On Fri, Oct 12, 2018 at 11:43:13AM +0000, Marcel Ziswiler wrote:
> On Fri, 2018-10-12 at 11:43 +0100, Russell King - ARM Linux wrote:

> > The existing DT description is reasonable, and introducing ficticious
> > regulators in DT to work around the implementation is not reasonable.

> I don't think it is that fictitious as it makes it crystal clear that
> there is something shared with all its pros and cons. E.g. what happens
> if one of them regulators wants to turn off while the other one still
> needs power? The regular regulator dependency tree would nicely make
> this all clear.

We already have code to handle that via refcounting on the GPIO once we
identify that it's the same GPIO.  If we make a shared virtual parent
regulator that'll break other things where we're tracking what the
actual physical parent for voltage reasons like adjusting parent
voltages up and down to improve efficiency or handling things that are
just dumb power switches rather than actual regulators.