core.c 174 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1
/*
2
 *  kernel/sched/core.c
Linus Torvalds's avatar
Linus Torvalds committed
3
 *
4
 *  Core kernel scheduler code and related syscalls
Linus Torvalds's avatar
Linus Torvalds committed
5 6 7
 *
 *  Copyright (C) 1991-2002  Linus Torvalds
 */
8
#include "sched.h"
Linus Torvalds's avatar
Linus Torvalds committed
9

10
#include <linux/nospec.h>
11

12 13
#include <linux/kcov.h>

14
#include <asm/switch_to.h>
15
#include <asm/tlb.h>
Linus Torvalds's avatar
Linus Torvalds committed
16

17
#include "../workqueue_internal.h"
18
#include "../smpboot.h"
19

20 21
#include "pelt.h"

22
#define CREATE_TRACE_POINTS
23
#include <trace/events/sched.h>
24

25
DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
26

27
#if defined(CONFIG_SCHED_DEBUG) && defined(CONFIG_JUMP_LABEL)
28 29
/*
 * Debugging: various feature bits
30 31 32 33
 *
 * 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.
34
 */
35 36
#define SCHED_FEAT(name, enabled)	\
	(1UL << __SCHED_FEAT_##name) * enabled |
37
const_debug unsigned int sysctl_sched_features =
38
#include "features.h"
39 40
	0;
#undef SCHED_FEAT
41
#endif
42

43 44 45 46 47 48
/*
 * 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;

Peter Zijlstra's avatar
Peter Zijlstra committed
49
/*
50
 * period over which we measure -rt task CPU usage in us.
Peter Zijlstra's avatar
Peter Zijlstra committed
51 52
 * default: 1s
 */
Peter Zijlstra's avatar
Peter Zijlstra committed
53
unsigned int sysctl_sched_rt_period = 1000000;
Peter Zijlstra's avatar
Peter Zijlstra committed
54

55
__read_mostly int scheduler_running;
56

Peter Zijlstra's avatar
Peter Zijlstra committed
57 58 59 60 61
/*
 * part of the period that we allow rt tasks to run in us.
 * default: 0.95s
 */
int sysctl_sched_rt_runtime = 950000;
Peter Zijlstra's avatar
Peter Zijlstra committed
62

63 64 65
/*
 * __task_rq_lock - lock the rq @p resides on.
 */
66
struct rq *__task_rq_lock(struct task_struct *p, struct rq_flags *rf)
67 68 69 70 71 72 73 74 75 76
	__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))) {
77
			rq_pin_lock(rq, rf);
78 79 80 81 82 83 84 85 86 87 88 89
			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.
 */
90
struct rq *task_rq_lock(struct task_struct *p, struct rq_flags *rf)
91 92 93 94 95 96
	__acquires(p->pi_lock)
	__acquires(rq->lock)
{
	struct rq *rq;

	for (;;) {
97
		raw_spin_lock_irqsave(&p->pi_lock, rf->flags);
98 99 100 101 102 103 104 105 106 107 108 109
		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)
		 *
110
		 * If we observe the old CPU in task_rq_lock(), the acquire of
111 112
		 * the old rq->lock will fully serialize against the stores.
		 *
113 114 115
		 * If we observe the new CPU in task_rq_lock(), the address
		 * dependency headed by '[L] rq = task_rq()' and the acquire
		 * will pair with the WMB to ensure we then also see migrating.
116 117
		 */
		if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) {
118
			rq_pin_lock(rq, rf);
119 120 121
			return rq;
		}
		raw_spin_unlock(&rq->lock);
122
		raw_spin_unlock_irqrestore(&p->pi_lock, rf->flags);
123 124 125 126 127 128

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

129 130 131 132 133 134 135 136 137 138
/*
 * 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...
 */
139 140
	s64 __maybe_unused steal = 0, irq_delta = 0;

141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179
#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;

180
#ifdef CONFIG_HAVE_SCHED_AVG_IRQ
181
	if ((irq_delta + steal) && sched_feat(NONTASK_CAPACITY))
182
		update_irq_load_avg(rq, irq_delta + steal);
183
#endif
184
	update_rq_clock_pelt(rq, delta);
185 186 187 188 189 190 191 192 193 194 195 196
}

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
197 198
	if (sched_feat(WARN_DOUBLE_CLOCK))
		SCHED_WARN_ON(rq->clock_update_flags & RQCF_UPDATED);
199 200
	rq->clock_update_flags |= RQCF_UPDATED;
#endif
201

202 203 204 205 206 207 208 209
	delta = sched_clock_cpu(cpu_of(rq)) - rq->clock;
	if (delta < 0)
		return;
	rq->clock += delta;
	update_rq_clock_task(rq, delta);
}


210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227
#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);
228
	struct rq_flags rf;
229 230 231

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

232
	rq_lock(rq, &rf);
233
	update_rq_clock(rq);
234
	rq->curr->sched_class->task_tick(rq, rq->curr, 1);
235
	rq_unlock(rq, &rf);
236 237 238 239

	return HRTIMER_NORESTART;
}

240
#ifdef CONFIG_SMP
Peter Zijlstra's avatar
Peter Zijlstra committed
241

242
static void __hrtick_restart(struct rq *rq)
Peter Zijlstra's avatar
Peter Zijlstra committed
243 244 245
{
	struct hrtimer *timer = &rq->hrtick_timer;

246
	hrtimer_start_expires(timer, HRTIMER_MODE_ABS_PINNED);
Peter Zijlstra's avatar
Peter Zijlstra committed
247 248
}

249 250 251 252
/*
 * called from hardirq (IPI) context
 */
static void __hrtick_start(void *arg)
253
{
254
	struct rq *rq = arg;
255
	struct rq_flags rf;
256

257
	rq_lock(rq, &rf);
Peter Zijlstra's avatar
Peter Zijlstra committed
258
	__hrtick_restart(rq);
259
	rq->hrtick_csd_pending = 0;
260
	rq_unlock(rq, &rf);
261 262
}

263 264 265 266 267
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
268
void hrtick_start(struct rq *rq, u64 delay)
269
{
270
	struct hrtimer *timer = &rq->hrtick_timer;
271 272 273 274 275 276 277 278 279
	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);
280

281
	hrtimer_set_expires(timer, time);
282 283

	if (rq == this_rq()) {
Peter Zijlstra's avatar
Peter Zijlstra committed
284
		__hrtick_restart(rq);
285
	} else if (!rq->hrtick_csd_pending) {
286
		smp_call_function_single_async(cpu_of(rq), &rq->hrtick_csd);
287 288
		rq->hrtick_csd_pending = 1;
	}
289 290
}

291 292 293 294 295 296
#else
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
297
void hrtick_start(struct rq *rq, u64 delay)
298
{
299 300 301 302 303
	/*
	 * Don't schedule slices shorter than 10000ns, that just
	 * doesn't make sense. Rely on vruntime for fairness.
	 */
	delay = max_t(u64, delay, 10000LL);
304 305
	hrtimer_start(&rq->hrtick_timer, ns_to_ktime(delay),
		      HRTIMER_MODE_REL_PINNED);
306 307
}
#endif /* CONFIG_SMP */
308

309
static void hrtick_rq_init(struct rq *rq)
310
{
311 312
#ifdef CONFIG_SMP
	rq->hrtick_csd_pending = 0;
313

314 315 316 317
	rq->hrtick_csd.flags = 0;
	rq->hrtick_csd.func = __hrtick_start;
	rq->hrtick_csd.info = rq;
#endif
318

319 320
	hrtimer_init(&rq->hrtick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	rq->hrtick_timer.function = hrtick;
321
}
Andrew Morton's avatar
Andrew Morton committed
322
#else	/* CONFIG_SCHED_HRTICK */
323 324 325 326
static inline void hrtick_clear(struct rq *rq)
{
}

327
static inline void hrtick_rq_init(struct rq *rq)
328 329
{
}
Andrew Morton's avatar
Andrew Morton committed
330
#endif	/* CONFIG_SCHED_HRTICK */
331

332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349
/*
 * 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;								\
})

350
#if defined(CONFIG_SMP) && defined(TIF_POLLING_NRFLAG)
351 352 353 354 355 356 357 358 359 360
/*
 * 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);
}
361 362 363 364 365 366 367 368 369 370

/*
 * 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);
371
	typeof(ti->flags) old, val = READ_ONCE(ti->flags);
372 373 374 375 376 377 378 379 380 381 382 383 384 385

	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;
}

386 387 388 389 390 391
#else
static bool set_nr_and_not_polling(struct task_struct *p)
{
	set_tsk_need_resched(p);
	return true;
}
392 393 394 395 396 397 398

#ifdef CONFIG_SMP
static bool set_nr_if_polling(struct task_struct *p)
{
	return false;
}
#endif
399 400
#endif

401
static bool __wake_q_add(struct wake_q_head *head, struct task_struct *task)
402 403 404 405 406 407 408 409
{
	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.
	 *
410 411
	 * In order to ensure that a pending wakeup will observe our pending
	 * state, even in the failed case, an explicit smp_mb() must be used.
412
	 */
413
	smp_mb__before_atomic();
414
	if (unlikely(cmpxchg_relaxed(&node->next, NULL, WAKE_Q_TAIL)))
415
		return false;
416 417 418 419 420 421

	/*
	 * The head is context local, there can be no concurrency.
	 */
	*head->lastp = node;
	head->lastp = &node->next;
422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463
	return true;
}

/**
 * wake_q_add() - queue a wakeup for 'later' waking.
 * @head: the wake_q_head to add @task to
 * @task: the task to queue for 'later' wakeup
 *
 * Queue a task for later wakeup, most likely by the wake_up_q() call in the
 * same context, _HOWEVER_ this is not guaranteed, the wakeup can come
 * instantly.
 *
 * This function must be used as-if it were wake_up_process(); IOW the task
 * must be ready to be woken at this location.
 */
void wake_q_add(struct wake_q_head *head, struct task_struct *task)
{
	if (__wake_q_add(head, task))
		get_task_struct(task);
}

/**
 * wake_q_add_safe() - safely queue a wakeup for 'later' waking.
 * @head: the wake_q_head to add @task to
 * @task: the task to queue for 'later' wakeup
 *
 * Queue a task for later wakeup, most likely by the wake_up_q() call in the
 * same context, _HOWEVER_ this is not guaranteed, the wakeup can come
 * instantly.
 *
 * This function must be used as-if it were wake_up_process(); IOW the task
 * must be ready to be woken at this location.
 *
 * This function is essentially a task-safe equivalent to wake_q_add(). Callers
 * that already hold reference to @task can call the 'safe' version and trust
 * wake_q to do the right thing depending whether or not the @task is already
 * queued for wakeup.
 */
void wake_q_add_safe(struct wake_q_head *head, struct task_struct *task)
{
	if (!__wake_q_add(head, task))
		put_task_struct(task);
464 465 466 467 468 469 470 471 472 473 474
}

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);
475
		/* Task can safely be re-inserted now: */
476 477 478 479
		node = node->next;
		task->wake_q.next = NULL;

		/*
480 481
		 * wake_up_process() executes a full barrier, which pairs with
		 * the queueing in wake_q_add() so as not to miss wakeups.
482 483 484 485 486 487
		 */
		wake_up_process(task);
		put_task_struct(task);
	}
}

488
/*
489
 * resched_curr - mark rq's current task 'to be rescheduled now'.
490 491 492 493 494
 *
 * 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.
 */
495
void resched_curr(struct rq *rq)
496
{
497
	struct task_struct *curr = rq->curr;
498 499
	int cpu;

500
	lockdep_assert_held(&rq->lock);
501

502
	if (test_tsk_need_resched(curr))
503 504
		return;

505
	cpu = cpu_of(rq);
506

507
	if (cpu == smp_processor_id()) {
508
		set_tsk_need_resched(curr);
509
		set_preempt_need_resched();
510
		return;
511
	}
512

513
	if (set_nr_and_not_polling(curr))
514
		smp_send_reschedule(cpu);
515 516
	else
		trace_sched_wake_idle_without_ipi(cpu);
517 518
}

519
void resched_cpu(int cpu)
520 521 522 523
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

524
	raw_spin_lock_irqsave(&rq->lock, flags);
525 526
	if (cpu_online(cpu) || cpu == smp_processor_id())
		resched_curr(rq);
527
	raw_spin_unlock_irqrestore(&rq->lock, flags);
528
}
529

530
#ifdef CONFIG_SMP
531
#ifdef CONFIG_NO_HZ_COMMON
532
/*
533 534
 * In the semi idle case, use the nearest busy CPU for migrating timers
 * from an idle CPU.  This is good for power-savings.
535 536
 *
 * We don't do similar optimization for completely idle system, as
537 538
 * 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).
539
 */
540
int get_nohz_timer_target(void)
541
{
542
	int i, cpu = smp_processor_id();
543 544
	struct sched_domain *sd;

545
	if (!idle_cpu(cpu) && housekeeping_cpu(cpu, HK_FLAG_TIMER))
546 547
		return cpu;

548
	rcu_read_lock();
549
	for_each_domain(cpu, sd) {
550
		for_each_cpu(i, sched_domain_span(sd)) {
551 552 553
			if (cpu == i)
				continue;

554
			if (!idle_cpu(i) && housekeeping_cpu(i, HK_FLAG_TIMER)) {
555 556 557 558
				cpu = i;
				goto unlock;
			}
		}
559
	}
560

561 562
	if (!housekeeping_cpu(cpu, HK_FLAG_TIMER))
		cpu = housekeeping_any_cpu(HK_FLAG_TIMER);
563 564
unlock:
	rcu_read_unlock();
565 566
	return cpu;
}
567

568 569 570 571 572 573 574 575 576 577
/*
 * 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.
 */
578
static void wake_up_idle_cpu(int cpu)
579 580 581 582 583 584
{
	struct rq *rq = cpu_rq(cpu);

	if (cpu == smp_processor_id())
		return;

585
	if (set_nr_and_not_polling(rq->idle))
586
		smp_send_reschedule(cpu);
587 588
	else
		trace_sched_wake_idle_without_ipi(cpu);
589 590
}

591
static bool wake_up_full_nohz_cpu(int cpu)
592
{
593 594 595 596 597 598
	/*
	 * 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.
	 */
599 600
	if (cpu_is_offline(cpu))
		return true;  /* Don't try to wake offline CPUs. */
601
	if (tick_nohz_full_cpu(cpu)) {
602 603
		if (cpu != smp_processor_id() ||
		    tick_nohz_tick_stopped())
604
			tick_nohz_full_kick_cpu(cpu);
605 606 607 608 609 610
		return true;
	}

	return false;
}

611 612 613 614 615
/*
 * 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.
 */
616 617
void wake_up_nohz_cpu(int cpu)
{
618
	if (!wake_up_full_nohz_cpu(cpu))
619 620 621
		wake_up_idle_cpu(cpu);
}

622
static inline bool got_nohz_idle_kick(void)
623
{
624
	int cpu = smp_processor_id();
625

626
	if (!(atomic_read(nohz_flags(cpu)) & NOHZ_KICK_MASK))
627 628 629 630 631 632 633 634 635
		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
	 */
636
	atomic_andnot(NOHZ_KICK_MASK, nohz_flags(cpu));
637
	return false;
638 639
}

640
#else /* CONFIG_NO_HZ_COMMON */
641

642
static inline bool got_nohz_idle_kick(void)
643
{
644
	return false;
645 646
}

647
#endif /* CONFIG_NO_HZ_COMMON */
648

649
#ifdef CONFIG_NO_HZ_FULL
650
bool sched_can_stop_tick(struct rq *rq)
651
{
652 653 654 655 656 657
	int fifo_nr_running;

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

658
	/*
659 660
	 * If there are more than one RR tasks, we need the tick to effect the
	 * actual RR behaviour.
661
	 */
662 663 664 665 666
	if (rq->rt.rr_nr_running) {
		if (rq->rt.rr_nr_running == 1)
			return true;
		else
			return false;
667 668
	}

669 670 671 672 673 674 675 676 677 678 679 680 681 682
	/*
	 * 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)
683
		return false;
684

685
	return true;
686 687
}
#endif /* CONFIG_NO_HZ_FULL */
688
#endif /* CONFIG_SMP */
689

690 691
#if defined(CONFIG_RT_GROUP_SCHED) || (defined(CONFIG_FAIR_GROUP_SCHED) && \
			(defined(CONFIG_SMP) || defined(CONFIG_CFS_BANDWIDTH)))
692
/*
693 694 695 696
 * Iterate task_group tree rooted at *from, calling @down when first entering a
 * node and @up when leaving it for the final time.
 *
 * Caller must hold rcu_lock or sufficient equivalent.
697
 */
698
int walk_tg_tree_from(struct task_group *from,
699
			     tg_visitor down, tg_visitor up, void *data)
700 701
{
	struct task_group *parent, *child;
702
	int ret;
703

704 705
	parent = from;

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

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

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

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

735
static void set_load_weight(struct task_struct *p, bool update_load)
736
{
737 738 739
	int prio = p->static_prio - MAX_RT_PRIO;
	struct load_weight *load = &p->se.load;

Ingo Molnar's avatar
Ingo Molnar committed
740 741 742
	/*
	 * SCHED_IDLE tasks get minimal weight:
	 */
743
	if (task_has_idle_policy(p)) {
744
		load->weight = scale_load(WEIGHT_IDLEPRIO);
745
		load->inv_weight = WMULT_IDLEPRIO;
746
		p->se.runnable_weight = load->weight;
Ingo Molnar's avatar
Ingo Molnar committed
747 748
		return;
	}
749

750 751 752 753 754 755 756 757 758
	/*
	 * 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];
759
		p->se.runnable_weight = load->weight;
760
	}
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
	if (!(flags & ENQUEUE_RESTORE)) {
769
		sched_info_queued(rq, p);
770 771
		psi_enqueue(p, flags & ENQUEUE_WAKEUP);
	}
772

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

776
static inline void dequeue_task(struct rq *rq, struct task_struct *p, int flags)
777
{
778 779 780
	if (!(flags & DEQUEUE_NOCLOCK))
		update_rq_clock(rq);

781
	if (!(flags & DEQUEUE_SAVE)) {
782
		sched_info_dequeued(rq, p);
783 784
		psi_dequeue(p, flags & DEQUEUE_SLEEP);
	}
785

786
	p->sched_class->dequeue_task(rq, p, flags);
787 788
}

789
void activate_task(struct rq *rq, struct task_struct *p, int flags)
790 791 792 793
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible--;

794
	enqueue_task(rq, p, flags);
795 796
}

797
void deactivate_task(struct rq *rq, struct task_struct *p, int flags)
798 799 800 801
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible++;

802
	dequeue_task(rq, p, flags);
803 804
}

805
/*
Ingo Molnar's avatar
Ingo Molnar committed
806
 * __normal_prio - return the priority that is based on the static prio
807 808 809
 */
static inline int __normal_prio(struct task_struct *p)
{
Ingo Molnar's avatar
Ingo Molnar committed
810
	return p->static_prio;
811 812
}

813 814 815 816 817 818 819
/*
 * 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.
 */
820
static inline int normal_prio(struct task_struct *p)
821 822 823
{
	int prio;

824 825 826
	if (task_has_dl_policy(p))
		prio = MAX_DL_PRIO-1;
	else if (task_has_rt_policy(p))
827 828 829 830 831 832 833 834 835 836 837 838 839
		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.
 */
840
static int effective_prio(struct task_struct *p)
841 842 843 844 845 846 847 848 849 850 851 852
{
	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;
}

Linus Torvalds's avatar
Linus Torvalds committed
853 854 855
/**
 * task_curr - is this task currently executing on a CPU?
 * @p: the task in question.
856 857
 *
 * Return: 1 if the task is currently executing. 0 otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
858
 */
859
inline int task_curr(const struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
860 861 862 863
{
	return cpu_curr(task_cpu(p)) == p;
}

864
/*
865 866 867 868 869
 * 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().
870
 */
871 872
static inline void check_class_changed(struct rq *rq, struct task_struct *p,
				       const struct sched_class *prev_class,
873
				       int oldprio)
874 875 876
{
	if (prev_class != p->sched_class) {
		if (prev_class->switched_from)
877
			prev_class->switched_from(rq, p);
878

879
		p->sched_class->switched_to(rq, p);
880
	} else if (oldprio != p->prio || dl_task(p))
881
		p->sched_class->prio_changed(rq, p, oldprio);
882 883
}

884
void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags)
885 886 887 888 889 890 891 892 893 894
{
	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) {
895
				resched_curr(rq);
896 897 898 899 900 901 902 903 904
				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.
	 */
905
	if (task_on_rq_queued(rq->curr) && test_tsk_need_resched(rq->curr))
906
		rq_clock_skip_update(rq);
907 908
}

Linus Torvalds's avatar
Linus Torvalds committed
909
#ifdef CONFIG_SMP
910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936

static inline bool is_per_cpu_kthread(struct task_struct *p)
{
	if (!(p->flags & PF_KTHREAD))
		return false;

	if (p->nr_cpus_allowed != 1)
		return false;

	return true;
}

/*
 * Per-CPU kthreads are allowed to run on !actie && online CPUs, see
 * __set_cpus_allowed_ptr() and select_fallback_rq().
 */
static inline bool is_cpu_allowed(struct task_struct *p, int cpu)
{
	if (!cpumask_test_cpu(cpu, &p->cpus_allowed))
		return false;

	if (is_per_cpu_kthread(p))
		return cpu_online(cpu);

	return cpu_active(cpu);
}

Peter Zijlstra's avatar
Peter Zijlstra committed
937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955
/*
 * 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.
 */
956 957
static struct rq *move_queued_task(struct rq *rq, struct rq_flags *rf,
				   struct task_struct *p, int new_cpu)
Peter Zijlstra's avatar
Peter Zijlstra committed
958 959 960
{
	lockdep_assert_held(&rq->lock);

961
	WRITE_ONCE(p->on_rq, TASK_ON_RQ_MIGRATING);
962
	dequeue_task(rq, p, DEQUEUE_NOCLOCK);
Peter Zijlstra's avatar
Peter Zijlstra committed
963
	set_task_cpu(p, new_cpu);
964
	rq_unlock(rq, rf);
Peter Zijlstra's avatar
Peter Zijlstra committed
965 966 967

	rq = cpu_rq(new_cpu);

968
	rq_lock(rq, rf);
Peter Zijlstra's avatar
Peter Zijlstra committed
969 970
	BUG_ON(task_cpu(p) != new_cpu);
	enqueue_task(rq, p, 0);
971
	p->on_rq = TASK_ON_RQ_QUEUED;
Peter Zijlstra's avatar
Peter Zijlstra committed
972 973 974 975 976 977 978 979 980 981 982
	check_preempt_curr(rq, p, 0);

	return rq;
}

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

/*
983
 * Move (not current) task off this CPU, onto the destination CPU. We're doing
Peter Zijlstra's avatar
Peter Zijlstra committed
984 985 986 987 988 989 990
 * 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.
 */
991 992
static struct rq *__migrate_task(struct rq *rq, struct rq_flags *rf,
				 struct task_struct *p, int dest_cpu)
Peter Zijlstra's avatar
Peter Zijlstra committed
993 994
{
	/* Affinity changed (again). */
995
	if (!is_cpu_allowed(p, dest_cpu))
996
		return rq;
Peter Zijlstra's avatar
Peter Zijlstra committed
997

998
	update_rq_clock(rq);
999
	rq = move_queued_task(rq, rf, p, dest_cpu);
1000 1001

	return rq;
Peter Zijlstra's avatar
Peter Zijlstra committed
1002 1003 1004 1005 1006 1007 1008 1009 1010 1011
}

/*
 * 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;
1012 1013
	struct task_struct *p = arg->task;
	struct rq *rq = this_rq();
1014
	struct rq_flags rf;
Peter Zijlstra's avatar
Peter Zijlstra committed
1015 1016

	/*
1017 1018
	 * The original target CPU might have gone down and we might
	 * be on another CPU but it doesn't matter.
Peter Zijlstra's avatar
Peter Zijlstra committed
1019 1020 1021 1022 1023 1024 1025 1026
	 */
	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();
1027 1028

	raw_spin_lock(&p->pi_lock);
1029
	rq_lock(rq, &rf);
1030 1031 1032 1033 1034
	/*
	 * 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.
	 */
1035 1036
	if (task_rq(p) == rq) {
		if (task_on_rq_queued(p))
1037
			rq = __migrate_task(rq, &rf, p, arg->dest_cpu);
1038 1039 1040
		else
			p->wake_cpu = arg->dest_cpu;
	}
1041
	rq_unlock(rq, &rf);
1042 1043
	raw_spin_unlock(&p->pi_lock);

Peter Zijlstra's avatar
Peter Zijlstra committed
1044 1045 1046 1047
	local_irq_enable();
	return 0;
}

1048 1049 1050 1051 1052
/*
 * 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)
Peter Zijlstra's avatar
Peter Zijlstra committed
1053 1054 1055 1056 1057
{
	cpumask_copy(&p->cpus_allowed, new_mask);
	p->nr_cpus_allowed = cpumask_weight(new_mask);
}

1058 1059
void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
{
1060 1061 1062
	struct rq *rq = task_rq(p);
	bool queued, running;

1063
	lockdep_assert_held(&p->pi_lock);
1064 1065 1066 1067 1068 1069 1070 1071 1072 1073

	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);
1074
		dequeue_task(rq, p, DEQUEUE_SAVE | DEQUEUE_NOCLOCK);
1075 1076 1077 1078
	}
	if (running)
		put_prev_task(rq, p);

1079
	p->sched_class->set_cpus_allowed(p, new_mask);
1080 1081

	if (queued)
1082
		enqueue_task(rq, p, ENQUEUE_RESTORE | ENQUEUE_NOCLOCK);
1083
	if (running)
1084
		set_curr_task(rq, p);
1085 1086
}

Peter Zijlstra's avatar
Peter Zijlstra committed
1087 1088 1089 1090 1091 1092 1093 1094 1095
/*
 * 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.
 */
1096 1097
static int __set_cpus_allowed_ptr(struct task_struct *p,
				  const struct cpumask *new_mask, bool check)
Peter Zijlstra's avatar
Peter Zijlstra committed
1098
{
1099
	const struct cpumask *cpu_valid_mask = cpu_active_mask;
Peter Zijlstra's avatar
Peter Zijlstra committed
1100
	unsigned int dest_cpu;
1101 1102
	struct rq_flags rf;
	struct rq *rq;
Peter Zijlstra's avatar
Peter Zijlstra committed
1103 1104
	int ret = 0;

