core.c 177.7 KB
Newer Older
L
Linus Torvalds 已提交
1
/*
2
 *  kernel/sched/core.c
L
Linus Torvalds 已提交
3
 *
I
Ingo Molnar 已提交
4
 *  Core kernel scheduler code and related syscalls
L
Linus Torvalds 已提交
5 6 7
 *
 *  Copyright (C) 1991-2002  Linus Torvalds
 */
8
#include <linux/sched.h>
L
Linus Torvalds 已提交
9
#include <linux/cpuset.h>
10
#include <linux/delayacct.h>
11
#include <linux/init_task.h>
12
#include <linux/context_tracking.h>
13 14 15 16 17 18

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

24
#include <asm/switch_to.h>
25
#include <asm/tlb.h>
L
Linus Torvalds 已提交
26

27
#include "sched.h"
28
#include "../workqueue_internal.h"
29
#include "../smpboot.h"
30

31
#define CREATE_TRACE_POINTS
32
#include <trace/events/sched.h>
33

34
DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
35

I
Ingo Molnar 已提交
36 37 38
/*
 * Debugging: various feature bits
 */
P
Peter Zijlstra 已提交
39 40 41 42

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

I
Ingo Molnar 已提交
43
const_debug unsigned int sysctl_sched_features =
44
#include "features.h"
P
Peter Zijlstra 已提交
45 46 47 48
	0;

#undef SCHED_FEAT

49 50 51 52 53 54
/*
 * 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;

55 56 57 58 59 60 61 62
/*
 * 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
63
/*
I
Ingo Molnar 已提交
64
 * period over which we measure -rt task CPU usage in us.
P
Peter Zijlstra 已提交
65 66
 * default: 1s
 */
P
Peter Zijlstra 已提交
67
unsigned int sysctl_sched_rt_period = 1000000;
P
Peter Zijlstra 已提交
68

69
__read_mostly int scheduler_running;
70

P
Peter Zijlstra 已提交
71 72 73 74 75
/*
 * 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
76

I
Ingo Molnar 已提交
77
/* CPUs with isolated domains */
78 79
cpumask_var_t cpu_isolated_map;

L
Linus Torvalds 已提交
80
/*
81
 * this_rq_lock - lock this runqueue and disable interrupts.
L
Linus Torvalds 已提交
82
 */
A
Alexey Dobriyan 已提交
83
static struct rq *this_rq_lock(void)
L
Linus Torvalds 已提交
84 85
	__acquires(rq->lock)
{
86
	struct rq *rq;
L
Linus Torvalds 已提交
87 88 89

	local_irq_disable();
	rq = this_rq();
90
	raw_spin_lock(&rq->lock);
L
Linus Torvalds 已提交
91 92 93 94

	return rq;
}

95 96 97
/*
 * __task_rq_lock - lock the rq @p resides on.
 */
98
struct rq *__task_rq_lock(struct task_struct *p, struct rq_flags *rf)
99 100 101 102 103 104 105 106 107 108
	__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))) {
109
			rq_pin_lock(rq, rf);
110 111 112 113 114 115 116 117 118 119 120 121
			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.
 */
122
struct rq *task_rq_lock(struct task_struct *p, struct rq_flags *rf)
123 124 125 126 127 128
	__acquires(p->pi_lock)
	__acquires(rq->lock)
{
	struct rq *rq;

	for (;;) {
129
		raw_spin_lock_irqsave(&p->pi_lock, rf->flags);
130 131 132 133 134 135 136 137 138 139 140 141 142 143 144
		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
145
		 * If we observe the new CPU in task_rq_lock, the acquire will
146 147 148
		 * pair with the WMB to ensure we must then also see migrating.
		 */
		if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) {
149
			rq_pin_lock(rq, rf);
150 151 152
			return rq;
		}
		raw_spin_unlock(&rq->lock);
153
		raw_spin_unlock_irqrestore(&p->pi_lock, rf->flags);
154 155 156 157 158 159

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

160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237
/*
 * 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
238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258
#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());

259
	raw_spin_lock(&rq->lock);
260
	update_rq_clock(rq);
P
Peter Zijlstra 已提交
261
	rq->curr->sched_class->task_tick(rq, rq->curr, 1);
262
	raw_spin_unlock(&rq->lock);
P
Peter Zijlstra 已提交
263 264 265 266

	return HRTIMER_NORESTART;
}

267
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
268

269
static void __hrtick_restart(struct rq *rq)
P
Peter Zijlstra 已提交
270 271 272
{
	struct hrtimer *timer = &rq->hrtick_timer;

273
	hrtimer_start_expires(timer, HRTIMER_MODE_ABS_PINNED);
P
Peter Zijlstra 已提交
274 275
}

276 277 278 279
/*
 * called from hardirq (IPI) context
 */
static void __hrtick_start(void *arg)
280
{
281
	struct rq *rq = arg;
282

283
	raw_spin_lock(&rq->lock);
P
Peter Zijlstra 已提交
284
	__hrtick_restart(rq);
285
	rq->hrtick_csd_pending = 0;
286
	raw_spin_unlock(&rq->lock);
287 288
}

289 290 291 292 293
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
294
void hrtick_start(struct rq *rq, u64 delay)
295
{
296
	struct hrtimer *timer = &rq->hrtick_timer;
297 298 299 300 301 302 303 304 305
	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);
306

307
	hrtimer_set_expires(timer, time);
308 309

	if (rq == this_rq()) {
P
Peter Zijlstra 已提交
310
		__hrtick_restart(rq);
311
	} else if (!rq->hrtick_csd_pending) {
312
		smp_call_function_single_async(cpu_of(rq), &rq->hrtick_csd);
313 314
		rq->hrtick_csd_pending = 1;
	}
315 316
}

317 318 319 320 321 322
#else
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
323
void hrtick_start(struct rq *rq, u64 delay)
324
{
W
Wanpeng Li 已提交
325 326 327 328 329
	/*
	 * Don't schedule slices shorter than 10000ns, that just
	 * doesn't make sense. Rely on vruntime for fairness.
	 */
	delay = max_t(u64, delay, 10000LL);
330 331
	hrtimer_start(&rq->hrtick_timer, ns_to_ktime(delay),
		      HRTIMER_MODE_REL_PINNED);
332 333
}
#endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
334

335
static void init_rq_hrtick(struct rq *rq)
P
Peter Zijlstra 已提交
336
{
337 338
#ifdef CONFIG_SMP
	rq->hrtick_csd_pending = 0;
P
Peter Zijlstra 已提交
339

340 341 342 343
	rq->hrtick_csd.flags = 0;
	rq->hrtick_csd.func = __hrtick_start;
	rq->hrtick_csd.info = rq;
#endif
P
Peter Zijlstra 已提交
344

345 346
	hrtimer_init(&rq->hrtick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	rq->hrtick_timer.function = hrtick;
P
Peter Zijlstra 已提交
347
}
A
Andrew Morton 已提交
348
#else	/* CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
349 350 351 352 353 354 355
static inline void hrtick_clear(struct rq *rq)
{
}

static inline void init_rq_hrtick(struct rq *rq)
{
}
A
Andrew Morton 已提交
356
#endif	/* CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
357

358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375
/*
 * 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;								\
})

376
#if defined(CONFIG_SMP) && defined(TIF_POLLING_NRFLAG)
377 378 379 380 381 382 383 384 385 386
/*
 * 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);
}
387 388 389 390 391 392 393 394 395 396

/*
 * 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);
397
	typeof(ti->flags) old, val = READ_ONCE(ti->flags);
398 399 400 401 402 403 404 405 406 407 408 409 410 411

	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;
}

412 413 414 415 416 417
#else
static bool set_nr_and_not_polling(struct task_struct *p)
{
	set_tsk_need_resched(p);
	return true;
}
418 419 420 421 422 423 424

#ifdef CONFIG_SMP
static bool set_nr_if_polling(struct task_struct *p)
{
	return false;
}
#endif
425 426
#endif

427 428 429 430 431 432 433 434 435 436
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
437
	 * barrier implied by the wakeup in wake_up_q().
438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459
	 */
	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
460
		/* Task can safely be re-inserted now: */
461 462 463 464 465 466 467 468 469 470 471 472
		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
473
/*
474
 * resched_curr - mark rq's current task 'to be rescheduled now'.
I
Ingo Molnar 已提交
475 476 477 478 479
 *
 * 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.
 */
480
void resched_curr(struct rq *rq)
I
Ingo Molnar 已提交
481
{
482
	struct task_struct *curr = rq->curr;
I
Ingo Molnar 已提交
483 484
	int cpu;

485
	lockdep_assert_held(&rq->lock);
I
Ingo Molnar 已提交
486

487
	if (test_tsk_need_resched(curr))
I
Ingo Molnar 已提交
488 489
		return;

490
	cpu = cpu_of(rq);
491

492
	if (cpu == smp_processor_id()) {
493
		set_tsk_need_resched(curr);
494
		set_preempt_need_resched();
I
Ingo Molnar 已提交
495
		return;
496
	}
I
Ingo Molnar 已提交
497

498
	if (set_nr_and_not_polling(curr))
I
Ingo Molnar 已提交
499
		smp_send_reschedule(cpu);
500 501
	else
		trace_sched_wake_idle_without_ipi(cpu);
I
Ingo Molnar 已提交
502 503
}

504
void resched_cpu(int cpu)
I
Ingo Molnar 已提交
505 506 507 508
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

509
	if (!raw_spin_trylock_irqsave(&rq->lock, flags))
I
Ingo Molnar 已提交
510
		return;
511
	resched_curr(rq);
512
	raw_spin_unlock_irqrestore(&rq->lock, flags);
I
Ingo Molnar 已提交
513
}
514

515
#ifdef CONFIG_SMP
516
#ifdef CONFIG_NO_HZ_COMMON
517
/*
I
Ingo Molnar 已提交
518 519
 * In the semi idle case, use the nearest busy CPU for migrating timers
 * from an idle CPU.  This is good for power-savings.
520 521
 *
 * We don't do similar optimization for completely idle system, as
I
Ingo Molnar 已提交
522 523
 * 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).
524
 */
525
int get_nohz_timer_target(void)
526
{
527
	int i, cpu = smp_processor_id();
528 529
	struct sched_domain *sd;

530
	if (!idle_cpu(cpu) && is_housekeeping_cpu(cpu))
531 532
		return cpu;

533
	rcu_read_lock();
534
	for_each_domain(cpu, sd) {
535
		for_each_cpu(i, sched_domain_span(sd)) {
536 537 538 539
			if (cpu == i)
				continue;

			if (!idle_cpu(i) && is_housekeeping_cpu(i)) {
540 541 542 543
				cpu = i;
				goto unlock;
			}
		}
544
	}
545 546 547

	if (!is_housekeeping_cpu(cpu))
		cpu = housekeeping_any_cpu();
548 549
unlock:
	rcu_read_unlock();
550 551
	return cpu;
}
I
Ingo Molnar 已提交
552

553 554 555 556 557 558 559 560 561 562
/*
 * 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.
 */
563
static void wake_up_idle_cpu(int cpu)
564 565 566 567 568 569
{
	struct rq *rq = cpu_rq(cpu);

	if (cpu == smp_processor_id())
		return;

570
	if (set_nr_and_not_polling(rq->idle))
571
		smp_send_reschedule(cpu);
572 573
	else
		trace_sched_wake_idle_without_ipi(cpu);
574 575
}

576
static bool wake_up_full_nohz_cpu(int cpu)
577
{
578 579 580 581 582 583
	/*
	 * 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.
	 */
584 585
	if (cpu_is_offline(cpu))
		return true;  /* Don't try to wake offline CPUs. */
586
	if (tick_nohz_full_cpu(cpu)) {
587 588
		if (cpu != smp_processor_id() ||
		    tick_nohz_tick_stopped())
589
			tick_nohz_full_kick_cpu(cpu);
590 591 592 593 594 595
		return true;
	}

	return false;
}

596 597 598 599 600
/*
 * 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.
 */
601 602
void wake_up_nohz_cpu(int cpu)
{
603
	if (!wake_up_full_nohz_cpu(cpu))
604 605 606
		wake_up_idle_cpu(cpu);
}

607
static inline bool got_nohz_idle_kick(void)
608
{
609
	int cpu = smp_processor_id();
610 611 612 613 614 615 616 617 618 619 620 621 622

	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;
623 624
}

625
#else /* CONFIG_NO_HZ_COMMON */
626

627
static inline bool got_nohz_idle_kick(void)
P
Peter Zijlstra 已提交
628
{
629
	return false;
P
Peter Zijlstra 已提交
630 631
}

632
#endif /* CONFIG_NO_HZ_COMMON */
633

634
#ifdef CONFIG_NO_HZ_FULL
635
bool sched_can_stop_tick(struct rq *rq)
636
{
637 638 639 640 641 642
	int fifo_nr_running;

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

643
	/*
644 645
	 * If there are more than one RR tasks, we need the tick to effect the
	 * actual RR behaviour.
646
	 */
647 648 649 650 651
	if (rq->rt.rr_nr_running) {
		if (rq->rt.rr_nr_running == 1)
			return true;
		else
			return false;
652 653
	}

654 655 656 657 658 659 660 661 662 663 664 665 666 667
	/*
	 * 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)
668
		return false;
669

670
	return true;
671 672
}
#endif /* CONFIG_NO_HZ_FULL */
673

674
void sched_avg_update(struct rq *rq)
675
{
676 677
	s64 period = sched_avg_period();

678
	while ((s64)(rq_clock(rq) - rq->age_stamp) > period) {
679 680 681 682 683 684
		/*
		 * 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));
685 686 687
		rq->age_stamp += period;
		rq->rt_avg /= 2;
	}
688 689
}

690
#endif /* CONFIG_SMP */
691

692 693
#if defined(CONFIG_RT_GROUP_SCHED) || (defined(CONFIG_FAIR_GROUP_SCHED) && \
			(defined(CONFIG_SMP) || defined(CONFIG_CFS_BANDWIDTH)))
694
/*
695 696 697 698
 * 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.
699
 */
700
int walk_tg_tree_from(struct task_group *from,
701
			     tg_visitor down, tg_visitor up, void *data)
702 703
{
	struct task_group *parent, *child;
P
Peter Zijlstra 已提交
704
	int ret;
705

706 707
	parent = from;

708
down:
P
Peter Zijlstra 已提交
709 710
	ret = (*down)(parent, data);
	if (ret)
711
		goto out;
712 713 714 715 716 717 718
	list_for_each_entry_rcu(child, &parent->children, siblings) {
		parent = child;
		goto down;

up:
		continue;
	}
P
Peter Zijlstra 已提交
719
	ret = (*up)(parent, data);
720 721
	if (ret || parent == from)
		goto out;
722 723 724 725 726

	child = parent;
	parent = parent->parent;
	if (parent)
		goto up;
727
out:
P
Peter Zijlstra 已提交
728
	return ret;
729 730
}

731
int tg_nop(struct task_group *tg, void *data)
P
Peter Zijlstra 已提交
732
{
733
	return 0;
P
Peter Zijlstra 已提交
734
}
735 736
#endif

737 738
static void set_load_weight(struct task_struct *p)
{
N
Nikhil Rao 已提交
739 740 741
	int prio = p->static_prio - MAX_RT_PRIO;
	struct load_weight *load = &p->se.load;

I
Ingo Molnar 已提交
742 743 744
	/*
	 * SCHED_IDLE tasks get minimal weight:
	 */
745
	if (idle_policy(p->policy)) {
746
		load->weight = scale_load(WEIGHT_IDLEPRIO);
N
Nikhil Rao 已提交
747
		load->inv_weight = WMULT_IDLEPRIO;
I
Ingo Molnar 已提交
748 749
		return;
	}
750

751 752
	load->weight = scale_load(sched_prio_to_weight[prio]);
	load->inv_weight = sched_prio_to_wmult[prio];
753 754
}

755
static inline void enqueue_task(struct rq *rq, struct task_struct *p, int flags)
756
{
757
	update_rq_clock(rq);
758 759
	if (!(flags & ENQUEUE_RESTORE))
		sched_info_queued(rq, p);
760
	p->sched_class->enqueue_task(rq, p, flags);
761 762
}

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

771
void activate_task(struct rq *rq, struct task_struct *p, int flags)
772 773 774 775
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible--;

776
	enqueue_task(rq, p, flags);
777 778
}

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

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

787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816
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;
	}
}

817
/*
I
Ingo Molnar 已提交
818
 * __normal_prio - return the priority that is based on the static prio
819 820 821
 */
static inline int __normal_prio(struct task_struct *p)
{
I
Ingo Molnar 已提交
822
	return p->static_prio;
823 824
}

825 826 827 828 829 830 831
/*
 * 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.
 */
832
static inline int normal_prio(struct task_struct *p)
833 834 835
{
	int prio;

836 837 838
	if (task_has_dl_policy(p))
		prio = MAX_DL_PRIO-1;
	else if (task_has_rt_policy(p))
839 840 841 842 843 844 845 846 847 848 849 850 851
		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.
 */
852
static int effective_prio(struct task_struct *p)
853 854 855 856 857 858 859 860 861 862 863 864
{
	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
865 866 867
/**
 * task_curr - is this task currently executing on a CPU?
 * @p: the task in question.
868 869
 *
 * Return: 1 if the task is currently executing. 0 otherwise.
L
Linus Torvalds 已提交
870
 */
871
inline int task_curr(const struct task_struct *p)
L
Linus Torvalds 已提交
872 873 874 875
{
	return cpu_curr(task_cpu(p)) == p;
}

876
/*
877 878 879 880 881
 * 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().
882
 */
883 884
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
885
				       int oldprio)
886 887 888
{
	if (prev_class != p->sched_class) {
		if (prev_class->switched_from)
P
Peter Zijlstra 已提交
889
			prev_class->switched_from(rq, p);
890

P
Peter Zijlstra 已提交
891
		p->sched_class->switched_to(rq, p);
892
	} else if (oldprio != p->prio || dl_task(p))
P
Peter Zijlstra 已提交
893
		p->sched_class->prio_changed(rq, p, oldprio);
894 895
}

896
void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags)
897 898 899 900 901 902 903 904 905 906
{
	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) {
907
				resched_curr(rq);
908 909 910 911 912 913 914 915 916
				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.
	 */
917
	if (task_on_rq_queued(rq->curr) && test_tsk_need_resched(rq->curr))
918
		rq_clock_skip_update(rq, true);
919 920
}

L
Linus Torvalds 已提交
921
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940
/*
 * 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.
 */
941
static struct rq *move_queued_task(struct rq *rq, struct task_struct *p, int new_cpu)
P
Peter Zijlstra 已提交
942 943 944 945
{
	lockdep_assert_held(&rq->lock);

	p->on_rq = TASK_ON_RQ_MIGRATING;
946
	dequeue_task(rq, p, 0);
P
Peter Zijlstra 已提交
947 948 949 950 951 952 953 954
	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);
955
	p->on_rq = TASK_ON_RQ_QUEUED;
P
Peter Zijlstra 已提交
956 957 958 959 960 961 962 963 964 965 966
	check_preempt_curr(rq, p, 0);

	return rq;
}

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

/*
I
Ingo Molnar 已提交
967
 * Move (not current) task off this CPU, onto the destination CPU. We're doing
P
Peter Zijlstra 已提交
968 969 970 971 972 973 974
 * 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.
 */
975
static struct rq *__migrate_task(struct rq *rq, struct task_struct *p, int dest_cpu)
P
Peter Zijlstra 已提交
976 977
{
	if (unlikely(!cpu_active(dest_cpu)))
978
		return rq;
P
Peter Zijlstra 已提交
979 980 981

	/* Affinity changed (again). */
	if (!cpumask_test_cpu(dest_cpu, tsk_cpus_allowed(p)))
982
		return rq;
P
Peter Zijlstra 已提交
983

984 985 986
	rq = move_queued_task(rq, p, dest_cpu);

	return rq;
P
Peter Zijlstra 已提交
987 988 989 990 991 992 993 994 995 996
}

/*
 * 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;
997 998
	struct task_struct *p = arg->task;
	struct rq *rq = this_rq();
P
Peter Zijlstra 已提交
999 1000

	/*
I
Ingo Molnar 已提交
1001 1002
	 * 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
1003 1004 1005 1006 1007 1008 1009 1010
	 */
	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();
1011 1012 1013 1014 1015 1016 1017 1018

	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.
	 */
1019 1020 1021 1022 1023 1024
	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;
	}
1025 1026 1027
	raw_spin_unlock(&rq->lock);
	raw_spin_unlock(&p->pi_lock);

P
Peter Zijlstra 已提交
1028 1029 1030 1031
	local_irq_enable();
	return 0;
}

1032 1033 1034 1035 1036
/*
 * 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
1037 1038 1039 1040 1041
{
	cpumask_copy(&p->cpus_allowed, new_mask);
	p->nr_cpus_allowed = cpumask_weight(new_mask);
}

1042 1043
void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
{
1044 1045 1046
	struct rq *rq = task_rq(p);
	bool queued, running;

1047
	lockdep_assert_held(&p->pi_lock);
1048 1049 1050 1051 1052 1053 1054 1055 1056 1057

	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);
1058
		dequeue_task(rq, p, DEQUEUE_SAVE);
1059 1060 1061 1062
	}
	if (running)
		put_prev_task(rq, p);

1063
	p->sched_class->set_cpus_allowed(p, new_mask);
1064 1065

	if (queued)
1066
		enqueue_task(rq, p, ENQUEUE_RESTORE);
1067
	if (running)
1068
		set_curr_task(rq, p);
1069 1070
}

P
Peter Zijlstra 已提交
1071 1072 1073 1074 1075 1076 1077 1078 1079
/*
 * 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.
 */
1080 1081
static int __set_cpus_allowed_ptr(struct task_struct *p,
				  const struct cpumask *new_mask, bool check)
P
Peter Zijlstra 已提交
1082
{
1083
	const struct cpumask *cpu_valid_mask = cpu_active_mask;
P
Peter Zijlstra 已提交
1084
	unsigned int dest_cpu;
1085 1086
	struct rq_flags rf;
	struct rq *rq;
P
Peter Zijlstra 已提交
1087 1088
	int ret = 0;

1089
	rq = task_rq_lock(p, &rf);
P
Peter Zijlstra 已提交
1090

1091 1092 1093 1094 1095 1096 1097
	if (p->flags & PF_KTHREAD) {
		/*
		 * Kernel threads are allowed on online && !active CPUs
		 */
		cpu_valid_mask = cpu_online_mask;
	}

1098 1099 1100 1101 1102 1103 1104 1105 1106
	/*
	 * 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
1107 1108 1109
	if (cpumask_equal(&p->cpus_allowed, new_mask))
		goto out;

1110
	if (!cpumask_intersects(new_mask, cpu_valid_mask)) {
P
Peter Zijlstra 已提交
1111 1112 1113 1114 1115 1116
		ret = -EINVAL;
		goto out;
	}

	do_set_cpus_allowed(p, new_mask);

1117 1118 1119
	if (p->flags & PF_KTHREAD) {
		/*
		 * For kernel threads that do indeed end up on online &&
I
Ingo Molnar 已提交
1120
		 * !active we want to ensure they are strict per-CPU threads.
1121 1122 1123 1124 1125 1126
		 */
		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
1127 1128 1129 1130
	/* 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;

1131
	dest_cpu = cpumask_any_and(cpu_valid_mask, new_mask);
P
Peter Zijlstra 已提交
1132 1133 1134
	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. */
1135
		task_rq_unlock(rq, p, &rf);
P
Peter Zijlstra 已提交
1136 1137 1138
		stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg);
		tlb_migrate_finish(p->mm);
		return 0;
1139 1140 1141 1142 1143
	} else if (task_on_rq_queued(p)) {
		/*
		 * OK, since we're going to drop the lock immediately
		 * afterwards anyway.
		 */
1144
		rq_unpin_lock(rq, &rf);
1145
		rq = move_queued_task(rq, p, dest_cpu);
1146
		rq_repin_lock(rq, &rf);
1147
	}
P
Peter Zijlstra 已提交
1148
out:
1149
	task_rq_unlock(rq, p, &rf);
P
Peter Zijlstra 已提交
1150 1151 1152

	return ret;
}
1153 1154 1155 1156 1157

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
1158 1159
EXPORT_SYMBOL_GPL(set_cpus_allowed_ptr);

I
Ingo Molnar 已提交
1160
void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
I
Ingo Molnar 已提交
1161
{
1162 1163 1164 1165 1166
#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
1167
	WARN_ON_ONCE(p->state != TASK_RUNNING && p->state != TASK_WAKING &&
O
Oleg Nesterov 已提交
1168
			!p->on_rq);
1169

1170 1171 1172 1173 1174 1175 1176 1177 1178
	/*
	 * 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)));

1179
#ifdef CONFIG_LOCKDEP
1180 1181 1182 1183 1184
	/*
	 * 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
1185
	 * see task_group().
1186 1187 1188 1189
	 *
	 * Furthermore, all task_rq users should acquire both locks, see
	 * task_rq_lock().
	 */
1190 1191 1192
	WARN_ON_ONCE(debug_locks && !(lockdep_is_held(&p->pi_lock) ||
				      lockdep_is_held(&task_rq(p)->lock)));
#endif
1193 1194
#endif

1195
	trace_sched_migrate_task(p, new_cpu);
1196

1197
	if (task_cpu(p) != new_cpu) {
1198
		if (p->sched_class->migrate_task_rq)
1199
			p->sched_class->migrate_task_rq(p);
1200
		p->se.nr_migrations++;
1201
		perf_event_task_migrate(p);
1202
	}
I
Ingo Molnar 已提交
1203 1204

	__set_task_cpu(p, new_cpu);
I
Ingo Molnar 已提交
1205 1206
}

1207 1208
static void __migrate_swap_task(struct task_struct *p, int cpu)
{
1209
	if (task_on_rq_queued(p)) {
1210 1211 1212 1213 1214
		struct rq *src_rq, *dst_rq;

