X-Git-Url: https://err.no/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=kernel%2Fsched_fair.c;h=01859f662ab7c834826f9197e93ebe9d44ad548a;hb=82798a17ad40df827d465329a20ace80497f9b32;hp=ae2d4b08e782c4e6237208593865ceb93020cf13;hpb=5f6d858ecca78f71755859a346d845e302973cd1;p=linux-2.6

diff --git a/kernel/sched_fair.c b/kernel/sched_fair.c
index ae2d4b08e7..01859f662a 100644
--- a/kernel/sched_fair.c
+++ b/kernel/sched_fair.c
@@ -25,14 +25,12 @@
  * (default: 20ms, units: nanoseconds)
  *
  * NOTE: this latency value is not the same as the concept of
- * 'timeslice length' - timeslices in CFS are of variable length.
- * (to see the precise effective timeslice length of your workload,
- *  run vmstat and monitor the context-switches field)
+ * 'timeslice length' - timeslices in CFS are of variable length
+ * and have no persistent notion like in traditional, time-slice
+ * based scheduling concepts.
  *
- * On SMP systems the value of this is multiplied by the log2 of the
- * number of CPUs. (i.e. factor 2x on 2-way systems, 3x on 4-way
- * systems, 4x on 8-way systems, 5x on 16-way systems, etc.)
- * Targeted preemption latency for CPU-bound tasks:
+ * (to see the precise effective timeslice length of your workload,
+ *  run vmstat and monitor the context-switches (cs) field)
  */
 const_debug unsigned int sysctl_sched_latency = 20000000ULL;
 
@@ -58,25 +56,25 @@ unsigned int __read_mostly sysctl_sched_compat_yield;
 
 /*
  * SCHED_BATCH wake-up granularity.
- * (default: 25 msec, units: nanoseconds)
+ * (default: 10 msec, units: nanoseconds)
  *
  * This option delays the preemption effects of decoupled workloads
  * and reduces their over-scheduling. Synchronous workloads will still
  * have immediate wakeup/sleep latencies.
  */
-const_debug unsigned int sysctl_sched_batch_wakeup_granularity = 25000000UL;
+const_debug unsigned int sysctl_sched_batch_wakeup_granularity = 10000000UL;
 
 /*
  * SCHED_OTHER wake-up granularity.
- * (default: 1 msec, units: nanoseconds)
+ * (default: 10 msec, units: nanoseconds)
  *
  * This option delays the preemption effects of decoupled workloads
  * and reduces their over-scheduling. Synchronous workloads will still
  * have immediate wakeup/sleep latencies.
  */
-const_debug unsigned int sysctl_sched_wakeup_granularity = 2000000UL;
+const_debug unsigned int sysctl_sched_wakeup_granularity = 10000000UL;
 
-extern struct sched_class fair_sched_class;
+const_debug unsigned int sysctl_sched_migration_cost = 500000UL;
 
 /**************************************************************
  * CFS operations on generic schedulable entities:
@@ -114,8 +112,7 @@ static inline struct task_struct *task_of(struct sched_entity *se)
  * Scheduling class tree data structure manipulation methods:
  */
 
-static inline u64
-max_vruntime(u64 min_vruntime, u64 vruntime)
+static inline u64 max_vruntime(u64 min_vruntime, u64 vruntime)
 {
 	s64 delta = (s64)(vruntime - min_vruntime);
 	if (delta > 0)
@@ -124,8 +121,7 @@ max_vruntime(u64 min_vruntime, u64 vruntime)
 	return min_vruntime;
 }
 
-static inline u64
-min_vruntime(u64 min_vruntime, u64 vruntime)
+static inline u64 min_vruntime(u64 min_vruntime, u64 vruntime)
 {
 	s64 delta = (s64)(vruntime - min_vruntime);
 	if (delta < 0)
@@ -134,8 +130,7 @@ min_vruntime(u64 min_vruntime, u64 vruntime)
 	return min_vruntime;
 }
 
-static inline s64
-entity_key(struct cfs_rq *cfs_rq, struct sched_entity *se)
+static inline s64 entity_key(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
 	return se->vruntime - cfs_rq->min_vruntime;
 }
@@ -143,8 +138,7 @@ entity_key(struct cfs_rq *cfs_rq, struct sched_entity *se)
 /*
  * Enqueue an entity into the rb-tree:
  */
-static void
-__enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
+static void __enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
 	struct rb_node **link = &cfs_rq->tasks_timeline.rb_node;
 	struct rb_node *parent = NULL;
@@ -181,8 +175,7 @@ __enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
 	rb_insert_color(&se->run_node, &cfs_rq->tasks_timeline);
 }
 
-static void
-__dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
+static void __dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
 	if (cfs_rq->rb_leftmost == &se->run_node)
 		cfs_rq->rb_leftmost = rb_next(&se->run_node);
@@ -219,6 +212,15 @@ static inline struct sched_entity *__pick_last_entity(struct cfs_rq *cfs_rq)
  * Scheduling class statistics methods:
  */
 
+
+/*
+ * The idea is to set a period in which each task runs once.
+ *
+ * When there are too many tasks (sysctl_sched_nr_latency) we have to stretch
+ * this period because otherwise the slices get too small.
+ *
+ * p = (nr <= nl) ? l : l*nr/nl
+ */
 static u64 __sched_period(unsigned long nr_running)
 {
 	u64 period = sysctl_sched_latency;
@@ -232,27 +234,45 @@ static u64 __sched_period(unsigned long nr_running)
 	return period;
 }
 
+/*
+ * We calculate the wall-time slice from the period by taking a part
+ * proportional to the weight.
+ *
+ * s = p*w/rw
+ */
 static u64 sched_slice(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
-	u64 period = __sched_period(cfs_rq->nr_running);
+	u64 slice = __sched_period(cfs_rq->nr_running);
 
-	period *= se->load.weight;
-	do_div(period, cfs_rq->load.weight);
+	slice *= se->load.weight;
+	do_div(slice, cfs_rq->load.weight);
 
-	return period;
+	return slice;
 }
 
-static u64 __sched_vslice(unsigned long nr_running)
+/*
+ * We calculate the vruntime slice.
+ *
+ * vs = s/w = p/rw
+ */
+static u64 __sched_vslice(unsigned long rq_weight, unsigned long nr_running)
 {
-	unsigned long period = sysctl_sched_latency;
-	unsigned long nr_latency = sysctl_sched_nr_latency;
+	u64 vslice = __sched_period(nr_running);
 
-	if (unlikely(nr_running > nr_latency))
-		nr_running = nr_latency;
+	do_div(vslice, rq_weight);
 
-	period /= nr_running;
+	return vslice;
+}
 
-	return (u64)period;
+static u64 sched_vslice(struct cfs_rq *cfs_rq)
+{
+	return __sched_vslice(cfs_rq->load.weight, cfs_rq->nr_running);
+}
+
+static u64 sched_vslice_add(struct cfs_rq *cfs_rq, struct sched_entity *se)
+{
+	return __sched_vslice(cfs_rq->load.weight + se->load.weight,
+			cfs_rq->nr_running + 1);
 }
 
 /*
@@ -317,17 +337,6 @@ update_stats_wait_start(struct cfs_rq *cfs_rq, struct sched_entity *se)
 	schedstat_set(se->wait_start, rq_of(cfs_rq)->clock);
 }
 
