core.c 171.7 KB
Newer Older
L
Linus Torvalds 已提交
1
/*
2
 *  kernel/sched/core.c
L
Linus Torvalds 已提交
3
 *
I
Ingo Molnar 已提交
4
 *  Core kernel scheduler code and related syscalls
L
Linus Torvalds 已提交
5 6 7
 *
 *  Copyright (C) 1991-2002  Linus Torvalds
 */
8
#include <linux/sched.h>
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
#include <linux/compat.h>
20 21 22 23 24 25

#include <linux/blkdev.h>
#include <linux/kprobes.h>
#include <linux/mmu_context.h>
#include <linux/module.h>
#include <linux/nmi.h>
26
#include <linux/prefetch.h>
27 28 29
#include <linux/profile.h>
#include <linux/security.h>
#include <linux/syscalls.h>
30
#include <linux/sched/isolation.h>
L
Linus Torvalds 已提交
31

32
#include <asm/switch_to.h>
33
#include <asm/tlb.h>
34 35 36
#ifdef CONFIG_PARAVIRT
#include <asm/paravirt.h>
#endif
L
Linus Torvalds 已提交
37

38
#include "sched.h"
39
#include "../workqueue_internal.h"
40
#include "../smpboot.h"
41

42
#define CREATE_TRACE_POINTS
43
#include <trace/events/sched.h>
44

45
DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
46

47
#if defined(CONFIG_SCHED_DEBUG) && defined(HAVE_JUMP_LABEL)
I
Ingo Molnar 已提交
48 49
/*
 * Debugging: various feature bits
50 51 52 53
 *
 * If SCHED_DEBUG is disabled, each compilation unit has its own copy of
 * sysctl_sched_features, defined in sched.h, to allow constants propagation
 * at compile time and compiler optimization based on features default.
I
Ingo Molnar 已提交
54
 */
P
Peter Zijlstra 已提交
55 56
#define SCHED_FEAT(name, enabled)	\
	(1UL << __SCHED_FEAT_##name) * enabled |
I
Ingo Molnar 已提交
57
const_debug unsigned int sysctl_sched_features =
58
#include "features.h"
P
Peter Zijlstra 已提交
59 60
	0;
#undef SCHED_FEAT
61
#endif
P
Peter Zijlstra 已提交
62

63 64 65 66 67 68
/*
 * 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;

69 70 71 72 73 74 75 76
/*
 * 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
77
/*
I
Ingo Molnar 已提交
78
 * period over which we measure -rt task CPU usage in us.
P
Peter Zijlstra 已提交
79 80
 * default: 1s
 */
P
Peter Zijlstra 已提交
81
unsigned int sysctl_sched_rt_period = 1000000;
P
Peter Zijlstra 已提交
82

83
__read_mostly int scheduler_running;
84

P
Peter Zijlstra 已提交
85 86 87 88 89
/*
 * 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
90

91 92 93
/*
 * __task_rq_lock - lock the rq @p resides on.
 */
94
struct rq *__task_rq_lock(struct task_struct *p, struct rq_flags *rf)
95 96 97 98 99 100 101 102 103 104
	__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))) {
105
			rq_pin_lock(rq, rf);
106 107 108 109 110 111 112 113 114 115 116 117
			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.
 */
118
struct rq *task_rq_lock(struct task_struct *p, struct rq_flags *rf)
119 120 121 122 123 124
	__acquires(p->pi_lock)
	__acquires(rq->lock)
{
	struct rq *rq;

	for (;;) {
125
		raw_spin_lock_irqsave(&p->pi_lock, rf->flags);
126 127 128 129 130 131 132 133 134 135 136 137 138 139 140
		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
141
		 * If we observe the new CPU in task_rq_lock, the acquire will
142 143 144
		 * pair with the WMB to ensure we must then also see migrating.
		 */
		if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) {
145
			rq_pin_lock(rq, rf);
146 147 148
			return rq;
		}
		raw_spin_unlock(&rq->lock);
149
		raw_spin_unlock_irqrestore(&p->pi_lock, rf->flags);
150 151 152 153 154 155

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

156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223
/*
 * 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
224 225
	if (sched_feat(WARN_DOUBLE_CLOCK))
		SCHED_WARN_ON(rq->clock_update_flags & RQCF_UPDATED);
226 227
	rq->clock_update_flags |= RQCF_UPDATED;
#endif
228

229 230 231 232 233 234 235 236
	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
237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254
#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);
255
	struct rq_flags rf;
P
Peter Zijlstra 已提交
256 257 258

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

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

	return HRTIMER_NORESTART;
}

267
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
268

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

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

276 277 278 279
/*
 * called from hardirq (IPI) context
 */
static void __hrtick_start(void *arg)
280
{
281
	struct rq *rq = arg;
282
	struct rq_flags rf;
283

284
	rq_lock(rq, &rf);
P
Peter Zijlstra 已提交
285
	__hrtick_restart(rq);
286
	rq->hrtick_csd_pending = 0;
287
	rq_unlock(rq, &rf);
288 289
}

290 291 292 293 294
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
295
void hrtick_start(struct rq *rq, u64 delay)
296
{
297
	struct hrtimer *timer = &rq->hrtick_timer;
298 299 300 301 302 303 304 305 306
	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);
307

308
	hrtimer_set_expires(timer, time);
309 310

	if (rq == this_rq()) {
P
Peter Zijlstra 已提交
311
		__hrtick_restart(rq);
312
	} else if (!rq->hrtick_csd_pending) {
313
		smp_call_function_single_async(cpu_of(rq), &rq->hrtick_csd);
314 315
		rq->hrtick_csd_pending = 1;
	}
316 317
}

318 319 320 321 322 323
#else
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
324
void hrtick_start(struct rq *rq, u64 delay)
325
{
W
Wanpeng Li 已提交
326 327 328 329 330
	/*
	 * Don't schedule slices shorter than 10000ns, that just
	 * doesn't make sense. Rely on vruntime for fairness.
	 */
	delay = max_t(u64, delay, 10000LL);
331 332
	hrtimer_start(&rq->hrtick_timer, ns_to_ktime(delay),
		      HRTIMER_MODE_REL_PINNED);
333 334
}
#endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
335

336
static void init_rq_hrtick(struct rq *rq)
P
Peter Zijlstra 已提交
337
{
338 339
#ifdef CONFIG_SMP
	rq->hrtick_csd_pending = 0;
P
Peter Zijlstra 已提交
340

341 342 343 344
	rq->hrtick_csd.flags = 0;
	rq->hrtick_csd.func = __hrtick_start;
	rq->hrtick_csd.info = rq;
#endif
P
Peter Zijlstra 已提交
345

346 347
	hrtimer_init(&rq->hrtick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	rq->hrtick_timer.function = hrtick;
P
Peter Zijlstra 已提交
348
}
A
Andrew Morton 已提交
349
#else	/* CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
350 351 352 353 354 355 356
static inline void hrtick_clear(struct rq *rq)
{
}

static inline void init_rq_hrtick(struct rq *rq)
{
}
A
Andrew Morton 已提交
357
#endif	/* CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
358

359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376
/*
 * 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;								\
})

377
#if defined(CONFIG_SMP) && defined(TIF_POLLING_NRFLAG)
378 379 380 381 382 383 384 385 386 387
/*
 * 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);
}
388 389 390 391 392 393 394 395 396 397

/*
 * 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);
398
	typeof(ti->flags) old, val = READ_ONCE(ti->flags);
399 400 401 402 403 404 405 406 407 408 409 410 411 412

	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;
}

413 414 415 416 417 418
#else
static bool set_nr_and_not_polling(struct task_struct *p)
{
	set_tsk_need_resched(p);
	return true;
}
419 420 421 422 423 424 425

#ifdef CONFIG_SMP
static bool set_nr_if_polling(struct task_struct *p)
{
	return false;
}
#endif
426 427
#endif

428 429 430 431 432 433 434 435 436 437
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
438
	 * barrier implied by the wakeup in wake_up_q().
439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460
	 */
	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
461
		/* Task can safely be re-inserted now: */
462 463 464 465 466 467 468 469 470 471 472 473
		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
474
/*
475
 * resched_curr - mark rq's current task 'to be rescheduled now'.
I
Ingo Molnar 已提交
476 477 478 479 480
 *
 * 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.
 */
481
void resched_curr(struct rq *rq)
I
Ingo Molnar 已提交
482
{
483
	struct task_struct *curr = rq->curr;
I
Ingo Molnar 已提交
484 485
	int cpu;

486
	lockdep_assert_held(&rq->lock);
I
Ingo Molnar 已提交
487

488
	if (test_tsk_need_resched(curr))
I
Ingo Molnar 已提交
489 490
		return;

491
	cpu = cpu_of(rq);
492

493
	if (cpu == smp_processor_id()) {
494
		set_tsk_need_resched(curr);
495
		set_preempt_need_resched();
I
Ingo Molnar 已提交
496
		return;
497
	}
I
Ingo Molnar 已提交
498

499
	if (set_nr_and_not_polling(curr))
I
Ingo Molnar 已提交
500
		smp_send_reschedule(cpu);
501 502
	else
		trace_sched_wake_idle_without_ipi(cpu);
I
Ingo Molnar 已提交
503 504
}

505
void resched_cpu(int cpu)
I
Ingo Molnar 已提交
506 507 508 509
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

510
	raw_spin_lock_irqsave(&rq->lock, flags);
511
	resched_curr(rq);
512
	raw_spin_unlock_irqrestore(&rq->lock, flags);
I
Ingo Molnar 已提交
513
}
514

515
#ifdef CONFIG_SMP
516
#ifdef CONFIG_NO_HZ_COMMON
517
/*
I
Ingo Molnar 已提交
518 519
 * In the semi idle case, use the nearest busy CPU for migrating timers
 * from an idle CPU.  This is good for power-savings.
520 521
 *
 * We don't do similar optimization for completely idle system, as
I
Ingo Molnar 已提交
522 523
 * selecting an idle CPU will add more delays to the timers than intended
 * (as that CPU's timer base may not be uptodate wrt jiffies etc).
524
 */
525
int get_nohz_timer_target(void)
526
{
527
	int i, cpu = smp_processor_id();
528 529
	struct sched_domain *sd;

530
	if (!idle_cpu(cpu) && housekeeping_cpu(cpu, HK_FLAG_TIMER))
531 532
		return cpu;

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

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

546 547
	if (!housekeeping_cpu(cpu, HK_FLAG_TIMER))
		cpu = housekeeping_any_cpu(HK_FLAG_TIMER);
548 549
unlock:
	rcu_read_unlock();
550 551
	return cpu;
}
I
Ingo Molnar 已提交
552

553 554 555 556 557 558 559 560 561 562
/*
 * When add_timer_on() enqueues a timer into the timer wheel of an
 * idle CPU then this timer might expire before the next timer event
 * which is scheduled to wake up that CPU. In case of a completely
 * idle system the next event might even be infinite time into the
 * future. wake_up_idle_cpu() ensures that the CPU is woken up and
 * leaves the inner idle loop so the newly added timer is taken into
 * account when the CPU goes back to idle and evaluates the timer
 * wheel for the next timer event.
 */
563
static void wake_up_idle_cpu(int cpu)
564 565 566 567 568 569
{
	struct rq *rq = cpu_rq(cpu);

	if (cpu == smp_processor_id())
		return;

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

576
static bool wake_up_full_nohz_cpu(int cpu)
577
{
578 579 580 581 582 583
	/*
	 * We just need the target to call irq_exit() and re-evaluate
	 * the next tick. The nohz full kick at least implies that.
	 * If needed we can still optimize that later with an
	 * empty IRQ.
	 */
584 585
	if (cpu_is_offline(cpu))
		return true;  /* Don't try to wake offline CPUs. */
586
	if (tick_nohz_full_cpu(cpu)) {
587 588
		if (cpu != smp_processor_id() ||
		    tick_nohz_tick_stopped())
589
			tick_nohz_full_kick_cpu(cpu);
590 591 592 593 594 595
		return true;
	}

	return false;
}

596 597 598 599 600
/*
 * Wake up the specified CPU.  If the CPU is going offline, it is the
 * caller's responsibility to deal with the lost wakeup, for example,
 * by hooking into the CPU_DEAD notifier like timers and hrtimers do.
 */
601 602
void wake_up_nohz_cpu(int cpu)
{
603
	if (!wake_up_full_nohz_cpu(cpu))
604 605 606
		wake_up_idle_cpu(cpu);
}

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

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

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

	/*
	 * We can't run Idle Load Balance on this CPU for this time so we
	 * cancel it and clear NOHZ_BALANCE_KICK
	 */
	clear_bit(NOHZ_BALANCE_KICK, nohz_flags(cpu));
	return false;
623 624
}

625
#else /* CONFIG_NO_HZ_COMMON */
626

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

632
#endif /* CONFIG_NO_HZ_COMMON */
633

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

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

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

654 655 656 657 658 659 660 661 662 663 664 665 666 667
	/*
	 * If there's no RR tasks, but FIFO tasks, we can skip the tick, no
	 * forced preemption between FIFO tasks.
	 */
	fifo_nr_running = rq->rt.rt_nr_running - rq->rt.rr_nr_running;
	if (fifo_nr_running)
		return true;

	/*
	 * If there are no DL,RR/FIFO tasks, there must only be CFS tasks left;
	 * if there's more than one we need the tick for involuntary
	 * preemption.
	 */
	if (rq->nr_running > 1)
668
		return false;
669

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

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

678
	while ((s64)(rq_clock(rq) - rq->age_stamp) > period) {
679 680 681 682 683 684
		/*
		 * Inline assembly required to prevent the compiler
		 * optimising this loop into a divmod call.
		 * See __iter_div_u64_rem() for another example of this.
		 */
		asm("" : "+rm" (rq->age_stamp));
685 686 687
		rq->age_stamp += period;
		rq->rt_avg /= 2;
	}
688 689
}

690
#endif /* CONFIG_SMP */
691

692 693
#if defined(CONFIG_RT_GROUP_SCHED) || (defined(CONFIG_FAIR_GROUP_SCHED) && \
			(defined(CONFIG_SMP) || defined(CONFIG_CFS_BANDWIDTH)))
694
/*
695 696 697 698
 * Iterate task_group tree rooted at *from, calling @down when first entering a
 * node and @up when leaving it for the final time.
 *
 * Caller must hold rcu_lock or sufficient equivalent.
699
 */
700
int walk_tg_tree_from(struct task_group *from,
701
			     tg_visitor down, tg_visitor up, void *data)
702 703
{
	struct task_group *parent, *child;
P
Peter Zijlstra 已提交
704
	int ret;
705

706 707
	parent = from;

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

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

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

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

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

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

751 752 753 754 755 756 757 758 759 760
	/*
	 * SCHED_OTHER tasks have to update their load when changing their
	 * weight
	 */
	if (update_load && p->sched_class == &fair_sched_class) {
		reweight_task(p, prio);
	} else {
		load->weight = scale_load(sched_prio_to_weight[prio]);
		load->inv_weight = sched_prio_to_wmult[prio];
	}
761 762
}

763
static inline void enqueue_task(struct rq *rq, struct task_struct *p, int flags)
764
{
765 766 767
	if (!(flags & ENQUEUE_NOCLOCK))
		update_rq_clock(rq);

768 769
	if (!(flags & ENQUEUE_RESTORE))
		sched_info_queued(rq, p);
770

771
	p->sched_class->enqueue_task(rq, p, flags);
772 773
}

774
static inline void dequeue_task(struct rq *rq, struct task_struct *p, int flags)
775
{
776 777 778
	if (!(flags & DEQUEUE_NOCLOCK))
		update_rq_clock(rq);

779 780
	if (!(flags & DEQUEUE_SAVE))
		sched_info_dequeued(rq, p);
781

782
	p->sched_class->dequeue_task(rq, p, flags);
783 784
}

785
void activate_task(struct rq *rq, struct task_struct *p, int flags)
786 787 788 789
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible--;

790
	enqueue_task(rq, p, flags);
791 792
}

793
void deactivate_task(struct rq *rq, struct task_struct *p, int flags)
794 795 796 797
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible++;

798
	dequeue_task(rq, p, flags);
799 800
}

801
/*
I
Ingo Molnar 已提交
802
 * __normal_prio - return the priority that is based on the static prio
803 804 805
 */
static inline int __normal_prio(struct task_struct *p)
{
I
Ingo Molnar 已提交
806
	return p->static_prio;
807 808
}

809 810 811 812 813 814 815
/*
 * 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.
 */
816
static inline int normal_prio(struct task_struct *p)
817 818 819
{
	int prio;

820 821 822
	if (task_has_dl_policy(p))
		prio = MAX_DL_PRIO-1;
	else if (task_has_rt_policy(p))
823 824 825 826 827 828 829 830 831 832 833 834 835
		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.
 */
836
static int effective_prio(struct task_struct *p)
837 838 839 840 841 842 843 844 845 846 847 848
{
	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
849 850 851
/**
 * task_curr - is this task currently executing on a CPU?
 * @p: the task in question.
852 853
 *
 * Return: 1 if the task is currently executing. 0 otherwise.
L
Linus Torvalds 已提交
854
 */
855
inline int task_curr(const struct task_struct *p)
L
Linus Torvalds 已提交
856 857 858 859
{
	return cpu_curr(task_cpu(p)) == p;
}

860
/*
861 862 863 864 865
 * 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().
866
 */
867 868
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
869
				       int oldprio)
870 871 872
{
	if (prev_class != p->sched_class) {
		if (prev_class->switched_from)
P
Peter Zijlstra 已提交
873
			prev_class->switched_from(rq, p);
874

P
Peter Zijlstra 已提交
875
		p->sched_class->switched_to(rq, p);
876
	} else if (oldprio != p->prio || dl_task(p))
P
Peter Zijlstra 已提交
877
		p->sched_class->prio_changed(rq, p, oldprio);
878 879
}

880
void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags)
881 882 883 884 885 886 887 888 889 890
{
	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) {
891
				resched_curr(rq);
892 893 894 895 896 897 898 899 900
				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.
	 */
901
	if (task_on_rq_queued(rq->curr) && test_tsk_need_resched(rq->curr))
902
		rq_clock_skip_update(rq, true);
903 904
}

L
Linus Torvalds 已提交
905
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924
/*
 * 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.
 */
925 926
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
927 928 929 930
{
	lockdep_assert_held(&rq->lock);

	p->on_rq = TASK_ON_RQ_MIGRATING;
931
	dequeue_task(rq, p, DEQUEUE_NOCLOCK);
P
Peter Zijlstra 已提交
932
	set_task_cpu(p, new_cpu);
933
	rq_unlock(rq, rf);
P
Peter Zijlstra 已提交
934 935 936

	rq = cpu_rq(new_cpu);

937
	rq_lock(rq, rf);
P
Peter Zijlstra 已提交
938 939
	BUG_ON(task_cpu(p) != new_cpu);
	enqueue_task(rq, p, 0);
940
	p->on_rq = TASK_ON_RQ_QUEUED;
P
Peter Zijlstra 已提交
941 942 943 944 945 946 947 948 949 950 951
	check_preempt_curr(rq, p, 0);

	return rq;
}

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

/*
I
Ingo Molnar 已提交
952
 * Move (not current) task off this CPU, onto the destination CPU. We're doing
P
Peter Zijlstra 已提交
953 954 955 956 957 958 959
 * 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.
 */
960 961
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
962
{
963 964 965 966 967 968 969
	if (p->flags & PF_KTHREAD) {
		if (unlikely(!cpu_online(dest_cpu)))
			return rq;
	} else {
		if (unlikely(!cpu_active(dest_cpu)))
			return rq;
	}
P
Peter Zijlstra 已提交
970 971

	/* Affinity changed (again). */
972
	if (!cpumask_test_cpu(dest_cpu, &p->cpus_allowed))
973
		return rq;
P
Peter Zijlstra 已提交
974

975
	update_rq_clock(rq);
976
	rq = move_queued_task(rq, rf, p, dest_cpu);
977 978

	return rq;
P
Peter Zijlstra 已提交
979 980 981 982 983 984 985 986 987 988
}

/*
 * 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;
989 990
	struct task_struct *p = arg->task;
	struct rq *rq = this_rq();
991
	struct rq_flags rf;
P
Peter Zijlstra 已提交
992 993

	/*
I
Ingo Molnar 已提交
994 995
	 * 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
996 997 998 999 1000 1001 1002 1003
	 */
	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();
1004 1005

	raw_spin_lock(&p->pi_lock);
1006
	rq_lock(rq, &rf);
1007 1008 1009 1010 1011
	/*
	 * 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.
	 */
1012 1013
	if (task_rq(p) == rq) {
		if (task_on_rq_queued(p))
1014
			rq = __migrate_task(rq, &rf, p, arg->dest_cpu);
1015 1016 1017
		else
			p->wake_cpu = arg->dest_cpu;
	}
1018
	rq_unlock(rq, &rf);
1019 1020
	raw_spin_unlock(&p->pi_lock);

P
Peter Zijlstra 已提交
1021 1022 1023 1024
	local_irq_enable();
	return 0;
}

1025 1026 1027 1028 1029
/*
 * 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
1030 1031 1032 1033 1034
{
	cpumask_copy(&p->cpus_allowed, new_mask);
	p->nr_cpus_allowed = cpumask_weight(new_mask);
}

1035 1036
void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
{
1037 1038 1039
	struct rq *rq = task_rq(p);
	bool queued, running;

1040
	lockdep_assert_held(&p->pi_lock);
1041 1042 1043 1044 1045 1046 1047 1048 1049 1050

	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);
1051
		dequeue_task(rq, p, DEQUEUE_SAVE | DEQUEUE_NOCLOCK);
1052 1053 1054 1055
	}
	if (running)
		put_prev_task(rq, p);

1056
	p->sched_class->set_cpus_allowed(p, new_mask);
1057 1058

	if (queued)
1059
		enqueue_task(rq, p, ENQUEUE_RESTORE | ENQUEUE_NOCLOCK);
1060
	if (running)
1061
		set_curr_task(rq, p);
1062 1063
}

P
Peter Zijlstra 已提交
1064 1065 1066 1067 1068 1069 1070 1071 1072
/*
 * 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.
 */
1073 1074
static int __set_cpus_allowed_ptr(struct task_struct *p,
				  const struct cpumask *new_mask, bool check)
P
Peter Zijlstra 已提交
1075
{
1076
	const struct cpumask *cpu_valid_mask = cpu_active_mask;
P
Peter Zijlstra 已提交
1077
	unsigned int dest_cpu;
1078 1079
	struct rq_flags rf;
	struct rq *rq;
P
Peter Zijlstra 已提交
1080 1081
	int ret = 0;

1082
	rq = task_rq_lock(p, &rf);
1083
	update_rq_clock(rq);
P
Peter Zijlstra 已提交
1084

1085 1086 1087 1088 1089 1090 1091
	if (p->flags & PF_KTHREAD) {
		/*
		 * Kernel threads are allowed on online && !active CPUs
		 */
		cpu_valid_mask = cpu_online_mask;
	}

1092 1093 1094 1095 1096 1097 1098 1099 1100
	/*
	 * 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
1101 1102 1103
	if (cpumask_equal(&p->cpus_allowed, new_mask))
		goto out;

1104
	if (!cpumask_intersects(new_mask, cpu_valid_mask)) {
P
Peter Zijlstra 已提交
1105 1106 1107 1108 1109 1110
		ret = -EINVAL;
		goto out;
	}

	do_set_cpus_allowed(p, new_mask);

1111 1112 1113
	if (p->flags & PF_KTHREAD) {
		/*
		 * For kernel threads that do indeed end up on online &&
I
Ingo Molnar 已提交
1114
		 * !active we want to ensure they are strict per-CPU threads.
1115 1116 1117 1118 1119 1120
		 */
		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
1121 1122 1123 1124
	/* 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;

1125
	dest_cpu = cpumask_any_and(cpu_valid_mask, new_mask);
P
Peter Zijlstra 已提交
1126 1127 1128
	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. */
1129
		task_rq_unlock(rq, p, &rf);
P
Peter Zijlstra 已提交
1130 1131 1132
		stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg);
		tlb_migrate_finish(p->mm);
		return 0;
1133 1134 1135 1136 1137
	} else if (task_on_rq_queued(p)) {
		/*
		 * OK, since we're going to drop the lock immediately
		 * afterwards anyway.
		 */
1138
		rq = move_queued_task(rq, &rf, p, dest_cpu);
1139
	}
P
Peter Zijlstra 已提交
1140
out:
1141
	task_rq_unlock(rq, p, &rf);
P
Peter Zijlstra 已提交
1142 1143 1144

	return ret;
}
1145 1146 1147 1148 1149

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
1150 1151
EXPORT_SYMBOL_GPL(set_cpus_allowed_ptr);

I
Ingo Molnar 已提交
1152
void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
I
Ingo Molnar 已提交
1153
{
1154 1155 1156 1157 1158
#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
1159
	WARN_ON_ONCE(p->state != TASK_RUNNING && p->state != TASK_WAKING &&
O
Oleg Nesterov 已提交
1160
			!p->on_rq);
1161

1162 1163 1164 1165 1166 1167 1168 1169 1170
	/*
	 * 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)));

1171
#ifdef CONFIG_LOCKDEP
1172 1173 1174 1175 1176
	/*
	 * 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
1177
	 * see task_group().
1178 1179 1180 1181
	 *
	 * Furthermore, all task_rq users should acquire both locks, see
	 * task_rq_lock().
	 */
1182 1183 1184
	WARN_ON_ONCE(debug_locks && !(lockdep_is_held(&p->pi_lock) ||
				      lockdep_is_held(&task_rq(p)->lock)));
#endif
1185 1186 1187 1188
	/*
	 * Clearly, migrating tasks to offline CPUs is a fairly daft thing.
	 */
	WARN_ON_ONCE(!cpu_online(new_cpu));
1189 1190
#endif

1191
	trace_sched_migrate_task(p, new_cpu);
1192

1193
	if (task_cpu(p) != new_cpu) {
1194
		if (p->sched_class->migrate_task_rq)
1195
			p->sched_class->migrate_task_rq(p);
1196
		p->se.nr_migrations++;
1197
		perf_event_task_migrate(p);
1198
	}
I
Ingo Molnar 已提交
1199 1200

	__set_task_cpu(p, new_cpu);
I
Ingo Molnar 已提交
1201 1202
}

1203 1204
static void __migrate_swap_task(struct task_struct *p, int cpu)
{
1205
	if (task_on_rq_queued(p)) {
1206
		struct rq *src_rq, *dst_rq;
1207
		struct rq_flags srf, drf;
1208 1209 1210 1211

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

1212 1213 1214
		rq_pin_lock(src_rq, &srf);
		rq_pin_lock(dst_rq, &drf);

1215
		p->on_rq = TASK_ON_RQ_MIGRATING;
1216 1217 1218
		deactivate_task(src_rq, p, 0);
		set_task_cpu(p, cpu);
		activate_task(dst_rq, p, 0);
1219
		p->on_rq = TASK_ON_RQ_QUEUED;
1220
		check_preempt_curr(dst_rq, p, 0);
1221 1222 1223 1224

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

1225 1226 1227 1228
	} else {
		/*
		 * Task isn't running anymore; make it appear like we migrated
		 * it before it went to sleep. This means on wakeup we make the
I
Ingo Molnar 已提交
1229
		 * previous CPU our target instead of where it really is.
1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245
		 */
		p->wake_cpu = cpu;
	}
}

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

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

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

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

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

1256 1257 1258 1259 1260 1261
	if (task_cpu(arg->dst_task) != arg->dst_cpu)
		goto unlock;

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

1262
	if (!cpumask_test_cpu(arg->dst_cpu, &arg->src_task->cpus_allowed))
1263 1264
		goto unlock;

1265
	if (!cpumask_test_cpu(arg->src_cpu, &arg->dst_task->cpus_allowed))
1266 1267 1268 1269 1270 1271 1272 1273 1274
		goto unlock;

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

	ret = 0;

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

	return ret;
}

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

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

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

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

1306
	if (!cpumask_test_cpu(arg.dst_cpu, &arg.src_task->cpus_allowed))
1307 1308
		goto out;

1309
	if (!cpumask_test_cpu(arg.src_cpu, &arg.dst_task->cpus_allowed))
1310 1311
		goto out;

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

out:
	return ret;
}

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

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

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

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

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

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

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

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

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

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

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

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

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

1481
	/*
I
Ingo Molnar 已提交
1482 1483 1484
	 * If the node that the CPU is on has been offlined, cpu_to_node()
	 * will return -1. There is no CPU on the node, and we should
	 * select the CPU on the other node.
1485 1486 1487 1488 1489 1490 1491 1492
	 */
	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;
1493
			if (cpumask_test_cpu(dest_cpu, &p->cpus_allowed))
1494 1495
				return dest_cpu;
		}
1496
	}
1497

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

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

		case fail:
			BUG();
			break;
		}
	}

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

	return dest_cpu;
}

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

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

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

	return cpu;
1572
}
1573 1574 1575 1576 1577 1578

static void update_avg(u64 *avg, u64 sample)
{
	s64 diff = sample - *avg;
	*avg += diff >> 3;
}
1579

1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609
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;
	}
}

1610 1611 1612 1613 1614 1615 1616 1617
#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
1618
#endif /* CONFIG_SMP */
1619

P
Peter Zijlstra 已提交
1620
static void
1621
ttwu_stat(struct task_struct *p, int cpu, int wake_flags)
T
Tejun Heo 已提交
1622
{
1623
	struct rq *rq;
1624

1625 1626 1627 1628
	if (!schedstat_enabled())
		return;

	rq = this_rq();
P
Peter Zijlstra 已提交
1629

1630 1631
#ifdef CONFIG_SMP
	if (cpu == rq->cpu) {
1632 1633
		schedstat_inc(rq->ttwu_local);
		schedstat_inc(p->se.statistics.nr_wakeups_local);
P
Peter Zijlstra 已提交
1634 1635 1636
	} else {
		struct sched_domain *sd;

1637
		schedstat_inc(p->se.statistics.nr_wakeups_remote);
1638
		rcu_read_lock();
1639
		for_each_domain(rq->cpu, sd) {
P
Peter Zijlstra 已提交
1640
			if (cpumask_test_cpu(cpu, sched_domain_span(sd))) {
1641
				schedstat_inc(sd->ttwu_wake_remote);
P
Peter Zijlstra 已提交
1642 1643 1644
				break;
			}
		}
1645
		rcu_read_unlock();
P
Peter Zijlstra 已提交
1646
	}
1647 1648

	if (wake_flags & WF_MIGRATED)
1649
		schedstat_inc(p->se.statistics.nr_wakeups_migrate);
P
Peter Zijlstra 已提交
1650 1651
#endif /* CONFIG_SMP */

1652 1653
	schedstat_inc(rq->ttwu_count);
	schedstat_inc(p->se.statistics.nr_wakeups);
P
Peter Zijlstra 已提交
1654 1655

	if (wake_flags & WF_SYNC)
1656
		schedstat_inc(p->se.statistics.nr_wakeups_sync);
P
Peter Zijlstra 已提交
1657 1658
}

1659
static inline void ttwu_activate(struct rq *rq, struct task_struct *p, int en_flags)
P
Peter Zijlstra 已提交
1660
{
T
Tejun Heo 已提交
1661
	activate_task(rq, p, en_flags);
1662
	p->on_rq = TASK_ON_RQ_QUEUED;
1663

I
Ingo Molnar 已提交
1664
	/* If a worker is waking up, notify the workqueue: */
1665 1666
	if (p->flags & PF_WQ_WORKER)
		wq_worker_waking_up(p, cpu_of(rq));
T
Tejun Heo 已提交
1667 1668
}

1669 1670 1671
/*
 * Mark the task runnable and perform wakeup-preemption.
 */
1672
static void ttwu_do_wakeup(struct rq *rq, struct task_struct *p, int wake_flags,
1673
			   struct rq_flags *rf)
T
Tejun Heo 已提交
1674 1675 1676
{
	check_preempt_curr(rq, p, wake_flags);
	p->state = TASK_RUNNING;
1677 1678
	trace_sched_wakeup(p);

T
Tejun Heo 已提交
1679
#ifdef CONFIG_SMP
1680 1681
	if (p->sched_class->task_woken) {
		/*
1682 1683
		 * Our task @p is fully woken up and running; so its safe to
		 * drop the rq->lock, hereafter rq is only used for statistics.
1684
		 */
1685
		rq_unpin_lock(rq, rf);
T
Tejun Heo 已提交
1686
		p->sched_class->task_woken(rq, p);
1687
		rq_repin_lock(rq, rf);
1688
	}
T
Tejun Heo 已提交
1689

1690
	if (rq->idle_stamp) {
1691
		u64 delta = rq_clock(rq) - rq->idle_stamp;
1692
		u64 max = 2*rq->max_idle_balance_cost;
T
Tejun Heo 已提交
1693

1694 1695 1696
		update_avg(&rq->avg_idle, delta);

		if (rq->avg_idle > max)
T
Tejun Heo 已提交
1697
			rq->avg_idle = max;
1698

T
Tejun Heo 已提交
1699 1700 1701 1702 1703
		rq->idle_stamp = 0;
	}
#endif
}

1704
static void
1705
ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags,
1706
		 struct rq_flags *rf)
1707
{
1708
	int en_flags = ENQUEUE_WAKEUP | ENQUEUE_NOCLOCK;
1709

1710 1711
	lockdep_assert_held(&rq->lock);

1712 1713 1714
#ifdef CONFIG_SMP
	if (p->sched_contributes_to_load)
		rq->nr_uninterruptible--;
1715 1716

	if (wake_flags & WF_MIGRATED)
1717
		en_flags |= ENQUEUE_MIGRATED;
1718 1719
#endif

1720
	ttwu_activate(rq, p, en_flags);
1721
	ttwu_do_wakeup(rq, p, wake_flags, rf);
1722 1723 1724 1725 1726 1727 1728 1729 1730 1731
}

/*
 * 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)
{
1732
	struct rq_flags rf;
1733 1734 1735
	struct rq *rq;
	int ret = 0;

1736
	rq = __task_rq_lock(p, &rf);
1737
	if (task_on_rq_queued(p)) {
1738 1739
		/* check_preempt_curr() may use rq clock */
		update_rq_clock(rq);
1740
		ttwu_do_wakeup(rq, p, wake_flags, &rf);
1741 1742
		ret = 1;
	}
1743
	__task_rq_unlock(rq, &rf);
1744 1745 1746 1747

	return ret;
}

1748
#ifdef CONFIG_SMP
1749
void sched_ttwu_pending(void)
1750 1751
{
	struct rq *rq = this_rq();
P
Peter Zijlstra 已提交
1752
	struct llist_node *llist = llist_del_all(&rq->wake_list);
1753
	struct task_struct *p, *t;
1754
	struct rq_flags rf;
1755

1756 1757 1758
	if (!llist)
		return;

1759
	rq_lock_irqsave(rq, &rf);
1760
	update_rq_clock(rq);
1761

1762 1763
	llist_for_each_entry_safe(p, t, llist, wake_entry)
		ttwu_do_activate(rq, p, p->sched_remote_wakeup ? WF_MIGRATED : 0, &rf);
1764

1765
	rq_unlock_irqrestore(rq, &rf);
1766 1767 1768 1769
}

void scheduler_ipi(void)
{
1770 1771 1772 1773 1774
	/*
	 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
	 * TIF_NEED_RESCHED remotely (for the first time) will also send
	 * this IPI.
	 */
1775
	preempt_fold_need_resched();
1776

1777
	if (llist_empty(&this_rq()->wake_list) && !got_nohz_idle_kick())
1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793
		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
1794
	sched_ttwu_pending();
1795 1796 1797 1798

	/*
	 * Check if someone kicked us for doing the nohz idle load balance.
	 */
1799
	if (unlikely(got_nohz_idle_kick())) {
1800
		this_rq()->idle_balance = 1;
1801
		raise_softirq_irqoff(SCHED_SOFTIRQ);
1802
	}
1803
	irq_exit();
1804 1805
}

P
Peter Zijlstra 已提交
1806
static void ttwu_queue_remote(struct task_struct *p, int cpu, int wake_flags)
1807
{
1808 1809
	struct rq *rq = cpu_rq(cpu);

P
Peter Zijlstra 已提交
1810 1811
	p->sched_remote_wakeup = !!(wake_flags & WF_MIGRATED);

1812 1813 1814 1815 1816 1817
	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);
	}
1818
}
1819

1820 1821 1822
void wake_up_if_idle(int cpu)
{
	struct rq *rq = cpu_rq(cpu);
1823
	struct rq_flags rf;
1824

1825 1826 1827 1828
	rcu_read_lock();

	if (!is_idle_task(rcu_dereference(rq->curr)))
		goto out;
1829 1830 1831 1832

	if (set_nr_if_polling(rq->idle)) {
		trace_sched_wake_idle_without_ipi(cpu);
	} else {
1833
		rq_lock_irqsave(rq, &rf);
1834 1835
		if (is_idle_task(rq->curr))
			smp_send_reschedule(cpu);
I
Ingo Molnar 已提交
1836
		/* Else CPU is not idle, do nothing here: */
1837
		rq_unlock_irqrestore(rq, &rf);
1838
	}
1839 1840 1841

out:
	rcu_read_unlock();
1842 1843
}

1844
bool cpus_share_cache(int this_cpu, int that_cpu)
1845 1846 1847
{
	return per_cpu(sd_llc_id, this_cpu) == per_cpu(sd_llc_id, that_cpu);
}
1848
#endif /* CONFIG_SMP */
1849

1850
static void ttwu_queue(struct task_struct *p, int cpu, int wake_flags)
1851 1852
{
	struct rq *rq = cpu_rq(cpu);
1853
	struct rq_flags rf;
1854

1855
#if defined(CONFIG_SMP)
1856
	if (sched_feat(TTWU_QUEUE) && !cpus_share_cache(smp_processor_id(), cpu)) {
I
Ingo Molnar 已提交
1857
		sched_clock_cpu(cpu); /* Sync clocks across CPUs */
P
Peter Zijlstra 已提交
1858
		ttwu_queue_remote(p, cpu, wake_flags);
1859 1860 1861 1862
		return;
	}
#endif

1863
	rq_lock(rq, &rf);
1864
	update_rq_clock(rq);
1865
	ttwu_do_activate(rq, p, wake_flags, &rf);
1866
	rq_unlock(rq, &rf);
T
Tejun Heo 已提交
1867 1868
}

1869 1870 1871 1872 1873 1874
/*
 * 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
1875 1876
 * migrates, all its activity on its old CPU [c0] happens-before any subsequent
 * execution on its new CPU [c1].
1877 1878 1879 1880 1881 1882 1883 1884 1885 1886
 *
 * 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
1887
 * Note: the CPU doing B need not be c0 or c1
1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918
 *
 * 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)
1919
 *   2) smp_cond_load_acquire(!X->on_cpu)
1920 1921 1922 1923 1924 1925 1926 1927 1928 1929
 *
 * 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);
 *
1930
 *                    smp_cond_load_acquire(&X->on_cpu, !VAL);
1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955
 *                    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,
1956
 * since the waking CPU is the one issueing the ACQUIRE (smp_cond_load_acquire).
1957 1958 1959
 *
 */

T
Tejun Heo 已提交
1960
/**
L
Linus Torvalds 已提交
1961
 * try_to_wake_up - wake up a thread
T
Tejun Heo 已提交
1962
 * @p: the thread to be awakened
L
Linus Torvalds 已提交
1963
 * @state: the mask of task states that can be woken
T
Tejun Heo 已提交
1964
 * @wake_flags: wake modifier flags (WF_*)
L
Linus Torvalds 已提交
1965
 *
1966
 * If (@state & @p->state) @p->state = TASK_RUNNING.
L
Linus Torvalds 已提交
1967
 *
1968 1969 1970 1971 1972 1973 1974
 * 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
1975
 */
1976 1977
static int
try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
L
Linus Torvalds 已提交
1978 1979
{
	unsigned long flags;
1980
	int cpu, success = 0;
P
Peter Zijlstra 已提交
1981

1982 1983 1984 1985 1986 1987
	/*
	 * 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.
	 */
1988
	raw_spin_lock_irqsave(&p->pi_lock, flags);
1989
	smp_mb__after_spinlock();
P
Peter Zijlstra 已提交
1990
	if (!(p->state & state))
L
Linus Torvalds 已提交
1991 1992
		goto out;

1993 1994
	trace_sched_waking(p);

I
Ingo Molnar 已提交
1995 1996
	/* We're going to change ->state: */
	success = 1;
L
Linus Torvalds 已提交
1997 1998
	cpu = task_cpu(p);

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
	/*
	 * 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();
2021 2022
	if (p->on_rq && ttwu_remote(p, wake_flags))
		goto stat;
L
Linus Torvalds 已提交
2023 2024

#ifdef CONFIG_SMP
2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043
	/*
	 * 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
2044
	/*
I
Ingo Molnar 已提交
2045
	 * If the owning (remote) CPU is still in the middle of schedule() with
2046
	 * this task as prev, wait until its done referencing the task.
2047
	 *
2048
	 * Pairs with the smp_store_release() in finish_task().
2049 2050 2051
	 *
	 * This ensures that tasks getting woken will be fully ordered against
	 * their previous state and preserve Program Order.
2052
	 */
2053
	smp_cond_load_acquire(&p->on_cpu, !VAL);
L
Linus Torvalds 已提交
2054

2055
	p->sched_contributes_to_load = !!task_contributes_to_load(p);
P
Peter Zijlstra 已提交
2056
	p->state = TASK_WAKING;
2057

2058 2059 2060 2061 2062
	if (p->in_iowait) {
		delayacct_blkio_end();
		atomic_dec(&task_rq(p)->nr_iowait);
	}

2063
	cpu = select_task_rq(p, p->wake_cpu, SD_BALANCE_WAKE, wake_flags);
2064 2065
	if (task_cpu(p) != cpu) {
		wake_flags |= WF_MIGRATED;
2066
		set_task_cpu(p, cpu);
2067
	}
2068 2069 2070 2071 2072 2073 2074 2075

#else /* CONFIG_SMP */

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

L
Linus Torvalds 已提交
2076 2077
#endif /* CONFIG_SMP */

2078
	ttwu_queue(p, cpu, wake_flags);
2079
stat:
2080
	ttwu_stat(p, cpu, wake_flags);
L
Linus Torvalds 已提交
2081
out:
2082
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
2083 2084 2085 2086

	return success;
}

T
Tejun Heo 已提交
2087 2088 2089
/**
 * try_to_wake_up_local - try to wake up a local task with rq lock held
 * @p: the thread to be awakened
2090
 * @rf: request-queue flags for pinning
T
Tejun Heo 已提交
2091
 *
2092
 * Put @p on the run-queue if it's not already there. The caller must
T
Tejun Heo 已提交
2093
 * ensure that this_rq() is locked, @p is bound to this_rq() and not
2094
 * the current task.
T
Tejun Heo 已提交
2095
 */
2096
static void try_to_wake_up_local(struct task_struct *p, struct rq_flags *rf)
T
Tejun Heo 已提交
2097 2098 2099
{
	struct rq *rq = task_rq(p);

2100 2101 2102 2103
	if (WARN_ON_ONCE(rq != this_rq()) ||
	    WARN_ON_ONCE(p == current))
		return;

T
Tejun Heo 已提交
2104 2105
	lockdep_assert_held(&rq->lock);

2106
	if (!raw_spin_trylock(&p->pi_lock)) {
2107 2108 2109 2110 2111 2112
		/*
		 * 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.
		 */
2113
		rq_unlock(rq, rf);
2114
		raw_spin_lock(&p->pi_lock);
2115
		rq_relock(rq, rf);
2116 2117
	}

T
Tejun Heo 已提交
2118
	if (!(p->state & TASK_NORMAL))
2119
		goto out;
T
Tejun Heo 已提交
2120

2121 2122
	trace_sched_waking(p);

2123 2124 2125 2126 2127
	if (!task_on_rq_queued(p)) {
		if (p->in_iowait) {
			delayacct_blkio_end();
			atomic_dec(&rq->nr_iowait);
		}
2128
		ttwu_activate(rq, p, ENQUEUE_WAKEUP | ENQUEUE_NOCLOCK);
2129
	}
P
Peter Zijlstra 已提交
2130

2131
	ttwu_do_wakeup(rq, p, 0, rf);
2132
	ttwu_stat(p, smp_processor_id(), 0);
2133 2134
out:
	raw_spin_unlock(&p->pi_lock);
T
Tejun Heo 已提交
2135 2136
}

2137 2138 2139 2140 2141
/**
 * 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
2142 2143 2144
 * processes.
 *
 * Return: 1 if the process was woken up, 0 if it was already running.
2145 2146 2147 2148
 *
 * 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.
 */
2149
int wake_up_process(struct task_struct *p)
L
Linus Torvalds 已提交
2150
{
2151
	return try_to_wake_up(p, TASK_NORMAL, 0);
L
Linus Torvalds 已提交
2152 2153 2154
}
EXPORT_SYMBOL(wake_up_process);

2155
int wake_up_state(struct task_struct *p, unsigned int state)
L
Linus Torvalds 已提交
2156 2157 2158 2159 2160 2161 2162
{
	return try_to_wake_up(p, state, 0);
}

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

	p->se.on_rq			= 0;
I
Ingo Molnar 已提交
2171 2172
	p->se.exec_start		= 0;
	p->se.sum_exec_runtime		= 0;
2173
	p->se.prev_sum_exec_runtime	= 0;
2174
	p->se.nr_migrations		= 0;
P
Peter Zijlstra 已提交
2175
	p->se.vruntime			= 0;
P
Peter Zijlstra 已提交
2176
	INIT_LIST_HEAD(&p->se.group_node);
I
Ingo Molnar 已提交
2177

2178 2179 2180 2181
#ifdef CONFIG_FAIR_GROUP_SCHED
	p->se.cfs_rq			= NULL;
#endif

I
Ingo Molnar 已提交
2182
#ifdef CONFIG_SCHEDSTATS
2183
	/* Even if schedstat is disabled, there should not be garbage */
2184
	memset(&p->se.statistics, 0, sizeof(p->se.statistics));
I
Ingo Molnar 已提交
2185
#endif
N
Nick Piggin 已提交
2186

2187
	RB_CLEAR_NODE(&p->dl.rb_node);
2188
	init_dl_task_timer(&p->dl);
2189
	init_dl_inactive_task_timer(&p->dl);
2190
	__dl_clear_params(p);
2191

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

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

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

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

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

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

2225 2226
DEFINE_STATIC_KEY_FALSE(sched_numa_balancing);

2227
#ifdef CONFIG_NUMA_BALANCING
2228

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

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

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

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

2260 2261
#ifdef CONFIG_SCHEDSTATS

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

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

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

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

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

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

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

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

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

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

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

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

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

2358 2359 2360 2361
	/*
	 * Revert to default priority/policy on fork if requested.
	 */
	if (unlikely(p->sched_reset_on_fork)) {
2362
		if (task_has_dl_policy(p) || task_has_rt_policy(p)) {
2363
			p->policy = SCHED_NORMAL;
2364
			p->static_prio = NICE_TO_PRIO(0);
2365 2366 2367 2368 2369
			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);
2370
		set_load_weight(p, false);
2371

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

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

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

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

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

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

2424 2425 2426
unsigned long to_ratio(u64 period, u64 runtime)
{
	if (runtime == RUNTIME_INF)
2427
		return BW_UNIT;
2428 2429 2430 2431 2432 2433 2434 2435 2436

	/*
	 * 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;

2437
	return div64_u64(runtime << BW_SHIFT, period);
2438 2439
}

L
Linus Torvalds 已提交
2440 2441 2442 2443 2444 2445 2446
/*
 * 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.
 */
2447
void wake_up_new_task(struct task_struct *p)
L
Linus Torvalds 已提交
2448
{
2449
	struct rq_flags rf;
I
Ingo Molnar 已提交
2450
	struct rq *rq;
2451

