Subject: System freezes under X.
To: None <netbsd-help@netbsd.org, port-i386@netbsd.org>
From: Richard Rauch <rkr@rkr.kcnet.com>
List: netbsd-help
Date: 08/01/1999 14:14:05
I'm not sure how general of a problem this is, so I'm posting to both
port-i386 and netbsd-help.

Ever since upgrading to 1.4 in May, I've seen my NetBSD box lock up under
semi-predictable circumstances.  The general requirements seem to be:

  Run X.
  Do something that puts a heavy CPU strain on the system.


For the CPU strain, I can use programs such as SETI@Home (or however one
wants to spell that) and Quake II (the LINUX version).  Apache is a
possible candidate, though I haven't had it up much.

Initially, I thought that it was the Quake II game, but as I said,
SETI@Home will do it.  On the other hand, if I run SETI without starting
X, there's no problem.  Since the SETI client has _no_ graphics support,
I hesitantly conclude that if it can (semi-reliably) crash the system
while the system is running X, there is a serious internal bug in X.

I still assumed that it was a ``nothing'', since I had configured to use
the XF86_S3V X server.  S3V is not documented, and seems to be
deprecated---but IS included in 1.4's distribution.  Recently, I fooled
around with getting the XF86_SVGA server to work (so far only in 8-bit
mode).  Now, to my disappointment, the SVGA server crashes just like the
S3V server.

I cannot seem to get the Quake II program to run without X (perhaps I need
to download some further libraries?), so I can't be certain that Quake II,
like the SETI client, causes X to crash even when the software is
fundamentally sound.


 /~~~ uname -a

NetBSD rkr.kcnet.com 1.4 NetBSD 1.4 (olibGENERIC) #10: Tue Jun 29 01:07:38
CDT 1999     root@rkr.kcnet.com:/usr/src/sys/arch/i386/compile/olibGENERIC
i386

 \___ uname -a


 /~~~ Symptoms

The system locks up tight.  It seems to suspend network communications,
and certainly freezes the X console (even <ctrl>-<alt>-<backspace> won't
do anything).  In one case, a sound sample that was playing (in Quake II)
seemed to get stuck in an endless cycle.

As near as I can tell, the machine is completely frozen at this point, and
the only way out is to cycle the power button on the CPU.

 \___ Symptoms

Has anyone seen similar problems?  This does NOT happen during normal use
(e.g., the daily events at 3am run normally; email is sent & received;
one can telnet or FTP, etc...).  It ONLY seems to happen when one makes
heavy use of the CPU while X is up and running.

(As circumstantial evidence that X has to be running, I have managed to
get one SETI data set processed with just the wscons up.  But with X
running, it locks up (as I recall) within 15 or 30 minutes.)


  "I probably don't know what I'm talking about."  --rkr@rkr.kcnet.com