Re: [PATCH 2/2] ASoC: SDCA: Add port constraints into sdca_asoc_set_constraints()

[Pierre-Louis Bossart <pierre-louis.bossart@linux.dev>]

>>> -	if (entity->type != SDCA_ENTITY_TYPE_IT)
>>> -		return 0;
>>> +	/* FIXME: Currently only a single unshared port is supported */
>>
>> I must admit I have no idea how we'd handle ports that can be connected
>> to multiple terminals. It's allowed in the spec, even between Functions
>> which are then no longer independent, and I was told it was due to
>> hardware restrictions, but still I hope we never have to deal with this...
> 
> lol yeah me too. For now I am mostly just putting these FIXME's
> in at all the places where this is going to become relevant. I
> must confess I haven't reviewed this part of the spec in detail.
> It feels like sharing a couple of ports for a single terminal is
> probably tractable if a little annoying. But yeah, if we got as
> far as sharing a port across functions that seems like a total
> nightmare. We would probably need to do a bit of re-architecting
> and add a port abstraction at the device level, but hopefully
> people are wise enough to not build hardware like that.
I re-read the spec and I think the main usage for the Terminal to ports
mapping is bandwidth management, possibly across multiple lanes. But
that's a 1:N terminal to port mapping, not M:1 or M:N. It's complicated
but not earth-shattering. Also the mapping is supposed to be pre-made in
DisCo tables, not something that requires much intelligence in drivers -
until the tables are wrong but that's a different story.

I didn't see any wording on sharing ports across functions, I am 100%
sure this was talked about in the WG but maybe it didn't make it in the
final 1.0 version. That said it's not described as forbidden either so a
well-intended creative hardware designer will probably make that
"optimization" at some point...