-static inline unsigned long
-calc_weighted(unsigned long delta, struct sched_entity *se)
-{
-	unsigned long weight = se->load.weight;
-
-	if (unlikely(weight != NICE_0_LOAD))
-		return (u64)delta * se->load.weight >> NICE_0_SHIFT;
-	else
-		return delta;
-}
-
 /*
  * Task is being enqueued - update stats:
  */
@@ -352,7 +361,6 @@ update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se)
 static inline void
 update_stats_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
-	update_curr(cfs_rq);
 	/*
 	 * Mark the end of the wait period if dequeueing a
 	 * waiting task:
@@ -373,15 +381,6 @@ update_stats_curr_start(struct cfs_rq *cfs_rq, struct sched_entity *se)
 	se->exec_start = rq_of(cfs_rq)->clock;
 }
 
-/*
- * We are descheduling a task - update its stats:
- */
-static inline void
-update_stats_curr_end(struct cfs_rq *cfs_rq, struct sched_entity *se)
-{
-	se->exec_start = 0;
-}
-
 /**************************************************
  * Scheduling class queueing methods:
  */
@@ -464,20 +463,21 @@ place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial)
 
 	vruntime = cfs_rq->min_vruntime;
 
-	if (sched_feat(USE_TREE_AVG)) {
+	if (sched_feat(TREE_AVG)) {
 		struct sched_entity *last = __pick_last_entity(cfs_rq);
 		if (last) {
 			vruntime += last->vruntime;
 			vruntime >>= 1;
 		}
 	} else if (sched_feat(APPROX_AVG) && cfs_rq->nr_running)
-		vruntime += __sched_vslice(cfs_rq->nr_running)/2;
+		vruntime += sched_vslice(cfs_rq)/2;
 
 	if (initial && sched_feat(START_DEBIT))
-		vruntime += __sched_vslice(cfs_rq->nr_running + 1);
+		vruntime += sched_vslice_add(cfs_rq, se);
 
 	if (!initial) {
-		if (sched_feat(NEW_FAIR_SLEEPERS))
+		if (sched_feat(NEW_FAIR_SLEEPERS) && entity_is_task(se) &&
+				task_of(se)->policy != SCHED_BATCH)
 			vruntime -= sysctl_sched_latency;
 
 		vruntime = max_t(s64, vruntime, se->vruntime);
@@ -491,12 +491,11 @@ static void
 enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int wakeup)
 {
 	/*
-	 * Update the fair clock.
+	 * Update run-time statistics of the 'current'.
 	 */
 	update_curr(cfs_rq);
 
 	if (wakeup) {
-		/* se->vruntime += cfs_rq->min_vruntime; */
 		place_entity(cfs_rq, se, 0);
 		enqueue_sleeper(cfs_rq, se);
 	}
@@ -511,8 +510,14 @@ enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int wakeup)
 static void
 dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int sleep)
 {
+	/*
+	 * Update run-time statistics of the 'current'.
+	 */
+	update_curr(cfs_rq);
+
 	update_stats_dequeue(cfs_rq, se);
 	if (sleep) {
+		se->peer_preempt = 0;
 #ifdef CONFIG_SCHEDSTATS
 		if (entity_is_task(se)) {
 			struct task_struct *tsk = task_of(se);
@@ -540,8 +545,10 @@ check_preempt_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr)
 
 	ideal_runtime = sched_slice(cfs_rq, curr);
 	delta_exec = curr->sum_exec_runtime - curr->prev_sum_exec_runtime;
-	if (delta_exec > ideal_runtime)
+	if (delta_exec > ideal_runtime ||
+			(sched_feat(PREEMPT_RESTRICT) && curr->peer_preempt))
 		resched_task(rq_of(cfs_rq)->curr);
+	curr->peer_preempt = 0;
 }
 
 static void
@@ -576,9 +583,12 @@ set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
 
 static struct sched_entity *pick_next_entity(struct cfs_rq *cfs_rq)
 {
-	struct sched_entity *se = __pick_next_entity(cfs_rq);
+	struct sched_entity *se = NULL;
 
-	set_next_entity(cfs_rq, se);
+	if (first_fair(cfs_rq)) {
+		se = __pick_next_entity(cfs_rq);
+		set_next_entity(cfs_rq, se);
+	}
 
 	return se;
 }
@@ -592,8 +602,6 @@ static void put_prev_entity(struct cfs_rq *cfs_rq, struct sched_entity *prev)
 	if (prev->on_rq)
 		update_curr(cfs_rq);
 
