core.c 180.1 KB
Newer Older
L
Linus Torvalds 已提交
1
/*
2
 *  kernel/sched/core.c
L
Linus Torvalds 已提交
3
 *
I
Ingo Molnar 已提交
4
 *  Core kernel scheduler code and related syscalls
L
Linus Torvalds 已提交
5 6 7
 *
 *  Copyright (C) 1991-2002  Linus Torvalds
 */
8
#include <linux/sched.h>
9
#include <linux/sched/clock.h>
10
#include <uapi/linux/sched/types.h>
11
#include <linux/sched/loadavg.h>
12
#include <linux/sched/hotplug.h>
L
Linus Torvalds 已提交
13
#include <linux/cpuset.h>
14
#include <linux/delayacct.h>
15
#include <linux/init_task.h>
16
#include <linux/context_tracking.h>
17
#include <linux/rcupdate_wait.h>
18 19 20 21 22 23

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

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

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

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

42
DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
43

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

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

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

#undef SCHED_FEAT

57 58 59 60 61 62
/*
 * Number of tasks to iterate in a single balance run.
 * Limited because this is done with IRQs disabled.
 */
const_debug unsigned int sysctl_sched_nr_migrate = 32;

63 64 65 66 67 68 69 70
/*
 * period over which we average the RT time consumption, measured
 * in ms.
 *
 * default: 1s
 */
const_debug unsigned int sysctl_sched_time_avg = MSEC_PER_SEC;

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

77
__read_mostly int scheduler_running;
78

P
Peter Zijlstra 已提交
79 80 81 82 83
/*
 * part of the period that we allow rt tasks to run in us.
 * default: 0.95s
 */
int sysctl_sched_rt_runtime = 950000;
P
Peter Zijlstra 已提交
84

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

88 89 90
/*
 * __task_rq_lock - lock the rq @p resides on.
 */
91
struct rq *__task_rq_lock(struct task_struct *p, struct rq_flags *rf)
92 93 94 95 96 97 98 99 100 101
	__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))) {
102
			rq_pin_lock(rq, rf);
103 104 105 106 107 108 109 110 111 112 113 114
			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.
 */
115
struct rq *task_rq_lock(struct task_struct *p, struct rq_flags *rf)
116 117 118 119 120 121
	__acquires(p->pi_lock)
	__acquires(rq->lock)
{
	struct rq *rq;

	for (;;) {
122
		raw_spin_lock_irqsave(&p->pi_lock, rf->flags);
123 124 125 126 127 128 129 130 131 132 133 134 135 136 137
		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
138
		 * If we observe the new CPU in task_rq_lock, the acquire will
139 140 141
		 * pair with the WMB to ensure we must then also see migrating.
		 */
		if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) {
142
			rq_pin_lock(rq, rf);
143 144 145
			return rq;
		}
		raw_spin_unlock(&rq->lock);
146
		raw_spin_unlock_irqrestore(&p->pi_lock, rf->flags);
147 148 149 150 151 152

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

153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220
/*
 * 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
221 222
	if (sched_feat(WARN_DOUBLE_CLOCK))
		SCHED_WARN_ON(rq->clock_update_flags & RQCF_UPDATED);
223 224
	rq->clock_update_flags |= RQCF_UPDATED;
#endif
225

226 227 228 229 230 231 232 233
	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
234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251
#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);
252
	struct rq_flags rf;
P
Peter Zijlstra 已提交
253 254 255

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

256
	rq_lock(rq, &rf);
257
	update_rq_clock(rq);
P
Peter Zijlstra 已提交
258
	rq->curr->sched_class->task_tick(rq, rq->curr, 1);
259
	rq_unlock(rq, &rf);
P
Peter Zijlstra 已提交
260 261 262 263

	return HRTIMER_NORESTART;
}

264
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
265

266
static void __hrtick_restart(struct rq *rq)
P
Peter Zijlstra 已提交
267 268 269
{
	struct hrtimer *timer = &rq->hrtick_timer;

270
	hrtimer_start_expires(timer, HRTIMER_MODE_ABS_PINNED);
P
Peter Zijlstra 已提交
271 272
}

273 274 275 276
/*
 * called from hardirq (IPI) context
 */
static void __hrtick_start(void *arg)
277
{
278
	struct rq *rq = arg;
279
	struct rq_flags rf;
280

281
	rq_lock(rq, &rf);
P
Peter Zijlstra 已提交
282
	__hrtick_restart(rq);
283
	rq->hrtick_csd_pending = 0;
284
	rq_unlock(rq, &rf);
285 286
}

287 288 289 290 291
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
292
void hrtick_start(struct rq *rq, u64 delay)
293
{
294
	struct hrtimer *timer = &rq->hrtick_timer;
295 296 297 298 299 300 301 302 303
	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);
304

305
	hrtimer_set_expires(timer, time);
306 307

	if (rq == this_rq()) {
P
Peter Zijlstra 已提交
308
		__hrtick_restart(rq);
309
	} else if (!rq->hrtick_csd_pending) {
310
		smp_call_function_single_async(cpu_of(rq), &rq->hrtick_csd);
311 312
		rq->hrtick_csd_pending = 1;
	}
313 314
}

315 316 317 318 319 320
#else
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
321
void hrtick_start(struct rq *rq, u64 delay)
322
{
W
Wanpeng Li 已提交
323 324 325 326 327
	/*
	 * Don't schedule slices shorter than 10000ns, that just
	 * doesn't make sense. Rely on vruntime for fairness.
	 */
	delay = max_t(u64, delay, 10000LL);
328 329
	hrtimer_start(&rq->hrtick_timer, ns_to_ktime(delay),
		      HRTIMER_MODE_REL_PINNED);
330 331
}
#endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
332

333
static void init_rq_hrtick(struct rq *rq)
P
Peter Zijlstra 已提交
334
{
335 336
#ifdef CONFIG_SMP
	rq->hrtick_csd_pending = 0;
P
Peter Zijlstra 已提交
337

338 339 340 341
	rq->hrtick_csd.flags = 0;
	rq->hrtick_csd.func = __hrtick_start;
	rq->hrtick_csd.info = rq;
#endif
P
Peter Zijlstra 已提交
342

343 344
	hrtimer_init(&rq->hrtick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	rq->hrtick_timer.function = hrtick;
P
Peter Zijlstra 已提交
345
}
A
Andrew Morton 已提交
346
#else	/* CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
347 348 349 350 351 352 353
static inline void hrtick_clear(struct rq *rq)
{
}

static inline void init_rq_hrtick(struct rq *rq)
{
}
A
Andrew Morton 已提交
354
#endif	/* CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
355

356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373
/*
 * 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;								\
})

374
#if defined(CONFIG_SMP) && defined(TIF_POLLING_NRFLAG)
375 376 377 378 379 380 381 382 383 384
/*
 * 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);
}
385 386 387 388 389 390 391 392 393 394

/*
 * 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);
395
	typeof(ti->flags) old, val = READ_ONCE(ti->flags);
396 397 398 399 400 401 402 403 404 405 406 407 408 409

	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;
}

410 411 412 413 414 415
#else
static bool set_nr_and_not_polling(struct task_struct *p)
{
	set_tsk_need_resched(p);
	return true;
}
416 417 418 419 420 421 422

#ifdef CONFIG_SMP
static bool set_nr_if_polling(struct task_struct *p)
{
	return false;
}
#endif
423 424
#endif

425 426 427 428 429 430 431 432 433 434
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
435
	 * barrier implied by the wakeup in wake_up_q().
436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457
	 */
	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
458
		/* Task can safely be re-inserted now: */
459 460 461 462 463 464 465 466 467 468 469 470
		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
471
/*
472
 * resched_curr - mark rq's current task 'to be rescheduled now'.
I
Ingo Molnar 已提交
473 474 475 476 477
 *
 * 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.
 */
478
void resched_curr(struct rq *rq)
I
Ingo Molnar 已提交
479
{
480
	struct task_struct *curr = rq->curr;
I
Ingo Molnar 已提交
481 482
	int cpu;

483
	lockdep_assert_held(&rq->lock);
I
Ingo Molnar 已提交
484

485
	if (test_tsk_need_resched(curr))
I
Ingo Molnar 已提交
486 487
		return;

488
	cpu = cpu_of(rq);
489

490
	if (cpu == smp_processor_id()) {
491
		set_tsk_need_resched(curr);
492
		set_preempt_need_resched();
I
Ingo Molnar 已提交
493
		return;
494
	}
I
Ingo Molnar 已提交
495

496
	if (set_nr_and_not_polling(curr))
I
Ingo Molnar 已提交
497
		smp_send_reschedule(cpu);
498 499
	else
		trace_sched_wake_idle_without_ipi(cpu);
I
Ingo Molnar 已提交
500 501
}

502
void resched_cpu(int cpu)
I
Ingo Molnar 已提交
503 504 505 506
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

507
	if (!raw_spin_trylock_irqsave(&rq->lock, flags))
I
Ingo Molnar 已提交
508
		return;
509
	resched_curr(rq);
510
	raw_spin_unlock_irqrestore(&rq->lock, flags);
I
Ingo Molnar 已提交
511
}
512

513
#ifdef CONFIG_SMP
514
#ifdef CONFIG_NO_HZ_COMMON
515
/*
I
Ingo Molnar 已提交
516 517
 * In the semi idle case, use the nearest busy CPU for migrating timers
 * from an idle CPU.  This is good for power-savings.
518 519
 *
 * We don't do similar optimization for completely idle system, as
I
Ingo Molnar 已提交
520 521
 * 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).
522
 */
523
int get_nohz_timer_target(void)
524
{
525
	int i, cpu = smp_processor_id();
526 527
	struct sched_domain *sd;

528
	if (!idle_cpu(cpu) && is_housekeeping_cpu(cpu))
529 530
		return cpu;

531
	rcu_read_lock();
532
	for_each_domain(cpu, sd) {
533
		for_each_cpu(i, sched_domain_span(sd)) {
534 535 536 537
			if (cpu == i)
				continue;

			if (!idle_cpu(i) && is_housekeeping_cpu(i)) {
538 539 540 541
				cpu = i;
				goto unlock;
			}
		}
542
	}
543 544 545

	if (!is_housekeeping_cpu(cpu))
		cpu = housekeeping_any_cpu();
546 547
unlock:
	rcu_read_unlock();
548 549
	return cpu;
}
I
Ingo Molnar 已提交
550

551 552 553 554 555 556 557 558 559 560
/*
 * 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.
 */
561
static void wake_up_idle_cpu(int cpu)
562 563 564 565 566 567
{
	struct rq *rq = cpu_rq(cpu);

	if (cpu == smp_processor_id())
		return;

568
	if (set_nr_and_not_polling(rq->idle))
569
		smp_send_reschedule(cpu);
570 571
	else
		trace_sched_wake_idle_without_ipi(cpu);
572 573
}

574
static bool wake_up_full_nohz_cpu(int cpu)
575
{
576 577 578 579 580 581
	/*
	 * 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.
	 */
582 583
	if (cpu_is_offline(cpu))
		return true;  /* Don't try to wake offline CPUs. */
584
	if (tick_nohz_full_cpu(cpu)) {
585 586
		if (cpu != smp_processor_id() ||
		    tick_nohz_tick_stopped())
587
			tick_nohz_full_kick_cpu(cpu);
588 589 590 591 592 593
		return true;
	}

	return false;
}

594 595 596 597 598
/*
 * 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.
 */
599 600
void wake_up_nohz_cpu(int cpu)
{
601
	if (!wake_up_full_nohz_cpu(cpu))
602 603 604
		wake_up_idle_cpu(cpu);
}

605
static inline bool got_nohz_idle_kick(void)
606
{
607
	int cpu = smp_processor_id();
608 609 610 611 612 613 614 615 616 617 618 619 620

	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;
621 622
}

623
#else /* CONFIG_NO_HZ_COMMON */
624

625
static inline bool got_nohz_idle_kick(void)
P
Peter Zijlstra 已提交
626
{
627
	return false;
P
Peter Zijlstra 已提交
628 629
}

630
#endif /* CONFIG_NO_HZ_COMMON */
631

632
#ifdef CONFIG_NO_HZ_FULL
633
bool sched_can_stop_tick(struct rq *rq)
634
{
635 636 637 638 639 640
	int fifo_nr_running;

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

641
	/*
642 643
	 * If there are more than one RR tasks, we need the tick to effect the
	 * actual RR behaviour.
644
	 */
645 646 647 648 649
	if (rq->rt.rr_nr_running) {
		if (rq->rt.rr_nr_running == 1)
			return true;
		else
			return false;
650 651
	}

652 653 654 655 656 657 658 659 660 661 662 663 664 665
	/*
	 * 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)
666
		return false;
667

668
	return true;
669 670
}
#endif /* CONFIG_NO_HZ_FULL */
671

672
void sched_avg_update(struct rq *rq)
673
{
674 675
	s64 period = sched_avg_period();

676
	while ((s64)(rq_clock(rq) - rq->age_stamp) > period) {
677 678 679 680 681 682
		/*
		 * 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));
683 684 685
		rq->age_stamp += period;
		rq->rt_avg /= 2;
	}
686 687
}

688
#endif /* CONFIG_SMP */
689

690 691
#if defined(CONFIG_RT_GROUP_SCHED) || (defined(CONFIG_FAIR_GROUP_SCHED) && \
			(defined(CONFIG_SMP) || defined(CONFIG_CFS_BANDWIDTH)))
692
/*
693 694 695 696
 * 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.
697
 */
698
int walk_tg_tree_from(struct task_group *from,
699
			     tg_visitor down, tg_visitor up, void *data)
700 701
{
	struct task_group *parent, *child;
P
Peter Zijlstra 已提交
702
	int ret;
703

704 705
	parent = from;

706
down:
P
Peter Zijlstra 已提交
707 708
	ret = (*down)(parent, data);
	if (ret)
709
		goto out;
710 711 712 713 714 715 716
	list_for_each_entry_rcu(child, &parent->children, siblings) {
		parent = child;
		goto down;

up:
		continue;
	}
P
Peter Zijlstra 已提交
717
	ret = (*up)(parent, data);
718 719
	if (ret || parent == from)
		goto out;
720 721 722 723 724

	child = parent;
	parent = parent->parent;
	if (parent)
		goto up;
725
out:
P
Peter Zijlstra 已提交
726
	return ret;
727 728
}

729
int tg_nop(struct task_group *tg, void *data)
P
Peter Zijlstra 已提交
730
{
731
	return 0;
P
Peter Zijlstra 已提交
732
}
733 734
#endif

735 736
static void set_load_weight(struct task_struct *p)
{
N
Nikhil Rao 已提交
737 738 739
	int prio = p->static_prio - MAX_RT_PRIO;
	struct load_weight *load = &p->se.load;

I
Ingo Molnar 已提交
740 741 742
	/*
	 * SCHED_IDLE tasks get minimal weight:
	 */
743
	if (idle_policy(p->policy)) {
744
		load->weight = scale_load(WEIGHT_IDLEPRIO);
N
Nikhil Rao 已提交
745
		load->inv_weight = WMULT_IDLEPRIO;
I
Ingo Molnar 已提交
746 747
		return;
	}
748

749 750
	load->weight = scale_load(sched_prio_to_weight[prio]);
	load->inv_weight = sched_prio_to_wmult[prio];
751 752
}

753
static inline void enqueue_task(struct rq *rq, struct task_struct *p, int flags)
754
{
755 756 757
	if (!(flags & ENQUEUE_NOCLOCK))
		update_rq_clock(rq);

758 759
	if (!(flags & ENQUEUE_RESTORE))
		sched_info_queued(rq, p);
760

761
	p->sched_class->enqueue_task(rq, p, flags);
762 763
}

764
static inline void dequeue_task(struct rq *rq, struct task_struct *p, int flags)
765
{
766 767 768
	if (!(flags & DEQUEUE_NOCLOCK))
		update_rq_clock(rq);

769 770
	if (!(flags & DEQUEUE_SAVE))
		sched_info_dequeued(rq, p);
771

772
	p->sched_class->dequeue_task(rq, p, flags);
773 774
}

775
void activate_task(struct rq *rq, struct task_struct *p, int flags)
776 777 778 779
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible--;

780
	enqueue_task(rq, p, flags);
781 782
}

783
void deactivate_task(struct rq *rq, struct task_struct *p, int flags)
784 785 786 787
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible++;

788
	dequeue_task(rq, p, flags);
789 790
}

791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820
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;
	}
}

821
/*
I
Ingo Molnar 已提交
822
 * __normal_prio - return the priority that is based on the static prio
823 824 825
 */
static inline int __normal_prio(struct task_struct *p)
{
I
Ingo Molnar 已提交
826
	return p->static_prio;
827 828
}

829 830 831 832 833 834 835
/*
 * 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.
 */
836
static inline int normal_prio(struct task_struct *p)
837 838 839
{
	int prio;

840 841 842
	if (task_has_dl_policy(p))
		prio = MAX_DL_PRIO-1;
	else if (task_has_rt_policy(p))
843 844 845 846 847 848 849 850 851 852 853 854 855
		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.
 */
856
static int effective_prio(struct task_struct *p)
857 858 859 860 861 862 863 864 865 866 867 868
{
	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
869 870 871
/**
 * task_curr - is this task currently executing on a CPU?
 * @p: the task in question.
872 873
 *
 * Return: 1 if the task is currently executing. 0 otherwise.
L
Linus Torvalds 已提交
874
 */
875
inline int task_curr(const struct task_struct *p)
L
Linus Torvalds 已提交
876 877 878 879
{
	return cpu_curr(task_cpu(p)) == p;
}

880
/*
881 882 883 884 885
 * 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().
886
 */
887 888
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
889
				       int oldprio)
890 891 892
{
	if (prev_class != p->sched_class) {
		if (prev_class->switched_from)
P
Peter Zijlstra 已提交
893
			prev_class->switched_from(rq, p);
894

P
Peter Zijlstra 已提交
895
		p->sched_class->switched_to(rq, p);
896
	} else if (oldprio != p->prio || dl_task(p))
P
Peter Zijlstra 已提交
897
		p->sched_class->prio_changed(rq, p, oldprio);
898 899
}

900
void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags)
901 902 903 904 905 906 907 908 909 910
{
	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) {
911
				resched_curr(rq);
912 913 914 915 916 917 918 919 920
				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.
	 */
921
	if (task_on_rq_queued(rq->curr) && test_tsk_need_resched(rq->curr))
922
		rq_clock_skip_update(rq, true);
923 924
}

L
Linus Torvalds 已提交
925
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944
/*
 * 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.
 */
945 946
static struct rq *move_queued_task(struct rq *rq, struct rq_flags *rf,
				   struct task_struct *p, int new_cpu)
P
Peter Zijlstra 已提交
947 948 949 950
{
	lockdep_assert_held(&rq->lock);

	p->on_rq = TASK_ON_RQ_MIGRATING;
951
	dequeue_task(rq, p, DEQUEUE_NOCLOCK);
P
Peter Zijlstra 已提交
952
	set_task_cpu(p, new_cpu);
953
	rq_unlock(rq, rf);
P
Peter Zijlstra 已提交
954 955 956

	rq = cpu_rq(new_cpu);

957
	rq_lock(rq, rf);
P
Peter Zijlstra 已提交
958 959
	BUG_ON(task_cpu(p) != new_cpu);
	enqueue_task(rq, p, 0);
960
	p->on_rq = TASK_ON_RQ_QUEUED;
P
Peter Zijlstra 已提交
961 962 963 964 965 966 967 968 969 970 971
	check_preempt_curr(rq, p, 0);

	return rq;
}

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

/*
I
Ingo Molnar 已提交
972
 * Move (not current) task off this CPU, onto the destination CPU. We're doing
P
Peter Zijlstra 已提交
973 974 975 976 977 978 979
 * 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.
 */
980 981
static struct rq *__migrate_task(struct rq *rq, struct rq_flags *rf,
				 struct task_struct *p, int dest_cpu)
P
Peter Zijlstra 已提交
982 983
{
	if (unlikely(!cpu_active(dest_cpu)))
984
		return rq;
P
Peter Zijlstra 已提交
985 986

	/* Affinity changed (again). */
987
	if (!cpumask_test_cpu(dest_cpu, &p->cpus_allowed))
988
		return rq;
P
Peter Zijlstra 已提交
989

990
	update_rq_clock(rq);
991
	rq = move_queued_task(rq, rf, p, dest_cpu);
992 993

	return rq;
P
Peter Zijlstra 已提交
994 995 996 997 998 999 1000 1001 1002 1003
}

/*
 * 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;
1004 1005
	struct task_struct *p = arg->task;
	struct rq *rq = this_rq();
1006
	struct rq_flags rf;
P
Peter Zijlstra 已提交
1007 1008

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

	raw_spin_lock(&p->pi_lock);
1021
	rq_lock(rq, &rf);
1022 1023 1024 1025 1026
	/*
	 * If task_rq(p) != rq, it cannot be migrated here, because we're
	 * holding rq->lock, if p->on_rq == 0 it cannot get enqueued because
	 * we're holding p->pi_lock.
	 */
1027 1028
	if (task_rq(p) == rq) {
		if (task_on_rq_queued(p))
1029
			rq = __migrate_task(rq, &rf, p, arg->dest_cpu);
1030 1031 1032
		else
			p->wake_cpu = arg->dest_cpu;
	}
1033
	rq_unlock(rq, &rf);
1034 1035
	raw_spin_unlock(&p->pi_lock);

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

1040 1041 1042 1043 1044
/*
 * sched_class::set_cpus_allowed must do the below, but is not required to
 * actually call this function.
 */
void set_cpus_allowed_common(struct task_struct *p, const struct cpumask *new_mask)
P
Peter Zijlstra 已提交
1045 1046 1047 1048 1049
{
	cpumask_copy(&p->cpus_allowed, new_mask);
	p->nr_cpus_allowed = cpumask_weight(new_mask);
}

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

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

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

	if (queued) {
		/*
		 * Because __kthread_bind() calls this on blocked tasks without
		 * holding rq->lock.
		 */
		lockdep_assert_held(&rq->lock);
1066
		dequeue_task(rq, p, DEQUEUE_SAVE | DEQUEUE_NOCLOCK);
1067 1068 1069 1070
	}
	if (running)
		put_prev_task(rq, p);

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

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

P
Peter Zijlstra 已提交
1079 1080 1081 1082 1083 1084 1085 1086 1087
/*
 * Change a given task's CPU affinity. Migrate the thread to a
 * proper CPU and schedule it away if the CPU it's executing on
 * is removed from the allowed bitmask.
 *
 * NOTE: the caller must have a valid reference to the task, the
 * task must not exit() & deallocate itself prematurely. The
 * call is not atomic; no spinlocks may be held.
 */
1088 1089
static int __set_cpus_allowed_ptr(struct task_struct *p,
				  const struct cpumask *new_mask, bool check)
P
Peter Zijlstra 已提交
1090
{
1091
	const struct cpumask *cpu_valid_mask = cpu_active_mask;
P
Peter Zijlstra 已提交
1092
	unsigned int dest_cpu;
1093 1094
	struct rq_flags rf;
	struct rq *rq;
P
Peter Zijlstra 已提交
1095 1096
	int ret = 0;

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

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

1107 1108 1109 1110 1111 1112 1113 1114 1115
	/*
	 * Must re-check here, to close a race against __kthread_bind(),
	 * sched_setaffinity() is not guaranteed to observe the flag.
	 */
	if (check && (p->flags & PF_NO_SETAFFINITY)) {
		ret = -EINVAL;
		goto out;
	}

P
Peter Zijlstra 已提交
1116 1117 1118
	if (cpumask_equal(&p->cpus_allowed, new_mask))
		goto out;

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

	do_set_cpus_allowed(p, new_mask);

1126 1127 1128
	if (p->flags & PF_KTHREAD) {
		/*
		 * For kernel threads that do indeed end up on online &&
I
Ingo Molnar 已提交
1129
		 * !active we want to ensure they are strict per-CPU threads.
1130 1131 1132 1133 1134 1135
		 */
		WARN_ON(cpumask_intersects(new_mask, cpu_online_mask) &&
			!cpumask_intersects(new_mask, cpu_active_mask) &&
			p->nr_cpus_allowed != 1);
	}

P
Peter Zijlstra 已提交
1136 1137 1138 1139
	/* Can the task run on the task's current CPU? If so, we're done */
	if (cpumask_test_cpu(task_cpu(p), new_mask))
		goto out;

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

	return ret;
}
1160 1161 1162 1163 1164

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
1165 1166
EXPORT_SYMBOL_GPL(set_cpus_allowed_ptr);

I
Ingo Molnar 已提交
1167
void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
I
Ingo Molnar 已提交
1168
{
1169 1170 1171 1172 1173
#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
1174
	WARN_ON_ONCE(p->state != TASK_RUNNING && p->state != TASK_WAKING &&
O
Oleg Nesterov 已提交
1175
			!p->on_rq);
1176

1177 1178 1179 1180 1181 1182 1183 1184 1185
	/*
	 * 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)));

1186
#ifdef CONFIG_LOCKDEP
1187 1188 1189 1190 1191
	/*
	 * 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
1192
	 * see task_group().
1193 1194 1195 1196
	 *
	 * Furthermore, all task_rq users should acquire both locks, see
	 * task_rq_lock().
	 */
1197 1198 1199
	WARN_ON_ONCE(debug_locks && !(lockdep_is_held(&p->pi_lock) ||
				      lockdep_is_held(&task_rq(p)->lock)));
#endif
1200 1201
#endif

1202
	trace_sched_migrate_task(p, new_cpu);
1203

1204
	if (task_cpu(p) != new_cpu) {
1205
		if (p->sched_class->migrate_task_rq)
1206
			p->sched_class->migrate_task_rq(p);
1207
		p->se.nr_migrations++;
1208
		perf_event_task_migrate(p);
1209
	}
I
Ingo Molnar 已提交
1210 1211

	__set_task_cpu(p, new_cpu);
I
Ingo Molnar 已提交
1212 1213
}

1214 1215
static void __migrate_swap_task(struct task_struct *p, int cpu)
{
1216
	if (task_on_rq_queued(p)) {
1217
		struct rq *src_rq, *dst_rq;
1218
		struct rq_flags srf, drf;
1219 1220 1221 1222

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

1223 1224 1225
		rq_pin_lock(src_rq, &srf);
		rq_pin_lock(dst_rq, &drf);

