Bug 88771 - GPU HANG: ecode -1:0x00000000, reason: Ring hung, action: reset
Summary: GPU HANG: ecode -1:0x00000000, reason: Ring hung, action: reset
Status: RESOLVED WORKSFORME
Alias: None
Product: libva
Classification: Unclassified
Component: intel (show other bugs)
Version: unspecified
Hardware: x86-64 (AMD64) Linux (All)
: medium normal
Assignee: haihao
QA Contact: Sean V Kelley
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2015-01-24 10:39 UTC by Oleksii Shevchuk
Modified: 2015-11-24 06:10 UTC (History)
1 user (show)

See Also:
i915 platform:
i915 features:


Attachments
GPU crash dump (690.95 KB, text/plain)
2015-01-24 10:39 UTC, Oleksii Shevchuk
Details
Crash dump #2 (690.06 KB, text/plain)
2015-01-24 10:39 UTC, Oleksii Shevchuk
Details
Crash dump #3 (639.41 KB, text/plain)
2015-01-25 10:03 UTC, Oleksii Shevchuk
Details
Crash dump #4 (638.07 KB, text/plain)
2015-01-25 10:25 UTC, Oleksii Shevchuk
Details
My kernel .config file (112.04 KB, text/plain)
2015-02-12 05:21 UTC, Oleksii Shevchuk
Details

Description Oleksii Shevchuk 2015-01-24 10:39:14 UTC
Created attachment 112767 [details]
GPU crash dump

Happens twice with firefox-35 with this site: zenrus.ru. The system was under the load.

Linux WinterMute 3.18.0SIGN #6 SMP Sat Jan 17 07:20:28 EET 2015 x86_64 Intel(R) Core(TM) i7-3520M CPU @ 2.90GHz GenuineIntel GNU/Linux

00:02.0 VGA compatible controller: Intel Corporation 3rd Gen Core processor Graphics Controller (rev 09)

x11-drivers/xf86-video-intel-2.99.917

[48393.528091] [drm] no progress on render ring
[48393.528811] [drm] GPU HANG: ecode -1:0x00000000, reason: Ring hung, action: reset
[48393.528812] [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
[48393.528813] [drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel
[48393.528814] [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue.
[48393.528815] [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it.
[48393.528816] [drm] GPU crash dump saved to /sys/class/drm/card0/error
Comment 1 Oleksii Shevchuk 2015-01-24 10:39:43 UTC
Created attachment 112768 [details]
Crash dump #2
Comment 2 Chris Wilson 2015-01-24 15:52:34 UTC
Very suspect chained batches.
Comment 3 Oleksii Shevchuk 2015-01-25 10:03:19 UTC
Created attachment 112789 [details]
Crash dump #3

All crashes are under firefox 35. This time system was not under the load and the site was different.
Comment 4 Oleksii Shevchuk 2015-01-25 10:25:46 UTC
Created attachment 112790 [details]
Crash dump #4

This one with xf86-video-intel git (latest).

Like Crash dump #3 - firefox, the same site
Comment 5 haihao 2015-01-26 05:00:33 UTC
In libva-intel-driver, the chained batch buffer on render ring is only used for video encoding & video post post processing.  I don't think user is using video encoding & video post processing in this case.
Comment 6 haihao 2015-01-26 05:03:28 UTC
Did you install libva-intel-driver or gstreamer-vaapi in your system ? If yes, could you uninstall libva-intel-driver or gstreamer-vaapi ?
Comment 7 Oleksii Shevchuk 2015-01-26 22:05:13 UTC
Yes, I have both installed. So far I didn't have any problems. Probably because latest firefox was built with gstreamer, and I have gst-vaapi plugin. I'm going to remove it. 

But, anyway, something wrong with this..
Comment 8 haihao 2015-01-27 01:07:13 UTC
Can you reproduce the original issue without libva-intel-driver or gst-vaapi ?
Comment 9 Oleksii Shevchuk 2015-01-31 13:51:53 UTC
Without gst-vaapi there were no hangs anymore
Comment 10 haihao 2015-02-02 08:30:19 UTC
Could you provide the version of libva/libva-intel-driver and gst-vaapi ?
Comment 11 Oleksii Shevchuk 2015-02-02 20:20:39 UTC
libva-intel-driver: bf324e3440246a487997a1624ac862e3e4027f9e
libva: 1.5
media-plugins/gst-plugins-vaapi: 0.5.9
Comment 12 haihao 2015-02-10 07:15:00 UTC
I can't find the same of your machine (Ivy bridge, PCI ID 0x166), but I tried firefox3.5 on my machine (Ivy bridge, PCI ID 0x162) and can't reproduce this issue.
Comment 13 haihao 2015-02-10 07:16:23 UTC
(In reply to haihao from comment #12)
> I can't find the same of your machine (Ivy bridge, PCI ID 0x166), but I
> tried firefox3.5 on my machine (Ivy bridge, PCI ID 0x162) and can't
> reproduce this issue.

Typo. it is firefox35
Comment 14 haihao 2015-02-10 07:42:40 UTC
Could you give a try with gstreamer-vaapi 0.5.10 ? gstreamer-vaapi 0.5.10 should fix lots of decoding issue.
Comment 15 Oleksii Shevchuk 2015-02-10 21:22:51 UTC
Same behaviour with 0.5.10
Comment 16 haihao 2015-02-12 01:04:15 UTC
I tried firefox35 on Ivybridge (PCI ID 0x166) and can't reproduce this issue too. I make sure VAAPI is in use in my testing.  Do you always reproduce this issue ?
Comment 17 Oleksii Shevchuk 2015-02-12 05:21:36 UTC
Created attachment 113385 [details]
My kernel .config file

Yes. Be sure that your firefox uses gstreamer, vaapi and so on. Also, maybe this depends on kernel configuration?
Comment 18 Oleksii Shevchuk 2015-02-12 05:25:27 UTC
Also I use this version of libraries (libva-intel-driver deps):
x11-libs/libdrm-2.4.59
x11-libs/libpciaccess-0.13.2
Comment 19 haihao 2015-03-02 07:48:45 UTC
I can't reproduce this issue with your .config
Comment 20 haihao 2015-11-24 04:57:53 UTC
Is it still an issue for you?
Comment 21 Oleksii Shevchuk 2015-11-24 06:10:59 UTC
Didn't get this for a long time. 
So - no.


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.