1105
	rq = task_rq_lock(p, &rf);
1106
	update_rq_clock(rq);
Peter Zijlstra's avatar
Peter Zijlstra committed
1107

1108 1109 1110 1111 1112 1113 1114
	if (p->flags & PF_KTHREAD) {
		/*
		 * Kernel threads are allowed on online && !active CPUs
		 */
		cpu_valid_mask = cpu_online_mask;
	}

1115 1116 1117 1118 1119 1120 1121 1122 1123
	/*
	 * 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;
	}

Peter Zijlstra's avatar
Peter Zijlstra committed
1124 1125 1126
	if (cpumask_equal(&p->cpus_allowed, new_mask))
		goto out;

1127
	if (!cpumask_intersects(new_mask, cpu_valid_mask)) {
Peter Zijlstra's avatar
Peter Zijlstra committed
1128 1129 1130 1131 1132 1133
		ret = -EINVAL;
		goto out;
	}

	do_set_cpus_allowed(p, new_mask);

1134 1135 1136
	if (p->flags & PF_KTHREAD) {
		/*
		 * For kernel threads that do indeed end up on online &&
1137
		 * !active we want to ensure they are strict per-CPU threads.
1138 1139 1140 1141 1142 1143
		 */
		WARN_ON(cpumask_intersects(new_mask, cpu_online_mask) &&
			!cpumask_intersects(new_mask, cpu_active_mask) &&
			p->nr_cpus_allowed != 1);
	}

Peter Zijlstra's avatar
Peter Zijlstra committed
1144 1145 1146 1147
	/* 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;

1148
	dest_cpu = cpumask_any_and(cpu_valid_mask, new_mask);
Peter Zijlstra's avatar
Peter Zijlstra committed
1149 1150 1151
	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. */
1152
		task_rq_unlock(rq, p, &rf);
Peter Zijlstra's avatar
Peter Zijlstra committed
1153 1154 1155
		stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg);
		tlb_migrate_finish(p->mm);
		return 0;
1156 1157 1158 1159 1160
	} else if (task_on_rq_queued(p)) {
		/*
		 * OK, since we're going to drop the lock immediately
		 * afterwards anyway.
		 */
1161
		rq = move_queued_task(rq, &rf, p, dest_cpu);
1162
	}
Peter Zijlstra's avatar
Peter Zijlstra committed
1163
out:
1164
	task_rq_unlock(rq, p, &rf);
Peter Zijlstra's avatar
Peter Zijlstra committed
1165 1166 1167

	return ret;
}
1168 1169 1170 1171 1172

int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
{
	return __set_cpus_allowed_ptr(p, new_mask, false);
}
Peter Zijlstra's avatar
Peter Zijlstra committed
1173 1174
EXPORT_SYMBOL_GPL(set_cpus_allowed_ptr);

Ingo Molnar's avatar
Ingo Molnar committed
1175
void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
Ingo Molnar's avatar
Ingo Molnar committed
1176
{
1177 1178 1179 1180 1181
#ifdef CONFIG_SCHED_DEBUG
	/*
	 * We should never call set_task_cpu() on a blocked task,
	 * ttwu() will sort out the placement.
	 */
Peter Zijlstra's avatar
Peter Zijlstra committed
1182
	WARN_ON_ONCE(p->state != TASK_RUNNING && p->state != TASK_WAKING &&
1183
			!p->on_rq);
1184

1185 1186 1187 1188 1189 1190 1191 1192 1193
	/*
	 * 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)));

1194
#ifdef CONFIG_LOCKDEP
1195 1196 1197 1198 1199
	/*
	 * 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,
1200
	 * see task_group().
1201 1202 1203 1204
	 *
	 * Furthermore, all task_rq users should acquire both locks, see
	 * task_rq_lock().
	 */
1205 1206 1207
	WARN_ON_ONCE(debug_locks && !(lockdep_is_held(&p->pi_lock) ||
				      lockdep_is_held(&task_rq(p)->lock)));
#endif
1208 1209 1210 1211
	/*
	 * Clearly, migrating tasks to offline CPUs is a fairly daft thing.
	 */
	WARN_ON_ONCE(!cpu_online(new_cpu));
1212 1213
#endif

1214
	trace_sched_migrate_task(p, new_cpu);
1215

1216
	if (task_cpu(p) != new_cpu) {
1217
		if (p->sched_class->migrate_task_rq)
1218
			p->sched_class->migrate_task_rq(p, new_cpu);
1219
		p->se.nr_migrations++;
1220
		rseq_migrate(p);
1221
		perf_event_task_migrate(p);
1222
	}
Ingo Molnar's avatar
Ingo Molnar committed
1223 1224

	__set_task_cpu(p, new_cpu);
Ingo Molnar's avatar
Ingo Molnar committed
1225 1226
}

1227
#ifdef CONFIG_NUMA_BALANCING
1228 1229
static void __migrate_swap_task(struct task_struct *p, int cpu)
{
1230
	if (task_on_rq_queued(p)) {
1231
		struct rq *src_rq, *dst_rq;
1232
		struct rq_flags srf, drf;
1233 1234 1235 1236

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

1237 1238 1239
		rq_pin_lock(src_rq, &srf);
		rq_pin_lock(dst_rq, &drf);

1240
		p->on_rq = TASK_ON_RQ_MIGRATING;
1241 1242 1243
		deactivate_task(src_rq, p, 0);
		set_task_cpu(p, cpu);
		activate_task(dst_rq, p, 0);
1244
		p->on_rq = TASK_ON_RQ_QUEUED;
1245
		check_preempt_curr(dst_rq, p, 0);
1246 1247 1248 1249

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

1250 1251 1252 1253
	} 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
1254
		 * previous CPU our target instead of where it really is.
1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270
		 */
		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;

1271 1272 1273
	if (!cpu_active(arg->src_cpu) || !cpu_active(arg->dst_cpu))
		return -EAGAIN;

1274 1275 1276
	src_rq = cpu_rq(arg->src_cpu);
	dst_rq = cpu_rq(arg->dst_cpu);

1277 1278
	double_raw_lock(&arg->src_task->pi_lock,
			&arg->dst_task->pi_lock);
1279
	double_rq_lock(src_rq, dst_rq);
1280

1281 1282 1283 1284 1285 1286
	if (task_cpu(arg->dst_task) != arg->dst_cpu)
		goto unlock;

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

1287
	if (!cpumask_test_cpu(arg->dst_cpu, &arg->src_task->cpus_allowed))
1288 1289
		goto unlock;

1290
	if (!cpumask_test_cpu(arg->src_cpu, &arg->dst_task->cpus_allowed))
1291 1292 1293 1294 1295 1296 1297 1298 1299
		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);
1300 1301
	raw_spin_unlock(&arg->dst_task->pi_lock);
	raw_spin_unlock(&arg->src_task->pi_lock);
1302 1303 1304 1305 1306 1307 1308

	return ret;
}

/*
 * Cross migrate two tasks
 */
1309 1310
int migrate_swap(struct task_struct *cur, struct task_struct *p,
		int target_cpu, int curr_cpu)
1311 1312 1313 1314 1315 1316
{
	struct migration_swap_arg arg;
	int ret = -EINVAL;

	arg = (struct migration_swap_arg){
		.src_task = cur,
1317
		.src_cpu = curr_cpu,
1318
		.dst_task = p,
1319
		.dst_cpu = target_cpu,
1320 1321 1322 1323 1324
	};

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

1325 1326 1327 1328
	/*
	 * 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.
	 */
1329 1330 1331
	if (!cpu_active(arg.src_cpu) || !cpu_active(arg.dst_cpu))
		goto out;

1332
	if (!cpumask_test_cpu(arg.dst_cpu, &arg.src_task->cpus_allowed))
1333 1334
		goto out;

1335
	if (!cpumask_test_cpu(arg.src_cpu, &arg.dst_task->cpus_allowed))
1336 1337
		goto out;

1338
	trace_sched_swap_numa(cur, arg.src_cpu, p, arg.dst_cpu);
1339 1340 1341 1342 1343
	ret = stop_two_cpus(arg.dst_cpu, arg.src_cpu, migrate_swap_stop, &arg);

out:
	return ret;
}
1344
#endif /* CONFIG_NUMA_BALANCING */
1345

Linus Torvalds's avatar
Linus Torvalds committed
1346 1347 1348
/*
 * wait_task_inactive - wait for a thread to unschedule.
 *
1349 1350 1351 1352 1353 1354 1355
 * 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.
 *
Linus Torvalds's avatar
Linus Torvalds committed
1356 1357 1358 1359 1360 1361
 * 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.
 */
1362
unsigned long wait_task_inactive(struct task_struct *p, long match_state)
Linus Torvalds's avatar
Linus Torvalds committed
1363
{
1364
	int running, queued;
1365
	struct rq_flags rf;
1366
	unsigned long ncsw;
1367
	struct rq *rq;
Linus Torvalds's avatar
Linus Torvalds committed
1368

1369 1370 1371 1372 1373 1374 1375 1376
	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);
1377

1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388
		/*
		 * 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!
		 */
1389 1390 1391
		while (task_running(rq, p)) {
			if (match_state && unlikely(p->state != match_state))
				return 0;
1392
			cpu_relax();
1393
		}
1394

1395 1396 1397 1398 1399
		/*
		 * 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.
		 */
1400
		rq = task_rq_lock(p, &rf);
1401
		trace_sched_wait_task(p);
1402
		running = task_running(rq, p);
1403
		queued = task_on_rq_queued(p);
1404
		ncsw = 0;
1405
		if (!match_state || p->state == match_state)
1406
			ncsw = p->nvcsw | LONG_MIN; /* sets MSB */
1407
		task_rq_unlock(rq, p, &rf);
1408

1409 1410 1411 1412 1413 1414
		/*
		 * If it changed from the expected state, bail out now.
		 */
		if (unlikely(!ncsw))
			break;

1415 1416 1417 1418 1419 1420 1421 1422 1423 1424
		/*
		 * 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;
		}
1425

1426 1427 1428 1429 1430
		/*
		 * It's not enough that it's not actively running,
		 * it must be off the runqueue _entirely_, and not
		 * preempted!
		 *
1431
		 * So if it was still runnable (but just not actively
1432 1433 1434
		 * running right now), it's preempted, and we should
		 * yield - it could be a while.
		 */
1435
		if (unlikely(queued)) {
1436
			ktime_t to = NSEC_PER_SEC / HZ;
1437 1438 1439

			set_current_state(TASK_UNINTERRUPTIBLE);
			schedule_hrtimeout(&to, HRTIMER_MODE_REL);
1440 1441
			continue;
		}
1442

1443 1444 1445 1446 1447 1448 1449
		/*
		 * 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;
	}
1450 1451

	return ncsw;
Linus Torvalds's avatar
Linus Torvalds committed
1452 1453 1454 1455 1456 1457 1458 1459 1460
}

/***
 * 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.)
 *
Lucas De Marchi's avatar
Lucas De Marchi committed
1461
 * NOTE: this function doesn't have to take the runqueue lock,
Linus Torvalds's avatar
Linus Torvalds committed
1462 1463 1464 1465 1466
 * 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.
 */
1467
void kick_process(struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
1468 1469 1470 1471 1472 1473 1474 1475 1476
{
	int cpu;

	preempt_disable();
	cpu = task_cpu(p);
	if ((cpu != smp_processor_id()) && task_curr(p))
		smp_send_reschedule(cpu);
	preempt_enable();
}
Rusty Russell's avatar
Rusty Russell committed
1477
EXPORT_SYMBOL_GPL(kick_process);
Linus Torvalds's avatar
Linus Torvalds committed
1478

1479
/*
1480
 * ->cpus_allowed is protected by both rq->lock and p->pi_lock
1481 1482 1483 1484 1485
 *
 * A few notes on cpu_active vs cpu_online:
 *
 *  - cpu_active must be a subset of cpu_online
 *
1486
 *  - on CPU-up we allow per-CPU kthreads on the online && !active CPU,
1487
 *    see __set_cpus_allowed_ptr(). At this point the newly online
1488
 *    CPU isn't yet part of the sched domains, and balancing will not
1489 1490
 *    see it.
 *
1491
 *  - on CPU-down we clear cpu_active() to mask the sched domains and
1492
 *    avoid the load balancer to place new tasks on the to be removed
1493
 *    CPU. Existing tasks will remain running there and will be taken
1494 1495 1496 1497 1498 1499
 *    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.
1500
 */
1501 1502
static int select_fallback_rq(int cpu, struct task_struct *p)
{
1503 1504
	int nid = cpu_to_node(cpu);
	const struct cpumask *nodemask = NULL;
1505 1506
	enum { cpuset, possible, fail } state = cpuset;
	int dest_cpu;
1507

1508
	/*
1509 1510 1511
	 * 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.
1512 1513 1514 1515 1516 1517 1518 1519
	 */
	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;
1520
			if (cpumask_test_cpu(dest_cpu, &p->cpus_allowed))
1521 1522
				return dest_cpu;
		}
1523
	}
1524

1525 1526
	for (;;) {
		/* Any allowed, online CPU? */
1527
		for_each_cpu(dest_cpu, &p->cpus_allowed) {
1528
			if (!is_cpu_allowed(p, dest_cpu))
1529
				continue;
1530

1531 1532
			goto out;
		}
1533

1534
		/* No more Mr. Nice Guy. */
1535 1536
		switch (state) {
		case cpuset:
1537 1538 1539 1540 1541
			if (IS_ENABLED(CONFIG_CPUSETS)) {
				cpuset_cpus_allowed_fallback(p);
				state = possible;
				break;
			}
1542
			/* Fall-through */
1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561
		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()) {
1562
			printk_deferred("process %d (%s) no longer affine to cpu%d\n",
1563 1564
					task_pid_nr(p), p->comm, cpu);
		}
1565 1566 1567 1568 1569
	}

	return dest_cpu;
}

1570
/*
1571
 * The caller (fork, wakeup) owns p->pi_lock, ->cpus_allowed is stable.
1572
 */
1573
static inline
1574
int select_task_rq(struct task_struct *p, int cpu, int sd_flags, int wake_flags)
1575
{
1576 1577
	lockdep_assert_held(&p->pi_lock);

1578
	if (p->nr_cpus_allowed > 1)
1579
		cpu = p->sched_class->select_task_rq(p, cpu, sd_flags, wake_flags);
1580
	else
1581
		cpu = cpumask_any(&p->cpus_allowed);
1582 1583 1584 1585

	/*
	 * 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
1586
	 * CPU.
1587 1588 1589 1590 1591 1592
	 *
	 * 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 ]
	 */
1593
	if (unlikely(!is_cpu_allowed(p, cpu)))
1594
		cpu = select_fallback_rq(task_cpu(p), p);
1595 1596

	return cpu;
1597
}
1598 1599 1600 1601 1602 1603

static void update_avg(u64 *avg, u64 sample)
{
	s64 diff = sample - *avg;
	*avg += diff >> 3;
}
1604

1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634
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;
	}
}

1635 1636 1637 1638 1639 1640 1641 1642
#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);
}

Peter Zijlstra's avatar
Peter Zijlstra committed
1643
#endif /* CONFIG_SMP */
1644

1645
static void
1646
ttwu_stat(struct task_struct *p, int cpu, int wake_flags)
Tejun Heo's avatar
Tejun Heo committed
1647
{
1648
	struct rq *rq;
1649

1650 1651 1652 1653
	if (!schedstat_enabled())
		return;

	rq = this_rq();
1654

1655 1656
#ifdef CONFIG_SMP
	if (cpu == rq->cpu) {
1657 1658
		__schedstat_inc(rq->ttwu_local);
		__schedstat_inc(p->se.statistics.nr_wakeups_local);
1659 1660 1661
	} else {
		struct sched_domain *sd;

1662
		__schedstat_inc(p->se.statistics.nr_wakeups_remote);
1663
		rcu_read_lock();
1664
		for_each_domain(rq->cpu, sd) {
1665
			if (cpumask_test_cpu(cpu, sched_domain_span(sd))) {
1666
				__schedstat_inc(sd->ttwu_wake_remote);
1667 1668 1669
				break;
			}
		}
1670
		rcu_read_unlock();
1671
	}
1672 1673

	if (wake_flags & WF_MIGRATED)
1674
		__schedstat_inc(p->se.statistics.nr_wakeups_migrate);
1675 1676
#endif /* CONFIG_SMP */

1677 1678
	__schedstat_inc(rq->ttwu_count);
	__schedstat_inc(p->se.statistics.nr_wakeups);
1679 1680

	if (wake_flags & WF_SYNC)
1681
		__schedstat_inc(p->se.statistics.nr_wakeups_sync);
1682 1683
}

1684 1685 1686
/*
 * Mark the task runnable and perform wakeup-preemption.
 */
1687
static void ttwu_do_wakeup(struct rq *rq, struct task_struct *p, int wake_flags,
1688
			   struct rq_flags *rf)
Tejun Heo's avatar
Tejun Heo committed
1689 1690 1691
{
	check_preempt_curr(rq, p, wake_flags);
	p->state = TASK_RUNNING;
1692 1693
	trace_sched_wakeup(p);

Tejun Heo's avatar
Tejun Heo committed
1694
#ifdef CONFIG_SMP
1695 1696
	if (p->sched_class->task_woken) {
		/*
1697 1698
		 * Our task @p is fully woken up and running; so its safe to
		 * drop the rq->lock, hereafter rq is only used for statistics.
1699
		 */
1700
		rq_unpin_lock(rq, rf);
Tejun Heo's avatar
Tejun Heo committed
1701
		p->sched_class->task_woken(rq, p);
1702
		rq_repin_lock(rq, rf);
1703
	}
Tejun Heo's avatar
Tejun Heo committed
1704

1705
	if (rq->idle_stamp) {
1706
		u64 delta = rq_clock(rq) - rq->idle_stamp;
1707
		u64 max = 2*rq->max_idle_balance_cost;
Tejun Heo's avatar
Tejun Heo committed
1708

1709 1710 1711
		update_avg(&rq->avg_idle, delta);

		if (rq->avg_idle > max)
Tejun Heo's avatar
Tejun Heo committed
1712
			rq->avg_idle = max;
1713

Tejun Heo's avatar
Tejun Heo committed
1714 1715 1716 1717 1718
		rq->idle_stamp = 0;
	}
#endif
}

1719
static void
1720
ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags,
1721
		 struct rq_flags *rf)
1722
{
1723
	int en_flags = ENQUEUE_WAKEUP | ENQUEUE_NOCLOCK;
1724

1725 1726
	lockdep_assert_held(&rq->lock);

1727 1728 1729
#ifdef CONFIG_SMP
	if (p->sched_contributes_to_load)
		rq->nr_uninterruptible--;
1730 1731

	if (wake_flags & WF_MIGRATED)
1732
		en_flags |= ENQUEUE_MIGRATED;
1733 1734
#endif

1735 1736
	activate_task(rq, p, en_flags);
	p->on_rq = TASK_ON_RQ_QUEUED;
1737
	ttwu_do_wakeup(rq, p, wake_flags, rf);
1738 1739 1740 1741 1742 1743 1744 1745 1746 1747
}

/*
 * 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)
{
1748
	struct rq_flags rf;
1749 1750 1751
	struct rq *rq;
	int ret = 0;

1752
	rq = __task_rq_lock(p, &rf);
1753
	if (task_on_rq_queued(p)) {
1754 1755
		/* check_preempt_curr() may use rq clock */
		update_rq_clock(rq);
1756
		ttwu_do_wakeup(rq, p, wake_flags, &rf);
1757 1758
		ret = 1;
	}
1759
	__task_rq_unlock(rq, &rf);
1760 1761 1762 1763

	return ret;
}

1764
#ifdef CONFIG_SMP
1765
void sched_ttwu_pending(void)
1766 1767
{
	struct rq *rq = this_rq();
1768
	struct llist_node *llist = llist_del_all(&rq->wake_list);
1769
	struct task_struct *p, *t;
1770
	struct rq_flags rf;
1771

1772 1773 1774
	if (!llist)
		return;

1775
	rq_lock_irqsave(rq, &rf);
1776
	update_rq_clock(rq);
1777

1778 1779
	llist_for_each_entry_safe(p, t, llist, wake_entry)
		ttwu_do_activate(rq, p, p->sched_remote_wakeup ? WF_MIGRATED : 0, &rf);
1780

1781
	rq_unlock_irqrestore(rq, &rf);
1782 1783 1784 1785
}

void scheduler_ipi(void)
{
1786 1787 1788 1789 1790
	/*
	 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
	 * TIF_NEED_RESCHED remotely (for the first time) will also send
	 * this IPI.
	 */
1791
	preempt_fold_need_resched();
1792

1793
	if (llist_empty(&this_rq()->wake_list) && !got_nohz_idle_kick())
1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809
		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();
1810
	sched_ttwu_pending();
1811 1812 1813 1814

	/*
	 * Check if someone kicked us for doing the nohz idle load balance.
	 */
1815
	if (unlikely(got_nohz_idle_kick())) {
1816
		this_rq()->idle_balance = 1;
1817
		raise_softirq_irqoff(SCHED_SOFTIRQ);
1818
	}
1819
	irq_exit();
1820 1821
}

1822
static void ttwu_queue_remote(struct task_struct *p, int cpu, int wake_flags)
1823
{
1824 1825
	struct rq *rq = cpu_rq(cpu);

1826 1827
	p->sched_remote_wakeup = !!(wake_flags & WF_MIGRATED);

1828 1829 1830 1831 1832 1833
	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);
	}
1834
}
1835

1836 1837 1838
void wake_up_if_idle(int cpu)
{
	struct rq *rq = cpu_rq(cpu);
1839
	struct rq_flags rf;
1840

1841 1842 1843 1844
	rcu_read_lock();

	if (!is_idle_task(rcu_dereference(rq->curr)))
		goto out;
1845 1846 1847 1848

	if (set_nr_if_polling(rq->idle)) {
		trace_sched_wake_idle_without_ipi(cpu);
	} else {
1849
		rq_lock_irqsave(rq, &rf);
1850 1851
		if (is_idle_task(rq->curr))
			smp_send_reschedule(cpu);
1852
		/* Else CPU is not idle, do nothing here: */
1853
		rq_unlock_irqrestore(rq, &rf);
1854
	}
1855 1856 1857

out:
	rcu_read_unlock();
1858 1859
}

1860
bool cpus_share_cache(int this_cpu, int that_cpu)
1861 1862 1863
{
	return per_cpu(sd_llc_id, this_cpu) == per_cpu(sd_llc_id, that_cpu);
}
1864
#endif /* CONFIG_SMP */
1865

1866
static void ttwu_queue(struct task_struct *p, int cpu, int wake_flags)
1867 1868
{
	struct rq *rq = cpu_rq(cpu);
1869
	struct rq_flags rf;
1870

1871
#if defined(CONFIG_SMP)
1872
	if (sched_feat(TTWU_QUEUE) && !cpus_share_cache(smp_processor_id(), cpu)) {
1873
		sched_clock_cpu(cpu); /* Sync clocks across CPUs */
1874
		ttwu_queue_remote(p, cpu, wake_flags);
1875 1876 1877 1878
		return;
	}
#endif

1879
	rq_lock(rq, &rf);
1880
	update_rq_clock(rq);
1881
	ttwu_do_activate(rq, p, wake_flags, &rf);
1882
	rq_unlock(rq, &rf);
Tejun Heo's avatar
Tejun Heo committed
1883 1884
}

1885 1886 1887 1888 1889 1890
/*
 * Notes on Program-Order guarantees on SMP systems.
 *
 *  MIGRATION
 *
 * The basic program-order guarantee on SMP systems is that when a task [t]
1891 1892
 * migrates, all its activity on its old CPU [c0] happens-before any subsequent
 * execution on its new CPU [c1].
1893 1894 1895 1896 1897 1898 1899 1900
 *
 * 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
 *
1901
 * Release/acquire chaining guarantees that B happens after A and C after B.
1902
 * Note: the CPU doing B need not be c0 or c1
1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933
 *
 * 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)
1934
 *   2) smp_cond_load_acquire(!X->on_cpu)
1935 1936 1937 1938 1939 1940 1941 1942 1943 1944
 *
 * 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);
 *
1945
 *                    smp_cond_load_acquire(&X->on_cpu, !VAL);
1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962
 *                    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
 *
 *
1963 1964 1965
 * However, for wakeups there is a second guarantee we must provide, namely we
 * must ensure that CONDITION=1 done by the caller can not be reordered with
 * accesses to the task state; see try_to_wake_up() and set_current_state().
1966 1967
 */

Tejun Heo's avatar
Tejun Heo committed
1968
/**
Linus Torvalds's avatar
Linus Torvalds committed
1969
 * try_to_wake_up - wake up a thread
Tejun Heo's avatar
Tejun Heo committed
1970
 * @p: the thread to be awakened
Linus Torvalds's avatar
Linus Torvalds committed
1971
 * @state: the mask of task states that can be woken
Tejun Heo's avatar
Tejun Heo committed
1972
 * @wake_flags: wake modifier flags (WF_*)
Linus Torvalds's avatar
Linus Torvalds committed
1973
 *
1974
 * If (@state & @p->state) @p->state = TASK_RUNNING.
Linus Torvalds's avatar
Linus Torvalds committed
1975
 *
1976 1977 1978 1979 1980
 * 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().
 *
1981 1982 1983
 * This function executes a full memory barrier before accessing the task
 * state; see set_current_state().
 *
1984 1985
 * Return: %true if @p->state changes (an actual wakeup was done),
 *	   %false otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
1986
 */
1987 1988
static int
try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
Linus Torvalds's avatar
Linus Torvalds committed
1989 1990
{
	unsigned long flags;
1991
	int cpu, success = 0;
1992

1993 1994 1995 1996 1997 1998
	/*
	 * 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.
	 */
1999
	raw_spin_lock_irqsave(&p->pi_lock, flags);
2000
	smp_mb__after_spinlock();
Peter Zijlstra's avatar
Peter Zijlstra committed
2001
	if (!(p->state & state))
Linus Torvalds's avatar
Linus Torvalds committed
2002 2003
		goto out;