1226
		p->on_rq = TASK_ON_RQ_MIGRATING;
1227 1228 1229
		deactivate_task(src_rq, p, 0);
		set_task_cpu(p, cpu);
		activate_task(dst_rq, p, 0);
1230
		p->on_rq = TASK_ON_RQ_QUEUED;
1231
		check_preempt_curr(dst_rq, p, 0);
1232 1233 1234 1235

		rq_unpin_lock(dst_rq, &drf);
		rq_unpin_lock(src_rq, &srf);

1236 1237 1238 1239
	} 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
1240
		 * previous CPU our target instead of where it really is.
1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256
		 */
		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;

1257 1258 1259
	if (!cpu_active(arg->src_cpu) || !cpu_active(arg->dst_cpu))
		return -EAGAIN;

1260 1261 1262
	src_rq = cpu_rq(arg->src_cpu);
	dst_rq = cpu_rq(arg->dst_cpu);

1263 1264
	double_raw_lock(&arg->src_task->pi_lock,
			&arg->dst_task->pi_lock);
1265
	double_rq_lock(src_rq, dst_rq);
1266

1267 1268 1269 1270 1271 1272
	if (task_cpu(arg->dst_task) != arg->dst_cpu)
		goto unlock;

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

1273
	if (!cpumask_test_cpu(arg->dst_cpu, &arg->src_task->cpus_allowed))
1274 1275
		goto unlock;

1276
	if (!cpumask_test_cpu(arg->src_cpu, &arg->dst_task->cpus_allowed))
1277 1278 1279 1280 1281 1282 1283 1284 1285
		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);
1286 1287
	raw_spin_unlock(&arg->dst_task->pi_lock);
	raw_spin_unlock(&arg->src_task->pi_lock);
1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309

	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;

1310 1311 1312 1313
	/*
	 * 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.
	 */
1314 1315 1316
	if (!cpu_active(arg.src_cpu) || !cpu_active(arg.dst_cpu))
		goto out;

1317
	if (!cpumask_test_cpu(arg.dst_cpu, &arg.src_task->cpus_allowed))
1318 1319
		goto out;

1320
	if (!cpumask_test_cpu(arg.src_cpu, &arg.dst_task->cpus_allowed))
1321 1322
		goto out;

1323
	trace_sched_swap_numa(cur, arg.src_cpu, p, arg.dst_cpu);
1324 1325 1326 1327 1328 1329
	ret = stop_two_cpus(arg.dst_cpu, arg.src_cpu, migrate_swap_stop, &arg);

out:
	return ret;
}

L
Linus Torvalds 已提交
1330 1331 1332
/*
 * wait_task_inactive - wait for a thread to unschedule.
 *
R
Roland McGrath 已提交
1333 1334 1335 1336 1337 1338 1339
 * 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
1340 1341 1342 1343 1344 1345
 * 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
1346
unsigned long wait_task_inactive(struct task_struct *p, long match_state)
L
Linus Torvalds 已提交
1347
{
1348
	int running, queued;
1349
	struct rq_flags rf;
R
Roland McGrath 已提交
1350
	unsigned long ncsw;
1351
	struct rq *rq;
L
Linus Torvalds 已提交
1352

1353 1354 1355 1356 1357 1358 1359 1360
	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);
1361

1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372
		/*
		 * 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
1373 1374 1375
		while (task_running(rq, p)) {
			if (match_state && unlikely(p->state != match_state))
				return 0;
1376
			cpu_relax();
R
Roland McGrath 已提交
1377
		}
1378

1379 1380 1381 1382 1383
		/*
		 * 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.
		 */
1384
		rq = task_rq_lock(p, &rf);
1385
		trace_sched_wait_task(p);
1386
		running = task_running(rq, p);
1387
		queued = task_on_rq_queued(p);
R
Roland McGrath 已提交
1388
		ncsw = 0;
1389
		if (!match_state || p->state == match_state)
1390
			ncsw = p->nvcsw | LONG_MIN; /* sets MSB */
1391
		task_rq_unlock(rq, p, &rf);
1392

R
Roland McGrath 已提交
1393 1394 1395 1396 1397 1398
		/*
		 * If it changed from the expected state, bail out now.
		 */
		if (unlikely(!ncsw))
			break;

1399 1400 1401 1402 1403 1404 1405 1406 1407 1408
		/*
		 * 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;
		}
1409

1410 1411 1412 1413 1414
		/*
		 * It's not enough that it's not actively running,
		 * it must be off the runqueue _entirely_, and not
		 * preempted!
		 *
1415
		 * So if it was still runnable (but just not actively
1416 1417 1418
		 * running right now), it's preempted, and we should
		 * yield - it could be a while.
		 */
1419
		if (unlikely(queued)) {
T
Thomas Gleixner 已提交
1420
			ktime_t to = NSEC_PER_SEC / HZ;
1421 1422 1423

			set_current_state(TASK_UNINTERRUPTIBLE);
			schedule_hrtimeout(&to, HRTIMER_MODE_REL);
1424 1425
			continue;
		}
1426

1427 1428 1429 1430 1431 1432 1433
		/*
		 * 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
1434 1435

	return ncsw;
L
Linus Torvalds 已提交
1436 1437 1438 1439 1440 1441 1442 1443 1444
}

/***
 * 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
1445
 * NOTE: this function doesn't have to take the runqueue lock,
L
Linus Torvalds 已提交
1446 1447 1448 1449 1450
 * 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.
 */
1451
void kick_process(struct task_struct *p)
L
Linus Torvalds 已提交
1452 1453 1454 1455 1456 1457 1458 1459 1460
{
	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
1461
EXPORT_SYMBOL_GPL(kick_process);
L
Linus Torvalds 已提交
1462

1463
/*
1464
 * ->cpus_allowed is protected by both rq->lock and p->pi_lock
1465 1466 1467 1468 1469 1470 1471
 *
 * 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
1472
 *    CPU isn't yet part of the sched domains, and balancing will not
1473 1474
 *    see it.
 *
I
Ingo Molnar 已提交
1475
 *  - on CPU-down we clear cpu_active() to mask the sched domains and
1476
 *    avoid the load balancer to place new tasks on the to be removed
I
Ingo Molnar 已提交
1477
 *    CPU. Existing tasks will remain running there and will be taken
1478 1479 1480 1481 1482 1483
 *    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.
1484
 */
1485 1486
static int select_fallback_rq(int cpu, struct task_struct *p)
{
1487 1488
	int nid = cpu_to_node(cpu);
	const struct cpumask *nodemask = NULL;
1489 1490
	enum { cpuset, possible, fail } state = cpuset;
	int dest_cpu;
1491

1492
	/*
I
Ingo Molnar 已提交
1493 1494 1495
	 * 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.
1496 1497 1498 1499 1500 1501 1502 1503
	 */
	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;
1504
			if (cpumask_test_cpu(dest_cpu, &p->cpus_allowed))
1505 1506
				return dest_cpu;
		}
1507
	}
1508

1509 1510
	for (;;) {
		/* Any allowed, online CPU? */
1511
		for_each_cpu(dest_cpu, &p->cpus_allowed) {
1512 1513 1514
			if (!(p->flags & PF_KTHREAD) && !cpu_active(dest_cpu))
				continue;
			if (!cpu_online(dest_cpu))
1515 1516 1517
				continue;
			goto out;
		}
1518

1519
		/* No more Mr. Nice Guy. */
1520 1521
		switch (state) {
		case cpuset:
1522 1523 1524 1525 1526
			if (IS_ENABLED(CONFIG_CPUSETS)) {
				cpuset_cpus_allowed_fallback(p);
				state = possible;
				break;
			}
I
Ingo Molnar 已提交
1527
			/* Fall-through */
1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546
		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()) {
1547
			printk_deferred("process %d (%s) no longer affine to cpu%d\n",
1548 1549
					task_pid_nr(p), p->comm, cpu);
		}
1550 1551 1552 1553 1554
	}

	return dest_cpu;
}

1555
/*
1556
 * The caller (fork, wakeup) owns p->pi_lock, ->cpus_allowed is stable.
1557
 */
1558
static inline
1559
int select_task_rq(struct task_struct *p, int cpu, int sd_flags, int wake_flags)
1560
{
1561 1562
	lockdep_assert_held(&p->pi_lock);

1563
	if (p->nr_cpus_allowed > 1)
1564
		cpu = p->sched_class->select_task_rq(p, cpu, sd_flags, wake_flags);
1565
	else
1566
		cpu = cpumask_any(&p->cpus_allowed);
1567 1568 1569 1570

	/*
	 * 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
1571
	 * CPU.
1572 1573 1574 1575 1576 1577
	 *
	 * 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 ]
	 */
1578
	if (unlikely(!cpumask_test_cpu(cpu, &p->cpus_allowed) ||
P
Peter Zijlstra 已提交
1579
		     !cpu_online(cpu)))
1580
		cpu = select_fallback_rq(task_cpu(p), p);
1581 1582

	return cpu;
1583
}
1584 1585 1586 1587 1588 1589

static void update_avg(u64 *avg, u64 sample)
{
	s64 diff = sample - *avg;
	*avg += diff >> 3;
}
1590 1591 1592 1593 1594 1595 1596 1597 1598

#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
1599
#endif /* CONFIG_SMP */
1600

P
Peter Zijlstra 已提交
1601
static void
1602
ttwu_stat(struct task_struct *p, int cpu, int wake_flags)
T
Tejun Heo 已提交
1603
{
1604
	struct rq *rq;
1605

1606 1607 1608 1609
	if (!schedstat_enabled())
		return;

	rq = this_rq();
P
Peter Zijlstra 已提交
1610

1611 1612
#ifdef CONFIG_SMP
	if (cpu == rq->cpu) {
1613 1614
		schedstat_inc(rq->ttwu_local);
		schedstat_inc(p->se.statistics.nr_wakeups_local);
P
Peter Zijlstra 已提交
1615 1616 1617
	} else {
		struct sched_domain *sd;

1618
		schedstat_inc(p->se.statistics.nr_wakeups_remote);
1619
		rcu_read_lock();
1620
		for_each_domain(rq->cpu, sd) {
P
Peter Zijlstra 已提交
1621
			if (cpumask_test_cpu(cpu, sched_domain_span(sd))) {
1622
				schedstat_inc(sd->ttwu_wake_remote);
P
Peter Zijlstra 已提交
1623 1624 1625
				break;
			}
		}
1626
		rcu_read_unlock();
P
Peter Zijlstra 已提交
1627
	}
1628 1629

	if (wake_flags & WF_MIGRATED)
1630
		schedstat_inc(p->se.statistics.nr_wakeups_migrate);
P
Peter Zijlstra 已提交
1631 1632
#endif /* CONFIG_SMP */

1633 1634
	schedstat_inc(rq->ttwu_count);
	schedstat_inc(p->se.statistics.nr_wakeups);
P
Peter Zijlstra 已提交
1635 1636

	if (wake_flags & WF_SYNC)
1637
		schedstat_inc(p->se.statistics.nr_wakeups_sync);
P
Peter Zijlstra 已提交
1638 1639
}

1640
static inline void ttwu_activate(struct rq *rq, struct task_struct *p, int en_flags)
P
Peter Zijlstra 已提交
1641
{
T
Tejun Heo 已提交
1642
	activate_task(rq, p, en_flags);
1643
	p->on_rq = TASK_ON_RQ_QUEUED;
1644

I
Ingo Molnar 已提交
1645
	/* If a worker is waking up, notify the workqueue: */
1646 1647
	if (p->flags & PF_WQ_WORKER)
		wq_worker_waking_up(p, cpu_of(rq));
T
Tejun Heo 已提交
1648 1649
}

1650 1651 1652
/*
 * Mark the task runnable and perform wakeup-preemption.
 */
1653
static void ttwu_do_wakeup(struct rq *rq, struct task_struct *p, int wake_flags,
1654
			   struct rq_flags *rf)
T
Tejun Heo 已提交
1655 1656 1657
{
	check_preempt_curr(rq, p, wake_flags);
	p->state = TASK_RUNNING;
1658 1659
	trace_sched_wakeup(p);

T
Tejun Heo 已提交
1660
#ifdef CONFIG_SMP
1661 1662
	if (p->sched_class->task_woken) {
		/*
1663 1664
		 * Our task @p is fully woken up and running; so its safe to
		 * drop the rq->lock, hereafter rq is only used for statistics.
1665
		 */
1666
		rq_unpin_lock(rq, rf);
T
Tejun Heo 已提交
1667
		p->sched_class->task_woken(rq, p);
1668
		rq_repin_lock(rq, rf);
1669
	}
T
Tejun Heo 已提交
1670

1671
	if (rq->idle_stamp) {
1672
		u64 delta = rq_clock(rq) - rq->idle_stamp;
1673
		u64 max = 2*rq->max_idle_balance_cost;
T
Tejun Heo 已提交
1674

1675 1676 1677
		update_avg(&rq->avg_idle, delta);

		if (rq->avg_idle > max)
T
Tejun Heo 已提交
1678
			rq->avg_idle = max;
1679

T
Tejun Heo 已提交
1680 1681 1682 1683 1684
		rq->idle_stamp = 0;
	}
#endif
}

1685
static void
1686
ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags,
1687
		 struct rq_flags *rf)
1688
{
1689
	int en_flags = ENQUEUE_WAKEUP | ENQUEUE_NOCLOCK;
1690

1691 1692
	lockdep_assert_held(&rq->lock);

1693 1694 1695
#ifdef CONFIG_SMP
	if (p->sched_contributes_to_load)
		rq->nr_uninterruptible--;
1696 1697

	if (wake_flags & WF_MIGRATED)
1698
		en_flags |= ENQUEUE_MIGRATED;
1699 1700
#endif

1701
	ttwu_activate(rq, p, en_flags);
1702
	ttwu_do_wakeup(rq, p, wake_flags, rf);
1703 1704 1705 1706 1707 1708 1709 1710 1711 1712
}

/*
 * 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)
{
1713
	struct rq_flags rf;
1714 1715 1716
	struct rq *rq;
	int ret = 0;

1717
	rq = __task_rq_lock(p, &rf);
1718
	if (task_on_rq_queued(p)) {
1719 1720
		/* check_preempt_curr() may use rq clock */
		update_rq_clock(rq);
1721
		ttwu_do_wakeup(rq, p, wake_flags, &rf);
1722 1723
		ret = 1;
	}
1724
	__task_rq_unlock(rq, &rf);
1725 1726 1727 1728

	return ret;
}

1729
#ifdef CONFIG_SMP
1730
void sched_ttwu_pending(void)
1731 1732
{
	struct rq *rq = this_rq();
P
Peter Zijlstra 已提交
1733
	struct llist_node *llist = llist_del_all(&rq->wake_list);
1734
	struct task_struct *p, *t;
1735
	struct rq_flags rf;
1736

1737 1738 1739
	if (!llist)
		return;

1740
	rq_lock_irqsave(rq, &rf);
1741
	update_rq_clock(rq);
1742

1743 1744
	llist_for_each_entry_safe(p, t, llist, wake_entry)
		ttwu_do_activate(rq, p, p->sched_remote_wakeup ? WF_MIGRATED : 0, &rf);
1745

1746
	rq_unlock_irqrestore(rq, &rf);
1747 1748 1749 1750
}

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

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

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

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

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

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

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

1801 1802 1803
void wake_up_if_idle(int cpu)
{
	struct rq *rq = cpu_rq(cpu);
1804
	struct rq_flags rf;
1805

1806 1807 1808 1809
	rcu_read_lock();

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

	if (set_nr_if_polling(rq->idle)) {
		trace_sched_wake_idle_without_ipi(cpu);
	} else {
1814
		rq_lock_irqsave(rq, &rf);
1815 1816
		if (is_idle_task(rq->curr))
			smp_send_reschedule(cpu);
I
Ingo Molnar 已提交
1817
		/* Else CPU is not idle, do nothing here: */
1818
		rq_unlock_irqrestore(rq, &rf);
1819
	}
1820 1821 1822

out:
	rcu_read_unlock();
1823 1824
}

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

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

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

1844
	rq_lock(rq, &rf);
1845
	update_rq_clock(rq);
1846
	ttwu_do_activate(rq, p, wake_flags, &rf);
1847
	rq_unlock(rq, &rf);
T
Tejun Heo 已提交
1848 1849
}

1850 1851 1852 1853 1854 1855
/*
 * 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
1856 1857
 * migrates, all its activity on its old CPU [c0] happens-before any subsequent
 * execution on its new CPU [c1].
1858 1859 1860 1861 1862 1863 1864 1865 1866 1867
 *
 * 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
1868
 * Note: the CPU doing B need not be c0 or c1
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 1897 1898 1899
 *
 * 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)
1900
 *   2) smp_cond_load_acquire(!X->on_cpu)
1901 1902 1903 1904 1905 1906 1907 1908 1909 1910
 *
 * 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);
 *
1911
 *                    smp_cond_load_acquire(&X->on_cpu, !VAL);
1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936
 *                    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,
1937
 * since the waking CPU is the one issueing the ACQUIRE (smp_cond_load_acquire).
1938 1939 1940
 *
 */

T
Tejun Heo 已提交
1941
/**
L
Linus Torvalds 已提交
1942
 * try_to_wake_up - wake up a thread
T
Tejun Heo 已提交
1943
 * @p: the thread to be awakened
L
Linus Torvalds 已提交
1944
 * @state: the mask of task states that can be woken
T
Tejun Heo 已提交
1945
 * @wake_flags: wake modifier flags (WF_*)
L
Linus Torvalds 已提交
1946
 *
1947
 * If (@state & @p->state) @p->state = TASK_RUNNING.
L
Linus Torvalds 已提交
1948
 *
1949 1950 1951 1952 1953 1954 1955
 * 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
1956
 */
1957 1958
static int
try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
L
Linus Torvalds 已提交
1959 1960
{
	unsigned long flags;
1961
	int cpu, success = 0;
P
Peter Zijlstra 已提交
1962

1963 1964 1965 1966 1967 1968 1969
	/*
	 * 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();
1970
	raw_spin_lock_irqsave(&p->pi_lock, flags);
P
Peter Zijlstra 已提交
1971
	if (!(p->state & state))
L
Linus Torvalds 已提交
1972 1973
		goto out;

1974 1975
	trace_sched_waking(p);

I
Ingo Molnar 已提交
1976 1977
	/* We're going to change ->state: */
	success = 1;
L
Linus Torvalds 已提交
1978 1979
	cpu = task_cpu(p);

1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001
	/*
	 * 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();
2002 2003
	if (p->on_rq && ttwu_remote(p, wake_flags))
		goto stat;
L
Linus Torvalds 已提交
2004 2005

#ifdef CONFIG_SMP
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024
	/*
	 * 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
2025
	/*
I
Ingo Molnar 已提交
2026
	 * If the owning (remote) CPU is still in the middle of schedule() with
2027
	 * this task as prev, wait until its done referencing the task.
2028 2029 2030 2031 2032
	 *
	 * 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.
2033
	 */
2034
	smp_cond_load_acquire(&p->on_cpu, !VAL);
L
Linus Torvalds 已提交
2035

2036
	p->sched_contributes_to_load = !!task_contributes_to_load(p);
P
Peter Zijlstra 已提交
2037
	p->state = TASK_WAKING;
2038

2039 2040 2041 2042 2043
	if (p->in_iowait) {
		delayacct_blkio_end();
		atomic_dec(&task_rq(p)->nr_iowait);
	}

2044
	cpu = select_task_rq(p, p->wake_cpu, SD_BALANCE_WAKE, wake_flags);
2045 2046
	if (task_cpu(p) != cpu) {
		wake_flags |= WF_MIGRATED;
2047
		set_task_cpu(p, cpu);
2048
	}
2049 2050 2051 2052 2053 2054 2055 2056

#else /* CONFIG_SMP */

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

L
Linus Torvalds 已提交
2057 2058
#endif /* CONFIG_SMP */

2059
	ttwu_queue(p, cpu, wake_flags);
2060
stat:
2061
	ttwu_stat(p, cpu, wake_flags);
L
Linus Torvalds 已提交
2062
out:
2063
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
2064 2065 2066 2067

	return success;
}

T
Tejun Heo 已提交
2068 2069 2070
/**
 * try_to_wake_up_local - try to wake up a local task with rq lock held
 * @p: the thread to be awakened
2071
 * @cookie: context's cookie for pinning
T
Tejun Heo 已提交
2072
 *
2073
 * Put @p on the run-queue if it's not already there. The caller must
T
Tejun Heo 已提交
2074
 * ensure that this_rq() is locked, @p is bound to this_rq() and not
2075
 * the current task.
T
Tejun Heo 已提交
2076
 */
2077
static void try_to_wake_up_local(struct task_struct *p, struct rq_flags *rf)
T
Tejun Heo 已提交
2078 2079 2080
{
	struct rq *rq = task_rq(p);

2081 2082 2083 2084
	if (WARN_ON_ONCE(rq != this_rq()) ||
	    WARN_ON_ONCE(p == current))
		return;

T
Tejun Heo 已提交
2085 2086
	lockdep_assert_held(&rq->lock);

2087
	if (!raw_spin_trylock(&p->pi_lock)) {
2088 2089 2090 2091 2092 2093
		/*
		 * 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.
		 */
2094
		rq_unlock(rq, rf);
2095
		raw_spin_lock(&p->pi_lock);
2096
		rq_relock(rq, rf);
2097 2098
	}

T
Tejun Heo 已提交
2099
	if (!(p->state & TASK_NORMAL))
2100
		goto out;
T
Tejun Heo 已提交
2101

2102 2103
	trace_sched_waking(p);

2104 2105 2106 2107 2108
	if (!task_on_rq_queued(p)) {
		if (p->in_iowait) {
			delayacct_blkio_end();
			atomic_dec(&rq->nr_iowait);
		}
2109
		ttwu_activate(rq, p, ENQUEUE_WAKEUP | ENQUEUE_NOCLOCK);
2110
	}
P
Peter Zijlstra 已提交
2111

2112
	ttwu_do_wakeup(rq, p, 0, rf);
2113
	ttwu_stat(p, smp_processor_id(), 0);
2114 2115
out:
	raw_spin_unlock(&p->pi_lock);
T
Tejun Heo 已提交
2116 2117
}

2118 2119 2120 2121 2122
/**
 * 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
2123 2124 2125
 * processes.
 *
 * Return: 1 if the process was woken up, 0 if it was already running.
2126 2127 2128 2129
 *
 * 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.
 */
2130
int wake_up_process(struct task_struct *p)
L
Linus Torvalds 已提交
2131
{
2132
	return try_to_wake_up(p, TASK_NORMAL, 0);
L
Linus Torvalds 已提交
2133 2134 2135
}
EXPORT_SYMBOL(wake_up_process);

2136
int wake_up_state(struct task_struct *p, unsigned int state)
L
Linus Torvalds 已提交
2137 2138 2139 2140
{
	return try_to_wake_up(p, state, 0);
}

2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152
/*
 * 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;
2153 2154 2155

	dl_se->dl_throttled = 0;
	dl_se->dl_yielded = 0;
2156
	dl_se->dl_non_contending = 0;
2157 2158
}

L
Linus Torvalds 已提交
2159 2160 2161
/*
 * Perform scheduler related setup for a newly forked process p.
 * p is forked by current.
I
Ingo Molnar 已提交
2162 2163 2164
 *
 * __sched_fork() is basic setup used by init_idle() too:
 */
2165
static void __sched_fork(unsigned long clone_flags, struct task_struct *p)
I
Ingo Molnar 已提交
2166
{
P
Peter Zijlstra 已提交
2167 2168 2169
	p->on_rq			= 0;

	p->se.on_rq			= 0;
I
Ingo Molnar 已提交
2170 2171
	p->se.exec_start		= 0;
	p->se.sum_exec_runtime		= 0;
2172
	p->se.prev_sum_exec_runtime	= 0;
2173
	p->se.nr_migrations		= 0;
P
Peter Zijlstra 已提交
2174
	p->se.vruntime			= 0;
P
Peter Zijlstra 已提交
2175
	INIT_LIST_HEAD(&p->se.group_node);
I
Ingo Molnar 已提交
2176

2177 2178 2179 2180
#ifdef CONFIG_FAIR_GROUP_SCHED
	p->se.cfs_rq			= NULL;
#endif

I
Ingo Molnar 已提交
2181
#ifdef CONFIG_SCHEDSTATS
2182
	/* Even if schedstat is disabled, there should not be garbage */
2183
	memset(&p->se.statistics, 0, sizeof(p->se.statistics));
I
Ingo Molnar 已提交
2184
#endif
N
Nick Piggin 已提交
2185

2186
	RB_CLEAR_NODE(&p->dl.rb_node);
2187
	init_dl_task_timer(&p->dl);
2188
	init_dl_inactive_task_timer(&p->dl);
2189
	__dl_clear_params(p);
2190

P
Peter Zijlstra 已提交
2191
	INIT_LIST_HEAD(&p->rt.run_list);
2192 2193 2194 2195
	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
2196

2197 2198 2199
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&p->preempt_notifiers);
#endif
2200 2201 2202

#ifdef CONFIG_NUMA_BALANCING
	if (p->mm && atomic_read(&p->mm->mm_users) == 1) {
2203
		p->mm->numa_next_scan = jiffies + msecs_to_jiffies(sysctl_numa_balancing_scan_delay);
2204 2205 2206
		p->mm->numa_scan_seq = 0;
	}

2207 2208 2209 2210 2211
	if (clone_flags & CLONE_VM)
		p->numa_preferred_nid = current->numa_preferred_nid;
	else
		p->numa_preferred_nid = -1;

2212 2213
	p->node_stamp = 0ULL;
	p->numa_scan_seq = p->mm ? p->mm->numa_scan_seq : 0;
2214
	p->numa_scan_period = sysctl_numa_balancing_scan_delay;
2215
	p->numa_work.next = &p->numa_work;
2216
	p->numa_faults = NULL;
2217 2218
	p->last_task_numa_placement = 0;
	p->last_sum_exec_runtime = 0;
2219 2220

	p->numa_group = NULL;
2221
#endif /* CONFIG_NUMA_BALANCING */
I
Ingo Molnar 已提交
2222 2223
}

2224 2225
DEFINE_STATIC_KEY_FALSE(sched_numa_balancing);

2226
#ifdef CONFIG_NUMA_BALANCING
2227

2228 2229 2230
void set_numabalancing_state(bool enabled)
{
	if (enabled)
2231
		static_branch_enable(&sched_numa_balancing);
2232
	else
2233
		static_branch_disable(&sched_numa_balancing);
2234
}
2235 2236 2237 2238 2239 2240 2241

