[alsa-devel] [PATCH 1/4] ASoC: rt286: rt286_mic_detect() uses component

[Lars-Peter Clausen]

On 10/26/2016 05:50 AM, Vinod Koul wrote:
> On Mon, Oct 24, 2016 at 11:19:17AM +0200, Lars-Peter Clausen wrote:
>> On 10/24/2016 09:43 AM, Kuninori Morimoto wrote:
>>> Hi
>>>
>>>>>> From: Kuninori Morimoto <kuninori.morimoto.gx at renesas.com>
>>>>>>
>>>>>> Intel broadwell is using card->codec_dev_list, but it will be
>>>>>> replaced to component list soon. Then, it will not be able to use
>>>>>> "codec".
>>>>>
>>>>> Can't it be deduced by a simple container_of()?
>>>>
>>>> There is the snd_soc_component_to_codec() helper for this.
>>>
>>> The points are
>>>  1) We would like to remove "codec" related functions
>>
>> The focus at the moment is to remove all special handling of CODECs from the
>> ASoC core. And trying to do this without being too invasive and causing too
>> many changes all over the place.
>>
>>>  2) We can get "codec" pointer by above function/macro,
>>>     but "codec" function/macro is using codec->component inside.
>>>
>>
>> That's not a problem though.
>>
>> This is slightly unrelated to this series, but these drivers should not be
>> poking the internal data structures of snd_soc_card in the first place. The
>> best way to get the pointer to the CODEC is by using the DAI link init
>> callback. This also means that there is no need to run a lookup each time
>> the CODEC is needed. An alternative is to add a helper function in the core
>> that allows to lookup a component by name.
>>
>> Maybe somebody from the Intel side can look into fixing this. The affected
>> boards are cht_bsw_rt5672 and broadwell, which both access the cards
>> codec_dev_list field.
> 
> This also exists in some customer SKL machines.
> 
> I agree that this may not be best implementation so I can send a patch for
> this.
> 
> As Lars suggested we can use DAI link init callback, but then I dont feel it
> is right to use rtd->codec to get codec pointer, again we will be looking
> into rtd internals. So would make sense to combine two suggestion and add an
> API:
> 
> struct snd_soc_codec *snd_soc_get_codec(struct snd_soc_pcm_runtime *rtd)
> {
> 	return rtd->codec;
> }

I've started working on this series[1] a while ago with the intention of
eventually removing the CODEC field from the rtd struct. Maybe you can use
it as a base.

- Lars

[1]
http://git.metafoo.de/?p=linux-2.6;a=shortlog;h=refs/heads/asoc-remove-pcm-codec