Re: [PATCHv3 1/3] pwm: make the PWM_POLARITY flag in DTB optional

[Sascha Hauer <s.hauer@pengutronix.de>]

On Wed, Apr 09, 2014 at 09:22:50AM +0200, Thierry Reding wrote:
> On Wed, Apr 09, 2014 at 08:12:09AM +0200, Sascha Hauer wrote:
> > On Tue, Apr 08, 2014 at 01:43:22PM -0700, Tim Kryger wrote:
> > > On Mon, Apr 7, 2014 at 10:02 PM, Lothar Waßmann <LW@karo-electronics.de> wrote:
> > > > Thierry Reding wrote:
> > > 
> > > >> No. You cannot emulate polarity inversion in software.
> > > >>
> > > > Why not?
> > > >
> > > > duty_ns = period_ns - duty_ns;
> > > 
> > > Since I made the same mistake, I will pass along the pointer Thierry gave me.
> > > 
> > > In include/linux/pwm.h the second difference for an inverted signal is
> > > described.
> > > 
> > > /**
> > >  * enum pwm_polarity - polarity of a PWM signal
> > >  * @PWM_POLARITY_NORMAL: a high signal for the duration of the duty-
> > >  * cycle, followed by a low signal for the remainder of the pulse
> > >  * period
> > >  * @PWM_POLARITY_INVERSED: a low signal for the duration of the duty-
> > >  * cycle, followed by a high signal for the remainder of the pulse
> > >  * period
> > >  */
> > > enum pwm_polarity {
> > > PWM_POLARITY_NORMAL,
> > > PWM_POLARITY_INVERSED,
> > > };
> > > 
> > > Of course, I suspect not all PWM hardware respects this definition of
> > > inverted output.
> > > 
> > > Either way, hacking the duty in software certainly would get the
> > > high/low order wrong.
> > 
> > This only relevant if you have some reference signal the PWM must be
> > relative to, for example if you combine multiple PWMs for motor control.
> > For PWMs used for backlight or beepers a signal inversion in software is
> > perfectly fine. And I also think that it makes sense to put it once into
> > the framework instead of bothering all consumer drivers with the
> > inversion.
> 
> The PWM framework itself doesn't have enough knowledge about what a PWM
> is being used for. Therefore it cannot determine whether emulating
> polarity inversion by reversing the duty cycle will be appropriate.
> 
> Putting such functionality into the core will prevent PWM channels from
> being used for cases where the signal polarity does matter

The PWM core is in no way prepared for handling such situations. Should
we want to add support it a PWM_POLARITY_INVERSED flag would be the
least of our problems. It could be renamed to
PWM_POLARITY_INVERSED_ASYNC for the beeper/led drivers which do not need
synchronization.

Sascha

-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |