Bug 93788 - [SKL][DMC][GLES 3.1 CTS] dmc inflicting gt hangs with cts-runner
Summary: [SKL][DMC][GLES 3.1 CTS] dmc inflicting gt hangs with cts-runner
Status: NEW
Alias: None
Product: Mesa
Classification: Unclassified
Component: Drivers/DRI/i965 (show other bugs)
Version: 11.0
Hardware: Other All
: medium normal
Assignee: Mika Kuoppala
QA Contact: Intel 3D Bugs Mailing List
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2016-01-20 12:04 UTC by Marta Löfstedt
Modified: 2016-04-04 16:23 UTC (History)
1 user (show)

See Also:
i915 platform:
i915 features:


Attachments
dmesg from when the problem occur. (62.57 KB, text/plain)
2016-01-21 08:53 UTC, Marta Löfstedt
Details

Note You need to log in before you can comment on or make changes to this bug.
Description Marta Löfstedt 2016-01-20 12:04:40 UTC
Software versions:
    4.4.0-rc8+
    OpenGL version string: 3.0 Mesa 11.2.0-devel (git-4475d8f)

GPU hardware:
    OpenGL renderer string: Mesa DRI Intel(R) HD Graphics 530 (Skylake GT2)
    00:02.0 VGA compatible controller [0300]: Intel Corporation Sky Lake Integrated Graphics [8086:1912] (rev 06)

CPU hardware:
    x86_64
    Intel(R) Core(TM) i5-6600K CPU @ 3.50GHz

command: cts-runner
cts-version: Intel internal branch:gles31_rebase_khronos_commit_01f3503b3502 git@ae6a221332
------------------------
When running cts runner on SKL somewhat "random" tests fail. I am still gathering data on this issue, but so far:

The number of failing tests on a full cts-runner run are equal to getting a new:

"[66951.923100] [drm] stuck on render ring
[66951.923838] [drm] GPU HANG: ecode 9:0:0x85dfffff, in cts-runner [14838], reason: Ring hung, action: reset
[66951.926772] drm/i915: Resetting chip after gpu hang
[66953.923521] [drm] RC6 on"

in dmesg.

The tests appear to fail on random configs. To check the result of the cts-runner I use the scripts/

So far after 4 runs I have only see these test fail on respective config:
----- run 1 ----
Verifying config-18-run-0.qpa
FAIL: ES2-CTS.gtf.GL.operators.multiplicative_vert: Fail (Fail)
Verifying config-25-run-0.qpa
FAIL: ES31-CTS.layout_binding.block_layout_binding_block_ComputeShader: Fail (Fail)
Verifying config-82-run-0.qpa
FAIL: ES31-CTS.layout_binding.block_layout_binding_block_ComputeShader: Fail (Fail)
Verifying config-89-run-0.qpa
FAIL: ES2-CTS.gtf.GL.operators.multiplicative_vert: Fail (Fail)
----- run 2 ----
Verifying config-18-run-0.qpa
FAIL: ES3-CTS.gtf.GL3Tests.npot_textures.npot_tex3d_render: Fail (Fail)
Verifying config-25-run-0.qpa
FAIL: ES3-CTS.gtf.GL3Tests.transform_feedback.transform_feedback_misc: Fail (Fail)
----- run 3 ----
Verifying config-18-run-1.qpa
FAIL: ES3-CTS.gtf.GL3Tests.transform_feedback.transform_feedback_misc: Fail (Fail)
Verifying config-25-run-0.qpa
FAIL: ES31-CTS.layout_binding.block_layout_binding_block_ComputeShader: Fail (Fail)
Verifying config-78-run-1.qpa
FAIL: ES3-CTS.gtf.GL3Tests.transform_feedback.transform_feedback_misc: Fail (Fail)
Verifying config-82-run-0.qpa
FAIL: ES31-CTS.layout_binding.block_layout_binding_block_ComputeShader: Fail (Fail)
----- run 4 ----
Verifying config-82-run-0.qpa
FAIL: ES2-CTS.gtf.GL.operators.multiplicative_vert: Fail (Fail)
Verifying config-89-run-0.qpa
FAIL: ES31-CTS.layout_binding.block_layout_binding_block_ComputeShader: Fail (Fail)
Comment 1 Marta Löfstedt 2016-01-20 15:16:16 UTC
With drm-nightly kernel:

commit c783b5011894af49992f2095cb2848b6cf8ebc57
Author: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Date:   Wed Jan 20 10:12:32 2016 +0000

    drm-intel-nightly: 2016y-01m-20d-10h-12m-03s UTC integration manifest

and by disabling the DMC firmware from loading. I can't reproduce this issue.
Comment 2 Marta Löfstedt 2016-01-20 16:14:55 UTC
The issue is not reproducible either if DMC firmware is not loaded or DMC is loaded and I prevent the display from timing out.

I.e. this does not actually block the OpenGL ES 3.1 CTS from passing on SKL.

It should be duplicated to some DMC kernel bug.
Comment 3 Marta Löfstedt 2016-01-21 08:53:49 UTC
Created attachment 121169 [details]
dmesg from when the problem occur.

dmesg from when the problem occur.
Comment 4 Marta Löfstedt 2016-02-02 09:23:50 UTC
This issue is reproducible with dmc version 1.26
Comment 5 Patrik Jakobsson 2016-02-19 12:11:18 UTC
I ran the tests again with Mesa 11.2.0-devel git-5d87a7c but got conflicting results with our previous findings. This time we had random fails on "ES31-CTS.layout_binding.block_layout_binding_block_ComputeShader" with both display on and off. This makes me think that we just got lucky with the previous testing and this is unrelated to DMC. More investigation is needed.


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.