald" Date: Tue Oct 15 08:28:50 PDT 1996 Subject: Tornado,WindowsNT,MVME2600 We are looking at using Tornado under Windows NT 4.0 with the MVME 2600 = =3D Processor. Does anyone have any experience with this system? I am most interested in the BSP for the MVME 2600 processor, and how =3D Tornado performs under Windows NT 4.0. Any information and comments would be appreciated. Thanks in advance, Ron Aguilar Manuel Lujan Jr. Neutron Scattering Center MS H805 Los Alamos National Lab Los Alamos, NM 87545 USA Phone: (505)667-8369 Fax: (505)665-2676 E-Mail: rgaguilar@lanl.gov From beeman1@popcorn.llnl.gov Tue Oct 15 15:16:16 1996 From: Bart Beeman Date: Tue Oct 15 15:16:18 PDT 1996 Subject: Industry Packs Hello, I've just gotten my system up and running, mv162 (with several industry packs) and Tornado. Everything seems to be running OK, I can boot and even get a windsh window. My problem is I have not figured out how to talk to my industry packs. If I stop the autoboot and play with memory (fff58000 - IP(A)) I can get to the locations and get the expected response from the IP. The problems start when trying to do the same thing from windsh. ->d 0xfff58000 WTX Error 0x100ca (AGENT_MEM_ACCESS_ERROR) value = -1 = (0xffffffff) At the bottom of /config/mv162/config.h there is some verbage eluding to support is determined by INDUSTRY_PACK_C_FILE & INDUSTRY_PACK_H_FILE but I can't seem to find any o fthe files they talk about. Is this something that would needed to be taken care of at install time? Could someone please point me in the right direction? Thanks Bart P.S. I just registered with the exploder and don't know if it has gone through yet, please respond directly to beeman1@llnl.gov Bart Beeman Lawrence Livermore National Laboratory beeman1@popcorn.llnl.gov (510) 422-5938 voice From royg@semantic.com Tue Oct 15 19:00:15 1996 From: Roy Gordon Date: Tue Oct 15 19:00:18 PDT 1996 Subject: C++ on 5.2 We want to develop a mixed C/C++ application on 5.2 using both VxSim and then a MIPs BSP with a 4650 processor. Naturally, we want source level debugging for the BSP. Has anyone ever done this successfully? If not for the 4650 then for the 4600? If so, what products did you use? If you had problems with the development tools chosen, what were they? Other success/failures with a C/C++ mix on 5.2? TIA (very much). -- roy (Please forgive if this has already been posted, but it looked to me that my first attempt bounced.) From c.j.slominski@larc.nasa.gov Wed Oct 16 04:29:08 1996 From: Christopher Slominski Date: Wed Oct 16 04:29:10 PDT 1996 Subject: PowerPC 604 software -- Has anyone developed VxWorks software drivers for the PowerPC 604 board's VGA graphics output, mouse and keyboard devices? *----------------------------------------* | Christopher J. Slominski | | Computer Sciences Corporation (CSC) | | 3217 N. Armistead Ave. | | Hampton, VA 23666 | | (757) 766-8258 | | c.j.slominski@larc.nasa.gov | | NASA Mail Stop 931 | *----------------------------------------* From MikeMcQuade@connect.com Wed Oct 16 07:16:31 1996 From: Mike McQuade Date: Wed Oct 16 07:16:33 PDT 1996 Subject: Re: PowerPC 604 software At 04:29 AM 10/16/96 PDT, you wrote: >Submitted-by c.j.slominski@larc.nasa.gov Wed Oct 16 04:29:08 1996 >Submitted-by: Christopher Slominski > >-- Has anyone developed VxWorks software drivers for the PowerPC 604 >board's VGA graphics output, mouse and keyboard devices? > Ive been using a product called SDL from Raster Graphics. You can get a demo for the 1603 / 1604 at: http://www.rastergr.com/ SDL is very easy to use, I was able to get a framework user interface up and running very quickly. MikeMcQuade@connect.com From tschlemm@nrao.edu Wed Oct 16 07:29:55 1996 From: Tessy Schlemmer Date: Wed Oct 16 07:29:57 PDT 1996 Subject: Open Position The National Radio Astronomy Observatory has the following JOB VACANCY: TITLE: PROGRAMMER - AIPS++ PROJECT LOCATION: Charlottesville, Virginia DESCRIPTION: The AIPS++ project is an international collaboration by a consortium of radio observatories with the goal of developing a software package for the processing of radio astronomical data. The project is managed by the NRAO. The duties will be to participate in the design and coding of interactive, GUI-based data reduction software written using C++ and glish (an interpreted and distributed language with a binding to Tk-widgets). QUALIFICATIONS: Substantial experience in C++ and software engineering is required. Experience developing distributed, network applications in the Unix environment is desirable. A background in astronomy would be advantageous. APPLICATION DEADLINE: November 1, 1996. Interested persons should contact: National Radio Astronomy Observatory P.O. Box 0 Socorro, New Mexico 87801 Attn: Judi Lowell Fax: (505) 835-7027 Email: Jlowell@nrao.edu Applicants should provide: -Letter of application; -a current curriculum vitae; -the names of three references whom we may contact as necessary. Further information is available for both the AIPS++ Project and NRAO from the NRAO Home Page http://www.nrao.edu. EOE M/F/V/H From MikeMcQuade@connect.com Wed Oct 16 09:37:24 1996 From: Mike McQuade Date: Wed Oct 16 09:37:26 PDT 1996 Subject: PowerPC 1603/4 flash roms. Has anyone out there burned their own boot roms for Motorola 1603/4 boards ? if so, can you share how you are doing it ? what device programmer, can the parts be flashed by mot-bug ? thanks. MikeMcQuade@connect.com From MSCHMIDT@spar.ca Wed Oct 16 12:59:44 1996 From: Michael Schmidt Date: Wed Oct 16 13:02:21 PDT 1996 Subject: subscribe subscribe From ctas_majors@acm.org Wed Oct 16 15:50:50 1996 From: ctas_majors@acm.org Date: Wed Oct 16 15:50:52 PDT 1996 Subject: RE: PowerPC 1603/4 flash roms. Has anyone out there burned their own boot roms for Motorola 1603/4 boards ? if so, can you share how you are doing it ? what device programmer, can the parts be flashed by mot-bug ? thanks. MikeMcQuade@connect.com -----------------End of Original Message----------------- We have copied the boot rom we got from Wind River into other devices. We're using a Data I/O ChipLab programmer with a special adapter for the 29F040 package. We programmed over the mot-bug roms that came with the boards. If you need more detailed into, please email me. Michael Majors Raytheon E-Systems ------------------------------------- E-mail: ctas_majors@acm.org Date: 10/16/96 Time: 5:45:57 PM This message was sent by Chameleon ------------------------------------- From daemon@csg.lbl.gov Thu Oct 17 04:00:38 1996 From: daemon@csg.lbl.gov Date: Thu Oct 17 04:00:40 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Thu Oct 17 04:00:35 PDT 1996 Subject: Re: TCP over a serial link Subject: Re: tcp/ip performance survey ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: TCP over a serial link Date: 3 Oct 1996 03:33:28 GMT From: lazarus@sun-valley.Stanford.EDU (Howard Wang) Organization: Aerospace Robotics Lab, Stanford University Message-ID: <52vc68$g2v@nntp.Stanford.EDU> Oops, >gives IP over serial links. That means your code that does TCP/UDP >doesn't have to change. You only need to change routes and gateways and >such at the operating system level when switching between TCP and UDP. I meant between SLIP and Ethernet. Howard --------------------------- Newsgroups: comp.os.vxworks Subject: Re: tcp/ip performance survey Date: Tue, 15 Oct 1996 16:16:09 -0500 From: David Perkinson Organization: TRW, Inc. Message-ID: <3263FF19.41C6@boris.msfc.nasa.gov> References: <199610102316.TAA26967@alioth.CCCHP> Paul Banta wrote: > > I'd like to take an quick survey of experience with tcp/ip in high bandwidth > applications. Particularly where VxWorks targets are sending as much data > as they can muster through a network interface. > > What is the average output bandwidth? > What is the peak output bandwidth? > Does the system perform well over a long period of time? > Are there any special tcp parameter modifications (e.g. adjusting the send and > receive sliding window sizes)? > > I'll collect the results and distribute them by request. > > My curiosity about this stems from my current problem: > I have a Radstone PPC604 (100MHz PowerPC) with an ATM adapter on it. The > driver is home grown (with some help from various people around the net). > The other end of the fiber is plugged into a sun ultra 1. The unix system > has a process that eats data from a socket as fast as it can. The vxworks > target crams data through a socket as fast as it can. > > The system runs steady at around 4.9 MBytes/s. It will occasionally peak > at just over 5MBytes/s (the driver has yet to be tweaked so I'm hoping that > this number will go up. I still have CPU left at this rate). But, it > doesn't run steady for long. After 30-60 seconds it begins to slowly > degrade and the data rate ultimately drops dramatically. I'm guessing that > I'm stuck in some funky tcp state which, of course, could be happening due > to a driver bug, but I'd like to hear about other experiences. > > Thanks, > > -- > Paul Banta voice: 757-269-5829 > TJNAF, MS 12H fax: 757-269-5800 > 12000 Jefferson Avenue email: banta@cebaf.gov > Newport News, VA 23606 Thomas Jefferson National Accelerator Facility We are not using the TCP/IP stack but we would like to in the future on a 100baseT network. Using the raw API to the Intel ethernet chips on our Synergy 040, we can sustain any data rate at any packet size that our HP LAN analyzer can generate. This include 10 Mbit/s at 256 byte packets and up. We have sustained 8.3 Mbit/s with a 100 byte packet size. The data is buffered 20k at a time and sent across the backplane to a Huerikon Nitro 60 where is is logged to a SCSI disk. This tells me that the hardware can handle it. The question is, can the TCP/IP header be processed with something other than the main CPU so that the TCP stack does not choke out other tasks. - -- =================================================================== David Perkinson, TRW | Opinions that may be expressed in | this email are my own and not perkinsond@boris.msfc.nasa.gov | neccessarily that of TRW. =================================================================== --------------------------- End of New-News digest ********************** From han2326@aminet.co.kr Thu Oct 17 07:57:38 1996 From: "Han, Jeong Soo" Date: Thu Oct 17 07:58:04 PDT 1996 Subject: serial booting Hello, I want to know how to boot using serial link(RS232) and how to set the console speed. My host is sun solaris 2.4(have no slip protocol ?), and using vxWorks 5.2. I know that this system have no slip protocol. If so, could you tell me where I get the protocol? Your detail description would be highly appreciated. Thanks, J.S.HAN ------------------- H D P I C Research Center From daemon@csg.lbl.gov Fri Oct 18 04:00:38 1996 From: daemon@csg.lbl.gov Date: Fri Oct 18 04:00:40 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Fri Oct 18 04:00:36 PDT 1996 Subject: Q: VME card operation Subject: AT&T Assembler1 Subject: Has anyone developed a VxWorks Ethernet driver for the 68EN302? Subject: Re: mailbox/location monitor interrupts? ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Q: VME card operation Date: Thu, 17 Oct 1996 15:52:43 -0400 From: CK Wong Organization: Nortel Technology Message-ID: <32668E8B.5E35@nortel.ca> Please help me on the following questions : 1. On the VME bus, is there a master card that controls the bus like the CPU control the system ISA bus on a PC ? 2. Can any VME bus card, raise interrupt to another VME card ? Thank you in advance. --------------------------- Newsgroups: comp.os.vxworks Subject: AT&T Assembler1 Date: 17 Oct 1996 19:09:32 -0400 From: scotttsx@aol.com (ScottTSX) Organization: America Online, Inc. (1-800-827-6364) Message-ID: <546ebc$o5n@newsbf02.news.aol.com> Reply-To: scotttsx@aol.com (ScottTSX) Sender: root@newsbf02.news.aol.com Hello: Can anyone tell me where to find AT&T System V 386 Assembler syntax ( used by GNU ). I wrote some test routines in Microsoft MASM and now we want to continue testing under VXworks. I can't get 'stc', 'cmc', and 'bt' to assemble correctly. Somebody give me a hint. Thanks for any help. I'm grateful. Scott Shaw --------------------------- Newsgroups: comp.os.vxworks Subject: Has anyone developed a VxWorks Ethernet driver for the 68EN302? Date: 17 Oct 1996 13:51:21 -0700 From: ralerche@lindy.stanford.edu (Robert A. Lerche) Organization: MSB Associates Message-ID: <546689$98r@lindy.stanford.edu> I have a client planning to use VxWorks and the 68EN302 in an embedded system design. Wind River doesn't seem to have this driver available yet. I looked at Motorola's documentation of both the 68EN302 and the 68EN360; there are similarities but they are far from identical -- the 360 (an older design, I guess) swiped one of the SCC's, whereas the 302 incorporates an Ethernet protocol machine. Can anyone out there help me? Thanks. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: mailbox/location monitor interrupts? Date: Thu, 17 Oct 1996 22:31:25 GMT From: fuda@norden.com (Thomas Fuda) Organization: Northrop Grumman Corporation Message-ID: References: <53h858$j0f@library.clarkson.edu> Sender: usenet@tron.bwi.wec.com (USEnet News Poster) kangy@cauchy.clarkson.edu (Yimin Kang) wrote: >Hi, there, >I need some information of how to use the mailbox/localint monitor >interrupts. >Please send response to yimin@array.ca. >Thanks. What kind of information are you looking for specifically? What kind of target processor are you using? We are in the process of implementing mailbox interrupts on a VME based 486 single board computer and may be able to assist. --------------------------- End of New-News digest ********************** From mea@mclean.sparta.com Fri Oct 18 06:47:09 1996 From: Mike Anderson Date: Fri Oct 18 06:47:12 PDT 1996 Subject: Re: VME card operation At 04:00 AM 10/18/96 PDT, you wrote: >Subject: Q: VME card operation >Date: Thu, 17 Oct 1996 15:52:43 -0400 >From: CK Wong >Organization: Nortel Technology >Message-ID: <32668E8B.5E35@nortel.ca> > >Please help me on the following questions : > >1. On the VME bus, is there a master card that controls the bus > like the CPU control the system ISA bus on a PC ? > >2. Can any VME bus card, raise interrupt to another VME card ? > Unlike typical ISA implementations, the VMEbus is a multi-master, asynchronous bus architecture (PCI bus is synchronous). In the VMEbus, there is a slot 0 "controller" that is responsible for bus arbitration (4 bus grant levels in priority, round-robin and single-level arbitration mode) and for asserting the SYSCLK signal (a 16MHz reference clock) onto the bus. This controller function is typically built into most modern SBCs, although it used to be found on a separate boards such as the Mot MVME-25 system controller board. However, it should be noted that the SYSCON feature is not typically a "CPU" per se. Provided that the interrupt responder has unmasked the interrupt level being generated, yes. Again, the interrupt structure in VMEbus is quite a bit more complex (i.e., flexible) than what you'll find in an ISA bus architecture. Interrupts on the VMEbus are typically followed by a vector (i.e., a bit of data) that describes essentially "who" generated the interrupt -- not just that an interrupt occured as in ISA. As for mailbox interrupts, think of them as each CPU on the bus monitoring a specific location in the VME address space (that CPU's "mailbox"). When that address is strobed on the bus, the CPU that's watching that mailbox will generate an on-card (i.e., local) interrupt. Some mailboxes can then be read and the data in the mailbox used to dispatch an interrupt vector just like the typical VMEbus interrupt. This ability varies from vendor to vendor and from card-type to card-type from the same vendor. Ask your vendor if their mailboxes can be read. Location monitors, on the other hand, are essentially "broadcast" mailboxes. That is, you can set some SBCs up so that several CPUs are monitoring the same address strobe. When that address occurs on the bus, all of the SBCs that are set up to watch the location generate an on-card interrupt at the same time. This broadcast capability cannot be duplicated with conventional VMEbus interrupts (they can have more than one generator but only one responder). HTH, =============================================================================== __ Real-Time System Development, Integration, Training and Services //\\ // \\ Mike Anderson // /\ \\ Chief Engineer Voice : (703) 448-0210 ext. 235 // / \ \\ SPARTA, Inc. FAX : (703) 734-3323 // \ \\ 7926 Jones Branch Drive EMAIL : mea@mclean.sparta.com \\ \ // Suite 900 Web : http://www.mclean.sparta.com \\ \ / // McLean, VA 22102 \\ \/ // "Software development is like making \\ // a baby... You can't make a baby in one \\// month by impregnating nine women. -- "Pride in Performance" Some things just take time." =============================================================================== From syed@sunriver.com Fri Oct 18 10:35:08 1996 From: syed@sunriver.com (Syed Naseem) Date: Fri Oct 18 10:35:25 PDT 1996 Subject: subscribe -- Syed Naseem From shenry@absolom.eng.vmic.com Fri Oct 18 11:51:33 1996 From: Shawn Henry Date: Fri Oct 18 11:54:30 PDT 1996 Subject: Digital Semiconductor 21142 Driver Hello, I am looking for a Digital Semiconductor 21142 ethernet driver. A VxWorks version would be great, source code for a UNIX version would be nice, source code for anything else would be good. thanks in advance Shawn Henry email shawn.henry@vmic.com From grano@serval.kla.com Fri Oct 18 19:03:25 1996 From: grano@serval.kla.com (David A. Grano x6495) Date: Fri Oct 18 19:03:27 PDT 1996 Subject: subscribe David A. Grano KLA Instruments M/S F1-W700 P.O. Box 49055 160 Rio Robles San Jose, CA 95161-9055 Phone: 408-468-6495 Fax: 408-434-4273 E-mail: d_grano@kla.com Inspection equipment Custom 680x0 Custom PowerPC From fareed@aloha.nmsu.edu Sat Oct 19 18:20:28 1996 From: Fareed Masarrat Date: Sat Oct 19 18:20:30 PDT 1996 Subject: Subscibe ---------------------------------------------------------------------------- /////////////////////////// Fareed Masarrat |||| /\ ||| ||||||||||| / 2600 E.Idaho #203 P | | | | || | | | / Las Cruces NM 88011 || |__| ||| || || | | / OR | | | |\ | | | | / ---- | | | | \ ||||||| ||| / Center For Space Telemetring /////////////////////////// Klipsch School of Elect & Comp Engg New Mexico State University Email: fareed@aloha.nmsu.edu Las Cruces NM 88011 masarrat@cs.nmsu.edu WWW: (505) 522-5923(home) http://web.nmsu.edu/~fmasarra (505) 646-1911(office) (505) 646-6417(Lab) ---------------------------------------------------------------------------- From leif@tsl.uu.se Sun Oct 20 08:34:44 1996 From: leif@tsl.uu.se Date: Sun Oct 20 08:34:46 PDT 1996 Subject: Re: Digital Semiconductor 21142 Driver The Linux distribution have drivers for 21040,21041,21140. Check http://sunsite.unc.edu/mdw/HOWTO/Ethernet-HOWTO-5.html if this is of any help. /Leif leif@tsl.uu.se From muebel@Myriadlogic.com Sun Oct 20 14:13:58 1996 From: muebel@Myriadlogic.com (Mark Uebel) Date: Sun Oct 20 14:14:00 PDT 1996 Subject: Need a VxWorks Programmer? I am an experienced VxWorks programmer (5 years) looking to "hang my own shingle" in Charlotte, NC. Currently, I'm looking for contract work in North and South Carolina. My experience ranges from motion control and robotic systems, to high-speed VME data transfer boards. If you're in need of a talented, real-time system developer, please call me at 202-547-8426. I will be happy to discuss your application and forward my resume on to you. Mark Uebel ******************************************************** * * * Mark Uebel * * muebel@myriadlogic.com * * (301) 588-1900 [office] * * * ******************************************************** From malcolm.cifuentes@nautronix.com.au Sun Oct 20 20:17:32 1996 From: malcolm.cifuentes@nautronix.com.au (Malcolm Cifuentes) Date: Sun Oct 20 20:17:36 