[PATCH] sched:Prefer numa hotness over cache hotness

[Srikar Dronamraju]

The current load balancer may not try to prevent a task from moving out
of a preferred node to a less preferred node. The reason for this being:

- Since sched features NUMA and NUMA_RESIST_LOWER are disabled by
  default, migrate_degrades_locality() always returns false.

- Even if NUMA_RESIST_LOWER were to be enabled, if its cache hot,
  migrate_degrades_locality() never gets called.

The above behaviour can mean that tasks can move out of their preferred
node but they may be eventually be brought back to their preferred node
by numa balancer (due to higher numa faults).

To avoid the above, this commit merges migrate_degrades_locality() and
migrate_improves_locality(). It also replaces 3 sched features NUMA,
NUMA_FAVOUR_HIGHER and NUMA_RESIST_LOWER by a single sched feature NUMA.

Signed-off-by: Srikar Dronamraju <srikar@xxxxxxxxxxxxxxxxxx>
---
 kernel/sched/fair.c     | 96 ++++++++++++++++++-------------------------------
 kernel/sched/features.h | 18 +++-------
 2 files changed, 40 insertions(+), 74 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index ffeaa41..ddde6fd 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -5467,90 +5467,64 @@ static int task_hot(struct task_struct *p, struct lb_env *env)
 }
 
 #ifdef CONFIG_NUMA_BALANCING
-/* Returns true if the destination node has incurred more faults */
-static bool migrate_improves_locality(struct task_struct *p, struct lb_env *env)
-{
-	struct numa_group *numa_group = rcu_dereference(p->numa_group);
-	int src_nid, dst_nid;
-
-	if (!sched_feat(NUMA_FAVOUR_HIGHER) || !p->numa_faults ||
-	    !(env->sd->flags & SD_NUMA)) {
-		return false;
-	}
-
-	src_nid = cpu_to_node(env->src_cpu);
-	dst_nid = cpu_to_node(env->dst_cpu);
-
-	if (src_nid == dst_nid)
-		return false;
-
-	if (numa_group) {
-		/* Task is already in the group's interleave set. */
-		if (node_isset(src_nid, numa_group->active_nodes))
-			return false;
-
-		/* Task is moving into the group's interleave set. */
-		if (node_isset(dst_nid, numa_group->active_nodes))
-			return true;
-
-		return group_faults(p, dst_nid) > group_faults(p, src_nid);
-	}
-
-	/* Encourage migration to the preferred node. */
-	if (dst_nid == p->numa_preferred_nid)
-		return true;
-
-	return task_faults(p, dst_nid) > task_faults(p, src_nid);
-}
 
+/*
+ * Returns 1, if task migration degrades locality
+ * Returns 0, if task migration improves locality i.e migration preferred.
+ * Returns -1, if task migration is not affected by locality.
+ */
 
-static bool migrate_degrades_locality(struct task_struct *p, struct lb_env *env)
+static int migrate_degrades_locality(struct task_struct *p, struct lb_env *env)
 {
 	struct numa_group *numa_group = rcu_dereference(p->numa_group);
 	int src_nid, dst_nid;
 
-	if (!sched_feat(NUMA) || !sched_feat(NUMA_RESIST_LOWER))
-		return false;
-
 	if (!p->numa_faults || !(env->sd->flags & SD_NUMA))
-		return false;
+		return -1;
+
+	if (!sched_feat(NUMA))
+		return -1;
 
 	src_nid = cpu_to_node(env->src_cpu);
 	dst_nid = cpu_to_node(env->dst_cpu);
 
 	if (src_nid == dst_nid)
-		return false;
+		return -1;
 
 	if (numa_group) {
 		/* Task is moving within/into the group's interleave set. */
-		if (node_isset(dst_nid, numa_group->active_nodes))
-			return false;
+		if (node_isset(dst_nid, numa_group->active_nodes) &&
+				!node_isset(src_nid, numa_group->active_nodes))
+			return 0;
 