#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;
2242
	int state = static_branch_likely(&sched_numa_balancing);
2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257

	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
2258

2259 2260
#ifdef CONFIG_SCHEDSTATS

2261
DEFINE_STATIC_KEY_FALSE(sched_schedstats);
2262
static bool __initdata __sched_schedstats = false;
2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285

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;

2286 2287 2288 2289 2290
	/*
	 * 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.
	 */
2291
	if (!strcmp(str, "enable")) {
2292
		__sched_schedstats = true;
2293 2294
		ret = 1;
	} else if (!strcmp(str, "disable")) {
2295
		__sched_schedstats = false;
2296 2297 2298 2299 2300 2301 2302 2303 2304 2305
		ret = 1;
	}
out:
	if (!ret)
		pr_warn("Unable to parse schedstats=\n");

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

2306 2307 2308 2309 2310
static void __init init_schedstats(void)
{
	set_schedstats(__sched_schedstats);
}

2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330
#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;
}
2331 2332 2333 2334
#endif /* CONFIG_PROC_SYSCTL */
#else  /* !CONFIG_SCHEDSTATS */
static inline void init_schedstats(void) {}
#endif /* CONFIG_SCHEDSTATS */
I
Ingo Molnar 已提交
2335 2336 2337 2338

/*
 * fork()/clone()-time setup:
 */
2339
int sched_fork(unsigned long clone_flags, struct task_struct *p)
I
Ingo Molnar 已提交
2340
{
2341
	unsigned long flags;
I
Ingo Molnar 已提交
2342 2343
	int cpu = get_cpu();

2344
	__sched_fork(clone_flags, p);
2345
	/*
2346
	 * We mark the process as NEW here. This guarantees that
2347 2348 2349
	 * nobody will actually run it, and a signal or other external
	 * event cannot wake it up and insert it on the runqueue either.
	 */
2350
	p->state = TASK_NEW;
I
Ingo Molnar 已提交
2351

2352 2353 2354 2355 2356
	/*
	 * Make sure we do not leak PI boosting priority to the child.
	 */
	p->prio = current->normal_prio;

2357 2358 2359 2360
	/*
	 * Revert to default priority/policy on fork if requested.
	 */
	if (unlikely(p->sched_reset_on_fork)) {
2361
		if (task_has_dl_policy(p) || task_has_rt_policy(p)) {
2362
			p->policy = SCHED_NORMAL;
2363
			p->static_prio = NICE_TO_PRIO(0);
2364 2365 2366 2367 2368 2369
			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);
2370

2371 2372 2373 2374 2375 2376
		/*
		 * We don't need the reset flag anymore after the fork. It has
		 * fulfilled its duty:
		 */
		p->sched_reset_on_fork = 0;
	}
2377

2378 2379 2380 2381 2382 2383
	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
2384
		p->sched_class = &fair_sched_class;
2385
	}
2386

2387
	init_entity_runnable_average(&p->se);
P
Peter Zijlstra 已提交
2388

2389 2390 2391 2392 2393 2394 2395
	/*
	 * 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.
	 */
2396
	raw_spin_lock_irqsave(&p->pi_lock, flags);
2397
	/*
I
Ingo Molnar 已提交
2398
	 * We're setting the CPU for the first time, we don't migrate,
2399 2400 2401 2402 2403
	 * so use __set_task_cpu().
	 */
	__set_task_cpu(p, cpu);
	if (p->sched_class->task_fork)
		p->sched_class->task_fork(p);
2404
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
2405

2406
#ifdef CONFIG_SCHED_INFO
I
Ingo Molnar 已提交
2407
	if (likely(sched_info_on()))
2408
		memset(&p->sched_info, 0, sizeof(p->sched_info));
L
Linus Torvalds 已提交
2409
#endif
P
Peter Zijlstra 已提交
2410 2411
#if defined(CONFIG_SMP)
	p->on_cpu = 0;
2412
#endif
2413
	init_task_preempt_count(p);
2414
#ifdef CONFIG_SMP
2415
	plist_node_init(&p->pushable_tasks, MAX_PRIO);
2416
	RB_CLEAR_NODE(&p->pushable_dl_tasks);
2417
#endif
2418

N
Nick Piggin 已提交
2419
	put_cpu();
2420
	return 0;
L
Linus Torvalds 已提交
2421 2422
}

2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441
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)
{
2442 2443
	RCU_LOCKDEP_WARN(!rcu_read_lock_sched_held(),
			 "sched RCU must be held");
2444 2445 2446
	return &cpu_rq(i)->rd->dl_bw;
}

2447
static inline int dl_bw_cpus(int i)
2448
{
2449 2450 2451
	struct root_domain *rd = cpu_rq(i)->rd;
	int cpus = 0;

2452 2453
	RCU_LOCKDEP_WARN(!rcu_read_lock_sched_held(),
			 "sched RCU must be held");
2454 2455 2456 2457
	for_each_cpu_and(i, rd->span, cpu_active_mask)
		cpus++;

	return cpus;
2458 2459 2460 2461 2462 2463 2464
}
#else
inline struct dl_bw *dl_bw_of(int i)
{
	return &cpu_rq(i)->dl.dl_bw;
}

2465
static inline int dl_bw_cpus(int i)
2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483
{
	return 1;
}
#endif

/*
 * We must be sure that accepting a new task (or allowing changing the
 * parameters of an existing one) is consistent with the bandwidth
 * constraints. If yes, this function also accordingly updates the currently
 * allocated bandwidth to reflect the new situation.
 *
 * This function is called while holding p's rq->lock.
 */
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
	if (dl_policy(policy) && !task_has_dl_policy(p) &&
	    !__dl_overflow(dl_b, cpus, 0, new_bw)) {
2502 2503
		if (hrtimer_active(&p->dl.inactive_timer))
			__dl_clear(dl_b, p->dl.dl_bw);
2504 2505 2506 2507
		__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)) {
2508 2509 2510 2511 2512 2513 2514
		/*
		 * XXX this is slightly incorrect: when the task
		 * utilization decreases, we should delay the total
		 * utilization change until the task's 0-lag point.
		 * But this would require to set the task's "inactive
		 * timer" when the task is not inactive.
		 */
2515 2516
		__dl_clear(dl_b, p->dl.dl_bw);
		__dl_add(dl_b, new_bw);
2517
		dl_change_utilization(p, new_bw);
2518 2519
		err = 0;
	} else if (!dl_policy(policy) && task_has_dl_policy(p)) {
2520 2521 2522 2523 2524
		/*
		 * Do not decrease the total deadline utilization here,
		 * switched_from_dl() will take care to do it at the correct
		 * (0-lag) time.
		 */
2525 2526 2527 2528 2529 2530 2531 2532 2533
		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
2534 2535 2536 2537 2538 2539 2540
/*
 * 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.
 */
2541
void wake_up_new_task(struct task_struct *p)
L
Linus Torvalds 已提交
2542
{
2543
	struct rq_flags rf;
I
Ingo Molnar 已提交
2544
	struct rq *rq;
2545

2546
	raw_spin_lock_irqsave(&p->pi_lock, rf.flags);
2547
	p->state = TASK_RUNNING;
2548 2549 2550 2551
#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
2552
	 *  - any previously selected CPU might disappear through hotplug
2553 2554 2555
	 *
	 * Use __set_task_cpu() to avoid calling sched_class::migrate_task_rq,
	 * as we're not fully set-up yet.
2556
	 */
2557
	__set_task_cpu(p, select_task_rq(p, task_cpu(p), SD_BALANCE_FORK, 0));
2558
#endif
2559
	rq = __task_rq_lock(p, &rf);
2560
	update_rq_clock(rq);
2561
	post_init_entity_util_avg(&p->se);
2562

2563
	activate_task(rq, p, ENQUEUE_NOCLOCK);
2564
	p->on_rq = TASK_ON_RQ_QUEUED;
2565
	trace_sched_wakeup_new(p);
P
Peter Zijlstra 已提交
2566
	check_preempt_curr(rq, p, WF_FORK);
2567
#ifdef CONFIG_SMP
2568 2569 2570 2571 2572
	if (p->sched_class->task_woken) {
		/*
		 * Nothing relies on rq->lock after this, so its fine to
		 * drop it.
		 */
2573
		rq_unpin_lock(rq, &rf);
2574
		p->sched_class->task_woken(rq, p);
2575
		rq_repin_lock(rq, &rf);
2576
	}
2577
#endif
2578
	task_rq_unlock(rq, p, &rf);
L
Linus Torvalds 已提交
2579 2580
}

2581 2582
#ifdef CONFIG_PREEMPT_NOTIFIERS

2583 2584
static struct static_key preempt_notifier_key = STATIC_KEY_INIT_FALSE;

2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596
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);

2597
/**
2598
 * preempt_notifier_register - tell me when current is being preempted & rescheduled
R
Randy Dunlap 已提交
2599
 * @notifier: notifier struct to register
2600 2601 2602
 */
void preempt_notifier_register(struct preempt_notifier *notifier)
{
2603 2604 2605
	if (!static_key_false(&preempt_notifier_key))
		WARN(1, "registering preempt_notifier while notifiers disabled\n");

2606 2607 2608 2609 2610 2611
	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
2612
 * @notifier: notifier struct to unregister
2613
 *
2614
 * This is *not* safe to call from within a preemption notifier.
2615 2616 2617 2618 2619 2620 2621
 */
void preempt_notifier_unregister(struct preempt_notifier *notifier)
{
	hlist_del(&notifier->link);
}
EXPORT_SYMBOL_GPL(preempt_notifier_unregister);

2622
static void __fire_sched_in_preempt_notifiers(struct task_struct *curr)
2623 2624 2625
{
	struct preempt_notifier *notifier;

2626
	hlist_for_each_entry(notifier, &curr->preempt_notifiers, link)
2627 2628 2629
		notifier->ops->sched_in(notifier, raw_smp_processor_id());
}

2630 2631 2632 2633 2634 2635
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);
}

2636
static void
2637 2638
__fire_sched_out_preempt_notifiers(struct task_struct *curr,
				   struct task_struct *next)
2639 2640 2641
{
	struct preempt_notifier *notifier;

2642
	hlist_for_each_entry(notifier, &curr->preempt_notifiers, link)
2643 2644 2645
		notifier->ops->sched_out(notifier, next);
}

2646 2647 2648 2649 2650 2651 2652 2653
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);
}

2654
#else /* !CONFIG_PREEMPT_NOTIFIERS */
2655

2656
static inline void fire_sched_in_preempt_notifiers(struct task_struct *curr)
2657 2658 2659
{
}

2660
static inline void
2661 2662 2663 2664 2665
fire_sched_out_preempt_notifiers(struct task_struct *curr,
				 struct task_struct *next)
{
}

2666
#endif /* CONFIG_PREEMPT_NOTIFIERS */
2667

2668 2669 2670
/**
 * prepare_task_switch - prepare to switch tasks
 * @rq: the runqueue preparing to switch
R
Randy Dunlap 已提交
2671
 * @prev: the current task that is being switched out
2672 2673 2674 2675 2676 2677 2678 2679 2680
 * @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.
 */
2681 2682 2683
static inline void
prepare_task_switch(struct rq *rq, struct task_struct *prev,
		    struct task_struct *next)
2684
{
2685
	sched_info_switch(rq, prev, next);
2686
	perf_event_task_sched_out(prev, next);
2687
	fire_sched_out_preempt_notifiers(prev, next);
2688 2689 2690 2691
	prepare_lock_switch(rq, next);
	prepare_arch_switch(next);
}

L
Linus Torvalds 已提交
2692 2693 2694 2695
/**
 * finish_task_switch - clean up after a task-switch
 * @prev: the thread we just switched away from.
 *
2696 2697 2698 2699
 * 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
2700 2701
 *
 * Note that we may have delayed dropping an mm in context_switch(). If
I
Ingo Molnar 已提交
2702
 * so, we finish that here outside of the runqueue lock. (Doing it
L
Linus Torvalds 已提交
2703 2704
 * with the lock held can cause deadlocks; see schedule() for
 * details.)
2705 2706 2707 2708 2709
 *
 * 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
2710
 */
2711
static struct rq *finish_task_switch(struct task_struct *prev)
L
Linus Torvalds 已提交
2712 2713
	__releases(rq->lock)
{
2714
	struct rq *rq = this_rq();
L
Linus Torvalds 已提交
2715
	struct mm_struct *mm = rq->prev_mm;
O
Oleg Nesterov 已提交
2716
	long prev_state;
L
Linus Torvalds 已提交
2717

2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728
	/*
	 * 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.
	 */
2729 2730 2731 2732
	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);
2733

L
Linus Torvalds 已提交
2734 2735 2736 2737
	rq->prev_mm = NULL;

	/*
	 * A task struct has one reference for the use as "current".
2738
	 * If a task dies, then it sets TASK_DEAD in tsk->state and calls
O
Oleg Nesterov 已提交
2739 2740
	 * schedule one last time. The schedule call will never return, and
	 * the scheduled task must drop that reference.
2741 2742 2743 2744 2745
	 *
	 * 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
2746
	 */
O
Oleg Nesterov 已提交
2747
	prev_state = prev->state;
2748
	vtime_task_switch(prev);
2749
	perf_event_task_sched_in(prev, current);
2750
	finish_lock_switch(rq, prev);
2751
	finish_arch_post_lock_switch();
S
Steven Rostedt 已提交
2752

2753
	fire_sched_in_preempt_notifiers(current);
L
Linus Torvalds 已提交
2754 2755
	if (mm)
		mmdrop(mm);
2756
	if (unlikely(prev_state == TASK_DEAD)) {
2757 2758 2759
		if (prev->sched_class->task_dead)
			prev->sched_class->task_dead(prev);

2760 2761 2762
		/*
		 * Remove function-return probe instances associated with this
		 * task and put them back on the free list.
I
Ingo Molnar 已提交
2763
		 */
2764
		kprobe_flush_task(prev);
2765 2766 2767 2768

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

L
Linus Torvalds 已提交
2769
		put_task_struct(prev);
2770
	}
2771

2772
	tick_nohz_task_switch();
2773
	return rq;
L
Linus Torvalds 已提交
2774 2775
}

2776 2777 2778
#ifdef CONFIG_SMP

/* rq->lock is NOT held, but preemption is disabled */
2779
static void __balance_callback(struct rq *rq)
2780
{
2781 2782 2783
	struct callback_head *head, *next;
	void (*func)(struct rq *rq);
	unsigned long flags;
2784

2785 2786 2787 2788 2789 2790 2791 2792
	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;
2793

2794
		func(rq);
2795
	}
2796 2797 2798 2799 2800 2801 2802
	raw_spin_unlock_irqrestore(&rq->lock, flags);
}

static inline void balance_callback(struct rq *rq)
{
	if (unlikely(rq->balance_callback))
		__balance_callback(rq);
2803 2804 2805
}

#else
2806

2807
static inline void balance_callback(struct rq *rq)
2808
{
L
Linus Torvalds 已提交
2809 2810
}

2811 2812
#endif

L
Linus Torvalds 已提交
2813 2814 2815 2816
/**
 * schedule_tail - first thing a freshly forked thread must call.
 * @prev: the thread we just switched away from.
 */
2817
asmlinkage __visible void schedule_tail(struct task_struct *prev)
L
Linus Torvalds 已提交
2818 2819
	__releases(rq->lock)
{
2820
	struct rq *rq;
2821

2822 2823 2824 2825 2826 2827 2828 2829 2830
	/*
	 * 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).
	 */

2831
	rq = finish_task_switch(prev);
2832
	balance_callback(rq);
2833
	preempt_enable();
2834

L
Linus Torvalds 已提交
2835
	if (current->set_child_tid)
2836
		put_user(task_pid_vnr(current), current->set_child_tid);
L
Linus Torvalds 已提交
2837 2838 2839
}

/*
2840
 * context_switch - switch to the new MM and the new thread's register state.
L
Linus Torvalds 已提交
2841
 */
2842
static __always_inline struct rq *
2843
context_switch(struct rq *rq, struct task_struct *prev,
2844
	       struct task_struct *next, struct rq_flags *rf)
L
Linus Torvalds 已提交
2845
{
I
Ingo Molnar 已提交
2846
	struct mm_struct *mm, *oldmm;
L
Linus Torvalds 已提交
2847

2848
	prepare_task_switch(rq, prev, next);
2849

I
Ingo Molnar 已提交
2850 2851
	mm = next->mm;
	oldmm = prev->active_mm;
2852 2853 2854 2855 2856
	/*
	 * For paravirt, this is coupled with an exit in switch_to to
	 * combine the page table reload and the switch backend into
	 * one hypercall.
	 */
2857
	arch_start_context_switch(prev);
2858

2859
	if (!mm) {
L
Linus Torvalds 已提交
2860
		next->active_mm = oldmm;
V
Vegard Nossum 已提交
2861
		mmgrab(oldmm);
L
Linus Torvalds 已提交
2862 2863
		enter_lazy_tlb(oldmm, next);
	} else
2864
		switch_mm_irqs_off(oldmm, mm, next);
L
Linus Torvalds 已提交
2865

2866
	if (!prev->mm) {
L
Linus Torvalds 已提交
2867 2868 2869
		prev->active_mm = NULL;
		rq->prev_mm = oldmm;
	}
2870

2871
	rq->clock_update_flags &= ~(RQCF_ACT_SKIP|RQCF_REQ_SKIP);
2872

2873 2874 2875 2876 2877 2878
	/*
	 * 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:
	 */
2879
	rq_unpin_lock(rq, rf);
2880
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
L
Linus Torvalds 已提交
2881 2882 2883

	/* Here we just switch the register state and the stack. */
	switch_to(prev, next, prev);
I
Ingo Molnar 已提交
2884
	barrier();
2885 2886

	return finish_task_switch(prev);
L
Linus Torvalds 已提交
2887 2888 2889
}

/*
2890
 * nr_running and nr_context_switches:
L
Linus Torvalds 已提交
2891 2892
 *
 * externally visible scheduler statistics: current number of runnable
2893
 * threads, total number of context switches performed since bootup.
L
Linus Torvalds 已提交
2894 2895 2896 2897 2898 2899 2900 2901 2902
 */
unsigned long nr_running(void)
{
	unsigned long i, sum = 0;

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

	return sum;
2903
}
L
Linus Torvalds 已提交
2904

2905
/*
I
Ingo Molnar 已提交
2906
 * Check if only the current task is running on the CPU.
2907 2908 2909 2910 2911 2912 2913 2914 2915 2916
 *
 * 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)
2917 2918 2919
 */
bool single_task_running(void)
{
2920
	return raw_rq()->nr_running == 1;
2921 2922 2923
}
EXPORT_SYMBOL(single_task_running);

L
Linus Torvalds 已提交
2924
unsigned long long nr_context_switches(void)
2925
{
2926 2927
	int i;
	unsigned long long sum = 0;
2928

2929
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2930
		sum += cpu_rq(i)->nr_switches;
2931

L
Linus Torvalds 已提交
2932 2933
	return sum;
}
2934

2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964
/*
 * 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
2965 2966 2967
unsigned long nr_iowait(void)
{
	unsigned long i, sum = 0;
2968

2969
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2970
		sum += atomic_read(&cpu_rq(i)->nr_iowait);
2971

L
Linus Torvalds 已提交
2972 2973
	return sum;
}
2974

2975 2976 2977 2978 2979 2980 2981
/*
 * 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.
 */

2982
unsigned long nr_iowait_cpu(int cpu)
2983
{
2984
	struct rq *this = cpu_rq(cpu);
2985 2986
	return atomic_read(&this->nr_iowait);
}
2987

2988 2989
void get_iowait_load(unsigned long *nr_waiters, unsigned long *load)
{
2990 2991 2992
	struct rq *rq = this_rq();
	*nr_waiters = atomic_read(&rq->nr_iowait);
	*load = rq->load.weight;
2993 2994
}

I
Ingo Molnar 已提交
2995
#ifdef CONFIG_SMP
2996

2997
/*
P
Peter Zijlstra 已提交
2998 2999
 * sched_exec - execve() is a valuable balancing opportunity, because at
 * this point the task has the smallest effective memory and cache footprint.
3000
 */
P
Peter Zijlstra 已提交
3001
void sched_exec(void)
3002
{
P
Peter Zijlstra 已提交
3003
	struct task_struct *p = current;
L
Linus Torvalds 已提交
3004
	unsigned long flags;
3005
	int dest_cpu;
3006

3007
	raw_spin_lock_irqsave(&p->pi_lock, flags);
3008
	dest_cpu = p->sched_class->select_task_rq(p, task_cpu(p), SD_BALANCE_EXEC, 0);
3009 3010
	if (dest_cpu == smp_processor_id())
		goto unlock;
P
Peter Zijlstra 已提交
3011

3012
	if (likely(cpu_active(dest_cpu))) {
3013
		struct migration_arg arg = { p, dest_cpu };
3014

3015 3016
		raw_spin_unlock_irqrestore(&p->pi_lock, flags);
		stop_one_cpu(task_cpu(p), migration_cpu_stop, &arg);
L
Linus Torvalds 已提交
3017 3018
		return;
	}
3019
unlock:
3020
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
3021
}
I
Ingo Molnar 已提交
3022

L
Linus Torvalds 已提交
3023 3024 3025
#endif

DEFINE_PER_CPU(struct kernel_stat, kstat);
3026
DEFINE_PER_CPU(struct kernel_cpustat, kernel_cpustat);
L
Linus Torvalds 已提交
3027 3028

EXPORT_PER_CPU_SYMBOL(kstat);
3029
EXPORT_PER_CPU_SYMBOL(kernel_cpustat);
L
Linus Torvalds 已提交
3030

3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047
/*
 * 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);
}

3048 3049 3050 3051 3052 3053 3054
/*
 * 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)
{
3055
	struct rq_flags rf;
3056
	struct rq *rq;
3057
	u64 ns;
3058

3059 3060 3061 3062 3063 3064
#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
3065 3066
	 * 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
3067
	 * indistinguishable from the read occurring a few cycles earlier.
3068 3069
	 * If we see ->on_cpu without ->on_rq, the task is leaving, and has
	 * been accounted, so we're correct here as well.
3070
	 */
3071
	if (!p->on_cpu || !task_on_rq_queued(p))
3072 3073 3074
		return p->se.sum_exec_runtime;
#endif

3075
	rq = task_rq_lock(p, &rf);
3076 3077 3078 3079 3080 3081
	/*
	 * 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)) {
3082
		prefetch_curr_exec_start(p);
3083 3084 3085 3086
		update_rq_clock(rq);
		p->sched_class->update_curr(rq);
	}
	ns = p->se.sum_exec_runtime;
3087
	task_rq_unlock(rq, p, &rf);
3088 3089 3090

	return ns;
}
3091

3092 3093 3094 3095 3096 3097 3098 3099
/*
 * 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
3100
	struct task_struct *curr = rq->curr;
3101
	struct rq_flags rf;
3102 3103

	sched_clock_tick();
I
Ingo Molnar 已提交
3104

3105 3106
	rq_lock(rq, &rf);

3107
	update_rq_clock(rq);
P
Peter Zijlstra 已提交
3108
	curr->sched_class->task_tick(rq, curr, 0);
3109
	cpu_load_update_active(rq);
3110
	calc_global_load_tick(rq);
3111 3112

	rq_unlock(rq, &rf);
3113

3114
	perf_event_task_tick();
3115

3116
#ifdef CONFIG_SMP
3117
	rq->idle_balance = idle_cpu(cpu);
3118
	trigger_load_balance(rq);
3119
#endif
3120
	rq_last_tick_reset(rq);
L
Linus Torvalds 已提交
3121 3122
}

3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133
#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.
3134 3135
 *
 * Return: Maximum deferment in nanoseconds.
3136 3137 3138 3139
 */
u64 scheduler_tick_max_deferment(void)
{
	struct rq *rq = this_rq();
3140
	unsigned long next, now = READ_ONCE(jiffies);
3141 3142 3143 3144 3145 3146

	next = rq->last_sched_tick + HZ;

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

3147
	return jiffies_to_nsecs(next - now);
L
Linus Torvalds 已提交
3148
}
3149
#endif
L
Linus Torvalds 已提交
3150

3151 3152
#if defined(CONFIG_PREEMPT) && (defined(CONFIG_DEBUG_PREEMPT) || \
				defined(CONFIG_PREEMPT_TRACER))
3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166
/*
 * 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);
	}
}
3167

3168
void preempt_count_add(int val)
L
Linus Torvalds 已提交
3169
{
3170
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3171 3172 3173
	/*
	 * Underflow?
	 */
3174 3175
	if (DEBUG_LOCKS_WARN_ON((preempt_count() < 0)))
		return;
3176
#endif
3177
	__preempt_count_add(val);
3178
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3179 3180 3181
	/*
	 * Spinlock count overflowing soon?
	 */
3182 3183
	DEBUG_LOCKS_WARN_ON((preempt_count() & PREEMPT_MASK) >=
				PREEMPT_MASK - 10);
3184
#endif
3185
	preempt_latency_start(val);
L
Linus Torvalds 已提交
3186
}
3187
EXPORT_SYMBOL(preempt_count_add);
3188
NOKPROBE_SYMBOL(preempt_count_add);
L
Linus Torvalds 已提交
3189

3190 3191 3192 3193 3194 3195 3196 3197 3198 3199
/*
 * 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());
}

3200
void preempt_count_sub(int val)
L
Linus Torvalds 已提交
3201
{
3202
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3203 3204 3205
	/*
	 * Underflow?
	 */
3206
	if (DEBUG_LOCKS_WARN_ON(val > preempt_count()))
3207
		return;
L
Linus Torvalds 已提交
3208 3209 3210
	/*
	 * Is the spinlock portion underflowing?
	 */
3211 3212 3213
	if (DEBUG_LOCKS_WARN_ON((val < PREEMPT_MASK) &&
			!(preempt_count() & PREEMPT_MASK)))
		return;
3214
#endif
3215

3216
	preempt_latency_stop(val);
3217
	__preempt_count_sub(val);
L
Linus Torvalds 已提交
3218
}
3219
EXPORT_SYMBOL(preempt_count_sub);
3220
NOKPROBE_SYMBOL(preempt_count_sub);
L
Linus Torvalds 已提交
3221

