fair.c 158 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
/*
 * Completely Fair Scheduling (CFS) Class (SCHED_NORMAL/SCHED_BATCH)
 *
 *  Copyright (C) 2007 Red Hat, Inc., Ingo Molnar <mingo@redhat.com>
 *
 *  Interactivity improvements by Mike Galbraith
 *  (C) 2007 Mike Galbraith <efault@gmx.de>
 *
 *  Various enhancements by Dmitry Adamushko.
 *  (C) 2007 Dmitry Adamushko <dmitry.adamushko@gmail.com>
 *
 *  Group scheduling enhancements by Srivatsa Vaddagiri
 *  Copyright IBM Corporation, 2007
 *  Author: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
 *
 *  Scaled math optimizations by Thomas Gleixner
 *  Copyright (C) 2007, Thomas Gleixner <tglx@linutronix.de>
18 19 20
 *
 *  Adaptive scheduling granularity, math enhancements by Peter Zijlstra
 *  Copyright (C) 2007 Red Hat, Inc., Peter Zijlstra <pzijlstr@redhat.com>
21 22
 */

Arjan van de Ven's avatar
Arjan van de Ven committed
23
#include <linux/latencytop.h>
24
#include <linux/sched.h>
25
#include <linux/cpumask.h>
26 27 28
#include <linux/slab.h>
#include <linux/profile.h>
#include <linux/interrupt.h>
29
#include <linux/mempolicy.h>
30
#include <linux/migrate.h>
31
#include <linux/task_work.h>
32 33 34 35

#include <trace/events/sched.h>

#include "sched.h"
Arjan van de Ven's avatar
Arjan van de Ven committed
36

37
/*
38
 * Targeted preemption latency for CPU-bound tasks:
39
 * (default: 6ms * (1 + ilog(ncpus)), units: nanoseconds)
40
 *
41
 * NOTE: this latency value is not the same as the concept of
Ingo Molnar's avatar
Ingo Molnar committed
42 43 44
 * 'timeslice length' - timeslices in CFS are of variable length
 * and have no persistent notion like in traditional, time-slice
 * based scheduling concepts.
45
 *
Ingo Molnar's avatar
Ingo Molnar committed
46 47
 * (to see the precise effective timeslice length of your workload,
 *  run vmstat and monitor the context-switches (cs) field)
48
 */
49 50
unsigned int sysctl_sched_latency = 6000000ULL;
unsigned int normalized_sysctl_sched_latency = 6000000ULL;
51

52 53 54 55 56 57 58 59 60 61 62 63
/*
 * The initial- and re-scaling of tunables is configurable
 * (default SCHED_TUNABLESCALING_LOG = *(1+ilog(ncpus))
 *
 * Options are:
 * SCHED_TUNABLESCALING_NONE - unscaled, always *1
 * SCHED_TUNABLESCALING_LOG - scaled logarithmical, *1+ilog(ncpus)
 * SCHED_TUNABLESCALING_LINEAR - scaled linear, *ncpus
 */
enum sched_tunable_scaling sysctl_sched_tunable_scaling
	= SCHED_TUNABLESCALING_LOG;

64
/*
65
 * Minimal preemption granularity for CPU-bound tasks:
66
 * (default: 0.75 msec * (1 + ilog(ncpus)), units: nanoseconds)
67
 */
68 69
unsigned int sysctl_sched_min_granularity = 750000ULL;
unsigned int normalized_sysctl_sched_min_granularity = 750000ULL;
70 71

/*
72 73
 * is kept at sysctl_sched_latency / sysctl_sched_min_granularity
 */
74
static unsigned int sched_nr_latency = 8;
75 76

/*
77
 * After fork, child runs first. If set to 0 (default) then
78
 * parent will (try to) run first.
79
 */
80
unsigned int sysctl_sched_child_runs_first __read_mostly;
81 82 83

/*
 * SCHED_OTHER wake-up granularity.
84
 * (default: 1 msec * (1 + ilog(ncpus)), units: nanoseconds)
85 86 87 88 89
 *
 * This option delays the preemption effects of decoupled workloads
 * and reduces their over-scheduling. Synchronous workloads will still
 * have immediate wakeup/sleep latencies.
 */
90
unsigned int sysctl_sched_wakeup_granularity = 1000000UL;
91
unsigned int normalized_sysctl_sched_wakeup_granularity = 1000000UL;
92

93 94
const_debug unsigned int sysctl_sched_migration_cost = 500000UL;

95 96 97 98 99 100 101
/*
 * The exponential sliding  window over which load is averaged for shares
 * distribution.
 * (default: 10msec)
 */
unsigned int __read_mostly sysctl_sched_shares_window = 10000000UL;

102 103 104 105 106 107 108 109 110 111 112 113 114 115
#ifdef CONFIG_CFS_BANDWIDTH
/*
 * Amount of runtime to allocate from global (tg) to local (per-cfs_rq) pool
 * each time a cfs_rq requests quota.
 *
 * Note: in the case that the slice exceeds the runtime remaining (either due
 * to consumption or the quota being specified to be smaller than the slice)
 * we will always only issue the remaining available time.
 *
 * default: 5 msec, units: microseconds
  */
unsigned int sysctl_sched_cfs_bandwidth_slice = 5000UL;
#endif

116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133
static inline void update_load_add(struct load_weight *lw, unsigned long inc)
{
	lw->weight += inc;
	lw->inv_weight = 0;
}

static inline void update_load_sub(struct load_weight *lw, unsigned long dec)
{
	lw->weight -= dec;
	lw->inv_weight = 0;
}

static inline void update_load_set(struct load_weight *lw, unsigned long w)
{
	lw->weight = w;
	lw->inv_weight = 0;
}

134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237
/*
 * Increase the granularity value when there are more CPUs,
 * because with more CPUs the 'effective latency' as visible
 * to users decreases. But the relationship is not linear,
 * so pick a second-best guess by going with the log2 of the
 * number of CPUs.
 *
 * This idea comes from the SD scheduler of Con Kolivas:
 */
static int get_update_sysctl_factor(void)
{
	unsigned int cpus = min_t(int, num_online_cpus(), 8);
	unsigned int factor;

	switch (sysctl_sched_tunable_scaling) {
	case SCHED_TUNABLESCALING_NONE:
		factor = 1;
		break;
	case SCHED_TUNABLESCALING_LINEAR:
		factor = cpus;
		break;
	case SCHED_TUNABLESCALING_LOG:
	default:
		factor = 1 + ilog2(cpus);
		break;
	}

	return factor;
}

static void update_sysctl(void)
{
	unsigned int factor = get_update_sysctl_factor();

#define SET_SYSCTL(name) \
	(sysctl_##name = (factor) * normalized_sysctl_##name)
	SET_SYSCTL(sched_min_granularity);
	SET_SYSCTL(sched_latency);
	SET_SYSCTL(sched_wakeup_granularity);
#undef SET_SYSCTL
}

void sched_init_granularity(void)
{
	update_sysctl();
}

#if BITS_PER_LONG == 32
# define WMULT_CONST	(~0UL)
#else
# define WMULT_CONST	(1UL << 32)
#endif

#define WMULT_SHIFT	32

/*
 * Shift right and round:
 */
#define SRR(x, y) (((x) + (1UL << ((y) - 1))) >> (y))

/*
 * delta *= weight / lw
 */
static unsigned long
calc_delta_mine(unsigned long delta_exec, unsigned long weight,
		struct load_weight *lw)
{
	u64 tmp;

	/*
	 * weight can be less than 2^SCHED_LOAD_RESOLUTION for task group sched
	 * entities since MIN_SHARES = 2. Treat weight as 1 if less than
	 * 2^SCHED_LOAD_RESOLUTION.
	 */
	if (likely(weight > (1UL << SCHED_LOAD_RESOLUTION)))
		tmp = (u64)delta_exec * scale_load_down(weight);
	else
		tmp = (u64)delta_exec;

	if (!lw->inv_weight) {
		unsigned long w = scale_load_down(lw->weight);

		if (BITS_PER_LONG > 32 && unlikely(w >= WMULT_CONST))
			lw->inv_weight = 1;
		else if (unlikely(!w))
			lw->inv_weight = WMULT_CONST;
		else
			lw->inv_weight = WMULT_CONST / w;
	}

	/*
	 * Check whether we'd overflow the 64-bit multiplication:
	 */
	if (unlikely(tmp > WMULT_CONST))
		tmp = SRR(SRR(tmp, WMULT_SHIFT/2) * lw->inv_weight,
			WMULT_SHIFT/2);
	else
		tmp = SRR(tmp * lw->inv_weight, WMULT_SHIFT);

	return (unsigned long)min(tmp, (u64)(unsigned long)LONG_MAX);
}


const struct sched_class fair_sched_class;
238

239 240 241 242
/**************************************************************
 * CFS operations on generic schedulable entities:
 */

243
#ifdef CONFIG_FAIR_GROUP_SCHED
244

245
/* cpu runqueue to which this cfs_rq is attached */
246 247
static inline struct rq *rq_of(struct cfs_rq *cfs_rq)
{
248
	return cfs_rq->rq;
249 250
}

251 252
/* An entity is a task if it doesn't "own" a runqueue */
#define entity_is_task(se)	(!se->my_q)
253

254 255 256 257 258 259 260 261
static inline struct task_struct *task_of(struct sched_entity *se)
{
#ifdef CONFIG_SCHED_DEBUG
	WARN_ON_ONCE(!entity_is_task(se));
#endif
	return container_of(se, struct task_struct, se);
}

262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282
/* Walk up scheduling entities hierarchy */
#define for_each_sched_entity(se) \
		for (; se; se = se->parent)

static inline struct cfs_rq *task_cfs_rq(struct task_struct *p)
{
	return p->se.cfs_rq;
}

/* runqueue on which this entity is (to be) queued */
static inline struct cfs_rq *cfs_rq_of(struct sched_entity *se)
{
	return se->cfs_rq;
}

/* runqueue "owned" by this group */
static inline struct cfs_rq *group_cfs_rq(struct sched_entity *grp)
{
	return grp->my_q;
}

283 284
static void update_cfs_rq_blocked_load(struct cfs_rq *cfs_rq,
				       int force_update);
285

286 287 288
static inline void list_add_leaf_cfs_rq(struct cfs_rq *cfs_rq)
{
	if (!cfs_rq->on_list) {
289 290 291 292 293 294 295 296 297 298 299 300
		/*
		 * Ensure we either appear before our parent (if already
		 * enqueued) or force our parent to appear after us when it is
		 * enqueued.  The fact that we always enqueue bottom-up
		 * reduces this to two cases.
		 */
		if (cfs_rq->tg->parent &&
		    cfs_rq->tg->parent->cfs_rq[cpu_of(rq_of(cfs_rq))]->on_list) {
			list_add_rcu(&cfs_rq->leaf_cfs_rq_list,
				&rq_of(cfs_rq)->leaf_cfs_rq_list);
		} else {
			list_add_tail_rcu(&cfs_rq->leaf_cfs_rq_list,
301
				&rq_of(cfs_rq)->leaf_cfs_rq_list);
302
		}
303 304

		cfs_rq->on_list = 1;
305
		/* We should have no load, but we need to update last_decay. */
306
		update_cfs_rq_blocked_load(cfs_rq, 0);
307 308 309 310 311 312 313 314 315 316 317
	}
}

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;
	}
}

318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336
/* 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)

/* 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 (se->cfs_rq == pse->cfs_rq)
		return 1;

	return 0;
}

static inline struct sched_entity *parent_entity(struct sched_entity *se)
{
	return se->parent;
}

337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379
/* return depth at which a sched entity is present in the hierarchy */
static inline int depth_se(struct sched_entity *se)
{
	int depth = 0;

	for_each_sched_entity(se)
		depth++;

	return depth;
}

static void
find_matching_se(struct sched_entity **se, struct sched_entity **pse)
{
	int se_depth, pse_depth;

	/*
	 * preemption test can be made between sibling entities who are in the
	 * same cfs_rq i.e who have a common parent. Walk up the hierarchy of
	 * both tasks until we find their ancestors who are siblings of common
	 * parent.
	 */

	/* First walk up until both entities are at same depth */
	se_depth = depth_se(*se);
	pse_depth = depth_se(*pse);

	while (se_depth > pse_depth) {
		se_depth--;
		*se = parent_entity(*se);
	}

	while (pse_depth > se_depth) {
		pse_depth--;
		*pse = parent_entity(*pse);
	}

	while (!is_same_group(*se, *pse)) {
		*se = parent_entity(*se);
		*pse = parent_entity(*pse);
	}
}

380 381 382 383 384 385
#else	/* !CONFIG_FAIR_GROUP_SCHED */

static inline struct task_struct *task_of(struct sched_entity *se)
{
	return container_of(se, struct task_struct, se);
}
386

387 388 389
static inline struct rq *rq_of(struct cfs_rq *cfs_rq)
{
	return container_of(cfs_rq, struct rq, cfs);
390 391 392 393
}

#define entity_is_task(se)	1

394 395
#define for_each_sched_entity(se) \
		for (; se; se = NULL)
396

397
static inline struct cfs_rq *task_cfs_rq(struct task_struct *p)
398
{
399
	return &task_rq(p)->cfs;
400 401
}

402 403 404 405 406 407 408 409 410 411 412 413 414 415
static inline struct cfs_rq *cfs_rq_of(struct sched_entity *se)
{
	struct task_struct *p = task_of(se);
	struct rq *rq = task_rq(p);

	return &rq->cfs;
}

/* runqueue "owned" by this group */
static inline struct cfs_rq *group_cfs_rq(struct sched_entity *grp)
{
	return NULL;
}

416 417 418 419 420 421 422 423
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)
{
}

424 425 426 427 428 429 430 431 432 433 434 435 436 437
#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 sched_entity *se, struct sched_entity *pse)
{
	return 1;
}

static inline struct sched_entity *parent_entity(struct sched_entity *se)
{
	return NULL;
}

438 439 440 441 442
static inline void
find_matching_se(struct sched_entity **se, struct sched_entity **pse)
{
}

443 444
#endif	/* CONFIG_FAIR_GROUP_SCHED */

445 446
static __always_inline
void account_cfs_rq_runtime(struct cfs_rq *cfs_rq, unsigned long delta_exec);
447 448 449 450 451

/**************************************************************
 * Scheduling class tree data structure manipulation methods:
 */

452
static inline u64 max_vruntime(u64 max_vruntime, u64 vruntime)
453
{
454
	s64 delta = (s64)(vruntime - max_vruntime);
455
	if (delta > 0)
456
		max_vruntime = vruntime;
457

458
	return max_vruntime;
459 460
}

461
static inline u64 min_vruntime(u64 min_vruntime, u64 vruntime)
462 463 464 465 466 467 468 469
{
	s64 delta = (s64)(vruntime - min_vruntime);
	if (delta < 0)
		min_vruntime = vruntime;

	return min_vruntime;
}

470 471 472 473 474 475
static inline int entity_before(struct sched_entity *a,
				struct sched_entity *b)
{
	return (s64)(a->vruntime - b->vruntime) < 0;
}

476 477 478 479 480 481 482 483 484 485 486 487
static void update_min_vruntime(struct cfs_rq *cfs_rq)
{
	u64 vruntime = cfs_rq->min_vruntime;

	if (cfs_rq->curr)
		vruntime = cfs_rq->curr->vruntime;

	if (cfs_rq->rb_leftmost) {
		struct sched_entity *se = rb_entry(cfs_rq->rb_leftmost,
						   struct sched_entity,
						   run_node);

488
		if (!cfs_rq->curr)
489 490 491 492 493
			vruntime = se->vruntime;
		else
			vruntime = min_vruntime(vruntime, se->vruntime);
	}

494
	/* ensure we never gain time by being placed backwards. */
495
	cfs_rq->min_vruntime = max_vruntime(cfs_rq->min_vruntime, vruntime);
496 497 498 499
#ifndef CONFIG_64BIT
	smp_wmb();
	cfs_rq->min_vruntime_copy = cfs_rq->min_vruntime;
#endif
500 501
}

502 503 504
/*
 * Enqueue an entity into the rb-tree:
 */
505
static void __enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521
{
	struct rb_node **link = &cfs_rq->tasks_timeline.rb_node;
	struct rb_node *parent = NULL;
	struct sched_entity *entry;
	int leftmost = 1;

	/*
	 * Find the right place in the rbtree:
	 */
	while (*link) {
		parent = *link;
		entry = rb_entry(parent, struct sched_entity, run_node);
		/*
		 * We dont care about collisions. Nodes with
		 * the same key stay together.
		 */
522
		if (entity_before(se, entry)) {
523 524 525 526 527 528 529 530 531 532 533
			link = &parent->rb_left;
		} else {
			link = &parent->rb_right;
			leftmost = 0;
		}
	}

	/*
	 * Maintain a cache of leftmost tree entries (it is frequently
	 * used):
	 */
534
	if (leftmost)
Ingo Molnar's avatar
Ingo Molnar committed
535
		cfs_rq->rb_leftmost = &se->run_node;
536 537 538 539 540

	rb_link_node(&se->run_node, parent, link);
	rb_insert_color(&se->run_node, &cfs_rq->tasks_timeline);
}

541
static void __dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
542
{
Peter Zijlstra's avatar
Peter Zijlstra committed
543 544 545 546 547 548
	if (cfs_rq->rb_leftmost == &se->run_node) {
		struct rb_node *next_node;

		next_node = rb_next(&se->run_node);
		cfs_rq->rb_leftmost = next_node;
	}
Ingo Molnar's avatar
Ingo Molnar committed
549

550 551 552
	rb_erase(&se->run_node, &cfs_rq->tasks_timeline);
}

553
struct sched_entity *__pick_first_entity(struct cfs_rq *cfs_rq)
554
{
555 556 557 558 559 560
	struct rb_node *left = cfs_rq->rb_leftmost;

	if (!left)
		return NULL;

	return rb_entry(left, struct sched_entity, run_node);
561 562
}

563 564 565 566 567 568 569 570 571 572 573
static struct sched_entity *__pick_next_entity(struct sched_entity *se)
{
	struct rb_node *next = rb_next(&se->run_node);

	if (!next)
		return NULL;

	return rb_entry(next, struct sched_entity, run_node);
}

#ifdef CONFIG_SCHED_DEBUG
574
struct sched_entity *__pick_last_entity(struct cfs_rq *cfs_rq)
575
{
576
	struct rb_node *last = rb_last(&cfs_rq->tasks_timeline);
577

578 579
	if (!last)
		return NULL;
580 581

	return rb_entry(last, struct sched_entity, run_node);
582 583
}

584 585 586 587
/**************************************************************
 * Scheduling class statistics methods:
 */

588
int sched_proc_update_handler(struct ctl_table *table, int write,
589
		void __user *buffer, size_t *lenp,
590 591
		loff_t *ppos)
{
592
	int ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
593
	int factor = get_update_sysctl_factor();
594 595 596 597 598 599 600

	if (ret || !write)
		return ret;

	sched_nr_latency = DIV_ROUND_UP(sysctl_sched_latency,
					sysctl_sched_min_granularity);

601 602 603 604 605 606 607
#define WRT_SYSCTL(name) \
	(normalized_sysctl_##name = sysctl_##name / (factor))
	WRT_SYSCTL(sched_min_granularity);
	WRT_SYSCTL(sched_latency);
	WRT_SYSCTL(sched_wakeup_granularity);
#undef WRT_SYSCTL

608 609 610
	return 0;
}
#endif
611

612
/*
613
 * delta /= w
614 615 616 617
 */
static inline unsigned long
calc_delta_fair(unsigned long delta, struct sched_entity *se)
{
618 619
	if (unlikely(se->load.weight != NICE_0_LOAD))
		delta = calc_delta_mine(delta, NICE_0_LOAD, &se->load);
620 621 622 623

	return delta;
}

624 625 626
/*
 * The idea is to set a period in which each task runs once.
 *
627
 * When there are too many tasks (sched_nr_latency) we have to stretch
628 629 630 631
 * this period because otherwise the slices get too small.
 *
 * p = (nr <= nl) ? l : l*nr/nl
 */
632 633 634
static u64 __sched_period(unsigned long nr_running)
{
	u64 period = sysctl_sched_latency;
635
	unsigned long nr_latency = sched_nr_latency;
636 637

	if (unlikely(nr_running > nr_latency)) {
638
		period = sysctl_sched_min_granularity;
639 640 641 642 643 644
		period *= nr_running;
	}

	return period;
}

645 646 647 648
/*
 * We calculate the wall-time slice from the period by taking a part
 * proportional to the weight.
 *
649
 * s = p*P[w/rw]
650
 */
651
static u64 sched_slice(struct cfs_rq *cfs_rq, struct sched_entity *se)
652
{
Mike Galbraith's avatar
Mike Galbraith committed
653
	u64 slice = __sched_period(cfs_rq->nr_running + !se->on_rq);
654

Mike Galbraith's avatar
Mike Galbraith committed
655
	for_each_sched_entity(se) {
Lin Ming's avatar
Lin Ming committed
656
		struct load_weight *load;
657
		struct load_weight lw;
Lin Ming's avatar
Lin Ming committed
658 659 660

		cfs_rq = cfs_rq_of(se);
		load = &cfs_rq->load;
661

Mike Galbraith's avatar
Mike Galbraith committed
662
		if (unlikely(!se->on_rq)) {
663
			lw = cfs_rq->load;
Mike Galbraith's avatar
Mike Galbraith committed
664 665 666 667 668 669 670

			update_load_add(&lw, se->load.weight);
			load = &lw;
		}
		slice = calc_delta_mine(slice, se->load.weight, load);
	}
	return slice;
671 672
}

673
/*
Andrei Epure's avatar
Andrei Epure committed
674
 * We calculate the vruntime slice of a to-be-inserted task.
675
 *
676
 * vs = s/w
677
 */
678
static u64 sched_vslice(struct cfs_rq *cfs_rq, struct sched_entity *se)
Peter Zijlstra's avatar
Peter Zijlstra committed
679
{
680
	return calc_delta_fair(sched_slice(cfs_rq, se), se);
681 682
}

683 684 685 686 687
/*
 * Update the current task's runtime statistics. Skip current tasks that
 * are not in our scheduling class.
 */
static inline void
Ingo Molnar's avatar
Ingo Molnar committed
688 689
__update_curr(struct cfs_rq *cfs_rq, struct sched_entity *curr,
	      unsigned long delta_exec)
690
{
691
	unsigned long delta_exec_weighted;
692

693 694
	schedstat_set(curr->statistics.exec_max,
		      max((u64)delta_exec, curr->statistics.exec_max));
695 696

	curr->sum_exec_runtime += delta_exec;
697
	schedstat_add(cfs_rq, exec_clock, delta_exec);
698
	delta_exec_weighted = calc_delta_fair(delta_exec, curr);
699

Ingo Molnar's avatar
Ingo Molnar committed
700
	curr->vruntime += delta_exec_weighted;
701
	update_min_vruntime(cfs_rq);
702 703
}

704
static void update_curr(struct cfs_rq *cfs_rq)
705
{
706
	struct sched_entity *curr = cfs_rq->curr;
707
	u64 now = rq_clock_task(rq_of(cfs_rq));
708 709 710 711 712 713 714 715 716 717
	unsigned long delta_exec;

	if (unlikely(!curr))
		return;

	/*
	 * Get the amount of time the current task was running
	 * since the last time we changed load (this cannot
	 * overflow on 32 bits):
	 */
Ingo Molnar's avatar
Ingo Molnar committed
718
	delta_exec = (unsigned long)(now - curr->exec_start);
719 720
	if (!delta_exec)
		return;
721

Ingo Molnar's avatar
Ingo Molnar committed
722 723
	__update_curr(cfs_rq, curr, delta_exec);
	curr->exec_start = now;
724 725 726 727

	if (entity_is_task(curr)) {
		struct task_struct *curtask = task_of(curr);

728
		trace_sched_stat_runtime(curtask, delta_exec, curr->vruntime);
729
		cpuacct_charge(curtask, delta_exec);
730
		account_group_exec_runtime(curtask, delta_exec);
731
	}
732 733

	account_cfs_rq_runtime(cfs_rq, delta_exec);
734 735 736
}

static inline void
737
update_stats_wait_start(struct cfs_rq *cfs_rq, struct sched_entity *se)
738
{
739
	schedstat_set(se->statistics.wait_start, rq_clock(rq_of(cfs_rq)));
740 741 742 743 744
}

/*
 * Task is being enqueued - update stats:
 */
745
static void update_stats_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se)
746 747 748 749 750
{
	/*
	 * Are we enqueueing a waiting task? (for current tasks
	 * a dequeue/enqueue event is a NOP)
	 */
751
	if (se != cfs_rq->curr)
752
		update_stats_wait_start(cfs_rq, se);
753 754 755
}

static void
756
update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se)
757
{
758
	schedstat_set(se->statistics.wait_max, max(se->statistics.wait_max,
759
			rq_clock(rq_of(cfs_rq)) - se->statistics.wait_start));
760 761
	schedstat_set(se->statistics.wait_count, se->statistics.wait_count + 1);
	schedstat_set(se->statistics.wait_sum, se->statistics.wait_sum +
762
			rq_clock(rq_of(cfs_rq)) - se->statistics.wait_start);
763 764 765
#ifdef CONFIG_SCHEDSTATS
	if (entity_is_task(se)) {
		trace_sched_stat_wait(task_of(se),
766
			rq_clock(rq_of(cfs_rq)) - se->statistics.wait_start);
767 768
	}
#endif
769
	schedstat_set(se->statistics.wait_start, 0);
770 771 772
}

static inline void
773
update_stats_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se)
774 775 776 777 778
{
	/*
	 * Mark the end of the wait period if dequeueing a
	 * waiting task:
	 */
779
	if (se != cfs_rq->curr)
780
		update_stats_wait_end(cfs_rq, se);
781 782 783 784 785 786
}

/*
 * We are picking a new current task - update its stats:
 */
static inline void
787
update_stats_curr_start(struct cfs_rq *cfs_rq, struct sched_entity *se)
788 789 790 791
{
	/*
	 * We are starting a new run period:
	 */
792
	se->exec_start = rq_clock_task(rq_of(cfs_rq));
793 794 795 796 797 798
}

/**************************************************
 * Scheduling class queueing methods:
 */

799 800
#ifdef CONFIG_NUMA_BALANCING
/*
801
 * numa task sample period in ms
802
 */
803
unsigned int sysctl_numa_balancing_scan_period_min = 100;
804 805
unsigned int sysctl_numa_balancing_scan_period_max = 100*50;
unsigned int sysctl_numa_balancing_scan_period_reset = 100*600;
806 807 808

/* Portion of address space to scan in MB */
unsigned int sysctl_numa_balancing_scan_size = 256;
809

810 811 812
/* Scan @scan_size MB every @scan_period after an initial @scan_delay in ms */
unsigned int sysctl_numa_balancing_scan_delay = 1000;

813 814
static void task_numa_placement(struct task_struct *p)
{
815
	int seq;
816

817 818 819
	if (!p->mm)	/* for example, ksmd faulting in a user's mm */
		return;
	seq = ACCESS_ONCE(p->mm->numa_scan_seq);
820 821 822 823 824 825 826 827 828 829
	if (p->numa_scan_seq == seq)
		return;
	p->numa_scan_seq = seq;

	/* FIXME: Scheduling placement policy hints go here */
}

/*
 * Got a PROT_NONE fault for a page on @node.
 */
830
void task_numa_fault(int node, int pages, bool migrated)
831 832 833
{
	struct task_struct *p = current;

834 835 836
	if (!sched_feat_numa(NUMA))
		return;

837 838
	/* FIXME: Allocate task-specific structure for placement policy here */

839
	/*
840 841
	 * If pages are properly placed (did not migrate) then scan slower.
	 * This is reset periodically in case of phase changes
842
	 */
843 844 845
        if (!migrated)
		p->numa_scan_period = min(sysctl_numa_balancing_scan_period_max,
			p->numa_scan_period + jiffies_to_msecs(10));
846

847 848 849
	task_numa_placement(p);
}

850 851 852 853 854 855
static void reset_ptenuma_scan(struct task_struct *p)
{
	ACCESS_ONCE(p->mm->numa_scan_seq)++;
	p->mm->numa_scan_offset = 0;
}

856 857 858 859 860 861 862 863 864
/*
 * The expensive part of numa migration is done from task_work context.
 * Triggered from task_tick_numa().
 */
void task_numa_work(struct callback_head *work)
{
	unsigned long migrate, next_scan, now = jiffies;
	struct task_struct *p = current;
	struct mm_struct *mm = p->mm;
865
	struct vm_area_struct *vma;
866 867
	unsigned long start, end;
	long pages;
868 869 870 871 872 873 874 875 876 877 878 879 880 881 882

