core.c 180.2 KB
Newer Older
L
Linus Torvalds 已提交
1
/*
2
 *  kernel/sched/core.c
L
Linus Torvalds 已提交
3
 *
I
Ingo Molnar 已提交
4
 *  Core kernel scheduler code and related syscalls
L
Linus Torvalds 已提交
5 6 7
 *
 *  Copyright (C) 1991-2002  Linus Torvalds
 */
8
#include <linux/sched.h>
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>
13
#include <linux/wait_bit.h>
L
Linus Torvalds 已提交
14
#include <linux/cpuset.h>
15
#include <linux/delayacct.h>
16
#include <linux/init_task.h>
17
#include <linux/context_tracking.h>
18
#include <linux/rcupdate_wait.h>
19 20 21 22 23 24

#include <linux/blkdev.h>
#include <linux/kprobes.h>
#include <linux/mmu_context.h>
#include <linux/module.h>
#include <linux/nmi.h>
25
#include <linux/prefetch.h>
26 27 28
#include <linux/profile.h>
#include <linux/security.h>
#include <linux/syscalls.h>
L
Linus Torvalds 已提交
29

30
#include <asm/switch_to.h>
31
#include <asm/tlb.h>
32 33 34
#ifdef CONFIG_PARAVIRT
#include <asm/paravirt.h>
#endif
L
Linus Torvalds 已提交
35

36
#include "sched.h"
37
#include "../workqueue_internal.h"
38
#include "../smpboot.h"
39

40
#define CREATE_TRACE_POINTS
41
#include <trace/events/sched.h>
42

43
DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
44

I
Ingo Molnar 已提交
45 46 47
/*
 * Debugging: various feature bits
 */
P
Peter Zijlstra 已提交
48 49 50 51

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

I
Ingo Molnar 已提交
52
const_debug unsigned int sysctl_sched_features =
53
#include "features.h"
P
Peter Zijlstra 已提交
54 55 56 57
	0;

#undef SCHED_FEAT

58 59 60 61 62 63
/*
 * 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;

64 65 66 67 68 69 70 71
/*
 * 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
72
/*
I
Ingo Molnar 已提交
73
 * period over which we measure -rt task CPU usage in us.
P
Peter Zijlstra 已提交
74 75
 * default: 1s
 */
P
Peter Zijlstra 已提交
76
unsigned int sysctl_sched_rt_period = 1000000;
P
Peter Zijlstra 已提交
77

78
__read_mostly int scheduler_running;
79

P
Peter Zijlstra 已提交
80 81 82 83 84
/*
 * 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
85

I
Ingo Molnar 已提交
86
/* CPUs with isolated domains */
87 88
cpumask_var_t cpu_isolated_map;

89 90 91
/*
 * __task_rq_lock - lock the rq @p resides on.
 */
92
struct rq *__task_rq_lock(struct task_struct *p, struct rq_flags *rf)
93 94 95 96 97 98 99 100 101 102
	__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))) {
103
			rq_pin_lock(rq, rf);
104 105 106 107 108 109 110 111 112 113 114 115
			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.
 */
116
struct rq *task_rq_lock(struct task_struct *p, struct rq_flags *rf)
117 118 119 120 121 122
	__acquires(p->pi_lock)
	__acquires(rq->lock)
{
	struct rq *rq;

	for (;;) {
123
		raw_spin_lock_irqsave(&p->pi_lock, rf->flags);
124 125 126 127 128 129 130 131 132 133 134 135 136 137 138
		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
139
		 * If we observe the new CPU in task_rq_lock, the acquire will
140 141 142
		 * pair with the WMB to ensure we must then also see migrating.
		 */
		if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) {
143
			rq_pin_lock(rq, rf);
144 145 146
			return rq;
		}
		raw_spin_unlock(&rq->lock);
147
		raw_spin_unlock_irqrestore(&p->pi_lock, rf->flags);
148 149 150 151 152 153

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

154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221
/*
 * 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
222 223
	if (sched_feat(WARN_DOUBLE_CLOCK))
		SCHED_WARN_ON(rq->clock_update_flags & RQCF_UPDATED);
224 225
	rq->clock_update_flags |= RQCF_UPDATED;
#endif
226

227 228 229 230 231 232 233 234
	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
235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252
#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);
253
	struct rq_flags rf;
P
Peter Zijlstra 已提交
254 255 256

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

257
	rq_lock(rq, &rf);
258
	update_rq_clock(rq);
P
Peter Zijlstra 已提交
259
	rq->curr->sched_class->task_tick(rq, rq->curr, 1);
260
	rq_unlock(rq, &rf);
P
Peter Zijlstra 已提交
261 262 263 264

	return HRTIMER_NORESTART;
}

265
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
266

267
static void __hrtick_restart(struct rq *rq)
P
Peter Zijlstra 已提交
268 269 270
{
	struct hrtimer *timer = &rq->hrtick_timer;

271
	hrtimer_start_expires(timer, HRTIMER_MODE_ABS_PINNED);
P
Peter Zijlstra 已提交
272 273
}

274 275 276 277
/*
 * called from hardirq (IPI) context
 */
static void __hrtick_start(void *arg)
278
{
279
	struct rq *rq = arg;
280
	struct rq_flags rf;
281

282
	rq_lock(rq, &rf);
P
Peter Zijlstra 已提交
283
	__hrtick_restart(rq);
284
	rq->hrtick_csd_pending = 0;
285
	rq_unlock(rq, &rf);
286 287
}

288 289 290 291 292
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
293
void hrtick_start(struct rq *rq, u64 delay)
294
{
295
	struct hrtimer *timer = &rq->hrtick_timer;
296 297 298 299 300 301 302 303 304
	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);
305

306
	hrtimer_set_expires(timer, time);
307 308

	if (rq == this_rq()) {
P
Peter Zijlstra 已提交
309
		__hrtick_restart(rq);
310
	} else if (!rq->hrtick_csd_pending) {
311
		smp_call_function_single_async(cpu_of(rq), &rq->hrtick_csd);
312 313
		rq->hrtick_csd_pending = 1;
	}
314 315
}

316 317 318 319 320 321
#else
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
322
void hrtick_start(struct rq *rq, u64 delay)
323
{
W
Wanpeng Li 已提交
324 325 326 327 328
	/*
	 * Don't schedule slices shorter than 10000ns, that just
	 * doesn't make sense. Rely on vruntime for fairness.
	 */
	delay = max_t(u64, delay, 10000LL);
329 330
	hrtimer_start(&rq->hrtick_timer, ns_to_ktime(delay),
		      HRTIMER_MODE_REL_PINNED);
331 332
}
#endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
333

334
static void init_rq_hrtick(struct rq *rq)
P
Peter Zijlstra 已提交
335
{
336 337
#ifdef CONFIG_SMP
	rq->hrtick_csd_pending = 0;
P
Peter Zijlstra 已提交
338

339 340 341 342
	rq->hrtick_csd.flags = 0;
	rq->hrtick_csd.func = __hrtick_start;
	rq->hrtick_csd.info = rq;
#endif
P
Peter Zijlstra 已提交
343

344 345
	hrtimer_init(&rq->hrtick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	rq->hrtick_timer.function = hrtick;
P
Peter Zijlstra 已提交
346
}
A
Andrew Morton 已提交
347
#else	/* CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
348 349 350 351 352 353 354
static inline void hrtick_clear(struct rq *rq)
{
}

static inline void init_rq_hrtick(struct rq *rq)
{
}
A
Andrew Morton 已提交
355
#endif	/* CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
356

357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374
/*
 * 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;								\
})

375
#if defined(CONFIG_SMP) && defined(TIF_POLLING_NRFLAG)
376 377 378 379 380 381 382 383 384 385
/*
 * 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);
}
386 387 388 389 390 391 392 393 394 395

/*
 * 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);
396
	typeof(ti->flags) old, val = READ_ONCE(ti->flags);
397 398 399 400 401 402 403 404 405 406 407 408 409 410

	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;
}

411 412 413 414 415 416
#else
static bool set_nr_and_not_polling(struct task_struct *p)
{
	set_tsk_need_resched(p);
	return true;
}
417 418 419 420 421 422 423

#ifdef CONFIG_SMP
static bool set_nr_if_polling(struct task_struct *p)
{
	return false;
}
#endif
424 425
#endif

426 427 428 429 430 431 432 433 434 435
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
436
	 * barrier implied by the wakeup in wake_up_q().
437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458
	 */
	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
459
		/* Task can safely be re-inserted now: */
460 461 462 463 464 465 466 467 468 469 470 471
		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
472
/*
473
 * resched_curr - mark rq's current task 'to be rescheduled now'.
I
Ingo Molnar 已提交
474 475 476 477 478
 *
 * 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.
 */
479
void resched_curr(struct rq *rq)
I
Ingo Molnar 已提交
480
{
481
	struct task_struct *curr = rq->curr;
I
Ingo Molnar 已提交
482 483
	int cpu;

484
	lockdep_assert_held(&rq->lock);
I
Ingo Molnar 已提交
485

486
	if (test_tsk_need_resched(curr))
I
Ingo Molnar 已提交
487 488
		return;

489
	cpu = cpu_of(rq);
490

491
	if (cpu == smp_processor_id()) {
492
		set_tsk_need_resched(curr);
493
		set_preempt_need_resched();
I
Ingo Molnar 已提交
494
		return;
495
	}
I
Ingo Molnar 已提交
496

497
	if (set_nr_and_not_polling(curr))
I
Ingo Molnar 已提交
498
		smp_send_reschedule(cpu);
499 500
	else
		trace_sched_wake_idle_without_ipi(cpu);
I
Ingo Molnar 已提交
501 502
}

503
void resched_cpu(int cpu)
I
Ingo Molnar 已提交
504 505 506 507
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

508
	if (!raw_spin_trylock_irqsave(&rq->lock, flags))
I
Ingo Molnar 已提交
509
		return;
510
	resched_curr(rq);
511
	raw_spin_unlock_irqrestore(&rq->lock, flags);
I
Ingo Molnar 已提交
512
}
513

514
#ifdef CONFIG_SMP
515
#ifdef CONFIG_NO_HZ_COMMON
516
/*
I
Ingo Molnar 已提交
517 518
 * In the semi idle case, use the nearest busy CPU for migrating timers
 * from an idle CPU.  This is good for power-savings.
519 520
 *
 * We don't do similar optimization for completely idle system, as
I
Ingo Molnar 已提交
521 522
 * 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).
523
 */
524
int get_nohz_timer_target(void)
525
{
526
	int i, cpu = smp_processor_id();
527 528
	struct sched_domain *sd;

529
	if (!idle_cpu(cpu) && is_housekeeping_cpu(cpu))
530 531
		return cpu;

532
	rcu_read_lock();
533
	for_each_domain(cpu, sd) {
534
		for_each_cpu(i, sched_domain_span(sd)) {
535 536 537 538
			if (cpu == i)
				continue;

			if (!idle_cpu(i) && is_housekeeping_cpu(i)) {
539 540 541 542
				cpu = i;
				goto unlock;
			}
		}
543
	}
544 545 546

	if (!is_housekeeping_cpu(cpu))
		cpu = housekeeping_any_cpu();
547 548
unlock:
	rcu_read_unlock();
549 550
	return cpu;
}
I
Ingo Molnar 已提交
551

552 553 554 555 556 557 558 559 560 561
/*
 * 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.
 */
562
static void wake_up_idle_cpu(int cpu)
563 564 565 566 567 568
{
	struct rq *rq = cpu_rq(cpu);

	if (cpu == smp_processor_id())
		return;

569
	if (set_nr_and_not_polling(rq->idle))
570
		smp_send_reschedule(cpu);
571 572
	else
		trace_sched_wake_idle_without_ipi(cpu);
573 574
}

575
static bool wake_up_full_nohz_cpu(int cpu)
576
{
577 578 579 580 581 582
	/*
	 * 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.
	 */
583 584
	if (cpu_is_offline(cpu))
		return true;  /* Don't try to wake offline CPUs. */
585
	if (tick_nohz_full_cpu(cpu)) {
586 587
		if (cpu != smp_processor_id() ||
		    tick_nohz_tick_stopped())
588
			tick_nohz_full_kick_cpu(cpu);
589 590 591 592 593 594
		return true;
	}

	return false;
}

595 596 597 598 599
/*
 * 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.
 */
600 601
void wake_up_nohz_cpu(int cpu)
{
602
	if (!wake_up_full_nohz_cpu(cpu))
603 604 605
		wake_up_idle_cpu(cpu);
}

606
static inline bool got_nohz_idle_kick(void)
607
{
608
	int cpu = smp_processor_id();
609 610 611 612 613 614 615 616 617 618 619 620 621

	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;
622 623
}

624
#else /* CONFIG_NO_HZ_COMMON */
625

626
static inline bool got_nohz_idle_kick(void)
P
Peter Zijlstra 已提交
627
{
628
	return false;
P
Peter Zijlstra 已提交
629 630
}

631
#endif /* CONFIG_NO_HZ_COMMON */
632

633
#ifdef CONFIG_NO_HZ_FULL
634
bool sched_can_stop_tick(struct rq *rq)
635
{
636 637 638 639 640 641
	int fifo_nr_running;

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

642
	/*
643 644
	 * If there are more than one RR tasks, we need the tick to effect the
	 * actual RR behaviour.
645
	 */
646 647 648 649 650
	if (rq->rt.rr_nr_running) {
		if (rq->rt.rr_nr_running == 1)
			return true;
		else
			return false;
651 652
	}

653 654 655 656 657 658 659 660 661 662 663 664 665 666
	/*
	 * 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)
667
		return false;
668

669
	return true;
670 671
}
#endif /* CONFIG_NO_HZ_FULL */
672

673
void sched_avg_update(struct rq *rq)
674
{
675 676
	s64 period = sched_avg_period();

677
	while ((s64)(rq_clock(rq) - rq->age_stamp) > period) {
678 679 680 681 682 683
		/*
		 * 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));
684 685 686
		rq->age_stamp += period;
		rq->rt_avg /= 2;
	}
687 688
}

689
#endif /* CONFIG_SMP */
690

691 692
#if defined(CONFIG_RT_GROUP_SCHED) || (defined(CONFIG_FAIR_GROUP_SCHED) && \
			(defined(CONFIG_SMP) || defined(CONFIG_CFS_BANDWIDTH)))
693
/*
694 695 696 697
 * 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.
698
 */
699
int walk_tg_tree_from(struct task_group *from,
700
			     tg_visitor down, tg_visitor up, void *data)
701 702
{
	struct task_group *parent, *child;
P
Peter Zijlstra 已提交
703
	int ret;
704

705 706
	parent = from;

707
down:
P
Peter Zijlstra 已提交
708 709
	ret = (*down)(parent, data);
	if (ret)
710
		goto out;
711 712 713 714 715 716 717
	list_for_each_entry_rcu(child, &parent->children, siblings) {
		parent = child;
		goto down;

up:
		continue;
	}
P
Peter Zijlstra 已提交
718
	ret = (*up)(parent, data);
719 720
	if (ret || parent == from)
		goto out;
721 722 723 724 725

	child = parent;
	parent = parent->parent;
	if (parent)
		goto up;
726
out:
P
Peter Zijlstra 已提交
727
	return ret;
728 729
}

730
int tg_nop(struct task_group *tg, void *data)
P
Peter Zijlstra 已提交
731
{
732
	return 0;
P
Peter Zijlstra 已提交
733
}
734 735
#endif

736 737
static void set_load_weight(struct task_struct *p)
{
N
Nikhil Rao 已提交
738 739 740
	int prio = p->static_prio - MAX_RT_PRIO;
	struct load_weight *load = &p->se.load;

I
Ingo Molnar 已提交
741 742 743
	/*
	 * SCHED_IDLE tasks get minimal weight:
	 */
744
	if (idle_policy(p->policy)) {
745
		load->weight = scale_load(WEIGHT_IDLEPRIO);
N
Nikhil Rao 已提交
746
		load->inv_weight = WMULT_IDLEPRIO;
I
Ingo Molnar 已提交
747 748
		return;
	}
749

750 751
	load->weight = scale_load(sched_prio_to_weight[prio]);
	load->inv_weight = sched_prio_to_wmult[prio];
752 753
}

754
static inline void enqueue_task(struct rq *rq, struct task_struct *p, int flags)
755
{
756 757 758
	if (!(flags & ENQUEUE_NOCLOCK))
		update_rq_clock(rq);

759 760
	if (!(flags & ENQUEUE_RESTORE))
		sched_info_queued(rq, p);
761

762
	p->sched_class->enqueue_task(rq, p, flags);
763 764
}

765
static inline void dequeue_task(struct rq *rq, struct task_struct *p, int flags)
766
{
767 768 769
	if (!(flags & DEQUEUE_NOCLOCK))
		update_rq_clock(rq);

770 771
	if (!(flags & DEQUEUE_SAVE))
		sched_info_dequeued(rq, p);
772

773
	p->sched_class->dequeue_task(rq, p, flags);
774 775
}

776
void activate_task(struct rq *rq, struct task_struct *p, int flags)
777 778 779 780
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible--;

781
	enqueue_task(rq, p, flags);
782 783
}

784
void deactivate_task(struct rq *rq, struct task_struct *p, int flags)
785 786 787 788
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible++;

789
	dequeue_task(rq, p, flags);
790 791
}

792
/*
I
Ingo Molnar 已提交
793
 * __normal_prio - return the priority that is based on the static prio
794 795 796
 */
static inline int __normal_prio(struct task_struct *p)
{
I
Ingo Molnar 已提交
797
	return p->static_prio;
798 799
}

800 801 802 803 804 805 806
/*
 * 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.
 */
807
static inline int normal_prio(struct task_struct *p)
808 809 810
{
	int prio;

811 812 813
	if (task_has_dl_policy(p))
		prio = MAX_DL_PRIO-1;
	else if (task_has_rt_policy(p))
814 815 816 817 818 819 820 821 822 823 824 825 826
		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.
 */
827
static int effective_prio(struct task_struct *p)
828 829 830 831 832 833 834 835 836 837 838 839
{
	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
840 841 842
/**
 * task_curr - is this task currently executing on a CPU?
 * @p: the task in question.
843 844
 *
 * Return: 1 if the task is currently executing. 0 otherwise.
L
Linus Torvalds 已提交
845
 */
846
inline int task_curr(const struct task_struct *p)
L
Linus Torvalds 已提交
847 848 849 850
{
	return cpu_curr(task_cpu(p)) == p;
}

851
/*
852 853 854 855 856
 * 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().
857
 */
858 859
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
860
				       int oldprio)
861 862 863
{
	if (prev_class != p->sched_class) {
		if (prev_class->switched_from)
P
Peter Zijlstra 已提交
864
			prev_class->switched_from(rq, p);
865

P
Peter Zijlstra 已提交
866
		p->sched_class->switched_to(rq, p);
867
	} else if (oldprio != p->prio || dl_task(p))
P
Peter Zijlstra 已提交
868
		p->sched_class->prio_changed(rq, p, oldprio);
869 870
}

871
void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags)
872 873 874 875 876 877 878 879 880 881
{
	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) {
882
				resched_curr(rq);
883 884 885 886 887 888 889 890 891
				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.
	 */
892
	if (task_on_rq_queued(rq->curr) && test_tsk_need_resched(rq->curr))
893
		rq_clock_skip_update(rq, true);
894 895
}

L
Linus Torvalds 已提交
896
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915
/*
 * 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.
 */
916 917
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
918 919 920 921
{
	lockdep_assert_held(&rq->lock);

	p->on_rq = TASK_ON_RQ_MIGRATING;
922
	dequeue_task(rq, p, DEQUEUE_NOCLOCK);
P
Peter Zijlstra 已提交
923
	set_task_cpu(p, new_cpu);
924
	rq_unlock(rq, rf);
P
Peter Zijlstra 已提交
925 926 927

	rq = cpu_rq(new_cpu);

928
	rq_lock(rq, rf);
P
Peter Zijlstra 已提交
929 930
	BUG_ON(task_cpu(p) != new_cpu);
	enqueue_task(rq, p, 0);
931
	p->on_rq = TASK_ON_RQ_QUEUED;
P
Peter Zijlstra 已提交
932 933 934 935 936 937 938 939 940 941 942
	check_preempt_curr(rq, p, 0);

	return rq;
}

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

/*
I
Ingo Molnar 已提交
943
 * Move (not current) task off this CPU, onto the destination CPU. We're doing
P
Peter Zijlstra 已提交
944 945 946 947 948 949 950
 * 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.
 */
951 952
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
953 954
{
	if (unlikely(!cpu_active(dest_cpu)))
955
		return rq;
P
Peter Zijlstra 已提交
956 957

	/* Affinity changed (again). */
958
	if (!cpumask_test_cpu(dest_cpu, &p->cpus_allowed))
959
		return rq;
P
Peter Zijlstra 已提交
960

961
	update_rq_clock(rq);
962
	rq = move_queued_task(rq, rf, p, dest_cpu);
963 964

	return rq;
P
Peter Zijlstra 已提交
965 966 967 968 969 970 971 972 973 974
}

/*
 * 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;
975 976
	struct task_struct *p = arg->task;
	struct rq *rq = this_rq();
977
	struct rq_flags rf;
P
Peter Zijlstra 已提交
978 979

	/*
I
Ingo Molnar 已提交
980 981
	 * 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
982 983 984 985 986 987 988 989
	 */
	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();
990 991

	raw_spin_lock(&p->pi_lock);
992
	rq_lock(rq, &rf);
993 994 995 996 997
	/*
	 * 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.
	 */
998 999
	if (task_rq(p) == rq) {
		if (task_on_rq_queued(p))
1000
			rq = __migrate_task(rq, &rf, p, arg->dest_cpu);
1001 1002 1003
		else
			p->wake_cpu = arg->dest_cpu;
	}
1004
	rq_unlock(rq, &rf);
1005 1006
	raw_spin_unlock(&p->pi_lock);

P
Peter Zijlstra 已提交
1007 1008 1009 1010
	local_irq_enable();
	return 0;
}

1011 1012 1013 1014 1015
/*
 * 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
1016 1017 1018 1019 1020
{
	cpumask_copy(&p->cpus_allowed, new_mask);
	p->nr_cpus_allowed = cpumask_weight(new_mask);
}

1021 1022
void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
{
1023 1024 1025
	struct rq *rq = task_rq(p);
	bool queued, running;

1026
	lockdep_assert_held(&p->pi_lock);
1027 1028 1029 1030 1031 1032 1033 1034 1035 1036

	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);
1037
		dequeue_task(rq, p, DEQUEUE_SAVE | DEQUEUE_NOCLOCK);
1038 1039 1040 1041
	}
	if (running)
		put_prev_task(rq, p);

1042
	p->sched_class->set_cpus_allowed(p, new_mask);
1043 1044

	if (queued)
1045
		enqueue_task(rq, p, ENQUEUE_RESTORE | ENQUEUE_NOCLOCK);
1046
	if (running)
1047
		set_curr_task(rq, p);
1048 1049
}

P
Peter Zijlstra 已提交
1050 1051 1052 1053 1054 1055 1056 1057 1058
/*
 * 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.
 */
1059 1060
static int __set_cpus_allowed_ptr(struct task_struct *p,
				  const struct cpumask *new_mask, bool check)
P
Peter Zijlstra 已提交
1061
{
1062
	const struct cpumask *cpu_valid_mask = cpu_active_mask;
P
Peter Zijlstra 已提交
1063
	unsigned int dest_cpu;
1064 1065
	struct rq_flags rf;
	struct rq *rq;
P
Peter Zijlstra 已提交
1066 1067
	int ret = 0;

1068
	rq = task_rq_lock(p, &rf);
1069
	update_rq_clock(rq);
P
Peter Zijlstra 已提交
1070

1071 1072 1073 1074 1075 1076 1077
	if (p->flags & PF_KTHREAD) {
		/*
		 * Kernel threads are allowed on online && !active CPUs
		 */
		cpu_valid_mask = cpu_online_mask;
	}

1078 1079 1080 1081 1082 1083 1084 1085 1086
	/*
	 * 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
1087 1088 1089
	if (cpumask_equal(&p->cpus_allowed, new_mask))
		goto out;

1090
	if (!cpumask_intersects(new_mask, cpu_valid_mask)) {
P
Peter Zijlstra 已提交
1091 1092 1093 1094 1095 1096
		ret = -EINVAL;
		goto out;
	}

	do_set_cpus_allowed(p, new_mask);

1097 1098 1099
	if (p->flags & PF_KTHREAD) {
		/*
		 * For kernel threads that do indeed end up on online &&
I
Ingo Molnar 已提交
1100
		 * !active we want to ensure they are strict per-CPU threads.
1101 1102 1103 1104 1105 1106
		 */
		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
1107 1108 1109 1110
	/* 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;

1111
	dest_cpu = cpumask_any_and(cpu_valid_mask, new_mask);
P
Peter Zijlstra 已提交
1112 1113 1114
	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. */
1115
		task_rq_unlock(rq, p, &rf);
P
Peter Zijlstra 已提交
1116 1117 1118
		stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg);
		tlb_migrate_finish(p->mm);
		return 0;
1119 1120 1121 1122 1123
	} else if (task_on_rq_queued(p)) {
		/*
		 * OK, since we're going to drop the lock immediately
		 * afterwards anyway.
		 */
1124
		rq = move_queued_task(rq, &rf, p, dest_cpu);
1125
	}
P
Peter Zijlstra 已提交
1126
out:
1127
	task_rq_unlock(rq, p, &rf);
P
Peter Zijlstra 已提交
1128 1129 1130

	return ret;
}
1131 1132 1133 1134 1135

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
1136 1137
EXPORT_SYMBOL_GPL(set_cpus_allowed_ptr);

I
Ingo Molnar 已提交
1138
void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
I
Ingo Molnar 已提交
1139
{
1140 1141 1142 1143 1144
#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
1145
	WARN_ON_ONCE(p->state != TASK_RUNNING && p->state != TASK_WAKING &&
O
Oleg Nesterov 已提交
1146
			!p->on_rq);
1147

1148 1149 1150 1151 1152 1153 1154 1155 1156
	/*
	 * 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)));

