Re: [patch v2 1/2] sched: Use resched IPI to kick off the nohz idle balance

From: Suresh Siddha <suresh.b.siddha@intel.com>
To: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>,
	Venki Pallipadi <venki@google.com>, Ingo Molnar <mingo@elte.hu>,
	Prarit Bhargava <prarit@redhat.com>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"stable@kernel.org" <stable@kernel.org>
Subject: Re: [patch v2 1/2] sched: Use resched IPI to kick off the nohz idle balance
Date: Tue, 04 Oct 2011 11:03:32 -0700	[thread overview]
Message-ID: <1317751412.11592.170.camel@sbsiddha-desk.sc.intel.com> (raw)
In-Reply-To: <1317719121.25926.15.camel@twins>

On Tue, 2011-10-04 at 02:05 -0700, Peter Zijlstra wrote:
> On Mon, 2011-10-03 at 15:09 -0700, Suresh Siddha wrote:
> > +               smp_mb();
> > +               /*
> > +                * Use smp_send_reschedule() instead of resched_cpu().
> > +                * This way we generate a sched IPI on the target cpu which
> > +                * is idle. And the softirq performing nohz idle load balance
> > +                * will be run before returning from the IPI.
> > +                */
> > +               smp_send_reschedule(ilb_cpu); 
> 
> Now the only thing missing is where the matching barrier is, does
> receiving an interrupt imply a mb? If so that wants mentioning
> somewhere, otherwise people will wonder how we can ever read the right
> value.

This is similar to resched_task()? Isn't it?

I am expecting the interrupt handlers will be already having the
necessary memory barriers, if the arch requires it.

thanks,
suresh