	WARN_ON_ONCE(p != container_of(work, struct task_struct, numa_work));

	work->next = work; /* protect against double add */
	/*
	 * Who cares about NUMA placement when they're dying.
	 *
	 * NOTE: make sure not to dereference p->mm before this check,
	 * exit_task_work() happens _after_ exit_mm() so we could be called
	 * without p->mm even though we still had it when we enqueued this
	 * work.
	 */
	if (p->flags & PF_EXITING)
		return;

883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900
	/*
	 * We do not care about task placement until a task runs on a node
	 * other than the first one used by the address space. This is
	 * largely because migrations are driven by what CPU the task
	 * is running on. If it's never scheduled on another node, it'll
	 * not migrate so why bother trapping the fault.
	 */
	if (mm->first_nid == NUMA_PTE_SCAN_INIT)
		mm->first_nid = numa_node_id();
	if (mm->first_nid != NUMA_PTE_SCAN_ACTIVE) {
		/* Are we running on a new node yet? */
		if (numa_node_id() == mm->first_nid &&
		    !sched_feat_numa(NUMA_FORCE))
			return;

		mm->first_nid = NUMA_PTE_SCAN_ACTIVE;
	}

901 902 903 904 905 906 907 908 909 910 911 912 913
	/*
	 * Reset the scan period if enough time has gone by. Objective is that
	 * scanning will be reduced if pages are properly placed. As tasks
	 * can enter different phases this needs to be re-examined. Lacking
	 * proper tracking of reference behaviour, this blunt hammer is used.
	 */
	migrate = mm->numa_next_reset;
	if (time_after(now, migrate)) {
		p->numa_scan_period = sysctl_numa_balancing_scan_period_min;
		next_scan = now + msecs_to_jiffies(sysctl_numa_balancing_scan_period_reset);
		xchg(&mm->numa_next_reset, next_scan);
	}

914 915 916 917 918 919 920 921 922 923
	/*
	 * Enforce maximal scan/migration frequency..
	 */
	migrate = mm->numa_next_scan;
	if (time_before(now, migrate))
		return;

	if (p->numa_scan_period == 0)
		p->numa_scan_period = sysctl_numa_balancing_scan_period_min;

924
	next_scan = now + msecs_to_jiffies(p->numa_scan_period);
925 926 927
	if (cmpxchg(&mm->numa_next_scan, migrate, next_scan) != migrate)
		return;

928 929 930 931 932 933 934 935
	/*
	 * Do not set pte_numa if the current running node is rate-limited.
	 * This loses statistics on the fault but if we are unwilling to
	 * migrate to this node, it is less likely we can do useful work
	 */
	if (migrate_ratelimited(numa_node_id()))
		return;

936 937 938 939 940
	start = mm->numa_scan_offset;
	pages = sysctl_numa_balancing_scan_size;
	pages <<= 20 - PAGE_SHIFT; /* MB in pages */
	if (!pages)
		return;
941

942
	down_read(&mm->mmap_sem);
943
	vma = find_vma(mm, start);
944 945
	if (!vma) {
		reset_ptenuma_scan(p);
946
		start = 0;
947 948
		vma = mm->mmap;
	}
949
	for (; vma; vma = vma->vm_next) {
950 951 952 953
		if (!vma_migratable(vma))
			continue;

		/* Skip small VMAs. They are not likely to be of relevance */
954
		if (vma->vm_end - vma->vm_start < HPAGE_SIZE)
955 956
			continue;

957 958 959 960 961
		do {
			start = max(start, vma->vm_start);
			end = ALIGN(start + (pages << PAGE_SHIFT), HPAGE_SIZE);
			end = min(end, vma->vm_end);
			pages -= change_prot_numa(vma, start, end);
962

963 964 965 966
			start = end;
			if (pages <= 0)
				goto out;
		} while (end != vma->vm_end);
967
	}
968

969
out:
970 971 972 973 974 975 976
	/*
	 * It is possible to reach the end of the VMA list but the last few VMAs are
	 * not guaranteed to the vma_migratable. If they are not, we would find the
	 * !migratable VMA on the next scan but not reset the scanner to the start
	 * so check it now.
	 */
	if (vma)
977
		mm->numa_scan_offset = start;
978 979 980
	else
		reset_ptenuma_scan(p);
	up_read(&mm->mmap_sem);
981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006
}

/*
 * Drive the periodic memory faults..
 */
void task_tick_numa(struct rq *rq, struct task_struct *curr)
{
	struct callback_head *work = &curr->numa_work;
	u64 period, now;

	/*
	 * We don't care about NUMA placement if we don't have memory.
	 */
	if (!curr->mm || (curr->flags & PF_EXITING) || work->next != work)
		return;

	/*
	 * Using runtime rather than walltime has the dual advantage that
	 * we (mostly) drive the selection from busy threads and that the
	 * task needs to have done some actual work before we bother with
	 * NUMA placement.
	 */
	now = curr->se.sum_exec_runtime;
	period = (u64)curr->numa_scan_period * NSEC_PER_MSEC;

	if (now - curr->node_stamp > period) {
1007 1008
		if (!curr->node_stamp)
			curr->numa_scan_period = sysctl_numa_balancing_scan_period_min;
1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022
		curr->node_stamp = now;

		if (!time_before(jiffies, curr->mm->numa_next_scan)) {
			init_task_work(work, task_numa_work); /* TODO: move this into sched_fork() */
			task_work_add(curr, work, true);
		}
	}
}
#else
static void task_tick_numa(struct rq *rq, struct task_struct *curr)
{
}
#endif /* CONFIG_NUMA_BALANCING */

1023 1024 1025 1026
static void
account_entity_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se)
{
	update_load_add(&cfs_rq->load, se->load.weight);
1027
	if (!parent_entity(se))
1028
		update_load_add(&rq_of(cfs_rq)->load, se->load.weight);
1029 1030
#ifdef CONFIG_SMP
	if (entity_is_task(se))
1031
		list_add(&se->group_node, &rq_of(cfs_rq)->cfs_tasks);
1032
#endif
1033 1034 1035 1036 1037 1038 1039
	cfs_rq->nr_running++;
}

static void
account_entity_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se)
{
	update_load_sub(&cfs_rq->load, se->load.weight);
1040
	if (!parent_entity(se))
1041
		update_load_sub(&rq_of(cfs_rq)->load, se->load.weight);
1042
	if (entity_is_task(se))
1043
		list_del_init(&se->group_node);
1044 1045 1046
	cfs_rq->nr_running--;
}

1047 1048
#ifdef CONFIG_FAIR_GROUP_SCHED
# ifdef CONFIG_SMP
1049 1050 1051 1052 1053 1054 1055 1056 1057
static inline long calc_tg_weight(struct task_group *tg, struct cfs_rq *cfs_rq)
{
	long tg_weight;

	/*
	 * Use this CPU's actual weight instead of the last load_contribution
	 * to gain a more accurate current total weight. See
	 * update_cfs_rq_load_contribution().
	 */
1058 1059
	tg_weight = atomic64_read(&tg->load_avg);
	tg_weight -= cfs_rq->tg_load_contrib;
1060 1061 1062 1063 1064
	tg_weight += cfs_rq->load.weight;

	return tg_weight;
}

1065
static long calc_cfs_shares(struct cfs_rq *cfs_rq, struct task_group *tg)
1066
{
1067
	long tg_weight, load, shares;
1068

1069
	tg_weight = calc_tg_weight(tg, cfs_rq);
1070
	load = cfs_rq->load.weight;
1071 1072

	shares = (tg->shares * load);
1073 1074
	if (tg_weight)
		shares /= tg_weight;
1075 1076 1077 1078 1079 1080 1081 1082 1083

	if (shares < MIN_SHARES)
		shares = MIN_SHARES;
	if (shares > tg->shares)
		shares = tg->shares;

	return shares;
}
# else /* CONFIG_SMP */
1084
static inline long calc_cfs_shares(struct cfs_rq *cfs_rq, struct task_group *tg)
1085 1086 1087 1088
{
	return tg->shares;
}
# endif /* CONFIG_SMP */
1089 1090 1091
static void reweight_entity(struct cfs_rq *cfs_rq, struct sched_entity *se,
			    unsigned long weight)
{
1092 1093 1094 1095
	if (se->on_rq) {
		/* commit outstanding execution time */
		if (cfs_rq->curr == se)
			update_curr(cfs_rq);
1096
		account_entity_dequeue(cfs_rq, se);
1097
	}
1098 1099 1100 1101 1102 1103 1104

	update_load_set(&se->load, weight);

	if (se->on_rq)
		account_entity_enqueue(cfs_rq, se);
}

1105 1106
static inline int throttled_hierarchy(struct cfs_rq *cfs_rq);

1107
static void update_cfs_shares(struct cfs_rq *cfs_rq)
1108 1109 1110
{
	struct task_group *tg;
	struct sched_entity *se;
1111
	long shares;
1112 1113 1114

	tg = cfs_rq->tg;
	se = tg->se[cpu_of(rq_of(cfs_rq))];
1115
	if (!se || throttled_hierarchy(cfs_rq))
1116
		return;
1117 1118 1119 1120
#ifndef CONFIG_SMP
	if (likely(se->load.weight == tg->shares))
		return;
#endif
1121
	shares = calc_cfs_shares(cfs_rq, tg);
1122 1123 1124 1125

	reweight_entity(cfs_rq_of(se), se, shares);
}
#else /* CONFIG_FAIR_GROUP_SCHED */
1126
static inline void update_cfs_shares(struct cfs_rq *cfs_rq)
1127 1128 1129 1130
{
}
#endif /* CONFIG_FAIR_GROUP_SCHED */

1131
#ifdef CONFIG_SMP
1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159
/*
 * We choose a half-life close to 1 scheduling period.
 * Note: The tables below are dependent on this value.
 */
#define LOAD_AVG_PERIOD 32
#define LOAD_AVG_MAX 47742 /* maximum possible load avg */
#define LOAD_AVG_MAX_N 345 /* number of full periods to produce LOAD_MAX_AVG */

/* Precomputed fixed inverse multiplies for multiplication by y^n */
static const u32 runnable_avg_yN_inv[] = {
	0xffffffff, 0xfa83b2da, 0xf5257d14, 0xefe4b99a, 0xeac0c6e6, 0xe5b906e6,
	0xe0ccdeeb, 0xdbfbb796, 0xd744fcc9, 0xd2a81d91, 0xce248c14, 0xc9b9bd85,
	0xc5672a10, 0xc12c4cc9, 0xbd08a39e, 0xb8fbaf46, 0xb504f333, 0xb123f581,
	0xad583ee9, 0xa9a15ab4, 0xa5fed6a9, 0xa2704302, 0x9ef5325f, 0x9b8d39b9,
	0x9837f050, 0x94f4efa8, 0x91c3d373, 0x8ea4398a, 0x8b95c1e3, 0x88980e80,
	0x85aac367, 0x82cd8698,
};

/*
 * Precomputed \Sum y^k { 1<=k<=n }.  These are floor(true_value) to prevent
 * over-estimates when re-combining.
 */
static const u32 runnable_avg_yN_sum[] = {
	    0, 1002, 1982, 2941, 3880, 4798, 5697, 6576, 7437, 8279, 9103,
	 9909,10698,11470,12226,12966,13690,14398,15091,15769,16433,17082,
	17718,18340,18949,19545,20128,20698,21256,21802,22336,22859,23371,
};

1160 1161 1162 1163 1164 1165
/*
 * Approximate:
 *   val * y^n,    where y^32 ~= 0.5 (~1 scheduling period)
 */
static __always_inline u64 decay_load(u64 val, u64 n)
{
1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185
	unsigned int local_n;

	if (!n)
		return val;
	else if (unlikely(n > LOAD_AVG_PERIOD * 63))
		return 0;

	/* after bounds checking we can collapse to 32-bit */
	local_n = n;

	/*
	 * As y^PERIOD = 1/2, we can combine
	 *    y^n = 1/2^(n/PERIOD) * k^(n%PERIOD)
	 * With a look-up table which covers k^n (n<PERIOD)
	 *
	 * To achieve constant time decay_load.
	 */
	if (unlikely(local_n >= LOAD_AVG_PERIOD)) {
		val >>= local_n / LOAD_AVG_PERIOD;
		local_n %= LOAD_AVG_PERIOD;
1186 1187
	}

1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218
	val *= runnable_avg_yN_inv[local_n];
	/* We don't use SRR here since we always want to round down. */
	return val >> 32;
}

/*
 * For updates fully spanning n periods, the contribution to runnable
 * average will be: \Sum 1024*y^n
 *
 * We can compute this reasonably efficiently by combining:
 *   y^PERIOD = 1/2 with precomputed \Sum 1024*y^n {for  n <PERIOD}
 */
static u32 __compute_runnable_contrib(u64 n)
{
	u32 contrib = 0;

	if (likely(n <= LOAD_AVG_PERIOD))
		return runnable_avg_yN_sum[n];
	else if (unlikely(n >= LOAD_AVG_MAX_N))
		return LOAD_AVG_MAX;

	/* Compute \Sum k^n combining precomputed values for k^i, \Sum k^j */
	do {
		contrib /= 2; /* y^LOAD_AVG_PERIOD = 1/2 */
		contrib += runnable_avg_yN_sum[LOAD_AVG_PERIOD];

		n -= LOAD_AVG_PERIOD;
	} while (n > LOAD_AVG_PERIOD);

	contrib = decay_load(contrib, n);
	return contrib + runnable_avg_yN_sum[n];
1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252
}

/*
 * We can represent the historical contribution to runnable average as the
 * coefficients of a geometric series.  To do this we sub-divide our runnable
 * history into segments of approximately 1ms (1024us); label the segment that
 * occurred N-ms ago p_N, with p_0 corresponding to the current period, e.g.
 *
 * [<- 1024us ->|<- 1024us ->|<- 1024us ->| ...
 *      p0            p1           p2
 *     (now)       (~1ms ago)  (~2ms ago)
 *
 * Let u_i denote the fraction of p_i that the entity was runnable.
 *
 * We then designate the fractions u_i as our co-efficients, yielding the
 * following representation of historical load:
 *   u_0 + u_1*y + u_2*y^2 + u_3*y^3 + ...
 *
 * We choose y based on the with of a reasonably scheduling period, fixing:
 *   y^32 = 0.5
 *
 * This means that the contribution to load ~32ms ago (u_32) will be weighted
 * approximately half as much as the contribution to load within the last ms
 * (u_0).
 *
 * When a period "rolls over" and we have new u_0`, multiplying the previous
 * sum again by y is sufficient to update:
 *   load_avg = u_0` + y*(u_0 + u_1*y + u_2*y^2 + ... )
 *            = u_0 + u_1*y + u_2*y^2 + ... [re-labeling u_i --> u_{i+1}]
 */
static __always_inline int __update_entity_runnable_avg(u64 now,
							struct sched_avg *sa,
							int runnable)
{
1253 1254
	u64 delta, periods;
	u32 runnable_contrib;
1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287
	int delta_w, decayed = 0;

	delta = now - sa->last_runnable_update;
	/*
	 * This should only happen when time goes backwards, which it
	 * unfortunately does during sched clock init when we swap over to TSC.
	 */
	if ((s64)delta < 0) {
		sa->last_runnable_update = now;
		return 0;
	}

	/*
	 * Use 1024ns as the unit of measurement since it's a reasonable
	 * approximation of 1us and fast to compute.
	 */
	delta >>= 10;
	if (!delta)
		return 0;
	sa->last_runnable_update = now;

	/* delta_w is the amount already accumulated against our next period */
	delta_w = sa->runnable_avg_period % 1024;
	if (delta + delta_w >= 1024) {
		/* period roll-over */
		decayed = 1;

		/*
		 * Now that we know we're crossing a period boundary, figure
		 * out how much from delta we need to complete the current
		 * period and accrue it.
		 */
		delta_w = 1024 - delta_w;
1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307
		if (runnable)
			sa->runnable_avg_sum += delta_w;
		sa->runnable_avg_period += delta_w;

		delta -= delta_w;

		/* Figure out how many additional periods this update spans */
		periods = delta / 1024;
		delta %= 1024;

		sa->runnable_avg_sum = decay_load(sa->runnable_avg_sum,
						  periods + 1);
		sa->runnable_avg_period = decay_load(sa->runnable_avg_period,
						     periods + 1);

		/* Efficiently calculate \sum (1..n_period) 1024*y^i */
		runnable_contrib = __compute_runnable_contrib(periods);
		if (runnable)
			sa->runnable_avg_sum += runnable_contrib;
		sa->runnable_avg_period += runnable_contrib;
1308 1309 1310 1311 1312 1313 1314 1315 1316 1317
	}

	/* Remainder of delta accrued against u_0` */
	if (runnable)
		sa->runnable_avg_sum += delta;
	sa->runnable_avg_period += delta;

	return decayed;
}

1318
/* Synchronize an entity's decay with its parenting cfs_rq.*/
1319
static inline u64 __synchronize_entity_decay(struct sched_entity *se)
1320 1321 1322 1323 1324 1325
{
	struct cfs_rq *cfs_rq = cfs_rq_of(se);
	u64 decays = atomic64_read(&cfs_rq->decay_counter);

	decays -= se->avg.decay_count;
	if (!decays)
1326
		return 0;
1327 1328 1329

	se->avg.load_avg_contrib = decay_load(se->avg.load_avg_contrib, decays);
	se->avg.decay_count = 0;
1330 1331

	return decays;
1332 1333
}

1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348
#ifdef CONFIG_FAIR_GROUP_SCHED
static inline void __update_cfs_rq_tg_load_contrib(struct cfs_rq *cfs_rq,
						 int force_update)
{
	struct task_group *tg = cfs_rq->tg;
	s64 tg_contrib;

	tg_contrib = cfs_rq->runnable_load_avg + cfs_rq->blocked_load_avg;
	tg_contrib -= cfs_rq->tg_load_contrib;

	if (force_update || abs64(tg_contrib) > cfs_rq->tg_load_contrib / 8) {
		atomic64_add(tg_contrib, &tg->load_avg);
		cfs_rq->tg_load_contrib += tg_contrib;
	}
}
1349

1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370
/*
 * Aggregate cfs_rq runnable averages into an equivalent task_group
 * representation for computing load contributions.
 */
static inline void __update_tg_runnable_avg(struct sched_avg *sa,
						  struct cfs_rq *cfs_rq)
{
	struct task_group *tg = cfs_rq->tg;
	long contrib;

	/* The fraction of a cpu used by this cfs_rq */
	contrib = div_u64(sa->runnable_avg_sum << NICE_0_SHIFT,
			  sa->runnable_avg_period + 1);
	contrib -= cfs_rq->tg_runnable_contrib;

	if (abs(contrib) > cfs_rq->tg_runnable_contrib / 64) {
		atomic_add(contrib, &tg->runnable_avg);
		cfs_rq->tg_runnable_contrib += contrib;
	}
}

1371 1372 1373 1374
static inline void __update_group_entity_contrib(struct sched_entity *se)
{
	struct cfs_rq *cfs_rq = group_cfs_rq(se);
	struct task_group *tg = cfs_rq->tg;
1375 1376
	int runnable_avg;

1377 1378 1379 1380 1381
	u64 contrib;

	contrib = cfs_rq->tg_load_contrib * tg->shares;
	se->avg.load_avg_contrib = div64_u64(contrib,
					     atomic64_read(&tg->load_avg) + 1);
1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410

	/*
	 * For group entities we need to compute a correction term in the case
	 * that they are consuming <1 cpu so that we would contribute the same
	 * load as a task of equal weight.
	 *
	 * Explicitly co-ordinating this measurement would be expensive, but
	 * fortunately the sum of each cpus contribution forms a usable
	 * lower-bound on the true value.
	 *
	 * Consider the aggregate of 2 contributions.  Either they are disjoint
	 * (and the sum represents true value) or they are disjoint and we are
	 * understating by the aggregate of their overlap.
	 *
	 * Extending this to N cpus, for a given overlap, the maximum amount we
	 * understand is then n_i(n_i+1)/2 * w_i where n_i is the number of
	 * cpus that overlap for this interval and w_i is the interval width.
	 *
	 * On a small machine; the first term is well-bounded which bounds the
	 * total error since w_i is a subset of the period.  Whereas on a
	 * larger machine, while this first term can be larger, if w_i is the
	 * of consequential size guaranteed to see n_i*w_i quickly converge to
	 * our upper bound of 1-cpu.
	 */
	runnable_avg = atomic_read(&tg->runnable_avg);
	if (runnable_avg < NICE_0_LOAD) {
		se->avg.load_avg_contrib *= runnable_avg;
		se->avg.load_avg_contrib >>= NICE_0_SHIFT;
	}
1411
}
1412 1413 1414
#else
static inline void __update_cfs_rq_tg_load_contrib(struct cfs_rq *cfs_rq,
						 int force_update) {}
1415 1416
static inline void __update_tg_runnable_avg(struct sched_avg *sa,
						  struct cfs_rq *cfs_rq) {}
1417
static inline void __update_group_entity_contrib(struct sched_entity *se) {}
1418 1419
#endif

1420 1421 1422 1423 1424 1425 1426 1427 1428 1429
static inline void __update_task_entity_contrib(struct sched_entity *se)
{
	u32 contrib;

	/* avoid overflowing a 32-bit type w/ SCHED_LOAD_SCALE */
	contrib = se->avg.runnable_avg_sum * scale_load_down(se->load.weight);
	contrib /= (se->avg.runnable_avg_period + 1);
	se->avg.load_avg_contrib = scale_load(contrib);
}

1430 1431 1432 1433 1434
/* Compute the current contribution to load_avg by se, return any delta */
static long __update_entity_load_avg_contrib(struct sched_entity *se)
{
	long old_contrib = se->avg.load_avg_contrib;

1435 1436 1437
	if (entity_is_task(se)) {
		__update_task_entity_contrib(se);
	} else {
1438
		__update_tg_runnable_avg(&se->avg, group_cfs_rq(se));
1439 1440
		__update_group_entity_contrib(se);
	}
1441 1442 1443 1444

	return se->avg.load_avg_contrib - old_contrib;
}

1445 1446 1447 1448 1449 1450 1451 1452 1453
static inline void subtract_blocked_load_contrib(struct cfs_rq *cfs_rq,
						 long load_contrib)
{
	if (likely(load_contrib < cfs_rq->blocked_load_avg))
		cfs_rq->blocked_load_avg -= load_contrib;
	else
		cfs_rq->blocked_load_avg = 0;
}

1454 1455
static inline u64 cfs_rq_clock_task(struct cfs_rq *cfs_rq);

1456
/* Update a sched_entity's runnable average */
1457 1458
static inline void update_entity_load_avg(struct sched_entity *se,
					  int update_cfs_rq)
1459
{
1460 1461
	struct cfs_rq *cfs_rq = cfs_rq_of(se);
	long contrib_delta;
1462
	u64 now;
1463

1464 1465 1466 1467 1468 1469 1470 1471 1472 1473
	/*
	 * For a group entity we need to use their owned cfs_rq_clock_task() in
	 * case they are the parent of a throttled hierarchy.
	 */
	if (entity_is_task(se))
		now = cfs_rq_clock_task(cfs_rq);
	else
		now = cfs_rq_clock_task(group_cfs_rq(se));

	if (!__update_entity_runnable_avg(now, &se->avg, se->on_rq))
1474 1475 1476
		return;

	contrib_delta = __update_entity_load_avg_contrib(se);
1477 1478 1479 1480

	if (!update_cfs_rq)
		return;

1481 1482
	if (se->on_rq)
		cfs_rq->runnable_load_avg += contrib_delta;
1483 1484 1485 1486 1487 1488 1489 1490
	else
		subtract_blocked_load_contrib(cfs_rq, -contrib_delta);
}

/*
 * Decay the load contributed by all blocked children and account this so that
 * their contribution may appropriately discounted when they wake up.
 */
1491
static void update_cfs_rq_blocked_load(struct cfs_rq *cfs_rq, int force_update)
1492
{
1493
	u64 now = cfs_rq_clock_task(cfs_rq) >> 20;
1494 1495 1496
	u64 decays;

	decays = now - cfs_rq->last_decay;
1497
	if (!decays && !force_update)
1498 1499
		return;

1500 1501 1502 1503
	if (atomic64_read(&cfs_rq->removed_load)) {
		u64 removed_load = atomic64_xchg(&cfs_rq->removed_load, 0);
		subtract_blocked_load_contrib(cfs_rq, removed_load);
	}
1504

1505 1506 1507 1508 1509 1510
	if (decays) {
		cfs_rq->blocked_load_avg = decay_load(cfs_rq->blocked_load_avg,
						      decays);
		atomic64_add(decays, &cfs_rq->decay_counter);
		cfs_rq->last_decay = now;
	}
1511 1512

	__update_cfs_rq_tg_load_contrib(cfs_rq, force_update);
1513
}
1514 1515 1516

static inline void update_rq_runnable_avg(struct rq *rq, int runnable)
{
1517
	__update_entity_runnable_avg(rq_clock_task(rq), &rq->avg, runnable);
1518
	__update_tg_runnable_avg(&rq->avg, &rq->cfs);
1519
}
1520 1521 1522

/* Add the load generated by se into cfs_rq's child load-average */
static inline void enqueue_entity_load_avg(struct cfs_rq *cfs_rq,
1523 1524
						  struct sched_entity *se,
						  int wakeup)
1525
{
1526 1527 1528 1529 1530 1531
	/*
	 * We track migrations using entity decay_count <= 0, on a wake-up
	 * migration we use a negative decay count to track the remote decays
	 * accumulated while sleeping.
	 */
	if (unlikely(se->avg.decay_count <= 0)) {
1532
		se->avg.last_runnable_update = rq_clock_task(rq_of(cfs_rq));
1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547
		if (se->avg.decay_count) {
			/*
			 * In a wake-up migration we have to approximate the
			 * time sleeping.  This is because we can't synchronize
			 * clock_task between the two cpus, and it is not
			 * guaranteed to be read-safe.  Instead, we can
			 * approximate this using our carried decays, which are
			 * explicitly atomically readable.
			 */
			se->avg.last_runnable_update -= (-se->avg.decay_count)
							<< 20;
			update_entity_load_avg(se, 0);
			/* Indicate that we're now synchronized and on-rq */
			se->avg.decay_count = 0;
		}
1548 1549 1550 1551 1552
		wakeup = 0;
	} else {
		__synchronize_entity_decay(se);
	}

1553 1554
	/* migrated tasks did not contribute to our blocked load */
	if (wakeup) {
1555
		subtract_blocked_load_contrib(cfs_rq, se->avg.load_avg_contrib);
1556 1557
		update_entity_load_avg(se, 0);
	}
1558

1559
	cfs_rq->runnable_load_avg += se->avg.load_avg_contrib;
1560 1561
	/* we force update consideration on load-balancer moves */
	update_cfs_rq_blocked_load(cfs_rq, !wakeup);
1562 1563
}

1564 1565 1566 1567 1568
/*
 * Remove se's load from this cfs_rq child load-average, if the entity is
 * transitioning to a blocked state we track its projected decay using
 * blocked_load_avg.
 */
1569
static inline void dequeue_entity_load_avg(struct cfs_rq *cfs_rq,
1570 1571
						  struct sched_entity *se,
						  int sleep)
1572
{
1573
	update_entity_load_avg(se, 1);
1574 1575
	/* we force update consideration on load-balancer moves */
	update_cfs_rq_blocked_load(cfs_rq, !sleep);
1576

1577
	cfs_rq->runnable_load_avg -= se->avg.load_avg_contrib;
1578 1579 1580 1581
	if (sleep) {
		cfs_rq->blocked_load_avg += se->avg.load_avg_contrib;
		se->avg.decay_count = atomic64_read(&cfs_rq->decay_counter);
	} /* migrations, e.g. sleep=0 leave decay_count == 0 */
1582
}
1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603

/*
 * Update the rq's load with the elapsed running time before entering
 * idle. if the last scheduled task is not a CFS task, idle_enter will
 * be the only way to update the runnable statistic.
 */
void idle_enter_fair(struct rq *this_rq)
{
	update_rq_runnable_avg(this_rq, 1);
}

