Sun-Spots-Request@RICE.EDU (Vicky Riffle) (09/04/87)
SUN-SPOTS DIGEST Friday, 4 September 1987 Volume 5 : Issue 39 Today's Topics: Re: What does "text: table is full" mean (3) Re: Sun 3.4 problems Modems and Security VME disk controllers for Suns RCS and compatible make for Suns Re: Root non-security on Sun workstations 800 numbers from the UK Bug in SunView 3.4 ASPLOS-II advance program Need advice on multiple-disk server Modem wont hang up Suns and high speed modems? Sun-3 executables on a Sun-4? shelltool .vs. parity bit? Ciprico VMEbus controllers? Public Domain version of SPICE for Sun3? ---------------------------------------------------------------------- Date: Wed, 2 Sep 87 17:24:11 EDT From: bzs@bu-cs.bu.edu (Barry Shein) Subject: Re: What does "text: table is full" mean (1) There are two approaches and a little bit of advice: 1. Just increase MAXUSERS in the appropriate /sys/conf config file, build a new kernel and re-boot. If your users use a lot of processes and you're not horribly short on memory I'd try double or at least 150% of the actual number of users logged on. This will increase the size of the text table (the number of pure procedures simultaneous being run). The daemons etc tend to eat up quite a few slots. 2. Edit /sys/conf/param.c, change the initialization for 'int ntext'. As opposed to the above this will increase the text table size w/o increasing the size of everything else which is calculated from MAXUSERS. This may or may not be a losing battle (ie. you might soon run out of something else, I'd try (1) first, that's all I ever do). Advice: You can inadvertantly build Sun3.4 kernels that will immediately panic with "panic: sys pt too small" (something like that) so use the standard precautions (copy the old vmunix to /ovmunix before installing a new one). If this happens you almost certainly have copied lots of devices into your config file that you're not using, clean it up and re-build and you should be ok. -Barry Shein, Boston University ------------------------------ Date: Wed, 02 Sep 87 21:14:30 PDT From: Tim Ehrhart <ehrhart@spam.istc.sri.com> Subject: Re: What does "text: table is full" mean (2) I have a fix for this we have been using since 1.4. With all the stuff getting spun of by inetd for networking we quickly blew out our text table and got the same message. The text table is the mechanism that keeps track of the text portions (i.e. code) of executables being stored in the swap area. It creates an entry for every UNIQUE binary being run. That is, if you have 10 csh processes running there is still only ONE text table entry being used. But with all the different diverse processes being run on a multi-user networked UNIX box you can exhaust the table real quick. I have modified /sys/conf/param.c to make the table size larger when I grind out a kernel. Sun's default parameter is to use a hard coded 24 + MAXUSERS (defined in /sys/conf/YOUR_MACHINE). Most folks set MAXUSERS to 4 or 8. In this case I feel the increase of MAXUSERS should cause some kind of multiplitive increase, not addition. Here are the context diffs of /sys/conf/param.c and /sys/param.c.orig 50,54d49 < #ifdef INET < #define NETSLOP 40 < #else < #define NETSLOP 0 < #endif 57c52 < int ntext = 24 + MAXUSERS + NETSLOP; --- > int ntext = 24 + MAXUSERS; Basically, if I have defined INET in /sys/conf/MY_MACHINE file I bump ntext by NETSLOP (couldn't think of a better name). This sets it to a much more realistic value. This file IS for you to put your site specific system configuration parameters into. Don't flame Sun, these were only ballpark figures meant to be tuned. I say this because I've told other folks this and they were scared Sun wouldn't "support" this kernel configuration because param.c had be changed, God forbid. Tim Ehrhart Information Sciences and Technology Center SRI International [[ Thanks to Dick St.Peters at ge-crd for a similar suggestion. He also recommends the command "pstat -T" to check on kernel table usage. --wnl ]] ------------------------------ Date: Thu, 03 Sep 87 10:33:17 -0700 From: Craig Rolandelli <craig@rome.uci.edu> Subject: Re: What does "text: table is full" mean (3) It has nothing to do with memory, you could have gigabytes of memory and still have the same problem. The problem is that you have maxusers set to small. Sun distributes the kernal with maxusers set to 4. We have have changed this to 32. But that's not all, by raising maxusers to 32, you will exercise another bug in the Sun kernel and your system will panic when it tries to boot. Sun has a patch for the kernel, just call your Sun rep. and ask for the maxusers.bzero patch. The maxusers variable can be changed in two places. 