Re: [PATCH v2 01/15] dt-bindings: gpio: convert bindings for NXP PCA953x family to dtschema

[Grygorii Strashko]

On 11/09/2020 09:52, Krzysztof Kozlowski wrote:
> On Fri, 11 Sep 2020 at 08:24, Joel Stanley <joel@xxxxxxxxx> wrote:
> > On Thu, 10 Sep 2020 at 17:57, Krzysztof Kozlowski <krzk@xxxxxxxxxx> wrote:
> > > Convert the NXP PCA953x family of GPIO expanders bindings to device tree
> > > schema.
> > >
> > > Signed-off-by: Krzysztof Kozlowski <krzk@xxxxxxxxxx>
> >
> > > +patternProperties:
> > > +  "^(hog-[0-9]+|.+-hog(-[0-9]+)?)$":
> > > +    type: object
> > > +    properties:
> > > +      gpio-hog: true
> > > +      gpios: true
> > > +      input: true
> > > +      output-high: true
> > > +      output-low: true
> > > +      line-name: true
> > > +
> > > +    required:
> > > +      - gpio-hog
> > > +      - gpios
> > > +
> >
> > > +            usb3-sata-sel-hog {
> > > +                gpio-hog;
> > > +                gpios = <4 GPIO_ACTIVE_HIGH>;
> > > +                output-low;
> > > +                line-name = "usb3_sata_sel";
> >
> > I would prefer we didn't require the addition of hte -hog prefix. It's
> > mostly just a matter of taste, but I can think of a few more concrete
> > reasons:
> >
> > We don't require -high or -low prefixes, so the node name doesn't need
> > to describe the properties that will be found below.
>
> Thanks for the comments.
>
> It is not about properties (high or low) but the role of a device
> node. The node names should represent a generic class of device (ePAPR
> and device tree spec) and "hog" is such class.
>
> The Documentation/devicetree/bindings/gpio/fsl-imx-gpio.yaml already
> uses such naming so the best would be to unify.

In my opinion, It's not right to define this on per gpio-controller and introduce such
per gpio-controller restrictions.

More over, there is already generic schema for gpio hogs: gpio-hog.yaml
Originally, gpio bindings were defined without restricting gpio hog node names and,
generic schema follows this.

I think, the generic "gpio-hogs" sub-node may be introduced to place gpio hogs child nodes,
if gpio hogs node names restriction need to be introduces (*which i'm not sure is reasonable*).

gpio@20 {
	gpio-hogs {
		yyy-hog {
                        gpio-hog;
                        gpios
		}
	}

But this require as gpio code as generic gpio schema update (with backward compatibility in mind).


> > Changing around node names for existing boards carries with it the
> > chance of userspace breakage (as sysfs paths change). I would prefer
> > we avoid that if possible.
>
> The impact on userspace is indeed important, but are you sure that
> hogs are visible to user-space via sysfs and configurable? I guess you
> think of deprecated CONFIG_GPIO_SYSFS?
>
> Rob,
> Any hints from you about hog-naming?
>
> Best regards,
> Krzysztof

--
Best regards,
grygorii