core.c 178.2 KB
Newer Older
L
Linus Torvalds 已提交
1
/*
2
 *  kernel/sched/core.c
L
Linus Torvalds 已提交
3
 *
I
Ingo Molnar 已提交
4
 *  Core kernel scheduler code and related syscalls
L
Linus Torvalds 已提交
5 6 7
 *
 *  Copyright (C) 1991-2002  Linus Torvalds
 */
8
#include <linux/sched.h>
L
Linus Torvalds 已提交
9
#include <linux/cpuset.h>
10
#include <linux/delayacct.h>
11
#include <linux/init_task.h>
12
#include <linux/context_tracking.h>
13 14 15 16 17 18

#include <linux/blkdev.h>
#include <linux/kprobes.h>
#include <linux/mmu_context.h>
#include <linux/module.h>
#include <linux/nmi.h>
19
#include <linux/prefetch.h>
20 21 22
#include <linux/profile.h>
#include <linux/security.h>
#include <linux/syscalls.h>
L
Linus Torvalds 已提交
23

24
#include <asm/switch_to.h>
25
#include <asm/tlb.h>
26 27 28
#ifdef CONFIG_PARAVIRT
#include <asm/paravirt.h>
#endif
L
Linus Torvalds 已提交
29

30
#include "sched.h"
31
#include "../workqueue_internal.h"
32
#include "../smpboot.h"
33

34
#define CREATE_TRACE_POINTS
35
#include <trace/events/sched.h>
36

37
DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
38

I
Ingo Molnar 已提交
39 40 41
/*
 * Debugging: various feature bits
 */
P
Peter Zijlstra 已提交
42 43 44 45

#define SCHED_FEAT(name, enabled)	\
	(1UL << __SCHED_FEAT_##name) * enabled |

I
Ingo Molnar 已提交
46
const_debug unsigned int sysctl_sched_features =
47
#include "features.h"
P
Peter Zijlstra 已提交
48 49 50 51
	0;

#undef SCHED_FEAT

52 53 54 55 56 57
/*
 * Number of tasks to iterate in a single balance run.
 * Limited because this is done with IRQs disabled.
 */
const_debug unsigned int sysctl_sched_nr_migrate = 32;

58 59 60 61 62 63 64 65
/*
 * period over which we average the RT time consumption, measured
 * in ms.
 *
 * default: 1s
 */
const_debug unsigned int sysctl_sched_time_avg = MSEC_PER_SEC;

P
Peter Zijlstra 已提交
66
/*
I
Ingo Molnar 已提交
67
 * period over which we measure -rt task CPU usage in us.
P
Peter Zijlstra 已提交
68 69
 * default: 1s
 */
P
Peter Zijlstra 已提交
70
unsigned int sysctl_sched_rt_period = 1000000;
P
Peter Zijlstra 已提交
71

72
__read_mostly int scheduler_running;
73

P
Peter Zijlstra 已提交
74 75 76 77 78
/*
 * part of the period that we allow rt tasks to run in us.
 * default: 0.95s
 */
int sysctl_sched_rt_runtime = 950000;
P
Peter Zijlstra 已提交
79

I
Ingo Molnar 已提交
80
/* CPUs with isolated domains */
81 82
cpumask_var_t cpu_isolated_map;

L
Linus Torvalds 已提交
83
/*
84
 * this_rq_lock - lock this runqueue and disable interrupts.
L
Linus Torvalds 已提交
85
 */
A
Alexey Dobriyan 已提交
86
static struct rq *this_rq_lock(void)
L
Linus Torvalds 已提交
87 88
	__acquires(rq->lock)
{
89
	struct rq *rq;
L
Linus Torvalds 已提交
90 91 92

	local_irq_disable();
	rq = this_rq();
93
	raw_spin_lock(&rq->lock);
L
Linus Torvalds 已提交
94 95 96 97

	return rq;
}

98 99 100
/*
 * __task_rq_lock - lock the rq @p resides on.
 */
101
struct rq *__task_rq_lock(struct task_struct *p, struct rq_flags *rf)
102 103 104 105 106 107 108 109 110 111
	__acquires(rq->lock)
{
	struct rq *rq;

	lockdep_assert_held(&p->pi_lock);

	for (;;) {
		rq = task_rq(p);
		raw_spin_lock(&rq->lock);
		if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) {
112
			rq_pin_lock(rq, rf);
113 114 115 116 117 118 119 120 121 122 123 124
			return rq;
		}
		raw_spin_unlock(&rq->lock);

		while (unlikely(task_on_rq_migrating(p)))
			cpu_relax();
	}
}

/*
 * task_rq_lock - lock p->pi_lock and lock the rq @p resides on.
 */
125
struct rq *task_rq_lock(struct task_struct *p, struct rq_flags *rf)
126 127 128 129 130 131
	__acquires(p->pi_lock)
	__acquires(rq->lock)
{
	struct rq *rq;

	for (;;) {
132
		raw_spin_lock_irqsave(&p->pi_lock, rf->flags);
133 134 135 136 137 138 139 140 141 142 143 144 145 146 147
		rq = task_rq(p);
		raw_spin_lock(&rq->lock);
		/*
		 *	move_queued_task()		task_rq_lock()
		 *
		 *	ACQUIRE (rq->lock)
		 *	[S] ->on_rq = MIGRATING		[L] rq = task_rq()
		 *	WMB (__set_task_cpu())		ACQUIRE (rq->lock);
		 *	[S] ->cpu = new_cpu		[L] task_rq()
		 *					[L] ->on_rq
		 *	RELEASE (rq->lock)
		 *
		 * If we observe the old cpu in task_rq_lock, the acquire of
		 * the old rq->lock will fully serialize against the stores.
		 *
I
Ingo Molnar 已提交
148
		 * If we observe the new CPU in task_rq_lock, the acquire will
149 150 151
		 * pair with the WMB to ensure we must then also see migrating.
		 */
		if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) {
152
			rq_pin_lock(rq, rf);
153 154 155
			return rq;
		}
		raw_spin_unlock(&rq->lock);
156
		raw_spin_unlock_irqrestore(&p->pi_lock, rf->flags);
157 158 159 160 161 162

		while (unlikely(task_on_rq_migrating(p)))
			cpu_relax();
	}
}

163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240
/*
 * RQ-clock updating methods:
 */

static void update_rq_clock_task(struct rq *rq, s64 delta)
{
/*
 * In theory, the compile should just see 0 here, and optimize out the call
 * to sched_rt_avg_update. But I don't trust it...
 */
#if defined(CONFIG_IRQ_TIME_ACCOUNTING) || defined(CONFIG_PARAVIRT_TIME_ACCOUNTING)
	s64 steal = 0, irq_delta = 0;
#endif
#ifdef CONFIG_IRQ_TIME_ACCOUNTING
	irq_delta = irq_time_read(cpu_of(rq)) - rq->prev_irq_time;

	/*
	 * Since irq_time is only updated on {soft,}irq_exit, we might run into
	 * this case when a previous update_rq_clock() happened inside a
	 * {soft,}irq region.
	 *
	 * When this happens, we stop ->clock_task and only update the
	 * prev_irq_time stamp to account for the part that fit, so that a next
	 * update will consume the rest. This ensures ->clock_task is
	 * monotonic.
	 *
	 * It does however cause some slight miss-attribution of {soft,}irq
	 * time, a more accurate solution would be to update the irq_time using
	 * the current rq->clock timestamp, except that would require using
	 * atomic ops.
	 */
	if (irq_delta > delta)
		irq_delta = delta;

	rq->prev_irq_time += irq_delta;
	delta -= irq_delta;
#endif
#ifdef CONFIG_PARAVIRT_TIME_ACCOUNTING
	if (static_key_false((&paravirt_steal_rq_enabled))) {
		steal = paravirt_steal_clock(cpu_of(rq));
		steal -= rq->prev_steal_time_rq;

		if (unlikely(steal > delta))
			steal = delta;

		rq->prev_steal_time_rq += steal;
		delta -= steal;
	}
#endif

	rq->clock_task += delta;

#if defined(CONFIG_IRQ_TIME_ACCOUNTING) || defined(CONFIG_PARAVIRT_TIME_ACCOUNTING)
	if ((irq_delta + steal) && sched_feat(NONTASK_CAPACITY))
		sched_rt_avg_update(rq, irq_delta + steal);
#endif
}

void update_rq_clock(struct rq *rq)
{
	s64 delta;

	lockdep_assert_held(&rq->lock);

	if (rq->clock_update_flags & RQCF_ACT_SKIP)
		return;

#ifdef CONFIG_SCHED_DEBUG
	rq->clock_update_flags |= RQCF_UPDATED;
#endif
	delta = sched_clock_cpu(cpu_of(rq)) - rq->clock;
	if (delta < 0)
		return;
	rq->clock += delta;
	update_rq_clock_task(rq, delta);
}


P
Peter Zijlstra 已提交
241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261
#ifdef CONFIG_SCHED_HRTICK
/*
 * Use HR-timers to deliver accurate preemption points.
 */

static void hrtick_clear(struct rq *rq)
{
	if (hrtimer_active(&rq->hrtick_timer))
		hrtimer_cancel(&rq->hrtick_timer);
}

/*
 * High-resolution timer tick.
 * Runs from hardirq context with interrupts disabled.
 */
static enum hrtimer_restart hrtick(struct hrtimer *timer)
{
	struct rq *rq = container_of(timer, struct rq, hrtick_timer);

	WARN_ON_ONCE(cpu_of(rq) != smp_processor_id());

262
	raw_spin_lock(&rq->lock);
263
	update_rq_clock(rq);
P
Peter Zijlstra 已提交
264
	rq->curr->sched_class->task_tick(rq, rq->curr, 1);
265
	raw_spin_unlock(&rq->lock);
P
Peter Zijlstra 已提交
266 267 268 269

	return HRTIMER_NORESTART;
}

270
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
271

272
static void __hrtick_restart(struct rq *rq)
P
Peter Zijlstra 已提交
273 274 275
{
	struct hrtimer *timer = &rq->hrtick_timer;

276
	hrtimer_start_expires(timer, HRTIMER_MODE_ABS_PINNED);
P
Peter Zijlstra 已提交
277 278
}

279 280 281 282
/*
 * called from hardirq (IPI) context
 */
static void __hrtick_start(void *arg)
283
{
284
	struct rq *rq = arg;
285

286
	raw_spin_lock(&rq->lock);
P
Peter Zijlstra 已提交
287
	__hrtick_restart(rq);
288
	rq->hrtick_csd_pending = 0;
289
	raw_spin_unlock(&rq->lock);
290 291
}

292 293 294 295 296
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
297
void hrtick_start(struct rq *rq, u64 delay)
298
{
299
	struct hrtimer *timer = &rq->hrtick_timer;
300 301 302 303 304 305 306 307 308
	ktime_t time;
	s64 delta;

	/*
	 * Don't schedule slices shorter than 10000ns, that just
	 * doesn't make sense and can cause timer DoS.
	 */
	delta = max_t(s64, delay, 10000LL);
	time = ktime_add_ns(timer->base->get_time(), delta);
309

310
	hrtimer_set_expires(timer, time);
311 312

	if (rq == this_rq()) {
P
Peter Zijlstra 已提交
313
		__hrtick_restart(rq);
314
	} else if (!rq->hrtick_csd_pending) {
315
		smp_call_function_single_async(cpu_of(rq), &rq->hrtick_csd);
316 317
		rq->hrtick_csd_pending = 1;
	}
318 319
}

320 321 322 323 324 325
#else
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
326
void hrtick_start(struct rq *rq, u64 delay)
327
{
W
Wanpeng Li 已提交
328 329 330 331 332
	/*
	 * Don't schedule slices shorter than 10000ns, that just
	 * doesn't make sense. Rely on vruntime for fairness.
	 */
	delay = max_t(u64, delay, 10000LL);
333 334
	hrtimer_start(&rq->hrtick_timer, ns_to_ktime(delay),
		      HRTIMER_MODE_REL_PINNED);
335 336
}
#endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
337

338
static void init_rq_hrtick(struct rq *rq)
P
Peter Zijlstra 已提交
339
{
340 341
#ifdef CONFIG_SMP
	rq->hrtick_csd_pending = 0;
P
Peter Zijlstra 已提交
342

343 344 345 346
	rq->hrtick_csd.flags = 0;
	rq->hrtick_csd.func = __hrtick_start;
	rq->hrtick_csd.info = rq;
#endif
P
Peter Zijlstra 已提交
347

348 349
	hrtimer_init(&rq->hrtick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	rq->hrtick_timer.function = hrtick;
P
Peter Zijlstra 已提交
350
}
A
Andrew Morton 已提交
351
#else	/* CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
352 353 354 355 356 357 358
static inline void hrtick_clear(struct rq *rq)
{
}

static inline void init_rq_hrtick(struct rq *rq)
{
}
A
Andrew Morton 已提交
359
#endif	/* CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
360

361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378
/*
 * cmpxchg based fetch_or, macro so it works for different integer types
 */
#define fetch_or(ptr, mask)						\
	({								\
		typeof(ptr) _ptr = (ptr);				\
		typeof(mask) _mask = (mask);				\
		typeof(*_ptr) _old, _val = *_ptr;			\
									\
		for (;;) {						\
			_old = cmpxchg(_ptr, _val, _val | _mask);	\
			if (_old == _val)				\
				break;					\
			_val = _old;					\
		}							\
	_old;								\
})

379
#if defined(CONFIG_SMP) && defined(TIF_POLLING_NRFLAG)
380 381 382 383 384 385 386 387 388 389
/*
 * Atomically set TIF_NEED_RESCHED and test for TIF_POLLING_NRFLAG,
 * this avoids any races wrt polling state changes and thereby avoids
 * spurious IPIs.
 */
static bool set_nr_and_not_polling(struct task_struct *p)
{
	struct thread_info *ti = task_thread_info(p);
	return !(fetch_or(&ti->flags, _TIF_NEED_RESCHED) & _TIF_POLLING_NRFLAG);
}
390 391 392 393 394 395 396 397 398 399

/*
 * Atomically set TIF_NEED_RESCHED if TIF_POLLING_NRFLAG is set.
 *
 * If this returns true, then the idle task promises to call
 * sched_ttwu_pending() and reschedule soon.
 */
static bool set_nr_if_polling(struct task_struct *p)
{
	struct thread_info *ti = task_thread_info(p);
400
	typeof(ti->flags) old, val = READ_ONCE(ti->flags);
401 402 403 404 405 406 407 408 409 410 411 412 413 414

	for (;;) {
		if (!(val & _TIF_POLLING_NRFLAG))
			return false;
		if (val & _TIF_NEED_RESCHED)
			return true;
		old = cmpxchg(&ti->flags, val, val | _TIF_NEED_RESCHED);
		if (old == val)
			break;
		val = old;
	}
	return true;
}

415 416 417 418 419 420
#else
static bool set_nr_and_not_polling(struct task_struct *p)
{
	set_tsk_need_resched(p);
	return true;
}
421 422 423 424 425 426 427

#ifdef CONFIG_SMP
static bool set_nr_if_polling(struct task_struct *p)
{
	return false;
}
#endif
428 429
#endif

430 431 432 433 434 435 436 437 438 439
void wake_q_add(struct wake_q_head *head, struct task_struct *task)
{
	struct wake_q_node *node = &task->wake_q;

	/*
	 * Atomically grab the task, if ->wake_q is !nil already it means
	 * its already queued (either by us or someone else) and will get the
	 * wakeup due to that.
	 *
	 * This cmpxchg() implies a full barrier, which pairs with the write
440
	 * barrier implied by the wakeup in wake_up_q().
441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462
	 */
	if (cmpxchg(&node->next, NULL, WAKE_Q_TAIL))
		return;

	get_task_struct(task);

	/*
	 * The head is context local, there can be no concurrency.
	 */
	*head->lastp = node;
	head->lastp = &node->next;
}

void wake_up_q(struct wake_q_head *head)
{
	struct wake_q_node *node = head->first;

	while (node != WAKE_Q_TAIL) {
		struct task_struct *task;

		task = container_of(node, struct task_struct, wake_q);
		BUG_ON(!task);
I
Ingo Molnar 已提交
463
		/* Task can safely be re-inserted now: */
464 465 466 467 468 469 470 471 472 473 474 475
		node = node->next;
		task->wake_q.next = NULL;

		/*
		 * wake_up_process() implies a wmb() to pair with the queueing
		 * in wake_q_add() so as not to miss wakeups.
		 */
		wake_up_process(task);
		put_task_struct(task);
	}
}

I
Ingo Molnar 已提交
476
/*
477
 * resched_curr - mark rq's current task 'to be rescheduled now'.
I
Ingo Molnar 已提交
478 479 480 481 482
 *
 * On UP this means the setting of the need_resched flag, on SMP it
 * might also involve a cross-CPU call to trigger the scheduler on
 * the target CPU.
 */
483
void resched_curr(struct rq *rq)
I
Ingo Molnar 已提交
484
{
485
	struct task_struct *curr = rq->curr;
I
Ingo Molnar 已提交
486 487
	int cpu;

488
	lockdep_assert_held(&rq->lock);
I
Ingo Molnar 已提交
489

490
	if (test_tsk_need_resched(curr))
I
Ingo Molnar 已提交
491 492
		return;

493
	cpu = cpu_of(rq);
494

495
	if (cpu == smp_processor_id()) {
496
		set_tsk_need_resched(curr);
497
		set_preempt_need_resched();
I
Ingo Molnar 已提交
498
		return;
499
	}
I
Ingo Molnar 已提交
500

501
	if (set_nr_and_not_polling(curr))
I
Ingo Molnar 已提交
502
		smp_send_reschedule(cpu);
503 504
	else
		trace_sched_wake_idle_without_ipi(cpu);
I
Ingo Molnar 已提交
505 506
}

507
void resched_cpu(int cpu)
I
Ingo Molnar 已提交
508 509 510 511
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

512
	if (!raw_spin_trylock_irqsave(&rq->lock, flags))
I
Ingo Molnar 已提交
513
		return;
514
	resched_curr(rq);
515
	raw_spin_unlock_irqrestore(&rq->lock, flags);
I
Ingo Molnar 已提交
516
}
517

518
#ifdef CONFIG_SMP
519
#ifdef CONFIG_NO_HZ_COMMON
520
/*
I
Ingo Molnar 已提交
521 522
 * In the semi idle case, use the nearest busy CPU for migrating timers
 * from an idle CPU.  This is good for power-savings.
523 524
 *
 * We don't do similar optimization for completely idle system, as
I
Ingo Molnar 已提交
525 526
 * selecting an idle CPU will add more delays to the timers than intended
 * (as that CPU's timer base may not be uptodate wrt jiffies etc).
527
 */
528
int get_nohz_timer_target(void)
529
{
530
	int i, cpu = smp_processor_id();
531 532
	struct sched_domain *sd;

533
	if (!idle_cpu(cpu) && is_housekeeping_cpu(cpu))
534 535
		return cpu;

536
	rcu_read_lock();
537
	for_each_domain(cpu, sd) {
538
		for_each_cpu(i, sched_domain_span(sd)) {
539 540 541 542
			if (cpu == i)
				continue;

			if (!idle_cpu(i) && is_housekeeping_cpu(i)) {
543 544 545 546
				cpu = i;
				goto unlock;
			}
		}
547
	}
548 549 550

	if (!is_housekeeping_cpu(cpu))
		cpu = housekeeping_any_cpu();
551 552
unlock:
	rcu_read_unlock();
553 554
	return cpu;
}
I
Ingo Molnar 已提交
555

556 557 558 559 560 561 562 563 564 565
/*
 * When add_timer_on() enqueues a timer into the timer wheel of an
 * idle CPU then this timer might expire before the next timer event
 * which is scheduled to wake up that CPU. In case of a completely
 * idle system the next event might even be infinite time into the
 * future. wake_up_idle_cpu() ensures that the CPU is woken up and
 * leaves the inner idle loop so the newly added timer is taken into
 * account when the CPU goes back to idle and evaluates the timer
 * wheel for the next timer event.
 */
566
static void wake_up_idle_cpu(int cpu)
567 568 569 570 571 572
{
	struct rq *rq = cpu_rq(cpu);

	if (cpu == smp_processor_id())
		return;

573
	if (set_nr_and_not_polling(rq->idle))
574
		smp_send_reschedule(cpu);
575 576
	else
		trace_sched_wake_idle_without_ipi(cpu);
577 578
}

579
static bool wake_up_full_nohz_cpu(int cpu)
580
{
581 582 583 584 585 586
	/*
	 * We just need the target to call irq_exit() and re-evaluate
	 * the next tick. The nohz full kick at least implies that.
	 * If needed we can still optimize that later with an
	 * empty IRQ.
	 */
587 588
	if (cpu_is_offline(cpu))
		return true;  /* Don't try to wake offline CPUs. */
589
	if (tick_nohz_full_cpu(cpu)) {
590 591
		if (cpu != smp_processor_id() ||
		    tick_nohz_tick_stopped())
592
			tick_nohz_full_kick_cpu(cpu);
593 594 595 596 597 598
		return true;
	}

	return false;
}

599 600 601 602 603
/*
 * Wake up the specified CPU.  If the CPU is going offline, it is the
 * caller's responsibility to deal with the lost wakeup, for example,
 * by hooking into the CPU_DEAD notifier like timers and hrtimers do.
 */
604 605
void wake_up_nohz_cpu(int cpu)
{
606
	if (!wake_up_full_nohz_cpu(cpu))
607 608 609
		wake_up_idle_cpu(cpu);
}

610
static inline bool got_nohz_idle_kick(void)
611
{
612
	int cpu = smp_processor_id();
613 614 615 616 617 618 619 620 621 622 623 624 625

	if (!test_bit(NOHZ_BALANCE_KICK, nohz_flags(cpu)))
		return false;

	if (idle_cpu(cpu) && !need_resched())
		return true;

	/*
	 * We can't run Idle Load Balance on this CPU for this time so we
	 * cancel it and clear NOHZ_BALANCE_KICK
	 */
	clear_bit(NOHZ_BALANCE_KICK, nohz_flags(cpu));
	return false;
626 627
}

628
#else /* CONFIG_NO_HZ_COMMON */
629

630
static inline bool got_nohz_idle_kick(void)
P
Peter Zijlstra 已提交
631
{
632
	return false;
P
Peter Zijlstra 已提交
633 634
}

635
#endif /* CONFIG_NO_HZ_COMMON */
636

637
#ifdef CONFIG_NO_HZ_FULL
638
bool sched_can_stop_tick(struct rq *rq)
639
{
640 641 642 643 644 645
	int fifo_nr_running;

	/* Deadline tasks, even if single, need the tick */
	if (rq->dl.dl_nr_running)
		return false;

646
	/*
647 648
	 * If there are more than one RR tasks, we need the tick to effect the
	 * actual RR behaviour.
649
	 */
650 651 652 653 654
	if (rq->rt.rr_nr_running) {
		if (rq->rt.rr_nr_running == 1)
			return true;
		else
			return false;
655 656
	}

657 658 659 660 661 662 663 664 665 666 667 668 669 670
	/*
	 * If there's no RR tasks, but FIFO tasks, we can skip the tick, no
	 * forced preemption between FIFO tasks.
	 */
	fifo_nr_running = rq->rt.rt_nr_running - rq->rt.rr_nr_running;
	if (fifo_nr_running)
		return true;

	/*
	 * If there are no DL,RR/FIFO tasks, there must only be CFS tasks left;
	 * if there's more than one we need the tick for involuntary
	 * preemption.
	 */
	if (rq->nr_running > 1)
671
		return false;
672

673
	return true;
674 675
}
#endif /* CONFIG_NO_HZ_FULL */
676

677
void sched_avg_update(struct rq *rq)
678
{
679 680
	s64 period = sched_avg_period();

681
	while ((s64)(rq_clock(rq) - rq->age_stamp) > period) {
682 683 684 685 686 687
		/*
		 * 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" (rq->age_stamp));
688 689 690
		rq->age_stamp += period;
		rq->rt_avg /= 2;
	}
691 692
}

693
#endif /* CONFIG_SMP */
694

695 696
#if defined(CONFIG_RT_GROUP_SCHED) || (defined(CONFIG_FAIR_GROUP_SCHED) && \
			(defined(CONFIG_SMP) || defined(CONFIG_CFS_BANDWIDTH)))
697
/*
698 699 700 701
 * Iterate task_group tree rooted at *from, calling @down when first entering a
 * node and @up when leaving it for the final time.
 *
 * Caller must hold rcu_lock or sufficient equivalent.
702
 */
703
int walk_tg_tree_from(struct task_group *from,
704
			     tg_visitor down, tg_visitor up, void *data)
705 706
{
	struct task_group *parent, *child;
P
Peter Zijlstra 已提交
707
	int ret;
708

709 710
	parent = from;

711
down:
P
Peter Zijlstra 已提交
712 713
	ret = (*down)(parent, data);
	if (ret)
714
		goto out;
715 716 717 718 719 720 721
	list_for_each_entry_rcu(child, &parent->children, siblings) {
		parent = child;
		goto down;

up:
		continue;
	}
P
Peter Zijlstra 已提交
722
	ret = (*up)(parent, data);
723 724
	if (ret || parent == from)
		goto out;
725 726 727 728 729

	child = parent;
	parent = parent->parent;
	if (parent)
		goto up;
730
out:
P
Peter Zijlstra 已提交
731
	return ret;
732 733
}

734
int tg_nop(struct task_group *tg, void *data)
P
Peter Zijlstra 已提交
735
{
736
	return 0;
P
Peter Zijlstra 已提交
737
}
738 739
#endif

740 741
static void set_load_weight(struct task_struct *p)
{
N
Nikhil Rao 已提交
742 743 744
	int prio = p->static_prio - MAX_RT_PRIO;
	struct load_weight *load = &p->se.load;

I
Ingo Molnar 已提交
745 746 747
	/*
	 * SCHED_IDLE tasks get minimal weight:
	 */
748
	if (idle_policy(p->policy)) {
749
		load->weight = scale_load(WEIGHT_IDLEPRIO);
N
Nikhil Rao 已提交
750
		load->inv_weight = WMULT_IDLEPRIO;
I
Ingo Molnar 已提交
751 752
		return;
	}
753

754 755
	load->weight = scale_load(sched_prio_to_weight[prio]);
	load->inv_weight = sched_prio_to_wmult[prio];
756 757
}

758
static inline void enqueue_task(struct rq *rq, struct task_struct *p, int flags)
759
{
760
	update_rq_clock(rq);
761 762
	if (!(flags & ENQUEUE_RESTORE))
		sched_info_queued(rq, p);
763
	p->sched_class->enqueue_task(rq, p, flags);
764 765
}

766
static inline void dequeue_task(struct rq *rq, struct task_struct *p, int flags)
767
{
768
	update_rq_clock(rq);
769 770
	if (!(flags & DEQUEUE_SAVE))
		sched_info_dequeued(rq, p);
771
	p->sched_class->dequeue_task(rq, p, flags);
772 773
}

774
void activate_task(struct rq *rq, struct task_struct *p, int flags)
775 776 777 778
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible--;

779
	enqueue_task(rq, p, flags);
780 781
}

782
void deactivate_task(struct rq *rq, struct task_struct *p, int flags)
783 784 785 786
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible++;

787
	dequeue_task(rq, p, flags);
788 789
}

790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819
void sched_set_stop_task(int cpu, struct task_struct *stop)
{
	struct sched_param param = { .sched_priority = MAX_RT_PRIO - 1 };
	struct task_struct *old_stop = cpu_rq(cpu)->stop;

	if (stop) {
		/*
		 * Make it appear like a SCHED_FIFO task, its something
		 * userspace knows about and won't get confused about.
		 *
		 * Also, it will make PI more or less work without too
		 * much confusion -- but then, stop work should not
		 * rely on PI working anyway.
		 */
		sched_setscheduler_nocheck(stop, SCHED_FIFO, &param);

		stop->sched_class = &stop_sched_class;
	}

	cpu_rq(cpu)->stop = stop;

	if (old_stop) {
		/*
		 * Reset it back to a normal scheduling class so that
		 * it can die in pieces.
		 */
		old_stop->sched_class = &rt_sched_class;
	}
}

820
/*
I
Ingo Molnar 已提交
821
 * __normal_prio - return the priority that is based on the static prio
822 823 824
 */
static inline int __normal_prio(struct task_struct *p)
{
I
Ingo Molnar 已提交
825
	return p->static_prio;
826 827
}

828 829 830 831 832 833 834
/*
 * Calculate the expected normal priority: i.e. priority
 * without taking RT-inheritance into account. Might be
 * boosted by interactivity modifiers. Changes upon fork,
 * setprio syscalls, and whenever the interactivity
 * estimator recalculates.
 */
835
static inline int normal_prio(struct task_struct *p)
836 837 838
{
	int prio;

839 840 841
	if (task_has_dl_policy(p))
		prio = MAX_DL_PRIO-1;
	else if (task_has_rt_policy(p))
842 843 844 845 846 847 848 849 850 851 852 853 854
		prio = MAX_RT_PRIO-1 - p->rt_priority;
	else
		prio = __normal_prio(p);
	return prio;
}

/*
 * Calculate the current priority, i.e. the priority
 * taken into account by the scheduler. This value might
 * be boosted by RT tasks, or might be boosted by
 * interactivity modifiers. Will be RT if the task got
 * RT-boosted. If not then it returns p->normal_prio.
 */
855
static int effective_prio(struct task_struct *p)
856 857 858 859 860 861 862 863 864 865 866 867
{
	p->normal_prio = normal_prio(p);
	/*
	 * If we are RT tasks or we were boosted to RT priority,
	 * keep the priority unchanged. Otherwise, update priority
	 * to the normal priority:
	 */
	if (!rt_prio(p->prio))
		return p->normal_prio;
	return p->prio;
}

L
Linus Torvalds 已提交
868 869 870
/**
 * task_curr - is this task currently executing on a CPU?
 * @p: the task in question.
871 872
 *
 * Return: 1 if the task is currently executing. 0 otherwise.
L
Linus Torvalds 已提交
873
 */
874
inline int task_curr(const struct task_struct *p)
L
Linus Torvalds 已提交
875 876 877 878
{
	return cpu_curr(task_cpu(p)) == p;
}

879
/*
880 881 882 883 884
 * switched_from, switched_to and prio_changed must _NOT_ drop rq->lock,
 * use the balance_callback list if you want balancing.
 *
 * this means any call to check_class_changed() must be followed by a call to
 * balance_callback().
885
 */
886 887
static inline void check_class_changed(struct rq *rq, struct task_struct *p,
				       const struct sched_class *prev_class,
P
Peter Zijlstra 已提交
888
				       int oldprio)
889 890 891
{
	if (prev_class != p->sched_class) {
		if (prev_class->switched_from)
P
Peter Zijlstra 已提交
892
			prev_class->switched_from(rq, p);
893

P
Peter Zijlstra 已提交
894
		p->sched_class->switched_to(rq, p);
895
	} else if (oldprio != p->prio || dl_task(p))
P
Peter Zijlstra 已提交
896
		p->sched_class->prio_changed(rq, p, oldprio);
897 898
}

899
void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags)
900 901 902 903 904 905 906 907 908 909
{
	const struct sched_class *class;

	if (p->sched_class == rq->curr->sched_class) {
		rq->curr->sched_class->check_preempt_curr(rq, p, flags);
	} else {
		for_each_class(class) {
			if (class == rq->curr->sched_class)
				break;
			if (class == p->sched_class) {
910
				resched_curr(rq);
911 912 913 914 915 916 917 918 919
				break;
			}
		}
	}

	/*
	 * A queue event has occurred, and we're going to schedule.  In
	 * this case, we can save a useless back to back clock update.
	 */
920
	if (task_on_rq_queued(rq->curr) && test_tsk_need_resched(rq->curr))
921
		rq_clock_skip_update(rq, true);
922 923
}

L
Linus Torvalds 已提交
924
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943
/*
 * This is how migration works:
 *
 * 1) we invoke migration_cpu_stop() on the target CPU using
 *    stop_one_cpu().
 * 2) stopper starts to run (implicitly forcing the migrated thread
 *    off the CPU)
 * 3) it checks whether the migrated task is still in the wrong runqueue.
 * 4) if it's in the wrong runqueue then the migration thread removes
 *    it and puts it into the right queue.
 * 5) stopper completes and stop_one_cpu() returns and the migration
 *    is done.
 */

/*
 * move_queued_task - move a queued task to new rq.
 *
 * Returns (locked) new rq. Old rq's lock is released.
 */
944
static struct rq *move_queued_task(struct rq *rq, struct task_struct *p, int new_cpu)
P
Peter Zijlstra 已提交
945 946 947 948
{
	lockdep_assert_held(&rq->lock);

	p->on_rq = TASK_ON_RQ_MIGRATING;
949
	dequeue_task(rq, p, 0);
P
Peter Zijlstra 已提交
950 951 952 953 954 955 956 957
	set_task_cpu(p, new_cpu);
	raw_spin_unlock(&rq->lock);

	rq = cpu_rq(new_cpu);

	raw_spin_lock(&rq->lock);
	BUG_ON(task_cpu(p) != new_cpu);
	enqueue_task(rq, p, 0);
958
	p->on_rq = TASK_ON_RQ_QUEUED;
P
Peter Zijlstra 已提交
959 960 961 962 963 964 965 966 967 968 969
	check_preempt_curr(rq, p, 0);

	return rq;
}

struct migration_arg {
	struct task_struct *task;
	int dest_cpu;
};

/*
I
Ingo Molnar 已提交
970
 * Move (not current) task off this CPU, onto the destination CPU. We're doing
P
Peter Zijlstra 已提交
971 972 973 974 975 976 977
 * this because either it can't run here any more (set_cpus_allowed()
 * away from this CPU, or CPU going down), or because we're
 * attempting to rebalance this task on exec (sched_exec).
 *
 * So we race with normal scheduler movements, but that's OK, as long
 * as the task is no longer on this CPU.
 */
978
static struct rq *__migrate_task(struct rq *rq, struct task_struct *p, int dest_cpu)
P
Peter Zijlstra 已提交
979 980
{
	if (unlikely(!cpu_active(dest_cpu)))
981
		return rq;
P
Peter Zijlstra 已提交
982 983 984

	/* Affinity changed (again). */
	if (!cpumask_test_cpu(dest_cpu, tsk_cpus_allowed(p)))
985
		return rq;
P
Peter Zijlstra 已提交
986

987 988 989
	rq = move_queued_task(rq, p, dest_cpu);

	return rq;
P
Peter Zijlstra 已提交
990 991 992 993 994 995 996 997 998 999
}

/*
 * migration_cpu_stop - this will be executed by a highprio stopper thread
 * and performs thread migration by bumping thread off CPU then
 * 'pushing' onto another runqueue.
 */
static int migration_cpu_stop(void *data)
{
	struct migration_arg *arg = data;
1000 1001
	struct task_struct *p = arg->task;
	struct rq *rq = this_rq();
P
Peter Zijlstra 已提交
1002 1003

	/*
I
Ingo Molnar 已提交
1004 1005
	 * The original target CPU might have gone down and we might
	 * be on another CPU but it doesn't matter.
P
Peter Zijlstra 已提交
1006 1007 1008 1009 1010 1011 1012 1013
	 */
	local_irq_disable();
	/*
	 * We need to explicitly wake pending tasks before running
	 * __migrate_task() such that we will not miss enforcing cpus_allowed
	 * during wakeups, see set_cpus_allowed_ptr()'s TASK_WAKING test.
	 */
	sched_ttwu_pending();
1014 1015 1016 1017 1018 1019 1020 1021

	raw_spin_lock(&p->pi_lock);
	raw_spin_lock(&rq->lock);
	/*
	 * If task_rq(p) != rq, it cannot be migrated here, because we're
	 * holding rq->lock, if p->on_rq == 0 it cannot get enqueued because
	 * we're holding p->pi_lock.
	 */
1022 1023 1024 1025 1026 1027
	if (task_rq(p) == rq) {
		if (task_on_rq_queued(p))
			rq = __migrate_task(rq, p, arg->dest_cpu);
		else
			p->wake_cpu = arg->dest_cpu;
	}
1028 1029 1030
	raw_spin_unlock(&rq->lock);
	raw_spin_unlock(&p->pi_lock);