1157
#ifdef CONFIG_LOCKDEP
1158 1159 1160 1161 1162
	/*
	 * 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
1163
	 * see task_group().
1164 1165 1166 1167
	 *
	 * Furthermore, all task_rq users should acquire both locks, see
	 * task_rq_lock().
	 */
1168 1169 1170
	WARN_ON_ONCE(debug_locks && !(lockdep_is_held(&p->pi_lock) ||
				      lockdep_is_held(&task_rq(p)->lock)));
#endif
1171 1172
#endif

1173
	trace_sched_migrate_task(p, new_cpu);
1174

1175
	if (task_cpu(p) != new_cpu) {
1176
		if (p->sched_class->migrate_task_rq)
1177
			p->sched_class->migrate_task_rq(p);
1178
		p->se.nr_migrations++;
1179
		perf_event_task_migrate(p);
1180
	}
I
Ingo Molnar 已提交
1181 1182

	__set_task_cpu(p, new_cpu);
I
Ingo Molnar 已提交
1183 1184
}

1185 1186
static void __migrate_swap_task(struct task_struct *p, int cpu)
{
1187
	if (task_on_rq_queued(p)) {
1188
		struct rq *src_rq, *dst_rq;
1189
		struct rq_flags srf, drf;
1190 1191 1192 1193

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

1194 1195 1196
		rq_pin_lock(src_rq, &srf);
		rq_pin_lock(dst_rq, &drf);

1197
		p->on_rq = TASK_ON_RQ_MIGRATING;
1198 1199 1200
		deactivate_task(src_rq, p, 0);
		set_task_cpu(p, cpu);
		activate_task(dst_rq, p, 0);
1201
		p->on_rq = TASK_ON_RQ_QUEUED;
1202
		check_preempt_curr(dst_rq, p, 0);
1203 1204 1205 1206

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

1207 1208 1209 1210
	} 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
1211
		 * previous CPU our target instead of where it really is.
1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227
		 */
		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;

1228 1229 1230
	if (!cpu_active(arg->src_cpu) || !cpu_active(arg->dst_cpu))
		return -EAGAIN;

1231 1232 1233
	src_rq = cpu_rq(arg->src_cpu);
	dst_rq = cpu_rq(arg->dst_cpu);

1234 1235
	double_raw_lock(&arg->src_task->pi_lock,
			&arg->dst_task->pi_lock);
1236
	double_rq_lock(src_rq, dst_rq);
1237

1238 1239 1240 1241 1242 1243
	if (task_cpu(arg->dst_task) != arg->dst_cpu)
		goto unlock;

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

1244
	if (!cpumask_test_cpu(arg->dst_cpu, &arg->src_task->cpus_allowed))
1245 1246
		goto unlock;

1247
	if (!cpumask_test_cpu(arg->src_cpu, &arg->dst_task->cpus_allowed))
1248 1249 1250 1251 1252 1253 1254 1255 1256
		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);
1257 1258
	raw_spin_unlock(&arg->dst_task->pi_lock);
	raw_spin_unlock(&arg->src_task->pi_lock);
1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280

	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;

1281 1282 1283 1284
	/*
	 * 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.
	 */
1285 1286 1287
	if (!cpu_active(arg.src_cpu) || !cpu_active(arg.dst_cpu))
		goto out;

1288
	if (!cpumask_test_cpu(arg.dst_cpu, &arg.src_task->cpus_allowed))
1289 1290
		goto out;

1291
	if (!cpumask_test_cpu(arg.src_cpu, &arg.dst_task->cpus_allowed))
1292 1293
		goto out;

1294
	trace_sched_swap_numa(cur, arg.src_cpu, p, arg.dst_cpu);
1295 1296 1297 1298 1299 1300
	ret = stop_two_cpus(arg.dst_cpu, arg.src_cpu, migrate_swap_stop, &arg);

out:
	return ret;
}

L
Linus Torvalds 已提交
1301 1302 1303
/*
 * wait_task_inactive - wait for a thread to unschedule.
 *
R
Roland McGrath 已提交
1304 1305 1306 1307 1308 1309 1310
 * 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
1311 1312 1313 1314 1315 1316
 * 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
1317
unsigned long wait_task_inactive(struct task_struct *p, long match_state)
L
Linus Torvalds 已提交
1318
{
1319
	int running, queued;
1320
	struct rq_flags rf;
R
Roland McGrath 已提交
1321
	unsigned long ncsw;
1322
	struct rq *rq;
L
Linus Torvalds 已提交
1323

1324 1325 1326 1327 1328 1329 1330 1331
	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);
1332

1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343
		/*
		 * 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
1344 1345 1346
		while (task_running(rq, p)) {
			if (match_state && unlikely(p->state != match_state))
				return 0;
1347
			cpu_relax();
R
Roland McGrath 已提交
1348
		}
1349

1350 1351 1352 1353 1354
		/*
		 * 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.
		 */
1355
		rq = task_rq_lock(p, &rf);
1356
		trace_sched_wait_task(p);
1357
		running = task_running(rq, p);
1358
		queued = task_on_rq_queued(p);
R
Roland McGrath 已提交
1359
		ncsw = 0;
1360
		if (!match_state || p->state == match_state)
1361
			ncsw = p->nvcsw | LONG_MIN; /* sets MSB */
1362
		task_rq_unlock(rq, p, &rf);
1363

R
Roland McGrath 已提交
1364 1365 1366 1367 1368 1369
		/*
		 * If it changed from the expected state, bail out now.
		 */
		if (unlikely(!ncsw))
			break;

1370 1371 1372 1373 1374 1375 1376 1377 1378 1379
		/*
		 * 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;
		}
1380

1381 1382 1383 1384 1385
		/*
		 * It's not enough that it's not actively running,
		 * it must be off the runqueue _entirely_, and not
		 * preempted!
		 *
1386
		 * So if it was still runnable (but just not actively
1387 1388 1389
		 * running right now), it's preempted, and we should
		 * yield - it could be a while.
		 */
1390
		if (unlikely(queued)) {
T
Thomas Gleixner 已提交
1391
			ktime_t to = NSEC_PER_SEC / HZ;
1392 1393 1394

			set_current_state(TASK_UNINTERRUPTIBLE);
			schedule_hrtimeout(&to, HRTIMER_MODE_REL);
1395 1396
			continue;
		}
1397

1398 1399 1400 1401 1402 1403 1404
		/*
		 * 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
1405 1406

	return ncsw;
L
Linus Torvalds 已提交
1407 1408 1409 1410 1411 1412 1413 1414 1415
}

/***
 * 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
1416
 * NOTE: this function doesn't have to take the runqueue lock,
L
Linus Torvalds 已提交
1417 1418 1419 1420 1421
 * 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.
 */
1422
void kick_process(struct task_struct *p)
L
Linus Torvalds 已提交
1423 1424 1425 1426 1427 1428 1429 1430 1431
{
	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
1432
EXPORT_SYMBOL_GPL(kick_process);
L
Linus Torvalds 已提交
1433

1434
/*
1435
 * ->cpus_allowed is protected by both rq->lock and p->pi_lock
1436 1437 1438 1439 1440 1441 1442
 *
 * 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
1443
 *    CPU isn't yet part of the sched domains, and balancing will not
1444 1445
 *    see it.
 *
I
Ingo Molnar 已提交
1446
 *  - on CPU-down we clear cpu_active() to mask the sched domains and
1447
 *    avoid the load balancer to place new tasks on the to be removed
I
Ingo Molnar 已提交
1448
 *    CPU. Existing tasks will remain running there and will be taken
1449 1450 1451 1452 1453 1454
 *    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.
1455
 */
1456 1457
static int select_fallback_rq(int cpu, struct task_struct *p)
{
1458 1459
	int nid = cpu_to_node(cpu);
	const struct cpumask *nodemask = NULL;
1460 1461
	enum { cpuset, possible, fail } state = cpuset;
	int dest_cpu;
1462

1463
	/*
I
Ingo Molnar 已提交
1464 1465 1466
	 * 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.
1467 1468 1469 1470 1471 1472 1473 1474
	 */
	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;
1475
			if (cpumask_test_cpu(dest_cpu, &p->cpus_allowed))
1476 1477
				return dest_cpu;
		}
1478
	}
1479

1480 1481
	for (;;) {
		/* Any allowed, online CPU? */
1482
		for_each_cpu(dest_cpu, &p->cpus_allowed) {
1483 1484 1485
			if (!(p->flags & PF_KTHREAD) && !cpu_active(dest_cpu))
				continue;
			if (!cpu_online(dest_cpu))
1486 1487 1488
				continue;
			goto out;
		}
1489

1490
		/* No more Mr. Nice Guy. */
1491 1492
		switch (state) {
		case cpuset:
1493 1494 1495 1496 1497
			if (IS_ENABLED(CONFIG_CPUSETS)) {
				cpuset_cpus_allowed_fallback(p);
				state = possible;
				break;
			}
I
Ingo Molnar 已提交
1498
			/* Fall-through */
1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517
		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()) {
1518
			printk_deferred("process %d (%s) no longer affine to cpu%d\n",
1519 1520
					task_pid_nr(p), p->comm, cpu);
		}
1521 1522 1523 1524 1525
	}

	return dest_cpu;
}

1526
/*
1527
 * The caller (fork, wakeup) owns p->pi_lock, ->cpus_allowed is stable.
1528
 */
1529
static inline
1530
int select_task_rq(struct task_struct *p, int cpu, int sd_flags, int wake_flags)
1531
{
1532 1533
	lockdep_assert_held(&p->pi_lock);

1534
	if (p->nr_cpus_allowed > 1)
1535
		cpu = p->sched_class->select_task_rq(p, cpu, sd_flags, wake_flags);
1536
	else
1537
		cpu = cpumask_any(&p->cpus_allowed);
1538 1539 1540 1541

	/*
	 * 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
1542
	 * CPU.
1543 1544 1545 1546 1547 1548
	 *
	 * 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 ]
	 */
1549
	if (unlikely(!cpumask_test_cpu(cpu, &p->cpus_allowed) ||
P
Peter Zijlstra 已提交
1550
		     !cpu_online(cpu)))
1551
		cpu = select_fallback_rq(task_cpu(p), p);
1552 1553

	return cpu;
1554
}
1555 1556 1557 1558 1559 1560

static void update_avg(u64 *avg, u64 sample)
{
	s64 diff = sample - *avg;
	*avg += diff >> 3;
}
1561

1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591
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;
	}
}

1592 1593 1594 1595 1596 1597 1598 1599
#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
1600
#endif /* CONFIG_SMP */
1601

P
Peter Zijlstra 已提交
1602
static void
1603
ttwu_stat(struct task_struct *p, int cpu, int wake_flags)
T
Tejun Heo 已提交
1604
{
1605
	struct rq *rq;
1606

1607 1608 1609 1610
	if (!schedstat_enabled())
		return;

	rq = this_rq();
P
Peter Zijlstra 已提交
1611

1612 1613
#ifdef CONFIG_SMP
	if (cpu == rq->cpu) {
1614 1615
		schedstat_inc(rq->ttwu_local);
		schedstat_inc(p->se.statistics.nr_wakeups_local);
P
Peter Zijlstra 已提交
1616 1617 1618
	} else {
		struct sched_domain *sd;

1619
		schedstat_inc(p->se.statistics.nr_wakeups_remote);
1620
		rcu_read_lock();
1621
		for_each_domain(rq->cpu, sd) {
P
Peter Zijlstra 已提交
1622
			if (cpumask_test_cpu(cpu, sched_domain_span(sd))) {
1623
				schedstat_inc(sd->ttwu_wake_remote);
P
Peter Zijlstra 已提交
1624 1625 1626
				break;
			}
		}
1627
		rcu_read_unlock();
P
Peter Zijlstra 已提交
1628
	}
1629 1630

	if (wake_flags & WF_MIGRATED)
1631
		schedstat_inc(p->se.statistics.nr_wakeups_migrate);
P
Peter Zijlstra 已提交
1632 1633
#endif /* CONFIG_SMP */

1634 1635
	schedstat_inc(rq->ttwu_count);
	schedstat_inc(p->se.statistics.nr_wakeups);
P
Peter Zijlstra 已提交
1636 1637

	if (wake_flags & WF_SYNC)
1638
		schedstat_inc(p->se.statistics.nr_wakeups_sync);
P
Peter Zijlstra 已提交
1639 1640
}

1641
static inline void ttwu_activate(struct rq *rq, struct task_struct *p, int en_flags)
P
Peter Zijlstra 已提交
1642
{
T
Tejun Heo 已提交
1643
	activate_task(rq, p, en_flags);
1644
	p->on_rq = TASK_ON_RQ_QUEUED;
1645

I
Ingo Molnar 已提交
1646
	/* If a worker is waking up, notify the workqueue: */
1647 1648
	if (p->flags & PF_WQ_WORKER)
		wq_worker_waking_up(p, cpu_of(rq));
T
Tejun Heo 已提交
1649 1650
}

1651 1652 1653
/*
 * Mark the task runnable and perform wakeup-preemption.
 */
1654
static void ttwu_do_wakeup(struct rq *rq, struct task_struct *p, int wake_flags,
1655
			   struct rq_flags *rf)
T
Tejun Heo 已提交
1656 1657 1658
{
	check_preempt_curr(rq, p, wake_flags);
	p->state = TASK_RUNNING;
1659 1660
	trace_sched_wakeup(p);

T
Tejun Heo 已提交
1661
#ifdef CONFIG_SMP
1662 1663
	if (p->sched_class->task_woken) {
		/*
1664 1665
		 * Our task @p is fully woken up and running; so its safe to
		 * drop the rq->lock, hereafter rq is only used for statistics.
1666
		 */
1667
		rq_unpin_lock(rq, rf);
T
Tejun Heo 已提交
1668
		p->sched_class->task_woken(rq, p);
1669
		rq_repin_lock(rq, rf);
1670
	}
T
Tejun Heo 已提交
1671

1672
	if (rq->idle_stamp) {
1673
		u64 delta = rq_clock(rq) - rq->idle_stamp;
1674
		u64 max = 2*rq->max_idle_balance_cost;
T
Tejun Heo 已提交
1675

1676 1677 1678
		update_avg(&rq->avg_idle, delta);

		if (rq->avg_idle > max)
T
Tejun Heo 已提交
1679
			rq->avg_idle = max;
1680

T
Tejun Heo 已提交
1681 1682 1683 1684 1685
		rq->idle_stamp = 0;
	}
#endif
}

1686
static void
1687
ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags,
1688
		 struct rq_flags *rf)
1689
{
1690
	int en_flags = ENQUEUE_WAKEUP | ENQUEUE_NOCLOCK;
1691

1692 1693
	lockdep_assert_held(&rq->lock);

1694 1695 1696
#ifdef CONFIG_SMP
	if (p->sched_contributes_to_load)
		rq->nr_uninterruptible--;
1697 1698

	if (wake_flags & WF_MIGRATED)
1699
		en_flags |= ENQUEUE_MIGRATED;
1700 1701
#endif

1702
	ttwu_activate(rq, p, en_flags);
1703
	ttwu_do_wakeup(rq, p, wake_flags, rf);
1704 1705 1706 1707 1708 1709 1710 1711 1712 1713
}

/*
 * 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)
{
1714
	struct rq_flags rf;
1715 1716 1717
	struct rq *rq;
	int ret = 0;

1718
	rq = __task_rq_lock(p, &rf);
1719
	if (task_on_rq_queued(p)) {
1720 1721
		/* check_preempt_curr() may use rq clock */
		update_rq_clock(rq);
1722
		ttwu_do_wakeup(rq, p, wake_flags, &rf);
1723 1724
		ret = 1;
	}
1725
	__task_rq_unlock(rq, &rf);
1726 1727 1728 1729

	return ret;
}

1730
#ifdef CONFIG_SMP
1731
void sched_ttwu_pending(void)
1732 1733
{
	struct rq *rq = this_rq();
P
Peter Zijlstra 已提交
1734
	struct llist_node *llist = llist_del_all(&rq->wake_list);
1735
	struct task_struct *p, *t;
1736
	struct rq_flags rf;
1737

1738 1739 1740
	if (!llist)
		return;

1741
	rq_lock_irqsave(rq, &rf);
1742
	update_rq_clock(rq);
1743

1744 1745
	llist_for_each_entry_safe(p, t, llist, wake_entry)
		ttwu_do_activate(rq, p, p->sched_remote_wakeup ? WF_MIGRATED : 0, &rf);
1746

1747
	rq_unlock_irqrestore(rq, &rf);
1748 1749 1750 1751
}

void scheduler_ipi(void)
{
1752 1753 1754 1755 1756
	/*
	 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
	 * TIF_NEED_RESCHED remotely (for the first time) will also send
	 * this IPI.
	 */
1757
	preempt_fold_need_resched();
1758

1759
	if (llist_empty(&this_rq()->wake_list) && !got_nohz_idle_kick())
1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775
		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
1776
	sched_ttwu_pending();
1777 1778 1779 1780

	/*
	 * Check if someone kicked us for doing the nohz idle load balance.
	 */
1781
	if (unlikely(got_nohz_idle_kick())) {
1782
		this_rq()->idle_balance = 1;
1783
		raise_softirq_irqoff(SCHED_SOFTIRQ);
1784
	}
1785
	irq_exit();
1786 1787
}

P
Peter Zijlstra 已提交
1788
static void ttwu_queue_remote(struct task_struct *p, int cpu, int wake_flags)
1789
{
1790 1791
	struct rq *rq = cpu_rq(cpu);

P
Peter Zijlstra 已提交
1792 1793
	p->sched_remote_wakeup = !!(wake_flags & WF_MIGRATED);

1794 1795 1796 1797 1798 1799
	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);
	}
1800
}
1801

1802 1803 1804
void wake_up_if_idle(int cpu)
{
	struct rq *rq = cpu_rq(cpu);
1805
	struct rq_flags rf;
1806

1807 1808 1809 1810
	rcu_read_lock();

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

	if (set_nr_if_polling(rq->idle)) {
		trace_sched_wake_idle_without_ipi(cpu);
	} else {
1815
		rq_lock_irqsave(rq, &rf);
1816 1817
		if (is_idle_task(rq->curr))
			smp_send_reschedule(cpu);
I
Ingo Molnar 已提交
1818
		/* Else CPU is not idle, do nothing here: */
1819
		rq_unlock_irqrestore(rq, &rf);
1820
	}
1821 1822 1823

out:
	rcu_read_unlock();
1824 1825
}

1826
bool cpus_share_cache(int this_cpu, int that_cpu)
1827 1828 1829
{
	return per_cpu(sd_llc_id, this_cpu) == per_cpu(sd_llc_id, that_cpu);
}
1830
#endif /* CONFIG_SMP */
1831

1832
static void ttwu_queue(struct task_struct *p, int cpu, int wake_flags)
1833 1834
{
	struct rq *rq = cpu_rq(cpu);
1835
	struct rq_flags rf;
1836

1837
#if defined(CONFIG_SMP)
1838
	if (sched_feat(TTWU_QUEUE) && !cpus_share_cache(smp_processor_id(), cpu)) {
I
Ingo Molnar 已提交
1839
		sched_clock_cpu(cpu); /* Sync clocks across CPUs */
P
Peter Zijlstra 已提交
1840
		ttwu_queue_remote(p, cpu, wake_flags);
1841 1842 1843 1844
		return;
	}
#endif

1845
	rq_lock(rq, &rf);
1846
	update_rq_clock(rq);
1847
	ttwu_do_activate(rq, p, wake_flags, &rf);
1848
	rq_unlock(rq, &rf);
T
Tejun Heo 已提交
1849 1850
}

1851 1852 1853 1854 1855 1856
/*
 * Notes on Program-Order guarantees on SMP systems.
 *
 *  MIGRATION
 *
 * The basic program-order guarantee on SMP systems is that when a task [t]
I
Ingo Molnar 已提交
1857 1858
 * migrates, all its activity on its old CPU [c0] happens-before any subsequent
 * execution on its new CPU [c1].
1859 1860 1861 1862 1863 1864 1865 1866 1867 1868
 *
 * For migration (of runnable tasks) this is provided by the following means:
 *
 *  A) UNLOCK of the rq(c0)->lock scheduling out task t
 *  B) migration for t is required to synchronize *both* rq(c0)->lock and
 *     rq(c1)->lock (if not at the same time, then in that order).
 *  C) LOCK of the rq(c1)->lock scheduling in task
 *
 * Transitivity guarantees that B happens after A and C after B.
 * Note: we only require RCpc transitivity.
I
Ingo Molnar 已提交
1869
 * Note: the CPU doing B need not be c0 or c1
1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900
 *
 * Example:
 *
 *   CPU0            CPU1            CPU2
 *
 *   LOCK rq(0)->lock
 *   sched-out X
 *   sched-in Y
 *   UNLOCK rq(0)->lock
 *
 *                                   LOCK rq(0)->lock // orders against CPU0
 *                                   dequeue X
 *                                   UNLOCK rq(0)->lock
 *
 *                                   LOCK rq(1)->lock
 *                                   enqueue X
 *                                   UNLOCK rq(1)->lock
 *
 *                   LOCK rq(1)->lock // orders against CPU2
 *                   sched-out Z
 *                   sched-in X
 *                   UNLOCK rq(1)->lock
 *
 *
 *  BLOCKING -- aka. SLEEP + WAKEUP
 *
 * For blocking we (obviously) need to provide the same guarantee as for
 * migration. However the means are completely different as there is no lock
 * chain to provide order. Instead we do:
 *
 *   1) smp_store_release(X->on_cpu, 0)
1901
 *   2) smp_cond_load_acquire(!X->on_cpu)
1902 1903 1904 1905 1906 1907 1908 1909 1910 1911
 *
 * Example:
 *
 *   CPU0 (schedule)  CPU1 (try_to_wake_up) CPU2 (schedule)
 *
 *   LOCK rq(0)->lock LOCK X->pi_lock
 *   dequeue X
 *   sched-out X
 *   smp_store_release(X->on_cpu, 0);
 *
1912
 *                    smp_cond_load_acquire(&X->on_cpu, !VAL);
1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937
 *                    X->state = WAKING
 *                    set_task_cpu(X,2)
 *
 *                    LOCK rq(2)->lock
 *                    enqueue X
 *                    X->state = RUNNING
 *                    UNLOCK rq(2)->lock
 *
 *                                          LOCK rq(2)->lock // orders against CPU1
 *                                          sched-out Z
 *                                          sched-in X
 *                                          UNLOCK rq(2)->lock
 *
 *                    UNLOCK X->pi_lock
 *   UNLOCK rq(0)->lock
 *
 *
 * However; for wakeups there is a second guarantee we must provide, namely we
 * must observe the state that lead to our wakeup. That is, not only must our
 * task observe its own prior state, it must also observe the stores prior to
 * its wakeup.
 *
 * This means that any means of doing remote wakeups must order the CPU doing
 * the wakeup against the CPU the task is going to end up running on. This,
 * however, is already required for the regular Program-Order guarantee above,
1938
 * since the waking CPU is the one issueing the ACQUIRE (smp_cond_load_acquire).
1939 1940 1941
 *
 */

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

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

1975 1976
	trace_sched_waking(p);

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

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

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

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

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

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

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

#else /* CONFIG_SMP */

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

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

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

	return success;
}

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

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

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

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

T
Tejun Heo 已提交
2100
	if (!(p->state & TASK_NORMAL))
2101
		goto out;
T
Tejun Heo 已提交
2102

2103 2104
	trace_sched_waking(p);

2105 2106 2107 2108 2109
	if (!task_on_rq_queued(p)) {
		if (p->in_iowait) {
			delayacct_blkio_end();
			atomic_dec(&rq->nr_iowait);
		}
2110
		ttwu_activate(rq, p, ENQUEUE_WAKEUP | ENQUEUE_NOCLOCK);
2111
	}
P
Peter Zijlstra 已提交
2112

2113
	ttwu_do_wakeup(rq, p, 0, rf);
2114
	ttwu_stat(p, smp_processor_id(), 0);
2115 2116
out:
	raw_spin_unlock(&p->pi_lock);
T
Tejun Heo 已提交
2117 2118
}

2119 2120 2121 2122 2123
/**
 * 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
2124 2125 2126
 * processes.
 *
 * Return: 1 if the process was woken up, 0 if it was already running.
2127 2128 2129 2130
 *
 * 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.
 */
2131
int wake_up_process(struct task_struct *p)
L
Linus Torvalds 已提交
2132
{
2133
	return try_to_wake_up(p, TASK_NORMAL, 0);
L
Linus Torvalds 已提交
2134 2135 2136
}
EXPORT_SYMBOL(wake_up_process);

2137
int wake_up_state(struct task_struct *p, unsigned int state)
L
Linus Torvalds 已提交
2138 2139 2140 2141
{
	return try_to_wake_up(p, state, 0);
}

2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153
/*
 * 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;
2154
	dl_se->dl_density = 0;
2155 2156 2157

	dl_se->dl_throttled = 0;
	dl_se->dl_yielded = 0;
2158
	dl_se->dl_non_contending = 0;
2159 2160
}

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

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

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

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

2188
	RB_CLEAR_NODE(&p->dl.rb_node);
2189
	init_dl_task_timer(&p->dl);
2190
	init_dl_inactive_task_timer(&p->dl);
2191
	__dl_clear_params(p);
2192

P
Peter Zijlstra 已提交
2193
	INIT_LIST_HEAD(&p->rt.run_list);
2194 2195 2196 2197
	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
2198

2199 2200 2201
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&p->preempt_notifiers);
#endif
2202 2203 2204

#ifdef CONFIG_NUMA_BALANCING
	if (p->mm && atomic_read(&p->mm->mm_users) == 1) {
2205
		p->mm->numa_next_scan = jiffies + msecs_to_jiffies(sysctl_numa_balancing_scan_delay);
2206 2207 2208
		p->mm->numa_scan_seq = 0;
	}

2209 2210 2211 2212 2213
	if (clone_flags & CLONE_VM)
		p->numa_preferred_nid = current->numa_preferred_nid;
	else
		p->numa_preferred_nid = -1;