1) The configuration file in /usr/sys/conf 2) The file param.c in /usr/sys/conf Currently param.c sets the number of text tables (variable name "ntext" in param.c) to 24 + maxusers. You could just change the 24 to 52, but its easier to change the configuration file. Craig Rolandelli (craig@ics.uci.edu, craig!ucivax!ucbvax) Computer Science Department University of California, Irvine Irvine Ca. 92717 (714) 856-4222 ------------------------------ Date: 3 Sep 87 21:18:53 GMT From: jmr@philabs.philips.com (Joanne Mannarino) Subject: Re: Sun 3.4 problems Thanks for all your responses to my 3.4 problem. We have finally found a solution. The 3.4 release introduced subnetting. The way the networking software for 3.4 works is that the diskless client comes up and requests a subnet mask. It waits for a response from the server. The TCP/IP software (we're running XLAN but problems have occured with Wollongong also) responds with a subnet mask that is invalid. The diskless client is supposed to check it, determine that it is invalid, disregard it, and wait for the server to respond with the correct subnet mask. BUT with the 3.4 release, the client wasn't ignoring the invalid subnet mask so thus it would hang. Sun has provided a patch to the kernel software to take care of this. If anyone else is still having this problem, you should call Sun support and ask for the fix to bug id 1006127. This solution was sent to me by Bill Nowicki, Network Software Developer at Sun. According to him, they have a guard against this in release 3.5. -Joanne Mannarino joanne mannarino uunet!philabs!jmr philips laboratories or (914)945-6008 jmr@philabs.philips.com ------------------------------ Date: 2 Sep 87 12:24:00 MST From: tjsober@sandia-2.arpa Subject: Modems and Security In response to ALDER1@BRANDEIS.BITNET: The problem you describe should not occur if the kernel has been modified to enable hardware carrier detect (see chapter 5 Adding Hardware to your System in System Administration Manual). By changing the flags on the serial port, a logout is forced when the modem notifies the computer that the carrier is lost (via the DTR line, which should be set on the modem for hardware DTR). Also, the carrier should be set to follow the remote system. This will eliminate a large number of unauthorized logins, but is not 100% secure. To increase security, we have two different systems. The first is a Cermetek Security Modem. The Cermetek requests a password from each caller and then grants access or hangs up and dials a preprogrammed callback number. This system has been the most reliable, although it only operates up to 1200 baud. The second system uses a secure front end made by Tact Technology, called a Computer Sentry. The Computer Sentry functions in a similar manner to the Cermetek, but instead the codes are numeric and are entered via a touch-tone phone. The Computer Sentry also logs all calls to a printer and has a variety of shutdown modes if an excessive number of failed attempts occur. It will also work at any baud rate. Now for the bad news. Although the Computer Sentry works, and we are using it, it is a marginal piece of equipment. One example is that it doesn't reset the date at the end of the month (did you know there are 33 days in february?). Occasionally, it just goes off in left field and dies. When talking to the company, I had to explain to their applications engineer how their system worked. The only reasons we still use it is because of the higher baud rate and a single modem can be used for dialup and dialout, which can't be done easily with the Cermetek. ------------------------------ Date: 02 Sep 87 18:13:08 CDT (Wed) From: Todd Nugent <nugent@anubis.uchicago.edu> Subject: