[PATCH v2 2/2] dt-bindings: display: simple-framebuffer: Document physical width and height properties

[Krzysztof Kozlowski]

On 24/01/2023 23:19, Rayyan Ansari wrote:
> On 23/01/2023 17:53, Rob Herring wrote:
>> On Sun, Jan 22, 2023 at 05:25:38PM +0000, Rayyan Ansari wrote:
>>> On 22/01/2023 15:36, Rob Herring wrote:
>>>> On Sat, Jan 21, 2023 at 9:36 AM Rayyan Ansari <rayyan at ansari.sh> wrote:
>>>>>
>>>>
>>>> Why do you need this change?
>>>>
>>>> The 'simple-framebuffer' contains data on how the bootloader
>>>> configured the display. The bootloader doesn't configure the display
>>>> size, so this information doesn't belong here. The information should
>>>> already be in the panel node, so also no point in duplicating it here.
>>>>
>>>>> Signed-off-by: Rayyan Ansari <rayyan at ansari.sh>
>>>>> ---
>>>>>    .../devicetree/bindings/display/simple-framebuffer.yaml   | 8 ++++++++
>>>>>    1 file changed, 8 insertions(+)
>>>
>>> Hi Rob,
>>>
>>> There is the usecase that Hans has mentioned, but I have also mentioned
>>> another usecase previously.
>>>
>>> Adding the width-mm and height-mm properties allows user interfaces such as
>>> Phosh (https://puri.sm/posts/phosh-overview/) to scale correctly to the
>>> screen. In my case, a panel node is not available and the aforementioned
>>> interface is in fact running on the SimpleDRM driver (which binds to the
>>> simple-framebuffer device).
>>
>> Why is the panel node not available? Why not add it? Presumably it is
>> not there because you aren't (yet) using the simple-panel driver (and
>> others that would need). But presumably you will eventually as I'd
>> imagine turning the screen off and back on might be a desired feature.
> 
> It requires more than using the simple-panel driver: first the SoC side 
> display hardware needs to be brought up, then a panel driver that 
> implements the proper DCS initialisation sequence needs to be written 
> (which is currently not fully known).
> 
>>
>> So why add a temporary DT property that's tied to your *current* kernel? > The DT should not be tightly coupled to the kernel.
> 
> I'm not sure what you mean by it being "tightly coupled" to the kernel.

It means that you used current Linux driver support (or lack) for some
hardware as an argument for bindings. If you add later the driver, the
bindings should be changed? Answer is: not. Bindings should be
independent of Linux drivers, thus whatever kernel is missing now, is
not an argument in favor of this property.

Best regards,
Krzysztof