intel-technology-enabling-for-openshift icon indicating copy to clipboard operation
intel-technology-enabling-for-openshift copied to clipboard

VAAPI can't use other rate control profiles of GPU

Open brgavino opened this issue 1 year ago • 7 comments

When deploying in Openshift, only one rate control profile Is available:

image

We have installed the gpu driver from 1.1.0 (with firmware)

version: intel-data-center-gpu-driver-container:2.0.0-5.14.0-284.28.1.el9_2.x86_64 and can access the GPU with OpenCL and other utilities with no issue.

It seems that one possible cause of the limitation may listed here: https://github.com/intel/media-driver?tab=readme-ov-file#known-issues-and-limitations

The GPU configuration sets the mode of enable_guc to 3 (see attached doc syskerneldebugdri1i915_.txt), but according to https://wiki.archlinux.org/title/intel_graphics#Enable_GuC_/_HuC_firmware_loading we should see the message i915 0000:00:02.0: [drm] HuC firmware i915/icl_huc_9.0.0.bin version 9.0 authenticated:yes

but instead,

image

It may be possible that HuC fw is not loaded correctly, and the rate control isn't available

brgavino avatar Jan 08 '24 17:01 brgavino

Test Dockerfile to see VAAPI through vainfo command: vainfo.zip

brgavino avatar Jan 08 '24 17:01 brgavino

Note: This happens on Openshift 4.12.7, as well as OpenShift 4.13.11

brgavino avatar Jan 08 '24 18:01 brgavino

Thanks for submitting the issue @brgavino. We have not support media driver so far. Without the media driver, the driver container you have used works with no issues right? @uMartinXu any thoughts here?

vbedida79 avatar Jan 09 '24 17:01 vbedida79

Thanks, @brgavino @vbedida79, The Flex GPU media features are in plan but have not been supported yet. We can create a feature request github issue and address this feature request. And partners like @brgavino can give more clear requirments there.

uMartinXu avatar Jan 11 '24 19:01 uMartinXu

That's fine, let's create the feature request - it would be helpful to add a note to documentation that media driver support isn't available in OpenShift yet. The dependency (GPU Plugin) does support it on other platforms, otherwise - so it is unclear.

brgavino avatar Jan 11 '24 19:01 brgavino

looks like there is some issue with "mei_gsc" kernel module in the backport driver. [ 155.993631] mei_gsc i915.mei-gscfi.19712: FW not ready: resetting: dev_state = 2 pxp = 0

Any idea to fix it quickly?

djie1 avatar Jan 12 '24 06:01 djie1

@djie1 please refer to this https://github.com/intel/intel-data-center-gpu-driver-for-openshift/issues/68

hershpa avatar Jan 16 '24 17:01 hershpa