Re: [RFC PATCH 3/3] mmc: jz4740: Use pm_sleep_ptr() macro

[Ulf Hansson]

On Tue, 11 Feb 2020 at 17:03, Paul Cercueil <paul@xxxxxxxxxxxxxxx> wrote:
>
> Use the newly introduced pm_sleep_ptr() macro to simplify the code.
>
> Signed-off-by: Paul Cercueil <paul@xxxxxxxxxxxxxxx>
> ---
>  drivers/mmc/host/jz4740_mmc.c | 12 +++---------
>  1 file changed, 3 insertions(+), 9 deletions(-)
>
> diff --git a/drivers/mmc/host/jz4740_mmc.c b/drivers/mmc/host/jz4740_mmc.c
> index fbae87d1f017..09554f9831de 100644
> --- a/drivers/mmc/host/jz4740_mmc.c
> +++ b/drivers/mmc/host/jz4740_mmc.c
> @@ -1099,24 +1099,18 @@ static int jz4740_mmc_remove(struct platform_device *pdev)
>         return 0;
>  }
>
> -#ifdef CONFIG_PM_SLEEP
> -
> -static int jz4740_mmc_suspend(struct device *dev)
> +static int __maybe_unused jz4740_mmc_suspend(struct device *dev)
>  {
>         return pinctrl_pm_select_sleep_state(dev);
>  }
>
> -static int jz4740_mmc_resume(struct device *dev)
> +static int __maybe_unused jz4740_mmc_resume(struct device *dev)
>  {
>         return pinctrl_select_default_state(dev);
>  }
>
>  static SIMPLE_DEV_PM_OPS(jz4740_mmc_pm_ops, jz4740_mmc_suspend,
>         jz4740_mmc_resume);
> -#define JZ4740_MMC_PM_OPS (&jz4740_mmc_pm_ops)
> -#else
> -#define JZ4740_MMC_PM_OPS NULL
> -#endif

All of the above code can be simplified in this way, without having to
convert into using the new pm_sleep_ptr() macro, below.

The only "penalty" would be that, the struct dev_pm_ops
(jz4740_mmc_pm_ops) would then be referenced even when CONFIG_PM* is
unset, thus the compiler would be able to throw it away.

Just wanted to point this out.

>
>  static struct platform_driver jz4740_mmc_driver = {
>         .probe = jz4740_mmc_probe,
> @@ -1124,7 +1118,7 @@ static struct platform_driver jz4740_mmc_driver = {
>         .driver = {
>                 .name = "jz4740-mmc",
>                 .of_match_table = of_match_ptr(jz4740_mmc_of_match),
> -               .pm = JZ4740_MMC_PM_OPS,
> +               .pm = pm_sleep_ptr(&jz4740_mmc_pm_ops),

If the driver would have runtime suspend/resume callbacks, then it
would need the use the pm_ptr() macro instead, I guess.

>         },
>  };
>
> --
> 2.25.0
>

My overall feeling is that this series improves the code/behaviour,
but I am also a bit worried about adding yet another pair of macros
for dealing with CONFIG_PM* callbacks as it could add more confusion.

An option could be to introduce only the pm_ptr() macro, then skip the
optimization that pm_sleep_ptr() gives. This could make it easier to
use, as you wouldn't need to decide between two macros. Just a
thought.

I don't know what Rafael's thinks about this, let's see if he has some
other ideas.

Kind regards
Uffe