Re: [PATCH v3 1/2] dt-bindings: i2c-mux: Add property for settle time

[Rob Herring <robh@kernel.org>]

On Mon, Nov 01, 2021 at 10:32:01PM +0100, Horatiu Vultur wrote:
> The 11/01/2021 15:32, Peter Rosin wrote:
> 
> Hi Peter,
> 
> > 
> > On 2021-11-01 13:25, Horatiu Vultur wrote:
> > > Some HW requires some time for the signals to settle after the muxing is
> > > changed. Allow this time to be specified in device tree.
> > >
> > > Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
> > > ---
> > >  Documentation/devicetree/bindings/i2c/i2c-mux.yaml | 6 ++++++
> > >  1 file changed, 6 insertions(+)
> > >
> > > diff --git a/Documentation/devicetree/bindings/i2c/i2c-mux.yaml b/Documentation/devicetree/bindings/i2c/i2c-mux.yaml
> > > index 24cac36037f5..4628ff6340c1 100644
> > > --- a/Documentation/devicetree/bindings/i2c/i2c-mux.yaml
> > > +++ b/Documentation/devicetree/bindings/i2c/i2c-mux.yaml
> > > @@ -29,6 +29,12 @@ properties:
> > >    '#size-cells':
> > >      const: 0
> > >
> > > +  settle-time-us:
> > > +    default: 0
> > > +    description:
> > > +      The time required for the signals to settle. Currently only the
> > > +      i2c-mux-gpmux driver supports this optional binding.
> > 
> > The information about how i2c-mux-gpmux is special is bound to go stale,
> > and I don't think we should mention such specific details in the binding.
> > What I meant was a generic warnings about optional bindings perhaps not
> > being supported by all drivers, along the lines of this from i2c.txt:
> > 
> > "These properties may not be supported by all drivers. However, if a driver
> >  wants to support one of the below features, it should adapt these bindings."
> > 
> > However, I now notice that this sentence makes no sense. It looks like it
> > should be s/adapt/adopt/.
> > 
> > And, in the i2c-mux.yaml case it can simply say "Optional properties"
> > instead of "These properites" (which refers to a subset of properties
> > immediately below the text) since with a yaml binding it is always
> > clear which properties are optional and which are required. Lastly, I
> > guess this warning belongs in the description.
> > 
> > > +
> > >  patternProperties:
> > >    '^i2c@[0-9a-f]+$':
> > >      $ref: /schemas/i2c/i2c-controller.yaml
> > >
> > 
> > Since this is the first optional property, you now need to specify what
> > properties are required, which is everything but settle-time-us. If you
> > don't, all properties are required. Which is not what we want...
> > 
> > Something like this should do it, I think:
> > 
> > required:
> >   - compatible
> >   - '#address-cells'
> >   - '#size-cells'
> 
> Thanks for a detail explanation but I am still struggling with these
> bindings. Were you thinking to have something like this?
> 
> ---
> diff --git a/Documentation/devicetree/bindings/i2c/i2c-mux.yaml b/Documentation/devicetree/bindings/i2c/i2c-mux.yaml
> index 24cac36037f5..c9fde1bb0fea 100644
> --- a/Documentation/devicetree/bindings/i2c/i2c-mux.yaml
> +++ b/Documentation/devicetree/bindings/i2c/i2c-mux.yaml
> @@ -19,6 +19,9 @@ description: |+
>    populating the i2c child busses.  If an 'i2c-mux' subnode is present, only
>    subnodes of this will be considered as i2c child busses.
> 
> +  Optional properties may not be supported by all drivers. However, if a driver
> +  wants to support one of the below features, it should adopt these bindings.
> +
>  properties:
>    $nodename:
>      pattern: '^(i2c-?)?mux'
> @@ -29,6 +32,11 @@ properties:
>    '#size-cells':
>      const: 0
> 
> +  settle-time-us:
> +    default: 0
> +    description:
> +      The time required for the signals to settle.
> +
>  patternProperties:
>    '^i2c@[0-9a-f]+$':
>      $ref: /schemas/i2c/i2c-controller.yaml
> @@ -41,6 +49,11 @@ patternProperties:
> 
>  additionalProperties: true
> 
> +required:
> +  - compatible

compatible should not be required here.

> +  - '#address-cells'
> +  - '#size-cells'
> +
>  examples:
>    - |
>      /*
> ---
> 
> If I have this then my problem is with the required properties because then I
> start to get new warnings once I run:
> 
> make ARCH=arm CROSS_COMPILE=arm-linux- dtbs_check DT_SCHEMA_FILES=Documentation/devicetree/bindings/i2c/i2c-mux.yaml
> 
> For example, one of new the warnings is this:
> 
> /home/hvultur/linux/arch/arm/boot/dts/am335x-icev2.dt.yaml: mux-mii-hog: 'compatible' is a required property
> 	From schema: /home/hvultur/linux/Documentation/devicetree/bindings/i2c/i2c-mux.yaml
> /home/hvultur/linux/arch/arm/boot/dts/am335x-icev2.dt.yaml: mux-mii-hog: '#address-cells' is a required property
> 	From schema: /home/hvultur/linux/Documentation/devicetree/bindings/i2c/i2c-mux.yaml
> /home/hvultur/linux/arch/arm/boot/dts/am335x-icev2.dt.yaml: mux-mii-hog: '#size-cells' is a required property
> 	From schema: /home/hvultur/linux/Documentation/devicetree/bindings/i2c/i2c-mux.yaml

This is because of the $nodename pattern being pretty lax and matches 
on mux-mii-hog by mistake. We have 2 options. Change the nodename 
pattern to '^(i2c-?)?mux(@.*)?$' or add 'select: false'. The former 
would still match on 'mux' or 'mux@.*' which might still have problems. 
For the latter, we just need to make sure all the i2c-mux schemas have a 
$ref to this schema. Also, with that change we'd stop checking 'i2c-mux' 
nodes that don't yet have a specific schema. That said, I do lean toward 
the latter option.

Rob