Re: [PATCH v5 5/9] drm/msm/hdmi: turn mode_set into atomic_enable

[Abhinav Kumar <quic_abhinavk@quicinc.com>]

On 6/20/2024 1:32 PM, Dmitry Baryshkov wrote:
> On Thu, Jun 20, 2024 at 01:27:15PM GMT, Abhinav Kumar wrote:
>>
>>
>> On 6/7/2024 6:23 AM, Dmitry Baryshkov wrote:
>>> The mode_set callback is deprecated, it doesn't get the
>>> drm_bridge_state, just mode-related argumetns. Turn it into the
>>> atomic_enable callback as suggested by the documentation.
>>>
>>
>> mode_set is deprecated but atomic_mode_set is not.
> 
> There is no atomic_mode_set() in drm_bridge_funcs. Also:
> 

Please excuse me. I thought since encoder has atomic_mode_set(), bridge 
has one too.

> * This is deprecated, do not use!
> * New drivers shall set their mode in the
> * &drm_bridge_funcs.atomic_enable operation.
>

Yes I saw this note but it also says "new drivers" and not really 
enforcing migrating existing ones which are using modeset to atomic_enable.

My concern is that today the timing engine setup happens in encoder's 
enable() and the hdmi's timing is programmed in mode_set().

Ideally, we should program hdmi's timing registers first before the 
encoder's timing.

Although timing engine is not enabled yet, till post_kickoff, this is 
changing the sequence.

If this really required for rest of this series?

>>
>> I would rather use atomic_mode_set because moving to atomic_enable() would
>> be incorrect.
>>
>> That would be called after encoder's enable and hence changes the sequence.
>> That was not the intention of this patch.
>>
>> NAK.
>>
>>> Acked-by: Maxime Ripard <mripard@kernel.org>
>>> Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
>>> ---
>>>    drivers/gpu/drm/msm/hdmi/hdmi_bridge.c | 33 ++++++++++++++++++++++++++-------
>>>    1 file changed, 26 insertions(+), 7 deletions(-)
> 
>