[BUG] Glitch in wav file with DMIC interface FIFO B

[kkarask]

Describe the bug
Glitches was found in test 05_04_TestDmicFifoBSimpleStreamOperator48000hz32b32b4ch

Topology

 +----------+      +------------------+
    |    OS    |----->|   AUDIO DEVICE   |-----------------+
    +----------+      +------------------+                 |
                                                           |
                  +---------------------------+            |
    +----------+  | +------+   +------------+ |  +------+  |
    | Host cap |<---|Buffer|<--| Dai Fifo_B |<---| DMIC |<-+
    +----------+  | +------+   +------------+ |  +------+
                  +---------------------------+

To Reproduce
Run test 05_04_TestDmicFifoBSimpleStreamOperator48000hz32b32b4ch with Diagnostic driver
with parameters:
--playback_iterations=75

Reproduction Rate
1/75

Environment

1. Firmware branch name and commit

   • SOF: Branch: main, Hash: d6e29ce

2. Name of the platform(s) on which the bug is observed.

   • Platform: APL RVP

Screenshots or console output

Logs:
05_04_TestDmicFifoBSimpleStreamOperator48000hz32b32b4ch.zip

[lgirdwood]

@kkarask I assume this works with FIFO A ? Can you stress ?

[kkarask]

@lgirdwood yes, it works with FIFO A. I run stress tests and reproduction rate with failed FIFO B is 5/200.

[lgirdwood]

@singalsu any difference here between A (working) and B (glitch 5/200) ?

[singalsu]

@lgirdwood The A and B paths in HW are identical except DMA configuration.

@kkarask If the configuration of A and B is swapped in topology, does the issue swap too?

[kkarask]

@singalsu In that test we have configured only FIFO B and there are glitches.
We have another test where only FIFO A is configured and it works well.
We have also a test where FIFO A and FIFO B are configured and the problem doesn't exist.

[singalsu]

I was thinking a case where FIFO A is configured to 16 kHz, does the FIFO A make then glitches?

[keyonjie]

@lgirdwood we have run out the 1.8 time, so this have to be pushed out to 1.9.

[kkarask]

@singalsu In case where FIFO A is configured as 16 kHz problem doesn't occur.

[wszypelt]

unsupported platform
on TGL issue does not occur
Tested FW hash: 4943aaa