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

1215
		p->on_rq = TASK_ON_RQ_MIGRATING;
1216 1217 1218
		deactivate_task(src_rq, p, 0);
		set_task_cpu(p, cpu);
		activate_task(dst_rq, p, 0);
1219
		p->on_rq = TASK_ON_RQ_QUEUED;
1220 1221 1222 1223 1224
		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
1225
		 * previous CPU our target instead of where it really is.
1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241
		 */
		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;

1242 1243 1244
	if (!cpu_active(arg->src_cpu) || !cpu_active(arg->dst_cpu))
		return -EAGAIN;

1245 1246 1247
	src_rq = cpu_rq(arg->src_cpu);
	dst_rq = cpu_rq(arg->dst_cpu);

1248 1249
	double_raw_lock(&arg->src_task->pi_lock,
			&arg->dst_task->pi_lock);
1250
	double_rq_lock(src_rq, dst_rq);
1251

1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270
	if (task_cpu(arg->dst_task) != arg->dst_cpu)
		goto unlock;

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

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

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

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

	ret = 0;

unlock:
	double_rq_unlock(src_rq, dst_rq);
1271 1272
	raw_spin_unlock(&arg->dst_task->pi_lock);
	raw_spin_unlock(&arg->src_task->pi_lock);
1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294

	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;

1295 1296 1297 1298
	/*
	 * 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.
	 */
1299 1300 1301 1302 1303 1304 1305 1306 1307
	if (!cpu_active(arg.src_cpu) || !cpu_active(arg.dst_cpu))
		goto out;

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

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

1308
	trace_sched_swap_numa(cur, arg.src_cpu, p, arg.dst_cpu);
1309 1310 1311 1312 1313 1314
	ret = stop_two_cpus(arg.dst_cpu, arg.src_cpu, migrate_swap_stop, &arg);

out:
	return ret;
}

L
Linus Torvalds 已提交
1315 1316 1317
/*
 * wait_task_inactive - wait for a thread to unschedule.
 *
R
Roland McGrath 已提交
1318 1319 1320 1321 1322 1323 1324
 * 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
1325 1326 1327 1328 1329 1330
 * 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
1331
unsigned long wait_task_inactive(struct task_struct *p, long match_state)
L
Linus Torvalds 已提交
1332
{
1333
	int running, queued;
1334
	struct rq_flags rf;
R
Roland McGrath 已提交
1335
	unsigned long ncsw;
1336
	struct rq *rq;
L
Linus Torvalds 已提交
1337

1338 1339 1340 1341 1342 1343 1344 1345
	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);
1346

1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357
		/*
		 * 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
1358 1359 1360
		while (task_running(rq, p)) {
			if (match_state && unlikely(p->state != match_state))
				return 0;
1361
			cpu_relax();
R
Roland McGrath 已提交
1362
		}
1363

1364 1365 1366 1367 1368
		/*
		 * 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.
		 */
1369
		rq = task_rq_lock(p, &rf);
1370
		trace_sched_wait_task(p);
1371
		running = task_running(rq, p);
1372
		queued = task_on_rq_queued(p);
R
Roland McGrath 已提交
1373
		ncsw = 0;
1374
		if (!match_state || p->state == match_state)
1375
			ncsw = p->nvcsw | LONG_MIN; /* sets MSB */
1376
		task_rq_unlock(rq, p, &rf);
1377

R
Roland McGrath 已提交
1378 1379 1380 1381 1382 1383
		/*
		 * If it changed from the expected state, bail out now.
		 */
		if (unlikely(!ncsw))
			break;

1384 1385 1386 1387 1388 1389 1390 1391 1392 1393
		/*
		 * 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;
		}
1394

1395 1396 1397 1398 1399
		/*
		 * It's not enough that it's not actively running,
		 * it must be off the runqueue _entirely_, and not
		 * preempted!
		 *
1400
		 * So if it was still runnable (but just not actively
1401 1402 1403
		 * running right now), it's preempted, and we should
		 * yield - it could be a while.
		 */
1404
		if (unlikely(queued)) {
T
Thomas Gleixner 已提交
1405
			ktime_t to = NSEC_PER_SEC / HZ;
1406 1407 1408

			set_current_state(TASK_UNINTERRUPTIBLE);
			schedule_hrtimeout(&to, HRTIMER_MODE_REL);
1409 1410
			continue;
		}
1411

1412 1413 1414 1415 1416 1417 1418
		/*
		 * 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
1419 1420

	return ncsw;
L
Linus Torvalds 已提交
1421 1422 1423 1424 1425 1426 1427 1428 1429
}

/***
 * 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
1430
 * NOTE: this function doesn't have to take the runqueue lock,
L
Linus Torvalds 已提交
1431 1432 1433 1434 1435
 * 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.
 */
1436
void kick_process(struct task_struct *p)
L
Linus Torvalds 已提交
1437 1438 1439 1440 1441 1442 1443 1444 1445
{
	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
1446
EXPORT_SYMBOL_GPL(kick_process);
L
Linus Torvalds 已提交
1447

1448
/*
1449
 * ->cpus_allowed is protected by both rq->lock and p->pi_lock
1450 1451 1452 1453 1454 1455 1456
 *
 * 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
1457
 *    CPU isn't yet part of the sched domains, and balancing will not
1458 1459
 *    see it.
 *
I
Ingo Molnar 已提交
1460
 *  - on CPU-down we clear cpu_active() to mask the sched domains and
1461
 *    avoid the load balancer to place new tasks on the to be removed
I
Ingo Molnar 已提交
1462
 *    CPU. Existing tasks will remain running there and will be taken
1463 1464 1465 1466 1467 1468
 *    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.
1469
 */
1470 1471
static int select_fallback_rq(int cpu, struct task_struct *p)
{
1472 1473
	int nid = cpu_to_node(cpu);
	const struct cpumask *nodemask = NULL;
1474 1475
	enum { cpuset, possible, fail } state = cpuset;
	int dest_cpu;
1476

1477
	/*
I
Ingo Molnar 已提交
1478 1479 1480
	 * 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.
1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491
	 */
	if (nid != -1) {
		nodemask = cpumask_of_node(nid);

		/* Look for allowed, online CPU in same node. */
		for_each_cpu(dest_cpu, nodemask) {
			if (!cpu_active(dest_cpu))
				continue;
			if (cpumask_test_cpu(dest_cpu, tsk_cpus_allowed(p)))
				return dest_cpu;
		}
1492
	}
1493

1494 1495
	for (;;) {
		/* Any allowed, online CPU? */
1496
		for_each_cpu(dest_cpu, tsk_cpus_allowed(p)) {
1497 1498 1499
			if (!(p->flags & PF_KTHREAD) && !cpu_active(dest_cpu))
				continue;
			if (!cpu_online(dest_cpu))
1500 1501 1502
				continue;
			goto out;
		}
1503

1504
		/* No more Mr. Nice Guy. */
1505 1506
		switch (state) {
		case cpuset:
1507 1508 1509 1510 1511
			if (IS_ENABLED(CONFIG_CPUSETS)) {
				cpuset_cpus_allowed_fallback(p);
				state = possible;
				break;
			}
I
Ingo Molnar 已提交
1512
			/* Fall-through */
1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531
		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()) {
1532
			printk_deferred("process %d (%s) no longer affine to cpu%d\n",
1533 1534
					task_pid_nr(p), p->comm, cpu);
		}
1535 1536 1537 1538 1539
	}

	return dest_cpu;
}

1540
/*
1541
 * The caller (fork, wakeup) owns p->pi_lock, ->cpus_allowed is stable.
1542
 */
1543
static inline
1544
int select_task_rq(struct task_struct *p, int cpu, int sd_flags, int wake_flags)
1545
{
1546 1547
	lockdep_assert_held(&p->pi_lock);

1548
	if (tsk_nr_cpus_allowed(p) > 1)
1549
		cpu = p->sched_class->select_task_rq(p, cpu, sd_flags, wake_flags);
1550 1551
	else
		cpu = cpumask_any(tsk_cpus_allowed(p));
1552 1553 1554 1555

	/*
	 * 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
1556
	 * CPU.
1557 1558 1559 1560 1561 1562
	 *
	 * 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 ]
	 */
1563
	if (unlikely(!cpumask_test_cpu(cpu, tsk_cpus_allowed(p)) ||
P
Peter Zijlstra 已提交
1564
		     !cpu_online(cpu)))
1565
		cpu = select_fallback_rq(task_cpu(p), p);
1566 1567

	return cpu;
1568
}
1569 1570 1571 1572 1573 1574

static void update_avg(u64 *avg, u64 sample)
{
	s64 diff = sample - *avg;
	*avg += diff >> 3;
}
1575 1576 1577 1578 1579 1580 1581 1582 1583

#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
1584
#endif /* CONFIG_SMP */
1585

P
Peter Zijlstra 已提交
1586
static void
1587
ttwu_stat(struct task_struct *p, int cpu, int wake_flags)
T
Tejun Heo 已提交
1588
{
1589
	struct rq *rq;
1590

1591 1592 1593 1594
	if (!schedstat_enabled())
		return;

	rq = this_rq();
P
Peter Zijlstra 已提交
1595

1596 1597
#ifdef CONFIG_SMP
	if (cpu == rq->cpu) {
1598 1599
		schedstat_inc(rq->ttwu_local);
		schedstat_inc(p->se.statistics.nr_wakeups_local);
P
Peter Zijlstra 已提交
1600 1601 1602
	} else {
		struct sched_domain *sd;

1603
		schedstat_inc(p->se.statistics.nr_wakeups_remote);
1604
		rcu_read_lock();
1605
		for_each_domain(rq->cpu, sd) {
P
Peter Zijlstra 已提交
1606
			if (cpumask_test_cpu(cpu, sched_domain_span(sd))) {
1607
				schedstat_inc(sd->ttwu_wake_remote);
P
Peter Zijlstra 已提交
1608 1609 1610
				break;
			}
		}
1611
		rcu_read_unlock();
P
Peter Zijlstra 已提交
1612
	}
1613 1614

	if (wake_flags & WF_MIGRATED)
1615
		schedstat_inc(p->se.statistics.nr_wakeups_migrate);
P
Peter Zijlstra 已提交
1616 1617
#endif /* CONFIG_SMP */

1618 1619
	schedstat_inc(rq->ttwu_count);
	schedstat_inc(p->se.statistics.nr_wakeups);
P
Peter Zijlstra 已提交
1620 1621

	if (wake_flags & WF_SYNC)
1622
		schedstat_inc(p->se.statistics.nr_wakeups_sync);
P
Peter Zijlstra 已提交
1623 1624
}

1625
static inline void ttwu_activate(struct rq *rq, struct task_struct *p, int en_flags)
P
Peter Zijlstra 已提交
1626
{
T
Tejun Heo 已提交
1627
	activate_task(rq, p, en_flags);
1628
	p->on_rq = TASK_ON_RQ_QUEUED;
1629

I
Ingo Molnar 已提交
1630
	/* If a worker is waking up, notify the workqueue: */
1631 1632
	if (p->flags & PF_WQ_WORKER)
		wq_worker_waking_up(p, cpu_of(rq));
T
Tejun Heo 已提交
1633 1634
}

1635 1636 1637
/*
 * Mark the task runnable and perform wakeup-preemption.
 */
1638
static void ttwu_do_wakeup(struct rq *rq, struct task_struct *p, int wake_flags,
1639
			   struct rq_flags *rf)
T
Tejun Heo 已提交
1640 1641 1642
{
	check_preempt_curr(rq, p, wake_flags);
	p->state = TASK_RUNNING;
1643 1644
	trace_sched_wakeup(p);

T
Tejun Heo 已提交
1645
#ifdef CONFIG_SMP
1646 1647
	if (p->sched_class->task_woken) {
		/*
1648 1649
		 * Our task @p is fully woken up and running; so its safe to
		 * drop the rq->lock, hereafter rq is only used for statistics.
1650
		 */
1651
		rq_unpin_lock(rq, rf);
T
Tejun Heo 已提交
1652
		p->sched_class->task_woken(rq, p);
1653
		rq_repin_lock(rq, rf);
1654
	}
T
Tejun Heo 已提交
1655

1656
	if (rq->idle_stamp) {
1657
		u64 delta = rq_clock(rq) - rq->idle_stamp;
1658
		u64 max = 2*rq->max_idle_balance_cost;
T
Tejun Heo 已提交
1659

1660 1661 1662
		update_avg(&rq->avg_idle, delta);

		if (rq->avg_idle > max)
T
Tejun Heo 已提交
1663
			rq->avg_idle = max;
1664

T
Tejun Heo 已提交
1665 1666 1667 1668 1669
		rq->idle_stamp = 0;
	}
#endif
}

1670
static void
1671
ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags,
1672
		 struct rq_flags *rf)
1673
{
1674 1675
	int en_flags = ENQUEUE_WAKEUP;

1676 1677
	lockdep_assert_held(&rq->lock);

1678 1679 1680
#ifdef CONFIG_SMP
	if (p->sched_contributes_to_load)
		rq->nr_uninterruptible--;
1681 1682

	if (wake_flags & WF_MIGRATED)
1683
		en_flags |= ENQUEUE_MIGRATED;
1684 1685
#endif

1686
	ttwu_activate(rq, p, en_flags);
1687
	ttwu_do_wakeup(rq, p, wake_flags, rf);
1688 1689 1690 1691 1692 1693 1694 1695 1696 1697
}

/*
 * 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)
{
1698
	struct rq_flags rf;
1699 1700 1701
	struct rq *rq;
	int ret = 0;

1702
	rq = __task_rq_lock(p, &rf);
1703
	if (task_on_rq_queued(p)) {
1704 1705
		/* check_preempt_curr() may use rq clock */
		update_rq_clock(rq);
1706
		ttwu_do_wakeup(rq, p, wake_flags, &rf);
1707 1708
		ret = 1;
	}
1709
	__task_rq_unlock(rq, &rf);
1710 1711 1712 1713

	return ret;
}

1714
#ifdef CONFIG_SMP
1715
void sched_ttwu_pending(void)
1716 1717
{
	struct rq *rq = this_rq();
P
Peter Zijlstra 已提交
1718 1719
	struct llist_node *llist = llist_del_all(&rq->wake_list);
	struct task_struct *p;
1720
	unsigned long flags;
1721
	struct rq_flags rf;
1722

1723 1724 1725 1726
	if (!llist)
		return;

	raw_spin_lock_irqsave(&rq->lock, flags);
1727
	rq_pin_lock(rq, &rf);
1728

P
Peter Zijlstra 已提交
1729
	while (llist) {
P
Peter Zijlstra 已提交
1730 1731
		int wake_flags = 0;

P
Peter Zijlstra 已提交
1732 1733
		p = llist_entry(llist, struct task_struct, wake_entry);
		llist = llist_next(llist);
P
Peter Zijlstra 已提交
1734 1735 1736 1737

		if (p->sched_remote_wakeup)
			wake_flags = WF_MIGRATED;

1738
		ttwu_do_activate(rq, p, wake_flags, &rf);
1739 1740
	}

1741
	rq_unpin_lock(rq, &rf);
1742
	raw_spin_unlock_irqrestore(&rq->lock, flags);
1743 1744 1745 1746
}

void scheduler_ipi(void)
{
1747 1748 1749 1750 1751
	/*
	 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
	 * TIF_NEED_RESCHED remotely (for the first time) will also send
	 * this IPI.
	 */
1752
	preempt_fold_need_resched();
1753

1754
	if (llist_empty(&this_rq()->wake_list) && !got_nohz_idle_kick())
1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770
		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
1771
	sched_ttwu_pending();
1772 1773 1774 1775

	/*
	 * Check if someone kicked us for doing the nohz idle load balance.
	 */
1776
	if (unlikely(got_nohz_idle_kick())) {
1777
		this_rq()->idle_balance = 1;
1778
		raise_softirq_irqoff(SCHED_SOFTIRQ);
1779
	}
1780
	irq_exit();
1781 1782
}

P
Peter Zijlstra 已提交
1783
static void ttwu_queue_remote(struct task_struct *p, int cpu, int wake_flags)
1784
{
1785 1786
	struct rq *rq = cpu_rq(cpu);

P
Peter Zijlstra 已提交
1787 1788
	p->sched_remote_wakeup = !!(wake_flags & WF_MIGRATED);

1789 1790 1791 1792 1793 1794
	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);
	}
1795
}
1796

1797 1798 1799 1800 1801
void wake_up_if_idle(int cpu)
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

1802 1803 1804 1805
	rcu_read_lock();

	if (!is_idle_task(rcu_dereference(rq->curr)))
		goto out;
1806 1807 1808 1809 1810 1811 1812

	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
1813
		/* Else CPU is not idle, do nothing here: */
1814 1815
		raw_spin_unlock_irqrestore(&rq->lock, flags);
	}
1816 1817 1818

out:
	rcu_read_unlock();
1819 1820
}

1821
bool cpus_share_cache(int this_cpu, int that_cpu)
1822 1823 1824
{
	return per_cpu(sd_llc_id, this_cpu) == per_cpu(sd_llc_id, that_cpu);
}
1825
#endif /* CONFIG_SMP */
1826

1827
static void ttwu_queue(struct task_struct *p, int cpu, int wake_flags)
1828 1829
{
	struct rq *rq = cpu_rq(cpu);
1830
	struct rq_flags rf;
1831

1832
#if defined(CONFIG_SMP)
1833
	if (sched_feat(TTWU_QUEUE) && !cpus_share_cache(smp_processor_id(), cpu)) {
I
Ingo Molnar 已提交
1834
		sched_clock_cpu(cpu); /* Sync clocks across CPUs */
P
Peter Zijlstra 已提交
1835
		ttwu_queue_remote(p, cpu, wake_flags);
1836 1837 1838 1839
		return;
	}
#endif

1840
	raw_spin_lock(&rq->lock);
1841 1842 1843
	rq_pin_lock(rq, &rf);
	ttwu_do_activate(rq, p, wake_flags, &rf);
	rq_unpin_lock(rq, &rf);
1844
	raw_spin_unlock(&rq->lock);
T
Tejun Heo 已提交
1845 1846
}

1847 1848 1849 1850 1851 1852
/*
 * 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
1853 1854
 * migrates, all its activity on its old CPU [c0] happens-before any subsequent
 * execution on its new CPU [c1].
1855 1856 1857 1858 1859 1860 1861 1862 1863 1864
 *
 * 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
1865
 * Note: the CPU doing B need not be c0 or c1
1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896
 *
 * 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)
1897
 *   2) smp_cond_load_acquire(!X->on_cpu)
1898 1899 1900 1901 1902 1903 1904 1905 1906 1907
 *
 * 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);
 *
1908
 *                    smp_cond_load_acquire(&X->on_cpu, !VAL);
1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933
 *                    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,
1934
 * since the waking CPU is the one issueing the ACQUIRE (smp_cond_load_acquire).
1935 1936 1937
 *
 */

T
Tejun Heo 已提交
1938
/**
L
Linus Torvalds 已提交
1939
 * try_to_wake_up - wake up a thread
T
Tejun Heo 已提交
1940
 * @p: the thread to be awakened
L
Linus Torvalds 已提交
1941
 * @state: the mask of task states that can be woken
T
Tejun Heo 已提交
1942
 * @wake_flags: wake modifier flags (WF_*)
L
Linus Torvalds 已提交
1943
 *
1944
 * If (@state & @p->state) @p->state = TASK_RUNNING.
L
Linus Torvalds 已提交
1945
 *
1946 1947 1948 1949 1950 1951 1952
 * 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
1953
 */
1954 1955
static int
try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
L
Linus Torvalds 已提交
1956 1957
{
	unsigned long flags;
1958
	int cpu, success = 0;
P
Peter Zijlstra 已提交
1959

1960 1961 1962 1963 1964 1965 1966
	/*
	 * 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();
1967
	raw_spin_lock_irqsave(&p->pi_lock, flags);
P
Peter Zijlstra 已提交
1968
	if (!(p->state & state))
L
Linus Torvalds 已提交
1969 1970
		goto out;

1971 1972
	trace_sched_waking(p);

I
Ingo Molnar 已提交
1973 1974
	/* We're going to change ->state: */
	success = 1;
L
Linus Torvalds 已提交
1975 1976
	cpu = task_cpu(p);

1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998
	/*
	 * 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();
1999 2000
	if (p->on_rq && ttwu_remote(p, wake_flags))
		goto stat;
L
Linus Torvalds 已提交
2001 2002

#ifdef CONFIG_SMP
2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021
	/*
	 * 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
2022
	/*
I
Ingo Molnar 已提交
2023
	 * If the owning (remote) CPU is still in the middle of schedule() with
2024
	 * this task as prev, wait until its done referencing the task.
2025 2026 2027 2028 2029
	 *
	 * 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.
2030
	 */
2031
	smp_cond_load_acquire(&p->on_cpu, !VAL);
L
Linus Torvalds 已提交
2032

2033
	p->sched_contributes_to_load = !!task_contributes_to_load(p);
P
Peter Zijlstra 已提交
2034
	p->state = TASK_WAKING;
2035

2036 2037 2038 2039 2040
	if (p->in_iowait) {
		delayacct_blkio_end();
		atomic_dec(&task_rq(p)->nr_iowait);
	}

2041
	cpu = select_task_rq(p, p->wake_cpu, SD_BALANCE_WAKE, wake_flags);
2042 2043
	if (task_cpu(p) != cpu) {
		wake_flags |= WF_MIGRATED;
2044
		set_task_cpu(p, cpu);
2045
	}
2046 2047 2048 2049 2050 2051 2052 2053

#else /* CONFIG_SMP */

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

L
Linus Torvalds 已提交
2054 2055
#endif /* CONFIG_SMP */

2056
	ttwu_queue(p, cpu, wake_flags);
2057
stat:
2058
	ttwu_stat(p, cpu, wake_flags);
L
Linus Torvalds 已提交
2059
out:
2060
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
2061 2062 2063 2064

	return success;
}

T
Tejun Heo 已提交
2065 2066 2067
/**
 * try_to_wake_up_local - try to wake up a local task with rq lock held
 * @p: the thread to be awakened
2068
 * @cookie: context's cookie for pinning
T
Tejun Heo 已提交
2069
 *
2070
 * Put @p on the run-queue if it's not already there. The caller must
T
Tejun Heo 已提交
2071
 * ensure that this_rq() is locked, @p is bound to this_rq() and not
2072
 * the current task.
T
Tejun Heo 已提交
2073
 */
2074
static void try_to_wake_up_local(struct task_struct *p, struct rq_flags *rf)
T
Tejun Heo 已提交
2075 2076 2077
{
	struct rq *rq = task_rq(p);

2078 2079 2080 2081
	if (WARN_ON_ONCE(rq != this_rq()) ||
	    WARN_ON_ONCE(p == current))
		return;

T
Tejun Heo 已提交
2082 2083
	lockdep_assert_held(&rq->lock);

2084
	if (!raw_spin_trylock(&p->pi_lock)) {
2085 2086 2087 2088 2089 2090
		/*
		 * 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.
		 */
2091
		rq_unpin_lock(rq, rf);
2092 2093 2094
		raw_spin_unlock(&rq->lock);
		raw_spin_lock(&p->pi_lock);
		raw_spin_lock(&rq->lock);
2095
		rq_repin_lock(rq, rf);
2096 2097
	}

T
Tejun Heo 已提交
2098
	if (!(p->state & TASK_NORMAL))
2099
		goto out;
T
Tejun Heo 已提交
2100

2101 2102
	trace_sched_waking(p);

2103 2104 2105 2106 2107
	if (!task_on_rq_queued(p)) {
		if (p->in_iowait) {
			delayacct_blkio_end();
			atomic_dec(&rq->nr_iowait);
		}
P
Peter Zijlstra 已提交
2108
		ttwu_activate(rq, p, ENQUEUE_WAKEUP);
2109
	}
P
Peter Zijlstra 已提交
2110

2111
	ttwu_do_wakeup(rq, p, 0, rf);
2112
	ttwu_stat(p, smp_processor_id(), 0);
2113 2114
out:
	raw_spin_unlock(&p->pi_lock);
T
Tejun Heo 已提交
2115 2116
}

2117 2118 2119 2120 2121
/**
 * 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
2122 2123 2124
 * processes.
 *
 * Return: 1 if the process was woken up, 0 if it was already running.
2125 2126 2127 2128
 *
 * 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.
 */
2129
int wake_up_process(struct task_struct *p)
L
Linus Torvalds 已提交
2130
{
2131
	return try_to_wake_up(p, TASK_NORMAL, 0);
L
Linus Torvalds 已提交
2132 2133 2134
}
EXPORT_SYMBOL(wake_up_process);

2135
int wake_up_state(struct task_struct *p, unsigned int state)
L
Linus Torvalds 已提交
2136 2137 2138 2139
{
	return try_to_wake_up(p, state, 0);
}

2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151
/*
 * 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;
2152 2153 2154

	dl_se->dl_throttled = 0;
	dl_se->dl_yielded = 0;
2155 2156
}

L
Linus Torvalds 已提交
2157 2158 2159
/*
 * Perform scheduler related setup for a newly forked process p.
 * p is forked by current.
I
Ingo Molnar 已提交
2160 2161 2162
 *
 * __sched_fork() is basic setup used by init_idle() too:
 */
2163
static void __sched_fork(unsigned long clone_flags, struct task_struct *p)
I
Ingo Molnar 已提交
2164
{
P
Peter Zijlstra 已提交
2165 2166 2167
	p->on_rq			= 0;

	p->se.on_rq			= 0;
I
Ingo Molnar 已提交
2168 2169
	p->se.exec_start		= 0;
	p->se.sum_exec_runtime		= 0;
2170
	p->se.prev_sum_exec_runtime	= 0;
2171
	p->se.nr_migrations		= 0;
P
Peter Zijlstra 已提交
2172
	p->se.vruntime			= 0;
P
Peter Zijlstra 已提交
2173
	INIT_LIST_HEAD(&p->se.group_node);
I
Ingo Molnar 已提交
2174

