log and screenshot:
https://pantheon.corp.google.com/storage/browser/chromiumos-test-logs/bugfiles/cr/687290/

1

marcheu@ is this something that would fall on your plate? If not, can you please suggest a more appropriate owner?

marcheu@ Can you please suggest an alternative owner?

Daniele, I see the following:

[4215:4227:0209/152524.515565:ERROR:hardware_display_plane_manager.cc(253)] Failed to find a free plane for crtc 19
[4215:4227:0209/152524.515671:ERROR:crtc_controller.cc(102)] Failed to assign overlay planes for crtc 19: Invalid argument

I think that's related to overlay work, right?

dcastagna@ can you please review and let us know if this is something you would work on? this is currently beta blocking.

I'm looking into this.

The first ChromeOS version where we see this issue is R57-9148.0.0, before that it seems to work fine.

I suspect this is related to ChromeOS code and not Chrome code since if I run ToT Chrome on top of R57-9147.0.0 I can't reproduce the issue.

Just looking at https://crosland.corp.google.com/log/9147.0.0..9148.0.0 I don't see what could have caused it. 

I'll investigate a little bit more.

Also, this is not only on Butterfly, as Stephane noticed this probably applies to all SandyBridge devices. I'm reproducing the problem on a Lumpy.

I was wrong in #8. The regression was introduced by a Chrome CL (crrev.com/2603863002)

crrev.com/2693903003 should fix the issue.

 Issue 691752  has been merged into this issue.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/e2b768afd564f860a3dc32db6e0114ca6e0db9f6

commit e2b768afd564f860a3dc32db6e0114ca6e0db9f6
Author: dcastagna <dcastagna@chromium.org>
Date: Mon Feb 13 22:41:25 2017

ozone: Fix HardwareDisplayController::AddCrtc.

crrev.com/2603863002 replaced ScopedPtrHashMap with unordered_map.

The CL replaced the statement
  owned_hardware_planes_.add(drm, plane_list)
with
  owned_hardware_planes_[drm.get()] = plane_list

While ScopedPtrHashMap::add did nothing if the key was already present,
the new operation on the unordered_map is now always resetting the list
of planes every time the statement is executed. This new unintented
behavior introduced a regression ( crbug.com/687290 ).

This CL restores the original intended behavior.

BUG= 687290 

Review-Url: https://codereview.chromium.org/2693903003
Cr-Commit-Position: refs/heads/master@{#450125}

[modify] https://crrev.com/e2b768afd564f860a3dc32db6e0114ca6e0db9f6/ui/ozone/platform/drm/gpu/hardware_display_controller.cc

Approving merge to M57 Chrome OS.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/9e6372cfc110717015ac2249bf13961987536fca

commit 9e6372cfc110717015ac2249bf13961987536fca
Author: Daniele Castagna <dcastagna@chromium.org>
Date: Tue Feb 14 18:48:47 2017

ozone: Fix HardwareDisplayController::AddCrtc.

crrev.com/2603863002 replaced ScopedPtrHashMap with unordered_map.

The CL replaced the statement
  owned_hardware_planes_.add(drm, plane_list)
with
  owned_hardware_planes_[drm.get()] = plane_list

While ScopedPtrHashMap::add did nothing if the key was already present,
the new operation on the unordered_map is now always resetting the list
of planes every time the statement is executed. This new unintented
behavior introduced a regression ( crbug.com/687290 ).

This CL restores the original intended behavior.

BUG= 687290 

Review-Url: https://codereview.chromium.org/2693903003
Cr-Commit-Position: refs/heads/master@{#450125}
(cherry picked from commit e2b768afd564f860a3dc32db6e0114ca6e0db9f6)

Review-Url: https://codereview.chromium.org/2693043006 .
Cr-Commit-Position: refs/branch-heads/2987@{#504}
Cr-Branched-From: ad51088c0e8776e8dcd963dbe752c4035ba6dab6-refs/heads/master@{#444943}

[modify] https://crrev.com/9e6372cfc110717015ac2249bf13961987536fca/ui/ozone/platform/drm/gpu/hardware_display_controller.cc

kalin@ can you please validate this fix and close once build is ready?

It's still reproducible in 9202.27.0/57.0.2987.52. I'll check tomorrow in later build.  

Verified the fix in 9202.30.0/57.0.2987.54.