PDT 1996 Subject: IOSTREAMS Hi all, I have just installed the FCS version of tornado for PPC 604, running on solaris 2.5. I have noticed that iostreams(GNU) doesn't seem to exist in this version of the library released. If I look at the library shipped with the BETA version, I find an object called cstreams.o which doesn't seem to be apart of the new shipped version. Has any one else encountered this problem and how did you work around it? Does any one know why this change has taken place? -- Malcolm Cifuentes ( Nautronix LTD malcom.cifuentes@nautronix.com.au PH : +61-09-430-5900 FAX : +61-09-430-5901 ) From daemon@csg.lbl.gov Mon Oct 21 04:00:35 1996 From: daemon@csg.lbl.gov Date: Mon Oct 21 04:00:38 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Mon Oct 21 04:00:33 PDT 1996 Subject: Re: Problem with "bcopy" ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Problem with "bcopy" Date: Thu, 17 Oct 1996 15:44:31 GMT From: breti@bit3.com (Bret Indrelee) Organization: Bit 3 Computer Corp Message-ID: References: <32641D36.41C67EA6@gdesystems.com> <54353u$c0s@badger.heurikon.com> In article <54353u$c0s@badger.heurikon.com>, Ted Friedl wrote: >In article <32641D36.41C67EA6@gdesystems.com>, >Michael Levasseur wrote: >:Hello, >: >:I'm using VxWorks 5.2 on a Heurikon Baja R4700 board and I'm having >:problems (difficulties) with the "bcopy" routine. Does anyone have >:the source for "bcopy"? [ snip ] >:When I write a word (a 16-bit object), "bcopy" is writting two >:bytes and not the one 16-bit word I'd expect to see written. The >:man pages say that "bcopy" uses the MOST efficient method! Two byte >:writes aren't the most efficient! This is just an inconvienience >:and I can live with this. > >I checked the source of 5.2/Mips bcopy (src/arc/mips/bALib.s). Your >observation is correct. Because CPU cycles are burned to figure out >alignments, any transfer under 4 bytes is transferred in bytes. The efficiency is evaluated assuming the destination is cacheable memory. This is a common assumption for the bcopy() implementations I've seen on various operating systems. Try using bcopyWords() instead. Although the documentation doesn't state it, it appears that VxWorks assumes that Word is 16-bit and Long is 32-bits. - -Bret - -- Bret Indrelee breti@bit3.com #include --------------------------- End of New-News digest ********************** From daemon@csg.lbl.gov Tue Oct 22 04:00:28 1996 From: daemon@csg.lbl.gov Date: Tue Oct 22 04:00:30 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Tue Oct 22 04:00:25 PDT 1996 Subject: rpcgen & Solaris 2.5.1 Subject: Re: Q: VME card operation Subject: Re: IOSTREAMS Subject: Re: Tornado and Source Control Subject: Re: pc-486 serial ports 3+4 ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: rpcgen & Solaris 2.5.1 Date: 8 Oct 1996 11:48:52 GMT From: dpf@orion.ksc.nasa.gov (dpf) Organization: NASA - Kenndy Space Center Message-ID: <53df34$8p9@itchy.ksc.nasa.gov> Reply-To: dpf@orion.ksc.nasa.gov Does anyone have a compiled version of "rpcgen" for VxWorks version 5.1.1 on the Solaris 2.5.1 platform? - -- David P. Floyd I-NET Space Services, Inc. dpf@orion.ksc.nasa.gov --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Q: VME card operation Date: 18 Oct 1996 12:41:04 GMT From: tfm@sei.cmu.edu (Theodore Marz) Organization: The Software Engineering Institute Message-ID: <547tt0$9hf@news.sei.cmu.edu> References: <32668E8B.5E35@nortel.ca> In article <32668E8B.5E35@nortel.ca>, CK Wong writes: |> Please help me on the following questions : |> |> 1. On the VME bus, is there a master card that controls the bus |> like the CPU control the system ISA bus on a PC ? |> |> 2. Can any VME bus card, raise interrupt to another VME card ? |> |> Thank you in advance. 1) VME is multi-mastered. Any card can post data on the bus, once it get's a bug grant. The card in Slot 0 (the leftmost slot) performs the bus arbitration. 2) Yes. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: IOSTREAMS Date: 21 Oct 1996 22:19:11 GMT From: j.gordon@dial.pipex.com (John Gordon) Organization: N/A Message-ID: <54gssv$19g@soap.news.pipex.net> References: <199610210319.LAA10282@synaps.nixltd.com.au> In article <199610210319.LAA10282@synaps.nixltd.com.au>, malcolm.cifuentes@nautronix.com.au says... > >Hi all, > I have just installed the FCS version of tornado for PPC 604, running > on solaris 2.5. I have noticed that iostreams(GNU) doesn't seem to exist > in this version of the library released. If I look at the library > shipped with the BETA version, I find an object called cstreams.o which > doesn't seem to be apart of the new shipped version. > > Has any one else encountered this problem and how did you work around it? > > Does any one know why this change has taken place? >-- >Malcolm Cifuentes >( > Nautronix LTD > malcom.cifuentes@nautronix.com.au > PH : +61-09-430-5900 > FAX : +61-09-430-5901 >) Hi, You are correct in that the iostreams support is not in FCS Tornado, but according to our distributor in the UK it is AT&T iostreams that is supplied by WRS to avoid a potential licensing problem with the GNU version of this library. As a result of this you need to buy an additional package containing a set of class libraries, one of which is the iostreams library you *need* to do any real work (or for that matter even hello world!) in C++! John... --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Tornado and Source Control Date: Tue, 08 Oct 1996 22:07:07 -0600 From: Wes Peters Organization: Softweyr LLC Message-ID: <325B24EB.41C67EA6@xmission.com> References: <9610041610.AA08358@lbl.gov> <3256A61E.4717@peacefulstar.oakland.ca.us> H.J. Bae wrote: > > I have used CVS successfully on both Solaris and WinNT for > a fairly large size project using Tornado. The simplest (perhaps > not the best) way to allow access to the CVS repository > over the LAN to workstation clients is to put the repository > on a WinNT server and add NFS server to the WinNT server. > This way, WinNT clients can access the repository over SMB > and Solaris clients can access the repository over NFS. > CVS has its own remote access protocol based on RSH, but > the free versions of CVS available for WinNT do not seem to > work well over RSH (it could be just pilot error). Ditto; we use HP-UX and Win95. Our CVS server resides on one of the HPs for now, we're building a PC with FreeBSD to run it on in the near future. We use NFS to share the CVSROOT directory among the HPs, and Samba for the Win95 machines. The WinNT executables for CVS (client, at least) seem to work acceptably under Win95. Personally, I really hate the idea of shelling out all that money for WinNT and NT-NFS for a server, when I can get better performance on the same PC hardward for free with FreeBSD, which includes NFS client and server, Samba SMB server, and CVS. ;^) - -- Wes Peters | Softweyr | Where am I, and what am I doing in this handbasket? Consulting | softweyr@xmission.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: pc-486 serial ports 3+4 Date: Tue, 22 Oct 1996 02:15:31 GMT From: afraser@hookup.net (Andrew Fraser) Organization: HookUp Communication Corporation, Ottawa, Ontario, CANADA Message-ID: <326c2af0.83458317@nnrp.ott.hookup.net> References: <326B827B.4D87@mail.imech.uni-duisburg.de> Stefan Vogel wrote: >Hi, > >I would like to use the serial ports 3 and 4 >on a pc-board. Has anyone experience in this >Subject?? > >Stefan > You will need to buy a serial I/O card which, of course, has the two ports, but the card must support the "high interrupts". This means it will look like a 16 bit card, plugging into the second EISA connector where INT 12, 13, 14 are. It won't actually be a 16-bit data card though. On the I/O card, configure the two ports to be COM3 and COM4, and to use two of those high interrupts. COM1 and COM2 will remain using INT3 and INT4. Each port now has it's own interrupt, as required. If you're using Windows 3, you also need to configure those ports via the Control Panel, I/O Ports, Advanced options. Leave the I/O address at the default, just set the interrupt level. Win95 probably has a similar configuration area, since the cards are usually not plug'n'play. One problem you might have is a conflict with the video card, if it supports XGA mode. When IBM designed that, they stupidly used the I/O address assigned for COM4 as one of the video card addresses. This means Windows will lock up very quickly if COM4 is installed (even if you're not using XGA). Either change the video card, or disable COM4 via it's jumper. Some of the new multi-format SVGA and other high res cards include XGA. And for anyone wanting to add more com ports to a laptop, Quatech has a nice PCMCIA card including two serial ports that I like. - -Andy - --------------------------------------------------------------------- ANDREW FRASER DY 4 Systems Inc email afraser@dy4.com Senior Hardware Designer 333 Palladium Dr. voice (613)599-9199x251 Product Support Group Kanata, Ontario fax (613)599-7777 Canada K2V 1A6 support support@dy4.com http://www.dy4.com sales sales@dy4.com - ------------------- CUSTOMER FIRST, QUALITY ALWAYS ------------------ --------------------------- End of New-News digest ********************** From SECKA@hqgate.orbital.com Tue Oct 22 06:05:19 1996 From: SECKA@hqgate.orbital.com Date: Tue Oct 22 06:05:23 PDT 1996 Subject: logMsg() and printf() with No Virtual Console Configuration: VxWorks 5.3/Tornado on a custom 68302-based board. Question: What is the behavior of logMsg/logTask and printf when there is no virtual console? Does the data fall into the "bit bucket" or will the software eventually hang, pended on a full output queue? ================================================================= = Suzie Eckardt = email: secka@hqgate.orbital.com = = Orbital Sciences Corp. = Phone: (703) 406-5478 = = 21700 Atlantic Boulevard = Fax: (703) 406-5553 = = Dulles, VA 20166 = ================================================================= From lyorg@Scorpio.Com Tue Oct 22 06:30:26 1996 From: lyorg@Scorpio.Com (Lyor Goldstein) Date: Tue Oct 22 06:30:28 PDT 1996 Subject: TAR & GZIP porting Hello, Has anyone ported TAR & GZIP to VxWorks ? Example: gzipPack (int in_fd, int out_fd) reads from "in_fd", packs and writes into "out_fd" gzipUnpack(int in_fd, int out_fd) ditto, only unpacks Thanks a lot, ***************************************************************** * Lyor Goldstein. Email : lyorg@scorpio.com * * Scorpio Communications Ltd. Phone : 972-3-5339654 ext 115 * * Software group Fax : 972-3-5339518 * * 1c Yoni Netanyahu st. * * Or Yehuda 60250 * * P.O.Box 564, Israel * ***************************************************************** From tja@radstone.co.uk Tue Oct 22 07:20:25 1996 From: Tim Allen Date: Tue Oct 22 07:20:27 PDT 1996 Subject: Re: logMsg() and printf() with No Virtual Console > Question: What is the behavior of logMsg/logTask and printf when > there is no virtual console? Does the data fall into the "bit bucket" > or will the software eventually hang, pended on a full output queue? The default standard I/O device for VxWorks is the serial console. When using the Tornado host shell (WindSh), instead of the target shell, printf and logMsg output data appears on the serial console. To redirect all I/O to WindSh (or virtual console, if it exists), either execute the TCL script given in the Tornado User's Guide, section 5.11.3 Tcl: Tornado Shell Initialization (or the shell commands given in section 5.4.4 Standard Input and Output). That way, if you're using the target shell the TCL script won't be executed, so all I/O remains on the serial console. If you're using WindSh, with no virtual console the output is sent to WindSh, and if you're using WindSh with the virtual console, the output will go to the virtual console. I think! +------------------------------------------------------------------+ | Tim Allen phone: +44 (0)1327 359444 | | Radstone Technology PLC, UK. fax: +44 (0)1327 358112 | +------------------------------------------------------------------+ From jjjurek@fly.aeromet.com Tue Oct 22 09:14:26 1996 From: "Jon Jurek" Date: Tue Oct 22 09:14:36 PDT 1996 Subject: SLIP interface between Windows95 and an MVME162 There have been a number of messages dealing with TCP over a serial link over the last few weeks, but I have not seen any that helps with my problem. I am using a MVME162 with Tornado on a Windows95 machine. I am trying to implement a SLIP protocol to boot VxWorks and have not had any success so far. In particular I would like information on setting up the host side. Any clues on software to use and how to set it up would be appreciated. ______________________________________ Jon Jurek Aeromet, Inc. Software Group Leader 112 Beechcraft Drive (918) 299-2621 Jones Riverside Airport jjjurek@fly.aeromet.com Tulsa, OK 74132 http://www.aeromet.com/ From kressig@ti.com Tue Oct 22 10:59:08 1996 From: Bob Kressig Date: Tue Oct 22 10:59:10 PDT 1996 Subject: re: TAR & GZIP porting > Submitted-by: lyorg@Scorpio.Com (Lyor Goldstein) > > Hello, > Has anyone ported TAR & GZIP to VxWorks ? > > Example: gzipPack (int in_fd, int out_fd) > reads from "in_fd", packs and writes into "out_fd" > > gzipUnpack(int in_fd, int out_fd) > ditto, only unpacks > > > Thanks a lot, > > ***************************************************************** > * Lyor Goldstein. Email : lyorg@scorpio.com * > * Scorpio Communications Ltd. Phone : 972-3-5339654 ext 115 * > * Software group Fax : 972-3-5339518 * > * 1c Yoni Netanyahu st. * > * Or Yehuda 60250 * > * P.O.Box 564, Israel * > ***************************************************************** > Hi, I ported gzip to vxWorks with little problem. I wrote a wrapper that created command line like arguments (argc & **argv) and called the gzip main program. The other thing I had to do was write a routine that re-initialized all global variables to their defaults before calling the gzip main routine. Good Luck, -- Bob Kressig Msgid: BOBK email: kressig@ti.com Texas Instruments voice: 972-952-3489 fax: 972-952-3435 From luettgen@nis.lanl.gov Tue Oct 22 12:58:51 1996 From: Allen Luettgen Date: Tue Oct 22 12:58:53 PDT 1996 Subject: unsubscribe please unsubscribe me. Thank--Allen From dmidkiff@trandes.com Tue Oct 22 14:15:03 1996 From: "DENNIS MIDKIFF" Date: Tue Oct 22 14:15:05 PDT 1996 Subject: Tornado NE2000 Ethernet Card Configuration Hello Vxers, I have a rack mount 486 pc with a NE2000 PCI Ethernet Card and I want to know if the driver is included in the Tornado drivers. If not how do I include the driver that comes with the NIC card into the BSP before I build the new image. Thanks in advance. Dennis From bobh@mbari.org Tue Oct 22 16:28:14 1996 From: Bob Herlien Date: Tue Oct 22 16:28:16 PDT 1996 Subject: DTTR regs on 68040 How do you use the DTTR{0,1} registers on the 68040 to map in physical memory without using the page table mechanism of sysPhysMemDesc? CPU: MVME-162 (68040) VxWorks 5.1.1 We have a board with 128 MB of address space that we're currently mapping in with sysPhysMemDesc, but the creation of all those page tables takes longer than we can afford. I know that it's possible to use one of the TTR registers instead, but I'm not sure where to do this. My first stab was to put it in sysALib.s, since it needs a line of assembly code for setting DTTR. Lo and behold, sysALib for the '162 *already* sets DTTR0 and 1, using them to map in the *entire* address space. What's that about? So, I then figured I could just delete the sysPhysMemDesc entry, but nope, we get bus errors when accessing the memory. OK, so somebody's messing with DTTR. After a little snooping, I think the DTTRs are reset by either cacheEnable() or possibly by one of the functions called by usrMmuInit. I figure I could just add "cacheDTTR0Set(0xnnnn)" somewhere in usrRoot() after usrMmuInit, but since I don't really understand who's using the DTTRs and why, I thought I'd ask here. cachDTTR0Set(), by the way, is an undocumented function I found while snooping around the rest of this stuff. So, could somebody who's done this before tell me if I'm on the right track? Where exactly should I add a line of code to set DTTR{0,1}? Which DTTR register should I use? Or does the system manage these registers, and I'm out of luck? a T d H v A a N n K c S e -------------------------------------------------------------- Bob Herlien MBARI (Monterey Bay Aquarium Research Institute) bobh@mbari.org From froeber@bbn.com Tue Oct 22 18:55:40 1996 From: Fred Roeber Date: Tue Oct 22 18:55:42 PDT 1996 Subject: Re: DTTR regs on 68040 On Tue, 22 Oct 1996, Bob Herlien asked: > How do you use the DTTR{0,1} registers on the 68040 to map in physical > memory without using the page table mechanism of sysPhysMemDesc? > > CPU: MVME-162 (68040) > VxWorks 5.1.1 We use these registers in place of MMU under VxWorks 5.2 on MVME162 to not only speed up initialization time but to also speed up execution since address translation using the transparent translation registers is faster than that done using the MMU. It also yields more deterministic operation and avoids page table overhead. I'll indicate the hooks we have used below. > My first stab was to put it in sysALib.s, since it needs a line of > assembly code for setting DTTR. Lo and behold, sysALib for the '162 > *already* sets DTTR0 and 1, using them to map in the *entire* address > space. What's that about? sysALib code sets up the TTR registers so that the MMU isn't needed during the initial boot process. The code there explains how the registers are set up so that the first 16MB is cached (which covers local RAM) and everything else isn't. > So, I then figured I could just delete the sysPhysMemDesc entry, but > nope, we get bus errors when accessing the memory. OK, so somebody's > messing with DTTR. After a little snooping, I think the DTTRs are > reset by either cacheEnable() or possibly by one of the functions > called by usrMmuInit. I found that they are reset by the mmuEnable function to do nothing (since the MMU is being set up to handle cache control). > I figure I could just add "cacheDTTR0Set(0xnnnn)" somewhere in usrRoot() > after usrMmuInit, but since I don't really understand who's using the > DTTRs and why, I thought I'd ask here. cachDTTR0Set(), by the way, > is an undocumented function I found while snooping around the rest of > this stuff. What we do is to replace the call to usrMmuInit in usrConfig.c with a call to our own sysMmuInit function (we don't want to bother using the normal MMU setup at all). For the MVME162, the code for sysMmuInit is as follows: /* Set up to allow efficient multiprocessor access with data cache enabled * on all processors. We use the MMU on the backplane master processor * to allow him to selectively make parts of his memory uncached using * cacheDmaAlloc. This is used by the shared memory driver to make his * dual ported memory writeable by other processors. Most of the rest of * the translations are handled by the transparent translation registers. * On slave processors, the MMU isn't even turned on. */ extern STATUS usrBootLineCrack (char * bootString, BOOT_PARAMS *pParams); extern void usrBootLineInit (int startType); extern void cacheDTTR0Set(UINT); extern void cacheDTTR1Set(UINT); extern void usrMmuInit(void); extern int sysStartType; /* The sysMmuInit function is called before terminal IO is set up so we * can't display error messages directly. Instead, we set bits in the * variable below which we later decode to display status messages. */ int BbnMmuSetupState; void sysMmuInit(void) { BOOT_PARAMS params; unsigned memsize; /* local memory size in bytes */ BbnMmuSetupState = 0; /* crack the boot line here to get access to processor # */ usrBootLineInit (sysStartType); if (usrBootLineCrack ((char *) BOOT_LINE_ADRS, ¶ms) != OK) { /* indicate error cracking boot line */ BbnMmuSetupState |= 1; } if (params.procNum == 0) { /* master */ /* flag that we set up for the master processor */ BbnMmuSetupState |= 2; /* set up MMU. We basically, only have it mapping local memory and * the IO pages but none of VME memory. We don't want to use the * MMU to control all of VME since that results in huge page tables * and slow initialization of the MMU (can take minutes). */ usrMmuInit(); /* use the transparent translation register 0 and 1 to override the * MMU and make some off card memory uncached. Can't make it all * uncached due to the way the transparent translation registers * work. Set the registers so that the upper half of memory and * the range 0x20000000 - 0x3fffffff and 0x60000000 - 0x7fffffff * are uncached and serialized. Have to still use sysPhysMemDesc * to get access to areas below 0x20000000 or between 0x40000000 * and 0x5fffffff. */ cacheDTTR0Set(0x205FC040); cacheDTTR1Set(0x807FC040); } else { /* slave processor */ /* flag that we set up for the slave processor */ BbnMmuSetupState |= 4; /* handle cache control in slave with the transparent translation * registers only. We don't need to set up the MMU since these * processors don't dual port any memory. We don't want the overhead * of the MMU. This setup mimics that done in sysALib.s. We set * up DTT0 to make on card memory cached. */ memsize = sysPhysMemDesc[0].len; memsize >>= 8; memsize &= 0x70000; if (memsize != 0) memsize -= 0x10000; cacheDTTR0Set(memsize | 0x0000C000); /* set DTT1 to make all of memory uncached and serialized. Local * memory will remain cached since DTT0 takes precedence over DTT1. */ cacheDTTR1Set(0x00FFC040); } } Note that this function is called before the IO system is set up so can't use printf to display messages. In our production code we check the BbnMmuSetupState flags and print the correct messages after the IO system is set up (changing the MMU initialization to after the IO initialization does not work). The code above needs the following routine added to sysALib.s: /*********************************************************************** * * cacheDTTR1Set - write data transparent translation register 1 * * cacheLib.s already provides cacheDTTR0Set * * void cacheDTTR1Set (UINT registerVal) */ .globl _cacheDTTR1Set _cacheDTTR1Set: movel sp@(4),d0 movec d0,dtt1 .word 0xf518 /* pflusha */ rts Note, that the way this code sets up address translation for the backplane master processor is a little strange since it combines the MMU and the transparent translation registers. This is required so that parts of on-card memory can be set up as non-cached as required for proper operation of the backplane communication driver or the VxMP package. To get the real speedup you want from not using the MMU at all, just disable the special check for the master processor and always use the slave setup code. Of course, in this case, don't try to share the CPU's local memory across the VME bus. Hope this helps. Good luck. Fred | Fred J Roeber, BBN Systems & Technologies | | 50 Enterprise Place Middletown, RI 02842-5202 | | froeber@bbn.com 401-849-2543 (X48) | | See http://www.bbn.com/tv for TraceMaker info | From daemon@csg.lbl.gov Wed Oct 23 04:00:38 1996 From: daemon@csg.lbl.gov Date: Wed Oct 23 04:04:21 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Wed Oct 23 04:00:36 PDT 1996 Subject: Re: tcp/ip performance survey Subject: Re: VME card operation Subject: PPC and Green Hills Subject: Re: C/C++ on 5.2 Subject: Problem with "bcopy" Subject: Re: TAR & GZIP porting Subject: Integrating IrDA driver with VxWorks ... Subject: Interphase 4524 Subject: Re: IOSTREAMS Subject: GNU tar ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: tcp/ip performance survey Date: Tue, 22 Oct 1996 10:37:59 +0200 From: michaelv@qualcomm.com (Michael Vakulenko) Organization: QUALCOMM Israel Message-ID: References: <199610102316.TAA26967@alioth.CCCHP> <53n9jr$435@idiom.com> > From: Paul Banta > Date: Thu Oct 10 16:17:55 PDT 1996 > Subject: tcp/ip performance survey > I'd like to take an quick survey of experience with tcp/ip in high bandwidth > applications. Particularly where VxWorks targets are sending as much data > as they can muster through a network interface. > > What is the average output bandwidth? > What is the peak output bandwidth? > Does the system perform well over a long period of time? > Are there any special tcp parameter modifications (e.g. adjusting the send and > receive sliding window sizes)? > > > I'll collect the results and distribute them by request. > > My curiosity about this stems from my current problem: > I have a Radstone PPC604 (100MHz PowerPC) with an ATM adapter on it. The > driver is home grown (with some help from various people around the net). > The other end of the fiber is plugged into a sun ultra 1. The unix system > has a process that eats data from a socket as fast as it can. The vxworks > target crams data through a socket as fast as it can. > > The system runs steady at around 4.9 MBytes/s. It will occasionally peak > at just over 5MBytes/s (the driver has yet to be tweaked so I'm hoping that > this number will go up. I still have CPU left at this rate). But, it > doesn't run steady for long. After 30-60 seconds it begins to slowly > degrade and the data rate ultimately drops dramatically. I'm guessing that > I'm stuck in some funky tcp state which, of course, could be happening due > to a driver bug, but I'd like to hear about other experiences. > > Thanks, > > -- > Paul Banta voice: 757-269-5829 > TJNAF, MS 12H fax: 757-269-5800 > 12000 Jefferson Avenue email: banta@cebaf.gov > Newport News, VA 23606 Thomas Jefferson National Accelerator Facility > As far as I understand from the original message, Paul's application uses TCP stream sockets. The following is right for TCP sockets only. In order to improve TCP performance, you can increase TCP window sizes by setting (use setsockopt) Rx/Tx buffers to max values. Note, that this should be done at both sides of the TCP connection, target and Unix. The default TCP window size is 16K for most TCP implementations. The TCP window size is important, because TCP cannot send more than one window of data per round trip delay. If the TCP window is less than the channel's delay*bandwidth product, the channel will not be fully utilized. ( This phenomena is known as "long fat pipes"). Since you are using high bandwidth channel, it is probably that 16K default window size is not enough for full channel utilization. The second thing is that TCP shrinks its Tx window when segment loss occures. TCP assumes that the reason for segment loss is network congestion and immediately closes the Tx window, thus decreasing the number of segments being sent to the network. Then TCP gradualy opens the Tx window. If you have some problems in the driver or buffer overflow, causing IP packet loss at either side of the connection, then most probably TCP will never have a chance to open it's window to max size. Use ipstatShow or your driver statistics to check if there any packet loss in the IP layer at the target side of the connection. Third, increase all the numbers related to mbufs allocation. The default values (at least for vxworks v5.2) are not good enough for steady state performance. It is difficult to recommend any specific values, the best way is to experiment with the numbers and using vxworks mbufShow routine. VxWorks manuals has pretty good explanation of mbuf stuff. Fourth, try to play with tNettask and the Rx/Tx tasks priorities, setting the Rx/Tx tasks priority above and below the tNettask. BTW, same things can happen at the Unix side, as well. In general, you'll get more accurate measures of VxWorks TCP stack performance opening TCP connection between two vxworks targets of the same type. In order to test raw stack and driver performance, I'd recommend to use UDP, which is much simpler for analysis. Cheers, ___________________________________________________________ Michael Vakulenko, : Bytes: michaelv@qualcomm.com Senior Engineer, : Voice: +972-4-857-7999 Qualcomm Israel, Ltd. : Paper: +972-4-857-7998 ___________________________________________________________ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VME card operation Date: Tue, 22 Oct 1996 12:55:57 -0400 From: CK Wong Organization: Nortel Technology Message-ID: <326CD74A.6E82@nortel.ca> References: <2.2.32.19961018135557.00759e4c@mailhost.mclean.sparta.com> Thank you very much for so many gentlepeople provide me information on the VME interrupt function. As a following up question, could someone enlight me on 1. how information can be exchanged using the mailbox ? Is there a size limitation ? 2. How inter-processor communication is done ? Is there something like shared memory can be used ? If so is the shared memory a mapped IO address space that you can write to it and data can be read from other side ? Many thanks. --------------------------- Newsgroups: comp.os.vxworks Subject: PPC and Green Hills Date: 18 Oct 1996 18:29:34 -0700 From: mwette@mr-ed.jpl.nasa.gov (Matt Wette) Organization: Jet Propulsion Laboratory Message-ID: <549atu$noj@mr-ed.jpl.nasa.gov> I remember a while back someone was having problems compiling utilities for VxWorks PPC target using the Green Hills compiler. I now have gotten back to this. The magic you want is ccvxppc ... -Xppcfarcalls ... - -- matthew.r.wette@jpl.nasa.gov --------------------------- Newsgroups: comp.os.vxworks Subject: Re: C/C++ on 5.2 Date: Fri, 18 Oct 1996 09:19:27 -0200 From: Bob Strickland Organization: Lockheed Martin Missiles & Space Message-ID: <326767BF.5535@lmsc.lockheed.com> References: <541c35$7g9@news.aimnet.com> Reply-To: bstrickland@lmsc.lockheed.com We have a similar (nothing is ever the same) setup. We have the Heurikon kkv3500 MIPS processor, a mix of C/C++, and use wrs 5.2, but we don't use the vxSim. Our C++ is the windRivers/Centerline configuration. The way we delt with the C/C++ mix was to treat all code as C++ and thus use the same makefile/compiler setup for all the code. This has worked quite well for us. Good luck Roy Gordon wrote: > > We want to develop a mixed C/C++ application on 5.2 using both > VxSim and then a MIPs BSP with a 4650 processor. Naturally, we want > source level debugging for the BSP. > > Has anyone ever done this successfully? If not for the 4650 then for the > 4600? If so, what products did you use? If you had problems with > the development tools chosen, what were they? > > Other success/failures with a C/C++ mix on 5.2? > > TIA (very much). > > -- roy --------------------------- Newsgroups: comp.os.vxworks Subject: Problem with "bcopy" Date: Tue, 15 Oct 1996 16:24:38 -0700 From: Michael Levasseur Organization: GDE Systems Inc. Message-ID: <32641D36.41C67EA6@gdesystems.com> Hello, I'm using VxWorks 5.2 on a Heurikon Baja R4700 board and I'm having problems (difficulties) with the "bcopy" routine. Does anyone have the source for "bcopy"? The problems I'm having are: I'm writting across the VME64 Bus to an Extended slave address that is on a custom board that we have. When I write a word (a 16-bit object), "bcopy" is writting two bytes and not the one 16-bit word I'd expect to see written. The man pages say that "bcopy" uses the MOST efficient method! Two byte writes aren't the most efficient! This is just an inconvienience and I can live with this. The other problem is when I write an object that is a longword (a 32-bit object), "bcopy" is writting the same longword twice across the VME Bus. The same address and data are sent over the VME bus two times. Unfortunately the device that we are writting from is destructive. That is after a write from the address, the contents are destroyed. Therefore writting twice causes a FAILURE. This is also a concern because we are using the shared memory option and TCP/IP to pass messages around and I'm concerned that every message will be coppied twice because the VxWorks socket implementation utilizes "bcopy". Another problem I've seen is with VME BERR problems. Bus Errors that are raised when bcopy is used to read an invalid VME slave address cause a DATA_ERROR and cause the application to die. For a write a Bus Error is raised and the VIC64 on the Baja board can be seen without to examine to determine the status of the write operation, but the read operation is unrecoverable. Has anyone else had any problems, fixes, or updates to the VxWorks 5.2 version of bcopy for the Mips architecture? thanx for any information. p.s. We've mailed Wind River's most of this information but haven't gotten a reply yet. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: TAR & GZIP porting Date: Tue, 22 Oct 1996 13:08:45 -0800 From: russell.brill@jpl.nasa.gov (Russ Brill) Organization: Jet Propulsion Laboratory Message-ID: References: <9610221317.AA26576@Scorpio.Com> In article <9610221317.AA26576@Scorpio.Com>, lyorg@Scorpio.Com (Lyor Goldstein) wrote: > Hello, > Has anyone ported TAR & GZIP to VxWorks ? > > Example: gzipPack (int in_fd, int out_fd) > reads from "in_fd", packs and writes into "out_fd" > > gzipUnpack(int in_fd, int out_fd) > ditto, only unpacks > > I have been pondering this question (regarding gzip) myself. To rephrase the question, how can I change the program gzip into a function. In general, for any UNIX program, is there any reason you can't just edit the source, and rename main()? --------------------------- Newsgroups: comp.os.vxworks Subject: Integrating IrDA driver with VxWorks ... Date: 22 Oct 1996 21:39:39 GMT From: "John Sambrook" Organization: Common Sense Systems, Inc. Message-ID: <01bbc061$f3f4fba0$d6f176cc@johns.oz.net> Hello, I am working on a project that will create a handheld device that has to support IrDA v1.1. We are planning to use VxWorks, and I am wondering if anyone in this group has any experience with respect to developing this kind of support under VxWorks. We are planning on buying the IrDA driver. I know of one company, ACTiSYS, which makes such a beast. If you know of others, or have other experience with ACTiSYS, I would appreciate any information you would be willing to share. Also, descriptions of any specific strategies that you employed (to good or bad effect) would be appreciated. Finally, if you can contribute significant IrDA experience and would be interested in a subcontracting relationship to do some or all of this work, please let me know. I would expect you to have IrDA experience (not just driver work) and be able to put together a defensible business case for subcontracting the work. Thanks, John Sambrook Common Sense Systems, Inc. 206.806.8744 --------------------------- Newsgroups: comp.os.vxworks Subject: Interphase 4524 Date: Wed, 09 Oct 1996 08:18:33 +0100 From: Paul Hatcher Organization: Hepworth & Hatcher Message-ID: <325B51C9.1B4@dial.pipex.com> Is there a VxWorks driver for the Interphase 4524 network card? If not, then are there VxWorks drivers avaiable for the TI ThunderLan chip set? Thanks, P.Hatcher hephat@dial.pipex.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: IOSTREAMS Date: Tue, 22 Oct 1996 19:01:52 -0700 From: Todd Hoff Organization: Possibility Outpost Message-ID: <326D7C90.31DA@possibility.com> References: <199610210319.LAA10282@synaps.nixltd.com.au> <54gssv$19g@soap.news.pipex.net> Reply-To: tmh@possibility.com John Gordon wrote: >As a result of this you need to buy an > additional package containing a set of class libraries, one of which > is the iostreams library you *need* to do any real work (or for that > matter even hello world!) in C++! I understand what you are saying, yet some projects may find they can do without. Good old printf still works in case you need to tell the world hello. And the gnu class libraries are pretty minimal, certainly it contains nothing you cannot find on the net or make yourself. Iostreams are the possible exception. If you have a lot of memory iostreams are fine. If you work in a limitted memory environment you will not want to make much use of iostreams in any case. They bloat your code and are slow t'boot. Most of the features of iostreams are easily simulatable in your own custom iostreams type package. There is nothing special about '<<' and '>>'. You can use them too. - -------------------------------------------------------- tmh@possibility.com | The loyalty of small men can be http://www.possibility.com | bought cheaply, for greed has no | pride. - Michael Kube-McDowell --------------------------- Newsgroups: comp.os.vxworks Subject: GNU tar Date: Wed, 16 Oct 1996 14:27:04 -0400 From: "Vipul A. Shah" Organization: GE-Harris Railway Electronics Message-ID: <326528F8.E00@ge-harris.com> Has anyone ported these utility to VxWorks ? thanx - -- Vipul Shah voice : 407-242-5390 email : vshah@ge-harris.com --------------------------- End of New-News digest ********************** From leonid@rst.co.il Wed Oct 23 05:51:05 1996 From: leonid@rst.co.il (Leonid Rosenboim) Date: Wed Oct 23 05:51:08 PDT 1996 Subject: Tornado Symbol Table Synchronization This problem has been brought up previoysly - if you load objects in to VxWorks 5.3 using that old-fasioned target-resident loader and symbol table, the Tornado's target server symbol table does not get updated, hence the Tornado tools can not recignize these modules. An experimental solution that we have done, will synchronize the symbol tables each time the target loader is invoked to load an object. The code consists of a TCL script and a "C" module, and is availbale on our Web site under "Free Software". The module is free and unsupported, however you may send your comments to us. ----------------------------------------------------------------------- Leonid Rosenboim Phone: +972-3-559-8144 R S T Software Industries Ltd. Mobile: +972-50-307-142 P.O.Box 11502, AZUR 58017, Israel Fax: +972-3-559-8244 WWW: http://www.rst.co.il E-Mail: leonid@rst.co.il From yfan@SPAR.CA Wed Oct 23 07:41:42 1996 From: Yibing Fan Date: Wed Oct 23 07:41:44 PDT 1996 Subject: Inter-processor comunication (ipc) >CK Wong wrote: >2. How inter-processor communication is done ? Is there something > like shared memory can be used ? If so is the shared memory > a mapped IO address space that you can write to it and data > can be read from other side ? Yes. Shared memory is the most efficient way for ipc. Here is how you do it: 1. localAddr = smMemMalloc() to locate sm object. This function returns a local addr of obj. 2. smNameAdd(smName, ...) add sm obj to sm data base when you add, you need smObjLocalToGlobal(localAddr) to map local addr to global one. when you use it from other process 1. smNameFind(smName, (void **)&pGlb, ...) pGlb is the global addr of pointer to sm obj 2. localAddr = smObjGlobalToLocal(pGlb); now you have the local pointer to the sm obj. Do what ever you wnat eith the obj. You also need shared memory semiphore to protect the sm obj. Hope it helps Yibing Fan From han2326@aminet.co.kr Wed Oct 23 07:57:03 1996 From: "Han, Jeong Soo" Date: Wed Oct 23 07:57:04 PDT 1996 Subject: Q: memShow() Hi, My question is memShow() function(vxWorks 5.2). What are differences between current status(free & alloc) and cumulative status(alloc) in the memShow() results ? What is the relation with my target's max. memory? During the runnings of my application, cumulative(alloc bytes) is continuously increasing and over my target's max. memory, still with normal operation. Is it potential problem ? Thanks in advance, J.S. Han ------------------------- H D P I C Research Center From davida@laplace.idec.sdl.usu.edu Wed Oct 23 08:36:42 1996 From: davida@laplace.idec.sdl.usu.edu (David Anderson) Date: Wed Oct 23 08:36:45 PDT 1996 Subject: calling objects from the shell Is there anyone using VxWorks 5.3/Tornado that knows if and how you can call an objects method directly from a shell? We are currently using VxWorks 5.2/GNU C++ and the only way we can call an objects method from the shell is to write a task that calls the method and then to run the task from the shell. We would like to be able to call the method directly and not have to have a piece of code for