/*
 * Update the rq's load with the elapsed idle time before a task is
 * scheduled. if the newly scheduled task is not a CFS task, idle_exit will
 * be the only way to update the runnable statistic.
 */
void idle_exit_fair(struct rq *this_rq)
{
	update_rq_runnable_avg(this_rq, 0);
}

1604
#else
1605 1606
static inline void update_entity_load_avg(struct sched_entity *se,
					  int update_cfs_rq) {}
1607
static inline void update_rq_runnable_avg(struct rq *rq, int runnable) {}
1608
static inline void enqueue_entity_load_avg(struct cfs_rq *cfs_rq,
1609 1610
					   struct sched_entity *se,
					   int wakeup) {}
1611
static inline void dequeue_entity_load_avg(struct cfs_rq *cfs_rq,
1612 1613
					   struct sched_entity *se,
					   int sleep) {}
1614 1615
static inline void update_cfs_rq_blocked_load(struct cfs_rq *cfs_rq,
					      int force_update) {}
1616 1617
#endif

1618
static void enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
1619 1620
{
#ifdef CONFIG_SCHEDSTATS
1621 1622 1623 1624 1625
	struct task_struct *tsk = NULL;

	if (entity_is_task(se))
		tsk = task_of(se);

1626
	if (se->statistics.sleep_start) {
1627
		u64 delta = rq_clock(rq_of(cfs_rq)) - se->statistics.sleep_start;
1628 1629 1630 1631

		if ((s64)delta < 0)
			delta = 0;

1632 1633
		if (unlikely(delta > se->statistics.sleep_max))
			se->statistics.sleep_max = delta;
1634

1635
		se->statistics.sleep_start = 0;
1636
		se->statistics.sum_sleep_runtime += delta;
Arjan van de Ven's avatar
Arjan van de Ven committed
1637

1638
		if (tsk) {
1639
			account_scheduler_latency(tsk, delta >> 10, 1);
1640 1641
			trace_sched_stat_sleep(tsk, delta);
		}
1642
	}
1643
	if (se->statistics.block_start) {
1644
		u64 delta = rq_clock(rq_of(cfs_rq)) - se->statistics.block_start;
1645 1646 1647 1648

		if ((s64)delta < 0)
			delta = 0;

1649 1650
		if (unlikely(delta > se->statistics.block_max))
			se->statistics.block_max = delta;
1651

1652
		se->statistics.block_start = 0;
1653
		se->statistics.sum_sleep_runtime += delta;
Ingo Molnar's avatar
Ingo Molnar committed
1654

1655
		if (tsk) {
1656
			if (tsk->in_iowait) {
1657 1658
				se->statistics.iowait_sum += delta;
				se->statistics.iowait_count++;
1659
				trace_sched_stat_iowait(tsk, delta);
1660 1661
			}

1662 1663
			trace_sched_stat_blocked(tsk, delta);

1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674
			/*
			 * Blocking time is in units of nanosecs, so shift by
			 * 20 to get a milliseconds-range estimation of the
			 * amount of time that the task spent sleeping:
			 */
			if (unlikely(prof_on == SLEEP_PROFILING)) {
				profile_hits(SLEEP_PROFILING,
						(void *)get_wchan(tsk),
						delta >> 20);
			}
			account_scheduler_latency(tsk, delta >> 10, 0);
Ingo Molnar's avatar
Ingo Molnar committed
1675
		}
1676 1677 1678 1679
	}
#endif
}

Peter Zijlstra's avatar
Peter Zijlstra committed
1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692
static void check_spread(struct cfs_rq *cfs_rq, struct sched_entity *se)
{
#ifdef CONFIG_SCHED_DEBUG
	s64 d = se->vruntime - cfs_rq->min_vruntime;

	if (d < 0)
		d = -d;

	if (d > 3*sysctl_sched_latency)
		schedstat_inc(cfs_rq, nr_spread_over);
#endif
}

1693 1694 1695
static void
place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial)
{
1696
	u64 vruntime = cfs_rq->min_vruntime;
1697

1698 1699 1700 1701 1702 1703
	/*
	 * The 'current' period is already promised to the current tasks,
	 * however the extra weight of the new task will slow them down a
	 * little, place the new task so that it fits in the slot that
	 * stays open at the end.
	 */
1704
	if (initial && sched_feat(START_DEBIT))
1705
		vruntime += sched_vslice(cfs_rq, se);
1706

1707
	/* sleeps up to a single latency don't count. */
1708
	if (!initial) {
1709
		unsigned long thresh = sysctl_sched_latency;
1710

1711 1712 1713 1714 1715 1716
		/*
		 * Halve their sleep time's effect, to allow
		 * for a gentler effect of sleepers:
		 */
		if (sched_feat(GENTLE_FAIR_SLEEPERS))
			thresh >>= 1;
1717

1718
		vruntime -= thresh;
1719 1720
	}

1721
	/* ensure we never gain time by being placed backwards. */
1722
	se->vruntime = max_vruntime(se->vruntime, vruntime);
1723 1724
}

1725 1726
static void check_enqueue_throttle(struct cfs_rq *cfs_rq);

1727
static void
1728
enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
1729
{
1730 1731 1732 1733
	/*
	 * Update the normalized vruntime before updating min_vruntime
	 * through callig update_curr().
	 */
1734
	if (!(flags & ENQUEUE_WAKEUP) || (flags & ENQUEUE_WAKING))
1735 1736
		se->vruntime += cfs_rq->min_vruntime;

1737
	/*
1738
	 * Update run-time statistics of the 'current'.
1739
	 */
1740
	update_curr(cfs_rq);
1741
	enqueue_entity_load_avg(cfs_rq, se, flags & ENQUEUE_WAKEUP);
1742 1743
	account_entity_enqueue(cfs_rq, se);
	update_cfs_shares(cfs_rq);
1744

1745
	if (flags & ENQUEUE_WAKEUP) {
1746
		place_entity(cfs_rq, se, 0);
1747
		enqueue_sleeper(cfs_rq, se);
Ingo Molnar's avatar
Ingo Molnar committed
1748
	}
1749

1750
	update_stats_enqueue(cfs_rq, se);
Peter Zijlstra's avatar
Peter Zijlstra committed
1751
	check_spread(cfs_rq, se);
1752 1753
	if (se != cfs_rq->curr)
		__enqueue_entity(cfs_rq, se);
1754
	se->on_rq = 1;
1755

1756
	if (cfs_rq->nr_running == 1) {
1757
		list_add_leaf_cfs_rq(cfs_rq);
1758 1759
		check_enqueue_throttle(cfs_rq);
	}
1760 1761
}

1762
static void __clear_buddies_last(struct sched_entity *se)
1763
{
1764 1765 1766 1767 1768 1769 1770 1771
	for_each_sched_entity(se) {
		struct cfs_rq *cfs_rq = cfs_rq_of(se);
		if (cfs_rq->last == se)
			cfs_rq->last = NULL;
		else
			break;
	}
}
1772

1773 1774 1775 1776 1777 1778 1779 1780 1781
static void __clear_buddies_next(struct sched_entity *se)
{
	for_each_sched_entity(se) {
		struct cfs_rq *cfs_rq = cfs_rq_of(se);
		if (cfs_rq->next == se)
			cfs_rq->next = NULL;
		else
			break;
	}
1782 1783
}

1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794
static void __clear_buddies_skip(struct sched_entity *se)
{
	for_each_sched_entity(se) {
		struct cfs_rq *cfs_rq = cfs_rq_of(se);
		if (cfs_rq->skip == se)
			cfs_rq->skip = NULL;
		else
			break;
	}
}

1795 1796
static void clear_buddies(struct cfs_rq *cfs_rq, struct sched_entity *se)
{
1797 1798 1799 1800 1801
	if (cfs_rq->last == se)
		__clear_buddies_last(se);

	if (cfs_rq->next == se)
		__clear_buddies_next(se);
1802 1803 1804

	if (cfs_rq->skip == se)
		__clear_buddies_skip(se);
1805 1806
}

1807
static __always_inline void return_cfs_rq_runtime(struct cfs_rq *cfs_rq);
1808

1809
static void
1810
dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
1811
{
1812 1813 1814 1815
	/*
	 * Update run-time statistics of the 'current'.
	 */
	update_curr(cfs_rq);
1816
	dequeue_entity_load_avg(cfs_rq, se, flags & DEQUEUE_SLEEP);
1817

1818
	update_stats_dequeue(cfs_rq, se);
1819
	if (flags & DEQUEUE_SLEEP) {
Peter Zijlstra's avatar
Peter Zijlstra committed
1820
#ifdef CONFIG_SCHEDSTATS
1821 1822 1823 1824
		if (entity_is_task(se)) {
			struct task_struct *tsk = task_of(se);

			if (tsk->state & TASK_INTERRUPTIBLE)
1825
				se->statistics.sleep_start = rq_clock(rq_of(cfs_rq));
1826
			if (tsk->state & TASK_UNINTERRUPTIBLE)
1827
				se->statistics.block_start = rq_clock(rq_of(cfs_rq));
1828
		}
1829
#endif
Peter Zijlstra's avatar
Peter Zijlstra committed
1830 1831
	}

1832
	clear_buddies(cfs_rq, se);
1833

1834
	if (se != cfs_rq->curr)
1835
		__dequeue_entity(cfs_rq, se);
1836
	se->on_rq = 0;
1837
	account_entity_dequeue(cfs_rq, se);
1838 1839 1840 1841 1842 1843

	/*
	 * Normalize the entity after updating the min_vruntime because the
	 * update can refer to the ->curr item and we need to reflect this
	 * movement in our normalized position.
	 */
1844
	if (!(flags & DEQUEUE_SLEEP))
1845
		se->vruntime -= cfs_rq->min_vruntime;
1846

1847 1848 1849
	/* return excess runtime on last dequeue */
	return_cfs_rq_runtime(cfs_rq);

1850
	update_min_vruntime(cfs_rq);
1851
	update_cfs_shares(cfs_rq);
1852 1853 1854 1855 1856
}

/*
 * Preempt the current task with a newly woken task if needed:
 */
1857
static void
1858
check_preempt_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr)
1859
{
1860
	unsigned long ideal_runtime, delta_exec;
1861 1862
	struct sched_entity *se;
	s64 delta;
1863

1864
	ideal_runtime = sched_slice(cfs_rq, curr);
1865
	delta_exec = curr->sum_exec_runtime - curr->prev_sum_exec_runtime;
1866
	if (delta_exec > ideal_runtime) {
1867
		resched_task(rq_of(cfs_rq)->curr);
1868 1869 1870 1871 1872
		/*
		 * The current task ran long enough, ensure it doesn't get
		 * re-elected due to buddy favours.
		 */
		clear_buddies(cfs_rq, curr);
1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883
		return;
	}

	/*
	 * Ensure that a task that missed wakeup preemption by a
	 * narrow margin doesn't have to wait for a full slice.
	 * This also mitigates buddy induced latencies under load.
	 */
	if (delta_exec < sysctl_sched_min_granularity)
		return;

1884 1885
	se = __pick_first_entity(cfs_rq);
	delta = curr->vruntime - se->vruntime;
1886

1887 1888
	if (delta < 0)
		return;
1889

1890 1891
	if (delta > ideal_runtime)
		resched_task(rq_of(cfs_rq)->curr);
1892 1893
}

1894
static void
1895
set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
1896
{
1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907
	/* 'current' is not kept within the tree. */
	if (se->on_rq) {
		/*
		 * Any task has to be enqueued before it get to execute on
		 * a CPU. So account for the time it spent waiting on the
		 * runqueue.
		 */
		update_stats_wait_end(cfs_rq, se);
		__dequeue_entity(cfs_rq, se);
	}

1908
	update_stats_curr_start(cfs_rq, se);
1909
	cfs_rq->curr = se;
1910 1911 1912 1913 1914 1915
#ifdef CONFIG_SCHEDSTATS
	/*
	 * Track our maximum slice length, if the CPU's load is at
	 * least twice that of our own weight (i.e. dont track it
	 * when there are only lesser-weight tasks around):
	 */
1916
	if (rq_of(cfs_rq)->load.weight >= 2*se->load.weight) {
1917
		se->statistics.slice_max = max(se->statistics.slice_max,
1918 1919 1920
			se->sum_exec_runtime - se->prev_sum_exec_runtime);
	}
#endif
1921
	se->prev_sum_exec_runtime = se->sum_exec_runtime;
1922 1923
}

1924 1925 1926
static int
wakeup_preempt_entity(struct sched_entity *curr, struct sched_entity *se);

1927 1928 1929 1930 1931 1932 1933
/*
 * Pick the next process, keeping these things in mind, in this order:
 * 1) keep things fair between processes/task groups
 * 2) pick the "next" process, since someone really wants that to run
 * 3) pick the "last" process, for cache locality
 * 4) do not run the "skip" process, if something else is available
 */
1934
static struct sched_entity *pick_next_entity(struct cfs_rq *cfs_rq)
1935
{
1936
	struct sched_entity *se = __pick_first_entity(cfs_rq);
1937
	struct sched_entity *left = se;
1938

1939 1940 1941 1942 1943 1944 1945 1946 1947
	/*
	 * Avoid running the skip buddy, if running something else can
	 * be done without getting too unfair.
	 */
	if (cfs_rq->skip == se) {
		struct sched_entity *second = __pick_next_entity(se);
		if (second && wakeup_preempt_entity(second, left) < 1)
			se = second;
	}
1948

1949 1950 1951 1952 1953 1954
	/*
	 * Prefer last buddy, try to return the CPU to a preempted task.
	 */
	if (cfs_rq->last && wakeup_preempt_entity(cfs_rq->last, left) < 1)
		se = cfs_rq->last;

1955 1956 1957 1958 1959 1960
	/*
	 * Someone really wants this to run. If it's not unfair, run it.
	 */
	if (cfs_rq->next && wakeup_preempt_entity(cfs_rq->next, left) < 1)
		se = cfs_rq->next;

1961
	clear_buddies(cfs_rq, se);
1962 1963

	return se;
1964 1965
}

1966 1967
static void check_cfs_rq_runtime(struct cfs_rq *cfs_rq);

1968
static void put_prev_entity(struct cfs_rq *cfs_rq, struct sched_entity *prev)
1969 1970 1971 1972 1973 1974
{
	/*
	 * If still on the runqueue then deactivate_task()
	 * was not called and update_curr() has to be done:
	 */
	if (prev->on_rq)
1975
		update_curr(cfs_rq);
1976

1977 1978 1979
	/* throttle cfs_rqs exceeding runtime */
	check_cfs_rq_runtime(cfs_rq);

Peter Zijlstra's avatar
Peter Zijlstra committed
1980
	check_spread(cfs_rq, prev);
1981
	if (prev->on_rq) {
1982
		update_stats_wait_start(cfs_rq, prev);
1983 1984
		/* Put 'current' back into the tree. */
		__enqueue_entity(cfs_rq, prev);
1985
		/* in !on_rq case, update occurred at dequeue */
1986
		update_entity_load_avg(prev, 1);
1987
	}
1988
	cfs_rq->curr = NULL;
1989 1990
}

1991 1992
static void
entity_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr, int queued)
1993 1994
{
	/*
1995
	 * Update run-time statistics of the 'current'.
1996
	 */
1997
	update_curr(cfs_rq);
1998

1999 2000 2001
	/*
	 * Ensure that runnable average is periodically updated.
	 */
2002
	update_entity_load_avg(curr, 1);
2003
	update_cfs_rq_blocked_load(cfs_rq, 1);
2004

2005 2006 2007 2008 2009
#ifdef CONFIG_SCHED_HRTICK
	/*
	 * queued ticks are scheduled to match the slice, so don't bother
	 * validating it and just reschedule.
	 */
2010 2011 2012 2013
	if (queued) {
		resched_task(rq_of(cfs_rq)->curr);
		return;
	}
2014 2015 2016 2017 2018 2019 2020 2021
	/*
	 * don't let the period tick interfere with the hrtick preemption
	 */
	if (!sched_feat(DOUBLE_TICK) &&
			hrtimer_active(&rq_of(cfs_rq)->hrtick_timer))
		return;
#endif

Yong Zhang's avatar
Yong Zhang committed
2022
	if (cfs_rq->nr_running > 1)
2023
		check_preempt_tick(cfs_rq, curr);
2024 2025
}

2026 2027 2028 2029 2030 2031

/**************************************************
 * CFS bandwidth control machinery
 */

#ifdef CONFIG_CFS_BANDWIDTH
2032 2033

#ifdef HAVE_JUMP_LABEL
2034
static struct static_key __cfs_bandwidth_used;
2035 2036 2037

static inline bool cfs_bandwidth_used(void)
{
2038
	return static_key_false(&__cfs_bandwidth_used);
2039 2040 2041 2042 2043 2044
}

void account_cfs_bandwidth_used(int enabled, int was_enabled)
{
	/* only need to count groups transitioning between enabled/!enabled */
	if (enabled && !was_enabled)
2045
		static_key_slow_inc(&__cfs_bandwidth_used);
2046
	else if (!enabled && was_enabled)
2047
		static_key_slow_dec(&__cfs_bandwidth_used);
2048 2049 2050 2051 2052 2053 2054 2055 2056 2057
}
#else /* HAVE_JUMP_LABEL */
static bool cfs_bandwidth_used(void)
{
	return true;
}

void account_cfs_bandwidth_used(int enabled, int was_enabled) {}
#endif /* HAVE_JUMP_LABEL */

2058 2059 2060 2061 2062 2063 2064 2065
/*
 * default period for cfs group bandwidth.
 * default: 0.1s, units: nanoseconds
 */
static inline u64 default_cfs_period(void)
{
	return 100000000ULL;
}
2066 2067 2068 2069 2070 2071

static inline u64 sched_cfs_bandwidth_slice(void)
{
	return (u64)sysctl_sched_cfs_bandwidth_slice * NSEC_PER_USEC;
}

Paul Turner's avatar
Paul Turner committed
2072 2073 2074 2075 2076 2077 2078
/*
 * Replenish runtime according to assigned quota and update expiration time.
 * We use sched_clock_cpu directly instead of rq->clock to avoid adding
 * additional synchronization around rq->lock.
 *
 * requires cfs_b->lock
 */
2079
void __refill_cfs_bandwidth_runtime(struct cfs_bandwidth *cfs_b)
Paul Turner's avatar
Paul Turner committed
2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090
{
	u64 now;

	if (cfs_b->quota == RUNTIME_INF)
		return;

	now = sched_clock_cpu(smp_processor_id());
	cfs_b->runtime = cfs_b->quota;
	cfs_b->runtime_expires = now + ktime_to_ns(cfs_b->period);
}

2091 2092 2093 2094 2095
static inline struct cfs_bandwidth *tg_cfs_bandwidth(struct task_group *tg)
{
	return &tg->cfs_bandwidth;
}

2096 2097 2098 2099 2100 2101
/* rq->task_clock normalized against any time this cfs_rq has spent throttled */
static inline u64 cfs_rq_clock_task(struct cfs_rq *cfs_rq)
{
	if (unlikely(cfs_rq->throttle_count))
		return cfs_rq->throttled_clock_task;

2102
	return rq_clock_task(rq_of(cfs_rq)) - cfs_rq->throttled_clock_task_time;
2103 2104
}

2105 2106
/* returns 0 on failure to allocate runtime */
static int assign_cfs_rq_runtime(struct cfs_rq *cfs_rq)
2107 2108 2109
{
	struct task_group *tg = cfs_rq->tg;
	struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(tg);
Paul Turner's avatar
Paul Turner committed
2110
	u64 amount = 0, min_amount, expires;
2111 2112 2113 2114 2115 2116 2117

	/* note: this is a positive sum as runtime_remaining <= 0 */
	min_amount = sched_cfs_bandwidth_slice() - cfs_rq->runtime_remaining;

	raw_spin_lock(&cfs_b->lock);
	if (cfs_b->quota == RUNTIME_INF)
		amount = min_amount;
2118
	else {
Paul Turner's avatar
Paul Turner committed
2119 2120 2121 2122 2123 2124 2125 2126
		/*
		 * If the bandwidth pool has become inactive, then at least one
		 * period must have elapsed since the last consumption.
		 * Refresh the global state and ensure bandwidth timer becomes
		 * active.
		 */
		if (!cfs_b->timer_active) {
			__refill_cfs_bandwidth_runtime(cfs_b);
2127
			__start_cfs_bandwidth(cfs_b);
Paul Turner's avatar
Paul Turner committed
2128
		}
2129 2130 2131 2132 2133 2134

		if (cfs_b->runtime > 0) {
			amount = min(cfs_b->runtime, min_amount);
			cfs_b->runtime -= amount;
			cfs_b->idle = 0;
		}
2135
	}
Paul Turner's avatar
Paul Turner committed
2136
	expires = cfs_b->runtime_expires;
2137 2138 2139
	raw_spin_unlock(&cfs_b->lock);

	cfs_rq->runtime_remaining += amount;
Paul Turner's avatar
Paul Turner committed
2140 2141 2142 2143 2144 2145 2146
	/*
	 * we may have advanced our local expiration to account for allowed
	 * spread between our sched_clock and the one on which runtime was
	 * issued.
	 */
	if ((s64)(expires - cfs_rq->runtime_expires) > 0)
		cfs_rq->runtime_expires = expires;
2147 2148

	return cfs_rq->runtime_remaining > 0;
2149 2150
}

Paul Turner's avatar
Paul Turner committed
2151 2152 2153 2154 2155
/*
 * Note: This depends on the synchronization provided by sched_clock and the
 * fact that rq->clock snapshots this value.
 */
static void expire_cfs_rq_runtime(struct cfs_rq *cfs_rq)
2156
{
Paul Turner's avatar
Paul Turner committed
2157 2158 2159
	struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(cfs_rq->tg);

	/* if the deadline is ahead of our clock, nothing to do */
2160
	if (likely((s64)(rq_clock(rq_of(cfs_rq)) - cfs_rq->runtime_expires) < 0))
2161 2162
		return;

Paul Turner's avatar
Paul Turner committed
2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187
	if (cfs_rq->runtime_remaining < 0)
		return;

	/*
	 * If the local deadline has passed we have to consider the
	 * possibility that our sched_clock is 'fast' and the global deadline
	 * has not truly expired.
	 *
	 * Fortunately we can check determine whether this the case by checking
	 * whether the global deadline has advanced.
	 */

	if ((s64)(cfs_rq->runtime_expires - cfs_b->runtime_expires) >= 0) {
		/* extend local deadline, drift is bounded above by 2 ticks */
		cfs_rq->runtime_expires += TICK_NSEC;
	} else {
		/* global deadline is ahead, expiration has passed */
		cfs_rq->runtime_remaining = 0;
	}
}

static void __account_cfs_rq_runtime(struct cfs_rq *cfs_rq,
				     unsigned long delta_exec)
{
	/* dock delta_exec before expiring quota (as it could span periods) */
2188
	cfs_rq->runtime_remaining -= delta_exec;
Paul Turner's avatar
Paul Turner committed
2189 2190 2191
	expire_cfs_rq_runtime(cfs_rq);

	if (likely(cfs_rq->runtime_remaining > 0))
2192 2193
		return;

2194 2195 2196 2197 2198 2199
	/*
	 * if we're unable to extend our runtime we resched so that the active
	 * hierarchy can be throttled
	 */
	if (!assign_cfs_rq_runtime(cfs_rq) && likely(cfs_rq->curr))
		resched_task(rq_of(cfs_rq)->curr);
2200 2201
}

2202 2203
static __always_inline
void account_cfs_rq_runtime(struct cfs_rq *cfs_rq, unsigned long delta_exec)
2204
{
2205
	if (!cfs_bandwidth_used() || !cfs_rq->runtime_enabled)
2206 2207 2208 2209 2210
		return;

	__account_cfs_rq_runtime(cfs_rq, delta_exec);
}

2211 2212
static inline int cfs_rq_throttled(struct cfs_rq *cfs_rq)
{
2213
	return cfs_bandwidth_used() && cfs_rq->throttled;
2214 2215
}

2216 2217 2218
/* check whether cfs_rq, or any parent, is throttled */
static inline int throttled_hierarchy(struct cfs_rq *cfs_rq)
{
2219
	return cfs_bandwidth_used() && cfs_rq->throttle_count;
2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247
}

/*
 * Ensure that neither of the group entities corresponding to src_cpu or
 * dest_cpu are members of a throttled hierarchy when performing group
 * load-balance operations.
 */
static inline int throttled_lb_pair(struct task_group *tg,
				    int src_cpu, int dest_cpu)
{
	struct cfs_rq *src_cfs_rq, *dest_cfs_rq;

	src_cfs_rq = tg->cfs_rq[src_cpu];
	dest_cfs_rq = tg->cfs_rq[dest_cpu];

	return throttled_hierarchy(src_cfs_rq) ||
	       throttled_hierarchy(dest_cfs_rq);
}

/* updated child weight may affect parent so we have to do this bottom up */
static int tg_unthrottle_up(struct task_group *tg, void *data)
{
	struct rq *rq = data;
	struct cfs_rq *cfs_rq = tg->cfs_rq[cpu_of(rq)];

	cfs_rq->throttle_count--;
#ifdef CONFIG_SMP
	if (!cfs_rq->throttle_count) {
2248
		/* adjust cfs_rq_clock_task() */
2249
		cfs_rq->throttled_clock_task_time += rq_clock_task(rq) -
2250
					     cfs_rq->throttled_clock_task;
2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261
	}
#endif

	return 0;
}

static int tg_throttle_down(struct task_group *tg, void *data)
{
	struct rq *rq = data;
	struct cfs_rq *cfs_rq = tg->cfs_rq[cpu_of(rq)];

2262 2263
	/* group is entering throttled state, stop time */
	if (!cfs_rq->throttle_count)
2264
		cfs_rq->throttled_clock_task = rq_clock_task(rq);
2265 2266 2267 2268 2269
	cfs_rq->throttle_count++;

	return 0;
}

2270
static void throttle_cfs_rq(struct cfs_rq *cfs_rq)
2271 2272 2273 2274 2275 2276 2277 2278
{
	struct rq *rq = rq_of(cfs_rq);
	struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(cfs_rq->tg);
	struct sched_entity *se;
	long task_delta, dequeue = 1;

	se = cfs_rq->tg->se[cpu_of(rq_of(cfs_rq))];

2279
	/* freeze hierarchy runnable averages while throttled */
2280 2281 2282
	rcu_read_lock();
	walk_tg_tree_from(cfs_rq->tg, tg_throttle_down, tg_nop, (void *)rq);
	rcu_read_unlock();
2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302

	task_delta = cfs_rq->h_nr_running;
	for_each_sched_entity(se) {
		struct cfs_rq *qcfs_rq = cfs_rq_of(se);
		/* throttled entity or throttle-on-deactivate */
		if (!se->on_rq)
			break;

		if (dequeue)
			dequeue_entity(qcfs_rq, se, DEQUEUE_SLEEP);
		qcfs_rq->h_nr_running -= task_delta;

		if (qcfs_rq->load.weight)
			dequeue = 0;
	}

	if (!se)
		rq->nr_running -= task_delta;

	cfs_rq->throttled = 1;
2303
	cfs_rq->throttled_clock = rq_clock(rq);
2304 2305 2306 2307 2308
	raw_spin_lock(&cfs_b->lock);
	list_add_tail_rcu(&cfs_rq->throttled_list, &cfs_b->throttled_cfs_rq);
	raw_spin_unlock(&cfs_b->lock);
}

2309
void unthrottle_cfs_rq(struct cfs_rq *cfs_rq)
2310 2311 2312 2313 2314 2315 2316
{
	struct rq *rq = rq_of(cfs_rq);
	struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(cfs_rq->tg);
	struct sched_entity *se;
	int enqueue = 1;
	long task_delta;

2317
	se = cfs_rq->tg->se[cpu_of(rq)];
2318 2319

	cfs_rq->throttled = 0;
2320 2321 2322

	update_rq_clock(rq);

2323
	raw_spin_lock(&cfs_b->lock);
2324
	cfs_b->throttled_time += rq_clock(rq) - cfs_rq->throttled_clock;
2325 2326 2327
	list_del_rcu(&cfs_rq->throttled_list);
	raw_spin_unlock(&cfs_b->lock);

2328 2329 2330
	/* update hierarchical throttle state */
	walk_tg_tree_from(cfs_rq->tg, tg_nop, tg_unthrottle_up, (void *)rq);

