VAAPI/AMD Rembrandt: artifacts, video stops, crashes
Categories
(Core :: Audio/Video: Playback, defect, P3)
Tracking
()
| Tracking | Status | |
|---|---|---|
| firefox111 | --- | disabled |
People
(Reporter: mwolf, Unassigned)
References
(Blocks 1 open bug)
Details
Attachments
(2 files)
Stats of the video watched on FF 113 a1 where the video froze.
User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:109.0) Gecko/20100101 Firefox/110.0
Steps to reproduce:
- open youtube
- play any video
- wait for artifacts / video stopping (audio continues to play)
3a (sometimes the entire amdgpu driver hangs
Actual results:
artifacts, video stopping for a few seconds while audio is still playing, crashes of amdgpu driver.
I already filed a bug report at freedesktop.org:
https://gitlab.freedesktop.org/mesa/mesa/-/issues/8303
It contains the infos about my system and the error itself.
Expected results:
normal playback without stutter, hangs or crashes
|
||
Comment 1•3 years ago
|
||
The Bugbug bot thinks this bug should belong to the 'Core::Audio/Video: Playback' component, and is moving the bug to that component. Please correct in case you think the bot is wrong.
Please attach your about:support page. May be dupe of Bug 1802844.
Can you try latest nightly?
https://fedoraproject.org/wiki/How_to_debug_Firefox_problems#Testing_Mozilla_binaries
Thanks.
(In reply to Martin Stránský [:stransky] (ni? me) from comment #3)
Can you try latest nightly?
https://fedoraproject.org/wiki/How_to_debug_Firefox_problems#Testing_Mozilla_binaries
Thanks.
I will give it a try after work.
Can you try to disable HW video decoding?
https://fedoraproject.org/wiki/Firefox_Hardware_acceleration#Video_decoding
Thanks.
without HW video decoding, the problem is gone.
(In reply to Martin from comment #7)
without HW video decoding, the problem is gone.
Good. Then please try the latest nightly. We need to check if your GPU works ok with disabled zero copy playback.
I tried in my lunch-break.
With the latest nightly it feels better, but I think I had a few rare hickups with 60FPS videos.
We'd need to update blocklist then, Thanks.
|
Reporter | |
Comment 11•3 years ago
|
||
it makes sense, since the gpu in the other bug report is also RDNA2
|
|
||
Comment 12•3 years ago
|
||
The severity field is not set for this bug.
:jimm, could you have a look please?
For more information, please visit auto_nag documentation.
Looks like VA-API drivers are not ready yet.
Martin, can you please try latest nightly?
https://fedoraproject.org/wiki/How_to_debug_Firefox_problems#Testing_Mozilla_binaries
Thanks.
|
||
Updated•3 years ago
|
|
|
Reporter | |
Comment 15•3 years ago
|
||
I will run it tonight. About the crashes, they are very random and rare. I am not sure, if I can capture one.
But I will focus on the video issues. Ty for your effort!
|
|
||
Comment 16•3 years ago
|
||
The bug has a release status flag that shows some version of Firefox is affected, thus it will be considered confirmed.
|
|
Reporter | |
Comment 17•3 years ago
|
||
I tested with the latest nightly and I have no hickups or artifacts.
|
|
Reporter | |
Comment 18•3 years ago
|
||
I have to revert my comment above. It happens much rarer but the video sometimes freezes.
(In reply to Martin from comment #18)
I have to revert my comment above. It happens much rarer but the video sometimes freezes.
So looks like Bug 1802844 helped only partially here.
|
|
Reporter | |
Comment 20•3 years ago
|
||
I watched 6 videos, where no freezes occured, but then I tested these shorts or reels, where I ran into a freeze basically every fith "short".
I am not sure if there is a coincidence. I will test tonight more.
Can you test mpv video player in VA-API mode?
mpv --hwdec=vaapi test_clip
|
|
Reporter | |
Comment 22•3 years ago
|
||
I will test with your suggested settings tomorrow.
Btw. it happened today with this particular regular video on youtube as well:
https://www.youtube.com/watch?v=BbybZyj_5PM
VP9 + Opus Codec
|
|
Reporter | |
Comment 23•3 years ago
|
||
One more question:
Does it matter what Video I watch on the player? I think I might have a mp4 .
|
|
Reporter | |
Comment 24•3 years ago
|
||
On youtube I had major issues with this video (firefox nightly)
https://www.youtube.com/watch?v=5TMx5ibD6pI
downloaded with yt-dlp and played it in mpv
the issues persisted also with
mpv --hwdec=vaapi video.webm
(+) Video --vid=1 (*) (vp9 1920x1080 23.976fps)
(+) Audio --aid=1 --alang=eng (*) (opus 2ch 48000Hz)
[vo/gpu/wayland] GNOME's wayland compositor lacks support for the idle inhibit protocol. This means the screen can blank during playback.
Using hardware decoding (vaapi).
AO: [pipewire] 48000Hz stereo 2ch floatp
VO: [gpu] 1920x1080 vaapi[nv12]
AV: 00:00:33 / 00:02:46 (20%) A-V: 0.000
If you see the bugs with mpv it's definitely issue with your gfx drivers so Mesa folks should fix that.
|
|
Reporter | |
Comment 26•3 years ago
|
||
I understand, would it be possible, that you write that on the mesa bug report please?
https://gitlab.freedesktop.org/mesa/mesa/-/issues/8303
(In reply to Martin from comment #26)
I understand, would it be possible, that you write that on the mesa bug report please?
https://gitlab.freedesktop.org/mesa/mesa/-/issues/8303
There's no point to poke them as I don't have such hardware.
Mesa 23.1.1 contains AMD fix for surfaces (https://gitlab.freedesktop.org/mesa/mesa/-/issues/8996). Can you try that? It's already shipped in Fedora 38 for instance.
Thanks.
|
|
Reporter | |
Comment 29•3 years ago
|
||
I will try over the weekend.
|
|
||
Comment 31•3 years ago
|
||
(In reply to Martin from comment #30)
sadly the problem is still present.
Does the problem still occur after setting media.ffmpeg.vaapi.force-surface-zero-copy=1, webgl.threadsafe-gl.force-disabled=true, gfx.canvas.accelerated=false (plus if you test Nightly: webgl.out-of-process.async-present.force-sync=true to avoid bug bug 1831548) on about:config and restarting Firefox?
|
|
Reporter | |
Comment 32•3 years ago
|
||
I will give it a try in the next few days.
|
|
Reporter | |
Comment 33•2 years ago
|
||
sorry, I was sick. I'm testing now. Will let you know (I used every setting except the one for nightly)
|
|
Reporter | |
Comment 34•2 years ago
|
||
@Darkspirit
sadly it does not help.
|
||
Comment 35•2 years ago
|
||
I expect this should be fixed by the updated VCN binaries for Rembrandt that had fixed some decoding problems reported in https://gitlab.freedesktop.org/mesa/mesa/-/issues/8044
Here is the commit:
https://gitlab.com/kernel-firmware/linux-firmware/-/commit/977332782302476e1c863b09b840f463d0378807
Martin, can you please re-test?
Thanks.
|
|
Reporter | |
Comment 37•2 years ago
|
||
I will give it a try.
I reenabled VP9. Will let you know in a few days.
|
|
Reporter | |
Comment 38•2 years ago
|
||
I tested for a few weeks now.
The problem seems to be gone.
Thank you!
|
||
Comment 39•1 year ago
|
||
Hi Martin, if problem is gone, can we close this issue?
Okay, closing.
Description
•