Re: [PATCH 1/4] mfd: devicetree: bindings: Add Qualcomm RPM regulator subnodes

[Bjorn Andersson <bjorn.andersson@sonymobile.com>]

On Tue 03 Mar 04:47 PST 2015, Mark Brown wrote:

> On Mon, Mar 02, 2015 at 08:25:37PM -0800, Bjorn Andersson wrote:
> 
> > +- compatible:
> > +	Usage: required
> > +	Value type: <string>
> > +	Definition: must be one of:
> > +		    "qcom,rpm-pm8058-regulators"
> > +		    "qcom,rpm-pm8901-regulators"
> > +		    "qcom,rpm-pm8921-regulators"
> 
> Why do these subnodes have a compatible - do they ever appear except as
> a child of a parent of the same type of device?

The relationship with the parent node is 1:M;

in the case of 8960/8064 there is only one PMIC controlled by the RPM,
hence the dt will look like:

rpm {
	compatible = "qcom,rpm-apq8960";

	regulators {
		compatible = "qcom,rpm-pm8921-regulators";
		...
	};
};

But for 8660, and later for e.g. 8974 we have something like:

rpm {
	compatible = "qcom,rpm-msm8660";

	pm8058-regulators {
		compatible = "qcom,rpm-pm8058-regulators";

		vdd_xxx-supply = <&pm8058_s4>;
		...
	};

	pm8901-regulators {
		compatible = "qcom,rpm-pm8901-regulators";
		...
	};
};

I intended to match these by name, having one rpm-regulator device
instance and using desc->regulators_node to match regulators in the
right node - with of_node being the rpm node.

But this doesn't really map to reality, as supplies are a property of
pm8058 and pm8901 and not of the rpm. I ended up writing some custom
device registering code to instantiate the right number of regulator
children and give each of them the right of_node etc.


But as several other of-based platforms have a compatible in their
regulators node I consider that a viable and cleaner solution.

Regards,
Bjorn