core.c 198 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0-only
Linus Torvalds's avatar
Linus Torvalds committed
2
/*
3
 *  kernel/sched/core.c
Linus Torvalds's avatar
Linus Torvalds committed
4
 *
5
 *  Core kernel scheduler code and related syscalls
Linus Torvalds's avatar
Linus Torvalds committed
6 7 8
 *
 *  Copyright (C) 1991-2002  Linus Torvalds
 */
9
#include "sched.h"
Linus Torvalds's avatar
Linus Torvalds committed
10

11
#include <linux/nospec.h>
12

13 14
#include <linux/kcov.h>

15
#include <asm/switch_to.h>
16
#include <asm/tlb.h>
Linus Torvalds's avatar
Linus Torvalds committed
17

18
#include "../workqueue_internal.h"
19
#include "../../fs/io-wq.h"
20
#include "../smpboot.h"
21

22 23
#include "pelt.h"

24
#define CREATE_TRACE_POINTS
25
#include <trace/events/sched.h>
26

27 28 29 30 31 32 33 34 35 36 37
/*
 * Export tracepoints that act as a bare tracehook (ie: have no trace event
 * associated with them) to allow external modules to probe them.
 */
EXPORT_TRACEPOINT_SYMBOL_GPL(pelt_cfs_tp);
EXPORT_TRACEPOINT_SYMBOL_GPL(pelt_rt_tp);
EXPORT_TRACEPOINT_SYMBOL_GPL(pelt_dl_tp);
EXPORT_TRACEPOINT_SYMBOL_GPL(pelt_irq_tp);
EXPORT_TRACEPOINT_SYMBOL_GPL(pelt_se_tp);
EXPORT_TRACEPOINT_SYMBOL_GPL(sched_overutilized_tp);

38
DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
39

40
#if defined(CONFIG_SCHED_DEBUG) && defined(CONFIG_JUMP_LABEL)
41 42
/*
 * Debugging: various feature bits
43 44 45 46
 *
 * 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.
47
 */
48 49
#define SCHED_FEAT(name, enabled)	\
	(1UL << __SCHED_FEAT_##name) * enabled |
50
const_debug unsigned int sysctl_sched_features =
51
#include "features.h"
52 53
	0;
#undef SCHED_FEAT
54
#endif
55

56 57 58 59 60 61
/*
 * 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
62
/*
63
 * period over which we measure -rt task CPU usage in us.
Peter Zijlstra's avatar
Peter Zijlstra committed
64 65
 * default: 1s
 */
Peter Zijlstra's avatar
Peter Zijlstra committed
66
unsigned int sysctl_sched_rt_period = 1000000;
Peter Zijlstra's avatar
Peter Zijlstra committed
67

68
__read_mostly int scheduler_running;
69

Peter Zijlstra's avatar
Peter Zijlstra committed
70 71 72 73 74
/*
 * 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
75

76 77 78
/*
 * __task_rq_lock - lock the rq @p resides on.
 */
79
struct rq *__task_rq_lock(struct task_struct *p, struct rq_flags *rf)
80 81 82 83 84 85 86 87 88 89
	__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))) {
90
			rq_pin_lock(rq, rf);
91 92 93 94 95 96 97 98 99 100 101 102
			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.
 */
103
struct rq *task_rq_lock(struct task_struct *p, struct rq_flags *rf)
104 105 106 107 108 109
	__acquires(p->pi_lock)
	__acquires(rq->lock)
{
	struct rq *rq;

	for (;;) {
110
		raw_spin_lock_irqsave(&p->pi_lock, rf->flags);
111 112 113 114 115 116 117 118 119 120 121 122
		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)
		 *
123
		 * If we observe the old CPU in task_rq_lock(), the acquire of
124 125
		 * the old rq->lock will fully serialize against the stores.
		 *
126 127 128
		 * 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.
129 130
		 */
		if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) {
131
			rq_pin_lock(rq, rf);
132 133 134
			return rq;
		}
		raw_spin_unlock(&rq->lock);
135
		raw_spin_unlock_irqrestore(&p->pi_lock, rf->flags);
136 137 138 139 140 141

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

142 143 144 145 146 147 148 149 150 151
/*
 * 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...
 */
152 153
	s64 __maybe_unused steal = 0, irq_delta = 0;

154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192
#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;

193
#ifdef CONFIG_HAVE_SCHED_AVG_IRQ
194
	if ((irq_delta + steal) && sched_feat(NONTASK_CAPACITY))
195
		update_irq_load_avg(rq, irq_delta + steal);
196
#endif
197
	update_rq_clock_pelt(rq, delta);
198 199 200 201 202 203 204 205 206 207 208 209
}

