Hi - identifying cause of crash - a how to please

[Michael Powell]

David Southwell wrote:

> Hi
> 
> I have a specific situation which causes a system crash on freebsd 7.2 p3
> amd64 on intel quad core.  Can someone teach me how to trace the cause?
> 
> The crash is repeatable in the following circumstances:
> 
> 
> (a) User logs in
> (b) % startx
> (c) kde4 loads and works the session
> (d) user logs out x session terminates.
> (e) user attempts to start a new x session with:
> (f) % startx
> (g) system crashes immediately
> 
> System requires rebooting to single user mode. Run fsck -y and then go
> multiuser. Whereupon the cycle can be repeated.
> 
> 
> This event did not occur until kde was upgraded to kde 4.3.5.
> 
> The video card is a winfast PX7800GT providing openGL with dual DVI. The
> crash problem was not present before upgrading to 3.4.5 so whilst not
> ruling out the card I am not assuming it is the video card.
> 
> It would be helpful if the procedure could be identified clearly. I will
> post the results available on the web so someone who knows how to
> interpret them could take a look.
> 
> Finally should I be asking this question on another maillist?
> 

This is a good place to start, and if someone knows a better spot it will 
probably be indicated. I can't give you an in depth response, but maybe 
something to start with. The Kernel Debugging section of the Developer's 
Handbook may serve as an introduction to a few basics:

http://www.freebsd.org/doc/en_US.ISO8859-1/books/developers-handbook/

You can also start right away by examining the .xsession-errors file you 
will find in the users home directory. Also, in /var/log look for the 
Xorg.0.log and Xorg.0.log.old. After restarting from a crash, if you startx 
the first one will only contain info on that startup; the second one (.old) 
may contain some info on what happened at the time of the previous crash, if 
it is indeed the X server crashing. Sometimes any errors to stdout may 
appear in /var/log/messages. 

Sometimes there may be informative error messages present which can be 
Googled, and other times nothing useful. The same with the .xsession-errors 
file - it usually contains information relevant to the applications which 
were running on top of X.

If you are running the nvidia driver and see some evidence that it may be 
responsible for the crash, you could try substituting the nv driver as a 
test. This might help isolate the problem to the nvidia driver. But where to 
go from here is a good question. Sometimes if there is some problem wrt the 
nvidia driver after some kind of upgrade doing a make, make deinstall, and 
make reinstall of the nvidia driver port occasionally fixes something, but 
this is a long shot.

Nvidia also has a web forum you might hunt around in. Also, if you are doing 
startx to start KDE with a .xinitrc file in the users home dir (with 
startkde in here), try it without so the default TWM window manager comes up 
instead of KDE. Then do the restart test and see if it crashes.  This is a 
good way to separate the problem from being X related and/or KDE related.

-Mike