fair.c 133.2 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 29 30 31 32
#include <linux/slab.h>
#include <linux/profile.h>
#include <linux/interrupt.h>

#include <trace/events/sched.h>

#include "sched.h"
A
Arjan van de Ven 已提交
33

34
/*
35
 * Targeted preemption latency for CPU-bound tasks:
36
 * (default: 6ms * (1 + ilog(ncpus)), units: nanoseconds)
37
 *
38
 * NOTE: this latency value is not the same as the concept of
I
Ingo Molnar 已提交
39 40 41
 * 'timeslice length' - timeslices in CFS are of variable length
 * and have no persistent notion like in traditional, time-slice
 * based scheduling concepts.
42
 *
I
Ingo Molnar 已提交
43 44
 * (to see the precise effective timeslice length of your workload,
 *  run vmstat and monitor the context-switches (cs) field)
45
 */
46 47
unsigned int sysctl_sched_latency = 6000000ULL;
unsigned int normalized_sysctl_sched_latency = 6000000ULL;
48

49 50 51 52 53 54 55 56 57 58 59 60
/*
 * 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;

61
/*
62
 * Minimal preemption granularity for CPU-bound tasks:
63
 * (default: 0.75 msec * (1 + ilog(ncpus)), units: nanoseconds)
64
 */
65 66
unsigned int sysctl_sched_min_granularity = 750000ULL;
unsigned int normalized_sysctl_sched_min_granularity = 750000ULL;
67 68

/*
69 70
 * is kept at sysctl_sched_latency / sysctl_sched_min_granularity
 */
71
static unsigned int sched_nr_latency = 8;
72 73

/*
74
 * After fork, child runs first. If set to 0 (default) then
75
 * parent will (try to) run first.
76
 */
77
unsigned int sysctl_sched_child_runs_first __read_mostly;
78 79 80

/*
 * SCHED_OTHER wake-up granularity.
81
 * (default: 1 msec * (1 + ilog(ncpus)), units: nanoseconds)
82 83 84 85 86
 *
 * This option delays the preemption effects of decoupled workloads
 * and reduces their over-scheduling. Synchronous workloads will still
 * have immediate wakeup/sleep latencies.
 */
87
unsigned int sysctl_sched_wakeup_granularity = 1000000UL;
88
unsigned int normalized_sysctl_sched_wakeup_granularity = 1000000UL;
89

90 91
const_debug unsigned int sysctl_sched_migration_cost = 500000UL;

92 93 94 95 96 97 98
/*
 * The exponential sliding  window over which load is averaged for shares
 * distribution.
 * (default: 10msec)
 */
unsigned int __read_mostly sysctl_sched_shares_window = 10000000UL;

99 100 101 102 103 104 105 106 107 108 109 110 111 112
#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

113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216
/*
 * 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;
217

218 219 220 221
/**************************************************************
 * CFS operations on generic schedulable entities:
 */

222
#ifdef CONFIG_FAIR_GROUP_SCHED
223

224
/* cpu runqueue to which this cfs_rq is attached */
225 226
static inline struct rq *rq_of(struct cfs_rq *cfs_rq)
{
227
	return cfs_rq->rq;
228 229
}

230 231
/* An entity is a task if it doesn't "own" a runqueue */
#define entity_is_task(se)	(!se->my_q)
232

233 234 235 236 237 238 239 240
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
241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261
/* 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;
}

262 263 264
static inline void list_add_leaf_cfs_rq(struct cfs_rq *cfs_rq)
{
	if (!cfs_rq->on_list) {
265 266 267 268 269 270 271 272 273 274 275 276
		/*
		 * 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,
277
				&rq_of(cfs_rq)->leaf_cfs_rq_list);
278
		}
279 280 281 282 283 284 285 286 287 288 289 290 291

		cfs_rq->on_list = 1;
	}
}

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
292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310
/* 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;
}

311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353
/* 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);
	}
}

354 355 356 357 358 359
#else	/* !CONFIG_FAIR_GROUP_SCHED */

static inline struct task_struct *task_of(struct sched_entity *se)
{
	return container_of(se, struct task_struct, se);
}
360

361 362 363
static inline struct rq *rq_of(struct cfs_rq *cfs_rq)
{
	return container_of(cfs_rq, struct rq, cfs);
364 365 366 367
}

#define entity_is_task(se)	1

P
Peter Zijlstra 已提交
368 369
#define for_each_sched_entity(se) \
		for (; se; se = NULL)
370

P
Peter Zijlstra 已提交
371
static inline struct cfs_rq *task_cfs_rq(struct task_struct *p)
372
{
P
Peter Zijlstra 已提交
373
	return &task_rq(p)->cfs;
374 375
}

P
Peter Zijlstra 已提交
376 377 378 379 380 381 382 383 384 385 386 387 388 389
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;
}

390 391 392 393 394 395 396 397
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
398 399 400 401 402 403 404 405 406 407 408 409 410 411
#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;
}

412 413 414 415 416
static inline void
find_matching_se(struct sched_entity **se, struct sched_entity **pse)
{
}

P
Peter Zijlstra 已提交
417 418
#endif	/* CONFIG_FAIR_GROUP_SCHED */

419 420
static __always_inline
void account_cfs_rq_runtime(struct cfs_rq *cfs_rq, unsigned long delta_exec);
421 422 423 424 425

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

426
static inline u64 max_vruntime(u64 min_vruntime, u64 vruntime)
427
{
428 429
	s64 delta = (s64)(vruntime - min_vruntime);
	if (delta > 0)
430 431 432 433 434
		min_vruntime = vruntime;

	return min_vruntime;
}

435
static inline u64 min_vruntime(u64 min_vruntime, u64 vruntime)
P
Peter Zijlstra 已提交
436 437 438 439 440 441 442 443
{
	s64 delta = (s64)(vruntime - min_vruntime);
	if (delta < 0)
		min_vruntime = vruntime;

	return min_vruntime;
}

444 445 446 447 448 449
static inline int entity_before(struct sched_entity *a,
				struct sched_entity *b)
{
	return (s64)(a->vruntime - b->vruntime) < 0;
}

450 451 452 453 454 455 456 457 458 459 460 461
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
462
		if (!cfs_rq->curr)
463 464 465 466 467 468
			vruntime = se->vruntime;
		else
			vruntime = min_vruntime(vruntime, se->vruntime);
	}

	cfs_rq->min_vruntime = max_vruntime(cfs_rq->min_vruntime, vruntime);
469 470 471 472
#ifndef CONFIG_64BIT
	smp_wmb();
	cfs_rq->min_vruntime_copy = cfs_rq->min_vruntime;
#endif
473 474
}

475 476 477
/*
 * Enqueue an entity into the rb-tree:
 */
478
static void __enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494
{
	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.
		 */
495
		if (entity_before(se, entry)) {
496 497 498 499 500 501 502 503 504 505 506
			link = &parent->rb_left;
		} else {
			link = &parent->rb_right;
			leftmost = 0;
		}
	}

	/*
	 * Maintain a cache of leftmost tree entries (it is frequently
	 * used):
	 */
507
	if (leftmost)
I
Ingo Molnar 已提交
508
		cfs_rq->rb_leftmost = &se->run_node;
509 510 511 512 513

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

514
static void __dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
515
{
P
Peter Zijlstra 已提交
516 517 518 519 520 521
	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
522

523 524 525
	rb_erase(&se->run_node, &cfs_rq->tasks_timeline);
}

526
struct sched_entity *__pick_first_entity(struct cfs_rq *cfs_rq)
527
{
528 529 530 531 532 533
	struct rb_node *left = cfs_rq->rb_leftmost;

	if (!left)
		return NULL;

	return rb_entry(left, struct sched_entity, run_node);
534 535
}

536 537 538 539 540 541 542 543 544 545 546
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
547
struct sched_entity *__pick_last_entity(struct cfs_rq *cfs_rq)
548
{
549
	struct rb_node *last = rb_last(&cfs_rq->tasks_timeline);
550

551 552
	if (!last)
		return NULL;
553 554

	return rb_entry(last, struct sched_entity, run_node);
555 556
}

557 558 559 560
/**************************************************************
 * Scheduling class statistics methods:
 */

561
int sched_proc_update_handler(struct ctl_table *table, int write,
562
		void __user *buffer, size_t *lenp,
563 564
		loff_t *ppos)
{
565
	int ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
566
	int factor = get_update_sysctl_factor();
567 568 569 570 571 572 573

	if (ret || !write)
		return ret;

	sched_nr_latency = DIV_ROUND_UP(sysctl_sched_latency,
					sysctl_sched_min_granularity);

574 575 576 577 578 579 580
#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

581 582 583
	return 0;
}
#endif
584

585
/*
586
 * delta /= w
587 588 589 590
 */
static inline unsigned long
calc_delta_fair(unsigned long delta, struct sched_entity *se)
{
591 592
	if (unlikely(se->load.weight != NICE_0_LOAD))
		delta = calc_delta_mine(delta, NICE_0_LOAD, &se->load);
593 594 595 596

	return delta;
}

597 598 599 600 601 602 603 604
/*
 * The idea is to set a period in which each task runs once.
 *
 * When there are too many tasks (sysctl_sched_nr_latency) we have to stretch
 * this period because otherwise the slices get too small.
 *
 * p = (nr <= nl) ? l : l*nr/nl
 */
605 606 607
static u64 __sched_period(unsigned long nr_running)
{
	u64 period = sysctl_sched_latency;
608
	unsigned long nr_latency = sched_nr_latency;
609 610

	if (unlikely(nr_running > nr_latency)) {
611
		period = sysctl_sched_min_granularity;
612 613 614 615 616 617
		period *= nr_running;
	}

	return period;
}

618 619 620 621
/*
 * We calculate the wall-time slice from the period by taking a part
 * proportional to the weight.
 *
622
 * s = p*P[w/rw]
623
 */
P
Peter Zijlstra 已提交
624
static u64 sched_slice(struct cfs_rq *cfs_rq, struct sched_entity *se)
625
{
M
Mike Galbraith 已提交
626
	u64 slice = __sched_period(cfs_rq->nr_running + !se->on_rq);
627

M
Mike Galbraith 已提交
628
	for_each_sched_entity(se) {
L
Lin Ming 已提交
629
		struct load_weight *load;
630
		struct load_weight lw;
L
Lin Ming 已提交
631 632 633

		cfs_rq = cfs_rq_of(se);
		load = &cfs_rq->load;
634

M
Mike Galbraith 已提交
635
		if (unlikely(!se->on_rq)) {
636
			lw = cfs_rq->load;
M
Mike Galbraith 已提交
637 638 639 640 641 642 643

			update_load_add(&lw, se->load.weight);
			load = &lw;
		}
		slice = calc_delta_mine(slice, se->load.weight, load);
	}
	return slice;
644 645
}

646
/*
647
 * We calculate the vruntime slice of a to be inserted task
648
 *
649
 * vs = s/w
650
 */
651
static u64 sched_vslice(struct cfs_rq *cfs_rq, struct sched_entity *se)
P
Peter Zijlstra 已提交
652
{
653
	return calc_delta_fair(sched_slice(cfs_rq, se), se);
654 655
}

656
static void update_cfs_load(struct cfs_rq *cfs_rq, int global_update);
657
static void update_cfs_shares(struct cfs_rq *cfs_rq);
658

659 660 661 662 663
/*
 * 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
664 665
__update_curr(struct cfs_rq *cfs_rq, struct sched_entity *curr,
	      unsigned long delta_exec)
666
{
667
	unsigned long delta_exec_weighted;
668

669 670
	schedstat_set(curr->statistics.exec_max,
		      max((u64)delta_exec, curr->statistics.exec_max));
671 672

	curr->sum_exec_runtime += delta_exec;
673
	schedstat_add(cfs_rq, exec_clock, delta_exec);
674
	delta_exec_weighted = calc_delta_fair(delta_exec, curr);
675

I
Ingo Molnar 已提交
676
	curr->vruntime += delta_exec_weighted;
677
	update_min_vruntime(cfs_rq);
678

P
Peter Zijlstra 已提交
679
#if defined CONFIG_SMP && defined CONFIG_FAIR_GROUP_SCHED
680 681
	cfs_rq->load_unacc_exec_time += delta_exec;
#endif
682 683
}

684
static void update_curr(struct cfs_rq *cfs_rq)
685
{
686
	struct sched_entity *curr = cfs_rq->curr;
687
	u64 now = rq_of(cfs_rq)->clock_task;
688 689 690 691 692 693 694 695 696 697
	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
698
	delta_exec = (unsigned long)(now - curr->exec_start);
P
Peter Zijlstra 已提交
699 700
	if (!delta_exec)
		return;
701

I
Ingo Molnar 已提交
702 703
	__update_curr(cfs_rq, curr, delta_exec);
	curr->exec_start = now;
704 705 706 707

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

708
		trace_sched_stat_runtime(curtask, delta_exec, curr->vruntime);
709
		cpuacct_charge(curtask, delta_exec);
710
		account_group_exec_runtime(curtask, delta_exec);
711
	}
712 713

	account_cfs_rq_runtime(cfs_rq, delta_exec);
714 715 716
}

static inline void
717
update_stats_wait_start(struct cfs_rq *cfs_rq, struct sched_entity *se)
718
{
719
	schedstat_set(se->statistics.wait_start, rq_of(cfs_rq)->clock);
720 721 722 723 724
}

/*
 * Task is being enqueued - update stats:
 */
725
static void update_stats_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se)
726 727 728 729 730
{
	/*
	 * Are we enqueueing a waiting task? (for current tasks
	 * a dequeue/enqueue event is a NOP)
	 */
731
	if (se != cfs_rq->curr)
732
		update_stats_wait_start(cfs_rq, se);
733 734 735
}

static void
736
update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se)
737
{
738 739 740 741 742
	schedstat_set(se->statistics.wait_max, max(se->statistics.wait_max,
			rq_of(cfs_rq)->clock - se->statistics.wait_start));
	schedstat_set(se->statistics.wait_count, se->statistics.wait_count + 1);
	schedstat_set(se->statistics.wait_sum, se->statistics.wait_sum +
			rq_of(cfs_rq)->clock - se->statistics.wait_start);
743 744 745
#ifdef CONFIG_SCHEDSTATS
	if (entity_is_task(se)) {
		trace_sched_stat_wait(task_of(se),
746
			rq_of(cfs_rq)->clock - se->statistics.wait_start);
747 748
	}
#endif
749
	schedstat_set(se->statistics.wait_start, 0);
750 751 752
}

static inline void
753
update_stats_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se)
754 755 756 757 758
{
	/*
	 * Mark the end of the wait period if dequeueing a
	 * waiting task:
	 */
759
	if (se != cfs_rq->curr)
760
		update_stats_wait_end(cfs_rq, se);
761 762 763 764 765 766
}

/*
 * We are picking a new current task - update its stats:
 */
static inline void
767
update_stats_curr_start(struct cfs_rq *cfs_rq, struct sched_entity *se)
768 769 770 771
{
	/*
	 * We are starting a new run period:
	 */
772
	se->exec_start = rq_of(cfs_rq)->clock_task;
773 774 775 776 777 778
}

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

779 780 781 782
static void
account_entity_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se)
{
	update_load_add(&cfs_rq->load, se->load.weight);
783
	if (!parent_entity(se))
784
		update_load_add(&rq_of(cfs_rq)->load, se->load.weight);
785 786
#ifdef CONFIG_SMP
	if (entity_is_task(se))
787
		list_add(&se->group_node, &rq_of(cfs_rq)->cfs_tasks);
788
#endif
789 790 791 792 793 794 795
	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);
796
	if (!parent_entity(se))
797
		update_load_sub(&rq_of(cfs_rq)->load, se->load.weight);
798
	if (entity_is_task(se))
799
		list_del_init(&se->group_node);
800 801 802
	cfs_rq->nr_running--;
}

