[PATCH 1/2] i2c: mux: add device tree support

[David Daney]

On 04/23/2012 09:13 AM, Stephen Warren wrote:
> On 04/23/2012 05:15 AM, Wolfram Sang wrote:
>> On Tue, Apr 17, 2012 at 12:49:04PM -0600, Stephen Warren wrote:
>>> From: Stephen Warren<swarren at nvidia.com>
>>>
>>> * Define core portions of the DT binding for I2C bus muxes.
>>> * Enhance i2c_add_mux_adapter():
>>> ** Add parameters required for DT support. Update all callers.
>>> ** Set the appropriate adap->dev.of_node for the child bus.
>>> ** Call of_i2c_register_devices() for the child bus.
>>>
>>> Signed-off-by: Stephen Warren<swarren at nvidia.com>
>>
>> David Daney (CCed) posted another variant [1]. Just looking at the
>> patches (and not really using them), I tend to like the approach using
>> <reg>  better. But I am open for discussion, so I'd appreciate your
>> feedback.
>>
>> Regards,
>>
>>     Wolfram
>>
>> [1] http://lkml.org/lkml/2012/4/12/423
>
> Ah, that does look like a reasonable binding.
>

It was arrived at by iterating through several versions with Grant and Rob.

You make at least the third person (after me and Lars-Peter Clausen) 
wanting to use the device tree to configure the I2C muxes.  So at a 
minimum, it shows a need for this.

David Daney.

> I had meant to call out to reviewers the potentially unusual use of
> explicitly named sub-nodes, rather than using the usual reg-based matching.
>
> The main reason I chose named sub-nodes for the busses was so the
> sub-nodes would match the pinctrl named states. However, I think we can
> make the pinctrl numbering match rather than the pinctrl naming instead.
> The only issue is the "idle" state; if we allow it to exist anywhere in
> the pinctrl-names list, it'll make the pinctrl numbering mismatch the
> sub-node numbering. I think we can solve this by forcing the idle state
> to be listed last in pinctrl-names (if it's listed at all). I'll update
> my patches based on that David's patch.