Hi Andrew, thanks for your reply.

Sorry for the delay in replying - I was waiting for it to crash again :)

And it has.

(I tried mem=nopentium but it still locks up).

This time it locked up just after I started a 'find . -name sa1100*' on a
big directory structure. No mouse pointer either. Not sure if this is
relevant tho. It is possible that this is related because the last time it
locked up was when I hit 'tab' with the advanced bash autocompletion
script (and I imagine that does some sort of filesystem search too).

I did an strace -p 241  (241 is the pid of the XFree86 process at this
time). I get this:

--- SIGALRM (Alarm clock) ---
sigreturn()                             = ? (mask now [])
--- SIGALRM (Alarm clock) ---
sigreturn()                             = ? (mask now [])
--- SIGALRM (Alarm clock) ---
sigreturn()                             = ? (mask now [])
--- SIGALRM (Alarm clock) ---
sigreturn()
....
[repeats ad nauseum, and very quickly too]

I also tried ltrace - same syscall over and over again. Nothing else of
interest there it would seem.

Also (let it run for 10 seconds or so):

:~$ sudo strace -c -p 241
% time     seconds  usecs/call     calls    errors syscall
------ ----------- ----------- --------- --------- ----------------
100.00    0.002046           3       590           sigreturn
------ ----------- ----------- --------- --------- ----------------
100.00    0.002046                   590           total


> Next time you get to ssh in to the machine in this state,
> you could try strace'ing the X process, which might give us a clue as to
> where the problem is.
>
> There is a modified version of gdb, which understands XFree86
> driver modules, an old version (for gdb 4.17) at:
>       ftp://ftp.xfree86.org/pub/xpert/gdb/
> and a newer one (for gdb 5.1..) at:
>       ftp://people.redhat.com/mharris/gdb-xfree86/
> You can use the standard gdb, but unless you have a statically
> built server, gdb wont understand the Xfree86 modules, so wont
> help if it is looping inside a driver.

How exactly should I use this?

> Can you try attaching the debug to the looping process, and get some idea
> of where it looping ? If you can get a core dump, try to see where that is
> too, but it sounds as if the processes doesn't die.

Can't get it to die - unless you know of a way of forcing a process to die
and core-dump?

Thanks for the help - any more thoughts?

Cheers,
David.



_______________________________________________
Xpert mailing list
[EMAIL PROTECTED]
http://XFree86.Org/mailman/listinfo/xpert

Reply via email to