Bug 92555 - GPU lockup crashing the system on Cayman with HyperZ
Summary: GPU lockup crashing the system on Cayman with HyperZ
Status: RESOLVED MOVED
Alias: None
Product: Mesa
Classification: Unclassified
Component: Drivers/Gallium/r600 (show other bugs)
Version: 11.0
Hardware: x86-64 (AMD64) Linux (All)
: medium normal
Assignee: Default DRI bug account
QA Contact: Default DRI bug account
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2015-10-20 11:42 UTC by Thomas Rohloff
Modified: 2019-09-18 19:19 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
dmesg part 1 (png file) (158.47 KB, image/png)
2015-10-20 11:42 UTC, Thomas Rohloff
Details
dmesg part 2 (png file) (149.83 KB, image/png)
2015-10-20 11:43 UTC, Thomas Rohloff
Details
radeontop output while the screen shows garbage (Metro 2033 Redux) (82.02 KB, image/png)
2015-10-20 11:44 UTC, Thomas Rohloff
Details

Description Thomas Rohloff 2015-10-20 11:42:42 UTC
Created attachment 119002 [details]
dmesg part 1 (png file)

Reproduced with Left 4 Dead 2, Distance and Metro 2033 Redux.

First the screen either freezes (Distance) or shows garbage, looks like many small brown rectangles on the screen (Left 4 Dead 2, Metro 2033 Redux), then, after around 30 seconds, the whole system freezes (keyboard/mouse doesn't react, ssh connections drop, finally watchdog kicks in and reboots the system).

In between the 30 seconds I was able to grab some data via ssh. Screenshots attached.

This is on a HD 6950.
Comment 1 Thomas Rohloff 2015-10-20 11:43:19 UTC
Created attachment 119003 [details]
dmesg part 2 (png file)
Comment 2 Thomas Rohloff 2015-10-20 11:44:33 UTC
Created attachment 119004 [details]
radeontop output while the screen shows garbage (Metro 2033 Redux)
Comment 3 Thomas Rohloff 2015-10-20 11:47:34 UTC
I also thought that the GPU might simply overheat but sensors told it has 60°C which should be fine.

Kernel in use: 4.0.4 with patches (see https://bugzilla.kernel.org/show_bug.cgi?id=99041 )
Comment 4 Thomas Rohloff 2015-10-20 11:56:43 UTC
Forgot to say: It's very random. Sometimes it happens after 20 minutes of gameplay, sometimes I can play for hours without problems.
Comment 5 Thomas Rohloff 2015-12-04 19:37:44 UTC
Seems to be a problem with HyperZ. R600_DEBUG=nohyperz looks like a workaround.
Comment 7 Thomas Rohloff 2016-01-06 07:28:42 UTC
Looks good. Thanks to you and the hw people.
Comment 8 Thomas Rohloff 2016-01-10 13:30:57 UTC
I was too fast on this, it's still there.
Comment 9 Marek Olšák 2016-10-23 19:36:22 UTC
This might help:
https://cgit.freedesktop.org/mesa/mesa/commit/?id=d4d9ec55c589156df4edc227a86b4a8c41048d58

Although it's an unrelated feature, it changes the allocation function from pipe_buffer_create to r600_aligned_buffer_create. I believe that can fix HyperZ hangs on big GPU.
Comment 10 GitLab Migration User 2019-09-18 19:19:46 UTC
-- GitLab Migration Automatic Message --

This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity.

You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/mesa/mesa/issues/557.


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.