fair.c 138.3 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;
}

265 266 267
static inline void list_add_leaf_cfs_rq(struct cfs_rq *cfs_rq)
{
	if (!cfs_rq->on_list) {
268 269 270 271 272 273 274 275 276 277 278 279
		/*
		 * 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,
280
				&rq_of(cfs_rq)->leaf_cfs_rq_list);
281
		}
282 283 284 285 286 287 288 289 290 291 292 293 294

		cfs_rq->on_list = 1;
	}
}

static inline void list_del_leaf_cfs_rq(struct cfs_rq *cfs_rq)
{
	if (cfs_rq->on_list) {
		list_del_rcu(&cfs_rq->leaf_cfs_rq_list);
		cfs_rq->on_list = 0;
	}
}

P
Peter Zijlstra 已提交
295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313
/* 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;
}

314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356
/* 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);
	}
}

357 358 359 360 361 362
#else	/* !CONFIG_FAIR_GROUP_SCHED */

static inline struct task_struct *task_of(struct sched_entity *se)
{
	return container_of(se, struct task_struct, se);
}
363

364 365 366
static inline struct rq *rq_of(struct cfs_rq *cfs_rq)
{
	return container_of(cfs_rq, struct rq, cfs);
367 368 369 370
}

#define entity_is_task(se)	1

P
Peter Zijlstra 已提交
371 372
#define for_each_sched_entity(se) \
		for (; se; se = NULL)
373

P
Peter Zijlstra 已提交
374
static inline struct cfs_rq *task_cfs_rq(struct task_struct *p)
375
{
P
Peter Zijlstra 已提交
376
	return &task_rq(p)->cfs;
377 378
}

P
Peter Zijlstra 已提交
379 380 381 382 383 384 385 386 387 388 389 390 391 392
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;
}

393 394 395 396 397 398 399 400
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
401 402 403 404 405 406 407 408 409 410 411 412 413 414
#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;
}

415 416 417 418 419
static inline void
find_matching_se(struct sched_entity **se, struct sched_entity **pse)
{
}

P
Peter Zijlstra 已提交
420 421
#endif	/* CONFIG_FAIR_GROUP_SCHED */

422 423
static __always_inline
void account_cfs_rq_runtime(struct cfs_rq *cfs_rq, unsigned long delta_exec);
424 425 426 427 428

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

429
static inline u64 max_vruntime(u64 min_vruntime, u64 vruntime)
430
{
431 432
	s64 delta = (s64)(vruntime - min_vruntime);
	if (delta > 0)
433 434 435 436 437
		min_vruntime = vruntime;

	return min_vruntime;
}

438
static inline u64 min_vruntime(u64 min_vruntime, u64 vruntime)
P
Peter Zijlstra 已提交
439 440 441 442 443 444 445 446
{
	s64 delta = (s64)(vruntime - min_vruntime);
	if (delta < 0)
		min_vruntime = vruntime;

	return min_vruntime;
}

447 448 449 450 451 452
static inline int entity_before(struct sched_entity *a,
				struct sched_entity *b)
{
	return (s64)(a->vruntime - b->vruntime) < 0;
}

453 454 455 456 457 458 459 460 461 462 463 464
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
465
		if (!cfs_rq->curr)
466 467 468 469 470 471
			vruntime = se->vruntime;
		else
			vruntime = min_vruntime(vruntime, se->vruntime);
	}

	cfs_rq->min_vruntime = max_vruntime(cfs_rq->min_vruntime, vruntime);
472 473 474 475
#ifndef CONFIG_64BIT
	smp_wmb();
	cfs_rq->min_vruntime_copy = cfs_rq->min_vruntime;
#endif
476 477
}

478 479 480
/*
 * Enqueue an entity into the rb-tree:
 */
481
static void __enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497
{
	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.
		 */
498
		if (entity_before(se, entry)) {
499 500 501 502 503 504 505 506 507 508 509
			link = &parent->rb_left;
		} else {
			link = &parent->rb_right;
			leftmost = 0;
		}
	}

	/*
	 * Maintain a cache of leftmost tree entries (it is frequently
	 * used):
	 */
510
	if (leftmost)
I
Ingo Molnar 已提交
511
		cfs_rq->rb_leftmost = &se->run_node;
512 513 514 515 516

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

517
static void __dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
518
{
P
Peter Zijlstra 已提交
519 520 521 522 523 524
	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
525

526 527 528
	rb_erase(&se->run_node, &cfs_rq->tasks_timeline);
}

529
struct sched_entity *__pick_first_entity(struct cfs_rq *cfs_rq)
530
{
531 532 533 534 535 536
	struct rb_node *left = cfs_rq->rb_leftmost;

	if (!left)
		return NULL;

	return rb_entry(left, struct sched_entity, run_node);
537 538
}

539 540 541 542 543 544 545 546 547 548 549
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
550
struct sched_entity *__pick_last_entity(struct cfs_rq *cfs_rq)
551
{
552
	struct rb_node *last = rb_last(&cfs_rq->tasks_timeline);
553

554 555
	if (!last)
		return NULL;
556 557

	return rb_entry(last, struct sched_entity, run_node);
558 559
}

560 561 562 563
/**************************************************************
 * Scheduling class statistics methods:
 */

564
int sched_proc_update_handler(struct ctl_table *table, int write,
565
		void __user *buffer, size_t *lenp,
566 567
		loff_t *ppos)
{
568
	int ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
569
	int factor = get_update_sysctl_factor();
570 571 572 573 574 575 576

	if (ret || !write)
		return ret;

	sched_nr_latency = DIV_ROUND_UP(sysctl_sched_latency,
					sysctl_sched_min_granularity);

577 578 579 580 581 582 583
#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

584 585 586
	return 0;
}
#endif
587

588
/*
589
 * delta /= w
590 591 592 593
 */
static inline unsigned long
calc_delta_fair(unsigned long delta, struct sched_entity *se)
{
594 595
	if (unlikely(se->load.weight != NICE_0_LOAD))
		delta = calc_delta_mine(delta, NICE_0_LOAD, &se->load);
596 597 598 599

	return delta;
}

600 601 602
/*
 * The idea is to set a period in which each task runs once.
 *
603
 * When there are too many tasks (sched_nr_latency) we have to stretch
604 605 606 607
 * this period because otherwise the slices get too small.
 *
 * p = (nr <= nl) ? l : l*nr/nl
 */
608 609 610
static u64 __sched_period(unsigned long nr_running)
{
	u64 period = sysctl_sched_latency;
611
	unsigned long nr_latency = sched_nr_latency;
612 613

	if (unlikely(nr_running > nr_latency)) {
614
		period = sysctl_sched_min_granularity;
615 616 617 618 619 620
		period *= nr_running;
	}

	return period;
}

621 622 623 624
/*
 * We calculate the wall-time slice from the period by taking a part
 * proportional to the weight.
 *
625
 * s = p*P[w/rw]
626
 */
P
Peter Zijlstra 已提交
627
static u64 sched_slice(struct cfs_rq *cfs_rq, struct sched_entity *se)
628
{
M
Mike Galbraith 已提交
629
	u64 slice = __sched_period(cfs_rq->nr_running + !se->on_rq);
630

M
Mike Galbraith 已提交
631
	for_each_sched_entity(se) {
L
Lin Ming 已提交
632
		struct load_weight *load;
633
		struct load_weight lw;
L
Lin Ming 已提交
634 635 636

		cfs_rq = cfs_rq_of(se);
		load = &cfs_rq->load;
637

M
Mike Galbraith 已提交
638
		if (unlikely(!se->on_rq)) {
639
			lw = cfs_rq->load;
M
Mike Galbraith 已提交
640 641 642 643 644 645 646

			update_load_add(&lw, se->load.weight);
			load = &lw;
		}
		slice = calc_delta_mine(slice, se->load.weight, load);
	}
	return slice;
647 648
}

649
/*
650
 * We calculate the vruntime slice of a to be inserted task
651
 *
652
 * vs = s/w
653
 */
654
static u64 sched_vslice(struct cfs_rq *cfs_rq, struct sched_entity *se)
P
Peter Zijlstra 已提交
655
{
656
	return calc_delta_fair(sched_slice(cfs_rq, se), se);
657 658
}

659
static void update_cfs_load(struct cfs_rq *cfs_rq, int global_update);
660
static void update_cfs_shares(struct cfs_rq *cfs_rq);
661

662 663 664 665 666
/*
 * 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
667 668
__update_curr(struct cfs_rq *cfs_rq, struct sched_entity *curr,
	      unsigned long delta_exec)
669
{
670
	unsigned long delta_exec_weighted;
671

672 673
	schedstat_set(curr->statistics.exec_max,
		      max((u64)delta_exec, curr->statistics.exec_max));
674 675

	curr->sum_exec_runtime += delta_exec;
676
	schedstat_add(cfs_rq, exec_clock, delta_exec);
677
	delta_exec_weighted = calc_delta_fair(delta_exec, curr);
678

I
Ingo Molnar 已提交
679
	curr->vruntime += delta_exec_weighted;
680
	update_min_vruntime(cfs_rq);
681

P
Peter Zijlstra 已提交
682
#if defined CONFIG_SMP && defined CONFIG_FAIR_GROUP_SCHED
683 684
	cfs_rq->load_unacc_exec_time += delta_exec;
#endif
685 686
}

687
static void update_curr(struct cfs_rq *cfs_rq)
688
{
689
	struct sched_entity *curr = cfs_rq->curr;
690
	u64 now = rq_of(cfs_rq)->clock_task;
691 692 693 694 695 696 697 698 699 700
	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
701
	delta_exec = (unsigned long)(now - curr->exec_start);
P
Peter Zijlstra 已提交
702 703
	if (!delta_exec)
		return;
704

I
Ingo Molnar 已提交
705 706
	__update_curr(cfs_rq, curr, delta_exec);
	curr->exec_start = now;
707 708 709 710

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

711
		trace_sched_stat_runtime(curtask, delta_exec, curr->vruntime);
712
		cpuacct_charge(curtask, delta_exec);
713
		account_group_exec_runtime(curtask, delta_exec);
714
	}
715 716

	account_cfs_rq_runtime(cfs_rq, delta_exec);
717 718 719
}

static inline void
720
update_stats_wait_start(struct cfs_rq *cfs_rq, struct sched_entity *se)
721
{
722
	schedstat_set(se->statistics.wait_start, rq_of(cfs_rq)->clock);
723 724 725 726 727
}

/*
 * Task is being enqueued - update stats:
 */
728
static void update_stats_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se)
729 730 731 732 733
{
	/*
	 * Are we enqueueing a waiting task? (for current tasks
	 * a dequeue/enqueue event is a NOP)
	 */
734
	if (se != cfs_rq->curr)
735
		update_stats_wait_start(cfs_rq, se);
736 737 738
}

static void
739
update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se)
740
{
741 742 743 744 745
	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);
746 747 748
#ifdef CONFIG_SCHEDSTATS
	if (entity_is_task(se)) {
		trace_sched_stat_wait(task_of(se),
749
			rq_of(cfs_rq)->clock - se->statistics.wait_start);
750 751
	}
#endif
752
	schedstat_set(se->statistics.wait_start, 0);
753 754 755
}

static inline void
756
update_stats_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se)
757 758 759 760 761
{
	/*
	 * Mark the end of the wait period if dequeueing a
	 * waiting task:
	 */
762
	if (se != cfs_rq->curr)
763
		update_stats_wait_end(cfs_rq, se);
764 765 766 767 768 769
}

/*
 * We are picking a new current task - update its stats:
 */
static inline void
770
update_stats_curr_start(struct cfs_rq *cfs_rq, struct sched_entity *se)
771 772 773 774
{
	/*
	 * We are starting a new run period:
	 */
775
	se->exec_start = rq_of(cfs_rq)->clock_task;
776 777 778 779 780 781
}

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

782 783
#ifdef CONFIG_NUMA_BALANCING
/*
784
 * numa task sample period in ms
785
 */
786
unsigned int sysctl_numa_balancing_scan_period_min = 100;
787 788
unsigned int sysctl_numa_balancing_scan_period_max = 100*50;
unsigned int sysctl_numa_balancing_scan_period_reset = 100*600;
789 790 791

/* Portion of address space to scan in MB */
unsigned int sysctl_numa_balancing_scan_size = 256;
792

793 794 795
/* Scan @scan_size MB every @scan_period after an initial @scan_delay in ms */
unsigned int sysctl_numa_balancing_scan_delay = 1000;

796 797 798 799 800 801 802 803 804 805 806 807 808 809
static void task_numa_placement(struct task_struct *p)
{
	int seq = ACCESS_ONCE(p->mm->numa_scan_seq);

	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.
 */
810
void task_numa_fault(int node, int pages, bool migrated)
811 812 813 814 815
{
	struct task_struct *p = current;

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

816
	/*
817 818
	 * If pages are properly placed (did not migrate) then scan slower.
	 * This is reset periodically in case of phase changes
819
	 */
820 821 822
        if (!migrated)
		p->numa_scan_period = min(sysctl_numa_balancing_scan_period_max,
			p->numa_scan_period + jiffies_to_msecs(10));
823

824 825 826
	task_numa_placement(p);
}

827 828 829 830 831 832
static void reset_ptenuma_scan(struct task_struct *p)
{
	ACCESS_ONCE(p->mm->numa_scan_seq)++;
	p->mm->numa_scan_offset = 0;
}

833 834 835 836 837 838 839 840 841
/*
 * 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;
842
	struct vm_area_struct *vma;
843 844
	unsigned long start, end;
	long pages;
845 846 847 848 849 850 851 852 853 854 855 856 857 858 859

	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;

860 861 862 863 864 865 866 867 868 869 870 871 872
	/*
	 * 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);
	}

873 874 875 876 877 878 879 880 881 882
	/*
	 * 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;

883
	next_scan = now + msecs_to_jiffies(p->numa_scan_period);
884 885 886
	if (cmpxchg(&mm->numa_next_scan, migrate, next_scan) != migrate)
		return;

887 888 889 890 891 892 893 894
	/*
	 * 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;

895 896 897 898 899
	start = mm->numa_scan_offset;
	pages = sysctl_numa_balancing_scan_size;
	pages <<= 20 - PAGE_SHIFT; /* MB in pages */
	if (!pages)
		return;
900

901
	down_read(&mm->mmap_sem);
902
	vma = find_vma(mm, start);
903 904
	if (!vma) {
		reset_ptenuma_scan(p);
905
		start = 0;
906 907
		vma = mm->mmap;
	}
908
	for (; vma; vma = vma->vm_next) {
909 910 911 912 913 914 915
		if (!vma_migratable(vma))
			continue;

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

916 917 918 919 920
		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);
921

922 923 924 925
			start = end;
			if (pages <= 0)
				goto out;
		} while (end != vma->vm_end);
926
	}
927

928
out:
929 930 931 932 933 934 935
	/*
	 * 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)
936
		mm->numa_scan_offset = start;
937 938 939
	else
		reset_ptenuma_scan(p);
	up_read(&mm->mmap_sem);
940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965
}

/*
 * 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) {
966 967
		if (!curr->node_stamp)
			curr->numa_scan_period = sysctl_numa_balancing_scan_period_min;
968 969 970 971 972 973 974 975 976 977 978 979 980 981
		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 */

982 983 984 985
static void
account_entity_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se)
{
	update_load_add(&cfs_rq->load, se->load.weight);
986
	if (!parent_entity(se))
987
		update_load_add(&rq_of(cfs_rq)->load, se->load.weight);
988 989
#ifdef CONFIG_SMP
	if (entity_is_task(se))
990
		list_add(&se->group_node, &rq_of(cfs_rq)->cfs_tasks);
991
#endif
992 993 994 995 996 997 998
	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);
999
	if (!parent_entity(se))
1000
		update_load_sub(&rq_of(cfs_rq)->load, se->load.weight);
1001
	if (entity_is_task(se))
1002
		list_del_init(&se->group_node);
1003 1004 1005
	cfs_rq->nr_running--;
}

