From owner-freebsd-questions  Sun Nov  2 16:53:23 1997
Return-Path: <owner-freebsd-questions>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.7/8.8.7) id QAA22293
          for questions-outgoing; Sun, 2 Nov 1997 16:53:23 -0800 (PST)
          (envelope-from owner-freebsd-questions)
Received: from word.smith.net.au (vh1.gsoft.com.au [203.38.152.122])
          by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id QAA22259;
          Sun, 2 Nov 1997 16:53:05 -0800 (PST)
          (envelope-from mike@word.smith.net.au)
Received: from word.smith.net.au (localhost.gsoft.com.au [127.0.0.1])
	by word.smith.net.au (8.8.7/8.8.5) with ESMTP id LAA00564;
	Mon, 3 Nov 1997 11:15:32 +1030 (CST)
Message-Id: <199711030045.LAA00564@word.smith.net.au>
X-Mailer: exmh version 2.0zeta 7/24/97
To: Donald Burr <dburr@POBoxes.com>
cc: FreeBSD Questions <freebsd-questions@FreeBSD.ORG>,
        FreeBSD Hardware <freebsd-hardware@FreeBSD.ORG>
Subject: Re: programs dying with SIGBUS after long uptime 
In-reply-to: Your message of "Sun, 02 Nov 1997 03:33:21 -0800."
             <XFMail.971102034223.dburr@POBoxes.com> 
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Date: Mon, 03 Nov 1997 11:15:32 +1030
From: Mike Smith <mike@smith.net.au>
Sender: owner-freebsd-questions@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk


> After the system has been up and in use for a lengthy period of
> time (from several hours to a whole day or more), programs start dying with
> signal 10's (SIBGUS).

You should be checking that none of your system components (eg. CPU, 
disk) are overheating.  If you have a fan-cooled processor, you should 
definitely check that the fan is still working.

Then you should check the output of 'vmstat -m', and look for 
abnormally large allocations.

> Many programs, however, still work fine.  For example, right now, I CANNOT
> start StarOffice or Executor (SIGBUS), but I CAN compile WINE (a rather
> large package, IMHO), use Netscape (the 3.04Gold BSDi version), and read
> and compose mail using XFMail.

Note that once a text image has been corrupted during execution, 
repeated execution of the same image will run the (corrupted) sticky 
copy in core until same is flushed; this basically means that once a 
program has died due to memory corruption you need to reboot.

> I've tried reloading them from scratch, with similar results.  Could my
> memory or CPU be going bad, or possibly overheating?  (it has been very
...
> documentation)  Could this be due to some BIOS configurations that aren't
> set right (memory timings, cache timings, PCI bus stuff, ...?)

Yes and yes.

mike