Re: [PATCH v3 1/2] dt-bindings: hwmon: amc6821: add fan and PWM output

[Guenter Roeck <linux@roeck-us.net>]

On 3/31/25 09:16, Francesco Dolcini wrote:
> Hello Guenter,
> 
> 
> On Mon, Mar 31, 2025 at 09:01:08AM -0700, Guenter Roeck wrote:
>> On 3/31/25 08:52, Francesco Dolcini wrote:
>>> From: Francesco Dolcini <francesco.dolcini@toradex.com>
>>>
>>> Add properties to describe the fan and the PWM controller output.
>>>
>>> Link: https://www.ti.com/lit/gpn/amc6821
>>> Signed-off-by: Francesco Dolcini <francesco.dolcini@toradex.com>
>>> ---
>>> v3:
>>>    - explicitly describe the fan, use standard PWM and FAN bindings
>>>    - pwm.yaml cannot be referenced, because of the $nodename pattern that is
>>>      enforced there
>>> v2: https://lore.kernel.org/all/20250224180801.128685-2-francesco@dolcini.it/
>>>    - no changes
>>> v1: https://lore.kernel.org/all/20250218165633.106867-2-francesco@dolcini.it/
>>> ---
>>>    .../devicetree/bindings/hwmon/ti,amc6821.yaml      | 14 +++++++++++++-
>>>    1 file changed, 13 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/Documentation/devicetree/bindings/hwmon/ti,amc6821.yaml b/Documentation/devicetree/bindings/hwmon/ti,amc6821.yaml
>>> index 5d33f1a23d03..94aca9c378e6 100644
>>> --- a/Documentation/devicetree/bindings/hwmon/ti,amc6821.yaml
>>> +++ b/Documentation/devicetree/bindings/hwmon/ti,amc6821.yaml
>>> @@ -28,6 +28,13 @@ properties:
>>>      i2c-mux:
>>>        type: object
>>> +  fan:
>>> +    $ref: fan-common.yaml#
>>> +    unevaluatedProperties: false
>>> +
>>> +  "#pwm-cells":
>>> +    const: 3
>>> +
>>>    required:
>>>      - compatible
>>>      - reg
>>> @@ -50,9 +57,14 @@ examples:
>>>            #address-cells = <1>;
>>>            #size-cells = <0>;
>>> -        fan@18 {
>>> +        fan_controller: fan@18 {
>>>                compatible = "ti,amc6821";
>>>                reg = <0x18>;
>>> +            #pwm-cells = <3>;
>>> +
>>> +            fan {
>>> +                pwms = <&fan_controller 0 40000 0>;
>>
>> There is no explanation for the parameters. I guess that the first is the index and the
>> second is the frequency. The index is not used and can be dropped (there is just a single
>> channel), and the frequency only makes sense if it can actually be configured.
> 
> The PWM period can be configured, the HW supports it, so I think is a
> good idea to properly describe the HW now that I am modifying the
> binding and not have the situation in which we need it in 1 month.
> 
> The instance is just one, it's correct, but there is an assumption
> pretty much everywhere that these 3 #cells are
>   - PWM instance
>   - PWM period in ns
>   - PWM flags
> 
> I decided to keep the first cell there, with the PWM instance, for this
> reason, even as you correctly wrote it would not be required here.
> 
> With that said, looking at existing binding, the description is almost
> never present.
> 

That is not a valid argument.

> I would proceed this way for v4:
>   - keep the first cell as PWM instance
>   - add the description
> 

We just had another driver where we dropped the pwm instance number because
it is unnecessary. See commit a25633a00906. I very much prefer to keep it that
way if it is not needed. Everything else is just confusing. It is bad enough
that the sub-node is needed. We should not make it worse.

Thanks,
Guenter