Re: [PATCH 2/3] Input: rotary-encoder-counter: add DT bindings

[Kamel Bouhara]

On Tue, Apr 07, 2020 at 11:41:59AM +0200, Maxime Ripard wrote:
> Hi Kamel,
>

Hi Maxime,

> The prefix for device tree bindings is usually dt-bindings:
> $framework: $title
>
> So a title like "dt-bindings: input: Add a counter-based rotary
> encoder binding" would be better.
>

OK, to be fixed then.

> On Mon, Apr 06, 2020 at 05:58:05PM +0200, Kamel Bouhara wrote:
> > Add dt binding for the counter variant of the rotary encoder driver.
> >
> > Signed-off-by: Kamel Bouhara <kamel.bouhara@xxxxxxxxxxx>
> > ---
> >  .../input/rotary-encoder-counter.yaml         | 67 +++++++++++++++++++
> >  1 file changed, 67 insertions(+)
> >  create mode 100644 Documentation/devicetree/bindings/input/rotary-encoder-counter.yaml
> >
> > diff --git a/Documentation/devicetree/bindings/input/rotary-encoder-counter.yaml b/Documentation/devicetree/bindings/input/rotary-encoder-counter.yaml
> > new file mode 100644
> > index 000000000000..a59f7c1faf0c
> > --- /dev/null
> > +++ b/Documentation/devicetree/bindings/input/rotary-encoder-counter.yaml
> > @@ -0,0 +1,67 @@
> > +# SPDX-License-Identifier: GPL-2.0
>
> Bindings are usually used by other OS's, so you should consider
> putting it under a more permissive license, usually that would be GPL2
> and the BSD-2-Clause
>

Well to be honest I just looked into an existing binding and I guess
the wrong one :).

> > +%YAML 1.2
> > +---
> > +$id: http://devicetree.org/schemas/input/rotary-encoder-counter.yaml#
> > +$schema: http://devicetree.org/meta-schemas/core.yaml#
> > +
> > +title: Rotary Encoder Counter
> > +
> > +maintainers:
> > +  - Kamel Bouhara <kamel.bouhara@xxxxxxxxxxx>
> > +
> > +description:
> > +  Registers a Rotary encoder connected through a counter device.
>
> You shouldn't really describe the action here, but more what the
> binding is about. The registration will not depend on the presence of
> the node following that binding, but rather on whether or not the OS
> that uses it has support for it.
>

Then shall it be better with just :
"A rotary encoder device using a generic counter interface." ?

> > +properties:
> > +  compatible:
> > +    const: rotary-encoder-counter
> > +
> > +  counter:
> > +    description: Phandle for the counter device providing rotary position.
>
> This should have a type
>
> > +  linux-axis:
> > +    description: The input subsystem axis to map to this rotary encoder.
> > +    type: boolean
> > +
> > +  qdec-mode:
> > +    description: |
> > +      Quadrature decoder function to set in the counter device.
> > +      3: x1-PHA
> > +      4: x1-PHB
> > +      5: x2-PHA
> > +      6: x2-PHB
> > +      7: x4-PHA and PHB
>
> That range (even though it's a bit odd) should be expressed through an
> enum so that you can check that the values are actually within that
> range.
>

Indeed, that make sens to check it from the binding.

Will fix it in v2.

> > +  steps:
> > +    description: Number of steps in a full turnaround of the encoder.
>
> Muli-line strings should have either quotes around them, or a | or >
> like you did for the description. | will keep the \n, > will make that
> a single string.
>
> This should also have a type
>
> > +      Only relevant for absolute axis.
>
> This should be expressed through a if / then clause, or a dependencies one
>
> >                                         Defaults to 24 which is a typical
> > +      value for such devices.
>
> This should be expressed through a default property.
>

The devil is in the details and yet quite lot of them to fix.

Thanks.

> > +  relative-axis:
> > +    description: Register a relative axis rather than an absolute one.
> > +    type: boolean
> > +
> > +  rollover:
> > +    description: Automatic rollover when the rotary value becomes greater
> > +      than the specified steps or smaller than 0. For absolute axis only.
> > +    type: boolean
>
> Same story than steps for the dependency. Also, what is is the
> behaviour when this property isn't set?
>

OK, if rollover isn't set then the count is unbounded, of course this
shall be described here.

> > +  poll-interval:
> > +    description: Poll interval at which the position is read from the counter
> > +      device (default 500ms).
>
> It should have a type too, and a default property
>
> > +
> > +required:
> > +  - compatible
> > +  - counter
> > +  - qdec-mode
> > +
> > +examples:
> > +  - |
> > +    rotary@0 {
> > +        compatible = "rotary-encoder-counter";
>
> A unit-address (the part after @) only makes sense for a node if
> there's a matching reg property in the node. This will trigger a DTC
> warning, so you should remove the @0
>

Ok I'll fix it then.

Thanks again.

> Maxime



--
Kamel Bouhara, Bootlin
Embedded Linux and kernel engineering
https://bootlin.com