2214 2215
	p->node_stamp = 0ULL;
	p->numa_scan_seq = p->mm ? p->mm->numa_scan_seq : 0;
2216
	p->numa_scan_period = sysctl_numa_balancing_scan_delay;
2217
	p->numa_work.next = &p->numa_work;
2218
	p->numa_faults = NULL;
2219 2220
	p->last_task_numa_placement = 0;
	p->last_sum_exec_runtime = 0;
2221 2222

	p->numa_group = NULL;
2223
#endif /* CONFIG_NUMA_BALANCING */
I
Ingo Molnar 已提交
2224 2225
}

2226 2227
DEFINE_STATIC_KEY_FALSE(sched_numa_balancing);

2228
#ifdef CONFIG_NUMA_BALANCING
2229

2230 2231 2232
void set_numabalancing_state(bool enabled)
{
	if (enabled)
2233
		static_branch_enable(&sched_numa_balancing);
2234
	else
2235
		static_branch_disable(&sched_numa_balancing);
2236
}
2237 2238 2239 2240 2241 2242 2243

#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;
2244
	int state = static_branch_likely(&sched_numa_balancing);
2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259

	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
2260

2261 2262
#ifdef CONFIG_SCHEDSTATS

2263
DEFINE_STATIC_KEY_FALSE(sched_schedstats);
2264
static bool __initdata __sched_schedstats = false;
2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287

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;

2288 2289 2290 2291 2292
	/*
	 * 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.
	 */
2293
	if (!strcmp(str, "enable")) {
2294
		__sched_schedstats = true;
2295 2296
		ret = 1;
	} else if (!strcmp(str, "disable")) {
2297
		__sched_schedstats = false;
2298 2299 2300 2301 2302 2303 2304 2305 2306 2307
		ret = 1;
	}
out:
	if (!ret)
		pr_warn("Unable to parse schedstats=\n");

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

2308 2309 2310 2311 2312
static void __init init_schedstats(void)
{
	set_schedstats(__sched_schedstats);
}

2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332
#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;
}
2333 2334 2335 2336
#endif /* CONFIG_PROC_SYSCTL */
#else  /* !CONFIG_SCHEDSTATS */
static inline void init_schedstats(void) {}
#endif /* CONFIG_SCHEDSTATS */
I
Ingo Molnar 已提交
2337 2338 2339 2340

/*
 * fork()/clone()-time setup:
 */
2341
int sched_fork(unsigned long clone_flags, struct task_struct *p)
I
Ingo Molnar 已提交
2342
{
2343
	unsigned long flags;
I
Ingo Molnar 已提交
2344 2345
	int cpu = get_cpu();

2346
	__sched_fork(clone_flags, p);
2347
	/*
2348
	 * We mark the process as NEW here. This guarantees that
2349 2350 2351
	 * nobody will actually run it, and a signal or other external
	 * event cannot wake it up and insert it on the runqueue either.
	 */
2352
	p->state = TASK_NEW;
I
Ingo Molnar 已提交
2353

2354 2355 2356 2357 2358
	/*
	 * Make sure we do not leak PI boosting priority to the child.
	 */
	p->prio = current->normal_prio;

2359 2360 2361 2362
	/*
	 * Revert to default priority/policy on fork if requested.
	 */
	if (unlikely(p->sched_reset_on_fork)) {
2363
		if (task_has_dl_policy(p) || task_has_rt_policy(p)) {
2364
			p->policy = SCHED_NORMAL;
2365
			p->static_prio = NICE_TO_PRIO(0);
2366 2367 2368 2369 2370 2371
			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);
2372

2373 2374 2375 2376 2377 2378
		/*
		 * We don't need the reset flag anymore after the fork. It has
		 * fulfilled its duty:
		 */
		p->sched_reset_on_fork = 0;
	}
2379

2380 2381 2382 2383 2384 2385
	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
2386
		p->sched_class = &fair_sched_class;
2387
	}
2388

2389
	init_entity_runnable_average(&p->se);
P
Peter Zijlstra 已提交
2390

2391 2392 2393 2394 2395 2396 2397
	/*
	 * 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.
	 */
2398
	raw_spin_lock_irqsave(&p->pi_lock, flags);
2399
	/*
I
Ingo Molnar 已提交
2400
	 * We're setting the CPU for the first time, we don't migrate,
2401 2402 2403 2404 2405
	 * so use __set_task_cpu().
	 */
	__set_task_cpu(p, cpu);
	if (p->sched_class->task_fork)
		p->sched_class->task_fork(p);
2406
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
2407

2408
#ifdef CONFIG_SCHED_INFO
I
Ingo Molnar 已提交
2409
	if (likely(sched_info_on()))
2410
		memset(&p->sched_info, 0, sizeof(p->sched_info));
L
Linus Torvalds 已提交
2411
#endif
P
Peter Zijlstra 已提交
2412 2413
#if defined(CONFIG_SMP)
	p->on_cpu = 0;
2414
#endif
2415
	init_task_preempt_count(p);
2416
#ifdef CONFIG_SMP
2417
	plist_node_init(&p->pushable_tasks, MAX_PRIO);
2418
	RB_CLEAR_NODE(&p->pushable_dl_tasks);
2419
#endif
2420

N
Nick Piggin 已提交
2421
	put_cpu();
2422
	return 0;
L
Linus Torvalds 已提交
2423 2424
}

2425 2426 2427
unsigned long to_ratio(u64 period, u64 runtime)
{
	if (runtime == RUNTIME_INF)
2428
		return BW_UNIT;
2429 2430 2431 2432 2433 2434 2435 2436 2437

	/*
	 * 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;

2438
	return div64_u64(runtime << BW_SHIFT, period);
2439 2440 2441 2442 2443
}

#ifdef CONFIG_SMP
inline struct dl_bw *dl_bw_of(int i)
{
2444 2445
	RCU_LOCKDEP_WARN(!rcu_read_lock_sched_held(),
			 "sched RCU must be held");
2446 2447 2448
	return &cpu_rq(i)->rd->dl_bw;
}

2449
inline int dl_bw_cpus(int i)
2450
{
2451 2452 2453
	struct root_domain *rd = cpu_rq(i)->rd;
	int cpus = 0;

2454 2455
	RCU_LOCKDEP_WARN(!rcu_read_lock_sched_held(),
			 "sched RCU must be held");
2456 2457 2458 2459
	for_each_cpu_and(i, rd->span, cpu_active_mask)
		cpus++;

	return cpus;
2460 2461 2462 2463 2464 2465 2466
}
#else
inline struct dl_bw *dl_bw_of(int i)
{
	return &cpu_rq(i)->dl.dl_bw;
}

2467
inline int dl_bw_cpus(int i)
2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485
{
	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));
2486
	u64 period = attr->sched_period ?: attr->sched_deadline;
2487 2488
	u64 runtime = attr->sched_runtime;
	u64 new_bw = dl_policy(policy) ? to_ratio(period, runtime) : 0;
2489
	int cpus, err = -1;
2490

2491 2492
	/* !deadline task may carry old deadline bandwidth */
	if (new_bw == p->dl.dl_bw && task_has_dl_policy(p))
2493 2494 2495 2496 2497 2498 2499 2500
		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);
2501
	cpus = dl_bw_cpus(task_cpu(p));
2502 2503
	if (dl_policy(policy) && !task_has_dl_policy(p) &&
	    !__dl_overflow(dl_b, cpus, 0, new_bw)) {
2504
		if (hrtimer_active(&p->dl.inactive_timer))
2505 2506
			__dl_clear(dl_b, p->dl.dl_bw, cpus);
		__dl_add(dl_b, new_bw, cpus);
2507 2508 2509
		err = 0;
	} else if (dl_policy(policy) && task_has_dl_policy(p) &&
		   !__dl_overflow(dl_b, cpus, p->dl.dl_bw, new_bw)) {
2510 2511 2512 2513 2514 2515 2516
		/*
		 * 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.
		 */
2517 2518
		__dl_clear(dl_b, p->dl.dl_bw, cpus);
		__dl_add(dl_b, new_bw, cpus);
2519
		dl_change_utilization(p, new_bw);
2520 2521
		err = 0;
	} else if (!dl_policy(policy) && task_has_dl_policy(p)) {
2522 2523 2524 2525 2526
		/*
		 * Do not decrease the total deadline utilization here,
		 * switched_from_dl() will take care to do it at the correct
		 * (0-lag) time.
		 */
2527 2528 2529 2530 2531 2532 2533 2534 2535
		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
2536 2537 2538 2539 2540 2541 2542
/*
 * 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.
 */
2543
void wake_up_new_task(struct task_struct *p)
L
Linus Torvalds 已提交
2544
{
2545
	struct rq_flags rf;
I
Ingo Molnar 已提交
2546
	struct rq *rq;
2547

2548
	raw_spin_lock_irqsave(&p->pi_lock, rf.flags);
2549
	p->state = TASK_RUNNING;
2550 2551 2552 2553
#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
2554
	 *  - any previously selected CPU might disappear through hotplug
2555 2556 2557
	 *
	 * Use __set_task_cpu() to avoid calling sched_class::migrate_task_rq,
	 * as we're not fully set-up yet.
2558
	 */
2559
	__set_task_cpu(p, select_task_rq(p, task_cpu(p), SD_BALANCE_FORK, 0));
2560
#endif
2561
	rq = __task_rq_lock(p, &rf);
2562
	update_rq_clock(rq);
2563
	post_init_entity_util_avg(&p->se);
2564

2565
	activate_task(rq, p, ENQUEUE_NOCLOCK);
2566
	p->on_rq = TASK_ON_RQ_QUEUED;
2567
	trace_sched_wakeup_new(p);
P
Peter Zijlstra 已提交
2568
	check_preempt_curr(rq, p, WF_FORK);
2569
#ifdef CONFIG_SMP
2570 2571 2572 2573 2574
	if (p->sched_class->task_woken) {
		/*
		 * Nothing relies on rq->lock after this, so its fine to
		 * drop it.
		 */
2575
		rq_unpin_lock(rq, &rf);
2576
		p->sched_class->task_woken(rq, p);
2577
		rq_repin_lock(rq, &rf);
2578
	}
2579
#endif
2580
	task_rq_unlock(rq, p, &rf);
L
Linus Torvalds 已提交
2581 2582
}

2583 2584
#ifdef CONFIG_PREEMPT_NOTIFIERS

2585 2586
static struct static_key preempt_notifier_key = STATIC_KEY_INIT_FALSE;

2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598
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);

2599
/**
2600
 * preempt_notifier_register - tell me when current is being preempted & rescheduled
R
Randy Dunlap 已提交
2601
 * @notifier: notifier struct to register
2602 2603 2604
 */
void preempt_notifier_register(struct preempt_notifier *notifier)
{
2605 2606 2607
	if (!static_key_false(&preempt_notifier_key))
		WARN(1, "registering preempt_notifier while notifiers disabled\n");

2608 2609 2610 2611 2612 2613
	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
2614
 * @notifier: notifier struct to unregister
2615
 *
2616
 * This is *not* safe to call from within a preemption notifier.
2617 2618 2619 2620 2621 2622 2623
 */
void preempt_notifier_unregister(struct preempt_notifier *notifier)
{
	hlist_del(&notifier->link);
}
EXPORT_SYMBOL_GPL(preempt_notifier_unregister);

2624
static void __fire_sched_in_preempt_notifiers(struct task_struct *curr)
2625 2626 2627
{
	struct preempt_notifier *notifier;

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

2632 2633 2634 2635 2636 2637
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);
}

2638
static void
2639 2640
__fire_sched_out_preempt_notifiers(struct task_struct *curr,
				   struct task_struct *next)
2641 2642 2643
{
	struct preempt_notifier *notifier;

2644
	hlist_for_each_entry(notifier, &curr->preempt_notifiers, link)
2645 2646 2647
		notifier->ops->sched_out(notifier, next);
}

2648 2649 2650 2651 2652 2653 2654 2655
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);
}

2656
#else /* !CONFIG_PREEMPT_NOTIFIERS */
2657

2658
static inline void fire_sched_in_preempt_notifiers(struct task_struct *curr)
2659 2660 2661
{
}

2662
static inline void
2663 2664 2665 2666 2667
fire_sched_out_preempt_notifiers(struct task_struct *curr,
				 struct task_struct *next)
{
}

2668
#endif /* CONFIG_PREEMPT_NOTIFIERS */
2669

2670 2671 2672
/**
 * prepare_task_switch - prepare to switch tasks
 * @rq: the runqueue preparing to switch
R
Randy Dunlap 已提交
2673
 * @prev: the current task that is being switched out
2674 2675 2676 2677 2678 2679 2680 2681 2682
 * @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.
 */
2683 2684 2685
static inline void
prepare_task_switch(struct rq *rq, struct task_struct *prev,
		    struct task_struct *next)
2686
{
2687
	sched_info_switch(rq, prev, next);
2688
	perf_event_task_sched_out(prev, next);
2689
	fire_sched_out_preempt_notifiers(prev, next);
2690 2691 2692 2693
	prepare_lock_switch(rq, next);
	prepare_arch_switch(next);
}

L
Linus Torvalds 已提交
2694 2695 2696 2697
/**
 * finish_task_switch - clean up after a task-switch
 * @prev: the thread we just switched away from.
 *
2698 2699 2700 2701
 * 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
2702 2703
 *
 * Note that we may have delayed dropping an mm in context_switch(). If
I
Ingo Molnar 已提交
2704
 * so, we finish that here outside of the runqueue lock. (Doing it
L
Linus Torvalds 已提交
2705 2706
 * with the lock held can cause deadlocks; see schedule() for
 * details.)
2707 2708 2709 2710 2711
 *
 * 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
2712
 */
2713
static struct rq *finish_task_switch(struct task_struct *prev)
L
Linus Torvalds 已提交
2714 2715
	__releases(rq->lock)
{
2716
	struct rq *rq = this_rq();
L
Linus Torvalds 已提交
2717
	struct mm_struct *mm = rq->prev_mm;
O
Oleg Nesterov 已提交
2718
	long prev_state;
L
Linus Torvalds 已提交
2719

2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730
	/*
	 * 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.
	 */
2731 2732 2733 2734
	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);
2735

L
Linus Torvalds 已提交
2736 2737 2738 2739
	rq->prev_mm = NULL;

	/*
	 * A task struct has one reference for the use as "current".
2740
	 * If a task dies, then it sets TASK_DEAD in tsk->state and calls
O
Oleg Nesterov 已提交
2741 2742
	 * schedule one last time. The schedule call will never return, and
	 * the scheduled task must drop that reference.
2743 2744 2745 2746 2747
	 *
	 * 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
2748
	 */
O
Oleg Nesterov 已提交
2749
	prev_state = prev->state;
2750
	vtime_task_switch(prev);
2751
	perf_event_task_sched_in(prev, current);
2752
	finish_lock_switch(rq, prev);
2753
	finish_arch_post_lock_switch();
S
Steven Rostedt 已提交
2754

2755
	fire_sched_in_preempt_notifiers(current);
L
Linus Torvalds 已提交
2756 2757
	if (mm)
		mmdrop(mm);
2758
	if (unlikely(prev_state == TASK_DEAD)) {
2759 2760 2761
		if (prev->sched_class->task_dead)
			prev->sched_class->task_dead(prev);

2762 2763 2764
		/*
		 * Remove function-return probe instances associated with this
		 * task and put them back on the free list.
I
Ingo Molnar 已提交
2765
		 */
2766
		kprobe_flush_task(prev);
2767 2768 2769 2770

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

L
Linus Torvalds 已提交
2771
		put_task_struct(prev);
2772
	}
2773

2774
	tick_nohz_task_switch();
2775
	return rq;
L
Linus Torvalds 已提交
2776 2777
}

2778 2779 2780
#ifdef CONFIG_SMP

/* rq->lock is NOT held, but preemption is disabled */
2781
static void __balance_callback(struct rq *rq)
2782
{
2783 2784 2785
	struct callback_head *head, *next;
	void (*func)(struct rq *rq);
	unsigned long flags;
2786

2787 2788 2789 2790 2791 2792 2793 2794
	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;
2795

2796
		func(rq);
2797
	}
2798 2799 2800 2801 2802 2803 2804
	raw_spin_unlock_irqrestore(&rq->lock, flags);
}

static inline void balance_callback(struct rq *rq)
{
	if (unlikely(rq->balance_callback))
		__balance_callback(rq);
2805 2806 2807
}

#else
2808

2809
static inline void balance_callback(struct rq *rq)
2810
{
L
Linus Torvalds 已提交
2811 2812
}

2813 2814
#endif

L
Linus Torvalds 已提交
2815 2816 2817 2818
/**
 * schedule_tail - first thing a freshly forked thread must call.
 * @prev: the thread we just switched away from.
 */
2819
asmlinkage __visible void schedule_tail(struct task_struct *prev)
L
Linus Torvalds 已提交
2820 2821
	__releases(rq->lock)
{
2822
	struct rq *rq;
2823

2824 2825 2826 2827 2828 2829 2830 2831 2832
	/*
	 * 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).
	 */

2833
	rq = finish_task_switch(prev);
2834
	balance_callback(rq);
2835
	preempt_enable();
2836

L
Linus Torvalds 已提交
2837
	if (current->set_child_tid)
2838
		put_user(task_pid_vnr(current), current->set_child_tid);
L
Linus Torvalds 已提交
2839 2840 2841
}

/*
2842
 * context_switch - switch to the new MM and the new thread's register state.
L
Linus Torvalds 已提交
2843
 */
2844
static __always_inline struct rq *
2845
context_switch(struct rq *rq, struct task_struct *prev,
2846
	       struct task_struct *next, struct rq_flags *rf)
L
Linus Torvalds 已提交
2847
{
I
Ingo Molnar 已提交
2848
	struct mm_struct *mm, *oldmm;
L
Linus Torvalds 已提交
2849

2850
	prepare_task_switch(rq, prev, next);
2851

I
Ingo Molnar 已提交
2852 2853
	mm = next->mm;
	oldmm = prev->active_mm;
2854 2855 2856 2857 2858
	/*
	 * For paravirt, this is coupled with an exit in switch_to to
	 * combine the page table reload and the switch backend into
	 * one hypercall.
	 */
2859
	arch_start_context_switch(prev);
2860

2861
	if (!mm) {
L
Linus Torvalds 已提交
2862
		next->active_mm = oldmm;
V
Vegard Nossum 已提交
2863
		mmgrab(oldmm);
L
Linus Torvalds 已提交
2864 2865
		enter_lazy_tlb(oldmm, next);
	} else
2866
		switch_mm_irqs_off(oldmm, mm, next);
L
Linus Torvalds 已提交
2867

2868
	if (!prev->mm) {
L
Linus Torvalds 已提交
2869 2870 2871
		prev->active_mm = NULL;
		rq->prev_mm = oldmm;
	}
2872

2873
	rq->clock_update_flags &= ~(RQCF_ACT_SKIP|RQCF_REQ_SKIP);
2874

2875 2876 2877 2878 2879 2880
	/*
	 * 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:
	 */
2881
	rq_unpin_lock(rq, rf);
2882
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
L
Linus Torvalds 已提交
2883 2884 2885

	/* Here we just switch the register state and the stack. */
	switch_to(prev, next, prev);
I
Ingo Molnar 已提交
2886
	barrier();
2887 2888

	return finish_task_switch(prev);
L
Linus Torvalds 已提交
2889 2890 2891
}

/*
2892
 * nr_running and nr_context_switches:
L
Linus Torvalds 已提交
2893 2894
 *
 * externally visible scheduler statistics: current number of runnable
2895
 * threads, total number of context switches performed since bootup.
L
Linus Torvalds 已提交
2896 2897 2898 2899 2900 2901 2902 2903 2904
 */
unsigned long nr_running(void)
{
	unsigned long i, sum = 0;

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

	return sum;
2905
}
L
Linus Torvalds 已提交
2906

2907
/*
I
Ingo Molnar 已提交
2908
 * Check if only the current task is running on the CPU.
2909 2910 2911 2912 2913 2914 2915 2916 2917 2918
 *
 * 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)
2919 2920 2921
 */
bool single_task_running(void)
{
2922
	return raw_rq()->nr_running == 1;
2923 2924 2925
}
EXPORT_SYMBOL(single_task_running);

L
Linus Torvalds 已提交
2926
unsigned long long nr_context_switches(void)
2927
{
2928 2929
	int i;
	unsigned long long sum = 0;
2930

2931
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2932
		sum += cpu_rq(i)->nr_switches;
2933

L
Linus Torvalds 已提交
2934 2935
	return sum;
}
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 2965 2966
/*
 * 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
2967 2968 2969
unsigned long nr_iowait(void)
{
	unsigned long i, sum = 0;
2970

2971
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2972
		sum += atomic_read(&cpu_rq(i)->nr_iowait);
2973

L
Linus Torvalds 已提交
2974 2975
	return sum;
}
2976

2977 2978 2979 2980 2981 2982 2983
/*
 * 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.
 */

2984
unsigned long nr_iowait_cpu(int cpu)
2985
{
2986
	struct rq *this = cpu_rq(cpu);
2987 2988
	return atomic_read(&this->nr_iowait);
}
2989

2990 2991
void get_iowait_load(unsigned long *nr_waiters, unsigned long *load)
{
2992 2993 2994
	struct rq *rq = this_rq();
	*nr_waiters = atomic_read(&rq->nr_iowait);
	*load = rq->load.weight;
2995 2996
}

I
Ingo Molnar 已提交
2997
#ifdef CONFIG_SMP
2998

2999
/*
P
Peter Zijlstra 已提交
3000 3001
 * sched_exec - execve() is a valuable balancing opportunity, because at
 * this point the task has the smallest effective memory and cache footprint.
3002
 */
P
Peter Zijlstra 已提交
3003
void sched_exec(void)
3004
{
P
Peter Zijlstra 已提交
3005
	struct task_struct *p = current;
L
Linus Torvalds 已提交
3006
	unsigned long flags;
3007
	int dest_cpu;
3008

3009
	raw_spin_lock_irqsave(&p->pi_lock, flags);
3010
	dest_cpu = p->sched_class->select_task_rq(p, task_cpu(p), SD_BALANCE_EXEC, 0);
3011 3012
	if (dest_cpu == smp_processor_id())
		goto unlock;
P
Peter Zijlstra 已提交
3013

3014
	if (likely(cpu_active(dest_cpu))) {
3015
		struct migration_arg arg = { p, dest_cpu };
3016

3017 3018
		raw_spin_unlock_irqrestore(&p->pi_lock, flags);
		stop_one_cpu(task_cpu(p), migration_cpu_stop, &arg);
L
Linus Torvalds 已提交
3019 3020
		return;
	}
3021
unlock:
3022
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
3023
}
I
Ingo Molnar 已提交
3024

L
Linus Torvalds 已提交
3025 3026 3027
#endif

DEFINE_PER_CPU(struct kernel_stat, kstat);
3028
DEFINE_PER_CPU(struct kernel_cpustat, kernel_cpustat);
L
Linus Torvalds 已提交
3029 3030

EXPORT_PER_CPU_SYMBOL(kstat);
3031
EXPORT_PER_CPU_SYMBOL(kernel_cpustat);
L
Linus Torvalds 已提交
3032

3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049
/*
 * 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);
}

3050 3051 3052 3053 3054 3055 3056
/*
 * 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)
{
3057
	struct rq_flags rf;
3058
	struct rq *rq;
3059
	u64 ns;
3060

3061 3062 3063 3064 3065 3066
#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
3067 3068
	 * 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
3069
	 * indistinguishable from the read occurring a few cycles earlier.
3070 3071
	 * If we see ->on_cpu without ->on_rq, the task is leaving, and has
	 * been accounted, so we're correct here as well.
3072
	 */
3073
	if (!p->on_cpu || !task_on_rq_queued(p))
3074 3075 3076
		return p->se.sum_exec_runtime;
#endif

3077
	rq = task_rq_lock(p, &rf);
3078 3079 3080 3081 3082 3083
	/*
	 * 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)) {
3084
		prefetch_curr_exec_start(p);
3085 3086 3087 3088
		update_rq_clock(rq);
		p->sched_class->update_curr(rq);
	}
	ns = p->se.sum_exec_runtime;
3089
	task_rq_unlock(rq, p, &rf);
3090 3091 3092

	return ns;
}
3093

3094 3095 3096 3097 3098 3099 3100 3101
/*
 * 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
3102
	struct task_struct *curr = rq->curr;
3103
	struct rq_flags rf;
3104 3105

	sched_clock_tick();
I
Ingo Molnar 已提交
3106

3107 3108
	rq_lock(rq, &rf);

3109
	update_rq_clock(rq);
P
Peter Zijlstra 已提交
3110
	curr->sched_class->task_tick(rq, curr, 0);
3111
	cpu_load_update_active(rq);
3112
	calc_global_load_tick(rq);
3113 3114

	rq_unlock(rq, &rf);
3115

3116
	perf_event_task_tick();
3117

3118
#ifdef CONFIG_SMP
3119
	rq->idle_balance = idle_cpu(cpu);
3120
	trigger_load_balance(rq);
3121
#endif
3122
	rq_last_tick_reset(rq);
L
Linus Torvalds 已提交
3123 3124
}

3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135
#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.
3136 3137
 *
 * Return: Maximum deferment in nanoseconds.
3138 3139 3140 3141
 */
u64 scheduler_tick_max_deferment(void)
{
	struct rq *rq = this_rq();
3142
	unsigned long next, now = READ_ONCE(jiffies);
3143 3144 3145 3146 3147 3148

	next = rq->last_sched_tick + HZ;

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

3149
	return jiffies_to_nsecs(next - now);
L
Linus Torvalds 已提交
3150
}
3151
#endif
L
Linus Torvalds 已提交
3152

3153 3154
#if defined(CONFIG_PREEMPT) && (defined(CONFIG_DEBUG_PREEMPT) || \
				defined(CONFIG_PREEMPT_TRACER))
3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168
/*
 * 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);
	}
}
3169

