Re: [PATCH/RFC v4 3/3] DT: Add documentation for gp2ap020a00f sensor

[Pawel Moll <pawel.moll@arm.com>]

On Mon, 2013-08-19 at 11:55 +0100, Sylwester Nawrocki wrote:
> On 08/16/2013 03:28 PM, Pawel Moll wrote:
> > On Fri, 2013-08-16 at 14:12 +0100, Jacek Anaszewski wrote:
> >> diff --git a/Documentation/devicetree/bindings/iio/light/gp2ap020a00f.txt b/Documentation/devicetree/bindings/iio/light/gp2ap020a00f.txt
> >> new file mode 100644
> >> index 0000000..e4f377d
> >> --- /dev/null
> >> +++ b/Documentation/devicetree/bindings/iio/light/gp2ap020a00f.txt
> >> @@ -0,0 +1,20 @@
> >> +* Sharp GP2AP020A00F I2C Proximity/ALS sensor
> >> +
> >> +Required properties:
> >> +
> >> +  - compatible : should be "sharp,gp2ap020a00f"
> >> +  - reg : the I2C address of the light sensor
> >> +  - interrupt-parent : phandle to the parent interrupt controller
> >> +  - interrupts : should be INT interrupt pin
> 
> This seems a bit misleading, perhaps:
> 
> interrupts : should be interrupt number the INT pin interrupt is routed to
> 
> ?
> > Any particular reason for the interrupt-parent being required? Couldn't
> > it be inherited, as is happens in most cases?
> 
> I think this was intended for the cases when the sensor is connected
> to a GPIO and the GPIO interrupt controller is used. For example on
> the Exynos SoCs with each GPIO bank is associated separate interrupt
> controller instance (see arch/arm/boot/dts/exynos4210-pinctrl.dtsi
> for details). Thus one of those interrupt controllers need to be
> selected by dts, rather than inheriting the GIC.
> 
> I'm not sure if "interrupt-parent" needs to be part of DT binding of
> this sensor device. Perhaps it could be an optional property or should
> it be dropped completely ? I guess it should be considered as part
> of the interrupt controller binding ?

Yes, I believe so. At least this is the de-facto standard for existing
device bindings documentations. Hopefully the DT-schemas-being-forged
will clarify the matter once for all.

Pawel