803
#ifdef CONFIG_FAIR_GROUP_SCHED
804 805
/* we need this in update_cfs_load and load-balance functions below */
static inline int throttled_hierarchy(struct cfs_rq *cfs_rq);
806
# ifdef CONFIG_SMP
807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822
static void update_cfs_rq_load_contribution(struct cfs_rq *cfs_rq,
					    int global_update)
{
	struct task_group *tg = cfs_rq->tg;
	long load_avg;

	load_avg = div64_u64(cfs_rq->load_avg, cfs_rq->load_period+1);
	load_avg -= cfs_rq->load_contribution;

	if (global_update || abs(load_avg) > cfs_rq->load_contribution / 8) {
		atomic_add(load_avg, &tg->load_weight);
		cfs_rq->load_contribution += load_avg;
	}
}

static void update_cfs_load(struct cfs_rq *cfs_rq, int global_update)
P
Peter Zijlstra 已提交
823
{
824
	u64 period = sysctl_sched_shares_window;
P
Peter Zijlstra 已提交
825
	u64 now, delta;
826
	unsigned long load = cfs_rq->load.weight;
P
Peter Zijlstra 已提交
827

828
	if (cfs_rq->tg == &root_task_group || throttled_hierarchy(cfs_rq))
P
Peter Zijlstra 已提交
829 830
		return;

831
	now = rq_of(cfs_rq)->clock_task;
P
Peter Zijlstra 已提交
832 833
	delta = now - cfs_rq->load_stamp;

834 835 836 837 838
	/* truncate load history at 4 idle periods */
	if (cfs_rq->load_stamp > cfs_rq->load_last &&
	    now - cfs_rq->load_last > 4 * period) {
		cfs_rq->load_period = 0;
		cfs_rq->load_avg = 0;
839
		delta = period - 1;
840 841
	}

P
Peter Zijlstra 已提交
842
	cfs_rq->load_stamp = now;
843
	cfs_rq->load_unacc_exec_time = 0;
P
Peter Zijlstra 已提交
844
	cfs_rq->load_period += delta;
845 846 847 848
	if (load) {
		cfs_rq->load_last = now;
		cfs_rq->load_avg += delta * load;
	}
P
Peter Zijlstra 已提交
849

850 851 852 853 854
	/* consider updating load contribution on each fold or truncate */
	if (global_update || cfs_rq->load_period > period
	    || !cfs_rq->load_period)
		update_cfs_rq_load_contribution(cfs_rq, global_update);

P
Peter Zijlstra 已提交
855 856 857 858 859 860 861 862 863 864
	while (cfs_rq->load_period > period) {
		/*
		 * Inline assembly required to prevent the compiler
		 * optimising this loop into a divmod call.
		 * See __iter_div_u64_rem() for another example of this.
		 */
		asm("" : "+rm" (cfs_rq->load_period));
		cfs_rq->load_period /= 2;
		cfs_rq->load_avg /= 2;
	}
865

866 867
	if (!cfs_rq->curr && !cfs_rq->nr_running && !cfs_rq->load_avg)
		list_del_leaf_cfs_rq(cfs_rq);
P
Peter Zijlstra 已提交
868 869
}

870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885
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().
	 */
	tg_weight = atomic_read(&tg->load_weight);
	tg_weight -= cfs_rq->load_contribution;
	tg_weight += cfs_rq->load.weight;

	return tg_weight;
}

886
static long calc_cfs_shares(struct cfs_rq *cfs_rq, struct task_group *tg)
887
{
888
	long tg_weight, load, shares;
889

890
	tg_weight = calc_tg_weight(tg, cfs_rq);
891
	load = cfs_rq->load.weight;
892 893

	shares = (tg->shares * load);
894 895
	if (tg_weight)
		shares /= tg_weight;
896 897 898 899 900 901 902 903 904 905 906 907 908

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

	return shares;
}

static void update_entity_shares_tick(struct cfs_rq *cfs_rq)
{
	if (cfs_rq->load_unacc_exec_time > sysctl_sched_shares_window) {
		update_cfs_load(cfs_rq, 0);
909
		update_cfs_shares(cfs_rq);
910 911 912 913 914 915 916
	}
}
# else /* CONFIG_SMP */
static void update_cfs_load(struct cfs_rq *cfs_rq, int global_update)
{
}

917
static inline long calc_cfs_shares(struct cfs_rq *cfs_rq, struct task_group *tg)
918 919 920 921 922 923 924 925
{
	return tg->shares;
}

static inline void update_entity_shares_tick(struct cfs_rq *cfs_rq)
{
}
# endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
926 927 928
static void reweight_entity(struct cfs_rq *cfs_rq, struct sched_entity *se,
			    unsigned long weight)
{
929 930 931 932
	if (se->on_rq) {
		/* commit outstanding execution time */
		if (cfs_rq->curr == se)
			update_curr(cfs_rq);
P
Peter Zijlstra 已提交
933
		account_entity_dequeue(cfs_rq, se);
934
	}
P
Peter Zijlstra 已提交
935 936 937 938 939 940 941

	update_load_set(&se->load, weight);

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

942
static void update_cfs_shares(struct cfs_rq *cfs_rq)
P
Peter Zijlstra 已提交
943 944 945
{
	struct task_group *tg;
	struct sched_entity *se;
946
	long shares;
P
Peter Zijlstra 已提交
947 948 949

	tg = cfs_rq->tg;
	se = tg->se[cpu_of(rq_of(cfs_rq))];
950
	if (!se || throttled_hierarchy(cfs_rq))
P
Peter Zijlstra 已提交
951
		return;
952 953 954 955
#ifndef CONFIG_SMP
	if (likely(se->load.weight == tg->shares))
		return;
#endif
956
	shares = calc_cfs_shares(cfs_rq, tg);
P
Peter Zijlstra 已提交
957 958 959 960

	reweight_entity(cfs_rq_of(se), se, shares);
}
#else /* CONFIG_FAIR_GROUP_SCHED */
961
static void update_cfs_load(struct cfs_rq *cfs_rq, int global_update)
P
Peter Zijlstra 已提交
962 963 964
{
}

965
static inline void update_cfs_shares(struct cfs_rq *cfs_rq)
P
Peter Zijlstra 已提交
966 967
{
}
968 969 970 971

static inline void update_entity_shares_tick(struct cfs_rq *cfs_rq)
{
}
P
Peter Zijlstra 已提交
972 973
#endif /* CONFIG_FAIR_GROUP_SCHED */

974
static void enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
975 976
{
#ifdef CONFIG_SCHEDSTATS
977 978 979 980 981
	struct task_struct *tsk = NULL;

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

982 983
	if (se->statistics.sleep_start) {
		u64 delta = rq_of(cfs_rq)->clock - se->statistics.sleep_start;
984 985 986 987

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

988 989
		if (unlikely(delta > se->statistics.sleep_max))
			se->statistics.sleep_max = delta;
990

991
		se->statistics.sleep_start = 0;
992
		se->statistics.sum_sleep_runtime += delta;
A
Arjan van de Ven 已提交
993

994
		if (tsk) {
995
			account_scheduler_latency(tsk, delta >> 10, 1);
996 997
			trace_sched_stat_sleep(tsk, delta);
		}
998
	}
999 1000
	if (se->statistics.block_start) {
		u64 delta = rq_of(cfs_rq)->clock - se->statistics.block_start;
1001 1002 1003 1004

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

1005 1006
		if (unlikely(delta > se->statistics.block_max))
			se->statistics.block_max = delta;
1007

1008
		se->statistics.block_start = 0;
1009
		se->statistics.sum_sleep_runtime += delta;
I
Ingo Molnar 已提交
1010

1011
		if (tsk) {
1012
			if (tsk->in_iowait) {
1013 1014
				se->statistics.iowait_sum += delta;
				se->statistics.iowait_count++;
1015
				trace_sched_stat_iowait(tsk, delta);
1016 1017
			}

1018 1019
			trace_sched_stat_blocked(tsk, delta);

1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030
			/*
			 * 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
1031
		}
1032 1033 1034 1035
	}
#endif
}

P
Peter Zijlstra 已提交
1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048
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
}

1049 1050 1051
static void
place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial)
{
1052
	u64 vruntime = cfs_rq->min_vruntime;
P
Peter Zijlstra 已提交
1053

1054 1055 1056 1057 1058 1059
	/*
	 * 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
1060
	if (initial && sched_feat(START_DEBIT))
1061
		vruntime += sched_vslice(cfs_rq, se);
1062

1063
	/* sleeps up to a single latency don't count. */
1064
	if (!initial) {
1065
		unsigned long thresh = sysctl_sched_latency;
1066

1067 1068 1069 1070 1071 1072
		/*
		 * Halve their sleep time's effect, to allow
		 * for a gentler effect of sleepers:
		 */
		if (sched_feat(GENTLE_FAIR_SLEEPERS))
			thresh >>= 1;
1073

1074
		vruntime -= thresh;
1075 1076
	}

1077 1078 1079
	/* ensure we never gain time by being placed backwards. */
	vruntime = max_vruntime(se->vruntime, vruntime);

P
Peter Zijlstra 已提交
1080
	se->vruntime = vruntime;
1081 1082
}

1083 1084
static void check_enqueue_throttle(struct cfs_rq *cfs_rq);

1085
static void
1086
enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
1087
{
1088 1089 1090 1091
	/*
	 * Update the normalized vruntime before updating min_vruntime
	 * through callig update_curr().
	 */
1092
	if (!(flags & ENQUEUE_WAKEUP) || (flags & ENQUEUE_WAKING))
1093 1094
		se->vruntime += cfs_rq->min_vruntime;

1095
	/*
1096
	 * Update run-time statistics of the 'current'.
1097
	 */
1098
	update_curr(cfs_rq);
1099
	update_cfs_load(cfs_rq, 0);
P
Peter Zijlstra 已提交
1100
	account_entity_enqueue(cfs_rq, se);
1101
	update_cfs_shares(cfs_rq);
1102

1103
	if (flags & ENQUEUE_WAKEUP) {
1104
		place_entity(cfs_rq, se, 0);
1105
		enqueue_sleeper(cfs_rq, se);
I
Ingo Molnar 已提交
1106
	}
1107

1108
	update_stats_enqueue(cfs_rq, se);
P
Peter Zijlstra 已提交
1109
	check_spread(cfs_rq, se);
1110 1111
	if (se != cfs_rq->curr)
		__enqueue_entity(cfs_rq, se);
P
Peter Zijlstra 已提交
1112
	se->on_rq = 1;
1113

1114
	if (cfs_rq->nr_running == 1) {
1115
		list_add_leaf_cfs_rq(cfs_rq);
1116 1117
		check_enqueue_throttle(cfs_rq);
	}
1118 1119
}

1120
static void __clear_buddies_last(struct sched_entity *se)
P
Peter Zijlstra 已提交
1121
{
1122 1123 1124 1125 1126 1127 1128 1129
	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
1130

1131 1132 1133 1134 1135 1136 1137 1138 1139
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
1140 1141
}

1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152
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
1153 1154
static void clear_buddies(struct cfs_rq *cfs_rq, struct sched_entity *se)
{
1155 1156 1157 1158 1159
	if (cfs_rq->last == se)
		__clear_buddies_last(se);

	if (cfs_rq->next == se)
		__clear_buddies_next(se);
1160 1161 1162

	if (cfs_rq->skip == se)
		__clear_buddies_skip(se);
P
Peter Zijlstra 已提交
1163 1164
}

1165
static __always_inline void return_cfs_rq_runtime(struct cfs_rq *cfs_rq);
1166

1167
static void
1168
dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
1169
{
1170 1171 1172 1173 1174
	/*
	 * Update run-time statistics of the 'current'.
	 */
	update_curr(cfs_rq);

1175
	update_stats_dequeue(cfs_rq, se);
1176
	if (flags & DEQUEUE_SLEEP) {
P
Peter Zijlstra 已提交
1177
#ifdef CONFIG_SCHEDSTATS
1178 1179 1180 1181
		if (entity_is_task(se)) {
			struct task_struct *tsk = task_of(se);

			if (tsk->state & TASK_INTERRUPTIBLE)
1182
				se->statistics.sleep_start = rq_of(cfs_rq)->clock;
1183
			if (tsk->state & TASK_UNINTERRUPTIBLE)
1184
				se->statistics.block_start = rq_of(cfs_rq)->clock;
1185
		}
1186
#endif
P
Peter Zijlstra 已提交
1187 1188
	}

P
Peter Zijlstra 已提交
1189
	clear_buddies(cfs_rq, se);
P
Peter Zijlstra 已提交
1190

1191
	if (se != cfs_rq->curr)
1192
		__dequeue_entity(cfs_rq, se);
P
Peter Zijlstra 已提交
1193
	se->on_rq = 0;
1194
	update_cfs_load(cfs_rq, 0);
1195
	account_entity_dequeue(cfs_rq, se);
1196 1197 1198 1199 1200 1201

	/*
	 * 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.
	 */
1202
	if (!(flags & DEQUEUE_SLEEP))
1203
		se->vruntime -= cfs_rq->min_vruntime;
1204

1205 1206 1207
	/* return excess runtime on last dequeue */
	return_cfs_rq_runtime(cfs_rq);

1208 1209
	update_min_vruntime(cfs_rq);
	update_cfs_shares(cfs_rq);
1210 1211 1212 1213 1214
}

/*
 * Preempt the current task with a newly woken task if needed:
 */
1215
static void
I
Ingo Molnar 已提交
1216
check_preempt_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr)
1217
{
1218
	unsigned long ideal_runtime, delta_exec;
1219 1220
	struct sched_entity *se;
	s64 delta;
1221

P
Peter Zijlstra 已提交
1222
	ideal_runtime = sched_slice(cfs_rq, curr);
1223
	delta_exec = curr->sum_exec_runtime - curr->prev_sum_exec_runtime;
1224
	if (delta_exec > ideal_runtime) {
1225
		resched_task(rq_of(cfs_rq)->curr);
1226 1227 1228 1229 1230
		/*
		 * The current task ran long enough, ensure it doesn't get
		 * re-elected due to buddy favours.
		 */
		clear_buddies(cfs_rq, curr);
1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241
		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;

1242 1243
	se = __pick_first_entity(cfs_rq);
	delta = curr->vruntime - se->vruntime;
1244

1245 1246
	if (delta < 0)
		return;
1247

1248 1249
	if (delta > ideal_runtime)
		resched_task(rq_of(cfs_rq)->curr);
1250 1251
}

1252
static void
1253
set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
1254
{
1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265
	/* '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);
	}

1266
	update_stats_curr_start(cfs_rq, se);
1267
	cfs_rq->curr = se;
I
Ingo Molnar 已提交
1268 1269 1270 1271 1272 1273
#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):
	 */
1274
	if (rq_of(cfs_rq)->load.weight >= 2*se->load.weight) {
1275
		se->statistics.slice_max = max(se->statistics.slice_max,
I
Ingo Molnar 已提交
1276 1277 1278
			se->sum_exec_runtime - se->prev_sum_exec_runtime);
	}
#endif
1279
	se->prev_sum_exec_runtime = se->sum_exec_runtime;
1280 1281
}

1282 1283 1284
static int
wakeup_preempt_entity(struct sched_entity *curr, struct sched_entity *se);

1285 1286 1287 1288 1289 1290 1291
/*
 * 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
 */
1292
static struct sched_entity *pick_next_entity(struct cfs_rq *cfs_rq)
1293
{
1294
	struct sched_entity *se = __pick_first_entity(cfs_rq);
1295
	struct sched_entity *left = se;
1296

