On Fri, Jan 11, 2019 at 1:54 PM Thomas Petazzoni
[off-list ref] wrote:

On Fri, 11 Jan 2019 10:54:20 +0100, Linus Walleij wrote:

quoted

My stance is that the driver is responsible of enabling and managing
runtime PM for its hardware block(s).

Runtime PM in the core should only be added if the core needs to
be aware about it, such as is the case when e.g. a block device
needs to drain its write buffer before going to runtime sleep.

I fail so see why the GPIO core need to be aware about this.

In this very same thread at
https://www.spinics.net/lists/arm-kernel/msg600515.html, you kind of
proposed to handle this in the core in fact :-) Though indeed you said
that the core could provide helpers.

Yeah allright, I have never been good with consistency but what I guess
I would mean to say (today) is that the driver needs to be in the driver
seat (heh) and opting in to any runtime PM support.

This is in contrast with "midlayer" where all drivers are forced to behave
"as if" they had runtime PM (i.e. calls are done to the runtime PM helpers
even if the device doesn't really activate runtime PM).

Yours,
Linus Walleij

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel