Reporting directly at the suggestion in the #debian-devel IRC channel.
Some more details in http://bugs.debian.org/573264
Basically, X dies with "Failed to submit batchbuffer: Input/output error" in Xorg.log and with this in dmesg:
[ 4345.716012] [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer elapsed... GPU hung
[ 4345.716027] render error detected, EIR: 0x00000000
[ 4345.716045] [drm:i915_do_wait_request] *ERROR* i915_do_wait_request returns -5 (awaiting 245106 at 245105)
[ 4348.220013] [drm:i915_gem_idle] *ERROR* hardware wedged
[ 4357.031224] [drm:i915_gem_entervt_ioctl] *ERROR* Reenabling wedged hardware, good luck
[ 4357.380008] [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer elapsed... GPU hung
[ 4357.380017] render error detected, EIR: 0x00000000
[ 4357.380673] [drm:i915_do_wait_request] *ERROR* i915_do_wait_request returns -5 (awaiting 245109 at 245105)
[ 4359.236011] [drm:i915_gem_idle] *ERROR* hardware wedged
This is with xserver-xorg-video-intel 2:2.9.1-2 as packaged by Debian and the 2.6.33 (Debian: 2.6.33-1~experimental.2) kernel.
Depending on how I trigger the bug, the errors are slightly different. Enabling compositing (KDE4) or switching to a text console (i915.modeset is off) both certainly trigger it (symptom: screen goes dark); just using X also shows the issue after 20-40min.
I tried newer intel driver (packaged by Debian: 2:2.10.902-1) and newest libdrm (c1c8bbf8), but for lack of time not in all combinations. This exact combination showed a frozen display (instead of a black one) with a "working" mouse pointer. Not sure if that's an improvement ;-)
Sorry for the spotty information/version coverage - I just don't get enough time to work on this as much as I'd like...
This is likely a dup of 26345, the i845 has a severe chipset bug that we've yet to address. If you get the chance, update to a recent kernel, 2.6.34-rc2, and upload the /sys/kernel/debug/dri/0/i915_error_state following a hang.
Without additional info, I'm assuming this is a dup of the nasty 845G bug.
*** This bug has been marked as a duplicate of bug 26345 ***
On Sun, Apr 17, 2011 at 01:01:14 -0700, firstname.lastname@example.org wrote:
> --- Comment #2 from Chris Wilson <email@example.com> 2011-04-17 01:01:14 PDT ---
> Without additional info, I'm assuming this is a dup of the nasty 845G bug.
See http://www.debian.org/News/2011/20110423 fwiw