Re: [PATCH 23/44] ASoC: codecs: max*: use .auto_selectable_formats

On 2026-08-04 19:37, Mark Brown wrote:
> On Tue, Aug 04, 2026 at 12:42:26AM +0000, Kuninori Morimoto wrote:
>> We can use .auto_selectable_formats. Let's adds it.
> 
>> +static const u64 max9860_selectable_formats =
>> +	SND_SOC_POSSIBLE_DAIFMT_I2S	|
>> +	SND_SOC_POSSIBLE_DAIFMT_LEFT_J	|
>> +	SND_SOC_POSSIBLE_DAIFMT_DSP_A	|
>> +	SND_SOC_POSSIBLE_DAIFMT_DSP_B	|
>> +	SND_SOC_POSSIBLE_DAIFMT_NB_NF	|
>> +	SND_SOC_POSSIBLE_DAIFMT_NB_IF	|
>> +	SND_SOC_POSSIBLE_DAIFMT_IB_NF	|
>> +	SND_SOC_POSSIBLE_DAIFMT_IB_IF;
> 
> This has the same issue the Apple controller had where not all clock
> inversion setups are supported with all formats, I2S style and DSP style
> are different with DSP only supporting normal frame.

Hi!

I agree with this, but I worry that more such limitations are ignored.

E.g. max98090 has this:

	/* Remove 24-bit format support if it is not in right justified mode. */
	if ((fmt & SND_SOC_DAIFMT_FORMAT_MASK) != SND_SOC_DAIFMT_RIGHT_J) {
		substream->runtime->hw.formats = SNDRV_PCM_FMTBIT_S16_LE;
		snd_pcm_hw_constraint_msbits(substream->runtime, 0, 16, 16);
	}

Should RIGHT_J be preferred over I2S/LEFT_J/DSP_A for that codec?

I started looking at max98088 before proceeding to max98090 when the
above problem was spotted. I then stopped looking. It wasn't exactly
hard to find an example.

On the other hand, I'm not familiar with exactly how the auto selection
works...

Cheers,
Peter