fair.c 159.1 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
 */

A
Arjan van de Ven 已提交
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"
A
Arjan van de Ven 已提交
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
I
Ingo Molnar 已提交
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
 *
I
Ingo Molnar 已提交
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);
}

P
Peter Zijlstra 已提交
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;
	}
}

P
Peter Zijlstra 已提交
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

P
Peter Zijlstra 已提交
394 395
#define for_each_sched_entity(se) \
		for (; se; se = NULL)
396

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

P
Peter Zijlstra 已提交
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)
{
}

P
Peter Zijlstra 已提交
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)
{
}

P
Peter Zijlstra 已提交
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)
P
Peter Zijlstra 已提交
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);

P
Peter Zijlstra 已提交
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)
I
Ingo Molnar 已提交
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
{
P
Peter Zijlstra 已提交
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;
	}
I
Ingo Molnar 已提交
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
 */
P
Peter Zijlstra 已提交
651
static u64 sched_slice(struct cfs_rq *cfs_rq, struct sched_entity *se)
652
{
M
Mike Galbraith 已提交
653
	u64 slice = __sched_period(cfs_rq->nr_running + !se->on_rq);
654

M
Mike Galbraith 已提交
655
	for_each_sched_entity(se) {
L
Lin Ming 已提交
656
		struct load_weight *load;
657
		struct load_weight lw;
L
Lin Ming 已提交
658 659 660

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

M
Mike Galbraith 已提交
662
		if (unlikely(!se->on_rq)) {
663
			lw = cfs_rq->load;
M
Mike Galbraith 已提交
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
/*
A
Andrei Epure 已提交
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)
P
Peter Zijlstra 已提交
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
I
Ingo Molnar 已提交
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

I
Ingo Molnar 已提交
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):
	 */
I
Ingo Molnar 已提交
718
	delta_exec = (unsigned long)(now - curr->exec_start);
P
Peter Zijlstra 已提交
719 720
	if (!delta_exec)
		return;
721

I
Ingo Molnar 已提交
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 */
P
Peter Zijlstra 已提交
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);
P
Peter Zijlstra 已提交
1096
		account_entity_dequeue(cfs_rq, se);
1097
	}
P
Peter Zijlstra 已提交
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)
P
Peter Zijlstra 已提交
1108 1109 1110
{
	struct task_group *tg;
	struct sched_entity *se;
1111
	long shares;
P
Peter Zijlstra 已提交
1112 1113 1114

	tg = cfs_rq->tg;
	se = tg->se[cpu_of(rq_of(cfs_rq))];
1115
	if (!se || throttled_hierarchy(cfs_rq))
P
Peter Zijlstra 已提交
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);
P
Peter Zijlstra 已提交
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)
P
Peter Zijlstra 已提交
1127 1128 1129 1130
{
}
#endif /* CONFIG_FAIR_GROUP_SCHED */

1131 1132
/* Only depends on SMP, FAIR_GROUP_SCHED may be removed when useful in lb */
#if defined(CONFIG_SMP) && defined(CONFIG_FAIR_GROUP_SCHED)
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 1160
/*
 * 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,
};

1161 1162 1163 1164 1165 1166
/*
 * Approximate:
 *   val * y^n,    where y^32 ~= 0.5 (~1 scheduling period)
 */
static __always_inline u64 decay_load(u64 val, u64 n)
{
1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186
	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;
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 1219
	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];
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 1253
}

/*
 * 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)
{
1254 1255
	u64 delta, periods;
	u32 runnable_contrib;
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 1288
	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;
1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308
		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;
1309 1310 1311 1312 1313 1314 1315 1316 1317 1318
	}

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

	return decayed;
}

1319
/* Synchronize an entity's decay with its parenting cfs_rq.*/
1320
static inline u64 __synchronize_entity_decay(struct sched_entity *se)
1321 1322 1323 1324 1325 1326
{
	struct cfs_rq *cfs_rq = cfs_rq_of(se);
	u64 decays = atomic64_read(&cfs_rq->decay_counter);

	decays -= se->avg.decay_count;
	if (!decays)
1327
		return 0;
1328 1329 1330

	se->avg.load_avg_contrib = decay_load(se->avg.load_avg_contrib, decays);
	se->avg.decay_count = 0;
1331 1332

	return decays;
1333 1334
}

1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349
#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;
	}
}
1350

1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371
/*
 * 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;
	}
}

1372 1373 1374 1375
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;
1376 1377
	int runnable_avg;

1378 1379 1380 1381 1382
	u64 contrib;

	contrib = cfs_rq->tg_load_contrib * tg->shares;
	se->avg.load_avg_contrib = div64_u64(contrib,
					     atomic64_read(&tg->load_avg) + 1);
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 1411

	/*
	 * 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;
	}
1412
}
1413 1414 1415
#else
static inline void __update_cfs_rq_tg_load_contrib(struct cfs_rq *cfs_rq,
						 int force_update) {}
1416 1417
static inline void __update_tg_runnable_avg(struct sched_avg *sa,
						  struct cfs_rq *cfs_rq) {}
1418
static inline void __update_group_entity_contrib(struct sched_entity *se) {}
1419 1420
#endif

1421 1422 1423 1424 1425 1426 1427 1428 1429 1430
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);
}

1431 1432 1433 1434 1435
/* 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;

1436 1437 1438
	if (entity_is_task(se)) {
		__update_task_entity_contrib(se);
	} else {
1439
		__update_tg_runnable_avg(&se->avg, group_cfs_rq(se));
1440 1441
		__update_group_entity_contrib(se);
	}
1442 1443 1444 1445

	return se->avg.load_avg_contrib - old_contrib;
}

1446 1447 1448 1449 1450 1451 1452 1453 1454
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;
}

1455 1456
static inline u64 cfs_rq_clock_task(struct cfs_rq *cfs_rq);

1457
/* Update a sched_entity's runnable average */
1458 1459
static inline void update_entity_load_avg(struct sched_entity *se,
					  int update_cfs_rq)
1460
{
1461 1462
	struct cfs_rq *cfs_rq = cfs_rq_of(se);
	long contrib_delta;
1463
	u64 now;
1464

1465 1466 1467 1468 1469 1470 1471 1472 1473 1474
	/*
	 * 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))
1475 1476 1477
		return;

	contrib_delta = __update_entity_load_avg_contrib(se);
1478 1479 1480 1481

	if (!update_cfs_rq)
		return;

1482 1483
	if (se->on_rq)
		cfs_rq->runnable_load_avg += contrib_delta;
1484 1485 1486 1487 1488 1489 1490 1491
	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.
 */
1492
static void update_cfs_rq_blocked_load(struct cfs_rq *cfs_rq, int force_update)
1493
{
1494
	u64 now = cfs_rq_clock_task(cfs_rq) >> 20;
1495 1496 1497
	u64 decays;

	decays = now - cfs_rq->last_decay;
1498
	if (!decays && !force_update)
1499 1500
		return;

1501 1502 1503 1504
	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);
	}
1505

1506 1507 1508 1509 1510 1511
	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;
	}
1512 1513

	__update_cfs_rq_tg_load_contrib(cfs_rq, force_update);
1514
}
1515 1516 1517

static inline void update_rq_runnable_avg(struct rq *rq, int runnable)
{
1518
	__update_entity_runnable_avg(rq_clock_task(rq), &rq->avg, runnable);
1519
	__update_tg_runnable_avg(&rq->avg, &rq->cfs);
1520
}
1521 1522 1523

/* 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,
1524 1525
						  struct sched_entity *se,
						  int wakeup)
1526
{
1527 1528 1529 1530 1531 1532
	/*
	 * 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)) {
1533
		se->avg.last_runnable_update = rq_clock_task(rq_of(cfs_rq));
1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548
		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;
		}
1549 1550 1551 1552 1553
		wakeup = 0;
	} else {
		__synchronize_entity_decay(se);
	}

1554 1555
	/* migrated tasks did not contribute to our blocked load */
	if (wakeup) {
1556
		subtract_blocked_load_contrib(cfs_rq, se->avg.load_avg_contrib);
1557 1558
		update_entity_load_avg(se, 0);
	}
1559

1560
	cfs_rq->runnable_load_avg += se->avg.load_avg_contrib;
1561 1562
	/* we force update consideration on load-balancer moves */
	update_cfs_rq_blocked_load(cfs_rq, !wakeup);
1563 1564
}

1565 1566 1567 1568 1569
/*
 * 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.
 */
1570
static inline void dequeue_entity_load_avg(struct cfs_rq *cfs_rq,
1571 1572
						  struct sched_entity *se,
						  int sleep)
1573
{
1574
	update_entity_load_avg(se, 1);
1575 1576
	/* we force update consideration on load-balancer moves */
	update_cfs_rq_blocked_load(cfs_rq, !sleep);
1577

1578
	cfs_rq->runnable_load_avg -= se->avg.load_avg_contrib;
1579 1580 1581 1582
	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 */
1583
}
1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604

/*
 * 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);
}

1605
#else
1606 1607
static inline void update_entity_load_avg(struct sched_entity *se,
					  int update_cfs_rq) {}
1608
static inline void update_rq_runnable_avg(struct rq *rq, int runnable) {}
1609
static inline void enqueue_entity_load_avg(struct cfs_rq *cfs_rq,
1610 1611
					   struct sched_entity *se,
					   int wakeup) {}
1612
static inline void dequeue_entity_load_avg(struct cfs_rq *cfs_rq,
1613 1614
					   struct sched_entity *se,
					   int sleep) {}
1615 1616
static inline void update_cfs_rq_blocked_load(struct cfs_rq *cfs_rq,
					      int force_update) {}
1617 1618
#endif

1619
static void enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
1620 1621
{
#ifdef CONFIG_SCHEDSTATS
1622 1623 1624 1625 1626
	struct task_struct *tsk = NULL;

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

1627
	if (se->statistics.sleep_start) {
1628
		u64 delta = rq_clock(rq_of(cfs_rq)) - se->statistics.sleep_start;
1629 1630 1631 1632

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

1633 1634
		if (unlikely(delta > se->statistics.sleep_max))
			se->statistics.sleep_max = delta;
1635

1636
		se->statistics.sleep_start = 0;
1637
		se->statistics.sum_sleep_runtime += delta;
A
Arjan van de Ven 已提交
1638

1639
		if (tsk) {
1640
			account_scheduler_latency(tsk, delta >> 10, 1);
1641 1642
			trace_sched_stat_sleep(tsk, delta);
		}
1643
	}
1644
	if (se->statistics.block_start) {
1645
		u64 delta = rq_clock(rq_of(cfs_rq)) - se->statistics.block_start;
1646 1647 1648 1649

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

1650 1651
		if (unlikely(delta > se->statistics.block_max))
			se->statistics.block_max = delta;
1652

1653
		se->statistics.block_start = 0;
1654
		se->statistics.sum_sleep_runtime += delta;
I
Ingo Molnar 已提交
1655

1656
		if (tsk) {
1657
			if (tsk->in_iowait) {
1658 1659
				se->statistics.iowait_sum += delta;
				se->statistics.iowait_count++;
1660
				trace_sched_stat_iowait(tsk, delta);
1661 1662
			}

1663 1664
			trace_sched_stat_blocked(tsk, delta);

1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675
			/*
			 * 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);
I
Ingo Molnar 已提交
1676
		}
1677 1678 1679 1680
	}
#endif
}

P
Peter Zijlstra 已提交
1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693
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
}

1694 1695 1696
static void
place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial)
{
1697
	u64 vruntime = cfs_rq->min_vruntime;
P
Peter Zijlstra 已提交
1698

1699 1700 1701 1702 1703 1704
	/*
	 * 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.
	 */
P
Peter Zijlstra 已提交
1705
	if (initial && sched_feat(START_DEBIT))
1706
		vruntime += sched_vslice(cfs_rq, se);
1707

1708
	/* sleeps up to a single latency don't count. */
1709
	if (!initial) {
1710
		unsigned long thresh = sysctl_sched_latency;
1711

1712 1713 1714 1715 1716 1717
		/*
		 * Halve their sleep time's effect, to allow
		 * for a gentler effect of sleepers:
		 */
		if (sched_feat(GENTLE_FAIR_SLEEPERS))
			thresh >>= 1;
1718

1719
		vruntime -= thresh;
1720 1721
	}

1722
	/* ensure we never gain time by being placed backwards. */
1723
	se->vruntime = max_vruntime(se->vruntime, vruntime);
1724 1725
}

1726 1727
static void check_enqueue_throttle(struct cfs_rq *cfs_rq);

1728
static void
1729
enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
1730
{
1731 1732 1733 1734
	/*
	 * Update the normalized vruntime before updating min_vruntime
	 * through callig update_curr().
	 */
1735
	if (!(flags & ENQUEUE_WAKEUP) || (flags & ENQUEUE_WAKING))
1736 1737
		se->vruntime += cfs_rq->min_vruntime;

1738
	/*
1739
	 * Update run-time statistics of the 'current'.
1740
	 */
1741
	update_curr(cfs_rq);
1742
	enqueue_entity_load_avg(cfs_rq, se, flags & ENQUEUE_WAKEUP);
1743 1744
	account_entity_enqueue(cfs_rq, se);
	update_cfs_shares(cfs_rq);
1745

1746
	if (flags & ENQUEUE_WAKEUP) {
1747
		place_entity(cfs_rq, se, 0);
1748
		enqueue_sleeper(cfs_rq, se);
I
Ingo Molnar 已提交
1749
	}
1750

1751
	update_stats_enqueue(cfs_rq, se);
P
Peter Zijlstra 已提交
1752
	check_spread(cfs_rq, se);
1753 1754
	if (se != cfs_rq->curr)
		__enqueue_entity(cfs_rq, se);
P
Peter Zijlstra 已提交
1755
	se->on_rq = 1;
1756

1757
	if (cfs_rq->nr_running == 1) {
1758
		list_add_leaf_cfs_rq(cfs_rq);
1759 1760
		check_enqueue_throttle(cfs_rq);
	}
1761 1762
}

1763
static void __clear_buddies_last(struct sched_entity *se)
P
Peter Zijlstra 已提交
1764
{
1765 1766 1767 1768 1769 1770 1771 1772
	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;
	}
}
P
Peter Zijlstra 已提交
1773

1774 1775 1776 1777 1778 1779 1780 1781 1782
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;
	}
P
Peter Zijlstra 已提交
1783 1784
}

1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795
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;
	}
}

