core.c 178.8 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>
9
#include <linux/sched/clock.h>
10
#include <uapi/linux/sched/types.h>
11
#include <linux/sched/loadavg.h>
12
#include <linux/sched/hotplug.h>
L
Linus Torvalds 已提交
13
#include <linux/cpuset.h>
14
#include <linux/delayacct.h>
15
#include <linux/init_task.h>
16
#include <linux/context_tracking.h>
17
#include <linux/rcupdate_wait.h>
18 19 20 21 22 23

#include <linux/blkdev.h>
#include <linux/kprobes.h>
#include <linux/mmu_context.h>
#include <linux/module.h>
#include <linux/nmi.h>
24
#include <linux/prefetch.h>
25 26 27
#include <linux/profile.h>
#include <linux/security.h>
#include <linux/syscalls.h>
L
Linus Torvalds 已提交
28

29
#include <asm/switch_to.h>
30
#include <asm/tlb.h>
31 32 33
#ifdef CONFIG_PARAVIRT
#include <asm/paravirt.h>
#endif
L
Linus Torvalds 已提交
34

35
#include "sched.h"
36
#include "../workqueue_internal.h"
37
#include "../smpboot.h"
38

39
#define CREATE_TRACE_POINTS
40
#include <trace/events/sched.h>
41

42
DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
43

I
Ingo Molnar 已提交
44 45 46
/*
 * Debugging: various feature bits
 */
P
Peter Zijlstra 已提交
47 48 49 50

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

I
Ingo Molnar 已提交
51
const_debug unsigned int sysctl_sched_features =
52
#include "features.h"
P
Peter Zijlstra 已提交
53 54 55 56
	0;

#undef SCHED_FEAT

57 58 59 60 61 62
/*
 * 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;

63 64 65 66 67 68 69 70
/*
 * 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
71
/*
I
Ingo Molnar 已提交
72
 * period over which we measure -rt task CPU usage in us.
P
Peter Zijlstra 已提交
73 74
 * default: 1s
 */
P
Peter Zijlstra 已提交
75
unsigned int sysctl_sched_rt_period = 1000000;
P
Peter Zijlstra 已提交
76

77
__read_mostly int scheduler_running;
78

P
Peter Zijlstra 已提交
79 80 81 82 83
/*
 * 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
84

I
Ingo Molnar 已提交
85
/* CPUs with isolated domains */
86 87
cpumask_var_t cpu_isolated_map;

L
Linus Torvalds 已提交
88
/*
89
 * this_rq_lock - lock this runqueue and disable interrupts.
L
Linus Torvalds 已提交
90
 */
A
Alexey Dobriyan 已提交
91
static struct rq *this_rq_lock(void)
L
Linus Torvalds 已提交
92 93
	__acquires(rq->lock)
{
94
	struct rq *rq;
L
Linus Torvalds 已提交
95 96 97

	local_irq_disable();
	rq = this_rq();
98
	raw_spin_lock(&rq->lock);
L
Linus Torvalds 已提交
99 100 101 102

	return rq;
}

103 104 105
/*
 * __task_rq_lock - lock the rq @p resides on.
 */
106
struct rq *__task_rq_lock(struct task_struct *p, struct rq_flags *rf)
107 108 109 110 111 112 113 114 115 116
	__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))) {
117
			rq_pin_lock(rq, rf);
118 119 120 121 122 123 124 125 126 127 128 129
			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.
 */
130
struct rq *task_rq_lock(struct task_struct *p, struct rq_flags *rf)
131 132 133 134 135 136
	__acquires(p->pi_lock)
	__acquires(rq->lock)
{
	struct rq *rq;

	for (;;) {
137
		raw_spin_lock_irqsave(&p->pi_lock, rf->flags);
138 139 140 141 142 143 144 145 146 147 148 149 150 151 152
		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
153
		 * If we observe the new CPU in task_rq_lock, the acquire will
154 155 156
		 * pair with the WMB to ensure we must then also see migrating.
		 */
		if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) {
157
			rq_pin_lock(rq, rf);
158 159 160
			return rq;
		}
		raw_spin_unlock(&rq->lock);
161
		raw_spin_unlock_irqrestore(&p->pi_lock, rf->flags);
162 163 164 165 166 167

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

168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245
/*
 * 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
246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266
#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());

267
	raw_spin_lock(&rq->lock);
268
	update_rq_clock(rq);
P
Peter Zijlstra 已提交
269
	rq->curr->sched_class->task_tick(rq, rq->curr, 1);
270
	raw_spin_unlock(&rq->lock);
P
Peter Zijlstra 已提交
271 272 273 274

	return HRTIMER_NORESTART;
}

275
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
276

277
static void __hrtick_restart(struct rq *rq)
P
Peter Zijlstra 已提交
278 279 280
{
	struct hrtimer *timer = &rq->hrtick_timer;

281
	hrtimer_start_expires(timer, HRTIMER_MODE_ABS_PINNED);
P
Peter Zijlstra 已提交
282 283
}

284 285 286 287
/*
 * called from hardirq (IPI) context
 */
static void __hrtick_start(void *arg)
288
{
289
	struct rq *rq = arg;
290

291
	raw_spin_lock(&rq->lock);
P
Peter Zijlstra 已提交
292
	__hrtick_restart(rq);
293
	rq->hrtick_csd_pending = 0;
294
	raw_spin_unlock(&rq->lock);
295 296
}

297 298 299 300 301
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
302
void hrtick_start(struct rq *rq, u64 delay)
303
{
304
	struct hrtimer *timer = &rq->hrtick_timer;
305 306 307 308 309 310 311 312 313
	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);
314

315
	hrtimer_set_expires(timer, time);
316 317

	if (rq == this_rq()) {
P
Peter Zijlstra 已提交
318
		__hrtick_restart(rq);
319
	} else if (!rq->hrtick_csd_pending) {
320
		smp_call_function_single_async(cpu_of(rq), &rq->hrtick_csd);
321 322
		rq->hrtick_csd_pending = 1;
	}
323 324
}

325 326 327 328 329 330
#else
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
331
void hrtick_start(struct rq *rq, u64 delay)
332
{
W
Wanpeng Li 已提交
333 334 335 336 337
	/*
	 * Don't schedule slices shorter than 10000ns, that just
	 * doesn't make sense. Rely on vruntime for fairness.
	 */
	delay = max_t(u64, delay, 10000LL);
338 339
	hrtimer_start(&rq->hrtick_timer, ns_to_ktime(delay),
		      HRTIMER_MODE_REL_PINNED);
340 341
}
#endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
342

343
static void init_rq_hrtick(struct rq *rq)
P
Peter Zijlstra 已提交
344
{
345 346
#ifdef CONFIG_SMP
	rq->hrtick_csd_pending = 0;
P
Peter Zijlstra 已提交
347

348 349 350 351
	rq->hrtick_csd.flags = 0;
	rq->hrtick_csd.func = __hrtick_start;
	rq->hrtick_csd.info = rq;
#endif
P
Peter Zijlstra 已提交
352

353 354
	hrtimer_init(&rq->hrtick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	rq->hrtick_timer.function = hrtick;
P
Peter Zijlstra 已提交
355
}
A
Andrew Morton 已提交
356
#else	/* CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
357 358 359 360 361 362 363
static inline void hrtick_clear(struct rq *rq)
{
}

static inline void init_rq_hrtick(struct rq *rq)
{
}
A
Andrew Morton 已提交
364
#endif	/* CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
365

366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383
/*
 * 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;								\
})

384
#if defined(CONFIG_SMP) && defined(TIF_POLLING_NRFLAG)
385 386 387 388 389 390 391 392 393 394
/*
 * 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);
}
395 396 397 398 399 400 401 402 403 404

/*
 * 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);
405
	typeof(ti->flags) old, val = READ_ONCE(ti->flags);
406 407 408 409 410 411 412 413 414 415 416 417 418 419

	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;
}

420 421 422 423 424 425
#else
static bool set_nr_and_not_polling(struct task_struct *p)
{
	set_tsk_need_resched(p);
	return true;
}
426 427 428 429 430 431 432

#ifdef CONFIG_SMP
static bool set_nr_if_polling(struct task_struct *p)
{
	return false;
}
#endif
433 434
#endif

435 436 437 438 439 440 441 442 443 444
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
445
	 * barrier implied by the wakeup in wake_up_q().
446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467
	 */
	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
468
		/* Task can safely be re-inserted now: */
469 470 471 472 473 474 475 476 477 478 479 480
		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
481
/*
482
 * resched_curr - mark rq's current task 'to be rescheduled now'.
I
Ingo Molnar 已提交
483 484 485 486 487
 *
 * 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.
 */
488
void resched_curr(struct rq *rq)
I
Ingo Molnar 已提交
489
{
490
	struct task_struct *curr = rq->curr;
I
Ingo Molnar 已提交
491 492
	int cpu;

493
	lockdep_assert_held(&rq->lock);
I
Ingo Molnar 已提交
494

495
	if (test_tsk_need_resched(curr))
I
Ingo Molnar 已提交
496 497
		return;

498
	cpu = cpu_of(rq);
499

500
	if (cpu == smp_processor_id()) {
501
		set_tsk_need_resched(curr);
502
		set_preempt_need_resched();
I
Ingo Molnar 已提交
503
		return;
504
	}
I
Ingo Molnar 已提交
505

506
	if (set_nr_and_not_polling(curr))
I
Ingo Molnar 已提交
507
		smp_send_reschedule(cpu);
508 509
	else
		trace_sched_wake_idle_without_ipi(cpu);
I
Ingo Molnar 已提交
510 511
}

512
void resched_cpu(int cpu)
I
Ingo Molnar 已提交
513 514 515 516
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

517
	if (!raw_spin_trylock_irqsave(&rq->lock, flags))
I
Ingo Molnar 已提交
518
		return;
519
	resched_curr(rq);
520
	raw_spin_unlock_irqrestore(&rq->lock, flags);
I
Ingo Molnar 已提交
521
}
522

523
#ifdef CONFIG_SMP
524
#ifdef CONFIG_NO_HZ_COMMON
525
/*
I
Ingo Molnar 已提交
526 527
 * In the semi idle case, use the nearest busy CPU for migrating timers
 * from an idle CPU.  This is good for power-savings.
528 529
 *
 * We don't do similar optimization for completely idle system, as
I
Ingo Molnar 已提交
530 531
 * 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).
532
 */
533
int get_nohz_timer_target(void)
534
{
535
	int i, cpu = smp_processor_id();
536 537
	struct sched_domain *sd;

538
	if (!idle_cpu(cpu) && is_housekeeping_cpu(cpu))
539 540
		return cpu;

541
	rcu_read_lock();
542
	for_each_domain(cpu, sd) {
543
		for_each_cpu(i, sched_domain_span(sd)) {
544 545 546 547
			if (cpu == i)
				continue;

			if (!idle_cpu(i) && is_housekeeping_cpu(i)) {
548 549 550 551
				cpu = i;
				goto unlock;
			}
		}
552
	}
553 554 555

	if (!is_housekeeping_cpu(cpu))
		cpu = housekeeping_any_cpu();
556 557
unlock:
	rcu_read_unlock();
558 559
	return cpu;
}
I
Ingo Molnar 已提交
560

561 562 563 564 565 566 567 568 569 570
/*
 * 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.
 */
571
static void wake_up_idle_cpu(int cpu)
572 573 574 575 576 577
{
	struct rq *rq = cpu_rq(cpu);

	if (cpu == smp_processor_id())
		return;

578
	if (set_nr_and_not_polling(rq->idle))
579
		smp_send_reschedule(cpu);
580 581
	else
		trace_sched_wake_idle_without_ipi(cpu);
582 583
}

584
static bool wake_up_full_nohz_cpu(int cpu)
585
{
586 587 588 589 590 591
	/*
	 * 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.
	 */
592 593
	if (cpu_is_offline(cpu))
		return true;  /* Don't try to wake offline CPUs. */
594
	if (tick_nohz_full_cpu(cpu)) {
595 596
		if (cpu != smp_processor_id() ||
		    tick_nohz_tick_stopped())
597
			tick_nohz_full_kick_cpu(cpu);
598 599 600 601 602 603
		return true;
	}

	return false;
}

604 605 606 607 608
/*
 * 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.
 */
609 610
void wake_up_nohz_cpu(int cpu)
{
611
	if (!wake_up_full_nohz_cpu(cpu))
612 613 614
		wake_up_idle_cpu(cpu);
}

615
static inline bool got_nohz_idle_kick(void)
616
{
617
	int cpu = smp_processor_id();
618 619 620 621 622 623 624 625 626 627 628 629 630

	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;
631 632
}

633
#else /* CONFIG_NO_HZ_COMMON */
634

635
static inline bool got_nohz_idle_kick(void)
P
Peter Zijlstra 已提交
636
{
637
	return false;
P
Peter Zijlstra 已提交
638 639
}

640
#endif /* CONFIG_NO_HZ_COMMON */
641

642
#ifdef CONFIG_NO_HZ_FULL
643
bool sched_can_stop_tick(struct rq *rq)
644
{
645 646 647 648 649 650
	int fifo_nr_running;

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

651
	/*
652 653
	 * If there are more than one RR tasks, we need the tick to effect the
	 * actual RR behaviour.
654
	 */
655 656 657 658 659
	if (rq->rt.rr_nr_running) {
		if (rq->rt.rr_nr_running == 1)
			return true;
		else
			return false;
660 661
	}

662 663 664 665 666 667 668 669 670 671 672 673 674 675
	/*
	 * 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)
676
		return false;
677

678
	return true;
679 680
}
#endif /* CONFIG_NO_HZ_FULL */
681

682
void sched_avg_update(struct rq *rq)
683
{
684 685
	s64 period = sched_avg_period();

686
	while ((s64)(rq_clock(rq) - rq->age_stamp) > period) {
687 688 689 690 691 692
		/*
		 * 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));
693 694 695
		rq->age_stamp += period;
		rq->rt_avg /= 2;
	}
696 697
}

698
#endif /* CONFIG_SMP */
699

700 701
#if defined(CONFIG_RT_GROUP_SCHED) || (defined(CONFIG_FAIR_GROUP_SCHED) && \
			(defined(CONFIG_SMP) || defined(CONFIG_CFS_BANDWIDTH)))
702
/*
703 704 705 706
 * 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.
707
 */
708
int walk_tg_tree_from(struct task_group *from,
709
			     tg_visitor down, tg_visitor up, void *data)
710 711
{
	struct task_group *parent, *child;
P
Peter Zijlstra 已提交
712
	int ret;
713

714 715
	parent = from;

716
down:
P
Peter Zijlstra 已提交
717 718
	ret = (*down)(parent, data);
	if (ret)
719
		goto out;
720 721 722 723 724 725 726
	list_for_each_entry_rcu(child, &parent->children, siblings) {
		parent = child;
		goto down;

up:
		continue;
	}
P
Peter Zijlstra 已提交
727
	ret = (*up)(parent, data);
728 729
	if (ret || parent == from)
		goto out;
730 731 732 733 734

	child = parent;
	parent = parent->parent;
	if (parent)
		goto up;
735
out:
P
Peter Zijlstra 已提交
736
	return ret;
737 738
}

739
int tg_nop(struct task_group *tg, void *data)
P
Peter Zijlstra 已提交
740
{
741
	return 0;
P
Peter Zijlstra 已提交
742
}
743 744
#endif

745 746
static void set_load_weight(struct task_struct *p)
{
N
Nikhil Rao 已提交
747 748 749
	int prio = p->static_prio - MAX_RT_PRIO;
	struct load_weight *load = &p->se.load;

I
Ingo Molnar 已提交
750 751 752
	/*
	 * SCHED_IDLE tasks get minimal weight:
	 */
753
	if (idle_policy(p->policy)) {
754
		load->weight = scale_load(WEIGHT_IDLEPRIO);
N
Nikhil Rao 已提交
755
		load->inv_weight = WMULT_IDLEPRIO;
I
Ingo Molnar 已提交
756 757
		return;
	}
758

759 760
	load->weight = scale_load(sched_prio_to_weight[prio]);
	load->inv_weight = sched_prio_to_wmult[prio];
761 762
}

763
static inline void enqueue_task(struct rq *rq, struct task_struct *p, int flags)
764
{
765
	update_rq_clock(rq);
766 767
	if (!(flags & ENQUEUE_RESTORE))
		sched_info_queued(rq, p);
768
	p->sched_class->enqueue_task(rq, p, flags);
769 770
}

771
static inline void dequeue_task(struct rq *rq, struct task_struct *p, int flags)
772
{
773
	update_rq_clock(rq);
774 775
	if (!(flags & DEQUEUE_SAVE))
		sched_info_dequeued(rq, p);
776
	p->sched_class->dequeue_task(rq, p, flags);
777 778
}

779
void activate_task(struct rq *rq, struct task_struct *p, int flags)
780 781 782 783
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible--;

784
	enqueue_task(rq, p, flags);
785 786
}

787
void deactivate_task(struct rq *rq, struct task_struct *p, int flags)
788 789 790 791
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible++;

792
	dequeue_task(rq, p, flags);
793 794
}

795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824
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;
	}
}

825
/*
I
Ingo Molnar 已提交
826
 * __normal_prio - return the priority that is based on the static prio
827 828 829
 */
static inline int __normal_prio(struct task_struct *p)
{
I
Ingo Molnar 已提交
830
	return p->static_prio;
831 832
}

833 834 835 836 837 838 839
/*
 * 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.
 */
840
static inline int normal_prio(struct task_struct *p)
841 842 843
{
	int prio;

844 845 846
	if (task_has_dl_policy(p))
		prio = MAX_DL_PRIO-1;
	else if (task_has_rt_policy(p))
847 848 849 850 851 852 853 854 855 856 857 858 859
		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.
 */
860
static int effective_prio(struct task_struct *p)
861 862 863 864 865 866 867 868 869 870 871 872
{
	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
873 874 875
/**
 * task_curr - is this task currently executing on a CPU?
 * @p: the task in question.
876 877
 *
 * Return: 1 if the task is currently executing. 0 otherwise.
L
Linus Torvalds 已提交
878
 */
879
inline int task_curr(const struct task_struct *p)
L
Linus Torvalds 已提交
880 881 882 883
{
	return cpu_curr(task_cpu(p)) == p;
}

884
/*
885 886 887 888 889
 * 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().
890
 */
891 892
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
893
				       int oldprio)
894 895 896
{
	if (prev_class != p->sched_class) {
		if (prev_class->switched_from)
P
Peter Zijlstra 已提交
897
			prev_class->switched_from(rq, p);
898

P
Peter Zijlstra 已提交
899
		p->sched_class->switched_to(rq, p);
900
	} else if (oldprio != p->prio || dl_task(p))
P
Peter Zijlstra 已提交
901
		p->sched_class->prio_changed(rq, p, oldprio);
902 903
}

904
void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags)
905 906 907 908 909 910 911 912 913 914
{
	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) {
915
				resched_curr(rq);
916 917 918 919 920 921 922 923 924
				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.
	 */
925
	if (task_on_rq_queued(rq->curr) && test_tsk_need_resched(rq->curr))
926
		rq_clock_skip_update(rq, true);
927 928
}

L
Linus Torvalds 已提交
929
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948
/*
 * 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.
 */
949
static struct rq *move_queued_task(struct rq *rq, struct task_struct *p, int new_cpu)
P
Peter Zijlstra 已提交
950 951 952 953
{
	lockdep_assert_held(&rq->lock);

	p->on_rq = TASK_ON_RQ_MIGRATING;
954
	dequeue_task(rq, p, 0);
P
Peter Zijlstra 已提交
955 956 957 958 959 960 961 962
	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);
963
	p->on_rq = TASK_ON_RQ_QUEUED;
P
Peter Zijlstra 已提交
964 965 966 967 968 969 970 971 972 973 974
	check_preempt_curr(rq, p, 0);

	return rq;
}

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

/*
I
Ingo Molnar 已提交
975
 * Move (not current) task off this CPU, onto the destination CPU. We're doing
P
Peter Zijlstra 已提交
976 977 978 979 980 981 982
 * 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.
 */
983
static struct rq *__migrate_task(struct rq *rq, struct task_struct *p, int dest_cpu)
P
Peter Zijlstra 已提交
984 985
{
	if (unlikely(!cpu_active(dest_cpu)))
986
		return rq;
P
Peter Zijlstra 已提交
987 988

	/* Affinity changed (again). */
989
	if (!cpumask_test_cpu(dest_cpu, &p->cpus_allowed))
990
		return rq;
P
Peter Zijlstra 已提交
991

992 993 994
	rq = move_queued_task(rq, p, dest_cpu);

	return rq;
P
Peter Zijlstra 已提交
995 996 997 998 999 1000 1001 1002 1003 1004
}

/*
 * 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;
1005 1006
	struct task_struct *p = arg->task;
	struct rq *rq = this_rq();
P
Peter Zijlstra 已提交
1007 1008

	/*
I
Ingo Molnar 已提交
1009 1010
	 * 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
1011 1012 1013 1014 1015 1016 1017 1018
	 */
	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();
1019 1020 1021 1022 1023 1024 1025 1026

	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.
	 */
1027 1028 1029 1030 1031 1032
	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;
	}
1033 1034 1035
	raw_spin_unlock(&rq->lock);
	raw_spin_unlock(&p->pi_lock);

P
Peter Zijlstra 已提交
1036 1037 1038 1039
	local_irq_enable();
	return 0;
}

1040 1041 1042 1043 1044
/*
 * 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
1045 1046 1047 1048 1049
{
	cpumask_copy(&p->cpus_allowed, new_mask);
	p->nr_cpus_allowed = cpumask_weight(new_mask);
}

1050 1051
void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
{
1052 1053 1054
	struct rq *rq = task_rq(p);
	bool queued, running;

1055
	lockdep_assert_held(&p->pi_lock);
1056 1057 1058 1059 1060 1061 1062 1063 1064 1065

	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);
1066
		dequeue_task(rq, p, DEQUEUE_SAVE);
1067 1068 1069 1070
	}
	if (running)
		put_prev_task(rq, p);

1071
	p->sched_class->set_cpus_allowed(p, new_mask);
1072 1073

	if (queued)
1074
		enqueue_task(rq, p, ENQUEUE_RESTORE);
1075
	if (running)
1076
		set_curr_task(rq, p);
1077 1078
}

P
Peter Zijlstra 已提交
1079 1080 1081 1082 1083 1084 1085 1086 1087
/*
 * 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.
 */
1088 1089
static int __set_cpus_allowed_ptr(struct task_struct *p,
				  const struct cpumask *new_mask, bool check)
P
Peter Zijlstra 已提交
1090
{
1091
	const struct cpumask *cpu_valid_mask = cpu_active_mask;
P
Peter Zijlstra 已提交
1092
	unsigned int dest_cpu;
1093 1094
	struct rq_flags rf;
	struct rq *rq;
P
Peter Zijlstra 已提交
1095 1096
	int ret = 0;

1097
	rq = task_rq_lock(p, &rf);
1098
	update_rq_clock(rq);
P
Peter Zijlstra 已提交
1099

1100 1101 1102 1103 1104 1105 1106
	if (p->flags & PF_KTHREAD) {
		/*
		 * Kernel threads are allowed on online && !active CPUs
		 */
		cpu_valid_mask = cpu_online_mask;
	}

1107 1108 1109 1110 1111 1112 1113 1114 1115
	/*
	 * 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
1116 1117 1118
	if (cpumask_equal(&p->cpus_allowed, new_mask))
		goto out;

1119
	if (!cpumask_intersects(new_mask, cpu_valid_mask)) {
P
Peter Zijlstra 已提交
1120 1121 1122 1123 1124 1125
		ret = -EINVAL;
		goto out;
	}

	do_set_cpus_allowed(p, new_mask);

1126 1127 1128
	if (p->flags & PF_KTHREAD) {
		/*
		 * For kernel threads that do indeed end up on online &&
I
Ingo Molnar 已提交
1129
		 * !active we want to ensure they are strict per-CPU threads.
1130 1131 1132 1133 1134 1135
		 */
		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
1136 1137 1138 1139
	/* 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;

1140
	dest_cpu = cpumask_any_and(cpu_valid_mask, new_mask);
P
Peter Zijlstra 已提交
1141 1142 1143
	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. */
1144
		task_rq_unlock(rq, p, &rf);
P
Peter Zijlstra 已提交
1145 1146 1147
		stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg);
		tlb_migrate_finish(p->mm);
		return 0;
1148 1149 1150 1151 1152
	} else if (task_on_rq_queued(p)) {
		/*
		 * OK, since we're going to drop the lock immediately
		 * afterwards anyway.
		 */
1153
		rq_unpin_lock(rq, &rf);
1154
		rq = move_queued_task(rq, p, dest_cpu);
1155
		rq_repin_lock(rq, &rf);
1156
	}
P
Peter Zijlstra 已提交
1157
out:
1158
	task_rq_unlock(rq, p, &rf);
P
Peter Zijlstra 已提交
1159 1160 1161

	return ret;
}
1162 1163 1164 1165 1166

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
1167 1168
EXPORT_SYMBOL_GPL(set_cpus_allowed_ptr);

I
Ingo Molnar 已提交
1169
void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
I
Ingo Molnar 已提交
1170
{
1171 1172 1173 1174 1175
#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
1176
	WARN_ON_ONCE(p->state != TASK_RUNNING && p->state != TASK_WAKING &&
O
Oleg Nesterov 已提交
1177
			!p->on_rq);
1178

1179 1180 1181 1182 1183 1184 1185 1186 1187
	/*
	 * 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)));

1188
#ifdef CONFIG_LOCKDEP
1189 1190 1191 1192 1193
	/*
	 * 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
1194
	 * see task_group().
1195 1196 1197 1198
	 *
	 * Furthermore, all task_rq users should acquire both locks, see
	 * task_rq_lock().
	 */
1199 1200 1201
	WARN_ON_ONCE(debug_locks && !(lockdep_is_held(&p->pi_lock) ||
				      lockdep_is_held(&task_rq(p)->lock)));
#endif
1202 1203
#endif

