DMIC:ipc error and DSP panic when doing pause / release

[ZhendanYang]

This issue happens when run capture through dmic.

The reproducibility is almost 100%.
How to reproduce:

1. arecord -D hw:0,6 -c 4 -r 48000 -f 32_le /dev/null -vv -i
2. pause / release

Environment:
on UP^2
sof-dev: 08f7ddbcdd7eb4fcbf43cef09272215ef53dba7c
sof: c1123ec
soft: cfff9e0
tplg file name: sof-apl-nocodec.tplg
log:
dmesg.log
memorywindow.log
rmbox.log

[singalsu]

You have an old rmbox binary since it doesn't show dmic trace texts. It would be easier to read if you could update the rmbox originated logs.

[singalsu]

I tried to repeat this but I don't get DSP panic (SOF and SOFT master from today, some days old kernel). Does it happen with first space key press or do you need press pause a lot of times or very rapidly or something?

[ZhendanYang]

@singalsu
I just tested this issue and found out that when I used "-c 1" or "-c 2" in "arecord -D hw:0,6 -r 48000 -f 32_le /dev/null -vv -i" there was no such problem, but when I used "-c 4", once I first pressed space key, the dsp would panic. And I have already update reproduce steps and logs.

[singalsu]

This trace output shows that the topology has configured DMIC for 2ch (num_pdm_active == 1, a pdm controller provides two channels so this should be 2 for 4ch) . It's no wonder that it fails. I think the bug is that 4ch capture for 2ch configuration is not rejected cleanly.

0xf70 [0.085074]	delta [0.000002]	dmic dsc
0xf80 [0.085078]	delta [0.000004]	value 0x0000000000000001
0xf90 [0.085079]	delta [0.000001]	value 0x000000000007a120
0xfa0 [0.085081]	delta [0.000001]	value 0x0000000000493e00
0xfb0 [0.085082]	delta [0.000001]	value 0x000000000000bb80
0xfc0 [0.085083]	delta [0.000001]	value 0x0000000000000000
0xfd0 [0.085084]	delta [0.000001]	value 0x0000000000000020
0xfe0 [0.085086]	delta [0.000001]	value 0x0000000000000020
0xff0 [0.085087]	delta [0.000001]	value 0x0000000000000028
0x1000 [0.085089]	delta [0.000001]	value 0x000000000000003c
0x1010 [0.085090]	delta [0.000001]	value 0x0000000000000001 <---- this value

[keqiaozhang]

Since the DMIC is configured as 2ch in topology and this bug can only reproduce with 4ch, so let's close this bug.