Bug 105483

Yep, that's the one failure I have no explanation for :(

Here is one from when we run the shards testlist on all BAT machines. The overview of the results are here:
https://intel-gfx-ci.01.org/tree/drm-tip/drmtip.html

https://intel-gfx-ci.01.org/tree/drm-tip/drmtip_1/fi-skl-6770hq/igt@perf@oa-exponents.html

https://intel-gfx-ci.01.org/tree/drm-tip/drmtip_4/fi-skl-gvtdvm/igt@perf@oa-exponents.html
https://intel-gfx-ci.01.org/tree/drm-tip/drmtip_4/fi-bsw-n3050/igt@perf@oa-exponents.html

(perf:1521) CRITICAL: Test assertion failure function test_oa_exponents, file ../tests/perf.c:1783:
(perf:1521) CRITICAL: Failed assertion: n_timer_reports - 3 <= matches
(perf:1521) CRITICAL: error: 27 > 26
Subtest oa-exponents failed.

Thanks, I sent out a patch to improve the test a bit further for this particular issue : https://lists.freedesktop.org/archives/igt-dev/2018-March/001806.html

Waiting for review.

(In reply to Lionel Landwerlin from comment #4)
> Thanks, I sent out a patch to improve the test a bit further for this
> particular issue :
> https://lists.freedesktop.org/archives/igt-dev/2018-March/001806.html
> 
> Waiting for review.

Since this was reproduced on drmtip runs, i.e. the shards testlist run on BAT machines. I would require an IGT hack patch bringing in the affected tests to tests/intel-ci/fast-feedback.testlist to get the tests run on the BAT machines.

https://intel-gfx-ci.01.org/tree/drm-tip/drmtip_11/fi-bxt-j4205/igt@perf@oa-exponents.html	

(perf:1839) CRITICAL: Test assertion failure function test_oa_exponents, file ../tests/perf.c:1783:
(perf:1839) CRITICAL: Failed assertion: n_timer_reports - 3 <= matches
(perf:1839) CRITICAL: error: 27 > 26
Subtest oa-exponents failed.

Also seen on KBL: https://intel-gfx-ci.01.org/tree/drm-tip/drmtip_76/fi-kbl-7560u/igt@perf@oa-exponents.html

(perf:2665) CRITICAL: Test assertion failure function test_oa_exponents, file ../tests/perf.c:1785:
(perf:2665) CRITICAL: Failed assertion: n_timer_reports - 3 <= matches
(perf:2665) CRITICAL: error: 27 > 26
Subtest oa-exponents failed.

https://intel-gfx-ci.01.org/tree/drm-tip/drmtip_128/fi-apl-guc/igt@perf@oa-exponents.html

Starting subtest: oa-exponents
(perf:1345) CRITICAL: Test assertion failure function test_oa_exponents, file ../tests/perf.c:1799:
(perf:1345) CRITICAL: Failed assertion: n_timer_reports - 3 <= matches
(perf:1345) CRITICAL: error: 27 > 25
Subtest oa-exponents failed.

Also seen in BDW
https://intel-gfx-ci.01.org/tree/drm-tip/drmtip_133/fi-bdw-gvtdvm/igt@perf@oa-
exponents.html

	
Starting subtest: oa-exponents
(perf:1112) CRITICAL: Test assertion failure function test_oa_exponents, file ../tests/perf.c:1799:
(perf:1112) CRITICAL: Failed assertion: n_timer_reports - 3 <= matches
(perf:1112) CRITICAL: error: 27 > 25
Subtest oa-exponents failed.

A CI Bug Log filter associated to this bug has been updated:

{- BSW BDW SKL BXT APL KBL GLK: igt@perf@oa-exponents - Failed assertion: n_timer_reports - 3 <= matches -}
{+ BSW BDW SKL BXT APL KBL GLK: igt@perf@oa-exponents - Failed assertion: n_timer_reports - 3 <= matches +}


  No new failures caught with the new filter

The test verifies different sample rates to capture reports into the OA buffer. Rates range approx. from 3.33 us to 109 ms. Every iteration is double the previous rate. Once 30 reports are captures, the timestamp delta between reports is expected to be equal to the sample rate. A tolerance of 3 errors in timestamp deltas (in a total of 29 deltas collected) is allowed.

A quick repro on APL suggests that the tolerance of 3 error deltas is a magic number. Changing it to 4 increases the tolerance for the mismatches on platforms like APL and allows the test to pass.

Machines: apl
Overall Repro rate: 22%

User Impact: None
Severity: Low
Pririty: Low

-- GitLab Migration Automatic Message --

This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity.

You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/drm/intel/issues/84.