On 02/05/2022 12:40, Ansuel Smith wrote: > > The idea is that you put the clk name in 'clock-output-names' and the > driver needs to have support for it (and set the clk name based on the > name defined in the dts) > > This driver doesn't have support for it and is actually hardcoded. > So you are right and I should just drop it. > > But now another question... Since #clock-cells was added as a > requirement for clock-output-names, should I drop also that? > > In theory #clock-cells should always be declared for clock providers, is > it right to add it in the conversion commit or I should put this change > in another commit? (since it's now an addition and now something required > to fix a bot warning) These are not the best bindings to convert, if you are not into the qcom DTS and drivers. :) It looks like the bindings were added to match current Linux implementation and in this implementation the device is not used in DTS as a clock provider (even though it registers a clock) but as a syscon. I am not even sure if it is used as a clock provider outside of DTS (through using a fixed clock name in some clock consumer). Probably this should be made either a proper clock controller or something stripped down to the point matching current usage (accepting the fact that bindings are incomplete). Anyway your choice should be made according to how this device and its driver fit to entire system. IOW, it's not a simple binding conversion and you should not just convert it to make dtbs_check happy. Best regards, Krzysztof