1006
#ifdef CONFIG_FAIR_GROUP_SCHED
1007 1008
/* we need this in update_cfs_load and load-balance functions below */
static inline int throttled_hierarchy(struct cfs_rq *cfs_rq);
1009
# ifdef CONFIG_SMP
1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025
static void update_cfs_rq_load_contribution(struct cfs_rq *cfs_rq,
					    int global_update)
{
	struct task_group *tg = cfs_rq->tg;
	long load_avg;

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

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

static void update_cfs_load(struct cfs_rq *cfs_rq, int global_update)
P
Peter Zijlstra 已提交
1026
{
1027
	u64 period = sysctl_sched_shares_window;
P
Peter Zijlstra 已提交
1028
	u64 now, delta;
1029
	unsigned long load = cfs_rq->load.weight;
P
Peter Zijlstra 已提交
1030

1031
	if (cfs_rq->tg == &root_task_group || throttled_hierarchy(cfs_rq))
P
Peter Zijlstra 已提交
1032 1033
		return;

1034
	now = rq_of(cfs_rq)->clock_task;
P
Peter Zijlstra 已提交
1035 1036
	delta = now - cfs_rq->load_stamp;

1037 1038 1039 1040 1041
	/* truncate load history at 4 idle periods */
	if (cfs_rq->load_stamp > cfs_rq->load_last &&
	    now - cfs_rq->load_last > 4 * period) {
		cfs_rq->load_period = 0;
		cfs_rq->load_avg = 0;
1042
		delta = period - 1;
1043 1044
	}

P
Peter Zijlstra 已提交
1045
	cfs_rq->load_stamp = now;
1046
	cfs_rq->load_unacc_exec_time = 0;
P
Peter Zijlstra 已提交
1047
	cfs_rq->load_period += delta;
1048 1049 1050 1051
	if (load) {
		cfs_rq->load_last = now;
		cfs_rq->load_avg += delta * load;
	}
P
Peter Zijlstra 已提交
1052

1053 1054 1055 1056 1057
	/* consider updating load contribution on each fold or truncate */
	if (global_update || cfs_rq->load_period > period
	    || !cfs_rq->load_period)
		update_cfs_rq_load_contribution(cfs_rq, global_update);

P
Peter Zijlstra 已提交
1058 1059 1060 1061 1062 1063 1064 1065 1066 1067
	while (cfs_rq->load_period > period) {
		/*
		 * Inline assembly required to prevent the compiler
		 * optimising this loop into a divmod call.
		 * See __iter_div_u64_rem() for another example of this.
		 */
		asm("" : "+rm" (cfs_rq->load_period));
		cfs_rq->load_period /= 2;
		cfs_rq->load_avg /= 2;
	}
1068

1069 1070
	if (!cfs_rq->curr && !cfs_rq->nr_running && !cfs_rq->load_avg)
		list_del_leaf_cfs_rq(cfs_rq);
P
Peter Zijlstra 已提交
1071 1072
}

1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088
static inline long calc_tg_weight(struct task_group *tg, struct cfs_rq *cfs_rq)
{
	long tg_weight;

	/*
	 * Use this CPU's actual weight instead of the last load_contribution
	 * to gain a more accurate current total weight. See
	 * update_cfs_rq_load_contribution().
	 */
	tg_weight = atomic_read(&tg->load_weight);
	tg_weight -= cfs_rq->load_contribution;
	tg_weight += cfs_rq->load.weight;

	return tg_weight;
}

1089
static long calc_cfs_shares(struct cfs_rq *cfs_rq, struct task_group *tg)
1090
{
1091
	long tg_weight, load, shares;
1092

1093
	tg_weight = calc_tg_weight(tg, cfs_rq);
1094
	load = cfs_rq->load.weight;
1095 1096

	shares = (tg->shares * load);
1097 1098
	if (tg_weight)
		shares /= tg_weight;
1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111

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

	return shares;
}

static void update_entity_shares_tick(struct cfs_rq *cfs_rq)
{
	if (cfs_rq->load_unacc_exec_time > sysctl_sched_shares_window) {
		update_cfs_load(cfs_rq, 0);
1112
		update_cfs_shares(cfs_rq);
1113 1114 1115 1116 1117 1118 1119
	}
}
# else /* CONFIG_SMP */
static void update_cfs_load(struct cfs_rq *cfs_rq, int global_update)
{
}

1120
static inline long calc_cfs_shares(struct cfs_rq *cfs_rq, struct task_group *tg)
1121 1122 1123 1124 1125 1126 1127 1128
{
	return tg->shares;
}

static inline void update_entity_shares_tick(struct cfs_rq *cfs_rq)
{
}
# endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
1129 1130 1131
static void reweight_entity(struct cfs_rq *cfs_rq, struct sched_entity *se,
			    unsigned long weight)
{
1132 1133 1134 1135
	if (se->on_rq) {
		/* commit outstanding execution time */
		if (cfs_rq->curr == se)
			update_curr(cfs_rq);
P
Peter Zijlstra 已提交
1136
		account_entity_dequeue(cfs_rq, se);
1137
	}
P
Peter Zijlstra 已提交
1138 1139 1140 1141 1142 1143 1144

	update_load_set(&se->load, weight);

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

1145
static void update_cfs_shares(struct cfs_rq *cfs_rq)
P
Peter Zijlstra 已提交
1146 1147 1148
{
	struct task_group *tg;
	struct sched_entity *se;
1149
	long shares;
P
Peter Zijlstra 已提交
1150 1151 1152

	tg = cfs_rq->tg;
	se = tg->se[cpu_of(rq_of(cfs_rq))];
1153
	if (!se || throttled_hierarchy(cfs_rq))
P
Peter Zijlstra 已提交
1154
		return;
1155 1156 1157 1158
#ifndef CONFIG_SMP
	if (likely(se->load.weight == tg->shares))
		return;
#endif
1159
	shares = calc_cfs_shares(cfs_rq, tg);
P
Peter Zijlstra 已提交
1160 1161 1162 1163

	reweight_entity(cfs_rq_of(se), se, shares);
}
#else /* CONFIG_FAIR_GROUP_SCHED */
1164
static void update_cfs_load(struct cfs_rq *cfs_rq, int global_update)
P
Peter Zijlstra 已提交
1165 1166 1167
{
}

1168
static inline void update_cfs_shares(struct cfs_rq *cfs_rq)
P
Peter Zijlstra 已提交
1169 1170
{
}
1171 1172 1173 1174

static inline void update_entity_shares_tick(struct cfs_rq *cfs_rq)
{
}
P
Peter Zijlstra 已提交
1175 1176
#endif /* CONFIG_FAIR_GROUP_SCHED */

1177
static void enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
1178 1179
{
#ifdef CONFIG_SCHEDSTATS
1180 1181 1182 1183 1184
	struct task_struct *tsk = NULL;

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

1185 1186
	if (se->statistics.sleep_start) {
		u64 delta = rq_of(cfs_rq)->clock - se->statistics.sleep_start;
1187 1188 1189 1190

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

1191 1192
		if (unlikely(delta > se->statistics.sleep_max))
			se->statistics.sleep_max = delta;
1193

1194
		se->statistics.sleep_start = 0;
1195
		se->statistics.sum_sleep_runtime += delta;
A
Arjan van de Ven 已提交
1196

1197
		if (tsk) {
1198
			account_scheduler_latency(tsk, delta >> 10, 1);
1199 1200
			trace_sched_stat_sleep(tsk, delta);
		}
1201
	}
1202 1203
	if (se->statistics.block_start) {
		u64 delta = rq_of(cfs_rq)->clock - se->statistics.block_start;
1204 1205 1206 1207

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

1208 1209
		if (unlikely(delta > se->statistics.block_max))
			se->statistics.block_max = delta;
1210

1211
		se->statistics.block_start = 0;
1212
		se->statistics.sum_sleep_runtime += delta;
I
Ingo Molnar 已提交
1213

1214
		if (tsk) {
1215
			if (tsk->in_iowait) {
1216 1217
				se->statistics.iowait_sum += delta;
				se->statistics.iowait_count++;
1218
				trace_sched_stat_iowait(tsk, delta);
1219 1220
			}

1221 1222
			trace_sched_stat_blocked(tsk, delta);

1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233
			/*
			 * 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
1234
		}
1235 1236 1237 1238
	}
#endif
}

P
Peter Zijlstra 已提交
1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251
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
}

1252 1253 1254
static void
place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial)
{
1255
	u64 vruntime = cfs_rq->min_vruntime;
P
Peter Zijlstra 已提交
1256

1257 1258 1259 1260 1261 1262
	/*
	 * 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
1263
	if (initial && sched_feat(START_DEBIT))
1264
		vruntime += sched_vslice(cfs_rq, se);
1265

1266
	/* sleeps up to a single latency don't count. */
1267
	if (!initial) {
1268
		unsigned long thresh = sysctl_sched_latency;
1269

1270 1271 1272 1273 1274 1275
		/*
		 * Halve their sleep time's effect, to allow
		 * for a gentler effect of sleepers:
		 */
		if (sched_feat(GENTLE_FAIR_SLEEPERS))
			thresh >>= 1;
1276

1277
		vruntime -= thresh;
1278 1279
	}

1280 1281 1282
	/* ensure we never gain time by being placed backwards. */
	vruntime = max_vruntime(se->vruntime, vruntime);

P
Peter Zijlstra 已提交
1283
	se->vruntime = vruntime;
1284 1285
}

1286 1287
static void check_enqueue_throttle(struct cfs_rq *cfs_rq);

1288
static void
1289
enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
1290
{
1291 1292 1293 1294
	/*
	 * Update the normalized vruntime before updating min_vruntime
	 * through callig update_curr().
	 */
1295
	if (!(flags & ENQUEUE_WAKEUP) || (flags & ENQUEUE_WAKING))
1296 1297
		se->vruntime += cfs_rq->min_vruntime;

1298
	/*
1299
	 * Update run-time statistics of the 'current'.
1300
	 */
1301
	update_curr(cfs_rq);
1302
	update_cfs_load(cfs_rq, 0);
P
Peter Zijlstra 已提交
1303
	account_entity_enqueue(cfs_rq, se);
1304
	update_cfs_shares(cfs_rq);
1305

1306
	if (flags & ENQUEUE_WAKEUP) {
1307
		place_entity(cfs_rq, se, 0);
1308
		enqueue_sleeper(cfs_rq, se);
I
Ingo Molnar 已提交
1309
	}
1310

1311
	update_stats_enqueue(cfs_rq, se);
P
Peter Zijlstra 已提交
1312
	check_spread(cfs_rq, se);
1313 1314
	if (se != cfs_rq->curr)
		__enqueue_entity(cfs_rq, se);
P
Peter Zijlstra 已提交
1315
	se->on_rq = 1;
1316

1317
	if (cfs_rq->nr_running == 1) {
1318
		list_add_leaf_cfs_rq(cfs_rq);
1319 1320
		check_enqueue_throttle(cfs_rq);
	}
1321 1322
}

1323
static void __clear_buddies_last(struct sched_entity *se)
P
Peter Zijlstra 已提交
1324
{
1325 1326 1327 1328 1329 1330 1331 1332
	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
1333

1334 1335 1336 1337 1338 1339 1340 1341 1342
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
1343 1344
}

1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355
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
1356 1357
static void clear_buddies(struct cfs_rq *cfs_rq, struct sched_entity *se)
{
1358 1359 1360 1361 1362
	if (cfs_rq->last == se)
		__clear_buddies_last(se);

	if (cfs_rq->next == se)
		__clear_buddies_next(se);
1363 1364 1365

	if (cfs_rq->skip == se)
		__clear_buddies_skip(se);
P
Peter Zijlstra 已提交
1366 1367
}

1368
static __always_inline void return_cfs_rq_runtime(struct cfs_rq *cfs_rq);
1369

1370
static void
1371
dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
1372
{
1373 1374 1375 1376 1377
	/*
	 * Update run-time statistics of the 'current'.
	 */
	update_curr(cfs_rq);

1378
	update_stats_dequeue(cfs_rq, se);
1379
	if (flags & DEQUEUE_SLEEP) {
P
Peter Zijlstra 已提交
1380
#ifdef CONFIG_SCHEDSTATS
1381 1382 1383 1384
		if (entity_is_task(se)) {
			struct task_struct *tsk = task_of(se);

			if (tsk->state & TASK_INTERRUPTIBLE)
1385
				se->statistics.sleep_start = rq_of(cfs_rq)->clock;
1386
			if (tsk->state & TASK_UNINTERRUPTIBLE)
1387
				se->statistics.block_start = rq_of(cfs_rq)->clock;
1388
		}
1389
#endif
P
Peter Zijlstra 已提交
1390 1391
	}

P
Peter Zijlstra 已提交
1392
	clear_buddies(cfs_rq, se);
P
Peter Zijlstra 已提交
1393

1394
	if (se != cfs_rq->curr)
1395
		__dequeue_entity(cfs_rq, se);
P
Peter Zijlstra 已提交
1396
	se->on_rq = 0;
1397
	update_cfs_load(cfs_rq, 0);
1398
	account_entity_dequeue(cfs_rq, se);
1399 1400 1401 1402 1403 1404

	/*
	 * 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.
	 */
1405
	if (!(flags & DEQUEUE_SLEEP))
1406
		se->vruntime -= cfs_rq->min_vruntime;
1407

1408 1409 1410
	/* return excess runtime on last dequeue */
	return_cfs_rq_runtime(cfs_rq);

1411 1412
	update_min_vruntime(cfs_rq);
	update_cfs_shares(cfs_rq);
1413 1414 1415 1416 1417
}

/*
 * Preempt the current task with a newly woken task if needed:
 */
1418
static void
I
Ingo Molnar 已提交
1419
check_preempt_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr)
1420
{
1421
	unsigned long ideal_runtime, delta_exec;
1422 1423
	struct sched_entity *se;
	s64 delta;
1424

P
Peter Zijlstra 已提交
1425
	ideal_runtime = sched_slice(cfs_rq, curr);
1426
	delta_exec = curr->sum_exec_runtime - curr->prev_sum_exec_runtime;
1427
	if (delta_exec > ideal_runtime) {
1428
		resched_task(rq_of(cfs_rq)->curr);
1429 1430 1431 1432 1433
		/*
		 * The current task ran long enough, ensure it doesn't get
		 * re-elected due to buddy favours.
		 */
		clear_buddies(cfs_rq, curr);
1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444
		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;

1445 1446
	se = __pick_first_entity(cfs_rq);
	delta = curr->vruntime - se->vruntime;
1447

1448 1449
	if (delta < 0)
		return;
1450

1451 1452
	if (delta > ideal_runtime)
		resched_task(rq_of(cfs_rq)->curr);
1453 1454
}

1455
static void
1456
set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
1457
{
1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468
	/* '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);
	}

1469
	update_stats_curr_start(cfs_rq, se);
1470
	cfs_rq->curr = se;
I
Ingo Molnar 已提交
1471 1472 1473 1474 1475 1476
#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):
	 */
1477
	if (rq_of(cfs_rq)->load.weight >= 2*se->load.weight) {
1478
		se->statistics.slice_max = max(se->statistics.slice_max,
I
Ingo Molnar 已提交
1479 1480 1481
			se->sum_exec_runtime - se->prev_sum_exec_runtime);
	}
#endif
1482
	se->prev_sum_exec_runtime = se->sum_exec_runtime;
1483 1484
}