-	update_stats_curr_end(cfs_rq, prev);
-
 	check_spread(cfs_rq, prev);
 	if (prev->on_rq) {
 		update_stats_wait_start(cfs_rq, prev);
@@ -610,7 +618,7 @@ static void entity_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr)
 	 */
 	update_curr(cfs_rq);
 
-	if (cfs_rq->nr_running > 1)
+	if (cfs_rq->nr_running > 1 || !sched_feat(WAKEUP_PREEMPT))
 		check_preempt_tick(cfs_rq, curr);
 }
 
@@ -653,15 +661,21 @@ static inline struct cfs_rq *cpu_cfs_rq(struct cfs_rq *cfs_rq, int this_cpu)
 #define for_each_leaf_cfs_rq(rq, cfs_rq) \
 	list_for_each_entry(cfs_rq, &rq->leaf_cfs_rq_list, leaf_cfs_rq_list)
 
-/* Do the two (enqueued) tasks belong to the same group ? */
-static inline int is_same_group(struct task_struct *curr, struct task_struct *p)
+/* Do the two (enqueued) entities belong to the same group ? */
+static inline int
+is_same_group(struct sched_entity *se, struct sched_entity *pse)
 {
-	if (curr->se.cfs_rq == p->se.cfs_rq)
+	if (se->cfs_rq == pse->cfs_rq)
 		return 1;
 
 	return 0;
 }
 
+static inline struct sched_entity *parent_entity(struct sched_entity *se)
+{
+	return se->parent;
+}
+
 #else	/* CONFIG_FAIR_GROUP_SCHED */
 
 #define for_each_sched_entity(se) \
@@ -694,11 +708,17 @@ static inline struct cfs_rq *cpu_cfs_rq(struct cfs_rq *cfs_rq, int this_cpu)
 #define for_each_leaf_cfs_rq(rq, cfs_rq) \
 		for (cfs_rq = &rq->cfs; cfs_rq; cfs_rq = NULL)
 
-static inline int is_same_group(struct task_struct *curr, struct task_struct *p)
+static inline int
+is_same_group(struct sched_entity *se, struct sched_entity *pse)
 {
 	return 1;
 }
 
+static inline struct sched_entity *parent_entity(struct sched_entity *se)
+{
+	return NULL;
+}
+
 #endif	/* CONFIG_FAIR_GROUP_SCHED */
 
 /*
@@ -716,6 +736,7 @@ static void enqueue_task_fair(struct rq *rq, struct task_struct *p, int wakeup)
 			break;
 		cfs_rq = cfs_rq_of(se);
 		enqueue_entity(cfs_rq, se, wakeup);
+		wakeup = 1;
 	}
 }
 
@@ -735,6 +756,7 @@ static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int sleep)
 		/* Don't dequeue parent if it has other entities besides us */
 		if (cfs_rq->load.weight)
 			break;
+		sleep = 1;
 	}
 }
 
@@ -757,8 +779,7 @@ static void yield_task_fair(struct rq *rq)
 	if (likely(!sysctl_sched_compat_yield)) {
 		__update_rq_clock(rq);
 		/*
-		 * Dequeue and enqueue the task to update its
-		 * position within the tree:
+		 * Update run-time statistics of the 'current'.
 		 */
 		update_curr(cfs_rq);
 