3222 3223 3224
#else
static inline void preempt_latency_start(int val) { }
static inline void preempt_latency_stop(int val) { }
L
Linus Torvalds 已提交
3225 3226
#endif

3227 3228 3229 3230 3231 3232 3233 3234 3235
static inline unsigned long get_preempt_disable_ip(struct task_struct *p)
{
#ifdef CONFIG_DEBUG_PREEMPT
	return p->preempt_disable_ip;
#else
	return 0;
#endif
}

L
Linus Torvalds 已提交
3236
/*
I
Ingo Molnar 已提交
3237
 * Print scheduling while atomic bug:
L
Linus Torvalds 已提交
3238
 */
I
Ingo Molnar 已提交
3239
static noinline void __schedule_bug(struct task_struct *prev)
L
Linus Torvalds 已提交
3240
{
3241 3242 3243
	/* Save this before calling printk(), since that will clobber it */
	unsigned long preempt_disable_ip = get_preempt_disable_ip(current);

3244 3245 3246
	if (oops_in_progress)
		return;

P
Peter Zijlstra 已提交
3247 3248
	printk(KERN_ERR "BUG: scheduling while atomic: %s/%d/0x%08x\n",
		prev->comm, prev->pid, preempt_count());
3249

I
Ingo Molnar 已提交
3250
	debug_show_held_locks(prev);
3251
	print_modules();
I
Ingo Molnar 已提交
3252 3253
	if (irqs_disabled())
		print_irqtrace_events(prev);
3254 3255
	if (IS_ENABLED(CONFIG_DEBUG_PREEMPT)
	    && in_atomic_preempt_off()) {
3256
		pr_err("Preemption disabled at:");
3257
		print_ip_sym(preempt_disable_ip);
3258 3259
		pr_cont("\n");
	}
3260 3261 3262
	if (panic_on_warn)
		panic("scheduling while atomic\n");

3263
	dump_stack();
3264
	add_taint(TAINT_WARN, LOCKDEP_STILL_OK);
I
Ingo Molnar 已提交
3265
}
L
Linus Torvalds 已提交
3266

I
Ingo Molnar 已提交
3267 3268 3269 3270 3271
/*
 * Various schedule()-time debugging checks and statistics:
 */
static inline void schedule_debug(struct task_struct *prev)
{
3272
#ifdef CONFIG_SCHED_STACK_END_CHECK
J
Jann Horn 已提交
3273 3274
	if (task_stack_end_corrupted(prev))
		panic("corrupted stack end detected inside scheduler\n");
3275
#endif
3276

3277
	if (unlikely(in_atomic_preempt_off())) {
I
Ingo Molnar 已提交
3278
		__schedule_bug(prev);
3279 3280
		preempt_count_set(PREEMPT_DISABLED);
	}
3281
	rcu_sleep_check();
I
Ingo Molnar 已提交
3282

L
Linus Torvalds 已提交
3283 3284
	profile_hit(SCHED_PROFILING, __builtin_return_address(0));

3285
	schedstat_inc(this_rq()->sched_count);
I
Ingo Molnar 已提交
3286 3287 3288 3289 3290 3291
}

/*
 * Pick up the highest-prio task:
 */
static inline struct task_struct *
3292
pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
I
Ingo Molnar 已提交
3293
{
3294
	const struct sched_class *class;
I
Ingo Molnar 已提交
3295
	struct task_struct *p;
L
Linus Torvalds 已提交
3296 3297

	/*
3298 3299 3300 3301
	 * Optimization: we know that if all tasks are in the fair class we can
	 * call that function directly, but only if the @prev task wasn't of a
	 * higher scheduling class, because otherwise those loose the
	 * opportunity to pull in more work from other CPUs.
L
Linus Torvalds 已提交
3302
	 */
3303 3304 3305 3306
	if (likely((prev->sched_class == &idle_sched_class ||
		    prev->sched_class == &fair_sched_class) &&
		   rq->nr_running == rq->cfs.h_nr_running)) {

3307
		p = fair_sched_class.pick_next_task(rq, prev, rf);
3308 3309 3310
		if (unlikely(p == RETRY_TASK))
			goto again;

I
Ingo Molnar 已提交
3311
		/* Assumes fair_sched_class->next == idle_sched_class */
3312
		if (unlikely(!p))
3313
			p = idle_sched_class.pick_next_task(rq, prev, rf);
3314 3315

		return p;
L
Linus Torvalds 已提交
3316 3317
	}

3318
again:
3319
	for_each_class(class) {
3320
		p = class->pick_next_task(rq, prev, rf);
3321 3322 3323
		if (p) {
			if (unlikely(p == RETRY_TASK))
				goto again;
I
Ingo Molnar 已提交
3324
			return p;
3325
		}
I
Ingo Molnar 已提交
3326
	}
3327

I
Ingo Molnar 已提交
3328 3329
	/* The idle class should always have a runnable task: */
	BUG();
I
Ingo Molnar 已提交
3330
}
L
Linus Torvalds 已提交
3331

I
Ingo Molnar 已提交
3332
/*
3333
 * __schedule() is the main scheduler function.
3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367
 *
 * 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
3368
 *
3369
 * WARNING: must be called with preemption disabled!
I
Ingo Molnar 已提交
3370
 */
3371
static void __sched notrace __schedule(bool preempt)
I
Ingo Molnar 已提交
3372 3373
{
	struct task_struct *prev, *next;
3374
	unsigned long *switch_count;
3375
	struct rq_flags rf;
I
Ingo Molnar 已提交
3376
	struct rq *rq;
3377
	int cpu;
I
Ingo Molnar 已提交
3378 3379 3380 3381 3382 3383

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

	schedule_debug(prev);
L
Linus Torvalds 已提交
3384

3385
	if (sched_feat(HRTICK))
M
Mike Galbraith 已提交
3386
		hrtick_clear(rq);
P
Peter Zijlstra 已提交
3387

3388
	local_irq_disable();
3389
	rcu_note_context_switch(preempt);
3390

3391 3392 3393 3394 3395 3396
	/*
	 * 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();
3397
	rq_lock(rq, &rf);
L
Linus Torvalds 已提交
3398

I
Ingo Molnar 已提交
3399 3400
	/* Promote REQ to ACT */
	rq->clock_update_flags <<= 1;
3401
	update_rq_clock(rq);
3402

3403
	switch_count = &prev->nivcsw;
3404
	if (!preempt && prev->state) {
T
Tejun Heo 已提交
3405
		if (unlikely(signal_pending_state(prev->state, prev))) {
L
Linus Torvalds 已提交
3406
			prev->state = TASK_RUNNING;
T
Tejun Heo 已提交
3407
		} else {
3408
			deactivate_task(rq, prev, DEQUEUE_SLEEP | DEQUEUE_NOCLOCK);
3409 3410
			prev->on_rq = 0;

3411 3412 3413 3414 3415
			if (prev->in_iowait) {
				atomic_inc(&rq->nr_iowait);
				delayacct_blkio_start();
			}

T
Tejun Heo 已提交
3416
			/*
3417 3418 3419
			 * 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
3420 3421 3422 3423
			 */
			if (prev->flags & PF_WQ_WORKER) {
				struct task_struct *to_wakeup;

3424
				to_wakeup = wq_worker_sleeping(prev);
T
Tejun Heo 已提交
3425
				if (to_wakeup)
3426
					try_to_wake_up_local(to_wakeup, &rf);
T
Tejun Heo 已提交
3427 3428
			}
		}
I
Ingo Molnar 已提交
3429
		switch_count = &prev->nvcsw;
L
Linus Torvalds 已提交
3430 3431
	}

3432
	next = pick_next_task(rq, prev, &rf);
3433
	clear_tsk_need_resched(prev);
3434
	clear_preempt_need_resched();
L
Linus Torvalds 已提交
3435 3436 3437 3438 3439 3440

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

3441
		trace_sched_switch(preempt, prev, next);
I
Ingo Molnar 已提交
3442 3443 3444

		/* Also unlocks the rq: */
		rq = context_switch(rq, prev, next, &rf);
3445
	} else {
3446
		rq->clock_update_flags &= ~(RQCF_ACT_SKIP|RQCF_REQ_SKIP);
3447
		rq_unlock_irq(rq, &rf);
3448
	}
L
Linus Torvalds 已提交
3449

3450
	balance_callback(rq);
L
Linus Torvalds 已提交
3451
}
3452

3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469
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
3470
	/* Causes final put_task_struct in finish_task_switch(): */
3471
	__set_current_state(TASK_DEAD);
I
Ingo Molnar 已提交
3472 3473 3474 3475

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

3476 3477
	__schedule(false);
	BUG();
I
Ingo Molnar 已提交
3478 3479

	/* Avoid "noreturn function does return" - but don't continue if BUG() is a NOP: */
3480
	for (;;)
I
Ingo Molnar 已提交
3481
		cpu_relax();
3482 3483
}

3484 3485
static inline void sched_submit_work(struct task_struct *tsk)
{
3486
	if (!tsk->state || tsk_is_pi_blocked(tsk))
3487 3488 3489 3490 3491 3492 3493 3494 3495
		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);
}

3496
asmlinkage __visible void __sched schedule(void)
3497
{
3498 3499 3500
	struct task_struct *tsk = current;

	sched_submit_work(tsk);
3501
	do {
3502
		preempt_disable();
3503
		__schedule(false);
3504
		sched_preempt_enable_no_resched();
3505
	} while (need_resched());
3506
}
L
Linus Torvalds 已提交
3507 3508
EXPORT_SYMBOL(schedule);

3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533
/*
 * synchronize_rcu_tasks() makes sure that no task is stuck in preempted
 * state (have scheduled out non-voluntarily) by making sure that all
 * tasks have either left the run queue or have gone into user space.
 * As idle tasks do not do either, they must not ever be preempted
 * (schedule out non-voluntarily).
 *
 * schedule_idle() is similar to schedule_preempt_disable() except that it
 * never enables preemption because it does not call sched_submit_work().
 */
void __sched schedule_idle(void)
{
	/*
	 * As this skips calling sched_submit_work(), which the idle task does
	 * regardless because that function is a nop when the task is in a
	 * TASK_RUNNING state, make sure this isn't used someplace that the
	 * current task can be in any other state. Note, idle is always in the
	 * TASK_RUNNING state.
	 */
	WARN_ON_ONCE(current->state);
	do {
		__schedule(false);
	} while (need_resched());
}

3534
#ifdef CONFIG_CONTEXT_TRACKING
3535
asmlinkage __visible void __sched schedule_user(void)
3536 3537 3538 3539 3540 3541
{
	/*
	 * 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.
3542 3543
	 *
	 * NB: There are buggy callers of this function.  Ideally we
3544
	 * should warn if prev_state != CONTEXT_USER, but that will trigger
3545
	 * too frequently to make sense yet.
3546
	 */
3547
	enum ctx_state prev_state = exception_enter();
3548
	schedule();
3549
	exception_exit(prev_state);
3550 3551 3552
}
#endif

3553 3554 3555 3556 3557 3558 3559
/**
 * schedule_preempt_disabled - called with preemption disabled
 *
 * Returns with preemption disabled. Note: preempt_count must be 1
 */
void __sched schedule_preempt_disabled(void)
{
3560
	sched_preempt_enable_no_resched();
3561 3562 3563 3564
	schedule();
	preempt_disable();
}

3565
static void __sched notrace preempt_schedule_common(void)
3566 3567
{
	do {
3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580
		/*
		 * 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.
		 */
3581
		preempt_disable_notrace();
3582
		preempt_latency_start(1);
3583
		__schedule(true);
3584
		preempt_latency_stop(1);
3585
		preempt_enable_no_resched_notrace();
3586 3587 3588 3589 3590 3591 3592 3593

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

L
Linus Torvalds 已提交
3594 3595
#ifdef CONFIG_PREEMPT
/*
3596
 * this is the entry point to schedule() from in-kernel preemption
I
Ingo Molnar 已提交
3597
 * off of preempt_enable. Kernel preemptions off return from interrupt
L
Linus Torvalds 已提交
3598 3599
 * occur there and call schedule directly.
 */
3600
asmlinkage __visible void __sched notrace preempt_schedule(void)
L
Linus Torvalds 已提交
3601 3602 3603
{
	/*
	 * If there is a non-zero preempt_count or interrupts are disabled,
I
Ingo Molnar 已提交
3604
	 * we do not want to preempt the current task. Just return..
L
Linus Torvalds 已提交
3605
	 */
3606
	if (likely(!preemptible()))
L
Linus Torvalds 已提交
3607 3608
		return;

3609
	preempt_schedule_common();
L
Linus Torvalds 已提交
3610
}
3611
NOKPROBE_SYMBOL(preempt_schedule);
L
Linus Torvalds 已提交
3612
EXPORT_SYMBOL(preempt_schedule);
3613 3614

/**
3615
 * preempt_schedule_notrace - preempt_schedule called by tracing
3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627
 *
 * 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.
 */
3628
asmlinkage __visible void __sched notrace preempt_schedule_notrace(void)
3629 3630 3631 3632 3633 3634 3635
{
	enum ctx_state prev_ctx;

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

	do {
3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648
		/*
		 * 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.
		 */
3649
		preempt_disable_notrace();
3650
		preempt_latency_start(1);
3651 3652 3653 3654 3655 3656
		/*
		 * Needs preempt disabled in case user_exit() is traced
		 * and the tracer calls preempt_enable_notrace() causing
		 * an infinite recursion.
		 */
		prev_ctx = exception_enter();
3657
		__schedule(true);
3658 3659
		exception_exit(prev_ctx);

3660
		preempt_latency_stop(1);
3661
		preempt_enable_no_resched_notrace();
3662 3663
	} while (need_resched());
}
3664
EXPORT_SYMBOL_GPL(preempt_schedule_notrace);
3665

3666
#endif /* CONFIG_PREEMPT */
L
Linus Torvalds 已提交
3667 3668

/*
3669
 * this is the entry point to schedule() from kernel preemption
L
Linus Torvalds 已提交
3670 3671 3672 3673
 * off of irq context.
 * Note, that this is called and return with irqs disabled. This will
 * protect us against recursive calling from irq.
 */
3674
asmlinkage __visible void __sched preempt_schedule_irq(void)
L
Linus Torvalds 已提交
3675
{
3676
	enum ctx_state prev_state;
3677

3678
	/* Catch callers which need to be fixed */
3679
	BUG_ON(preempt_count() || !irqs_disabled());
L
Linus Torvalds 已提交
3680

3681 3682
	prev_state = exception_enter();

3683
	do {
3684
		preempt_disable();
3685
		local_irq_enable();
3686
		__schedule(true);
3687
		local_irq_disable();
3688
		sched_preempt_enable_no_resched();
3689
	} while (need_resched());
3690 3691

	exception_exit(prev_state);
L
Linus Torvalds 已提交
3692 3693
}

P
Peter Zijlstra 已提交
3694
int default_wake_function(wait_queue_t *curr, unsigned mode, int wake_flags,
I
Ingo Molnar 已提交
3695
			  void *key)
L
Linus Torvalds 已提交
3696
{
P
Peter Zijlstra 已提交
3697
	return try_to_wake_up(curr->private, mode, wake_flags);
L
Linus Torvalds 已提交
3698 3699 3700
}
EXPORT_SYMBOL(default_wake_function);

3701 3702
#ifdef CONFIG_RT_MUTEXES

3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717
static inline int __rt_effective_prio(struct task_struct *pi_task, int prio)
{
	if (pi_task)
		prio = min(prio, pi_task->prio);

	return prio;
}

static inline int rt_effective_prio(struct task_struct *p, int prio)
{
	struct task_struct *pi_task = rt_mutex_get_top_task(p);

	return __rt_effective_prio(pi_task, prio);
}

3718 3719
/*
 * rt_mutex_setprio - set the current priority of a task
3720 3721
 * @p: task to boost
 * @pi_task: donor task
3722 3723 3724 3725
 *
 * This function changes the 'effective' priority of a task. It does
 * not touch ->normal_prio like __setscheduler().
 *
3726 3727
 * Used by the rt_mutex code to implement priority inheritance
 * logic. Call site only calls if the priority of the task changed.
3728
 */
3729
void rt_mutex_setprio(struct task_struct *p, struct task_struct *pi_task)
3730
{
3731
	int prio, oldprio, queued, running, queue_flag =
3732
		DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
3733
	const struct sched_class *prev_class;
3734 3735
	struct rq_flags rf;
	struct rq *rq;
3736

3737 3738 3739 3740 3741 3742 3743 3744
	/* XXX used to be waiter->prio, not waiter->task->prio */
	prio = __rt_effective_prio(pi_task, p->normal_prio);

	/*
	 * If nothing changed; bail early.
	 */
	if (p->pi_top_task == pi_task && prio == p->prio && !dl_prio(prio))
		return;
3745

3746
	rq = __task_rq_lock(p, &rf);
3747
	update_rq_clock(rq);
3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764
	/*
	 * Set under pi_lock && rq->lock, such that the value can be used under
	 * either lock.
	 *
	 * Note that there is loads of tricky to make this pointer cache work
	 * right. rt_mutex_slowunlock()+rt_mutex_postunlock() work together to
	 * ensure a task is de-boosted (pi_task is set to NULL) before the
	 * task is allowed to run again (and can exit). This ensures the pointer
	 * points to a blocked task -- which guaratees the task is present.
	 */
	p->pi_top_task = pi_task;

	/*
	 * For FIFO/RR we only need to set prio, if that matches we're done.
	 */
	if (prio == p->prio && !dl_prio(prio))
		goto out_unlock;
3765

3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783
	/*
	 * 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;
	}

3784
	trace_sched_pi_setprio(p, pi_task);
3785
	oldprio = p->prio;
3786 3787 3788 3789

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

3790
	prev_class = p->sched_class;
3791
	queued = task_on_rq_queued(p);
3792
	running = task_current(rq, p);
3793
	if (queued)
3794
		dequeue_task(rq, p, queue_flag);
3795
	if (running)
3796
		put_prev_task(rq, p);
I
Ingo Molnar 已提交
3797

3798 3799 3800 3801 3802 3803 3804 3805 3806 3807
	/*
	 * 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)) {
3808 3809
		if (!dl_prio(p->normal_prio) ||
		    (pi_task && dl_entity_preempt(&pi_task->dl, &p->dl))) {
3810
			p->dl.dl_boosted = 1;
3811
			queue_flag |= ENQUEUE_REPLENISH;
3812 3813
		} else
			p->dl.dl_boosted = 0;
3814
		p->sched_class = &dl_sched_class;
3815 3816 3817 3818
	} else if (rt_prio(prio)) {
		if (dl_prio(oldprio))
			p->dl.dl_boosted = 0;
		if (oldprio < prio)
3819
			queue_flag |= ENQUEUE_HEAD;
I
Ingo Molnar 已提交
3820
		p->sched_class = &rt_sched_class;
3821 3822 3823
	} else {
		if (dl_prio(oldprio))
			p->dl.dl_boosted = 0;
3824 3825
		if (rt_prio(oldprio))
			p->rt.timeout = 0;
I
Ingo Molnar 已提交
3826
		p->sched_class = &fair_sched_class;
3827
	}
I
Ingo Molnar 已提交
3828

3829 3830
	p->prio = prio;

3831
	if (queued)
3832
		enqueue_task(rq, p, queue_flag);
3833
	if (running)
3834
		set_curr_task(rq, p);
3835

P
Peter Zijlstra 已提交
3836
	check_class_changed(rq, p, prev_class, oldprio);
3837
out_unlock:
I
Ingo Molnar 已提交
3838 3839
	/* Avoid rq from going away on us: */
	preempt_disable();
3840
	__task_rq_unlock(rq, &rf);
3841 3842 3843

	balance_callback(rq);
	preempt_enable();
3844
}
3845 3846 3847 3848 3849
#else
static inline int rt_effective_prio(struct task_struct *p, int prio)
{
	return prio;
}
3850
#endif
3851

3852
void set_user_nice(struct task_struct *p, long nice)
L
Linus Torvalds 已提交
3853
{
P
Peter Zijlstra 已提交
3854 3855
	bool queued, running;
	int old_prio, delta;
3856
	struct rq_flags rf;
3857
	struct rq *rq;
L
Linus Torvalds 已提交
3858

3859
	if (task_nice(p) == nice || nice < MIN_NICE || nice > MAX_NICE)
L
Linus Torvalds 已提交
3860 3861 3862 3863 3864
		return;
	/*
	 * We have to be careful, if called from sys_setpriority(),
	 * the task might be in the middle of scheduling on another CPU.
	 */
3865
	rq = task_rq_lock(p, &rf);
3866 3867
	update_rq_clock(rq);

L
Linus Torvalds 已提交
3868 3869 3870 3871
	/*
	 * 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
3872
	 * SCHED_DEADLINE, SCHED_FIFO or SCHED_RR:
L
Linus Torvalds 已提交
3873
	 */
3874
	if (task_has_dl_policy(p) || task_has_rt_policy(p)) {
L
Linus Torvalds 已提交
3875 3876 3877
		p->static_prio = NICE_TO_PRIO(nice);
		goto out_unlock;
	}
3878
	queued = task_on_rq_queued(p);
P
Peter Zijlstra 已提交
3879
	running = task_current(rq, p);
3880
	if (queued)
3881
		dequeue_task(rq, p, DEQUEUE_SAVE | DEQUEUE_NOCLOCK);
P
Peter Zijlstra 已提交
3882 3883
	if (running)
		put_prev_task(rq, p);
L
Linus Torvalds 已提交
3884 3885

	p->static_prio = NICE_TO_PRIO(nice);
3886
	set_load_weight(p);
3887 3888 3889
	old_prio = p->prio;
	p->prio = effective_prio(p);
	delta = p->prio - old_prio;
L
Linus Torvalds 已提交
3890

3891
	if (queued) {
3892
		enqueue_task(rq, p, ENQUEUE_RESTORE | ENQUEUE_NOCLOCK);
L
Linus Torvalds 已提交
3893
		/*
3894 3895
		 * If the task increased its priority or is running and
		 * lowered its priority, then reschedule its CPU:
L
Linus Torvalds 已提交
3896
		 */
3897
		if (delta < 0 || (delta > 0 && task_running(rq, p)))
3898
			resched_curr(rq);
L
Linus Torvalds 已提交
3899
	}
P
Peter Zijlstra 已提交
3900 3901
	if (running)
		set_curr_task(rq, p);
L
Linus Torvalds 已提交
3902
out_unlock:
3903
	task_rq_unlock(rq, p, &rf);
L
Linus Torvalds 已提交
3904 3905 3906
}
EXPORT_SYMBOL(set_user_nice);

M
Matt Mackall 已提交
3907 3908 3909 3910 3911
/*
 * can_nice - check if a task can reduce its nice value
 * @p: task
 * @nice: nice value
 */
3912
int can_nice(const struct task_struct *p, const int nice)
M
Matt Mackall 已提交
3913
{
I
Ingo Molnar 已提交
3914
	/* Convert nice value [19,-20] to rlimit style value [1,40]: */
3915
	int nice_rlim = nice_to_rlimit(nice);
3916

3917
	return (nice_rlim <= task_rlimit(p, RLIMIT_NICE) ||
M
Matt Mackall 已提交
3918 3919 3920
		capable(CAP_SYS_NICE));
}

L
Linus Torvalds 已提交
3921 3922 3923 3924 3925 3926 3927 3928 3929
#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.
 */
3930
SYSCALL_DEFINE1(nice, int, increment)
L
Linus Torvalds 已提交
3931
{
3932
	long nice, retval;
L
Linus Torvalds 已提交
3933 3934 3935 3936 3937 3938

	/*
	 * 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.
	 */
3939
	increment = clamp(increment, -NICE_WIDTH, NICE_WIDTH);
3940
	nice = task_nice(current) + increment;
L
Linus Torvalds 已提交
3941

3942
	nice = clamp_val(nice, MIN_NICE, MAX_NICE);
M
Matt Mackall 已提交
3943 3944 3945
	if (increment < 0 && !can_nice(current, nice))
		return -EPERM;

L
Linus Torvalds 已提交
3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959
	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.
 *
3960
 * Return: The priority value as seen by users in /proc.
L
Linus Torvalds 已提交
3961 3962 3963
 * RT tasks are offset by -200. Normal tasks are centered
 * around 0, value goes from -16 to +15.
 */
3964
int task_prio(const struct task_struct *p)
L
Linus Torvalds 已提交
3965 3966 3967 3968 3969
{
	return p->prio - MAX_RT_PRIO;
}

/**
I
Ingo Molnar 已提交
3970
 * idle_cpu - is a given CPU idle currently?
L
Linus Torvalds 已提交
3971
 * @cpu: the processor in question.
3972 3973
 *
 * Return: 1 if the CPU is currently idle. 0 otherwise.
L
Linus Torvalds 已提交
3974 3975 3976
 */
int idle_cpu(int cpu)
{
T
Thomas Gleixner 已提交
3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990
	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
3991 3992 3993
}

/**
I
Ingo Molnar 已提交
3994
 * idle_task - return the idle task for a given CPU.
L
Linus Torvalds 已提交
3995
 * @cpu: the processor in question.
3996
 *
I
Ingo Molnar 已提交
3997
 * Return: The idle task for the CPU @cpu.
L
Linus Torvalds 已提交
3998
 */
3999
struct task_struct *idle_task(int cpu)
L
Linus Torvalds 已提交
4000 4001 4002 4003 4004 4005 4006
{
	return cpu_rq(cpu)->idle;
}

/**
 * find_process_by_pid - find a process with a matching PID value.
 * @pid: the pid in question.
4007 4008
 *
 * The task of @pid, if found. %NULL otherwise.
L
Linus Torvalds 已提交
4009
 */
A
Alexey Dobriyan 已提交
4010
static struct task_struct *find_process_by_pid(pid_t pid)
L
Linus Torvalds 已提交
4011
{
4012
	return pid ? find_task_by_vpid(pid) : current;
L
Linus Torvalds 已提交
4013 4014
}

4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029
/*
 * 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;
4030
	dl_se->dl_period = attr->sched_period ?: dl_se->dl_deadline;
4031
	dl_se->flags = attr->sched_flags;
4032
	dl_se->dl_bw = to_ratio(dl_se->dl_period, dl_se->dl_runtime);
4033 4034
}

4035 4036 4037 4038 4039 4040
/*
 * sched_setparam() passes in -1 for its policy, to let the functions
 * it calls know not to change it.
 */
#define SETPARAM_POLICY	-1