2004 2005
	trace_sched_waking(p);

2006 2007
	/* We're going to change ->state: */
	success = 1;
Linus Torvalds's avatar
Linus Torvalds committed
2008 2009
	cpu = task_cpu(p);

2010 2011 2012 2013 2014
	/*
	 * 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.
	 *
2015 2016 2017 2018 2019 2020 2021 2022
	 * sched_ttwu_pending()			try_to_wake_up()
	 *   STORE p->on_rq = 1			  LOAD p->state
	 *   UNLOCK rq->lock
	 *
	 * __schedule() (switch to task 'p')
	 *   LOCK rq->lock			  smp_rmb();
	 *   smp_mb__after_spinlock();
	 *   UNLOCK rq->lock
2023 2024
	 *
	 * [task p]
2025
	 *   STORE p->state = UNINTERRUPTIBLE	  LOAD p->on_rq
2026
	 *
2027 2028
	 * Pairs with the LOCK+smp_mb__after_spinlock() on rq->lock in
	 * __schedule().  See the comment for smp_mb__after_spinlock().
2029 2030
	 */
	smp_rmb();
2031 2032
	if (p->on_rq && ttwu_remote(p, wake_flags))
		goto stat;
Linus Torvalds's avatar
Linus Torvalds committed
2033 2034

#ifdef CONFIG_SMP
2035 2036 2037 2038 2039 2040 2041
	/*
	 * 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.
	 *
2042 2043 2044 2045 2046 2047 2048 2049
	 * __schedule() (switch to task 'p')	try_to_wake_up()
	 *   STORE p->on_cpu = 1		  LOAD p->on_rq
	 *   UNLOCK rq->lock
	 *
	 * __schedule() (put 'p' to sleep)
	 *   LOCK rq->lock			  smp_rmb();
	 *   smp_mb__after_spinlock();
	 *   STORE p->on_rq = 0			  LOAD p->on_cpu
2050
	 *
2051 2052
	 * Pairs with the LOCK+smp_mb__after_spinlock() on rq->lock in
	 * __schedule().  See the comment for smp_mb__after_spinlock().
2053 2054 2055
	 */
	smp_rmb();

Peter Zijlstra's avatar
Peter Zijlstra committed
2056
	/*
2057
	 * If the owning (remote) CPU is still in the middle of schedule() with
2058
	 * this task as prev, wait until its done referencing the task.
2059
	 *
2060
	 * Pairs with the smp_store_release() in finish_task().
2061 2062 2063
	 *
	 * This ensures that tasks getting woken will be fully ordered against
	 * their previous state and preserve Program Order.
2064
	 */
2065
	smp_cond_load_acquire(&p->on_cpu, !VAL);
Linus Torvalds's avatar
Linus Torvalds committed
2066

2067
	p->sched_contributes_to_load = !!task_contributes_to_load(p);
Peter Zijlstra's avatar
Peter Zijlstra committed
2068
	p->state = TASK_WAKING;
2069

2070
	if (p->in_iowait) {
2071
		delayacct_blkio_end(p);
2072 2073 2074
		atomic_dec(&task_rq(p)->nr_iowait);
	}

2075
	cpu = select_task_rq(p, p->wake_cpu, SD_BALANCE_WAKE, wake_flags);
2076 2077
	if (task_cpu(p) != cpu) {
		wake_flags |= WF_MIGRATED;
2078
		psi_ttwu_dequeue(p);
2079
		set_task_cpu(p, cpu);
2080
	}
2081 2082 2083 2084

#else /* CONFIG_SMP */

	if (p->in_iowait) {
2085
		delayacct_blkio_end(p);
2086 2087 2088
		atomic_dec(&task_rq(p)->nr_iowait);
	}

Linus Torvalds's avatar
Linus Torvalds committed
2089 2090
#endif /* CONFIG_SMP */

2091
	ttwu_queue(p, cpu, wake_flags);
2092
stat:
2093
	ttwu_stat(p, cpu, wake_flags);
Linus Torvalds's avatar
Linus Torvalds committed
2094
out:
2095
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
2096 2097 2098 2099

	return success;
}

2100 2101 2102 2103 2104
/**
 * 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
2105 2106 2107
 * processes.
 *
 * Return: 1 if the process was woken up, 0 if it was already running.
2108
 *
2109
 * This function executes a full memory barrier before accessing the task state.
2110
 */
2111
int wake_up_process(struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
2112
{
2113
	return try_to_wake_up(p, TASK_NORMAL, 0);
Linus Torvalds's avatar
Linus Torvalds committed
2114 2115 2116
}
EXPORT_SYMBOL(wake_up_process);

2117
int wake_up_state(struct task_struct *p, unsigned int state)
Linus Torvalds's avatar
Linus Torvalds committed
2118 2119 2120 2121 2122 2123 2124
{
	return try_to_wake_up(p, state, 0);
}

/*
 * Perform scheduler related setup for a newly forked process p.
 * p is forked by current.
Ingo Molnar's avatar
Ingo Molnar committed
2125 2126 2127
 *
 * __sched_fork() is basic setup used by init_idle() too:
 */
2128
static void __sched_fork(unsigned long clone_flags, struct task_struct *p)
Ingo Molnar's avatar
Ingo Molnar committed
2129
{
Peter Zijlstra's avatar
Peter Zijlstra committed
2130 2131 2132
	p->on_rq			= 0;

	p->se.on_rq			= 0;
Ingo Molnar's avatar
Ingo Molnar committed
2133 2134
	p->se.exec_start		= 0;
	p->se.sum_exec_runtime		= 0;
2135
	p->se.prev_sum_exec_runtime	= 0;
2136
	p->se.nr_migrations		= 0;
2137
	p->se.vruntime			= 0;
Peter Zijlstra's avatar
Peter Zijlstra committed
2138
	INIT_LIST_HEAD(&p->se.group_node);
2139

2140 2141 2142 2143
#ifdef CONFIG_FAIR_GROUP_SCHED
	p->se.cfs_rq			= NULL;
#endif

2144
#ifdef CONFIG_SCHEDSTATS
2145
	/* Even if schedstat is disabled, there should not be garbage */
2146
	memset(&p->se.statistics, 0, sizeof(p->se.statistics));
2147
#endif
2148

2149
	RB_CLEAR_NODE(&p->dl.rb_node);
2150
	init_dl_task_timer(&p->dl);
2151
	init_dl_inactive_task_timer(&p->dl);
2152
	__dl_clear_params(p);
2153

Peter Zijlstra's avatar
Peter Zijlstra committed
2154
	INIT_LIST_HEAD(&p->rt.run_list);
2155 2156 2157 2158
	p->rt.timeout		= 0;
	p->rt.time_slice	= sched_rr_timeslice;
	p->rt.on_rq		= 0;
	p->rt.on_list		= 0;
2159

2160 2161 2162
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&p->preempt_notifiers);
#endif
2163

2164 2165 2166
#ifdef CONFIG_COMPACTION
	p->capture_control = NULL;
#endif
2167
	init_numa_balancing(clone_flags, p);
Ingo Molnar's avatar
Ingo Molnar committed
2168 2169
}

2170 2171
DEFINE_STATIC_KEY_FALSE(sched_numa_balancing);

2172
#ifdef CONFIG_NUMA_BALANCING
2173

2174 2175 2176
void set_numabalancing_state(bool enabled)
{
	if (enabled)
2177
		static_branch_enable(&sched_numa_balancing);
2178
	else
2179
		static_branch_disable(&sched_numa_balancing);
2180
}
2181 2182 2183 2184 2185 2186 2187

#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;
2188
	int state = static_branch_likely(&sched_numa_balancing);
2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203

	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
Ingo Molnar's avatar
Ingo Molnar committed
2204

2205 2206
#ifdef CONFIG_SCHEDSTATS

2207
DEFINE_STATIC_KEY_FALSE(sched_schedstats);
2208
static bool __initdata __sched_schedstats = false;
2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231

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;

2232 2233 2234 2235 2236
	/*
	 * 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.
	 */
2237
	if (!strcmp(str, "enable")) {
2238
		__sched_schedstats = true;
2239 2240
		ret = 1;
	} else if (!strcmp(str, "disable")) {
2241
		__sched_schedstats = false;
2242 2243 2244 2245 2246 2247 2248 2249 2250 2251
		ret = 1;
	}
out:
	if (!ret)
		pr_warn("Unable to parse schedstats=\n");

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

2252 2253 2254 2255 2256
static void __init init_schedstats(void)
{
	set_schedstats(__sched_schedstats);
}

2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276
#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;
}
2277 2278 2279 2280
#endif /* CONFIG_PROC_SYSCTL */
#else  /* !CONFIG_SCHEDSTATS */
static inline void init_schedstats(void) {}
#endif /* CONFIG_SCHEDSTATS */
Ingo Molnar's avatar
Ingo Molnar committed
2281 2282 2283 2284

/*
 * fork()/clone()-time setup:
 */
2285
int sched_fork(unsigned long clone_flags, struct task_struct *p)
Ingo Molnar's avatar
Ingo Molnar committed
2286
{
2287
	unsigned long flags;
Ingo Molnar's avatar
Ingo Molnar committed
2288

2289
	__sched_fork(clone_flags, p);
2290
	/*
2291
	 * We mark the process as NEW here. This guarantees that
2292 2293 2294
	 * nobody will actually run it, and a signal or other external
	 * event cannot wake it up and insert it on the runqueue either.
	 */
2295
	p->state = TASK_NEW;
Ingo Molnar's avatar
Ingo Molnar committed
2296

2297 2298 2299 2300 2301
	/*
	 * Make sure we do not leak PI boosting priority to the child.
	 */
	p->prio = current->normal_prio;

2302 2303 2304 2305
	/*
	 * Revert to default priority/policy on fork if requested.
	 */
	if (unlikely(p->sched_reset_on_fork)) {
2306
		if (task_has_dl_policy(p) || task_has_rt_policy(p)) {
2307
			p->policy = SCHED_NORMAL;
2308
			p->static_prio = NICE_TO_PRIO(0);
2309 2310 2311 2312 2313
			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);
2314
		set_load_weight(p, false);
2315

2316 2317 2318 2319 2320 2321
		/*
		 * We don't need the reset flag anymore after the fork. It has
		 * fulfilled its duty:
		 */
		p->sched_reset_on_fork = 0;
	}
2322

2323
	if (dl_prio(p->prio))
2324
		return -EAGAIN;
2325
	else if (rt_prio(p->prio))
2326
		p->sched_class = &rt_sched_class;
2327
	else
2328
		p->sched_class = &fair_sched_class;
2329

2330
	init_entity_runnable_average(&p->se);
2331

2332 2333 2334 2335 2336 2337 2338
	/*
	 * 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.
	 */
2339
	raw_spin_lock_irqsave(&p->pi_lock, flags);
2340
	/*
2341
	 * We're setting the CPU for the first time, we don't migrate,
2342 2343
	 * so use __set_task_cpu().
	 */
2344
	__set_task_cpu(p, smp_processor_id());
2345 2346
	if (p->sched_class->task_fork)
		p->sched_class->task_fork(p);
2347
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
2348

2349
#ifdef CONFIG_SCHED_INFO
Ingo Molnar's avatar
Ingo Molnar committed
2350
	if (likely(sched_info_on()))
2351
		memset(&p->sched_info, 0, sizeof(p->sched_info));
Linus Torvalds's avatar
Linus Torvalds committed
2352
#endif
2353 2354
#if defined(CONFIG_SMP)
	p->on_cpu = 0;
2355
#endif
2356
	init_task_preempt_count(p);
2357
#ifdef CONFIG_SMP
2358
	plist_node_init(&p->pushable_tasks, MAX_PRIO);
2359
	RB_CLEAR_NODE(&p->pushable_dl_tasks);
2360
#endif
2361
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
2362 2363
}

2364 2365 2366
unsigned long to_ratio(u64 period, u64 runtime)
{
	if (runtime == RUNTIME_INF)
2367
		return BW_UNIT;
2368 2369 2370 2371 2372 2373 2374 2375 2376

	/*
	 * 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;

2377
	return div64_u64(runtime << BW_SHIFT, period);
2378 2379
}

Linus Torvalds's avatar
Linus Torvalds committed
2380 2381 2382 2383 2384 2385 2386
/*
 * 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.
 */
2387
void wake_up_new_task(struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
2388
{
2389
	struct rq_flags rf;
Ingo Molnar's avatar
Ingo Molnar committed
2390
	struct rq *rq;
2391

2392
	raw_spin_lock_irqsave(&p->pi_lock, rf.flags);
2393
	p->state = TASK_RUNNING;
2394 2395 2396 2397
#ifdef CONFIG_SMP
	/*
	 * Fork balancing, do it here and not earlier because:
	 *  - cpus_allowed can change in the fork path
2398
	 *  - any previously selected CPU might disappear through hotplug
2399 2400 2401
	 *
	 * Use __set_task_cpu() to avoid calling sched_class::migrate_task_rq,
	 * as we're not fully set-up yet.
2402
	 */
2403
	p->recent_used_cpu = task_cpu(p);
2404
	__set_task_cpu(p, select_task_rq(p, task_cpu(p), SD_BALANCE_FORK, 0));
2405
#endif
2406
	rq = __task_rq_lock(p, &rf);
2407
	update_rq_clock(rq);
2408
	post_init_entity_util_avg(p);
2409

2410
	activate_task(rq, p, ENQUEUE_NOCLOCK);
2411
	p->on_rq = TASK_ON_RQ_QUEUED;
2412
	trace_sched_wakeup_new(p);
Peter Zijlstra's avatar
Peter Zijlstra committed
2413
	check_preempt_curr(rq, p, WF_FORK);
2414
#ifdef CONFIG_SMP
2415 2416 2417 2418 2419
	if (p->sched_class->task_woken) {
		/*
		 * Nothing relies on rq->lock after this, so its fine to
		 * drop it.
		 */
2420
		rq_unpin_lock(rq, &rf);
2421
		p->sched_class->task_woken(rq, p);
2422
		rq_repin_lock(rq, &rf);
2423
	}
2424
#endif
2425
	task_rq_unlock(rq, p, &rf);
Linus Torvalds's avatar
Linus Torvalds committed
2426 2427
}

2428 2429
#ifdef CONFIG_PREEMPT_NOTIFIERS

2430
static DEFINE_STATIC_KEY_FALSE(preempt_notifier_key);
2431

2432 2433
void preempt_notifier_inc(void)
{
2434
	static_branch_inc(&preempt_notifier_key);
2435 2436 2437 2438 2439
}
EXPORT_SYMBOL_GPL(preempt_notifier_inc);

void preempt_notifier_dec(void)
{
2440
	static_branch_dec(&preempt_notifier_key);
2441 2442 2443
}
EXPORT_SYMBOL_GPL(preempt_notifier_dec);

2444
/**
2445
 * preempt_notifier_register - tell me when current is being preempted & rescheduled
2446
 * @notifier: notifier struct to register
2447 2448 2449
 */
void preempt_notifier_register(struct preempt_notifier *notifier)
{
2450
	if (!static_branch_unlikely(&preempt_notifier_key))
2451 2452
		WARN(1, "registering preempt_notifier while notifiers disabled\n");

2453 2454 2455 2456 2457 2458
	hlist_add_head(&notifier->link, &current->preempt_notifiers);
}
EXPORT_SYMBOL_GPL(preempt_notifier_register);

/**
 * preempt_notifier_unregister - no longer interested in preemption notifications
2459
 * @notifier: notifier struct to unregister
2460
 *
2461
 * This is *not* safe to call from within a preemption notifier.
2462 2463 2464 2465 2466 2467 2468
 */
void preempt_notifier_unregister(struct preempt_notifier *notifier)
{
	hlist_del(&notifier->link);
}
EXPORT_SYMBOL_GPL(preempt_notifier_unregister);

2469
static void __fire_sched_in_preempt_notifiers(struct task_struct *curr)
2470 2471 2472
{
	struct preempt_notifier *notifier;

2473
	hlist_for_each_entry(notifier, &curr->preempt_notifiers, link)
2474 2475 2476
		notifier->ops->sched_in(notifier, raw_smp_processor_id());
}

2477 2478
static __always_inline void fire_sched_in_preempt_notifiers(struct task_struct *curr)
{
2479
	if (static_branch_unlikely(&preempt_notifier_key))
2480 2481 2482
		__fire_sched_in_preempt_notifiers(curr);
}

2483
static void
2484 2485
__fire_sched_out_preempt_notifiers(struct task_struct *curr,
				   struct task_struct *next)
2486 2487 2488
{
	struct preempt_notifier *notifier;

2489
	hlist_for_each_entry(notifier, &curr->preempt_notifiers, link)
2490 2491 2492
		notifier->ops->sched_out(notifier, next);
}

2493 2494 2495 2496
static __always_inline void
fire_sched_out_preempt_notifiers(struct task_struct *curr,
				 struct task_struct *next)
{
2497
	if (static_branch_unlikely(&preempt_notifier_key))
2498 2499 2500
		__fire_sched_out_preempt_notifiers(curr, next);
}

2501
#else /* !CONFIG_PREEMPT_NOTIFIERS */
2502

2503
static inline void fire_sched_in_preempt_notifiers(struct task_struct *curr)
2504 2505 2506
{
}

2507
static inline void
2508 2509 2510 2511 2512
fire_sched_out_preempt_notifiers(struct task_struct *curr,
				 struct task_struct *next)
{
}

2513
#endif /* CONFIG_PREEMPT_NOTIFIERS */
2514

2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542
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
}

2543 2544
static inline void
prepare_lock_switch(struct rq *rq, struct task_struct *next, struct rq_flags *rf)
2545
{
2546 2547 2548 2549 2550 2551 2552 2553
	/*
	 * 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:
	 */
	rq_unpin_lock(rq, rf);
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
2554 2555
#ifdef CONFIG_DEBUG_SPINLOCK
	/* this is a valid case when another task releases the spinlock */
2556
	rq->lock.owner = next;
2557
#endif
2558 2559 2560 2561
}

static inline void finish_lock_switch(struct rq *rq)
{
2562 2563 2564 2565 2566 2567 2568 2569 2570
	/*
	 * 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);
}

2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582
/*
 * NOP if the arch has not defined these:
 */

#ifndef prepare_arch_switch
# define prepare_arch_switch(next)	do { } while (0)
#endif

#ifndef finish_arch_post_lock_switch
# define finish_arch_post_lock_switch()	do { } while (0)
#endif

2583 2584 2585
/**
 * prepare_task_switch - prepare to switch tasks
 * @rq: the runqueue preparing to switch
2586
 * @prev: the current task that is being switched out
2587 2588 2589 2590 2591 2592 2593 2594 2595
 * @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.
 */
2596 2597 2598
static inline void
prepare_task_switch(struct rq *rq, struct task_struct *prev,
		    struct task_struct *next)
2599
{
2600
	kcov_prepare_switch(prev);
2601
	sched_info_switch(rq, prev, next);
2602
	perf_event_task_sched_out(prev, next);
2603
	rseq_preempt(prev);
2604
	fire_sched_out_preempt_notifiers(prev, next);
2605
	prepare_task(next);
2606 2607 2608
	prepare_arch_switch(next);
}

Linus Torvalds's avatar
Linus Torvalds committed
2609 2610 2611 2612
/**
 * finish_task_switch - clean up after a task-switch
 * @prev: the thread we just switched away from.
 *
2613 2614 2615 2616
 * 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.
Linus Torvalds's avatar
Linus Torvalds committed
2617 2618
 *
 * Note that we may have delayed dropping an mm in context_switch(). If
Ingo Molnar's avatar
Ingo Molnar committed
2619
 * so, we finish that here outside of the runqueue lock. (Doing it
Linus Torvalds's avatar
Linus Torvalds committed
2620 2621
 * with the lock held can cause deadlocks; see schedule() for
 * details.)
2622 2623 2624 2625 2626
 *
 * 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.
Linus Torvalds's avatar
Linus Torvalds committed
2627
 */
2628
static struct rq *finish_task_switch(struct task_struct *prev)
Linus Torvalds's avatar
Linus Torvalds committed
2629 2630
	__releases(rq->lock)
{
2631
	struct rq *rq = this_rq();
Linus Torvalds's avatar
Linus Torvalds committed
2632
	struct mm_struct *mm = rq->prev_mm;
Oleg Nesterov's avatar
Oleg Nesterov committed
2633
	long prev_state;
Linus Torvalds's avatar
Linus Torvalds committed
2634

2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645
	/*
	 * 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.
	 */
2646 2647 2648 2649
	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);
2650

Linus Torvalds's avatar
Linus Torvalds committed
2651 2652 2653 2654
	rq->prev_mm = NULL;

	/*
	 * A task struct has one reference for the use as "current".
2655
	 * If a task dies, then it sets TASK_DEAD in tsk->state and calls
Oleg Nesterov's avatar
Oleg Nesterov committed
2656 2657
	 * schedule one last time. The schedule call will never return, and
	 * the scheduled task must drop that reference.
2658 2659
	 *
	 * We must observe prev->state before clearing prev->on_cpu (in
2660
	 * finish_task), otherwise a concurrent wakeup can get prev
2661 2662
	 * running on another CPU and we could rave with its RUNNING -> DEAD
	 * transition, resulting in a double drop.
Linus Torvalds's avatar
Linus Torvalds committed
2663
	 */
Oleg Nesterov's avatar
Oleg Nesterov committed
2664
	prev_state = prev->state;
2665
	vtime_task_switch(prev);
2666
	perf_event_task_sched_in(prev, current);
2667 2668
	finish_task(prev);
	finish_lock_switch(rq);
2669
	finish_arch_post_lock_switch();
2670
	kcov_finish_switch(current);
Steven Rostedt's avatar
Steven Rostedt committed
2671

2672
	fire_sched_in_preempt_notifiers(current);
2673
	/*
2674 2675 2676 2677 2678 2679 2680 2681 2682 2683
	 * When switching through a kernel thread, the loop in
	 * membarrier_{private,global}_expedited() may have observed that
	 * kernel thread and not issued an IPI. It is therefore possible to
	 * schedule between user->kernel->user threads without passing though
	 * switch_mm(). Membarrier requires a barrier after storing to
	 * rq->curr, before returning to userspace, so provide them here:
	 *
	 * - a full memory barrier for {PRIVATE,GLOBAL}_EXPEDITED, implicitly
	 *   provided by mmdrop(),
	 * - a sync_core for SYNC_CORE.
2684
	 */
2685 2686
	if (mm) {
		membarrier_mm_sync_core_before_usermode(mm);
Linus Torvalds's avatar
Linus Torvalds committed
2687
		mmdrop(mm);
2688
	}
2689 2690 2691
	if (unlikely(prev_state == TASK_DEAD)) {
		if (prev->sched_class->task_dead)
			prev->sched_class->task_dead(prev);
2692

2693 2694 2695 2696 2697 2698 2699 2700 2701 2702
		/*
		 * Remove function-return probe instances associated with this
		 * task and put them back on the free list.
		 */
		kprobe_flush_task(prev);

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

		put_task_struct(prev);
2703
	}
2704

2705
	tick_nohz_task_switch();
2706
	return rq;
Linus Torvalds's avatar
Linus Torvalds committed
2707 2708
}

2709 2710 2711
#ifdef CONFIG_SMP

/* rq->lock is NOT held, but preemption is disabled */
2712
static void __balance_callback(struct rq *rq)
2713
{
2714 2715 2716
	struct callback_head *head, *next;
	void (*func)(struct rq *rq);
	unsigned long flags;
2717

2718 2719 2720 2721 2722 2723 2724 2725
	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;
2726

2727
		func(rq);
2728
	}
2729 2730 2731 2732 2733 2734 2735
	raw_spin_unlock_irqrestore(&rq->lock, flags);
}

static inline void balance_callback(struct rq *rq)
{
	if (unlikely(rq->balance_callback))
		__balance_callback(rq);
2736 2737 2738
}

#else
2739

2740
static inline void balance_callback(struct rq *rq)
2741
{
Linus Torvalds's avatar
Linus Torvalds committed
2742 2743
}

2744 2745
#endif

Linus Torvalds's avatar
Linus Torvalds committed
2746 2747 2748 2749
/**
 * schedule_tail - first thing a freshly forked thread must call.
 * @prev: the thread we just switched away from.
 */
2750
asmlinkage __visible void schedule_tail(struct task_struct *prev)
Linus Torvalds's avatar
Linus Torvalds committed
2751 2752
	__releases(rq->lock)
{
2753
	struct rq *rq;
2754

2755 2756 2757 2758 2759 2760 2761 2762 2763
	/*
	 * 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).
	 */

2764
	rq = finish_task_switch(prev);
2765
	balance_callback(rq);
2766
	preempt_enable();
2767

Linus Torvalds's avatar
Linus Torvalds committed
2768
	if (current->set_child_tid)
2769
		put_user(task_pid_vnr(current), current->set_child_tid);
2770 2771

	calculate_sigpending();
Linus Torvalds's avatar
Linus Torvalds committed
2772 2773 2774
}

/*
2775
 * context_switch - switch to the new MM and the new thread's register state.
Linus Torvalds's avatar
Linus Torvalds committed
2776
 */
2777
static __always_inline struct rq *
2778
context_switch(struct rq *rq, struct task_struct *prev,
2779
	       struct task_struct *next, struct rq_flags *rf)
Linus Torvalds's avatar
Linus Torvalds committed
2780
{
Ingo Molnar's avatar
Ingo Molnar committed
2781
	struct mm_struct *mm, *oldmm;
Linus Torvalds's avatar
Linus Torvalds committed
2782

2783
	prepare_task_switch(rq, prev, next);
2784

Ingo Molnar's avatar
Ingo Molnar committed
2785 2786
	mm = next->mm;
	oldmm = prev->active_mm;
2787 2788 2789 2790 2791
	/*
	 * For paravirt, this is coupled with an exit in switch_to to
	 * combine the page table reload and the switch backend into
	 * one hypercall.
	 */
2792
	arch_start_context_switch(prev);
2793

