Steps to reproduce:

On Debian sid/unstable with Nightly 20230123094444, watch YouTube video with h264ify.

Actual results:

The fan started spinning. Looking at intel_gpu_top, the video engine is not used. The messages below are shown on the console:

libva info: VA-API version 1.17.0
libva info: Trying to open /usr/lib/x86_64-linux-gnu/dri/iHD_drv_video.so
libva info: Found init function __vaDriverInit_1_17
libva info: va_openDriver() returns 0
Unsupported modifier, resource creation failed.

Expected results:

The video engine should be used as in the past.

No idea, if it’s a regression in Nightly or due to upgrading intel-media-va-driver-non-free from 22.6.6+ds1-1 to 23.1.0+ds1-1.

VA-API decoding still works with mpv 0.35.0-4.

$ vainfo
libva info: VA-API version 1.17.0
libva info: Trying to open /usr/lib/x86_64-linux-gnu/dri/iHD_drv_video.so
libva info: Found init function __vaDriverInit_1_17
libva info: va_openDriver() returns 0
vainfo: VA-API version: 1.17 (libva 2.12.0)
vainfo: Driver version: Intel iHD driver for Intel(R) Gen Graphics - 23.1.0 ()
vainfo: Supported profile and entrypoints
      VAProfileNone                   :	VAEntrypointVideoProc
      VAProfileNone                   :	VAEntrypointStats
      VAProfileMPEG2Simple            :	VAEntrypointVLD
      VAProfileMPEG2Simple            :	VAEntrypointEncSlice
      VAProfileMPEG2Main              :	VAEntrypointVLD
      VAProfileMPEG2Main              :	VAEntrypointEncSlice
      VAProfileH264Main               :	VAEntrypointVLD
      VAProfileH264Main               :	VAEntrypointEncSlice
      VAProfileH264Main               :	VAEntrypointFEI
      VAProfileH264High               :	VAEntrypointVLD
      VAProfileH264High               :	VAEntrypointEncSlice
      VAProfileH264High               :	VAEntrypointFEI
      VAProfileVC1Simple              :	VAEntrypointVLD
      VAProfileVC1Main                :	VAEntrypointVLD
      VAProfileVC1Advanced            :	VAEntrypointVLD
      VAProfileJPEGBaseline           :	VAEntrypointVLD
      VAProfileH264ConstrainedBaseline:	VAEntrypointVLD
      VAProfileH264ConstrainedBaseline:	VAEntrypointEncSlice
      VAProfileH264ConstrainedBaseline:	VAEntrypointFEI
      VAProfileVP8Version0_3          :	VAEntrypointVLD

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.

Everything still works with firefox 109.0-1.

This was reported also in bug #1767243 1, but said to be a Mesa Iris problem. No idea, why it works here with Firefox 109.0 from Debian sid/unstable, but not Nightly.

Gnome Wayland, Debian Testing, Intel Iris Graphics 6100 (BDW GT3), Mesa 22.3.3
mozregression --good 2023-01-01 --bad 20230123094444 -P stdout -a https://bug1619882.bmoattachments.org/attachment.cgi?id=9149605

6:28.58 INFO: Last good revision: e226ed35355cea5a48f3199c3af504f8498e0aca
6:28.58 INFO: First bad revision: a6db250874298a18905648a37f29f6266e926bd7
6:28.58 INFO: Pushlog:
https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=e226ed35355cea5a48f3199c3af504f8498e0aca&tochange=a6db250874298a18905648a37f29f6266e926bd7

a6db250874298a18905648a37f29f6266e926bd7 stransky — Bug 1809162 [Linux] Utilize dmabuf surface copy in VideoFramePool
1cb0dda1e2c032b81733629a27bed6067bed31ab stransky — Bug 1809162 [Linux] Implement dmabuf YUV surface copy r=emilio
1a21a0a5258323775dfb8af3c7aab9bd8bfcebfb stransky — Bug 1809162 [Linux] Fixed ClaimSnapshotGLContext() logs r=emilio
82b73378753bbbd69ec84584ecc3f7cd6ef50f78 stransky — Bug 1809162 [Linux] Implement GetAndConfigureDMABufDevice() to configure dmabuf backend without gfxPlatform on RDD process r=emilio

:stransky, since you are the author of the regressor, bug 1809162, could you take a look?

For more information, please visit auto_nag documentation.

Yes, that should be "fixed" by Bug 1811210 (so try latest nightly) but we still need to make it working on Intel then.

Tested with H264 and VP8:
bug 1809162 introduced&enforced one-copy which fails.
Yes, bug 1811210 prefers zero-copy again which works. There was room, the heuristic didn't switch over to one-copy.

Indeed it seems to work again with Nightly, 111.0a1 (20230126212606). Thank you all.

Should I mark this bug as resolved?

No, we should make it work on Intel anyway to make sure we can copy surfaces in case we need it. Will post patches soon.

Depends on D168043

:stransky this is the final week of 111 nightly, 111 goes to beta next week.
Do you plan on landing these patches in central this week?

(In reply to Donal Meehan [:dmeehan] from comment #13)

:stransky this is the final week of 111 nightly, 111 goes to beta next week.
Do you plan on landing these patches in central this week?

Yes, I do. The patches here depends on https://phabricator.services.mozilla.com/D167917, will try to solve that ASAP.

Backed out for causing bp-hybrid bustages in FFmpegVideoFramePool.cpp.

• Backout link
• Push with failures
• Failure Log
• Failure line: gmake[4]: *** [/builds/worker/checkouts/gecko/config/rules.mk:670: FFmpegVideoFramePool.o] Error 1

Updated, Thanks.

https://hg.mozilla.org/mozilla-central/rev/ad002eb6e765
https://hg.mozilla.org/mozilla-central/rev/ff9ad7a4fb13