@@ -788,8 +809,9 @@ static void yield_task_fair(struct rq *rq)
 static void check_preempt_wakeup(struct rq *rq, struct task_struct *p)
 {
 	struct task_struct *curr = rq->curr;
-	struct cfs_rq *cfs_rq = task_cfs_rq(curr), *pcfs_rq;
+	struct cfs_rq *cfs_rq = task_cfs_rq(curr);
 	struct sched_entity *se = &curr->se, *pse = &p->se;
+	s64 delta, gran;
 
 	if (unlikely(rt_prio(p->prio))) {
 		update_rq_clock(rq);
@@ -797,21 +819,30 @@ static void check_preempt_wakeup(struct rq *rq, struct task_struct *p)
 		resched_task(curr);
 		return;
 	}
+	/*
+	 * Batch tasks do not preempt (their preemption is driven by
+	 * the tick):
+	 */
+	if (unlikely(p->policy == SCHED_BATCH))
+		return;
 
-	for_each_sched_entity(se) {
-		cfs_rq = cfs_rq_of(se);
-		pcfs_rq = cfs_rq_of(pse);
+	if (sched_feat(WAKEUP_PREEMPT)) {
+		while (!is_same_group(se, pse)) {
+			se = parent_entity(se);
+			pse = parent_entity(pse);
+		}
 
-		if (cfs_rq == pcfs_rq) {
-			s64 delta = se->vruntime - pse->vruntime;
+		delta = se->vruntime - pse->vruntime;
+		gran = sysctl_sched_wakeup_granularity;
+		if (unlikely(se->load.weight != NICE_0_LOAD))
+			gran = calc_delta_fair(gran, &se->load);
 
-			if (delta > (s64)sysctl_sched_wakeup_granularity)
+		if (delta > gran) {
+			int now = !sched_feat(PREEMPT_RESTRICT);
+
+			if (now || p->prio < curr->prio || !se->peer_preempt++)
 				resched_task(curr);
-			break;
 		}
-#ifdef CONFIG_FAIR_GROUP_SCHED
-		pse = pse->parent;
-#endif
 	}
 }
 
@@ -845,6 +876,7 @@ static void put_prev_task_fair(struct rq *rq, struct task_struct *prev)
 	}
 }
 
+#ifdef CONFIG_SMP
 /**************************************************
  * Fair scheduling class load-balancing methods:
  */
@@ -856,7 +888,7 @@ static void put_prev_task_fair(struct rq *rq, struct task_struct *prev)
  * achieve that by always pre-iterating before returning
  * the current task:
  */
-static inline struct task_struct *
+static struct task_struct *
 __load_balance_iterator(struct cfs_rq *cfs_rq, struct rb_node *curr)
 {
 	struct task_struct *p;
@@ -905,12 +937,11 @@ static int cfs_rq_best_prio(struct cfs_rq *cfs_rq)
 
 static unsigned long
 load_balance_fair(struct rq *this_rq, int this_cpu, struct rq *busiest,
-		  unsigned long max_nr_move, unsigned long max_load_move,
+		  unsigned long max_load_move,
 		  struct sched_domain *sd, enum cpu_idle_type idle,
 		  int *all_pinned, int *this_best_prio)
 {
 	struct cfs_rq *busy_cfs_rq;
-	unsigned long load_moved, total_nr_moved = 0, nr_moved;
 	long rem_load_move = max_load_move;
 	struct rq_iterator cfs_rq_iterator;
 
@@ -938,25 +969,48 @@ load_balance_fair(struct rq *this_rq, int this_cpu, struct rq *busiest,
 #else
 # define maxload rem_load_move
 #endif
-		/* pass busy_cfs_rq argument into
+		/*
+		 * pass busy_cfs_rq argument into
 		 * load_balance_[start|next]_fair iterators
 		 */
 		cfs_rq_iterator.arg = busy_cfs_rq;
-		nr_moved = balance_tasks(this_rq, this_cpu, busiest,
-				max_nr_move, maxload, sd, idle, all_pinned,
-				&load_moved, this_best_prio, &cfs_rq_iterator);
+		rem_load_move -= balance_tasks(this_rq, this_cpu, busiest,
+					       maxload, sd, idle, all_pinned,
+					       this_best_prio,
+					       &cfs_rq_iterator);
 
