Bug 59941

Summary:	Frequent GPU hangs since v3.9
Product:	Drivers	Reporter:	Chow Loong Jin (hyperair)
Component:	Video(DRI - Intel)	Assignee:	intel-gfx-bugs (intel-gfx-bugs)
Status:	RESOLVED INVALID
Severity:	normal	CC:	chris, daniel, hyperair, intel-gfx-bugs
Priority:	P1
Hardware:	All
OS:	Linux
Kernel Version:		Subsystem:
Regression:	No	Bisected commit-id:
Attachments:	i915_error_state from 3.10-rc6 glxinfo -l output

Description Chow Loong Jin 2013-06-20 07:39:54 UTC

Since v3.9, when running Ubuntu 13.04 with the X and Mesa stack from https://launchpad.net/~xorg-edgers/+archive/ppa, I have been getting frequent hangs on my Thinkpad E220S running with a Sandy Bridge (8086:0116/Core i5-2537M) GPU:

[10121.411125] [drm:i915_hangcheck_hung] *ERROR* Hangcheck timer elapsed... GPU hung
[10121.411131] [drm] capturing error event; look for more information in /sys/kernel/debug/dri/0/i915_error_state
--
[53978.637042] [drm:i915_hangcheck_hung] *ERROR* Hangcheck timer elapsed... GPU hung
--
[68285.759379] [drm:i915_hangcheck_hung] *ERROR* Hangcheck timer elapsed... GPU hung
[68285.759404] [drm:kick_ring] *ERROR* Kicking stuck semaphore on render ring
--
[69825.557728] [drm:i915_hangcheck_hung] *ERROR* Hangcheck timer elapsed... GPU hung
[69825.557740] [drm] capturing error event; look for more information in /sys/kernel/debug/dri/0/i915_error_state

In v3.9.x, X would freeze up, and I would have to kill X via Alt+SysRq+K in order to get out of the situation, but in v3.10 it looks like it's at least able to unhang itself without crashing X.

Comment 1 Daniel Vetter 2013-06-24 16:46:52 UTC

We need the error state file from debugfs (please attach as gzip to fit into bugzilla size limits).

Comment 2 Chow Loong Jin 2013-06-24 17:04:45 UTC

Here you go: http://people.ubuntu.com/~hyperair/i915-error-state

Comment 3 Daniel Vetter 2013-06-24 17:36:02 UTC

We seem to be executing (or trying to) complete garbage ... or the error state capture code is doing something funny. I'm confused.

Chris?

Comment 4 Chris Wilson 2013-06-24 19:47:08 UTC

That's the old i965g gallium driver.

Comment 5 Chow Loong Jin 2013-06-24 23:52:01 UTC

What? No, I'm not running i965g. I don't even have that installed.

Comment 6 Chow Loong Jin 2013-06-25 06:21:15 UTC

Created attachment 105931 [details]
i915_error_state from 3.10-rc6

If it helps, here's one from 3.10-rc6.

Comment 7 Chris Wilson 2013-06-25 06:43:12 UTC

At this moment, glxinfo would be more useful. Those batches are profoundly incorrect (sending commands for the wrong generation of GPUs).

Comment 8 Chow Loong Jin 2013-06-25 07:11:13 UTC

Created attachment 105941 [details]
glxinfo -l output

Comment 9 Chris Wilson 2013-06-25 07:18:02 UTC

That last error state is much cleaner; just the usual mesa blorp death.

Comment 10 Daniel Vetter 2013-06-25 09:01:39 UTC

Yeah, this smells like a mesa bug. Please retest with latest git mesa (if possible) and file a bug on bugs.freedesktop.org against Mesa -> Drivers/DRI/i965 with the error_state attached.