2175 2176 2177 2178
#ifdef CONFIG_FAIR_GROUP_SCHED
	p->se.cfs_rq			= NULL;
#endif

I
Ingo Molnar 已提交
2179
#ifdef CONFIG_SCHEDSTATS
2180
	/* Even if schedstat is disabled, there should not be garbage */
2181
	memset(&p->se.statistics, 0, sizeof(p->se.statistics));
I
Ingo Molnar 已提交
2182
#endif
N
Nick Piggin 已提交
2183

2184
	RB_CLEAR_NODE(&p->dl.rb_node);
2185
	init_dl_task_timer(&p->dl);
2186
	__dl_clear_params(p);
2187

P
Peter Zijlstra 已提交
2188
	INIT_LIST_HEAD(&p->rt.run_list);
2189 2190 2191 2192
	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
2193

2194 2195 2196
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&p->preempt_notifiers);
#endif
2197 2198 2199

#ifdef CONFIG_NUMA_BALANCING
	if (p->mm && atomic_read(&p->mm->mm_users) == 1) {
2200
		p->mm->numa_next_scan = jiffies + msecs_to_jiffies(sysctl_numa_balancing_scan_delay);
2201 2202 2203
		p->mm->numa_scan_seq = 0;
	}

2204 2205 2206 2207 2208
	if (clone_flags & CLONE_VM)
		p->numa_preferred_nid = current->numa_preferred_nid;
	else
		p->numa_preferred_nid = -1;

2209 2210
	p->node_stamp = 0ULL;
	p->numa_scan_seq = p->mm ? p->mm->numa_scan_seq : 0;
2211
	p->numa_scan_period = sysctl_numa_balancing_scan_delay;
2212
	p->numa_work.next = &p->numa_work;
2213
	p->numa_faults = NULL;
2214 2215
	p->last_task_numa_placement = 0;
	p->last_sum_exec_runtime = 0;
2216 2217

	p->numa_group = NULL;
2218
#endif /* CONFIG_NUMA_BALANCING */
I
Ingo Molnar 已提交
2219 2220
}

2221 2222
DEFINE_STATIC_KEY_FALSE(sched_numa_balancing);

2223
#ifdef CONFIG_NUMA_BALANCING
2224

2225 2226 2227
void set_numabalancing_state(bool enabled)
{
	if (enabled)
2228
		static_branch_enable(&sched_numa_balancing);
2229
	else
2230
		static_branch_disable(&sched_numa_balancing);
2231
}
2232 2233 2234 2235 2236 2237 2238

#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;
2239
	int state = static_branch_likely(&sched_numa_balancing);
2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254

	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
2255

2256 2257
#ifdef CONFIG_SCHEDSTATS

2258
DEFINE_STATIC_KEY_FALSE(sched_schedstats);
2259
static bool __initdata __sched_schedstats = false;
2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282

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;

2283 2284 2285 2286 2287
	/*
	 * 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.
	 */
2288
	if (!strcmp(str, "enable")) {
2289
		__sched_schedstats = true;
2290 2291
		ret = 1;
	} else if (!strcmp(str, "disable")) {
2292
		__sched_schedstats = false;
2293 2294 2295 2296 2297 2298 2299 2300 2301 2302
		ret = 1;
	}
out:
	if (!ret)
		pr_warn("Unable to parse schedstats=\n");

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

2303 2304 2305 2306 2307
static void __init init_schedstats(void)
{
	set_schedstats(__sched_schedstats);
}

2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327
#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;
}
2328 2329 2330 2331
#endif /* CONFIG_PROC_SYSCTL */
#else  /* !CONFIG_SCHEDSTATS */
static inline void init_schedstats(void) {}
#endif /* CONFIG_SCHEDSTATS */
I
Ingo Molnar 已提交
2332 2333 2334 2335

/*
 * fork()/clone()-time setup:
 */
2336
int sched_fork(unsigned long clone_flags, struct task_struct *p)
I
Ingo Molnar 已提交
2337
{
2338
	unsigned long flags;
I
Ingo Molnar 已提交
2339 2340
	int cpu = get_cpu();

2341
	__sched_fork(clone_flags, p);
2342
	/*
2343
	 * We mark the process as NEW here. This guarantees that
2344 2345 2346
	 * nobody will actually run it, and a signal or other external
	 * event cannot wake it up and insert it on the runqueue either.
	 */
2347
	p->state = TASK_NEW;
I
Ingo Molnar 已提交
2348

2349 2350 2351 2352 2353
	/*
	 * Make sure we do not leak PI boosting priority to the child.
	 */
	p->prio = current->normal_prio;

2354 2355 2356 2357
	/*
	 * Revert to default priority/policy on fork if requested.
	 */
	if (unlikely(p->sched_reset_on_fork)) {
2358
		if (task_has_dl_policy(p) || task_has_rt_policy(p)) {
2359
			p->policy = SCHED_NORMAL;
2360
			p->static_prio = NICE_TO_PRIO(0);
2361 2362 2363 2364 2365 2366
			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);
2367

2368 2369 2370 2371 2372 2373
		/*
		 * We don't need the reset flag anymore after the fork. It has
		 * fulfilled its duty:
		 */
		p->sched_reset_on_fork = 0;
	}
2374

2375 2376 2377 2378 2379 2380
	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
2381
		p->sched_class = &fair_sched_class;
2382
	}
2383

2384
	init_entity_runnable_average(&p->se);
P
Peter Zijlstra 已提交
2385

2386 2387 2388 2389 2390 2391 2392
	/*
	 * 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.
	 */
2393
	raw_spin_lock_irqsave(&p->pi_lock, flags);
2394
	/*
I
Ingo Molnar 已提交
2395
	 * We're setting the CPU for the first time, we don't migrate,
2396 2397 2398 2399 2400
	 * so use __set_task_cpu().
	 */
	__set_task_cpu(p, cpu);
	if (p->sched_class->task_fork)
		p->sched_class->task_fork(p);
2401
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
2402

2403
#ifdef CONFIG_SCHED_INFO
I
Ingo Molnar 已提交
2404
	if (likely(sched_info_on()))
2405
		memset(&p->sched_info, 0, sizeof(p->sched_info));
L
Linus Torvalds 已提交
2406
#endif
P
Peter Zijlstra 已提交
2407 2408
#if defined(CONFIG_SMP)
	p->on_cpu = 0;
2409
#endif
2410
	init_task_preempt_count(p);
2411
#ifdef CONFIG_SMP
2412
	plist_node_init(&p->pushable_tasks, MAX_PRIO);
2413
	RB_CLEAR_NODE(&p->pushable_dl_tasks);
2414
#endif
2415

N
Nick Piggin 已提交
2416
	put_cpu();
2417
	return 0;
L
Linus Torvalds 已提交
2418 2419
}

2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438
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)
{
2439 2440
	RCU_LOCKDEP_WARN(!rcu_read_lock_sched_held(),
			 "sched RCU must be held");
2441 2442 2443
	return &cpu_rq(i)->rd->dl_bw;
}

2444
static inline int dl_bw_cpus(int i)
2445
{
2446 2447 2448
	struct root_domain *rd = cpu_rq(i)->rd;
	int cpus = 0;

2449 2450
	RCU_LOCKDEP_WARN(!rcu_read_lock_sched_held(),
			 "sched RCU must be held");
2451 2452 2453 2454
	for_each_cpu_and(i, rd->span, cpu_active_mask)
		cpus++;

	return cpus;
2455 2456 2457 2458 2459 2460 2461
}
#else
inline struct dl_bw *dl_bw_of(int i)
{
	return &cpu_rq(i)->dl.dl_bw;
}

2462
static inline int dl_bw_cpus(int i)
2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474
{
	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.
2475 2476 2477
 *
 * XXX we should delay bw change until the task's 0-lag point, see
 * __setparam_dl().
2478 2479 2480 2481 2482 2483
 */
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));
2484
	u64 period = attr->sched_period ?: attr->sched_deadline;
2485 2486
	u64 runtime = attr->sched_runtime;
	u64 new_bw = dl_policy(policy) ? to_ratio(period, runtime) : 0;
2487
	int cpus, err = -1;
2488

2489 2490
	/* !deadline task may carry old deadline bandwidth */
	if (new_bw == p->dl.dl_bw && task_has_dl_policy(p))
2491 2492 2493 2494 2495 2496 2497 2498
		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);
2499
	cpus = dl_bw_cpus(task_cpu(p));
2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519
	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
2520 2521 2522 2523 2524 2525 2526
/*
 * 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.
 */
2527
void wake_up_new_task(struct task_struct *p)
L
Linus Torvalds 已提交
2528
{
2529
	struct rq_flags rf;
I
Ingo Molnar 已提交
2530
	struct rq *rq;
2531

2532
	raw_spin_lock_irqsave(&p->pi_lock, rf.flags);
2533
	p->state = TASK_RUNNING;
2534 2535 2536 2537
#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
2538
	 *  - any previously selected CPU might disappear through hotplug
2539 2540 2541
	 *
	 * Use __set_task_cpu() to avoid calling sched_class::migrate_task_rq,
	 * as we're not fully set-up yet.
2542
	 */
2543
	__set_task_cpu(p, select_task_rq(p, task_cpu(p), SD_BALANCE_FORK, 0));
2544
#endif
2545
	rq = __task_rq_lock(p, &rf);
2546
	update_rq_clock(rq);
2547
	post_init_entity_util_avg(&p->se);
2548

P
Peter Zijlstra 已提交
2549
	activate_task(rq, p, 0);
2550
	p->on_rq = TASK_ON_RQ_QUEUED;
2551
	trace_sched_wakeup_new(p);
P
Peter Zijlstra 已提交
2552
	check_preempt_curr(rq, p, WF_FORK);
2553
#ifdef CONFIG_SMP
2554 2555 2556 2557 2558
	if (p->sched_class->task_woken) {
		/*
		 * Nothing relies on rq->lock after this, so its fine to
		 * drop it.
		 */
2559
		rq_unpin_lock(rq, &rf);
2560
		p->sched_class->task_woken(rq, p);
2561
		rq_repin_lock(rq, &rf);
2562
	}
2563
#endif
2564
	task_rq_unlock(rq, p, &rf);
L
Linus Torvalds 已提交
2565 2566
}

2567 2568
#ifdef CONFIG_PREEMPT_NOTIFIERS

2569 2570
static struct static_key preempt_notifier_key = STATIC_KEY_INIT_FALSE;

2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582
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);

2583
/**
2584
 * preempt_notifier_register - tell me when current is being preempted & rescheduled
R
Randy Dunlap 已提交
2585
 * @notifier: notifier struct to register
2586 2587 2588
 */
void preempt_notifier_register(struct preempt_notifier *notifier)
{
2589 2590 2591
	if (!static_key_false(&preempt_notifier_key))
		WARN(1, "registering preempt_notifier while notifiers disabled\n");

2592 2593 2594 2595 2596 2597
	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
2598
 * @notifier: notifier struct to unregister
2599
 *
2600
 * This is *not* safe to call from within a preemption notifier.
2601 2602 2603 2604 2605 2606 2607
 */
void preempt_notifier_unregister(struct preempt_notifier *notifier)
{
	hlist_del(&notifier->link);
}
EXPORT_SYMBOL_GPL(preempt_notifier_unregister);

2608
static void __fire_sched_in_preempt_notifiers(struct task_struct *curr)
2609 2610 2611
{
	struct preempt_notifier *notifier;

2612
	hlist_for_each_entry(notifier, &curr->preempt_notifiers, link)
2613 2614 2615
		notifier->ops->sched_in(notifier, raw_smp_processor_id());
}

2616 2617 2618 2619 2620 2621
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);
}

2622
static void
2623 2624
__fire_sched_out_preempt_notifiers(struct task_struct *curr,
				   struct task_struct *next)
2625 2626 2627
{
	struct preempt_notifier *notifier;

2628
	hlist_for_each_entry(notifier, &curr->preempt_notifiers, link)
2629 2630 2631
		notifier->ops->sched_out(notifier, next);
}

2632 2633 2634 2635 2636 2637 2638 2639
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);
}

2640
#else /* !CONFIG_PREEMPT_NOTIFIERS */
2641

2642
static inline void fire_sched_in_preempt_notifiers(struct task_struct *curr)
2643 2644 2645
{
}

2646
static inline void
2647 2648 2649 2650 2651
fire_sched_out_preempt_notifiers(struct task_struct *curr,
				 struct task_struct *next)
{
}

2652
#endif /* CONFIG_PREEMPT_NOTIFIERS */
2653

2654 2655 2656
/**
 * prepare_task_switch - prepare to switch tasks
 * @rq: the runqueue preparing to switch
R
Randy Dunlap 已提交
2657
 * @prev: the current task that is being switched out
2658 2659 2660 2661 2662 2663 2664 2665 2666
 * @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.
 */
2667 2668 2669
static inline void
prepare_task_switch(struct rq *rq, struct task_struct *prev,
		    struct task_struct *next)
2670
{
2671
	sched_info_switch(rq, prev, next);
2672
	perf_event_task_sched_out(prev, next);
2673
	fire_sched_out_preempt_notifiers(prev, next);
2674 2675 2676 2677
	prepare_lock_switch(rq, next);
	prepare_arch_switch(next);
}

L
Linus Torvalds 已提交
2678 2679 2680 2681
/**
 * finish_task_switch - clean up after a task-switch
 * @prev: the thread we just switched away from.
 *
2682 2683 2684 2685
 * 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
2686 2687
 *
 * Note that we may have delayed dropping an mm in context_switch(). If
I
Ingo Molnar 已提交
2688
 * so, we finish that here outside of the runqueue lock. (Doing it
L
Linus Torvalds 已提交
2689 2690
 * with the lock held can cause deadlocks; see schedule() for
 * details.)
2691 2692 2693 2694 2695
 *
 * 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
2696
 */
2697
static struct rq *finish_task_switch(struct task_struct *prev)
L
Linus Torvalds 已提交
2698 2699
	__releases(rq->lock)
{
2700
	struct rq *rq = this_rq();
L
Linus Torvalds 已提交
2701
	struct mm_struct *mm = rq->prev_mm;
O
Oleg Nesterov 已提交
2702
	long prev_state;
L
Linus Torvalds 已提交
2703

2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714
	/*
	 * 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.
	 */
2715 2716 2717 2718
	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);
2719

L
Linus Torvalds 已提交
2720 2721 2722 2723
	rq->prev_mm = NULL;

	/*
	 * A task struct has one reference for the use as "current".
2724
	 * If a task dies, then it sets TASK_DEAD in tsk->state and calls
O
Oleg Nesterov 已提交
2725 2726
	 * schedule one last time. The schedule call will never return, and
	 * the scheduled task must drop that reference.
2727 2728 2729 2730 2731
	 *
	 * 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
2732
	 */
O
Oleg Nesterov 已提交
2733
	prev_state = prev->state;
2734
	vtime_task_switch(prev);
2735
	perf_event_task_sched_in(prev, current);
2736
	finish_lock_switch(rq, prev);
2737
	finish_arch_post_lock_switch();
S
Steven Rostedt 已提交
2738

2739
	fire_sched_in_preempt_notifiers(current);
L
Linus Torvalds 已提交
2740 2741
	if (mm)
		mmdrop(mm);
2742
	if (unlikely(prev_state == TASK_DEAD)) {
2743 2744 2745
		if (prev->sched_class->task_dead)
			prev->sched_class->task_dead(prev);

2746 2747 2748
		/*
		 * Remove function-return probe instances associated with this
		 * task and put them back on the free list.
I
Ingo Molnar 已提交
2749
		 */
2750
		kprobe_flush_task(prev);
2751 2752 2753 2754

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

L
Linus Torvalds 已提交
2755
		put_task_struct(prev);
2756
	}
2757

2758
	tick_nohz_task_switch();
2759
	return rq;
L
Linus Torvalds 已提交
2760 2761
}

2762 2763 2764
#ifdef CONFIG_SMP

/* rq->lock is NOT held, but preemption is disabled */
2765
static void __balance_callback(struct rq *rq)
2766
{
2767 2768 2769
	struct callback_head *head, *next;
	void (*func)(struct rq *rq);
	unsigned long flags;
2770

2771 2772 2773 2774 2775 2776 2777 2778
	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;
2779

2780
		func(rq);
2781
	}
2782 2783 2784 2785 2786 2787 2788
	raw_spin_unlock_irqrestore(&rq->lock, flags);
}

static inline void balance_callback(struct rq *rq)
{
	if (unlikely(rq->balance_callback))
		__balance_callback(rq);
2789 2790 2791
}

#else
2792

2793
static inline void balance_callback(struct rq *rq)
2794
{
L
Linus Torvalds 已提交
2795 2796
}

2797 2798
#endif

L
Linus Torvalds 已提交
2799 2800 2801 2802
/**
 * schedule_tail - first thing a freshly forked thread must call.
 * @prev: the thread we just switched away from.
 */
2803
asmlinkage __visible void schedule_tail(struct task_struct *prev)
L
Linus Torvalds 已提交
2804 2805
	__releases(rq->lock)
{
2806
	struct rq *rq;
2807

2808 2809 2810 2811 2812 2813 2814 2815 2816
	/*
	 * 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).
	 */

2817
	rq = finish_task_switch(prev);
2818
	balance_callback(rq);
2819
	preempt_enable();
2820

L
Linus Torvalds 已提交
2821
	if (current->set_child_tid)
2822
		put_user(task_pid_vnr(current), current->set_child_tid);
L
Linus Torvalds 已提交
2823 2824 2825
}

/*
2826
 * context_switch - switch to the new MM and the new thread's register state.
L
Linus Torvalds 已提交
2827
 */
2828
static __always_inline struct rq *
2829
context_switch(struct rq *rq, struct task_struct *prev,
2830
	       struct task_struct *next, struct rq_flags *rf)
L
Linus Torvalds 已提交
2831
{
I
Ingo Molnar 已提交
2832
	struct mm_struct *mm, *oldmm;
L
Linus Torvalds 已提交
2833

2834
	prepare_task_switch(rq, prev, next);
2835

I
Ingo Molnar 已提交
2836 2837
	mm = next->mm;
	oldmm = prev->active_mm;
2838 2839 2840 2841 2842
	/*
	 * For paravirt, this is coupled with an exit in switch_to to
	 * combine the page table reload and the switch backend into
	 * one hypercall.
	 */
2843
	arch_start_context_switch(prev);
2844

2845
	if (!mm) {
L
Linus Torvalds 已提交
2846 2847 2848 2849
		next->active_mm = oldmm;
		atomic_inc(&oldmm->mm_count);
		enter_lazy_tlb(oldmm, next);
	} else
2850
		switch_mm_irqs_off(oldmm, mm, next);
L
Linus Torvalds 已提交
2851

2852
	if (!prev->mm) {
L
Linus Torvalds 已提交
2853 2854 2855
		prev->active_mm = NULL;
		rq->prev_mm = oldmm;
	}
2856

2857
	rq->clock_update_flags &= ~(RQCF_ACT_SKIP|RQCF_REQ_SKIP);
2858

2859 2860 2861 2862 2863 2864
	/*
	 * 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:
	 */
2865
	rq_unpin_lock(rq, rf);
2866
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
L
Linus Torvalds 已提交
2867 2868 2869

	/* Here we just switch the register state and the stack. */
	switch_to(prev, next, prev);
I
Ingo Molnar 已提交
2870
	barrier();
2871 2872

	return finish_task_switch(prev);
L
Linus Torvalds 已提交
2873 2874 2875
}

/*
2876
 * nr_running and nr_context_switches:
L
Linus Torvalds 已提交
2877 2878
 *
 * externally visible scheduler statistics: current number of runnable
2879
 * threads, total number of context switches performed since bootup.
L
Linus Torvalds 已提交
2880 2881 2882 2883 2884 2885 2886 2887 2888
 */
unsigned long nr_running(void)
{
	unsigned long i, sum = 0;

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

	return sum;
2889
}
L
Linus Torvalds 已提交
2890

2891
/*
I
Ingo Molnar 已提交
2892
 * Check if only the current task is running on the CPU.
2893 2894 2895 2896 2897 2898 2899 2900 2901 2902
 *
 * 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)
2903 2904 2905
 */
bool single_task_running(void)
{
2906
	return raw_rq()->nr_running == 1;
2907 2908 2909
}
EXPORT_SYMBOL(single_task_running);

L
Linus Torvalds 已提交
2910
unsigned long long nr_context_switches(void)
2911
{
2912 2913
	int i;
	unsigned long long sum = 0;
2914

2915
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2916
		sum += cpu_rq(i)->nr_switches;
2917

L
Linus Torvalds 已提交
2918 2919
	return sum;
}
2920

2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950
/*
 * 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
2951 2952 2953
unsigned long nr_iowait(void)
{
	unsigned long i, sum = 0;
2954

2955
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2956
		sum += atomic_read(&cpu_rq(i)->nr_iowait);
2957

L
Linus Torvalds 已提交
2958 2959
	return sum;
}
2960

2961 2962 2963 2964 2965 2966 2967
/*
 * 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.
 */

2968
unsigned long nr_iowait_cpu(int cpu)
2969
{
2970
	struct rq *this = cpu_rq(cpu);
2971 2972
	return atomic_read(&this->nr_iowait);
}
2973

2974 2975
void get_iowait_load(unsigned long *nr_waiters, unsigned long *load)
{
2976 2977 2978
	struct rq *rq = this_rq();
	*nr_waiters = atomic_read(&rq->nr_iowait);
	*load = rq->load.weight;
2979 2980
}

I
Ingo Molnar 已提交
2981
#ifdef CONFIG_SMP
2982

2983
/*
P
Peter Zijlstra 已提交
2984 2985
 * sched_exec - execve() is a valuable balancing opportunity, because at
 * this point the task has the smallest effective memory and cache footprint.
2986
 */
P
Peter Zijlstra 已提交
2987
void sched_exec(void)
2988
{
P
Peter Zijlstra 已提交
2989
	struct task_struct *p = current;
L
Linus Torvalds 已提交
2990
	unsigned long flags;
2991
	int dest_cpu;
2992

2993
	raw_spin_lock_irqsave(&p->pi_lock, flags);
2994
	dest_cpu = p->sched_class->select_task_rq(p, task_cpu(p), SD_BALANCE_EXEC, 0);
2995 2996
	if (dest_cpu == smp_processor_id())
		goto unlock;
P
Peter Zijlstra 已提交
2997

2998
	if (likely(cpu_active(dest_cpu))) {
2999
		struct migration_arg arg = { p, dest_cpu };
3000

3001 3002
		raw_spin_unlock_irqrestore(&p->pi_lock, flags);
		stop_one_cpu(task_cpu(p), migration_cpu_stop, &arg);
L
Linus Torvalds 已提交
3003 3004
		return;
	}
3005
unlock:
3006
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
3007
}
I
Ingo Molnar 已提交
3008

L
Linus Torvalds 已提交
3009 3010 3011
#endif

DEFINE_PER_CPU(struct kernel_stat, kstat);
3012
DEFINE_PER_CPU(struct kernel_cpustat, kernel_cpustat);
L
Linus Torvalds 已提交
3013 3014

EXPORT_PER_CPU_SYMBOL(kstat);
3015
EXPORT_PER_CPU_SYMBOL(kernel_cpustat);
L
Linus Torvalds 已提交
3016

3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033
/*
 * 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);
}

3034 3035 3036 3037 3038 3039 3040
/*
 * 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)
{
3041
	struct rq_flags rf;
3042
	struct rq *rq;
3043
	u64 ns;
3044

3045 3046 3047 3048 3049 3050
#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
3051 3052
	 * 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
3053
	 * indistinguishable from the read occurring a few cycles earlier.
3054 3055
	 * If we see ->on_cpu without ->on_rq, the task is leaving, and has
	 * been accounted, so we're correct here as well.
3056
	 */
3057
	if (!p->on_cpu || !task_on_rq_queued(p))
3058 3059 3060
		return p->se.sum_exec_runtime;
#endif

3061
	rq = task_rq_lock(p, &rf);
3062 3063 3064 3065 3066 3067
	/*
	 * 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)) {
3068
		prefetch_curr_exec_start(p);
3069 3070 3071 3072
		update_rq_clock(rq);
		p->sched_class->update_curr(rq);
	}
	ns = p->se.sum_exec_runtime;
3073
	task_rq_unlock(rq, p, &rf);
3074 3075 3076

	return ns;
}
3077

3078 3079 3080 3081 3082 3083 3084 3085
/*
 * 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
3086
	struct task_struct *curr = rq->curr;
3087 3088

	sched_clock_tick();
I
Ingo Molnar 已提交
3089

3090
	raw_spin_lock(&rq->lock);
3091
	update_rq_clock(rq);
P
Peter Zijlstra 已提交
3092
	curr->sched_class->task_tick(rq, curr, 0);
3093
	cpu_load_update_active(rq);
3094
	calc_global_load_tick(rq);
3095
	raw_spin_unlock(&rq->lock);
3096

3097
	perf_event_task_tick();
3098

3099
#ifdef CONFIG_SMP
3100
	rq->idle_balance = idle_cpu(cpu);
3101
	trigger_load_balance(rq);
3102
#endif
3103
	rq_last_tick_reset(rq);
L
Linus Torvalds 已提交
3104 3105
}

3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116
#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.
3117 3118
 *
 * Return: Maximum deferment in nanoseconds.
3119 3120 3121 3122
 */
u64 scheduler_tick_max_deferment(void)
{
	struct rq *rq = this_rq();
3123
	unsigned long next, now = READ_ONCE(jiffies);
3124 3125 3126 3127 3128 3129

	next = rq->last_sched_tick + HZ;