P
Peter Zijlstra 已提交
1031 1032 1033 1034
	local_irq_enable();
	return 0;
}

1035 1036 1037 1038 1039
/*
 * sched_class::set_cpus_allowed must do the below, but is not required to
 * actually call this function.
 */
void set_cpus_allowed_common(struct task_struct *p, const struct cpumask *new_mask)
P
Peter Zijlstra 已提交
1040 1041 1042 1043 1044
{
	cpumask_copy(&p->cpus_allowed, new_mask);
	p->nr_cpus_allowed = cpumask_weight(new_mask);
}

1045 1046
void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
{
1047 1048 1049
	struct rq *rq = task_rq(p);
	bool queued, running;

1050
	lockdep_assert_held(&p->pi_lock);
1051 1052 1053 1054 1055 1056 1057 1058 1059 1060

	queued = task_on_rq_queued(p);
	running = task_current(rq, p);

	if (queued) {
		/*
		 * Because __kthread_bind() calls this on blocked tasks without
		 * holding rq->lock.
		 */
		lockdep_assert_held(&rq->lock);
1061
		dequeue_task(rq, p, DEQUEUE_SAVE);
1062 1063 1064 1065
	}
	if (running)
		put_prev_task(rq, p);

1066
	p->sched_class->set_cpus_allowed(p, new_mask);
1067 1068

	if (queued)
1069
		enqueue_task(rq, p, ENQUEUE_RESTORE);
1070
	if (running)
1071
		set_curr_task(rq, p);
1072 1073
}

P
Peter Zijlstra 已提交
1074 1075 1076 1077 1078 1079 1080 1081 1082
/*
 * Change a given task's CPU affinity. Migrate the thread to a
 * proper CPU and schedule it away if the CPU it's executing on
 * is removed from the allowed bitmask.
 *
 * NOTE: the caller must have a valid reference to the task, the
 * task must not exit() & deallocate itself prematurely. The
 * call is not atomic; no spinlocks may be held.
 */
1083 1084
static int __set_cpus_allowed_ptr(struct task_struct *p,
				  const struct cpumask *new_mask, bool check)
P
Peter Zijlstra 已提交
1085
{
1086
	const struct cpumask *cpu_valid_mask = cpu_active_mask;
P
Peter Zijlstra 已提交
1087
	unsigned int dest_cpu;
1088 1089
	struct rq_flags rf;
	struct rq *rq;
P
Peter Zijlstra 已提交
1090 1091
	int ret = 0;

1092
	rq = task_rq_lock(p, &rf);
1093
	update_rq_clock(rq);
P
Peter Zijlstra 已提交
1094

1095 1096 1097 1098 1099 1100 1101
	if (p->flags & PF_KTHREAD) {
		/*
		 * Kernel threads are allowed on online && !active CPUs
		 */
		cpu_valid_mask = cpu_online_mask;
	}

1102 1103 1104 1105 1106 1107 1108 1109 1110
	/*
	 * Must re-check here, to close a race against __kthread_bind(),
	 * sched_setaffinity() is not guaranteed to observe the flag.
	 */
	if (check && (p->flags & PF_NO_SETAFFINITY)) {
		ret = -EINVAL;
		goto out;
	}

P
Peter Zijlstra 已提交
1111 1112 1113
	if (cpumask_equal(&p->cpus_allowed, new_mask))
		goto out;

1114
	if (!cpumask_intersects(new_mask, cpu_valid_mask)) {
P
Peter Zijlstra 已提交
1115 1116 1117 1118 1119 1120
		ret = -EINVAL;
		goto out;
	}

	do_set_cpus_allowed(p, new_mask);

1121 1122 1123
	if (p->flags & PF_KTHREAD) {
		/*
		 * For kernel threads that do indeed end up on online &&
I
Ingo Molnar 已提交
1124
		 * !active we want to ensure they are strict per-CPU threads.
1125 1126 1127 1128 1129 1130
		 */
		WARN_ON(cpumask_intersects(new_mask, cpu_online_mask) &&
			!cpumask_intersects(new_mask, cpu_active_mask) &&
			p->nr_cpus_allowed != 1);
	}

P
Peter Zijlstra 已提交
1131 1132 1133 1134
	/* Can the task run on the task's current CPU? If so, we're done */
	if (cpumask_test_cpu(task_cpu(p), new_mask))
		goto out;

1135
	dest_cpu = cpumask_any_and(cpu_valid_mask, new_mask);
P
Peter Zijlstra 已提交
1136 1137 1138
	if (task_running(rq, p) || p->state == TASK_WAKING) {
		struct migration_arg arg = { p, dest_cpu };
		/* Need help from migration thread: drop lock and wait. */
1139
		task_rq_unlock(rq, p, &rf);
P
Peter Zijlstra 已提交
1140 1141 1142
		stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg);
		tlb_migrate_finish(p->mm);
		return 0;
1143 1144 1145 1146 1147
	} else if (task_on_rq_queued(p)) {
		/*
		 * OK, since we're going to drop the lock immediately
		 * afterwards anyway.
		 */
1148
		rq_unpin_lock(rq, &rf);
1149
		rq = move_queued_task(rq, p, dest_cpu);
1150
		rq_repin_lock(rq, &rf);
1151
	}
P
Peter Zijlstra 已提交
1152
out:
1153
	task_rq_unlock(rq, p, &rf);
P
Peter Zijlstra 已提交
1154 1155 1156

	return ret;
}
1157 1158 1159 1160 1161

int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
{
	return __set_cpus_allowed_ptr(p, new_mask, false);
}
P
Peter Zijlstra 已提交
1162 1163
EXPORT_SYMBOL_GPL(set_cpus_allowed_ptr);

I
Ingo Molnar 已提交
1164
void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
I
Ingo Molnar 已提交
1165
{
1166 1167 1168 1169 1170
#ifdef CONFIG_SCHED_DEBUG
	/*
	 * We should never call set_task_cpu() on a blocked task,
	 * ttwu() will sort out the placement.
	 */
P
Peter Zijlstra 已提交
1171
	WARN_ON_ONCE(p->state != TASK_RUNNING && p->state != TASK_WAKING &&
O
Oleg Nesterov 已提交
1172
			!p->on_rq);
1173

1174 1175 1176 1177 1178 1179 1180 1181 1182
	/*
	 * Migrating fair class task must have p->on_rq = TASK_ON_RQ_MIGRATING,
	 * because schedstat_wait_{start,end} rebase migrating task's wait_start
	 * time relying on p->on_rq.
	 */
	WARN_ON_ONCE(p->state == TASK_RUNNING &&
		     p->sched_class == &fair_sched_class &&
		     (p->on_rq && !task_on_rq_migrating(p)));

1183
#ifdef CONFIG_LOCKDEP
1184 1185 1186 1187 1188
	/*
	 * The caller should hold either p->pi_lock or rq->lock, when changing
	 * a task's CPU. ->pi_lock for waking tasks, rq->lock for runnable tasks.
	 *
	 * sched_move_task() holds both and thus holding either pins the cgroup,
P
Peter Zijlstra 已提交
1189
	 * see task_group().
1190 1191 1192 1193
	 *
	 * Furthermore, all task_rq users should acquire both locks, see
	 * task_rq_lock().
	 */
1194 1195 1196
	WARN_ON_ONCE(debug_locks && !(lockdep_is_held(&p->pi_lock) ||
				      lockdep_is_held(&task_rq(p)->lock)));
#endif
1197 1198
#endif

1199
	trace_sched_migrate_task(p, new_cpu);
1200

1201
	if (task_cpu(p) != new_cpu) {
1202
		if (p->sched_class->migrate_task_rq)
1203
			p->sched_class->migrate_task_rq(p);
1204
		p->se.nr_migrations++;
1205
		perf_event_task_migrate(p);
1206
	}
I
Ingo Molnar 已提交
1207 1208

	__set_task_cpu(p, new_cpu);
I
Ingo Molnar 已提交
1209 1210
}

1211 1212
static void __migrate_swap_task(struct task_struct *p, int cpu)
{
1213
	if (task_on_rq_queued(p)) {
1214 1215 1216 1217 1218
		struct rq *src_rq, *dst_rq;

		src_rq = task_rq(p);
		dst_rq = cpu_rq(cpu);

1219
		p->on_rq = TASK_ON_RQ_MIGRATING;
1220 1221 1222
		deactivate_task(src_rq, p, 0);
		set_task_cpu(p, cpu);
		activate_task(dst_rq, p, 0);
1223
		p->on_rq = TASK_ON_RQ_QUEUED;
1224 1225 1226 1227 1228
		check_preempt_curr(dst_rq, p, 0);
	} else {
		/*
		 * Task isn't running anymore; make it appear like we migrated
		 * it before it went to sleep. This means on wakeup we make the
I
Ingo Molnar 已提交
1229
		 * previous CPU our target instead of where it really is.
1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245
		 */
		p->wake_cpu = cpu;
	}
}

struct migration_swap_arg {
	struct task_struct *src_task, *dst_task;
	int src_cpu, dst_cpu;
};

static int migrate_swap_stop(void *data)
{
	struct migration_swap_arg *arg = data;
	struct rq *src_rq, *dst_rq;
	int ret = -EAGAIN;

1246 1247 1248
	if (!cpu_active(arg->src_cpu) || !cpu_active(arg->dst_cpu))
		return -EAGAIN;

1249 1250 1251
	src_rq = cpu_rq(arg->src_cpu);
	dst_rq = cpu_rq(arg->dst_cpu);

1252 1253
	double_raw_lock(&arg->src_task->pi_lock,
			&arg->dst_task->pi_lock);
1254
	double_rq_lock(src_rq, dst_rq);
1255

1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274
	if (task_cpu(arg->dst_task) != arg->dst_cpu)
		goto unlock;

	if (task_cpu(arg->src_task) != arg->src_cpu)
		goto unlock;

	if (!cpumask_test_cpu(arg->dst_cpu, tsk_cpus_allowed(arg->src_task)))
		goto unlock;

	if (!cpumask_test_cpu(arg->src_cpu, tsk_cpus_allowed(arg->dst_task)))
		goto unlock;

	__migrate_swap_task(arg->src_task, arg->dst_cpu);
	__migrate_swap_task(arg->dst_task, arg->src_cpu);

	ret = 0;

unlock:
	double_rq_unlock(src_rq, dst_rq);
1275 1276
	raw_spin_unlock(&arg->dst_task->pi_lock);
	raw_spin_unlock(&arg->src_task->pi_lock);
1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298

	return ret;
}

/*
 * Cross migrate two tasks
 */
int migrate_swap(struct task_struct *cur, struct task_struct *p)
{
	struct migration_swap_arg arg;
	int ret = -EINVAL;

	arg = (struct migration_swap_arg){
		.src_task = cur,
		.src_cpu = task_cpu(cur),
		.dst_task = p,
		.dst_cpu = task_cpu(p),
	};

	if (arg.src_cpu == arg.dst_cpu)
		goto out;

1299 1300 1301 1302
	/*
	 * These three tests are all lockless; this is OK since all of them
	 * will be re-checked with proper locks held further down the line.
	 */
1303 1304 1305 1306 1307 1308 1309 1310 1311
	if (!cpu_active(arg.src_cpu) || !cpu_active(arg.dst_cpu))
		goto out;

	if (!cpumask_test_cpu(arg.dst_cpu, tsk_cpus_allowed(arg.src_task)))
		goto out;

	if (!cpumask_test_cpu(arg.src_cpu, tsk_cpus_allowed(arg.dst_task)))
		goto out;

1312
	trace_sched_swap_numa(cur, arg.src_cpu, p, arg.dst_cpu);
1313 1314 1315 1316 1317 1318
	ret = stop_two_cpus(arg.dst_cpu, arg.src_cpu, migrate_swap_stop, &arg);

out:
	return ret;
}

L
Linus Torvalds 已提交
1319 1320 1321
/*
 * wait_task_inactive - wait for a thread to unschedule.
 *
R
Roland McGrath 已提交
1322 1323 1324 1325 1326 1327 1328
 * If @match_state is nonzero, it's the @p->state value just checked and
 * not expected to change.  If it changes, i.e. @p might have woken up,
 * then return zero.  When we succeed in waiting for @p to be off its CPU,
 * we return a positive number (its total switch count).  If a second call
 * a short while later returns the same number, the caller can be sure that
 * @p has remained unscheduled the whole time.
 *
L
Linus Torvalds 已提交
1329 1330 1331 1332 1333 1334
 * The caller must ensure that the task *will* unschedule sometime soon,
 * else this function might spin for a *long* time. This function can't
 * be called with interrupts off, or it may introduce deadlock with
 * smp_call_function() if an IPI is sent by the same process we are
 * waiting to become inactive.
 */
R
Roland McGrath 已提交
1335
unsigned long wait_task_inactive(struct task_struct *p, long match_state)
L
Linus Torvalds 已提交
1336
{
1337
	int running, queued;
1338
	struct rq_flags rf;
R
Roland McGrath 已提交
1339
	unsigned long ncsw;
1340
	struct rq *rq;
L
Linus Torvalds 已提交
1341

1342 1343 1344 1345 1346 1347 1348 1349
	for (;;) {
		/*
		 * We do the initial early heuristics without holding
		 * any task-queue locks at all. We'll only try to get
		 * the runqueue lock when things look like they will
		 * work out!
		 */
		rq = task_rq(p);
1350

1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361
		/*
		 * If the task is actively running on another CPU
		 * still, just relax and busy-wait without holding
		 * any locks.
		 *
		 * NOTE! Since we don't hold any locks, it's not
		 * even sure that "rq" stays as the right runqueue!
		 * But we don't care, since "task_running()" will
		 * return false if the runqueue has changed and p
		 * is actually now running somewhere else!
		 */
R
Roland McGrath 已提交
1362 1363 1364
		while (task_running(rq, p)) {
			if (match_state && unlikely(p->state != match_state))
				return 0;
1365
			cpu_relax();
R
Roland McGrath 已提交
1366
		}
1367

1368 1369 1370 1371 1372
		/*
		 * Ok, time to look more closely! We need the rq
		 * lock now, to be *sure*. If we're wrong, we'll
		 * just go back and repeat.
		 */
1373
		rq = task_rq_lock(p, &rf);
1374
		trace_sched_wait_task(p);
1375
		running = task_running(rq, p);
1376
		queued = task_on_rq_queued(p);
R
Roland McGrath 已提交
1377
		ncsw = 0;
1378
		if (!match_state || p->state == match_state)
1379
			ncsw = p->nvcsw | LONG_MIN; /* sets MSB */
1380
		task_rq_unlock(rq, p, &rf);
1381

R
Roland McGrath 已提交
1382 1383 1384 1385 1386 1387
		/*
		 * If it changed from the expected state, bail out now.
		 */
		if (unlikely(!ncsw))
			break;

1388 1389 1390 1391 1392 1393 1394 1395 1396 1397
		/*
		 * Was it really running after all now that we
		 * checked with the proper locks actually held?
		 *
		 * Oops. Go back and try again..
		 */
		if (unlikely(running)) {
			cpu_relax();
			continue;
		}
1398

1399 1400 1401 1402 1403
		/*
		 * It's not enough that it's not actively running,
		 * it must be off the runqueue _entirely_, and not
		 * preempted!
		 *
1404
		 * So if it was still runnable (but just not actively
1405 1406 1407
		 * running right now), it's preempted, and we should
		 * yield - it could be a while.
		 */
1408
		if (unlikely(queued)) {
T
Thomas Gleixner 已提交
1409
			ktime_t to = NSEC_PER_SEC / HZ;
1410 1411 1412

			set_current_state(TASK_UNINTERRUPTIBLE);
			schedule_hrtimeout(&to, HRTIMER_MODE_REL);
1413 1414
			continue;
		}
1415

1416 1417 1418 1419 1420 1421 1422
		/*
		 * Ahh, all good. It wasn't running, and it wasn't
		 * runnable, which means that it will never become
		 * running in the future either. We're all done!
		 */
		break;
	}
R
Roland McGrath 已提交
1423 1424

	return ncsw;
L
Linus Torvalds 已提交
1425 1426 1427 1428 1429 1430 1431 1432 1433
}

/***
 * kick_process - kick a running thread to enter/exit the kernel
 * @p: the to-be-kicked thread
 *
 * Cause a process which is running on another CPU to enter
 * kernel-mode, without any delay. (to get signals handled.)
 *
L
Lucas De Marchi 已提交
1434
 * NOTE: this function doesn't have to take the runqueue lock,
L
Linus Torvalds 已提交
1435 1436 1437 1438 1439
 * because all it wants to ensure is that the remote task enters
 * the kernel. If the IPI races and the task has been migrated
 * to another CPU then no harm is done and the purpose has been
 * achieved as well.
 */
1440
void kick_process(struct task_struct *p)
L
Linus Torvalds 已提交
1441 1442 1443 1444 1445 1446 1447 1448 1449
{
	int cpu;

	preempt_disable();
	cpu = task_cpu(p);
	if ((cpu != smp_processor_id()) && task_curr(p))
		smp_send_reschedule(cpu);
	preempt_enable();
}
R
Rusty Russell 已提交
1450
EXPORT_SYMBOL_GPL(kick_process);
L
Linus Torvalds 已提交
1451

1452
/*
1453
 * ->cpus_allowed is protected by both rq->lock and p->pi_lock
1454 1455 1456 1457 1458 1459 1460
 *
 * A few notes on cpu_active vs cpu_online:
 *
 *  - cpu_active must be a subset of cpu_online
 *
 *  - on cpu-up we allow per-cpu kthreads on the online && !active cpu,
 *    see __set_cpus_allowed_ptr(). At this point the newly online
I
Ingo Molnar 已提交
1461
 *    CPU isn't yet part of the sched domains, and balancing will not
1462 1463
 *    see it.
 *
I
Ingo Molnar 已提交
1464
 *  - on CPU-down we clear cpu_active() to mask the sched domains and
1465
 *    avoid the load balancer to place new tasks on the to be removed
I
Ingo Molnar 已提交
1466
 *    CPU. Existing tasks will remain running there and will be taken
1467 1468 1469 1470 1471 1472
 *    off.
 *
 * This means that fallback selection must not select !active CPUs.
 * And can assume that any active CPU must be online. Conversely
 * select_task_rq() below may allow selection of !active CPUs in order
 * to satisfy the above rules.
1473
 */
1474 1475
static int select_fallback_rq(int cpu, struct task_struct *p)
{
1476 1477
	int nid = cpu_to_node(cpu);
	const struct cpumask *nodemask = NULL;
1478 1479
	enum { cpuset, possible, fail } state = cpuset;
	int dest_cpu;
1480

1481
	/*
I
Ingo Molnar 已提交
1482 1483 1484
	 * If the node that the CPU is on has been offlined, cpu_to_node()
	 * will return -1. There is no CPU on the node, and we should
	 * select the CPU on the other node.
1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495
	 */
	if (nid != -1) {
		nodemask = cpumask_of_node(nid);

		/* Look for allowed, online CPU in same node. */
		for_each_cpu(dest_cpu, nodemask) {
			if (!cpu_active(dest_cpu))
				continue;
			if (cpumask_test_cpu(dest_cpu, tsk_cpus_allowed(p)))
				return dest_cpu;
		}
1496
	}
1497

1498 1499
	for (;;) {
		/* Any allowed, online CPU? */
1500
		for_each_cpu(dest_cpu, tsk_cpus_allowed(p)) {
1501 1502 1503
			if (!(p->flags & PF_KTHREAD) && !cpu_active(dest_cpu))
				continue;
			if (!cpu_online(dest_cpu))
1504 1505 1506
				continue;
			goto out;
		}
1507

1508
		/* No more Mr. Nice Guy. */
1509 1510
		switch (state) {
		case cpuset:
1511 1512 1513 1514 1515
			if (IS_ENABLED(CONFIG_CPUSETS)) {
				cpuset_cpus_allowed_fallback(p);
				state = possible;
				break;
			}
I
Ingo Molnar 已提交
1516
			/* Fall-through */
1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535
		case possible:
			do_set_cpus_allowed(p, cpu_possible_mask);
			state = fail;
			break;

		case fail:
			BUG();
			break;
		}
	}

out:
	if (state != cpuset) {
		/*
		 * Don't tell them about moving exiting tasks or
		 * kernel threads (both mm NULL), since they never
		 * leave kernel.
		 */
		if (p->mm && printk_ratelimit()) {
1536
			printk_deferred("process %d (%s) no longer affine to cpu%d\n",
1537 1538
					task_pid_nr(p), p->comm, cpu);
		}
1539 1540 1541 1542 1543
	}

	return dest_cpu;
}

1544
/*
1545
 * The caller (fork, wakeup) owns p->pi_lock, ->cpus_allowed is stable.
1546
 */
1547
static inline
1548
int select_task_rq(struct task_struct *p, int cpu, int sd_flags, int wake_flags)
1549
{
1550 1551
	lockdep_assert_held(&p->pi_lock);

1552
	if (tsk_nr_cpus_allowed(p) > 1)
1553
		cpu = p->sched_class->select_task_rq(p, cpu, sd_flags, wake_flags);
1554 1555
	else
		cpu = cpumask_any(tsk_cpus_allowed(p));
1556 1557 1558 1559

	/*
	 * In order not to call set_task_cpu() on a blocking task we need
	 * to rely on ttwu() to place the task on a valid ->cpus_allowed
I
Ingo Molnar 已提交
1560
	 * CPU.
1561 1562 1563 1564 1565 1566
	 *
	 * Since this is common to all placement strategies, this lives here.
	 *
	 * [ this allows ->select_task() to simply return task_cpu(p) and
	 *   not worry about this generic constraint ]
	 */
1567
	if (unlikely(!cpumask_test_cpu(cpu, tsk_cpus_allowed(p)) ||
P
Peter Zijlstra 已提交
1568
		     !cpu_online(cpu)))
1569
		cpu = select_fallback_rq(task_cpu(p), p);
1570 1571

	return cpu;
1572
}
1573 1574 1575 1576 1577 1578

static void update_avg(u64 *avg, u64 sample)
{
	s64 diff = sample - *avg;
	*avg += diff >> 3;
}
1579 1580 1581 1582 1583 1584 1585 1586 1587

#else

static inline int __set_cpus_allowed_ptr(struct task_struct *p,
					 const struct cpumask *new_mask, bool check)
{
	return set_cpus_allowed_ptr(p, new_mask);
}

P
Peter Zijlstra 已提交
1588
#endif /* CONFIG_SMP */
1589

P
Peter Zijlstra 已提交
1590
static void
1591
ttwu_stat(struct task_struct *p, int cpu, int wake_flags)
T
Tejun Heo 已提交
1592
{
1593
	struct rq *rq;
1594

1595 1596 1597 1598
	if (!schedstat_enabled())
		return;

	rq = this_rq();
P
Peter Zijlstra 已提交
1599

1600 1601
#ifdef CONFIG_SMP
	if (cpu == rq->cpu) {
1602 1603
		schedstat_inc(rq->ttwu_local);
		schedstat_inc(p->se.statistics.nr_wakeups_local);
P
Peter Zijlstra 已提交
1604 1605 1606
	} else {
		struct sched_domain *sd;

1607
		schedstat_inc(p->se.statistics.nr_wakeups_remote);
1608
		rcu_read_lock();
1609
		for_each_domain(rq->cpu, sd) {
P
Peter Zijlstra 已提交
1610
			if (cpumask_test_cpu(cpu, sched_domain_span(sd))) {
1611
				schedstat_inc(sd->ttwu_wake_remote);
P
Peter Zijlstra 已提交
1612 1613 1614
				break;
			}
		}
1615
		rcu_read_unlock();
P
Peter Zijlstra 已提交
1616
	}
1617 1618

	if (wake_flags & WF_MIGRATED)
1619
		schedstat_inc(p->se.statistics.nr_wakeups_migrate);
P
Peter Zijlstra 已提交
1620 1621
#endif /* CONFIG_SMP */

1622 1623
	schedstat_inc(rq->ttwu_count);
	schedstat_inc(p->se.statistics.nr_wakeups);
P
Peter Zijlstra 已提交
1624 1625

	if (wake_flags & WF_SYNC)
1626
		schedstat_inc(p->se.statistics.nr_wakeups_sync);
P
Peter Zijlstra 已提交
1627 1628
}

1629
static inline void ttwu_activate(struct rq *rq, struct task_struct *p, int en_flags)
P
Peter Zijlstra 已提交
1630
{
T
Tejun Heo 已提交
1631
	activate_task(rq, p, en_flags);
1632
	p->on_rq = TASK_ON_RQ_QUEUED;
1633

I
Ingo Molnar 已提交
1634
	/* If a worker is waking up, notify the workqueue: */
1635 1636
	if (p->flags & PF_WQ_WORKER)
		wq_worker_waking_up(p, cpu_of(rq));
T
Tejun Heo 已提交
1637 1638
}

1639 1640 1641
/*
 * Mark the task runnable and perform wakeup-preemption.
 */
1642
static void ttwu_do_wakeup(struct rq *rq, struct task_struct *p, int wake_flags,
1643
			   struct rq_flags *rf)
T
Tejun Heo 已提交
1644 1645 1646
{
	check_preempt_curr(rq, p, wake_flags);
	p->state = TASK_RUNNING;
1647 1648
	trace_sched_wakeup(p);

T
Tejun Heo 已提交
1649
#ifdef CONFIG_SMP
1650 1651
	if (p->sched_class->task_woken) {
		/*
1652 1653
		 * Our task @p is fully woken up and running; so its safe to
		 * drop the rq->lock, hereafter rq is only used for statistics.
1654
		 */
1655
		rq_unpin_lock(rq, rf);
T
Tejun Heo 已提交
1656
		p->sched_class->task_woken(rq, p);
1657
		rq_repin_lock(rq, rf);
1658
	}
T
Tejun Heo 已提交
1659

1660
	if (rq->idle_stamp) {
1661
		u64 delta = rq_clock(rq) - rq->idle_stamp;
1662
		u64 max = 2*rq->max_idle_balance_cost;
T
Tejun Heo 已提交
1663

1664 1665 1666
		update_avg(&rq->avg_idle, delta);

		if (rq->avg_idle > max)
T
Tejun Heo 已提交
1667
			rq->avg_idle = max;
1668

T
Tejun Heo 已提交
1669 1670 1671 1672 1673
		rq->idle_stamp = 0;
	}
#endif
}

1674
static void
1675
ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags,
1676
		 struct rq_flags *rf)
1677
{
1678 1679
	int en_flags = ENQUEUE_WAKEUP;

1680 1681
	lockdep_assert_held(&rq->lock);

1682 1683 1684
#ifdef CONFIG_SMP
	if (p->sched_contributes_to_load)
		rq->nr_uninterruptible--;
1685 1686

	if (wake_flags & WF_MIGRATED)
1687
		en_flags |= ENQUEUE_MIGRATED;
1688 1689
#endif

1690
	ttwu_activate(rq, p, en_flags);
1691
	ttwu_do_wakeup(rq, p, wake_flags, rf);
1692 1693 1694 1695 1696 1697 1698 1699 1700 1701
}

/*
 * Called in case the task @p isn't fully descheduled from its runqueue,
 * in this case we must do a remote wakeup. Its a 'light' wakeup though,
 * since all we need to do is flip p->state to TASK_RUNNING, since
 * the task is still ->on_rq.
 */
static int ttwu_remote(struct task_struct *p, int wake_flags)
{
1702
	struct rq_flags rf;
1703 1704 1705
	struct rq *rq;
	int ret = 0;

1706
	rq = __task_rq_lock(p, &rf);
1707
	if (task_on_rq_queued(p)) {
1708 1709
		/* check_preempt_curr() may use rq clock */
		update_rq_clock(rq);
1710
		ttwu_do_wakeup(rq, p, wake_flags, &rf);
1711 1712
		ret = 1;
	}
1713
	__task_rq_unlock(rq, &rf);
1714 1715 1716 1717

	return ret;
}

1718
#ifdef CONFIG_SMP
1719
void sched_ttwu_pending(void)
1720 1721
{
	struct rq *rq = this_rq();
P
Peter Zijlstra 已提交
1722 1723
	struct llist_node *llist = llist_del_all(&rq->wake_list);
	struct task_struct *p;
1724
	unsigned long flags;
1725
	struct rq_flags rf;
1726

1727 1728 1729 1730
	if (!llist)
		return;

	raw_spin_lock_irqsave(&rq->lock, flags);
1731
	rq_pin_lock(rq, &rf);
1732

P
Peter Zijlstra 已提交
1733
	while (llist) {
P
Peter Zijlstra 已提交
1734 1735
		int wake_flags = 0;

P
Peter Zijlstra 已提交
1736 1737
		p = llist_entry(llist, struct task_struct, wake_entry);
		llist = llist_next(llist);
P
Peter Zijlstra 已提交
1738 1739 1740 1741

		if (p->sched_remote_wakeup)
			wake_flags = WF_MIGRATED;

1742
		ttwu_do_activate(rq, p, wake_flags, &rf);
1743 1744
	}

1745
	rq_unpin_lock(rq, &rf);
1746
	raw_spin_unlock_irqrestore(&rq->lock, flags);
1747 1748 1749 1750
}

void scheduler_ipi(void)
{
1751 1752 1753 1754 1755
	/*
	 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
	 * TIF_NEED_RESCHED remotely (for the first time) will also send
	 * this IPI.
	 */
1756
	preempt_fold_need_resched();
1757

1758
	if (llist_empty(&this_rq()->wake_list) && !got_nohz_idle_kick())
1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774
		return;

	/*
	 * Not all reschedule IPI handlers call irq_enter/irq_exit, since
	 * traditionally all their work was done from the interrupt return
	 * path. Now that we actually do some work, we need to make sure
	 * we do call them.
	 *
	 * Some archs already do call them, luckily irq_enter/exit nest
	 * properly.
	 *
	 * Arguably we should visit all archs and update all handlers,
	 * however a fair share of IPIs are still resched only so this would
	 * somewhat pessimize the simple resched case.
	 */
	irq_enter();
P
Peter Zijlstra 已提交
1775
	sched_ttwu_pending();
1776 1777 1778 1779

	/*
	 * Check if someone kicked us for doing the nohz idle load balance.
	 */
1780
	if (unlikely(got_nohz_idle_kick())) {
1781
		this_rq()->idle_balance = 1;
1782
		raise_softirq_irqoff(SCHED_SOFTIRQ);
1783
	}
1784
	irq_exit();
1785 1786
}

P
Peter Zijlstra 已提交
1787
static void ttwu_queue_remote(struct task_struct *p, int cpu, int wake_flags)
1788
{
1789 1790
	struct rq *rq = cpu_rq(cpu);

P
Peter Zijlstra 已提交
1791 1792
	p->sched_remote_wakeup = !!(wake_flags & WF_MIGRATED);

1793 1794 1795 1796 1797 1798
	if (llist_add(&p->wake_entry, &cpu_rq(cpu)->wake_list)) {
		if (!set_nr_if_polling(rq->idle))
			smp_send_reschedule(cpu);
		else
			trace_sched_wake_idle_without_ipi(cpu);
	}
1799
}
1800

1801 1802 1803 1804 1805
void wake_up_if_idle(int cpu)
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

1806 1807 1808 1809
	rcu_read_lock();

	if (!is_idle_task(rcu_dereference(rq->curr)))
		goto out;
1810 1811 1812 1813 1814 1815 1816

	if (set_nr_if_polling(rq->idle)) {
		trace_sched_wake_idle_without_ipi(cpu);
	} else {
		raw_spin_lock_irqsave(&rq->lock, flags);
		if (is_idle_task(rq->curr))
			smp_send_reschedule(cpu);
I
Ingo Molnar 已提交
1817
		/* Else CPU is not idle, do nothing here: */
1818 1819
		raw_spin_unlock_irqrestore(&rq->lock, flags);
	}
1820 1821 1822

out:
	rcu_read_unlock();
1823 1824
}

1825
bool cpus_share_cache(int this_cpu, int that_cpu)
1826 1827 1828
{
	return per_cpu(sd_llc_id, this_cpu) == per_cpu(sd_llc_id, that_cpu);
}
1829
#endif /* CONFIG_SMP */
1830

1831
static void ttwu_queue(struct task_struct *p, int cpu, int wake_flags)
1832 1833
{
	struct rq *rq = cpu_rq(cpu);
1834
	struct rq_flags rf;
1835

1836
#if defined(CONFIG_SMP)
1837
	if (sched_feat(TTWU_QUEUE) && !cpus_share_cache(smp_processor_id(), cpu)) {
I
Ingo Molnar 已提交
1838
		sched_clock_cpu(cpu); /* Sync clocks across CPUs */
P
Peter Zijlstra 已提交
1839
		ttwu_queue_remote(p, cpu, wake_flags);
1840 1841 1842 1843
		return;
	}
#endif

1844
	raw_spin_lock(&rq->lock);
1845 1846 1847
	rq_pin_lock(rq, &rf);
	ttwu_do_activate(rq, p, wake_flags, &rf);
	rq_unpin_lock(rq, &rf);
1848
	raw_spin_unlock(&rq->lock);
T
Tejun Heo 已提交
1849 1850
}

1851 1852 1853 1854 1855 1856
/*
 * Notes on Program-Order guarantees on SMP systems.
 *
 *  MIGRATION
 *
 * The basic program-order guarantee on SMP systems is that when a task [t]
I
Ingo Molnar 已提交
1857 1858
 * migrates, all its activity on its old CPU [c0] happens-before any subsequent
 * execution on its new CPU [c1].
1859 1860 1861 1862 1863 1864 1865 1866 1867 1868
 *
 * For migration (of runnable tasks) this is provided by the following means:
 *
 *  A) UNLOCK of the rq(c0)->lock scheduling out task t
 *  B) migration for t is required to synchronize *both* rq(c0)->lock and
 *     rq(c1)->lock (if not at the same time, then in that order).
 *  C) LOCK of the rq(c1)->lock scheduling in task
 *
 * Transitivity guarantees that B happens after A and C after B.
 * Note: we only require RCpc transitivity.
I
Ingo Molnar 已提交
1869
 * Note: the CPU doing B need not be c0 or c1
1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900
 *
 * Example:
 *
 *   CPU0            CPU1            CPU2
 *
 *   LOCK rq(0)->lock
 *   sched-out X
 *   sched-in Y
 *   UNLOCK rq(0)->lock
 *
 *                                   LOCK rq(0)->lock // orders against CPU0
 *                                   dequeue X
 *                                   UNLOCK rq(0)->lock
 *
 *                                   LOCK rq(1)->lock
 *                                   enqueue X
 *                                   UNLOCK rq(1)->lock
 *
 *                   LOCK rq(1)->lock // orders against CPU2
 *                   sched-out Z
 *                   sched-in X
 *                   UNLOCK rq(1)->lock
 *
 *
 *  BLOCKING -- aka. SLEEP + WAKEUP
 *
 * For blocking we (obviously) need to provide the same guarantee as for
 * migration. However the means are completely different as there is no lock
 * chain to provide order. Instead we do:
 *
 *   1) smp_store_release(X->on_cpu, 0)
1901
 *   2) smp_cond_load_acquire(!X->on_cpu)
1902 1903 1904 1905 1906 1907 1908 1909 1910 1911
 *
 * Example:
 *
 *   CPU0 (schedule)  CPU1 (try_to_wake_up) CPU2 (schedule)
 *
 *   LOCK rq(0)->lock LOCK X->pi_lock
 *   dequeue X
 *   sched-out X
 *   smp_store_release(X->on_cpu, 0);
 *
1912
 *                    smp_cond_load_acquire(&X->on_cpu, !VAL);
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
 *                    X->state = WAKING
 *                    set_task_cpu(X,2)
 *
 *                    LOCK rq(2)->lock
 *                    enqueue X
 *                    X->state = RUNNING
 *                    UNLOCK rq(2)->lock
 *
 *                                          LOCK rq(2)->lock // orders against CPU1
 *                                          sched-out Z
 *                                          sched-in X
 *                                          UNLOCK rq(2)->lock
 *
 *                    UNLOCK X->pi_lock
 *   UNLOCK rq(0)->lock
 *
 *
 * However; for wakeups there is a second guarantee we must provide, namely we
 * must observe the state that lead to our wakeup. That is, not only must our
 * task observe its own prior state, it must also observe the stores prior to
 * its wakeup.
 *
 * This means that any means of doing remote wakeups must order the CPU doing
 * the wakeup against the CPU the task is going to end up running on. This,
 * however, is already required for the regular Program-Order guarantee above,
1938
 * since the waking CPU is the one issueing the ACQUIRE (smp_cond_load_acquire).
1939 1940 1941
 *
 */