2452
	raw_spin_lock_irqsave(&p->pi_lock, rf.flags);
2453
	p->state = TASK_RUNNING;
2454 2455 2456 2457
#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
2458
	 *  - any previously selected CPU might disappear through hotplug
2459 2460 2461
	 *
	 * Use __set_task_cpu() to avoid calling sched_class::migrate_task_rq,
	 * as we're not fully set-up yet.
2462
	 */
2463
	__set_task_cpu(p, select_task_rq(p, task_cpu(p), SD_BALANCE_FORK, 0));
2464
#endif
2465
	rq = __task_rq_lock(p, &rf);
2466
	update_rq_clock(rq);
2467
	post_init_entity_util_avg(&p->se);
2468

2469
	activate_task(rq, p, ENQUEUE_NOCLOCK);
2470
	p->on_rq = TASK_ON_RQ_QUEUED;
2471
	trace_sched_wakeup_new(p);
P
Peter Zijlstra 已提交
2472
	check_preempt_curr(rq, p, WF_FORK);
2473
#ifdef CONFIG_SMP
2474 2475 2476 2477 2478
	if (p->sched_class->task_woken) {
		/*
		 * Nothing relies on rq->lock after this, so its fine to
		 * drop it.
		 */
2479
		rq_unpin_lock(rq, &rf);
2480
		p->sched_class->task_woken(rq, p);
2481
		rq_repin_lock(rq, &rf);
2482
	}
2483
#endif
2484
	task_rq_unlock(rq, p, &rf);
L
Linus Torvalds 已提交
2485 2486
}

2487 2488
#ifdef CONFIG_PREEMPT_NOTIFIERS

2489 2490
static struct static_key preempt_notifier_key = STATIC_KEY_INIT_FALSE;

2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502
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);

2503
/**
2504
 * preempt_notifier_register - tell me when current is being preempted & rescheduled
R
Randy Dunlap 已提交
2505
 * @notifier: notifier struct to register
2506 2507 2508
 */
void preempt_notifier_register(struct preempt_notifier *notifier)
{
2509 2510 2511
	if (!static_key_false(&preempt_notifier_key))
		WARN(1, "registering preempt_notifier while notifiers disabled\n");

2512 2513 2514 2515 2516 2517
	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
2518
 * @notifier: notifier struct to unregister
2519
 *
2520
 * This is *not* safe to call from within a preemption notifier.
2521 2522 2523 2524 2525 2526 2527
 */
void preempt_notifier_unregister(struct preempt_notifier *notifier)
{
	hlist_del(&notifier->link);
}
EXPORT_SYMBOL_GPL(preempt_notifier_unregister);

2528
static void __fire_sched_in_preempt_notifiers(struct task_struct *curr)
2529 2530 2531
{
	struct preempt_notifier *notifier;

2532
	hlist_for_each_entry(notifier, &curr->preempt_notifiers, link)
2533 2534 2535
		notifier->ops->sched_in(notifier, raw_smp_processor_id());
}

2536 2537 2538 2539 2540 2541
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);
}

2542
static void
2543 2544
__fire_sched_out_preempt_notifiers(struct task_struct *curr,
				   struct task_struct *next)
2545 2546 2547
{
	struct preempt_notifier *notifier;

2548
	hlist_for_each_entry(notifier, &curr->preempt_notifiers, link)
2549 2550 2551
		notifier->ops->sched_out(notifier, next);
}

2552 2553 2554 2555 2556 2557 2558 2559
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);
}

2560
#else /* !CONFIG_PREEMPT_NOTIFIERS */
2561

2562
static inline void fire_sched_in_preempt_notifiers(struct task_struct *curr)
2563 2564 2565
{
}

2566
static inline void
2567 2568 2569 2570 2571
fire_sched_out_preempt_notifiers(struct task_struct *curr,
				 struct task_struct *next)
{
}

2572
#endif /* CONFIG_PREEMPT_NOTIFIERS */
2573

2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617
static inline void prepare_task(struct task_struct *next)
{
#ifdef CONFIG_SMP
	/*
	 * Claim the task as running, we do this before switching to it
	 * such that any running task will have this set.
	 */
	next->on_cpu = 1;
#endif
}

static inline void finish_task(struct task_struct *prev)
{
#ifdef CONFIG_SMP
	/*
	 * After ->on_cpu is cleared, the task can be moved to a different CPU.
	 * We must ensure this doesn't happen until the switch is completely
	 * finished.
	 *
	 * In particular, the load of prev->state in finish_task_switch() must
	 * happen before this.
	 *
	 * Pairs with the smp_cond_load_acquire() in try_to_wake_up().
	 */
	smp_store_release(&prev->on_cpu, 0);
#endif
}

static inline void finish_lock_switch(struct rq *rq)
{
#ifdef CONFIG_DEBUG_SPINLOCK
	/* this is a valid case when another task releases the spinlock */
	rq->lock.owner = current;
#endif
	/*
	 * If we are tracking spinlock dependencies then we have to
	 * fix up the runqueue lock - which gets 'carried over' from
	 * prev into current:
	 */
	spin_acquire(&rq->lock.dep_map, 0, 0, _THIS_IP_);

	raw_spin_unlock_irq(&rq->lock);
}

2618 2619 2620
/**
 * prepare_task_switch - prepare to switch tasks
 * @rq: the runqueue preparing to switch
R
Randy Dunlap 已提交
2621
 * @prev: the current task that is being switched out
2622 2623 2624 2625 2626 2627 2628 2629 2630
 * @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.
 */
2631 2632 2633
static inline void
prepare_task_switch(struct rq *rq, struct task_struct *prev,
		    struct task_struct *next)
2634
{
2635
	sched_info_switch(rq, prev, next);
2636
	perf_event_task_sched_out(prev, next);
2637
	fire_sched_out_preempt_notifiers(prev, next);
2638
	prepare_task(next);
2639 2640 2641
	prepare_arch_switch(next);
}

L
Linus Torvalds 已提交
2642 2643 2644 2645
/**
 * finish_task_switch - clean up after a task-switch
 * @prev: the thread we just switched away from.
 *
2646 2647 2648 2649
 * 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
2650 2651
 *
 * Note that we may have delayed dropping an mm in context_switch(). If
I
Ingo Molnar 已提交
2652
 * so, we finish that here outside of the runqueue lock. (Doing it
L
Linus Torvalds 已提交
2653 2654
 * with the lock held can cause deadlocks; see schedule() for
 * details.)
2655 2656 2657 2658 2659
 *
 * 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
2660
 */
2661
static struct rq *finish_task_switch(struct task_struct *prev)
L
Linus Torvalds 已提交
2662 2663
	__releases(rq->lock)
{
2664
	struct rq *rq = this_rq();
L
Linus Torvalds 已提交
2665
	struct mm_struct *mm = rq->prev_mm;
O
Oleg Nesterov 已提交
2666
	long prev_state;
L
Linus Torvalds 已提交
2667

2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678
	/*
	 * 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.
	 */
2679 2680 2681 2682
	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);
2683

L
Linus Torvalds 已提交
2684 2685 2686 2687
	rq->prev_mm = NULL;

	/*
	 * A task struct has one reference for the use as "current".
2688
	 * If a task dies, then it sets TASK_DEAD in tsk->state and calls
O
Oleg Nesterov 已提交
2689 2690
	 * schedule one last time. The schedule call will never return, and
	 * the scheduled task must drop that reference.
2691 2692
	 *
	 * We must observe prev->state before clearing prev->on_cpu (in
2693
	 * finish_task), otherwise a concurrent wakeup can get prev
2694 2695
	 * running on another CPU and we could rave with its RUNNING -> DEAD
	 * transition, resulting in a double drop.
L
Linus Torvalds 已提交
2696
	 */
O
Oleg Nesterov 已提交
2697
	prev_state = prev->state;
2698
	vtime_task_switch(prev);
2699
	perf_event_task_sched_in(prev, current);
2700 2701 2702 2703 2704 2705 2706 2707 2708 2709
	/*
	 * The membarrier system call requires a full memory barrier
	 * after storing to rq->curr, before going back to user-space.
	 *
	 * TODO: This smp_mb__after_unlock_lock can go away if PPC end
	 * up adding a full barrier to switch_mm(), or we should figure
	 * out if a smp_mb__after_unlock_lock is really the proper API
	 * to use.
	 */
	smp_mb__after_unlock_lock();
2710 2711
	finish_task(prev);
	finish_lock_switch(rq);
2712
	finish_arch_post_lock_switch();
S
Steven Rostedt 已提交
2713

2714
	fire_sched_in_preempt_notifiers(current);
L
Linus Torvalds 已提交
2715 2716
	if (mm)
		mmdrop(mm);
2717
	if (unlikely(prev_state == TASK_DEAD)) {
2718 2719 2720
		if (prev->sched_class->task_dead)
			prev->sched_class->task_dead(prev);

2721 2722 2723
		/*
		 * Remove function-return probe instances associated with this
		 * task and put them back on the free list.
I
Ingo Molnar 已提交
2724
		 */
2725
		kprobe_flush_task(prev);
2726 2727 2728 2729

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

L
Linus Torvalds 已提交
2730
		put_task_struct(prev);
2731
	}
2732

2733
	tick_nohz_task_switch();
2734
	return rq;
L
Linus Torvalds 已提交
2735 2736
}

2737 2738 2739
#ifdef CONFIG_SMP

/* rq->lock is NOT held, but preemption is disabled */
2740
static void __balance_callback(struct rq *rq)
2741
{
2742 2743 2744
	struct callback_head *head, *next;
	void (*func)(struct rq *rq);
	unsigned long flags;
2745

2746 2747 2748 2749 2750 2751 2752 2753
	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;
2754

2755
		func(rq);
2756
	}
2757 2758 2759 2760 2761 2762 2763
	raw_spin_unlock_irqrestore(&rq->lock, flags);
}

static inline void balance_callback(struct rq *rq)
{
	if (unlikely(rq->balance_callback))
		__balance_callback(rq);
2764 2765 2766
}

#else
2767

2768
static inline void balance_callback(struct rq *rq)
2769
{
L
Linus Torvalds 已提交
2770 2771
}

2772 2773
#endif

L
Linus Torvalds 已提交
2774 2775 2776 2777
/**
 * schedule_tail - first thing a freshly forked thread must call.
 * @prev: the thread we just switched away from.
 */
2778
asmlinkage __visible void schedule_tail(struct task_struct *prev)
L
Linus Torvalds 已提交
2779 2780
	__releases(rq->lock)
{
2781
	struct rq *rq;
2782

2783 2784 2785 2786 2787 2788 2789 2790 2791
	/*
	 * 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).
	 */

2792
	rq = finish_task_switch(prev);
2793
	balance_callback(rq);
2794
	preempt_enable();
2795

L
Linus Torvalds 已提交
2796
	if (current->set_child_tid)
2797
		put_user(task_pid_vnr(current), current->set_child_tid);
L
Linus Torvalds 已提交
2798 2799 2800
}

/*
2801
 * context_switch - switch to the new MM and the new thread's register state.
L
Linus Torvalds 已提交
2802
 */
2803
static __always_inline struct rq *
2804
context_switch(struct rq *rq, struct task_struct *prev,
2805
	       struct task_struct *next, struct rq_flags *rf)
L
Linus Torvalds 已提交
2806
{
I
Ingo Molnar 已提交
2807
	struct mm_struct *mm, *oldmm;
L
Linus Torvalds 已提交
2808

2809
	prepare_task_switch(rq, prev, next);
2810

I
Ingo Molnar 已提交
2811 2812
	mm = next->mm;
	oldmm = prev->active_mm;
2813 2814 2815 2816 2817
	/*
	 * For paravirt, this is coupled with an exit in switch_to to
	 * combine the page table reload and the switch backend into
	 * one hypercall.
	 */
2818
	arch_start_context_switch(prev);
2819

2820
	if (!mm) {
L
Linus Torvalds 已提交
2821
		next->active_mm = oldmm;
V
Vegard Nossum 已提交
2822
		mmgrab(oldmm);
L
Linus Torvalds 已提交
2823 2824
		enter_lazy_tlb(oldmm, next);
	} else
2825
		switch_mm_irqs_off(oldmm, mm, next);
L
Linus Torvalds 已提交
2826

2827
	if (!prev->mm) {
L
Linus Torvalds 已提交
2828 2829 2830
		prev->active_mm = NULL;
		rq->prev_mm = oldmm;
	}
2831

2832
	rq->clock_update_flags &= ~(RQCF_ACT_SKIP|RQCF_REQ_SKIP);
2833

2834 2835 2836 2837 2838 2839
	/*
	 * 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:
	 */
2840
	rq_unpin_lock(rq, rf);
2841
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
L
Linus Torvalds 已提交
2842 2843 2844

	/* Here we just switch the register state and the stack. */
	switch_to(prev, next, prev);
I
Ingo Molnar 已提交
2845
	barrier();
2846 2847

	return finish_task_switch(prev);
L
Linus Torvalds 已提交
2848 2849 2850
}

/*
2851
 * nr_running and nr_context_switches:
L
Linus Torvalds 已提交
2852 2853
 *
 * externally visible scheduler statistics: current number of runnable
2854
 * threads, total number of context switches performed since bootup.
L
Linus Torvalds 已提交
2855 2856 2857 2858 2859 2860 2861 2862 2863
 */
unsigned long nr_running(void)
{
	unsigned long i, sum = 0;

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

	return sum;
2864
}
L
Linus Torvalds 已提交
2865

2866
/*
I
Ingo Molnar 已提交
2867
 * Check if only the current task is running on the CPU.
2868 2869 2870 2871 2872 2873 2874 2875 2876 2877
 *
 * 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)
2878 2879 2880
 */
bool single_task_running(void)
{
2881
	return raw_rq()->nr_running == 1;
2882 2883 2884
}
EXPORT_SYMBOL(single_task_running);

L
Linus Torvalds 已提交
2885
unsigned long long nr_context_switches(void)
2886
{
2887 2888
	int i;
	unsigned long long sum = 0;
2889

2890
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2891
		sum += cpu_rq(i)->nr_switches;
2892

L
Linus Torvalds 已提交
2893 2894
	return sum;
}
2895

2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925
/*
 * 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
2926 2927 2928
unsigned long nr_iowait(void)
{
	unsigned long i, sum = 0;
2929

2930
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2931
		sum += atomic_read(&cpu_rq(i)->nr_iowait);
2932

L
Linus Torvalds 已提交
2933 2934
	return sum;
}
2935

2936 2937 2938 2939 2940 2941 2942
/*
 * 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.
 */

2943
unsigned long nr_iowait_cpu(int cpu)
2944
{
2945
	struct rq *this = cpu_rq(cpu);
2946 2947
	return atomic_read(&this->nr_iowait);
}
2948

2949 2950
void get_iowait_load(unsigned long *nr_waiters, unsigned long *load)
{
2951 2952 2953
	struct rq *rq = this_rq();
	*nr_waiters = atomic_read(&rq->nr_iowait);
	*load = rq->load.weight;
2954 2955
}

I
Ingo Molnar 已提交
2956
#ifdef CONFIG_SMP
2957

2958
/*
P
Peter Zijlstra 已提交
2959 2960
 * sched_exec - execve() is a valuable balancing opportunity, because at
 * this point the task has the smallest effective memory and cache footprint.
2961
 */
P
Peter Zijlstra 已提交
2962
void sched_exec(void)
2963
{
P
Peter Zijlstra 已提交
2964
	struct task_struct *p = current;
L
Linus Torvalds 已提交
2965
	unsigned long flags;
2966
	int dest_cpu;
2967

2968
	raw_spin_lock_irqsave(&p->pi_lock, flags);
2969
	dest_cpu = p->sched_class->select_task_rq(p, task_cpu(p), SD_BALANCE_EXEC, 0);
2970 2971
	if (dest_cpu == smp_processor_id())
		goto unlock;
P
Peter Zijlstra 已提交
2972

2973
	if (likely(cpu_active(dest_cpu))) {
2974
		struct migration_arg arg = { p, dest_cpu };
2975

2976 2977
		raw_spin_unlock_irqrestore(&p->pi_lock, flags);
		stop_one_cpu(task_cpu(p), migration_cpu_stop, &arg);
L
Linus Torvalds 已提交
2978 2979
		return;
	}
2980
unlock:
2981
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
2982
}
I
Ingo Molnar 已提交
2983

L
Linus Torvalds 已提交
2984 2985 2986
#endif

DEFINE_PER_CPU(struct kernel_stat, kstat);
2987
DEFINE_PER_CPU(struct kernel_cpustat, kernel_cpustat);
L
Linus Torvalds 已提交
2988 2989

EXPORT_PER_CPU_SYMBOL(kstat);
2990
EXPORT_PER_CPU_SYMBOL(kernel_cpustat);
L
Linus Torvalds 已提交
2991

2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008
/*
 * 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);
}

3009 3010 3011 3012 3013 3014 3015
/*
 * 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)
{
3016
	struct rq_flags rf;
3017
	struct rq *rq;
3018
	u64 ns;
3019

3020 3021 3022 3023 3024 3025
#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
3026 3027
	 * 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
3028
	 * indistinguishable from the read occurring a few cycles earlier.
3029 3030
	 * If we see ->on_cpu without ->on_rq, the task is leaving, and has
	 * been accounted, so we're correct here as well.
3031
	 */
3032
	if (!p->on_cpu || !task_on_rq_queued(p))
3033 3034 3035
		return p->se.sum_exec_runtime;
#endif

3036
	rq = task_rq_lock(p, &rf);
3037 3038 3039 3040 3041 3042
	/*
	 * 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)) {
3043
		prefetch_curr_exec_start(p);
3044 3045 3046 3047
		update_rq_clock(rq);
		p->sched_class->update_curr(rq);
	}
	ns = p->se.sum_exec_runtime;
3048
	task_rq_unlock(rq, p, &rf);
3049 3050 3051

	return ns;
}
3052

3053 3054 3055 3056 3057 3058 3059 3060
/*
 * 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
3061
	struct task_struct *curr = rq->curr;
3062
	struct rq_flags rf;
3063 3064

	sched_clock_tick();
I
Ingo Molnar 已提交
3065

3066 3067
	rq_lock(rq, &rf);

3068
	update_rq_clock(rq);
P
Peter Zijlstra 已提交
3069
	curr->sched_class->task_tick(rq, curr, 0);
3070
	cpu_load_update_active(rq);
3071
	calc_global_load_tick(rq);
3072 3073

	rq_unlock(rq, &rf);
3074

3075
	perf_event_task_tick();
3076

3077
#ifdef CONFIG_SMP
3078
	rq->idle_balance = idle_cpu(cpu);
3079
	trigger_load_balance(rq);
3080
#endif
3081
	rq_last_tick_reset(rq);
L
Linus Torvalds 已提交
3082 3083
}

3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094
#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.
3095 3096
 *
 * Return: Maximum deferment in nanoseconds.
3097 3098 3099 3100
 */
u64 scheduler_tick_max_deferment(void)
{
	struct rq *rq = this_rq();
3101
	unsigned long next, now = READ_ONCE(jiffies);
3102 3103 3104 3105 3106 3107

	next = rq->last_sched_tick + HZ;

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

3108
	return jiffies_to_nsecs(next - now);
L
Linus Torvalds 已提交
3109
}
3110
#endif
L
Linus Torvalds 已提交
3111

3112 3113
#if defined(CONFIG_PREEMPT) && (defined(CONFIG_DEBUG_PREEMPT) || \
				defined(CONFIG_PREEMPT_TRACER))
3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127
/*
 * 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);
	}
}
3128

3129
void preempt_count_add(int val)
L
Linus Torvalds 已提交
3130
{
3131
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3132 3133 3134
	/*
	 * Underflow?
	 */
3135 3136
	if (DEBUG_LOCKS_WARN_ON((preempt_count() < 0)))
		return;
3137
#endif
3138
	__preempt_count_add(val);
3139
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3140 3141 3142
	/*
	 * Spinlock count overflowing soon?
	 */
3143 3144
	DEBUG_LOCKS_WARN_ON((preempt_count() & PREEMPT_MASK) >=
				PREEMPT_MASK - 10);
3145
#endif
3146
	preempt_latency_start(val);
L
Linus Torvalds 已提交
3147
}
3148
EXPORT_SYMBOL(preempt_count_add);
3149
NOKPROBE_SYMBOL(preempt_count_add);
L
Linus Torvalds 已提交
3150

3151 3152 3153 3154 3155 3156 3157 3158 3159 3160
/*
 * 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());
}

3161
void preempt_count_sub(int val)
L
Linus Torvalds 已提交
3162
{
3163
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3164 3165 3166
	/*
	 * Underflow?
	 */
3167
	if (DEBUG_LOCKS_WARN_ON(val > preempt_count()))
3168
		return;
L
Linus Torvalds 已提交
3169 3170 3171
	/*
	 * Is the spinlock portion underflowing?
	 */
3172 3173 3174
	if (DEBUG_LOCKS_WARN_ON((val < PREEMPT_MASK) &&
			!(preempt_count() & PREEMPT_MASK)))
		return;
3175
#endif
3176

3177
	preempt_latency_stop(val);
3178
	__preempt_count_sub(val);
L
Linus Torvalds 已提交
3179
}
3180
EXPORT_SYMBOL(preempt_count_sub);
3181
NOKPROBE_SYMBOL(preempt_count_sub);
L
Linus Torvalds 已提交
3182

3183 3184 3185
#else
static inline void preempt_latency_start(int val) { }
static inline void preempt_latency_stop(int val) { }
L
Linus Torvalds 已提交
3186 3187
#endif

3188 3189 3190 3191 3192 3193 3194 3195 3196
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
3197
/*
I
Ingo Molnar 已提交
3198
 * Print scheduling while atomic bug:
L
Linus Torvalds 已提交
3199
 */
I
Ingo Molnar 已提交
3200
static noinline void __schedule_bug(struct task_struct *prev)
L
Linus Torvalds 已提交
3201
{
3202 3203 3204
	/* Save this before calling printk(), since that will clobber it */
	unsigned long preempt_disable_ip = get_preempt_disable_ip(current);

3205 3206 3207
	if (oops_in_progress)
		return;

P
Peter Zijlstra 已提交
3208 3209
	printk(KERN_ERR "BUG: scheduling while atomic: %s/%d/0x%08x\n",
		prev->comm, prev->pid, preempt_count());
3210

I
Ingo Molnar 已提交
3211
	debug_show_held_locks(prev);
3212
	print_modules();
I
Ingo Molnar 已提交
3213 3214
	if (irqs_disabled())
		print_irqtrace_events(prev);
3215 3216
	if (IS_ENABLED(CONFIG_DEBUG_PREEMPT)
	    && in_atomic_preempt_off()) {
3217
		pr_err("Preemption disabled at:");
3218
		print_ip_sym(preempt_disable_ip);
3219 3220
		pr_cont("\n");
	}
3221 3222 3223
	if (panic_on_warn)
		panic("scheduling while atomic\n");

3224
	dump_stack();
3225
	add_taint(TAINT_WARN, LOCKDEP_STILL_OK);
I
Ingo Molnar 已提交
3226
}
L
Linus Torvalds 已提交
3227

I
Ingo Molnar 已提交
3228 3229 3230 3231 3232
/*
 * Various schedule()-time debugging checks and statistics:
 */
static inline void schedule_debug(struct task_struct *prev)
{
3233
#ifdef CONFIG_SCHED_STACK_END_CHECK
J
Jann Horn 已提交
3234 3235
	if (task_stack_end_corrupted(prev))
		panic("corrupted stack end detected inside scheduler\n");
3236
#endif
3237

3238
	if (unlikely(in_atomic_preempt_off())) {
I
Ingo Molnar 已提交
3239
		__schedule_bug(prev);
3240 3241
		preempt_count_set(PREEMPT_DISABLED);
	}
3242
	rcu_sleep_check();
I
Ingo Molnar 已提交
3243

L
Linus Torvalds 已提交
3244 3245
	profile_hit(SCHED_PROFILING, __builtin_return_address(0));

3246
	schedstat_inc(this_rq()->sched_count);
I
Ingo Molnar 已提交
3247 3248 3249 3250 3251 3252
}

