On 4/13/20 9:29 AM, Lu, Brent wrote:
That looks like an error caught by the ALSA conformance tool?
What are the odds that we have a similar issue with the other broadwell drivers which don't have a constraint on the number of channels either on their 'System PCM' dailink?
Thanks -Pierre
Yes. That's why I am sending patch for this old BDW platform...
So far as I know only 'buddy' supports 2/4-channel recording while other BDW Chrome products should support stereo recording only. Therefore, this defect should only be triggered by the ALSA conformance tool.
I am think about implementing the constraint in FE DAI's startup() callback instead of DAI Link's callback. Since the channels_max is 4 for the capture stream, ALSA conformance tool will always test 3-channel recording on any platforms using this driver. Does it make sense to you?
Looking back at previous threads, you indicated that the number of channels supported in propagated from BE to FE, so a similar patch to add 2ch constraints for bwd-rt5677 was dropped ("ASoC: bdw-rt5677: channel constraint support")
Actually I am not sure it was dropped since later you submitted another patch ("ASoC: bdw-rt5677: enable runtime channel merge"), and my feedback was that it seemed simpler to add constraints on all machine drivers.
And now this patch only addresses bdw-rt5650.c but with the initial solution suggested for bdw-rt5677.c
It seems like a generic problem on all Broadwell devices so let's solve with one a single patchset.
Shouldn't we just add the 2ch constraints for broadwell.c and bdw-rt5677.c, and the 2 or 4ch constraint for bdw-rt5650.c? Would this work for you?
Thanks -Pierre