VAAPI: FEATURE_FAILURE_VIDEO_DECODING_MISSING if media.ffmpeg.vaapi.enabled:true or media.hardware-video-decoding.force-enabled:true are set. VAAPI only works if it's enabled by default.
Categories
(Core :: Audio/Video: Playback, defect)
Tracking
()
Tracking | Status | |
---|---|---|
firefox-esr102 | --- | unaffected |
firefox113 | --- | unaffected |
firefox114 | --- | unaffected |
firefox115 | --- | verified |
People
(Reporter: elsandosgrande, Assigned: stransky)
References
(Blocks 1 open bug, Regression)
Details
(Keywords: nightly-community, regression)
Attachments
(3 files)
A few days ago, I noticed that Firefox had inexplicably stopped utilising hardware video decoding via VA-API — no Video
utilisation in intel_gpu_top and no messages from libva mentioning opening the iHD VA-API driver in either the journal or the terminal output. I played a video in VLC to double-check if the iHD driver was still functional and it very much was.
Today, I decided to look into it further and took a look at about:support
. There, I noticed some new rows in the decision log, [codec name]_HW_DECODE
, which all reported the same error, FEATURE_FAILURE_VIDEO_DECODING_MISSING
; ironically enough, the codec support information list right below the decision log lists all but AV1 as having hardware support, which is in line with what Ice Lake supports. After reaching out in the GFX Matrix room, I was pointed towards a ticket opened by Martin in which whatever method of probing for hardware codec capabilities was replaced by something called “vaapitest”. Whilst I have not used mozregression to verify that this is the root cause, I haven't touched any configuration options in the past two weeks and the overall timeline fits, so I was instructed to file a report without hesitation.
I will reproduce this issue with a fresh profile tomorrow or possibly after tomorrow.
Reporter | ||
Updated•1 year ago
|
Reporter | ||
Updated•1 year ago
|
Reporter | ||
Comment 1•1 year ago
|
||
Comment 2•1 year ago
|
||
:stransky, since you are the author of the regressor, bug 1831038, could you take a look?
For more information, please visit BugBot documentation.
Comment 3•1 year ago
|
||
Thanks for the report! Confirmed on KDE Wayland, Debian Testing, Intel.
H264 VAAPI is not used if VAAPI is force-enabled. It only works if it's enabled by default (in Nightly).
Regression range is correct.
last good: mozregression --repo autoland --launch 752dc76b26eb54e7dfa875c749e91526496688b6 --pref media.hardware-video-decoding.force-enabled:true -a about:support -a https://bug1619882.bmoattachments.org/attachment.cgi?id=9149605
first bad: mozregression --repo autoland --launch 6e6af58dbf2488e49eb9c1ae8b8df039c47f0d2e --pref media.hardware-video-decoding.force-enabled:true -a about:support -a https://bug1619882.bmoattachments.org/attachment.cgi?id=9149605
(In reply to Sandi Vujaković [:elsandosgrande] from comment #0)
Created attachment 9334225 [details]
about:support
text
Cleanup: Please consider setting these prefs back to their defaults:
gfx.swap-interval.glx
gfx.use-iosurface-textures
gfx.use-surfacetexture-textures
gfx.webrender.all
gfx.webrender.batched-texture-uploads
gfx.webrender.enabled
gfx.webrender.fallback.software
gfx.webrender.picture-tile-height
gfx.webrender.picture-tile-width
gfx.webrender.precache-shaders
gfx.webrender.program-binary-disk
gfx.webrender.software.d3d11
layers.force-shmem-tiles
layers.gpu-process.enabled
layers.gpu-process.force-enabled
layers.mlgpu.enabled
layers.omtp.enabled
media.ffmpeg.vaapi.enabled
media.ffvpx.enabled
media.ffvpx.mp3.enabled
media.gpu-process-decoder
media.rdd-ffvpx.enabled
media.utility-ffvpx.enabled
webgl.force-enabled
webgl.msaa-force
webgl.out-of-process.force
webgl.threadsafe-gl.force-enabled
widget.dmabuf-textures.enabled
Assignee | ||
Updated•1 year ago
|
Assignee | ||
Comment 4•1 year ago
|
||
Comment 7•1 year ago
|
||
bugherder |
Updated•1 year ago
|
Description
•