Bug 46474 - [965gm uxa] GPU hang when running glslideshow for too long
Summary: [965gm uxa] GPU hang when running glslideshow for too long
Status: RESOLVED FIXED
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/intel (show other bugs)
Version: 7.7 (2012.06)
Hardware: Other All
: medium normal
Assignee: Chris Wilson
QA Contact: Intel GFX Bugs mailing list
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2012-02-22 12:04 UTC by Yves-Alexis
Modified: 2013-01-16 18:03 UTC (History)
1 user (show)

See Also:
i915 platform:
i915 features:


Attachments
extract from syslog (58.16 KB, text/plain)
2012-02-22 12:04 UTC, Yves-Alexis
no flags Details
dmesg extract (54.35 KB, text/plain)
2012-02-22 12:04 UTC, Yves-Alexis
no flags Details
error state (760.18 KB, text/plain)
2012-02-22 13:03 UTC, Yves-Alexis
no flags Details
new log (28.61 KB, text/plain)
2012-02-22 23:16 UTC, Yves-Alexis
no flags Details
erro state (760.28 KB, text/plain)
2012-02-22 23:21 UTC, Yves-Alexis
no flags Details
Xorg log (34.44 KB, text/plain)
2012-02-23 04:48 UTC, Yves-Alexis
no flags Details

Description Yves-Alexis 2012-02-22 12:04:09 UTC
Created attachment 57477 [details]
extract from syslog

Hey,

I have an issue with my Intel GM965 card (on Thinkpad T61, running Debian sid).

Basically when letting my screensaver (which is glsideshow displaying pictures on both LVDS and VGA screen) run for too long, the GPU seems to hang.

I experienced it few times when I forgot to close the laptop lid for the night, when waking up the displays are black, no answer when touching the keyboard, but the laptop is still alive and can be accessed by ssh.

There, I have the attached logs (dmesg and syslog)

I have debugfs mounted so I can show some debugging information if you tell me what you want.
Comment 1 Yves-Alexis 2012-02-22 12:04:46 UTC
Created attachment 57478 [details]
dmesg extract
Comment 2 Chris Wilson 2012-02-22 12:29:29 UTC
Feb 21 22:05:19 hidalgo kernel: [236215.660109] [drm] capturing error event; look for more information in /debug/dri/0/i915_error_state

Please attach the error-state :)
Comment 3 Yves-Alexis 2012-02-22 13:03:52 UTC
Created attachment 57481 [details]
error state

Here it is.
Comment 4 Chris Wilson 2012-02-22 13:54:45 UTC
Nothing appears obviously wrong. :|
Comment 5 Yves-Alexis 2012-02-22 14:20:46 UTC
That might be because the problem appeared last night or early this morning, but I got the error state this evening. I've rebooted the laptop and will let it idle tonight, and see if it happens again. Then I'll provide the new error state.
Comment 6 Chris Wilson 2012-02-22 14:30:26 UTC
The error-state looks genuine, I just can't spot an actual error. If it does happen again, please do attach further error-states and lets hope the bug is more obvious in one of those, or a pattern becomes evident.
Comment 7 Yves-Alexis 2012-02-22 23:16:41 UTC
Created attachment 57505 [details]
new log

So it did happen again, so here's a new log.
Comment 8 Yves-Alexis 2012-02-22 23:21:12 UTC
Created attachment 57506 [details]
erro state

And the new error state. Unfortunately I forgot to mount debugfs *before* it happened, so it might be invalid.
Comment 9 Chris Wilson 2012-02-23 01:39:51 UTC
Can you also post an Xorg.log following one of the hangs? There's a slim chance it contains some warnings from before the hang.
Comment 10 Chris Wilson 2012-02-23 01:51:16 UTC
The error state itself demonstrates the same hang, so it is consistent even if I have no clue as to the cause.

The shotgun debugging response would be to run with sna (download xf86-video-intel.git and ./configure --enable-sna) to see if that makes any difference.
Comment 11 Yves-Alexis 2012-02-23 04:48:47 UTC
Created attachment 57522 [details]
Xorg log

(In reply to comment #10)
> The error state itself demonstrates the same hang, so it is consistent even if
> I have no clue as to the cause.

Here it is.
> 
> The shotgun debugging response would be to run with sna (download
> xf86-video-intel.git and ./configure --enable-sna) to see if that makes any
> difference.

I'll try to do that (first by using the same build options as the Debian package just to be sure, then with sna).
Comment 12 Yves-Alexis 2012-02-24 06:53:59 UTC
I've upgraded to 2.18 with SNA enabled and will report back
Comment 13 Yves-Alexis 2012-02-25 10:01:56 UTC
So I tried 2.18 with and without SNA: with SNA, problem seems fixed, without it still happens.
Comment 14 Chris Wilson 2012-03-02 03:49:18 UTC
Yves-Alexis please keep using SNA for the time being, and please do report any issues found. Thanks.
Comment 15 Chris Wilson 2013-01-16 18:03:47 UTC
I believe this to be:

commit 262b6d363fcff16359c93bd58c297f961f6e6273
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Tue Jan 15 16:17:54 2013 +0000

    drm/i915: Invalidate the relocation presumed_offsets along the slow path


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.