T
Tejun Heo 已提交
1942
/**
L
Linus Torvalds 已提交
1943
 * try_to_wake_up - wake up a thread
T
Tejun Heo 已提交
1944
 * @p: the thread to be awakened
L
Linus Torvalds 已提交
1945
 * @state: the mask of task states that can be woken
T
Tejun Heo 已提交
1946
 * @wake_flags: wake modifier flags (WF_*)
L
Linus Torvalds 已提交
1947
 *
1948
 * If (@state & @p->state) @p->state = TASK_RUNNING.
L
Linus Torvalds 已提交
1949
 *
1950 1951 1952 1953 1954 1955 1956
 * If the task was not queued/runnable, also place it back on a runqueue.
 *
 * Atomic against schedule() which would dequeue a task, also see
 * set_current_state().
 *
 * Return: %true if @p->state changes (an actual wakeup was done),
 *	   %false otherwise.
L
Linus Torvalds 已提交
1957
 */
1958 1959
static int
try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
L
Linus Torvalds 已提交
1960 1961
{
	unsigned long flags;
1962
	int cpu, success = 0;
P
Peter Zijlstra 已提交
1963

1964 1965 1966 1967 1968 1969 1970
	/*
	 * If we are going to wake up a thread waiting for CONDITION we
	 * need to ensure that CONDITION=1 done by the caller can not be
	 * reordered with p->state check below. This pairs with mb() in
	 * set_current_state() the waiting thread does.
	 */
	smp_mb__before_spinlock();
1971
	raw_spin_lock_irqsave(&p->pi_lock, flags);
P
Peter Zijlstra 已提交
1972
	if (!(p->state & state))
L
Linus Torvalds 已提交
1973 1974
		goto out;

1975 1976
	trace_sched_waking(p);

I
Ingo Molnar 已提交
1977 1978
	/* We're going to change ->state: */
	success = 1;
L
Linus Torvalds 已提交
1979 1980
	cpu = task_cpu(p);

1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002
	/*
	 * Ensure we load p->on_rq _after_ p->state, otherwise it would
	 * be possible to, falsely, observe p->on_rq == 0 and get stuck
	 * in smp_cond_load_acquire() below.
	 *
	 * sched_ttwu_pending()                 try_to_wake_up()
	 *   [S] p->on_rq = 1;                  [L] P->state
	 *       UNLOCK rq->lock  -----.
	 *                              \
	 *				 +---   RMB
	 * schedule()                   /
	 *       LOCK rq->lock    -----'
	 *       UNLOCK rq->lock
	 *
	 * [task p]
	 *   [S] p->state = UNINTERRUPTIBLE     [L] p->on_rq
	 *
	 * Pairs with the UNLOCK+LOCK on rq->lock from the
	 * last wakeup of our task and the schedule that got our task
	 * current.
	 */
	smp_rmb();
2003 2004
	if (p->on_rq && ttwu_remote(p, wake_flags))
		goto stat;
L
Linus Torvalds 已提交
2005 2006

#ifdef CONFIG_SMP
2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025
	/*
	 * Ensure we load p->on_cpu _after_ p->on_rq, otherwise it would be
	 * possible to, falsely, observe p->on_cpu == 0.
	 *
	 * One must be running (->on_cpu == 1) in order to remove oneself
	 * from the runqueue.
	 *
	 *  [S] ->on_cpu = 1;	[L] ->on_rq
	 *      UNLOCK rq->lock
	 *			RMB
	 *      LOCK   rq->lock
	 *  [S] ->on_rq = 0;    [L] ->on_cpu
	 *
	 * Pairs with the full barrier implied in the UNLOCK+LOCK on rq->lock
	 * from the consecutive calls to schedule(); the first switching to our
	 * task, the second putting it to sleep.
	 */
	smp_rmb();

P
Peter Zijlstra 已提交
2026
	/*
I
Ingo Molnar 已提交
2027
	 * If the owning (remote) CPU is still in the middle of schedule() with
2028
	 * this task as prev, wait until its done referencing the task.
2029 2030 2031 2032 2033
	 *
	 * Pairs with the smp_store_release() in finish_lock_switch().
	 *
	 * This ensures that tasks getting woken will be fully ordered against
	 * their previous state and preserve Program Order.
2034
	 */
2035
	smp_cond_load_acquire(&p->on_cpu, !VAL);
L
Linus Torvalds 已提交
2036

2037
	p->sched_contributes_to_load = !!task_contributes_to_load(p);
P
Peter Zijlstra 已提交
2038
	p->state = TASK_WAKING;
2039

2040 2041 2042 2043 2044
	if (p->in_iowait) {
		delayacct_blkio_end();
		atomic_dec(&task_rq(p)->nr_iowait);
	}

2045
	cpu = select_task_rq(p, p->wake_cpu, SD_BALANCE_WAKE, wake_flags);
2046 2047
	if (task_cpu(p) != cpu) {
		wake_flags |= WF_MIGRATED;
2048
		set_task_cpu(p, cpu);
2049
	}
2050 2051 2052 2053 2054 2055 2056 2057

#else /* CONFIG_SMP */

	if (p->in_iowait) {
		delayacct_blkio_end();
		atomic_dec(&task_rq(p)->nr_iowait);
	}

L
Linus Torvalds 已提交
2058 2059
#endif /* CONFIG_SMP */

2060
	ttwu_queue(p, cpu, wake_flags);
2061
stat:
2062
	ttwu_stat(p, cpu, wake_flags);
L
Linus Torvalds 已提交
2063
out:
2064
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
2065 2066 2067 2068

	return success;
}

T
Tejun Heo 已提交
2069 2070 2071
/**
 * try_to_wake_up_local - try to wake up a local task with rq lock held
 * @p: the thread to be awakened
2072
 * @cookie: context's cookie for pinning
T
Tejun Heo 已提交
2073
 *
2074
 * Put @p on the run-queue if it's not already there. The caller must
T
Tejun Heo 已提交
2075
 * ensure that this_rq() is locked, @p is bound to this_rq() and not
2076
 * the current task.
T
Tejun Heo 已提交
2077
 */
2078
static void try_to_wake_up_local(struct task_struct *p, struct rq_flags *rf)
T
Tejun Heo 已提交
2079 2080 2081
{
	struct rq *rq = task_rq(p);

2082 2083 2084 2085
	if (WARN_ON_ONCE(rq != this_rq()) ||
	    WARN_ON_ONCE(p == current))
		return;

T
Tejun Heo 已提交
2086 2087
	lockdep_assert_held(&rq->lock);

2088
	if (!raw_spin_trylock(&p->pi_lock)) {
2089 2090 2091 2092 2093 2094
		/*
		 * This is OK, because current is on_cpu, which avoids it being
		 * picked for load-balance and preemption/IRQs are still
		 * disabled avoiding further scheduler activity on it and we've
		 * not yet picked a replacement task.
		 */
2095
		rq_unpin_lock(rq, rf);
2096 2097 2098
		raw_spin_unlock(&rq->lock);
		raw_spin_lock(&p->pi_lock);
		raw_spin_lock(&rq->lock);
2099
		rq_repin_lock(rq, rf);
2100 2101
	}

T
Tejun Heo 已提交
2102
	if (!(p->state & TASK_NORMAL))
2103
		goto out;
T
Tejun Heo 已提交
2104

2105 2106
	trace_sched_waking(p);

2107 2108 2109 2110 2111
	if (!task_on_rq_queued(p)) {
		if (p->in_iowait) {
			delayacct_blkio_end();
			atomic_dec(&rq->nr_iowait);
		}
P
Peter Zijlstra 已提交
2112
		ttwu_activate(rq, p, ENQUEUE_WAKEUP);
2113
	}
P
Peter Zijlstra 已提交
2114

2115
	ttwu_do_wakeup(rq, p, 0, rf);
2116
	ttwu_stat(p, smp_processor_id(), 0);
2117 2118
out:
	raw_spin_unlock(&p->pi_lock);
T
Tejun Heo 已提交
2119 2120
}

2121 2122 2123 2124 2125
/**
 * wake_up_process - Wake up a specific process
 * @p: The process to be woken up.
 *
 * Attempt to wake up the nominated process and move it to the set of runnable
2126 2127 2128
 * processes.
 *
 * Return: 1 if the process was woken up, 0 if it was already running.
2129 2130 2131 2132
 *
 * It may be assumed that this function implies a write memory barrier before
 * changing the task state if and only if any tasks are woken up.
 */
2133
int wake_up_process(struct task_struct *p)
L
Linus Torvalds 已提交
2134
{
2135
	return try_to_wake_up(p, TASK_NORMAL, 0);
L
Linus Torvalds 已提交
2136 2137 2138
}
EXPORT_SYMBOL(wake_up_process);

2139
int wake_up_state(struct task_struct *p, unsigned int state)
L
Linus Torvalds 已提交
2140 2141 2142 2143
{
	return try_to_wake_up(p, state, 0);
}

2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155
/*
 * This function clears the sched_dl_entity static params.
 */
void __dl_clear_params(struct task_struct *p)
{
	struct sched_dl_entity *dl_se = &p->dl;

	dl_se->dl_runtime = 0;
	dl_se->dl_deadline = 0;
	dl_se->dl_period = 0;
	dl_se->flags = 0;
	dl_se->dl_bw = 0;
2156 2157 2158

	dl_se->dl_throttled = 0;
	dl_se->dl_yielded = 0;
2159 2160
}

L
Linus Torvalds 已提交
2161 2162 2163
/*
 * Perform scheduler related setup for a newly forked process p.
 * p is forked by current.
I
Ingo Molnar 已提交
2164 2165 2166
 *
 * __sched_fork() is basic setup used by init_idle() too:
 */
2167
static void __sched_fork(unsigned long clone_flags, struct task_struct *p)
I
Ingo Molnar 已提交
2168
{
P
Peter Zijlstra 已提交
2169 2170 2171
	p->on_rq			= 0;

	p->se.on_rq			= 0;
I
Ingo Molnar 已提交
2172 2173
	p->se.exec_start		= 0;
	p->se.sum_exec_runtime		= 0;
2174
	p->se.prev_sum_exec_runtime	= 0;
2175
	p->se.nr_migrations		= 0;
P
Peter Zijlstra 已提交
2176
	p->se.vruntime			= 0;
P
Peter Zijlstra 已提交
2177
	INIT_LIST_HEAD(&p->se.group_node);
I
Ingo Molnar 已提交
2178

2179 2180 2181 2182
#ifdef CONFIG_FAIR_GROUP_SCHED
	p->se.cfs_rq			= NULL;
#endif

I
Ingo Molnar 已提交
2183
#ifdef CONFIG_SCHEDSTATS
2184
	/* Even if schedstat is disabled, there should not be garbage */
2185
	memset(&p->se.statistics, 0, sizeof(p->se.statistics));
I
Ingo Molnar 已提交
2186
#endif
N
Nick Piggin 已提交
2187

2188
	RB_CLEAR_NODE(&p->dl.rb_node);
2189
	init_dl_task_timer(&p->dl);
2190
	__dl_clear_params(p);
2191

P
Peter Zijlstra 已提交
2192
	INIT_LIST_HEAD(&p->rt.run_list);
2193 2194 2195 2196
	p->rt.timeout		= 0;
	p->rt.time_slice	= sched_rr_timeslice;
	p->rt.on_rq		= 0;
	p->rt.on_list		= 0;
N
Nick Piggin 已提交
2197

2198 2199 2200
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&p->preempt_notifiers);
#endif
2201 2202 2203

#ifdef CONFIG_NUMA_BALANCING
	if (p->mm && atomic_read(&p->mm->mm_users) == 1) {
2204
		p->mm->numa_next_scan = jiffies + msecs_to_jiffies(sysctl_numa_balancing_scan_delay);
2205 2206 2207
		p->mm->numa_scan_seq = 0;
	}

2208 2209 2210 2211 2212
	if (clone_flags & CLONE_VM)
		p->numa_preferred_nid = current->numa_preferred_nid;
	else
		p->numa_preferred_nid = -1;

2213 2214
	p->node_stamp = 0ULL;
	p->numa_scan_seq = p->mm ? p->mm->numa_scan_seq : 0;
2215
	p->numa_scan_period = sysctl_numa_balancing_scan_delay;
2216
	p->numa_work.next = &p->numa_work;
2217
	p->numa_faults = NULL;
2218 2219
	p->last_task_numa_placement = 0;
	p->last_sum_exec_runtime = 0;
2220 2221

	p->numa_group = NULL;
2222
#endif /* CONFIG_NUMA_BALANCING */
I
Ingo Molnar 已提交
2223 2224
}

2225 2226
DEFINE_STATIC_KEY_FALSE(sched_numa_balancing);

2227
#ifdef CONFIG_NUMA_BALANCING
2228

2229 2230 2231
void set_numabalancing_state(bool enabled)
{
	if (enabled)
2232
		static_branch_enable(&sched_numa_balancing);
2233
	else
2234
		static_branch_disable(&sched_numa_balancing);
2235
}
2236 2237 2238 2239 2240 2241 2242

#ifdef CONFIG_PROC_SYSCTL
int sysctl_numa_balancing(struct ctl_table *table, int write,
			 void __user *buffer, size_t *lenp, loff_t *ppos)
{
	struct ctl_table t;
	int err;
2243
	int state = static_branch_likely(&sched_numa_balancing);
2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258

	if (write && !capable(CAP_SYS_ADMIN))
		return -EPERM;

	t = *table;
	t.data = &state;
	err = proc_dointvec_minmax(&t, write, buffer, lenp, ppos);
	if (err < 0)
		return err;
	if (write)
		set_numabalancing_state(state);
	return err;
}
#endif
#endif
I
Ingo Molnar 已提交
2259

2260 2261
#ifdef CONFIG_SCHEDSTATS

2262
DEFINE_STATIC_KEY_FALSE(sched_schedstats);
2263
static bool __initdata __sched_schedstats = false;
2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286

static void set_schedstats(bool enabled)
{
	if (enabled)
		static_branch_enable(&sched_schedstats);
	else
		static_branch_disable(&sched_schedstats);
}

void force_schedstat_enabled(void)
{
	if (!schedstat_enabled()) {
		pr_info("kernel profiling enabled schedstats, disable via kernel.sched_schedstats.\n");
		static_branch_enable(&sched_schedstats);
	}
}

static int __init setup_schedstats(char *str)
{
	int ret = 0;
	if (!str)
		goto out;

2287 2288 2289 2290 2291
	/*
	 * This code is called before jump labels have been set up, so we can't
	 * change the static branch directly just yet.  Instead set a temporary
	 * variable so init_schedstats() can do it later.
	 */
2292
	if (!strcmp(str, "enable")) {
2293
		__sched_schedstats = true;
2294 2295
		ret = 1;
	} else if (!strcmp(str, "disable")) {
2296
		__sched_schedstats = false;
2297 2298 2299 2300 2301 2302 2303 2304 2305 2306
		ret = 1;
	}
out:
	if (!ret)
		pr_warn("Unable to parse schedstats=\n");

	return ret;
}
__setup("schedstats=", setup_schedstats);

2307 2308 2309 2310 2311
static void __init init_schedstats(void)
{
	set_schedstats(__sched_schedstats);
}

2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331
#ifdef CONFIG_PROC_SYSCTL
int sysctl_schedstats(struct ctl_table *table, int write,
			 void __user *buffer, size_t *lenp, loff_t *ppos)
{
	struct ctl_table t;
	int err;
	int state = static_branch_likely(&sched_schedstats);

	if (write && !capable(CAP_SYS_ADMIN))
		return -EPERM;

	t = *table;
	t.data = &state;
	err = proc_dointvec_minmax(&t, write, buffer, lenp, ppos);
	if (err < 0)
		return err;
	if (write)
		set_schedstats(state);
	return err;
}
2332 2333 2334 2335
#endif /* CONFIG_PROC_SYSCTL */
#else  /* !CONFIG_SCHEDSTATS */
static inline void init_schedstats(void) {}
#endif /* CONFIG_SCHEDSTATS */
I
Ingo Molnar 已提交
2336 2337 2338 2339

/*
 * fork()/clone()-time setup:
 */
2340
int sched_fork(unsigned long clone_flags, struct task_struct *p)
I
Ingo Molnar 已提交
2341
{
2342
	unsigned long flags;
I
Ingo Molnar 已提交
2343 2344
	int cpu = get_cpu();

2345
	__sched_fork(clone_flags, p);
2346
	/*
2347
	 * We mark the process as NEW here. This guarantees that
2348 2349 2350
	 * nobody will actually run it, and a signal or other external
	 * event cannot wake it up and insert it on the runqueue either.
	 */
2351
	p->state = TASK_NEW;
I
Ingo Molnar 已提交
2352

2353 2354 2355 2356 2357
	/*
	 * Make sure we do not leak PI boosting priority to the child.
	 */
	p->prio = current->normal_prio;

2358 2359 2360 2361
	/*
	 * Revert to default priority/policy on fork if requested.
	 */
	if (unlikely(p->sched_reset_on_fork)) {
2362
		if (task_has_dl_policy(p) || task_has_rt_policy(p)) {
2363
			p->policy = SCHED_NORMAL;
2364
			p->static_prio = NICE_TO_PRIO(0);
2365 2366 2367 2368 2369 2370
			p->rt_priority = 0;
		} else if (PRIO_TO_NICE(p->static_prio) < 0)
			p->static_prio = NICE_TO_PRIO(0);

		p->prio = p->normal_prio = __normal_prio(p);
		set_load_weight(p);
2371

2372 2373 2374 2375 2376 2377
		/*
		 * We don't need the reset flag anymore after the fork. It has
		 * fulfilled its duty:
		 */
		p->sched_reset_on_fork = 0;
	}
2378

2379 2380 2381 2382 2383 2384
	if (dl_prio(p->prio)) {
		put_cpu();
		return -EAGAIN;
	} else if (rt_prio(p->prio)) {
		p->sched_class = &rt_sched_class;
	} else {
H
Hiroshi Shimamoto 已提交
2385
		p->sched_class = &fair_sched_class;
2386
	}
2387

2388
	init_entity_runnable_average(&p->se);
P
Peter Zijlstra 已提交
2389

2390 2391 2392 2393 2394 2395 2396
	/*
	 * The child is not yet in the pid-hash so no cgroup attach races,
	 * and the cgroup is pinned to this child due to cgroup_fork()
	 * is ran before sched_fork().
	 *
	 * Silence PROVE_RCU.
	 */
2397
	raw_spin_lock_irqsave(&p->pi_lock, flags);
2398
	/*
I
Ingo Molnar 已提交
2399
	 * We're setting the CPU for the first time, we don't migrate,
2400 2401 2402 2403 2404
	 * so use __set_task_cpu().
	 */
	__set_task_cpu(p, cpu);
	if (p->sched_class->task_fork)
		p->sched_class->task_fork(p);
2405
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
2406

2407
#ifdef CONFIG_SCHED_INFO
I
Ingo Molnar 已提交
2408
	if (likely(sched_info_on()))
2409
		memset(&p->sched_info, 0, sizeof(p->sched_info));
L
Linus Torvalds 已提交
2410
#endif
P
Peter Zijlstra 已提交
2411 2412
#if defined(CONFIG_SMP)
	p->on_cpu = 0;
2413
#endif
2414
	init_task_preempt_count(p);
2415
#ifdef CONFIG_SMP
2416
	plist_node_init(&p->pushable_tasks, MAX_PRIO);
2417
	RB_CLEAR_NODE(&p->pushable_dl_tasks);
2418
#endif
2419

N
Nick Piggin 已提交
2420
	put_cpu();
2421
	return 0;
L
Linus Torvalds 已提交
2422 2423
}

2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442
unsigned long to_ratio(u64 period, u64 runtime)
{
	if (runtime == RUNTIME_INF)
		return 1ULL << 20;

	/*
	 * Doing this here saves a lot of checks in all
	 * the calling paths, and returning zero seems
	 * safe for them anyway.
	 */
	if (period == 0)
		return 0;

	return div64_u64(runtime << 20, period);
}

#ifdef CONFIG_SMP
inline struct dl_bw *dl_bw_of(int i)
{
2443 2444
	RCU_LOCKDEP_WARN(!rcu_read_lock_sched_held(),
			 "sched RCU must be held");
2445 2446 2447
	return &cpu_rq(i)->rd->dl_bw;
}

2448
static inline int dl_bw_cpus(int i)
2449
{
2450 2451 2452
	struct root_domain *rd = cpu_rq(i)->rd;
	int cpus = 0;

2453 2454
	RCU_LOCKDEP_WARN(!rcu_read_lock_sched_held(),
			 "sched RCU must be held");
2455 2456 2457 2458
	for_each_cpu_and(i, rd->span, cpu_active_mask)
		cpus++;

	return cpus;
2459 2460 2461 2462 2463 2464 2465
}
#else
inline struct dl_bw *dl_bw_of(int i)
{
	return &cpu_rq(i)->dl.dl_bw;
}

2466
static inline int dl_bw_cpus(int i)
2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478
{
	return 1;
}
#endif

/*
 * We must be sure that accepting a new task (or allowing changing the
 * parameters of an existing one) is consistent with the bandwidth
 * constraints. If yes, this function also accordingly updates the currently
 * allocated bandwidth to reflect the new situation.
 *
 * This function is called while holding p's rq->lock.
2479 2480 2481
 *
 * XXX we should delay bw change until the task's 0-lag point, see
 * __setparam_dl().
2482 2483 2484 2485 2486 2487
 */
static int dl_overflow(struct task_struct *p, int policy,
		       const struct sched_attr *attr)
{

	struct dl_bw *dl_b = dl_bw_of(task_cpu(p));
2488
	u64 period = attr->sched_period ?: attr->sched_deadline;
2489 2490
	u64 runtime = attr->sched_runtime;
	u64 new_bw = dl_policy(policy) ? to_ratio(period, runtime) : 0;
2491
	int cpus, err = -1;
2492

2493 2494
	/* !deadline task may carry old deadline bandwidth */
	if (new_bw == p->dl.dl_bw && task_has_dl_policy(p))
2495 2496 2497 2498 2499 2500 2501 2502
		return 0;

	/*
	 * Either if a task, enters, leave, or stays -deadline but changes
	 * its parameters, we may need to update accordingly the total
	 * allocated bandwidth of the container.
	 */
	raw_spin_lock(&dl_b->lock);
2503
	cpus = dl_bw_cpus(task_cpu(p));
2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523
	if (dl_policy(policy) && !task_has_dl_policy(p) &&
	    !__dl_overflow(dl_b, cpus, 0, new_bw)) {
		__dl_add(dl_b, new_bw);
		err = 0;
	} else if (dl_policy(policy) && task_has_dl_policy(p) &&
		   !__dl_overflow(dl_b, cpus, p->dl.dl_bw, new_bw)) {
		__dl_clear(dl_b, p->dl.dl_bw);
		__dl_add(dl_b, new_bw);
		err = 0;
	} else if (!dl_policy(policy) && task_has_dl_policy(p)) {
		__dl_clear(dl_b, p->dl.dl_bw);
		err = 0;
	}
	raw_spin_unlock(&dl_b->lock);

	return err;
}

extern void init_dl_bw(struct dl_bw *dl_b);

L
Linus Torvalds 已提交
2524 2525 2526 2527 2528 2529 2530
/*
 * wake_up_new_task - wake up a newly created task for the first time.
 *
 * This function will do some initial scheduler statistics housekeeping
 * that must be done for every newly created context, then puts the task
 * on the runqueue and wakes it.
 */
2531
void wake_up_new_task(struct task_struct *p)
L
Linus Torvalds 已提交
2532
{
2533
	struct rq_flags rf;
I
Ingo Molnar 已提交
2534
	struct rq *rq;
2535

2536
	raw_spin_lock_irqsave(&p->pi_lock, rf.flags);
2537
	p->state = TASK_RUNNING;
2538 2539 2540 2541
#ifdef CONFIG_SMP
	/*
	 * Fork balancing, do it here and not earlier because:
	 *  - cpus_allowed can change in the fork path
I
Ingo Molnar 已提交
2542
	 *  - any previously selected CPU might disappear through hotplug
2543 2544 2545
	 *
	 * Use __set_task_cpu() to avoid calling sched_class::migrate_task_rq,
	 * as we're not fully set-up yet.
2546
	 */
2547
	__set_task_cpu(p, select_task_rq(p, task_cpu(p), SD_BALANCE_FORK, 0));
2548
#endif
2549
	rq = __task_rq_lock(p, &rf);
2550
	update_rq_clock(rq);
2551
	post_init_entity_util_avg(&p->se);
2552

P
Peter Zijlstra 已提交
2553
	activate_task(rq, p, 0);
2554
	p->on_rq = TASK_ON_RQ_QUEUED;
2555
	trace_sched_wakeup_new(p);
P
Peter Zijlstra 已提交
2556
	check_preempt_curr(rq, p, WF_FORK);
2557
#ifdef CONFIG_SMP
2558 2559 2560 2561 2562
	if (p->sched_class->task_woken) {
		/*
		 * Nothing relies on rq->lock after this, so its fine to
		 * drop it.
		 */
2563
		rq_unpin_lock(rq, &rf);
2564
		p->sched_class->task_woken(rq, p);
2565
		rq_repin_lock(rq, &rf);
2566
	}
2567
#endif
2568
	task_rq_unlock(rq, p, &rf);
L
Linus Torvalds 已提交
2569 2570
}

2571 2572
#ifdef CONFIG_PREEMPT_NOTIFIERS

2573 2574
static struct static_key preempt_notifier_key = STATIC_KEY_INIT_FALSE;

2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586
void preempt_notifier_inc(void)
{
	static_key_slow_inc(&preempt_notifier_key);
}
EXPORT_SYMBOL_GPL(preempt_notifier_inc);

void preempt_notifier_dec(void)
{
	static_key_slow_dec(&preempt_notifier_key);
}
EXPORT_SYMBOL_GPL(preempt_notifier_dec);

2587
/**
2588
 * preempt_notifier_register - tell me when current is being preempted & rescheduled
R
Randy Dunlap 已提交
2589
 * @notifier: notifier struct to register
2590 2591 2592
 */
void preempt_notifier_register(struct preempt_notifier *notifier)
{
2593 2594 2595
	if (!static_key_false(&preempt_notifier_key))
		WARN(1, "registering preempt_notifier while notifiers disabled\n");

2596 2597 2598 2599 2600 2601
	hlist_add_head(&notifier->link, &current->preempt_notifiers);
}
EXPORT_SYMBOL_GPL(preempt_notifier_register);

/**
 * preempt_notifier_unregister - no longer interested in preemption notifications
R
Randy Dunlap 已提交
2602
 * @notifier: notifier struct to unregister
2603
 *
2604
 * This is *not* safe to call from within a preemption notifier.
2605 2606 2607 2608 2609 2610 2611
 */
void preempt_notifier_unregister(struct preempt_notifier *notifier)
{
	hlist_del(&notifier->link);
}
EXPORT_SYMBOL_GPL(preempt_notifier_unregister);

2612
static void __fire_sched_in_preempt_notifiers(struct task_struct *curr)
2613 2614 2615
{
	struct preempt_notifier *notifier;

2616
	hlist_for_each_entry(notifier, &curr->preempt_notifiers, link)
2617 2618 2619
		notifier->ops->sched_in(notifier, raw_smp_processor_id());
}

2620 2621 2622 2623 2624 2625
static __always_inline void fire_sched_in_preempt_notifiers(struct task_struct *curr)
{
	if (static_key_false(&preempt_notifier_key))
		__fire_sched_in_preempt_notifiers(curr);
}

2626
static void
2627 2628
__fire_sched_out_preempt_notifiers(struct task_struct *curr,
				   struct task_struct *next)
2629 2630 2631
{
	struct preempt_notifier *notifier;

2632
	hlist_for_each_entry(notifier, &curr->preempt_notifiers, link)
2633 2634 2635
		notifier->ops->sched_out(notifier, next);
}

2636 2637 2638 2639 2640 2641 2642 2643
static __always_inline void
fire_sched_out_preempt_notifiers(struct task_struct *curr,
				 struct task_struct *next)
{
	if (static_key_false(&preempt_notifier_key))
		__fire_sched_out_preempt_notifiers(curr, next);
}

2644
#else /* !CONFIG_PREEMPT_NOTIFIERS */
2645

2646
static inline void fire_sched_in_preempt_notifiers(struct task_struct *curr)
2647 2648 2649
{
}

2650
static inline void
2651 2652 2653 2654 2655
fire_sched_out_preempt_notifiers(struct task_struct *curr,
				 struct task_struct *next)
{
}

2656
#endif /* CONFIG_PREEMPT_NOTIFIERS */
2657

2658 2659 2660
/**
 * prepare_task_switch - prepare to switch tasks
 * @rq: the runqueue preparing to switch
R
Randy Dunlap 已提交
2661
 * @prev: the current task that is being switched out
2662 2663 2664 2665 2666 2667 2668 2669 2670
 * @next: the task we are going to switch to.
 *
 * This is called with the rq lock held and interrupts off. It must
 * be paired with a subsequent finish_task_switch after the context
 * switch.
 *
 * prepare_task_switch sets up locking and calls architecture specific
 * hooks.
 */
2671 2672 2673
static inline void
prepare_task_switch(struct rq *rq, struct task_struct *prev,
		    struct task_struct *next)
2674
{
2675
	sched_info_switch(rq, prev, next);
2676
	perf_event_task_sched_out(prev, next);
2677
	fire_sched_out_preempt_notifiers(prev, next);
2678 2679 2680 2681
	prepare_lock_switch(rq, next);
	prepare_arch_switch(next);
}

L
Linus Torvalds 已提交
2682 2683 2684 2685
/**
 * finish_task_switch - clean up after a task-switch
 * @prev: the thread we just switched away from.
 *
2686 2687 2688 2689
 * finish_task_switch must be called after the context switch, paired
 * with a prepare_task_switch call before the context switch.
 * finish_task_switch will reconcile locking set up by prepare_task_switch,
 * and do any other architecture-specific cleanup actions.
L
Linus Torvalds 已提交
2690 2691
 *
 * Note that we may have delayed dropping an mm in context_switch(). If
I
Ingo Molnar 已提交
2692
 * so, we finish that here outside of the runqueue lock. (Doing it
L
Linus Torvalds 已提交
2693 2694
 * with the lock held can cause deadlocks; see schedule() for
 * details.)
2695 2696 2697 2698 2699
 *
 * The context switch have flipped the stack from under us and restored the
 * local variables which were saved when this task called schedule() in the
 * past. prev == current is still correct but we need to recalculate this_rq
 * because prev may have moved to another CPU.
L
Linus Torvalds 已提交
2700
 */
2701
static struct rq *finish_task_switch(struct task_struct *prev)
L
Linus Torvalds 已提交
2702 2703
	__releases(rq->lock)
{
2704
	struct rq *rq = this_rq();
L
Linus Torvalds 已提交
2705
	struct mm_struct *mm = rq->prev_mm;
O
Oleg Nesterov 已提交
2706
	long prev_state;
L
Linus Torvalds 已提交
2707

2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718
	/*
	 * The previous task will have left us with a preempt_count of 2
	 * because it left us after:
	 *
	 *	schedule()
	 *	  preempt_disable();			// 1
	 *	  __schedule()
	 *	    raw_spin_lock_irq(&rq->lock)	// 2
	 *
	 * Also, see FORK_PREEMPT_COUNT.
	 */
2719 2720 2721 2722
	if (WARN_ONCE(preempt_count() != 2*PREEMPT_DISABLE_OFFSET,
		      "corrupted preempt_count: %s/%d/0x%x\n",
		      current->comm, current->pid, preempt_count()))
		preempt_count_set(FORK_PREEMPT_COUNT);
2723

L
Linus Torvalds 已提交
2724 2725 2726 2727
	rq->prev_mm = NULL;

	/*
	 * A task struct has one reference for the use as "current".
2728
	 * If a task dies, then it sets TASK_DEAD in tsk->state and calls
O
Oleg Nesterov 已提交
2729 2730
	 * schedule one last time. The schedule call will never return, and
	 * the scheduled task must drop that reference.
2731 2732 2733 2734 2735
	 *
	 * We must observe prev->state before clearing prev->on_cpu (in
	 * finish_lock_switch), otherwise a concurrent wakeup can get prev
	 * running on another CPU and we could rave with its RUNNING -> DEAD
	 * transition, resulting in a double drop.
L
Linus Torvalds 已提交
2736
	 */
O
Oleg Nesterov 已提交
2737
	prev_state = prev->state;
2738
	vtime_task_switch(prev);
2739
	perf_event_task_sched_in(prev, current);
2740
	finish_lock_switch(rq, prev);
2741
	finish_arch_post_lock_switch();
S
Steven Rostedt 已提交
2742

2743
	fire_sched_in_preempt_notifiers(current);
L
Linus Torvalds 已提交
2744 2745
	if (mm)
		mmdrop(mm);
2746
	if (unlikely(prev_state == TASK_DEAD)) {
2747 2748 2749
		if (prev->sched_class->task_dead)
			prev->sched_class->task_dead(prev);

2750 2751 2752
		/*
		 * Remove function-return probe instances associated with this
		 * task and put them back on the free list.
I
Ingo Molnar 已提交
2753
		 */
2754
		kprobe_flush_task(prev);
2755 2756 2757 2758

		/* Task is done with its stack. */
		put_task_stack(prev);

L
Linus Torvalds 已提交
2759
		put_task_struct(prev);
2760
	}
2761

2762
	tick_nohz_task_switch();
2763
	return rq;
L
Linus Torvalds 已提交
2764 2765
}

2766 2767 2768
#ifdef CONFIG_SMP

/* rq->lock is NOT held, but preemption is disabled */
2769
static void __balance_callback(struct rq *rq)
2770
{
2771 2772 2773
	struct callback_head *head, *next;
	void (*func)(struct rq *rq);
	unsigned long flags;
2774

2775 2776 2777 2778 2779 2780 2781 2782
	raw_spin_lock_irqsave(&rq->lock, flags);
	head = rq->balance_callback;
	rq->balance_callback = NULL;
	while (head) {
		func = (void (*)(struct rq *))head->func;
		next = head->next;
		head->next = NULL;
		head = next;
2783

2784
		func(rq);
2785
	}
2786 2787 2788 2789 2790 2791 2792
	raw_spin_unlock_irqrestore(&rq->lock, flags);
}

static inline void balance_callback(struct rq *rq)
{
	if (unlikely(rq->balance_callback))
		__balance_callback(rq);
2793 2794 2795
}

#else
2796

2797
static inline void balance_callback(struct rq *rq)
2798
{
L
Linus Torvalds 已提交
2799 2800
}

2801 2802
#endif

L
Linus Torvalds 已提交
2803 2804 2805 2806
/**
 * schedule_tail - first thing a freshly forked thread must call.
 * @prev: the thread we just switched away from.
 */
2807
asmlinkage __visible void schedule_tail(struct task_struct *prev)
L
Linus Torvalds 已提交
2808 2809
	__releases(rq->lock)
{
2810
	struct rq *rq;
2811

2812 2813 2814 2815 2816 2817 2818 2819 2820
	/*
	 * New tasks start with FORK_PREEMPT_COUNT, see there and
	 * finish_task_switch() for details.
	 *
	 * finish_task_switch() will drop rq->lock() and lower preempt_count
	 * and the preempt_enable() will end up enabling preemption (on
	 * PREEMPT_COUNT kernels).
	 */

2821
	rq = finish_task_switch(prev);
2822
	balance_callback(rq);
2823
	preempt_enable();
2824

L
Linus Torvalds 已提交
2825
	if (current->set_child_tid)
2826
		put_user(task_pid_vnr(current), current->set_child_tid);
L
Linus Torvalds 已提交
2827 2828 2829
}

/*
2830
 * context_switch - switch to the new MM and the new thread's register state.
L
Linus Torvalds 已提交
2831
 */
2832
static __always_inline struct rq *
2833
context_switch(struct rq *rq, struct task_struct *prev,
2834
	       struct task_struct *next, struct rq_flags *rf)
L
Linus Torvalds 已提交
2835
{
I
Ingo Molnar 已提交
2836
	struct mm_struct *mm, *oldmm;
L
Linus Torvalds 已提交
2837

2838
	prepare_task_switch(rq, prev, next);
2839

I
Ingo Molnar 已提交
2840 2841
	mm = next->mm;
	oldmm = prev->active_mm;
2842 2843 2844 2845 2846
	/*
	 * For paravirt, this is coupled with an exit in switch_to to
	 * combine the page table reload and the switch backend into
	 * one hypercall.
	 */
2847
	arch_start_context_switch(prev);
2848

