Bug 21232 - [GM965 GEM KMS] Running rss-glx hacks lock up GPU
Summary: [GM965 GEM KMS] Running rss-glx hacks lock up GPU
Status: RESOLVED FIXED
Alias: None
Product: Mesa
Classification: Unclassified
Component: Drivers/DRI/i965 (show other bugs)
Version: unspecified
Hardware: Other All
: high critical
Assignee: Eric Anholt
QA Contact:
URL:
Whiteboard:
Keywords: NEEDINFO
Depends on:
Blocks:
 
Reported: 2009-04-16 10:49 UTC by Ben Gamari
Modified: 2009-08-08 07:47 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
GPU dump from crash while running lattice (93.06 KB, application/x-bzip)
2009-04-16 10:57 UTC, Ben Gamari
Details
GPU dump from crash while running hyperspace (92.76 KB, application/x-bzip)
2009-04-16 10:58 UTC, Ben Gamari
Details

Description Ben Gamari 2009-04-16 10:49:18 UTC
Attempting to run the rss-glx hacks has the tendency of occasionally locking up the GPU. This has been reproduced with spirographx, lattice, and hyperspace. GPU dumps attached.
Comment 1 Ben Gamari 2009-04-16 10:57:02 UTC
Created attachment 24879 [details]
GPU dump from crash while running lattice
Comment 2 Ben Gamari 2009-04-16 10:58:04 UTC
Created attachment 24880 [details]
GPU dump from crash while running hyperspace
Comment 3 Ben Gamari 2009-04-16 10:59:27 UTC
$ xorg-versions.sh 
Xorg components as of Thu Apr 16 13:58:33 EDT 2009
drm: 	1173e7abdcdf758a2403ce921076080c6672c054
xf86-video-intel: 	ebb8d6a13a18138b31ad119be2d3807a1e4010b3
mesa: 	e704de8cb62092f7402cfe99064fcd692e492086
xserver: 	49bd35c28245d2261f17887f03a23deddf57d1e9

Linux mercury.localdomain 2.6.29work #34 SMP PREEMPT Thu Apr 16 12:25:16 EDT 2009 x86_64 x86_64 x86_64 GNU/Linux
Effectively drm-intel-next (with PAT and tiling fixes)
Comment 4 Ben Gamari 2009-04-16 11:02:26 UTC
I should probably note that it rarely takes more than 5 attempts of starting/killing any given series of rss-glx hacks to crash the GPU.
Comment 5 Eric Anholt 2009-08-03 19:33:58 UTC
Is this still a problem for you?  I've run:

while true; do mesagl /usr/lib/xscreensaver/lattice&; sleep 1; killall lattice; done

and same for hyperspace, for several minutes each on 965GM (T61) and G45, with no troubles.
Comment 6 Eric Anholt 2009-08-07 18:52:09 UTC
OK, I've pushed a commit series that should fix GPU hangs with DRI2 in use.
The symptoms in the GPU dumps are unreliable, as sometimes the corruption
would occur in a state buffer not dumped in the intel_gpu_dump output, but
I've definitely seen the symptoms of this in a number of the reports.  Please
retest with this set of changes, and reopen if the problem continues.

(in this case, your dump shows a depth buffer that would have drawn around 4 pages outside of its buffer)

xf86-video-intel:
commit e8f0763d405a8152c74c28792c52fe12c1d41dd5
Author: Eric Anholt <eric@anholt.net>
Date:   Fri Aug 7 18:24:44 2009 -0700

    Fix math in the tiling alignment fix.

commit 222b52ef16895823fbf3a0fc0be4eb23b930ed1b
Author: Eric Anholt <eric@anholt.net>
Date:   Fri Aug 7 18:05:29 2009 -0700

    Align tiled pixmap height so we don't address beyond the end of our buffers.

Mesa:
commit ceb8afcca5b0a52b005a782ea54b301beaee1a15
Author: Eric Anholt <eric@anholt.net>
Date:   Fri Aug 7 18:09:31 2009 -0700

    intel: Align region height as required for tiled regions.

    Otherwise, we would address beyond the end of our buffers.  Fixes reliable
    GPU segfault with texture_tiling=true and oglconform shadow.c.

    Bug #22406.
Comment 7 Ben Gamari 2009-08-08 07:47:10 UTC
Cool, I'll take a look at it. Thanks!

- Ben


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.