2794 2795 2796 2797 2798 2799 2800
	/*
	 * If mm is non-NULL, we pass through switch_mm(). If mm is
	 * NULL, we will pass through mmdrop() in finish_task_switch().
	 * Both of these contain the full memory barrier required by
	 * membarrier after storing to rq->curr, before returning to
	 * user-space.
	 */
2801
	if (!mm) {
Linus Torvalds's avatar
Linus Torvalds committed
2802
		next->active_mm = oldmm;
2803
		mmgrab(oldmm);
Linus Torvalds's avatar
Linus Torvalds committed
2804 2805
		enter_lazy_tlb(oldmm, next);
	} else
2806
		switch_mm_irqs_off(oldmm, mm, next);
Linus Torvalds's avatar
Linus Torvalds committed
2807

2808
	if (!prev->mm) {
Linus Torvalds's avatar
Linus Torvalds committed
2809 2810 2811
		prev->active_mm = NULL;
		rq->prev_mm = oldmm;
	}
2812

2813
	rq->clock_update_flags &= ~(RQCF_ACT_SKIP|RQCF_REQ_SKIP);
2814

2815
	prepare_lock_switch(rq, next, rf);
Linus Torvalds's avatar
Linus Torvalds committed
2816 2817 2818

	/* Here we just switch the register state and the stack. */
	switch_to(prev, next, prev);
Ingo Molnar's avatar
Ingo Molnar committed
2819
	barrier();
2820 2821

	return finish_task_switch(prev);
Linus Torvalds's avatar
Linus Torvalds committed
2822 2823 2824
}

/*
2825
 * nr_running and nr_context_switches:
Linus Torvalds's avatar
Linus Torvalds committed
2826 2827
 *
 * externally visible scheduler statistics: current number of runnable
2828
 * threads, total number of context switches performed since bootup.
Linus Torvalds's avatar
Linus Torvalds committed
2829 2830 2831 2832 2833 2834 2835 2836 2837
 */
unsigned long nr_running(void)
{
	unsigned long i, sum = 0;

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

	return sum;
2838
}
Linus Torvalds's avatar
Linus Torvalds committed
2839

2840
/*
2841
 * Check if only the current task is running on the CPU.
2842 2843 2844 2845 2846
 *
 * 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:
 *
2847
 * - from a non-preemptible section (of course)
2848 2849 2850 2851
 *
 * - from a thread that is bound to a single CPU
 *
 * - in a loop with very short iterations (e.g. a polling loop)
2852 2853 2854
 */
bool single_task_running(void)
{
2855
	return raw_rq()->nr_running == 1;
2856 2857 2858
}
EXPORT_SYMBOL(single_task_running);

Linus Torvalds's avatar
Linus Torvalds committed
2859
unsigned long long nr_context_switches(void)
2860
{
2861 2862
	int i;
	unsigned long long sum = 0;
2863

2864
	for_each_possible_cpu(i)
Linus Torvalds's avatar
Linus Torvalds committed
2865
		sum += cpu_rq(i)->nr_switches;
2866

Linus Torvalds's avatar
Linus Torvalds committed
2867 2868
	return sum;
}
2869

2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881
/*
 * Consumers of these two interfaces, like for example the cpuidle menu
 * governor, are using nonsensical data. Preferring shallow idle state selection
 * for a CPU that has IO-wait which might not even end up running the task when
 * it does become runnable.
 */

unsigned long nr_iowait_cpu(int cpu)
{
	return atomic_read(&cpu_rq(cpu)->nr_iowait);
}

2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911
/*
 * 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'.
 */

Linus Torvalds's avatar
Linus Torvalds committed
2912 2913 2914
unsigned long nr_iowait(void)
{
	unsigned long i, sum = 0;
2915

2916
	for_each_possible_cpu(i)
2917
		sum += nr_iowait_cpu(i);
2918

Linus Torvalds's avatar
Linus Torvalds committed
2919 2920
	return sum;
}
2921

Ingo Molnar's avatar
Ingo Molnar committed
2922
#ifdef CONFIG_SMP
2923

2924
/*
2925 2926
 * sched_exec - execve() is a valuable balancing opportunity, because at
 * this point the task has the smallest effective memory and cache footprint.
2927
 */
2928
void sched_exec(void)
2929
{
2930
	struct task_struct *p = current;
Linus Torvalds's avatar
Linus Torvalds committed
2931
	unsigned long flags;
2932
	int dest_cpu;
2933

2934
	raw_spin_lock_irqsave(&p->pi_lock, flags);
2935
	dest_cpu = p->sched_class->select_task_rq(p, task_cpu(p), SD_BALANCE_EXEC, 0);
2936 2937
	if (dest_cpu == smp_processor_id())
		goto unlock;
2938

2939
	if (likely(cpu_active(dest_cpu))) {
2940
		struct migration_arg arg = { p, dest_cpu };
2941

2942 2943
		raw_spin_unlock_irqrestore(&p->pi_lock, flags);
		stop_one_cpu(task_cpu(p), migration_cpu_stop, &arg);
Linus Torvalds's avatar
Linus Torvalds committed
2944 2945
		return;
	}
2946
unlock:
2947
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
2948
}
Ingo Molnar's avatar
Ingo Molnar committed
2949

Linus Torvalds's avatar
Linus Torvalds committed
2950 2951 2952
#endif

DEFINE_PER_CPU(struct kernel_stat, kstat);
2953
DEFINE_PER_CPU(struct kernel_cpustat, kernel_cpustat);
Linus Torvalds's avatar
Linus Torvalds committed
2954 2955

EXPORT_PER_CPU_SYMBOL(kstat);
2956
EXPORT_PER_CPU_SYMBOL(kernel_cpustat);
Linus Torvalds's avatar
Linus Torvalds committed
2957

2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974
/*
 * 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);
}

2975 2976 2977 2978 2979 2980 2981
/*
 * 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)
{
2982
	struct rq_flags rf;
2983
	struct rq *rq;
2984
	u64 ns;
2985

2986 2987
#if defined(CONFIG_64BIT) && defined(CONFIG_SMP)
	/*
2988
	 * 64-bit doesn't need locks to atomically read a 64-bit value.
2989 2990 2991
	 * So we have a optimization chance when the task's delta_exec is 0.
	 * Reading ->on_cpu is racy, but this is ok.
	 *
2992 2993
	 * 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
2994
	 * indistinguishable from the read occurring a few cycles earlier.
2995 2996
	 * If we see ->on_cpu without ->on_rq, the task is leaving, and has
	 * been accounted, so we're correct here as well.
2997
	 */
2998
	if (!p->on_cpu || !task_on_rq_queued(p))
2999 3000 3001
		return p->se.sum_exec_runtime;
#endif

3002
	rq = task_rq_lock(p, &rf);
3003 3004 3005 3006 3007 3008
	/*
	 * 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)) {
3009
		prefetch_curr_exec_start(p);
3010 3011 3012 3013
		update_rq_clock(rq);
		p->sched_class->update_curr(rq);
	}
	ns = p->se.sum_exec_runtime;
3014
	task_rq_unlock(rq, p, &rf);
3015 3016 3017

	return ns;
}
3018

3019 3020 3021 3022 3023 3024 3025 3026
/*
 * 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);
Ingo Molnar's avatar
Ingo Molnar committed
3027
	struct task_struct *curr = rq->curr;
3028
	struct rq_flags rf;
3029 3030

	sched_clock_tick();
Ingo Molnar's avatar
Ingo Molnar committed
3031

3032 3033
	rq_lock(rq, &rf);

3034
	update_rq_clock(rq);
Peter Zijlstra's avatar
Peter Zijlstra committed
3035
	curr->sched_class->task_tick(rq, curr, 0);
3036
	cpu_load_update_active(rq);
3037
	calc_global_load_tick(rq);
3038
	psi_task_tick(rq);
3039 3040

	rq_unlock(rq, &rf);
3041

3042
	perf_event_task_tick();
3043

3044
#ifdef CONFIG_SMP
3045
	rq->idle_balance = idle_cpu(cpu);
3046
	trigger_load_balance(rq);
3047
#endif
Linus Torvalds's avatar
Linus Torvalds committed
3048 3049
}

3050
#ifdef CONFIG_NO_HZ_FULL
3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064

struct tick_work {
	int			cpu;
	struct delayed_work	work;
};

static struct tick_work __percpu *tick_work_cpu;

static void sched_tick_remote(struct work_struct *work)
{
	struct delayed_work *dwork = to_delayed_work(work);
	struct tick_work *twork = container_of(dwork, struct tick_work, work);
	int cpu = twork->cpu;
	struct rq *rq = cpu_rq(cpu);
3065
	struct task_struct *curr;
3066
	struct rq_flags rf;
3067
	u64 delta;
3068 3069 3070 3071 3072 3073 3074 3075

	/*
	 * Handle the tick only if it appears the remote CPU is running in full
	 * dynticks mode. The check is racy by nature, but missing a tick or
	 * having one too much is no big deal because the scheduler tick updates
	 * statistics and checks timeslices in a time-independent way, regardless
	 * of when exactly it is running.
	 */
3076 3077
	if (idle_cpu(cpu) || !tick_nohz_tick_stopped_cpu(cpu))
		goto out_requeue;
3078

3079 3080 3081 3082
	rq_lock_irq(rq, &rf);
	curr = rq->curr;
	if (is_idle_task(curr))
		goto out_unlock;
3083

3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095
	update_rq_clock(rq);
	delta = rq_clock_task(rq) - curr->se.exec_start;

	/*
	 * Make sure the next tick runs within a reasonable
	 * amount of time.
	 */
	WARN_ON_ONCE(delta > (u64)NSEC_PER_SEC * 3);
	curr->sched_class->task_tick(rq, curr, 0);

out_unlock:
	rq_unlock_irq(rq, &rf);
3096

3097
out_requeue:
3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146
	/*
	 * Run the remote tick once per second (1Hz). This arbitrary
	 * frequency is large enough to avoid overload but short enough
	 * to keep scheduler internal stats reasonably up to date.
	 */
	queue_delayed_work(system_unbound_wq, dwork, HZ);
}

static void sched_tick_start(int cpu)
{
	struct tick_work *twork;

	if (housekeeping_cpu(cpu, HK_FLAG_TICK))
		return;

	WARN_ON_ONCE(!tick_work_cpu);

	twork = per_cpu_ptr(tick_work_cpu, cpu);
	twork->cpu = cpu;
	INIT_DELAYED_WORK(&twork->work, sched_tick_remote);
	queue_delayed_work(system_unbound_wq, &twork->work, HZ);
}

#ifdef CONFIG_HOTPLUG_CPU
static void sched_tick_stop(int cpu)
{
	struct tick_work *twork;

	if (housekeeping_cpu(cpu, HK_FLAG_TICK))
		return;

	WARN_ON_ONCE(!tick_work_cpu);

	twork = per_cpu_ptr(tick_work_cpu, cpu);
	cancel_delayed_work_sync(&twork->work);
}
#endif /* CONFIG_HOTPLUG_CPU */

int __init sched_tick_offload_init(void)
{
	tick_work_cpu = alloc_percpu(struct tick_work);
	BUG_ON(!tick_work_cpu);

	return 0;
}

#else /* !CONFIG_NO_HZ_FULL */
static inline void sched_tick_start(int cpu) { }
static inline void sched_tick_stop(int cpu) { }
3147
#endif
Linus Torvalds's avatar
Linus Torvalds committed
3148

3149
#if defined(CONFIG_PREEMPT) && (defined(CONFIG_DEBUG_PREEMPT) || \
3150
				defined(CONFIG_TRACE_PREEMPT_TOGGLE))
3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164
/*
 * 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);
	}
}
3165

3166
void preempt_count_add(int val)
Linus Torvalds's avatar
Linus Torvalds committed
3167
{
3168
#ifdef CONFIG_DEBUG_PREEMPT
Linus Torvalds's avatar
Linus Torvalds committed
3169 3170 3171
	/*
	 * Underflow?
	 */
3172 3173
	if (DEBUG_LOCKS_WARN_ON((preempt_count() < 0)))
		return;
3174
#endif
3175
	__preempt_count_add(val);
3176
#ifdef CONFIG_DEBUG_PREEMPT
Linus Torvalds's avatar
Linus Torvalds committed
3177 3178 3179
	/*
	 * Spinlock count overflowing soon?
	 */
3180 3181
	DEBUG_LOCKS_WARN_ON((preempt_count() & PREEMPT_MASK) >=
				PREEMPT_MASK - 10);
3182
#endif
3183
	preempt_latency_start(val);
Linus Torvalds's avatar
Linus Torvalds committed
3184
}
3185
EXPORT_SYMBOL(preempt_count_add);
3186
NOKPROBE_SYMBOL(preempt_count_add);
Linus Torvalds's avatar
Linus Torvalds committed
3187

3188 3189 3190 3191 3192 3193 3194 3195 3196 3197
/*
 * 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());
}

3198
void preempt_count_sub(int val)
Linus Torvalds's avatar
Linus Torvalds committed
3199
{
3200
#ifdef CONFIG_DEBUG_PREEMPT
Linus Torvalds's avatar
Linus Torvalds committed
3201 3202 3203
	/*
	 * Underflow?
	 */
3204
	if (DEBUG_LOCKS_WARN_ON(val > preempt_count()))
3205
		return;
Linus Torvalds's avatar
Linus Torvalds committed
3206 3207 3208
	/*
	 * Is the spinlock portion underflowing?
	 */
3209 3210 3211
	if (DEBUG_LOCKS_WARN_ON((val < PREEMPT_MASK) &&
			!(preempt_count() & PREEMPT_MASK)))
		return;
3212
#endif
3213

3214
	preempt_latency_stop(val);
3215
	__preempt_count_sub(val);
Linus Torvalds's avatar
Linus Torvalds committed
3216
}
3217
EXPORT_SYMBOL(preempt_count_sub);
3218
NOKPROBE_SYMBOL(preempt_count_sub);
Linus Torvalds's avatar
Linus Torvalds committed
3219

3220 3221 3222
#else
static inline void preempt_latency_start(int val) { }
static inline void preempt_latency_stop(int val) { }
Linus Torvalds's avatar
Linus Torvalds committed
3223 3224
#endif

3225 3226 3227 3228 3229 3230 3231 3232 3233
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
}

Linus Torvalds's avatar
Linus Torvalds committed
3234
/*
Ingo Molnar's avatar
Ingo Molnar committed
3235
 * Print scheduling while atomic bug:
Linus Torvalds's avatar
Linus Torvalds committed
3236
 */
Ingo Molnar's avatar
Ingo Molnar committed
3237
static noinline void __schedule_bug(struct task_struct *prev)
Linus Torvalds's avatar
Linus Torvalds committed
3238
{
3239 3240 3241
	/* Save this before calling printk(), since that will clobber it */
	unsigned long preempt_disable_ip = get_preempt_disable_ip(current);

3242 3243 3244
	if (oops_in_progress)
		return;

3245 3246
	printk(KERN_ERR "BUG: scheduling while atomic: %s/%d/0x%08x\n",
		prev->comm, prev->pid, preempt_count());
3247

Ingo Molnar's avatar
Ingo Molnar committed
3248
	debug_show_held_locks(prev);
3249
	print_modules();
Ingo Molnar's avatar
Ingo Molnar committed
3250 3251
	if (irqs_disabled())
		print_irqtrace_events(prev);
3252 3253
	if (IS_ENABLED(CONFIG_DEBUG_PREEMPT)
	    && in_atomic_preempt_off()) {
3254
		pr_err("Preemption disabled at:");
3255
		print_ip_sym(preempt_disable_ip);
3256 3257
		pr_cont("\n");
	}
3258 3259 3260
	if (panic_on_warn)
		panic("scheduling while atomic\n");

3261
	dump_stack();
3262
	add_taint(TAINT_WARN, LOCKDEP_STILL_OK);
Ingo Molnar's avatar
Ingo Molnar committed
3263
}
Linus Torvalds's avatar
Linus Torvalds committed
3264

Ingo Molnar's avatar
Ingo Molnar committed
3265 3266 3267 3268 3269
/*
 * Various schedule()-time debugging checks and statistics:
 */
static inline void schedule_debug(struct task_struct *prev)
{
3270
#ifdef CONFIG_SCHED_STACK_END_CHECK
3271 3272
	if (task_stack_end_corrupted(prev))
		panic("corrupted stack end detected inside scheduler\n");
3273
#endif
3274

3275
	if (unlikely(in_atomic_preempt_off())) {
Ingo Molnar's avatar
Ingo Molnar committed
3276
		__schedule_bug(prev);
3277 3278
		preempt_count_set(PREEMPT_DISABLED);
	}
3279
	rcu_sleep_check();
Ingo Molnar's avatar
Ingo Molnar committed
3280

Linus Torvalds's avatar
Linus Torvalds committed
3281 3282
	profile_hit(SCHED_PROFILING, __builtin_return_address(0));

3283
	schedstat_inc(this_rq()->sched_count);
Ingo Molnar's avatar
Ingo Molnar committed
3284 3285 3286 3287 3288 3289
}

/*
 * Pick up the highest-prio task:
 */
static inline struct task_struct *
3290
pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
Ingo Molnar's avatar
Ingo Molnar committed
3291
{
3292
	const struct sched_class *class;
Ingo Molnar's avatar
Ingo Molnar committed
3293
	struct task_struct *p;
Linus Torvalds's avatar
Linus Torvalds committed
3294 3295

	/*
3296 3297 3298 3299
	 * 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.
Linus Torvalds's avatar
Linus Torvalds committed
3300
	 */
3301 3302 3303 3304
	if (likely((prev->sched_class == &idle_sched_class ||
		    prev->sched_class == &fair_sched_class) &&
		   rq->nr_running == rq->cfs.h_nr_running)) {

3305
		p = fair_sched_class.pick_next_task(rq, prev, rf);
3306 3307 3308
		if (unlikely(p == RETRY_TASK))
			goto again;

3309
		/* Assumes fair_sched_class->next == idle_sched_class */
3310
		if (unlikely(!p))
3311
			p = idle_sched_class.pick_next_task(rq, prev, rf);
3312 3313

		return p;
Linus Torvalds's avatar
Linus Torvalds committed
3314 3315
	}

3316
again:
3317
	for_each_class(class) {
3318
		p = class->pick_next_task(rq, prev, rf);
3319 3320 3321
		if (p) {
			if (unlikely(p == RETRY_TASK))
				goto again;
Ingo Molnar's avatar
Ingo Molnar committed
3322
			return p;
3323
		}
Ingo Molnar's avatar
Ingo Molnar committed
3324
	}
3325

3326 3327
	/* The idle class should always have a runnable task: */
	BUG();
Ingo Molnar's avatar
Ingo Molnar committed
3328
}
Linus Torvalds's avatar
Linus Torvalds committed
3329

Ingo Molnar's avatar
Ingo Molnar committed
3330
/*
3331
 * __schedule() is the main scheduler function.
3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365
 *
 * 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
3366
 *
3367
 * WARNING: must be called with preemption disabled!
Ingo Molnar's avatar
Ingo Molnar committed
3368
 */
3369
static void __sched notrace __schedule(bool preempt)
Ingo Molnar's avatar
Ingo Molnar committed
3370 3371
{
	struct task_struct *prev, *next;
3372
	unsigned long *switch_count;
3373
	struct rq_flags rf;
Ingo Molnar's avatar
Ingo Molnar committed
3374
	struct rq *rq;
3375
	int cpu;
Ingo Molnar's avatar
Ingo Molnar committed
3376 3377 3378 3379 3380 3381

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

	schedule_debug(prev);
Linus Torvalds's avatar
Linus Torvalds committed
3382

3383
	if (sched_feat(HRTICK))
Mike Galbraith's avatar
Mike Galbraith committed
3384
		hrtick_clear(rq);
3385

3386
	local_irq_disable();
3387
	rcu_note_context_switch(preempt);
3388

3389 3390 3391 3392
	/*
	 * 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().
3393 3394 3395
	 *
	 * The membarrier system call requires a full memory barrier
	 * after coming from user-space, before storing to rq->curr.
3396
	 */
3397
	rq_lock(rq, &rf);
3398
	smp_mb__after_spinlock();
Linus Torvalds's avatar
Linus Torvalds committed
3399

3400 3401
	/* Promote REQ to ACT */
	rq->clock_update_flags <<= 1;
3402
	update_rq_clock(rq);
3403

3404
	switch_count = &prev->nivcsw;
3405
	if (!preempt && prev->state) {
3406
		if (signal_pending_state(prev->state, prev)) {
Linus Torvalds's avatar
Linus Torvalds committed
3407
			prev->state = TASK_RUNNING;
Tejun Heo's avatar
Tejun Heo committed
3408
		} else {
3409
			deactivate_task(rq, prev, DEQUEUE_SLEEP | DEQUEUE_NOCLOCK);
3410 3411
			prev->on_rq = 0;

3412 3413 3414 3415
			if (prev->in_iowait) {
				atomic_inc(&rq->nr_iowait);
				delayacct_blkio_start();
			}
Tejun Heo's avatar
Tejun Heo committed
3416
		}
Ingo Molnar's avatar
Ingo Molnar committed
3417
		switch_count = &prev->nvcsw;
Linus Torvalds's avatar
Linus Torvalds committed
3418 3419
	}

3420
	next = pick_next_task(rq, prev, &rf);
3421
	clear_tsk_need_resched(prev);
3422
	clear_preempt_need_resched();
Linus Torvalds's avatar
Linus Torvalds committed
3423 3424 3425 3426

	if (likely(prev != next)) {
		rq->nr_switches++;
		rq->curr = next;
3427 3428 3429
		/*
		 * The membarrier system call requires each architecture
		 * to have a full memory barrier after updating
3430 3431 3432 3433 3434 3435 3436 3437 3438 3439
		 * rq->curr, before returning to user-space.
		 *
		 * Here are the schemes providing that barrier on the
		 * various architectures:
		 * - mm ? switch_mm() : mmdrop() for x86, s390, sparc, PowerPC.
		 *   switch_mm() rely on membarrier_arch_switch_mm() on PowerPC.
		 * - finish_lock_switch() for weakly-ordered
		 *   architectures where spin_unlock is a full barrier,
		 * - switch_to() for arm64 (weakly-ordered, spin_unlock
		 *   is a RELEASE barrier),
3440
		 */
Linus Torvalds's avatar
Linus Torvalds committed
3441 3442
		++*switch_count;

3443
		trace_sched_switch(preempt, prev, next);
3444 3445 3446

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

3452
	balance_callback(rq);
Linus Torvalds's avatar
Linus Torvalds committed
3453
}
3454

3455 3456
void __noreturn do_task_dead(void)
{
3457
	/* Causes final put_task_struct in finish_task_switch(): */
3458
	set_special_state(TASK_DEAD);
3459 3460 3461 3462

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

3463 3464
	__schedule(false);
	BUG();
3465 3466

	/* Avoid "noreturn function does return" - but don't continue if BUG() is a NOP: */
3467
	for (;;)
3468
		cpu_relax();
3469 3470
}

3471 3472
static inline void sched_submit_work(struct task_struct *tsk)
{
3473
	if (!tsk->state || tsk_is_pi_blocked(tsk))
3474
		return;
3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488

	/*
	 * If a worker went to sleep, notify and ask workqueue whether
	 * it wants to wake up a task to maintain concurrency.
	 * As this function is called inside the schedule() context,
	 * we disable preemption to avoid it calling schedule() again
	 * in the possible wakeup of a kworker.
	 */
	if (tsk->flags & PF_WQ_WORKER) {
		preempt_disable();
		wq_worker_sleeping(tsk);
		preempt_enable_no_resched();
	}

3489 3490 3491 3492 3493 3494 3495 3496
	/*
	 * 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);
}

3497 3498 3499 3500 3501 3502
static void sched_update_worker(struct task_struct *tsk)
{
	if (tsk->flags & PF_WQ_WORKER)
		wq_worker_running(tsk);
}

3503
asmlinkage __visible void __sched schedule(void)
3504
{
3505 3506 3507
	struct task_struct *tsk = current;

	sched_submit_work(tsk);
3508
	do {
3509
		preempt_disable();
3510
		__schedule(false);
3511
		sched_preempt_enable_no_resched();
3512
	} while (need_resched());
3513
	sched_update_worker(tsk);
3514
}
Linus Torvalds's avatar
Linus Torvalds committed
3515 3516
EXPORT_SYMBOL(schedule);

3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541
/*
 * 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());
}

3542
#ifdef CONFIG_CONTEXT_TRACKING
3543
asmlinkage __visible void __sched schedule_user(void)
3544 3545 3546 3547 3548 3549
{
	/*
	 * 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.
3550 3551
	 *
	 * NB: There are buggy callers of this function.  Ideally we
3552
	 * should warn if prev_state != CONTEXT_USER, but that will trigger
3553
	 * too frequently to make sense yet.
3554
	 */
3555
	enum ctx_state prev_state = exception_enter();
3556
	schedule();
3557
	exception_exit(prev_state);
3558 3559 3560
}
#endif

3561 3562 3563 3564 3565 3566 3567
/**
 * schedule_preempt_disabled - called with preemption disabled
 *
 * Returns with preemption disabled. Note: preempt_count must be 1
 */
void __sched schedule_preempt_disabled(void)
{
3568
	sched_preempt_enable_no_resched();
3569 3570 3571 3572
	schedule();
	preempt_disable();
}

3573
static void __sched notrace preempt_schedule_common(void)
3574 3575
{
	do {
3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588
		/*
		 * 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.
		 */
3589
		preempt_disable_notrace();
3590
		preempt_latency_start(1);
3591
		__schedule(true);
3592
		preempt_latency_stop(1);
3593
		preempt_enable_no_resched_notrace();
3594 3595 3596 3597 3598 3599 3600 3601

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

Linus Torvalds's avatar
Linus Torvalds committed
3602 3603
#ifdef CONFIG_PREEMPT
/*
3604
 * this is the entry point to schedule() from in-kernel preemption
Ingo Molnar's avatar
Ingo Molnar committed
3605
 * off of preempt_enable. Kernel preemptions off return from interrupt
Linus Torvalds's avatar
Linus Torvalds committed
3606 3607
 * occur there and call schedule directly.
 */
3608
asmlinkage __visible void __sched notrace preempt_schedule(void)
Linus Torvalds's avatar
Linus Torvalds committed
3609 3610 3611
{
	/*
	 * If there is a non-zero preempt_count or interrupts are disabled,
Ingo Molnar's avatar
Ingo Molnar committed
3612
	 * we do not want to preempt the current task. Just return..
Linus Torvalds's avatar
Linus Torvalds committed
3613
	 */
3614
	if (likely(!preemptible()))
Linus Torvalds's avatar
Linus Torvalds committed
3615 3616
		return;