1297 1298 1299 1300 1301 1302 1303 1304 1305
	/*
	 * 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;
	}
1306

1307 1308 1309 1310 1311 1312
	/*
	 * 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;

1313 1314 1315 1316 1317 1318
	/*
	 * 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;

1319
	clear_buddies(cfs_rq, se);
P
Peter Zijlstra 已提交
1320 1321

	return se;
1322 1323
}

1324 1325
static void check_cfs_rq_runtime(struct cfs_rq *cfs_rq);

1326
static void put_prev_entity(struct cfs_rq *cfs_rq, struct sched_entity *prev)
1327 1328 1329 1330 1331 1332
{
	/*
	 * If still on the runqueue then deactivate_task()
	 * was not called and update_curr() has to be done:
	 */
	if (prev->on_rq)
1333
		update_curr(cfs_rq);
1334

1335 1336 1337
	/* throttle cfs_rqs exceeding runtime */
	check_cfs_rq_runtime(cfs_rq);

P
Peter Zijlstra 已提交
1338
	check_spread(cfs_rq, prev);
1339
	if (prev->on_rq) {
1340
		update_stats_wait_start(cfs_rq, prev);
1341 1342 1343
		/* Put 'current' back into the tree. */
		__enqueue_entity(cfs_rq, prev);
	}
1344
	cfs_rq->curr = NULL;
1345 1346
}

P
Peter Zijlstra 已提交
1347 1348
static void
entity_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr, int queued)
1349 1350
{
	/*
1351
	 * Update run-time statistics of the 'current'.
1352
	 */
1353
	update_curr(cfs_rq);
1354

1355 1356 1357 1358 1359
	/*
	 * Update share accounting for long-running entities.
	 */
	update_entity_shares_tick(cfs_rq);

P
Peter Zijlstra 已提交
1360 1361 1362 1363 1364
#ifdef CONFIG_SCHED_HRTICK
	/*
	 * queued ticks are scheduled to match the slice, so don't bother
	 * validating it and just reschedule.
	 */
1365 1366 1367 1368
	if (queued) {
		resched_task(rq_of(cfs_rq)->curr);
		return;
	}
P
Peter Zijlstra 已提交
1369 1370 1371 1372 1373 1374 1375 1376
	/*
	 * 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
1377
	if (cfs_rq->nr_running > 1)
I
Ingo Molnar 已提交
1378
		check_preempt_tick(cfs_rq, curr);
1379 1380
}

1381 1382 1383 1384 1385 1386

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

#ifdef CONFIG_CFS_BANDWIDTH
1387 1388

#ifdef HAVE_JUMP_LABEL
1389
static struct static_key __cfs_bandwidth_used;
1390 1391 1392

static inline bool cfs_bandwidth_used(void)
{
1393
	return static_key_false(&__cfs_bandwidth_used);
1394 1395 1396 1397 1398 1399
}

void account_cfs_bandwidth_used(int enabled, int was_enabled)
{
	/* only need to count groups transitioning between enabled/!enabled */
	if (enabled && !was_enabled)
1400
		static_key_slow_inc(&__cfs_bandwidth_used);
1401
	else if (!enabled && was_enabled)
1402
		static_key_slow_dec(&__cfs_bandwidth_used);
1403 1404 1405 1406 1407 1408 1409 1410 1411 1412
}
#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 */

1413 1414 1415 1416 1417 1418 1419 1420
/*
 * default period for cfs group bandwidth.
 * default: 0.1s, units: nanoseconds
 */
static inline u64 default_cfs_period(void)
{
	return 100000000ULL;
}
1421 1422 1423 1424 1425 1426

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

P
Paul Turner 已提交
1427 1428 1429 1430 1431 1432 1433
/*
 * 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
 */
1434
void __refill_cfs_bandwidth_runtime(struct cfs_bandwidth *cfs_b)
P
Paul Turner 已提交
1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445
{
	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);
}

1446 1447 1448 1449 1450
static inline struct cfs_bandwidth *tg_cfs_bandwidth(struct task_group *tg)
{
	return &tg->cfs_bandwidth;
}

1451 1452
/* returns 0 on failure to allocate runtime */
static int assign_cfs_rq_runtime(struct cfs_rq *cfs_rq)
1453 1454 1455
{
	struct task_group *tg = cfs_rq->tg;
	struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(tg);
P
Paul Turner 已提交
1456
	u64 amount = 0, min_amount, expires;
1457 1458 1459 1460 1461 1462 1463

	/* 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;
1464
	else {
P
Paul Turner 已提交
1465 1466 1467 1468 1469 1470 1471 1472
		/*
		 * 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);
1473
			__start_cfs_bandwidth(cfs_b);
P
Paul Turner 已提交
1474
		}
1475 1476 1477 1478 1479 1480

		if (cfs_b->runtime > 0) {
			amount = min(cfs_b->runtime, min_amount);
			cfs_b->runtime -= amount;
			cfs_b->idle = 0;
		}
1481
	}
P
Paul Turner 已提交
1482
	expires = cfs_b->runtime_expires;
1483 1484 1485
	raw_spin_unlock(&cfs_b->lock);

	cfs_rq->runtime_remaining += amount;
P
Paul Turner 已提交
1486 1487 1488 1489 1490 1491 1492
	/*
	 * 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;
1493 1494

	return cfs_rq->runtime_remaining > 0;
1495 1496
}

P
Paul Turner 已提交
1497 1498 1499 1500 1501
/*
 * 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)
1502
{
P
Paul Turner 已提交
1503 1504 1505 1506 1507
	struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(cfs_rq->tg);
	struct rq *rq = rq_of(cfs_rq);

	/* if the deadline is ahead of our clock, nothing to do */
	if (likely((s64)(rq->clock - cfs_rq->runtime_expires) < 0))
1508 1509
		return;

P
Paul Turner 已提交
1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534
	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) */
1535
	cfs_rq->runtime_remaining -= delta_exec;
P
Paul Turner 已提交
1536 1537 1538
	expire_cfs_rq_runtime(cfs_rq);

	if (likely(cfs_rq->runtime_remaining > 0))
1539 1540
		return;

1541 1542 1543 1544 1545 1546
	/*
	 * 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);
1547 1548
}

1549 1550
static __always_inline
void account_cfs_rq_runtime(struct cfs_rq *cfs_rq, unsigned long delta_exec)
1551
{
1552
	if (!cfs_bandwidth_used() || !cfs_rq->runtime_enabled)
1553 1554 1555 1556 1557
		return;

	__account_cfs_rq_runtime(cfs_rq, delta_exec);
}

1558 1559
static inline int cfs_rq_throttled(struct cfs_rq *cfs_rq)
{
1560
	return cfs_bandwidth_used() && cfs_rq->throttled;
1561 1562
}

1563 1564 1565
/* check whether cfs_rq, or any parent, is throttled */
static inline int throttled_hierarchy(struct cfs_rq *cfs_rq)
{
1566
	return cfs_bandwidth_used() && cfs_rq->throttle_count;
1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621
}

/*
 * 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) {
		u64 delta = rq->clock_task - cfs_rq->load_stamp;

		/* leaving throttled state, advance shares averaging windows */
		cfs_rq->load_stamp += delta;
		cfs_rq->load_last += delta;

		/* update entity weight now that we are on_rq again */
		update_cfs_shares(cfs_rq);
	}
#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)];

	/* group is entering throttled state, record last load */
	if (!cfs_rq->throttle_count)
		update_cfs_load(cfs_rq, 0);
	cfs_rq->throttle_count++;

	return 0;
}

1622
static void throttle_cfs_rq(struct cfs_rq *cfs_rq)
1623 1624 1625 1626 1627 1628 1629 1630 1631
{
	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))];

	/* account load preceding throttle */
1632 1633 1634
	rcu_read_lock();
	walk_tg_tree_from(cfs_rq->tg, tg_throttle_down, tg_nop, (void *)rq);
	rcu_read_unlock();
1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654

	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;
1655
	cfs_rq->throttled_timestamp = rq->clock;
1656 1657 1658 1659 1660
	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);
}

1661
void unthrottle_cfs_rq(struct cfs_rq *cfs_rq)
1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672
{
	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;
	raw_spin_lock(&cfs_b->lock);
1673
	cfs_b->throttled_time += rq->clock - cfs_rq->throttled_timestamp;
1674 1675
	list_del_rcu(&cfs_rq->throttled_list);
	raw_spin_unlock(&cfs_b->lock);
1676
	cfs_rq->throttled_timestamp = 0;
1677

1678 1679 1680 1681
	update_rq_clock(rq);
	/* update hierarchical throttle state */
	walk_tg_tree_from(cfs_rq->tg, tg_nop, tg_unthrottle_up, (void *)rq);

1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744
	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;
}

1745 1746 1747 1748 1749 1750 1751 1752
/*
 * 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)
{
1753 1754
	u64 runtime, runtime_expires;
	int idle = 1, throttled;
1755 1756 1757 1758 1759 1760

	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;

1761 1762 1763
	throttled = !list_empty(&cfs_b->throttled_cfs_rq);
	/* idle depends on !throttled (for the case of a large deficit) */
	idle = cfs_b->idle && !throttled;
1764
	cfs_b->nr_periods += overrun;
1765

P
Paul Turner 已提交
1766 1767 1768 1769 1770 1771
	/* if we're going inactive then everything else can be deferred */
	if (idle)
		goto out_unlock;

	__refill_cfs_bandwidth_runtime(cfs_b);

1772 1773 1774 1775 1776 1777
	if (!throttled) {
		/* mark as potentially idle for the upcoming period */
		cfs_b->idle = 1;
		goto out_unlock;
	}

1778 1779 1780
	/* account preceding periods in which throttling occurred */
	cfs_b->nr_throttled += overrun;

1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804
	/*
	 * 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);
	}
1805

1806 1807 1808 1809 1810 1811 1812 1813 1814
	/* 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;
1815 1816 1817 1818 1819 1820 1821
out_unlock:
	if (idle)
		cfs_b->timer_active = 0;
	raw_spin_unlock(&cfs_b->lock);

	return idle;
}
1822

1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886
/* 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)
{
1887 1888 1889
	if (!cfs_bandwidth_used())
		return;

1890
	if (!cfs_rq->runtime_enabled || cfs_rq->nr_running)
1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927
		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);
}

1928 1929 1930 1931 1932 1933 1934
/*
 * 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)
{
1935 1936 1937
	if (!cfs_bandwidth_used())
		return;

1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954
	/* 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)
{
1955 1956 1957
	if (!cfs_bandwidth_used())
		return;

1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969
	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);
}
1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054

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);
}

2055
static void unthrottle_offline_cfs_rqs(struct rq *rq)
2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075
{
	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 */
2076 2077
static __always_inline
void account_cfs_rq_runtime(struct cfs_rq *cfs_rq, unsigned long delta_exec) {}
2078 2079
static void check_cfs_rq_runtime(struct cfs_rq *cfs_rq) {}
static void check_enqueue_throttle(struct cfs_rq *cfs_rq) {}
2080
static __always_inline void return_cfs_rq_runtime(struct cfs_rq *cfs_rq) {}
2081 2082 2083 2084 2085

static inline int cfs_rq_throttled(struct cfs_rq *cfs_rq)
{
	return 0;
}
2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096

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;
}
2097 2098 2099 2100 2101

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) {}
2102 2103
#endif

2104 2105 2106 2107 2108
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) {}
2109
static inline void unthrottle_offline_cfs_rqs(struct rq *rq) {}
2110 2111 2112

#endif /* CONFIG_CFS_BANDWIDTH */

2113 2114 2115 2116
/**************************************************
 * CFS operations on tasks:
 */

P
Peter Zijlstra 已提交
2117 2118 2119 2120 2121 2122 2123 2124
#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);

2125
	if (cfs_rq->nr_running > 1) {
P
Peter Zijlstra 已提交
2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139
		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.
		 */
2140
		if (rq->curr != p)
2141
			delta = max_t(s64, 10000LL, delta);
P
Peter Zijlstra 已提交
2142

2143
		hrtick_start(rq, delta);
P
Peter Zijlstra 已提交
2144 2145
	}
}
2146 2147 2148 2149 2150 2151 2152 2153 2154 2155

/*
 * 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;

2156
	if (!hrtick_enabled(rq) || curr->sched_class != &fair_sched_class)
2157 2158 2159 2160 2161
		return;

	if (cfs_rq_of(&curr->se)->nr_running < sched_nr_latency)
		hrtick_start_fair(rq, curr);
}
2162
#else /* !CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
2163 2164 2165 2166
static inline void
hrtick_start_fair(struct rq *rq, struct task_struct *p)
{
}
2167 2168 2169 2170

static inline void hrtick_update(struct rq *rq)
{
}
P
Peter Zijlstra 已提交
2171 2172
#endif

2173 2174 2175 2176 2177
/*
 * 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:
 */
2178
static void
2179
enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags)
2180 2181
{
	struct cfs_rq *cfs_rq;
2182
	struct sched_entity *se = &p->se;
2183 2184

	for_each_sched_entity(se) {
2185
		if (se->on_rq)
2186 2187
			break;
		cfs_rq = cfs_rq_of(se);
2188
		enqueue_entity(cfs_rq, se, flags);
2189 2190 2191 2192 2193 2194 2195 2196 2197

		/*
		 * 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;
2198
		cfs_rq->h_nr_running++;
2199

2200
		flags = ENQUEUE_WAKEUP;
2201
	}
P
Peter Zijlstra 已提交
2202

P
Peter Zijlstra 已提交
2203
	for_each_sched_entity(se) {
2204
		cfs_rq = cfs_rq_of(se);
2205
		cfs_rq->h_nr_running++;
P
Peter Zijlstra 已提交
2206

2207 2208 2209
		if (cfs_rq_throttled(cfs_rq))
			break;

2210
		update_cfs_load(cfs_rq, 0);
2211
		update_cfs_shares(cfs_rq);
P
Peter Zijlstra 已提交
2212 2213
	}

2214 2215
	if (!se)
		inc_nr_running(rq);
2216
	hrtick_update(rq);
2217 2218
}

2219 2220
static void set_next_buddy(struct sched_entity *se);

2221 2222 2223 2224 2225
/*
 * The dequeue_task method is called before nr_running is
 * decreased. We remove the task from the rbtree and
 * update the fair scheduling stats:
 */
2226
static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags)
2227 2228
{
	struct cfs_rq *cfs_rq;
2229
	struct sched_entity *se = &p->se;
2230
	int task_sleep = flags & DEQUEUE_SLEEP;
2231 2232 2233

	for_each_sched_entity(se) {
		cfs_rq = cfs_rq_of(se);
2234
		dequeue_entity(cfs_rq, se, flags);
2235 2236 2237 2238 2239 2240 2241 2242 2243

		/*
		 * 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;
2244
		cfs_rq->h_nr_running--;
P
Peter Zijlstra 已提交
2245

2246
		/* Don't dequeue parent if it has other entities besides us */
2247 2248 2249 2250 2251 2252 2253
		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));
2254 2255 2256

			/* avoid re-evaluating load for this entity */
			se = parent_entity(se);
2257
			break;
2258
		}
2259
		flags |= DEQUEUE_SLEEP;
2260
	}
P
Peter Zijlstra 已提交
2261

P
Peter Zijlstra 已提交
2262
	for_each_sched_entity(se) {
2263
		cfs_rq = cfs_rq_of(se);
2264
		cfs_rq->h_nr_running--;
P
Peter Zijlstra 已提交
2265

2266 2267 2268
		if (cfs_rq_throttled(cfs_rq))
			break;

2269
		update_cfs_load(cfs_rq, 0);
2270
		update_cfs_shares(cfs_rq);
P
Peter Zijlstra 已提交
2271 2272
	}