every method whos sole purpose is to call the method. Thanks in advance, Dave. From pardo@rti.com Wed Oct 23 13:35:27 1996 From: Gerardo Pardo-Castellote Date: Wed Oct 23 13:35:29 PDT 1996 Subject: Re: Q: memShow() >> My question is memShow() function(vxWorks 5.2). >> >> What are differences between current status(free & alloc) and cumulative >> status(alloc) in the memShow() results ? >> What is the relation with my target's max. memory? The cummulative count is increased every time anybody allocates memory, it is not decremented when memory is freed so it doesn't reflect the curently allocated memory. Current count reflects the allocated memory (i.e. it is incremented on allocations, and decremented on frees). >> During the runnings of my application, cumulative(alloc bytes) is >> continuously increasing and over my target's max. memory, still with >> normal operation. >> Is it potential problem ? No, as long as the current count doesn't increase you don't have a leak... -Gerardo =========================================================================== = = = = Gerardo Pardo-Castellote = email: pardo@rti.com = = Real-Time Innovations, Inc. = Phone: (408) 720-8312 = = 155A Moffett Park Drive, Suite 111 = Fax: (408) 734-5009 = = Sunnyvale, CA 94089 = http://www.rti.com = = = = =========================================================================== From pardo@rti.com Wed Oct 23 13:40:03 1996 From: Gerardo Pardo-Castellote Date: Wed Oct 23 13:40:05 PDT 1996 Subject: Re: calling objects from the shell >> >> Is there anyone using VxWorks 5.3/Tornado that knows if and how you can call an >> objects method directly from a shell? We are currently using VxWorks 5.2/GNU C++ >> and the only way we can call an objects method from the shell is to write a task >> that calls the method and then to run the task from the shell. We would like to >> be able to call the method directly and not have to have a piece of code for >> every method whos sole purpose is to call the method. You can certainly call the methods from the windsh. The only problem here may lie in the synchronization of symbol tables between the target and host. If you load objects through windsh you can call those functions from windsh, if you load them to the target directly (e.g. using rlogin, assuming that you have a target-based shell, symbol table, and loader), you will be able to call them from the target-shell. -Gerardo =========================================================================== = = = = Gerardo Pardo-Castellote = email: pardo@rti.com = = Real-Time Innovations, Inc. = Phone: (408) 720-8312 = = 155A Moffett Park Drive, Suite 111 = Fax: (408) 734-5009 = = Sunnyvale, CA 94089 = http://www.rti.com = = = = =========================================================================== From cchen@yurie.com Wed Oct 23 17:29:26 1996 From: cchen@yurie.com Date: Wed Oct 23 17:29:29 PDT 1996 Subject: ZIP utility on VxWorks Hi, I heard from a friend that someone ported "zip" utility functions to VXworks. Unfortunately, my friend delete that messqge and can't give me the details. Can anyone has that information give me a cpoy? Thanks!! Chin J. Chen, Yurie Systems Inc. From leonid@rst.co.il Thu Oct 24 01:23:35 1996 From: leonid@rst.co.il (Leonid Rosenboim) Date: Thu Oct 24 01:23:39 PDT 1996 Subject: PowerPC Assembly Mnemonics One of our new customers have started to use Tornado for PowerPC, and since we eint got much experience with PowerPC yet, I would like to get some assistance from more experienced VxWorks/PowerPC users. The Assembly mnemonics that are generated by the GNU compiler, and displayed by GDB disassembly, do not correlate well with the assembly mnemonics which are published in the PowerPC processor manual. We could not find any documentation on this subject, and we would prefer not to deal with the GNU sources for this. Does anyone have a pointer to some documentation explaining the relations between the GNU "as" mnemonics for PowerPC instructions and the ones documented by the processor manufacturer ? Thanks ! ----------------------------------------------------------------------- Leonid Rosenboim Phone: +972-3-559-8144 R S T Software Industries Ltd. Mobile: +972-50-307-142 P.O.Box 11502, AZUR 58017, Israel Fax: +972-3-559-8244 WWW: http://www.rst.co.il E-Mail: leonid@rst.co.il From daemon@csg.lbl.gov Thu Oct 24 04:00:24 1996 From: daemon@csg.lbl.gov Date: Thu Oct 24 04:00:29 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Thu Oct 24 04:00:21 PDT 1996 Subject: GZIP & TAR porting Subject: Re: SLIP interface between Windows95 and an MVME162 Subject: Any DejaGNU / VxWorks users out there??? Subject: Any DejaGNU / VxWorks users out there Subject: DejaGNU Subject: sad Subject: problem binding socket to target's inet address ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: GZIP & TAR porting Date: Wed, 23 Oct 1996 07:35:46 GMT From: lyorg@Scorpio.Com (Lyor Goldstein) Organization: Scorpio Communications Ltd. Message-ID: Reply-To: lyorg@Scorpio.Com Sender: news@Scorpio.Com - --- Has anyone ported GZIP & TAR to VxWorks ? I would appreciate such source code, and would be willing to exchange other useful code I wrote... ***************************************************************** * Lyor Goldstein. Email : lyorg@scorpio.com * * Scorpio Communications Ltd. Phone : 972-3-5339654 ext 115 * * Software group Fax : 972-3-5339518 * * 1c Yoni Netanyahu st. * * Or Yehuda 60250 * * P.O.Box 564, Israel * ***************************************************************** --------------------------- Newsgroups: comp.os.vxworks Subject: Re: SLIP interface between Windows95 and an MVME162 Date: 23 Oct 1996 08:59:10 -0600 From: bgeer@xmission.xmission.com (bgeer) Organization: XMission Internet (801 539 0900) Message-ID: <54lbru$8k@xmission.xmission.com> References: <1366140033-19540572@fly.aeromet.com> "Jon Jurek" writes: >There have been a number of messages dealing with TCP over a serial link >over >the last few weeks, but I have not seen any that helps with my problem. >I am using a MVME162 with Tornado on a Windows95 machine. I am trying to >implement a SLIP protocol to boot VxWorks and have not had any success so >far. >In particular I would like information on setting up the host side. Any >clues >on software to use and how to set it up would be appreciated. Hmmmm... Does this require a tftp daemon on the Win95 host? Also a SLIP daemon? Does Tornado supply these? I'm a hardcore Unix type & hate the thought of having to support VxWorks from Win95 (or any other "bondage & discipline pseudo- operating system from the Evil Empire :-) However, I'd like to know the answers to these questions... - -- <> Robert Geer & Donna Tomky / * <> <> bgeer@xmission.com | _o * o * o <> <> Salt Lake City, Utah | -\<, * <\ <> U S A | O/ O __ /__, /> <> --------------------------- Newsgroups: comp.os.vxworks,comp.software.testing Subject: Any DejaGNU / VxWorks users out there??? Date: Wed, 23 Oct 1996 17:11:14 GMT From: gbruce@ti.com (Gene Bruce) Organization: Texas Instruments Message-ID: <54ljhj$9ht$1@superb.csc.ti.com> I'd like to here from anyone using dejaGNU to test Vxworks applications. Gene Bruce Texas Instruments I want to die like Grandpa - Systems Group in my sleep. Electronic Systems Division _/_/_/_/_/ _/_/_/ Ocean Surveillance Radar Not like his passengers - _/ _/ 2501 W. University, MS 8056 screaming in the car. _/ _/ McKinney, Texas 75070 _/ _/ Phone/Voice Mail: (972)952-4937 _/ _/_/_/ FAX: (972)952-4275 All expressions of opinion are strictly my own and do not reflect those of my employer. --------------------------- Newsgroups: comp.os.vxworks Subject: Any DejaGNU / VxWorks users out there Date: Wed, 23 Oct 1996 17:11:15 GMT From: gbruce@ti.com (Gene Bruce) Organization: Texas Instruments Message-ID: <54ljhj$9ht$2@superb.csc.ti.com> I'd like to hear from anyone using DejaGNU to test VxWorks applications. Gene Bruce Texas Instruments I want to die like Grandpa - Systems Group in my sleep. Electronic Systems Division _/_/_/_/_/ _/_/_/ Ocean Surveillance Radar Not like his passengers - _/ _/ 2501 W. University, MS 8056 screaming in the car. _/ _/ McKinney, Texas 75070 _/ _/ Phone/Voice Mail: (972)952-4937 _/ _/_/_/ FAX: (972)952-4275 All expressions of opinion are strictly my own and do not reflect those of my employer. --------------------------- Newsgroups: comp.os.vxworks,comp.software.testing Subject: DejaGNU Date: Wed, 23 Oct 1996 17:11:15 GMT From: gbruce@ti.com (Gene Bruce) Organization: Texas Instruments Message-ID: <54ljhj$9ht$3@superb.csc.ti.com> Is there anyone out there with DejaGNU experience? I have just discovered this tool and it seems that the info that comes in the distribution is for the experienced user only - a reference doc without any tutorial info. Is there a source of tutorials? Where? This tool seems like it might be for me but I need a way to evaluate it without spending a month to do so. Gene Bruce Texas Instruments I want to die like Grandpa - Systems Group in my sleep. Electronic Systems Division _/_/_/_/_/ _/_/_/ Ocean Surveillance Radar Not like his passengers - _/ _/ 2501 W. University, MS 8056 screaming in the car. _/ _/ McKinney, Texas 75070 _/ _/ Phone/Voice Mail: (972)952-4937 _/ _/_/_/ FAX: (972)952-4275 All expressions of opinion are strictly my own and do not reflect those of my employer. --------------------------- Newsgroups: comp.os.vxworks Subject: sad Date: 17 Oct 1996 17:49:15 GMT From: srejto@ll.mit.edu (Stephen Rejto) Organization: MIT Lincoln Lab Message-ID: <545rir$r5n@llnews.ll.mit.edu> asd --------------------------- Newsgroups: comp.os.vxworks Subject: problem binding socket to target's inet address Date: 10 Oct 1996 21:07:15 GMT From: mkamp@ix.netcom.com(Marcela A Kamp ) Organization: Netcom Message-ID: <53joi3$5j6@sjx-ixn3.ix.netcom.com> I am having a problem binding a UDP socket to my target system's IP address. It works if I specify INADDR_ANY as the address. The error returned is 0x31 which is something like "ENETADDRNOTAVAIL" I am getting the host IP address from hostGetAddr {?} which I can verify is the correct address, and have tried both with and without the htonl() call. This is very similar to an example in the programmers guide, but it doesn't work! other info: vxWorks version 5.3 target 486 target IPaddr 192.9.200.3 if flags ? Sorry -- I don't have access to the system/data right now I will send more data if requested. IMPORTANT: Please send response to tomf@acetsw.amat.com Thanks, T.F --------------------------- End of New-News digest ********************** From intrepid!minidds@uunet.uu.net Thu Oct 24 06:44:18 1996 From: intrepid!minidds@uunet.uu.net (Lee Kappel) Date: Thu Oct 24 06:45:48 PDT 1996 Subject: Re: PowerPC Assembly Mnemonics Leonid Rosenboim writes: > One of our new customers have started to use Tornado for PowerPC, and > since we eint got much experience with PowerPC yet, I would like > to get some assistance from more experienced VxWorks/PowerPC users. > > The Assembly mnemonics that are generated by the GNU compiler, and displayed > by GDB disassembly, do not correlate well with the assembly mnemonics > which are published in the PowerPC processor manual. We could not > find any documentation on this subject, and we would prefer not to > deal with the GNU sources for this. > > Does anyone have a pointer to some documentation explaining the relations > between the GNU "as" mnemonics for PowerPC instructions and the ones > documented by the processor manufacturer ? > > Thanks ! > ----------------------------------------------------------------------- > Leonid Rosenboim Phone: +972-3-559-8144 > R S T Software Industries Ltd. Mobile: +972-50-307-142 > P.O.Box 11502, AZUR 58017, Israel Fax: +972-3-559-8244 > WWW: http://www.rst.co.il E-Mail: leonid@rst.co.il > We had similar problem with mnemonics generated by our Ada compiler. Have you checked the "Simplified Mnemonics" for the PPC documented in Appendix F of the "PowerPC Microprocessor Family: The Programming Environments" published by Motorola? >>>>>>>>>>>>>>>>>>>~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ >>>> Lee Kappel >>>~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ >>>>>>>>>>>>>>>>>>>~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~ Raytheon E-Systems Communications Division ~ ~ P.O. Box 12248, M.S. 61 ~ ~ St. Petersburg, FL 33733-2248 ~ ~ Phone: 813-381-2000 ext. 7209 ~ ~ Fax: 813-347-2890 ~ ~ E-Mail: kappel@eci-cec.com ~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ From MikeMcQuade@connect.com Thu Oct 24 07:30:49 1996 From: Mike McQuade Date: Thu Oct 24 07:32:19 PDT 1996 Subject: Re: PowerPC Assembly Mnemonics At 01:23 AM 10/24/96 PDT, you wrote: >Submitted-by leonid@rst.co.il Thu Oct 24 01:23:35 1996 >Submitted-by: leonid@rst.co.il (Leonid Rosenboim) > >The Assembly mnemonics that are generated by the GNU compiler, and displayed >by GDB disassembly, do not correlate well with the assembly mnemonics >which are published in the PowerPC processor manual. We could not >find any documentation on this subject, and we would prefer not to >deal with the GNU sources for this. > GNU uses what is called Extended Instruction Forms. I reccomend the book "Optimizing PowerPC Code" by Gary Kacmarcik, ISBN # 0-201-40839-2 Addison-Wesley. In this book it says: 3.3 Extended Instruction Forms "Extended instruction forms are extra mnemonics that the assembler accepts and converts into valid instructions" ... "for example, Rotate Left Word Immediate then AND with Mask (rlwinm) becomes something more understandable (for example, Shift Left Immediate)" another example of an extended mnemonic is lis (load immediate shifted) this is an extended form of addis. I hope this helps. MikeMcQuade@connect.com From mea@mclean.sparta.com Thu Oct 24 07:46:59 1996 From: Mike Anderson Date: Thu Oct 24 07:48:29 PDT 1996 Subject: Re: Slip interface between Win95 and MVME162 On Thu, 24 Oct 1996, the vxWorks Users Group Exploder wrote: > "Jon Jurek" writes: > > >There have been a number of messages dealing with TCP over a serial link > >over > >the last few weeks, but I have not seen any that helps with my problem. > >I am using a MVME162 with Tornado on a Windows95 machine. I am trying to > >implement a SLIP protocol to boot VxWorks and have not had any success so > >far. > >In particular I would like information on setting up the host side. Any > >clues > >on software to use and how to set it up would be appreciated. Well... First you'll have to get a *real* IP stack for Win95. The one from FTP Software (OnNet32 V2.0 or higher) does a pretty good job and it supplies an FTP server (as well as NFS server and client :-). Once you have that installed, you may have to modify your boot ROMs to automatically turn on the compression options for CSLIP (and definately boost the default 9600 Baud rate on the SLIP boot). As I recall, Win95 supports CSLIP and PPP. VxWorks supports SLIP and CSLIP (but only if you set the CSLIP flags in slipInit in the boot ROM code). Once you have the baud rates set straight and are sure you have CSLIP configured on both sides (try it out in VxWorks first, then try it from the bootroms), use netDevCreate to verify that you can indeed create an FTP device and load code across the link. BTW, don't forget to start the FTP server process on the Win95 side ;-). If that all works, then try it from the bootroms. In summary, break the problem into smaller steps. Try a hardwire between the two systems and use the VxWorks shell to do the setups on the vxWorks side by hand. Then, as you prove you can do it from the O/S level, then try to duplicate that setup from the ROMs. HTH, =============================================================================== __ Real-Time System Development, Integration, Training and Services //\\ // \\ Mike Anderson // /\ \\ Chief Engineer Voice : (703) 448-0210 ext. 235 // / \ \\ SPARTA, Inc. FAX : (703) 734-3323 // \ \\ 7926 Jones Branch Drive EMAIL : mea@mclean.sparta.com \\ \ // Suite 900 Web : http://www.mclean.sparta.com \\ \ / // McLean, VA 22102 \\ \/ // "Software development is like making \\ // a baby... You can't make a baby in one \\// month by impregnating nine women. -- "Pride in Performance" Some things just take time." =============================================================================== From mea@mclean.sparta.com Thu Oct 24 08:22:23 1996 From: Mike Anderson Date: Thu Oct 24 08:22:26 PDT 1996 Subject: Re: Slip interface between Win95 and MVME162 On Thu, 24 Oct 1996, the vxWorks Users Group Exploder wrote: > "Jon Jurek" writes: > > >There have been a number of messages dealing with TCP over a serial link > >over > >the last few weeks, but I have not seen any that helps with my problem. > >I am using a MVME162 with Tornado on a Windows95 machine. I am trying to > >implement a SLIP protocol to boot VxWorks and have not had any success so > >far. > >In particular I would like information on setting up the host side. Any > >clues > >on software to use and how to set it up would be appreciated. Well... First you'll have to get a *real* IP stack for Win95. The one from FTP Software (OnNet32 V2.0 or higher) does a pretty good job and it supplies an FTP server (as well as NFS server and client :-). Once you have that installed, you may have to modify your boot ROMs to automatically turn on the compression options for CSLIP (and definately boost the default 9600 Baud rate on the SLIP boot). As I recall, Win95 supports CSLIP and PPP. VxWorks supports SLIP and CSLIP (but only if you set the CSLIP flags in slipInit in the boot ROM code). Once you have the baud rates set straight and are sure you have CSLIP configured on both sides (try it out in VxWorks first, then try it from the bootroms), use netDevCreate to verify that you can indeed create an FTP device and load code across the link. BTW, don't forget to start the FTP server process on the Win95 side ;-). If that all works, then try it from the bootroms. In summary, break the problem into smaller steps. Try a hardwire between the two systems and use the VxWorks shell to do the setups on the vxWorks side by hand. Then, as you prove you can do it from the O/S level, then try to duplicate that setup from the ROMs. HTH, =============================================================================== __ Real-Time System Development, Integration, Training and Services //\\ // \\ Mike Anderson // /\ \\ Chief Engineer Voice : (703) 448-0210 ext. 235 // / \ \\ SPARTA, Inc. FAX : (703) 734-3323 // \ \\ 7926 Jones Branch Drive EMAIL : mea@mclean.sparta.com \\ \ // Suite 900 Web : http://www.mclean.sparta.com \\ \ / // McLean, VA 22102 \\ \/ // "Software development is like making \\ // a baby... You can't make a baby in one \\// month by impregnating nine women. -- "Pride in Performance" Some things just take time." =============================================================================== From rickc@protocol.com Thu Oct 24 10:07:45 1996 From: Rick Combest Date: Thu Oct 24 10:07:47 PDT 1996 Subject: Re[2]: calling objects from the shell >> >> Is there anyone using VxWorks 5.3/Tornado that knows if and how you can call an >> objects method directly from a shell? We are currently using VxWorks 5.2/GNU C++ >> and the only way we can call an objects method from the shell is to write a task >> that calls the method and then to run the task from the shell. We would like to >> be able to call the method directly and not have to have a piece of code for >> every method whos sole purpose is to call the method. You can call the methods from the shell. You must first execute a short program which loads the object references into variables (pointers or refernces). You can then call the methods by using the method name including the pointer or reference to the object as the first parameter. For example: given a class Object with methods int test1(int) and void test2(double), In short program: Object* foo; foo = new Object(); In shell: test1(foo,10) test2(foo,112.123) Hope this helps. Rick Combest Protocol systems From daemon@csg.lbl.gov Fri Oct 25 04:00:16 1996 From: daemon@csg.lbl.gov Date: Fri Oct 25 04:00:20 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Fri Oct 25 04:00:13 PDT 1996 Subject: Video driver on VxWorks Subject: Re: SLIP interface between Windows95 and an MVME162 Subject: Problems with CrossWind Subject: Re: SLIP interface between Windows95 and an MVME162 Subject: subscribe ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Video driver on VxWorks Date: Thu, 24 Oct 1996 16:51:23 +1000 From: Rami Smair Organization: Digital Equipment Corporation Message-ID: <326F11EB.47BD@ozy.dec.com> I'm likely to be writing video driver on VxWorks, I need any information/documentation on the subject as I know close to nothing about it. Any help would be appreciated. Rami - ------------------------------------------------------------------ .