VME disk controllers for Suns > A Xylogics salesman told me that his company has, and that Sun will later > this year release, a 32 bit disk controller for the VME bus. He says that > the current 16 bit controller is just a Multibus-VME converter. The rumor I have heard is that Sun has contracted Xylogics to make a Eurocard version of the Xylogics 752 VME (32 bit) disk controller which can be used on the Suns without an adapter. This will be sold exclusively through Sun. If you buy your own Xylogics 752 card, you will also need to buy a VME-to-"triple height extended Eurocard VME" adapter. I will be interested to see if Sun prices their VME controller so high that it is cheaper to buy the xy752 with an adapter. Todd Nugent - Univ of Chicago ------------------------------ Date: Thu, 3 Sep 87 11:44:06 PDT From: Hibbert.pa@xerox.com Subject: RCS and compatible make for Suns RCS can be retrieved on the arpa internet by anonymous ftp (give "anonymous" as your username, and your real name as a password) to purdue.edu. There is also a version of make there that knows about RCS. It's incompletely documented, but there are builtin rules for ".CO" and ".CLEANUP" that allow you to change respectively the way files are checked out (telling where the RCS directory is for instance), and the way checked out files are deleted after everything is built. Both are in the directory /pub/RCS, and they're available in tar and flar format. In RCS I had to edit files.c by changing "#ifdef vax" to "#if defined(vax) || defined(sun)" in two places. It looks like suns and vaxen have similar structures defined in a.out.h. The standard version of make on our vax (running 4.3BSD) has this fix in it. I suspect it came that way from Berkeley. This fix may have made it into the source, since I sent Walter Tichy a mesage about it. I also installed a change in rcs.c and rcsbase.h that I apparently got in sun-spots. It defines a macro STRCMP that fixes an incompatibility in strcmp. The macro to be added to rcsbase.h is: #define STRCMP(a,b) ((a)? ( (b) ? strcmp((a),(b)) : 1 )\ : ( (b) ? -1 : 0) ) in rcs.c I changed line 993 to use the macro rather than the function. It's the first use of strcmp in breaklock(). I don't know why all the uses of strcmp in rcs.c aren't changed, but that's what I have in my code, and it runs just fine. I also had to make a change in make. Apparently someone is using a homebrew replacement for rm that takes a -n flag that means "don't squirrel away a copy of the file before deleting it so I can find it if I decide I didn't really mean to delete the file." The new make has a RMFLAG definition of -fn builtin that has to be changed to -f in order for the automatic checkout to not blow up during cleanup. This change is in rules.c, and I haven't told Walter about it yet, so you'll definitely have to fix it yourself. Chris P.S.: I don't know anything about availability to sites that can't get to the internet. Can someone who does send a followup to sun-spots? [[ I will se what can be done about placing a copy in the sun-spots archives. This would make it available to non-Internet people. --wnl ]] ------------------------------ Date: Thu, 3 Sep 87 16:11:20 EDT From: Guy Middleton <gamiddleton%orchid.waterloo.edu@relay.cs.net> Subject: Re: Root non-security on Sun workstations It seems a few people want clients that won't come up single-user. This is how we did it; the following are changes to /etc/init. If the machine is a client, and if /.secure exists, then init will go straight to multiuser. [[ The entire diff is available from the host "titan.rice.edu" via anonymous FTP as the name "sun-source/init.c.diff". Those without Internet access can obtain a copy by sending the request "send sun-source init.c.diff" to the address "archive-server@rice.edu". For further information on the use of the archive server, send it the word "help". --wnl ]] ------------------------------ Date: Fri, 4 Sep 87 18:18:11 -0100 From: Jeff Dalton <jeff%aiva.edinburgh.ac.uk@cs.ucl.ac.uk> Subject: 800 numbers from the UK >> ([The NIC] user assistance number is 800-235-3155.) > That number doesn't seem to work from Scotland :-) > [[ It might if you put the appropriate international