[plbossart]

Without this cycle, HDaudio capability parsing fails on some SoundWire-enabled devices.

We are not sure if this sequence is required or not, but it's definitively used by the Skylake driver which successfully parses these capabilities.

Adding as a test to see if this patch has any side effects on existing devices. It would be really nice if people can test on additional devices not covered by CI.

[ranj063]

@plbossart the logic makes sense but how about moving hda_dsp_ctrl_get_caps() to after the controller init in hda_init_caps()?

[plbossart]

@plbossart the logic makes sense but how about moving hda_dsp_ctrl_get_caps() to after the controller init in hda_init_caps()?

I thought about it, but the first thing hda_init_caps() does is to check the bus->ppcap, which is precisely set by hda_dsp_ctrl_get_caps()...

What we could do is split hda_dsp_ctrl_init_chip in two, and do the reset before reading the capabilities, but then we'll have questions on his_dsp_ctrl_misc_clock_gating() - not sure when it's needed or not.

[ranj063]

I thought about it, but the first thing hda_init_caps() does is to check the bus->ppcap, which is precisely set by hda_dsp_ctrl_get_caps()...

@plbossart yes but the check does nothing if it is not there?

[plbossart]

I thought about it, but the first thing hda_init_caps() does is to check the bus->ppcap, which is precisely set by hda_dsp_ctrl_get_caps()...

@plbossart yes but the check does nothing if it is not there?

Indeed, I missed this.
I don't have a clue what "PP capability, will probe DSP later" is supposed to mean and what the expected flow was. That looks like historical comments from @lgirdwood and @keyonjie

[ranj063]

I don't have a clue what "PP capability, will probe DSP later" is supposed to mean and what the expected flow was. That looks like historical comments from @lgirdwood and @keyonjie

@plbossart this comment is stale now I think. In the skylake driver, the bus->ppcap is used to check if the DSP capability exists and then only go on to load the firmware. But in the case of SOF, we dont quite seem to do that. Irrespective of whether this is set or not, we go on and load the firmware anyway.

[bardliao]

This looks good to me. It is the minimal change for now.

[kv2019i]

Looks good!

[plbossart]

closing, will resubmit when we have better information on the hardware needs