P
Peter Zijlstra 已提交
1796 1797
static void clear_buddies(struct cfs_rq *cfs_rq, struct sched_entity *se)
{
1798 1799 1800 1801 1802
	if (cfs_rq->last == se)
		__clear_buddies_last(se);

	if (cfs_rq->next == se)
		__clear_buddies_next(se);
1803 1804 1805

	if (cfs_rq->skip == se)
		__clear_buddies_skip(se);
P
Peter Zijlstra 已提交
1806 1807
}

1808
static __always_inline void return_cfs_rq_runtime(struct cfs_rq *cfs_rq);
1809

1810
static void
1811
dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
1812
{
1813 1814 1815 1816
	/*
	 * Update run-time statistics of the 'current'.
	 */
	update_curr(cfs_rq);
1817
	dequeue_entity_load_avg(cfs_rq, se, flags & DEQUEUE_SLEEP);
1818

1819
	update_stats_dequeue(cfs_rq, se);
1820
	if (flags & DEQUEUE_SLEEP) {
P
Peter Zijlstra 已提交
1821
#ifdef CONFIG_SCHEDSTATS
1822 1823 1824 1825
		if (entity_is_task(se)) {
			struct task_struct *tsk = task_of(se);

			if (tsk->state & TASK_INTERRUPTIBLE)
1826
				se->statistics.sleep_start = rq_clock(rq_of(cfs_rq));
1827
			if (tsk->state & TASK_UNINTERRUPTIBLE)
1828
				se->statistics.block_start = rq_clock(rq_of(cfs_rq));
1829
		}
1830
#endif
P
Peter Zijlstra 已提交
1831 1832
	}

P
Peter Zijlstra 已提交
1833
	clear_buddies(cfs_rq, se);
P
Peter Zijlstra 已提交
1834

1835
	if (se != cfs_rq->curr)
1836
		__dequeue_entity(cfs_rq, se);
1837
	se->on_rq = 0;
1838
	account_entity_dequeue(cfs_rq, se);
1839 1840 1841 1842 1843 1844

	/*
	 * 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.
	 */
1845
	if (!(flags & DEQUEUE_SLEEP))
1846
		se->vruntime -= cfs_rq->min_vruntime;
1847

1848 1849 1850
	/* return excess runtime on last dequeue */
	return_cfs_rq_runtime(cfs_rq);

1851
	update_min_vruntime(cfs_rq);
1852
	update_cfs_shares(cfs_rq);
1853 1854 1855 1856 1857
}

/*
 * Preempt the current task with a newly woken task if needed:
 */
1858
static void
I
Ingo Molnar 已提交
1859
check_preempt_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr)
1860
{
1861
	unsigned long ideal_runtime, delta_exec;
1862 1863
	struct sched_entity *se;
	s64 delta;
1864

P
Peter Zijlstra 已提交
1865
	ideal_runtime = sched_slice(cfs_rq, curr);
1866
	delta_exec = curr->sum_exec_runtime - curr->prev_sum_exec_runtime;
1867
	if (delta_exec > ideal_runtime) {
1868
		resched_task(rq_of(cfs_rq)->curr);
1869 1870 1871 1872 1873
		/*
		 * The current task ran long enough, ensure it doesn't get
		 * re-elected due to buddy favours.
		 */
		clear_buddies(cfs_rq, curr);
1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884
		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;

1885 1886
	se = __pick_first_entity(cfs_rq);
	delta = curr->vruntime - se->vruntime;
1887

1888 1889
	if (delta < 0)
		return;
1890

1891 1892
	if (delta > ideal_runtime)
		resched_task(rq_of(cfs_rq)->curr);
1893 1894
}

1895
static void
1896
set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
1897
{
1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908
	/* '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);
	}

1909
	update_stats_curr_start(cfs_rq, se);
1910
	cfs_rq->curr = se;
I
Ingo Molnar 已提交
1911 1912 1913 1914 1915 1916
#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):
	 */
1917
	if (rq_of(cfs_rq)->load.weight >= 2*se->load.weight) {
1918
		se->statistics.slice_max = max(se->statistics.slice_max,
I
Ingo Molnar 已提交
1919 1920 1921
			se->sum_exec_runtime - se->prev_sum_exec_runtime);
	}
#endif
1922
	se->prev_sum_exec_runtime = se->sum_exec_runtime;
1923 1924
}

1925 1926 1927
static int
wakeup_preempt_entity(struct sched_entity *curr, struct sched_entity *se);

1928 1929 1930 1931 1932 1933 1934
/*
 * 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
 */
1935
static struct sched_entity *pick_next_entity(struct cfs_rq *cfs_rq)
1936
{
1937
	struct sched_entity *se = __pick_first_entity(cfs_rq);
1938
	struct sched_entity *left = se;
1939

1940 1941 1942 1943 1944 1945 1946 1947 1948
	/*
	 * 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;
	}
1949

1950 1951 1952 1953 1954 1955
	/*
	 * 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;

1956 1957 1958 1959 1960 1961
	/*
	 * 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;

1962
	clear_buddies(cfs_rq, se);
P
Peter Zijlstra 已提交
1963 1964

	return se;
1965 1966
}

1967 1968
static void check_cfs_rq_runtime(struct cfs_rq *cfs_rq);

1969
static void put_prev_entity(struct cfs_rq *cfs_rq, struct sched_entity *prev)
1970 1971 1972 1973 1974 1975
{
	/*
	 * If still on the runqueue then deactivate_task()
	 * was not called and update_curr() has to be done:
	 */
	if (prev->on_rq)
1976
		update_curr(cfs_rq);
1977

1978 1979 1980
	/* throttle cfs_rqs exceeding runtime */
	check_cfs_rq_runtime(cfs_rq);

P
Peter Zijlstra 已提交
1981
	check_spread(cfs_rq, prev);
1982
	if (prev->on_rq) {
1983
		update_stats_wait_start(cfs_rq, prev);
1984 1985
		/* Put 'current' back into the tree. */
		__enqueue_entity(cfs_rq, prev);
1986
		/* in !on_rq case, update occurred at dequeue */
1987
		update_entity_load_avg(prev, 1);
1988
	}
1989
	cfs_rq->curr = NULL;
1990 1991
}

P
Peter Zijlstra 已提交
1992 1993
static void
entity_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr, int queued)
1994 1995
{
	/*
1996
	 * Update run-time statistics of the 'current'.
1997
	 */
1998
	update_curr(cfs_rq);
1999

2000 2001 2002
	/*
	 * Ensure that runnable average is periodically updated.
	 */
2003
	update_entity_load_avg(curr, 1);
2004
	update_cfs_rq_blocked_load(cfs_rq, 1);
2005

P
Peter Zijlstra 已提交
2006 2007 2008 2009 2010
#ifdef CONFIG_SCHED_HRTICK
	/*
	 * queued ticks are scheduled to match the slice, so don't bother
	 * validating it and just reschedule.
	 */
2011 2012 2013 2014
	if (queued) {
		resched_task(rq_of(cfs_rq)->curr);
		return;
	}
P
Peter Zijlstra 已提交
2015 2016 2017 2018 2019 2020 2021 2022
	/*
	 * 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

Y
Yong Zhang 已提交
2023
	if (cfs_rq->nr_running > 1)
I
Ingo Molnar 已提交
2024
		check_preempt_tick(cfs_rq, curr);
2025 2026
}

2027 2028 2029 2030 2031 2032

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

#ifdef CONFIG_CFS_BANDWIDTH
2033 2034

#ifdef HAVE_JUMP_LABEL
2035
static struct static_key __cfs_bandwidth_used;
2036 2037 2038

static inline bool cfs_bandwidth_used(void)
{
2039
	return static_key_false(&__cfs_bandwidth_used);
2040 2041 2042 2043 2044 2045
}

void account_cfs_bandwidth_used(int enabled, int was_enabled)
{
	/* only need to count groups transitioning between enabled/!enabled */
	if (enabled && !was_enabled)
2046
		static_key_slow_inc(&__cfs_bandwidth_used);
2047
	else if (!enabled && was_enabled)
2048
		static_key_slow_dec(&__cfs_bandwidth_used);
2049 2050 2051 2052 2053 2054 2055 2056 2057 2058
}
#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 */

2059 2060 2061 2062 2063 2064 2065 2066
/*
 * default period for cfs group bandwidth.
 * default: 0.1s, units: nanoseconds
 */
static inline u64 default_cfs_period(void)
{
	return 100000000ULL;
}
2067 2068 2069 2070 2071 2072

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

P
Paul Turner 已提交
2073 2074 2075 2076 2077 2078 2079
/*
 * 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
 */
2080
void __refill_cfs_bandwidth_runtime(struct cfs_bandwidth *cfs_b)
P
Paul Turner 已提交
2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091
{
	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);
}

2092 2093 2094 2095 2096
static inline struct cfs_bandwidth *tg_cfs_bandwidth(struct task_group *tg)
{
	return &tg->cfs_bandwidth;
}

2097 2098 2099 2100 2101 2102
/* 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;

2103
	return rq_clock_task(rq_of(cfs_rq)) - cfs_rq->throttled_clock_task_time;
2104 2105
}

2106 2107
/* returns 0 on failure to allocate runtime */
static int assign_cfs_rq_runtime(struct cfs_rq *cfs_rq)
2108 2109 2110
{
	struct task_group *tg = cfs_rq->tg;
	struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(tg);
P
Paul Turner 已提交
2111
	u64 amount = 0, min_amount, expires;
2112 2113 2114 2115 2116 2117 2118

	/* 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;
2119
	else {
P
Paul Turner 已提交
2120 2121 2122 2123 2124 2125 2126 2127
		/*
		 * 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);
2128
			__start_cfs_bandwidth(cfs_b);
P
Paul Turner 已提交
2129
		}
2130 2131 2132 2133 2134 2135

		if (cfs_b->runtime > 0) {
			amount = min(cfs_b->runtime, min_amount);
			cfs_b->runtime -= amount;
			cfs_b->idle = 0;
		}
2136
	}
P
Paul Turner 已提交
2137
	expires = cfs_b->runtime_expires;
2138 2139 2140
	raw_spin_unlock(&cfs_b->lock);

	cfs_rq->runtime_remaining += amount;
P
Paul Turner 已提交
2141 2142 2143 2144 2145 2146 2147
	/*
	 * 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;
2148 2149

	return cfs_rq->runtime_remaining > 0;
2150 2151
}

P
Paul Turner 已提交
2152 2153 2154 2155 2156
/*
 * 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)
2157
{
P
Paul Turner 已提交
2158 2159 2160
	struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(cfs_rq->tg);

	/* if the deadline is ahead of our clock, nothing to do */
2161
	if (likely((s64)(rq_clock(rq_of(cfs_rq)) - cfs_rq->runtime_expires) < 0))
2162 2163
		return;

P
Paul Turner 已提交
2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188
	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) */
2189
	cfs_rq->runtime_remaining -= delta_exec;
P
Paul Turner 已提交
2190 2191 2192
	expire_cfs_rq_runtime(cfs_rq);

	if (likely(cfs_rq->runtime_remaining > 0))
2193 2194
		return;

2195 2196 2197 2198 2199 2200
	/*
	 * 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);
2201 2202
}

2203 2204
static __always_inline
void account_cfs_rq_runtime(struct cfs_rq *cfs_rq, unsigned long delta_exec)
2205
{
2206
	if (!cfs_bandwidth_used() || !cfs_rq->runtime_enabled)
2207 2208 2209 2210 2211
		return;

	__account_cfs_rq_runtime(cfs_rq, delta_exec);
}

2212 2213
static inline int cfs_rq_throttled(struct cfs_rq *cfs_rq)
{
2214
	return cfs_bandwidth_used() && cfs_rq->throttled;
2215 2216
}

2217 2218 2219
/* check whether cfs_rq, or any parent, is throttled */
static inline int throttled_hierarchy(struct cfs_rq *cfs_rq)
{
2220
	return cfs_bandwidth_used() && cfs_rq->throttle_count;
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 2248
}

/*
 * 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) {
2249
		/* adjust cfs_rq_clock_task() */
2250
		cfs_rq->throttled_clock_task_time += rq_clock_task(rq) -
2251
					     cfs_rq->throttled_clock_task;
2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262
	}
#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)];

2263 2264
	/* group is entering throttled state, stop time */
	if (!cfs_rq->throttle_count)
2265
		cfs_rq->throttled_clock_task = rq_clock_task(rq);
2266 2267 2268 2269 2270
	cfs_rq->throttle_count++;

	return 0;
}

2271
static void throttle_cfs_rq(struct cfs_rq *cfs_rq)
2272 2273 2274 2275 2276 2277 2278 2279
{
	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))];

2280
	/* freeze hierarchy runnable averages while throttled */
2281 2282 2283
	rcu_read_lock();
	walk_tg_tree_from(cfs_rq->tg, tg_throttle_down, tg_nop, (void *)rq);
	rcu_read_unlock();
2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303

	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;
2304
	cfs_rq->throttled_clock = rq_clock(rq);
2305 2306 2307 2308 2309
	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);
}

2310
void unthrottle_cfs_rq(struct cfs_rq *cfs_rq)
2311 2312 2313 2314 2315 2316 2317 2318 2319 2320
{
	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;

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

	cfs_rq->throttled = 0;
2321 2322 2323

	update_rq_clock(rq);

2324
	raw_spin_lock(&cfs_b->lock);
2325
	cfs_b->throttled_time += rq_clock(rq) - cfs_rq->throttled_clock;
2326 2327 2328
	list_del_rcu(&cfs_rq->throttled_list);
	raw_spin_unlock(&cfs_b->lock);

2329 2330 2331
	/* update hierarchical throttle state */
	walk_tg_tree_from(cfs_rq->tg, tg_nop, tg_unthrottle_up, (void *)rq);

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 2394
	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;
}

2395 2396 2397 2398 2399 2400 2401 2402
/*
 * 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)
{
2403 2404
	u64 runtime, runtime_expires;
	int idle = 1, throttled;
2405 2406 2407 2408 2409 2410

	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;

2411 2412 2413
	throttled = !list_empty(&cfs_b->throttled_cfs_rq);
	/* idle depends on !throttled (for the case of a large deficit) */
	idle = cfs_b->idle && !throttled;
2414
	cfs_b->nr_periods += overrun;
2415

P
Paul Turner 已提交
2416 2417 2418 2419 2420 2421
	/* if we're going inactive then everything else can be deferred */
	if (idle)
		goto out_unlock;