1204
	trace_sched_migrate_task(p, new_cpu);
1205

1206
	if (task_cpu(p) != new_cpu) {
1207
		if (p->sched_class->migrate_task_rq)
1208
			p->sched_class->migrate_task_rq(p);
1209
		p->se.nr_migrations++;
1210
		perf_event_task_migrate(p);
1211
	}
I
Ingo Molnar 已提交
1212 1213

	__set_task_cpu(p, new_cpu);
I
Ingo Molnar 已提交
1214 1215
}

1216 1217
static void __migrate_swap_task(struct task_struct *p, int cpu)
{
1218
	if (task_on_rq_queued(p)) {
1219 1220 1221 1222 1223
		struct rq *src_rq, *dst_rq;

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

1224
		p->on_rq = TASK_ON_RQ_MIGRATING;
1225 1226 1227
		deactivate_task(src_rq, p, 0);
		set_task_cpu(p, cpu);
		activate_task(dst_rq, p, 0);
1228
		p->on_rq = TASK_ON_RQ_QUEUED;
1229 1230 1231 1232 1233
		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
1234
		 * previous CPU our target instead of where it really is.
1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250
		 */
		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;

1251 1252 1253
	if (!cpu_active(arg->src_cpu) || !cpu_active(arg->dst_cpu))
		return -EAGAIN;

1254 1255 1256
	src_rq = cpu_rq(arg->src_cpu);
	dst_rq = cpu_rq(arg->dst_cpu);

1257 1258
	double_raw_lock(&arg->src_task->pi_lock,
			&arg->dst_task->pi_lock);
1259
	double_rq_lock(src_rq, dst_rq);
1260

1261 1262 1263 1264 1265 1266
	if (task_cpu(arg->dst_task) != arg->dst_cpu)
		goto unlock;

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

1267
	if (!cpumask_test_cpu(arg->dst_cpu, &arg->src_task->cpus_allowed))
1268 1269
		goto unlock;

1270
	if (!cpumask_test_cpu(arg->src_cpu, &arg->dst_task->cpus_allowed))
1271 1272 1273 1274 1275 1276 1277 1278 1279
		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);
1280 1281
	raw_spin_unlock(&arg->dst_task->pi_lock);
	raw_spin_unlock(&arg->src_task->pi_lock);
1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303

	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;

1304 1305 1306 1307
	/*
	 * 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.
	 */
1308 1309 1310
	if (!cpu_active(arg.src_cpu) || !cpu_active(arg.dst_cpu))
		goto out;

1311
	if (!cpumask_test_cpu(arg.dst_cpu, &arg.src_task->cpus_allowed))
1312 1313
		goto out;

1314
	if (!cpumask_test_cpu(arg.src_cpu, &arg.dst_task->cpus_allowed))
1315 1316
		goto out;

1317
	trace_sched_swap_numa(cur, arg.src_cpu, p, arg.dst_cpu);
1318 1319 1320 1321 1322 1323
	ret = stop_two_cpus(arg.dst_cpu, arg.src_cpu, migrate_swap_stop, &arg);

out:
	return ret;
}

L
Linus Torvalds 已提交
1324 1325 1326
/*
 * wait_task_inactive - wait for a thread to unschedule.
 *
R
Roland McGrath 已提交
1327 1328 1329 1330 1331 1332 1333
 * 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
1334 1335 1336 1337 1338 1339
 * 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
1340
unsigned long wait_task_inactive(struct task_struct *p, long match_state)
L
Linus Torvalds 已提交
1341
{
1342
	int running, queued;
1343
	struct rq_flags rf;
R
Roland McGrath 已提交
1344
	unsigned long ncsw;
1345
	struct rq *rq;
L
Linus Torvalds 已提交
1346

1347 1348 1349 1350 1351 1352 1353 1354
	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);
1355

1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366
		/*
		 * 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
1367 1368 1369
		while (task_running(rq, p)) {
			if (match_state && unlikely(p->state != match_state))
				return 0;
1370
			cpu_relax();
R
Roland McGrath 已提交
1371
		}
1372

1373 1374 1375 1376 1377
		/*
		 * 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.
		 */
1378
		rq = task_rq_lock(p, &rf);
1379
		trace_sched_wait_task(p);
1380
		running = task_running(rq, p);
1381
		queued = task_on_rq_queued(p);
R
Roland McGrath 已提交
1382
		ncsw = 0;
1383
		if (!match_state || p->state == match_state)
1384
			ncsw = p->nvcsw | LONG_MIN; /* sets MSB */
1385
		task_rq_unlock(rq, p, &rf);
1386

R
Roland McGrath 已提交
1387 1388 1389 1390 1391 1392
		/*
		 * If it changed from the expected state, bail out now.
		 */
		if (unlikely(!ncsw))
			break;

1393 1394 1395 1396 1397 1398 1399 1400 1401 1402
		/*
		 * 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;
		}
1403

1404 1405 1406 1407 1408
		/*
		 * It's not enough that it's not actively running,
		 * it must be off the runqueue _entirely_, and not
		 * preempted!
		 *
1409
		 * So if it was still runnable (but just not actively
1410 1411 1412
		 * running right now), it's preempted, and we should
		 * yield - it could be a while.
		 */
1413
		if (unlikely(queued)) {
T
Thomas Gleixner 已提交
1414
			ktime_t to = NSEC_PER_SEC / HZ;
1415 1416 1417

			set_current_state(TASK_UNINTERRUPTIBLE);
			schedule_hrtimeout(&to, HRTIMER_MODE_REL);
1418 1419
			continue;
		}
1420

1421 1422 1423 1424 1425 1426 1427
		/*
		 * 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
1428 1429

	return ncsw;
L
Linus Torvalds 已提交
1430 1431 1432 1433 1434 1435 1436 1437 1438
}

/***
 * 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
1439
 * NOTE: this function doesn't have to take the runqueue lock,
L
Linus Torvalds 已提交
1440 1441 1442 1443 1444
 * 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.
 */
1445
void kick_process(struct task_struct *p)
L
Linus Torvalds 已提交
1446 1447 1448 1449 1450 1451 1452 1453 1454
{
	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
1455
EXPORT_SYMBOL_GPL(kick_process);
L
Linus Torvalds 已提交
1456

1457
/*
1458
 * ->cpus_allowed is protected by both rq->lock and p->pi_lock
1459 1460 1461 1462 1463 1464 1465
 *
 * 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
1466
 *    CPU isn't yet part of the sched domains, and balancing will not
1467 1468
 *    see it.
 *
I
Ingo Molnar 已提交
1469
 *  - on CPU-down we clear cpu_active() to mask the sched domains and
1470
 *    avoid the load balancer to place new tasks on the to be removed
I
Ingo Molnar 已提交
1471
 *    CPU. Existing tasks will remain running there and will be taken
1472 1473 1474 1475 1476 1477
 *    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.
1478
 */
1479 1480
static int select_fallback_rq(int cpu, struct task_struct *p)
{
1481 1482
	int nid = cpu_to_node(cpu);
	const struct cpumask *nodemask = NULL;
1483 1484
	enum { cpuset, possible, fail } state = cpuset;
	int dest_cpu;
1485

1486
	/*
I
Ingo Molnar 已提交
1487 1488 1489
	 * 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.
1490 1491 1492 1493 1494 1495 1496 1497
	 */
	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;
1498
			if (cpumask_test_cpu(dest_cpu, &p->cpus_allowed))
1499 1500
				return dest_cpu;
		}
1501
	}
1502

1503 1504
	for (;;) {
		/* Any allowed, online CPU? */
1505
		for_each_cpu(dest_cpu, &p->cpus_allowed) {
1506 1507 1508
			if (!(p->flags & PF_KTHREAD) && !cpu_active(dest_cpu))
				continue;
			if (!cpu_online(dest_cpu))
1509 1510 1511
				continue;
			goto out;
		}
1512

1513
		/* No more Mr. Nice Guy. */
1514 1515
		switch (state) {
		case cpuset:
1516 1517 1518 1519 1520
			if (IS_ENABLED(CONFIG_CPUSETS)) {
				cpuset_cpus_allowed_fallback(p);
				state = possible;
				break;
			}
I
Ingo Molnar 已提交
1521
			/* Fall-through */
1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540
		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()) {
1541
			printk_deferred("process %d (%s) no longer affine to cpu%d\n",
1542 1543
					task_pid_nr(p), p->comm, cpu);
		}
1544 1545 1546 1547 1548
	}

	return dest_cpu;
}

1549
/*
1550
 * The caller (fork, wakeup) owns p->pi_lock, ->cpus_allowed is stable.
1551
 */
1552
static inline
1553
int select_task_rq(struct task_struct *p, int cpu, int sd_flags, int wake_flags)
1554
{
1555 1556
	lockdep_assert_held(&p->pi_lock);

1557
	if (p->nr_cpus_allowed > 1)
1558
		cpu = p->sched_class->select_task_rq(p, cpu, sd_flags, wake_flags);
1559
	else
1560
		cpu = cpumask_any(&p->cpus_allowed);
1561 1562 1563 1564

	/*
	 * 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
1565
	 * CPU.
1566 1567 1568 1569 1570 1571
	 *
	 * 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 ]
	 */
1572
	if (unlikely(!cpumask_test_cpu(cpu, &p->cpus_allowed) ||
P
Peter Zijlstra 已提交
1573
		     !cpu_online(cpu)))
1574
		cpu = select_fallback_rq(task_cpu(p), p);
1575 1576

	return cpu;
1577
}
1578 1579 1580 1581 1582 1583

static void update_avg(u64 *avg, u64 sample)
{
	s64 diff = sample - *avg;
	*avg += diff >> 3;
}
1584 1585 1586 1587 1588 1589 1590 1591 1592

#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
1593
#endif /* CONFIG_SMP */
1594

P
Peter Zijlstra 已提交
1595
static void
1596
ttwu_stat(struct task_struct *p, int cpu, int wake_flags)
T
Tejun Heo 已提交
1597
{
1598
	struct rq *rq;
1599

1600 1601 1602 1603
	if (!schedstat_enabled())
		return;

	rq = this_rq();
P
Peter Zijlstra 已提交
1604

1605 1606
#ifdef CONFIG_SMP
	if (cpu == rq->cpu) {
1607 1608
		schedstat_inc(rq->ttwu_local);
		schedstat_inc(p->se.statistics.nr_wakeups_local);
P
Peter Zijlstra 已提交
1609 1610 1611
	} else {
		struct sched_domain *sd;

1612
		schedstat_inc(p->se.statistics.nr_wakeups_remote);
1613
		rcu_read_lock();
1614
		for_each_domain(rq->cpu, sd) {
P
Peter Zijlstra 已提交
1615
			if (cpumask_test_cpu(cpu, sched_domain_span(sd))) {
1616
				schedstat_inc(sd->ttwu_wake_remote);
P
Peter Zijlstra 已提交
1617 1618 1619
				break;
			}
		}
1620
		rcu_read_unlock();
P
Peter Zijlstra 已提交
1621
	}
1622 1623

	if (wake_flags & WF_MIGRATED)
1624
		schedstat_inc(p->se.statistics.nr_wakeups_migrate);
P
Peter Zijlstra 已提交
1625 1626
#endif /* CONFIG_SMP */

1627 1628
	schedstat_inc(rq->ttwu_count);
	schedstat_inc(p->se.statistics.nr_wakeups);
P
Peter Zijlstra 已提交
1629 1630

	if (wake_flags & WF_SYNC)
1631
		schedstat_inc(p->se.statistics.nr_wakeups_sync);
P
Peter Zijlstra 已提交
1632 1633
}

1634
static inline void ttwu_activate(struct rq *rq, struct task_struct *p, int en_flags)
P
Peter Zijlstra 已提交
1635
{
T
Tejun Heo 已提交
1636
	activate_task(rq, p, en_flags);
1637
	p->on_rq = TASK_ON_RQ_QUEUED;
1638

I
Ingo Molnar 已提交
1639
	/* If a worker is waking up, notify the workqueue: */
1640 1641
	if (p->flags & PF_WQ_WORKER)
		wq_worker_waking_up(p, cpu_of(rq));
T
Tejun Heo 已提交
1642 1643
}

1644 1645 1646
/*
 * Mark the task runnable and perform wakeup-preemption.
 */
1647
static void ttwu_do_wakeup(struct rq *rq, struct task_struct *p, int wake_flags,
1648
			   struct rq_flags *rf)
T
Tejun Heo 已提交
1649 1650 1651
{
	check_preempt_curr(rq, p, wake_flags);
	p->state = TASK_RUNNING;
1652 1653
	trace_sched_wakeup(p);

T
Tejun Heo 已提交
1654
#ifdef CONFIG_SMP
1655 1656
	if (p->sched_class->task_woken) {
		/*
1657 1658
		 * Our task @p is fully woken up and running; so its safe to
		 * drop the rq->lock, hereafter rq is only used for statistics.
1659
		 */
1660
		rq_unpin_lock(rq, rf);
T
Tejun Heo 已提交
1661
		p->sched_class->task_woken(rq, p);
1662
		rq_repin_lock(rq, rf);
1663
	}
T
Tejun Heo 已提交
1664

1665
	if (rq->idle_stamp) {
1666
		u64 delta = rq_clock(rq) - rq->idle_stamp;
1667
		u64 max = 2*rq->max_idle_balance_cost;
T
Tejun Heo 已提交
1668

1669 1670 1671
		update_avg(&rq->avg_idle, delta);

		if (rq->avg_idle > max)
T
Tejun Heo 已提交
1672
			rq->avg_idle = max;
1673

T
Tejun Heo 已提交
1674 1675 1676 1677 1678
		rq->idle_stamp = 0;
	}
#endif
}

1679
static void
1680
ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags,
1681
		 struct rq_flags *rf)
1682
{
1683 1684
	int en_flags = ENQUEUE_WAKEUP;

1685 1686
	lockdep_assert_held(&rq->lock);

1687 1688 1689
#ifdef CONFIG_SMP
	if (p->sched_contributes_to_load)
		rq->nr_uninterruptible--;
1690 1691

	if (wake_flags & WF_MIGRATED)
1692
		en_flags |= ENQUEUE_MIGRATED;
1693 1694
#endif

1695
	ttwu_activate(rq, p, en_flags);
1696
	ttwu_do_wakeup(rq, p, wake_flags, rf);
1697 1698 1699 1700 1701 1702 1703 1704 1705 1706
}

/*
 * 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)
{
1707
	struct rq_flags rf;
1708 1709 1710
	struct rq *rq;
	int ret = 0;

1711
	rq = __task_rq_lock(p, &rf);
1712
	if (task_on_rq_queued(p)) {
1713 1714
		/* check_preempt_curr() may use rq clock */
		update_rq_clock(rq);
1715
		ttwu_do_wakeup(rq, p, wake_flags, &rf);
1716 1717
		ret = 1;
	}
1718
	__task_rq_unlock(rq, &rf);
1719 1720 1721 1722

	return ret;
}

1723
#ifdef CONFIG_SMP
1724
void sched_ttwu_pending(void)
1725 1726
{
	struct rq *rq = this_rq();
P
Peter Zijlstra 已提交
1727 1728
	struct llist_node *llist = llist_del_all(&rq->wake_list);
	struct task_struct *p;
1729
	unsigned long flags;
1730
	struct rq_flags rf;
1731

1732 1733 1734 1735
	if (!llist)
		return;

	raw_spin_lock_irqsave(&rq->lock, flags);
1736
	rq_pin_lock(rq, &rf);
1737

P
Peter Zijlstra 已提交
1738
	while (llist) {
P
Peter Zijlstra 已提交
1739 1740
		int wake_flags = 0;

P
Peter Zijlstra 已提交
1741 1742
		p = llist_entry(llist, struct task_struct, wake_entry);
		llist = llist_next(llist);
P
Peter Zijlstra 已提交
1743 1744 1745 1746

		if (p->sched_remote_wakeup)
			wake_flags = WF_MIGRATED;

1747
		ttwu_do_activate(rq, p, wake_flags, &rf);
1748 1749
	}

1750
	rq_unpin_lock(rq, &rf);
1751
	raw_spin_unlock_irqrestore(&rq->lock, flags);
1752 1753 1754 1755
}

void scheduler_ipi(void)
{
1756 1757 1758 1759 1760
	/*
	 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
	 * TIF_NEED_RESCHED remotely (for the first time) will also send
	 * this IPI.
	 */
1761
	preempt_fold_need_resched();
1762

1763
	if (llist_empty(&this_rq()->wake_list) && !got_nohz_idle_kick())
1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779
		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
1780
	sched_ttwu_pending();
1781 1782 1783 1784

	/*
	 * Check if someone kicked us for doing the nohz idle load balance.
	 */
1785
	if (unlikely(got_nohz_idle_kick())) {
1786
		this_rq()->idle_balance = 1;
1787
		raise_softirq_irqoff(SCHED_SOFTIRQ);
1788
	}
1789
	irq_exit();
1790 1791
}

P
Peter Zijlstra 已提交
1792
static void ttwu_queue_remote(struct task_struct *p, int cpu, int wake_flags)
1793
{
1794 1795
	struct rq *rq = cpu_rq(cpu);

P
Peter Zijlstra 已提交
1796 1797
	p->sched_remote_wakeup = !!(wake_flags & WF_MIGRATED);

1798 1799 1800 1801 1802 1803
	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);
	}
1804
}
1805

1806 1807 1808 1809 1810
void wake_up_if_idle(int cpu)
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

1811 1812 1813 1814
	rcu_read_lock();

	if (!is_idle_task(rcu_dereference(rq->curr)))
		goto out;
1815 1816 1817 1818 1819 1820 1821

	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
1822
		/* Else CPU is not idle, do nothing here: */
1823 1824
		raw_spin_unlock_irqrestore(&rq->lock, flags);
	}
1825 1826 1827

out:
	rcu_read_unlock();
1828 1829
}

1830
bool cpus_share_cache(int this_cpu, int that_cpu)
1831 1832 1833
{
	return per_cpu(sd_llc_id, this_cpu) == per_cpu(sd_llc_id, that_cpu);
}
1834
#endif /* CONFIG_SMP */
1835

1836
static void ttwu_queue(struct task_struct *p, int cpu, int wake_flags)
1837 1838
{
	struct rq *rq = cpu_rq(cpu);
1839
	struct rq_flags rf;
1840

1841
#if defined(CONFIG_SMP)
1842
	if (sched_feat(TTWU_QUEUE) && !cpus_share_cache(smp_processor_id(), cpu)) {
I
Ingo Molnar 已提交
1843
		sched_clock_cpu(cpu); /* Sync clocks across CPUs */
P
Peter Zijlstra 已提交
1844
		ttwu_queue_remote(p, cpu, wake_flags);
1845 1846 1847 1848
		return;
	}
#endif

1849
	raw_spin_lock(&rq->lock);
1850 1851 1852
	rq_pin_lock(rq, &rf);
	ttwu_do_activate(rq, p, wake_flags, &rf);
	rq_unpin_lock(rq, &rf);
1853
	raw_spin_unlock(&rq->lock);
T
Tejun Heo 已提交
1854 1855
}

1856 1857 1858 1859 1860 1861
/*
 * 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
1862 1863
 * migrates, all its activity on its old CPU [c0] happens-before any subsequent
 * execution on its new CPU [c1].
1864 1865 1866 1867 1868 1869 1870 1871 1872 1873
 *
 * 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
1874
 * Note: the CPU doing B need not be c0 or c1
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 1901 1902 1903 1904 1905
 *
 * 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)
1906
 *   2) smp_cond_load_acquire(!X->on_cpu)
1907 1908 1909 1910 1911 1912 1913 1914 1915 1916
 *
 * 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);
 *
1917
 *                    smp_cond_load_acquire(&X->on_cpu, !VAL);
1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942
 *                    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,
1943
 * since the waking CPU is the one issueing the ACQUIRE (smp_cond_load_acquire).
1944 1945 1946
 *
 */

T
Tejun Heo 已提交
1947
/**
L
Linus Torvalds 已提交
1948
 * try_to_wake_up - wake up a thread
T
Tejun Heo 已提交
1949
 * @p: the thread to be awakened
L
Linus Torvalds 已提交
1950
 * @state: the mask of task states that can be woken
T
Tejun Heo 已提交
1951
 * @wake_flags: wake modifier flags (WF_*)
L
Linus Torvalds 已提交
1952
 *
1953
 * If (@state & @p->state) @p->state = TASK_RUNNING.
L
Linus Torvalds 已提交
1954
 *
1955 1956 1957 1958 1959 1960 1961
 * 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
1962
 */
1963 1964
static int
try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
L
Linus Torvalds 已提交
1965 1966
{
	unsigned long flags;
1967
	int cpu, success = 0;
P
Peter Zijlstra 已提交
1968

1969 1970 1971 1972 1973 1974 1975
	/*
	 * 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();
1976
	raw_spin_lock_irqsave(&p->pi_lock, flags);
P
Peter Zijlstra 已提交
1977
	if (!(p->state & state))
L
Linus Torvalds 已提交
1978 1979
		goto out;

1980 1981
	trace_sched_waking(p);

I
Ingo Molnar 已提交
1982 1983
	/* We're going to change ->state: */
	success = 1;
L
Linus Torvalds 已提交
1984 1985
	cpu = task_cpu(p);

1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
	/*
	 * 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();
2008 2009
	if (p->on_rq && ttwu_remote(p, wake_flags))
		goto stat;
L
Linus Torvalds 已提交
2010 2011

#ifdef CONFIG_SMP
2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030
	/*
	 * 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
2031
	/*
I
Ingo Molnar 已提交
2032
	 * If the owning (remote) CPU is still in the middle of schedule() with
2033
	 * this task as prev, wait until its done referencing the task.
2034 2035 2036 2037 2038
	 *
	 * 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.
2039
	 */
2040
	smp_cond_load_acquire(&p->on_cpu, !VAL);
L
Linus Torvalds 已提交
2041

2042
	p->sched_contributes_to_load = !!task_contributes_to_load(p);
P
Peter Zijlstra 已提交
2043
	p->state = TASK_WAKING;
2044

2045 2046 2047 2048 2049
	if (p->in_iowait) {
		delayacct_blkio_end();
		atomic_dec(&task_rq(p)->nr_iowait);
	}

2050
	cpu = select_task_rq(p, p->wake_cpu, SD_BALANCE_WAKE, wake_flags);
2051 2052
	if (task_cpu(p) != cpu) {
		wake_flags |= WF_MIGRATED;
2053
		set_task_cpu(p, cpu);
2054
	}
2055 2056 2057 2058 2059 2060 2061 2062

#else /* CONFIG_SMP */

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

L
Linus Torvalds 已提交
2063 2064
#endif /* CONFIG_SMP */

2065
	ttwu_queue(p, cpu, wake_flags);
2066
stat:
2067
	ttwu_stat(p, cpu, wake_flags);
L
Linus Torvalds 已提交
2068
out:
2069
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
2070 2071 2072 2073

	return success;
}

T
Tejun Heo 已提交
2074 2075 2076
/**
 * try_to_wake_up_local - try to wake up a local task with rq lock held
 * @p: the thread to be awakened
2077
 * @cookie: context's cookie for pinning
T
Tejun Heo 已提交
2078
 *
2079
 * Put @p on the run-queue if it's not already there. The caller must
T
Tejun Heo 已提交
2080
 * ensure that this_rq() is locked, @p is bound to this_rq() and not
2081
 * the current task.
T
Tejun Heo 已提交
2082
 */
2083
static void try_to_wake_up_local(struct task_struct *p, struct rq_flags *rf)
T
Tejun Heo 已提交
2084 2085 2086
{
	struct rq *rq = task_rq(p);

2087 2088 2089 2090
	if (WARN_ON_ONCE(rq != this_rq()) ||
	    WARN_ON_ONCE(p == current))
		return;

T
Tejun Heo 已提交
2091 2092
	lockdep_assert_held(&rq->lock);

2093
	if (!raw_spin_trylock(&p->pi_lock)) {
2094 2095 2096 2097 2098 2099
		/*
		 * 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.
		 */
2100
		rq_unpin_lock(rq, rf);
2101 2102 2103
		raw_spin_unlock(&rq->lock);
		raw_spin_lock(&p->pi_lock);
		raw_spin_lock(&rq->lock);
2104
		rq_repin_lock(rq, rf);
2105 2106
	}

T
Tejun Heo 已提交
2107
	if (!(p->state & TASK_NORMAL))
2108
		goto out;
T
Tejun Heo 已提交
2109

2110 2111
	trace_sched_waking(p);

2112 2113 2114 2115 2116
	if (!task_on_rq_queued(p)) {
		if (p->in_iowait) {
			delayacct_blkio_end();
			atomic_dec(&rq->nr_iowait);
		}
P
Peter Zijlstra 已提交
2117
		ttwu_activate(rq, p, ENQUEUE_WAKEUP);
2118
	}
P
Peter Zijlstra 已提交
2119

2120
	ttwu_do_wakeup(rq, p, 0, rf);
2121
	ttwu_stat(p, smp_processor_id(), 0);
2122 2123
out:
	raw_spin_unlock(&p->pi_lock);
T
Tejun Heo 已提交
2124 2125
}

2126 2127 2128 2129 2130
/**
 * 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
2131 2132 2133
 * processes.
 *
 * Return: 1 if the process was woken up, 0 if it was already running.
2134 2135 2136 2137
 *
 * 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.
 */
2138
int wake_up_process(struct task_struct *p)
L
Linus Torvalds 已提交
2139
{
2140
	return try_to_wake_up(p, TASK_NORMAL, 0);
L
Linus Torvalds 已提交
2141 2142 2143
}
EXPORT_SYMBOL(wake_up_process);

2144
int wake_up_state(struct task_struct *p, unsigned int state)
L
Linus Torvalds 已提交
2145 2146 2147 2148
{
	return try_to_wake_up(p, state, 0);
}

