fair.c 158.5 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
/*
 * Completely Fair Scheduling (CFS) Class (SCHED_NORMAL/SCHED_BATCH)
 *
 *  Copyright (C) 2007 Red Hat, Inc., Ingo Molnar <mingo@redhat.com>
 *
 *  Interactivity improvements by Mike Galbraith
 *  (C) 2007 Mike Galbraith <efault@gmx.de>
 *
 *  Various enhancements by Dmitry Adamushko.
 *  (C) 2007 Dmitry Adamushko <dmitry.adamushko@gmail.com>
 *
 *  Group scheduling enhancements by Srivatsa Vaddagiri
 *  Copyright IBM Corporation, 2007
 *  Author: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
 *
 *  Scaled math optimizations by Thomas Gleixner
 *  Copyright (C) 2007, Thomas Gleixner <tglx@linutronix.de>
18 19 20
 *
 *  Adaptive scheduling granularity, math enhancements by Peter Zijlstra
 *  Copyright (C) 2007 Red Hat, Inc., Peter Zijlstra <pzijlstr@redhat.com>
21 22
 */

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

#include <trace/events/sched.h>

#include "sched.h"
A
Arjan van de Ven 已提交
36

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

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

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

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

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

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

93 94
const_debug unsigned int sysctl_sched_migration_cost = 500000UL;

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

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

116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219
/*
 * 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;
220

221 222 223 224
/**************************************************************
 * CFS operations on generic schedulable entities:
 */

225
#ifdef CONFIG_FAIR_GROUP_SCHED
226

227
/* cpu runqueue to which this cfs_rq is attached */
228 229
static inline struct rq *rq_of(struct cfs_rq *cfs_rq)
{
230
	return cfs_rq->rq;
231 232
}

233 234
/* An entity is a task if it doesn't "own" a runqueue */
#define entity_is_task(se)	(!se->my_q)
235

236 237 238 239 240 241 242 243
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
244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264
/* 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;
}

265 266
static void update_cfs_rq_blocked_load(struct cfs_rq *cfs_rq,
				       int force_update);
267

268 269 270
static inline void list_add_leaf_cfs_rq(struct cfs_rq *cfs_rq)
{
	if (!cfs_rq->on_list) {
271 272 273 274 275 276 277 278 279 280 281 282
		/*
		 * 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,
283
				&rq_of(cfs_rq)->leaf_cfs_rq_list);
284
		}
285 286

		cfs_rq->on_list = 1;
287
		/* We should have no load, but we need to update last_decay. */
288
		update_cfs_rq_blocked_load(cfs_rq, 0);
289 290 291 292 293 294 295 296 297 298 299
	}
}

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
300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318
/* 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;
}

319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361
/* 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);
	}
}

362 363 364 365 366 367
#else	/* !CONFIG_FAIR_GROUP_SCHED */

static inline struct task_struct *task_of(struct sched_entity *se)
{
	return container_of(se, struct task_struct, se);
}
368

369 370 371
static inline struct rq *rq_of(struct cfs_rq *cfs_rq)
{
	return container_of(cfs_rq, struct rq, cfs);
372 373 374 375
}

#define entity_is_task(se)	1

P
Peter Zijlstra 已提交
376 377
#define for_each_sched_entity(se) \
		for (; se; se = NULL)
378

P
Peter Zijlstra 已提交
379
static inline struct cfs_rq *task_cfs_rq(struct task_struct *p)
380
{
P
Peter Zijlstra 已提交
381
	return &task_rq(p)->cfs;
382 383
}

P
Peter Zijlstra 已提交
384 385 386 387 388 389 390 391 392 393 394 395 396 397
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;
}

398 399 400 401 402 403 404 405
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
406 407 408 409 410 411 412 413 414 415 416 417 418 419
#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;
}

420 421 422 423 424
static inline void
find_matching_se(struct sched_entity **se, struct sched_entity **pse)
{
}

P
Peter Zijlstra 已提交
425 426
#endif	/* CONFIG_FAIR_GROUP_SCHED */

427 428
static __always_inline
void account_cfs_rq_runtime(struct cfs_rq *cfs_rq, unsigned long delta_exec);
429 430 431 432 433

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

434
static inline u64 max_vruntime(u64 max_vruntime, u64 vruntime)
435
{
436
	s64 delta = (s64)(vruntime - max_vruntime);
437
	if (delta > 0)
438
		max_vruntime = vruntime;
439

440
	return max_vruntime;
441 442
}

443
static inline u64 min_vruntime(u64 min_vruntime, u64 vruntime)
P
Peter Zijlstra 已提交
444 445 446 447 448 449 450 451
{
	s64 delta = (s64)(vruntime - min_vruntime);
	if (delta < 0)
		min_vruntime = vruntime;

	return min_vruntime;
}

452 453 454 455 456 457
static inline int entity_before(struct sched_entity *a,
				struct sched_entity *b)
{
	return (s64)(a->vruntime - b->vruntime) < 0;
}

458 459 460 461 462 463 464 465 466 467 468 469
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
470
		if (!cfs_rq->curr)
471 472 473 474 475
			vruntime = se->vruntime;
		else
			vruntime = min_vruntime(vruntime, se->vruntime);
	}

476
	/* ensure we never gain time by being placed backwards. */
477
	cfs_rq->min_vruntime = max_vruntime(cfs_rq->min_vruntime, vruntime);
478 479 480 481
#ifndef CONFIG_64BIT
	smp_wmb();
	cfs_rq->min_vruntime_copy = cfs_rq->min_vruntime;
#endif
482 483
}

484 485 486
/*
 * Enqueue an entity into the rb-tree:
 */
487
static void __enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503
{
	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.
		 */
504
		if (entity_before(se, entry)) {
505 506 507 508 509 510 511 512 513 514 515
			link = &parent->rb_left;
		} else {
			link = &parent->rb_right;
			leftmost = 0;
		}
	}

	/*
	 * Maintain a cache of leftmost tree entries (it is frequently
	 * used):
	 */
516
	if (leftmost)
I
Ingo Molnar 已提交
517
		cfs_rq->rb_leftmost = &se->run_node;
518 519 520 521 522

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

523
static void __dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
524
{
P
Peter Zijlstra 已提交
525 526 527 528 529 530
	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
531

532 533 534
	rb_erase(&se->run_node, &cfs_rq->tasks_timeline);
}

535
struct sched_entity *__pick_first_entity(struct cfs_rq *cfs_rq)
536
{
537 538 539 540 541 542
	struct rb_node *left = cfs_rq->rb_leftmost;

	if (!left)
		return NULL;

	return rb_entry(left, struct sched_entity, run_node);
543 544
}

545 546 547 548 549 550 551 552 553 554 555
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
556
struct sched_entity *__pick_last_entity(struct cfs_rq *cfs_rq)
557
{
558
	struct rb_node *last = rb_last(&cfs_rq->tasks_timeline);
559

560 561
	if (!last)
		return NULL;
562 563

	return rb_entry(last, struct sched_entity, run_node);
564 565
}

566 567 568 569
/**************************************************************
 * Scheduling class statistics methods:
 */

570
int sched_proc_update_handler(struct ctl_table *table, int write,
571
		void __user *buffer, size_t *lenp,
572 573
		loff_t *ppos)
{
574
	int ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
575
	int factor = get_update_sysctl_factor();
576 577 578 579 580 581 582

	if (ret || !write)
		return ret;

	sched_nr_latency = DIV_ROUND_UP(sysctl_sched_latency,
					sysctl_sched_min_granularity);

583 584 585 586 587 588 589
#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

590 591 592
	return 0;
}
#endif
593

594
/*
595
 * delta /= w
596 597 598 599
 */
static inline unsigned long
calc_delta_fair(unsigned long delta, struct sched_entity *se)
{
600 601
	if (unlikely(se->load.weight != NICE_0_LOAD))
		delta = calc_delta_mine(delta, NICE_0_LOAD, &se->load);
602 603 604 605

	return delta;
}

606 607 608
/*
 * The idea is to set a period in which each task runs once.
 *
609
 * When there are too many tasks (sched_nr_latency) we have to stretch
610 611 612 613
 * this period because otherwise the slices get too small.
 *
 * p = (nr <= nl) ? l : l*nr/nl
 */
614 615 616
static u64 __sched_period(unsigned long nr_running)
{
	u64 period = sysctl_sched_latency;
617
	unsigned long nr_latency = sched_nr_latency;
618 619

	if (unlikely(nr_running > nr_latency)) {
620
		period = sysctl_sched_min_granularity;
621 622 623 624 625 626
		period *= nr_running;
	}

	return period;
}

627 628 629 630
/*
 * We calculate the wall-time slice from the period by taking a part
 * proportional to the weight.
 *
631
 * s = p*P[w/rw]
632
 */
P
Peter Zijlstra 已提交
633
static u64 sched_slice(struct cfs_rq *cfs_rq, struct sched_entity *se)
634
{
M
Mike Galbraith 已提交
635
	u64 slice = __sched_period(cfs_rq->nr_running + !se->on_rq);
636

M
Mike Galbraith 已提交
637
	for_each_sched_entity(se) {
L
Lin Ming 已提交
638
		struct load_weight *load;
639
		struct load_weight lw;
L
Lin Ming 已提交
640 641 642

		cfs_rq = cfs_rq_of(se);
		load = &cfs_rq->load;
643

M
Mike Galbraith 已提交
644
		if (unlikely(!se->on_rq)) {
645
			lw = cfs_rq->load;
M
Mike Galbraith 已提交
646 647 648 649 650 651 652

			update_load_add(&lw, se->load.weight);
			load = &lw;
		}
		slice = calc_delta_mine(slice, se->load.weight, load);
	}
	return slice;
653 654
}

655
/*
A
Andrei Epure 已提交
656
 * We calculate the vruntime slice of a to-be-inserted task.
657
 *
658
 * vs = s/w
659
 */
660
static u64 sched_vslice(struct cfs_rq *cfs_rq, struct sched_entity *se)
P
Peter Zijlstra 已提交
661
{
662
	return calc_delta_fair(sched_slice(cfs_rq, se), se);
663 664
}

665 666 667 668 669
/*
 * 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
670 671
__update_curr(struct cfs_rq *cfs_rq, struct sched_entity *curr,
	      unsigned long delta_exec)
672
{
673
	unsigned long delta_exec_weighted;
674

675 676
	schedstat_set(curr->statistics.exec_max,
		      max((u64)delta_exec, curr->statistics.exec_max));
677 678

	curr->sum_exec_runtime += delta_exec;
679
	schedstat_add(cfs_rq, exec_clock, delta_exec);
680
	delta_exec_weighted = calc_delta_fair(delta_exec, curr);
681

I
Ingo Molnar 已提交
682
	curr->vruntime += delta_exec_weighted;
683
	update_min_vruntime(cfs_rq);
684 685
}

686
static void update_curr(struct cfs_rq *cfs_rq)
687
{
688
	struct sched_entity *curr = cfs_rq->curr;
689
	u64 now = rq_of(cfs_rq)->clock_task;
690 691 692 693 694 695 696 697 698 699
	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
700
	delta_exec = (unsigned long)(now - curr->exec_start);
P
Peter Zijlstra 已提交
701 702
	if (!delta_exec)
		return;
703

I
Ingo Molnar 已提交
704 705
	__update_curr(cfs_rq, curr, delta_exec);
	curr->exec_start = now;
706 707 708 709

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

710
		trace_sched_stat_runtime(curtask, delta_exec, curr->vruntime);
711
		cpuacct_charge(curtask, delta_exec);
712
		account_group_exec_runtime(curtask, delta_exec);
713
	}
714 715

	account_cfs_rq_runtime(cfs_rq, delta_exec);
716 717 718
}

static inline void
719
update_stats_wait_start(struct cfs_rq *cfs_rq, struct sched_entity *se)
720
{
721
	schedstat_set(se->statistics.wait_start, rq_of(cfs_rq)->clock);
722 723 724 725 726
}

/*
 * Task is being enqueued - update stats:
 */
727
static void update_stats_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se)
728 729 730 731 732
{
	/*
	 * Are we enqueueing a waiting task? (for current tasks
	 * a dequeue/enqueue event is a NOP)
	 */
733
	if (se != cfs_rq->curr)
734
		update_stats_wait_start(cfs_rq, se);
735 736 737
}

static void
738
update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se)
739
{
740 741 742 743 744
	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);
745 746 747
#ifdef CONFIG_SCHEDSTATS
	if (entity_is_task(se)) {
		trace_sched_stat_wait(task_of(se),
748
			rq_of(cfs_rq)->clock - se->statistics.wait_start);
749 750
	}
#endif
751
	schedstat_set(se->statistics.wait_start, 0);
752 753 754
}

static inline void
755
update_stats_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se)
756 757 758 759 760
{
	/*
	 * Mark the end of the wait period if dequeueing a
	 * waiting task:
	 */
761
	if (se != cfs_rq->curr)
762
		update_stats_wait_end(cfs_rq, se);
763 764 765 766 767 768
}

/*
 * We are picking a new current task - update its stats:
 */
static inline void
769
update_stats_curr_start(struct cfs_rq *cfs_rq, struct sched_entity *se)
770 771 772 773
{
	/*
	 * We are starting a new run period:
	 */
774
	se->exec_start = rq_of(cfs_rq)->clock_task;
775 776 777 778 779 780
}

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

781 782
#ifdef CONFIG_NUMA_BALANCING
/*
783
 * numa task sample period in ms
784
 */
785
unsigned int sysctl_numa_balancing_scan_period_min = 100;
786 787
unsigned int sysctl_numa_balancing_scan_period_max = 100*50;
unsigned int sysctl_numa_balancing_scan_period_reset = 100*600;
788 789 790

/* Portion of address space to scan in MB */
unsigned int sysctl_numa_balancing_scan_size = 256;
791

792 793 794
/* Scan @scan_size MB every @scan_period after an initial @scan_delay in ms */
unsigned int sysctl_numa_balancing_scan_delay = 1000;

795 796
static void task_numa_placement(struct task_struct *p)
{
797
	int seq;
798

799 800 801
	if (!p->mm)	/* for example, ksmd faulting in a user's mm */
		return;
	seq = ACCESS_ONCE(p->mm->numa_scan_seq);
802 803 804 805 806 807 808 809 810 811
	if (p->numa_scan_seq == seq)
		return;
	p->numa_scan_seq = seq;

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

/*
 * Got a PROT_NONE fault for a page on @node.
 */
812
void task_numa_fault(int node, int pages, bool migrated)
813 814 815
{
	struct task_struct *p = current;

816 817 818
	if (!sched_feat_numa(NUMA))
		return;

819 820
	/* FIXME: Allocate task-specific structure for placement policy here */

821
	/*
822 823
	 * If pages are properly placed (did not migrate) then scan slower.
	 * This is reset periodically in case of phase changes
824
	 */
825 826 827
        if (!migrated)
		p->numa_scan_period = min(sysctl_numa_balancing_scan_period_max,
			p->numa_scan_period + jiffies_to_msecs(10));
828

829 830 831
	task_numa_placement(p);
}

832 833 834 835 836 837
static void reset_ptenuma_scan(struct task_struct *p)
{
	ACCESS_ONCE(p->mm->numa_scan_seq)++;
	p->mm->numa_scan_offset = 0;
}

838 839 840 841 842 843 844 845 846
/*
 * The expensive part of numa migration is done from task_work context.
 * Triggered from task_tick_numa().
 */
void task_numa_work(struct callback_head *work)
{
	unsigned long migrate, next_scan, now = jiffies;
	struct task_struct *p = current;
	struct mm_struct *mm = p->mm;
847
	struct vm_area_struct *vma;
848 849
	unsigned long start, end;
	long pages;
850 851 852 853 854 855 856 857 858 859 860 861 862 863 864

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

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

865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882
	/*
	 * We do not care about task placement until a task runs on a node
	 * other than the first one used by the address space. This is
	 * largely because migrations are driven by what CPU the task
	 * is running on. If it's never scheduled on another node, it'll
	 * not migrate so why bother trapping the fault.
	 */
	if (mm->first_nid == NUMA_PTE_SCAN_INIT)
		mm->first_nid = numa_node_id();
	if (mm->first_nid != NUMA_PTE_SCAN_ACTIVE) {
		/* Are we running on a new node yet? */
		if (numa_node_id() == mm->first_nid &&
		    !sched_feat_numa(NUMA_FORCE))
			return;

		mm->first_nid = NUMA_PTE_SCAN_ACTIVE;
	}

883 884 885 886 887 888 889 890 891 892 893 894 895
	/*
	 * Reset the scan period if enough time has gone by. Objective is that
	 * scanning will be reduced if pages are properly placed. As tasks
	 * can enter different phases this needs to be re-examined. Lacking
	 * proper tracking of reference behaviour, this blunt hammer is used.
	 */
	migrate = mm->numa_next_reset;
	if (time_after(now, migrate)) {
		p->numa_scan_period = sysctl_numa_balancing_scan_period_min;
		next_scan = now + msecs_to_jiffies(sysctl_numa_balancing_scan_period_reset);
		xchg(&mm->numa_next_reset, next_scan);
	}

896 897 898 899 900 901 902 903 904 905
	/*
	 * Enforce maximal scan/migration frequency..
	 */
	migrate = mm->numa_next_scan;
	if (time_before(now, migrate))
		return;

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

906
	next_scan = now + msecs_to_jiffies(p->numa_scan_period);
907 908 909
	if (cmpxchg(&mm->numa_next_scan, migrate, next_scan) != migrate)
		return;

910 911 912 913 914 915 916 917
	/*
	 * Do not set pte_numa if the current running node is rate-limited.
	 * This loses statistics on the fault but if we are unwilling to
	 * migrate to this node, it is less likely we can do useful work
	 */
	if (migrate_ratelimited(numa_node_id()))
		return;

918 919 920 921 922
	start = mm->numa_scan_offset;
	pages = sysctl_numa_balancing_scan_size;
	pages <<= 20 - PAGE_SHIFT; /* MB in pages */
	if (!pages)
		return;
923

924
	down_read(&mm->mmap_sem);
925
	vma = find_vma(mm, start);
926 927
	if (!vma) {
		reset_ptenuma_scan(p);
928
		start = 0;
929 930
		vma = mm->mmap;
	}
931
	for (; vma; vma = vma->vm_next) {
932 933 934 935
		if (!vma_migratable(vma))
			continue;

		/* Skip small VMAs. They are not likely to be of relevance */
936
		if (vma->vm_end - vma->vm_start < HPAGE_SIZE)
937 938
			continue;

939 940 941 942 943
		do {
			start = max(start, vma->vm_start);
			end = ALIGN(start + (pages << PAGE_SHIFT), HPAGE_SIZE);
			end = min(end, vma->vm_end);
			pages -= change_prot_numa(vma, start, end);
944

945 946 947 948
			start = end;
			if (pages <= 0)
				goto out;
		} while (end != vma->vm_end);
949
	}
950