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

3130
	return jiffies_to_nsecs(next - now);
L
Linus Torvalds 已提交
3131
}
3132
#endif
L
Linus Torvalds 已提交
3133

3134 3135
#if defined(CONFIG_PREEMPT) && (defined(CONFIG_DEBUG_PREEMPT) || \
				defined(CONFIG_PREEMPT_TRACER))
3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149
/*
 * 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);
	}
}
3150

3151
void preempt_count_add(int val)
L
Linus Torvalds 已提交
3152
{
3153
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3154 3155 3156
	/*
	 * Underflow?
	 */
3157 3158
	if (DEBUG_LOCKS_WARN_ON((preempt_count() < 0)))
		return;
3159
#endif
3160
	__preempt_count_add(val);
3161
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3162 3163 3164
	/*
	 * Spinlock count overflowing soon?
	 */
3165 3166
	DEBUG_LOCKS_WARN_ON((preempt_count() & PREEMPT_MASK) >=
				PREEMPT_MASK - 10);
3167
#endif
3168
	preempt_latency_start(val);
L
Linus Torvalds 已提交
3169
}
3170
EXPORT_SYMBOL(preempt_count_add);
3171
NOKPROBE_SYMBOL(preempt_count_add);
L
Linus Torvalds 已提交
3172

3173 3174 3175 3176 3177 3178 3179 3180 3181 3182
/*
 * 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());
}

3183
void preempt_count_sub(int val)
L
Linus Torvalds 已提交
3184
{
3185
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3186 3187 3188
	/*
	 * Underflow?
	 */
3189
	if (DEBUG_LOCKS_WARN_ON(val > preempt_count()))
3190
		return;
L
Linus Torvalds 已提交
3191 3192 3193
	/*
	 * Is the spinlock portion underflowing?
	 */
3194 3195 3196
	if (DEBUG_LOCKS_WARN_ON((val < PREEMPT_MASK) &&
			!(preempt_count() & PREEMPT_MASK)))
		return;
3197
#endif
3198

3199
	preempt_latency_stop(val);
3200
	__preempt_count_sub(val);
L
Linus Torvalds 已提交
3201
}
3202
EXPORT_SYMBOL(preempt_count_sub);
3203
NOKPROBE_SYMBOL(preempt_count_sub);
L
Linus Torvalds 已提交
3204

3205 3206 3207
#else
static inline void preempt_latency_start(int val) { }
static inline void preempt_latency_stop(int val) { }
L
Linus Torvalds 已提交
3208 3209 3210
#endif

/*
I
Ingo Molnar 已提交
3211
 * Print scheduling while atomic bug:
L
Linus Torvalds 已提交
3212
 */
I
Ingo Molnar 已提交
3213
static noinline void __schedule_bug(struct task_struct *prev)
L
Linus Torvalds 已提交
3214
{
3215 3216 3217
	/* Save this before calling printk(), since that will clobber it */
	unsigned long preempt_disable_ip = get_preempt_disable_ip(current);

3218 3219 3220
	if (oops_in_progress)
		return;

P
Peter Zijlstra 已提交
3221 3222
	printk(KERN_ERR "BUG: scheduling while atomic: %s/%d/0x%08x\n",
		prev->comm, prev->pid, preempt_count());
3223

I
Ingo Molnar 已提交
3224
	debug_show_held_locks(prev);
3225
	print_modules();
I
Ingo Molnar 已提交
3226 3227
	if (irqs_disabled())
		print_irqtrace_events(prev);
3228 3229
	if (IS_ENABLED(CONFIG_DEBUG_PREEMPT)
	    && in_atomic_preempt_off()) {
3230
		pr_err("Preemption disabled at:");
3231
		print_ip_sym(preempt_disable_ip);
3232 3233
		pr_cont("\n");
	}
3234 3235 3236
	if (panic_on_warn)
		panic("scheduling while atomic\n");

3237
	dump_stack();
3238
	add_taint(TAINT_WARN, LOCKDEP_STILL_OK);
I
Ingo Molnar 已提交
3239
}
L
Linus Torvalds 已提交
3240

I
Ingo Molnar 已提交
3241 3242 3243 3244 3245
/*
 * Various schedule()-time debugging checks and statistics:
 */
static inline void schedule_debug(struct task_struct *prev)
{
3246
#ifdef CONFIG_SCHED_STACK_END_CHECK
J
Jann Horn 已提交
3247 3248
	if (task_stack_end_corrupted(prev))
		panic("corrupted stack end detected inside scheduler\n");
3249
#endif
3250

3251
	if (unlikely(in_atomic_preempt_off())) {
I
Ingo Molnar 已提交
3252
		__schedule_bug(prev);
3253 3254
		preempt_count_set(PREEMPT_DISABLED);
	}
3255
	rcu_sleep_check();
I
Ingo Molnar 已提交
3256

L
Linus Torvalds 已提交
3257 3258
	profile_hit(SCHED_PROFILING, __builtin_return_address(0));

3259
	schedstat_inc(this_rq()->sched_count);
I
Ingo Molnar 已提交
3260 3261 3262 3263 3264 3265
}

/*
 * Pick up the highest-prio task:
 */
static inline struct task_struct *
3266
pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
I
Ingo Molnar 已提交
3267
{
3268
	const struct sched_class *class;
I
Ingo Molnar 已提交
3269
	struct task_struct *p;
L
Linus Torvalds 已提交
3270 3271

	/*
I
Ingo Molnar 已提交
3272 3273
	 * Optimization: we know that if all tasks are in
	 * the fair class we can call that function directly:
L
Linus Torvalds 已提交
3274
	 */
3275
	if (likely(rq->nr_running == rq->cfs.h_nr_running)) {
3276
		p = fair_sched_class.pick_next_task(rq, prev, rf);
3277 3278 3279
		if (unlikely(p == RETRY_TASK))
			goto again;

I
Ingo Molnar 已提交
3280
		/* Assumes fair_sched_class->next == idle_sched_class */
3281
		if (unlikely(!p))
3282
			p = idle_sched_class.pick_next_task(rq, prev, rf);
3283 3284

		return p;
L
Linus Torvalds 已提交
3285 3286
	}

3287
again:
3288
	for_each_class(class) {
3289
		p = class->pick_next_task(rq, prev, rf);
3290 3291 3292
		if (p) {
			if (unlikely(p == RETRY_TASK))
				goto again;
I
Ingo Molnar 已提交
3293
			return p;
3294
		}
I
Ingo Molnar 已提交
3295
	}
3296

I
Ingo Molnar 已提交
3297 3298
	/* The idle class should always have a runnable task: */
	BUG();
I
Ingo Molnar 已提交
3299
}
L
Linus Torvalds 已提交
3300

I
Ingo Molnar 已提交
3301
/*
3302
 * __schedule() is the main scheduler function.
3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336
 *
 * 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
3337
 *
3338
 * WARNING: must be called with preemption disabled!
I
Ingo Molnar 已提交
3339
 */
3340
static void __sched notrace __schedule(bool preempt)
I
Ingo Molnar 已提交
3341 3342
{
	struct task_struct *prev, *next;
3343
	unsigned long *switch_count;
3344
	struct rq_flags rf;
I
Ingo Molnar 已提交
3345
	struct rq *rq;
3346
	int cpu;
I
Ingo Molnar 已提交
3347 3348 3349 3350 3351 3352

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

	schedule_debug(prev);
L
Linus Torvalds 已提交
3353

3354
	if (sched_feat(HRTICK))
M
Mike Galbraith 已提交
3355
		hrtick_clear(rq);
P
Peter Zijlstra 已提交
3356

3357 3358 3359
	local_irq_disable();
	rcu_note_context_switch();

3360 3361 3362 3363 3364 3365
	/*
	 * 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();
3366
	raw_spin_lock(&rq->lock);
3367
	rq_pin_lock(rq, &rf);
L
Linus Torvalds 已提交
3368

I
Ingo Molnar 已提交
3369 3370
	/* Promote REQ to ACT */
	rq->clock_update_flags <<= 1;
3371

3372
	switch_count = &prev->nivcsw;
3373
	if (!preempt && prev->state) {
T
Tejun Heo 已提交
3374
		if (unlikely(signal_pending_state(prev->state, prev))) {
L
Linus Torvalds 已提交
3375
			prev->state = TASK_RUNNING;
T
Tejun Heo 已提交
3376
		} else {
3377 3378 3379
			deactivate_task(rq, prev, DEQUEUE_SLEEP);
			prev->on_rq = 0;

3380 3381 3382 3383 3384
			if (prev->in_iowait) {
				atomic_inc(&rq->nr_iowait);
				delayacct_blkio_start();
			}

T
Tejun Heo 已提交
3385
			/*
3386 3387 3388
			 * 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
3389 3390 3391 3392
			 */
			if (prev->flags & PF_WQ_WORKER) {
				struct task_struct *to_wakeup;

3393
				to_wakeup = wq_worker_sleeping(prev);
T
Tejun Heo 已提交
3394
				if (to_wakeup)
3395
					try_to_wake_up_local(to_wakeup, &rf);
T
Tejun Heo 已提交
3396 3397
			}
		}
I
Ingo Molnar 已提交
3398
		switch_count = &prev->nvcsw;
L
Linus Torvalds 已提交
3399 3400
	}

3401
	if (task_on_rq_queued(prev))
3402 3403
		update_rq_clock(rq);

3404
	next = pick_next_task(rq, prev, &rf);
3405
	clear_tsk_need_resched(prev);
3406
	clear_preempt_need_resched();
L
Linus Torvalds 已提交
3407 3408 3409 3410 3411 3412

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

3413
		trace_sched_switch(preempt, prev, next);
I
Ingo Molnar 已提交
3414 3415 3416

		/* Also unlocks the rq: */
		rq = context_switch(rq, prev, next, &rf);
3417
	} else {
3418
		rq->clock_update_flags &= ~(RQCF_ACT_SKIP|RQCF_REQ_SKIP);
3419
		rq_unpin_lock(rq, &rf);
3420
		raw_spin_unlock_irq(&rq->lock);
3421
	}
L
Linus Torvalds 已提交
3422

3423
	balance_callback(rq);
L
Linus Torvalds 已提交
3424
}
3425

3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442
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
3443
	/* Causes final put_task_struct in finish_task_switch(): */
3444
	__set_current_state(TASK_DEAD);
I
Ingo Molnar 已提交
3445 3446 3447 3448

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

3449 3450
	__schedule(false);
	BUG();
I
Ingo Molnar 已提交
3451 3452

	/* Avoid "noreturn function does return" - but don't continue if BUG() is a NOP: */
3453
	for (;;)
I
Ingo Molnar 已提交
3454
		cpu_relax();
3455 3456
}

3457 3458
static inline void sched_submit_work(struct task_struct *tsk)
{
3459
	if (!tsk->state || tsk_is_pi_blocked(tsk))
3460 3461 3462 3463 3464 3465 3466 3467 3468
		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);
}

3469
asmlinkage __visible void __sched schedule(void)
3470
{
3471 3472 3473
	struct task_struct *tsk = current;

	sched_submit_work(tsk);
3474
	do {
3475
		preempt_disable();
3476
		__schedule(false);
3477
		sched_preempt_enable_no_resched();
3478
	} while (need_resched());
3479
}
L
Linus Torvalds 已提交
3480 3481
EXPORT_SYMBOL(schedule);

3482
#ifdef CONFIG_CONTEXT_TRACKING
3483
asmlinkage __visible void __sched schedule_user(void)
3484 3485 3486 3487 3488 3489
{
	/*
	 * 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.
3490 3491
	 *
	 * NB: There are buggy callers of this function.  Ideally we
3492
	 * should warn if prev_state != CONTEXT_USER, but that will trigger
3493
	 * too frequently to make sense yet.
3494
	 */
3495
	enum ctx_state prev_state = exception_enter();
3496
	schedule();
3497
	exception_exit(prev_state);
3498 3499 3500
}
#endif

3501 3502 3503 3504 3505 3506 3507
/**
 * schedule_preempt_disabled - called with preemption disabled
 *
 * Returns with preemption disabled. Note: preempt_count must be 1
 */
void __sched schedule_preempt_disabled(void)
{
3508
	sched_preempt_enable_no_resched();
3509 3510 3511 3512
	schedule();
	preempt_disable();
}

3513
static void __sched notrace preempt_schedule_common(void)
3514 3515
{
	do {
3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528
		/*
		 * 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.
		 */
3529
		preempt_disable_notrace();
3530
		preempt_latency_start(1);
3531
		__schedule(true);
3532
		preempt_latency_stop(1);
3533
		preempt_enable_no_resched_notrace();
3534 3535 3536 3537 3538 3539 3540 3541

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

L
Linus Torvalds 已提交
3542 3543
#ifdef CONFIG_PREEMPT
/*
3544
 * this is the entry point to schedule() from in-kernel preemption
I
Ingo Molnar 已提交
3545
 * off of preempt_enable. Kernel preemptions off return from interrupt
L
Linus Torvalds 已提交
3546 3547
 * occur there and call schedule directly.
 */
3548
asmlinkage __visible void __sched notrace preempt_schedule(void)
L
Linus Torvalds 已提交
3549 3550 3551
{
	/*
	 * If there is a non-zero preempt_count or interrupts are disabled,
I
Ingo Molnar 已提交
3552
	 * we do not want to preempt the current task. Just return..
L
Linus Torvalds 已提交
3553
	 */
3554
	if (likely(!preemptible()))
L
Linus Torvalds 已提交
3555 3556
		return;

3557
	preempt_schedule_common();
L
Linus Torvalds 已提交
3558
}
3559
NOKPROBE_SYMBOL(preempt_schedule);
L
Linus Torvalds 已提交
3560
EXPORT_SYMBOL(preempt_schedule);
3561 3562

/**
3563
 * preempt_schedule_notrace - preempt_schedule called by tracing
3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575
 *
 * 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.
 */
3576
asmlinkage __visible void __sched notrace preempt_schedule_notrace(void)
3577 3578 3579 3580 3581 3582 3583
{
	enum ctx_state prev_ctx;

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

	do {
3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596
		/*
		 * 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.
		 */
3597
		preempt_disable_notrace();
3598
		preempt_latency_start(1);
3599 3600 3601 3602 3603 3604
		/*
		 * Needs preempt disabled in case user_exit() is traced
		 * and the tracer calls preempt_enable_notrace() causing
		 * an infinite recursion.
		 */
		prev_ctx = exception_enter();
3605
		__schedule(true);
3606 3607
		exception_exit(prev_ctx);

3608
		preempt_latency_stop(1);
3609
		preempt_enable_no_resched_notrace();
3610 3611
	} while (need_resched());
}
3612
EXPORT_SYMBOL_GPL(preempt_schedule_notrace);
3613

3614
#endif /* CONFIG_PREEMPT */
L
Linus Torvalds 已提交
3615 3616

/*
3617
 * this is the entry point to schedule() from kernel preemption
L
Linus Torvalds 已提交
3618 3619 3620 3621
 * off of irq context.
 * Note, that this is called and return with irqs disabled. This will
 * protect us against recursive calling from irq.
 */
3622
asmlinkage __visible void __sched preempt_schedule_irq(void)
L
Linus Torvalds 已提交
3623
{
3624
	enum ctx_state prev_state;
3625

3626
	/* Catch callers which need to be fixed */
3627
	BUG_ON(preempt_count() || !irqs_disabled());
L
Linus Torvalds 已提交
3628

3629 3630
	prev_state = exception_enter();

3631
	do {
3632
		preempt_disable();
3633
		local_irq_enable();
3634
		__schedule(true);
3635
		local_irq_disable();
3636
		sched_preempt_enable_no_resched();
3637
	} while (need_resched());
3638 3639

	exception_exit(prev_state);
L
Linus Torvalds 已提交
3640 3641
}

P
Peter Zijlstra 已提交
3642
int default_wake_function(wait_queue_t *curr, unsigned mode, int wake_flags,
I
Ingo Molnar 已提交
3643
			  void *key)
L
Linus Torvalds 已提交
3644
{
P
Peter Zijlstra 已提交
3645
	return try_to_wake_up(curr->private, mode, wake_flags);
L
Linus Torvalds 已提交
3646 3647 3648
}
EXPORT_SYMBOL(default_wake_function);

3649 3650 3651 3652 3653 3654 3655 3656 3657 3658
#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().
 *
3659 3660
 * Used by the rt_mutex code to implement priority inheritance
 * logic. Call site only calls if the priority of the task changed.
3661
 */
3662
void rt_mutex_setprio(struct task_struct *p, int prio)
3663
{
3664
	int oldprio, queued, running, queue_flag = DEQUEUE_SAVE | DEQUEUE_MOVE;
3665
	const struct sched_class *prev_class;
3666 3667
	struct rq_flags rf;
	struct rq *rq;
3668

3669
	BUG_ON(prio > MAX_PRIO);
3670

3671
	rq = __task_rq_lock(p, &rf);
3672
	update_rq_clock(rq);
3673

3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691
	/*
	 * 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;
	}

3692
	trace_sched_pi_setprio(p, prio);
3693
	oldprio = p->prio;
3694 3695 3696 3697

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

3698
	prev_class = p->sched_class;
3699
	queued = task_on_rq_queued(p);
3700
	running = task_current(rq, p);
3701
	if (queued)
3702
		dequeue_task(rq, p, queue_flag);
3703
	if (running)
3704
		put_prev_task(rq, p);
I
Ingo Molnar 已提交
3705

3706 3707 3708 3709 3710 3711 3712 3713 3714 3715
	/*
	 * 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)) {
3716 3717 3718
		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))) {
3719
			p->dl.dl_boosted = 1;
3720
			queue_flag |= ENQUEUE_REPLENISH;
3721 3722
		} else
			p->dl.dl_boosted = 0;
3723
		p->sched_class = &dl_sched_class;
3724 3725 3726 3727
	} else if (rt_prio(prio)) {
		if (dl_prio(oldprio))
			p->dl.dl_boosted = 0;
		if (oldprio < prio)
3728
			queue_flag |= ENQUEUE_HEAD;
I
Ingo Molnar 已提交
3729
		p->sched_class = &rt_sched_class;
3730 3731 3732
	} else {
		if (dl_prio(oldprio))
			p->dl.dl_boosted = 0;
3733 3734
		if (rt_prio(oldprio))
			p->rt.timeout = 0;
I
Ingo Molnar 已提交
3735
		p->sched_class = &fair_sched_class;
3736
	}
I
Ingo Molnar 已提交
3737

3738 3739
	p->prio = prio;

3740
	if (queued)
3741
		enqueue_task(rq, p, queue_flag);
3742
	if (running)
3743
		set_curr_task(rq, p);
3744

P
Peter Zijlstra 已提交
3745
	check_class_changed(rq, p, prev_class, oldprio);
3746
out_unlock:
I
Ingo Molnar 已提交
3747 3748
	/* Avoid rq from going away on us: */
	preempt_disable();
3749
	__task_rq_unlock(rq, &rf);
3750 3751 3752

	balance_callback(rq);
	preempt_enable();
3753 3754
}
#endif
3755

3756
void set_user_nice(struct task_struct *p, long nice)
L
Linus Torvalds 已提交
3757
{
P
Peter Zijlstra 已提交
3758 3759
	bool queued, running;
	int old_prio, delta;
3760
	struct rq_flags rf;
3761
	struct rq *rq;
L
Linus Torvalds 已提交
3762

3763
	if (task_nice(p) == nice || nice < MIN_NICE || nice > MAX_NICE)
L
Linus Torvalds 已提交
3764 3765 3766 3767 3768
		return;
	/*
	 * We have to be careful, if called from sys_setpriority(),
	 * the task might be in the middle of scheduling on another CPU.
	 */
3769
	rq = task_rq_lock(p, &rf);
3770 3771
	update_rq_clock(rq);

L
Linus Torvalds 已提交
3772 3773 3774 3775
	/*
	 * 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
3776
	 * SCHED_DEADLINE, SCHED_FIFO or SCHED_RR:
L
Linus Torvalds 已提交
3777
	 */
3778
	if (task_has_dl_policy(p) || task_has_rt_policy(p)) {
L
Linus Torvalds 已提交
3779 3780 3781
		p->static_prio = NICE_TO_PRIO(nice);
		goto out_unlock;
	}
3782
	queued = task_on_rq_queued(p);
P
Peter Zijlstra 已提交
3783
	running = task_current(rq, p);
3784
	if (queued)
3785
		dequeue_task(rq, p, DEQUEUE_SAVE);
P
Peter Zijlstra 已提交
3786 3787
	if (running)
		put_prev_task(rq, p);
L
Linus Torvalds 已提交
3788 3789

	p->static_prio = NICE_TO_PRIO(nice);
3790
	set_load_weight(p);
3791 3792 3793
	old_prio = p->prio;
	p->prio = effective_prio(p);
	delta = p->prio - old_prio;
L
Linus Torvalds 已提交
3794

3795
	if (queued) {
3796
		enqueue_task(rq, p, ENQUEUE_RESTORE);
L
Linus Torvalds 已提交
3797
		/*
3798 3799
		 * If the task increased its priority or is running and
		 * lowered its priority, then reschedule its CPU:
L
Linus Torvalds 已提交
3800
		 */
3801
		if (delta < 0 || (delta > 0 && task_running(rq, p)))
3802
			resched_curr(rq);
L
Linus Torvalds 已提交
3803
	}
P
Peter Zijlstra 已提交
3804 3805
	if (running)
		set_curr_task(rq, p);
L
Linus Torvalds 已提交
3806
out_unlock:
3807
	task_rq_unlock(rq, p, &rf);
L
Linus Torvalds 已提交
3808 3809 3810
}
EXPORT_SYMBOL(set_user_nice);

M
Matt Mackall 已提交
3811 3812 3813 3814 3815
/*
 * can_nice - check if a task can reduce its nice value
 * @p: task
 * @nice: nice value
 */
3816
int can_nice(const struct task_struct *p, const int nice)
M
Matt Mackall 已提交
3817
{
I
Ingo Molnar 已提交
3818
	/* Convert nice value [19,-20] to rlimit style value [1,40]: */
3819
	int nice_rlim = nice_to_rlimit(nice);
3820

3821
	return (nice_rlim <= task_rlimit(p, RLIMIT_NICE) ||
M
Matt Mackall 已提交
3822 3823 3824
		capable(CAP_SYS_NICE));
}

L
Linus Torvalds 已提交
3825 3826 3827 3828 3829 3830 3831 3832 3833
#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.
 */
3834
SYSCALL_DEFINE1(nice, int, increment)
L
Linus Torvalds 已提交
3835
{
3836
	long nice, retval;
L
Linus Torvalds 已提交
3837 3838 3839 3840 3841 3842

	/*
	 * 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.
	 */
3843
	increment = clamp(increment, -NICE_WIDTH, NICE_WIDTH);
3844
	nice = task_nice(current) + increment;
L
Linus Torvalds 已提交
3845

3846
	nice = clamp_val(nice, MIN_NICE, MAX_NICE);
M
Matt Mackall 已提交
3847 3848 3849
	if (increment < 0 && !can_nice(current, nice))
		return -EPERM;

L
Linus Torvalds 已提交
3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862 3863
	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.
 *
3864
 * Return: The priority value as seen by users in /proc.
L
Linus Torvalds 已提交
3865 3866 3867
 * RT tasks are offset by -200. Normal tasks are centered
 * around 0, value goes from -16 to +15.
 */
3868
int task_prio(const struct task_struct *p)
L
Linus Torvalds 已提交
3869 3870 3871 3872 3873
{
	return p->prio - MAX_RT_PRIO;
}

/**
I
Ingo Molnar 已提交
3874
 * idle_cpu - is a given CPU idle currently?
L
Linus Torvalds 已提交
3875
 * @cpu: the processor in question.
3876 3877
 *
 * Return: 1 if the CPU is currently idle. 0 otherwise.
L
Linus Torvalds 已提交
3878 3879 3880
 */
int idle_cpu(int cpu)
{
T
Thomas Gleixner 已提交
3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894
	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
3895 3896 3897
}

/**
I
Ingo Molnar 已提交
3898
 * idle_task - return the idle task for a given CPU.
L
Linus Torvalds 已提交
3899
 * @cpu: the processor in question.
3900
 *
I
Ingo Molnar 已提交
3901
 * Return: The idle task for the CPU @cpu.
L
Linus Torvalds 已提交
3902
 */
3903
struct task_struct *idle_task(int cpu)
L
Linus Torvalds 已提交
3904 3905 3906 3907 3908 3909 3910
{
	return cpu_rq(cpu)->idle;
}

/**
 * find_process_by_pid - find a process with a matching PID value.
 * @pid: the pid in question.
3911 3912
 *
 * The task of @pid, if found. %NULL otherwise.
L
Linus Torvalds 已提交
3913
 */
A
Alexey Dobriyan 已提交
3914
static struct task_struct *find_process_by_pid(pid_t pid)
L
Linus Torvalds 已提交
3915
{
3916
	return pid ? find_task_by_vpid(pid) : current;
L
Linus Torvalds 已提交
3917 3918
}

3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933
/*
 * 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;
3934
	dl_se->dl_period = attr->sched_period ?: dl_se->dl_deadline;
3935
	dl_se->flags = attr->sched_flags;
3936
	dl_se->dl_bw = to_ratio(dl_se->dl_period, dl_se->dl_runtime);
3937 3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956

	/*
	 * 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.
	 */
3957 3958
}

3959 3960 3961 3962 3963 3964
/*
 * sched_setparam() passes in -1 for its policy, to let the functions
 * it calls know not to change it.
 */
#define SETPARAM_POLICY	-1

3965 3966
static void __setscheduler_params(struct task_struct *p,
		const struct sched_attr *attr)
L
Linus Torvalds 已提交
3967
{
3968 3969
	int policy = attr->sched_policy;

3970
	if (policy == SETPARAM_POLICY)
3971 3972
		policy = p->policy;