2273 2274
	if (!se)
		dec_nr_running(rq);
2275
	hrtick_update(rq);
2276 2277
}

2278
#ifdef CONFIG_SMP
2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333
/* 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;
}

2334

2335
static void task_waking_fair(struct task_struct *p)
2336 2337 2338
{
	struct sched_entity *se = &p->se;
	struct cfs_rq *cfs_rq = cfs_rq_of(se);
2339 2340 2341 2342
	u64 min_vruntime;

#ifndef CONFIG_64BIT
	u64 min_vruntime_copy;
2343

2344 2345 2346 2347 2348 2349 2350 2351
	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
2352

2353
	se->vruntime -= min_vruntime;
2354 2355
}

2356
#ifdef CONFIG_FAIR_GROUP_SCHED
2357 2358 2359 2360 2361 2362
/*
 * 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.
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 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405
 *
 * 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.
2406
 */
P
Peter Zijlstra 已提交
2407
static long effective_load(struct task_group *tg, int cpu, long wl, long wg)
2408
{
P
Peter Zijlstra 已提交
2409
	struct sched_entity *se = tg->se[cpu];
2410

2411
	if (!tg->parent)	/* the trivial, non-cgroup case */
2412 2413
		return wl;

P
Peter Zijlstra 已提交
2414
	for_each_sched_entity(se) {
2415
		long w, W;
P
Peter Zijlstra 已提交
2416

2417
		tg = se->my_q->tg;
2418

2419 2420 2421 2422
		/*
		 * W = @wg + \Sum rw_j
		 */
		W = wg + calc_tg_weight(tg, se->my_q);
P
Peter Zijlstra 已提交
2423

2424 2425 2426 2427
		/*
		 * w = rw_i + @wl
		 */
		w = se->my_q->load.weight + wl;
2428

2429 2430 2431 2432 2433
		/*
		 * wl = S * s'_i; see (2)
		 */
		if (W > 0 && w < W)
			wl = (w * tg->shares) / W;
2434 2435
		else
			wl = tg->shares;
2436

2437 2438 2439 2440 2441
		/*
		 * 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().
		 */
2442 2443
		if (wl < MIN_SHARES)
			wl = MIN_SHARES;
2444 2445 2446 2447

		/*
		 * wl = dw_i = S * (s'_i - s_i); see (3)
		 */
2448
		wl -= se->load.weight;
2449 2450 2451 2452 2453 2454 2455 2456

		/*
		 * 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
2457 2458
		wg = 0;
	}
2459

P
Peter Zijlstra 已提交
2460
	return wl;
2461 2462
}
#else
P
Peter Zijlstra 已提交
2463

2464 2465
static inline unsigned long effective_load(struct task_group *tg, int cpu,
		unsigned long wl, unsigned long wg)
P
Peter Zijlstra 已提交
2466
{
2467
	return wl;
2468
}
P
Peter Zijlstra 已提交
2469

2470 2471
#endif

2472
static int wake_affine(struct sched_domain *sd, struct task_struct *p, int sync)
2473
{
2474
	s64 this_load, load;
2475
	int idx, this_cpu, prev_cpu;
2476
	unsigned long tl_per_task;
2477
	struct task_group *tg;
2478
	unsigned long weight;
2479
	int balanced;
2480

2481 2482 2483 2484 2485
	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);
2486

2487 2488 2489 2490 2491
	/*
	 * If sync wakeup then subtract the (maximum possible)
	 * effect of the currently running task from the load
	 * of the current CPU:
	 */
2492 2493 2494 2495
	if (sync) {
		tg = task_group(current);
		weight = current->se.load.weight;

2496
		this_load += effective_load(tg, this_cpu, -weight, -weight);
2497 2498
		load += effective_load(tg, prev_cpu, 0, -weight);
	}
2499

2500 2501
	tg = task_group(p);
	weight = p->se.load.weight;
2502

2503 2504
	/*
	 * In low-load situations, where prev_cpu is idle and this_cpu is idle
2505 2506 2507
	 * 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.
2508 2509 2510 2511
	 *
	 * Otherwise check if either cpus are near enough in load to allow this
	 * task to be woken on this_cpu.
	 */
2512 2513
	if (this_load > 0) {
		s64 this_eff_load, prev_eff_load;
2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526

		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;
2527

2528
	/*
I
Ingo Molnar 已提交
2529 2530 2531
	 * If the currently running task will sleep within
	 * a reasonable amount of time then attract this newly
	 * woken task:
2532
	 */
2533 2534
	if (sync && balanced)
		return 1;
2535

2536
	schedstat_inc(p, se.statistics.nr_wakeups_affine_attempts);
2537 2538
	tl_per_task = cpu_avg_load_per_task(this_cpu);

2539 2540 2541
	if (balanced ||
	    (this_load <= load &&
	     this_load + target_load(prev_cpu, idx) <= tl_per_task)) {
2542 2543 2544 2545 2546
		/*
		 * This domain has SD_WAKE_AFFINE and
		 * p is cache cold in this domain, and
		 * there is no bad imbalance.
		 */
2547
		schedstat_inc(sd, ttwu_move_affine);
2548
		schedstat_inc(p, se.statistics.nr_wakeups_affine);
2549 2550 2551 2552 2553 2554

		return 1;
	}
	return 0;
}

2555 2556 2557 2558 2559
/*
 * find_idlest_group finds and returns the least busy CPU group within the
 * domain.
 */
static struct sched_group *
P
Peter Zijlstra 已提交
2560
find_idlest_group(struct sched_domain *sd, struct task_struct *p,
2561
		  int this_cpu, int load_idx)
2562
{
2563
	struct sched_group *idlest = NULL, *group = sd->groups;
2564 2565
	unsigned long min_load = ULONG_MAX, this_load = 0;
	int imbalance = 100 + (sd->imbalance_pct-100)/2;
2566

2567 2568 2569 2570
	do {
		unsigned long load, avg_load;
		int local_group;
		int i;
2571

2572 2573
		/* Skip over this group if it has no CPUs allowed */
		if (!cpumask_intersects(sched_group_cpus(group),
2574
					tsk_cpus_allowed(p)))
2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593
			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 */
2594
		avg_load = (avg_load * SCHED_POWER_SCALE) / group->sgp->power;
2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619

		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 */
2620
	for_each_cpu_and(i, sched_group_cpus(group), tsk_cpus_allowed(p)) {
2621 2622 2623 2624 2625
		load = weighted_cpuload(i);

		if (load < min_load || (load == min_load && i == this_cpu)) {
			min_load = load;
			idlest = i;
2626 2627 2628
		}
	}

2629 2630
	return idlest;
}
2631

2632 2633 2634
/*
 * Try and locate an idle CPU in the sched_domain.
 */
2635
static int select_idle_sibling(struct task_struct *p, int target)
2636 2637 2638
{
	int cpu = smp_processor_id();
	int prev_cpu = task_cpu(p);
2639
	struct sched_domain *sd;
2640 2641

	/*
2642 2643
	 * If the task is going to be woken-up on this cpu and if it is
	 * already idle, then it is the right target.
2644
	 */
2645 2646 2647 2648 2649 2650 2651 2652
	if (target == cpu && idle_cpu(cpu))
		return cpu;

	/*
	 * If the task is going to be woken-up on the cpu where it previously
	 * ran and if it is currently idle, then it the right target.
	 */
	if (target == prev_cpu && idle_cpu(prev_cpu))
2653
		return prev_cpu;
2654 2655

	/*
2656
	 * Otherwise, check assigned siblings to find an elegible idle cpu.
2657
	 */
2658
	sd = rcu_dereference(per_cpu(sd_llc, target));
2659

2660 2661 2662 2663 2664
	for_each_lower_domain(sd) {
		if (!cpumask_test_cpu(sd->idle_buddy, tsk_cpus_allowed(p)))
			continue;
		if (idle_cpu(sd->idle_buddy))
			return sd->idle_buddy;
2665
	}
2666

2667 2668 2669
	return target;
}

2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680
/*
 * 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.
 */
2681
static int
2682
select_task_rq_fair(struct task_struct *p, int sd_flag, int wake_flags)
2683
{
2684
	struct sched_domain *tmp, *affine_sd = NULL, *sd = NULL;
2685 2686 2687
	int cpu = smp_processor_id();
	int prev_cpu = task_cpu(p);
	int new_cpu = cpu;
2688
	int want_affine = 0;
2689
	int want_sd = 1;
2690
	int sync = wake_flags & WF_SYNC;
2691

2692
	if (p->nr_cpus_allowed == 1)
2693 2694
		return prev_cpu;

2695
	if (sd_flag & SD_BALANCE_WAKE) {
2696
		if (cpumask_test_cpu(cpu, tsk_cpus_allowed(p)))
2697 2698 2699
			want_affine = 1;
		new_cpu = prev_cpu;
	}
2700

2701
	rcu_read_lock();
2702
	for_each_domain(cpu, tmp) {
2703 2704 2705
		if (!(tmp->flags & SD_LOAD_BALANCE))
			continue;

2706
		/*
2707 2708
		 * If power savings logic is enabled for a domain, see if we
		 * are not overloaded, if so, don't balance wider.
2709
		 */
2710
		if (tmp->flags & (SD_PREFER_LOCAL)) {
2711 2712 2713 2714 2715 2716 2717 2718 2719 2720
			unsigned long power = 0;
			unsigned long nr_running = 0;
			unsigned long capacity;
			int i;

			for_each_cpu(i, sched_domain_span(tmp)) {
				power += power_of(i);
				nr_running += cpu_rq(i)->cfs.nr_running;
			}

2721
			capacity = DIV_ROUND_CLOSEST(power, SCHED_POWER_SCALE);
2722

P
Peter Zijlstra 已提交
2723
			if (nr_running < capacity)
2724
				want_sd = 0;
2725
		}
2726

2727
		/*
2728 2729
		 * If both cpu and prev_cpu are part of this domain,
		 * cpu is a valid SD_WAKE_AFFINE target.
2730
		 */
2731 2732 2733 2734
		if (want_affine && (tmp->flags & SD_WAKE_AFFINE) &&
		    cpumask_test_cpu(prev_cpu, sched_domain_span(tmp))) {
			affine_sd = tmp;
			want_affine = 0;
2735 2736
		}

2737 2738 2739
		if (!want_sd && !want_affine)
			break;

2740
		if (!(tmp->flags & sd_flag))
2741 2742
			continue;

2743 2744 2745 2746
		if (want_sd)
			sd = tmp;
	}

2747
	if (affine_sd) {
2748
		if (cpu == prev_cpu || wake_affine(affine_sd, p, sync))
2749 2750 2751 2752
			prev_cpu = cpu;

		new_cpu = select_idle_sibling(p, prev_cpu);
		goto unlock;
2753
	}
2754

2755
	while (sd) {
2756
		int load_idx = sd->forkexec_idx;
2757
		struct sched_group *group;
2758
		int weight;
2759

2760
		if (!(sd->flags & sd_flag)) {
2761 2762 2763
			sd = sd->child;
			continue;
		}
2764

2765 2766
		if (sd_flag & SD_BALANCE_WAKE)
			load_idx = sd->wake_idx;
2767

2768
		group = find_idlest_group(sd, p, cpu, load_idx);
2769 2770 2771 2772
		if (!group) {
			sd = sd->child;
			continue;
		}
I
Ingo Molnar 已提交
2773

2774
		new_cpu = find_idlest_cpu(group, p, cpu);
2775 2776 2777 2778
		if (new_cpu == -1 || new_cpu == cpu) {
			/* Now try balancing at a lower domain level of cpu */
			sd = sd->child;
			continue;
2779
		}
2780 2781 2782

		/* Now try balancing at a lower domain level of new_cpu */
		cpu = new_cpu;
2783
		weight = sd->span_weight;
2784 2785
		sd = NULL;
		for_each_domain(cpu, tmp) {
2786
			if (weight <= tmp->span_weight)
2787
				break;
2788
			if (tmp->flags & sd_flag)
2789 2790 2791
				sd = tmp;
		}
		/* while loop will break here if sd == NULL */
2792
	}
2793 2794
unlock:
	rcu_read_unlock();
2795

2796
	return new_cpu;
2797 2798 2799
}
#endif /* CONFIG_SMP */

P
Peter Zijlstra 已提交
2800 2801
static unsigned long
wakeup_gran(struct sched_entity *curr, struct sched_entity *se)
2802 2803 2804 2805
{
	unsigned long gran = sysctl_sched_wakeup_granularity;

	/*
P
Peter Zijlstra 已提交
2806 2807
	 * Since its curr running now, convert the gran from real-time
	 * to virtual-time in his units.
M
Mike Galbraith 已提交
2808 2809 2810 2811 2812 2813 2814 2815 2816
	 *
	 * 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.
2817
	 */
2818
	return calc_delta_fair(gran, se);
2819 2820
}

2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842
/*
 * 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
2843
	gran = wakeup_gran(curr, se);
2844 2845 2846 2847 2848 2849
	if (vdiff > gran)
		return 1;

	return 0;
}

2850 2851
static void set_last_buddy(struct sched_entity *se)
{
2852 2853 2854 2855 2856
	if (entity_is_task(se) && unlikely(task_of(se)->policy == SCHED_IDLE))
		return;

	for_each_sched_entity(se)
		cfs_rq_of(se)->last = se;
2857 2858 2859 2860
}

static void set_next_buddy(struct sched_entity *se)
{
2861 2862 2863 2864 2865
	if (entity_is_task(se) && unlikely(task_of(se)->policy == SCHED_IDLE))
		return;

	for_each_sched_entity(se)
		cfs_rq_of(se)->next = se;
2866 2867
}

2868 2869
static void set_skip_buddy(struct sched_entity *se)
{
2870 2871
	for_each_sched_entity(se)
		cfs_rq_of(se)->skip = se;
2872 2873
}

2874 2875 2876
/*
 * Preempt the current task with a newly woken task if needed:
 */
2877
static void check_preempt_wakeup(struct rq *rq, struct task_struct *p, int wake_flags)
2878 2879
{
	struct task_struct *curr = rq->curr;
2880
	struct sched_entity *se = &curr->se, *pse = &p->se;
2881
	struct cfs_rq *cfs_rq = task_cfs_rq(curr);
2882
	int scale = cfs_rq->nr_running >= sched_nr_latency;
2883
	int next_buddy_marked = 0;
2884

I
Ingo Molnar 已提交
2885 2886 2887
	if (unlikely(se == pse))
		return;

2888
	/*
2889
	 * This is possible from callers such as move_task(), in which we
2890 2891 2892 2893 2894 2895 2896
	 * 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;

2897
	if (sched_feat(NEXT_BUDDY) && scale && !(wake_flags & WF_FORK)) {
M
Mike Galbraith 已提交
2898
		set_next_buddy(pse);
2899 2900
		next_buddy_marked = 1;
	}
P
Peter Zijlstra 已提交
2901

2902 2903 2904
	/*
	 * We can come here with TIF_NEED_RESCHED already set from new task
	 * wake up path.
2905 2906 2907 2908 2909 2910
	 *
	 * 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.
2911 2912 2913 2914
	 */
	if (test_tsk_need_resched(curr))
		return;

2915 2916 2917 2918 2919
	/* Idle tasks are by definition preempted by non-idle tasks. */
	if (unlikely(curr->policy == SCHED_IDLE) &&
	    likely(p->policy != SCHED_IDLE))
		goto preempt;

2920
	/*
2921 2922
	 * Batch and idle tasks do not preempt non-idle tasks (their preemption
	 * is driven by the tick):
2923
	 */
2924
	if (unlikely(p->policy != SCHED_NORMAL))
2925
		return;
2926