	__refill_cfs_bandwidth_runtime(cfs_b);

2422 2423 2424 2425 2426 2427
	if (!throttled) {
		/* mark as potentially idle for the upcoming period */
		cfs_b->idle = 1;
		goto out_unlock;
	}

2428 2429 2430
	/* account preceding periods in which throttling occurred */
	cfs_b->nr_throttled += overrun;

2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454
	/*
	 * 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);
	}
2455

2456 2457 2458 2459 2460 2461 2462 2463 2464
	/* 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;
2465 2466 2467 2468 2469 2470 2471
out_unlock:
	if (idle)
		cfs_b->timer_active = 0;
	raw_spin_unlock(&cfs_b->lock);

	return idle;
}
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 2536
/* 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)
{
2537 2538 2539
	if (!cfs_bandwidth_used())
		return;

2540
	if (!cfs_rq->runtime_enabled || cfs_rq->nr_running)
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 2577
		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);
}

2578 2579 2580 2581 2582 2583 2584
/*
 * 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)
{
2585 2586 2587
	if (!cfs_bandwidth_used())
		return;

2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604
	/* 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)
{
2605 2606 2607
	if (!cfs_bandwidth_used())
		return;

2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619
	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);
}
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 2700 2701 2702 2703 2704

static inline u64 default_cfs_period(void);
static int do_sched_cfs_period_timer(struct cfs_bandwidth *cfs_b, int overrun);
static void do_sched_cfs_slack_timer(struct cfs_bandwidth *cfs_b);

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);
}

2705
static void __maybe_unused unthrottle_offline_cfs_rqs(struct rq *rq)
2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725
{
	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 */
2726 2727
static inline u64 cfs_rq_clock_task(struct cfs_rq *cfs_rq)
{
2728
	return rq_clock_task(rq_of(cfs_rq));
2729 2730 2731 2732
}

static void account_cfs_rq_runtime(struct cfs_rq *cfs_rq,
				     unsigned long delta_exec) {}
2733 2734
static void check_cfs_rq_runtime(struct cfs_rq *cfs_rq) {}
static void check_enqueue_throttle(struct cfs_rq *cfs_rq) {}
2735
static __always_inline void return_cfs_rq_runtime(struct cfs_rq *cfs_rq) {}
2736 2737 2738 2739 2740

static inline int cfs_rq_throttled(struct cfs_rq *cfs_rq)
{
	return 0;
}
2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751

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;
}
2752 2753 2754 2755 2756

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) {}
2757 2758
#endif

2759 2760 2761 2762 2763
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) {}
2764
static inline void unthrottle_offline_cfs_rqs(struct rq *rq) {}
2765 2766 2767

#endif /* CONFIG_CFS_BANDWIDTH */

2768 2769 2770 2771
/**************************************************
 * CFS operations on tasks:
 */

P
Peter Zijlstra 已提交
2772 2773 2774 2775 2776 2777 2778 2779
#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);

2780
	if (cfs_rq->nr_running > 1) {
P
Peter Zijlstra 已提交
2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794
		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.
		 */
2795
		if (rq->curr != p)
2796
			delta = max_t(s64, 10000LL, delta);
P
Peter Zijlstra 已提交
2797

2798
		hrtick_start(rq, delta);
P
Peter Zijlstra 已提交
2799 2800
	}
}
2801 2802 2803 2804 2805 2806 2807 2808 2809 2810

/*
 * 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;

2811
	if (!hrtick_enabled(rq) || curr->sched_class != &fair_sched_class)
2812 2813 2814 2815 2816
		return;

	if (cfs_rq_of(&curr->se)->nr_running < sched_nr_latency)
		hrtick_start_fair(rq, curr);
}
2817
#else /* !CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
2818 2819 2820 2821
static inline void
hrtick_start_fair(struct rq *rq, struct task_struct *p)
{
}
2822 2823 2824 2825

static inline void hrtick_update(struct rq *rq)
{
}
P
Peter Zijlstra 已提交
2826 2827
#endif

2828 2829 2830 2831 2832
/*
 * 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:
 */
2833
static void
2834
enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags)
2835 2836
{
	struct cfs_rq *cfs_rq;
2837
	struct sched_entity *se = &p->se;
2838 2839

	for_each_sched_entity(se) {
2840
		if (se->on_rq)
2841 2842
			break;
		cfs_rq = cfs_rq_of(se);
2843
		enqueue_entity(cfs_rq, se, flags);
2844 2845 2846 2847 2848 2849 2850 2851 2852

		/*
		 * 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;
2853
		cfs_rq->h_nr_running++;
2854

2855
		flags = ENQUEUE_WAKEUP;
2856
	}
P
Peter Zijlstra 已提交
2857

P
Peter Zijlstra 已提交
2858
	for_each_sched_entity(se) {
2859
		cfs_rq = cfs_rq_of(se);
2860
		cfs_rq->h_nr_running++;
P
Peter Zijlstra 已提交
2861

2862 2863 2864
		if (cfs_rq_throttled(cfs_rq))
			break;

2865
		update_cfs_shares(cfs_rq);
2866
		update_entity_load_avg(se, 1);
P
Peter Zijlstra 已提交
2867 2868
	}

2869 2870
	if (!se) {
		update_rq_runnable_avg(rq, rq->nr_running);
2871
		inc_nr_running(rq);
2872
	}
2873
	hrtick_update(rq);
2874 2875
}

2876 2877
static void set_next_buddy(struct sched_entity *se);

2878 2879 2880 2881 2882
/*
 * The dequeue_task method is called before nr_running is
 * decreased. We remove the task from the rbtree and
 * update the fair scheduling stats:
 */
2883
static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags)
2884 2885
{
	struct cfs_rq *cfs_rq;
2886
	struct sched_entity *se = &p->se;
2887
	int task_sleep = flags & DEQUEUE_SLEEP;
2888 2889 2890

	for_each_sched_entity(se) {
		cfs_rq = cfs_rq_of(se);
2891
		dequeue_entity(cfs_rq, se, flags);
2892 2893 2894 2895 2896 2897 2898 2899 2900

		/*
		 * 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;
2901
		cfs_rq->h_nr_running--;
P
Peter Zijlstra 已提交
2902

2903
		/* Don't dequeue parent if it has other entities besides us */
2904 2905 2906 2907 2908 2909 2910
		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));
2911 2912 2913

			/* avoid re-evaluating load for this entity */
			se = parent_entity(se);
2914
			break;
2915
		}
2916
		flags |= DEQUEUE_SLEEP;
2917
	}
P
Peter Zijlstra 已提交
2918

P
Peter Zijlstra 已提交
2919
	for_each_sched_entity(se) {
2920
		cfs_rq = cfs_rq_of(se);
2921
		cfs_rq->h_nr_running--;
P
Peter Zijlstra 已提交
2922

2923 2924 2925
		if (cfs_rq_throttled(cfs_rq))
			break;

2926
		update_cfs_shares(cfs_rq);
2927
		update_entity_load_avg(se, 1);
P
Peter Zijlstra 已提交
2928 2929
	}

2930
	if (!se) {
2931
		dec_nr_running(rq);
2932 2933
		update_rq_runnable_avg(rq, 1);
	}
2934
	hrtick_update(rq);
2935 2936
}

2937
#ifdef CONFIG_SMP
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 2988 2989 2990 2991 2992
/* 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;
}

2993

2994
static void task_waking_fair(struct task_struct *p)
2995 2996 2997
{
	struct sched_entity *se = &p->se;
	struct cfs_rq *cfs_rq = cfs_rq_of(se);
2998 2999 3000 3001
	u64 min_vruntime;

#ifndef CONFIG_64BIT
	u64 min_vruntime_copy;
3002

3003 3004 3005 3006 3007 3008 3009 3010
	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
3011

3012
	se->vruntime -= min_vruntime;
3013 3014
}

3015
#ifdef CONFIG_FAIR_GROUP_SCHED
3016 3017 3018 3019 3020 3021
/*
 * 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.
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 3060 3061 3062 3063 3064
 *
 * 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.
3065
 */
P
Peter Zijlstra 已提交
3066
static long effective_load(struct task_group *tg, int cpu, long wl, long wg)
3067
{
P
Peter Zijlstra 已提交
3068
	struct sched_entity *se = tg->se[cpu];
3069

3070
	if (!tg->parent)	/* the trivial, non-cgroup case */
3071 3072
		return wl;

P
Peter Zijlstra 已提交
3073
	for_each_sched_entity(se) {
3074
		long w, W;
P
Peter Zijlstra 已提交
3075

3076
		tg = se->my_q->tg;
3077

3078 3079 3080 3081
		/*
		 * W = @wg + \Sum rw_j
		 */
		W = wg + calc_tg_weight(tg, se->my_q);
P
Peter Zijlstra 已提交
3082

3083 3084 3085 3086
		/*
		 * w = rw_i + @wl
		 */
		w = se->my_q->load.weight + wl;
3087

3088 3089 3090 3091 3092
		/*
		 * wl = S * s'_i; see (2)
		 */
		if (W > 0 && w < W)
			wl = (w * tg->shares) / W;
3093 3094
		else
			wl = tg->shares;
3095

3096 3097 3098 3099 3100
		/*
		 * 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().
		 */
3101 3102
		if (wl < MIN_SHARES)
			wl = MIN_SHARES;
3103 3104 3105 3106

		/*
		 * wl = dw_i = S * (s'_i - s_i); see (3)
		 */
3107
		wl -= se->load.weight;
3108 3109 3110 3111 3112 3113 3114 3115

		/*
		 * 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.
		 */
P
Peter Zijlstra 已提交
3116 3117
		wg = 0;
	}
3118

P
Peter Zijlstra 已提交
3119
	return wl;
3120 3121
}
#else
P
Peter Zijlstra 已提交
3122

3123 3124
static inline unsigned long effective_load(struct task_group *tg, int cpu,
		unsigned long wl, unsigned long wg)
P
Peter Zijlstra 已提交
3125
{
3126
	return wl;
3127
}
P
Peter Zijlstra 已提交
3128

3129 3130
#endif

3131
static int wake_affine(struct sched_domain *sd, struct task_struct *p, int sync)
3132
{
3133
	s64 this_load, load;
3134
	int idx, this_cpu, prev_cpu;
3135
	unsigned long tl_per_task;
3136
	struct task_group *tg;
3137
	unsigned long weight;
3138
	int balanced;
3139

3140 3141 3142 3143 3144
	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);
3145

3146 3147 3148 3149 3150
	/*
	 * If sync wakeup then subtract the (maximum possible)
	 * effect of the currently running task from the load
	 * of the current CPU:
	 */
3151 3152 3153 3154
	if (sync) {
		tg = task_group(current);
		weight = current->se.load.weight;

3155
		this_load += effective_load(tg, this_cpu, -weight, -weight);
3156 3157
		load += effective_load(tg, prev_cpu, 0, -weight);
	}
3158

3159 3160
	tg = task_group(p);
	weight = p->se.load.weight;
3161

3162 3163
	/*
	 * In low-load situations, where prev_cpu is idle and this_cpu is idle
3164 3165 3166
	 * 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.
3167 3168 3169 3170
	 *
	 * Otherwise check if either cpus are near enough in load to allow this
	 * task to be woken on this_cpu.
	 */
3171 3172
	if (this_load > 0) {
		s64 this_eff_load, prev_eff_load;
3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185

		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;
3186

3187
	/*
I
Ingo Molnar 已提交
3188 3189 3190
	 * If the currently running task will sleep within
	 * a reasonable amount of time then attract this newly
	 * woken task:
3191
	 */
3192 3193
	if (sync && balanced)
		return 1;
3194

3195
	schedstat_inc(p, se.statistics.nr_wakeups_affine_attempts);
3196 3197
	tl_per_task = cpu_avg_load_per_task(this_cpu);

3198 3199 3200
	if (balanced ||
	    (this_load <= load &&
	     this_load + target_load(prev_cpu, idx) <= tl_per_task)) {
3201 3202 3203 3204 3205
		/*
		 * This domain has SD_WAKE_AFFINE and
		 * p is cache cold in this domain, and
		 * there is no bad imbalance.
		 */
3206
		schedstat_inc(sd, ttwu_move_affine);
3207
		schedstat_inc(p, se.statistics.nr_wakeups_affine);
3208 3209 3210 3211 3212 3213

		return 1;
	}
	return 0;
}

3214 3215 3216 3217 3218
/*
 * find_idlest_group finds and returns the least busy CPU group within the
 * domain.
 */
static struct sched_group *
P
Peter Zijlstra 已提交
3219
find_idlest_group(struct sched_domain *sd, struct task_struct *p,
3220
		  int this_cpu, int load_idx)
3221
{
3222
	struct sched_group *idlest = NULL, *group = sd->groups;
3223 3224
	unsigned long min_load = ULONG_MAX, this_load = 0;
	int imbalance = 100 + (sd->imbalance_pct-100)/2;
3225

3226 3227 3228 3229
	do {
		unsigned long load, avg_load;
		int local_group;
		int i;
3230

3231 3232
		/* Skip over this group if it has no CPUs allowed */
		if (!cpumask_intersects(sched_group_cpus(group),
3233
					tsk_cpus_allowed(p)))
3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252
			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 */
3253
		avg_load = (avg_load * SCHED_POWER_SCALE) / group->sgp->power;
3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278

		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 */
3279
	for_each_cpu_and(i, sched_group_cpus(group), tsk_cpus_allowed(p)) {
3280 3281 3282 3283 3284
		load = weighted_cpuload(i);

		if (load < min_load || (load == min_load && i == this_cpu)) {
			min_load = load;
			idlest = i;
3285 3286 3287
		}
	}

3288 3289
	return idlest;
}
3290

3291 3292 3293
/*
 * Try and locate an idle CPU in the sched_domain.
 */
3294
static int select_idle_sibling(struct task_struct *p, int target)
3295
{
3296
	struct sched_domain *sd;
3297
	struct sched_group *sg;
3298
	int i = task_cpu(p);
3299

3300 3301
	if (idle_cpu(target))
		return target;
3302 3303

	/*
3304
	 * If the prevous cpu is cache affine and idle, don't be stupid.
3305
	 */
3306 3307
	if (i != target && cpus_share_cache(i, target) && idle_cpu(i))
		return i;
3308 3309

