Re: [PATCH v5 2/7] pwm: pca9685: Support hardware readout

[Sven Van Asbroeck <thesven73@gmail.com>]

On Mon, Jan 11, 2021 at 3:35 PM Uwe Kleine-König
<u.kleine-koenig@pengutronix.de> wrote:
>
> My position here is: A consumer should disable a PWM before calling
> pwm_put. The driver should however not enforce this and so should not
> modify the hardware state in .free().
>
> Also .probe should not change the PWM configuration.

I agree that this is the most user-friendly behaviour.

The problem however with the pca9685 is that it has many degrees of
freedom: there are many possible register values which produce the same
physical chip outputs.

This could lead to a situation where, if .probe() does not reset the register
values, subsequent writes may lead to different outputs than expected.

One possible solution is to write .get_state() so that it always reads the
correct state, even if "unconventional" register settings are present, i.e.
those written by an outside entity, e.g. a bootloader. Then write that
state back using driver conventions.

This may be trickier than it sounds - after all we've learnt that the pca9685
looks simple on the surface, but is actually quite challenging to get right.

Clemens, Uwe, what do you think?

Sven