sched_fair.c 25.6 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 23
 */

/*
24 25
 * Targeted preemption latency for CPU-bound tasks:
 * (default: 20ms, units: nanoseconds)
26
 *
27 28 29 30
 * NOTE: this latency value is not the same as the concept of
 * 'timeslice length' - timeslices in CFS are of variable length.
 * (to see the precise effective timeslice length of your workload,
 *  run vmstat and monitor the context-switches field)
31 32 33 34
 *
 * On SMP systems the value of this is multiplied by the log2 of the
 * number of CPUs. (i.e. factor 2x on 2-way systems, 3x on 4-way
 * systems, 4x on 8-way systems, 5x on 16-way systems, etc.)
35
 * Targeted preemption latency for CPU-bound tasks:
36
 */
37 38 39 40 41 42 43
const_debug unsigned int sysctl_sched_latency = 20000000ULL;

/*
 * After fork, child runs first. (default) If set to 0 then
 * parent will (try to) run first.
 */
const_debug unsigned int sysctl_sched_child_runs_first = 1;
44 45 46 47 48

/*
 * Minimal preemption granularity for CPU-bound tasks:
 * (default: 2 msec, units: nanoseconds)
 */
49
const_debug unsigned int sysctl_sched_nr_latency = 20;
50

51 52 53 54 55 56 57 58
/*
 * sys_sched_yield() compat mode
 *
 * This option switches the agressive yield implementation of the
 * old scheduler back on.
 */
unsigned int __read_mostly sysctl_sched_compat_yield;

59 60
/*
 * SCHED_BATCH wake-up granularity.
I
Ingo Molnar 已提交
61
 * (default: 10 msec, units: nanoseconds)
62 63 64 65 66
 *
 * This option delays the preemption effects of decoupled workloads
 * and reduces their over-scheduling. Synchronous workloads will still
 * have immediate wakeup/sleep latencies.
 */
I
Ingo Molnar 已提交
67
const_debug unsigned int sysctl_sched_batch_wakeup_granularity = 10000000UL;
68 69 70

/*
 * SCHED_OTHER wake-up granularity.
I
Ingo Molnar 已提交
71
 * (default: 10 msec, units: nanoseconds)
72 73 74 75 76
 *
 * This option delays the preemption effects of decoupled workloads
 * and reduces their over-scheduling. Synchronous workloads will still
 * have immediate wakeup/sleep latencies.
 */
I
Ingo Molnar 已提交
77
const_debug unsigned int sysctl_sched_wakeup_granularity = 10000000UL;
78 79 80 81 82

/**************************************************************
 * CFS operations on generic schedulable entities:
 */

83
#ifdef CONFIG_FAIR_GROUP_SCHED
84

85
/* cpu runqueue to which this cfs_rq is attached */
86 87
static inline struct rq *rq_of(struct cfs_rq *cfs_rq)
{
88
	return cfs_rq->rq;
89 90
}

91 92
/* An entity is a task if it doesn't "own" a runqueue */
#define entity_is_task(se)	(!se->my_q)
93

94
#else	/* CONFIG_FAIR_GROUP_SCHED */
95

96 97 98
static inline struct rq *rq_of(struct cfs_rq *cfs_rq)
{
	return container_of(cfs_rq, struct rq, cfs);
99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114
}

#define entity_is_task(se)	1

#endif	/* CONFIG_FAIR_GROUP_SCHED */

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


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

115
static inline u64 max_vruntime(u64 min_vruntime, u64 vruntime)
116
{
117 118
	s64 delta = (s64)(vruntime - min_vruntime);
	if (delta > 0)
119 120 121 122 123
		min_vruntime = vruntime;

	return min_vruntime;
}

124
static inline u64 min_vruntime(u64 min_vruntime, u64 vruntime)
P
Peter Zijlstra 已提交
125 126 127 128 129 130 131 132
{
	s64 delta = (s64)(vruntime - min_vruntime);
	if (delta < 0)
		min_vruntime = vruntime;

	return min_vruntime;
}

133
static inline s64 entity_key(struct cfs_rq *cfs_rq, struct sched_entity *se)
134
{
135
	return se->vruntime - cfs_rq->min_vruntime;
136 137
}

