[SDW] DSP panic w/ sof-test suspend-resume on Dell SKU 09C6 (CML single amp) #2875

Activity

1. plbossart commented on Apr 27, 2021

   

   plbossart

   on Apr 27, 2021

   MemberAuthor

   More actions

   This seems to be a kernel regression, if I go back to the sound/upstream-20210426 tag things are fine.
2. plbossart commented on Apr 27, 2021

   

   plbossart

   on Apr 27, 2021

   MemberAuthor

   More actions

   sof-dev-rebase-20210426 is broken, bisecting now
3. plbossart commented on Apr 27, 2021

   

   plbossart

   on Apr 27, 2021

   MemberAuthor

   More actions

   patch to sof-test to speed-up the tests.
   sof-test.diff.txt

   This seems to occur after either a playback/capture or module load/unload. If I do a suspend only things are fine. I don't know which exact sequence causes the issue.
4. plbossart commented on Apr 27, 2021

   

   plbossart

   on Apr 27, 2021

   Last edited by plbossart

   MemberAuthor

   More actions

   git bisect tests on topic/sof-dev-rebase branch

   339397348db61ffe3ee9e2202c66ffae694d1977 is the first bad commit
   commit 339397348db61ffe3ee9e2202c66ffae694d1977
   Author: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
   Date:   Fri Mar 12 10:40:59 2021 -0800
   
       ASoC: SOF: Intel: hda: make sure DAI widget is set up before IPC
       
       With the implementation of the dynamic pipeline feature, widgets
       will only be setup when a PCM is opened during the
       hw_params ioctl. The BE hw_params callback is responsible for
       sending the DAI_CONFIG for the DAI widgets in the DSP.
       With dynamic pipelines, the DAI widgets will need to set up
       first before sending the DAI_CONFIG IPC in the BE hw_params.
       
       Update the BE hw_params/hw_free callbacks for all ALH, HDA and SSP
       DAIs to set up/free the DAI widget before/after DAI_CONFIG IPC.
       
       Signed-off-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
       Reviewed-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>
       Reviewed-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
       Reviewed-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
   
    sound/soc/sof/intel/hda-dai.c | 160 ++++++++++++++++++++++++------------------
    sound/soc/sof/intel/hda.c     | 144 ++++++++++++++++++++++++-------------
    sound/soc/sof/intel/hda.h     |   4 ++
    sound/soc/sof/sof-audio.h     |   1 -
    sound/soc/sof/topology.c      |   3 -
    5 files changed, 194 insertions(+), 118 deletions(-)
   
   git bisect start
   # good: [a49194360d81ee84d39672e2c6ce8ed62cb3d95d] CI: GitHub: set up build test for SOF
   git bisect good a49194360d81ee84d39672e2c6ce8ed62cb3d95d
   # good: [f221de624b0b3956287b707911d37be7443edd15] ipc: debug: Add shared memory heap to memory scan
   git bisect good f221de624b0b3956287b707911d37be7443edd15
   # bad: [f1560a519e633b6b8b6c1455f332750a14ef91c8] ASoC: rt715-sdca: fix clock stop prepare timeout issue
   git bisect bad f1560a519e633b6b8b6c1455f332750a14ef91c8
   # good: [e901af346028f25037331fe29d462c1f6b43041e] ASoC: SOF: restore kcontrols for widget during set up
   git bisect good e901af346028f25037331fe29d462c1f6b43041e
   # bad: [c06aca9b8b7602bb3008b9bc3b6a231e3d471afc] ASoC: SOF: Add support for dynamic pipelines
   git bisect bad c06aca9b8b7602bb3008b9bc3b6a231e3d471afc
   # good: [d4650862e582fd37f4eefd1befd106beb73057f5] ASoC: SOF: Introduce widget use_count
   git bisect good d4650862e582fd37f4eefd1befd106beb73057f5
   # bad: [339397348db61ffe3ee9e2202c66ffae694d1977] ASoC: SOF: Intel: hda: make sure DAI widget is set up before IPC
   git bisect bad 339397348db61ffe3ee9e2202c66ffae694d1977
   # first bad commit: [339397348db61ffe3ee9e2202c66ffae694d1977] ASoC: SOF: Intel: hda: make sure DAI widget is set up before IPC
   

   @bardliao @RanderWang @ranj063 @fredoh9 FYI
5. plbossart commented on Apr 27, 2021

   

   plbossart

   on Apr 27, 2021

   MemberAuthor

   More actions

   Removing PulseAudio removes the IPC error and DSP panic. We're probably missing something in the resume sequence after the dynamic pipeline rework.
6. plbossart commented on Apr 28, 2021

   

   plbossart

   on Apr 28, 2021

   MemberAuthor

   More actions

   Well, the "good news" is that I can reproduce this problem on a TGL Dell device (SKU 0A3E), using SOF v1.7 however since it's using an Intel signed key.

   Same DSP panic with PulseAudio doing weird things. It's not a case of broken hardware but more a broken state machine.

   @ranj063 I think given the correlation with dynamic pipeline changes you may want to setup your device and take a look.
7. plbossart

   added

   bugSomething isn't workingSomething isn't working

   CMLApplies to Comet Lake platformApplies to Comet Lake platform

   IPC error

   P1Blocker bugs or important featuresBlocker bugs or important features

   SDWApplies to SoundWire bus for codec connectionApplies to SoundWire bus for codec connection

   TGLApplies to Tiger Lake platformApplies to Tiger Lake platform

   on Apr 28, 2021
8. plbossart commented on Apr 28, 2021

   

   plbossart

   on Apr 28, 2021

   MemberAuthor

   More actions

   TGL dmesg log for reference
   dmesg.txt
9. bardliao commented on Apr 28, 2021

   

   bardliao

   on Apr 28, 2021

   Collaborator

   More actions

   Well, the "good news" is that I can reproduce this problem on a TGL Dell device (SKU 0A3E)

   I can also see the issue on TGL Dell device (SKU 0A3E)
   dmesg.txt
10. ranj063

    self-assigned this

    on May 5, 2021
11. kv2019i

    closed this as completedin #2888on May 12, 2021