VAAPI stopped working in FF99: FFmpeg 5.0/KDE/X11/Mesa 21 i965/Manjaro, Crash in [@ av_buffer_ref ]
Categories
(Core :: Audio/Video: Playback, defect, P3)
Tracking
()
People
(Reporter: fabriciol.castillo, Unassigned)
References
(Blocks 1 open bug)
Details
(Keywords: crash)
Crash Data
User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:99.0) Gecko/20100101 Firefox/99.0
Steps to reproduce:
I updated my system and went from FF98 to F99 and video acceleration was working fine (with the MOZ_DISABLE_RDD_SANDBOX=1 flag). Chromium v100 also stopped hardware acceleration but mpv and vlc work fine.
With h264fy installed and enabled, open a video in youtube and check gpu usage using intel_gpu_top
Actual results:
No activity registered in video decode indicator in intel_gpu_top, higher cpu usual than expected
Expected results:
Firefox should have used vaapi for decoding h264 streams as reported by vainfo and and as mpv and vlc do (both mpv and vlc use gpu for decoding video)
vainfo: VA-API version: 1.13 (libva 2.14.0)
vainfo: Driver version: Intel i965 driver for Intel(R) Sandybridge Mobile - 2.4.1
vainfo: Supported profile and entrypoints
VAProfileMPEG2Simple : VAEntrypointVLD
VAProfileMPEG2Main : VAEntrypointVLD
VAProfileH264ConstrainedBaseline: VAEntrypointVLD
VAProfileH264ConstrainedBaseline: VAEntrypointEncSlice
VAProfileH264Main : VAEntrypointVLD
VAProfileH264Main : VAEntrypointEncSlice
VAProfileH264High : VAEntrypointVLD
VAProfileH264High : VAEntrypointEncSlice
VAProfileH264StereoHigh : VAEntrypointVLD
VAProfileVC1Simple : VAEntrypointVLD
VAProfileVC1Main : VAEntrypointVLD
VAProfileVC1Advanced : VAEntrypointVLD
VAProfileNone : VAEntrypointVideoProc
This is my sistem info:
about:support -> https://pastebin.com/m1xURW0z
Packages: extra/libva 2.13.0-2 [installed]
Video Acceleration (VA) API for Linux
extra/libva-intel-driver 2.4.1-2 [installed]
VA-API implementation for Intel G45 and HD Graphics family
extra/libva-mesa-driver 21.3.8-1 [installed]
multilib/lib32-libva-intel-driver 2.4.1-1 [installed]
VA-API implementation for Intel G45 and HD Graphics family (32-bit)
multilib/lib32-libva-mesa-driver 21.3.8-2 [installed]
VA-API implementation for gallium (32-bit)
multilib/lib32-libva-vdpau-driver 0.7.4-6 [installed]
VDPAU backend for VA API (32-bit)
community/intel-gpu-tools 1.26-1 [installed]
Tools for development and testing of the Intel DRM driver
multilib/lib32-libva-intel-driver 2.4.1-1 [installed]
VA-API implementation for Intel G45 and HD Graphics family (32-bit)
multilib/lib32-vulkan-intel 21.3.8-2 [installed]
Intel's Vulkan mesa driver (32-bit)
community/libva-utils 2.14.0-1 [installed]
Intel VA-API Media Applications and Scripts for libva
community/intel-gpu-tools 1.26-1 [installed]
Tools for development and testing of the Intel DRM driver
|
||
Comment 1•2 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.
Hi, thanks for the info! Could you try starting Firefox with the following commandline?
MOZ_DISABLE_RDD_SANDBOX=1 MOZ_LOG="DmaBuf:5,PlatformDecoderModule:5" firefox
Then, on Youtube start a video that should be hardware-accelerated and then quickly pause it. If you could report back with the log, that would be helpful to diagnose. There'll be a lot of output, including VAAPI information -- for instance, on my machine I have among other info:
...
[AVHWDeviceContext @ 0x7f0163e5d4c0] VAAPI driver: Intel iHD driver for Intel(R) Gen Graphics - 22.2.1 ().
[AVHWDeviceContext @ 0x7f0163e5d4c0] Driver not found in known nonstandard list, using standard behaviour.
[h264 @ 0x7f0163e52500] nal_unit_type: 7(SPS), nal_ref_idc: 3
[h264 @ 0x7f0163e52500] nal_unit_type: 8(PPS), nal_ref_idc: 3
[RDD 2729526: MediaPDecoder #1]: D/PlatformDecoderModule FFMPEG: FFmpegVideoDecoder::GetAcceleratedFormats()
[RDD 2729526: MediaPDecoder #1]: D/PlatformDecoderModule FFMPEG: Profile H264Main:
[RDD 2729526: MediaPDecoder #1]: D/PlatformDecoderModule FFMPEG: codec h264 format nv12 3 12
[RDD 2729526: MediaPDecoder #1]: D/PlatformDecoderModule FFMPEG: codec h264 format nv12 3 12
...
|
Reporter | |
Comment 3•2 years ago
|
||
Hello, thanks for the reply. I ran the command you gave me with the additional MOZ_LOG_FILE=firefox.log to have that on file cause my cli was clipping the log and it was not complete. It returned several logfiles of which only a couple had anything in them (i asume other child processes weren't making use of the decoder).