2927
	find_matching_se(&se, &pse);
2928
	update_curr(cfs_rq_of(se));
2929
	BUG_ON(!pse);
2930 2931 2932 2933 2934 2935 2936
	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);
2937
		goto preempt;
2938
	}
2939

2940
	return;
2941

2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957
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);
2958 2959
}

2960
static struct task_struct *pick_next_task_fair(struct rq *rq)
2961
{
P
Peter Zijlstra 已提交
2962
	struct task_struct *p;
2963 2964 2965
	struct cfs_rq *cfs_rq = &rq->cfs;
	struct sched_entity *se;

2966
	if (!cfs_rq->nr_running)
2967 2968 2969
		return NULL;

	do {
2970
		se = pick_next_entity(cfs_rq);
2971
		set_next_entity(cfs_rq, se);
2972 2973 2974
		cfs_rq = group_cfs_rq(se);
	} while (cfs_rq);

P
Peter Zijlstra 已提交
2975
	p = task_of(se);
2976 2977
	if (hrtick_enabled(rq))
		hrtick_start_fair(rq, p);
P
Peter Zijlstra 已提交
2978 2979

	return p;
2980 2981 2982 2983 2984
}

/*
 * Account for a descheduled task:
 */
2985
static void put_prev_task_fair(struct rq *rq, struct task_struct *prev)
2986 2987 2988 2989 2990 2991
{
	struct sched_entity *se = &prev->se;
	struct cfs_rq *cfs_rq;

	for_each_sched_entity(se) {
		cfs_rq = cfs_rq_of(se);
2992
		put_prev_entity(cfs_rq, se);
2993 2994 2995
	}
}

2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020
/*
 * 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);
3021 3022 3023 3024 3025 3026
		/*
		 * 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;
3027 3028 3029 3030 3031
	}

	set_skip_buddy(se);
}

3032 3033 3034 3035
static bool yield_to_task_fair(struct rq *rq, struct task_struct *p, bool preempt)
{
	struct sched_entity *se = &p->se;

3036 3037
	/* throttled hierarchies are not runnable */
	if (!se->on_rq || throttled_hierarchy(cfs_rq_of(se)))
3038 3039 3040 3041 3042 3043 3044 3045 3046 3047
		return false;

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

	yield_task_fair(rq);

	return true;
}

3048
#ifdef CONFIG_SMP
3049 3050 3051 3052
/**************************************************
 * Fair scheduling class load-balancing methods:
 */

3053 3054
static unsigned long __read_mostly max_load_balance_interval = HZ/10;

3055
#define LBF_ALL_PINNED	0x01
3056
#define LBF_NEED_BREAK	0x02
3057
#define LBF_SOME_PINNED 0x04
3058 3059 3060 3061 3062

struct lb_env {
	struct sched_domain	*sd;

	struct rq		*src_rq;
3063
	int			src_cpu;
3064 3065 3066 3067

	int			dst_cpu;
	struct rq		*dst_rq;

3068 3069
	struct cpumask		*dst_grpmask;
	int			new_dst_cpu;
3070
	enum cpu_idle_type	idle;
3071
	long			imbalance;
3072 3073 3074
	/* The set of CPUs under consideration for load-balancing */
	struct cpumask		*cpus;

3075
	unsigned int		flags;
3076 3077 3078 3079

	unsigned int		loop;
	unsigned int		loop_break;
	unsigned int		loop_max;
3080 3081
};

3082
/*
3083
 * move_task - move a task from one runqueue to another runqueue.
3084 3085
 * Both runqueues must be locked.
 */
3086
static void move_task(struct task_struct *p, struct lb_env *env)
3087
{
3088 3089 3090 3091
	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);
3092 3093
}

3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125
/*
 * 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;
}

3126 3127 3128 3129
/*
 * can_migrate_task - may task p from runqueue rq be migrated to this_cpu?
 */
static
3130
int can_migrate_task(struct task_struct *p, struct lb_env *env)
3131 3132 3133 3134 3135 3136 3137 3138
{
	int tsk_cache_hot = 0;
	/*
	 * We do not migrate tasks that are:
	 * 1) running (obviously), or
	 * 2) cannot be migrated to this CPU due to cpus_allowed, or
	 * 3) are cache-hot on their current CPU.
	 */
3139
	if (!cpumask_test_cpu(env->dst_cpu, tsk_cpus_allowed(p))) {
3140 3141
		int new_dst_cpu;

3142
		schedstat_inc(p, se.statistics.nr_failed_migrations_affine);
3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160

		/*
		 * 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;

		new_dst_cpu = cpumask_first_and(env->dst_grpmask,
						tsk_cpus_allowed(p));
		if (new_dst_cpu < nr_cpu_ids) {
			env->flags |= LBF_SOME_PINNED;
			env->new_dst_cpu = new_dst_cpu;
		}
3161 3162
		return 0;
	}
3163 3164

	/* Record that we found atleast one task that could run on dst_cpu */
3165
	env->flags &= ~LBF_ALL_PINNED;
3166

3167
	if (task_running(env->src_rq, p)) {
3168
		schedstat_inc(p, se.statistics.nr_failed_migrations_running);
3169 3170 3171 3172 3173 3174 3175 3176 3177
		return 0;
	}

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

3178
	tsk_cache_hot = task_hot(p, env->src_rq->clock_task, env->sd);
3179
	if (!tsk_cache_hot ||
3180
		env->sd->nr_balance_failed > env->sd->cache_nice_tries) {
3181 3182
#ifdef CONFIG_SCHEDSTATS
		if (tsk_cache_hot) {
3183
			schedstat_inc(env->sd, lb_hot_gained[env->idle]);
3184
			schedstat_inc(p, se.statistics.nr_forced_migrations);
3185 3186 3187 3188 3189 3190
		}
#endif
		return 1;
	}

	if (tsk_cache_hot) {
3191
		schedstat_inc(p, se.statistics.nr_failed_migrations_hot);
3192 3193 3194 3195 3196
		return 0;
	}
	return 1;
}

3197 3198 3199 3200 3201 3202 3203
/*
 * 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.
 */
3204
static int move_one_task(struct lb_env *env)
3205 3206 3207
{
	struct task_struct *p, *n;

3208 3209 3210
	list_for_each_entry_safe(p, n, &env->src_rq->cfs_tasks, se.group_node) {
		if (throttled_lb_pair(task_group(p), env->src_rq->cpu, env->dst_cpu))
			continue;
3211

3212 3213
		if (!can_migrate_task(p, env))
			continue;
3214

3215 3216 3217 3218 3219 3220 3221 3222
		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;
3223 3224 3225 3226
	}
	return 0;
}

3227 3228
static unsigned long task_h_load(struct task_struct *p);

3229 3230
static const unsigned int sched_nr_migrate_break = 32;

3231
/*
3232
 * move_tasks tries to move up to imbalance weighted load from busiest to
3233 3234 3235 3236 3237 3238
 * 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)
3239
{
3240 3241
	struct list_head *tasks = &env->src_rq->cfs_tasks;
	struct task_struct *p;
3242 3243
	unsigned long load;
	int pulled = 0;
3244

3245
	if (env->imbalance <= 0)
3246
		return 0;
3247

3248 3249
	while (!list_empty(tasks)) {
		p = list_first_entry(tasks, struct task_struct, se.group_node);
3250

3251 3252
		env->loop++;
		/* We've more or less seen every task there is, call it quits */
3253
		if (env->loop > env->loop_max)
3254
			break;
3255 3256

		/* take a breather every nr_migrate tasks */
3257
		if (env->loop > env->loop_break) {
3258
			env->loop_break += sched_nr_migrate_break;
3259
			env->flags |= LBF_NEED_BREAK;
3260
			break;
3261
		}
3262

3263
		if (throttled_lb_pair(task_group(p), env->src_cpu, env->dst_cpu))
3264 3265 3266
			goto next;

		load = task_h_load(p);
3267

3268
		if (sched_feat(LB_MIN) && load < 16 && !env->sd->nr_balance_failed)
3269 3270
			goto next;

3271
		if ((load / 2) > env->imbalance)
3272
			goto next;
3273

3274 3275
		if (!can_migrate_task(p, env))
			goto next;
3276

3277
		move_task(p, env);
3278
		pulled++;
3279
		env->imbalance -= load;
3280 3281

#ifdef CONFIG_PREEMPT
3282 3283 3284 3285 3286
		/*
		 * NEWIDLE balancing is a source of latency, so preemptible
		 * kernels will stop after the first task is pulled to minimize
		 * the critical section.
		 */
3287
		if (env->idle == CPU_NEWLY_IDLE)
3288
			break;
3289 3290
#endif

3291 3292 3293 3294
		/*
		 * We only want to steal up to the prescribed amount of
		 * weighted load.
		 */
3295
		if (env->imbalance <= 0)
3296
			break;
3297 3298 3299

		continue;
next:
3300
		list_move_tail(&p->se.group_node, tasks);
3301
	}
3302

3303
	/*
3304 3305 3306
	 * 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().
3307
	 */
3308
	schedstat_add(env->sd, lb_gained[env->idle], pulled);
3309

3310
	return pulled;
3311 3312
}

P
Peter Zijlstra 已提交
3313
#ifdef CONFIG_FAIR_GROUP_SCHED
3314 3315 3316
/*
 * update tg->load_weight by folding this cpu's load_avg
 */
3317
static int update_shares_cpu(struct task_group *tg, int cpu)
3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331
{
	struct cfs_rq *cfs_rq;
	unsigned long flags;
	struct rq *rq;

	if (!tg->se[cpu])
		return 0;

	rq = cpu_rq(cpu);
	cfs_rq = tg->cfs_rq[cpu];

	raw_spin_lock_irqsave(&rq->lock, flags);

	update_rq_clock(rq);
3332
	update_cfs_load(cfs_rq, 1);
3333 3334 3335 3336 3337

	/*
	 * We need to update shares after updating tg->load_weight in
	 * order to adjust the weight of groups with long running tasks.
	 */
3338
	update_cfs_shares(cfs_rq);
3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350

	raw_spin_unlock_irqrestore(&rq->lock, flags);

	return 0;
}

static void update_shares(int cpu)
{
	struct cfs_rq *cfs_rq;
	struct rq *rq = cpu_rq(cpu);

	rcu_read_lock();
3351 3352 3353 3354
	/*
	 * Iterates the task_group tree in a bottom up fashion, see
	 * list_add_leaf_cfs_rq() for details.
	 */
3355 3356 3357 3358 3359
	for_each_leaf_cfs_rq(rq, cfs_rq) {
		/* throttled entities do not contribute to load */
		if (throttled_hierarchy(cfs_rq))
			continue;

3360
		update_shares_cpu(cfs_rq->tg, cpu);
3361
	}
3362 3363 3364
	rcu_read_unlock();
}

3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389
/*
 * 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)
{
3390 3391 3392 3393 3394 3395 3396 3397
	struct rq *rq = cpu_rq(cpu);
	unsigned long now = jiffies;

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

	rq->h_load_throttle = now;

3398
	rcu_read_lock();
3399
	walk_tg_tree(tg_load_down, tg_nop, (void *)cpu);
3400
	rcu_read_unlock();
3401 3402
}

3403
static unsigned long task_h_load(struct task_struct *p)
P
Peter Zijlstra 已提交
3404
{
3405 3406
	struct cfs_rq *cfs_rq = task_cfs_rq(p);
	unsigned long load;
P
Peter Zijlstra 已提交
3407

3408 3409
	load = p->se.load.weight;
	load = div_u64(load * cfs_rq->h_load, cfs_rq->load.weight + 1);
P
Peter Zijlstra 已提交
3410

3411
	return load;
P
Peter Zijlstra 已提交
3412 3413
}
#else
3414 3415 3416 3417
static inline void update_shares(int cpu)
{
}

3418
static inline void update_h_load(long cpu)
P
Peter Zijlstra 已提交
3419 3420 3421
{
}

3422
static unsigned long task_h_load(struct task_struct *p)
3423
{
3424
	return p->se.load.weight;
3425
}
P
Peter Zijlstra 已提交
3426
#endif
3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443

/********** 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;
3444
	unsigned long this_has_capacity;
3445
	unsigned int  this_idle_cpus;
3446 3447

	/* Statistics of the busiest group */
3448
	unsigned int  busiest_idle_cpus;
3449 3450 3451
	unsigned long max_load;
	unsigned long busiest_load_per_task;
	unsigned long busiest_nr_running;
3452
	unsigned long busiest_group_capacity;
3453
	unsigned long busiest_has_capacity;
3454
	unsigned int  busiest_group_weight;
3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467

	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;
3468 3469
	unsigned long idle_cpus;
	unsigned long group_weight;
3470
	int group_imb; /* Is there an imbalance in the group ? */
3471
	int group_has_capacity; /* Is there extra capacity in the group? */
3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501
};

/**
 * 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;
}

unsigned long default_scale_freq_power(struct sched_domain *sd, int cpu)
{
3502
	return SCHED_POWER_SCALE;
3503 3504 3505 3506 3507 3508 3509 3510 3511
}

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

unsigned long default_scale_smt_power(struct sched_domain *sd, int cpu)
{
3512
	unsigned long weight = sd->span_weight;
3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527
	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);
}

unsigned long scale_rt_power(int cpu)
{
	struct rq *rq = cpu_rq(cpu);
3528
	u64 total, available, age_stamp, avg;
3529

3530 3531 3532 3533 3534 3535 3536 3537
	/*
	 * 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);

	total = sched_avg_period() + (rq->clock - age_stamp);
3538

3539
	if (unlikely(total < avg)) {
3540 3541 3542
		/* Ensures that power won't end up being negative */
		available = 0;
	} else {
3543
		available = total - avg;
3544
	}
3545

3546 3547
	if (unlikely((s64)total < SCHED_POWER_SCALE))
		total = SCHED_POWER_SCALE;
3548

3549
	total >>= SCHED_POWER_SHIFT;
3550 3551 3552 3553 3554 3555

	return div_u64(available, total);
}

static void update_cpu_power(struct sched_domain *sd, int cpu)
{
3556
	unsigned long weight = sd->span_weight;
3557
	unsigned long power = SCHED_POWER_SCALE;
3558 3559 3560 3561 3562 3563 3564 3565
	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);

3566
		power >>= SCHED_POWER_SHIFT;
3567 3568
	}

3569
	sdg->sgp->power_orig = power;
3570 3571 3572 3573 3574 3575

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

3576
	power >>= SCHED_POWER_SHIFT;
3577

3578
	power *= scale_rt_power(cpu);
3579
	power >>= SCHED_POWER_SHIFT;
3580 3581 3582 3583

	if (!power)
		power = 1;

3584
	cpu_rq(cpu)->cpu_power = power;
3585
	sdg->sgp->power = power;
3586 3587
}

3588
void update_group_power(struct sched_domain *sd, int cpu)
3589 3590 3591 3592
{
	struct sched_domain *child = sd->child;
	struct sched_group *group, *sdg = sd->groups;
	unsigned long power;
3593 3594 3595 3596 3597
	unsigned long interval;

	interval = msecs_to_jiffies(sd->balance_interval);
	interval = clamp(interval, 1UL, max_load_balance_interval);
	sdg->sgp->next_update = jiffies + interval;
3598 3599 3600 3601 3602 3603 3604 3605

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

	power = 0;

P
Peter Zijlstra 已提交
3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625
	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);
	}
3626

3627
	sdg->sgp->power_orig = sdg->sgp->power = power;
3628 3629
}

3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640
/*
 * 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)
{
	/*
3641
	 * Only siblings can have significantly less than SCHED_POWER_SCALE
3642
	 */
P
Peter Zijlstra 已提交
3643
	if (!(sd->flags & SD_SHARE_CPUPOWER))
3644 3645 3646 3647 3648
		return 0;