3170
void preempt_count_add(int val)
L
Linus Torvalds 已提交
3171
{
3172
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3173 3174 3175
	/*
	 * Underflow?
	 */
3176 3177
	if (DEBUG_LOCKS_WARN_ON((preempt_count() < 0)))
		return;
3178
#endif
3179
	__preempt_count_add(val);
3180
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3181 3182 3183
	/*
	 * Spinlock count overflowing soon?
	 */
3184 3185
	DEBUG_LOCKS_WARN_ON((preempt_count() & PREEMPT_MASK) >=
				PREEMPT_MASK - 10);
3186
#endif
3187
	preempt_latency_start(val);
L
Linus Torvalds 已提交
3188
}
3189
EXPORT_SYMBOL(preempt_count_add);
3190
NOKPROBE_SYMBOL(preempt_count_add);
L
Linus Torvalds 已提交
3191

3192 3193 3194 3195 3196 3197 3198 3199 3200 3201
/*
 * 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());
}

3202
void preempt_count_sub(int val)
L
Linus Torvalds 已提交
3203
{
3204
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3205 3206 3207
	/*
	 * Underflow?
	 */
3208
	if (DEBUG_LOCKS_WARN_ON(val > preempt_count()))
3209
		return;
L
Linus Torvalds 已提交
3210 3211 3212
	/*
	 * Is the spinlock portion underflowing?
	 */
3213 3214 3215
	if (DEBUG_LOCKS_WARN_ON((val < PREEMPT_MASK) &&
			!(preempt_count() & PREEMPT_MASK)))
		return;
3216
#endif
3217

3218
	preempt_latency_stop(val);
3219
	__preempt_count_sub(val);
L
Linus Torvalds 已提交
3220
}
3221
EXPORT_SYMBOL(preempt_count_sub);
3222
NOKPROBE_SYMBOL(preempt_count_sub);
L
Linus Torvalds 已提交
3223

3224 3225 3226
#else
static inline void preempt_latency_start(int val) { }
static inline void preempt_latency_stop(int val) { }
L
Linus Torvalds 已提交
3227 3228
#endif

3229 3230 3231 3232 3233 3234 3235 3236 3237
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
3238
/*
I
Ingo Molnar 已提交
3239
 * Print scheduling while atomic bug:
L
Linus Torvalds 已提交
3240
 */
I
Ingo Molnar 已提交
3241
static noinline void __schedule_bug(struct task_struct *prev)
L
Linus Torvalds 已提交
3242
{
3243 3244 3245
	/* Save this before calling printk(), since that will clobber it */
	unsigned long preempt_disable_ip = get_preempt_disable_ip(current);

3246 3247 3248
	if (oops_in_progress)
		return;

P
Peter Zijlstra 已提交
3249 3250
	printk(KERN_ERR "BUG: scheduling while atomic: %s/%d/0x%08x\n",
		prev->comm, prev->pid, preempt_count());
3251

I
Ingo Molnar 已提交
3252
	debug_show_held_locks(prev);
3253
	print_modules();
I
Ingo Molnar 已提交
3254 3255
	if (irqs_disabled())
		print_irqtrace_events(prev);
3256 3257
	if (IS_ENABLED(CONFIG_DEBUG_PREEMPT)
	    && in_atomic_preempt_off()) {
3258
		pr_err("Preemption disabled at:");
3259
		print_ip_sym(preempt_disable_ip);
3260 3261
		pr_cont("\n");
	}
3262 3263 3264
	if (panic_on_warn)
		panic("scheduling while atomic\n");

3265
	dump_stack();
3266
	add_taint(TAINT_WARN, LOCKDEP_STILL_OK);
I
Ingo Molnar 已提交
3267
}
L
Linus Torvalds 已提交
3268

I
Ingo Molnar 已提交
3269 3270 3271 3272 3273
/*
 * Various schedule()-time debugging checks and statistics:
 */
static inline void schedule_debug(struct task_struct *prev)
{
3274
#ifdef CONFIG_SCHED_STACK_END_CHECK
J
Jann Horn 已提交
3275 3276
	if (task_stack_end_corrupted(prev))
		panic("corrupted stack end detected inside scheduler\n");
3277
#endif
3278

3279
	if (unlikely(in_atomic_preempt_off())) {
I
Ingo Molnar 已提交
3280
		__schedule_bug(prev);
3281 3282
		preempt_count_set(PREEMPT_DISABLED);
	}
3283
	rcu_sleep_check();
I
Ingo Molnar 已提交
3284

L
Linus Torvalds 已提交
3285 3286
	profile_hit(SCHED_PROFILING, __builtin_return_address(0));

3287
	schedstat_inc(this_rq()->sched_count);
I
Ingo Molnar 已提交
3288 3289 3290 3291 3292 3293
}

/*
 * Pick up the highest-prio task:
 */
static inline struct task_struct *
3294
pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
I
Ingo Molnar 已提交
3295
{
3296
	const struct sched_class *class;
I
Ingo Molnar 已提交
3297
	struct task_struct *p;
L
Linus Torvalds 已提交
3298 3299

	/*
3300 3301 3302 3303
	 * 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
3304
	 */
3305 3306 3307 3308
	if (likely((prev->sched_class == &idle_sched_class ||
		    prev->sched_class == &fair_sched_class) &&
		   rq->nr_running == rq->cfs.h_nr_running)) {

3309
		p = fair_sched_class.pick_next_task(rq, prev, rf);
3310 3311 3312
		if (unlikely(p == RETRY_TASK))
			goto again;

I
Ingo Molnar 已提交
3313
		/* Assumes fair_sched_class->next == idle_sched_class */
3314
		if (unlikely(!p))
3315
			p = idle_sched_class.pick_next_task(rq, prev, rf);
3316 3317

		return p;
L
Linus Torvalds 已提交
3318 3319
	}

3320
again:
3321
	for_each_class(class) {
3322
		p = class->pick_next_task(rq, prev, rf);
3323 3324 3325
		if (p) {
			if (unlikely(p == RETRY_TASK))
				goto again;
I
Ingo Molnar 已提交
3326
			return p;
3327
		}
I
Ingo Molnar 已提交
3328
	}
3329

I
Ingo Molnar 已提交
3330 3331
	/* The idle class should always have a runnable task: */
	BUG();
I
Ingo Molnar 已提交
3332
}
L
Linus Torvalds 已提交
3333

I
Ingo Molnar 已提交
3334
/*
3335
 * __schedule() is the main scheduler function.
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 3368 3369
 *
 * 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
3370
 *
3371
 * WARNING: must be called with preemption disabled!
I
Ingo Molnar 已提交
3372
 */
3373
static void __sched notrace __schedule(bool preempt)
I
Ingo Molnar 已提交
3374 3375
{
	struct task_struct *prev, *next;
3376
	unsigned long *switch_count;
3377
	struct rq_flags rf;
I
Ingo Molnar 已提交
3378
	struct rq *rq;
3379
	int cpu;
I
Ingo Molnar 已提交
3380 3381 3382 3383 3384 3385

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

	schedule_debug(prev);
L
Linus Torvalds 已提交
3386

3387
	if (sched_feat(HRTICK))
M
Mike Galbraith 已提交
3388
		hrtick_clear(rq);
P
Peter Zijlstra 已提交
3389

3390
	local_irq_disable();
3391
	rcu_note_context_switch(preempt);
3392

3393 3394 3395 3396 3397 3398
	/*
	 * 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();
3399
	rq_lock(rq, &rf);
L
Linus Torvalds 已提交
3400

I
Ingo Molnar 已提交
3401 3402
	/* Promote REQ to ACT */
	rq->clock_update_flags <<= 1;
3403
	update_rq_clock(rq);
3404

3405
	switch_count = &prev->nivcsw;
3406
	if (!preempt && prev->state) {
T
Tejun Heo 已提交
3407
		if (unlikely(signal_pending_state(prev->state, prev))) {
L
Linus Torvalds 已提交
3408
			prev->state = TASK_RUNNING;
T
Tejun Heo 已提交
3409
		} else {
3410
			deactivate_task(rq, prev, DEQUEUE_SLEEP | DEQUEUE_NOCLOCK);
3411 3412
			prev->on_rq = 0;

3413 3414 3415 3416 3417
			if (prev->in_iowait) {
				atomic_inc(&rq->nr_iowait);
				delayacct_blkio_start();
			}

T
Tejun Heo 已提交
3418
			/*
3419 3420 3421
			 * 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
3422 3423 3424 3425
			 */
			if (prev->flags & PF_WQ_WORKER) {
				struct task_struct *to_wakeup;

3426
				to_wakeup = wq_worker_sleeping(prev);
T
Tejun Heo 已提交
3427
				if (to_wakeup)
3428
					try_to_wake_up_local(to_wakeup, &rf);
T
Tejun Heo 已提交
3429 3430
			}
		}
I
Ingo Molnar 已提交
3431
		switch_count = &prev->nvcsw;
L
Linus Torvalds 已提交
3432 3433
	}

3434
	next = pick_next_task(rq, prev, &rf);
3435
	clear_tsk_need_resched(prev);
3436
	clear_preempt_need_resched();
L
Linus Torvalds 已提交
3437 3438 3439 3440 3441 3442

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

3443
		trace_sched_switch(preempt, prev, next);
I
Ingo Molnar 已提交
3444 3445 3446

		/* Also unlocks the rq: */
		rq = context_switch(rq, prev, next, &rf);
3447
	} else {
3448
		rq->clock_update_flags &= ~(RQCF_ACT_SKIP|RQCF_REQ_SKIP);
3449
		rq_unlock_irq(rq, &rf);
3450
	}
L
Linus Torvalds 已提交
3451

3452
	balance_callback(rq);
L
Linus Torvalds 已提交
3453
}
3454

3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471
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
3472
	/* Causes final put_task_struct in finish_task_switch(): */
3473
	__set_current_state(TASK_DEAD);
I
Ingo Molnar 已提交
3474 3475 3476 3477

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

3478 3479
	__schedule(false);
	BUG();
I
Ingo Molnar 已提交
3480 3481

	/* Avoid "noreturn function does return" - but don't continue if BUG() is a NOP: */
3482
	for (;;)
I
Ingo Molnar 已提交
3483
		cpu_relax();
3484 3485
}

3486 3487
static inline void sched_submit_work(struct task_struct *tsk)
{
3488
	if (!tsk->state || tsk_is_pi_blocked(tsk))
3489 3490 3491 3492 3493 3494 3495 3496 3497
		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);
}

3498
asmlinkage __visible void __sched schedule(void)
3499
{
3500 3501 3502
	struct task_struct *tsk = current;

	sched_submit_work(tsk);
3503
	do {
3504
		preempt_disable();
3505
		__schedule(false);
3506
		sched_preempt_enable_no_resched();
3507
	} while (need_resched());
3508
}
L
Linus Torvalds 已提交
3509 3510
EXPORT_SYMBOL(schedule);

3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535
/*
 * 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());
}

3536
#ifdef CONFIG_CONTEXT_TRACKING
3537
asmlinkage __visible void __sched schedule_user(void)
3538 3539 3540 3541 3542 3543
{
	/*
	 * 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.
3544 3545
	 *
	 * NB: There are buggy callers of this function.  Ideally we
3546
	 * should warn if prev_state != CONTEXT_USER, but that will trigger
3547
	 * too frequently to make sense yet.
3548
	 */
3549
	enum ctx_state prev_state = exception_enter();
3550
	schedule();
3551
	exception_exit(prev_state);
3552 3553 3554
}
#endif

3555 3556 3557 3558 3559 3560 3561
/**
 * schedule_preempt_disabled - called with preemption disabled
 *
 * Returns with preemption disabled. Note: preempt_count must be 1
 */
void __sched schedule_preempt_disabled(void)
{
3562
	sched_preempt_enable_no_resched();
3563 3564 3565 3566
	schedule();
	preempt_disable();
}

3567
static void __sched notrace preempt_schedule_common(void)
3568 3569
{
	do {
3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582
		/*
		 * 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.
		 */
3583
		preempt_disable_notrace();
3584
		preempt_latency_start(1);
3585
		__schedule(true);
3586
		preempt_latency_stop(1);
3587
		preempt_enable_no_resched_notrace();
3588 3589 3590 3591 3592 3593 3594 3595

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

L
Linus Torvalds 已提交
3596 3597
#ifdef CONFIG_PREEMPT
/*
3598
 * this is the entry point to schedule() from in-kernel preemption
I
Ingo Molnar 已提交
3599
 * off of preempt_enable. Kernel preemptions off return from interrupt
L
Linus Torvalds 已提交
3600 3601
 * occur there and call schedule directly.
 */
3602
asmlinkage __visible void __sched notrace preempt_schedule(void)
L
Linus Torvalds 已提交
3603 3604 3605
{
	/*
	 * If there is a non-zero preempt_count or interrupts are disabled,
I
Ingo Molnar 已提交
3606
	 * we do not want to preempt the current task. Just return..
L
Linus Torvalds 已提交
3607
	 */
3608
	if (likely(!preemptible()))
L
Linus Torvalds 已提交
3609 3610
		return;

3611
	preempt_schedule_common();
L
Linus Torvalds 已提交
3612
}
3613
NOKPROBE_SYMBOL(preempt_schedule);
L
Linus Torvalds 已提交
3614
EXPORT_SYMBOL(preempt_schedule);
3615 3616

/**
3617
 * preempt_schedule_notrace - preempt_schedule called by tracing
3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629
 *
 * 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.
 */
3630
asmlinkage __visible void __sched notrace preempt_schedule_notrace(void)
3631 3632 3633 3634 3635 3636 3637
{
	enum ctx_state prev_ctx;

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

	do {
3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650
		/*
		 * 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.
		 */
3651
		preempt_disable_notrace();
3652
		preempt_latency_start(1);
3653 3654 3655 3656 3657 3658
		/*
		 * Needs preempt disabled in case user_exit() is traced
		 * and the tracer calls preempt_enable_notrace() causing
		 * an infinite recursion.
		 */
		prev_ctx = exception_enter();
3659
		__schedule(true);
3660 3661
		exception_exit(prev_ctx);

3662
		preempt_latency_stop(1);
3663
		preempt_enable_no_resched_notrace();
3664 3665
	} while (need_resched());
}
3666
EXPORT_SYMBOL_GPL(preempt_schedule_notrace);
3667

3668
#endif /* CONFIG_PREEMPT */
L
Linus Torvalds 已提交
3669 3670

/*
3671
 * this is the entry point to schedule() from kernel preemption
L
Linus Torvalds 已提交
3672 3673 3674 3675
 * off of irq context.
 * Note, that this is called and return with irqs disabled. This will
 * protect us against recursive calling from irq.
 */
3676
asmlinkage __visible void __sched preempt_schedule_irq(void)
L
Linus Torvalds 已提交
3677
{
3678
	enum ctx_state prev_state;
3679

3680
	/* Catch callers which need to be fixed */
3681
	BUG_ON(preempt_count() || !irqs_disabled());
L
Linus Torvalds 已提交
3682

3683 3684
	prev_state = exception_enter();

3685
	do {
3686
		preempt_disable();
3687
		local_irq_enable();
3688
		__schedule(true);
3689
		local_irq_disable();
3690
		sched_preempt_enable_no_resched();
3691
	} while (need_resched());
3692 3693

	exception_exit(prev_state);
L
Linus Torvalds 已提交
3694 3695
}

3696
int default_wake_function(wait_queue_entry_t *curr, unsigned mode, int wake_flags,
I
Ingo Molnar 已提交
3697
			  void *key)
L
Linus Torvalds 已提交
3698
{
P
Peter Zijlstra 已提交
3699
	return try_to_wake_up(curr->private, mode, wake_flags);
L
Linus Torvalds 已提交
3700 3701 3702
}
EXPORT_SYMBOL(default_wake_function);

3703 3704
#ifdef CONFIG_RT_MUTEXES

3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719
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);
}

3720 3721
/*
 * rt_mutex_setprio - set the current priority of a task
3722 3723
 * @p: task to boost
 * @pi_task: donor task
3724 3725 3726 3727
 *
 * This function changes the 'effective' priority of a task. It does
 * not touch ->normal_prio like __setscheduler().
 *
3728 3729
 * Used by the rt_mutex code to implement priority inheritance
 * logic. Call site only calls if the priority of the task changed.
3730
 */
3731
void rt_mutex_setprio(struct task_struct *p, struct task_struct *pi_task)
3732
{
3733
	int prio, oldprio, queued, running, queue_flag =
3734
		DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
3735
	const struct sched_class *prev_class;
3736 3737
	struct rq_flags rf;
	struct rq *rq;
3738

3739 3740 3741 3742 3743 3744 3745 3746
	/* 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;
3747

3748
	rq = __task_rq_lock(p, &rf);
3749
	update_rq_clock(rq);
3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766
	/*
	 * 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;
3767

3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785
	/*
	 * 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;
	}

3786
	trace_sched_pi_setprio(p, pi_task);
3787
	oldprio = p->prio;
3788 3789 3790 3791

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

3792
	prev_class = p->sched_class;
3793
	queued = task_on_rq_queued(p);
3794
	running = task_current(rq, p);
3795
	if (queued)
3796
		dequeue_task(rq, p, queue_flag);
3797
	if (running)
3798
		put_prev_task(rq, p);
I
Ingo Molnar 已提交
3799

3800 3801 3802 3803 3804 3805 3806 3807 3808 3809
	/*
	 * 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)) {
3810 3811
		if (!dl_prio(p->normal_prio) ||
		    (pi_task && dl_entity_preempt(&pi_task->dl, &p->dl))) {
3812
			p->dl.dl_boosted = 1;
3813
			queue_flag |= ENQUEUE_REPLENISH;
3814 3815
		} else
			p->dl.dl_boosted = 0;
3816
		p->sched_class = &dl_sched_class;
3817 3818 3819 3820
	} else if (rt_prio(prio)) {
		if (dl_prio(oldprio))
			p->dl.dl_boosted = 0;
		if (oldprio < prio)
3821
			queue_flag |= ENQUEUE_HEAD;
I
Ingo Molnar 已提交
3822
		p->sched_class = &rt_sched_class;
3823 3824 3825
	} else {
		if (dl_prio(oldprio))
			p->dl.dl_boosted = 0;
3826 3827
		if (rt_prio(oldprio))
			p->rt.timeout = 0;
I
Ingo Molnar 已提交
3828
		p->sched_class = &fair_sched_class;
3829
	}
I
Ingo Molnar 已提交
3830

3831 3832
	p->prio = prio;

3833
	if (queued)
3834
		enqueue_task(rq, p, queue_flag);
3835
	if (running)
3836
		set_curr_task(rq, p);
3837

P
Peter Zijlstra 已提交
3838
	check_class_changed(rq, p, prev_class, oldprio);
3839
out_unlock:
I
Ingo Molnar 已提交
3840 3841
	/* Avoid rq from going away on us: */
	preempt_disable();
3842
	__task_rq_unlock(rq, &rf);
3843 3844 3845

	balance_callback(rq);
	preempt_enable();
3846
}
3847 3848 3849 3850 3851
#else
static inline int rt_effective_prio(struct task_struct *p, int prio)
{
	return prio;
}
3852
#endif
3853

3854
void set_user_nice(struct task_struct *p, long nice)
L
Linus Torvalds 已提交
3855
{
P
Peter Zijlstra 已提交
3856 3857
	bool queued, running;
	int old_prio, delta;
3858
	struct rq_flags rf;
3859
	struct rq *rq;
L
Linus Torvalds 已提交
3860

3861
	if (task_nice(p) == nice || nice < MIN_NICE || nice > MAX_NICE)
L
Linus Torvalds 已提交
3862 3863 3864 3865 3866
		return;
	/*
	 * We have to be careful, if called from sys_setpriority(),
	 * the task might be in the middle of scheduling on another CPU.
	 */
3867
	rq = task_rq_lock(p, &rf);
3868 3869
	update_rq_clock(rq);

L
Linus Torvalds 已提交
3870 3871 3872 3873
	/*
	 * 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
3874
	 * SCHED_DEADLINE, SCHED_FIFO or SCHED_RR:
L
Linus Torvalds 已提交
3875
	 */
3876
	if (task_has_dl_policy(p) || task_has_rt_policy(p)) {
L
Linus Torvalds 已提交
3877 3878 3879
		p->static_prio = NICE_TO_PRIO(nice);
		goto out_unlock;
	}
3880
	queued = task_on_rq_queued(p);
P
Peter Zijlstra 已提交
3881
	running = task_current(rq, p);
3882
	if (queued)
3883
		dequeue_task(rq, p, DEQUEUE_SAVE | DEQUEUE_NOCLOCK);
P
Peter Zijlstra 已提交
3884 3885
	if (running)
		put_prev_task(rq, p);
L
Linus Torvalds 已提交
3886 3887

	p->static_prio = NICE_TO_PRIO(nice);
3888
	set_load_weight(p);
3889 3890 3891
	old_prio = p->prio;
	p->prio = effective_prio(p);
	delta = p->prio - old_prio;
L
Linus Torvalds 已提交
3892

3893
	if (queued) {
3894
		enqueue_task(rq, p, ENQUEUE_RESTORE | ENQUEUE_NOCLOCK);
L
Linus Torvalds 已提交
3895
		/*
3896 3897
		 * If the task increased its priority or is running and
		 * lowered its priority, then reschedule its CPU:
L
Linus Torvalds 已提交
3898
		 */
3899
		if (delta < 0 || (delta > 0 && task_running(rq, p)))
3900
			resched_curr(rq);
L
Linus Torvalds 已提交
3901
	}
P
Peter Zijlstra 已提交
3902 3903
	if (running)
		set_curr_task(rq, p);
L
Linus Torvalds 已提交
3904
out_unlock:
3905
	task_rq_unlock(rq, p, &rf);
L
Linus Torvalds 已提交
3906 3907 3908
}
EXPORT_SYMBOL(set_user_nice);

M
Matt Mackall 已提交
3909 3910 3911 3912 3913
/*
 * can_nice - check if a task can reduce its nice value
 * @p: task
 * @nice: nice value
 */
3914
int can_nice(const struct task_struct *p, const int nice)
M
Matt Mackall 已提交
3915
{
I
Ingo Molnar 已提交
3916
	/* Convert nice value [19,-20] to rlimit style value [1,40]: */
3917
	int nice_rlim = nice_to_rlimit(nice);
3918

3919
	return (nice_rlim <= task_rlimit(p, RLIMIT_NICE) ||
M
Matt Mackall 已提交
3920 3921 3922
		capable(CAP_SYS_NICE));
}

L
Linus Torvalds 已提交
3923 3924 3925 3926 3927 3928 3929 3930 3931
#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.
 */
3932
SYSCALL_DEFINE1(nice, int, increment)
L
Linus Torvalds 已提交
3933
{
3934
	long nice, retval;
L
Linus Torvalds 已提交
3935 3936 3937 3938 3939 3940

	/*
	 * 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.
	 */
3941
	increment = clamp(increment, -NICE_WIDTH, NICE_WIDTH);
3942
	nice = task_nice(current) + increment;
L
Linus Torvalds 已提交
3943

3944
	nice = clamp_val(nice, MIN_NICE, MAX_NICE);
M
Matt Mackall 已提交
3945 3946 3947
	if (increment < 0 && !can_nice(current, nice))
		return -EPERM;

L
Linus Torvalds 已提交
3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961
	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.
 *
3962
 * Return: The priority value as seen by users in /proc.
L
Linus Torvalds 已提交
3963 3964 3965
 * RT tasks are offset by -200. Normal tasks are centered
 * around 0, value goes from -16 to +15.
 */
3966
int task_prio(const struct task_struct *p)
L
Linus Torvalds 已提交
3967 3968 3969 3970 3971
{
	return p->prio - MAX_RT_PRIO;
}

/**
I
Ingo Molnar 已提交
3972
 * idle_cpu - is a given CPU idle currently?
L
Linus Torvalds 已提交
3973
 * @cpu: the processor in question.
3974 3975
 *
 * Return: 1 if the CPU is currently idle. 0 otherwise.
L
Linus Torvalds 已提交
3976 3977 3978
 */
int idle_cpu(int cpu)
{
T
Thomas Gleixner 已提交
3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992
	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
3993 3994 3995
}

/**
I
Ingo Molnar 已提交
3996
 * idle_task - return the idle task for a given CPU.
L
Linus Torvalds 已提交
3997
 * @cpu: the processor in question.
3998
 *
I
Ingo Molnar 已提交
3999
 * Return: The idle task for the CPU @cpu.
L
Linus Torvalds 已提交
4000
 */
4001
struct task_struct *idle_task(int cpu)
L
Linus Torvalds 已提交
4002 4003 4004 4005 4006 4007 4008
{
	return cpu_rq(cpu)->idle;
}

/**
 * find_process_by_pid - find a process with a matching PID value.
 * @pid: the pid in question.
4009 4010
 *
 * The task of @pid, if found. %NULL otherwise.
L
Linus Torvalds 已提交
4011
 */
A
Alexey Dobriyan 已提交
4012
static struct task_struct *find_process_by_pid(pid_t pid)
L
Linus Torvalds 已提交
4013
{
4014
	return pid ? find_task_by_vpid(pid) : current;
L
Linus Torvalds 已提交
4015 4016
}

4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031
/*
 * 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;
4032
	dl_se->dl_period = attr->sched_period ?: dl_se->dl_deadline;
4033
	dl_se->flags = attr->sched_flags;
4034
	dl_se->dl_bw = to_ratio(dl_se->dl_period, dl_se->dl_runtime);
4035
	dl_se->dl_density = to_ratio(dl_se->dl_deadline, dl_se->dl_runtime);
4036 4037
}

4038 4039 4040 4041 4042 4043
/*
 * sched_setparam() passes in -1 for its policy, to let the functions
 * it calls know not to change it.
 */
#define SETPARAM_POLICY	-1

4044 4045
static void __setscheduler_params(struct task_struct *p,
		const struct sched_attr *attr)
L
Linus Torvalds 已提交
4046
{
4047 4048
	int policy = attr->sched_policy;

4049
	if (policy == SETPARAM_POLICY)
4050 4051
		policy = p->policy;

L
Linus Torvalds 已提交
4052
	p->policy = policy;
4053

4054 4055
	if (dl_policy(policy))
		__setparam_dl(p, attr);
4056
	else if (fair_policy(policy))
4057 4058
		p->static_prio = NICE_TO_PRIO(attr->sched_nice);

