Video decode fails with "operation failed" on Tiger Lake

[lhw]

System information

• CPU information(cat /proc/cpuinfo | grep "model name" | uniq): 11th Gen Intel(R) Core(TM) i7-1165G7
• GPU information(lspci -nn | grep -E 'VGA|isplay): Intel Corporation TigerLake-LP GT2 [Iris Xe Graphics] [8086:9a49]
• Display server if rendering to display(X or wayland): wayland

Issue behavior

Describe the current behavior

The va-api library returns with code 1 operation failed. See the vainfo log below. Usage via mpv and web browsers is also impossible. i.e. with mpv

[vaapi] libva: /usr/lib/x86_64-linux-gnu/dri/iHD_drv_video.so init failed

but runs fine with the opengl backend which then bypasses the video decoder

         ENGINES     BUSY                                                                                        MI_SEMA MI_WAIT
       Render/3D   29.24% |████████████████████████▋                                                           |      0%      0%
         Blitter    0.00% |                                                                                    |      0%      0%
           Video    0.00% |                                                                                    |      0%      0%
    VideoEnhance    0.00% |                                                                                    |      0%      0%

Describe the expected behavior

Debug information

• What's libva/libva-utils/gmmlib/media-driver version? libva 2.14.0 / libva-utils 2.12.0 / gmmlib 22.1.1 / media-driver 22.2.1 (current packages on Ubuntu 22.04)
• Could you confirm whether GPU hardware exist or not by ls /dev/dri? card0 renderD128
• Could you provide vainfo log if possible by vainfo >vainfo.log 2>&1?

  % LIBVA_DRIVERS_PATH=/usr/lib/x86_64-linux-gnu/dri LIBVA_DRIVER_NAME=iHD vainfo
  libva info: VA-API version 1.14.0
  libva info: User environment variable requested driver 'iHD'
  libva info: Trying to open /usr/lib/x86_64-linux-gnu/dri/iHD_drv_video.so
  libva info: Found init function __vaDriverInit_1_14
  libva error: /usr/lib/x86_64-linux-gnu/dri/iHD_drv_video.so init failed
  libva info: va_openDriver() returns 1
  vaInitialize failed with error code 1 (operation failed),exit
  

• Could you provide libva trace log if possible? Run cmd export LIBVA_TRACE=/tmp/libva_trace.log first then execute the case.

  [55000.599878][ctx       none]==========va_TraceInitialize
  [55000.599881][ctx       none]=========vaInitialize ret = VA_STATUS_ERROR_UNKNOWN, unknown libva error 
  [55000.599884][ctx       none]==========va_TraceTerminate
  [55000.599901][ctx       none]=========vaTerminate ret = VA_STATUS_SUCCESS, success (no error) 
  

• Could you attach dmesg log if it's GPU hang by dmesg >dmesg.log 2>&1? /
• Do you want to contribute a patch to fix the issue? (yes/no): no

[rathann]

Same here on Fedora 36 pre-release with:

intel-gmmlib-22.1.0-1.fc36
intel-media-driver-22.1.1-2.fc36.x86_64
libva-2.14.0-1.fc36.x86_64

Downgrading to 21.x series makes it work again, so this looks like a regression.

Similar hardware (same GPU):

$ lspci -nn |grep VGA
00:02.0 VGA compatible controller [0300]: Intel Corporation TigerLake-LP GT2 [Iris Xe Graphics] [8086:9a49] (rev 01)

My libva trace is identical to what was posted above.

Here's my dmesg: kernel-5.17.0-dmesg.txt.

[rathann]

It looks like a simple rebuild of intel-media-driver package fixed things for me. VA-API is working again with:

intel-gmmlib-22.1.0-1.fc36.x86_64
intel-media-driver-22.1.1-3.fc36.x86_64

Downstream bug report: https://bugzilla.rpmfusion.org/show_bug.cgi?id=6248 .

[lhw]

I can report the same for Ubuntu 22.04. A simple recompile of both packages solves the problem. So my guess would be that this is not an upstream issue per se. Maybe just indirect due to changes with the ABI. Feel free to close the issue if you consider the problem solved.

[seb128]

confirmed that rebuilding intel-media-driver resolve the issue on Ubuntu, I uploaded a package rebuild. Is libva supposed to maintain ABI compability or what is happening there?

[rathann]

Unfortunately, the issue is back after updating intel-gmmlib from 22.1.0 to 22.1.2 without rebuilding intel-media-driver. Rebuild helps, so there's definitely some ABI incompatibility between minor intel-gmmlib versions.

[seb128]

the new update fixed the ABI to be compatible with older versions but it means if the media driver was built with the buggy .0 then it needs another rebuild now

[rathann]

Oh, so 22.1.0 was buggy? Too bad, but thanks for letting us know. Which versions are ABI-compatible, then? 22.1.2 isn't compatible with 22.0.3, either.

[seb128]

see intel/gmmlib#95 , it states that 22.0.3 to 22.1.1 are incompatible with the old ABI

[rathann]

Thanks for the link. It explains the issue well. I can confirm that media-driver built with either gmmlib 22.0.2 or 22.1.2 works with either, too.

[lhw]

I think we can go ahead and close the issue then