-_|\ Rami Smair Software Engineering Australia / * <- Software Engineer Digital Equipment Corporation \_.-._/ Email: smair@ozy.dec.com Research Park, Bond University v Fax: [61] 7 5575 0100 Gold Coast, 4229 QLD AUSTRALIA - ------------------------------------------------------------------ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: SLIP interface between Windows95 and an MVME162 Date: 24 Oct 1996 08:43:06 -0600 From: bgeer@xmission.xmission.com (bgeer) Organization: XMission Internet (801 539 0900) Message-ID: <54nv9q$k3o@xmission.xmission.com> References: <1366140033-19540572@fly.aeromet.com> <54lbru$8k@xmission.xmission.com> bgeer@xmission.xmission.com (bgeer) writes: >"Jon Jurek" writes: > >I am using a MVME162 with Tornado on a Windows95 machine. >I'm a hardcore Unix type & hate the thought of having to support >VxWorks from Win95 (or any other "bondage & discipline pseudo- >operating system from the Evil Empire :-) However, I'd like to know >the answers to these questions... Actually I solved my needs very simply by installing Linux, which of course comes with everything required except a cc68k...:-) - -- <> Robert Geer & Donna Tomky / * <> <> bgeer@xmission.com | _o * o * o <> <> Salt Lake City, Utah | -\<, * <\ <> U S A | O/ O __ /__, /> <> --------------------------- Newsgroups: comp.os.vxworks Subject: Problems with CrossWind Date: Thu, 24 Oct 1996 18:00:20 +0200 From: "Natalie Rogozovsky (SDH-S)" Organization: Telematics International, Inc. Message-ID: <326F9294.41C67EA6@ecitele.com> I'm working with vxworks5.3 with simulator and have a problems with CrossWind. For example: 1. I can't to go up to source level object representation when the program is running and stopped on the breakpoint. 2. I can't get the correct source when I perform the command "line function_name". I get the not relevant source part from the other file(in the same time I can get this source by the "line file_name: string" command. I think that the source of the problems can be the size of the object. Thank you, Natalya natalier@ecitele.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: SLIP interface between Windows95 and an MVME162 Date: Thu, 24 Oct 1996 10:48:35 +0100 From: "H.J. Bae" Organization: Peaceful Star Message-ID: <326F3B73.527@idiom.com> References: <1366140033-19540572@fly.aeromet.com> <54lbru$8k@xmission.xmission.com> <54nv9q$k3o@xmission.xmission.com> Reply-To: hjb@peacefulstar.oakland.ca.us > Actually I solved my needs very simply by installing Linux, which of > course comes with everything required except a cc68k...:-) It's easy to build GNU toolchain on Linux for m68k. If you're lazy you can just grab re-compiled binaries from ftp://ftp.pfnet.com --------------------------- Newsgroups: comp.os.vxworks Subject: subscribe Date: Fri, 25 Oct 1996 01:57:11 -0700 From: Ben Liu Organization: DCI HiNet Message-ID: <327080E7.2FEC@ms1.hinet.net> - -- Ben Liu --------------------------- End of New-News digest ********************** From han2326@aminet.co.kr Fri Oct 25 05:33:40 1996 From: "Han, Jeong Soo" Date: Fri Oct 25 05:33:42 PDT 1996 Subject: Q:Memory check Hi, Does anyone know What kind of functions dose the vxWorks have in order to check memory status(ROM, DRAM and SRAM)? I found 'moduleCheck()' and the system library 'sysMemProbe'. How these functions to check the memory status, and what are the results ? Any hints would be highly appreciated. From dgilleland@sntc.com Fri Oct 25 06:09:26 1996 From: dgilleland@sntc.com (Dan Gilleland) Date: Fri Oct 25 06:09:29 PDT 1996 Subject: I960 Stack Trace Back I'm trying to understand some VxWorks on I960 behavior. I want to capture an accurate stack history upon receiving a fault or failing a sanity check on an I960Hx target. From an exception handler, I can get the stack pointer from the current task control block (or the task referenced in the fault record). However, it looks to me like the stack frames do not have accurate register information (none of the parameters passed into the routines in the stack trace show up in either the global or local registers in the stack). I know I have valid stack pointers since the previous frame pointer and return instruction pointers are valid in each frame. Where are the function parameters? Thanks for any pointers, -dan --- Dan Gilleland SNT, Inc Email : dan@sntc.com Phone: (612) 404-0070 Ext 211 FAX : (612) 404-0075 From davida@laplace.idec.sdl.usu.edu Fri Oct 25 08:02:42 1996 From: davida@laplace.idec.sdl.usu.edu (David Anderson) Date: Fri Oct 25 08:04:10 PDT 1996 Subject: Re: Re[2]: calling objects from the shell Thanks for the help From davida@laplace.idec.sdl.usu.edu Fri Oct 25 08:06:54 1996 From: davida@laplace.idec.sdl.usu.edu (David Anderson) Date: Fri Oct 25 08:06:56 PDT 1996 Subject: Re: calling objects from the shell Thanks for the info. Dave. From aisie.aisinc.com!aisinc.com!pew@msen.com Fri Oct 25 09:55:28 1996 From: pew@aisinc.com (Paul E. Wilt) Date: Fri Oct 25 09:55:31 PDT 1996 Subject: Re: I960 Stack Trace Back Dan: > > > I'm trying to understand some VxWorks on I960 behavior. > > I want to capture an accurate stack history upon receiving > a fault or failing a sanity check on an I960Hx target. > > >From an exception handler, I can get the stack pointer from > the current task control block (or the task referenced in the > fault record). However, it looks to me like the stack frames > do not have accurate register information (none of the > parameters passed into the routines in the stack trace > show up in either the global or local registers in the stack). The i960 contains a local register set cache. Consequently, unless those local register sets cause a memory spill, a number of the internal register values will not be flushed to memory. The 'flushreg' instruction forces the CPU to put the stack contents to host memory. mon960 handles this in its fault handler by performing a flushreg and then copying the register contents into an internal register contents array. > > I know I have valid stack pointers since the previous frame > pointer and return instruction pointers are valid in each > frame. > > Where are the function parameters? The i960 C procedure call mechanism by default loads the function parameters into the global registers. Thus the first argument will be located in register g0, the second in g1, ... up to register g11. The global register g12 gets used for varargs types of functions. I am not certain how vxWorks handles stack switching when a task generates an exception. I think that when you execute "ti" on the specified task, the last local/global register values are displayed. Looking at the global registers included in that dump may give you some idea what happened. > > Thanks for any pointers, > -dan Hope this helps! > > --- > Dan Gilleland > SNT, Inc > Email : dan@sntc.com > Phone: (612) 404-0070 Ext 211 > FAX : (612) 404-0075 Thanks, Paul -- ============================================================================= Paul Wilt, Sr Software Engineer (313) 332-7044 FAX : (313) 332-7077 Applied Intelligent Systems, Inc. pew@aisinc.com 110 Parkland Plaza Ann Arbor, MI 48103 ============================================================================= From getker@sentientnet.com Fri Oct 25 09:59:39 1996 From: getker@sentientnet.com (Jim Getker) Date: Fri Oct 25 09:59:41 PDT 1996 Subject: RE:I960 Stack Trace Back Dan Gilleland wrote: >>From an exception handler, I can get the stack pointer from >>the current task control block (or the task referenced in the >>fault record). However, it looks to me like the stack frames >>do not have accurate register information (none of the >>parameters passed into the routines in the stack trace >>show up in either the global or local registers in the stack). >> >>I know I have valid stack pointers since the previous frame >>pointer and return instruction pointers are valid in each >>frame. >> >>Where are the function parameters? I don't know about the HX processor, but with an i960CF the register sets are cached in internal SRAM. Perhaps you need to execute a flushreg instruction before looking at the stack. Good Luck, Jim Jim Getker Sentient Networks, Inc getker@sentientnetcom From daemon@csg.lbl.gov Sat Oct 26 04:00:27 1996 From: daemon@csg.lbl.gov Date: Sat Oct 26 04:00:29 PDT 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sat Oct 26 04:00:23 PDT 1996 Subject: Re: Q: memShow() Subject: what can be called from signal handlers? Subject: Re: Inter-processor comunication (ipc) Subject: Looking for SMB server for VxWorks (x86) Subject: Re: AT&T Assembler1 ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Q: memShow() Date: 25 Oct 1996 03:56:04 GMT From: emshoff@omniport.net Organization: Deja News Usenet Posting Service Message-ID: <846213710.22423@dejanews.com> References: <3270648C.3000@aminet.co.kr> In article <3270648C.3000@aminet.co.kr>, "Han, Jeong Soo" wrote: > > Hi, > My question is memShow() function(vxWorks 5.2). > What are differences between current status(free & alloc) and cumulative > status(alloc) in the memShow() results ? Current Status indicates just what it implies - amount free and amount alloc'd. The total of the two should equal sysMemTop() minus the bottom of memory in the pool. The cumulative amount is a running total increased by every call to malloc(), but not decreased by calls to free(). The usefulness of the "cumulative" measure is somewhat limited. One thing to note about it however, is that if your application does a lot of malloc's and free's, memory can become fragmented. The level of fragmentation is indicated in the "max free block" column of memShow(). - ----------------------------------------------------------------------- This article was posted to Usenet via the Posting Service at Deja News: http://www.dejanews.com/ [Search, Post, and Read Usenet News!] --------------------------- Newsgroups: comp.os.vxworks Subject: what can be called from signal handlers? Date: 25 Oct 1996 15:39:40 GMT From: varlese@tech.ascom.ch (Chris Varlese) Organization: Ascom Hasler AG, Berne, Switzerland Message-ID: <54qmvt$6fp@ascomax.hasler.ascom.ch> Does someone have experience of what features can or cannot be called from inside a signal handler (using sigLib) on VxWorks? For example, for most UNIX systems, it is not safe to use malloc or free in signal handlers, because they are not re-entrant. We are using the timer library (timerLib), which uses signals to trigger the timeout event. Inside the signal handler I would like to delete the timer, but I believe that timer_delete() is calling free(). I am not sure that is safe from a signal handler? OK, on VxWorks you would say malloc and free are re-entrant. However, if they implement mutual exclusion between tasks with semaphores, this protection does not work when a task is interruptible with a signal. This we have experienced. Anyone travelled along this road? Chris Varlese --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Inter-processor comunication (ipc) Date: 25 Oct 1996 13:02:00 GMT From: tfm@sei.cmu.edu (Theodore Marz) Organization: The Software Engineering Institute Message-ID: <54qdo8$383@news.sei.cmu.edu> References: <96Oct23.104125edt.26519-1@janus.spar.ca> In article <96Oct23.104125edt.26519-1@janus.spar.ca>, Yibing Fan writes: |> |> >CK Wong wrote: |> |> >2. How inter-processor communication is done ? Is there something |> > like shared memory can be used ? If so is the shared memory |> > a mapped IO address space that you can write to it and data |> > can be read from other side ? |> |> Yes. Shared memory is the most efficient way for ipc. |> Here is how you do it: |> 1. localAddr = smMemMalloc() to locate sm object. |> This function returns a local addr of obj. |> 2. smNameAdd(smName, ...) add sm obj to sm data base |> when you add, you need smObjLocalToGlobal(localAddr) to map |> local addr to global one. |> |> when you use it from other process |> 1. smNameFind(smName, (void **)&pGlb, ...) |> pGlb is the global addr of pointer to sm obj |> 2. localAddr = smObjGlobalToLocal(pGlb); |> now you have the local pointer to the sm obj. |> Do what ever you wnat eith the obj. |> |> You also need shared memory semiphore to protect the sm obj. |> |> Hope it helps |> Yibing Fan |> However, since VxWorks has no task level memory protection, you could create a global array in one program/task, declair it extern in another task and just use it. This, of course, works in a single processor system. I have no experience with multiprocessor VxWorks systems. There, you may have to create a global object, etc. --------------------------- Newsgroups: comp.os.vxworks Subject: Looking for SMB server for VxWorks (x86) Date: 25 Oct 1996 04:41:03 GMT From: jfl@netistech.com (Jason Lin) Organization: Aimnet Information Services Message-ID: <54pgcv$989@news.aimnet.com> Hello, Is there any third party SMB server available for VxWorks on the x86 platform? I'm also looking for NETBui and IPX/SPX protocol stack implementation on the same platform. Any information is appreciated. Thanks in advance. Jason Lin ========== EMail: jfl@netistech.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: AT&T Assembler1 Date: 18 Oct 1996 14:31:10 -0700 From: stevemw@jetsons.wv.tek.com (Stephen M. Wynne) Organization: Tektronix, Inc. (CPID) Message-ID: References: <546ebc$o5n@newsbf02.news.aol.com> Sender: stevemw@jetsons.wv.tek.com In article <546ebc$o5n@newsbf02.news.aol.com> scotttsx@aol.com (ScottTSX) writes: Can anyone tell me where to find AT&T System V 386 Assembler syntax ( used by GNU ). Scott, Try looking at this file: http://cal011111.student.utwente.nl/ldp/HOWTO/mini/Assembly It has several pointers that may help you. Steve Wynne, Tektronix CPID Software Tools. Phone/FAX: (503) 685-3526/685-4901 M/S 63-356, P.O. Box 1000, Wilsonville, OR 97070-1000 --------------------------- End of New-News digest ********************** From han2326@aminet.co.kr Sat Oct 26 06:25:25 1996 From: "Han, Jeong Soo" Date: Sat Oct 26 06:25:27 PDT 1996 Subject: MC68302 : vxWorks.st_rom.hex Hi, I have got some problem to make a rommable image to my target(PEP VSBC4 board, with CPU MC68302). I have made vxWorks_rom.hex with my application and wrote to EPROM(4Mbit) with PROM programmer. The target board is well operating with the rommable image on a power-on(initialization state). But, when I push the RESET button(reset status) in a front panel, Any operation does not occur. Maybe the CPU 68302 dose not find a start address on a reset. I am wondering wht not? I have seen a passage at a BSP Porting Kit manual as follow. "The Motorala M683xx family is a special release. It runs on busless boards that communicate with the outside world via serial line; for this reason, the boot ROMs contain an embedded standalone vxWorks shell(vxWorks.st_rom.hex)" Should I make the vxWorks.st_rom.hex instead of vxWorks_rom.hex for my target? Any helps would be appreciated. J.S. Han From peters@mirage.phx.mcd.mot.com Sat Oct 26 07:20:54 1996 From: peters@mirage.phx.mcd.mot.com (via Siren Mail Auto-Reply facility) Date: Sat Oct 26 07:20:56 PDT 1996 Subject: I'm away from Siren Mail I'll be out of the office Oct. 28-30 - on business travel - returning on the 31st. Dave Peters From daemon@csg.lbl.gov Sun Oct 27 04:00:35 1996 From: daemon@csg.lbl.gov Date: Sun Oct 27 04:03:18 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Sun Oct 27 04:00:33 PST 1996 Subject: FTP Between VxWorks front ends Subject: Re: INterrupt levels on MVME162 board ?? Subject: Re: Flash Disk Driver for VxWorks ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: FTP Between VxWorks front ends Date: Sat, 26 Oct 1996 11:27:56 -0400 From: lavergr@LSKMP002.atsc.allied.com Organization: NASA Goddard Space Flight Center -- Greenbelt, Maryland USA Message-ID: <32722DFC.1921@LSKMP002.atsc.allied.com> We have two ethernet strings of Hp9000 workstations running HP-UX with VxWorks front ends on each. We're looking for a way to ftp from one FE to the other. Right now we ftp from FE1 to the workstation, then ftp from workstation 1 to workstation 2, then ftp from workstation 2 to FE2. The process works, but is a roundabout way of doing things. Any suggestions? Rob --------------------------- Newsgroups: comp.os.vxworks Subject: Re: INterrupt levels on MVME162 board ?? Date: Fri, 18 Oct 1996 16:38:24 -0400 From: "J. P. Marple" Organization: Johns Hopkins University Applied Physics Lab, Laurel, MD, USA Message-ID: <3267EAC0.375F@aplexus.jhuapl.edu> References: <545rmh$r5n@llnews.ll.mit.edu> <546oo7$se6@news.pacifier.com> Steve Doiel wrote: > > > You need to get your hands on teh ;-) MVME 162 hardware reference manual. It > does a good job of explaining the priority of interrupts. I used the manual > recently to set up DMA on the 162. > > As I recall... the hardware gives local interrupts the highest priority, and the > VME bus the lowest priority. Unfortunately you may be running into a hardware > limitation. The MVME162FX Embedded Controller Installation and Use Guide outlines a fixed priority scheme for local bus arbitration. Table 1-2 says: Device Priority Note - --------------------------------------- LAN 0 Highest IP DMA 1 SCSI 2 VMEbus 3 Next Lowest MC68040 4 Lowest JPM --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Flash Disk Driver for VxWorks Date: Mon, 14 Oct 1996 10:50:37 GMT From: rlykrha@rly.abb.se (Krister Hagman) Organization: ABB Network Partner AB Message-ID: <53t5vv$qs0@sdaw04.seinf.abb.se> References: <53214o$fdv@sf18.dseg.ti.com> <533mni$1pt@uuneo.neosoft.com> <3255EFB0.4DB@peacefulstar.oakland.ca.us> ------- | Actually, I'm surprised that Leonid Rosenboim hasn't already | contributed to this thread. His company has a VxWorks flash file | system, which he sells (in source) for about $12k. | His e-mail address is "leonid@rst.co.il". | Nigel - He has already! /Krister Hagman --------------------------- End of New-News digest ********************** From rterry@cacihsv.com Mon Oct 28 05:11:15 1996 From: "Russell Terry" Date: Mon Oct 28 05:11:17 PST 1996 Subject: Re: MC68302 : vxWorks.st_rom.hex Try burning bootrom.hex in your EPROM. ---------------------------------------------------- Russell M. Terry Jr. CACI-ASG Senior Systems Engineer 1555 The Boardwalk InfraRed Sensor Systems Huntsville, AL 35816 Embedded Systems 205.830.4782 x133, Custom Hardware/Software 205.837.2086, fax VME, VxWorks, SCSI From rawson.gavind@traf.com Mon Oct 28 08:10:39 1996 From: rawson.gavind@traf.com Date: Mon Oct 28 08:10:41 PST 1996 Subject: Heap Management under VxWorks Does anyone have any *cunning* heap management code that I can borrow. My heap gets fragmented and never recovers. thanks, Gav (rawson.gavind@traf.com) From jhend@vaj4432.tu.hac.com Mon Oct 28 09:17:16 1996 From: John R Henderson Date: Mon Oct 28 09:17:19 PST 1996 Subject: Shared memory We are using the VxMP package in version 5.2. I would like to modify the behavior of the VxMP such that the call to usrSmObjInit does not cause the bootup to wait for processor #0 to write a sequence number to shared