/*
 * Pick up the highest-prio task:
 */
static inline struct task_struct *
3253
pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
I
Ingo Molnar 已提交
3254
{
3255
	const struct sched_class *class;
I
Ingo Molnar 已提交
3256
	struct task_struct *p;
L
Linus Torvalds 已提交
3257 3258

	/*
3259 3260 3261 3262
	 * 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
3263
	 */
3264 3265 3266 3267
	if (likely((prev->sched_class == &idle_sched_class ||
		    prev->sched_class == &fair_sched_class) &&
		   rq->nr_running == rq->cfs.h_nr_running)) {

3268
		p = fair_sched_class.pick_next_task(rq, prev, rf);
3269 3270 3271
		if (unlikely(p == RETRY_TASK))
			goto again;

I
Ingo Molnar 已提交
3272
		/* Assumes fair_sched_class->next == idle_sched_class */
3273
		if (unlikely(!p))
3274
			p = idle_sched_class.pick_next_task(rq, prev, rf);
3275 3276

		return p;
L
Linus Torvalds 已提交
3277 3278
	}

3279
again:
3280
	for_each_class(class) {
3281
		p = class->pick_next_task(rq, prev, rf);
3282 3283 3284
		if (p) {
			if (unlikely(p == RETRY_TASK))
				goto again;
I
Ingo Molnar 已提交
3285
			return p;
3286
		}
I
Ingo Molnar 已提交
3287
	}
3288

I
Ingo Molnar 已提交
3289 3290
	/* The idle class should always have a runnable task: */
	BUG();
I
Ingo Molnar 已提交
3291
}
L
Linus Torvalds 已提交
3292

I
Ingo Molnar 已提交
3293
/*
3294
 * __schedule() is the main scheduler function.
3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328
 *
 * 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
3329
 *
3330
 * WARNING: must be called with preemption disabled!
I
Ingo Molnar 已提交
3331
 */
3332
static void __sched notrace __schedule(bool preempt)
I
Ingo Molnar 已提交
3333 3334
{
	struct task_struct *prev, *next;
3335
	unsigned long *switch_count;
3336
	struct rq_flags rf;
I
Ingo Molnar 已提交
3337
	struct rq *rq;
3338
	int cpu;
I
Ingo Molnar 已提交
3339 3340 3341 3342 3343 3344

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

	schedule_debug(prev);
L
Linus Torvalds 已提交
3345

3346
	if (sched_feat(HRTICK))
M
Mike Galbraith 已提交
3347
		hrtick_clear(rq);
P
Peter Zijlstra 已提交
3348

3349
	local_irq_disable();
3350
	rcu_note_context_switch(preempt);
3351

3352 3353 3354 3355 3356
	/*
	 * 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().
	 */
3357
	rq_lock(rq, &rf);
3358
	smp_mb__after_spinlock();
L
Linus Torvalds 已提交
3359

I
Ingo Molnar 已提交
3360 3361
	/* Promote REQ to ACT */
	rq->clock_update_flags <<= 1;
3362
	update_rq_clock(rq);
3363

3364
	switch_count = &prev->nivcsw;
3365
	if (!preempt && prev->state) {
T
Tejun Heo 已提交
3366
		if (unlikely(signal_pending_state(prev->state, prev))) {
L
Linus Torvalds 已提交
3367
			prev->state = TASK_RUNNING;
T
Tejun Heo 已提交
3368
		} else {
3369
			deactivate_task(rq, prev, DEQUEUE_SLEEP | DEQUEUE_NOCLOCK);
3370 3371
			prev->on_rq = 0;

3372 3373 3374 3375 3376
			if (prev->in_iowait) {
				atomic_inc(&rq->nr_iowait);
				delayacct_blkio_start();
			}

T
Tejun Heo 已提交
3377
			/*
3378 3379 3380
			 * 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
3381 3382 3383 3384
			 */
			if (prev->flags & PF_WQ_WORKER) {
				struct task_struct *to_wakeup;

3385
				to_wakeup = wq_worker_sleeping(prev);
T
Tejun Heo 已提交
3386
				if (to_wakeup)
3387
					try_to_wake_up_local(to_wakeup, &rf);
T
Tejun Heo 已提交
3388 3389
			}
		}
I
Ingo Molnar 已提交
3390
		switch_count = &prev->nvcsw;
L
Linus Torvalds 已提交
3391 3392
	}

3393
	next = pick_next_task(rq, prev, &rf);
3394
	clear_tsk_need_resched(prev);
3395
	clear_preempt_need_resched();
L
Linus Torvalds 已提交
3396 3397 3398 3399

	if (likely(prev != next)) {
		rq->nr_switches++;
		rq->curr = next;
3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414
		/*
		 * The membarrier system call requires each architecture
		 * to have a full memory barrier after updating
		 * rq->curr, before returning to user-space. For TSO
		 * (e.g. x86), the architecture must provide its own
		 * barrier in switch_mm(). For weakly ordered machines
		 * for which spin_unlock() acts as a full memory
		 * barrier, finish_lock_switch() in common code takes
		 * care of this barrier. For weakly ordered machines for
		 * which spin_unlock() acts as a RELEASE barrier (only
		 * arm64 and PowerPC), arm64 has a full barrier in
		 * switch_to(), and PowerPC has
		 * smp_mb__after_unlock_lock() before
		 * finish_lock_switch().
		 */
L
Linus Torvalds 已提交
3415 3416
		++*switch_count;

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

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

3426
	balance_callback(rq);
L
Linus Torvalds 已提交
3427
}
3428

3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442
void __noreturn do_task_dead(void)
{
	/*
	 * The setting of TASK_RUNNING by try_to_wake_up() may be delayed
	 * when the following two conditions become true.
	 *   - There is race condition of mmap_sem (It is acquired by
	 *     exit_mm()), and
	 *   - SMI occurs before setting TASK_RUNINNG.
	 *     (or hypervisor of virtual machine switches to other guest)
	 *  As a result, we may become TASK_RUNNING after becoming TASK_DEAD
	 *
	 * To avoid it, we have to wait for releasing tsk->pi_lock which
	 * is held by try_to_wake_up()
	 */
3443 3444
	raw_spin_lock_irq(&current->pi_lock);
	raw_spin_unlock_irq(&current->pi_lock);
3445

I
Ingo Molnar 已提交
3446
	/* Causes final put_task_struct in finish_task_switch(): */
3447
	__set_current_state(TASK_DEAD);
I
Ingo Molnar 已提交
3448 3449 3450 3451

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

3452 3453
	__schedule(false);
	BUG();
I
Ingo Molnar 已提交
3454 3455

	/* Avoid "noreturn function does return" - but don't continue if BUG() is a NOP: */
3456
	for (;;)
I
Ingo Molnar 已提交
3457
		cpu_relax();
3458 3459
}

3460 3461
static inline void sched_submit_work(struct task_struct *tsk)
{
3462
	if (!tsk->state || tsk_is_pi_blocked(tsk))
3463 3464 3465 3466 3467 3468 3469 3470 3471
		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);
}

3472
asmlinkage __visible void __sched schedule(void)
3473
{
3474 3475 3476
	struct task_struct *tsk = current;

	sched_submit_work(tsk);
3477
	do {
3478
		preempt_disable();
3479
		__schedule(false);
3480
		sched_preempt_enable_no_resched();
3481
	} while (need_resched());
3482
}
L
Linus Torvalds 已提交
3483 3484
EXPORT_SYMBOL(schedule);

3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509
/*
 * 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());
}

3510
#ifdef CONFIG_CONTEXT_TRACKING
3511
asmlinkage __visible void __sched schedule_user(void)
3512 3513 3514 3515 3516 3517
{
	/*
	 * 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.
3518 3519
	 *
	 * NB: There are buggy callers of this function.  Ideally we
3520
	 * should warn if prev_state != CONTEXT_USER, but that will trigger
3521
	 * too frequently to make sense yet.
3522
	 */
3523
	enum ctx_state prev_state = exception_enter();
3524
	schedule();
3525
	exception_exit(prev_state);
3526 3527 3528
}
#endif

3529 3530 3531 3532 3533 3534 3535
/**
 * schedule_preempt_disabled - called with preemption disabled
 *
 * Returns with preemption disabled. Note: preempt_count must be 1
 */
void __sched schedule_preempt_disabled(void)
{
3536
	sched_preempt_enable_no_resched();
3537 3538 3539 3540
	schedule();
	preempt_disable();
}

3541
static void __sched notrace preempt_schedule_common(void)
3542 3543
{
	do {
3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556
		/*
		 * 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.
		 */
3557
		preempt_disable_notrace();
3558
		preempt_latency_start(1);
3559
		__schedule(true);
3560
		preempt_latency_stop(1);
3561
		preempt_enable_no_resched_notrace();
3562 3563 3564 3565 3566 3567 3568 3569

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

L
Linus Torvalds 已提交
3570 3571
#ifdef CONFIG_PREEMPT
/*
3572
 * this is the entry point to schedule() from in-kernel preemption
I
Ingo Molnar 已提交
3573
 * off of preempt_enable. Kernel preemptions off return from interrupt
L
Linus Torvalds 已提交
3574 3575
 * occur there and call schedule directly.
 */
3576
asmlinkage __visible void __sched notrace preempt_schedule(void)
L
Linus Torvalds 已提交
3577 3578 3579
{
	/*
	 * If there is a non-zero preempt_count or interrupts are disabled,
I
Ingo Molnar 已提交
3580
	 * we do not want to preempt the current task. Just return..
L
Linus Torvalds 已提交
3581
	 */
3582
	if (likely(!preemptible()))
L
Linus Torvalds 已提交
3583 3584
		return;

3585
	preempt_schedule_common();
L
Linus Torvalds 已提交
3586
}
3587
NOKPROBE_SYMBOL(preempt_schedule);
L
Linus Torvalds 已提交
3588
EXPORT_SYMBOL(preempt_schedule);
3589 3590

/**
3591
 * preempt_schedule_notrace - preempt_schedule called by tracing
3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603
 *
 * 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.
 */
3604
asmlinkage __visible void __sched notrace preempt_schedule_notrace(void)
3605 3606 3607 3608 3609 3610 3611
{
	enum ctx_state prev_ctx;

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

	do {
3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624
		/*
		 * 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.
		 */
3625
		preempt_disable_notrace();
3626
		preempt_latency_start(1);
3627 3628 3629 3630 3631 3632
		/*
		 * Needs preempt disabled in case user_exit() is traced
		 * and the tracer calls preempt_enable_notrace() causing
		 * an infinite recursion.
		 */
		prev_ctx = exception_enter();
3633
		__schedule(true);
3634 3635
		exception_exit(prev_ctx);

3636
		preempt_latency_stop(1);
3637
		preempt_enable_no_resched_notrace();
3638 3639
	} while (need_resched());
}
3640
EXPORT_SYMBOL_GPL(preempt_schedule_notrace);
3641

3642
#endif /* CONFIG_PREEMPT */
L
Linus Torvalds 已提交
3643 3644

/*
3645
 * this is the entry point to schedule() from kernel preemption
L
Linus Torvalds 已提交
3646 3647 3648 3649
 * off of irq context.
 * Note, that this is called and return with irqs disabled. This will
 * protect us against recursive calling from irq.
 */
3650
asmlinkage __visible void __sched preempt_schedule_irq(void)
L
Linus Torvalds 已提交
3651
{
3652
	enum ctx_state prev_state;
3653

3654
	/* Catch callers which need to be fixed */
3655
	BUG_ON(preempt_count() || !irqs_disabled());
L
Linus Torvalds 已提交
3656

3657 3658
	prev_state = exception_enter();

3659
	do {
3660
		preempt_disable();
3661
		local_irq_enable();
3662
		__schedule(true);
3663
		local_irq_disable();
3664
		sched_preempt_enable_no_resched();
3665
	} while (need_resched());
3666 3667

	exception_exit(prev_state);
L
Linus Torvalds 已提交
3668 3669
}

3670
int default_wake_function(wait_queue_entry_t *curr, unsigned mode, int wake_flags,
I
Ingo Molnar 已提交
3671
			  void *key)
L
Linus Torvalds 已提交
3672
{
P
Peter Zijlstra 已提交
3673
	return try_to_wake_up(curr->private, mode, wake_flags);
L
Linus Torvalds 已提交
3674 3675 3676
}
EXPORT_SYMBOL(default_wake_function);

3677 3678
#ifdef CONFIG_RT_MUTEXES

3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693
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);
}

3694 3695
/*
 * rt_mutex_setprio - set the current priority of a task
3696 3697
 * @p: task to boost
 * @pi_task: donor task
3698 3699 3700 3701
 *
 * This function changes the 'effective' priority of a task. It does
 * not touch ->normal_prio like __setscheduler().
 *
3702 3703
 * Used by the rt_mutex code to implement priority inheritance
 * logic. Call site only calls if the priority of the task changed.
3704
 */
3705
void rt_mutex_setprio(struct task_struct *p, struct task_struct *pi_task)
3706
{
3707
	int prio, oldprio, queued, running, queue_flag =
3708
		DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
3709
	const struct sched_class *prev_class;
3710 3711
	struct rq_flags rf;
	struct rq *rq;
3712

3713 3714 3715 3716 3717 3718 3719 3720
	/* 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;
3721

3722
	rq = __task_rq_lock(p, &rf);
3723
	update_rq_clock(rq);
3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740
	/*
	 * 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;
3741

3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759
	/*
	 * 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;
	}

3760
	trace_sched_pi_setprio(p, pi_task);
3761
	oldprio = p->prio;
3762 3763 3764 3765

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

3766
	prev_class = p->sched_class;
3767
	queued = task_on_rq_queued(p);
3768
	running = task_current(rq, p);
3769
	if (queued)
3770
		dequeue_task(rq, p, queue_flag);
3771
	if (running)
3772
		put_prev_task(rq, p);
I
Ingo Molnar 已提交
3773

3774 3775 3776 3777 3778 3779 3780 3781 3782 3783
	/*
	 * 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)) {
3784 3785
		if (!dl_prio(p->normal_prio) ||
		    (pi_task && dl_entity_preempt(&pi_task->dl, &p->dl))) {
3786
			p->dl.dl_boosted = 1;
3787
			queue_flag |= ENQUEUE_REPLENISH;
3788 3789
		} else
			p->dl.dl_boosted = 0;
3790
		p->sched_class = &dl_sched_class;
3791 3792 3793 3794
	} else if (rt_prio(prio)) {
		if (dl_prio(oldprio))
			p->dl.dl_boosted = 0;
		if (oldprio < prio)
3795
			queue_flag |= ENQUEUE_HEAD;
I
Ingo Molnar 已提交
3796
		p->sched_class = &rt_sched_class;
3797 3798 3799
	} else {
		if (dl_prio(oldprio))
			p->dl.dl_boosted = 0;
3800 3801
		if (rt_prio(oldprio))
			p->rt.timeout = 0;
I
Ingo Molnar 已提交
3802
		p->sched_class = &fair_sched_class;
3803
	}
I
Ingo Molnar 已提交
3804

3805 3806
	p->prio = prio;

3807
	if (queued)
3808
		enqueue_task(rq, p, queue_flag);
3809
	if (running)
3810
		set_curr_task(rq, p);
3811

P
Peter Zijlstra 已提交
3812
	check_class_changed(rq, p, prev_class, oldprio);
3813
out_unlock:
I
Ingo Molnar 已提交
3814 3815
	/* Avoid rq from going away on us: */
	preempt_disable();
3816
	__task_rq_unlock(rq, &rf);
3817 3818 3819

	balance_callback(rq);
	preempt_enable();
3820
}
3821 3822 3823 3824 3825
#else
static inline int rt_effective_prio(struct task_struct *p, int prio)
{
	return prio;
}
3826
#endif
3827

3828
void set_user_nice(struct task_struct *p, long nice)
L
Linus Torvalds 已提交
3829
{
P
Peter Zijlstra 已提交
3830 3831
	bool queued, running;
	int old_prio, delta;
3832
	struct rq_flags rf;
3833
	struct rq *rq;
L
Linus Torvalds 已提交
3834

3835
	if (task_nice(p) == nice || nice < MIN_NICE || nice > MAX_NICE)
L
Linus Torvalds 已提交
3836 3837 3838 3839 3840
		return;
	/*
	 * We have to be careful, if called from sys_setpriority(),
	 * the task might be in the middle of scheduling on another CPU.
	 */
3841
	rq = task_rq_lock(p, &rf);
3842 3843
	update_rq_clock(rq);

L
Linus Torvalds 已提交
3844 3845 3846 3847
	/*
	 * 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
3848
	 * SCHED_DEADLINE, SCHED_FIFO or SCHED_RR:
L
Linus Torvalds 已提交
3849
	 */
3850
	if (task_has_dl_policy(p) || task_has_rt_policy(p)) {
L
Linus Torvalds 已提交
3851 3852 3853
		p->static_prio = NICE_TO_PRIO(nice);
		goto out_unlock;
	}
3854
	queued = task_on_rq_queued(p);
P
Peter Zijlstra 已提交
3855
	running = task_current(rq, p);
3856
	if (queued)
3857
		dequeue_task(rq, p, DEQUEUE_SAVE | DEQUEUE_NOCLOCK);
P
Peter Zijlstra 已提交
3858 3859
	if (running)
		put_prev_task(rq, p);
L
Linus Torvalds 已提交
3860 3861

	p->static_prio = NICE_TO_PRIO(nice);
3862
	set_load_weight(p, true);
3863 3864 3865
	old_prio = p->prio;
	p->prio = effective_prio(p);
	delta = p->prio - old_prio;
L
Linus Torvalds 已提交
3866

3867
	if (queued) {
3868
		enqueue_task(rq, p, ENQUEUE_RESTORE | ENQUEUE_NOCLOCK);
L
Linus Torvalds 已提交
3869
		/*
3870 3871
		 * If the task increased its priority or is running and
		 * lowered its priority, then reschedule its CPU:
L
Linus Torvalds 已提交
3872
		 */
3873
		if (delta < 0 || (delta > 0 && task_running(rq, p)))
3874
			resched_curr(rq);
L
Linus Torvalds 已提交
3875
	}
P
Peter Zijlstra 已提交
3876 3877
	if (running)
		set_curr_task(rq, p);
L
Linus Torvalds 已提交
3878
out_unlock:
3879
	task_rq_unlock(rq, p, &rf);
L
Linus Torvalds 已提交
3880 3881 3882
}
EXPORT_SYMBOL(set_user_nice);

M
Matt Mackall 已提交
3883 3884 3885 3886 3887
/*
 * can_nice - check if a task can reduce its nice value
 * @p: task
 * @nice: nice value
 */
3888
int can_nice(const struct task_struct *p, const int nice)
M
Matt Mackall 已提交
3889
{
I
Ingo Molnar 已提交
3890
	/* Convert nice value [19,-20] to rlimit style value [1,40]: */
3891
	int nice_rlim = nice_to_rlimit(nice);
3892

3893
	return (nice_rlim <= task_rlimit(p, RLIMIT_NICE) ||
M
Matt Mackall 已提交
3894 3895 3896
		capable(CAP_SYS_NICE));
}

L
Linus Torvalds 已提交
3897 3898 3899 3900 3901 3902 3903 3904 3905
#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.
 */
3906
SYSCALL_DEFINE1(nice, int, increment)
L
Linus Torvalds 已提交
3907
{
3908
	long nice, retval;
L
Linus Torvalds 已提交
3909 3910 3911 3912 3913 3914

	/*
	 * 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.
	 */
3915
	increment = clamp(increment, -NICE_WIDTH, NICE_WIDTH);
3916
	nice = task_nice(current) + increment;
L
Linus Torvalds 已提交
3917

3918
	nice = clamp_val(nice, MIN_NICE, MAX_NICE);
M
Matt Mackall 已提交
3919 3920 3921
	if (increment < 0 && !can_nice(current, nice))
		return -EPERM;

L
Linus Torvalds 已提交
3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935
	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.
 *
3936
 * Return: The priority value as seen by users in /proc.
L
Linus Torvalds 已提交
3937 3938 3939
 * RT tasks are offset by -200. Normal tasks are centered
 * around 0, value goes from -16 to +15.
 */
3940
int task_prio(const struct task_struct *p)
L
Linus Torvalds 已提交
3941 3942 3943 3944 3945
{
	return p->prio - MAX_RT_PRIO;
}

/**
I
Ingo Molnar 已提交
3946
 * idle_cpu - is a given CPU idle currently?
L
Linus Torvalds 已提交
3947
 * @cpu: the processor in question.
3948 3949
 *
 * Return: 1 if the CPU is currently idle. 0 otherwise.
L
Linus Torvalds 已提交
3950 3951 3952
 */
int idle_cpu(int cpu)
{
T
Thomas Gleixner 已提交
3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966
	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
3967 3968 3969
}

/**
I
Ingo Molnar 已提交
3970
 * idle_task - return the idle task for a given CPU.
L
Linus Torvalds 已提交
3971
 * @cpu: the processor in question.
3972
 *
I
Ingo Molnar 已提交
3973
 * Return: The idle task for the CPU @cpu.
L
Linus Torvalds 已提交
3974
 */
3975
struct task_struct *idle_task(int cpu)
L
Linus Torvalds 已提交
3976 3977 3978 3979 3980 3981 3982
{
	return cpu_rq(cpu)->idle;
}

/**
 * find_process_by_pid - find a process with a matching PID value.
 * @pid: the pid in question.
3983 3984
 *
 * The task of @pid, if found. %NULL otherwise.
L
Linus Torvalds 已提交
3985
 */
A
Alexey Dobriyan 已提交
3986
static struct task_struct *find_process_by_pid(pid_t pid)
L
Linus Torvalds 已提交
3987
{
3988
	return pid ? find_task_by_vpid(pid) : current;
L
Linus Torvalds 已提交
3989 3990
}

3991 3992 3993 3994 3995 3996
/*
 * sched_setparam() passes in -1 for its policy, to let the functions
 * it calls know not to change it.
 */
#define SETPARAM_POLICY	-1

3997 3998
static void __setscheduler_params(struct task_struct *p,
		const struct sched_attr *attr)
L
Linus Torvalds 已提交
3999
{
4000 4001
	int policy = attr->sched_policy;

4002
	if (policy == SETPARAM_POLICY)
4003 4004
		policy = p->policy;

L
Linus Torvalds 已提交
4005
	p->policy = policy;
4006

4007 4008
	if (dl_policy(policy))
		__setparam_dl(p, attr);
4009
	else if (fair_policy(policy))
4010 4011
		p->static_prio = NICE_TO_PRIO(attr->sched_nice);

4012 4013 4014 4015 4016 4017
	/*
	 * __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;
4018
	p->normal_prio = normal_prio(p);
4019
	set_load_weight(p, true);
4020
}
4021

4022 4023
/* Actually do priority change: must hold pi & rq lock. */
static void __setscheduler(struct rq *rq, struct task_struct *p,
4024
			   const struct sched_attr *attr, bool keep_boost)
4025 4026
{
	__setscheduler_params(p, attr);
4027

4028
	/*
4029 4030
	 * Keep a potential priority boosting if called from
	 * sched_setscheduler().
4031
	 */
4032
	p->prio = normal_prio(p);
4033
	if (keep_boost)
4034
		p->prio = rt_effective_prio(p, p->prio);
4035

4036 4037 4038
	if (dl_prio(p->prio))
		p->sched_class = &dl_sched_class;
	else if (rt_prio(p->prio))
4039 4040 4041
		p->sched_class = &rt_sched_class;
	else
		p->sched_class = &fair_sched_class;
L
Linus Torvalds 已提交
4042
}
4043

4044
/*
I
Ingo Molnar 已提交
4045
 * Check the target process has a UID that matches the current process's:
4046 4047 4048 4049 4050 4051 4052 4053
 */
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);
4054 4055
	match = (uid_eq(cred->euid, pcred->euid) ||
		 uid_eq(cred->euid, pcred->uid));
4056 4057 4058 4059
	rcu_read_unlock();
	return match;
}

4060 4061
static int __sched_setscheduler(struct task_struct *p,
				const struct sched_attr *attr,
4062
				bool user, bool pi)
L
Linus Torvalds 已提交
4063
{
4064 4065
	int newprio = dl_policy(attr->sched_policy) ? MAX_DL_PRIO - 1 :
		      MAX_RT_PRIO - 1 - attr->sched_priority;
4066
	int retval, oldprio, oldpolicy = -1, queued, running;
4067
	int new_effective_prio, policy = attr->sched_policy;
4068
	const struct sched_class *prev_class;
4069
	struct rq_flags rf;
4070
	int reset_on_fork;
4071
	int queue_flags = DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
4072
	struct rq *rq;
L
Linus Torvalds 已提交
4073

4074 4075
	/* The pi code expects interrupts enabled */
	BUG_ON(pi && in_interrupt());
L
Linus Torvalds 已提交
4076
recheck:
I
Ingo Molnar 已提交
4077
	/* Double check policy once rq lock held: */
4078 4079
	if (policy < 0) {
		reset_on_fork = p->sched_reset_on_fork;
L
Linus Torvalds 已提交
4080
		policy = oldpolicy = p->policy;
4081
	} else {
4082
		reset_on_fork = !!(attr->sched_flags & SCHED_FLAG_RESET_ON_FORK);
4083

4084
		if (!valid_policy(policy))
4085 4086 4087
			return -EINVAL;
	}

4088 4089
	if (attr->sched_flags &
		~(SCHED_FLAG_RESET_ON_FORK | SCHED_FLAG_RECLAIM))
4090 4091
		return -EINVAL;

L
Linus Torvalds 已提交
4092 4093
	/*
	 * Valid priorities for SCHED_FIFO and SCHED_RR are
I
Ingo Molnar 已提交
4094 4095
	 * 1..MAX_USER_RT_PRIO-1, valid priority for SCHED_NORMAL,
	 * SCHED_BATCH and SCHED_IDLE is 0.
L
Linus Torvalds 已提交
4096
	 */
4097
	if ((p->mm && attr->sched_priority > MAX_USER_RT_PRIO-1) ||
4098
	    (!p->mm && attr->sched_priority > MAX_RT_PRIO-1))
L
Linus Torvalds 已提交
4099
		return -EINVAL;
4100 4101
	if ((dl_policy(policy) && !__checkparam_dl(attr)) ||
	    (rt_policy(policy) != (attr->sched_priority != 0)))
L
Linus Torvalds 已提交
4102 4103
		return -EINVAL;

4104 4105 4106
	/*
	 * Allow unprivileged RT tasks to decrease priority:
	 */
4107
	if (user && !capable(CAP_SYS_NICE)) {
4108
		if (fair_policy(policy)) {
4109
			if (attr->sched_nice < task_nice(p) &&
4110
			    !can_nice(p, attr->sched_nice))
4111 4112 4113
				return -EPERM;
		}

4114
		if (rt_policy(policy)) {
4115 4116
			unsigned long rlim_rtprio =
					task_rlimit(p, RLIMIT_RTPRIO);
4117

I
Ingo Molnar 已提交
4118
			/* Can't set/change the rt policy: */
4119 4120 4121
			if (policy != p->policy && !rlim_rtprio)
				return -EPERM;

I
Ingo Molnar 已提交
4122
			/* Can't increase priority: */
4123 4124
			if (attr->sched_priority > p->rt_priority &&
			    attr->sched_priority > rlim_rtprio)
4125 4126
				return -EPERM;
		}
4127

4128 4129 4130 4131 4132 4133 4134 4135 4136
		 /*
		  * 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
4137
		/*
4138 4139
		 * 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
4140
		 */
4141
		if (idle_policy(p->policy) && !idle_policy(policy)) {
4142
			if (!can_nice(p, task_nice(p)))
4143 4144
				return -EPERM;
		}
4145

I
Ingo Molnar 已提交
4146
		/* Can't change other user's priorities: */
4147
		if (!check_same_owner(p))
4148
			return -EPERM;
4149

I
Ingo Molnar 已提交
4150
		/* Normal users shall not reset the sched_reset_on_fork flag: */
4151 4152
		if (p->sched_reset_on_fork && !reset_on_fork)
			return -EPERM;
4153
	}
L
Linus Torvalds 已提交
4154

4155
	if (user) {
4156
		retval = security_task_setscheduler(p);
4157 4158 4159 4160
		if (retval)
			return retval;
	}

4161
	/*
I
Ingo Molnar 已提交
4162
	 * Make sure no PI-waiters arrive (or leave) while we are
4163
	 * changing the priority of the task:
4164
	 *
L
Lucas De Marchi 已提交
4165
	 * To be able to change p->policy safely, the appropriate
L
Linus Torvalds 已提交
4166 4167
	 * runqueue lock must be held.
	 */
4168
	rq = task_rq_lock(p, &rf);
4169
	update_rq_clock(rq);
4170

4171
	/*
I
Ingo Molnar 已提交
4172
	 * Changing the policy of the stop threads its a very bad idea:
4173 4174
	 */
	if (p == rq->stop) {
4175
		task_rq_unlock(rq, p, &rf);
4176 4177 4178
		return -EINVAL;
	}

4179
	/*
4180 4181
	 * If not changing anything there's no need to proceed further,
	 * but store a possible modification of reset_on_fork.
4182
	 */
4183
	if (unlikely(policy == p->policy)) {
4184
		if (fair_policy(policy) && attr->sched_nice != task_nice(p))
4185 4186 4187
			goto change;
		if (rt_policy(policy) && attr->sched_priority != p->rt_priority)
			goto change;
4188
		if (dl_policy(policy) && dl_param_changed(p, attr))
4189
			goto change;
4190

4191
		p->sched_reset_on_fork = reset_on_fork;
4192
		task_rq_unlock(rq, p, &rf);
4193 4194
		return 0;
	}
4195
change:
4196

4197
	if (user) {
4198
#ifdef CONFIG_RT_GROUP_SCHED
4199 4200 4201 4202 4203
		/*
		 * Do not allow realtime tasks into groups that have no runtime
		 * assigned.
		 */
		if (rt_bandwidth_enabled() && rt_policy(policy) &&
4204 4205
				task_group(p)->rt_bandwidth.rt_runtime == 0 &&
				!task_group_is_autogroup(task_group(p))) {
4206
			task_rq_unlock(rq, p, &rf);
4207 4208 4209
			return -EPERM;
		}
#endif
4210 4211 4212 4213 4214 4215 4216 4217 4218
#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.
			 */
4219 4220
			if (!cpumask_subset(span, &p->cpus_allowed) ||
			    rq->rd->dl_bw.bw == 0) {
4221
				task_rq_unlock(rq, p, &rf);
4222 4223 4224 4225 4226
				return -EPERM;
			}
		}
#endif
	}
4227

I
Ingo Molnar 已提交
4228
	/* Re-check policy now with rq lock held: */
L
Linus Torvalds 已提交
4229 4230
	if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) {
		policy = oldpolicy = -1;
4231
		task_rq_unlock(rq, p, &rf);
L
Linus Torvalds 已提交
4232 4233
		goto recheck;
	}
4234 4235 4236 4237 4238 4239

	/*
	 * If setscheduling to SCHED_DEADLINE (or changing the parameters
	 * of a SCHED_DEADLINE task) we need to check if enough bandwidth
	 * is available.
	 */
4240
	if ((dl_policy(policy) || dl_task(p)) && sched_dl_overflow(p, policy, attr)) {
4241
		task_rq_unlock(rq, p, &rf);
4242 4243 4244
		return -EBUSY;
	}

4245 4246 4247
	p->sched_reset_on_fork = reset_on_fork;
	oldprio = p->prio;

4248 4249 4250 4251 4252 4253 4254 4255
	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.
		 */
4256
		new_effective_prio = rt_effective_prio(p, newprio);
4257 4258
		if (new_effective_prio == oldprio)
			queue_flags &= ~DEQUEUE_MOVE;
4259 4260
	}

4261
	queued = task_on_rq_queued(p);
4262
	running = task_current(rq, p);
4263
	if (queued)
4264
		dequeue_task(rq, p, queue_flags);
4265
	if (running)
4266
		put_prev_task(rq, p);
4267

4268
	prev_class = p->sched_class;
4269
	__setscheduler(rq, p, attr, pi);
4270

4271
	if (queued) {
4272 4273 4274 4275
		/*
		 * We enqueue to tail when the priority of a task is
		 * increased (user space view).
		 */
4276 4277
		if (oldprio < p->prio)
			queue_flags |= ENQUEUE_HEAD;
4278

4279
		enqueue_task(rq, p, queue_flags);
4280
	}
4281
	if (running)
4282
		set_curr_task(rq, p);
4283

P
Peter Zijlstra 已提交
4284
	check_class_changed(rq, p, prev_class, oldprio);
I
Ingo Molnar 已提交
4285 4286 4287

	/* Avoid rq from going away on us: */
	preempt_disable();
4288
	task_rq_unlock(rq, p, &rf);
4289

4290 4291
	if (pi)
		rt_mutex_adjust_pi(p);
4292

I
Ingo Molnar 已提交
4293
	/* Run balance callbacks after we've adjusted the PI chain: */
4294 4295
	balance_callback(rq);
	preempt_enable();
4296

L
Linus Torvalds 已提交
4297 4298
	return 0;
}
4299

4300 4301 4302 4303 4304 4305 4306 4307 4308
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),
	};

4309 4310
	/* Fixup the legacy SCHED_RESET_ON_FORK hack. */
	if ((policy != SETPARAM_POLICY) && (policy & SCHED_RESET_ON_FORK)) {
4311 4312 4313 4314 4315
		attr.sched_flags |= SCHED_FLAG_RESET_ON_FORK;
		policy &= ~SCHED_RESET_ON_FORK;
		attr.sched_policy = policy;
	}

4316
	return __sched_setscheduler(p, &attr, check, true);
4317
}
4318 4319 4320 4321 4322 4323
/**
 * 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.
 *
4324 4325
 * Return: 0 on success. An error code otherwise.
 *
4326 4327 4328
 * NOTE that the task may be already dead.
 */
int sched_setscheduler(struct task_struct *p, int policy,
4329
		       const struct sched_param *param)
4330
{
4331
	return _sched_setscheduler(p, policy, param, true);
4332
}
L
Linus Torvalds 已提交
4333 4334
EXPORT_SYMBOL_GPL(sched_setscheduler);

4335 4336
int sched_setattr(struct task_struct *p, const struct sched_attr *attr)
{
4337
	return __sched_setscheduler(p, attr, true, true);
4338 4339 4340
}
EXPORT_SYMBOL_GPL(sched_setattr);

4341 4342 4343 4344 4345 4346 4347 4348 4349 4350
/**
 * 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.
4351 4352
 *
 * Return: 0 on success. An error code otherwise.
4353 4354
 */
int sched_setscheduler_nocheck(struct task_struct *p, int policy,
4355
			       const struct sched_param *param)
4356
{
4357
	return _sched_setscheduler(p, policy, param, false);
4358
}
4359
EXPORT_SYMBOL_GPL(sched_setscheduler_nocheck);
4360

I
Ingo Molnar 已提交
4361 4362
static int
do_sched_setscheduler(pid_t pid, int policy, struct sched_param __user *param)
L
Linus Torvalds 已提交
4363 4364 4365
{
	struct sched_param lparam;
	struct task_struct *p;
4366
	int retval;
L
Linus Torvalds 已提交
4367 4368 4369 4370 4371

	if (!param || pid < 0)
		return -EINVAL;
	if (copy_from_user(&lparam, param, sizeof(struct sched_param)))
		return -EFAULT;
4372 4373 4374

	rcu_read_lock();
	retval = -ESRCH;
L
Linus Torvalds 已提交
4375
	p = find_process_by_pid(pid);
4376 4377 4378
	if (p != NULL)
		retval = sched_setscheduler(p, policy, &lparam);
	rcu_read_unlock();
4379

L
Linus Torvalds 已提交
4380 4381 4382
	return retval;
}

4383 4384 4385
/*
 * Mimics kernel/events/core.c perf_copy_attr().
 */
I
Ingo Molnar 已提交
4386
static int sched_copy_attr(struct sched_attr __user *uattr, struct sched_attr *attr)
4387 4388 4389 4390 4391 4392 4393
{
	u32 size;
	int ret;

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

I
Ingo Molnar 已提交
4394
	/* Zero the full structure, so that a short copy will be nice: */
4395 4396 4397 4398 4399 4400
	memset(attr, 0, sizeof(*attr));

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

I
Ingo Molnar 已提交
4401 4402
	/* Bail out on silly large: */
	if (size > PAGE_SIZE)
4403 4404
		goto err_size;

I
Ingo Molnar 已提交
4405 4406
	/* ABI compatibility quirk: */
	if (!size)
4407 4408 4409 4410 4411 4412 4413 4414 4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440
		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
4441
	 * XXX: Do we want to be lenient like existing syscalls; or do we want
4442 4443
	 * to be strict and return an error on out-of-bounds values?
	 */
4444
	attr->sched_nice = clamp(attr->sched_nice, MIN_NICE, MAX_NICE);
4445

4446
	return 0;
4447 4448 4449

err_size:
	put_user(sizeof(*attr), &uattr->size);
4450
	return -E2BIG;
4451 4452
}

L
Linus Torvalds 已提交
4453 4454 4455 4456 4457
/**
 * 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.
4458 4459
 *
 * Return: 0 on success. An error code otherwise.
L
Linus Torvalds 已提交
4460
 */
I
Ingo Molnar 已提交
4461
SYSCALL_DEFINE3(sched_setscheduler, pid_t, pid, int, policy, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4462
{
4463 4464 4465
	if (policy < 0)
		return -EINVAL;

L
Linus Torvalds 已提交
4466 4467 4468 4469 4470 4471 4472
	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.
4473 4474
 *
 * Return: 0 on success. An error code otherwise.
L
Linus Torvalds 已提交
4475
 */
4476
SYSCALL_DEFINE2(sched_setparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4477
{
4478
	return do_sched_setscheduler(pid, SETPARAM_POLICY, param);
L
Linus Torvalds 已提交
4479 4480
}

4481 4482 4483
/**
 * sys_sched_setattr - same as above, but with extended sched_attr
 * @pid: the pid in question.
J
Juri Lelli 已提交
4484
 * @uattr: structure containing the extended parameters.
4485
 * @flags: for future extension.
4486
 */
4487 4488
SYSCALL_DEFINE3(sched_setattr, pid_t, pid, struct sched_attr __user *, uattr,
			       unsigned int, flags)
4489 4490 4491 4492 4493
{
	struct sched_attr attr;
	struct task_struct *p;
	int retval;

4494
	if (!uattr || pid < 0 || flags)
4495 4496
		return -EINVAL;

4497 4498 4499
	retval = sched_copy_attr(uattr, &attr);
	if (retval)
		return retval;
4500

4501
	if ((int)attr.sched_policy < 0)
4502
		return -EINVAL;
4503 4504 4505 4506 4507 4508 4509 4510 4511 4512 4513

	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
4514 4515 4516
/**
 * sys_sched_getscheduler - get the policy (scheduling class) of a thread
 * @pid: the pid in question.
4517 4518 4519
 *
 * Return: On success, the policy of the thread. Otherwise, a negative error
 * code.
L
Linus Torvalds 已提交
4520
 */
4521
SYSCALL_DEFINE1(sched_getscheduler, pid_t, pid)
L
Linus Torvalds 已提交
4522
{
4523
	struct task_struct *p;
4524
	int retval;
L
Linus Torvalds 已提交
4525 4526

	if (pid < 0)
4527
		return -EINVAL;
L
Linus Torvalds 已提交
4528 4529

	retval = -ESRCH;
4530
	rcu_read_lock();
L
Linus Torvalds 已提交
4531 4532 4533 4534
	p = find_process_by_pid(pid);
	if (p) {
		retval = security_task_getscheduler(p);
		if (!retval)
4535 4536
			retval = p->policy
				| (p->sched_reset_on_fork ? SCHED_RESET_ON_FORK : 0);
L
Linus Torvalds 已提交
4537
	}
4538
	rcu_read_unlock();
L
Linus Torvalds 已提交
4539 4540 4541 4542
	return retval;
}

/**
4543
 * sys_sched_getparam - get the RT priority of a thread
L
Linus Torvalds 已提交
4544 4545
 * @pid: the pid in question.
 * @param: structure containing the RT priority.
4546 4547 4548
 *
 * Return: On success, 0 and the RT priority is in @param. Otherwise, an error
 * code.
L
Linus Torvalds 已提交
4549
 */
4550
SYSCALL_DEFINE2(sched_getparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4551
{
4552
	struct sched_param lp = { .sched_priority = 0 };
4553
	struct task_struct *p;
4554
	int retval;
L
Linus Torvalds 已提交
4555 4556

	if (!param || pid < 0)
4557
		return -EINVAL;
L
Linus Torvalds 已提交
4558

4559
	rcu_read_lock();
L
Linus Torvalds 已提交
4560 4561 4562 4563 4564 4565 4566 4567 4568
	p = find_process_by_pid(pid);
	retval = -ESRCH;
	if (!p)
		goto out_unlock;

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

4569 4570
	if (task_has_rt_policy(p))
		lp.sched_priority = p->rt_priority;
4571
	rcu_read_unlock();
L
Linus Torvalds 已提交
4572 4573 4574 4575 4576 4577 4578 4579 4580

	/*
	 * 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:
4581
	rcu_read_unlock();
L
Linus Torvalds 已提交
4582 4583 4584
	return retval;
}

4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604 4605 4606 4607
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)
4608
				return -EFBIG;
4609 4610 4611 4612 4613
		}

		attr->size = usize;
	}

4614
	ret = copy_to_user(uattr, attr, attr->size);
4615 4616 4617
	if (ret)
		return -EFAULT;

4618
	return 0;
4619 4620 4621
}

/**
4622
 * sys_sched_getattr - similar to sched_getparam, but with sched_attr
4623
 * @pid: the pid in question.
J
Juri Lelli 已提交
4624
 * @uattr: structure containing the extended parameters.
4625
 * @size: sizeof(attr) for fwd/bwd comp.
4626
 * @flags: for future extension.
4627
 */
4628 4629
SYSCALL_DEFINE4(sched_getattr, pid_t, pid, struct sched_attr __user *, uattr,
		unsigned int, size, unsigned int, flags)
4630 4631 4632 4633 4634 4635 4636 4637
{
	struct sched_attr attr = {
		.size = sizeof(struct sched_attr),
	};
	struct task_struct *p;
	int retval;

	if (!uattr || pid < 0 || size > PAGE_SIZE ||
4638
	    size < SCHED_ATTR_SIZE_VER0 || flags)
4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650 4651
		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;
4652 4653
	if (p->sched_reset_on_fork)
		attr.sched_flags |= SCHED_FLAG_RESET_ON_FORK;
4654 4655 4656
	if (task_has_dl_policy(p))
		__getparam_dl(p, &attr);
	else if (task_has_rt_policy(p))
4657 4658
		attr.sched_priority = p->rt_priority;
	else
4659
		attr.sched_nice = task_nice(p);
4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670

	rcu_read_unlock();

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

out_unlock:
	rcu_read_unlock();
	return retval;
}

4671
long sched_setaffinity(pid_t pid, const struct cpumask *in_mask)
L
Linus Torvalds 已提交
4672
{
4673
	cpumask_var_t cpus_allowed, new_mask;
4674 4675
	struct task_struct *p;
	int retval;
L
Linus Torvalds 已提交
4676

4677
	rcu_read_lock();
L
Linus Torvalds 已提交
4678 4679 4680

	p = find_process_by_pid(pid);
	if (!p) {
4681
		rcu_read_unlock();
L
Linus Torvalds 已提交
4682 4683 4684
		return -ESRCH;
	}

4685
	/* Prevent p going away */
L
Linus Torvalds 已提交
4686
	get_task_struct(p);
4687
	rcu_read_unlock();
L
Linus Torvalds 已提交
4688

4689 4690 4691 4692
	if (p->flags & PF_NO_SETAFFINITY) {
		retval = -EINVAL;
		goto out_put_task;
	}
4693 4694 4695 4696 4697 4698 4699 4700
	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
4701
	retval = -EPERM;
E
Eric W. Biederman 已提交
4702 4703 4704 4705
	if (!check_same_owner(p)) {
		rcu_read_lock();
		if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
			rcu_read_unlock();
4706
			goto out_free_new_mask;
E
Eric W. Biederman 已提交
4707 4708 4709
		}
		rcu_read_unlock();
	}
L
Linus Torvalds 已提交
4710

4711
	retval = security_task_setscheduler(p);
4712
	if (retval)
4713
		goto out_free_new_mask;
4714

4715 4716 4717 4718

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

4719 4720 4721 4722 4723 4724 4725
	/*
	 * 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
4726 4727 4728
	if (task_has_dl_policy(p) && dl_bandwidth_enabled()) {
		rcu_read_lock();
		if (!cpumask_subset(task_rq(p)->rd->span, new_mask)) {
4729
			retval = -EBUSY;
4730
			rcu_read_unlock();
4731
			goto out_free_new_mask;
4732
		}
4733
		rcu_read_unlock();
4734 4735
	}
#endif
P
Peter Zijlstra 已提交
4736
again:
4737
	retval = __set_cpus_allowed_ptr(p, new_mask, true);
L
Linus Torvalds 已提交
4738

P
Paul Menage 已提交
4739
	if (!retval) {
4740 4741
		cpuset_cpus_allowed(p, cpus_allowed);
		if (!cpumask_subset(new_mask, cpus_allowed)) {
P
Paul Menage 已提交
4742 4743 4744 4745 4746
			/*
			 * We must have raced with a concurrent cpuset
			 * update. Just reset the cpus_allowed to the
			 * cpuset's cpus_allowed
			 */
4747
			cpumask_copy(new_mask, cpus_allowed);
P
Paul Menage 已提交
4748 4749 4750
			goto again;
		}
	}
4751
out_free_new_mask:
4752 4753 4754 4755
	free_cpumask_var(new_mask);
out_free_cpus_allowed:
	free_cpumask_var(cpus_allowed);
out_put_task:
L
Linus Torvalds 已提交
4756 4757 4758 4759 4760
	put_task_struct(p);
	return retval;
}

static int get_user_cpu_mask(unsigned long __user *user_mask_ptr, unsigned len,
4761
			     struct cpumask *new_mask)
L
Linus Torvalds 已提交
4762
{
4763 4764 4765 4766 4767
	if (len < cpumask_size())
		cpumask_clear(new_mask);
	else if (len > cpumask_size())
		len = cpumask_size();

L
Linus Torvalds 已提交
4768 4769 4770 4771
	return copy_from_user(new_mask, user_mask_ptr, len) ? -EFAULT : 0;
}

/**
I
Ingo Molnar 已提交
4772
 * sys_sched_setaffinity - set the CPU affinity of a process
L
Linus Torvalds 已提交
4773 4774
 * @pid: pid of the process
 * @len: length in bytes of the bitmask pointed to by user_mask_ptr
I
Ingo Molnar 已提交
4775
 * @user_mask_ptr: user-space pointer to the new CPU mask
4776 4777
 *
 * Return: 0 on success. An error code otherwise.
L
Linus Torvalds 已提交
4778
 */
4779 4780
SYSCALL_DEFINE3(sched_setaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
4781
{
4782
	cpumask_var_t new_mask;
L
Linus Torvalds 已提交
4783 4784
	int retval;

4785 4786
	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
4787

4788 4789 4790 4791 4792
	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
4793 4794
}

4795
long sched_getaffinity(pid_t pid, struct cpumask *mask)
L
Linus Torvalds 已提交
4796
{
4797
	struct task_struct *p;
4798
	unsigned long flags;
L
Linus Torvalds 已提交
4799 4800
	int retval;

4801
	rcu_read_lock();
L
Linus Torvalds 已提交
4802 4803 4804 4805 4806 4807

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

4808 4809 4810 4811
	retval = security_task_getscheduler(p);
	if (retval)
		goto out_unlock;