2849
	if (!mm) {
L
Linus Torvalds 已提交
2850
		next->active_mm = oldmm;
V
Vegard Nossum 已提交
2851
		mmgrab(oldmm);
L
Linus Torvalds 已提交
2852 2853
		enter_lazy_tlb(oldmm, next);
	} else
2854
		switch_mm_irqs_off(oldmm, mm, next);
L
Linus Torvalds 已提交
2855

2856
	if (!prev->mm) {
L
Linus Torvalds 已提交
2857 2858 2859
		prev->active_mm = NULL;
		rq->prev_mm = oldmm;
	}
2860

2861
	rq->clock_update_flags &= ~(RQCF_ACT_SKIP|RQCF_REQ_SKIP);
2862

2863 2864 2865 2866 2867 2868
	/*
	 * Since the runqueue lock will be released by the next
	 * task (which is an invalid locking op but in the case
	 * of the scheduler it's an obvious special-case), so we
	 * do an early lockdep release here:
	 */
2869
	rq_unpin_lock(rq, rf);
2870
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
L
Linus Torvalds 已提交
2871 2872 2873

	/* Here we just switch the register state and the stack. */
	switch_to(prev, next, prev);
I
Ingo Molnar 已提交
2874
	barrier();
2875 2876

	return finish_task_switch(prev);
L
Linus Torvalds 已提交
2877 2878 2879
}

/*
2880
 * nr_running and nr_context_switches:
L
Linus Torvalds 已提交
2881 2882
 *
 * externally visible scheduler statistics: current number of runnable
2883
 * threads, total number of context switches performed since bootup.
L
Linus Torvalds 已提交
2884 2885 2886 2887 2888 2889 2890 2891 2892
 */
unsigned long nr_running(void)
{
	unsigned long i, sum = 0;

	for_each_online_cpu(i)
		sum += cpu_rq(i)->nr_running;

	return sum;
2893
}
L
Linus Torvalds 已提交
2894

2895
/*
I
Ingo Molnar 已提交
2896
 * Check if only the current task is running on the CPU.
2897 2898 2899 2900 2901 2902 2903 2904 2905 2906
 *
 * Caution: this function does not check that the caller has disabled
 * preemption, thus the result might have a time-of-check-to-time-of-use
 * race.  The caller is responsible to use it correctly, for example:
 *
 * - from a non-preemptable section (of course)
 *
 * - from a thread that is bound to a single CPU
 *
 * - in a loop with very short iterations (e.g. a polling loop)
2907 2908 2909
 */
bool single_task_running(void)
{
2910
	return raw_rq()->nr_running == 1;
2911 2912 2913
}
EXPORT_SYMBOL(single_task_running);

L
Linus Torvalds 已提交
2914
unsigned long long nr_context_switches(void)
2915
{
2916 2917
	int i;
	unsigned long long sum = 0;
2918

2919
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2920
		sum += cpu_rq(i)->nr_switches;
2921

L
Linus Torvalds 已提交
2922 2923
	return sum;
}
2924

2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954
/*
 * IO-wait accounting, and how its mostly bollocks (on SMP).
 *
 * The idea behind IO-wait account is to account the idle time that we could
 * have spend running if it were not for IO. That is, if we were to improve the
 * storage performance, we'd have a proportional reduction in IO-wait time.
 *
 * This all works nicely on UP, where, when a task blocks on IO, we account
 * idle time as IO-wait, because if the storage were faster, it could've been
 * running and we'd not be idle.
 *
 * This has been extended to SMP, by doing the same for each CPU. This however
 * is broken.
 *
 * Imagine for instance the case where two tasks block on one CPU, only the one
 * CPU will have IO-wait accounted, while the other has regular idle. Even
 * though, if the storage were faster, both could've ran at the same time,
 * utilising both CPUs.
 *
 * This means, that when looking globally, the current IO-wait accounting on
 * SMP is a lower bound, by reason of under accounting.
 *
 * Worse, since the numbers are provided per CPU, they are sometimes
 * interpreted per CPU, and that is nonsensical. A blocked task isn't strictly
 * associated with any one particular CPU, it can wake to another CPU than it
 * blocked on. This means the per CPU IO-wait number is meaningless.
 *
 * Task CPU affinities can make all that even more 'interesting'.
 */

L
Linus Torvalds 已提交
2955 2956 2957
unsigned long nr_iowait(void)
{
	unsigned long i, sum = 0;
2958

2959
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2960
		sum += atomic_read(&cpu_rq(i)->nr_iowait);
2961

L
Linus Torvalds 已提交
2962 2963
	return sum;
}
2964

2965 2966 2967 2968 2969 2970 2971
/*
 * Consumers of these two interfaces, like for example the cpufreq menu
 * governor are using nonsensical data. Boosting frequency for a CPU that has
 * IO-wait which might not even end up running the task when it does become
 * runnable.
 */

2972
unsigned long nr_iowait_cpu(int cpu)
2973
{
2974
	struct rq *this = cpu_rq(cpu);
2975 2976
	return atomic_read(&this->nr_iowait);
}
2977

2978 2979
void get_iowait_load(unsigned long *nr_waiters, unsigned long *load)
{
2980 2981 2982
	struct rq *rq = this_rq();
	*nr_waiters = atomic_read(&rq->nr_iowait);
	*load = rq->load.weight;
2983 2984
}

I
Ingo Molnar 已提交
2985
#ifdef CONFIG_SMP
2986

2987
/*
P
Peter Zijlstra 已提交
2988 2989
 * sched_exec - execve() is a valuable balancing opportunity, because at
 * this point the task has the smallest effective memory and cache footprint.
2990
 */
P
Peter Zijlstra 已提交
2991
void sched_exec(void)
2992
{
P
Peter Zijlstra 已提交
2993
	struct task_struct *p = current;
L
Linus Torvalds 已提交
2994
	unsigned long flags;
2995
	int dest_cpu;
2996

2997
	raw_spin_lock_irqsave(&p->pi_lock, flags);
2998
	dest_cpu = p->sched_class->select_task_rq(p, task_cpu(p), SD_BALANCE_EXEC, 0);
2999 3000
	if (dest_cpu == smp_processor_id())
		goto unlock;
P
Peter Zijlstra 已提交
3001

3002
	if (likely(cpu_active(dest_cpu))) {
3003
		struct migration_arg arg = { p, dest_cpu };
3004

3005 3006
		raw_spin_unlock_irqrestore(&p->pi_lock, flags);
		stop_one_cpu(task_cpu(p), migration_cpu_stop, &arg);
L
Linus Torvalds 已提交
3007 3008
		return;
	}
3009
unlock:
3010
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
3011
}
I
Ingo Molnar 已提交
3012

L
Linus Torvalds 已提交
3013 3014 3015
#endif

DEFINE_PER_CPU(struct kernel_stat, kstat);
3016
DEFINE_PER_CPU(struct kernel_cpustat, kernel_cpustat);
L
Linus Torvalds 已提交
3017 3018

EXPORT_PER_CPU_SYMBOL(kstat);
3019
EXPORT_PER_CPU_SYMBOL(kernel_cpustat);
L
Linus Torvalds 已提交
3020

3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037
/*
 * The function fair_sched_class.update_curr accesses the struct curr
 * and its field curr->exec_start; when called from task_sched_runtime(),
 * we observe a high rate of cache misses in practice.
 * Prefetching this data results in improved performance.
 */
static inline void prefetch_curr_exec_start(struct task_struct *p)
{
#ifdef CONFIG_FAIR_GROUP_SCHED
	struct sched_entity *curr = (&p->se)->cfs_rq->curr;
#else
	struct sched_entity *curr = (&task_rq(p)->cfs)->curr;
#endif
	prefetch(curr);
	prefetch(&curr->exec_start);
}

3038 3039 3040 3041 3042 3043 3044
/*
 * Return accounted runtime for the task.
 * In case the task is currently running, return the runtime plus current's
 * pending runtime that have not been accounted yet.
 */
unsigned long long task_sched_runtime(struct task_struct *p)
{
3045
	struct rq_flags rf;
3046
	struct rq *rq;
3047
	u64 ns;
3048

3049 3050 3051 3052 3053 3054
#if defined(CONFIG_64BIT) && defined(CONFIG_SMP)
	/*
	 * 64-bit doesn't need locks to atomically read a 64bit value.
	 * So we have a optimization chance when the task's delta_exec is 0.
	 * Reading ->on_cpu is racy, but this is ok.
	 *
I
Ingo Molnar 已提交
3055 3056
	 * If we race with it leaving CPU, we'll take a lock. So we're correct.
	 * If we race with it entering CPU, unaccounted time is 0. This is
3057
	 * indistinguishable from the read occurring a few cycles earlier.
3058 3059
	 * If we see ->on_cpu without ->on_rq, the task is leaving, and has
	 * been accounted, so we're correct here as well.
3060
	 */
3061
	if (!p->on_cpu || !task_on_rq_queued(p))
3062 3063 3064
		return p->se.sum_exec_runtime;
#endif

3065
	rq = task_rq_lock(p, &rf);
3066 3067 3068 3069 3070 3071
	/*
	 * Must be ->curr _and_ ->on_rq.  If dequeued, we would
	 * project cycles that may never be accounted to this
	 * thread, breaking clock_gettime().
	 */
	if (task_current(rq, p) && task_on_rq_queued(p)) {
3072
		prefetch_curr_exec_start(p);
3073 3074 3075 3076
		update_rq_clock(rq);
		p->sched_class->update_curr(rq);
	}
	ns = p->se.sum_exec_runtime;
3077
	task_rq_unlock(rq, p, &rf);
3078 3079 3080

	return ns;
}
3081

3082 3083 3084 3085 3086 3087 3088 3089
/*
 * This function gets called by the timer code, with HZ frequency.
 * We call it with interrupts disabled.
 */
void scheduler_tick(void)
{
	int cpu = smp_processor_id();
	struct rq *rq = cpu_rq(cpu);
I
Ingo Molnar 已提交
3090
	struct task_struct *curr = rq->curr;
3091 3092

	sched_clock_tick();
I
Ingo Molnar 已提交
3093

3094
	raw_spin_lock(&rq->lock);
3095
	update_rq_clock(rq);
P
Peter Zijlstra 已提交
3096
	curr->sched_class->task_tick(rq, curr, 0);
3097
	cpu_load_update_active(rq);
3098
	calc_global_load_tick(rq);
3099
	raw_spin_unlock(&rq->lock);
3100

3101
	perf_event_task_tick();
3102

3103
#ifdef CONFIG_SMP
3104
	rq->idle_balance = idle_cpu(cpu);
3105
	trigger_load_balance(rq);
3106
#endif
3107
	rq_last_tick_reset(rq);
L
Linus Torvalds 已提交
3108 3109
}

3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120
#ifdef CONFIG_NO_HZ_FULL
/**
 * scheduler_tick_max_deferment
 *
 * Keep at least one tick per second when a single
 * active task is running because the scheduler doesn't
 * yet completely support full dynticks environment.
 *
 * This makes sure that uptime, CFS vruntime, load
 * balancing, etc... continue to move forward, even
 * with a very low granularity.
3121 3122
 *
 * Return: Maximum deferment in nanoseconds.
3123 3124 3125 3126
 */
u64 scheduler_tick_max_deferment(void)
{
	struct rq *rq = this_rq();
3127
	unsigned long next, now = READ_ONCE(jiffies);
3128 3129 3130 3131 3132 3133

	next = rq->last_sched_tick + HZ;

	if (time_before_eq(next, now))
		return 0;

3134
	return jiffies_to_nsecs(next - now);
L
Linus Torvalds 已提交
3135
}
3136
#endif
L
Linus Torvalds 已提交
3137

3138 3139
#if defined(CONFIG_PREEMPT) && (defined(CONFIG_DEBUG_PREEMPT) || \
				defined(CONFIG_PREEMPT_TRACER))
3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153
/*
 * If the value passed in is equal to the current preempt count
 * then we just disabled preemption. Start timing the latency.
 */
static inline void preempt_latency_start(int val)
{
	if (preempt_count() == val) {
		unsigned long ip = get_lock_parent_ip();
#ifdef CONFIG_DEBUG_PREEMPT
		current->preempt_disable_ip = ip;
#endif
		trace_preempt_off(CALLER_ADDR0, ip);
	}
}
3154

3155
void preempt_count_add(int val)
L
Linus Torvalds 已提交
3156
{
3157
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3158 3159 3160
	/*
	 * Underflow?
	 */
3161 3162
	if (DEBUG_LOCKS_WARN_ON((preempt_count() < 0)))
		return;
3163
#endif
3164
	__preempt_count_add(val);
3165
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3166 3167 3168
	/*
	 * Spinlock count overflowing soon?
	 */
3169 3170
	DEBUG_LOCKS_WARN_ON((preempt_count() & PREEMPT_MASK) >=
				PREEMPT_MASK - 10);
3171
#endif
3172
	preempt_latency_start(val);
L
Linus Torvalds 已提交
3173
}
3174
EXPORT_SYMBOL(preempt_count_add);
3175
NOKPROBE_SYMBOL(preempt_count_add);
L
Linus Torvalds 已提交
3176

3177 3178 3179 3180 3181 3182 3183 3184 3185 3186
/*
 * If the value passed in equals to the current preempt count
 * then we just enabled preemption. Stop timing the latency.
 */
static inline void preempt_latency_stop(int val)
{
	if (preempt_count() == val)
		trace_preempt_on(CALLER_ADDR0, get_lock_parent_ip());
}

3187
void preempt_count_sub(int val)
L
Linus Torvalds 已提交
3188
{
3189
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3190 3191 3192
	/*
	 * Underflow?
	 */
3193
	if (DEBUG_LOCKS_WARN_ON(val > preempt_count()))
3194
		return;
L
Linus Torvalds 已提交
3195 3196 3197
	/*
	 * Is the spinlock portion underflowing?
	 */
3198 3199 3200
	if (DEBUG_LOCKS_WARN_ON((val < PREEMPT_MASK) &&
			!(preempt_count() & PREEMPT_MASK)))
		return;
3201
#endif
3202

3203
	preempt_latency_stop(val);
3204
	__preempt_count_sub(val);
L
Linus Torvalds 已提交
3205
}
3206
EXPORT_SYMBOL(preempt_count_sub);
3207
NOKPROBE_SYMBOL(preempt_count_sub);
L
Linus Torvalds 已提交
3208

3209 3210 3211
#else
static inline void preempt_latency_start(int val) { }
static inline void preempt_latency_stop(int val) { }
L
Linus Torvalds 已提交
3212 3213 3214
#endif

/*
I
Ingo Molnar 已提交
3215
 * Print scheduling while atomic bug:
L
Linus Torvalds 已提交
3216
 */
I
Ingo Molnar 已提交
3217
static noinline void __schedule_bug(struct task_struct *prev)
L
Linus Torvalds 已提交
3218
{
3219 3220 3221
	/* Save this before calling printk(), since that will clobber it */
	unsigned long preempt_disable_ip = get_preempt_disable_ip(current);

3222 3223 3224
	if (oops_in_progress)
		return;

P
Peter Zijlstra 已提交
3225 3226
	printk(KERN_ERR "BUG: scheduling while atomic: %s/%d/0x%08x\n",
		prev->comm, prev->pid, preempt_count());
3227

I
Ingo Molnar 已提交
3228
	debug_show_held_locks(prev);
3229
	print_modules();
I
Ingo Molnar 已提交
3230 3231
	if (irqs_disabled())
		print_irqtrace_events(prev);
3232 3233
	if (IS_ENABLED(CONFIG_DEBUG_PREEMPT)
	    && in_atomic_preempt_off()) {
3234
		pr_err("Preemption disabled at:");
3235
		print_ip_sym(preempt_disable_ip);
3236 3237
		pr_cont("\n");
	}
3238 3239 3240
	if (panic_on_warn)
		panic("scheduling while atomic\n");

3241
	dump_stack();
3242
	add_taint(TAINT_WARN, LOCKDEP_STILL_OK);
I
Ingo Molnar 已提交
3243
}
L
Linus Torvalds 已提交
3244

I
Ingo Molnar 已提交
3245 3246 3247 3248 3249
/*
 * Various schedule()-time debugging checks and statistics:
 */
static inline void schedule_debug(struct task_struct *prev)
{
3250
#ifdef CONFIG_SCHED_STACK_END_CHECK
J
Jann Horn 已提交
3251 3252
	if (task_stack_end_corrupted(prev))
		panic("corrupted stack end detected inside scheduler\n");
3253
#endif
3254

3255
	if (unlikely(in_atomic_preempt_off())) {
I
Ingo Molnar 已提交
3256
		__schedule_bug(prev);
3257 3258
		preempt_count_set(PREEMPT_DISABLED);
	}
3259
	rcu_sleep_check();
I
Ingo Molnar 已提交
3260

L
Linus Torvalds 已提交
3261 3262
	profile_hit(SCHED_PROFILING, __builtin_return_address(0));

3263
	schedstat_inc(this_rq()->sched_count);
I
Ingo Molnar 已提交
3264 3265 3266 3267 3268 3269
}

/*
 * Pick up the highest-prio task:
 */
static inline struct task_struct *
3270
pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
I
Ingo Molnar 已提交
3271
{
3272
	const struct sched_class *class;
I
Ingo Molnar 已提交
3273
	struct task_struct *p;
L
Linus Torvalds 已提交
3274 3275

	/*
I
Ingo Molnar 已提交
3276 3277
	 * Optimization: we know that if all tasks are in
	 * the fair class we can call that function directly:
L
Linus Torvalds 已提交
3278
	 */
3279
	if (likely(rq->nr_running == rq->cfs.h_nr_running)) {
3280
		p = fair_sched_class.pick_next_task(rq, prev, rf);
3281 3282 3283
		if (unlikely(p == RETRY_TASK))
			goto again;

I
Ingo Molnar 已提交
3284
		/* Assumes fair_sched_class->next == idle_sched_class */
3285
		if (unlikely(!p))
3286
			p = idle_sched_class.pick_next_task(rq, prev, rf);
3287 3288

		return p;
L
Linus Torvalds 已提交
3289 3290
	}

3291
again:
3292
	for_each_class(class) {
3293
		p = class->pick_next_task(rq, prev, rf);
3294 3295 3296
		if (p) {
			if (unlikely(p == RETRY_TASK))
				goto again;
I
Ingo Molnar 已提交
3297
			return p;
3298
		}
I
Ingo Molnar 已提交
3299
	}
3300

I
Ingo Molnar 已提交
3301 3302
	/* The idle class should always have a runnable task: */
	BUG();
I
Ingo Molnar 已提交
3303
}
L
Linus Torvalds 已提交
3304

I
Ingo Molnar 已提交
3305
/*
3306
 * __schedule() is the main scheduler function.
3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340
 *
 * The main means of driving the scheduler and thus entering this function are:
 *
 *   1. Explicit blocking: mutex, semaphore, waitqueue, etc.
 *
 *   2. TIF_NEED_RESCHED flag is checked on interrupt and userspace return
 *      paths. For example, see arch/x86/entry_64.S.
 *
 *      To drive preemption between tasks, the scheduler sets the flag in timer
 *      interrupt handler scheduler_tick().
 *
 *   3. Wakeups don't really cause entry into schedule(). They add a
 *      task to the run-queue and that's it.
 *
 *      Now, if the new task added to the run-queue preempts the current
 *      task, then the wakeup sets TIF_NEED_RESCHED and schedule() gets
 *      called on the nearest possible occasion:
 *
 *       - If the kernel is preemptible (CONFIG_PREEMPT=y):
 *
 *         - in syscall or exception context, at the next outmost
 *           preempt_enable(). (this might be as soon as the wake_up()'s
 *           spin_unlock()!)
 *
 *         - in IRQ context, return from interrupt-handler to
 *           preemptible context
 *
 *       - If the kernel is not preemptible (CONFIG_PREEMPT is not set)
 *         then at the next:
 *
 *          - cond_resched() call
 *          - explicit schedule() call
 *          - return from syscall or exception to user-space
 *          - return from interrupt-handler to user-space
3341
 *
3342
 * WARNING: must be called with preemption disabled!
I
Ingo Molnar 已提交
3343
 */
3344
static void __sched notrace __schedule(bool preempt)
I
Ingo Molnar 已提交
3345 3346
{
	struct task_struct *prev, *next;
3347
	unsigned long *switch_count;
3348
	struct rq_flags rf;
I
Ingo Molnar 已提交
3349
	struct rq *rq;
3350
	int cpu;
I
Ingo Molnar 已提交
3351 3352 3353 3354 3355 3356

	cpu = smp_processor_id();
	rq = cpu_rq(cpu);
	prev = rq->curr;

	schedule_debug(prev);
L
Linus Torvalds 已提交
3357

3358
	if (sched_feat(HRTICK))
M
Mike Galbraith 已提交
3359
		hrtick_clear(rq);
P
Peter Zijlstra 已提交
3360

3361 3362 3363
	local_irq_disable();
	rcu_note_context_switch();

3364 3365 3366 3367 3368 3369
	/*
	 * Make sure that signal_pending_state()->signal_pending() below
	 * can't be reordered with __set_current_state(TASK_INTERRUPTIBLE)
	 * done by the caller to avoid the race with signal_wake_up().
	 */
	smp_mb__before_spinlock();
3370
	raw_spin_lock(&rq->lock);
3371
	rq_pin_lock(rq, &rf);
L
Linus Torvalds 已提交
3372

I
Ingo Molnar 已提交
3373 3374
	/* Promote REQ to ACT */
	rq->clock_update_flags <<= 1;
3375

3376
	switch_count = &prev->nivcsw;
3377
	if (!preempt && prev->state) {
T
Tejun Heo 已提交
3378
		if (unlikely(signal_pending_state(prev->state, prev))) {
L
Linus Torvalds 已提交
3379
			prev->state = TASK_RUNNING;
T
Tejun Heo 已提交
3380
		} else {
3381 3382 3383
			deactivate_task(rq, prev, DEQUEUE_SLEEP);
			prev->on_rq = 0;

3384 3385 3386 3387 3388
			if (prev->in_iowait) {
				atomic_inc(&rq->nr_iowait);
				delayacct_blkio_start();
			}

T
Tejun Heo 已提交
3389
			/*
3390 3391 3392
			 * If a worker went to sleep, notify and ask workqueue
			 * whether it wants to wake up a task to maintain
			 * concurrency.
T
Tejun Heo 已提交
3393 3394 3395 3396
			 */
			if (prev->flags & PF_WQ_WORKER) {
				struct task_struct *to_wakeup;

3397
				to_wakeup = wq_worker_sleeping(prev);
T
Tejun Heo 已提交
3398
				if (to_wakeup)
3399
					try_to_wake_up_local(to_wakeup, &rf);
T
Tejun Heo 已提交
3400 3401
			}
		}
I
Ingo Molnar 已提交
3402
		switch_count = &prev->nvcsw;
L
Linus Torvalds 已提交
3403 3404
	}

3405
	if (task_on_rq_queued(prev))
3406 3407
		update_rq_clock(rq);

3408
	next = pick_next_task(rq, prev, &rf);
3409
	clear_tsk_need_resched(prev);
3410
	clear_preempt_need_resched();
L
Linus Torvalds 已提交
3411 3412 3413 3414 3415 3416

	if (likely(prev != next)) {
		rq->nr_switches++;
		rq->curr = next;
		++*switch_count;

3417
		trace_sched_switch(preempt, prev, next);
I
Ingo Molnar 已提交
3418 3419 3420

		/* Also unlocks the rq: */
		rq = context_switch(rq, prev, next, &rf);
3421
	} else {
3422
		rq->clock_update_flags &= ~(RQCF_ACT_SKIP|RQCF_REQ_SKIP);
3423
		rq_unpin_lock(rq, &rf);
3424
		raw_spin_unlock_irq(&rq->lock);
3425
	}
L
Linus Torvalds 已提交
3426

3427
	balance_callback(rq);
L
Linus Torvalds 已提交
3428
}
3429

3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446
void __noreturn do_task_dead(void)
{
	/*
	 * The setting of TASK_RUNNING by try_to_wake_up() may be delayed
	 * when the following two conditions become true.
	 *   - There is race condition of mmap_sem (It is acquired by
	 *     exit_mm()), and
	 *   - SMI occurs before setting TASK_RUNINNG.
	 *     (or hypervisor of virtual machine switches to other guest)
	 *  As a result, we may become TASK_RUNNING after becoming TASK_DEAD
	 *
	 * To avoid it, we have to wait for releasing tsk->pi_lock which
	 * is held by try_to_wake_up()
	 */
	smp_mb();
	raw_spin_unlock_wait(&current->pi_lock);

I
Ingo Molnar 已提交
3447
	/* Causes final put_task_struct in finish_task_switch(): */
3448
	__set_current_state(TASK_DEAD);
I
Ingo Molnar 已提交
3449 3450 3451 3452

	/* Tell freezer to ignore us: */
	current->flags |= PF_NOFREEZE;

3453 3454
	__schedule(false);
	BUG();
I
Ingo Molnar 已提交
3455 3456

	/* Avoid "noreturn function does return" - but don't continue if BUG() is a NOP: */
3457
	for (;;)
I
Ingo Molnar 已提交
3458
		cpu_relax();
3459 3460
}

3461 3462
static inline void sched_submit_work(struct task_struct *tsk)
{
3463
	if (!tsk->state || tsk_is_pi_blocked(tsk))
3464 3465 3466 3467 3468 3469 3470 3471 3472
		return;
	/*
	 * If we are going to sleep and we have plugged IO queued,
	 * make sure to submit it to avoid deadlocks.
	 */
	if (blk_needs_flush_plug(tsk))
		blk_schedule_flush_plug(tsk);
}

3473
asmlinkage __visible void __sched schedule(void)
3474
{
3475 3476 3477
	struct task_struct *tsk = current;

	sched_submit_work(tsk);
3478
	do {
3479
		preempt_disable();
3480
		__schedule(false);
3481
		sched_preempt_enable_no_resched();
3482
	} while (need_resched());
3483
}
L
Linus Torvalds 已提交
3484 3485
EXPORT_SYMBOL(schedule);

3486
#ifdef CONFIG_CONTEXT_TRACKING
3487
asmlinkage __visible void __sched schedule_user(void)
3488 3489 3490 3491 3492 3493
{
	/*
	 * If we come here after a random call to set_need_resched(),
	 * or we have been woken up remotely but the IPI has not yet arrived,
	 * we haven't yet exited the RCU idle mode. Do it here manually until
	 * we find a better solution.
3494 3495
	 *
	 * NB: There are buggy callers of this function.  Ideally we
3496
	 * should warn if prev_state != CONTEXT_USER, but that will trigger
3497
	 * too frequently to make sense yet.
3498
	 */
3499
	enum ctx_state prev_state = exception_enter();
3500
	schedule();
3501
	exception_exit(prev_state);
3502 3503 3504
}
#endif

3505 3506 3507 3508 3509 3510 3511
/**
 * schedule_preempt_disabled - called with preemption disabled
 *
 * Returns with preemption disabled. Note: preempt_count must be 1
 */
void __sched schedule_preempt_disabled(void)
{
3512
	sched_preempt_enable_no_resched();
3513 3514 3515 3516
	schedule();
	preempt_disable();
}

3517
static void __sched notrace preempt_schedule_common(void)
3518 3519
{
	do {
3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532
		/*
		 * Because the function tracer can trace preempt_count_sub()
		 * and it also uses preempt_enable/disable_notrace(), if
		 * NEED_RESCHED is set, the preempt_enable_notrace() called
		 * by the function tracer will call this function again and
		 * cause infinite recursion.
		 *
		 * Preemption must be disabled here before the function
		 * tracer can trace. Break up preempt_disable() into two
		 * calls. One to disable preemption without fear of being
		 * traced. The other to still record the preemption latency,
		 * which can also be traced by the function tracer.
		 */
3533
		preempt_disable_notrace();
3534
		preempt_latency_start(1);
3535
		__schedule(true);
3536
		preempt_latency_stop(1);
3537
		preempt_enable_no_resched_notrace();
3538 3539 3540 3541 3542 3543 3544 3545

		/*
		 * Check again in case we missed a preemption opportunity
		 * between schedule and now.
		 */
	} while (need_resched());
}

L
Linus Torvalds 已提交
3546 3547
#ifdef CONFIG_PREEMPT
/*
3548
 * this is the entry point to schedule() from in-kernel preemption
I
Ingo Molnar 已提交
3549
 * off of preempt_enable. Kernel preemptions off return from interrupt
L
Linus Torvalds 已提交
3550 3551
 * occur there and call schedule directly.
 */
3552
asmlinkage __visible void __sched notrace preempt_schedule(void)
L
Linus Torvalds 已提交
3553 3554 3555
{
	/*
	 * If there is a non-zero preempt_count or interrupts are disabled,
I
Ingo Molnar 已提交
3556
	 * we do not want to preempt the current task. Just return..
L
Linus Torvalds 已提交
3557
	 */
3558
	if (likely(!preemptible()))
L
Linus Torvalds 已提交
3559 3560
		return;

3561
	preempt_schedule_common();
L
Linus Torvalds 已提交
3562
}
3563
NOKPROBE_SYMBOL(preempt_schedule);
L
Linus Torvalds 已提交
3564
EXPORT_SYMBOL(preempt_schedule);
3565 3566

/**
3567
 * preempt_schedule_notrace - preempt_schedule called by tracing
3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579
 *
 * The tracing infrastructure uses preempt_enable_notrace to prevent
 * recursion and tracing preempt enabling caused by the tracing
 * infrastructure itself. But as tracing can happen in areas coming
 * from userspace or just about to enter userspace, a preempt enable
 * can occur before user_exit() is called. This will cause the scheduler
 * to be called when the system is still in usermode.
 *
 * To prevent this, the preempt_enable_notrace will use this function
 * instead of preempt_schedule() to exit user context if needed before
 * calling the scheduler.
 */
3580
asmlinkage __visible void __sched notrace preempt_schedule_notrace(void)
3581 3582 3583 3584 3585 3586 3587
{
	enum ctx_state prev_ctx;

	if (likely(!preemptible()))
		return;

	do {
3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600
		/*
		 * Because the function tracer can trace preempt_count_sub()
		 * and it also uses preempt_enable/disable_notrace(), if
		 * NEED_RESCHED is set, the preempt_enable_notrace() called
		 * by the function tracer will call this function again and
		 * cause infinite recursion.
		 *
		 * Preemption must be disabled here before the function
		 * tracer can trace. Break up preempt_disable() into two
		 * calls. One to disable preemption without fear of being
		 * traced. The other to still record the preemption latency,
		 * which can also be traced by the function tracer.
		 */
3601
		preempt_disable_notrace();
3602
		preempt_latency_start(1);
3603 3604 3605 3606 3607 3608
		/*
		 * Needs preempt disabled in case user_exit() is traced
		 * and the tracer calls preempt_enable_notrace() causing
		 * an infinite recursion.
		 */
		prev_ctx = exception_enter();
3609
		__schedule(true);
3610 3611
		exception_exit(prev_ctx);

3612
		preempt_latency_stop(1);
3613
		preempt_enable_no_resched_notrace();
3614 3615
	} while (need_resched());
}
3616
EXPORT_SYMBOL_GPL(preempt_schedule_notrace);
3617

3618
#endif /* CONFIG_PREEMPT */
L
Linus Torvalds 已提交
3619 3620

/*
3621
 * this is the entry point to schedule() from kernel preemption
L
Linus Torvalds 已提交
3622 3623 3624 3625
 * off of irq context.
 * Note, that this is called and return with irqs disabled. This will
 * protect us against recursive calling from irq.
 */
3626
asmlinkage __visible void __sched preempt_schedule_irq(void)
L
Linus Torvalds 已提交
3627
{
3628
	enum ctx_state prev_state;
3629

3630
	/* Catch callers which need to be fixed */
3631
	BUG_ON(preempt_count() || !irqs_disabled());
L
Linus Torvalds 已提交
3632

3633 3634
	prev_state = exception_enter();

3635
	do {
3636
		preempt_disable();
3637
		local_irq_enable();
3638
		__schedule(true);
3639
		local_irq_disable();
3640
		sched_preempt_enable_no_resched();
3641
	} while (need_resched());
3642 3643

	exception_exit(prev_state);
L
Linus Torvalds 已提交
3644 3645
}

P
Peter Zijlstra 已提交
3646
int default_wake_function(wait_queue_t *curr, unsigned mode, int wake_flags,
I
Ingo Molnar 已提交
3647
			  void *key)
L
Linus Torvalds 已提交
3648
{
P
Peter Zijlstra 已提交
3649
	return try_to_wake_up(curr->private, mode, wake_flags);
L
Linus Torvalds 已提交
3650 3651 3652
}
EXPORT_SYMBOL(default_wake_function);

3653 3654 3655 3656 3657 3658 3659 3660 3661 3662
#ifdef CONFIG_RT_MUTEXES

/*
 * rt_mutex_setprio - set the current priority of a task
 * @p: task
 * @prio: prio value (kernel-internal form)
 *
 * This function changes the 'effective' priority of a task. It does
 * not touch ->normal_prio like __setscheduler().
 *
3663 3664
 * Used by the rt_mutex code to implement priority inheritance
 * logic. Call site only calls if the priority of the task changed.
3665
 */
3666
void rt_mutex_setprio(struct task_struct *p, int prio)
3667
{
3668
	int oldprio, queued, running, queue_flag = DEQUEUE_SAVE | DEQUEUE_MOVE;
3669
	const struct sched_class *prev_class;
3670 3671
	struct rq_flags rf;
	struct rq *rq;
3672

3673
	BUG_ON(prio > MAX_PRIO);
3674

3675
	rq = __task_rq_lock(p, &rf);
3676
	update_rq_clock(rq);
3677

3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695
	/*
	 * Idle task boosting is a nono in general. There is one
	 * exception, when PREEMPT_RT and NOHZ is active:
	 *
	 * The idle task calls get_next_timer_interrupt() and holds
	 * the timer wheel base->lock on the CPU and another CPU wants
	 * to access the timer (probably to cancel it). We can safely
	 * ignore the boosting request, as the idle CPU runs this code
	 * with interrupts disabled and will complete the lock
	 * protected section without being interrupted. So there is no
	 * real need to boost.
	 */
	if (unlikely(p == rq->idle)) {
		WARN_ON(p != rq->curr);
		WARN_ON(p->pi_blocked_on);
		goto out_unlock;
	}

3696
	trace_sched_pi_setprio(p, prio);
3697
	oldprio = p->prio;
3698 3699 3700 3701

	if (oldprio == prio)
		queue_flag &= ~DEQUEUE_MOVE;

3702
	prev_class = p->sched_class;
3703
	queued = task_on_rq_queued(p);
3704
	running = task_current(rq, p);
3705
	if (queued)
3706
		dequeue_task(rq, p, queue_flag);
3707
	if (running)
3708
		put_prev_task(rq, p);
I
Ingo Molnar 已提交
3709

3710 3711 3712 3713 3714 3715 3716 3717 3718 3719
	/*
	 * Boosting condition are:
	 * 1. -rt task is running and holds mutex A
	 *      --> -dl task blocks on mutex A
	 *
	 * 2. -dl task is running and holds mutex A
	 *      --> -dl task blocks on mutex A and could preempt the
	 *          running task
	 */
	if (dl_prio(prio)) {
3720 3721 3722
		struct task_struct *pi_task = rt_mutex_get_top_task(p);
		if (!dl_prio(p->normal_prio) ||
		    (pi_task && dl_entity_preempt(&pi_task->dl, &p->dl))) {
3723
			p->dl.dl_boosted = 1;
3724
			queue_flag |= ENQUEUE_REPLENISH;
3725 3726
		} else
			p->dl.dl_boosted = 0;
3727
		p->sched_class = &dl_sched_class;
3728 3729 3730 3731
	} else if (rt_prio(prio)) {
		if (dl_prio(oldprio))
			p->dl.dl_boosted = 0;
		if (oldprio < prio)
3732
			queue_flag |= ENQUEUE_HEAD;
I
Ingo Molnar 已提交
3733
		p->sched_class = &rt_sched_class;
3734 3735 3736
	} else {
		if (dl_prio(oldprio))
			p->dl.dl_boosted = 0;
3737 3738
		if (rt_prio(oldprio))
			p->rt.timeout = 0;
I
Ingo Molnar 已提交
3739
		p->sched_class = &fair_sched_class;
3740
	}
I
Ingo Molnar 已提交
3741