3617
	preempt_schedule_common();
Linus Torvalds's avatar
Linus Torvalds committed
3618
}
3619
NOKPROBE_SYMBOL(preempt_schedule);
Linus Torvalds's avatar
Linus Torvalds committed
3620
EXPORT_SYMBOL(preempt_schedule);
3621 3622

/**
3623
 * preempt_schedule_notrace - preempt_schedule called by tracing
3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635
 *
 * 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.
 */
3636
asmlinkage __visible void __sched notrace preempt_schedule_notrace(void)
3637 3638 3639 3640 3641 3642 3643
{
	enum ctx_state prev_ctx;

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

	do {
3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656
		/*
		 * 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.
		 */
3657
		preempt_disable_notrace();
3658
		preempt_latency_start(1);
3659 3660 3661 3662 3663 3664
		/*
		 * Needs preempt disabled in case user_exit() is traced
		 * and the tracer calls preempt_enable_notrace() causing
		 * an infinite recursion.
		 */
		prev_ctx = exception_enter();
3665
		__schedule(true);
3666 3667
		exception_exit(prev_ctx);

3668
		preempt_latency_stop(1);
3669
		preempt_enable_no_resched_notrace();
3670 3671
	} while (need_resched());
}
3672
EXPORT_SYMBOL_GPL(preempt_schedule_notrace);
3673

3674
#endif /* CONFIG_PREEMPT */
Linus Torvalds's avatar
Linus Torvalds committed
3675 3676

/*
3677
 * this is the entry point to schedule() from kernel preemption
Linus Torvalds's avatar
Linus Torvalds committed
3678 3679 3680 3681
 * off of irq context.
 * Note, that this is called and return with irqs disabled. This will
 * protect us against recursive calling from irq.
 */
3682
asmlinkage __visible void __sched preempt_schedule_irq(void)
Linus Torvalds's avatar
Linus Torvalds committed
3683
{
3684
	enum ctx_state prev_state;
3685

3686
	/* Catch callers which need to be fixed */
3687
	BUG_ON(preempt_count() || !irqs_disabled());
Linus Torvalds's avatar
Linus Torvalds committed
3688

3689 3690
	prev_state = exception_enter();

3691
	do {
3692
		preempt_disable();
3693
		local_irq_enable();
3694
		__schedule(true);
3695
		local_irq_disable();
3696
		sched_preempt_enable_no_resched();
3697
	} while (need_resched());
3698 3699

	exception_exit(prev_state);
Linus Torvalds's avatar
Linus Torvalds committed
3700 3701
}

3702
int default_wake_function(wait_queue_entry_t *curr, unsigned mode, int wake_flags,
Ingo Molnar's avatar
Ingo Molnar committed
3703
			  void *key)
Linus Torvalds's avatar
Linus Torvalds committed
3704
{
3705
	return try_to_wake_up(curr->private, mode, wake_flags);
Linus Torvalds's avatar
Linus Torvalds committed
3706 3707 3708
}
EXPORT_SYMBOL(default_wake_function);

3709 3710
#ifdef CONFIG_RT_MUTEXES

3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725
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);
}

3726 3727
/*
 * rt_mutex_setprio - set the current priority of a task
3728 3729
 * @p: task to boost
 * @pi_task: donor task
3730 3731 3732 3733
 *
 * This function changes the 'effective' priority of a task. It does
 * not touch ->normal_prio like __setscheduler().
 *
3734 3735
 * Used by the rt_mutex code to implement priority inheritance
 * logic. Call site only calls if the priority of the task changed.
3736
 */
3737
void rt_mutex_setprio(struct task_struct *p, struct task_struct *pi_task)
3738
{
3739
	int prio, oldprio, queued, running, queue_flag =
3740
		DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
3741
	const struct sched_class *prev_class;
3742 3743
	struct rq_flags rf;
	struct rq *rq;
3744

3745 3746 3747 3748 3749 3750 3751 3752
	/* 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;
3753

3754
	rq = __task_rq_lock(p, &rf);
3755
	update_rq_clock(rq);
3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772
	/*
	 * 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;
3773

3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791
	/*
	 * 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;
	}

3792
	trace_sched_pi_setprio(p, pi_task);
3793
	oldprio = p->prio;
3794 3795 3796 3797

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

3798
	prev_class = p->sched_class;
3799
	queued = task_on_rq_queued(p);
3800
	running = task_current(rq, p);
3801
	if (queued)
3802
		dequeue_task(rq, p, queue_flag);
3803
	if (running)
3804
		put_prev_task(rq, p);
Ingo Molnar's avatar
Ingo Molnar committed
3805

3806 3807 3808 3809 3810 3811 3812 3813 3814 3815
	/*
	 * 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)) {
3816 3817
		if (!dl_prio(p->normal_prio) ||
		    (pi_task && dl_entity_preempt(&pi_task->dl, &p->dl))) {
3818
			p->dl.dl_boosted = 1;
3819
			queue_flag |= ENQUEUE_REPLENISH;
3820 3821
		} else
			p->dl.dl_boosted = 0;
3822
		p->sched_class = &dl_sched_class;
3823 3824 3825 3826
	} else if (rt_prio(prio)) {
		if (dl_prio(oldprio))
			p->dl.dl_boosted = 0;
		if (oldprio < prio)
3827
			queue_flag |= ENQUEUE_HEAD;
Ingo Molnar's avatar
Ingo Molnar committed
3828
		p->sched_class = &rt_sched_class;
3829 3830 3831
	} else {
		if (dl_prio(oldprio))
			p->dl.dl_boosted = 0;
3832 3833
		if (rt_prio(oldprio))
			p->rt.timeout = 0;
Ingo Molnar's avatar
Ingo Molnar committed
3834
		p->sched_class = &fair_sched_class;
3835
	}
Ingo Molnar's avatar
Ingo Molnar committed
3836

3837 3838
	p->prio = prio;

3839
	if (queued)
3840
		enqueue_task(rq, p, queue_flag);
3841
	if (running)
3842
		set_curr_task(rq, p);
3843

3844
	check_class_changed(rq, p, prev_class, oldprio);
3845
out_unlock:
3846 3847
	/* Avoid rq from going away on us: */
	preempt_disable();
3848
	__task_rq_unlock(rq, &rf);
3849 3850 3851

	balance_callback(rq);
	preempt_enable();
3852
}
3853 3854 3855 3856 3857
#else
static inline int rt_effective_prio(struct task_struct *p, int prio)
{
	return prio;
}
3858
#endif
3859

3860
void set_user_nice(struct task_struct *p, long nice)
Linus Torvalds's avatar
Linus Torvalds committed
3861
{
3862 3863
	bool queued, running;
	int old_prio, delta;
3864
	struct rq_flags rf;
3865
	struct rq *rq;
Linus Torvalds's avatar
Linus Torvalds committed
3866

3867
	if (task_nice(p) == nice || nice < MIN_NICE || nice > MAX_NICE)
Linus Torvalds's avatar
Linus Torvalds committed
3868 3869 3870 3871 3872
		return;
	/*
	 * We have to be careful, if called from sys_setpriority(),
	 * the task might be in the middle of scheduling on another CPU.
	 */
3873
	rq = task_rq_lock(p, &rf);
3874 3875
	update_rq_clock(rq);

Linus Torvalds's avatar
Linus Torvalds committed
3876 3877 3878 3879
	/*
	 * 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
3880
	 * SCHED_DEADLINE, SCHED_FIFO or SCHED_RR:
Linus Torvalds's avatar
Linus Torvalds committed
3881
	 */
3882
	if (task_has_dl_policy(p) || task_has_rt_policy(p)) {
Linus Torvalds's avatar
Linus Torvalds committed
3883 3884 3885
		p->static_prio = NICE_TO_PRIO(nice);
		goto out_unlock;
	}
3886
	queued = task_on_rq_queued(p);
3887
	running = task_current(rq, p);
3888
	if (queued)
3889
		dequeue_task(rq, p, DEQUEUE_SAVE | DEQUEUE_NOCLOCK);
3890 3891
	if (running)
		put_prev_task(rq, p);
Linus Torvalds's avatar
Linus Torvalds committed
3892 3893

	p->static_prio = NICE_TO_PRIO(nice);
3894
	set_load_weight(p, true);
3895 3896 3897
	old_prio = p->prio;
	p->prio = effective_prio(p);
	delta = p->prio - old_prio;
Linus Torvalds's avatar
Linus Torvalds committed
3898

3899
	if (queued) {
3900
		enqueue_task(rq, p, ENQUEUE_RESTORE | ENQUEUE_NOCLOCK);
Linus Torvalds's avatar
Linus Torvalds committed
3901
		/*
3902 3903
		 * If the task increased its priority or is running and
		 * lowered its priority, then reschedule its CPU:
Linus Torvalds's avatar
Linus Torvalds committed
3904
		 */
3905
		if (delta < 0 || (delta > 0 && task_running(rq, p)))
3906
			resched_curr(rq);
Linus Torvalds's avatar
Linus Torvalds committed
3907
	}
3908 3909
	if (running)
		set_curr_task(rq, p);
Linus Torvalds's avatar
Linus Torvalds committed
3910
out_unlock:
3911
	task_rq_unlock(rq, p, &rf);
Linus Torvalds's avatar
Linus Torvalds committed
3912 3913 3914
}
EXPORT_SYMBOL(set_user_nice);

3915 3916 3917 3918 3919
/*
 * can_nice - check if a task can reduce its nice value
 * @p: task
 * @nice: nice value
 */
3920
int can_nice(const struct task_struct *p, const int nice)
3921
{
3922
	/* Convert nice value [19,-20] to rlimit style value [1,40]: */
3923
	int nice_rlim = nice_to_rlimit(nice);
3924

3925
	return (nice_rlim <= task_rlimit(p, RLIMIT_NICE) ||
3926 3927 3928
		capable(CAP_SYS_NICE));
}

Linus Torvalds's avatar
Linus Torvalds committed
3929 3930 3931 3932 3933 3934 3935 3936 3937
#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.
 */
3938
SYSCALL_DEFINE1(nice, int, increment)
Linus Torvalds's avatar
Linus Torvalds committed
3939
{
3940
	long nice, retval;
Linus Torvalds's avatar
Linus Torvalds committed
3941 3942 3943 3944 3945 3946

	/*
	 * 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.
	 */
3947
	increment = clamp(increment, -NICE_WIDTH, NICE_WIDTH);
3948
	nice = task_nice(current) + increment;
Linus Torvalds's avatar
Linus Torvalds committed
3949

3950
	nice = clamp_val(nice, MIN_NICE, MAX_NICE);
3951 3952 3953
	if (increment < 0 && !can_nice(current, nice))
		return -EPERM;

Linus Torvalds's avatar
Linus Torvalds committed
3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967
	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.
 *
3968
 * Return: The priority value as seen by users in /proc.
Linus Torvalds's avatar
Linus Torvalds committed
3969 3970 3971
 * RT tasks are offset by -200. Normal tasks are centered
 * around 0, value goes from -16 to +15.
 */
3972
int task_prio(const struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
3973 3974 3975 3976 3977
{
	return p->prio - MAX_RT_PRIO;
}

/**
3978
 * idle_cpu - is a given CPU idle currently?
Linus Torvalds's avatar
Linus Torvalds committed
3979
 * @cpu: the processor in question.
3980 3981
 *
 * Return: 1 if the CPU is currently idle. 0 otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
3982 3983 3984
 */
int idle_cpu(int cpu)
{
Thomas Gleixner's avatar
Thomas Gleixner committed
3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998
	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;
Linus Torvalds's avatar
Linus Torvalds committed
3999 4000
}

4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011
/**
 * available_idle_cpu - is a given CPU idle for enqueuing work.
 * @cpu: the CPU in question.
 *
 * Return: 1 if the CPU is currently idle. 0 otherwise.
 */
int available_idle_cpu(int cpu)
{
	if (!idle_cpu(cpu))
		return 0;

4012 4013 4014
	if (vcpu_is_preempted(cpu))
		return 0;

Thomas Gleixner's avatar
Thomas Gleixner committed
4015
	return 1;
Linus Torvalds's avatar
Linus Torvalds committed
4016 4017 4018
}

/**
4019
 * idle_task - return the idle task for a given CPU.
Linus Torvalds's avatar
Linus Torvalds committed
4020
 * @cpu: the processor in question.
4021
 *
4022
 * Return: The idle task for the CPU @cpu.
Linus Torvalds's avatar
Linus Torvalds committed
4023
 */
4024
struct task_struct *idle_task(int cpu)
Linus Torvalds's avatar
Linus Torvalds committed
4025 4026 4027 4028 4029 4030 4031
{
	return cpu_rq(cpu)->idle;
}

/**
 * find_process_by_pid - find a process with a matching PID value.
 * @pid: the pid in question.
4032 4033
 *
 * The task of @pid, if found. %NULL otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
4034
 */
Alexey Dobriyan's avatar
Alexey Dobriyan committed
4035
static struct task_struct *find_process_by_pid(pid_t pid)
Linus Torvalds's avatar
Linus Torvalds committed
4036
{
4037
	return pid ? find_task_by_vpid(pid) : current;
Linus Torvalds's avatar
Linus Torvalds committed
4038 4039
}

4040 4041 4042 4043 4044 4045
/*
 * sched_setparam() passes in -1 for its policy, to let the functions
 * it calls know not to change it.
 */
#define SETPARAM_POLICY	-1

4046 4047
static void __setscheduler_params(struct task_struct *p,
		const struct sched_attr *attr)
Linus Torvalds's avatar
Linus Torvalds committed
4048
{
4049 4050
	int policy = attr->sched_policy;

4051
	if (policy == SETPARAM_POLICY)
4052 4053
		policy = p->policy;

Linus Torvalds's avatar
Linus Torvalds committed
4054
	p->policy = policy;
4055

4056 4057
	if (dl_policy(policy))
		__setparam_dl(p, attr);
4058
	else if (fair_policy(policy))
4059 4060
		p->static_prio = NICE_TO_PRIO(attr->sched_nice);

4061 4062 4063 4064 4065 4066
	/*
	 * __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;
4067
	p->normal_prio = normal_prio(p);
4068
	set_load_weight(p, true);
4069
}
4070

4071 4072
/* Actually do priority change: must hold pi & rq lock. */
static void __setscheduler(struct rq *rq, struct task_struct *p,
4073
			   const struct sched_attr *attr, bool keep_boost)
4074 4075
{
	__setscheduler_params(p, attr);
4076

4077
	/*
4078 4079
	 * Keep a potential priority boosting if called from
	 * sched_setscheduler().
4080
	 */
4081
	p->prio = normal_prio(p);
4082
	if (keep_boost)
4083
		p->prio = rt_effective_prio(p, p->prio);
4084

4085 4086 4087
	if (dl_prio(p->prio))
		p->sched_class = &dl_sched_class;
	else if (rt_prio(p->prio))
4088 4089 4090
		p->sched_class = &rt_sched_class;
	else
		p->sched_class = &fair_sched_class;
Linus Torvalds's avatar
Linus Torvalds committed
4091
}
4092

4093
/*
4094
 * Check the target process has a UID that matches the current process's:
4095 4096 4097 4098 4099 4100 4101 4102
 */
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);
4103 4104
	match = (uid_eq(cred->euid, pcred->euid) ||
		 uid_eq(cred->euid, pcred->uid));
4105 4106 4107 4108
	rcu_read_unlock();
	return match;
}

4109 4110
static int __sched_setscheduler(struct task_struct *p,
				const struct sched_attr *attr,
4111
				bool user, bool pi)
Linus Torvalds's avatar
Linus Torvalds committed
4112
{
4113 4114
	int newprio = dl_policy(attr->sched_policy) ? MAX_DL_PRIO - 1 :
		      MAX_RT_PRIO - 1 - attr->sched_priority;
4115
	int retval, oldprio, oldpolicy = -1, queued, running;
4116
	int new_effective_prio, policy = attr->sched_policy;
4117
	const struct sched_class *prev_class;
4118
	struct rq_flags rf;
4119
	int reset_on_fork;
4120
	int queue_flags = DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
4121
	struct rq *rq;
Linus Torvalds's avatar
Linus Torvalds committed
4122

4123 4124
	/* The pi code expects interrupts enabled */
	BUG_ON(pi && in_interrupt());
Linus Torvalds's avatar
Linus Torvalds committed
4125
recheck:
4126
	/* Double check policy once rq lock held: */
4127 4128
	if (policy < 0) {
		reset_on_fork = p->sched_reset_on_fork;
Linus Torvalds's avatar
Linus Torvalds committed
4129
		policy = oldpolicy = p->policy;
4130
	} else {
4131
		reset_on_fork = !!(attr->sched_flags & SCHED_FLAG_RESET_ON_FORK);
4132

4133
		if (!valid_policy(policy))
4134 4135 4136
			return -EINVAL;
	}

4137
	if (attr->sched_flags & ~(SCHED_FLAG_ALL | SCHED_FLAG_SUGOV))
4138 4139
		return -EINVAL;

Linus Torvalds's avatar
Linus Torvalds committed
4140 4141
	/*
	 * Valid priorities for SCHED_FIFO and SCHED_RR are
Ingo Molnar's avatar
Ingo Molnar committed
4142 4143
	 * 1..MAX_USER_RT_PRIO-1, valid priority for SCHED_NORMAL,
	 * SCHED_BATCH and SCHED_IDLE is 0.
Linus Torvalds's avatar
Linus Torvalds committed
4144
	 */
4145
	if ((p->mm && attr->sched_priority > MAX_USER_RT_PRIO-1) ||
4146
	    (!p->mm && attr->sched_priority > MAX_RT_PRIO-1))
Linus Torvalds's avatar
Linus Torvalds committed
4147
		return -EINVAL;
4148 4149
	if ((dl_policy(policy) && !__checkparam_dl(attr)) ||
	    (rt_policy(policy) != (attr->sched_priority != 0)))
Linus Torvalds's avatar
Linus Torvalds committed
4150 4151
		return -EINVAL;

4152 4153 4154
	/*
	 * Allow unprivileged RT tasks to decrease priority:
	 */
4155
	if (user && !capable(CAP_SYS_NICE)) {
4156
		if (fair_policy(policy)) {
4157
			if (attr->sched_nice < task_nice(p) &&
4158
			    !can_nice(p, attr->sched_nice))
4159 4160 4161
				return -EPERM;
		}

4162
		if (rt_policy(policy)) {
4163 4164
			unsigned long rlim_rtprio =
					task_rlimit(p, RLIMIT_RTPRIO);
4165

4166
			/* Can't set/change the rt policy: */
4167 4168 4169
			if (policy != p->policy && !rlim_rtprio)
				return -EPERM;

4170
			/* Can't increase priority: */
4171 4172
			if (attr->sched_priority > p->rt_priority &&
			    attr->sched_priority > rlim_rtprio)
4173 4174
				return -EPERM;
		}
4175

4176 4177 4178 4179 4180 4181 4182 4183 4184
		 /*
		  * 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;

Ingo Molnar's avatar
Ingo Molnar committed
4185
		/*
4186 4187
		 * Treat SCHED_IDLE as nice 20. Only allow a switch to
		 * SCHED_NORMAL if the RLIMIT_NICE would normally permit it.
Ingo Molnar's avatar
Ingo Molnar committed
4188
		 */
4189
		if (task_has_idle_policy(p) && !idle_policy(policy)) {
4190
			if (!can_nice(p, task_nice(p)))
4191 4192
				return -EPERM;
		}
4193

4194
		/* Can't change other user's priorities: */
4195
		if (!check_same_owner(p))
4196
			return -EPERM;
4197

4198
		/* Normal users shall not reset the sched_reset_on_fork flag: */
4199 4200
		if (p->sched_reset_on_fork && !reset_on_fork)
			return -EPERM;
4201
	}
Linus Torvalds's avatar
Linus Torvalds committed
4202

4203
	if (user) {
4204 4205 4206
		if (attr->sched_flags & SCHED_FLAG_SUGOV)
			return -EINVAL;

4207
		retval = security_task_setscheduler(p);
4208 4209 4210 4211
		if (retval)
			return retval;
	}

4212
	/*
4213
	 * Make sure no PI-waiters arrive (or leave) while we are
4214
	 * changing the priority of the task:
4215
	 *
Lucas De Marchi's avatar
Lucas De Marchi committed
4216
	 * To be able to change p->policy safely, the appropriate
Linus Torvalds's avatar
Linus Torvalds committed
4217 4218
	 * runqueue lock must be held.
	 */
4219
	rq = task_rq_lock(p, &rf);
4220
	update_rq_clock(rq);
4221

4222
	/*
4223
	 * Changing the policy of the stop threads its a very bad idea:
4224 4225
	 */
	if (p == rq->stop) {
4226
		task_rq_unlock(rq, p, &rf);
4227 4228 4229
		return -EINVAL;
	}

4230
	/*
4231 4232
	 * If not changing anything there's no need to proceed further,
	 * but store a possible modification of reset_on_fork.
4233
	 */
4234
	if (unlikely(policy == p->policy)) {
4235
		if (fair_policy(policy) && attr->sched_nice != task_nice(p))
4236 4237 4238
			goto change;
		if (rt_policy(policy) && attr->sched_priority != p->rt_priority)
			goto change;
4239
		if (dl_policy(policy) && dl_param_changed(p, attr))
4240
			goto change;
4241

4242
		p->sched_reset_on_fork = reset_on_fork;
4243
		task_rq_unlock(rq, p, &rf);
4244 4245
		return 0;
	}
4246
change:
4247

4248
	if (user) {
4249
#ifdef CONFIG_RT_GROUP_SCHED
4250 4251 4252 4253 4254
		/*
		 * Do not allow realtime tasks into groups that have no runtime
		 * assigned.
		 */
		if (rt_bandwidth_enabled() && rt_policy(policy) &&
4255 4256
				task_group(p)->rt_bandwidth.rt_runtime == 0 &&
				!task_group_is_autogroup(task_group(p))) {
4257
			task_rq_unlock(rq, p, &rf);
4258 4259 4260
			return -EPERM;
		}
#endif
4261
#ifdef CONFIG_SMP
4262 4263
		if (dl_bandwidth_enabled() && dl_policy(policy) &&
				!(attr->sched_flags & SCHED_FLAG_SUGOV)) {
4264 4265 4266 4267 4268 4269 4270
			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.
			 */
4271 4272
			if (!cpumask_subset(span, &p->cpus_allowed) ||
			    rq->rd->dl_bw.bw == 0) {
4273
				task_rq_unlock(rq, p, &rf);
4274 4275 4276 4277 4278
				return -EPERM;
			}
		}
#endif
	}
4279

4280
	/* Re-check policy now with rq lock held: */
Linus Torvalds's avatar
Linus Torvalds committed
4281 4282
	if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) {
		policy = oldpolicy = -1;
4283
		task_rq_unlock(rq, p, &rf);
Linus Torvalds's avatar
Linus Torvalds committed
4284 4285
		goto recheck;
	}
4286 4287 4288 4289 4290 4291

	/*
	 * If setscheduling to SCHED_DEADLINE (or changing the parameters
	 * of a SCHED_DEADLINE task) we need to check if enough bandwidth
	 * is available.
	 */
4292
	if ((dl_policy(policy) || dl_task(p)) && sched_dl_overflow(p, policy, attr)) {
4293
		task_rq_unlock(rq, p, &rf);
4294 4295 4296
		return -EBUSY;
	}

4297 4298 4299
	p->sched_reset_on_fork = reset_on_fork;
	oldprio = p->prio;

4300 4301 4302 4303 4304 4305 4306 4307
	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.
		 */
4308
		new_effective_prio = rt_effective_prio(p, newprio);
4309 4310
		if (new_effective_prio == oldprio)
			queue_flags &= ~DEQUEUE_MOVE;
4311 4312
	}

4313
	queued = task_on_rq_queued(p);
4314
	running = task_current(rq, p);
4315
	if (queued)
4316
		dequeue_task(rq, p, queue_flags);
4317
	if (running)
4318
		put_prev_task(rq, p);
4319

4320
	prev_class = p->sched_class;
4321
	__setscheduler(rq, p, attr, pi);
4322

4323
	if (queued) {
4324 4325 4326 4327
		/*
		 * We enqueue to tail when the priority of a task is
		 * increased (user space view).
		 */
4328 4329
		if (oldprio < p->prio)
			queue_flags |= ENQUEUE_HEAD;
4330

4331
		enqueue_task(rq, p, queue_flags);
4332
	}
4333
	if (running)
4334
		set_curr_task(rq, p);
4335

4336
	check_class_changed(rq, p, prev_class, oldprio);
4337 4338 4339

	/* Avoid rq from going away on us: */
	preempt_disable();
4340
	task_rq_unlock(rq, p, &rf);
4341

4342 4343
	if (pi)
		rt_mutex_adjust_pi(p);
4344

4345
	/* Run balance callbacks after we've adjusted the PI chain: */
4346 4347
	balance_callback(rq);
	preempt_enable();
4348

Linus Torvalds's avatar
Linus Torvalds committed
4349 4350
	return 0;
}
4351

4352 4353 4354 4355 4356 4357 4358 4359 4360
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),
	};

4361 4362
	/* Fixup the legacy SCHED_RESET_ON_FORK hack. */
	if ((policy != SETPARAM_POLICY) && (policy & SCHED_RESET_ON_FORK)) {
4363 4364 4365 4366 4367
		attr.sched_flags |= SCHED_FLAG_RESET_ON_FORK;
		policy &= ~SCHED_RESET_ON_FORK;
		attr.sched_policy = policy;
	}

4368
	return __sched_setscheduler(p, &attr, check, true);
4369
}
4370 4371 4372 4373 4374 4375
/**
 * 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.
 *
4376 4377
 * Return: 0 on success. An error code otherwise.
 *
4378 4379 4380
 * NOTE that the task may be already dead.
 */
int sched_setscheduler(struct task_struct *p, int policy,
4381
		       const struct sched_param *param)
4382
{
4383
	return _sched_setscheduler(p, policy, param, true);
4384
}
Linus Torvalds's avatar
Linus Torvalds committed
4385 4386
EXPORT_SYMBOL_GPL(sched_setscheduler);