2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393
	if (!cfs_rq->load.weight)
		return;

	task_delta = cfs_rq->h_nr_running;
	for_each_sched_entity(se) {
		if (se->on_rq)
			enqueue = 0;

		cfs_rq = cfs_rq_of(se);
		if (enqueue)
			enqueue_entity(cfs_rq, se, ENQUEUE_WAKEUP);
		cfs_rq->h_nr_running += task_delta;

		if (cfs_rq_throttled(cfs_rq))
			break;
	}

	if (!se)
		rq->nr_running += task_delta;

	/* determine whether we need to wake up potentially idle cpu */
	if (rq->curr == rq->idle && rq->cfs.nr_running)
		resched_task(rq->curr);
}

static u64 distribute_cfs_runtime(struct cfs_bandwidth *cfs_b,
		u64 remaining, u64 expires)
{
	struct cfs_rq *cfs_rq;
	u64 runtime = remaining;

	rcu_read_lock();
	list_for_each_entry_rcu(cfs_rq, &cfs_b->throttled_cfs_rq,
				throttled_list) {
		struct rq *rq = rq_of(cfs_rq);

		raw_spin_lock(&rq->lock);
		if (!cfs_rq_throttled(cfs_rq))
			goto next;

		runtime = -cfs_rq->runtime_remaining + 1;
		if (runtime > remaining)
			runtime = remaining;
		remaining -= runtime;

		cfs_rq->runtime_remaining += runtime;
		cfs_rq->runtime_expires = expires;

		/* we check whether we're throttled above */
		if (cfs_rq->runtime_remaining > 0)
			unthrottle_cfs_rq(cfs_rq);

next:
		raw_spin_unlock(&rq->lock);

		if (!remaining)
			break;
	}
	rcu_read_unlock();

	return remaining;
}

2394 2395 2396 2397 2398 2399 2400 2401
/*
 * Responsible for refilling a task_group's bandwidth and unthrottling its
 * cfs_rqs as appropriate. If there has been no activity within the last
 * period the timer is deactivated until scheduling resumes; cfs_b->idle is
 * used to track this state.
 */
static int do_sched_cfs_period_timer(struct cfs_bandwidth *cfs_b, int overrun)
{
2402 2403
	u64 runtime, runtime_expires;
	int idle = 1, throttled;
2404 2405 2406 2407 2408 2409

	raw_spin_lock(&cfs_b->lock);
	/* no need to continue the timer with no bandwidth constraint */
	if (cfs_b->quota == RUNTIME_INF)
		goto out_unlock;

2410 2411 2412
	throttled = !list_empty(&cfs_b->throttled_cfs_rq);
	/* idle depends on !throttled (for the case of a large deficit) */
	idle = cfs_b->idle && !throttled;
2413
	cfs_b->nr_periods += overrun;
2414

Paul Turner's avatar
Paul Turner committed
2415 2416 2417 2418 2419 2420
	/* if we're going inactive then everything else can be deferred */
	if (idle)
		goto out_unlock;

	__refill_cfs_bandwidth_runtime(cfs_b);

2421 2422 2423 2424 2425 2426
	if (!throttled) {
		/* mark as potentially idle for the upcoming period */
		cfs_b->idle = 1;
		goto out_unlock;
	}

2427 2428 2429
	/* account preceding periods in which throttling occurred */
	cfs_b->nr_throttled += overrun;

2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453
	/*
	 * There are throttled entities so we must first use the new bandwidth
	 * to unthrottle them before making it generally available.  This
	 * ensures that all existing debts will be paid before a new cfs_rq is
	 * allowed to run.
	 */
	runtime = cfs_b->runtime;
	runtime_expires = cfs_b->runtime_expires;
	cfs_b->runtime = 0;

	/*
	 * This check is repeated as we are holding onto the new bandwidth
	 * while we unthrottle.  This can potentially race with an unthrottled
	 * group trying to acquire new bandwidth from the global pool.
	 */
	while (throttled && runtime > 0) {
		raw_spin_unlock(&cfs_b->lock);
		/* we can't nest cfs_b->lock while distributing bandwidth */
		runtime = distribute_cfs_runtime(cfs_b, runtime,
						 runtime_expires);
		raw_spin_lock(&cfs_b->lock);

		throttled = !list_empty(&cfs_b->throttled_cfs_rq);
	}
2454

2455 2456 2457 2458 2459 2460 2461 2462 2463
	/* return (any) remaining runtime */
	cfs_b->runtime = runtime;
	/*
	 * While we are ensured activity in the period following an
	 * unthrottle, this also covers the case in which the new bandwidth is
	 * insufficient to cover the existing bandwidth deficit.  (Forcing the
	 * timer to remain active while there are any throttled entities.)
	 */
	cfs_b->idle = 0;
2464 2465 2466 2467 2468 2469 2470
out_unlock:
	if (idle)
		cfs_b->timer_active = 0;
	raw_spin_unlock(&cfs_b->lock);

	return idle;
}
2471

2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535
/* a cfs_rq won't donate quota below this amount */
static const u64 min_cfs_rq_runtime = 1 * NSEC_PER_MSEC;
/* minimum remaining period time to redistribute slack quota */
static const u64 min_bandwidth_expiration = 2 * NSEC_PER_MSEC;
/* how long we wait to gather additional slack before distributing */
static const u64 cfs_bandwidth_slack_period = 5 * NSEC_PER_MSEC;

/* are we near the end of the current quota period? */
static int runtime_refresh_within(struct cfs_bandwidth *cfs_b, u64 min_expire)
{
	struct hrtimer *refresh_timer = &cfs_b->period_timer;
	u64 remaining;

	/* if the call-back is running a quota refresh is already occurring */
	if (hrtimer_callback_running(refresh_timer))
		return 1;

	/* is a quota refresh about to occur? */
	remaining = ktime_to_ns(hrtimer_expires_remaining(refresh_timer));
	if (remaining < min_expire)
		return 1;

	return 0;
}

static void start_cfs_slack_bandwidth(struct cfs_bandwidth *cfs_b)
{
	u64 min_left = cfs_bandwidth_slack_period + min_bandwidth_expiration;

	/* if there's a quota refresh soon don't bother with slack */
	if (runtime_refresh_within(cfs_b, min_left))
		return;

	start_bandwidth_timer(&cfs_b->slack_timer,
				ns_to_ktime(cfs_bandwidth_slack_period));
}

/* we know any runtime found here is valid as update_curr() precedes return */
static void __return_cfs_rq_runtime(struct cfs_rq *cfs_rq)
{
	struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(cfs_rq->tg);
	s64 slack_runtime = cfs_rq->runtime_remaining - min_cfs_rq_runtime;

	if (slack_runtime <= 0)
		return;

	raw_spin_lock(&cfs_b->lock);
	if (cfs_b->quota != RUNTIME_INF &&
	    cfs_rq->runtime_expires == cfs_b->runtime_expires) {
		cfs_b->runtime += slack_runtime;

		/* we are under rq->lock, defer unthrottling using a timer */
		if (cfs_b->runtime > sched_cfs_bandwidth_slice() &&
		    !list_empty(&cfs_b->throttled_cfs_rq))
			start_cfs_slack_bandwidth(cfs_b);
	}
	raw_spin_unlock(&cfs_b->lock);

	/* even if it's not valid for return we don't want to try again */
	cfs_rq->runtime_remaining -= slack_runtime;
}

static __always_inline void return_cfs_rq_runtime(struct cfs_rq *cfs_rq)
{
2536 2537 2538
	if (!cfs_bandwidth_used())
		return;

2539
	if (!cfs_rq->runtime_enabled || cfs_rq->nr_running)
2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576
		return;

	__return_cfs_rq_runtime(cfs_rq);
}

/*
 * This is done with a timer (instead of inline with bandwidth return) since
 * it's necessary to juggle rq->locks to unthrottle their respective cfs_rqs.
 */
static void do_sched_cfs_slack_timer(struct cfs_bandwidth *cfs_b)
{
	u64 runtime = 0, slice = sched_cfs_bandwidth_slice();
	u64 expires;

	/* confirm we're still not at a refresh boundary */
	if (runtime_refresh_within(cfs_b, min_bandwidth_expiration))
		return;

	raw_spin_lock(&cfs_b->lock);
	if (cfs_b->quota != RUNTIME_INF && cfs_b->runtime > slice) {
		runtime = cfs_b->runtime;
		cfs_b->runtime = 0;
	}
	expires = cfs_b->runtime_expires;
	raw_spin_unlock(&cfs_b->lock);

	if (!runtime)
		return;

	runtime = distribute_cfs_runtime(cfs_b, runtime, expires);

	raw_spin_lock(&cfs_b->lock);
	if (expires == cfs_b->runtime_expires)
		cfs_b->runtime = runtime;
	raw_spin_unlock(&cfs_b->lock);
}

2577 2578 2579 2580 2581 2582 2583
/*
 * When a group wakes up we want to make sure that its quota is not already
 * expired/exceeded, otherwise it may be allowed to steal additional ticks of
 * runtime as update_curr() throttling can not not trigger until it's on-rq.
 */
static void check_enqueue_throttle(struct cfs_rq *cfs_rq)
{
2584 2585 2586
	if (!cfs_bandwidth_used())
		return;

2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603
	/* an active group must be handled by the update_curr()->put() path */
	if (!cfs_rq->runtime_enabled || cfs_rq->curr)
		return;

	/* ensure the group is not already throttled */
	if (cfs_rq_throttled(cfs_rq))
		return;

	/* update runtime allocation */
	account_cfs_rq_runtime(cfs_rq, 0);
	if (cfs_rq->runtime_remaining <= 0)
		throttle_cfs_rq(cfs_rq);
}

/* conditionally throttle active cfs_rq's from put_prev_entity() */
static void check_cfs_rq_runtime(struct cfs_rq *cfs_rq)
{
2604 2605 2606
	if (!cfs_bandwidth_used())
		return;

2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618
	if (likely(!cfs_rq->runtime_enabled || cfs_rq->runtime_remaining > 0))
		return;

	/*
	 * it's possible for a throttled entity to be forced into a running
	 * state (e.g. set_curr_task), in this case we're finished.
	 */
	if (cfs_rq_throttled(cfs_rq))
		return;

	throttle_cfs_rq(cfs_rq);
}
2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699

static enum hrtimer_restart sched_cfs_slack_timer(struct hrtimer *timer)
{
	struct cfs_bandwidth *cfs_b =
		container_of(timer, struct cfs_bandwidth, slack_timer);
	do_sched_cfs_slack_timer(cfs_b);

	return HRTIMER_NORESTART;
}

static enum hrtimer_restart sched_cfs_period_timer(struct hrtimer *timer)
{
	struct cfs_bandwidth *cfs_b =
		container_of(timer, struct cfs_bandwidth, period_timer);
	ktime_t now;
	int overrun;
	int idle = 0;

	for (;;) {
		now = hrtimer_cb_get_time(timer);
		overrun = hrtimer_forward(timer, now, cfs_b->period);

		if (!overrun)
			break;

		idle = do_sched_cfs_period_timer(cfs_b, overrun);
	}

	return idle ? HRTIMER_NORESTART : HRTIMER_RESTART;
}

void init_cfs_bandwidth(struct cfs_bandwidth *cfs_b)
{
	raw_spin_lock_init(&cfs_b->lock);
	cfs_b->runtime = 0;
	cfs_b->quota = RUNTIME_INF;
	cfs_b->period = ns_to_ktime(default_cfs_period());

	INIT_LIST_HEAD(&cfs_b->throttled_cfs_rq);
	hrtimer_init(&cfs_b->period_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	cfs_b->period_timer.function = sched_cfs_period_timer;
	hrtimer_init(&cfs_b->slack_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	cfs_b->slack_timer.function = sched_cfs_slack_timer;
}

static void init_cfs_rq_runtime(struct cfs_rq *cfs_rq)
{
	cfs_rq->runtime_enabled = 0;
	INIT_LIST_HEAD(&cfs_rq->throttled_list);
}

/* requires cfs_b->lock, may release to reprogram timer */
void __start_cfs_bandwidth(struct cfs_bandwidth *cfs_b)
{
	/*
	 * The timer may be active because we're trying to set a new bandwidth
	 * period or because we're racing with the tear-down path
	 * (timer_active==0 becomes visible before the hrtimer call-back
	 * terminates).  In either case we ensure that it's re-programmed
	 */
	while (unlikely(hrtimer_active(&cfs_b->period_timer))) {
		raw_spin_unlock(&cfs_b->lock);
		/* ensure cfs_b->lock is available while we wait */
		hrtimer_cancel(&cfs_b->period_timer);

		raw_spin_lock(&cfs_b->lock);
		/* if someone else restarted the timer then we're done */
		if (cfs_b->timer_active)
			return;
	}

	cfs_b->timer_active = 1;
	start_bandwidth_timer(&cfs_b->period_timer, cfs_b->period);
}

static void destroy_cfs_bandwidth(struct cfs_bandwidth *cfs_b)
{
	hrtimer_cancel(&cfs_b->period_timer);
	hrtimer_cancel(&cfs_b->slack_timer);
}

2700
static void __maybe_unused unthrottle_offline_cfs_rqs(struct rq *rq)
2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720
{
	struct cfs_rq *cfs_rq;

	for_each_leaf_cfs_rq(rq, cfs_rq) {
		struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(cfs_rq->tg);

		if (!cfs_rq->runtime_enabled)
			continue;

		/*
		 * clock_task is not advancing so we just need to make sure
		 * there's some valid quota amount
		 */
		cfs_rq->runtime_remaining = cfs_b->quota;
		if (cfs_rq_throttled(cfs_rq))
			unthrottle_cfs_rq(cfs_rq);
	}
}

#else /* CONFIG_CFS_BANDWIDTH */
2721 2722
static inline u64 cfs_rq_clock_task(struct cfs_rq *cfs_rq)
{
2723
	return rq_clock_task(rq_of(cfs_rq));
2724 2725 2726 2727
}

static void account_cfs_rq_runtime(struct cfs_rq *cfs_rq,
				     unsigned long delta_exec) {}
2728 2729
static void check_cfs_rq_runtime(struct cfs_rq *cfs_rq) {}
static void check_enqueue_throttle(struct cfs_rq *cfs_rq) {}
2730
static __always_inline void return_cfs_rq_runtime(struct cfs_rq *cfs_rq) {}
2731 2732 2733 2734 2735

static inline int cfs_rq_throttled(struct cfs_rq *cfs_rq)
{
	return 0;
}
2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746

static inline int throttled_hierarchy(struct cfs_rq *cfs_rq)
{
	return 0;
}

static inline int throttled_lb_pair(struct task_group *tg,
				    int src_cpu, int dest_cpu)
{
	return 0;
}
2747 2748 2749 2750 2751

void init_cfs_bandwidth(struct cfs_bandwidth *cfs_b) {}

#ifdef CONFIG_FAIR_GROUP_SCHED
static void init_cfs_rq_runtime(struct cfs_rq *cfs_rq) {}
2752 2753
#endif

2754 2755 2756 2757 2758
static inline struct cfs_bandwidth *tg_cfs_bandwidth(struct task_group *tg)
{
	return NULL;
}
static inline void destroy_cfs_bandwidth(struct cfs_bandwidth *cfs_b) {}
2759
static inline void unthrottle_offline_cfs_rqs(struct rq *rq) {}
2760 2761 2762

#endif /* CONFIG_CFS_BANDWIDTH */

2763 2764 2765 2766
/**************************************************
 * CFS operations on tasks:
 */

2767 2768 2769 2770 2771 2772 2773 2774
#ifdef CONFIG_SCHED_HRTICK
static void hrtick_start_fair(struct rq *rq, struct task_struct *p)
{
	struct sched_entity *se = &p->se;
	struct cfs_rq *cfs_rq = cfs_rq_of(se);

	WARN_ON(task_rq(p) != rq);

2775
	if (cfs_rq->nr_running > 1) {
2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789
		u64 slice = sched_slice(cfs_rq, se);
		u64 ran = se->sum_exec_runtime - se->prev_sum_exec_runtime;
		s64 delta = slice - ran;

		if (delta < 0) {
			if (rq->curr == p)
				resched_task(p);
			return;
		}

		/*
		 * Don't schedule slices shorter than 10000ns, that just
		 * doesn't make sense. Rely on vruntime for fairness.
		 */
2790
		if (rq->curr != p)
2791
			delta = max_t(s64, 10000LL, delta);
2792

2793
		hrtick_start(rq, delta);
2794 2795
	}
}
2796 2797 2798 2799 2800 2801 2802 2803 2804 2805

/*
 * called from enqueue/dequeue and updates the hrtick when the
 * current task is from our class and nr_running is low enough
 * to matter.
 */
static void hrtick_update(struct rq *rq)
{
	struct task_struct *curr = rq->curr;

2806
	if (!hrtick_enabled(rq) || curr->sched_class != &fair_sched_class)
2807 2808 2809 2810 2811
		return;

	if (cfs_rq_of(&curr->se)->nr_running < sched_nr_latency)
		hrtick_start_fair(rq, curr);
}
2812
#else /* !CONFIG_SCHED_HRTICK */
2813 2814 2815 2816
static inline void
hrtick_start_fair(struct rq *rq, struct task_struct *p)
{
}
2817 2818 2819 2820

static inline void hrtick_update(struct rq *rq)
{
}
2821 2822
#endif

2823 2824 2825 2826 2827
/*
 * The enqueue_task method is called before nr_running is
 * increased. Here we update the fair scheduling stats and
 * then put the task into the rbtree:
 */
2828
static void
2829
enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags)
2830 2831
{
	struct cfs_rq *cfs_rq;
2832
	struct sched_entity *se = &p->se;
2833 2834

	for_each_sched_entity(se) {
2835
		if (se->on_rq)
2836 2837
			break;
		cfs_rq = cfs_rq_of(se);
2838
		enqueue_entity(cfs_rq, se, flags);
2839 2840 2841 2842 2843 2844 2845 2846 2847

		/*
		 * end evaluation on encountering a throttled cfs_rq
		 *
		 * note: in the case of encountering a throttled cfs_rq we will
		 * post the final h_nr_running increment below.
		*/
		if (cfs_rq_throttled(cfs_rq))
			break;
2848
		cfs_rq->h_nr_running++;
2849

2850
		flags = ENQUEUE_WAKEUP;
2851
	}
2852

2853
	for_each_sched_entity(se) {
2854
		cfs_rq = cfs_rq_of(se);
2855
		cfs_rq->h_nr_running++;
2856

2857 2858 2859
		if (cfs_rq_throttled(cfs_rq))
			break;

2860
		update_cfs_shares(cfs_rq);
2861
		update_entity_load_avg(se, 1);
2862 2863
	}

2864 2865
	if (!se) {
		update_rq_runnable_avg(rq, rq->nr_running);
2866
		inc_nr_running(rq);
2867
	}
2868
	hrtick_update(rq);
2869 2870
}

2871 2872
static void set_next_buddy(struct sched_entity *se);

2873 2874 2875 2876 2877
/*
 * The dequeue_task method is called before nr_running is
 * decreased. We remove the task from the rbtree and
 * update the fair scheduling stats:
 */
2878
static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags)
2879 2880
{
	struct cfs_rq *cfs_rq;
2881
	struct sched_entity *se = &p->se;
2882
	int task_sleep = flags & DEQUEUE_SLEEP;
2883 2884 2885

	for_each_sched_entity(se) {
		cfs_rq = cfs_rq_of(se);
2886
		dequeue_entity(cfs_rq, se, flags);
2887 2888 2889 2890 2891 2892 2893 2894 2895

		/*
		 * end evaluation on encountering a throttled cfs_rq
		 *
		 * note: in the case of encountering a throttled cfs_rq we will
		 * post the final h_nr_running decrement below.
		*/
		if (cfs_rq_throttled(cfs_rq))
			break;
2896
		cfs_rq->h_nr_running--;
2897

2898
		/* Don't dequeue parent if it has other entities besides us */
2899 2900 2901 2902 2903 2904 2905
		if (cfs_rq->load.weight) {
			/*
			 * Bias pick_next to pick a task from this cfs_rq, as
			 * p is sleeping when it is within its sched_slice.
			 */
			if (task_sleep && parent_entity(se))
				set_next_buddy(parent_entity(se));
2906 2907 2908

			/* avoid re-evaluating load for this entity */
			se = parent_entity(se);
2909
			break;
2910
		}
2911
		flags |= DEQUEUE_SLEEP;
2912
	}
2913

2914
	for_each_sched_entity(se) {
2915
		cfs_rq = cfs_rq_of(se);
2916
		cfs_rq->h_nr_running--;
2917

2918 2919 2920
		if (cfs_rq_throttled(cfs_rq))
			break;

2921
		update_cfs_shares(cfs_rq);
2922
		update_entity_load_avg(se, 1);
2923 2924
	}

2925
	if (!se) {
2926
		dec_nr_running(rq);
2927 2928
		update_rq_runnable_avg(rq, 1);
	}
2929
	hrtick_update(rq);
2930 2931
}

2932
#ifdef CONFIG_SMP
2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987
/* Used instead of source_load when we know the type == 0 */
static unsigned long weighted_cpuload(const int cpu)
{
	return cpu_rq(cpu)->load.weight;
}

/*
 * Return a low guess at the load of a migration-source cpu weighted
 * according to the scheduling class and "nice" value.
 *
 * We want to under-estimate the load of migration sources, to
 * balance conservatively.
 */
static unsigned long source_load(int cpu, int type)
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long total = weighted_cpuload(cpu);

	if (type == 0 || !sched_feat(LB_BIAS))
		return total;

	return min(rq->cpu_load[type-1], total);
}

/*
 * Return a high guess at the load of a migration-target cpu weighted
 * according to the scheduling class and "nice" value.
 */
static unsigned long target_load(int cpu, int type)
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long total = weighted_cpuload(cpu);

	if (type == 0 || !sched_feat(LB_BIAS))
		return total;

	return max(rq->cpu_load[type-1], total);
}

static unsigned long power_of(int cpu)
{
	return cpu_rq(cpu)->cpu_power;
}

static unsigned long cpu_avg_load_per_task(int cpu)
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long nr_running = ACCESS_ONCE(rq->nr_running);

	if (nr_running)
		return rq->load.weight / nr_running;

	return 0;
}

2988

2989
static void task_waking_fair(struct task_struct *p)
2990 2991 2992
{
	struct sched_entity *se = &p->se;
	struct cfs_rq *cfs_rq = cfs_rq_of(se);
2993 2994 2995 2996
	u64 min_vruntime;

#ifndef CONFIG_64BIT
	u64 min_vruntime_copy;
2997

2998 2999 3000 3001 3002 3003 3004 3005
	do {
		min_vruntime_copy = cfs_rq->min_vruntime_copy;
		smp_rmb();
		min_vruntime = cfs_rq->min_vruntime;
	} while (min_vruntime != min_vruntime_copy);
#else
	min_vruntime = cfs_rq->min_vruntime;
#endif
3006

3007
	se->vruntime -= min_vruntime;
3008 3009
}

3010
#ifdef CONFIG_FAIR_GROUP_SCHED
3011 3012 3013 3014 3015 3016
/*
 * effective_load() calculates the load change as seen from the root_task_group
 *
 * Adding load to a group doesn't make a group heavier, but can cause movement
 * of group shares between cpus. Assuming the shares were perfectly aligned one
 * can calculate the shift in shares.
3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059
 *
 * Calculate the effective load difference if @wl is added (subtracted) to @tg
 * on this @cpu and results in a total addition (subtraction) of @wg to the
 * total group weight.
 *
 * Given a runqueue weight distribution (rw_i) we can compute a shares
 * distribution (s_i) using:
 *
 *   s_i = rw_i / \Sum rw_j						(1)
 *
 * Suppose we have 4 CPUs and our @tg is a direct child of the root group and
 * has 7 equal weight tasks, distributed as below (rw_i), with the resulting
 * shares distribution (s_i):
 *
 *   rw_i = {   2,   4,   1,   0 }
 *   s_i  = { 2/7, 4/7, 1/7,   0 }
 *
 * As per wake_affine() we're interested in the load of two CPUs (the CPU the
 * task used to run on and the CPU the waker is running on), we need to
 * compute the effect of waking a task on either CPU and, in case of a sync
 * wakeup, compute the effect of the current task going to sleep.
 *
 * So for a change of @wl to the local @cpu with an overall group weight change
 * of @wl we can compute the new shares distribution (s'_i) using:
 *
 *   s'_i = (rw_i + @wl) / (@wg + \Sum rw_j)				(2)
 *
 * Suppose we're interested in CPUs 0 and 1, and want to compute the load
 * differences in waking a task to CPU 0. The additional task changes the
 * weight and shares distributions like:
 *
 *   rw'_i = {   3,   4,   1,   0 }
 *   s'_i  = { 3/8, 4/8, 1/8,   0 }
 *
 * We can then compute the difference in effective weight by using:
 *
 *   dw_i = S * (s'_i - s_i)						(3)
 *
 * Where 'S' is the group weight as seen by its parent.
 *
 * Therefore the effective change in loads on CPU 0 would be 5/56 (3/8 - 2/7)
 * times the weight of the group. The effect on CPU 1 would be -4/56 (4/8 -
 * 4/7) times the weight of the group.
3060
 */
3061
static long effective_load(struct task_group *tg, int cpu, long wl, long wg)
3062
{
Peter Zijlstra's avatar
Peter Zijlstra committed
3063
	struct sched_entity *se = tg->se[cpu];
3064

3065
	if (!tg->parent)	/* the trivial, non-cgroup case */
3066 3067
		return wl;

Peter Zijlstra's avatar
Peter Zijlstra committed
3068
	for_each_sched_entity(se) {
3069
		long w, W;
Peter Zijlstra's avatar
Peter Zijlstra committed
3070

3071
		tg = se->my_q->tg;
3072

3073 3074 3075 3076
		/*
		 * W = @wg + \Sum rw_j
		 */
		W = wg + calc_tg_weight(tg, se->my_q);
Peter Zijlstra's avatar
Peter Zijlstra committed
3077

3078 3079 3080 3081
		/*
		 * w = rw_i + @wl
		 */
		w = se->my_q->load.weight + wl;
3082

3083 3084 3085 3086 3087
		/*
		 * wl = S * s'_i; see (2)
		 */
		if (W > 0 && w < W)
			wl = (w * tg->shares) / W;
3088 3089
		else
			wl = tg->shares;
3090

3091 3092 3093 3094 3095
		/*
		 * Per the above, wl is the new se->load.weight value; since
		 * those are clipped to [MIN_SHARES, ...) do so now. See
		 * calc_cfs_shares().
		 */
3096 3097
		if (wl < MIN_SHARES)
			wl = MIN_SHARES;
3098 3099 3100 3101

		/*
		 * wl = dw_i = S * (s'_i - s_i); see (3)
		 */
3102
		wl -= se->load.weight;
3103 3104 3105 3106 3107 3108 3109 3110

		/*
		 * Recursively apply this logic to all parent groups to compute
		 * the final effective load change on the root group. Since
		 * only the @tg group gets extra weight, all parent groups can
		 * only redistribute existing shares. @wl is the shift in shares
		 * resulting from this level per the above.
		 */
Peter Zijlstra's avatar
Peter Zijlstra committed
3111 3112
		wg = 0;
	}
3113

Peter Zijlstra's avatar
Peter Zijlstra committed
3114
	return wl;
3115 3116
}
#else
Peter Zijlstra's avatar
Peter Zijlstra committed
3117

3118 3119
static inline unsigned long effective_load(struct task_group *tg, int cpu,
		unsigned long wl, unsigned long wg)
Peter Zijlstra's avatar
Peter Zijlstra committed
3120
{
3121
	return wl;
3122
}
Peter Zijlstra's avatar
Peter Zijlstra committed
3123

3124 3125
#endif

3126
static int wake_affine(struct sched_domain *sd, struct task_struct *p, int sync)
3127
{
3128
	s64 this_load, load;
3129
	int idx, this_cpu, prev_cpu;
3130
	unsigned long tl_per_task;
3131
	struct task_group *tg;
3132
	unsigned long weight;
3133
	int balanced;
3134

3135 3136 3137 3138 3139
	idx	  = sd->wake_idx;
	this_cpu  = smp_processor_id();
	prev_cpu  = task_cpu(p);
	load	  = source_load(prev_cpu, idx);
	this_load = target_load(this_cpu, idx);
3140

3141 3142 3143 3144 3145
	/*
	 * If sync wakeup then subtract the (maximum possible)
	 * effect of the currently running task from the load
	 * of the current CPU:
	 */
3146 3147 3148 3149
	if (sync) {
		tg = task_group(current);
		weight = current->se.load.weight;

3150
		this_load += effective_load(tg, this_cpu, -weight, -weight);
3151 3152
		load += effective_load(tg, prev_cpu, 0, -weight);
	}
3153

