sched: Implement on-demand (active) cfs_rq list
Make certain load-balance actions scale per number of active cgroups
instead of the number of existing cgroups.
This makes wakeup/sleep paths more expensive, but is a win for systems
where the vast majority of existing cgroups are idle.
Signed-off-by: Paul Turner <pjt@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20101115234937.666535048@google.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
diff --git a/kernel/sched_fair.c b/kernel/sched_fair.c
index d86544b..0560e72 100644
--- a/kernel/sched_fair.c
+++ b/kernel/sched_fair.c
@@ -143,6 +143,24 @@
return cfs_rq->tg->cfs_rq[this_cpu];
}
+static inline void list_add_leaf_cfs_rq(struct cfs_rq *cfs_rq)
+{
+ if (!cfs_rq->on_list) {
+ list_add_rcu(&cfs_rq->leaf_cfs_rq_list,
+ &rq_of(cfs_rq)->leaf_cfs_rq_list);
+
+ cfs_rq->on_list = 1;
+ }
+}
+
+static inline void list_del_leaf_cfs_rq(struct cfs_rq *cfs_rq)
+{
+ if (cfs_rq->on_list) {
+ list_del_rcu(&cfs_rq->leaf_cfs_rq_list);
+ cfs_rq->on_list = 0;
+ }
+}
+
/* Iterate thr' all leaf cfs_rq's on a runqueue */
#define for_each_leaf_cfs_rq(rq, cfs_rq) \
list_for_each_entry_rcu(cfs_rq, &rq->leaf_cfs_rq_list, leaf_cfs_rq_list)
@@ -246,6 +264,14 @@
return &cpu_rq(this_cpu)->cfs;
}
+static inline void list_add_leaf_cfs_rq(struct cfs_rq *cfs_rq)
+{
+}
+
+static inline void list_del_leaf_cfs_rq(struct cfs_rq *cfs_rq)
+{
+}
+
#define for_each_leaf_cfs_rq(rq, cfs_rq) \
for (cfs_rq = &rq->cfs; cfs_rq; cfs_rq = NULL)
@@ -648,7 +674,7 @@
}
#if defined CONFIG_SMP && defined CONFIG_FAIR_GROUP_SCHED
-static void update_cfs_load(struct cfs_rq *cfs_rq)
+static void update_cfs_load(struct cfs_rq *cfs_rq, int lb)
{
u64 period = sched_avg_period();
u64 now, delta;
@@ -673,6 +699,11 @@
cfs_rq->load_period /= 2;
cfs_rq->load_avg /= 2;
}
+
+ if (lb && !cfs_rq->nr_running) {
+ if (cfs_rq->load_avg < (period / 8))
+ list_del_leaf_cfs_rq(cfs_rq);
+ }
}
static void reweight_entity(struct cfs_rq *cfs_rq, struct sched_entity *se,
@@ -719,7 +750,7 @@
reweight_entity(cfs_rq_of(se), se, shares);
}
#else /* CONFIG_FAIR_GROUP_SCHED */
-static inline void update_cfs_load(struct cfs_rq *cfs_rq)
+static inline void update_cfs_load(struct cfs_rq *cfs_rq, int lb)
{
}
@@ -849,7 +880,7 @@
* Update run-time statistics of the 'current'.
*/
update_curr(cfs_rq);
- update_cfs_load(cfs_rq);
+ update_cfs_load(cfs_rq, 0);
account_entity_enqueue(cfs_rq, se);
update_cfs_shares(cfs_rq);
@@ -863,6 +894,9 @@
if (se != cfs_rq->curr)
__enqueue_entity(cfs_rq, se);
se->on_rq = 1;
+
+ if (cfs_rq->nr_running == 1)
+ list_add_leaf_cfs_rq(cfs_rq);
}
static void __clear_buddies(struct cfs_rq *cfs_rq, struct sched_entity *se)
@@ -907,7 +941,7 @@
if (se != cfs_rq->curr)
__dequeue_entity(cfs_rq, se);
se->on_rq = 0;
- update_cfs_load(cfs_rq);
+ update_cfs_load(cfs_rq, 0);
account_entity_dequeue(cfs_rq, se);
update_min_vruntime(cfs_rq);
update_cfs_shares(cfs_rq);
@@ -1142,7 +1176,7 @@
for_each_sched_entity(se) {
struct cfs_rq *cfs_rq = cfs_rq_of(se);
- update_cfs_load(cfs_rq);
+ update_cfs_load(cfs_rq, 0);
update_cfs_shares(cfs_rq);
}
@@ -1172,7 +1206,7 @@
for_each_sched_entity(se) {
struct cfs_rq *cfs_rq = cfs_rq_of(se);
- update_cfs_load(cfs_rq);
+ update_cfs_load(cfs_rq, 0);
update_cfs_shares(cfs_rq);
}
