New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 833580 link

Starred by 3 users

Issue metadata

Status: Assigned
Owner:
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Chrome
Pri: 2
Type: Bug



Sign in to add a comment

[Broadwell] graphics_Idle.arc Failed: Did not reach 0 gem actives.

Project Member Reported by ihf@chromium.org, Apr 16 2018

Issue description

On Broadwell only, and graphics_Idle.arc only (not without .arc) we periodically see intermittent failures (1 in 20) stating
Failed: Did not reach 0 gem actives.

https://cros-goldeneye.corp.google.com/chromeos/healthmonitoring/testDetails?testName=graphics_Idle
https://pantheon.corp.google.com/storage/browser/chromeos-autotest-results/192190009-chromeos-test/chromeos4-row8-rack4-host20

04/14 15:59:36.311 INFO |     graphics_Idle:0358| Running verify_graphics_gem_idle
04/14 16:00:36.594 ERROR|     graphics_Idle:0074| Error: Did not reach 0 gem actives. 
04/14 16:00:36.604 INFO |     graphics_Idle:0078| Content of /sys/kernel/debug/dri/0/i915_gem_active
Active:
   ffff880075186c00:  Y       128KiB 36 00 59365 0 0 snooped or LLC dirty (ggtt offset: 078f2000, size: 00020000) (render ring)
   ffff88005d621c80:            4KiB 02 02 59365 59365 0 snooped or LLC dirty (ggtt offset: 011c6000, size: 00001000) (render ring)
   ffff88004e8af080:           16KiB 37 00 59365 0 0 snooped or LLC purgeable (ggtt offset: 04ed7000, size: 00004000) (render ring)
   ffff8800723ab800:           32KiB 37 00 59365 0 0 snooped or LLC dirty (ggtt offset: 03172000, size: 00008000) (render ring)
   ffff880075e5e480: pXg     4224KiB 42 00 59365 59365 0 uncached dirty (pinned x 1) (display) (fence: 11) (ggtt offset: 0409d000, size: 00420000) (p mappable) (render ring)
   ffff88004e902d80:  Y       256KiB 36 00 59365 0 0 snooped or LLC dirty (ggtt offset: 03c61000, size: 00040000) (render ring)
   ffff88007702fe00:           64KiB 37 00 59365 0 0 snooped or LLC (ggtt offset: 052e2000, size: 00010000) (render ring)
   ffff880003c8bb00:            4KiB 37 00 59365 0 0 snooped or LLC dirty (ggtt offset: 015eb000, size: 00001000) (render ring)
   ffff880003c8a480:            4KiB 37 00 59365 0 0 snooped or LLC dirty (ggtt offset: 015ec000, size: 00001000) (render ring)
   ffff880075187200:  Y       256KiB 36 00 59365 0 0 snooped or LLC dirty (ggtt offset: 07912000, size: 00040000) (render ring)
   ffff88003a992000:           20KiB 3f 00 59365 0 0 snooped or LLC dirty (ggtt offset: 04d07000, size: 00005000) (render ring)
Total 11 objects, 5128192 bytes, 5128192 GTT size

 
We should probably modify the test to dump a snapshot of "top" on failure, so that we can know what's actually running.
Ilja informed me that this is already the case :) In one of the logs I found a possible culprit:

 7373 665379    20   0 1176148  70556  50432 S  94.4  3.6   0:04.57 com.androi+


Looking for pid 7373 in the file, it is com.android.vending:

5 665379 7373  6022  20   0 1176148 70556 ep_pol Sl  ?          0:04 com.android.vending

So it seems like the test is doing its job and detecting a runaway process.
Status: Assigned (was: Untriaged)
This bug has an owner, thus, it's been triaged. Changing status to "assigned".

Sign in to add a comment