1 files changed, 183 insertions, 82 deletions

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 7e9bd0b1fa9..9855e87d671 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -1095,6 +1095,34 @@ static void task_numa_assign(struct task_numa_env *env,
 	env->best_cpu = env->dst_cpu;
 }
 
+static bool load_too_imbalanced(long orig_src_load, long orig_dst_load,
+				long src_load, long dst_load,
+				struct task_numa_env *env)
+{
+	long imb, old_imb;
+
+	/* We care about the slope of the imbalance, not the direction. */
+	if (dst_load < src_load)
+		swap(dst_load, src_load);
+
+	/* Is the difference below the threshold? */
+	imb = dst_load * 100 - src_load * env->imbalance_pct;
+	if (imb <= 0)
+		return false;
+
+	/*
+	 * The imbalance is above the allowed threshold.
+	 * Compare it with the old imbalance.
+	 */
+	if (orig_dst_load < orig_src_load)
+		swap(orig_dst_load, orig_src_load);
+
+	old_imb = orig_dst_load * 100 - orig_src_load * env->imbalance_pct;
+
+	/* Would this change make things worse? */
+	return (imb > old_imb);
+}
+
 /*
  * This checks if the overall compute and NUMA accesses of the system would
  * be improved if the source tasks was migrated to the target dst_cpu taking
@@ -1107,7 +1135,8 @@ static void task_numa_compare(struct task_numa_env *env,
 	struct rq *src_rq = cpu_rq(env->src_cpu);
 	struct rq *dst_rq = cpu_rq(env->dst_cpu);
 	struct task_struct *cur;
-	long dst_load, src_load;
+	long orig_src_load, src_load;
+	long orig_dst_load, dst_load;
 	long load;
 	long imp = (groupimp > 0) ? groupimp : taskimp;
 
@@ -1181,13 +1210,13 @@ static void task_numa_compare(struct task_numa_env *env,
 	 * In the overloaded case, try and keep the load balanced.
 	 */
 balance:
-	dst_load = env->dst_stats.load;
-	src_load = env->src_stats.load;
+	orig_dst_load = env->dst_stats.load;
+	orig_src_load = env->src_stats.load;
 
 	/* XXX missing power terms */
 	load = task_h_load(env->p);
-	dst_load += load;
-	src_load -= load;
+	dst_load = orig_dst_load + load;
+	src_load = orig_src_load - load;
 
 	if (cur) {
 		load = task_h_load(cur);
@@ -1195,11 +1224,8 @@ balance:
 		src_load += load;
 	}
 
-	/* make src_load the smaller */
-	if (dst_load < src_load)
-		swap(dst_load, src_load);
-
-	if (src_load * env->imbalance_pct < dst_load * 100)
+	if (load_too_imbalanced(orig_src_load, orig_dst_load,
+				src_load, dst_load, env))
 		goto unlock;
 
 assign:
@@ -1301,7 +1327,16 @@ static int task_numa_migrate(struct task_struct *p)
 	if (env.best_cpu == -1)
 		return -EAGAIN;
 
-	sched_setnuma(p, env.dst_nid);
+	/*
+	 * If the task is part of a workload that spans multiple NUMA nodes,
+	 * and is migrating into one of the workload's active nodes, remember
+	 * this node as the task's preferred numa node, so the workload can
+	 * settle down.
+	 * A task that migrated to a second choice node will be better off
+	 * trying for a better one later. Do not set the preferred node here.
+	 */
+	if (p->numa_group && node_isset(env.dst_nid, p->numa_group->active_nodes))
+		sched_setnuma(p, env.dst_nid);
 
 	/*
 	 * Reset the scan period if the task is being rescheduled on an
@@ -1326,12 +1361,15 @@ static int task_numa_migrate(struct task_struct *p)
 /* Attempt to migrate a task to a CPU on the preferred node. */
 static void numa_migrate_preferred(struct task_struct *p)
 {
+	unsigned long interval = HZ;
+
 	/* This task has no NUMA fault statistics yet */
 	if (unlikely(p->numa_preferred_nid == -1 || !p->numa_faults_memory))
 		return;
 
 	/* Periodically retry migrating the task to the preferred node */
-	p->numa_migrate_retry = jiffies + HZ;
+	interval = min(interval, msecs_to_jiffies(p->numa_scan_period) / 16);
+	p->numa_migrate_retry = jiffies + interval;
 