2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160
/*
 * 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;
2161 2162 2163

	dl_se->dl_throttled = 0;
	dl_se->dl_yielded = 0;
2164 2165
}

L
Linus Torvalds 已提交
2166 2167 2168
/*
 * Perform scheduler related setup for a newly forked process p.
 * p is forked by current.
I
Ingo Molnar 已提交
2169 2170 2171
 *
 * __sched_fork() is basic setup used by init_idle() too:
 */
2172
static void __sched_fork(unsigned long clone_flags, struct task_struct *p)
I
Ingo Molnar 已提交
2173
{
P
Peter Zijlstra 已提交
2174 2175 2176
	p->on_rq			= 0;

	p->se.on_rq			= 0;
I
Ingo Molnar 已提交
2177 2178
	p->se.exec_start		= 0;
	p->se.sum_exec_runtime		= 0;
2179
	p->se.prev_sum_exec_runtime	= 0;
2180
	p->se.nr_migrations		= 0;
P
Peter Zijlstra 已提交
2181
	p->se.vruntime			= 0;
P
Peter Zijlstra 已提交
2182
	INIT_LIST_HEAD(&p->se.group_node);
I
Ingo Molnar 已提交
2183

2184 2185 2186 2187
#ifdef CONFIG_FAIR_GROUP_SCHED
	p->se.cfs_rq			= NULL;
#endif

I
Ingo Molnar 已提交
2188
#ifdef CONFIG_SCHEDSTATS
2189
	/* Even if schedstat is disabled, there should not be garbage */
2190
	memset(&p->se.statistics, 0, sizeof(p->se.statistics));
I
Ingo Molnar 已提交
2191
#endif
N
Nick Piggin 已提交
2192

2193
	RB_CLEAR_NODE(&p->dl.rb_node);
2194
	init_dl_task_timer(&p->dl);
2195
	__dl_clear_params(p);
2196

P
Peter Zijlstra 已提交
2197
	INIT_LIST_HEAD(&p->rt.run_list);
2198 2199 2200 2201
	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
2202

2203 2204 2205
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&p->preempt_notifiers);
#endif
2206 2207 2208

#ifdef CONFIG_NUMA_BALANCING
	if (p->mm && atomic_read(&p->mm->mm_users) == 1) {
2209
		p->mm->numa_next_scan = jiffies + msecs_to_jiffies(sysctl_numa_balancing_scan_delay);
2210 2211 2212
		p->mm->numa_scan_seq = 0;
	}

2213 2214 2215 2216 2217
	if (clone_flags & CLONE_VM)
		p->numa_preferred_nid = current->numa_preferred_nid;
	else
		p->numa_preferred_nid = -1;

2218 2219
	p->node_stamp = 0ULL;
	p->numa_scan_seq = p->mm ? p->mm->numa_scan_seq : 0;
2220
	p->numa_scan_period = sysctl_numa_balancing_scan_delay;
2221
	p->numa_work.next = &p->numa_work;
2222
	p->numa_faults = NULL;
2223 2224
	p->last_task_numa_placement = 0;
	p->last_sum_exec_runtime = 0;
2225 2226

	p->numa_group = NULL;
2227
#endif /* CONFIG_NUMA_BALANCING */
I
Ingo Molnar 已提交
2228 2229
}

2230 2231
DEFINE_STATIC_KEY_FALSE(sched_numa_balancing);

2232
#ifdef CONFIG_NUMA_BALANCING
2233

2234 2235 2236
void set_numabalancing_state(bool enabled)
{
	if (enabled)
2237
		static_branch_enable(&sched_numa_balancing);
2238
	else
2239
		static_branch_disable(&sched_numa_balancing);
2240
}
2241 2242 2243 2244 2245 2246 2247

#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;
2248
	int state = static_branch_likely(&sched_numa_balancing);
2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263

	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
2264

2265 2266
#ifdef CONFIG_SCHEDSTATS

2267
DEFINE_STATIC_KEY_FALSE(sched_schedstats);
2268
static bool __initdata __sched_schedstats = false;
2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291

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;

2292 2293 2294 2295 2296
	/*
	 * 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.
	 */
2297
	if (!strcmp(str, "enable")) {
2298
		__sched_schedstats = true;
2299 2300
		ret = 1;
	} else if (!strcmp(str, "disable")) {
2301
		__sched_schedstats = false;
2302 2303 2304 2305 2306 2307 2308 2309 2310 2311
		ret = 1;
	}
out:
	if (!ret)
		pr_warn("Unable to parse schedstats=\n");

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

2312 2313 2314 2315 2316
static void __init init_schedstats(void)
{
	set_schedstats(__sched_schedstats);
}

2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336
#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;
}
2337 2338 2339 2340
#endif /* CONFIG_PROC_SYSCTL */
#else  /* !CONFIG_SCHEDSTATS */
static inline void init_schedstats(void) {}
#endif /* CONFIG_SCHEDSTATS */
I
Ingo Molnar 已提交
2341 2342 2343 2344

/*
 * fork()/clone()-time setup:
 */
2345
int sched_fork(unsigned long clone_flags, struct task_struct *p)
I
Ingo Molnar 已提交
2346
{
2347
	unsigned long flags;
I
Ingo Molnar 已提交
2348 2349
	int cpu = get_cpu();

2350
	__sched_fork(clone_flags, p);
2351
	/*
2352
	 * We mark the process as NEW here. This guarantees that
2353 2354 2355
	 * nobody will actually run it, and a signal or other external
	 * event cannot wake it up and insert it on the runqueue either.
	 */
2356
	p->state = TASK_NEW;
I
Ingo Molnar 已提交
2357

2358 2359 2360 2361 2362
	/*
	 * Make sure we do not leak PI boosting priority to the child.
	 */
	p->prio = current->normal_prio;

2363 2364 2365 2366
	/*
	 * Revert to default priority/policy on fork if requested.
	 */
	if (unlikely(p->sched_reset_on_fork)) {
2367
		if (task_has_dl_policy(p) || task_has_rt_policy(p)) {
2368
			p->policy = SCHED_NORMAL;
2369
			p->static_prio = NICE_TO_PRIO(0);
2370 2371 2372 2373 2374 2375
			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);
2376

2377 2378 2379 2380 2381 2382
		/*
		 * We don't need the reset flag anymore after the fork. It has
		 * fulfilled its duty:
		 */
		p->sched_reset_on_fork = 0;
	}
2383

2384 2385 2386 2387 2388 2389
	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
2390
		p->sched_class = &fair_sched_class;
2391
	}
2392

2393
	init_entity_runnable_average(&p->se);
P
Peter Zijlstra 已提交
2394

2395 2396 2397 2398 2399 2400 2401
	/*
	 * 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.
	 */
2402
	raw_spin_lock_irqsave(&p->pi_lock, flags);
2403
	/*
I
Ingo Molnar 已提交
2404
	 * We're setting the CPU for the first time, we don't migrate,
2405 2406 2407 2408 2409
	 * so use __set_task_cpu().
	 */
	__set_task_cpu(p, cpu);
	if (p->sched_class->task_fork)
		p->sched_class->task_fork(p);
2410
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
2411

2412
#ifdef CONFIG_SCHED_INFO
I
Ingo Molnar 已提交
2413
	if (likely(sched_info_on()))
2414
		memset(&p->sched_info, 0, sizeof(p->sched_info));
L
Linus Torvalds 已提交
2415
#endif
P
Peter Zijlstra 已提交
2416 2417
#if defined(CONFIG_SMP)
	p->on_cpu = 0;
2418
#endif
2419
	init_task_preempt_count(p);
2420
#ifdef CONFIG_SMP
2421
	plist_node_init(&p->pushable_tasks, MAX_PRIO);
2422
	RB_CLEAR_NODE(&p->pushable_dl_tasks);
2423
#endif
2424

N
Nick Piggin 已提交
2425
	put_cpu();
2426
	return 0;
L
Linus Torvalds 已提交
2427 2428
}

2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447
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)
{
2448 2449
	RCU_LOCKDEP_WARN(!rcu_read_lock_sched_held(),
			 "sched RCU must be held");
2450 2451 2452
	return &cpu_rq(i)->rd->dl_bw;
}

2453
static inline int dl_bw_cpus(int i)
2454
{
2455 2456 2457
	struct root_domain *rd = cpu_rq(i)->rd;
	int cpus = 0;

2458 2459
	RCU_LOCKDEP_WARN(!rcu_read_lock_sched_held(),
			 "sched RCU must be held");
2460 2461 2462 2463
	for_each_cpu_and(i, rd->span, cpu_active_mask)
		cpus++;

	return cpus;
2464 2465 2466 2467 2468 2469 2470
}
#else
inline struct dl_bw *dl_bw_of(int i)
{
	return &cpu_rq(i)->dl.dl_bw;
}

2471
static inline int dl_bw_cpus(int i)
2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483
{
	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.
2484 2485 2486
 *
 * XXX we should delay bw change until the task's 0-lag point, see
 * __setparam_dl().
2487 2488 2489 2490 2491 2492
 */
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));
2493
	u64 period = attr->sched_period ?: attr->sched_deadline;
2494 2495
	u64 runtime = attr->sched_runtime;
	u64 new_bw = dl_policy(policy) ? to_ratio(period, runtime) : 0;
2496
	int cpus, err = -1;
2497

2498 2499
	/* !deadline task may carry old deadline bandwidth */
	if (new_bw == p->dl.dl_bw && task_has_dl_policy(p))
2500 2501 2502 2503 2504 2505 2506 2507
		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);
2508
	cpus = dl_bw_cpus(task_cpu(p));
2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528
	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
2529 2530 2531 2532 2533 2534 2535
/*
 * 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.
 */
2536
void wake_up_new_task(struct task_struct *p)
L
Linus Torvalds 已提交
2537
{
2538
	struct rq_flags rf;
I
Ingo Molnar 已提交
2539
	struct rq *rq;
2540

2541
	raw_spin_lock_irqsave(&p->pi_lock, rf.flags);
2542
	p->state = TASK_RUNNING;
2543 2544 2545 2546
#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
2547
	 *  - any previously selected CPU might disappear through hotplug
2548 2549 2550
	 *
	 * Use __set_task_cpu() to avoid calling sched_class::migrate_task_rq,
	 * as we're not fully set-up yet.
2551
	 */
2552
	__set_task_cpu(p, select_task_rq(p, task_cpu(p), SD_BALANCE_FORK, 0));
2553
#endif
2554
	rq = __task_rq_lock(p, &rf);
2555
	update_rq_clock(rq);
2556
	post_init_entity_util_avg(&p->se);
2557

P
Peter Zijlstra 已提交
2558
	activate_task(rq, p, 0);
2559
	p->on_rq = TASK_ON_RQ_QUEUED;
2560
	trace_sched_wakeup_new(p);
P
Peter Zijlstra 已提交
2561
	check_preempt_curr(rq, p, WF_FORK);
2562
#ifdef CONFIG_SMP
2563 2564 2565 2566 2567
	if (p->sched_class->task_woken) {
		/*
		 * Nothing relies on rq->lock after this, so its fine to
		 * drop it.
		 */
2568
		rq_unpin_lock(rq, &rf);
2569
		p->sched_class->task_woken(rq, p);
2570
		rq_repin_lock(rq, &rf);
2571
	}
2572
#endif
2573
	task_rq_unlock(rq, p, &rf);
L
Linus Torvalds 已提交
2574 2575
}

2576 2577
#ifdef CONFIG_PREEMPT_NOTIFIERS

2578 2579
static struct static_key preempt_notifier_key = STATIC_KEY_INIT_FALSE;

2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591
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);

2592
/**
2593
 * preempt_notifier_register - tell me when current is being preempted & rescheduled
R
Randy Dunlap 已提交
2594
 * @notifier: notifier struct to register
2595 2596 2597
 */
void preempt_notifier_register(struct preempt_notifier *notifier)
{
2598 2599 2600
	if (!static_key_false(&preempt_notifier_key))
		WARN(1, "registering preempt_notifier while notifiers disabled\n");

2601 2602 2603 2604 2605 2606
	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
2607
 * @notifier: notifier struct to unregister
2608
 *
2609
 * This is *not* safe to call from within a preemption notifier.
2610 2611 2612 2613 2614 2615 2616
 */
void preempt_notifier_unregister(struct preempt_notifier *notifier)
{
	hlist_del(&notifier->link);
}
EXPORT_SYMBOL_GPL(preempt_notifier_unregister);

2617
static void __fire_sched_in_preempt_notifiers(struct task_struct *curr)
2618 2619 2620
{
	struct preempt_notifier *notifier;

2621
	hlist_for_each_entry(notifier, &curr->preempt_notifiers, link)
2622 2623 2624
		notifier->ops->sched_in(notifier, raw_smp_processor_id());
}

2625 2626 2627 2628 2629 2630
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);
}

2631
static void
2632 2633
__fire_sched_out_preempt_notifiers(struct task_struct *curr,
				   struct task_struct *next)
2634 2635 2636
{
	struct preempt_notifier *notifier;

2637
	hlist_for_each_entry(notifier, &curr->preempt_notifiers, link)
2638 2639 2640
		notifier->ops->sched_out(notifier, next);
}

2641 2642 2643 2644 2645 2646 2647 2648
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);
}

2649
#else /* !CONFIG_PREEMPT_NOTIFIERS */
2650

2651
static inline void fire_sched_in_preempt_notifiers(struct task_struct *curr)
2652 2653 2654
{
}

2655
static inline void
2656 2657 2658 2659 2660
fire_sched_out_preempt_notifiers(struct task_struct *curr,
				 struct task_struct *next)
{
}

2661
#endif /* CONFIG_PREEMPT_NOTIFIERS */
2662

2663 2664 2665
/**
 * prepare_task_switch - prepare to switch tasks
 * @rq: the runqueue preparing to switch
R
Randy Dunlap 已提交
2666
 * @prev: the current task that is being switched out
2667 2668 2669 2670 2671 2672 2673 2674 2675
 * @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.
 */
2676 2677 2678
static inline void
prepare_task_switch(struct rq *rq, struct task_struct *prev,
		    struct task_struct *next)
2679
{
2680
	sched_info_switch(rq, prev, next);
2681
	perf_event_task_sched_out(prev, next);
2682
	fire_sched_out_preempt_notifiers(prev, next);
2683 2684 2685 2686
	prepare_lock_switch(rq, next);
	prepare_arch_switch(next);
}

L
Linus Torvalds 已提交
2687 2688 2689 2690
/**
 * finish_task_switch - clean up after a task-switch
 * @prev: the thread we just switched away from.
 *
2691 2692 2693 2694
 * 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
2695 2696
 *
 * Note that we may have delayed dropping an mm in context_switch(). If
I
Ingo Molnar 已提交
2697
 * so, we finish that here outside of the runqueue lock. (Doing it
L
Linus Torvalds 已提交
2698 2699
 * with the lock held can cause deadlocks; see schedule() for
 * details.)
2700 2701 2702 2703 2704
 *
 * 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
2705
 */
2706
static struct rq *finish_task_switch(struct task_struct *prev)
L
Linus Torvalds 已提交
2707 2708
	__releases(rq->lock)
{
2709
	struct rq *rq = this_rq();
L
Linus Torvalds 已提交
2710
	struct mm_struct *mm = rq->prev_mm;
O
Oleg Nesterov 已提交
2711
	long prev_state;
L
Linus Torvalds 已提交
2712

2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723
	/*
	 * 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.
	 */
2724 2725 2726 2727
	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);
2728

L
Linus Torvalds 已提交
2729 2730 2731 2732
	rq->prev_mm = NULL;

	/*
	 * A task struct has one reference for the use as "current".
2733
	 * If a task dies, then it sets TASK_DEAD in tsk->state and calls
O
Oleg Nesterov 已提交
2734 2735
	 * schedule one last time. The schedule call will never return, and
	 * the scheduled task must drop that reference.
2736 2737 2738 2739 2740
	 *
	 * 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
2741
	 */
O
Oleg Nesterov 已提交
2742
	prev_state = prev->state;
2743
	vtime_task_switch(prev);
2744
	perf_event_task_sched_in(prev, current);
2745
	finish_lock_switch(rq, prev);
2746
	finish_arch_post_lock_switch();
S
Steven Rostedt 已提交
2747

2748
	fire_sched_in_preempt_notifiers(current);
L
Linus Torvalds 已提交
2749 2750
	if (mm)
		mmdrop(mm);
2751
	if (unlikely(prev_state == TASK_DEAD)) {
2752 2753 2754
		if (prev->sched_class->task_dead)
			prev->sched_class->task_dead(prev);

2755 2756 2757
		/*
		 * Remove function-return probe instances associated with this
		 * task and put them back on the free list.
I
Ingo Molnar 已提交
2758
		 */
2759
		kprobe_flush_task(prev);
2760 2761 2762 2763

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

L
Linus Torvalds 已提交
2764
		put_task_struct(prev);
2765
	}
2766

2767
	tick_nohz_task_switch();
2768
	return rq;
L
Linus Torvalds 已提交
2769 2770
}

2771 2772 2773
#ifdef CONFIG_SMP

/* rq->lock is NOT held, but preemption is disabled */
2774
static void __balance_callback(struct rq *rq)
2775
{
2776 2777 2778
	struct callback_head *head, *next;
	void (*func)(struct rq *rq);
	unsigned long flags;
2779

2780 2781 2782 2783 2784 2785 2786 2787
	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;
2788

2789
		func(rq);
2790
	}
2791 2792 2793 2794 2795 2796 2797
	raw_spin_unlock_irqrestore(&rq->lock, flags);
}

static inline void balance_callback(struct rq *rq)
{
	if (unlikely(rq->balance_callback))
		__balance_callback(rq);
2798 2799 2800
}

#else
2801

2802
static inline void balance_callback(struct rq *rq)
2803
{
L
Linus Torvalds 已提交
2804 2805
}

2806 2807
#endif

L
Linus Torvalds 已提交
2808 2809 2810 2811
/**
 * schedule_tail - first thing a freshly forked thread must call.
 * @prev: the thread we just switched away from.
 */
2812
asmlinkage __visible void schedule_tail(struct task_struct *prev)
L
Linus Torvalds 已提交
2813 2814
	__releases(rq->lock)
{
2815
	struct rq *rq;
2816

2817 2818 2819 2820 2821 2822 2823 2824 2825
	/*
	 * 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).
	 */

2826
	rq = finish_task_switch(prev);
2827
	balance_callback(rq);
2828
	preempt_enable();
2829

L
Linus Torvalds 已提交
2830
	if (current->set_child_tid)
2831
		put_user(task_pid_vnr(current), current->set_child_tid);
L
Linus Torvalds 已提交
2832 2833 2834
}

/*
2835
 * context_switch - switch to the new MM and the new thread's register state.
L
Linus Torvalds 已提交
2836
 */
2837
static __always_inline struct rq *
2838
context_switch(struct rq *rq, struct task_struct *prev,
2839
	       struct task_struct *next, struct rq_flags *rf)
L
Linus Torvalds 已提交
2840
{
I
Ingo Molnar 已提交
2841
	struct mm_struct *mm, *oldmm;
L
Linus Torvalds 已提交
2842

2843
	prepare_task_switch(rq, prev, next);
2844

I
Ingo Molnar 已提交
2845 2846
	mm = next->mm;
	oldmm = prev->active_mm;
2847 2848 2849 2850 2851
	/*
	 * For paravirt, this is coupled with an exit in switch_to to
	 * combine the page table reload and the switch backend into
	 * one hypercall.
	 */
2852
	arch_start_context_switch(prev);
2853

2854
	if (!mm) {
L
Linus Torvalds 已提交
2855
		next->active_mm = oldmm;
V
Vegard Nossum 已提交
2856
		mmgrab(oldmm);
L
Linus Torvalds 已提交
2857 2858
		enter_lazy_tlb(oldmm, next);
	} else
2859
		switch_mm_irqs_off(oldmm, mm, next);
L
Linus Torvalds 已提交
2860

2861
	if (!prev->mm) {
L
Linus Torvalds 已提交
2862 2863 2864
		prev->active_mm = NULL;
		rq->prev_mm = oldmm;
	}
2865

2866
	rq->clock_update_flags &= ~(RQCF_ACT_SKIP|RQCF_REQ_SKIP);
2867

2868 2869 2870 2871 2872 2873
	/*
	 * 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:
	 */
2874
	rq_unpin_lock(rq, rf);
2875
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
L
Linus Torvalds 已提交
2876 2877 2878

	/* Here we just switch the register state and the stack. */
	switch_to(prev, next, prev);
I
Ingo Molnar 已提交
2879
	barrier();
2880 2881

	return finish_task_switch(prev);
L
Linus Torvalds 已提交
2882 2883 2884
}

/*
2885
 * nr_running and nr_context_switches:
L
Linus Torvalds 已提交
2886 2887
 *
 * externally visible scheduler statistics: current number of runnable
2888
 * threads, total number of context switches performed since bootup.
L
Linus Torvalds 已提交
2889 2890 2891 2892 2893 2894 2895 2896 2897
 */
unsigned long nr_running(void)
{
	unsigned long i, sum = 0;

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

	return sum;
2898
}
L
Linus Torvalds 已提交
2899

2900
/*
I
Ingo Molnar 已提交
2901
 * Check if only the current task is running on the CPU.
2902 2903 2904 2905 2906 2907 2908 2909 2910 2911
 *
 * 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)
2912 2913 2914
 */
bool single_task_running(void)
{
2915
	return raw_rq()->nr_running == 1;
2916 2917 2918
}
EXPORT_SYMBOL(single_task_running);

L
Linus Torvalds 已提交
2919
unsigned long long nr_context_switches(void)
2920
{
2921 2922
	int i;
	unsigned long long sum = 0;
2923

2924
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2925
		sum += cpu_rq(i)->nr_switches;
2926

L
Linus Torvalds 已提交
2927 2928
	return sum;
}
2929

2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959
/*
 * 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
2960 2961 2962
unsigned long nr_iowait(void)
{
	unsigned long i, sum = 0;
2963

2964
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2965
		sum += atomic_read(&cpu_rq(i)->nr_iowait);
2966

L
Linus Torvalds 已提交
2967 2968
	return sum;
}
2969

2970 2971 2972 2973 2974 2975 2976
/*
 * 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.
 */

2977
unsigned long nr_iowait_cpu(int cpu)
2978
{
2979
	struct rq *this = cpu_rq(cpu);
2980 2981
	return atomic_read(&this->nr_iowait);
}
2982

2983 2984
void get_iowait_load(unsigned long *nr_waiters, unsigned long *load)
{
2985 2986 2987
	struct rq *rq = this_rq();
	*nr_waiters = atomic_read(&rq->nr_iowait);
	*load = rq->load.weight;
2988 2989
}

I
Ingo Molnar 已提交
2990
#ifdef CONFIG_SMP
2991

2992
/*
P
Peter Zijlstra 已提交
2993 2994
 * sched_exec - execve() is a valuable balancing opportunity, because at
 * this point the task has the smallest effective memory and cache footprint.
2995
 */
P
Peter Zijlstra 已提交
2996
void sched_exec(void)
2997
{
P
Peter Zijlstra 已提交
2998
	struct task_struct *p = current;
L
Linus Torvalds 已提交
2999
	unsigned long flags;
3000
	int dest_cpu;
3001

3002
	raw_spin_lock_irqsave(&p->pi_lock, flags);
3003
	dest_cpu = p->sched_class->select_task_rq(p, task_cpu(p), SD_BALANCE_EXEC, 0);
3004 3005
	if (dest_cpu == smp_processor_id())
		goto unlock;
P
Peter Zijlstra 已提交
3006

3007
	if (likely(cpu_active(dest_cpu))) {
3008
		struct migration_arg arg = { p, dest_cpu };
3009

3010 3011
		raw_spin_unlock_irqrestore(&p->pi_lock, flags);
		stop_one_cpu(task_cpu(p), migration_cpu_stop, &arg);
L
Linus Torvalds 已提交
3012 3013
		return;
	}
3014
unlock:
3015
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
3016
}
I
Ingo Molnar 已提交
3017

L
Linus Torvalds 已提交
3018 3019 3020
#endif

DEFINE_PER_CPU(struct kernel_stat, kstat);
3021
DEFINE_PER_CPU(struct kernel_cpustat, kernel_cpustat);
L
Linus Torvalds 已提交
3022 3023

EXPORT_PER_CPU_SYMBOL(kstat);
3024
EXPORT_PER_CPU_SYMBOL(kernel_cpustat);
L
Linus Torvalds 已提交
3025

3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042
/*
 * 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);
}

3043 3044 3045 3046 3047 3048 3049
/*
 * 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)
{
3050
	struct rq_flags rf;
3051
	struct rq *rq;
3052
	u64 ns;
3053

3054 3055 3056 3057 3058 3059
#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
3060 3061
	 * 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
3062
	 * indistinguishable from the read occurring a few cycles earlier.
3063 3064
	 * If we see ->on_cpu without ->on_rq, the task is leaving, and has
	 * been accounted, so we're correct here as well.
3065
	 */
3066
	if (!p->on_cpu || !task_on_rq_queued(p))
3067 3068 3069
		return p->se.sum_exec_runtime;
#endif

3070
	rq = task_rq_lock(p, &rf);
3071 3072 3073 3074 3075 3076
	/*
	 * 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)) {
3077
		prefetch_curr_exec_start(p);
3078 3079 3080 3081
		update_rq_clock(rq);
		p->sched_class->update_curr(rq);
	}
	ns = p->se.sum_exec_runtime;
3082
	task_rq_unlock(rq, p, &rf);
3083 3084 3085

	return ns;
}
3086

3087 3088 3089 3090 3091 3092 3093 3094
/*
 * 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
3095
	struct task_struct *curr = rq->curr;
3096 3097

	sched_clock_tick();
I
Ingo Molnar 已提交
3098

3099
	raw_spin_lock(&rq->lock);
3100
	update_rq_clock(rq);
P
Peter Zijlstra 已提交
3101
	curr->sched_class->task_tick(rq, curr, 0);
3102
	cpu_load_update_active(rq);
3103
	calc_global_load_tick(rq);
3104
	raw_spin_unlock(&rq->lock);
3105

3106
	perf_event_task_tick();
3107

3108
#ifdef CONFIG_SMP
3109
	rq->idle_balance = idle_cpu(cpu);
3110
	trigger_load_balance(rq);
3111
#endif
3112
	rq_last_tick_reset(rq);
L
Linus Torvalds 已提交
3113 3114
}

