Bug#661253: nvidia-glx: NVIDIA 295.20 causes random and unexpected X termination and hence logout

[Andreas Beckmann]

On 2012-02-28 11:57, Julian Gilbey wrote:
> On Sat, Feb 25, 2012 at 03:21:47PM +0000, Russel Winder wrote:
>> Using Debian Unstable fully up to date had been, up to the 290.xxx -> 295.20 upgrade, entirely fine.  The
>> 295.20 package set seems to cause, apparently at random, X to terminate causing a logout. This renders the
>> machine completely unusable since it cannot be predicted when there will be a logout.  I suspect though it
>> is mainly when rendering images rather than things to do with the GUI.  I haven't experimented extensively
>> though as I am having to use other machines so as to achieve productive things.
> 
> My experience is even more severe: after X dies, it causes the machine
> to basically hang - it ceases to respond to ssh connection requests or
> http connection requests, although pings still work; the keyboard
> becomes unresponsive, and so on.  I have been forced to do a hard
> reboot twice in two days.

Hi Russel, hi Julian,

did you have a look at the nvidia forum to see whether this probelm has
been reported there?
http://www.nvnews.net/vbulletin/forumdisplay.php?f=14&order=desc
There is also an issue with 295.20 segfaulting in libnvidia-tls:
http://bugs.debian.org/660189 - does this math your problem?

Russel, can you confirm that the machine becomes unresponsive after this
crash? Or can you just relogin and continue to work?

Julian, can you setup a serial console or netconsole to capture the
kernel messages in case of such a crash?
But first ensure that you have matching versions of the kernel and the
corresponding kernel headers installed (matching also all the the debian
revision part) and rebuild the module (and reboot).


Andreas