[Bug]: FFmpeg vp9_qsv encoder MFXVideoCORE_SyncOperation return MFX_ERR_NOT_ENOUGH_BUFFER

[feiwan1]

Which component impacted?

Encode

Is it regression? Good in old configuration?

None

What happened?

1. In Linux use master code of gmmlib/libva/media-driver/onevpl/onevpl-gpu-rt/ffmpeg.

2. Run the 1:N cmd: ffmpeg -v verbose \ -init_hw_device vaapi=intel:/dev/dri/renderD128 \ -hwaccel vaapi -hwaccel_output_format vaapi \ -hwaccel_device intel -filter_hw_device intel \ -loglevel info \ -fflags +genpts \ -i "meridian_hdr.mp4" \ -filter_complex \ "[0:v]hwmap=derive_device=qsv,format=qsv,vpp_qsv=deinterlace=2:async_depth=1,split=5[s1][s2][s3][s4][s5]; \ [s1]vpp_qsv=w=1920:h=1080:async_depth=1:scale_mode=1[v1]; \ [s2]vpp_qsv=w=1280:h=720:async_depth=1:scale_mode=1[v2]; \ [s3]vpp_qsv=w=960:h=540:async_depth=1:scale_mode=1[v3]; \ [s4]vpp_qsv=w=768:h=432:async_depth=1:scale_mode=1[v4]; \ [s5]vpp_qsv=w=640:h=360:async_depth=1:scale_mode=1[v5]" \ -map "[v1]" -c:v vp9_qsv -b:v 5000k -minrate:v 5000k -maxrate:v 5000k -bufsize:v 1250k -rc_init_occupancy:v 625k \ -g:v 120 -async_depth:v 1 -refs:v 5 -bf:v 0 -preset:v medium -strict:v 1 -low_power:v 1 \ -flags +global_header+cgop -f mp4 -y "1080p-vp9_qsv.mp4" \ -map "[v2]" -c:v vp9_qsv -b:v 4000k -minrate:v 4000k -maxrate:v 4000k -bufsize:v 1000k -rc_init_occupancy:v 500k \ -g:v 120 -async_depth:v 1 -refs:v 5 -bf:v 0 -preset:v medium -strict:v 1 -low_power:v 1 \ -flags +global_header+cgop -f mp4 -y "720p-vp9_qsv.mp4" \ -map "[v3]" -c:v vp9_qsv -b:v 2500k -minrate:v 2500k -maxrate:v 2500k -bufsize:v 625k -rc_init_occupancy:v 312k \ -g:v 120 -async_depth:v 1 -refs:v 5 -bf:v 0 -preset:v medium -strict:v 1 -low_power:v 1 \ -flags +global_header+cgop -f mp4 -y "540p-vp9_qsv.mp4" \ -map "[v4]" -c:v vp9_qsv -b:v 1500k -minrate:v 1500k -maxrate:v 1500k -bufsize:v 375k -rc_init_occupancy:v 187k \ -g:v 120 -async_depth:v 1 -refs:v 5 -bf:v 0 -preset:v medium -strict:v 1 -low_power:v 1 \ -flags +global_header+cgop -f mp4 -y "432p-vp9_qsv.mp4" \ -map "[v5]" -c:v vp9_qsv -b:v 1000k -minrate:v 1000k -maxrate:v 1000k -bufsize:v 250k -rc_init_occupancy:v 125k \ -g:v 120 -async_depth:v 1 -refs:v 5 -bf:v 0 -preset:v medium -strict:v 1 -low_power:v 1 \ -flags +global_header+cgop -f mp4 -y "360p-vp9_qsv.mp4"

3. See the error log:

[vp9_qsv @ 0x5a82e1403540] Invalid FrameType:0 [vost#4:0/vp9_qsv @ 0x5a82e1402840] [enc:vp9_qsv @ 0x5a82e1403440] Error submitting video frame to the encoder [vost#4:0/vp9_qsv @ 0x5a82e1402840] [enc:vp9_qsv @ 0x5a82e1403440] Error encoding a frame: Invalid data found when processing input [vost#0:0/vp9_qsv @ 0x5a82e13ee2c0] Task finished with error code: -1094995529 (Invalid data found when processing input) [vost#0:0/vp9_qsv @ 0x5a82e13ee2c0] Terminating thread with return code -1094995529 (Invalid data found when processing input) frame=21360 fps= 38 q=-0.0 Lq=-0.0 q=-0.0 q=-0.0 q=-0.0 size= 220461KiB time=00:05:56.33 bitrate=5068.3kbits/s speed=0.635x Conversion failed!

4. And checked the err is caused by: https://github.com/FFmpeg/FFmpeg/blob/96d45c3b212689f82bff2530c3637405df9e9369/libavcodec/qsvenc.c#L2663 The MFXVideoCORE_SyncOperation returns MFX_ERR_NOT_ENOUGH_BUFFER. And the vpl doc doesn't mention this API can and why return MFX_ERR_NOT_ENOUGH_BUFFER: https://intel.github.io/libvpl/latest/API_ref/VPL_func_vidcore.html#_CPPv426MFXVideoCORE_SyncOperation10mfxSession12mfxSyncPoint6mfxU32

5. The issue only happens on vp9_qsv. hevc_qsv and h264_qsv looks good. And this issue is a re-map of https://github.com/intel/cartwheel-ffmpeg/issues/360

What's the usage scenario when you are seeing the problem?

Transcode for media delivery

What impacted?

No response

Debug Information

No response

Do you want to contribute a patch to fix the issue?

None

[Brainiarc7]

Any updates here?

[intel-mediadev]

Auto Created VSMGWL-77181 for further analysis.

[Brainiarc7]

Any updates here?

[Sherry-Lin]

Any issues if only 1:1 cmd?

[Brainiarc7]

Yes, the same problem occurs

On Mon, 3 Mar 2025, 4:32 am Sherry Lin, @.***> wrote:

Any issues if only 1:1 cmd?

— Reply to this email directly, view it on GitHub https://github.com/intel/vpl-gpu-rt/issues/353#issuecomment-2693057181, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAR7NB22AHOWEFLV7HOUG4D2SOWMFAVCNFSM6AAAAABRRD7UOSVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMOJTGA2TOMJYGE . You are receiving this because you commented.Message ID: @.***> [image: Sherry-Lin]Sherry-Lin left a comment (intel/vpl-gpu-rt#353) https://github.com/intel/vpl-gpu-rt/issues/353#issuecomment-2693057181

Any issues if only 1:1 cmd?

— Reply to this email directly, view it on GitHub https://github.com/intel/vpl-gpu-rt/issues/353#issuecomment-2693057181, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAR7NB22AHOWEFLV7HOUG4D2SOWMFAVCNFSM6AAAAABRRD7UOSVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMOJTGA2TOMJYGE . You are receiving this because you commented.Message ID: @.***>