memory, and for that matter, never write a sequence to shared memory. Currently VxMP does so every second. Our application essentially performs this function and I don't want to duplicate it. Does anyone know how I can do what I want with VxMP? Furthermore, is there any problem with delaying the call to usrSmObjInit until after our application is up and running? Thank you, John R Henderson HMSC From ddavies@worf.dvg.com Mon Oct 28 09:19:06 1996 From: ddavies@worf.dvg.com (Douglas Davies) Date: Mon Oct 28 09:19:08 PST 1996 Subject: Re: Heap Management under VxWorks > Submitted-by rawson.gavind@traf.com Mon Oct 28 08:10:39 1996 > > Does anyone have any *cunning* heap management code that I can borrow. > My heap gets fragmented and never recovers. Kernighan & Ritchie "The C Programming Language", 2nd Edition has an example of a heap manager. You will probably need to modify it to work from a large pre-allocated heap, rather than let it request system heap chunks when it runs out of local available heap. An easy modification... It's a good place to start, and the code is small and very easy to follow. The application does automatic combination of adjacent blocks upon deallocation, which should keep fragmentation down. -Doug Douglas Davies Director of Software Engineering Utah Scientific, Inc. From jcarissi@BBN.COM Mon Oct 28 10:53:45 1996 From: Joe Carissimi Date: Mon Oct 28 10:53:47 PST 1996 Subject: SVME163 DLCS Question I was wondering if anyone has ran into this same problem. We have a DY4 SVME163 (with the MAX 220 daugther card) running vxWorks 5.2. We have a need to boot using either of the ethernet connections provided by the board (front panel SONIC port or the cs-enet from the daughter card). Also, applications have to use the serial ports on the daughter card. Access to the daughter card (serial ports) is via the DY4 DLCS package (part of DY4's CS220 suite). Whenever we boot using the Sonic port (front panel Enet connection), and then initialize the serial ports from our application (via dlcsDrv), all if fine. But when we switch to Quebooting from the daughter card (cs-enet), the DLCS initialization sequence dies (namely the ioctl call). The CS220 patch is in (as provided by DY4). We called DY4 tech support and are looking into the problem. Has anyone out there been able to get a configuration like this to work and are there any tricks?? I appreciate responses on this topic. ---- Joe Carissimi BBN Systems & Technologies 50 Enterprise Place, Suite 201 Middletown, RI 02842 jcarissi@bbn.com 401-849-2543 (X25) From marc@onion.jhuapl.edu Mon Oct 28 11:25:52 1996 From: marc@onion.jhuapl.edu (Marcus H. Gates) Date: Mon Oct 28 11:25:58 PST 1996 Subject: "known bug" in gnu complier for 68040? We are using vxWorks 5.1.1 / gnu 2.2.3.1 for development of code to run on 68040. I'm sure that sometime a while ago I saw a "known bug" described relative to floating-point multiplies when using -O flag. Now that I want to go back and see this ( to see if it is fixed in 5.2 ), I can't find it. Does anyone else remember this problem. As I recall it had something to do with multiplying extremely small numbers. Compile flags we are using are: -ansi -nostdinc -fno-builtin -fstrength-reduce -m68040 -DCPU=MC68040 I would really like to be able to use -O2 flag as code runs a lot faster when optimized ( and my board is running out of time to get everything done in real time ). Thanks Marc Gates From js@eroch.mc.xerox.com Mon Oct 28 14:29:50 1996 From: js@eroch.mc.xerox.com (Jeffrey R. Szczepanski) Date: Mon Oct 28 14:30:00 PST 1996 Subject: Re: Heap Management under VxWorks >> Does anyone have any *cunning* heap management code that I can borrow. >> My heap gets fragmented and never recovers. >Kernighan & Ritchie "The C Programming Language", 2nd Edition has >an example of a heap manager. You will probably need to modify it >to work from a large pre-allocated heap, rather than let it request >system heap chunks when it runs out of local available heap. An easy >modification... It's a good place to start, and the code is small >and very easy to follow. The application does automatic combination >of adjacent blocks upon deallocation, which should keep fragmentation >down. > >-Doug >Douglas Davies >Director of Software Engineering >Utah Scientific, Inc. The heap manager Doug referenced in K&R is a very basic. The automatic recombination is normally called "coalesced free" and done by all but the most "cheezy" of allocators. I am sure that VxWorks coalesces free blocks. As Doug points out, this is a good starting point to see how a heap is implemented, but I would bet that it won't solve your particular problem. Heap managers are a science of their own. There was a proof done to show that no one heap manager algorithm will be optimimal for all applications. Therefore, to achieve optimum performance, one can conclude that a custom heap manager is required, or atleast the selection of "a good match" from a standard algorithm is required. There are different designs of heap managers to achieve different goals. - Constant allocation time - Constant free time - Minimum free or allocation time - Fixed allocation sizes, or 2^N allocation sizes - etc,... The simple "generic" allocators like the K&R reference and the one used in VxWorks (and nearly all OSes), assume the following to work well: - Individual allocation much less that total pool of memory ie: allocated blocks small relative to total memory size - Total usage of heap memory will be less than approx 80% or 90% of the available RAM....(depending on each allocation size) If you are violating these "rules of thumb" then, you probably need to do some research on heap allocation techniques and plan on coding your own memory allocator, or atleast selecting a different algorithm than the VxWorks one. For research on heap algorithms, start with any and all references by David Knuth, especially "The Art of Computer Programming" series and references found within. If you don't want to code your own routine, you can look into modifing the application to break large malloc's into many smaller mallocs, or work on making the size of each allocation similar or identical sizes to minimize fragmentation. Also, permanent mallocs (no free before app ends) should be done all at once, up front, so they don't artificially fragment the heap. Hope that helps some, Jeff ------------------------------------------------------------------------------ Jeffrey R. Szczepanski Phone: (716) 264-6613 Xerox Engineering Systems FAX: (716) 264-6612 300 Main Street East Rochester, NY 14445 Email: js@eroch.mc.xerox.com ------------------------------------------------------------------------------ From stan@rti.com Mon Oct 28 17:37:37 1996 From: Stan Schneider Date: Mon Oct 28 17:37:39 PST 1996 Subject: Re: Heap Management under VxWorks >> Submitted-by js@eroch.mc.xerox.com Mon Oct 28 14:29:50 1996 >> Submitted-by: js@eroch.mc.xerox.com (Jeffrey R. Szczepanski) >> >> >> >> Does anyone have any *cunning* heap management code that I can borrow. >> >> My heap gets fragmented and never recovers. >> If you can live with fixed-size buffers (e.g. you're allocating the same type structure over and over), then RTILib's FastBuffer will do it. It's very fast, and has no fragmentation issues. Otherwise, there's another memory allocation scheme marketed by Seaweed Systems. It may help; I think you can find them at info@seaweed.com -- Stan stan@rti.com From brettk@llnl.gov Tue Oct 29 11:12:14 1996 From: "Brett M. Kettering" Date: Tue Oct 29 11:12:32 PST 1996 Subject: Synchronizing Time I am interested in finding out what applications or products may be available out there for VxWorks and UNIX time-share OSs (e.g. Solaris) to synchronize the clocks. Our need is only to the 1/100 of a second, though a protocol such as NTP is capable of even better synchronization than that. I know that OSF's DCE includes the Distributed Time Service to do this type of job. Also, CORBA defines an optional time service. Anyone out there have experience trying to make sure that all their VxWorks and host computer nodes all think it is the same time? Thanks for your help. Brett ----------------------------------------------------------------------- Brett M. Kettering LLNL P. O. Box 808 L-493 Livermore, CA. 94550 E-mail: brettk@llnl.gov Voice: (510) 423-3467 FAX: (510) 422-1930 ----------------------------------------------------------------------- From uchenick@tis4000.com Tue Oct 29 11:59:18 1996 From: Gordon Uchenick Date: Tue Oct 29 11:59:20 PST 1996 Subject: errno = 0x1a9 When booting vxWorks from the Ethernet, I often see a failure with the message: error loading file, errno = 0x1a9 In vxWorks 5.1, no code in errno.h exceeds 71. What does errno = 0x1a9 mean? TIA, Gordon Uchenick uchenick@tis4000.com From johill@lanl.gov Tue Oct 29 13:35:37 1996 From: Jeff Hill Date: Tue Oct 29 13:35:39 PST 1996 Subject: C++ exceptions Hello all, The following was extracted from the "NEWS" file in the gcc 2.7.2 distribution. > Support for exception handling has been improved; more targets are now > supported, and throws will use the RTTI mechanism to match against the > catch parameter type. Optimization is NOT SUPPORTED with > -fhandle-exceptions; no need to report this as a bug. It appears that using C++ exceptions with gnu C++ 2.7.2 turns off all compiler optimization. Does anyone out there know if there is a gnu C++ release that does not have this limitation (or if gnu is even working on the problem). Jeff -- ______________________________________________________________________ Jeffrey O. Hill Internet johill@lanl.gov LANL MS H820 Voice 505 665 1831 Los Alamos, NM 87545 USA FAX 505 665 5107 From davida@laplace.idec.sdl.usu.edu Tue Oct 29 15:03:50 1996 From: davida@laplace.idec.sdl.usu.edu (David Anderson) Date: Tue Oct 29 15:03:53 PST 1996 Subject: Re: errno = 0x1a9 Hi Gordon, We had the same problem. It has to do with the tcp wait interval on the Sun not being long enough and timing out. Heurikon sent us the following solution: Recreated: rcp the vxWorks and vxWorks.sym to sparc:/tmp and boot via ftp Fix: Change the device driver setting for tcp using ndd (for more info do a man on ndd) 1) Log on to host system as the root user: 2) Enter the following commands in order: a) ndd -get /dev/tcp tcp_close_wait_interval b) ndd -set /dev/tcp tcp_close_wait_interval 5000 c) ndd -get /dev/tcp tcp_close_wait_interval On each -get ndd will return the current value of tcp_close_wait_interval, at first it is 25000...I think, and after the the -set command it will return the value of 5000. hope this helps, Dave. From dpioneer@kkucc.konkuk.ac.kr Tue Oct 29 23:22:05 1996 From: KIM WOO JIN Date: Tue Oct 29 23:22:17 PST 1996 Subject: registration to the vxworks users group! my name is Woojin-Kim. I bave registrated to vxworks users group at past day, but i have received that i'm not menber. Plese send me the method that what i have to do. From daemon@csg.lbl.gov Wed Oct 30 04:00:30 1996 From: daemon@csg.lbl.gov Date: Wed Oct 30 04:00:34 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Wed Oct 30 04:00:28 PST 1996 Subject: Non Terminating TCP/IP Sockets Subject: Interrupts on the PC platform - help Subject: test Subject: Interrupts on PC platform-help Subject: Re: VME interrupt levels & vectors Subject: Re: Heap Management under VxWorks Subject: Recompiling for 5.1 -> 5.2? Subject: Multiple file read problem in 5.3 Subject: Software Archives Subject: Problems mounting NFS files Subject: vxworks x86 driver for multi-port ISA serial board Subject: Re: Tcl/Tk Subject: Re: Problems mounting NFS files Subject: IIIIIIIIII $80 IIIIIIIIII MS OFFICE PROFESSIONAL 95 IIIIIIIIII $80 IIIIIIIIII Subject: Re: No Serial H/W Flow Control? Subject: VxSim Subject: Motorola Flash Programming Subject: Re: VME interrupt levels & vectors Subject: Re: Interrupts on the PC platform - help Subject: VxSim Subject: Re: Motorola Flash Programming ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Non Terminating TCP/IP Sockets Date: Mon, 28 Oct 1996 14:04:22 -0500 From: Ray Gilbert Organization: General Instruments Communications Division Message-ID: <327503B6.7860@gic.gi.com> Reply-To: rgilbet@gic.gi.com Hello, I am currently experiencing a problem with non-terminating tcp/ip sockets. We are running VxWorks 5.2 on a MC68360 board. The board acts as a server for a SCO Unix client. The client opens about 20 connections to the server, and each connection is handled by a different handler spawned on the 360. Now, each handler task blocks waiting on its socket with fioRead. After the server has been up and the client connected for some hours (exact number not yet determined) and you shutdown the client, 17-18 of the connects will terminate/close gracefully, but 2-3 will remain. The handler tasks remain block-waiting on sockets that are no longer valid. Has anyone else experienced this problem???? Thanks Ray --------------------------- Newsgroups: comp.os.vxworks Subject: Interrupts on the PC platform - help Date: Mon, 28 Oct 1996 12:30:26 -0800 From: eric olsen Organization: stanford university Message-ID: <327517E2.3161F610@leland.stanford.edu> Hello, We are using vxworks on the pc platform. We have an isa card which generates interrupts at any rate from 1-2khz. It can be set to use interrupt 10-12,14, or 15. We are unable to connect our ISR to the IRQ. We have tried using intConnect, and we have even tried using intVecSet, and writing the priority/EOI's directly to the PIC. It just doesn't seem to work. We know that the isa card is working, but vxworks will just not respond to the interrupt. We think it may have something to do with the vector associated with the IRQ. What vector is associated with IRQ 10??? We also don't understand how the macro INUM_TO_IVEC works. Any help will be appreciated. Thanks, - -------------------------------------------------------------- "How can I lose to such an idiot?" Aaron Nimzovich(1886-1935) eric olsen eolsen@leland.stanford.edu - -------------------------------------------------------------- --------------------------- Newsgroups: comp.os.vxworks Subject: test Date: Mon, 28 Oct 1996 12:28:33 -0800 From: eric olsen Organization: stanford university Message-ID: <32751771.1918C6A7@leland.stanford.edu> test - -------------------------------------------------------------- "How can I lose to such an idiot?" Aaron Nimzovich(1886-1935) eric olsen eolsen@leland.stanford.edu - -------------------------------------------------------------- --------------------------- Newsgroups: comp.os.vxworks Subject: Interrupts on PC platform-help Date: Mon, 28 Oct 1996 12:20:39 -0800 From: eric olsen Organization: stanford university Message-ID: <32751597.678BEC15@leland.stanford.edu> Hello, We are using vxworks on the pc platform. we have an isa card which generates interrupts at any rate from 1-2khz. It can be set to use interrupt 10-12,14, or 15. We are unable to connect our ISR to the IRQ. We have tried using intConnect, and we have even tried using intVecSet, and writing the priority/EOI's directly to the PIC. It just doesn't seem to work. We know that the isa card is working, but vxworks will just not respond to the interrupt. We think it may have something to do with the vector associated with the IRQ. What vector is associated with IRQ 10??? We also don't understand how the macro INUM_TO_IVEC works. Any help will be appreciated. Thanks, - -------------------------------------------------------------- "How can I lose to such an idiot?" Aaron Nimzovich(1886-1935) eric olsen eolsen@leland.stanford.edu - -------------------------------------------------------------- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VME interrupt levels & vectors Date: Mon, 28 Oct 1996 14:38:05 -0500 From: "Bruce K. Gitlin" Message-ID: <32750B9D.EC6@wl.corning.com> References: <01bbc25a$6bec15e0$f371e8c7@rgross.venus.star.net> Ron Gross wrote: > > We have a MVME-133XT (68020) running 4.0.2VxWorks (working on moving to > 5.2) and have installed a new card in out VME rack (DeltaTau PMAC-VME). > This card provides three interrupt vectors on one interrupt level, > depending on what the interrupt was for. I have not setup interrupts on > VxWorks before, and the docs (4.0.2 and 5.2) do not seem to talk about > levels. > > 1. How do I tell VxWorks about the interrupt level (3 or 4 can be used)? > Does VxWorks just trap all level interrupts and automatically get the > vector number and use that to 'vector' to the selected routine? > > 2. Do I use a seperate routine to handle each interrupt vector? > > -- > Ron Gross > Precision Data Services 1a. sysIntEnable() enables an interrupt level (1 thru 7). The call is in sysLib. b. You use intConnect() to connect a routine to a vector. 2. You can use separate routines or you can use a single routine. The third parameter of intConnect() is a parameter to pass to the routine. This can be used to determine who triggered the interrupt. The best way to do it depends on the application. If the code for all three interrupts are similar, I would use a case statement and switch on the parameter passed in. We've done that in the past and it seemed to work OK. Hope that helps, Bruce - -- ============================== || Bruce K. Gitlin || || CTG / Corning || /)| (910)397-8204 |(\ / )| gitlin_bk@wl.corning.com |( \ ( (==============================) ) (((\ \) /,) / ) (/ //)) (\\\\ \_/ / \ \_/ ///// \ / \ / \ _/ \_ / - ----/ /--------------------------\ \---- / / \ \ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Heap Management under VxWorks Date: Mon, 28 Oct 1996 12:38:45 -0800 From: John Finley Organization: Netcom Message-ID: <327519D5.22ED@finley.com> References: <9610281718.AA03901@worf.dvg.com> Reply-To: john@finley.com Douglas Davies wrote: > > > Submitted-by rawson.gavind@traf.com Mon Oct 28 08:10:39 1996 > > > > Does anyone have any *cunning* heap management code that I can borrow. > > My heap gets fragmented and never recovers. > > Kernighan & Ritchie "The C Programming Language", 2nd Edition has > an example of a heap manager. You will probably need to modify it > to work from a large pre-allocated heap, rather than let it request > system heap chunks when it runs out of local available heap. An easy > modification... It's a good place to start, and the code is small > and very easy to follow. The application does automatic combination > of adjacent blocks upon deallocation, which should keep fragmentation > down. .... VxWorks does the coagulation of freed blocks. Perhaps there is something else in there worth using? I think VxWorks is first-fit, so just going to best-fit might help, at the cost of speed. A simple first-fit or best-fit allocator is pretty simple, although I realize that getting one already done is even simpler :). I seem to recall that someone has a commercial memlib replacement out there; could it be RTI? or Sparta? John - -------------------------------------------------- John Finley Engineering Consultant john@finley.com Real Time Systems (619) 689-0032 Networking Protocols - -------------------------------------------------- --------------------------- Newsgroups: comp.os.vxworks Subject: Recompiling for 5.1 -> 5.2? Date: 16 Oct 1996 00:36:41 GMT From: litwin@robotics.jpl.nasa.gov (Todd Litwin) Organization: Jet Propulsion Laboratory, Pasadena, CA Message-ID: <541amp$cn7@netline-fddi.jpl.nasa.gov> Is it necessary to recompile modules when moving from VxWorks 5.1 to 5.2? I'm running on a 68060, and I have third-party software which was designed for 5.1. Am I in trouble? Where in the VxWorks documentation can I look to find out if recompiling is necessary for a new version of the software? (I think I knew this a few years ago, but my memory seems to be decaying....) - -- Todd Litwin Jet Propulsion Laboratory (818) 354-5028 Todd.E.Litwin@jpl.nasa.gov --------------------------- Newsgroups: comp.os.vxworks Subject: Multiple file read problem in 5.3 Date: Tue, 29 Oct 1996 09:38:57 -0800 From: Dave Zaleta Organization: Reticular Systems, Inc. Message-ID: <32764131.41C67EA6@reticular.com> Wanted to find out if anyone out there has had any problems with either opening and closing many (>10) files or opening and closing a single file many times in 5.3. The problem that we have with our Radstone 68-64 board is that when the program below is run it indefinitely pends at or around the 8th iteration. The have also confirmed that the same thing happens with a CPU-40 board, however, it does not happen with the POWER-PC board and Wind River was unable to reproduce the problem or give any advice on how to fix the problem. Radstone was able to find out that the problem appears to be with the Unix server deamon (in our case a Sun sparc20 running SunOS 4.1.4). Apparently the Unix deamon 'inetd' had stopped listening to the port, and thus stopped receiving. They indicated that this was a safety mechanism...apparently Unix thinks that the same command is being repeated and there is something wrong. A log message in /var/adm/messages shows an error like this: Oct 16 11:24:57 dvorak inetd[147]: shell/tcp server failing (looping), service terminated Has anyone run into this problem? If so how did you deal with it and what needs to be changed in order to get it to work? Any help would be greatly appreciated. dave z zaleta@reticular.com --------------------------- Newsgroups: comp.os.vxworks Subject: Software Archives Date: Tue, 29 Oct 1996 13:03:14 -0800 From: eric olsen Organization: stanford university Message-ID: <32767112.7636E565@leland.stanford.edu> I understand that there is an archive containing publically available libraries, etc for vxworks. I am looking for a port of the curses library(to the pc platform). Where is this archive located?????? thanks, - -------------------------------------------------------------- "How can I lose to such an idiot?" Aaron Nimzovich(1886-1935) eric olsen eolsen@leland.stanford.edu - -------------------------------------------------------------- --------------------------- Newsgroups: comp.os.vxworks Subject: Problems mounting NFS files Date: 28 Oct 1996 11:59:46 -0500 From: David Uhrenholdt Organization: Sanders SSD Message-ID: Sender: duhrenho@hickory I am having trouble NFS Mounting directories that are NFS mounted on the NFS host. If my VxWorks NFS host is HP-UX 9.0.7 perror produces S_errno_ENOSYS, If the NFS host is SunOS 4.1.3, or Solaris 2.5, I get S_rpcLib_RPC_TIMEDOUT. If the filesystem is local to the NFS host there isn't any problem. The HP-UX systems have UDP checksumming on. Does anyone have insight into this problem? Dave - -- Dave Uhrenholdt duhrenho@vette.sanders.lockheed.com Phone (603) 885-7398 --------------------------- Newsgroups: comp.os.vxworks Subject: vxworks x86 driver for multi-port ISA serial board Date: Tue, 29 Oct 1996 23:02:47 GMT From: akimoto@netcom.com (Steve Akimoto) Organization: NETCOM On-line Communication Services (408 261-4700 guest) Message-ID: Sender: akimoto@netcom6.netcom.com Keywords: VxWorks, x86 driver, multi-port serial board (ISA) We are looking for a supplier of an x86 VxWorks driver and an ISA multi-port serial board. Can anyone provide a reference? Need: VxWorks x86 driver ISA multi-port serial board with buffered UART RS-232, RS-422, & RS-485 ports 8 serial ports (56 kbps max per channel) 115.2 kbps (max aggregate rate) Thanks --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Tcl/Tk Date: 30 Oct 1996 01:04:19 GMT From: Jim Spears Organization: Hughes Aircraft Company Message-ID: <5569ij$esc@hacgate2.hac.com> References: <3275C320.893@llnl.gov> you betcha, we are using Tcl/Tk on vxWorks here at Hughes Aircraft Company, fullerton, ca. jim spears sr staff engineer hac --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Problems mounting NFS files Date: 29 Oct 1996 15:37:11 GMT From: hartl@Informatik.TU-Muenchen.DE (Anton Hartl) Organization: Technische Universitaet Muenchen, Germany Message-ID: <5558b7$5hj@sunsystem5.informatik.tu-muenchen.de> References: David Uhrenholdt writes: >I am having trouble NFS Mounting directories that >are NFS mounted on the NFS host. By definition of NFS this is NOT allowed. You have to mount always from the host that physically provides the filesystem you intend to mount. If NFS-mounting an NFS-mounted filesystem were possible than any access limitations the original exporter might have defined can be rendered useless on the NFS-mounting host. >Does anyone have insight into this problem? Hope that has clarified the problem. - -Toni --------------------------- Newsgroups: comp.os.os2.bugs,comp.os.os2.comm,comp.os.os2.games,comp.os.os2.mail-news,comp.os.os2.marketplace,comp.os.os2.misc,comp.os.os2.multimedia,comp.os.os2.networking.misc,comp.os.os2.networking.tcp-ip,comp.os.os2.networking.www,comp.os.os2.programmer.misc,comp.os.os2.programmer.oop,comp.os.os2.programmer.porting,comp.os.os2.programmer.tools,comp.os.os2.setup,comp.os.os2.setup.misc,comp.os.os2.setup.storage,comp.os.os2.setup.video,comp.os.os2.utilities,comp.os.os9,comp.os.parix,comp.os.qnx,comp.os.rs Subject: IIIIIIIIII $80 IIIIIIIIII MS OFFICE PROFESSIONAL 95 IIIIIIIIII $80 IIIIIIIIII Date: 30 Oct 96 06:08:56 UTC From: noemail@aol.com (M Blank) Message-ID: <32eee33a.9722500@news2.compulink.com> Due to personal financial hardships I find myself in the unenviable position of trying to get back a fraction of what I paid for this program. My retailer refuses returns on 'opened' software, even if I haven't installed the package. Microsoft couldn't care less about my situation, and numerous pleas for an exception to their return policy have fallen on deaf ears. Believe it or not, I need this money more than any program, and I thought someone could at least benefit from this awkward situation. To clarify, this is a store-bought full retail package which contains 2 cd's (unopened), a license agreement and manual, and includes the full versions of WORD and EXCEL (versions 7.0) POWERPOINT, SCHEDULE, ACCESS and BOOKSHELF. This is a stand alone product and not an upgrade (you don't need older versions to run it). It was my intention to put these programs on a shelf somewhere until I needed them, but I do not have that luxury. My resources are extremely limited and I am at the point where this is the only responsible thing to do. I'm just hoping there is someone out there who can use these programs. Please contact me if you or someone you know might be interested. I do not have my own email address yet, however you can write to me at: M. Blank Box 54 Toronto, Ontario M5A 1N1 Thank you and God Bless. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: No Serial H/W Flow Control? Date: Tue, 29 Oct 1996 19:52:24 +0000 From: "H.J. Bae" Organization: Peaceful Star Message-ID: <32766078.2789@idiom.com> References: <3276B402.95B@teleport.com> Reply-To: hjb@peacefulstar.oakland.ca.us it's true. --------------------------- Newsgroups: comp.os.vxworks Subject: VxSim Date: Wed, 30 Oct 1996 10:48:54 +0200 From: "Natalie Rogozovsky (SDH-S)" Organization: Telematics International, Inc. Message-ID: <32771676.31DFF4F5@ecitele.com> References: <541c35$7g9@news.aimnet.com> We want to develop our application on 5.2 using VxSim and CrossWind on the SUN-SunOs platform. Has anyone experience with it? We have the problems because of the large size of our object. Thank you. Nataly --------------------------- Newsgroups: comp.os.vxworks Subject: Motorola Flash Programming Date: Sun, 20 Oct 1996 21:21:12 +0100 From: John Gordon Organization: UUNet PIPEX server (post doesn't reflect views of UUNet PIPEX) Message-ID: <326A89B8.4C01@dial.pipex.com> Hi, Has anybody written a VxWorks app to re-program the boot ROM image into the flash on MVME1604/2604 boards? I am getting fed up taking all the devices out to do them on our Data I/O programmer! If not I might have to start working on one - has anybody the algorithm for writing to AMD 29f040 flash? John... --------------------------- Newsgroups: comp.os.vxworks Subject: Re: VME interrupt levels & vectors Date: Sun, 27 Oct 1996 10:38:55 GMT From: krister@unidata.se (Krister Lagerstrom) Organization: Tripnet AB Message-ID: <54vh5s$rlj@minox.tripnet.se> References: <01bbc25a$6bec15e0$f371e8c7@rgross.venus.star.net> "Ron Gross" wrote: >We have a MVME-133XT (68020) running 4.0.2VxWorks (working on moving to >5.2) and have installed a new card in out VME rack (DeltaTau PMAC-VME). >This card provides three interrupt vectors on one interrupt level, >depending on what the interrupt was for. I have not setup interrupts on >VxWorks before, and the docs (4.0.2 and 5.2) do not seem to talk about >levels. >1. How do I tell VxWorks about the interrupt level (3 or 4 can be used)? Try sysIntEnable (3) to enable handling of VME bus interrupt 3. >Does VxWorks just trap all level interrupts and automatically get the >vector number and use that to 'vector' to the selected routine? Yes. >2. Do I use a seperate routine to handle each interrupt vector? You have the choice of using one ISR (interrupt service routine) per vector, or you could use one ISR to handle all of them. Use intConnect() to connect a vectorized interrupt to an ISR written in C. You're welcome to email me if you have more questions. / Krister --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Interrupts on the PC platform - help Date: 28 Oct 1996 21:29:00 -0800 From: dzhu@enlil.jpl.nasa.gov (David Zhu) Organization: jpl Message-ID: References: <327517E2.3161F610@leland.stanford.edu> Sender: dzhu@enlil.jpl.nasa.gov In article <327517E2.3161F610@leland.stanford.edu> eric olsen writes: > From: eric olsen > Newsgroups: comp.os.vxworks > Date: Mon, 28 Oct 1996 12:30:26 -0800 > Organization: stanford university > Path: netline-fddi.jpl.nasa.gov!nntp1.jpl.nasa.gov!iag.net!enews.sgi.com!news.sgi.com!nntp-hub2.barrnet.net!news.Stanford.EDU!usenet > Lines: 28 > NNTP-Posting-Host: rachel.stanford.edu > Mime-Version: 1.0 > Content-Type: text/plain; charset=us-ascii > Content-Transfer-Encoding: 7bit > X-Mailer: Mozilla 3.0Gold (X11; I; Linux 2.0.6 i586) > > Hello, > > We are using vxworks on the pc platform. We have an > isa card which generates interrupts at any rate from > 1-2khz. It can be set to use interrupt 10-12,14, or 15. > > We are unable to connect our ISR to the IRQ. We have > tried using intConnect, and we have even tried using > intVecSet, and writing the priority/EOI's directly to > the PIC. It just doesn't seem to work. > > We know that the isa card is working, but vxworks will > just not respond to the interrupt. We think it may have > something to do with the vector associated with the IRQ. > What vector is associated with IRQ 10??? We also don't > understand how the macro INUM_TO_IVEC works. > > Any help will be appreciated. > > Thanks, > > -------------------------------------------------------------- > "How can I lose to such an idiot?" > Aaron Nimzovich(1886-1935) > > eric olsen > eolsen@leland.stanford.edu > -------------------------------------------------------------- > The PC interrupt has to be converted to i86 interrupt. As defined in pc.h, INT_VEC_IRQ0 (the timer) is 0x20 since interrupts 0 - 31 are for internal use or reserved. So IRQ1 is 0x21, IRQ2 is 0x22 etc. INUM_TO_IVEC then maps interrupt number to interrupt vector which is an index into the interrupt descriptor table (IDT). INUM_TO_IVEC just scales interrupt number by eight which is the size of an interrupt descriptor. David - -- - ---------------------------------------------------------------------- David Q. Zhu Jet Propulsion Laboratory e-mail: David.Zhu@jpl.nasa.gov MS 198-235 phone: 818-354-9058 (office) 4800 Oak Grove Dr. Pasadena, CA 91109 - ---------------------------------------------------------------------- --------------------------- Newsgroups: comp.os.vxworks Subject: VxSim Date: Wed, 30 Oct 1996 13:10:03 +0200 From: "Natalie Rogozovsky (SDH-S)" Organization: Telematics International, Inc. Message-ID: <3277378B.237C228A@ecitele.com> We want to develop our application on 5.2 using VxSim and CrossWind on the SUN-SunOs platform. Has anyone experience with it? We have the problems because of the large size of our object. Thank you. Nataly --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Motorola Flash Programming Date: 29 Oct 1996 23:42:58 -0800 From: chaos@idiom.com (H.J. Bae) Organization: Peaceful Star, Oakland, CA Message-ID: <5570u2$3ck@idiom.com> References: <326A89B8.4C01@dial.pipex.com> In article <326A89B8.4C01@dial.pipex.com>, John Gordon wrote: > >If not I might have to start working on one - has anybody the >algorithm for writing to AMD 29f040 flash? > AMD 29f040 is a very simple flash device. You send command sequences to the address and data bus in pre-defined patterns like 0x5555, 0x2aaa, etc. The details are all well-specified in the datasheet. Wind River has a driver for this part, I think. - -- http://peacefulstar.oakland.ca.us mailto:hjb@peacefulstar.oakland.ca.us --------------------------- End of New-News digest ********************** From johill@lanl.gov Wed Oct 30 05:49:54 1996 From: Jeff Hill Date: Wed Oct 30 05:50:13 PST 1996 Subject: Re: comp.os.vxworks newsdigest > Hello, > I am currently experiencing a problem with non-terminating tcp/ip > sockets. We are running VxWorks 5.2 on a MC68360 board. The board acts > as a server for a SCO Unix client. The client opens about 20 > connections to the server, and each connection is handled by a different > handler spawned on the 360. Now, each handler task blocks waiting on > its socket with fioRead. After the server has been up and the client > connected for some hours (exact number not yet determined) and you > shutdown the client, 17-18 of the connects will terminate/close > gracefully, but 2-3 will remain. The handler tasks remain block-waiting > on sockets that are no longer valid. Has anyone else experienced this > problem???? > Use of socket option SO_KEEPALIVE will cause TCP/IP to shut down stale circuits after some (usually long) delay. Here is an extract from "man setsockopt". SO_KEEPALIVE enables the periodic transmission of messages on a connected socket. Should the connected party fail to respond to these messages, the con- nection is considered broken. A process attempting to write to the socket receives a SIGPIPE signal and the write opera- tion returns an error. By default, a process exits when it receives SIGPIPE. A read operation on the socket returns an error but does not generate SIGPIPE. If the process is waiting in select(2) when the connection is broken, select() returns true for any read or write events selected for the socket. Jeff -- ______________________________________________________________________ Jeffrey O. Hill Internet johill@lanl.gov LANL MS H820 Voice 505 665 1831 Los Alamos, NM 87545 USA FAX 505 665 5107 From AS_Bridge@MSMAIL3.HAC.COM Wed Oct 30 06:30:43 1996 From: "AS Bridge" Date: Wed Oct 30 06:30:46 PST 1996 Subject: Undeliverable Mail Unknown Microsoft mail form. Approximate representation follows. Message: comp.os.vxworks newsdigest Sent: Wed, Oct 30, 1996 5:28 AM To: Quiroz, Gabriel On Server: 631 Segment2 Date: Wed, Oct 30, 1996 6:28 AM Reason: Could not be delivered because the destination Microsoft Mail server could not be found. From cjs@cscsgi.larc.nasa.gov Wed Oct 30 09:47:13 1996 From: "Christopher J. slominski" Date: Wed Oct 30 09:47:15 PST 1996 Subject: Re: comp.os.vxworks newsdigest You can get the curses library from vxworks_archive@ncar.ucar.edu. Check out the Users Group section on the Wind Rivers home page. http://www.wrs.com/usrgroup.html *----------------------------------------* | Christopher J. Slominski | | Computer Sciences Corporation (CSC) | | 3217 N. Armistead Ave. | | Hampton, VA 23666 | | (757) 766-8258 | | c.j.slominski@larc.nasa.gov | | NASA Mail Stop 931 | *----------------------------------------* From froeber@bbn.com Wed Oct 30 18:33:11 1996 From: Fred Roeber Date: Wed Oct 30 18:33:13 PST 1996 Subject: Re: C++ exceptions On Tue, 29 Oct 1996, Jeff Hill wrote: > The following was extracted from the "NEWS" file in the gcc 2.7.2 > distribution. > > > Support for exception handling has been improved; more targets are now > > supported, and throws will use the RTTI mechanism to match against the > > catch parameter type. Optimization is NOT SUPPORTED with > > -fhandle-exceptions; no need to report this as a bug. > > It appears that using C++ exceptions with gnu C++ 2.7.2 turns off > all compiler optimization. Does anyone out there know if there is > a gnu C++ release that does not have this limitation (or if gnu > is even working on the problem). There was a very interesting editorial by PJ Plaugher at the end of the November issue of Embedded Systems magazine about a new emerging "Embedded C++" standard. This looks like something that could become quite applicable to us types developing embedded systems under VxWorks. He said that Cygnus was a supporter behind this standard and that a version of the GNU C++ compiler supporting it was close (if not ready). The interesting thing is that the new standard, which attempts to specify a subset of C++ that is capable of being efficient for embedded applications, precludes the use of either exceptions or RTTI mechanisms. In a sense, Embedded C++ could be considered as an answer to this question and is something that "GNU", or more properly Cygnus, would seem to be working on. People interested in C++ for VxWOrks applications might want to check out the editorial. Fred | Fred J Roeber, BBN Systems & Technologies | | 50 Enterprise Place Middletown, RI 02842-5202 | | froeber@bbn.com 401-849-2543 (X48) | | See http://www.bbn.com/tv for TraceMaker info | From leonid@rst.co.il Thu Oct 31 00:21:01 1996 From: leonid@rst.co.il (Leonid Rosenboim) Date: Thu Oct 31 00:21:04 PST 1996 Subject: Re: Interrupts on the PC platform - help > We are using vxworks on the pc platform. We have an > isa card which generates interrupts at any rate from > 1-2khz. It can be set to use interrupt 10-12,14, or 15. Funny you should ask, just yesterday I went through the same issue with one of our local customers. Here is what you should do: Assu,e the IRQ number of your device is defined is MY_IRQ, and your ISR is name myIsr(), here is what you do: extern UINT sysVectorIRQ0; intConnect( INUM_TO)IVEC( sysVectorIRQ0+MY_IRQ), myIsr, 0); sysIntEnablePIC( MY_IRQ); You can see examples in target/config/pc486/sysLib.c of other devices. ----------------------------------------------------------------------- Leonid Rosenboim Phone: +972-3-559-8144 R S T Software Industries Ltd. Mobile: +972-50-307-142 P.O.Box 11502, AZUR 58017, Israel Fax: +972-3-559-8244 WWW: http://www.rst.co.il E-Mail: leonid@rst.co.il From daemon@csg.lbl.gov Thu Oct 31 04:00:20 1996 From: daemon@csg.lbl.gov Date: Thu Oct 31 04:00:23 PST 1996 Subject: comp.os.vxworks newsdigest Comp.Os.Vxworks Daily Digest Thu Oct 31 04:00:17 PST 1996 Subject: Re: Multiple file read problem in 5.3 Subject: Tornado on WinNT Subject: No Serial H/W Flow Control? Subject: Kbhit Subject: Re: Synchronizing Time Subject: Re: Non Terminating TCP/IP Sockets Subject: Re: Tornado on WinNT Subject: varargs on PowerPC with 5.2 Subject: Tcl/Tk Subject: C++ names Subject: Re: C++ names Subject: Re: Non Terminating TCP/IP Sockets Subject: Re: Non Terminating TCP/IP Sockets Subject: Immediate Job Openings - RI & CT Subject: Re: Tornado on WinNT Subject: Re: FTP Between VxWorks front ends Subject: PCMCIA ethernet driver for VxWorks ------------------------------------------------------- Newsgroups: comp.os.vxworks Subject: Re: Multiple file read problem in 5.3 Date: Tue, 29 Oct 1996 21:06:37 -0500 From: rypma@waterloo.hp.com (Ted Rypma) Organization: Hewlett-Packard Panacom Division Message-ID: References: <32764131.41C67EA6@reticular.com> In article <32764131.41C67EA6@reticular.com>, Dave Zaleta wrote: >Radstone was able to find out that the problem appears to be with the >Unix