Bug 95197 - [BSW/CHT] GPU HANG: ecode 8:2:0xfffffffe, in DVDPlayerVideo [1027], reason: Ring hung, action: reset in bsd ring
Summary: [BSW/CHT] GPU HANG: ecode 8:2:0xfffffffe, in DVDPlayerVideo [1027], reason: R...
Status: CLOSED FIXED
Alias: None
Product: libva
Classification: Unclassified
Component: intel (show other bugs)
Version: unspecified
Hardware: Other All
: medium normal
Assignee: ykzhao
QA Contact: Sean V Kelley
URL:
Whiteboard:
Keywords: regression
Depends on:
Blocks:
 
Reported: 2016-04-28 22:08 UTC by Igor Mammedov
Modified: 2016-12-14 18:28 UTC (History)
1 user (show)

See Also:
i915 platform: BSW/CHT
i915 features: GPU hang


Attachments
dmesg (224.49 KB, text/plain)
2016-04-28 22:08 UTC, Igor Mammedov
Details
error log (197.04 KB, text/plain)
2016-04-28 22:09 UTC, Igor Mammedov
Details

Description Igor Mammedov 2016-04-28 22:08:58 UTC
Created attachment 123331 [details]
dmesg

getting unrecoverable GPU hang with error:

Apr 28 23:37:00 dhcppc14 kernel: [drm] GPU HANG: ecode 8:2:0xfffffffe, in DVDPlayerVideo [1027], reason: Ring hung, action: reset
Apr 28 23:37:00 dhcppc14 kernel: [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
Apr 28 23:37:00 dhcppc14 kernel: [drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel
Apr 28 23:37:00 dhcppc14 kernel: [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue.
Apr 28 23:37:00 dhcppc14 kernel: [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it.
Apr 28 23:37:00 dhcppc14 kernel: [drm] GPU crash dump saved to /sys/class/drm/card0/error
Apr 28 23:37:00 dhcppc14 kernel: [drm:i915_reset_and_wakeup] resetting chip
Apr 28 23:37:00 dhcppc14 kernel: drm/i915: Resetting chip after gpu hang

Reproducible 100% after starting video playback via vaapi backend.
Kernel: 4.6.0-0.rc5.git1.1.fc25.x86_64
kodi: 16.0
libva/libva-intel 1.6.2-1

It 'works' with kernel-core-4.6.0-0.rc0.git1.1.fc25, where occasionally video hangs with:
[drm:intel_cpu_fifo_underrun_irq_handler [i915]] *ERROR* CPU pipe A FIFO underrun  
but it's recoverable by stopping playback or skipping to playing the next video file.
Comment 1 Igor Mammedov 2016-04-28 22:09:37 UTC
Created attachment 123333 [details]
error log
Comment 2 yann 2016-12-07 09:15:53 UTC
We seem to have neglected the bug a bit, apologies.

The issue is occurring in bsd ring. Reassigning to libva component.

There were improvements pushed in kernel and libva that will benefit to your system, so please re-test with latest kernel & libva and mark as REOPENED if you can reproduce (and attach fresh gpu error dump & kernel log) and RESOLVED/* if you cannot reproduce.
Comment 3 haihao 2016-12-07 15:23:00 UTC
Besides kodi, can you reproduce this issue with other vaapi based player?
Comment 4 haihao 2016-12-07 15:27:23 UTC
Did you reproduce this issue with a specific video file? If yes, is it possible to share the video file?
Comment 5 Igor Mammedov 2016-12-14 18:14:13 UTC
It seems to be working fine (without visible effects) since 4.8 kernel.
Currently I run Fedora 24 with following components:

kernel: 4.8.11-200.fc24.x86_64
libva-1.7.3-1.fc24.x86_64
libva-intel-driver-1.7.3-1.fc24.x86_64
kodi-16.1-3.fc24.x86_64
ffmpeg-libs-3.1.5-1.fc24.x86_64

Uptime since last update/reboot is 5 days and there is only one error in dmesg
which doesn't seem to affect anything:

[drm:intel_pipe_update_end [i915]] *ERROR* Atomic update failure on pipe A (start=81437 end=81438) time 290 us, min 1074, max 1079, scanline start 1067, end 1067

So I'm closing this BZ as fixed.


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.