Re: [PATCH] pinctrl: at91-pio4: add support for drive-strength property

[Rob Herring]

On Mon, Jan 22, 2018 at 09:37:38AM +0100, Linus Walleij wrote:
> On Thu, Jan 18, 2018 at 5:02 PM, Ludovic Desroches
> <ludovic.desroches@xxxxxxxxxxxxx> wrote:
> 
> > Add support for the drive-strength property. Usually its value is
> > expressed in mA. Since the numeric value depends on VDDIOP voltage,
> > the controller uses low, medium and high to define the drive-strengh.
> 
> Aha I see. That's complex. It certainly results in a certain mA drive
> strength in the end, but what you're saying is that this is not usually
> what we configure.
> 
> > The PIO controller accepts two values for the low drive: 0 or 1. Most
> > of the time, we don't care about the drive strength, there is no need
> > to change it, so 0 is considered as the default value.
> 
> Do you mean default value as in "whatever the hardware was set
> up as at boot time"?
> 
> > The low-drive
> > value won't be advertised through pinconf-pins file excepted if it
> 
> except?
> 
> > has been set explicitly in the device tree ie if its value is
> > different from 0.
> >
> > Signed-off-by: Ludovic Desroches <ludovic.desroches@xxxxxxxxxxxxx>
> 
> OK I think I get it.
> 
> >  Optional properties:
> >  - GENERIC_PINCONFIG: generic pinconfig options to use, bias-disable,
> > -bias-pull-down, bias-pull-up, drive-open-drain, input-schmitt-enable,
> > -input-debounce, output-low, output-high.
> > +bias-pull-down, bias-pull-up, drive-open-drain, drive-strength,
> > +input-schmitt-enable, input-debounce, output-low, output-high.
> (...)
> > +                       drive-strength = <ATMEL_PIO_DRVSTR_LO>;
> 
> So you say you support this argument and it will be something like
> 
> include/dt-bindings/pinctrl/at91.h:#define ATMEL_PIO_DRVSTR_LO  1
> include/dt-bindings/pinctrl/at91.h:#define ATMEL_PIO_DRVSTR_ME  2
> include/dt-bindings/pinctrl/at91.h:#define ATMEL_PIO_DRVSTR_HI  3
> 
> But the definition if generic drive strength is actually in mA.

Yes, and the reason we put unit suffixes on properties is to avoid 
differing units.
 
> I think it is OK to deviate from stating it in mA, but you should
> write this in the DT bindings so people do not get confused.

I don't think it is okay. If "drive-strength" doesn't work, then use 
a vendor specific property ("atmel,drive-strength"). Of course, I might 
forget this in the next version and tell you to use a standard property 
in mA. 

Rob