[ALUG] How to diagnose this crash on exit from X?

Steve Fosdick lists at pelvoux.nildram.co.uk
Tue Nov 11 18:53:12 GMT 2008


On 11/11/2008 17:54:18, Chris G wrote:

> ...
> It all installed OK without any problems and ran OK.  However since I
> set the graphics resolution to 1600x1200 (the display's default) it
> crashes *totally* on exiting from X.  It's a total, hard, crash -
> CTRL/ALT/Fn won't switch to a text console, I can't ssh into the
> system (which I can before it has crashed) and I can't ping the
> system.
> 
> There is one other symptom, when I try and switch to a text mode
> console (using CTRL/ALT/Fn) even when the system is running OK I get 
> a message from my monitor saying "this mode isn't supported" which 
> seems a little odd on a monitor capable of 1600x1200 at more than 
> 60Hz.

I would take a guess that either X or one of the kernel drives does not 
understand all the registers on the graphics card so the card is not 
being put in a consistent state when X exits or hands the card back to 
the console/framebuffer driver to switch to the text console.  
Depending on what happens next the card is left with the wrong settings 
(i.e. not those for the text console, hence the message) or the kernel 
is fooled into doing something that results in the crash.

I do not know how you would debug this in the sense of knowing exactly 
what is happening.  There will be some info in the X log file but 
obviously once the the kernel is dead it doesn't write to disk anymore.

I has a case of something similar when text mode used the framebuffer 
device and whether or not there was a crash depended on the exact 
settings of that framebuffer mode, i.e. what resolution, depth and 
refresh rate.  I did not try different settings in X to see if it was 
just a certain combination that was problematic because I spend most of 
my time in X and didn't want to compromise on that.  Instead I reverted 
to having the text mode use the old (non framebuffer) console driver 
and that was reliable.  Since then problem seems to have been resolved 
and framebuffer and X co-exist though I do not know if it was an update 
to X or the kernel that fixed it.

Good Luck,
Steve.




More information about the main mailing list