Bug 38322 - ColorTiling + DualHead (xrandr) + screensaver -> GPU lockup on RV620
Summary: ColorTiling + DualHead (xrandr) + screensaver -> GPU lockup on RV620
Status: RESOLVED FIXED
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/Radeon (show other bugs)
Version: 7.6 (2010.12)
Hardware: x86-64 (AMD64) Linux (All)
: medium critical
Assignee: xf86-video-ati maintainers
QA Contact: Xorg Project Team
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2011-06-14 16:30 UTC by Jan Kriho
Modified: 2011-08-10 10:38 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
Xorg log from the lockup (135.49 KB, text/plain)
2011-07-28 05:03 UTC, Jan Kriho
no flags Details
/var/log/messages (257.01 KB, text/plain)
2011-07-28 06:45 UTC, Jan Kriho
no flags Details
`xrandr --verbose` output (6.59 KB, text/plain)
2011-08-09 04:49 UTC, Jan Kriho
no flags Details
Xorg backtrace (1.96 KB, text/plain)
2011-08-09 06:10 UTC, Jan Kriho
no flags Details

Description Jan Kriho 2011-06-14 16:30:57 UTC
While running KDE 4.6.3 (Compositing enabled) on two monitors (configured with XRandr), when Goban screensaver starts, the whole graphic system freezes (box is still SSH-able tough) and only SysRq combo can reboot the machine. 

With single display, everything is fine. 


System: Debian Wheezy/Sid
Kernel: 2.6.38
Xorg: 7.6
xserver: 1.10.2
Driver: radeon 6.14.2 (ColorTiling enabled)
Renderer: Mesa 7.10.2 r600g
Graphic card: Mobility Radeon HD3430
Comment 1 Jan Kriho 2011-07-28 05:03:18 UTC
Created attachment 49663 [details]
Xorg log from the lockup

Bug still present with git version of radeon driver, attaching the Xorg.0.log
Comment 2 Alex Deucher 2011-07-28 05:55:39 UTC
Please attach your dmesg output.
Comment 3 Jan Kriho 2011-07-28 06:45:58 UTC
Created attachment 49665 [details]
/var/log/messages

Attaching /var/log/messages (hope that is what dmesg outputs :)), infinite loop happened couple of minutes before Jul 28, 13:53
Comment 4 Alex Deucher 2011-07-28 06:58:53 UTC
Does a newer kernel help?  2.6.39 or 3.0?
Comment 5 Jan Kriho 2011-07-28 07:19:57 UTC
It does not, I am using 3.0.0 from Debian experimental repository, it still persists.
Comment 6 Michel Dänzer 2011-08-09 04:02:19 UTC
(In reply to comment #6)
> While running KDE 4.6.3 (Compositing enabled)

Does disabling compositing or tiling work around the problem?

> on two monitors (configured with XRandr),

Please attach the output of

xrandr --verbose

when both monitors are enabled.

> when Goban screensaver starts, the whole graphic system freezes (box
> is still SSH-able tough) and only SysRq combo can reboot the machine. 

If you ssh in, is the X server (and goban, ...) process still alive? If yes, can you attach gdb to the X server and get a backtrace?

> Renderer: Mesa 7.10.2 r600g

Would a newer version of this help by any chance?
Comment 7 Jan Kriho 2011-08-09 04:49:05 UTC
Created attachment 50066 [details]
`xrandr --verbose` output

Attaching the `xrandr --verbose` output when both monitors are enabled.
Comment 8 Jan Kriho 2011-08-09 05:00:18 UTC
(In reply to comment #6)
> (In reply to comment #6)
> > While running KDE 4.6.3 (Compositing enabled)
> 
> Does disabling compositing or tiling work around the problem?

Disabling compositing does not help, as I turn it off when using second monitor for performance reasons. 

Before enabling Color Tiling I haven't encountered the problem. Now I have it turned on again for performance reasons.

> 
> > on two monitors (configured with XRandr),
> 
> Please attach the output of
> 
> xrandr --verbose
> 
> when both monitors are enabled.

Please see the last attachments.

> 
> > when Goban screensaver starts, the whole graphic system freezes (box
> > is still SSH-able tough) and only SysRq combo can reboot the machine. 
> 
> If you ssh in, is the X server (and goban, ...) process still alive? If yes,
> can you attach gdb to the X server and get a backtrace?

I will do so when I encounter the loop again. It is now not-so-predictable and various applications can trigger it in non-deterministic way.

Also I have upgraded my ati drivers yesterday (to the git head version) and I don't know yet if the problem persists.

> 
> > Renderer: Mesa 7.10.2 r600g
> 
> Would a newer version of this help by any chance?


Upgrading to 7.11 branch (from Debian experimental repository) did not help.
Comment 9 Jan Kriho 2011-08-09 06:10:41 UTC
Created attachment 50069 [details]
Xorg backtrace

Xorg backtrace from the crash (caused by running screensaver and quitting it successfully twice and starting Digikam afterwards)
Comment 10 Michel Dänzer 2011-08-09 06:32:26 UTC
Would http://lists.x.org/archives/xorg-devel/2011-August/024291.html help by any chance?
Comment 11 Jan Kriho 2011-08-09 09:29:16 UTC
Thanks, I tested the patch, and so far it looks stable.
Comment 12 Michel Dänzer 2011-08-09 09:41:03 UTC
(In reply to comment #11)
> Thanks, I tested the patch, and so far it looks stable.

Thanks, can you please test http://lists.x.org/archives/xorg-devel/2011-August/024302.html as well?
Comment 13 Jan Kriho 2011-08-09 10:59:51 UTC
(In reply to comment #12)
> (In reply to comment #11)
> > Thanks, I tested the patch, and so far it looks stable.
> 
> Thanks, can you please test
> http://lists.x.org/archives/xorg-devel/2011-August/024302.html as well?

Yes, this patch looks stable too.
Comment 14 Michel Dänzer 2011-08-10 08:58:35 UTC
The fix has landed in xserver master, and I nominated it for the 1.10 branch.
Comment 15 Jan Kriho 2011-08-10 10:38:23 UTC
Thanks.


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.