951
out:
952 953 954 955 956 957 958
	/*
	 * It is possible to reach the end of the VMA list but the last few VMAs are
	 * not guaranteed to the vma_migratable. If they are not, we would find the
	 * !migratable VMA on the next scan but not reset the scanner to the start
	 * so check it now.
	 */
	if (vma)
959
		mm->numa_scan_offset = start;
960 961 962
	else
		reset_ptenuma_scan(p);
	up_read(&mm->mmap_sem);
963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988
}

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

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

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

	if (now - curr->node_stamp > period) {
989 990
		if (!curr->node_stamp)
			curr->numa_scan_period = sysctl_numa_balancing_scan_period_min;
991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004
		curr->node_stamp = now;

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

1005 1006 1007 1008
static void
account_entity_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se)
{
	update_load_add(&cfs_rq->load, se->load.weight);
1009
	if (!parent_entity(se))
1010
		update_load_add(&rq_of(cfs_rq)->load, se->load.weight);
1011 1012
#ifdef CONFIG_SMP
	if (entity_is_task(se))
1013
		list_add(&se->group_node, &rq_of(cfs_rq)->cfs_tasks);
1014
#endif
1015 1016 1017 1018 1019 1020 1021
	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);
1022
	if (!parent_entity(se))
1023
		update_load_sub(&rq_of(cfs_rq)->load, se->load.weight);
1024
	if (entity_is_task(se))
1025
		list_del_init(&se->group_node);
1026 1027 1028
	cfs_rq->nr_running--;
}

1029 1030
#ifdef CONFIG_FAIR_GROUP_SCHED
# ifdef CONFIG_SMP
1031 1032 1033 1034 1035 1036 1037 1038 1039
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().
	 */
1040 1041
	tg_weight = atomic64_read(&tg->load_avg);
	tg_weight -= cfs_rq->tg_load_contrib;
1042 1043 1044 1045 1046
	tg_weight += cfs_rq->load.weight;

	return tg_weight;
}

1047
static long calc_cfs_shares(struct cfs_rq *cfs_rq, struct task_group *tg)
1048
{
1049
	long tg_weight, load, shares;
1050

1051
	tg_weight = calc_tg_weight(tg, cfs_rq);
1052
	load = cfs_rq->load.weight;
1053 1054

	shares = (tg->shares * load);
1055 1056
	if (tg_weight)
		shares /= tg_weight;
1057 1058 1059 1060 1061 1062 1063 1064 1065

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

	return shares;
}
# else /* CONFIG_SMP */
1066
static inline long calc_cfs_shares(struct cfs_rq *cfs_rq, struct task_group *tg)
1067 1068 1069 1070
{
	return tg->shares;
}
# endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
1071 1072 1073
static void reweight_entity(struct cfs_rq *cfs_rq, struct sched_entity *se,
			    unsigned long weight)
{
1074 1075 1076 1077
	if (se->on_rq) {
		/* commit outstanding execution time */
		if (cfs_rq->curr == se)
			update_curr(cfs_rq);
P
Peter Zijlstra 已提交
1078
		account_entity_dequeue(cfs_rq, se);
1079
	}
P
Peter Zijlstra 已提交
1080 1081 1082 1083 1084 1085 1086

	update_load_set(&se->load, weight);

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

1087 1088
static inline int throttled_hierarchy(struct cfs_rq *cfs_rq);

1089
static void update_cfs_shares(struct cfs_rq *cfs_rq)
P
Peter Zijlstra 已提交
1090 1091 1092
{
	struct task_group *tg;
	struct sched_entity *se;
1093
	long shares;
P
Peter Zijlstra 已提交
1094 1095 1096

	tg = cfs_rq->tg;
	se = tg->se[cpu_of(rq_of(cfs_rq))];
1097
	if (!se || throttled_hierarchy(cfs_rq))
P
Peter Zijlstra 已提交
1098
		return;
1099 1100 1101 1102
#ifndef CONFIG_SMP
	if (likely(se->load.weight == tg->shares))
		return;
#endif
1103
	shares = calc_cfs_shares(cfs_rq, tg);
P
Peter Zijlstra 已提交
1104 1105 1106 1107

	reweight_entity(cfs_rq_of(se), se, shares);
}
#else /* CONFIG_FAIR_GROUP_SCHED */
1108
static inline void update_cfs_shares(struct cfs_rq *cfs_rq)
P
Peter Zijlstra 已提交
1109 1110 1111 1112
{
}
#endif /* CONFIG_FAIR_GROUP_SCHED */

1113 1114
/* Only depends on SMP, FAIR_GROUP_SCHED may be removed when useful in lb */
#if defined(CONFIG_SMP) && defined(CONFIG_FAIR_GROUP_SCHED)
1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142
/*
 * We choose a half-life close to 1 scheduling period.
 * Note: The tables below are dependent on this value.
 */
#define LOAD_AVG_PERIOD 32
#define LOAD_AVG_MAX 47742 /* maximum possible load avg */
#define LOAD_AVG_MAX_N 345 /* number of full periods to produce LOAD_MAX_AVG */

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

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

1143 1144 1145 1146 1147 1148
/*
 * Approximate:
 *   val * y^n,    where y^32 ~= 0.5 (~1 scheduling period)
 */
static __always_inline u64 decay_load(u64 val, u64 n)
{
1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168
	unsigned int local_n;

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

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

	/*
	 * As y^PERIOD = 1/2, we can combine
	 *    y^n = 1/2^(n/PERIOD) * k^(n%PERIOD)
	 * With a look-up table which covers k^n (n<PERIOD)
	 *
	 * To achieve constant time decay_load.
	 */
	if (unlikely(local_n >= LOAD_AVG_PERIOD)) {
		val >>= local_n / LOAD_AVG_PERIOD;
		local_n %= LOAD_AVG_PERIOD;
1169 1170
	}

1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201
	val *= runnable_avg_yN_inv[local_n];
	/* We don't use SRR here since we always want to round down. */
	return val >> 32;
}

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

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

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

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

	contrib = decay_load(contrib, n);
	return contrib + runnable_avg_yN_sum[n];
1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235
}

/*
 * We can represent the historical contribution to runnable average as the
 * coefficients of a geometric series.  To do this we sub-divide our runnable
 * history into segments of approximately 1ms (1024us); label the segment that
 * occurred N-ms ago p_N, with p_0 corresponding to the current period, e.g.
 *
 * [<- 1024us ->|<- 1024us ->|<- 1024us ->| ...
 *      p0            p1           p2
 *     (now)       (~1ms ago)  (~2ms ago)
 *
 * Let u_i denote the fraction of p_i that the entity was runnable.
 *
 * We then designate the fractions u_i as our co-efficients, yielding the
 * following representation of historical load:
 *   u_0 + u_1*y + u_2*y^2 + u_3*y^3 + ...
 *
 * We choose y based on the with of a reasonably scheduling period, fixing:
 *   y^32 = 0.5
 *
 * This means that the contribution to load ~32ms ago (u_32) will be weighted
 * approximately half as much as the contribution to load within the last ms
 * (u_0).
 *
 * When a period "rolls over" and we have new u_0`, multiplying the previous
 * sum again by y is sufficient to update:
 *   load_avg = u_0` + y*(u_0 + u_1*y + u_2*y^2 + ... )
 *            = u_0 + u_1*y + u_2*y^2 + ... [re-labeling u_i --> u_{i+1}]
 */
static __always_inline int __update_entity_runnable_avg(u64 now,
							struct sched_avg *sa,
							int runnable)
{
1236 1237
	u64 delta, periods;
	u32 runnable_contrib;
1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270
	int delta_w, decayed = 0;

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

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

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

		/*
		 * Now that we know we're crossing a period boundary, figure
		 * out how much from delta we need to complete the current
		 * period and accrue it.
		 */
		delta_w = 1024 - delta_w;
1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290
		if (runnable)
			sa->runnable_avg_sum += delta_w;
		sa->runnable_avg_period += delta_w;

		delta -= delta_w;

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

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

		/* Efficiently calculate \sum (1..n_period) 1024*y^i */
		runnable_contrib = __compute_runnable_contrib(periods);
		if (runnable)
			sa->runnable_avg_sum += runnable_contrib;
		sa->runnable_avg_period += runnable_contrib;
1291 1292 1293 1294 1295 1296 1297 1298 1299 1300
	}

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

	return decayed;
}

1301
/* Synchronize an entity's decay with its parenting cfs_rq.*/
1302
static inline u64 __synchronize_entity_decay(struct sched_entity *se)
1303 1304 1305 1306 1307 1308
{
	struct cfs_rq *cfs_rq = cfs_rq_of(se);
	u64 decays = atomic64_read(&cfs_rq->decay_counter);

	decays -= se->avg.decay_count;
	if (!decays)
1309
		return 0;
1310 1311 1312

	se->avg.load_avg_contrib = decay_load(se->avg.load_avg_contrib, decays);
	se->avg.decay_count = 0;
1313 1314

	return decays;
1315 1316
}

1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331
#ifdef CONFIG_FAIR_GROUP_SCHED
static inline void __update_cfs_rq_tg_load_contrib(struct cfs_rq *cfs_rq,
						 int force_update)
{
	struct task_group *tg = cfs_rq->tg;
	s64 tg_contrib;

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

	if (force_update || abs64(tg_contrib) > cfs_rq->tg_load_contrib / 8) {
		atomic64_add(tg_contrib, &tg->load_avg);
		cfs_rq->tg_load_contrib += tg_contrib;
	}
}
1332

1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353
/*
 * Aggregate cfs_rq runnable averages into an equivalent task_group
 * representation for computing load contributions.
 */
static inline void __update_tg_runnable_avg(struct sched_avg *sa,
						  struct cfs_rq *cfs_rq)
{
	struct task_group *tg = cfs_rq->tg;
	long contrib;

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

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

1354 1355 1356 1357
static inline void __update_group_entity_contrib(struct sched_entity *se)
{
	struct cfs_rq *cfs_rq = group_cfs_rq(se);
	struct task_group *tg = cfs_rq->tg;
1358 1359
	int runnable_avg;

1360 1361 1362 1363 1364
	u64 contrib;

	contrib = cfs_rq->tg_load_contrib * tg->shares;
	se->avg.load_avg_contrib = div64_u64(contrib,
					     atomic64_read(&tg->load_avg) + 1);
1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393

	/*
	 * For group entities we need to compute a correction term in the case
	 * that they are consuming <1 cpu so that we would contribute the same
	 * load as a task of equal weight.
	 *
	 * Explicitly co-ordinating this measurement would be expensive, but
	 * fortunately the sum of each cpus contribution forms a usable
	 * lower-bound on the true value.
	 *
	 * Consider the aggregate of 2 contributions.  Either they are disjoint
	 * (and the sum represents true value) or they are disjoint and we are
	 * understating by the aggregate of their overlap.
	 *
	 * Extending this to N cpus, for a given overlap, the maximum amount we
	 * understand is then n_i(n_i+1)/2 * w_i where n_i is the number of
	 * cpus that overlap for this interval and w_i is the interval width.
	 *
	 * On a small machine; the first term is well-bounded which bounds the
	 * total error since w_i is a subset of the period.  Whereas on a
	 * larger machine, while this first term can be larger, if w_i is the
	 * of consequential size guaranteed to see n_i*w_i quickly converge to
	 * our upper bound of 1-cpu.
	 */
	runnable_avg = atomic_read(&tg->runnable_avg);
	if (runnable_avg < NICE_0_LOAD) {
		se->avg.load_avg_contrib *= runnable_avg;
		se->avg.load_avg_contrib >>= NICE_0_SHIFT;
	}
1394
}
1395 1396 1397
#else
static inline void __update_cfs_rq_tg_load_contrib(struct cfs_rq *cfs_rq,
						 int force_update) {}
1398 1399
static inline void __update_tg_runnable_avg(struct sched_avg *sa,
						  struct cfs_rq *cfs_rq) {}
1400
static inline void __update_group_entity_contrib(struct sched_entity *se) {}
1401 1402
#endif

1403 1404 1405 1406 1407 1408 1409 1410 1411 1412
static inline void __update_task_entity_contrib(struct sched_entity *se)
{
	u32 contrib;

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

1413 1414 1415 1416 1417
/* Compute the current contribution to load_avg by se, return any delta */
static long __update_entity_load_avg_contrib(struct sched_entity *se)
{
	long old_contrib = se->avg.load_avg_contrib;

1418 1419 1420
	if (entity_is_task(se)) {
		__update_task_entity_contrib(se);
	} else {
1421
		__update_tg_runnable_avg(&se->avg, group_cfs_rq(se));
1422 1423
		__update_group_entity_contrib(se);
	}
1424 1425 1426 1427

	return se->avg.load_avg_contrib - old_contrib;
}

1428 1429 1430 1431 1432 1433 1434 1435 1436
static inline void subtract_blocked_load_contrib(struct cfs_rq *cfs_rq,
						 long load_contrib)
{
	if (likely(load_contrib < cfs_rq->blocked_load_avg))
		cfs_rq->blocked_load_avg -= load_contrib;
	else
		cfs_rq->blocked_load_avg = 0;
}

1437 1438
static inline u64 cfs_rq_clock_task(struct cfs_rq *cfs_rq);

1439
/* Update a sched_entity's runnable average */
1440 1441
static inline void update_entity_load_avg(struct sched_entity *se,
					  int update_cfs_rq)
1442
{
1443 1444
	struct cfs_rq *cfs_rq = cfs_rq_of(se);
	long contrib_delta;
1445
	u64 now;
1446

1447 1448 1449 1450 1451 1452 1453 1454 1455 1456
	/*
	 * For a group entity we need to use their owned cfs_rq_clock_task() in
	 * case they are the parent of a throttled hierarchy.
	 */
	if (entity_is_task(se))
		now = cfs_rq_clock_task(cfs_rq);
	else
		now = cfs_rq_clock_task(group_cfs_rq(se));

	if (!__update_entity_runnable_avg(now, &se->avg, se->on_rq))
1457 1458 1459
		return;

	contrib_delta = __update_entity_load_avg_contrib(se);
1460 1461 1462 1463

	if (!update_cfs_rq)
		return;

1464 1465
	if (se->on_rq)
		cfs_rq->runnable_load_avg += contrib_delta;
1466 1467 1468 1469 1470 1471 1472 1473
	else
		subtract_blocked_load_contrib(cfs_rq, -contrib_delta);
}

/*
 * Decay the load contributed by all blocked children and account this so that
 * their contribution may appropriately discounted when they wake up.
 */
1474
static void update_cfs_rq_blocked_load(struct cfs_rq *cfs_rq, int force_update)
1475
{
1476
	u64 now = cfs_rq_clock_task(cfs_rq) >> 20;
1477 1478 1479
	u64 decays;

	decays = now - cfs_rq->last_decay;
1480
	if (!decays && !force_update)
1481 1482
		return;

1483 1484 1485 1486
	if (atomic64_read(&cfs_rq->removed_load)) {
		u64 removed_load = atomic64_xchg(&cfs_rq->removed_load, 0);
		subtract_blocked_load_contrib(cfs_rq, removed_load);
	}
1487

1488 1489 1490 1491 1492 1493
	if (decays) {
		cfs_rq->blocked_load_avg = decay_load(cfs_rq->blocked_load_avg,
						      decays);
		atomic64_add(decays, &cfs_rq->decay_counter);
		cfs_rq->last_decay = now;
	}
1494 1495

	__update_cfs_rq_tg_load_contrib(cfs_rq, force_update);
1496
}
1497 1498 1499 1500

static inline void update_rq_runnable_avg(struct rq *rq, int runnable)
{
	__update_entity_runnable_avg(rq->clock_task, &rq->avg, runnable);
1501
	__update_tg_runnable_avg(&rq->avg, &rq->cfs);
1502
}
1503 1504 1505

/* Add the load generated by se into cfs_rq's child load-average */
static inline void enqueue_entity_load_avg(struct cfs_rq *cfs_rq,
1506 1507
						  struct sched_entity *se,
						  int wakeup)
1508
{
1509 1510 1511 1512 1513 1514
	/*
	 * We track migrations using entity decay_count <= 0, on a wake-up
	 * migration we use a negative decay count to track the remote decays
	 * accumulated while sleeping.
	 */
	if (unlikely(se->avg.decay_count <= 0)) {
1515
		se->avg.last_runnable_update = rq_of(cfs_rq)->clock_task;
1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530
		if (se->avg.decay_count) {
			/*
			 * In a wake-up migration we have to approximate the
			 * time sleeping.  This is because we can't synchronize
			 * clock_task between the two cpus, and it is not
			 * guaranteed to be read-safe.  Instead, we can
			 * approximate this using our carried decays, which are
			 * explicitly atomically readable.
			 */
			se->avg.last_runnable_update -= (-se->avg.decay_count)
							<< 20;
			update_entity_load_avg(se, 0);
			/* Indicate that we're now synchronized and on-rq */
			se->avg.decay_count = 0;
		}
1531 1532 1533 1534 1535
		wakeup = 0;
	} else {
		__synchronize_entity_decay(se);
	}

1536 1537
	/* migrated tasks did not contribute to our blocked load */
	if (wakeup) {
1538
		subtract_blocked_load_contrib(cfs_rq, se->avg.load_avg_contrib);
1539 1540
		update_entity_load_avg(se, 0);
	}
1541

1542
	cfs_rq->runnable_load_avg += se->avg.load_avg_contrib;
1543 1544
	/* we force update consideration on load-balancer moves */
	update_cfs_rq_blocked_load(cfs_rq, !wakeup);
1545 1546
}

1547 1548 1549 1550 1551
/*
 * Remove se's load from this cfs_rq child load-average, if the entity is
 * transitioning to a blocked state we track its projected decay using
 * blocked_load_avg.
 */
1552
static inline void dequeue_entity_load_avg(struct cfs_rq *cfs_rq,
1553 1554
						  struct sched_entity *se,
						  int sleep)
1555
{
1556
	update_entity_load_avg(se, 1);
1557 1558
	/* we force update consideration on load-balancer moves */
	update_cfs_rq_blocked_load(cfs_rq, !sleep);
1559

1560
	cfs_rq->runnable_load_avg -= se->avg.load_avg_contrib;
1561 1562 1563 1564
	if (sleep) {
		cfs_rq->blocked_load_avg += se->avg.load_avg_contrib;
		se->avg.decay_count = atomic64_read(&cfs_rq->decay_counter);
	} /* migrations, e.g. sleep=0 leave decay_count == 0 */
1565
}
1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586

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

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

1587
#else
1588 1589
static inline void update_entity_load_avg(struct sched_entity *se,
					  int update_cfs_rq) {}
1590
static inline void update_rq_runnable_avg(struct rq *rq, int runnable) {}
1591
static inline void enqueue_entity_load_avg(struct cfs_rq *cfs_rq,
1592 1593
					   struct sched_entity *se,
					   int wakeup) {}
1594
static inline void dequeue_entity_load_avg(struct cfs_rq *cfs_rq,
1595 1596
					   struct sched_entity *se,
					   int sleep) {}
1597 1598
static inline void update_cfs_rq_blocked_load(struct cfs_rq *cfs_rq,
					      int force_update) {}
1599 1600
#endif

