[RFC PATCH 2/3] pinctrl: at91: add support for generic pinconf

[Stephen Warren]

On 08/27/2013 12:40 AM, boris brezillon wrote:
> On 27/08/2013 05:57, Stephen Warren wrote:
>> On 08/26/2013 11:17 AM, boris brezillon wrote:
>>> On 26/08/2013 18:53, Stephen Warren wrote:
>>>> On 08/24/2013 03:37 PM, Boris BREZILLON wrote:
>>>>> Add support for generic pin configuration to pinctrl-at91 driver.
>>>>> diff --git
>>>>> a/Documentation/devicetree/bindings/pinctrl/atmel,at91-pinctrl.txt
>>>>> b/Documentation/devicetree/bindings/pinctrl/atmel,at91-pinctrl.txt
>>>>>    Required properties for iomux controller:
>>>>> -- compatible: "atmel,at91rm9200-pinctrl"
>>>>> +- compatible: "atmel,at91rm9200-pinctrl" or
>>>>> "atmel,at91sam9x5-pinctrl".
>>>> You seem to also be adding a second chip name to the list here,
>>>> which is
>>>> more than the patch subject/description imply you're doing...
>>> This is an update of the documentation:
>>> "atmel,at91sam9x5-pinctrl" compatible is already used in the pinctrl
>>> driver but the documention
>>> was not updated.
>>>
>>> But I agree, this should not be part of this series.
>>>
>>>>> +  Add "generic-pinconf" to the compatible string list to use the
>>>>> generic pin
>>>>> +  configuration syntax.
>>>> "generic-pinconf" is too generic of a compatible value for this binding
>>>> to define.
>>>>
>>>> Instead, I think you want to either:
>>>>
>>>> a)
>>>>
>>>> Use compatible="atmel,at91rm9200-pinctrl" for the old binding,
>>>> use compatible="atmel,at91rm9200-pinctrl-generic" for the new binding
>>>>
>>>> or:
>>>>
>>>> b)
>>>>
>>>> Define Boolean property atmel,generic-pinconf (perhaps a better name
>>>> could be chosen?). If it's not present, parse the node assuming the old
>>>> binding. If it is present, parse the node assuming the new binding.
>>>>
>>> Okay.
>>>
>>> I thought this property string could be generic as it may concern other
>>> drivers too
>>> (in order to keep compatibility with old dt ABI and add support the
>>> generic pinconf binding).
>>>
>>> Anyway, I prefer the first proposition.
>>>
>>> pinctrl single driver is already using these names:
>>>
>>> |compatible = "pinctrl-single" for non generic pinconf binding
>>> ||compatible = "pinconf-single" ||for generic pinconf binding|
>>>
>>> So I think we should use something similar:
>>>
>>> |compatible = "atmel,at91xx-pinctrl" for non generic pinconf binding
>>> ||compatible = "|||atmel,at91xx-|pinconf" ||for generic pinconf binding|
>>>
>>> What do you think ?
>> Hmmm. It is a little odd to switch out the compatible value and invent a
>> new binding for the same HW. Isn't it possible to define both sets of
>> properties in the binding, and have drivers look for either?
>>
> 
> Do you mean something like:
> 
> atmel,pins = <xxx>;                /* current dt binding */
> atmel,generic-pins = <yyy>;    /* new dt binding */
> 
> If that's what you had in mind, it will be a little bit tricky to
> handle, because AFAIK the pinconf_ops
> callbacks do not give me any element I could use to deduce the type of
> pinconf (generic or
> native).
> This implies I have to know early during the probe process which kind of
> binding is in use.
> 
> Please tell me if I missed some key points, and this can be easily done.

It's probably most compatible to keep all the existing properties, and
just add new properties for new features.