RELEVANT: I also notice that the video would play only sometimes. It seems as if it's switching resolutions for some reason and it ends up with a lower one (<1080p) by default and sometimes this switching would result in an error and youtube would display a message saying there was an payback error. Again, this didn't happen in FF98
LOGS:
firefox.log.child-5.moz_log
firefox.log.child-8.moz_log
firefox.log.child-11.moz_log
firefox.log.child-13.moz_log
firefox.log.child-14.moz_log
|
|
Reporter | |
Comment 4•2 years ago
|
||
I tried again and got the output form the CLI directly (it's all together now and I can already see error messages)
Thank you for the info! I saw some lines that look interesting, among others:
[Child 24794, MediaDecoderStateMachine #1] WARNING: Decoder=7f0c2b2da600 Decode error: NS_ERROR_OUT_OF_MEMORY (0x8007000e): file /build/firefox/src/firefox-99.0.1/dom/media/MediaDecoderStateMachine.cpp:3577
[Parent 24608: Main Thread]: D/PlatformDecoderModule RDD shutdown before creating content bridge
... and VAAPI being reported as disabled despite it being force enabled in your config:
[RDD 24953: Main Thread]: D/PlatformDecoderModule PDMInitializer, Init PDMs in RDD process
[RDD 24953: Main Thread]: D/PlatformDecoderModule FFMPEG: Couldn't load function av_frame_get_colorspace
[RDD 24953: Main Thread]: D/PlatformDecoderModule FFMPEG: Couldn't load function av_frame_get_color_range
[RDD 24953: Main Thread]: D/PlatformDecoderModule Agnostic decoder supports requested type
...
[RDD 24953: MediaSupervisor #1]: D/PlatformDecoderModule FFMPEG: FFmpegVideoDecoder::FFmpegVideoDecoder MIME video/avc Codec ID 27
[RDD 24953: MediaSupervisor #1]: D/PlatformDecoderModule FFMPEG: VAAPI is disabled by parent decoder module.
Martin, would you have any ideas here?
|
||
Comment 6•2 years ago
•
|
||
Report ID: bp-167d52e5-2613-48af-8ec0-10c600220427
FFmpeg 5.0 [@ av_buffer_ref ]
This is the same as bug 1759137 (its patch is going to be backed out by bug 1762725).
Description: Mesa DRI Intel(R) HD Graphics 3000 (SNB GT2)
Driver Vendor: mesa/i965
Driver Version: 21.3.8.0
MOZ_DISABLE_RDD_SANDBOX: 1
gfx.canvas.accelerated: true
|
||
Comment 7•2 years ago
|
||
(In reply to az [:az] from comment #5)
Thank you for the info! I saw some lines that look interesting, among others:
[Child 24794, MediaDecoderStateMachine #1] WARNING: Decoder=7f0c2b2da600 Decode error: NS_ERROR_OUT_OF_MEMORY (0x8007000e): file /build/firefox/src/firefox-99.0.1/dom/media/MediaDecoderStateMachine.cpp:3577
It's Bug 1761217.
The reason why RDD crashes is perhaps Bug 1762725 / Bug 1759137.
|
|
||
Comment 8•2 years ago
|
||
(fabriciol.castillo from comment #0)
about:support -> https://pastebin.com/m1xURW0z
Crash Reports for the Last 3 Days
Report ID: bp-167d52e5-2613-48af-8ec0-10c600220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-aa6c9d56-ebe3-4697-be18-5d2850220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-7883de56-0341-4ce7-9c47-2a0da0220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-90ec9a86-92c8-4d69-977b-e895c0220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-329d850b-0bdc-42ca-9f4c-d62d20220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-e9c46418-a300-4631-b695-90c910220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-804328a8-1c69-4b77-b870-bf3820220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-144f2756-e963-435b-90bf-0a96b0220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-4977d7dd-cd36-430e-95c0-011cf0220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-fd0ba362-932a-4539-b2d4-af33e0220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-bdda99f2-eda5-45fd-88ba-2a5340220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-54eb7ee0-c6c5-4b9f-92ac-753cb0220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-a82adc60-2d2b-4e22-84ce-34ef10220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-37a9f384-a7de-40ef-95c5-4421d0220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-5281d007-9883-4a1b-9d59-27e910220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-48d5e061-aa32-44c5-a968-9acb10220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-8205a944-25db-4bf4-822e-588640220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-249df353-14c9-4174-8ee8-74fcc0220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-c461f39d-4c6a-4c9d-8f0d-23a4b0220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-a6648735-9e49-468a-adf9-5583a0220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-f9beda3a-b6f3-4c49-81a0-e82df0220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-1a49b93d-3845-4c3c-ab42-aa8a00220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-d6992ec6-1efb-4ff4-a980-6270d0220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-687544fe-a578-4a79-9850-8013c0220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-5799b28b-b0b7-48bb-967b-b51e10220427 [@ av_buffer_ref ]
Submitted: 1 hour agoReport ID: bp-9b6e3a7a-47a8-4ed5-9889-ae4370220427 [@ av_buffer_ref ]
Submitted: 1 hour ago
Description
•