4059 4060 4061 4062 4063 4064
	/*
	 * __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;
4065
	p->normal_prio = normal_prio(p);
4066 4067
	set_load_weight(p);
}
4068

4069 4070
/* Actually do priority change: must hold pi & rq lock. */
static void __setscheduler(struct rq *rq, struct task_struct *p,
4071
			   const struct sched_attr *attr, bool keep_boost)
4072 4073
{
	__setscheduler_params(p, attr);
4074

4075
	/*
4076 4077
	 * Keep a potential priority boosting if called from
	 * sched_setscheduler().
4078
	 */
4079
	p->prio = normal_prio(p);
4080
	if (keep_boost)
4081
		p->prio = rt_effective_prio(p, p->prio);
4082

4083 4084 4085
	if (dl_prio(p->prio))
		p->sched_class = &dl_sched_class;
	else if (rt_prio(p->prio))
4086 4087 4088
		p->sched_class = &rt_sched_class;
	else
		p->sched_class = &fair_sched_class;
L
Linus Torvalds 已提交
4089
}
4090 4091 4092 4093 4094 4095 4096 4097 4098

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;
4099
	attr->sched_period = dl_se->dl_period;
4100 4101 4102 4103 4104 4105
	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
4106
 * than the runtime, as well as the period of being zero or
4107
 * greater than deadline. Furthermore, we have to be sure that
4108 4109 4110 4111
 * 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).
4112 4113 4114 4115
 */
static bool
__checkparam_dl(const struct sched_attr *attr)
{
4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141
	/* 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;
4142 4143
}

4144
/*
I
Ingo Molnar 已提交
4145
 * Check the target process has a UID that matches the current process's:
4146 4147 4148 4149 4150 4151 4152 4153
 */
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);
4154 4155
	match = (uid_eq(cred->euid, pcred->euid) ||
		 uid_eq(cred->euid, pcred->uid));
4156 4157 4158 4159
	rcu_read_unlock();
	return match;
}

I
Ingo Molnar 已提交
4160
static bool dl_param_changed(struct task_struct *p, const struct sched_attr *attr)
4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172
{
	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;
}

4173 4174
static int __sched_setscheduler(struct task_struct *p,
				const struct sched_attr *attr,
4175
				bool user, bool pi)
L
Linus Torvalds 已提交
4176
{
4177 4178
	int newprio = dl_policy(attr->sched_policy) ? MAX_DL_PRIO - 1 :
		      MAX_RT_PRIO - 1 - attr->sched_priority;
4179
	int retval, oldprio, oldpolicy = -1, queued, running;
4180
	int new_effective_prio, policy = attr->sched_policy;
4181
	const struct sched_class *prev_class;
4182
	struct rq_flags rf;
4183
	int reset_on_fork;
4184
	int queue_flags = DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
4185
	struct rq *rq;
L
Linus Torvalds 已提交
4186

4187 4188
	/* The pi code expects interrupts enabled */
	BUG_ON(pi && in_interrupt());
L
Linus Torvalds 已提交
4189
recheck:
I
Ingo Molnar 已提交
4190
	/* Double check policy once rq lock held: */
4191 4192
	if (policy < 0) {
		reset_on_fork = p->sched_reset_on_fork;
L
Linus Torvalds 已提交
4193
		policy = oldpolicy = p->policy;
4194
	} else {
4195
		reset_on_fork = !!(attr->sched_flags & SCHED_FLAG_RESET_ON_FORK);
4196

4197
		if (!valid_policy(policy))
4198 4199 4200
			return -EINVAL;
	}

4201 4202
	if (attr->sched_flags &
		~(SCHED_FLAG_RESET_ON_FORK | SCHED_FLAG_RECLAIM))
4203 4204
		return -EINVAL;

L
Linus Torvalds 已提交
4205 4206
	/*
	 * Valid priorities for SCHED_FIFO and SCHED_RR are
I
Ingo Molnar 已提交
4207 4208
	 * 1..MAX_USER_RT_PRIO-1, valid priority for SCHED_NORMAL,
	 * SCHED_BATCH and SCHED_IDLE is 0.
L
Linus Torvalds 已提交
4209
	 */
4210
	if ((p->mm && attr->sched_priority > MAX_USER_RT_PRIO-1) ||
4211
	    (!p->mm && attr->sched_priority > MAX_RT_PRIO-1))
L
Linus Torvalds 已提交
4212
		return -EINVAL;
4213 4214
	if ((dl_policy(policy) && !__checkparam_dl(attr)) ||
	    (rt_policy(policy) != (attr->sched_priority != 0)))
L
Linus Torvalds 已提交
4215 4216
		return -EINVAL;

4217 4218 4219
	/*
	 * Allow unprivileged RT tasks to decrease priority:
	 */
4220
	if (user && !capable(CAP_SYS_NICE)) {
4221
		if (fair_policy(policy)) {
4222
			if (attr->sched_nice < task_nice(p) &&
4223
			    !can_nice(p, attr->sched_nice))
4224 4225 4226
				return -EPERM;
		}

4227
		if (rt_policy(policy)) {
4228 4229
			unsigned long rlim_rtprio =
					task_rlimit(p, RLIMIT_RTPRIO);
4230

I
Ingo Molnar 已提交
4231
			/* Can't set/change the rt policy: */
4232 4233 4234
			if (policy != p->policy && !rlim_rtprio)
				return -EPERM;

I
Ingo Molnar 已提交
4235
			/* Can't increase priority: */
4236 4237
			if (attr->sched_priority > p->rt_priority &&
			    attr->sched_priority > rlim_rtprio)
4238 4239
				return -EPERM;
		}
4240

4241 4242 4243 4244 4245 4246 4247 4248 4249
		 /*
		  * 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
4250
		/*
4251 4252
		 * 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
4253
		 */
4254
		if (idle_policy(p->policy) && !idle_policy(policy)) {
4255
			if (!can_nice(p, task_nice(p)))
4256 4257
				return -EPERM;
		}
4258

I
Ingo Molnar 已提交
4259
		/* Can't change other user's priorities: */
4260
		if (!check_same_owner(p))
4261
			return -EPERM;
4262

I
Ingo Molnar 已提交
4263
		/* Normal users shall not reset the sched_reset_on_fork flag: */
4264 4265
		if (p->sched_reset_on_fork && !reset_on_fork)
			return -EPERM;
4266
	}
L
Linus Torvalds 已提交
4267

4268
	if (user) {
4269
		retval = security_task_setscheduler(p);
4270 4271 4272 4273
		if (retval)
			return retval;
	}

4274
	/*
I
Ingo Molnar 已提交
4275
	 * Make sure no PI-waiters arrive (or leave) while we are
4276
	 * changing the priority of the task:
4277
	 *
L
Lucas De Marchi 已提交
4278
	 * To be able to change p->policy safely, the appropriate
L
Linus Torvalds 已提交
4279 4280
	 * runqueue lock must be held.
	 */
4281
	rq = task_rq_lock(p, &rf);
4282
	update_rq_clock(rq);
4283

4284
	/*
I
Ingo Molnar 已提交
4285
	 * Changing the policy of the stop threads its a very bad idea:
4286 4287
	 */
	if (p == rq->stop) {
4288
		task_rq_unlock(rq, p, &rf);
4289 4290 4291
		return -EINVAL;
	}

4292
	/*
4293 4294
	 * If not changing anything there's no need to proceed further,
	 * but store a possible modification of reset_on_fork.
4295
	 */
4296
	if (unlikely(policy == p->policy)) {
4297
		if (fair_policy(policy) && attr->sched_nice != task_nice(p))
4298 4299 4300
			goto change;
		if (rt_policy(policy) && attr->sched_priority != p->rt_priority)
			goto change;
4301
		if (dl_policy(policy) && dl_param_changed(p, attr))
4302
			goto change;
4303

4304
		p->sched_reset_on_fork = reset_on_fork;
4305
		task_rq_unlock(rq, p, &rf);
4306 4307
		return 0;
	}
4308
change:
4309

4310
	if (user) {
4311
#ifdef CONFIG_RT_GROUP_SCHED
4312 4313 4314 4315 4316
		/*
		 * Do not allow realtime tasks into groups that have no runtime
		 * assigned.
		 */
		if (rt_bandwidth_enabled() && rt_policy(policy) &&
4317 4318
				task_group(p)->rt_bandwidth.rt_runtime == 0 &&
				!task_group_is_autogroup(task_group(p))) {
4319
			task_rq_unlock(rq, p, &rf);
4320 4321 4322
			return -EPERM;
		}
#endif
4323 4324 4325 4326 4327 4328 4329 4330 4331
#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.
			 */
4332 4333
			if (!cpumask_subset(span, &p->cpus_allowed) ||
			    rq->rd->dl_bw.bw == 0) {
4334
				task_rq_unlock(rq, p, &rf);
4335 4336 4337 4338 4339
				return -EPERM;
			}
		}
#endif
	}
4340

I
Ingo Molnar 已提交
4341
	/* Re-check policy now with rq lock held: */
L
Linus Torvalds 已提交
4342 4343
	if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) {
		policy = oldpolicy = -1;
4344
		task_rq_unlock(rq, p, &rf);
L
Linus Torvalds 已提交
4345 4346
		goto recheck;
	}
4347 4348 4349 4350 4351 4352

	/*
	 * If setscheduling to SCHED_DEADLINE (or changing the parameters
	 * of a SCHED_DEADLINE task) we need to check if enough bandwidth
	 * is available.
	 */
4353
	if ((dl_policy(policy) || dl_task(p)) && dl_overflow(p, policy, attr)) {
4354
		task_rq_unlock(rq, p, &rf);
4355 4356 4357
		return -EBUSY;
	}

4358 4359 4360
	p->sched_reset_on_fork = reset_on_fork;
	oldprio = p->prio;

4361 4362 4363 4364 4365 4366 4367 4368
	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.
		 */
4369
		new_effective_prio = rt_effective_prio(p, newprio);
4370 4371
		if (new_effective_prio == oldprio)
			queue_flags &= ~DEQUEUE_MOVE;
4372 4373
	}

4374
	queued = task_on_rq_queued(p);
4375
	running = task_current(rq, p);
4376
	if (queued)
4377
		dequeue_task(rq, p, queue_flags);
4378
	if (running)
4379
		put_prev_task(rq, p);
4380

4381
	prev_class = p->sched_class;
4382
	__setscheduler(rq, p, attr, pi);
4383

4384
	if (queued) {
4385 4386 4387 4388
		/*
		 * We enqueue to tail when the priority of a task is
		 * increased (user space view).
		 */
4389 4390
		if (oldprio < p->prio)
			queue_flags |= ENQUEUE_HEAD;
4391

4392
		enqueue_task(rq, p, queue_flags);
4393
	}
4394
	if (running)
4395
		set_curr_task(rq, p);
4396

P
Peter Zijlstra 已提交
4397
	check_class_changed(rq, p, prev_class, oldprio);
I
Ingo Molnar 已提交
4398 4399 4400

	/* Avoid rq from going away on us: */
	preempt_disable();
4401
	task_rq_unlock(rq, p, &rf);
4402

4403 4404
	if (pi)
		rt_mutex_adjust_pi(p);
4405

I
Ingo Molnar 已提交
4406
	/* Run balance callbacks after we've adjusted the PI chain: */
4407 4408
	balance_callback(rq);
	preempt_enable();
4409

L
Linus Torvalds 已提交
4410 4411
	return 0;
}
4412

4413 4414 4415 4416 4417 4418 4419 4420 4421
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),
	};

4422 4423
	/* Fixup the legacy SCHED_RESET_ON_FORK hack. */
	if ((policy != SETPARAM_POLICY) && (policy & SCHED_RESET_ON_FORK)) {
4424 4425 4426 4427 4428
		attr.sched_flags |= SCHED_FLAG_RESET_ON_FORK;
		policy &= ~SCHED_RESET_ON_FORK;
		attr.sched_policy = policy;
	}

4429
	return __sched_setscheduler(p, &attr, check, true);
4430
}
4431 4432 4433 4434 4435 4436
/**
 * 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.
 *
4437 4438
 * Return: 0 on success. An error code otherwise.
 *
4439 4440 4441
 * NOTE that the task may be already dead.
 */
int sched_setscheduler(struct task_struct *p, int policy,
4442
		       const struct sched_param *param)
4443
{
4444
	return _sched_setscheduler(p, policy, param, true);
4445
}
L
Linus Torvalds 已提交
4446 4447
EXPORT_SYMBOL_GPL(sched_setscheduler);

4448 4449
int sched_setattr(struct task_struct *p, const struct sched_attr *attr)
{
4450
	return __sched_setscheduler(p, attr, true, true);
4451 4452 4453
}
EXPORT_SYMBOL_GPL(sched_setattr);

4454 4455 4456 4457 4458 4459 4460 4461 4462 4463
/**
 * 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.
4464 4465
 *
 * Return: 0 on success. An error code otherwise.
4466 4467
 */
int sched_setscheduler_nocheck(struct task_struct *p, int policy,
4468
			       const struct sched_param *param)
4469
{
4470
	return _sched_setscheduler(p, policy, param, false);
4471
}
4472
EXPORT_SYMBOL_GPL(sched_setscheduler_nocheck);
4473

I
Ingo Molnar 已提交
4474 4475
static int
do_sched_setscheduler(pid_t pid, int policy, struct sched_param __user *param)
L
Linus Torvalds 已提交
4476 4477 4478
{
	struct sched_param lparam;
	struct task_struct *p;
4479
	int retval;
L
Linus Torvalds 已提交
4480 4481 4482 4483 4484

	if (!param || pid < 0)
		return -EINVAL;
	if (copy_from_user(&lparam, param, sizeof(struct sched_param)))
		return -EFAULT;
4485 4486 4487

	rcu_read_lock();
	retval = -ESRCH;
L
Linus Torvalds 已提交
4488
	p = find_process_by_pid(pid);
4489 4490 4491
	if (p != NULL)
		retval = sched_setscheduler(p, policy, &lparam);
	rcu_read_unlock();
4492

L
Linus Torvalds 已提交
4493 4494 4495
	return retval;
}

4496 4497 4498
/*
 * Mimics kernel/events/core.c perf_copy_attr().
 */
I
Ingo Molnar 已提交
4499
static int sched_copy_attr(struct sched_attr __user *uattr, struct sched_attr *attr)
4500 4501 4502 4503 4504 4505 4506
{
	u32 size;
	int ret;

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

I
Ingo Molnar 已提交
4507
	/* Zero the full structure, so that a short copy will be nice: */
4508 4509 4510 4511 4512 4513
	memset(attr, 0, sizeof(*attr));

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

I
Ingo Molnar 已提交
4514 4515
	/* Bail out on silly large: */
	if (size > PAGE_SIZE)
4516 4517
		goto err_size;

I
Ingo Molnar 已提交
4518 4519
	/* ABI compatibility quirk: */
	if (!size)
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 4550 4551 4552 4553
		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
4554
	 * XXX: Do we want to be lenient like existing syscalls; or do we want
4555 4556
	 * to be strict and return an error on out-of-bounds values?
	 */
4557
	attr->sched_nice = clamp(attr->sched_nice, MIN_NICE, MAX_NICE);
4558

4559
	return 0;
4560 4561 4562

err_size:
	put_user(sizeof(*attr), &uattr->size);
4563
	return -E2BIG;
4564 4565
}

L
Linus Torvalds 已提交
4566 4567 4568 4569 4570
/**
 * 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.
4571 4572
 *
 * Return: 0 on success. An error code otherwise.
L
Linus Torvalds 已提交
4573
 */
I
Ingo Molnar 已提交
4574
SYSCALL_DEFINE3(sched_setscheduler, pid_t, pid, int, policy, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4575
{
4576 4577 4578
	if (policy < 0)
		return -EINVAL;

L
Linus Torvalds 已提交
4579 4580 4581 4582 4583 4584 4585
	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.
4586 4587
 *
 * Return: 0 on success. An error code otherwise.
L
Linus Torvalds 已提交
4588
 */
4589
SYSCALL_DEFINE2(sched_setparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4590
{
4591
	return do_sched_setscheduler(pid, SETPARAM_POLICY, param);
L
Linus Torvalds 已提交
4592 4593
}

4594 4595 4596
/**
 * sys_sched_setattr - same as above, but with extended sched_attr
 * @pid: the pid in question.
J
Juri Lelli 已提交
4597
 * @uattr: structure containing the extended parameters.
4598
 * @flags: for future extension.
4599
 */
4600 4601
SYSCALL_DEFINE3(sched_setattr, pid_t, pid, struct sched_attr __user *, uattr,
			       unsigned int, flags)
4602 4603 4604 4605 4606
{
	struct sched_attr attr;
	struct task_struct *p;
	int retval;

4607
	if (!uattr || pid < 0 || flags)
4608 4609
		return -EINVAL;

4610 4611 4612
	retval = sched_copy_attr(uattr, &attr);
	if (retval)
		return retval;
4613

4614
	if ((int)attr.sched_policy < 0)
4615
		return -EINVAL;
4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626

	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
4627 4628 4629
/**
 * sys_sched_getscheduler - get the policy (scheduling class) of a thread
 * @pid: the pid in question.
4630 4631 4632
 *
 * Return: On success, the policy of the thread. Otherwise, a negative error
 * code.
L
Linus Torvalds 已提交
4633
 */
4634
SYSCALL_DEFINE1(sched_getscheduler, pid_t, pid)
L
Linus Torvalds 已提交
4635
{
4636
	struct task_struct *p;
4637
	int retval;
L
Linus Torvalds 已提交
4638 4639

	if (pid < 0)
4640
		return -EINVAL;
L
Linus Torvalds 已提交
4641 4642

	retval = -ESRCH;
4643
	rcu_read_lock();
L
Linus Torvalds 已提交
4644 4645 4646 4647
	p = find_process_by_pid(pid);
	if (p) {
		retval = security_task_getscheduler(p);
		if (!retval)
4648 4649
			retval = p->policy
				| (p->sched_reset_on_fork ? SCHED_RESET_ON_FORK : 0);
L
Linus Torvalds 已提交
4650
	}
4651
	rcu_read_unlock();
L
Linus Torvalds 已提交
4652 4653 4654 4655
	return retval;
}

/**
4656
 * sys_sched_getparam - get the RT priority of a thread
L
Linus Torvalds 已提交
4657 4658
 * @pid: the pid in question.
 * @param: structure containing the RT priority.
4659 4660 4661
 *
 * Return: On success, 0 and the RT priority is in @param. Otherwise, an error
 * code.
L
Linus Torvalds 已提交
4662
 */
4663
SYSCALL_DEFINE2(sched_getparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4664
{
4665
	struct sched_param lp = { .sched_priority = 0 };
4666
	struct task_struct *p;
4667
	int retval;
L
Linus Torvalds 已提交
4668 4669

	if (!param || pid < 0)
4670
		return -EINVAL;
L
Linus Torvalds 已提交
4671

4672
	rcu_read_lock();
L
Linus Torvalds 已提交
4673 4674 4675 4676 4677 4678 4679 4680 4681
	p = find_process_by_pid(pid);
	retval = -ESRCH;
	if (!p)
		goto out_unlock;

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

4682 4683
	if (task_has_rt_policy(p))
		lp.sched_priority = p->rt_priority;
4684
	rcu_read_unlock();
L
Linus Torvalds 已提交
4685 4686 4687 4688 4689 4690 4691 4692 4693

	/*
	 * 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:
4694
	rcu_read_unlock();
L
Linus Torvalds 已提交
4695 4696 4697
	return retval;
}

4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715 4716 4717 4718 4719 4720
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)
4721
				return -EFBIG;
4722 4723 4724 4725 4726
		}

		attr->size = usize;
	}

4727
	ret = copy_to_user(uattr, attr, attr->size);
4728 4729 4730
	if (ret)
		return -EFAULT;

4731
	return 0;
4732 4733 4734
}

/**
4735
 * sys_sched_getattr - similar to sched_getparam, but with sched_attr
4736
 * @pid: the pid in question.
J
Juri Lelli 已提交
4737
 * @uattr: structure containing the extended parameters.
4738
 * @size: sizeof(attr) for fwd/bwd comp.
4739
 * @flags: for future extension.
4740
 */
4741 4742
SYSCALL_DEFINE4(sched_getattr, pid_t, pid, struct sched_attr __user *, uattr,
		unsigned int, size, unsigned int, flags)
4743 4744 4745 4746 4747 4748 4749 4750
{
	struct sched_attr attr = {
		.size = sizeof(struct sched_attr),
	};
	struct task_struct *p;
	int retval;

	if (!uattr || pid < 0 || size > PAGE_SIZE ||
4751
	    size < SCHED_ATTR_SIZE_VER0 || flags)
4752 4753 4754 4755 4756 4757 4758 4759 4760 4761 4762 4763 4764
		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;
4765 4766
	if (p->sched_reset_on_fork)
		attr.sched_flags |= SCHED_FLAG_RESET_ON_FORK;
4767 4768 4769
	if (task_has_dl_policy(p))
		__getparam_dl(p, &attr);
	else if (task_has_rt_policy(p))
4770 4771
		attr.sched_priority = p->rt_priority;
	else
4772
		attr.sched_nice = task_nice(p);
4773 4774 4775 4776 4777 4778 4779 4780 4781 4782 4783

	rcu_read_unlock();

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

out_unlock:
	rcu_read_unlock();
	return retval;
}

4784
long sched_setaffinity(pid_t pid, const struct cpumask *in_mask)
L
Linus Torvalds 已提交
4785
{
4786
	cpumask_var_t cpus_allowed, new_mask;
4787 4788
	struct task_struct *p;
	int retval;
L
Linus Torvalds 已提交
4789

4790
	rcu_read_lock();
L
Linus Torvalds 已提交
4791 4792 4793

	p = find_process_by_pid(pid);
	if (!p) {
4794
		rcu_read_unlock();
L
Linus Torvalds 已提交
4795 4796 4797
		return -ESRCH;
	}

4798
	/* Prevent p going away */
L
Linus Torvalds 已提交
4799
	get_task_struct(p);
4800
	rcu_read_unlock();
L
Linus Torvalds 已提交
4801

4802 4803 4804 4805
	if (p->flags & PF_NO_SETAFFINITY) {
		retval = -EINVAL;
		goto out_put_task;
	}
4806 4807 4808 4809 4810 4811 4812 4813
	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
4814
	retval = -EPERM;
E
Eric W. Biederman 已提交
4815 4816 4817 4818
	if (!check_same_owner(p)) {
		rcu_read_lock();
		if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
			rcu_read_unlock();
4819
			goto out_free_new_mask;
E
Eric W. Biederman 已提交
4820 4821 4822
		}
		rcu_read_unlock();
	}
L
Linus Torvalds 已提交
4823

4824
	retval = security_task_setscheduler(p);
4825
	if (retval)
4826
		goto out_free_new_mask;
4827

4828 4829 4830 4831

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

4832 4833 4834 4835 4836 4837 4838
	/*
	 * 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
4839 4840 4841
	if (task_has_dl_policy(p) && dl_bandwidth_enabled()) {
		rcu_read_lock();
		if (!cpumask_subset(task_rq(p)->rd->span, new_mask)) {
4842
			retval = -EBUSY;
4843
			rcu_read_unlock();
4844
			goto out_free_new_mask;
4845
		}
4846
		rcu_read_unlock();
4847 4848
	}
#endif
P
Peter Zijlstra 已提交
4849
again:
4850
	retval = __set_cpus_allowed_ptr(p, new_mask, true);
L
Linus Torvalds 已提交
4851

P
Paul Menage 已提交
4852
	if (!retval) {
4853 4854
		cpuset_cpus_allowed(p, cpus_allowed);
		if (!cpumask_subset(new_mask, cpus_allowed)) {
P
Paul Menage 已提交
4855 4856 4857 4858 4859
			/*
			 * We must have raced with a concurrent cpuset
			 * update. Just reset the cpus_allowed to the
			 * cpuset's cpus_allowed
			 */
4860
			cpumask_copy(new_mask, cpus_allowed);
P
Paul Menage 已提交
4861 4862 4863
			goto again;
		}
	}
4864
out_free_new_mask:
4865 4866 4867 4868
	free_cpumask_var(new_mask);
out_free_cpus_allowed:
	free_cpumask_var(cpus_allowed);
out_put_task:
L
Linus Torvalds 已提交
4869 4870 4871 4872 4873
	put_task_struct(p);
	return retval;
}

static int get_user_cpu_mask(unsigned long __user *user_mask_ptr, unsigned len,
4874
			     struct cpumask *new_mask)
L
Linus Torvalds 已提交
4875
{
4876 4877 4878 4879 4880
	if (len < cpumask_size())
		cpumask_clear(new_mask);
	else if (len > cpumask_size())
		len = cpumask_size();

L
Linus Torvalds 已提交
4881 4882 4883 4884
	return copy_from_user(new_mask, user_mask_ptr, len) ? -EFAULT : 0;
}

/**
I
Ingo Molnar 已提交
4885
 * sys_sched_setaffinity - set the CPU affinity of a process
L
Linus Torvalds 已提交
4886 4887
 * @pid: pid of the process
 * @len: length in bytes of the bitmask pointed to by user_mask_ptr
I
Ingo Molnar 已提交
4888
 * @user_mask_ptr: user-space pointer to the new CPU mask
4889 4890
 *
 * Return: 0 on success. An error code otherwise.
L
Linus Torvalds 已提交
4891
 */
4892 4893
SYSCALL_DEFINE3(sched_setaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
4894
{
4895
	cpumask_var_t new_mask;
L
Linus Torvalds 已提交
4896 4897
	int retval;

4898 4899
	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
4900

4901 4902 4903 4904 4905
	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
4906 4907
}

4908
long sched_getaffinity(pid_t pid, struct cpumask *mask)
L
Linus Torvalds 已提交
4909
{
4910
	struct task_struct *p;
4911
	unsigned long flags;
L
Linus Torvalds 已提交
4912 4913
	int retval;

4914
	rcu_read_lock();
L
Linus Torvalds 已提交
4915 4916 4917 4918 4919 4920

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

4921 4922 4923 4924
	retval = security_task_getscheduler(p);
	if (retval)
		goto out_unlock;

