Re: [PATCH v12 2/2] drm/amdgpu: Enable async flip on overlay planes

["André Almeida" <andrealmeid@igalia.com>]

Em 29/01/2025 11:36, Xaver Hugl escreveu:
> Am Mo., 27. Jan. 2025 um 21:00 Uhr schrieb André Almeida
> <andrealmeid@igalia.com>:
>>
>> amdgpu can handle async flips on overlay planes, so allow it for atomic
>> async checks.
>>
>> Signed-off-by: André Almeida <andrealmeid@igalia.com>
>> ---
>>   drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_plane.c | 10 ++++++----
>>   1 file changed, 6 insertions(+), 4 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_plane.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_plane.c
>> index 774cc3f4f3fd9a964fe48c66eb596d2f6dfee602..6bfed3d1530e6610eea025b477f409ee505870da 100644
>> --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_plane.c
>> +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_plane.c
>> @@ -1258,21 +1258,23 @@ static int amdgpu_dm_plane_atomic_check(struct drm_plane *plane,
>>   }
>>
>>   static int amdgpu_dm_plane_atomic_async_check(struct drm_plane *plane,
>> -                                             struct drm_atomic_state *state)
>> +                                             struct drm_atomic_state *state, bool flip)
>>   {
>>          struct drm_crtc_state *new_crtc_state;
>>          struct drm_plane_state *new_plane_state;
>>          struct dm_crtc_state *dm_new_crtc_state;
>>
>> -       /* Only support async updates on cursor planes. */
>> -       if (plane->type != DRM_PLANE_TYPE_CURSOR)
>> +       if (flip) {
>> +               if (plane->type != DRM_PLANE_TYPE_OVERLAY)
>> +                       return -EINVAL;
>> +       } else if (plane->type != DRM_PLANE_TYPE_CURSOR)
>>                  return -EINVAL;
> 
> This changes the logic for cursor updates, flipping on the cursor
> plane allowed async updates before. Is that intentional?
> 

It's not the intention of this patch to disable async updates on cursor 
planes... but I don't think it's happening here? Async plane updates and 
async page flips are different things.

Any function that used to call amdgpu_dm_plane_atomic_async_check() for 
an async update on a cursor plane will continue to being able to do that.

For callers of _atomic_async_check() from a page flip path (like 
drm_atomic_set_property()), those couldn't flip a cursor plane, and will 
continue to be like that for now.

At least this is my analysis, please let me know if I got something wrong.