[PATCH v3 RESEND 21/24] drm/exynos/decon5433: add local path support

[Krzysztof Kozlowski]

On 03/03/2022 17:11, Marek Szyprowski wrote:
> Hi Krzysztof,
> 
> On 03.03.2022 17:03, Krzysztof Kozlowski wrote:
>> On 02/03/2022 02:00, Inki Dae wrote:
>>> 22. 2. 7. 01:51에 Krzysztof Kozlowski 이(가) 쓴 글:
>>>> On 25/03/2019 08:13, Andrzej Hajda wrote:
>>>>> GSCALERs in Exynos5433 have local path to DECON and DECON_TV.
>>>>> They can be used as extra planes with support for non-RGB formats and scaling.
>>>>> To enable it on DECON update_plane and disable_plane callback should
>>>>> be modified. Moreover DSD mux should be set accordingly, and finally
>>>>> atomic_check callback should be used to limit the number of active planes.
>>>>>
>>>>> Signed-off-by: Andrzej Hajda <a.hajda at samsung.com>
>>>>> ---
>>>>>   drivers/gpu/drm/exynos/exynos5433_drm_decon.c | 80 +++++++++++++++----
>>>>>   drivers/gpu/drm/exynos/regs-decon5433.h       |  6 ++
>>>>>   2 files changed, 72 insertions(+), 14 deletions(-)
>>>>>
>>>>> diff --git a/drivers/gpu/drm/ex
>>>> Hi guys!
>>>>
>>>> I am working on DRM bindings conversion to DT schema format and I found
>>>> this set only partially applied. I merged the DTS patches ("dsd" clock),
>>>> but I think the driver and bindings were not picked up.
>>>>
>>>> Nevertheless I am going to include the "dsd" clock in the new bindings,
>>>> so the DTS passes DT schema checks. Let me know if other approach
>>>> (revert of DTS change) should be taken.
>>>>
>>> Sorry for late response.
>>>
>>> As of now, "dsd" is a dead property not used anywhere.
>>> This patch series makes real user not to work correctly due to ABI change.
>>> How about reverting it until this patch series is merged after fixing the real user problem?
>> The Exynos5433 DECON bindings were already merged by Rob, so someone
>> would need to send a revert. However this does not answer the actual
>> question - whether the "dsd" clock is necessary, whether it is there
>> (routed to DECON). If it is, it should stay in the bindings.
> 
> It is routed to DECON hardware and enabling it is needed to make the so 
> called 'local path' (when DECON takes image data directly from the 
> GSCALER hardware block instead of the memory buffer) working.

Awesome, thanks for confirmation! Bindings are good then.

Best regards,
Krzysztof