Re: [PATCH 01/12] ALSA: hda: intel: Reduce CONFIG_PM dependencies

[Andy Shevchenko <andriy.shevchenko@xxxxxxxxx>]

On Thu, May 09, 2024 at 09:05:58AM +0200, Takashi Iwai wrote:
> On Wed, 08 May 2024 19:45:42 +0200,
> Andy Shevchenko wrote:
> > On Mon, May 06, 2024 at 06:13:44PM +0200, Takashi Iwai wrote:

> > > snd-hda-intel contains lots of CONFIG_PM dependent code although
> > > CONFIG_PM is almost mandatory nowadays, and it makes the code
> > > unnecessarily complex.
> > > 
> > > Let's reduce the dependencies of CONFIG_PM in snd-hda-intel driver
> > > code.  I left a few module options to be dependent on CONFIG_PM (which
> > > are visible to users), but other places are either enabled or
> > > optimized by compiler automatically.

...

> > > +static int __maybe_unused azx_resume(struct device *dev)
> > 
> > __maybe)unused is discouraged nowadays.
> > We have new PM macros (w/o SET_ prefix) along with pm_ptr() / pm_sleep_ptr()
> > macros. They are preferred. In complicated cases the PTR_IF() can be used
> > directly.
> 
> Yeah, it was a dilemma there.  There seems no standard macro to use
> pm_ptr() for runtime_suspend (there is only RUNTIME_PM_OPS()), so for
> avoiding __maybe_unused, I'd have to expand them manually instead.

I'm not sure I got the use case. If we have runtime PM, we use pm_ptr(),
otherwise pm_sleep_ptr().

-- 
With Best Regards,
Andy Shevchenko