[BUG][CML]Can't detect Mic jack of combo headset

[qinwan-hp]

Describe the bug
Plug in headset, the headphone jack can be detected, but the mic jack can seldom to be detected.
If prohibit the codec runtime auto-suspend, the headphone and mic jack can be detected normally.
If change to hda driver, the headphone jack and mic jack can be detected normally.

To Reproduce
Plugin headset

Reproduction Rate
Almost near to 100%

Expected behavior
Both headphone and mic jack can be detected.

Impact
Cannot record sound if plugin headset

Environment

1. Branch name and commit hash of the 2 repositories: sof (firmware/topology) and linux (kernel driver).
   • Kernel: a55aa89 https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git Tag: 5.3-rc6
   • SOF: fcf6c8c https://github.com/thesofproject/sof.git Tag: v.1.4.1
2. Name of the topology file
   • Topology: sof-hda-generic-2ch.tplg
3. Name of the platform(s) on which the bug is observed.
   • Platform: A notebook with CometLake LP, codec ALC3247 and combo headset

codecinfo.txt
dmesg.txt
sof.txt

[juimonen]

@qinwan-hp #1689, this is probably the same...(?)

[lgirdwood]

@qinwan-hp any chance you could try #1689 and see if it fixes the issue for you. I think Its a duplicate.

[qinwan-hp]

@lgirdwood @juimonen I tried this method before and I still have chance without mic detected.
When wait dsp and codec suspended, plugin headset can duplicate this issue sometimes.
There is only unsol event for headphone jack, no unsol event for mic jack.
That's why I disable auto-suspend.
But if I disable auto-suspend, we found the DSP panic when kill/run pulseaudio several times.

I upload the dsp panic kernel log and sof trace.
dsp_firmware_panic.txt
sof_trace.txt
sof_etrace.txt

[lgirdwood]

@qinwan-hp thanks for the info, this now looks like 2 issues :-

1. Failure to detect the mic jack. @kv2019i any comments ?
2. FW is constantly busy - hence we get the panic. @qinwan-hp we need more information here, can you switch on FW IPC debug within the kernel (under SOF debug kconfig). This will tell us what commands pulseaudio is sending to the FW when the Mic is not detected.

[kv2019i]

@lgirdwood wrote:

1. Failure to detect the mic jack. @kv2019i any comments ?

I'm pretty certain this is fixed by #1689 . We have the fix candidate (but no agreement how this should be fixed in upstream yet). @qinwan-hp testing feedback on that pull-request is welcome. FYI @juimonen

[qinwan-hp]

@juimonen Original we have hda with CONFIG_SND_HDA_POWER_SAVE_DEFAULT =0 on our platform, so hda is OK. Change config to CONFIG_SND_HDA_POWER_SAVE_DEFAULT = 2, I did see the same issue happend. But the failure rate is very low and can be igonored.

[qinwan-hp]

2. he kernel (under SOF debug kconfig). This will tell us what commands pulseaudio is sending to the FW when the Mic is not detected.
@lgirdwood It occurs when prohibit codec auto-suspend. Mic can be detected if there is no codec runtime suspend/resume.

[juimonen]

@qinwan-hp so the value 0 for suspend disables the PM altogether, so the codec never goes to sleep -> detect works. I guess with certain values, like you mentioned 2s the detection might still fail occasionally. if pm_autosuspend is set, legacy HDA will set 3s suspend delay, so even longer than 2s.

There is a blacklist in legacy HDA for certain platforms that will never use PM, so they will be always on. So this platform looks/sounds like a candidate for a quirk, if it fails even with 2s delay.

Now with SOF we are currently not calling at all the suspend delay setting, so codec will go to suspend immediately/asap. I think this is a bug and we need to fix it in SOF. What value to set from SOF is a good question (so should we have quirks for platforms which don't work with 2s etc.).

[qinwan-hp]

@juimonen I tried the kernel v5.4.1. The mic jack detction can work and still has failure but hard to reproduce. Our current kernel is v5.3.6.

[juimonen]

@qinwan-hp ok thanks, did you test the new kernel with thesofproject/sof#1689 or without it?

[qinwan-hp]

@juimonen With thesofproject/sof#1689. But I cannot figure out what's the difference between v5.3.6 and v5.4.1 that may lead to this issue. Do you have any comment ?

[juimonen]

@qinwan-hp hmm ok, I can't think anything from top of my head... maybe we have some changes in the power management, but not sure... @ranj063 any idea?

[slawblauciak]

@bkokoszx mind taking a look at it?

[bkokoszx]

@lgirdwood @juimonen
I have a questions, as I'm not kernel expert. Is it correct that we have here in logs things like: skl_hda_dsp_generic, snd_hda_codec_realtek. It seemed to me that we usually blacklist them during kernel building. Why does skylake appear in dmesg logs:
Digital mics found on Skylake+ platform, using SOF driver? Also there appear some topology mismatches:
Topology: ABI 3:12:0 Kernel ABI 3:8:0
warn: topology ABI is more recent than kernel