4925
	raw_spin_lock_irqsave(&p->pi_lock, flags);
4926
	cpumask_and(mask, &p->cpus_allowed, cpu_active_mask);
4927
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
4928 4929

out_unlock:
4930
	rcu_read_unlock();
L
Linus Torvalds 已提交
4931

4932
	return retval;
L
Linus Torvalds 已提交
4933 4934 4935
}

/**
I
Ingo Molnar 已提交
4936
 * sys_sched_getaffinity - get the CPU affinity of a process
L
Linus Torvalds 已提交
4937 4938
 * @pid: pid of the process
 * @len: length in bytes of the bitmask pointed to by user_mask_ptr
I
Ingo Molnar 已提交
4939
 * @user_mask_ptr: user-space pointer to hold the current CPU mask
4940
 *
4941 4942
 * Return: size of CPU mask copied to user_mask_ptr on success. An
 * error code otherwise.
L
Linus Torvalds 已提交
4943
 */
4944 4945
SYSCALL_DEFINE3(sched_getaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
4946 4947
{
	int ret;
4948
	cpumask_var_t mask;
L
Linus Torvalds 已提交
4949

A
Anton Blanchard 已提交
4950
	if ((len * BITS_PER_BYTE) < nr_cpu_ids)
4951 4952
		return -EINVAL;
	if (len & (sizeof(unsigned long)-1))
L
Linus Torvalds 已提交
4953 4954
		return -EINVAL;

4955 4956
	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
4957

4958 4959
	ret = sched_getaffinity(pid, mask);
	if (ret == 0) {
4960
		size_t retlen = min_t(size_t, len, cpumask_size());
4961 4962

		if (copy_to_user(user_mask_ptr, mask, retlen))
4963 4964
			ret = -EFAULT;
		else
4965
			ret = retlen;
4966 4967
	}
	free_cpumask_var(mask);
L
Linus Torvalds 已提交
4968

4969
	return ret;
L
Linus Torvalds 已提交
4970 4971 4972 4973 4974
}

/**
 * sys_sched_yield - yield the current processor to other threads.
 *
I
Ingo Molnar 已提交
4975 4976
 * This function yields the current CPU to other tasks. If there are no
 * other threads running on this CPU then this function will return.
4977 4978
 *
 * Return: 0.
L
Linus Torvalds 已提交
4979
 */
4980
SYSCALL_DEFINE0(sched_yield)
L
Linus Torvalds 已提交
4981
{
4982 4983 4984 4985 4986 4987
	struct rq_flags rf;
	struct rq *rq;

	local_irq_disable();
	rq = this_rq();
	rq_lock(rq, &rf);
L
Linus Torvalds 已提交
4988

4989
	schedstat_inc(rq->yld_count);
4990
	current->sched_class->yield_task(rq);
L
Linus Torvalds 已提交
4991 4992 4993 4994 4995

	/*
	 * Since we are going to call schedule() anyway, there's
	 * no need to preempt or enable interrupts:
	 */
4996 4997
	preempt_disable();
	rq_unlock(rq, &rf);
4998
	sched_preempt_enable_no_resched();
L
Linus Torvalds 已提交
4999 5000 5001 5002 5003 5004

	schedule();

	return 0;
}

5005
#ifndef CONFIG_PREEMPT
5006
int __sched _cond_resched(void)
L
Linus Torvalds 已提交
5007
{
5008
	if (should_resched(0)) {
5009
		preempt_schedule_common();
L
Linus Torvalds 已提交
5010 5011 5012 5013
		return 1;
	}
	return 0;
}
5014
EXPORT_SYMBOL(_cond_resched);
5015
#endif
L
Linus Torvalds 已提交
5016 5017

/*
5018
 * __cond_resched_lock() - if a reschedule is pending, drop the given lock,
L
Linus Torvalds 已提交
5019 5020
 * call schedule, and on return reacquire the lock.
 *
I
Ingo Molnar 已提交
5021
 * This works OK both with and without CONFIG_PREEMPT. We do strange low-level
L
Linus Torvalds 已提交
5022 5023 5024
 * operations here to prevent schedule() from being called twice (once via
 * spin_unlock(), once by hand).
 */
5025
int __cond_resched_lock(spinlock_t *lock)
L
Linus Torvalds 已提交
5026
{
5027
	int resched = should_resched(PREEMPT_LOCK_OFFSET);
J
Jan Kara 已提交
5028 5029
	int ret = 0;

5030 5031
	lockdep_assert_held(lock);

5032
	if (spin_needbreak(lock) || resched) {
L
Linus Torvalds 已提交
5033
		spin_unlock(lock);
P
Peter Zijlstra 已提交
5034
		if (resched)
5035
			preempt_schedule_common();
N
Nick Piggin 已提交
5036 5037
		else
			cpu_relax();
J
Jan Kara 已提交
5038
		ret = 1;
L
Linus Torvalds 已提交
5039 5040
		spin_lock(lock);
	}
J
Jan Kara 已提交
5041
	return ret;
L
Linus Torvalds 已提交
5042
}
5043
EXPORT_SYMBOL(__cond_resched_lock);
L
Linus Torvalds 已提交
5044

5045
int __sched __cond_resched_softirq(void)
L
Linus Torvalds 已提交
5046 5047 5048
{
	BUG_ON(!in_softirq());

5049
	if (should_resched(SOFTIRQ_DISABLE_OFFSET)) {
5050
		local_bh_enable();
5051
		preempt_schedule_common();
L
Linus Torvalds 已提交
5052 5053 5054 5055 5056
		local_bh_disable();
		return 1;
	}
	return 0;
}
5057
EXPORT_SYMBOL(__cond_resched_softirq);
L
Linus Torvalds 已提交
5058 5059 5060 5061

/**
 * yield - yield the current processor to other threads.
 *
P
Peter Zijlstra 已提交
5062 5063 5064 5065 5066 5067 5068 5069 5070
 * 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
5071
 *	yield();
P
Peter Zijlstra 已提交
5072 5073 5074 5075 5076 5077 5078 5079
 *
 * 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
5080 5081 5082 5083 5084 5085 5086 5087
 */
void __sched yield(void)
{
	set_current_state(TASK_RUNNING);
	sys_sched_yield();
}
EXPORT_SYMBOL(yield);

5088 5089 5090 5091
/**
 * 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
5092 5093
 * @p: target task
 * @preempt: whether task preemption is allowed or not
5094 5095 5096 5097
 *
 * 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.
 *
5098
 * Return:
5099 5100 5101
 *	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.
5102
 */
5103
int __sched yield_to(struct task_struct *p, bool preempt)
5104 5105 5106 5107
{
	struct task_struct *curr = current;
	struct rq *rq, *p_rq;
	unsigned long flags;
5108
	int yielded = 0;
5109 5110 5111 5112 5113 5114

	local_irq_save(flags);
	rq = this_rq();

again:
	p_rq = task_rq(p);
5115 5116 5117 5118 5119 5120 5121 5122 5123
	/*
	 * 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;
	}

5124
	double_rq_lock(rq, p_rq);
5125
	if (task_rq(p) != p_rq) {
5126 5127 5128 5129 5130
		double_rq_unlock(rq, p_rq);
		goto again;
	}

	if (!curr->sched_class->yield_to_task)
5131
		goto out_unlock;
5132 5133

	if (curr->sched_class != p->sched_class)
5134
		goto out_unlock;
5135 5136

	if (task_running(p_rq, p) || p->state)
5137
		goto out_unlock;
5138 5139

	yielded = curr->sched_class->yield_to_task(rq, p, preempt);
5140
	if (yielded) {
5141
		schedstat_inc(rq->yld_count);
5142 5143 5144 5145 5146
		/*
		 * Make p's CPU reschedule; pick_next_entity takes care of
		 * fairness.
		 */
		if (preempt && rq != p_rq)
5147
			resched_curr(p_rq);
5148
	}
5149

5150
out_unlock:
5151
	double_rq_unlock(rq, p_rq);
5152
out_irq:
5153 5154
	local_irq_restore(flags);

5155
	if (yielded > 0)
5156 5157 5158 5159 5160 5161
		schedule();

	return yielded;
}
EXPORT_SYMBOL_GPL(yield_to);

5162 5163 5164 5165 5166 5167 5168 5169 5170 5171 5172 5173 5174 5175 5176
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
5177
/*
I
Ingo Molnar 已提交
5178
 * This task is about to go to sleep on IO. Increment rq->nr_iowait so
L
Linus Torvalds 已提交
5179 5180 5181 5182
 * that process accounting knows that this is a task in IO wait state.
 */
long __sched io_schedule_timeout(long timeout)
{
5183
	int token;
L
Linus Torvalds 已提交
5184 5185
	long ret;

5186
	token = io_schedule_prepare();
L
Linus Torvalds 已提交
5187
	ret = schedule_timeout(timeout);
5188
	io_schedule_finish(token);
5189

L
Linus Torvalds 已提交
5190 5191
	return ret;
}
5192
EXPORT_SYMBOL(io_schedule_timeout);
L
Linus Torvalds 已提交
5193

5194 5195 5196 5197 5198 5199 5200 5201 5202 5203
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
5204 5205 5206 5207
/**
 * sys_sched_get_priority_max - return maximum RT priority.
 * @policy: scheduling class.
 *
5208 5209 5210
 * 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
5211
 */
5212
SYSCALL_DEFINE1(sched_get_priority_max, int, policy)
L
Linus Torvalds 已提交
5213 5214 5215 5216 5217 5218 5219 5220
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = MAX_USER_RT_PRIO-1;
		break;
5221
	case SCHED_DEADLINE:
L
Linus Torvalds 已提交
5222
	case SCHED_NORMAL:
5223
	case SCHED_BATCH:
I
Ingo Molnar 已提交
5224
	case SCHED_IDLE:
L
Linus Torvalds 已提交
5225 5226 5227 5228 5229 5230 5231 5232 5233 5234
		ret = 0;
		break;
	}
	return ret;
}

/**
 * sys_sched_get_priority_min - return minimum RT priority.
 * @policy: scheduling class.
 *
5235 5236 5237
 * 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
5238
 */
5239
SYSCALL_DEFINE1(sched_get_priority_min, int, policy)
L
Linus Torvalds 已提交
5240 5241 5242 5243 5244 5245 5246 5247
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = 1;
		break;
5248
	case SCHED_DEADLINE:
L
Linus Torvalds 已提交
5249
	case SCHED_NORMAL:
5250
	case SCHED_BATCH:
I
Ingo Molnar 已提交
5251
	case SCHED_IDLE:
L
Linus Torvalds 已提交
5252 5253 5254 5255 5256 5257 5258 5259 5260 5261 5262 5263
		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.
5264 5265 5266
 *
 * Return: On success, 0 and the timeslice is in @interval. Otherwise,
 * an error code.
L
Linus Torvalds 已提交
5267
 */
5268
SYSCALL_DEFINE2(sched_rr_get_interval, pid_t, pid,
5269
		struct timespec __user *, interval)
L
Linus Torvalds 已提交
5270
{
5271
	struct task_struct *p;
D
Dmitry Adamushko 已提交
5272
	unsigned int time_slice;
5273 5274
	struct rq_flags rf;
	struct timespec t;
5275
	struct rq *rq;
5276
	int retval;
L
Linus Torvalds 已提交
5277 5278

	if (pid < 0)
5279
		return -EINVAL;
L
Linus Torvalds 已提交
5280 5281

	retval = -ESRCH;
5282
	rcu_read_lock();
L
Linus Torvalds 已提交
5283 5284 5285 5286 5287 5288 5289 5290
	p = find_process_by_pid(pid);
	if (!p)
		goto out_unlock;

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

5291
	rq = task_rq_lock(p, &rf);
5292 5293 5294
	time_slice = 0;
	if (p->sched_class->get_rr_interval)
		time_slice = p->sched_class->get_rr_interval(rq, p);
5295
	task_rq_unlock(rq, p, &rf);
D
Dmitry Adamushko 已提交
5296

5297
	rcu_read_unlock();
D
Dmitry Adamushko 已提交
5298
	jiffies_to_timespec(time_slice, &t);
L
Linus Torvalds 已提交
5299 5300
	retval = copy_to_user(interval, &t, sizeof(t)) ? -EFAULT : 0;
	return retval;
5301

L
Linus Torvalds 已提交
5302
out_unlock:
5303
	rcu_read_unlock();
L
Linus Torvalds 已提交
5304 5305 5306
	return retval;
}

5307
static const char stat_nam[] = TASK_STATE_TO_CHAR_STR;
5308

5309
void sched_show_task(struct task_struct *p)
L
Linus Torvalds 已提交
5310 5311
{
	unsigned long free = 0;
5312
	int ppid;
5313
	unsigned long state = p->state;
L
Linus Torvalds 已提交
5314

5315 5316 5317
	/* 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);

5318 5319
	if (!try_get_task_stack(p))
		return;
5320 5321
	if (state)
		state = __ffs(state) + 1;
5322
	printk(KERN_INFO "%-15.15s %c", p->comm,
5323
		state < sizeof(stat_nam) - 1 ? stat_nam[state] : '?');
L
Linus Torvalds 已提交
5324
	if (state == TASK_RUNNING)
P
Peter Zijlstra 已提交
5325
		printk(KERN_CONT "  running task    ");
L
Linus Torvalds 已提交
5326
#ifdef CONFIG_DEBUG_STACK_USAGE
5327
	free = stack_not_used(p);
L
Linus Torvalds 已提交
5328
#endif
5329
	ppid = 0;
5330
	rcu_read_lock();
5331 5332
	if (pid_alive(p))
		ppid = task_pid_nr(rcu_dereference(p->real_parent));
5333
	rcu_read_unlock();
P
Peter Zijlstra 已提交
5334
	printk(KERN_CONT "%5lu %5d %6d 0x%08lx\n", free,
5335
		task_pid_nr(p), ppid,
5336
		(unsigned long)task_thread_info(p)->flags);
L
Linus Torvalds 已提交
5337

5338
	print_worker_info(KERN_INFO, p);
5339
	show_stack(p, NULL);
5340
	put_task_stack(p);
L
Linus Torvalds 已提交
5341 5342
}

I
Ingo Molnar 已提交
5343
void show_state_filter(unsigned long state_filter)
L
Linus Torvalds 已提交
5344
{
5345
	struct task_struct *g, *p;
L
Linus Torvalds 已提交
5346

5347
#if BITS_PER_LONG == 32
P
Peter Zijlstra 已提交
5348 5349
	printk(KERN_INFO
		"  task                PC stack   pid father\n");
L
Linus Torvalds 已提交
5350
#else
P
Peter Zijlstra 已提交
5351 5352
	printk(KERN_INFO
		"  task                        PC stack   pid father\n");
L
Linus Torvalds 已提交
5353
#endif
5354
	rcu_read_lock();
5355
	for_each_process_thread(g, p) {
L
Linus Torvalds 已提交
5356 5357
		/*
		 * reset the NMI-timeout, listing all files on a slow
L
Lucas De Marchi 已提交
5358
		 * console might take a lot of time:
5359 5360 5361
		 * 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
5362 5363
		 */
		touch_nmi_watchdog();
5364
		touch_all_softlockup_watchdogs();
I
Ingo Molnar 已提交
5365
		if (!state_filter || (p->state & state_filter))
5366
			sched_show_task(p);
5367
	}
L
Linus Torvalds 已提交
5368

I
Ingo Molnar 已提交
5369
#ifdef CONFIG_SCHED_DEBUG
5370 5371
	if (!state_filter)
		sysrq_sched_debug_show();
I
Ingo Molnar 已提交
5372
#endif
5373
	rcu_read_unlock();
I
Ingo Molnar 已提交
5374 5375 5376
	/*
	 * Only show locks if all tasks are dumped:
	 */
5377
	if (!state_filter)
I
Ingo Molnar 已提交
5378
		debug_show_all_locks();
L
Linus Torvalds 已提交
5379 5380
}

5381
void init_idle_bootup_task(struct task_struct *idle)
I
Ingo Molnar 已提交
5382
{
I
Ingo Molnar 已提交
5383
	idle->sched_class = &idle_sched_class;
I
Ingo Molnar 已提交
5384 5385
}

5386 5387 5388
/**
 * init_idle - set up an idle thread for a given CPU
 * @idle: task in question
I
Ingo Molnar 已提交
5389
 * @cpu: CPU the idle task belongs to
5390 5391 5392 5393
 *
 * NOTE: this function does not set the idle thread's NEED_RESCHED
 * flag, to make booting more robust.
 */
5394
void init_idle(struct task_struct *idle, int cpu)
L
Linus Torvalds 已提交
5395
{
5396
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
5397 5398
	unsigned long flags;

5399 5400
	raw_spin_lock_irqsave(&idle->pi_lock, flags);
	raw_spin_lock(&rq->lock);
5401

5402
	__sched_fork(0, idle);
5403
	idle->state = TASK_RUNNING;
I
Ingo Molnar 已提交
5404
	idle->se.exec_start = sched_clock();
5405
	idle->flags |= PF_IDLE;
I
Ingo Molnar 已提交
5406

5407 5408
	kasan_unpoison_task_stack(idle);

5409 5410 5411 5412 5413 5414 5415 5416 5417
#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
5418 5419
	/*
	 * We're having a chicken and egg problem, even though we are
I
Ingo Molnar 已提交
5420
	 * holding rq->lock, the CPU isn't yet set to this CPU so the
5421 5422 5423 5424 5425 5426 5427 5428
	 * 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
5429
	__set_task_cpu(idle, cpu);
5430
	rcu_read_unlock();
L
Linus Torvalds 已提交
5431 5432

	rq->curr = rq->idle = idle;
5433
	idle->on_rq = TASK_ON_RQ_QUEUED;
5434
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
5435
	idle->on_cpu = 1;
5436
#endif
5437 5438
	raw_spin_unlock(&rq->lock);
	raw_spin_unlock_irqrestore(&idle->pi_lock, flags);
L
Linus Torvalds 已提交
5439 5440

	/* Set the preempt count _outside_ the spinlocks! */
5441
	init_idle_preempt_count(idle, cpu);
5442

I
Ingo Molnar 已提交
5443 5444 5445 5446
	/*
	 * The idle tasks have their own, simple scheduling class:
	 */
	idle->sched_class = &idle_sched_class;
5447
	ftrace_graph_init_idle_task(idle, cpu);
5448
	vtime_init_idle(idle, cpu);
5449
#ifdef CONFIG_SMP
5450 5451
	sprintf(idle->comm, "%s/%d", INIT_TASK_COMM, cpu);
#endif
I
Ingo Molnar 已提交
5452 5453
}

5454 5455
#ifdef CONFIG_SMP

5456 5457 5458 5459 5460 5461 5462
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;

5463 5464 5465
	if (!cpumask_weight(cur))
		return ret;

5466
	rcu_read_lock_sched();
5467 5468 5469 5470 5471 5472 5473 5474
	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);
5475
	rcu_read_unlock_sched();
5476 5477 5478 5479

	return ret;
}

5480 5481 5482 5483 5484 5485 5486
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
5487
	 * to a new cpuset; we don't want to change their CPU
5488 5489 5490 5491 5492 5493 5494 5495 5496 5497 5498 5499 5500 5501 5502
	 * 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;
	}

	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);
5503
		struct dl_bw *dl_b;
5504 5505 5506 5507
		bool overflow;
		int cpus;
		unsigned long flags;

5508 5509
		rcu_read_lock_sched();
		dl_b = dl_bw_of(dest_cpu);
5510 5511 5512 5513 5514 5515 5516 5517 5518 5519 5520 5521
		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()).
			 */
5522
			__dl_add(dl_b, p->dl.dl_bw, cpus);
5523 5524
		}
		raw_spin_unlock_irqrestore(&dl_b->lock, flags);
5525
		rcu_read_unlock_sched();
5526 5527

	}
5528

5529 5530 5531 5532
out:
	return ret;
}

5533
bool sched_smp_initialized __read_mostly;
5534

5535 5536 5537 5538 5539 5540 5541 5542 5543 5544
#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;

5545
	if (!cpumask_test_cpu(target_cpu, &p->cpus_allowed))
5546 5547 5548 5549
		return -EINVAL;

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

5550
	trace_sched_move_numa(p, curr_cpu, target_cpu);
5551 5552
	return stop_one_cpu(curr_cpu, migration_cpu_stop, &arg);
}
5553 5554 5555 5556 5557 5558 5559

/*
 * 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)
{
5560
	bool queued, running;
5561 5562
	struct rq_flags rf;
	struct rq *rq;
5563

5564
	rq = task_rq_lock(p, &rf);
5565
	queued = task_on_rq_queued(p);
5566 5567
	running = task_current(rq, p);

5568
	if (queued)
5569
		dequeue_task(rq, p, DEQUEUE_SAVE);
5570
	if (running)
5571
		put_prev_task(rq, p);
5572 5573 5574

	p->numa_preferred_nid = nid;

5575
	if (queued)
5576
		enqueue_task(rq, p, ENQUEUE_RESTORE | ENQUEUE_NOCLOCK);
5577
	if (running)
5578
		set_curr_task(rq, p);
5579
	task_rq_unlock(rq, p, &rf);
5580
}
P
Peter Zijlstra 已提交
5581
#endif /* CONFIG_NUMA_BALANCING */
5582

L
Linus Torvalds 已提交
5583
#ifdef CONFIG_HOTPLUG_CPU
5584
/*
I
Ingo Molnar 已提交
5585
 * Ensure that the idle task is using init_mm right before its CPU goes
5586
 * offline.
5587
 */
5588
void idle_task_exit(void)
L
Linus Torvalds 已提交
5589
{
5590
	struct mm_struct *mm = current->active_mm;
5591

5592
	BUG_ON(cpu_online(smp_processor_id()));
5593

5594
	if (mm != &init_mm) {
5595
		switch_mm(mm, &init_mm, current);
5596 5597
		finish_arch_post_lock_switch();
	}
5598
	mmdrop(mm);
L
Linus Torvalds 已提交
5599 5600 5601
}

/*
5602 5603
 * 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
5604 5605 5606
 * 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.
5607 5608
 *
 * Also see the comment "Global load-average calculations".
L
Linus Torvalds 已提交
5609
 */
5610
static void calc_load_migrate(struct rq *rq)
L
Linus Torvalds 已提交
5611
{
5612
	long delta = calc_load_fold_active(rq, 1);
5613 5614
	if (delta)
		atomic_long_add(delta, &calc_load_tasks);
L
Linus Torvalds 已提交
5615 5616
}

5617 5618 5619 5620 5621 5622 5623 5624 5625 5626 5627 5628 5629 5630 5631 5632
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,
};

5633
/*
5634 5635 5636 5637 5638 5639
 * 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
5640
 */
5641
static void migrate_tasks(struct rq *dead_rq, struct rq_flags *rf)
L
Linus Torvalds 已提交
5642
{
5643
	struct rq *rq = dead_rq;
5644
	struct task_struct *next, *stop = rq->stop;
5645
	struct rq_flags orf = *rf;
5646
	int dest_cpu;
L
Linus Torvalds 已提交
5647 5648

	/*
5649 5650 5651 5652 5653 5654 5655
	 * 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
5656
	 */
5657
	rq->stop = NULL;
5658

5659 5660 5661 5662 5663 5664 5665
	/*
	 * 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);

5666
	for (;;) {
5667 5668
		/*
		 * There's this thread running, bail when that's the only
I
Ingo Molnar 已提交
5669
		 * remaining thread:
5670 5671
		 */
		if (rq->nr_running == 1)
I
Ingo Molnar 已提交
5672
			break;
5673

5674
		/*
I
Ingo Molnar 已提交
5675
		 * pick_next_task() assumes pinned rq->lock:
5676
		 */
5677
		next = pick_next_task(rq, &fake_task, rf);
5678
		BUG_ON(!next);
D
Dmitry Adamushko 已提交
5679
		next->sched_class->put_prev_task(rq, next);
5680

W
Wanpeng Li 已提交
5681 5682 5683 5684 5685 5686 5687 5688 5689
		/*
		 * 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.
		 */
5690
		rq_unlock(rq, rf);
W
Wanpeng Li 已提交
5691
		raw_spin_lock(&next->pi_lock);
5692
		rq_relock(rq, rf);
W
Wanpeng Li 已提交
5693 5694 5695 5696 5697 5698 5699 5700 5701 5702 5703

		/*
		 * 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;
		}

5704
		/* Find suitable destination for @next, with force if needed. */
5705
		dest_cpu = select_fallback_rq(dead_rq->cpu, next);
5706
		rq = __migrate_task(rq, rf, next, dest_cpu);
5707
		if (rq != dead_rq) {
5708
			rq_unlock(rq, rf);
5709
			rq = dead_rq;
5710 5711
			*rf = orf;
			rq_relock(rq, rf);
5712
		}
W
Wanpeng Li 已提交
5713
		raw_spin_unlock(&next->pi_lock);
L
Linus Torvalds 已提交
5714
	}
5715

5716
	rq->stop = stop;
5717
}
L
Linus Torvalds 已提交
5718 5719
#endif /* CONFIG_HOTPLUG_CPU */

5720
void set_rq_online(struct rq *rq)
5721 5722 5723 5724
{
	if (!rq->online) {
		const struct sched_class *class;

5725
		cpumask_set_cpu(rq->cpu, rq->rd->online);
5726 5727 5728 5729 5730 5731 5732 5733 5734
		rq->online = 1;

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

5735
void set_rq_offline(struct rq *rq)
5736 5737 5738 5739 5740 5741 5742 5743 5744
{
	if (rq->online) {
		const struct sched_class *class;

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

5745
		cpumask_clear_cpu(rq->cpu, rq->rd->online);
5746 5747 5748 5749
		rq->online = 0;
	}
}

5750
static void set_cpu_rq_start_time(unsigned int cpu)
L
Linus Torvalds 已提交
5751
{
5752
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
5753

5754 5755 5756
	rq->age_stamp = sched_clock_cpu(cpu);
}

I
Ingo Molnar 已提交
5757 5758 5759 5760
/*
 * used to mark begin/end of suspend/resume:
 */
static int num_cpus_frozen;
5761