1485 1486 1487
static int
wakeup_preempt_entity(struct sched_entity *curr, struct sched_entity *se);

1488 1489 1490 1491 1492 1493 1494
/*
 * 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
 */
1495
static struct sched_entity *pick_next_entity(struct cfs_rq *cfs_rq)
1496
{
1497
	struct sched_entity *se = __pick_first_entity(cfs_rq);
1498
	struct sched_entity *left = se;
1499

1500 1501 1502 1503 1504 1505 1506 1507 1508
	/*
	 * 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;
	}
1509

1510 1511 1512 1513 1514 1515
	/*
	 * 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;

1516 1517 1518 1519 1520 1521
	/*
	 * 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;

1522
	clear_buddies(cfs_rq, se);
P
Peter Zijlstra 已提交
1523 1524

	return se;
1525 1526
}

1527 1528
static void check_cfs_rq_runtime(struct cfs_rq *cfs_rq);

1529
static void put_prev_entity(struct cfs_rq *cfs_rq, struct sched_entity *prev)
1530 1531 1532 1533 1534 1535
{
	/*
	 * If still on the runqueue then deactivate_task()
	 * was not called and update_curr() has to be done:
	 */
	if (prev->on_rq)
1536
		update_curr(cfs_rq);
1537

1538 1539 1540
	/* throttle cfs_rqs exceeding runtime */
	check_cfs_rq_runtime(cfs_rq);

P
Peter Zijlstra 已提交
1541
	check_spread(cfs_rq, prev);
1542
	if (prev->on_rq) {
1543
		update_stats_wait_start(cfs_rq, prev);
1544 1545 1546
		/* Put 'current' back into the tree. */
		__enqueue_entity(cfs_rq, prev);
	}
1547
	cfs_rq->curr = NULL;
1548 1549
}

P
Peter Zijlstra 已提交
1550 1551
static void
entity_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr, int queued)
1552 1553
{
	/*
1554
	 * Update run-time statistics of the 'current'.
1555
	 */
1556
	update_curr(cfs_rq);
1557

1558 1559 1560 1561 1562
	/*
	 * Update share accounting for long-running entities.
	 */
	update_entity_shares_tick(cfs_rq);

P
Peter Zijlstra 已提交
1563 1564 1565 1566 1567
#ifdef CONFIG_SCHED_HRTICK
	/*
	 * queued ticks are scheduled to match the slice, so don't bother
	 * validating it and just reschedule.
	 */
1568 1569 1570 1571
	if (queued) {
		resched_task(rq_of(cfs_rq)->curr);
		return;
	}
P
Peter Zijlstra 已提交
1572 1573 1574 1575 1576 1577 1578 1579
	/*
	 * 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
1580
	if (cfs_rq->nr_running > 1)
I
Ingo Molnar 已提交
1581
		check_preempt_tick(cfs_rq, curr);
1582 1583
}

1584 1585 1586 1587 1588 1589

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

#ifdef CONFIG_CFS_BANDWIDTH
1590 1591

#ifdef HAVE_JUMP_LABEL
1592
static struct static_key __cfs_bandwidth_used;
1593 1594 1595

static inline bool cfs_bandwidth_used(void)
{
1596
	return static_key_false(&__cfs_bandwidth_used);
1597 1598 1599 1600 1601 1602
}

void account_cfs_bandwidth_used(int enabled, int was_enabled)
{
	/* only need to count groups transitioning between enabled/!enabled */
	if (enabled && !was_enabled)
1603
		static_key_slow_inc(&__cfs_bandwidth_used);
1604
	else if (!enabled && was_enabled)
1605
		static_key_slow_dec(&__cfs_bandwidth_used);
1606 1607 1608 1609 1610 1611 1612 1613 1614 1615
}
#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 */

1616 1617 1618 1619 1620 1621 1622 1623
/*
 * default period for cfs group bandwidth.
 * default: 0.1s, units: nanoseconds
 */
static inline u64 default_cfs_period(void)
{
	return 100000000ULL;
}
1624 1625 1626 1627 1628 1629

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

P
Paul Turner 已提交
1630 1631 1632 1633 1634 1635 1636
/*
 * 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
 */
1637
void __refill_cfs_bandwidth_runtime(struct cfs_bandwidth *cfs_b)
P
Paul Turner 已提交
1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648
{
	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);
}

1649 1650 1651 1652 1653
static inline struct cfs_bandwidth *tg_cfs_bandwidth(struct task_group *tg)
{
	return &tg->cfs_bandwidth;
}

1654 1655
/* returns 0 on failure to allocate runtime */
static int assign_cfs_rq_runtime(struct cfs_rq *cfs_rq)
1656 1657 1658
{
	struct task_group *tg = cfs_rq->tg;
	struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(tg);
P
Paul Turner 已提交
1659
	u64 amount = 0, min_amount, expires;
1660 1661 1662 1663 1664 1665 1666

	/* 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;
1667
	else {
P
Paul Turner 已提交
1668 1669 1670 1671 1672 1673 1674 1675
		/*
		 * 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);
1676
			__start_cfs_bandwidth(cfs_b);
P
Paul Turner 已提交
1677
		}
1678 1679 1680 1681 1682 1683

		if (cfs_b->runtime > 0) {
			amount = min(cfs_b->runtime, min_amount);
			cfs_b->runtime -= amount;
			cfs_b->idle = 0;
		}
1684
	}
P
Paul Turner 已提交
1685
	expires = cfs_b->runtime_expires;
1686 1687 1688
	raw_spin_unlock(&cfs_b->lock);

	cfs_rq->runtime_remaining += amount;
P
Paul Turner 已提交
1689 1690 1691 1692 1693 1694 1695
	/*
	 * 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;
1696 1697

	return cfs_rq->runtime_remaining > 0;
1698 1699
}

P
Paul Turner 已提交
1700 1701 1702 1703 1704
/*
 * 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)
1705
{
P
Paul Turner 已提交
1706 1707 1708 1709 1710
	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))
1711 1712
		return;

P
Paul Turner 已提交
1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737
	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) */
1738
	cfs_rq->runtime_remaining -= delta_exec;
P
Paul Turner 已提交
1739 1740 1741
	expire_cfs_rq_runtime(cfs_rq);

	if (likely(cfs_rq->runtime_remaining > 0))
1742 1743
		return;

1744 1745 1746 1747 1748 1749
	/*
	 * 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);
1750 1751
}

1752 1753
static __always_inline
void account_cfs_rq_runtime(struct cfs_rq *cfs_rq, unsigned long delta_exec)
1754
{
1755
	if (!cfs_bandwidth_used() || !cfs_rq->runtime_enabled)
1756 1757 1758 1759 1760
		return;

	__account_cfs_rq_runtime(cfs_rq, delta_exec);
}

1761 1762
static inline int cfs_rq_throttled(struct cfs_rq *cfs_rq)
{
1763
	return cfs_bandwidth_used() && cfs_rq->throttled;
1764 1765
}

1766 1767 1768
/* check whether cfs_rq, or any parent, is throttled */
static inline int throttled_hierarchy(struct cfs_rq *cfs_rq)
{
1769
	return cfs_bandwidth_used() && cfs_rq->throttle_count;
1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824
}

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

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

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

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

	cfs_rq->throttle_count--;
#ifdef CONFIG_SMP
	if (!cfs_rq->throttle_count) {
		u64 delta = rq->clock_task - cfs_rq->load_stamp;

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

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

	return 0;
}

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

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

	return 0;
}

1825
static void throttle_cfs_rq(struct cfs_rq *cfs_rq)
1826 1827 1828 1829 1830 1831 1832 1833 1834
{
	struct rq *rq = rq_of(cfs_rq);
	struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(cfs_rq->tg);
	struct sched_entity *se;
	long task_delta, dequeue = 1;

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

	/* account load preceding throttle */
1835 1836 1837
	rcu_read_lock();
	walk_tg_tree_from(cfs_rq->tg, tg_throttle_down, tg_nop, (void *)rq);
	rcu_read_unlock();
1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857

	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;
1858
	cfs_rq->throttled_timestamp = rq->clock;
1859 1860 1861 1862 1863
	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);
}

1864
void unthrottle_cfs_rq(struct cfs_rq *cfs_rq)
1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875
{
	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);
1876
	cfs_b->throttled_time += rq->clock - cfs_rq->throttled_timestamp;
1877 1878
	list_del_rcu(&cfs_rq->throttled_list);
	raw_spin_unlock(&cfs_b->lock);
1879
	cfs_rq->throttled_timestamp = 0;
1880

1881 1882 1883 1884
	update_rq_clock(rq);
	/* update hierarchical throttle state */
	walk_tg_tree_from(cfs_rq->tg, tg_nop, tg_unthrottle_up, (void *)rq);

1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947
	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;
}

1948 1949 1950 1951 1952 1953 1954 1955
/*
 * 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)
{
1956 1957
	u64 runtime, runtime_expires;
	int idle = 1, throttled;
1958 1959 1960 1961 1962 1963

	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;

1964 1965 1966
	throttled = !list_empty(&cfs_b->throttled_cfs_rq);
	/* idle depends on !throttled (for the case of a large deficit) */
	idle = cfs_b->idle && !throttled;
1967
	cfs_b->nr_periods += overrun;
1968

P
Paul Turner 已提交
1969 1970 1971 1972 1973 1974
	/* if we're going inactive then everything else can be deferred */
	if (idle)
		goto out_unlock;

	__refill_cfs_bandwidth_runtime(cfs_b);

1975 1976 1977 1978 1979 1980
	if (!throttled) {
		/* mark as potentially idle for the upcoming period */
		cfs_b->idle = 1;
		goto out_unlock;
	}

1981 1982 1983
	/* account preceding periods in which throttling occurred */
	cfs_b->nr_throttled += overrun;

1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
	/*
	 * 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);
	}
2008

2009 2010 2011 2012 2013 2014 2015 2016 2017
	/* 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;
2018 2019 2020 2021 2022 2023 2024
out_unlock:
	if (idle)
		cfs_b->timer_active = 0;
	raw_spin_unlock(&cfs_b->lock);

	return idle;
}
2025

2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089
/* 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)
{
2090 2091 2092
	if (!cfs_bandwidth_used())
		return;

2093
	if (!cfs_rq->runtime_enabled || cfs_rq->nr_running)
2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130
		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);
}

2131 2132 2133 2134 2135 2136 2137
/*
 * 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)
{
2138 2139 2140
	if (!cfs_bandwidth_used())
		return;

2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157
	/* 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)
{
2158 2159 2160
	if (!cfs_bandwidth_used())
		return;

2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172
	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);
}
2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 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 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257

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);
}

2258
static void unthrottle_offline_cfs_rqs(struct rq *rq)
2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278
{
	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 */
2279 2280
static __always_inline
void account_cfs_rq_runtime(struct cfs_rq *cfs_rq, unsigned long delta_exec) {}
2281 2282
static void check_cfs_rq_runtime(struct cfs_rq *cfs_rq) {}
static void check_enqueue_throttle(struct cfs_rq *cfs_rq) {}
2283
static __always_inline void return_cfs_rq_runtime(struct cfs_rq *cfs_rq) {}
2284 2285 2286 2287 2288

static inline int cfs_rq_throttled(struct cfs_rq *cfs_rq)
{
	return 0;
}
2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299

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;
}
2300 2301 2302 2303 2304

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) {}
2305 2306
#endif

2307 2308 2309 2310 2311
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) {}
2312
static inline void unthrottle_offline_cfs_rqs(struct rq *rq) {}
2313 2314 2315

#endif /* CONFIG_CFS_BANDWIDTH */

2316 2317 2318 2319
/**************************************************
 * CFS operations on tasks:
 */

P
Peter Zijlstra 已提交
2320 2321 2322 2323 2324 2325 2326 2327
#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);

2328
	if (cfs_rq->nr_running > 1) {
P
Peter Zijlstra 已提交
2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342
		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.
		 */
2343
		if (rq->curr != p)
2344
			delta = max_t(s64, 10000LL, delta);
P
Peter Zijlstra 已提交
2345

2346
		hrtick_start(rq, delta);
P
Peter Zijlstra 已提交
2347 2348
	}
}
2349 2350 2351 2352 2353 2354 2355 2356 2357 2358

/*
 * 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;

2359
	if (!hrtick_enabled(rq) || curr->sched_class != &fair_sched_class)
2360 2361 2362 2363 2364
		return;

	if (cfs_rq_of(&curr->se)->nr_running < sched_nr_latency)
		hrtick_start_fair(rq, curr);
}
2365
#else /* !CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
2366 2367 2368 2369
static inline void
hrtick_start_fair(struct rq *rq, struct task_struct *p)
{
}
2370 2371 2372 2373

static inline void hrtick_update(struct rq *rq)
{
}
P
Peter Zijlstra 已提交
2374 2375
#endif

2376 2377 2378 2379 2380
/*
 * 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:
 */
2381
static void
2382
enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags)
2383 2384
{
	struct cfs_rq *cfs_rq;
2385
	struct sched_entity *se = &p->se;
2386 2387

	for_each_sched_entity(se) {
2388
		if (se->on_rq)
2389 2390
			break;
		cfs_rq = cfs_rq_of(se);
2391
		enqueue_entity(cfs_rq, se, flags);
2392 2393 2394 2395 2396 2397 2398 2399 2400

		/*
		 * 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;
2401
		cfs_rq->h_nr_running++;
2402

2403
		flags = ENQUEUE_WAKEUP;
2404
	}
P
Peter Zijlstra 已提交
2405

P
Peter Zijlstra 已提交
2406
	for_each_sched_entity(se) {
2407
		cfs_rq = cfs_rq_of(se);
2408
		cfs_rq->h_nr_running++;
P
Peter Zijlstra 已提交
2409

2410 2411 2412
		if (cfs_rq_throttled(cfs_rq))
			break;

2413
		update_cfs_load(cfs_rq, 0);
2414
		update_cfs_shares(cfs_rq);
P
Peter Zijlstra 已提交
2415 2416
	}

2417 2418
	if (!se)
		inc_nr_running(rq);
2419
	hrtick_update(rq);
2420 2421
}

2422 2423
static void set_next_buddy(struct sched_entity *se);

2424 2425 2426 2427 2428
/*
 * The dequeue_task method is called before nr_running is
 * decreased. We remove the task from the rbtree and
 * update the fair scheduling stats:
 */
2429
static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags)
2430 2431
{
	struct cfs_rq *cfs_rq;
2432
	struct sched_entity *se = &p->se;
2433
	int task_sleep = flags & DEQUEUE_SLEEP;
2434 2435 2436

	for_each_sched_entity(se) {
		cfs_rq = cfs_rq_of(se);
2437
		dequeue_entity(cfs_rq, se, flags);
2438 2439 2440 2441 2442 2443 2444 2445 2446

		/*
		 * 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;
2447
		cfs_rq->h_nr_running--;
P
Peter Zijlstra 已提交
2448

2449
		/* Don't dequeue parent if it has other entities besides us */
2450 2451 2452 2453 2454 2455 2456
		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));
2457 2458 2459

			/* avoid re-evaluating load for this entity */
			se = parent_entity(se);
2460
			break;
2461
		}
2462
		flags |= DEQUEUE_SLEEP;
2463
	}
P
Peter Zijlstra 已提交
2464

P
Peter Zijlstra 已提交
2465
	for_each_sched_entity(se) {
2466
		cfs_rq = cfs_rq_of(se);
2467
		cfs_rq->h_nr_running--;
P
Peter Zijlstra 已提交
2468

2469 2470 2471
		if (cfs_rq_throttled(cfs_rq))
			break;

2472
		update_cfs_load(cfs_rq, 0);
2473
		update_cfs_shares(cfs_rq);
P
Peter Zijlstra 已提交
2474 2475
	}