138 139 140
/*
 * Enqueue an entity into the rb-tree:
 */
141
static void __enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
142 143 144 145
{
	struct rb_node **link = &cfs_rq->tasks_timeline.rb_node;
	struct rb_node *parent = NULL;
	struct sched_entity *entry;
146
	s64 key = entity_key(cfs_rq, se);
147 148 149 150 151 152 153 154 155 156 157 158
	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.
		 */
159
		if (key < entity_key(cfs_rq, entry)) {
160 161 162 163 164 165 166 167 168 169 170 171
			link = &parent->rb_left;
		} else {
			link = &parent->rb_right;
			leftmost = 0;
		}
	}

	/*
	 * Maintain a cache of leftmost tree entries (it is frequently
	 * used):
	 */
	if (leftmost)
I
Ingo Molnar 已提交
172
		cfs_rq->rb_leftmost = &se->run_node;
173 174 175 176 177

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

178
static void __dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
179 180
{
	if (cfs_rq->rb_leftmost == &se->run_node)
I
Ingo Molnar 已提交
181
		cfs_rq->rb_leftmost = rb_next(&se->run_node);
I
Ingo Molnar 已提交
182

183 184 185 186 187 188 189 190 191 192 193 194 195
	rb_erase(&se->run_node, &cfs_rq->tasks_timeline);
}

static inline struct rb_node *first_fair(struct cfs_rq *cfs_rq)
{
	return cfs_rq->rb_leftmost;
}

static struct sched_entity *__pick_next_entity(struct cfs_rq *cfs_rq)
{
	return rb_entry(first_fair(cfs_rq), struct sched_entity, run_node);
}

196 197 198 199 200 201 202 203 204 205 206 207 208 209 210
static inline struct sched_entity *__pick_last_entity(struct cfs_rq *cfs_rq)
{
	struct rb_node **link = &cfs_rq->tasks_timeline.rb_node;
	struct sched_entity *se = NULL;
	struct rb_node *parent;

	while (*link) {
		parent = *link;
		se = rb_entry(parent, struct sched_entity, run_node);
		link = &parent->rb_right;
	}

	return se;
}

211 212 213 214
/**************************************************************
 * Scheduling class statistics methods:
 */

215 216 217 218 219 220 221 222 223

/*
 * The idea is to set a period in which each task runs once.
 *
 * When there are too many tasks (sysctl_sched_nr_latency) we have to stretch
 * this period because otherwise the slices get too small.
 *
 * p = (nr <= nl) ? l : l*nr/nl
 */
224 225 226
static u64 __sched_period(unsigned long nr_running)
{
	u64 period = sysctl_sched_latency;
227
	unsigned long nr_latency = sysctl_sched_nr_latency;
228 229 230 231 232 233 234 235 236

	if (unlikely(nr_running > nr_latency)) {
		period *= nr_running;
		do_div(period, nr_latency);
	}

	return period;
}

237 238 239 240 241 242
/*
 * We calculate the wall-time slice from the period by taking a part
 * proportional to the weight.
 *
 * s = p*w/rw
 */
P
Peter Zijlstra 已提交
243
static u64 sched_slice(struct cfs_rq *cfs_rq, struct sched_entity *se)
244
{
245
	u64 slice = __sched_period(cfs_rq->nr_running);
246

247 248
	slice *= se->load.weight;
	do_div(slice, cfs_rq->load.weight);
249

250
	return slice;
251 252
}

253 254 255 256 257 258
/*
 * We calculate the vruntime slice.
 *
 * vs = s/w = p/rw
 */
static u64 __sched_vslice(unsigned long rq_weight, unsigned long nr_running)
P
Peter Zijlstra 已提交
259
{
260
	u64 vslice = __sched_period(nr_running);
P
Peter Zijlstra 已提交
261

262
	do_div(vslice, rq_weight);
P
Peter Zijlstra 已提交
263

264 265
	return vslice;
}
266

267 268 269 270 271 272 273 274 275
static u64 sched_vslice(struct cfs_rq *cfs_rq)
{
	return __sched_vslice(cfs_rq->load.weight, cfs_rq->nr_running);
}

static u64 sched_vslice_add(struct cfs_rq *cfs_rq, struct sched_entity *se)
{
	return __sched_vslice(cfs_rq->load.weight + se->load.weight,
			cfs_rq->nr_running + 1);
P
Peter Zijlstra 已提交
276 277
}

278 279 280 281 282
/*
 * 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
283 284
__update_curr(struct cfs_rq *cfs_rq, struct sched_entity *curr,
	      unsigned long delta_exec)
285
{
286
	unsigned long delta_exec_weighted;
P
Peter Zijlstra 已提交
287
	u64 vruntime;
288

