Re: [PATCH v4] ALSA: core: Fix deadlock when shutdown a frozen userspace

[Takashi Iwai]

On Mon, 28 Nov 2022 17:49:20 +0100,
Pierre-Louis Bossart wrote:
> 
> 
> 
> On 11/28/22 07:42, Ricardo Ribalda wrote:
> > During kexec(), the userspace is frozen. Therefore we cannot wait for it
> > to complete.
> > 
> > Avoid running snd_sof_machine_unregister during shutdown.
> > 
> > This fixes:
> > 
> > [   84.943749] Freezing user space processes ... (elapsed 0.111 seconds) done.
> > [  246.784446] INFO: task kexec-lite:5123 blocked for more than 122 seconds.
> > [  246.819035] Call Trace:
> > [  246.821782]  <TASK>
> > [  246.824186]  __schedule+0x5f9/0x1263
> > [  246.828231]  schedule+0x87/0xc5
> > [  246.831779]  snd_card_disconnect_sync+0xb5/0x127
> > ...
> > [  246.889249]  snd_sof_device_shutdown+0xb4/0x150
> > [  246.899317]  pci_device_shutdown+0x37/0x61
> > [  246.903990]  device_shutdown+0x14c/0x1d6
> > [  246.908391]  kernel_kexec+0x45/0xb9
> > 
> > And:
> > 
> > [  246.893222] INFO: task kexec-lite:4891 blocked for more than 122 seconds.
> > [  246.927709] Call Trace:
> > [  246.930461]  <TASK>
> > [  246.932819]  __schedule+0x5f9/0x1263
> > [  246.936855]  ? fsnotify_grab_connector+0x5c/0x70
> > [  246.942045]  schedule+0x87/0xc5
> > [  246.945567]  schedule_timeout+0x49/0xf3
> > [  246.949877]  wait_for_completion+0x86/0xe8
> > [  246.954463]  snd_card_free+0x68/0x89
> > ...
> > [  247.001080]  platform_device_unregister+0x12/0x35
> > 
> > Cc: stable@xxxxxxxxxxxxxxx
> > Fixes: 83bfc7e793b5 ("ASoC: SOF: core: unregister clients and machine drivers in .shutdown")
> > Signed-off-by: Ricardo Ribalda <ribalda@xxxxxxxxxxxx>
> > ---
> > To: Pierre-Louis Bossart <pierre-louis.bossart@xxxxxxxxxxxxxxx>
> > To: Liam Girdwood <lgirdwood@xxxxxxxxx>
> > To: Peter Ujfalusi <peter.ujfalusi@xxxxxxxxxxxxxxx>
> > To: Bard Liao <yung-chuan.liao@xxxxxxxxxxxxxxx>
> > To: Ranjani Sridharan <ranjani.sridharan@xxxxxxxxxxxxxxx>
> > To: Kai Vehmanen <kai.vehmanen@xxxxxxxxxxxxxxx>
> > To: Daniel Baluta <daniel.baluta@xxxxxxx>
> > To: Mark Brown <broonie@xxxxxxxxxx>
> > To: Jaroslav Kysela <perex@xxxxxxxx>
> > To: Takashi Iwai <tiwai@xxxxxxxx>
> > Cc: sound-open-firmware@xxxxxxxxxxxxxxxx
> > Cc: alsa-devel@xxxxxxxxxxxxxxxx
> > Cc: linux-kernel@xxxxxxxxxxxxxxx
> > ---
> > Changes in v4:
> > - Do not call snd_sof_machine_unregister from shutdown.
> > - Link to v3: https://lore.kernel.org/r/20221127-snd-freeze-v3-0-a2eda731ca14@xxxxxxxxxxxx
> > 
> > Changes in v3:
> > - Wrap pm_freezing in a function
> > - Link to v2: https://lore.kernel.org/r/20221127-snd-freeze-v2-0-d8a425ea9663@xxxxxxxxxxxx
> > 
> > Changes in v2:
> > - Only use pm_freezing if CONFIG_FREEZER 
> > - Link to v1: https://lore.kernel.org/r/20221127-snd-freeze-v1-0-57461a366ec2@xxxxxxxxxxxx
> > ---
> >  sound/soc/sof/core.c | 7 ++-----
> >  1 file changed, 2 insertions(+), 5 deletions(-)
> > 
> > diff --git a/sound/soc/sof/core.c b/sound/soc/sof/core.c
> > index 3e6141d03770..9616ba607ded 100644
> > --- a/sound/soc/sof/core.c
> > +++ b/sound/soc/sof/core.c
> > @@ -475,19 +475,16 @@ EXPORT_SYMBOL(snd_sof_device_remove);
> >  int snd_sof_device_shutdown(struct device *dev)
> >  {
> >  	struct snd_sof_dev *sdev = dev_get_drvdata(dev);
> > -	struct snd_sof_pdata *pdata = sdev->pdata;
> >  
> >  	if (IS_ENABLED(CONFIG_SND_SOC_SOF_PROBE_WORK_QUEUE))
> >  		cancel_work_sync(&sdev->probe_work);
> >  
> >  	/*
> > -	 * make sure clients and machine driver(s) are unregistered to force
> > -	 * all userspace devices to be closed prior to the DSP shutdown sequence
> > +	 * make sure clients are unregistered prior to the DSP shutdown
> > +	 * sequence.
> >  	 */
> >  	sof_unregister_clients(sdev);
> >  
> > -	snd_sof_machine_unregister(sdev, pdata);
> > -
> 
> The comment clearly says that we do want all userspace devices to be
> closed. This was added in 83bfc7e793b5 ("ASoC: SOF: core: unregister
> clients and machine drivers in .shutdown") precisely to avoid a platform
> hang if the devices are used after the shutdown completes.

The problem is that it wants the *close* of the user-space programs
unnecessarily.  Basically the shutdown can be seen as a sort of device
hot unplug; i.e. the disconnection of the device files and the cleanup
of device state are the main task.  The difference is that the hot
unplug (unbind) usually follows the sync for the all processes being
closed (so that you can release all resources gracefully), while this
step is skipped for the shutdown (no need for resource-free).


Takashi