Hello,

On 08/26/2010 11:46 AM, Nick Piggin wrote:
quoted text
> Oh, I thought we quiesce / preempt all online cpus before adding
> another one. That sucks if we don't because then you need a big
> heavy get_online_cpus when a simple preempt_disable would have
> worked.
> 
> Why is that? Don't tell me realtime people want some latency "guarantee"
> while onlining CPUs? :)

Probably similar rationale of not doing stop_machine() on module
unload, I suppose.  Onlining something is usually considered hotter
path than offlining.  Performance penalty caused by the difference
between possible and online cpumask or cpu onlining probably only
matters for very large machines and on those machines stop-machine is
very expensive.  If there's a pressing need, doing stop_machine for
onlining too is definitely an option.

quoted text
>> So, yeah, given that there's no cpu notifier implemented, the use of
>> for_each_online_cpu for brlock seems fishy to me.  It probably should
>> use for_each_possible_cpu().
> 
> It should if that's the case, yes.

IMHO, in most configurations the difference between possible and
online cpumasks doesn't matter much (they're the same during normal
operation), so just using possible cpumask should be fine.  It's
already a pretty heavy path, right?

Thanks.

-- 
tejun
--
unsubscribe notice
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/