L
Linus Torvalds 已提交
3973
	p->policy = policy;
3974

3975 3976
	if (dl_policy(policy))
		__setparam_dl(p, attr);
3977
	else if (fair_policy(policy))
3978 3979
		p->static_prio = NICE_TO_PRIO(attr->sched_nice);

3980 3981 3982 3983 3984 3985
	/*
	 * __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;
3986
	p->normal_prio = normal_prio(p);
3987 3988
	set_load_weight(p);
}
3989

3990 3991
/* Actually do priority change: must hold pi & rq lock. */
static void __setscheduler(struct rq *rq, struct task_struct *p,
3992
			   const struct sched_attr *attr, bool keep_boost)
3993 3994
{
	__setscheduler_params(p, attr);
3995

3996
	/*
3997 3998
	 * Keep a potential priority boosting if called from
	 * sched_setscheduler().
3999
	 */
4000 4001 4002 4003
	if (keep_boost)
		p->prio = rt_mutex_get_effective_prio(p, normal_prio(p));
	else
		p->prio = normal_prio(p);
4004

4005 4006 4007
	if (dl_prio(p->prio))
		p->sched_class = &dl_sched_class;
	else if (rt_prio(p->prio))
4008 4009 4010
		p->sched_class = &rt_sched_class;
	else
		p->sched_class = &fair_sched_class;
L
Linus Torvalds 已提交
4011
}
4012 4013 4014 4015 4016 4017 4018 4019 4020

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;
4021
	attr->sched_period = dl_se->dl_period;
4022 4023 4024 4025 4026 4027
	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
4028
 * than the runtime, as well as the period of being zero or
4029
 * greater than deadline. Furthermore, we have to be sure that
4030 4031 4032 4033
 * 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).
4034 4035 4036 4037
 */
static bool
__checkparam_dl(const struct sched_attr *attr)
{
4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063
	/* 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;
4064 4065
}

4066
/*
I
Ingo Molnar 已提交
4067
 * Check the target process has a UID that matches the current process's:
4068 4069 4070 4071 4072 4073 4074 4075
 */
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);
4076 4077
	match = (uid_eq(cred->euid, pcred->euid) ||
		 uid_eq(cred->euid, pcred->uid));
4078 4079 4080 4081
	rcu_read_unlock();
	return match;
}

I
Ingo Molnar 已提交
4082
static bool dl_param_changed(struct task_struct *p, const struct sched_attr *attr)
4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094
{
	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;
}

4095 4096
static int __sched_setscheduler(struct task_struct *p,
				const struct sched_attr *attr,
4097
				bool user, bool pi)
L
Linus Torvalds 已提交
4098
{
4099 4100
	int newprio = dl_policy(attr->sched_policy) ? MAX_DL_PRIO - 1 :
		      MAX_RT_PRIO - 1 - attr->sched_priority;
4101
	int retval, oldprio, oldpolicy = -1, queued, running;
4102
	int new_effective_prio, policy = attr->sched_policy;
4103
	const struct sched_class *prev_class;
4104
	struct rq_flags rf;
4105
	int reset_on_fork;
4106
	int queue_flags = DEQUEUE_SAVE | DEQUEUE_MOVE;
4107
	struct rq *rq;
L
Linus Torvalds 已提交
4108

I
Ingo Molnar 已提交
4109
	/* May grab non-irq protected spin_locks: */
4110
	BUG_ON(in_interrupt());
L
Linus Torvalds 已提交
4111
recheck:
I
Ingo Molnar 已提交
4112
	/* Double check policy once rq lock held: */
4113 4114
	if (policy < 0) {
		reset_on_fork = p->sched_reset_on_fork;
L
Linus Torvalds 已提交
4115
		policy = oldpolicy = p->policy;
4116
	} else {
4117
		reset_on_fork = !!(attr->sched_flags & SCHED_FLAG_RESET_ON_FORK);
4118

4119
		if (!valid_policy(policy))
4120 4121 4122
			return -EINVAL;
	}

4123 4124 4125
	if (attr->sched_flags & ~(SCHED_FLAG_RESET_ON_FORK))
		return -EINVAL;

L
Linus Torvalds 已提交
4126 4127
	/*
	 * Valid priorities for SCHED_FIFO and SCHED_RR are
I
Ingo Molnar 已提交
4128 4129
	 * 1..MAX_USER_RT_PRIO-1, valid priority for SCHED_NORMAL,
	 * SCHED_BATCH and SCHED_IDLE is 0.
L
Linus Torvalds 已提交
4130
	 */
4131
	if ((p->mm && attr->sched_priority > MAX_USER_RT_PRIO-1) ||
4132
	    (!p->mm && attr->sched_priority > MAX_RT_PRIO-1))
L
Linus Torvalds 已提交
4133
		return -EINVAL;
4134 4135
	if ((dl_policy(policy) && !__checkparam_dl(attr)) ||
	    (rt_policy(policy) != (attr->sched_priority != 0)))
L
Linus Torvalds 已提交
4136 4137
		return -EINVAL;

4138 4139 4140
	/*
	 * Allow unprivileged RT tasks to decrease priority:
	 */
4141
	if (user && !capable(CAP_SYS_NICE)) {
4142
		if (fair_policy(policy)) {
4143
			if (attr->sched_nice < task_nice(p) &&
4144
			    !can_nice(p, attr->sched_nice))
4145 4146 4147
				return -EPERM;
		}

4148
		if (rt_policy(policy)) {
4149 4150
			unsigned long rlim_rtprio =
					task_rlimit(p, RLIMIT_RTPRIO);
4151

I
Ingo Molnar 已提交
4152
			/* Can't set/change the rt policy: */
4153 4154 4155
			if (policy != p->policy && !rlim_rtprio)
				return -EPERM;

I
Ingo Molnar 已提交
4156
			/* Can't increase priority: */
4157 4158
			if (attr->sched_priority > p->rt_priority &&
			    attr->sched_priority > rlim_rtprio)
4159 4160
				return -EPERM;
		}
4161

4162 4163 4164 4165 4166 4167 4168 4169 4170
		 /*
		  * 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
4171
		/*
4172 4173
		 * 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
4174
		 */
4175
		if (idle_policy(p->policy) && !idle_policy(policy)) {
4176
			if (!can_nice(p, task_nice(p)))
4177 4178
				return -EPERM;
		}
4179

I
Ingo Molnar 已提交
4180
		/* Can't change other user's priorities: */
4181
		if (!check_same_owner(p))
4182
			return -EPERM;
4183

I
Ingo Molnar 已提交
4184
		/* Normal users shall not reset the sched_reset_on_fork flag: */
4185 4186
		if (p->sched_reset_on_fork && !reset_on_fork)
			return -EPERM;
4187
	}
L
Linus Torvalds 已提交
4188

4189
	if (user) {
4190
		retval = security_task_setscheduler(p);
4191 4192 4193 4194
		if (retval)
			return retval;
	}

4195
	/*
I
Ingo Molnar 已提交
4196
	 * Make sure no PI-waiters arrive (or leave) while we are
4197
	 * changing the priority of the task:
4198
	 *
L
Lucas De Marchi 已提交
4199
	 * To be able to change p->policy safely, the appropriate
L
Linus Torvalds 已提交
4200 4201
	 * runqueue lock must be held.
	 */
4202
	rq = task_rq_lock(p, &rf);
4203
	update_rq_clock(rq);
4204

4205
	/*
I
Ingo Molnar 已提交
4206
	 * Changing the policy of the stop threads its a very bad idea:
4207 4208
	 */
	if (p == rq->stop) {
4209
		task_rq_unlock(rq, p, &rf);
4210 4211 4212
		return -EINVAL;
	}

4213
	/*
4214 4215
	 * If not changing anything there's no need to proceed further,
	 * but store a possible modification of reset_on_fork.
4216
	 */
4217
	if (unlikely(policy == p->policy)) {
4218
		if (fair_policy(policy) && attr->sched_nice != task_nice(p))
4219 4220 4221
			goto change;
		if (rt_policy(policy) && attr->sched_priority != p->rt_priority)
			goto change;
4222
		if (dl_policy(policy) && dl_param_changed(p, attr))
4223
			goto change;
4224

4225
		p->sched_reset_on_fork = reset_on_fork;
4226
		task_rq_unlock(rq, p, &rf);
4227 4228
		return 0;
	}
4229
change:
4230

4231
	if (user) {
4232
#ifdef CONFIG_RT_GROUP_SCHED
4233 4234 4235 4236 4237
		/*
		 * Do not allow realtime tasks into groups that have no runtime
		 * assigned.
		 */
		if (rt_bandwidth_enabled() && rt_policy(policy) &&
4238 4239
				task_group(p)->rt_bandwidth.rt_runtime == 0 &&
				!task_group_is_autogroup(task_group(p))) {
4240
			task_rq_unlock(rq, p, &rf);
4241 4242 4243
			return -EPERM;
		}
#endif
4244 4245 4246 4247 4248 4249 4250 4251 4252
#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.
			 */
4253 4254
			if (!cpumask_subset(span, &p->cpus_allowed) ||
			    rq->rd->dl_bw.bw == 0) {
4255
				task_rq_unlock(rq, p, &rf);
4256 4257 4258 4259 4260
				return -EPERM;
			}
		}
#endif
	}
4261

I
Ingo Molnar 已提交
4262
	/* Re-check policy now with rq lock held: */
L
Linus Torvalds 已提交
4263 4264
	if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) {
		policy = oldpolicy = -1;
4265
		task_rq_unlock(rq, p, &rf);
L
Linus Torvalds 已提交
4266 4267
		goto recheck;
	}
4268 4269 4270 4271 4272 4273

	/*
	 * If setscheduling to SCHED_DEADLINE (or changing the parameters
	 * of a SCHED_DEADLINE task) we need to check if enough bandwidth
	 * is available.
	 */
4274
	if ((dl_policy(policy) || dl_task(p)) && dl_overflow(p, policy, attr)) {
4275
		task_rq_unlock(rq, p, &rf);
4276 4277 4278
		return -EBUSY;
	}

4279 4280 4281
	p->sched_reset_on_fork = reset_on_fork;
	oldprio = p->prio;

4282 4283 4284 4285 4286 4287 4288 4289 4290
	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);
4291 4292
		if (new_effective_prio == oldprio)
			queue_flags &= ~DEQUEUE_MOVE;
4293 4294
	}

4295
	queued = task_on_rq_queued(p);
4296
	running = task_current(rq, p);
4297
	if (queued)
4298
		dequeue_task(rq, p, queue_flags);
4299
	if (running)
4300
		put_prev_task(rq, p);
4301

4302
	prev_class = p->sched_class;
4303
	__setscheduler(rq, p, attr, pi);
4304

4305
	if (queued) {
4306 4307 4308 4309
		/*
		 * We enqueue to tail when the priority of a task is
		 * increased (user space view).
		 */
4310 4311
		if (oldprio < p->prio)
			queue_flags |= ENQUEUE_HEAD;
4312

4313
		enqueue_task(rq, p, queue_flags);
4314
	}
4315
	if (running)
4316
		set_curr_task(rq, p);
4317

P
Peter Zijlstra 已提交
4318
	check_class_changed(rq, p, prev_class, oldprio);
I
Ingo Molnar 已提交
4319 4320 4321

	/* Avoid rq from going away on us: */
	preempt_disable();
4322
	task_rq_unlock(rq, p, &rf);
4323

4324 4325
	if (pi)
		rt_mutex_adjust_pi(p);
4326

I
Ingo Molnar 已提交
4327
	/* Run balance callbacks after we've adjusted the PI chain: */
4328 4329
	balance_callback(rq);
	preempt_enable();
4330

L
Linus Torvalds 已提交
4331 4332
	return 0;
}
4333

4334 4335 4336 4337 4338 4339 4340 4341 4342
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),
	};

4343 4344
	/* Fixup the legacy SCHED_RESET_ON_FORK hack. */
	if ((policy != SETPARAM_POLICY) && (policy & SCHED_RESET_ON_FORK)) {
4345 4346 4347 4348 4349
		attr.sched_flags |= SCHED_FLAG_RESET_ON_FORK;
		policy &= ~SCHED_RESET_ON_FORK;
		attr.sched_policy = policy;
	}

4350
	return __sched_setscheduler(p, &attr, check, true);
4351
}
4352 4353 4354 4355 4356 4357
/**
 * 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.
 *
4358 4359
 * Return: 0 on success. An error code otherwise.
 *
4360 4361 4362
 * NOTE that the task may be already dead.
 */
int sched_setscheduler(struct task_struct *p, int policy,
4363
		       const struct sched_param *param)
4364
{
4365
	return _sched_setscheduler(p, policy, param, true);
4366
}
L
Linus Torvalds 已提交
4367 4368
EXPORT_SYMBOL_GPL(sched_setscheduler);

4369 4370
int sched_setattr(struct task_struct *p, const struct sched_attr *attr)
{
4371
	return __sched_setscheduler(p, attr, true, true);
4372 4373 4374
}
EXPORT_SYMBOL_GPL(sched_setattr);

4375 4376 4377 4378 4379 4380 4381 4382 4383 4384
/**
 * 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.
4385 4386
 *
 * Return: 0 on success. An error code otherwise.
4387 4388
 */
int sched_setscheduler_nocheck(struct task_struct *p, int policy,
4389
			       const struct sched_param *param)
4390
{
4391
	return _sched_setscheduler(p, policy, param, false);
4392
}
4393
EXPORT_SYMBOL_GPL(sched_setscheduler_nocheck);
4394

I
Ingo Molnar 已提交
4395 4396
static int
do_sched_setscheduler(pid_t pid, int policy, struct sched_param __user *param)
L
Linus Torvalds 已提交
4397 4398 4399
{
	struct sched_param lparam;
	struct task_struct *p;
4400
	int retval;
L
Linus Torvalds 已提交
4401 4402 4403 4404 4405

	if (!param || pid < 0)
		return -EINVAL;
	if (copy_from_user(&lparam, param, sizeof(struct sched_param)))
		return -EFAULT;
4406 4407 4408

	rcu_read_lock();
	retval = -ESRCH;
L
Linus Torvalds 已提交
4409
	p = find_process_by_pid(pid);
4410 4411 4412
	if (p != NULL)
		retval = sched_setscheduler(p, policy, &lparam);
	rcu_read_unlock();
4413

L
Linus Torvalds 已提交
4414 4415 4416
	return retval;
}

4417 4418 4419
/*
 * Mimics kernel/events/core.c perf_copy_attr().
 */
I
Ingo Molnar 已提交
4420
static int sched_copy_attr(struct sched_attr __user *uattr, struct sched_attr *attr)
4421 4422 4423 4424 4425 4426 4427
{
	u32 size;
	int ret;

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

I
Ingo Molnar 已提交
4428
	/* Zero the full structure, so that a short copy will be nice: */
4429 4430 4431 4432 4433 4434
	memset(attr, 0, sizeof(*attr));

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

I
Ingo Molnar 已提交
4435 4436
	/* Bail out on silly large: */
	if (size > PAGE_SIZE)
4437 4438
		goto err_size;

I
Ingo Molnar 已提交
4439 4440
	/* ABI compatibility quirk: */
	if (!size)
4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474
		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
4475
	 * XXX: Do we want to be lenient like existing syscalls; or do we want
4476 4477
	 * to be strict and return an error on out-of-bounds values?
	 */
4478
	attr->sched_nice = clamp(attr->sched_nice, MIN_NICE, MAX_NICE);
4479

4480
	return 0;
4481 4482 4483

err_size:
	put_user(sizeof(*attr), &uattr->size);
4484
	return -E2BIG;
4485 4486
}

L
Linus Torvalds 已提交
4487 4488 4489 4490 4491
/**
 * 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.
4492 4493
 *
 * Return: 0 on success. An error code otherwise.
L
Linus Torvalds 已提交
4494
 */
I
Ingo Molnar 已提交
4495
SYSCALL_DEFINE3(sched_setscheduler, pid_t, pid, int, policy, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4496
{
4497 4498 4499
	if (policy < 0)
		return -EINVAL;

L
Linus Torvalds 已提交
4500 4501 4502 4503 4504 4505 4506
	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.
4507 4508
 *
 * Return: 0 on success. An error code otherwise.
L
Linus Torvalds 已提交
4509
 */
4510
SYSCALL_DEFINE2(sched_setparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4511
{
4512
	return do_sched_setscheduler(pid, SETPARAM_POLICY, param);
L
Linus Torvalds 已提交
4513 4514
}

4515 4516 4517
/**
 * sys_sched_setattr - same as above, but with extended sched_attr
 * @pid: the pid in question.
J
Juri Lelli 已提交
4518
 * @uattr: structure containing the extended parameters.
4519
 * @flags: for future extension.
4520
 */
4521 4522
SYSCALL_DEFINE3(sched_setattr, pid_t, pid, struct sched_attr __user *, uattr,
			       unsigned int, flags)
4523 4524 4525 4526 4527
{
	struct sched_attr attr;
	struct task_struct *p;
	int retval;

4528
	if (!uattr || pid < 0 || flags)
4529 4530
		return -EINVAL;

4531 4532 4533
	retval = sched_copy_attr(uattr, &attr);
	if (retval)
		return retval;
4534

4535
	if ((int)attr.sched_policy < 0)
4536
		return -EINVAL;
4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547

	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
4548 4549 4550
/**
 * sys_sched_getscheduler - get the policy (scheduling class) of a thread
 * @pid: the pid in question.
4551 4552 4553
 *
 * Return: On success, the policy of the thread. Otherwise, a negative error
 * code.
L
Linus Torvalds 已提交
4554
 */
4555
SYSCALL_DEFINE1(sched_getscheduler, pid_t, pid)
L
Linus Torvalds 已提交
4556
{
4557
	struct task_struct *p;
4558
	int retval;
L
Linus Torvalds 已提交
4559 4560

	if (pid < 0)
4561
		return -EINVAL;
L
Linus Torvalds 已提交
4562 4563

	retval = -ESRCH;
4564
	rcu_read_lock();
L
Linus Torvalds 已提交
4565 4566 4567 4568
	p = find_process_by_pid(pid);
	if (p) {
		retval = security_task_getscheduler(p);
		if (!retval)
4569 4570
			retval = p->policy
				| (p->sched_reset_on_fork ? SCHED_RESET_ON_FORK : 0);
L
Linus Torvalds 已提交
4571
	}
4572
	rcu_read_unlock();
L
Linus Torvalds 已提交
4573 4574 4575 4576
	return retval;
}

/**
4577
 * sys_sched_getparam - get the RT priority of a thread
L
Linus Torvalds 已提交
4578 4579
 * @pid: the pid in question.
 * @param: structure containing the RT priority.
4580 4581 4582
 *
 * Return: On success, 0 and the RT priority is in @param. Otherwise, an error
 * code.
L
Linus Torvalds 已提交
4583
 */
4584
SYSCALL_DEFINE2(sched_getparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4585
{
4586
	struct sched_param lp = { .sched_priority = 0 };
4587
	struct task_struct *p;
4588
	int retval;
L
Linus Torvalds 已提交
4589 4590

	if (!param || pid < 0)
4591
		return -EINVAL;
L
Linus Torvalds 已提交
4592

4593
	rcu_read_lock();
L
Linus Torvalds 已提交
4594 4595 4596 4597 4598 4599 4600 4601 4602
	p = find_process_by_pid(pid);
	retval = -ESRCH;
	if (!p)
		goto out_unlock;

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

4603 4604
	if (task_has_rt_policy(p))
		lp.sched_priority = p->rt_priority;
4605
	rcu_read_unlock();
L
Linus Torvalds 已提交
4606 4607 4608 4609 4610 4611 4612 4613 4614

	/*
	 * 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:
4615
	rcu_read_unlock();
L
Linus Torvalds 已提交
4616 4617 4618
	return retval;
}

4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641
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)
4642
				return -EFBIG;
4643 4644 4645 4646 4647
		}

		attr->size = usize;
	}

4648
	ret = copy_to_user(uattr, attr, attr->size);
4649 4650 4651
	if (ret)
		return -EFAULT;

4652
	return 0;
4653 4654 4655
}

/**
4656
 * sys_sched_getattr - similar to sched_getparam, but with sched_attr
4657
 * @pid: the pid in question.
J
Juri Lelli 已提交
4658
 * @uattr: structure containing the extended parameters.
4659
 * @size: sizeof(attr) for fwd/bwd comp.
4660
 * @flags: for future extension.
4661
 */
4662 4663
SYSCALL_DEFINE4(sched_getattr, pid_t, pid, struct sched_attr __user *, uattr,
		unsigned int, size, unsigned int, flags)
4664 4665 4666 4667 4668 4669 4670 4671
{
	struct sched_attr attr = {
		.size = sizeof(struct sched_attr),
	};
	struct task_struct *p;
	int retval;

	if (!uattr || pid < 0 || size > PAGE_SIZE ||
4672
	    size < SCHED_ATTR_SIZE_VER0 || flags)
4673 4674 4675 4676 4677 4678 4679 4680 4681 4682 4683 4684 4685
		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;
4686 4687
	if (p->sched_reset_on_fork)
		attr.sched_flags |= SCHED_FLAG_RESET_ON_FORK;
4688 4689 4690
	if (task_has_dl_policy(p))
		__getparam_dl(p, &attr);
	else if (task_has_rt_policy(p))
4691 4692
		attr.sched_priority = p->rt_priority;
	else
4693
		attr.sched_nice = task_nice(p);
4694 4695 4696 4697 4698 4699 4700 4701 4702 4703 4704

	rcu_read_unlock();

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

out_unlock:
	rcu_read_unlock();
	return retval;
}

4705
long sched_setaffinity(pid_t pid, const struct cpumask *in_mask)
L
Linus Torvalds 已提交
4706
{
4707
	cpumask_var_t cpus_allowed, new_mask;
4708 4709
	struct task_struct *p;
	int retval;
L
Linus Torvalds 已提交
4710

4711
	rcu_read_lock();
L
Linus Torvalds 已提交
4712 4713 4714

	p = find_process_by_pid(pid);
	if (!p) {
4715
		rcu_read_unlock();
L
Linus Torvalds 已提交
4716 4717 4718
		return -ESRCH;
	}

4719
	/* Prevent p going away */
L
Linus Torvalds 已提交
4720
	get_task_struct(p);
4721
	rcu_read_unlock();
L
Linus Torvalds 已提交
4722

4723 4724 4725 4726
	if (p->flags & PF_NO_SETAFFINITY) {
		retval = -EINVAL;
		goto out_put_task;
	}
4727 4728 4729 4730 4731 4732 4733 4734
	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
4735
	retval = -EPERM;
E
Eric W. Biederman 已提交
4736 4737 4738 4739
	if (!check_same_owner(p)) {
		rcu_read_lock();
		if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
			rcu_read_unlock();
4740
			goto out_free_new_mask;
E
Eric W. Biederman 已提交
4741 4742 4743
		}
		rcu_read_unlock();
	}
L
Linus Torvalds 已提交
4744

4745
	retval = security_task_setscheduler(p);
4746
	if (retval)
4747
		goto out_free_new_mask;
4748

4749 4750 4751 4752

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

4753 4754 4755 4756 4757 4758 4759
	/*
	 * 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
4760 4761 4762
	if (task_has_dl_policy(p) && dl_bandwidth_enabled()) {
		rcu_read_lock();
		if (!cpumask_subset(task_rq(p)->rd->span, new_mask)) {
4763
			retval = -EBUSY;
4764
			rcu_read_unlock();
4765
			goto out_free_new_mask;
4766
		}
4767
		rcu_read_unlock();
4768 4769
	}
#endif
P
Peter Zijlstra 已提交
4770
again:
4771
	retval = __set_cpus_allowed_ptr(p, new_mask, true);
L
Linus Torvalds 已提交
4772

P
Paul Menage 已提交
4773
	if (!retval) {
4774 4775
		cpuset_cpus_allowed(p, cpus_allowed);
		if (!cpumask_subset(new_mask, cpus_allowed)) {
P
Paul Menage 已提交
4776 4777 4778 4779 4780
			/*
			 * We must have raced with a concurrent cpuset
			 * update. Just reset the cpus_allowed to the
			 * cpuset's cpus_allowed
			 */
4781
			cpumask_copy(new_mask, cpus_allowed);
P
Paul Menage 已提交
4782 4783 4784
			goto again;
		}
	}
4785
out_free_new_mask:
4786 4787 4788 4789
	free_cpumask_var(new_mask);
out_free_cpus_allowed:
	free_cpumask_var(cpus_allowed);
out_put_task:
L
Linus Torvalds 已提交
4790 4791 4792 4793 4794
	put_task_struct(p);
	return retval;
}

static int get_user_cpu_mask(unsigned long __user *user_mask_ptr, unsigned len,
4795
			     struct cpumask *new_mask)
L
Linus Torvalds 已提交
4796
{
4797 4798 4799 4800 4801
	if (len < cpumask_size())
		cpumask_clear(new_mask);
	else if (len > cpumask_size())
		len = cpumask_size();

L
Linus Torvalds 已提交
4802 4803 4804 4805
	return copy_from_user(new_mask, user_mask_ptr, len) ? -EFAULT : 0;
}

/**
I
Ingo Molnar 已提交
4806
 * sys_sched_setaffinity - set the CPU affinity of a process
L
Linus Torvalds 已提交
4807 4808
 * @pid: pid of the process
 * @len: length in bytes of the bitmask pointed to by user_mask_ptr
I
Ingo Molnar 已提交
4809
 * @user_mask_ptr: user-space pointer to the new CPU mask
4810 4811
 *
 * Return: 0 on success. An error code otherwise.
L
Linus Torvalds 已提交
4812
 */
4813 4814
SYSCALL_DEFINE3(sched_setaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
4815
{
4816
	cpumask_var_t new_mask;
L
Linus Torvalds 已提交
4817 4818
	int retval;

4819 4820
	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
4821

4822 4823 4824 4825 4826
	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
4827 4828
}

