Bug 93635

Summary:	[BDW] Updating to linux 4.3 produces intermittent piglit.spec.arb_shader_image_load_store.qualifiers failures
Product:	DRI	Reporter:	Mark Janes <mark.a.janes>
Component:	DRM/Intel	Assignee:	Intel GFX Bugs mailing list <intel-gfx-bugs>
Status:	CLOSED FIXED	QA Contact:	Intel GFX Bugs mailing list <intel-gfx-bugs>
Severity:	normal
Priority:	medium	CC:	intel-gfx-bugs
Version:	DRI git
Hardware:	x86 (IA32)
OS:	Linux (All)
Whiteboard:
i915 platform:	BDW	i915 features:	GEM/Other

Description Mark Janes 2016-01-07 20:56:31 UTC

System is BDWGT3e
Mesa/DRM versions are tip of master
test passes reliably on Linux 4.1 (stock debian testing kernel)
Stock 4.3.0 kernel fails this test intermittently

Test fails about 20% of the time, when run in Mesa CI (parallel execution, gbm back end).

Sample output:
/tmp/build_root/m64/lib/piglit/bin/arb_shader_image_load_store-qualifiers -auto -fbo
piglit: debug: Requested an OpenGL 3.2 Core Context, and received a matching 3.3 context

Probe value at (1, 0, 0, 0)
  Expected: 21.000000
  Observed: 0.000000
Probe value at (12, 0, 0, 0)
  Expected: 1.000000
  Observed: 0.000000

Let me know if more information is needed to reproduce this.

Comment 1 Chris Wilson 2016-01-07 21:59:54 UTC

What's the failure rate when running the testcase standalone? (Confirming that it fails by itself would also be useful.)

Comment 2 Chris Wilson 2016-01-07 22:17:54 UTC

Only BDWGT3e?

Hmm, my GT3 is just an Iris 6100 not GT3e. That doesn't seem to be tripping up on -nightly.

Comment 3 Mark Janes 2016-01-07 22:40:46 UTC

BDWGT3e was the test machine I used.  I can verify BDWGT2, I expect it to be the same.

I have had terrible luck trying nightly in the past.  Is it healthy ATM, or is there a known-good sha I can use?

I will run it individually as you requested.

Comment 4 Mark Janes 2016-01-08 01:42:28 UTC

running by itself, it fails on BDWGT2 about 10% of the time.

The subtest causing the failure is:
PIGLIT: {"subtest": {"r16ui/permissive layout qualifiers/strict access qualifiers/permissive binding test" : "pass"}}
Probe value at (1, 0, 0, 0)
  Expected: 21.000000
  Observed: 0.000000

Comment 5 Mark Janes 2016-01-08 01:44:24 UTC

(In reply to Mark Janes from comment #4)
> running by itself, it fails on BDWGT2 about 10% of the time.
> 
> The subtest causing the failure is:
> PIGLIT: {"subtest": {"r16ui/permissive layout qualifiers/strict access
> qualifiers/permissive binding test" : "pass"}}
> Probe value at (1, 0, 0, 0)
>   Expected: 21.000000
>   Observed: 0.000000

BTW, this is still with stock Linux 4.3

Comment 6 Jani Nikula 2016-04-25 09:50:12 UTC

Mark, still an issue with latest kernels?

Comment 7 Mark Janes 2016-04-25 13:59:08 UTC

This bug does not reproduce on later kernels.

Comment 8 Jani Nikula 2016-04-25 14:14:39 UTC

Thanks for the follow-up!

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.