Bug 309438 – [GStreamer] Flaky media test failures due to video frame map failures in avviddec with buffers allocated by the GL buffer pool

• Status: NEW
• Component: Media

[Claudio Saavedra]

https://results.webkit.org/?suite=layout-tests&test=media%2Fremote-control-command-scrubbing.html&platform=GTK&platform=WPE

https://build.webkit.org/results/GTK-Linux-64-bit-Release-Tests/308879@main%20(23819)/results.html

0:00:08.781789309 650221 0x7f0900001120 ERROR              videometa gstvideometa.c:424:default_map: plane 1, no memory at offset 99840
0:00:08.781802564 650221 0x7f0900001120 ERROR                default video-frame.c:168:gst_video_frame_map_id: failed to map video frame plane 1
0:00:08.785723698 650221 0x557320cd5760 ERROR           webkitcommon GStreamerCommon.cpp:1087:operator():<regular-media-player-8> Got message: error message: 0x7f06a0027ba0, time 99:99:99.999999999, seq-num 6880, element 'avdec_h264-6', GstMessageError, gerror=(GError)NULL, debug=(string)"../../../../../jhbuild/checkout/gstreamer/subprojects/gst-libav/ext/libav/gstavviddec.c\(1888\):\ get_output_buffer\ \(\):\ /GstPlayBin:regular-media-player-8/GstURIDecodeBin:uridecodebin8/GstDecodeBin:decodebin8/avdec_h264:avdec_h264-6:\012The\ video\ memory\ allocated\ from\ downstream\ pool\ could\ not\ mapped\ for\ read\ and\ write.";
0:00:08.785742483 650221 0x557320cd5760 ERROR      webkitmediaplayer MediaPlayerPrivateGStreamer.cpp:2132:handleMessage:<regular-media-player-8> Cannot access memory for read and write operation. (url=file:///sdk/webkit/LayoutTests/media/content/test.mp4) (code=7)

[Philippe Normand]

*** Bug 224105 has been marked as a duplicate of this bug. ***

[Philippe Normand]

*** Bug 309436 has been marked as a duplicate of this bug. ***

[Philippe Normand]

*** Bug 308265 has been marked as a duplicate of this bug. ***

[Philippe Normand]

I wonder if this is related with bug 310895 ...

[Philippe Normand]

*** Bug 311093 has been marked as a duplicate of this bug. ***

[Philippe Normand]

results link: https://results.webkit.org/?suite=layout-tests&suite=layout-tests&suite=layout-tests&suite=layout-tests&suite=layout-tests&test=media%2Fvideo-src-blob-replay.html&test=media%2Fmedia-source%2Fmedia-source-reopen.html&test=media%2Fvideo-load-preload-metadata.html&test=compositing%2Fvideo%2Fvideo-clip-change-src.html&test=media%2Fremote-control-command-scrubbing.html&platform=GTK&platform=WPE

[Claudio Saavedra]

*** Bug 311320 has been marked as a duplicate of this bug. ***

[Diego Pino]

Several tests failing with this error were filed under bug webkit.org/b/308438.

These other tests are also flaky timeout (>0.5%), failing with a similar stderr:

```
media/media-controller-timeupdate.html
media/media-source/media-source-webm-configuration-framerate.html
media/modern-media-controls/media-controller/media-controller-click-on-video-controls-should-not-pause.html
media/playlist-inherits-user-gesture.html
media/remote-control-command-seek.html
media/remove-from-document-with-text-track.html
media/track/track-cue-missing.html
media/track/track-long-captions-file.html
media/track/webvtt-inline.html
media/video-currentTime-duration.html
media/video-isplayingtoautomotiveheadunit.html
media/video-main-content-allow-then-scroll.html
media/video-main-content-deny-display-none.html
media/video-main-content-deny-too-small.html
media/video-paused-0-rate.html
media/video-pause-empty-events.html
media/video-replaces-poster.html
media/video-seek-by-small-increment.html
media/video-seeking.html
media/video-unmuted-after-play-holds-sleep-assertion.html
```

