Firefox Freezes Computer
Categories
(Core :: Graphics, defect, P3)
Tracking
()
| Performance Impact | none |
People
(Reporter: sdaswani, Unassigned)
Details
Comment 1•7 years ago
|
||
Updated•7 years ago
|
Comment 2•7 years ago
|
||
Comment 6•7 years ago
|
||
Comment 7•7 years ago
|
||
| Reporter | ||
Comment 10•7 years ago
|
||
| Reporter | ||
Comment 11•7 years ago
|
||
| Reporter | ||
Comment 13•7 years ago
|
||
| Reporter | ||
Comment 14•7 years ago
|
||
Comment 15•7 years ago
|
||
Comment 16•7 years ago
|
||
Comment 17•7 years ago
|
||
Comment 18•7 years ago
|
||
| Reporter | ||
Comment 19•7 years ago
|
||
Nope, this profile seems to have disappeared. Back on me to try to get another profile. :/
| Reporter | ||
Comment 20•7 years ago
|
||
Great news - I opened up FF afresh this evening, opened gmail, open bugzilla, and then I encountered the freeze, and this time I have the profile!
See a screenshot and a gz of the profile here: https://drive.google.com/file/d/1Cabps0pKlOhFfHo232JE7DShzhDdeUMW/view?usp=sharing
https://drive.google.com/file/d/1rmE8Q2bAVLLwA8c5peuT_l_lfqEiXdPo/view?usp=sharing
Comment 21•7 years ago
|
||
I see the main process is waiting for something from the APZ thread (or process), see also the marker "PAPZInputBridge::Msg_ReceiveKeyboardInputEvent".
Can we profile the APZ thread somehow?
Comment 22•7 years ago
|
||
Thanks, Susheel!
Maybe kats has thoughts on comment 21.
Comment 23•7 years ago
|
||
I put the profile here for easier access: https://perfht.ml/2VLTChM
The main thread is indeed stuck in PAPZInputBridge::Msg_ReceiveKeyboardInputEvent for ~50 seconds, preceded by a ~10s block in FlushRendering. Both of these attempt to talk to the GPU process, and in the profile it looks like GPU process threads only start after the big hang. (Although technically the APZ message goes to the main thread in the GPU process, which isn't included in the profile).
As a first guess there might be a GPU process restart happening here, and the profile is only picking up stuff from the "new" GPU process and throwing away the old stuff. Markus, do you know if that's what happens if you are collecting a profile across a GPU process reset?
At any rate, the hang ends once the new GPU process is up, so the more interesting question is what happens in the "old" GPU process, and what triggers that, but I don't think we can answer that from this profile.
It would also be useful to look at about:support after one of these hangs and see if there are any useful error messages reported in the graphics section.
Comment 24•7 years ago
|
||
As a first guess there might be a GPU process restart happening here, and the profile is only picking up stuff from the "new" GPU process and throwing away the old stuff. Markus, do you know if that's what happens if you are collecting a profile across a GPU process reset?
We only get the profile from a GPU process that's shutting down if it shuts down regularly and has a chance to report its shutdown profile. However, if the GPU process crashes, the profile information from it is lost.
Maybe there's a crash report on about:crashes for the GPU process crash? It might help.
Comment 25•7 years ago
|
||
This sounds like a GPU process misbehaviour and then crash, so moving to the Graphics component. Our responsiveness is poor, but I suspect we should probably try to avoid the misbehaviour and crash rather than trying to make our reaction to this misbehaviour better. So marking qf- for now.
Updated•7 years ago
|
Updated•7 years ago
|
Updated•4 years ago
|
Updated•3 years ago
|
Description
•