On Sunday 21 February 2010, Brian King wrote:
> On 02/21/2010 04:27 PM, Rafael J. Wysocki wrote:
> > On Sunday 21 February 2010, Brian King wrote:
> >> On 02/21/2010 04:08 PM, Rafael J. Wysocki wrote:
> >>> I'm not a big fan of __attribute__ ((weak)), though. While we already use that
> >>> in the suspend code, I'm not particularly comfortable with it.
> >>>
> >>> Have you considered any alternative approaches?
> >>
> >> I suppose another option would be to implement this similar to how
> >> arch_free_page and arch_alloc_page do. Something like this:
> >>
> >> #ifndef CONFIG_HAVE_ARCH_SUSPEND_CPUS
> >> static inline int arch_suspend_disable_nonboot_cpus(void)
> >> {
> >> return disable_nonboot_cpus();
> >> }
> >>
> >> static inline void arch_suspend_enable_nonboot_cpus(void)
> >> {
> >> return enable_nonboot_cpus()'
> >> }
> >> #else
> >> extern int arch_suspend_disable_nonboot_cpus(void);
> >> extern void arch_suspend_enable_nonboot_cpus(void);
> >> #endif
> >>
> >> I figured I would just be consistent with arch_suspend_disable_irqs /
> >> arch_suspend_enable_irqs.
> >
> > I just think that doing arch_suspend_[enable|disable]_irqs() this way was
> > a mistake.
>
> Do you prefer the example above? I can send an updated patch. If not,
> any other suggestions you might have as to the way you would like this
> done would be greatly appreciated.
disable_nonboot_cpus() is also called by kernel_power_off(). Is that fine
with your architecture?
Rafael
_______________________________________________
linux-pm mailing list
linux-pm@xxxxxxxxxxxxxxxxxxxxxxxxxx
https://lists.linux-foundation.org/mailman/listinfo/linux-pm
