Quoting Andrea della Porta (2024-11-15 03:31:45) > On 10:16 Thu 31 Oct , Andrea della Porta wrote: > > On 08:23 Tue 29 Oct , Krzysztof Kozlowski wrote: > > > > + '#clock-cells': > > > > + description: > > > > + The index in the assigned-clocks is mapped to the output clock as per > > > > + definitions in include/dt-bindings/clock/raspberrypi,rp1-clocks.h. > > > > > > You still describe how current driver matches assigned-clocks to your > > > output clocks. That's not the property of clock-cells and that's not how > > > assigned-clocks work. > > > > This description is taken by another upstream binding, please see > > Documentation/devicetree/bindings/clock/renesas,5p35023.yaml > > > > Its purpose is to let the user know how clock-cell number specified > > in assigned-clocks is mapped to the clock provided by this generator. > > Since some of these clocks are shared among peripherals, their frequency > > cannot be set by consumers, so it's the provider itself (i.e. the clock > > device described with this binding) that should take care of them. > > The renesas example has assigned-clocks specified though, please see below. > > > > > > > > There are no assigned clocks in your DTS, so this is really irrelevant > > > (or not correct, choose). > > > > In the first revision of this patchset (please see [1] and following messages) > > I had the assigned-clocks setup in the example while trying to explain their > > purpose, but Conor said those didn't seem to be relevant, hence I dropped them. > > Maybe I had to be more incisive on that. > > So, I'd be inclined to retain the description as it is and reintroduce some > > assigned-clocks in the example as in the renesas one, would it be ok for you? > > Since I'm on the verge of producing a new patchset revision, may I kindly ask > some comments on this? Is it ok for you? > Everyone knows how #clock-cells works. It shouldn't need a description about how it works. It should just point at the header file with the numbers if anything.