dialing prefix in > front of it! --wnl ]] Unfortunately, the problem is the 800 rather than the missing prefix. As far as I have been able to determine, these numbers are for use only inside the US, and it is not possible to call then from Scotland even with the assistance of the international operator. Jeff Dalton, JANET: J.Dalton@uk.ac.ed AI Applications Institute, ARPA: J.Dalton%uk.ac.ed@nss.cs.ucl.ac.uk Edinburgh University. UUCP: ...!ukc!ed.ac.uk!J.Dalton [[ I did not know this at the time. I naturally assumed that any number, including 800 numbers, would be accessible via 01+ dialing. Oh well. --wnl ]] ------------------------------ Date: Thu, 3 Sep 87 10:47:27 CDT From: Michael Begeman <begeman@mcc.com> Subject: Bug in SunView 3.4 SunView 3.4 introduces a window_create() bug: In 3.2, you could create a TEXTSW with a particular font by specifying: textsw = window_create(base_frame, TEXTSW, WIN_FONT, my_font, 0); In 3.4, the window_create() call apparently ignores the WIN_FONT attribute and the textsw will be assigned the DEFAULT_FONT. A workaround for 3.4 is to split the assignment of WIN_FONT out of the window_create() call as follows: textsw = window_create(base_frame, TEXTSW, 0); window_set(textsw, WIN_FONT, my_font, 0); Note that in 3.2 one had to split the WIN_COLUMNS attribute out of the window_create() when assigning a non-DEFAULT_FONT to the TEXTSW. At least we're moving to a point of consistency: now *all* of the font stuff has to be outside of the window_create() for it to work! Sheez.... [[ Did you send this bug report to the "hotline"? If not, you should. --wnl ]] 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 ------------------------------ Date: 2 Sep 87 18:18:13 GMT From: mfreeman@cascade.stanford.edu (Martin Freeman) Subject: ASPLOS-II advance program SECOND INTERNATIONAL CONFERENCE ON ARCHITECTURAL SUPPORT FOR PROGRAMMING LANGUAGES AND OPERATING SYSTEMS (ASPLOS-II) Palo Alto, California, October 5-8, 1987 Sponsored by ACM and Computer Society of the IEEE CONFERENCE CHAIRMEN General: Martin Freeman, Stanford University & Philips Research Labs Program: Randy Katz, U.C. Berkeley Finance: Dennis Reinhardt, DAIR Computer Systems Publicity: Jim Flournoy, Consultant [[ The entire program is must too large to include here (even without the registration forms at the end). It is available in the archives as "sun-spots/asplos2". Archive-server users can use the request "send public asplos2". --wnl ]] ------------------------------ Date: Tue, 1 Sep 87 16:10:18 PDT From: ultra!wayne@ames.arpa (Wayne Hathaway) Subject: Need advice on multiple-disk server We have eight diskless Sun 3/50's on an Ethernet with one Sun 3/280 server. Currently the server has one 600 MB Eagle, and we have partitioned it pretty much as suggested by Sun. We are acquiring a second 600 MB Eagle, and were wondering if anybody out there had any particular recommendations on what to put where on what disks? One obvious idea is to try to "load share" by balancing the total activity on the two disks (although we have no idea how to MEASURE such activity). Another (unrelated) thought is to move the clients' swap spaces to the new disk. Since this is undoubtedly old hat to many of you, we'd appreciate any advice anybody can share. And by the way, our job mix is pretty much program and driver development in C and (non-Sun) assembler: editing, compiling, kernel makes, debug. No huge programs, no AI, no LISP. Oh, and an occasional iconedit to sweep away cobwebs ... AdTHANKSvance! Wayne Hathaway ultra!wayne@Ames.ARPA Ultra Network Technologies 2140 Bering drive with a domain server: San Jose, CA 95131 wayne@Ultra.COM 408-922-0100 ------------------------------ Date: 3 Sep 87 19:45:18 GMT From: jwabik@ub.d.umn.edu (Jeff Wabik) Subject: Modem wont hang up I need help from you Sytem Administrator types out there: I just installed a modem on my 3/160. I use the modem for dial-ins, dial-outs (tip, kermit), and UUCP. In all cases, I am unable to get the getty (or whomever is responsbile) to drop DTR to the modem, inhibiting the modem from hanging up properly. Here are cases that I've noticed: 1) On dial in, when I "logout", rather than hanging up, I get another "login:". 