Re: [PATCH] pwm-pca9685: Allow any of the 16 PWMs to be used as a GPIO

[Thierry Reding <thierry.reding@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 1630 bytes --]

On Thu, Oct 20, 2016 at 03:08:53PM +0200, Clemens Gruber wrote:
> On Thu, Oct 20, 2016 at 02:51:44PM +0200, Thierry Reding wrote:
> > I'm reluctant to add this to the PWM core because it's effectively
> > duplicating something for which a proper subsystem already exists.
> > 
> > If adding pinmux is considered overkill, maybe doing so needs to be
> > simplified. Surely if this is applicable to more than one PWM controller
> > some helpers could be extracted to make it easier to add.
> > 
> > But if it isn't generally useful I don't think it makes sense to add it
> > to the PWM core either. So I think our choices here are to register via
> > pinmux and in the process add helpers to make that easier (remove the
> > overkill) or to keep this to the driver until we start seeing a pattern
> > emerge.
> 
> But the fact that this one pwm chip bypasses PWM control and acts as a
> GPO is an implementation detail of the chip, this feature is helpful for
> other PWM chips too, as long as they don't toggle.
> Would this really be a good fit in the pinmux subsystem?
> 
> I really like the idea of Mika to add a gpio_set callback in pwm_ops and
> let the pwm core export a GPIO chip. For now we could add this callback
> only to the pwm-pca9685 driver but others could opt-in if needed.

I don't think it's as simple as that. There could be chips where only a
subset of PWM channels could act as a GPIO, or there could be any other
amount of quirks. A single callback is unlikely to cut it, so I think a
better alternative is to keep this in drivers until some patterns start
to emerge.

Thierry

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 801 bytes --]