Bug 101089

Summary:	[BAT][CTG] igt@gem_exec_suspend@basic-s3's execution time tripled after the pre-rc1 backmerge
Product:	DRI	Reporter:	Martin Peres <martin.peres>
Component:	DRM/Intel	Assignee:	Intel GFX Bugs mailing list <intel-gfx-bugs>
Status:	CLOSED WONTFIX	QA Contact:	Intel GFX Bugs mailing list <intel-gfx-bugs>
Severity:	critical
Priority:	high	CC:	intel-gfx-bugs
Version:	DRI git	Keywords:	bisect_pending
Hardware:	Other
OS:	All
Whiteboard:	ReadyForDev
i915 platform:	GM45	i915 features:	power/suspend-resume

Description Martin Peres 2017-05-18 12:48:21 UTC

The execution time of the igt@gem_exec_suspend@basic-s3 test on fi-ctg-l9400 went from 29.68s to 93.66s between CI_DRM_2605 and CI_DRM_2606. This is a separate problem from https://bugs.freedesktop.org/show_bug.cgi?id=101074 because it appeared a before.

Because of this problem, the IGT runs now randomly time out.

Here is the list of commits that were made between them: https://intel-gfx-ci.01.org/CI/CI_DRM_2606/commits_short.log

Comment 1 Ricardo 2017-05-23 16:30:48 UTC

Question: if this is a BAT bug should not be Highest? or because is random is considered only Medium?

Comment 2 Martin Peres 2017-05-23 16:32:38 UTC

Good observation Ricardo, I just forgot to make the change or to press "Save Changes"!

Comment 3 Petri Latvala 2017-05-30 11:34:42 UTC

Investigated this with a ctg-l9400 that was supposed to be an identical machine. Couldn't reproduce the issue at all.

With further investigation it was revealed that the suspend on the farm's ctg machine did not in fact take more time, but the machine itself had a clock skew during suspend. The CI timeouts are unrelated to the reported runtime changes.

Something happened in the kernel between CI_DRM_2605 and CI_DRM_2606 that caused the reported times to increase, but as even the reported time before was incorrect, the ctg in the CI farm has now been replaced with another one to unblock CI usage.

Marking this bug as NEEDINFO for bisecting the time increase between 2605 and 2606.

Comment 4 Martin Peres 2017-05-30 14:26:11 UTC

*** Bug 101074 has been marked as a duplicate of this bug. ***

Comment 5 Elizabeth 2017-08-18 20:55:59 UTC

Good afternoon, I noticed there is no update on this case since May. Is there any advance with this case? Thank you.
(In reply to Petri Latvala from comment #3)
> ...Something happened in the kernel between CI_DRM_2605 and CI_DRM_2606 that
> caused the reported times to increase, but as even the reported time before
> was incorrect, the ctg in the CI farm has now been replaced with another one
> to unblock CI usage.
> 
> Marking this bug as NEEDINFO for bisecting the time increase between 2605
> and 2606.

Comment 6 Petri Latvala 2017-10-10 14:45:08 UTC

Marking wontfix, the original issue is gone and bisecting the change in reported times without the affected hardware.

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.