3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125
#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.
3126 3127
 *
 * Return: Maximum deferment in nanoseconds.
3128 3129 3130 3131
 */
u64 scheduler_tick_max_deferment(void)
{
	struct rq *rq = this_rq();
3132
	unsigned long next, now = READ_ONCE(jiffies);
3133 3134 3135 3136 3137 3138

	next = rq->last_sched_tick + HZ;

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

3139
	return jiffies_to_nsecs(next - now);
L
Linus Torvalds 已提交
3140
}
3141
#endif
L
Linus Torvalds 已提交
3142

3143 3144
#if defined(CONFIG_PREEMPT) && (defined(CONFIG_DEBUG_PREEMPT) || \
				defined(CONFIG_PREEMPT_TRACER))
3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158
/*
 * 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);
	}
}
3159

3160
void preempt_count_add(int val)
L
Linus Torvalds 已提交
3161
{
3162
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3163 3164 3165
	/*
	 * Underflow?
	 */
3166 3167
	if (DEBUG_LOCKS_WARN_ON((preempt_count() < 0)))
		return;
3168
#endif
3169
	__preempt_count_add(val);
3170
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3171 3172 3173
	/*
	 * Spinlock count overflowing soon?
	 */
3174 3175
	DEBUG_LOCKS_WARN_ON((preempt_count() & PREEMPT_MASK) >=
				PREEMPT_MASK - 10);
3176
#endif
3177
	preempt_latency_start(val);
L
Linus Torvalds 已提交
3178
}
3179
EXPORT_SYMBOL(preempt_count_add);
3180
NOKPROBE_SYMBOL(preempt_count_add);
L
Linus Torvalds 已提交
3181

3182 3183 3184 3185 3186 3187 3188 3189 3190 3191
/*
 * 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());
}

3192
void preempt_count_sub(int val)
L
Linus Torvalds 已提交
3193
{
3194
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3195 3196 3197
	/*
	 * Underflow?
	 */
3198
	if (DEBUG_LOCKS_WARN_ON(val > preempt_count()))
3199
		return;
L
Linus Torvalds 已提交
3200 3201 3202
	/*
	 * Is the spinlock portion underflowing?
	 */
3203 3204 3205
	if (DEBUG_LOCKS_WARN_ON((val < PREEMPT_MASK) &&
			!(preempt_count() & PREEMPT_MASK)))
		return;
3206
#endif
3207

3208
	preempt_latency_stop(val);
3209
	__preempt_count_sub(val);
L
Linus Torvalds 已提交
3210
}
3211
EXPORT_SYMBOL(preempt_count_sub);
3212
NOKPROBE_SYMBOL(preempt_count_sub);
L
Linus Torvalds 已提交
3213

3214 3215 3216
#else
static inline void preempt_latency_start(int val) { }
static inline void preempt_latency_stop(int val) { }
L
Linus Torvalds 已提交
3217 3218
#endif

3219 3220 3221 3222 3223 3224 3225 3226 3227
static inline unsigned long get_preempt_disable_ip(struct task_struct *p)
{
#ifdef CONFIG_DEBUG_PREEMPT
	return p->preempt_disable_ip;
#else
	return 0;
#endif
}

L
Linus Torvalds 已提交
3228
/*
I
Ingo Molnar 已提交
3229
 * Print scheduling while atomic bug:
L
Linus Torvalds 已提交
3230
 */
I
Ingo Molnar 已提交
3231
static noinline void __schedule_bug(struct task_struct *prev)
L
Linus Torvalds 已提交
3232
{
3233 3234 3235
	/* Save this before calling printk(), since that will clobber it */
	unsigned long preempt_disable_ip = get_preempt_disable_ip(current);

3236 3237 3238
	if (oops_in_progress)
		return;

P
Peter Zijlstra 已提交
3239 3240
	printk(KERN_ERR "BUG: scheduling while atomic: %s/%d/0x%08x\n",
		prev->comm, prev->pid, preempt_count());
3241

I
Ingo Molnar 已提交
3242
	debug_show_held_locks(prev);
3243
	print_modules();
I
Ingo Molnar 已提交
3244 3245
	if (irqs_disabled())
		print_irqtrace_events(prev);
3246 3247
	if (IS_ENABLED(CONFIG_DEBUG_PREEMPT)
	    && in_atomic_preempt_off()) {
3248
		pr_err("Preemption disabled at:");
3249
		print_ip_sym(preempt_disable_ip);
3250 3251
		pr_cont("\n");
	}
3252 3253 3254
	if (panic_on_warn)
		panic("scheduling while atomic\n");

3255
	dump_stack();
3256
	add_taint(TAINT_WARN, LOCKDEP_STILL_OK);
I
Ingo Molnar 已提交
3257
}
L
Linus Torvalds 已提交
3258

I
Ingo Molnar 已提交
3259 3260 3261 3262 3263
/*
 * Various schedule()-time debugging checks and statistics:
 */
static inline void schedule_debug(struct task_struct *prev)
{
3264
#ifdef CONFIG_SCHED_STACK_END_CHECK
J
Jann Horn 已提交
3265 3266
	if (task_stack_end_corrupted(prev))
		panic("corrupted stack end detected inside scheduler\n");
3267
#endif
3268

3269
	if (unlikely(in_atomic_preempt_off())) {
I
Ingo Molnar 已提交
3270
		__schedule_bug(prev);
3271 3272
		preempt_count_set(PREEMPT_DISABLED);
	}
3273
	rcu_sleep_check();
I
Ingo Molnar 已提交
3274

L
Linus Torvalds 已提交
3275 3276
	profile_hit(SCHED_PROFILING, __builtin_return_address(0));

3277
	schedstat_inc(this_rq()->sched_count);
I
Ingo Molnar 已提交
3278 3279 3280 3281 3282 3283
}

/*
 * Pick up the highest-prio task:
 */
static inline struct task_struct *
3284
pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
I
Ingo Molnar 已提交
3285
{
3286
	const struct sched_class *class;
I
Ingo Molnar 已提交
3287
	struct task_struct *p;
L
Linus Torvalds 已提交
3288 3289

	/*
3290 3291 3292 3293
	 * Optimization: we know that if all tasks are in the fair class we can
	 * call that function directly, but only if the @prev task wasn't of a
	 * higher scheduling class, because otherwise those loose the
	 * opportunity to pull in more work from other CPUs.
L
Linus Torvalds 已提交
3294
	 */
3295 3296 3297 3298
	if (likely((prev->sched_class == &idle_sched_class ||
		    prev->sched_class == &fair_sched_class) &&
		   rq->nr_running == rq->cfs.h_nr_running)) {

3299
		p = fair_sched_class.pick_next_task(rq, prev, rf);
3300 3301 3302
		if (unlikely(p == RETRY_TASK))
			goto again;

I
Ingo Molnar 已提交
3303
		/* Assumes fair_sched_class->next == idle_sched_class */
3304
		if (unlikely(!p))
3305
			p = idle_sched_class.pick_next_task(rq, prev, rf);
3306 3307

		return p;
L
Linus Torvalds 已提交
3308 3309
	}

3310
again:
3311
	for_each_class(class) {
3312
		p = class->pick_next_task(rq, prev, rf);
3313 3314 3315
		if (p) {
			if (unlikely(p == RETRY_TASK))
				goto again;
I
Ingo Molnar 已提交
3316
			return p;
3317
		}
I
Ingo Molnar 已提交
3318
	}
3319

I
Ingo Molnar 已提交
3320 3321
	/* The idle class should always have a runnable task: */
	BUG();
I
Ingo Molnar 已提交
3322
}
L
Linus Torvalds 已提交
3323

I
Ingo Molnar 已提交
3324
/*
3325
 * __schedule() is the main scheduler function.
3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359
 *
 * 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
3360
 *
3361
 * WARNING: must be called with preemption disabled!
I
Ingo Molnar 已提交
3362
 */
3363
static void __sched notrace __schedule(bool preempt)
I
Ingo Molnar 已提交
3364 3365
{
	struct task_struct *prev, *next;
3366
	unsigned long *switch_count;
3367
	struct rq_flags rf;
I
Ingo Molnar 已提交
3368
	struct rq *rq;
3369
	int cpu;
I
Ingo Molnar 已提交
3370 3371 3372 3373 3374 3375

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

	schedule_debug(prev);
L
Linus Torvalds 已提交
3376

3377
	if (sched_feat(HRTICK))
M
Mike Galbraith 已提交
3378
		hrtick_clear(rq);
P
Peter Zijlstra 已提交
3379

3380 3381 3382
	local_irq_disable();
	rcu_note_context_switch();

3383 3384 3385 3386 3387 3388
	/*
	 * 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();
3389
	raw_spin_lock(&rq->lock);
3390
	rq_pin_lock(rq, &rf);
L
Linus Torvalds 已提交
3391

I
Ingo Molnar 已提交
3392 3393
	/* Promote REQ to ACT */
	rq->clock_update_flags <<= 1;
3394

3395
	switch_count = &prev->nivcsw;
3396
	if (!preempt && prev->state) {
T
Tejun Heo 已提交
3397
		if (unlikely(signal_pending_state(prev->state, prev))) {
L
Linus Torvalds 已提交
3398
			prev->state = TASK_RUNNING;
T
Tejun Heo 已提交
3399
		} else {
3400 3401 3402
			deactivate_task(rq, prev, DEQUEUE_SLEEP);
			prev->on_rq = 0;

3403 3404 3405 3406 3407
			if (prev->in_iowait) {
				atomic_inc(&rq->nr_iowait);
				delayacct_blkio_start();
			}

T
Tejun Heo 已提交
3408
			/*
3409 3410 3411
			 * 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
3412 3413 3414 3415
			 */
			if (prev->flags & PF_WQ_WORKER) {
				struct task_struct *to_wakeup;

3416
				to_wakeup = wq_worker_sleeping(prev);
T
Tejun Heo 已提交
3417
				if (to_wakeup)
3418
					try_to_wake_up_local(to_wakeup, &rf);
T
Tejun Heo 已提交
3419 3420
			}
		}
I
Ingo Molnar 已提交
3421
		switch_count = &prev->nvcsw;
L
Linus Torvalds 已提交
3422 3423
	}

3424
	if (task_on_rq_queued(prev))
3425 3426
		update_rq_clock(rq);

3427
	next = pick_next_task(rq, prev, &rf);
3428
	clear_tsk_need_resched(prev);
3429
	clear_preempt_need_resched();
L
Linus Torvalds 已提交
3430 3431 3432 3433 3434 3435

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

3436
		trace_sched_switch(preempt, prev, next);
I
Ingo Molnar 已提交
3437 3438 3439

		/* Also unlocks the rq: */
		rq = context_switch(rq, prev, next, &rf);
3440
	} else {
3441
		rq->clock_update_flags &= ~(RQCF_ACT_SKIP|RQCF_REQ_SKIP);
3442
		rq_unpin_lock(rq, &rf);
3443
		raw_spin_unlock_irq(&rq->lock);
3444
	}
L
Linus Torvalds 已提交
3445

3446
	balance_callback(rq);
L
Linus Torvalds 已提交
3447
}
3448

3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465
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
3466
	/* Causes final put_task_struct in finish_task_switch(): */
3467
	__set_current_state(TASK_DEAD);
I
Ingo Molnar 已提交
3468 3469 3470 3471

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

3472 3473
	__schedule(false);
	BUG();
I
Ingo Molnar 已提交
3474 3475

	/* Avoid "noreturn function does return" - but don't continue if BUG() is a NOP: */
3476
	for (;;)
I
Ingo Molnar 已提交
3477
		cpu_relax();
3478 3479
}

3480 3481
static inline void sched_submit_work(struct task_struct *tsk)
{
3482
	if (!tsk->state || tsk_is_pi_blocked(tsk))
3483 3484 3485 3486 3487 3488 3489 3490 3491
		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);
}

3492
asmlinkage __visible void __sched schedule(void)
3493
{
3494 3495 3496
	struct task_struct *tsk = current;

	sched_submit_work(tsk);
3497
	do {
3498
		preempt_disable();
3499
		__schedule(false);
3500
		sched_preempt_enable_no_resched();
3501
	} while (need_resched());
3502
}
L
Linus Torvalds 已提交
3503 3504
EXPORT_SYMBOL(schedule);

3505
#ifdef CONFIG_CONTEXT_TRACKING
3506
asmlinkage __visible void __sched schedule_user(void)
3507 3508 3509 3510 3511 3512
{
	/*
	 * 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.
3513 3514
	 *
	 * NB: There are buggy callers of this function.  Ideally we
3515
	 * should warn if prev_state != CONTEXT_USER, but that will trigger
3516
	 * too frequently to make sense yet.
3517
	 */
3518
	enum ctx_state prev_state = exception_enter();
3519
	schedule();
3520
	exception_exit(prev_state);
3521 3522 3523
}
#endif

3524 3525 3526 3527 3528 3529 3530
/**
 * schedule_preempt_disabled - called with preemption disabled
 *
 * Returns with preemption disabled. Note: preempt_count must be 1
 */
void __sched schedule_preempt_disabled(void)
{
3531
	sched_preempt_enable_no_resched();
3532 3533 3534 3535
	schedule();
	preempt_disable();
}

3536
static void __sched notrace preempt_schedule_common(void)
3537 3538
{
	do {
3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551
		/*
		 * 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.
		 */
3552
		preempt_disable_notrace();
3553
		preempt_latency_start(1);
3554
		__schedule(true);
3555
		preempt_latency_stop(1);
3556
		preempt_enable_no_resched_notrace();
3557 3558 3559 3560 3561 3562 3563 3564

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

L
Linus Torvalds 已提交
3565 3566
#ifdef CONFIG_PREEMPT
/*
3567
 * this is the entry point to schedule() from in-kernel preemption
I
Ingo Molnar 已提交
3568
 * off of preempt_enable. Kernel preemptions off return from interrupt
L
Linus Torvalds 已提交
3569 3570
 * occur there and call schedule directly.
 */
3571
asmlinkage __visible void __sched notrace preempt_schedule(void)
L
Linus Torvalds 已提交
3572 3573 3574
{
	/*
	 * If there is a non-zero preempt_count or interrupts are disabled,
I
Ingo Molnar 已提交
3575
	 * we do not want to preempt the current task. Just return..
L
Linus Torvalds 已提交
3576
	 */
3577
	if (likely(!preemptible()))
L
Linus Torvalds 已提交
3578 3579
		return;

3580
	preempt_schedule_common();
L
Linus Torvalds 已提交
3581
}
3582
NOKPROBE_SYMBOL(preempt_schedule);
L
Linus Torvalds 已提交
3583
EXPORT_SYMBOL(preempt_schedule);
3584 3585

/**
3586
 * preempt_schedule_notrace - preempt_schedule called by tracing
3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598
 *
 * 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.
 */
3599
asmlinkage __visible void __sched notrace preempt_schedule_notrace(void)
3600 3601 3602 3603 3604 3605 3606
{
	enum ctx_state prev_ctx;

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

	do {
3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619
		/*
		 * 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.
		 */
3620
		preempt_disable_notrace();
3621
		preempt_latency_start(1);
3622 3623 3624 3625 3626 3627
		/*
		 * Needs preempt disabled in case user_exit() is traced
		 * and the tracer calls preempt_enable_notrace() causing
		 * an infinite recursion.
		 */
		prev_ctx = exception_enter();
3628
		__schedule(true);
3629 3630
		exception_exit(prev_ctx);

3631
		preempt_latency_stop(1);
3632
		preempt_enable_no_resched_notrace();
3633 3634
	} while (need_resched());
}
3635
EXPORT_SYMBOL_GPL(preempt_schedule_notrace);
3636

3637
#endif /* CONFIG_PREEMPT */
L
Linus Torvalds 已提交
3638 3639

/*
3640
 * this is the entry point to schedule() from kernel preemption
L
Linus Torvalds 已提交
3641 3642 3643 3644
 * off of irq context.
 * Note, that this is called and return with irqs disabled. This will
 * protect us against recursive calling from irq.
 */
3645
asmlinkage __visible void __sched preempt_schedule_irq(void)
L
Linus Torvalds 已提交
3646
{
3647
	enum ctx_state prev_state;
3648

3649
	/* Catch callers which need to be fixed */
3650
	BUG_ON(preempt_count() || !irqs_disabled());
L
Linus Torvalds 已提交
3651

3652 3653
	prev_state = exception_enter();

3654
	do {
3655
		preempt_disable();
3656
		local_irq_enable();
3657
		__schedule(true);
3658
		local_irq_disable();
3659
		sched_preempt_enable_no_resched();
3660
	} while (need_resched());
3661 3662

	exception_exit(prev_state);
L
Linus Torvalds 已提交
3663 3664
}

P
Peter Zijlstra 已提交
3665
int default_wake_function(wait_queue_t *curr, unsigned mode, int wake_flags,
I
Ingo Molnar 已提交
3666
			  void *key)
L
Linus Torvalds 已提交
3667
{
P
Peter Zijlstra 已提交
3668
	return try_to_wake_up(curr->private, mode, wake_flags);
L
Linus Torvalds 已提交
3669 3670 3671
}
EXPORT_SYMBOL(default_wake_function);

3672 3673 3674 3675 3676 3677 3678 3679 3680 3681
#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().
 *
3682 3683
 * Used by the rt_mutex code to implement priority inheritance
 * logic. Call site only calls if the priority of the task changed.
3684
 */
3685
void rt_mutex_setprio(struct task_struct *p, int prio)
3686
{
3687
	int oldprio, queued, running, queue_flag = DEQUEUE_SAVE | DEQUEUE_MOVE;
3688
	const struct sched_class *prev_class;
3689 3690
	struct rq_flags rf;
	struct rq *rq;
3691

3692
	BUG_ON(prio > MAX_PRIO);
3693

3694
	rq = __task_rq_lock(p, &rf);
3695
	update_rq_clock(rq);
3696

3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714
	/*
	 * 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;
	}

3715
	trace_sched_pi_setprio(p, prio);
3716
	oldprio = p->prio;
3717 3718 3719 3720

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

3721
	prev_class = p->sched_class;
3722
	queued = task_on_rq_queued(p);
3723
	running = task_current(rq, p);
3724
	if (queued)
3725
		dequeue_task(rq, p, queue_flag);
3726
	if (running)
3727
		put_prev_task(rq, p);
I
Ingo Molnar 已提交
3728

3729 3730 3731 3732 3733 3734 3735 3736 3737 3738
	/*
	 * 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)) {
3739 3740 3741
		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))) {
3742
			p->dl.dl_boosted = 1;
3743
			queue_flag |= ENQUEUE_REPLENISH;
3744 3745
		} else
			p->dl.dl_boosted = 0;
3746
		p->sched_class = &dl_sched_class;
3747 3748 3749 3750
	} else if (rt_prio(prio)) {
		if (dl_prio(oldprio))
			p->dl.dl_boosted = 0;
		if (oldprio < prio)
3751
			queue_flag |= ENQUEUE_HEAD;
I
Ingo Molnar 已提交
3752
		p->sched_class = &rt_sched_class;
3753 3754 3755
	} else {
		if (dl_prio(oldprio))
			p->dl.dl_boosted = 0;
3756 3757
		if (rt_prio(oldprio))
			p->rt.timeout = 0;
I
Ingo Molnar 已提交
3758
		p->sched_class = &fair_sched_class;
3759
	}
I
Ingo Molnar 已提交
3760

3761 3762
	p->prio = prio;

3763
	if (queued)
3764
		enqueue_task(rq, p, queue_flag);
3765
	if (running)
3766
		set_curr_task(rq, p);
3767

P
Peter Zijlstra 已提交
3768
	check_class_changed(rq, p, prev_class, oldprio);
3769
out_unlock:
I
Ingo Molnar 已提交
3770 3771
	/* Avoid rq from going away on us: */
	preempt_disable();
3772
	__task_rq_unlock(rq, &rf);
3773 3774 3775

	balance_callback(rq);
	preempt_enable();
3776 3777
}
#endif
3778

3779
void set_user_nice(struct task_struct *p, long nice)
L
Linus Torvalds 已提交
3780
{
P
Peter Zijlstra 已提交
3781 3782
	bool queued, running;
	int old_prio, delta;
3783
	struct rq_flags rf;
3784
	struct rq *rq;
L
Linus Torvalds 已提交
3785

3786
	if (task_nice(p) == nice || nice < MIN_NICE || nice > MAX_NICE)
L
Linus Torvalds 已提交
3787 3788 3789 3790 3791
		return;
	/*
	 * We have to be careful, if called from sys_setpriority(),
	 * the task might be in the middle of scheduling on another CPU.
	 */
3792
	rq = task_rq_lock(p, &rf);
3793 3794
	update_rq_clock(rq);

L
Linus Torvalds 已提交
3795 3796 3797 3798
	/*
	 * 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
3799
	 * SCHED_DEADLINE, SCHED_FIFO or SCHED_RR:
L
Linus Torvalds 已提交
3800
	 */
3801
	if (task_has_dl_policy(p) || task_has_rt_policy(p)) {
L
Linus Torvalds 已提交
3802 3803 3804
		p->static_prio = NICE_TO_PRIO(nice);
		goto out_unlock;
	}
3805
	queued = task_on_rq_queued(p);
P
Peter Zijlstra 已提交
3806
	running = task_current(rq, p);
3807
	if (queued)
3808
		dequeue_task(rq, p, DEQUEUE_SAVE);
P
Peter Zijlstra 已提交
3809 3810
	if (running)
		put_prev_task(rq, p);
L
Linus Torvalds 已提交
3811 3812

	p->static_prio = NICE_TO_PRIO(nice);
3813
	set_load_weight(p);
3814 3815 3816
	old_prio = p->prio;
	p->prio = effective_prio(p);
	delta = p->prio - old_prio;
L
Linus Torvalds 已提交
3817

3818
	if (queued) {
3819
		enqueue_task(rq, p, ENQUEUE_RESTORE);
L
Linus Torvalds 已提交
3820
		/*
3821 3822
		 * If the task increased its priority or is running and
		 * lowered its priority, then reschedule its CPU:
L
Linus Torvalds 已提交
3823
		 */
3824
		if (delta < 0 || (delta > 0 && task_running(rq, p)))
3825
			resched_curr(rq);
L
Linus Torvalds 已提交
3826
	}
P
Peter Zijlstra 已提交
3827 3828
	if (running)
		set_curr_task(rq, p);
L
Linus Torvalds 已提交
3829
out_unlock:
3830
	task_rq_unlock(rq, p, &rf);
L
Linus Torvalds 已提交
3831 3832 3833
}
EXPORT_SYMBOL(set_user_nice);

M
Matt Mackall 已提交
3834 3835 3836 3837 3838
/*
 * can_nice - check if a task can reduce its nice value
 * @p: task
 * @nice: nice value
 */
3839
int can_nice(const struct task_struct *p, const int nice)
M
Matt Mackall 已提交
3840
{
I
Ingo Molnar 已提交
3841
	/* Convert nice value [19,-20] to rlimit style value [1,40]: */
3842
	int nice_rlim = nice_to_rlimit(nice);
3843

3844
	return (nice_rlim <= task_rlimit(p, RLIMIT_NICE) ||
M
Matt Mackall 已提交
3845 3846 3847
		capable(CAP_SYS_NICE));
}

L
Linus Torvalds 已提交
3848 3849 3850 3851 3852 3853 3854 3855 3856
#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.
 */
3857
SYSCALL_DEFINE1(nice, int, increment)
L
Linus Torvalds 已提交
3858
{
3859
	long nice, retval;
L
Linus Torvalds 已提交
3860 3861 3862 3863 3864 3865

	/*
	 * 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.
	 */
3866
	increment = clamp(increment, -NICE_WIDTH, NICE_WIDTH);
3867
	nice = task_nice(current) + increment;
L
Linus Torvalds 已提交
3868

3869
	nice = clamp_val(nice, MIN_NICE, MAX_NICE);
M
Matt Mackall 已提交
3870 3871 3872
	if (increment < 0 && !can_nice(current, nice))
		return -EPERM;

L
Linus Torvalds 已提交
3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886
	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.
 *
3887
 * Return: The priority value as seen by users in /proc.
L
Linus Torvalds 已提交
3888 3889 3890
 * RT tasks are offset by -200. Normal tasks are centered
 * around 0, value goes from -16 to +15.
 */
3891
int task_prio(const struct task_struct *p)
L
Linus Torvalds 已提交
3892 3893 3894 3895 3896
{
	return p->prio - MAX_RT_PRIO;
}

/**
I
Ingo Molnar 已提交
3897
 * idle_cpu - is a given CPU idle currently?
L
Linus Torvalds 已提交
3898
 * @cpu: the processor in question.
3899 3900
 *
 * Return: 1 if the CPU is currently idle. 0 otherwise.
L
Linus Torvalds 已提交
3901 3902 3903
 */
int idle_cpu(int cpu)
{
T
Thomas Gleixner 已提交
3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917
	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
3918 3919 3920
}

/**
I
Ingo Molnar 已提交
3921
 * idle_task - return the idle task for a given CPU.
L
Linus Torvalds 已提交
3922
 * @cpu: the processor in question.
3923
 *
I
Ingo Molnar 已提交
3924
 * Return: The idle task for the CPU @cpu.
L
Linus Torvalds 已提交
3925
 */
3926
struct task_struct *idle_task(int cpu)
L
Linus Torvalds 已提交
3927 3928 3929 3930 3931 3932 3933
{
	return cpu_rq(cpu)->idle;
}

/**
 * find_process_by_pid - find a process with a matching PID value.
 * @pid: the pid in question.
3934 3935
 *
 * The task of @pid, if found. %NULL otherwise.
L
Linus Torvalds 已提交
3936
 */
A
Alexey Dobriyan 已提交
3937
static struct task_struct *find_process_by_pid(pid_t pid)
L
Linus Torvalds 已提交
3938
{
3939
	return pid ? find_task_by_vpid(pid) : current;
L
Linus Torvalds 已提交
3940 3941
}

3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956
/*
 * 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;
3957
	dl_se->dl_period = attr->sched_period ?: dl_se->dl_deadline;
3958
	dl_se->flags = attr->sched_flags;
3959
	dl_se->dl_bw = to_ratio(dl_se->dl_period, dl_se->dl_runtime);
3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979

