Dear Maintainer,
For a few months (maybe since wheezy has become stable, but I’m not
sure), I have been experiencing a seemingly random bug with X, on my
Dell Latitude 5420: now and then, around once or twice a month, X
freezes except for the mouse cursor. The keyboard is completely unusable
(can’t even switch to console, or restart, only SysRq magic combinations
work), the mouse cursor moves, but no interaction is possible with text,
buttons or windows. At that stage, there is no error message in syslog
nor in Xorg.log.
Still I can login via SSH. Then if I restart the DM `sudo invoke-rc.d
lightdm restart` the display flickers, goes to a black console with a
blinking caret and stop there. Running `ps` shows Xorg and some of my
user X clients are still alive. The Xorg process can be killed with
signal 9 (plain 15 does nothing). Only then do I obtain two lines in
dmesg:
kernel: [46367.563343] [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer elapsed... GPU hung
kernel: [46367.563416] [drm] capturing error event; look for more information in /debug/dri/0/i915_error_state
Then I can start a new session, e.g. `sudo invoke-rc.d lightdm start`,
which gets me back to normal.
The Xorg.log of the problematic session shows a mix of "[mi] EQ
overflow" and backtraces in Xorg.log (see below). I also attach the
contents of /sys/kernel/debug/dri/0/i915_error_state, as requested by
the drm error message.
This might be duplicate of other bugs (especially #680514, #680515 and
#703276), but I prefer you to be judge of that.
Please let me know if you would like more info on this.
Best,
iouri.