L
Linus Torvalds 已提交
5762
/*
5763 5764 5765
 * Update cpusets according to cpu_active mask.  If cpusets are
 * disabled, cpuset_update_active_cpus() becomes a simple wrapper
 * around partition_sched_domains().
5766 5767 5768
 *
 * 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
5769
 */
5770
static void cpuset_cpu_active(void)
5771
{
5772
	if (cpuhp_tasks_frozen) {
5773 5774 5775 5776 5777 5778 5779 5780 5781
		/*
		 * 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);
5782
			return;
5783 5784 5785 5786 5787 5788
		}
		/*
		 * This is the last CPU online operation. So fall through and
		 * restore the original sched domains by considering the
		 * cpuset configurations.
		 */
5789
	}
5790
	cpuset_update_active_cpus();
5791
}
5792

5793
static int cpuset_cpu_inactive(unsigned int cpu)
5794
{
5795 5796
	unsigned long flags;
	struct dl_bw *dl_b;
5797 5798
	bool overflow;
	int cpus;
5799

5800
	if (!cpuhp_tasks_frozen) {
5801 5802
		rcu_read_lock_sched();
		dl_b = dl_bw_of(cpu);
5803

5804 5805 5806 5807
		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);
5808

5809
		rcu_read_unlock_sched();
5810

5811
		if (overflow)
5812
			return -EBUSY;
5813
		cpuset_update_active_cpus();
5814
	} else {
5815 5816
		num_cpus_frozen++;
		partition_sched_domains(1, NULL, NULL);
5817
	}
5818
	return 0;
5819 5820
}

5821
int sched_cpu_activate(unsigned int cpu)
5822
{
5823
	struct rq *rq = cpu_rq(cpu);
5824
	struct rq_flags rf;
5825

5826
	set_cpu_active(cpu, true);
5827

5828
	if (sched_smp_initialized) {
5829
		sched_domains_numa_masks_set(cpu);
5830
		cpuset_cpu_active();
5831
	}
5832 5833 5834 5835 5836

	/*
	 * 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
5837
	 *    after all CPUs have been brought up.
5838 5839 5840 5841
	 *
	 * 2) At runtime, if cpuset_cpu_active() fails to rebuild the
	 *    domains.
	 */
5842
	rq_lock_irqsave(rq, &rf);
5843 5844 5845 5846
	if (rq->rd) {
		BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
		set_rq_online(rq);
	}
5847
	rq_unlock_irqrestore(rq, &rf);
5848 5849 5850

	update_max_interval();

5851
	return 0;
5852 5853
}

5854
int sched_cpu_deactivate(unsigned int cpu)
5855 5856 5857
{
	int ret;

5858
	set_cpu_active(cpu, false);
5859 5860 5861 5862 5863 5864 5865 5866 5867 5868 5869 5870 5871 5872
	/*
	 * 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();
5873 5874 5875 5876 5877 5878 5879 5880

	if (!sched_smp_initialized)
		return 0;

	ret = cpuset_cpu_inactive(cpu);
	if (ret) {
		set_cpu_active(cpu, true);
		return ret;
5881
	}
5882 5883
	sched_domains_numa_masks_clear(cpu);
	return 0;
5884 5885
}

5886 5887 5888 5889 5890 5891 5892 5893
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();
}

5894 5895 5896
int sched_cpu_starting(unsigned int cpu)
{
	set_cpu_rq_start_time(cpu);
5897
	sched_rq_cpu_starting(cpu);
5898
	return 0;
5899 5900
}

5901 5902 5903 5904
#ifdef CONFIG_HOTPLUG_CPU
int sched_cpu_dying(unsigned int cpu)
{
	struct rq *rq = cpu_rq(cpu);
5905
	struct rq_flags rf;
5906 5907 5908

	/* Handle pending wakeups and then migrate everything off */
	sched_ttwu_pending();
5909 5910

	rq_lock_irqsave(rq, &rf);
5911 5912 5913 5914
	if (rq->rd) {
		BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
		set_rq_offline(rq);
	}
5915
	migrate_tasks(rq, &rf);
5916
	BUG_ON(rq->nr_running != 1);
5917 5918
	rq_unlock_irqrestore(rq, &rf);

5919 5920
	calc_load_migrate(rq);
	update_max_interval();
5921
	nohz_balance_exit_idle(cpu);
5922
	hrtick_clear(rq);
5923 5924 5925 5926
	return 0;
}
#endif

P
Peter Zijlstra 已提交
5927 5928 5929 5930 5931 5932 5933 5934 5935 5936 5937 5938 5939 5940 5941 5942
#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
5943 5944
void __init sched_init_smp(void)
{
5945 5946 5947
	cpumask_var_t non_isolated_cpus;

	alloc_cpumask_var(&non_isolated_cpus, GFP_KERNEL);
5948

5949 5950
	sched_init_numa();

5951 5952
	/*
	 * There's no userspace yet to cause hotplug operations; hence all the
I
Ingo Molnar 已提交
5953
	 * CPU masks are stable and all blatant races in the below code cannot
5954 5955
	 * happen.
	 */
5956
	mutex_lock(&sched_domains_mutex);
P
Peter Zijlstra 已提交
5957
	sched_init_domains(cpu_active_mask);
5958 5959 5960
	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);
5961
	mutex_unlock(&sched_domains_mutex);
5962

5963
	/* Move init over to a non-isolated CPU */
5964
	if (set_cpus_allowed_ptr(current, non_isolated_cpus) < 0)
5965
		BUG();
I
Ingo Molnar 已提交
5966
	sched_init_granularity();
5967
	free_cpumask_var(non_isolated_cpus);
5968

5969
	init_sched_rt_class();
5970
	init_sched_dl_class();
P
Peter Zijlstra 已提交
5971 5972 5973

	sched_init_smt();

5974
	sched_smp_initialized = true;
L
Linus Torvalds 已提交
5975
}
5976 5977 5978

static int __init migration_init(void)
{
5979
	sched_rq_cpu_starting(smp_processor_id());
5980
	return 0;
L
Linus Torvalds 已提交
5981
}
5982 5983
early_initcall(migration_init);

L
Linus Torvalds 已提交
5984 5985 5986
#else
void __init sched_init_smp(void)
{
I
Ingo Molnar 已提交
5987
	sched_init_granularity();
L
Linus Torvalds 已提交
5988 5989 5990 5991 5992 5993 5994 5995 5996 5997
}
#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);
}

5998
#ifdef CONFIG_CGROUP_SCHED
5999 6000 6001 6002
/*
 * Default task group.
 * Every task in system belongs to this group at bootup.
 */
6003
struct task_group root_task_group;
6004
LIST_HEAD(task_groups);
6005 6006 6007

/* Cacheline aligned slab cache for task_group */
static struct kmem_cache *task_group_cache __read_mostly;
6008
#endif
P
Peter Zijlstra 已提交
6009

6010
DECLARE_PER_CPU(cpumask_var_t, load_balance_mask);
6011
DECLARE_PER_CPU(cpumask_var_t, select_idle_mask);
P
Peter Zijlstra 已提交
6012

L
Linus Torvalds 已提交
6013 6014
void __init sched_init(void)
{
I
Ingo Molnar 已提交
6015
	int i, j;
6016 6017
	unsigned long alloc_size = 0, ptr;

6018
	sched_clock_init();
6019
	wait_bit_init();
6020

6021 6022 6023 6024 6025 6026 6027
#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) {
6028
		ptr = (unsigned long)kzalloc(alloc_size, GFP_NOWAIT);
6029 6030

#ifdef CONFIG_FAIR_GROUP_SCHED
6031
		root_task_group.se = (struct sched_entity **)ptr;
6032 6033
		ptr += nr_cpu_ids * sizeof(void **);

6034
		root_task_group.cfs_rq = (struct cfs_rq **)ptr;
6035
		ptr += nr_cpu_ids * sizeof(void **);
6036

6037
#endif /* CONFIG_FAIR_GROUP_SCHED */
6038
#ifdef CONFIG_RT_GROUP_SCHED
6039
		root_task_group.rt_se = (struct sched_rt_entity **)ptr;
6040 6041
		ptr += nr_cpu_ids * sizeof(void **);

6042
		root_task_group.rt_rq = (struct rt_rq **)ptr;
6043 6044
		ptr += nr_cpu_ids * sizeof(void **);

6045
#endif /* CONFIG_RT_GROUP_SCHED */
6046
	}
6047
#ifdef CONFIG_CPUMASK_OFFSTACK
6048 6049 6050
	for_each_possible_cpu(i) {
		per_cpu(load_balance_mask, i) = (cpumask_var_t)kzalloc_node(
			cpumask_size(), GFP_KERNEL, cpu_to_node(i));
6051 6052
		per_cpu(select_idle_mask, i) = (cpumask_var_t)kzalloc_node(
			cpumask_size(), GFP_KERNEL, cpu_to_node(i));
6053
	}
6054
#endif /* CONFIG_CPUMASK_OFFSTACK */
I
Ingo Molnar 已提交
6055

I
Ingo Molnar 已提交
6056 6057
	init_rt_bandwidth(&def_rt_bandwidth, global_rt_period(), global_rt_runtime());
	init_dl_bandwidth(&def_dl_bandwidth, global_rt_period(), global_rt_runtime());
6058

G
Gregory Haskins 已提交
6059 6060 6061 6062
#ifdef CONFIG_SMP
	init_defrootdomain();
#endif

6063
#ifdef CONFIG_RT_GROUP_SCHED
6064
	init_rt_bandwidth(&root_task_group.rt_bandwidth,
6065
			global_rt_period(), global_rt_runtime());
6066
#endif /* CONFIG_RT_GROUP_SCHED */
6067

D
Dhaval Giani 已提交
6068
#ifdef CONFIG_CGROUP_SCHED
6069 6070
	task_group_cache = KMEM_CACHE(task_group, 0);

6071 6072
	list_add(&root_task_group.list, &task_groups);
	INIT_LIST_HEAD(&root_task_group.children);
6073
	INIT_LIST_HEAD(&root_task_group.siblings);
6074
	autogroup_init(&init_task);
D
Dhaval Giani 已提交
6075
#endif /* CONFIG_CGROUP_SCHED */
P
Peter Zijlstra 已提交
6076

6077
	for_each_possible_cpu(i) {
6078
		struct rq *rq;
L
Linus Torvalds 已提交
6079 6080

		rq = cpu_rq(i);
6081
		raw_spin_lock_init(&rq->lock);
N
Nick Piggin 已提交
6082
		rq->nr_running = 0;
6083 6084
		rq->calc_load_active = 0;
		rq->calc_load_update = jiffies + LOAD_FREQ;
6085
		init_cfs_rq(&rq->cfs);
6086 6087
		init_rt_rq(&rq->rt);
		init_dl_rq(&rq->dl);
I
Ingo Molnar 已提交
6088
#ifdef CONFIG_FAIR_GROUP_SCHED
6089
		root_task_group.shares = ROOT_TASK_GROUP_LOAD;
P
Peter Zijlstra 已提交
6090
		INIT_LIST_HEAD(&rq->leaf_cfs_rq_list);
6091
		rq->tmp_alone_branch = &rq->leaf_cfs_rq_list;
D
Dhaval Giani 已提交
6092
		/*
I
Ingo Molnar 已提交
6093
		 * How much CPU bandwidth does root_task_group get?
D
Dhaval Giani 已提交
6094 6095
		 *
		 * In case of task-groups formed thr' the cgroup filesystem, it
I
Ingo Molnar 已提交
6096 6097
		 * gets 100% of the CPU resources in the system. This overall
		 * system CPU resource is divided among the tasks of
6098
		 * root_task_group and its child task-groups in a fair manner,
D
Dhaval Giani 已提交
6099 6100 6101
		 * based on each entity's (task or task-group's) weight
		 * (se->load.weight).
		 *
6102
		 * In other words, if root_task_group has 10 tasks of weight
D
Dhaval Giani 已提交
6103
		 * 1024) and two child groups A0 and A1 (of weight 1024 each),
I
Ingo Molnar 已提交
6104
		 * then A0's share of the CPU resource is:
D
Dhaval Giani 已提交
6105
		 *
6106
		 *	A0's bandwidth = 1024 / (10*1024 + 1024 + 1024) = 8.33%
D
Dhaval Giani 已提交
6107
		 *
6108 6109
		 * 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
6110
		 */
6111
		init_cfs_bandwidth(&root_task_group.cfs_bandwidth);
6112
		init_tg_cfs_entry(&root_task_group, &rq->cfs, NULL, i, NULL);
D
Dhaval Giani 已提交
6113 6114 6115
#endif /* CONFIG_FAIR_GROUP_SCHED */

		rq->rt.rt_runtime = def_rt_bandwidth.rt_runtime;
6116
#ifdef CONFIG_RT_GROUP_SCHED
6117
		init_tg_rt_entry(&root_task_group, &rq->rt, NULL, i, NULL);
I
Ingo Molnar 已提交
6118
#endif
L
Linus Torvalds 已提交
6119

I
Ingo Molnar 已提交
6120 6121
		for (j = 0; j < CPU_LOAD_IDX_MAX; j++)
			rq->cpu_load[j] = 0;
6122

L
Linus Torvalds 已提交
6123
#ifdef CONFIG_SMP
N
Nick Piggin 已提交
6124
		rq->sd = NULL;
G
Gregory Haskins 已提交
6125
		rq->rd = NULL;
6126
		rq->cpu_capacity = rq->cpu_capacity_orig = SCHED_CAPACITY_SCALE;
6127
		rq->balance_callback = NULL;
L
Linus Torvalds 已提交
6128
		rq->active_balance = 0;
I
Ingo Molnar 已提交
6129
		rq->next_balance = jiffies;
L
Linus Torvalds 已提交
6130
		rq->push_cpu = 0;
6131
		rq->cpu = i;
6132
		rq->online = 0;
6133 6134
		rq->idle_stamp = 0;
		rq->avg_idle = 2*sysctl_sched_migration_cost;
6135
		rq->max_idle_balance_cost = sysctl_sched_migration_cost;
6136 6137 6138

		INIT_LIST_HEAD(&rq->cfs_tasks);

6139
		rq_attach_root(rq, &def_root_domain);
6140
#ifdef CONFIG_NO_HZ_COMMON
6141
		rq->last_load_update_tick = jiffies;
6142
		rq->nohz_flags = 0;
6143
#endif
6144 6145 6146
#ifdef CONFIG_NO_HZ_FULL
		rq->last_sched_tick = 0;
#endif
6147
#endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
6148
		init_rq_hrtick(rq);
L
Linus Torvalds 已提交
6149 6150 6151
		atomic_set(&rq->nr_iowait, 0);
	}

6152
	set_load_weight(&init_task);
6153

L
Linus Torvalds 已提交
6154 6155 6156
	/*
	 * The boot idle thread does lazy MMU switching as well:
	 */
V
Vegard Nossum 已提交
6157
	mmgrab(&init_mm);
L
Linus Torvalds 已提交
6158 6159 6160 6161 6162 6163 6164 6165 6166
	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());
6167 6168 6169

	calc_load_update = jiffies + LOAD_FREQ;

6170
#ifdef CONFIG_SMP
R
Rusty Russell 已提交
6171 6172 6173
	/* May be allocated at isolcpus cmdline parse time */
	if (cpu_isolated_map == NULL)
		zalloc_cpumask_var(&cpu_isolated_map, GFP_NOWAIT);
6174
	idle_thread_set_boot_cpu();
6175
	set_cpu_rq_start_time(smp_processor_id());
6176 6177
#endif
	init_sched_fair_class();
6178

6179 6180
	init_schedstats();

6181
	scheduler_running = 1;
L
Linus Torvalds 已提交
6182 6183
}

6184
#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
6185 6186
static inline int preempt_count_equals(int preempt_offset)
{
6187
	int nested = preempt_count() + rcu_preempt_depth();
6188

A
Arnd Bergmann 已提交
6189
	return (nested == preempt_offset);
6190 6191
}

6192
void __might_sleep(const char *file, int line, int preempt_offset)
L
Linus Torvalds 已提交
6193
{
P
Peter Zijlstra 已提交
6194 6195 6196 6197 6198
	/*
	 * 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.
	 */
6199
	WARN_ONCE(current->state != TASK_RUNNING && current->task_state_change,
P
Peter Zijlstra 已提交
6200 6201 6202 6203
			"do not call blocking ops when !TASK_RUNNING; "
			"state=%lx set at [<%p>] %pS\n",
			current->state,
			(void *)current->task_state_change,
6204
			(void *)current->task_state_change);
P
Peter Zijlstra 已提交
6205

6206 6207 6208 6209 6210
	___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
6211
{
I
Ingo Molnar 已提交
6212 6213 6214
	/* Ratelimiting timestamp: */
	static unsigned long prev_jiffy;

6215
	unsigned long preempt_disable_ip;
L
Linus Torvalds 已提交
6216

I
Ingo Molnar 已提交
6217 6218 6219
	/* WARN_ON_ONCE() by default, no rate limit required: */
	rcu_sleep_check();

6220 6221
	if ((preempt_count_equals(preempt_offset) && !irqs_disabled() &&
	     !is_idle_task(current)) ||
6222 6223
	    system_state == SYSTEM_BOOTING || system_state > SYSTEM_RUNNING ||
	    oops_in_progress)
I
Ingo Molnar 已提交
6224
		return;
6225

I
Ingo Molnar 已提交
6226 6227 6228 6229
	if (time_before(jiffies, prev_jiffy + HZ) && prev_jiffy)
		return;
	prev_jiffy = jiffies;

I
Ingo Molnar 已提交
6230
	/* Save this before calling printk(), since that will clobber it: */
6231 6232
	preempt_disable_ip = get_preempt_disable_ip(current);

P
Peter Zijlstra 已提交
6233 6234 6235 6236 6237 6238 6239
	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
6240

6241 6242 6243
	if (task_stack_end_corrupted(current))
		printk(KERN_EMERG "Thread overran stack, or stack corrupted\n");

I
Ingo Molnar 已提交
6244 6245 6246
	debug_show_held_locks(current);
	if (irqs_disabled())
		print_irqtrace_events(current);
6247 6248
	if (IS_ENABLED(CONFIG_DEBUG_PREEMPT)
	    && !preempt_count_equals(preempt_offset)) {
6249
		pr_err("Preemption disabled at:");
6250
		print_ip_sym(preempt_disable_ip);
6251 6252
		pr_cont("\n");
	}
I
Ingo Molnar 已提交
6253
	dump_stack();
6254
	add_taint(TAINT_WARN, LOCKDEP_STILL_OK);
L
Linus Torvalds 已提交
6255
}
6256
EXPORT_SYMBOL(___might_sleep);
L
Linus Torvalds 已提交
6257 6258 6259
#endif

#ifdef CONFIG_MAGIC_SYSRQ
6260
void normalize_rt_tasks(void)
6261
{
6262
	struct task_struct *g, *p;
6263 6264 6265
	struct sched_attr attr = {
		.sched_policy = SCHED_NORMAL,
	};
L
Linus Torvalds 已提交
6266

6267
	read_lock(&tasklist_lock);
6268
	for_each_process_thread(g, p) {
6269 6270 6271
		/*
		 * Only normalize user tasks:
		 */
6272
		if (p->flags & PF_KTHREAD)
6273 6274
			continue;

6275 6276 6277 6278
		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
6279

6280
		if (!dl_task(p) && !rt_task(p)) {
I
Ingo Molnar 已提交
6281 6282 6283 6284
			/*
			 * Renice negative nice level userspace
			 * tasks back to 0:
			 */
6285
			if (task_nice(p) < 0)
I
Ingo Molnar 已提交
6286
				set_user_nice(p, 0);
L
Linus Torvalds 已提交
6287
			continue;
I
Ingo Molnar 已提交
6288
		}
L
Linus Torvalds 已提交
6289

6290
		__sched_setscheduler(p, &attr, false, false);
6291
	}
6292
	read_unlock(&tasklist_lock);
L
Linus Torvalds 已提交
6293 6294 6295
}

#endif /* CONFIG_MAGIC_SYSRQ */
6296

6297
#if defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB)
6298
/*
6299
 * These functions are only useful for the IA64 MCA handling, or kdb.
6300 6301 6302 6303 6304 6305 6306 6307 6308
 *
 * 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
6309
 * curr_task - return the current task for a given CPU.
6310 6311 6312
 * @cpu: the processor in question.
 *
 * ONLY VALID WHEN THE WHOLE SYSTEM IS STOPPED!
6313 6314
 *
 * Return: The current task for @cpu.
6315
 */
6316
struct task_struct *curr_task(int cpu)
6317 6318 6319 6320
{
	return cpu_curr(cpu);
}

6321 6322 6323
#endif /* defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB) */

#ifdef CONFIG_IA64
6324
/**
I
Ingo Molnar 已提交
6325
 * set_curr_task - set the current task for a given CPU.
6326 6327 6328 6329
 * @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
6330
 * are serviced on a separate stack. It allows the architecture to switch the
I
Ingo Molnar 已提交
6331
 * notion of the current task on a CPU in a non-blocking manner. This function
6332 6333 6334 6335 6336 6337 6338
 * 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!
 */
6339
void ia64_set_curr_task(int cpu, struct task_struct *p)
6340 6341 6342 6343 6344
{
	cpu_curr(cpu) = p;
}

#endif
S
Srivatsa Vaddagiri 已提交
6345

D
Dhaval Giani 已提交
6346
#ifdef CONFIG_CGROUP_SCHED
6347 6348 6349
/* task_group_lock serializes the addition/removal of task groups */
static DEFINE_SPINLOCK(task_group_lock);

6350
static void sched_free_group(struct task_group *tg)
6351 6352 6353
{
	free_fair_sched_group(tg);
	free_rt_sched_group(tg);
6354
	autogroup_free(tg);
6355
	kmem_cache_free(task_group_cache, tg);
6356 6357 6358
}

/* allocate runqueue etc for a new task group */
6359
struct task_group *sched_create_group(struct task_group *parent)
6360 6361 6362
{
	struct task_group *tg;

6363
	tg = kmem_cache_alloc(task_group_cache, GFP_KERNEL | __GFP_ZERO);
6364 6365 6366
	if (!tg)
		return ERR_PTR(-ENOMEM);

6367
	if (!alloc_fair_sched_group(tg, parent))
6368 6369
		goto err;

6370
	if (!alloc_rt_sched_group(tg, parent))
6371 6372
		goto err;

6373 6374 6375
	return tg;

err:
6376
	sched_free_group(tg);
6377 6378 6379 6380 6381 6382 6383
	return ERR_PTR(-ENOMEM);
}

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

6384
	spin_lock_irqsave(&task_group_lock, flags);
P
Peter Zijlstra 已提交
6385
	list_add_rcu(&tg->list, &task_groups);
P
Peter Zijlstra 已提交
6386

I
Ingo Molnar 已提交
6387 6388
	/* Root should already exist: */
	WARN_ON(!parent);
P
Peter Zijlstra 已提交
6389 6390 6391

	tg->parent = parent;
	INIT_LIST_HEAD(&tg->children);
6392
	list_add_rcu(&tg->siblings, &parent->children);
6393
	spin_unlock_irqrestore(&task_group_lock, flags);
6394 6395

	online_fair_sched_group(tg);
S
Srivatsa Vaddagiri 已提交
6396 6397
}

6398
/* rcu callback to free various structures associated with a task group */
6399
static void sched_free_group_rcu(struct rcu_head *rhp)
S
Srivatsa Vaddagiri 已提交
6400
{
I
Ingo Molnar 已提交
6401
	/* Now it should be safe to free those cfs_rqs: */
6402
	sched_free_group(container_of(rhp, struct task_group, rcu));
S
Srivatsa Vaddagiri 已提交
6403 6404
}

6405
void sched_destroy_group(struct task_group *tg)
6406
{
I
Ingo Molnar 已提交
6407
	/* Wait for possible concurrent references to cfs_rqs complete: */
6408
	call_rcu(&tg->rcu, sched_free_group_rcu);
6409 6410 6411
}

void sched_offline_group(struct task_group *tg)
S
Srivatsa Vaddagiri 已提交
6412
{
6413
	unsigned long flags;
S
Srivatsa Vaddagiri 已提交
6414

I
Ingo Molnar 已提交
6415
	/* End participation in shares distribution: */
6416
	unregister_fair_sched_group(tg);
6417 6418

	spin_lock_irqsave(&task_group_lock, flags);
P
Peter Zijlstra 已提交
6419
	list_del_rcu(&tg->list);
P
Peter Zijlstra 已提交
6420
	list_del_rcu(&tg->siblings);
6421
	spin_unlock_irqrestore(&task_group_lock, flags);
S
Srivatsa Vaddagiri 已提交
6422 6423
}

6424
static void sched_change_group(struct task_struct *tsk, int type)
S
Srivatsa Vaddagiri 已提交
6425
{
P
Peter Zijlstra 已提交
6426
	struct task_group *tg;
S
Srivatsa Vaddagiri 已提交
6427

6428 6429 6430 6431 6432 6433
	/*
	 * 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
6434 6435 6436 6437
			  struct task_group, css);
	tg = autogroup_task_group(tsk, tg);
	tsk->sched_task_group = tg;

P
Peter Zijlstra 已提交
6438
#ifdef CONFIG_FAIR_GROUP_SCHED
6439 6440
	if (tsk->sched_class->task_change_group)
		tsk->sched_class->task_change_group(tsk, type);
6441
	else
P
Peter Zijlstra 已提交
6442
#endif
6443
		set_task_rq(tsk, task_cpu(tsk));
6444 6445 6446 6447 6448 6449 6450 6451 6452 6453 6454
}

/*
 * 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)
{
6455 6456
	int queued, running, queue_flags =
		DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
6457 6458 6459 6460
	struct rq_flags rf;
	struct rq *rq;

	rq = task_rq_lock(tsk, &rf);
6461
	update_rq_clock(rq);
6462 6463 6464 6465 6466

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

	if (queued)
6467
		dequeue_task(rq, tsk, queue_flags);
6468
	if (running)
6469 6470 6471
		put_prev_task(rq, tsk);

	sched_change_group(tsk, TASK_MOVE_GROUP);
P
Peter Zijlstra 已提交
6472

6473
	if (queued)
6474
		enqueue_task(rq, tsk, queue_flags);
6475
	if (running)
6476
		set_curr_task(rq, tsk);
S
Srivatsa Vaddagiri 已提交
6477

