On 24/02/2022 20:14:51+0200, Sakari Ailus wrote: > Hi Mark, > > On Thu, Feb 24, 2022 at 03:33:12PM +0000, Mark Brown wrote: > > On Thu, Feb 24, 2022 at 03:58:04PM +0100, Hans de Goede wrote: > > > > > As Mark already mentioned the regulator subsystem has shown to > > > be a bit problematic here, but you don't seem to need that? > > > > I believe clocks are also potentially problematic for similar reasons > > (ACPI wants to handle those as part of the device level power management > > and/or should have native abstractions for them, and I think we also > > have board file provisions that work well for them and are less error > > prone than translating into an abstract data structure). > > Per ACPI spec, what corresponds to clocks and regulators in DT is handled > through power resources. This is generally how things work in ACPI based > systems but there are cases out there where regulators and/or clocks are > exposed to software directly. This concerns e.g. camera sensors and lens > voice coils on some systems while rest of the devices in the system are > powered on and off the usual ACPI way. > > So controlling regulators or clocks directly on an ACPI based system > wouldn't be exactly something new. All you need to do in that case is to > ensure that there's exactly one way regulators and clocks are controlled > for a given device. For software nodes this is a non-issue. > > This does have the limitation that a clock or a regulator is either > controlled through power resources or relevant drivers, but that tends to > be the case in practice. But I presume it wouldn't be different with board > files. > In this use case, we don't need anything that is actually described in ACPI as all the clocks we need to control are on the device itself so I don't think this is relevant here. -- Alexandre Belloni, co-owner and COO, Bootlin Embedded Linux and Kernel engineering https://bootlin.com
