On Tue, Mar 18, 2014 at 06:06:03PM -0700, Tim Kryger wrote:
> On Tue, Mar 18, 2014 at 4:47 PM, Tim Kryger <tim.kryger@xxxxxxxxxx> wrote:
> > On Tue, Mar 18, 2014 at 2:52 PM, Thierry Reding <thierry.reding@xxxxxxxxx> wrote:
> >> On Wed, Mar 12, 2014 at 01:15:43PM -0700, Tim Kryger wrote:
[...]
> >>> +static int kona_pwmc_set_polarity(struct pwm_chip *chip, struct pwm_device *pwm,
> >>> + enum pwm_polarity polarity)
> >>> +{
> >>> + /*
> >>> + * The framework only allows the polarity to be changed when a PWM is
> >>> + * disabled so no immediate action is required here. When a channel is
> >>> + * enabled, the polarity gets handled as part of the re-config step.
> >>> + */
> >>> +
> >>> + return 0;
> >>> +}
> >>
> >> See above. If you don't want to implement the hardware support for
> >> inversed polarity, then simply don't implement this.
> >
> > I had originally planned to omit polarity support but because it
> > affects the binding (which is treated as ABI), it wouldn't be possible
> > to add it in later without defining a new compatible string.
>
> I would like to get this right but it occurred to me that there may be
> a way to defer the implementation of this feature without disrupting
> the binding.
>
> Would it be acceptable to continue using #pwm-cells = <3> and
> of_pwm_xlate_with_flags but return -EINVAL from kona_pwmc_set_polarity
> if PWM_POLARITY_INVERSED is specified?
This was recently discussed for the pwm-imx driver. And you can easily
support #pwm-cells = <2> and #pwm-cells = <3> with the same binding. So
you could start with #pwm-cells = <2>, leaving out .set_polarity() and
implement it later on, extending the binding in a backwards-compatible
way to support the polarity flag.
Thierry
Attachment:
pgpiZhN9eclUW.pgp
Description: PGP signature