3154 3155
	tg = task_group(p);
	weight = p->se.load.weight;
3156

3157 3158
	/*
	 * In low-load situations, where prev_cpu is idle and this_cpu is idle
3159 3160 3161
	 * due to the sync cause above having dropped this_load to 0, we'll
	 * always have an imbalance, but there's really nothing you can do
	 * about that, so that's good too.
3162 3163 3164 3165
	 *
	 * Otherwise check if either cpus are near enough in load to allow this
	 * task to be woken on this_cpu.
	 */
3166 3167
	if (this_load > 0) {
		s64 this_eff_load, prev_eff_load;
3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180

		this_eff_load = 100;
		this_eff_load *= power_of(prev_cpu);
		this_eff_load *= this_load +
			effective_load(tg, this_cpu, weight, weight);

		prev_eff_load = 100 + (sd->imbalance_pct - 100) / 2;
		prev_eff_load *= power_of(this_cpu);
		prev_eff_load *= load + effective_load(tg, prev_cpu, 0, weight);

		balanced = this_eff_load <= prev_eff_load;
	} else
		balanced = true;
3181

3182
	/*
Ingo Molnar's avatar
Ingo Molnar committed
3183 3184 3185
	 * If the currently running task will sleep within
	 * a reasonable amount of time then attract this newly
	 * woken task:
3186
	 */
3187 3188
	if (sync && balanced)
		return 1;
3189

3190
	schedstat_inc(p, se.statistics.nr_wakeups_affine_attempts);
3191 3192
	tl_per_task = cpu_avg_load_per_task(this_cpu);

3193 3194 3195
	if (balanced ||
	    (this_load <= load &&
	     this_load + target_load(prev_cpu, idx) <= tl_per_task)) {
3196 3197 3198 3199 3200
		/*
		 * This domain has SD_WAKE_AFFINE and
		 * p is cache cold in this domain, and
		 * there is no bad imbalance.
		 */
3201
		schedstat_inc(sd, ttwu_move_affine);
3202
		schedstat_inc(p, se.statistics.nr_wakeups_affine);
3203 3204 3205 3206 3207 3208

		return 1;
	}
	return 0;
}

3209 3210 3211 3212 3213
/*
 * find_idlest_group finds and returns the least busy CPU group within the
 * domain.
 */
static struct sched_group *
Peter Zijlstra's avatar
Peter Zijlstra committed
3214
find_idlest_group(struct sched_domain *sd, struct task_struct *p,
3215
		  int this_cpu, int load_idx)
3216
{
3217
	struct sched_group *idlest = NULL, *group = sd->groups;
3218 3219
	unsigned long min_load = ULONG_MAX, this_load = 0;
	int imbalance = 100 + (sd->imbalance_pct-100)/2;
3220

3221 3222 3223 3224
	do {
		unsigned long load, avg_load;
		int local_group;
		int i;
3225

3226 3227
		/* Skip over this group if it has no CPUs allowed */
		if (!cpumask_intersects(sched_group_cpus(group),
3228
					tsk_cpus_allowed(p)))
3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247
			continue;

		local_group = cpumask_test_cpu(this_cpu,
					       sched_group_cpus(group));

		/* Tally up the load of all CPUs in the group */
		avg_load = 0;

		for_each_cpu(i, sched_group_cpus(group)) {
			/* Bias balancing toward cpus of our domain */
			if (local_group)
				load = source_load(i, load_idx);
			else
				load = target_load(i, load_idx);

			avg_load += load;
		}

		/* Adjust by relative CPU power of the group */
3248
		avg_load = (avg_load * SCHED_POWER_SCALE) / group->sgp->power;
3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273

		if (local_group) {
			this_load = avg_load;
		} else if (avg_load < min_load) {
			min_load = avg_load;
			idlest = group;
		}
	} while (group = group->next, group != sd->groups);

	if (!idlest || 100*this_load < imbalance*min_load)
		return NULL;
	return idlest;
}

/*
 * find_idlest_cpu - find the idlest cpu among the cpus in group.
 */
static int
find_idlest_cpu(struct sched_group *group, struct task_struct *p, int this_cpu)
{
	unsigned long load, min_load = ULONG_MAX;
	int idlest = -1;
	int i;

	/* Traverse only the allowed CPUs */
3274
	for_each_cpu_and(i, sched_group_cpus(group), tsk_cpus_allowed(p)) {
3275 3276 3277 3278 3279
		load = weighted_cpuload(i);

		if (load < min_load || (load == min_load && i == this_cpu)) {
			min_load = load;
			idlest = i;
3280 3281 3282
		}
	}

3283 3284
	return idlest;
}
3285

3286 3287 3288
/*
 * Try and locate an idle CPU in the sched_domain.
 */
3289
static int select_idle_sibling(struct task_struct *p, int target)
3290
{
3291
	struct sched_domain *sd;
3292
	struct sched_group *sg;
3293
	int i = task_cpu(p);
3294

3295 3296
	if (idle_cpu(target))
		return target;
3297 3298

	/*
3299
	 * If the prevous cpu is cache affine and idle, don't be stupid.
3300
	 */
3301 3302
	if (i != target && cpus_share_cache(i, target) && idle_cpu(i))
		return i;
3303 3304

	/*
3305
	 * Otherwise, iterate the domains and find an elegible idle cpu.
3306
	 */
3307
	sd = rcu_dereference(per_cpu(sd_llc, target));
3308
	for_each_lower_domain(sd) {
3309 3310 3311 3312 3313 3314 3315
		sg = sd->groups;
		do {
			if (!cpumask_intersects(sched_group_cpus(sg),
						tsk_cpus_allowed(p)))
				goto next;

			for_each_cpu(i, sched_group_cpus(sg)) {
3316
				if (i == target || !idle_cpu(i))
3317 3318
					goto next;
			}
3319

3320 3321 3322 3323 3324 3325 3326 3327
			target = cpumask_first_and(sched_group_cpus(sg),
					tsk_cpus_allowed(p));
			goto done;
next:
			sg = sg->next;
		} while (sg != sd->groups);
	}
done:
3328 3329 3330
	return target;
}

3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341
/*
 * sched_balance_self: balance the current task (running on cpu) in domains
 * that have the 'flag' flag set. In practice, this is SD_BALANCE_FORK and
 * SD_BALANCE_EXEC.
 *
 * Balance, ie. select the least loaded group.
 *
 * Returns the target CPU number, or the same CPU if no balancing is needed.
 *
 * preempt must be disabled.
 */
3342
static int
3343
select_task_rq_fair(struct task_struct *p, int sd_flag, int wake_flags)
3344
{
3345
	struct sched_domain *tmp, *affine_sd = NULL, *sd = NULL;
3346 3347 3348
	int cpu = smp_processor_id();
	int prev_cpu = task_cpu(p);
	int new_cpu = cpu;
3349
	int want_affine = 0;
3350
	int sync = wake_flags & WF_SYNC;
3351

3352
	if (p->nr_cpus_allowed == 1)
3353 3354
		return prev_cpu;

3355
	if (sd_flag & SD_BALANCE_WAKE) {
3356
		if (cpumask_test_cpu(cpu, tsk_cpus_allowed(p)))
3357 3358 3359
			want_affine = 1;
		new_cpu = prev_cpu;
	}
3360

3361
	rcu_read_lock();
3362
	for_each_domain(cpu, tmp) {
3363 3364 3365
		if (!(tmp->flags & SD_LOAD_BALANCE))
			continue;

3366
		/*
3367 3368
		 * If both cpu and prev_cpu are part of this domain,
		 * cpu is a valid SD_WAKE_AFFINE target.
3369
		 */
3370 3371 3372
		if (want_affine && (tmp->flags & SD_WAKE_AFFINE) &&
		    cpumask_test_cpu(prev_cpu, sched_domain_span(tmp))) {
			affine_sd = tmp;
3373
			break;
3374
		}
3375

3376
		if (tmp->flags & sd_flag)
3377 3378 3379
			sd = tmp;
	}

3380
	if (affine_sd) {
3381
		if (cpu != prev_cpu && wake_affine(affine_sd, p, sync))
3382 3383 3384 3385
			prev_cpu = cpu;

		new_cpu = select_idle_sibling(p, prev_cpu);
		goto unlock;
3386
	}
3387

3388
	while (sd) {
3389
		int load_idx = sd->forkexec_idx;
3390
		struct sched_group *group;
3391
		int weight;
3392

3393
		if (!(sd->flags & sd_flag)) {
3394 3395 3396
			sd = sd->child;
			continue;
		}
3397

3398 3399
		if (sd_flag & SD_BALANCE_WAKE)
			load_idx = sd->wake_idx;
3400

3401
		group = find_idlest_group(sd, p, cpu, load_idx);
3402 3403 3404 3405
		if (!group) {
			sd = sd->child;
			continue;
		}
Ingo Molnar's avatar
Ingo Molnar committed
3406

3407
		new_cpu = find_idlest_cpu(group, p, cpu);
3408 3409 3410 3411
		if (new_cpu == -1 || new_cpu == cpu) {
			/* Now try balancing at a lower domain level of cpu */
			sd = sd->child;
			continue;
3412
		}
3413 3414 3415

		/* Now try balancing at a lower domain level of new_cpu */
		cpu = new_cpu;
3416
		weight = sd->span_weight;
3417 3418
		sd = NULL;
		for_each_domain(cpu, tmp) {
3419
			if (weight <= tmp->span_weight)
3420
				break;
3421
			if (tmp->flags & sd_flag)
3422 3423 3424
				sd = tmp;
		}
		/* while loop will break here if sd == NULL */
3425
	}
3426 3427
unlock:
	rcu_read_unlock();
3428

3429
	return new_cpu;
3430
}
3431 3432 3433 3434 3435 3436 3437 3438 3439 3440

/*
 * Called immediately before a task is migrated to a new cpu; task_cpu(p) and
 * cfs_rq_of(p) references at time of call are still valid and identify the
 * previous cpu.  However, the caller only guarantees p->pi_lock is held; no
 * other assumptions, including the state of rq->lock, should be made.
 */
static void
migrate_task_rq_fair(struct task_struct *p, int next_cpu)
{
3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453
	struct sched_entity *se = &p->se;
	struct cfs_rq *cfs_rq = cfs_rq_of(se);

	/*
	 * Load tracking: accumulate removed load so that it can be processed
	 * when we next update owning cfs_rq under rq->lock.  Tasks contribute
	 * to blocked load iff they have a positive decay-count.  It can never
	 * be negative here since on-rq tasks have decay-count == 0.
	 */
	if (se->avg.decay_count) {
		se->avg.decay_count = -__synchronize_entity_decay(se);
		atomic64_add(se->avg.load_avg_contrib, &cfs_rq->removed_load);
	}
3454
}
3455 3456
#endif /* CONFIG_SMP */

Peter Zijlstra's avatar
Peter Zijlstra committed
3457 3458
static unsigned long
wakeup_gran(struct sched_entity *curr, struct sched_entity *se)
3459 3460 3461 3462
{
	unsigned long gran = sysctl_sched_wakeup_granularity;

	/*
Peter Zijlstra's avatar
Peter Zijlstra committed
3463 3464
	 * Since its curr running now, convert the gran from real-time
	 * to virtual-time in his units.
3465 3466 3467 3468 3469 3470 3471 3472 3473
	 *
	 * By using 'se' instead of 'curr' we penalize light tasks, so
	 * they get preempted easier. That is, if 'se' < 'curr' then
	 * the resulting gran will be larger, therefore penalizing the
	 * lighter, if otoh 'se' > 'curr' then the resulting gran will
	 * be smaller, again penalizing the lighter task.
	 *
	 * This is especially important for buddies when the leftmost
	 * task is higher priority than the buddy.
3474
	 */
3475
	return calc_delta_fair(gran, se);
3476 3477
}

3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499
/*
 * Should 'se' preempt 'curr'.
 *
 *             |s1
 *        |s2
 *   |s3
 *         g
 *      |<--->|c
 *
 *  w(c, s1) = -1
 *  w(c, s2) =  0
 *  w(c, s3) =  1
 *
 */
static int
wakeup_preempt_entity(struct sched_entity *curr, struct sched_entity *se)
{
	s64 gran, vdiff = curr->vruntime - se->vruntime;

	if (vdiff <= 0)
		return -1;

Peter Zijlstra's avatar
Peter Zijlstra committed
3500
	gran = wakeup_gran(curr, se);
3501 3502 3503 3504 3505 3506
	if (vdiff > gran)
		return 1;

	return 0;
}

3507 3508
static void set_last_buddy(struct sched_entity *se)
{
3509 3510 3511 3512 3513
	if (entity_is_task(se) && unlikely(task_of(se)->policy == SCHED_IDLE))
		return;

	for_each_sched_entity(se)
		cfs_rq_of(se)->last = se;
3514 3515 3516 3517
}

static void set_next_buddy(struct sched_entity *se)
{
3518 3519 3520 3521 3522
	if (entity_is_task(se) && unlikely(task_of(se)->policy == SCHED_IDLE))
		return;

	for_each_sched_entity(se)
		cfs_rq_of(se)->next = se;
3523 3524
}

3525 3526
static void set_skip_buddy(struct sched_entity *se)
{
3527 3528
	for_each_sched_entity(se)
		cfs_rq_of(se)->skip = se;
3529 3530
}

3531 3532 3533
/*
 * Preempt the current task with a newly woken task if needed:
 */
3534
static void check_preempt_wakeup(struct rq *rq, struct task_struct *p, int wake_flags)
3535 3536
{
	struct task_struct *curr = rq->curr;
3537
	struct sched_entity *se = &curr->se, *pse = &p->se;
3538
	struct cfs_rq *cfs_rq = task_cfs_rq(curr);
3539
	int scale = cfs_rq->nr_running >= sched_nr_latency;
3540
	int next_buddy_marked = 0;
3541

Ingo Molnar's avatar
Ingo Molnar committed
3542 3543 3544
	if (unlikely(se == pse))
		return;

3545
	/*
3546
	 * This is possible from callers such as move_task(), in which we
3547 3548 3549 3550 3551 3552 3553
	 * unconditionally check_prempt_curr() after an enqueue (which may have
	 * lead to a throttle).  This both saves work and prevents false
	 * next-buddy nomination below.
	 */
	if (unlikely(throttled_hierarchy(cfs_rq_of(pse))))
		return;

3554
	if (sched_feat(NEXT_BUDDY) && scale && !(wake_flags & WF_FORK)) {
3555
		set_next_buddy(pse);
3556 3557
		next_buddy_marked = 1;
	}
3558

3559 3560 3561
	/*
	 * We can come here with TIF_NEED_RESCHED already set from new task
	 * wake up path.
3562 3563 3564 3565 3566 3567
	 *
	 * Note: this also catches the edge-case of curr being in a throttled
	 * group (e.g. via set_curr_task), since update_curr() (in the
	 * enqueue of curr) will have resulted in resched being set.  This
	 * prevents us from potentially nominating it as a false LAST_BUDDY
	 * below.
3568 3569 3570 3571
	 */
	if (test_tsk_need_resched(curr))
		return;

3572 3573 3574 3575 3576
	/* Idle tasks are by definition preempted by non-idle tasks. */
	if (unlikely(curr->policy == SCHED_IDLE) &&
	    likely(p->policy != SCHED_IDLE))
		goto preempt;

3577
	/*
3578 3579
	 * Batch and idle tasks do not preempt non-idle tasks (their preemption
	 * is driven by the tick):
3580
	 */
3581
	if (unlikely(p->policy != SCHED_NORMAL) || !sched_feat(WAKEUP_PREEMPTION))
3582
		return;
3583

3584
	find_matching_se(&se, &pse);
3585
	update_curr(cfs_rq_of(se));
3586
	BUG_ON(!pse);
3587 3588 3589 3590 3591 3592 3593
	if (wakeup_preempt_entity(se, pse) == 1) {
		/*
		 * Bias pick_next to pick the sched entity that is
		 * triggering this preemption.
		 */
		if (!next_buddy_marked)
			set_next_buddy(pse);
3594
		goto preempt;
3595
	}
3596

3597
	return;
3598

3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614
preempt:
	resched_task(curr);
	/*
	 * Only set the backward buddy when the current task is still
	 * on the rq. This can happen when a wakeup gets interleaved
	 * with schedule on the ->pre_schedule() or idle_balance()
	 * point, either of which can * drop the rq lock.
	 *
	 * Also, during early boot the idle thread is in the fair class,
	 * for obvious reasons its a bad idea to schedule back to it.
	 */
	if (unlikely(!se->on_rq || curr == rq->idle))
		return;

	if (sched_feat(LAST_BUDDY) && scale && entity_is_task(se))
		set_last_buddy(se);
3615 3616
}

3617
static struct task_struct *pick_next_task_fair(struct rq *rq)
3618
{
3619
	struct task_struct *p;
3620 3621 3622
	struct cfs_rq *cfs_rq = &rq->cfs;
	struct sched_entity *se;

3623
	if (!cfs_rq->nr_running)
3624 3625 3626
		return NULL;

	do {
3627
		se = pick_next_entity(cfs_rq);
3628
		set_next_entity(cfs_rq, se);
3629 3630 3631
		cfs_rq = group_cfs_rq(se);
	} while (cfs_rq);

3632
	p = task_of(se);
3633 3634
	if (hrtick_enabled(rq))
		hrtick_start_fair(rq, p);
3635 3636

	return p;
3637 3638 3639 3640 3641
}

/*
 * Account for a descheduled task:
 */
3642
static void put_prev_task_fair(struct rq *rq, struct task_struct *prev)
3643 3644 3645 3646 3647 3648
{
	struct sched_entity *se = &prev->se;
	struct cfs_rq *cfs_rq;

	for_each_sched_entity(se) {
		cfs_rq = cfs_rq_of(se);
3649
		put_prev_entity(cfs_rq, se);
3650 3651 3652
	}
}

3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677
/*
 * sched_yield() is very simple
 *
 * The magic of dealing with the ->skip buddy is in pick_next_entity.
 */
static void yield_task_fair(struct rq *rq)
{
	struct task_struct *curr = rq->curr;
	struct cfs_rq *cfs_rq = task_cfs_rq(curr);
	struct sched_entity *se = &curr->se;

	/*
	 * Are we the only task in the tree?
	 */
	if (unlikely(rq->nr_running == 1))
		return;

	clear_buddies(cfs_rq, se);

	if (curr->policy != SCHED_BATCH) {
		update_rq_clock(rq);
		/*
		 * Update run-time statistics of the 'current'.
		 */
		update_curr(cfs_rq);
3678 3679 3680 3681 3682 3683
		/*
		 * Tell update_rq_clock() that we've just updated,
		 * so we don't do microscopic update in schedule()
		 * and double the fastpath cost.
		 */
		 rq->skip_clock_update = 1;
3684 3685 3686 3687 3688
	}

	set_skip_buddy(se);
}

3689 3690 3691 3692
static bool yield_to_task_fair(struct rq *rq, struct task_struct *p, bool preempt)
{
	struct sched_entity *se = &p->se;

3693 3694
	/* throttled hierarchies are not runnable */
	if (!se->on_rq || throttled_hierarchy(cfs_rq_of(se)))
3695 3696 3697 3698 3699 3700 3701 3702 3703 3704
		return false;

	/* Tell the scheduler that we'd really like pse to run next. */
	set_next_buddy(se);

	yield_task_fair(rq);

	return true;
}

3705
#ifdef CONFIG_SMP
3706
/**************************************************
3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822
 * Fair scheduling class load-balancing methods.
 *
 * BASICS
 *
 * The purpose of load-balancing is to achieve the same basic fairness the
 * per-cpu scheduler provides, namely provide a proportional amount of compute
 * time to each task. This is expressed in the following equation:
 *
 *   W_i,n/P_i == W_j,n/P_j for all i,j                               (1)
 *
 * Where W_i,n is the n-th weight average for cpu i. The instantaneous weight
 * W_i,0 is defined as:
 *
 *   W_i,0 = \Sum_j w_i,j                                             (2)
 *
 * Where w_i,j is the weight of the j-th runnable task on cpu i. This weight
 * is derived from the nice value as per prio_to_weight[].
 *
 * The weight average is an exponential decay average of the instantaneous
 * weight:
 *
 *   W'_i,n = (2^n - 1) / 2^n * W_i,n + 1 / 2^n * W_i,0               (3)
 *
 * P_i is the cpu power (or compute capacity) of cpu i, typically it is the
 * fraction of 'recent' time available for SCHED_OTHER task execution. But it
 * can also include other factors [XXX].
 *
 * To achieve this balance we define a measure of imbalance which follows
 * directly from (1):
 *
 *   imb_i,j = max{ avg(W/P), W_i/P_i } - min{ avg(W/P), W_j/P_j }    (4)
 *
 * We them move tasks around to minimize the imbalance. In the continuous
 * function space it is obvious this converges, in the discrete case we get
 * a few fun cases generally called infeasible weight scenarios.
 *
 * [XXX expand on:
 *     - infeasible weights;
 *     - local vs global optima in the discrete case. ]
 *
 *
 * SCHED DOMAINS
 *
 * In order to solve the imbalance equation (4), and avoid the obvious O(n^2)
 * for all i,j solution, we create a tree of cpus that follows the hardware
 * topology where each level pairs two lower groups (or better). This results
 * in O(log n) layers. Furthermore we reduce the number of cpus going up the
 * tree to only the first of the previous level and we decrease the frequency
 * of load-balance at each level inv. proportional to the number of cpus in
 * the groups.
 *
 * This yields:
 *
 *     log_2 n     1     n
 *   \Sum       { --- * --- * 2^i } = O(n)                            (5)
 *     i = 0      2^i   2^i
 *                               `- size of each group
 *         |         |     `- number of cpus doing load-balance
 *         |         `- freq
 *         `- sum over all levels
 *
 * Coupled with a limit on how many tasks we can migrate every balance pass,
 * this makes (5) the runtime complexity of the balancer.
 *
 * An important property here is that each CPU is still (indirectly) connected
 * to every other cpu in at most O(log n) steps:
 *
 * The adjacency matrix of the resulting graph is given by:
 *
 *             log_2 n     
 *   A_i,j = \Union     (i % 2^k == 0) && i / 2^(k+1) == j / 2^(k+1)  (6)
 *             k = 0
 *
 * And you'll find that:
 *
 *   A^(log_2 n)_i,j != 0  for all i,j                                (7)
 *
 * Showing there's indeed a path between every cpu in at most O(log n) steps.
 * The task movement gives a factor of O(m), giving a convergence complexity
 * of:
 *
 *   O(nm log n),  n := nr_cpus, m := nr_tasks                        (8)
 *
 *
 * WORK CONSERVING
 *
 * In order to avoid CPUs going idle while there's still work to do, new idle
 * balancing is more aggressive and has the newly idle cpu iterate up the domain
 * tree itself instead of relying on other CPUs to bring it work.
 *
 * This adds some complexity to both (5) and (8) but it reduces the total idle
 * time.
 *
 * [XXX more?]
 *
 *
 * CGROUPS
 *
 * Cgroups make a horror show out of (2), instead of a simple sum we get:
 *
 *                                s_k,i
 *   W_i,0 = \Sum_j \Prod_k w_k * -----                               (9)
 *                                 S_k
 *
 * Where
 *
 *   s_k,i = \Sum_j w_i,j,k  and  S_k = \Sum_i s_k,i                 (10)
 *
 * w_i,j,k is the weight of the j-th runnable task in the k-th cgroup on cpu i.
 *
 * The big problem is S_k, its a global sum needed to compute a local (W_i)
 * property.
 *
 * [XXX write more on how we solve this.. _after_ merging pjt's patches that
 *      rewrite all of this once again.]
 */ 
3823

3824 3825
static unsigned long __read_mostly max_load_balance_interval = HZ/10;

3826
#define LBF_ALL_PINNED	0x01
3827
#define LBF_NEED_BREAK	0x02
3828
#define LBF_SOME_PINNED 0x04
3829 3830 3831 3832 3833

struct lb_env {
	struct sched_domain	*sd;

	struct rq		*src_rq;
3834
	int			src_cpu;
3835 3836 3837 3838

	int			dst_cpu;
	struct rq		*dst_rq;

3839 3840
	struct cpumask		*dst_grpmask;
	int			new_dst_cpu;
3841
	enum cpu_idle_type	idle;
3842
	long			imbalance;
3843 3844 3845
	/* The set of CPUs under consideration for load-balancing */
	struct cpumask		*cpus;

3846
	unsigned int		flags;
3847 3848 3849 3850

	unsigned int		loop;
	unsigned int		loop_break;
	unsigned int		loop_max;
3851 3852
};

3853
/*
3854
 * move_task - move a task from one runqueue to another runqueue.
3855 3856
 * Both runqueues must be locked.
 */
3857
static void move_task(struct task_struct *p, struct lb_env *env)
3858
{
3859 3860 3861 3862
	deactivate_task(env->src_rq, p, 0);
	set_task_cpu(p, env->dst_cpu);
	activate_task(env->dst_rq, p, 0);
	check_preempt_curr(env->dst_rq, p, 0);
3863 3864
}

3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896
/*
 * Is this task likely cache-hot:
 */
static int
task_hot(struct task_struct *p, u64 now, struct sched_domain *sd)
{
	s64 delta;

	if (p->sched_class != &fair_sched_class)
		return 0;

	if (unlikely(p->policy == SCHED_IDLE))
		return 0;

	/*
	 * Buddy candidates are cache hot:
	 */
	if (sched_feat(CACHE_HOT_BUDDY) && this_rq()->nr_running &&
			(&p->se == cfs_rq_of(&p->se)->next ||
			 &p->se == cfs_rq_of(&p->se)->last))
		return 1;

	if (sysctl_sched_migration_cost == -1)
		return 1;
	if (sysctl_sched_migration_cost == 0)
		return 0;

	delta = now - p->se.exec_start;

	return delta < (s64)sysctl_sched_migration_cost;
}

3897 3898 3899 3900
/*
 * can_migrate_task - may task p from runqueue rq be migrated to this_cpu?
 */
static
3901
int can_migrate_task(struct task_struct *p, struct lb_env *env)
3902 3903 3904 3905
{
	int tsk_cache_hot = 0;
	/*
	 * We do not migrate tasks that are:
3906
	 * 1) throttled_lb_pair, or
3907
	 * 2) cannot be migrated to this CPU due to cpus_allowed, or
3908 3909
	 * 3) running (obviously), or
	 * 4) are cache-hot on their current CPU.
3910
	 */
3911 3912 3913
	if (throttled_lb_pair(task_group(p), env->src_cpu, env->dst_cpu))
		return 0;

3914
	if (!cpumask_test_cpu(env->dst_cpu, tsk_cpus_allowed(p))) {
3915
		int cpu;
3916

3917
		schedstat_inc(p, se.statistics.nr_failed_migrations_affine);
3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929

		/*
		 * Remember if this task can be migrated to any other cpu in
		 * our sched_group. We may want to revisit it if we couldn't
		 * meet load balance goals by pulling other tasks on src_cpu.
		 *
		 * Also avoid computing new_dst_cpu if we have already computed
		 * one in current iteration.
		 */
		if (!env->dst_grpmask || (env->flags & LBF_SOME_PINNED))
			return 0;

3930 3931 3932 3933 3934 3935 3936
		/* Prevent to re-select dst_cpu via env's cpus */
		for_each_cpu_and(cpu, env->dst_grpmask, env->cpus) {
			if (cpumask_test_cpu(cpu, tsk_cpus_allowed(p))) {
				env->flags |= LBF_SOME_PINNED;
				env->new_dst_cpu = cpu;
				break;
			}
3937
		}
3938

3939 3940
		return 0;
	}
3941 3942

	/* Record that we found atleast one task that could run on dst_cpu */
3943
	env->flags &= ~LBF_ALL_PINNED;
3944

3945
	if (task_running(env->src_rq, p)) {
3946
		schedstat_inc(p, se.statistics.nr_failed_migrations_running);
3947 3948 3949 3950 3951 3952 3953 3954 3955
		return 0;
	}

	/*
	 * Aggressive migration if:
	 * 1) task is cache cold, or
	 * 2) too many balance attempts have failed.
	 */

3956
	tsk_cache_hot = task_hot(p, rq_clock_task(env->src_rq), env->sd);
3957
	if (!tsk_cache_hot ||
3958
		env->sd->nr_balance_failed > env->sd->cache_nice_tries) {
3959