2476 2477
	if (!se)
		dec_nr_running(rq);
2478
	hrtick_update(rq);
2479 2480
}

2481
#ifdef CONFIG_SMP
2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536
/* 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;
}

2537

2538
static void task_waking_fair(struct task_struct *p)
2539 2540 2541
{
	struct sched_entity *se = &p->se;
	struct cfs_rq *cfs_rq = cfs_rq_of(se);
2542 2543 2544 2545
	u64 min_vruntime;

#ifndef CONFIG_64BIT
	u64 min_vruntime_copy;
2546

2547 2548 2549 2550 2551 2552 2553 2554
	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
2555

2556
	se->vruntime -= min_vruntime;
2557 2558
}

2559
#ifdef CONFIG_FAIR_GROUP_SCHED
2560 2561 2562 2563 2564 2565
/*
 * 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.
2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608
 *
 * 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.
2609
 */
P
Peter Zijlstra 已提交
2610
static long effective_load(struct task_group *tg, int cpu, long wl, long wg)
2611
{
P
Peter Zijlstra 已提交
2612
	struct sched_entity *se = tg->se[cpu];
2613

2614
	if (!tg->parent)	/* the trivial, non-cgroup case */
2615 2616
		return wl;

P
Peter Zijlstra 已提交
2617
	for_each_sched_entity(se) {
2618
		long w, W;
P
Peter Zijlstra 已提交
2619

2620
		tg = se->my_q->tg;
2621

2622 2623 2624 2625
		/*
		 * W = @wg + \Sum rw_j
		 */
		W = wg + calc_tg_weight(tg, se->my_q);
P
Peter Zijlstra 已提交
2626

2627 2628 2629 2630
		/*
		 * w = rw_i + @wl
		 */
		w = se->my_q->load.weight + wl;
2631

2632 2633 2634 2635 2636
		/*
		 * wl = S * s'_i; see (2)
		 */
		if (W > 0 && w < W)
			wl = (w * tg->shares) / W;
2637 2638
		else
			wl = tg->shares;
2639

2640 2641 2642 2643 2644
		/*
		 * 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().
		 */
2645 2646
		if (wl < MIN_SHARES)
			wl = MIN_SHARES;
2647 2648 2649 2650

		/*
		 * wl = dw_i = S * (s'_i - s_i); see (3)
		 */
2651
		wl -= se->load.weight;
2652 2653 2654 2655 2656 2657 2658 2659

		/*
		 * 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
2660 2661
		wg = 0;
	}
2662

P
Peter Zijlstra 已提交
2663
	return wl;
2664 2665
}
#else
P
Peter Zijlstra 已提交
2666

2667 2668
static inline unsigned long effective_load(struct task_group *tg, int cpu,
		unsigned long wl, unsigned long wg)
P
Peter Zijlstra 已提交
2669
{
2670
	return wl;
2671
}
P
Peter Zijlstra 已提交
2672

2673 2674
#endif

2675
static int wake_affine(struct sched_domain *sd, struct task_struct *p, int sync)
2676
{
2677
	s64 this_load, load;
2678
	int idx, this_cpu, prev_cpu;
2679
	unsigned long tl_per_task;
2680
	struct task_group *tg;
2681
	unsigned long weight;
2682
	int balanced;
2683

2684 2685 2686 2687 2688
	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);
2689

2690 2691 2692 2693 2694
	/*
	 * If sync wakeup then subtract the (maximum possible)
	 * effect of the currently running task from the load
	 * of the current CPU:
	 */
2695 2696 2697 2698
	if (sync) {
		tg = task_group(current);
		weight = current->se.load.weight;

2699
		this_load += effective_load(tg, this_cpu, -weight, -weight);
2700 2701
		load += effective_load(tg, prev_cpu, 0, -weight);
	}
2702

2703 2704
	tg = task_group(p);
	weight = p->se.load.weight;
2705

2706 2707
	/*
	 * In low-load situations, where prev_cpu is idle and this_cpu is idle
2708 2709 2710
	 * 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.
2711 2712 2713 2714
	 *
	 * Otherwise check if either cpus are near enough in load to allow this
	 * task to be woken on this_cpu.
	 */
2715 2716
	if (this_load > 0) {
		s64 this_eff_load, prev_eff_load;
2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729

		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;
2730

2731
	/*
I
Ingo Molnar 已提交
2732 2733 2734
	 * If the currently running task will sleep within
	 * a reasonable amount of time then attract this newly
	 * woken task:
2735
	 */
2736 2737
	if (sync && balanced)
		return 1;
2738

2739
	schedstat_inc(p, se.statistics.nr_wakeups_affine_attempts);
2740 2741
	tl_per_task = cpu_avg_load_per_task(this_cpu);

2742 2743 2744
	if (balanced ||
	    (this_load <= load &&
	     this_load + target_load(prev_cpu, idx) <= tl_per_task)) {
2745 2746 2747 2748 2749
		/*
		 * This domain has SD_WAKE_AFFINE and
		 * p is cache cold in this domain, and
		 * there is no bad imbalance.
		 */
2750
		schedstat_inc(sd, ttwu_move_affine);
2751
		schedstat_inc(p, se.statistics.nr_wakeups_affine);
2752 2753 2754 2755 2756 2757

		return 1;
	}
	return 0;
}

2758 2759 2760 2761 2762
/*
 * find_idlest_group finds and returns the least busy CPU group within the
 * domain.
 */
static struct sched_group *
P
Peter Zijlstra 已提交
2763
find_idlest_group(struct sched_domain *sd, struct task_struct *p,
2764
		  int this_cpu, int load_idx)
2765
{
2766
	struct sched_group *idlest = NULL, *group = sd->groups;
2767 2768
	unsigned long min_load = ULONG_MAX, this_load = 0;
	int imbalance = 100 + (sd->imbalance_pct-100)/2;
2769

2770 2771 2772 2773
	do {
		unsigned long load, avg_load;
		int local_group;
		int i;
2774

2775 2776
		/* Skip over this group if it has no CPUs allowed */
		if (!cpumask_intersects(sched_group_cpus(group),
2777
					tsk_cpus_allowed(p)))
2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796
			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 */
2797
		avg_load = (avg_load * SCHED_POWER_SCALE) / group->sgp->power;
2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822

		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 */
2823
	for_each_cpu_and(i, sched_group_cpus(group), tsk_cpus_allowed(p)) {
2824 2825 2826 2827 2828
		load = weighted_cpuload(i);

		if (load < min_load || (load == min_load && i == this_cpu)) {
			min_load = load;
			idlest = i;
2829 2830 2831
		}
	}

2832 2833
	return idlest;
}
2834

2835 2836 2837
/*
 * Try and locate an idle CPU in the sched_domain.
 */
2838
static int select_idle_sibling(struct task_struct *p, int target)
2839 2840 2841
{
	int cpu = smp_processor_id();
	int prev_cpu = task_cpu(p);
2842
	struct sched_domain *sd;
2843 2844
	struct sched_group *sg;
	int i;
2845 2846

	/*
2847 2848
	 * If the task is going to be woken-up on this cpu and if it is
	 * already idle, then it is the right target.
2849
	 */
2850 2851 2852 2853 2854 2855 2856 2857
	if (target == cpu && idle_cpu(cpu))
		return cpu;

	/*
	 * If the task is going to be woken-up on the cpu where it previously
	 * ran and if it is currently idle, then it the right target.
	 */
	if (target == prev_cpu && idle_cpu(prev_cpu))
2858
		return prev_cpu;
2859 2860

	/*
2861
	 * Otherwise, iterate the domains and find an elegible idle cpu.
2862
	 */
2863
	sd = rcu_dereference(per_cpu(sd_llc, target));
2864
	for_each_lower_domain(sd) {
2865 2866 2867 2868 2869 2870 2871 2872 2873 2874
		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)) {
				if (!idle_cpu(i))
					goto next;
			}
2875

2876 2877 2878 2879 2880 2881 2882 2883
			target = cpumask_first_and(sched_group_cpus(sg),
					tsk_cpus_allowed(p));
			goto done;
next:
			sg = sg->next;
		} while (sg != sd->groups);
	}
done:
2884 2885 2886
	return target;
}

2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897
/*
 * 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.
 */
2898
static int
2899
select_task_rq_fair(struct task_struct *p, int sd_flag, int wake_flags)
2900
{
2901
	struct sched_domain *tmp, *affine_sd = NULL, *sd = NULL;
2902 2903 2904
	int cpu = smp_processor_id();
	int prev_cpu = task_cpu(p);
	int new_cpu = cpu;
2905
	int want_affine = 0;
2906
	int sync = wake_flags & WF_SYNC;
2907

2908
	if (p->nr_cpus_allowed == 1)
2909 2910
		return prev_cpu;

2911
	if (sd_flag & SD_BALANCE_WAKE) {
2912
		if (cpumask_test_cpu(cpu, tsk_cpus_allowed(p)))
2913 2914 2915
			want_affine = 1;
		new_cpu = prev_cpu;
	}
2916

2917
	rcu_read_lock();
2918
	for_each_domain(cpu, tmp) {
2919 2920 2921
		if (!(tmp->flags & SD_LOAD_BALANCE))
			continue;

2922
		/*
2923 2924
		 * If both cpu and prev_cpu are part of this domain,
		 * cpu is a valid SD_WAKE_AFFINE target.
2925
		 */
2926 2927 2928
		if (want_affine && (tmp->flags & SD_WAKE_AFFINE) &&
		    cpumask_test_cpu(prev_cpu, sched_domain_span(tmp))) {
			affine_sd = tmp;
2929
			break;
2930
		}
2931

2932
		if (tmp->flags & sd_flag)
2933 2934 2935
			sd = tmp;
	}

2936
	if (affine_sd) {
2937
		if (cpu != prev_cpu && wake_affine(affine_sd, p, sync))
2938 2939 2940 2941
			prev_cpu = cpu;

		new_cpu = select_idle_sibling(p, prev_cpu);
		goto unlock;
2942
	}
2943

2944
	while (sd) {
2945
		int load_idx = sd->forkexec_idx;
2946
		struct sched_group *group;
2947
		int weight;
2948

2949
		if (!(sd->flags & sd_flag)) {
2950 2951 2952
			sd = sd->child;
			continue;
		}
2953

2954 2955
		if (sd_flag & SD_BALANCE_WAKE)
			load_idx = sd->wake_idx;
2956

2957
		group = find_idlest_group(sd, p, cpu, load_idx);
2958 2959 2960 2961
		if (!group) {
			sd = sd->child;
			continue;
		}
I
Ingo Molnar 已提交
2962

2963
		new_cpu = find_idlest_cpu(group, p, cpu);
2964 2965 2966 2967
		if (new_cpu == -1 || new_cpu == cpu) {
			/* Now try balancing at a lower domain level of cpu */
			sd = sd->child;
			continue;
2968
		}
2969 2970 2971

		/* Now try balancing at a lower domain level of new_cpu */
		cpu = new_cpu;
2972
		weight = sd->span_weight;
2973 2974
		sd = NULL;
		for_each_domain(cpu, tmp) {
2975
			if (weight <= tmp->span_weight)
2976
				break;
2977
			if (tmp->flags & sd_flag)
2978 2979 2980
				sd = tmp;
		}
		/* while loop will break here if sd == NULL */
2981
	}
2982 2983
unlock:
	rcu_read_unlock();
2984

2985
	return new_cpu;
2986 2987 2988
}
#endif /* CONFIG_SMP */

P
Peter Zijlstra 已提交
2989 2990
static unsigned long
wakeup_gran(struct sched_entity *curr, struct sched_entity *se)
2991 2992 2993 2994
{
	unsigned long gran = sysctl_sched_wakeup_granularity;

	/*
P
Peter Zijlstra 已提交
2995 2996
	 * Since its curr running now, convert the gran from real-time
	 * to virtual-time in his units.
M
Mike Galbraith 已提交
2997 2998 2999 3000 3001 3002 3003 3004 3005
	 *
	 * 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.
3006
	 */
3007
	return calc_delta_fair(gran, se);
3008 3009
}

3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031
/*
 * 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
3032
	gran = wakeup_gran(curr, se);
3033 3034 3035 3036 3037 3038
	if (vdiff > gran)
		return 1;

	return 0;
}

3039 3040
static void set_last_buddy(struct sched_entity *se)
{
3041 3042 3043 3044 3045
	if (entity_is_task(se) && unlikely(task_of(se)->policy == SCHED_IDLE))
		return;

	for_each_sched_entity(se)
		cfs_rq_of(se)->last = se;
3046 3047 3048 3049
}

static void set_next_buddy(struct sched_entity *se)
{
3050 3051 3052 3053 3054
	if (entity_is_task(se) && unlikely(task_of(se)->policy == SCHED_IDLE))
		return;

	for_each_sched_entity(se)
		cfs_rq_of(se)->next = se;
3055 3056
}

3057 3058
static void set_skip_buddy(struct sched_entity *se)
{
3059 3060
	for_each_sched_entity(se)
		cfs_rq_of(se)->skip = se;
3061 3062
}

3063 3064 3065
/*
 * Preempt the current task with a newly woken task if needed:
 */
3066
static void check_preempt_wakeup(struct rq *rq, struct task_struct *p, int wake_flags)
3067 3068
{
	struct task_struct *curr = rq->curr;
3069
	struct sched_entity *se = &curr->se, *pse = &p->se;
3070
	struct cfs_rq *cfs_rq = task_cfs_rq(curr);
3071
	int scale = cfs_rq->nr_running >= sched_nr_latency;
3072
	int next_buddy_marked = 0;
3073

I
Ingo Molnar 已提交
3074 3075 3076
	if (unlikely(se == pse))
		return;

3077
	/*
3078
	 * This is possible from callers such as move_task(), in which we
3079 3080 3081 3082 3083 3084 3085
	 * 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;

3086
	if (sched_feat(NEXT_BUDDY) && scale && !(wake_flags & WF_FORK)) {
M
Mike Galbraith 已提交
3087
		set_next_buddy(pse);
3088 3089
		next_buddy_marked = 1;
	}
P
Peter Zijlstra 已提交
3090

3091 3092 3093
	/*
	 * We can come here with TIF_NEED_RESCHED already set from new task
	 * wake up path.
3094 3095 3096 3097 3098 3099
	 *
	 * 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.
3100 3101 3102 3103
	 */
	if (test_tsk_need_resched(curr))
		return;

3104 3105 3106 3107 3108
	/* Idle tasks are by definition preempted by non-idle tasks. */
	if (unlikely(curr->policy == SCHED_IDLE) &&
	    likely(p->policy != SCHED_IDLE))
		goto preempt;

3109
	/*
3110 3111
	 * Batch and idle tasks do not preempt non-idle tasks (their preemption
	 * is driven by the tick):
3112
	 */
3113
	if (unlikely(p->policy != SCHED_NORMAL))
3114
		return;
3115

3116
	find_matching_se(&se, &pse);
3117
	update_curr(cfs_rq_of(se));
3118
	BUG_ON(!pse);
3119 3120 3121 3122 3123 3124 3125
	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);
3126
		goto preempt;
3127
	}
3128

3129
	return;
3130

3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146
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);
3147 3148
}

3149
static struct task_struct *pick_next_task_fair(struct rq *rq)
3150
{
P
Peter Zijlstra 已提交
3151
	struct task_struct *p;
3152 3153 3154
	struct cfs_rq *cfs_rq = &rq->cfs;
	struct sched_entity *se;

3155
	if (!cfs_rq->nr_running)
3156 3157 3158
		return NULL;

	do {
3159
		se = pick_next_entity(cfs_rq);
3160
		set_next_entity(cfs_rq, se);
3161 3162 3163
		cfs_rq = group_cfs_rq(se);
	} while (cfs_rq);

