>I've seen a few of these, and they _all_ seem to be related to module
>use. I simply cannot debug panic's that happen in a module (as you can
>see, the stack trace and EIP are not even found by ksymoops), especially
>as I don't even know _what_ module it is.
>
>(It _looks_ like this might be a mouse module, as it's gpm that crashes,
>but which one?)
>
>I'd love to fix this, but I need more information. When does it seem to
>happen? The above looks like it happened while trying to open the mouse
>(?) device, but it obviously doesn't happen all the time, or your
>machine wouldn't stay up for even one day..
Linus, I have been having similar problems. I just compiled 1.3.73
yesterday before I left work, and left the system running with X going
overnight. After compiling the kernel, I rebooted and then started up xdm
and left the system in that state. I came in this morning and started using
it with essentially one app, 'knews', which seem to run fine for about an
hour or so. Then I was changing news groups and all of a sudden, knews
dissapears! I went back to the xterm window I started knews from and hit
enter which notified me of a Segmentation Fault (which BTW, is the message I
recieve on my system just prior to everything that has been described here
happens). Shortly after that, my mouse stopped working and I tried
CTRL-ALT-F[1..4] in an attempt to get back to my standard console and the
system dies with the typical 'attractive' display of colored ascii
characters/symbols arranged in vertical bars on the screen. At this point I
can't even do the three finger salute to reboot. One thing I would mention
is that yesterday while I was running 1.3.71 (before building 1.3.73) the
same exact thing happened except in this case, I happened to be rlog'd into
another machine on our net. When everything locked up as before and I tried
access my console, I was still able to access the console which was running
my rlogin process to the other system and use it without problems. The
other consoles however all had 'Oops' errors on them (from either hitting
enter or basically trying anything).
>
>For stability, you might try just compiling in the mouse driver instead
>of using a module, but to be frank I'd much rather see the problem
>fixed. And for that I'd need you to test it out a bit more and try to
>find a pattern _when_ it happens, and more information on what your
>system setup is (ie which module, how you load it etc etc)
FWIW, the only drivers I use which aren't compiled in are 'ppp' and the
various dos/msdos/vfat modules and a couple of others like lp, etc. For my
basic kernel however, I compile in my mouse driver, my cdrom driver, NFS
driver, etc. because it is too cumbersome to try and work with the system if
a module load fails on one of these needed drivers.
Steve Ginn
sginn@tfg.com.inter.net