1601
static void enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
1602 1603
{
#ifdef CONFIG_SCHEDSTATS
1604 1605 1606 1607 1608
	struct task_struct *tsk = NULL;

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

1609 1610
	if (se->statistics.sleep_start) {
		u64 delta = rq_of(cfs_rq)->clock - se->statistics.sleep_start;
1611 1612 1613 1614

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

1615 1616
		if (unlikely(delta > se->statistics.sleep_max))
			se->statistics.sleep_max = delta;
1617

1618
		se->statistics.sleep_start = 0;
1619
		se->statistics.sum_sleep_runtime += delta;
A
Arjan van de Ven 已提交
1620

1621
		if (tsk) {
1622
			account_scheduler_latency(tsk, delta >> 10, 1);
1623 1624
			trace_sched_stat_sleep(tsk, delta);
		}
1625
	}
1626 1627
	if (se->statistics.block_start) {
		u64 delta = rq_of(cfs_rq)->clock - se->statistics.block_start;
1628 1629 1630 1631

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

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

1635
		se->statistics.block_start = 0;
1636
		se->statistics.sum_sleep_runtime += delta;
I
Ingo Molnar 已提交
1637

1638
		if (tsk) {
1639
			if (tsk->in_iowait) {
1640 1641
				se->statistics.iowait_sum += delta;
				se->statistics.iowait_count++;
1642
				trace_sched_stat_iowait(tsk, delta);
1643 1644
			}

1645 1646
			trace_sched_stat_blocked(tsk, delta);

1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657
			/*
			 * 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
1658
		}
1659 1660 1661 1662
	}
#endif
}

P
Peter Zijlstra 已提交
1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675
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
}

1676 1677 1678
static void
place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial)
{
1679
	u64 vruntime = cfs_rq->min_vruntime;
P
Peter Zijlstra 已提交
1680

1681 1682 1683 1684 1685 1686
	/*
	 * 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
1687
	if (initial && sched_feat(START_DEBIT))
1688
		vruntime += sched_vslice(cfs_rq, se);
1689

1690
	/* sleeps up to a single latency don't count. */
1691
	if (!initial) {
1692
		unsigned long thresh = sysctl_sched_latency;
1693

1694 1695 1696 1697 1698 1699
		/*
		 * Halve their sleep time's effect, to allow
		 * for a gentler effect of sleepers:
		 */
		if (sched_feat(GENTLE_FAIR_SLEEPERS))
			thresh >>= 1;
1700

1701
		vruntime -= thresh;
1702 1703
	}

1704
	/* ensure we never gain time by being placed backwards. */
1705
	se->vruntime = max_vruntime(se->vruntime, vruntime);
1706 1707
}

1708 1709
static void check_enqueue_throttle(struct cfs_rq *cfs_rq);

1710
static void
1711
enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
1712
{
1713 1714 1715 1716
	/*
	 * Update the normalized vruntime before updating min_vruntime
	 * through callig update_curr().
	 */
1717
	if (!(flags & ENQUEUE_WAKEUP) || (flags & ENQUEUE_WAKING))
1718 1719
		se->vruntime += cfs_rq->min_vruntime;

1720
	/*
1721
	 * Update run-time statistics of the 'current'.
1722
	 */
1723
	update_curr(cfs_rq);
1724
	enqueue_entity_load_avg(cfs_rq, se, flags & ENQUEUE_WAKEUP);
1725 1726
	account_entity_enqueue(cfs_rq, se);
	update_cfs_shares(cfs_rq);
1727

1728
	if (flags & ENQUEUE_WAKEUP) {
1729
		place_entity(cfs_rq, se, 0);
1730
		enqueue_sleeper(cfs_rq, se);
I
Ingo Molnar 已提交
1731
	}
1732

1733
	update_stats_enqueue(cfs_rq, se);
P
Peter Zijlstra 已提交
1734
	check_spread(cfs_rq, se);
1735 1736
	if (se != cfs_rq->curr)
		__enqueue_entity(cfs_rq, se);
P
Peter Zijlstra 已提交
1737
	se->on_rq = 1;
1738

1739
	if (cfs_rq->nr_running == 1) {
1740
		list_add_leaf_cfs_rq(cfs_rq);
1741 1742
		check_enqueue_throttle(cfs_rq);
	}
1743 1744
}

1745
static void __clear_buddies_last(struct sched_entity *se)
P
Peter Zijlstra 已提交
1746
{
1747 1748 1749 1750 1751 1752 1753 1754
	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
1755

1756 1757 1758 1759 1760 1761 1762 1763 1764
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
1765 1766
}

1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777
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
1778 1779
static void clear_buddies(struct cfs_rq *cfs_rq, struct sched_entity *se)
{
1780 1781 1782 1783 1784
	if (cfs_rq->last == se)
		__clear_buddies_last(se);

	if (cfs_rq->next == se)
		__clear_buddies_next(se);
1785 1786 1787

	if (cfs_rq->skip == se)
		__clear_buddies_skip(se);
P
Peter Zijlstra 已提交
1788 1789
}

1790
static __always_inline void return_cfs_rq_runtime(struct cfs_rq *cfs_rq);
1791

1792
static void
1793
dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
1794
{
1795 1796 1797 1798
	/*
	 * Update run-time statistics of the 'current'.
	 */
	update_curr(cfs_rq);
1799
	dequeue_entity_load_avg(cfs_rq, se, flags & DEQUEUE_SLEEP);
1800

1801
	update_stats_dequeue(cfs_rq, se);
1802
	if (flags & DEQUEUE_SLEEP) {
P
Peter Zijlstra 已提交
1803
#ifdef CONFIG_SCHEDSTATS
1804 1805 1806 1807
		if (entity_is_task(se)) {
			struct task_struct *tsk = task_of(se);

			if (tsk->state & TASK_INTERRUPTIBLE)
1808
				se->statistics.sleep_start = rq_of(cfs_rq)->clock;
1809
			if (tsk->state & TASK_UNINTERRUPTIBLE)
1810
				se->statistics.block_start = rq_of(cfs_rq)->clock;
1811
		}
1812
#endif
P
Peter Zijlstra 已提交
1813 1814
	}

P
Peter Zijlstra 已提交
1815
	clear_buddies(cfs_rq, se);
P
Peter Zijlstra 已提交
1816

1817
	if (se != cfs_rq->curr)
1818
		__dequeue_entity(cfs_rq, se);
1819
	se->on_rq = 0;
1820
	account_entity_dequeue(cfs_rq, se);
1821 1822 1823 1824 1825 1826

	/*
	 * 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.
	 */
1827
	if (!(flags & DEQUEUE_SLEEP))
1828
		se->vruntime -= cfs_rq->min_vruntime;
1829

1830 1831 1832
	/* return excess runtime on last dequeue */
	return_cfs_rq_runtime(cfs_rq);

1833
	update_min_vruntime(cfs_rq);
1834
	update_cfs_shares(cfs_rq);
1835 1836 1837 1838 1839
}

/*
 * Preempt the current task with a newly woken task if needed:
 */
1840
static void
I
Ingo Molnar 已提交
1841
check_preempt_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr)
1842
{
1843
	unsigned long ideal_runtime, delta_exec;
1844 1845
	struct sched_entity *se;
	s64 delta;
1846

P
Peter Zijlstra 已提交
1847
	ideal_runtime = sched_slice(cfs_rq, curr);
1848
	delta_exec = curr->sum_exec_runtime - curr->prev_sum_exec_runtime;
1849
	if (delta_exec > ideal_runtime) {
1850
		resched_task(rq_of(cfs_rq)->curr);
1851 1852 1853 1854 1855
		/*
		 * The current task ran long enough, ensure it doesn't get
		 * re-elected due to buddy favours.
		 */
		clear_buddies(cfs_rq, curr);
1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866
		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;

1867 1868
	se = __pick_first_entity(cfs_rq);
	delta = curr->vruntime - se->vruntime;
1869

1870 1871
	if (delta < 0)
		return;
1872

1873 1874
	if (delta > ideal_runtime)
		resched_task(rq_of(cfs_rq)->curr);
1875 1876
}

1877
static void
1878
set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
1879
{
1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890
	/* '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);
	}

1891
	update_stats_curr_start(cfs_rq, se);
1892
	cfs_rq->curr = se;
I
Ingo Molnar 已提交
1893 1894 1895 1896 1897 1898
#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):
	 */
1899
	if (rq_of(cfs_rq)->load.weight >= 2*se->load.weight) {
1900
		se->statistics.slice_max = max(se->statistics.slice_max,
I
Ingo Molnar 已提交
1901 1902 1903
			se->sum_exec_runtime - se->prev_sum_exec_runtime);
	}
#endif
1904
	se->prev_sum_exec_runtime = se->sum_exec_runtime;
1905 1906
}

1907 1908 1909
static int
wakeup_preempt_entity(struct sched_entity *curr, struct sched_entity *se);

1910 1911 1912 1913 1914 1915 1916
/*
 * 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
 */
1917
static struct sched_entity *pick_next_entity(struct cfs_rq *cfs_rq)
1918
{
1919
	struct sched_entity *se = __pick_first_entity(cfs_rq);
1920
	struct sched_entity *left = se;
1921

1922 1923 1924 1925 1926 1927 1928 1929 1930
	/*
	 * 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;
	}
1931

1932 1933 1934 1935 1936 1937
	/*
	 * 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;

1938 1939 1940 1941 1942 1943
	/*
	 * 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;

1944
	clear_buddies(cfs_rq, se);
P
Peter Zijlstra 已提交
1945 1946

	return se;
1947 1948
}

1949 1950
static void check_cfs_rq_runtime(struct cfs_rq *cfs_rq);

1951
static void put_prev_entity(struct cfs_rq *cfs_rq, struct sched_entity *prev)
1952 1953 1954 1955 1956 1957
{
	/*
	 * If still on the runqueue then deactivate_task()
	 * was not called and update_curr() has to be done:
	 */
	if (prev->on_rq)
1958
		update_curr(cfs_rq);
1959

1960 1961 1962
	/* throttle cfs_rqs exceeding runtime */
	check_cfs_rq_runtime(cfs_rq);

P
Peter Zijlstra 已提交
1963
	check_spread(cfs_rq, prev);
1964
	if (prev->on_rq) {
1965
		update_stats_wait_start(cfs_rq, prev);
1966 1967
		/* Put 'current' back into the tree. */
		__enqueue_entity(cfs_rq, prev);
1968
		/* in !on_rq case, update occurred at dequeue */
1969
		update_entity_load_avg(prev, 1);
1970
	}
1971
	cfs_rq->curr = NULL;
1972 1973
}

P
Peter Zijlstra 已提交
1974 1975
static void
entity_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr, int queued)
1976 1977
{
	/*
1978
	 * Update run-time statistics of the 'current'.
1979
	 */
1980
	update_curr(cfs_rq);
1981

1982 1983 1984
	/*
	 * Ensure that runnable average is periodically updated.
	 */
1985
	update_entity_load_avg(curr, 1);
1986
	update_cfs_rq_blocked_load(cfs_rq, 1);
1987

P
Peter Zijlstra 已提交
1988 1989 1990 1991 1992
#ifdef CONFIG_SCHED_HRTICK
	/*
	 * queued ticks are scheduled to match the slice, so don't bother
	 * validating it and just reschedule.
	 */
1993 1994 1995 1996
	if (queued) {
		resched_task(rq_of(cfs_rq)->curr);
		return;
	}
P
Peter Zijlstra 已提交
1997 1998 1999 2000 2001 2002 2003 2004
	/*
	 * 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
2005
	if (cfs_rq->nr_running > 1)
I
Ingo Molnar 已提交
2006
		check_preempt_tick(cfs_rq, curr);
2007 2008
}

2009 2010 2011 2012 2013 2014

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

#ifdef CONFIG_CFS_BANDWIDTH
2015 2016

#ifdef HAVE_JUMP_LABEL
2017
static struct static_key __cfs_bandwidth_used;
2018 2019 2020

static inline bool cfs_bandwidth_used(void)
{
2021
	return static_key_false(&__cfs_bandwidth_used);
2022 2023 2024 2025 2026 2027
}

void account_cfs_bandwidth_used(int enabled, int was_enabled)
{
	/* only need to count groups transitioning between enabled/!enabled */
	if (enabled && !was_enabled)
2028
		static_key_slow_inc(&__cfs_bandwidth_used);
2029
	else if (!enabled && was_enabled)
2030
		static_key_slow_dec(&__cfs_bandwidth_used);
2031 2032 2033 2034 2035 2036 2037 2038 2039 2040
}
#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 */

2041 2042 2043 2044 2045 2046 2047 2048
/*
 * default period for cfs group bandwidth.
 * default: 0.1s, units: nanoseconds
 */
static inline u64 default_cfs_period(void)
{
	return 100000000ULL;
}
2049 2050 2051 2052 2053 2054

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

P
Paul Turner 已提交
2055 2056 2057 2058 2059 2060 2061
/*
 * 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
 */
2062
void __refill_cfs_bandwidth_runtime(struct cfs_bandwidth *cfs_b)
P
Paul Turner 已提交
2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073
{
	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);
}

2074 2075 2076 2077 2078
static inline struct cfs_bandwidth *tg_cfs_bandwidth(struct task_group *tg)
{
	return &tg->cfs_bandwidth;
}

2079 2080 2081 2082 2083 2084 2085 2086 2087
/* rq->task_clock normalized against any time this cfs_rq has spent throttled */
static inline u64 cfs_rq_clock_task(struct cfs_rq *cfs_rq)
{
	if (unlikely(cfs_rq->throttle_count))
		return cfs_rq->throttled_clock_task;

	return rq_of(cfs_rq)->clock_task - cfs_rq->throttled_clock_task_time;
}

2088 2089
/* returns 0 on failure to allocate runtime */
static int assign_cfs_rq_runtime(struct cfs_rq *cfs_rq)
2090 2091 2092
{
	struct task_group *tg = cfs_rq->tg;
	struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(tg);
P
Paul Turner 已提交
2093
	u64 amount = 0, min_amount, expires;
2094 2095 2096 2097 2098 2099 2100

	/* 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;
2101
	else {
P
Paul Turner 已提交
2102 2103 2104 2105 2106 2107 2108 2109
		/*
		 * 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);
2110
			__start_cfs_bandwidth(cfs_b);
P
Paul Turner 已提交
2111
		}
2112 2113 2114 2115 2116 2117

		if (cfs_b->runtime > 0) {
			amount = min(cfs_b->runtime, min_amount);
			cfs_b->runtime -= amount;
			cfs_b->idle = 0;
		}
2118
	}
P
Paul Turner 已提交
2119
	expires = cfs_b->runtime_expires;
2120 2121 2122
	raw_spin_unlock(&cfs_b->lock);

	cfs_rq->runtime_remaining += amount;
P
Paul Turner 已提交
2123 2124 2125 2126 2127 2128 2129
	/*
	 * 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;
2130 2131

	return cfs_rq->runtime_remaining > 0;
2132 2133
}

P
Paul Turner 已提交
2134 2135 2136 2137 2138
/*
 * 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)
2139
{
P
Paul Turner 已提交
2140 2141 2142 2143 2144
	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))
2145 2146
		return;

P
Paul Turner 已提交
2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171
	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) */
2172
	cfs_rq->runtime_remaining -= delta_exec;
P
Paul Turner 已提交
2173 2174 2175
	expire_cfs_rq_runtime(cfs_rq);

	if (likely(cfs_rq->runtime_remaining > 0))
2176 2177
		return;

2178 2179 2180 2181 2182 2183
	/*
	 * 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);
2184 2185
}

2186 2187
static __always_inline
void account_cfs_rq_runtime(struct cfs_rq *cfs_rq, unsigned long delta_exec)
2188
{
2189
	if (!cfs_bandwidth_used() || !cfs_rq->runtime_enabled)
2190 2191 2192 2193 2194
		return;

	__account_cfs_rq_runtime(cfs_rq, delta_exec);
}

2195 2196
static inline int cfs_rq_throttled(struct cfs_rq *cfs_rq)
{
2197
	return cfs_bandwidth_used() && cfs_rq->throttled;
2198 2199
}

2200 2201 2202
/* check whether cfs_rq, or any parent, is throttled */
static inline int throttled_hierarchy(struct cfs_rq *cfs_rq)
{
2203
	return cfs_bandwidth_used() && cfs_rq->throttle_count;
2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231
}

/*
 * 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) {
2232 2233 2234
		/* adjust cfs_rq_clock_task() */
		cfs_rq->throttled_clock_task_time += rq->clock_task -
					     cfs_rq->throttled_clock_task;
2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245
	}
#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)];

2246 2247
	/* group is entering throttled state, stop time */
	if (!cfs_rq->throttle_count)
2248
		cfs_rq->throttled_clock_task = rq->clock_task;
2249 2250 2251 2252 2253
	cfs_rq->throttle_count++;

	return 0;
}

2254
static void throttle_cfs_rq(struct cfs_rq *cfs_rq)
2255 2256 2257 2258 2259 2260 2261 2262
{
	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))];

2263
	/* freeze hierarchy runnable averages while throttled */
2264 2265 2266
	rcu_read_lock();
	walk_tg_tree_from(cfs_rq->tg, tg_throttle_down, tg_nop, (void *)rq);
	rcu_read_unlock();
2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286

	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;
2287
	cfs_rq->throttled_clock = rq->clock;
2288 2289 2290 2291 2292
	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);
}

2293
void unthrottle_cfs_rq(struct cfs_rq *cfs_rq)
2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304
{
	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);
2305
	cfs_b->throttled_time += rq->clock - cfs_rq->throttled_clock;
2306 2307 2308
	list_del_rcu(&cfs_rq->throttled_list);
	raw_spin_unlock(&cfs_b->lock);

2309 2310 2311 2312
	update_rq_clock(rq);
	/* update hierarchical throttle state */
	walk_tg_tree_from(cfs_rq->tg, tg_nop, tg_unthrottle_up, (void *)rq);

2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375
	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;
}

2376 2377 2378 2379 2380 2381 2382 2383
/*
 * 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)
{
2384 2385
	u64 runtime, runtime_expires;
	int idle = 1, throttled;
2386 2387 2388 2389 2390 2391

	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;

2392 2393 2394
	throttled = !list_empty(&cfs_b->throttled_cfs_rq);
	/* idle depends on !throttled (for the case of a large deficit) */
	idle = cfs_b->idle && !throttled;
2395
	cfs_b->nr_periods += overrun;
2396

P
Paul Turner 已提交
2397 2398 2399 2400 2401 2402
	/* if we're going inactive then everything else can be deferred */
	if (idle)
		goto out_unlock;

	__refill_cfs_bandwidth_runtime(cfs_b);

2403 2404 2405 2406 2407 2408
	if (!throttled) {
		/* mark as potentially idle for the upcoming period */
		cfs_b->idle = 1;
		goto out_unlock;
	}

2409 2410 2411
	/* account preceding periods in which throttling occurred */
	cfs_b->nr_throttled += overrun;

2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435
	/*
	 * 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);
	}
2436

