[BUG][BYT][BDW] There are host position update warnings always

[keyonjie]

There are many host position update warnings from sof-logger according to recent daily test on Intel's BDW and BYT platforms:

[   291465955.814017] (          62.499996) c0 ll-schedule        ./schedule/ll_schedule.c:355  INFO num_tasks 1 total_num_tasks 1
[   292697318.629670] (     1231362.875000) c0 pipe         1.10  ....../pipeline-stream.c:473  WARN pipeline_get_timestamp(): DAI position update failed
[   294113987.635877] (     1416669.000000) c0 pipe         1.10  ....../pipeline-stream.c:473  WARN pipeline_get_timestamp(): DAI position update failed
[   295530656.277500] (     1416668.625000) c0 pipe         1.10  ....../pipeline-stream.c:473  WARN pipeline_get_timestamp(): DAI position update failed
[   296963985.022628] (     1433328.750000) c0 pipe         1.10  ....../pipeline-stream.c:473  WARN pipeline_get_timestamp(): DAI position update failed
[   298380654.133001] (     1416669.125000) c0 pipe         1.10  ....../pipeline-stream.c:473  WARN pipeline_get_timestamp(): DAI position update failed
[   299797322.774625] (     1416668.625000) c0 pipe         1.10  ....../pipeline-stream.c:473  WARN pipeline_get_timestamp(): DAI position update failed
[   301230651.623919] (     1433328.875000) c0 pipe         1.10  ....../pipeline-stream.c:473  WARN pipeline_get_timestamp(): DAI position update failed
[   302647320.525959] (     1416668.875000) c0 pipe         1.10  ....../pipeline-stream.c:473  WARN pipeline_get_timestamp(): DAI position update failed
[   304063989.271749] (     1416668.750000) c0 pipe         1.10  ....../pipeline-stream.c:473  WARN pipeline_get_timestamp(): DAI position update failed
[   305497318.121044] (     1433328.875000) c0 pipe         1.10  ....../pipeline-stream.c:473  WARN pipeline_get_timestamp(): DAI position update failed
[   306913987.127251] (     1416669.000000) c0 pipe         1.10  ....../pipeline-stream.c:473  WARN pipeline_get_timestamp(): DAI position update failed
[   308330655.664707] (     1416668.500000) c0 pipe         1.10  ....../pipeline-stream.c:473  WARN pipeline_get_timestamp(): DAI position update failed

To Reproduce
Just run any playback or capture streams on BDW/BYT, and check sof-logger logs.

Reproduction Rate
It happens in about every 500ms~2s when the stream is active

Expected behavior
We didn't have this warning previous.

Impact
Not sure if it will impact any audio quality, but this warning should be taken care anyway.

Environment

1. Branch name and commit hash of the 2 repositories: sof (firmware/topology) and linux (kernel driver).
   • Kernel: topic/sof-dev Commit: a9c617ef
   • SOF: main Commit: cd48b89
2. Name of the topology file
   • Topology: sof-bdw-rt286.tplg on BDW and sof-byt-nocodec.tplg on BYT.
3. Name of the platform(s) on which the bug is observed.
   • Platform: BDW/BYT.

[miRoox]

After bisect, I found this warning is first introduced by b82e0ef, but it seems to bring the underlying issue to the surface, not the root cause.

[marc-hb]

The DMA on BYT and BDW has always had a fairly different set of issues: #4333

https://sof-ci.01.org/sofpr/PR5261/build11791/devicetest/?model=BDW_WSB_RT286&testcase=check-alsabat-headset-playback

[plbossart]

@keyonjie Which topologies trigger this warning?

IIRC the BYT topologies use the 'host' and 'dai' pipelines while BDW is more 'linear' with a single pipeline. This could be quite different in terms of pipeline walks.

[keyonjie]

@keyonjie Which topologies trigger this warning?

sof-bdw-rt286.tplg and sof-byt-nocodec.tplg.

[lgirdwood]

@keyonjie are you able to improve this error ? IIRC we see this when a singe component returns an unexpected return code in the pipeline position walk. We could add a

trace_err("comp X has unexpected state Y when getting position");

[lgirdwood]

@keyonjie did we send a fix for this, IIRC there was a recent update that addressed this warning ?

[plbossart]

still happening, see https://sof-ci.01.org/linuxpr/PR3312/build6800/devicetest/?model=BYT_MB_NOCODEC&testcase=check-kmod-load-unload-after-playback

[marc-hb]

I grepped through test results and I confirm this started to happen systematically on BYT and BDW on Oct 15 which is the exact date b82e0ef was merged.

P1 because it affects all test results.