On Fri, 23 May 2025, Luke Jones wrote:

> On Fri, 23 May 2025, at 3:25 PM, Ilpo Järvinen wrote:
> > On Fri, 23 May 2025, Luke Jones wrote:
> >
> >> The patch "Refactor Ally suspend/resume" introduced an
> >
> > The commit feea7bd6b02d ("...")
> >
> >> acpi_s2idle_dev_ops for use with ROG Ally which caused a build error
> >> if CONFIG_SUSPEND was not defined.
> >> 
> >> Signed-off-by: Luke Jones <luke@ljones.dev>
> >> Reported-by: kernel test robot <lkp@intel.com>
> >> Closes: https://lore.kernel.org/oe-kbuild-all/202505090418.DaeaXe4i-lkp@intel.com/
> >> Fixes: feea7bd6b02d ("platform/x86: asus-wmi: Refactor Ally suspend/resume")
> >> ---
> >>  drivers/platform/x86/asus-wmi.c | 9 +++++++++
> >>  1 file changed, 9 insertions(+)
> >> 
> >> diff --git a/drivers/platform/x86/asus-wmi.c b/drivers/platform/x86/asus-wmi.c
> >> index 27f11643a00d..087318e0d595 100644
> >> --- a/drivers/platform/x86/asus-wmi.c
> >> +++ b/drivers/platform/x86/asus-wmi.c
> >> @@ -5005,6 +5005,7 @@ static int asus_hotk_restore(struct device *device)
> >>  	return 0;
> >>  }
> >>  
> >> +#if defined(CONFIG_SUSPEND)
> >>  static void asus_ally_s2idle_restore(void)
> >>  {
> >>  	if (use_ally_mcu_hack == ASUS_WMI_ALLY_MCU_HACK_ENABLED) {
> >> @@ -5013,6 +5014,7 @@ static void asus_ally_s2idle_restore(void)
> >>  		msleep(ASUS_USB0_PWR_EC0_CSEE_WAIT);
> >>  	}
> >>  }
> >> +#endif /* CONFIG_SUSPEND */
> >
> > Move this function below asus_hotk_prepare() next to ops, so that only one 
> > #if block is needed for them.
> >
> 
> Done
> 
> >>  
> >>  static int asus_hotk_prepare(struct device *device)
> >>  {
> >> @@ -5025,9 +5027,11 @@ static int asus_hotk_prepare(struct device *device)
> >>  }
> >>  
> >>  /* Use only for Ally devices due to the wake_on_ac */
> >> +#if defined(CONFIG_SUSPEND)
> >>  static struct acpi_s2idle_dev_ops asus_ally_s2idle_dev_ops = {
> >>  	.restore = asus_ally_s2idle_restore,
> >>  };
> >> +#endif /* CONFIG_SUSPEND */
> >>  
> >>  static const struct dev_pm_ops asus_pm_ops = {
> >>  	.thaw = asus_hotk_thaw,
> >> @@ -5060,9 +5064,11 @@ static int asus_wmi_probe(struct platform_device *pdev)
> >>  			return ret;
> >>  	}
> >>  
> >> +	#if defined(CONFIG_SUSPEND)
> >>  	ret = acpi_register_lps0_dev(&asus_ally_s2idle_dev_ops);
> >>  	if (ret)
> >>  		pr_warn("failed to register LPS0 sleep handler in asus-wmi\n");
> >> +	#endif /* CONFIG_SUSPEND */
> >>  
> >>  	return asus_wmi_add(pdev);
> >>  }
> >> @@ -5096,7 +5102,10 @@ EXPORT_SYMBOL_GPL(asus_wmi_register_driver);
> >>  
> >>  void asus_wmi_unregister_driver(struct asus_wmi_driver *driver)
> >>  {
> >> +	#if defined(CONFIG_SUSPEND)
> >>  	acpi_unregister_lps0_dev(&asus_ally_s2idle_dev_ops);
> >> +	#endif /* CONFIG_SUSPEND */
> >
> > I'd have preferred these reg/unreg be solved with wrappers (see 
> > pmc_atom.c).
> 
> Can do, but I don't really understand the difference it makes in this context. Should I log that  CONFIG_SUSPEND was not enabled? Otherwise it's just as good as not there, right?

I just want to limit the use of ifdeffery to minimum and prefer using 
#else + a stub over sprinkling ifdefs all over the place. And it's 
orthogonal whether the CONFIG_SUSPEND=n should print a warning or not.

I'll take the v3.

-- 
 i.