Re: [Freedreno] [PATCH] Revert "drm/msm/dp: Remove INIT_SETUP delay"

[Abhinav Kumar <quic_abhinavk@quicinc.com>]

Hi Leonard

On 5/22/2023 7:39 PM, Leonard Lausen wrote:
> Abhinav Kumar <quic_abhinavk@quicinc.com> writes:
>>>>> There is no need to add the 100ms delay back yet.
>>>>>
>>>>> thanks for posting this but NAK on this patch till we post the fix this
>>>>> week.
>>>>>
>>>>> Appreciate a bit of patience till then.
>>>>
>>>> This regression is already part of the 6.3 stable release series. Will
>>>> the new patch qualify for inclusion in 6.3.y? Or will it be part of 6.4
>>>> and this revert should go into 6.3.y?
>>>
>>> This is a tough situation, as landing a revert will break x13s, as noted
>>> by Bjorn. Given that the workaround is known at this moment, I would
>>> like to wait for the patch from Abhinav to appear, then we can decide
>>> which of the fixes should go to the stable kernel.
> 
> I wasn't able to find new patches, though may have missed them. Is there
> a decision yet how to proceed with this regression? 6.2 now being EOL
> may make this a good moment to decide on the next steps.
> 

Yes, the new patch to fix this issue is here

https://patchwork.freedesktop.org/patch/538601/?series=118148&rev=3

Apologies if you were not CCed on this, if a next version is CCed, will 
ask kuogee to cc you.

Meanwhile, will be great if you can verify if it works for you and 
provide Tested-by tags.

>>>> [  275.025497] [drm:dpu_encoder_phys_vid_wait_for_commit_done:488]
>>>> [dpu error]vblank timeout
>>>> [  275.025514] [drm:dpu_kms_wait_for_commit_done:510] [dpu error]wait
>>>> for commit done returned -110
>>>> [  275.064141] [drm:dpu_encoder_frame_done_timeout:2382] [dpu
>>>> error]enc33 frame done timeout
>>
>> This is a different crash but the root-cause of both the issues is the
>> bridge hpd_enable/disable series.
>>
>> https://patchwork.freedesktop.org/patch/514414/
>>
>> This is breaking the sequence and logic of internal hpd as per my
>> discussion with kuogee.
>>
>> We are analyzing the issue and the fix internally first and once we
>> figure out all the details will post it.
> 
> Thank you!