Hi!

(Please keep all the recipiens in Cc)

On Sun, Dec 02, 2018 at 12:13:21AM +0800, Hao Zhang wrote:

Maxime Ripard [off-list ref] 于2018年11月27日周二 下午6:33写道：

quoted

On Tue, Nov 27, 2018 at 09:35:23AM +0100, Uwe Kleine-König wrote:

quoted

Hello,

On Tue, Nov 27, 2018 at 08:52:26AM +0100, Maxime Ripard wrote:

quoted

On Mon, Nov 26, 2018 at 12:18:59AM +0800, Hao Zhang wrote:

quoted

+  - clocks: From common clock binding, handle to the parent clock.
+  - clock-names: Must contain the clock names described just above.

[...]

You seem to have used mux-0 and mux-1 for the clock names. I guess we
don't have to use a name there, we can simply use the position to find
out (as long as it's documented in the binding)

I also wondered if the driver relies on the fact that the second clock
is the faster running one. Is this sensible?

Not really, I'm not sure we can make those expectations in the DT
binding, especially since clock rate can change at runtime.

How about just add one clock on DT, most of the time, 24MHZ is enough
(apb1 is 100MHZ)
other one just use as a optional.
clock rate change at runtime would make the same pair pwm channel
uncontrollable,
because previous one would be change by the new one different setting.

The DT is a hardware representation. If the hardware block can use
both clocks, it should be described.

Now, you can totally use only one clock of these 2 in your driver if that's
easier / more reasonable.

Maxime

-- 
Maxime Ripard, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com