Re: [PATCH 2/3] dt-bindings: phy: brcm,ns-usb2-phy: bind single CRU reg

[Rafał Miłecki]

On 06.03.2021 22:52, Rob Herring wrote:
> On Fri, Feb 26, 2021 at 12:45:00PM +0100, Rafał Miłecki wrote:
> > From: Rafał Miłecki <rafal@xxxxxxxxxx>
> >
> > The old binding was using whole DMU space. It was an overkill. DMU is a
> > big block which contains e.g. CRU which contains e.g. PLLs, PHY, pinctrl
> > and thermal blocks.
> >
> > Rework the binding to directly use a single CRU register that controls
> > USB 2.0 PHY. It's still required to reference CRU generic clkset
> > register so add a syscon for that.
> >
> > For a full DMU & CRU description see arch/arm/boot/dts/bcm5301x.dtsi .
> >
> > The old binding is deprecated now.
> >
> > Signed-off-by: Rafał Miłecki <rafal@xxxxxxxxxx>
> > ---
> > This has been verified using dt_binding_check
> >
> > I'd really like to get Rob's ack to make sure I don't do anything stupid
> >
> > It's a reworked version of my abonded 2019 patch:
> > [PATCH V2 1/2] dt-bindings: bcm-ns-usb2-phy: rework binding to use CRU syscon
> > https://patchwork.ozlabs.org/project/devicetree-bindings/patch/20190108123907.19816-1-zajec5@xxxxxxxxx/
> > ---
> >   .../bindings/phy/brcm,ns-usb2-phy.yaml        | 46 +++++++++++++++----
> >   1 file changed, 36 insertions(+), 10 deletions(-)
> >
> > diff --git a/Documentation/devicetree/bindings/phy/brcm,ns-usb2-phy.yaml b/Documentation/devicetree/bindings/phy/brcm,ns-usb2-phy.yaml
> > index b8b683ce8fa9..8e056d4d205a 100644
> > --- a/Documentation/devicetree/bindings/phy/brcm,ns-usb2-phy.yaml
> > +++ b/Documentation/devicetree/bindings/phy/brcm,ns-usb2-phy.yaml
> > @@ -16,11 +16,20 @@ properties:
> >       const: brcm,ns-usb2-phy
> >   
> >     reg:
> > -    maxItems: 1
> > -    description: DMU (Device Management Unit) address range
> > +    anyOf:
> > +      - maxItems: 1
> > +        description: PHY control register
> > +      - maxItems: 1
> > +        description: DMU (Device Management Unit) address range
> > +        deprecated: true
> >   
> >     reg-names:
> >       const: dmu
> > +    deprecated: true
> > +
> > +  brcm,syscon-clkset:
> > +    description: phandle to syscon for clkset register
> > +    $ref: /schemas/types.yaml#/definitions/phandle
>
> Don't really need this as it's just a compatible node of the parent
> node.

Thanks


> >   
> >     clocks:
> >       maxItems: 1
> > @@ -34,22 +43,39 @@ properties:
> >   
> >   required:
> >     - reg
> > -  - reg-names
> >     - clocks
> >     - clock-names
> >     - "#phy-cells"
> >   
> > +oneOf:
> > +  - required:
> > +      - brcm,syscon-clkset
> > +  - required:
> > +      - reg-names
> > +
> >   additionalProperties: false
> >   
> >   examples:
> >     - |
> >       #include <dt-bindings/clock/bcm-nsp.h>
> >   
> > -    usb2-phy@1800c000 {
> > -        compatible = "brcm,ns-usb2-phy";
> > -        reg = <0x1800c000 0x1000>;
> > -        reg-names = "dmu";
> > -        clocks = <&genpll BCM_NSP_GENPLL_USB_PHY_REF_CLK>;
> > -        clock-names = "phy-ref-clk";
> > -        #phy-cells = <0>;
> > +    cru-bus@1800c100 {
> > +        compatible = "simple-bus";
>
> A specific compatible is needed for this block.

I've just sent independently following patch:
[PATCH robh next] dt-bindings: bus: add Broadcom CRU


> > +        ranges = <0 0x1800c100 0x1a4>;
> > +        #address-cells = <1>;
> > +        #size-cells = <1>;
> > +
> > +        usb2-phy@64 {
> > +            compatible = "brcm,ns-usb2-phy";
> > +            reg = <0x64 0x4>;
> > +            brcm,syscon-clkset = <&clkset>;
> > +            clocks = <&genpll BCM_NSP_GENPLL_USB_PHY_REF_CLK>;
> > +            clock-names = "phy-ref-clk";
> > +            #phy-cells = <0>;
> > +        };
> > +
> > +        clkset: syscon@80 {
> > +            compatible = "brcm,cru-clkset", "syscon";
> > +            reg = <0x80 0x4>;
> > +        };
>
> Is this going to expand to 0x1a4/4 child nodes? The problem with one
> node per register is I don't know when it ends and you have to
> constantly update your DT.

I'm sorry, I don't fully understand that expanding question.

Most of CRU subblocks are standalone devices occupying multiple
registers. The last one is thermal@1800c2c0. That "brcm,cru-clkset"
seems to be an exception in the CRU space. clkset is a single
register so it should never become reg = <0x80 0x8>;

The finally documented CRU is expected to look like:

cru-bus@1800c100 {
	compatible = "brcm,cru", "simple-bus";
	reg = <0x1800c100 0x1d0>;
	ranges = <0 0x1800c100 0x1d0>;
	#address-cells = <1>;
	#size-cells = <1>;

	lcpll0@0 {
		compatible = "brcm,nsp-lcpll0";
		reg = <0x0 0x14>;
		(...)
	};

	genpll@40 {
		compatible = "brcm,nsp-genpll";
		reg = <0x40 0x24>;
		(...)
	};

	usb2-phy@64 {
		compatible = "brcm,ns-usb2-phy";
		reg = <0x64 0x4>;
		brcm,syscon-clkset = <&clkset>;
		clocks = <&genpll BCM_NSP_GENPLL_USB_PHY_REF_CLK>;
		clock-names = "phy-ref-clk";
		#phy-cells = <0>;
	};

	clkset: syscon@80 {
		compatible = "brcm,cru-clkset", "syscon";
		reg = <0x80 0x4>;
	};

	pin-controller@c0 {
		compatible = "brcm,bcm4708-pinmux";
		reg = <0xc0 0x24>;
		(...)
	};

	thermal@1c0 {
		compatible = "brcm,ns-thermal";
		reg = <0x1c0 0x10>;
		(...)
	};
};