Re: [PATCH 1/2] cpumask: centralize cpu_online_map and cpu_possible_map

[Rusty Russell <rusty@rustcorp.com.au>]

On Thursday 20 November 2008 04:47:44 Russell King wrote:
> On Thu, Nov 20, 2008 at 01:07:16AM +1030, Rusty Russell wrote:
> > diff -r c8ab7f6fc5e8 kernel/cpu.c
> > --- a/kernel/cpu.c	Fri Oct 31 10:48:30 2008 +1100
> > +++ b/kernel/cpu.c	Fri Oct 31 11:22:29 2008 +1100
> > @@ -24,19 +24,20 @@ cpumask_t cpu_present_map __read_mostly;
> >  cpumask_t cpu_present_map __read_mostly;
> >  EXPORT_SYMBOL(cpu_present_map);
> >
> > -#ifndef CONFIG_SMP
> > -
> >  /*
> >   * Represents all cpu's that are currently online.
> >   */
> > -cpumask_t cpu_online_map __read_mostly = CPU_MASK_ALL;
> > +cpumask_t cpu_online_map __read_mostly;
>
> Just a question: in the uniprocessor case, does this mean that
> cpu_online_map becomes zero or do we mark cpu0 as online somewhere?
> I couldn't see it in this patch.
>
> I'm just wondering from a review point of view whether this change
> of initialization could have undesirable side effects.

Just checked; we actually set it in boot_cpu_init() already (init/main.c).
Note that all the cpu iterators on UP ignore the mask anyway.

So I think a worthwhile cleanup.

Thanks,
Rusty.