289
	schedstat_set(curr->exec_max, max((u64)delta_exec, curr->exec_max));
290 291

	curr->sum_exec_runtime += delta_exec;
292
	schedstat_add(cfs_rq, exec_clock, delta_exec);
I
Ingo Molnar 已提交
293 294 295 296 297 298
	delta_exec_weighted = delta_exec;
	if (unlikely(curr->load.weight != NICE_0_LOAD)) {
		delta_exec_weighted = calc_delta_fair(delta_exec_weighted,
							&curr->load);
	}
	curr->vruntime += delta_exec_weighted;
299 300 301 302 303 304

	/*
	 * maintain cfs_rq->min_vruntime to be a monotonic increasing
	 * value tracking the leftmost vruntime in the tree.
	 */
	if (first_fair(cfs_rq)) {
P
Peter Zijlstra 已提交
305 306
		vruntime = min_vruntime(curr->vruntime,
				__pick_next_entity(cfs_rq)->vruntime);
307
	} else
P
Peter Zijlstra 已提交
308
		vruntime = curr->vruntime;
309 310

	cfs_rq->min_vruntime =
P
Peter Zijlstra 已提交
311
		max_vruntime(cfs_rq->min_vruntime, vruntime);
312 313
}

314
static void update_curr(struct cfs_rq *cfs_rq)
315
{
316
	struct sched_entity *curr = cfs_rq->curr;
I
Ingo Molnar 已提交
317
	u64 now = rq_of(cfs_rq)->clock;
318 319 320 321 322 323 324 325 326 327
	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
328
	delta_exec = (unsigned long)(now - curr->exec_start);
329

I
Ingo Molnar 已提交
330 331
	__update_curr(cfs_rq, curr, delta_exec);
	curr->exec_start = now;
332 333 334
}

static inline void
335
update_stats_wait_start(struct cfs_rq *cfs_rq, struct sched_entity *se)
336
{
337
	schedstat_set(se->wait_start, rq_of(cfs_rq)->clock);
338 339 340 341 342
}

/*
 * Task is being enqueued - update stats:
 */
343
static void update_stats_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se)
344 345 346 347 348
{
	/*
	 * Are we enqueueing a waiting task? (for current tasks
	 * a dequeue/enqueue event is a NOP)
	 */
349
	if (se != cfs_rq->curr)
350
		update_stats_wait_start(cfs_rq, se);
351 352 353
}

static void
354
update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se)
355
{
356 357
	schedstat_set(se->wait_max, max(se->wait_max,
			rq_of(cfs_rq)->clock - se->wait_start));
I
Ingo Molnar 已提交
358
	schedstat_set(se->wait_start, 0);
359 360 361
}

static inline void
362
update_stats_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se)
363 364 365 366 367
{
	/*
	 * Mark the end of the wait period if dequeueing a
	 * waiting task:
	 */
368
	if (se != cfs_rq->curr)
369
		update_stats_wait_end(cfs_rq, se);
370 371 372 373 374 375
}

/*
 * We are picking a new current task - update its stats:
 */
static inline void
376
update_stats_curr_start(struct cfs_rq *cfs_rq, struct sched_entity *se)
377 378 379 380
{
	/*
	 * We are starting a new run period:
	 */
381
	se->exec_start = rq_of(cfs_rq)->clock;
382 383 384 385 386 387
}

/*
 * We are descheduling a task - update its stats:
 */
static inline void
388
update_stats_curr_end(struct cfs_rq *cfs_rq, struct sched_entity *se)
389 390 391 392 393 394 395 396
{
	se->exec_start = 0;
}

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

397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412
static void
account_entity_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se)
{
	update_load_add(&cfs_rq->load, se->load.weight);
	cfs_rq->nr_running++;
	se->on_rq = 1;
}

static void
account_entity_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se)
{
	update_load_sub(&cfs_rq->load, se->load.weight);
	cfs_rq->nr_running--;
	se->on_rq = 0;
}

413
static void enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
414 415 416
{
#ifdef CONFIG_SCHEDSTATS
	if (se->sleep_start) {
417
		u64 delta = rq_of(cfs_rq)->clock - se->sleep_start;
418 419 420 421 422 423 424 425 426 427 428

