Re: [PATCH v5 13/13] ASoC: qcom: q6apm: Add support for early buffer mapping on DSP

[Srinivas Kandagatla]

On 3/17/26 6:03 PM, Mark Brown wrote:
> On Mon, Mar 09, 2026 at 06:51:37AM +0000, Srinivas Kandagatla wrote:
> 
>> +static void q6apm_dai_pcm_destruct(struct snd_soc_component *component, struct snd_pcm *pcm)
>> +{
>> +	struct snd_pcm_substream *substream;
>> +	struct snd_soc_pcm_runtime *soc_prtd;
>> +	struct snd_soc_dai *cpu_dai;
>> +	int graph_id;
>> +
>> +	if (pcm->streams[SNDRV_PCM_STREAM_CAPTURE].substream)
>> +		substream = pcm->streams[SNDRV_PCM_STREAM_PLAYBACK].substream;
>> +	else
>> +		substream = pcm->streams[SNDRV_PCM_STREAM_CAPTURE].substream;
> 
> This looks weird, we're destructing the playback substream if the
> capture one is present and the capture stream otherwise?  If that's
> correct the logic is very non-obvious, especially in the case where the
> capture stream isn't presenta and...


Looks like this is a typo in the patch it should be

if (pcm->streams[SNDRV_PCM_STREAM_CAPTURE].substream)
	substream = pcm->streams[SNDRV_PCM_STREAM_CAPTURE].substream;
else
	substream = pcm->streams[SNDRV_PCM_STREAM_PLAYBACK].substream;


will fix this in next spin, along with converting the destruct to pcm_free

--srini

> 
>> +	soc_prtd = snd_soc_substream_to_rtd(substream);
>> +	cpu_dai = snd_soc_rtd_to_cpu(soc_prtd, 0);
> 
> ...we immediately try convert it into a runtime and CPU DAI?