Date: Tue, 15 Oct 2013 12:35:07 +0200
From: Peter Zijlstra <peterz@infradead.org>
To: Juri Lelli <juri.lelli@gmail.com>
Cc: tglx@linutronix.de, mingo@redhat.com, rostedt@goodmis.org, oleg@redhat.com,
        fweisbec@gmail.com, darren@dvhart.com, johan.eker@ericsson.com,
        p.faure@akatech.ch, linux-kernel@vger.kernel.org,
        claudio@evidence.eu.com, michael@amarulasolutions.com,
        fchecconi@gmail.com, tommaso.cucinotta@sssup.it,
        nicola.manica@disi.unitn.it, luca.abeni@unitn.it,
        dhaval.giani@gmail.com, hgu1972@gmail.com, paulmck@linux.vnet.ibm.com,
        raistlin@linux.it, insop.song@gmail.com, liming.wang@windriver.com,
        jkacur@redhat.com, harald.gustafsson@ericsson.com,
        vincent.guittot@linaro.org,
        bruce.ashfield@windriver.com--no-chain-reply-to
Subject: Re: [PATCH 04/14] sched: SCHED_DEADLINE SMP-related data structures
 & logic.
Message-ID: <20131015103507.GF10651@twins.programming.kicks-ass.net>
References: <1381747426-31334-1-git-send-email-juri.lelli@gmail.com>
 <1381747426-31334-5-git-send-email-juri.lelli@gmail.com>
 <20131014120306.GH3081@twins.programming.kicks-ass.net>
 <525D0C91.1090700@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <525D0C91.1090700@gmail.com>
User-Agent: Mutt/1.5.21 (2012-12-30)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 2553
Lines: 77

On Tue, Oct 15, 2013 at 11:36:17AM +0200, Juri Lelli wrote:
> On 10/14/2013 02:03 PM, Peter Zijlstra wrote:
> > On Mon, Oct 14, 2013 at 12:43:36PM +0200, Juri Lelli wrote:
> >> +static inline void dl_set_overload(struct rq *rq)
> >> +{
> >> +	if (!rq->online)
> >> +		return;
> >> +
> >> +	cpumask_set_cpu(rq->cpu, rq->rd->dlo_mask);
> >> +	/*
> >> +	 * Must be visible before the overload count is
> >> +	 * set (as in sched_rt.c).
> >> +	 */
> >> +	wmb();
> >> +	atomic_inc(&rq->rd->dlo_count);
> >> +}
> > 
> > Please, make that smp_wmb() and modify the comment to point to the
> > matching barrier ; I couldn't find one! Which suggests something is
> > amiss.
> > 
> > Ideally we'd have something like smp_wmb__after_set_bit() but alas.
> > 
> 
> The only user of this function is pull_dl_task (that tries to pull only if at
> least one runqueue of the root_domain is overloaded). Surely makes sense to
> ensure that changes in the dlo_mask have to be visible before we check if we
> should look at that mask. Am I right if I say that the matching barrier is
> constituted by the spin_lock on this_rq acquired by schedule() before calling
> pre_schedule()?
> 
> Same thing in rt_set_overload(), do we need to modify the comment also there?

So I haven't looked at the dl code, but for the RT code the below is
required.

Without that smp_rmb() in there we could actually miss seeing the
rto_mask bit.

---
 kernel/sched/rt.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index e9304cdc26fe..a848f526b941 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -246,8 +246,10 @@ static inline void rt_set_overload(struct rq *rq)
 	 * if we should look at the mask. It would be a shame
 	 * if we looked at the mask, but the mask was not
 	 * updated yet.
+	 *
+	 * Matched by the barrier in pull_rt_task().
 	 */
-	wmb();
+	smp_wmb();
 	atomic_inc(&rq->rd->rto_count);
 }
 
@@ -1626,6 +1628,12 @@ static int pull_rt_task(struct rq *this_rq)
 	if (likely(!rt_overloaded(this_rq)))
 		return 0;
 
+	/*
+	 * Match the barrier from rt_set_overloaded; this guarantees that if we
+	 * see overloaded we must also see the rto_mask bit.
+	 */
+	smp_rmb();
+
 	for_each_cpu(cpu, this_rq->rd->rto_mask) {
 		if (this_cpu == cpu)
 			continue;
--
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/