Bug 104096

(In reply to Marta Löfstedt from comment #0)
> https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_3452/shard-glkb3/
> igt@kms_ccs@pipe-c-crc-primary-rotation-180.html
> 
> (kms_ccs:3892) CRITICAL: Test assertion failure function try_config, file
> kms_ccs.c:439:
> (kms_ccs:3892) CRITICAL: Failed assertion: ret == 0
> (kms_ccs:3892) CRITICAL: Last errno: 22, Invalid argument
> (kms_ccs:3892) CRITICAL: error: -22 != 0
> Subtest pipe-C-crc-primary-rotation-180 failed.

The log shows:

<7>[   80.778305] [drm:skl_check_plane_surface [i915]] No RC support on pipe C

Which is caused by the verification for pipe C in intel_display.c:skl_check_plane_surface(), which, despite the name, only executes for GEN > 10 or Geminilake.  The important hunk is:

>	if (crtc->pipe == PIPE_C) {
>		DRM_DEBUG_KMS("No RC support on pipe C\n");
>		return -EINVAL;
>	}

What I cannot tell is whether GLK should support RC for pipe C or if the test is bogus.  I don't have the documentation for GLK yet :)

So, I emailed Ben Widawsky at Intel to confirm whether the testcase is bogus or if we need to support pipe_c ccs modifier in the kernel.  I'm waiting on him.

Patch commited:

f0cbd8bd877f ("i915: Reject CCS modifiers for pipe C on Geminilake")

Which solved the issue in CI:

https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_3561/shard-glkb1/igt@kms_ccs@pipe-c-crc-primary-rotation-180.html