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

Issue 774304 link

Starred by 2 users

Issue metadata

Status: Archived
Owner: ----
Closed: Oct 15
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Chrome
Pri: 2
Type: Bug-Regression



Sign in to add a comment

CRTC reconfiguration freezes primary plane (R63-9991 regression)

Reported by marc.her...@intel.com, Oct 12 2017

Issue description

UserAgent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_12_6) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/61.0.3163.100 Safari/537.36
Platform: 9991.0.0 (Official Build) dev-channel reef test

Steps to reproduce the problem:
There are (at least) three different ways to reproduce the same issue:
- Switch to frecon with Ctrl-Alt-F2
- docking
- suspend/resume

They all involve some CRTC reconfiguration and they all end in the same frozen state

What is the expected behavior?
The primary plane doesn't freeze

What went wrong?
The primary plane shows a frozen image. The UI is still active, for instance it's still possible to pause and resume a video by clicking on it and hear the sound pausing and resuming.

Did this work before? Yes R63.9990

Chrome version: 63.0.3228.0  Channel: dev
OS Version: 9991.0
Flash Version: NA

The same behaviour has been reproduced on eve, reef and Kevin. Since Intel has no access to Kevin images, the latest on the dev channel was used: R63 build 10009 instead of 9991. However the behaviour is exactly the same.

I tried to compare the 9990 and 9991 manifests there but no change jumped out:
https://chromium.googlesource.com/chromiumos/manifest-versions/+/master/chrome-LKGM-candidates/buildspecs/63/
 
> - Switch to frecon with Ctrl-Alt-F2
> - docking/undocking
> - suspend/resume

Suspend/resume is the least likely to reproduce while switching to frecon reproduces after 10 tries or less.


On Intel eve and reef the following ERROR message is observed every time the freeze happens:

#tailf /var/log/chrome/chrome
[3231:3246:1012/133712.384260:WARNING:screen_manager.cc(114)] Display controller (crtc=35) already present.
[3231:3246:1012/133712.386604:VERBOSE1:drm_display.cc(102)] DRM configuring: device=/sys/devices/pci0000:00/0000:00:02.0/drm/card0 crtc=35 connector=53 origin=0,0 size=0x0
[3231:3246:1012/133712.387745:ERROR:hardware_display_plane_manager_atomic.cc(78)] Failed to commit properties for page flip.: Device or resource busy (16)

This is one of the few changes between 9990 and 9991 but it claims to be a no-op:
https://chromium-review.googlesource.com/c/chromiumos/overlays/chromiumos-overlay/+/685477/3/chromeos-base/chromeos-chrome/chromeos-chrome-9999.ebuild
Cc: puthik@chromium.org marc...@chromium.org bhthompson@chromium.org
This sounds like b/67714318 ?

Comment 3 by puthik@chromium.org, Oct 12 2017

Re #2.

This is M63 but that is M62. So different bug.
I was just talking to Marc, this happens in 62 also.
Russ reproduced this same issue on eve with the latest R62 about 4-5 days ago.

Comment 6 by puthik@chromium.org, Oct 13 2017

Probably duplicate of  Issue 769570  (fixed in R63-10021.0.0 / R62-9901.46.0)

Did you see it with newer version?
Components: OS>Kernel>Graphics
It looks similar, however R63-9991 is the very first image that reproduces on both eve and reef whereas  issue 769570  was reported against R63-9981.0.0

R63-9991 is dated from September 30th whereas  Issue 769570  was reported on September 28th.

Comment 9 by puthik@chromium.org, Oct 13 2017

Cc: dcasta...@chromium.org
+Daniele 

Did you know anything about log in #1 ?
Reproduced with 3230 (reef-release/R63-10020.0.0)
Note reproduction seems faster when "stressing" the system more with everything below:
- external monitor
- video playing
- Ctrl-Alt-F

CanNOT reproduce with 3238 (reef-release/R63-10029.0.0) built today, so I guess it's fixed. 

I still can't make sense of the dates and versions (like R63-9981) in  issue 769570  however, can someone? Maybe the reporter was testing a manual build?

A "HOWTO compare .xml manifests files" wouldn't hurt either.
https://chromium.googlesource.com/chromiumos/manifest-versions/+/master/chrome-LKGM-candidates/buildspecs/63/

> can't make sense of the dates and versions reported in  issue 769570  

I think the versions reported at the very top of  issue 769570  were just wrong. Everything else matches so this must be a duplicate, please close as such.

> A "HOWTO compare .xml manifests files"

Skipping many steps and simplifying the rest a lot (hello, David Wheeler :-)

diff -u 9990.0.0-rc4.xml 9991.0.0-rc2.xml
- path="src/third_party/chromiumos-overlay" revision="347e21ffc412...
+ path="src/third_party/chromiumos-overlay" revision="5ebe9b18de50...

cd src/third_party/chromiumos-overlay
git diff 347e21ffc..5ebe9b18de50 -- metadata/md5-cache/chromeos-base/
rename from metadata/md5-cache/chromeos-base/chromeos-chrome-63.0.3223.0_rc-r1
rename to   metadata/md5-cache/chromeos-base/chromeos-chrome-63.0.3228.0_rc-r2

Is that logic correct?
Project Member

Comment 12 by sheriffbot@chromium.org, Oct 15

Status: Archived (was: Unconfirmed)
Issue has not been modified or commented on in the last 365 days, please re-open or file a new bug if this is still an issue.

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot

Sign in to add a comment