2437 2438 2439 2440 2441 2442 2443 2444 2445
	/* 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;
2446 2447 2448 2449 2450 2451 2452
out_unlock:
	if (idle)
		cfs_b->timer_active = 0;
	raw_spin_unlock(&cfs_b->lock);

	return idle;
}
2453

2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517
/* 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)
{
2518 2519 2520
	if (!cfs_bandwidth_used())
		return;

2521
	if (!cfs_rq->runtime_enabled || cfs_rq->nr_running)
2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558
		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);
}

2559 2560 2561 2562 2563 2564 2565
/*
 * 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)
{
2566 2567 2568
	if (!cfs_bandwidth_used())
		return;

2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585
	/* 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)
{
2586 2587 2588
	if (!cfs_bandwidth_used())
		return;

2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600
	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);
}
2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685

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);
}

2686
static void __maybe_unused unthrottle_offline_cfs_rqs(struct rq *rq)
2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706
{
	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 */
2707 2708 2709 2710 2711 2712 2713
static inline u64 cfs_rq_clock_task(struct cfs_rq *cfs_rq)
{
	return rq_of(cfs_rq)->clock_task;
}

static void account_cfs_rq_runtime(struct cfs_rq *cfs_rq,
				     unsigned long delta_exec) {}
2714 2715
static void check_cfs_rq_runtime(struct cfs_rq *cfs_rq) {}
static void check_enqueue_throttle(struct cfs_rq *cfs_rq) {}
2716
static __always_inline void return_cfs_rq_runtime(struct cfs_rq *cfs_rq) {}
2717 2718 2719 2720 2721

static inline int cfs_rq_throttled(struct cfs_rq *cfs_rq)
{
	return 0;
}
2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732

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;
}
2733 2734 2735 2736 2737

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) {}
2738 2739
#endif

2740 2741 2742 2743 2744
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) {}
2745
static inline void unthrottle_offline_cfs_rqs(struct rq *rq) {}
2746 2747 2748

#endif /* CONFIG_CFS_BANDWIDTH */

2749 2750 2751 2752
/**************************************************
 * CFS operations on tasks:
 */

P
Peter Zijlstra 已提交
2753 2754 2755 2756 2757 2758 2759 2760
#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);

2761
	if (cfs_rq->nr_running > 1) {
P
Peter Zijlstra 已提交
2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775
		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.
		 */
2776
		if (rq->curr != p)
2777
			delta = max_t(s64, 10000LL, delta);
P
Peter Zijlstra 已提交
2778

2779
		hrtick_start(rq, delta);
P
Peter Zijlstra 已提交
2780 2781
	}
}
2782 2783 2784 2785 2786 2787 2788 2789 2790 2791

/*
 * 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;

2792
	if (!hrtick_enabled(rq) || curr->sched_class != &fair_sched_class)
2793 2794 2795 2796 2797
		return;

	if (cfs_rq_of(&curr->se)->nr_running < sched_nr_latency)
		hrtick_start_fair(rq, curr);
}
2798
#else /* !CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
2799 2800 2801 2802
static inline void
hrtick_start_fair(struct rq *rq, struct task_struct *p)
{
}
2803 2804 2805 2806

static inline void hrtick_update(struct rq *rq)
{
}
P
Peter Zijlstra 已提交
2807 2808
#endif

2809 2810 2811 2812 2813
/*
 * 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:
 */
2814
static void
2815
enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags)
2816 2817
{
	struct cfs_rq *cfs_rq;
2818
	struct sched_entity *se = &p->se;
2819 2820

	for_each_sched_entity(se) {
2821
		if (se->on_rq)
2822 2823
			break;
		cfs_rq = cfs_rq_of(se);
2824
		enqueue_entity(cfs_rq, se, flags);
2825 2826 2827 2828 2829 2830 2831 2832 2833

		/*
		 * 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;
2834
		cfs_rq->h_nr_running++;
2835

2836
		flags = ENQUEUE_WAKEUP;
2837
	}
P
Peter Zijlstra 已提交
2838

P
Peter Zijlstra 已提交
2839
	for_each_sched_entity(se) {
2840
		cfs_rq = cfs_rq_of(se);
2841
		cfs_rq->h_nr_running++;
P
Peter Zijlstra 已提交
2842

2843 2844 2845
		if (cfs_rq_throttled(cfs_rq))
			break;

2846
		update_cfs_shares(cfs_rq);
2847
		update_entity_load_avg(se, 1);
P
Peter Zijlstra 已提交
2848 2849
	}

2850 2851
	if (!se) {
		update_rq_runnable_avg(rq, rq->nr_running);
2852
		inc_nr_running(rq);
2853
	}
2854
	hrtick_update(rq);
2855 2856
}

2857 2858
static void set_next_buddy(struct sched_entity *se);

2859 2860 2861 2862 2863
/*
 * The dequeue_task method is called before nr_running is
 * decreased. We remove the task from the rbtree and
 * update the fair scheduling stats:
 */
2864
static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags)
2865 2866
{
	struct cfs_rq *cfs_rq;
2867
	struct sched_entity *se = &p->se;
2868
	int task_sleep = flags & DEQUEUE_SLEEP;
2869 2870 2871

	for_each_sched_entity(se) {
		cfs_rq = cfs_rq_of(se);
2872
		dequeue_entity(cfs_rq, se, flags);
2873 2874 2875 2876 2877 2878 2879 2880 2881

		/*
		 * 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;
2882
		cfs_rq->h_nr_running--;
P
Peter Zijlstra 已提交
2883

2884
		/* Don't dequeue parent if it has other entities besides us */
2885 2886 2887 2888 2889 2890 2891
		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));
2892 2893 2894

			/* avoid re-evaluating load for this entity */
			se = parent_entity(se);
2895
			break;
2896
		}
2897
		flags |= DEQUEUE_SLEEP;
2898
	}
P
Peter Zijlstra 已提交
2899

P
Peter Zijlstra 已提交
2900
	for_each_sched_entity(se) {
2901
		cfs_rq = cfs_rq_of(se);
2902
		cfs_rq->h_nr_running--;
P
Peter Zijlstra 已提交
2903

2904 2905 2906
		if (cfs_rq_throttled(cfs_rq))
			break;

2907
		update_cfs_shares(cfs_rq);
2908
		update_entity_load_avg(se, 1);
P
Peter Zijlstra 已提交
2909 2910
	}

2911
	if (!se) {
2912
		dec_nr_running(rq);
2913 2914
		update_rq_runnable_avg(rq, 1);
	}
2915
	hrtick_update(rq);
2916 2917
}

2918
#ifdef CONFIG_SMP
2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973
/* 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;
}

2974

2975
static void task_waking_fair(struct task_struct *p)
2976 2977 2978
{
	struct sched_entity *se = &p->se;
	struct cfs_rq *cfs_rq = cfs_rq_of(se);
2979 2980 2981 2982
	u64 min_vruntime;

#ifndef CONFIG_64BIT
	u64 min_vruntime_copy;
2983

2984 2985 2986 2987 2988 2989 2990 2991
	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
2992

2993
	se->vruntime -= min_vruntime;
2994 2995
}

2996
#ifdef CONFIG_FAIR_GROUP_SCHED
2997 2998 2999 3000 3001 3002
/*
 * 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.
3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045
 *
 * 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.
3046
 */
P
Peter Zijlstra 已提交
3047
static long effective_load(struct task_group *tg, int cpu, long wl, long wg)
3048
{
P
Peter Zijlstra 已提交
3049
	struct sched_entity *se = tg->se[cpu];
3050

3051
	if (!tg->parent)	/* the trivial, non-cgroup case */
3052 3053
		return wl;

P
Peter Zijlstra 已提交
3054
	for_each_sched_entity(se) {
3055
		long w, W;
P
Peter Zijlstra 已提交
3056

3057
		tg = se->my_q->tg;
3058

3059 3060 3061 3062
		/*
		 * W = @wg + \Sum rw_j
		 */
		W = wg + calc_tg_weight(tg, se->my_q);
P
Peter Zijlstra 已提交
3063

3064 3065 3066 3067
		/*
		 * w = rw_i + @wl
		 */
		w = se->my_q->load.weight + wl;
3068

3069 3070 3071 3072 3073
		/*
		 * wl = S * s'_i; see (2)
		 */
		if (W > 0 && w < W)
			wl = (w * tg->shares) / W;
3074 3075
		else
			wl = tg->shares;
3076

3077 3078 3079 3080 3081
		/*
		 * 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().
		 */
3082 3083
		if (wl < MIN_SHARES)
			wl = MIN_SHARES;
3084 3085 3086 3087

		/*
		 * wl = dw_i = S * (s'_i - s_i); see (3)
		 */
3088
		wl -= se->load.weight;
3089 3090 3091 3092 3093 3094 3095 3096

		/*
		 * 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
3097 3098
		wg = 0;
	}
3099

P
Peter Zijlstra 已提交
3100
	return wl;
3101 3102
}
#else
P
Peter Zijlstra 已提交
3103

3104 3105
static inline unsigned long effective_load(struct task_group *tg, int cpu,
		unsigned long wl, unsigned long wg)
P
Peter Zijlstra 已提交
3106
{
3107
	return wl;
3108
}
P
Peter Zijlstra 已提交
3109

3110 3111
#endif

3112
static int wake_affine(struct sched_domain *sd, struct task_struct *p, int sync)
3113
{
3114
	s64 this_load, load;
3115
	int idx, this_cpu, prev_cpu;
3116
	unsigned long tl_per_task;
3117
	struct task_group *tg;
3118
	unsigned long weight;
3119
	int balanced;
3120

3121 3122 3123 3124 3125
	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);
3126

3127 3128 3129 3130 3131
	/*
	 * If sync wakeup then subtract the (maximum possible)
	 * effect of the currently running task from the load
	 * of the current CPU:
	 */
3132 3133 3134 3135
	if (sync) {
		tg = task_group(current);
		weight = current->se.load.weight;

3136
		this_load += effective_load(tg, this_cpu, -weight, -weight);
3137 3138
		load += effective_load(tg, prev_cpu, 0, -weight);
	}
3139

3140 3141
	tg = task_group(p);
	weight = p->se.load.weight;
3142

3143 3144
	/*
	 * In low-load situations, where prev_cpu is idle and this_cpu is idle
3145 3146 3147
	 * 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.
3148 3149 3150 3151
	 *
	 * Otherwise check if either cpus are near enough in load to allow this
	 * task to be woken on this_cpu.
	 */
3152 3153
	if (this_load > 0) {
		s64 this_eff_load, prev_eff_load;
3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166

		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;
3167

3168
	/*
I
Ingo Molnar 已提交
3169 3170 3171
	 * If the currently running task will sleep within
	 * a reasonable amount of time then attract this newly
	 * woken task:
3172
	 */
3173 3174
	if (sync && balanced)
		return 1;
3175

3176
	schedstat_inc(p, se.statistics.nr_wakeups_affine_attempts);
3177 3178
	tl_per_task = cpu_avg_load_per_task(this_cpu);

3179 3180 3181
	if (balanced ||
	    (this_load <= load &&
	     this_load + target_load(prev_cpu, idx) <= tl_per_task)) {
3182 3183 3184 3185 3186
		/*
		 * This domain has SD_WAKE_AFFINE and
		 * p is cache cold in this domain, and
		 * there is no bad imbalance.
		 */
3187
		schedstat_inc(sd, ttwu_move_affine);
3188
		schedstat_inc(p, se.statistics.nr_wakeups_affine);
3189 3190 3191 3192 3193 3194

		return 1;
	}
	return 0;
}

3195 3196 3197 3198 3199
/*
 * find_idlest_group finds and returns the least busy CPU group within the
 * domain.
 */
static struct sched_group *
P
Peter Zijlstra 已提交
3200
find_idlest_group(struct sched_domain *sd, struct task_struct *p,
3201
		  int this_cpu, int load_idx)
3202
{
3203
	struct sched_group *idlest = NULL, *group = sd->groups;
3204 3205
	unsigned long min_load = ULONG_MAX, this_load = 0;
	int imbalance = 100 + (sd->imbalance_pct-100)/2;
3206

3207 3208 3209 3210
	do {
		unsigned long load, avg_load;
		int local_group;
		int i;
3211

3212 3213
		/* Skip over this group if it has no CPUs allowed */
		if (!cpumask_intersects(sched_group_cpus(group),
3214
					tsk_cpus_allowed(p)))
3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233
			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 */
3234
		avg_load = (avg_load * SCHED_POWER_SCALE) / group->sgp->power;
3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259

		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 */
3260
	for_each_cpu_and(i, sched_group_cpus(group), tsk_cpus_allowed(p)) {
3261 3262 3263 3264 3265
		load = weighted_cpuload(i);

		if (load < min_load || (load == min_load && i == this_cpu)) {
			min_load = load;
			idlest = i;
3266 3267 3268
		}
	}

3269 3270
	return idlest;
}
3271

3272 3273 3274
/*
 * Try and locate an idle CPU in the sched_domain.
 */
3275
static int select_idle_sibling(struct task_struct *p, int target)
3276
{
3277
	struct sched_domain *sd;
3278
	struct sched_group *sg;
3279
	int i = task_cpu(p);
3280

3281 3282
	if (idle_cpu(target))
		return target;
3283 3284

	/*
3285
	 * If the prevous cpu is cache affine and idle, don't be stupid.
3286
	 */
3287 3288
	if (i != target && cpus_share_cache(i, target) && idle_cpu(i))
		return i;
3289 3290

	/*
3291
	 * Otherwise, iterate the domains and find an elegible idle cpu.
3292
	 */
3293
	sd = rcu_dereference(per_cpu(sd_llc, target));
3294
	for_each_lower_domain(sd) {
3295 3296 3297 3298 3299 3300 3301
		sg = sd->groups;
		do {
			if (!cpumask_intersects(sched_group_cpus(sg),
						tsk_cpus_allowed(p)))
				goto next;

			for_each_cpu(i, sched_group_cpus(sg)) {
3302
				if (i == target || !idle_cpu(i))
3303 3304
					goto next;
			}
3305

3306 3307 3308 3309 3310 3311 3312 3313
			target = cpumask_first_and(sched_group_cpus(sg),
					tsk_cpus_allowed(p));
			goto done;
next:
			sg = sg->next;
		} while (sg != sd->groups);
	}
done:
3314 3315 3316
	return target;
}

3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327
/*
 * 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.
 */
3328
static int
3329
select_task_rq_fair(struct task_struct *p, int sd_flag, int wake_flags)
3330
{
3331
	struct sched_domain *tmp, *affine_sd = NULL, *sd = NULL;
3332 3333 3334
	int cpu = smp_processor_id();
	int prev_cpu = task_cpu(p);
	int new_cpu = cpu;
3335
	int want_affine = 0;
3336
	int sync = wake_flags & WF_SYNC;
3337

3338
	if (p->nr_cpus_allowed == 1)
3339 3340
		return prev_cpu;

3341
	if (sd_flag & SD_BALANCE_WAKE) {
3342
		if (cpumask_test_cpu(cpu, tsk_cpus_allowed(p)))
3343 3344 3345
			want_affine = 1;
		new_cpu = prev_cpu;
	}
3346

3347
	rcu_read_lock();
3348
	for_each_domain(cpu, tmp) {
3349 3350 3351
		if (!(tmp->flags & SD_LOAD_BALANCE))
			continue;

3352
		/*
3353 3354
		 * If both cpu and prev_cpu are part of this domain,
		 * cpu is a valid SD_WAKE_AFFINE target.
3355
		 */
3356 3357 3358
		if (want_affine && (tmp->flags & SD_WAKE_AFFINE) &&
		    cpumask_test_cpu(prev_cpu, sched_domain_span(tmp))) {
			affine_sd = tmp;
3359
			break;
3360
		}
3361

3362
		if (tmp->flags & sd_flag)
3363 3364 3365
			sd = tmp;
	}

3366
	if (affine_sd) {
3367
		if (cpu != prev_cpu && wake_affine(affine_sd, p, sync))
3368 3369 3370 3371
			prev_cpu = cpu;

		new_cpu = select_idle_sibling(p, prev_cpu);
		goto unlock;
3372
	}
3373

3374
	while (sd) {
3375
		int load_idx = sd->forkexec_idx;
3376
		struct sched_group *group;
3377
		int weight;
3378

3379
		if (!(sd->flags & sd_flag)) {
3380 3381 3382
			sd = sd->child;
			continue;
		}
3383

3384 3385
		if (sd_flag & SD_BALANCE_WAKE)
			load_idx = sd->wake_idx;
3386

3387
		group = find_idlest_group(sd, p, cpu, load_idx);
3388 3389 3390 3391
		if (!group) {
			sd = sd->child;
			continue;
		}
I
Ingo Molnar 已提交
3392

3393
		new_cpu = find_idlest_cpu(group, p, cpu);
3394 3395 3396 3397
		if (new_cpu == -1 || new_cpu == cpu) {
			/* Now try balancing at a lower domain level of cpu */
			sd = sd->child;
			continue;
3398
		}
3399 3400 3401

		/* Now try balancing at a lower domain level of new_cpu */
		cpu = new_cpu;
3402
		weight = sd->span_weight;
3403 3404
		sd = NULL;
		for_each_domain(cpu, tmp) {
3405
			if (weight <= tmp->span_weight)
3406
				break;
3407
			if (tmp->flags & sd_flag)
3408 3409 3410
				sd = tmp;
		}
		/* while loop will break here if sd == NULL */
3411
	}
3412 3413
unlock:
	rcu_read_unlock();
3414

3415
	return new_cpu;
3416
}
3417

3418 3419 3420 3421 3422 3423
/*
 * Load-tracking only depends on SMP, FAIR_GROUP_SCHED dependency below may be
 * removed when useful for applications beyond shares distribution (e.g.
 * load-balance).
 */
#ifdef CONFIG_FAIR_GROUP_SCHED
3424 3425 3426 3427 3428 3429 3430 3431 3432
/*
 * Called immediately before a task is migrated to a new cpu; task_cpu(p) and
 * cfs_rq_of(p) references at time of call are still valid and identify the
 * previous cpu.  However, the caller only guarantees p->pi_lock is held; no
 * other assumptions, including the state of rq->lock, should be made.
 */
static void
migrate_task_rq_fair(struct task_struct *p, int next_cpu)
{
3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445
	struct sched_entity *se = &p->se;
	struct cfs_rq *cfs_rq = cfs_rq_of(se);

	/*
	 * Load tracking: accumulate removed load so that it can be processed
	 * when we next update owning cfs_rq under rq->lock.  Tasks contribute
	 * to blocked load iff they have a positive decay-count.  It can never
	 * be negative here since on-rq tasks have decay-count == 0.
	 */
	if (se->avg.decay_count) {
		se->avg.decay_count = -__synchronize_entity_decay(se);
		atomic64_add(se->avg.load_avg_contrib, &cfs_rq->removed_load);
	}
3446
}
3447
#endif
3448 3449
#endif /* CONFIG_SMP */

P
Peter Zijlstra 已提交
3450 3451
static unsigned long
wakeup_gran(struct sched_entity *curr, struct sched_entity *se)
3452 3453 3454 3455
{
	unsigned long gran = sysctl_sched_wakeup_granularity;

