[PATCH 2/5] pinctrl: clarify some dt pinconfig options

[Linus Walleij]

On Thu, Jun 20, 2013 at 12:10 AM, Stephen Warren <swarren at wwwdotorg.org> wrote:
> On 06/14/2013 09:42 AM, Heiko Stübner wrote:

>> diff --git a/Documentation/devicetree/bindings/pinctrl/pinctrl-bindings.txt b/Documentation/devicetree/bindings/pinctrl/pinctrl-bindings.txt
>
>> -low-power-mode               - low power mode
>> +low-power-enable     - enable low power mode
>> +low-power-disable    - disable low power mode
>
> Hmmm. That's changing the binding definition. What if somebody already
> wrote their device tree according previous definition?

It's not merged so see it as alterations to a WIP in the
turners workshop or something.

> It seems to be that tri-states are preferable for pinctrl DT:
>
> no entry: do nothing
> = 0: disable
> = 1: enable

Better with explict enable/disable strings instead of
<0> or <1> I think, but the semantic effect would be the
same I guess, the upside with *enable/*disable strings is
that we do not have to handle cases like
tristate = <2>; ...

>> +Arguments for parameters:
>> +
>> +- bias-pull-up, -down and -pin-default take as optional argument 0 to disable
>> +  the pull, on hardware supporting it the pull strength in Ohm. bias-disable
>> +  will also disable any active pull.
>
> Does this agree with the latest definition of the kernel-internal
> meaning of 0 for pull-up/down?

No that is wrong. Heiko, care to fix this binding doc?

>> +- input-schmitt takes as argument the adjustable hysteresis in a
>> +  driver-specific format
>> +
>> +- input-debounce takes the debounce time as argument or 0 to disable debouncing
>> +
>> +- power-source argument is the custom value describing the source to select
>> +
>> +- slew-rate takes as argument the target rate in a driver-specific format
>
> If those things have driver-specific (note: should be
> DT-binding-specific, not driver-specific) values, then I'm not convinced
> that having a generic parameter name for them is a good idea; it makes
> things look the same when they aren't. By forcing each binding to
> include the vendor prefix on those properties and hence define a custom
> property name, you're making it clear that the semantics may be different.

Hmmm I don't think they're used right now, let's deal with them when
we have something to showcase them with. Patches to delete the
unclear bindings will be considered...

Yours,
Linus Walleij