	/*
	 * 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.
	 */
3980 3981
}

3982 3983 3984 3985 3986 3987
/*
 * sched_setparam() passes in -1 for its policy, to let the functions
 * it calls know not to change it.
 */
#define SETPARAM_POLICY	-1

3988 3989
static void __setscheduler_params(struct task_struct *p,
		const struct sched_attr *attr)
L
Linus Torvalds 已提交
3990
{
3991 3992
	int policy = attr->sched_policy;

3993
	if (policy == SETPARAM_POLICY)
3994 3995
		policy = p->policy;

L
Linus Torvalds 已提交
3996
	p->policy = policy;
3997

3998 3999
	if (dl_policy(policy))
		__setparam_dl(p, attr);
4000
	else if (fair_policy(policy))
4001 4002
		p->static_prio = NICE_TO_PRIO(attr->sched_nice);

4003 4004 4005 4006 4007 4008
	/*
	 * __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;
4009
	p->normal_prio = normal_prio(p);
4010 4011
	set_load_weight(p);
}
4012

4013 4014
/* Actually do priority change: must hold pi & rq lock. */
static void __setscheduler(struct rq *rq, struct task_struct *p,
4015
			   const struct sched_attr *attr, bool keep_boost)
4016 4017
{
	__setscheduler_params(p, attr);
4018

4019
	/*
4020 4021
	 * Keep a potential priority boosting if called from
	 * sched_setscheduler().
4022
	 */
4023 4024 4025 4026
	if (keep_boost)
		p->prio = rt_mutex_get_effective_prio(p, normal_prio(p));
	else
		p->prio = normal_prio(p);
4027

4028 4029 4030
	if (dl_prio(p->prio))
		p->sched_class = &dl_sched_class;
	else if (rt_prio(p->prio))
4031 4032 4033
		p->sched_class = &rt_sched_class;
	else
		p->sched_class = &fair_sched_class;
L
Linus Torvalds 已提交
4034
}
4035 4036 4037 4038 4039 4040 4041 4042 4043

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;
4044
	attr->sched_period = dl_se->dl_period;
4045 4046 4047 4048 4049 4050
	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
4051
 * than the runtime, as well as the period of being zero or
4052
 * greater than deadline. Furthermore, we have to be sure that
4053 4054 4055 4056
 * 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).
4057 4058 4059 4060
 */
static bool
__checkparam_dl(const struct sched_attr *attr)
{
4061 4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086
	/* 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;
4087 4088
}

4089
/*
I
Ingo Molnar 已提交
4090
 * Check the target process has a UID that matches the current process's:
4091 4092 4093 4094 4095 4096 4097 4098
 */
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);
4099 4100
	match = (uid_eq(cred->euid, pcred->euid) ||
		 uid_eq(cred->euid, pcred->uid));
4101 4102 4103 4104
	rcu_read_unlock();
	return match;
}

I
Ingo Molnar 已提交
4105
static bool dl_param_changed(struct task_struct *p, const struct sched_attr *attr)
4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117
{
	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;
}

4118 4119
static int __sched_setscheduler(struct task_struct *p,
				const struct sched_attr *attr,
4120
				bool user, bool pi)
L
Linus Torvalds 已提交
4121
{
4122 4123
	int newprio = dl_policy(attr->sched_policy) ? MAX_DL_PRIO - 1 :
		      MAX_RT_PRIO - 1 - attr->sched_priority;
4124
	int retval, oldprio, oldpolicy = -1, queued, running;
4125
	int new_effective_prio, policy = attr->sched_policy;
4126
	const struct sched_class *prev_class;
4127
	struct rq_flags rf;
4128
	int reset_on_fork;
4129
	int queue_flags = DEQUEUE_SAVE | DEQUEUE_MOVE;
4130
	struct rq *rq;
L
Linus Torvalds 已提交
4131

I
Ingo Molnar 已提交
4132
	/* May grab non-irq protected spin_locks: */
4133
	BUG_ON(in_interrupt());
L
Linus Torvalds 已提交
4134
recheck:
I
Ingo Molnar 已提交
4135
	/* Double check policy once rq lock held: */
4136 4137
	if (policy < 0) {
		reset_on_fork = p->sched_reset_on_fork;
L
Linus Torvalds 已提交
4138
		policy = oldpolicy = p->policy;
4139
	} else {
4140
		reset_on_fork = !!(attr->sched_flags & SCHED_FLAG_RESET_ON_FORK);
4141

4142
		if (!valid_policy(policy))
4143 4144 4145
			return -EINVAL;
	}

4146 4147 4148
	if (attr->sched_flags & ~(SCHED_FLAG_RESET_ON_FORK))
		return -EINVAL;

L
Linus Torvalds 已提交
4149 4150
	/*
	 * Valid priorities for SCHED_FIFO and SCHED_RR are
I
Ingo Molnar 已提交
4151 4152
	 * 1..MAX_USER_RT_PRIO-1, valid priority for SCHED_NORMAL,
	 * SCHED_BATCH and SCHED_IDLE is 0.
L
Linus Torvalds 已提交
4153
	 */
4154
	if ((p->mm && attr->sched_priority > MAX_USER_RT_PRIO-1) ||
4155
	    (!p->mm && attr->sched_priority > MAX_RT_PRIO-1))
L
Linus Torvalds 已提交
4156
		return -EINVAL;
4157 4158
	if ((dl_policy(policy) && !__checkparam_dl(attr)) ||
	    (rt_policy(policy) != (attr->sched_priority != 0)))
L
Linus Torvalds 已提交
4159 4160
		return -EINVAL;

4161 4162 4163
	/*
	 * Allow unprivileged RT tasks to decrease priority:
	 */
4164
	if (user && !capable(CAP_SYS_NICE)) {
4165
		if (fair_policy(policy)) {
4166
			if (attr->sched_nice < task_nice(p) &&
4167
			    !can_nice(p, attr->sched_nice))
4168 4169 4170
				return -EPERM;
		}

4171
		if (rt_policy(policy)) {
4172 4173
			unsigned long rlim_rtprio =
					task_rlimit(p, RLIMIT_RTPRIO);
4174

I
Ingo Molnar 已提交
4175
			/* Can't set/change the rt policy: */
4176 4177 4178
			if (policy != p->policy && !rlim_rtprio)
				return -EPERM;

I
Ingo Molnar 已提交
4179
			/* Can't increase priority: */
4180 4181
			if (attr->sched_priority > p->rt_priority &&
			    attr->sched_priority > rlim_rtprio)
4182 4183
				return -EPERM;
		}
4184

4185 4186 4187 4188 4189 4190 4191 4192 4193
		 /*
		  * 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
4194
		/*
4195 4196
		 * 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
4197
		 */
4198
		if (idle_policy(p->policy) && !idle_policy(policy)) {
4199
			if (!can_nice(p, task_nice(p)))
4200 4201
				return -EPERM;
		}
4202

I
Ingo Molnar 已提交
4203
		/* Can't change other user's priorities: */
4204
		if (!check_same_owner(p))
4205
			return -EPERM;
4206

I
Ingo Molnar 已提交
4207
		/* Normal users shall not reset the sched_reset_on_fork flag: */
4208 4209
		if (p->sched_reset_on_fork && !reset_on_fork)
			return -EPERM;
4210
	}
L
Linus Torvalds 已提交
4211

4212
	if (user) {
4213
		retval = security_task_setscheduler(p);
4214 4215 4216 4217
		if (retval)
			return retval;
	}

4218
	/*
I
Ingo Molnar 已提交
4219
	 * Make sure no PI-waiters arrive (or leave) while we are
4220
	 * changing the priority of the task:
4221
	 *
L
Lucas De Marchi 已提交
4222
	 * To be able to change p->policy safely, the appropriate
L
Linus Torvalds 已提交
4223 4224
	 * runqueue lock must be held.
	 */
4225
	rq = task_rq_lock(p, &rf);
4226
	update_rq_clock(rq);
4227

4228
	/*
I
Ingo Molnar 已提交
4229
	 * Changing the policy of the stop threads its a very bad idea:
4230 4231
	 */
	if (p == rq->stop) {
4232
		task_rq_unlock(rq, p, &rf);
4233 4234 4235
		return -EINVAL;
	}

4236
	/*
4237 4238
	 * If not changing anything there's no need to proceed further,
	 * but store a possible modification of reset_on_fork.
4239
	 */
4240
	if (unlikely(policy == p->policy)) {
4241
		if (fair_policy(policy) && attr->sched_nice != task_nice(p))
4242 4243 4244
			goto change;
		if (rt_policy(policy) && attr->sched_priority != p->rt_priority)
			goto change;
4245
		if (dl_policy(policy) && dl_param_changed(p, attr))
4246
			goto change;
4247

4248
		p->sched_reset_on_fork = reset_on_fork;
4249
		task_rq_unlock(rq, p, &rf);
4250 4251
		return 0;
	}
4252
change:
4253

4254
	if (user) {
4255
#ifdef CONFIG_RT_GROUP_SCHED
4256 4257 4258 4259 4260
		/*
		 * Do not allow realtime tasks into groups that have no runtime
		 * assigned.
		 */
		if (rt_bandwidth_enabled() && rt_policy(policy) &&
4261 4262
				task_group(p)->rt_bandwidth.rt_runtime == 0 &&
				!task_group_is_autogroup(task_group(p))) {
4263
			task_rq_unlock(rq, p, &rf);
4264 4265 4266
			return -EPERM;
		}
#endif
4267 4268 4269 4270 4271 4272 4273 4274 4275
#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.
			 */
4276 4277
			if (!cpumask_subset(span, &p->cpus_allowed) ||
			    rq->rd->dl_bw.bw == 0) {
4278
				task_rq_unlock(rq, p, &rf);
4279 4280 4281 4282 4283
				return -EPERM;
			}
		}
#endif
	}
4284

I
Ingo Molnar 已提交
4285
	/* Re-check policy now with rq lock held: */
L
Linus Torvalds 已提交
4286 4287
	if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) {
		policy = oldpolicy = -1;
4288
		task_rq_unlock(rq, p, &rf);
L
Linus Torvalds 已提交
4289 4290
		goto recheck;
	}
4291 4292 4293 4294 4295 4296

	/*
	 * If setscheduling to SCHED_DEADLINE (or changing the parameters
	 * of a SCHED_DEADLINE task) we need to check if enough bandwidth
	 * is available.
	 */
4297
	if ((dl_policy(policy) || dl_task(p)) && dl_overflow(p, policy, attr)) {
4298
		task_rq_unlock(rq, p, &rf);
4299 4300 4301
		return -EBUSY;
	}

4302 4303 4304
	p->sched_reset_on_fork = reset_on_fork;
	oldprio = p->prio;

4305 4306 4307 4308 4309 4310 4311 4312 4313
	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);
4314 4315
		if (new_effective_prio == oldprio)
			queue_flags &= ~DEQUEUE_MOVE;
4316 4317
	}

4318
	queued = task_on_rq_queued(p);
4319
	running = task_current(rq, p);
4320
	if (queued)
4321
		dequeue_task(rq, p, queue_flags);
4322
	if (running)
4323
		put_prev_task(rq, p);
4324

4325
	prev_class = p->sched_class;
4326
	__setscheduler(rq, p, attr, pi);
4327

4328
	if (queued) {
4329 4330 4331 4332
		/*
		 * We enqueue to tail when the priority of a task is
		 * increased (user space view).
		 */
4333 4334
		if (oldprio < p->prio)
			queue_flags |= ENQUEUE_HEAD;
4335

4336
		enqueue_task(rq, p, queue_flags);
4337
	}
4338
	if (running)
4339
		set_curr_task(rq, p);
4340

P
Peter Zijlstra 已提交
4341
	check_class_changed(rq, p, prev_class, oldprio);
I
Ingo Molnar 已提交
4342 4343 4344

	/* Avoid rq from going away on us: */
	preempt_disable();
4345
	task_rq_unlock(rq, p, &rf);
4346

4347 4348
	if (pi)
		rt_mutex_adjust_pi(p);
4349

I
Ingo Molnar 已提交
4350
	/* Run balance callbacks after we've adjusted the PI chain: */
4351 4352
	balance_callback(rq);
	preempt_enable();
4353

L
Linus Torvalds 已提交
4354 4355
	return 0;
}
4356

4357 4358 4359 4360 4361 4362 4363 4364 4365
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),
	};

4366 4367
	/* Fixup the legacy SCHED_RESET_ON_FORK hack. */
	if ((policy != SETPARAM_POLICY) && (policy & SCHED_RESET_ON_FORK)) {
4368 4369 4370 4371 4372
		attr.sched_flags |= SCHED_FLAG_RESET_ON_FORK;
		policy &= ~SCHED_RESET_ON_FORK;
		attr.sched_policy = policy;
	}

4373
	return __sched_setscheduler(p, &attr, check, true);
4374
}
4375 4376 4377 4378 4379 4380
/**
 * 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.
 *
4381 4382
 * Return: 0 on success. An error code otherwise.
 *
4383 4384 4385
 * NOTE that the task may be already dead.
 */
int sched_setscheduler(struct task_struct *p, int policy,
4386
		       const struct sched_param *param)
4387
{
4388
	return _sched_setscheduler(p, policy, param, true);
4389
}
L
Linus Torvalds 已提交
4390 4391
EXPORT_SYMBOL_GPL(sched_setscheduler);

4392 4393
int sched_setattr(struct task_struct *p, const struct sched_attr *attr)
{
4394
	return __sched_setscheduler(p, attr, true, true);
4395 4396 4397
}
EXPORT_SYMBOL_GPL(sched_setattr);

4398 4399 4400 4401 4402 4403 4404 4405 4406 4407
/**
 * 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.
4408 4409
 *
 * Return: 0 on success. An error code otherwise.
4410 4411
 */
int sched_setscheduler_nocheck(struct task_struct *p, int policy,
4412
			       const struct sched_param *param)
4413
{
4414
	return _sched_setscheduler(p, policy, param, false);
4415
}
4416
EXPORT_SYMBOL_GPL(sched_setscheduler_nocheck);
4417

I
Ingo Molnar 已提交
4418 4419
static int
do_sched_setscheduler(pid_t pid, int policy, struct sched_param __user *param)
L
Linus Torvalds 已提交
4420 4421 4422
{
	struct sched_param lparam;
	struct task_struct *p;
4423
	int retval;
L
Linus Torvalds 已提交
4424 4425 4426 4427 4428

	if (!param || pid < 0)
		return -EINVAL;
	if (copy_from_user(&lparam, param, sizeof(struct sched_param)))
		return -EFAULT;
4429 4430 4431

	rcu_read_lock();
	retval = -ESRCH;
L
Linus Torvalds 已提交
4432
	p = find_process_by_pid(pid);
4433 4434 4435
	if (p != NULL)
		retval = sched_setscheduler(p, policy, &lparam);
	rcu_read_unlock();
4436

L
Linus Torvalds 已提交
4437 4438 4439
	return retval;
}

4440 4441 4442
/*
 * Mimics kernel/events/core.c perf_copy_attr().
 */
I
Ingo Molnar 已提交
4443
static int sched_copy_attr(struct sched_attr __user *uattr, struct sched_attr *attr)
4444 4445 4446 4447 4448 4449 4450
{
	u32 size;
	int ret;

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

I
Ingo Molnar 已提交
4451
	/* Zero the full structure, so that a short copy will be nice: */
4452 4453 4454 4455 4456 4457
	memset(attr, 0, sizeof(*attr));

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

I
Ingo Molnar 已提交
4458 4459
	/* Bail out on silly large: */
	if (size > PAGE_SIZE)
4460 4461
		goto err_size;

I
Ingo Molnar 已提交
4462 4463
	/* ABI compatibility quirk: */
	if (!size)
4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497
		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
4498
	 * XXX: Do we want to be lenient like existing syscalls; or do we want
4499 4500
	 * to be strict and return an error on out-of-bounds values?
	 */
4501
	attr->sched_nice = clamp(attr->sched_nice, MIN_NICE, MAX_NICE);
4502

4503
	return 0;
4504 4505 4506

err_size:
	put_user(sizeof(*attr), &uattr->size);
4507
	return -E2BIG;
4508 4509
}

L
Linus Torvalds 已提交
4510 4511 4512 4513 4514
/**
 * 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.
4515 4516
 *
 * Return: 0 on success. An error code otherwise.
L
Linus Torvalds 已提交
4517
 */
I
Ingo Molnar 已提交
4518
SYSCALL_DEFINE3(sched_setscheduler, pid_t, pid, int, policy, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4519
{
4520 4521 4522
	if (policy < 0)
		return -EINVAL;

L
Linus Torvalds 已提交
4523 4524 4525 4526 4527 4528 4529
	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.
4530 4531
 *
 * Return: 0 on success. An error code otherwise.
L
Linus Torvalds 已提交
4532
 */
4533
SYSCALL_DEFINE2(sched_setparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4534
{
4535
	return do_sched_setscheduler(pid, SETPARAM_POLICY, param);
L
Linus Torvalds 已提交
4536 4537
}

4538 4539 4540
/**
 * sys_sched_setattr - same as above, but with extended sched_attr
 * @pid: the pid in question.
J
Juri Lelli 已提交
4541
 * @uattr: structure containing the extended parameters.
4542
 * @flags: for future extension.
4543
 */
4544 4545
SYSCALL_DEFINE3(sched_setattr, pid_t, pid, struct sched_attr __user *, uattr,
			       unsigned int, flags)
4546 4547 4548 4549 4550
{
	struct sched_attr attr;
	struct task_struct *p;
	int retval;

4551
	if (!uattr || pid < 0 || flags)
4552 4553
		return -EINVAL;

4554 4555 4556
	retval = sched_copy_attr(uattr, &attr);
	if (retval)
		return retval;
4557

4558
	if ((int)attr.sched_policy < 0)
4559
		return -EINVAL;
4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570

	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
4571 4572 4573
/**
 * sys_sched_getscheduler - get the policy (scheduling class) of a thread
 * @pid: the pid in question.
4574 4575 4576
 *
 * Return: On success, the policy of the thread. Otherwise, a negative error
 * code.
L
Linus Torvalds 已提交
4577
 */
4578
SYSCALL_DEFINE1(sched_getscheduler, pid_t, pid)
L
Linus Torvalds 已提交
4579
{
4580
	struct task_struct *p;
4581
	int retval;
L
Linus Torvalds 已提交
4582 4583

	if (pid < 0)
4584
		return -EINVAL;
L
Linus Torvalds 已提交
4585 4586

	retval = -ESRCH;
4587
	rcu_read_lock();
L
Linus Torvalds 已提交
4588 4589 4590 4591
	p = find_process_by_pid(pid);
	if (p) {
		retval = security_task_getscheduler(p);
		if (!retval)
4592 4593
			retval = p->policy
				| (p->sched_reset_on_fork ? SCHED_RESET_ON_FORK : 0);
L
Linus Torvalds 已提交
4594
	}
4595
	rcu_read_unlock();
L
Linus Torvalds 已提交
4596 4597 4598 4599
	return retval;
}

/**
4600
 * sys_sched_getparam - get the RT priority of a thread
L
Linus Torvalds 已提交
4601 4602
 * @pid: the pid in question.
 * @param: structure containing the RT priority.
4603 4604 4605
 *
 * Return: On success, 0 and the RT priority is in @param. Otherwise, an error
 * code.
L
Linus Torvalds 已提交
4606
 */
4607
SYSCALL_DEFINE2(sched_getparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4608
{
4609
	struct sched_param lp = { .sched_priority = 0 };
4610
	struct task_struct *p;
4611
	int retval;
L
Linus Torvalds 已提交
4612 4613

	if (!param || pid < 0)
4614
		return -EINVAL;
L
Linus Torvalds 已提交
4615

4616
	rcu_read_lock();
L
Linus Torvalds 已提交
4617 4618 4619 4620 4621 4622 4623 4624 4625
	p = find_process_by_pid(pid);
	retval = -ESRCH;
	if (!p)
		goto out_unlock;

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

4626 4627
	if (task_has_rt_policy(p))
		lp.sched_priority = p->rt_priority;
4628
	rcu_read_unlock();
L
Linus Torvalds 已提交
4629 4630 4631 4632 4633 4634 4635 4636 4637

	/*
	 * 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:
4638
	rcu_read_unlock();
L
Linus Torvalds 已提交
4639 4640 4641
	return retval;
}

4642 4643 4644 4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664
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)
4665
				return -EFBIG;
4666 4667 4668 4669 4670
		}

		attr->size = usize;
	}

4671
	ret = copy_to_user(uattr, attr, attr->size);
4672 4673 4674
	if (ret)
		return -EFAULT;

4675
	return 0;
4676 4677 4678
}

/**
4679
 * sys_sched_getattr - similar to sched_getparam, but with sched_attr
4680
 * @pid: the pid in question.
J
Juri Lelli 已提交
4681
 * @uattr: structure containing the extended parameters.
4682
 * @size: sizeof(attr) for fwd/bwd comp.
4683
 * @flags: for future extension.
4684
 */
4685 4686
SYSCALL_DEFINE4(sched_getattr, pid_t, pid, struct sched_attr __user *, uattr,
		unsigned int, size, unsigned int, flags)
4687 4688 4689 4690 4691 4692 4693 4694
{
	struct sched_attr attr = {
		.size = sizeof(struct sched_attr),
	};
	struct task_struct *p;
	int retval;

	if (!uattr || pid < 0 || size > PAGE_SIZE ||
4695
	    size < SCHED_ATTR_SIZE_VER0 || flags)
4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708
		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;
4709 4710
	if (p->sched_reset_on_fork)
		attr.sched_flags |= SCHED_FLAG_RESET_ON_FORK;
4711 4712 4713
	if (task_has_dl_policy(p))
		__getparam_dl(p, &attr);
	else if (task_has_rt_policy(p))
4714 4715
		attr.sched_priority = p->rt_priority;
	else
4716
		attr.sched_nice = task_nice(p);
4717 4718 4719 4720 4721 4722 4723 4724 4725 4726 4727

	rcu_read_unlock();

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

out_unlock:
	rcu_read_unlock();
	return retval;
}

4728
long sched_setaffinity(pid_t pid, const struct cpumask *in_mask)
L
Linus Torvalds 已提交
4729
{
4730
	cpumask_var_t cpus_allowed, new_mask;
4731 4732
	struct task_struct *p;
	int retval;
L
Linus Torvalds 已提交
4733

4734
	rcu_read_lock();
L
Linus Torvalds 已提交
4735 4736 4737

	p = find_process_by_pid(pid);
	if (!p) {
4738
		rcu_read_unlock();
L
Linus Torvalds 已提交
4739 4740 4741
		return -ESRCH;
	}

4742
	/* Prevent p going away */
L
Linus Torvalds 已提交
4743
	get_task_struct(p);
4744
	rcu_read_unlock();
L
Linus Torvalds 已提交
4745

4746 4747 4748 4749
	if (p->flags & PF_NO_SETAFFINITY) {
		retval = -EINVAL;
		goto out_put_task;
	}
4750 4751 4752 4753 4754 4755 4756 4757
	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
4758
	retval = -EPERM;
E
Eric W. Biederman 已提交
4759 4760 4761 4762
	if (!check_same_owner(p)) {
		rcu_read_lock();
		if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
			rcu_read_unlock();
4763
			goto out_free_new_mask;
E
Eric W. Biederman 已提交
4764 4765 4766
		}
		rcu_read_unlock();
	}
L
Linus Torvalds 已提交
4767

4768
	retval = security_task_setscheduler(p);
4769
	if (retval)
4770
		goto out_free_new_mask;
4771

4772 4773 4774 4775

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

4776 4777 4778 4779 4780 4781 4782
	/*
	 * 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
4783 4784 4785
	if (task_has_dl_policy(p) && dl_bandwidth_enabled()) {
		rcu_read_lock();
		if (!cpumask_subset(task_rq(p)->rd->span, new_mask)) {
4786
			retval = -EBUSY;
4787
			rcu_read_unlock();
4788
			goto out_free_new_mask;
4789
		}
4790
		rcu_read_unlock();
4791 4792
	}
#endif
P
Peter Zijlstra 已提交
4793
again:
4794
	retval = __set_cpus_allowed_ptr(p, new_mask, true);
L
Linus Torvalds 已提交
4795

P
Paul Menage 已提交
4796
	if (!retval) {
4797 4798
		cpuset_cpus_allowed(p, cpus_allowed);
		if (!cpumask_subset(new_mask, cpus_allowed)) {
P
Paul Menage 已提交
4799 4800 4801 4802 4803
			/*
			 * We must have raced with a concurrent cpuset
			 * update. Just reset the cpus_allowed to the
			 * cpuset's cpus_allowed
			 */
4804
			cpumask_copy(new_mask, cpus_allowed);
P
Paul Menage 已提交
4805 4806 4807
			goto again;
		}
	}
4808
out_free_new_mask:
4809 4810 4811 4812
	free_cpumask_var(new_mask);
out_free_cpus_allowed:
	free_cpumask_var(cpus_allowed);
out_put_task:
L
Linus Torvalds 已提交
4813 4814 4815 4816 4817
	put_task_struct(p);
	return retval;
}

static int get_user_cpu_mask(unsigned long __user *user_mask_ptr, unsigned len,
4818
			     struct cpumask *new_mask)
L
Linus Torvalds 已提交
4819
{
4820 4821 4822 4823 4824
	if (len < cpumask_size())
		cpumask_clear(new_mask);
	else if (len > cpumask_size())
		len = cpumask_size();

L
Linus Torvalds 已提交
4825 4826 4827 4828
	return copy_from_user(new_mask, user_mask_ptr, len) ? -EFAULT : 0;
}

/**
I
Ingo Molnar 已提交
4829
 * sys_sched_setaffinity - set the CPU affinity of a process
L
Linus Torvalds 已提交
4830 4831
 * @pid: pid of the process
 * @len: length in bytes of the bitmask pointed to by user_mask_ptr
I
Ingo Molnar 已提交
4832
 * @user_mask_ptr: user-space pointer to the new CPU mask
4833 4834
 *
 * Return: 0 on success. An error code otherwise.
L
Linus Torvalds 已提交
4835
 */
4836 4837
SYSCALL_DEFINE3(sched_setaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
4838
{
4839
	cpumask_var_t new_mask;
L
Linus Torvalds 已提交
4840 4841
	int retval;