4387 4388
int sched_setattr(struct task_struct *p, const struct sched_attr *attr)
{
4389
	return __sched_setscheduler(p, attr, true, true);
4390 4391 4392
}
EXPORT_SYMBOL_GPL(sched_setattr);

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

4398 4399 4400 4401 4402 4403 4404 4405 4406 4407
/**
 * sched_setscheduler_nocheck - change the scheduling policy and/or RT priority of a thread from kernelspace.
 * @p: the task in question.
 * @policy: new policy.
 * @param: structure containing the new RT priority.
 *
 * Just like sched_setscheduler, only don't bother checking if the
 * current context has permission.  For example, this is needed in
 * stop_machine(): we create temporary high priority worker threads,
 * but our caller might not have that capability.
4408 4409
 *
 * Return: 0 on success. An error code otherwise.
4410 4411
 */
int sched_setscheduler_nocheck(struct task_struct *p, int policy,
4412
			       const struct sched_param *param)
4413
{
4414
	return _sched_setscheduler(p, policy, param, false);
4415
}
4416
EXPORT_SYMBOL_GPL(sched_setscheduler_nocheck);
4417

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

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

	rcu_read_lock();
	retval = -ESRCH;
Linus Torvalds's avatar
Linus Torvalds committed
4432
	p = find_process_by_pid(pid);
4433 4434 4435
	if (p != NULL)
		retval = sched_setscheduler(p, policy, &lparam);
	rcu_read_unlock();
4436

Linus Torvalds's avatar
Linus Torvalds committed
4437 4438 4439
	return retval;
}

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

4448
	if (!access_ok(uattr, SCHED_ATTR_SIZE_VER0))
4449 4450
		return -EFAULT;

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

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

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

4462 4463
	/* ABI compatibility quirk: */
	if (!size)
4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497
		size = SCHED_ATTR_SIZE_VER0;

	if (size < SCHED_ATTR_SIZE_VER0)
		goto err_size;

	/*
	 * If we're handed a bigger struct than we know of,
	 * ensure all the unknown bits are 0 - i.e. new
	 * user-space does not rely on any kernel feature
	 * extensions we dont know about yet.
	 */
	if (size > sizeof(*attr)) {
		unsigned char __user *addr;
		unsigned char __user *end;
		unsigned char val;

		addr = (void __user *)uattr + sizeof(*attr);
		end  = (void __user *)uattr + size;

		for (; addr < end; addr++) {
			ret = get_user(val, addr);
			if (ret)
				return ret;
			if (val)
				goto err_size;
		}
		size = sizeof(*attr);
	}

	ret = copy_from_user(attr, uattr, size);
	if (ret)
		return -EFAULT;

	/*
4498
	 * XXX: Do we want to be lenient like existing syscalls; or do we want
4499 4500
	 * to be strict and return an error on out-of-bounds values?
	 */
4501
	attr->sched_nice = clamp(attr->sched_nice, MIN_NICE, MAX_NICE);
4502

4503
	return 0;
4504 4505 4506

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

Linus Torvalds's avatar
Linus Torvalds committed
4510 4511 4512 4513 4514
/**
 * sys_sched_setscheduler - set/change the scheduler policy and RT priority
 * @pid: the pid in question.
 * @policy: new policy.
 * @param: structure containing the new RT priority.
4515 4516
 *
 * Return: 0 on success. An error code otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
4517
 */
4518
SYSCALL_DEFINE3(sched_setscheduler, pid_t, pid, int, policy, struct sched_param __user *, param)
Linus Torvalds's avatar
Linus Torvalds committed
4519
{
4520 4521 4522
	if (policy < 0)
		return -EINVAL;

Linus Torvalds's avatar
Linus Torvalds committed
4523 4524 4525 4526 4527 4528 4529
	return do_sched_setscheduler(pid, policy, param);
}

/**
 * sys_sched_setparam - set/change the RT priority of a thread
 * @pid: the pid in question.
 * @param: structure containing the new RT priority.
4530 4531
 *
 * Return: 0 on success. An error code otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
4532
 */
4533
SYSCALL_DEFINE2(sched_setparam, pid_t, pid, struct sched_param __user *, param)
Linus Torvalds's avatar
Linus Torvalds committed
4534
{
4535
	return do_sched_setscheduler(pid, SETPARAM_POLICY, param);
Linus Torvalds's avatar
Linus Torvalds committed
4536 4537
}

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

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

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

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

	rcu_read_lock();
	retval = -ESRCH;
	p = find_process_by_pid(pid);
	if (p != NULL)
		retval = sched_setattr(p, &attr);
	rcu_read_unlock();

	return retval;
}

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

	if (pid < 0)
4584
		return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
4585 4586

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

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

	if (!param || pid < 0)
4614
		return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
4615

4616
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
4617 4618 4619 4620 4621 4622 4623 4624 4625
	p = find_process_by_pid(pid);
	retval = -ESRCH;
	if (!p)
		goto out_unlock;

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

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

	/*
	 * This one might sleep, we cannot do it with a spinlock held ...
	 */
	retval = copy_to_user(param, &lp, sizeof(*param)) ? -EFAULT : 0;

	return retval;

out_unlock:
4638
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
4639 4640 4641
	return retval;
}

4642 4643 4644 4645 4646 4647
static int sched_read_attr(struct sched_attr __user *uattr,
			   struct sched_attr *attr,
			   unsigned int usize)
{
	int ret;

4648
	if (!access_ok(uattr, usize))
4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664
		return -EFAULT;

	/*
	 * If we're handed a smaller struct than we know of,
	 * ensure all the unknown bits are 0 - i.e. old
	 * user-space does not get uncomplete information.
	 */
	if (usize < sizeof(*attr)) {
		unsigned char *addr;
		unsigned char *end;

		addr = (void *)attr + usize;
		end  = (void *)attr + sizeof(*attr);

		for (; addr < end; addr++) {
			if (*addr)
4665
				return -EFBIG;
4666 4667 4668 4669 4670
		}

		attr->size = usize;
	}

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

4675
	return 0;
4676 4677 4678
}

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

	if (!uattr || pid < 0 || size > PAGE_SIZE ||
4695
	    size < SCHED_ATTR_SIZE_VER0 || flags)
4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708
		return -EINVAL;

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

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

	attr.sched_policy = p->policy;
4709 4710
	if (p->sched_reset_on_fork)
		attr.sched_flags |= SCHED_FLAG_RESET_ON_FORK;
4711 4712 4713
	if (task_has_dl_policy(p))
		__getparam_dl(p, &attr);
	else if (task_has_rt_policy(p))
4714 4715
		attr.sched_priority = p->rt_priority;
	else
4716
		attr.sched_nice = task_nice(p);
4717 4718 4719 4720 4721 4722 4723 4724 4725 4726 4727

	rcu_read_unlock();

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

out_unlock:
	rcu_read_unlock();
	return retval;
}

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

4734
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
4735 4736 4737

	p = find_process_by_pid(pid);
	if (!p) {
4738
		rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
4739 4740 4741
		return -ESRCH;
	}

4742
	/* Prevent p going away */
Linus Torvalds's avatar
Linus Torvalds committed
4743
	get_task_struct(p);
4744
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
4745

4746 4747 4748 4749
	if (p->flags & PF_NO_SETAFFINITY) {
		retval = -EINVAL;
		goto out_put_task;
	}
4750 4751 4752 4753 4754 4755 4756 4757
	if (!alloc_cpumask_var(&cpus_allowed, GFP_KERNEL)) {
		retval = -ENOMEM;
		goto out_put_task;
	}
	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL)) {
		retval = -ENOMEM;
		goto out_free_cpus_allowed;
	}
Linus Torvalds's avatar
Linus Torvalds committed
4758
	retval = -EPERM;
4759 4760 4761 4762
	if (!check_same_owner(p)) {
		rcu_read_lock();
		if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
			rcu_read_unlock();
4763
			goto out_free_new_mask;
4764 4765 4766
		}
		rcu_read_unlock();
	}
Linus Torvalds's avatar
Linus Torvalds committed
4767

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

4772 4773 4774 4775

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

4776 4777 4778 4779 4780 4781 4782
	/*
	 * Since bandwidth control happens on root_domain basis,
	 * if admission test is enabled, we only admit -deadline
	 * tasks allowed to run on all the CPUs in the task's
	 * root_domain.
	 */
#ifdef CONFIG_SMP
4783 4784 4785
	if (task_has_dl_policy(p) && dl_bandwidth_enabled()) {
		rcu_read_lock();
		if (!cpumask_subset(task_rq(p)->rd->span, new_mask)) {
4786
			retval = -EBUSY;
4787
			rcu_read_unlock();
4788
			goto out_free_new_mask;
4789
		}
4790
		rcu_read_unlock();
4791 4792
	}
#endif
Peter Zijlstra's avatar
Peter Zijlstra committed
4793
again:
4794
	retval = __set_cpus_allowed_ptr(p, new_mask, true);
Linus Torvalds's avatar
Linus Torvalds committed
4795

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

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

Linus Torvalds's avatar
Linus Torvalds committed
4825 4826 4827 4828
	return copy_from_user(new_mask, user_mask_ptr, len) ? -EFAULT : 0;
}

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

4842 4843
	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL))
		return -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
4844

4845 4846 4847 4848 4849
	retval = get_user_cpu_mask(user_mask_ptr, len, new_mask);
	if (retval == 0)
		retval = sched_setaffinity(pid, new_mask);
	free_cpumask_var(new_mask);
	return retval;
Linus Torvalds's avatar
Linus Torvalds committed
4850 4851
}

4852
long sched_getaffinity(pid_t pid, struct cpumask *mask)
Linus Torvalds's avatar
Linus Torvalds committed
4853
{
4854
	struct task_struct *p;
4855
	unsigned long flags;
Linus Torvalds's avatar
Linus Torvalds committed
4856 4857
	int retval;

4858
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
4859 4860 4861 4862 4863 4864

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

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

4869
	raw_spin_lock_irqsave(&p->pi_lock, flags);
4870
	cpumask_and(mask, &p->cpus_allowed, cpu_active_mask);
4871
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
4872 4873

out_unlock:
4874
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
4875

4876
	return retval;
Linus Torvalds's avatar
Linus Torvalds committed
4877 4878 4879
}

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

4894
	if ((len * BITS_PER_BYTE) < nr_cpu_ids)
4895 4896
		return -EINVAL;
	if (len & (sizeof(unsigned long)-1))
Linus Torvalds's avatar
Linus Torvalds committed
4897 4898
		return -EINVAL;

4899 4900
	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
		return -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
4901

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

		if (copy_to_user(user_mask_ptr, mask, retlen))
4907 4908
			ret = -EFAULT;
		else
4909
			ret = retlen;
4910 4911
	}
	free_cpumask_var(mask);
Linus Torvalds's avatar
Linus Torvalds committed
4912

4913
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
4914 4915 4916 4917 4918
}

/**
 * sys_sched_yield - yield the current processor to other threads.
 *
Ingo Molnar's avatar
Ingo Molnar committed
4919 4920
 * This function yields the current CPU to other tasks. If there are no
 * other threads running on this CPU then this function will return.
4921 4922
 *
 * Return: 0.
Linus Torvalds's avatar
Linus Torvalds committed
4923
 */
4924
static void do_sched_yield(void)
Linus Torvalds's avatar
Linus Torvalds committed
4925
{
4926 4927 4928
	struct rq_flags rf;
	struct rq *rq;

4929
	rq = this_rq_lock_irq(&rf);
Linus Torvalds's avatar
Linus Torvalds committed
4930

4931
	schedstat_inc(rq->yld_count);
4932
	current->sched_class->yield_task(rq);
Linus Torvalds's avatar
Linus Torvalds committed
4933 4934 4935 4936 4937

	/*
	 * Since we are going to call schedule() anyway, there's
	 * no need to preempt or enable interrupts:
	 */
4938 4939
	preempt_disable();
	rq_unlock(rq, &rf);
4940
	sched_preempt_enable_no_resched();
Linus Torvalds's avatar
Linus Torvalds committed
4941 4942

	schedule();
4943
}
Linus Torvalds's avatar
Linus Torvalds committed
4944

4945 4946 4947
SYSCALL_DEFINE0(sched_yield)
{
	do_sched_yield();
Linus Torvalds's avatar
Linus Torvalds committed
4948 4949 4950
	return 0;
}

4951
#ifndef CONFIG_PREEMPT
4952
int __sched _cond_resched(void)
Linus Torvalds's avatar
Linus Torvalds committed
4953
{
4954
	if (should_resched(0)) {
4955
		preempt_schedule_common();
Linus Torvalds's avatar
Linus Torvalds committed
4956 4957
		return 1;
	}
4958
	rcu_all_qs();
Linus Torvalds's avatar
Linus Torvalds committed
4959 4960
	return 0;
}
4961
EXPORT_SYMBOL(_cond_resched);
4962
#endif
Linus Torvalds's avatar
Linus Torvalds committed
4963 4964

/*
4965
 * __cond_resched_lock() - if a reschedule is pending, drop the given lock,
Linus Torvalds's avatar
Linus Torvalds committed
4966 4967
 * call schedule, and on return reacquire the lock.
 *
Ingo Molnar's avatar
Ingo Molnar committed
4968
 * This works OK both with and without CONFIG_PREEMPT. We do strange low-level
Linus Torvalds's avatar
Linus Torvalds committed
4969 4970 4971
 * operations here to prevent schedule() from being called twice (once via
 * spin_unlock(), once by hand).
 */
4972
int __cond_resched_lock(spinlock_t *lock)
Linus Torvalds's avatar
Linus Torvalds committed
4973
{
4974
	int resched = should_resched(PREEMPT_LOCK_OFFSET);
Jan Kara's avatar
Jan Kara committed
4975 4976
	int ret = 0;

4977 4978
	lockdep_assert_held(lock);

4979
	if (spin_needbreak(lock) || resched) {
Linus Torvalds's avatar
Linus Torvalds committed
4980
		spin_unlock(lock);
4981
		if (resched)
4982
			preempt_schedule_common();
Nick Piggin's avatar
Nick Piggin committed
4983 4984
		else
			cpu_relax();
Jan Kara's avatar
Jan Kara committed
4985
		ret = 1;
Linus Torvalds's avatar
Linus Torvalds committed
4986 4987
		spin_lock(lock);
	}
Jan Kara's avatar
Jan Kara committed
4988
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
4989
}
4990
EXPORT_SYMBOL(__cond_resched_lock);
Linus Torvalds's avatar
Linus Torvalds committed
4991 4992 4993 4994

/**
 * yield - yield the current processor to other threads.
 *
Peter Zijlstra's avatar
Peter Zijlstra committed
4995 4996 4997 4998 4999 5000 5001 5002 5003
 * 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)
5004
 *	yield();
Peter Zijlstra's avatar
Peter Zijlstra committed
5005 5006 5007 5008 5009 5010 5011 5012
 *
 * 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!
Linus Torvalds's avatar
Linus Torvalds committed
5013 5014 5015 5016
 */
void __sched yield(void)
{
	set_current_state(TASK_RUNNING);
5017
	do_sched_yield();
Linus Torvalds's avatar
Linus Torvalds committed
5018 5019 5020
}
EXPORT_SYMBOL(yield);

5021 5022 5023 5024
/**
 * yield_to - yield the current processor to another thread in
 * your thread group, or accelerate that thread toward the
 * processor it's on.
5025 5026
 * @p: target task
 * @preempt: whether task preemption is allowed or not
5027 5028 5029 5030
 *
 * 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.
 *
5031
 * Return:
5032 5033 5034
 *	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.
5035
 */
5036
int __sched yield_to(struct task_struct *p, bool preempt)
5037 5038 5039 5040
{
	struct task_struct *curr = current;
	struct rq *rq, *p_rq;
	unsigned long flags;
5041
	int yielded = 0;
5042 5043 5044 5045 5046 5047

	local_irq_save(flags);
	rq = this_rq();

again:
	p_rq = task_rq(p);
5048 5049 5050 5051 5052 5053 5054 5055 5056
	/*
	 * 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;
	}

5057
	double_rq_lock(rq, p_rq);
5058
	if (task_rq(p) != p_rq) {
5059 5060 5061 5062 5063
		double_rq_unlock(rq, p_rq);
		goto again;
	}

	if (!curr->sched_class->yield_to_task)
5064
		goto out_unlock;
5065 5066

	if (curr->sched_class != p->sched_class)
5067
		goto out_unlock;
5068 5069

	if (task_running(p_rq, p) || p->state)
5070
		goto out_unlock;
5071 5072

	yielded = curr->sched_class->yield_to_task(rq, p, preempt);
5073
	if (yielded) {
5074
		schedstat_inc(rq->yld_count);
5075 5076 5077 5078 5079
		/*
		 * Make p's CPU reschedule; pick_next_entity takes care of
		 * fairness.
		 */
		if (preempt && rq != p_rq)
5080
			resched_curr(p_rq);
5081
	}
5082

5083
out_unlock:
5084
	double_rq_unlock(rq, p_rq);
5085
out_irq:
5086 5087
	local_irq_restore(flags);

5088
	if (yielded > 0)
5089 5090 5091 5092 5093 5094
		schedule();

	return yielded;
}
EXPORT_SYMBOL_GPL(yield_to);

5095 5096 5097 5098 5099 5100 5101 5102 5103 5104 5105 5106 5107 5108 5109
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;
}

Linus Torvalds's avatar
Linus Torvalds committed
5110
/*
Ingo Molnar's avatar
Ingo Molnar committed
5111
 * This task is about to go to sleep on IO. Increment rq->nr_iowait so
Linus Torvalds's avatar
Linus Torvalds committed
5112 5113 5114 5115
 * that process accounting knows that this is a task in IO wait state.
 */
long __sched io_schedule_timeout(long timeout)
{
5116
	int token;
Linus Torvalds's avatar
Linus Torvalds committed
5117 5118
	long ret;

5119
	token = io_schedule_prepare();
Linus Torvalds's avatar
Linus Torvalds committed
5120
	ret = schedule_timeout(timeout);
5121
	io_schedule_finish(token);
5122

Linus Torvalds's avatar
Linus Torvalds committed
5123 5124
	return ret;
}
5125
EXPORT_SYMBOL(io_schedule_timeout);
Linus Torvalds's avatar
Linus Torvalds committed
5126

5127 5128 5129 5130 5131 5132 5133 5134 5135 5136
void io_schedule(void)
{
	int token;

	token = io_schedule_prepare();
	schedule();
	io_schedule_finish(token);
}
EXPORT_SYMBOL(io_schedule);

Linus Torvalds's avatar
Linus Torvalds committed
5137 5138 5139 5140
/**
 * sys_sched_get_priority_max - return maximum RT priority.
 * @policy: scheduling class.
 *
5141 5142 5143
 * 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.
Linus Torvalds's avatar
Linus Torvalds committed
5144
 */
5145
SYSCALL_DEFINE1(sched_get_priority_max, int, policy)
Linus Torvalds's avatar
Linus Torvalds committed
5146 5147 5148 5149 5150 5151 5152 5153
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = MAX_USER_RT_PRIO-1;
		break;
5154
	case SCHED_DEADLINE:
Linus Torvalds's avatar
Linus Torvalds committed
5155
	case SCHED_NORMAL:
5156
	case SCHED_BATCH:
Ingo Molnar's avatar
Ingo Molnar committed
5157
	case SCHED_IDLE:
Linus Torvalds's avatar
Linus Torvalds committed
5158 5159 5160 5161 5162 5163 5164 5165 5166 5167
		ret = 0;
		break;
	}
	return ret;
}

/**
 * sys_sched_get_priority_min - return minimum RT priority.
 * @policy: scheduling class.
 *
5168 5169 5170
 * 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.
Linus Torvalds's avatar
Linus Torvalds committed
5171
 */
5172
SYSCALL_DEFINE1(sched_get_priority_min, int, policy)
Linus Torvalds's avatar
Linus Torvalds committed
5173 5174 5175 5176 5177 5178 5179 5180
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = 1;
		break;
5181
	case SCHED_DEADLINE:
Linus Torvalds's avatar
Linus Torvalds committed
5182
	case SCHED_NORMAL:
5183
	case SCHED_BATCH:
Ingo Molnar's avatar
Ingo Molnar committed
5184
	case SCHED_IDLE:
Linus Torvalds's avatar
Linus Torvalds committed
5185 5186 5187 5188 5189
		ret = 0;
	}
	return ret;
}

5190
static int sched_rr_get_interval(pid_t pid, struct timespec64 *t)
Linus Torvalds's avatar
Linus Torvalds committed
5191
{
5192
	struct task_struct *p;
Dmitry Adamushko's avatar
Dmitry Adamushko committed
5193
	unsigned int time_slice;
5194
	struct rq_flags rf;
5195
	struct rq *rq;
5196
	int retval;
Linus Torvalds's avatar
Linus Torvalds committed
5197 5198

	if (pid < 0)
5199
		return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
5200 5201

	retval = -ESRCH;
5202
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
5203 5204 5205 5206 5207 5208 5209 5210
	p = find_process_by_pid(pid);
	if (!p)
		goto out_unlock;

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

5211
	rq = task_rq_lock(p, &rf);
5212 5213 5214
	time_slice = 0;
	if (p->sched_class->get_rr_interval)
		time_slice = p->sched_class->get_rr_interval(rq, p);
5215
	task_rq_unlock(rq, p, &rf);
Dmitry Adamushko's avatar
Dmitry Adamushko committed
5216

5217
	rcu_read_unlock();
5218 5219
	jiffies_to_timespec64(time_slice, t);
	return 0;
5220

Linus Torvalds's avatar
Linus Torvalds committed
5221
out_unlock:
5222
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
5223 5224 5225
	return retval;
}

5226 5227 5228 5229 5230 5231 5232 5233 5234 5235 5236
/**
 * 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.
 */
5237
SYSCALL_DEFINE2(sched_rr_get_interval, pid_t, pid,
5238
		struct __kernel_timespec __user *, interval)
5239 5240 5241 5242 5243 5244 5245 5246 5247 5248
{
	struct timespec64 t;
	int retval = sched_rr_get_interval(pid, &t);

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

	return retval;
}

5249
#ifdef CONFIG_COMPAT_32BIT_TIME
5250 5251
SYSCALL_DEFINE2(sched_rr_get_interval_time32, pid_t, pid,
		struct old_timespec32 __user *, interval)
5252 5253 5254 5255 5256
{
	struct timespec64 t;
	int retval = sched_rr_get_interval(pid, &t);

	if (retval == 0)
5257
		retval = put_old_timespec32(&t, interval);
5258 5259 5260 5261
	return retval;
}
#endif

5262
void sched_show_task(struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
5263 5264
{
	unsigned long free = 0;
5265
	int ppid;
5266

5267 5268
	if (!try_get_task_stack(p))
		return;
5269 5270 5271 5272

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

	if (p->state == TASK_RUNNING)
5273
		printk(KERN_CONT "  running task    ");
Linus Torvalds's avatar
Linus Torvalds committed
5274
#ifdef CONFIG_DEBUG_STACK_USAGE
5275
	free = stack_not_used(p);
Linus Torvalds's avatar
Linus Torvalds committed
5276
#endif
5277
	ppid = 0;
5278
	rcu_read_lock();
5279 5280
	if (pid_alive(p))
		ppid = task_pid_nr(rcu_dereference(p->real_parent));
5281
	rcu_read_unlock();
5282
	printk(KERN_CONT "%5lu %5d %6d 0x%08lx\n", free,
5283
		task_pid_nr(p), ppid,
5284
		(unsigned long)task_thread_info(p)->flags);
Linus Torvalds's avatar
Linus Torvalds committed
5285

5286
	print_worker_info(KERN_INFO, p);
5287
	show_stack(p, NULL);
5288
	put_task_stack(p);
Linus Torvalds's avatar
Linus Torvalds committed
5289
}
5290
EXPORT_SYMBOL_GPL(sched_show_task);
Linus Torvalds's avatar
Linus Torvalds committed
5291

5292 5293 5294 5295 5296 5297 5298 5299 5300 5301 5302 5303 5304 5305 5306 5307 5308 5309 5310 5311 5312 5313
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;
}


5314
void show_state_filter(unsigned long state_filter)
Linus Torvalds's avatar
Linus Torvalds committed
5315
{
5316
	struct task_struct *g, *p;
Linus Torvalds's avatar
Linus Torvalds committed
5317

5318
#if BITS_PER_LONG == 32
5319 5320
	printk(KERN_INFO
		"  task                PC stack   pid father\n");
Linus Torvalds's avatar
Linus Torvalds committed
5321
#else
5322 5323
	printk(KERN_INFO
		"  task                        PC stack   pid father\n");
Linus Torvalds's avatar
Linus Torvalds committed
5324
#endif
5325
	rcu_read_lock();
5326
	for_each_process_thread(g, p) {
Linus Torvalds's avatar
Linus Torvalds committed
5327 5328
		/*
		 * reset the NMI-timeout, listing all files on a slow
Lucas De Marchi's avatar
Lucas De Marchi committed
5329
		 * console might take a lot of time:
5330 5331 5332
		 * Also, reset softlockup watchdogs on all CPUs, because
		 * another CPU might be blocked waiting for us to process
		 * an IPI.
Linus Torvalds's avatar
Linus Torvalds committed
5333 5334
		 */
		touch_nmi_watchdog();
5335
		touch_all_softlockup_watchdogs();
5336
		if (state_filter_match(state_filter, p))
5337
			sched_show_task(p);
5338
	}
Linus Torvalds's avatar
Linus Torvalds committed
5339

Ingo Molnar's avatar
Ingo Molnar committed
5340
#ifdef CONFIG_SCHED_DEBUG
5341 5342
	if (!state_filter)
		sysrq_sched_debug_show();
Ingo Molnar's avatar
Ingo Molnar committed
5343
#endif
5344
	rcu_read_unlock();
5345 5346 5347
	/*
	 * Only show locks if all tasks are dumped:
	 */
5348
	if (!state_filter)
5349
		debug_show_all_locks();
Linus Torvalds's avatar
Linus Torvalds committed
5350 5351
}

