mbox series

[SRU,H/G/OEM-5.10,0/6] power off stress test will hang on the TGL machines

Message ID 20210318034922.7546-1-hui.wang@canonical.com
Headers show
Series power off stress test will hang on the TGL machines | expand

Message

Hui Wang March 18, 2021, 3:49 a.m. UTC
BugLink: https://bugs.launchpad.net/bugs/1919930

Intel suggested that we do 2 actions to fix this problem, the 1st is
merging 5 kernel patches, this only applies to H and OEM-5.10 since
there is no tgl.c in the groovy kernel yet. the 2nd is change a kernel
config, this change applies to H, G and OEM-5.10.

https://github.com/thesofproject/linux/issues/2781

[Impact]
When we run poweroff/on stress test on some lenovo TGL laptop, the
system will randomly hang, and when this issue happens, the dmesg
shows the sof audio driver fails.

[Fix]
Intel recommend that we backport 5 kernel patches and change a
kernel config.

[Test]
After applying the changes, and test on TGL/cml/whl machines,
the audio function works as good as before, and the poweroff stress
test didn't hang anymore.


[Where problems could occur]
The kernel patches probably could introduce issues when system
powre off or reboot on TGL machines, but this possibility is low
since we have tested these patches on different TGL machines.

the kernel option change could introduce power consumption
regression, but it only affects power saving and package_cstate values
when any capture stream is active, while no impact if all capture
streams are inactive. that is to say, in theory it will not impact
the power consumption in short idle or long idle. And I checked the
system cound enter package_c10 after this change.


Hui Wang (1):
  UBUNTU: [Config] set SND_SOC_SOF_HDA_ALWAYS_ENABLE_DMI_L1 to n

Keyon Jie (4):
  ASoC: SOF: add .shutdown() callback to snd_sof_dsp_ops
  ASoC: SOF: add snd_sof_device_shutdown() helper for shutdown
  ASoC: SOF: sof-pci-dev: add .shutdown() callback
  ASoC: SOF: Intel: tgl: do thorough remove at .shutdown() callback

Libin Yang (1):
  ASoC: SOF: Intel: TGL: set shutdown callback to hda_dsp_shutdown

 debian.master/config/annotations          | 3 ++-
 debian.master/config/config.common.ubuntu | 2 +-
 sound/soc/sof/core.c                      | 8 ++++++++
 sound/soc/sof/intel/hda-dsp.c             | 6 ++++++
 sound/soc/sof/intel/hda.h                 | 1 +
 sound/soc/sof/intel/tgl.c                 | 3 ++-
 sound/soc/sof/ops.h                       | 8 ++++++++
 sound/soc/sof/sof-pci-dev.c               | 6 ++++++
 sound/soc/sof/sof-priv.h                  | 4 +++-
 9 files changed, 37 insertions(+), 4 deletions(-)

Comments

Hui Wang March 18, 2021, 4:36 a.m. UTC | #1
The 0005-xxx.patch is not in upstream kernel yet, forgot to add ubuntu 
sauce. Please ignore this thread and I will send a v2 version.

Thanks.

On 3/18/21 11:49 AM, Hui Wang wrote:
> BugLink: https://bugs.launchpad.net/bugs/1919930
>
> Intel suggested that we do 2 actions to fix this problem, the 1st is
> merging 5 kernel patches, this only applies to H and OEM-5.10 since
> there is no tgl.c in the groovy kernel yet. the 2nd is change a kernel
> config, this change applies to H, G and OEM-5.10.
>
> https://github.com/thesofproject/linux/issues/2781
>
> [Impact]
> When we run poweroff/on stress test on some lenovo TGL laptop, the
> system will randomly hang, and when this issue happens, the dmesg
> shows the sof audio driver fails.
>
> [Fix]
> Intel recommend that we backport 5 kernel patches and change a
> kernel config.
>
> [Test]
> After applying the changes, and test on TGL/cml/whl machines,
> the audio function works as good as before, and the poweroff stress
> test didn't hang anymore.
>
>
> [Where problems could occur]
> The kernel patches probably could introduce issues when system
> powre off or reboot on TGL machines, but this possibility is low
> since we have tested these patches on different TGL machines.
>
> the kernel option change could introduce power consumption
> regression, but it only affects power saving and package_cstate values
> when any capture stream is active, while no impact if all capture
> streams are inactive. that is to say, in theory it will not impact
> the power consumption in short idle or long idle. And I checked the
> system cound enter package_c10 after this change.
>
>
> Hui Wang (1):
>    UBUNTU: [Config] set SND_SOC_SOF_HDA_ALWAYS_ENABLE_DMI_L1 to n
>
> Keyon Jie (4):
>    ASoC: SOF: add .shutdown() callback to snd_sof_dsp_ops
>    ASoC: SOF: add snd_sof_device_shutdown() helper for shutdown
>    ASoC: SOF: sof-pci-dev: add .shutdown() callback
>    ASoC: SOF: Intel: tgl: do thorough remove at .shutdown() callback
>
> Libin Yang (1):
>    ASoC: SOF: Intel: TGL: set shutdown callback to hda_dsp_shutdown
>
>   debian.master/config/annotations          | 3 ++-
>   debian.master/config/config.common.ubuntu | 2 +-
>   sound/soc/sof/core.c                      | 8 ++++++++
>   sound/soc/sof/intel/hda-dsp.c             | 6 ++++++
>   sound/soc/sof/intel/hda.h                 | 1 +
>   sound/soc/sof/intel/tgl.c                 | 3 ++-
>   sound/soc/sof/ops.h                       | 8 ++++++++
>   sound/soc/sof/sof-pci-dev.c               | 6 ++++++
>   sound/soc/sof/sof-priv.h                  | 4 +++-
>   9 files changed, 37 insertions(+), 4 deletions(-)
>