P
Peter Zijlstra 已提交
3164
	p = task_of(se);
3165 3166
	if (hrtick_enabled(rq))
		hrtick_start_fair(rq, p);
P
Peter Zijlstra 已提交
3167 3168

	return p;
3169 3170 3171 3172 3173
}

/*
 * Account for a descheduled task:
 */
3174
static void put_prev_task_fair(struct rq *rq, struct task_struct *prev)
3175 3176 3177 3178 3179 3180
{
	struct sched_entity *se = &prev->se;
	struct cfs_rq *cfs_rq;

	for_each_sched_entity(se) {
		cfs_rq = cfs_rq_of(se);
3181
		put_prev_entity(cfs_rq, se);
3182 3183 3184
	}
}

3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209
/*
 * 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);
3210 3211 3212 3213 3214 3215
		/*
		 * 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;
3216 3217 3218 3219 3220
	}

	set_skip_buddy(se);
}

3221 3222 3223 3224
static bool yield_to_task_fair(struct rq *rq, struct task_struct *p, bool preempt)
{
	struct sched_entity *se = &p->se;

3225 3226
	/* throttled hierarchies are not runnable */
	if (!se->on_rq || throttled_hierarchy(cfs_rq_of(se)))
3227 3228 3229 3230 3231 3232 3233 3234 3235 3236
		return false;

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

	yield_task_fair(rq);

	return true;
}

3237
#ifdef CONFIG_SMP
3238 3239 3240 3241
/**************************************************
 * Fair scheduling class load-balancing methods:
 */

3242 3243
static unsigned long __read_mostly max_load_balance_interval = HZ/10;

3244
#define LBF_ALL_PINNED	0x01
3245
#define LBF_NEED_BREAK	0x02
3246
#define LBF_SOME_PINNED 0x04
3247 3248 3249 3250 3251

struct lb_env {
	struct sched_domain	*sd;

	struct rq		*src_rq;
3252
	int			src_cpu;
3253 3254 3255 3256

	int			dst_cpu;
	struct rq		*dst_rq;

3257 3258
	struct cpumask		*dst_grpmask;
	int			new_dst_cpu;
3259
	enum cpu_idle_type	idle;
3260
	long			imbalance;
3261 3262 3263
	/* The set of CPUs under consideration for load-balancing */
	struct cpumask		*cpus;

3264
	unsigned int		flags;
3265 3266 3267 3268

	unsigned int		loop;
	unsigned int		loop_break;
	unsigned int		loop_max;
3269 3270
};

3271
/*
3272
 * move_task - move a task from one runqueue to another runqueue.
3273 3274
 * Both runqueues must be locked.
 */
3275
static void move_task(struct task_struct *p, struct lb_env *env)
3276
{
3277 3278 3279 3280
	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);
3281 3282
}

3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314
/*
 * 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;
}

3315 3316 3317 3318
/*
 * can_migrate_task - may task p from runqueue rq be migrated to this_cpu?
 */
static
3319
int can_migrate_task(struct task_struct *p, struct lb_env *env)
3320 3321 3322 3323 3324 3325 3326 3327
{
	int tsk_cache_hot = 0;
	/*
	 * We do not migrate tasks that are:
	 * 1) running (obviously), or
	 * 2) cannot be migrated to this CPU due to cpus_allowed, or
	 * 3) are cache-hot on their current CPU.
	 */
3328
	if (!cpumask_test_cpu(env->dst_cpu, tsk_cpus_allowed(p))) {
3329 3330
		int new_dst_cpu;

3331
		schedstat_inc(p, se.statistics.nr_failed_migrations_affine);
3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349

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

		new_dst_cpu = cpumask_first_and(env->dst_grpmask,
						tsk_cpus_allowed(p));
		if (new_dst_cpu < nr_cpu_ids) {
			env->flags |= LBF_SOME_PINNED;
			env->new_dst_cpu = new_dst_cpu;
		}
3350 3351
		return 0;
	}
3352 3353

	/* Record that we found atleast one task that could run on dst_cpu */
3354
	env->flags &= ~LBF_ALL_PINNED;
3355

3356
	if (task_running(env->src_rq, p)) {
3357
		schedstat_inc(p, se.statistics.nr_failed_migrations_running);
3358 3359 3360 3361 3362 3363 3364 3365 3366
		return 0;
	}

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

3367
	tsk_cache_hot = task_hot(p, env->src_rq->clock_task, env->sd);
3368
	if (!tsk_cache_hot ||
3369
		env->sd->nr_balance_failed > env->sd->cache_nice_tries) {
3370 3371
#ifdef CONFIG_SCHEDSTATS
		if (tsk_cache_hot) {
3372
			schedstat_inc(env->sd, lb_hot_gained[env->idle]);
3373
			schedstat_inc(p, se.statistics.nr_forced_migrations);
3374 3375 3376 3377 3378 3379
		}
#endif
		return 1;
	}

	if (tsk_cache_hot) {
3380
		schedstat_inc(p, se.statistics.nr_failed_migrations_hot);
3381 3382 3383 3384 3385
		return 0;
	}
	return 1;
}

3386 3387 3388 3389 3390 3391 3392
/*
 * 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.
 */
3393
static int move_one_task(struct lb_env *env)
3394 3395 3396
{
	struct task_struct *p, *n;

3397 3398 3399
	list_for_each_entry_safe(p, n, &env->src_rq->cfs_tasks, se.group_node) {
		if (throttled_lb_pair(task_group(p), env->src_rq->cpu, env->dst_cpu))
			continue;
3400

3401 3402
		if (!can_migrate_task(p, env))
			continue;
3403

3404 3405 3406 3407 3408 3409 3410 3411
		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;
3412 3413 3414 3415
	}
	return 0;
}

3416 3417
static unsigned long task_h_load(struct task_struct *p);

3418 3419
static const unsigned int sched_nr_migrate_break = 32;

3420
/*
3421
 * move_tasks tries to move up to imbalance weighted load from busiest to
3422 3423 3424 3425 3426 3427
 * 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)
3428
{
3429 3430
	struct list_head *tasks = &env->src_rq->cfs_tasks;
	struct task_struct *p;
3431 3432
	unsigned long load;
	int pulled = 0;
3433

3434
	if (env->imbalance <= 0)
3435
		return 0;
3436

3437 3438
	while (!list_empty(tasks)) {
		p = list_first_entry(tasks, struct task_struct, se.group_node);
3439

3440 3441
		env->loop++;
		/* We've more or less seen every task there is, call it quits */
3442
		if (env->loop > env->loop_max)
3443
			break;
3444 3445

		/* take a breather every nr_migrate tasks */
3446
		if (env->loop > env->loop_break) {
3447
			env->loop_break += sched_nr_migrate_break;
3448
			env->flags |= LBF_NEED_BREAK;
3449
			break;
3450
		}
3451

3452
		if (throttled_lb_pair(task_group(p), env->src_cpu, env->dst_cpu))
3453 3454 3455
			goto next;

		load = task_h_load(p);
3456

3457
		if (sched_feat(LB_MIN) && load < 16 && !env->sd->nr_balance_failed)
3458 3459
			goto next;

3460
		if ((load / 2) > env->imbalance)
3461
			goto next;
3462

3463 3464
		if (!can_migrate_task(p, env))
			goto next;
3465

3466
		move_task(p, env);
3467
		pulled++;
3468
		env->imbalance -= load;
3469 3470

#ifdef CONFIG_PREEMPT
3471 3472 3473 3474 3475
		/*
		 * NEWIDLE balancing is a source of latency, so preemptible
		 * kernels will stop after the first task is pulled to minimize
		 * the critical section.
		 */
3476
		if (env->idle == CPU_NEWLY_IDLE)
3477
			break;
3478 3479
#endif

3480 3481 3482 3483
		/*
		 * We only want to steal up to the prescribed amount of
		 * weighted load.
		 */
3484
		if (env->imbalance <= 0)
3485
			break;
3486 3487 3488

		continue;
next:
3489
		list_move_tail(&p->se.group_node, tasks);
3490
	}
3491

3492
	/*
3493 3494 3495
	 * 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().
3496
	 */
3497
	schedstat_add(env->sd, lb_gained[env->idle], pulled);
3498

3499
	return pulled;
3500 3501
}

P
Peter Zijlstra 已提交
3502
#ifdef CONFIG_FAIR_GROUP_SCHED
3503 3504 3505
/*
 * update tg->load_weight by folding this cpu's load_avg
 */
3506
static int update_shares_cpu(struct task_group *tg, int cpu)
3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520
{
	struct cfs_rq *cfs_rq;
	unsigned long flags;
	struct rq *rq;

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

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

	raw_spin_lock_irqsave(&rq->lock, flags);

	update_rq_clock(rq);
3521
	update_cfs_load(cfs_rq, 1);
3522 3523 3524 3525 3526

	/*
	 * We need to update shares after updating tg->load_weight in
	 * order to adjust the weight of groups with long running tasks.
	 */
3527
	update_cfs_shares(cfs_rq);
3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539

	raw_spin_unlock_irqrestore(&rq->lock, flags);

	return 0;
}

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

	rcu_read_lock();
3540 3541 3542 3543
	/*
	 * Iterates the task_group tree in a bottom up fashion, see
	 * list_add_leaf_cfs_rq() for details.
	 */
3544 3545 3546 3547 3548
	for_each_leaf_cfs_rq(rq, cfs_rq) {
		/* throttled entities do not contribute to load */
		if (throttled_hierarchy(cfs_rq))
			continue;

3549
		update_shares_cpu(cfs_rq->tg, cpu);
3550
	}
3551 3552 3553
	rcu_read_unlock();
}

3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578
/*
 * 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)
{
3579 3580 3581 3582 3583 3584 3585 3586
	struct rq *rq = cpu_rq(cpu);
	unsigned long now = jiffies;

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

	rq->h_load_throttle = now;

3587
	rcu_read_lock();
3588
	walk_tg_tree(tg_load_down, tg_nop, (void *)cpu);
3589
	rcu_read_unlock();
3590 3591
}

3592
static unsigned long task_h_load(struct task_struct *p)
P
Peter Zijlstra 已提交
3593
{
3594 3595
	struct cfs_rq *cfs_rq = task_cfs_rq(p);
	unsigned long load;
P
Peter Zijlstra 已提交
3596

3597 3598
	load = p->se.load.weight;
	load = div_u64(load * cfs_rq->h_load, cfs_rq->load.weight + 1);
P
Peter Zijlstra 已提交
3599

3600
	return load;
P
Peter Zijlstra 已提交
3601 3602
}
#else
3603 3604 3605 3606
static inline void update_shares(int cpu)
{
}

3607
static inline void update_h_load(long cpu)
P
Peter Zijlstra 已提交
3608 3609 3610
{
}

3611
static unsigned long task_h_load(struct task_struct *p)
3612
{
3613
	return p->se.load.weight;
3614
}
P
Peter Zijlstra 已提交
3615
#endif
3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632

/********** 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;
3633
	unsigned long this_has_capacity;
3634
	unsigned int  this_idle_cpus;
3635 3636

	/* Statistics of the busiest group */
3637
	unsigned int  busiest_idle_cpus;
3638 3639 3640
	unsigned long max_load;
	unsigned long busiest_load_per_task;
	unsigned long busiest_nr_running;
3641
	unsigned long busiest_group_capacity;
3642
	unsigned long busiest_has_capacity;
3643
	unsigned int  busiest_group_weight;
3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656

	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;
3657 3658
	unsigned long idle_cpus;
	unsigned long group_weight;
3659
	int group_imb; /* Is there an imbalance in the group ? */
3660
	int group_has_capacity; /* Is there extra capacity in the group? */
3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690
};

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

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

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

	return load_idx;
}

unsigned long default_scale_freq_power(struct sched_domain *sd, int cpu)
{
3691
	return SCHED_POWER_SCALE;
3692 3693 3694 3695 3696 3697 3698 3699 3700
}

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

unsigned long default_scale_smt_power(struct sched_domain *sd, int cpu)
{
3701
	unsigned long weight = sd->span_weight;
3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716
	unsigned long smt_gain = sd->smt_gain;

	smt_gain /= weight;

	return smt_gain;
}

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

unsigned long scale_rt_power(int cpu)
{
	struct rq *rq = cpu_rq(cpu);
3717
	u64 total, available, age_stamp, avg;
3718

3719 3720 3721 3722 3723 3724 3725 3726
	/*
	 * 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);
3727

3728
	if (unlikely(total < avg)) {
3729 3730 3731
		/* Ensures that power won't end up being negative */
		available = 0;
	} else {
3732
		available = total - avg;
3733
	}
3734

3735 3736
	if (unlikely((s64)total < SCHED_POWER_SCALE))
		total = SCHED_POWER_SCALE;
3737

3738
	total >>= SCHED_POWER_SHIFT;
3739 3740 3741 3742 3743 3744

	return div_u64(available, total);
}

static void update_cpu_power(struct sched_domain *sd, int cpu)
{
3745
	unsigned long weight = sd->span_weight;
3746
	unsigned long power = SCHED_POWER_SCALE;
3747 3748 3749 3750 3751 3752 3753 3754
	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);

3755
		power >>= SCHED_POWER_SHIFT;
3756 3757
	}

3758
	sdg->sgp->power_orig = power;
3759 3760 3761 3762 3763 3764

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

3765
	power >>= SCHED_POWER_SHIFT;
3766

3767
	power *= scale_rt_power(cpu);
3768
	power >>= SCHED_POWER_SHIFT;
3769 3770 3771 3772

	if (!power)
		power = 1;

3773
	cpu_rq(cpu)->cpu_power = power;
3774
	sdg->sgp->power = power;
3775 3776
}

3777
void update_group_power(struct sched_domain *sd, int cpu)
3778 3779 3780 3781
{
	struct sched_domain *child = sd->child;
	struct sched_group *group, *sdg = sd->groups;
	unsigned long power;
3782 3783 3784 3785 3786
	unsigned long interval;

	interval = msecs_to_jiffies(sd->balance_interval);
	interval = clamp(interval, 1UL, max_load_balance_interval);
	sdg->sgp->next_update = jiffies + interval;
3787 3788 3789 3790 3791 3792 3793 3794

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

	power = 0;

P
Peter Zijlstra 已提交
3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814
	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);
	}
3815

3816
	sdg->sgp->power_orig = sdg->sgp->power = power;
3817 3818
}

3819 3820 3821 3822 3823 3824 3825 3826 3827 3828 3829
/*
 * 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)
{
	/*
3830
	 * Only siblings can have significantly less than SCHED_POWER_SCALE
3831
	 */
P
Peter Zijlstra 已提交
3832
	if (!(sd->flags & SD_SHARE_CPUPOWER))
3833 3834 3835 3836 3837
		return 0;

	/*
	 * If ~90% of the cpu_power is still there, we're good.
	 */
3838
	if (group->sgp->power * 32 > group->sgp->power_orig * 29)
3839 3840 3841 3842 3843
		return 1;

	return 0;
}

3844 3845
/**
 * update_sg_lb_stats - Update sched_group's statistics for load balancing.
3846
 * @env: The load balancing environment.
3847 3848 3849 3850 3851 3852
 * @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.
 */
3853 3854
static inline void update_sg_lb_stats(struct lb_env *env,
			struct sched_group *group, int load_idx,
3855
			int local_group, int *balance, struct sg_lb_stats *sgs)
3856
{
3857 3858
	unsigned long nr_running, max_nr_running, min_nr_running;
	unsigned long load, max_cpu_load, min_cpu_load;
3859
	unsigned int balance_cpu = -1, first_idle_cpu = 0;
3860
	unsigned long avg_load_per_task = 0;
3861
	int i;
3862

