On Fri, 21 Aug 2015, Pierre-Louis Bossart wrote:

On 8/21/15 2:11 AM, Ricard Wanderlof wrote:

...

On Thu, 20 Aug 2015, Pierre-Louis Bossart wrote:

...

...

/** * snd_soc_runtime_activate() - Increment active count for PCM runtime components * @rtd: ASoC PCM runtime that is activated @@ -371,6 +389,20 @@ static void soc_pcm_init_runtime_hw(struct snd_pcm_substream *substream)

/* first calculate min/max only for CODECs in the DAI link */
for (i = 0; i < rtd->num_codecs; i++) {

• /*
  

•  * Skip CODECs which don't support the current stream type.
  

•  * Otherwise, since the rate, channel, and format values will
  

•  * zero in that case, we would have no usable settings left,
  

•  * causing the resulting setup to fail.
  

•  * At least one CODEC should match, otherwise we should have
  

•  * bailed out on a higher level, since there would be no
  

•  * CODEC to support the transfer direction in that case.
  

•  */
  

• if (!snd_soc_dai_stream_valid(rtd->codec_dais[i],
  

• 			      substream->stream))
  

Maybe I misunderstood but shouldn't there be some sort of verification that the codecs can use the same number of slots between playback and capture if they share the same LRCLK/FS? e.g. it's not uncommon to have 4 mic capture and 2 ch playback. If the capture and playback is handled by different chips you'd still need to maintain some level of consistency.

You're probably right, although it may be that that sort of thing is handled at a higher level, e.g. the machine driver in such a setup would configure both codecs to use TDM before we even get to this function, I don't know.

I think we'd have to take care of that situation if and when it arises. The setup I have here doesn't allow me to test it.

ok, sounds fine. maybe you could add this to the comments or commit message then so that this assumption is known.

Ok, good point. Will do.

/Ricard