Re: [PATCH 3/3] ASoC: rt5645: add device tree support

[Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>]

On 5/7/15 1:36 PM, Mark Brown wrote:
> On Thu, May 07, 2015 at 01:28:39PM -0500, Pierre-Louis Bossart wrote:
>> On 5/7/15 7:51 AM, Mark Brown wrote:
>
>>> You shouldn't be using of_get_named_gpio() for DT stuff either, use
>>> gpiod_get() which follows the usual pattern of taking a string which is
>>> used to do the lookup with whatever firmware is in use.
>
>> But to Bard's credit the use of_get_named_gpio() is pretty common - i see 18
>> occurrences in soc/codecs alone, others will have the same problem...
>> we talked internally (RafaelW, Darren Hart, LiamG and me) about reaching out
>> to gpio and audio maintainers and aligning some sort of coordinated change
>> to the gpiod framework w/ guidance to developers, did that thread start?
>
> I'm not sure there's any particular need for coordination here, the APIs
> are in place already - the gpiod_ APIs will already transparently look
> up both ACPI and DT (see __gpiod_get_index() for the implementation) and
> new drivers should really be using gpiod_ anyway regardless of trying to
> do both ACPI and DT.

Agree, but there wasn't a clear message provided to the readers of this 
mailing list. A 'should' is a recommendation that provides no real 
incentive to move to the new gpiod framework. The message would be 
clearer if maintainers stated that new contributions using 
of_get_named_gpio() will no longer be merged in the mainline (maybe 
after a specific milestone). That would set the direction for everyone.