 		/* Task is moving out of the group's interleave set. */
-		if (node_isset(src_nid, numa_group->active_nodes))
-			return true;
+		if (!node_isset(dst_nid, numa_group->active_nodes) &&
+				node_isset(src_nid, numa_group->active_nodes))
+			return 1;
 
 		return group_faults(p, dst_nid) < group_faults(p, src_nid);
 	}
 
 	/* Migrating away from the preferred node is always bad. */
-	if (src_nid == p->numa_preferred_nid)
-		return true;
+	if (src_nid == p->numa_preferred_nid) {
+		if (env->src_rq->nr_running > env->src_rq->nr_preferred_running)
+			return 1;
+		else
+			return -1;
+	}
+
+	/* Encourage migration to the preferred node. */
+	if (dst_nid == p->numa_preferred_nid)
+		return 0;
 
 	return task_faults(p, dst_nid) < task_faults(p, src_nid);
 }
 
 #else
-static inline bool migrate_improves_locality(struct task_struct *p,
+static inline int migrate_degrades_locality(struct task_struct *p,
 					     struct lb_env *env)
 {
-	return false;
-}
-
-static inline bool migrate_degrades_locality(struct task_struct *p,
-					     struct lb_env *env)
-{
-	return false;
+	return -1;
 }
 #endif
 
@@ -5560,7 +5534,7 @@ static inline bool migrate_degrades_locality(struct task_struct *p,
 static
 int can_migrate_task(struct task_struct *p, struct lb_env *env)
 {
-	int tsk_cache_hot = 0;
+	int tsk_cache_hot;
 
 	lockdep_assert_held(&env->src_rq->lock);
 
@@ -5618,13 +5592,13 @@ int can_migrate_task(struct task_struct *p, struct lb_env *env)
 	 * 2) task is cache cold, or
 	 * 3) too many balance attempts have failed.
 	 */
-	tsk_cache_hot = task_hot(p, env);
-	if (!tsk_cache_hot)
-		tsk_cache_hot = migrate_degrades_locality(p, env);
+	tsk_cache_hot = migrate_degrades_locality(p, env);
+	if (tsk_cache_hot == -1)
+		tsk_cache_hot = task_hot(p, env);
 
-	if (migrate_improves_locality(p, env) || !tsk_cache_hot ||
+	if (tsk_cache_hot <= 0 ||
 	    env->sd->nr_balance_failed > env->sd->cache_nice_tries) {
-		if (tsk_cache_hot) {
+		if (tsk_cache_hot == 1) {
 			schedstat_inc(env->sd, lb_hot_gained[env->idle]);
 			schedstat_inc(p, se.statistics.nr_forced_migrations);
 		}
diff --git a/kernel/sched/features.h b/kernel/sched/features.h
index 91e33cd..83a50e7 100644
--- a/kernel/sched/features.h
+++ b/kernel/sched/features.h
@@ -79,20 +79,12 @@ SCHED_FEAT(LB_MIN, false)
  * numa_balancing=
  */
 #ifdef CONFIG_NUMA_BALANCING
-SCHED_FEAT(NUMA,	false)
 
 /*
- * NUMA_FAVOUR_HIGHER will favor moving tasks towards nodes where a
- * higher number of hinting faults are recorded during active load
- * balancing.
+ * NUMA will favor moving tasks towards nodes where a higher number of
+ * hinting faults are recorded during active load balancing. It will
+ * resist moving tasks towards nodes where a lower number of hinting
+ * faults have been recorded.
  */
-SCHED_FEAT(NUMA_FAVOUR_HIGHER, true)
-
-/*
- * NUMA_RESIST_LOWER will resist moving tasks towards nodes where a
- * lower number of hinting faults have been recorded. As this has
- * the potential to prevent a task ever migrating to a new node
- * due to CPU overload it is disabled by default.
- */
-SCHED_FEAT(NUMA_RESIST_LOWER, false)
+SCHED_FEAT(NUMA,	true)
 #endif
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/