Re: [PATCH 1/2] dt-bindings: phy: Add Sophgo CV1800 USB phy

[Inochi Amaoto]

On Mon, Apr 22, 2024 at 05:21:26PM GMT, Conor Dooley wrote:
> On Sat, Apr 20, 2024 at 09:39:18AM +0800, Inochi Amaoto wrote:
> > On Fri, Apr 19, 2024 at 03:26:53PM GMT, Conor Dooley wrote:
> > > On Fri, Apr 12, 2024 at 03:22:24PM +0800, Inochi Amaoto wrote:
> > > > The USB phy of Sophgo CV18XX series SoC needs to sense a pin called
> > > > "VBUS_DET" to get the right operation mode. If this pin is not
> > > > connected, it only supports setting the mode manually.
> > > > 
> > > > Add USB phy bindings for Sophgo CV18XX/SG200X series SoC.
> > > > 
> > > > Signed-off-by: Inochi Amaoto <inochiama@xxxxxxxxxxx>
> > > > ---
> > > >  .../bindings/phy/sophgo,cv1800-usb-phy.yaml   | 90 +++++++++++++++++++
> > > >  1 file changed, 90 insertions(+)
> > > >  create mode 100644 Documentation/devicetree/bindings/phy/sophgo,cv1800-usb-phy.yaml
> > > > 
> > > > diff --git a/Documentation/devicetree/bindings/phy/sophgo,cv1800-usb-phy.yaml b/Documentation/devicetree/bindings/phy/sophgo,cv1800-usb-phy.yaml
> > > > new file mode 100644
> > > > index 000000000000..cb394ac5d8c4
> > > > --- /dev/null
> > > > +++ b/Documentation/devicetree/bindings/phy/sophgo,cv1800-usb-phy.yaml
> > > > @@ -0,0 +1,90 @@
> > > > +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
> > > > +%YAML 1.2
> > > > +---
> > > > +$id: http://devicetree.org/schemas/phy/sophgo,cv1800-usb-phy.yaml#
> > > > +$schema: http://devicetree.org/meta-schemas/core.yaml#
> > > > +
> > > > +title: Sophgo CV18XX/SG200X USB 2.0 PHY
> > > > +
> > > > +maintainers:
> > > > +  - Inochi Amaoto <inochiama@xxxxxxxxxxx>
> > > > +
> > > > +properties:
> > > > +  compatible:
> > > > +    const: sophgo,cv1800-usb-phy
> > > > +
> > > > +  reg:
> > > > +    maxItems: 1
> > > > +
> > > > +  "#phy-cells":
> > > > +    const: 0
> > > > +
> > > > +  clocks:
> > > > +    items:
> > > > +      - description: PHY clock
> > > > +      - description: PHY app clock
> > > > +      - description: PHY stb clock
> > > > +      - description: PHY lpm clock
> > > > +
> > > > +  clock-names:
> > > > +    items:
> > > > +      - const: phy
> > > > +      - const: app
> > > > +      - const: stb
> > > > +      - const: lpm
> > > > +
> > > > +  dr_mode:
> > > > +    description: PHY device mode when initing.
> > > 
> > > "initing" isn't a word, "initialising" is the correct word. Or
> > > "initializing" if you speak American. But if it is just the value during
> > > initialisation, why do we need to know - we can just overwrite it in
> > > software, right?
> > > 
> > > Are you sure that this is limited to initialisation? I would have
> > > thought that it describes the configuration that the board is in, and is
> > > a fixed property of the system?
> > > 
> > > > +    enum: [host, peripheral, otg]
> > > 
> > > Rob, I know this is a phy rather than a controller, so referencing
> > > usb-drd.yaml doesn't really make sense, but there are several PHYs using
> > > dr_mode so it feels like there should be something to reference here
> > > rather than defining the property anew.
> > > 
> > 
> > Yes, you are right. Using dr_mode in initialisation is not necessary.
> > We can just let it go and using the default mode. In fact, for most
> > boards with this SoC, host mode is just enough. And it is just easy 
> > to overwrite the mode value in the driver if we want another mode. 
> > For the OTG, it can just check the `vbus_det-gpios`. I will remove 
> > this property, thanks.
> 
> Just to be clear, it's valid to have a dr_mode property in cases that
> you cannot detect what the mode is dynamically. What I was questioning
> was the wording about only valid for init.
> 

OK, for the USB phy of CV1800, it always needs to start at host mode.
Because it needs to switch to both mode when initializing. As a result,
the "dr_mode" property is just added to decide which mode is set after
initializing (mostly for functionality). This is why I say it is for 
initializing.

Now, it is clean that setting dr_mode is just a function hint, and user 
can just overwrite the mode. So I decided to remove this.

> > > > +  vbus_det-gpios:
> > > > +    description: GPIO to the USB OTG VBUS detect pin. This should not be
> > > > +      defined if vbus_det gpio and switch gpio are connected.
> > > 
> > > I don't understand the second sentence here.
> 
> Ah, with your explanation below I now understand what you mean here. I
> think this needs to be re-written - I think it would be easier to
> understand with s/gpio/pin/ in the second line.
> 
> > > > +    maxItems: 1
> > > > +
> > > > +  sophgo,switch-gpios:
> > > > +    description: GPIO for the phy to control connected switch.
> > > > +    maxItems: 2
> > > 
> > > You've got two items here, they should be described. /But/ the property
> > > above says "switch gpio", which is singular. Which is it?
> > > 
> > 
> > `switch-gpios` is gpios to controll USB switch connected to the
> > phy. Sophgo recommends that phy use a switch to separate device
> > port and host port if it supports both. I know this is weird,
> > but many board follows this design, such as Huashan PI and 
> > Milkv Duo S. As for item number, it is just an array of gpios
> > to process one by one, I mark it as two just because Milkv 
> > Duo S use two gpios to controll the USB switch.
> 
> Right, but what I'm looking for is a description for what each GPIO
> does, so that someone can know how the dts should be written.
> 
> > For vbus_det gpio description, There is because the design of 
> > Milk-v Duo S, which connect the switch gpio and VBUS detect 
> > gpio. In this case the OTG function is broken and we can just 
> > change the mode by toggling switch gpios. So I suggest not 
> > defining this property.
> 
> Okay. See my comment on it above.
> 

Thanks, I will add some necessary comments for these two properties.

> Thanks,
> Conor.