	/*
P
Peter Zijlstra 已提交
3456 3457
	 * Since its curr running now, convert the gran from real-time
	 * to virtual-time in his units.
M
Mike Galbraith 已提交
3458 3459 3460 3461 3462 3463 3464 3465 3466
	 *
	 * 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.
3467
	 */
3468
	return calc_delta_fair(gran, se);
3469 3470
}

3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492
/*
 * 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
3493
	gran = wakeup_gran(curr, se);
3494 3495 3496 3497 3498 3499
	if (vdiff > gran)
		return 1;

	return 0;
}

3500 3501
static void set_last_buddy(struct sched_entity *se)
{
3502 3503 3504 3505 3506
	if (entity_is_task(se) && unlikely(task_of(se)->policy == SCHED_IDLE))
		return;

	for_each_sched_entity(se)
		cfs_rq_of(se)->last = se;
3507 3508 3509 3510
}

static void set_next_buddy(struct sched_entity *se)
{
3511 3512 3513 3514 3515
	if (entity_is_task(se) && unlikely(task_of(se)->policy == SCHED_IDLE))
		return;

	for_each_sched_entity(se)
		cfs_rq_of(se)->next = se;
3516 3517
}

3518 3519
static void set_skip_buddy(struct sched_entity *se)
{
3520 3521
	for_each_sched_entity(se)
		cfs_rq_of(se)->skip = se;
3522 3523
}

3524 3525 3526
/*
 * Preempt the current task with a newly woken task if needed:
 */
3527
static void check_preempt_wakeup(struct rq *rq, struct task_struct *p, int wake_flags)
3528 3529
{
	struct task_struct *curr = rq->curr;
3530
	struct sched_entity *se = &curr->se, *pse = &p->se;
3531
	struct cfs_rq *cfs_rq = task_cfs_rq(curr);
3532
	int scale = cfs_rq->nr_running >= sched_nr_latency;
3533
	int next_buddy_marked = 0;
3534

I
Ingo Molnar 已提交
3535 3536 3537
	if (unlikely(se == pse))
		return;

3538
	/*
3539
	 * This is possible from callers such as move_task(), in which we
3540 3541 3542 3543 3544 3545 3546
	 * 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;

3547
	if (sched_feat(NEXT_BUDDY) && scale && !(wake_flags & WF_FORK)) {
M
Mike Galbraith 已提交
3548
		set_next_buddy(pse);
3549 3550
		next_buddy_marked = 1;
	}
P
Peter Zijlstra 已提交
3551

3552 3553 3554
	/*
	 * We can come here with TIF_NEED_RESCHED already set from new task
	 * wake up path.
3555 3556 3557 3558 3559 3560
	 *
	 * 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.
3561 3562 3563 3564
	 */
	if (test_tsk_need_resched(curr))
		return;

3565 3566 3567 3568 3569
	/* Idle tasks are by definition preempted by non-idle tasks. */
	if (unlikely(curr->policy == SCHED_IDLE) &&
	    likely(p->policy != SCHED_IDLE))
		goto preempt;

3570
	/*
3571 3572
	 * Batch and idle tasks do not preempt non-idle tasks (their preemption
	 * is driven by the tick):
3573
	 */
3574
	if (unlikely(p->policy != SCHED_NORMAL) || !sched_feat(WAKEUP_PREEMPTION))
3575
		return;
3576

3577
	find_matching_se(&se, &pse);
3578
	update_curr(cfs_rq_of(se));
3579
	BUG_ON(!pse);
3580 3581 3582 3583 3584 3585 3586
	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);
3587
		goto preempt;
3588
	}
3589

3590
	return;
3591

3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607
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);
3608 3609
}

3610
static struct task_struct *pick_next_task_fair(struct rq *rq)
3611
{
P
Peter Zijlstra 已提交
3612
	struct task_struct *p;
3613 3614 3615
	struct cfs_rq *cfs_rq = &rq->cfs;
	struct sched_entity *se;

3616
	if (!cfs_rq->nr_running)
3617 3618 3619
		return NULL;

	do {
3620
		se = pick_next_entity(cfs_rq);
3621
		set_next_entity(cfs_rq, se);
3622 3623 3624
		cfs_rq = group_cfs_rq(se);
	} while (cfs_rq);

P
Peter Zijlstra 已提交
3625
	p = task_of(se);
3626 3627
	if (hrtick_enabled(rq))
		hrtick_start_fair(rq, p);
P
Peter Zijlstra 已提交
3628 3629

	return p;
3630 3631 3632 3633 3634
}

/*
 * Account for a descheduled task:
 */
3635
static void put_prev_task_fair(struct rq *rq, struct task_struct *prev)
3636 3637 3638 3639 3640 3641
{
	struct sched_entity *se = &prev->se;
	struct cfs_rq *cfs_rq;

	for_each_sched_entity(se) {
		cfs_rq = cfs_rq_of(se);
3642
		put_prev_entity(cfs_rq, se);
3643 3644 3645
	}
}

3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670
/*
 * 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);
3671 3672 3673 3674 3675 3676
		/*
		 * 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;
3677 3678 3679 3680 3681
	}

	set_skip_buddy(se);
}

3682 3683 3684 3685
static bool yield_to_task_fair(struct rq *rq, struct task_struct *p, bool preempt)
{
	struct sched_entity *se = &p->se;

3686 3687
	/* throttled hierarchies are not runnable */
	if (!se->on_rq || throttled_hierarchy(cfs_rq_of(se)))
3688 3689 3690 3691 3692 3693 3694 3695 3696 3697
		return false;

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

	yield_task_fair(rq);

	return true;
}

3698
#ifdef CONFIG_SMP
3699
/**************************************************
P
Peter Zijlstra 已提交
3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815
 * Fair scheduling class load-balancing methods.
 *
 * BASICS
 *
 * The purpose of load-balancing is to achieve the same basic fairness the
 * per-cpu scheduler provides, namely provide a proportional amount of compute
 * time to each task. This is expressed in the following equation:
 *
 *   W_i,n/P_i == W_j,n/P_j for all i,j                               (1)
 *
 * Where W_i,n is the n-th weight average for cpu i. The instantaneous weight
 * W_i,0 is defined as:
 *
 *   W_i,0 = \Sum_j w_i,j                                             (2)
 *
 * Where w_i,j is the weight of the j-th runnable task on cpu i. This weight
 * is derived from the nice value as per prio_to_weight[].
 *
 * The weight average is an exponential decay average of the instantaneous
 * weight:
 *
 *   W'_i,n = (2^n - 1) / 2^n * W_i,n + 1 / 2^n * W_i,0               (3)
 *
 * P_i is the cpu power (or compute capacity) of cpu i, typically it is the
 * fraction of 'recent' time available for SCHED_OTHER task execution. But it
 * can also include other factors [XXX].
 *
 * To achieve this balance we define a measure of imbalance which follows
 * directly from (1):
 *
 *   imb_i,j = max{ avg(W/P), W_i/P_i } - min{ avg(W/P), W_j/P_j }    (4)
 *
 * We them move tasks around to minimize the imbalance. In the continuous
 * function space it is obvious this converges, in the discrete case we get
 * a few fun cases generally called infeasible weight scenarios.
 *
 * [XXX expand on:
 *     - infeasible weights;
 *     - local vs global optima in the discrete case. ]
 *
 *
 * SCHED DOMAINS
 *
 * In order to solve the imbalance equation (4), and avoid the obvious O(n^2)
 * for all i,j solution, we create a tree of cpus that follows the hardware
 * topology where each level pairs two lower groups (or better). This results
 * in O(log n) layers. Furthermore we reduce the number of cpus going up the
 * tree to only the first of the previous level and we decrease the frequency
 * of load-balance at each level inv. proportional to the number of cpus in
 * the groups.
 *
 * This yields:
 *
 *     log_2 n     1     n
 *   \Sum       { --- * --- * 2^i } = O(n)                            (5)
 *     i = 0      2^i   2^i
 *                               `- size of each group
 *         |         |     `- number of cpus doing load-balance
 *         |         `- freq
 *         `- sum over all levels
 *
 * Coupled with a limit on how many tasks we can migrate every balance pass,
 * this makes (5) the runtime complexity of the balancer.
 *
 * An important property here is that each CPU is still (indirectly) connected
 * to every other cpu in at most O(log n) steps:
 *
 * The adjacency matrix of the resulting graph is given by:
 *
 *             log_2 n     
 *   A_i,j = \Union     (i % 2^k == 0) && i / 2^(k+1) == j / 2^(k+1)  (6)
 *             k = 0
 *
 * And you'll find that:
 *
 *   A^(log_2 n)_i,j != 0  for all i,j                                (7)
 *
 * Showing there's indeed a path between every cpu in at most O(log n) steps.
 * The task movement gives a factor of O(m), giving a convergence complexity
 * of:
 *
 *   O(nm log n),  n := nr_cpus, m := nr_tasks                        (8)
 *
 *
 * WORK CONSERVING
 *
 * In order to avoid CPUs going idle while there's still work to do, new idle
 * balancing is more aggressive and has the newly idle cpu iterate up the domain
 * tree itself instead of relying on other CPUs to bring it work.
 *
 * This adds some complexity to both (5) and (8) but it reduces the total idle
 * time.
 *
 * [XXX more?]
 *
 *
 * CGROUPS
 *
 * Cgroups make a horror show out of (2), instead of a simple sum we get:
 *
 *                                s_k,i
 *   W_i,0 = \Sum_j \Prod_k w_k * -----                               (9)
 *                                 S_k
 *
 * Where
 *
 *   s_k,i = \Sum_j w_i,j,k  and  S_k = \Sum_i s_k,i                 (10)
 *
 * w_i,j,k is the weight of the j-th runnable task in the k-th cgroup on cpu i.
 *
 * The big problem is S_k, its a global sum needed to compute a local (W_i)
 * property.
 *
 * [XXX write more on how we solve this.. _after_ merging pjt's patches that
 *      rewrite all of this once again.]
 */ 
3816

3817 3818
static unsigned long __read_mostly max_load_balance_interval = HZ/10;

3819
#define LBF_ALL_PINNED	0x01
3820
#define LBF_NEED_BREAK	0x02
3821
#define LBF_SOME_PINNED 0x04
3822 3823 3824 3825 3826

struct lb_env {
	struct sched_domain	*sd;

	struct rq		*src_rq;
3827
	int			src_cpu;
3828 3829 3830 3831

	int			dst_cpu;
	struct rq		*dst_rq;

3832 3833
	struct cpumask		*dst_grpmask;
	int			new_dst_cpu;
3834
	enum cpu_idle_type	idle;
3835
	long			imbalance;
3836 3837 3838
	/* The set of CPUs under consideration for load-balancing */
	struct cpumask		*cpus;

3839
	unsigned int		flags;
3840 3841 3842 3843

	unsigned int		loop;
	unsigned int		loop_break;
	unsigned int		loop_max;
3844 3845
};

3846
/*
3847
 * move_task - move a task from one runqueue to another runqueue.
3848 3849
 * Both runqueues must be locked.
 */
3850
static void move_task(struct task_struct *p, struct lb_env *env)
3851
{
3852 3853 3854 3855
	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);
3856 3857
}

3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889
/*
 * 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;
}

3890 3891 3892 3893
/*
 * can_migrate_task - may task p from runqueue rq be migrated to this_cpu?
 */
static
3894
int can_migrate_task(struct task_struct *p, struct lb_env *env)
3895 3896 3897 3898
{
	int tsk_cache_hot = 0;
	/*
	 * We do not migrate tasks that are:
3899
	 * 1) throttled_lb_pair, or
3900
	 * 2) cannot be migrated to this CPU due to cpus_allowed, or
3901 3902
	 * 3) running (obviously), or
	 * 4) are cache-hot on their current CPU.
3903
	 */
3904 3905 3906
	if (throttled_lb_pair(task_group(p), env->src_cpu, env->dst_cpu))
		return 0;

3907
	if (!cpumask_test_cpu(env->dst_cpu, tsk_cpus_allowed(p))) {
3908
		int cpu;
3909

3910
		schedstat_inc(p, se.statistics.nr_failed_migrations_affine);
3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922

		/*
		 * 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;

3923 3924 3925 3926 3927 3928 3929
		/* Prevent to re-select dst_cpu via env's cpus */
		for_each_cpu_and(cpu, env->dst_grpmask, env->cpus) {
			if (cpumask_test_cpu(cpu, tsk_cpus_allowed(p))) {
				env->flags |= LBF_SOME_PINNED;
				env->new_dst_cpu = cpu;
				break;
			}
3930
		}
3931

3932 3933
		return 0;
	}
3934 3935

	/* Record that we found atleast one task that could run on dst_cpu */
3936
	env->flags &= ~LBF_ALL_PINNED;
3937

3938
	if (task_running(env->src_rq, p)) {
3939
		schedstat_inc(p, se.statistics.nr_failed_migrations_running);
3940 3941 3942 3943 3944 3945 3946 3947 3948
		return 0;
	}

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

3949
	tsk_cache_hot = task_hot(p, env->src_rq->clock_task, env->sd);
3950
	if (!tsk_cache_hot ||
3951
		env->sd->nr_balance_failed > env->sd->cache_nice_tries) {
Z
Zhang Hang 已提交
3952

3953
		if (tsk_cache_hot) {
3954
			schedstat_inc(env->sd, lb_hot_gained[env->idle]);
3955
			schedstat_inc(p, se.statistics.nr_forced_migrations);
3956
		}
Z
Zhang Hang 已提交
3957

3958 3959 3960
		return 1;
	}

Z
Zhang Hang 已提交
3961 3962
	schedstat_inc(p, se.statistics.nr_failed_migrations_hot);
	return 0;
3963 3964
}

3965 3966 3967 3968 3969 3970 3971
/*
 * 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.
 */
3972
static int move_one_task(struct lb_env *env)
3973 3974 3975
{
	struct task_struct *p, *n;

3976 3977 3978
	list_for_each_entry_safe(p, n, &env->src_rq->cfs_tasks, se.group_node) {
		if (!can_migrate_task(p, env))
			continue;
3979

3980 3981 3982 3983 3984 3985 3986 3987
		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;
3988 3989 3990 3991
	}
	return 0;
}

3992 3993
static unsigned long task_h_load(struct task_struct *p);

3994 3995
static const unsigned int sched_nr_migrate_break = 32;

3996
/*
3997
 * move_tasks tries to move up to imbalance weighted load from busiest to
3998 3999 4000 4001 4002 4003
 * 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)
4004
{
4005 4006
	struct list_head *tasks = &env->src_rq->cfs_tasks;
	struct task_struct *p;
4007 4008
	unsigned long load;
	int pulled = 0;
4009

4010
	if (env->imbalance <= 0)
4011
		return 0;
4012

4013 4014
	while (!list_empty(tasks)) {
		p = list_first_entry(tasks, struct task_struct, se.group_node);
4015

4016 4017
		env->loop++;
		/* We've more or less seen every task there is, call it quits */
4018
		if (env->loop > env->loop_max)
4019
			break;
4020 4021

		/* take a breather every nr_migrate tasks */
4022
		if (env->loop > env->loop_break) {
4023
			env->loop_break += sched_nr_migrate_break;
4024
			env->flags |= LBF_NEED_BREAK;
4025
			break;
4026
		}
4027

4028
		if (!can_migrate_task(p, env))
4029 4030 4031
			goto next;

		load = task_h_load(p);
4032

4033
		if (sched_feat(LB_MIN) && load < 16 && !env->sd->nr_balance_failed)
4034 4035
			goto next;

4036
		if ((load / 2) > env->imbalance)
4037
			goto next;
4038

4039
		move_task(p, env);
4040
		pulled++;
4041
		env->imbalance -= load;
4042 4043

#ifdef CONFIG_PREEMPT
4044 4045 4046 4047 4048
		/*
		 * NEWIDLE balancing is a source of latency, so preemptible
		 * kernels will stop after the first task is pulled to minimize
		 * the critical section.
		 */
4049
		if (env->idle == CPU_NEWLY_IDLE)
4050
			break;
4051 4052
#endif

4053 4054 4055 4056
		/*
		 * We only want to steal up to the prescribed amount of
		 * weighted load.
		 */
4057
		if (env->imbalance <= 0)
4058
			break;
4059 4060 4061

		continue;
next:
4062
		list_move_tail(&p->se.group_node, tasks);
4063
	}
4064

4065
	/*
4066 4067 4068
	 * 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().
4069
	 */
4070
	schedstat_add(env->sd, lb_gained[env->idle], pulled);
4071

4072
	return pulled;
4073 4074
}

P
Peter Zijlstra 已提交
4075
#ifdef CONFIG_FAIR_GROUP_SCHED
4076 4077 4078
/*
 * update tg->load_weight by folding this cpu's load_avg
 */
4079
static void __update_blocked_averages_cpu(struct task_group *tg, int cpu)
4080
{
4081 4082
	struct sched_entity *se = tg->se[cpu];
	struct cfs_rq *cfs_rq = tg->cfs_rq[cpu];
4083

4084 4085 4086
	/* throttled entities do not contribute to load */
	if (throttled_hierarchy(cfs_rq))
		return;
4087

4088
	update_cfs_rq_blocked_load(cfs_rq, 1);
4089

4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103
	if (se) {
		update_entity_load_avg(se, 1);
		/*
		 * We pivot on our runnable average having decayed to zero for
		 * list removal.  This generally implies that all our children
		 * have also been removed (modulo rounding error or bandwidth
		 * control); however, such cases are rare and we can fix these
		 * at enqueue.
		 *
		 * TODO: fix up out-of-order children on enqueue.
		 */
		if (!se->avg.runnable_avg_sum && !cfs_rq->nr_running)
			list_del_leaf_cfs_rq(cfs_rq);
	} else {
4104
		struct rq *rq = rq_of(cfs_rq);
4105 4106
		update_rq_runnable_avg(rq, rq->nr_running);
	}
4107 4108
}

4109
static void update_blocked_averages(int cpu)
4110 4111
{
	struct rq *rq = cpu_rq(cpu);
4112 4113
	struct cfs_rq *cfs_rq;
	unsigned long flags;
4114

4115 4116
	raw_spin_lock_irqsave(&rq->lock, flags);
	update_rq_clock(rq);
4117 4118 4119 4120
	/*
	 * Iterates the task_group tree in a bottom up fashion, see
	 * list_add_leaf_cfs_rq() for details.
	 */
4121
	for_each_leaf_cfs_rq(rq, cfs_rq) {
4122 4123 4124 4125 4126 4127
		/*
		 * Note: We may want to consider periodically releasing
		 * rq->lock about these updates so that creating many task
		 * groups does not result in continually extending hold time.
		 */
		__update_blocked_averages_cpu(cfs_rq->tg, rq->cpu);
4128
	}
4129 4130

	raw_spin_unlock_irqrestore(&rq->lock, flags);
4131 4132
}

4133 4134 4135 4136 4137 4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153 4154 4155 4156 4157
/*
 * 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)
{
4158 4159 4160 4161 4162 4163 4164 4165
	struct rq *rq = cpu_rq(cpu);
	unsigned long now = jiffies;