3863
	if (local_group)
P
Peter Zijlstra 已提交
3864
		balance_cpu = group_balance_cpu(group);
3865 3866 3867 3868

	/* Tally up the load of all CPUs in the group */
	max_cpu_load = 0;
	min_cpu_load = ~0UL;
3869
	max_nr_running = 0;
3870
	min_nr_running = ~0UL;
3871

3872
	for_each_cpu_and(i, sched_group_cpus(group), env->cpus) {
3873 3874
		struct rq *rq = cpu_rq(i);

3875 3876
		nr_running = rq->nr_running;

3877 3878
		/* Bias balancing toward cpus of our domain */
		if (local_group) {
P
Peter Zijlstra 已提交
3879 3880
			if (idle_cpu(i) && !first_idle_cpu &&
					cpumask_test_cpu(i, sched_group_mask(group))) {
3881
				first_idle_cpu = 1;
3882 3883
				balance_cpu = i;
			}
3884 3885

			load = target_load(i, load_idx);
3886 3887
		} else {
			load = source_load(i, load_idx);
3888
			if (load > max_cpu_load)
3889 3890 3891
				max_cpu_load = load;
			if (min_cpu_load > load)
				min_cpu_load = load;
3892 3893 3894 3895 3896

			if (nr_running > max_nr_running)
				max_nr_running = nr_running;
			if (min_nr_running > nr_running)
				min_nr_running = nr_running;
3897 3898 3899
		}

		sgs->group_load += load;
3900
		sgs->sum_nr_running += nr_running;
3901
		sgs->sum_weighted_load += weighted_cpuload(i);
3902 3903
		if (idle_cpu(i))
			sgs->idle_cpus++;
3904 3905 3906 3907 3908 3909 3910 3911
	}

	/*
	 * 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.
	 */
3912
	if (local_group) {
3913
		if (env->idle != CPU_NEWLY_IDLE) {
3914
			if (balance_cpu != env->dst_cpu) {
3915 3916 3917
				*balance = 0;
				return;
			}
3918
			update_group_power(env->sd, env->dst_cpu);
3919
		} else if (time_after_eq(jiffies, group->sgp->next_update))
3920
			update_group_power(env->sd, env->dst_cpu);
3921 3922 3923
	}

	/* Adjust by relative CPU power of the group */
3924
	sgs->avg_load = (sgs->group_load*SCHED_POWER_SCALE) / group->sgp->power;
3925 3926 3927

	/*
	 * Consider the group unbalanced when the imbalance is larger
P
Peter Zijlstra 已提交
3928
	 * than the average weight of a task.
3929 3930 3931 3932 3933 3934
	 *
	 * 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?
	 */
3935 3936
	if (sgs->sum_nr_running)
		avg_load_per_task = sgs->sum_weighted_load / sgs->sum_nr_running;
3937

3938 3939
	if ((max_cpu_load - min_cpu_load) >= avg_load_per_task &&
	    (max_nr_running - min_nr_running) > 1)
3940 3941
		sgs->group_imb = 1;

3942
	sgs->group_capacity = DIV_ROUND_CLOSEST(group->sgp->power,
3943
						SCHED_POWER_SCALE);
3944
	if (!sgs->group_capacity)
3945
		sgs->group_capacity = fix_small_capacity(env->sd, group);
3946
	sgs->group_weight = group->group_weight;
3947 3948 3949

	if (sgs->group_capacity > sgs->sum_nr_running)
		sgs->group_has_capacity = 1;
3950 3951
}

3952 3953
/**
 * update_sd_pick_busiest - return 1 on busiest group
3954
 * @env: The load balancing environment.
3955 3956
 * @sds: sched_domain statistics
 * @sg: sched_group candidate to be checked for being the busiest
3957
 * @sgs: sched_group statistics
3958 3959 3960 3961
 *
 * Determine if @sg is a busier group than the previously selected
 * busiest group.
 */
3962
static bool update_sd_pick_busiest(struct lb_env *env,
3963 3964
				   struct sd_lb_stats *sds,
				   struct sched_group *sg,
3965
				   struct sg_lb_stats *sgs)
3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980
{
	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.
	 */
3981 3982
	if ((env->sd->flags & SD_ASYM_PACKING) && sgs->sum_nr_running &&
	    env->dst_cpu < group_first_cpu(sg)) {
3983 3984 3985 3986 3987 3988 3989 3990 3991 3992
		if (!sds->busiest)
			return true;

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

	return false;
}

3993
/**
3994
 * update_sd_lb_stats - Update sched_domain's statistics for load balancing.
3995
 * @env: The load balancing environment.
3996 3997 3998
 * @balance: Should we balance.
 * @sds: variable to hold the statistics for this sched_domain.
 */
3999
static inline void update_sd_lb_stats(struct lb_env *env,
4000
					int *balance, struct sd_lb_stats *sds)
4001
{
4002 4003
	struct sched_domain *child = env->sd->child;
	struct sched_group *sg = env->sd->groups;
4004 4005 4006 4007 4008 4009
	struct sg_lb_stats sgs;
	int load_idx, prefer_sibling = 0;

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

4010
	load_idx = get_sd_load_idx(env->sd, env->idle);
4011 4012 4013 4014

	do {
		int local_group;

4015
		local_group = cpumask_test_cpu(env->dst_cpu, sched_group_cpus(sg));
4016
		memset(&sgs, 0, sizeof(sgs));
4017
		update_sg_lb_stats(env, sg, load_idx, local_group, balance, &sgs);
4018

P
Peter Zijlstra 已提交
4019
		if (local_group && !(*balance))
4020 4021 4022
			return;

		sds->total_load += sgs.group_load;
4023
		sds->total_pwr += sg->sgp->power;
4024 4025 4026

		/*
		 * In case the child domain prefers tasks go to siblings
4027
		 * first, lower the sg capacity to one so that we'll try
4028 4029 4030 4031 4032 4033
		 * 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).
4034
		 */
4035
		if (prefer_sibling && !local_group && sds->this_has_capacity)
4036 4037 4038 4039
			sgs.group_capacity = min(sgs.group_capacity, 1UL);

		if (local_group) {
			sds->this_load = sgs.avg_load;
4040
			sds->this = sg;
4041 4042
			sds->this_nr_running = sgs.sum_nr_running;
			sds->this_load_per_task = sgs.sum_weighted_load;
4043
			sds->this_has_capacity = sgs.group_has_capacity;
4044
			sds->this_idle_cpus = sgs.idle_cpus;
4045
		} else if (update_sd_pick_busiest(env, sds, sg, &sgs)) {
4046
			sds->max_load = sgs.avg_load;
4047
			sds->busiest = sg;
4048
			sds->busiest_nr_running = sgs.sum_nr_running;
4049
			sds->busiest_idle_cpus = sgs.idle_cpus;
4050
			sds->busiest_group_capacity = sgs.group_capacity;
4051
			sds->busiest_load_per_task = sgs.sum_weighted_load;
4052
			sds->busiest_has_capacity = sgs.group_has_capacity;
4053
			sds->busiest_group_weight = sgs.group_weight;
4054 4055 4056
			sds->group_imb = sgs.group_imb;
		}

4057
		sg = sg->next;
4058
	} while (sg != env->sd->groups);
4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077
}

/**
 * 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.
 *
4078 4079 4080
 * Returns 1 when packing is required and a task should be moved to
 * this CPU.  The amount of the imbalance is returned in *imbalance.
 *
4081
 * @env: The load balancing environment.
4082 4083
 * @sds: Statistics of the sched_domain which is to be packed
 */
4084
static int check_asym_packing(struct lb_env *env, struct sd_lb_stats *sds)
4085 4086 4087
{
	int busiest_cpu;

4088
	if (!(env->sd->flags & SD_ASYM_PACKING))
4089 4090 4091 4092 4093 4094
		return 0;

	if (!sds->busiest)
		return 0;

	busiest_cpu = group_first_cpu(sds->busiest);
4095
	if (env->dst_cpu > busiest_cpu)
4096 4097
		return 0;

4098 4099 4100
	env->imbalance = DIV_ROUND_CLOSEST(
		sds->max_load * sds->busiest->sgp->power, SCHED_POWER_SCALE);

4101
	return 1;
4102 4103 4104 4105 4106 4107
}

/**
 * fix_small_imbalance - Calculate the minor imbalance that exists
 *			amongst the groups of a sched_domain, during
 *			load balancing.
4108
 * @env: The load balancing environment.
4109 4110
 * @sds: Statistics of the sched_domain whose imbalance is to be calculated.
 */
4111 4112
static inline
void fix_small_imbalance(struct lb_env *env, struct sd_lb_stats *sds)
4113 4114 4115
{
	unsigned long tmp, pwr_now = 0, pwr_move = 0;
	unsigned int imbn = 2;
4116
	unsigned long scaled_busy_load_per_task;
4117 4118 4119 4120 4121 4122

	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;
4123
	} else {
4124
		sds->this_load_per_task =
4125 4126
			cpu_avg_load_per_task(env->dst_cpu);
	}
4127

4128
	scaled_busy_load_per_task = sds->busiest_load_per_task
4129
					 * SCHED_POWER_SCALE;
4130
	scaled_busy_load_per_task /= sds->busiest->sgp->power;
4131 4132 4133

	if (sds->max_load - sds->this_load + scaled_busy_load_per_task >=
			(scaled_busy_load_per_task * imbn)) {
4134
		env->imbalance = sds->busiest_load_per_task;
4135 4136 4137 4138 4139 4140 4141 4142 4143
		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.
	 */

4144
	pwr_now += sds->busiest->sgp->power *
4145
			min(sds->busiest_load_per_task, sds->max_load);
4146
	pwr_now += sds->this->sgp->power *
4147
			min(sds->this_load_per_task, sds->this_load);
4148
	pwr_now /= SCHED_POWER_SCALE;
4149 4150

	/* Amount of load we'd subtract */
4151
	tmp = (sds->busiest_load_per_task * SCHED_POWER_SCALE) /
4152
		sds->busiest->sgp->power;
4153
	if (sds->max_load > tmp)
4154
		pwr_move += sds->busiest->sgp->power *
4155 4156 4157
			min(sds->busiest_load_per_task, sds->max_load - tmp);

	/* Amount of load we'd add */
4158
	if (sds->max_load * sds->busiest->sgp->power <
4159
		sds->busiest_load_per_task * SCHED_POWER_SCALE)
4160 4161
		tmp = (sds->max_load * sds->busiest->sgp->power) /
			sds->this->sgp->power;
4162
	else
4163
		tmp = (sds->busiest_load_per_task * SCHED_POWER_SCALE) /
4164 4165
			sds->this->sgp->power;
	pwr_move += sds->this->sgp->power *
4166
			min(sds->this_load_per_task, sds->this_load + tmp);
4167
	pwr_move /= SCHED_POWER_SCALE;
4168 4169 4170

	/* Move if we gain throughput */
	if (pwr_move > pwr_now)
4171
		env->imbalance = sds->busiest_load_per_task;
4172 4173 4174 4175 4176
}

/**
 * calculate_imbalance - Calculate the amount of imbalance present within the
 *			 groups of a given sched_domain during load balance.
4177
 * @env: load balance environment
4178 4179
 * @sds: statistics of the sched_domain whose imbalance is to be calculated.
 */
4180
static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *sds)
4181
{
4182 4183 4184 4185 4186 4187 4188 4189
	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);
	}

4190 4191 4192 4193 4194 4195
	/*
	 * 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) {
4196 4197
		env->imbalance = 0;
		return fix_small_imbalance(env, sds);
4198 4199
	}

4200 4201 4202 4203 4204 4205 4206
	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);

4207
		load_above_capacity *= (SCHED_LOAD_SCALE * SCHED_POWER_SCALE);
4208

4209
		load_above_capacity /= sds->busiest->sgp->power;
4210 4211 4212 4213 4214 4215 4216 4217 4218 4219 4220 4221 4222
	}

	/*
	 * 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);
4223 4224

	/* How much load to actually move to equalise the imbalance */
4225
	env->imbalance = min(max_pull * sds->busiest->sgp->power,
4226
		(sds->avg_load - sds->this_load) * sds->this->sgp->power)
4227
			/ SCHED_POWER_SCALE;
4228 4229 4230

	/*
	 * if *imbalance is less than the average load per runnable task
L
Lucas De Marchi 已提交
4231
	 * there is no guarantee that any tasks will be moved so we'll have
4232 4233 4234
	 * a think about bumping its value to force at least one task to be
	 * moved
	 */
4235 4236
	if (env->imbalance < sds->busiest_load_per_task)
		return fix_small_imbalance(env, sds);
4237 4238

}
4239

4240 4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251
/******* 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.
 *
4252
 * @env: The load balancing environment.
4253 4254 4255 4256 4257 4258 4259 4260 4261
 * @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 *
4262
find_busiest_group(struct lb_env *env, int *balance)
4263 4264 4265 4266 4267 4268 4269 4270 4271
{
	struct sd_lb_stats sds;

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

	/*
	 * Compute the various statistics relavent for load balancing at
	 * this level.
	 */
4272
	update_sd_lb_stats(env, balance, &sds);
4273

4274 4275 4276
	/*
	 * this_cpu is not the appropriate cpu to perform load balancing at
	 * this level.
4277
	 */
P
Peter Zijlstra 已提交
4278
	if (!(*balance))
4279 4280
		goto ret;

4281 4282
	if ((env->idle == CPU_IDLE || env->idle == CPU_NEWLY_IDLE) &&
	    check_asym_packing(env, &sds))
4283 4284
		return sds.busiest;

4285
	/* There is no busy sibling group to pull tasks from */
4286 4287 4288
	if (!sds.busiest || sds.busiest_nr_running == 0)
		goto out_balanced;

4289
	sds.avg_load = (SCHED_POWER_SCALE * sds.total_load) / sds.total_pwr;
4290

P
Peter Zijlstra 已提交
4291 4292 4293 4294 4295 4296 4297 4298
	/*
	 * 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;

4299
	/* SD_BALANCE_NEWIDLE trumps SMP nice when underutilized */
4300
	if (env->idle == CPU_NEWLY_IDLE && sds.this_has_capacity &&
4301 4302 4303
			!sds.busiest_has_capacity)
		goto force_balance;

4304 4305 4306 4307
	/*
	 * If the local group is more busy than the selected busiest group
	 * don't try and pull any tasks.
	 */
4308 4309 4310
	if (sds.this_load >= sds.max_load)
		goto out_balanced;

4311 4312 4313 4314
	/*
	 * Don't pull any tasks if this group is already above the domain
	 * average load.
	 */
4315 4316 4317
	if (sds.this_load >= sds.avg_load)
		goto out_balanced;

4318
	if (env->idle == CPU_IDLE) {
4319 4320 4321 4322 4323 4324
		/*
		 * 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.
		 */
4325
		if ((sds.this_idle_cpus <= sds.busiest_idle_cpus + 1) &&
4326 4327
		    sds.busiest_nr_running <= sds.busiest_group_weight)
			goto out_balanced;
4328 4329 4330 4331 4332
	} else {
		/*
		 * In the CPU_NEWLY_IDLE, CPU_NOT_IDLE cases, use
		 * imbalance_pct to be conservative.
		 */
4333
		if (100 * sds.max_load <= env->sd->imbalance_pct * sds.this_load)
4334
			goto out_balanced;
4335
	}
4336

4337
force_balance:
4338
	/* Looks like there is an imbalance. Compute it */
4339
	calculate_imbalance(env, &sds);
4340 4341 4342 4343
	return sds.busiest;

out_balanced:
ret:
4344
	env->imbalance = 0;
4345 4346 4347 4348 4349 4350
	return NULL;
}

