Sun-Spots-Request@RICE.EDU.UUCP (11/19/87)
SUN-SPOTS DIGEST Wednesday, 18 November 1987 Volume 5 : Issue 61 Today's Topics: Administrivia: change in Bitnet service Re: Problem destroying panels in SunView 3.2 (2) Re: Documentation for dbx symbols Re: XENIX cartridge tapes compatible w/Sun Optimising Servers uudecode problem: the trivial solution Re: fractool demo (2) Help with SunOS 3.4 problems - Thank you Korn Shell on Sun 3s? SCSI data corruption problem on a Sun 3/160 Disabling the cache on a Sun 3/280? Help with Raw IP sockets on Sun 3 Lifetime of Sun B&W monitors? Extra displays? Disk enclosures for 5" disks? ethernet-appletalk connection? ---------------------------------------------------------------------- Date: Wed, 18 Nov 87 14:31:55 CST From: William LeFebvre <phil@Rice.edu> Subject: Administrivia: change in Bitnet service Now don't panic! This is good news, not bad. The people that run the machine which is the Bitnet node here at Rice have kindly set up a Bitnet list server for Sun Spots to use. If all goes as planned, this issue will be the first one sent out to the Bitnet through that list server. If this is successful, it means that Sun Spots digests will no longer be dependent on the doomed WiscVM gateway. Bitnet subscribers should continue to submit articles to the digest by sending them to the ARPANet address "sun-spots@rice.edu". How you get those messages onto the ARPANet is a problem you will have to solve. But the existence of the list server means that Bitnet people will be able to add or remove their names automatically. Unfortunately, I am not familiar with the details of how this is done (just yet, anyway). The list has been set up at the Bitnet node "RICE", and initially contains all the Bitnet addresses in our main list. If you know of any problems caused by this change, send mail to "Sun-Spots-Request@Rice.edu" describing the problem. William LeFebvre Department of Computer Science Rice University <phil@Rice.edu> ------------------------------ Date: Sun, 08 Nov 87 16:45:01 PST From: Vern Paxson <vern%lbl-helios@lbl-rtsg.arpa> Subject: Re: Problem destroying panels in SunView 3.2 (1) [Eric Barber writes about sporadic core dumps when destroying panels in popups. His popups also have text sub-windows.] There's a bug related to showing cursors in text sun-windows in 3.2 which causes random memory trashing. Here's a patch to get rid of it (at the cost of never having your cursors go away): #! /bin/csh # makes the ev_xy_in_view() routine always return EV_XY_ABOVE. This is # necessary because if we let it do its thing, one of its descendants can # wind up doing a bad ft_set(), trashing memory. adb -w $1 <<! _ev_xy_in_view+0x22?w 4e71 ! Hope it helps. Vern Vern Paxson vern@lbl-csam.arpa Real Time Systems ucbvax!lbl-csam.arpa!vern Lawrence Berkeley Laboratory (415) 486-6411 ------------------------------ Date: Mon, 9 Nov 87 15:19:35 CST From: Michael Begeman <begeman@mcc.com> Subject: Re: Problem destroying panels in SunView 3.2 (2) I reported this bug to Sun on January 15, 1987. I also provided a sample program to cause it. On February 13th, I got the following workaround. Please note two things: THIS FIX RELIES ON CALLS TO UNDOCUMENTED INTERNAL SUNVIEW 3.2 FUNCTIONS, and, this was promised to be fixed in release 3.4. I have not verified this. I would suggest upgrading to 3.4 over using this fix! +----------------------------------------------------------------+ | Michael L. Begeman, MCC, 9390 Research Blvd., Austin TX 78759 | | (512)338-3308 begeman@mcc.com | | {gatech,harvard,pyramid,seismo}!ut-sally!im4u!milano!begeman | +----------------------------------------------------------------+ [[ I really wish I could include the sample program, but it's just too big. It is in the archives as "sun-source/textswfix.c" (although the file was originally named blinktimer.c). It can be retrieved via anonymous FTP from the host "titan.rice.edu" or via the archive server. For more information about the archive server, send a mail message containing the word "help" to the address "archive-server@rice.edu". --wnl ]] ------------------------------ Date: Mon, 9 Nov 87 15:47:07 EST From: Don Hopkins <don@brillig.umd.edu> Subject: Re: Documentation for dbx symbols I have two Sun technical reports, "DBX and DBXTool Interfaces," by Steve S. Muchnick, and and "DBXTool, A Window-Based Symbolic Debugger For Sun Workstations," by Evan Adams & Steven S. Muchnick. I think they're just what you're looking for. You can order them by calling Sun at 1-800-821-4643 (in CA: 1-800-821-4642). -Don ------------------------------ Date: Wed, 11 Nov 87 22:45:22 EST From: dyer@spdcc.com (Steve Dyer) Subject: Re: XENIX cartridge tapes compatible w/Sun In-Reply-To: <1987.11.10.12.18.00.695.14582@rice.edu> Both the Archive and Wangtek cartridge tape drives which are supported by SCO in XENIX 286 and XENIX 386 v.2.2 will read and