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

		if (unlikely(delta > se->sleep_max))
			se->sleep_max = delta;

		se->sleep_start = 0;
		se->sum_sleep_runtime += delta;
	}
	if (se->block_start) {
429
		u64 delta = rq_of(cfs_rq)->clock - se->block_start;
430 431 432 433 434 435 436 437 438

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

		if (unlikely(delta > se->block_max))
			se->block_max = delta;

		se->block_start = 0;
		se->sum_sleep_runtime += delta;
I
Ingo Molnar 已提交
439 440 441 442 443 444 445

		/*
		 * 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)) {
I
Ingo Molnar 已提交
446 447
			struct task_struct *tsk = task_of(se);

I
Ingo Molnar 已提交
448 449 450
			profile_hits(SLEEP_PROFILING, (void *)get_wchan(tsk),
				     delta >> 20);
		}
451 452 453 454
	}
#endif
}

P
Peter Zijlstra 已提交
455 456 457 458 459 460 461 462 463 464 465 466 467
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
}

468 469 470
static void
place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial)
{
P
Peter Zijlstra 已提交
471
	u64 vruntime;
472

P
Peter Zijlstra 已提交
473
	vruntime = cfs_rq->min_vruntime;
P
Peter Zijlstra 已提交
474

475
	if (sched_feat(TREE_AVG)) {
P
Peter Zijlstra 已提交
476 477
		struct sched_entity *last = __pick_last_entity(cfs_rq);
		if (last) {
P
Peter Zijlstra 已提交
478 479
			vruntime += last->vruntime;
			vruntime >>= 1;
P
Peter Zijlstra 已提交
480
		}
P
Peter Zijlstra 已提交
481
	} else if (sched_feat(APPROX_AVG) && cfs_rq->nr_running)
482
		vruntime += sched_vslice(cfs_rq)/2;
P
Peter Zijlstra 已提交
483 484

	if (initial && sched_feat(START_DEBIT))
485
		vruntime += sched_vslice_add(cfs_rq, se);
486

I
Ingo Molnar 已提交
487
	if (!initial) {
488 489
		if (sched_feat(NEW_FAIR_SLEEPERS) && entity_is_task(se) &&
				task_of(se)->policy != SCHED_BATCH)
490 491
			vruntime -= sysctl_sched_latency;

492
		vruntime = max_t(s64, vruntime, se->vruntime);
493 494
	}

P
Peter Zijlstra 已提交
495 496
	se->vruntime = vruntime;

497 498
}

499
static void
500
enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int wakeup)
501 502
{
	/*
503
	 * Update run-time statistics of the 'current'.
504
	 */
505
	update_curr(cfs_rq);
506

I
Ingo Molnar 已提交
507
	if (wakeup) {
508
		place_entity(cfs_rq, se, 0);
509
		enqueue_sleeper(cfs_rq, se);
I
Ingo Molnar 已提交
510
	}
511

512
	update_stats_enqueue(cfs_rq, se);
P
Peter Zijlstra 已提交
513
	check_spread(cfs_rq, se);
514 515
	if (se != cfs_rq->curr)
		__enqueue_entity(cfs_rq, se);
516
	account_entity_enqueue(cfs_rq, se);
517 518 519
}

static void
520
dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int sleep)
521
{
522 523 524 525 526
	/*
	 * Update run-time statistics of the 'current'.
	 */
	update_curr(cfs_rq);

527
	update_stats_dequeue(cfs_rq, se);
528
	if (sleep) {
P
Peter Zijlstra 已提交
529
#ifdef CONFIG_SCHEDSTATS
530 531 532 533
		if (entity_is_task(se)) {
			struct task_struct *tsk = task_of(se);

			if (tsk->state & TASK_INTERRUPTIBLE)
534
				se->sleep_start = rq_of(cfs_rq)->clock;
535
			if (tsk->state & TASK_UNINTERRUPTIBLE)
536
				se->block_start = rq_of(cfs_rq)->clock;
537
		}
538
#endif
P
Peter Zijlstra 已提交
539 540
	}

541
	if (se != cfs_rq->curr)
542 543
		__dequeue_entity(cfs_rq, se);
	account_entity_dequeue(cfs_rq, se);
544 545 546 547 548
}

/*
 * Preempt the current task with a newly woken task if needed:
 */
549
static void
I
Ingo Molnar 已提交
550
check_preempt_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr)
551
{
552 553
	unsigned long ideal_runtime, delta_exec;

