Re: [PATCH 1/3] dt-bindings: Add support for Broadcom USB pin map driver

[Alan Cooper]

On Tue, Aug 25, 2020 at 11:46 AM Rob Herring <robh@xxxxxxxxxx> wrote:
>
> +Linus W
>
> On Tue, Aug 25, 2020 at 6:26 AM Alan Cooper <alcooperx@xxxxxxxxx> wrote:
> >
> > On Mon, Aug 24, 2020 at 7:30 PM Rob Herring <robh@xxxxxxxxxx> wrote:
> > >
> > > On Wed, Aug 12, 2020 at 04:20:16PM -0400, Al Cooper wrote:
> > > > Add DT bindings for the Broadcom USB pin map driver. This driver allows
> > > > some USB input and output signals to be mapped to any GPIO instead
> > > > of the normal dedicated pins to/from the XHCI controller.
> > >
> > > Is this a driver or h/w block because bindings are for h/w blocks?
> >
> > This is a hardware block. I'll remove "driver" from the description.
>
> Another question, this kind of looks like a pin mux controller. Is
> that not a fit for this? If not, why?

This driver is not doing any pin-muxing of a physical pin on the chip.
Instead it's using standard gpio's, through gpiolib, and propagating
the gpio state for in-coming signals to a special register that feeds
into a XHCI host controller register and it's propagating the state of
out-going signals from the special register fed by a XHCI controller
register to a gpio. Both directions are interrupt driven and
continually mirroring the state between the XHCI host controller
registers and the gpios. I don't see any pinmux/pinctrl driver doing
this kind of thing.

Thanks
Al

>
> > > > Signed-off-by: Al Cooper <alcooperx@xxxxxxxxx>
> > > > ---
> > > >  .../bindings/usb/brcm,usb-pinmap.yaml         | 63 +++++++++++++++++++
> > > >  1 file changed, 63 insertions(+)
> > > >  create mode 100644 Documentation/devicetree/bindings/usb/brcm,usb-pinmap.yaml
> > > >
> > > > diff --git a/Documentation/devicetree/bindings/usb/brcm,usb-pinmap.yaml b/Documentation/devicetree/bindings/usb/brcm,usb-pinmap.yaml
> > > > new file mode 100644
> > > > index 000000000000..19cf6ad36373
> > > > --- /dev/null
> > > > +++ b/Documentation/devicetree/bindings/usb/brcm,usb-pinmap.yaml
> > > > @@ -0,0 +1,63 @@
> > > > +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
> > > > +%YAML 1.2
> > > > +---
> > > > +$id: http://devicetree.org/schemas/usb/brcm,usb-pinmap.yaml#
> > > > +$schema: http://devicetree.org/meta-schemas/core.yaml#
> > > > +
> > > > +title: Broadcom USB pin map Controller Device Tree Bindings
> > > > +
> > > > +maintainers:
> > > > +  - Al Cooper <alcooperx@xxxxxxxxx>
> > > > +
> > > > +properties:
> > > > +  compatible:
> > > > +      items:
> > > > +          - const: brcm,usb-pinmap
> > >
> > > 2 space indentation please.
> >
> > Fixed.
> >
> > >
> > > > +
> > > > +  reg:
> > > > +    maxItems: 1
> > > > +
> > > > +  interrupts:
> > > > +    maxItems: 1
> > > > +    description: Must be defined if any out-gpios are specified.
> > >
> > > 'dependencies' can express this in schema.
> >
> > Okay.
> >
> > >
> > > > +
> > > > +  in-gpios:
> > > > +    description: Array of one or more GPIO pins used for input signals.
> > >
> > > You need to define how many GPIOs are valid.
> >
> > I tried to avoid doing this because there is a possibility that future
> > chips will have a few more signals added and the driver was written so
> > new signals can be added entirely in device tree without any changes
> > to the driver. If this is unacceptable, I can add the current max in
> > and out valid gpios.
>
> A 'should be enough for a while' value is fine. The driver doesn't
> have to have a max. I'd expect the binding to be updated for new SoCs
> anyways.
>
> > >
> > > > +
> > > > +  in-names:
> > > > +    description: Array of input signal names, one per gpio in in-gpios.
> > >
> > > No, this isn't how we name GPIOs. The part before '-gpios' is how.
> >
> > This is the meant to be a description of how each gpio is being used
> > to help with error messages in the driver.
> > What if I use "brcmstb,in-functions" instead?
>
> 'brcmstb' is not a vendor. But brcm,in-functions is fine.
>
> > > > +
> > > > +  in-masks:
> > > > +    description: Array of enable and mask pairs, one per gpio in-gpios.
> > >
> > > Needs a vendor prefix.
> >
> > I'll change it to "brcmstb,in-masks"
> >
> > >
> > > > +
> > > > +  out-gpios:
> > > > +    description: Array of one or more GPIO pins used for output signals.
> > > > +
> > > > +  out-names:
> > > > +    description: Array of output signal names, one per gpio in out-gpios.
> > > > +
> > > > +  out-masks:
> > > > +    description: Array of enable, value, changed and clear masks, one
> > > > +      per gpio in out-gpios.
> > > > +
> > > > +required:
> > > > +  - compatible
> > > > +  - reg
> > > > +
> > > > +additionalProperties: false
> > > > +
> > > > +examples:
> > > > +  - |
> > > > +    usb_pinmap: usb-pinmap@22000d0 {
> > > > +        compatible = "brcm,usb-pinmap";
> > > > +        reg = <0x22000d0 0x4>;
> > > > +        in-gpios = <&gpio 18 0>, <&gpio 19 0>;
> > > > +        in-names = "VBUS", "PWRFLT";
> > > > +        in-masks = <0x8000 0x40000 0x10000 0x80000>;
> > > > +        out-gpios = <&gpio 20 0>;
> > > > +        out-names = "PWRON";
> > > > +        out-masks = <0x20000 0x800000 0x400000 0x200000>;
> > > > +        interrupts = <0x0 0xb2 0x4>;
> > > > +    };
> > > > +
> > > > +...
> > > > --
> > > > 2.17.1
> > > >