5352 5353 5354
/**
 * init_idle - set up an idle thread for a given CPU
 * @idle: task in question
5355
 * @cpu: CPU the idle task belongs to
5356 5357 5358 5359
 *
 * NOTE: this function does not set the idle thread's NEED_RESCHED
 * flag, to make booting more robust.
 */
5360
void init_idle(struct task_struct *idle, int cpu)
Linus Torvalds's avatar
Linus Torvalds committed
5361
{
5362
	struct rq *rq = cpu_rq(cpu);
Linus Torvalds's avatar
Linus Torvalds committed
5363 5364
	unsigned long flags;

5365 5366
	raw_spin_lock_irqsave(&idle->pi_lock, flags);
	raw_spin_lock(&rq->lock);
5367

5368
	__sched_fork(0, idle);
5369
	idle->state = TASK_RUNNING;
Ingo Molnar's avatar
Ingo Molnar committed
5370
	idle->se.exec_start = sched_clock();
5371
	idle->flags |= PF_IDLE;
Ingo Molnar's avatar
Ingo Molnar committed
5372

5373 5374
	kasan_unpoison_task_stack(idle);

5375 5376 5377 5378 5379 5380 5381 5382 5383
#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
5384 5385
	/*
	 * We're having a chicken and egg problem, even though we are
5386
	 * holding rq->lock, the CPU isn't yet set to this CPU so the
5387 5388 5389 5390 5391 5392 5393 5394
	 * 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();
Ingo Molnar's avatar
Ingo Molnar committed
5395
	__set_task_cpu(idle, cpu);
5396
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
5397 5398

	rq->curr = rq->idle = idle;
5399
	idle->on_rq = TASK_ON_RQ_QUEUED;
5400
#ifdef CONFIG_SMP
5401
	idle->on_cpu = 1;
5402
#endif
5403 5404
	raw_spin_unlock(&rq->lock);
	raw_spin_unlock_irqrestore(&idle->pi_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
5405 5406

	/* Set the preempt count _outside_ the spinlocks! */
5407
	init_idle_preempt_count(idle, cpu);
5408

Ingo Molnar's avatar
Ingo Molnar committed
5409 5410 5411 5412
	/*
	 * The idle tasks have their own, simple scheduling class:
	 */
	idle->sched_class = &idle_sched_class;
5413
	ftrace_graph_init_idle_task(idle, cpu);
5414
	vtime_init_idle(idle, cpu);
5415
#ifdef CONFIG_SMP
5416 5417
	sprintf(idle->comm, "%s/%d", INIT_TASK_COMM, cpu);
#endif
5418 5419
}

5420 5421
#ifdef CONFIG_SMP

5422 5423 5424
int cpuset_cpumask_can_shrink(const struct cpumask *cur,
			      const struct cpumask *trial)
{
5425
	int ret = 1;
5426

5427 5428 5429
	if (!cpumask_weight(cur))
		return ret;

5430
	ret = dl_cpuset_cpumask_can_shrink(cur, trial);
5431 5432 5433 5434

	return ret;
}

5435 5436 5437 5438 5439 5440 5441
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
5442
	 * to a new cpuset; we don't want to change their CPU
5443 5444 5445 5446 5447 5448 5449 5450 5451 5452 5453 5454
	 * 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,
5455 5456
					      cs_cpus_allowed))
		ret = dl_task_can_attach(p, cs_cpus_allowed);
5457 5458 5459 5460 5461

out:
	return ret;
}

5462
bool sched_smp_initialized __read_mostly;
5463

5464 5465 5466 5467 5468 5469 5470 5471 5472 5473
#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;

5474
	if (!cpumask_test_cpu(target_cpu, &p->cpus_allowed))
5475 5476 5477 5478
		return -EINVAL;

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

5479
	trace_sched_move_numa(p, curr_cpu, target_cpu);
5480 5481
	return stop_one_cpu(curr_cpu, migration_cpu_stop, &arg);
}
5482 5483 5484 5485 5486 5487 5488

/*
 * 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)
{
5489
	bool queued, running;
5490 5491
	struct rq_flags rf;
	struct rq *rq;
5492

5493
	rq = task_rq_lock(p, &rf);
5494
	queued = task_on_rq_queued(p);
5495 5496
	running = task_current(rq, p);

5497
	if (queued)
5498
		dequeue_task(rq, p, DEQUEUE_SAVE);
5499
	if (running)
5500
		put_prev_task(rq, p);
5501 5502 5503

	p->numa_preferred_nid = nid;

5504
	if (queued)
5505
		enqueue_task(rq, p, ENQUEUE_RESTORE | ENQUEUE_NOCLOCK);
5506
	if (running)
5507
		set_curr_task(rq, p);
5508
	task_rq_unlock(rq, p, &rf);
5509
}
Peter Zijlstra's avatar
Peter Zijlstra committed
5510
#endif /* CONFIG_NUMA_BALANCING */
5511

Linus Torvalds's avatar
Linus Torvalds committed
5512
#ifdef CONFIG_HOTPLUG_CPU
5513
/*
5514
 * Ensure that the idle task is using init_mm right before its CPU goes
5515
 * offline.
5516
 */
5517
void idle_task_exit(void)
Linus Torvalds's avatar
Linus Torvalds committed
5518
{
5519
	struct mm_struct *mm = current->active_mm;
5520

5521
	BUG_ON(cpu_online(smp_processor_id()));
5522

5523
	if (mm != &init_mm) {
5524
		switch_mm(mm, &init_mm, current);
5525
		current->active_mm = &init_mm;
5526 5527
		finish_arch_post_lock_switch();
	}
5528
	mmdrop(mm);
Linus Torvalds's avatar
Linus Torvalds committed
5529 5530 5531
}

/*
5532 5533
 * 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
5534 5535 5536
 * 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.
5537 5538
 *
 * Also see the comment "Global load-average calculations".
Linus Torvalds's avatar
Linus Torvalds committed
5539
 */
5540
static void calc_load_migrate(struct rq *rq)
Linus Torvalds's avatar
Linus Torvalds committed
5541
{
5542
	long delta = calc_load_fold_active(rq, 1);
5543 5544
	if (delta)
		atomic_long_add(delta, &calc_load_tasks);
Linus Torvalds's avatar
Linus Torvalds committed
5545 5546
}

5547 5548 5549 5550 5551 5552 5553 5554 5555 5556 5557 5558 5559 5560 5561 5562
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,
};

5563
/*
5564 5565 5566 5567 5568 5569
 * 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.
Linus Torvalds's avatar
Linus Torvalds committed
5570
 */
5571
static void migrate_tasks(struct rq *dead_rq, struct rq_flags *rf)
Linus Torvalds's avatar
Linus Torvalds committed
5572
{
5573
	struct rq *rq = dead_rq;
5574
	struct task_struct *next, *stop = rq->stop;
5575
	struct rq_flags orf = *rf;
5576
	int dest_cpu;
Linus Torvalds's avatar
Linus Torvalds committed
5577 5578

	/*
5579 5580 5581 5582 5583 5584 5585
	 * 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.
Linus Torvalds's avatar
Linus Torvalds committed
5586
	 */
5587
	rq->stop = NULL;
5588

5589 5590 5591 5592 5593 5594 5595
	/*
	 * 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);

5596
	for (;;) {
5597 5598
		/*
		 * There's this thread running, bail when that's the only
5599
		 * remaining thread:
5600 5601
		 */
		if (rq->nr_running == 1)
Ingo Molnar's avatar
Ingo Molnar committed
5602
			break;
5603

5604
		/*
5605
		 * pick_next_task() assumes pinned rq->lock:
5606
		 */
5607
		next = pick_next_task(rq, &fake_task, rf);
5608
		BUG_ON(!next);
5609
		put_prev_task(rq, next);
5610

5611 5612 5613 5614 5615 5616 5617 5618 5619
		/*
		 * 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.
		 */
5620
		rq_unlock(rq, rf);
5621
		raw_spin_lock(&next->pi_lock);
5622
		rq_relock(rq, rf);
5623 5624 5625 5626 5627 5628 5629 5630 5631 5632 5633

		/*
		 * 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;
		}

5634
		/* Find suitable destination for @next, with force if needed. */
5635
		dest_cpu = select_fallback_rq(dead_rq->cpu, next);
5636
		rq = __migrate_task(rq, rf, next, dest_cpu);
5637
		if (rq != dead_rq) {
5638
			rq_unlock(rq, rf);
5639
			rq = dead_rq;
5640 5641
			*rf = orf;
			rq_relock(rq, rf);
5642
		}
5643
		raw_spin_unlock(&next->pi_lock);
Linus Torvalds's avatar
Linus Torvalds committed
5644
	}
5645

5646
	rq->stop = stop;
5647
}
Linus Torvalds's avatar
Linus Torvalds committed
5648 5649
#endif /* CONFIG_HOTPLUG_CPU */

5650
void set_rq_online(struct rq *rq)
5651 5652 5653 5654
{
	if (!rq->online) {
		const struct sched_class *class;

5655
		cpumask_set_cpu(rq->cpu, rq->rd->online);
5656 5657 5658 5659 5660 5661 5662 5663 5664
		rq->online = 1;

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

5665
void set_rq_offline(struct rq *rq)
5666 5667 5668 5669 5670 5671 5672 5673 5674
{
	if (rq->online) {
		const struct sched_class *class;

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

5675
		cpumask_clear_cpu(rq->cpu, rq->rd->online);
5676 5677 5678 5679
		rq->online = 0;
	}
}

5680 5681 5682 5683
/*
 * used to mark begin/end of suspend/resume:
 */
static int num_cpus_frozen;
5684

Linus Torvalds's avatar
Linus Torvalds committed
5685
/*
5686 5687 5688
 * Update cpusets according to cpu_active mask.  If cpusets are
 * disabled, cpuset_update_active_cpus() becomes a simple wrapper
 * around partition_sched_domains().
5689 5690 5691
 *
 * 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.
Linus Torvalds's avatar
Linus Torvalds committed
5692
 */
5693
static void cpuset_cpu_active(void)
5694
{
5695
	if (cpuhp_tasks_frozen) {
5696 5697 5698 5699 5700 5701
		/*
		 * 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.
		 */
5702 5703
		partition_sched_domains(1, NULL, NULL);
		if (--num_cpus_frozen)
5704
			return;
5705 5706 5707 5708 5709
		/*
		 * This is the last CPU online operation. So fall through and
		 * restore the original sched domains by considering the
		 * cpuset configurations.
		 */
5710
		cpuset_force_rebuild();
5711
	}
5712
	cpuset_update_active_cpus();
5713
}
5714

5715
static int cpuset_cpu_inactive(unsigned int cpu)
5716
{
5717
	if (!cpuhp_tasks_frozen) {
5718
		if (dl_cpu_busy(cpu))
5719
			return -EBUSY;
5720
		cpuset_update_active_cpus();
5721
	} else {
5722 5723
		num_cpus_frozen++;
		partition_sched_domains(1, NULL, NULL);
5724
	}
5725
	return 0;
5726 5727
}

5728
int sched_cpu_activate(unsigned int cpu)
5729
{
5730
	struct rq *rq = cpu_rq(cpu);
5731
	struct rq_flags rf;
5732

5733 5734
#ifdef CONFIG_SCHED_SMT
	/*
5735
	 * When going up, increment the number of cores with SMT present.
5736
	 */
5737 5738
	if (cpumask_weight(cpu_smt_mask(cpu)) == 2)
		static_branch_inc_cpuslocked(&sched_smt_present);
5739
#endif
5740
	set_cpu_active(cpu, true);
5741

5742
	if (sched_smp_initialized) {
5743
		sched_domains_numa_masks_set(cpu);
5744
		cpuset_cpu_active();
5745
	}
5746 5747 5748 5749 5750

	/*
	 * Put the rq online, if not already. This happens:
	 *
	 * 1) In the early boot process, because we build the real domains
5751
	 *    after all CPUs have been brought up.
5752 5753 5754 5755
	 *
	 * 2) At runtime, if cpuset_cpu_active() fails to rebuild the
	 *    domains.
	 */
5756
	rq_lock_irqsave(rq, &rf);
5757 5758 5759 5760
	if (rq->rd) {
		BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
		set_rq_online(rq);
	}
5761
	rq_unlock_irqrestore(rq, &rf);
5762 5763 5764

	update_max_interval();

5765
	return 0;
5766 5767
}

5768
int sched_cpu_deactivate(unsigned int cpu)
5769 5770 5771
{
	int ret;

5772
	set_cpu_active(cpu, false);
5773 5774 5775 5776 5777 5778 5779
	/*
	 * 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.
	 */
5780
	synchronize_rcu();
5781

5782 5783 5784 5785 5786 5787 5788 5789
#ifdef CONFIG_SCHED_SMT
	/*
	 * When going down, decrement the number of cores with SMT present.
	 */
	if (cpumask_weight(cpu_smt_mask(cpu)) == 2)
		static_branch_dec_cpuslocked(&sched_smt_present);
#endif

5790 5791 5792 5793 5794 5795 5796
	if (!sched_smp_initialized)
		return 0;

	ret = cpuset_cpu_inactive(cpu);
	if (ret) {
		set_cpu_active(cpu, true);
		return ret;
5797
	}
5798 5799
	sched_domains_numa_masks_clear(cpu);
	return 0;
5800 5801
}

5802 5803 5804 5805 5806 5807 5808 5809
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();
}

5810 5811
int sched_cpu_starting(unsigned int cpu)
{
5812
	sched_rq_cpu_starting(cpu);
5813
	sched_tick_start(cpu);
5814
	return 0;
5815 5816
}

5817 5818 5819 5820
#ifdef CONFIG_HOTPLUG_CPU
int sched_cpu_dying(unsigned int cpu)
{
	struct rq *rq = cpu_rq(cpu);
5821
	struct rq_flags rf;
5822 5823 5824

	/* Handle pending wakeups and then migrate everything off */
	sched_ttwu_pending();
5825
	sched_tick_stop(cpu);
5826 5827

	rq_lock_irqsave(rq, &rf);
5828 5829 5830 5831
	if (rq->rd) {
		BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
		set_rq_offline(rq);
	}
5832
	migrate_tasks(rq, &rf);
5833
	BUG_ON(rq->nr_running != 1);
5834 5835
	rq_unlock_irqrestore(rq, &rf);

5836 5837
	calc_load_migrate(rq);
	update_max_interval();
5838
	nohz_balance_exit_idle(rq);
5839
	hrtick_clear(rq);
5840 5841 5842 5843
	return 0;
}
#endif

Linus Torvalds's avatar
Linus Torvalds committed
5844 5845
void __init sched_init_smp(void)
{
5846 5847
	sched_init_numa();

5848 5849
	/*
	 * There's no userspace yet to cause hotplug operations; hence all the
5850
	 * CPU masks are stable and all blatant races in the below code cannot
5851
	 * happen.
5852
	 */
5853
	mutex_lock(&sched_domains_mutex);
5854
	sched_init_domains(cpu_active_mask);
5855
	mutex_unlock(&sched_domains_mutex);
5856

5857
	/* Move init over to a non-isolated CPU */
5858
	if (set_cpus_allowed_ptr(current, housekeeping_cpumask(HK_FLAG_DOMAIN)) < 0)
5859
		BUG();
5860
	sched_init_granularity();
5861

5862
	init_sched_rt_class();
5863
	init_sched_dl_class();
5864

5865
	sched_smp_initialized = true;
Linus Torvalds's avatar
Linus Torvalds committed
5866
}
5867 5868 5869

static int __init migration_init(void)
{
5870
	sched_rq_cpu_starting(smp_processor_id());
5871
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
5872
}
5873 5874
early_initcall(migration_init);

Linus Torvalds's avatar
Linus Torvalds committed
5875 5876 5877
#else
void __init sched_init_smp(void)
{
5878
	sched_init_granularity();
Linus Torvalds's avatar
Linus Torvalds committed
5879 5880 5881 5882 5883 5884 5885 5886 5887 5888
}
#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);
}

5889
#ifdef CONFIG_CGROUP_SCHED
5890 5891 5892 5893
/*
 * Default task group.
 * Every task in system belongs to this group at bootup.
 */
5894
struct task_group root_task_group;
5895
LIST_HEAD(task_groups);
5896 5897 5898

/* Cacheline aligned slab cache for task_group */
static struct kmem_cache *task_group_cache __read_mostly;
5899
#endif
Peter Zijlstra's avatar
Peter Zijlstra committed
5900

5901
DECLARE_PER_CPU(cpumask_var_t, load_balance_mask);
5902
DECLARE_PER_CPU(cpumask_var_t, select_idle_mask);
Peter Zijlstra's avatar
Peter Zijlstra committed
5903

Linus Torvalds's avatar
Linus Torvalds committed
5904 5905
void __init sched_init(void)
{
Ingo Molnar's avatar
Ingo Molnar committed
5906
	int i, j;
5907 5908
	unsigned long alloc_size = 0, ptr;

5909
	wait_bit_init();
5910

5911 5912 5913 5914 5915 5916 5917
#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) {
5918
		ptr = (unsigned long)kzalloc(alloc_size, GFP_NOWAIT);
5919 5920

#ifdef CONFIG_FAIR_GROUP_SCHED
5921
		root_task_group.se = (struct sched_entity **)ptr;
5922 5923
		ptr += nr_cpu_ids * sizeof(void **);

5924
		root_task_group.cfs_rq = (struct cfs_rq **)ptr;
5925
		ptr += nr_cpu_ids * sizeof(void **);
5926

5927
#endif /* CONFIG_FAIR_GROUP_SCHED */
5928
#ifdef CONFIG_RT_GROUP_SCHED
5929
		root_task_group.rt_se = (struct sched_rt_entity **)ptr;
5930 5931
		ptr += nr_cpu_ids * sizeof(void **);

5932
		root_task_group.rt_rq = (struct rt_rq **)ptr;
5933 5934
		ptr += nr_cpu_ids * sizeof(void **);

5935
#endif /* CONFIG_RT_GROUP_SCHED */
5936
	}
5937
#ifdef CONFIG_CPUMASK_OFFSTACK
5938 5939 5940
	for_each_possible_cpu(i) {
		per_cpu(load_balance_mask, i) = (cpumask_var_t)kzalloc_node(
			cpumask_size(), GFP_KERNEL, cpu_to_node(i));
5941 5942
		per_cpu(select_idle_mask, i) = (cpumask_var_t)kzalloc_node(
			cpumask_size(), GFP_KERNEL, cpu_to_node(i));
5943
	}
5944
#endif /* CONFIG_CPUMASK_OFFSTACK */
Ingo Molnar's avatar
Ingo Molnar committed
5945

5946 5947
	init_rt_bandwidth(&def_rt_bandwidth, global_rt_period(), global_rt_runtime());
	init_dl_bandwidth(&def_dl_bandwidth, global_rt_period(), global_rt_runtime());
5948

5949 5950 5951 5952
#ifdef CONFIG_SMP
	init_defrootdomain();
#endif

5953
#ifdef CONFIG_RT_GROUP_SCHED
5954
	init_rt_bandwidth(&root_task_group.rt_bandwidth,
5955
			global_rt_period(), global_rt_runtime());
5956
#endif /* CONFIG_RT_GROUP_SCHED */
5957

Dhaval Giani's avatar
Dhaval Giani committed
5958
#ifdef CONFIG_CGROUP_SCHED
5959 5960
	task_group_cache = KMEM_CACHE(task_group, 0);

5961 5962
	list_add(&root_task_group.list, &task_groups);
	INIT_LIST_HEAD(&root_task_group.children);
5963
	INIT_LIST_HEAD(&root_task_group.siblings);
5964
	autogroup_init(&init_task);
Dhaval Giani's avatar
Dhaval Giani committed
5965
#endif /* CONFIG_CGROUP_SCHED */
Peter Zijlstra's avatar
Peter Zijlstra committed
5966

5967
	for_each_possible_cpu(i) {
5968
		struct rq *rq;
Linus Torvalds's avatar
Linus Torvalds committed
5969 5970

		rq = cpu_rq(i);
5971
		raw_spin_lock_init(&rq->lock);
Nick Piggin's avatar
Nick Piggin committed
5972
		rq->nr_running = 0;
5973 5974
		rq->calc_load_active = 0;
		rq->calc_load_update = jiffies + LOAD_FREQ;
5975
		init_cfs_rq(&rq->cfs);
5976 5977
		init_rt_rq(&rq->rt);
		init_dl_rq(&rq->dl);
Ingo Molnar's avatar
Ingo Molnar committed
5978
#ifdef CONFIG_FAIR_GROUP_SCHED
5979
		root_task_group.shares = ROOT_TASK_GROUP_LOAD;
Peter Zijlstra's avatar
Peter Zijlstra committed
5980
		INIT_LIST_HEAD(&rq->leaf_cfs_rq_list);
5981
		rq->tmp_alone_branch = &rq->leaf_cfs_rq_list;
Dhaval Giani's avatar
Dhaval Giani committed
5982
		/*
5983
		 * How much CPU bandwidth does root_task_group get?
Dhaval Giani's avatar
Dhaval Giani committed
5984 5985
		 *
		 * In case of task-groups formed thr' the cgroup filesystem, it
5986 5987
		 * gets 100% of the CPU resources in the system. This overall
		 * system CPU resource is divided among the tasks of
5988
		 * root_task_group and its child task-groups in a fair manner,
Dhaval Giani's avatar
Dhaval Giani committed
5989 5990 5991
		 * based on each entity's (task or task-group's) weight
		 * (se->load.weight).
		 *
5992
		 * In other words, if root_task_group has 10 tasks of weight
Dhaval Giani's avatar
Dhaval Giani committed
5993
		 * 1024) and two child groups A0 and A1 (of weight 1024 each),
5994
		 * then A0's share of the CPU resource is:
Dhaval Giani's avatar
Dhaval Giani committed
5995
		 *
5996
		 *	A0's bandwidth = 1024 / (10*1024 + 1024 + 1024) = 8.33%
Dhaval Giani's avatar
Dhaval Giani committed
5997
		 *
5998 5999
		 * We achieve this by letting root_task_group's tasks sit
		 * directly in rq->cfs (i.e root_task_group->se[] = NULL).
Dhaval Giani's avatar
Dhaval Giani committed
6000
		 */
6001
		init_cfs_bandwidth(&root_task_group.cfs_bandwidth);
6002
		init_tg_cfs_entry(&root_task_group, &rq->cfs, NULL, i, NULL);
Dhaval Giani's avatar
Dhaval Giani committed
6003 6004 6005
#endif /* CONFIG_FAIR_GROUP_SCHED */

		rq->rt.rt_runtime = def_rt_bandwidth.rt_runtime;
6006
#ifdef CONFIG_RT_GROUP_SCHED
6007
		init_tg_rt_entry(&root_task_group, &rq->rt, NULL, i, NULL);
Ingo Molnar's avatar
Ingo Molnar committed
6008
#endif
Linus Torvalds's avatar
Linus Torvalds committed
6009

Ingo Molnar's avatar
Ingo Molnar committed
6010 6011
		for (j = 0; j < CPU_LOAD_IDX_MAX; j++)
			rq->cpu_load[j] = 0;
6012

Linus Torvalds's avatar
Linus Torvalds committed
6013
#ifdef CONFIG_SMP
Nick Piggin's avatar
Nick Piggin committed
6014
		rq->sd = NULL;
6015
		rq->rd = NULL;
6016
		rq->cpu_capacity = rq->cpu_capacity_orig = SCHED_CAPACITY_SCALE;
6017
		rq->balance_callback = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
6018
		rq->active_balance = 0;
Ingo Molnar's avatar
Ingo Molnar committed
6019
		rq->next_balance = jiffies;
Linus Torvalds's avatar
Linus Torvalds committed
6020
		rq->push_cpu = 0;
6021
		rq->cpu = i;
6022
		rq->online = 0;
6023 6024
		rq->idle_stamp = 0;
		rq->avg_idle = 2*sysctl_sched_migration_cost;
6025
		rq->max_idle_balance_cost = sysctl_sched_migration_cost;
6026 6027 6028

		INIT_LIST_HEAD(&rq->cfs_tasks);

6029
		rq_attach_root(rq, &def_root_domain);
6030
#ifdef CONFIG_NO_HZ_COMMON
6031
		rq->last_load_update_tick = jiffies;
6032
		rq->last_blocked_load_update_tick = jiffies;
6033
		atomic_set(&rq->nohz_flags, 0);
6034
#endif
6035
#endif /* CONFIG_SMP */
6036
		hrtick_rq_init(rq);
Linus Torvalds's avatar
Linus Torvalds committed
6037 6038 6039
		atomic_set(&rq->nr_iowait, 0);
	}

6040
	set_load_weight(&init_task, false);
6041

Linus Torvalds's avatar
Linus Torvalds committed
6042 6043 6044
	/*
	 * The boot idle thread does lazy MMU switching as well:
	 */
6045
	mmgrab(&init_mm);
Linus Torvalds's avatar
Linus Torvalds committed
6046 6047 6048 6049 6050 6051 6052 6053 6054
	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());
6055 6056 6057

	calc_load_update = jiffies + LOAD_FREQ;

6058
#ifdef CONFIG_SMP
6059
	idle_thread_set_boot_cpu();
6060 6061
#endif
	init_sched_fair_class();
6062

6063 6064
	init_schedstats();

6065 6066
	psi_init();

6067
	scheduler_running = 1;
Linus Torvalds's avatar
Linus Torvalds committed
6068 6069
}

6070
#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
6071 6072
static inline int preempt_count_equals(int preempt_offset)
{
6073
	int nested = preempt_count() + rcu_preempt_depth();
6074

Arnd Bergmann's avatar
Arnd Bergmann committed
6075
	return (nested == preempt_offset);
6076 6077
}

6078
void __might_sleep(const char *file, int line, int preempt_offset)
Linus Torvalds's avatar
Linus Torvalds committed
6079
{
Peter Zijlstra's avatar
Peter Zijlstra committed
6080 6081 6082 6083 6084
	/*
	 * 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.
	 */
6085
	WARN_ONCE(current->state != TASK_RUNNING && current->task_state_change,
Peter Zijlstra's avatar
Peter Zijlstra committed
6086 6087 6088 6089
			"do not call blocking ops when !TASK_RUNNING; "
			"state=%lx set at [<%p>] %pS\n",
			current->state,
			(void *)current->task_state_change,
6090
			(void *)current->task_state_change);
Peter Zijlstra's avatar
Peter Zijlstra committed
6091

