[ltp] User processes disappearing while idle

Richard Neill linux-thinkpad@linux-thinkpad.org
Sun, 05 Jun 2011 01:35:08 +0100


On 05/06/11 00:33, Jeffrey L. Taylor wrote
>
> Jun  4 09:59:01 viajero kernel: [73516.813960] [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer elapsed... GPU hung
> Jun  4 09:59:01 viajero kernel: [73516.815783] [drm:i915_do_wait_request] *ERROR* i915_do_wait_request returns -11 (awaiting 2449443 at 2449439, next 2449444)
> Jun  4 09:59:01 viajero kernel: [73516.815903] [drm:i915_reset] *ERROR* Failed to reset chip.
> Jun  4 10:01:01 viajero /USR/SBIN/CRON[20684]: (jeff) CMD (cd /home/jeff/Rails/amethyst2production; ruby enqueue_refresh.rb)
> Jun  4 10:01:47 viajero polkitd(authority=local): Unregistered Authentication Agent for unix-session:/org/freedesktop/ConsoleKit/Session1 (system bus name :1.30, object path /org/kde/PolicyKit1/AuthenticationAgent, locale en_US.UTF-8) (disconnected from bus)
> Jun  4 10:01:47 viajero kdm[1175]: X server for display :0 terminated unexpectedly
>
> The first line is always present before a hang.  The next two are frequently
> present, but not always.  The last line (X server for display :0 terminated
> unexpectedly) is frequently but not always present.
>
> Googling the error message reveals that it appears to be an intermittent bug
> in the kernel and/or userspace xorg-drv-intel component that comes and goes
> with kernel versions.  It occurs in Fedora and Arch, as well as OpenSuSE, and
> probably others.  Lenovo T520 is mentioned by name, at least once.  The Dell
> Latitude E6320 is also.  AFAICT, it is a Sandybridge problem.

I'm not convinced that this is the real cause of your problem - because 
a GPU error shouldn't make the other non-X processes crash.

However, you should be able to get a temporary fix by forcing the X-org 
driver to VESA - save your /etc/X11/xorg.conf (if it exists), and 
recreate it (or use your distro's X-configuration tool). You'll loose 
accelerated graphics, but the machine should still function pretty well 
- at least long enough to discover whether this is really the bug.
[Tip: use the space-bar for vertical jump-scrolling; it's quicker]

Richard