	/*
3310
	 * Otherwise, iterate the domains and find an elegible idle cpu.
3311
	 */
3312
	sd = rcu_dereference(per_cpu(sd_llc, target));
3313
	for_each_lower_domain(sd) {
3314 3315 3316 3317 3318 3319 3320
		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)) {
3321
				if (i == target || !idle_cpu(i))
3322 3323
					goto next;
			}
3324

3325 3326 3327 3328 3329 3330 3331 3332
			target = cpumask_first_and(sched_group_cpus(sg),
					tsk_cpus_allowed(p));
			goto done;
next:
			sg = sg->next;
		} while (sg != sd->groups);
	}
done:
3333 3334 3335
	return target;
}

3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346
/*
 * 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.
 */
3347
static int
3348
select_task_rq_fair(struct task_struct *p, int sd_flag, int wake_flags)
3349
{
3350
	struct sched_domain *tmp, *affine_sd = NULL, *sd = NULL;
3351 3352 3353
	int cpu = smp_processor_id();
	int prev_cpu = task_cpu(p);
	int new_cpu = cpu;
3354
	int want_affine = 0;
3355
	int sync = wake_flags & WF_SYNC;
3356

3357
	if (p->nr_cpus_allowed == 1)
3358 3359
		return prev_cpu;

3360
	if (sd_flag & SD_BALANCE_WAKE) {
3361
		if (cpumask_test_cpu(cpu, tsk_cpus_allowed(p)))
3362 3363 3364
			want_affine = 1;
		new_cpu = prev_cpu;
	}
3365

3366
	rcu_read_lock();
3367
	for_each_domain(cpu, tmp) {
3368 3369 3370
		if (!(tmp->flags & SD_LOAD_BALANCE))
			continue;

3371
		/*
3372 3373
		 * If both cpu and prev_cpu are part of this domain,
		 * cpu is a valid SD_WAKE_AFFINE target.
3374
		 */
3375 3376 3377
		if (want_affine && (tmp->flags & SD_WAKE_AFFINE) &&
		    cpumask_test_cpu(prev_cpu, sched_domain_span(tmp))) {
			affine_sd = tmp;
3378
			break;
3379
		}
3380

3381
		if (tmp->flags & sd_flag)
3382 3383 3384
			sd = tmp;
	}

3385
	if (affine_sd) {
3386
		if (cpu != prev_cpu && wake_affine(affine_sd, p, sync))
3387 3388 3389 3390
			prev_cpu = cpu;

		new_cpu = select_idle_sibling(p, prev_cpu);
		goto unlock;
3391
	}
3392

3393
	while (sd) {
3394
		int load_idx = sd->forkexec_idx;
3395
		struct sched_group *group;
3396
		int weight;
3397

3398
		if (!(sd->flags & sd_flag)) {
3399 3400 3401
			sd = sd->child;
			continue;
		}
3402

3403 3404
		if (sd_flag & SD_BALANCE_WAKE)
			load_idx = sd->wake_idx;
3405

3406
		group = find_idlest_group(sd, p, cpu, load_idx);
3407 3408 3409 3410
		if (!group) {
			sd = sd->child;
			continue;
		}
I
Ingo Molnar 已提交
3411

3412
		new_cpu = find_idlest_cpu(group, p, cpu);
3413 3414 3415 3416
		if (new_cpu == -1 || new_cpu == cpu) {
			/* Now try balancing at a lower domain level of cpu */
			sd = sd->child;
			continue;
3417
		}
3418 3419 3420

		/* Now try balancing at a lower domain level of new_cpu */
		cpu = new_cpu;
3421
		weight = sd->span_weight;
3422 3423
		sd = NULL;
		for_each_domain(cpu, tmp) {
3424
			if (weight <= tmp->span_weight)
3425
				break;
3426
			if (tmp->flags & sd_flag)
3427 3428 3429
				sd = tmp;
		}
		/* while loop will break here if sd == NULL */
3430
	}
3431 3432
unlock:
	rcu_read_unlock();
3433

3434
	return new_cpu;
3435
}
3436

3437 3438 3439 3440 3441 3442
/*
 * Load-tracking only depends on SMP, FAIR_GROUP_SCHED dependency below may be
 * removed when useful for applications beyond shares distribution (e.g.
 * load-balance).
 */
#ifdef CONFIG_FAIR_GROUP_SCHED
3443 3444 3445 3446 3447 3448 3449 3450 3451
/*
 * 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)
{
3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464
	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);
	}
3465
}
3466
#endif
3467 3468
#endif /* CONFIG_SMP */

P
Peter Zijlstra 已提交
3469 3470
static unsigned long
wakeup_gran(struct sched_entity *curr, struct sched_entity *se)
3471 3472 3473 3474
{
	unsigned long gran = sysctl_sched_wakeup_granularity;

	/*
P
Peter Zijlstra 已提交
3475 3476
	 * Since its curr running now, convert the gran from real-time
	 * to virtual-time in his units.
M
Mike Galbraith 已提交
3477 3478 3479 3480 3481 3482 3483 3484 3485
	 *
	 * 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.
3486
	 */
3487
	return calc_delta_fair(gran, se);
3488 3489
}

3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511
/*
 * 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;

P
Peter Zijlstra 已提交
3512
	gran = wakeup_gran(curr, se);
3513 3514 3515 3516 3517 3518
	if (vdiff > gran)
		return 1;

	return 0;
}

3519 3520
static void set_last_buddy(struct sched_entity *se)
{
3521 3522 3523 3524 3525
	if (entity_is_task(se) && unlikely(task_of(se)->policy == SCHED_IDLE))
		return;

	for_each_sched_entity(se)
		cfs_rq_of(se)->last = se;
3526 3527 3528 3529
}

static void set_next_buddy(struct sched_entity *se)
{
3530 3531 3532 3533 3534
	if (entity_is_task(se) && unlikely(task_of(se)->policy == SCHED_IDLE))
		return;

	for_each_sched_entity(se)
		cfs_rq_of(se)->next = se;
3535 3536
}

3537 3538
static void set_skip_buddy(struct sched_entity *se)
{
3539 3540
	for_each_sched_entity(se)
		cfs_rq_of(se)->skip = se;
3541 3542
}

3543 3544 3545
/*
 * Preempt the current task with a newly woken task if needed:
 */
3546
static void check_preempt_wakeup(struct rq *rq, struct task_struct *p, int wake_flags)
3547 3548
{
	struct task_struct *curr = rq->curr;
3549
	struct sched_entity *se = &curr->se, *pse = &p->se;
3550
	struct cfs_rq *cfs_rq = task_cfs_rq(curr);
3551
	int scale = cfs_rq->nr_running >= sched_nr_latency;
3552
	int next_buddy_marked = 0;
3553

I
Ingo Molnar 已提交
3554 3555 3556
	if (unlikely(se == pse))
		return;

3557
	/*
3558
	 * This is possible from callers such as move_task(), in which we
3559 3560 3561 3562 3563 3564 3565
	 * 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;

3566
	if (sched_feat(NEXT_BUDDY) && scale && !(wake_flags & WF_FORK)) {
M
Mike Galbraith 已提交
3567
		set_next_buddy(pse);
3568 3569
		next_buddy_marked = 1;
	}
P
Peter Zijlstra 已提交
3570

3571 3572 3573
	/*
	 * We can come here with TIF_NEED_RESCHED already set from new task
	 * wake up path.
3574 3575 3576 3577 3578 3579
	 *
	 * 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.
3580 3581 3582 3583
	 */
	if (test_tsk_need_resched(curr))
		return;

3584 3585 3586 3587 3588
	/* Idle tasks are by definition preempted by non-idle tasks. */
	if (unlikely(curr->policy == SCHED_IDLE) &&
	    likely(p->policy != SCHED_IDLE))
		goto preempt;

3589
	/*
3590 3591
	 * Batch and idle tasks do not preempt non-idle tasks (their preemption
	 * is driven by the tick):
3592
	 */
3593
	if (unlikely(p->policy != SCHED_NORMAL) || !sched_feat(WAKEUP_PREEMPTION))
3594
		return;
3595

3596
	find_matching_se(&se, &pse);
3597
	update_curr(cfs_rq_of(se));
3598
	BUG_ON(!pse);
3599 3600 3601 3602 3603 3604 3605
	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);
3606
		goto preempt;
3607
	}
3608

3609
	return;
3610

3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626
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);
3627 3628
}

3629
static struct task_struct *pick_next_task_fair(struct rq *rq)
3630
{
P
Peter Zijlstra 已提交
3631
	struct task_struct *p;
3632 3633 3634
	struct cfs_rq *cfs_rq = &rq->cfs;
	struct sched_entity *se;

3635
	if (!cfs_rq->nr_running)
3636 3637 3638
		return NULL;

	do {
3639
		se = pick_next_entity(cfs_rq);
3640
		set_next_entity(cfs_rq, se);
3641 3642 3643
		cfs_rq = group_cfs_rq(se);
	} while (cfs_rq);

P
Peter Zijlstra 已提交
3644
	p = task_of(se);
3645 3646
	if (hrtick_enabled(rq))
		hrtick_start_fair(rq, p);
P
Peter Zijlstra 已提交
3647 3648

	return p;
3649 3650 3651 3652 3653
}

/*
 * Account for a descheduled task:
 */
3654
static void put_prev_task_fair(struct rq *rq, struct task_struct *prev)
3655 3656 3657 3658 3659 3660
{
	struct sched_entity *se = &prev->se;
	struct cfs_rq *cfs_rq;

	for_each_sched_entity(se) {
		cfs_rq = cfs_rq_of(se);
3661
		put_prev_entity(cfs_rq, se);
3662 3663 3664
	}
}

3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689
/*
 * 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);
3690 3691 3692 3693 3694 3695
		/*
		 * 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;
3696 3697 3698 3699 3700
	}

	set_skip_buddy(se);
}

3701 3702 3703 3704
static bool yield_to_task_fair(struct rq *rq, struct task_struct *p, bool preempt)
{
	struct sched_entity *se = &p->se;

3705 3706
	/* throttled hierarchies are not runnable */
	if (!se->on_rq || throttled_hierarchy(cfs_rq_of(se)))
3707 3708 3709 3710 3711 3712 3713 3714 3715 3716
		return false;

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

	yield_task_fair(rq);

	return true;
}

3717
#ifdef CONFIG_SMP
3718
/**************************************************
P
Peter Zijlstra 已提交
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 3823 3824 3825 3826 3827 3828 3829 3830 3831 3832 3833 3834
 * 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.]
 */ 
3835

3836 3837
static unsigned long __read_mostly max_load_balance_interval = HZ/10;

3838
#define LBF_ALL_PINNED	0x01
3839
#define LBF_NEED_BREAK	0x02
3840
#define LBF_SOME_PINNED 0x04
3841 3842 3843 3844 3845

struct lb_env {
	struct sched_domain	*sd;

	struct rq		*src_rq;
3846
	int			src_cpu;
3847 3848 3849 3850

	int			dst_cpu;
	struct rq		*dst_rq;

3851 3852
	struct cpumask		*dst_grpmask;
	int			new_dst_cpu;
3853
	enum cpu_idle_type	idle;
3854
	long			imbalance;
3855 3856 3857
	/* The set of CPUs under consideration for load-balancing */
	struct cpumask		*cpus;

3858
	unsigned int		flags;
3859 3860 3861 3862

	unsigned int		loop;
	unsigned int		loop_break;
	unsigned int		loop_max;
3863 3864
};

3865
/*
3866
 * move_task - move a task from one runqueue to another runqueue.
3867 3868
 * Both runqueues must be locked.
 */
3869
static void move_task(struct task_struct *p, struct lb_env *env)
3870
{
3871 3872 3873 3874
	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);
3875 3876
}

3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908
/*
 * 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;
}

3909 3910 3911 3912
/*
 * can_migrate_task - may task p from runqueue rq be migrated to this_cpu?
 */
static
3913
int can_migrate_task(struct task_struct *p, struct lb_env *env)
3914 3915 3916 3917
{
	int tsk_cache_hot = 0;
	/*
	 * We do not migrate tasks that are:
3918
	 * 1) throttled_lb_pair, or
3919
	 * 2) cannot be migrated to this CPU due to cpus_allowed, or
3920 3921
	 * 3) running (obviously), or
	 * 4) are cache-hot on their current CPU.
3922
	 */
3923 3924 3925
	if (throttled_lb_pair(task_group(p), env->src_cpu, env->dst_cpu))
		return 0;

3926
	if (!cpumask_test_cpu(env->dst_cpu, tsk_cpus_allowed(p))) {
3927
		int cpu;
3928

3929
		schedstat_inc(p, se.statistics.nr_failed_migrations_affine);
3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941

		/*
		 * 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;

3942 3943 3944 3945 3946 3947 3948
		/* 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;
			}
3949
		}
3950

3951 3952
		return 0;
	}
3953 3954

	/* Record that we found atleast one task that could run on dst_cpu */
3955
	env->flags &= ~LBF_ALL_PINNED;
3956

3957
	if (task_running(env->src_rq, p)) {
3958
		schedstat_inc(p, se.statistics.nr_failed_migrations_running);
3959 3960 3961 3962 3963 3964 3965 3966 3967
		return 0;
	}

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

3968
	tsk_cache_hot = task_hot(p, rq_clock_task(env->src_rq), env->sd);
3969
	if (!tsk_cache_hot ||
3970
		env->sd->nr_balance_failed > env->sd->cache_nice_tries) {
Z
Zhang Hang 已提交
3971

3972
		if (tsk_cache_hot) {
3973
			schedstat_inc(env->sd, lb_hot_gained[env->idle]);
3974
			schedstat_inc(p, se.statistics.nr_forced_migrations);
3975
		}
Z
Zhang Hang 已提交
3976

3977 3978 3979
		return 1;
	}

Z
Zhang Hang 已提交
3980 3981
	schedstat_inc(p, se.statistics.nr_failed_migrations_hot);
	return 0;
3982 3983
}

3984 3985 3986 3987 3988 3989 3990
/*
 * 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.
 */
3991
static int move_one_task(struct lb_env *env)
3992 3993 3994
{
	struct task_struct *p, *n;

3995 3996 3997
	list_for_each_entry_safe(p, n, &env->src_rq->cfs_tasks, se.group_node) {
		if (!can_migrate_task(p, env))
			continue;
3998

3999 4000 4001 4002 4003 4004 4005 4006
		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;
4007 4008 4009 4010
	}
	return 0;
}

