Bug 74256 - [snb] chromium triggers a gpu hang
Summary: [snb] chromium triggers a gpu hang
Status: NEEDINFO
Alias: None
Product: Mesa
Classification: Unclassified
Component: Drivers/DRI/i965 (show other bugs)
Version: 10.0
Hardware: Other All
: medium normal
Assignee: Intel 3D Bugs Mailing List
QA Contact: Intel 3D Bugs Mailing List
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2014-01-31 05:18 UTC by Joe Kappus
Modified: 2017-03-22 02:18 UTC (History)
1 user (show)

See Also:
i915 platform:
i915 features:


Attachments
gpu crash dump (2.39 MB, text/plain)
2014-01-31 05:18 UTC, Joe Kappus
Details
Patch to always emit workaround flushes, rather than being stingy about it (4.06 KB, patch)
2014-01-31 17:01 UTC, Kenneth Graunke
Details | Splinter Review

Note You need to log in before you can comment on or make changes to this bug.
Description Joe Kappus 2014-01-31 05:18:59 UTC
Created attachment 93097 [details]
gpu crash dump

[ 9224.436175] Watchdog[18586]: segfault at 0 ip 00007fcdd2d19868 sp 00007fcdbf522810 error 6 in chrome[7fcdcf65b000+50e9000]
[ 9232.579353] [drm] stuck on render ring
[ 9232.579356] [drm] GPU crash dump saved to /sys/class/drm/card0/error
[ 9232.579356] [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
[ 9232.579357] [drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel
[ 9232.579367] [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue.
[ 9232.579367] [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it.
[ 9232.605360] [drm:i915_set_reset_status] *ERROR* render ring hung inside bo (0xcd02000 ctx 2) at 0xcd03a84

above from dmesg, screen freezes and X drops off.
Comment 1 Chris Wilson 2014-01-31 08:10:02 UTC
What happens to X? Clean shutdown from clients dieing or does it crash?
Comment 2 Joe Kappus 2014-01-31 08:17:10 UTC
Jan 31 00:08:47 superbia kernel: Watchdog[18586]: segfault at 0 ip 00007fcdd2d19868 sp 00007fcdbf522810 error 6 in chrome[7fcdcf65b000+50e9000]
Jan 31 00:08:55 superbia kernel: [drm] stuck on render ring
Jan 31 00:08:55 superbia kernel: [drm] GPU crash dump saved to /sys/class/drm/card0/error
Jan 31 00:08:55 superbia kernel: [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
Jan 31 00:08:55 superbia kernel: [drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel
Jan 31 00:08:55 superbia kernel: [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue.
Jan 31 00:08:55 superbia kernel: [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it.
Jan 31 00:08:55 superbia kernel: [drm:i915_set_reset_status] *ERROR* render ring hung inside bo (0xcd02000 ctx 2) at 0xcd03a84
Jan 31 00:09:09 superbia acpid: client 2435[0:100] has disconnected
Jan 31 00:09:10 superbia acpid: client connected from 20490[0:100]
Jan 31 00:09:10 superbia acpid: 1 client rule loaded

This is from syslog on the same event. I think X had a clean shutdown this time, the other times it didn't always take down X, I could flip to console and back to X to get my screen back up (just appeared black with cursor until I did this).
Comment 3 Chris Wilson 2014-01-31 09:42:11 UTC
Without an exploding X, lets focus on the hanging batch.

* note that marcheau has disabled hw contexts in CrOS due to unidentified hangs.
Comment 4 Kenneth Graunke 2014-01-31 17:01:01 UTC
Created attachment 93147 [details] [review]
Patch to always emit workaround flushes, rather than being stingy about it

Hi Joe,

Can you try the attached patch (on your Sandybridge machine)?  I believe it may help with the hangs.  Either way, I'd be interested to know the results.

Thanks!
--Ken


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.