2015-12-16 10:11 GMT+09:00 Sebastian Reichel [off-list ref]:

Hi,

On Tue, Dec 15, 2015 at 04:53:31PM -0800, Eric Anholt wrote:

quoted

quoted

quoted

quoted

What motivated the location of this power domain driver in
arch/arm/mach-bcm? Should not we have this in drivers/power/ or
somewhere in drivers/ at the very least?

ls stronly suggests that power contains drivers for power supplies and
batteries, not power domains.

Indeed it's used for fuel gauges and chargers, but also for
reboot/powerdown and adaptive voltage scaling, so another
subdirectory for power-domains wouldn't be that odd.

quoted

quoted

quoted

There are 6 power domain drivers in
arch/arm, 3 in drivers/clk, and 3 in drivers/soc.

If we ever have to support a different architecture which happens to use
a similar power domain, then we want it to be in a location which makes
it easy for sharing it in the first place. As it stands today, it does
not seem useful to me to have this code in arch/arm/mach-bcm/ at all.

Maybe there is room from a drivers/power/domains/ of some kind?

I like the idea, but let's include generic power domain maintainers
in this discussion, as I suggested here (I got a power domain driver
patch for drivers/power just a few days ago):

https://lkml.org/lkml/2015/12/15/815

Also somebody would have to step up to maintain that directory.

This could go into drivers/soc. We put there a lot of mach-specific
stuff which we want to make a little more generic (like generic enough
multiplatform, multiarchitecture etc). Rockchip has its own power
domains there. Dove and Mediatek seem as well but I am not sure. Some
other architectures keep this still in arm/mach (exynos, ux500, zx,
imx, s34c64xx, shmobile) but this looks more of like a legacy choice.

However, since generic power domains have its own maintainers entry
and reside under drivers/base/power, maybe making a separate directory
for power domains drivers makes sense?

Best regards,
Krzysztof