4011 4012
static unsigned long task_h_load(struct task_struct *p);

4013 4014
static const unsigned int sched_nr_migrate_break = 32;

4015
/*
4016
 * move_tasks tries to move up to imbalance weighted load from busiest to
4017 4018 4019 4020 4021 4022
 * 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)
4023
{
4024 4025
	struct list_head *tasks = &env->src_rq->cfs_tasks;
	struct task_struct *p;
4026 4027
	unsigned long load;
	int pulled = 0;
4028

4029
	if (env->imbalance <= 0)
4030
		return 0;
4031

4032 4033
	while (!list_empty(tasks)) {
		p = list_first_entry(tasks, struct task_struct, se.group_node);
4034

4035 4036
		env->loop++;
		/* We've more or less seen every task there is, call it quits */
4037
		if (env->loop > env->loop_max)
4038
			break;
4039 4040

		/* take a breather every nr_migrate tasks */
4041
		if (env->loop > env->loop_break) {
4042
			env->loop_break += sched_nr_migrate_break;
4043
			env->flags |= LBF_NEED_BREAK;
4044
			break;
4045
		}
4046

4047
		if (!can_migrate_task(p, env))
4048 4049 4050
			goto next;

		load = task_h_load(p);
4051

4052
		if (sched_feat(LB_MIN) && load < 16 && !env->sd->nr_balance_failed)
4053 4054
			goto next;

4055
		if ((load / 2) > env->imbalance)
4056
			goto next;
4057

4058
		move_task(p, env);
4059
		pulled++;
4060
		env->imbalance -= load;
4061 4062

#ifdef CONFIG_PREEMPT
4063 4064 4065 4066 4067
		/*
		 * NEWIDLE balancing is a source of latency, so preemptible
		 * kernels will stop after the first task is pulled to minimize
		 * the critical section.
		 */
4068
		if (env->idle == CPU_NEWLY_IDLE)
4069
			break;
4070 4071
#endif

4072 4073 4074 4075
		/*
		 * We only want to steal up to the prescribed amount of
		 * weighted load.
		 */
4076
		if (env->imbalance <= 0)
4077
			break;
4078 4079 4080

		continue;
next:
4081
		list_move_tail(&p->se.group_node, tasks);
4082
	}
4083

4084
	/*
4085 4086 4087
	 * 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().
4088
	 */
4089
	schedstat_add(env->sd, lb_gained[env->idle], pulled);
4090

4091
	return pulled;
4092 4093
}

P
Peter Zijlstra 已提交
4094
#ifdef CONFIG_FAIR_GROUP_SCHED
4095 4096 4097
/*
 * update tg->load_weight by folding this cpu's load_avg
 */
4098
static void __update_blocked_averages_cpu(struct task_group *tg, int cpu)
4099
{
4100 4101
	struct sched_entity *se = tg->se[cpu];
	struct cfs_rq *cfs_rq = tg->cfs_rq[cpu];
4102

4103 4104 4105
	/* throttled entities do not contribute to load */
	if (throttled_hierarchy(cfs_rq))
		return;
4106

4107
	update_cfs_rq_blocked_load(cfs_rq, 1);
4108

4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122
	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 {
4123
		struct rq *rq = rq_of(cfs_rq);
4124 4125
		update_rq_runnable_avg(rq, rq->nr_running);
	}
4126 4127
}

4128
static void update_blocked_averages(int cpu)
4129 4130
{
	struct rq *rq = cpu_rq(cpu);
4131 4132
	struct cfs_rq *cfs_rq;
	unsigned long flags;
4133

4134 4135
	raw_spin_lock_irqsave(&rq->lock, flags);
	update_rq_clock(rq);
4136 4137 4138 4139
	/*
	 * Iterates the task_group tree in a bottom up fashion, see
	 * list_add_leaf_cfs_rq() for details.
	 */
4140
	for_each_leaf_cfs_rq(rq, cfs_rq) {
4141 4142 4143 4144 4145 4146
		/*
		 * 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);
4147
	}
4148 4149

	raw_spin_unlock_irqrestore(&rq->lock, flags);
4150 4151
}

4152 4153 4154 4155 4156 4157 4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174 4175 4176
/*
 * 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)
{
4177 4178 4179 4180 4181 4182 4183 4184
	struct rq *rq = cpu_rq(cpu);
	unsigned long now = jiffies;

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

	rq->h_load_throttle = now;

4185
	rcu_read_lock();
4186
	walk_tg_tree(tg_load_down, tg_nop, (void *)cpu);
4187
	rcu_read_unlock();
4188 4189
}

4190
static unsigned long task_h_load(struct task_struct *p)
P
Peter Zijlstra 已提交
4191
{
4192 4193
	struct cfs_rq *cfs_rq = task_cfs_rq(p);
	unsigned long load;
P
Peter Zijlstra 已提交
4194

4195 4196
	load = p->se.load.weight;
	load = div_u64(load * cfs_rq->h_load, cfs_rq->load.weight + 1);
P
Peter Zijlstra 已提交
4197

4198
	return load;
P
Peter Zijlstra 已提交
4199 4200
}
#else
4201
static inline void update_blocked_averages(int cpu)
4202 4203 4204
{
}

4205
static inline void update_h_load(long cpu)
P
Peter Zijlstra 已提交
4206 4207 4208
{
}

4209
static unsigned long task_h_load(struct task_struct *p)
4210
{
4211
	return p->se.load.weight;
4212
}
P
Peter Zijlstra 已提交
4213
#endif
4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230

/********** 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;
4231
	unsigned long this_has_capacity;
4232
	unsigned int  this_idle_cpus;
4233 4234

	/* Statistics of the busiest group */
4235
	unsigned int  busiest_idle_cpus;
4236 4237 4238
	unsigned long max_load;
	unsigned long busiest_load_per_task;
	unsigned long busiest_nr_running;
4239
	unsigned long busiest_group_capacity;
4240
	unsigned long busiest_has_capacity;
4241
	unsigned int  busiest_group_weight;
4242 4243 4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254

	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;
4255 4256
	unsigned long idle_cpus;
	unsigned long group_weight;
4257
	int group_imb; /* Is there an imbalance in the group ? */
4258
	int group_has_capacity; /* Is there extra capacity in the group? */
4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286
};

/**
 * 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;
}

4287
static unsigned long default_scale_freq_power(struct sched_domain *sd, int cpu)
4288
{
4289
	return SCHED_POWER_SCALE;
4290 4291 4292 4293 4294 4295 4296
}

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

4297
static unsigned long default_scale_smt_power(struct sched_domain *sd, int cpu)
4298
{
4299
	unsigned long weight = sd->span_weight;
4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311
	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);
}

4312
static unsigned long scale_rt_power(int cpu)
4313 4314
{
	struct rq *rq = cpu_rq(cpu);
4315
	u64 total, available, age_stamp, avg;
4316

4317 4318 4319 4320 4321 4322 4323
	/*
	 * 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);

4324
	total = sched_avg_period() + (rq_clock(rq) - age_stamp);
4325

4326
	if (unlikely(total < avg)) {
4327 4328 4329
		/* Ensures that power won't end up being negative */
		available = 0;
	} else {
4330
		available = total - avg;
4331
	}
4332

4333 4334
	if (unlikely((s64)total < SCHED_POWER_SCALE))
		total = SCHED_POWER_SCALE;
4335

4336
	total >>= SCHED_POWER_SHIFT;
4337 4338 4339 4340 4341 4342

	return div_u64(available, total);
}

static void update_cpu_power(struct sched_domain *sd, int cpu)
{
4343
	unsigned long weight = sd->span_weight;
4344
	unsigned long power = SCHED_POWER_SCALE;
4345 4346 4347 4348 4349 4350 4351 4352
	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);

4353
		power >>= SCHED_POWER_SHIFT;
4354 4355
	}

4356
	sdg->sgp->power_orig = power;
4357 4358 4359 4360 4361 4362

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

4363
	power >>= SCHED_POWER_SHIFT;
4364

4365
	power *= scale_rt_power(cpu);
4366
	power >>= SCHED_POWER_SHIFT;
4367 4368 4369 4370

	if (!power)
		power = 1;

4371
	cpu_rq(cpu)->cpu_power = power;
4372
	sdg->sgp->power = power;
4373 4374
}

4375
void update_group_power(struct sched_domain *sd, int cpu)
4376 4377 4378 4379
{
	struct sched_domain *child = sd->child;
	struct sched_group *group, *sdg = sd->groups;
	unsigned long power;
4380 4381 4382 4383 4384
	unsigned long interval;

	interval = msecs_to_jiffies(sd->balance_interval);
	interval = clamp(interval, 1UL, max_load_balance_interval);
	sdg->sgp->next_update = jiffies + interval;
4385 4386 4387 4388 4389 4390 4391 4392

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

	power = 0;

P
Peter Zijlstra 已提交
4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404 4405 4406 4407 4408 4409 4410 4411 4412
	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);
	}
4413

4414
	sdg->sgp->power_orig = sdg->sgp->power = power;
4415 4416
}

4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427
/*
 * 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)
{
	/*
4428
	 * Only siblings can have significantly less than SCHED_POWER_SCALE
4429
	 */
P
Peter Zijlstra 已提交
4430
	if (!(sd->flags & SD_SHARE_CPUPOWER))
4431 4432 4433 4434 4435
		return 0;

	/*
	 * If ~90% of the cpu_power is still there, we're good.
	 */
4436
	if (group->sgp->power * 32 > group->sgp->power_orig * 29)
4437 4438 4439 4440 4441
		return 1;

	return 0;
}

4442 4443
/**
 * update_sg_lb_stats - Update sched_group's statistics for load balancing.
4444
 * @env: The load balancing environment.
4445 4446 4447 4448 4449 4450
 * @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.
 */
4451 4452
static inline void update_sg_lb_stats(struct lb_env *env,
			struct sched_group *group, int load_idx,
4453
			int local_group, int *balance, struct sg_lb_stats *sgs)
4454
{
4455 4456
	unsigned long nr_running, max_nr_running, min_nr_running;
	unsigned long load, max_cpu_load, min_cpu_load;
4457
	unsigned int balance_cpu = -1, first_idle_cpu = 0;
4458
	unsigned long avg_load_per_task = 0;
4459
	int i;
4460

4461
	if (local_group)
P
Peter Zijlstra 已提交
4462
		balance_cpu = group_balance_cpu(group);
4463 4464 4465 4466

	/* Tally up the load of all CPUs in the group */
	max_cpu_load = 0;
	min_cpu_load = ~0UL;
4467
	max_nr_running = 0;
4468
	min_nr_running = ~0UL;
4469

4470
	for_each_cpu_and(i, sched_group_cpus(group), env->cpus) {
4471 4472
		struct rq *rq = cpu_rq(i);

4473 4474
		nr_running = rq->nr_running;

4475 4476
		/* Bias balancing toward cpus of our domain */
		if (local_group) {
P
Peter Zijlstra 已提交
4477 4478
			if (idle_cpu(i) && !first_idle_cpu &&
					cpumask_test_cpu(i, sched_group_mask(group))) {
4479
				first_idle_cpu = 1;
4480 4481
				balance_cpu = i;
			}
4482 4483

			load = target_load(i, load_idx);
4484 4485
		} else {
			load = source_load(i, load_idx);
4486
			if (load > max_cpu_load)
4487 4488 4489
				max_cpu_load = load;
			if (min_cpu_load > load)
				min_cpu_load = load;
4490 4491 4492 4493 4494

			if (nr_running > max_nr_running)
				max_nr_running = nr_running;
			if (min_nr_running > nr_running)
				min_nr_running = nr_running;
4495 4496 4497
		}

		sgs->group_load += load;
4498
		sgs->sum_nr_running += nr_running;
4499
		sgs->sum_weighted_load += weighted_cpuload(i);
4500 4501
		if (idle_cpu(i))
			sgs->idle_cpus++;
4502 4503 4504 4505 4506 4507 4508 4509
	}

	/*
	 * 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.
	 */
4510
	if (local_group) {
4511
		if (env->idle != CPU_NEWLY_IDLE) {
4512
			if (balance_cpu != env->dst_cpu) {
4513 4514 4515
				*balance = 0;
				return;
			}
4516
			update_group_power(env->sd, env->dst_cpu);
4517
		} else if (time_after_eq(jiffies, group->sgp->next_update))
4518
			update_group_power(env->sd, env->dst_cpu);
4519 4520 4521
	}

	/* Adjust by relative CPU power of the group */
4522
	sgs->avg_load = (sgs->group_load*SCHED_POWER_SCALE) / group->sgp->power;
4523 4524 4525

	/*
	 * Consider the group unbalanced when the imbalance is larger
P
Peter Zijlstra 已提交
4526
	 * than the average weight of a task.
4527 4528 4529 4530 4531 4532
	 *
	 * 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?
	 */
4533 4534
	if (sgs->sum_nr_running)
		avg_load_per_task = sgs->sum_weighted_load / sgs->sum_nr_running;
4535

4536 4537
	if ((max_cpu_load - min_cpu_load) >= avg_load_per_task &&
	    (max_nr_running - min_nr_running) > 1)
4538 4539
		sgs->group_imb = 1;

4540
	sgs->group_capacity = DIV_ROUND_CLOSEST(group->sgp->power,
4541
						SCHED_POWER_SCALE);
4542
	if (!sgs->group_capacity)
4543
		sgs->group_capacity = fix_small_capacity(env->sd, group);
4544
	sgs->group_weight = group->group_weight;
4545 4546 4547

	if (sgs->group_capacity > sgs->sum_nr_running)
		sgs->group_has_capacity = 1;
4548 4549
}

4550 4551
/**
 * update_sd_pick_busiest - return 1 on busiest group
4552
 * @env: The load balancing environment.
4553 4554
 * @sds: sched_domain statistics
 * @sg: sched_group candidate to be checked for being the busiest
4555
 * @sgs: sched_group statistics
4556 4557 4558 4559
 *
 * Determine if @sg is a busier group than the previously selected
 * busiest group.
 */
4560
static bool update_sd_pick_busiest(struct lb_env *env,
4561 4562
				   struct sd_lb_stats *sds,
				   struct sched_group *sg,
4563
				   struct sg_lb_stats *sgs)
4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578
{
	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.
	 */
4579 4580
	if ((env->sd->flags & SD_ASYM_PACKING) && sgs->sum_nr_running &&
	    env->dst_cpu < group_first_cpu(sg)) {
4581 4582 4583 4584 4585 4586 4587 4588 4589 4590
		if (!sds->busiest)
			return true;