write tapes which can be read on a Sun 3 using /dev/rmt8. Steve Dyer dyer@harvard.harvard.edu dyer@spdcc.COM aka {ihnp4,harvard,linus,ima,bbn,m2c}!spdcc!dyer ------------------------------ Date: Mon, 9 Nov 87 20:44:26 EST From: umix!itivax!lokkur!scs@rutgers.edu (Steve Simmons) Subject: Optimising Servers Pardon me for getting into this discussion a little late, but we've certianly had (and have now) a lot of experience with this stuff: >> Set up a hierarchy of Ethernets using gateway machines. Keep the diskless >> machines on very lightly loaded Ethernet trunks. If you have a lot of >> Suns, put them in clusters and restrict NFS to machines within a cluster. >> Don't export *and* import NFS filesystems on the same machine. > >This is a mixture of a little good advice, some bad advice, and some >unnecessary restrictions. Do cluster diskless nodes, but instead of "a >hierarchy of Ethernets", set up a single hierarchical Ethernet. Put each >cluster on an Ethernet segment and use *bridges*, not gateways, to connect >each segment to a main trunk. Our experience is somewhat different in both directions. We use separate enet segments all tied together on a common delni by gateways. Separate heirarchies are nice, but sometimes the physical constraints of your facility make it impossible. In addition, we have two user groups that don't really need or want to mix (even though their desks do)! So each group gets their own enet seg, and the gateway delni is a pseudo-main. >The important thing is to confine *paging* (ND) traffic to a single >segment, but the emphasis on "very lightly loaded" is overdone. Within a >cluster, the bottleneck is far more likely to be the ND server than the >network. Absolutely. Unfortunately we have found that 3 of our software developers running our *very* large programs along with debuggers, debugging code, many funky bitmaps, etc, etc, etc, drive a server into such a frenzy of swapping/paging that it's essentially swamped. This *will* get better with 4.0, but not enough to really help. We find that in periods of light load 70% of our traffic is ND, and 85-90% under heavy load. >If you don't use gateways, there is no need at all to confine NFS to >machines within a cluster. Nor is there any need not to both import and >export NFS filesystems on a given machine. Do limit any server's imported >filesystems to be read-only, mounted 'soft'. Increasing the timeouts in the mounts might help too, but we've not experimented with this. We've wished for the ability to export certian filesystsems read-only, too. >(Actually, not importing any filesystems on a server could help limit its >use for things other than serving, a big performance plus, but it can be a >pain for the server's SA.) Our solution is much simpler and doesn't hassle the SA as much. We have created 2 netgroups, "everybody" and "sys_staff". In the passwd files on the server the last 2 lines are +sys_staff::.... -everybody::.... Now all the hoi polloi are shut out, but me and the staff can get it. Downside: ~user names are all junked, 'ls -l' gives you uids rather than names. >Our 160 Suns share a single hierarchical Ethernet (about five segments >plus main trunk) with a like number of other machines, including many >mostly-diskless Symbolics Lisp Machines, with NFS cross-mounts all over. >My Sun mounts at least one filesystem from every segment, from Suns, an >Ultrix VAX, and a Lisp Machine. You win for size! We use 140 suns on four ethernet segments and a pseudo- trunk (actually a delni). Lot's of crossmounts, and people go thru the pseudo-trunk to get to our (horrors!) decnet. >Thanks to our network hardware elves, this all works well. Could we import some? :-) Steve Simmmons Schlumberger Technologies, CAD/CAM Division ... ihnp4!itivax!lokkur!scs ------------------------------ From: symcom!maeder@uxc.cso.uiuc.edu Date: Mon, 9 Nov 87 16:26:19 CST Subject: uudecode problem: the trivial solution In a note in v5n56 JON%UCLASTRO.BITNET@wiscvm.wisc.edu described a problem with uudecode. As was correctly diagnosed by the moderator this happens because uudecode is suid uucp and can therefore not create files in the user's directory. Having to chmod +w and -w your directory everytime you use uudecode is not a very good solution however and also poses security problems. I simply made a second copy of uudecode without the bit set and called it 'decode'. Using this instead of uudecode everything works fine. Roman E. Maeder Dept. of Mathematics UUCP: ...!uiucuxc!symcom!maeder Univ. of Illinois at Urbana-Champaign Internet: maeder@symcom.math.uiuc.edu [[ Someone should just figure out if turning off the setuid bit for uudecode would cause any uucp security problems. If it doesn't, then just leave uudecode not setuid. Of course, uucp isn't all that secure to start with.... --wnl ]] ------------------------------ Date: Tue, 10 Nov 87 22:08:27 CST From: steve@spock.ncsa.uiuc.edu (Steve Christensen) Subject: Re: fractool demo (1) In response to Jim Sharpe in v5n58: The "Fractool" you are asking about is called rpcmand. It is a program to compute the Mandelbrot set on many or one machine over a network using RPC's. It was written by someone at Sun and is part of their demo package. It works on Sun-2, Sun-3 and Sun-4 machines. To use it on a Sun-4 certain minor changes need to be make to the floating point optimization flags in the Makefile. As far as I know the only way to obtain it is from Sun. It runs on my Sun-4/260 about 4 times faster than on my Sun3/160 with FPA. It is especially spectacular on the new 19" color monitors that Sun uses on the 4's. I have had it using as many as 10 Suns simultaneously. There is a tool called Fractool that computes fractals rather than the Mandelbrot set. This is also a Sun internal demo I think. I may be that these demos will be on the next Sun User Group tape, but I do not know for sure. Steve Christensen Research Scientist NCSA, Illinois ------------------------------ Date: Tue, 10 Nov 87 22:09:38 PST From: hoptoad!gnu@cgl.ucsf.edu (John Gilmore) Subject: Re: fractool demo (2) The fractool demo was written by Phil Heller of Sun. It demonstrates both color graphics and remote procedure call, by having up to eight machines anywhere on the net compute parts of the Mandelbrot Set. It's especially fun when you have a Convex, Alliant, Cray, or similar machine accessible, but it's tolerable on a single Sun as well. Phil and his management were gracious enough to release it (and some other demos you may have seen at trade shows) to the Sun User Group tape. You can get a copy of it by ordering the new "87.1" User Group tape. We'll have the tapes for sale at the upcoming SUG conference, or you can order by mail -- see the directions and summary of the tape in the next README newsletter. John Gilmore for the Sun User Group tape committee ------------------------------ Date: Tue, 10 Nov 87 09:57:46 EST From: Dan Trinkle <trinkle@purdue.edu> Subject: Help with SunOS 3.4 problems - Thank you I asked for help from Sun-Spots readers on two problems we were having with a couple of our Suns. The first was running out of swap space on Suns (specifically Sun-3/110) with localdisks. The answer appeared in a digest (v5n54) before my question did (talk about response!). The problem is with the Sun kernel not properly setting up swap space. When swap is configured on sd0 and sd1 and only sd0 exists, the kernel thinks it has a huge swap area on the second disk. Changing the line in the configuration file to just "swap on sd0" and generating a new kernel fixed the problem. Thank you to Mark Dionne at Interleaf. The second concerned a Sun-3/50 with an ADAPTEC controller and a Micropolis 1325 disk not botting 3.4. I have had several suggestions (3 email, 1 phone call) that it is a problem with old versions of hardware (old ROM or SCSI controller interface). We are checking into this. This sounds most plausible. If it does not fix the problem, you may hear from me again [:-) Thank you for all the help. Daniel Trinkle trinkle@cs.purdue.edu ARPA Computer Science Department trinkle%purdue.edu@relay.cs.net CSNET Purdue University {ucbvax,decvax,ihnp4}!purdue!trinkle UUCP West Lafayette, IN 47907 (317) 494-7844 PHONE ------------------------------ From: prlb2!marlair@uunet.uu.net (Vincent Marlair) Date: 10 Nov 87 12:43:54 GMT Subject: Korn Shell on Sun 3s? Has anybody been able to run on a Sun 3 the Korn Shell distributed by AT&T without any modification to the code? ------------------------------ Date: Wed, 11 Nov 87 12:40:03 EST From: wmt%sonny@proteon.com (Willie Tsang) Subject: SCSI data corruption problem on a Sun 3/160 Has anyone experienced SCSI data corruption problem when running "foreign" peripheral boards on a 3/160? Particularly with DMA devices that may content with the SCSI controller for the bus? We have two 3/160's that have different revisions SCSI boards: 05-rev B and 07-rev A Does anyone know of a newer (latest and greatest) rev? --- Willie Tsang (Wmt@Proteon.Com) ------------------------------ Date: Wed, 11 Nov 87 12:56:05 EST From: wmt%sonny@proteon.com (Willie Tsang) Subject: Disabling the cache on a Sun 3/280? How can I disable the write-cache on a Sun3/280? I'm writing a device driver for a network device (the ProNET-80) which works fine on a Sun3/160 but I get DMA timed out when running the same driver on the 280. We plugged a MVME133 CPU board onto the bus and ran kadb, right after the DVMA address for a kernel mbuf is acquired from mballoc(), which calls mbsetup(), we can read the mbuf from the MVME133 CPU by querying the contents starting at the DVMA address. Fine. Then when we write to those same locations from the MVME133, the MVME133 reads back updated contents, but kadb still reads the same old values. I discovered that if I pop