For instance, test in the case of 'media/media-controller-timeupdate.html':

https://results.webkit.org/?suite=layout-tests&test=media%2Fmedia-controller-timeupdate.html&platform=GTK&platform=WPE

stderr:

```
0:00:08.642527865 607321 0xffff480012f0 ERROR              videometa gstvideometa.c:424:default_map: plane 1, no memory at offset 83200
0:00:08.642545225 607321 0xffff480012f0 ERROR                default video-frame.c:168:gst_video_frame_map_id: failed to map video frame plane 1
0:00:08.643018868 607321 0xaaaaf3a2d3b0 ERROR           webkitcommon GStreamerCommon.cpp:1086:operator():<regular-video-16> Got message: error message: 0xfffc740391b0, time 99:99:99.999999999, seq-num 13323, element 'avdec_h264-12', GstMessageError, gerror=(GError)NULL, debug=(string)"../../../../../jhbuild/checkout/gstreamer/subprojects/gst-libav/ext/libav/gstavviddec.c\(1888\):\ get_output_buffer\ \(\):\ /GstPlayBin:regular-video-16/GstURIDecodeBin:uridecodebin16/GstDecodeBin:decodebin16/avdec_h264:avdec_h264-12:\012The\ video\ memory\ allocated\ from\ downstream\ pool\ could\ not\ mapped\ for\ read\ and\ write.";
0:00:08.643040109 607321 0xaaaaf3a2d3b0 ERROR      webkitmediaplayer MediaPlayerPrivateGStreamer.cpp:2139:handleMessage:<regular-video-16> Cannot access memory for read and write operation. (url=file:///sdk/webkit/LayoutTests/media/content/test.mp4) (code=7)
```

[Philippe Normand]

How much RAM do we have on these bots?

Is the post-commit bot fleet using a uniform set of machines? (same amount of CPU vCores and RAM)?

[Philippe Normand]

I was able to reproduce this in wksdk only, I'm not sure what's going on, the decoder sometimes fails to map buffers allocated by the GL pool... Maybe some issue in Mesa...

Anyway, this won't be an issue anymore when we update to GStreamer 1.28 where glupload uses another allocator (udmabuf).

[Philippe Normand]

*** Bug 307294 has been marked as a duplicate of this bug. ***

[Philippe Normand]

*** Bug 306471 has been marked as a duplicate of this bug. ***

[Philippe Normand]

media/video-buffered.html [ Pass Failure Timeout ] should become media/video-buffered.html [ Pass Failure ] after fixing this...

[Philippe Normand]

*** Bug 311863 has been marked as a duplicate of this bug. ***

[Philippe Normand]

*** Bug 308146 has been marked as a duplicate of this bug. ***

[Philippe Normand]

*** Bug 307666 has been marked as a duplicate of this bug. ***

[Philippe Normand]

webkit.org/b/122508 compositing/video/video-reflection.html [ Pass Timeout ] should be removed after fixing this.

[Philippe Normand]

*** Bug 310278 has been marked as a duplicate of this bug. ***

[Philippe Normand]

*** Bug 310274 has been marked as a duplicate of this bug. ***

[Philippe Normand]

webkit.org/b/169918 compositing/video/video-object-position.html [ Failure Timeout ] should become webkit.org/b/169918 compositing/video/video-object-position.html [ Failure ] after fixing this.

[Philippe Normand]

webkit.org/b/264680 media/video-page-visibility-restriction.html [ Pass Timeout ] should be removed after fixing this.

[Philippe Normand]

webkit.org/b/262949 media/media-garbage-collection.html [ Pass Timeout Failure ] should become webkit.org/b/262949 media/media-garbage-collection.html [ Pass Failure ] after fixing this.

[Philippe Normand]

webkit.org/b/261024 media/video-playsinline.html [ Failure Timeout Pass ] also affected by this issue.

[Philippe Normand]

Before opening new bugs about flaky timeout media tests please check if there is a -stderr.txt file for the test in the results folder, containing the error mentioned above...

[Diego Pino]

*** Bug 311860 has been marked as a duplicate of this bug. ***

[Diego Pino]

*** Bug 311862 has been marked as a duplicate of this bug. ***