On 2020-03-06 16:49, Pierre-Louis Bossart wrote:
But a more pressing issue is that this breaks platforms using SOF firmware:
[ 28.751756] skl_hda_dsp_generic skl_hda_dsp_generic: ASoC: no sink widget found for dmic01_hifi [ 28.751987] skl_hda_dsp_generic skl_hda_dsp_generic: ASoC: Failed to add route DMIC01 Rx -> direct -> dmic01_hifi
... maybe you can align the topology to mathc so we can reuse the same widget mapping for both SOF and SST firmwares..?
Yeah, I thought this would break userspace and installed topologies and that just confirms it. Adding hard-coded routes is really not recommended.
the alternate solution is what I suggested in another thread "No sound since 5.4 on skl_n88l25_s4567", we could mark this machine driver as having an incomplete topology and remove the topology checks in the core.
I couldn't really test my initial patch but that that Cezary unintentionally broke SOF actually that gives me a tool to test the solution.
Glad to help Pierre ^)^ and thanks for the review Kai.
Guys, we've simply taken long-standing working example such as skl_rt286 or bxt_rt298 and applied the missing diff between skl_hda_dsp's and said machine boards DAPM routes. skl-pcm.c exposes BE: DMIC01 Rx which Intel's SST topologies link against via dmic01_hifi. That has always been the case. No bad intentions, the exact opposite is true: taken old path approach to make sure nothing is broken. Turns out SOF does things differently. Thanks for spotting/ testing this out on your end.
Not a problem to adjust topology on our end, though. In fact, I've already done that on your requested, tested it out and it works just fine. In consequence: - this patch will be dropped from the series - topology patch provided for alsa-ucm-conf will be updated accordingly (dmic01_hifi widget will cease to exist)
Czarek