Sounds like OOM or something like that.

Reporter: please provide us with the full contents of chrome://gpu after reproducing the bug - I'm guessing there should be some log entries that tell us what happened.

attached requested logs after reproduce issue

Deleted: logs.txt 14.4 KB

logs.txt 14.4 KB View Download

Daniele, I think that's a duplicate of the addfb bug:

[10119:10136:1212/054726.998696:ERROR:gbm_buffer.cc(73)] : AddFramebuffer2 failed: Invalid argument (22)
[10119:10136:1212/054727.002005:ERROR:hardware_display_plane_manager_atomic.cc(78)] : Failed to commit properties for page flip.: Invalid argument (22)

Maybe this gives us a repro case?

Thank you for providing more feedback. Adding requester "kainino@chromium.org" to the cc list and removing "Needs-Feedback" label.

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

Daniele, I think that's a duplicate of the addfb bug:

May i know which issue is this , where its getting tracked.

issue reproduce in below scenario

"Rat! WebGL hit snag" issue observed when we do HDMI plug/unplug or switching between extended to mirror mode with any gfx flash demo running
below are the sample flash demo:
http://www.ambiera.com/coppercube/demo.php?demo=backyard&mode=flash
http://9elements.com/io/projects/html5/canvas/
http://www.clicktorelease.com/code/gif/

from logs we dont see any failure of  AddFramebuffer2 failed

I think we might be observing two issues here.
The first one might be that the GPU process crashes, maybe for OOM, maybe for other reasons.

Then we try to reimport buffers, and at that time we hit the second issue where we fail with "AddFramebuffer2 failed", that might end up in failing to pageflip, similar to what we observed in crbug.com/789292

I suspected the failure of adding a framebuffer is triggered by the GPU crash, but the first GPU crash is not caused by failing to add the framebuffer.

Lowest memory observed at 1216th iteration during one of our experiments.

MemTotal: 3937312 kB
MemFree: 196720 kB
MemAvailable: 94872 kB

After this immediate memory recovered . From top command logs, webgl stopped at this point.

Add framebuffer2 issue is sporadically observed while opening 50 different web sites on Chrome platforms that results in screen blank for few secs and  it automatically recovers back

[VERBOSE1:drm_display.cc(105)] DRM configuring: device=/sys/devices/pci0000:00/0000:00:02.0/drm/card0 crtc=36 connector=58 origin=0,0 size=2400x1600
[ERROR:gbm_buffer.cc(73)] AddFramebuffer2 failed: Invalid argument (22)
[ERROR:hardware_display_plane_manager_atomic.cc(78)] Failed to commit properties for page flip.: Invalid argument (22)

ping

Are we always running out of memory when you see this? If so we could have a leak, in which case we should track it down...

This looks like a dup of bug 789292. This should be fixed in the latest 65 build (10323.31.0 or later). It hasn't landed in ToT yet because of an accidental revert, but as soon as a 66 build with Chrome 66.0.3344.0 or later becomes available that should fix it.

I'm going to close this as a dup, please reopen if the specified version or later don't fix the problem.