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

	rq->h_load_throttle = now;

4166
	rcu_read_lock();
4167
	walk_tg_tree(tg_load_down, tg_nop, (void *)cpu);
4168
	rcu_read_unlock();
4169 4170
}

4171
static unsigned long task_h_load(struct task_struct *p)
P
Peter Zijlstra 已提交
4172
{
4173 4174
	struct cfs_rq *cfs_rq = task_cfs_rq(p);
	unsigned long load;
P
Peter Zijlstra 已提交
4175

4176 4177
	load = p->se.load.weight;
	load = div_u64(load * cfs_rq->h_load, cfs_rq->load.weight + 1);
P
Peter Zijlstra 已提交
4178

4179
	return load;
P
Peter Zijlstra 已提交
4180 4181
}
#else
4182
static inline void update_blocked_averages(int cpu)
4183 4184 4185
{
}

4186
static inline void update_h_load(long cpu)
P
Peter Zijlstra 已提交
4187 4188 4189
{
}

4190
static unsigned long task_h_load(struct task_struct *p)
4191
{
4192
	return p->se.load.weight;
4193
}
P
Peter Zijlstra 已提交
4194
#endif
4195 4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210 4211

/********** 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;
4212
	unsigned long this_has_capacity;
4213
	unsigned int  this_idle_cpus;
4214 4215

	/* Statistics of the busiest group */
4216
	unsigned int  busiest_idle_cpus;
4217 4218 4219
	unsigned long max_load;
	unsigned long busiest_load_per_task;
	unsigned long busiest_nr_running;
4220
	unsigned long busiest_group_capacity;
4221
	unsigned long busiest_has_capacity;
4222
	unsigned int  busiest_group_weight;
4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235

	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;
4236 4237
	unsigned long idle_cpus;
	unsigned long group_weight;
4238
	int group_imb; /* Is there an imbalance in the group ? */
4239
	int group_has_capacity; /* Is there extra capacity in the group? */
4240 4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267
};

/**
 * 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;
}

4268
static unsigned long default_scale_freq_power(struct sched_domain *sd, int cpu)
4269
{
4270
	return SCHED_POWER_SCALE;
4271 4272 4273 4274 4275 4276 4277
}

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

4278
static unsigned long default_scale_smt_power(struct sched_domain *sd, int cpu)
4279
{
4280
	unsigned long weight = sd->span_weight;
4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292
	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);
}

4293
static unsigned long scale_rt_power(int cpu)
4294 4295
{
	struct rq *rq = cpu_rq(cpu);
4296
	u64 total, available, age_stamp, avg;
4297

4298 4299 4300 4301 4302 4303 4304 4305
	/*
	 * 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);
4306

4307
	if (unlikely(total < avg)) {
4308 4309 4310
		/* Ensures that power won't end up being negative */
		available = 0;
	} else {
4311
		available = total - avg;
4312
	}
4313

4314 4315
	if (unlikely((s64)total < SCHED_POWER_SCALE))
		total = SCHED_POWER_SCALE;
4316

4317
	total >>= SCHED_POWER_SHIFT;
4318 4319 4320 4321 4322 4323

	return div_u64(available, total);
}

static void update_cpu_power(struct sched_domain *sd, int cpu)
{
4324
	unsigned long weight = sd->span_weight;
4325
	unsigned long power = SCHED_POWER_SCALE;
4326 4327 4328 4329 4330 4331 4332 4333
	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);

4334
		power >>= SCHED_POWER_SHIFT;
4335 4336
	}

4337
	sdg->sgp->power_orig = power;
4338 4339 4340 4341 4342 4343

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

4344
	power >>= SCHED_POWER_SHIFT;
4345

4346
	power *= scale_rt_power(cpu);
4347
	power >>= SCHED_POWER_SHIFT;
4348 4349 4350 4351

	if (!power)
		power = 1;

4352
	cpu_rq(cpu)->cpu_power = power;
4353
	sdg->sgp->power = power;
4354 4355
}

4356
void update_group_power(struct sched_domain *sd, int cpu)
4357 4358 4359 4360
{
	struct sched_domain *child = sd->child;
	struct sched_group *group, *sdg = sd->groups;
	unsigned long power;
4361 4362 4363 4364 4365
	unsigned long interval;

	interval = msecs_to_jiffies(sd->balance_interval);
	interval = clamp(interval, 1UL, max_load_balance_interval);
	sdg->sgp->next_update = jiffies + interval;
4366 4367 4368 4369 4370 4371 4372 4373

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

	power = 0;

P
Peter Zijlstra 已提交
4374 4375 4376 4377 4378 4379 4380 4381 4382 4383 4384 4385 4386 4387 4388 4389 4390 4391 4392 4393
	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);
	}
4394

4395
	sdg->sgp->power_orig = sdg->sgp->power = power;
4396 4397
}

4398 4399 4400 4401 4402 4403 4404 4405 4406 4407 4408
/*
 * 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)
{
	/*
4409
	 * Only siblings can have significantly less than SCHED_POWER_SCALE
4410
	 */
P
Peter Zijlstra 已提交
4411
	if (!(sd->flags & SD_SHARE_CPUPOWER))
4412 4413 4414 4415 4416
		return 0;

	/*
	 * If ~90% of the cpu_power is still there, we're good.
	 */
4417
	if (group->sgp->power * 32 > group->sgp->power_orig * 29)
4418 4419 4420 4421 4422
		return 1;

	return 0;
}

4423 4424
/**
 * update_sg_lb_stats - Update sched_group's statistics for load balancing.
4425
 * @env: The load balancing environment.
4426 4427 4428 4429 4430 4431
 * @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.
 */
4432 4433
static inline void update_sg_lb_stats(struct lb_env *env,
			struct sched_group *group, int load_idx,
4434
			int local_group, int *balance, struct sg_lb_stats *sgs)
4435
{
4436 4437
	unsigned long nr_running, max_nr_running, min_nr_running;
	unsigned long load, max_cpu_load, min_cpu_load;
4438
	unsigned int balance_cpu = -1, first_idle_cpu = 0;
4439
	unsigned long avg_load_per_task = 0;
4440
	int i;
4441

4442
	if (local_group)
P
Peter Zijlstra 已提交
4443
		balance_cpu = group_balance_cpu(group);
4444 4445 4446 4447

	/* Tally up the load of all CPUs in the group */
	max_cpu_load = 0;
	min_cpu_load = ~0UL;
4448
	max_nr_running = 0;
4449
	min_nr_running = ~0UL;
4450

4451
	for_each_cpu_and(i, sched_group_cpus(group), env->cpus) {
4452 4453
		struct rq *rq = cpu_rq(i);

4454 4455
		nr_running = rq->nr_running;

4456 4457
		/* Bias balancing toward cpus of our domain */
		if (local_group) {
P
Peter Zijlstra 已提交
4458 4459
			if (idle_cpu(i) && !first_idle_cpu &&
					cpumask_test_cpu(i, sched_group_mask(group))) {
4460
				first_idle_cpu = 1;
4461 4462
				balance_cpu = i;
			}
4463 4464

			load = target_load(i, load_idx);
4465 4466
		} else {
			load = source_load(i, load_idx);
4467
			if (load > max_cpu_load)
4468 4469 4470
				max_cpu_load = load;
			if (min_cpu_load > load)
				min_cpu_load = load;
4471 4472 4473 4474 4475

			if (nr_running > max_nr_running)
				max_nr_running = nr_running;
			if (min_nr_running > nr_running)
				min_nr_running = nr_running;
4476 4477 4478
		}

		sgs->group_load += load;
4479
		sgs->sum_nr_running += nr_running;
4480
		sgs->sum_weighted_load += weighted_cpuload(i);
4481 4482
		if (idle_cpu(i))
			sgs->idle_cpus++;
4483 4484 4485 4486 4487 4488 4489 4490
	}

	/*
	 * 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.
	 */
4491
	if (local_group) {
4492
		if (env->idle != CPU_NEWLY_IDLE) {
4493
			if (balance_cpu != env->dst_cpu) {
4494 4495 4496
				*balance = 0;
				return;
			}
4497
			update_group_power(env->sd, env->dst_cpu);
4498
		} else if (time_after_eq(jiffies, group->sgp->next_update))
4499
			update_group_power(env->sd, env->dst_cpu);
4500 4501 4502
	}

	/* Adjust by relative CPU power of the group */
4503
	sgs->avg_load = (sgs->group_load*SCHED_POWER_SCALE) / group->sgp->power;
4504 4505 4506

	/*
	 * Consider the group unbalanced when the imbalance is larger
P
Peter Zijlstra 已提交
4507
	 * than the average weight of a task.
4508 4509 4510 4511 4512 4513
	 *
	 * 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?
	 */
4514 4515
	if (sgs->sum_nr_running)
		avg_load_per_task = sgs->sum_weighted_load / sgs->sum_nr_running;
4516

4517 4518
	if ((max_cpu_load - min_cpu_load) >= avg_load_per_task &&
	    (max_nr_running - min_nr_running) > 1)
4519 4520
		sgs->group_imb = 1;

4521
	sgs->group_capacity = DIV_ROUND_CLOSEST(group->sgp->power,
4522
						SCHED_POWER_SCALE);
4523
	if (!sgs->group_capacity)
4524
		sgs->group_capacity = fix_small_capacity(env->sd, group);
4525
	sgs->group_weight = group->group_weight;
4526 4527 4528

	if (sgs->group_capacity > sgs->sum_nr_running)
		sgs->group_has_capacity = 1;
4529 4530
}

4531 4532
/**
 * update_sd_pick_busiest - return 1 on busiest group
4533
 * @env: The load balancing environment.
4534 4535
 * @sds: sched_domain statistics
 * @sg: sched_group candidate to be checked for being the busiest
4536
 * @sgs: sched_group statistics
4537 4538 4539 4540
 *
 * Determine if @sg is a busier group than the previously selected
 * busiest group.
 */
4541
static bool update_sd_pick_busiest(struct lb_env *env,
4542 4543
				   struct sd_lb_stats *sds,
				   struct sched_group *sg,
4544
				   struct sg_lb_stats *sgs)
4545 4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559
{
	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.
	 */
4560 4561
	if ((env->sd->flags & SD_ASYM_PACKING) && sgs->sum_nr_running &&
	    env->dst_cpu < group_first_cpu(sg)) {
4562 4563 4564 4565 4566 4567 4568 4569 4570 4571
		if (!sds->busiest)
			return true;

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

	return false;
}

4572
/**
4573
 * update_sd_lb_stats - Update sched_domain's statistics for load balancing.
4574
 * @env: The load balancing environment.
4575 4576 4577
 * @balance: Should we balance.
 * @sds: variable to hold the statistics for this sched_domain.
 */
4578
static inline void update_sd_lb_stats(struct lb_env *env,
4579
					int *balance, struct sd_lb_stats *sds)
4580
{
4581 4582
	struct sched_domain *child = env->sd->child;
	struct sched_group *sg = env->sd->groups;
4583 4584 4585 4586 4587 4588
	struct sg_lb_stats sgs;
	int load_idx, prefer_sibling = 0;

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

4589
	load_idx = get_sd_load_idx(env->sd, env->idle);
4590 4591 4592 4593

	do {
		int local_group;

4594
		local_group = cpumask_test_cpu(env->dst_cpu, sched_group_cpus(sg));
4595
		memset(&sgs, 0, sizeof(sgs));
4596
		update_sg_lb_stats(env, sg, load_idx, local_group, balance, &sgs);
4597

P
Peter Zijlstra 已提交
4598
		if (local_group && !(*balance))
4599 4600 4601
			return;

		sds->total_load += sgs.group_load;
4602
		sds->total_pwr += sg->sgp->power;
4603 4604 4605

		/*
		 * In case the child domain prefers tasks go to siblings
4606
		 * first, lower the sg capacity to one so that we'll try
4607 4608 4609 4610 4611 4612
		 * 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).
4613
		 */
4614
		if (prefer_sibling && !local_group && sds->this_has_capacity)
4615 4616 4617 4618
			sgs.group_capacity = min(sgs.group_capacity, 1UL);

		if (local_group) {
			sds->this_load = sgs.avg_load;
4619
			sds->this = sg;
4620 4621
			sds->this_nr_running = sgs.sum_nr_running;
			sds->this_load_per_task = sgs.sum_weighted_load;
4622
			sds->this_has_capacity = sgs.group_has_capacity;
4623
			sds->this_idle_cpus = sgs.idle_cpus;
4624
		} else if (update_sd_pick_busiest(env, sds, sg, &sgs)) {
4625
			sds->max_load = sgs.avg_load;
4626
			sds->busiest = sg;
4627
			sds->busiest_nr_running = sgs.sum_nr_running;
4628
			sds->busiest_idle_cpus = sgs.idle_cpus;
4629
			sds->busiest_group_capacity = sgs.group_capacity;
4630
			sds->busiest_load_per_task = sgs.sum_weighted_load;
4631
			sds->busiest_has_capacity = sgs.group_has_capacity;
4632
			sds->busiest_group_weight = sgs.group_weight;
4633 4634 4635
			sds->group_imb = sgs.group_imb;
		}

4636
		sg = sg->next;
4637
	} while (sg != env->sd->groups);
4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656
}

/**
 * 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.
 *
4657 4658 4659
 * Returns 1 when packing is required and a task should be moved to
 * this CPU.  The amount of the imbalance is returned in *imbalance.
 *
4660
 * @env: The load balancing environment.
4661 4662
 * @sds: Statistics of the sched_domain which is to be packed
 */
4663
static int check_asym_packing(struct lb_env *env, struct sd_lb_stats *sds)
4664 4665 4666
{
	int busiest_cpu;

4667
	if (!(env->sd->flags & SD_ASYM_PACKING))
4668 4669 4670 4671 4672 4673
		return 0;

	if (!sds->busiest)
		return 0;

	busiest_cpu = group_first_cpu(sds->busiest);
4674
	if (env->dst_cpu > busiest_cpu)
4675 4676
		return 0;

4677 4678 4679
	env->imbalance = DIV_ROUND_CLOSEST(
		sds->max_load * sds->busiest->sgp->power, SCHED_POWER_SCALE);

4680
	return 1;
4681 4682 4683 4684 4685 4686
}

/**
 * fix_small_imbalance - Calculate the minor imbalance that exists
 *			amongst the groups of a sched_domain, during
 *			load balancing.
4687
 * @env: The load balancing environment.
4688 4689
 * @sds: Statistics of the sched_domain whose imbalance is to be calculated.
 */
4690 4691
static inline
void fix_small_imbalance(struct lb_env *env, struct sd_lb_stats *sds)
4692 4693 4694
{
	unsigned long tmp, pwr_now = 0, pwr_move = 0;
	unsigned int imbn = 2;
4695
	unsigned long scaled_busy_load_per_task;
4696 4697 4698 4699 4700 4701

	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;
4702
	} else {
4703
		sds->this_load_per_task =
4704 4705
			cpu_avg_load_per_task(env->dst_cpu);
	}
4706

4707
	scaled_busy_load_per_task = sds->busiest_load_per_task
4708
					 * SCHED_POWER_SCALE;
4709
	scaled_busy_load_per_task /= sds->busiest->sgp->power;
4710 4711 4712

	if (sds->max_load - sds->this_load + scaled_busy_load_per_task >=
			(scaled_busy_load_per_task * imbn)) {
4713
		env->imbalance = sds->busiest_load_per_task;
4714 4715 4716 4717 4718 4719 4720 4721 4722
		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.
	 */

4723
	pwr_now += sds->busiest->sgp->power *
4724
			min(sds->busiest_load_per_task, sds->max_load);
4725
	pwr_now += sds->this->sgp->power *
4726
			min(sds->this_load_per_task, sds->this_load);
4727
	pwr_now /= SCHED_POWER_SCALE;
4728 4729

	/* Amount of load we'd subtract */
4730
	tmp = (sds->busiest_load_per_task * SCHED_POWER_SCALE) /
4731
		sds->busiest->sgp->power;
4732
	if (sds->max_load > tmp)
4733
		pwr_move += sds->busiest->sgp->power *
4734 4735 4736
			min(sds->busiest_load_per_task, sds->max_load - tmp);

	/* Amount of load we'd add */
4737
	if (sds->max_load * sds->busiest->sgp->power <
4738
		sds->busiest_load_per_task * SCHED_POWER_SCALE)
4739 4740
		tmp = (sds->max_load * sds->busiest->sgp->power) /
			sds->this->sgp->power;
4741
	else
4742
		tmp = (sds->busiest_load_per_task * SCHED_POWER_SCALE) /
4743 4744
			sds->this->sgp->power;
	pwr_move += sds->this->sgp->power *
4745
			min(sds->this_load_per_task, sds->this_load + tmp);
4746
	pwr_move /= SCHED_POWER_SCALE;
4747 4748 4749

	/* Move if we gain throughput */
	if (pwr_move > pwr_now)
4750
		env->imbalance = sds->busiest_load_per_task;
4751 4752 4753 4754 4755
}

/**
 * calculate_imbalance - Calculate the amount of imbalance present within the
 *			 groups of a given sched_domain during load balance.
4756
 * @env: load balance environment
4757 4758
 * @sds: statistics of the sched_domain whose imbalance is to be calculated.
 */
4759
static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *sds)
4760
{
4761 4762 4763 4764 4765 4766 4767 4768
	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);
	}

4769 4770 4771 4772 4773 4774
	/*
	 * 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) {
4775 4776
		env->imbalance = 0;
		return fix_small_imbalance(env, sds);
4777 4778
	}

4779 4780 4781 4782 4783 4784 4785
	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);

4786
		load_above_capacity *= (SCHED_LOAD_SCALE * SCHED_POWER_SCALE);
4787

4788
		load_above_capacity /= sds->busiest->sgp->power;
4789 4790 4791 4792 4793 4794 4795 4796 4797 4798 4799 4800 4801
	}

	/*
	 * 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);
4802 4803

	/* How much load to actually move to equalise the imbalance */
4804
	env->imbalance = min(max_pull * sds->busiest->sgp->power,
4805
		(sds->avg_load - sds->this_load) * sds->this->sgp->power)
4806
			/ SCHED_POWER_SCALE;
4807 4808 4809

	/*
	 * if *imbalance is less than the average load per runnable task
L
Lucas De Marchi 已提交
4810
	 * there is no guarantee that any tasks will be moved so we'll have
4811 4812 4813
	 * a think about bumping its value to force at least one task to be
	 * moved
	 */
4814 4815
	if (env->imbalance < sds->busiest_load_per_task)
		return fix_small_imbalance(env, sds);
4816 4817

}
4818

4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830
/******* 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.
 *
4831
 * @env: The load balancing environment.
4832 4833 4834 4835 4836 4837 4838 4839 4840
 * @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 *
4841
find_busiest_group(struct lb_env *env, int *balance)
4842 4843 4844 4845 4846 4847 4848 4849 4850
{
	struct sd_lb_stats sds;

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

	/*
	 * Compute the various statistics relavent for load balancing at
	 * this level.
	 */
4851
	update_sd_lb_stats(env, balance, &sds);
4852