4842 4843
	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
4844

4845 4846 4847 4848 4849
	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
4850 4851
}

4852
long sched_getaffinity(pid_t pid, struct cpumask *mask)
L
Linus Torvalds 已提交
4853
{
4854
	struct task_struct *p;
4855
	unsigned long flags;
L
Linus Torvalds 已提交
4856 4857
	int retval;

4858
	rcu_read_lock();
L
Linus Torvalds 已提交
4859 4860 4861 4862 4863 4864

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

4865 4866 4867 4868
	retval = security_task_getscheduler(p);
	if (retval)
		goto out_unlock;

4869
	raw_spin_lock_irqsave(&p->pi_lock, flags);
4870
	cpumask_and(mask, &p->cpus_allowed, cpu_active_mask);
4871
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
4872 4873

out_unlock:
4874
	rcu_read_unlock();
L
Linus Torvalds 已提交
4875

4876
	return retval;
L
Linus Torvalds 已提交
4877 4878 4879
}

/**
I
Ingo Molnar 已提交
4880
 * sys_sched_getaffinity - get the CPU affinity of a process
L
Linus Torvalds 已提交
4881 4882
 * @pid: pid of the process
 * @len: length in bytes of the bitmask pointed to by user_mask_ptr
I
Ingo Molnar 已提交
4883
 * @user_mask_ptr: user-space pointer to hold the current CPU mask
4884
 *
4885 4886
 * Return: size of CPU mask copied to user_mask_ptr on success. An
 * error code otherwise.
L
Linus Torvalds 已提交
4887
 */
4888 4889
SYSCALL_DEFINE3(sched_getaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
4890 4891
{
	int ret;
4892
	cpumask_var_t mask;
L
Linus Torvalds 已提交
4893

A
Anton Blanchard 已提交
4894
	if ((len * BITS_PER_BYTE) < nr_cpu_ids)
4895 4896
		return -EINVAL;
	if (len & (sizeof(unsigned long)-1))
L
Linus Torvalds 已提交
4897 4898
		return -EINVAL;

4899 4900
	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
4901

4902 4903
	ret = sched_getaffinity(pid, mask);
	if (ret == 0) {
4904
		size_t retlen = min_t(size_t, len, cpumask_size());
4905 4906

		if (copy_to_user(user_mask_ptr, mask, retlen))
4907 4908
			ret = -EFAULT;
		else
4909
			ret = retlen;
4910 4911
	}
	free_cpumask_var(mask);
L
Linus Torvalds 已提交
4912

4913
	return ret;
L
Linus Torvalds 已提交
4914 4915 4916 4917 4918
}

/**
 * sys_sched_yield - yield the current processor to other threads.
 *
I
Ingo Molnar 已提交
4919 4920
 * This function yields the current CPU to other tasks. If there are no
 * other threads running on this CPU then this function will return.
4921 4922
 *
 * Return: 0.
L
Linus Torvalds 已提交
4923
 */
4924
SYSCALL_DEFINE0(sched_yield)
L
Linus Torvalds 已提交
4925
{
4926
	struct rq *rq = this_rq_lock();
L
Linus Torvalds 已提交
4927

4928
	schedstat_inc(rq->yld_count);
4929
	current->sched_class->yield_task(rq);
L
Linus Torvalds 已提交
4930 4931 4932 4933 4934 4935

	/*
	 * Since we are going to call schedule() anyway, there's
	 * no need to preempt or enable interrupts:
	 */
	__release(rq->lock);
4936
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
4937
	do_raw_spin_unlock(&rq->lock);
4938
	sched_preempt_enable_no_resched();
L
Linus Torvalds 已提交
4939 4940 4941 4942 4943 4944

	schedule();

	return 0;
}

4945
#ifndef CONFIG_PREEMPT
4946
int __sched _cond_resched(void)
L
Linus Torvalds 已提交
4947
{
4948
	if (should_resched(0)) {
4949
		preempt_schedule_common();
L
Linus Torvalds 已提交
4950 4951 4952 4953
		return 1;
	}
	return 0;
}
4954
EXPORT_SYMBOL(_cond_resched);
4955
#endif
L
Linus Torvalds 已提交
4956 4957

/*
4958
 * __cond_resched_lock() - if a reschedule is pending, drop the given lock,
L
Linus Torvalds 已提交
4959 4960
 * call schedule, and on return reacquire the lock.
 *
I
Ingo Molnar 已提交
4961
 * This works OK both with and without CONFIG_PREEMPT. We do strange low-level
L
Linus Torvalds 已提交
4962 4963 4964
 * operations here to prevent schedule() from being called twice (once via
 * spin_unlock(), once by hand).
 */
4965
int __cond_resched_lock(spinlock_t *lock)
L
Linus Torvalds 已提交
4966
{
4967
	int resched = should_resched(PREEMPT_LOCK_OFFSET);
J
Jan Kara 已提交
4968 4969
	int ret = 0;

4970 4971
	lockdep_assert_held(lock);

4972
	if (spin_needbreak(lock) || resched) {
L
Linus Torvalds 已提交
4973
		spin_unlock(lock);
P
Peter Zijlstra 已提交
4974
		if (resched)
4975
			preempt_schedule_common();
N
Nick Piggin 已提交
4976 4977
		else
			cpu_relax();
J
Jan Kara 已提交
4978
		ret = 1;
L
Linus Torvalds 已提交
4979 4980
		spin_lock(lock);
	}
J
Jan Kara 已提交
4981
	return ret;
L
Linus Torvalds 已提交
4982
}
4983
EXPORT_SYMBOL(__cond_resched_lock);
L
Linus Torvalds 已提交
4984

4985
int __sched __cond_resched_softirq(void)
L
Linus Torvalds 已提交
4986 4987 4988
{
	BUG_ON(!in_softirq());

4989
	if (should_resched(SOFTIRQ_DISABLE_OFFSET)) {
4990
		local_bh_enable();
4991
		preempt_schedule_common();
L
Linus Torvalds 已提交
4992 4993 4994 4995 4996
		local_bh_disable();
		return 1;
	}
	return 0;
}
4997
EXPORT_SYMBOL(__cond_resched_softirq);
L
Linus Torvalds 已提交
4998 4999 5000 5001

/**
 * yield - yield the current processor to other threads.
 *
P
Peter Zijlstra 已提交
5002 5003 5004 5005 5006 5007 5008 5009 5010
 * 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
5011
 *	yield();
P
Peter Zijlstra 已提交
5012 5013 5014 5015 5016 5017 5018 5019
 *
 * 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
5020 5021 5022 5023 5024 5025 5026 5027
 */
void __sched yield(void)
{
	set_current_state(TASK_RUNNING);
	sys_sched_yield();
}
EXPORT_SYMBOL(yield);

5028 5029 5030 5031
/**
 * 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
5032 5033
 * @p: target task
 * @preempt: whether task preemption is allowed or not
5034 5035 5036 5037
 *
 * 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.
 *
5038
 * Return:
5039 5040 5041
 *	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.
5042
 */
5043
int __sched yield_to(struct task_struct *p, bool preempt)
5044 5045 5046 5047
{
	struct task_struct *curr = current;
	struct rq *rq, *p_rq;
	unsigned long flags;
5048
	int yielded = 0;
5049 5050 5051 5052 5053 5054

	local_irq_save(flags);
	rq = this_rq();

again:
	p_rq = task_rq(p);
5055 5056 5057 5058 5059 5060 5061 5062 5063
	/*
	 * 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;
	}

5064
	double_rq_lock(rq, p_rq);
5065
	if (task_rq(p) != p_rq) {
5066 5067 5068 5069 5070
		double_rq_unlock(rq, p_rq);
		goto again;
	}

	if (!curr->sched_class->yield_to_task)
5071
		goto out_unlock;
5072 5073

	if (curr->sched_class != p->sched_class)
5074
		goto out_unlock;
5075 5076

	if (task_running(p_rq, p) || p->state)
5077
		goto out_unlock;
5078 5079

	yielded = curr->sched_class->yield_to_task(rq, p, preempt);
5080
	if (yielded) {
5081
		schedstat_inc(rq->yld_count);
5082 5083 5084 5085 5086
		/*
		 * Make p's CPU reschedule; pick_next_entity takes care of
		 * fairness.
		 */
		if (preempt && rq != p_rq)
5087
			resched_curr(p_rq);
5088
	}
5089

5090
out_unlock:
5091
	double_rq_unlock(rq, p_rq);
5092
out_irq:
5093 5094
	local_irq_restore(flags);

5095
	if (yielded > 0)
5096 5097 5098 5099 5100 5101
		schedule();

	return yielded;
}
EXPORT_SYMBOL_GPL(yield_to);

5102 5103 5104 5105 5106 5107 5108 5109 5110 5111 5112 5113 5114 5115 5116
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
5117
/*
I
Ingo Molnar 已提交
5118
 * This task is about to go to sleep on IO. Increment rq->nr_iowait so
L
Linus Torvalds 已提交
5119 5120 5121 5122
 * that process accounting knows that this is a task in IO wait state.
 */
long __sched io_schedule_timeout(long timeout)
{
5123
	int token;
L
Linus Torvalds 已提交
5124 5125
	long ret;

5126
	token = io_schedule_prepare();
L
Linus Torvalds 已提交
5127
	ret = schedule_timeout(timeout);
5128
	io_schedule_finish(token);
5129

L
Linus Torvalds 已提交
5130 5131
	return ret;
}
5132
EXPORT_SYMBOL(io_schedule_timeout);
L
Linus Torvalds 已提交
5133

5134 5135 5136 5137 5138 5139 5140 5141 5142 5143
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
5144 5145 5146 5147
/**
 * sys_sched_get_priority_max - return maximum RT priority.
 * @policy: scheduling class.
 *
5148 5149 5150
 * 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
5151
 */
5152
SYSCALL_DEFINE1(sched_get_priority_max, int, policy)
L
Linus Torvalds 已提交
5153 5154 5155 5156 5157 5158 5159 5160
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = MAX_USER_RT_PRIO-1;
		break;
5161
	case SCHED_DEADLINE:
L
Linus Torvalds 已提交
5162
	case SCHED_NORMAL:
5163
	case SCHED_BATCH:
I
Ingo Molnar 已提交
5164
	case SCHED_IDLE:
L
Linus Torvalds 已提交
5165 5166 5167 5168 5169 5170 5171 5172 5173 5174
		ret = 0;
		break;
	}
	return ret;
}

/**
 * sys_sched_get_priority_min - return minimum RT priority.
 * @policy: scheduling class.
 *
5175 5176 5177
 * 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
5178
 */
5179
SYSCALL_DEFINE1(sched_get_priority_min, int, policy)
L
Linus Torvalds 已提交
5180 5181 5182 5183 5184 5185 5186 5187
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = 1;
		break;
5188
	case SCHED_DEADLINE:
L
Linus Torvalds 已提交
5189
	case SCHED_NORMAL:
5190
	case SCHED_BATCH:
I
Ingo Molnar 已提交
5191
	case SCHED_IDLE:
L
Linus Torvalds 已提交
5192 5193 5194 5195 5196 5197 5198 5199 5200 5201 5202 5203
		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.
5204 5205 5206
 *
 * Return: On success, 0 and the timeslice is in @interval. Otherwise,
 * an error code.
L
Linus Torvalds 已提交
5207
 */
5208
SYSCALL_DEFINE2(sched_rr_get_interval, pid_t, pid,
5209
		struct timespec __user *, interval)
L
Linus Torvalds 已提交
5210
{
5211
	struct task_struct *p;
D
Dmitry Adamushko 已提交
5212
	unsigned int time_slice;
5213 5214
	struct rq_flags rf;
	struct timespec t;
5215
	struct rq *rq;
5216
	int retval;
L
Linus Torvalds 已提交
5217 5218

	if (pid < 0)
5219
		return -EINVAL;
L
Linus Torvalds 已提交
5220 5221

	retval = -ESRCH;
5222
	rcu_read_lock();
L
Linus Torvalds 已提交
5223 5224 5225 5226 5227 5228 5229 5230
	p = find_process_by_pid(pid);
	if (!p)
		goto out_unlock;

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

5231
	rq = task_rq_lock(p, &rf);
5232 5233 5234
	time_slice = 0;
	if (p->sched_class->get_rr_interval)
		time_slice = p->sched_class->get_rr_interval(rq, p);
5235
	task_rq_unlock(rq, p, &rf);
D
Dmitry Adamushko 已提交
5236

5237
	rcu_read_unlock();
D
Dmitry Adamushko 已提交
5238
	jiffies_to_timespec(time_slice, &t);
L
Linus Torvalds 已提交
5239 5240
	retval = copy_to_user(interval, &t, sizeof(t)) ? -EFAULT : 0;
	return retval;
5241

L
Linus Torvalds 已提交
5242
out_unlock:
5243
	rcu_read_unlock();
L
Linus Torvalds 已提交
5244 5245 5246
	return retval;
}

5247
static const char stat_nam[] = TASK_STATE_TO_CHAR_STR;
5248

5249
void sched_show_task(struct task_struct *p)
L
Linus Torvalds 已提交
5250 5251
{
	unsigned long free = 0;
5252
	int ppid;
5253
	unsigned long state = p->state;
L
Linus Torvalds 已提交
5254

5255 5256 5257
	/* 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);

5258 5259
	if (!try_get_task_stack(p))
		return;
5260 5261
	if (state)
		state = __ffs(state) + 1;
5262
	printk(KERN_INFO "%-15.15s %c", p->comm,
5263
		state < sizeof(stat_nam) - 1 ? stat_nam[state] : '?');
L
Linus Torvalds 已提交
5264
	if (state == TASK_RUNNING)
P
Peter Zijlstra 已提交
5265
		printk(KERN_CONT "  running task    ");
L
Linus Torvalds 已提交
5266
#ifdef CONFIG_DEBUG_STACK_USAGE
5267
	free = stack_not_used(p);
L
Linus Torvalds 已提交
5268
#endif
5269
	ppid = 0;
5270
	rcu_read_lock();
5271 5272
	if (pid_alive(p))
		ppid = task_pid_nr(rcu_dereference(p->real_parent));
5273
	rcu_read_unlock();
P
Peter Zijlstra 已提交
5274
	printk(KERN_CONT "%5lu %5d %6d 0x%08lx\n", free,
5275
		task_pid_nr(p), ppid,
5276
		(unsigned long)task_thread_info(p)->flags);
L
Linus Torvalds 已提交
5277

5278
	print_worker_info(KERN_INFO, p);
5279
	show_stack(p, NULL);
5280
	put_task_stack(p);
L
Linus Torvalds 已提交
5281 5282
}

I
Ingo Molnar 已提交
5283
void show_state_filter(unsigned long state_filter)
L
Linus Torvalds 已提交
5284
{
5285
	struct task_struct *g, *p;
L
Linus Torvalds 已提交
5286

5287
#if BITS_PER_LONG == 32
P
Peter Zijlstra 已提交
5288 5289
	printk(KERN_INFO
		"  task                PC stack   pid father\n");
L
Linus Torvalds 已提交
5290
#else
P
Peter Zijlstra 已提交
5291 5292
	printk(KERN_INFO
		"  task                        PC stack   pid father\n");
L
Linus Torvalds 已提交
5293
#endif
5294
	rcu_read_lock();
5295
	for_each_process_thread(g, p) {
L
Linus Torvalds 已提交
5296 5297
		/*
		 * reset the NMI-timeout, listing all files on a slow
L
Lucas De Marchi 已提交
5298
		 * console might take a lot of time:
5299 5300 5301
		 * 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
5302 5303
		 */
		touch_nmi_watchdog();
5304
		touch_all_softlockup_watchdogs();
I
Ingo Molnar 已提交
5305
		if (!state_filter || (p->state & state_filter))
5306
			sched_show_task(p);
5307
	}
L
Linus Torvalds 已提交
5308

I
Ingo Molnar 已提交
5309
#ifdef CONFIG_SCHED_DEBUG
5310 5311
	if (!state_filter)
		sysrq_sched_debug_show();
I
Ingo Molnar 已提交
5312
#endif
5313
	rcu_read_unlock();
I
Ingo Molnar 已提交
5314 5315 5316
	/*
	 * Only show locks if all tasks are dumped:
	 */
5317
	if (!state_filter)
I
Ingo Molnar 已提交
5318
		debug_show_all_locks();
L
Linus Torvalds 已提交
5319 5320
}

5321
void init_idle_bootup_task(struct task_struct *idle)
I
Ingo Molnar 已提交
5322
{
I
Ingo Molnar 已提交
5323
	idle->sched_class = &idle_sched_class;
I
Ingo Molnar 已提交
5324 5325
}

5326 5327 5328
/**
 * init_idle - set up an idle thread for a given CPU
 * @idle: task in question
I
Ingo Molnar 已提交
5329
 * @cpu: CPU the idle task belongs to
5330 5331 5332 5333
 *
 * NOTE: this function does not set the idle thread's NEED_RESCHED
 * flag, to make booting more robust.
 */
5334
void init_idle(struct task_struct *idle, int cpu)
L
Linus Torvalds 已提交
5335
{
5336
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
5337 5338
	unsigned long flags;

5339 5340
	raw_spin_lock_irqsave(&idle->pi_lock, flags);
	raw_spin_lock(&rq->lock);
5341

5342
	__sched_fork(0, idle);
5343
	idle->state = TASK_RUNNING;
I
Ingo Molnar 已提交
5344
	idle->se.exec_start = sched_clock();
5345
	idle->flags |= PF_IDLE;
I
Ingo Molnar 已提交
5346

5347 5348
	kasan_unpoison_task_stack(idle);

5349 5350 5351 5352 5353 5354 5355 5356 5357
#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
5358 5359
	/*
	 * We're having a chicken and egg problem, even though we are
I
Ingo Molnar 已提交
5360
	 * holding rq->lock, the CPU isn't yet set to this CPU so the
5361 5362 5363 5364 5365 5366 5367 5368
	 * 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
5369
	__set_task_cpu(idle, cpu);
5370
	rcu_read_unlock();
L
Linus Torvalds 已提交
5371 5372

	rq->curr = rq->idle = idle;
5373
	idle->on_rq = TASK_ON_RQ_QUEUED;
5374
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
5375
	idle->on_cpu = 1;
5376
#endif
5377 5378
	raw_spin_unlock(&rq->lock);
	raw_spin_unlock_irqrestore(&idle->pi_lock, flags);
L
Linus Torvalds 已提交
5379 5380

	/* Set the preempt count _outside_ the spinlocks! */
5381
	init_idle_preempt_count(idle, cpu);
5382

I
Ingo Molnar 已提交
5383 5384 5385 5386
	/*
	 * The idle tasks have their own, simple scheduling class:
	 */
	idle->sched_class = &idle_sched_class;
5387
	ftrace_graph_init_idle_task(idle, cpu);
5388
	vtime_init_idle(idle, cpu);
5389
#ifdef CONFIG_SMP
5390 5391
	sprintf(idle->comm, "%s/%d", INIT_TASK_COMM, cpu);
#endif
I
Ingo Molnar 已提交
5392 5393
}

5394 5395 5396 5397 5398 5399 5400
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;

5401 5402 5403
	if (!cpumask_weight(cur))
		return ret;

5404
	rcu_read_lock_sched();
5405 5406 5407 5408 5409 5410 5411 5412
	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);
5413
	rcu_read_unlock_sched();
5414 5415 5416 5417

	return ret;
}

5418 5419 5420 5421 5422 5423 5424
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
5425
	 * to a new cpuset; we don't want to change their CPU
5426 5427 5428 5429 5430 5431 5432 5433 5434 5435 5436 5437 5438 5439 5440 5441
	 * 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);
5442
		struct dl_bw *dl_b;
5443 5444 5445 5446
		bool overflow;
		int cpus;
		unsigned long flags;

5447 5448
		rcu_read_lock_sched();
		dl_b = dl_bw_of(dest_cpu);
5449 5450 5451 5452 5453 5454 5455 5456 5457 5458 5459 5460 5461 5462 5463
		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);
5464
		rcu_read_unlock_sched();
5465 5466 5467 5468 5469 5470 5471

	}
#endif
out:
	return ret;
}

L
Linus Torvalds 已提交
5472 5473
#ifdef CONFIG_SMP

5474
bool sched_smp_initialized __read_mostly;
5475

5476 5477 5478 5479 5480 5481 5482 5483 5484 5485
#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;

5486
	if (!cpumask_test_cpu(target_cpu, &p->cpus_allowed))
5487 5488 5489 5490
		return -EINVAL;

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

5491
	trace_sched_move_numa(p, curr_cpu, target_cpu);
5492 5493
	return stop_one_cpu(curr_cpu, migration_cpu_stop, &arg);
}
5494 5495 5496 5497 5498 5499 5500

/*
 * 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)
{
5501
	bool queued, running;
5502 5503
	struct rq_flags rf;
	struct rq *rq;
5504

5505
	rq = task_rq_lock(p, &rf);
5506
	queued = task_on_rq_queued(p);
5507 5508
	running = task_current(rq, p);

5509
	if (queued)
5510
		dequeue_task(rq, p, DEQUEUE_SAVE);
5511
	if (running)
5512
		put_prev_task(rq, p);
5513 5514 5515

	p->numa_preferred_nid = nid;

5516
	if (queued)
5517
		enqueue_task(rq, p, ENQUEUE_RESTORE);
5518
	if (running)
5519
		set_curr_task(rq, p);
5520
	task_rq_unlock(rq, p, &rf);
5521
}
P
Peter Zijlstra 已提交
5522
#endif /* CONFIG_NUMA_BALANCING */
5523

L
Linus Torvalds 已提交
5524
#ifdef CONFIG_HOTPLUG_CPU
5525
/*
I
Ingo Molnar 已提交
5526
 * Ensure that the idle task is using init_mm right before its CPU goes
5527
 * offline.
5528
 */
5529
void idle_task_exit(void)
L
Linus Torvalds 已提交
5530
{
5531
	struct mm_struct *mm = current->active_mm;
5532

5533
	BUG_ON(cpu_online(smp_processor_id()));
5534

5535
	if (mm != &init_mm) {
5536
		switch_mm_irqs_off(mm, &init_mm, current);
5537 5538
		finish_arch_post_lock_switch();
	}
5539
	mmdrop(mm);
L
Linus Torvalds 已提交
5540 5541 5542
}

/*
5543 5544
 * 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
5545 5546 5547
 * 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.
5548 5549
 *
 * Also see the comment "Global load-average calculations".
L
Linus Torvalds 已提交
5550
 */
5551
static void calc_load_migrate(struct rq *rq)
L
Linus Torvalds 已提交
5552
{
5553
	long delta = calc_load_fold_active(rq, 1);
5554 5555
	if (delta)
		atomic_long_add(delta, &calc_load_tasks);
L
Linus Torvalds 已提交
5556 5557
}

5558 5559 5560 5561 5562 5563 5564 5565 5566 5567 5568 5569 5570 5571 5572 5573
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,
};

5574
/*
5575 5576 5577 5578 5579 5580
 * 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
5581
 */
5582
static void migrate_tasks(struct rq *dead_rq)
L
Linus Torvalds 已提交
5583
{
5584
	struct rq *rq = dead_rq;
5585
	struct task_struct *next, *stop = rq->stop;
5586
	struct rq_flags rf;
5587
	int dest_cpu;
L
Linus Torvalds 已提交
5588 5589

	/*
5590 5591 5592 5593 5594 5595 5596
	 * 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
5597
	 */
5598
	rq->stop = NULL;
5599

5600 5601 5602 5603 5604
	/*
	 * put_prev_task() and pick_next_task() sched
	 * class method both need to have an up-to-date
	 * value of rq->clock[_task]
	 */
5605
	rq_pin_lock(rq, &rf);
5606
	update_rq_clock(rq);
5607
	rq_unpin_lock(rq, &rf);
5608

5609
	for (;;) {
5610 5611
		/*
		 * There's this thread running, bail when that's the only
I
Ingo Molnar 已提交
5612
		 * remaining thread:
5613 5614
		 */
		if (rq->nr_running == 1)
I
Ingo Molnar 已提交
5615
			break;
5616

5617
		/*
I
Ingo Molnar 已提交
5618
		 * pick_next_task() assumes pinned rq->lock:
5619
		 */
5620
		rq_repin_lock(rq, &rf);
5621
		next = pick_next_task(rq, &fake_task, &rf);
5622
		BUG_ON(!next);
D
Dmitry Adamushko 已提交
5623
		next->sched_class->put_prev_task(rq, next);
5624

W
Wanpeng Li 已提交
5625 5626 5627 5628 5629 5630 5631 5632 5633
		/*
		 * 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.
		 */
5634
		rq_unpin_lock(rq, &rf);
W
Wanpeng Li 已提交
5635 5636 5637 5638 5639 5640 5641 5642 5643 5644 5645 5646 5647 5648
		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;
		}

5649
		/* Find suitable destination for @next, with force if needed. */
5650
		dest_cpu = select_fallback_rq(dead_rq->cpu, next);
5651

5652 5653 5654 5655 5656 5657
		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
5658
		raw_spin_unlock(&next->pi_lock);
L
Linus Torvalds 已提交
5659
	}
5660

5661
	rq->stop = stop;
5662
}
L
Linus Torvalds 已提交
5663 5664
#endif /* CONFIG_HOTPLUG_CPU */

5665
void set_rq_online(struct rq *rq)
5666 5667 5668 5669
{
	if (!rq->online) {
		const struct sched_class *class;

5670
		cpumask_set_cpu(rq->cpu, rq->rd->online);
5671 5672 5673 5674 5675 5676 5677 5678 5679
		rq->online = 1;

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

5680
void set_rq_offline(struct rq *rq)
5681 5682 5683 5684 5685 5686 5687 5688 5689
{
	if (rq->online) {
		const struct sched_class *class;

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

5690
		cpumask_clear_cpu(rq->cpu, rq->rd->online);
5691 5692 5693 5694
		rq->online = 0;
	}
}

5695
static void set_cpu_rq_start_time(unsigned int cpu)
L
Linus Torvalds 已提交
5696
{
5697
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
5698