/*
 * find_busiest_queue - find the busiest runqueue among the cpus in group.
 */
4351
static struct rq *find_busiest_queue(struct lb_env *env,
4352
				     struct sched_group *group)
4353 4354 4355 4356 4357 4358 4359
{
	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);
4360 4361
		unsigned long capacity = DIV_ROUND_CLOSEST(power,
							   SCHED_POWER_SCALE);
4362 4363
		unsigned long wl;

4364
		if (!capacity)
4365
			capacity = fix_small_capacity(env->sd, group);
4366

4367
		if (!cpumask_test_cpu(i, env->cpus))
4368 4369 4370
			continue;

		rq = cpu_rq(i);
4371
		wl = weighted_cpuload(i);
4372

4373 4374 4375 4376
		/*
		 * When comparing with imbalance, use weighted_cpuload()
		 * which is not scaled with the cpu power.
		 */
4377
		if (capacity && rq->nr_running == 1 && wl > env->imbalance)
4378 4379
			continue;

4380 4381 4382 4383 4384 4385
		/*
		 * 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.
		 */
4386
		wl = (wl * SCHED_POWER_SCALE) / power;
4387

4388 4389 4390 4391 4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403
		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. */
4404
DEFINE_PER_CPU(cpumask_var_t, load_balance_tmpmask);
4405

4406
static int need_active_balance(struct lb_env *env)
4407
{
4408 4409 4410
	struct sched_domain *sd = env->sd;

	if (env->idle == CPU_NEWLY_IDLE) {
4411 4412 4413 4414 4415 4416

		/*
		 * ASYM_PACKING needs to force migrate tasks from busy but
		 * higher numbered CPUs in order to pack all tasks in the
		 * lowest numbered CPUs.
		 */
4417
		if ((sd->flags & SD_ASYM_PACKING) && env->src_cpu > env->dst_cpu)
4418
			return 1;
4419 4420 4421 4422 4423
	}

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

4424 4425
static int active_load_balance_cpu_stop(void *data);

4426 4427 4428 4429 4430 4431 4432 4433
/*
 * 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)
{
4434 4435
	int ld_moved, cur_ld_moved, active_balance = 0;
	int lb_iterations, max_lb_iterations;
4436 4437 4438 4439 4440
	struct sched_group *group;
	struct rq *busiest;
	unsigned long flags;
	struct cpumask *cpus = __get_cpu_var(load_balance_tmpmask);

4441 4442
	struct lb_env env = {
		.sd		= sd,
4443 4444
		.dst_cpu	= this_cpu,
		.dst_rq		= this_rq,
4445
		.dst_grpmask    = sched_group_cpus(sd->groups),
4446
		.idle		= idle,
4447
		.loop_break	= sched_nr_migrate_break,
4448
		.cpus		= cpus,
4449 4450
	};

4451
	cpumask_copy(cpus, cpu_active_mask);
4452
	max_lb_iterations = cpumask_weight(env.dst_grpmask);
4453 4454 4455 4456

	schedstat_inc(sd, lb_count[idle]);

redo:
4457
	group = find_busiest_group(&env, balance);
4458 4459 4460 4461 4462 4463 4464 4465 4466

	if (*balance == 0)
		goto out_balanced;

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

4467
	busiest = find_busiest_queue(&env, group);
4468 4469 4470 4471 4472
	if (!busiest) {
		schedstat_inc(sd, lb_nobusyq[idle]);
		goto out_balanced;
	}

4473
	BUG_ON(busiest == env.dst_rq);
4474

4475
	schedstat_add(sd, lb_imbalance[idle], env.imbalance);
4476 4477

	ld_moved = 0;
4478
	lb_iterations = 1;
4479 4480 4481 4482 4483 4484 4485
	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.
		 */
4486
		env.flags |= LBF_ALL_PINNED;
4487 4488 4489
		env.src_cpu   = busiest->cpu;
		env.src_rq    = busiest;
		env.loop_max  = min(sysctl_sched_nr_migrate, busiest->nr_running);
4490

4491
		update_h_load(env.src_cpu);
4492
more_balance:
4493
		local_irq_save(flags);
4494
		double_rq_lock(env.dst_rq, busiest);
4495 4496 4497 4498 4499 4500 4501

		/*
		 * 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;
4502
		double_rq_unlock(env.dst_rq, busiest);
4503 4504
		local_irq_restore(flags);

4505 4506 4507 4508 4509
		if (env.flags & LBF_NEED_BREAK) {
			env.flags &= ~LBF_NEED_BREAK;
			goto more_balance;
		}

4510 4511 4512
		/*
		 * some other cpu did the load balance for us.
		 */
4513 4514 4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536 4537
		if (cur_ld_moved && env.dst_cpu != smp_processor_id())
			resched_cpu(env.dst_cpu);

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

4538
			env.dst_rq	 = cpu_rq(env.new_dst_cpu);
4539 4540 4541 4542 4543 4544 4545 4546 4547 4548
			env.dst_cpu	 = env.new_dst_cpu;
			env.flags	&= ~LBF_SOME_PINNED;
			env.loop	 = 0;
			env.loop_break	 = sched_nr_migrate_break;
			/*
			 * Go back to "more_balance" rather than "redo" since we
			 * need to continue with same src_cpu.
			 */
			goto more_balance;
		}
4549 4550

		/* All tasks on this runqueue were pinned by CPU affinity */
4551
		if (unlikely(env.flags & LBF_ALL_PINNED)) {
4552
			cpumask_clear_cpu(cpu_of(busiest), cpus);
4553 4554 4555
			if (!cpumask_empty(cpus)) {
				env.loop = 0;
				env.loop_break = sched_nr_migrate_break;
4556
				goto redo;
4557
			}
4558 4559 4560 4561 4562 4563
			goto out_balanced;
		}
	}

	if (!ld_moved) {
		schedstat_inc(sd, lb_failed[idle]);
4564 4565 4566 4567 4568 4569 4570 4571
		/*
		 * 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++;
4572

4573
		if (need_active_balance(&env)) {
4574 4575
			raw_spin_lock_irqsave(&busiest->lock, flags);

4576 4577 4578
			/* don't kick the active_load_balance_cpu_stop,
			 * if the curr task on busiest cpu can't be
			 * moved to this_cpu
4579 4580
			 */
			if (!cpumask_test_cpu(this_cpu,
4581
					tsk_cpus_allowed(busiest->curr))) {
4582 4583
				raw_spin_unlock_irqrestore(&busiest->lock,
							    flags);
4584
				env.flags |= LBF_ALL_PINNED;
4585 4586 4587
				goto out_one_pinned;
			}

4588 4589 4590 4591 4592
			/*
			 * ->active_balance synchronizes accesses to
			 * ->active_balance_work.  Once set, it's cleared
			 * only after active load balance is finished.
			 */
4593 4594 4595 4596 4597 4598
			if (!busiest->active_balance) {
				busiest->active_balance = 1;
				busiest->push_cpu = this_cpu;
				active_balance = 1;
			}
			raw_spin_unlock_irqrestore(&busiest->lock, flags);
4599

4600
			if (active_balance) {
4601 4602 4603
				stop_one_cpu_nowait(cpu_of(busiest),
					active_load_balance_cpu_stop, busiest,
					&busiest->active_balance_work);
4604
			}
4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637

			/*
			 * 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 */
4638
	if (((env.flags & LBF_ALL_PINNED) &&
4639
			sd->balance_interval < MAX_PINNED_INTERVAL) ||
4640 4641 4642
			(sd->balance_interval < sd->max_interval))
		sd->balance_interval *= 2;

4643
	ld_moved = 0;
4644 4645 4646 4647 4648 4649 4650 4651
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.
 */
4652
void idle_balance(int this_cpu, struct rq *this_rq)
4653 4654 4655 4656 4657 4658 4659 4660 4661 4662
{
	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;

4663 4664 4665 4666 4667
	/*
	 * Drop the rq->lock, but keep IRQ/preempt disabled.
	 */
	raw_spin_unlock(&this_rq->lock);

P
Paul Turner 已提交
4668
	update_shares(this_cpu);
4669
	rcu_read_lock();
4670 4671
	for_each_domain(this_cpu, sd) {
		unsigned long interval;
4672
		int balance = 1;
4673 4674 4675 4676

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

4677
		if (sd->flags & SD_BALANCE_NEWIDLE) {
4678
			/* If we've pulled tasks over stop searching: */
4679 4680 4681
			pulled_task = load_balance(this_cpu, this_rq,
						   sd, CPU_NEWLY_IDLE, &balance);
		}
4682 4683 4684 4685

		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
4686 4687
		if (pulled_task) {
			this_rq->idle_stamp = 0;
4688
			break;
N
Nikhil Rao 已提交
4689
		}
4690
	}
4691
	rcu_read_unlock();
4692 4693 4694

	raw_spin_lock(&this_rq->lock);

4695 4696 4697 4698 4699 4700 4701 4702 4703 4704
	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;
	}
}

/*
4705 4706 4707 4708
 * 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.
4709
 */
4710
static int active_load_balance_cpu_stop(void *data)
4711
{
4712 4713
	struct rq *busiest_rq = data;
	int busiest_cpu = cpu_of(busiest_rq);
4714
	int target_cpu = busiest_rq->push_cpu;
4715
	struct rq *target_rq = cpu_rq(target_cpu);
4716
	struct sched_domain *sd;
4717 4718 4719 4720 4721 4722 4723

	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;
4724 4725 4726

	/* Is there any task to move? */
	if (busiest_rq->nr_running <= 1)
4727
		goto out_unlock;
4728 4729 4730 4731 4732 4733 4734 4735 4736 4737 4738 4739

	/*
	 * 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. */
4740
	rcu_read_lock();
4741 4742 4743 4744 4745 4746 4747
	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)) {
4748 4749
		struct lb_env env = {
			.sd		= sd,
4750 4751 4752 4753
			.dst_cpu	= target_cpu,
			.dst_rq		= target_rq,
			.src_cpu	= busiest_rq->cpu,
			.src_rq		= busiest_rq,
4754 4755 4756
			.idle		= CPU_IDLE,
		};

4757 4758
		schedstat_inc(sd, alb_count);

4759
		if (move_one_task(&env))
4760 4761 4762 4763
			schedstat_inc(sd, alb_pushed);
		else
			schedstat_inc(sd, alb_failed);
	}
4764
	rcu_read_unlock();
4765
	double_unlock_balance(busiest_rq, target_rq);
4766 4767 4768 4769
out_unlock:
	busiest_rq->active_balance = 0;
	raw_spin_unlock_irq(&busiest_rq->lock);
	return 0;
4770 4771 4772
}

#ifdef CONFIG_NO_HZ
4773 4774 4775 4776 4777 4778
/*
 * 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.
 */
4779
static struct {
4780
	cpumask_var_t idle_cpus_mask;
4781
	atomic_t nr_cpus;
4782 4783
	unsigned long next_balance;     /* in jiffy units */
} nohz ____cacheline_aligned;
4784

4785
static inline int find_new_ilb(int call_cpu)
4786
{
4787
	int ilb = cpumask_first(nohz.idle_cpus_mask);
4788

4789 4790 4791 4792
	if (ilb < nr_cpu_ids && idle_cpu(ilb))
		return ilb;

	return nr_cpu_ids;
4793 4794
}

4795 4796 4797 4798 4799 4800 4801 4802 4803 4804 4805
/*
 * 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++;

4806
	ilb_cpu = find_new_ilb(cpu);
4807

4808 4809
	if (ilb_cpu >= nr_cpu_ids)
		return;
4810

4811
	if (test_and_set_bit(NOHZ_BALANCE_KICK, nohz_flags(ilb_cpu)))
4812 4813 4814 4815 4816 4817 4818 4819
		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);
4820 4821 4822
	return;
}

4823
static inline void nohz_balance_exit_idle(int cpu)
4824 4825 4826 4827 4828 4829 4830 4831
{
	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));
	}
}

4832 4833 4834 4835 4836 4837 4838 4839 4840 4841 4842 4843 4844 4845 4846 4847 4848 4849 4850 4851 4852 4853 4854 4855 4856 4857 4858 4859 4860 4861
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();
}

4862
/*
4863
 * This routine will record that the cpu is going idle with tick stopped.
4864
 * This info will be used in performing idle load balancing in the future.
4865
 */
4866
void nohz_balance_enter_idle(int cpu)
4867
{
4868 4869 4870 4871 4872 4873
	/*
	 * If this cpu is going down, then nothing needs to be done.
	 */
	if (!cpu_active(cpu))
		return;

4874 4875
	if (test_bit(NOHZ_TICK_STOPPED, nohz_flags(cpu)))
		return;
4876

4877 4878 4879
	cpumask_set_cpu(cpu, nohz.idle_cpus_mask);
	atomic_inc(&nohz.nr_cpus);
	set_bit(NOHZ_TICK_STOPPED, nohz_flags(cpu));
4880
}
4881 4882 4883 4884 4885 4886

static int __cpuinit sched_ilb_notifier(struct notifier_block *nfb,
					unsigned long action, void *hcpu)
{
	switch (action & ~CPU_TASKS_FROZEN) {
	case CPU_DYING:
4887
		nohz_balance_exit_idle(smp_processor_id());
4888 4889 4890 4891 4892
		return NOTIFY_OK;
	default:
		return NOTIFY_DONE;
	}
}
4893 4894 4895 4896
#endif

static DEFINE_SPINLOCK(balancing);

4897 4898 4899 4900
/*
 * 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.
 */
4901
void update_max_interval(void)
4902 4903 4904 4905
{
	max_load_balance_interval = HZ*num_online_cpus()/10;
}

4906 4907 4908 4909 4910 4911 4912 4913 4914 4915 4916
/*
 * It checks each scheduling domain to see if it is due to be balanced,
 * and initiates a balancing operation if so.
 *
 * Balancing parameters are set up in arch_init_sched_domains.
 */
static void rebalance_domains(int cpu, enum cpu_idle_type idle)
{
	int balance = 1;
	struct rq *rq = cpu_rq(cpu);
	unsigned long interval;
4917
	struct sched_domain *sd;
4918 4919 4920 4921 4922
	/* Earliest time when we have to do rebalance again */
	unsigned long next_balance = jiffies + 60*HZ;
	int update_next_balance = 0;
	int need_serialize;

P
Peter Zijlstra 已提交
4923 4924
	update_shares(cpu);

4925
	rcu_read_lock();
4926 4927 4928 4929 4930 4931 4932 4933 4934 4935
	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);
4936
		interval = clamp(interval, 1UL, max_load_balance_interval);
4937 4938 4939 4940 4941 4942 4943 4944 4945 4946 4947 4948

		need_serialize = sd->flags & SD_SERIALIZE;

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

		if (time_after_eq(jiffies, sd->last_balance + interval)) {
			if (load_balance(cpu, rq, sd, idle, &balance)) {
				/*
				 * We've pulled tasks over so either we're no
4949
				 * longer idle.
4950 4951 4952 4953 4954 4955 4956 4957 4958 4959 4960 4961 4962 4963 4964 4965 4966 4967 4968 4969 4970
				 */
				idle = CPU_NOT_IDLE;
			}
			sd->last_balance = jiffies;
		}
		if (need_serialize)
			spin_unlock(&balancing);
out:
		if (time_after(next_balance, sd->last_balance + interval)) {
			next_balance = sd->last_balance + interval;
			update_next_balance = 1;
		}

		/*
		 * Stop the load balance at this level. There is another
		 * CPU in our sched group which is doing load balancing more
		 * actively.
		 */
		if (!balance)
			break;
	}
4971
	rcu_read_unlock();
4972 4973 4974 4975 4976 4977 4978 4979 4980 4981

	/*
	 * 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;
}

4982
#ifdef CONFIG_NO_HZ
4983
/*
4984
 * In CONFIG_NO_HZ case, the idle balance kickee will do the
4985 4986
 * rebalancing for all the cpus for whom scheduler ticks are stopped.
 */
4987 4988 4989 4990 4991 4992
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;