 	/* Success if task is already running on preferred CPU */
 	if (task_node(p) == p->numa_preferred_nid)
@@ -1497,7 +1535,7 @@ static void task_numa_placement(struct task_struct *p)
 	/* If the task is part of a group prevent parallel updates to group stats */
 	if (p->numa_group) {
 		group_lock = &p->numa_group->lock;
-		spin_lock(group_lock);
+		spin_lock_irq(group_lock);
 	}
 
 	/* Find the node with the highest number of faults */
@@ -1572,7 +1610,7 @@ static void task_numa_placement(struct task_struct *p)
 			}
 		}
 
-		spin_unlock(group_lock);
+		spin_unlock_irq(group_lock);
 	}
 
 	/* Preferred node as the node with the most faults */
@@ -1677,7 +1715,8 @@ static void task_numa_group(struct task_struct *p, int cpupid, int flags,
 	if (!join)
 		return;
 
-	double_lock(&my_grp->lock, &grp->lock);
+	BUG_ON(irqs_disabled());
+	double_lock_irq(&my_grp->lock, &grp->lock);
 
 	for (i = 0; i < NR_NUMA_HINT_FAULT_STATS * nr_node_ids; i++) {
 		my_grp->faults[i] -= p->numa_faults_memory[i];
@@ -1691,7 +1730,7 @@ static void task_numa_group(struct task_struct *p, int cpupid, int flags,
 	grp->nr_tasks++;
 
 	spin_unlock(&my_grp->lock);
-	spin_unlock(&grp->lock);
+	spin_unlock_irq(&grp->lock);
 
 	rcu_assign_pointer(p->numa_group, grp);
 
@@ -1706,18 +1745,19 @@ no_join:
 void task_numa_free(struct task_struct *p)
 {
 	struct numa_group *grp = p->numa_group;
-	int i;
 	void *numa_faults = p->numa_faults_memory;
+	unsigned long flags;
+	int i;
 
 	if (grp) {
-		spin_lock(&grp->lock);
+		spin_lock_irqsave(&grp->lock, flags);
 		for (i = 0; i < NR_NUMA_HINT_FAULT_STATS * nr_node_ids; i++)
 			grp->faults[i] -= p->numa_faults_memory[i];
 		grp->total_faults -= p->total_numa_faults;
 
 		list_del(&p->numa_entry);
 		grp->nr_tasks--;
-		spin_unlock(&grp->lock);
+		spin_unlock_irqrestore(&grp->lock, flags);
 		rcu_assign_pointer(p->numa_group, NULL);
 		put_numa_group(grp);
 	}
@@ -1737,6 +1777,7 @@ void task_numa_fault(int last_cpupid, int mem_node, int pages, int flags)
 	struct task_struct *p = current;
 	bool migrated = flags & TNF_MIGRATED;
 	int cpu_node = task_node(current);
+	int local = !!(flags & TNF_FAULT_LOCAL);
 	int priv;
 
 	if (!numabalancing_enabled)
@@ -1785,6 +1826,17 @@ void task_numa_fault(int last_cpupid, int mem_node, int pages, int flags)
 			task_numa_group(p, last_cpupid, flags, &priv);
 	}
 
+	/*
+	 * If a workload spans multiple NUMA nodes, a shared fault that
+	 * occurs wholly within the set of nodes that the workload is
+	 * actively using should be counted as local. This allows the
+	 * scan rate to slow down when a workload has settled down.
+	 */
+	if (!priv && !local && p->numa_group &&
+			node_isset(cpu_node, p->numa_group->active_nodes) &&
+			node_isset(mem_node, p->numa_group->active_nodes))
+		local = 1;
+
 	task_numa_placement(p);
 
 	/*
@@ -1799,7 +1851,7 @@ void task_numa_fault(int last_cpupid, int mem_node, int pages, int flags)
 
 	p->numa_faults_buffer_memory[task_faults_idx(mem_node, priv)] += pages;
 	p->numa_faults_buffer_cpu[task_faults_idx(cpu_node, priv)] += pages;
-	p->numa_faults_locality[!!(flags & TNF_FAULT_LOCAL)] += pages;
+	p->numa_faults_locality[local] += pages;
 }
 
 static void reset_ptenuma_scan(struct task_struct *p)
@@ -3128,7 +3180,7 @@ static int assign_cfs_rq_runtime(struct cfs_rq *cfs_rq)
 		 */
 		if (!cfs_b->timer_active) {
 			__refill_cfs_bandwidth_runtime(cfs_b);
-			__start_cfs_bandwidth(cfs_b);
+			__start_cfs_bandwidth(cfs_b, false);
 		}
 