5699 5700 5701
	rq->age_stamp = sched_clock_cpu(cpu);
}

I
Ingo Molnar 已提交
5702 5703 5704 5705
/*
 * used to mark begin/end of suspend/resume:
 */
static int num_cpus_frozen;
5706

L
Linus Torvalds 已提交
5707
/*
5708 5709 5710
 * Update cpusets according to cpu_active mask.  If cpusets are
 * disabled, cpuset_update_active_cpus() becomes a simple wrapper
 * around partition_sched_domains().
5711 5712 5713
 *
 * 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
5714
 */
5715
static void cpuset_cpu_active(void)
5716
{
5717
	if (cpuhp_tasks_frozen) {
5718 5719 5720 5721 5722 5723 5724 5725 5726
		/*
		 * 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);
5727
			return;
5728 5729 5730 5731 5732 5733
		}
		/*
		 * This is the last CPU online operation. So fall through and
		 * restore the original sched domains by considering the
		 * cpuset configurations.
		 */
5734
	}
5735
	cpuset_update_active_cpus(true);
5736
}
5737

5738
static int cpuset_cpu_inactive(unsigned int cpu)
5739
{
5740 5741
	unsigned long flags;
	struct dl_bw *dl_b;
5742 5743
	bool overflow;
	int cpus;
5744

5745
	if (!cpuhp_tasks_frozen) {
5746 5747
		rcu_read_lock_sched();
		dl_b = dl_bw_of(cpu);
5748

5749 5750 5751 5752
		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);
5753

5754
		rcu_read_unlock_sched();
5755

5756
		if (overflow)
5757
			return -EBUSY;
5758
		cpuset_update_active_cpus(false);
5759
	} else {
5760 5761
		num_cpus_frozen++;
		partition_sched_domains(1, NULL, NULL);
5762
	}
5763
	return 0;
5764 5765
}

5766
int sched_cpu_activate(unsigned int cpu)
5767
{
5768 5769 5770
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

5771
	set_cpu_active(cpu, true);
5772

5773
	if (sched_smp_initialized) {
5774
		sched_domains_numa_masks_set(cpu);
5775
		cpuset_cpu_active();
5776
	}
5777 5778 5779 5780 5781

	/*
	 * 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
5782
	 *    after all CPUs have been brought up.
5783 5784 5785 5786 5787 5788 5789 5790 5791 5792 5793 5794 5795
	 *
	 * 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();

5796
	return 0;
5797 5798
}

5799
int sched_cpu_deactivate(unsigned int cpu)
5800 5801 5802
{
	int ret;

5803
	set_cpu_active(cpu, false);
5804 5805 5806 5807 5808 5809 5810 5811 5812 5813 5814 5815 5816 5817
	/*
	 * 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();
5818 5819 5820 5821 5822 5823 5824 5825

	if (!sched_smp_initialized)
		return 0;

	ret = cpuset_cpu_inactive(cpu);
	if (ret) {
		set_cpu_active(cpu, true);
		return ret;
5826
	}
5827 5828
	sched_domains_numa_masks_clear(cpu);
	return 0;
5829 5830
}

5831 5832 5833 5834 5835 5836 5837 5838
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();
}

5839 5840 5841
int sched_cpu_starting(unsigned int cpu)
{
	set_cpu_rq_start_time(cpu);
5842
	sched_rq_cpu_starting(cpu);
5843
	return 0;
5844 5845
}

5846 5847 5848 5849 5850 5851 5852 5853 5854 5855 5856 5857 5858 5859 5860 5861 5862 5863
#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();
5864
	nohz_balance_exit_idle(cpu);
5865
	hrtick_clear(rq);
5866 5867 5868 5869
	return 0;
}
#endif

P
Peter Zijlstra 已提交
5870 5871 5872 5873 5874 5875 5876 5877 5878 5879 5880 5881 5882 5883 5884 5885
#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
5886 5887
void __init sched_init_smp(void)
{
5888 5889 5890
	cpumask_var_t non_isolated_cpus;

	alloc_cpumask_var(&non_isolated_cpus, GFP_KERNEL);
5891
	alloc_cpumask_var(&fallback_doms, GFP_KERNEL);
5892

5893 5894
	sched_init_numa();

5895 5896
	/*
	 * There's no userspace yet to cause hotplug operations; hence all the
I
Ingo Molnar 已提交
5897
	 * CPU masks are stable and all blatant races in the below code cannot
5898 5899
	 * happen.
	 */
5900
	mutex_lock(&sched_domains_mutex);
5901
	init_sched_domains(cpu_active_mask);
5902 5903 5904
	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);
5905
	mutex_unlock(&sched_domains_mutex);
5906

5907
	/* Move init over to a non-isolated CPU */
5908
	if (set_cpus_allowed_ptr(current, non_isolated_cpus) < 0)
5909
		BUG();
I
Ingo Molnar 已提交
5910
	sched_init_granularity();
5911
	free_cpumask_var(non_isolated_cpus);
5912

5913
	init_sched_rt_class();
5914
	init_sched_dl_class();
P
Peter Zijlstra 已提交
5915 5916

	sched_init_smt();
5917
	sched_clock_init_late();
P
Peter Zijlstra 已提交
5918

5919
	sched_smp_initialized = true;
L
Linus Torvalds 已提交
5920
}
5921 5922 5923

static int __init migration_init(void)
{
5924
	sched_rq_cpu_starting(smp_processor_id());
5925
	return 0;
L
Linus Torvalds 已提交
5926
}
5927 5928
early_initcall(migration_init);

L
Linus Torvalds 已提交
5929 5930 5931
#else
void __init sched_init_smp(void)
{
I
Ingo Molnar 已提交
5932
	sched_init_granularity();
5933
	sched_clock_init_late();
L
Linus Torvalds 已提交
5934 5935 5936 5937 5938 5939 5940 5941 5942 5943
}
#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);
}

5944
#ifdef CONFIG_CGROUP_SCHED
5945 5946 5947 5948
/*
 * Default task group.
 * Every task in system belongs to this group at bootup.
 */
5949
struct task_group root_task_group;
5950
LIST_HEAD(task_groups);
5951 5952 5953

/* Cacheline aligned slab cache for task_group */
static struct kmem_cache *task_group_cache __read_mostly;
5954
#endif
P
Peter Zijlstra 已提交
5955

5956
DECLARE_PER_CPU(cpumask_var_t, load_balance_mask);
5957
DECLARE_PER_CPU(cpumask_var_t, select_idle_mask);
P
Peter Zijlstra 已提交
5958

5959 5960 5961 5962 5963 5964 5965 5966 5967 5968 5969 5970 5971
#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
5972 5973
void __init sched_init(void)
{
I
Ingo Molnar 已提交
5974
	int i, j;
5975 5976
	unsigned long alloc_size = 0, ptr;

5977 5978
	sched_clock_init();

5979 5980 5981
	for (i = 0; i < WAIT_TABLE_SIZE; i++)
		init_waitqueue_head(bit_wait_table + i);

5982 5983 5984 5985 5986 5987 5988
#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) {
5989
		ptr = (unsigned long)kzalloc(alloc_size, GFP_NOWAIT);
5990 5991

#ifdef CONFIG_FAIR_GROUP_SCHED
5992
		root_task_group.se = (struct sched_entity **)ptr;
5993 5994
		ptr += nr_cpu_ids * sizeof(void **);

5995
		root_task_group.cfs_rq = (struct cfs_rq **)ptr;
5996
		ptr += nr_cpu_ids * sizeof(void **);
5997

5998
#endif /* CONFIG_FAIR_GROUP_SCHED */
5999
#ifdef CONFIG_RT_GROUP_SCHED
6000
		root_task_group.rt_se = (struct sched_rt_entity **)ptr;
6001 6002
		ptr += nr_cpu_ids * sizeof(void **);

6003
		root_task_group.rt_rq = (struct rt_rq **)ptr;
6004 6005
		ptr += nr_cpu_ids * sizeof(void **);

6006
#endif /* CONFIG_RT_GROUP_SCHED */
6007
	}
6008
#ifdef CONFIG_CPUMASK_OFFSTACK
6009 6010 6011
	for_each_possible_cpu(i) {
		per_cpu(load_balance_mask, i) = (cpumask_var_t)kzalloc_node(
			cpumask_size(), GFP_KERNEL, cpu_to_node(i));
6012 6013
		per_cpu(select_idle_mask, i) = (cpumask_var_t)kzalloc_node(
			cpumask_size(), GFP_KERNEL, cpu_to_node(i));
6014
	}
6015
#endif /* CONFIG_CPUMASK_OFFSTACK */
I
Ingo Molnar 已提交
6016

I
Ingo Molnar 已提交
6017 6018
	init_rt_bandwidth(&def_rt_bandwidth, global_rt_period(), global_rt_runtime());
	init_dl_bandwidth(&def_dl_bandwidth, global_rt_period(), global_rt_runtime());
6019

G
Gregory Haskins 已提交
6020 6021 6022 6023
#ifdef CONFIG_SMP
	init_defrootdomain();
#endif

6024
#ifdef CONFIG_RT_GROUP_SCHED
6025
	init_rt_bandwidth(&root_task_group.rt_bandwidth,
6026
			global_rt_period(), global_rt_runtime());
6027
#endif /* CONFIG_RT_GROUP_SCHED */
6028

D
Dhaval Giani 已提交
6029
#ifdef CONFIG_CGROUP_SCHED
6030 6031
	task_group_cache = KMEM_CACHE(task_group, 0);

6032 6033
	list_add(&root_task_group.list, &task_groups);
	INIT_LIST_HEAD(&root_task_group.children);
6034
	INIT_LIST_HEAD(&root_task_group.siblings);
6035
	autogroup_init(&init_task);
D
Dhaval Giani 已提交
6036
#endif /* CONFIG_CGROUP_SCHED */
P
Peter Zijlstra 已提交
6037

6038
	for_each_possible_cpu(i) {
6039
		struct rq *rq;
L
Linus Torvalds 已提交
6040 6041

		rq = cpu_rq(i);
6042
		raw_spin_lock_init(&rq->lock);
N
Nick Piggin 已提交
6043
		rq->nr_running = 0;
6044 6045
		rq->calc_load_active = 0;
		rq->calc_load_update = jiffies + LOAD_FREQ;
6046
		init_cfs_rq(&rq->cfs);
6047 6048
		init_rt_rq(&rq->rt);
		init_dl_rq(&rq->dl);
I
Ingo Molnar 已提交
6049
#ifdef CONFIG_FAIR_GROUP_SCHED
6050
		root_task_group.shares = ROOT_TASK_GROUP_LOAD;
P
Peter Zijlstra 已提交
6051
		INIT_LIST_HEAD(&rq->leaf_cfs_rq_list);
6052
		rq->tmp_alone_branch = &rq->leaf_cfs_rq_list;
D
Dhaval Giani 已提交
6053
		/*
I
Ingo Molnar 已提交
6054
		 * How much CPU bandwidth does root_task_group get?
D
Dhaval Giani 已提交
6055 6056
		 *
		 * In case of task-groups formed thr' the cgroup filesystem, it
I
Ingo Molnar 已提交
6057 6058
		 * gets 100% of the CPU resources in the system. This overall
		 * system CPU resource is divided among the tasks of
6059
		 * root_task_group and its child task-groups in a fair manner,
D
Dhaval Giani 已提交
6060 6061 6062
		 * based on each entity's (task or task-group's) weight
		 * (se->load.weight).
		 *
6063
		 * In other words, if root_task_group has 10 tasks of weight
D
Dhaval Giani 已提交
6064
		 * 1024) and two child groups A0 and A1 (of weight 1024 each),
I
Ingo Molnar 已提交
6065
		 * then A0's share of the CPU resource is:
D
Dhaval Giani 已提交
6066
		 *
6067
		 *	A0's bandwidth = 1024 / (10*1024 + 1024 + 1024) = 8.33%
D
Dhaval Giani 已提交
6068
		 *
6069 6070
		 * 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
6071
		 */
6072
		init_cfs_bandwidth(&root_task_group.cfs_bandwidth);
6073
		init_tg_cfs_entry(&root_task_group, &rq->cfs, NULL, i, NULL);
D
Dhaval Giani 已提交
6074 6075 6076
#endif /* CONFIG_FAIR_GROUP_SCHED */

		rq->rt.rt_runtime = def_rt_bandwidth.rt_runtime;
6077
#ifdef CONFIG_RT_GROUP_SCHED
6078
		init_tg_rt_entry(&root_task_group, &rq->rt, NULL, i, NULL);
I
Ingo Molnar 已提交
6079
#endif
L
Linus Torvalds 已提交
6080

I
Ingo Molnar 已提交
6081 6082
		for (j = 0; j < CPU_LOAD_IDX_MAX; j++)
			rq->cpu_load[j] = 0;
6083

L
Linus Torvalds 已提交
6084
#ifdef CONFIG_SMP
N
Nick Piggin 已提交
6085
		rq->sd = NULL;
G
Gregory Haskins 已提交
6086
		rq->rd = NULL;
6087
		rq->cpu_capacity = rq->cpu_capacity_orig = SCHED_CAPACITY_SCALE;
6088
		rq->balance_callback = NULL;
L
Linus Torvalds 已提交
6089
		rq->active_balance = 0;
I
Ingo Molnar 已提交
6090
		rq->next_balance = jiffies;
L
Linus Torvalds 已提交
6091
		rq->push_cpu = 0;
6092
		rq->cpu = i;
6093
		rq->online = 0;
6094 6095
		rq->idle_stamp = 0;
		rq->avg_idle = 2*sysctl_sched_migration_cost;
6096
		rq->max_idle_balance_cost = sysctl_sched_migration_cost;
6097 6098 6099

		INIT_LIST_HEAD(&rq->cfs_tasks);

6100
		rq_attach_root(rq, &def_root_domain);
6101
#ifdef CONFIG_NO_HZ_COMMON
6102
		rq->last_load_update_tick = jiffies;
6103
		rq->nohz_flags = 0;
6104
#endif
6105 6106 6107
#ifdef CONFIG_NO_HZ_FULL
		rq->last_sched_tick = 0;
#endif
6108
#endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
6109
		init_rq_hrtick(rq);
L
Linus Torvalds 已提交
6110 6111 6112
		atomic_set(&rq->nr_iowait, 0);
	}

6113
	set_load_weight(&init_task);
6114

L
Linus Torvalds 已提交
6115 6116 6117
	/*
	 * The boot idle thread does lazy MMU switching as well:
	 */
V
Vegard Nossum 已提交
6118
	mmgrab(&init_mm);
L
Linus Torvalds 已提交
6119 6120 6121 6122 6123 6124 6125 6126 6127
	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());
6128 6129 6130

	calc_load_update = jiffies + LOAD_FREQ;

6131
#ifdef CONFIG_SMP
6132
	zalloc_cpumask_var(&sched_domains_tmpmask, GFP_NOWAIT);
R
Rusty Russell 已提交
6133 6134 6135
	/* May be allocated at isolcpus cmdline parse time */
	if (cpu_isolated_map == NULL)
		zalloc_cpumask_var(&cpu_isolated_map, GFP_NOWAIT);
6136
	idle_thread_set_boot_cpu();
6137
	set_cpu_rq_start_time(smp_processor_id());
6138 6139
#endif
	init_sched_fair_class();
6140

6141 6142
	init_schedstats();

6143
	scheduler_running = 1;
L
Linus Torvalds 已提交
6144 6145
}

6146
#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
6147 6148
static inline int preempt_count_equals(int preempt_offset)
{
6149
	int nested = preempt_count() + rcu_preempt_depth();
6150

A
Arnd Bergmann 已提交
6151
	return (nested == preempt_offset);
6152 6153
}

6154
void __might_sleep(const char *file, int line, int preempt_offset)
L
Linus Torvalds 已提交
6155
{
P
Peter Zijlstra 已提交
6156 6157 6158 6159 6160
	/*
	 * 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.
	 */
6161
	WARN_ONCE(current->state != TASK_RUNNING && current->task_state_change,
P
Peter Zijlstra 已提交
6162 6163 6164 6165
			"do not call blocking ops when !TASK_RUNNING; "
			"state=%lx set at [<%p>] %pS\n",
			current->state,
			(void *)current->task_state_change,
6166
			(void *)current->task_state_change);
P
Peter Zijlstra 已提交
6167

6168 6169 6170 6171 6172
	___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
6173
{
I
Ingo Molnar 已提交
6174 6175 6176
	/* Ratelimiting timestamp: */
	static unsigned long prev_jiffy;

6177
	unsigned long preempt_disable_ip;
L
Linus Torvalds 已提交
6178

I
Ingo Molnar 已提交
6179 6180 6181
	/* WARN_ON_ONCE() by default, no rate limit required: */
	rcu_sleep_check();

6182 6183
	if ((preempt_count_equals(preempt_offset) && !irqs_disabled() &&
	     !is_idle_task(current)) ||
6184
	    system_state != SYSTEM_RUNNING || oops_in_progress)
I
Ingo Molnar 已提交
6185 6186 6187 6188 6189
		return;
	if (time_before(jiffies, prev_jiffy + HZ) && prev_jiffy)
		return;
	prev_jiffy = jiffies;

I
Ingo Molnar 已提交
6190
	/* Save this before calling printk(), since that will clobber it: */
6191 6192
	preempt_disable_ip = get_preempt_disable_ip(current);

P
Peter Zijlstra 已提交
6193 6194 6195 6196 6197 6198 6199
	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
6200

6201 6202 6203
	if (task_stack_end_corrupted(current))
		printk(KERN_EMERG "Thread overran stack, or stack corrupted\n");

I
Ingo Molnar 已提交
6204 6205 6206
	debug_show_held_locks(current);
	if (irqs_disabled())
		print_irqtrace_events(current);
6207 6208
	if (IS_ENABLED(CONFIG_DEBUG_PREEMPT)
	    && !preempt_count_equals(preempt_offset)) {
6209
		pr_err("Preemption disabled at:");
6210
		print_ip_sym(preempt_disable_ip);
6211 6212
		pr_cont("\n");
	}
I
Ingo Molnar 已提交
6213
	dump_stack();
6214
	add_taint(TAINT_WARN, LOCKDEP_STILL_OK);
L
Linus Torvalds 已提交
6215
}
6216
EXPORT_SYMBOL(___might_sleep);
L
Linus Torvalds 已提交
6217 6218 6219
#endif

#ifdef CONFIG_MAGIC_SYSRQ
6220
void normalize_rt_tasks(void)
6221
{
6222
	struct task_struct *g, *p;
6223 6224 6225
	struct sched_attr attr = {
		.sched_policy = SCHED_NORMAL,
	};
L
Linus Torvalds 已提交
6226

6227
	read_lock(&tasklist_lock);
6228
	for_each_process_thread(g, p) {
6229 6230 6231
		/*
		 * Only normalize user tasks:
		 */
6232
		if (p->flags & PF_KTHREAD)
6233 6234
			continue;

6235 6236 6237 6238
		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
6239

6240
		if (!dl_task(p) && !rt_task(p)) {
I
Ingo Molnar 已提交
6241 6242 6243 6244
			/*
			 * Renice negative nice level userspace
			 * tasks back to 0:
			 */
6245
			if (task_nice(p) < 0)
I
Ingo Molnar 已提交
6246
				set_user_nice(p, 0);
L
Linus Torvalds 已提交
6247
			continue;
I
Ingo Molnar 已提交
6248
		}
L
Linus Torvalds 已提交
6249

6250
		__sched_setscheduler(p, &attr, false, false);
6251
	}
6252
	read_unlock(&tasklist_lock);
L
Linus Torvalds 已提交
6253 6254 6255
}

#endif /* CONFIG_MAGIC_SYSRQ */
6256

6257
#if defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB)
6258
/*
6259
 * These functions are only useful for the IA64 MCA handling, or kdb.
6260 6261 6262 6263 6264 6265 6266 6267 6268
 *
 * 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
6269
 * curr_task - return the current task for a given CPU.
6270 6271 6272
 * @cpu: the processor in question.
 *
 * ONLY VALID WHEN THE WHOLE SYSTEM IS STOPPED!
6273 6274
 *
 * Return: The current task for @cpu.
6275
 */
6276
struct task_struct *curr_task(int cpu)
6277 6278 6279 6280
{
	return cpu_curr(cpu);
}

6281 6282 6283
#endif /* defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB) */

#ifdef CONFIG_IA64
6284
/**
I
Ingo Molnar 已提交
6285
 * set_curr_task - set the current task for a given CPU.
6286 6287 6288 6289
 * @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
6290
 * are serviced on a separate stack. It allows the architecture to switch the
I
Ingo Molnar 已提交
6291
 * notion of the current task on a CPU in a non-blocking manner. This function
6292 6293 6294 6295 6296 6297 6298
 * 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!
 */
6299
void ia64_set_curr_task(int cpu, struct task_struct *p)
6300 6301 6302 6303 6304
{
	cpu_curr(cpu) = p;
}

#endif
S
Srivatsa Vaddagiri 已提交
6305

D
Dhaval Giani 已提交
6306
#ifdef CONFIG_CGROUP_SCHED
6307 6308 6309
/* task_group_lock serializes the addition/removal of task groups */
static DEFINE_SPINLOCK(task_group_lock);

6310
static void sched_free_group(struct task_group *tg)
6311 6312 6313
{
	free_fair_sched_group(tg);
	free_rt_sched_group(tg);
6314
	autogroup_free(tg);
6315
	kmem_cache_free(task_group_cache, tg);
6316 6317 6318
}

/* allocate runqueue etc for a new task group */
6319
struct task_group *sched_create_group(struct task_group *parent)
6320 6321 6322
{
	struct task_group *tg;

6323
	tg = kmem_cache_alloc(task_group_cache, GFP_KERNEL | __GFP_ZERO);
6324 6325 6326
	if (!tg)
		return ERR_PTR(-ENOMEM);

6327
	if (!alloc_fair_sched_group(tg, parent))
6328 6329
		goto err;

6330
	if (!alloc_rt_sched_group(tg, parent))
6331 6332
		goto err;

6333 6334 6335
	return tg;

err:
6336
	sched_free_group(tg);
6337 6338 6339 6340 6341 6342 6343
	return ERR_PTR(-ENOMEM);
}

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

6344
	spin_lock_irqsave(&task_group_lock, flags);
P
Peter Zijlstra 已提交
6345
	list_add_rcu(&tg->list, &task_groups);
P
Peter Zijlstra 已提交
6346

I
Ingo Molnar 已提交
6347 6348
	/* Root should already exist: */
	WARN_ON(!parent);
P
Peter Zijlstra 已提交
6349 6350 6351

	tg->parent = parent;
	INIT_LIST_HEAD(&tg->children);
6352
	list_add_rcu(&tg->siblings, &parent->children);
6353
	spin_unlock_irqrestore(&task_group_lock, flags);
6354 6355

	online_fair_sched_group(tg);
S
Srivatsa Vaddagiri 已提交
6356 6357
}

6358
/* rcu callback to free various structures associated with a task group */
6359
static void sched_free_group_rcu(struct rcu_head *rhp)
S
Srivatsa Vaddagiri 已提交
6360
{
I
Ingo Molnar 已提交
6361
	/* Now it should be safe to free those cfs_rqs: */
6362
	sched_free_group(container_of(rhp, struct task_group, rcu));
S
Srivatsa Vaddagiri 已提交
6363 6364
}

6365
void sched_destroy_group(struct task_group *tg)
6366
{
I
Ingo Molnar 已提交
6367
	/* Wait for possible concurrent references to cfs_rqs complete: */
6368
	call_rcu(&tg->rcu, sched_free_group_rcu);
6369 6370 6371
}

void sched_offline_group(struct task_group *tg)
S
Srivatsa Vaddagiri 已提交
6372
{
6373
	unsigned long flags;
S
Srivatsa Vaddagiri 已提交
6374

I
Ingo Molnar 已提交
6375
	/* End participation in shares distribution: */
6376
	unregister_fair_sched_group(tg);
6377 6378

	spin_lock_irqsave(&task_group_lock, flags);
P
Peter Zijlstra 已提交
6379
	list_del_rcu(&tg->list);
P
Peter Zijlstra 已提交
6380
	list_del_rcu(&tg->siblings);
6381
	spin_unlock_irqrestore(&task_group_lock, flags);
S
Srivatsa Vaddagiri 已提交
6382 6383
}

6384
static void sched_change_group(struct task_struct *tsk, int type)
S
Srivatsa Vaddagiri 已提交
6385
{
P
Peter Zijlstra 已提交
6386
	struct task_group *tg;
S
Srivatsa Vaddagiri 已提交
6387

6388 6389 6390 6391 6392 6393
	/*
	 * 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
6394 6395 6396 6397
			  struct task_group, css);
	tg = autogroup_task_group(tsk, tg);
	tsk->sched_task_group = tg;

P
Peter Zijlstra 已提交
6398
#ifdef CONFIG_FAIR_GROUP_SCHED
6399 6400
	if (tsk->sched_class->task_change_group)
		tsk->sched_class->task_change_group(tsk, type);
6401
	else
P
Peter Zijlstra 已提交
6402
#endif
6403
		set_task_rq(tsk, task_cpu(tsk));
6404 6405 6406 6407 6408 6409 6410 6411 6412 6413 6414 6415 6416 6417 6418 6419
}

/*
 * 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);
6420
	update_rq_clock(rq);
6421 6422 6423 6424 6425 6426

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

	if (queued)
		dequeue_task(rq, tsk, DEQUEUE_SAVE | DEQUEUE_MOVE);
6427
	if (running)
6428 6429 6430
		put_prev_task(rq, tsk);

	sched_change_group(tsk, TASK_MOVE_GROUP);
P
Peter Zijlstra 已提交
6431

6432
	if (queued)
6433
		enqueue_task(rq, tsk, ENQUEUE_RESTORE | ENQUEUE_MOVE);
6434
	if (running)
6435
		set_curr_task(rq, tsk);
S
Srivatsa Vaddagiri 已提交
6436