	/*
	 * If ~90% of the cpu_power is still there, we're good.
	 */
3649
	if (group->sgp->power * 32 > group->sgp->power_orig * 29)
3650 3651 3652 3653 3654
		return 1;

	return 0;
}

3655 3656
/**
 * update_sg_lb_stats - Update sched_group's statistics for load balancing.
3657
 * @env: The load balancing environment.
3658 3659 3660 3661 3662 3663
 * @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.
 */
3664 3665
static inline void update_sg_lb_stats(struct lb_env *env,
			struct sched_group *group, int load_idx,
3666
			int local_group, int *balance, struct sg_lb_stats *sgs)
3667
{
3668 3669
	unsigned long nr_running, max_nr_running, min_nr_running;
	unsigned long load, max_cpu_load, min_cpu_load;
3670
	unsigned int balance_cpu = -1, first_idle_cpu = 0;
3671
	unsigned long avg_load_per_task = 0;
3672
	int i;
3673

3674
	if (local_group)
P
Peter Zijlstra 已提交
3675
		balance_cpu = group_balance_cpu(group);
3676 3677 3678 3679

	/* Tally up the load of all CPUs in the group */
	max_cpu_load = 0;
	min_cpu_load = ~0UL;
3680
	max_nr_running = 0;
3681
	min_nr_running = ~0UL;
3682

3683
	for_each_cpu_and(i, sched_group_cpus(group), env->cpus) {
3684 3685
		struct rq *rq = cpu_rq(i);

3686 3687
		nr_running = rq->nr_running;

3688 3689
		/* Bias balancing toward cpus of our domain */
		if (local_group) {
P
Peter Zijlstra 已提交
3690 3691
			if (idle_cpu(i) && !first_idle_cpu &&
					cpumask_test_cpu(i, sched_group_mask(group))) {
3692
				first_idle_cpu = 1;
3693 3694
				balance_cpu = i;
			}
3695 3696

			load = target_load(i, load_idx);
3697 3698
		} else {
			load = source_load(i, load_idx);
3699
			if (load > max_cpu_load)
3700 3701 3702
				max_cpu_load = load;
			if (min_cpu_load > load)
				min_cpu_load = load;
3703 3704 3705 3706 3707

			if (nr_running > max_nr_running)
				max_nr_running = nr_running;
			if (min_nr_running > nr_running)
				min_nr_running = nr_running;
3708 3709 3710
		}

		sgs->group_load += load;
3711
		sgs->sum_nr_running += nr_running;
3712
		sgs->sum_weighted_load += weighted_cpuload(i);
3713 3714
		if (idle_cpu(i))
			sgs->idle_cpus++;
3715 3716 3717 3718 3719 3720 3721 3722
	}

	/*
	 * 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.
	 */
3723
	if (local_group) {
3724
		if (env->idle != CPU_NEWLY_IDLE) {
3725
			if (balance_cpu != env->dst_cpu) {
3726 3727 3728
				*balance = 0;
				return;
			}
3729
			update_group_power(env->sd, env->dst_cpu);
3730
		} else if (time_after_eq(jiffies, group->sgp->next_update))
3731
			update_group_power(env->sd, env->dst_cpu);
3732 3733 3734
	}

	/* Adjust by relative CPU power of the group */
3735
	sgs->avg_load = (sgs->group_load*SCHED_POWER_SCALE) / group->sgp->power;
3736 3737 3738

	/*
	 * Consider the group unbalanced when the imbalance is larger
P
Peter Zijlstra 已提交
3739
	 * than the average weight of a task.
3740 3741 3742 3743 3744 3745
	 *
	 * 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?
	 */
3746 3747
	if (sgs->sum_nr_running)
		avg_load_per_task = sgs->sum_weighted_load / sgs->sum_nr_running;
3748

3749 3750
	if ((max_cpu_load - min_cpu_load) >= avg_load_per_task &&
	    (max_nr_running - min_nr_running) > 1)
3751 3752
		sgs->group_imb = 1;

3753
	sgs->group_capacity = DIV_ROUND_CLOSEST(group->sgp->power,
3754
						SCHED_POWER_SCALE);
3755
	if (!sgs->group_capacity)
3756
		sgs->group_capacity = fix_small_capacity(env->sd, group);
3757
	sgs->group_weight = group->group_weight;
3758 3759 3760

	if (sgs->group_capacity > sgs->sum_nr_running)
		sgs->group_has_capacity = 1;
3761 3762
}

3763 3764
/**
 * update_sd_pick_busiest - return 1 on busiest group
3765
 * @env: The load balancing environment.
3766 3767
 * @sds: sched_domain statistics
 * @sg: sched_group candidate to be checked for being the busiest
3768
 * @sgs: sched_group statistics
3769 3770 3771 3772
 *
 * Determine if @sg is a busier group than the previously selected
 * busiest group.
 */
3773
static bool update_sd_pick_busiest(struct lb_env *env,
3774 3775
				   struct sd_lb_stats *sds,
				   struct sched_group *sg,
3776
				   struct sg_lb_stats *sgs)
3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791
{
	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.
	 */
3792 3793
	if ((env->sd->flags & SD_ASYM_PACKING) && sgs->sum_nr_running &&
	    env->dst_cpu < group_first_cpu(sg)) {
3794 3795 3796 3797 3798 3799 3800 3801 3802 3803
		if (!sds->busiest)
			return true;

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

	return false;
}

3804
/**
3805
 * update_sd_lb_stats - Update sched_domain's statistics for load balancing.
3806
 * @env: The load balancing environment.
3807 3808 3809
 * @balance: Should we balance.
 * @sds: variable to hold the statistics for this sched_domain.
 */
3810
static inline void update_sd_lb_stats(struct lb_env *env,
3811
					int *balance, struct sd_lb_stats *sds)
3812
{
3813 3814
	struct sched_domain *child = env->sd->child;
	struct sched_group *sg = env->sd->groups;
3815 3816 3817 3818 3819 3820
	struct sg_lb_stats sgs;
	int load_idx, prefer_sibling = 0;

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

3821
	load_idx = get_sd_load_idx(env->sd, env->idle);
3822 3823 3824 3825

	do {
		int local_group;

3826
		local_group = cpumask_test_cpu(env->dst_cpu, sched_group_cpus(sg));
3827
		memset(&sgs, 0, sizeof(sgs));
3828
		update_sg_lb_stats(env, sg, load_idx, local_group, balance, &sgs);
3829

P
Peter Zijlstra 已提交
3830
		if (local_group && !(*balance))
3831 3832 3833
			return;

		sds->total_load += sgs.group_load;
3834
		sds->total_pwr += sg->sgp->power;
3835 3836 3837

		/*
		 * In case the child domain prefers tasks go to siblings
3838
		 * first, lower the sg capacity to one so that we'll try
3839 3840 3841 3842 3843 3844
		 * 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).
3845
		 */
3846
		if (prefer_sibling && !local_group && sds->this_has_capacity)
3847 3848 3849 3850
			sgs.group_capacity = min(sgs.group_capacity, 1UL);

		if (local_group) {
			sds->this_load = sgs.avg_load;
3851
			sds->this = sg;
3852 3853
			sds->this_nr_running = sgs.sum_nr_running;
			sds->this_load_per_task = sgs.sum_weighted_load;
3854
			sds->this_has_capacity = sgs.group_has_capacity;
3855
			sds->this_idle_cpus = sgs.idle_cpus;
3856
		} else if (update_sd_pick_busiest(env, sds, sg, &sgs)) {
3857
			sds->max_load = sgs.avg_load;
3858
			sds->busiest = sg;
3859
			sds->busiest_nr_running = sgs.sum_nr_running;
3860
			sds->busiest_idle_cpus = sgs.idle_cpus;
3861
			sds->busiest_group_capacity = sgs.group_capacity;
3862
			sds->busiest_load_per_task = sgs.sum_weighted_load;
3863
			sds->busiest_has_capacity = sgs.group_has_capacity;
3864
			sds->busiest_group_weight = sgs.group_weight;
3865 3866 3867
			sds->group_imb = sgs.group_imb;
		}

3868
		sg = sg->next;
3869
	} while (sg != env->sd->groups);
3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888
}

/**
 * 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.
 *
3889 3890 3891
 * Returns 1 when packing is required and a task should be moved to
 * this CPU.  The amount of the imbalance is returned in *imbalance.
 *
3892
 * @env: The load balancing environment.
3893 3894
 * @sds: Statistics of the sched_domain which is to be packed
 */
3895
static int check_asym_packing(struct lb_env *env, struct sd_lb_stats *sds)
3896 3897 3898
{
	int busiest_cpu;

3899
	if (!(env->sd->flags & SD_ASYM_PACKING))
3900 3901 3902 3903 3904 3905
		return 0;

	if (!sds->busiest)
		return 0;

	busiest_cpu = group_first_cpu(sds->busiest);
3906
	if (env->dst_cpu > busiest_cpu)
3907 3908
		return 0;

3909 3910 3911
	env->imbalance = DIV_ROUND_CLOSEST(
		sds->max_load * sds->busiest->sgp->power, SCHED_POWER_SCALE);

3912
	return 1;
3913 3914 3915 3916 3917 3918
}

/**
 * fix_small_imbalance - Calculate the minor imbalance that exists
 *			amongst the groups of a sched_domain, during
 *			load balancing.
3919
 * @env: The load balancing environment.
3920 3921
 * @sds: Statistics of the sched_domain whose imbalance is to be calculated.
 */
3922 3923
static inline
void fix_small_imbalance(struct lb_env *env, struct sd_lb_stats *sds)
3924 3925 3926
{
	unsigned long tmp, pwr_now = 0, pwr_move = 0;
	unsigned int imbn = 2;
3927
	unsigned long scaled_busy_load_per_task;
3928 3929 3930 3931 3932 3933

	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;
3934
	} else {
3935
		sds->this_load_per_task =
3936 3937
			cpu_avg_load_per_task(env->dst_cpu);
	}
3938

3939
	scaled_busy_load_per_task = sds->busiest_load_per_task
3940
					 * SCHED_POWER_SCALE;
3941
	scaled_busy_load_per_task /= sds->busiest->sgp->power;
3942 3943 3944

	if (sds->max_load - sds->this_load + scaled_busy_load_per_task >=
			(scaled_busy_load_per_task * imbn)) {
3945
		env->imbalance = sds->busiest_load_per_task;
3946 3947 3948 3949 3950 3951 3952 3953 3954
		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.
	 */

3955
	pwr_now += sds->busiest->sgp->power *
3956
			min(sds->busiest_load_per_task, sds->max_load);
3957
	pwr_now += sds->this->sgp->power *
3958
			min(sds->this_load_per_task, sds->this_load);
3959
	pwr_now /= SCHED_POWER_SCALE;
3960 3961

	/* Amount of load we'd subtract */
3962
	tmp = (sds->busiest_load_per_task * SCHED_POWER_SCALE) /
3963
		sds->busiest->sgp->power;
3964
	if (sds->max_load > tmp)
3965
		pwr_move += sds->busiest->sgp->power *
3966 3967 3968
			min(sds->busiest_load_per_task, sds->max_load - tmp);

	/* Amount of load we'd add */
3969
	if (sds->max_load * sds->busiest->sgp->power <
3970
		sds->busiest_load_per_task * SCHED_POWER_SCALE)
3971 3972
		tmp = (sds->max_load * sds->busiest->sgp->power) /
			sds->this->sgp->power;
3973
	else
3974
		tmp = (sds->busiest_load_per_task * SCHED_POWER_SCALE) /
3975 3976
			sds->this->sgp->power;
	pwr_move += sds->this->sgp->power *
3977
			min(sds->this_load_per_task, sds->this_load + tmp);
3978
	pwr_move /= SCHED_POWER_SCALE;
3979 3980 3981

	/* Move if we gain throughput */
	if (pwr_move > pwr_now)
3982
		env->imbalance = sds->busiest_load_per_task;
3983 3984 3985 3986 3987
}

/**
 * calculate_imbalance - Calculate the amount of imbalance present within the
 *			 groups of a given sched_domain during load balance.
3988
 * @env: load balance environment
3989 3990
 * @sds: statistics of the sched_domain whose imbalance is to be calculated.
 */
3991
static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *sds)
3992
{
3993 3994 3995 3996 3997 3998 3999 4000
	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);
	}

4001 4002 4003 4004 4005 4006
	/*
	 * 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) {
4007 4008
		env->imbalance = 0;
		return fix_small_imbalance(env, sds);
4009 4010
	}

4011 4012 4013 4014 4015 4016 4017
	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);

4018
		load_above_capacity *= (SCHED_LOAD_SCALE * SCHED_POWER_SCALE);
4019

4020
		load_above_capacity /= sds->busiest->sgp->power;
4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033
	}

	/*
	 * 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);
4034 4035

	/* How much load to actually move to equalise the imbalance */
4036
	env->imbalance = min(max_pull * sds->busiest->sgp->power,
4037
		(sds->avg_load - sds->this_load) * sds->this->sgp->power)
4038
			/ SCHED_POWER_SCALE;
4039 4040 4041

	/*
	 * if *imbalance is less than the average load per runnable task
L
Lucas De Marchi 已提交
4042
	 * there is no guarantee that any tasks will be moved so we'll have
4043 4044 4045
	 * a think about bumping its value to force at least one task to be
	 * moved
	 */
4046 4047
	if (env->imbalance < sds->busiest_load_per_task)
		return fix_small_imbalance(env, sds);
4048 4049

}
4050

4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062
/******* 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.
 *
4063
 * @env: The load balancing environment.
4064 4065 4066 4067 4068 4069 4070 4071 4072
 * @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 *
4073
find_busiest_group(struct lb_env *env, int *balance)
4074 4075 4076 4077 4078 4079 4080 4081 4082
{
	struct sd_lb_stats sds;

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

	/*
	 * Compute the various statistics relavent for load balancing at
	 * this level.
	 */
4083
	update_sd_lb_stats(env, balance, &sds);
4084

4085 4086 4087
	/*
	 * this_cpu is not the appropriate cpu to perform load balancing at
	 * this level.
4088
	 */
P
Peter Zijlstra 已提交
4089
	if (!(*balance))
4090 4091
		goto ret;

4092 4093
	if ((env->idle == CPU_IDLE || env->idle == CPU_NEWLY_IDLE) &&
	    check_asym_packing(env, &sds))
4094 4095
		return sds.busiest;

4096
	/* There is no busy sibling group to pull tasks from */
4097 4098 4099
	if (!sds.busiest || sds.busiest_nr_running == 0)
		goto out_balanced;

4100
	sds.avg_load = (SCHED_POWER_SCALE * sds.total_load) / sds.total_pwr;
4101

P
Peter Zijlstra 已提交
4102 4103 4104 4105 4106 4107 4108 4109
	/*
	 * 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;

4110
	/* SD_BALANCE_NEWIDLE trumps SMP nice when underutilized */
4111
	if (env->idle == CPU_NEWLY_IDLE && sds.this_has_capacity &&
4112 4113 4114
			!sds.busiest_has_capacity)
		goto force_balance;

4115 4116 4117 4118
	/*
	 * If the local group is more busy than the selected busiest group
	 * don't try and pull any tasks.
	 */
4119 4120 4121
	if (sds.this_load >= sds.max_load)
		goto out_balanced;

4122 4123 4124 4125
	/*
	 * Don't pull any tasks if this group is already above the domain
	 * average load.
	 */
4126 4127 4128
	if (sds.this_load >= sds.avg_load)
		goto out_balanced;

