Date: Mon, 26 Jul 1999 18:58:11 +0930 From: Greg Lehey <grog@lemis.com> To: Biju Susmer <bee@wipinfo.soft.net> Cc: FreeBSD Questions <questions@FreeBSD.org> Subject: Re: "3.2-RELEASE #0" crashed Message-ID: <19990726185811.Y51019@freebie.lemis.com> In-Reply-To: <000e01bed73a$b76b7360$88291fac@wipro.tcpn.com>; from Biju Susmer on Mon, Jul 26, 1999 at 01:14:38PM %2B0530 References: <19990726155307.U51019@freebie.lemis.com> <000e01bed73a$b76b7360$88291fac@wipro.tcpn.com>
next in thread | previous in thread | raw e-mail | index | archive | help
[Format recovered--see http://www.lemis.com/email/email-format.html] On Monday, 26 July 1999 at 13:14:38 +0530, Biju Susmer wrote: > Thanks for the info. I think it is deep... Don't want to jump in to this ;) read > on.. > > >> [Format recovered--see http://www.lemis.com/email/email-format.html] >> [Moved to FreeBSD-questions; there's no evidence of a bug here] > > OK, I too am a developer. They should have the last word in these cases. Others may agree with you. We'll see if they do. But I believe this message belongs on -questions, and I'm posting this followup there. > On Monday, 26 July 1999 at 9:13:45 +0530, Biju Susmer wrote: >> Hi, >> >> The following program caused my system to crash and reboot. I was >> doing a stress test on my machine by keeping it 200% busy... My uid is >> 1001 [not root ;) ]. I attached all details which i think may be useful. >> The system has swap of 37MB on /dev/wd0s2b. >> >> Category: kern >> Sev: serious >> Pri: High >> Class: sw_bug > >> It would really help if you sent this message in a legible format. >> It's very difficult to read this. > > ?? what went wrong? The first message had overlong lines, such as many Microsoft-based mailers generate. This one omitted the quote of the original text, which makes it very difficult to read. See the web page for a discussion. >>> 3 out of 6 times it rebooted. In all cases, the message "swap_pager: >>> out of swap space" was repeated at least 7 times. The messages >>> generated were different. >>> >>> * first time i could not catch (un-expected) >>> * second and third time it showed some thing, and gave me 15 >>> seconds, but i could not do anything. I remember it is some thing "de >>> allocated many times" or so. PLEASE TELL ME HOW CAN I CATCH THE >>> MESSAGES IF I WANT TO GIVE IT TO U!! >> >> You can't in this point. You need to enable dumping. Set dumpdev >> in /etc/rc.conf to the name of a swap partition which is at least the >> size of physical memory. For example, >> >> dumpdev=/dev/wd0b >> >> This will cause the system to write the dump when it crashes. In >> order to enable it, you need a debug kernel. See the handbook for >> further details. > > I was doing that :) Your sysctl output says that you weren't. >>> The rest 3 times, it went to some loop, i believe. Nothing >>> worked. The chars i typed were echod on the terminal in which i >>> started the program. Also switching screens (v0-v2) were fine, but >>> nothing echod on those screens. Intr or quit char didn't have any >>> effect. at last after one minute i resetted. >>> >>> i tried changing kern.securelevel to 2 but that also didn't help. >>> >>> >>> $ cat crash.c >>> /* program to create "out of CPU" error */ >>> #include<sys/types.h> >>> #include<unistd.h> >>> >>> main() >>> { >>> while(1) >>> if(fork()< 0) /* multiply... */ >>> while(1); /* if no more processes, eat up CPU */ >>> } >> >> It's interesting that you should have run out of swap before you ran >> out of processes, but basically I don't think what you're doing is a >> reasonable use of the system. No UNIX system handles this situation >> well, and FreeBSD is no exception. Since you supply a lot of >> information, but nothing on swap usage, I suspect that your swap >> partitions are too small. > > Point taken. I said may swap is 37MB and it went out of > swap... Yes, that's too small for what you're trying to do. FreeBSD's VM system uses a lot of swap, more than you may think is necessary. This enables it to deliver better performance, and swap is cheap. I run 450 MB swap on my machine. Currently it's using 63 MB of it. > I was not able to kill the processes to bring back my system. That > is what i need. I don't want to reboot! I object the statement about > UNIX. Linux handled this perfectly. I suspect that Linux ran out of processes before it ran out of swap. But if you run out of swap with Linux, you'll have the same problems. I've noticed that Linux uses much less swap than FreeBSD. Don't assume that this is always an advantage. > I'll try with more swap space one day, probably. But i still feel > this is a kernel problem. Yes, it's a kernel problem. I did say " No UNIX system handles this situation well, and FreeBSD is no exception.". If you can come up with a sure-fire system which handles all swap depletion problems gracefully, people will kiss your feet. In the meantime, there's a workaround, and I'd suggest you use it. Greg -- When replying to this message, please copy the original recipients. For more information, see http://www.lemis.com/questions.html See complete headers for address, home page and phone numbers finger grog@lemis.com for PGP public key To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19990726185811.Y51019>