6437
	task_rq_unlock(rq, tsk, &rf);
S
Srivatsa Vaddagiri 已提交
6438
}
D
Dhaval Giani 已提交
6439
#endif /* CONFIG_CGROUP_SCHED */
S
Srivatsa Vaddagiri 已提交
6440

6441 6442 6443 6444 6445
#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
6446

P
Peter Zijlstra 已提交
6447 6448
/* Must be called with tasklist_lock held */
static inline int tg_has_rt_tasks(struct task_group *tg)
6449
{
P
Peter Zijlstra 已提交
6450
	struct task_struct *g, *p;
6451

6452 6453 6454 6455 6456 6457
	/*
	 * Autogroups do not have RT tasks; see autogroup_create().
	 */
	if (task_group_is_autogroup(tg))
		return 0;

6458
	for_each_process_thread(g, p) {
6459
		if (rt_task(p) && task_group(p) == tg)
P
Peter Zijlstra 已提交
6460
			return 1;
6461
	}
6462

P
Peter Zijlstra 已提交
6463 6464
	return 0;
}
6465

P
Peter Zijlstra 已提交
6466 6467 6468 6469 6470
struct rt_schedulable_data {
	struct task_group *tg;
	u64 rt_period;
	u64 rt_runtime;
};
6471

6472
static int tg_rt_schedulable(struct task_group *tg, void *data)
P
Peter Zijlstra 已提交
6473 6474 6475 6476 6477
{
	struct rt_schedulable_data *d = data;
	struct task_group *child;
	unsigned long total, sum = 0;
	u64 period, runtime;
6478

P
Peter Zijlstra 已提交
6479 6480
	period = ktime_to_ns(tg->rt_bandwidth.rt_period);
	runtime = tg->rt_bandwidth.rt_runtime;
6481

P
Peter Zijlstra 已提交
6482 6483 6484
	if (tg == d->tg) {
		period = d->rt_period;
		runtime = d->rt_runtime;
6485 6486
	}

6487 6488 6489 6490 6491
	/*
	 * Cannot have more runtime than the period.
	 */
	if (runtime > period && runtime != RUNTIME_INF)
		return -EINVAL;
P
Peter Zijlstra 已提交
6492

6493 6494 6495
	/*
	 * Ensure we don't starve existing RT tasks.
	 */
P
Peter Zijlstra 已提交
6496 6497
	if (rt_bandwidth_enabled() && !runtime && tg_has_rt_tasks(tg))
		return -EBUSY;
P
Peter Zijlstra 已提交
6498

P
Peter Zijlstra 已提交
6499
	total = to_ratio(period, runtime);
P
Peter Zijlstra 已提交
6500

6501 6502 6503 6504 6505
	/*
	 * 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
6506

6507 6508 6509
	/*
	 * The sum of our children's runtime should not exceed our own.
	 */
P
Peter Zijlstra 已提交
6510 6511 6512
	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
6513

P
Peter Zijlstra 已提交
6514 6515 6516 6517
		if (child == d->tg) {
			period = d->rt_period;
			runtime = d->rt_runtime;
		}
P
Peter Zijlstra 已提交
6518

P
Peter Zijlstra 已提交
6519
		sum += to_ratio(period, runtime);
P
Peter Zijlstra 已提交
6520
	}
P
Peter Zijlstra 已提交
6521

P
Peter Zijlstra 已提交
6522 6523 6524 6525
	if (sum > total)
		return -EINVAL;

	return 0;
P
Peter Zijlstra 已提交
6526 6527
}

P
Peter Zijlstra 已提交
6528
static int __rt_schedulable(struct task_group *tg, u64 period, u64 runtime)
6529
{
6530 6531
	int ret;

P
Peter Zijlstra 已提交
6532 6533 6534 6535 6536 6537
	struct rt_schedulable_data data = {
		.tg = tg,
		.rt_period = period,
		.rt_runtime = runtime,
	};

6538 6539 6540 6541 6542
	rcu_read_lock();
	ret = walk_tg_tree(tg_rt_schedulable, tg_nop, &data);
	rcu_read_unlock();

	return ret;
6543 6544
}

6545
static int tg_set_rt_bandwidth(struct task_group *tg,
6546
		u64 rt_period, u64 rt_runtime)
P
Peter Zijlstra 已提交
6547
{
P
Peter Zijlstra 已提交
6548
	int i, err = 0;
P
Peter Zijlstra 已提交
6549

6550 6551 6552 6553 6554 6555 6556 6557 6558 6559 6560
	/*
	 * 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
6561
	mutex_lock(&rt_constraints_mutex);
6562
	read_lock(&tasklist_lock);
P
Peter Zijlstra 已提交
6563 6564
	err = __rt_schedulable(tg, rt_period, rt_runtime);
	if (err)
P
Peter Zijlstra 已提交
6565
		goto unlock;
P
Peter Zijlstra 已提交
6566

6567
	raw_spin_lock_irq(&tg->rt_bandwidth.rt_runtime_lock);
6568 6569
	tg->rt_bandwidth.rt_period = ns_to_ktime(rt_period);
	tg->rt_bandwidth.rt_runtime = rt_runtime;
P
Peter Zijlstra 已提交
6570 6571 6572 6573

	for_each_possible_cpu(i) {
		struct rt_rq *rt_rq = tg->rt_rq[i];

6574
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
6575
		rt_rq->rt_runtime = rt_runtime;
6576
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
6577
	}
6578
	raw_spin_unlock_irq(&tg->rt_bandwidth.rt_runtime_lock);
P
Peter Zijlstra 已提交
6579
unlock:
6580
	read_unlock(&tasklist_lock);
P
Peter Zijlstra 已提交
6581 6582 6583
	mutex_unlock(&rt_constraints_mutex);

	return err;
P
Peter Zijlstra 已提交
6584 6585
}

6586
static int sched_group_set_rt_runtime(struct task_group *tg, long rt_runtime_us)
6587 6588 6589 6590 6591 6592 6593 6594
{
	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;

6595
	return tg_set_rt_bandwidth(tg, rt_period, rt_runtime);
6596 6597
}

6598
static long sched_group_rt_runtime(struct task_group *tg)
P
Peter Zijlstra 已提交
6599 6600 6601
{
	u64 rt_runtime_us;

6602
	if (tg->rt_bandwidth.rt_runtime == RUNTIME_INF)
P
Peter Zijlstra 已提交
6603 6604
		return -1;

6605
	rt_runtime_us = tg->rt_bandwidth.rt_runtime;
P
Peter Zijlstra 已提交
6606 6607 6608
	do_div(rt_runtime_us, NSEC_PER_USEC);
	return rt_runtime_us;
}
6609

6610
static int sched_group_set_rt_period(struct task_group *tg, u64 rt_period_us)
6611 6612 6613
{
	u64 rt_runtime, rt_period;

6614
	rt_period = rt_period_us * NSEC_PER_USEC;
6615 6616
	rt_runtime = tg->rt_bandwidth.rt_runtime;

6617
	return tg_set_rt_bandwidth(tg, rt_period, rt_runtime);
6618 6619
}

6620
static long sched_group_rt_period(struct task_group *tg)
6621 6622 6623 6624 6625 6626 6627
{
	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;
}
6628
#endif /* CONFIG_RT_GROUP_SCHED */
6629

6630
#ifdef CONFIG_RT_GROUP_SCHED
6631 6632 6633 6634 6635
static int sched_rt_global_constraints(void)
{
	int ret = 0;

	mutex_lock(&rt_constraints_mutex);
P
Peter Zijlstra 已提交
6636
	read_lock(&tasklist_lock);
6637
	ret = __rt_schedulable(NULL, 0, 0);
P
Peter Zijlstra 已提交
6638
	read_unlock(&tasklist_lock);
6639 6640 6641 6642
	mutex_unlock(&rt_constraints_mutex);

	return ret;
}
6643

6644
static int sched_rt_can_attach(struct task_group *tg, struct task_struct *tsk)
6645 6646 6647 6648 6649 6650 6651 6652
{
	/* 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;
}

6653
#else /* !CONFIG_RT_GROUP_SCHED */
6654 6655
static int sched_rt_global_constraints(void)
{
P
Peter Zijlstra 已提交
6656
	unsigned long flags;
6657
	int i;
6658

6659
	raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags);
P
Peter Zijlstra 已提交
6660 6661 6662
	for_each_possible_cpu(i) {
		struct rt_rq *rt_rq = &cpu_rq(i)->rt;

6663
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
6664
		rt_rq->rt_runtime = global_rt_runtime();
6665
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
6666
	}
6667
	raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags);
P
Peter Zijlstra 已提交
6668

6669
	return 0;
6670
}
6671
#endif /* CONFIG_RT_GROUP_SCHED */
6672

6673
static int sched_dl_global_validate(void)
6674
{
6675 6676
	u64 runtime = global_rt_runtime();
	u64 period = global_rt_period();
6677
	u64 new_bw = to_ratio(period, runtime);
6678
	struct dl_bw *dl_b;
6679
	int cpu, ret = 0;
6680
	unsigned long flags;
6681 6682 6683 6684 6685 6686 6687 6688 6689 6690

	/*
	 * 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!
	 */
6691
	for_each_possible_cpu(cpu) {
6692 6693
		rcu_read_lock_sched();
		dl_b = dl_bw_of(cpu);
6694

6695
		raw_spin_lock_irqsave(&dl_b->lock, flags);
6696 6697
		if (new_bw < dl_b->total_bw)
			ret = -EBUSY;
6698
		raw_spin_unlock_irqrestore(&dl_b->lock, flags);
6699

6700 6701
		rcu_read_unlock_sched();

6702 6703
		if (ret)
			break;
6704 6705
	}

6706
	return ret;
6707 6708
}

6709
static void sched_dl_do_global(void)
6710
{
6711
	u64 new_bw = -1;
6712
	struct dl_bw *dl_b;
6713
	int cpu;
6714
	unsigned long flags;
6715

6716 6717 6718 6719 6720 6721 6722 6723 6724 6725
	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) {
6726 6727
		rcu_read_lock_sched();
		dl_b = dl_bw_of(cpu);
6728

6729
		raw_spin_lock_irqsave(&dl_b->lock, flags);
6730
		dl_b->bw = new_bw;
6731
		raw_spin_unlock_irqrestore(&dl_b->lock, flags);
6732 6733

		rcu_read_unlock_sched();
6734
	}
6735 6736 6737 6738 6739 6740 6741
}

static int sched_rt_global_validate(void)
{
	if (sysctl_sched_rt_period <= 0)
		return -EINVAL;

6742 6743
	if ((sysctl_sched_rt_runtime != RUNTIME_INF) &&
		(sysctl_sched_rt_runtime > sysctl_sched_rt_period))
6744 6745 6746 6747 6748 6749 6750 6751 6752
		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());
6753 6754
}

6755
int sched_rt_handler(struct ctl_table *table, int write,
6756
		void __user *buffer, size_t *lenp,
6757 6758 6759 6760
		loff_t *ppos)
{
	int old_period, old_runtime;
	static DEFINE_MUTEX(mutex);
6761
	int ret;
6762 6763 6764 6765 6766

	mutex_lock(&mutex);
	old_period = sysctl_sched_rt_period;
	old_runtime = sysctl_sched_rt_runtime;

6767
	ret = proc_dointvec(table, write, buffer, lenp, ppos);
6768 6769

	if (!ret && write) {
6770 6771 6772 6773
		ret = sched_rt_global_validate();
		if (ret)
			goto undo;

6774
		ret = sched_dl_global_validate();
6775 6776 6777
		if (ret)
			goto undo;

6778
		ret = sched_rt_global_constraints();
6779 6780 6781 6782 6783 6784 6785 6786 6787 6788
		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;
6789 6790 6791 6792 6793
	}
	mutex_unlock(&mutex);

	return ret;
}
6794

6795
int sched_rr_handler(struct ctl_table *table, int write,
6796 6797 6798 6799 6800 6801 6802 6803
		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
6804 6805 6806 6807
	/*
	 * Make sure that internally we keep jiffies.
	 * Also, writing zero resets the timeslice to default:
	 */
6808
	if (!ret && write) {
6809 6810 6811
		sched_rr_timeslice =
			sysctl_sched_rr_timeslice <= 0 ? RR_TIMESLICE :
			msecs_to_jiffies(sysctl_sched_rr_timeslice);
6812 6813 6814 6815 6816
	}
	mutex_unlock(&mutex);
	return ret;
}

6817
#ifdef CONFIG_CGROUP_SCHED
6818

6819
static inline struct task_group *css_tg(struct cgroup_subsys_state *css)
6820
{
6821
	return css ? container_of(css, struct task_group, css) : NULL;
6822 6823
}

6824 6825
static struct cgroup_subsys_state *
cpu_cgroup_css_alloc(struct cgroup_subsys_state *parent_css)
6826
{
6827 6828
	struct task_group *parent = css_tg(parent_css);
	struct task_group *tg;
6829

6830
	if (!parent) {
6831
		/* This is early initialization for the top cgroup */
6832
		return &root_task_group.css;
6833 6834
	}

6835
	tg = sched_create_group(parent);
6836 6837 6838 6839 6840 6841
	if (IS_ERR(tg))
		return ERR_PTR(-ENOMEM);

	return &tg->css;
}

6842 6843 6844 6845 6846 6847 6848 6849 6850 6851 6852
/* 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;
}

6853
static void cpu_cgroup_css_released(struct cgroup_subsys_state *css)
6854
{
6855
	struct task_group *tg = css_tg(css);
6856

6857
	sched_offline_group(tg);
6858 6859
}

6860
static void cpu_cgroup_css_free(struct cgroup_subsys_state *css)
6861
{
6862
	struct task_group *tg = css_tg(css);
6863

6864 6865 6866 6867
	/*
	 * Relies on the RCU grace period between css_released() and this.
	 */
	sched_free_group(tg);
6868 6869
}

6870 6871 6872 6873
/*
 * 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.
 */
6874
static void cpu_cgroup_fork(struct task_struct *task)
6875
{
6876 6877 6878 6879 6880
	struct rq_flags rf;
	struct rq *rq;

	rq = task_rq_lock(task, &rf);

6881
	update_rq_clock(rq);
6882 6883 6884
	sched_change_group(task, TASK_SET_GROUP);

	task_rq_unlock(rq, task, &rf);
6885 6886
}

6887
static int cpu_cgroup_can_attach(struct cgroup_taskset *tset)
6888
{
6889
	struct task_struct *task;
6890
	struct cgroup_subsys_state *css;
6891
	int ret = 0;
6892

6893
	cgroup_taskset_for_each(task, css, tset) {
6894
#ifdef CONFIG_RT_GROUP_SCHED
6895
		if (!sched_rt_can_attach(css_tg(css), task))
6896
			return -EINVAL;
6897
#else
6898 6899 6900
		/* We don't support RT-tasks being in separate groups */
		if (task->sched_class != &fair_sched_class)
			return -EINVAL;
6901
#endif
6902 6903 6904 6905 6906 6907 6908 6909 6910 6911 6912 6913 6914 6915 6916 6917
		/*
		 * 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;
6918
	}
6919
	return ret;
6920
}
6921

6922
static void cpu_cgroup_attach(struct cgroup_taskset *tset)
6923
{
6924
	struct task_struct *task;
6925
	struct cgroup_subsys_state *css;
6926

6927
	cgroup_taskset_for_each(task, css, tset)
6928
		sched_move_task(task);
6929 6930
}

6931
#ifdef CONFIG_FAIR_GROUP_SCHED
6932 6933
static int cpu_shares_write_u64(struct cgroup_subsys_state *css,
				struct cftype *cftype, u64 shareval)
6934
{
6935
	return sched_group_set_shares(css_tg(css), scale_load(shareval));
6936 6937
}

6938 6939
static u64 cpu_shares_read_u64(struct cgroup_subsys_state *css,
			       struct cftype *cft)
6940
{
6941
	struct task_group *tg = css_tg(css);
6942

6943
	return (u64) scale_load_down(tg->shares);
6944
}
6945 6946

#ifdef CONFIG_CFS_BANDWIDTH
6947 6948
static DEFINE_MUTEX(cfs_constraints_mutex);

6949 6950 6951
const u64 max_cfs_quota_period = 1 * NSEC_PER_SEC; /* 1s */
const u64 min_cfs_quota_period = 1 * NSEC_PER_MSEC; /* 1ms */

6952 6953
static int __cfs_schedulable(struct task_group *tg, u64 period, u64 runtime);

6954 6955
static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota)
{
6956
	int i, ret = 0, runtime_enabled, runtime_was_enabled;
6957
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
6958 6959 6960 6961 6962 6963 6964 6965 6966 6967 6968 6969 6970 6971 6972 6973 6974 6975 6976 6977

	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;

6978 6979 6980 6981 6982
	/*
	 * Prevent race between setting of cfs_rq->runtime_enabled and
	 * unthrottle_offline_cfs_rqs().
	 */
	get_online_cpus();
6983 6984 6985 6986 6987
	mutex_lock(&cfs_constraints_mutex);
	ret = __cfs_schedulable(tg, period, quota);
	if (ret)
		goto out_unlock;

6988
	runtime_enabled = quota != RUNTIME_INF;
6989
	runtime_was_enabled = cfs_b->quota != RUNTIME_INF;
6990 6991 6992 6993 6994 6995
	/*
	 * 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();
6996 6997 6998
	raw_spin_lock_irq(&cfs_b->lock);
	cfs_b->period = ns_to_ktime(period);
	cfs_b->quota = quota;
6999

P
Paul Turner 已提交
7000
	__refill_cfs_bandwidth_runtime(cfs_b);
I
Ingo Molnar 已提交
7001 7002

	/* Restart the period timer (if active) to handle new period expiry: */
P
Peter Zijlstra 已提交
7003 7004
	if (runtime_enabled)
		start_cfs_bandwidth(cfs_b);
I
Ingo Molnar 已提交
7005

7006 7007
	raw_spin_unlock_irq(&cfs_b->lock);

7008
	for_each_online_cpu(i) {
7009
		struct cfs_rq *cfs_rq = tg->cfs_rq[i];
7010
		struct rq *rq = cfs_rq->rq;
7011 7012

		raw_spin_lock_irq(&rq->lock);
7013
		cfs_rq->runtime_enabled = runtime_enabled;
7014
		cfs_rq->runtime_remaining = 0;
7015

7016
		if (cfs_rq->throttled)
7017
			unthrottle_cfs_rq(cfs_rq);
7018 7019
		raw_spin_unlock_irq(&rq->lock);
	}
7020 7021
	if (runtime_was_enabled && !runtime_enabled)
		cfs_bandwidth_usage_dec();
7022 7023
out_unlock:
	mutex_unlock(&cfs_constraints_mutex);
7024
	put_online_cpus();
7025

7026
	return ret;
7027 7028 7029 7030 7031 7032
}

int tg_set_cfs_quota(struct task_group *tg, long cfs_quota_us)
{
	u64 quota, period;

7033
	period = ktime_to_ns(tg->cfs_bandwidth.period);
7034 7035 7036 7037 7038 7039 7040 7041 7042 7043 7044 7045
	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;

7046
	if (tg->cfs_bandwidth.quota == RUNTIME_INF)
7047 7048
		return -1;

7049
	quota_us = tg->cfs_bandwidth.quota;
7050 7051 7052 7053 7054 7055 7056 7057 7058 7059
	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;
7060
	quota = tg->cfs_bandwidth.quota;
7061 7062 7063 7064 7065 7066 7067 7068

	return tg_set_cfs_bandwidth(tg, period, quota);
}

long tg_get_cfs_period(struct task_group *tg)
{
	u64 cfs_period_us;

7069
	cfs_period_us = ktime_to_ns(tg->cfs_bandwidth.period);
7070 7071 7072 7073 7074
	do_div(cfs_period_us, NSEC_PER_USEC);

	return cfs_period_us;
}

7075 7076
static s64 cpu_cfs_quota_read_s64(struct cgroup_subsys_state *css,
				  struct cftype *cft)
7077
{
7078
	return tg_get_cfs_quota(css_tg(css));
7079 7080
}

7081 7082
static int cpu_cfs_quota_write_s64(struct cgroup_subsys_state *css,
				   struct cftype *cftype, s64 cfs_quota_us)
7083
{
7084
	return tg_set_cfs_quota(css_tg(css), cfs_quota_us);
7085 7086
}

7087 7088
static u64 cpu_cfs_period_read_u64(struct cgroup_subsys_state *css,
				   struct cftype *cft)
7089
{
7090
	return tg_get_cfs_period(css_tg(css));
7091 7092
}

7093 7094
static int cpu_cfs_period_write_u64(struct cgroup_subsys_state *css,
				    struct cftype *cftype, u64 cfs_period_us)
7095
{
7096
	return tg_set_cfs_period(css_tg(css), cfs_period_us);
7097 7098
}

7099 7100 7101 7102 7103 7104 7105 7106 7107 7108 7109 7110 7111 7112 7113 7114 7115 7116 7117 7118 7119 7120 7121 7122 7123 7124 7125 7126 7127 7128 7129 7130
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;
7131
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
7132 7133 7134 7135 7136
	s64 quota = 0, parent_quota = -1;

	if (!tg->parent) {
		quota = RUNTIME_INF;
	} else {
7137
		struct cfs_bandwidth *parent_b = &tg->parent->cfs_bandwidth;
7138 7139

		quota = normalize_cfs_quota(tg, d);
7140
		parent_quota = parent_b->hierarchical_quota;
7141 7142

		/*
I
Ingo Molnar 已提交
7143 7144
		 * Ensure max(child_quota) <= parent_quota, inherit when no
		 * limit is set:
7145 7146 7147 7148 7149 7150
		 */
		if (quota == RUNTIME_INF)
			quota = parent_quota;
		else if (parent_quota != RUNTIME_INF && quota > parent_quota)
			return -EINVAL;
	}
7151
	cfs_b->hierarchical_quota = quota;
7152 7153 7154 7155 7156 7157

	return 0;
}

static int __cfs_schedulable(struct task_group *tg, u64 period, u64 quota)
{
7158
	int ret;
7159 7160 7161 7162 7163 7164 7165 7166 7167 7168 7169
	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);
	}

7170 7171 7172 7173 7174
	rcu_read_lock();
	ret = walk_tg_tree(tg_cfs_schedulable_down, tg_nop, &data);
	rcu_read_unlock();

	return ret;
7175
}
7176

7177
static int cpu_stats_show(struct seq_file *sf, void *v)
7178
{
7179
	struct task_group *tg = css_tg(seq_css(sf));
7180
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
7181

7182 7183 7184
	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);
7185 7186 7187

	return 0;
}
7188
#endif /* CONFIG_CFS_BANDWIDTH */
7189
#endif /* CONFIG_FAIR_GROUP_SCHED */
7190

7191
#ifdef CONFIG_RT_GROUP_SCHED
7192 7193
static int cpu_rt_runtime_write(struct cgroup_subsys_state *css,
				struct cftype *cft, s64 val)
P
Peter Zijlstra 已提交
7194
{
7195
	return sched_group_set_rt_runtime(css_tg(css), val);
P
Peter Zijlstra 已提交
7196 7197
}

7198 7199
static s64 cpu_rt_runtime_read(struct cgroup_subsys_state *css,
			       struct cftype *cft)
P
Peter Zijlstra 已提交
7200
{
7201
	return sched_group_rt_runtime(css_tg(css));
P
Peter Zijlstra 已提交
7202
}
7203

7204 7205
static int cpu_rt_period_write_uint(struct cgroup_subsys_state *css,
				    struct cftype *cftype, u64 rt_period_us)
7206
{
7207
	return sched_group_set_rt_period(css_tg(css), rt_period_us);
7208 7209
}

7210 7211
static u64 cpu_rt_period_read_uint(struct cgroup_subsys_state *css,
				   struct cftype *cft)
7212
{
7213
	return sched_group_rt_period(css_tg(css));
7214
}
7215
#endif /* CONFIG_RT_GROUP_SCHED */
P
Peter Zijlstra 已提交
7216

7217
static struct cftype cpu_files[] = {
7218
#ifdef CONFIG_FAIR_GROUP_SCHED
7219 7220
	{
		.name = "shares",
7221 7222
		.read_u64 = cpu_shares_read_u64,
		.write_u64 = cpu_shares_write_u64,
7223
	},
7224
#endif
7225 7226 7227 7228 7229 7230 7231 7232 7233 7234 7235
#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,
	},
7236 7237
	{
		.name = "stat",
7238
		.seq_show = cpu_stats_show,
7239
	},
7240
#endif
7241
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
7242
	{
P
Peter Zijlstra 已提交
7243
		.name = "rt_runtime_us",
7244 7245
		.read_s64 = cpu_rt_runtime_read,
		.write_s64 = cpu_rt_runtime_write,
P
Peter Zijlstra 已提交
7246
	},
7247 7248
	{
		.name = "rt_period_us",
7249 7250
		.read_u64 = cpu_rt_period_read_uint,
		.write_u64 = cpu_rt_period_write_uint,
7251
	},
7252
#endif
I
Ingo Molnar 已提交
7253
	{ }	/* Terminate */
7254 7255
};

7256
struct cgroup_subsys cpu_cgrp_subsys = {
7257
	.css_alloc	= cpu_cgroup_css_alloc,
7258
	.css_online	= cpu_cgroup_css_online,
7259
	.css_released	= cpu_cgroup_css_released,
7260
	.css_free	= cpu_cgroup_css_free,
7261
	.fork		= cpu_cgroup_fork,
7262 7263
	.can_attach	= cpu_cgroup_can_attach,
	.attach		= cpu_cgroup_attach,
7264
	.legacy_cftypes	= cpu_files,
7265
	.early_init	= true,
7266 7267
};

7268
#endif	/* CONFIG_CGROUP_SCHED */
7269

7270 7271 7272 7273 7274
void dump_cpu_task(int cpu)
{
	pr_info("Task dump for CPU %d:\n", cpu);
	sched_show_task(cpu_curr(cpu));
}
7275 7276 7277 7278 7279 7280 7281 7282 7283 7284 7285 7286 7287 7288 7289 7290 7291 7292 7293 7294 7295 7296 7297 7298 7299 7300 7301 7302 7303 7304 7305 7306 7307 7308 7309 7310 7311 7312 7313 7314 7315

/*
 * 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,
};