4041 4042
static void __setscheduler_params(struct task_struct *p,
		const struct sched_attr *attr)
L
Linus Torvalds 已提交
4043
{
4044 4045
	int policy = attr->sched_policy;

4046
	if (policy == SETPARAM_POLICY)
4047 4048
		policy = p->policy;

L
Linus Torvalds 已提交
4049
	p->policy = policy;
4050

4051 4052
	if (dl_policy(policy))
		__setparam_dl(p, attr);
4053
	else if (fair_policy(policy))
4054 4055
		p->static_prio = NICE_TO_PRIO(attr->sched_nice);

4056 4057 4058 4059 4060 4061
	/*
	 * __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;
4062
	p->normal_prio = normal_prio(p);
4063 4064
	set_load_weight(p);
}
4065

4066 4067
/* Actually do priority change: must hold pi & rq lock. */
static void __setscheduler(struct rq *rq, struct task_struct *p,
4068
			   const struct sched_attr *attr, bool keep_boost)
4069 4070
{
	__setscheduler_params(p, attr);
4071

4072
	/*
4073 4074
	 * Keep a potential priority boosting if called from
	 * sched_setscheduler().
4075
	 */
4076
	p->prio = normal_prio(p);
4077
	if (keep_boost)
4078
		p->prio = rt_effective_prio(p, p->prio);
4079

4080 4081 4082
	if (dl_prio(p->prio))
		p->sched_class = &dl_sched_class;
	else if (rt_prio(p->prio))
4083 4084 4085
		p->sched_class = &rt_sched_class;
	else
		p->sched_class = &fair_sched_class;
L
Linus Torvalds 已提交
4086
}
4087 4088 4089 4090 4091 4092 4093 4094 4095

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;
4096
	attr->sched_period = dl_se->dl_period;
4097 4098 4099 4100 4101 4102
	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
4103
 * than the runtime, as well as the period of being zero or
4104
 * greater than deadline. Furthermore, we have to be sure that
4105 4106 4107 4108
 * 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).
4109 4110 4111 4112
 */
static bool
__checkparam_dl(const struct sched_attr *attr)
{
4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138
	/* 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;
4139 4140
}

4141
/*
I
Ingo Molnar 已提交
4142
 * Check the target process has a UID that matches the current process's:
4143 4144 4145 4146 4147 4148 4149 4150
 */
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);
4151 4152
	match = (uid_eq(cred->euid, pcred->euid) ||
		 uid_eq(cred->euid, pcred->uid));
4153 4154 4155 4156
	rcu_read_unlock();
	return match;
}

I
Ingo Molnar 已提交
4157
static bool dl_param_changed(struct task_struct *p, const struct sched_attr *attr)
4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169
{
	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;
}

4170 4171
static int __sched_setscheduler(struct task_struct *p,
				const struct sched_attr *attr,
4172
				bool user, bool pi)
L
Linus Torvalds 已提交
4173
{
4174 4175
	int newprio = dl_policy(attr->sched_policy) ? MAX_DL_PRIO - 1 :
		      MAX_RT_PRIO - 1 - attr->sched_priority;
4176
	int retval, oldprio, oldpolicy = -1, queued, running;
4177
	int new_effective_prio, policy = attr->sched_policy;
4178
	const struct sched_class *prev_class;
4179
	struct rq_flags rf;
4180
	int reset_on_fork;
4181
	int queue_flags = DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
4182
	struct rq *rq;
L
Linus Torvalds 已提交
4183

4184 4185
	/* The pi code expects interrupts enabled */
	BUG_ON(pi && in_interrupt());
L
Linus Torvalds 已提交
4186
recheck:
I
Ingo Molnar 已提交
4187
	/* Double check policy once rq lock held: */
4188 4189
	if (policy < 0) {
		reset_on_fork = p->sched_reset_on_fork;
L
Linus Torvalds 已提交
4190
		policy = oldpolicy = p->policy;
4191
	} else {
4192
		reset_on_fork = !!(attr->sched_flags & SCHED_FLAG_RESET_ON_FORK);
4193

4194
		if (!valid_policy(policy))
4195 4196 4197
			return -EINVAL;
	}

4198 4199 4200
	if (attr->sched_flags & ~(SCHED_FLAG_RESET_ON_FORK))
		return -EINVAL;

L
Linus Torvalds 已提交
4201 4202
	/*
	 * Valid priorities for SCHED_FIFO and SCHED_RR are
I
Ingo Molnar 已提交
4203 4204
	 * 1..MAX_USER_RT_PRIO-1, valid priority for SCHED_NORMAL,
	 * SCHED_BATCH and SCHED_IDLE is 0.
L
Linus Torvalds 已提交
4205
	 */
4206
	if ((p->mm && attr->sched_priority > MAX_USER_RT_PRIO-1) ||
4207
	    (!p->mm && attr->sched_priority > MAX_RT_PRIO-1))
L
Linus Torvalds 已提交
4208
		return -EINVAL;
4209 4210
	if ((dl_policy(policy) && !__checkparam_dl(attr)) ||
	    (rt_policy(policy) != (attr->sched_priority != 0)))
L
Linus Torvalds 已提交
4211 4212
		return -EINVAL;

4213 4214 4215
	/*
	 * Allow unprivileged RT tasks to decrease priority:
	 */
4216
	if (user && !capable(CAP_SYS_NICE)) {
4217
		if (fair_policy(policy)) {
4218
			if (attr->sched_nice < task_nice(p) &&
4219
			    !can_nice(p, attr->sched_nice))
4220 4221 4222
				return -EPERM;
		}

4223
		if (rt_policy(policy)) {
4224 4225
			unsigned long rlim_rtprio =
					task_rlimit(p, RLIMIT_RTPRIO);
4226

I
Ingo Molnar 已提交
4227
			/* Can't set/change the rt policy: */
4228 4229 4230
			if (policy != p->policy && !rlim_rtprio)
				return -EPERM;

I
Ingo Molnar 已提交
4231
			/* Can't increase priority: */
4232 4233
			if (attr->sched_priority > p->rt_priority &&
			    attr->sched_priority > rlim_rtprio)
4234 4235
				return -EPERM;
		}
4236

4237 4238 4239 4240 4241 4242 4243 4244 4245
		 /*
		  * 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
4246
		/*
4247 4248
		 * 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
4249
		 */
4250
		if (idle_policy(p->policy) && !idle_policy(policy)) {
4251
			if (!can_nice(p, task_nice(p)))
4252 4253
				return -EPERM;
		}
4254

I
Ingo Molnar 已提交
4255
		/* Can't change other user's priorities: */
4256
		if (!check_same_owner(p))
4257
			return -EPERM;
4258

I
Ingo Molnar 已提交
4259
		/* Normal users shall not reset the sched_reset_on_fork flag: */
4260 4261
		if (p->sched_reset_on_fork && !reset_on_fork)
			return -EPERM;
4262
	}
L
Linus Torvalds 已提交
4263

4264
	if (user) {
4265
		retval = security_task_setscheduler(p);
4266 4267 4268 4269
		if (retval)
			return retval;
	}

4270
	/*
I
Ingo Molnar 已提交
4271
	 * Make sure no PI-waiters arrive (or leave) while we are
4272
	 * changing the priority of the task:
4273
	 *
L
Lucas De Marchi 已提交
4274
	 * To be able to change p->policy safely, the appropriate
L
Linus Torvalds 已提交
4275 4276
	 * runqueue lock must be held.
	 */
4277
	rq = task_rq_lock(p, &rf);
4278
	update_rq_clock(rq);
4279

4280
	/*
I
Ingo Molnar 已提交
4281
	 * Changing the policy of the stop threads its a very bad idea:
4282 4283
	 */
	if (p == rq->stop) {
4284
		task_rq_unlock(rq, p, &rf);
4285 4286 4287
		return -EINVAL;
	}

4288
	/*
4289 4290
	 * If not changing anything there's no need to proceed further,
	 * but store a possible modification of reset_on_fork.
4291
	 */
4292
	if (unlikely(policy == p->policy)) {
4293
		if (fair_policy(policy) && attr->sched_nice != task_nice(p))
4294 4295 4296
			goto change;
		if (rt_policy(policy) && attr->sched_priority != p->rt_priority)
			goto change;
4297
		if (dl_policy(policy) && dl_param_changed(p, attr))
4298
			goto change;
4299

4300
		p->sched_reset_on_fork = reset_on_fork;
4301
		task_rq_unlock(rq, p, &rf);
4302 4303
		return 0;
	}
4304
change:
4305

4306
	if (user) {
4307
#ifdef CONFIG_RT_GROUP_SCHED
4308 4309 4310 4311 4312
		/*
		 * Do not allow realtime tasks into groups that have no runtime
		 * assigned.
		 */
		if (rt_bandwidth_enabled() && rt_policy(policy) &&
4313 4314
				task_group(p)->rt_bandwidth.rt_runtime == 0 &&
				!task_group_is_autogroup(task_group(p))) {
4315
			task_rq_unlock(rq, p, &rf);
4316 4317 4318
			return -EPERM;
		}
#endif
4319 4320 4321 4322 4323 4324 4325 4326 4327
#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.
			 */
4328 4329
			if (!cpumask_subset(span, &p->cpus_allowed) ||
			    rq->rd->dl_bw.bw == 0) {
4330
				task_rq_unlock(rq, p, &rf);
4331 4332 4333 4334 4335
				return -EPERM;
			}
		}
#endif
	}
4336

I
Ingo Molnar 已提交
4337
	/* Re-check policy now with rq lock held: */
L
Linus Torvalds 已提交
4338 4339
	if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) {
		policy = oldpolicy = -1;
4340
		task_rq_unlock(rq, p, &rf);
L
Linus Torvalds 已提交
4341 4342
		goto recheck;
	}
4343 4344 4345 4346 4347 4348

	/*
	 * If setscheduling to SCHED_DEADLINE (or changing the parameters
	 * of a SCHED_DEADLINE task) we need to check if enough bandwidth
	 * is available.
	 */
4349
	if ((dl_policy(policy) || dl_task(p)) && dl_overflow(p, policy, attr)) {
4350
		task_rq_unlock(rq, p, &rf);
4351 4352 4353
		return -EBUSY;
	}

4354 4355 4356
	p->sched_reset_on_fork = reset_on_fork;
	oldprio = p->prio;

4357 4358 4359 4360 4361 4362 4363 4364
	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.
		 */
4365
		new_effective_prio = rt_effective_prio(p, newprio);
4366 4367
		if (new_effective_prio == oldprio)
			queue_flags &= ~DEQUEUE_MOVE;
4368 4369
	}

4370
	queued = task_on_rq_queued(p);
4371
	running = task_current(rq, p);
4372
	if (queued)
4373
		dequeue_task(rq, p, queue_flags);
4374
	if (running)
4375
		put_prev_task(rq, p);
4376

4377
	prev_class = p->sched_class;
4378
	__setscheduler(rq, p, attr, pi);
4379

4380
	if (queued) {
4381 4382 4383 4384
		/*
		 * We enqueue to tail when the priority of a task is
		 * increased (user space view).
		 */
4385 4386
		if (oldprio < p->prio)
			queue_flags |= ENQUEUE_HEAD;
4387

4388
		enqueue_task(rq, p, queue_flags);
4389
	}
4390
	if (running)
4391
		set_curr_task(rq, p);
4392

P
Peter Zijlstra 已提交
4393
	check_class_changed(rq, p, prev_class, oldprio);
I
Ingo Molnar 已提交
4394 4395 4396

	/* Avoid rq from going away on us: */
	preempt_disable();
4397
	task_rq_unlock(rq, p, &rf);
4398

4399 4400
	if (pi)
		rt_mutex_adjust_pi(p);
4401

I
Ingo Molnar 已提交
4402
	/* Run balance callbacks after we've adjusted the PI chain: */
4403 4404
	balance_callback(rq);
	preempt_enable();
4405

L
Linus Torvalds 已提交
4406 4407
	return 0;
}
4408

4409 4410 4411 4412 4413 4414 4415 4416 4417
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),
	};

4418 4419
	/* Fixup the legacy SCHED_RESET_ON_FORK hack. */
	if ((policy != SETPARAM_POLICY) && (policy & SCHED_RESET_ON_FORK)) {
4420 4421 4422 4423 4424
		attr.sched_flags |= SCHED_FLAG_RESET_ON_FORK;
		policy &= ~SCHED_RESET_ON_FORK;
		attr.sched_policy = policy;
	}

4425
	return __sched_setscheduler(p, &attr, check, true);
4426
}
4427 4428 4429 4430 4431 4432
/**
 * 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.
 *
4433 4434
 * Return: 0 on success. An error code otherwise.
 *
4435 4436 4437
 * NOTE that the task may be already dead.
 */
int sched_setscheduler(struct task_struct *p, int policy,
4438
		       const struct sched_param *param)
4439
{
4440
	return _sched_setscheduler(p, policy, param, true);
4441
}
L
Linus Torvalds 已提交
4442 4443
EXPORT_SYMBOL_GPL(sched_setscheduler);

4444 4445
int sched_setattr(struct task_struct *p, const struct sched_attr *attr)
{
4446
	return __sched_setscheduler(p, attr, true, true);
4447 4448 4449
}
EXPORT_SYMBOL_GPL(sched_setattr);

4450 4451 4452 4453 4454 4455 4456 4457 4458 4459
/**
 * 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.
4460 4461
 *
 * Return: 0 on success. An error code otherwise.
4462 4463
 */
int sched_setscheduler_nocheck(struct task_struct *p, int policy,
4464
			       const struct sched_param *param)
4465
{
4466
	return _sched_setscheduler(p, policy, param, false);
4467
}
4468
EXPORT_SYMBOL_GPL(sched_setscheduler_nocheck);
4469

I
Ingo Molnar 已提交
4470 4471
static int
do_sched_setscheduler(pid_t pid, int policy, struct sched_param __user *param)
L
Linus Torvalds 已提交
4472 4473 4474
{
	struct sched_param lparam;
	struct task_struct *p;
4475
	int retval;
L
Linus Torvalds 已提交
4476 4477 4478 4479 4480

	if (!param || pid < 0)
		return -EINVAL;
	if (copy_from_user(&lparam, param, sizeof(struct sched_param)))
		return -EFAULT;
4481 4482 4483

	rcu_read_lock();
	retval = -ESRCH;
L
Linus Torvalds 已提交
4484
	p = find_process_by_pid(pid);
4485 4486 4487
	if (p != NULL)
		retval = sched_setscheduler(p, policy, &lparam);
	rcu_read_unlock();
4488

L
Linus Torvalds 已提交
4489 4490 4491
	return retval;
}

4492 4493 4494
/*
 * Mimics kernel/events/core.c perf_copy_attr().
 */
I
Ingo Molnar 已提交
4495
static int sched_copy_attr(struct sched_attr __user *uattr, struct sched_attr *attr)
4496 4497 4498 4499 4500 4501 4502
{
	u32 size;
	int ret;

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

I
Ingo Molnar 已提交
4503
	/* Zero the full structure, so that a short copy will be nice: */
4504 4505 4506 4507 4508 4509
	memset(attr, 0, sizeof(*attr));

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

I
Ingo Molnar 已提交
4510 4511
	/* Bail out on silly large: */
	if (size > PAGE_SIZE)
4512 4513
		goto err_size;

I
Ingo Molnar 已提交
4514 4515
	/* ABI compatibility quirk: */
	if (!size)
4516 4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536 4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547 4548 4549
		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
4550
	 * XXX: Do we want to be lenient like existing syscalls; or do we want
4551 4552
	 * to be strict and return an error on out-of-bounds values?
	 */
4553
	attr->sched_nice = clamp(attr->sched_nice, MIN_NICE, MAX_NICE);
4554

4555
	return 0;
4556 4557 4558

err_size:
	put_user(sizeof(*attr), &uattr->size);
4559
	return -E2BIG;
4560 4561
}

L
Linus Torvalds 已提交
4562 4563 4564 4565 4566
/**
 * 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.
4567 4568
 *
 * Return: 0 on success. An error code otherwise.
L
Linus Torvalds 已提交
4569
 */
I
Ingo Molnar 已提交
4570
SYSCALL_DEFINE3(sched_setscheduler, pid_t, pid, int, policy, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4571
{
4572 4573 4574
	if (policy < 0)
		return -EINVAL;

L
Linus Torvalds 已提交
4575 4576 4577 4578 4579 4580 4581
	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.
4582 4583
 *
 * Return: 0 on success. An error code otherwise.
L
Linus Torvalds 已提交
4584
 */
4585
SYSCALL_DEFINE2(sched_setparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4586
{
4587
	return do_sched_setscheduler(pid, SETPARAM_POLICY, param);
L
Linus Torvalds 已提交
4588 4589
}

4590 4591 4592
/**
 * sys_sched_setattr - same as above, but with extended sched_attr
 * @pid: the pid in question.
J
Juri Lelli 已提交
4593
 * @uattr: structure containing the extended parameters.
4594
 * @flags: for future extension.
4595
 */
4596 4597
SYSCALL_DEFINE3(sched_setattr, pid_t, pid, struct sched_attr __user *, uattr,
			       unsigned int, flags)
4598 4599 4600 4601 4602
{
	struct sched_attr attr;
	struct task_struct *p;
	int retval;

4603
	if (!uattr || pid < 0 || flags)
4604 4605
		return -EINVAL;

4606 4607 4608
	retval = sched_copy_attr(uattr, &attr);
	if (retval)
		return retval;
4609

4610
	if ((int)attr.sched_policy < 0)
4611
		return -EINVAL;
4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622

	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
4623 4624 4625
/**
 * sys_sched_getscheduler - get the policy (scheduling class) of a thread
 * @pid: the pid in question.
4626 4627 4628
 *
 * Return: On success, the policy of the thread. Otherwise, a negative error
 * code.
L
Linus Torvalds 已提交
4629
 */
4630
SYSCALL_DEFINE1(sched_getscheduler, pid_t, pid)
L
Linus Torvalds 已提交
4631
{
4632
	struct task_struct *p;
4633
	int retval;
L
Linus Torvalds 已提交
4634 4635

	if (pid < 0)
4636
		return -EINVAL;
L
Linus Torvalds 已提交
4637 4638

	retval = -ESRCH;
4639
	rcu_read_lock();
L
Linus Torvalds 已提交
4640 4641 4642 4643
	p = find_process_by_pid(pid);
	if (p) {
		retval = security_task_getscheduler(p);
		if (!retval)
4644 4645
			retval = p->policy
				| (p->sched_reset_on_fork ? SCHED_RESET_ON_FORK : 0);
L
Linus Torvalds 已提交
4646
	}
4647
	rcu_read_unlock();
L
Linus Torvalds 已提交
4648 4649 4650 4651
	return retval;
}

/**
4652
 * sys_sched_getparam - get the RT priority of a thread
L
Linus Torvalds 已提交
4653 4654
 * @pid: the pid in question.
 * @param: structure containing the RT priority.
4655 4656 4657
 *
 * Return: On success, 0 and the RT priority is in @param. Otherwise, an error
 * code.
L
Linus Torvalds 已提交
4658
 */
4659
SYSCALL_DEFINE2(sched_getparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4660
{
4661
	struct sched_param lp = { .sched_priority = 0 };
4662
	struct task_struct *p;
4663
	int retval;
L
Linus Torvalds 已提交
4664 4665

	if (!param || pid < 0)
4666
		return -EINVAL;
L
Linus Torvalds 已提交
4667

4668
	rcu_read_lock();
L
Linus Torvalds 已提交
4669 4670 4671 4672 4673 4674 4675 4676 4677
	p = find_process_by_pid(pid);
	retval = -ESRCH;
	if (!p)
		goto out_unlock;

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

4678 4679
	if (task_has_rt_policy(p))
		lp.sched_priority = p->rt_priority;
4680
	rcu_read_unlock();
L
Linus Torvalds 已提交
4681 4682 4683 4684 4685 4686 4687 4688 4689

	/*
	 * 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:
4690
	rcu_read_unlock();
L
Linus Torvalds 已提交
4691 4692 4693
	return retval;
}

4694 4695 4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715 4716
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)
4717
				return -EFBIG;
4718 4719 4720 4721 4722
		}

		attr->size = usize;
	}

4723
	ret = copy_to_user(uattr, attr, attr->size);
4724 4725 4726
	if (ret)
		return -EFAULT;

4727
	return 0;
4728 4729 4730
}

/**
4731
 * sys_sched_getattr - similar to sched_getparam, but with sched_attr
4732
 * @pid: the pid in question.
J
Juri Lelli 已提交
4733
 * @uattr: structure containing the extended parameters.
4734
 * @size: sizeof(attr) for fwd/bwd comp.
4735
 * @flags: for future extension.
4736
 */
4737 4738
SYSCALL_DEFINE4(sched_getattr, pid_t, pid, struct sched_attr __user *, uattr,
		unsigned int, size, unsigned int, flags)
4739 4740 4741 4742 4743 4744 4745 4746
{
	struct sched_attr attr = {
		.size = sizeof(struct sched_attr),
	};
	struct task_struct *p;
	int retval;

	if (!uattr || pid < 0 || size > PAGE_SIZE ||
4747
	    size < SCHED_ATTR_SIZE_VER0 || flags)
4748 4749 4750 4751 4752 4753 4754 4755 4756 4757 4758 4759 4760
		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;
4761 4762
	if (p->sched_reset_on_fork)
		attr.sched_flags |= SCHED_FLAG_RESET_ON_FORK;
4763 4764 4765
	if (task_has_dl_policy(p))
		__getparam_dl(p, &attr);
	else if (task_has_rt_policy(p))
4766 4767
		attr.sched_priority = p->rt_priority;
	else
4768
		attr.sched_nice = task_nice(p);
4769 4770 4771 4772 4773 4774 4775 4776 4777 4778 4779

	rcu_read_unlock();

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

out_unlock:
	rcu_read_unlock();
	return retval;
}

4780
long sched_setaffinity(pid_t pid, const struct cpumask *in_mask)
L
Linus Torvalds 已提交
4781
{
4782
	cpumask_var_t cpus_allowed, new_mask;
4783 4784
	struct task_struct *p;
	int retval;
L
Linus Torvalds 已提交
4785

4786
	rcu_read_lock();
L
Linus Torvalds 已提交
4787 4788 4789

	p = find_process_by_pid(pid);
	if (!p) {
4790
		rcu_read_unlock();
L
Linus Torvalds 已提交
4791 4792 4793
		return -ESRCH;
	}

4794
	/* Prevent p going away */
L
Linus Torvalds 已提交
4795
	get_task_struct(p);
4796
	rcu_read_unlock();
L
Linus Torvalds 已提交
4797

4798 4799 4800 4801
	if (p->flags & PF_NO_SETAFFINITY) {
		retval = -EINVAL;
		goto out_put_task;
	}
4802 4803 4804 4805 4806 4807 4808 4809
	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
4810
	retval = -EPERM;
E
Eric W. Biederman 已提交
4811 4812 4813 4814
	if (!check_same_owner(p)) {
		rcu_read_lock();
		if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
			rcu_read_unlock();
4815
			goto out_free_new_mask;
E
Eric W. Biederman 已提交
4816 4817 4818
		}
		rcu_read_unlock();
	}
L
Linus Torvalds 已提交
4819

4820
	retval = security_task_setscheduler(p);
4821
	if (retval)
4822
		goto out_free_new_mask;
4823

4824 4825 4826 4827

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

4828 4829 4830 4831 4832 4833 4834
	/*
	 * 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
4835 4836 4837
	if (task_has_dl_policy(p) && dl_bandwidth_enabled()) {
		rcu_read_lock();
		if (!cpumask_subset(task_rq(p)->rd->span, new_mask)) {
4838
			retval = -EBUSY;
4839
			rcu_read_unlock();
4840
			goto out_free_new_mask;
4841
		}
4842
		rcu_read_unlock();
4843 4844
	}
#endif
P
Peter Zijlstra 已提交
4845
again:
4846
	retval = __set_cpus_allowed_ptr(p, new_mask, true);
L
Linus Torvalds 已提交
4847

P
Paul Menage 已提交
4848
	if (!retval) {
4849 4850
		cpuset_cpus_allowed(p, cpus_allowed);
		if (!cpumask_subset(new_mask, cpus_allowed)) {
P
Paul Menage 已提交
4851 4852 4853 4854 4855
			/*
			 * We must have raced with a concurrent cpuset
			 * update. Just reset the cpus_allowed to the
			 * cpuset's cpus_allowed
			 */
4856
			cpumask_copy(new_mask, cpus_allowed);
P
Paul Menage 已提交
4857 4858 4859
			goto again;
		}
	}
4860
out_free_new_mask:
4861 4862 4863 4864
	free_cpumask_var(new_mask);
out_free_cpus_allowed:
	free_cpumask_var(cpus_allowed);
out_put_task:
L
Linus Torvalds 已提交
4865 4866 4867 4868 4869
	put_task_struct(p);
	return retval;
}

static int get_user_cpu_mask(unsigned long __user *user_mask_ptr, unsigned len,
4870
			     struct cpumask *new_mask)
L
Linus Torvalds 已提交
4871
{
4872 4873 4874 4875 4876
	if (len < cpumask_size())
		cpumask_clear(new_mask);
	else if (len > cpumask_size())
		len = cpumask_size();

L
Linus Torvalds 已提交
4877 4878 4879 4880
	return copy_from_user(new_mask, user_mask_ptr, len) ? -EFAULT : 0;
}

/**
I
Ingo Molnar 已提交
4881
 * sys_sched_setaffinity - set the CPU affinity of a process
L
Linus Torvalds 已提交
4882 4883
 * @pid: pid of the process
 * @len: length in bytes of the bitmask pointed to by user_mask_ptr
I
Ingo Molnar 已提交
4884
 * @user_mask_ptr: user-space pointer to the new CPU mask
4885 4886
 *
 * Return: 0 on success. An error code otherwise.
L
Linus Torvalds 已提交
4887
 */
4888 4889
SYSCALL_DEFINE3(sched_setaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
4890
{
4891
	cpumask_var_t new_mask;
L
Linus Torvalds 已提交
4892 4893
	int retval;

4894 4895
	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
4896

4897 4898 4899 4900 4901
	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
4902 4903
}

4904
long sched_getaffinity(pid_t pid, struct cpumask *mask)
L
Linus Torvalds 已提交
4905
{
4906
	struct task_struct *p;
4907
	unsigned long flags;
L
Linus Torvalds 已提交
4908 4909
	int retval;