3960
		if (tsk_cache_hot) {
3961
			schedstat_inc(env->sd, lb_hot_gained[env->idle]);
3962
			schedstat_inc(p, se.statistics.nr_forced_migrations);
3963
		}
3964

3965 3966 3967
		return 1;
	}

3968 3969
	schedstat_inc(p, se.statistics.nr_failed_migrations_hot);
	return 0;
3970 3971
}

3972 3973 3974 3975 3976 3977 3978
/*
 * move_one_task tries to move exactly one task from busiest to this_rq, as
 * part of active balancing operations within "domain".
 * Returns 1 if successful and 0 otherwise.
 *
 * Called with both runqueues locked.
 */
3979
static int move_one_task(struct lb_env *env)
3980 3981 3982
{
	struct task_struct *p, *n;

3983 3984 3985
	list_for_each_entry_safe(p, n, &env->src_rq->cfs_tasks, se.group_node) {
		if (!can_migrate_task(p, env))
			continue;
3986

3987 3988 3989 3990 3991 3992 3993 3994
		move_task(p, env);
		/*
		 * Right now, this is only the second place move_task()
		 * is called, so we can safely collect move_task()
		 * stats here rather than inside move_task().
		 */
		schedstat_inc(env->sd, lb_gained[env->idle]);
		return 1;
3995 3996 3997 3998
	}
	return 0;
}

3999 4000
static unsigned long task_h_load(struct task_struct *p);

4001 4002
static const unsigned int sched_nr_migrate_break = 32;

4003
/*
4004
 * move_tasks tries to move up to imbalance weighted load from busiest to
4005 4006 4007 4008 4009 4010
 * this_rq, as part of a balancing operation within domain "sd".
 * Returns 1 if successful and 0 otherwise.
 *
 * Called with both runqueues locked.
 */
static int move_tasks(struct lb_env *env)
4011
{
4012 4013
	struct list_head *tasks = &env->src_rq->cfs_tasks;
	struct task_struct *p;
4014 4015
	unsigned long load;
	int pulled = 0;
4016

4017
	if (env->imbalance <= 0)
4018
		return 0;
4019

4020 4021
	while (!list_empty(tasks)) {
		p = list_first_entry(tasks, struct task_struct, se.group_node);
4022

4023 4024
		env->loop++;
		/* We've more or less seen every task there is, call it quits */
4025
		if (env->loop > env->loop_max)
4026
			break;
4027 4028

		/* take a breather every nr_migrate tasks */
4029
		if (env->loop > env->loop_break) {
4030
			env->loop_break += sched_nr_migrate_break;
4031
			env->flags |= LBF_NEED_BREAK;
4032
			break;
4033
		}
4034

4035
		if (!can_migrate_task(p, env))
4036 4037 4038
			goto next;

		load = task_h_load(p);
4039

4040
		if (sched_feat(LB_MIN) && load < 16 && !env->sd->nr_balance_failed)
4041 4042
			goto next;

4043
		if ((load / 2) > env->imbalance)
4044
			goto next;
4045

4046
		move_task(p, env);
4047
		pulled++;
4048
		env->imbalance -= load;
4049 4050

#ifdef CONFIG_PREEMPT
4051 4052 4053 4054 4055
		/*
		 * NEWIDLE balancing is a source of latency, so preemptible
		 * kernels will stop after the first task is pulled to minimize
		 * the critical section.
		 */
4056
		if (env->idle == CPU_NEWLY_IDLE)
4057
			break;
4058 4059
#endif

4060 4061 4062 4063
		/*
		 * We only want to steal up to the prescribed amount of
		 * weighted load.
		 */
4064
		if (env->imbalance <= 0)
4065
			break;
4066 4067 4068

		continue;
next:
4069
		list_move_tail(&p->se.group_node, tasks);
4070
	}
4071

4072
	/*
4073 4074 4075
	 * Right now, this is one of only two places move_task() is called,
	 * so we can safely collect move_task() stats here rather than
	 * inside move_task().
4076
	 */
4077
	schedstat_add(env->sd, lb_gained[env->idle], pulled);
4078

4079
	return pulled;
4080 4081
}

4082
#ifdef CONFIG_FAIR_GROUP_SCHED
4083 4084 4085
/*
 * update tg->load_weight by folding this cpu's load_avg
 */
4086
static void __update_blocked_averages_cpu(struct task_group *tg, int cpu)
4087
{
4088 4089
	struct sched_entity *se = tg->se[cpu];
	struct cfs_rq *cfs_rq = tg->cfs_rq[cpu];
4090

4091 4092 4093
	/* throttled entities do not contribute to load */
	if (throttled_hierarchy(cfs_rq))
		return;
4094

4095
	update_cfs_rq_blocked_load(cfs_rq, 1);
4096

4097 4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108 4109 4110
	if (se) {
		update_entity_load_avg(se, 1);
		/*
		 * We pivot on our runnable average having decayed to zero for
		 * list removal.  This generally implies that all our children
		 * have also been removed (modulo rounding error or bandwidth
		 * control); however, such cases are rare and we can fix these
		 * at enqueue.
		 *
		 * TODO: fix up out-of-order children on enqueue.
		 */
		if (!se->avg.runnable_avg_sum && !cfs_rq->nr_running)
			list_del_leaf_cfs_rq(cfs_rq);
	} else {
4111
		struct rq *rq = rq_of(cfs_rq);
4112 4113
		update_rq_runnable_avg(rq, rq->nr_running);
	}
4114 4115
}

4116
static void update_blocked_averages(int cpu)
4117 4118
{
	struct rq *rq = cpu_rq(cpu);
4119 4120
	struct cfs_rq *cfs_rq;
	unsigned long flags;
4121

4122 4123
	raw_spin_lock_irqsave(&rq->lock, flags);
	update_rq_clock(rq);
4124 4125 4126 4127
	/*
	 * Iterates the task_group tree in a bottom up fashion, see
	 * list_add_leaf_cfs_rq() for details.
	 */
4128
	for_each_leaf_cfs_rq(rq, cfs_rq) {
4129 4130 4131 4132 4133 4134
		/*
		 * Note: We may want to consider periodically releasing
		 * rq->lock about these updates so that creating many task
		 * groups does not result in continually extending hold time.
		 */
		__update_blocked_averages_cpu(cfs_rq->tg, rq->cpu);
4135
	}
4136 4137

	raw_spin_unlock_irqrestore(&rq->lock, flags);
4138 4139
}

4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153 4154 4155 4156 4157 4158 4159 4160 4161 4162 4163 4164
/*
 * Compute the cpu's hierarchical load factor for each task group.
 * This needs to be done in a top-down fashion because the load of a child
 * group is a fraction of its parents load.
 */
static int tg_load_down(struct task_group *tg, void *data)
{
	unsigned long load;
	long cpu = (long)data;

	if (!tg->parent) {
		load = cpu_rq(cpu)->load.weight;
	} else {
		load = tg->parent->cfs_rq[cpu]->h_load;
		load *= tg->se[cpu]->load.weight;
		load /= tg->parent->cfs_rq[cpu]->load.weight + 1;
	}

	tg->cfs_rq[cpu]->h_load = load;

	return 0;
}

static void update_h_load(long cpu)
{
4165 4166 4167 4168 4169 4170 4171 4172
	struct rq *rq = cpu_rq(cpu);
	unsigned long now = jiffies;

	if (rq->h_load_throttle == now)
		return;

	rq->h_load_throttle = now;

4173
	rcu_read_lock();
4174
	walk_tg_tree(tg_load_down, tg_nop, (void *)cpu);
4175
	rcu_read_unlock();
4176 4177
}

4178
static unsigned long task_h_load(struct task_struct *p)
4179
{
4180 4181
	struct cfs_rq *cfs_rq = task_cfs_rq(p);
	unsigned long load;
4182

4183 4184
	load = p->se.load.weight;
	load = div_u64(load * cfs_rq->h_load, cfs_rq->load.weight + 1);
4185

4186
	return load;
4187 4188
}
#else
4189
static inline void update_blocked_averages(int cpu)
4190 4191 4192
{
}

4193
static inline void update_h_load(long cpu)
4194 4195 4196
{
}

4197
static unsigned long task_h_load(struct task_struct *p)
4198
{
4199
	return p->se.load.weight;
4200
}
4201
#endif
4202 4203 4204 4205 4206 4207 4208 4209 4210 4211 4212 4213 4214 4215 4216 4217 4218

/********** Helpers for find_busiest_group ************************/
/*
 * sd_lb_stats - Structure to store the statistics of a sched_domain
 * 		during load balancing.
 */
struct sd_lb_stats {
	struct sched_group *busiest; /* Busiest group in this sd */
	struct sched_group *this;  /* Local group in this sd */
	unsigned long total_load;  /* Total load of all groups in sd */
	unsigned long total_pwr;   /*	Total power of all groups in sd */
	unsigned long avg_load;	   /* Average load across all groups in sd */

	/** Statistics of this group */
	unsigned long this_load;
	unsigned long this_load_per_task;
	unsigned long this_nr_running;
4219
	unsigned long this_has_capacity;
4220
	unsigned int  this_idle_cpus;
4221 4222

	/* Statistics of the busiest group */
4223
	unsigned int  busiest_idle_cpus;
4224 4225 4226
	unsigned long max_load;
	unsigned long busiest_load_per_task;
	unsigned long busiest_nr_running;
4227
	unsigned long busiest_group_capacity;
4228
	unsigned long busiest_has_capacity;
4229
	unsigned int  busiest_group_weight;
4230 4231 4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242

	int group_imb; /* Is there imbalance in this sd */
};

/*
 * sg_lb_stats - stats of a sched_group required for load_balancing
 */
struct sg_lb_stats {
	unsigned long avg_load; /*Avg load across the CPUs of the group */
	unsigned long group_load; /* Total load over the CPUs of the group */
	unsigned long sum_nr_running; /* Nr tasks running in the group */
	unsigned long sum_weighted_load; /* Weighted load of group's tasks */
	unsigned long group_capacity;
4243 4244
	unsigned long idle_cpus;
	unsigned long group_weight;
4245
	int group_imb; /* Is there an imbalance in the group ? */
4246
	int group_has_capacity; /* Is there extra capacity in the group? */
4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274
};

/**
 * get_sd_load_idx - Obtain the load index for a given sched domain.
 * @sd: The sched_domain whose load_idx is to be obtained.
 * @idle: The Idle status of the CPU for whose sd load_icx is obtained.
 */
static inline int get_sd_load_idx(struct sched_domain *sd,
					enum cpu_idle_type idle)
{
	int load_idx;

	switch (idle) {
	case CPU_NOT_IDLE:
		load_idx = sd->busy_idx;
		break;

	case CPU_NEWLY_IDLE:
		load_idx = sd->newidle_idx;
		break;
	default:
		load_idx = sd->idle_idx;
		break;
	}

	return load_idx;
}

4275
static unsigned long default_scale_freq_power(struct sched_domain *sd, int cpu)
4276
{
4277
	return SCHED_POWER_SCALE;
4278 4279 4280 4281 4282 4283 4284
}

unsigned long __weak arch_scale_freq_power(struct sched_domain *sd, int cpu)
{
	return default_scale_freq_power(sd, cpu);
}

4285
static unsigned long default_scale_smt_power(struct sched_domain *sd, int cpu)
4286
{
4287
	unsigned long weight = sd->span_weight;
4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299
	unsigned long smt_gain = sd->smt_gain;

	smt_gain /= weight;

	return smt_gain;
}

unsigned long __weak arch_scale_smt_power(struct sched_domain *sd, int cpu)
{
	return default_scale_smt_power(sd, cpu);
}

4300
static unsigned long scale_rt_power(int cpu)
4301 4302
{
	struct rq *rq = cpu_rq(cpu);
4303
	u64 total, available, age_stamp, avg;
4304

4305 4306 4307 4308 4309 4310 4311
	/*
	 * Since we're reading these variables without serialization make sure
	 * we read them once before doing sanity checks on them.
	 */
	age_stamp = ACCESS_ONCE(rq->age_stamp);
	avg = ACCESS_ONCE(rq->rt_avg);

4312
	total = sched_avg_period() + (rq_clock(rq) - age_stamp);
4313

4314
	if (unlikely(total < avg)) {
4315 4316 4317
		/* Ensures that power won't end up being negative */
		available = 0;
	} else {
4318
		available = total - avg;
4319
	}
4320

4321 4322
	if (unlikely((s64)total < SCHED_POWER_SCALE))
		total = SCHED_POWER_SCALE;
4323

4324
	total >>= SCHED_POWER_SHIFT;
4325 4326 4327 4328 4329 4330

	return div_u64(available, total);
}

static void update_cpu_power(struct sched_domain *sd, int cpu)
{
4331
	unsigned long weight = sd->span_weight;
4332
	unsigned long power = SCHED_POWER_SCALE;
4333 4334 4335 4336 4337 4338 4339 4340
	struct sched_group *sdg = sd->groups;

	if ((sd->flags & SD_SHARE_CPUPOWER) && weight > 1) {
		if (sched_feat(ARCH_POWER))
			power *= arch_scale_smt_power(sd, cpu);
		else
			power *= default_scale_smt_power(sd, cpu);

4341
		power >>= SCHED_POWER_SHIFT;
4342 4343
	}

4344
	sdg->sgp->power_orig = power;
4345 4346 4347 4348 4349 4350

	if (sched_feat(ARCH_POWER))
		power *= arch_scale_freq_power(sd, cpu);
	else
		power *= default_scale_freq_power(sd, cpu);

4351
	power >>= SCHED_POWER_SHIFT;
4352

4353
	power *= scale_rt_power(cpu);
4354
	power >>= SCHED_POWER_SHIFT;
4355 4356 4357 4358

	if (!power)
		power = 1;

4359
	cpu_rq(cpu)->cpu_power = power;
4360
	sdg->sgp->power = power;
4361 4362
}

4363
void update_group_power(struct sched_domain *sd, int cpu)
4364 4365 4366 4367
{
	struct sched_domain *child = sd->child;
	struct sched_group *group, *sdg = sd->groups;
	unsigned long power;
4368 4369 4370 4371 4372
	unsigned long interval;

	interval = msecs_to_jiffies(sd->balance_interval);
	interval = clamp(interval, 1UL, max_load_balance_interval);
	sdg->sgp->next_update = jiffies + interval;
4373 4374 4375 4376 4377 4378 4379 4380

	if (!child) {
		update_cpu_power(sd, cpu);
		return;
	}

	power = 0;

Peter Zijlstra's avatar
Peter Zijlstra committed
4381 4382 4383 4384 4385 4386 4387 4388 4389 4390 4391 4392 4393 4394 4395 4396 4397 4398 4399 4400
	if (child->flags & SD_OVERLAP) {
		/*
		 * SD_OVERLAP domains cannot assume that child groups
		 * span the current group.
		 */

		for_each_cpu(cpu, sched_group_cpus(sdg))
			power += power_of(cpu);
	} else  {
		/*
		 * !SD_OVERLAP domains can assume that child groups
		 * span the current group.
		 */ 

		group = child->groups;
		do {
			power += group->sgp->power;
			group = group->next;
		} while (group != child->groups);
	}
4401

4402
	sdg->sgp->power_orig = sdg->sgp->power = power;
4403 4404
}

4405 4406 4407 4408 4409 4410 4411 4412 4413 4414 4415
/*
 * Try and fix up capacity for tiny siblings, this is needed when
 * things like SD_ASYM_PACKING need f_b_g to select another sibling
 * which on its own isn't powerful enough.
 *
 * See update_sd_pick_busiest() and check_asym_packing().
 */
static inline int
fix_small_capacity(struct sched_domain *sd, struct sched_group *group)
{
	/*
4416
	 * Only siblings can have significantly less than SCHED_POWER_SCALE
4417
	 */
4418
	if (!(sd->flags & SD_SHARE_CPUPOWER))
4419 4420 4421 4422 4423
		return 0;

	/*
	 * If ~90% of the cpu_power is still there, we're good.
	 */
4424
	if (group->sgp->power * 32 > group->sgp->power_orig * 29)
4425 4426 4427 4428 4429
		return 1;

	return 0;
}

4430 4431
/**
 * update_sg_lb_stats - Update sched_group's statistics for load balancing.
4432
 * @env: The load balancing environment.
4433 4434 4435 4436 4437 4438
 * @group: sched_group whose statistics are to be updated.
 * @load_idx: Load index of sched_domain of this_cpu for load calc.
 * @local_group: Does group contain this_cpu.
 * @balance: Should we balance.
 * @sgs: variable to hold the statistics for this group.
 */
4439 4440
static inline void update_sg_lb_stats(struct lb_env *env,
			struct sched_group *group, int load_idx,
4441
			int local_group, int *balance, struct sg_lb_stats *sgs)
4442
{
4443 4444
	unsigned long nr_running, max_nr_running, min_nr_running;
	unsigned long load, max_cpu_load, min_cpu_load;
4445
	unsigned int balance_cpu = -1, first_idle_cpu = 0;
4446
	unsigned long avg_load_per_task = 0;
4447
	int i;
4448

4449
	if (local_group)
Peter Zijlstra's avatar
Peter Zijlstra committed
4450
		balance_cpu = group_balance_cpu(group);
4451 4452 4453 4454

	/* Tally up the load of all CPUs in the group */
	max_cpu_load = 0;
	min_cpu_load = ~0UL;
4455
	max_nr_running = 0;
4456
	min_nr_running = ~0UL;
4457

4458
	for_each_cpu_and(i, sched_group_cpus(group), env->cpus) {
4459 4460
		struct rq *rq = cpu_rq(i);

4461 4462
		nr_running = rq->nr_running;

4463 4464
		/* Bias balancing toward cpus of our domain */
		if (local_group) {
Peter Zijlstra's avatar
Peter Zijlstra committed
4465 4466
			if (idle_cpu(i) && !first_idle_cpu &&
					cpumask_test_cpu(i, sched_group_mask(group))) {
4467
				first_idle_cpu = 1;
4468 4469
				balance_cpu = i;
			}
4470 4471

			load = target_load(i, load_idx);
4472 4473
		} else {
			load = source_load(i, load_idx);
4474
			if (load > max_cpu_load)
4475 4476 4477
				max_cpu_load = load;
			if (min_cpu_load > load)
				min_cpu_load = load;
4478 4479 4480 4481 4482

			if (nr_running > max_nr_running)
				max_nr_running = nr_running;
			if (min_nr_running > nr_running)
				min_nr_running = nr_running;
4483 4484 4485
		}

		sgs->group_load += load;
4486
		sgs->sum_nr_running += nr_running;
4487
		sgs->sum_weighted_load += weighted_cpuload(i);
4488 4489
		if (idle_cpu(i))
			sgs->idle_cpus++;
4490 4491 4492 4493 4494 4495 4496 4497
	}

	/*
	 * First idle cpu or the first cpu(busiest) in this sched group
	 * is eligible for doing load balancing at this and above
	 * domains. In the newly idle case, we will allow all the cpu's
	 * to do the newly idle load balance.
	 */
4498
	if (local_group) {
4499
		if (env->idle != CPU_NEWLY_IDLE) {
4500
			if (balance_cpu != env->dst_cpu) {
4501 4502 4503
				*balance = 0;
				return;
			}
4504
			update_group_power(env->sd, env->dst_cpu);
4505
		} else if (time_after_eq(jiffies, group->sgp->next_update))
4506
			update_group_power(env->sd, env->dst_cpu);
4507 4508 4509
	}

	/* Adjust by relative CPU power of the group */
4510
	sgs->avg_load = (sgs->group_load*SCHED_POWER_SCALE) / group->sgp->power;
4511 4512 4513

	/*
	 * Consider the group unbalanced when the imbalance is larger
4514
	 * than the average weight of a task.
4515 4516 4517 4518 4519 4520
	 *
	 * APZ: with cgroup the avg task weight can vary wildly and
	 *      might not be a suitable number - should we keep a
	 *      normalized nr_running number somewhere that negates
	 *      the hierarchy?
	 */
4521 4522
	if (sgs->sum_nr_running)
		avg_load_per_task = sgs->sum_weighted_load / sgs->sum_nr_running;
4523

4524 4525
	if ((max_cpu_load - min_cpu_load) >= avg_load_per_task &&
	    (max_nr_running - min_nr_running) > 1)
4526 4527
		sgs->group_imb = 1;

4528
	sgs->group_capacity = DIV_ROUND_CLOSEST(group->sgp->power,
4529
						SCHED_POWER_SCALE);
4530
	if (!sgs->group_capacity)
4531
		sgs->group_capacity = fix_small_capacity(env->sd, group);
4532
	sgs->group_weight = group->group_weight;
4533 4534 4535

	if (sgs->group_capacity > sgs->sum_nr_running)
		sgs->group_has_capacity = 1;
4536 4537
}

4538 4539
/**
 * update_sd_pick_busiest - return 1 on busiest group
4540
 * @env: The load balancing environment.
4541 4542
 * @sds: sched_domain statistics
 * @sg: sched_group candidate to be checked for being the busiest
4543
 * @sgs: sched_group statistics
4544 4545 4546 4547
 *
 * Determine if @sg is a busier group than the previously selected
 * busiest group.
 */
4548
static bool update_sd_pick_busiest(struct lb_env *env,
4549 4550
				   struct sd_lb_stats *sds,
				   struct sched_group *sg,
4551
				   struct sg_lb_stats *sgs)
4552 4553 4554 4555 4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566
{
	if (sgs->avg_load <= sds->max_load)
		return false;

	if (sgs->sum_nr_running > sgs->group_capacity)
		return true;

	if (sgs->group_imb)
		return true;

	/*
	 * ASYM_PACKING needs to move all the work to the lowest
	 * numbered CPUs in the group, therefore mark all groups
	 * higher than ourself as busy.
	 */
4567 4568
	if ((env->sd->flags & SD_ASYM_PACKING) && sgs->sum_nr_running &&
	    env->dst_cpu < group_first_cpu(sg)) {
4569 4570 4571 4572 4573 4574 4575 4576 4577 4578
		if (!sds->busiest)
			return true;

		if (group_first_cpu(sds->busiest) > group_first_cpu(sg))
			return true;
	}

	return false;
}

4579
/**
4580
 * update_sd_lb_stats - Update sched_domain's statistics for load balancing.
4581
 * @env: The load balancing environment.
4582 4583 4584
 * @balance: Should we balance.
 * @sds: variable to hold the statistics for this sched_domain.
 */
4585
static inline void update_sd_lb_stats(struct lb_env *env,
4586
					int *balance, struct sd_lb_stats *sds)
4587
{
4588 4589
	struct sched_domain *child = env->sd->child;
	struct sched_group *sg = env->sd->groups;
4590 4591 4592 4593 4594 4595
	struct sg_lb_stats sgs;
	int load_idx, prefer_sibling = 0;

	if (child && child->flags & SD_PREFER_SIBLING)
		prefer_sibling = 1;

4596
	load_idx = get_sd_load_idx(env->sd, env->idle);
4597 4598 4599 4600

	do {
		int local_group;

4601
		local_group = cpumask_test_cpu(env->dst_cpu, sched_group_cpus(sg));
4602
		memset(&sgs, 0, sizeof(sgs));
4603
		update_sg_lb_stats(env, sg, load_idx, local_group, balance, &sgs);
4604

4605
		if (local_group && !(*balance))
4606 4607 4608
			return;

		sds->total_load += sgs.group_load;
4609
		sds->total_pwr += sg->sgp->power;
4610 4611 4612

		/*
		 * In case the child domain prefers tasks go to siblings
4613
		 * first, lower the sg capacity to one so that we'll try
4614 4615 4616 4617 4618 4619
		 * and move all the excess tasks away. We lower the capacity
		 * of a group only if the local group has the capacity to fit
		 * these excess tasks, i.e. nr_running < group_capacity. The
		 * extra check prevents the case where you always pull from the
		 * heaviest group when it is already under-utilized (possible
		 * with a large weight task outweighs the tasks on the system).
4620
		 */
4621
		if (prefer_sibling && !local_group && sds->this_has_capacity)
4622 4623 4624 4625
			sgs.group_capacity = min(sgs.group_capacity, 1UL);

		if (local_group) {
			sds->this_load = sgs.avg_load;
4626
			sds->this = sg;
4627 4628
			sds->this_nr_running = sgs.sum_nr_running;
			sds->this_load_per_task = sgs.sum_weighted_load;
4629
			sds->this_has_capacity = sgs.group_has_capacity;
4630
			sds->this_idle_cpus = sgs.idle_cpus;
4631
		} else if (update_sd_pick_busiest(env, sds, sg, &sgs)) {
4632
			sds->max_load = sgs.avg_load;
4633
			sds->busiest = sg;
4634
			sds->busiest_nr_running = sgs.sum_nr_running;
4635
			sds->busiest_idle_cpus = sgs.idle_cpus;
4636
			sds->busiest_group_capacity = sgs.group_capacity;
4637
			sds->busiest_load_per_task = sgs.sum_weighted_load;
4638
			sds->busiest_has_capacity = sgs.group_has_capacity;
4639
			sds->busiest_group_weight = sgs.group_weight;
4640 4641 4642
			sds->group_imb = sgs.group_imb;
		}

4643
		sg = sg->next;
4644
	} while (sg != env->sd->groups);
4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663
}

/**
 * check_asym_packing - Check to see if the group is packed into the
 *			sched doman.
 *
 * This is primarily intended to used at the sibling level.  Some
 * cores like POWER7 prefer to use lower numbered SMT threads.  In the
 * case of POWER7, it can move to lower SMT modes only when higher
 * threads are idle.  When in lower SMT modes, the threads will
 * perform better since they share less core resources.  Hence when we
 * have idle threads, we want them to be the higher ones.
 *
 * This packing function is run on idle threads.  It checks to see if
 * the busiest CPU in this domain (core in the P7 case) has a higher
 * CPU number than the packing function is being run on.  Here we are
 * assuming lower CPU number will be equivalent to lower a SMT thread
 * number.
 *
4664 4665 4666
 * Returns 1 when packing is required and a task should be moved to
 * this CPU.  The amount of the imbalance is returned in *imbalance.
 *
4667
 * @env: The load balancing environment.
4668 4669
 * @sds: Statistics of the sched_domain which is to be packed
 */
4670
static int check_asym_packing(struct lb_env *env, struct sd_lb_stats *sds)
4671 4672 4673
{
	int busiest_cpu;

4674
	if (!(env->sd->flags & SD_ASYM_PACKING))
4675 4676 4677 4678 4679 4680
		return 0;

	if (!sds->busiest)
		return 0;

	busiest_cpu = group_first_cpu(sds->busiest);
4681
	if (env->dst_cpu > busiest_cpu)
4682 4683
		return 0;

4684 4685 4686
	env->imbalance = DIV_ROUND_CLOSEST(
		sds->max_load * sds->busiest->sgp->power, SCHED_POWER_SCALE);

4687
	return 1;
4688 4689 4690 4691 4692 4693
}

/**
 * fix_small_imbalance - Calculate the minor imbalance that exists
 *			amongst the groups of a sched_domain, during
 *			load balancing.
4694
 * @env: The load balancing environment.
4695 4696
 * @sds: Statistics of the sched_domain whose imbalance is to be calculated.
 */
4697 4698
static inline
void fix_small_imbalance(struct lb_env *env, struct sd_lb_stats *sds)
4699 4700 4701
{
	unsigned long tmp, pwr_now = 0, pwr_move = 0;
	unsigned int imbn = 2;
4702
	unsigned long scaled_busy_load_per_task;
4703 4704 4705 4706 4707 4708

	if (sds->this_nr_running) {
		sds->this_load_per_task /= sds->this_nr_running;
		if (sds->busiest_load_per_task >
				sds->this_load_per_task)
			imbn = 1;
4709
	} else {
4710
		sds->this_load_per_task =
4711 4712
			cpu_avg_load_per_task(env->dst_cpu);
	}
4713

4714
	scaled_busy_load_per_task = sds->busiest_load_per_task
4715
					 * SCHED_POWER_SCALE;
4716
	scaled_busy_load_per_task /= sds->busiest->sgp->power;
4717 4718 4719

	if (sds->max_load - sds->this_load + scaled_busy_load_per_task >=
			(scaled_busy_load_per_task * imbn)) {
4720
		env->imbalance = sds->busiest_load_per_task;
4721 4722 4723 4724 4725 4726 4727 4728 4729
		return;
	}

	/*
	 * OK, we don't have enough imbalance to justify moving tasks,
	 * however we may be able to increase total CPU power used by
	 * moving them.
	 */

