Re: [PATCH v2 6/9] ASoC: tegra: add Tegra186 based DSPK driver

[Sameer Pujar]

On 2/10/2020 5:52 PM, Jon Hunter wrote:
> On 10/02/2020 11:15, Sameer Pujar wrote:
> > On 2/7/2020 11:52 PM, Dmitry Osipenko wrote:
> > > External email: Use caution opening links or attachments
> > >
> > >
> > > 07.02.2020 14:26, Sameer Pujar ÐÐÑÐÑ:
> > > > On 2/6/2020 10:45 PM, Dmitry Osipenko wrote:
> > > > > External email: Use caution opening links or attachments
> > > > >
> > > > >
> > > > > 30.01.2020 13:33, Sameer Pujar ÐÐÑÐÑ:
> > > > > > +static const struct dev_pm_ops tegra186_dspk_pm_ops = {
> > > > > > +ÂÂÂÂ SET_RUNTIME_PM_OPS(tegra186_dspk_runtime_suspend,
> > > > > > +ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ tegra186_dspk_runtime_resume, NULL)
> > > > > > +ÂÂÂÂ SET_LATE_SYSTEM_SLEEP_PM_OPS(pm_runtime_force_suspend,
> > > > > > +ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ pm_runtime_force_resume)
> > > > > > +};
> > > > > Could you please explain why drivers need the "late" system sleep?
> > > > It was done to ensure core drivers are suspended first and defer the
> > > > codec driver suspend
> > > Suspend order is opposite to the drivers registration order. If there is
> > > no real problem with that, then you should use the default suspend
> > > level. Please don't try to fix a non-existent problems.
> > No. This was done specifically to allow sound core to first stop any
> > ongoing audio activity during normal suspend and ensure a safe suspend
> > of AHUB devices by doing a LATE suspend.
> What Dmitry is saying is that if the DSPK driver is registered after the
> sound core then we will not need to suspend in the late phase. The DSPK
> device should only be registered once the sound core is loaded, because
> otherwise we should fail to register it with the sound core. So I don't
> think we need this to be late afterall.

I was originally thinking if DMA is the main reason for using LATE 
suspend for audio drivers as well. I did a small sanity check and 
appears normal suspend is fine. I will update the drivers with normal 
suspend. If we come across any issue later, it can be addressed 
separately. Thanks Dmitry and Jon.
> Jon