4993 4994 4995
	if (idle != CPU_IDLE ||
	    !test_bit(NOHZ_BALANCE_KICK, nohz_flags(this_cpu)))
		goto end;
4996 4997

	for_each_cpu(balance_cpu, nohz.idle_cpus_mask) {
4998
		if (balance_cpu == this_cpu || !idle_cpu(balance_cpu))
4999 5000 5001 5002 5003 5004 5005
			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.
		 */
5006
		if (need_resched())
5007 5008
			break;

V
Vincent Guittot 已提交
5009 5010 5011 5012 5013 5014
		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);
5015 5016 5017 5018 5019 5020 5021

		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;
5022 5023
end:
	clear_bit(NOHZ_BALANCE_KICK, nohz_flags(this_cpu));
5024 5025 5026
}

/*
5027 5028 5029 5030 5031 5032 5033
 * 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.
5034 5035 5036 5037
 */
static inline int nohz_kick_needed(struct rq *rq, int cpu)
{
	unsigned long now = jiffies;
5038
	struct sched_domain *sd;
5039

5040
	if (unlikely(idle_cpu(cpu)))
5041 5042
		return 0;

5043 5044 5045 5046
       /*
	* 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.
	*/
5047
	set_cpu_sd_state_busy();
5048
	nohz_balance_exit_idle(cpu);
5049 5050 5051 5052 5053 5054 5055

	/*
	 * None are in tickless mode and hence no need for NOHZ idle load
	 * balancing.
	 */
	if (likely(!atomic_read(&nohz.nr_cpus)))
		return 0;
5056 5057

	if (time_before(now, nohz.next_balance))
5058 5059
		return 0;

5060 5061
	if (rq->nr_running >= 2)
		goto need_kick;
5062

5063
	rcu_read_lock();
5064 5065 5066 5067
	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);
5068

5069
		if (sd->flags & SD_SHARE_PKG_RESOURCES && nr_busy > 1)
5070
			goto need_kick_unlock;
5071 5072 5073 5074

		if (sd->flags & SD_ASYM_PACKING && nr_busy != sg->group_weight
		    && (cpumask_first_and(nohz.idle_cpus_mask,
					  sched_domain_span(sd)) < cpu))
5075
			goto need_kick_unlock;
5076 5077 5078

		if (!(sd->flags & (SD_SHARE_PKG_RESOURCES | SD_ASYM_PACKING)))
			break;
5079
	}
5080
	rcu_read_unlock();
5081
	return 0;
5082 5083 5084

need_kick_unlock:
	rcu_read_unlock();
5085 5086
need_kick:
	return 1;
5087 5088 5089 5090 5091 5092 5093 5094 5095
}
#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).
 */
5096 5097 5098 5099
static void run_rebalance_domains(struct softirq_action *h)
{
	int this_cpu = smp_processor_id();
	struct rq *this_rq = cpu_rq(this_cpu);
5100
	enum cpu_idle_type idle = this_rq->idle_balance ?
5101 5102 5103 5104 5105
						CPU_IDLE : CPU_NOT_IDLE;

	rebalance_domains(this_cpu, idle);

	/*
5106
	 * If this cpu has a pending nohz_balance_kick, then do the
5107 5108 5109
	 * balancing on behalf of the other idle cpus whose ticks are
	 * stopped.
	 */
5110
	nohz_idle_balance(this_cpu, idle);
5111 5112 5113 5114
}

static inline int on_null_domain(int cpu)
{
5115
	return !rcu_dereference_sched(cpu_rq(cpu)->sd);
5116 5117 5118 5119 5120
}

/*
 * Trigger the SCHED_SOFTIRQ if it is time to do periodic load balancing.
 */
5121
void trigger_load_balance(struct rq *rq, int cpu)
5122 5123 5124 5125 5126
{
	/* 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);
5127
#ifdef CONFIG_NO_HZ
5128
	if (nohz_kick_needed(rq, cpu) && likely(!on_null_domain(cpu)))
5129 5130
		nohz_balancer_kick(cpu);
#endif
5131 5132
}

5133 5134 5135 5136 5137 5138 5139 5140
static void rq_online_fair(struct rq *rq)
{
	update_sysctl();
}

static void rq_offline_fair(struct rq *rq)
{
	update_sysctl();
5141 5142 5143

	/* Ensure any throttled groups are reachable by pick_next_task */
	unthrottle_offline_cfs_rqs(rq);
5144 5145
}

5146
#endif /* CONFIG_SMP */
5147

5148 5149 5150
/*
 * scheduler tick hitting a task of our scheduling class:
 */
P
Peter Zijlstra 已提交
5151
static void task_tick_fair(struct rq *rq, struct task_struct *curr, int queued)
5152 5153 5154 5155 5156 5157
{
	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
5158
		entity_tick(cfs_rq, se, queued);
5159
	}
5160 5161 5162

	if (sched_feat_numa(NUMA))
		task_tick_numa(rq, curr);
5163 5164 5165
}

/*
P
Peter Zijlstra 已提交
5166 5167 5168
 * called on fork with the child task as argument from the parent's context
 *  - child not yet on the tasklist
 *  - preemption disabled
5169
 */
P
Peter Zijlstra 已提交
5170
static void task_fork_fair(struct task_struct *p)
5171
{
5172 5173
	struct cfs_rq *cfs_rq;
	struct sched_entity *se = &p->se, *curr;
5174
	int this_cpu = smp_processor_id();
P
Peter Zijlstra 已提交
5175 5176 5177
	struct rq *rq = this_rq();
	unsigned long flags;

5178
	raw_spin_lock_irqsave(&rq->lock, flags);
5179

5180 5181
	update_rq_clock(rq);

5182 5183 5184
	cfs_rq = task_cfs_rq(current);
	curr = cfs_rq->curr;

5185 5186
	if (unlikely(task_cpu(p) != this_cpu)) {
		rcu_read_lock();
P
Peter Zijlstra 已提交
5187
		__set_task_cpu(p, this_cpu);
5188 5189
		rcu_read_unlock();
	}
5190

5191
	update_curr(cfs_rq);
P
Peter Zijlstra 已提交
5192

5193 5194
	if (curr)
		se->vruntime = curr->vruntime;
5195
	place_entity(cfs_rq, se, 1);
5196

P
Peter Zijlstra 已提交
5197
	if (sysctl_sched_child_runs_first && curr && entity_before(curr, se)) {
D
Dmitry Adamushko 已提交
5198
		/*
5199 5200 5201
		 * Upon rescheduling, sched_class::put_prev_task() will place
		 * 'current' within the tree based on its new key value.
		 */
5202
		swap(curr->vruntime, se->vruntime);
5203
		resched_task(rq->curr);
5204
	}
5205

5206 5207
	se->vruntime -= cfs_rq->min_vruntime;

5208
	raw_spin_unlock_irqrestore(&rq->lock, flags);
5209 5210
}

5211 5212 5213 5214
/*
 * Priority of the task has changed. Check to see if we preempt
 * the current task.
 */
P
Peter Zijlstra 已提交
5215 5216
static void
prio_changed_fair(struct rq *rq, struct task_struct *p, int oldprio)
5217
{
P
Peter Zijlstra 已提交
5218 5219 5220
	if (!p->se.on_rq)
		return;

5221 5222 5223 5224 5225
	/*
	 * 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
5226
	if (rq->curr == p) {
5227 5228 5229
		if (p->prio > oldprio)
			resched_task(rq->curr);
	} else
5230
		check_preempt_curr(rq, p, 0);
5231 5232
}

P
Peter Zijlstra 已提交
5233 5234 5235 5236 5237 5238 5239 5240 5241 5242 5243 5244 5245 5246 5247 5248 5249 5250 5251 5252 5253 5254 5255 5256
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;
	}
}

5257 5258 5259
/*
 * We switched to the sched_fair class.
 */
P
Peter Zijlstra 已提交
5260
static void switched_to_fair(struct rq *rq, struct task_struct *p)
5261
{
P
Peter Zijlstra 已提交
5262 5263 5264
	if (!p->se.on_rq)
		return;

5265 5266 5267 5268 5269
	/*
	 * 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
5270
	if (rq->curr == p)
5271 5272
		resched_task(rq->curr);
	else
5273
		check_preempt_curr(rq, p, 0);
5274 5275
}

5276 5277 5278 5279 5280 5281 5282 5283 5284
/* 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;

5285 5286 5287 5288 5289 5290 5291
	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);
	}
5292 5293
}

5294 5295 5296 5297 5298 5299 5300 5301 5302
void init_cfs_rq(struct cfs_rq *cfs_rq)
{
	cfs_rq->tasks_timeline = RB_ROOT;
	cfs_rq->min_vruntime = (u64)(-(1LL << 20));
#ifndef CONFIG_64BIT
	cfs_rq->min_vruntime_copy = cfs_rq->min_vruntime;
#endif
}

P
Peter Zijlstra 已提交
5303
#ifdef CONFIG_FAIR_GROUP_SCHED
5304
static void task_move_group_fair(struct task_struct *p, int on_rq)
P
Peter Zijlstra 已提交
5305
{
5306 5307 5308 5309 5310 5311 5312 5313 5314 5315 5316 5317 5318
	/*
	 * 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.
	 */
5319 5320 5321 5322 5323 5324
	/*
	 * 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().
5325 5326
	 * - Moving a task which has been woken up by try_to_wake_up() and
	 *   waiting for actually being woken up by sched_ttwu_pending().
5327 5328 5329 5330
	 *
	 * To prevent boost or penalty in the new cfs_rq caused by delta
	 * min_vruntime between the two cfs_rqs, we skip vruntime adjustment.
	 */
5331
	if (!on_rq && (!p->se.sum_exec_runtime || p->state == TASK_WAKING))
5332 5333
		on_rq = 1;

5334 5335 5336
	if (!on_rq)
		p->se.vruntime -= cfs_rq_of(&p->se)->min_vruntime;
	set_task_rq(p, task_cpu(p));
5337
	if (!on_rq)
5338
		p->se.vruntime += cfs_rq_of(&p->se)->min_vruntime;
P
Peter Zijlstra 已提交
5339
}
5340 5341 5342 5343 5344 5345 5346 5347 5348 5349 5350 5351 5352 5353 5354 5355 5356 5357 5358 5359 5360 5361 5362 5363 5364 5365 5366 5367 5368 5369 5370 5371 5372 5373 5374 5375 5376 5377 5378 5379 5380 5381 5382 5383 5384 5385 5386 5387 5388 5389 5390 5391 5392 5393 5394 5395 5396 5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410 5411 5412 5413 5414 5415 5416 5417 5418 5419 5420 5421 5422 5423 5424 5425

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

	destroy_cfs_bandwidth(tg_cfs_bandwidth(tg));

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

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

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

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

	tg->shares = NICE_0_LOAD;

	init_cfs_bandwidth(tg_cfs_bandwidth(tg));

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

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

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

	return 1;

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

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

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

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

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

	cfs_rq->tg = tg;
	cfs_rq->rq = rq;
#ifdef CONFIG_SMP
	/* allow initial update_cfs_load() to truncate */
	cfs_rq->load_stamp = 1;
P
Peter Zijlstra 已提交
5426
#endif
5427 5428 5429 5430 5431 5432 5433 5434 5435 5436 5437 5438 5439 5440 5441 5442 5443 5444 5445 5446 5447 5448 5449 5450 5451 5452 5453 5454 5455 5456 5457 5458 5459 5460 5461 5462 5463 5464 5465 5466 5467 5468 5469 5470 5471 5472 5473 5474 5475 5476 5477 5478 5479 5480 5481 5482 5483 5484 5485 5486 5487 5488 5489 5490 5491 5492 5493 5494
	init_cfs_rq_runtime(cfs_rq);

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

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

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

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

static DEFINE_MUTEX(shares_mutex);

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

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

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

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

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

		se = tg->se[i];
		/* Propagate contribution to hierarchy */
		raw_spin_lock_irqsave(&rq->lock, flags);
		for_each_sched_entity(se)
			update_cfs_shares(group_cfs_rq(se));
		raw_spin_unlock_irqrestore(&rq->lock, flags);
	}

done:
	mutex_unlock(&shares_mutex);
	return 0;
}
#else /* CONFIG_FAIR_GROUP_SCHED */

void free_fair_sched_group(struct task_group *tg) { }

int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent)
{
	return 1;
}

void unregister_fair_sched_group(struct task_group *tg, int cpu) { }

#endif /* CONFIG_FAIR_GROUP_SCHED */

P
Peter Zijlstra 已提交
5495

5496
static unsigned int get_rr_interval_fair(struct rq *rq, struct task_struct *task)
5497 5498 5499 5500 5501 5502 5503 5504 5505 5506 5507 5508 5509 5510
{
	struct sched_entity *se = &task->se;
	unsigned int rr_interval = 0;

	/*
	 * Time slice is 0 for SCHED_OTHER tasks that are on an otherwise
	 * idle runqueue:
	 */
	if (rq->cfs.load.weight)
		rr_interval = NS_TO_JIFFIES(sched_slice(&rq->cfs, se));

	return rr_interval;
}

5511 5512 5513
/*
 * All the scheduling class methods:
 */
5514
const struct sched_class fair_sched_class = {
5515
	.next			= &idle_sched_class,
5516 5517 5518
	.enqueue_task		= enqueue_task_fair,
	.dequeue_task		= dequeue_task_fair,
	.yield_task		= yield_task_fair,
5519
	.yield_to_task		= yield_to_task_fair,
5520

I
Ingo Molnar 已提交
5521
	.check_preempt_curr	= check_preempt_wakeup,
5522 5523 5524 5525

	.pick_next_task		= pick_next_task_fair,
	.put_prev_task		= put_prev_task_fair,

5526
#ifdef CONFIG_SMP
L
Li Zefan 已提交
5527 5528
	.select_task_rq		= select_task_rq_fair,

5529 5530
	.rq_online		= rq_online_fair,
	.rq_offline		= rq_offline_fair,
5531 5532

	.task_waking		= task_waking_fair,
5533
#endif
5534

5535
	.set_curr_task          = set_curr_task_fair,
5536
	.task_tick		= task_tick_fair,
P
Peter Zijlstra 已提交
5537
	.task_fork		= task_fork_fair,
5538 5539

	.prio_changed		= prio_changed_fair,
P
Peter Zijlstra 已提交
5540
	.switched_from		= switched_from_fair,
5541
	.switched_to		= switched_to_fair,
P
Peter Zijlstra 已提交
5542

5543 5544
	.get_rr_interval	= get_rr_interval_fair,

P
Peter Zijlstra 已提交
5545
#ifdef CONFIG_FAIR_GROUP_SCHED
5546
	.task_move_group	= task_move_group_fair,
P
Peter Zijlstra 已提交
5547
#endif
5548 5549 5550
};

#ifdef CONFIG_SCHED_DEBUG
5551
void print_cfs_stats(struct seq_file *m, int cpu)
5552 5553 5554
{
	struct cfs_rq *cfs_rq;

5555
	rcu_read_lock();
5556
	for_each_leaf_cfs_rq(cpu_rq(cpu), cfs_rq)
5557
		print_cfs_rq(m, cpu, cfs_rq);
5558
	rcu_read_unlock();
5559 5560
}
#endif
5561 5562 5563 5564 5565 5566 5567

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

#ifdef CONFIG_NO_HZ
5568
	nohz.next_balance = jiffies;
5569
	zalloc_cpumask_var(&nohz.idle_cpus_mask, GFP_NOWAIT);
5570
	cpu_notifier(sched_ilb_notifier, 0);
5571 5572 5573 5574
#endif
#endif /* SMP */

}