4730
	pwr_now += sds->busiest->sgp->power *
4731
			min(sds->busiest_load_per_task, sds->max_load);
4732
	pwr_now += sds->this->sgp->power *
4733
			min(sds->this_load_per_task, sds->this_load);
4734
	pwr_now /= SCHED_POWER_SCALE;
4735 4736

	/* Amount of load we'd subtract */
4737
	tmp = (sds->busiest_load_per_task * SCHED_POWER_SCALE) /
4738
		sds->busiest->sgp->power;
4739
	if (sds->max_load > tmp)
4740
		pwr_move += sds->busiest->sgp->power *
4741 4742 4743
			min(sds->busiest_load_per_task, sds->max_load - tmp);

	/* Amount of load we'd add */
4744
	if (sds->max_load * sds->busiest->sgp->power <
4745
		sds->busiest_load_per_task * SCHED_POWER_SCALE)
4746 4747
		tmp = (sds->max_load * sds->busiest->sgp->power) /
			sds->this->sgp->power;
4748
	else
4749
		tmp = (sds->busiest_load_per_task * SCHED_POWER_SCALE) /
4750 4751
			sds->this->sgp->power;
	pwr_move += sds->this->sgp->power *
4752
			min(sds->this_load_per_task, sds->this_load + tmp);
4753
	pwr_move /= SCHED_POWER_SCALE;
4754 4755 4756

	/* Move if we gain throughput */
	if (pwr_move > pwr_now)
4757
		env->imbalance = sds->busiest_load_per_task;
4758 4759 4760 4761 4762
}

/**
 * calculate_imbalance - Calculate the amount of imbalance present within the
 *			 groups of a given sched_domain during load balance.
4763
 * @env: load balance environment
4764 4765
 * @sds: statistics of the sched_domain whose imbalance is to be calculated.
 */
4766
static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *sds)
4767
{
4768 4769 4770 4771 4772 4773 4774 4775
	unsigned long max_pull, load_above_capacity = ~0UL;

	sds->busiest_load_per_task /= sds->busiest_nr_running;
	if (sds->group_imb) {
		sds->busiest_load_per_task =
			min(sds->busiest_load_per_task, sds->avg_load);
	}

4776 4777 4778 4779 4780 4781
	/*
	 * In the presence of smp nice balancing, certain scenarios can have
	 * max load less than avg load(as we skip the groups at or below
	 * its cpu_power, while calculating max_load..)
	 */
	if (sds->max_load < sds->avg_load) {
4782 4783
		env->imbalance = 0;
		return fix_small_imbalance(env, sds);
4784 4785
	}

4786 4787 4788 4789 4790 4791 4792
	if (!sds->group_imb) {
		/*
		 * Don't want to pull so many tasks that a group would go idle.
		 */
		load_above_capacity = (sds->busiest_nr_running -
						sds->busiest_group_capacity);

4793
		load_above_capacity *= (SCHED_LOAD_SCALE * SCHED_POWER_SCALE);
4794

4795
		load_above_capacity /= sds->busiest->sgp->power;
4796 4797 4798 4799 4800 4801 4802 4803 4804 4805 4806 4807 4808
	}

	/*
	 * We're trying to get all the cpus to the average_load, so we don't
	 * want to push ourselves above the average load, nor do we wish to
	 * reduce the max loaded cpu below the average load. At the same time,
	 * we also don't want to reduce the group load below the group capacity
	 * (so that we can implement power-savings policies etc). Thus we look
	 * for the minimum possible imbalance.
	 * Be careful of negative numbers as they'll appear as very large values
	 * with unsigned longs.
	 */
	max_pull = min(sds->max_load - sds->avg_load, load_above_capacity);
4809 4810

	/* How much load to actually move to equalise the imbalance */
4811
	env->imbalance = min(max_pull * sds->busiest->sgp->power,
4812
		(sds->avg_load - sds->this_load) * sds->this->sgp->power)
4813
			/ SCHED_POWER_SCALE;
4814 4815 4816

	/*
	 * if *imbalance is less than the average load per runnable task
Lucas De Marchi's avatar
Lucas De Marchi committed
4817
	 * there is no guarantee that any tasks will be moved so we'll have
4818 4819 4820
	 * a think about bumping its value to force at least one task to be
	 * moved
	 */
4821 4822
	if (env->imbalance < sds->busiest_load_per_task)
		return fix_small_imbalance(env, sds);
4823 4824

}
4825

4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836 4837
/******* find_busiest_group() helpers end here *********************/

/**
 * find_busiest_group - Returns the busiest group within the sched_domain
 * if there is an imbalance. If there isn't an imbalance, and
 * the user has opted for power-savings, it returns a group whose
 * CPUs can be put to idle by rebalancing those tasks elsewhere, if
 * such a group exists.
 *
 * Also calculates the amount of weighted load which should be moved
 * to restore balance.
 *
4838
 * @env: The load balancing environment.
4839 4840 4841 4842 4843 4844 4845 4846 4847
 * @balance: Pointer to a variable indicating if this_cpu
 *	is the appropriate cpu to perform load balancing at this_level.
 *
 * Returns:	- the busiest group if imbalance exists.
 *		- If no imbalance and user has opted for power-savings balance,
 *		   return the least loaded group whose CPUs can be
 *		   put to idle by rebalancing its tasks onto our group.
 */
static struct sched_group *
4848
find_busiest_group(struct lb_env *env, int *balance)
4849 4850 4851 4852 4853 4854 4855 4856 4857
{
	struct sd_lb_stats sds;

	memset(&sds, 0, sizeof(sds));

	/*
	 * Compute the various statistics relavent for load balancing at
	 * this level.
	 */
4858
	update_sd_lb_stats(env, balance, &sds);
4859

4860 4861 4862
	/*
	 * this_cpu is not the appropriate cpu to perform load balancing at
	 * this level.
4863
	 */
4864
	if (!(*balance))
4865 4866
		goto ret;

4867 4868
	if ((env->idle == CPU_IDLE || env->idle == CPU_NEWLY_IDLE) &&
	    check_asym_packing(env, &sds))
4869 4870
		return sds.busiest;

4871
	/* There is no busy sibling group to pull tasks from */
4872 4873 4874
	if (!sds.busiest || sds.busiest_nr_running == 0)
		goto out_balanced;

4875
	sds.avg_load = (SCHED_POWER_SCALE * sds.total_load) / sds.total_pwr;
4876

4877 4878 4879 4880 4881 4882 4883 4884
	/*
	 * If the busiest group is imbalanced the below checks don't
	 * work because they assumes all things are equal, which typically
	 * isn't true due to cpus_allowed constraints and the like.
	 */
	if (sds.group_imb)
		goto force_balance;

4885
	/* SD_BALANCE_NEWIDLE trumps SMP nice when underutilized */
4886
	if (env->idle == CPU_NEWLY_IDLE && sds.this_has_capacity &&
4887 4888 4889
			!sds.busiest_has_capacity)
		goto force_balance;

4890 4891 4892 4893
	/*
	 * If the local group is more busy than the selected busiest group
	 * don't try and pull any tasks.
	 */
4894 4895 4896
	if (sds.this_load >= sds.max_load)
		goto out_balanced;

4897 4898 4899 4900
	/*
	 * Don't pull any tasks if this group is already above the domain
	 * average load.
	 */
4901 4902 4903
	if (sds.this_load >= sds.avg_load)
		goto out_balanced;

4904
	if (env->idle == CPU_IDLE) {
4905 4906 4907 4908 4909 4910
		/*
		 * This cpu is idle. If the busiest group load doesn't
		 * have more tasks than the number of available cpu's and
		 * there is no imbalance between this and busiest group
		 * wrt to idle cpu's, it is balanced.
		 */
4911
		if ((sds.this_idle_cpus <= sds.busiest_idle_cpus + 1) &&
4912 4913
		    sds.busiest_nr_running <= sds.busiest_group_weight)
			goto out_balanced;
4914 4915 4916 4917 4918
	} else {
		/*
		 * In the CPU_NEWLY_IDLE, CPU_NOT_IDLE cases, use
		 * imbalance_pct to be conservative.
		 */
4919
		if (100 * sds.max_load <= env->sd->imbalance_pct * sds.this_load)
4920
			goto out_balanced;
4921
	}
4922

4923
force_balance:
4924
	/* Looks like there is an imbalance. Compute it */
4925
	calculate_imbalance(env, &sds);
4926 4927 4928 4929
	return sds.busiest;

out_balanced:
ret:
4930
	env->imbalance = 0;
4931 4932 4933 4934 4935 4936
	return NULL;
}

/*
 * find_busiest_queue - find the busiest runqueue among the cpus in group.
 */
4937
static struct rq *find_busiest_queue(struct lb_env *env,
4938
				     struct sched_group *group)
4939 4940 4941 4942 4943 4944 4945
{
	struct rq *busiest = NULL, *rq;
	unsigned long max_load = 0;
	int i;

	for_each_cpu(i, sched_group_cpus(group)) {
		unsigned long power = power_of(i);
4946 4947
		unsigned long capacity = DIV_ROUND_CLOSEST(power,
							   SCHED_POWER_SCALE);
4948 4949
		unsigned long wl;

4950
		if (!capacity)
4951
			capacity = fix_small_capacity(env->sd, group);
4952

4953
		if (!cpumask_test_cpu(i, env->cpus))
4954 4955 4956
			continue;

		rq = cpu_rq(i);
4957
		wl = weighted_cpuload(i);
4958

4959 4960 4961 4962
		/*
		 * When comparing with imbalance, use weighted_cpuload()
		 * which is not scaled with the cpu power.
		 */
4963
		if (capacity && rq->nr_running == 1 && wl > env->imbalance)
4964 4965
			continue;

4966 4967 4968 4969 4970 4971
		/*
		 * For the load comparisons with the other cpu's, consider
		 * the weighted_cpuload() scaled with the cpu power, so that
		 * the load can be moved away from the cpu that is potentially
		 * running at a lower capacity.
		 */
4972
		wl = (wl * SCHED_POWER_SCALE) / power;
4973

4974 4975 4976 4977 4978 4979 4980 4981 4982 4983 4984 4985 4986 4987 4988 4989
		if (wl > max_load) {
			max_load = wl;
			busiest = rq;
		}
	}

	return busiest;
}

/*
 * Max backoff if we encounter pinned tasks. Pretty arbitrary value, but
 * so long as it is large enough.
 */
#define MAX_PINNED_INTERVAL	512

/* Working cpumask for load_balance and load_balance_newidle. */
4990
DEFINE_PER_CPU(cpumask_var_t, load_balance_mask);
4991

4992
static int need_active_balance(struct lb_env *env)
4993
{
4994 4995 4996
	struct sched_domain *sd = env->sd;

	if (env->idle == CPU_NEWLY_IDLE) {
4997 4998 4999 5000 5001 5002

		/*
		 * ASYM_PACKING needs to force migrate tasks from busy but
		 * higher numbered CPUs in order to pack all tasks in the
		 * lowest numbered CPUs.
		 */
5003
		if ((sd->flags & SD_ASYM_PACKING) && env->src_cpu > env->dst_cpu)
5004
			return 1;
5005 5006 5007 5008 5009
	}

	return unlikely(sd->nr_balance_failed > sd->cache_nice_tries+2);
}

5010 5011
static int active_load_balance_cpu_stop(void *data);

5012 5013 5014 5015 5016 5017 5018 5019
/*
 * Check this_cpu to ensure it is balanced within domain. Attempt to move
 * tasks if there is an imbalance.
 */
static int load_balance(int this_cpu, struct rq *this_rq,
			struct sched_domain *sd, enum cpu_idle_type idle,
			int *balance)
{
5020
	int ld_moved, cur_ld_moved, active_balance = 0;
5021 5022 5023
	struct sched_group *group;
	struct rq *busiest;
	unsigned long flags;
5024
	struct cpumask *cpus = __get_cpu_var(load_balance_mask);
5025

5026 5027
	struct lb_env env = {
		.sd		= sd,
5028 5029
		.dst_cpu	= this_cpu,
		.dst_rq		= this_rq,
5030
		.dst_grpmask    = sched_group_cpus(sd->groups),
5031
		.idle		= idle,
5032
		.loop_break	= sched_nr_migrate_break,
5033
		.cpus		= cpus,
5034 5035
	};

5036 5037 5038 5039
	/*
	 * For NEWLY_IDLE load_balancing, we don't need to consider
	 * other cpus in our group
	 */
5040
	if (idle == CPU_NEWLY_IDLE)
5041 5042
		env.dst_grpmask = NULL;

5043 5044 5045 5046 5047
	cpumask_copy(cpus, cpu_active_mask);

	schedstat_inc(sd, lb_count[idle]);

redo:
5048
	group = find_busiest_group(&env, balance);
5049 5050 5051 5052 5053 5054 5055 5056 5057

	if (*balance == 0)
		goto out_balanced;

	if (!group) {
		schedstat_inc(sd, lb_nobusyg[idle]);
		goto out_balanced;
	}

5058
	busiest = find_busiest_queue(&env, group);
5059 5060 5061 5062 5063
	if (!busiest) {
		schedstat_inc(sd, lb_nobusyq[idle]);
		goto out_balanced;
	}

5064
	BUG_ON(busiest == env.dst_rq);
5065

5066
	schedstat_add(sd, lb_imbalance[idle], env.imbalance);
5067 5068 5069 5070 5071 5072 5073 5074 5075

	ld_moved = 0;
	if (busiest->nr_running > 1) {
		/*
		 * Attempt to move tasks. If find_busiest_group has found
		 * an imbalance but busiest->nr_running <= 1, the group is
		 * still unbalanced. ld_moved simply stays zero, so it is
		 * correctly treated as an imbalance.
		 */
5076
		env.flags |= LBF_ALL_PINNED;
5077 5078 5079
		env.src_cpu   = busiest->cpu;
		env.src_rq    = busiest;
		env.loop_max  = min(sysctl_sched_nr_migrate, busiest->nr_running);
5080

5081
		update_h_load(env.src_cpu);
5082
more_balance:
5083
		local_irq_save(flags);
5084
		double_rq_lock(env.dst_rq, busiest);
5085 5086 5087 5088 5089 5090 5091

		/*
		 * cur_ld_moved - load moved in current iteration
		 * ld_moved     - cumulative load moved across iterations
		 */
		cur_ld_moved = move_tasks(&env);
		ld_moved += cur_ld_moved;
5092
		double_rq_unlock(env.dst_rq, busiest);
5093 5094 5095 5096 5097
		local_irq_restore(flags);

		/*
		 * some other cpu did the load balance for us.
		 */
5098 5099 5100
		if (cur_ld_moved && env.dst_cpu != smp_processor_id())
			resched_cpu(env.dst_cpu);

5101 5102 5103 5104 5105
		if (env.flags & LBF_NEED_BREAK) {
			env.flags &= ~LBF_NEED_BREAK;
			goto more_balance;
		}

5106 5107 5108 5109 5110 5111 5112 5113 5114 5115 5116 5117 5118 5119 5120 5121 5122 5123 5124
		/*
		 * Revisit (affine) tasks on src_cpu that couldn't be moved to
		 * us and move them to an alternate dst_cpu in our sched_group
		 * where they can run. The upper limit on how many times we
		 * iterate on same src_cpu is dependent on number of cpus in our
		 * sched_group.
		 *
		 * This changes load balance semantics a bit on who can move
		 * load to a given_cpu. In addition to the given_cpu itself
		 * (or a ilb_cpu acting on its behalf where given_cpu is
		 * nohz-idle), we now have balance_cpu in a position to move
		 * load to given_cpu. In rare situations, this may cause
		 * conflicts (balance_cpu and given_cpu/ilb_cpu deciding
		 * _independently_ and at _same_ time to move some load to
		 * given_cpu) causing exceess load to be moved to given_cpu.
		 * This however should not happen so much in practice and
		 * moreover subsequent load balance cycles should correct the
		 * excess load moved.
		 */
5125
		if ((env.flags & LBF_SOME_PINNED) && env.imbalance > 0) {
5126

5127
			env.dst_rq	 = cpu_rq(env.new_dst_cpu);
5128 5129 5130 5131
			env.dst_cpu	 = env.new_dst_cpu;
			env.flags	&= ~LBF_SOME_PINNED;
			env.loop	 = 0;
			env.loop_break	 = sched_nr_migrate_break;
5132 5133 5134 5135

			/* Prevent to re-select dst_cpu via env's cpus */
			cpumask_clear_cpu(env.dst_cpu, env.cpus);

5136 5137 5138 5139 5140 5141
			/*
			 * Go back to "more_balance" rather than "redo" since we
			 * need to continue with same src_cpu.
			 */
			goto more_balance;
		}
5142 5143

		/* All tasks on this runqueue were pinned by CPU affinity */
5144
		if (unlikely(env.flags & LBF_ALL_PINNED)) {
5145
			cpumask_clear_cpu(cpu_of(busiest), cpus);
5146 5147 5148
			if (!cpumask_empty(cpus)) {
				env.loop = 0;
				env.loop_break = sched_nr_migrate_break;
5149
				goto redo;
5150
			}
5151 5152 5153 5154 5155 5156
			goto out_balanced;
		}
	}

	if (!ld_moved) {
		schedstat_inc(sd, lb_failed[idle]);
5157 5158 5159 5160 5161 5162 5163 5164
		/*
		 * Increment the failure counter only on periodic balance.
		 * We do not want newidle balance, which can be very
		 * frequent, pollute the failure counter causing
		 * excessive cache_hot migrations and active balances.
		 */
		if (idle != CPU_NEWLY_IDLE)
			sd->nr_balance_failed++;
5165

5166
		if (need_active_balance(&env)) {
5167 5168
			raw_spin_lock_irqsave(&busiest->lock, flags);

5169 5170 5171
			/* don't kick the active_load_balance_cpu_stop,
			 * if the curr task on busiest cpu can't be
			 * moved to this_cpu
5172 5173
			 */
			if (!cpumask_test_cpu(this_cpu,
5174
					tsk_cpus_allowed(busiest->curr))) {
5175 5176
				raw_spin_unlock_irqrestore(&busiest->lock,
							    flags);
5177
				env.flags |= LBF_ALL_PINNED;
5178 5179 5180
				goto out_one_pinned;
			}

5181 5182 5183 5184 5185
			/*
			 * ->active_balance synchronizes accesses to
			 * ->active_balance_work.  Once set, it's cleared
			 * only after active load balance is finished.
			 */
5186 5187 5188 5189 5190 5191
			if (!busiest->active_balance) {
				busiest->active_balance = 1;
				busiest->push_cpu = this_cpu;
				active_balance = 1;
			}
			raw_spin_unlock_irqrestore(&busiest->lock, flags);
5192

5193
			if (active_balance) {
5194 5195 5196
				stop_one_cpu_nowait(cpu_of(busiest),
					active_load_balance_cpu_stop, busiest,
					&busiest->active_balance_work);
5197
			}
5198 5199 5200 5201 5202 5203 5204 5205 5206 5207 5208 5209 5210 5211 5212 5213 5214 5215 5216 5217 5218 5219 5220 5221 5222 5223 5224 5225 5226 5227 5228 5229 5230

			/*
			 * We've kicked active balancing, reset the failure
			 * counter.
			 */
			sd->nr_balance_failed = sd->cache_nice_tries+1;
		}
	} else
		sd->nr_balance_failed = 0;

	if (likely(!active_balance)) {
		/* We were unbalanced, so reset the balancing interval */
		sd->balance_interval = sd->min_interval;
	} else {
		/*
		 * If we've begun active balancing, start to back off. This
		 * case may not be covered by the all_pinned logic if there
		 * is only 1 task on the busy runqueue (because we don't call
		 * move_tasks).
		 */
		if (sd->balance_interval < sd->max_interval)
			sd->balance_interval *= 2;
	}

	goto out;

out_balanced:
	schedstat_inc(sd, lb_balanced[idle]);

	sd->nr_balance_failed = 0;

out_one_pinned:
	/* tune up the balancing interval */
5231
	if (((env.flags & LBF_ALL_PINNED) &&
5232
			sd->balance_interval < MAX_PINNED_INTERVAL) ||
5233 5234 5235
			(sd->balance_interval < sd->max_interval))
		sd->balance_interval *= 2;

5236
	ld_moved = 0;
5237 5238 5239 5240 5241 5242 5243 5244
out:
	return ld_moved;
}

/*
 * idle_balance is called by schedule() if this_cpu is about to become
 * idle. Attempts to pull tasks from other CPUs.
 */
5245
void idle_balance(int this_cpu, struct rq *this_rq)
5246 5247 5248 5249 5250
{
	struct sched_domain *sd;
	int pulled_task = 0;
	unsigned long next_balance = jiffies + HZ;

5251
	this_rq->idle_stamp = rq_clock(this_rq);
5252 5253 5254 5255

	if (this_rq->avg_idle < sysctl_sched_migration_cost)
		return;

5256 5257 5258 5259 5260
	/*
	 * Drop the rq->lock, but keep IRQ/preempt disabled.
	 */
	raw_spin_unlock(&this_rq->lock);

5261
	update_blocked_averages(this_cpu);
5262
	rcu_read_lock();
5263 5264
	for_each_domain(this_cpu, sd) {
		unsigned long interval;
5265
		int balance = 1;
5266 5267 5268 5269

		if (!(sd->flags & SD_LOAD_BALANCE))
			continue;

5270
		if (sd->flags & SD_BALANCE_NEWIDLE) {
5271
			/* If we've pulled tasks over stop searching: */
5272 5273 5274
			pulled_task = load_balance(this_cpu, this_rq,
						   sd, CPU_NEWLY_IDLE, &balance);
		}
5275 5276 5277 5278

		interval = msecs_to_jiffies(sd->balance_interval);
		if (time_after(next_balance, sd->last_balance + interval))
			next_balance = sd->last_balance + interval;
Nikhil Rao's avatar
Nikhil Rao committed
5279 5280
		if (pulled_task) {
			this_rq->idle_stamp = 0;
5281
			break;
Nikhil Rao's avatar
Nikhil Rao committed
5282
		}
5283
	}
5284
	rcu_read_unlock();
5285 5286 5287

	raw_spin_lock(&this_rq->lock);

5288 5289 5290 5291 5292 5293 5294 5295 5296 5297
	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.
		 */
		this_rq->next_balance = next_balance;
	}
}

/*
5298 5299 5300 5301
 * active_load_balance_cpu_stop is run by cpu stopper. It pushes
 * running tasks off the busiest CPU onto idle CPUs. It requires at
 * least 1 task to be running on each physical CPU where possible, and
 * avoids physical / logical imbalances.
5302
 */
5303
static int active_load_balance_cpu_stop(void *data)
5304
{
5305 5306
	struct rq *busiest_rq = data;
	int busiest_cpu = cpu_of(busiest_rq);
5307
	int target_cpu = busiest_rq->push_cpu;
5308
	struct rq *target_rq = cpu_rq(target_cpu);
5309
	struct sched_domain *sd;
5310 5311 5312 5313 5314 5315 5316

	raw_spin_lock_irq(&busiest_rq->lock);

	/* make sure the requested cpu hasn't gone down in the meantime */
	if (unlikely(busiest_cpu != smp_processor_id() ||
		     !busiest_rq->active_balance))
		goto out_unlock;
5317 5318 5319

	/* Is there any task to move? */
	if (busiest_rq->nr_running <= 1)
5320
		goto out_unlock;
5321 5322 5323 5324 5325 5326 5327 5328 5329 5330 5331 5332

	/*
	 * This condition is "impossible", if it occurs
	 * we need to fix it. Originally reported by
	 * Bjorn Helgaas on a 128-cpu setup.
	 */
	BUG_ON(busiest_rq == target_rq);

	/* move a task from busiest_rq to target_rq */
	double_lock_balance(busiest_rq, target_rq);

	/* Search for an sd spanning us and the target CPU. */
5333
	rcu_read_lock();
5334 5335 5336 5337 5338 5339 5340
	for_each_domain(target_cpu, sd) {
		if ((sd->flags & SD_LOAD_BALANCE) &&
		    cpumask_test_cpu(busiest_cpu, sched_domain_span(sd)))
				break;
	}

	if (likely(sd)) {
5341 5342
		struct lb_env env = {
			.sd		= sd,
5343 5344 5345 5346
			.dst_cpu	= target_cpu,
			.dst_rq		= target_rq,
			.src_cpu	= busiest_rq->cpu,
			.src_rq		= busiest_rq,
5347 5348 5349
			.idle		= CPU_IDLE,
		};

5350 5351
		schedstat_inc(sd, alb_count);

5352
		if (move_one_task(&env))
5353 5354 5355 5356
			schedstat_inc(sd, alb_pushed);
		else
			schedstat_inc(sd, alb_failed);
	}
5357
	rcu_read_unlock();
5358
	double_unlock_balance(busiest_rq, target_rq);
5359 5360 5361 5362
out_unlock:
	busiest_rq->active_balance = 0;
	raw_spin_unlock_irq(&busiest_rq->lock);
	return 0;
5363 5364
}

5365
#ifdef CONFIG_NO_HZ_COMMON
5366 5367 5368 5369 5370 5371
/*
 * idle load balancing details
 * - When one of the busy CPUs notice that there may be an idle rebalancing
 *   needed, they will kick the idle load balancer, which then does idle
 *   load balancing for all the idle CPUs.
 */
5372
static struct {
5373
	cpumask_var_t idle_cpus_mask;
5374
	atomic_t nr_cpus;
5375 5376
	unsigned long next_balance;     /* in jiffy units */
} nohz ____cacheline_aligned;
5377

5378
static inline int find_new_ilb(int call_cpu)
5379
{
5380
	int ilb = cpumask_first(nohz.idle_cpus_mask);
5381

5382 5383 5384 5385
	if (ilb < nr_cpu_ids && idle_cpu(ilb))
		return ilb;

	return nr_cpu_ids;
5386 5387
}

5388 5389 5390 5391 5392 5393 5394 5395 5396 5397 5398
/*
 * Kick a CPU to do the nohz balancing, if it is time for it. We pick the
 * nohz_load_balancer CPU (if there is one) otherwise fallback to any idle
 * CPU (if there is one).
 */
static void nohz_balancer_kick(int cpu)
{
	int ilb_cpu;

	nohz.next_balance++;

5399
	ilb_cpu = find_new_ilb(cpu);
5400

5401 5402
	if (ilb_cpu >= nr_cpu_ids)
		return;
5403

5404
	if (test_and_set_bit(NOHZ_BALANCE_KICK, nohz_flags(ilb_cpu)))
5405 5406 5407 5408 5409 5410 5411 5412
		return;
	/*
	 * Use smp_send_reschedule() instead of resched_cpu().
	 * This way we generate a sched IPI on the target cpu which
	 * is idle. And the softirq performing nohz idle load balance
	 * will be run before returning from the IPI.
	 */
	smp_send_reschedule(ilb_cpu);
5413 5414 5415
	return;
}

5416
static inline void nohz_balance_exit_idle(int cpu)
5417 5418 5419 5420 5421 5422 5423 5424
{
	if (unlikely(test_bit(NOHZ_TICK_STOPPED, nohz_flags(cpu)))) {
		cpumask_clear_cpu(cpu, nohz.idle_cpus_mask);
		atomic_dec(&nohz.nr_cpus);
		clear_bit(NOHZ_TICK_STOPPED, nohz_flags(cpu));
	}
}

5425 5426 5427 5428 5429
static inline void set_cpu_sd_state_busy(void)
{
	struct sched_domain *sd;

	rcu_read_lock();
Nathan Zimmer's avatar
Nathan Zimmer committed
5430
	sd = rcu_dereference_check_sched_domain(this_rq()->sd);
5431 5432 5433 5434 5435 5436

	if (!sd || !sd->nohz_idle)
		goto unlock;
	sd->nohz_idle = 0;

	for (; sd; sd = sd->parent)
5437
		atomic_inc(&sd->groups->sgp->nr_busy_cpus);
5438
unlock:
5439 5440 5441 5442 5443 5444 5445 5446
	rcu_read_unlock();
}

void set_cpu_sd_state_idle(void)
{
	struct sched_domain *sd;

	rcu_read_lock();
Nathan Zimmer's avatar
Nathan Zimmer committed
5447
	sd = rcu_dereference_check_sched_domain(this_rq()->sd);
5448 5449 5450 5451 5452 5453

	if (!sd || sd->nohz_idle)
		goto unlock;
	sd->nohz_idle = 1;

	for (; sd; sd = sd->parent)
5454
		atomic_dec(&sd->groups->sgp->nr_busy_cpus);
5455
unlock:
5456 5457 5458
	rcu_read_unlock();
}

