It's been a while since this thread started, and I still don't quite get the concepts or logic.
First, I don't understand what the problem with "aux" devices is. All the SoundWire stuff is based on the concept of DAI, so I guess I am
That is the actual problem! Some aux devices does not have dais.
Usually aux devices are used for things like analog amplifiers that clearly don't have a digital interface, thus it really makes no sense to have a DAI link connecting them. So I guess my question here would be what is the thinking on making the "smart amplifier" dailess? It feels like having a CODEC to CODEC DAI between the CODEC and the AMP would be a more obvious way to connect the two devices and would presumably avoid the issues being discussed around the patch.
Ah, now I get it, I missed the point about not having DAIs for the amplifier.
I will second Charles' point, the code you have in the machine driver at [1] gets a SoundWire stream context from the SLIMbus codec dai. It's a bit odd to mix layers this way.
And if I look at the code below, taken from [1], if you have more than one codec, then your code looks problematic: data->sruntime would be overriden and you'd use the info from the last codec dai on the dailink...
case SLIMBUS_0_RX...SLIMBUS_6_TX: for (i = 0 ; i < dai_link->num_codecs; i++) { [snip] data->sruntime[cpu_dai->id] = snd_soc_dai_get_sdw_stream(rtd->codec_dais[i], 0); << same destination for multiple codec_dais...
If you keep the amp dai-less, then the stream concept should be somehow allocated on the master (or machine) side and passed to the codec dais that are associated to the same 'stream'.
[1] https://git.linaro.org/people/srinivas.kandagatla/linux.git/tree/sound/soc/q...