3742 3743
	p->prio = prio;

3744
	if (queued)
3745
		enqueue_task(rq, p, queue_flag);
3746
	if (running)
3747
		set_curr_task(rq, p);
3748

P
Peter Zijlstra 已提交
3749
	check_class_changed(rq, p, prev_class, oldprio);
3750
out_unlock:
I
Ingo Molnar 已提交
3751 3752
	/* Avoid rq from going away on us: */
	preempt_disable();
3753
	__task_rq_unlock(rq, &rf);
3754 3755 3756

	balance_callback(rq);
	preempt_enable();
3757 3758
}
#endif
3759

3760
void set_user_nice(struct task_struct *p, long nice)
L
Linus Torvalds 已提交
3761
{
P
Peter Zijlstra 已提交
3762 3763
	bool queued, running;
	int old_prio, delta;
3764
	struct rq_flags rf;
3765
	struct rq *rq;
L
Linus Torvalds 已提交
3766

3767
	if (task_nice(p) == nice || nice < MIN_NICE || nice > MAX_NICE)
L
Linus Torvalds 已提交
3768 3769 3770 3771 3772
		return;
	/*
	 * We have to be careful, if called from sys_setpriority(),
	 * the task might be in the middle of scheduling on another CPU.
	 */
3773
	rq = task_rq_lock(p, &rf);
3774 3775
	update_rq_clock(rq);

L
Linus Torvalds 已提交
3776 3777 3778 3779
	/*
	 * The RT priorities are set via sched_setscheduler(), but we still
	 * allow the 'normal' nice value to be set - but as expected
	 * it wont have any effect on scheduling until the task is
3780
	 * SCHED_DEADLINE, SCHED_FIFO or SCHED_RR:
L
Linus Torvalds 已提交
3781
	 */
3782
	if (task_has_dl_policy(p) || task_has_rt_policy(p)) {
L
Linus Torvalds 已提交
3783 3784 3785
		p->static_prio = NICE_TO_PRIO(nice);
		goto out_unlock;
	}
3786
	queued = task_on_rq_queued(p);
P
Peter Zijlstra 已提交
3787
	running = task_current(rq, p);
3788
	if (queued)
3789
		dequeue_task(rq, p, DEQUEUE_SAVE);
P
Peter Zijlstra 已提交
3790 3791
	if (running)
		put_prev_task(rq, p);
L
Linus Torvalds 已提交
3792 3793

	p->static_prio = NICE_TO_PRIO(nice);
3794
	set_load_weight(p);
3795 3796 3797
	old_prio = p->prio;
	p->prio = effective_prio(p);
	delta = p->prio - old_prio;
L
Linus Torvalds 已提交
3798

3799
	if (queued) {
3800
		enqueue_task(rq, p, ENQUEUE_RESTORE);
L
Linus Torvalds 已提交
3801
		/*
3802 3803
		 * If the task increased its priority or is running and
		 * lowered its priority, then reschedule its CPU:
L
Linus Torvalds 已提交
3804
		 */
3805
		if (delta < 0 || (delta > 0 && task_running(rq, p)))
3806
			resched_curr(rq);
L
Linus Torvalds 已提交
3807
	}
P
Peter Zijlstra 已提交
3808 3809
	if (running)
		set_curr_task(rq, p);
L
Linus Torvalds 已提交
3810
out_unlock:
3811
	task_rq_unlock(rq, p, &rf);
L
Linus Torvalds 已提交
3812 3813 3814
}
EXPORT_SYMBOL(set_user_nice);

M
Matt Mackall 已提交
3815 3816 3817 3818 3819
/*
 * can_nice - check if a task can reduce its nice value
 * @p: task
 * @nice: nice value
 */
3820
int can_nice(const struct task_struct *p, const int nice)
M
Matt Mackall 已提交
3821
{
I
Ingo Molnar 已提交
3822
	/* Convert nice value [19,-20] to rlimit style value [1,40]: */
3823
	int nice_rlim = nice_to_rlimit(nice);
3824

3825
	return (nice_rlim <= task_rlimit(p, RLIMIT_NICE) ||
M
Matt Mackall 已提交
3826 3827 3828
		capable(CAP_SYS_NICE));
}

L
Linus Torvalds 已提交
3829 3830 3831 3832 3833 3834 3835 3836 3837
#ifdef __ARCH_WANT_SYS_NICE

/*
 * sys_nice - change the priority of the current process.
 * @increment: priority increment
 *
 * sys_setpriority is a more generic, but much slower function that
 * does similar things.
 */
3838
SYSCALL_DEFINE1(nice, int, increment)
L
Linus Torvalds 已提交
3839
{
3840
	long nice, retval;
L
Linus Torvalds 已提交
3841 3842 3843 3844 3845 3846

	/*
	 * Setpriority might change our priority at the same moment.
	 * We don't have to worry. Conceptually one call occurs first
	 * and we have a single winner.
	 */
3847
	increment = clamp(increment, -NICE_WIDTH, NICE_WIDTH);
3848
	nice = task_nice(current) + increment;
L
Linus Torvalds 已提交
3849

3850
	nice = clamp_val(nice, MIN_NICE, MAX_NICE);
M
Matt Mackall 已提交
3851 3852 3853
	if (increment < 0 && !can_nice(current, nice))
		return -EPERM;

L
Linus Torvalds 已提交
3854 3855 3856 3857 3858 3859 3860 3861 3862 3863 3864 3865 3866 3867
	retval = security_task_setnice(current, nice);
	if (retval)
		return retval;

	set_user_nice(current, nice);
	return 0;
}

#endif

/**
 * task_prio - return the priority value of a given task.
 * @p: the task in question.
 *
3868
 * Return: The priority value as seen by users in /proc.
L
Linus Torvalds 已提交
3869 3870 3871
 * RT tasks are offset by -200. Normal tasks are centered
 * around 0, value goes from -16 to +15.
 */
3872
int task_prio(const struct task_struct *p)
L
Linus Torvalds 已提交
3873 3874 3875 3876 3877
{
	return p->prio - MAX_RT_PRIO;
}

/**
I
Ingo Molnar 已提交
3878
 * idle_cpu - is a given CPU idle currently?
L
Linus Torvalds 已提交
3879
 * @cpu: the processor in question.
3880 3881
 *
 * Return: 1 if the CPU is currently idle. 0 otherwise.
L
Linus Torvalds 已提交
3882 3883 3884
 */
int idle_cpu(int cpu)
{
T
Thomas Gleixner 已提交
3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898
	struct rq *rq = cpu_rq(cpu);

	if (rq->curr != rq->idle)
		return 0;

	if (rq->nr_running)
		return 0;

#ifdef CONFIG_SMP
	if (!llist_empty(&rq->wake_list))
		return 0;
#endif

	return 1;
L
Linus Torvalds 已提交
3899 3900 3901
}

/**
I
Ingo Molnar 已提交
3902
 * idle_task - return the idle task for a given CPU.
L
Linus Torvalds 已提交
3903
 * @cpu: the processor in question.
3904
 *
I
Ingo Molnar 已提交
3905
 * Return: The idle task for the CPU @cpu.
L
Linus Torvalds 已提交
3906
 */
3907
struct task_struct *idle_task(int cpu)
L
Linus Torvalds 已提交
3908 3909 3910 3911 3912 3913 3914
{
	return cpu_rq(cpu)->idle;
}

/**
 * find_process_by_pid - find a process with a matching PID value.
 * @pid: the pid in question.
3915 3916
 *
 * The task of @pid, if found. %NULL otherwise.
L
Linus Torvalds 已提交
3917
 */
A
Alexey Dobriyan 已提交
3918
static struct task_struct *find_process_by_pid(pid_t pid)
L
Linus Torvalds 已提交
3919
{
3920
	return pid ? find_task_by_vpid(pid) : current;
L
Linus Torvalds 已提交
3921 3922
}

3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937
/*
 * This function initializes the sched_dl_entity of a newly becoming
 * SCHED_DEADLINE task.
 *
 * Only the static values are considered here, the actual runtime and the
 * absolute deadline will be properly calculated when the task is enqueued
 * for the first time with its new policy.
 */
static void
__setparam_dl(struct task_struct *p, const struct sched_attr *attr)
{
	struct sched_dl_entity *dl_se = &p->dl;

	dl_se->dl_runtime = attr->sched_runtime;
	dl_se->dl_deadline = attr->sched_deadline;
3938
	dl_se->dl_period = attr->sched_period ?: dl_se->dl_deadline;
3939
	dl_se->flags = attr->sched_flags;
3940
	dl_se->dl_bw = to_ratio(dl_se->dl_period, dl_se->dl_runtime);
3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960

	/*
	 * Changing the parameters of a task is 'tricky' and we're not doing
	 * the correct thing -- also see task_dead_dl() and switched_from_dl().
	 *
	 * What we SHOULD do is delay the bandwidth release until the 0-lag
	 * point. This would include retaining the task_struct until that time
	 * and change dl_overflow() to not immediately decrement the current
	 * amount.
	 *
	 * Instead we retain the current runtime/deadline and let the new
	 * parameters take effect after the current reservation period lapses.
	 * This is safe (albeit pessimistic) because the 0-lag point is always
	 * before the current scheduling deadline.
	 *
	 * We can still have temporary overloads because we do not delay the
	 * change in bandwidth until that time; so admission control is
	 * not on the safe side. It does however guarantee tasks will never
	 * consume more than promised.
	 */
3961 3962
}

3963 3964 3965 3966 3967 3968
/*
 * sched_setparam() passes in -1 for its policy, to let the functions
 * it calls know not to change it.
 */
#define SETPARAM_POLICY	-1

3969 3970
static void __setscheduler_params(struct task_struct *p,
		const struct sched_attr *attr)
L
Linus Torvalds 已提交
3971
{
3972 3973
	int policy = attr->sched_policy;

3974
	if (policy == SETPARAM_POLICY)
3975 3976
		policy = p->policy;

L
Linus Torvalds 已提交
3977
	p->policy = policy;
3978

3979 3980
	if (dl_policy(policy))
		__setparam_dl(p, attr);
3981
	else if (fair_policy(policy))
3982 3983
		p->static_prio = NICE_TO_PRIO(attr->sched_nice);

3984 3985 3986 3987 3988 3989
	/*
	 * __sched_setscheduler() ensures attr->sched_priority == 0 when
	 * !rt_policy. Always setting this ensures that things like
	 * getparam()/getattr() don't report silly values for !rt tasks.
	 */
	p->rt_priority = attr->sched_priority;
3990
	p->normal_prio = normal_prio(p);
3991 3992
	set_load_weight(p);
}
3993

3994 3995
/* Actually do priority change: must hold pi & rq lock. */
static void __setscheduler(struct rq *rq, struct task_struct *p,
3996
			   const struct sched_attr *attr, bool keep_boost)
3997 3998
{
	__setscheduler_params(p, attr);
3999

4000
	/*
4001 4002
	 * Keep a potential priority boosting if called from
	 * sched_setscheduler().
4003
	 */
4004 4005 4006 4007
	if (keep_boost)
		p->prio = rt_mutex_get_effective_prio(p, normal_prio(p));
	else
		p->prio = normal_prio(p);
4008

4009 4010 4011
	if (dl_prio(p->prio))
		p->sched_class = &dl_sched_class;
	else if (rt_prio(p->prio))
4012 4013 4014
		p->sched_class = &rt_sched_class;
	else
		p->sched_class = &fair_sched_class;
L
Linus Torvalds 已提交
4015
}
4016 4017 4018 4019 4020 4021 4022 4023 4024

static void
__getparam_dl(struct task_struct *p, struct sched_attr *attr)
{
	struct sched_dl_entity *dl_se = &p->dl;

	attr->sched_priority = p->rt_priority;
	attr->sched_runtime = dl_se->dl_runtime;
	attr->sched_deadline = dl_se->dl_deadline;
4025
	attr->sched_period = dl_se->dl_period;
4026 4027 4028 4029 4030 4031
	attr->sched_flags = dl_se->flags;
}

/*
 * This function validates the new parameters of a -deadline task.
 * We ask for the deadline not being zero, and greater or equal
4032
 * than the runtime, as well as the period of being zero or
4033
 * greater than deadline. Furthermore, we have to be sure that
4034 4035 4036 4037
 * user parameters are above the internal resolution of 1us (we
 * check sched_runtime only since it is always the smaller one) and
 * below 2^63 ns (we have to check both sched_deadline and
 * sched_period, as the latter can be zero).
4038 4039 4040 4041
 */
static bool
__checkparam_dl(const struct sched_attr *attr)
{
4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067
	/* deadline != 0 */
	if (attr->sched_deadline == 0)
		return false;

	/*
	 * Since we truncate DL_SCALE bits, make sure we're at least
	 * that big.
	 */
	if (attr->sched_runtime < (1ULL << DL_SCALE))
		return false;

	/*
	 * Since we use the MSB for wrap-around and sign issues, make
	 * sure it's not set (mind that period can be equal to zero).
	 */
	if (attr->sched_deadline & (1ULL << 63) ||
	    attr->sched_period & (1ULL << 63))
		return false;

	/* runtime <= deadline <= period (if period != 0) */
	if ((attr->sched_period != 0 &&
	     attr->sched_period < attr->sched_deadline) ||
	    attr->sched_deadline < attr->sched_runtime)
		return false;

	return true;
4068 4069
}

4070
/*
I
Ingo Molnar 已提交
4071
 * Check the target process has a UID that matches the current process's:
4072 4073 4074 4075 4076 4077 4078 4079
 */
static bool check_same_owner(struct task_struct *p)
{
	const struct cred *cred = current_cred(), *pcred;
	bool match;

	rcu_read_lock();
	pcred = __task_cred(p);
4080 4081
	match = (uid_eq(cred->euid, pcred->euid) ||
		 uid_eq(cred->euid, pcred->uid));
4082 4083 4084 4085
	rcu_read_unlock();
	return match;
}

I
Ingo Molnar 已提交
4086
static bool dl_param_changed(struct task_struct *p, const struct sched_attr *attr)
4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098
{
	struct sched_dl_entity *dl_se = &p->dl;

	if (dl_se->dl_runtime != attr->sched_runtime ||
		dl_se->dl_deadline != attr->sched_deadline ||
		dl_se->dl_period != attr->sched_period ||
		dl_se->flags != attr->sched_flags)
		return true;

	return false;
}

4099 4100
static int __sched_setscheduler(struct task_struct *p,
				const struct sched_attr *attr,
4101
				bool user, bool pi)
L
Linus Torvalds 已提交
4102
{
4103 4104
	int newprio = dl_policy(attr->sched_policy) ? MAX_DL_PRIO - 1 :
		      MAX_RT_PRIO - 1 - attr->sched_priority;
4105
	int retval, oldprio, oldpolicy = -1, queued, running;
4106
	int new_effective_prio, policy = attr->sched_policy;
4107
	const struct sched_class *prev_class;
4108
	struct rq_flags rf;
4109
	int reset_on_fork;
4110
	int queue_flags = DEQUEUE_SAVE | DEQUEUE_MOVE;
4111
	struct rq *rq;
L
Linus Torvalds 已提交
4112

I
Ingo Molnar 已提交
4113
	/* May grab non-irq protected spin_locks: */
4114
	BUG_ON(in_interrupt());
L
Linus Torvalds 已提交
4115
recheck:
I
Ingo Molnar 已提交
4116
	/* Double check policy once rq lock held: */
4117 4118
	if (policy < 0) {
		reset_on_fork = p->sched_reset_on_fork;
L
Linus Torvalds 已提交
4119
		policy = oldpolicy = p->policy;
4120
	} else {
4121
		reset_on_fork = !!(attr->sched_flags & SCHED_FLAG_RESET_ON_FORK);
4122

4123
		if (!valid_policy(policy))
4124 4125 4126
			return -EINVAL;
	}

4127 4128 4129
	if (attr->sched_flags & ~(SCHED_FLAG_RESET_ON_FORK))
		return -EINVAL;

L
Linus Torvalds 已提交
4130 4131
	/*
	 * Valid priorities for SCHED_FIFO and SCHED_RR are
I
Ingo Molnar 已提交
4132 4133
	 * 1..MAX_USER_RT_PRIO-1, valid priority for SCHED_NORMAL,
	 * SCHED_BATCH and SCHED_IDLE is 0.
L
Linus Torvalds 已提交
4134
	 */
4135
	if ((p->mm && attr->sched_priority > MAX_USER_RT_PRIO-1) ||
4136
	    (!p->mm && attr->sched_priority > MAX_RT_PRIO-1))
L
Linus Torvalds 已提交
4137
		return -EINVAL;
4138 4139
	if ((dl_policy(policy) && !__checkparam_dl(attr)) ||
	    (rt_policy(policy) != (attr->sched_priority != 0)))
L
Linus Torvalds 已提交
4140 4141
		return -EINVAL;

4142 4143 4144
	/*
	 * Allow unprivileged RT tasks to decrease priority:
	 */
4145
	if (user && !capable(CAP_SYS_NICE)) {
4146
		if (fair_policy(policy)) {
4147
			if (attr->sched_nice < task_nice(p) &&
4148
			    !can_nice(p, attr->sched_nice))
4149 4150 4151
				return -EPERM;
		}

4152
		if (rt_policy(policy)) {
4153 4154
			unsigned long rlim_rtprio =
					task_rlimit(p, RLIMIT_RTPRIO);
4155

I
Ingo Molnar 已提交
4156
			/* Can't set/change the rt policy: */
4157 4158 4159
			if (policy != p->policy && !rlim_rtprio)
				return -EPERM;

I
Ingo Molnar 已提交
4160
			/* Can't increase priority: */
4161 4162
			if (attr->sched_priority > p->rt_priority &&
			    attr->sched_priority > rlim_rtprio)
4163 4164
				return -EPERM;
		}
4165

4166 4167 4168 4169 4170 4171 4172 4173 4174
		 /*
		  * Can't set/change SCHED_DEADLINE policy at all for now
		  * (safest behavior); in the future we would like to allow
		  * unprivileged DL tasks to increase their relative deadline
		  * or reduce their runtime (both ways reducing utilization)
		  */
		if (dl_policy(policy))
			return -EPERM;

I
Ingo Molnar 已提交
4175
		/*
4176 4177
		 * Treat SCHED_IDLE as nice 20. Only allow a switch to
		 * SCHED_NORMAL if the RLIMIT_NICE would normally permit it.
I
Ingo Molnar 已提交
4178
		 */
4179
		if (idle_policy(p->policy) && !idle_policy(policy)) {
4180
			if (!can_nice(p, task_nice(p)))
4181 4182
				return -EPERM;
		}
4183

I
Ingo Molnar 已提交
4184
		/* Can't change other user's priorities: */
4185
		if (!check_same_owner(p))
4186
			return -EPERM;
4187

I
Ingo Molnar 已提交
4188
		/* Normal users shall not reset the sched_reset_on_fork flag: */
4189 4190
		if (p->sched_reset_on_fork && !reset_on_fork)
			return -EPERM;
4191
	}
L
Linus Torvalds 已提交
4192

4193
	if (user) {
4194
		retval = security_task_setscheduler(p);
4195 4196 4197 4198
		if (retval)
			return retval;
	}

4199
	/*
I
Ingo Molnar 已提交
4200
	 * Make sure no PI-waiters arrive (or leave) while we are
4201
	 * changing the priority of the task:
4202
	 *
L
Lucas De Marchi 已提交
4203
	 * To be able to change p->policy safely, the appropriate
L
Linus Torvalds 已提交
4204 4205
	 * runqueue lock must be held.
	 */
4206
	rq = task_rq_lock(p, &rf);
4207
	update_rq_clock(rq);
4208

4209
	/*
I
Ingo Molnar 已提交
4210
	 * Changing the policy of the stop threads its a very bad idea:
4211 4212
	 */
	if (p == rq->stop) {
4213
		task_rq_unlock(rq, p, &rf);
4214 4215 4216
		return -EINVAL;
	}

4217
	/*
4218 4219
	 * If not changing anything there's no need to proceed further,
	 * but store a possible modification of reset_on_fork.
4220
	 */
4221
	if (unlikely(policy == p->policy)) {
4222
		if (fair_policy(policy) && attr->sched_nice != task_nice(p))
4223 4224 4225
			goto change;
		if (rt_policy(policy) && attr->sched_priority != p->rt_priority)
			goto change;
4226
		if (dl_policy(policy) && dl_param_changed(p, attr))
4227
			goto change;
4228

4229
		p->sched_reset_on_fork = reset_on_fork;
4230
		task_rq_unlock(rq, p, &rf);
4231 4232
		return 0;
	}
4233
change:
4234

4235
	if (user) {
4236
#ifdef CONFIG_RT_GROUP_SCHED
4237 4238 4239 4240 4241
		/*
		 * Do not allow realtime tasks into groups that have no runtime
		 * assigned.
		 */
		if (rt_bandwidth_enabled() && rt_policy(policy) &&
4242 4243
				task_group(p)->rt_bandwidth.rt_runtime == 0 &&
				!task_group_is_autogroup(task_group(p))) {
4244
			task_rq_unlock(rq, p, &rf);
4245 4246 4247
			return -EPERM;
		}
#endif
4248 4249 4250 4251 4252 4253 4254 4255 4256
#ifdef CONFIG_SMP
		if (dl_bandwidth_enabled() && dl_policy(policy)) {
			cpumask_t *span = rq->rd->span;

			/*
			 * Don't allow tasks with an affinity mask smaller than
			 * the entire root_domain to become SCHED_DEADLINE. We
			 * will also fail if there's no bandwidth available.
			 */
4257 4258
			if (!cpumask_subset(span, &p->cpus_allowed) ||
			    rq->rd->dl_bw.bw == 0) {
4259
				task_rq_unlock(rq, p, &rf);
4260 4261 4262 4263 4264
				return -EPERM;
			}
		}
#endif
	}
4265

I
Ingo Molnar 已提交
4266
	/* Re-check policy now with rq lock held: */
L
Linus Torvalds 已提交
4267 4268
	if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) {
		policy = oldpolicy = -1;
4269
		task_rq_unlock(rq, p, &rf);
L
Linus Torvalds 已提交
4270 4271
		goto recheck;
	}
4272 4273 4274 4275 4276 4277

	/*
	 * If setscheduling to SCHED_DEADLINE (or changing the parameters
	 * of a SCHED_DEADLINE task) we need to check if enough bandwidth
	 * is available.
	 */
4278
	if ((dl_policy(policy) || dl_task(p)) && dl_overflow(p, policy, attr)) {
4279
		task_rq_unlock(rq, p, &rf);
4280 4281 4282
		return -EBUSY;
	}

4283 4284 4285
	p->sched_reset_on_fork = reset_on_fork;
	oldprio = p->prio;

4286 4287 4288 4289 4290 4291 4292 4293 4294
	if (pi) {
		/*
		 * Take priority boosted tasks into account. If the new
		 * effective priority is unchanged, we just store the new
		 * normal parameters and do not touch the scheduler class and
		 * the runqueue. This will be done when the task deboost
		 * itself.
		 */
		new_effective_prio = rt_mutex_get_effective_prio(p, newprio);
4295 4296
		if (new_effective_prio == oldprio)
			queue_flags &= ~DEQUEUE_MOVE;
4297 4298
	}

4299
	queued = task_on_rq_queued(p);
4300
	running = task_current(rq, p);
4301
	if (queued)
4302
		dequeue_task(rq, p, queue_flags);
4303
	if (running)
4304
		put_prev_task(rq, p);
4305

4306
	prev_class = p->sched_class;
4307
	__setscheduler(rq, p, attr, pi);
4308

4309
	if (queued) {
4310 4311 4312 4313
		/*
		 * We enqueue to tail when the priority of a task is
		 * increased (user space view).
		 */
4314 4315
		if (oldprio < p->prio)
			queue_flags |= ENQUEUE_HEAD;
4316

4317
		enqueue_task(rq, p, queue_flags);
4318
	}
4319
	if (running)
4320
		set_curr_task(rq, p);
4321

P
Peter Zijlstra 已提交
4322
	check_class_changed(rq, p, prev_class, oldprio);
I
Ingo Molnar 已提交
4323 4324 4325

	/* Avoid rq from going away on us: */
	preempt_disable();
4326
	task_rq_unlock(rq, p, &rf);
4327

4328 4329
	if (pi)
		rt_mutex_adjust_pi(p);
4330

I
Ingo Molnar 已提交
4331
	/* Run balance callbacks after we've adjusted the PI chain: */
4332 4333
	balance_callback(rq);
	preempt_enable();
4334

L
Linus Torvalds 已提交
4335 4336
	return 0;
}
4337

4338 4339 4340 4341 4342 4343 4344 4345 4346
static int _sched_setscheduler(struct task_struct *p, int policy,
			       const struct sched_param *param, bool check)
{
	struct sched_attr attr = {
		.sched_policy   = policy,
		.sched_priority = param->sched_priority,
		.sched_nice	= PRIO_TO_NICE(p->static_prio),
	};

4347 4348
	/* Fixup the legacy SCHED_RESET_ON_FORK hack. */
	if ((policy != SETPARAM_POLICY) && (policy & SCHED_RESET_ON_FORK)) {
4349 4350 4351 4352 4353
		attr.sched_flags |= SCHED_FLAG_RESET_ON_FORK;
		policy &= ~SCHED_RESET_ON_FORK;
		attr.sched_policy = policy;
	}

4354
	return __sched_setscheduler(p, &attr, check, true);
4355
}
4356 4357 4358 4359 4360 4361
/**
 * sched_setscheduler - change the scheduling policy and/or RT priority of a thread.
 * @p: the task in question.
 * @policy: new policy.
 * @param: structure containing the new RT priority.
 *
4362 4363
 * Return: 0 on success. An error code otherwise.
 *
4364 4365 4366
 * NOTE that the task may be already dead.
 */
int sched_setscheduler(struct task_struct *p, int policy,
4367
		       const struct sched_param *param)
4368
{
4369
	return _sched_setscheduler(p, policy, param, true);
4370
}
L
Linus Torvalds 已提交
4371 4372
EXPORT_SYMBOL_GPL(sched_setscheduler);

4373 4374
int sched_setattr(struct task_struct *p, const struct sched_attr *attr)
{
4375
	return __sched_setscheduler(p, attr, true, true);
4376 4377 4378
}
EXPORT_SYMBOL_GPL(sched_setattr);

4379 4380 4381 4382 4383 4384 4385 4386 4387 4388
/**
 * sched_setscheduler_nocheck - change the scheduling policy and/or RT priority of a thread from kernelspace.
 * @p: the task in question.
 * @policy: new policy.
 * @param: structure containing the new RT priority.
 *
 * Just like sched_setscheduler, only don't bother checking if the
 * current context has permission.  For example, this is needed in
 * stop_machine(): we create temporary high priority worker threads,
 * but our caller might not have that capability.
4389 4390
 *
 * Return: 0 on success. An error code otherwise.
4391 4392
 */
int sched_setscheduler_nocheck(struct task_struct *p, int policy,
4393
			       const struct sched_param *param)
4394
{
4395
	return _sched_setscheduler(p, policy, param, false);
4396
}
4397
EXPORT_SYMBOL_GPL(sched_setscheduler_nocheck);
4398

I
Ingo Molnar 已提交
4399 4400
static int
do_sched_setscheduler(pid_t pid, int policy, struct sched_param __user *param)
L
Linus Torvalds 已提交
4401 4402 4403
{
	struct sched_param lparam;
	struct task_struct *p;
4404
	int retval;
L
Linus Torvalds 已提交
4405 4406 4407 4408 4409

	if (!param || pid < 0)
		return -EINVAL;
	if (copy_from_user(&lparam, param, sizeof(struct sched_param)))
		return -EFAULT;
4410 4411 4412

	rcu_read_lock();
	retval = -ESRCH;
L
Linus Torvalds 已提交
4413
	p = find_process_by_pid(pid);
4414 4415 4416
	if (p != NULL)
		retval = sched_setscheduler(p, policy, &lparam);
	rcu_read_unlock();
4417

L
Linus Torvalds 已提交
4418 4419 4420
	return retval;
}

4421 4422 4423
/*
 * Mimics kernel/events/core.c perf_copy_attr().
 */
I
Ingo Molnar 已提交
4424
static int sched_copy_attr(struct sched_attr __user *uattr, struct sched_attr *attr)
4425 4426 4427 4428 4429 4430 4431
{
	u32 size;
	int ret;

	if (!access_ok(VERIFY_WRITE, uattr, SCHED_ATTR_SIZE_VER0))
		return -EFAULT;

I
Ingo Molnar 已提交
4432
	/* Zero the full structure, so that a short copy will be nice: */
4433 4434 4435 4436 4437 4438
	memset(attr, 0, sizeof(*attr));

	ret = get_user(size, &uattr->size);
	if (ret)
		return ret;

I
Ingo Molnar 已提交
4439 4440
	/* Bail out on silly large: */
	if (size > PAGE_SIZE)
4441 4442
		goto err_size;

I
Ingo Molnar 已提交
4443 4444
	/* ABI compatibility quirk: */
	if (!size)
4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478
		size = SCHED_ATTR_SIZE_VER0;

	if (size < SCHED_ATTR_SIZE_VER0)
		goto err_size;

	/*
	 * If we're handed a bigger struct than we know of,
	 * ensure all the unknown bits are 0 - i.e. new
	 * user-space does not rely on any kernel feature
	 * extensions we dont know about yet.
	 */
	if (size > sizeof(*attr)) {
		unsigned char __user *addr;
		unsigned char __user *end;
		unsigned char val;

		addr = (void __user *)uattr + sizeof(*attr);
		end  = (void __user *)uattr + size;

		for (; addr < end; addr++) {
			ret = get_user(val, addr);
			if (ret)
				return ret;
			if (val)
				goto err_size;
		}
		size = sizeof(*attr);
	}

	ret = copy_from_user(attr, uattr, size);
	if (ret)
		return -EFAULT;

	/*
I
Ingo Molnar 已提交
4479
	 * XXX: Do we want to be lenient like existing syscalls; or do we want
4480 4481
	 * to be strict and return an error on out-of-bounds values?
	 */
4482
	attr->sched_nice = clamp(attr->sched_nice, MIN_NICE, MAX_NICE);
4483

4484
	return 0;
4485 4486 4487

err_size:
	put_user(sizeof(*attr), &uattr->size);
4488
	return -E2BIG;
4489 4490
}

L
Linus Torvalds 已提交
4491 4492 4493 4494 4495
/**
 * sys_sched_setscheduler - set/change the scheduler policy and RT priority
 * @pid: the pid in question.
 * @policy: new policy.
 * @param: structure containing the new RT priority.
4496 4497
 *
 * Return: 0 on success. An error code otherwise.
L
Linus Torvalds 已提交
4498
 */
I
Ingo Molnar 已提交
4499
SYSCALL_DEFINE3(sched_setscheduler, pid_t, pid, int, policy, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4500
{
4501 4502 4503
	if (policy < 0)
		return -EINVAL;

L
Linus Torvalds 已提交
4504 4505 4506 4507 4508 4509 4510
	return do_sched_setscheduler(pid, policy, param);
}

/**
 * sys_sched_setparam - set/change the RT priority of a thread
 * @pid: the pid in question.
 * @param: structure containing the new RT priority.
4511 4512
 *
 * Return: 0 on success. An error code otherwise.
L
Linus Torvalds 已提交
4513
 */
4514
SYSCALL_DEFINE2(sched_setparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4515
{
4516
	return do_sched_setscheduler(pid, SETPARAM_POLICY, param);
L
Linus Torvalds 已提交
4517 4518
}

4519 4520 4521
/**
 * sys_sched_setattr - same as above, but with extended sched_attr
 * @pid: the pid in question.
J
Juri Lelli 已提交
4522
 * @uattr: structure containing the extended parameters.
4523
 * @flags: for future extension.
4524
 */
4525 4526
SYSCALL_DEFINE3(sched_setattr, pid_t, pid, struct sched_attr __user *, uattr,
			       unsigned int, flags)
4527 4528 4529 4530 4531
{
	struct sched_attr attr;
	struct task_struct *p;
	int retval;

4532
	if (!uattr || pid < 0 || flags)
4533 4534
		return -EINVAL;

4535 4536 4537
	retval = sched_copy_attr(uattr, &attr);
	if (retval)
		return retval;
4538

4539
	if ((int)attr.sched_policy < 0)
4540
		return -EINVAL;
4541 4542 4543 4544 4545 4546 4547 4548 4549 4550 4551

	rcu_read_lock();
	retval = -ESRCH;
	p = find_process_by_pid(pid);
	if (p != NULL)
		retval = sched_setattr(p, &attr);
	rcu_read_unlock();

	return retval;
}

L
Linus Torvalds 已提交
4552 4553 4554
/**
 * sys_sched_getscheduler - get the policy (scheduling class) of a thread
 * @pid: the pid in question.
4555 4556 4557
 *
 * Return: On success, the policy of the thread. Otherwise, a negative error
 * code.
L
Linus Torvalds 已提交
4558
 */
4559
SYSCALL_DEFINE1(sched_getscheduler, pid_t, pid)
L
Linus Torvalds 已提交
4560
{
4561
	struct task_struct *p;
4562
	int retval;
L
Linus Torvalds 已提交
4563 4564

	if (pid < 0)
4565
		return -EINVAL;
L
Linus Torvalds 已提交
4566 4567

	retval = -ESRCH;
4568
	rcu_read_lock();
L
Linus Torvalds 已提交
4569 4570 4571 4572
	p = find_process_by_pid(pid);
	if (p) {
		retval = security_task_getscheduler(p);
		if (!retval)
4573 4574
			retval = p->policy
				| (p->sched_reset_on_fork ? SCHED_RESET_ON_FORK : 0);
L
Linus Torvalds 已提交
4575
	}
4576
	rcu_read_unlock();
L
Linus Torvalds 已提交
4577 4578 4579 4580
	return retval;
}

/**
4581
 * sys_sched_getparam - get the RT priority of a thread
L
Linus Torvalds 已提交
4582 4583
 * @pid: the pid in question.
 * @param: structure containing the RT priority.
4584 4585 4586
 *
 * Return: On success, 0 and the RT priority is in @param. Otherwise, an error
 * code.
L
Linus Torvalds 已提交
4587
 */
4588
SYSCALL_DEFINE2(sched_getparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4589
{
4590
	struct sched_param lp = { .sched_priority = 0 };
4591
	struct task_struct *p;
4592
	int retval;
L
Linus Torvalds 已提交
4593 4594

	if (!param || pid < 0)
4595
		return -EINVAL;
L
Linus Torvalds 已提交
4596

4597
	rcu_read_lock();
L
Linus Torvalds 已提交
4598 4599 4600 4601 4602 4603 4604 4605 4606
	p = find_process_by_pid(pid);
	retval = -ESRCH;
	if (!p)
		goto out_unlock;

	retval = security_task_getscheduler(p);
	if (retval)
		goto out_unlock;

4607 4608
	if (task_has_rt_policy(p))
		lp.sched_priority = p->rt_priority;
4609
	rcu_read_unlock();
L
Linus Torvalds 已提交
4610 4611 4612 4613 4614 4615 4616 4617 4618

	/*
	 * This one might sleep, we cannot do it with a spinlock held ...
	 */
	retval = copy_to_user(param, &lp, sizeof(*param)) ? -EFAULT : 0;

	return retval;

out_unlock:
4619
	rcu_read_unlock();
L
Linus Torvalds 已提交
4620 4621 4622
	return retval;
}

4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643 4644 4645
static int sched_read_attr(struct sched_attr __user *uattr,
			   struct sched_attr *attr,
			   unsigned int usize)
{
	int ret;

	if (!access_ok(VERIFY_WRITE, uattr, usize))
		return -EFAULT;

	/*
	 * If we're handed a smaller struct than we know of,
	 * ensure all the unknown bits are 0 - i.e. old
	 * user-space does not get uncomplete information.
	 */
	if (usize < sizeof(*attr)) {
		unsigned char *addr;
		unsigned char *end;

		addr = (void *)attr + usize;
		end  = (void *)attr + sizeof(*attr);

		for (; addr < end; addr++) {
			if (*addr)
4646
				return -EFBIG;
4647 4648 4649 4650 4651
		}

		attr->size = usize;
	}

4652
	ret = copy_to_user(uattr, attr, attr->size);
4653 4654 4655
	if (ret)
		return -EFAULT;

