Re: [PATCH v4 2/4] mfd: Add Samsung Exynos Low Power Audio Subsystem driver

[Sylwester Nawrocki]

On 08/09/2016 10:50 PM, Lee Jones wrote:
> On Tue, 09 Aug 2016, Sylwester Nawrocki wrote:
> 
>> > On 08/09/2016 05:05 PM, Lee Jones wrote:
>>>> > >> +static SIMPLE_DEV_PM_OPS(lpass_pm_ops, exynos_lpass_suspend,
>>>>> > >> > +					exynos_lpass_resume);
>>> > > Put this up by the PM functions.
>> > 
>> > Sorry, I don't understand this comment, which PM functions do you
>> > mean exactly?
> *_suspend and *_resume.  PM == Power Management.
> 
> You also need to place the functions in side some CONFIG_PM_SLEEP
> guards.  Grep around, you'll see what I mean.

OK, but couldn't we just leave it with that SIMPLE_DEV_PM_OPS()
macro?  It saves us an #ifdef ugliness and if I actually grep
drivers/mfd most of the drivers use either SIMPLE_DEV_PM_OPS() or
SET_SYSTEM_SLEEP_PM_OPS. Are you asking me to change the above to:

static const struct dev_pm_ops exynos_lpass_pm_ops = {
#ifdef CONFIG_PM_SLEEP
	.suspend = exynos_lpass_suspend,
	.resume = exynos_lpass_resume,
	.freeze = exynos_lpass_suspend,
	.thaw = exynos_lpass_resume,
	.poweroff = exynos_lpass_suspend,
	.restore = exynos_lpass_resume,
#endif
}

or perhaps

static const struct dev_pm_ops exynos_lpass_pm_ops = {
	SET_SYSTEM_SLEEP_PM_OPS(exynos_lpass_suspend, exynos_lpass_resume)
}

?

I agree about the CONFIG_PM_SLEEP guards.  There is very little
chance though the LPASS driver will ever be used with
CONFIG_PM_SLEEP disabled. How about adding __maybe_unused attribute
to the *_suspend/*_resume functions then?  I can see this pattern
used already in drivers/mfd.

-- 
Thanks,
Sylwester