Messages in this thread | | | Subject | Re: [RFC PATCH 00/10] Cleanup ipi_call_lock[_irq]()/ipi_call_unlock[_irq]() | From | Peter Zijlstra <> | Date | Tue, 29 May 2012 10:32:15 +0200 |
| |
On Tue, 2012-05-29 at 10:28 +0200, Peter Zijlstra wrote: > On Tue, 2012-05-29 at 15:15 +0800, Yong Zhang wrote: > > As discussed with Srivatsa [1], it seems there is no need to keep > > ipi_call_[un]lock_irq() when cpu bring-up/down. Because: > > > > 1) call_function.lock used in smp_call_function_many() is just to protect > > call_function.queue and &data->refs, cpu_online_mask is outside of the > > lock. And it's not necessary to protect cpu_online_mask, > > because data->cpumask is pre-calculate and even if a cpu is brougt up > > when calling arch_send_call_function_ipi_mask(), it's harmless because > > validation test in generic_smp_call_function_interrupt() will take care > > of it. > > > > 2) For cpu down issue, stop_machine() will guarantee that no concurrent > > smp_call_fuction() is processing. > > But that lock was only taken around setting a cpu online, so the offline > case is pretty much irrelevant for these patches, right?
Ah, I see, some archs also did it on offline.
> That said, is there an alternative to stop_machine on the down side? > > I guess flipping the cpu offline and then doing synchronize_sched() > should suffice.
| |