4656
	return 0;
4657 4658 4659
}

/**
4660
 * sys_sched_getattr - similar to sched_getparam, but with sched_attr
4661
 * @pid: the pid in question.
J
Juri Lelli 已提交
4662
 * @uattr: structure containing the extended parameters.
4663
 * @size: sizeof(attr) for fwd/bwd comp.
4664
 * @flags: for future extension.
4665
 */
4666 4667
SYSCALL_DEFINE4(sched_getattr, pid_t, pid, struct sched_attr __user *, uattr,
		unsigned int, size, unsigned int, flags)
4668 4669 4670 4671 4672 4673 4674 4675
{
	struct sched_attr attr = {
		.size = sizeof(struct sched_attr),
	};
	struct task_struct *p;
	int retval;

	if (!uattr || pid < 0 || size > PAGE_SIZE ||
4676
	    size < SCHED_ATTR_SIZE_VER0 || flags)
4677 4678 4679 4680 4681 4682 4683 4684 4685 4686 4687 4688 4689
		return -EINVAL;

	rcu_read_lock();
	p = find_process_by_pid(pid);
	retval = -ESRCH;
	if (!p)
		goto out_unlock;

	retval = security_task_getscheduler(p);
	if (retval)
		goto out_unlock;

	attr.sched_policy = p->policy;
4690 4691
	if (p->sched_reset_on_fork)
		attr.sched_flags |= SCHED_FLAG_RESET_ON_FORK;
4692 4693 4694
	if (task_has_dl_policy(p))
		__getparam_dl(p, &attr);
	else if (task_has_rt_policy(p))
4695 4696
		attr.sched_priority = p->rt_priority;
	else
4697
		attr.sched_nice = task_nice(p);
4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708

	rcu_read_unlock();

	retval = sched_read_attr(uattr, &attr, size);
	return retval;

out_unlock:
	rcu_read_unlock();
	return retval;
}

4709
long sched_setaffinity(pid_t pid, const struct cpumask *in_mask)
L
Linus Torvalds 已提交
4710
{
4711
	cpumask_var_t cpus_allowed, new_mask;
4712 4713
	struct task_struct *p;
	int retval;
L
Linus Torvalds 已提交
4714

4715
	rcu_read_lock();
L
Linus Torvalds 已提交
4716 4717 4718

	p = find_process_by_pid(pid);
	if (!p) {
4719
		rcu_read_unlock();
L
Linus Torvalds 已提交
4720 4721 4722
		return -ESRCH;
	}

4723
	/* Prevent p going away */
L
Linus Torvalds 已提交
4724
	get_task_struct(p);
4725
	rcu_read_unlock();
L
Linus Torvalds 已提交
4726

4727 4728 4729 4730
	if (p->flags & PF_NO_SETAFFINITY) {
		retval = -EINVAL;
		goto out_put_task;
	}
4731 4732 4733 4734 4735 4736 4737 4738
	if (!alloc_cpumask_var(&cpus_allowed, GFP_KERNEL)) {
		retval = -ENOMEM;
		goto out_put_task;
	}
	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL)) {
		retval = -ENOMEM;
		goto out_free_cpus_allowed;
	}
L
Linus Torvalds 已提交
4739
	retval = -EPERM;
E
Eric W. Biederman 已提交
4740 4741 4742 4743
	if (!check_same_owner(p)) {
		rcu_read_lock();
		if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
			rcu_read_unlock();
4744
			goto out_free_new_mask;
E
Eric W. Biederman 已提交
4745 4746 4747
		}
		rcu_read_unlock();
	}
L
Linus Torvalds 已提交
4748

4749
	retval = security_task_setscheduler(p);
4750
	if (retval)
4751
		goto out_free_new_mask;
4752

4753 4754 4755 4756

	cpuset_cpus_allowed(p, cpus_allowed);
	cpumask_and(new_mask, in_mask, cpus_allowed);

4757 4758 4759 4760 4761 4762 4763
	/*
	 * Since bandwidth control happens on root_domain basis,
	 * if admission test is enabled, we only admit -deadline
	 * tasks allowed to run on all the CPUs in the task's
	 * root_domain.
	 */
#ifdef CONFIG_SMP
4764 4765 4766
	if (task_has_dl_policy(p) && dl_bandwidth_enabled()) {
		rcu_read_lock();
		if (!cpumask_subset(task_rq(p)->rd->span, new_mask)) {
4767
			retval = -EBUSY;
4768
			rcu_read_unlock();
4769
			goto out_free_new_mask;
4770
		}
4771
		rcu_read_unlock();
4772 4773
	}
#endif
P
Peter Zijlstra 已提交
4774
again:
4775
	retval = __set_cpus_allowed_ptr(p, new_mask, true);
L
Linus Torvalds 已提交
4776

P
Paul Menage 已提交
4777
	if (!retval) {
4778 4779
		cpuset_cpus_allowed(p, cpus_allowed);
		if (!cpumask_subset(new_mask, cpus_allowed)) {
P
Paul Menage 已提交
4780 4781 4782 4783 4784
			/*
			 * We must have raced with a concurrent cpuset
			 * update. Just reset the cpus_allowed to the
			 * cpuset's cpus_allowed
			 */
4785
			cpumask_copy(new_mask, cpus_allowed);
P
Paul Menage 已提交
4786 4787 4788
			goto again;
		}
	}
4789
out_free_new_mask:
4790 4791 4792 4793
	free_cpumask_var(new_mask);
out_free_cpus_allowed:
	free_cpumask_var(cpus_allowed);
out_put_task:
L
Linus Torvalds 已提交
4794 4795 4796 4797 4798
	put_task_struct(p);
	return retval;
}

static int get_user_cpu_mask(unsigned long __user *user_mask_ptr, unsigned len,
4799
			     struct cpumask *new_mask)
L
Linus Torvalds 已提交
4800
{
4801 4802 4803 4804 4805
	if (len < cpumask_size())
		cpumask_clear(new_mask);
	else if (len > cpumask_size())
		len = cpumask_size();

L
Linus Torvalds 已提交
4806 4807 4808 4809
	return copy_from_user(new_mask, user_mask_ptr, len) ? -EFAULT : 0;
}

/**
I
Ingo Molnar 已提交
4810
 * sys_sched_setaffinity - set the CPU affinity of a process
L
Linus Torvalds 已提交
4811 4812
 * @pid: pid of the process
 * @len: length in bytes of the bitmask pointed to by user_mask_ptr
I
Ingo Molnar 已提交
4813
 * @user_mask_ptr: user-space pointer to the new CPU mask
4814 4815
 *
 * Return: 0 on success. An error code otherwise.
L
Linus Torvalds 已提交
4816
 */
4817 4818
SYSCALL_DEFINE3(sched_setaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
4819
{
4820
	cpumask_var_t new_mask;
L
Linus Torvalds 已提交
4821 4822
	int retval;

4823 4824
	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
4825

4826 4827 4828 4829 4830
	retval = get_user_cpu_mask(user_mask_ptr, len, new_mask);
	if (retval == 0)
		retval = sched_setaffinity(pid, new_mask);
	free_cpumask_var(new_mask);
	return retval;
L
Linus Torvalds 已提交
4831 4832
}

4833
long sched_getaffinity(pid_t pid, struct cpumask *mask)
L
Linus Torvalds 已提交
4834
{
4835
	struct task_struct *p;
4836
	unsigned long flags;
L
Linus Torvalds 已提交
4837 4838
	int retval;

4839
	rcu_read_lock();
L
Linus Torvalds 已提交
4840 4841 4842 4843 4844 4845

	retval = -ESRCH;
	p = find_process_by_pid(pid);
	if (!p)
		goto out_unlock;

4846 4847 4848 4849
	retval = security_task_getscheduler(p);
	if (retval)
		goto out_unlock;

4850
	raw_spin_lock_irqsave(&p->pi_lock, flags);
4851
	cpumask_and(mask, &p->cpus_allowed, cpu_active_mask);
4852
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
4853 4854

out_unlock:
4855
	rcu_read_unlock();
L
Linus Torvalds 已提交
4856

4857
	return retval;
L
Linus Torvalds 已提交
4858 4859 4860
}

/**
I
Ingo Molnar 已提交
4861
 * sys_sched_getaffinity - get the CPU affinity of a process
L
Linus Torvalds 已提交
4862 4863
 * @pid: pid of the process
 * @len: length in bytes of the bitmask pointed to by user_mask_ptr
I
Ingo Molnar 已提交
4864
 * @user_mask_ptr: user-space pointer to hold the current CPU mask
4865
 *
4866 4867
 * Return: size of CPU mask copied to user_mask_ptr on success. An
 * error code otherwise.
L
Linus Torvalds 已提交
4868
 */
4869 4870
SYSCALL_DEFINE3(sched_getaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
4871 4872
{
	int ret;
4873
	cpumask_var_t mask;
L
Linus Torvalds 已提交
4874

A
Anton Blanchard 已提交
4875
	if ((len * BITS_PER_BYTE) < nr_cpu_ids)
4876 4877
		return -EINVAL;
	if (len & (sizeof(unsigned long)-1))
L
Linus Torvalds 已提交
4878 4879
		return -EINVAL;

4880 4881
	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
4882

4883 4884
	ret = sched_getaffinity(pid, mask);
	if (ret == 0) {
4885
		size_t retlen = min_t(size_t, len, cpumask_size());
4886 4887

		if (copy_to_user(user_mask_ptr, mask, retlen))
4888 4889
			ret = -EFAULT;
		else
4890
			ret = retlen;
4891 4892
	}
	free_cpumask_var(mask);
L
Linus Torvalds 已提交
4893

4894
	return ret;
L
Linus Torvalds 已提交
4895 4896 4897 4898 4899
}

/**
 * sys_sched_yield - yield the current processor to other threads.
 *
I
Ingo Molnar 已提交
4900 4901
 * This function yields the current CPU to other tasks. If there are no
 * other threads running on this CPU then this function will return.
4902 4903
 *
 * Return: 0.
L
Linus Torvalds 已提交
4904
 */
4905
SYSCALL_DEFINE0(sched_yield)
L
Linus Torvalds 已提交
4906
{
4907
	struct rq *rq = this_rq_lock();
L
Linus Torvalds 已提交
4908

4909
	schedstat_inc(rq->yld_count);
4910
	current->sched_class->yield_task(rq);
L
Linus Torvalds 已提交
4911 4912 4913 4914 4915 4916

	/*
	 * Since we are going to call schedule() anyway, there's
	 * no need to preempt or enable interrupts:
	 */
	__release(rq->lock);
4917
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
4918
	do_raw_spin_unlock(&rq->lock);
4919
	sched_preempt_enable_no_resched();
L
Linus Torvalds 已提交
4920 4921 4922 4923 4924 4925

	schedule();

	return 0;
}

4926
#ifndef CONFIG_PREEMPT
4927
int __sched _cond_resched(void)
L
Linus Torvalds 已提交
4928
{
4929
	if (should_resched(0)) {
4930
		preempt_schedule_common();
L
Linus Torvalds 已提交
4931 4932 4933 4934
		return 1;
	}
	return 0;
}
4935
EXPORT_SYMBOL(_cond_resched);
4936
#endif
L
Linus Torvalds 已提交
4937 4938

/*
4939
 * __cond_resched_lock() - if a reschedule is pending, drop the given lock,
L
Linus Torvalds 已提交
4940 4941
 * call schedule, and on return reacquire the lock.
 *
I
Ingo Molnar 已提交
4942
 * This works OK both with and without CONFIG_PREEMPT. We do strange low-level
L
Linus Torvalds 已提交
4943 4944 4945
 * operations here to prevent schedule() from being called twice (once via
 * spin_unlock(), once by hand).
 */
4946
int __cond_resched_lock(spinlock_t *lock)
L
Linus Torvalds 已提交
4947
{
4948
	int resched = should_resched(PREEMPT_LOCK_OFFSET);
J
Jan Kara 已提交
4949 4950
	int ret = 0;

4951 4952
	lockdep_assert_held(lock);

4953
	if (spin_needbreak(lock) || resched) {
L
Linus Torvalds 已提交
4954
		spin_unlock(lock);
P
Peter Zijlstra 已提交
4955
		if (resched)
4956
			preempt_schedule_common();
N
Nick Piggin 已提交
4957 4958
		else
			cpu_relax();
J
Jan Kara 已提交
4959
		ret = 1;
L
Linus Torvalds 已提交
4960 4961
		spin_lock(lock);
	}
J
Jan Kara 已提交
4962
	return ret;
L
Linus Torvalds 已提交
4963
}
4964
EXPORT_SYMBOL(__cond_resched_lock);
L
Linus Torvalds 已提交
4965

4966
int __sched __cond_resched_softirq(void)
L
Linus Torvalds 已提交
4967 4968 4969
{
	BUG_ON(!in_softirq());

4970
	if (should_resched(SOFTIRQ_DISABLE_OFFSET)) {
4971
		local_bh_enable();
4972
		preempt_schedule_common();
L
Linus Torvalds 已提交
4973 4974 4975 4976 4977
		local_bh_disable();
		return 1;
	}
	return 0;
}
4978
EXPORT_SYMBOL(__cond_resched_softirq);
L
Linus Torvalds 已提交
4979 4980 4981 4982

/**
 * yield - yield the current processor to other threads.
 *
P
Peter Zijlstra 已提交
4983 4984 4985 4986 4987 4988 4989 4990 4991
 * Do not ever use this function, there's a 99% chance you're doing it wrong.
 *
 * The scheduler is at all times free to pick the calling task as the most
 * eligible task to run, if removing the yield() call from your code breaks
 * it, its already broken.
 *
 * Typical broken usage is:
 *
 * while (!event)
I
Ingo Molnar 已提交
4992
 *	yield();
P
Peter Zijlstra 已提交
4993 4994 4995 4996 4997 4998 4999 5000
 *
 * where one assumes that yield() will let 'the other' process run that will
 * make event true. If the current task is a SCHED_FIFO task that will never
 * happen. Never use yield() as a progress guarantee!!
 *
 * If you want to use yield() to wait for something, use wait_event().
 * If you want to use yield() to be 'nice' for others, use cond_resched().
 * If you still want to use yield(), do not!
L
Linus Torvalds 已提交
5001 5002 5003 5004 5005 5006 5007 5008
 */
void __sched yield(void)
{
	set_current_state(TASK_RUNNING);
	sys_sched_yield();
}
EXPORT_SYMBOL(yield);

5009 5010 5011 5012
/**
 * yield_to - yield the current processor to another thread in
 * your thread group, or accelerate that thread toward the
 * processor it's on.
R
Randy Dunlap 已提交
5013 5014
 * @p: target task
 * @preempt: whether task preemption is allowed or not
5015 5016 5017 5018
 *
 * It's the caller's job to ensure that the target task struct
 * can't go away on us before we can do any checks.
 *
5019
 * Return:
5020 5021 5022
 *	true (>0) if we indeed boosted the target task.
 *	false (0) if we failed to boost the target.
 *	-ESRCH if there's no task to yield to.
5023
 */
5024
int __sched yield_to(struct task_struct *p, bool preempt)
5025 5026 5027 5028
{
	struct task_struct *curr = current;
	struct rq *rq, *p_rq;
	unsigned long flags;
5029
	int yielded = 0;
5030 5031 5032 5033 5034 5035

	local_irq_save(flags);
	rq = this_rq();

again:
	p_rq = task_rq(p);
5036 5037 5038 5039 5040 5041 5042 5043 5044
	/*
	 * If we're the only runnable task on the rq and target rq also
	 * has only one task, there's absolutely no point in yielding.
	 */
	if (rq->nr_running == 1 && p_rq->nr_running == 1) {
		yielded = -ESRCH;
		goto out_irq;
	}

5045
	double_rq_lock(rq, p_rq);
5046
	if (task_rq(p) != p_rq) {
5047 5048 5049 5050 5051
		double_rq_unlock(rq, p_rq);
		goto again;
	}

	if (!curr->sched_class->yield_to_task)
5052
		goto out_unlock;
5053 5054

	if (curr->sched_class != p->sched_class)
5055
		goto out_unlock;
5056 5057

	if (task_running(p_rq, p) || p->state)
5058
		goto out_unlock;
5059 5060

	yielded = curr->sched_class->yield_to_task(rq, p, preempt);
5061
	if (yielded) {
5062
		schedstat_inc(rq->yld_count);
5063 5064 5065 5066 5067
		/*
		 * Make p's CPU reschedule; pick_next_entity takes care of
		 * fairness.
		 */
		if (preempt && rq != p_rq)
5068
			resched_curr(p_rq);
5069
	}
5070

5071
out_unlock:
5072
	double_rq_unlock(rq, p_rq);
5073
out_irq:
5074 5075
	local_irq_restore(flags);

5076
	if (yielded > 0)
5077 5078 5079 5080 5081 5082
		schedule();

	return yielded;
}
EXPORT_SYMBOL_GPL(yield_to);

5083 5084 5085 5086 5087 5088 5089 5090 5091 5092 5093 5094 5095 5096 5097
int io_schedule_prepare(void)
{
	int old_iowait = current->in_iowait;

	current->in_iowait = 1;
	blk_schedule_flush_plug(current);

	return old_iowait;
}

void io_schedule_finish(int token)
{
	current->in_iowait = token;
}

L
Linus Torvalds 已提交
5098
/*
I
Ingo Molnar 已提交
5099
 * This task is about to go to sleep on IO. Increment rq->nr_iowait so
L
Linus Torvalds 已提交
5100 5101 5102 5103
 * that process accounting knows that this is a task in IO wait state.
 */
long __sched io_schedule_timeout(long timeout)
{
5104
	int token;
L
Linus Torvalds 已提交
5105 5106
	long ret;

5107
	token = io_schedule_prepare();
L
Linus Torvalds 已提交
5108
	ret = schedule_timeout(timeout);
5109
	io_schedule_finish(token);
5110

L
Linus Torvalds 已提交
5111 5112
	return ret;
}
5113
EXPORT_SYMBOL(io_schedule_timeout);
L
Linus Torvalds 已提交
5114

5115 5116 5117 5118 5119 5120 5121 5122 5123 5124
void io_schedule(void)
{
	int token;

	token = io_schedule_prepare();
	schedule();
	io_schedule_finish(token);
}
EXPORT_SYMBOL(io_schedule);

L
Linus Torvalds 已提交
5125 5126 5127 5128
/**
 * sys_sched_get_priority_max - return maximum RT priority.
 * @policy: scheduling class.
 *
5129 5130 5131
 * Return: On success, this syscall returns the maximum
 * rt_priority that can be used by a given scheduling class.
 * On failure, a negative error code is returned.
L
Linus Torvalds 已提交
5132
 */
5133
SYSCALL_DEFINE1(sched_get_priority_max, int, policy)
L
Linus Torvalds 已提交
5134 5135 5136 5137 5138 5139 5140 5141
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = MAX_USER_RT_PRIO-1;
		break;
5142
	case SCHED_DEADLINE:
L
Linus Torvalds 已提交
5143
	case SCHED_NORMAL:
5144
	case SCHED_BATCH:
I
Ingo Molnar 已提交
5145
	case SCHED_IDLE:
L
Linus Torvalds 已提交
5146 5147 5148 5149 5150 5151 5152 5153 5154 5155
		ret = 0;
		break;
	}
	return ret;
}

/**
 * sys_sched_get_priority_min - return minimum RT priority.
 * @policy: scheduling class.
 *
5156 5157 5158
 * Return: On success, this syscall returns the minimum
 * rt_priority that can be used by a given scheduling class.
 * On failure, a negative error code is returned.
L
Linus Torvalds 已提交
5159
 */
5160
SYSCALL_DEFINE1(sched_get_priority_min, int, policy)
L
Linus Torvalds 已提交
5161 5162 5163 5164 5165 5166 5167 5168
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = 1;
		break;
5169
	case SCHED_DEADLINE:
L
Linus Torvalds 已提交
5170
	case SCHED_NORMAL:
5171
	case SCHED_BATCH:
I
Ingo Molnar 已提交
5172
	case SCHED_IDLE:
L
Linus Torvalds 已提交
5173 5174 5175 5176 5177 5178 5179 5180 5181 5182 5183 5184
		ret = 0;
	}
	return ret;
}

/**
 * sys_sched_rr_get_interval - return the default timeslice of a process.
 * @pid: pid of the process.
 * @interval: userspace pointer to the timeslice value.
 *
 * this syscall writes the default timeslice value of a given process
 * into the user-space timespec buffer. A value of '0' means infinity.
5185 5186 5187
 *
 * Return: On success, 0 and the timeslice is in @interval. Otherwise,
 * an error code.
L
Linus Torvalds 已提交
5188
 */
5189
SYSCALL_DEFINE2(sched_rr_get_interval, pid_t, pid,
5190
		struct timespec __user *, interval)
L
Linus Torvalds 已提交
5191
{
5192
	struct task_struct *p;
D
Dmitry Adamushko 已提交
5193
	unsigned int time_slice;
5194 5195
	struct rq_flags rf;
	struct timespec t;
5196
	struct rq *rq;
5197
	int retval;
L
Linus Torvalds 已提交
5198 5199

	if (pid < 0)
5200
		return -EINVAL;
L
Linus Torvalds 已提交
5201 5202

	retval = -ESRCH;
5203
	rcu_read_lock();
L
Linus Torvalds 已提交
5204 5205 5206 5207 5208 5209 5210 5211
	p = find_process_by_pid(pid);
	if (!p)
		goto out_unlock;

	retval = security_task_getscheduler(p);
	if (retval)
		goto out_unlock;

5212
	rq = task_rq_lock(p, &rf);
5213 5214 5215
	time_slice = 0;
	if (p->sched_class->get_rr_interval)
		time_slice = p->sched_class->get_rr_interval(rq, p);
5216
	task_rq_unlock(rq, p, &rf);
D
Dmitry Adamushko 已提交
5217

5218
	rcu_read_unlock();
D
Dmitry Adamushko 已提交
5219
	jiffies_to_timespec(time_slice, &t);
L
Linus Torvalds 已提交
5220 5221
	retval = copy_to_user(interval, &t, sizeof(t)) ? -EFAULT : 0;
	return retval;
5222

L
Linus Torvalds 已提交
5223
out_unlock:
5224
	rcu_read_unlock();
L
Linus Torvalds 已提交
5225 5226 5227
	return retval;
}

5228
static const char stat_nam[] = TASK_STATE_TO_CHAR_STR;
5229

5230
void sched_show_task(struct task_struct *p)
L
Linus Torvalds 已提交
5231 5232
{
	unsigned long free = 0;
5233
	int ppid;
5234
	unsigned long state = p->state;
L
Linus Torvalds 已提交
5235

5236 5237 5238
	/* Make sure the string lines up properly with the number of task states: */
	BUILD_BUG_ON(sizeof(TASK_STATE_TO_CHAR_STR)-1 != ilog2(TASK_STATE_MAX)+1);

5239 5240
	if (!try_get_task_stack(p))
		return;
5241 5242
	if (state)
		state = __ffs(state) + 1;
5243
	printk(KERN_INFO "%-15.15s %c", p->comm,
5244
		state < sizeof(stat_nam) - 1 ? stat_nam[state] : '?');
L
Linus Torvalds 已提交
5245
	if (state == TASK_RUNNING)
P
Peter Zijlstra 已提交
5246
		printk(KERN_CONT "  running task    ");
L
Linus Torvalds 已提交
5247
#ifdef CONFIG_DEBUG_STACK_USAGE
5248
	free = stack_not_used(p);
L
Linus Torvalds 已提交
5249
#endif
5250
	ppid = 0;
5251
	rcu_read_lock();
5252 5253
	if (pid_alive(p))
		ppid = task_pid_nr(rcu_dereference(p->real_parent));
5254
	rcu_read_unlock();
P
Peter Zijlstra 已提交
5255
	printk(KERN_CONT "%5lu %5d %6d 0x%08lx\n", free,
5256
		task_pid_nr(p), ppid,
5257
		(unsigned long)task_thread_info(p)->flags);
L
Linus Torvalds 已提交
5258

5259
	print_worker_info(KERN_INFO, p);
5260
	show_stack(p, NULL);
5261
	put_task_stack(p);
L
Linus Torvalds 已提交
5262 5263
}

I
Ingo Molnar 已提交
5264
void show_state_filter(unsigned long state_filter)
L
Linus Torvalds 已提交
5265
{
5266
	struct task_struct *g, *p;
L
Linus Torvalds 已提交
5267

5268
#if BITS_PER_LONG == 32
P
Peter Zijlstra 已提交
5269 5270
	printk(KERN_INFO
		"  task                PC stack   pid father\n");
L
Linus Torvalds 已提交
5271
#else
P
Peter Zijlstra 已提交
5272 5273
	printk(KERN_INFO
		"  task                        PC stack   pid father\n");
L
Linus Torvalds 已提交
5274
#endif
5275
	rcu_read_lock();
5276
	for_each_process_thread(g, p) {
L
Linus Torvalds 已提交
5277 5278
		/*
		 * reset the NMI-timeout, listing all files on a slow
L
Lucas De Marchi 已提交
5279
		 * console might take a lot of time:
5280 5281 5282
		 * Also, reset softlockup watchdogs on all CPUs, because
		 * another CPU might be blocked waiting for us to process
		 * an IPI.
L
Linus Torvalds 已提交
5283 5284
		 */
		touch_nmi_watchdog();
5285
		touch_all_softlockup_watchdogs();
I
Ingo Molnar 已提交
5286
		if (!state_filter || (p->state & state_filter))
5287
			sched_show_task(p);
5288
	}
L
Linus Torvalds 已提交
5289

I
Ingo Molnar 已提交
5290
#ifdef CONFIG_SCHED_DEBUG
5291 5292
	if (!state_filter)
		sysrq_sched_debug_show();
I
Ingo Molnar 已提交
5293
#endif
5294
	rcu_read_unlock();
I
Ingo Molnar 已提交
5295 5296 5297
	/*
	 * Only show locks if all tasks are dumped:
	 */
5298
	if (!state_filter)
I
Ingo Molnar 已提交
5299
		debug_show_all_locks();
L
Linus Torvalds 已提交
5300 5301
}

5302
void init_idle_bootup_task(struct task_struct *idle)
I
Ingo Molnar 已提交
5303
{
I
Ingo Molnar 已提交
5304
	idle->sched_class = &idle_sched_class;
I
Ingo Molnar 已提交
5305 5306
}

5307 5308 5309
/**
 * init_idle - set up an idle thread for a given CPU
 * @idle: task in question
I
Ingo Molnar 已提交
5310
 * @cpu: CPU the idle task belongs to
5311 5312 5313 5314
 *
 * NOTE: this function does not set the idle thread's NEED_RESCHED
 * flag, to make booting more robust.
 */
5315
void init_idle(struct task_struct *idle, int cpu)
L
Linus Torvalds 已提交
5316
{
5317
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
5318 5319
	unsigned long flags;

5320 5321
	raw_spin_lock_irqsave(&idle->pi_lock, flags);
	raw_spin_lock(&rq->lock);
5322

5323
	__sched_fork(0, idle);
5324
	idle->state = TASK_RUNNING;
I
Ingo Molnar 已提交
5325
	idle->se.exec_start = sched_clock();
5326
	idle->flags |= PF_IDLE;
I
Ingo Molnar 已提交
5327

5328 5329
	kasan_unpoison_task_stack(idle);

5330 5331 5332 5333 5334 5335 5336 5337 5338
#ifdef CONFIG_SMP
	/*
	 * Its possible that init_idle() gets called multiple times on a task,
	 * in that case do_set_cpus_allowed() will not do the right thing.
	 *
	 * And since this is boot we can forgo the serialization.
	 */
	set_cpus_allowed_common(idle, cpumask_of(cpu));
#endif
5339 5340
	/*
	 * We're having a chicken and egg problem, even though we are
I
Ingo Molnar 已提交
5341
	 * holding rq->lock, the CPU isn't yet set to this CPU so the
5342 5343 5344 5345 5346 5347 5348 5349
	 * lockdep check in task_group() will fail.
	 *
	 * Similar case to sched_fork(). / Alternatively we could
	 * use task_rq_lock() here and obtain the other rq->lock.
	 *
	 * Silence PROVE_RCU
	 */
	rcu_read_lock();
I
Ingo Molnar 已提交
5350
	__set_task_cpu(idle, cpu);
5351
	rcu_read_unlock();
L
Linus Torvalds 已提交
5352 5353

	rq->curr = rq->idle = idle;
5354
	idle->on_rq = TASK_ON_RQ_QUEUED;
5355
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
5356
	idle->on_cpu = 1;
5357
#endif
5358 5359
	raw_spin_unlock(&rq->lock);
	raw_spin_unlock_irqrestore(&idle->pi_lock, flags);
L
Linus Torvalds 已提交
5360 5361

	/* Set the preempt count _outside_ the spinlocks! */
5362
	init_idle_preempt_count(idle, cpu);
5363

I
Ingo Molnar 已提交
5364 5365 5366 5367
	/*
	 * The idle tasks have their own, simple scheduling class:
	 */
	idle->sched_class = &idle_sched_class;
5368
	ftrace_graph_init_idle_task(idle, cpu);
5369
	vtime_init_idle(idle, cpu);
5370
#ifdef CONFIG_SMP
5371 5372
	sprintf(idle->comm, "%s/%d", INIT_TASK_COMM, cpu);
#endif
I
Ingo Molnar 已提交
5373 5374
}

5375 5376 5377 5378 5379 5380 5381
int cpuset_cpumask_can_shrink(const struct cpumask *cur,
			      const struct cpumask *trial)
{
	int ret = 1, trial_cpus;
	struct dl_bw *cur_dl_b;
	unsigned long flags;

5382 5383 5384
	if (!cpumask_weight(cur))
		return ret;

5385
	rcu_read_lock_sched();
5386 5387 5388 5389 5390 5391 5392 5393
	cur_dl_b = dl_bw_of(cpumask_any(cur));
	trial_cpus = cpumask_weight(trial);

	raw_spin_lock_irqsave(&cur_dl_b->lock, flags);
	if (cur_dl_b->bw != -1 &&
	    cur_dl_b->bw * trial_cpus < cur_dl_b->total_bw)
		ret = 0;
	raw_spin_unlock_irqrestore(&cur_dl_b->lock, flags);
5394
	rcu_read_unlock_sched();
5395 5396 5397 5398

	return ret;
}

5399 5400 5401 5402 5403 5404 5405
int task_can_attach(struct task_struct *p,
		    const struct cpumask *cs_cpus_allowed)
{
	int ret = 0;

	/*
	 * Kthreads which disallow setaffinity shouldn't be moved
I
Ingo Molnar 已提交
5406
	 * to a new cpuset; we don't want to change their CPU
5407 5408 5409 5410 5411 5412 5413 5414 5415 5416 5417 5418 5419 5420 5421 5422
	 * affinity and isolating such threads by their set of
	 * allowed nodes is unnecessary.  Thus, cpusets are not
	 * applicable for such threads.  This prevents checking for
	 * success of set_cpus_allowed_ptr() on all attached tasks
	 * before cpus_allowed may be changed.
	 */
	if (p->flags & PF_NO_SETAFFINITY) {
		ret = -EINVAL;
		goto out;
	}

#ifdef CONFIG_SMP
	if (dl_task(p) && !cpumask_intersects(task_rq(p)->rd->span,
					      cs_cpus_allowed)) {
		unsigned int dest_cpu = cpumask_any_and(cpu_active_mask,
							cs_cpus_allowed);
5423
		struct dl_bw *dl_b;
5424 5425 5426 5427
		bool overflow;
		int cpus;
		unsigned long flags;

5428 5429
		rcu_read_lock_sched();
		dl_b = dl_bw_of(dest_cpu);
5430 5431 5432 5433 5434 5435 5436 5437 5438 5439 5440 5441 5442 5443 5444
		raw_spin_lock_irqsave(&dl_b->lock, flags);
		cpus = dl_bw_cpus(dest_cpu);
		overflow = __dl_overflow(dl_b, cpus, 0, p->dl.dl_bw);
		if (overflow)
			ret = -EBUSY;
		else {
			/*
			 * We reserve space for this task in the destination
			 * root_domain, as we can't fail after this point.
			 * We will free resources in the source root_domain
			 * later on (see set_cpus_allowed_dl()).
			 */
			__dl_add(dl_b, p->dl.dl_bw);
		}
		raw_spin_unlock_irqrestore(&dl_b->lock, flags);
5445
		rcu_read_unlock_sched();
5446 5447 5448 5449 5450 5451 5452

	}
#endif
out:
	return ret;
}

L
Linus Torvalds 已提交
5453 5454
#ifdef CONFIG_SMP

5455
bool sched_smp_initialized __read_mostly;
5456

5457 5458 5459 5460 5461 5462 5463 5464 5465 5466 5467 5468 5469 5470 5471
#ifdef CONFIG_NUMA_BALANCING
/* Migrate current task p to target_cpu */
int migrate_task_to(struct task_struct *p, int target_cpu)
{
	struct migration_arg arg = { p, target_cpu };
	int curr_cpu = task_cpu(p);

	if (curr_cpu == target_cpu)
		return 0;

	if (!cpumask_test_cpu(target_cpu, tsk_cpus_allowed(p)))
		return -EINVAL;

	/* TODO: This is not properly updating schedstats */

5472
	trace_sched_move_numa(p, curr_cpu, target_cpu);
5473 5474
	return stop_one_cpu(curr_cpu, migration_cpu_stop, &arg);
}
5475 5476 5477 5478 5479 5480 5481

/*
 * Requeue a task on a given node and accurately track the number of NUMA
 * tasks on the runqueues
 */
void sched_setnuma(struct task_struct *p, int nid)
{
5482
	bool queued, running;
5483 5484
	struct rq_flags rf;
	struct rq *rq;
5485

5486
	rq = task_rq_lock(p, &rf);
5487
	queued = task_on_rq_queued(p);
5488 5489
	running = task_current(rq, p);

5490
	if (queued)
5491
		dequeue_task(rq, p, DEQUEUE_SAVE);
5492
	if (running)
5493
		put_prev_task(rq, p);
5494 5495 5496

	p->numa_preferred_nid = nid;

5497
	if (queued)
5498
		enqueue_task(rq, p, ENQUEUE_RESTORE);
5499
	if (running)
5500
		set_curr_task(rq, p);
5501
	task_rq_unlock(rq, p, &rf);
5502
}
P
Peter Zijlstra 已提交
5503
#endif /* CONFIG_NUMA_BALANCING */
5504

L
Linus Torvalds 已提交
5505
#ifdef CONFIG_HOTPLUG_CPU
5506
/*
I
Ingo Molnar 已提交
5507
 * Ensure that the idle task is using init_mm right before its CPU goes
5508
 * offline.
5509
 */
5510
void idle_task_exit(void)
L
Linus Torvalds 已提交
5511
{
5512
	struct mm_struct *mm = current->active_mm;
5513

5514
	BUG_ON(cpu_online(smp_processor_id()));
5515

5516
	if (mm != &init_mm) {
5517
		switch_mm_irqs_off(mm, &init_mm, current);
5518 5519
		finish_arch_post_lock_switch();
	}
5520
	mmdrop(mm);
L
Linus Torvalds 已提交
5521 5522 5523
}

/*
5524 5525
 * Since this CPU is going 'away' for a while, fold any nr_active delta
 * we might have. Assumes we're called after migrate_tasks() so that the
5526 5527 5528
 * nr_active count is stable. We need to take the teardown thread which
 * is calling this into account, so we hand in adjust = 1 to the load
 * calculation.
5529 5530
 *
 * Also see the comment "Global load-average calculations".
L
Linus Torvalds 已提交
5531
 */
5532
static void calc_load_migrate(struct rq *rq)
L
Linus Torvalds 已提交
5533
{
5534
	long delta = calc_load_fold_active(rq, 1);
5535 5536
	if (delta)
		atomic_long_add(delta, &calc_load_tasks);
L
Linus Torvalds 已提交
5537 5538
}

5539 5540 5541 5542 5543 5544 5545 5546 5547 5548 5549 5550 5551 5552 5553 5554
static void put_prev_task_fake(struct rq *rq, struct task_struct *prev)
{
}

static const struct sched_class fake_sched_class = {
	.put_prev_task = put_prev_task_fake,
};

static struct task_struct fake_task = {
	/*
	 * Avoid pull_{rt,dl}_task()
	 */
	.prio = MAX_PRIO + 1,
	.sched_class = &fake_sched_class,
};