4853 4854 4855
	/*
	 * this_cpu is not the appropriate cpu to perform load balancing at
	 * this level.
4856
	 */
P
Peter Zijlstra 已提交
4857
	if (!(*balance))
4858 4859
		goto ret;

4860 4861
	if ((env->idle == CPU_IDLE || env->idle == CPU_NEWLY_IDLE) &&
	    check_asym_packing(env, &sds))
4862 4863
		return sds.busiest;

4864
	/* There is no busy sibling group to pull tasks from */
4865 4866 4867
	if (!sds.busiest || sds.busiest_nr_running == 0)
		goto out_balanced;

4868
	sds.avg_load = (SCHED_POWER_SCALE * sds.total_load) / sds.total_pwr;
4869

P
Peter Zijlstra 已提交
4870 4871 4872 4873 4874 4875 4876 4877
	/*
	 * 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;

4878
	/* SD_BALANCE_NEWIDLE trumps SMP nice when underutilized */
4879
	if (env->idle == CPU_NEWLY_IDLE && sds.this_has_capacity &&
4880 4881 4882
			!sds.busiest_has_capacity)
		goto force_balance;

4883 4884 4885 4886
	/*
	 * If the local group is more busy than the selected busiest group
	 * don't try and pull any tasks.
	 */
4887 4888 4889
	if (sds.this_load >= sds.max_load)
		goto out_balanced;

4890 4891 4892 4893
	/*
	 * Don't pull any tasks if this group is already above the domain
	 * average load.
	 */
4894 4895 4896
	if (sds.this_load >= sds.avg_load)
		goto out_balanced;

4897
	if (env->idle == CPU_IDLE) {
4898 4899 4900 4901 4902 4903
		/*
		 * 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.
		 */
4904
		if ((sds.this_idle_cpus <= sds.busiest_idle_cpus + 1) &&
4905 4906
		    sds.busiest_nr_running <= sds.busiest_group_weight)
			goto out_balanced;
4907 4908 4909 4910 4911
	} else {
		/*
		 * In the CPU_NEWLY_IDLE, CPU_NOT_IDLE cases, use
		 * imbalance_pct to be conservative.
		 */
4912
		if (100 * sds.max_load <= env->sd->imbalance_pct * sds.this_load)
4913
			goto out_balanced;
4914
	}
4915

4916
force_balance:
4917
	/* Looks like there is an imbalance. Compute it */
4918
	calculate_imbalance(env, &sds);
4919 4920 4921 4922
	return sds.busiest;

out_balanced:
ret:
4923
	env->imbalance = 0;
4924 4925 4926 4927 4928 4929
	return NULL;
}

/*
 * find_busiest_queue - find the busiest runqueue among the cpus in group.
 */
4930
static struct rq *find_busiest_queue(struct lb_env *env,
4931
				     struct sched_group *group)
4932 4933 4934 4935 4936 4937 4938
{
	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);
4939 4940
		unsigned long capacity = DIV_ROUND_CLOSEST(power,
							   SCHED_POWER_SCALE);
4941 4942
		unsigned long wl;

4943
		if (!capacity)
4944
			capacity = fix_small_capacity(env->sd, group);
4945

4946
		if (!cpumask_test_cpu(i, env->cpus))
4947 4948 4949
			continue;

		rq = cpu_rq(i);
4950
		wl = weighted_cpuload(i);
4951

4952 4953 4954 4955
		/*
		 * When comparing with imbalance, use weighted_cpuload()
		 * which is not scaled with the cpu power.
		 */
4956
		if (capacity && rq->nr_running == 1 && wl > env->imbalance)
4957 4958
			continue;

4959 4960 4961 4962 4963 4964
		/*
		 * 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.
		 */
4965
		wl = (wl * SCHED_POWER_SCALE) / power;
4966

4967 4968 4969 4970 4971 4972 4973 4974 4975 4976 4977 4978 4979 4980 4981 4982
		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. */
4983
DEFINE_PER_CPU(cpumask_var_t, load_balance_mask);
4984

4985
static int need_active_balance(struct lb_env *env)
4986
{
4987 4988 4989
	struct sched_domain *sd = env->sd;

	if (env->idle == CPU_NEWLY_IDLE) {
4990 4991 4992 4993 4994 4995

		/*
		 * ASYM_PACKING needs to force migrate tasks from busy but
		 * higher numbered CPUs in order to pack all tasks in the
		 * lowest numbered CPUs.
		 */
4996
		if ((sd->flags & SD_ASYM_PACKING) && env->src_cpu > env->dst_cpu)
4997
			return 1;
4998 4999 5000 5001 5002
	}

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

5003 5004
static int active_load_balance_cpu_stop(void *data);

5005 5006 5007 5008 5009 5010 5011 5012
/*
 * 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)
{
5013
	int ld_moved, cur_ld_moved, active_balance = 0;
5014 5015 5016
	struct sched_group *group;
	struct rq *busiest;
	unsigned long flags;
5017
	struct cpumask *cpus = __get_cpu_var(load_balance_mask);
5018

5019 5020
	struct lb_env env = {
		.sd		= sd,
5021 5022
		.dst_cpu	= this_cpu,
		.dst_rq		= this_rq,
5023
		.dst_grpmask    = sched_group_cpus(sd->groups),
5024
		.idle		= idle,
5025
		.loop_break	= sched_nr_migrate_break,
5026
		.cpus		= cpus,
5027 5028
	};

5029 5030 5031 5032
	/*
	 * For NEWLY_IDLE load_balancing, we don't need to consider
	 * other cpus in our group
	 */
5033
	if (idle == CPU_NEWLY_IDLE)
5034 5035
		env.dst_grpmask = NULL;

5036 5037 5038 5039 5040
	cpumask_copy(cpus, cpu_active_mask);

	schedstat_inc(sd, lb_count[idle]);

redo:
5041
	group = find_busiest_group(&env, balance);
5042 5043 5044 5045 5046 5047 5048 5049 5050

	if (*balance == 0)
		goto out_balanced;

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

5051
	busiest = find_busiest_queue(&env, group);
5052 5053 5054 5055 5056
	if (!busiest) {
		schedstat_inc(sd, lb_nobusyq[idle]);
		goto out_balanced;
	}

5057
	BUG_ON(busiest == env.dst_rq);
5058

5059
	schedstat_add(sd, lb_imbalance[idle], env.imbalance);
5060 5061 5062 5063 5064 5065 5066 5067 5068

	ld_moved = 0;
	if (busiest->nr_running > 1) {
		/*
		 * Attempt to move tasks. If find_busiest_group has found
		 * an imbalance but busiest->nr_running <= 1, the group is
		 * still unbalanced. ld_moved simply stays zero, so it is
		 * correctly treated as an imbalance.
		 */
5069
		env.flags |= LBF_ALL_PINNED;
5070 5071 5072
		env.src_cpu   = busiest->cpu;
		env.src_rq    = busiest;
		env.loop_max  = min(sysctl_sched_nr_migrate, busiest->nr_running);
5073

5074
		update_h_load(env.src_cpu);
5075
more_balance:
5076
		local_irq_save(flags);
5077
		double_rq_lock(env.dst_rq, busiest);
5078 5079 5080 5081 5082 5083 5084

		/*
		 * 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;
5085
		double_rq_unlock(env.dst_rq, busiest);
5086 5087 5088 5089 5090
		local_irq_restore(flags);

		/*
		 * some other cpu did the load balance for us.
		 */
5091 5092 5093
		if (cur_ld_moved && env.dst_cpu != smp_processor_id())
			resched_cpu(env.dst_cpu);

5094 5095 5096 5097 5098
		if (env.flags & LBF_NEED_BREAK) {
			env.flags &= ~LBF_NEED_BREAK;
			goto more_balance;
		}

5099 5100 5101 5102 5103 5104 5105 5106 5107 5108 5109 5110 5111 5112 5113 5114 5115 5116 5117
		/*
		 * 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.
		 */
5118
		if ((env.flags & LBF_SOME_PINNED) && env.imbalance > 0) {
5119

5120
			env.dst_rq	 = cpu_rq(env.new_dst_cpu);
5121 5122 5123 5124
			env.dst_cpu	 = env.new_dst_cpu;
			env.flags	&= ~LBF_SOME_PINNED;
			env.loop	 = 0;
			env.loop_break	 = sched_nr_migrate_break;
5125 5126 5127 5128

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

5129 5130 5131 5132 5133 5134
			/*
			 * Go back to "more_balance" rather than "redo" since we
			 * need to continue with same src_cpu.
			 */
			goto more_balance;
		}
5135 5136

		/* All tasks on this runqueue were pinned by CPU affinity */
5137
		if (unlikely(env.flags & LBF_ALL_PINNED)) {
5138
			cpumask_clear_cpu(cpu_of(busiest), cpus);
5139 5140 5141
			if (!cpumask_empty(cpus)) {
				env.loop = 0;
				env.loop_break = sched_nr_migrate_break;
5142
				goto redo;
5143
			}
5144 5145 5146 5147 5148 5149
			goto out_balanced;
		}
	}

	if (!ld_moved) {
		schedstat_inc(sd, lb_failed[idle]);
5150 5151 5152 5153 5154 5155 5156 5157
		/*
		 * 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++;
5158

5159
		if (need_active_balance(&env)) {
5160 5161
			raw_spin_lock_irqsave(&busiest->lock, flags);

5162 5163 5164
			/* don't kick the active_load_balance_cpu_stop,
			 * if the curr task on busiest cpu can't be
			 * moved to this_cpu
5165 5166
			 */
			if (!cpumask_test_cpu(this_cpu,
5167
					tsk_cpus_allowed(busiest->curr))) {
5168 5169
				raw_spin_unlock_irqrestore(&busiest->lock,
							    flags);
5170
				env.flags |= LBF_ALL_PINNED;
5171 5172 5173
				goto out_one_pinned;
			}

5174 5175 5176 5177 5178
			/*
			 * ->active_balance synchronizes accesses to
			 * ->active_balance_work.  Once set, it's cleared
			 * only after active load balance is finished.
			 */
5179 5180 5181 5182 5183 5184
			if (!busiest->active_balance) {
				busiest->active_balance = 1;
				busiest->push_cpu = this_cpu;
				active_balance = 1;
			}
			raw_spin_unlock_irqrestore(&busiest->lock, flags);
5185

5186
			if (active_balance) {
5187 5188 5189
				stop_one_cpu_nowait(cpu_of(busiest),
					active_load_balance_cpu_stop, busiest,
					&busiest->active_balance_work);
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

			/*
			 * 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 */
5224
	if (((env.flags & LBF_ALL_PINNED) &&
5225
			sd->balance_interval < MAX_PINNED_INTERVAL) ||
5226 5227 5228
			(sd->balance_interval < sd->max_interval))
		sd->balance_interval *= 2;

5229
	ld_moved = 0;
5230 5231 5232 5233 5234 5235 5236 5237
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.
 */
5238
void idle_balance(int this_cpu, struct rq *this_rq)
5239 5240 5241 5242 5243 5244 5245 5246 5247 5248
{
	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;

5249 5250 5251 5252 5253
	/*
	 * Drop the rq->lock, but keep IRQ/preempt disabled.
	 */
	raw_spin_unlock(&this_rq->lock);

5254
	update_blocked_averages(this_cpu);
5255
	rcu_read_lock();
5256 5257
	for_each_domain(this_cpu, sd) {
		unsigned long interval;
5258
		int balance = 1;
5259 5260 5261 5262

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

5263
		if (sd->flags & SD_BALANCE_NEWIDLE) {
5264
			/* If we've pulled tasks over stop searching: */
5265 5266 5267
			pulled_task = load_balance(this_cpu, this_rq,
						   sd, CPU_NEWLY_IDLE, &balance);
		}
5268 5269 5270 5271

		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
5272 5273
		if (pulled_task) {
			this_rq->idle_stamp = 0;
5274
			break;
N
Nikhil Rao 已提交
5275
		}
5276
	}
5277
	rcu_read_unlock();
5278 5279 5280

	raw_spin_lock(&this_rq->lock);

5281 5282 5283 5284 5285 5286 5287 5288 5289 5290
	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;
	}
}

/*
5291 5292 5293 5294
 * 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.
5295
 */
5296
static int active_load_balance_cpu_stop(void *data)
5297
{
5298 5299
	struct rq *busiest_rq = data;
	int busiest_cpu = cpu_of(busiest_rq);
5300
	int target_cpu = busiest_rq->push_cpu;
5301
	struct rq *target_rq = cpu_rq(target_cpu);
5302
	struct sched_domain *sd;
5303 5304 5305 5306 5307 5308 5309

	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;
5310 5311 5312

	/* Is there any task to move? */
	if (busiest_rq->nr_running <= 1)
5313
		goto out_unlock;
5314 5315 5316 5317 5318 5319 5320 5321 5322 5323 5324 5325

	/*
	 * 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. */
5326
	rcu_read_lock();
5327 5328 5329 5330 5331 5332 5333
	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)) {
5334 5335
		struct lb_env env = {
			.sd		= sd,
5336 5337 5338 5339
			.dst_cpu	= target_cpu,
			.dst_rq		= target_rq,
			.src_cpu	= busiest_rq->cpu,
			.src_rq		= busiest_rq,
5340 5341 5342
			.idle		= CPU_IDLE,
		};

5343 5344
		schedstat_inc(sd, alb_count);

5345
		if (move_one_task(&env))
5346 5347 5348 5349
			schedstat_inc(sd, alb_pushed);
		else
			schedstat_inc(sd, alb_failed);
	}
5350
	rcu_read_unlock();
5351
	double_unlock_balance(busiest_rq, target_rq);
5352 5353 5354 5355
out_unlock:
	busiest_rq->active_balance = 0;
	raw_spin_unlock_irq(&busiest_rq->lock);
	return 0;
5356 5357 5358
}

#ifdef CONFIG_NO_HZ
5359 5360 5361 5362 5363 5364
/*
 * 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.
 */
5365
static struct {
5366
	cpumask_var_t idle_cpus_mask;
5367
	atomic_t nr_cpus;
5368 5369
	unsigned long next_balance;     /* in jiffy units */
} nohz ____cacheline_aligned;
5370

5371
static inline int find_new_ilb(int call_cpu)
5372
{
5373
	int ilb = cpumask_first(nohz.idle_cpus_mask);
5374

5375 5376 5377 5378
	if (ilb < nr_cpu_ids && idle_cpu(ilb))
		return ilb;

	return nr_cpu_ids;
5379 5380
}

5381 5382 5383 5384 5385 5386 5387 5388 5389 5390 5391
/*
 * 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++;

5392
	ilb_cpu = find_new_ilb(cpu);
5393

5394 5395
	if (ilb_cpu >= nr_cpu_ids)
		return;
5396

5397
	if (test_and_set_bit(NOHZ_BALANCE_KICK, nohz_flags(ilb_cpu)))
5398 5399 5400 5401 5402 5403 5404 5405
		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);
5406 5407 5408
	return;
}

5409
static inline void nohz_balance_exit_idle(int cpu)
5410 5411 5412 5413 5414 5415 5416 5417
{
	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));
	}
}

5418 5419 5420 5421 5422 5423 5424 5425 5426 5427 5428 5429 5430 5431 5432 5433 5434 5435 5436 5437 5438 5439 5440 5441 5442 5443 5444 5445 5446 5447
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();
}

5448
/*
5449
 * This routine will record that the cpu is going idle with tick stopped.
5450
 * This info will be used in performing idle load balancing in the future.
5451
 */
5452
void nohz_balance_enter_idle(int cpu)
5453
{
5454 5455 5456 5457 5458 5459
	/*
	 * If this cpu is going down, then nothing needs to be done.
	 */
	if (!cpu_active(cpu))
		return;

5460 5461
	if (test_bit(NOHZ_TICK_STOPPED, nohz_flags(cpu)))
		return;
5462

5463 5464 5465
	cpumask_set_cpu(cpu, nohz.idle_cpus_mask);
	atomic_inc(&nohz.nr_cpus);
	set_bit(NOHZ_TICK_STOPPED, nohz_flags(cpu));
5466
}
5467 5468 5469 5470 5471 5472

static int __cpuinit sched_ilb_notifier(struct notifier_block *nfb,
					unsigned long action, void *hcpu)
{
	switch (action & ~CPU_TASKS_FROZEN) {
	case CPU_DYING:
5473
		nohz_balance_exit_idle(smp_processor_id());
5474 5475 5476 5477 5478
		return NOTIFY_OK;
	default:
		return NOTIFY_DONE;
	}
}
5479 5480 5481 5482
#endif

static DEFINE_SPINLOCK(balancing);

5483 5484 5485 5486
/*
 * 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.
 */
5487
void update_max_interval(void)
5488 5489 5490 5491
{
	max_load_balance_interval = HZ*num_online_cpus()/10;
}

5492 5493 5494 5495
/*
 * It checks each scheduling domain to see if it is due to be balanced,
 * and initiates a balancing operation if so.
 *
5496
 * Balancing parameters are set up in init_sched_domains.
5497 5498 5499 5500 5501 5502
 */
static void rebalance_domains(int cpu, enum cpu_idle_type idle)
{
	int balance = 1;
	struct rq *rq = cpu_rq(cpu);
	unsigned long interval;
5503
	struct sched_domain *sd;
5504 5505 5506 5507 5508
	/* Earliest time when we have to do rebalance again */
	unsigned long next_balance = jiffies + 60*HZ;
	int update_next_balance = 0;
	int need_serialize;

5509
	update_blocked_averages(cpu);
P
Peter Zijlstra 已提交
5510

5511
	rcu_read_lock();
5512 5513 5514 5515 5516 5517 5518 5519 5520 5521
	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);
5522
		interval = clamp(interval, 1UL, max_load_balance_interval);
5523 5524 5525 5526 5527 5528 5529 5530 5531 5532 5533

		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)) {
				/*
5534 5535 5536
				 * The LBF_SOME_PINNED logic could have changed
				 * env->dst_cpu, so we can't know our idle
				 * state even if we migrated tasks. Update it.
5537
				 */
5538
				idle = idle_cpu(cpu) ? CPU_IDLE : CPU_NOT_IDLE;
5539 5540 5541 5542 5543 5544 5545 5546 5547 5548 5549 5550 5551 5552 5553 5554 5555 5556 5557
			}
			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;
	}
5558
	rcu_read_unlock();
5559 5560 5561 5562 5563 5564 5565 5566 5567 5568

	/*
	 * 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;
}

5569
#ifdef CONFIG_NO_HZ
5570
/*
5571
 * In CONFIG_NO_HZ case, the idle balance kickee will do the
5572 5573
 * rebalancing for all the cpus for whom scheduler ticks are stopped.
 */
5574 5575 5576 5577 5578 5579
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;

5580 5581 5582
	if (idle != CPU_IDLE ||
	    !test_bit(NOHZ_BALANCE_KICK, nohz_flags(this_cpu)))
		goto end;
5583 5584

	for_each_cpu(balance_cpu, nohz.idle_cpus_mask) {
5585
		if (balance_cpu == this_cpu || !idle_cpu(balance_cpu))
5586 5587 5588 5589 5590 5591 5592
			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.
		 */
5593
		if (need_resched())
5594 5595
			break;

