Re: [PATCH v8 3/3] arm64: dts: rockchip: add usb2-phy support for rk3399

[Heiko Stübner]

Hi Frank,

Am Donnerstag, 21. Juli 2016, 10:49:53 schrieb Frank Wang:
> >> @@ -69,6 +69,15 @@
> >> 
> >>                  regulator-max-microvolt = <3300000>;
> >>          
> >>          };
> >> 
> >> +       vbus_host: vbus-host-regulator {
> >> +               compatible = "regulator-fixed";
> >> +               enable-active-high;
> >> +               gpio = <&gpio4 25 GPIO_ACTIVE_HIGH>;
> >> +               pinctrl-names = "default";
> >> +               pinctrl-0 = <&host_vbus_drv>;
> >> +               regulator-name = "vbus_host";
> >> +       };
> >> +
> > 
> > To match my schematics, this would probably be "vcc5v0_host".
> > Technically there are two regulators but since they are the same
> > voltage and enabled by the same GPIO it seems like modeling it as one
> > regulator is fine.
> 
> Yep, you are right, I will rename it.
> 
> > If you really wanted to model things you could also include the input
> > supply (VCC5V0_SYS).  Not sure how much you care to model in EVB.
> 
> Actually, from
> "Documentation/devicetree/bindings/regulator/fixed-regulator.txt" show,
> input supply name is just optional property, and it seems that only do
> assign "vin" value for input_supply (the second member of struct
> fixed_voltage_config) if "vin-supply" is specified.
> 
> So is input supply name  (VCC5V0_SYS) required here? Would you like to
> give more comments please?

While vin-supply is optional, I think that is meant for real top-level 
regulators (our vcc_sys or whatever) that really don't have a parent 
regulator.

It is always nicer to model the whole power-tree [in a sane way], as it makes 
following the schematics a lot easier. If you mount a debugfs these days you 
can even get a nice tree graph of the regulator infrastructure ... where the 
parent-relationship is also needed to create something meaningful.


Heiko