out to the PROM monitor and disable the cache with the "n d" command, then the MVME133 can write through. But when I resume execution in kadb, I get bus errors and crash. I also tried vac_flush_kpage(), which doesn't solve the problem (although it does set the cache/don't cache flag bit in the page map entry.) --- Willie Tsang (Wmt@Proteon.Com) ------------------------------ Date: 11 Nov 87 17:03:38 GMT From: hope@gatech.edu (Theodore Hope at The ICS Lab) Subject: Help with Raw IP sockets on Sun 3 I am using raw IP sockets (socket (AF_INET, SOCK_RAW, IPPROTO_RAW)) on a Sun 3 (Sun OS 3.3). I use sendto and recvfrom, and everything works fine. However, as you know, everyone on the local net which is reading raw IP will get a copy of every transmitted packet. For now this is fine, as I'm the only one doing this. What I would like to do is use a raw IP socket which is bound to/from a specific IP (internet host) address. It seems like all I would have to do is fill in the destination IP address somewhere in the destination socket parameter ("to" socket, fifth parameter of sendto), but haven't figured out how to do that correctly. In the sendto, I fill in the sa_family of the destination socket ("raw IP") with AF_INET, but don't put anything in sa_data (see below). This works fine, but it's all broadcast (you even read a copy of your own packets). tosock.sa_family = AF_INET; sendto (sock, buf, sizesent, 0, &tosock, sizeof tosock); I've tried putting a destination IP address in sa_data, but it doesn't work. Any help?? Please respond via mail. Thank you very much. Theodore Hope School of Information and Computer Science Georgia Institute of Technology, Atlanta GA 30332 Internet: hope@gatech.edu uucp: ...!gatech!hope ------------------------------ Date: Mon, 9 Nov 87 13:17:02 EST From: Matt Landau <mlandau@diamond.bbn.com> Subject: Lifetime of Sun B&W monitors? Has anyone else experiences the phenomenon in which Sun 3 monochrome displays become almost useless after only about a year of active life? The first symptom is that the upper and lower corners of the display become warped, and anything that's displayed in them becomes unreadable due to the curvature of the image. After that, the corners become fuzzy and out-of-focus. Gradually, the whole screen becomes dim and unfocused. The whole process takes a month or two after the first symptoms appear. This has happened to virtually EVERY ONE of our Sun 3 monitors. Anyone have a clue what's going on here? Is this something we might be able to repair ourselves? The local field service engineers claim they've never heard of this problem, and want to sell us new monitors, which sounds pretty bogus to me. Information, opinions, and advice would be appreciated. Matt Landau Waiting for a flash of enlightenment mlandau@bbn.com in all this blood and thunder ------------------------------ From: hao!stcvax!stc-auts!kak@rutgers.edu (Kris Kugel) Date: 11 Nov 87 20:18:52 GMT Subject: Extra displays? We have developed an application which uses a Sun system as a fancy display, with minor interaction. Since the CPU is relatively unloaded, and we need an additional display, we were wondering, is there some way to add an additional monitor using suntools to a Sun? Is there some other way of reducing costs for an additional display beyond buying a vanilla Sun diskless workstation? Are there other (cheaper) monitors available? Kris A. Kugel Storage Tek: ...{hao,nbires,ihnp4}!stcvax!stc-auts!kak ------------------------------ Date: Wed, 11 Nov 87 11:16:34 -0500 (EST) From: "Anthony A. Datri" <aad+@andrew.cmu.edu> Subject: Disk enclosures for 5" disks? I've got a Maxtor 1140 on a sun3, and I'd like to find a generic enclosure with power supply for it. Actually, I'd like to find an enclosure that'll take two standard 5" drives. any pointers greatly appreciated. anthony a datri scribe systems (aka unilogic) ad0r@tb.cc.cmu.edu ------------------------------ Date: 11 Nov 87 22:11:33 GMT From: dukempd.uucp!gpm@cs.duke.edu (Guy Metcalfe) Subject: ethernet-appletalk connection? My situation is this: I have a data aquisition system consisting of Suns and ibm clones on an Ethernet. On the other side of the building are several Macintae and an Apple laserwriter with document preparation in mind. I would like to be able to use the Suns with the laserwriter. How do I connect the Ethernet and Appletalk together? I need to know hardware, software required and/or names and addresses of companies that make such connection products. I would especially like to hear from someone who has made this connection before. Thanks for any help. Guy Metcalfe despot@tucc.bitnet Duke University Dept. of Physics gpm@dukempd.uucp Durham, N.C. 27706 mcnc!duke!dukempd!gpm [[ This is marginal Sun-Spots material. Please send replies directly to him. --wnl ]] ------------------------------ End of SUN-Spots Digest ***********************