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

	return false;
}

4591
/**
4592
 * update_sd_lb_stats - Update sched_domain's statistics for load balancing.
4593
 * @env: The load balancing environment.
4594 4595 4596
 * @balance: Should we balance.
 * @sds: variable to hold the statistics for this sched_domain.
 */
4597
static inline void update_sd_lb_stats(struct lb_env *env,
4598
					int *balance, struct sd_lb_stats *sds)
4599
{
4600 4601
	struct sched_domain *child = env->sd->child;
	struct sched_group *sg = env->sd->groups;
4602 4603 4604 4605 4606 4607
	struct sg_lb_stats sgs;
	int load_idx, prefer_sibling = 0;

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

4608
	load_idx = get_sd_load_idx(env->sd, env->idle);
4609 4610 4611 4612

	do {
		int local_group;

4613
		local_group = cpumask_test_cpu(env->dst_cpu, sched_group_cpus(sg));
4614
		memset(&sgs, 0, sizeof(sgs));
4615
		update_sg_lb_stats(env, sg, load_idx, local_group, balance, &sgs);
4616

P
Peter Zijlstra 已提交
4617
		if (local_group && !(*balance))
4618 4619 4620
			return;

		sds->total_load += sgs.group_load;
4621
		sds->total_pwr += sg->sgp->power;
4622 4623 4624

		/*
		 * In case the child domain prefers tasks go to siblings
4625
		 * first, lower the sg capacity to one so that we'll try
4626 4627 4628 4629 4630 4631
		 * 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).
4632
		 */
4633
		if (prefer_sibling && !local_group && sds->this_has_capacity)
4634 4635 4636 4637
			sgs.group_capacity = min(sgs.group_capacity, 1UL);

		if (local_group) {
			sds->this_load = sgs.avg_load;
4638
			sds->this = sg;
4639 4640
			sds->this_nr_running = sgs.sum_nr_running;
			sds->this_load_per_task = sgs.sum_weighted_load;
4641
			sds->this_has_capacity = sgs.group_has_capacity;
4642
			sds->this_idle_cpus = sgs.idle_cpus;
4643
		} else if (update_sd_pick_busiest(env, sds, sg, &sgs)) {
4644
			sds->max_load = sgs.avg_load;
4645
			sds->busiest = sg;
4646
			sds->busiest_nr_running = sgs.sum_nr_running;
4647
			sds->busiest_idle_cpus = sgs.idle_cpus;
4648
			sds->busiest_group_capacity = sgs.group_capacity;
4649
			sds->busiest_load_per_task = sgs.sum_weighted_load;
4650
			sds->busiest_has_capacity = sgs.group_has_capacity;
4651
			sds->busiest_group_weight = sgs.group_weight;
4652 4653 4654
			sds->group_imb = sgs.group_imb;
		}

4655
		sg = sg->next;
4656
	} while (sg != env->sd->groups);
4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673 4674 4675
}

/**
 * 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.
 *
4676 4677 4678
 * Returns 1 when packing is required and a task should be moved to
 * this CPU.  The amount of the imbalance is returned in *imbalance.
 *
4679
 * @env: The load balancing environment.
4680 4681
 * @sds: Statistics of the sched_domain which is to be packed
 */
4682
static int check_asym_packing(struct lb_env *env, struct sd_lb_stats *sds)
4683 4684 4685
{
	int busiest_cpu;

4686
	if (!(env->sd->flags & SD_ASYM_PACKING))
4687 4688 4689 4690 4691 4692
		return 0;

	if (!sds->busiest)
		return 0;

	busiest_cpu = group_first_cpu(sds->busiest);
4693
	if (env->dst_cpu > busiest_cpu)
4694 4695
		return 0;

4696 4697 4698
	env->imbalance = DIV_ROUND_CLOSEST(
		sds->max_load * sds->busiest->sgp->power, SCHED_POWER_SCALE);

4699
	return 1;
4700 4701 4702 4703 4704 4705
}

/**
 * fix_small_imbalance - Calculate the minor imbalance that exists
 *			amongst the groups of a sched_domain, during
 *			load balancing.
4706
 * @env: The load balancing environment.
4707 4708
 * @sds: Statistics of the sched_domain whose imbalance is to be calculated.
 */
4709 4710
static inline
void fix_small_imbalance(struct lb_env *env, struct sd_lb_stats *sds)
4711 4712 4713
{
	unsigned long tmp, pwr_now = 0, pwr_move = 0;
	unsigned int imbn = 2;
4714
	unsigned long scaled_busy_load_per_task;
4715 4716 4717 4718 4719 4720

	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;
4721
	} else {
4722
		sds->this_load_per_task =
4723 4724
			cpu_avg_load_per_task(env->dst_cpu);
	}
4725

4726
	scaled_busy_load_per_task = sds->busiest_load_per_task
4727
					 * SCHED_POWER_SCALE;
4728
	scaled_busy_load_per_task /= sds->busiest->sgp->power;
4729 4730 4731

	if (sds->max_load - sds->this_load + scaled_busy_load_per_task >=
			(scaled_busy_load_per_task * imbn)) {
4732
		env->imbalance = sds->busiest_load_per_task;
4733 4734 4735 4736 4737 4738 4739 4740 4741
		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.
	 */

4742
	pwr_now += sds->busiest->sgp->power *
4743
			min(sds->busiest_load_per_task, sds->max_load);
4744
	pwr_now += sds->this->sgp->power *
4745
			min(sds->this_load_per_task, sds->this_load);
4746
	pwr_now /= SCHED_POWER_SCALE;
4747 4748

	/* Amount of load we'd subtract */
4749
	tmp = (sds->busiest_load_per_task * SCHED_POWER_SCALE) /
4750
		sds->busiest->sgp->power;
4751
	if (sds->max_load > tmp)
4752
		pwr_move += sds->busiest->sgp->power *
4753 4754 4755
			min(sds->busiest_load_per_task, sds->max_load - tmp);

	/* Amount of load we'd add */
4756
	if (sds->max_load * sds->busiest->sgp->power <
4757
		sds->busiest_load_per_task * SCHED_POWER_SCALE)
4758 4759
		tmp = (sds->max_load * sds->busiest->sgp->power) /
			sds->this->sgp->power;
4760
	else
4761
		tmp = (sds->busiest_load_per_task * SCHED_POWER_SCALE) /
4762 4763
			sds->this->sgp->power;
	pwr_move += sds->this->sgp->power *
4764
			min(sds->this_load_per_task, sds->this_load + tmp);
4765
	pwr_move /= SCHED_POWER_SCALE;
4766 4767 4768

	/* Move if we gain throughput */
	if (pwr_move > pwr_now)
4769
		env->imbalance = sds->busiest_load_per_task;
4770 4771 4772 4773 4774
}

/**
 * calculate_imbalance - Calculate the amount of imbalance present within the
 *			 groups of a given sched_domain during load balance.
4775
 * @env: load balance environment
4776 4777
 * @sds: statistics of the sched_domain whose imbalance is to be calculated.
 */
4778
static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *sds)
4779
{
4780 4781 4782 4783 4784 4785 4786 4787
	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);
	}

4788 4789 4790 4791 4792 4793
	/*
	 * 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) {
4794 4795
		env->imbalance = 0;
		return fix_small_imbalance(env, sds);
4796 4797
	}

4798 4799 4800 4801 4802 4803 4804
	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);

4805
		load_above_capacity *= (SCHED_LOAD_SCALE * SCHED_POWER_SCALE);
4806

4807
		load_above_capacity /= sds->busiest->sgp->power;
4808 4809 4810 4811 4812 4813 4814 4815 4816 4817 4818 4819 4820
	}

	/*
	 * 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);
4821 4822

	/* How much load to actually move to equalise the imbalance */
4823
	env->imbalance = min(max_pull * sds->busiest->sgp->power,
4824
		(sds->avg_load - sds->this_load) * sds->this->sgp->power)
4825
			/ SCHED_POWER_SCALE;
4826 4827 4828

	/*
	 * if *imbalance is less than the average load per runnable task
L
Lucas De Marchi 已提交
4829
	 * there is no guarantee that any tasks will be moved so we'll have
4830 4831 4832
	 * a think about bumping its value to force at least one task to be
	 * moved
	 */
4833 4834
	if (env->imbalance < sds->busiest_load_per_task)
		return fix_small_imbalance(env, sds);
4835 4836

}
4837

4838 4839 4840 4841 4842 4843 4844 4845 4846 4847 4848 4849
/******* 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.
 *
4850
 * @env: The load balancing environment.
4851 4852 4853 4854 4855 4856 4857 4858 4859
 * @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 *
4860
find_busiest_group(struct lb_env *env, int *balance)
4861 4862 4863 4864 4865 4866 4867 4868 4869
{
	struct sd_lb_stats sds;

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

	/*
	 * Compute the various statistics relavent for load balancing at
	 * this level.
	 */
4870
	update_sd_lb_stats(env, balance, &sds);
4871

4872 4873 4874
	/*
	 * this_cpu is not the appropriate cpu to perform load balancing at
	 * this level.
4875
	 */
P
Peter Zijlstra 已提交
4876
	if (!(*balance))
4877 4878
		goto ret;

4879 4880
	if ((env->idle == CPU_IDLE || env->idle == CPU_NEWLY_IDLE) &&
	    check_asym_packing(env, &sds))
4881 4882
		return sds.busiest;

4883
	/* There is no busy sibling group to pull tasks from */
4884 4885 4886
	if (!sds.busiest || sds.busiest_nr_running == 0)
		goto out_balanced;

4887
	sds.avg_load = (SCHED_POWER_SCALE * sds.total_load) / sds.total_pwr;
4888

P
Peter Zijlstra 已提交
4889 4890 4891 4892 4893 4894 4895 4896
	/*
	 * 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;

4897
	/* SD_BALANCE_NEWIDLE trumps SMP nice when underutilized */
4898
	if (env->idle == CPU_NEWLY_IDLE && sds.this_has_capacity &&
4899 4900 4901
			!sds.busiest_has_capacity)
		goto force_balance;

4902 4903 4904 4905
	/*
	 * If the local group is more busy than the selected busiest group
	 * don't try and pull any tasks.
	 */
4906 4907 4908
	if (sds.this_load >= sds.max_load)
		goto out_balanced;

4909 4910 4911 4912
	/*
	 * Don't pull any tasks if this group is already above the domain
	 * average load.
	 */
4913 4914 4915
	if (sds.this_load >= sds.avg_load)
		goto out_balanced;

4916
	if (env->idle == CPU_IDLE) {
4917 4918 4919 4920 4921 4922
		/*
		 * 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.
		 */
4923
		if ((sds.this_idle_cpus <= sds.busiest_idle_cpus + 1) &&
4924 4925
		    sds.busiest_nr_running <= sds.busiest_group_weight)
			goto out_balanced;
4926 4927 4928 4929 4930
	} else {
		/*
		 * In the CPU_NEWLY_IDLE, CPU_NOT_IDLE cases, use
		 * imbalance_pct to be conservative.
		 */
4931
		if (100 * sds.max_load <= env->sd->imbalance_pct * sds.this_load)
4932
			goto out_balanced;
4933
	}
4934

4935
force_balance:
4936
	/* Looks like there is an imbalance. Compute it */
4937
	calculate_imbalance(env, &sds);
4938 4939 4940 4941
	return sds.busiest;

out_balanced:
ret:
4942
	env->imbalance = 0;
4943 4944 4945 4946 4947 4948
	return NULL;
}

/*
 * find_busiest_queue - find the busiest runqueue among the cpus in group.
 */
4949
static struct rq *find_busiest_queue(struct lb_env *env,
4950
				     struct sched_group *group)
4951 4952 4953 4954 4955 4956 4957
{
	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);
4958 4959
		unsigned long capacity = DIV_ROUND_CLOSEST(power,
							   SCHED_POWER_SCALE);
4960 4961
		unsigned long wl;

4962
		if (!capacity)
4963
			capacity = fix_small_capacity(env->sd, group);
4964

4965
		if (!cpumask_test_cpu(i, env->cpus))
4966 4967 4968
			continue;

		rq = cpu_rq(i);
4969
		wl = weighted_cpuload(i);
4970

4971 4972 4973 4974
		/*
		 * When comparing with imbalance, use weighted_cpuload()
		 * which is not scaled with the cpu power.
		 */
4975
		if (capacity && rq->nr_running == 1 && wl > env->imbalance)
4976 4977
			continue;

4978 4979 4980 4981 4982 4983
		/*
		 * 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.
		 */
4984
		wl = (wl * SCHED_POWER_SCALE) / power;
4985

4986 4987 4988 4989 4990 4991 4992 4993 4994 4995 4996 4997 4998 4999 5000 5001
		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. */
5002
DEFINE_PER_CPU(cpumask_var_t, load_balance_mask);
5003

5004
static int need_active_balance(struct lb_env *env)
5005
{
5006 5007 5008
	struct sched_domain *sd = env->sd;

	if (env->idle == CPU_NEWLY_IDLE) {
5009 5010 5011 5012 5013 5014

		/*
		 * ASYM_PACKING needs to force migrate tasks from busy but
		 * higher numbered CPUs in order to pack all tasks in the
		 * lowest numbered CPUs.
		 */
5015
		if ((sd->flags & SD_ASYM_PACKING) && env->src_cpu > env->dst_cpu)
5016
			return 1;
5017 5018 5019 5020 5021
	}

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

5022 5023
static int active_load_balance_cpu_stop(void *data);

5024 5025 5026 5027 5028 5029 5030 5031
/*
 * 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)
{
5032
	int ld_moved, cur_ld_moved, active_balance = 0;
5033 5034 5035
	struct sched_group *group;
	struct rq *busiest;
	unsigned long flags;
5036
	struct cpumask *cpus = __get_cpu_var(load_balance_mask);
5037

5038 5039
	struct lb_env env = {
		.sd		= sd,
5040 5041
		.dst_cpu	= this_cpu,
		.dst_rq		= this_rq,
5042
		.dst_grpmask    = sched_group_cpus(sd->groups),
5043
		.idle		= idle,
5044
		.loop_break	= sched_nr_migrate_break,
5045
		.cpus		= cpus,
5046 5047
	};