4129
	if (env->idle == CPU_IDLE) {
4130 4131 4132 4133 4134 4135
		/*
		 * 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.
		 */
4136
		if ((sds.this_idle_cpus <= sds.busiest_idle_cpus + 1) &&
4137 4138
		    sds.busiest_nr_running <= sds.busiest_group_weight)
			goto out_balanced;
4139 4140 4141 4142 4143
	} else {
		/*
		 * In the CPU_NEWLY_IDLE, CPU_NOT_IDLE cases, use
		 * imbalance_pct to be conservative.
		 */
4144
		if (100 * sds.max_load <= env->sd->imbalance_pct * sds.this_load)
4145
			goto out_balanced;
4146
	}
4147

4148
force_balance:
4149
	/* Looks like there is an imbalance. Compute it */
4150
	calculate_imbalance(env, &sds);
4151 4152 4153 4154
	return sds.busiest;

out_balanced:
ret:
4155
	env->imbalance = 0;
4156 4157 4158 4159 4160 4161
	return NULL;
}

/*
 * find_busiest_queue - find the busiest runqueue among the cpus in group.
 */
4162
static struct rq *find_busiest_queue(struct lb_env *env,
4163
				     struct sched_group *group)
4164 4165 4166 4167 4168 4169 4170
{
	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);
4171 4172
		unsigned long capacity = DIV_ROUND_CLOSEST(power,
							   SCHED_POWER_SCALE);
4173 4174
		unsigned long wl;

4175
		if (!capacity)
4176
			capacity = fix_small_capacity(env->sd, group);
4177

4178
		if (!cpumask_test_cpu(i, env->cpus))
4179 4180 4181
			continue;

		rq = cpu_rq(i);
4182
		wl = weighted_cpuload(i);
4183

4184 4185 4186 4187
		/*
		 * When comparing with imbalance, use weighted_cpuload()
		 * which is not scaled with the cpu power.
		 */
4188
		if (capacity && rq->nr_running == 1 && wl > env->imbalance)
4189 4190
			continue;

4191 4192 4193 4194 4195 4196
		/*
		 * 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.
		 */
4197
		wl = (wl * SCHED_POWER_SCALE) / power;
4198

4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210 4211 4212 4213 4214
		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. */
4215
DEFINE_PER_CPU(cpumask_var_t, load_balance_tmpmask);
4216

4217
static int need_active_balance(struct lb_env *env)
4218
{
4219 4220 4221
	struct sched_domain *sd = env->sd;

	if (env->idle == CPU_NEWLY_IDLE) {
4222 4223 4224 4225 4226 4227

		/*
		 * ASYM_PACKING needs to force migrate tasks from busy but
		 * higher numbered CPUs in order to pack all tasks in the
		 * lowest numbered CPUs.
		 */
4228
		if ((sd->flags & SD_ASYM_PACKING) && env->src_cpu > env->dst_cpu)
4229
			return 1;
4230 4231 4232 4233 4234
	}

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

4235 4236
static int active_load_balance_cpu_stop(void *data);

4237 4238 4239 4240 4241 4242 4243 4244
/*
 * 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)
{
4245 4246
	int ld_moved, cur_ld_moved, active_balance = 0;
	int lb_iterations, max_lb_iterations;
4247 4248 4249 4250 4251
	struct sched_group *group;
	struct rq *busiest;
	unsigned long flags;
	struct cpumask *cpus = __get_cpu_var(load_balance_tmpmask);

4252 4253
	struct lb_env env = {
		.sd		= sd,
4254 4255
		.dst_cpu	= this_cpu,
		.dst_rq		= this_rq,
4256
		.dst_grpmask    = sched_group_cpus(sd->groups),
4257
		.idle		= idle,
4258
		.loop_break	= sched_nr_migrate_break,
4259
		.cpus		= cpus,
4260 4261
	};

4262
	cpumask_copy(cpus, cpu_active_mask);
4263
	max_lb_iterations = cpumask_weight(env.dst_grpmask);
4264 4265 4266 4267

	schedstat_inc(sd, lb_count[idle]);

redo:
4268
	group = find_busiest_group(&env, balance);
4269 4270 4271 4272 4273 4274 4275 4276 4277

	if (*balance == 0)
		goto out_balanced;

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

4278
	busiest = find_busiest_queue(&env, group);
4279 4280 4281 4282 4283 4284 4285
	if (!busiest) {
		schedstat_inc(sd, lb_nobusyq[idle]);
		goto out_balanced;
	}

	BUG_ON(busiest == this_rq);

4286
	schedstat_add(sd, lb_imbalance[idle], env.imbalance);
4287 4288

	ld_moved = 0;
4289
	lb_iterations = 1;
4290 4291 4292 4293 4294 4295 4296
	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.
		 */
4297
		env.flags |= LBF_ALL_PINNED;
4298 4299 4300
		env.src_cpu   = busiest->cpu;
		env.src_rq    = busiest;
		env.loop_max  = min(sysctl_sched_nr_migrate, busiest->nr_running);
4301

4302
		update_h_load(env.src_cpu);
4303
more_balance:
4304 4305
		local_irq_save(flags);
		double_rq_lock(this_rq, busiest);
4306 4307 4308 4309 4310 4311 4312

		/*
		 * 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;
4313 4314 4315
		double_rq_unlock(this_rq, busiest);
		local_irq_restore(flags);

4316 4317 4318 4319 4320
		if (env.flags & LBF_NEED_BREAK) {
			env.flags &= ~LBF_NEED_BREAK;
			goto more_balance;
		}

4321 4322 4323
		/*
		 * some other cpu did the load balance for us.
		 */
4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352 4353 4354 4355 4356 4357 4358 4359 4360
		if (cur_ld_moved && env.dst_cpu != smp_processor_id())
			resched_cpu(env.dst_cpu);

		/*
		 * 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.
		 */
		if ((env.flags & LBF_SOME_PINNED) && env.imbalance > 0 &&
				lb_iterations++ < max_lb_iterations) {

			this_rq		 = cpu_rq(env.new_dst_cpu);
			env.dst_rq	 = this_rq;
			env.dst_cpu	 = env.new_dst_cpu;
			env.flags	&= ~LBF_SOME_PINNED;
			env.loop	 = 0;
			env.loop_break	 = sched_nr_migrate_break;
			/*
			 * Go back to "more_balance" rather than "redo" since we
			 * need to continue with same src_cpu.
			 */
			goto more_balance;
		}
4361 4362

		/* All tasks on this runqueue were pinned by CPU affinity */
4363
		if (unlikely(env.flags & LBF_ALL_PINNED)) {
4364
			cpumask_clear_cpu(cpu_of(busiest), cpus);
4365 4366 4367
			if (!cpumask_empty(cpus)) {
				env.loop = 0;
				env.loop_break = sched_nr_migrate_break;
4368
				goto redo;
4369
			}
4370 4371 4372 4373 4374 4375
			goto out_balanced;
		}
	}

	if (!ld_moved) {
		schedstat_inc(sd, lb_failed[idle]);
4376 4377 4378 4379 4380 4381 4382 4383
		/*
		 * 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++;
4384

4385
		if (need_active_balance(&env)) {
4386 4387
			raw_spin_lock_irqsave(&busiest->lock, flags);

4388 4389 4390
			/* don't kick the active_load_balance_cpu_stop,
			 * if the curr task on busiest cpu can't be
			 * moved to this_cpu
4391 4392
			 */
			if (!cpumask_test_cpu(this_cpu,
4393
					tsk_cpus_allowed(busiest->curr))) {
4394 4395
				raw_spin_unlock_irqrestore(&busiest->lock,
							    flags);
4396
				env.flags |= LBF_ALL_PINNED;
4397 4398 4399
				goto out_one_pinned;
			}

4400 4401 4402 4403 4404
			/*
			 * ->active_balance synchronizes accesses to
			 * ->active_balance_work.  Once set, it's cleared
			 * only after active load balance is finished.
			 */
4405 4406 4407 4408 4409 4410
			if (!busiest->active_balance) {
				busiest->active_balance = 1;
				busiest->push_cpu = this_cpu;
				active_balance = 1;
			}
			raw_spin_unlock_irqrestore(&busiest->lock, flags);
4411

4412
			if (active_balance) {
4413 4414 4415
				stop_one_cpu_nowait(cpu_of(busiest),
					active_load_balance_cpu_stop, busiest,
					&busiest->active_balance_work);
4416
			}
4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449

			/*
			 * 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 */
4450
	if (((env.flags & LBF_ALL_PINNED) &&
4451
			sd->balance_interval < MAX_PINNED_INTERVAL) ||
4452 4453 4454
			(sd->balance_interval < sd->max_interval))
		sd->balance_interval *= 2;

4455
	ld_moved = 0;
4456 4457 4458 4459 4460 4461 4462 4463
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.
 */
4464
void idle_balance(int this_cpu, struct rq *this_rq)
4465 4466 4467 4468 4469 4470 4471 4472 4473 4474
{
	struct sched_domain *sd;
	int pulled_task = 0;
	unsigned long next_balance = jiffies + HZ;

	this_rq->idle_stamp = this_rq->clock;

	if (this_rq->avg_idle < sysctl_sched_migration_cost)
		return;

4475 4476 4477 4478 4479
	/*
	 * Drop the rq->lock, but keep IRQ/preempt disabled.
	 */
	raw_spin_unlock(&this_rq->lock);

P
Paul Turner 已提交
4480
	update_shares(this_cpu);
4481
	rcu_read_lock();
4482 4483
	for_each_domain(this_cpu, sd) {
		unsigned long interval;
4484
		int balance = 1;
4485 4486 4487 4488

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

4489
		if (sd->flags & SD_BALANCE_NEWIDLE) {
4490
			/* If we've pulled tasks over stop searching: */
4491 4492 4493
			pulled_task = load_balance(this_cpu, this_rq,
						   sd, CPU_NEWLY_IDLE, &balance);
		}
4494 4495 4496 4497

		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
4498 4499
		if (pulled_task) {
			this_rq->idle_stamp = 0;
4500
			break;
N
Nikhil Rao 已提交
4501
		}
4502
	}
4503
	rcu_read_unlock();
4504 4505 4506

	raw_spin_lock(&this_rq->lock);

4507 4508 4509 4510 4511 4512 4513 4514 4515 4516
	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;
	}
}

/*
4517 4518 4519 4520
 * 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.
4521
 */
4522
static int active_load_balance_cpu_stop(void *data)
4523
{
4524 4525
	struct rq *busiest_rq = data;
	int busiest_cpu = cpu_of(busiest_rq);
4526
	int target_cpu = busiest_rq->push_cpu;
4527
	struct rq *target_rq = cpu_rq(target_cpu);
4528
	struct sched_domain *sd;
4529 4530 4531 4532 4533 4534 4535

	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;
4536 4537 4538

	/* Is there any task to move? */
	if (busiest_rq->nr_running <= 1)
4539
		goto out_unlock;
4540 4541 4542 4543 4544 4545 4546 4547 4548 4549 4550 4551

	/*
	 * 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. */
4552
	rcu_read_lock();
4553 4554 4555 4556 4557 4558 4559
	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)) {
4560 4561
		struct lb_env env = {
			.sd		= sd,
4562 4563 4564 4565
			.dst_cpu	= target_cpu,
			.dst_rq		= target_rq,
			.src_cpu	= busiest_rq->cpu,
			.src_rq		= busiest_rq,
4566 4567 4568
			.idle		= CPU_IDLE,
		};

4569 4570
		schedstat_inc(sd, alb_count);

4571
		if (move_one_task(&env))
4572 4573 4574 4575
			schedstat_inc(sd, alb_pushed);
		else
			schedstat_inc(sd, alb_failed);
	}
4576
	rcu_read_unlock();
4577
	double_unlock_balance(busiest_rq, target_rq);
4578 4579 4580 4581
out_unlock:
	busiest_rq->active_balance = 0;
	raw_spin_unlock_irq(&busiest_rq->lock);
	return 0;
4582 4583 4584
}

#ifdef CONFIG_NO_HZ
4585 4586 4587 4588 4589 4590
/*
 * 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.
 */
4591
static struct {
4592
	cpumask_var_t idle_cpus_mask;
4593
	atomic_t nr_cpus;
4594 4595
	unsigned long next_balance;     /* in jiffy units */
} nohz ____cacheline_aligned;
4596

4597
static inline int find_new_ilb(int call_cpu)
4598
{
4599
	int ilb = cpumask_first(nohz.idle_cpus_mask);
4600

4601 4602 4603 4604
	if (ilb < nr_cpu_ids && idle_cpu(ilb))
		return ilb;

	return nr_cpu_ids;
4605 4606
}

4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617
/*
 * 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++;

4618
	ilb_cpu = find_new_ilb(cpu);
4619

4620 4621
	if (ilb_cpu >= nr_cpu_ids)
		return;
4622

4623
	if (test_and_set_bit(NOHZ_BALANCE_KICK, nohz_flags(ilb_cpu)))
4624 4625 4626 4627 4628 4629 4630 4631
		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);
4632 4633 4634
	return;
}

4635 4636 4637 4638 4639 4640 4641 4642 4643
static inline void clear_nohz_tick_stopped(int cpu)
{
	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));
	}
}

4644 4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673
static inline void set_cpu_sd_state_busy(void)
{
	struct sched_domain *sd;
	int cpu = smp_processor_id();

	if (!test_bit(NOHZ_IDLE, nohz_flags(cpu)))
		return;
	clear_bit(NOHZ_IDLE, nohz_flags(cpu));

	rcu_read_lock();
	for_each_domain(cpu, sd)
		atomic_inc(&sd->groups->sgp->nr_busy_cpus);
	rcu_read_unlock();
}

void set_cpu_sd_state_idle(void)
{
	struct sched_domain *sd;
	int cpu = smp_processor_id();

	if (test_bit(NOHZ_IDLE, nohz_flags(cpu)))
		return;
	set_bit(NOHZ_IDLE, nohz_flags(cpu));

	rcu_read_lock();
	for_each_domain(cpu, sd)
		atomic_dec(&sd->groups->sgp->nr_busy_cpus);
	rcu_read_unlock();
}

4674
/*
4675 4676
 * This routine will record that this cpu is going idle with tick stopped.
 * This info will be used in performing idle load balancing in the future.
4677
 */
4678
void select_nohz_load_balancer(int stop_tick)
4679 4680 4681
{
	int cpu = smp_processor_id();

4682 4683 4684 4685 4686 4687
	/*
	 * If this cpu is going down, then nothing needs to be done.
	 */
	if (!cpu_active(cpu))
		return;

4688
	if (stop_tick) {
4689
		if (test_bit(NOHZ_TICK_STOPPED, nohz_flags(cpu)))
4690
			return;
4691

4692
		cpumask_set_cpu(cpu, nohz.idle_cpus_mask);
4693
		atomic_inc(&nohz.nr_cpus);
4694
		set_bit(NOHZ_TICK_STOPPED, nohz_flags(cpu));
4695
	}
4696
	return;
4697
}
4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709

static int __cpuinit sched_ilb_notifier(struct notifier_block *nfb,
					unsigned long action, void *hcpu)
{
	switch (action & ~CPU_TASKS_FROZEN) {
	case CPU_DYING:
		clear_nohz_tick_stopped(smp_processor_id());
		return NOTIFY_OK;
	default:
		return NOTIFY_DONE;
	}
}
4710 4711 4712 4713
#endif

static DEFINE_SPINLOCK(balancing);

4714 4715 4716 4717
/*
 * 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.
 */
4718
void update_max_interval(void)
4719 4720 4721 4722
{
	max_load_balance_interval = HZ*num_online_cpus()/10;
}