4829
long sched_getaffinity(pid_t pid, struct cpumask *mask)
L
Linus Torvalds 已提交
4830
{
4831
	struct task_struct *p;
4832
	unsigned long flags;
L
Linus Torvalds 已提交
4833 4834
	int retval;

4835
	rcu_read_lock();
L
Linus Torvalds 已提交
4836 4837 4838 4839 4840 4841

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

4842 4843 4844 4845
	retval = security_task_getscheduler(p);
	if (retval)
		goto out_unlock;

4846
	raw_spin_lock_irqsave(&p->pi_lock, flags);
4847
	cpumask_and(mask, &p->cpus_allowed, cpu_active_mask);
4848
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
4849 4850

out_unlock:
4851
	rcu_read_unlock();
L
Linus Torvalds 已提交
4852

4853
	return retval;
L
Linus Torvalds 已提交
4854 4855 4856
}

/**
I
Ingo Molnar 已提交
4857
 * sys_sched_getaffinity - get the CPU affinity of a process
L
Linus Torvalds 已提交
4858 4859
 * @pid: pid of the process
 * @len: length in bytes of the bitmask pointed to by user_mask_ptr
I
Ingo Molnar 已提交
4860
 * @user_mask_ptr: user-space pointer to hold the current CPU mask
4861
 *
4862 4863
 * Return: size of CPU mask copied to user_mask_ptr on success. An
 * error code otherwise.
L
Linus Torvalds 已提交
4864
 */
4865 4866
SYSCALL_DEFINE3(sched_getaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
4867 4868
{
	int ret;
4869
	cpumask_var_t mask;
L
Linus Torvalds 已提交
4870

A
Anton Blanchard 已提交
4871
	if ((len * BITS_PER_BYTE) < nr_cpu_ids)
4872 4873
		return -EINVAL;
	if (len & (sizeof(unsigned long)-1))
L
Linus Torvalds 已提交
4874 4875
		return -EINVAL;

4876 4877
	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
4878

4879 4880
	ret = sched_getaffinity(pid, mask);
	if (ret == 0) {
4881
		size_t retlen = min_t(size_t, len, cpumask_size());
4882 4883

		if (copy_to_user(user_mask_ptr, mask, retlen))
4884 4885
			ret = -EFAULT;
		else
4886
			ret = retlen;
4887 4888
	}
	free_cpumask_var(mask);
L
Linus Torvalds 已提交
4889

4890
	return ret;
L
Linus Torvalds 已提交
4891 4892 4893 4894 4895
}

/**
 * sys_sched_yield - yield the current processor to other threads.
 *
I
Ingo Molnar 已提交
4896 4897
 * This function yields the current CPU to other tasks. If there are no
 * other threads running on this CPU then this function will return.
4898 4899
 *
 * Return: 0.
L
Linus Torvalds 已提交
4900
 */
4901
SYSCALL_DEFINE0(sched_yield)
L
Linus Torvalds 已提交
4902
{
4903
	struct rq *rq = this_rq_lock();
L
Linus Torvalds 已提交
4904

4905
	schedstat_inc(rq->yld_count);
4906
	current->sched_class->yield_task(rq);
L
Linus Torvalds 已提交
4907 4908 4909 4910 4911 4912

	/*
	 * Since we are going to call schedule() anyway, there's
	 * no need to preempt or enable interrupts:
	 */
	__release(rq->lock);
4913
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
4914
	do_raw_spin_unlock(&rq->lock);
4915
	sched_preempt_enable_no_resched();
L
Linus Torvalds 已提交
4916 4917 4918 4919 4920 4921

	schedule();

	return 0;
}

4922
#ifndef CONFIG_PREEMPT
4923
int __sched _cond_resched(void)
L
Linus Torvalds 已提交
4924
{
4925
	if (should_resched(0)) {
4926
		preempt_schedule_common();
L
Linus Torvalds 已提交
4927 4928 4929 4930
		return 1;
	}
	return 0;
}
4931
EXPORT_SYMBOL(_cond_resched);
4932
#endif
L
Linus Torvalds 已提交
4933 4934

/*
4935
 * __cond_resched_lock() - if a reschedule is pending, drop the given lock,
L
Linus Torvalds 已提交
4936 4937
 * call schedule, and on return reacquire the lock.
 *
I
Ingo Molnar 已提交
4938
 * This works OK both with and without CONFIG_PREEMPT. We do strange low-level
L
Linus Torvalds 已提交
4939 4940 4941
 * operations here to prevent schedule() from being called twice (once via
 * spin_unlock(), once by hand).
 */
4942
int __cond_resched_lock(spinlock_t *lock)
L
Linus Torvalds 已提交
4943
{
4944
	int resched = should_resched(PREEMPT_LOCK_OFFSET);
J
Jan Kara 已提交
4945 4946
	int ret = 0;

4947 4948
	lockdep_assert_held(lock);

4949
	if (spin_needbreak(lock) || resched) {
L
Linus Torvalds 已提交
4950
		spin_unlock(lock);
P
Peter Zijlstra 已提交
4951
		if (resched)
4952
			preempt_schedule_common();
N
Nick Piggin 已提交
4953 4954
		else
			cpu_relax();
J
Jan Kara 已提交
4955
		ret = 1;
L
Linus Torvalds 已提交
4956 4957
		spin_lock(lock);
	}
J
Jan Kara 已提交
4958
	return ret;
L
Linus Torvalds 已提交
4959
}
4960
EXPORT_SYMBOL(__cond_resched_lock);
L
Linus Torvalds 已提交
4961

4962
int __sched __cond_resched_softirq(void)
L
Linus Torvalds 已提交
4963 4964 4965
{
	BUG_ON(!in_softirq());

4966
	if (should_resched(SOFTIRQ_DISABLE_OFFSET)) {
4967
		local_bh_enable();
4968
		preempt_schedule_common();
L
Linus Torvalds 已提交
4969 4970 4971 4972 4973
		local_bh_disable();
		return 1;
	}
	return 0;
}
4974
EXPORT_SYMBOL(__cond_resched_softirq);
L
Linus Torvalds 已提交
4975 4976 4977 4978

/**
 * yield - yield the current processor to other threads.
 *
P
Peter Zijlstra 已提交
4979 4980 4981 4982 4983 4984 4985 4986 4987
 * 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
4988
 *	yield();
P
Peter Zijlstra 已提交
4989 4990 4991 4992 4993 4994 4995 4996
 *
 * 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
4997 4998 4999 5000 5001 5002 5003 5004
 */
void __sched yield(void)
{
	set_current_state(TASK_RUNNING);
	sys_sched_yield();
}
EXPORT_SYMBOL(yield);

5005 5006 5007 5008
/**
 * 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
5009 5010
 * @p: target task
 * @preempt: whether task preemption is allowed or not
5011 5012 5013 5014
 *
 * 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.
 *
5015
 * Return:
5016 5017 5018
 *	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.
5019
 */
5020
int __sched yield_to(struct task_struct *p, bool preempt)
5021 5022 5023 5024
{
	struct task_struct *curr = current;
	struct rq *rq, *p_rq;
	unsigned long flags;
5025
	int yielded = 0;
5026 5027 5028 5029 5030 5031

	local_irq_save(flags);
	rq = this_rq();

again:
	p_rq = task_rq(p);
5032 5033 5034 5035 5036 5037 5038 5039 5040
	/*
	 * 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;
	}

5041
	double_rq_lock(rq, p_rq);
5042
	if (task_rq(p) != p_rq) {
5043 5044 5045 5046 5047
		double_rq_unlock(rq, p_rq);
		goto again;
	}

	if (!curr->sched_class->yield_to_task)
5048
		goto out_unlock;
5049 5050

	if (curr->sched_class != p->sched_class)
5051
		goto out_unlock;
5052 5053

	if (task_running(p_rq, p) || p->state)
5054
		goto out_unlock;
5055 5056

	yielded = curr->sched_class->yield_to_task(rq, p, preempt);
5057
	if (yielded) {
5058
		schedstat_inc(rq->yld_count);
5059 5060 5061 5062 5063
		/*
		 * Make p's CPU reschedule; pick_next_entity takes care of
		 * fairness.
		 */
		if (preempt && rq != p_rq)
5064
			resched_curr(p_rq);
5065
	}
5066

5067
out_unlock:
5068
	double_rq_unlock(rq, p_rq);
5069
out_irq:
5070 5071
	local_irq_restore(flags);

5072
	if (yielded > 0)
5073 5074 5075 5076 5077 5078
		schedule();

	return yielded;
}
EXPORT_SYMBOL_GPL(yield_to);

5079 5080 5081 5082 5083 5084 5085 5086 5087 5088 5089 5090 5091 5092 5093
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
5094
/*
I
Ingo Molnar 已提交
5095
 * This task is about to go to sleep on IO. Increment rq->nr_iowait so
L
Linus Torvalds 已提交
5096 5097 5098 5099
 * that process accounting knows that this is a task in IO wait state.
 */
long __sched io_schedule_timeout(long timeout)
{
5100
	int token;
L
Linus Torvalds 已提交
5101 5102
	long ret;

5103
	token = io_schedule_prepare();
L
Linus Torvalds 已提交
5104
	ret = schedule_timeout(timeout);
5105
	io_schedule_finish(token);
5106

L
Linus Torvalds 已提交
5107 5108
	return ret;
}
5109
EXPORT_SYMBOL(io_schedule_timeout);
L
Linus Torvalds 已提交
5110

5111 5112 5113 5114 5115 5116 5117 5118 5119 5120
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
5121 5122 5123 5124
/**
 * sys_sched_get_priority_max - return maximum RT priority.
 * @policy: scheduling class.
 *
5125 5126 5127
 * 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
5128
 */
5129
SYSCALL_DEFINE1(sched_get_priority_max, int, policy)
L
Linus Torvalds 已提交
5130 5131 5132 5133 5134 5135 5136 5137
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = MAX_USER_RT_PRIO-1;
		break;
5138
	case SCHED_DEADLINE:
L
Linus Torvalds 已提交
5139
	case SCHED_NORMAL:
5140
	case SCHED_BATCH:
I
Ingo Molnar 已提交
5141
	case SCHED_IDLE:
L
Linus Torvalds 已提交
5142 5143 5144 5145 5146 5147 5148 5149 5150 5151
		ret = 0;
		break;
	}
	return ret;
}

/**
 * sys_sched_get_priority_min - return minimum RT priority.
 * @policy: scheduling class.
 *
5152 5153 5154
 * 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
5155
 */
5156
SYSCALL_DEFINE1(sched_get_priority_min, int, policy)
L
Linus Torvalds 已提交
5157 5158 5159 5160 5161 5162 5163 5164
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = 1;
		break;
5165
	case SCHED_DEADLINE:
L
Linus Torvalds 已提交
5166
	case SCHED_NORMAL:
5167
	case SCHED_BATCH:
I
Ingo Molnar 已提交
5168
	case SCHED_IDLE:
L
Linus Torvalds 已提交
5169 5170 5171 5172 5173 5174 5175 5176 5177 5178 5179 5180
		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.
5181 5182 5183
 *
 * Return: On success, 0 and the timeslice is in @interval. Otherwise,
 * an error code.
L
Linus Torvalds 已提交
5184
 */
5185
SYSCALL_DEFINE2(sched_rr_get_interval, pid_t, pid,
5186
		struct timespec __user *, interval)
L
Linus Torvalds 已提交
5187
{
5188
	struct task_struct *p;
D
Dmitry Adamushko 已提交
5189
	unsigned int time_slice;
5190 5191
	struct rq_flags rf;
	struct timespec t;
5192
	struct rq *rq;
5193
	int retval;
L
Linus Torvalds 已提交
5194 5195

	if (pid < 0)
5196
		return -EINVAL;
L
Linus Torvalds 已提交
5197 5198

	retval = -ESRCH;
5199
	rcu_read_lock();
L
Linus Torvalds 已提交
5200 5201 5202 5203 5204 5205 5206 5207
	p = find_process_by_pid(pid);
	if (!p)
		goto out_unlock;

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

5208
	rq = task_rq_lock(p, &rf);
5209 5210 5211
	time_slice = 0;
	if (p->sched_class->get_rr_interval)
		time_slice = p->sched_class->get_rr_interval(rq, p);
5212
	task_rq_unlock(rq, p, &rf);
D
Dmitry Adamushko 已提交
5213

5214
	rcu_read_unlock();
D
Dmitry Adamushko 已提交
5215
	jiffies_to_timespec(time_slice, &t);
L
Linus Torvalds 已提交
5216 5217
	retval = copy_to_user(interval, &t, sizeof(t)) ? -EFAULT : 0;
	return retval;
5218

L
Linus Torvalds 已提交
5219
out_unlock:
5220
	rcu_read_unlock();
L
Linus Torvalds 已提交
5221 5222 5223
	return retval;
}

5224
static const char stat_nam[] = TASK_STATE_TO_CHAR_STR;
5225

5226
void sched_show_task(struct task_struct *p)
L
Linus Torvalds 已提交
5227 5228
{
	unsigned long free = 0;
5229
	int ppid;
5230
	unsigned long state = p->state;
L
Linus Torvalds 已提交
5231

5232 5233
	if (!try_get_task_stack(p))
		return;
5234 5235
	if (state)
		state = __ffs(state) + 1;
5236
	printk(KERN_INFO "%-15.15s %c", p->comm,
5237
		state < sizeof(stat_nam) - 1 ? stat_nam[state] : '?');
L
Linus Torvalds 已提交
5238
	if (state == TASK_RUNNING)
P
Peter Zijlstra 已提交
5239
		printk(KERN_CONT "  running task    ");
L
Linus Torvalds 已提交
5240
#ifdef CONFIG_DEBUG_STACK_USAGE
5241
	free = stack_not_used(p);
L
Linus Torvalds 已提交
5242
#endif
5243
	ppid = 0;
5244
	rcu_read_lock();
5245 5246
	if (pid_alive(p))
		ppid = task_pid_nr(rcu_dereference(p->real_parent));
5247
	rcu_read_unlock();
P
Peter Zijlstra 已提交
5248
	printk(KERN_CONT "%5lu %5d %6d 0x%08lx\n", free,
5249
		task_pid_nr(p), ppid,
5250
		(unsigned long)task_thread_info(p)->flags);
L
Linus Torvalds 已提交
5251

5252
	print_worker_info(KERN_INFO, p);
5253
	show_stack(p, NULL);
5254
	put_task_stack(p);
L
Linus Torvalds 已提交
5255 5256
}

I
Ingo Molnar 已提交
5257
void show_state_filter(unsigned long state_filter)
L
Linus Torvalds 已提交
5258
{
5259
	struct task_struct *g, *p;
L
Linus Torvalds 已提交
5260

5261
#if BITS_PER_LONG == 32
P
Peter Zijlstra 已提交
5262 5263
	printk(KERN_INFO
		"  task                PC stack   pid father\n");
L
Linus Torvalds 已提交
5264
#else
P
Peter Zijlstra 已提交
5265 5266
	printk(KERN_INFO
		"  task                        PC stack   pid father\n");
L
Linus Torvalds 已提交
5267
#endif
5268
	rcu_read_lock();
5269
	for_each_process_thread(g, p) {
L
Linus Torvalds 已提交
5270 5271
		/*
		 * reset the NMI-timeout, listing all files on a slow
L
Lucas De Marchi 已提交
5272
		 * console might take a lot of time:
5273 5274 5275
		 * 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
5276 5277
		 */
		touch_nmi_watchdog();
5278
		touch_all_softlockup_watchdogs();
I
Ingo Molnar 已提交
5279
		if (!state_filter || (p->state & state_filter))
5280
			sched_show_task(p);
5281
	}
L
Linus Torvalds 已提交
5282

I
Ingo Molnar 已提交
5283
#ifdef CONFIG_SCHED_DEBUG
5284 5285
	if (!state_filter)
		sysrq_sched_debug_show();
I
Ingo Molnar 已提交
5286
#endif
5287
	rcu_read_unlock();
I
Ingo Molnar 已提交
5288 5289 5290
	/*
	 * Only show locks if all tasks are dumped:
	 */
5291
	if (!state_filter)
I
Ingo Molnar 已提交
5292
		debug_show_all_locks();
L
Linus Torvalds 已提交
5293 5294
}

5295
void init_idle_bootup_task(struct task_struct *idle)
I
Ingo Molnar 已提交
5296
{
I
Ingo Molnar 已提交
5297
	idle->sched_class = &idle_sched_class;
I
Ingo Molnar 已提交
5298 5299
}

5300 5301 5302
/**
 * init_idle - set up an idle thread for a given CPU
 * @idle: task in question
I
Ingo Molnar 已提交
5303
 * @cpu: CPU the idle task belongs to
5304 5305 5306 5307
 *
 * NOTE: this function does not set the idle thread's NEED_RESCHED
 * flag, to make booting more robust.
 */
5308
void init_idle(struct task_struct *idle, int cpu)
L
Linus Torvalds 已提交
5309
{
5310
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
5311 5312
	unsigned long flags;

5313 5314
	raw_spin_lock_irqsave(&idle->pi_lock, flags);
	raw_spin_lock(&rq->lock);
5315

5316
	__sched_fork(0, idle);
5317
	idle->state = TASK_RUNNING;
I
Ingo Molnar 已提交
5318
	idle->se.exec_start = sched_clock();
5319
	idle->flags |= PF_IDLE;
I
Ingo Molnar 已提交
5320

5321 5322
	kasan_unpoison_task_stack(idle);

5323 5324 5325 5326 5327 5328 5329 5330 5331
#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
5332 5333
	/*
	 * We're having a chicken and egg problem, even though we are
I
Ingo Molnar 已提交
5334
	 * holding rq->lock, the CPU isn't yet set to this CPU so the
5335 5336 5337 5338 5339 5340 5341 5342
	 * 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
5343
	__set_task_cpu(idle, cpu);
5344
	rcu_read_unlock();
L
Linus Torvalds 已提交
5345 5346

	rq->curr = rq->idle = idle;
5347
	idle->on_rq = TASK_ON_RQ_QUEUED;
5348
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
5349
	idle->on_cpu = 1;
5350
#endif
5351 5352
	raw_spin_unlock(&rq->lock);
	raw_spin_unlock_irqrestore(&idle->pi_lock, flags);
L
Linus Torvalds 已提交
5353 5354

	/* Set the preempt count _outside_ the spinlocks! */
5355
	init_idle_preempt_count(idle, cpu);
5356

I
Ingo Molnar 已提交
5357 5358 5359 5360
	/*
	 * The idle tasks have their own, simple scheduling class:
	 */
	idle->sched_class = &idle_sched_class;
5361
	ftrace_graph_init_idle_task(idle, cpu);
5362
	vtime_init_idle(idle, cpu);
5363
#ifdef CONFIG_SMP
5364 5365
	sprintf(idle->comm, "%s/%d", INIT_TASK_COMM, cpu);
#endif
I
Ingo Molnar 已提交
5366 5367
}

5368 5369 5370 5371 5372 5373 5374
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;

5375 5376 5377
	if (!cpumask_weight(cur))
		return ret;

5378
	rcu_read_lock_sched();
5379 5380 5381 5382 5383 5384 5385 5386
	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);
5387
	rcu_read_unlock_sched();
5388 5389 5390 5391

	return ret;
}

5392 5393 5394 5395 5396 5397 5398
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
5399
	 * to a new cpuset; we don't want to change their CPU
5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410 5411 5412 5413 5414 5415
	 * 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);
5416
		struct dl_bw *dl_b;
5417 5418 5419 5420
		bool overflow;
		int cpus;
		unsigned long flags;

5421 5422
		rcu_read_lock_sched();
		dl_b = dl_bw_of(dest_cpu);
5423 5424 5425 5426 5427 5428 5429 5430 5431 5432 5433 5434 5435 5436 5437
		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);
5438
		rcu_read_unlock_sched();
5439 5440 5441 5442 5443 5444 5445

	}
#endif
out:
	return ret;
}

L
Linus Torvalds 已提交
5446 5447
#ifdef CONFIG_SMP

5448
bool sched_smp_initialized __read_mostly;
5449

5450 5451 5452 5453 5454 5455 5456 5457 5458 5459 5460 5461 5462 5463 5464
#ifdef CONFIG_NUMA_BALANCING
/* Migrate current task p to target_cpu */
int migrate_task_to(struct task_struct *p, int target_cpu)
{
	struct migration_arg arg = { p, target_cpu };
	int curr_cpu = task_cpu(p);

	if (curr_cpu == target_cpu)
		return 0;

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

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

5465
	trace_sched_move_numa(p, curr_cpu, target_cpu);
5466 5467
	return stop_one_cpu(curr_cpu, migration_cpu_stop, &arg);
}
5468 5469 5470 5471 5472 5473 5474

/*
 * 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)
{
5475
	bool queued, running;
5476 5477
	struct rq_flags rf;
	struct rq *rq;
5478

5479
	rq = task_rq_lock(p, &rf);
5480
	queued = task_on_rq_queued(p);
5481 5482
	running = task_current(rq, p);

5483
	if (queued)
5484
		dequeue_task(rq, p, DEQUEUE_SAVE);
5485
	if (running)
5486
		put_prev_task(rq, p);
5487 5488 5489

	p->numa_preferred_nid = nid;

5490
	if (queued)
5491
		enqueue_task(rq, p, ENQUEUE_RESTORE);
5492
	if (running)
5493
		set_curr_task(rq, p);
5494
	task_rq_unlock(rq, p, &rf);
5495
}
P
Peter Zijlstra 已提交
5496
#endif /* CONFIG_NUMA_BALANCING */
5497

L
Linus Torvalds 已提交
5498
#ifdef CONFIG_HOTPLUG_CPU
5499
/*
I
Ingo Molnar 已提交
5500
 * Ensure that the idle task is using init_mm right before its CPU goes
5501
 * offline.
5502
 */
5503
void idle_task_exit(void)
L
Linus Torvalds 已提交
5504
{
5505
	struct mm_struct *mm = current->active_mm;
5506

5507
	BUG_ON(cpu_online(smp_processor_id()));
5508

5509
	if (mm != &init_mm) {
5510
		switch_mm_irqs_off(mm, &init_mm, current);
5511 5512
		finish_arch_post_lock_switch();
	}
5513
	mmdrop(mm);
L
Linus Torvalds 已提交
5514 5515 5516
}

/*
5517 5518
 * 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
5519 5520 5521
 * 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.
5522 5523
 *
 * Also see the comment "Global load-average calculations".
L
Linus Torvalds 已提交
5524
 */
5525
static void calc_load_migrate(struct rq *rq)
L
Linus Torvalds 已提交
5526
{
5527
	long delta = calc_load_fold_active(rq, 1);
5528 5529
	if (delta)
		atomic_long_add(delta, &calc_load_tasks);
L
Linus Torvalds 已提交
5530 5531
}

5532 5533 5534 5535 5536 5537 5538 5539 5540 5541 5542 5543 5544 5545 5546 5547
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,
};

5548
/*
5549 5550 5551 5552 5553 5554
 * 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
5555
 */
5556
static void migrate_tasks(struct rq *dead_rq)
L
Linus Torvalds 已提交
5557
{
5558
	struct rq *rq = dead_rq;
5559
	struct task_struct *next, *stop = rq->stop;
5560
	struct rq_flags rf;
5561
	int dest_cpu;
L
Linus Torvalds 已提交
5562 5563

	/*
5564 5565 5566 5567 5568 5569 5570
	 * 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
5571
	 */
5572
	rq->stop = NULL;
5573

5574 5575 5576 5577 5578
	/*
	 * put_prev_task() and pick_next_task() sched
	 * class method both need to have an up-to-date
	 * value of rq->clock[_task]
	 */
5579
	rq_pin_lock(rq, &rf);
5580
	update_rq_clock(rq);
5581
	rq_unpin_lock(rq, &rf);
5582

5583
	for (;;) {
5584 5585
		/*
		 * There's this thread running, bail when that's the only
I
Ingo Molnar 已提交
5586
		 * remaining thread:
5587 5588
		 */
		if (rq->nr_running == 1)
I
Ingo Molnar 已提交
5589
			break;
5590

5591
		/*
I
Ingo Molnar 已提交
5592
		 * pick_next_task() assumes pinned rq->lock:
5593
		 */
5594
		rq_repin_lock(rq, &rf);
5595
		next = pick_next_task(rq, &fake_task, &rf);
5596
		BUG_ON(!next);
D
Dmitry Adamushko 已提交
5597
		next->sched_class->put_prev_task(rq, next);
5598

W
Wanpeng Li 已提交
5599 5600 5601 5602 5603 5604 5605 5606 5607
		/*
		 * 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.
		 */
5608
		rq_unpin_lock(rq, &rf);
W
Wanpeng Li 已提交
5609 5610 5611 5612 5613 5614 5615 5616 5617 5618 5619 5620 5621 5622
		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;
		}

5623
		/* Find suitable destination for @next, with force if needed. */
5624
		dest_cpu = select_fallback_rq(dead_rq->cpu, next);
5625

5626 5627 5628 5629 5630 5631
		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
5632
		raw_spin_unlock(&next->pi_lock);
L
Linus Torvalds 已提交
5633
	}
5634

5635
	rq->stop = stop;
5636
}
L
Linus Torvalds 已提交
5637 5638
#endif /* CONFIG_HOTPLUG_CPU */

5639
void set_rq_online(struct rq *rq)
5640 5641 5642 5643
{
	if (!rq->online) {
		const struct sched_class *class;

5644
		cpumask_set_cpu(rq->cpu, rq->rd->online);
5645 5646 5647 5648 5649 5650 5651 5652 5653
		rq->online = 1;

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

5654
void set_rq_offline(struct rq *rq)
5655 5656 5657 5658 5659 5660 5661 5662 5663
{
	if (rq->online) {
		const struct sched_class *class;

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

5664
		cpumask_clear_cpu(rq->cpu, rq->rd->online);
5665 5666 5667 5668
		rq->online = 0;
	}
}