 		if (cfs_b->runtime > 0) {
@@ -3300,14 +3352,14 @@ static void throttle_cfs_rq(struct cfs_rq *cfs_rq)
 	}
 
 	if (!se)
-		rq->nr_running -= task_delta;
+		sub_nr_running(rq, task_delta);
 
 	cfs_rq->throttled = 1;
 	cfs_rq->throttled_clock = rq_clock(rq);
 	raw_spin_lock(&cfs_b->lock);
 	list_add_tail_rcu(&cfs_rq->throttled_list, &cfs_b->throttled_cfs_rq);
 	if (!cfs_b->timer_active)
-		__start_cfs_bandwidth(cfs_b);
+		__start_cfs_bandwidth(cfs_b, false);
 	raw_spin_unlock(&cfs_b->lock);
 }
 
@@ -3351,7 +3403,7 @@ void unthrottle_cfs_rq(struct cfs_rq *cfs_rq)
 	}
 
 	if (!se)
-		rq->nr_running += task_delta;
+		add_nr_running(rq, task_delta);
 
 	/* determine whether we need to wake up potentially idle cpu */
 	if (rq->curr == rq->idle && rq->cfs.nr_running)
@@ -3689,7 +3741,7 @@ static void init_cfs_rq_runtime(struct cfs_rq *cfs_rq)
 }
 
 /* requires cfs_b->lock, may release to reprogram timer */
-void __start_cfs_bandwidth(struct cfs_bandwidth *cfs_b)
+void __start_cfs_bandwidth(struct cfs_bandwidth *cfs_b, bool force)
 {
 	/*
 	 * The timer may be active because we're trying to set a new bandwidth
@@ -3704,7 +3756,7 @@ void __start_cfs_bandwidth(struct cfs_bandwidth *cfs_b)
 		cpu_relax();
 		raw_spin_lock(&cfs_b->lock);
 		/* if someone else restarted the timer then we're done */
-		if (cfs_b->timer_active)
+		if (!force && cfs_b->timer_active)
 			return;
 	}
 
@@ -3883,7 +3935,7 @@ enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags)
 
 	if (!se) {
 		update_rq_runnable_avg(rq, rq->nr_running);
-		inc_nr_running(rq);
+		add_nr_running(rq, 1);
 	}
 	hrtick_update(rq);
 }
@@ -3943,7 +3995,7 @@ static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags)
 	}
 
 	if (!se) {
-		dec_nr_running(rq);
+		sub_nr_running(rq, 1);
 		update_rq_runnable_avg(rq, 1);
 	}
 	hrtick_update(rq);
@@ -4014,7 +4066,7 @@ static void record_wakee(struct task_struct *p)
 	 * about the loss.
 	 */
 	if (jiffies > current->wakee_flip_decay_ts + HZ) {
-		current->wakee_flips = 0;
+		current->wakee_flips >>= 1;
 		current->wakee_flip_decay_ts = jiffies;
 	}
 
@@ -4448,10 +4500,10 @@ select_task_rq_fair(struct task_struct *p, int prev_cpu, int sd_flag, int wake_f
 			sd = tmp;
 	}
 
-	if (affine_sd) {
-		if (cpu != prev_cpu && wake_affine(affine_sd, p, sync))
-			prev_cpu = cpu;
+	if (affine_sd && cpu != prev_cpu && wake_affine(affine_sd, p, sync))
+		prev_cpu = cpu;
 
+	if (sd_flag & SD_BALANCE_WAKE) {
 		new_cpu = select_idle_sibling(p, prev_cpu);
 		goto unlock;
 	}
@@ -4519,6 +4571,9 @@ migrate_task_rq_fair(struct task_struct *p, int next_cpu)
 		atomic_long_add(se->avg.load_avg_contrib,
 						&cfs_rq->removed_load);
 	}
+
+	/* We have migrated, no longer consider this task hot */
+	se->exec_start = 0;
 }
 #endif /* CONFIG_SMP */
 
@@ -5069,6 +5124,7 @@ task_hot(struct task_struct *p, u64 now)
 /* 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_memory ||
@@ -5082,21 +5138,29 @@ static bool migrate_improves_locality(struct task_struct *p, struct lb_env *env)
 	if (src_nid == dst_nid)
 		return false;
 