5048 5049 5050 5051
	/*
	 * For NEWLY_IDLE load_balancing, we don't need to consider
	 * other cpus in our group
	 */
5052
	if (idle == CPU_NEWLY_IDLE)
5053 5054
		env.dst_grpmask = NULL;

5055 5056 5057 5058 5059
	cpumask_copy(cpus, cpu_active_mask);

	schedstat_inc(sd, lb_count[idle]);

redo:
5060
	group = find_busiest_group(&env, balance);
5061 5062 5063 5064 5065 5066 5067 5068 5069

	if (*balance == 0)
		goto out_balanced;

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

5070
	busiest = find_busiest_queue(&env, group);
5071 5072 5073 5074 5075
	if (!busiest) {
		schedstat_inc(sd, lb_nobusyq[idle]);
		goto out_balanced;
	}

5076
	BUG_ON(busiest == env.dst_rq);
5077

5078
	schedstat_add(sd, lb_imbalance[idle], env.imbalance);
5079 5080 5081 5082 5083 5084 5085 5086 5087

	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.
		 */
5088
		env.flags |= LBF_ALL_PINNED;
5089 5090 5091
		env.src_cpu   = busiest->cpu;
		env.src_rq    = busiest;
		env.loop_max  = min(sysctl_sched_nr_migrate, busiest->nr_running);
5092

5093
		update_h_load(env.src_cpu);
5094
more_balance:
5095
		local_irq_save(flags);
5096
		double_rq_lock(env.dst_rq, busiest);
5097 5098 5099 5100 5101 5102 5103

		/*
		 * 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;
5104
		double_rq_unlock(env.dst_rq, busiest);
5105 5106 5107 5108 5109
		local_irq_restore(flags);

		/*
		 * some other cpu did the load balance for us.
		 */
5110 5111 5112
		if (cur_ld_moved && env.dst_cpu != smp_processor_id())
			resched_cpu(env.dst_cpu);

5113 5114 5115 5116 5117
		if (env.flags & LBF_NEED_BREAK) {
			env.flags &= ~LBF_NEED_BREAK;
			goto more_balance;
		}

5118 5119 5120 5121 5122 5123 5124 5125 5126 5127 5128 5129 5130 5131 5132 5133 5134 5135 5136
		/*
		 * 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.
		 */
5137
		if ((env.flags & LBF_SOME_PINNED) && env.imbalance > 0) {
5138

5139
			env.dst_rq	 = cpu_rq(env.new_dst_cpu);
5140 5141 5142 5143
			env.dst_cpu	 = env.new_dst_cpu;
			env.flags	&= ~LBF_SOME_PINNED;
			env.loop	 = 0;
			env.loop_break	 = sched_nr_migrate_break;
5144 5145 5146 5147

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

5148 5149 5150 5151 5152 5153
			/*
			 * Go back to "more_balance" rather than "redo" since we
			 * need to continue with same src_cpu.
			 */
			goto more_balance;
		}
5154 5155

		/* All tasks on this runqueue were pinned by CPU affinity */
5156
		if (unlikely(env.flags & LBF_ALL_PINNED)) {
5157
			cpumask_clear_cpu(cpu_of(busiest), cpus);
5158 5159 5160
			if (!cpumask_empty(cpus)) {
				env.loop = 0;
				env.loop_break = sched_nr_migrate_break;
5161
				goto redo;
5162
			}
5163 5164 5165 5166 5167 5168
			goto out_balanced;
		}
	}

	if (!ld_moved) {
		schedstat_inc(sd, lb_failed[idle]);
5169 5170 5171 5172 5173 5174 5175 5176
		/*
		 * 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++;
5177

5178
		if (need_active_balance(&env)) {
5179 5180
			raw_spin_lock_irqsave(&busiest->lock, flags);

5181 5182 5183
			/* don't kick the active_load_balance_cpu_stop,
			 * if the curr task on busiest cpu can't be
			 * moved to this_cpu
5184 5185
			 */
			if (!cpumask_test_cpu(this_cpu,
5186
					tsk_cpus_allowed(busiest->curr))) {
5187 5188
				raw_spin_unlock_irqrestore(&busiest->lock,
							    flags);
5189
				env.flags |= LBF_ALL_PINNED;
5190 5191 5192
				goto out_one_pinned;
			}

5193 5194 5195 5196 5197
			/*
			 * ->active_balance synchronizes accesses to
			 * ->active_balance_work.  Once set, it's cleared
			 * only after active load balance is finished.
			 */
5198 5199 5200 5201 5202 5203
			if (!busiest->active_balance) {
				busiest->active_balance = 1;
				busiest->push_cpu = this_cpu;
				active_balance = 1;
			}
			raw_spin_unlock_irqrestore(&busiest->lock, flags);
5204

5205
			if (active_balance) {
5206 5207 5208
				stop_one_cpu_nowait(cpu_of(busiest),
					active_load_balance_cpu_stop, busiest,
					&busiest->active_balance_work);
5209
			}
5210 5211 5212 5213 5214 5215 5216 5217 5218 5219 5220 5221 5222 5223 5224 5225 5226 5227 5228 5229 5230 5231 5232 5233 5234 5235 5236 5237 5238 5239 5240 5241 5242

			/*
			 * 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 */
5243
	if (((env.flags & LBF_ALL_PINNED) &&
5244
			sd->balance_interval < MAX_PINNED_INTERVAL) ||
5245 5246 5247
			(sd->balance_interval < sd->max_interval))
		sd->balance_interval *= 2;

5248
	ld_moved = 0;
5249 5250 5251 5252 5253 5254 5255 5256
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.
 */
5257
void idle_balance(int this_cpu, struct rq *this_rq)
5258 5259 5260 5261 5262
{
	struct sched_domain *sd;
	int pulled_task = 0;
	unsigned long next_balance = jiffies + HZ;

5263
	this_rq->idle_stamp = rq_clock(this_rq);
5264 5265 5266 5267

	if (this_rq->avg_idle < sysctl_sched_migration_cost)
		return;

5268 5269 5270 5271 5272
	/*
	 * Drop the rq->lock, but keep IRQ/preempt disabled.
	 */
	raw_spin_unlock(&this_rq->lock);

5273
	update_blocked_averages(this_cpu);
5274
	rcu_read_lock();
5275 5276
	for_each_domain(this_cpu, sd) {
		unsigned long interval;
5277
		int balance = 1;
5278 5279 5280 5281

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

5282
		if (sd->flags & SD_BALANCE_NEWIDLE) {
5283
			/* If we've pulled tasks over stop searching: */
5284 5285 5286
			pulled_task = load_balance(this_cpu, this_rq,
						   sd, CPU_NEWLY_IDLE, &balance);
		}
5287 5288 5289 5290

		interval = msecs_to_jiffies(sd->balance_interval);
		if (time_after(next_balance, sd->last_balance + interval))
			next_balance = sd->last_balance + interval;
N
Nikhil Rao 已提交
5291 5292
		if (pulled_task) {
			this_rq->idle_stamp = 0;
5293
			break;
N
Nikhil Rao 已提交
5294
		}
5295
	}
5296
	rcu_read_unlock();
5297 5298 5299

	raw_spin_lock(&this_rq->lock);

5300 5301 5302 5303 5304 5305 5306 5307 5308 5309
	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;
	}
}

/*
5310 5311 5312 5313
 * 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.
5314
 */
5315
static int active_load_balance_cpu_stop(void *data)
5316
{
5317 5318
	struct rq *busiest_rq = data;
	int busiest_cpu = cpu_of(busiest_rq);
5319
	int target_cpu = busiest_rq->push_cpu;
5320
	struct rq *target_rq = cpu_rq(target_cpu);
5321
	struct sched_domain *sd;
5322 5323 5324 5325 5326 5327 5328

	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;
5329 5330 5331

	/* Is there any task to move? */
	if (busiest_rq->nr_running <= 1)
5332
		goto out_unlock;
5333 5334 5335 5336 5337 5338 5339 5340 5341 5342 5343 5344

	/*
	 * 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. */
5345
	rcu_read_lock();
5346 5347 5348 5349 5350 5351 5352
	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)) {
5353 5354
		struct lb_env env = {
			.sd		= sd,
5355 5356 5357 5358
			.dst_cpu	= target_cpu,
			.dst_rq		= target_rq,
			.src_cpu	= busiest_rq->cpu,
			.src_rq		= busiest_rq,
5359 5360 5361
			.idle		= CPU_IDLE,
		};

5362 5363
		schedstat_inc(sd, alb_count);

5364
		if (move_one_task(&env))
5365 5366 5367 5368
			schedstat_inc(sd, alb_pushed);
		else
			schedstat_inc(sd, alb_failed);
	}
5369
	rcu_read_unlock();
5370
	double_unlock_balance(busiest_rq, target_rq);
5371 5372 5373 5374
out_unlock:
	busiest_rq->active_balance = 0;
	raw_spin_unlock_irq(&busiest_rq->lock);
	return 0;
5375 5376
}

5377
#ifdef CONFIG_NO_HZ_COMMON
5378 5379 5380 5381 5382 5383
/*
 * 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.
 */
5384
static struct {
5385
	cpumask_var_t idle_cpus_mask;
5386
	atomic_t nr_cpus;
5387 5388
	unsigned long next_balance;     /* in jiffy units */
} nohz ____cacheline_aligned;
5389

5390
static inline int find_new_ilb(int call_cpu)
5391
{
5392
	int ilb = cpumask_first(nohz.idle_cpus_mask);
5393

5394 5395 5396 5397
	if (ilb < nr_cpu_ids && idle_cpu(ilb))
		return ilb;

	return nr_cpu_ids;
5398 5399
}

5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410
/*
 * 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++;

5411
	ilb_cpu = find_new_ilb(cpu);
5412

5413 5414
	if (ilb_cpu >= nr_cpu_ids)
		return;
5415

5416
	if (test_and_set_bit(NOHZ_BALANCE_KICK, nohz_flags(ilb_cpu)))
5417 5418 5419 5420 5421 5422 5423 5424
		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);
5425 5426 5427
	return;
}

5428
static inline void nohz_balance_exit_idle(int cpu)
5429 5430 5431 5432 5433 5434 5435 5436
{
	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));
	}
}

5437 5438 5439 5440 5441
static inline void set_cpu_sd_state_busy(void)
{
	struct sched_domain *sd;

	rcu_read_lock();
N
Nathan Zimmer 已提交
5442
	sd = rcu_dereference_check_sched_domain(this_rq()->sd);
V
Vincent Guittot 已提交
5443 5444 5445 5446 5447 5448

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

	for (; sd; sd = sd->parent)
5449
		atomic_inc(&sd->groups->sgp->nr_busy_cpus);
V
Vincent Guittot 已提交
5450
unlock:
5451 5452 5453 5454 5455 5456 5457 5458
	rcu_read_unlock();
}

void set_cpu_sd_state_idle(void)
{
	struct sched_domain *sd;

	rcu_read_lock();
N
Nathan Zimmer 已提交
5459
	sd = rcu_dereference_check_sched_domain(this_rq()->sd);
V
Vincent Guittot 已提交
5460 5461 5462 5463 5464 5465

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

	for (; sd; sd = sd->parent)
5466
		atomic_dec(&sd->groups->sgp->nr_busy_cpus);
V
Vincent Guittot 已提交
5467
unlock:
5468 5469 5470
	rcu_read_unlock();
}

5471
/*
5472
 * This routine will record that the cpu is going idle with tick stopped.
5473
 * This info will be used in performing idle load balancing in the future.
5474
 */
5475
void nohz_balance_enter_idle(int cpu)
5476
{
5477 5478 5479 5480 5481 5482
	/*
	 * If this cpu is going down, then nothing needs to be done.
	 */
	if (!cpu_active(cpu))
		return;

5483 5484
	if (test_bit(NOHZ_TICK_STOPPED, nohz_flags(cpu)))
		return;
5485

5486 5487 5488
	cpumask_set_cpu(cpu, nohz.idle_cpus_mask);
	atomic_inc(&nohz.nr_cpus);
	set_bit(NOHZ_TICK_STOPPED, nohz_flags(cpu));
5489
}
5490 5491 5492 5493 5494 5495

static int __cpuinit sched_ilb_notifier(struct notifier_block *nfb,
					unsigned long action, void *hcpu)
{
	switch (action & ~CPU_TASKS_FROZEN) {
	case CPU_DYING:
5496
		nohz_balance_exit_idle(smp_processor_id());
5497 5498 5499 5500 5501
		return NOTIFY_OK;
	default:
		return NOTIFY_DONE;
	}
}
5502 5503 5504 5505
#endif

static DEFINE_SPINLOCK(balancing);

5506 5507 5508 5509
/*
 * 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.
 */
5510
void update_max_interval(void)
5511 5512 5513 5514
{
	max_load_balance_interval = HZ*num_online_cpus()/10;
}

5515 5516 5517 5518
/*
 * It checks each scheduling domain to see if it is due to be balanced,
 * and initiates a balancing operation if so.
 *
5519
 * Balancing parameters are set up in init_sched_domains.
5520 5521 5522 5523 5524 5525
 */
static void rebalance_domains(int cpu, enum cpu_idle_type idle)
{
	int balance = 1;
	struct rq *rq = cpu_rq(cpu);
	unsigned long interval;
5526
	struct sched_domain *sd;
5527 5528 5529 5530 5531
	/* Earliest time when we have to do rebalance again */
	unsigned long next_balance = jiffies + 60*HZ;
	int update_next_balance = 0;
	int need_serialize;

5532
	update_blocked_averages(cpu);
P
Peter Zijlstra 已提交
5533

5534
	rcu_read_lock();
5535 5536 5537 5538 5539 5540 5541 5542 5543 5544
	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);
5545
		interval = clamp(interval, 1UL, max_load_balance_interval);
5546 5547 5548 5549 5550 5551 5552 5553 5554 5555 5556

		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)) {
				/*
5557 5558 5559
				 * 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.
5560
				 */
5561
				idle = idle_cpu(cpu) ? CPU_IDLE : CPU_NOT_IDLE;
5562 5563 5564 5565 5566 5567 5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578 5579 5580
			}
			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;
	}
5581
	rcu_read_unlock();
5582 5583 5584 5585 5586 5587 5588 5589 5590 5591

	/*
	 * 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;
}

5592
#ifdef CONFIG_NO_HZ_COMMON
5593
/*
5594
 * In CONFIG_NO_HZ_COMMON case, the idle balance kickee will do the
5595 5596
 * rebalancing for all the cpus for whom scheduler ticks are stopped.
 */
5597 5598 5599 5600 5601 5602
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;