P
Peter Zijlstra 已提交
554
	ideal_runtime = sched_slice(cfs_rq, curr);
555 556
	delta_exec = curr->sum_exec_runtime - curr->prev_sum_exec_runtime;
	if (delta_exec > ideal_runtime)
557 558 559
		resched_task(rq_of(cfs_rq)->curr);
}

560
static void
561
set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
562
{
563 564 565 566 567 568 569 570 571 572 573
	/* '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);
	}

574
	update_stats_curr_start(cfs_rq, se);
575
	cfs_rq->curr = se;
I
Ingo Molnar 已提交
576 577 578 579 580 581
#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):
	 */
582
	if (rq_of(cfs_rq)->load.weight >= 2*se->load.weight) {
I
Ingo Molnar 已提交
583 584 585 586
		se->slice_max = max(se->slice_max,
			se->sum_exec_runtime - se->prev_sum_exec_runtime);
	}
#endif
587
	se->prev_sum_exec_runtime = se->sum_exec_runtime;
588 589
}

590
static struct sched_entity *pick_next_entity(struct cfs_rq *cfs_rq)
591
{
D
Dmitry Adamushko 已提交
592
	struct sched_entity *se = NULL;
593

D
Dmitry Adamushko 已提交
594 595 596 597
	if (first_fair(cfs_rq)) {
		se = __pick_next_entity(cfs_rq);
		set_next_entity(cfs_rq, se);
	}
598 599 600 601

	return se;
}

602
static void put_prev_entity(struct cfs_rq *cfs_rq, struct sched_entity *prev)
603 604 605 606 607 608
{
	/*
	 * If still on the runqueue then deactivate_task()
	 * was not called and update_curr() has to be done:
	 */
	if (prev->on_rq)
609
		update_curr(cfs_rq);
610

611
	update_stats_curr_end(cfs_rq, prev);
612

P
Peter Zijlstra 已提交
613
	check_spread(cfs_rq, prev);
614
	if (prev->on_rq) {
615
		update_stats_wait_start(cfs_rq, prev);
616 617 618
		/* Put 'current' back into the tree. */
		__enqueue_entity(cfs_rq, prev);
	}
619
	cfs_rq->curr = NULL;
620 621 622 623 624
}

static void entity_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr)
{
	/*
625
	 * Update run-time statistics of the 'current'.
626
	 */
627
	update_curr(cfs_rq);
628

I
Ingo Molnar 已提交
629 630
	if (cfs_rq->nr_running > 1)
		check_preempt_tick(cfs_rq, curr);
631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664
}

/**************************************************
 * CFS operations on tasks:
 */

#ifdef CONFIG_FAIR_GROUP_SCHED

/* 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;
}

/* Given a group's cfs_rq on one cpu, return its corresponding cfs_rq on
 * another cpu ('this_cpu')
 */
static inline struct cfs_rq *cpu_cfs_rq(struct cfs_rq *cfs_rq, int this_cpu)
{
S
Srivatsa Vaddagiri 已提交
665
	return cfs_rq->tg->cfs_rq[this_cpu];
666 667 668 669 670 671
}

/* Iterate thr' all leaf cfs_rq's on a runqueue */
#define for_each_leaf_cfs_rq(rq, cfs_rq) \
	list_for_each_entry(cfs_rq, &rq->leaf_cfs_rq_list, leaf_cfs_rq_list)

672 673 674
/* Do the two (enqueued) entities belong to the same group ? */
static inline int
is_same_group(struct sched_entity *se, struct sched_entity *pse)
675
{
676
	if (se->cfs_rq == pse->cfs_rq)
677 678 679 680 681
		return 1;

	return 0;
}

682 683 684 685 686
static inline struct sched_entity *parent_entity(struct sched_entity *se)
{
	return se->parent;
}

687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718
#else	/* CONFIG_FAIR_GROUP_SCHED */

#define for_each_sched_entity(se) \
		for (; se; se = NULL)

static inline struct cfs_rq *task_cfs_rq(struct task_struct *p)
{
	return &task_rq(p)->cfs;
}

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;
}

static inline struct cfs_rq *cpu_cfs_rq(struct cfs_rq *cfs_rq, int this_cpu)
{
	return &cpu_rq(this_cpu)->cfs;
}

#define for_each_leaf_cfs_rq(rq, cfs_rq) \
		for (cfs_rq = &rq->cfs; cfs_rq; cfs_rq = NULL)