-	/* Always encourage migration to the preferred node. */
-	if (dst_nid == p->numa_preferred_nid)
-		return true;
+	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);
+	}
 
-	/* If both task and group weight improve, this move is a winner. */
-	if (task_weight(p, dst_nid) > task_weight(p, src_nid) &&
-	    group_weight(p, dst_nid) > group_weight(p, src_nid))
+	/* Encourage migration to the preferred node. */
+	if (dst_nid == p->numa_preferred_nid)
 		return true;
 
-	return false;
+	return task_faults(p, dst_nid) > task_faults(p, src_nid);
 }
 
 
 static bool 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))
@@ -5111,16 +5175,23 @@ static bool migrate_degrades_locality(struct task_struct *p, struct lb_env *env)
 	if (src_nid == dst_nid)
 		return false;
 
+	if (numa_group) {
+		/* Task is moving within/into the group's interleave set. */
+		if (node_isset(dst_nid, numa_group->active_nodes))
+			return false;
+
+		/* Task is moving out of the group's interleave set. */
+		if (node_isset(src_nid, numa_group->active_nodes))
+			return true;
+
+		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 either task or group weight get worse, don't do it. */
-	if (task_weight(p, dst_nid) < task_weight(p, src_nid) ||
-	    group_weight(p, dst_nid) < group_weight(p, src_nid))
-		return true;
-
-	return false;
+	return task_faults(p, dst_nid) < task_faults(p, src_nid);
 }
 
 #else
@@ -5563,6 +5634,7 @@ static unsigned long scale_rt_power(int cpu)
 {
 	struct rq *rq = cpu_rq(cpu);
 	u64 total, available, age_stamp, avg;
+	s64 delta;
 
 	/*
 	 * Since we're reading these variables without serialization make sure
@@ -5571,7 +5643,11 @@ static unsigned long scale_rt_power(int cpu)
 	age_stamp = ACCESS_ONCE(rq->age_stamp);
 	avg = ACCESS_ONCE(rq->rt_avg);
 
-	total = sched_avg_period() + (rq_clock(rq) - age_stamp);
+	delta = rq_clock(rq) - age_stamp;
+	if (unlikely(delta < 0))
+		delta = 0;
+
+	total = sched_avg_period() + delta;
 
 	if (unlikely(total < avg)) {
 		/* Ensures that power won't end up being negative */
@@ -6639,27 +6715,62 @@ out:
 	return ld_moved;
 }
 
+static inline unsigned long
+get_sd_balance_interval(struct sched_domain *sd, int cpu_busy)
+{
+	unsigned long interval = sd->balance_interval;
+
+	if (cpu_busy)
+		interval *= sd->busy_factor;
+
+	/* scale ms to jiffies */
+	interval = msecs_to_jiffies(interval);
+	interval = clamp(interval, 1UL, max_load_balance_interval);
+
+	return interval;
+}
+
+static inline void
+update_next_balance(struct sched_domain *sd, int cpu_busy, unsigned long *next_balance)
+{
+	unsigned long interval, next;
+
+	interval = get_sd_balance_interval(sd, cpu_busy);
+	next = sd->last_balance + interval;
+
+	if (time_after(*next_balance, next))
+		*next_balance = next;
+}
+
 /*
  * idle_balance is called by schedule() if this_cpu is about to become
  * idle. Attempts to pull tasks from other CPUs.
  */
 static int idle_balance(struct rq *this_rq)
 {
+	unsigned long next_balance = jiffies + HZ;
+	int this_cpu = this_rq->cpu;
 	struct sched_domain *sd;
 	int pulled_task = 0;
-	unsigned long next_balance = jiffies + HZ;
 	u64 curr_cost = 0;
-	int this_cpu = this_rq->cpu;
 
 	idle_enter_fair(this_rq);
+
 	/*
 	 * We must set idle_stamp _before_ calling idle_balance(), such that we
 	 * measure the duration of idle_balance() as idle time.
 	 */
 	this_rq->idle_stamp = rq_clock(this_rq);
 
-	if (this_rq->avg_idle < sysctl_sched_migration_cost)
+	if (this_rq->avg_idle < sysctl_sched_migration_cost) {
+		rcu_read_lock();
+		sd = rcu_dereference_check_sched_domain(this_rq->sd);
+		if (sd)
+			update_next_balance(sd, 0, &next_balance);
+		rcu_read_unlock();
+
 		goto out;
+	}
 
