Re: [PATCH v3 3/3] ARM: dts: stm32: fix several DT warnings on stm32mp15

[Marek Vasut <marex@denx.de>]

On 5/25/23 10:14, Raphael Gallais-Pou wrote:

Hi,

>> I think if you retain the stm32mp151.dtsi &ltdc { port { #address-cells = <1>;
>> #size-cells = <0>; }; }; part, then you wouldn't be getting any warnings
>> regarding LTDC , and you wouldn't have to remove the unit-address from
>> endpoint@0 .
>>
>> btw. I do use both endpoint@0/endpoint@1 in Avenger96 DTOs, but those are not
>> submitted yet, I have to clean them up a bit more first.
>>
>>> One way to do it would be to make the endpoint@0 go down in the device-tree with
>>> its dependencies, so that both endpoints are the same level without generating
>>> noise.
>>
>> I'm afraid I really don't quite understand which warning you're referring to.
>> Can you please share that warning and ideally how to trigger it (the
>> command-line incantation) ?
> 
> Using '$ make dtbs W=1', you can observe several of the followings:
> 
> arch/arm/boot/dts/stm32mp151.dtsi:1533.9-1536.6: Warning
> (avoid_unnecessary_addr_size): /soc/display-controller@5a001000/port:
> unnecessary #address-cells/#size-cells without "ranges" or child "reg" property
> arch/arm/boot/dts/stm32mp151.dtsi:1533.9-1536.6: Warning (graph_child_address):
> /soc/display-controller@5a001000/port: graph node has single child node
> 'endpoint@0', #address-cells/#size-cells are not necessary
> 
> This &ltdc { port { #address-cells = <1>; #size-cells = <0>; }; }; part is
> actually annoying. This is because there is several device-trees that only got
> one endpoint, and some other that includes two.
> 
> For instance: stm32mp15xx-dhcor-avenger96.dtsi vs stm32mp157c-dk2.dts.
> 
> I would like to remove to root part of address/size field and let only the lower
> device-trees with with multiple endpoints handle their own fields. I hope this
> explains a bit better my process.

After thinking about this some more, and digging through LTDC driver, 
and testing on EV1, I think dropping the LTDC node endpoint@N and 
reg=<N> altogether and just using port/endpoint (singular) is fine.

You might want to split the DSI node specific changes and the LTDC node 
specific changes into separate patches (LTDC specific change like you 
did in 1/3).