4910
	rcu_read_lock();
L
Linus Torvalds 已提交
4911 4912 4913 4914 4915 4916

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

4917 4918 4919 4920
	retval = security_task_getscheduler(p);
	if (retval)
		goto out_unlock;

4921
	raw_spin_lock_irqsave(&p->pi_lock, flags);
4922
	cpumask_and(mask, &p->cpus_allowed, cpu_active_mask);
4923
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
4924 4925

out_unlock:
4926
	rcu_read_unlock();
L
Linus Torvalds 已提交
4927

4928
	return retval;
L
Linus Torvalds 已提交
4929 4930 4931
}

/**
I
Ingo Molnar 已提交
4932
 * sys_sched_getaffinity - get the CPU affinity of a process
L
Linus Torvalds 已提交
4933 4934
 * @pid: pid of the process
 * @len: length in bytes of the bitmask pointed to by user_mask_ptr
I
Ingo Molnar 已提交
4935
 * @user_mask_ptr: user-space pointer to hold the current CPU mask
4936
 *
4937 4938
 * Return: size of CPU mask copied to user_mask_ptr on success. An
 * error code otherwise.
L
Linus Torvalds 已提交
4939
 */
4940 4941
SYSCALL_DEFINE3(sched_getaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
4942 4943
{
	int ret;
4944
	cpumask_var_t mask;
L
Linus Torvalds 已提交
4945

A
Anton Blanchard 已提交
4946
	if ((len * BITS_PER_BYTE) < nr_cpu_ids)
4947 4948
		return -EINVAL;
	if (len & (sizeof(unsigned long)-1))
L
Linus Torvalds 已提交
4949 4950
		return -EINVAL;

4951 4952
	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
4953

4954 4955
	ret = sched_getaffinity(pid, mask);
	if (ret == 0) {
4956
		size_t retlen = min_t(size_t, len, cpumask_size());
4957 4958

		if (copy_to_user(user_mask_ptr, mask, retlen))
4959 4960
			ret = -EFAULT;
		else
4961
			ret = retlen;
4962 4963
	}
	free_cpumask_var(mask);
L
Linus Torvalds 已提交
4964

4965
	return ret;
L
Linus Torvalds 已提交
4966 4967 4968 4969 4970
}

/**
 * sys_sched_yield - yield the current processor to other threads.
 *
I
Ingo Molnar 已提交
4971 4972
 * This function yields the current CPU to other tasks. If there are no
 * other threads running on this CPU then this function will return.
4973 4974
 *
 * Return: 0.
L
Linus Torvalds 已提交
4975
 */
4976
SYSCALL_DEFINE0(sched_yield)
L
Linus Torvalds 已提交
4977
{
4978 4979 4980 4981 4982 4983
	struct rq_flags rf;
	struct rq *rq;

	local_irq_disable();
	rq = this_rq();
	rq_lock(rq, &rf);
L
Linus Torvalds 已提交
4984

4985
	schedstat_inc(rq->yld_count);
4986
	current->sched_class->yield_task(rq);
L
Linus Torvalds 已提交
4987 4988 4989 4990 4991

	/*
	 * Since we are going to call schedule() anyway, there's
	 * no need to preempt or enable interrupts:
	 */
4992 4993
	preempt_disable();
	rq_unlock(rq, &rf);
4994
	sched_preempt_enable_no_resched();
L
Linus Torvalds 已提交
4995 4996 4997 4998 4999 5000

	schedule();

	return 0;
}

5001
#ifndef CONFIG_PREEMPT
5002
int __sched _cond_resched(void)
L
Linus Torvalds 已提交
5003
{
5004
	if (should_resched(0)) {
5005
		preempt_schedule_common();
L
Linus Torvalds 已提交
5006 5007 5008 5009
		return 1;
	}
	return 0;
}
5010
EXPORT_SYMBOL(_cond_resched);
5011
#endif
L
Linus Torvalds 已提交
5012 5013

/*
5014
 * __cond_resched_lock() - if a reschedule is pending, drop the given lock,
L
Linus Torvalds 已提交
5015 5016
 * call schedule, and on return reacquire the lock.
 *
I
Ingo Molnar 已提交
5017
 * This works OK both with and without CONFIG_PREEMPT. We do strange low-level
L
Linus Torvalds 已提交
5018 5019 5020
 * operations here to prevent schedule() from being called twice (once via
 * spin_unlock(), once by hand).
 */
5021
int __cond_resched_lock(spinlock_t *lock)
L
Linus Torvalds 已提交
5022
{
5023
	int resched = should_resched(PREEMPT_LOCK_OFFSET);
J
Jan Kara 已提交
5024 5025
	int ret = 0;

5026 5027
	lockdep_assert_held(lock);

5028
	if (spin_needbreak(lock) || resched) {
L
Linus Torvalds 已提交
5029
		spin_unlock(lock);
P
Peter Zijlstra 已提交
5030
		if (resched)
5031
			preempt_schedule_common();
N
Nick Piggin 已提交
5032 5033
		else
			cpu_relax();
J
Jan Kara 已提交
5034
		ret = 1;
L
Linus Torvalds 已提交
5035 5036
		spin_lock(lock);
	}
J
Jan Kara 已提交
5037
	return ret;
L
Linus Torvalds 已提交
5038
}
5039
EXPORT_SYMBOL(__cond_resched_lock);
L
Linus Torvalds 已提交
5040

5041
int __sched __cond_resched_softirq(void)
L
Linus Torvalds 已提交
5042 5043 5044
{
	BUG_ON(!in_softirq());

5045
	if (should_resched(SOFTIRQ_DISABLE_OFFSET)) {
5046
		local_bh_enable();
5047
		preempt_schedule_common();
L
Linus Torvalds 已提交
5048 5049 5050 5051 5052
		local_bh_disable();
		return 1;
	}
	return 0;
}
5053
EXPORT_SYMBOL(__cond_resched_softirq);
L
Linus Torvalds 已提交
5054 5055 5056 5057

/**
 * yield - yield the current processor to other threads.
 *
P
Peter Zijlstra 已提交
5058 5059 5060 5061 5062 5063 5064 5065 5066
 * 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
5067
 *	yield();
P
Peter Zijlstra 已提交
5068 5069 5070 5071 5072 5073 5074 5075
 *
 * 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
5076 5077 5078 5079 5080 5081 5082 5083
 */
void __sched yield(void)
{
	set_current_state(TASK_RUNNING);
	sys_sched_yield();
}
EXPORT_SYMBOL(yield);

5084 5085 5086 5087
/**
 * 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
5088 5089
 * @p: target task
 * @preempt: whether task preemption is allowed or not
5090 5091 5092 5093
 *
 * 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.
 *
5094
 * Return:
5095 5096 5097
 *	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.
5098
 */
5099
int __sched yield_to(struct task_struct *p, bool preempt)
5100 5101 5102 5103
{
	struct task_struct *curr = current;
	struct rq *rq, *p_rq;
	unsigned long flags;
5104
	int yielded = 0;
5105 5106 5107 5108 5109 5110

	local_irq_save(flags);
	rq = this_rq();

again:
	p_rq = task_rq(p);
5111 5112 5113 5114 5115 5116 5117 5118 5119
	/*
	 * 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;
	}

5120
	double_rq_lock(rq, p_rq);
5121
	if (task_rq(p) != p_rq) {
5122 5123 5124 5125 5126
		double_rq_unlock(rq, p_rq);
		goto again;
	}

	if (!curr->sched_class->yield_to_task)
5127
		goto out_unlock;
5128 5129

	if (curr->sched_class != p->sched_class)
5130
		goto out_unlock;
5131 5132

	if (task_running(p_rq, p) || p->state)
5133
		goto out_unlock;
5134 5135

	yielded = curr->sched_class->yield_to_task(rq, p, preempt);
5136
	if (yielded) {
5137
		schedstat_inc(rq->yld_count);
5138 5139 5140 5141 5142
		/*
		 * Make p's CPU reschedule; pick_next_entity takes care of
		 * fairness.
		 */
		if (preempt && rq != p_rq)
5143
			resched_curr(p_rq);
5144
	}
5145

5146
out_unlock:
5147
	double_rq_unlock(rq, p_rq);
5148
out_irq:
5149 5150
	local_irq_restore(flags);

5151
	if (yielded > 0)
5152 5153 5154 5155 5156 5157
		schedule();

	return yielded;
}
EXPORT_SYMBOL_GPL(yield_to);

5158 5159 5160 5161 5162 5163 5164 5165 5166 5167 5168 5169 5170 5171 5172
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
5173
/*
I
Ingo Molnar 已提交
5174
 * This task is about to go to sleep on IO. Increment rq->nr_iowait so
L
Linus Torvalds 已提交
5175 5176 5177 5178
 * that process accounting knows that this is a task in IO wait state.
 */
long __sched io_schedule_timeout(long timeout)
{
5179
	int token;
L
Linus Torvalds 已提交
5180 5181
	long ret;

5182
	token = io_schedule_prepare();
L
Linus Torvalds 已提交
5183
	ret = schedule_timeout(timeout);
5184
	io_schedule_finish(token);
5185

L
Linus Torvalds 已提交
5186 5187
	return ret;
}
5188
EXPORT_SYMBOL(io_schedule_timeout);
L
Linus Torvalds 已提交
5189

5190 5191 5192 5193 5194 5195 5196 5197 5198 5199
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
5200 5201 5202 5203
/**
 * sys_sched_get_priority_max - return maximum RT priority.
 * @policy: scheduling class.
 *
5204 5205 5206
 * 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
5207
 */
5208
SYSCALL_DEFINE1(sched_get_priority_max, int, policy)
L
Linus Torvalds 已提交
5209 5210 5211 5212 5213 5214 5215 5216
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = MAX_USER_RT_PRIO-1;
		break;
5217
	case SCHED_DEADLINE:
L
Linus Torvalds 已提交
5218
	case SCHED_NORMAL:
5219
	case SCHED_BATCH:
I
Ingo Molnar 已提交
5220
	case SCHED_IDLE:
L
Linus Torvalds 已提交
5221 5222 5223 5224 5225 5226 5227 5228 5229 5230
		ret = 0;
		break;
	}
	return ret;
}

/**
 * sys_sched_get_priority_min - return minimum RT priority.
 * @policy: scheduling class.
 *
5231 5232 5233
 * 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
5234
 */
5235
SYSCALL_DEFINE1(sched_get_priority_min, int, policy)
L
Linus Torvalds 已提交
5236 5237 5238 5239 5240 5241 5242 5243
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = 1;
		break;
5244
	case SCHED_DEADLINE:
L
Linus Torvalds 已提交
5245
	case SCHED_NORMAL:
5246
	case SCHED_BATCH:
I
Ingo Molnar 已提交
5247
	case SCHED_IDLE:
L
Linus Torvalds 已提交
5248 5249 5250 5251 5252 5253 5254 5255 5256 5257 5258 5259
		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.
5260 5261 5262
 *
 * Return: On success, 0 and the timeslice is in @interval. Otherwise,
 * an error code.
L
Linus Torvalds 已提交
5263
 */
5264
SYSCALL_DEFINE2(sched_rr_get_interval, pid_t, pid,
5265
		struct timespec __user *, interval)
L
Linus Torvalds 已提交
5266
{
5267
	struct task_struct *p;
D
Dmitry Adamushko 已提交
5268
	unsigned int time_slice;
5269 5270
	struct rq_flags rf;
	struct timespec t;
5271
	struct rq *rq;
5272
	int retval;
L
Linus Torvalds 已提交
5273 5274

	if (pid < 0)
5275
		return -EINVAL;
L
Linus Torvalds 已提交
5276 5277

	retval = -ESRCH;
5278
	rcu_read_lock();
L
Linus Torvalds 已提交
5279 5280 5281 5282 5283 5284 5285 5286
	p = find_process_by_pid(pid);
	if (!p)
		goto out_unlock;

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

5287
	rq = task_rq_lock(p, &rf);
5288 5289 5290
	time_slice = 0;
	if (p->sched_class->get_rr_interval)
		time_slice = p->sched_class->get_rr_interval(rq, p);
5291
	task_rq_unlock(rq, p, &rf);
D
Dmitry Adamushko 已提交
5292

5293
	rcu_read_unlock();
D
Dmitry Adamushko 已提交
5294
	jiffies_to_timespec(time_slice, &t);
L
Linus Torvalds 已提交
5295 5296
	retval = copy_to_user(interval, &t, sizeof(t)) ? -EFAULT : 0;
	return retval;
5297

L
Linus Torvalds 已提交
5298
out_unlock:
5299
	rcu_read_unlock();
L
Linus Torvalds 已提交
5300 5301 5302
	return retval;
}

5303
static const char stat_nam[] = TASK_STATE_TO_CHAR_STR;
5304

5305
void sched_show_task(struct task_struct *p)
L
Linus Torvalds 已提交
5306 5307
{
	unsigned long free = 0;
5308
	int ppid;
5309
	unsigned long state = p->state;
L
Linus Torvalds 已提交
5310

5311 5312 5313
	/* Make sure the string lines up properly with the number of task states: */
	BUILD_BUG_ON(sizeof(TASK_STATE_TO_CHAR_STR)-1 != ilog2(TASK_STATE_MAX)+1);

5314 5315
	if (!try_get_task_stack(p))
		return;
5316 5317
	if (state)
		state = __ffs(state) + 1;
5318
	printk(KERN_INFO "%-15.15s %c", p->comm,
5319
		state < sizeof(stat_nam) - 1 ? stat_nam[state] : '?');
L
Linus Torvalds 已提交
5320
	if (state == TASK_RUNNING)
P
Peter Zijlstra 已提交
5321
		printk(KERN_CONT "  running task    ");
L
Linus Torvalds 已提交
5322
#ifdef CONFIG_DEBUG_STACK_USAGE
5323
	free = stack_not_used(p);
L
Linus Torvalds 已提交
5324
#endif
5325
	ppid = 0;
5326
	rcu_read_lock();
5327 5328
	if (pid_alive(p))
		ppid = task_pid_nr(rcu_dereference(p->real_parent));
5329
	rcu_read_unlock();
P
Peter Zijlstra 已提交
5330
	printk(KERN_CONT "%5lu %5d %6d 0x%08lx\n", free,
5331
		task_pid_nr(p), ppid,
5332
		(unsigned long)task_thread_info(p)->flags);
L
Linus Torvalds 已提交
5333

5334
	print_worker_info(KERN_INFO, p);
5335
	show_stack(p, NULL);
5336
	put_task_stack(p);
L
Linus Torvalds 已提交
5337 5338
}

I
Ingo Molnar 已提交
5339
void show_state_filter(unsigned long state_filter)
L
Linus Torvalds 已提交
5340
{
5341
	struct task_struct *g, *p;
L
Linus Torvalds 已提交
5342

5343
#if BITS_PER_LONG == 32
P
Peter Zijlstra 已提交
5344 5345
	printk(KERN_INFO
		"  task                PC stack   pid father\n");
L
Linus Torvalds 已提交
5346
#else
P
Peter Zijlstra 已提交
5347 5348
	printk(KERN_INFO
		"  task                        PC stack   pid father\n");
L
Linus Torvalds 已提交
5349
#endif
5350
	rcu_read_lock();
5351
	for_each_process_thread(g, p) {
L
Linus Torvalds 已提交
5352 5353
		/*
		 * reset the NMI-timeout, listing all files on a slow
L
Lucas De Marchi 已提交
5354
		 * console might take a lot of time:
5355 5356 5357
		 * 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
5358 5359
		 */
		touch_nmi_watchdog();
5360
		touch_all_softlockup_watchdogs();
I
Ingo Molnar 已提交
5361
		if (!state_filter || (p->state & state_filter))
5362
			sched_show_task(p);
5363
	}
L
Linus Torvalds 已提交
5364

I
Ingo Molnar 已提交
5365
#ifdef CONFIG_SCHED_DEBUG
5366 5367
	if (!state_filter)
		sysrq_sched_debug_show();
I
Ingo Molnar 已提交
5368
#endif
5369
	rcu_read_unlock();
I
Ingo Molnar 已提交
5370 5371 5372
	/*
	 * Only show locks if all tasks are dumped:
	 */
5373
	if (!state_filter)
I
Ingo Molnar 已提交
5374
		debug_show_all_locks();
L
Linus Torvalds 已提交
5375 5376
}

5377
void init_idle_bootup_task(struct task_struct *idle)
I
Ingo Molnar 已提交
5378
{
I
Ingo Molnar 已提交
5379
	idle->sched_class = &idle_sched_class;
I
Ingo Molnar 已提交
5380 5381
}

5382 5383 5384
/**
 * init_idle - set up an idle thread for a given CPU
 * @idle: task in question
I
Ingo Molnar 已提交
5385
 * @cpu: CPU the idle task belongs to
5386 5387 5388 5389
 *
 * NOTE: this function does not set the idle thread's NEED_RESCHED
 * flag, to make booting more robust.
 */
5390
void init_idle(struct task_struct *idle, int cpu)
L
Linus Torvalds 已提交
5391
{
5392
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
5393 5394
	unsigned long flags;

5395 5396
	raw_spin_lock_irqsave(&idle->pi_lock, flags);
	raw_spin_lock(&rq->lock);
5397

5398
	__sched_fork(0, idle);
5399
	idle->state = TASK_RUNNING;
I
Ingo Molnar 已提交
5400
	idle->se.exec_start = sched_clock();
5401
	idle->flags |= PF_IDLE;
I
Ingo Molnar 已提交
5402

5403 5404
	kasan_unpoison_task_stack(idle);

5405 5406 5407 5408 5409 5410 5411 5412 5413
#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
5414 5415
	/*
	 * We're having a chicken and egg problem, even though we are
I
Ingo Molnar 已提交
5416
	 * holding rq->lock, the CPU isn't yet set to this CPU so the
5417 5418 5419 5420 5421 5422 5423 5424
	 * 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
5425
	__set_task_cpu(idle, cpu);
5426
	rcu_read_unlock();
L
Linus Torvalds 已提交
5427 5428

	rq->curr = rq->idle = idle;
5429
	idle->on_rq = TASK_ON_RQ_QUEUED;
5430
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
5431
	idle->on_cpu = 1;
5432
#endif
5433 5434
	raw_spin_unlock(&rq->lock);
	raw_spin_unlock_irqrestore(&idle->pi_lock, flags);
L
Linus Torvalds 已提交
5435 5436

	/* Set the preempt count _outside_ the spinlocks! */
5437
	init_idle_preempt_count(idle, cpu);
5438

I
Ingo Molnar 已提交
5439 5440 5441 5442
	/*
	 * The idle tasks have their own, simple scheduling class:
	 */
	idle->sched_class = &idle_sched_class;
5443
	ftrace_graph_init_idle_task(idle, cpu);
5444
	vtime_init_idle(idle, cpu);
5445
#ifdef CONFIG_SMP
5446 5447
	sprintf(idle->comm, "%s/%d", INIT_TASK_COMM, cpu);
#endif
I
Ingo Molnar 已提交
5448 5449
}

5450 5451 5452 5453 5454 5455 5456
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;

5457 5458 5459
	if (!cpumask_weight(cur))
		return ret;

5460
	rcu_read_lock_sched();
5461 5462 5463 5464 5465 5466 5467 5468
	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);
5469
	rcu_read_unlock_sched();
5470 5471 5472 5473

	return ret;
}

5474 5475 5476 5477 5478 5479 5480
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
5481
	 * to a new cpuset; we don't want to change their CPU
5482 5483 5484 5485 5486 5487 5488 5489 5490 5491 5492 5493 5494 5495 5496 5497
	 * 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);
5498
		struct dl_bw *dl_b;
5499 5500 5501 5502
		bool overflow;
		int cpus;
		unsigned long flags;

5503 5504
		rcu_read_lock_sched();
		dl_b = dl_bw_of(dest_cpu);
5505 5506 5507 5508 5509 5510 5511 5512 5513 5514 5515 5516 5517 5518 5519
		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);
5520
		rcu_read_unlock_sched();
5521 5522 5523 5524 5525 5526 5527

	}
#endif
out:
	return ret;
}

L
Linus Torvalds 已提交
5528 5529
#ifdef CONFIG_SMP

5530
bool sched_smp_initialized __read_mostly;
5531

5532 5533 5534 5535 5536 5537 5538 5539 5540 5541
#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;

5542
	if (!cpumask_test_cpu(target_cpu, &p->cpus_allowed))
5543 5544 5545 5546
		return -EINVAL;

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

5547
	trace_sched_move_numa(p, curr_cpu, target_cpu);
5548 5549
	return stop_one_cpu(curr_cpu, migration_cpu_stop, &arg);
}
5550 5551 5552 5553 5554 5555 5556

/*
 * 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)
{
5557
	bool queued, running;
5558 5559
	struct rq_flags rf;
	struct rq *rq;
5560

5561
	rq = task_rq_lock(p, &rf);
5562
	queued = task_on_rq_queued(p);
5563 5564
	running = task_current(rq, p);

5565
	if (queued)
5566
		dequeue_task(rq, p, DEQUEUE_SAVE);
5567
	if (running)
5568
		put_prev_task(rq, p);
5569 5570 5571

	p->numa_preferred_nid = nid;

5572
	if (queued)
5573
		enqueue_task(rq, p, ENQUEUE_RESTORE | ENQUEUE_NOCLOCK);
5574
	if (running)
5575
		set_curr_task(rq, p);
5576
	task_rq_unlock(rq, p, &rf);
5577
}
P
Peter Zijlstra 已提交
5578
#endif /* CONFIG_NUMA_BALANCING */
5579

L
Linus Torvalds 已提交
5580
#ifdef CONFIG_HOTPLUG_CPU
5581
/*
I
Ingo Molnar 已提交
5582
 * Ensure that the idle task is using init_mm right before its CPU goes
5583
 * offline.
5584
 */
5585
void idle_task_exit(void)
L
Linus Torvalds 已提交
5586
{
5587
	struct mm_struct *mm = current->active_mm;
5588

5589
	BUG_ON(cpu_online(smp_processor_id()));
5590

5591
	if (mm != &init_mm) {
5592
		switch_mm_irqs_off(mm, &init_mm, current);
5593 5594
		finish_arch_post_lock_switch();
	}
5595
	mmdrop(mm);
L
Linus Torvalds 已提交
5596 5597 5598
}

/*
5599 5600
 * 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
5601 5602 5603
 * 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.
5604 5605
 *
 * Also see the comment "Global load-average calculations".
L
Linus Torvalds 已提交
5606
 */
5607
static void calc_load_migrate(struct rq *rq)
L
Linus Torvalds 已提交
5608
{
5609
	long delta = calc_load_fold_active(rq, 1);
5610 5611
	if (delta)
		atomic_long_add(delta, &calc_load_tasks);
L
Linus Torvalds 已提交
5612 5613
}

5614 5615 5616 5617 5618 5619 5620 5621 5622 5623 5624 5625 5626 5627 5628 5629
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,
};

5630
/*
5631 5632 5633 5634 5635 5636
 * 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
5637
 */
5638
static void migrate_tasks(struct rq *dead_rq, struct rq_flags *rf)
L
Linus Torvalds 已提交
5639
{
5640
	struct rq *rq = dead_rq;
5641
	struct task_struct *next, *stop = rq->stop;
5642
	struct rq_flags orf = *rf;
5643
	int dest_cpu;
L
Linus Torvalds 已提交
5644 5645

	/*
5646 5647 5648 5649 5650 5651 5652
	 * 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
5653
	 */
5654
	rq->stop = NULL;
5655

5656 5657 5658 5659 5660 5661 5662
	/*
	 * put_prev_task() and pick_next_task() sched
	 * class method both need to have an up-to-date
	 * value of rq->clock[_task]
	 */
	update_rq_clock(rq);

5663
	for (;;) {
5664 5665
		/*
		 * There's this thread running, bail when that's the only
I
Ingo Molnar 已提交
5666
		 * remaining thread:
5667 5668
		 */
		if (rq->nr_running == 1)
I
Ingo Molnar 已提交
5669
			break;
5670

5671
		/*
I
Ingo Molnar 已提交
5672
		 * pick_next_task() assumes pinned rq->lock:
5673
		 */
5674
		next = pick_next_task(rq, &fake_task, rf);
5675
		BUG_ON(!next);
D
Dmitry Adamushko 已提交
5676
		next->sched_class->put_prev_task(rq, next);
5677

W
Wanpeng Li 已提交
5678 5679 5680 5681 5682 5683 5684 5685 5686
		/*
		 * 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.
		 */
5687
		rq_unlock(rq, rf);
W
Wanpeng Li 已提交
5688
		raw_spin_lock(&next->pi_lock);
5689
		rq_relock(rq, rf);
W
Wanpeng Li 已提交
5690 5691 5692 5693 5694 5695 5696 5697 5698 5699 5700

		/*
		 * 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;
		}

5701
		/* Find suitable destination for @next, with force if needed. */
5702
		dest_cpu = select_fallback_rq(dead_rq->cpu, next);
5703
		rq = __migrate_task(rq, rf, next, dest_cpu);
5704
		if (rq != dead_rq) {
5705
			rq_unlock(rq, rf);
5706
			rq = dead_rq;
5707 5708
			*rf = orf;
			rq_relock(rq, rf);
5709
		}
W
Wanpeng Li 已提交
5710
		raw_spin_unlock(&next->pi_lock);
L
Linus Torvalds 已提交
5711
	}
5712

5713
	rq->stop = stop;
5714
}
L
Linus Torvalds 已提交
5715 5716
#endif /* CONFIG_HOTPLUG_CPU */

5717
void set_rq_online(struct rq *rq)
5718 5719 5720 5721
{
	if (!rq->online) {
		const struct sched_class *class;

5722
		cpumask_set_cpu(rq->cpu, rq->rd->online);
5723 5724 5725 5726 5727 5728 5729 5730 5731
		rq->online = 1;

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

5732
void set_rq_offline(struct rq *rq)
5733 5734 5735 5736 5737 5738 5739 5740 5741
{
	if (rq->online) {
		const struct sched_class *class;

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

5742
		cpumask_clear_cpu(rq->cpu, rq->rd->online);
5743 5744 5745 5746
		rq->online = 0;
	}
}

5747
static void set_cpu_rq_start_time(unsigned int cpu)
L
Linus Torvalds 已提交
5748
{
5749
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
5750