719 720
static inline int
is_same_group(struct sched_entity *se, struct sched_entity *pse)
721 722 723 724
{
	return 1;
}

725 726 727 728 729
static inline struct sched_entity *parent_entity(struct sched_entity *se)
{
	return NULL;
}

730 731 732 733 734 735 736
#endif	/* CONFIG_FAIR_GROUP_SCHED */

/*
 * 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:
 */
737
static void enqueue_task_fair(struct rq *rq, struct task_struct *p, int wakeup)
738 739 740 741 742 743 744 745
{
	struct cfs_rq *cfs_rq;
	struct sched_entity *se = &p->se;

	for_each_sched_entity(se) {
		if (se->on_rq)
			break;
		cfs_rq = cfs_rq_of(se);
746
		enqueue_entity(cfs_rq, se, wakeup);
747
		wakeup = 1;
748 749 750 751 752 753 754 755
	}
}

/*
 * The dequeue_task method is called before nr_running is
 * decreased. We remove the task from the rbtree and
 * update the fair scheduling stats:
 */
756
static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int sleep)
757 758 759 760 761 762
{
	struct cfs_rq *cfs_rq;
	struct sched_entity *se = &p->se;

	for_each_sched_entity(se) {
		cfs_rq = cfs_rq_of(se);
763
		dequeue_entity(cfs_rq, se, sleep);
764 765 766
		/* Don't dequeue parent if it has other entities besides us */
		if (cfs_rq->load.weight)
			break;
767
		sleep = 1;
768 769 770 771
	}
}

/*
772 773 774
 * sched_yield() support is very simple - we dequeue and enqueue.
 *
 * If compat_yield is turned on then we requeue to the end of the tree.
775
 */
776
static void yield_task_fair(struct rq *rq)
777
{
S
Srivatsa Vaddagiri 已提交
778
	struct cfs_rq *cfs_rq = task_cfs_rq(rq->curr);
779
	struct sched_entity *rightmost, *se = &rq->curr->se;
780 781

	/*
782 783 784 785 786 787 788 789
	 * Are we the only task in the tree?
	 */
	if (unlikely(cfs_rq->nr_running == 1))
		return;

	if (likely(!sysctl_sched_compat_yield)) {
		__update_rq_clock(rq);
		/*
790
		 * Update run-time statistics of the 'current'.
791
		 */
D
Dmitry Adamushko 已提交
792
		update_curr(cfs_rq);
793 794 795 796 797

		return;
	}
	/*
	 * Find the rightmost entry in the rbtree:
798
	 */
D
Dmitry Adamushko 已提交
799
	rightmost = __pick_last_entity(cfs_rq);
800 801 802
	/*
	 * Already in the rightmost position?
	 */
D
Dmitry Adamushko 已提交
803
	if (unlikely(rightmost->vruntime < se->vruntime))
804 805 806 807
		return;

	/*
	 * Minimally necessary key value to be last in the tree:
D
Dmitry Adamushko 已提交
808 809
	 * Upon rescheduling, sched_class::put_prev_task() will place
	 * 'current' within the tree based on its new key value.
810
	 */
811
	se->vruntime = rightmost->vruntime + 1;
812 813 814 815 816
}

/*
 * Preempt the current task with a newly woken task if needed:
 */
I
Ingo Molnar 已提交
817
static void check_preempt_wakeup(struct rq *rq, struct task_struct *p)
818 819
{
	struct task_struct *curr = rq->curr;
820
	struct cfs_rq *cfs_rq = task_cfs_rq(curr);
821
	struct sched_entity *se = &curr->se, *pse = &p->se;
822
	s64 delta, gran;
823 824

	if (unlikely(rt_prio(p->prio))) {
I
Ingo Molnar 已提交
825
		update_rq_clock(rq);
826
		update_curr(cfs_rq);
827 828 829 830
		resched_task(curr);
		return;
	}

831 832 833 834
	while (!is_same_group(se, pse)) {
		se = parent_entity(se);
		pse = parent_entity(pse);
	}
835

836
	delta = se->vruntime - pse->vruntime;
837 838 839
	gran = sysctl_sched_wakeup_granularity;
	if (unlikely(se->load.weight != NICE_0_LOAD))
		gran = calc_delta_fair(gran, &se->load);
840

