On Sun, Jan 16, 2022 at 7:28 PM Jonathan Cameron <jic23@xxxxxxxxxx> wrote: > > > > > * Have a clock property per channel. Note that we this I moved to OF > > since we now have to use 'devm_get_clk_from_child()' which is using > > device_node. Note that I could use 'to_of_node()' but mixing of.h and > > property.h does not feel like a good idea. > > Ah, that's unfortunate given the clk is only needed in certain modes... > > Andy/Rafael/Rob, any thoughts on how we should handle this? Obviously > ACPI and clocks is generally a no go, but in this particular case we > aren't looking at a power management related use of clocks, but rather > using the clk framework to provide a way to control one of our inputs > used to generate the output dithered signal... If the device just its > own clock then we'd just control it directly with no problems, but it uses > and external source. > > We don't know of anyone actually looking at this device in conjunction with > ACPI so maybe just using dt specific calls for now rather than generic > firmware properties is the best we can do. The problem is the CCF is so OF-centric and maintainer(s) seems skeptical about switching it to use fwnode APIs (they wanted, last time I tried, to show how exactly it will be used, so here is your chance!), but OTOH switching to fwnode API is a half made road, because for ACPI we might need a glue layer which may be way too tricky to be considered as a part of this series. -- With Best Regards, Andy Shevchenko