5751 5752 5753
	rq->age_stamp = sched_clock_cpu(cpu);
}

I
Ingo Molnar 已提交
5754 5755 5756 5757
/*
 * used to mark begin/end of suspend/resume:
 */
static int num_cpus_frozen;
5758

L
Linus Torvalds 已提交
5759
/*
5760 5761 5762
 * Update cpusets according to cpu_active mask.  If cpusets are
 * disabled, cpuset_update_active_cpus() becomes a simple wrapper
 * around partition_sched_domains().
5763 5764 5765
 *
 * 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
5766
 */
5767
static void cpuset_cpu_active(void)
5768
{
5769
	if (cpuhp_tasks_frozen) {
5770 5771 5772 5773 5774 5775 5776 5777 5778
		/*
		 * 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);
5779
			return;
5780 5781 5782 5783 5784 5785
		}
		/*
		 * This is the last CPU online operation. So fall through and
		 * restore the original sched domains by considering the
		 * cpuset configurations.
		 */
5786
	}
5787
	cpuset_update_active_cpus();
5788
}
5789

5790
static int cpuset_cpu_inactive(unsigned int cpu)
5791
{
5792 5793
	unsigned long flags;
	struct dl_bw *dl_b;
5794 5795
	bool overflow;
	int cpus;
5796

5797
	if (!cpuhp_tasks_frozen) {
5798 5799
		rcu_read_lock_sched();
		dl_b = dl_bw_of(cpu);
5800

5801 5802 5803 5804
		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);
5805

5806
		rcu_read_unlock_sched();
5807

5808
		if (overflow)
5809
			return -EBUSY;
5810
		cpuset_update_active_cpus();
5811
	} else {
5812 5813
		num_cpus_frozen++;
		partition_sched_domains(1, NULL, NULL);
5814
	}
5815
	return 0;
5816 5817
}

5818
int sched_cpu_activate(unsigned int cpu)
5819
{
5820
	struct rq *rq = cpu_rq(cpu);
5821
	struct rq_flags rf;
5822

5823
	set_cpu_active(cpu, true);
5824

5825
	if (sched_smp_initialized) {
5826
		sched_domains_numa_masks_set(cpu);
5827
		cpuset_cpu_active();
5828
	}
5829 5830 5831 5832 5833

	/*
	 * 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
5834
	 *    after all CPUs have been brought up.
5835 5836 5837 5838
	 *
	 * 2) At runtime, if cpuset_cpu_active() fails to rebuild the
	 *    domains.
	 */
5839
	rq_lock_irqsave(rq, &rf);
5840 5841 5842 5843
	if (rq->rd) {
		BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
		set_rq_online(rq);
	}
5844
	rq_unlock_irqrestore(rq, &rf);
5845 5846 5847

	update_max_interval();

5848
	return 0;
5849 5850
}

5851
int sched_cpu_deactivate(unsigned int cpu)
5852 5853 5854
{
	int ret;

5855
	set_cpu_active(cpu, false);
5856 5857 5858 5859 5860 5861 5862 5863 5864 5865 5866 5867 5868 5869
	/*
	 * 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();
5870 5871 5872 5873 5874 5875 5876 5877

	if (!sched_smp_initialized)
		return 0;

	ret = cpuset_cpu_inactive(cpu);
	if (ret) {
		set_cpu_active(cpu, true);
		return ret;
5878
	}
5879 5880
	sched_domains_numa_masks_clear(cpu);
	return 0;
5881 5882
}

5883 5884 5885 5886 5887 5888 5889 5890
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();
}

5891 5892 5893
int sched_cpu_starting(unsigned int cpu)
{
	set_cpu_rq_start_time(cpu);
5894
	sched_rq_cpu_starting(cpu);
5895
	return 0;
5896 5897
}

5898 5899 5900 5901
#ifdef CONFIG_HOTPLUG_CPU
int sched_cpu_dying(unsigned int cpu)
{
	struct rq *rq = cpu_rq(cpu);
5902
	struct rq_flags rf;
5903 5904 5905

	/* Handle pending wakeups and then migrate everything off */
	sched_ttwu_pending();
5906 5907

	rq_lock_irqsave(rq, &rf);
5908 5909 5910 5911
	if (rq->rd) {
		BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
		set_rq_offline(rq);
	}
5912
	migrate_tasks(rq, &rf);
5913
	BUG_ON(rq->nr_running != 1);
5914 5915
	rq_unlock_irqrestore(rq, &rf);

5916 5917
	calc_load_migrate(rq);
	update_max_interval();
5918
	nohz_balance_exit_idle(cpu);
5919
	hrtick_clear(rq);
5920 5921 5922 5923
	return 0;
}
#endif

P
Peter Zijlstra 已提交
5924 5925 5926 5927 5928 5929 5930 5931 5932 5933 5934 5935 5936 5937 5938 5939
#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
5940 5941
void __init sched_init_smp(void)
{
5942 5943 5944
	cpumask_var_t non_isolated_cpus;

	alloc_cpumask_var(&non_isolated_cpus, GFP_KERNEL);
5945

5946 5947
	sched_init_numa();

5948 5949
	/*
	 * There's no userspace yet to cause hotplug operations; hence all the
I
Ingo Molnar 已提交
5950
	 * CPU masks are stable and all blatant races in the below code cannot
5951 5952
	 * happen.
	 */
5953
	mutex_lock(&sched_domains_mutex);
P
Peter Zijlstra 已提交
5954
	sched_init_domains(cpu_active_mask);
5955 5956 5957
	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);
5958
	mutex_unlock(&sched_domains_mutex);
5959

5960
	/* Move init over to a non-isolated CPU */
5961
	if (set_cpus_allowed_ptr(current, non_isolated_cpus) < 0)
5962
		BUG();
I
Ingo Molnar 已提交
5963
	sched_init_granularity();
5964
	free_cpumask_var(non_isolated_cpus);
5965

5966
	init_sched_rt_class();
5967
	init_sched_dl_class();
P
Peter Zijlstra 已提交
5968 5969 5970

	sched_init_smt();

5971
	sched_smp_initialized = true;
L
Linus Torvalds 已提交
5972
}
5973 5974 5975

static int __init migration_init(void)
{
5976
	sched_rq_cpu_starting(smp_processor_id());
5977
	return 0;
L
Linus Torvalds 已提交
5978
}
5979 5980
early_initcall(migration_init);

L
Linus Torvalds 已提交
5981 5982 5983
#else
void __init sched_init_smp(void)
{
I
Ingo Molnar 已提交
5984
	sched_init_granularity();
L
Linus Torvalds 已提交
5985 5986 5987 5988 5989 5990 5991 5992 5993 5994
}
#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);
}

5995
#ifdef CONFIG_CGROUP_SCHED
5996 5997 5998 5999
/*
 * Default task group.
 * Every task in system belongs to this group at bootup.
 */
6000
struct task_group root_task_group;
6001
LIST_HEAD(task_groups);
6002 6003 6004

/* Cacheline aligned slab cache for task_group */
static struct kmem_cache *task_group_cache __read_mostly;
6005
#endif
P
Peter Zijlstra 已提交
6006

6007
DECLARE_PER_CPU(cpumask_var_t, load_balance_mask);
6008
DECLARE_PER_CPU(cpumask_var_t, select_idle_mask);
P
Peter Zijlstra 已提交
6009

6010 6011 6012 6013 6014 6015 6016 6017 6018 6019 6020 6021 6022
#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
6023 6024
void __init sched_init(void)
{
I
Ingo Molnar 已提交
6025
	int i, j;
6026 6027
	unsigned long alloc_size = 0, ptr;

6028 6029
	sched_clock_init();

6030 6031 6032
	for (i = 0; i < WAIT_TABLE_SIZE; i++)
		init_waitqueue_head(bit_wait_table + i);

6033 6034 6035 6036 6037 6038 6039
#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) {
6040
		ptr = (unsigned long)kzalloc(alloc_size, GFP_NOWAIT);
6041 6042

#ifdef CONFIG_FAIR_GROUP_SCHED
6043
		root_task_group.se = (struct sched_entity **)ptr;
6044 6045
		ptr += nr_cpu_ids * sizeof(void **);

6046
		root_task_group.cfs_rq = (struct cfs_rq **)ptr;
6047
		ptr += nr_cpu_ids * sizeof(void **);
6048

6049
#endif /* CONFIG_FAIR_GROUP_SCHED */
6050
#ifdef CONFIG_RT_GROUP_SCHED
6051
		root_task_group.rt_se = (struct sched_rt_entity **)ptr;
6052 6053
		ptr += nr_cpu_ids * sizeof(void **);

6054
		root_task_group.rt_rq = (struct rt_rq **)ptr;
6055 6056
		ptr += nr_cpu_ids * sizeof(void **);

6057
#endif /* CONFIG_RT_GROUP_SCHED */
6058
	}
6059
#ifdef CONFIG_CPUMASK_OFFSTACK
6060 6061 6062
	for_each_possible_cpu(i) {
		per_cpu(load_balance_mask, i) = (cpumask_var_t)kzalloc_node(
			cpumask_size(), GFP_KERNEL, cpu_to_node(i));
6063 6064
		per_cpu(select_idle_mask, i) = (cpumask_var_t)kzalloc_node(
			cpumask_size(), GFP_KERNEL, cpu_to_node(i));
6065
	}
6066
#endif /* CONFIG_CPUMASK_OFFSTACK */
I
Ingo Molnar 已提交
6067

I
Ingo Molnar 已提交
6068 6069
	init_rt_bandwidth(&def_rt_bandwidth, global_rt_period(), global_rt_runtime());
	init_dl_bandwidth(&def_dl_bandwidth, global_rt_period(), global_rt_runtime());
6070

G
Gregory Haskins 已提交
6071 6072 6073 6074
#ifdef CONFIG_SMP
	init_defrootdomain();
#endif

6075
#ifdef CONFIG_RT_GROUP_SCHED
6076
	init_rt_bandwidth(&root_task_group.rt_bandwidth,
6077
			global_rt_period(), global_rt_runtime());
6078
#endif /* CONFIG_RT_GROUP_SCHED */
6079

D
Dhaval Giani 已提交
6080
#ifdef CONFIG_CGROUP_SCHED
6081 6082
	task_group_cache = KMEM_CACHE(task_group, 0);

6083 6084
	list_add(&root_task_group.list, &task_groups);
	INIT_LIST_HEAD(&root_task_group.children);
6085
	INIT_LIST_HEAD(&root_task_group.siblings);
6086
	autogroup_init(&init_task);
D
Dhaval Giani 已提交
6087
#endif /* CONFIG_CGROUP_SCHED */
P
Peter Zijlstra 已提交
6088

6089
	for_each_possible_cpu(i) {
6090
		struct rq *rq;
L
Linus Torvalds 已提交
6091 6092

		rq = cpu_rq(i);
6093
		raw_spin_lock_init(&rq->lock);
N
Nick Piggin 已提交
6094
		rq->nr_running = 0;
6095 6096
		rq->calc_load_active = 0;
		rq->calc_load_update = jiffies + LOAD_FREQ;
6097
		init_cfs_rq(&rq->cfs);
6098 6099
		init_rt_rq(&rq->rt);
		init_dl_rq(&rq->dl);
I
Ingo Molnar 已提交
6100
#ifdef CONFIG_FAIR_GROUP_SCHED
6101
		root_task_group.shares = ROOT_TASK_GROUP_LOAD;
P
Peter Zijlstra 已提交
6102
		INIT_LIST_HEAD(&rq->leaf_cfs_rq_list);
6103
		rq->tmp_alone_branch = &rq->leaf_cfs_rq_list;
D
Dhaval Giani 已提交
6104
		/*
I
Ingo Molnar 已提交
6105
		 * How much CPU bandwidth does root_task_group get?
D
Dhaval Giani 已提交
6106 6107
		 *
		 * In case of task-groups formed thr' the cgroup filesystem, it
I
Ingo Molnar 已提交
6108 6109
		 * gets 100% of the CPU resources in the system. This overall
		 * system CPU resource is divided among the tasks of
6110
		 * root_task_group and its child task-groups in a fair manner,
D
Dhaval Giani 已提交
6111 6112 6113
		 * based on each entity's (task or task-group's) weight
		 * (se->load.weight).
		 *
6114
		 * In other words, if root_task_group has 10 tasks of weight
D
Dhaval Giani 已提交
6115
		 * 1024) and two child groups A0 and A1 (of weight 1024 each),
I
Ingo Molnar 已提交
6116
		 * then A0's share of the CPU resource is:
D
Dhaval Giani 已提交
6117
		 *
6118
		 *	A0's bandwidth = 1024 / (10*1024 + 1024 + 1024) = 8.33%
D
Dhaval Giani 已提交
6119
		 *
6120 6121
		 * 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
6122
		 */
6123
		init_cfs_bandwidth(&root_task_group.cfs_bandwidth);
6124
		init_tg_cfs_entry(&root_task_group, &rq->cfs, NULL, i, NULL);
D
Dhaval Giani 已提交
6125 6126 6127
#endif /* CONFIG_FAIR_GROUP_SCHED */

		rq->rt.rt_runtime = def_rt_bandwidth.rt_runtime;
6128
#ifdef CONFIG_RT_GROUP_SCHED
6129
		init_tg_rt_entry(&root_task_group, &rq->rt, NULL, i, NULL);
I
Ingo Molnar 已提交
6130
#endif
L
Linus Torvalds 已提交
6131

I
Ingo Molnar 已提交
6132 6133
		for (j = 0; j < CPU_LOAD_IDX_MAX; j++)
			rq->cpu_load[j] = 0;
6134

L
Linus Torvalds 已提交
6135
#ifdef CONFIG_SMP
N
Nick Piggin 已提交
6136
		rq->sd = NULL;
G
Gregory Haskins 已提交
6137
		rq->rd = NULL;
6138
		rq->cpu_capacity = rq->cpu_capacity_orig = SCHED_CAPACITY_SCALE;
6139
		rq->balance_callback = NULL;
L
Linus Torvalds 已提交
6140
		rq->active_balance = 0;
I
Ingo Molnar 已提交
6141
		rq->next_balance = jiffies;
L
Linus Torvalds 已提交
6142
		rq->push_cpu = 0;
6143
		rq->cpu = i;
6144
		rq->online = 0;
6145 6146
		rq->idle_stamp = 0;
		rq->avg_idle = 2*sysctl_sched_migration_cost;
6147
		rq->max_idle_balance_cost = sysctl_sched_migration_cost;
6148 6149 6150

		INIT_LIST_HEAD(&rq->cfs_tasks);

6151
		rq_attach_root(rq, &def_root_domain);
6152
#ifdef CONFIG_NO_HZ_COMMON
6153
		rq->last_load_update_tick = jiffies;
6154
		rq->nohz_flags = 0;
6155
#endif
6156 6157 6158
#ifdef CONFIG_NO_HZ_FULL
		rq->last_sched_tick = 0;
#endif
6159
#endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
6160
		init_rq_hrtick(rq);
L
Linus Torvalds 已提交
6161 6162 6163
		atomic_set(&rq->nr_iowait, 0);
	}

6164
	set_load_weight(&init_task);
6165

L
Linus Torvalds 已提交
6166 6167 6168
	/*
	 * The boot idle thread does lazy MMU switching as well:
	 */
V
Vegard Nossum 已提交
6169
	mmgrab(&init_mm);
L
Linus Torvalds 已提交
6170 6171 6172 6173 6174 6175 6176 6177 6178
	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());
6179 6180 6181

	calc_load_update = jiffies + LOAD_FREQ;

6182
#ifdef CONFIG_SMP
R
Rusty Russell 已提交
6183 6184 6185
	/* May be allocated at isolcpus cmdline parse time */
	if (cpu_isolated_map == NULL)
		zalloc_cpumask_var(&cpu_isolated_map, GFP_NOWAIT);
6186
	idle_thread_set_boot_cpu();
6187
	set_cpu_rq_start_time(smp_processor_id());
6188 6189
#endif
	init_sched_fair_class();
6190

6191 6192
	init_schedstats();

6193
	scheduler_running = 1;
L
Linus Torvalds 已提交
6194 6195
}

6196
#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
6197 6198
static inline int preempt_count_equals(int preempt_offset)
{
6199
	int nested = preempt_count() + rcu_preempt_depth();
6200

A
Arnd Bergmann 已提交
6201
	return (nested == preempt_offset);
6202 6203
}

6204
void __might_sleep(const char *file, int line, int preempt_offset)
L
Linus Torvalds 已提交
6205
{
P
Peter Zijlstra 已提交
6206 6207 6208 6209 6210
	/*
	 * 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.
	 */
6211
	WARN_ONCE(current->state != TASK_RUNNING && current->task_state_change,
P
Peter Zijlstra 已提交
6212 6213 6214 6215
			"do not call blocking ops when !TASK_RUNNING; "
			"state=%lx set at [<%p>] %pS\n",
			current->state,
			(void *)current->task_state_change,
6216
			(void *)current->task_state_change);
P
Peter Zijlstra 已提交
6217

6218 6219 6220 6221 6222
	___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
6223
{
I
Ingo Molnar 已提交
6224 6225 6226
	/* Ratelimiting timestamp: */
	static unsigned long prev_jiffy;

6227
	unsigned long preempt_disable_ip;
L
Linus Torvalds 已提交
6228

I
Ingo Molnar 已提交
6229 6230 6231
	/* WARN_ON_ONCE() by default, no rate limit required: */
	rcu_sleep_check();

6232 6233
	if ((preempt_count_equals(preempt_offset) && !irqs_disabled() &&
	     !is_idle_task(current)) ||
6234 6235
	    system_state == SYSTEM_BOOTING || system_state > SYSTEM_RUNNING ||
	    oops_in_progress)
I
Ingo Molnar 已提交
6236
		return;
6237

I
Ingo Molnar 已提交
6238 6239 6240 6241
	if (time_before(jiffies, prev_jiffy + HZ) && prev_jiffy)
		return;
	prev_jiffy = jiffies;

I
Ingo Molnar 已提交
6242
	/* Save this before calling printk(), since that will clobber it: */
6243 6244
	preempt_disable_ip = get_preempt_disable_ip(current);

P
Peter Zijlstra 已提交
6245 6246 6247 6248 6249 6250 6251
	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
6252

6253 6254 6255
	if (task_stack_end_corrupted(current))
		printk(KERN_EMERG "Thread overran stack, or stack corrupted\n");

I
Ingo Molnar 已提交
6256 6257 6258
	debug_show_held_locks(current);
	if (irqs_disabled())
		print_irqtrace_events(current);
6259 6260
	if (IS_ENABLED(CONFIG_DEBUG_PREEMPT)
	    && !preempt_count_equals(preempt_offset)) {
6261
		pr_err("Preemption disabled at:");
6262
		print_ip_sym(preempt_disable_ip);
6263 6264
		pr_cont("\n");
	}
I
Ingo Molnar 已提交
6265
	dump_stack();
6266
	add_taint(TAINT_WARN, LOCKDEP_STILL_OK);
L
Linus Torvalds 已提交
6267
}
6268
EXPORT_SYMBOL(___might_sleep);
L
Linus Torvalds 已提交
6269 6270 6271
#endif

#ifdef CONFIG_MAGIC_SYSRQ
6272
void normalize_rt_tasks(void)
6273
{
6274
	struct task_struct *g, *p;
6275 6276 6277
	struct sched_attr attr = {
		.sched_policy = SCHED_NORMAL,
	};
L
Linus Torvalds 已提交
6278

6279
	read_lock(&tasklist_lock);
6280
	for_each_process_thread(g, p) {
6281 6282 6283
		/*
		 * Only normalize user tasks:
		 */
6284
		if (p->flags & PF_KTHREAD)
6285 6286
			continue;

6287 6288 6289 6290
		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
6291

6292
		if (!dl_task(p) && !rt_task(p)) {
I
Ingo Molnar 已提交
6293 6294 6295 6296
			/*
			 * Renice negative nice level userspace
			 * tasks back to 0:
			 */
6297
			if (task_nice(p) < 0)
I
Ingo Molnar 已提交
6298
				set_user_nice(p, 0);
L
Linus Torvalds 已提交
6299
			continue;
I
Ingo Molnar 已提交
6300
		}
L
Linus Torvalds 已提交
6301

6302
		__sched_setscheduler(p, &attr, false, false);
6303
	}
6304
	read_unlock(&tasklist_lock);
L
Linus Torvalds 已提交
6305 6306 6307
}

#endif /* CONFIG_MAGIC_SYSRQ */
6308

6309
#if defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB)
6310
/*
6311
 * These functions are only useful for the IA64 MCA handling, or kdb.
6312 6313 6314 6315 6316 6317 6318 6319 6320
 *
 * 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
6321
 * curr_task - return the current task for a given CPU.
6322 6323 6324
 * @cpu: the processor in question.
 *
 * ONLY VALID WHEN THE WHOLE SYSTEM IS STOPPED!
6325 6326
 *
 * Return: The current task for @cpu.
6327
 */
6328
struct task_struct *curr_task(int cpu)
6329 6330 6331 6332
{
	return cpu_curr(cpu);
}

6333 6334 6335
#endif /* defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB) */

#ifdef CONFIG_IA64
6336
/**
I
Ingo Molnar 已提交
6337
 * set_curr_task - set the current task for a given CPU.
6338 6339 6340 6341
 * @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
6342
 * are serviced on a separate stack. It allows the architecture to switch the
I
Ingo Molnar 已提交
6343
 * notion of the current task on a CPU in a non-blocking manner. This function
6344 6345 6346 6347 6348 6349 6350
 * 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!
 */
6351
void ia64_set_curr_task(int cpu, struct task_struct *p)
6352 6353 6354 6355 6356
{
	cpu_curr(cpu) = p;
}

#endif
S
Srivatsa Vaddagiri 已提交
6357

D
Dhaval Giani 已提交
6358
#ifdef CONFIG_CGROUP_SCHED
6359 6360 6361
/* task_group_lock serializes the addition/removal of task groups */
static DEFINE_SPINLOCK(task_group_lock);

6362
static void sched_free_group(struct task_group *tg)
6363 6364 6365
{
	free_fair_sched_group(tg);
	free_rt_sched_group(tg);
6366
	autogroup_free(tg);
6367
	kmem_cache_free(task_group_cache, tg);
6368 6369 6370
}

/* allocate runqueue etc for a new task group */
6371
struct task_group *sched_create_group(struct task_group *parent)
6372 6373 6374
{
	struct task_group *tg;

6375
	tg = kmem_cache_alloc(task_group_cache, GFP_KERNEL | __GFP_ZERO);
6376 6377 6378
	if (!tg)
		return ERR_PTR(-ENOMEM);

6379
	if (!alloc_fair_sched_group(tg, parent))
6380 6381
		goto err;

6382
	if (!alloc_rt_sched_group(tg, parent))
6383 6384
		goto err;

6385 6386 6387
	return tg;

err:
6388
	sched_free_group(tg);
6389 6390 6391 6392 6393 6394 6395
	return ERR_PTR(-ENOMEM);
}

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

6396
	spin_lock_irqsave(&task_group_lock, flags);
P
Peter Zijlstra 已提交
6397
	list_add_rcu(&tg->list, &task_groups);
P
Peter Zijlstra 已提交
6398

I
Ingo Molnar 已提交
6399 6400
	/* Root should already exist: */
	WARN_ON(!parent);
P
Peter Zijlstra 已提交
6401 6402 6403

	tg->parent = parent;
	INIT_LIST_HEAD(&tg->children);
6404
	list_add_rcu(&tg->siblings, &parent->children);
6405
	spin_unlock_irqrestore(&task_group_lock, flags);
6406 6407

	online_fair_sched_group(tg);
S
Srivatsa Vaddagiri 已提交
6408 6409
}

6410
/* rcu callback to free various structures associated with a task group */
6411
static void sched_free_group_rcu(struct rcu_head *rhp)
S
Srivatsa Vaddagiri 已提交
6412
{
I
Ingo Molnar 已提交
6413
	/* Now it should be safe to free those cfs_rqs: */
6414
	sched_free_group(container_of(rhp, struct task_group, rcu));
S
Srivatsa Vaddagiri 已提交
6415 6416
}

6417
void sched_destroy_group(struct task_group *tg)
6418
{
I
Ingo Molnar 已提交
6419
	/* Wait for possible concurrent references to cfs_rqs complete: */
6420
	call_rcu(&tg->rcu, sched_free_group_rcu);
6421 6422 6423
}

void sched_offline_group(struct task_group *tg)
S
Srivatsa Vaddagiri 已提交
6424
{
6425
	unsigned long flags;
S
Srivatsa Vaddagiri 已提交
6426

I
Ingo Molnar 已提交
6427
	/* End participation in shares distribution: */
6428
	unregister_fair_sched_group(tg);
6429 6430

	spin_lock_irqsave(&task_group_lock, flags);
P
Peter Zijlstra 已提交
6431
	list_del_rcu(&tg->list);
P
Peter Zijlstra 已提交
6432
	list_del_rcu(&tg->siblings);
6433
	spin_unlock_irqrestore(&task_group_lock, flags);
S
Srivatsa Vaddagiri 已提交
6434 6435
}

6436
static void sched_change_group(struct task_struct *tsk, int type)
S
Srivatsa Vaddagiri 已提交
6437
{
P
Peter Zijlstra 已提交
6438
	struct task_group *tg;
S
Srivatsa Vaddagiri 已提交
6439

6440 6441 6442 6443 6444 6445
	/*
	 * 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
6446 6447 6448 6449
			  struct task_group, css);
	tg = autogroup_task_group(tsk, tg);
	tsk->sched_task_group = tg;

P
Peter Zijlstra 已提交
6450
#ifdef CONFIG_FAIR_GROUP_SCHED
6451 6452
	if (tsk->sched_class->task_change_group)
		tsk->sched_class->task_change_group(tsk, type);
6453
	else
P
Peter Zijlstra 已提交
6454
#endif
6455
		set_task_rq(tsk, task_cpu(tsk));
6456 6457 6458 6459 6460 6461 6462 6463 6464 6465 6466
}

/*
 * 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)
{
6467 6468
	int queued, running, queue_flags =
		DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
6469 6470 6471 6472
	struct rq_flags rf;
	struct rq *rq;

	rq = task_rq_lock(tsk, &rf);
6473
	update_rq_clock(rq);
6474 6475 6476 6477 6478

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

	if (queued)
6479
		dequeue_task(rq, tsk, queue_flags);
6480
	if (running)
6481 6482 6483
		put_prev_task(rq, tsk);

	sched_change_group(tsk, TASK_MOVE_GROUP);
P
Peter Zijlstra 已提交
6484

6485
	if (queued)
6486
		enqueue_task(rq, tsk, queue_flags);
6487
	if (running)
6488
		set_curr_task(rq, tsk);
S
Srivatsa Vaddagiri 已提交
6489

