sched/fair: Add some serialization to the sched_domain load-balance walk

Since the sched_domain walk is completely unserialized (!SD_SERIALIZE)
it is possible that multiple cpus in the group get elected to do the
next level. Avoid this by adding some serialization.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/n/tip-vqh9ai6s0ewmeakjz80w4qz6@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>

2 files changed, 9 insertions, 2 deletions

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 0533a688ce22..6001e5c3b4e4 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -6060,6 +6060,7 @@ build_overlap_sched_groups(struct sched_domain *sd, int cpu)
 
                 sg->sgp = *per_cpu_ptr(sdd->sgp, cpumask_first(sg_span));
                 atomic_inc(&sg->sgp->ref);
+                sg->balance_cpu = -1;
 
                 if (cpumask_test_cpu(cpu, sg_span))
                         groups = sg;
@@ -6135,6 +6136,7 @@ build_sched_groups(struct sched_domain *sd, int cpu)
 
                 cpumask_clear(sched_group_cpus(sg));
                 sg->sgp->power = 0;
+                sg->balance_cpu = -1;
 
                 for_each_cpu(j, span) {
                         if (get_group(j, sdd, NULL) != group)
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 968ffee24721..cf86f74bcac2 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -3828,7 +3828,8 @@ static inline void update_sg_lb_stats(struct sched_domain *sd,
          */
         if (local_group) {
                 if (idle != CPU_NEWLY_IDLE) {
-                        if (balance_cpu != this_cpu) {
+                        if (balance_cpu != this_cpu ||
+                            cmpxchg(&group->balance_cpu, -1, balance_cpu) != -1) {
                                 *balance = 0;
                                 return;
                         }
@@ -4929,7 +4930,7 @@ static void rebalance_domains(int cpu, enum cpu_idle_type idle)
         int balance = 1;
         struct rq *rq = cpu_rq(cpu);
         unsigned long interval;
-        struct sched_domain *sd;
+        struct sched_domain *sd, *last = NULL;
         /* Earliest time when we have to do rebalance again */
         unsigned long next_balance = jiffies + 60*HZ;
         int update_next_balance = 0;
@@ -4939,6 +4940,7 @@ static void rebalance_domains(int cpu, enum cpu_idle_type idle)
 
         rcu_read_lock();
         for_each_domain(cpu, sd) {
+                last = sd;
                 if (!(sd->flags & SD_LOAD_BALANCE))
                         continue;
 
@@ -4983,6 +4985,9 @@ out:
                 if (!balance)
                         break;
         }
+        for (sd = last; sd; sd = sd->child)
+                (void)cmpxchg(&sd->groups->balance_cpu, cpu, -1);
+
         rcu_read_unlock();
 
         /*