4723 4724 4725 4726 4727 4728 4729 4730 4731 4732 4733
/*
 * It checks each scheduling domain to see if it is due to be balanced,
 * and initiates a balancing operation if so.
 *
 * Balancing parameters are set up in arch_init_sched_domains.
 */
static void rebalance_domains(int cpu, enum cpu_idle_type idle)
{
	int balance = 1;
	struct rq *rq = cpu_rq(cpu);
	unsigned long interval;
4734
	struct sched_domain *sd;
4735 4736 4737 4738 4739
	/* Earliest time when we have to do rebalance again */
	unsigned long next_balance = jiffies + 60*HZ;
	int update_next_balance = 0;
	int need_serialize;

P
Peter Zijlstra 已提交
4740 4741
	update_shares(cpu);

4742
	rcu_read_lock();
4743 4744 4745 4746 4747 4748 4749 4750 4751 4752
	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);
4753
		interval = clamp(interval, 1UL, max_load_balance_interval);
4754 4755 4756 4757 4758 4759 4760 4761 4762 4763 4764 4765

		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)) {
				/*
				 * We've pulled tasks over so either we're no
4766
				 * longer idle.
4767 4768 4769 4770 4771 4772 4773 4774 4775 4776 4777 4778 4779 4780 4781 4782 4783 4784 4785 4786 4787
				 */
				idle = CPU_NOT_IDLE;
			}
			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;
	}
4788
	rcu_read_unlock();
4789 4790 4791 4792 4793 4794 4795 4796 4797 4798

	/*
	 * 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;
}

4799
#ifdef CONFIG_NO_HZ
4800
/*
4801
 * In CONFIG_NO_HZ case, the idle balance kickee will do the
4802 4803
 * rebalancing for all the cpus for whom scheduler ticks are stopped.
 */
4804 4805 4806 4807 4808 4809
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;

4810 4811 4812
	if (idle != CPU_IDLE ||
	    !test_bit(NOHZ_BALANCE_KICK, nohz_flags(this_cpu)))
		goto end;
4813 4814

	for_each_cpu(balance_cpu, nohz.idle_cpus_mask) {
4815
		if (balance_cpu == this_cpu || !idle_cpu(balance_cpu))
4816 4817 4818 4819 4820 4821 4822
			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.
		 */
4823
		if (need_resched())
4824 4825 4826
			break;

		raw_spin_lock_irq(&this_rq->lock);
4827
		update_rq_clock(this_rq);
4828
		update_idle_cpu_load(this_rq);
4829 4830 4831 4832 4833 4834 4835 4836 4837
		raw_spin_unlock_irq(&this_rq->lock);

		rebalance_domains(balance_cpu, CPU_IDLE);

		rq = cpu_rq(balance_cpu);
		if (time_after(this_rq->next_balance, rq->next_balance))
			this_rq->next_balance = rq->next_balance;
	}
	nohz.next_balance = this_rq->next_balance;
4838 4839
end:
	clear_bit(NOHZ_BALANCE_KICK, nohz_flags(this_cpu));
4840 4841 4842
}

/*
4843 4844 4845 4846 4847 4848 4849
 * 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.
4850 4851 4852 4853
 */
static inline int nohz_kick_needed(struct rq *rq, int cpu)
{
	unsigned long now = jiffies;
4854
	struct sched_domain *sd;
4855

4856
	if (unlikely(idle_cpu(cpu)))
4857 4858
		return 0;

4859 4860 4861 4862
       /*
	* 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.
	*/
4863
	set_cpu_sd_state_busy();
4864
	clear_nohz_tick_stopped(cpu);
4865 4866 4867 4868 4869 4870 4871

	/*
	 * None are in tickless mode and hence no need for NOHZ idle load
	 * balancing.
	 */
	if (likely(!atomic_read(&nohz.nr_cpus)))
		return 0;
4872 4873

	if (time_before(now, nohz.next_balance))
4874 4875
		return 0;

4876 4877
	if (rq->nr_running >= 2)
		goto need_kick;
4878

4879
	rcu_read_lock();
4880 4881 4882 4883
	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);
4884

4885
		if (sd->flags & SD_SHARE_PKG_RESOURCES && nr_busy > 1)
4886
			goto need_kick_unlock;
4887 4888 4889 4890

		if (sd->flags & SD_ASYM_PACKING && nr_busy != sg->group_weight
		    && (cpumask_first_and(nohz.idle_cpus_mask,
					  sched_domain_span(sd)) < cpu))
4891
			goto need_kick_unlock;
4892 4893 4894

		if (!(sd->flags & (SD_SHARE_PKG_RESOURCES | SD_ASYM_PACKING)))
			break;
4895
	}
4896
	rcu_read_unlock();
4897
	return 0;
4898 4899 4900

need_kick_unlock:
	rcu_read_unlock();
4901 4902
need_kick:
	return 1;
4903 4904 4905 4906 4907 4908 4909 4910 4911
}
#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).
 */
4912 4913 4914 4915
static void run_rebalance_domains(struct softirq_action *h)
{
	int this_cpu = smp_processor_id();
	struct rq *this_rq = cpu_rq(this_cpu);
4916
	enum cpu_idle_type idle = this_rq->idle_balance ?
4917 4918 4919 4920 4921
						CPU_IDLE : CPU_NOT_IDLE;

	rebalance_domains(this_cpu, idle);

	/*
4922
	 * If this cpu has a pending nohz_balance_kick, then do the
4923 4924 4925
	 * balancing on behalf of the other idle cpus whose ticks are
	 * stopped.
	 */
4926
	nohz_idle_balance(this_cpu, idle);
4927 4928 4929 4930
}

static inline int on_null_domain(int cpu)
{
4931
	return !rcu_dereference_sched(cpu_rq(cpu)->sd);
4932 4933 4934 4935 4936
}

/*
 * Trigger the SCHED_SOFTIRQ if it is time to do periodic load balancing.
 */
4937
void trigger_load_balance(struct rq *rq, int cpu)
4938 4939 4940 4941 4942
{
	/* 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);
4943
#ifdef CONFIG_NO_HZ
4944
	if (nohz_kick_needed(rq, cpu) && likely(!on_null_domain(cpu)))
4945 4946
		nohz_balancer_kick(cpu);
#endif
4947 4948
}

4949 4950 4951 4952 4953 4954 4955 4956
static void rq_online_fair(struct rq *rq)
{
	update_sysctl();
}

static void rq_offline_fair(struct rq *rq)
{
	update_sysctl();
4957 4958 4959

	/* Ensure any throttled groups are reachable by pick_next_task */
	unthrottle_offline_cfs_rqs(rq);
4960 4961
}

4962
#endif /* CONFIG_SMP */
4963

4964 4965 4966
/*
 * scheduler tick hitting a task of our scheduling class:
 */
P
Peter Zijlstra 已提交
4967
static void task_tick_fair(struct rq *rq, struct task_struct *curr, int queued)
4968 4969 4970 4971 4972 4973
{
	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
4974
		entity_tick(cfs_rq, se, queued);
4975 4976 4977 4978
	}
}

/*
P
Peter Zijlstra 已提交
4979 4980 4981
 * called on fork with the child task as argument from the parent's context
 *  - child not yet on the tasklist
 *  - preemption disabled
4982
 */
P
Peter Zijlstra 已提交
4983
static void task_fork_fair(struct task_struct *p)
4984
{
4985 4986
	struct cfs_rq *cfs_rq;
	struct sched_entity *se = &p->se, *curr;
4987
	int this_cpu = smp_processor_id();
P
Peter Zijlstra 已提交
4988 4989 4990
	struct rq *rq = this_rq();
	unsigned long flags;

4991
	raw_spin_lock_irqsave(&rq->lock, flags);
4992

4993 4994
	update_rq_clock(rq);

4995 4996 4997
	cfs_rq = task_cfs_rq(current);
	curr = cfs_rq->curr;

4998 4999
	if (unlikely(task_cpu(p) != this_cpu)) {
		rcu_read_lock();
P
Peter Zijlstra 已提交
5000
		__set_task_cpu(p, this_cpu);
5001 5002
		rcu_read_unlock();
	}
5003

5004
	update_curr(cfs_rq);
P
Peter Zijlstra 已提交
5005

5006 5007
	if (curr)
		se->vruntime = curr->vruntime;
5008
	place_entity(cfs_rq, se, 1);
5009

P
Peter Zijlstra 已提交
5010
	if (sysctl_sched_child_runs_first && curr && entity_before(curr, se)) {
D
Dmitry Adamushko 已提交
5011
		/*
5012 5013 5014
		 * Upon rescheduling, sched_class::put_prev_task() will place
		 * 'current' within the tree based on its new key value.
		 */
5015
		swap(curr->vruntime, se->vruntime);
5016
		resched_task(rq->curr);
5017
	}
5018

5019 5020
	se->vruntime -= cfs_rq->min_vruntime;

5021
	raw_spin_unlock_irqrestore(&rq->lock, flags);
5022 5023
}

5024 5025 5026 5027
/*
 * Priority of the task has changed. Check to see if we preempt
 * the current task.
 */
P
Peter Zijlstra 已提交
5028 5029
static void
prio_changed_fair(struct rq *rq, struct task_struct *p, int oldprio)
5030
{
P
Peter Zijlstra 已提交
5031 5032 5033
	if (!p->se.on_rq)
		return;

5034 5035 5036 5037 5038
	/*
	 * 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
5039
	if (rq->curr == p) {
5040 5041 5042
		if (p->prio > oldprio)
			resched_task(rq->curr);
	} else
5043
		check_preempt_curr(rq, p, 0);
5044 5045
}

P
Peter Zijlstra 已提交
5046 5047 5048 5049 5050 5051 5052 5053 5054 5055 5056 5057 5058 5059 5060 5061 5062 5063 5064 5065 5066 5067 5068 5069
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;
	}
}

5070 5071 5072
/*
 * We switched to the sched_fair class.
 */
P
Peter Zijlstra 已提交
5073
static void switched_to_fair(struct rq *rq, struct task_struct *p)
5074
{
P
Peter Zijlstra 已提交
5075 5076 5077
	if (!p->se.on_rq)
		return;

5078 5079 5080 5081 5082
	/*
	 * 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
5083
	if (rq->curr == p)
5084 5085
		resched_task(rq->curr);
	else
5086
		check_preempt_curr(rq, p, 0);
5087 5088
}

5089 5090 5091 5092 5093 5094 5095 5096 5097
/* 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;

5098 5099 5100 5101 5102 5103 5104
	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);
	}
5105 5106
}

5107 5108 5109 5110 5111 5112 5113 5114 5115
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
}

P
Peter Zijlstra 已提交
5116
#ifdef CONFIG_FAIR_GROUP_SCHED
5117
static void task_move_group_fair(struct task_struct *p, int on_rq)
P
Peter Zijlstra 已提交
5118
{
5119 5120 5121 5122 5123 5124 5125 5126 5127 5128 5129 5130 5131
	/*
	 * 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.
	 */
5132 5133 5134 5135 5136 5137
	/*
	 * 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().
5138 5139
	 * - Moving a task which has been woken up by try_to_wake_up() and
	 *   waiting for actually being woken up by sched_ttwu_pending().
5140 5141 5142 5143
	 *
	 * To prevent boost or penalty in the new cfs_rq caused by delta
	 * min_vruntime between the two cfs_rqs, we skip vruntime adjustment.
	 */
5144
	if (!on_rq && (!p->se.sum_exec_runtime || p->state == TASK_WAKING))
5145 5146
		on_rq = 1;

5147 5148 5149
	if (!on_rq)
		p->se.vruntime -= cfs_rq_of(&p->se)->min_vruntime;
	set_task_rq(p, task_cpu(p));
5150
	if (!on_rq)
5151
		p->se.vruntime += cfs_rq_of(&p->se)->min_vruntime;
P
Peter Zijlstra 已提交
5152
}
5153 5154 5155 5156 5157 5158 5159 5160 5161 5162 5163 5164 5165 5166 5167 5168 5169 5170 5171 5172 5173 5174 5175 5176 5177 5178 5179 5180 5181 5182 5183 5184 5185 5186 5187 5188 5189 5190 5191 5192 5193 5194 5195 5196 5197 5198 5199 5200 5201 5202 5203 5204 5205 5206 5207 5208 5209 5210 5211 5212 5213 5214 5215 5216 5217 5218 5219 5220 5221 5222 5223 5224 5225 5226 5227 5228 5229 5230 5231 5232 5233 5234 5235 5236 5237 5238

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;
#ifdef CONFIG_SMP
	/* allow initial update_cfs_load() to truncate */
	cfs_rq->load_stamp = 1;
P
Peter Zijlstra 已提交
5239
#endif
5240 5241 5242 5243 5244 5245 5246 5247 5248 5249 5250 5251 5252 5253 5254 5255 5256 5257 5258 5259 5260 5261 5262 5263 5264 5265 5266 5267 5268 5269 5270 5271 5272 5273 5274 5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286 5287 5288 5289 5290 5291 5292 5293 5294 5295 5296 5297 5298 5299 5300 5301 5302 5303 5304 5305 5306 5307
	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);
		for_each_sched_entity(se)
			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
5308

5309
static unsigned int get_rr_interval_fair(struct rq *rq, struct task_struct *task)
5310 5311 5312 5313 5314 5315 5316 5317 5318 5319 5320 5321 5322 5323
{
	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)
		rr_interval = NS_TO_JIFFIES(sched_slice(&rq->cfs, se));

	return rr_interval;
}

5324 5325 5326
/*
 * All the scheduling class methods:
 */
5327
const struct sched_class fair_sched_class = {
5328
	.next			= &idle_sched_class,
5329 5330 5331
	.enqueue_task		= enqueue_task_fair,
	.dequeue_task		= dequeue_task_fair,
	.yield_task		= yield_task_fair,
5332
	.yield_to_task		= yield_to_task_fair,
5333

I
Ingo Molnar 已提交
5334
	.check_preempt_curr	= check_preempt_wakeup,
5335 5336 5337 5338

	.pick_next_task		= pick_next_task_fair,
	.put_prev_task		= put_prev_task_fair,

5339
#ifdef CONFIG_SMP
L
Li Zefan 已提交
5340 5341
	.select_task_rq		= select_task_rq_fair,

5342 5343
	.rq_online		= rq_online_fair,
	.rq_offline		= rq_offline_fair,
5344 5345

	.task_waking		= task_waking_fair,
5346
#endif
5347

5348
	.set_curr_task          = set_curr_task_fair,
5349
	.task_tick		= task_tick_fair,
P
Peter Zijlstra 已提交
5350
	.task_fork		= task_fork_fair,
5351 5352

	.prio_changed		= prio_changed_fair,
P
Peter Zijlstra 已提交
5353
	.switched_from		= switched_from_fair,
5354
	.switched_to		= switched_to_fair,
P
Peter Zijlstra 已提交
5355

5356 5357
	.get_rr_interval	= get_rr_interval_fair,

P
Peter Zijlstra 已提交
5358
#ifdef CONFIG_FAIR_GROUP_SCHED
5359
	.task_move_group	= task_move_group_fair,
P
Peter Zijlstra 已提交
5360
#endif
5361 5362 5363
};

#ifdef CONFIG_SCHED_DEBUG
5364
void print_cfs_stats(struct seq_file *m, int cpu)
5365 5366 5367
{
	struct cfs_rq *cfs_rq;

5368
	rcu_read_lock();
5369
	for_each_leaf_cfs_rq(cpu_rq(cpu), cfs_rq)
5370
		print_cfs_rq(m, cpu, cfs_rq);
5371
	rcu_read_unlock();
5372 5373
}
#endif
5374 5375 5376 5377 5378 5379 5380

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

#ifdef CONFIG_NO_HZ
5381
	nohz.next_balance = jiffies;
5382
	zalloc_cpumask_var(&nohz.idle_cpus_mask, GFP_NOWAIT);
5383
	cpu_notifier(sched_ilb_notifier, 0);
5384 5385 5386 5387
#endif
#endif /* SMP */

}