6490
	task_rq_unlock(rq, tsk, &rf);
S
Srivatsa Vaddagiri 已提交
6491
}
D
Dhaval Giani 已提交
6492
#endif /* CONFIG_CGROUP_SCHED */
S
Srivatsa Vaddagiri 已提交
6493

6494 6495 6496 6497 6498
#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
6499

P
Peter Zijlstra 已提交
6500 6501
/* Must be called with tasklist_lock held */
static inline int tg_has_rt_tasks(struct task_group *tg)
6502
{
P
Peter Zijlstra 已提交
6503
	struct task_struct *g, *p;
6504

6505 6506 6507 6508 6509 6510
	/*
	 * Autogroups do not have RT tasks; see autogroup_create().
	 */
	if (task_group_is_autogroup(tg))
		return 0;

6511
	for_each_process_thread(g, p) {
6512
		if (rt_task(p) && task_group(p) == tg)
P
Peter Zijlstra 已提交
6513
			return 1;
6514
	}
6515

P
Peter Zijlstra 已提交
6516 6517
	return 0;
}
6518

P
Peter Zijlstra 已提交
6519 6520 6521 6522 6523
struct rt_schedulable_data {
	struct task_group *tg;
	u64 rt_period;
	u64 rt_runtime;
};
6524

6525
static int tg_rt_schedulable(struct task_group *tg, void *data)
P
Peter Zijlstra 已提交
6526 6527 6528 6529 6530
{
	struct rt_schedulable_data *d = data;
	struct task_group *child;
	unsigned long total, sum = 0;
	u64 period, runtime;
6531

P
Peter Zijlstra 已提交
6532 6533
	period = ktime_to_ns(tg->rt_bandwidth.rt_period);
	runtime = tg->rt_bandwidth.rt_runtime;
6534

P
Peter Zijlstra 已提交
6535 6536 6537
	if (tg == d->tg) {
		period = d->rt_period;
		runtime = d->rt_runtime;
6538 6539
	}

6540 6541 6542 6543 6544
	/*
	 * Cannot have more runtime than the period.
	 */
	if (runtime > period && runtime != RUNTIME_INF)
		return -EINVAL;
P
Peter Zijlstra 已提交
6545

6546 6547 6548
	/*
	 * Ensure we don't starve existing RT tasks.
	 */
P
Peter Zijlstra 已提交
6549 6550
	if (rt_bandwidth_enabled() && !runtime && tg_has_rt_tasks(tg))
		return -EBUSY;
P
Peter Zijlstra 已提交
6551

P
Peter Zijlstra 已提交
6552
	total = to_ratio(period, runtime);
P
Peter Zijlstra 已提交
6553

6554 6555 6556 6557 6558
	/*
	 * 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
6559

6560 6561 6562
	/*
	 * The sum of our children's runtime should not exceed our own.
	 */
P
Peter Zijlstra 已提交
6563 6564 6565
	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
6566

P
Peter Zijlstra 已提交
6567 6568 6569 6570
		if (child == d->tg) {
			period = d->rt_period;
			runtime = d->rt_runtime;
		}
P
Peter Zijlstra 已提交
6571

P
Peter Zijlstra 已提交
6572
		sum += to_ratio(period, runtime);
P
Peter Zijlstra 已提交
6573
	}
P
Peter Zijlstra 已提交
6574

P
Peter Zijlstra 已提交
6575 6576 6577 6578
	if (sum > total)
		return -EINVAL;

	return 0;
P
Peter Zijlstra 已提交
6579 6580
}

P
Peter Zijlstra 已提交
6581
static int __rt_schedulable(struct task_group *tg, u64 period, u64 runtime)
6582
{
6583 6584
	int ret;

P
Peter Zijlstra 已提交
6585 6586 6587 6588 6589 6590
	struct rt_schedulable_data data = {
		.tg = tg,
		.rt_period = period,
		.rt_runtime = runtime,
	};

6591 6592 6593 6594 6595
	rcu_read_lock();
	ret = walk_tg_tree(tg_rt_schedulable, tg_nop, &data);
	rcu_read_unlock();

	return ret;
6596 6597
}

6598
static int tg_set_rt_bandwidth(struct task_group *tg,
6599
		u64 rt_period, u64 rt_runtime)
P
Peter Zijlstra 已提交
6600
{
P
Peter Zijlstra 已提交
6601
	int i, err = 0;
P
Peter Zijlstra 已提交
6602

6603 6604 6605 6606 6607 6608 6609 6610 6611 6612 6613
	/*
	 * 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
6614
	mutex_lock(&rt_constraints_mutex);
6615
	read_lock(&tasklist_lock);
P
Peter Zijlstra 已提交
6616 6617
	err = __rt_schedulable(tg, rt_period, rt_runtime);
	if (err)
P
Peter Zijlstra 已提交
6618
		goto unlock;
P
Peter Zijlstra 已提交
6619

6620
	raw_spin_lock_irq(&tg->rt_bandwidth.rt_runtime_lock);
6621 6622
	tg->rt_bandwidth.rt_period = ns_to_ktime(rt_period);
	tg->rt_bandwidth.rt_runtime = rt_runtime;
P
Peter Zijlstra 已提交
6623 6624 6625 6626

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

6627
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
6628
		rt_rq->rt_runtime = rt_runtime;
6629
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
6630
	}
6631
	raw_spin_unlock_irq(&tg->rt_bandwidth.rt_runtime_lock);
P
Peter Zijlstra 已提交
6632
unlock:
6633
	read_unlock(&tasklist_lock);
P
Peter Zijlstra 已提交
6634 6635 6636
	mutex_unlock(&rt_constraints_mutex);

	return err;
P
Peter Zijlstra 已提交
6637 6638
}

6639
static int sched_group_set_rt_runtime(struct task_group *tg, long rt_runtime_us)
6640 6641 6642 6643 6644 6645 6646 6647
{
	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;

6648
	return tg_set_rt_bandwidth(tg, rt_period, rt_runtime);
6649 6650
}

6651
static long sched_group_rt_runtime(struct task_group *tg)
P
Peter Zijlstra 已提交
6652 6653 6654
{
	u64 rt_runtime_us;

6655
	if (tg->rt_bandwidth.rt_runtime == RUNTIME_INF)
P
Peter Zijlstra 已提交
6656 6657
		return -1;

6658
	rt_runtime_us = tg->rt_bandwidth.rt_runtime;
P
Peter Zijlstra 已提交
6659 6660 6661
	do_div(rt_runtime_us, NSEC_PER_USEC);
	return rt_runtime_us;
}
6662

6663
static int sched_group_set_rt_period(struct task_group *tg, u64 rt_period_us)
6664 6665 6666
{
	u64 rt_runtime, rt_period;

6667
	rt_period = rt_period_us * NSEC_PER_USEC;
6668 6669
	rt_runtime = tg->rt_bandwidth.rt_runtime;

6670
	return tg_set_rt_bandwidth(tg, rt_period, rt_runtime);
6671 6672
}

6673
static long sched_group_rt_period(struct task_group *tg)
6674 6675 6676 6677 6678 6679 6680
{
	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;
}
6681
#endif /* CONFIG_RT_GROUP_SCHED */
6682

6683
#ifdef CONFIG_RT_GROUP_SCHED
6684 6685 6686 6687 6688
static int sched_rt_global_constraints(void)
{
	int ret = 0;

	mutex_lock(&rt_constraints_mutex);
P
Peter Zijlstra 已提交
6689
	read_lock(&tasklist_lock);
6690
	ret = __rt_schedulable(NULL, 0, 0);
P
Peter Zijlstra 已提交
6691
	read_unlock(&tasklist_lock);
6692 6693 6694 6695
	mutex_unlock(&rt_constraints_mutex);

	return ret;
}
6696

6697
static int sched_rt_can_attach(struct task_group *tg, struct task_struct *tsk)
6698 6699 6700 6701 6702 6703 6704 6705
{
	/* 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;
}

6706
#else /* !CONFIG_RT_GROUP_SCHED */
6707 6708
static int sched_rt_global_constraints(void)
{
P
Peter Zijlstra 已提交
6709
	unsigned long flags;
6710
	int i;
6711

6712
	raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags);
P
Peter Zijlstra 已提交
6713 6714 6715
	for_each_possible_cpu(i) {
		struct rt_rq *rt_rq = &cpu_rq(i)->rt;

6716
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
6717
		rt_rq->rt_runtime = global_rt_runtime();
6718
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
6719
	}
6720
	raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags);
P
Peter Zijlstra 已提交
6721

6722
	return 0;
6723
}
6724
#endif /* CONFIG_RT_GROUP_SCHED */
6725

6726
static int sched_dl_global_validate(void)
6727
{
6728 6729
	u64 runtime = global_rt_runtime();
	u64 period = global_rt_period();
6730
	u64 new_bw = to_ratio(period, runtime);
6731
	struct dl_bw *dl_b;
6732
	int cpu, ret = 0;
6733
	unsigned long flags;
6734 6735 6736 6737 6738 6739 6740 6741 6742 6743

	/*
	 * 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!
	 */
6744
	for_each_possible_cpu(cpu) {
6745 6746
		rcu_read_lock_sched();
		dl_b = dl_bw_of(cpu);
6747

6748
		raw_spin_lock_irqsave(&dl_b->lock, flags);
6749 6750
		if (new_bw < dl_b->total_bw)
			ret = -EBUSY;
6751
		raw_spin_unlock_irqrestore(&dl_b->lock, flags);
6752

6753 6754
		rcu_read_unlock_sched();

6755 6756
		if (ret)
			break;
6757 6758
	}

6759
	return ret;
6760 6761
}

6762
static void sched_dl_do_global(void)
6763
{
6764
	u64 new_bw = -1;
6765
	struct dl_bw *dl_b;
6766
	int cpu;
6767
	unsigned long flags;
6768

6769 6770 6771 6772 6773 6774 6775 6776 6777 6778
	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) {
6779 6780
		rcu_read_lock_sched();
		dl_b = dl_bw_of(cpu);
6781

6782
		raw_spin_lock_irqsave(&dl_b->lock, flags);
6783
		dl_b->bw = new_bw;
6784
		raw_spin_unlock_irqrestore(&dl_b->lock, flags);
6785 6786

		rcu_read_unlock_sched();
6787
	}
6788 6789 6790 6791 6792 6793 6794
}

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

6795 6796
	if ((sysctl_sched_rt_runtime != RUNTIME_INF) &&
		(sysctl_sched_rt_runtime > sysctl_sched_rt_period))
6797 6798 6799 6800 6801 6802 6803 6804 6805
		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());
6806 6807
}

6808
int sched_rt_handler(struct ctl_table *table, int write,
6809
		void __user *buffer, size_t *lenp,
6810 6811 6812 6813
		loff_t *ppos)
{
	int old_period, old_runtime;
	static DEFINE_MUTEX(mutex);
6814
	int ret;
6815 6816 6817 6818 6819

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

6820
	ret = proc_dointvec(table, write, buffer, lenp, ppos);
6821 6822

	if (!ret && write) {
6823 6824 6825 6826
		ret = sched_rt_global_validate();
		if (ret)
			goto undo;

6827
		ret = sched_dl_global_validate();
6828 6829 6830
		if (ret)
			goto undo;

6831
		ret = sched_rt_global_constraints();
6832 6833 6834 6835 6836 6837 6838 6839 6840 6841
		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;
6842 6843 6844 6845 6846
	}
	mutex_unlock(&mutex);

	return ret;
}
6847

6848
int sched_rr_handler(struct ctl_table *table, int write,
6849 6850 6851 6852 6853 6854 6855 6856
		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
6857 6858 6859 6860
	/*
	 * Make sure that internally we keep jiffies.
	 * Also, writing zero resets the timeslice to default:
	 */
6861
	if (!ret && write) {
6862 6863 6864
		sched_rr_timeslice =
			sysctl_sched_rr_timeslice <= 0 ? RR_TIMESLICE :
			msecs_to_jiffies(sysctl_sched_rr_timeslice);
6865 6866 6867 6868 6869
	}
	mutex_unlock(&mutex);
	return ret;
}

6870
#ifdef CONFIG_CGROUP_SCHED
6871

6872
static inline struct task_group *css_tg(struct cgroup_subsys_state *css)
6873
{
6874
	return css ? container_of(css, struct task_group, css) : NULL;
6875 6876
}

6877 6878
static struct cgroup_subsys_state *
cpu_cgroup_css_alloc(struct cgroup_subsys_state *parent_css)
6879
{
6880 6881
	struct task_group *parent = css_tg(parent_css);
	struct task_group *tg;
6882

6883
	if (!parent) {
6884
		/* This is early initialization for the top cgroup */
6885
		return &root_task_group.css;
6886 6887
	}

6888
	tg = sched_create_group(parent);
6889 6890 6891 6892 6893 6894
	if (IS_ERR(tg))
		return ERR_PTR(-ENOMEM);

	return &tg->css;
}

6895 6896 6897 6898 6899 6900 6901 6902 6903 6904 6905
/* Expose task group only after completing cgroup initialization */
static int cpu_cgroup_css_online(struct cgroup_subsys_state *css)
{
	struct task_group *tg = css_tg(css);
	struct task_group *parent = css_tg(css->parent);

	if (parent)
		sched_online_group(tg, parent);
	return 0;
}

6906
static void cpu_cgroup_css_released(struct cgroup_subsys_state *css)
6907
{
6908
	struct task_group *tg = css_tg(css);
6909

6910
	sched_offline_group(tg);
6911 6912
}

6913
static void cpu_cgroup_css_free(struct cgroup_subsys_state *css)
6914
{
6915
	struct task_group *tg = css_tg(css);
6916

6917 6918 6919 6920
	/*
	 * Relies on the RCU grace period between css_released() and this.
	 */
	sched_free_group(tg);
6921 6922
}

6923 6924 6925 6926
/*
 * 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.
 */
6927
static void cpu_cgroup_fork(struct task_struct *task)
6928
{
6929 6930 6931 6932 6933
	struct rq_flags rf;
	struct rq *rq;

	rq = task_rq_lock(task, &rf);

6934
	update_rq_clock(rq);
6935 6936 6937
	sched_change_group(task, TASK_SET_GROUP);

	task_rq_unlock(rq, task, &rf);
6938 6939
}

6940
static int cpu_cgroup_can_attach(struct cgroup_taskset *tset)
6941
{
6942
	struct task_struct *task;
6943
	struct cgroup_subsys_state *css;
6944
	int ret = 0;
6945

6946
	cgroup_taskset_for_each(task, css, tset) {
6947
#ifdef CONFIG_RT_GROUP_SCHED
6948
		if (!sched_rt_can_attach(css_tg(css), task))
6949
			return -EINVAL;
6950
#else
6951 6952 6953
		/* We don't support RT-tasks being in separate groups */
		if (task->sched_class != &fair_sched_class)
			return -EINVAL;
6954
#endif
6955 6956 6957 6958 6959 6960 6961 6962 6963 6964 6965 6966 6967 6968 6969 6970
		/*
		 * 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;
6971
	}
6972
	return ret;
6973
}
6974

6975
static void cpu_cgroup_attach(struct cgroup_taskset *tset)
6976
{
6977
	struct task_struct *task;
6978
	struct cgroup_subsys_state *css;
6979

6980
	cgroup_taskset_for_each(task, css, tset)
6981
		sched_move_task(task);
6982 6983
}

6984
#ifdef CONFIG_FAIR_GROUP_SCHED
6985 6986
static int cpu_shares_write_u64(struct cgroup_subsys_state *css,
				struct cftype *cftype, u64 shareval)
6987
{
6988
	return sched_group_set_shares(css_tg(css), scale_load(shareval));
6989 6990
}

6991 6992
static u64 cpu_shares_read_u64(struct cgroup_subsys_state *css,
			       struct cftype *cft)
6993
{
6994
	struct task_group *tg = css_tg(css);
6995

6996
	return (u64) scale_load_down(tg->shares);
6997
}
6998 6999

#ifdef CONFIG_CFS_BANDWIDTH
7000 7001
static DEFINE_MUTEX(cfs_constraints_mutex);

7002 7003 7004
const u64 max_cfs_quota_period = 1 * NSEC_PER_SEC; /* 1s */
const u64 min_cfs_quota_period = 1 * NSEC_PER_MSEC; /* 1ms */

7005 7006
static int __cfs_schedulable(struct task_group *tg, u64 period, u64 runtime);

7007 7008
static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota)
{
7009
	int i, ret = 0, runtime_enabled, runtime_was_enabled;
7010
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
7011 7012 7013 7014 7015 7016 7017 7018 7019 7020 7021 7022 7023 7024 7025 7026 7027 7028 7029 7030

	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;

7031 7032 7033 7034 7035
	/*
	 * Prevent race between setting of cfs_rq->runtime_enabled and
	 * unthrottle_offline_cfs_rqs().
	 */
	get_online_cpus();
7036 7037 7038 7039 7040
	mutex_lock(&cfs_constraints_mutex);
	ret = __cfs_schedulable(tg, period, quota);
	if (ret)
		goto out_unlock;

7041
	runtime_enabled = quota != RUNTIME_INF;
7042
	runtime_was_enabled = cfs_b->quota != RUNTIME_INF;
7043 7044 7045 7046 7047 7048
	/*
	 * 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();
7049 7050 7051
	raw_spin_lock_irq(&cfs_b->lock);
	cfs_b->period = ns_to_ktime(period);
	cfs_b->quota = quota;
7052

P
Paul Turner 已提交
7053
	__refill_cfs_bandwidth_runtime(cfs_b);
I
Ingo Molnar 已提交
7054 7055

	/* Restart the period timer (if active) to handle new period expiry: */
P
Peter Zijlstra 已提交
7056 7057
	if (runtime_enabled)
		start_cfs_bandwidth(cfs_b);
I
Ingo Molnar 已提交
7058

7059 7060
	raw_spin_unlock_irq(&cfs_b->lock);

7061
	for_each_online_cpu(i) {
7062
		struct cfs_rq *cfs_rq = tg->cfs_rq[i];
7063
		struct rq *rq = cfs_rq->rq;
7064
		struct rq_flags rf;
7065

7066
		rq_lock_irq(rq, &rf);
7067
		cfs_rq->runtime_enabled = runtime_enabled;
7068
		cfs_rq->runtime_remaining = 0;
7069

7070
		if (cfs_rq->throttled)
7071
			unthrottle_cfs_rq(cfs_rq);
7072
		rq_unlock_irq(rq, &rf);
7073
	}
7074 7075
	if (runtime_was_enabled && !runtime_enabled)
		cfs_bandwidth_usage_dec();
7076 7077
out_unlock:
	mutex_unlock(&cfs_constraints_mutex);
7078
	put_online_cpus();
7079

7080
	return ret;
7081 7082 7083 7084 7085 7086
}

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

7087
	period = ktime_to_ns(tg->cfs_bandwidth.period);
7088 7089 7090 7091 7092 7093 7094 7095 7096 7097 7098 7099
	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;

7100
	if (tg->cfs_bandwidth.quota == RUNTIME_INF)
7101 7102
		return -1;

7103
	quota_us = tg->cfs_bandwidth.quota;
7104 7105 7106 7107 7108 7109 7110 7111 7112 7113
	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;
7114
	quota = tg->cfs_bandwidth.quota;
7115 7116 7117 7118 7119 7120 7121 7122

	return tg_set_cfs_bandwidth(tg, period, quota);
}

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

7123
	cfs_period_us = ktime_to_ns(tg->cfs_bandwidth.period);
7124 7125 7126 7127 7128
	do_div(cfs_period_us, NSEC_PER_USEC);

	return cfs_period_us;
}

7129 7130
static s64 cpu_cfs_quota_read_s64(struct cgroup_subsys_state *css,
				  struct cftype *cft)
7131
{
7132
	return tg_get_cfs_quota(css_tg(css));
7133 7134
}

7135 7136
static int cpu_cfs_quota_write_s64(struct cgroup_subsys_state *css,
				   struct cftype *cftype, s64 cfs_quota_us)
7137
{
7138
	return tg_set_cfs_quota(css_tg(css), cfs_quota_us);
7139 7140
}

7141 7142
static u64 cpu_cfs_period_read_u64(struct cgroup_subsys_state *css,
				   struct cftype *cft)
7143
{
7144
	return tg_get_cfs_period(css_tg(css));
7145 7146
}

7147 7148
static int cpu_cfs_period_write_u64(struct cgroup_subsys_state *css,
				    struct cftype *cftype, u64 cfs_period_us)
7149
{
7150
	return tg_set_cfs_period(css_tg(css), cfs_period_us);
7151 7152
}

7153 7154 7155 7156 7157 7158 7159 7160 7161 7162 7163 7164 7165 7166 7167 7168 7169 7170 7171 7172 7173 7174 7175 7176 7177 7178 7179 7180 7181 7182 7183 7184
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;
7185
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
7186 7187 7188 7189 7190
	s64 quota = 0, parent_quota = -1;

	if (!tg->parent) {
		quota = RUNTIME_INF;
	} else {
7191
		struct cfs_bandwidth *parent_b = &tg->parent->cfs_bandwidth;
7192 7193

		quota = normalize_cfs_quota(tg, d);
7194
		parent_quota = parent_b->hierarchical_quota;
7195 7196

		/*
I
Ingo Molnar 已提交
7197 7198
		 * Ensure max(child_quota) <= parent_quota, inherit when no
		 * limit is set:
7199 7200 7201 7202 7203 7204
		 */
		if (quota == RUNTIME_INF)
			quota = parent_quota;
		else if (parent_quota != RUNTIME_INF && quota > parent_quota)
			return -EINVAL;
	}
7205
	cfs_b->hierarchical_quota = quota;
7206 7207 7208 7209 7210 7211

	return 0;
}

static int __cfs_schedulable(struct task_group *tg, u64 period, u64 quota)
{
7212
	int ret;
7213 7214 7215 7216 7217 7218 7219 7220 7221 7222 7223
	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);
	}

7224 7225 7226 7227 7228
	rcu_read_lock();
	ret = walk_tg_tree(tg_cfs_schedulable_down, tg_nop, &data);
	rcu_read_unlock();

	return ret;
7229
}
7230

7231
static int cpu_stats_show(struct seq_file *sf, void *v)
7232
{
7233
	struct task_group *tg = css_tg(seq_css(sf));
7234
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
7235

7236 7237 7238
	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);
7239 7240 7241

	return 0;
}
7242
#endif /* CONFIG_CFS_BANDWIDTH */
7243
#endif /* CONFIG_FAIR_GROUP_SCHED */
7244

7245
#ifdef CONFIG_RT_GROUP_SCHED
7246 7247
static int cpu_rt_runtime_write(struct cgroup_subsys_state *css,
				struct cftype *cft, s64 val)
P
Peter Zijlstra 已提交
7248
{
7249
	return sched_group_set_rt_runtime(css_tg(css), val);
P
Peter Zijlstra 已提交
7250 7251
}

7252 7253
static s64 cpu_rt_runtime_read(struct cgroup_subsys_state *css,
			       struct cftype *cft)
P
Peter Zijlstra 已提交
7254
{
7255
	return sched_group_rt_runtime(css_tg(css));
P
Peter Zijlstra 已提交
7256
}
7257

7258 7259
static int cpu_rt_period_write_uint(struct cgroup_subsys_state *css,
				    struct cftype *cftype, u64 rt_period_us)
7260
{
7261
	return sched_group_set_rt_period(css_tg(css), rt_period_us);
7262 7263
}

7264 7265
static u64 cpu_rt_period_read_uint(struct cgroup_subsys_state *css,
				   struct cftype *cft)
7266
{
7267
	return sched_group_rt_period(css_tg(css));
7268
}
7269
#endif /* CONFIG_RT_GROUP_SCHED */
P
Peter Zijlstra 已提交
7270

7271
static struct cftype cpu_files[] = {
7272
#ifdef CONFIG_FAIR_GROUP_SCHED
7273 7274
	{
		.name = "shares",
7275 7276
		.read_u64 = cpu_shares_read_u64,
		.write_u64 = cpu_shares_write_u64,
7277
	},
7278
#endif
7279 7280 7281 7282 7283 7284 7285 7286 7287 7288 7289
#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,
	},
7290 7291
	{
		.name = "stat",
7292
		.seq_show = cpu_stats_show,
7293
	},
7294
#endif
7295
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
7296
	{
P
Peter Zijlstra 已提交
7297
		.name = "rt_runtime_us",
7298 7299
		.read_s64 = cpu_rt_runtime_read,
		.write_s64 = cpu_rt_runtime_write,
P
Peter Zijlstra 已提交
7300
	},
7301 7302
	{
		.name = "rt_period_us",
7303 7304
		.read_u64 = cpu_rt_period_read_uint,
		.write_u64 = cpu_rt_period_write_uint,
7305
	},
7306
#endif
I
Ingo Molnar 已提交
7307
	{ }	/* Terminate */
7308 7309
};

7310
struct cgroup_subsys cpu_cgrp_subsys = {
7311
	.css_alloc	= cpu_cgroup_css_alloc,
7312
	.css_online	= cpu_cgroup_css_online,
7313
	.css_released	= cpu_cgroup_css_released,
7314
	.css_free	= cpu_cgroup_css_free,
7315
	.fork		= cpu_cgroup_fork,
7316 7317
	.can_attach	= cpu_cgroup_can_attach,
	.attach		= cpu_cgroup_attach,
7318
	.legacy_cftypes	= cpu_files,
7319
	.early_init	= true,
7320 7321
};

7322
#endif	/* CONFIG_CGROUP_SCHED */
7323

7324 7325 7326 7327 7328
void dump_cpu_task(int cpu)
{
	pr_info("Task dump for CPU %d:\n", cpu);
	sched_show_task(cpu_curr(cpu));
}
7329 7330 7331 7332 7333 7334 7335 7336 7337 7338 7339 7340 7341 7342 7343 7344 7345 7346 7347 7348 7349 7350 7351 7352 7353 7354 7355 7356 7357 7358 7359 7360 7361 7362 7363 7364 7365 7366 7367 7368 7369

/*
 * 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,
};