4812
	raw_spin_lock_irqsave(&p->pi_lock, flags);
4813
	cpumask_and(mask, &p->cpus_allowed, cpu_active_mask);
4814
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
4815 4816

out_unlock:
4817
	rcu_read_unlock();
L
Linus Torvalds 已提交
4818

4819
	return retval;
L
Linus Torvalds 已提交
4820 4821 4822
}

/**
I
Ingo Molnar 已提交
4823
 * sys_sched_getaffinity - get the CPU affinity of a process
L
Linus Torvalds 已提交
4824 4825
 * @pid: pid of the process
 * @len: length in bytes of the bitmask pointed to by user_mask_ptr
I
Ingo Molnar 已提交
4826
 * @user_mask_ptr: user-space pointer to hold the current CPU mask
4827
 *
4828 4829
 * Return: size of CPU mask copied to user_mask_ptr on success. An
 * error code otherwise.
L
Linus Torvalds 已提交
4830
 */
4831 4832
SYSCALL_DEFINE3(sched_getaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
4833 4834
{
	int ret;
4835
	cpumask_var_t mask;
L
Linus Torvalds 已提交
4836

A
Anton Blanchard 已提交
4837
	if ((len * BITS_PER_BYTE) < nr_cpu_ids)
4838 4839
		return -EINVAL;
	if (len & (sizeof(unsigned long)-1))
L
Linus Torvalds 已提交
4840 4841
		return -EINVAL;

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

4845 4846
	ret = sched_getaffinity(pid, mask);
	if (ret == 0) {
4847
		size_t retlen = min_t(size_t, len, cpumask_size());
4848 4849

		if (copy_to_user(user_mask_ptr, mask, retlen))
4850 4851
			ret = -EFAULT;
		else
4852
			ret = retlen;
4853 4854
	}
	free_cpumask_var(mask);
L
Linus Torvalds 已提交
4855

4856
	return ret;
L
Linus Torvalds 已提交
4857 4858 4859 4860 4861
}

/**
 * sys_sched_yield - yield the current processor to other threads.
 *
I
Ingo Molnar 已提交
4862 4863
 * This function yields the current CPU to other tasks. If there are no
 * other threads running on this CPU then this function will return.
4864 4865
 *
 * Return: 0.
L
Linus Torvalds 已提交
4866
 */
4867
SYSCALL_DEFINE0(sched_yield)
L
Linus Torvalds 已提交
4868
{
4869 4870 4871 4872 4873 4874
	struct rq_flags rf;
	struct rq *rq;

	local_irq_disable();
	rq = this_rq();
	rq_lock(rq, &rf);
L
Linus Torvalds 已提交
4875

4876
	schedstat_inc(rq->yld_count);
4877
	current->sched_class->yield_task(rq);
L
Linus Torvalds 已提交
4878 4879 4880 4881 4882

	/*
	 * Since we are going to call schedule() anyway, there's
	 * no need to preempt or enable interrupts:
	 */
4883 4884
	preempt_disable();
	rq_unlock(rq, &rf);
4885
	sched_preempt_enable_no_resched();
L
Linus Torvalds 已提交
4886 4887 4888 4889 4890 4891

	schedule();

	return 0;
}

4892
#ifndef CONFIG_PREEMPT
4893
int __sched _cond_resched(void)
L
Linus Torvalds 已提交
4894
{
4895
	if (should_resched(0)) {
4896
		preempt_schedule_common();
L
Linus Torvalds 已提交
4897 4898
		return 1;
	}
4899
	rcu_all_qs();
L
Linus Torvalds 已提交
4900 4901
	return 0;
}
4902
EXPORT_SYMBOL(_cond_resched);
4903
#endif
L
Linus Torvalds 已提交
4904 4905

/*
4906
 * __cond_resched_lock() - if a reschedule is pending, drop the given lock,
L
Linus Torvalds 已提交
4907 4908
 * call schedule, and on return reacquire the lock.
 *
I
Ingo Molnar 已提交
4909
 * This works OK both with and without CONFIG_PREEMPT. We do strange low-level
L
Linus Torvalds 已提交
4910 4911 4912
 * operations here to prevent schedule() from being called twice (once via
 * spin_unlock(), once by hand).
 */
4913
int __cond_resched_lock(spinlock_t *lock)
L
Linus Torvalds 已提交
4914
{
4915
	int resched = should_resched(PREEMPT_LOCK_OFFSET);
J
Jan Kara 已提交
4916 4917
	int ret = 0;

4918 4919
	lockdep_assert_held(lock);

4920
	if (spin_needbreak(lock) || resched) {
L
Linus Torvalds 已提交
4921
		spin_unlock(lock);
P
Peter Zijlstra 已提交
4922
		if (resched)
4923
			preempt_schedule_common();
N
Nick Piggin 已提交
4924 4925
		else
			cpu_relax();
J
Jan Kara 已提交
4926
		ret = 1;
L
Linus Torvalds 已提交
4927 4928
		spin_lock(lock);
	}
J
Jan Kara 已提交
4929
	return ret;
L
Linus Torvalds 已提交
4930
}
4931
EXPORT_SYMBOL(__cond_resched_lock);
L
Linus Torvalds 已提交
4932

4933
int __sched __cond_resched_softirq(void)
L
Linus Torvalds 已提交
4934 4935 4936
{
	BUG_ON(!in_softirq());

4937
	if (should_resched(SOFTIRQ_DISABLE_OFFSET)) {
4938
		local_bh_enable();
4939
		preempt_schedule_common();
L
Linus Torvalds 已提交
4940 4941 4942 4943 4944
		local_bh_disable();
		return 1;
	}
	return 0;
}
4945
EXPORT_SYMBOL(__cond_resched_softirq);
L
Linus Torvalds 已提交
4946 4947 4948 4949

/**
 * yield - yield the current processor to other threads.
 *
P
Peter Zijlstra 已提交
4950 4951 4952 4953 4954 4955 4956 4957 4958
 * 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
4959
 *	yield();
P
Peter Zijlstra 已提交
4960 4961 4962 4963 4964 4965 4966 4967
 *
 * 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
4968 4969 4970 4971 4972 4973 4974 4975
 */
void __sched yield(void)
{
	set_current_state(TASK_RUNNING);
	sys_sched_yield();
}
EXPORT_SYMBOL(yield);

4976 4977 4978 4979
/**
 * 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
4980 4981
 * @p: target task
 * @preempt: whether task preemption is allowed or not
4982 4983 4984 4985
 *
 * 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.
 *
4986
 * Return:
4987 4988 4989
 *	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.
4990
 */
4991
int __sched yield_to(struct task_struct *p, bool preempt)
4992 4993 4994 4995
{
	struct task_struct *curr = current;
	struct rq *rq, *p_rq;
	unsigned long flags;
4996
	int yielded = 0;
4997 4998 4999 5000 5001 5002

	local_irq_save(flags);
	rq = this_rq();

again:
	p_rq = task_rq(p);
5003 5004 5005 5006 5007 5008 5009 5010 5011
	/*
	 * 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;
	}

5012
	double_rq_lock(rq, p_rq);
5013
	if (task_rq(p) != p_rq) {
5014 5015 5016 5017 5018
		double_rq_unlock(rq, p_rq);
		goto again;
	}

	if (!curr->sched_class->yield_to_task)
5019
		goto out_unlock;
5020 5021

	if (curr->sched_class != p->sched_class)
5022
		goto out_unlock;
5023 5024

	if (task_running(p_rq, p) || p->state)
5025
		goto out_unlock;
5026 5027

	yielded = curr->sched_class->yield_to_task(rq, p, preempt);
5028
	if (yielded) {
5029
		schedstat_inc(rq->yld_count);
5030 5031 5032 5033 5034
		/*
		 * Make p's CPU reschedule; pick_next_entity takes care of
		 * fairness.
		 */
		if (preempt && rq != p_rq)
5035
			resched_curr(p_rq);
5036
	}
5037

5038
out_unlock:
5039
	double_rq_unlock(rq, p_rq);
5040
out_irq:
5041 5042
	local_irq_restore(flags);

5043
	if (yielded > 0)
5044 5045 5046 5047 5048 5049
		schedule();

	return yielded;
}
EXPORT_SYMBOL_GPL(yield_to);

5050 5051 5052 5053 5054 5055 5056 5057 5058 5059 5060 5061 5062 5063 5064
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
5065
/*
I
Ingo Molnar 已提交
5066
 * This task is about to go to sleep on IO. Increment rq->nr_iowait so
L
Linus Torvalds 已提交
5067 5068 5069 5070
 * that process accounting knows that this is a task in IO wait state.
 */
long __sched io_schedule_timeout(long timeout)
{
5071
	int token;
L
Linus Torvalds 已提交
5072 5073
	long ret;

5074
	token = io_schedule_prepare();
L
Linus Torvalds 已提交
5075
	ret = schedule_timeout(timeout);
5076
	io_schedule_finish(token);
5077

L
Linus Torvalds 已提交
5078 5079
	return ret;
}
5080
EXPORT_SYMBOL(io_schedule_timeout);
L
Linus Torvalds 已提交
5081

5082 5083 5084 5085 5086 5087 5088 5089 5090 5091
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
5092 5093 5094 5095
/**
 * sys_sched_get_priority_max - return maximum RT priority.
 * @policy: scheduling class.
 *
5096 5097 5098
 * 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
5099
 */
5100
SYSCALL_DEFINE1(sched_get_priority_max, int, policy)
L
Linus Torvalds 已提交
5101 5102 5103 5104 5105 5106 5107 5108
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = MAX_USER_RT_PRIO-1;
		break;
5109
	case SCHED_DEADLINE:
L
Linus Torvalds 已提交
5110
	case SCHED_NORMAL:
5111
	case SCHED_BATCH:
I
Ingo Molnar 已提交
5112
	case SCHED_IDLE:
L
Linus Torvalds 已提交
5113 5114 5115 5116 5117 5118 5119 5120 5121 5122
		ret = 0;
		break;
	}
	return ret;
}

/**
 * sys_sched_get_priority_min - return minimum RT priority.
 * @policy: scheduling class.
 *
5123 5124 5125
 * 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
5126
 */
5127
SYSCALL_DEFINE1(sched_get_priority_min, int, policy)
L
Linus Torvalds 已提交
5128 5129 5130 5131 5132 5133 5134 5135
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = 1;
		break;
5136
	case SCHED_DEADLINE:
L
Linus Torvalds 已提交
5137
	case SCHED_NORMAL:
5138
	case SCHED_BATCH:
I
Ingo Molnar 已提交
5139
	case SCHED_IDLE:
L
Linus Torvalds 已提交
5140 5141 5142 5143 5144
		ret = 0;
	}
	return ret;
}

5145
static int sched_rr_get_interval(pid_t pid, struct timespec64 *t)
L
Linus Torvalds 已提交
5146
{
5147
	struct task_struct *p;
D
Dmitry Adamushko 已提交
5148
	unsigned int time_slice;
5149
	struct rq_flags rf;
5150
	struct rq *rq;
5151
	int retval;
L
Linus Torvalds 已提交
5152 5153

	if (pid < 0)
5154
		return -EINVAL;
L
Linus Torvalds 已提交
5155 5156

	retval = -ESRCH;
5157
	rcu_read_lock();
L
Linus Torvalds 已提交
5158 5159 5160 5161 5162 5163 5164 5165
	p = find_process_by_pid(pid);
	if (!p)
		goto out_unlock;

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

5166
	rq = task_rq_lock(p, &rf);
5167 5168 5169
	time_slice = 0;
	if (p->sched_class->get_rr_interval)
		time_slice = p->sched_class->get_rr_interval(rq, p);
5170
	task_rq_unlock(rq, p, &rf);
D
Dmitry Adamushko 已提交
5171

5172
	rcu_read_unlock();
5173 5174
	jiffies_to_timespec64(time_slice, t);
	return 0;
5175

L
Linus Torvalds 已提交
5176
out_unlock:
5177
	rcu_read_unlock();
L
Linus Torvalds 已提交
5178 5179 5180
	return retval;
}

5181 5182 5183 5184 5185 5186 5187 5188 5189 5190 5191
/**
 * 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.
 *
 * Return: On success, 0 and the timeslice is in @interval. Otherwise,
 * an error code.
 */
5192 5193 5194 5195 5196 5197 5198 5199 5200 5201 5202 5203 5204 5205 5206 5207 5208 5209 5210 5211 5212 5213 5214 5215 5216 5217
SYSCALL_DEFINE2(sched_rr_get_interval, pid_t, pid,
		struct timespec __user *, interval)
{
	struct timespec64 t;
	int retval = sched_rr_get_interval(pid, &t);

	if (retval == 0)
		retval = put_timespec64(&t, interval);

	return retval;
}

#ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE2(sched_rr_get_interval,
		       compat_pid_t, pid,
		       struct compat_timespec __user *, interval)
{
	struct timespec64 t;
	int retval = sched_rr_get_interval(pid, &t);

	if (retval == 0)
		retval = compat_put_timespec64(&t, interval);
	return retval;
}
#endif

5218
void sched_show_task(struct task_struct *p)
L
Linus Torvalds 已提交
5219 5220
{
	unsigned long free = 0;
5221
	int ppid;
5222

5223 5224
	if (!try_get_task_stack(p))
		return;
5225 5226 5227 5228

	printk(KERN_INFO "%-15.15s %c", p->comm, task_state_to_char(p));

	if (p->state == TASK_RUNNING)
P
Peter Zijlstra 已提交
5229
		printk(KERN_CONT "  running task    ");
L
Linus Torvalds 已提交
5230
#ifdef CONFIG_DEBUG_STACK_USAGE
5231
	free = stack_not_used(p);
L
Linus Torvalds 已提交
5232
#endif
5233
	ppid = 0;
5234
	rcu_read_lock();
5235 5236
	if (pid_alive(p))
		ppid = task_pid_nr(rcu_dereference(p->real_parent));
5237
	rcu_read_unlock();
P
Peter Zijlstra 已提交
5238
	printk(KERN_CONT "%5lu %5d %6d 0x%08lx\n", free,
5239
		task_pid_nr(p), ppid,
5240
		(unsigned long)task_thread_info(p)->flags);
L
Linus Torvalds 已提交
5241

5242
	print_worker_info(KERN_INFO, p);
5243
	show_stack(p, NULL);
5244
	put_task_stack(p);
L
Linus Torvalds 已提交
5245
}
5246
EXPORT_SYMBOL_GPL(sched_show_task);
L
Linus Torvalds 已提交
5247

5248 5249 5250 5251 5252 5253 5254 5255 5256 5257 5258 5259 5260 5261 5262 5263 5264 5265 5266 5267 5268 5269
static inline bool
state_filter_match(unsigned long state_filter, struct task_struct *p)
{
	/* no filter, everything matches */
	if (!state_filter)
		return true;

	/* filter, but doesn't match */
	if (!(p->state & state_filter))
		return false;

	/*
	 * When looking for TASK_UNINTERRUPTIBLE skip TASK_IDLE (allows
	 * TASK_KILLABLE).
	 */
	if (state_filter == TASK_UNINTERRUPTIBLE && p->state == TASK_IDLE)
		return false;

	return true;
}


I
Ingo Molnar 已提交
5270
void show_state_filter(unsigned long state_filter)
L
Linus Torvalds 已提交
5271
{
5272
	struct task_struct *g, *p;
L
Linus Torvalds 已提交
5273

5274
#if BITS_PER_LONG == 32
P
Peter Zijlstra 已提交
5275 5276
	printk(KERN_INFO
		"  task                PC stack   pid father\n");
L
Linus Torvalds 已提交
5277
#else
P
Peter Zijlstra 已提交
5278 5279
	printk(KERN_INFO
		"  task                        PC stack   pid father\n");
L
Linus Torvalds 已提交
5280
#endif
5281
	rcu_read_lock();
5282
	for_each_process_thread(g, p) {
L
Linus Torvalds 已提交
5283 5284
		/*
		 * reset the NMI-timeout, listing all files on a slow
L
Lucas De Marchi 已提交
5285
		 * console might take a lot of time:
5286 5287 5288
		 * 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
5289 5290
		 */
		touch_nmi_watchdog();
5291
		touch_all_softlockup_watchdogs();
5292
		if (state_filter_match(state_filter, p))
5293
			sched_show_task(p);
5294
	}
L
Linus Torvalds 已提交
5295

I
Ingo Molnar 已提交
5296
#ifdef CONFIG_SCHED_DEBUG
5297 5298
	if (!state_filter)
		sysrq_sched_debug_show();
I
Ingo Molnar 已提交
5299
#endif
5300
	rcu_read_unlock();
I
Ingo Molnar 已提交
5301 5302 5303
	/*
	 * Only show locks if all tasks are dumped:
	 */
5304
	if (!state_filter)
I
Ingo Molnar 已提交
5305
		debug_show_all_locks();
L
Linus Torvalds 已提交
5306 5307
}

5308 5309 5310
/**
 * init_idle - set up an idle thread for a given CPU
 * @idle: task in question
I
Ingo Molnar 已提交
5311
 * @cpu: CPU the idle task belongs to
5312 5313 5314 5315
 *
 * NOTE: this function does not set the idle thread's NEED_RESCHED
 * flag, to make booting more robust.
 */
5316
void init_idle(struct task_struct *idle, int cpu)
L
Linus Torvalds 已提交
5317
{
5318
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
5319 5320
	unsigned long flags;

5321 5322
	raw_spin_lock_irqsave(&idle->pi_lock, flags);
	raw_spin_lock(&rq->lock);
5323

5324
	__sched_fork(0, idle);
5325
	idle->state = TASK_RUNNING;
I
Ingo Molnar 已提交
5326
	idle->se.exec_start = sched_clock();
5327
	idle->flags |= PF_IDLE;
I
Ingo Molnar 已提交
5328

5329 5330
	kasan_unpoison_task_stack(idle);

5331 5332 5333 5334 5335 5336 5337 5338 5339
#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
5340 5341
	/*
	 * We're having a chicken and egg problem, even though we are
I
Ingo Molnar 已提交
5342
	 * holding rq->lock, the CPU isn't yet set to this CPU so the
5343 5344 5345 5346 5347 5348 5349 5350
	 * 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
5351
	__set_task_cpu(idle, cpu);
5352
	rcu_read_unlock();
L
Linus Torvalds 已提交
5353 5354

	rq->curr = rq->idle = idle;
5355
	idle->on_rq = TASK_ON_RQ_QUEUED;
5356
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
5357
	idle->on_cpu = 1;
5358
#endif
5359 5360
	raw_spin_unlock(&rq->lock);
	raw_spin_unlock_irqrestore(&idle->pi_lock, flags);
L
Linus Torvalds 已提交
5361 5362

	/* Set the preempt count _outside_ the spinlocks! */
5363
	init_idle_preempt_count(idle, cpu);
5364

I
Ingo Molnar 已提交
5365 5366 5367 5368
	/*
	 * The idle tasks have their own, simple scheduling class:
	 */
	idle->sched_class = &idle_sched_class;
5369
	ftrace_graph_init_idle_task(idle, cpu);
5370
	vtime_init_idle(idle, cpu);
5371
#ifdef CONFIG_SMP
5372 5373
	sprintf(idle->comm, "%s/%d", INIT_TASK_COMM, cpu);
#endif
I
Ingo Molnar 已提交
5374 5375
}

5376 5377
#ifdef CONFIG_SMP

5378 5379 5380
int cpuset_cpumask_can_shrink(const struct cpumask *cur,
			      const struct cpumask *trial)
{
5381
	int ret = 1;
5382

5383 5384 5385
	if (!cpumask_weight(cur))
		return ret;

5386
	ret = dl_cpuset_cpumask_can_shrink(cur, trial);
5387 5388 5389 5390

	return ret;
}

5391 5392 5393 5394 5395 5396 5397
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
5398
	 * to a new cpuset; we don't want to change their CPU
5399 5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410
	 * 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,
5411 5412
					      cs_cpus_allowed))
		ret = dl_task_can_attach(p, cs_cpus_allowed);
5413 5414 5415 5416 5417

out:
	return ret;
}

5418
bool sched_smp_initialized __read_mostly;
5419

5420 5421 5422 5423 5424 5425 5426 5427 5428 5429
#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;

5430
	if (!cpumask_test_cpu(target_cpu, &p->cpus_allowed))
5431 5432 5433 5434
		return -EINVAL;

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

5435
	trace_sched_move_numa(p, curr_cpu, target_cpu);
5436 5437
	return stop_one_cpu(curr_cpu, migration_cpu_stop, &arg);
}
5438 5439 5440 5441 5442 5443 5444

/*
 * 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)
{
5445
	bool queued, running;
5446 5447
	struct rq_flags rf;
	struct rq *rq;
5448

5449
	rq = task_rq_lock(p, &rf);
5450
	queued = task_on_rq_queued(p);
5451 5452
	running = task_current(rq, p);

5453
	if (queued)
5454
		dequeue_task(rq, p, DEQUEUE_SAVE);
5455
	if (running)
5456
		put_prev_task(rq, p);
5457 5458 5459

	p->numa_preferred_nid = nid;

5460
	if (queued)
5461
		enqueue_task(rq, p, ENQUEUE_RESTORE | ENQUEUE_NOCLOCK);
5462
	if (running)
5463
		set_curr_task(rq, p);
5464
	task_rq_unlock(rq, p, &rf);
5465
}
P
Peter Zijlstra 已提交
5466
#endif /* CONFIG_NUMA_BALANCING */
5467

L
Linus Torvalds 已提交
5468
#ifdef CONFIG_HOTPLUG_CPU
5469
/*
I
Ingo Molnar 已提交
5470
 * Ensure that the idle task is using init_mm right before its CPU goes
5471
 * offline.
5472
 */
5473
void idle_task_exit(void)
L
Linus Torvalds 已提交
5474
{
5475
	struct mm_struct *mm = current->active_mm;
5476

5477
	BUG_ON(cpu_online(smp_processor_id()));
5478

5479
	if (mm != &init_mm) {
5480
		switch_mm(mm, &init_mm, current);
5481 5482
		finish_arch_post_lock_switch();
	}
5483
	mmdrop(mm);
L
Linus Torvalds 已提交
5484 5485 5486
}

/*
5487 5488
 * 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
5489 5490 5491
 * 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.
5492 5493
 *
 * Also see the comment "Global load-average calculations".
L
Linus Torvalds 已提交
5494
 */
5495
static void calc_load_migrate(struct rq *rq)
L
Linus Torvalds 已提交
5496
{
5497
	long delta = calc_load_fold_active(rq, 1);
5498 5499
	if (delta)
		atomic_long_add(delta, &calc_load_tasks);
L
Linus Torvalds 已提交
5500 5501
}

5502 5503 5504 5505 5506 5507 5508 5509 5510 5511 5512 5513 5514 5515 5516 5517
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,
};

5518
/*
5519 5520 5521 5522 5523 5524
 * 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
5525
 */
5526
static void migrate_tasks(struct rq *dead_rq, struct rq_flags *rf)
L
Linus Torvalds 已提交
5527
{
5528
	struct rq *rq = dead_rq;
5529
	struct task_struct *next, *stop = rq->stop;
5530
	struct rq_flags orf = *rf;
5531
	int dest_cpu;
L
Linus Torvalds 已提交
5532 5533

	/*
5534 5535 5536 5537 5538 5539 5540
	 * 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
5541
	 */
5542
	rq->stop = NULL;
5543

5544 5545 5546 5547 5548 5549 5550
	/*
	 * 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);

5551
	for (;;) {
5552 5553
		/*
		 * There's this thread running, bail when that's the only
I
Ingo Molnar 已提交
5554
		 * remaining thread:
5555 5556
		 */
		if (rq->nr_running == 1)
I
Ingo Molnar 已提交
5557
			break;
5558

5559
		/*
I
Ingo Molnar 已提交
5560
		 * pick_next_task() assumes pinned rq->lock:
5561
		 */
5562
		next = pick_next_task(rq, &fake_task, rf);
5563
		BUG_ON(!next);
V
Viresh Kumar 已提交
5564
		put_prev_task(rq, next);
5565

W
Wanpeng Li 已提交
5566 5567 5568 5569 5570 5571 5572 5573 5574
		/*
		 * 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.
		 */
5575
		rq_unlock(rq, rf);
W
Wanpeng Li 已提交
5576
		raw_spin_lock(&next->pi_lock);
5577
		rq_relock(rq, rf);
W
Wanpeng Li 已提交
5578 5579 5580 5581 5582 5583 5584 5585 5586 5587 5588