6092 6093 6094 6095 6096
	___might_sleep(file, line, preempt_offset);
}
EXPORT_SYMBOL(__might_sleep);

void ___might_sleep(const char *file, int line, int preempt_offset)
Linus Torvalds's avatar
Linus Torvalds committed
6097
{
6098 6099 6100
	/* Ratelimiting timestamp: */
	static unsigned long prev_jiffy;

6101
	unsigned long preempt_disable_ip;
Linus Torvalds's avatar
Linus Torvalds committed
6102

6103 6104 6105
	/* WARN_ON_ONCE() by default, no rate limit required: */
	rcu_sleep_check();

6106 6107
	if ((preempt_count_equals(preempt_offset) && !irqs_disabled() &&
	     !is_idle_task(current)) ||
6108 6109
	    system_state == SYSTEM_BOOTING || system_state > SYSTEM_RUNNING ||
	    oops_in_progress)
6110
		return;
6111

6112 6113 6114 6115
	if (time_before(jiffies, prev_jiffy + HZ) && prev_jiffy)
		return;
	prev_jiffy = jiffies;

6116
	/* Save this before calling printk(), since that will clobber it: */
6117 6118
	preempt_disable_ip = get_preempt_disable_ip(current);

6119 6120 6121 6122 6123 6124 6125
	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);
6126

6127 6128 6129
	if (task_stack_end_corrupted(current))
		printk(KERN_EMERG "Thread overran stack, or stack corrupted\n");

6130 6131 6132
	debug_show_held_locks(current);
	if (irqs_disabled())
		print_irqtrace_events(current);
6133 6134
	if (IS_ENABLED(CONFIG_DEBUG_PREEMPT)
	    && !preempt_count_equals(preempt_offset)) {
6135
		pr_err("Preemption disabled at:");
6136
		print_ip_sym(preempt_disable_ip);
6137 6138
		pr_cont("\n");
	}
6139
	dump_stack();
6140
	add_taint(TAINT_WARN, LOCKDEP_STILL_OK);
Linus Torvalds's avatar
Linus Torvalds committed
6141
}
6142
EXPORT_SYMBOL(___might_sleep);
6143 6144 6145 6146 6147 6148 6149 6150 6151 6152 6153 6154 6155 6156 6157 6158 6159 6160 6161 6162 6163 6164 6165 6166 6167 6168 6169 6170

void __cant_sleep(const char *file, int line, int preempt_offset)
{
	static unsigned long prev_jiffy;

	if (irqs_disabled())
		return;

	if (!IS_ENABLED(CONFIG_PREEMPT_COUNT))
		return;

	if (preempt_count() > preempt_offset)
		return;

	if (time_before(jiffies, prev_jiffy + HZ) && prev_jiffy)
		return;
	prev_jiffy = jiffies;

	printk(KERN_ERR "BUG: assuming atomic 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);

	debug_show_held_locks(current);
	dump_stack();
	add_taint(TAINT_WARN, LOCKDEP_STILL_OK);
}
EXPORT_SYMBOL_GPL(__cant_sleep);
Linus Torvalds's avatar
Linus Torvalds committed
6171 6172 6173
#endif

#ifdef CONFIG_MAGIC_SYSRQ
6174
void normalize_rt_tasks(void)
6175
{
6176
	struct task_struct *g, *p;
6177 6178 6179
	struct sched_attr attr = {
		.sched_policy = SCHED_NORMAL,
	};
Linus Torvalds's avatar
Linus Torvalds committed
6180

6181
	read_lock(&tasklist_lock);
6182
	for_each_process_thread(g, p) {
6183 6184 6185
		/*
		 * Only normalize user tasks:
		 */
6186
		if (p->flags & PF_KTHREAD)
6187 6188
			continue;

6189 6190 6191 6192
		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);
Ingo Molnar's avatar
Ingo Molnar committed
6193

6194
		if (!dl_task(p) && !rt_task(p)) {
Ingo Molnar's avatar
Ingo Molnar committed
6195 6196 6197 6198
			/*
			 * Renice negative nice level userspace
			 * tasks back to 0:
			 */
6199
			if (task_nice(p) < 0)
Ingo Molnar's avatar
Ingo Molnar committed
6200
				set_user_nice(p, 0);
Linus Torvalds's avatar
Linus Torvalds committed
6201
			continue;
Ingo Molnar's avatar
Ingo Molnar committed
6202
		}
Linus Torvalds's avatar
Linus Torvalds committed
6203

6204
		__sched_setscheduler(p, &attr, false, false);
6205
	}
6206
	read_unlock(&tasklist_lock);
Linus Torvalds's avatar
Linus Torvalds committed
6207 6208 6209
}

#endif /* CONFIG_MAGIC_SYSRQ */
6210

6211
#if defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB)
6212
/*
6213
 * These functions are only useful for the IA64 MCA handling, or kdb.
6214 6215 6216 6217 6218 6219 6220 6221 6222
 *
 * 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.
 */

/**
6223
 * curr_task - return the current task for a given CPU.
6224 6225 6226
 * @cpu: the processor in question.
 *
 * ONLY VALID WHEN THE WHOLE SYSTEM IS STOPPED!
6227 6228
 *
 * Return: The current task for @cpu.
6229
 */
6230
struct task_struct *curr_task(int cpu)
6231 6232 6233 6234
{
	return cpu_curr(cpu);
}

6235 6236 6237
#endif /* defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB) */

#ifdef CONFIG_IA64
6238
/**
6239
 * set_curr_task - set the current task for a given CPU.
6240 6241 6242 6243
 * @cpu: the processor in question.
 * @p: the task pointer to set.
 *
 * Description: This function must only be used when non-maskable interrupts
Ingo Molnar's avatar
Ingo Molnar committed
6244
 * are serviced on a separate stack. It allows the architecture to switch the
6245
 * notion of the current task on a CPU in a non-blocking manner. This function
6246 6247 6248 6249 6250 6251 6252
 * 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!
 */
6253
void ia64_set_curr_task(int cpu, struct task_struct *p)
6254 6255 6256 6257 6258
{
	cpu_curr(cpu) = p;
}

#endif
6259

Dhaval Giani's avatar
Dhaval Giani committed
6260
#ifdef CONFIG_CGROUP_SCHED
6261 6262 6263
/* task_group_lock serializes the addition/removal of task groups */
static DEFINE_SPINLOCK(task_group_lock);

6264
static void sched_free_group(struct task_group *tg)
6265 6266 6267
{
	free_fair_sched_group(tg);
	free_rt_sched_group(tg);
6268
	autogroup_free(tg);
6269
	kmem_cache_free(task_group_cache, tg);
6270 6271 6272
}

/* allocate runqueue etc for a new task group */
6273
struct task_group *sched_create_group(struct task_group *parent)
6274 6275 6276
{
	struct task_group *tg;

6277
	tg = kmem_cache_alloc(task_group_cache, GFP_KERNEL | __GFP_ZERO);
6278 6279 6280
	if (!tg)
		return ERR_PTR(-ENOMEM);

6281
	if (!alloc_fair_sched_group(tg, parent))
6282 6283
		goto err;

6284
	if (!alloc_rt_sched_group(tg, parent))
6285 6286
		goto err;

6287 6288 6289
	return tg;

err:
6290
	sched_free_group(tg);
6291 6292 6293 6294 6295 6296 6297
	return ERR_PTR(-ENOMEM);
}

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

6298
	spin_lock_irqsave(&task_group_lock, flags);
Peter Zijlstra's avatar
Peter Zijlstra committed
6299
	list_add_rcu(&tg->list, &task_groups);
Peter Zijlstra's avatar
Peter Zijlstra committed
6300

6301 6302
	/* Root should already exist: */
	WARN_ON(!parent);
Peter Zijlstra's avatar
Peter Zijlstra committed
6303 6304 6305

	tg->parent = parent;
	INIT_LIST_HEAD(&tg->children);
6306
	list_add_rcu(&tg->siblings, &parent->children);
6307
	spin_unlock_irqrestore(&task_group_lock, flags);
6308 6309

	online_fair_sched_group(tg);
6310 6311
}

6312
/* rcu callback to free various structures associated with a task group */
6313
static void sched_free_group_rcu(struct rcu_head *rhp)
6314
{
6315
	/* Now it should be safe to free those cfs_rqs: */
6316
	sched_free_group(container_of(rhp, struct task_group, rcu));
6317 6318
}

6319
void sched_destroy_group(struct task_group *tg)
6320
{
6321
	/* Wait for possible concurrent references to cfs_rqs complete: */
6322
	call_rcu(&tg->rcu, sched_free_group_rcu);
6323 6324 6325
}

void sched_offline_group(struct task_group *tg)
6326
{
6327
	unsigned long flags;
6328

6329
	/* End participation in shares distribution: */
6330
	unregister_fair_sched_group(tg);
6331 6332

	spin_lock_irqsave(&task_group_lock, flags);
Peter Zijlstra's avatar
Peter Zijlstra committed
6333
	list_del_rcu(&tg->list);
Peter Zijlstra's avatar
Peter Zijlstra committed
6334
	list_del_rcu(&tg->siblings);
6335
	spin_unlock_irqrestore(&task_group_lock, flags);
6336 6337
}

6338
static void sched_change_group(struct task_struct *tsk, int type)
6339
{
6340
	struct task_group *tg;
6341

6342 6343 6344 6345 6346 6347
	/*
	 * 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),
6348 6349 6350 6351
			  struct task_group, css);
	tg = autogroup_task_group(tsk, tg);
	tsk->sched_task_group = tg;

Peter Zijlstra's avatar
Peter Zijlstra committed
6352
#ifdef CONFIG_FAIR_GROUP_SCHED
6353 6354
	if (tsk->sched_class->task_change_group)
		tsk->sched_class->task_change_group(tsk, type);
6355
	else
Peter Zijlstra's avatar
Peter Zijlstra committed
6356
#endif
6357
		set_task_rq(tsk, task_cpu(tsk));
6358 6359 6360 6361 6362 6363 6364 6365 6366 6367 6368
}

/*
 * 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)
{
6369 6370
	int queued, running, queue_flags =
		DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
6371 6372 6373 6374
	struct rq_flags rf;
	struct rq *rq;

	rq = task_rq_lock(tsk, &rf);
6375
	update_rq_clock(rq);
6376 6377 6378 6379 6380

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

	if (queued)
6381
		dequeue_task(rq, tsk, queue_flags);
6382
	if (running)
6383 6384 6385
		put_prev_task(rq, tsk);

	sched_change_group(tsk, TASK_MOVE_GROUP);
Peter Zijlstra's avatar
Peter Zijlstra committed
6386

6387
	if (queued)
6388
		enqueue_task(rq, tsk, queue_flags);
6389
	if (running)
6390
		set_curr_task(rq, tsk);
6391

6392
	task_rq_unlock(rq, tsk, &rf);
6393
}
6394

6395
static inline struct task_group *css_tg(struct cgroup_subsys_state *css)
6396
{
6397
	return css ? container_of(css, struct task_group, css) : NULL;
6398 6399
}

6400 6401
static struct cgroup_subsys_state *
cpu_cgroup_css_alloc(struct cgroup_subsys_state *parent_css)
6402
{
6403 6404
	struct task_group *parent = css_tg(parent_css);
	struct task_group *tg;
6405

6406
	if (!parent) {
6407
		/* This is early initialization for the top cgroup */
6408
		return &root_task_group.css;
6409 6410
	}

6411
	tg = sched_create_group(parent);
6412 6413 6414 6415 6416 6417
	if (IS_ERR(tg))
		return ERR_PTR(-ENOMEM);

	return &tg->css;
}

6418 6419 6420 6421 6422 6423 6424 6425 6426 6427 6428
/* 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;
}

6429
static void cpu_cgroup_css_released(struct cgroup_subsys_state *css)
6430
{
6431
	struct task_group *tg = css_tg(css);
6432

6433
	sched_offline_group(tg);
6434 6435
}

6436
static void cpu_cgroup_css_free(struct cgroup_subsys_state *css)
6437
{
6438
	struct task_group *tg = css_tg(css);
6439

6440 6441 6442 6443
	/*
	 * Relies on the RCU grace period between css_released() and this.
	 */
	sched_free_group(tg);
6444 6445
}

6446 6447 6448 6449
/*
 * 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.
 */
6450
static void cpu_cgroup_fork(struct task_struct *task)
6451
{
6452 6453 6454 6455 6456
	struct rq_flags rf;
	struct rq *rq;

	rq = task_rq_lock(task, &rf);

6457
	update_rq_clock(rq);
6458 6459 6460
	sched_change_group(task, TASK_SET_GROUP);

	task_rq_unlock(rq, task, &rf);
6461 6462
}

6463
static int cpu_cgroup_can_attach(struct cgroup_taskset *tset)
6464
{
6465
	struct task_struct *task;
6466
	struct cgroup_subsys_state *css;
6467
	int ret = 0;
6468

6469
	cgroup_taskset_for_each(task, css, tset) {
6470
#ifdef CONFIG_RT_GROUP_SCHED
6471
		if (!sched_rt_can_attach(css_tg(css), task))
6472
			return -EINVAL;
6473
#else
6474 6475 6476
		/* We don't support RT-tasks being in separate groups */
		if (task->sched_class != &fair_sched_class)
			return -EINVAL;
6477
#endif
6478 6479 6480 6481 6482 6483 6484 6485 6486 6487 6488 6489 6490 6491 6492 6493
		/*
		 * 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;
6494
	}
6495
	return ret;
6496
}
6497

6498
static void cpu_cgroup_attach(struct cgroup_taskset *tset)
6499
{
6500
	struct task_struct *task;
6501
	struct cgroup_subsys_state *css;
6502

6503
	cgroup_taskset_for_each(task, css, tset)
6504
		sched_move_task(task);
6505 6506
}

6507
#ifdef CONFIG_FAIR_GROUP_SCHED
6508 6509
static int cpu_shares_write_u64(struct cgroup_subsys_state *css,
				struct cftype *cftype, u64 shareval)
6510
{
6511
	return sched_group_set_shares(css_tg(css), scale_load(shareval));
6512 6513
}

6514 6515
static u64 cpu_shares_read_u64(struct cgroup_subsys_state *css,
			       struct cftype *cft)
6516
{
6517
	struct task_group *tg = css_tg(css);
6518

6519
	return (u64) scale_load_down(tg->shares);
6520
}
6521 6522

#ifdef CONFIG_CFS_BANDWIDTH
6523 6524
static DEFINE_MUTEX(cfs_constraints_mutex);

6525 6526 6527
const u64 max_cfs_quota_period = 1 * NSEC_PER_SEC; /* 1s */
const u64 min_cfs_quota_period = 1 * NSEC_PER_MSEC; /* 1ms */

6528 6529
static int __cfs_schedulable(struct task_group *tg, u64 period, u64 runtime);

6530 6531
static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota)
{
6532
	int i, ret = 0, runtime_enabled, runtime_was_enabled;
6533
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
6534 6535 6536 6537 6538 6539 6540 6541 6542 6543 6544 6545 6546 6547 6548 6549 6550 6551 6552 6553

	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;

6554 6555 6556 6557 6558
	/*
	 * Prevent race between setting of cfs_rq->runtime_enabled and
	 * unthrottle_offline_cfs_rqs().
	 */
	get_online_cpus();
6559 6560 6561 6562 6563
	mutex_lock(&cfs_constraints_mutex);
	ret = __cfs_schedulable(tg, period, quota);
	if (ret)
		goto out_unlock;

6564
	runtime_enabled = quota != RUNTIME_INF;
6565
	runtime_was_enabled = cfs_b->quota != RUNTIME_INF;
6566 6567 6568 6569 6570 6571
	/*
	 * 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();
6572 6573 6574
	raw_spin_lock_irq(&cfs_b->lock);
	cfs_b->period = ns_to_ktime(period);
	cfs_b->quota = quota;
6575

Paul Turner's avatar
Paul Turner committed
6576
	__refill_cfs_bandwidth_runtime(cfs_b);
6577 6578

	/* Restart the period timer (if active) to handle new period expiry: */
6579 6580
	if (runtime_enabled)
		start_cfs_bandwidth(cfs_b);
6581

6582 6583
	raw_spin_unlock_irq(&cfs_b->lock);

6584
	for_each_online_cpu(i) {
6585
		struct cfs_rq *cfs_rq = tg->cfs_rq[i];
6586
		struct rq *rq = cfs_rq->rq;
6587
		struct rq_flags rf;
6588

6589
		rq_lock_irq(rq, &rf);
6590
		cfs_rq->runtime_enabled = runtime_enabled;
6591
		cfs_rq->runtime_remaining = 0;
6592

6593
		if (cfs_rq->throttled)
6594
			unthrottle_cfs_rq(cfs_rq);
6595
		rq_unlock_irq(rq, &rf);
6596
	}
6597 6598
	if (runtime_was_enabled && !runtime_enabled)
		cfs_bandwidth_usage_dec();
6599 6600
out_unlock:
	mutex_unlock(&cfs_constraints_mutex);
6601
	put_online_cpus();
6602

6603
	return ret;
6604 6605 6606 6607 6608 6609
}

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

6610
	period = ktime_to_ns(tg->cfs_bandwidth.period);
6611 6612 6613 6614 6615 6616 6617 6618 6619 6620 6621 6622
	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;

6623
	if (tg->cfs_bandwidth.quota == RUNTIME_INF)
6624 6625
		return -1;

6626
	quota_us = tg->cfs_bandwidth.quota;
6627 6628 6629 6630 6631 6632 6633 6634 6635 6636
	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;
6637
	quota = tg->cfs_bandwidth.quota;
6638 6639 6640 6641 6642 6643 6644 6645

	return tg_set_cfs_bandwidth(tg, period, quota);
}

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

6646
	cfs_period_us = ktime_to_ns(tg->cfs_bandwidth.period);
6647 6648 6649 6650 6651
	do_div(cfs_period_us, NSEC_PER_USEC);

	return cfs_period_us;
}

6652 6653
static s64 cpu_cfs_quota_read_s64(struct cgroup_subsys_state *css,
				  struct cftype *cft)
6654
{
6655
	return tg_get_cfs_quota(css_tg(css));
6656 6657
}

6658 6659
static int cpu_cfs_quota_write_s64(struct cgroup_subsys_state *css,
				   struct cftype *cftype, s64 cfs_quota_us)
6660
{
6661
	return tg_set_cfs_quota(css_tg(css), cfs_quota_us);
6662 6663
}

6664 6665
static u64 cpu_cfs_period_read_u64(struct cgroup_subsys_state *css,
				   struct cftype *cft)
6666
{
6667
	return tg_get_cfs_period(css_tg(css));
6668 6669
}

6670 6671
static int cpu_cfs_period_write_u64(struct cgroup_subsys_state *css,
				    struct cftype *cftype, u64 cfs_period_us)
6672
{
6673
	return tg_set_cfs_period(css_tg(css), cfs_period_us);
6674 6675
}

6676 6677 6678 6679 6680 6681 6682 6683 6684 6685 6686 6687 6688 6689 6690 6691 6692 6693 6694 6695 6696 6697 6698 6699 6700 6701 6702 6703 6704 6705 6706 6707
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;
6708
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
6709 6710 6711 6712 6713
	s64 quota = 0, parent_quota = -1;

	if (!tg->parent) {
		quota = RUNTIME_INF;
	} else {
6714
		struct cfs_bandwidth *parent_b = &tg->parent->cfs_bandwidth;
6715 6716

		quota = normalize_cfs_quota(tg, d);
6717
		parent_quota = parent_b->hierarchical_quota;
6718 6719

		/*
6720 6721
		 * Ensure max(child_quota) <= parent_quota.  On cgroup2,
		 * always take the min.  On cgroup1, only inherit when no
6722
		 * limit is set:
6723
		 */
6724 6725 6726 6727 6728 6729 6730 6731
		if (cgroup_subsys_on_dfl(cpu_cgrp_subsys)) {
			quota = min(quota, parent_quota);
		} else {
			if (quota == RUNTIME_INF)
				quota = parent_quota;
			else if (parent_quota != RUNTIME_INF && quota > parent_quota)
				return -EINVAL;
		}
6732
	}
6733
	cfs_b->hierarchical_quota = quota;
6734 6735 6736 6737 6738 6739

	return 0;
}

static int __cfs_schedulable(struct task_group *tg, u64 period, u64 quota)
{
6740
	int ret;
6741 6742 6743 6744 6745 6746 6747 6748 6749 6750 6751
	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);
	}

6752 6753 6754 6755 6756
	rcu_read_lock();
	ret = walk_tg_tree(tg_cfs_schedulable_down, tg_nop, &data);
	rcu_read_unlock();

	return ret;
6757
}
6758

6759
static int cpu_cfs_stat_show(struct seq_file *sf, void *v)
6760
{
6761
	struct task_group *tg = css_tg(seq_css(sf));
6762
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
6763

6764 6765 6766
	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);
6767

6768 6769 6770 6771 6772 6773 6774 6775 6776 6777
	if (schedstat_enabled() && tg != &root_task_group) {
		u64 ws = 0;
		int i;

		for_each_possible_cpu(i)
			ws += schedstat_val(tg->se[i]->statistics.wait_sum);

		seq_printf(sf, "wait_sum %llu\n", ws);
	}

6778 6779
	return 0;
}
6780
#endif /* CONFIG_CFS_BANDWIDTH */
6781
#endif /* CONFIG_FAIR_GROUP_SCHED */
6782

6783
#ifdef CONFIG_RT_GROUP_SCHED
6784 6785
static int cpu_rt_runtime_write(struct cgroup_subsys_state *css,
				struct cftype *cft, s64 val)
Peter Zijlstra's avatar
Peter Zijlstra committed
6786
{
6787
	return sched_group_set_rt_runtime(css_tg(css), val);
Peter Zijlstra's avatar
Peter Zijlstra committed
6788 6789
}

6790 6791
static s64 cpu_rt_runtime_read(struct cgroup_subsys_state *css,
			       struct cftype *cft)
Peter Zijlstra's avatar
Peter Zijlstra committed
6792
{
6793
	return sched_group_rt_runtime(css_tg(css));
Peter Zijlstra's avatar
Peter Zijlstra committed
6794
}
6795

6796 6797
static int cpu_rt_period_write_uint(struct cgroup_subsys_state *css,
				    struct cftype *cftype, u64 rt_period_us)
6798
{
6799
	return sched_group_set_rt_period(css_tg(css), rt_period_us);
6800 6801
}

6802 6803
static u64 cpu_rt_period_read_uint(struct cgroup_subsys_state *css,
				   struct cftype *cft)
6804
{
6805
	return sched_group_rt_period(css_tg(css));
6806
}
6807
#endif /* CONFIG_RT_GROUP_SCHED */
Peter Zijlstra's avatar
Peter Zijlstra committed
6808

6809
static struct cftype cpu_legacy_files[] = {
6810
#ifdef CONFIG_FAIR_GROUP_SCHED
6811 6812
	{
		.name = "shares",
6813 6814
		.read_u64 = cpu_shares_read_u64,
		.write_u64 = cpu_shares_write_u64,
6815
	},
6816
#endif
6817 6818 6819 6820 6821 6822 6823 6824 6825 6826 6827
#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,
	},
6828 6829
	{
		.name = "stat",
6830
		.seq_show = cpu_cfs_stat_show,
6831
	},
6832
#endif
6833
#ifdef CONFIG_RT_GROUP_SCHED
Peter Zijlstra's avatar
Peter Zijlstra committed
6834
	{
Peter Zijlstra's avatar
Peter Zijlstra committed
6835
		.name = "rt_runtime_us",
6836 6837
		.read_s64 = cpu_rt_runtime_read,
		.write_s64 = cpu_rt_runtime_write,
Peter Zijlstra's avatar
Peter Zijlstra committed
6838
	},
6839 6840
	{
		.name = "rt_period_us",
6841 6842
		.read_u64 = cpu_rt_period_read_uint,
		.write_u64 = cpu_rt_period_write_uint,
6843
	},
6844
#endif
6845
	{ }	/* Terminate */
6846 6847
};

6848 6849
static int cpu_extra_stat_show(struct seq_file *sf,
			       struct cgroup_subsys_state *css)
6850 6851 6852
{
#ifdef CONFIG_CFS_BANDWIDTH
	{
6853
		struct task_group *tg = css_tg(css);
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
		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;
6920
	int idx;
6921 6922 6923 6924

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

6925 6926 6927 6928
	idx = NICE_TO_PRIO(nice) - MAX_RT_PRIO;
	idx = array_index_nospec(idx, 40);
	weight = sched_prio_to_weight[idx];

6929 6930 6931 6932 6933 6934 6935 6936 6937 6938 6939 6940 6941 6942 6943 6944 6945 6946 6947 6948 6949
	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 */

6950
	if (sscanf(buf, "%20s %llu", tok, periodp) < 1)
6951 6952 6953 6954 6955 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 6997 6998 6999 7000 7001 7002 7003 7004 7005 7006 7007 7008 7009 7010 7011 7012 7013 7014
		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 */
};

7015
struct cgroup_subsys cpu_cgrp_subsys = {
7016
	.css_alloc	= cpu_cgroup_css_alloc,
7017
	.css_online	= cpu_cgroup_css_online,
7018
	.css_released	= cpu_cgroup_css_released,
7019
	.css_free	= cpu_cgroup_css_free,
7020
	.css_extra_stat_show = cpu_extra_stat_show,
7021
	.fork		= cpu_cgroup_fork,
7022 7023
	.can_attach	= cpu_cgroup_can_attach,
	.attach		= cpu_cgroup_attach,
7024
	.legacy_cftypes	= cpu_legacy_files,
7025
	.dfl_cftypes	= cpu_files,
7026
	.early_init	= true,
7027
	.threaded	= true,
7028 7029
};

7030
#endif	/* CONFIG_CGROUP_SCHED */
7031

7032 7033 7034 7035 7036
void dump_cpu_task(int cpu)
{
	pr_info("Task dump for CPU %d:\n", cpu);
	sched_show_task(cpu_curr(cpu));
}
7037 7038 7039 7040 7041 7042 7043 7044 7045 7046 7047 7048 7049 7050 7051 7052 7053 7054 7055 7056 7057 7058 7059 7060 7061 7062 7063 7064 7065 7066 7067 7068 7069 7070 7071 7072 7073 7074 7075 7076 7077

/*
 * 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,
};
7078 7079

#undef CREATE_TRACE_POINTS