Bug 25598 - [965GM] Corruption on resume from hibernation with xf86-video-intel-git
[965GM] Corruption on resume from hibernation with xf86-video-intel-git
Status: RESOLVED FIXED
Product: xorg
Classification: Unclassified
Component: Driver/intel
7.5 (2009.10)
x86 (IA32) Linux (All)
: highest critical
Assigned To: Carl Worth
Xorg Project Team
:
Depends on:
Blocks:
  Show dependency treegraph
 
Reported: 2009-12-11 12:41 UTC by Anonymous Helper
Modified: 2010-01-20 09:15 UTC (History)
4 users (show)

See Also:


Attachments
Xorg.0.log (59.33 KB, application/x-trash)
2009-12-13 22:30 UTC, Petar Velkovski
no flags Details
kern.log (571.51 KB, text/x-log)
2009-12-13 22:30 UTC, Petar Velkovski
no flags Details

Note You need to log in before you can comment on or make changes to this bug.
Description Anonymous Helper 2009-12-11 12:41:40 UTC
Hello, 
I've been using GIT snapshots of xf86-video-intel and got some graphics corruption after resuming from hibernation today. May last update was yesterday with commit 37f631d66 (Revert "uxa-glyphs: Enable TILING_X on glyph caches) - which I suspect was made to fix these kind of things, but my display got corrupted. I am sending this report because I hope this won't happen when version 2.10 of the driver is finally released.

Some 10 minutes after resuming I noticed graphic corruption and my display hanged, it was accompanied with these messages in the kernel log: 

[drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer elapsed... GPU hung
render error detected, EIR: 0x00000000
i915: Waking up sleeping processes
[drm:i915_gem_execbuffer] *ERROR* Execbuf while wedged
[drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer elapsed... GPU hung
render error detected, EIR: 0x00000000
i915: Waking up sleeping processes
[drm:i915_wait_request] *ERROR* i915_wait_request returns -5 (awaiting 523491 at 523490)
[drm:i915_gem_execbuffer] *ERROR* Execbuf while wedged

- and so on repeating it self until I killed X -


...and these messages in the X.org log:

(WW) intel(0): i830_uxa_prepare_access: gtt bo map failed: Input/output error
(WW) intel(0): i830_uxa_pixmap_swap_bo_with_image: bo map failed
(EE) intel(0): Failed to submit batch buffer, expect rendering corruption or even a frozen display: Input/output error.
(WW) intel(0): i830_uxa_prepare_access: gtt bo map failed: Input/output error
(EE) intel(0): Failed to submit batch buffer, expect rendering corruption or even a frozen display: Input/output error.
- and so on repeating it self until I killed X -

I am using X.org 7.5 with mesa 7.6 and libdrm 2.4.16.

Let me know how I could provide more information, thank you.
Comment 1 Petar Velkovski 2009-12-13 22:20:54 UTC
What kind of corruption are you experiencing? Font corruption or something more general? Developers usually ask for a screenshot in order to see what kind of corruption occurs. The patch that you are referring to was done because of bug 25406. You can access it at http://bugs.freedesktop.org/show_bug.cgi?id=25406, If your screen corruption is related to it.
Comment 2 Petar Velkovski 2009-12-13 22:29:35 UTC
I would like to document that instead of screen corruption, I am getting screen freezes that give same messages to the Xorg.0.log and kern.log. I am on mesa 7.7 git, X.org 7.4, libdrm 2.4.16. Linux kernel 2.6.32. The graphic chip is GM945. I am also sending attachments of Xorg.0.log and kern.log.
Comment 3 Petar Velkovski 2009-12-13 22:30:17 UTC
Created attachment 32052 [details]
Xorg.0.log
Comment 4 Petar Velkovski 2009-12-13 22:30:46 UTC
Created attachment 32053 [details]
kern.log
Comment 5 Anonymous Helper 2009-12-14 10:03:07 UTC
Hello Petar, 
indeed I noticed a font corruption in Emacs which signaled something is happening, but I couldn't do anything because all windows started lagging - like moving a window was delayed by 15 seconds and then followed by a display freeze.

I suspend and hibernate my machine a lot, this did not happen on resume from suspend, not once, but I did not hibernate since then because of it.

I forgot to add to my report that last Git snapshot I used before this one was from 24.11.2009, and didn't have these problems. I also use Linux 2.6.32 and of course have KMS enabled.
Comment 6 Robert Hooker (Sarvatt) 2009-12-25 10:39:20 UTC
(In reply to comment #0)

[264301.566382] (WW) intel(0): i830_uxa_pixmap_swap_bo_with_image: bo map failed
[264301.566535] (WW) intel(0): i830_uxa_prepare_access: gtt bo map failed: Input/output error
[264301.566658] (WW) intel(0): i830_uxa_prepare_access: gtt bo map failed: Input/output error
[264301.566777] (WW) intel(0): i830_uxa_prepare_access: gtt bo map failed: Input/output error
[264301.570390] (EE) intel(0): Failed to submit batch buffer, expect rendering corruption or even a frozen display: Input/output error.

I also am getting this if I get a hang after a suspend/resume cycle. This is similar to bug 25475, only the i830_uxa_pixmap_swap_bo_with_image: bo map failed line does not appear if the execbuff while wedged crash happens prior to a suspend/resume cycle. When this hang happens, I can't switch to a VT unlike the hangs in bug 25475. It is hard to reproduce this one to submit a dump because I am not usually in a situation where I can ssh in from another machine to do when it happens coupled with VT's not working.
Comment 7 Robert Hooker (Sarvatt) 2009-12-25 11:04:15 UTC
(In reply to comment #6)

I forgot to mention, but reverting libdrm to a 2.4.15 checkout from 11-25 up to commit 6f66de982ad6ee6967ec61a7399e600bdd9e5887 also fixes this problem for me as it does the bug 25475 problem, but that is not a good solution because 2.4.16 is needed to build intel now and 2.4.16 release from 12-03 also exibits this problem.
Comment 8 Anonymous Helper 2010-01-03 12:24:47 UTC
It has been (almost)a month since my initial report. I am now running the following: 

xf86-video-intel 2.9.99.902
intel-dri / mesa 7.7
libdrm 2.4.17
linux 2.6.32.2

I have not seen this bug in two weeks, I suspend daily and hibernate ocasionally.

The reports linked by other people suggest there are some problems with the driver, maybe even more than one, but at the moment I am unable to reproduce any of it on my machine. Being that there are multiple reports on the same issue, and I can't reporoduce the problem I have no objections to closing this bug report.
Comment 9 Eric Anholt 2010-01-20 09:15:02 UTC
Reporter says it's fixed.  If you're not the reporter and you're having an issue like this, please open your own bug.