Skip to content

[BUG] [TGL] [DSM] [I2S] No feedback data captured at times after SNDRV_PCM_TRIGGER_STOP/START events on pb stream #3428

Closed
Closed
[BUG] [TGL] [DSM] [I2S] No feedback data captured at times after SNDRV_PCM_TRIGGER_STOP/START events on pb stream#3428
Assignees
keyonjiebkokoszx
Labels
DSMDynamic Speaker ManagementP1Blocker bugs or important featuresTGLApplies to Tiger LakebugSomething isn't working as expected
Milestone
v1.6

Description

@RDharageswari
RDharageswari
opened on Sep 17, 2020

Steps to repro the issue:

  1. Do playback on speakers
    2.Start stop Start stop Start the pb stream [This start and stop refers to SNDRV_PCM_TRIGGER_STOP/START events]
    3.capture feedback data for 8 seconds
    4.Start stop Start stop Start the pb stream
    5.capture feedback data for 8 seconds
    Repeat the above for 5 times, in one of the feedback captured file, the Voltage/current data is missing.
    The result is not consistent across all the captured files

fw branch:
master

Activity

  1. RDharageswari
    added
    bugSomething isn't working as expected
    TGLApplies to Tiger Lake
    on Sep 17, 2020
  2. RDharageswari
    assigned
    bkokoszx
    and
    keyonjie
    on Sep 18, 2020

  3. RDharageswari commented on Sep 18, 2020

    @RDharageswari
    RDharageswari
    on Sep 18, 2020
    Author

    unable to provide more information due to this: #3432

  4. lgirdwood
    added this to the v1.6 milestone on Sep 18, 2020

  5. lgirdwood commented on Sep 18, 2020

    @lgirdwood
    lgirdwood
    on Sep 18, 2020
    Member

    @RDharageswari can you confirm if the exception is also on master ? Is this with real/test DSM ?

  6. RDharageswari commented on Sep 18, 2020

    @RDharageswari
    RDharageswari
    on Sep 18, 2020
    Author

    This issue can be reproduced on master as well. master uses the actual DSMcomponent.i don't have data to prove this issue can be reproduced without maxim DSM as the default smart_amp throws exception with feedback stream itself as in #3432

  7. lgirdwood commented on Sep 18, 2020

    @lgirdwood
    lgirdwood
    on Sep 18, 2020
    Member

    @RDharageswari have you tried running tools/coredumper, this will help shed light on if its related to DSM module or SOF code (it should be able to take us to the line of code) which is doing the unaligned memory read..

  8. mengdonglin
    added
    DSMDynamic Speaker Management
    on Sep 22, 2020
  9. slawblauciak
    added
    P1Blocker bugs or important features
    on Sep 22, 2020

  10. keyonjie commented on Sep 23, 2020

    @keyonjie
    keyonjie
    on Sep 23, 2020
    Contributor

    I can reproduce this with nocodec mode, @bkokoszx I can reproduce this with below steps:

    1. run aplay in one terminal, e.g. 'aplay -Dhw:0,2 -f s16_le -r48000 -c2 /dev/random -vvv -i'
    2. run arecord in another terminal, e.g. 'arecord -Dhw:0,2 -f s16_le -r48000 -c2 tmp.wav -vvv'. you should see the arecord level same with aplay (e.g. aplay terminal is 99, so is the arecord terminal).
    3. stop the arecord, and restart it, check if arecord level still correct.
    4. do step 3 several times, you will see sometimes the arecord level is 0, which means the capture/feedback data are 0s.

  11. bkokoszx commented on Sep 24, 2020

    @bkokoszx
    bkokoszx
    on Sep 24, 2020
    Collaborator

    @keyonjie @RDharageswari @lgirdwood
    I can reproduce issue on my RVP nocodec platform with smart amplifier test component. I can see that copying process looks fine - I mean capture pipeline (even in case when arecord level is 0) processes the right amount of data. I suspect that the data is equal to 0 - I think that maybe there is some more generic problem with SSP. I will debug it.

  12. lgirdwood commented on Sep 24, 2020

    @lgirdwood
    lgirdwood
    on Sep 24, 2020
    Member

    @bkokoszx thanks, I think we will miss v1.6- rc1 but we can hit rc final for this one.

  13. RDharageswari commented on Sep 24, 2020

    @RDharageswari
    RDharageswari
    on Sep 24, 2020
    Author

    Thanks @bkokoszx. Can you also check the PCM_START & PCM_STOP sequence once, to confirm if it is the same issue.
    issue
    This is a VIVI sequence, with maxim DSM module. But you can repro on smart_amp_test component also i believe.

  14. 3 remaining items

  15. keyonjie commented on Sep 27, 2020

    @keyonjie
    keyonjie
    on Sep 27, 2020
    Contributor

    Thanks for debugging and verification, @bkokoszx and @RDharageswari .

    I create a PR #3476 to fix this, please help to review.

  16. RDharageswari
    added a commit that references this issue on Oct 2, 2020

    ssp: fix the RX FIFO flushing logic

    14a4406
  17. lgirdwood
    added a commit that references this issue on Oct 2, 2020

    ssp: fix the RX FIFO flushing logic

    242be34
  18. bkokoszx
    added a commit that references this issue on Oct 2, 2020

    ssp: fix the RX FIFO flushing logic

    ceb1ad2
  19. lgirdwood
    added a commit that references this issue on Oct 6, 2020

    ssp: fix the RX FIFO flushing logic

    65d0a93
  20. RDharageswari
    added 2 commits that reference this issue on Oct 6, 2020

    ssp: fix the RX FIFO flushing logic

    e93ae3c

    ssp: fix the RX FIFO flushing logic

    93e3ef6

  21. keyonjie commented on Oct 12, 2020

    @keyonjie
    keyonjie
    on Oct 12, 2020
    Contributor

    @RDharageswari shall we close this now?

  22. RDharageswari
    closed this as completedon Oct 19, 2020
  23. RDharageswari
    added a commit that references this issue on Oct 23, 2020

    ssp: fix the RX FIFO flushing logic

    d7ea3b4
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

DSMDynamic Speaker ManagementP1Blocker bugs or important featuresTGLApplies to Tiger LakebugSomething isn't working as expected

Type

No type

Projects

No projects

    Milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions