sometimes takes ~5 seconds to _dma_fmt_to_dma_drm_fmts: assertion 'fmt != GST_VIDEO_FORMAT_UNKNOWN' failed

[Joe Barnett]

since upgrading to noble, sometimes the following bit of code works instantaneously, and sometimes it takes ~5 seconds to report:

** (gst-plugin-scanner:692358): CRITICAL **: 10:55:31.508: _dma_fmt_to_dma_drm_fmts: assertion 'fmt != GST_VIDEO_FORMAT_UNKNOWN' failed

code:
>>> import cv2
>>> vc = cv2.VideoCapture('/dev/video2')

when it takes a long time and reports the assertion error, I see the following in journalctl logs (so maybe an amd/kernel bug instead of an opencv bug? -- also fwiw am on a hybrid amd/intel system where DRI_PRIME=1 generally activates the amd card but this should be running w/out DRI_PRIME set):

Jul 01 10:55:29 taplop kernel: [drm] scheduler comp_1.2.1 is not ready, skipping
Jul 01 10:55:29 taplop kernel: [drm] PCIE GART of 256M enabled (table at 0x000000F400000000).
Jul 01 10:55:30 taplop kernel: amdgpu 0000:01:00.0: [drm:amdgpu_ring_test_helper [amdgpu]] *ERROR* ring comp_1.1.0 test failed (-110)
Jul 01 10:55:30 taplop kernel: amdgpu 0000:01:00.0: [drm:amdgpu_ring_test_helper [amdgpu]] *ERROR* ring comp_1.2.0 test failed (-110)
Jul 01 10:55:30 taplop kernel: amdgpu 0000:01:00.0: [drm:amdgpu_ring_test_helper [amdgpu]] *ERROR* ring comp_1.3.0 test failed (-110)
Jul 01 10:55:30 taplop kernel: amdgpu 0000:01:00.0: [drm:amdgpu_ring_test_helper [amdgpu]] *ERROR* ring comp_1.0.1 test failed (-110)
Jul 01 10:55:30 taplop kernel: amdgpu 0000:01:00.0: [drm:amdgpu_ring_test_helper [amdgpu]] *ERROR* ring comp_1.1.1 test failed (-110)
Jul 01 10:55:31 taplop kernel: amdgpu 0000:01:00.0: [drm:amdgpu_ring_test_helper [amdgpu]] *ERROR* ring comp_1.2.1 test failed (-110)
Jul 01 10:55:31 taplop kernel: [drm] UVD and UVD ENC initialized successfully.
Jul 01 10:55:31 taplop kernel: [drm] VCE initialized successfully.
Jul 01 10:55:31 taplop kernel: [drm] scheduler comp_1.1.0 is not ready, skipping
Jul 01 10:55:31 taplop kernel: [drm] scheduler comp_1.2.0 is not ready, skipping
Jul 01 10:55:31 taplop kernel: [drm] scheduler comp_1.3.0 is not ready, skipping
Jul 01 10:55:31 taplop kernel: [drm] scheduler comp_1.0.1 is not ready, skipping
<<< not ready skipping lines repeat a whole bunch of times >>>

ProblemType: Bug
DistroRelease: Ubuntu 24.04
Package: python3-opencv 4.6.0+dfsg-13.1ubuntu1
ProcVersionSignature: Ubuntu 6.8.0-35.35-generic 6.8.4
Uname: Linux 6.8.0-35-generic x86_64
ApportVersion: 2.28.1-0ubuntu3
Architecture: amd64
CasperMD5CheckResult: unknown
CurrentDesktop: GNOME
Date: Mon Jul 1 10:59:12 2024
InstallationDate: Installed on 2019-08-17 (1781 days ago)
InstallationMedia: Ubuntu 19.04 "Disco Dingo" - Alpha amd64 (20190305.1)
RebootRequiredPkgs: Error: path contained symlinks.
SourcePackage: opencv
UpgradeStatus: Upgraded to noble on 2024-03-23 (100 days ago)

[Joe Barnett (thejoe)]

can reproduce this without opencv using:

gst-launch-1.0 v4l2src device="/dev/video2" ! videoconvert ! ximagesink

[Joe Barnett (thejoe)]

it does appear that `mplayer tv:// -tv driver=v4l2:device=/dev/video2` doesn't have this problem, while the gst-launch command above does, so adding gst plugins package that includes the v4l2 src