V
Vincent Guittot 已提交
5596 5597 5598 5599 5600 5601
		rq = cpu_rq(balance_cpu);

		raw_spin_lock_irq(&rq->lock);
		update_rq_clock(rq);
		update_idle_cpu_load(rq);
		raw_spin_unlock_irq(&rq->lock);
5602 5603 5604 5605 5606 5607 5608

		rebalance_domains(balance_cpu, CPU_IDLE);

		if (time_after(this_rq->next_balance, rq->next_balance))
			this_rq->next_balance = rq->next_balance;
	}
	nohz.next_balance = this_rq->next_balance;
5609 5610
end:
	clear_bit(NOHZ_BALANCE_KICK, nohz_flags(this_cpu));
5611 5612 5613
}

/*
5614 5615 5616 5617 5618 5619 5620
 * 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.
5621 5622 5623 5624
 */
static inline int nohz_kick_needed(struct rq *rq, int cpu)
{
	unsigned long now = jiffies;
5625
	struct sched_domain *sd;
5626

5627
	if (unlikely(idle_cpu(cpu)))
5628 5629
		return 0;

5630 5631 5632 5633
       /*
	* 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.
	*/
5634
	set_cpu_sd_state_busy();
5635
	nohz_balance_exit_idle(cpu);
5636 5637 5638 5639 5640 5641 5642

	/*
	 * None are in tickless mode and hence no need for NOHZ idle load
	 * balancing.
	 */
	if (likely(!atomic_read(&nohz.nr_cpus)))
		return 0;
5643 5644

	if (time_before(now, nohz.next_balance))
5645 5646
		return 0;

5647 5648
	if (rq->nr_running >= 2)
		goto need_kick;
5649

5650
	rcu_read_lock();
5651 5652 5653 5654
	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);
5655

5656
		if (sd->flags & SD_SHARE_PKG_RESOURCES && nr_busy > 1)
5657
			goto need_kick_unlock;
5658 5659 5660 5661

		if (sd->flags & SD_ASYM_PACKING && nr_busy != sg->group_weight
		    && (cpumask_first_and(nohz.idle_cpus_mask,
					  sched_domain_span(sd)) < cpu))
5662
			goto need_kick_unlock;
5663 5664 5665

		if (!(sd->flags & (SD_SHARE_PKG_RESOURCES | SD_ASYM_PACKING)))
			break;
5666
	}
5667
	rcu_read_unlock();
5668
	return 0;
5669 5670 5671

need_kick_unlock:
	rcu_read_unlock();
5672 5673
need_kick:
	return 1;
5674 5675 5676 5677 5678 5679 5680 5681 5682
}
#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).
 */
5683 5684 5685 5686
static void run_rebalance_domains(struct softirq_action *h)
{
	int this_cpu = smp_processor_id();
	struct rq *this_rq = cpu_rq(this_cpu);
5687
	enum cpu_idle_type idle = this_rq->idle_balance ?
5688 5689 5690 5691 5692
						CPU_IDLE : CPU_NOT_IDLE;

	rebalance_domains(this_cpu, idle);

	/*
5693
	 * If this cpu has a pending nohz_balance_kick, then do the
5694 5695 5696
	 * balancing on behalf of the other idle cpus whose ticks are
	 * stopped.
	 */
5697
	nohz_idle_balance(this_cpu, idle);
5698 5699 5700 5701
}

static inline int on_null_domain(int cpu)
{
5702
	return !rcu_dereference_sched(cpu_rq(cpu)->sd);
5703 5704 5705 5706 5707
}

/*
 * Trigger the SCHED_SOFTIRQ if it is time to do periodic load balancing.
 */
5708
void trigger_load_balance(struct rq *rq, int cpu)
5709 5710 5711 5712 5713
{
	/* 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);
5714
#ifdef CONFIG_NO_HZ
5715
	if (nohz_kick_needed(rq, cpu) && likely(!on_null_domain(cpu)))
5716 5717
		nohz_balancer_kick(cpu);
#endif
5718 5719
}

5720 5721 5722 5723 5724 5725 5726 5727
static void rq_online_fair(struct rq *rq)
{
	update_sysctl();
}

static void rq_offline_fair(struct rq *rq)
{
	update_sysctl();
5728 5729 5730

	/* Ensure any throttled groups are reachable by pick_next_task */
	unthrottle_offline_cfs_rqs(rq);
5731 5732
}

5733
#endif /* CONFIG_SMP */
5734

5735 5736 5737
/*
 * scheduler tick hitting a task of our scheduling class:
 */
P
Peter Zijlstra 已提交
5738
static void task_tick_fair(struct rq *rq, struct task_struct *curr, int queued)
5739 5740 5741 5742 5743 5744
{
	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
5745
		entity_tick(cfs_rq, se, queued);
5746
	}
5747

5748 5749
	if (sched_feat_numa(NUMA))
		task_tick_numa(rq, curr);
5750

5751
	update_rq_runnable_avg(rq, 1);
5752 5753 5754
}

/*
P
Peter Zijlstra 已提交
5755 5756 5757
 * called on fork with the child task as argument from the parent's context
 *  - child not yet on the tasklist
 *  - preemption disabled
5758
 */
P
Peter Zijlstra 已提交
5759
static void task_fork_fair(struct task_struct *p)
5760
{
5761 5762
	struct cfs_rq *cfs_rq;
	struct sched_entity *se = &p->se, *curr;
5763
	int this_cpu = smp_processor_id();
P
Peter Zijlstra 已提交
5764 5765 5766
	struct rq *rq = this_rq();
	unsigned long flags;

5767
	raw_spin_lock_irqsave(&rq->lock, flags);
5768

5769 5770
	update_rq_clock(rq);

5771 5772 5773
	cfs_rq = task_cfs_rq(current);
	curr = cfs_rq->curr;

5774 5775
	if (unlikely(task_cpu(p) != this_cpu)) {
		rcu_read_lock();
P
Peter Zijlstra 已提交
5776
		__set_task_cpu(p, this_cpu);
5777 5778
		rcu_read_unlock();
	}
5779

5780
	update_curr(cfs_rq);
P
Peter Zijlstra 已提交
5781

5782 5783
	if (curr)
		se->vruntime = curr->vruntime;
5784
	place_entity(cfs_rq, se, 1);
5785

P
Peter Zijlstra 已提交
5786
	if (sysctl_sched_child_runs_first && curr && entity_before(curr, se)) {
D
Dmitry Adamushko 已提交
5787
		/*
5788 5789 5790
		 * Upon rescheduling, sched_class::put_prev_task() will place
		 * 'current' within the tree based on its new key value.
		 */
5791
		swap(curr->vruntime, se->vruntime);
5792
		resched_task(rq->curr);
5793
	}
5794

5795 5796
	se->vruntime -= cfs_rq->min_vruntime;

5797
	raw_spin_unlock_irqrestore(&rq->lock, flags);
5798 5799
}

5800 5801 5802 5803
/*
 * Priority of the task has changed. Check to see if we preempt
 * the current task.
 */
P
Peter Zijlstra 已提交
5804 5805
static void
prio_changed_fair(struct rq *rq, struct task_struct *p, int oldprio)
5806
{
P
Peter Zijlstra 已提交
5807 5808 5809
	if (!p->se.on_rq)
		return;

5810 5811 5812 5813 5814
	/*
	 * 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
5815
	if (rq->curr == p) {
5816 5817 5818
		if (p->prio > oldprio)
			resched_task(rq->curr);
	} else
5819
		check_preempt_curr(rq, p, 0);
5820 5821
}

P
Peter Zijlstra 已提交
5822 5823 5824 5825 5826 5827 5828 5829 5830 5831 5832 5833 5834 5835 5836 5837 5838 5839 5840 5841 5842 5843
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;
	}
5844 5845 5846 5847 5848 5849 5850 5851 5852 5853 5854 5855 5856 5857

#if defined(CONFIG_FAIR_GROUP_SCHED) && defined(CONFIG_SMP)
	/*
	* Remove our load from contribution when we leave sched_fair
	* and ensure we don't carry in an old decay_count if we
	* switch back.
	*/
	if (p->se.avg.decay_count) {
		struct cfs_rq *cfs_rq = cfs_rq_of(&p->se);
		__synchronize_entity_decay(&p->se);
		subtract_blocked_load_contrib(cfs_rq,
				p->se.avg.load_avg_contrib);
	}
#endif
P
Peter Zijlstra 已提交
5858 5859
}

5860 5861 5862
/*
 * We switched to the sched_fair class.
 */
P
Peter Zijlstra 已提交
5863
static void switched_to_fair(struct rq *rq, struct task_struct *p)
5864
{
P
Peter Zijlstra 已提交
5865 5866 5867
	if (!p->se.on_rq)
		return;

5868 5869 5870 5871 5872
	/*
	 * 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
5873
	if (rq->curr == p)
5874 5875
		resched_task(rq->curr);
	else
5876
		check_preempt_curr(rq, p, 0);
5877 5878
}

5879 5880 5881 5882 5883 5884 5885 5886 5887
/* 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;

5888 5889 5890 5891 5892 5893 5894
	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);
	}
5895 5896
}

5897 5898 5899 5900 5901 5902 5903
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
5904 5905
#if defined(CONFIG_FAIR_GROUP_SCHED) && defined(CONFIG_SMP)
	atomic64_set(&cfs_rq->decay_counter, 1);
5906
	atomic64_set(&cfs_rq->removed_load, 0);
5907
#endif
5908 5909
}

P
Peter Zijlstra 已提交
5910
#ifdef CONFIG_FAIR_GROUP_SCHED
5911
static void task_move_group_fair(struct task_struct *p, int on_rq)
P
Peter Zijlstra 已提交
5912
{
5913
	struct cfs_rq *cfs_rq;
5914 5915 5916 5917 5918 5919 5920 5921 5922 5923 5924 5925 5926
	/*
	 * 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.
	 */
5927 5928 5929 5930 5931 5932
	/*
	 * 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().
5933 5934
	 * - Moving a task which has been woken up by try_to_wake_up() and
	 *   waiting for actually being woken up by sched_ttwu_pending().
5935 5936 5937 5938
	 *
	 * To prevent boost or penalty in the new cfs_rq caused by delta
	 * min_vruntime between the two cfs_rqs, we skip vruntime adjustment.
	 */
5939
	if (!on_rq && (!p->se.sum_exec_runtime || p->state == TASK_WAKING))
5940 5941
		on_rq = 1;

5942 5943 5944
	if (!on_rq)
		p->se.vruntime -= cfs_rq_of(&p->se)->min_vruntime;
	set_task_rq(p, task_cpu(p));
5945 5946 5947 5948 5949 5950 5951 5952 5953 5954 5955 5956 5957
	if (!on_rq) {
		cfs_rq = cfs_rq_of(&p->se);
		p->se.vruntime += cfs_rq->min_vruntime;
#ifdef CONFIG_SMP
		/*
		 * migrate_task_rq_fair() will have removed our previous
		 * contribution, but we must synchronize for ongoing future
		 * decay.
		 */
		p->se.avg.decay_count = atomic64_read(&cfs_rq->decay_counter);
		cfs_rq->blocked_load_avg += p->se.avg.load_avg_contrib;
#endif
	}
P
Peter Zijlstra 已提交
5958
}
5959 5960 5961 5962 5963 5964 5965 5966 5967 5968 5969 5970 5971 5972 5973 5974 5975 5976 5977 5978 5979 5980 5981 5982 5983 5984 5985 5986 5987 5988 5989 5990 5991 5992 5993 5994 5995 5996 5997 5998 5999 6000 6001 6002 6003 6004 6005 6006 6007 6008 6009 6010 6011 6012 6013 6014 6015 6016 6017 6018 6019 6020 6021 6022 6023 6024 6025 6026 6027 6028 6029 6030 6031 6032 6033 6034 6035 6036 6037 6038 6039 6040 6041 6042 6043 6044 6045 6046 6047 6048 6049 6050 6051 6052 6053 6054 6055 6056 6057 6058 6059 6060 6061 6062 6063 6064 6065 6066 6067 6068 6069 6070 6071 6072 6073 6074 6075 6076 6077 6078 6079 6080 6081 6082 6083 6084 6085 6086 6087

void free_fair_sched_group(struct task_group *tg)
{
	int i;

	destroy_cfs_bandwidth(tg_cfs_bandwidth(tg));

	for_each_possible_cpu(i) {
		if (tg->cfs_rq)
			kfree(tg->cfs_rq[i]);
		if (tg->se)
			kfree(tg->se[i]);
	}

	kfree(tg->cfs_rq);
	kfree(tg->se);
}

int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent)
{
	struct cfs_rq *cfs_rq;
	struct sched_entity *se;
	int i;

	tg->cfs_rq = kzalloc(sizeof(cfs_rq) * nr_cpu_ids, GFP_KERNEL);
	if (!tg->cfs_rq)
		goto err;
	tg->se = kzalloc(sizeof(se) * nr_cpu_ids, GFP_KERNEL);
	if (!tg->se)
		goto err;

	tg->shares = NICE_0_LOAD;

	init_cfs_bandwidth(tg_cfs_bandwidth(tg));

	for_each_possible_cpu(i) {
		cfs_rq = kzalloc_node(sizeof(struct cfs_rq),
				      GFP_KERNEL, cpu_to_node(i));
		if (!cfs_rq)
			goto err;

		se = kzalloc_node(sizeof(struct sched_entity),
				  GFP_KERNEL, cpu_to_node(i));
		if (!se)
			goto err_free_rq;

		init_cfs_rq(cfs_rq);
		init_tg_cfs_entry(tg, cfs_rq, se, i, parent->se[i]);
	}

	return 1;

err_free_rq:
	kfree(cfs_rq);
err:
	return 0;
}

void unregister_fair_sched_group(struct task_group *tg, int cpu)
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

	/*
	* Only empty task groups can be destroyed; so we can speculatively
	* check on_list without danger of it being re-added.
	*/
	if (!tg->cfs_rq[cpu]->on_list)
		return;

	raw_spin_lock_irqsave(&rq->lock, flags);
	list_del_leaf_cfs_rq(tg->cfs_rq[cpu]);
	raw_spin_unlock_irqrestore(&rq->lock, flags);
}

void init_tg_cfs_entry(struct task_group *tg, struct cfs_rq *cfs_rq,
			struct sched_entity *se, int cpu,
			struct sched_entity *parent)
{
	struct rq *rq = cpu_rq(cpu);

	cfs_rq->tg = tg;
	cfs_rq->rq = rq;
	init_cfs_rq_runtime(cfs_rq);

	tg->cfs_rq[cpu] = cfs_rq;
	tg->se[cpu] = se;

	/* se could be NULL for root_task_group */
	if (!se)
		return;

	if (!parent)
		se->cfs_rq = &rq->cfs;
	else
		se->cfs_rq = parent->my_q;

	se->my_q = cfs_rq;
	update_load_set(&se->load, 0);
	se->parent = parent;
}

static DEFINE_MUTEX(shares_mutex);

int sched_group_set_shares(struct task_group *tg, unsigned long shares)
{
	int i;
	unsigned long flags;

	/*
	 * We can't change the weight of the root cgroup.
	 */
	if (!tg->se[0])
		return -EINVAL;

	shares = clamp(shares, scale_load(MIN_SHARES), scale_load(MAX_SHARES));

	mutex_lock(&shares_mutex);
	if (tg->shares == shares)
		goto done;

	tg->shares = shares;
	for_each_possible_cpu(i) {
		struct rq *rq = cpu_rq(i);
		struct sched_entity *se;

		se = tg->se[i];
		/* Propagate contribution to hierarchy */
		raw_spin_lock_irqsave(&rq->lock, flags);
6088
		for_each_sched_entity(se)
6089 6090 6091 6092 6093 6094 6095 6096 6097 6098 6099 6100 6101 6102 6103 6104 6105 6106 6107 6108 6109
			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
6110

6111
static unsigned int get_rr_interval_fair(struct rq *rq, struct task_struct *task)
6112 6113 6114 6115 6116 6117 6118 6119 6120
{
	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)
6121
		rr_interval = NS_TO_JIFFIES(sched_slice(cfs_rq_of(se), se));
6122 6123 6124 6125

	return rr_interval;
}

6126 6127 6128
/*
 * All the scheduling class methods:
 */
6129
const struct sched_class fair_sched_class = {
6130
	.next			= &idle_sched_class,
6131 6132 6133
	.enqueue_task		= enqueue_task_fair,
	.dequeue_task		= dequeue_task_fair,
	.yield_task		= yield_task_fair,
6134
	.yield_to_task		= yield_to_task_fair,
6135

I
Ingo Molnar 已提交
6136
	.check_preempt_curr	= check_preempt_wakeup,
6137 6138 6139 6140

	.pick_next_task		= pick_next_task_fair,
	.put_prev_task		= put_prev_task_fair,

6141
#ifdef CONFIG_SMP
L
Li Zefan 已提交
6142
	.select_task_rq		= select_task_rq_fair,
6143
#ifdef CONFIG_FAIR_GROUP_SCHED
6144
	.migrate_task_rq	= migrate_task_rq_fair,
6145
#endif
6146 6147
	.rq_online		= rq_online_fair,
	.rq_offline		= rq_offline_fair,
6148 6149

	.task_waking		= task_waking_fair,
6150
#endif
6151

6152
	.set_curr_task          = set_curr_task_fair,
6153
	.task_tick		= task_tick_fair,
P
Peter Zijlstra 已提交
6154
	.task_fork		= task_fork_fair,
6155 6156

	.prio_changed		= prio_changed_fair,
P
Peter Zijlstra 已提交
6157
	.switched_from		= switched_from_fair,
6158
	.switched_to		= switched_to_fair,
P
Peter Zijlstra 已提交
6159

6160 6161
	.get_rr_interval	= get_rr_interval_fair,

P
Peter Zijlstra 已提交
6162
#ifdef CONFIG_FAIR_GROUP_SCHED
6163
	.task_move_group	= task_move_group_fair,
P
Peter Zijlstra 已提交
6164
#endif
6165 6166 6167
};

#ifdef CONFIG_SCHED_DEBUG
6168
void print_cfs_stats(struct seq_file *m, int cpu)
6169 6170 6171
{
	struct cfs_rq *cfs_rq;

6172
	rcu_read_lock();
6173
	for_each_leaf_cfs_rq(cpu_rq(cpu), cfs_rq)
6174
		print_cfs_rq(m, cpu, cfs_rq);
6175
	rcu_read_unlock();
6176 6177
}
#endif
6178 6179 6180 6181 6182 6183 6184

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

#ifdef CONFIG_NO_HZ
6185
	nohz.next_balance = jiffies;
6186
	zalloc_cpumask_var(&nohz.idle_cpus_mask, GFP_NOWAIT);
6187
	cpu_notifier(sched_ilb_notifier, 0);
6188 6189 6190 6191
#endif
#endif /* SMP */

}