Re: [PATCH v3 4/4] ASoC: tegra: Specify components string for Nexus 7

[Dmitry Osipenko]

25.05.2021 19:24, Jaroslav Kysela пишет:
> On 24. 05. 21 16:00, Dmitry Osipenko wrote:
>> 24.05.2021 13:54, Jaroslav Kysela пишет:
>>> Dne 24. 05. 21 v 1:44 Dmitry Osipenko napsal(a):
>>>> Specify components string for Nexus 7 using the Intel BayTrail components
>>>> format. This may allow us to create a more generic UCM for RT5640 codec.
>>>>
>>>> Signed-off-by: Dmitry Osipenko <digetx@xxxxxxxxx>
>>>> ---
>>>>  sound/soc/tegra/tegra_asoc_machine.c | 19 +++++++++++++++++++
>>>>  1 file changed, 19 insertions(+)
>>>>
>>>> diff --git a/sound/soc/tegra/tegra_asoc_machine.c b/sound/soc/tegra/tegra_asoc_machine.c
>>>> index a81f2ebfc00c..87e0a47040a5 100644
>>>> --- a/sound/soc/tegra/tegra_asoc_machine.c
>>>> +++ b/sound/soc/tegra/tegra_asoc_machine.c
>>>> @@ -671,6 +671,24 @@ static const struct tegra_asoc_data tegra_rt5640_data = {
>>>>  	.add_hp_jack = true,
>>>>  };
>>>>  
>>>> +/*
>>>> + * Speaker: Connected to SPO L/R P/N pins, stereo.
>>>> + * Internal Microphone: Digital, connected to DMIC1_DAT IN2P/N pins.
>>>> + * Headphones: Connected to HPOL/R pins.
>>>> + * Headset Microphone: Unconnected.
>>>> + *
>>>> + * IF2_DAC/ADC are unpopulated.
>>>> + */
>>>> +static const struct tegra_asoc_data tegra_rt5640_grouper_data = {
>>>> +	.components = "codec:rt5640 cfg-spk:2 cfg-mic:dmic1 aif:1",
>>>
>>> Perhaps, it may be better to handle this string via the DT config (create new
>>> function like snd_soc_of_parse_card_name()).
>>>
>>> The "codec:rt5640" should be set in the driver (it's fixed), but the
>>> "cfg-spk:2 cfg-mic:dmic1 aif:1" part is suitable for the DT config.
>>
>> Not sure that this is possible. The DT is an ABI, while components
>> string has no specification. Any changes to the components string will
>> require DT update.
>>
>> I think it's much more feasible to have a database of components within
>> kernel driver, like Intel sound driver does it for ACPI devices.
> 
> There is no other possibility for ACPI. And while the components string is
> part of the ALSA ABI, it should not be changed in an incompatible way. Only
> new information should be added. The DT solution seems more flexible to me.
> For example the stereo/mono speaker or the AIF number can be set without the
> driver modification for rt5640.

Everything that goes into device-tree needs a firm specification. We
can't add it into device-tree since there is no specification for the
components string other than something that Intel driver made up for its
own needs. So either we could add the components string to the kernel
driver right now or not add it.

I think the best option would be to drop this patch for now.