841
	if (delta > gran)
842
		resched_task(curr);
843 844
}

845
static struct task_struct *pick_next_task_fair(struct rq *rq)
846 847 848 849 850 851 852 853
{
	struct cfs_rq *cfs_rq = &rq->cfs;
	struct sched_entity *se;

	if (unlikely(!cfs_rq->nr_running))
		return NULL;

	do {
854
		se = pick_next_entity(cfs_rq);
855 856 857 858 859 860 861 862 863
		cfs_rq = group_cfs_rq(se);
	} while (cfs_rq);

	return task_of(se);
}

/*
 * Account for a descheduled task:
 */
864
static void put_prev_task_fair(struct rq *rq, struct task_struct *prev)
865 866 867 868 869 870
{
	struct sched_entity *se = &prev->se;
	struct cfs_rq *cfs_rq;

	for_each_sched_entity(se) {
		cfs_rq = cfs_rq_of(se);
871
		put_prev_entity(cfs_rq, se);
872 873 874 875 876 877 878 879 880 881 882 883 884 885
	}
}

/**************************************************
 * Fair scheduling class load-balancing methods:
 */

/*
 * Load-balancing iterator. Note: while the runqueue stays locked
 * during the whole iteration, the current task might be
 * dequeued so the iterator has to be dequeue-safe. Here we
 * achieve that by always pre-iterating before returning
 * the current task:
 */
A
Alexey Dobriyan 已提交
886
static struct task_struct *
887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913
__load_balance_iterator(struct cfs_rq *cfs_rq, struct rb_node *curr)
{
	struct task_struct *p;

	if (!curr)
		return NULL;

	p = rb_entry(curr, struct task_struct, se.run_node);
	cfs_rq->rb_load_balance_curr = rb_next(curr);

	return p;
}

static struct task_struct *load_balance_start_fair(void *arg)
{
	struct cfs_rq *cfs_rq = arg;

	return __load_balance_iterator(cfs_rq, first_fair(cfs_rq));
}

static struct task_struct *load_balance_next_fair(void *arg)
{
	struct cfs_rq *cfs_rq = arg;

	return __load_balance_iterator(cfs_rq, cfs_rq->rb_load_balance_curr);
}

914
#ifdef CONFIG_FAIR_GROUP_SCHED
915 916 917 918 919 920 921 922
static int cfs_rq_best_prio(struct cfs_rq *cfs_rq)
{
	struct sched_entity *curr;
	struct task_struct *p;

	if (!cfs_rq->nr_running)
		return MAX_PRIO;

923 924 925 926
	curr = cfs_rq->curr;
	if (!curr)
		curr = __pick_next_entity(cfs_rq);

927 928 929 930
	p = task_of(curr);

	return p->prio;
}
931
#endif
932

P
Peter Williams 已提交
933
static unsigned long
934
load_balance_fair(struct rq *this_rq, int this_cpu, struct rq *busiest,
935 936 937
		  unsigned long max_nr_move, unsigned long max_load_move,
		  struct sched_domain *sd, enum cpu_idle_type idle,
		  int *all_pinned, int *this_best_prio)
938 939 940 941 942 943 944 945 946 947
{
	struct cfs_rq *busy_cfs_rq;
	unsigned long load_moved, total_nr_moved = 0, nr_moved;
	long rem_load_move = max_load_move;
	struct rq_iterator cfs_rq_iterator;

	cfs_rq_iterator.start = load_balance_start_fair;
	cfs_rq_iterator.next = load_balance_next_fair;

	for_each_leaf_cfs_rq(busiest, busy_cfs_rq) {
948
#ifdef CONFIG_FAIR_GROUP_SCHED
949
		struct cfs_rq *this_cfs_rq;
950
		long imbalance;
951 952 953 954
		unsigned long maxload;

		this_cfs_rq = cpu_cfs_rq(busy_cfs_rq, this_cpu);

955
		imbalance = busy_cfs_rq->load.weight - this_cfs_rq->load.weight;
956 957 958 959 960 961 962 963
		/* Don't pull if this_cfs_rq has more load than busy_cfs_rq */
		if (imbalance <= 0)
			continue;

		/* Don't pull more than imbalance/2 */
		imbalance /= 2;
		maxload = min(rem_load_move, imbalance);