5603 5604 5605
	if (idle != CPU_IDLE ||
	    !test_bit(NOHZ_BALANCE_KICK, nohz_flags(this_cpu)))
		goto end;
5606 5607

	for_each_cpu(balance_cpu, nohz.idle_cpus_mask) {
5608
		if (balance_cpu == this_cpu || !idle_cpu(balance_cpu))
5609 5610 5611 5612 5613 5614 5615
			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.
		 */
5616
		if (need_resched())
5617 5618
			break;

V
Vincent Guittot 已提交
5619 5620 5621 5622 5623 5624
		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);
5625 5626 5627 5628 5629 5630 5631

		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;
5632 5633
end:
	clear_bit(NOHZ_BALANCE_KICK, nohz_flags(this_cpu));
5634 5635 5636
}

/*
5637 5638 5639 5640 5641 5642 5643
 * 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.
5644 5645 5646 5647
 */
static inline int nohz_kick_needed(struct rq *rq, int cpu)
{
	unsigned long now = jiffies;
5648
	struct sched_domain *sd;
5649

5650
	if (unlikely(idle_cpu(cpu)))
5651 5652
		return 0;

5653 5654 5655 5656
       /*
	* 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.
	*/
5657
	set_cpu_sd_state_busy();
5658
	nohz_balance_exit_idle(cpu);
5659 5660 5661 5662 5663 5664 5665

	/*
	 * None are in tickless mode and hence no need for NOHZ idle load
	 * balancing.
	 */
	if (likely(!atomic_read(&nohz.nr_cpus)))
		return 0;
5666 5667

	if (time_before(now, nohz.next_balance))
5668 5669
		return 0;

5670 5671
	if (rq->nr_running >= 2)
		goto need_kick;
5672

5673
	rcu_read_lock();
5674 5675 5676 5677
	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);
5678

5679
		if (sd->flags & SD_SHARE_PKG_RESOURCES && nr_busy > 1)
5680
			goto need_kick_unlock;
5681 5682 5683 5684

		if (sd->flags & SD_ASYM_PACKING && nr_busy != sg->group_weight
		    && (cpumask_first_and(nohz.idle_cpus_mask,
					  sched_domain_span(sd)) < cpu))
5685
			goto need_kick_unlock;
5686 5687 5688

		if (!(sd->flags & (SD_SHARE_PKG_RESOURCES | SD_ASYM_PACKING)))
			break;
5689
	}
5690
	rcu_read_unlock();
5691
	return 0;
5692 5693 5694

need_kick_unlock:
	rcu_read_unlock();
5695 5696
need_kick:
	return 1;
5697 5698 5699 5700 5701 5702 5703 5704 5705
}
#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).
 */
5706 5707 5708 5709
static void run_rebalance_domains(struct softirq_action *h)
{
	int this_cpu = smp_processor_id();
	struct rq *this_rq = cpu_rq(this_cpu);
5710
	enum cpu_idle_type idle = this_rq->idle_balance ?
5711 5712 5713 5714 5715
						CPU_IDLE : CPU_NOT_IDLE;

	rebalance_domains(this_cpu, idle);

	/*
5716
	 * If this cpu has a pending nohz_balance_kick, then do the
5717 5718 5719
	 * balancing on behalf of the other idle cpus whose ticks are
	 * stopped.
	 */
5720
	nohz_idle_balance(this_cpu, idle);
5721 5722 5723 5724
}

static inline int on_null_domain(int cpu)
{
5725
	return !rcu_dereference_sched(cpu_rq(cpu)->sd);
5726 5727 5728 5729 5730
}

/*
 * Trigger the SCHED_SOFTIRQ if it is time to do periodic load balancing.
 */
5731
void trigger_load_balance(struct rq *rq, int cpu)
5732 5733 5734 5735 5736
{
	/* 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);
5737
#ifdef CONFIG_NO_HZ_COMMON
5738
	if (nohz_kick_needed(rq, cpu) && likely(!on_null_domain(cpu)))
5739 5740
		nohz_balancer_kick(cpu);
#endif
5741 5742
}

5743 5744 5745 5746 5747 5748 5749 5750
static void rq_online_fair(struct rq *rq)
{
	update_sysctl();
}

static void rq_offline_fair(struct rq *rq)
{
	update_sysctl();
5751 5752 5753

	/* Ensure any throttled groups are reachable by pick_next_task */
	unthrottle_offline_cfs_rqs(rq);
5754 5755
}

5756
#endif /* CONFIG_SMP */
5757

5758 5759 5760
/*
 * scheduler tick hitting a task of our scheduling class:
 */
P
Peter Zijlstra 已提交
5761
static void task_tick_fair(struct rq *rq, struct task_struct *curr, int queued)
5762 5763 5764 5765 5766 5767
{
	struct cfs_rq *cfs_rq;
	struct sched_entity *se = &curr->se;

	for_each_sched_entity(se) {
		cfs_rq = cfs_rq_of(se);
P
Peter Zijlstra 已提交
5768
		entity_tick(cfs_rq, se, queued);
5769
	}
5770

5771 5772
	if (sched_feat_numa(NUMA))
		task_tick_numa(rq, curr);
5773

5774
	update_rq_runnable_avg(rq, 1);
5775 5776 5777
}

/*
P
Peter Zijlstra 已提交
5778 5779 5780
 * called on fork with the child task as argument from the parent's context
 *  - child not yet on the tasklist
 *  - preemption disabled
5781
 */
P
Peter Zijlstra 已提交
5782
static void task_fork_fair(struct task_struct *p)
5783
{
5784 5785
	struct cfs_rq *cfs_rq;
	struct sched_entity *se = &p->se, *curr;
5786
	int this_cpu = smp_processor_id();
P
Peter Zijlstra 已提交
5787 5788 5789
	struct rq *rq = this_rq();
	unsigned long flags;

5790
	raw_spin_lock_irqsave(&rq->lock, flags);
5791

5792 5793
	update_rq_clock(rq);

5794 5795 5796
	cfs_rq = task_cfs_rq(current);
	curr = cfs_rq->curr;

5797 5798
	if (unlikely(task_cpu(p) != this_cpu)) {
		rcu_read_lock();
P
Peter Zijlstra 已提交
5799
		__set_task_cpu(p, this_cpu);
5800 5801
		rcu_read_unlock();
	}
5802

5803
	update_curr(cfs_rq);
P
Peter Zijlstra 已提交
5804

5805 5806
	if (curr)
		se->vruntime = curr->vruntime;
5807
	place_entity(cfs_rq, se, 1);
5808

P
Peter Zijlstra 已提交
5809
	if (sysctl_sched_child_runs_first && curr && entity_before(curr, se)) {
D
Dmitry Adamushko 已提交
5810
		/*
5811 5812 5813
		 * Upon rescheduling, sched_class::put_prev_task() will place
		 * 'current' within the tree based on its new key value.
		 */
5814
		swap(curr->vruntime, se->vruntime);
5815
		resched_task(rq->curr);
5816
	}
5817

5818 5819
	se->vruntime -= cfs_rq->min_vruntime;

5820
	raw_spin_unlock_irqrestore(&rq->lock, flags);
5821 5822
}

5823 5824 5825 5826
/*
 * Priority of the task has changed. Check to see if we preempt
 * the current task.
 */
P
Peter Zijlstra 已提交
5827 5828
static void
prio_changed_fair(struct rq *rq, struct task_struct *p, int oldprio)
5829
{
P
Peter Zijlstra 已提交
5830 5831 5832
	if (!p->se.on_rq)
		return;

5833 5834 5835 5836 5837
	/*
	 * 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
	 */
P
Peter Zijlstra 已提交
5838
	if (rq->curr == p) {
5839 5840 5841
		if (p->prio > oldprio)
			resched_task(rq->curr);
	} else
5842
		check_preempt_curr(rq, p, 0);
5843 5844
}

P
Peter Zijlstra 已提交
5845 5846 5847 5848 5849 5850 5851 5852 5853 5854 5855 5856 5857 5858 5859 5860 5861 5862 5863 5864 5865 5866
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;
	}
5867 5868 5869 5870 5871 5872 5873 5874 5875 5876 5877 5878 5879 5880

#if defined(CONFIG_FAIR_GROUP_SCHED) && defined(CONFIG_SMP)
	/*
	* 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
P
Peter Zijlstra 已提交
5881 5882
}

5883 5884 5885
/*
 * We switched to the sched_fair class.
 */
P
Peter Zijlstra 已提交
5886
static void switched_to_fair(struct rq *rq, struct task_struct *p)
5887
{
P
Peter Zijlstra 已提交
5888 5889 5890
	if (!p->se.on_rq)
		return;

5891 5892 5893 5894 5895
	/*
	 * 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.
	 */
P
Peter Zijlstra 已提交
5896
	if (rq->curr == p)
5897 5898
		resched_task(rq->curr);
	else
5899
		check_preempt_curr(rq, p, 0);
5900 5901
}

5902 5903 5904 5905 5906 5907 5908 5909 5910
/* 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;

5911 5912 5913 5914 5915 5916 5917
	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);
	}
5918 5919
}

5920 5921 5922 5923 5924 5925 5926
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
5927 5928
#if defined(CONFIG_FAIR_GROUP_SCHED) && defined(CONFIG_SMP)
	atomic64_set(&cfs_rq->decay_counter, 1);
5929
	atomic64_set(&cfs_rq->removed_load, 0);
5930
#endif
5931 5932
}

P
Peter Zijlstra 已提交
5933
#ifdef CONFIG_FAIR_GROUP_SCHED
5934
static void task_move_group_fair(struct task_struct *p, int on_rq)
P
Peter Zijlstra 已提交
5935
{
5936
	struct cfs_rq *cfs_rq;
5937 5938 5939 5940 5941 5942 5943 5944 5945 5946 5947 5948 5949
	/*
	 * 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.
	 */
5950 5951 5952 5953 5954 5955
	/*
	 * 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().
5956 5957
	 * - Moving a task which has been woken up by try_to_wake_up() and
	 *   waiting for actually being woken up by sched_ttwu_pending().
5958 5959 5960 5961
	 *
	 * To prevent boost or penalty in the new cfs_rq caused by delta
	 * min_vruntime between the two cfs_rqs, we skip vruntime adjustment.
	 */
5962
	if (!on_rq && (!p->se.sum_exec_runtime || p->state == TASK_WAKING))
5963 5964
		on_rq = 1;

5965 5966 5967
	if (!on_rq)
		p->se.vruntime -= cfs_rq_of(&p->se)->min_vruntime;
	set_task_rq(p, task_cpu(p));
5968 5969 5970 5971 5972 5973 5974 5975 5976 5977 5978 5979 5980
	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
	}
P
Peter Zijlstra 已提交
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 6099 6100 6101 6102 6103 6104 6105 6106 6107 6108 6109 6110

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);
6111 6112 6113

		/* Possible calls to update_curr() need rq clock */
		update_rq_clock(rq);
6114
		for_each_sched_entity(se)
6115 6116 6117 6118 6119 6120 6121 6122 6123 6124 6125 6126 6127 6128 6129 6130 6131 6132 6133 6134 6135
			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 */

P
Peter Zijlstra 已提交
6136

6137
static unsigned int get_rr_interval_fair(struct rq *rq, struct task_struct *task)
6138 6139 6140 6141 6142 6143 6144 6145 6146
{
	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)
6147
		rr_interval = NS_TO_JIFFIES(sched_slice(cfs_rq_of(se), se));
6148 6149 6150 6151

	return rr_interval;
}

6152 6153 6154
/*
 * All the scheduling class methods:
 */
6155
const struct sched_class fair_sched_class = {
6156
	.next			= &idle_sched_class,
6157 6158 6159
	.enqueue_task		= enqueue_task_fair,
	.dequeue_task		= dequeue_task_fair,
	.yield_task		= yield_task_fair,
6160
	.yield_to_task		= yield_to_task_fair,
6161

I
Ingo Molnar 已提交
6162
	.check_preempt_curr	= check_preempt_wakeup,
6163 6164 6165 6166

	.pick_next_task		= pick_next_task_fair,
	.put_prev_task		= put_prev_task_fair,

6167
#ifdef CONFIG_SMP
L
Li Zefan 已提交
6168
	.select_task_rq		= select_task_rq_fair,
6169
#ifdef CONFIG_FAIR_GROUP_SCHED
6170
	.migrate_task_rq	= migrate_task_rq_fair,
6171
#endif
6172 6173
	.rq_online		= rq_online_fair,
	.rq_offline		= rq_offline_fair,
6174 6175

	.task_waking		= task_waking_fair,
6176
#endif
6177

6178
	.set_curr_task          = set_curr_task_fair,
6179
	.task_tick		= task_tick_fair,
P
Peter Zijlstra 已提交
6180
	.task_fork		= task_fork_fair,
6181 6182

	.prio_changed		= prio_changed_fair,
P
Peter Zijlstra 已提交
6183
	.switched_from		= switched_from_fair,
6184
	.switched_to		= switched_to_fair,
P
Peter Zijlstra 已提交
6185

6186 6187
	.get_rr_interval	= get_rr_interval_fair,

P
Peter Zijlstra 已提交
6188
#ifdef CONFIG_FAIR_GROUP_SCHED
6189
	.task_move_group	= task_move_group_fair,
P
Peter Zijlstra 已提交
6190
#endif
6191 6192 6193
};

#ifdef CONFIG_SCHED_DEBUG
6194
void print_cfs_stats(struct seq_file *m, int cpu)
6195 6196 6197
{
	struct cfs_rq *cfs_rq;

6198
	rcu_read_lock();
6199
	for_each_leaf_cfs_rq(cpu_rq(cpu), cfs_rq)
6200
		print_cfs_rq(m, cpu, cfs_rq);
6201
	rcu_read_unlock();
6202 6203
}
#endif
6204 6205 6206 6207 6208 6209

__init void init_sched_fair_class(void)
{
#ifdef CONFIG_SMP
	open_softirq(SCHED_SOFTIRQ, run_rebalance_domains);

6210
#ifdef CONFIG_NO_HZ_COMMON
6211
	nohz.next_balance = jiffies;
6212
	zalloc_cpumask_var(&nohz.idle_cpus_mask, GFP_NOWAIT);
6213
	cpu_notifier(sched_ilb_notifier, 0);
6214 6215 6216 6217
#endif
#endif /* SMP */

}