[BUG] BYT: Random stereo channels swaps happen with sof-byt-nocodec.tplg

[singalsu]

Describe the bug
I have developed a test that plays different frequency sine waves in each channel to PCM0P and captures from PCM0C. The purpose of the test is volume control testing but it detects also basic issues like channel swap. The reported error indicates that the playback, ssp-loopback, or capture channels got swapped.

The topology contains also volume components and mixer in the playback path. Though since a similar topology passes in APL platform the bug in those components is unlikely. But I'll try to replicate with a minimized topology later.

To Reproduce
With PR thesofproject/sof-test#410 (to add the test) and sof PR #3509 (add the volume switch to nocodec topologies) run:
sof-test/test-case/check-volume-levels.sh
Reproduction Rate
For one sub-test (play-capture) about 50% occurrence. Below 2 of 3 sub-tests failed.

Expected behavior
No channels swapped, the test passed as it does in cAVS platforms.

Impact
Annoyance for stereo music, showstopper for applications where channels order is critical.

Environment

1. Branch name and commit hash of the 2 repositories: sof (firmware/topology) and linux (kernel driver).
   • Kernel: {SHA}
   • SOF: {SHA}
2. Name of the topology file
   • Topology: sof-byt-nocodec.tplg
3. Name of the platform(s) on which the bug is observed.
   • Platform: Baytrail

Screenshots or console output

2020-10-14 11:20:57 UTC [INFO] Measuring volume gains
File /tmp/tmp.iat6kH01et.wav:
pass (1/3)
File /tmp/tmp.EIsHmXIzcE.wav:
Channel 1 failed lower gain limit at  0.5 -  0.9s, gain -47.2 dB, min -10.5 dB
Channel 2 failed lower gain limit at  0.5 -  0.9s, gain -27.1 dB, min   9.5 dB
Channel 2 failed lower gain limit at  1.5 -  1.9s, gain -37.1 dB, min  -0.5 dB
fail (2/3)
Note: The test would pass with swapped channels.
File /tmp/tmp.3pEPEAF8DL.wav:
Channel 1 failed lower gain limit at  2.5 -  2.9s, gain -37.2 dB, min  -0.5 dB
Channel 1 failed lower gain limit at  3.5 -  3.9s, gain -37.2 dB, min  -0.5 dB
Channel 2 failed lower gain limit at  3.5 -  3.9s, gain -37.1 dB, min  -0.5 dB
fail (3/3)
Note: The test would pass with swapped channels.
FAIL
2020-10-14 11:21:11 UTC [ERROR] Error: Failed one or more tests in volume levels check.
2020-10-14 11:21:11 UTC [ERROR] Please inspect files: /tmp/tmp.iat6kH01et.wav, /tmp/tmp.EIsHmXIzcE.wav, and /tmp/tmp.3pEPEAF8DL.wav.
2020-10-14 11:21:11 UTC [ERROR] Starting func_exit_handler(), exit status=1, FUNCNAME stack:
2020-10-14 11:21:11 UTC [ERROR]  die()  @  /home/singalsu/git/sof-test/case-lib/lib.sh
2020-10-14 11:21:11 UTC [ERROR]  measure_levels()  @  sof-test/test-case/check-volume-levels.sh:328
2020-10-14 11:21:11 UTC [ERROR]  main()  @  sof-test/test-case/check-volume-levels.sh:84
2020-10-14 11:21:11 UTC [ERROR]  main()  @  sof-test/test-case/check-volume-levels.sh:336
2020-10-14 11:21:11 UTC [INFO] Test Result: FAIL!

Please also include the relevant sections from the firmware log and kernel log in the report (and attach the full logs for complete reference). Kernel log is taken from dmesg and firmware log from sof-logger. See https://thesofproject.github.io/latest/developer_guides/debugability/logger/index.html

[keyonjie]

@singalsu I figured out a fix for this issue here #3535, could you please help try if it works for you also?

[singalsu]

@keyonjie The channels swap issue happens still with BYT nocodec topology (LBM quirk is set). The kernel, firmware etc. are from git today.

Kernel Branch: topic/sof-dev
Kernel Commit: cfa9a863e20a
SOF Branch: master
SOF Commit: e7a339bd4

When I last tested near month ago #3535 it didn't help with LBM, didn't help with Tx-Rx wire and no LBM quirk.

[keyonjie]

Hi @singalsu can you help to try if the PR #3656 can help on BYT for external wired loopback?

For the internal LBM, the channel random swapped in the capture is the expected behavior.

[keyonjie]

Hi @singalsu unfortunately, after several hours debugging, it looks to me that the external wired mode looks quite similar to internal LBM for BYT SSP, the capture can't be stopped separately so the RX FIFO watermark(RFL) keep increasing even we have cleared the RSRE (while SSE is set), so, to me, this issue looks is not easy to be fixed.

[lgirdwood]

@lyakh fyi

[lyakh]

presumably fixed by #3777

[singalsu]

@lyakh I got still error and detected channels swap from check-volume-levels.sh. The same happens with LBM quirk an without with Tx-Rx jumper wire.

[keyonjie]

@singalsu @lyakh looks we have to close this as wont_fix, no advance effort is planned to work on it.

[lyakh]

do we have a list of "known won't-fix issues" to be tracked in case they do get fixed by accident or they have to be re-opened and we can benefit from earlier discussions?

[keyonjie]

yes, we have a label for that, just add it to this issue

[keyonjie]

closing this as wont_fix.