2) On dial-out, "~." from tip gives the appropriate "disconnecting" message, but fails to drop DTR and my modem interprets the dial-tone to be another hosts till I manually hang it up. 3) On UUCP, when uucico fails, again, the line never gets dropped thus inhibiting any more outcalls till the next day when I get to the system to recycle the modem. I've been fortunate in that the systems I UUCP to and from are smart enough to hang up at the end of a conversation, thus giving me a 50% sucess ratio to date. Bottom line is, HOW DO I MAKE THE SUN DROP DTR? I've got my modem configured to depend on DTR, and have made all appropriate changes to /usr/sys/conf/<SYSTEM> that the "adding hardware to your system" section of the Administrator's manual mentions. -Jeff P.S. My async connection is to the "ttya" port on the CPU board. ------------------------------ Date: Mon, 31 Aug 87 19:44:36 -0400 From: mayer.wbst@xerox.com Subject: Suns and high speed modems? Does anyone have any experience running a 9600 baud modem off of a Sun 3/60 or 3/110 workstation? I noticed that the serial ports on these (and other) Sun products are rated at 9,600 baud (input) and 19,200 baud (output). I have heard rumors (ONLY RUMORS... I HAVE NO PERSONAL EXPERIENCE WITH THIS) that they don't handle input over 2,400 baud very well. Is this true, and is there any way around the problem? What are the failure modes? For my application, I expect to be generating and receiving data as fast as the modem can handle it. Also, is there any way to receive data at 19,200 baud? Some of the modems on the market use data compression to get better than 9600 baud performance on a clean line. Note that the 3/60 has no slots, and the 3/110 has virtually no slots, so buying an "asynchronous line multiplexor" is not an option. -- Jim Mayer (mayer.wbst@xerox.com) ------------------------------ Date: Wed, 2 Sep 87 15:52:55 EST From: Steve M. Burinsky <smb@mimsy.umd.edu> Subject: Sun-3 executables on a Sun-4? Does anyone know if there is any chance of running Sun-3 executables on a Sun-4? Or am I going to have to upgrade all the commercial software I have? Steve M. Burinsky smb@mimsy.umd.edu [[ I seriously doubt that Sun-3 executables will run on a 4, since the architectures are radically different. Sun would have to provide a 68020 emulator that would run on a 4, and emulators like that are hard to get 100% correct. I would welcome comments from those who know differently. --wnl ]] ------------------------------ Date: 3 September 1987 1050 PDT (Thursday) From: bierma@nprdc.arpa (Larry Bierma) Subject: shelltool .vs. parity bit? Does anyone know a way to convince shelltool (when in raw mode) not to throw away characters with the 8th-bit set. --Larry ARPA: bierma@nprdc.arpa UUCP: {decvax,ucbvax,ihnp4}!sdcsvax!sdics!nprdc!bierma PSTN: (619) 225-2161 ------------------------------ Date: 3 Sep 87 17:33 +0500 From: naren <naren%systems.carleton.cdn%ubc.csnet@relay.cs.net> Subject: Ciprico VMEbus controllers? A local rep is trying to sell us a Rimfire 3200 VMEbus controller from Ciprico for our sun-3/160. They seem to provide the boot rom and the driver also. Anyone have any experience with this controller. A comparision with Xylogics 451? Also, has anyway tried using a smd disk drive on a sun-3/140. Thanks Narendra Mehta, Systems & Computer Eng. Department, Carleton University, Ottawa, Canada UUCP: {allegra,decvax,ihnp4,linus,utzoo}!watmath!clan!naren CDN: naren@systems.carleton.cdn ARPA: naren%systems.carleton.cdn%ubc.csnet@csnet-relay.arpa CSNET: naren%systems.carleton.cdn@ubc.csnet ------------------------------ Date: Fri, 4 Sep 87 08:45:25 EDT From: Comer Duncan <duncan%andy.bgsu.edu@relay.cs.net> Subject: Public Domain version of SPICE for Sun3? Is there a public domain version of SPICE which will run on the Sun3 machines and hopefully also runs in the Sun windows environment? The versions one finds in Catalyst seem beyond the means of most university department operating budgets. Comer Duncan Department of Physics and Astronomy Bowling Green State University Bowling Green, OH 43403 CSNET: duncan@bgsu.edu ATT: (419) 372-8108 ------------------------------ End of SUN-Spots Digest ***********************