964 965
		*this_best_prio = cfs_rq_best_prio(this_cfs_rq);
#else
966
# define maxload rem_load_move
967
#endif
968 969 970 971 972 973
		/* pass busy_cfs_rq argument into
		 * load_balance_[start|next]_fair iterators
		 */
		cfs_rq_iterator.arg = busy_cfs_rq;
		nr_moved = balance_tasks(this_rq, this_cpu, busiest,
				max_nr_move, maxload, sd, idle, all_pinned,
974
				&load_moved, this_best_prio, &cfs_rq_iterator);
975 976 977 978 979 980 981 982 983

		total_nr_moved += nr_moved;
		max_nr_move -= nr_moved;
		rem_load_move -= load_moved;

		if (max_nr_move <= 0 || rem_load_move <= 0)
			break;
	}

P
Peter Williams 已提交
984
	return max_load_move - rem_load_move;
985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000
}

/*
 * scheduler tick hitting a task of our scheduling class:
 */
static void task_tick_fair(struct rq *rq, struct task_struct *curr)
{
	struct cfs_rq *cfs_rq;
	struct sched_entity *se = &curr->se;

	for_each_sched_entity(se) {
		cfs_rq = cfs_rq_of(se);
		entity_tick(cfs_rq, se);
	}
}

1001 1002
#define swap(a,b) do { typeof(a) tmp = (a); (a) = (b); (b) = tmp; } while (0)

1003 1004 1005 1006 1007 1008 1009
/*
 * Share the fairness runtime between parent and child, thus the
 * total amount of pressure for CPU stays equal - new tasks
 * get a chance to run but frequent forkers are not allowed to
 * monopolize the CPU. Note: the parent runqueue is locked,
 * the child is not running yet.
 */
1010
static void task_new_fair(struct rq *rq, struct task_struct *p)
1011 1012
{
	struct cfs_rq *cfs_rq = task_cfs_rq(p);
1013
	struct sched_entity *se = &p->se, *curr = cfs_rq->curr;
1014
	int this_cpu = smp_processor_id();
1015 1016 1017

	sched_info_queued(p);

1018
	update_curr(cfs_rq);
1019
	place_entity(cfs_rq, se, 1);
1020

1021
	if (sysctl_sched_child_runs_first && this_cpu == task_cpu(p) &&
1022
			curr->vruntime < se->vruntime) {
D
Dmitry Adamushko 已提交
1023
		/*
1024 1025 1026
		 * Upon rescheduling, sched_class::put_prev_task() will place
		 * 'current' within the tree based on its new key value.
		 */
1027 1028
		swap(curr->vruntime, se->vruntime);
	}
1029

I
Ingo Molnar 已提交
1030
	update_stats_enqueue(cfs_rq, se);
P
Peter Zijlstra 已提交
1031 1032
	check_spread(cfs_rq, se);
	check_spread(cfs_rq, curr);
1033
	__enqueue_entity(cfs_rq, se);
1034
	account_entity_enqueue(cfs_rq, se);
1035
	resched_task(rq->curr);
1036 1037
}

1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050
/* 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;

	for_each_sched_entity(se)
		set_next_entity(cfs_rq_of(se), se);
}

1051 1052 1053
/*
 * All the scheduling class methods:
 */
1054 1055
static const struct sched_class fair_sched_class = {
	.next			= &idle_sched_class,
1056 1057 1058 1059
	.enqueue_task		= enqueue_task_fair,
	.dequeue_task		= dequeue_task_fair,
	.yield_task		= yield_task_fair,

I
Ingo Molnar 已提交
1060
	.check_preempt_curr	= check_preempt_wakeup,
1061 1062 1063 1064 1065 1066

	.pick_next_task		= pick_next_task_fair,
	.put_prev_task		= put_prev_task_fair,

	.load_balance		= load_balance_fair,

1067
	.set_curr_task          = set_curr_task_fair,
1068 1069 1070 1071 1072
	.task_tick		= task_tick_fair,
	.task_new		= task_new_fair,
};

#ifdef CONFIG_SCHED_DEBUG
1073
static void print_cfs_stats(struct seq_file *m, int cpu)
1074 1075 1076
{
	struct cfs_rq *cfs_rq;

S
Srivatsa Vaddagiri 已提交
1077 1078 1079
#ifdef CONFIG_FAIR_GROUP_SCHED
	print_cfs_rq(m, cpu, &cpu_rq(cpu)->cfs);
#endif
1080
	for_each_leaf_cfs_rq(cpu_rq(cpu), cfs_rq)
1081
		print_cfs_rq(m, cpu, cfs_rq);
1082 1083
}
#endif