-		total_nr_moved += nr_moved;
-		max_nr_move -= nr_moved;
-		rem_load_move -= load_moved;
-
-		if (max_nr_move <= 0 || rem_load_move <= 0)
+		if (rem_load_move <= 0)
 			break;
 	}
 
 	return max_load_move - rem_load_move;
 }
 
+static int
+move_one_task_fair(struct rq *this_rq, int this_cpu, struct rq *busiest,
+		   struct sched_domain *sd, enum cpu_idle_type idle)
+{
+	struct cfs_rq *busy_cfs_rq;
+	struct rq_iterator cfs_rq_iterator;
+
+	cfs_rq_iterator.start = load_balance_start_fair;
+	cfs_rq_iterator.next = load_balance_next_fair;
+
+	for_each_leaf_cfs_rq(busiest, busy_cfs_rq) {
+		/*
+		 * pass busy_cfs_rq argument into
+		 * load_balance_[start|next]_fair iterators
+		 */
+		cfs_rq_iterator.arg = busy_cfs_rq;
+		if (iter_move_one_task(this_rq, this_cpu, busiest, sd, idle,
+				       &cfs_rq_iterator))
+		    return 1;
+	}
+
+	return 0;
+}
+#endif
+
 /*
  * scheduler tick hitting a task of our scheduling class:
  */
@@ -971,7 +1025,7 @@ static void task_tick_fair(struct rq *rq, struct task_struct *curr)
 	}
 }
 
-#define swap(a,b) do { typeof(a) tmp = (a); (a) = (b); (b) = tmp; } while (0)
+#define swap(a, b) do { typeof(a) tmp = (a); (a) = (b); (b) = tmp; } while (0)
 
 /*
  * Share the fairness runtime between parent and child, thus the
@@ -984,13 +1038,14 @@ static void task_new_fair(struct rq *rq, struct task_struct *p)
 {
 	struct cfs_rq *cfs_rq = task_cfs_rq(p);
 	struct sched_entity *se = &p->se, *curr = cfs_rq->curr;
+	int this_cpu = smp_processor_id();
 
 	sched_info_queued(p);
 
 	update_curr(cfs_rq);
 	place_entity(cfs_rq, se, 1);
 
-	if (sysctl_sched_child_runs_first &&
+	if (sysctl_sched_child_runs_first && this_cpu == task_cpu(p) &&
 			curr->vruntime < se->vruntime) {
 		/*
 		 * Upon rescheduling, sched_class::put_prev_task() will place
@@ -999,11 +1054,8 @@ static void task_new_fair(struct rq *rq, struct task_struct *p)
 		swap(curr->vruntime, se->vruntime);
 	}
 
-	update_stats_enqueue(cfs_rq, se);
-	check_spread(cfs_rq, se);
-	check_spread(cfs_rq, curr);
-	__enqueue_entity(cfs_rq, se);
-	account_entity_enqueue(cfs_rq, se);
+	se->peer_preempt = 0;
+	enqueue_task_fair(rq, p, 0);
 	resched_task(rq->curr);
 }
 
@@ -1023,7 +1075,8 @@ static void set_curr_task_fair(struct rq *rq)
 /*
  * All the scheduling class methods:
  */
-struct sched_class fair_sched_class __read_mostly = {
+static const struct sched_class fair_sched_class = {
+	.next			= &idle_sched_class,
 	.enqueue_task		= enqueue_task_fair,
 	.dequeue_task		= dequeue_task_fair,
 	.yield_task		= yield_task_fair,
@@ -1033,7 +1086,10 @@ struct sched_class fair_sched_class __read_mostly = {
 	.pick_next_task		= pick_next_task_fair,
 	.put_prev_task		= put_prev_task_fair,
 
+#ifdef CONFIG_SMP
 	.load_balance		= load_balance_fair,
+	.move_one_task		= move_one_task_fair,
+#endif
 
 	.set_curr_task          = set_curr_task_fair,
 	.task_tick		= task_tick_fair,