5459
/*
5460
 * This routine will record that the cpu is going idle with tick stopped.
5461
 * This info will be used in performing idle load balancing in the future.
5462
 */
5463
void nohz_balance_enter_idle(int cpu)
5464
{
5465 5466 5467 5468 5469 5470
	/*
	 * If this cpu is going down, then nothing needs to be done.
	 */
	if (!cpu_active(cpu))
		return;

5471 5472
	if (test_bit(NOHZ_TICK_STOPPED, nohz_flags(cpu)))
		return;
5473

5474 5475 5476
	cpumask_set_cpu(cpu, nohz.idle_cpus_mask);
	atomic_inc(&nohz.nr_cpus);
	set_bit(NOHZ_TICK_STOPPED, nohz_flags(cpu));
5477
}
5478 5479 5480 5481 5482 5483

static int __cpuinit sched_ilb_notifier(struct notifier_block *nfb,
					unsigned long action, void *hcpu)
{
	switch (action & ~CPU_TASKS_FROZEN) {
	case CPU_DYING:
5484
		nohz_balance_exit_idle(smp_processor_id());
5485 5486 5487 5488 5489
		return NOTIFY_OK;
	default:
		return NOTIFY_DONE;
	}
}
5490 5491 5492 5493
#endif

static DEFINE_SPINLOCK(balancing);

5494 5495 5496 5497
/*
 * Scale the max load_balance interval with the number of CPUs in the system.
 * This trades load-balance latency on larger machines for less cross talk.
 */
5498
void update_max_interval(void)
5499 5500 5501 5502
{
	max_load_balance_interval = HZ*num_online_cpus()/10;
}

5503 5504 5505 5506
/*
 * It checks each scheduling domain to see if it is due to be balanced,
 * and initiates a balancing operation if so.
 *
5507
 * Balancing parameters are set up in init_sched_domains.
5508 5509 5510 5511 5512 5513
 */
static void rebalance_domains(int cpu, enum cpu_idle_type idle)
{
	int balance = 1;
	struct rq *rq = cpu_rq(cpu);
	unsigned long interval;
5514
	struct sched_domain *sd;
5515 5516 5517 5518 5519
	/* Earliest time when we have to do rebalance again */
	unsigned long next_balance = jiffies + 60*HZ;
	int update_next_balance = 0;
	int need_serialize;

5520
	update_blocked_averages(cpu);
5521

5522
	rcu_read_lock();
5523 5524 5525 5526 5527 5528 5529 5530 5531 5532
	for_each_domain(cpu, sd) {
		if (!(sd->flags & SD_LOAD_BALANCE))
			continue;

		interval = sd->balance_interval;
		if (idle != CPU_IDLE)
			interval *= sd->busy_factor;

		/* scale ms to jiffies */
		interval = msecs_to_jiffies(interval);
5533
		interval = clamp(interval, 1UL, max_load_balance_interval);
5534 5535 5536 5537 5538 5539 5540 5541 5542 5543 5544

		need_serialize = sd->flags & SD_SERIALIZE;

		if (need_serialize) {
			if (!spin_trylock(&balancing))
				goto out;
		}

		if (time_after_eq(jiffies, sd->last_balance + interval)) {
			if (load_balance(cpu, rq, sd, idle, &balance)) {
				/*
5545 5546 5547
				 * The LBF_SOME_PINNED logic could have changed
				 * env->dst_cpu, so we can't know our idle
				 * state even if we migrated tasks. Update it.
5548
				 */
5549
				idle = idle_cpu(cpu) ? CPU_IDLE : CPU_NOT_IDLE;
5550 5551 5552 5553 5554 5555 5556 5557 5558 5559 5560 5561 5562 5563 5564 5565 5566 5567 5568
			}
			sd->last_balance = jiffies;
		}
		if (need_serialize)
			spin_unlock(&balancing);
out:
		if (time_after(next_balance, sd->last_balance + interval)) {
			next_balance = sd->last_balance + interval;
			update_next_balance = 1;
		}

		/*
		 * Stop the load balance at this level. There is another
		 * CPU in our sched group which is doing load balancing more
		 * actively.
		 */
		if (!balance)
			break;
	}
5569
	rcu_read_unlock();
5570 5571 5572 5573 5574 5575 5576 5577 5578 5579

	/*
	 * next_balance will be updated only when there is a need.
	 * When the cpu is attached to null domain for ex, it will not be
	 * updated.
	 */
	if (likely(update_next_balance))
		rq->next_balance = next_balance;
}

5580
#ifdef CONFIG_NO_HZ_COMMON
5581
/*
5582
 * In CONFIG_NO_HZ_COMMON case, the idle balance kickee will do the
5583 5584
 * rebalancing for all the cpus for whom scheduler ticks are stopped.
 */
5585 5586 5587 5588 5589 5590
static void nohz_idle_balance(int this_cpu, enum cpu_idle_type idle)
{
	struct rq *this_rq = cpu_rq(this_cpu);
	struct rq *rq;
	int balance_cpu;

5591 5592 5593
	if (idle != CPU_IDLE ||
	    !test_bit(NOHZ_BALANCE_KICK, nohz_flags(this_cpu)))
		goto end;
5594 5595

	for_each_cpu(balance_cpu, nohz.idle_cpus_mask) {
5596
		if (balance_cpu == this_cpu || !idle_cpu(balance_cpu))
5597 5598 5599 5600 5601 5602 5603
			continue;

		/*
		 * If this cpu gets work to do, stop the load balancing
		 * work being done for other cpus. Next load
		 * balancing owner will pick it up.
		 */
5604
		if (need_resched())
5605 5606
			break;

5607 5608 5609 5610 5611 5612
		rq = cpu_rq(balance_cpu);

		raw_spin_lock_irq(&rq->lock);
		update_rq_clock(rq);
		update_idle_cpu_load(rq);
		raw_spin_unlock_irq(&rq->lock);
5613 5614 5615 5616 5617 5618 5619

		rebalance_domains(balance_cpu, CPU_IDLE);

		if (time_after(this_rq->next_balance, rq->next_balance))
			this_rq->next_balance = rq->next_balance;
	}
	nohz.next_balance = this_rq->next_balance;
5620 5621
end:
	clear_bit(NOHZ_BALANCE_KICK, nohz_flags(this_cpu));
5622 5623 5624
}

/*
5625 5626 5627 5628 5629 5630 5631
 * Current heuristic for kicking the idle load balancer in the presence
 * of an idle cpu is the system.
 *   - This rq has more than one task.
 *   - At any scheduler domain level, this cpu's scheduler group has multiple
 *     busy cpu's exceeding the group's power.
 *   - For SD_ASYM_PACKING, if the lower numbered cpu's in the scheduler
 *     domain span are idle.
5632 5633 5634 5635
 */
static inline int nohz_kick_needed(struct rq *rq, int cpu)
{
	unsigned long now = jiffies;
5636
	struct sched_domain *sd;
5637

5638
	if (unlikely(idle_cpu(cpu)))
5639 5640
		return 0;

5641 5642 5643 5644
       /*
	* We may be recently in ticked or tickless idle mode. At the first
	* busy tick after returning from idle, we will update the busy stats.
	*/
5645
	set_cpu_sd_state_busy();
5646
	nohz_balance_exit_idle(cpu);
5647 5648 5649 5650 5651 5652 5653

	/*
	 * None are in tickless mode and hence no need for NOHZ idle load
	 * balancing.
	 */
	if (likely(!atomic_read(&nohz.nr_cpus)))
		return 0;
5654 5655

	if (time_before(now, nohz.next_balance))
5656 5657
		return 0;

5658 5659
	if (rq->nr_running >= 2)
		goto need_kick;
5660

5661
	rcu_read_lock();
5662 5663 5664 5665
	for_each_domain(cpu, sd) {
		struct sched_group *sg = sd->groups;
		struct sched_group_power *sgp = sg->sgp;
		int nr_busy = atomic_read(&sgp->nr_busy_cpus);
5666

5667
		if (sd->flags & SD_SHARE_PKG_RESOURCES && nr_busy > 1)
5668
			goto need_kick_unlock;
5669 5670 5671 5672

		if (sd->flags & SD_ASYM_PACKING && nr_busy != sg->group_weight
		    && (cpumask_first_and(nohz.idle_cpus_mask,
					  sched_domain_span(sd)) < cpu))
5673
			goto need_kick_unlock;
5674 5675 5676

		if (!(sd->flags & (SD_SHARE_PKG_RESOURCES | SD_ASYM_PACKING)))
			break;
5677
	}
5678
	rcu_read_unlock();
5679
	return 0;
5680 5681 5682

need_kick_unlock:
	rcu_read_unlock();
5683 5684
need_kick:
	return 1;
5685 5686 5687 5688 5689 5690 5691 5692 5693
}
#else
static void nohz_idle_balance(int this_cpu, enum cpu_idle_type idle) { }
#endif

/*
 * run_rebalance_domains is triggered when needed from the scheduler tick.
 * Also triggered for nohz idle balancing (with nohz_balancing_kick set).
 */
5694 5695 5696 5697
static void run_rebalance_domains(struct softirq_action *h)
{
	int this_cpu = smp_processor_id();
	struct rq *this_rq = cpu_rq(this_cpu);
5698
	enum cpu_idle_type idle = this_rq->idle_balance ?
5699 5700 5701 5702 5703
						CPU_IDLE : CPU_NOT_IDLE;

	rebalance_domains(this_cpu, idle);

	/*
5704
	 * If this cpu has a pending nohz_balance_kick, then do the
5705 5706 5707
	 * balancing on behalf of the other idle cpus whose ticks are
	 * stopped.
	 */
5708
	nohz_idle_balance(this_cpu, idle);
5709 5710 5711 5712
}

static inline int on_null_domain(int cpu)
{
5713
	return !rcu_dereference_sched(cpu_rq(cpu)->sd);
5714 5715 5716 5717 5718
}

/*
 * Trigger the SCHED_SOFTIRQ if it is time to do periodic load balancing.
 */
5719
void trigger_load_balance(struct rq *rq, int cpu)
5720 5721 5722 5723 5724
{
	/* Don't need to rebalance while attached to NULL domain */
	if (time_after_eq(jiffies, rq->next_balance) &&
	    likely(!on_null_domain(cpu)))
		raise_softirq(SCHED_SOFTIRQ);
5725
#ifdef CONFIG_NO_HZ_COMMON
5726
	if (nohz_kick_needed(rq, cpu) && likely(!on_null_domain(cpu)))
5727 5728
		nohz_balancer_kick(cpu);
#endif
5729 5730
}

5731 5732 5733 5734 5735 5736 5737 5738
static void rq_online_fair(struct rq *rq)
{
	update_sysctl();
}

static void rq_offline_fair(struct rq *rq)
{
	update_sysctl();
5739 5740 5741

	/* Ensure any throttled groups are reachable by pick_next_task */
	unthrottle_offline_cfs_rqs(rq);
5742 5743
}

5744
#endif /* CONFIG_SMP */
5745

5746 5747 5748
/*
 * scheduler tick hitting a task of our scheduling class:
 */
5749
static void task_tick_fair(struct rq *rq, struct task_struct *curr, int queued)
5750 5751 5752 5753 5754 5755
{
	struct cfs_rq *cfs_rq;
	struct sched_entity *se = &curr->se;

	for_each_sched_entity(se) {
		cfs_rq = cfs_rq_of(se);
5756
		entity_tick(cfs_rq, se, queued);
5757
	}
5758

5759 5760
	if (sched_feat_numa(NUMA))
		task_tick_numa(rq, curr);
5761

5762
	update_rq_runnable_avg(rq, 1);
5763 5764 5765
}

/*
5766 5767 5768
 * called on fork with the child task as argument from the parent's context
 *  - child not yet on the tasklist
 *  - preemption disabled
5769
 */
5770
static void task_fork_fair(struct task_struct *p)
5771
{
5772 5773
	struct cfs_rq *cfs_rq;
	struct sched_entity *se = &p->se, *curr;
5774
	int this_cpu = smp_processor_id();
5775 5776 5777
	struct rq *rq = this_rq();
	unsigned long flags;

5778
	raw_spin_lock_irqsave(&rq->lock, flags);
5779

5780 5781
	update_rq_clock(rq);

5782 5783 5784
	cfs_rq = task_cfs_rq(current);
	curr = cfs_rq->curr;

5785 5786
	if (unlikely(task_cpu(p) != this_cpu)) {
		rcu_read_lock();
5787
		__set_task_cpu(p, this_cpu);
5788 5789
		rcu_read_unlock();
	}
5790

5791
	update_curr(cfs_rq);
5792

5793 5794
	if (curr)
		se->vruntime = curr->vruntime;
5795
	place_entity(cfs_rq, se, 1);
5796

5797
	if (sysctl_sched_child_runs_first && curr && entity_before(curr, se)) {
5798
		/*
5799 5800 5801
		 * Upon rescheduling, sched_class::put_prev_task() will place
		 * 'current' within the tree based on its new key value.
		 */
5802
		swap(curr->vruntime, se->vruntime);
5803
		resched_task(rq->curr);
5804
	}
5805

5806 5807
	se->vruntime -= cfs_rq->min_vruntime;

5808
	raw_spin_unlock_irqrestore(&rq->lock, flags);
5809 5810
}

5811 5812 5813 5814
/*
 * Priority of the task has changed. Check to see if we preempt
 * the current task.
 */
5815 5816
static void
prio_changed_fair(struct rq *rq, struct task_struct *p, int oldprio)
5817
{
5818 5819 5820
	if (!p->se.on_rq)
		return;

5821 5822 5823 5824 5825
	/*
	 * Reschedule if we are currently running on this runqueue and
	 * our priority decreased, or if we are not currently running on
	 * this runqueue and our priority is higher than the current's
	 */
5826
	if (rq->curr == p) {
5827 5828 5829
		if (p->prio > oldprio)
			resched_task(rq->curr);
	} else
5830
		check_preempt_curr(rq, p, 0);
5831 5832
}

5833 5834 5835 5836 5837 5838 5839 5840 5841 5842 5843 5844 5845 5846 5847 5848 5849 5850 5851 5852 5853 5854
static void switched_from_fair(struct rq *rq, struct task_struct *p)
{
	struct sched_entity *se = &p->se;
	struct cfs_rq *cfs_rq = cfs_rq_of(se);

	/*
	 * Ensure the task's vruntime is normalized, so that when its
	 * switched back to the fair class the enqueue_entity(.flags=0) will
	 * do the right thing.
	 *
	 * If it was on_rq, then the dequeue_entity(.flags=0) will already
	 * have normalized the vruntime, if it was !on_rq, then only when
	 * the task is sleeping will it still have non-normalized vruntime.
	 */
	if (!se->on_rq && p->state != TASK_RUNNING) {
		/*
		 * Fix up our vruntime so that the current sleep doesn't
		 * cause 'unlimited' sleep bonus.
		 */
		place_entity(cfs_rq, se, 0);
		se->vruntime -= cfs_rq->min_vruntime;
	}
5855

5856
#ifdef CONFIG_SMP
5857 5858 5859 5860 5861 5862 5863 5864 5865 5866 5867 5868
	/*
	* Remove our load from contribution when we leave sched_fair
	* and ensure we don't carry in an old decay_count if we
	* switch back.
	*/
	if (p->se.avg.decay_count) {
		struct cfs_rq *cfs_rq = cfs_rq_of(&p->se);
		__synchronize_entity_decay(&p->se);
		subtract_blocked_load_contrib(cfs_rq,
				p->se.avg.load_avg_contrib);
	}
#endif
5869 5870
}

5871 5872 5873
/*
 * We switched to the sched_fair class.
 */
5874
static void switched_to_fair(struct rq *rq, struct task_struct *p)
5875
{
5876 5877 5878
	if (!p->se.on_rq)
		return;

5879 5880 5881 5882 5883
	/*
	 * We were most likely switched from sched_rt, so
	 * kick off the schedule if running, otherwise just see
	 * if we can still preempt the current task.
	 */
5884
	if (rq->curr == p)
5885 5886
		resched_task(rq->curr);
	else
5887
		check_preempt_curr(rq, p, 0);
5888 5889
}

5890 5891 5892 5893 5894 5895 5896 5897 5898
/* Account for a task changing its policy or group.
 *
 * This routine is mostly called to set cfs_rq->curr field when a task
 * migrates between groups/classes.
 */
static void set_curr_task_fair(struct rq *rq)
{
	struct sched_entity *se = &rq->curr->se;

5899 5900 5901 5902 5903 5904 5905
	for_each_sched_entity(se) {
		struct cfs_rq *cfs_rq = cfs_rq_of(se);

		set_next_entity(cfs_rq, se);
		/* ensure bandwidth has been allocated on our new cfs_rq */
		account_cfs_rq_runtime(cfs_rq, 0);
	}
5906 5907
}

5908 5909 5910 5911 5912 5913 5914
void init_cfs_rq(struct cfs_rq *cfs_rq)
{
	cfs_rq->tasks_timeline = RB_ROOT;
	cfs_rq->min_vruntime = (u64)(-(1LL << 20));
#ifndef CONFIG_64BIT
	cfs_rq->min_vruntime_copy = cfs_rq->min_vruntime;
#endif
5915
#ifdef CONFIG_SMP
5916
	atomic64_set(&cfs_rq->decay_counter, 1);
5917
	atomic64_set(&cfs_rq->removed_load, 0);
5918
#endif
5919 5920
}

Peter Zijlstra's avatar
Peter Zijlstra committed
5921
#ifdef CONFIG_FAIR_GROUP_SCHED
5922
static void task_move_group_fair(struct task_struct *p, int on_rq)
Peter Zijlstra's avatar
Peter Zijlstra committed
5923
{
5924
	struct cfs_rq *cfs_rq;
5925 5926 5927 5928 5929 5930 5931 5932 5933 5934 5935 5936 5937
	/*
	 * If the task was not on the rq at the time of this cgroup movement
	 * it must have been asleep, sleeping tasks keep their ->vruntime
	 * absolute on their old rq until wakeup (needed for the fair sleeper
	 * bonus in place_entity()).
	 *
	 * If it was on the rq, we've just 'preempted' it, which does convert
	 * ->vruntime to a relative base.
	 *
	 * Make sure both cases convert their relative position when migrating
	 * to another cgroup's rq. This does somewhat interfere with the
	 * fair sleeper stuff for the first placement, but who cares.
	 */
5938 5939 5940 5941 5942 5943
	/*
	 * When !on_rq, vruntime of the task has usually NOT been normalized.
	 * But there are some cases where it has already been normalized:
	 *
	 * - Moving a forked child which is waiting for being woken up by
	 *   wake_up_new_task().
5944 5945
	 * - Moving a task which has been woken up by try_to_wake_up() and
	 *   waiting for actually being woken up by sched_ttwu_pending().
5946 5947 5948 5949
	 *
	 * To prevent boost or penalty in the new cfs_rq caused by delta
	 * min_vruntime between the two cfs_rqs, we skip vruntime adjustment.
	 */
5950
	if (!on_rq && (!p->se.sum_exec_runtime || p->state == TASK_WAKING))
5951 5952
		on_rq = 1;

5953 5954 5955
	if (!on_rq)
		p->se.vruntime -= cfs_rq_of(&p->se)->min_vruntime;
	set_task_rq(p, task_cpu(p));
5956 5957 5958 5959 5960 5961 5962 5963 5964 5965 5966 5967 5968
	if (!on_rq) {
		cfs_rq = cfs_rq_of(&p->se);
		p->se.vruntime += cfs_rq->min_vruntime;
#ifdef CONFIG_SMP
		/*
		 * migrate_task_rq_fair() will have removed our previous
		 * contribution, but we must synchronize for ongoing future
		 * decay.
		 */
		p->se.avg.decay_count = atomic64_read(&cfs_rq->decay_counter);
		cfs_rq->blocked_load_avg += p->se.avg.load_avg_contrib;
#endif
	}
Peter Zijlstra's avatar
Peter Zijlstra committed
5969
}
5970 5971 5972 5973 5974 5975 5976 5977 5978 5979 5980 5981 5982 5983 5984 5985 5986 5987 5988 5989 5990 5991 5992 5993 5994 5995 5996 5997 5998 5999 6000 6001 6002 6003 6004 6005 6006 6007 6008 6009 6010 6011 6012 6013 6014 6015 6016 6017 6018 6019 6020 6021 6022 6023 6024 6025 6026 6027 6028 6029 6030 6031 6032 6033 6034 6035 6036 6037 6038 6039 6040 6041 6042 6043 6044 6045 6046 6047 6048 6049 6050 6051 6052 6053 6054 6055 6056 6057 6058 6059 6060 6061 6062 6063 6064 6065 6066 6067 6068 6069 6070 6071 6072 6073 6074 6075 6076 6077 6078 6079 6080 6081 6082 6083 6084 6085 6086 6087 6088 6089 6090 6091 6092 6093 6094 6095 6096 6097 6098

void free_fair_sched_group(struct task_group *tg)
{
	int i;

	destroy_cfs_bandwidth(tg_cfs_bandwidth(tg));

	for_each_possible_cpu(i) {
		if (tg->cfs_rq)
			kfree(tg->cfs_rq[i]);
		if (tg->se)
			kfree(tg->se[i]);
	}

	kfree(tg->cfs_rq);
	kfree(tg->se);
}

int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent)
{
	struct cfs_rq *cfs_rq;
	struct sched_entity *se;
	int i;

	tg->cfs_rq = kzalloc(sizeof(cfs_rq) * nr_cpu_ids, GFP_KERNEL);
	if (!tg->cfs_rq)
		goto err;
	tg->se = kzalloc(sizeof(se) * nr_cpu_ids, GFP_KERNEL);
	if (!tg->se)
		goto err;

	tg->shares = NICE_0_LOAD;

	init_cfs_bandwidth(tg_cfs_bandwidth(tg));

	for_each_possible_cpu(i) {
		cfs_rq = kzalloc_node(sizeof(struct cfs_rq),
				      GFP_KERNEL, cpu_to_node(i));
		if (!cfs_rq)
			goto err;

		se = kzalloc_node(sizeof(struct sched_entity),
				  GFP_KERNEL, cpu_to_node(i));
		if (!se)
			goto err_free_rq;

		init_cfs_rq(cfs_rq);
		init_tg_cfs_entry(tg, cfs_rq, se, i, parent->se[i]);
	}

	return 1;

err_free_rq:
	kfree(cfs_rq);
err:
	return 0;
}

void unregister_fair_sched_group(struct task_group *tg, int cpu)
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

	/*
	* Only empty task groups can be destroyed; so we can speculatively
	* check on_list without danger of it being re-added.
	*/
	if (!tg->cfs_rq[cpu]->on_list)
		return;

	raw_spin_lock_irqsave(&rq->lock, flags);
	list_del_leaf_cfs_rq(tg->cfs_rq[cpu]);
	raw_spin_unlock_irqrestore(&rq->lock, flags);
}

void init_tg_cfs_entry(struct task_group *tg, struct cfs_rq *cfs_rq,
			struct sched_entity *se, int cpu,
			struct sched_entity *parent)
{
	struct rq *rq = cpu_rq(cpu);

	cfs_rq->tg = tg;
	cfs_rq->rq = rq;
	init_cfs_rq_runtime(cfs_rq);

	tg->cfs_rq[cpu] = cfs_rq;
	tg->se[cpu] = se;

	/* se could be NULL for root_task_group */
	if (!se)
		return;

	if (!parent)
		se->cfs_rq = &rq->cfs;
	else
		se->cfs_rq = parent->my_q;

	se->my_q = cfs_rq;
	update_load_set(&se->load, 0);
	se->parent = parent;
}

static DEFINE_MUTEX(shares_mutex);

int sched_group_set_shares(struct task_group *tg, unsigned long shares)
{
	int i;
	unsigned long flags;

	/*
	 * We can't change the weight of the root cgroup.
	 */
	if (!tg->se[0])
		return -EINVAL;

	shares = clamp(shares, scale_load(MIN_SHARES), scale_load(MAX_SHARES));

	mutex_lock(&shares_mutex);
	if (tg->shares == shares)
		goto done;

	tg->shares = shares;
	for_each_possible_cpu(i) {
		struct rq *rq = cpu_rq(i);
		struct sched_entity *se;

		se = tg->se[i];
		/* Propagate contribution to hierarchy */
		raw_spin_lock_irqsave(&rq->lock, flags);
6099 6100 6101

		/* Possible calls to update_curr() need rq clock */
		update_rq_clock(rq);
6102
		for_each_sched_entity(se)
6103 6104 6105 6106 6107 6108 6109 6110 6111 6112 6113 6114 6115 6116 6117 6118 6119 6120 6121 6122 6123
			update_cfs_shares(group_cfs_rq(se));
		raw_spin_unlock_irqrestore(&rq->lock, flags);
	}

done:
	mutex_unlock(&shares_mutex);
	return 0;
}
#else /* CONFIG_FAIR_GROUP_SCHED */

void free_fair_sched_group(struct task_group *tg) { }

int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent)
{
	return 1;
}

void unregister_fair_sched_group(struct task_group *tg, int cpu) { }

#endif /* CONFIG_FAIR_GROUP_SCHED */

Peter Zijlstra's avatar
Peter Zijlstra committed
6124

6125
static unsigned int get_rr_interval_fair(struct rq *rq, struct task_struct *task)
6126 6127 6128 6129 6130 6131 6132 6133 6134
{
	struct sched_entity *se = &task->se;
	unsigned int rr_interval = 0;

	/*
	 * Time slice is 0 for SCHED_OTHER tasks that are on an otherwise
	 * idle runqueue:
	 */
	if (rq->cfs.load.weight)
6135
		rr_interval = NS_TO_JIFFIES(sched_slice(cfs_rq_of(se), se));
6136 6137 6138 6139

	return rr_interval;
}

6140 6141 6142
/*
 * All the scheduling class methods:
 */
6143
const struct sched_class fair_sched_class = {
6144
	.next			= &idle_sched_class,
6145 6146 6147
	.enqueue_task		= enqueue_task_fair,
	.dequeue_task		= dequeue_task_fair,
	.yield_task		= yield_task_fair,
6148
	.yield_to_task		= yield_to_task_fair,
6149

6150
	.check_preempt_curr	= check_preempt_wakeup,
6151 6152 6153 6154

	.pick_next_task		= pick_next_task_fair,
	.put_prev_task		= put_prev_task_fair,

6155
#ifdef CONFIG_SMP
Li Zefan's avatar
Li Zefan committed
6156
	.select_task_rq		= select_task_rq_fair,
6157
	.migrate_task_rq	= migrate_task_rq_fair,
6158

6159 6160
	.rq_online		= rq_online_fair,
	.rq_offline		= rq_offline_fair,
6161 6162

	.task_waking		= task_waking_fair,
6163
#endif
6164

6165
	.set_curr_task          = set_curr_task_fair,
6166
	.task_tick		= task_tick_fair,
6167
	.task_fork		= task_fork_fair,
6168 6169

	.prio_changed		= prio_changed_fair,
6170
	.switched_from		= switched_from_fair,
6171
	.switched_to		= switched_to_fair,
Peter Zijlstra's avatar
Peter Zijlstra committed
6172

6173 6174
	.get_rr_interval	= get_rr_interval_fair,

Peter Zijlstra's avatar
Peter Zijlstra committed
6175
#ifdef CONFIG_FAIR_GROUP_SCHED
6176
	.task_move_group	= task_move_group_fair,
Peter Zijlstra's avatar
Peter Zijlstra committed
6177
#endif
6178 6179 6180
};

#ifdef CONFIG_SCHED_DEBUG
6181
void print_cfs_stats(struct seq_file *m, int cpu)
6182 6183 6184
{
	struct cfs_rq *cfs_rq;

6185
	rcu_read_lock();
6186
	for_each_leaf_cfs_rq(cpu_rq(cpu), cfs_rq)
6187
		print_cfs_rq(m, cpu, cfs_rq);
6188
	rcu_read_unlock();
6189 6190
}
#endif
6191 6192 6193 6194 6195 6196

__init void init_sched_fair_class(void)
{
#ifdef CONFIG_SMP
	open_softirq(SCHED_SOFTIRQ, run_rebalance_domains);

6197
#ifdef CONFIG_NO_HZ_COMMON
6198
	nohz.next_balance = jiffies;
6199
	zalloc_cpumask_var(&nohz.idle_cpus_mask, GFP_NOWAIT);
6200
	cpu_notifier(sched_ilb_notifier, 0);
6201 6202 6203 6204
#endif
#endif /* SMP */

}