5669
static void set_cpu_rq_start_time(unsigned int cpu)
L
Linus Torvalds 已提交
5670
{
5671
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
5672

5673 5674 5675
	rq->age_stamp = sched_clock_cpu(cpu);
}

I
Ingo Molnar 已提交
5676 5677 5678 5679
/*
 * used to mark begin/end of suspend/resume:
 */
static int num_cpus_frozen;
5680

L
Linus Torvalds 已提交
5681
/*
5682 5683 5684
 * Update cpusets according to cpu_active mask.  If cpusets are
 * disabled, cpuset_update_active_cpus() becomes a simple wrapper
 * around partition_sched_domains().
5685 5686 5687
 *
 * 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
5688
 */
5689
static void cpuset_cpu_active(void)
5690
{
5691
	if (cpuhp_tasks_frozen) {
5692 5693 5694 5695 5696 5697 5698 5699 5700
		/*
		 * 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);
5701
			return;
5702 5703 5704 5705 5706 5707
		}
		/*
		 * This is the last CPU online operation. So fall through and
		 * restore the original sched domains by considering the
		 * cpuset configurations.
		 */
5708
	}
5709
	cpuset_update_active_cpus(true);
5710
}
5711

5712
static int cpuset_cpu_inactive(unsigned int cpu)
5713
{
5714 5715
	unsigned long flags;
	struct dl_bw *dl_b;
5716 5717
	bool overflow;
	int cpus;
5718

5719
	if (!cpuhp_tasks_frozen) {
5720 5721
		rcu_read_lock_sched();
		dl_b = dl_bw_of(cpu);
5722

5723 5724 5725 5726
		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);
5727

5728
		rcu_read_unlock_sched();
5729

5730
		if (overflow)
5731
			return -EBUSY;
5732
		cpuset_update_active_cpus(false);
5733
	} else {
5734 5735
		num_cpus_frozen++;
		partition_sched_domains(1, NULL, NULL);
5736
	}
5737
	return 0;
5738 5739
}

5740
int sched_cpu_activate(unsigned int cpu)
5741
{
5742 5743 5744
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

5745
	set_cpu_active(cpu, true);
5746

5747
	if (sched_smp_initialized) {
5748
		sched_domains_numa_masks_set(cpu);
5749
		cpuset_cpu_active();
5750
	}
5751 5752 5753 5754 5755

	/*
	 * 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
5756
	 *    after all CPUs have been brought up.
5757 5758 5759 5760 5761 5762 5763 5764 5765 5766 5767 5768 5769
	 *
	 * 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();

5770
	return 0;
5771 5772
}

5773
int sched_cpu_deactivate(unsigned int cpu)
5774 5775 5776
{
	int ret;

5777
	set_cpu_active(cpu, false);
5778 5779 5780 5781 5782 5783 5784 5785 5786 5787 5788 5789 5790 5791
	/*
	 * 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();
5792 5793 5794 5795 5796 5797 5798 5799

	if (!sched_smp_initialized)
		return 0;

	ret = cpuset_cpu_inactive(cpu);
	if (ret) {
		set_cpu_active(cpu, true);
		return ret;
5800
	}
5801 5802
	sched_domains_numa_masks_clear(cpu);
	return 0;
5803 5804
}

5805 5806 5807 5808 5809 5810 5811 5812
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();
}

5813 5814 5815
int sched_cpu_starting(unsigned int cpu)
{
	set_cpu_rq_start_time(cpu);
5816
	sched_rq_cpu_starting(cpu);
5817
	return 0;
5818 5819
}

5820 5821 5822 5823 5824 5825 5826 5827 5828 5829 5830 5831 5832 5833 5834 5835 5836 5837
#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();
5838
	nohz_balance_exit_idle(cpu);
5839
	hrtick_clear(rq);
5840 5841 5842 5843
	return 0;
}
#endif

P
Peter Zijlstra 已提交
5844 5845 5846 5847 5848 5849 5850 5851 5852 5853 5854 5855 5856 5857 5858 5859
#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
5860 5861
void __init sched_init_smp(void)
{
5862 5863 5864
	cpumask_var_t non_isolated_cpus;

	alloc_cpumask_var(&non_isolated_cpus, GFP_KERNEL);
5865
	alloc_cpumask_var(&fallback_doms, GFP_KERNEL);
5866

5867 5868
	sched_init_numa();

5869 5870
	/*
	 * There's no userspace yet to cause hotplug operations; hence all the
I
Ingo Molnar 已提交
5871
	 * CPU masks are stable and all blatant races in the below code cannot
5872 5873
	 * happen.
	 */
5874
	mutex_lock(&sched_domains_mutex);
5875
	init_sched_domains(cpu_active_mask);
5876 5877 5878
	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);
5879
	mutex_unlock(&sched_domains_mutex);
5880

5881
	/* Move init over to a non-isolated CPU */
5882
	if (set_cpus_allowed_ptr(current, non_isolated_cpus) < 0)
5883
		BUG();
I
Ingo Molnar 已提交
5884
	sched_init_granularity();
5885
	free_cpumask_var(non_isolated_cpus);
5886

5887
	init_sched_rt_class();
5888
	init_sched_dl_class();
P
Peter Zijlstra 已提交
5889 5890

	sched_init_smt();
5891
	sched_clock_init_late();
P
Peter Zijlstra 已提交
5892

5893
	sched_smp_initialized = true;
L
Linus Torvalds 已提交
5894
}
5895 5896 5897

static int __init migration_init(void)
{
5898
	sched_rq_cpu_starting(smp_processor_id());
5899
	return 0;
L
Linus Torvalds 已提交
5900
}
5901 5902
early_initcall(migration_init);

L
Linus Torvalds 已提交
5903 5904 5905
#else
void __init sched_init_smp(void)
{
I
Ingo Molnar 已提交
5906
	sched_init_granularity();
5907
	sched_clock_init_late();
L
Linus Torvalds 已提交
5908 5909 5910 5911 5912 5913 5914 5915 5916 5917
}
#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);
}

5918
#ifdef CONFIG_CGROUP_SCHED
5919 5920 5921 5922
/*
 * Default task group.
 * Every task in system belongs to this group at bootup.
 */
5923
struct task_group root_task_group;
5924
LIST_HEAD(task_groups);
5925 5926 5927

/* Cacheline aligned slab cache for task_group */
static struct kmem_cache *task_group_cache __read_mostly;
5928
#endif
P
Peter Zijlstra 已提交
5929

5930
DECLARE_PER_CPU(cpumask_var_t, load_balance_mask);
5931
DECLARE_PER_CPU(cpumask_var_t, select_idle_mask);
P
Peter Zijlstra 已提交
5932

5933 5934 5935 5936 5937 5938 5939 5940 5941 5942 5943 5944 5945
#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
5946 5947
void __init sched_init(void)
{
I
Ingo Molnar 已提交
5948
	int i, j;
5949 5950
	unsigned long alloc_size = 0, ptr;

5951 5952
	sched_clock_init();

5953 5954 5955
	for (i = 0; i < WAIT_TABLE_SIZE; i++)
		init_waitqueue_head(bit_wait_table + i);

5956 5957 5958 5959 5960 5961 5962
#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) {
5963
		ptr = (unsigned long)kzalloc(alloc_size, GFP_NOWAIT);
5964 5965

#ifdef CONFIG_FAIR_GROUP_SCHED
5966
		root_task_group.se = (struct sched_entity **)ptr;
5967 5968
		ptr += nr_cpu_ids * sizeof(void **);

5969
		root_task_group.cfs_rq = (struct cfs_rq **)ptr;
5970
		ptr += nr_cpu_ids * sizeof(void **);
5971

5972
#endif /* CONFIG_FAIR_GROUP_SCHED */
5973
#ifdef CONFIG_RT_GROUP_SCHED
5974
		root_task_group.rt_se = (struct sched_rt_entity **)ptr;
5975 5976
		ptr += nr_cpu_ids * sizeof(void **);

5977
		root_task_group.rt_rq = (struct rt_rq **)ptr;
5978 5979
		ptr += nr_cpu_ids * sizeof(void **);

5980
#endif /* CONFIG_RT_GROUP_SCHED */
5981
	}
5982
#ifdef CONFIG_CPUMASK_OFFSTACK
5983 5984 5985
	for_each_possible_cpu(i) {
		per_cpu(load_balance_mask, i) = (cpumask_var_t)kzalloc_node(
			cpumask_size(), GFP_KERNEL, cpu_to_node(i));
5986 5987
		per_cpu(select_idle_mask, i) = (cpumask_var_t)kzalloc_node(
			cpumask_size(), GFP_KERNEL, cpu_to_node(i));
5988
	}
5989
#endif /* CONFIG_CPUMASK_OFFSTACK */
I
Ingo Molnar 已提交
5990

I
Ingo Molnar 已提交
5991 5992
	init_rt_bandwidth(&def_rt_bandwidth, global_rt_period(), global_rt_runtime());
	init_dl_bandwidth(&def_dl_bandwidth, global_rt_period(), global_rt_runtime());
5993

G
Gregory Haskins 已提交
5994 5995 5996 5997
#ifdef CONFIG_SMP
	init_defrootdomain();
#endif

5998
#ifdef CONFIG_RT_GROUP_SCHED
5999
	init_rt_bandwidth(&root_task_group.rt_bandwidth,
6000
			global_rt_period(), global_rt_runtime());
6001
#endif /* CONFIG_RT_GROUP_SCHED */
6002

D
Dhaval Giani 已提交
6003
#ifdef CONFIG_CGROUP_SCHED
6004 6005
	task_group_cache = KMEM_CACHE(task_group, 0);

6006 6007
	list_add(&root_task_group.list, &task_groups);
	INIT_LIST_HEAD(&root_task_group.children);
6008
	INIT_LIST_HEAD(&root_task_group.siblings);
6009
	autogroup_init(&init_task);
D
Dhaval Giani 已提交
6010
#endif /* CONFIG_CGROUP_SCHED */
P
Peter Zijlstra 已提交
6011

6012
	for_each_possible_cpu(i) {
6013
		struct rq *rq;
L
Linus Torvalds 已提交
6014 6015

		rq = cpu_rq(i);
6016
		raw_spin_lock_init(&rq->lock);
N
Nick Piggin 已提交
6017
		rq->nr_running = 0;
6018 6019
		rq->calc_load_active = 0;
		rq->calc_load_update = jiffies + LOAD_FREQ;
6020
		init_cfs_rq(&rq->cfs);
6021 6022
		init_rt_rq(&rq->rt);
		init_dl_rq(&rq->dl);
I
Ingo Molnar 已提交
6023
#ifdef CONFIG_FAIR_GROUP_SCHED
6024
		root_task_group.shares = ROOT_TASK_GROUP_LOAD;
P
Peter Zijlstra 已提交
6025
		INIT_LIST_HEAD(&rq->leaf_cfs_rq_list);
6026
		rq->tmp_alone_branch = &rq->leaf_cfs_rq_list;
D
Dhaval Giani 已提交
6027
		/*
I
Ingo Molnar 已提交
6028
		 * How much CPU bandwidth does root_task_group get?
D
Dhaval Giani 已提交
6029 6030
		 *
		 * In case of task-groups formed thr' the cgroup filesystem, it
I
Ingo Molnar 已提交
6031 6032
		 * gets 100% of the CPU resources in the system. This overall
		 * system CPU resource is divided among the tasks of
6033
		 * root_task_group and its child task-groups in a fair manner,
D
Dhaval Giani 已提交
6034 6035 6036
		 * based on each entity's (task or task-group's) weight
		 * (se->load.weight).
		 *
6037
		 * In other words, if root_task_group has 10 tasks of weight
D
Dhaval Giani 已提交
6038
		 * 1024) and two child groups A0 and A1 (of weight 1024 each),
I
Ingo Molnar 已提交
6039
		 * then A0's share of the CPU resource is:
D
Dhaval Giani 已提交
6040
		 *
6041
		 *	A0's bandwidth = 1024 / (10*1024 + 1024 + 1024) = 8.33%
D
Dhaval Giani 已提交
6042
		 *
6043 6044
		 * 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
6045
		 */
6046
		init_cfs_bandwidth(&root_task_group.cfs_bandwidth);
6047
		init_tg_cfs_entry(&root_task_group, &rq->cfs, NULL, i, NULL);
D
Dhaval Giani 已提交
6048 6049 6050
#endif /* CONFIG_FAIR_GROUP_SCHED */

		rq->rt.rt_runtime = def_rt_bandwidth.rt_runtime;
6051
#ifdef CONFIG_RT_GROUP_SCHED
6052
		init_tg_rt_entry(&root_task_group, &rq->rt, NULL, i, NULL);
I
Ingo Molnar 已提交
6053
#endif
L
Linus Torvalds 已提交
6054

I
Ingo Molnar 已提交
6055 6056
		for (j = 0; j < CPU_LOAD_IDX_MAX; j++)
			rq->cpu_load[j] = 0;
6057

L
Linus Torvalds 已提交
6058
#ifdef CONFIG_SMP
N
Nick Piggin 已提交
6059
		rq->sd = NULL;
G
Gregory Haskins 已提交
6060
		rq->rd = NULL;
6061
		rq->cpu_capacity = rq->cpu_capacity_orig = SCHED_CAPACITY_SCALE;
6062
		rq->balance_callback = NULL;
L
Linus Torvalds 已提交
6063
		rq->active_balance = 0;
I
Ingo Molnar 已提交
6064
		rq->next_balance = jiffies;
L
Linus Torvalds 已提交
6065
		rq->push_cpu = 0;
6066
		rq->cpu = i;
6067
		rq->online = 0;
6068 6069
		rq->idle_stamp = 0;
		rq->avg_idle = 2*sysctl_sched_migration_cost;
6070
		rq->max_idle_balance_cost = sysctl_sched_migration_cost;
6071 6072 6073

		INIT_LIST_HEAD(&rq->cfs_tasks);

6074
		rq_attach_root(rq, &def_root_domain);
6075
#ifdef CONFIG_NO_HZ_COMMON
6076
		rq->last_load_update_tick = jiffies;
6077
		rq->nohz_flags = 0;
6078
#endif
6079 6080 6081
#ifdef CONFIG_NO_HZ_FULL
		rq->last_sched_tick = 0;
#endif
6082
#endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
6083
		init_rq_hrtick(rq);
L
Linus Torvalds 已提交
6084 6085 6086
		atomic_set(&rq->nr_iowait, 0);
	}

6087
	set_load_weight(&init_task);
6088

L
Linus Torvalds 已提交
6089 6090 6091 6092 6093 6094 6095 6096 6097 6098 6099 6100 6101
	/*
	 * The boot idle thread does lazy MMU switching as well:
	 */
	atomic_inc(&init_mm.mm_count);
	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());
6102 6103 6104

	calc_load_update = jiffies + LOAD_FREQ;

6105
#ifdef CONFIG_SMP
6106
	zalloc_cpumask_var(&sched_domains_tmpmask, GFP_NOWAIT);
R
Rusty Russell 已提交
6107 6108 6109
	/* May be allocated at isolcpus cmdline parse time */
	if (cpu_isolated_map == NULL)
		zalloc_cpumask_var(&cpu_isolated_map, GFP_NOWAIT);
6110
	idle_thread_set_boot_cpu();
6111
	set_cpu_rq_start_time(smp_processor_id());
6112 6113
#endif
	init_sched_fair_class();
6114

6115 6116
	init_schedstats();

6117
	scheduler_running = 1;
L
Linus Torvalds 已提交
6118 6119
}

6120
#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
6121 6122
static inline int preempt_count_equals(int preempt_offset)
{
6123
	int nested = preempt_count() + rcu_preempt_depth();
6124

A
Arnd Bergmann 已提交
6125
	return (nested == preempt_offset);
6126 6127
}

6128
void __might_sleep(const char *file, int line, int preempt_offset)
L
Linus Torvalds 已提交
6129
{
P
Peter Zijlstra 已提交
6130 6131 6132 6133 6134
	/*
	 * 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.
	 */
6135
	WARN_ONCE(current->state != TASK_RUNNING && current->task_state_change,
P
Peter Zijlstra 已提交
6136 6137 6138 6139
			"do not call blocking ops when !TASK_RUNNING; "
			"state=%lx set at [<%p>] %pS\n",
			current->state,
			(void *)current->task_state_change,
6140
			(void *)current->task_state_change);
P
Peter Zijlstra 已提交
6141

6142 6143 6144 6145 6146
	___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
6147
{
I
Ingo Molnar 已提交
6148 6149 6150
	/* Ratelimiting timestamp: */
	static unsigned long prev_jiffy;

6151
	unsigned long preempt_disable_ip;
L
Linus Torvalds 已提交
6152

I
Ingo Molnar 已提交
6153 6154 6155
	/* WARN_ON_ONCE() by default, no rate limit required: */
	rcu_sleep_check();

6156 6157
	if ((preempt_count_equals(preempt_offset) && !irqs_disabled() &&
	     !is_idle_task(current)) ||
6158
	    system_state != SYSTEM_RUNNING || oops_in_progress)
I
Ingo Molnar 已提交
6159 6160 6161 6162 6163
		return;
	if (time_before(jiffies, prev_jiffy + HZ) && prev_jiffy)
		return;
	prev_jiffy = jiffies;

I
Ingo Molnar 已提交
6164
	/* Save this before calling printk(), since that will clobber it: */
6165 6166
	preempt_disable_ip = get_preempt_disable_ip(current);

P
Peter Zijlstra 已提交
6167 6168 6169 6170 6171 6172 6173
	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
6174

6175 6176 6177
	if (task_stack_end_corrupted(current))
		printk(KERN_EMERG "Thread overran stack, or stack corrupted\n");

I
Ingo Molnar 已提交
6178 6179 6180
	debug_show_held_locks(current);
	if (irqs_disabled())
		print_irqtrace_events(current);
6181 6182
	if (IS_ENABLED(CONFIG_DEBUG_PREEMPT)
	    && !preempt_count_equals(preempt_offset)) {
6183
		pr_err("Preemption disabled at:");
6184
		print_ip_sym(preempt_disable_ip);
6185 6186
		pr_cont("\n");
	}
I
Ingo Molnar 已提交
6187
	dump_stack();
6188
	add_taint(TAINT_WARN, LOCKDEP_STILL_OK);
L
Linus Torvalds 已提交
6189
}
6190
EXPORT_SYMBOL(___might_sleep);
L
Linus Torvalds 已提交
6191 6192 6193
#endif

#ifdef CONFIG_MAGIC_SYSRQ
6194
void normalize_rt_tasks(void)
6195
{
6196
	struct task_struct *g, *p;
6197 6198 6199
	struct sched_attr attr = {
		.sched_policy = SCHED_NORMAL,
	};
L
Linus Torvalds 已提交
6200

6201
	read_lock(&tasklist_lock);
6202
	for_each_process_thread(g, p) {
6203 6204 6205
		/*
		 * Only normalize user tasks:
		 */
6206
		if (p->flags & PF_KTHREAD)
6207 6208
			continue;

6209 6210 6211 6212
		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
6213

6214
		if (!dl_task(p) && !rt_task(p)) {
I
Ingo Molnar 已提交
6215 6216 6217 6218
			/*
			 * Renice negative nice level userspace
			 * tasks back to 0:
			 */
6219
			if (task_nice(p) < 0)
I
Ingo Molnar 已提交
6220
				set_user_nice(p, 0);
L
Linus Torvalds 已提交
6221
			continue;
I
Ingo Molnar 已提交
6222
		}
L
Linus Torvalds 已提交
6223

6224
		__sched_setscheduler(p, &attr, false, false);
6225
	}
6226
	read_unlock(&tasklist_lock);
L
Linus Torvalds 已提交
6227 6228 6229
}

#endif /* CONFIG_MAGIC_SYSRQ */
6230

6231
#if defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB)
6232
/*
6233
 * These functions are only useful for the IA64 MCA handling, or kdb.
6234 6235 6236 6237 6238 6239 6240 6241 6242
 *
 * 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
6243
 * curr_task - return the current task for a given CPU.
6244 6245 6246
 * @cpu: the processor in question.
 *
 * ONLY VALID WHEN THE WHOLE SYSTEM IS STOPPED!
6247 6248
 *
 * Return: The current task for @cpu.
6249
 */
6250
struct task_struct *curr_task(int cpu)
6251 6252 6253 6254
{
	return cpu_curr(cpu);
}

6255 6256 6257
#endif /* defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB) */

#ifdef CONFIG_IA64
6258
/**
I
Ingo Molnar 已提交
6259
 * set_curr_task - set the current task for a given CPU.
6260 6261 6262 6263
 * @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
6264
 * are serviced on a separate stack. It allows the architecture to switch the
I
Ingo Molnar 已提交
6265
 * notion of the current task on a CPU in a non-blocking manner. This function
6266 6267 6268 6269 6270 6271 6272
 * 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!
 */
6273
void ia64_set_curr_task(int cpu, struct task_struct *p)
6274 6275 6276 6277 6278
{
	cpu_curr(cpu) = p;
}

#endif
S
Srivatsa Vaddagiri 已提交
6279

D
Dhaval Giani 已提交
6280
#ifdef CONFIG_CGROUP_SCHED
6281 6282 6283
/* task_group_lock serializes the addition/removal of task groups */
static DEFINE_SPINLOCK(task_group_lock);

6284
static void sched_free_group(struct task_group *tg)
6285 6286 6287
{
	free_fair_sched_group(tg);
	free_rt_sched_group(tg);
6288
	autogroup_free(tg);
6289
	kmem_cache_free(task_group_cache, tg);
6290 6291 6292
}

/* allocate runqueue etc for a new task group */
6293
struct task_group *sched_create_group(struct task_group *parent)
6294 6295 6296
{
	struct task_group *tg;

6297
	tg = kmem_cache_alloc(task_group_cache, GFP_KERNEL | __GFP_ZERO);
6298 6299 6300
	if (!tg)
		return ERR_PTR(-ENOMEM);

6301
	if (!alloc_fair_sched_group(tg, parent))
6302 6303
		goto err;

6304
	if (!alloc_rt_sched_group(tg, parent))
6305 6306
		goto err;

6307 6308 6309
	return tg;

err:
6310
	sched_free_group(tg);
6311 6312 6313 6314 6315 6316 6317
	return ERR_PTR(-ENOMEM);
}

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

6318
	spin_lock_irqsave(&task_group_lock, flags);
P
Peter Zijlstra 已提交
6319
	list_add_rcu(&tg->list, &task_groups);
P
Peter Zijlstra 已提交
6320

I
Ingo Molnar 已提交
6321 6322
	/* Root should already exist: */
	WARN_ON(!parent);
P
Peter Zijlstra 已提交
6323 6324 6325

	tg->parent = parent;
	INIT_LIST_HEAD(&tg->children);
6326
	list_add_rcu(&tg->siblings, &parent->children);
6327
	spin_unlock_irqrestore(&task_group_lock, flags);
6328 6329

	online_fair_sched_group(tg);
S
Srivatsa Vaddagiri 已提交
6330 6331
}

6332
/* rcu callback to free various structures associated with a task group */
6333
static void sched_free_group_rcu(struct rcu_head *rhp)
S
Srivatsa Vaddagiri 已提交
6334
{
I
Ingo Molnar 已提交
6335
	/* Now it should be safe to free those cfs_rqs: */
6336
	sched_free_group(container_of(rhp, struct task_group, rcu));
S
Srivatsa Vaddagiri 已提交
6337 6338
}

6339
void sched_destroy_group(struct task_group *tg)
6340
{
I
Ingo Molnar 已提交
6341
	/* Wait for possible concurrent references to cfs_rqs complete: */
6342
	call_rcu(&tg->rcu, sched_free_group_rcu);
6343 6344 6345
}

void sched_offline_group(struct task_group *tg)
S
Srivatsa Vaddagiri 已提交
6346
{
6347
	unsigned long flags;
S
Srivatsa Vaddagiri 已提交
6348

I
Ingo Molnar 已提交
6349
	/* End participation in shares distribution: */
6350
	unregister_fair_sched_group(tg);
6351 6352

	spin_lock_irqsave(&task_group_lock, flags);
P
Peter Zijlstra 已提交
6353
	list_del_rcu(&tg->list);
P
Peter Zijlstra 已提交
6354
	list_del_rcu(&tg->siblings);
6355
	spin_unlock_irqrestore(&task_group_lock, flags);
S
Srivatsa Vaddagiri 已提交
6356 6357
}

6358
static void sched_change_group(struct task_struct *tsk, int type)
S
Srivatsa Vaddagiri 已提交
6359
{
P
Peter Zijlstra 已提交
6360
	struct task_group *tg;
S
Srivatsa Vaddagiri 已提交
6361

6362 6363 6364 6365 6366 6367
	/*
	 * 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
6368 6369 6370 6371
			  struct task_group, css);
	tg = autogroup_task_group(tsk, tg);
	tsk->sched_task_group = tg;

P
Peter Zijlstra 已提交
6372
#ifdef CONFIG_FAIR_GROUP_SCHED
6373 6374
	if (tsk->sched_class->task_change_group)
		tsk->sched_class->task_change_group(tsk, type);
6375
	else
P
Peter Zijlstra 已提交
6376
#endif
6377
		set_task_rq(tsk, task_cpu(tsk));
6378 6379 6380 6381 6382 6383 6384 6385 6386 6387 6388 6389 6390 6391 6392 6393
}

/*
 * 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);
6394
	update_rq_clock(rq);
6395 6396 6397 6398 6399 6400

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

	if (queued)
		dequeue_task(rq, tsk, DEQUEUE_SAVE | DEQUEUE_MOVE);
6401
	if (running)
6402 6403 6404
		put_prev_task(rq, tsk);

	sched_change_group(tsk, TASK_MOVE_GROUP);
P
Peter Zijlstra 已提交
6405

6406
	if (queued)
6407
		enqueue_task(rq, tsk, ENQUEUE_RESTORE | ENQUEUE_MOVE);
6408
	if (running)
6409
		set_curr_task(rq, tsk);
S
Srivatsa Vaddagiri 已提交
6410