5555
/*
5556 5557 5558 5559 5560 5561
 * Migrate all tasks from the rq, sleeping tasks will be migrated by
 * try_to_wake_up()->select_task_rq().
 *
 * Called with rq->lock held even though we'er in stop_machine() and
 * there's no concurrency possible, we hold the required locks anyway
 * because of lock validation efforts.
L
Linus Torvalds 已提交
5562
 */
5563
static void migrate_tasks(struct rq *dead_rq)
L
Linus Torvalds 已提交
5564
{
5565
	struct rq *rq = dead_rq;
5566
	struct task_struct *next, *stop = rq->stop;
5567
	struct rq_flags rf;
5568
	int dest_cpu;
L
Linus Torvalds 已提交
5569 5570

	/*
5571 5572 5573 5574 5575 5576 5577
	 * Fudge the rq selection such that the below task selection loop
	 * doesn't get stuck on the currently eligible stop task.
	 *
	 * We're currently inside stop_machine() and the rq is either stuck
	 * in the stop_machine_cpu_stop() loop, or we're executing this code,
	 * either way we should never end up calling schedule() until we're
	 * done here.
L
Linus Torvalds 已提交
5578
	 */
5579
	rq->stop = NULL;
5580

5581 5582 5583 5584 5585
	/*
	 * put_prev_task() and pick_next_task() sched
	 * class method both need to have an up-to-date
	 * value of rq->clock[_task]
	 */
5586
	rq_pin_lock(rq, &rf);
5587
	update_rq_clock(rq);
5588
	rq_unpin_lock(rq, &rf);
5589

5590
	for (;;) {
5591 5592
		/*
		 * There's this thread running, bail when that's the only
I
Ingo Molnar 已提交
5593
		 * remaining thread:
5594 5595
		 */
		if (rq->nr_running == 1)
I
Ingo Molnar 已提交
5596
			break;
5597

5598
		/*
I
Ingo Molnar 已提交
5599
		 * pick_next_task() assumes pinned rq->lock:
5600
		 */
5601
		rq_repin_lock(rq, &rf);
5602
		next = pick_next_task(rq, &fake_task, &rf);
5603
		BUG_ON(!next);
D
Dmitry Adamushko 已提交
5604
		next->sched_class->put_prev_task(rq, next);
5605

W
Wanpeng Li 已提交
5606 5607 5608 5609 5610 5611 5612 5613 5614
		/*
		 * Rules for changing task_struct::cpus_allowed are holding
		 * both pi_lock and rq->lock, such that holding either
		 * stabilizes the mask.
		 *
		 * Drop rq->lock is not quite as disastrous as it usually is
		 * because !cpu_active at this point, which means load-balance
		 * will not interfere. Also, stop-machine.
		 */
5615
		rq_unpin_lock(rq, &rf);
W
Wanpeng Li 已提交
5616 5617 5618 5619 5620 5621 5622 5623 5624 5625 5626 5627 5628 5629
		raw_spin_unlock(&rq->lock);
		raw_spin_lock(&next->pi_lock);
		raw_spin_lock(&rq->lock);

		/*
		 * Since we're inside stop-machine, _nothing_ should have
		 * changed the task, WARN if weird stuff happened, because in
		 * that case the above rq->lock drop is a fail too.
		 */
		if (WARN_ON(task_rq(next) != rq || !task_on_rq_queued(next))) {
			raw_spin_unlock(&next->pi_lock);
			continue;
		}

5630
		/* Find suitable destination for @next, with force if needed. */
5631
		dest_cpu = select_fallback_rq(dead_rq->cpu, next);
5632

5633 5634 5635 5636 5637 5638
		rq = __migrate_task(rq, next, dest_cpu);
		if (rq != dead_rq) {
			raw_spin_unlock(&rq->lock);
			rq = dead_rq;
			raw_spin_lock(&rq->lock);
		}
W
Wanpeng Li 已提交
5639
		raw_spin_unlock(&next->pi_lock);
L
Linus Torvalds 已提交
5640
	}
5641

5642
	rq->stop = stop;
5643
}
L
Linus Torvalds 已提交
5644 5645
#endif /* CONFIG_HOTPLUG_CPU */

5646
void set_rq_online(struct rq *rq)
5647 5648 5649 5650
{
	if (!rq->online) {
		const struct sched_class *class;

5651
		cpumask_set_cpu(rq->cpu, rq->rd->online);
5652 5653 5654 5655 5656 5657 5658 5659 5660
		rq->online = 1;

		for_each_class(class) {
			if (class->rq_online)
				class->rq_online(rq);
		}
	}
}

5661
void set_rq_offline(struct rq *rq)
5662 5663 5664 5665 5666 5667 5668 5669 5670
{
	if (rq->online) {
		const struct sched_class *class;

		for_each_class(class) {
			if (class->rq_offline)
				class->rq_offline(rq);
		}

5671
		cpumask_clear_cpu(rq->cpu, rq->rd->online);
5672 5673 5674 5675
		rq->online = 0;
	}
}

5676
static void set_cpu_rq_start_time(unsigned int cpu)
L
Linus Torvalds 已提交
5677
{
5678
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
5679

5680 5681 5682
	rq->age_stamp = sched_clock_cpu(cpu);
}

I
Ingo Molnar 已提交
5683 5684 5685 5686
/*
 * used to mark begin/end of suspend/resume:
 */
static int num_cpus_frozen;
5687

L
Linus Torvalds 已提交
5688
/*
5689 5690 5691
 * Update cpusets according to cpu_active mask.  If cpusets are
 * disabled, cpuset_update_active_cpus() becomes a simple wrapper
 * around partition_sched_domains().
5692 5693 5694
 *
 * If we come here as part of a suspend/resume, don't touch cpusets because we
 * want to restore it back to its original state upon resume anyway.
L
Linus Torvalds 已提交
5695
 */
5696
static void cpuset_cpu_active(void)
5697
{
5698
	if (cpuhp_tasks_frozen) {
5699 5700 5701 5702 5703 5704 5705 5706 5707
		/*
		 * num_cpus_frozen tracks how many CPUs are involved in suspend
		 * resume sequence. As long as this is not the last online
		 * operation in the resume sequence, just build a single sched
		 * domain, ignoring cpusets.
		 */
		num_cpus_frozen--;
		if (likely(num_cpus_frozen)) {
			partition_sched_domains(1, NULL, NULL);
5708
			return;
5709 5710 5711 5712 5713 5714
		}
		/*
		 * This is the last CPU online operation. So fall through and
		 * restore the original sched domains by considering the
		 * cpuset configurations.
		 */
5715
	}
5716
	cpuset_update_active_cpus(true);
5717
}
5718

5719
static int cpuset_cpu_inactive(unsigned int cpu)
5720
{
5721 5722
	unsigned long flags;
	struct dl_bw *dl_b;
5723 5724
	bool overflow;
	int cpus;
5725

5726
	if (!cpuhp_tasks_frozen) {
5727 5728
		rcu_read_lock_sched();
		dl_b = dl_bw_of(cpu);
5729

5730 5731 5732 5733
		raw_spin_lock_irqsave(&dl_b->lock, flags);
		cpus = dl_bw_cpus(cpu);
		overflow = __dl_overflow(dl_b, cpus, 0, 0);
		raw_spin_unlock_irqrestore(&dl_b->lock, flags);
5734

5735
		rcu_read_unlock_sched();
5736

5737
		if (overflow)
5738
			return -EBUSY;
5739
		cpuset_update_active_cpus(false);
5740
	} else {
5741 5742
		num_cpus_frozen++;
		partition_sched_domains(1, NULL, NULL);
5743
	}
5744
	return 0;
5745 5746
}

5747
int sched_cpu_activate(unsigned int cpu)
5748
{
5749 5750 5751
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

5752
	set_cpu_active(cpu, true);
5753

5754
	if (sched_smp_initialized) {
5755
		sched_domains_numa_masks_set(cpu);
5756
		cpuset_cpu_active();
5757
	}
5758 5759 5760 5761 5762

	/*
	 * Put the rq online, if not already. This happens:
	 *
	 * 1) In the early boot process, because we build the real domains
I
Ingo Molnar 已提交
5763
	 *    after all CPUs have been brought up.
5764 5765 5766 5767 5768 5769 5770 5771 5772 5773 5774 5775 5776
	 *
	 * 2) At runtime, if cpuset_cpu_active() fails to rebuild the
	 *    domains.
	 */
	raw_spin_lock_irqsave(&rq->lock, flags);
	if (rq->rd) {
		BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
		set_rq_online(rq);
	}
	raw_spin_unlock_irqrestore(&rq->lock, flags);

	update_max_interval();

5777
	return 0;
5778 5779
}

5780
int sched_cpu_deactivate(unsigned int cpu)
5781 5782 5783
{
	int ret;

5784
	set_cpu_active(cpu, false);
5785 5786 5787 5788 5789 5790 5791 5792 5793 5794 5795 5796 5797 5798
	/*
	 * We've cleared cpu_active_mask, wait for all preempt-disabled and RCU
	 * users of this state to go away such that all new such users will
	 * observe it.
	 *
	 * For CONFIG_PREEMPT we have preemptible RCU and its sync_rcu() might
	 * not imply sync_sched(), so wait for both.
	 *
	 * Do sync before park smpboot threads to take care the rcu boost case.
	 */
	if (IS_ENABLED(CONFIG_PREEMPT))
		synchronize_rcu_mult(call_rcu, call_rcu_sched);
	else
		synchronize_rcu();
5799 5800 5801 5802 5803 5804 5805 5806

	if (!sched_smp_initialized)
		return 0;

	ret = cpuset_cpu_inactive(cpu);
	if (ret) {
		set_cpu_active(cpu, true);
		return ret;
5807
	}
5808 5809
	sched_domains_numa_masks_clear(cpu);
	return 0;
5810 5811
}

5812 5813 5814 5815 5816 5817 5818 5819
static void sched_rq_cpu_starting(unsigned int cpu)
{
	struct rq *rq = cpu_rq(cpu);

	rq->calc_load_update = calc_load_update;
	update_max_interval();
}

5820 5821 5822
int sched_cpu_starting(unsigned int cpu)
{
	set_cpu_rq_start_time(cpu);
5823
	sched_rq_cpu_starting(cpu);
5824
	return 0;
5825 5826
}

5827 5828 5829 5830 5831 5832 5833 5834 5835 5836 5837 5838 5839 5840 5841 5842 5843 5844
#ifdef CONFIG_HOTPLUG_CPU
int sched_cpu_dying(unsigned int cpu)
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

	/* Handle pending wakeups and then migrate everything off */
	sched_ttwu_pending();
	raw_spin_lock_irqsave(&rq->lock, flags);
	if (rq->rd) {
		BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
		set_rq_offline(rq);
	}
	migrate_tasks(rq);
	BUG_ON(rq->nr_running != 1);
	raw_spin_unlock_irqrestore(&rq->lock, flags);
	calc_load_migrate(rq);
	update_max_interval();
5845
	nohz_balance_exit_idle(cpu);
5846
	hrtick_clear(rq);
5847 5848 5849 5850
	return 0;
}
#endif

P
Peter Zijlstra 已提交
5851 5852 5853 5854 5855 5856 5857 5858 5859 5860 5861 5862 5863 5864 5865 5866
#ifdef CONFIG_SCHED_SMT
DEFINE_STATIC_KEY_FALSE(sched_smt_present);

static void sched_init_smt(void)
{
	/*
	 * We've enumerated all CPUs and will assume that if any CPU
	 * has SMT siblings, CPU0 will too.
	 */
	if (cpumask_weight(cpu_smt_mask(0)) > 1)
		static_branch_enable(&sched_smt_present);
}
#else
static inline void sched_init_smt(void) { }
#endif

L
Linus Torvalds 已提交
5867 5868
void __init sched_init_smp(void)
{
5869 5870 5871
	cpumask_var_t non_isolated_cpus;

	alloc_cpumask_var(&non_isolated_cpus, GFP_KERNEL);
5872
	alloc_cpumask_var(&fallback_doms, GFP_KERNEL);
5873

5874 5875
	sched_init_numa();

5876 5877
	/*
	 * There's no userspace yet to cause hotplug operations; hence all the
I
Ingo Molnar 已提交
5878
	 * CPU masks are stable and all blatant races in the below code cannot
5879 5880
	 * happen.
	 */
5881
	mutex_lock(&sched_domains_mutex);
5882
	init_sched_domains(cpu_active_mask);
5883 5884 5885
	cpumask_andnot(non_isolated_cpus, cpu_possible_mask, cpu_isolated_map);
	if (cpumask_empty(non_isolated_cpus))
		cpumask_set_cpu(smp_processor_id(), non_isolated_cpus);
5886
	mutex_unlock(&sched_domains_mutex);
5887

5888
	/* Move init over to a non-isolated CPU */
5889
	if (set_cpus_allowed_ptr(current, non_isolated_cpus) < 0)
5890
		BUG();
I
Ingo Molnar 已提交
5891
	sched_init_granularity();
5892
	free_cpumask_var(non_isolated_cpus);
5893

5894
	init_sched_rt_class();
5895
	init_sched_dl_class();
P
Peter Zijlstra 已提交
5896 5897

	sched_init_smt();
5898
	sched_clock_init_late();
P
Peter Zijlstra 已提交
5899

5900
	sched_smp_initialized = true;
L
Linus Torvalds 已提交
5901
}
5902 5903 5904

static int __init migration_init(void)
{
5905
	sched_rq_cpu_starting(smp_processor_id());
5906
	return 0;
L
Linus Torvalds 已提交
5907
}
5908 5909
early_initcall(migration_init);

L
Linus Torvalds 已提交
5910 5911 5912
#else
void __init sched_init_smp(void)
{
I
Ingo Molnar 已提交
5913
	sched_init_granularity();
5914
	sched_clock_init_late();
L
Linus Torvalds 已提交
5915 5916 5917 5918 5919 5920 5921 5922 5923 5924
}
#endif /* CONFIG_SMP */

int in_sched_functions(unsigned long addr)
{
	return in_lock_functions(addr) ||
		(addr >= (unsigned long)__sched_text_start
		&& addr < (unsigned long)__sched_text_end);
}

5925
#ifdef CONFIG_CGROUP_SCHED
5926 5927 5928 5929
/*
 * Default task group.
 * Every task in system belongs to this group at bootup.
 */
5930
struct task_group root_task_group;
5931
LIST_HEAD(task_groups);
5932 5933 5934

/* Cacheline aligned slab cache for task_group */
static struct kmem_cache *task_group_cache __read_mostly;
5935
#endif
P
Peter Zijlstra 已提交
5936

5937
DECLARE_PER_CPU(cpumask_var_t, load_balance_mask);
5938
DECLARE_PER_CPU(cpumask_var_t, select_idle_mask);
P
Peter Zijlstra 已提交
5939

5940 5941 5942 5943 5944 5945 5946 5947 5948 5949 5950 5951 5952
#define WAIT_TABLE_BITS 8
#define WAIT_TABLE_SIZE (1 << WAIT_TABLE_BITS)
static wait_queue_head_t bit_wait_table[WAIT_TABLE_SIZE] __cacheline_aligned;

wait_queue_head_t *bit_waitqueue(void *word, int bit)
{
	const int shift = BITS_PER_LONG == 32 ? 5 : 6;
	unsigned long val = (unsigned long)word << shift | bit;

	return bit_wait_table + hash_long(val, WAIT_TABLE_BITS);
}
EXPORT_SYMBOL(bit_waitqueue);

L
Linus Torvalds 已提交
5953 5954
void __init sched_init(void)
{
I
Ingo Molnar 已提交
5955
	int i, j;
5956 5957
	unsigned long alloc_size = 0, ptr;

5958 5959
	sched_clock_init();

5960 5961 5962
	for (i = 0; i < WAIT_TABLE_SIZE; i++)
		init_waitqueue_head(bit_wait_table + i);

5963 5964 5965 5966 5967 5968 5969
#ifdef CONFIG_FAIR_GROUP_SCHED
	alloc_size += 2 * nr_cpu_ids * sizeof(void **);
#endif
#ifdef CONFIG_RT_GROUP_SCHED
	alloc_size += 2 * nr_cpu_ids * sizeof(void **);
#endif
	if (alloc_size) {
5970
		ptr = (unsigned long)kzalloc(alloc_size, GFP_NOWAIT);
5971 5972

#ifdef CONFIG_FAIR_GROUP_SCHED
5973
		root_task_group.se = (struct sched_entity **)ptr;
5974 5975
		ptr += nr_cpu_ids * sizeof(void **);

5976
		root_task_group.cfs_rq = (struct cfs_rq **)ptr;
5977
		ptr += nr_cpu_ids * sizeof(void **);
5978

5979
#endif /* CONFIG_FAIR_GROUP_SCHED */
5980
#ifdef CONFIG_RT_GROUP_SCHED
5981
		root_task_group.rt_se = (struct sched_rt_entity **)ptr;
5982 5983
		ptr += nr_cpu_ids * sizeof(void **);

5984
		root_task_group.rt_rq = (struct rt_rq **)ptr;
5985 5986
		ptr += nr_cpu_ids * sizeof(void **);

5987
#endif /* CONFIG_RT_GROUP_SCHED */
5988
	}
5989
#ifdef CONFIG_CPUMASK_OFFSTACK
5990 5991 5992
	for_each_possible_cpu(i) {
		per_cpu(load_balance_mask, i) = (cpumask_var_t)kzalloc_node(
			cpumask_size(), GFP_KERNEL, cpu_to_node(i));
5993 5994
		per_cpu(select_idle_mask, i) = (cpumask_var_t)kzalloc_node(
			cpumask_size(), GFP_KERNEL, cpu_to_node(i));
5995
	}
5996
#endif /* CONFIG_CPUMASK_OFFSTACK */
I
Ingo Molnar 已提交
5997

I
Ingo Molnar 已提交
5998 5999
	init_rt_bandwidth(&def_rt_bandwidth, global_rt_period(), global_rt_runtime());
	init_dl_bandwidth(&def_dl_bandwidth, global_rt_period(), global_rt_runtime());
6000

G
Gregory Haskins 已提交
6001 6002 6003 6004
#ifdef CONFIG_SMP
	init_defrootdomain();
#endif

6005
#ifdef CONFIG_RT_GROUP_SCHED
6006
	init_rt_bandwidth(&root_task_group.rt_bandwidth,
6007
			global_rt_period(), global_rt_runtime());
6008
#endif /* CONFIG_RT_GROUP_SCHED */
6009

D
Dhaval Giani 已提交
6010
#ifdef CONFIG_CGROUP_SCHED
6011 6012
	task_group_cache = KMEM_CACHE(task_group, 0);

6013 6014
	list_add(&root_task_group.list, &task_groups);
	INIT_LIST_HEAD(&root_task_group.children);
6015
	INIT_LIST_HEAD(&root_task_group.siblings);
6016
	autogroup_init(&init_task);
D
Dhaval Giani 已提交
6017
#endif /* CONFIG_CGROUP_SCHED */
P
Peter Zijlstra 已提交
6018

6019
	for_each_possible_cpu(i) {
6020
		struct rq *rq;
L
Linus Torvalds 已提交
6021 6022

		rq = cpu_rq(i);
6023
		raw_spin_lock_init(&rq->lock);
N
Nick Piggin 已提交
6024
		rq->nr_running = 0;
6025 6026
		rq->calc_load_active = 0;
		rq->calc_load_update = jiffies + LOAD_FREQ;
6027
		init_cfs_rq(&rq->cfs);
6028 6029
		init_rt_rq(&rq->rt);
		init_dl_rq(&rq->dl);
I
Ingo Molnar 已提交
6030
#ifdef CONFIG_FAIR_GROUP_SCHED
6031
		root_task_group.shares = ROOT_TASK_GROUP_LOAD;
P
Peter Zijlstra 已提交
6032
		INIT_LIST_HEAD(&rq->leaf_cfs_rq_list);
6033
		rq->tmp_alone_branch = &rq->leaf_cfs_rq_list;
D
Dhaval Giani 已提交
6034
		/*
I
Ingo Molnar 已提交
6035
		 * How much CPU bandwidth does root_task_group get?
D
Dhaval Giani 已提交
6036 6037
		 *
		 * In case of task-groups formed thr' the cgroup filesystem, it
I
Ingo Molnar 已提交
6038 6039
		 * gets 100% of the CPU resources in the system. This overall
		 * system CPU resource is divided among the tasks of
6040
		 * root_task_group and its child task-groups in a fair manner,
D
Dhaval Giani 已提交
6041 6042 6043
		 * based on each entity's (task or task-group's) weight
		 * (se->load.weight).
		 *
6044
		 * In other words, if root_task_group has 10 tasks of weight
D
Dhaval Giani 已提交
6045
		 * 1024) and two child groups A0 and A1 (of weight 1024 each),
I
Ingo Molnar 已提交
6046
		 * then A0's share of the CPU resource is:
D
Dhaval Giani 已提交
6047
		 *
6048
		 *	A0's bandwidth = 1024 / (10*1024 + 1024 + 1024) = 8.33%
D
Dhaval Giani 已提交
6049
		 *
6050 6051
		 * We achieve this by letting root_task_group's tasks sit
		 * directly in rq->cfs (i.e root_task_group->se[] = NULL).
D
Dhaval Giani 已提交
6052
		 */
6053
		init_cfs_bandwidth(&root_task_group.cfs_bandwidth);
6054
		init_tg_cfs_entry(&root_task_group, &rq->cfs, NULL, i, NULL);
D
Dhaval Giani 已提交
6055 6056 6057
#endif /* CONFIG_FAIR_GROUP_SCHED */

		rq->rt.rt_runtime = def_rt_bandwidth.rt_runtime;
6058
#ifdef CONFIG_RT_GROUP_SCHED
6059
		init_tg_rt_entry(&root_task_group, &rq->rt, NULL, i, NULL);
I
Ingo Molnar 已提交
6060
#endif
L
Linus Torvalds 已提交
6061

I
Ingo Molnar 已提交
6062 6063
		for (j = 0; j < CPU_LOAD_IDX_MAX; j++)
			rq->cpu_load[j] = 0;
6064

L
Linus Torvalds 已提交
6065
#ifdef CONFIG_SMP
N
Nick Piggin 已提交
6066
		rq->sd = NULL;
G
Gregory Haskins 已提交
6067
		rq->rd = NULL;
6068
		rq->cpu_capacity = rq->cpu_capacity_orig = SCHED_CAPACITY_SCALE;
6069
		rq->balance_callback = NULL;
L
Linus Torvalds 已提交
6070
		rq->active_balance = 0;
I
Ingo Molnar 已提交
6071
		rq->next_balance = jiffies;
L
Linus Torvalds 已提交
6072
		rq->push_cpu = 0;
6073
		rq->cpu = i;
6074
		rq->online = 0;
6075 6076
		rq->idle_stamp = 0;
		rq->avg_idle = 2*sysctl_sched_migration_cost;
6077
		rq->max_idle_balance_cost = sysctl_sched_migration_cost;
6078 6079 6080

		INIT_LIST_HEAD(&rq->cfs_tasks);

6081
		rq_attach_root(rq, &def_root_domain);
6082
#ifdef CONFIG_NO_HZ_COMMON
6083
		rq->last_load_update_tick = jiffies;
6084
		rq->nohz_flags = 0;
6085
#endif
6086 6087 6088
#ifdef CONFIG_NO_HZ_FULL
		rq->last_sched_tick = 0;
#endif
6089
#endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
6090
		init_rq_hrtick(rq);
L
Linus Torvalds 已提交
6091 6092 6093
		atomic_set(&rq->nr_iowait, 0);
	}

6094
	set_load_weight(&init_task);
6095

L
Linus Torvalds 已提交
6096 6097 6098
	/*
	 * The boot idle thread does lazy MMU switching as well:
	 */
V
Vegard Nossum 已提交
6099
	mmgrab(&init_mm);
L
Linus Torvalds 已提交
6100 6101 6102 6103 6104 6105 6106 6107 6108
	enter_lazy_tlb(&init_mm, current);

	/*
	 * Make us the idle thread. Technically, schedule() should not be
	 * called from this thread, however somewhere below it might be,
	 * but because we are the idle thread, we just pick up running again
	 * when this runqueue becomes "idle".
	 */
	init_idle(current, smp_processor_id());
6109 6110 6111

	calc_load_update = jiffies + LOAD_FREQ;

6112
#ifdef CONFIG_SMP
6113
	zalloc_cpumask_var(&sched_domains_tmpmask, GFP_NOWAIT);
R
Rusty Russell 已提交
6114 6115 6116
	/* May be allocated at isolcpus cmdline parse time */
	if (cpu_isolated_map == NULL)
		zalloc_cpumask_var(&cpu_isolated_map, GFP_NOWAIT);
6117
	idle_thread_set_boot_cpu();
6118
	set_cpu_rq_start_time(smp_processor_id());
6119 6120
#endif
	init_sched_fair_class();
6121

6122 6123
	init_schedstats();

6124
	scheduler_running = 1;
L
Linus Torvalds 已提交
6125 6126
}

6127
#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
6128 6129
static inline int preempt_count_equals(int preempt_offset)
{
6130
	int nested = preempt_count() + rcu_preempt_depth();
6131

A
Arnd Bergmann 已提交
6132
	return (nested == preempt_offset);
6133 6134
}

6135
void __might_sleep(const char *file, int line, int preempt_offset)
L
Linus Torvalds 已提交
6136
{
P
Peter Zijlstra 已提交
6137 6138 6139 6140 6141
	/*
	 * Blocking primitives will set (and therefore destroy) current->state,
	 * since we will exit with TASK_RUNNING make sure we enter with it,
	 * otherwise we will destroy state.
	 */
6142
	WARN_ONCE(current->state != TASK_RUNNING && current->task_state_change,
P
Peter Zijlstra 已提交
6143 6144 6145 6146
			"do not call blocking ops when !TASK_RUNNING; "
			"state=%lx set at [<%p>] %pS\n",
			current->state,
			(void *)current->task_state_change,
6147
			(void *)current->task_state_change);
P
Peter Zijlstra 已提交
6148

6149 6150 6151 6152 6153
	___might_sleep(file, line, preempt_offset);
}
EXPORT_SYMBOL(__might_sleep);

void ___might_sleep(const char *file, int line, int preempt_offset)
L
Linus Torvalds 已提交
6154
{
I
Ingo Molnar 已提交
6155 6156 6157
	/* Ratelimiting timestamp: */
	static unsigned long prev_jiffy;

6158
	unsigned long preempt_disable_ip;
L
Linus Torvalds 已提交
6159

I
Ingo Molnar 已提交
6160 6161 6162
	/* WARN_ON_ONCE() by default, no rate limit required: */
	rcu_sleep_check();

6163 6164
	if ((preempt_count_equals(preempt_offset) && !irqs_disabled() &&
	     !is_idle_task(current)) ||
6165
	    system_state != SYSTEM_RUNNING || oops_in_progress)
I
Ingo Molnar 已提交
6166 6167 6168 6169 6170
		return;
	if (time_before(jiffies, prev_jiffy + HZ) && prev_jiffy)
		return;
	prev_jiffy = jiffies;

I
Ingo Molnar 已提交
6171
	/* Save this before calling printk(), since that will clobber it: */
6172 6173
	preempt_disable_ip = get_preempt_disable_ip(current);

P
Peter Zijlstra 已提交
6174 6175 6176 6177 6178 6179 6180
	printk(KERN_ERR
		"BUG: sleeping function called from invalid context at %s:%d\n",
			file, line);
	printk(KERN_ERR
		"in_atomic(): %d, irqs_disabled(): %d, pid: %d, name: %s\n",
			in_atomic(), irqs_disabled(),
			current->pid, current->comm);
I
Ingo Molnar 已提交
6181

6182 6183 6184
	if (task_stack_end_corrupted(current))
		printk(KERN_EMERG "Thread overran stack, or stack corrupted\n");

I
Ingo Molnar 已提交
6185 6186 6187
	debug_show_held_locks(current);
	if (irqs_disabled())
		print_irqtrace_events(current);
6188 6189
	if (IS_ENABLED(CONFIG_DEBUG_PREEMPT)
	    && !preempt_count_equals(preempt_offset)) {
6190
		pr_err("Preemption disabled at:");
6191
		print_ip_sym(preempt_disable_ip);
6192 6193
		pr_cont("\n");
	}
I
Ingo Molnar 已提交
6194
	dump_stack();
6195
	add_taint(TAINT_WARN, LOCKDEP_STILL_OK);
L
Linus Torvalds 已提交
6196
}
6197
EXPORT_SYMBOL(___might_sleep);
L
Linus Torvalds 已提交
6198 6199 6200
#endif

#ifdef CONFIG_MAGIC_SYSRQ
6201
void normalize_rt_tasks(void)
6202
{
6203
	struct task_struct *g, *p;
6204 6205 6206
	struct sched_attr attr = {
		.sched_policy = SCHED_NORMAL,
	};
L
Linus Torvalds 已提交
6207

6208
	read_lock(&tasklist_lock);
6209
	for_each_process_thread(g, p) {
6210 6211 6212
		/*
		 * Only normalize user tasks:
		 */
6213
		if (p->flags & PF_KTHREAD)
6214 6215
			continue;

6216 6217 6218 6219
		p->se.exec_start = 0;
		schedstat_set(p->se.statistics.wait_start,  0);
		schedstat_set(p->se.statistics.sleep_start, 0);
		schedstat_set(p->se.statistics.block_start, 0);
I
Ingo Molnar 已提交
6220

6221
		if (!dl_task(p) && !rt_task(p)) {
I
Ingo Molnar 已提交
6222 6223 6224 6225
			/*
			 * Renice negative nice level userspace
			 * tasks back to 0:
			 */
6226
			if (task_nice(p) < 0)
I
Ingo Molnar 已提交
6227
				set_user_nice(p, 0);
L
Linus Torvalds 已提交
6228
			continue;
I
Ingo Molnar 已提交
6229
		}
L
Linus Torvalds 已提交
6230

6231
		__sched_setscheduler(p, &attr, false, false);
6232
	}
6233
	read_unlock(&tasklist_lock);
L
Linus Torvalds 已提交
6234 6235 6236
}

#endif /* CONFIG_MAGIC_SYSRQ */
6237

6238
#if defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB)
6239
/*
6240
 * These functions are only useful for the IA64 MCA handling, or kdb.
6241 6242 6243 6244 6245 6246 6247 6248 6249
 *
 * They can only be called when the whole system has been
 * stopped - every CPU needs to be quiescent, and no scheduling
 * activity can take place. Using them for anything else would
 * be a serious bug, and as a result, they aren't even visible
 * under any other configuration.
 */

/**
I
Ingo Molnar 已提交
6250
 * curr_task - return the current task for a given CPU.
6251 6252 6253
 * @cpu: the processor in question.
 *
 * ONLY VALID WHEN THE WHOLE SYSTEM IS STOPPED!
6254 6255
 *
 * Return: The current task for @cpu.
6256
 */
6257
struct task_struct *curr_task(int cpu)
6258 6259 6260 6261
{
	return cpu_curr(cpu);
}

6262 6263 6264
#endif /* defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB) */

#ifdef CONFIG_IA64
6265
/**
I
Ingo Molnar 已提交
6266
 * set_curr_task - set the current task for a given CPU.
6267 6268 6269 6270
 * @cpu: the processor in question.
 * @p: the task pointer to set.
 *
 * Description: This function must only be used when non-maskable interrupts
I
Ingo Molnar 已提交
6271
 * are serviced on a separate stack. It allows the architecture to switch the
I
Ingo Molnar 已提交
6272
 * notion of the current task on a CPU in a non-blocking manner. This function
6273 6274 6275 6276 6277 6278 6279
 * must be called with all CPU's synchronized, and interrupts disabled, the
 * and caller must save the original value of the current task (see
 * curr_task() above) and restore that value before reenabling interrupts and
 * re-starting the system.
 *
 * ONLY VALID WHEN THE WHOLE SYSTEM IS STOPPED!
 */
6280
void ia64_set_curr_task(int cpu, struct task_struct *p)
6281 6282 6283 6284 6285
{
	cpu_curr(cpu) = p;
}

#endif
S
Srivatsa Vaddagiri 已提交
6286

D
Dhaval Giani 已提交
6287
#ifdef CONFIG_CGROUP_SCHED
6288 6289 6290
/* task_group_lock serializes the addition/removal of task groups */
static DEFINE_SPINLOCK(task_group_lock);

6291
static void sched_free_group(struct task_group *tg)
6292 6293 6294
{
	free_fair_sched_group(tg);
	free_rt_sched_group(tg);
6295
	autogroup_free(tg);
6296
	kmem_cache_free(task_group_cache, tg);
6297 6298 6299
}

/* allocate runqueue etc for a new task group */
6300
struct task_group *sched_create_group(struct task_group *parent)
6301 6302 6303
{
	struct task_group *tg;

6304
	tg = kmem_cache_alloc(task_group_cache, GFP_KERNEL | __GFP_ZERO);
6305 6306 6307
	if (!tg)
		return ERR_PTR(-ENOMEM);

6308
	if (!alloc_fair_sched_group(tg, parent))
6309 6310
		goto err;

6311
	if (!alloc_rt_sched_group(tg, parent))
6312 6313
		goto err;

6314 6315 6316
	return tg;

err:
6317
	sched_free_group(tg);
6318 6319 6320 6321 6322 6323 6324
	return ERR_PTR(-ENOMEM);
}

void sched_online_group(struct task_group *tg, struct task_group *parent)
{
	unsigned long flags;

6325
	spin_lock_irqsave(&task_group_lock, flags);
P
Peter Zijlstra 已提交
6326
	list_add_rcu(&tg->list, &task_groups);
P
Peter Zijlstra 已提交
6327

I
Ingo Molnar 已提交
6328 6329
	/* Root should already exist: */
	WARN_ON(!parent);
P
Peter Zijlstra 已提交
6330 6331 6332

	tg->parent = parent;
	INIT_LIST_HEAD(&tg->children);
6333
	list_add_rcu(&tg->siblings, &parent->children);
6334
	spin_unlock_irqrestore(&task_group_lock, flags);
6335 6336

	online_fair_sched_group(tg);
S
Srivatsa Vaddagiri 已提交
6337 6338
}

6339
/* rcu callback to free various structures associated with a task group */
6340
static void sched_free_group_rcu(struct rcu_head *rhp)
S
Srivatsa Vaddagiri 已提交
6341
{
I
Ingo Molnar 已提交
6342
	/* Now it should be safe to free those cfs_rqs: */
6343
	sched_free_group(container_of(rhp, struct task_group, rcu));
S
Srivatsa Vaddagiri 已提交
6344 6345
}

6346
void sched_destroy_group(struct task_group *tg)
6347
{
I
Ingo Molnar 已提交
6348
	/* Wait for possible concurrent references to cfs_rqs complete: */
6349
	call_rcu(&tg->rcu, sched_free_group_rcu);
6350 6351 6352
}

void sched_offline_group(struct task_group *tg)
S
Srivatsa Vaddagiri 已提交
6353
{
6354
	unsigned long flags;
S
Srivatsa Vaddagiri 已提交
6355

I
Ingo Molnar 已提交
6356
	/* End participation in shares distribution: */
6357
	unregister_fair_sched_group(tg);
6358 6359

	spin_lock_irqsave(&task_group_lock, flags);
P
Peter Zijlstra 已提交
6360
	list_del_rcu(&tg->list);
P
Peter Zijlstra 已提交
6361
	list_del_rcu(&tg->siblings);
6362
	spin_unlock_irqrestore(&task_group_lock, flags);
S
Srivatsa Vaddagiri 已提交
6363 6364
}

6365
static void sched_change_group(struct task_struct *tsk, int type)
S
Srivatsa Vaddagiri 已提交
6366
{
P
Peter Zijlstra 已提交
6367
	struct task_group *tg;
S
Srivatsa Vaddagiri 已提交
6368

6369 6370 6371 6372 6373 6374
	/*
	 * All callers are synchronized by task_rq_lock(); we do not use RCU
	 * which is pointless here. Thus, we pass "true" to task_css_check()
	 * to prevent lockdep warnings.
	 */
	tg = container_of(task_css_check(tsk, cpu_cgrp_id, true),
P
Peter Zijlstra 已提交
6375 6376 6377 6378
			  struct task_group, css);
	tg = autogroup_task_group(tsk, tg);
	tsk->sched_task_group = tg;

P
Peter Zijlstra 已提交
6379
#ifdef CONFIG_FAIR_GROUP_SCHED
6380 6381
	if (tsk->sched_class->task_change_group)
		tsk->sched_class->task_change_group(tsk, type);
6382
	else
P
Peter Zijlstra 已提交
6383
#endif
6384
		set_task_rq(tsk, task_cpu(tsk));
6385 6386 6387 6388 6389 6390 6391 6392 6393 6394 6395 6396 6397 6398 6399 6400
}

/*
 * Change task's runqueue when it moves between groups.
 *
 * The caller of this function should have put the task in its new group by
 * now. This function just updates tsk->se.cfs_rq and tsk->se.parent to reflect
 * its new group.
 */
void sched_move_task(struct task_struct *tsk)
{
	int queued, running;
	struct rq_flags rf;
	struct rq *rq;

	rq = task_rq_lock(tsk, &rf);
6401
	update_rq_clock(rq);
6402 6403 6404 6405 6406 6407

	running = task_current(rq, tsk);
	queued = task_on_rq_queued(tsk);

	if (queued)
		dequeue_task(rq, tsk, DEQUEUE_SAVE | DEQUEUE_MOVE);
6408
	if (running)
6409 6410 6411
		put_prev_task(rq, tsk);

	sched_change_group(tsk, TASK_MOVE_GROUP);
P
Peter Zijlstra 已提交
6412

6413
	if (queued)
6414
		enqueue_task(rq, tsk, ENQUEUE_RESTORE | ENQUEUE_MOVE);
6415
	if (running)
6416
		set_curr_task(rq, tsk);
S
Srivatsa Vaddagiri 已提交
6417

6418
	task_rq_unlock(rq, tsk, &rf);
S
Srivatsa Vaddagiri 已提交
6419
}
D
Dhaval Giani 已提交
6420
#endif /* CONFIG_CGROUP_SCHED */
S
Srivatsa Vaddagiri 已提交
6421

6422 6423 6424 6425 6426
#ifdef CONFIG_RT_GROUP_SCHED
/*
 * Ensure that the real time constraints are schedulable.
 */
static DEFINE_MUTEX(rt_constraints_mutex);
P
Peter Zijlstra 已提交
6427

P
Peter Zijlstra 已提交
6428 6429
/* Must be called with tasklist_lock held */
static inline int tg_has_rt_tasks(struct task_group *tg)
6430
{
P
Peter Zijlstra 已提交
6431
	struct task_struct *g, *p;
6432

6433 6434 6435 6436 6437 6438
	/*
	 * Autogroups do not have RT tasks; see autogroup_create().
	 */
	if (task_group_is_autogroup(tg))
		return 0;

6439
	for_each_process_thread(g, p) {
6440
		if (rt_task(p) && task_group(p) == tg)
P
Peter Zijlstra 已提交
6441
			return 1;
6442
	}
6443

P
Peter Zijlstra 已提交
6444 6445
	return 0;
}
6446

P
Peter Zijlstra 已提交
6447 6448 6449 6450 6451
struct rt_schedulable_data {
	struct task_group *tg;
	u64 rt_period;
	u64 rt_runtime;
};
6452

6453
static int tg_rt_schedulable(struct task_group *tg, void *data)
P
Peter Zijlstra 已提交
6454 6455 6456 6457 6458
{
	struct rt_schedulable_data *d = data;
	struct task_group *child;
	unsigned long total, sum = 0;
	u64 period, runtime;
6459

P
Peter Zijlstra 已提交
6460 6461
	period = ktime_to_ns(tg->rt_bandwidth.rt_period);
	runtime = tg->rt_bandwidth.rt_runtime;
6462

P
Peter Zijlstra 已提交
6463 6464 6465
	if (tg == d->tg) {
		period = d->rt_period;
		runtime = d->rt_runtime;
6466 6467
	}

6468 6469 6470 6471 6472
	/*
	 * Cannot have more runtime than the period.
	 */
	if (runtime > period && runtime != RUNTIME_INF)
		return -EINVAL;
P
Peter Zijlstra 已提交
6473

6474 6475 6476
	/*
	 * Ensure we don't starve existing RT tasks.
	 */
P
Peter Zijlstra 已提交
6477 6478
	if (rt_bandwidth_enabled() && !runtime && tg_has_rt_tasks(tg))
		return -EBUSY;
P
Peter Zijlstra 已提交
6479

P
Peter Zijlstra 已提交
6480
	total = to_ratio(period, runtime);
P
Peter Zijlstra 已提交
6481

6482 6483 6484 6485 6486
	/*
	 * Nobody can have more than the global setting allows.
	 */
	if (total > to_ratio(global_rt_period(), global_rt_runtime()))
		return -EINVAL;
P
Peter Zijlstra 已提交
6487

6488 6489 6490
	/*
	 * The sum of our children's runtime should not exceed our own.
	 */
P
Peter Zijlstra 已提交
6491 6492 6493
	list_for_each_entry_rcu(child, &tg->children, siblings) {
		period = ktime_to_ns(child->rt_bandwidth.rt_period);
		runtime = child->rt_bandwidth.rt_runtime;
P
Peter Zijlstra 已提交
6494

P
Peter Zijlstra 已提交
6495 6496 6497 6498
		if (child == d->tg) {
			period = d->rt_period;
			runtime = d->rt_runtime;
		}
P
Peter Zijlstra 已提交
6499

P
Peter Zijlstra 已提交
6500
		sum += to_ratio(period, runtime);
P
Peter Zijlstra 已提交
6501
	}
P
Peter Zijlstra 已提交
6502

P
Peter Zijlstra 已提交
6503 6504 6505 6506
	if (sum > total)
		return -EINVAL;

	return 0;
P
Peter Zijlstra 已提交
6507 6508
}

P
Peter Zijlstra 已提交
6509
static int __rt_schedulable(struct task_group *tg, u64 period, u64 runtime)
6510
{
6511 6512
	int ret;

P
Peter Zijlstra 已提交
6513 6514 6515 6516 6517 6518
	struct rt_schedulable_data data = {
		.tg = tg,
		.rt_period = period,
		.rt_runtime = runtime,
	};

6519 6520 6521 6522 6523
	rcu_read_lock();
	ret = walk_tg_tree(tg_rt_schedulable, tg_nop, &data);
	rcu_read_unlock();

	return ret;
6524 6525
}

6526
static int tg_set_rt_bandwidth(struct task_group *tg,
6527
		u64 rt_period, u64 rt_runtime)
P
Peter Zijlstra 已提交
6528
{
P
Peter Zijlstra 已提交
6529
	int i, err = 0;
P
Peter Zijlstra 已提交
6530

6531 6532 6533 6534 6535 6536 6537 6538 6539 6540 6541
	/*
	 * Disallowing the root group RT runtime is BAD, it would disallow the
	 * kernel creating (and or operating) RT threads.
	 */
	if (tg == &root_task_group && rt_runtime == 0)
		return -EINVAL;

	/* No period doesn't make any sense. */
	if (rt_period == 0)
		return -EINVAL;

P
Peter Zijlstra 已提交
6542
	mutex_lock(&rt_constraints_mutex);
6543
	read_lock(&tasklist_lock);
P
Peter Zijlstra 已提交
6544 6545
	err = __rt_schedulable(tg, rt_period, rt_runtime);
	if (err)
P
Peter Zijlstra 已提交
6546
		goto unlock;
P
Peter Zijlstra 已提交
6547

6548
	raw_spin_lock_irq(&tg->rt_bandwidth.rt_runtime_lock);
6549 6550
	tg->rt_bandwidth.rt_period = ns_to_ktime(rt_period);
	tg->rt_bandwidth.rt_runtime = rt_runtime;
P
Peter Zijlstra 已提交
6551 6552 6553 6554

	for_each_possible_cpu(i) {
		struct rt_rq *rt_rq = tg->rt_rq[i];

6555
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
6556
		rt_rq->rt_runtime = rt_runtime;
6557
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
6558
	}
6559
	raw_spin_unlock_irq(&tg->rt_bandwidth.rt_runtime_lock);
P
Peter Zijlstra 已提交
6560
unlock:
6561
	read_unlock(&tasklist_lock);
P
Peter Zijlstra 已提交
6562 6563 6564
	mutex_unlock(&rt_constraints_mutex);

	return err;
P
Peter Zijlstra 已提交
6565 6566
}

6567
static int sched_group_set_rt_runtime(struct task_group *tg, long rt_runtime_us)
6568 6569 6570 6571 6572 6573 6574 6575
{
	u64 rt_runtime, rt_period;

	rt_period = ktime_to_ns(tg->rt_bandwidth.rt_period);
	rt_runtime = (u64)rt_runtime_us * NSEC_PER_USEC;
	if (rt_runtime_us < 0)
		rt_runtime = RUNTIME_INF;

6576
	return tg_set_rt_bandwidth(tg, rt_period, rt_runtime);
6577 6578
}

6579
static long sched_group_rt_runtime(struct task_group *tg)
P
Peter Zijlstra 已提交
6580 6581 6582
{
	u64 rt_runtime_us;

6583
	if (tg->rt_bandwidth.rt_runtime == RUNTIME_INF)
P
Peter Zijlstra 已提交
6584 6585
		return -1;

6586
	rt_runtime_us = tg->rt_bandwidth.rt_runtime;
P
Peter Zijlstra 已提交
6587 6588 6589
	do_div(rt_runtime_us, NSEC_PER_USEC);
	return rt_runtime_us;
}
6590

6591
static int sched_group_set_rt_period(struct task_group *tg, u64 rt_period_us)
6592 6593 6594
{
	u64 rt_runtime, rt_period;

6595
	rt_period = rt_period_us * NSEC_PER_USEC;
6596 6597
	rt_runtime = tg->rt_bandwidth.rt_runtime;

6598
	return tg_set_rt_bandwidth(tg, rt_period, rt_runtime);
6599 6600
}

6601
static long sched_group_rt_period(struct task_group *tg)
6602 6603 6604 6605 6606 6607 6608
{
	u64 rt_period_us;

	rt_period_us = ktime_to_ns(tg->rt_bandwidth.rt_period);
	do_div(rt_period_us, NSEC_PER_USEC);
	return rt_period_us;
}
6609
#endif /* CONFIG_RT_GROUP_SCHED */
6610

6611
#ifdef CONFIG_RT_GROUP_SCHED
6612 6613 6614 6615 6616
static int sched_rt_global_constraints(void)
{
	int ret = 0;

	mutex_lock(&rt_constraints_mutex);
P
Peter Zijlstra 已提交
6617
	read_lock(&tasklist_lock);
6618
	ret = __rt_schedulable(NULL, 0, 0);
P
Peter Zijlstra 已提交
6619
	read_unlock(&tasklist_lock);
6620 6621 6622 6623
	mutex_unlock(&rt_constraints_mutex);

	return ret;
}
6624

6625
static int sched_rt_can_attach(struct task_group *tg, struct task_struct *tsk)
6626 6627 6628 6629 6630 6631 6632 6633
{
	/* Don't accept realtime tasks when there is no way for them to run */
	if (rt_task(tsk) && tg->rt_bandwidth.rt_runtime == 0)
		return 0;

	return 1;
}

6634
#else /* !CONFIG_RT_GROUP_SCHED */
6635 6636
static int sched_rt_global_constraints(void)
{
P
Peter Zijlstra 已提交
6637
	unsigned long flags;
6638
	int i;
6639

6640
	raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags);
P
Peter Zijlstra 已提交
6641 6642 6643
	for_each_possible_cpu(i) {
		struct rt_rq *rt_rq = &cpu_rq(i)->rt;

6644
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
6645
		rt_rq->rt_runtime = global_rt_runtime();
6646
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
6647
	}
6648
	raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags);
P
Peter Zijlstra 已提交
6649

6650
	return 0;
6651
}
6652
#endif /* CONFIG_RT_GROUP_SCHED */
6653

6654
static int sched_dl_global_validate(void)
6655
{
6656 6657
	u64 runtime = global_rt_runtime();
	u64 period = global_rt_period();
6658
	u64 new_bw = to_ratio(period, runtime);
6659
	struct dl_bw *dl_b;
6660
	int cpu, ret = 0;
6661
	unsigned long flags;
6662 6663 6664 6665 6666 6667 6668 6669 6670 6671

	/*
	 * Here we want to check the bandwidth not being set to some
	 * value smaller than the currently allocated bandwidth in
	 * any of the root_domains.
	 *
	 * FIXME: Cycling on all the CPUs is overdoing, but simpler than
	 * cycling on root_domains... Discussion on different/better
	 * solutions is welcome!
	 */
6672
	for_each_possible_cpu(cpu) {
6673 6674
		rcu_read_lock_sched();
		dl_b = dl_bw_of(cpu);
6675

6676
		raw_spin_lock_irqsave(&dl_b->lock, flags);
6677 6678
		if (new_bw < dl_b->total_bw)
			ret = -EBUSY;
6679
		raw_spin_unlock_irqrestore(&dl_b->lock, flags);
6680

6681 6682
		rcu_read_unlock_sched();

6683 6684
		if (ret)
			break;
6685 6686
	}

6687
	return ret;
6688 6689
}

6690
static void sched_dl_do_global(void)
6691
{
6692
	u64 new_bw = -1;
6693
	struct dl_bw *dl_b;
6694
	int cpu;
6695
	unsigned long flags;
6696

6697 6698 6699 6700 6701 6702 6703 6704 6705 6706
	def_dl_bandwidth.dl_period = global_rt_period();
	def_dl_bandwidth.dl_runtime = global_rt_runtime();

	if (global_rt_runtime() != RUNTIME_INF)
		new_bw = to_ratio(global_rt_period(), global_rt_runtime());

	/*
	 * FIXME: As above...
	 */
	for_each_possible_cpu(cpu) {
6707 6708
		rcu_read_lock_sched();
		dl_b = dl_bw_of(cpu);
6709

6710
		raw_spin_lock_irqsave(&dl_b->lock, flags);
6711
		dl_b->bw = new_bw;
6712
		raw_spin_unlock_irqrestore(&dl_b->lock, flags);
6713 6714

		rcu_read_unlock_sched();
6715
	}
6716 6717 6718 6719 6720 6721 6722
}

static int sched_rt_global_validate(void)
{
	if (sysctl_sched_rt_period <= 0)
		return -EINVAL;

6723 6724
	if ((sysctl_sched_rt_runtime != RUNTIME_INF) &&
		(sysctl_sched_rt_runtime > sysctl_sched_rt_period))
6725 6726 6727 6728 6729 6730 6731 6732 6733
		return -EINVAL;

	return 0;
}

static void sched_rt_do_global(void)
{
	def_rt_bandwidth.rt_runtime = global_rt_runtime();
	def_rt_bandwidth.rt_period = ns_to_ktime(global_rt_period());
6734 6735
}

6736
int sched_rt_handler(struct ctl_table *table, int write,
6737
		void __user *buffer, size_t *lenp,
6738 6739 6740 6741
		loff_t *ppos)
{
	int old_period, old_runtime;
	static DEFINE_MUTEX(mutex);
6742
	int ret;
6743 6744 6745 6746 6747

	mutex_lock(&mutex);
	old_period = sysctl_sched_rt_period;
	old_runtime = sysctl_sched_rt_runtime;

6748
	ret = proc_dointvec(table, write, buffer, lenp, ppos);
6749 6750

	if (!ret && write) {
6751 6752 6753 6754
		ret = sched_rt_global_validate();
		if (ret)
			goto undo;

6755
		ret = sched_dl_global_validate();
6756 6757 6758
		if (ret)
			goto undo;

6759
		ret = sched_rt_global_constraints();
6760 6761 6762 6763 6764 6765 6766 6767 6768 6769
		if (ret)
			goto undo;

		sched_rt_do_global();
		sched_dl_do_global();
	}
	if (0) {
undo:
		sysctl_sched_rt_period = old_period;
		sysctl_sched_rt_runtime = old_runtime;
6770 6771 6772 6773 6774
	}
	mutex_unlock(&mutex);

	return ret;
}
6775

6776
int sched_rr_handler(struct ctl_table *table, int write,
6777 6778 6779 6780 6781 6782 6783 6784
		void __user *buffer, size_t *lenp,
		loff_t *ppos)
{
	int ret;
	static DEFINE_MUTEX(mutex);

	mutex_lock(&mutex);
	ret = proc_dointvec(table, write, buffer, lenp, ppos);
I
Ingo Molnar 已提交
6785 6786 6787 6788
	/*
	 * Make sure that internally we keep jiffies.
	 * Also, writing zero resets the timeslice to default:
	 */
6789
	if (!ret && write) {
6790 6791 6792
		sched_rr_timeslice =
			sysctl_sched_rr_timeslice <= 0 ? RR_TIMESLICE :
			msecs_to_jiffies(sysctl_sched_rr_timeslice);
6793 6794 6795 6796 6797
	}
	mutex_unlock(&mutex);
	return ret;
}

6798
#ifdef CONFIG_CGROUP_SCHED
6799

6800
static inline struct task_group *css_tg(struct cgroup_subsys_state *css)
6801
{
6802
	return css ? container_of(css, struct task_group, css) : NULL;
6803 6804
}

6805 6806
static struct cgroup_subsys_state *
cpu_cgroup_css_alloc(struct cgroup_subsys_state *parent_css)
6807
{
6808 6809
	struct task_group *parent = css_tg(parent_css);
	struct task_group *tg;
6810

6811
	if (!parent) {
6812
		/* This is early initialization for the top cgroup */
6813
		return &root_task_group.css;
6814 6815
	}

6816
	tg = sched_create_group(parent);
6817 6818 6819 6820 6821 6822
	if (IS_ERR(tg))
		return ERR_PTR(-ENOMEM);

	return &tg->css;
}

6823 6824 6825 6826 6827 6828 6829 6830 6831 6832 6833
/* Expose task group only after completing cgroup initialization */
static int cpu_cgroup_css_online(struct cgroup_subsys_state *css)
{
	struct task_group *tg = css_tg(css);
	struct task_group *parent = css_tg(css->parent);

	if (parent)
		sched_online_group(tg, parent);
	return 0;
}

6834
static void cpu_cgroup_css_released(struct cgroup_subsys_state *css)
6835
{
6836
	struct task_group *tg = css_tg(css);
6837

6838
	sched_offline_group(tg);
6839 6840
}

6841
static void cpu_cgroup_css_free(struct cgroup_subsys_state *css)
6842
{
6843
	struct task_group *tg = css_tg(css);
6844

6845 6846 6847 6848
	/*
	 * Relies on the RCU grace period between css_released() and this.
	 */
	sched_free_group(tg);
6849 6850
}

6851 6852 6853 6854
/*
 * This is called before wake_up_new_task(), therefore we really only
 * have to set its group bits, all the other stuff does not apply.
 */
6855
static void cpu_cgroup_fork(struct task_struct *task)
6856
{
6857 6858 6859 6860 6861
	struct rq_flags rf;
	struct rq *rq;

	rq = task_rq_lock(task, &rf);

6862
	update_rq_clock(rq);
6863 6864 6865
	sched_change_group(task, TASK_SET_GROUP);

	task_rq_unlock(rq, task, &rf);
6866 6867
}

6868
static int cpu_cgroup_can_attach(struct cgroup_taskset *tset)
6869
{
6870
	struct task_struct *task;
6871
	struct cgroup_subsys_state *css;
6872
	int ret = 0;
6873

6874
	cgroup_taskset_for_each(task, css, tset) {
6875
#ifdef CONFIG_RT_GROUP_SCHED
6876
		if (!sched_rt_can_attach(css_tg(css), task))
6877
			return -EINVAL;
6878
#else
6879 6880 6881
		/* We don't support RT-tasks being in separate groups */
		if (task->sched_class != &fair_sched_class)
			return -EINVAL;
6882
#endif
6883 6884 6885 6886 6887 6888 6889 6890 6891 6892 6893 6894 6895 6896 6897 6898
		/*
		 * Serialize against wake_up_new_task() such that if its
		 * running, we're sure to observe its full state.
		 */
		raw_spin_lock_irq(&task->pi_lock);
		/*
		 * Avoid calling sched_move_task() before wake_up_new_task()
		 * has happened. This would lead to problems with PELT, due to
		 * move wanting to detach+attach while we're not attached yet.
		 */
		if (task->state == TASK_NEW)
			ret = -EINVAL;
		raw_spin_unlock_irq(&task->pi_lock);

		if (ret)
			break;
6899
	}
6900
	return ret;
6901
}
6902

6903
static void cpu_cgroup_attach(struct cgroup_taskset *tset)
6904
{
6905
	struct task_struct *task;
6906
	struct cgroup_subsys_state *css;
6907

6908
	cgroup_taskset_for_each(task, css, tset)
6909
		sched_move_task(task);
6910 6911
}

6912
#ifdef CONFIG_FAIR_GROUP_SCHED
6913 6914
static int cpu_shares_write_u64(struct cgroup_subsys_state *css,
				struct cftype *cftype, u64 shareval)
6915
{
6916
	return sched_group_set_shares(css_tg(css), scale_load(shareval));
6917 6918
}

6919 6920
static u64 cpu_shares_read_u64(struct cgroup_subsys_state *css,
			       struct cftype *cft)
6921
{
6922
	struct task_group *tg = css_tg(css);
6923

6924
	return (u64) scale_load_down(tg->shares);
6925
}
6926 6927

#ifdef CONFIG_CFS_BANDWIDTH
6928 6929
static DEFINE_MUTEX(cfs_constraints_mutex);

6930 6931 6932
const u64 max_cfs_quota_period = 1 * NSEC_PER_SEC; /* 1s */
const u64 min_cfs_quota_period = 1 * NSEC_PER_MSEC; /* 1ms */

6933 6934
static int __cfs_schedulable(struct task_group *tg, u64 period, u64 runtime);

6935 6936
static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota)
{
6937
	int i, ret = 0, runtime_enabled, runtime_was_enabled;
6938
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
6939 6940 6941 6942 6943 6944 6945 6946 6947 6948 6949 6950 6951 6952 6953 6954 6955 6956 6957 6958

	if (tg == &root_task_group)
		return -EINVAL;

	/*
	 * Ensure we have at some amount of bandwidth every period.  This is
	 * to prevent reaching a state of large arrears when throttled via
	 * entity_tick() resulting in prolonged exit starvation.
	 */
	if (quota < min_cfs_quota_period || period < min_cfs_quota_period)
		return -EINVAL;

	/*
	 * Likewise, bound things on the otherside by preventing insane quota
	 * periods.  This also allows us to normalize in computing quota
	 * feasibility.
	 */
	if (period > max_cfs_quota_period)
		return -EINVAL;

6959 6960 6961 6962 6963
	/*
	 * Prevent race between setting of cfs_rq->runtime_enabled and
	 * unthrottle_offline_cfs_rqs().
	 */
	get_online_cpus();
6964 6965 6966 6967 6968
	mutex_lock(&cfs_constraints_mutex);
	ret = __cfs_schedulable(tg, period, quota);
	if (ret)
		goto out_unlock;

6969
	runtime_enabled = quota != RUNTIME_INF;
6970
	runtime_was_enabled = cfs_b->quota != RUNTIME_INF;
6971 6972 6973 6974 6975 6976
	/*
	 * If we need to toggle cfs_bandwidth_used, off->on must occur
	 * before making related changes, and on->off must occur afterwards
	 */
	if (runtime_enabled && !runtime_was_enabled)
		cfs_bandwidth_usage_inc();
6977 6978 6979
	raw_spin_lock_irq(&cfs_b->lock);
	cfs_b->period = ns_to_ktime(period);
	cfs_b->quota = quota;
6980

P
Paul Turner 已提交
6981
	__refill_cfs_bandwidth_runtime(cfs_b);
I
Ingo Molnar 已提交
6982 6983

	/* Restart the period timer (if active) to handle new period expiry: */
P
Peter Zijlstra 已提交
6984 6985
	if (runtime_enabled)
		start_cfs_bandwidth(cfs_b);
I
Ingo Molnar 已提交
6986

6987 6988
	raw_spin_unlock_irq(&cfs_b->lock);

6989
	for_each_online_cpu(i) {
6990
		struct cfs_rq *cfs_rq = tg->cfs_rq[i];
6991
		struct rq *rq = cfs_rq->rq;
6992 6993

		raw_spin_lock_irq(&rq->lock);
6994
		cfs_rq->runtime_enabled = runtime_enabled;
6995
		cfs_rq->runtime_remaining = 0;
6996

6997
		if (cfs_rq->throttled)
6998
			unthrottle_cfs_rq(cfs_rq);
6999 7000
		raw_spin_unlock_irq(&rq->lock);
	}
7001 7002
	if (runtime_was_enabled && !runtime_enabled)
		cfs_bandwidth_usage_dec();
7003 7004
out_unlock:
	mutex_unlock(&cfs_constraints_mutex);
7005
	put_online_cpus();
7006

7007
	return ret;
7008 7009 7010 7011 7012 7013
}

int tg_set_cfs_quota(struct task_group *tg, long cfs_quota_us)
{
	u64 quota, period;

7014
	period = ktime_to_ns(tg->cfs_bandwidth.period);
7015 7016 7017 7018 7019 7020 7021 7022 7023 7024 7025 7026
	if (cfs_quota_us < 0)
		quota = RUNTIME_INF;
	else
		quota = (u64)cfs_quota_us * NSEC_PER_USEC;

	return tg_set_cfs_bandwidth(tg, period, quota);
}

long tg_get_cfs_quota(struct task_group *tg)
{
	u64 quota_us;

7027
	if (tg->cfs_bandwidth.quota == RUNTIME_INF)
7028 7029
		return -1;

7030
	quota_us = tg->cfs_bandwidth.quota;
7031 7032 7033 7034 7035 7036 7037 7038 7039 7040
	do_div(quota_us, NSEC_PER_USEC);

	return quota_us;
}

int tg_set_cfs_period(struct task_group *tg, long cfs_period_us)
{
	u64 quota, period;

	period = (u64)cfs_period_us * NSEC_PER_USEC;
7041
	quota = tg->cfs_bandwidth.quota;
7042 7043 7044 7045 7046 7047 7048 7049

	return tg_set_cfs_bandwidth(tg, period, quota);
}

long tg_get_cfs_period(struct task_group *tg)
{
	u64 cfs_period_us;

7050
	cfs_period_us = ktime_to_ns(tg->cfs_bandwidth.period);
7051 7052 7053 7054 7055
	do_div(cfs_period_us, NSEC_PER_USEC);

	return cfs_period_us;
}

7056 7057
static s64 cpu_cfs_quota_read_s64(struct cgroup_subsys_state *css,
				  struct cftype *cft)
7058
{
7059
	return tg_get_cfs_quota(css_tg(css));
7060 7061
}

7062 7063
static int cpu_cfs_quota_write_s64(struct cgroup_subsys_state *css,
				   struct cftype *cftype, s64 cfs_quota_us)
7064
{
7065
	return tg_set_cfs_quota(css_tg(css), cfs_quota_us);
7066 7067
}

7068 7069
static u64 cpu_cfs_period_read_u64(struct cgroup_subsys_state *css,
				   struct cftype *cft)
7070
{
7071
	return tg_get_cfs_period(css_tg(css));
7072 7073
}

7074 7075
static int cpu_cfs_period_write_u64(struct cgroup_subsys_state *css,
				    struct cftype *cftype, u64 cfs_period_us)
7076
{
7077
	return tg_set_cfs_period(css_tg(css), cfs_period_us);
7078 7079
}

7080 7081 7082 7083 7084 7085 7086 7087 7088 7089 7090 7091 7092 7093 7094 7095 7096 7097 7098 7099 7100 7101 7102 7103 7104 7105 7106 7107 7108 7109 7110 7111
struct cfs_schedulable_data {
	struct task_group *tg;
	u64 period, quota;
};

/*
 * normalize group quota/period to be quota/max_period
 * note: units are usecs
 */
static u64 normalize_cfs_quota(struct task_group *tg,
			       struct cfs_schedulable_data *d)
{
	u64 quota, period;

	if (tg == d->tg) {
		period = d->period;
		quota = d->quota;
	} else {
		period = tg_get_cfs_period(tg);
		quota = tg_get_cfs_quota(tg);
	}

	/* note: these should typically be equivalent */
	if (quota == RUNTIME_INF || quota == -1)
		return RUNTIME_INF;

	return to_ratio(period, quota);
}

static int tg_cfs_schedulable_down(struct task_group *tg, void *data)
{
	struct cfs_schedulable_data *d = data;
7112
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
7113 7114 7115 7116 7117
	s64 quota = 0, parent_quota = -1;

	if (!tg->parent) {
		quota = RUNTIME_INF;
	} else {
7118
		struct cfs_bandwidth *parent_b = &tg->parent->cfs_bandwidth;
7119 7120

		quota = normalize_cfs_quota(tg, d);
7121
		parent_quota = parent_b->hierarchical_quota;
7122 7123

		/*
I
Ingo Molnar 已提交
7124 7125
		 * Ensure max(child_quota) <= parent_quota, inherit when no
		 * limit is set:
7126 7127 7128 7129 7130 7131
		 */
		if (quota == RUNTIME_INF)
			quota = parent_quota;
		else if (parent_quota != RUNTIME_INF && quota > parent_quota)
			return -EINVAL;
	}
7132
	cfs_b->hierarchical_quota = quota;
7133 7134 7135 7136 7137 7138

	return 0;
}

static int __cfs_schedulable(struct task_group *tg, u64 period, u64 quota)
{
7139
	int ret;
7140 7141 7142 7143 7144 7145 7146 7147 7148 7149 7150
	struct cfs_schedulable_data data = {
		.tg = tg,
		.period = period,
		.quota = quota,
	};

	if (quota != RUNTIME_INF) {
		do_div(data.period, NSEC_PER_USEC);
		do_div(data.quota, NSEC_PER_USEC);
	}

7151 7152 7153 7154 7155
	rcu_read_lock();
	ret = walk_tg_tree(tg_cfs_schedulable_down, tg_nop, &data);
	rcu_read_unlock();

	return ret;
7156
}
7157

7158
static int cpu_stats_show(struct seq_file *sf, void *v)
7159
{
7160
	struct task_group *tg = css_tg(seq_css(sf));
7161
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
7162

7163 7164 7165
	seq_printf(sf, "nr_periods %d\n", cfs_b->nr_periods);
	seq_printf(sf, "nr_throttled %d\n", cfs_b->nr_throttled);
	seq_printf(sf, "throttled_time %llu\n", cfs_b->throttled_time);
7166 7167 7168

	return 0;
}
7169
#endif /* CONFIG_CFS_BANDWIDTH */
7170
#endif /* CONFIG_FAIR_GROUP_SCHED */
7171

7172
#ifdef CONFIG_RT_GROUP_SCHED
7173 7174
static int cpu_rt_runtime_write(struct cgroup_subsys_state *css,
				struct cftype *cft, s64 val)
P
Peter Zijlstra 已提交
7175
{
7176
	return sched_group_set_rt_runtime(css_tg(css), val);
P
Peter Zijlstra 已提交
7177 7178
}

7179 7180
static s64 cpu_rt_runtime_read(struct cgroup_subsys_state *css,
			       struct cftype *cft)
P
Peter Zijlstra 已提交
7181
{
7182
	return sched_group_rt_runtime(css_tg(css));
P
Peter Zijlstra 已提交
7183
}
7184

7185 7186
static int cpu_rt_period_write_uint(struct cgroup_subsys_state *css,
				    struct cftype *cftype, u64 rt_period_us)
7187
{
7188
	return sched_group_set_rt_period(css_tg(css), rt_period_us);
7189 7190
}

7191 7192
static u64 cpu_rt_period_read_uint(struct cgroup_subsys_state *css,
				   struct cftype *cft)
7193
{
7194
	return sched_group_rt_period(css_tg(css));
7195
}
7196
#endif /* CONFIG_RT_GROUP_SCHED */
P
Peter Zijlstra 已提交
7197

7198
static struct cftype cpu_files[] = {
7199
#ifdef CONFIG_FAIR_GROUP_SCHED
7200 7201
	{
		.name = "shares",
7202 7203
		.read_u64 = cpu_shares_read_u64,
		.write_u64 = cpu_shares_write_u64,
7204
	},
7205
#endif
7206 7207 7208 7209 7210 7211 7212 7213 7214 7215 7216
#ifdef CONFIG_CFS_BANDWIDTH
	{
		.name = "cfs_quota_us",
		.read_s64 = cpu_cfs_quota_read_s64,
		.write_s64 = cpu_cfs_quota_write_s64,
	},
	{
		.name = "cfs_period_us",
		.read_u64 = cpu_cfs_period_read_u64,
		.write_u64 = cpu_cfs_period_write_u64,
	},
7217 7218
	{
		.name = "stat",
7219
		.seq_show = cpu_stats_show,
7220
	},
7221
#endif
7222
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
7223
	{
P
Peter Zijlstra 已提交
7224
		.name = "rt_runtime_us",
7225 7226
		.read_s64 = cpu_rt_runtime_read,
		.write_s64 = cpu_rt_runtime_write,
P
Peter Zijlstra 已提交
7227
	},
7228 7229
	{
		.name = "rt_period_us",
7230 7231
		.read_u64 = cpu_rt_period_read_uint,
		.write_u64 = cpu_rt_period_write_uint,
7232
	},
7233
#endif
I
Ingo Molnar 已提交
7234
	{ }	/* Terminate */
7235 7236
};

7237
struct cgroup_subsys cpu_cgrp_subsys = {
7238
	.css_alloc	= cpu_cgroup_css_alloc,
7239
	.css_online	= cpu_cgroup_css_online,
7240
	.css_released	= cpu_cgroup_css_released,
7241
	.css_free	= cpu_cgroup_css_free,
7242
	.fork		= cpu_cgroup_fork,
7243 7244
	.can_attach	= cpu_cgroup_can_attach,
	.attach		= cpu_cgroup_attach,
7245
	.legacy_cftypes	= cpu_files,
7246
	.early_init	= true,
7247 7248
};

7249
#endif	/* CONFIG_CGROUP_SCHED */
7250

7251 7252 7253 7254 7255
void dump_cpu_task(int cpu)
{
	pr_info("Task dump for CPU %d:\n", cpu);
	sched_show_task(cpu_curr(cpu));
}
7256 7257 7258 7259 7260 7261 7262 7263 7264 7265 7266 7267 7268 7269 7270 7271 7272 7273 7274 7275 7276 7277 7278 7279 7280 7281 7282 7283 7284 7285 7286 7287 7288 7289 7290 7291 7292 7293 7294 7295 7296

/*
 * Nice levels are multiplicative, with a gentle 10% change for every
 * nice level changed. I.e. when a CPU-bound task goes from nice 0 to
 * nice 1, it will get ~10% less CPU time than another CPU-bound task
 * that remained on nice 0.
 *
 * The "10% effect" is relative and cumulative: from _any_ nice level,
 * if you go up 1 level, it's -10% CPU usage, if you go down 1 level
 * it's +10% CPU usage. (to achieve that we use a multiplier of 1.25.
 * If a task goes up by ~10% and another task goes down by ~10% then
 * the relative distance between them is ~25%.)
 */
const int sched_prio_to_weight[40] = {
 /* -20 */     88761,     71755,     56483,     46273,     36291,
 /* -15 */     29154,     23254,     18705,     14949,     11916,
 /* -10 */      9548,      7620,      6100,      4904,      3906,
 /*  -5 */      3121,      2501,      1991,      1586,      1277,
 /*   0 */      1024,       820,       655,       526,       423,
 /*   5 */       335,       272,       215,       172,       137,
 /*  10 */       110,        87,        70,        56,        45,
 /*  15 */        36,        29,        23,        18,        15,
};

/*
 * Inverse (2^32/x) values of the sched_prio_to_weight[] array, precalculated.
 *
 * In cases where the weight does not change often, we can use the
 * precalculated inverse to speed up arithmetics by turning divisions
 * into multiplications:
 */
const u32 sched_prio_to_wmult[40] = {
 /* -20 */     48388,     59856,     76040,     92818,    118348,
 /* -15 */    147320,    184698,    229616,    287308,    360437,
 /* -10 */    449829,    563644,    704093,    875809,   1099582,
 /*  -5 */   1376151,   1717300,   2157191,   2708050,   3363326,
 /*   0 */   4194304,   5237765,   6557202,   8165337,  10153587,
 /*   5 */  12820798,  15790321,  19976592,  24970740,  31350126,
 /*  10 */  39045157,  49367440,  61356676,  76695844,  95443717,
 /*  15 */ 119304647, 148102320, 186737708, 238609294, 286331153,
};