On Sat, Jul 23, 2022 at 08:11:01PM +0100, Mark Brown wrote:

> The more I think about this the more I think that what's
> practically going on here fits better with the mode interface
> than the load one, normally I'd not like doing a run around the
> abstractions the underlying thing is offering but it really does
> seem like someone has to bite the bullet and do that runaround
> and it's probably going to be less work all round at the
> regulator level.

Actually thinking about it even more for most cases it may even
just be that we need to add some number to the load request while
the regulator is powered on, the existing system-load property
would probably do the trick here if putting things in DT.
Depending on if the firmware counts loads from things that voted
off or not the driver might need to do an update on the load when
enabling and disabling.  That is less of an end run around the
abstraction, though it does cause issues for the system-load's
intended use so perhaps a separate property might be safer.