6411
	task_rq_unlock(rq, tsk, &rf);
S
Srivatsa Vaddagiri 已提交
6412
}
D
Dhaval Giani 已提交
6413
#endif /* CONFIG_CGROUP_SCHED */
S
Srivatsa Vaddagiri 已提交
6414

6415 6416 6417 6418 6419
#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
6420

P
Peter Zijlstra 已提交
6421 6422
/* Must be called with tasklist_lock held */
static inline int tg_has_rt_tasks(struct task_group *tg)
6423
{
P
Peter Zijlstra 已提交
6424
	struct task_struct *g, *p;
6425

6426 6427 6428 6429 6430 6431
	/*
	 * Autogroups do not have RT tasks; see autogroup_create().
	 */
	if (task_group_is_autogroup(tg))
		return 0;

6432
	for_each_process_thread(g, p) {
6433
		if (rt_task(p) && task_group(p) == tg)
P
Peter Zijlstra 已提交
6434
			return 1;
6435
	}
6436

P
Peter Zijlstra 已提交
6437 6438
	return 0;
}
6439

P
Peter Zijlstra 已提交
6440 6441 6442 6443 6444
struct rt_schedulable_data {
	struct task_group *tg;
	u64 rt_period;
	u64 rt_runtime;
};
6445

6446
static int tg_rt_schedulable(struct task_group *tg, void *data)
P
Peter Zijlstra 已提交
6447 6448 6449 6450 6451
{
	struct rt_schedulable_data *d = data;
	struct task_group *child;
	unsigned long total, sum = 0;
	u64 period, runtime;
6452

P
Peter Zijlstra 已提交
6453 6454
	period = ktime_to_ns(tg->rt_bandwidth.rt_period);
	runtime = tg->rt_bandwidth.rt_runtime;
6455

P
Peter Zijlstra 已提交
6456 6457 6458
	if (tg == d->tg) {
		period = d->rt_period;
		runtime = d->rt_runtime;
6459 6460
	}

6461 6462 6463 6464 6465
	/*
	 * Cannot have more runtime than the period.
	 */
	if (runtime > period && runtime != RUNTIME_INF)
		return -EINVAL;
P
Peter Zijlstra 已提交
6466

6467 6468 6469
	/*
	 * Ensure we don't starve existing RT tasks.
	 */
P
Peter Zijlstra 已提交
6470 6471
	if (rt_bandwidth_enabled() && !runtime && tg_has_rt_tasks(tg))
		return -EBUSY;
P
Peter Zijlstra 已提交
6472

P
Peter Zijlstra 已提交
6473
	total = to_ratio(period, runtime);
P
Peter Zijlstra 已提交
6474

6475 6476 6477 6478 6479
	/*
	 * 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
6480

6481 6482 6483
	/*
	 * The sum of our children's runtime should not exceed our own.
	 */
P
Peter Zijlstra 已提交
6484 6485 6486
	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
6487

P
Peter Zijlstra 已提交
6488 6489 6490 6491
		if (child == d->tg) {
			period = d->rt_period;
			runtime = d->rt_runtime;
		}
P
Peter Zijlstra 已提交
6492

P
Peter Zijlstra 已提交
6493
		sum += to_ratio(period, runtime);
P
Peter Zijlstra 已提交
6494
	}
P
Peter Zijlstra 已提交
6495

P
Peter Zijlstra 已提交
6496 6497 6498 6499
	if (sum > total)
		return -EINVAL;

	return 0;
P
Peter Zijlstra 已提交
6500 6501
}

P
Peter Zijlstra 已提交
6502
static int __rt_schedulable(struct task_group *tg, u64 period, u64 runtime)
6503
{
6504 6505
	int ret;

P
Peter Zijlstra 已提交
6506 6507 6508 6509 6510 6511
	struct rt_schedulable_data data = {
		.tg = tg,
		.rt_period = period,
		.rt_runtime = runtime,
	};

6512 6513 6514 6515 6516
	rcu_read_lock();
	ret = walk_tg_tree(tg_rt_schedulable, tg_nop, &data);
	rcu_read_unlock();

	return ret;
6517 6518
}

6519
static int tg_set_rt_bandwidth(struct task_group *tg,
6520
		u64 rt_period, u64 rt_runtime)
P
Peter Zijlstra 已提交
6521
{
P
Peter Zijlstra 已提交
6522
	int i, err = 0;
P
Peter Zijlstra 已提交
6523

6524 6525 6526 6527 6528 6529 6530 6531 6532 6533 6534
	/*
	 * 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
6535
	mutex_lock(&rt_constraints_mutex);
6536
	read_lock(&tasklist_lock);
P
Peter Zijlstra 已提交
6537 6538
	err = __rt_schedulable(tg, rt_period, rt_runtime);
	if (err)
P
Peter Zijlstra 已提交
6539
		goto unlock;
P
Peter Zijlstra 已提交
6540

6541
	raw_spin_lock_irq(&tg->rt_bandwidth.rt_runtime_lock);
6542 6543
	tg->rt_bandwidth.rt_period = ns_to_ktime(rt_period);
	tg->rt_bandwidth.rt_runtime = rt_runtime;
P
Peter Zijlstra 已提交
6544 6545 6546 6547

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

6548
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
6549
		rt_rq->rt_runtime = rt_runtime;
6550
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
6551
	}
6552
	raw_spin_unlock_irq(&tg->rt_bandwidth.rt_runtime_lock);
P
Peter Zijlstra 已提交
6553
unlock:
6554
	read_unlock(&tasklist_lock);
P
Peter Zijlstra 已提交
6555 6556 6557
	mutex_unlock(&rt_constraints_mutex);

	return err;
P
Peter Zijlstra 已提交
6558 6559
}

6560
static int sched_group_set_rt_runtime(struct task_group *tg, long rt_runtime_us)
6561 6562 6563 6564 6565 6566 6567 6568
{
	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;

6569
	return tg_set_rt_bandwidth(tg, rt_period, rt_runtime);
6570 6571
}

6572
static long sched_group_rt_runtime(struct task_group *tg)
P
Peter Zijlstra 已提交
6573 6574 6575
{
	u64 rt_runtime_us;

6576
	if (tg->rt_bandwidth.rt_runtime == RUNTIME_INF)
P
Peter Zijlstra 已提交
6577 6578
		return -1;

6579
	rt_runtime_us = tg->rt_bandwidth.rt_runtime;
P
Peter Zijlstra 已提交
6580 6581 6582
	do_div(rt_runtime_us, NSEC_PER_USEC);
	return rt_runtime_us;
}
6583

6584
static int sched_group_set_rt_period(struct task_group *tg, u64 rt_period_us)
6585 6586 6587
{
	u64 rt_runtime, rt_period;

6588
	rt_period = rt_period_us * NSEC_PER_USEC;
6589 6590
	rt_runtime = tg->rt_bandwidth.rt_runtime;

6591
	return tg_set_rt_bandwidth(tg, rt_period, rt_runtime);
6592 6593
}

6594
static long sched_group_rt_period(struct task_group *tg)
6595 6596 6597 6598 6599 6600 6601
{
	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;
}
6602
#endif /* CONFIG_RT_GROUP_SCHED */
6603

6604
#ifdef CONFIG_RT_GROUP_SCHED
6605 6606 6607 6608 6609
static int sched_rt_global_constraints(void)
{
	int ret = 0;

	mutex_lock(&rt_constraints_mutex);
P
Peter Zijlstra 已提交
6610
	read_lock(&tasklist_lock);
6611
	ret = __rt_schedulable(NULL, 0, 0);
P
Peter Zijlstra 已提交
6612
	read_unlock(&tasklist_lock);
6613 6614 6615 6616
	mutex_unlock(&rt_constraints_mutex);

	return ret;
}
6617

6618
static int sched_rt_can_attach(struct task_group *tg, struct task_struct *tsk)
6619 6620 6621 6622 6623 6624 6625 6626
{
	/* 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;
}

6627
#else /* !CONFIG_RT_GROUP_SCHED */
6628 6629
static int sched_rt_global_constraints(void)
{
P
Peter Zijlstra 已提交
6630
	unsigned long flags;
6631
	int i;
6632

6633
	raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags);
P
Peter Zijlstra 已提交
6634 6635 6636
	for_each_possible_cpu(i) {
		struct rt_rq *rt_rq = &cpu_rq(i)->rt;

6637
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
6638
		rt_rq->rt_runtime = global_rt_runtime();
6639
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
6640
	}
6641
	raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags);
P
Peter Zijlstra 已提交
6642

6643
	return 0;
6644
}
6645
#endif /* CONFIG_RT_GROUP_SCHED */
6646

6647
static int sched_dl_global_validate(void)
6648
{
6649 6650
	u64 runtime = global_rt_runtime();
	u64 period = global_rt_period();
6651
	u64 new_bw = to_ratio(period, runtime);
6652
	struct dl_bw *dl_b;
6653
	int cpu, ret = 0;
6654
	unsigned long flags;
6655 6656 6657 6658 6659 6660 6661 6662 6663 6664

	/*
	 * 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!
	 */
6665
	for_each_possible_cpu(cpu) {
6666 6667
		rcu_read_lock_sched();
		dl_b = dl_bw_of(cpu);
6668

6669
		raw_spin_lock_irqsave(&dl_b->lock, flags);
6670 6671
		if (new_bw < dl_b->total_bw)
			ret = -EBUSY;
6672
		raw_spin_unlock_irqrestore(&dl_b->lock, flags);
6673

6674 6675
		rcu_read_unlock_sched();

6676 6677
		if (ret)
			break;
6678 6679
	}

6680
	return ret;
6681 6682
}

6683
static void sched_dl_do_global(void)
6684
{
6685
	u64 new_bw = -1;
6686
	struct dl_bw *dl_b;
6687
	int cpu;
6688
	unsigned long flags;
6689

6690 6691 6692 6693 6694 6695 6696 6697 6698 6699
	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) {
6700 6701
		rcu_read_lock_sched();
		dl_b = dl_bw_of(cpu);
6702

6703
		raw_spin_lock_irqsave(&dl_b->lock, flags);
6704
		dl_b->bw = new_bw;
6705
		raw_spin_unlock_irqrestore(&dl_b->lock, flags);
6706 6707

		rcu_read_unlock_sched();
6708
	}
6709 6710 6711 6712 6713 6714 6715
}

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

6716 6717
	if ((sysctl_sched_rt_runtime != RUNTIME_INF) &&
		(sysctl_sched_rt_runtime > sysctl_sched_rt_period))
6718 6719 6720 6721 6722 6723 6724 6725 6726
		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());
6727 6728
}

6729
int sched_rt_handler(struct ctl_table *table, int write,
6730
		void __user *buffer, size_t *lenp,
6731 6732 6733 6734
		loff_t *ppos)
{
	int old_period, old_runtime;
	static DEFINE_MUTEX(mutex);
6735
	int ret;
6736 6737 6738 6739 6740

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

6741
	ret = proc_dointvec(table, write, buffer, lenp, ppos);
6742 6743

	if (!ret && write) {
6744 6745 6746 6747
		ret = sched_rt_global_validate();
		if (ret)
			goto undo;

6748
		ret = sched_dl_global_validate();
6749 6750 6751
		if (ret)
			goto undo;

6752
		ret = sched_rt_global_constraints();
6753 6754 6755 6756 6757 6758 6759 6760 6761 6762
		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;
6763 6764 6765 6766 6767
	}
	mutex_unlock(&mutex);

	return ret;
}
6768

6769
int sched_rr_handler(struct ctl_table *table, int write,
6770 6771 6772 6773 6774 6775 6776 6777
		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
6778 6779 6780 6781
	/*
	 * Make sure that internally we keep jiffies.
	 * Also, writing zero resets the timeslice to default:
	 */
6782
	if (!ret && write) {
6783 6784 6785
		sched_rr_timeslice =
			sysctl_sched_rr_timeslice <= 0 ? RR_TIMESLICE :
			msecs_to_jiffies(sysctl_sched_rr_timeslice);
6786 6787 6788 6789 6790
	}
	mutex_unlock(&mutex);
	return ret;
}

6791
#ifdef CONFIG_CGROUP_SCHED
6792

6793
static inline struct task_group *css_tg(struct cgroup_subsys_state *css)
6794
{
6795
	return css ? container_of(css, struct task_group, css) : NULL;
6796 6797
}

6798 6799
static struct cgroup_subsys_state *
cpu_cgroup_css_alloc(struct cgroup_subsys_state *parent_css)
6800
{
6801 6802
	struct task_group *parent = css_tg(parent_css);
	struct task_group *tg;
6803

6804
	if (!parent) {
6805
		/* This is early initialization for the top cgroup */
6806
		return &root_task_group.css;
6807 6808
	}

6809
	tg = sched_create_group(parent);
6810 6811 6812
	if (IS_ERR(tg))
		return ERR_PTR(-ENOMEM);

6813 6814
	sched_online_group(tg, parent);

6815 6816 6817
	return &tg->css;
}

6818
static void cpu_cgroup_css_released(struct cgroup_subsys_state *css)
6819
{
6820
	struct task_group *tg = css_tg(css);
6821

6822
	sched_offline_group(tg);
6823 6824
}

6825
static void cpu_cgroup_css_free(struct cgroup_subsys_state *css)
6826
{
6827
	struct task_group *tg = css_tg(css);
6828

6829 6830 6831 6832
	/*
	 * Relies on the RCU grace period between css_released() and this.
	 */
	sched_free_group(tg);
6833 6834
}

6835 6836 6837 6838
/*
 * 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.
 */
6839
static void cpu_cgroup_fork(struct task_struct *task)
6840
{
6841 6842 6843 6844 6845
	struct rq_flags rf;
	struct rq *rq;

	rq = task_rq_lock(task, &rf);

6846
	update_rq_clock(rq);
6847 6848 6849
	sched_change_group(task, TASK_SET_GROUP);

	task_rq_unlock(rq, task, &rf);
6850 6851
}

6852
static int cpu_cgroup_can_attach(struct cgroup_taskset *tset)
6853
{
6854
	struct task_struct *task;
6855
	struct cgroup_subsys_state *css;
6856
	int ret = 0;
6857

6858
	cgroup_taskset_for_each(task, css, tset) {
6859
#ifdef CONFIG_RT_GROUP_SCHED
6860
		if (!sched_rt_can_attach(css_tg(css), task))
6861
			return -EINVAL;
6862
#else
6863 6864 6865
		/* We don't support RT-tasks being in separate groups */
		if (task->sched_class != &fair_sched_class)
			return -EINVAL;
6866
#endif
6867 6868 6869 6870 6871 6872 6873 6874 6875 6876 6877 6878 6879 6880 6881 6882
		/*
		 * 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;
6883
	}
6884
	return ret;
6885
}
6886

6887
static void cpu_cgroup_attach(struct cgroup_taskset *tset)
6888
{
6889
	struct task_struct *task;
6890
	struct cgroup_subsys_state *css;
6891

6892
	cgroup_taskset_for_each(task, css, tset)
6893
		sched_move_task(task);
6894 6895
}

6896
#ifdef CONFIG_FAIR_GROUP_SCHED
6897 6898
static int cpu_shares_write_u64(struct cgroup_subsys_state *css,
				struct cftype *cftype, u64 shareval)
6899
{
6900
	return sched_group_set_shares(css_tg(css), scale_load(shareval));
6901 6902
}

6903 6904
static u64 cpu_shares_read_u64(struct cgroup_subsys_state *css,
			       struct cftype *cft)
6905
{
6906
	struct task_group *tg = css_tg(css);
6907

6908
	return (u64) scale_load_down(tg->shares);
6909
}
6910 6911

#ifdef CONFIG_CFS_BANDWIDTH
6912 6913
static DEFINE_MUTEX(cfs_constraints_mutex);

6914 6915 6916
const u64 max_cfs_quota_period = 1 * NSEC_PER_SEC; /* 1s */
const u64 min_cfs_quota_period = 1 * NSEC_PER_MSEC; /* 1ms */

6917 6918
static int __cfs_schedulable(struct task_group *tg, u64 period, u64 runtime);

6919 6920
static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota)
{
6921
	int i, ret = 0, runtime_enabled, runtime_was_enabled;
6922
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
6923 6924 6925 6926 6927 6928 6929 6930 6931 6932 6933 6934 6935 6936 6937 6938 6939 6940 6941 6942

	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;

6943 6944 6945 6946 6947
	/*
	 * Prevent race between setting of cfs_rq->runtime_enabled and
	 * unthrottle_offline_cfs_rqs().
	 */
	get_online_cpus();
6948 6949 6950 6951 6952
	mutex_lock(&cfs_constraints_mutex);
	ret = __cfs_schedulable(tg, period, quota);
	if (ret)
		goto out_unlock;

6953
	runtime_enabled = quota != RUNTIME_INF;
6954
	runtime_was_enabled = cfs_b->quota != RUNTIME_INF;
6955 6956 6957 6958 6959 6960
	/*
	 * 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();
6961 6962 6963
	raw_spin_lock_irq(&cfs_b->lock);
	cfs_b->period = ns_to_ktime(period);
	cfs_b->quota = quota;
6964

P
Paul Turner 已提交
6965
	__refill_cfs_bandwidth_runtime(cfs_b);
I
Ingo Molnar 已提交
6966 6967

	/* Restart the period timer (if active) to handle new period expiry: */
P
Peter Zijlstra 已提交
6968 6969
	if (runtime_enabled)
		start_cfs_bandwidth(cfs_b);
I
Ingo Molnar 已提交
6970

6971 6972
	raw_spin_unlock_irq(&cfs_b->lock);

6973
	for_each_online_cpu(i) {
6974
		struct cfs_rq *cfs_rq = tg->cfs_rq[i];
6975
		struct rq *rq = cfs_rq->rq;
6976 6977

		raw_spin_lock_irq(&rq->lock);
6978
		cfs_rq->runtime_enabled = runtime_enabled;
6979
		cfs_rq->runtime_remaining = 0;
6980

6981
		if (cfs_rq->throttled)
6982
			unthrottle_cfs_rq(cfs_rq);
6983 6984
		raw_spin_unlock_irq(&rq->lock);
	}
6985 6986
	if (runtime_was_enabled && !runtime_enabled)
		cfs_bandwidth_usage_dec();
6987 6988
out_unlock:
	mutex_unlock(&cfs_constraints_mutex);
6989
	put_online_cpus();
6990

6991
	return ret;
6992 6993 6994 6995 6996 6997
}

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

6998
	period = ktime_to_ns(tg->cfs_bandwidth.period);
6999 7000 7001 7002 7003 7004 7005 7006 7007 7008 7009 7010
	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;

7011
	if (tg->cfs_bandwidth.quota == RUNTIME_INF)
7012 7013
		return -1;

7014
	quota_us = tg->cfs_bandwidth.quota;
7015 7016 7017 7018 7019 7020 7021 7022 7023 7024
	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;
7025
	quota = tg->cfs_bandwidth.quota;
7026 7027 7028 7029 7030 7031 7032 7033

	return tg_set_cfs_bandwidth(tg, period, quota);
}

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

7034
	cfs_period_us = ktime_to_ns(tg->cfs_bandwidth.period);
7035 7036 7037 7038 7039
	do_div(cfs_period_us, NSEC_PER_USEC);

	return cfs_period_us;
}

7040 7041
static s64 cpu_cfs_quota_read_s64(struct cgroup_subsys_state *css,
				  struct cftype *cft)
7042
{
7043
	return tg_get_cfs_quota(css_tg(css));
7044 7045
}

7046 7047
static int cpu_cfs_quota_write_s64(struct cgroup_subsys_state *css,
				   struct cftype *cftype, s64 cfs_quota_us)
7048
{
7049
	return tg_set_cfs_quota(css_tg(css), cfs_quota_us);
7050 7051
}

7052 7053
static u64 cpu_cfs_period_read_u64(struct cgroup_subsys_state *css,
				   struct cftype *cft)
7054
{
7055
	return tg_get_cfs_period(css_tg(css));
7056 7057
}

7058 7059
static int cpu_cfs_period_write_u64(struct cgroup_subsys_state *css,
				    struct cftype *cftype, u64 cfs_period_us)
7060
{
7061
	return tg_set_cfs_period(css_tg(css), cfs_period_us);
7062 7063
}

7064 7065 7066 7067 7068 7069 7070 7071 7072 7073 7074 7075 7076 7077 7078 7079 7080 7081 7082 7083 7084 7085 7086 7087 7088 7089 7090 7091 7092 7093 7094 7095
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;
7096
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
7097 7098 7099 7100 7101
	s64 quota = 0, parent_quota = -1;

	if (!tg->parent) {
		quota = RUNTIME_INF;
	} else {
7102
		struct cfs_bandwidth *parent_b = &tg->parent->cfs_bandwidth;
7103 7104

		quota = normalize_cfs_quota(tg, d);
7105
		parent_quota = parent_b->hierarchical_quota;
7106 7107

		/*
I
Ingo Molnar 已提交
7108 7109
		 * Ensure max(child_quota) <= parent_quota, inherit when no
		 * limit is set:
7110 7111 7112 7113 7114 7115
		 */
		if (quota == RUNTIME_INF)
			quota = parent_quota;
		else if (parent_quota != RUNTIME_INF && quota > parent_quota)
			return -EINVAL;
	}
7116
	cfs_b->hierarchical_quota = quota;
7117 7118 7119 7120 7121 7122

	return 0;
}

static int __cfs_schedulable(struct task_group *tg, u64 period, u64 quota)
{
7123
	int ret;
7124 7125 7126 7127 7128 7129 7130 7131 7132 7133 7134
	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);
	}

7135 7136 7137 7138 7139
	rcu_read_lock();
	ret = walk_tg_tree(tg_cfs_schedulable_down, tg_nop, &data);
	rcu_read_unlock();

	return ret;
7140
}
7141

7142
static int cpu_stats_show(struct seq_file *sf, void *v)
7143
{
7144
	struct task_group *tg = css_tg(seq_css(sf));
7145
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
7146

7147 7148 7149
	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);
7150 7151 7152

	return 0;
}
7153
#endif /* CONFIG_CFS_BANDWIDTH */
7154
#endif /* CONFIG_FAIR_GROUP_SCHED */
7155

7156
#ifdef CONFIG_RT_GROUP_SCHED
7157 7158
static int cpu_rt_runtime_write(struct cgroup_subsys_state *css,
				struct cftype *cft, s64 val)
P
Peter Zijlstra 已提交
7159
{
7160
	return sched_group_set_rt_runtime(css_tg(css), val);
P
Peter Zijlstra 已提交
7161 7162
}

7163 7164
static s64 cpu_rt_runtime_read(struct cgroup_subsys_state *css,
			       struct cftype *cft)
P
Peter Zijlstra 已提交
7165
{
7166
	return sched_group_rt_runtime(css_tg(css));
P
Peter Zijlstra 已提交
7167
}
7168

7169 7170
static int cpu_rt_period_write_uint(struct cgroup_subsys_state *css,
				    struct cftype *cftype, u64 rt_period_us)
7171
{
7172
	return sched_group_set_rt_period(css_tg(css), rt_period_us);
7173 7174
}

7175 7176
static u64 cpu_rt_period_read_uint(struct cgroup_subsys_state *css,
				   struct cftype *cft)
7177
{
7178
	return sched_group_rt_period(css_tg(css));
7179
}
7180
#endif /* CONFIG_RT_GROUP_SCHED */
P
Peter Zijlstra 已提交
7181

7182
static struct cftype cpu_files[] = {
7183
#ifdef CONFIG_FAIR_GROUP_SCHED
7184 7185
	{
		.name = "shares",
7186 7187
		.read_u64 = cpu_shares_read_u64,
		.write_u64 = cpu_shares_write_u64,
7188
	},
7189
#endif
7190 7191 7192 7193 7194 7195 7196 7197 7198 7199 7200
#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,
	},
7201 7202
	{
		.name = "stat",
7203
		.seq_show = cpu_stats_show,
7204
	},
7205
#endif
7206
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
7207
	{
P
Peter Zijlstra 已提交
7208
		.name = "rt_runtime_us",
7209 7210
		.read_s64 = cpu_rt_runtime_read,
		.write_s64 = cpu_rt_runtime_write,
P
Peter Zijlstra 已提交
7211
	},
7212 7213
	{
		.name = "rt_period_us",
7214 7215
		.read_u64 = cpu_rt_period_read_uint,
		.write_u64 = cpu_rt_period_write_uint,
7216
	},
7217
#endif
I
Ingo Molnar 已提交
7218
	{ }	/* Terminate */
7219 7220
};

7221
struct cgroup_subsys cpu_cgrp_subsys = {
7222
	.css_alloc	= cpu_cgroup_css_alloc,
7223
	.css_released	= cpu_cgroup_css_released,
7224
	.css_free	= cpu_cgroup_css_free,
7225
	.fork		= cpu_cgroup_fork,
7226 7227
	.can_attach	= cpu_cgroup_can_attach,
	.attach		= cpu_cgroup_attach,
7228
	.legacy_cftypes	= cpu_files,
7229
	.early_init	= true,
7230 7231
};

7232
#endif	/* CONFIG_CGROUP_SCHED */
7233

7234 7235 7236 7237 7238
void dump_cpu_task(int cpu)
{
	pr_info("Task dump for CPU %d:\n", cpu);
	sched_show_task(cpu_curr(cpu));
}
7239 7240 7241 7242 7243 7244 7245 7246 7247 7248 7249 7250 7251 7252 7253 7254 7255 7256 7257 7258 7259 7260 7261 7262 7263 7264 7265 7266 7267 7268 7269 7270 7271 7272 7273 7274 7275 7276 7277 7278 7279

/*
 * 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,
};