 	/*
 	 * Drop the rq->lock, but keep IRQ/preempt disabled.
@@ -6669,20 +6780,20 @@ static int idle_balance(struct rq *this_rq)
 	update_blocked_averages(this_cpu);
 	rcu_read_lock();
 	for_each_domain(this_cpu, sd) {
-		unsigned long interval;
 		int continue_balancing = 1;
 		u64 t0, domain_cost;
 
 		if (!(sd->flags & SD_LOAD_BALANCE))
 			continue;
 
-		if (this_rq->avg_idle < curr_cost + sd->max_newidle_lb_cost)
+		if (this_rq->avg_idle < curr_cost + sd->max_newidle_lb_cost) {
+			update_next_balance(sd, 0, &next_balance);
 			break;
+		}
 
 		if (sd->flags & SD_BALANCE_NEWIDLE) {
 			t0 = sched_clock_cpu(this_cpu);
 
-			/* If we've pulled tasks over stop searching: */
 			pulled_task = load_balance(this_cpu, this_rq,
 						   sd, CPU_NEWLY_IDLE,
 						   &continue_balancing);
@@ -6694,41 +6805,37 @@ static int idle_balance(struct rq *this_rq)
 			curr_cost += domain_cost;
 		}
 
-		interval = msecs_to_jiffies(sd->balance_interval);
-		if (time_after(next_balance, sd->last_balance + interval))
-			next_balance = sd->last_balance + interval;
-		if (pulled_task)
+		update_next_balance(sd, 0, &next_balance);
+
+		/*
+		 * Stop searching for tasks to pull if there are
+		 * now runnable tasks on this rq.
+		 */
+		if (pulled_task || this_rq->nr_running > 0)
 			break;
 	}
 	rcu_read_unlock();
 
 	raw_spin_lock(&this_rq->lock);
 
+	if (curr_cost > this_rq->max_idle_balance_cost)
+		this_rq->max_idle_balance_cost = curr_cost;
+
 	/*
-	 * While browsing the domains, we released the rq lock.
-	 * A task could have be enqueued in the meantime
+	 * While browsing the domains, we released the rq lock, a task could
+	 * have been enqueued in the meantime. Since we're not going idle,
+	 * pretend we pulled a task.
 	 */
-	if (this_rq->cfs.h_nr_running && !pulled_task) {
+	if (this_rq->cfs.h_nr_running && !pulled_task)
 		pulled_task = 1;
-		goto out;
-	}
 
-	if (pulled_task || time_after(jiffies, this_rq->next_balance)) {
-		/*
-		 * We are going idle. next_balance may be set based on
-		 * a busy processor. So reset next_balance.
-		 */
+out:
+	/* Move the next balance forward */
+	if (time_after(this_rq->next_balance, next_balance))
 		this_rq->next_balance = next_balance;
-	}
-
-	if (curr_cost > this_rq->max_idle_balance_cost)
-		this_rq->max_idle_balance_cost = curr_cost;
 
-out:
 	/* Is there a task of a high priority class? */
-	if (this_rq->nr_running != this_rq->cfs.h_nr_running &&
-	    (this_rq->dl.dl_nr_running ||
-	     (this_rq->rt.rt_nr_running && !rt_rq_throttled(&this_rq->rt))))
+	if (this_rq->nr_running != this_rq->cfs.h_nr_running)
 		pulled_task = -1;
 
 	if (pulled_task) {
@@ -7009,16 +7116,9 @@ static void rebalance_domains(struct rq *rq, enum cpu_idle_type idle)
 			break;
 		}
 
-		interval = sd->balance_interval;
-		if (idle != CPU_IDLE)
-			interval *= sd->busy_factor;
-
-		/* scale ms to jiffies */
-		interval = msecs_to_jiffies(interval);
-		interval = clamp(interval, 1UL, max_load_balance_interval);
+		interval = get_sd_balance_interval(sd, idle != CPU_IDLE);
 
 		need_serialize = sd->flags & SD_SERIALIZE;
-
 		if (need_serialize) {
 			if (!spin_trylock(&balancing))
 				goto out;
@@ -7034,6 +7134,7 @@ static void rebalance_domains(struct rq *rq, enum cpu_idle_type idle)
 				idle = idle_cpu(cpu) ? CPU_IDLE : CPU_NOT_IDLE;
 			}
 			sd->last_balance = jiffies;
+			interval = get_sd_balance_interval(sd, idle != CPU_IDLE);
 		}
 		if (need_serialize)
 			spin_unlock(&balancing);