		/*
		 * 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;
		}

5589
		/* Find suitable destination for @next, with force if needed. */
5590
		dest_cpu = select_fallback_rq(dead_rq->cpu, next);
5591
		rq = __migrate_task(rq, rf, next, dest_cpu);
5592
		if (rq != dead_rq) {
5593
			rq_unlock(rq, rf);
5594
			rq = dead_rq;
5595 5596
			*rf = orf;
			rq_relock(rq, rf);
5597
		}
W
Wanpeng Li 已提交
5598
		raw_spin_unlock(&next->pi_lock);
L
Linus Torvalds 已提交
5599
	}
5600

5601
	rq->stop = stop;
5602
}
L
Linus Torvalds 已提交
5603 5604
#endif /* CONFIG_HOTPLUG_CPU */

5605
void set_rq_online(struct rq *rq)
5606 5607 5608 5609
{
	if (!rq->online) {
		const struct sched_class *class;

5610
		cpumask_set_cpu(rq->cpu, rq->rd->online);
5611 5612 5613 5614 5615 5616 5617 5618 5619
		rq->online = 1;

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

5620
void set_rq_offline(struct rq *rq)
5621 5622 5623 5624 5625 5626 5627 5628 5629
{
	if (rq->online) {
		const struct sched_class *class;

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

5630
		cpumask_clear_cpu(rq->cpu, rq->rd->online);
5631 5632 5633 5634
		rq->online = 0;
	}
}

5635
static void set_cpu_rq_start_time(unsigned int cpu)
L
Linus Torvalds 已提交
5636
{
5637
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
5638

5639 5640 5641
	rq->age_stamp = sched_clock_cpu(cpu);
}

I
Ingo Molnar 已提交
5642 5643 5644 5645
/*
 * used to mark begin/end of suspend/resume:
 */
static int num_cpus_frozen;
5646

L
Linus Torvalds 已提交
5647
/*
5648 5649 5650
 * Update cpusets according to cpu_active mask.  If cpusets are
 * disabled, cpuset_update_active_cpus() becomes a simple wrapper
 * around partition_sched_domains().
5651 5652 5653
 *
 * 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
5654
 */
5655
static void cpuset_cpu_active(void)
5656
{
5657
	if (cpuhp_tasks_frozen) {
5658 5659 5660 5661 5662 5663
		/*
		 * 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.
		 */
5664 5665
		partition_sched_domains(1, NULL, NULL);
		if (--num_cpus_frozen)
5666
			return;
5667 5668 5669 5670 5671
		/*
		 * This is the last CPU online operation. So fall through and
		 * restore the original sched domains by considering the
		 * cpuset configurations.
		 */
5672
		cpuset_force_rebuild();
5673
	}
5674
	cpuset_update_active_cpus();
5675
}
5676

5677
static int cpuset_cpu_inactive(unsigned int cpu)
5678
{
5679
	if (!cpuhp_tasks_frozen) {
5680
		if (dl_cpu_busy(cpu))
5681
			return -EBUSY;
5682
		cpuset_update_active_cpus();
5683
	} else {
5684 5685
		num_cpus_frozen++;
		partition_sched_domains(1, NULL, NULL);
5686
	}
5687
	return 0;
5688 5689
}

5690
int sched_cpu_activate(unsigned int cpu)
5691
{
5692
	struct rq *rq = cpu_rq(cpu);
5693
	struct rq_flags rf;
5694

5695
	set_cpu_active(cpu, true);
5696

5697
	if (sched_smp_initialized) {
5698
		sched_domains_numa_masks_set(cpu);
5699
		cpuset_cpu_active();
5700
	}
5701 5702 5703 5704 5705

	/*
	 * 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
5706
	 *    after all CPUs have been brought up.
5707 5708 5709 5710
	 *
	 * 2) At runtime, if cpuset_cpu_active() fails to rebuild the
	 *    domains.
	 */
5711
	rq_lock_irqsave(rq, &rf);
5712 5713 5714 5715
	if (rq->rd) {
		BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
		set_rq_online(rq);
	}
5716
	rq_unlock_irqrestore(rq, &rf);
5717 5718 5719

	update_max_interval();

5720
	return 0;
5721 5722
}

5723
int sched_cpu_deactivate(unsigned int cpu)
5724 5725 5726
{
	int ret;

5727
	set_cpu_active(cpu, false);
5728 5729 5730 5731 5732 5733 5734
	/*
	 * 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.
	 *
	 * Do sync before park smpboot threads to take care the rcu boost case.
	 */
5735
	synchronize_rcu_mult(call_rcu, call_rcu_sched);
5736 5737 5738 5739 5740 5741 5742 5743

	if (!sched_smp_initialized)
		return 0;

	ret = cpuset_cpu_inactive(cpu);
	if (ret) {
		set_cpu_active(cpu, true);
		return ret;
5744
	}
5745 5746
	sched_domains_numa_masks_clear(cpu);
	return 0;
5747 5748
}

5749 5750 5751 5752 5753 5754 5755 5756
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();
}

5757 5758 5759
int sched_cpu_starting(unsigned int cpu)
{
	set_cpu_rq_start_time(cpu);
5760
	sched_rq_cpu_starting(cpu);
5761
	return 0;
5762 5763
}

5764 5765 5766 5767
#ifdef CONFIG_HOTPLUG_CPU
int sched_cpu_dying(unsigned int cpu)
{
	struct rq *rq = cpu_rq(cpu);
5768
	struct rq_flags rf;
5769 5770 5771

	/* Handle pending wakeups and then migrate everything off */
	sched_ttwu_pending();
5772 5773

	rq_lock_irqsave(rq, &rf);
5774 5775 5776 5777
	if (rq->rd) {
		BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
		set_rq_offline(rq);
	}
5778
	migrate_tasks(rq, &rf);
5779
	BUG_ON(rq->nr_running != 1);
5780 5781
	rq_unlock_irqrestore(rq, &rf);

5782 5783
	calc_load_migrate(rq);
	update_max_interval();
5784
	nohz_balance_exit_idle(cpu);
5785
	hrtick_clear(rq);
5786 5787 5788 5789
	return 0;
}
#endif

P
Peter Zijlstra 已提交
5790 5791 5792 5793 5794 5795 5796 5797 5798 5799 5800 5801 5802 5803 5804 5805
#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
5806 5807
void __init sched_init_smp(void)
{
5808 5809
	sched_init_numa();

5810 5811
	/*
	 * There's no userspace yet to cause hotplug operations; hence all the
I
Ingo Molnar 已提交
5812
	 * CPU masks are stable and all blatant races in the below code cannot
5813 5814
	 * happen.
	 */
5815
	mutex_lock(&sched_domains_mutex);
P
Peter Zijlstra 已提交
5816
	sched_init_domains(cpu_active_mask);
5817
	mutex_unlock(&sched_domains_mutex);
5818

5819
	/* Move init over to a non-isolated CPU */
5820
	if (set_cpus_allowed_ptr(current, housekeeping_cpumask(HK_FLAG_DOMAIN)) < 0)
5821
		BUG();
I
Ingo Molnar 已提交
5822
	sched_init_granularity();
5823

5824
	init_sched_rt_class();
5825
	init_sched_dl_class();
P
Peter Zijlstra 已提交
5826 5827 5828

	sched_init_smt();

5829
	sched_smp_initialized = true;
L
Linus Torvalds 已提交
5830
}
5831 5832 5833

static int __init migration_init(void)
{
5834
	sched_rq_cpu_starting(smp_processor_id());
5835
	return 0;
L
Linus Torvalds 已提交
5836
}
5837 5838
early_initcall(migration_init);

L
Linus Torvalds 已提交
5839 5840 5841
#else
void __init sched_init_smp(void)
{
I
Ingo Molnar 已提交
5842
	sched_init_granularity();
L
Linus Torvalds 已提交
5843 5844 5845 5846 5847 5848 5849 5850 5851 5852
}
#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);
}

5853
#ifdef CONFIG_CGROUP_SCHED
5854 5855 5856 5857
/*
 * Default task group.
 * Every task in system belongs to this group at bootup.
 */
5858
struct task_group root_task_group;
5859
LIST_HEAD(task_groups);
5860 5861 5862

/* Cacheline aligned slab cache for task_group */
static struct kmem_cache *task_group_cache __read_mostly;
5863
#endif
P
Peter Zijlstra 已提交
5864

5865
DECLARE_PER_CPU(cpumask_var_t, load_balance_mask);
5866
DECLARE_PER_CPU(cpumask_var_t, select_idle_mask);
P
Peter Zijlstra 已提交
5867

L
Linus Torvalds 已提交
5868 5869
void __init sched_init(void)
{
I
Ingo Molnar 已提交
5870
	int i, j;
5871 5872
	unsigned long alloc_size = 0, ptr;

5873
	sched_clock_init();
5874
	wait_bit_init();
5875

5876 5877 5878 5879 5880 5881 5882
#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) {
5883
		ptr = (unsigned long)kzalloc(alloc_size, GFP_NOWAIT);
5884 5885

#ifdef CONFIG_FAIR_GROUP_SCHED
5886
		root_task_group.se = (struct sched_entity **)ptr;
5887 5888
		ptr += nr_cpu_ids * sizeof(void **);

5889
		root_task_group.cfs_rq = (struct cfs_rq **)ptr;
5890
		ptr += nr_cpu_ids * sizeof(void **);
5891

5892
#endif /* CONFIG_FAIR_GROUP_SCHED */
5893
#ifdef CONFIG_RT_GROUP_SCHED
5894
		root_task_group.rt_se = (struct sched_rt_entity **)ptr;
5895 5896
		ptr += nr_cpu_ids * sizeof(void **);

5897
		root_task_group.rt_rq = (struct rt_rq **)ptr;
5898 5899
		ptr += nr_cpu_ids * sizeof(void **);

5900
#endif /* CONFIG_RT_GROUP_SCHED */
5901
	}
5902
#ifdef CONFIG_CPUMASK_OFFSTACK
5903 5904 5905
	for_each_possible_cpu(i) {
		per_cpu(load_balance_mask, i) = (cpumask_var_t)kzalloc_node(
			cpumask_size(), GFP_KERNEL, cpu_to_node(i));
5906 5907
		per_cpu(select_idle_mask, i) = (cpumask_var_t)kzalloc_node(
			cpumask_size(), GFP_KERNEL, cpu_to_node(i));
5908
	}
5909
#endif /* CONFIG_CPUMASK_OFFSTACK */
I
Ingo Molnar 已提交
5910

I
Ingo Molnar 已提交
5911 5912
	init_rt_bandwidth(&def_rt_bandwidth, global_rt_period(), global_rt_runtime());
	init_dl_bandwidth(&def_dl_bandwidth, global_rt_period(), global_rt_runtime());
5913

G
Gregory Haskins 已提交
5914 5915 5916 5917
#ifdef CONFIG_SMP
	init_defrootdomain();
#endif

5918
#ifdef CONFIG_RT_GROUP_SCHED
5919
	init_rt_bandwidth(&root_task_group.rt_bandwidth,
5920
			global_rt_period(), global_rt_runtime());
5921
#endif /* CONFIG_RT_GROUP_SCHED */
5922

D
Dhaval Giani 已提交
5923
#ifdef CONFIG_CGROUP_SCHED
5924 5925
	task_group_cache = KMEM_CACHE(task_group, 0);

5926 5927
	list_add(&root_task_group.list, &task_groups);
	INIT_LIST_HEAD(&root_task_group.children);
5928
	INIT_LIST_HEAD(&root_task_group.siblings);
5929
	autogroup_init(&init_task);
D
Dhaval Giani 已提交
5930
#endif /* CONFIG_CGROUP_SCHED */
P
Peter Zijlstra 已提交
5931

5932
	for_each_possible_cpu(i) {
5933
		struct rq *rq;
L
Linus Torvalds 已提交
5934 5935

		rq = cpu_rq(i);
5936
		raw_spin_lock_init(&rq->lock);
N
Nick Piggin 已提交
5937
		rq->nr_running = 0;
5938 5939
		rq->calc_load_active = 0;
		rq->calc_load_update = jiffies + LOAD_FREQ;
5940
		init_cfs_rq(&rq->cfs);
5941 5942
		init_rt_rq(&rq->rt);
		init_dl_rq(&rq->dl);
I
Ingo Molnar 已提交
5943
#ifdef CONFIG_FAIR_GROUP_SCHED
5944
		root_task_group.shares = ROOT_TASK_GROUP_LOAD;
P
Peter Zijlstra 已提交
5945
		INIT_LIST_HEAD(&rq->leaf_cfs_rq_list);
5946
		rq->tmp_alone_branch = &rq->leaf_cfs_rq_list;
D
Dhaval Giani 已提交
5947
		/*
I
Ingo Molnar 已提交
5948
		 * How much CPU bandwidth does root_task_group get?
D
Dhaval Giani 已提交
5949 5950
		 *
		 * In case of task-groups formed thr' the cgroup filesystem, it
I
Ingo Molnar 已提交
5951 5952
		 * gets 100% of the CPU resources in the system. This overall
		 * system CPU resource is divided among the tasks of
5953
		 * root_task_group and its child task-groups in a fair manner,
D
Dhaval Giani 已提交
5954 5955 5956
		 * based on each entity's (task or task-group's) weight
		 * (se->load.weight).
		 *
5957
		 * In other words, if root_task_group has 10 tasks of weight
D
Dhaval Giani 已提交
5958
		 * 1024) and two child groups A0 and A1 (of weight 1024 each),
I
Ingo Molnar 已提交
5959
		 * then A0's share of the CPU resource is:
D
Dhaval Giani 已提交
5960
		 *
5961
		 *	A0's bandwidth = 1024 / (10*1024 + 1024 + 1024) = 8.33%
D
Dhaval Giani 已提交
5962
		 *
5963 5964
		 * 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
5965
		 */
5966
		init_cfs_bandwidth(&root_task_group.cfs_bandwidth);
5967
		init_tg_cfs_entry(&root_task_group, &rq->cfs, NULL, i, NULL);
D
Dhaval Giani 已提交
5968 5969 5970
#endif /* CONFIG_FAIR_GROUP_SCHED */

		rq->rt.rt_runtime = def_rt_bandwidth.rt_runtime;
5971
#ifdef CONFIG_RT_GROUP_SCHED
5972
		init_tg_rt_entry(&root_task_group, &rq->rt, NULL, i, NULL);
I
Ingo Molnar 已提交
5973
#endif
L
Linus Torvalds 已提交
5974

I
Ingo Molnar 已提交
5975 5976
		for (j = 0; j < CPU_LOAD_IDX_MAX; j++)
			rq->cpu_load[j] = 0;
5977

L
Linus Torvalds 已提交
5978
#ifdef CONFIG_SMP
N
Nick Piggin 已提交
5979
		rq->sd = NULL;
G
Gregory Haskins 已提交
5980
		rq->rd = NULL;
5981
		rq->cpu_capacity = rq->cpu_capacity_orig = SCHED_CAPACITY_SCALE;
5982
		rq->balance_callback = NULL;
L
Linus Torvalds 已提交
5983
		rq->active_balance = 0;
I
Ingo Molnar 已提交
5984
		rq->next_balance = jiffies;
L
Linus Torvalds 已提交
5985
		rq->push_cpu = 0;
5986
		rq->cpu = i;
5987
		rq->online = 0;
5988 5989
		rq->idle_stamp = 0;
		rq->avg_idle = 2*sysctl_sched_migration_cost;
5990
		rq->max_idle_balance_cost = sysctl_sched_migration_cost;
5991 5992 5993

		INIT_LIST_HEAD(&rq->cfs_tasks);

5994
		rq_attach_root(rq, &def_root_domain);
5995
#ifdef CONFIG_NO_HZ_COMMON
5996
		rq->last_load_update_tick = jiffies;
5997
		rq->nohz_flags = 0;
5998
#endif
5999 6000 6001
#ifdef CONFIG_NO_HZ_FULL
		rq->last_sched_tick = 0;
#endif
6002
#endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
6003
		init_rq_hrtick(rq);
L
Linus Torvalds 已提交
6004 6005 6006
		atomic_set(&rq->nr_iowait, 0);
	}

6007
	set_load_weight(&init_task, false);
6008

L
Linus Torvalds 已提交
6009 6010 6011
	/*
	 * The boot idle thread does lazy MMU switching as well:
	 */
V
Vegard Nossum 已提交
6012
	mmgrab(&init_mm);
L
Linus Torvalds 已提交
6013 6014 6015 6016 6017 6018 6019 6020 6021
	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());
6022 6023 6024

	calc_load_update = jiffies + LOAD_FREQ;

6025
#ifdef CONFIG_SMP
6026
	idle_thread_set_boot_cpu();
6027
	set_cpu_rq_start_time(smp_processor_id());
6028 6029
#endif
	init_sched_fair_class();
6030

6031 6032
	init_schedstats();

6033
	scheduler_running = 1;
L
Linus Torvalds 已提交
6034 6035
}

6036
#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
6037 6038
static inline int preempt_count_equals(int preempt_offset)
{
6039
	int nested = preempt_count() + rcu_preempt_depth();
6040

A
Arnd Bergmann 已提交
6041
	return (nested == preempt_offset);
6042 6043
}

6044
void __might_sleep(const char *file, int line, int preempt_offset)
L
Linus Torvalds 已提交
6045
{
P
Peter Zijlstra 已提交
6046 6047 6048 6049 6050
	/*
	 * 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.
	 */
6051
	WARN_ONCE(current->state != TASK_RUNNING && current->task_state_change,
P
Peter Zijlstra 已提交
6052 6053 6054 6055
			"do not call blocking ops when !TASK_RUNNING; "
			"state=%lx set at [<%p>] %pS\n",
			current->state,
			(void *)current->task_state_change,
6056
			(void *)current->task_state_change);
P
Peter Zijlstra 已提交
6057

6058 6059 6060 6061 6062
	___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
6063
{
I
Ingo Molnar 已提交
6064 6065 6066
	/* Ratelimiting timestamp: */
	static unsigned long prev_jiffy;

6067
	unsigned long preempt_disable_ip;
L
Linus Torvalds 已提交
6068

I
Ingo Molnar 已提交
6069 6070 6071
	/* WARN_ON_ONCE() by default, no rate limit required: */
	rcu_sleep_check();

6072 6073
	if ((preempt_count_equals(preempt_offset) && !irqs_disabled() &&
	     !is_idle_task(current)) ||
6074 6075
	    system_state == SYSTEM_BOOTING || system_state > SYSTEM_RUNNING ||
	    oops_in_progress)
I
Ingo Molnar 已提交
6076
		return;
6077

I
Ingo Molnar 已提交
6078 6079 6080 6081
	if (time_before(jiffies, prev_jiffy + HZ) && prev_jiffy)
		return;
	prev_jiffy = jiffies;

I
Ingo Molnar 已提交
6082
	/* Save this before calling printk(), since that will clobber it: */
6083 6084
	preempt_disable_ip = get_preempt_disable_ip(current);

P
Peter Zijlstra 已提交
6085 6086 6087 6088 6089 6090 6091
	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
6092

6093 6094 6095
	if (task_stack_end_corrupted(current))
		printk(KERN_EMERG "Thread overran stack, or stack corrupted\n");

I
Ingo Molnar 已提交
6096 6097 6098
	debug_show_held_locks(current);
	if (irqs_disabled())
		print_irqtrace_events(current);
6099 6100
	if (IS_ENABLED(CONFIG_DEBUG_PREEMPT)
	    && !preempt_count_equals(preempt_offset)) {
6101
		pr_err("Preemption disabled at:");
6102
		print_ip_sym(preempt_disable_ip);
6103 6104
		pr_cont("\n");
	}
I
Ingo Molnar 已提交
6105
	dump_stack();
6106
	add_taint(TAINT_WARN, LOCKDEP_STILL_OK);
L
Linus Torvalds 已提交
6107
}
6108
EXPORT_SYMBOL(___might_sleep);
L
Linus Torvalds 已提交
6109 6110 6111
#endif

#ifdef CONFIG_MAGIC_SYSRQ
6112
void normalize_rt_tasks(void)
6113
{
6114
	struct task_struct *g, *p;
6115 6116 6117
	struct sched_attr attr = {
		.sched_policy = SCHED_NORMAL,
	};
L
Linus Torvalds 已提交
6118

6119
	read_lock(&tasklist_lock);
6120
	for_each_process_thread(g, p) {
6121 6122 6123
		/*
		 * Only normalize user tasks:
		 */
6124
		if (p->flags & PF_KTHREAD)
6125 6126
			continue;

6127 6128 6129 6130
		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
6131

6132
		if (!dl_task(p) && !rt_task(p)) {
I
Ingo Molnar 已提交
6133 6134 6135 6136
			/*
			 * Renice negative nice level userspace
			 * tasks back to 0:
			 */
6137
			if (task_nice(p) < 0)
I
Ingo Molnar 已提交
6138
				set_user_nice(p, 0);
L
Linus Torvalds 已提交
6139
			continue;
I
Ingo Molnar 已提交
6140
		}
L
Linus Torvalds 已提交
6141

6142
		__sched_setscheduler(p, &attr, false, false);
6143
	}
6144
	read_unlock(&tasklist_lock);
L
Linus Torvalds 已提交
6145 6146 6147
}

#endif /* CONFIG_MAGIC_SYSRQ */
6148

6149
#if defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB)
6150
/*
6151
 * These functions are only useful for the IA64 MCA handling, or kdb.
6152 6153 6154 6155 6156 6157 6158 6159 6160
 *
 * 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
6161
 * curr_task - return the current task for a given CPU.
6162 6163 6164
 * @cpu: the processor in question.
 *
 * ONLY VALID WHEN THE WHOLE SYSTEM IS STOPPED!
6165 6166
 *
 * Return: The current task for @cpu.
6167
 */
6168
struct task_struct *curr_task(int cpu)
6169 6170 6171 6172
{
	return cpu_curr(cpu);
}

6173 6174 6175
#endif /* defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB) */

#ifdef CONFIG_IA64
6176
/**
I
Ingo Molnar 已提交
6177
 * set_curr_task - set the current task for a given CPU.
6178 6179 6180 6181
 * @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
6182
 * are serviced on a separate stack. It allows the architecture to switch the
I
Ingo Molnar 已提交
6183
 * notion of the current task on a CPU in a non-blocking manner. This function
6184 6185 6186 6187 6188 6189 6190
 * 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!
 */
6191
void ia64_set_curr_task(int cpu, struct task_struct *p)
6192 6193 6194 6195 6196
{
	cpu_curr(cpu) = p;
}

#endif
S
Srivatsa Vaddagiri 已提交
6197

D
Dhaval Giani 已提交
6198
#ifdef CONFIG_CGROUP_SCHED
6199 6200 6201
/* task_group_lock serializes the addition/removal of task groups */
static DEFINE_SPINLOCK(task_group_lock);

6202
static void sched_free_group(struct task_group *tg)
6203 6204 6205
{
	free_fair_sched_group(tg);
	free_rt_sched_group(tg);
6206
	autogroup_free(tg);
6207
	kmem_cache_free(task_group_cache, tg);
6208 6209 6210
}

/* allocate runqueue etc for a new task group */
6211
struct task_group *sched_create_group(struct task_group *parent)
6212 6213 6214
{
	struct task_group *tg;

6215
	tg = kmem_cache_alloc(task_group_cache, GFP_KERNEL | __GFP_ZERO);
6216 6217 6218
	if (!tg)
		return ERR_PTR(-ENOMEM);

6219
	if (!alloc_fair_sched_group(tg, parent))
6220 6221
		goto err;

6222
	if (!alloc_rt_sched_group(tg, parent))
6223 6224
		goto err;

6225 6226 6227
	return tg;

err:
6228
	sched_free_group(tg);
6229 6230 6231 6232 6233 6234 6235
	return ERR_PTR(-ENOMEM);
}

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

6236
	spin_lock_irqsave(&task_group_lock, flags);
P
Peter Zijlstra 已提交
6237
	list_add_rcu(&tg->list, &task_groups);
P
Peter Zijlstra 已提交
6238

