Re: [PATCH i-g-t v4 1/4] tests/kms_writeback: clear writeback properties in teardown path

[Abhinav Kumar <quic_abhinavk@quicinc.com>]

On 10/14/2024 6:52 AM, Kamil Konieczny wrote:
> Hi Abhinav,
> On 2024-09-25 at 13:46:32 -0700, Abhinav Kumar wrote:
>>
>>
>> On 9/25/2024 1:37 PM, Jessica Zhang wrote:
>>> Drop the WRITEBACK_FB_ID property check and refactor detach_crtc() so
>>> that it properly clears all writeback-related properties before calling
>>> a NULL commit.
>>>
>>> The current WRITEBACK_FB_ID check in detach_crtc() will always return 0
>>> due to the return value being hardcoded to 0 in the framework [1].
>>> Because of this, detach_crtc() will never call the NULL commit to clean
>>> up the atomic state.
>>>
>>> Since the NULL commit is failing for cases where the WRITEBACK_FB_ID and
>>> OUT_FENCE_PTR haven't been cleared by a non-NULL commit, let's
>>> explicitly clear these properties before calling the NULL commit.
>>>
>>> [1] https://elixir.bootlin.com/linux/v6.10.6/source/drivers/gpu/drm/drm_atomic_uapi.c#L853
>>>
>>> Fixes: 7c3ceb08b ("tests/kms_writeback: Add helper method to detach pipe from output when test finishes")
>>> Signed-off-by: Jessica Zhang <quic_jesszhan@quicinc.com>
>>> ---
>>>    tests/kms_writeback.c | 14 +++++---------
>>>    1 file changed, 5 insertions(+), 9 deletions(-)
>>>
>>
>>
>> This fixes a long pending issue today where kms_writeback does not clear the
>> crtc it was using. Hence,
>>
>> Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
> 
> I will merge this now as in our CI this test skips.
> 
> As for rest patches in series please re-work and resend and make sure
> CI will be green (as 2/4 and 3/4 touch lib/kms).
> 
> Regards,
> Kamil
> 

Thanks Kamil, noted.

Abhinav