6478
	task_rq_unlock(rq, tsk, &rf);
S
Srivatsa Vaddagiri 已提交
6479
}
D
Dhaval Giani 已提交
6480
#endif /* CONFIG_CGROUP_SCHED */
S
Srivatsa Vaddagiri 已提交
6481

6482 6483 6484 6485 6486
#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
6487

P
Peter Zijlstra 已提交
6488 6489
/* Must be called with tasklist_lock held */
static inline int tg_has_rt_tasks(struct task_group *tg)
6490
{
P
Peter Zijlstra 已提交
6491
	struct task_struct *g, *p;
6492

6493 6494 6495 6496 6497 6498
	/*
	 * Autogroups do not have RT tasks; see autogroup_create().
	 */
	if (task_group_is_autogroup(tg))
		return 0;

6499
	for_each_process_thread(g, p) {
6500
		if (rt_task(p) && task_group(p) == tg)
P
Peter Zijlstra 已提交
6501
			return 1;
6502
	}
6503

P
Peter Zijlstra 已提交
6504 6505
	return 0;
}
6506

P
Peter Zijlstra 已提交
6507 6508 6509 6510 6511
struct rt_schedulable_data {
	struct task_group *tg;
	u64 rt_period;
	u64 rt_runtime;
};
6512

6513
static int tg_rt_schedulable(struct task_group *tg, void *data)
P
Peter Zijlstra 已提交
6514 6515 6516 6517 6518
{
	struct rt_schedulable_data *d = data;
	struct task_group *child;
	unsigned long total, sum = 0;
	u64 period, runtime;
6519

P
Peter Zijlstra 已提交
6520 6521
	period = ktime_to_ns(tg->rt_bandwidth.rt_period);
	runtime = tg->rt_bandwidth.rt_runtime;
6522

P
Peter Zijlstra 已提交
6523 6524 6525
	if (tg == d->tg) {
		period = d->rt_period;
		runtime = d->rt_runtime;
6526 6527
	}

6528 6529 6530 6531 6532
	/*
	 * Cannot have more runtime than the period.
	 */
	if (runtime > period && runtime != RUNTIME_INF)
		return -EINVAL;
P
Peter Zijlstra 已提交
6533

6534 6535 6536
	/*
	 * Ensure we don't starve existing RT tasks.
	 */
P
Peter Zijlstra 已提交
6537 6538
	if (rt_bandwidth_enabled() && !runtime && tg_has_rt_tasks(tg))
		return -EBUSY;
P
Peter Zijlstra 已提交
6539

P
Peter Zijlstra 已提交
6540
	total = to_ratio(period, runtime);
P
Peter Zijlstra 已提交
6541

6542 6543 6544 6545 6546
	/*
	 * 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
6547

6548 6549 6550
	/*
	 * The sum of our children's runtime should not exceed our own.
	 */
P
Peter Zijlstra 已提交
6551 6552 6553
	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
6554

P
Peter Zijlstra 已提交
6555 6556 6557 6558
		if (child == d->tg) {
			period = d->rt_period;
			runtime = d->rt_runtime;
		}
P
Peter Zijlstra 已提交
6559

P
Peter Zijlstra 已提交
6560
		sum += to_ratio(period, runtime);
P
Peter Zijlstra 已提交
6561
	}
P
Peter Zijlstra 已提交
6562

P
Peter Zijlstra 已提交
6563 6564 6565 6566
	if (sum > total)
		return -EINVAL;

	return 0;
P
Peter Zijlstra 已提交
6567 6568
}

P
Peter Zijlstra 已提交
6569
static int __rt_schedulable(struct task_group *tg, u64 period, u64 runtime)
6570
{
6571 6572
	int ret;

P
Peter Zijlstra 已提交
6573 6574 6575 6576 6577 6578
	struct rt_schedulable_data data = {
		.tg = tg,
		.rt_period = period,
		.rt_runtime = runtime,
	};

6579 6580 6581 6582 6583
	rcu_read_lock();
	ret = walk_tg_tree(tg_rt_schedulable, tg_nop, &data);
	rcu_read_unlock();

	return ret;
6584 6585
}

6586
static int tg_set_rt_bandwidth(struct task_group *tg,
6587
		u64 rt_period, u64 rt_runtime)
P
Peter Zijlstra 已提交
6588
{
P
Peter Zijlstra 已提交
6589
	int i, err = 0;
P
Peter Zijlstra 已提交
6590

6591 6592 6593 6594 6595 6596 6597 6598 6599 6600 6601
	/*
	 * 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
6602
	mutex_lock(&rt_constraints_mutex);
6603
	read_lock(&tasklist_lock);
P
Peter Zijlstra 已提交
6604 6605
	err = __rt_schedulable(tg, rt_period, rt_runtime);
	if (err)
P
Peter Zijlstra 已提交
6606
		goto unlock;
P
Peter Zijlstra 已提交
6607

6608
	raw_spin_lock_irq(&tg->rt_bandwidth.rt_runtime_lock);
6609 6610
	tg->rt_bandwidth.rt_period = ns_to_ktime(rt_period);
	tg->rt_bandwidth.rt_runtime = rt_runtime;
P
Peter Zijlstra 已提交
6611 6612 6613 6614

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

6615
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
6616
		rt_rq->rt_runtime = rt_runtime;
6617
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
6618
	}
6619
	raw_spin_unlock_irq(&tg->rt_bandwidth.rt_runtime_lock);
P
Peter Zijlstra 已提交
6620
unlock:
6621
	read_unlock(&tasklist_lock);
P
Peter Zijlstra 已提交
6622 6623 6624
	mutex_unlock(&rt_constraints_mutex);

	return err;
P
Peter Zijlstra 已提交
6625 6626
}

6627
static int sched_group_set_rt_runtime(struct task_group *tg, long rt_runtime_us)
6628 6629 6630 6631 6632 6633 6634 6635
{
	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;

6636
	return tg_set_rt_bandwidth(tg, rt_period, rt_runtime);
6637 6638
}

6639
static long sched_group_rt_runtime(struct task_group *tg)
P
Peter Zijlstra 已提交
6640 6641 6642
{
	u64 rt_runtime_us;

6643
	if (tg->rt_bandwidth.rt_runtime == RUNTIME_INF)
P
Peter Zijlstra 已提交
6644 6645
		return -1;

6646
	rt_runtime_us = tg->rt_bandwidth.rt_runtime;
P
Peter Zijlstra 已提交
6647 6648 6649
	do_div(rt_runtime_us, NSEC_PER_USEC);
	return rt_runtime_us;
}
6650

6651
static int sched_group_set_rt_period(struct task_group *tg, u64 rt_period_us)
6652 6653 6654
{
	u64 rt_runtime, rt_period;

6655
	rt_period = rt_period_us * NSEC_PER_USEC;
6656 6657
	rt_runtime = tg->rt_bandwidth.rt_runtime;

6658
	return tg_set_rt_bandwidth(tg, rt_period, rt_runtime);
6659 6660
}

6661
static long sched_group_rt_period(struct task_group *tg)
6662 6663 6664 6665 6666 6667 6668
{
	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;
}
6669
#endif /* CONFIG_RT_GROUP_SCHED */
6670

6671
#ifdef CONFIG_RT_GROUP_SCHED
6672 6673 6674 6675 6676
static int sched_rt_global_constraints(void)
{
	int ret = 0;

	mutex_lock(&rt_constraints_mutex);
P
Peter Zijlstra 已提交
6677
	read_lock(&tasklist_lock);
6678
	ret = __rt_schedulable(NULL, 0, 0);
P
Peter Zijlstra 已提交
6679
	read_unlock(&tasklist_lock);
6680 6681 6682 6683
	mutex_unlock(&rt_constraints_mutex);

	return ret;
}
6684

6685
static int sched_rt_can_attach(struct task_group *tg, struct task_struct *tsk)
6686 6687 6688 6689 6690 6691 6692 6693
{
	/* 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;
}

6694
#else /* !CONFIG_RT_GROUP_SCHED */
6695 6696
static int sched_rt_global_constraints(void)
{
P
Peter Zijlstra 已提交
6697
	unsigned long flags;
6698
	int i;
6699

6700
	raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags);
P
Peter Zijlstra 已提交
6701 6702 6703
	for_each_possible_cpu(i) {
		struct rt_rq *rt_rq = &cpu_rq(i)->rt;

6704
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
6705
		rt_rq->rt_runtime = global_rt_runtime();
6706
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
6707
	}
6708
	raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags);
P
Peter Zijlstra 已提交
6709

6710
	return 0;
6711
}
6712
#endif /* CONFIG_RT_GROUP_SCHED */
6713

6714
static int sched_dl_global_validate(void)
6715
{
6716 6717
	u64 runtime = global_rt_runtime();
	u64 period = global_rt_period();
6718
	u64 new_bw = to_ratio(period, runtime);
6719
	struct dl_bw *dl_b;
6720
	int cpu, ret = 0;
6721
	unsigned long flags;
6722 6723 6724 6725 6726 6727 6728 6729 6730 6731

	/*
	 * 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!
	 */
6732
	for_each_possible_cpu(cpu) {
6733 6734
		rcu_read_lock_sched();
		dl_b = dl_bw_of(cpu);
6735

6736
		raw_spin_lock_irqsave(&dl_b->lock, flags);
6737 6738
		if (new_bw < dl_b->total_bw)
			ret = -EBUSY;
6739
		raw_spin_unlock_irqrestore(&dl_b->lock, flags);
6740

6741 6742
		rcu_read_unlock_sched();

6743 6744
		if (ret)
			break;
6745 6746
	}

6747
	return ret;
6748 6749
}

6750 6751 6752 6753
void init_dl_rq_bw_ratio(struct dl_rq *dl_rq)
{
	if (global_rt_runtime() == RUNTIME_INF) {
		dl_rq->bw_ratio = 1 << RATIO_SHIFT;
6754
		dl_rq->extra_bw = 1 << BW_SHIFT;
6755 6756 6757
	} else {
		dl_rq->bw_ratio = to_ratio(global_rt_runtime(),
			  global_rt_period()) >> (BW_SHIFT - RATIO_SHIFT);
6758 6759
		dl_rq->extra_bw = to_ratio(global_rt_period(),
						    global_rt_runtime());
6760 6761 6762
	}
}

6763
static void sched_dl_do_global(void)
6764
{
6765
	u64 new_bw = -1;
6766
	struct dl_bw *dl_b;
6767
	int cpu;
6768
	unsigned long flags;
6769

6770 6771 6772 6773 6774 6775 6776 6777 6778 6779
	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) {
6780 6781
		rcu_read_lock_sched();
		dl_b = dl_bw_of(cpu);
6782

6783
		raw_spin_lock_irqsave(&dl_b->lock, flags);
6784
		dl_b->bw = new_bw;
6785
		raw_spin_unlock_irqrestore(&dl_b->lock, flags);
6786 6787

		rcu_read_unlock_sched();
6788
		init_dl_rq_bw_ratio(&cpu_rq(cpu)->dl);
6789
	}
6790 6791 6792 6793 6794 6795 6796
}

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

6797 6798
	if ((sysctl_sched_rt_runtime != RUNTIME_INF) &&
		(sysctl_sched_rt_runtime > sysctl_sched_rt_period))
6799 6800 6801 6802 6803 6804 6805 6806 6807
		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());
6808 6809
}

6810
int sched_rt_handler(struct ctl_table *table, int write,
6811
		void __user *buffer, size_t *lenp,
6812 6813 6814 6815
		loff_t *ppos)
{
	int old_period, old_runtime;
	static DEFINE_MUTEX(mutex);
6816
	int ret;
6817 6818 6819 6820 6821

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

6822
	ret = proc_dointvec(table, write, buffer, lenp, ppos);
6823 6824

	if (!ret && write) {
6825 6826 6827 6828
		ret = sched_rt_global_validate();
		if (ret)
			goto undo;

6829
		ret = sched_dl_global_validate();
6830 6831 6832
		if (ret)
			goto undo;

6833
		ret = sched_rt_global_constraints();
6834 6835 6836 6837 6838 6839 6840 6841 6842 6843
		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;
6844 6845 6846 6847 6848
	}
	mutex_unlock(&mutex);

	return ret;
}
6849

6850
int sched_rr_handler(struct ctl_table *table, int write,
6851 6852 6853 6854 6855 6856 6857 6858
		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
6859 6860 6861 6862
	/*
	 * Make sure that internally we keep jiffies.
	 * Also, writing zero resets the timeslice to default:
	 */
6863
	if (!ret && write) {
6864 6865 6866
		sched_rr_timeslice =
			sysctl_sched_rr_timeslice <= 0 ? RR_TIMESLICE :
			msecs_to_jiffies(sysctl_sched_rr_timeslice);
6867 6868 6869 6870 6871
	}
	mutex_unlock(&mutex);
	return ret;
}

6872
#ifdef CONFIG_CGROUP_SCHED
6873

6874
static inline struct task_group *css_tg(struct cgroup_subsys_state *css)
6875
{
6876
	return css ? container_of(css, struct task_group, css) : NULL;
6877 6878
}

6879 6880
static struct cgroup_subsys_state *
cpu_cgroup_css_alloc(struct cgroup_subsys_state *parent_css)
6881
{
6882 6883
	struct task_group *parent = css_tg(parent_css);
	struct task_group *tg;
6884

6885
	if (!parent) {
6886
		/* This is early initialization for the top cgroup */
6887
		return &root_task_group.css;
6888 6889
	}

6890
	tg = sched_create_group(parent);
6891 6892 6893 6894 6895 6896
	if (IS_ERR(tg))
		return ERR_PTR(-ENOMEM);

	return &tg->css;
}

6897 6898 6899 6900 6901 6902 6903 6904 6905 6906 6907
/* 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;
}

6908
static void cpu_cgroup_css_released(struct cgroup_subsys_state *css)
6909
{
6910
	struct task_group *tg = css_tg(css);
6911

6912
	sched_offline_group(tg);
6913 6914
}

6915
static void cpu_cgroup_css_free(struct cgroup_subsys_state *css)
6916
{
6917
	struct task_group *tg = css_tg(css);
6918

6919 6920 6921 6922
	/*
	 * Relies on the RCU grace period between css_released() and this.
	 */
	sched_free_group(tg);
6923 6924
}

6925 6926 6927 6928
/*
 * 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.
 */
6929
static void cpu_cgroup_fork(struct task_struct *task)
6930
{
6931 6932 6933 6934 6935
	struct rq_flags rf;
	struct rq *rq;

	rq = task_rq_lock(task, &rf);

6936
	update_rq_clock(rq);
6937 6938 6939
	sched_change_group(task, TASK_SET_GROUP);

	task_rq_unlock(rq, task, &rf);
6940 6941
}

6942
static int cpu_cgroup_can_attach(struct cgroup_taskset *tset)
6943
{
6944
	struct task_struct *task;
6945
	struct cgroup_subsys_state *css;
6946
	int ret = 0;
6947

6948
	cgroup_taskset_for_each(task, css, tset) {
6949
#ifdef CONFIG_RT_GROUP_SCHED
6950
		if (!sched_rt_can_attach(css_tg(css), task))
6951
			return -EINVAL;
6952
#else
6953 6954 6955
		/* We don't support RT-tasks being in separate groups */
		if (task->sched_class != &fair_sched_class)
			return -EINVAL;
6956
#endif
6957 6958 6959 6960 6961 6962 6963 6964 6965 6966 6967 6968 6969 6970 6971 6972
		/*
		 * 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;
6973
	}
6974
	return ret;
6975
}
6976

6977
static void cpu_cgroup_attach(struct cgroup_taskset *tset)
6978
{
6979
	struct task_struct *task;
6980
	struct cgroup_subsys_state *css;
6981

6982
	cgroup_taskset_for_each(task, css, tset)
6983
		sched_move_task(task);
6984 6985
}

6986
#ifdef CONFIG_FAIR_GROUP_SCHED
6987 6988
static int cpu_shares_write_u64(struct cgroup_subsys_state *css,
				struct cftype *cftype, u64 shareval)
6989
{
6990
	return sched_group_set_shares(css_tg(css), scale_load(shareval));
6991 6992
}

6993 6994
static u64 cpu_shares_read_u64(struct cgroup_subsys_state *css,
			       struct cftype *cft)
6995
{
6996
	struct task_group *tg = css_tg(css);
6997

6998
	return (u64) scale_load_down(tg->shares);
6999
}
7000 7001

#ifdef CONFIG_CFS_BANDWIDTH
7002 7003
static DEFINE_MUTEX(cfs_constraints_mutex);

7004 7005 7006
const u64 max_cfs_quota_period = 1 * NSEC_PER_SEC; /* 1s */
const u64 min_cfs_quota_period = 1 * NSEC_PER_MSEC; /* 1ms */

7007 7008
static int __cfs_schedulable(struct task_group *tg, u64 period, u64 runtime);

7009 7010
static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota)
{
7011
	int i, ret = 0, runtime_enabled, runtime_was_enabled;
7012
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
7013 7014 7015 7016 7017 7018 7019 7020 7021 7022 7023 7024 7025 7026 7027 7028 7029 7030 7031 7032

	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;

7033 7034 7035 7036 7037
	/*
	 * Prevent race between setting of cfs_rq->runtime_enabled and
	 * unthrottle_offline_cfs_rqs().
	 */
	get_online_cpus();
7038 7039 7040 7041 7042
	mutex_lock(&cfs_constraints_mutex);
	ret = __cfs_schedulable(tg, period, quota);
	if (ret)
		goto out_unlock;

7043
	runtime_enabled = quota != RUNTIME_INF;
7044
	runtime_was_enabled = cfs_b->quota != RUNTIME_INF;
7045 7046 7047 7048 7049 7050
	/*
	 * 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();
7051 7052 7053
	raw_spin_lock_irq(&cfs_b->lock);
	cfs_b->period = ns_to_ktime(period);
	cfs_b->quota = quota;
7054

P
Paul Turner 已提交
7055
	__refill_cfs_bandwidth_runtime(cfs_b);
I
Ingo Molnar 已提交
7056 7057

	/* Restart the period timer (if active) to handle new period expiry: */
P
Peter Zijlstra 已提交
7058 7059
	if (runtime_enabled)
		start_cfs_bandwidth(cfs_b);
I
Ingo Molnar 已提交
7060

7061 7062
	raw_spin_unlock_irq(&cfs_b->lock);

7063
	for_each_online_cpu(i) {
7064
		struct cfs_rq *cfs_rq = tg->cfs_rq[i];
7065
		struct rq *rq = cfs_rq->rq;
7066
		struct rq_flags rf;
7067

7068
		rq_lock_irq(rq, &rf);
7069
		cfs_rq->runtime_enabled = runtime_enabled;
7070
		cfs_rq->runtime_remaining = 0;
7071

7072
		if (cfs_rq->throttled)
7073
			unthrottle_cfs_rq(cfs_rq);
7074
		rq_unlock_irq(rq, &rf);
7075
	}
7076 7077
	if (runtime_was_enabled && !runtime_enabled)
		cfs_bandwidth_usage_dec();
7078 7079
out_unlock:
	mutex_unlock(&cfs_constraints_mutex);
7080
	put_online_cpus();
7081

7082
	return ret;
7083 7084 7085 7086 7087 7088
}

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

7089
	period = ktime_to_ns(tg->cfs_bandwidth.period);
7090 7091 7092 7093 7094 7095 7096 7097 7098 7099 7100 7101
	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;

7102
	if (tg->cfs_bandwidth.quota == RUNTIME_INF)
7103 7104
		return -1;

7105
	quota_us = tg->cfs_bandwidth.quota;
7106 7107 7108 7109 7110 7111 7112 7113 7114 7115
	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;
7116
	quota = tg->cfs_bandwidth.quota;
7117 7118 7119 7120 7121 7122 7123 7124

	return tg_set_cfs_bandwidth(tg, period, quota);
}

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

7125
	cfs_period_us = ktime_to_ns(tg->cfs_bandwidth.period);
7126 7127 7128 7129 7130
	do_div(cfs_period_us, NSEC_PER_USEC);

	return cfs_period_us;
}

7131 7132
static s64 cpu_cfs_quota_read_s64(struct cgroup_subsys_state *css,
				  struct cftype *cft)
7133
{
7134
	return tg_get_cfs_quota(css_tg(css));
7135 7136
}

7137 7138
static int cpu_cfs_quota_write_s64(struct cgroup_subsys_state *css,
				   struct cftype *cftype, s64 cfs_quota_us)
7139
{
7140
	return tg_set_cfs_quota(css_tg(css), cfs_quota_us);
7141 7142
}

7143 7144
static u64 cpu_cfs_period_read_u64(struct cgroup_subsys_state *css,
				   struct cftype *cft)
7145
{
7146
	return tg_get_cfs_period(css_tg(css));
7147 7148
}

7149 7150
static int cpu_cfs_period_write_u64(struct cgroup_subsys_state *css,
				    struct cftype *cftype, u64 cfs_period_us)
7151
{
7152
	return tg_set_cfs_period(css_tg(css), cfs_period_us);
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 7185 7186
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;
7187
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
7188 7189 7190 7191 7192
	s64 quota = 0, parent_quota = -1;

	if (!tg->parent) {
		quota = RUNTIME_INF;
	} else {
7193
		struct cfs_bandwidth *parent_b = &tg->parent->cfs_bandwidth;
7194 7195

		quota = normalize_cfs_quota(tg, d);
7196
		parent_quota = parent_b->hierarchical_quota;
7197 7198

		/*
I
Ingo Molnar 已提交
7199 7200
		 * Ensure max(child_quota) <= parent_quota, inherit when no
		 * limit is set:
7201 7202 7203 7204 7205 7206
		 */
		if (quota == RUNTIME_INF)
			quota = parent_quota;
		else if (parent_quota != RUNTIME_INF && quota > parent_quota)
			return -EINVAL;
	}
7207
	cfs_b->hierarchical_quota = quota;
7208 7209 7210 7211 7212 7213

	return 0;
}

static int __cfs_schedulable(struct task_group *tg, u64 period, u64 quota)
{
7214
	int ret;
7215 7216 7217 7218 7219 7220 7221 7222 7223 7224 7225
	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);
	}

7226 7227 7228 7229 7230
	rcu_read_lock();
	ret = walk_tg_tree(tg_cfs_schedulable_down, tg_nop, &data);
	rcu_read_unlock();

	return ret;
7231
}
7232

7233
static int cpu_stats_show(struct seq_file *sf, void *v)
7234
{
7235
	struct task_group *tg = css_tg(seq_css(sf));
7236
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
7237

7238 7239 7240
	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);
7241 7242 7243

	return 0;
}
7244
#endif /* CONFIG_CFS_BANDWIDTH */
7245
#endif /* CONFIG_FAIR_GROUP_SCHED */
7246

7247
#ifdef CONFIG_RT_GROUP_SCHED
7248 7249
static int cpu_rt_runtime_write(struct cgroup_subsys_state *css,
				struct cftype *cft, s64 val)
P
Peter Zijlstra 已提交
7250
{
7251
	return sched_group_set_rt_runtime(css_tg(css), val);
P
Peter Zijlstra 已提交
7252 7253
}

7254 7255
static s64 cpu_rt_runtime_read(struct cgroup_subsys_state *css,
			       struct cftype *cft)
P
Peter Zijlstra 已提交
7256
{
7257
	return sched_group_rt_runtime(css_tg(css));
P
Peter Zijlstra 已提交
7258
}
7259

7260 7261
static int cpu_rt_period_write_uint(struct cgroup_subsys_state *css,
				    struct cftype *cftype, u64 rt_period_us)
7262
{
7263
	return sched_group_set_rt_period(css_tg(css), rt_period_us);
7264 7265
}

7266 7267
static u64 cpu_rt_period_read_uint(struct cgroup_subsys_state *css,
				   struct cftype *cft)
7268
{
7269
	return sched_group_rt_period(css_tg(css));
7270
}
7271
#endif /* CONFIG_RT_GROUP_SCHED */
P
Peter Zijlstra 已提交
7272

7273
static struct cftype cpu_files[] = {
7274
#ifdef CONFIG_FAIR_GROUP_SCHED
7275 7276
	{
		.name = "shares",
7277 7278
		.read_u64 = cpu_shares_read_u64,
		.write_u64 = cpu_shares_write_u64,
7279
	},
7280
#endif
7281 7282 7283 7284 7285 7286 7287 7288 7289 7290 7291
#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,
	},
7292 7293
	{
		.name = "stat",
7294
		.seq_show = cpu_stats_show,
7295
	},
7296
#endif
7297
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
7298
	{
P
Peter Zijlstra 已提交
7299
		.name = "rt_runtime_us",
7300 7301
		.read_s64 = cpu_rt_runtime_read,
		.write_s64 = cpu_rt_runtime_write,
P
Peter Zijlstra 已提交
7302
	},
7303 7304
	{
		.name = "rt_period_us",
7305 7306
		.read_u64 = cpu_rt_period_read_uint,
		.write_u64 = cpu_rt_period_write_uint,
7307
	},
7308
#endif
I
Ingo Molnar 已提交
7309
	{ }	/* Terminate */
7310 7311
};

7312
struct cgroup_subsys cpu_cgrp_subsys = {
7313
	.css_alloc	= cpu_cgroup_css_alloc,
7314
	.css_online	= cpu_cgroup_css_online,
7315
	.css_released	= cpu_cgroup_css_released,
7316
	.css_free	= cpu_cgroup_css_free,
7317
	.fork		= cpu_cgroup_fork,
7318 7319
	.can_attach	= cpu_cgroup_can_attach,
	.attach		= cpu_cgroup_attach,
7320
	.legacy_cftypes	= cpu_files,
7321
	.early_init	= true,
7322 7323
};

7324
#endif	/* CONFIG_CGROUP_SCHED */
7325

7326 7327 7328 7329 7330
void dump_cpu_task(int cpu)
{
	pr_info("Task dump for CPU %d:\n", cpu);
	sched_show_task(cpu_curr(cpu));
}
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 7370 7371

/*
 * 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,
};