On Wed, Feb 21, 2024 at 12:13:59PM +0100, Geert Uytterhoeven wrote:
> The conversion to DEFINE_NOIRQ_DEV_PM_OPS() lost the ability of the
> compiler to optimize away the struct dev_pm_ops object when it is not
> needed.
>
> Fix this by replacing the use of pm_sleep_ptr() by a custom wrapper.
>
> Fixes: 727eb02eb753375e ("pinctrl: renesas: Switch to use DEFINE_NOIRQ_DEV_PM_OPS() helper")
Is this a new requirement (I mean more than 12 digits of SHA)?
...
> ---
> To be queued in renesas-pinctrl for v6.9.
>
> Alternatively, one could add a unified definition:
>
> #define pm_psci_sleep_ptr(_ptr) PTR_IF(IS_ENABLED(CONFIG_PM_SLEEP) && \
> IS_ENABLED(CONFIG_ARM_PSCI_FW), (_ptr))
>
> Since there are already separate sections for CONFIG_ARM_PSCI_FW enabled
> vs. disabled, I split in two and simplified the definition.
I'm fine with current approach, IIUC we discussed this that time and the
727eb02eb753 was not the first version (or I can be mistaken, I don't remember
exact driver that the discussion was about during the DEFINE_NOIRQ_DEV_PM_OPS()
conversions).
FWIW,
Reviewed-by: Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx>
--
With Best Regards,
Andy Shevchenko
