On 22/01/2024 11:11, Chen Wang wrote: >>>>>> + $ref: /schemas/clock/sophgo,sg2042-sysclk.yaml# >>>>>> + type: object >>>>> Why isn't this merged here? You do not need the child node really... >>>>> unless the clock inputs are specific to that clock controller and you >>>>> will have here more devices? But where are they in such case? >>>> I don't see more devices will be included later. It should be ok to >>>> merge them into one. >>> hi, Krzysztof, >>> >>> After some double check, I find we will have more devices in >>> system-control. For example, in the SYS_CTRL area, there is also a >>> section of registers used to control the "General Purpose Interrupt". >>> The pcie controller of sg2042 will use this interrupt controller which >>> is defined in SYS_CTRL, we will add it in later work. >>> >> I expect then all devices to be documented. > > hi, Krzysztof. > > First, I'm very sorry for having double-checked with you for this system > controller and child node issue, but this time I'm sure there should be > no more child nodes except the clock and interrupt controllers, though > there are some other registers in SYS_CTRL section, but we will not use > them till now. > > One question, when you say "to be documented", do you mean I need write > binding/yaml files for other child node? But they exceed the scope of > this patchset (this patchset is for clock support only). That's why I That's not true. The scope of this patch is to add DT binding description for your device. If you choose any other scope, I don't agree and I am not going to provide positive review. > suggest just add clock-controller in this patchset and to add the > interrupt controller in another patchset for pcie support. This > mechanism should be suitable for our expansion. How then are you going to solve the requirement: "DO attempt to make bindings complete even"? https://elixir.bootlin.com/linux/v6.1-rc1/source/Documentation/devicetree/bindings/writing-bindings.rst#L17 Best regards, Krzysztof
