The PSR Bat test fails (currently only hsw-xps12 has a psr panel). The testcase itself seems happy, but the kernel complains with the above message.
Is this a serious issue, or should we simply tune it down to DRM_DEBUG_KMS?
Correction, it seems to work well on bsw-nuc-2. Long-term history of this testcase on the CI server can be found at
(In reply to Daniel Vetter from comment #2)
> Correction, it seems to work well on bsw-nuc-2
A NUC wouldn't have eDP or PSR.
Oh, this is odd since the first check on this path is for the dpcd bit that tells sink crc is supported.
But anyway I'm going to tune down it to DRM_DEBUG_KMS and probably remove the "_basic" from sink crc since it isn't a basic feature or functionality, but just used for testing another features like psr and fbc. what do you think?
(In reply to Rodrigo Vivi from comment #4)
> Oh, this is odd since the first check on this path is for the dpcd bit that
> tells sink crc is supported.
> But anyway I'm going to tune down it to DRM_DEBUG_KMS and probably remove
> the "_basic" from sink crc since it isn't a basic feature or functionality,
> but just used for testing another features like psr and fbc. what do you
It's basic validation feature, if it doesn't work we can't validate the feature (PSR) itself. That's why imo it must be in BAT. Same reason we have a basic testcase for pipe CRC sanity in BAT, just to make sure that feature tests can give correct results.
wrt tuning down another option is that the kernel should maybe give the testcase more information (like return -ENOTTY to signal sink crc isn't working/present, so that the testcase can skip correctly.
Just to clarify: Currently the testcase returns SUCCESS, and since there's clearly no edp on a nuc that would be a bug in either the kernel or userspace in correctly skipping the testcase.
So, this machine do have eDP. Both hsw-xps12 and bdw-ultra have PSR capable eDP panels as we can see with archive/results/CI_IGT_test/igt@kms_psr_sink_crc@psr_basic.html
that is btw all green now.
Also, the test is returning success because the sink crc is actually being calculated... just times out when trying to disable back the sink crc calculation.
So, better to keep it as a debug_kms as Daniel suggested: https://patchwork.freedesktop.org/patch/70436/
Daniels argument on keeping it as basic made total sense to me so I now agree, so let's keep it there.
Author: Rodrigo Vivi <email@example.com>
Date: Fri Jan 29 14:44:59 2016 -0800
drm/i915: Sink CRC: tune down error message at stop to debug_kms.