server deamon (in our case a Sun sparc20 running SunOS 4.1.4). >Apparently the Unix deamon 'inetd' had stopped listening to the port, Dunno about SunOS, but for our HP-UX systems, we needed an inetd patch to fix this behavior. I must say, 8 iterations to shutdown is a little severe, no matter how close together! Another workaround is to bypass inetd and have the protocol daemon hang around, although that may not be practical or feasible. - -- - ----- Ted Rypma, Waterloo, Ontario, Canada === rypma@waterloo.hp.com --------------------------- Newsgroups: comp.os.vxworks Subject: Tornado on WinNT Date: Wed, 30 Oct 1996 13:47:57 GMT From: dhansen@btree.com (Dave Hansen) Organization: B-Tree Verification Systems, Inc. Message-ID: <557mjj$bti@cobra.Minn.Net> Reply-To: dhansen@btree.com Apologies if this appears multiple times. My news server has been "throttling" and I don't think my previous attempts have made it out... We will be beginning an evaluation of VxWorks and Tornado for a new product we're starting work on. My new computer will arrive with NT 4.0 installed. The Tornado docs specify NT 3.51, and previous messages in this group have indicated that Wind River is less than completely supportive of 4.0. So I have a 3.51 CD and the latest service pack downloaded, and I'm all ready to revert once the system arrives. But the sales guy and the AE from Wind River say this is no longer true (and in fact never was). Their only caution was to ensure the FTP server is installed on the Tornado host (this is not the default for 4.0). So how about it? Has anyone recently been turned away by Wind River support for having 4.0 installed rather than 3.51? Also, are there any other gems of wisdom regarding Tornado on NT out there? I know UN*X would probably be a better option for various reasons, but that option is not available to me. (at least I don't have to try Win95 :) Thanks in advance, -=Dave I can barely speak for myself, so I surely can't speak for B-Tree |==================================================================| | Dave Hansen | B-Tree Verification Systems, Inc. | | dhansen@btree.com | 5929 Baker Rd, Suite 475 | | (VOICE) 612-930-4112 | Minnetonka, MN 55345-5955 | | (FAX) 612-936-2187 | | |==================================================================| --------------------------- Newsgroups: comp.os.vxworks Subject: No Serial H/W Flow Control? Date: Tue, 29 Oct 1996 17:48:50 -0800 From: Frank Weissig Organization: Teleport - Portland's Public Access (503) 220-1016 Message-ID: <3276B402.95B@teleport.com> Reply-To: faw@teleport.com I'm using VxWorks on a MVME162. From my inspection of the tyLib docs and the z8530 files, it appears that RTS/CTS flow control is not supported. Is this true? - -- Frank Weissig Senior Staff Engineer STEP Technology, Inc. frank@steptech.com --------------------------- Newsgroups: comp.os.vxworks Subject: Kbhit Date: Tue, 29 Oct 1996 16:05:32 -0800 From: eric olsen Organization: stanford university Message-ID: <32769BCC.23FB72DA@leland.stanford.edu> Is there a kbhit function? If so, where? thanks, - -------------------------------------------------------------- "How can I lose to such an idiot?" Aaron Nimzovich(1886-1935) eric olsen eolsen@leland.stanford.edu - -------------------------------------------------------------- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Synchronizing Time Date: Wed, 30 Oct 1996 09:25:13 -0800 From: "H.J. Bae" Organization: Peaceful Star Message-ID: <32778F79.641B@peacefulstar.oakland.ca.us> References: Reply-To: hjb@peacefulstar.oakland.ca.us There is a port of NTP in vxWorks archive. - -- http://peacefulstar.oakland.ca.us mailto:hjb@peacefulstar.oakland.ca.us --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Non Terminating TCP/IP Sockets Date: Wed, 30 Oct 1996 09:33:20 -0800 From: "H.J. Bae" Organization: Peaceful Star Message-ID: <32779160.3AF8@peacefulstar.oakland.ca.us> References: <327503B6.7860@gic.gi.com> Reply-To: hjb@peacefulstar.oakland.ca.us It depends on what you mean when you "shutdown" the clients. If you properly shutdown client applications using shutdown() followed by close() socket calls, the server side should receive a TCP FIN, and read() on server socket will return with error, which will tell you when client goes away. If you just power down (withoug doing graceful shutdown of Unix OS) on the client side, server won't know anything about client going away. - -- http://peacefulstar.oakland.ca.us mailto:hjb@peacefulstar.oakland.ca.us --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Tornado on WinNT Date: Wed, 30 Oct 1996 09:23:45 -0800 From: "H.J. Bae" Organization: Peaceful Star Message-ID: <32778F21.4EF5@peacefulstar.oakland.ca.us> References: <557mjj$bti@cobra.Minn.Net> Reply-To: hjb@peacefulstar.oakland.ca.us I use WinNT4.0 and Tornado all the time. It works well. However, I do not use Tornado's IDE/GUI interface. I just use GNU compiler that comes with Tornado. I've found that Tornado GUI has a lot of bugs that makes it unusable under heavy use. Instead, I use emacs to edit and compile the programs, and use command line version of GNU gdb. Just like Unix. - -- http://peacefulstar.oakland.ca.us mailto:hjb@peacefulstar.oakland.ca.us --------------------------- Newsgroups: comp.os.vxworks Subject: varargs on PowerPC with 5.2 Date: Tue, 29 Oct 1996 16:27:30 GMT From: bonnett@sunland.gsfc.nasa.gov (Joe Vitale) Organization: NASA Goddard Space Flight Center -- Greenbelt, Maryland USA Message-ID: <555bdp$cms@post.gsfc.nasa.gov> Reply-To: bonnett@sunland.gsfc.nasa.gov We have run into a dead end trying to implement our own functions that take a variable number of arguments as input. We are using VxWorks 5.2 on an Ultra PowerPC 603 motherboard. We are using the Green Hills C/C++ Compiler and everything is hosted on a Solaris workstation. We've already implemented the function(s) and while they compile and don't crash the system, they also don't work. In talking with WRS tech support we have learned this is a known problem in 5.2 and one that will not be fixed since the problem is corrected in the latest 5.3 release. We are using the PowerPC environment for development while our "real" target hardware is being built. Our "real" hardware is a RAD6000 based board which is currently only supported by VxWorks in their 5.2 release. So upgrading to 5.3 would only be a temporary solution. Hence the dead end. Or is it? I'm wondering how many others out there may have encountered this problem and if any of you have a fix or work around? If we were writing code from scratch it would be less of an impact. However, we have quite a bit of legacy code we want to use and it uses vararg functions for outputting debug information. We'd rather not have to rework all that code. Thanks in advance, Joe ================================================================= Joe Vitale, NASA Goddard Space Flight Center, the Hammers Company Check out "OOG, The Object Orientation Game" by MCM Productions A 4-in-1 tiling puzzle game for Windows - http://www.mcmprod.com/ --------------------------- Newsgroups: comp.os.vxworks Subject: Tcl/Tk Date: Tue, 29 Oct 1996 08:41:10 +0000 From: Doug Owens Organization: Lawrence Livermore National Lab Message-ID: <3275C320.893@llnl.gov> Reply-To: owens2@llnl.gov Has Tcl/Tk been ported to vxWorks? Is anyone using it for user interface development? Thanks, Doug Owens (owens2@llnl.gov) --------------------------- Newsgroups: comp.os.vxworks Subject: C++ names Date: Tue, 29 Oct 1996 16:21:17 +0000 From: Hepworth & Hatcher Organization: UUNet PIPEX server (post doesn't reflect views of UUNet PIPEX) Message-ID: <32762EFD.13E0@dial.pipex.com> Us again ..... We are having problems unmangling names on our VxWorks system. If we compile the following program as a C++ program the fred name appears as fred__Fv in the symbol table. void fred(void) { /* body */ } When I load the munched .out file to the target and type lkup "fred" I get the output - ->lkup "fred" fred(void) text ("address") - -> Which I guess is correct. VxWorks says that fred is a function with argument void. If I now type "fred" on the shell command line I get the error message "undefined symbol". The only way I can run fred is to type fred__Fv. This does not seem to fit with the description in the VxWorks manual. Overloaded names exhibit the same problem. I never get prompted to select the correct name as described in the documentation. Has anyone else had this problem and how did they fix it. By the way we are not using the WindClass C++ libraries. Acording to the manuals Tornado/VxWorks is C++ ready !!!!! Host : PC, Win95, Tornado 1.0 Target : MVME1302, PowerPC, VxWorks P.Hatcher hephat@dial.pipex.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: C++ names Date: Thu, 31 Oct 1996 16:50:35 -0800 From: Graham Waters Organization: TRIUMF Message-ID: <3279495B.2BA3@triumf.ca> References: <32762EFD.13E0@dial.pipex.com> Hepworth & Hatcher wrote: > > Us again ..... > > We are having problems unmangling names on our VxWorks system. > > If we compile the following program as a C++ program the fred name appears as fred__Fv in the > symbol table. > > void fred(void) > { > /* body */ > } > > When I load the munched .out file to the target and type lkup "fred" I get the output > > ->lkup "fred" > fred(void) text ("address") > -> > > Which I guess is correct. VxWorks says that fred is a function with argument void. If I now > type "fred" on the shell command line I get the error message "undefined symbol". The only > way I can run fred is to type fred__Fv. Place any function prototype defs that are are not to be mangled within the extern "C" statement. This instructs the compile to treat the function as C code not C++. extern "C" { void fred(void); } +-----------------------------------------------------------------------+ | Graham Waters, Control System Eng Internet: waters@triumf.ca | | TRIUMF --- University of British Columbia, phone: (604)-222-1047 | | Vancouver, BC, Canada Fax: (604)-222-7307 | | http://www.triumf.ca | +-----------------------------------------------------------------------+ --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Non Terminating TCP/IP Sockets Date: 30 Oct 1996 10:35:41 -0700 From: softweyr@xmission.xmission.com (Softweyr LLC) Organization: XMission Internet (801 539 0900) Message-ID: <5583ld$itu@xmission.xmission.com> References: <327503B6.7860@gic.gi.com> Ray Gilbert wrote: : [...] The client opens about 20 : connections to the server, and each connection is handled by a different : handler spawned on the 360. Now, each handler task blocks waiting on : its socket with fioRead. After the server has been up and the client : connected for some hours (exact number not yet determined) and you : shutdown the client, 17-18 of the connects will terminate/close : gracefully, but 2-3 will remain. The handler tasks remain block-waiting : on sockets that are no longer valid. Has anyone else experienced this : problem???? The simplest way to handle this problem is to add an 'I'm done' operation in your protocol and have the client send this before closing the connection. - -- "Where am I, and what am I doing in this handbasket?" Wes Peters Softweyr LLC http://www.xmission.com/~softweyr softweyr@xmission.com - -- "Where am I, and what am I doing in this handbasket?" Wes Peters Softweyr LLC http://www.xmission.com/~softweyr softweyr@xmission.com --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Non Terminating TCP/IP Sockets Date: 30 Oct 1996 10:40:08 -0700 From: softweyr@xmission.xmission.com (Softweyr LLC) Organization: XMission Internet (801 539 0900) Message-ID: <5583to$jit@xmission.xmission.com> References: <327503B6.7860@gic.gi.com> Ray Gilbert wrote: : Hello, : I am currently experiencing a problem with non-terminating tcp/ip : sockets. We are running VxWorks 5.2 on a MC68360 board. The board acts : as a server for a SCO Unix client. The client opens about 20 : connections to the server, and each connection is handled by a different : handler spawned on the 360. Now, each handler task blocks waiting on : its socket with fioRead. After the server has been up and the client : connected for some hours (exact number not yet determined) and you : shutdown the client, 17-18 of the connects will terminate/close : gracefully, but 2-3 will remain. The handler tasks remain block-waiting : on sockets that are no longer valid. Has anyone else experienced this : problem???? : Thanks : Ray - -- "Where am I, and what am I doing in this handbasket?" Wes Peters Softweyr LLC http://www.xmission.com/~softweyr softweyr@xmission.com --------------------------- Newsgroups: misc.jobs.misc,misc.jobs.offered,ba.jobs.offered,biz.jobs.offered,comp.os.vxworks,alt.jobs,comp.lang.ada,comp.arch.embedded,comp.lang.C,comp.lang.C++ Subject: Immediate Job Openings - RI & CT Date: Wed, 30 Oct 1996 23:23:16 -0500 From: Terry Garvey Organization: Analysis & Technology, Inc. Message-ID: <327829B4.22D4@wsii.com> Reply-To: terry_garvey@ima.atinc.com Immediate job openings for professionals with solid experience in the the following areas: Ada, VxWorks, CORBA, Cadre Teamwork, system requirements analysis and design. Long term assignments (3+ years) to be performed in the Connecticut and Rhode Island areas. Check out our web page at www.aati.com. Please email resumes to terry_garvey@ima.atinc.com. - -- --------------------------- Newsgroups: comp.os.vxworks Subject: Re: Tornado on WinNT Date: 31 Oct 1996 03:38:14 GMT From: steved@pacifier.com@199.2.117.163 (Steve Doiel) Organization: Pacifier BBS, Vancouver, Wa. ((360) 693-0325) Message-ID: <5596v7$dgh@news.pacifier.com> References: <557mjj$bti@cobra.Minn.Net> Reply-To: steved@pacifier.com (Steve Doiel) Dave Hansen writes: >We will be beginning an evaluation of VxWorks and Tornado for a new >product we're starting work on. My new computer will arrive with NT >4.0 installed. The Tornado docs specify NT 3.51, and previous >messages in this group have indicated that Wind River is less than >completely supportive of 4.0. So I have a 3.51 CD and the latest >service pack downloaded, and I'm all ready to revert once the system >arrives. > >But the sales guy and the AE from Wind River say this is no longer >true (and in fact never was). Their only caution was to ensure the >FTP server is installed on the Tornado host (this is not the default >for 4.0). So how about it? Has anyone recently been turned away by >Wind River support for having 4.0 installed rather than 3.51? > >Also, are there any other gems of wisdom regarding Tornado on NT out >there? I know UN*X would probably be a better option for various >reasons, but that option is not available to me. (at least I don't >have to try Win95 :) Gem of wisdom. In naming your files, be consistant in the use of case. If you build a file named say Tmodel.c, and link the object as tmodel.o, it will link and be fine for everything except for the debugger. Also, with Tornado 1.0 using the GNU tools, you may only debug one task at a time (a pain). Having worked with other PC based environments (Borland, etc) Tornado is still fairly crude, and at times annoying, but workable. I started out using Tornado on NT 3.51 Service Pak 3, and moved to NT 4.0 with absolutely no problems. BTW, I would recommend >= 40M RAM for working in this environment. A fast Pentium doesn't hurt any either. Hope this helps. --------------------------- Newsgroups: comp.os.vxworks Subject: Re: FTP Between VxWorks front ends Date: Wed, 30 Oct 1996 22:02:36 -0600 From: "Grover C. Burkett, Jr." Organization: American Information Systems, Inc. Message-ID: <327824DC.1DC3@ais.net> References: <32722DFC.1921@LSKMP002.atsc.allied.com> lavergr@LSKMP002.atsc.allied.com wrote: > > We have two ethernet strings of Hp9000 workstations running HP-UX with > VxWorks front ends on each. We're looking for a way to ftp from one FE > to the other. Right now we ftp from FE1 to the workstation, then ftp > from workstation 1 to workstation 2, then ftp from workstation 2 to > FE2. The process works, but is a roundabout way of doing things. Any > suggestions? > > RobIf your able to change your IP router table on the workstations then on workstation 1 have workstation 2 be a gateway for FE2 on workstation 2 have workstation 1 be a gateway for FE1. Then on FE1 running VxWorks perform a routeAdd "(PE2 IP Address)","(Workstation 1 IP Address". On FE2 perform a routeAdd "(PE1 IP Address)","(Workstation 2 IP Address)". This should allow you to ftp from FE1 directly to FE2. Chuck --------------------------- Newsgroups: comp.os.vxworks Subject: PCMCIA ethernet driver for VxWorks Date: Thu, 31 Oct 96 07:29:39 GMT From: jessie@accessone.com (Jessie Chang) Organization: AccessOne Keywords: PCMCIA, ethernet, VxWorks Message-ID: <559kh1$d5a@kanga.accessone.com> Reply-To: jessie@accessone.com I'd appreciated if someone can share some light about where can I find PCMCIA ethernet driver and related information. Thanks, Jessie --------------------------- End of New-News digest ********************** From js@eroch.mc.xerox.com Thu Oct 31 07:11:08 1996 From: js@eroch.mc.xerox.com (Jeffrey R. Szczepanski) Date: Thu Oct 31 07:11:21 PST 1996 Subject: Floppy Driver Device Configuration We are trying to add floppy support to a Power PC BSP. The version of sysLib.c that we have does not include the "fdTypes[]" table to properly configure the floppy driver for DOS support. If your BSP includes the fdTypes[] lookup table in sysLib.c (such as the NEC765 driver on PC's), can you please forward a copy to me. Thanks, Jeff ------------------------------------------------------------------------------ Jeffrey R. Szczepanski Phone: (716) 264-6613 Xerox Engineering Systems FAX: (716) 264-6612 300 Main Street East Rochester, NY 14445 Email: js@eroch.mc.xerox.com ------------------------------------------------------------------------------ From andrew@extremenetworks.com Thu Oct 31 10:11:48 1996 From: Andrew Smith Date: Thu Oct 31 10:12:06 PST 1996 Subject: RE: Tornado on WinNT Just to second this comment: we're attempting to use tornado 1.0 under NT 4.0. We've tried both the 4.0 Beta2 and the real 4.0 and the Tornado 1.0 front end is definitely flaky still. I think it has some memory leaks. The port of the config/make tools is incomplete - some of the tools (e.g. mpp, mdepend) are not ported to NT yet and are not happy using bash as the shell - some POSIX difficulties I think. Also the interaction between tornado and other editors seems primitive and the "editor" (quotes are deliberate) included in Tornado is something of a waste of time. I've heard that interaction with CodeView editor is better. We're still thinking of ditching Tornado/NT and using our Unix server with X unless the flakiness improves. It's definitely still a poor relation to the Unix-hosted version. Andrew Smith ------------------------------------------------------------ Newsgroups: comp.os.vxworks Subject: Re: Tornado on WinNT Date: Wed, 30 Oct 1996 09:23:45 -0800 From: "H.J. Bae" Organization: Peaceful Star Message-ID: <32778F21.4EF5@peacefulstar.oakland.ca.us> References: <557mjj$bti@cobra.Minn.Net> Reply-To: hjb@peacefulstar.oakland.ca.us I use WinNT4.0 and Tornado all the time. It works well. However, I do not use Tornado's IDE/GUI interface. I just use GNU compiler that comes with Tornado. I've found that Tornado GUI has a lot of bugs that makes it unusable under heavy use. Instead, I use emacs to edit and compile the programs, and use command line version of GNU gdb. Just like Unix. From stan@rti.com Thu Oct 31 11:04:44 1996 From: Stan Schneider Date: Thu Oct 31 11:04:47 PST 1996 Subject: Re: Tcl/Tk >> Date: Tue, 29 Oct 1996 08:41:10 +0000 >> From: Doug Owens >> Has Tcl/Tk been ported to vxWorks? Is anyone using it for user interface >> development? Hi Doug, We successfully used Tcl/Tk (with Visicom's X port) on a project earlier this year. It was fairly painless. Contact me offline for details... -- Stan