Re: [PATCH x86] [15/16] Force __cpuinit on for CONFIG_PM without HOTPLUG_CPU

[Adrian Bunk <bunk@kernel.org>]

On Thu, Jan 03, 2008 at 07:43:43PM +0100, Andi Kleen wrote:
> On Thursday 03 January 2008 19:14:38 Adrian Bunk wrote:
> > On Thu, Jan 03, 2008 at 04:42:29PM +0100, Andi Kleen wrote:
> > > 
> > > This avoids the requirement to mark a lot of initialization functions not 
> > > __cpuinit just for resume from RAM.
> > > 
> > > More functions could be converted now, didn't do all.
> > >...
> > 
> > Shouldn't this aready be handled by the following?
> > 
> > config PM_SLEEP_SMP
> >         bool
> >         depends on SUSPEND_SMP_POSSIBLE || HIBERNATION_SMP_POSSIBLE
> >         depends on PM_SLEEP
> >         select HOTPLUG_CPU
> >         default y
> 
> Won't help for UP at least. 

I know that it's not popular to care about the kernel size, but your 
patch will cost precious memory in the common case of UP embedded 
systems with CONFIG_PM=y.

It seems the correct solution would be not to hijack __cpuinit
(as your patch does), but to create a new annotation.

> -Andi

cu
Adrian

BTW: Is there any good reason why your patch is x86 only?
     No matter how this gets handled, it should be an architecture 
     independent issue.

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed