[singalsu]

This patch addresses issue with device where BIOS reported 2ch DMIC
is not connected to default PDM0 pins but PDM1 pins instead.

Currently there is no known indication of such connection in BIOS so
the created topology needs to be manually copied over
sof-hda-generic-2ch.tplg on the impacted devices.

Signed-off-by: Seppo Ingalsuo seppo.ingalsuo@linux.intel.com

[kv2019i]

Looks good. We are looking into runtime logic how we can detect this scenario from NHLT/DMI data and have the kernel load this topology in these cases.

[lgirdwood]

@singalsu @kv2019i I assume we are good for v1.7 as a fix ? i.e. nothing is needed in kernel ?

[kv2019i]

@lgirdwood We do need kernel support as well, but the FW change can go in first and this enables early adopters to fix their systems by manually overriding the topology (like reported did on thesofproject/linux#2460).

[singalsu]

There was a good comment by @plbossart in the linux issue. Though not encountered, to be complete we would be also need sof-hda-generic-1ch-pdm1.tplg. I'll update this PR.

[singalsu]

@lgirdwood @kv2019i The PR update that I just did only added to CMakeLists.txt build of sof-hda-generic-1ch-pdm1.tplg.

[plbossart]

Could we instead have an IPC change to let the kernel do the PDM0 -> PDM1 change?

This feels like a quick and dirty change, but in the IPC I see this:

uint16_t id;		/**< PDM controller ID */

Could we override this value based on a user selection or kernel parameter?

[singalsu]

@plbossart We are seeing this issue now while a kernel based change takes long time to get into disto kernel binaries. Could this topology be released (in 1.7) to have a fix available for impacted devices? The topology could be removed from release when e.g. Ubuntu ships with kernel with this feature.

[plbossart]

@plbossart We are seeing this issue now while a kernel based change takes long time to get into disto kernel binaries. Could this topology be released (in 1.7) to have a fix available for impacted devices? The topology could be removed from release when e.g. Ubuntu ships with kernel with this feature.

You have a point @singalsu, but if we add a quick fix users still need to update their topologies by hand, it's complicated to explain. let's take a couple of days to check if we have an alternative.

[lgirdwood]

@plbossart @singalsu any update here, or shall we wait until after your DMIC call ?

[singalsu]

There will be a kernel option for PDMx forcing later or NHLT blob passing to driver to provide same guidance.

This could be closed or the generated topology build into development topologies. There seems to be a merge conflict as well. Maybe I'll edit this to development and decide a bit later if it should be closed.

[singalsu]

This version builds the topologies for PDM1 connected microphones to development. Let's wait what is the outcome of kernel discussion. The PDMx forcing with kernel option should be quite straightforward. The NHTL blob IPC is maybe something for v1.8.

[lgirdwood]

@singalsu any update ?