I
Ingo Molnar 已提交
6239 6240
	/* Root should already exist: */
	WARN_ON(!parent);
P
Peter Zijlstra 已提交
6241 6242 6243

	tg->parent = parent;
	INIT_LIST_HEAD(&tg->children);
6244
	list_add_rcu(&tg->siblings, &parent->children);
6245
	spin_unlock_irqrestore(&task_group_lock, flags);
6246 6247

	online_fair_sched_group(tg);
S
Srivatsa Vaddagiri 已提交
6248 6249
}

6250
/* rcu callback to free various structures associated with a task group */
6251
static void sched_free_group_rcu(struct rcu_head *rhp)
S
Srivatsa Vaddagiri 已提交
6252
{
I
Ingo Molnar 已提交
6253
	/* Now it should be safe to free those cfs_rqs: */
6254
	sched_free_group(container_of(rhp, struct task_group, rcu));
S
Srivatsa Vaddagiri 已提交
6255 6256
}

6257
void sched_destroy_group(struct task_group *tg)
6258
{
I
Ingo Molnar 已提交
6259
	/* Wait for possible concurrent references to cfs_rqs complete: */
6260
	call_rcu(&tg->rcu, sched_free_group_rcu);
6261 6262 6263
}

void sched_offline_group(struct task_group *tg)
S
Srivatsa Vaddagiri 已提交
6264
{
6265
	unsigned long flags;
S
Srivatsa Vaddagiri 已提交
6266

I
Ingo Molnar 已提交
6267
	/* End participation in shares distribution: */
6268
	unregister_fair_sched_group(tg);
6269 6270

	spin_lock_irqsave(&task_group_lock, flags);
P
Peter Zijlstra 已提交
6271
	list_del_rcu(&tg->list);
P
Peter Zijlstra 已提交
6272
	list_del_rcu(&tg->siblings);
6273
	spin_unlock_irqrestore(&task_group_lock, flags);
S
Srivatsa Vaddagiri 已提交
6274 6275
}

6276
static void sched_change_group(struct task_struct *tsk, int type)
S
Srivatsa Vaddagiri 已提交
6277
{
P
Peter Zijlstra 已提交
6278
	struct task_group *tg;
S
Srivatsa Vaddagiri 已提交
6279

6280 6281 6282 6283 6284 6285
	/*
	 * 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
6286 6287 6288 6289
			  struct task_group, css);
	tg = autogroup_task_group(tsk, tg);
	tsk->sched_task_group = tg;

P
Peter Zijlstra 已提交
6290
#ifdef CONFIG_FAIR_GROUP_SCHED
6291 6292
	if (tsk->sched_class->task_change_group)
		tsk->sched_class->task_change_group(tsk, type);
6293
	else
P
Peter Zijlstra 已提交
6294
#endif
6295
		set_task_rq(tsk, task_cpu(tsk));
6296 6297 6298 6299 6300 6301 6302 6303 6304 6305 6306
}

/*
 * 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)
{
6307 6308
	int queued, running, queue_flags =
		DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
6309 6310 6311 6312
	struct rq_flags rf;
	struct rq *rq;

	rq = task_rq_lock(tsk, &rf);
6313
	update_rq_clock(rq);
6314 6315 6316 6317 6318

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

	if (queued)
6319
		dequeue_task(rq, tsk, queue_flags);
6320
	if (running)
6321 6322 6323
		put_prev_task(rq, tsk);

	sched_change_group(tsk, TASK_MOVE_GROUP);
P
Peter Zijlstra 已提交
6324

6325
	if (queued)
6326
		enqueue_task(rq, tsk, queue_flags);
6327
	if (running)
6328
		set_curr_task(rq, tsk);
S
Srivatsa Vaddagiri 已提交
6329

6330
	task_rq_unlock(rq, tsk, &rf);
S
Srivatsa Vaddagiri 已提交
6331
}
6332

6333
static inline struct task_group *css_tg(struct cgroup_subsys_state *css)
6334
{
6335
	return css ? container_of(css, struct task_group, css) : NULL;
6336 6337
}

6338 6339
static struct cgroup_subsys_state *
cpu_cgroup_css_alloc(struct cgroup_subsys_state *parent_css)
6340
{
6341 6342
	struct task_group *parent = css_tg(parent_css);
	struct task_group *tg;
6343

6344
	if (!parent) {
6345
		/* This is early initialization for the top cgroup */
6346
		return &root_task_group.css;
6347 6348
	}

6349
	tg = sched_create_group(parent);
6350 6351 6352 6353 6354 6355
	if (IS_ERR(tg))
		return ERR_PTR(-ENOMEM);

	return &tg->css;
}

6356 6357 6358 6359 6360 6361 6362 6363 6364 6365 6366
/* 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;
}

6367
static void cpu_cgroup_css_released(struct cgroup_subsys_state *css)
6368
{
6369
	struct task_group *tg = css_tg(css);
6370

6371
	sched_offline_group(tg);
6372 6373
}

6374
static void cpu_cgroup_css_free(struct cgroup_subsys_state *css)
6375
{
6376
	struct task_group *tg = css_tg(css);
6377

6378 6379 6380 6381
	/*
	 * Relies on the RCU grace period between css_released() and this.
	 */
	sched_free_group(tg);
6382 6383
}

6384 6385 6386 6387
/*
 * 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.
 */
6388
static void cpu_cgroup_fork(struct task_struct *task)
6389
{
6390 6391 6392 6393 6394
	struct rq_flags rf;
	struct rq *rq;

	rq = task_rq_lock(task, &rf);

6395
	update_rq_clock(rq);
6396 6397 6398
	sched_change_group(task, TASK_SET_GROUP);

	task_rq_unlock(rq, task, &rf);
6399 6400
}

6401
static int cpu_cgroup_can_attach(struct cgroup_taskset *tset)
6402
{
6403
	struct task_struct *task;
6404
	struct cgroup_subsys_state *css;
6405
	int ret = 0;
6406

6407
	cgroup_taskset_for_each(task, css, tset) {
6408
#ifdef CONFIG_RT_GROUP_SCHED
6409
		if (!sched_rt_can_attach(css_tg(css), task))
6410
			return -EINVAL;
6411
#else
6412 6413 6414
		/* We don't support RT-tasks being in separate groups */
		if (task->sched_class != &fair_sched_class)
			return -EINVAL;
6415
#endif
6416 6417 6418 6419 6420 6421 6422 6423 6424 6425 6426 6427 6428 6429 6430 6431
		/*
		 * 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;
6432
	}
6433
	return ret;
6434
}
6435

6436
static void cpu_cgroup_attach(struct cgroup_taskset *tset)
6437
{
6438
	struct task_struct *task;
6439
	struct cgroup_subsys_state *css;
6440

6441
	cgroup_taskset_for_each(task, css, tset)
6442
		sched_move_task(task);
6443 6444
}

6445
#ifdef CONFIG_FAIR_GROUP_SCHED
6446 6447
static int cpu_shares_write_u64(struct cgroup_subsys_state *css,
				struct cftype *cftype, u64 shareval)
6448
{
6449
	return sched_group_set_shares(css_tg(css), scale_load(shareval));
6450 6451
}

6452 6453
static u64 cpu_shares_read_u64(struct cgroup_subsys_state *css,
			       struct cftype *cft)
6454
{
6455
	struct task_group *tg = css_tg(css);
6456

6457
	return (u64) scale_load_down(tg->shares);
6458
}
6459 6460

#ifdef CONFIG_CFS_BANDWIDTH
6461 6462
static DEFINE_MUTEX(cfs_constraints_mutex);

6463 6464 6465
const u64 max_cfs_quota_period = 1 * NSEC_PER_SEC; /* 1s */
const u64 min_cfs_quota_period = 1 * NSEC_PER_MSEC; /* 1ms */

6466 6467
static int __cfs_schedulable(struct task_group *tg, u64 period, u64 runtime);

6468 6469
static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota)
{
6470
	int i, ret = 0, runtime_enabled, runtime_was_enabled;
6471
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
6472 6473 6474 6475 6476 6477 6478 6479 6480 6481 6482 6483 6484 6485 6486 6487 6488 6489 6490 6491

	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;

6492 6493 6494 6495 6496
	/*
	 * Prevent race between setting of cfs_rq->runtime_enabled and
	 * unthrottle_offline_cfs_rqs().
	 */
	get_online_cpus();
6497 6498 6499 6500 6501
	mutex_lock(&cfs_constraints_mutex);
	ret = __cfs_schedulable(tg, period, quota);
	if (ret)
		goto out_unlock;

6502
	runtime_enabled = quota != RUNTIME_INF;
6503
	runtime_was_enabled = cfs_b->quota != RUNTIME_INF;
6504 6505 6506 6507 6508 6509
	/*
	 * 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();
6510 6511 6512
	raw_spin_lock_irq(&cfs_b->lock);
	cfs_b->period = ns_to_ktime(period);
	cfs_b->quota = quota;
6513

P
Paul Turner 已提交
6514
	__refill_cfs_bandwidth_runtime(cfs_b);
I
Ingo Molnar 已提交
6515 6516

	/* Restart the period timer (if active) to handle new period expiry: */
P
Peter Zijlstra 已提交
6517 6518
	if (runtime_enabled)
		start_cfs_bandwidth(cfs_b);
I
Ingo Molnar 已提交
6519

6520 6521
	raw_spin_unlock_irq(&cfs_b->lock);

6522
	for_each_online_cpu(i) {
6523
		struct cfs_rq *cfs_rq = tg->cfs_rq[i];
6524
		struct rq *rq = cfs_rq->rq;
6525
		struct rq_flags rf;
6526

6527
		rq_lock_irq(rq, &rf);
6528
		cfs_rq->runtime_enabled = runtime_enabled;
6529
		cfs_rq->runtime_remaining = 0;
6530

6531
		if (cfs_rq->throttled)
6532
			unthrottle_cfs_rq(cfs_rq);
6533
		rq_unlock_irq(rq, &rf);
6534
	}
6535 6536
	if (runtime_was_enabled && !runtime_enabled)
		cfs_bandwidth_usage_dec();
6537 6538
out_unlock:
	mutex_unlock(&cfs_constraints_mutex);
6539
	put_online_cpus();
6540

6541
	return ret;
6542 6543 6544 6545 6546 6547
}

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

6548
	period = ktime_to_ns(tg->cfs_bandwidth.period);
6549 6550 6551 6552 6553 6554 6555 6556 6557 6558 6559 6560
	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;

6561
	if (tg->cfs_bandwidth.quota == RUNTIME_INF)
6562 6563
		return -1;

6564
	quota_us = tg->cfs_bandwidth.quota;
6565 6566 6567 6568 6569 6570 6571 6572 6573 6574
	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;
6575
	quota = tg->cfs_bandwidth.quota;
6576 6577 6578 6579 6580 6581 6582 6583

	return tg_set_cfs_bandwidth(tg, period, quota);
}

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

6584
	cfs_period_us = ktime_to_ns(tg->cfs_bandwidth.period);
6585 6586 6587 6588 6589
	do_div(cfs_period_us, NSEC_PER_USEC);

	return cfs_period_us;
}

6590 6591
static s64 cpu_cfs_quota_read_s64(struct cgroup_subsys_state *css,
				  struct cftype *cft)
6592
{
6593
	return tg_get_cfs_quota(css_tg(css));
6594 6595
}

6596 6597
static int cpu_cfs_quota_write_s64(struct cgroup_subsys_state *css,
				   struct cftype *cftype, s64 cfs_quota_us)
6598
{
6599
	return tg_set_cfs_quota(css_tg(css), cfs_quota_us);
6600 6601
}

6602 6603
static u64 cpu_cfs_period_read_u64(struct cgroup_subsys_state *css,
				   struct cftype *cft)
6604
{
6605
	return tg_get_cfs_period(css_tg(css));
6606 6607
}

6608 6609
static int cpu_cfs_period_write_u64(struct cgroup_subsys_state *css,
				    struct cftype *cftype, u64 cfs_period_us)
6610
{
6611
	return tg_set_cfs_period(css_tg(css), cfs_period_us);
6612 6613
}

6614 6615 6616 6617 6618 6619 6620 6621 6622 6623 6624 6625 6626 6627 6628 6629 6630 6631 6632 6633 6634 6635 6636 6637 6638 6639 6640 6641 6642 6643 6644 6645
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;
6646
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
6647 6648 6649 6650 6651
	s64 quota = 0, parent_quota = -1;

	if (!tg->parent) {
		quota = RUNTIME_INF;
	} else {
6652
		struct cfs_bandwidth *parent_b = &tg->parent->cfs_bandwidth;
6653 6654

		quota = normalize_cfs_quota(tg, d);
6655
		parent_quota = parent_b->hierarchical_quota;
6656 6657

		/*
I
Ingo Molnar 已提交
6658 6659
		 * Ensure max(child_quota) <= parent_quota, inherit when no
		 * limit is set:
6660 6661 6662 6663 6664 6665
		 */
		if (quota == RUNTIME_INF)
			quota = parent_quota;
		else if (parent_quota != RUNTIME_INF && quota > parent_quota)
			return -EINVAL;
	}
6666
	cfs_b->hierarchical_quota = quota;
6667 6668 6669 6670 6671 6672

	return 0;
}

static int __cfs_schedulable(struct task_group *tg, u64 period, u64 quota)
{
6673
	int ret;
6674 6675 6676 6677 6678 6679 6680 6681 6682 6683 6684
	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);
	}

6685 6686 6687 6688 6689
	rcu_read_lock();
	ret = walk_tg_tree(tg_cfs_schedulable_down, tg_nop, &data);
	rcu_read_unlock();

	return ret;
6690
}
6691

6692
static int cpu_cfs_stat_show(struct seq_file *sf, void *v)
6693
{
6694
	struct task_group *tg = css_tg(seq_css(sf));
6695
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
6696

6697 6698 6699
	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);
6700 6701 6702

	return 0;
}
6703
#endif /* CONFIG_CFS_BANDWIDTH */
6704
#endif /* CONFIG_FAIR_GROUP_SCHED */
6705

6706
#ifdef CONFIG_RT_GROUP_SCHED
6707 6708
static int cpu_rt_runtime_write(struct cgroup_subsys_state *css,
				struct cftype *cft, s64 val)
P
Peter Zijlstra 已提交
6709
{
6710
	return sched_group_set_rt_runtime(css_tg(css), val);
P
Peter Zijlstra 已提交
6711 6712
}

6713 6714
static s64 cpu_rt_runtime_read(struct cgroup_subsys_state *css,
			       struct cftype *cft)
P
Peter Zijlstra 已提交
6715
{
6716
	return sched_group_rt_runtime(css_tg(css));
P
Peter Zijlstra 已提交
6717
}
6718

6719 6720
static int cpu_rt_period_write_uint(struct cgroup_subsys_state *css,
				    struct cftype *cftype, u64 rt_period_us)
6721
{
6722
	return sched_group_set_rt_period(css_tg(css), rt_period_us);
6723 6724
}

6725 6726
static u64 cpu_rt_period_read_uint(struct cgroup_subsys_state *css,
				   struct cftype *cft)
6727
{
6728
	return sched_group_rt_period(css_tg(css));
6729
}
6730
#endif /* CONFIG_RT_GROUP_SCHED */
P
Peter Zijlstra 已提交
6731

6732
static struct cftype cpu_legacy_files[] = {
6733
#ifdef CONFIG_FAIR_GROUP_SCHED
6734 6735
	{
		.name = "shares",
6736 6737
		.read_u64 = cpu_shares_read_u64,
		.write_u64 = cpu_shares_write_u64,
6738
	},
6739
#endif
6740 6741 6742 6743 6744 6745 6746 6747 6748 6749 6750
#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,
	},
6751 6752
	{
		.name = "stat",
6753
		.seq_show = cpu_cfs_stat_show,
6754
	},
6755
#endif
6756
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
6757
	{
P
Peter Zijlstra 已提交
6758
		.name = "rt_runtime_us",
6759 6760
		.read_s64 = cpu_rt_runtime_read,
		.write_s64 = cpu_rt_runtime_write,
P
Peter Zijlstra 已提交
6761
	},
6762 6763
	{
		.name = "rt_period_us",
6764 6765
		.read_u64 = cpu_rt_period_read_uint,
		.write_u64 = cpu_rt_period_write_uint,
6766
	},
6767
#endif
I
Ingo Molnar 已提交
6768
	{ }	/* Terminate */
6769 6770
};

6771 6772
static int cpu_extra_stat_show(struct seq_file *sf,
			       struct cgroup_subsys_state *css)
6773 6774 6775
{
#ifdef CONFIG_CFS_BANDWIDTH
	{
6776
		struct task_group *tg = css_tg(css);
6777 6778 6779 6780 6781 6782 6783 6784 6785 6786 6787 6788 6789 6790 6791 6792 6793 6794 6795 6796 6797 6798 6799 6800 6801 6802 6803 6804 6805 6806 6807 6808 6809 6810 6811 6812 6813 6814 6815 6816 6817 6818 6819 6820 6821 6822 6823 6824 6825 6826 6827 6828 6829 6830 6831 6832 6833 6834 6835 6836 6837 6838 6839 6840 6841 6842 6843 6844 6845 6846 6847 6848 6849 6850 6851 6852 6853 6854 6855 6856 6857 6858 6859 6860 6861 6862 6863 6864 6865 6866 6867 6868 6869 6870 6871 6872 6873 6874 6875 6876 6877 6878 6879 6880 6881 6882 6883 6884 6885 6886 6887 6888 6889 6890 6891 6892 6893 6894 6895 6896 6897 6898 6899 6900 6901 6902 6903 6904 6905 6906 6907 6908 6909 6910 6911 6912 6913 6914 6915 6916 6917 6918 6919 6920 6921 6922 6923 6924 6925 6926 6927 6928 6929 6930 6931 6932 6933
		struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
		u64 throttled_usec;

		throttled_usec = cfs_b->throttled_time;
		do_div(throttled_usec, NSEC_PER_USEC);

		seq_printf(sf, "nr_periods %d\n"
			   "nr_throttled %d\n"
			   "throttled_usec %llu\n",
			   cfs_b->nr_periods, cfs_b->nr_throttled,
			   throttled_usec);
	}
#endif
	return 0;
}

#ifdef CONFIG_FAIR_GROUP_SCHED
static u64 cpu_weight_read_u64(struct cgroup_subsys_state *css,
			       struct cftype *cft)
{
	struct task_group *tg = css_tg(css);
	u64 weight = scale_load_down(tg->shares);

	return DIV_ROUND_CLOSEST_ULL(weight * CGROUP_WEIGHT_DFL, 1024);
}

static int cpu_weight_write_u64(struct cgroup_subsys_state *css,
				struct cftype *cft, u64 weight)
{
	/*
	 * cgroup weight knobs should use the common MIN, DFL and MAX
	 * values which are 1, 100 and 10000 respectively.  While it loses
	 * a bit of range on both ends, it maps pretty well onto the shares
	 * value used by scheduler and the round-trip conversions preserve
	 * the original value over the entire range.
	 */
	if (weight < CGROUP_WEIGHT_MIN || weight > CGROUP_WEIGHT_MAX)
		return -ERANGE;

	weight = DIV_ROUND_CLOSEST_ULL(weight * 1024, CGROUP_WEIGHT_DFL);

	return sched_group_set_shares(css_tg(css), scale_load(weight));
}

static s64 cpu_weight_nice_read_s64(struct cgroup_subsys_state *css,
				    struct cftype *cft)
{
	unsigned long weight = scale_load_down(css_tg(css)->shares);
	int last_delta = INT_MAX;
	int prio, delta;

	/* find the closest nice value to the current weight */
	for (prio = 0; prio < ARRAY_SIZE(sched_prio_to_weight); prio++) {
		delta = abs(sched_prio_to_weight[prio] - weight);
		if (delta >= last_delta)
			break;
		last_delta = delta;
	}

	return PRIO_TO_NICE(prio - 1 + MAX_RT_PRIO);
}

static int cpu_weight_nice_write_s64(struct cgroup_subsys_state *css,
				     struct cftype *cft, s64 nice)
{
	unsigned long weight;

	if (nice < MIN_NICE || nice > MAX_NICE)
		return -ERANGE;

	weight = sched_prio_to_weight[NICE_TO_PRIO(nice) - MAX_RT_PRIO];
	return sched_group_set_shares(css_tg(css), scale_load(weight));
}
#endif

static void __maybe_unused cpu_period_quota_print(struct seq_file *sf,
						  long period, long quota)
{
	if (quota < 0)
		seq_puts(sf, "max");
	else
		seq_printf(sf, "%ld", quota);

	seq_printf(sf, " %ld\n", period);
}

/* caller should put the current value in *@periodp before calling */
static int __maybe_unused cpu_period_quota_parse(char *buf,
						 u64 *periodp, u64 *quotap)
{
	char tok[21];	/* U64_MAX */

	if (!sscanf(buf, "%s %llu", tok, periodp))
		return -EINVAL;

	*periodp *= NSEC_PER_USEC;

	if (sscanf(tok, "%llu", quotap))
		*quotap *= NSEC_PER_USEC;
	else if (!strcmp(tok, "max"))
		*quotap = RUNTIME_INF;
	else
		return -EINVAL;

	return 0;
}

#ifdef CONFIG_CFS_BANDWIDTH
static int cpu_max_show(struct seq_file *sf, void *v)
{
	struct task_group *tg = css_tg(seq_css(sf));

	cpu_period_quota_print(sf, tg_get_cfs_period(tg), tg_get_cfs_quota(tg));
	return 0;
}

static ssize_t cpu_max_write(struct kernfs_open_file *of,
			     char *buf, size_t nbytes, loff_t off)
{
	struct task_group *tg = css_tg(of_css(of));
	u64 period = tg_get_cfs_period(tg);
	u64 quota;
	int ret;

	ret = cpu_period_quota_parse(buf, &period, &quota);
	if (!ret)
		ret = tg_set_cfs_bandwidth(tg, period, quota);
	return ret ?: nbytes;
}
#endif

static struct cftype cpu_files[] = {
#ifdef CONFIG_FAIR_GROUP_SCHED
	{
		.name = "weight",
		.flags = CFTYPE_NOT_ON_ROOT,
		.read_u64 = cpu_weight_read_u64,
		.write_u64 = cpu_weight_write_u64,
	},
	{
		.name = "weight.nice",
		.flags = CFTYPE_NOT_ON_ROOT,
		.read_s64 = cpu_weight_nice_read_s64,
		.write_s64 = cpu_weight_nice_write_s64,
	},
#endif
#ifdef CONFIG_CFS_BANDWIDTH
	{
		.name = "max",
		.flags = CFTYPE_NOT_ON_ROOT,
		.seq_show = cpu_max_show,
		.write = cpu_max_write,
	},
#endif
	{ }	/* terminate */
};

6934
struct cgroup_subsys cpu_cgrp_subsys = {
6935
	.css_alloc	= cpu_cgroup_css_alloc,
6936
	.css_online	= cpu_cgroup_css_online,
6937
	.css_released	= cpu_cgroup_css_released,
6938
	.css_free	= cpu_cgroup_css_free,
6939
	.css_extra_stat_show = cpu_extra_stat_show,
6940
	.fork		= cpu_cgroup_fork,
6941 6942
	.can_attach	= cpu_cgroup_can_attach,
	.attach		= cpu_cgroup_attach,
6943
	.legacy_cftypes	= cpu_legacy_files,
6944
	.dfl_cftypes	= cpu_files,
6945
	.early_init	= true,
6946
	.threaded	= true,
6947 6948
};

6949
#endif	/* CONFIG_CGROUP_SCHED */
6950

6951 6952 6953 6954 6955
void dump_cpu_task(int cpu)
{
	pr_info("Task dump for CPU %d:\n", cpu);
	sched_show_task(cpu_curr(cpu));
}
6956 6957 6958 6959 6960 6961 6962 6963 6964 6965 6966 6967 6968 6969 6970 6971 6972 6973 6974 6975 6976 6977 6978 6979 6980 6981 6982 6983 6984 6985 6986 6987 6988 6989 6990 6991 6992 6993 6994 6995 6996

/*
 * 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,
};