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
210 211
	if (sched_feat(WARN_DOUBLE_CLOCK))
		SCHED_WARN_ON(rq->clock_update_flags & RQCF_UPDATED);
212 213
	rq->clock_update_flags |= RQCF_UPDATED;
#endif
214

215 216 217 218 219 220 221 222
	delta = sched_clock_cpu(cpu_of(rq)) - rq->clock;
	if (delta < 0)
		return;
	rq->clock += delta;
	update_rq_clock_task(rq, delta);
}


223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240
#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);
241
	struct rq_flags rf;
242 243 244

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

245
	rq_lock(rq, &rf);
246
	update_rq_clock(rq);
247
	rq->curr->sched_class->task_tick(rq, rq->curr, 1);
248
	rq_unlock(rq, &rf);
249 250 251 252

	return HRTIMER_NORESTART;
}

253
#ifdef CONFIG_SMP
Peter Zijlstra's avatar
Peter Zijlstra committed
254

255
static void __hrtick_restart(struct rq *rq)
Peter Zijlstra's avatar
Peter Zijlstra committed
256 257 258
{
	struct hrtimer *timer = &rq->hrtick_timer;

259
	hrtimer_start_expires(timer, HRTIMER_MODE_ABS_PINNED_HARD);
Peter Zijlstra's avatar
Peter Zijlstra committed
260 261
}

262 263 264 265
/*
 * called from hardirq (IPI) context
 */
static void __hrtick_start(void *arg)
266
{
267
	struct rq *rq = arg;
268
	struct rq_flags rf;
269

270
	rq_lock(rq, &rf);
Peter Zijlstra's avatar
Peter Zijlstra committed
271
	__hrtick_restart(rq);
272
	rq_unlock(rq, &rf);
273 274
}

275 276 277 278 279
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
280
void hrtick_start(struct rq *rq, u64 delay)
281
{
282
	struct hrtimer *timer = &rq->hrtick_timer;
283 284 285 286 287 288 289 290 291
	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);
292

293
	hrtimer_set_expires(timer, time);
294

295
	if (rq == this_rq())
Peter Zijlstra's avatar
Peter Zijlstra committed
296
		__hrtick_restart(rq);
297
	else
298
		smp_call_function_single_async(cpu_of(rq), &rq->hrtick_csd);
299 300
}

301 302 303 304 305 306
#else
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
307
void hrtick_start(struct rq *rq, u64 delay)
308
{
Wanpeng Li's avatar
Wanpeng Li committed
309 310 311 312 313
	/*
	 * Don't schedule slices shorter than 10000ns, that just
	 * doesn't make sense. Rely on vruntime for fairness.
	 */
	delay = max_t(u64, delay, 10000LL);
314
	hrtimer_start(&rq->hrtick_timer, ns_to_ktime(delay),
315
		      HRTIMER_MODE_REL_PINNED_HARD);
316 317
}
#endif /* CONFIG_SMP */
318

319
static void hrtick_rq_init(struct rq *rq)
320
{
321 322 323 324 325
#ifdef CONFIG_SMP
	rq->hrtick_csd.flags = 0;
	rq->hrtick_csd.func = __hrtick_start;
	rq->hrtick_csd.info = rq;
#endif
326

327
	hrtimer_init(&rq->hrtick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL_HARD);
328
	rq->hrtick_timer.function = hrtick;
329
}
Andrew Morton's avatar
Andrew Morton committed
330
#else	/* CONFIG_SCHED_HRTICK */
331 332 333 334
static inline void hrtick_clear(struct rq *rq)
{
}

335
static inline void hrtick_rq_init(struct rq *rq)
336 337
{
}
Andrew Morton's avatar
Andrew Morton committed
338
#endif	/* CONFIG_SCHED_HRTICK */
339

340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357
/*
 * 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;								\
})

358
#if defined(CONFIG_SMP) && defined(TIF_POLLING_NRFLAG)
359 360 361 362 363 364 365 366 367 368
/*
 * 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);
}
369 370 371 372 373 374 375 376 377 378

/*
 * 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);
379
	typeof(ti->flags) old, val = READ_ONCE(ti->flags);
380 381 382 383 384 385 386 387 388 389 390 391 392 393

	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;
}

394 395 396 397 398 399
#else
static bool set_nr_and_not_polling(struct task_struct *p)
{
	set_tsk_need_resched(p);
	return true;
}
400 401 402 403 404 405 406

#ifdef CONFIG_SMP
static bool set_nr_if_polling(struct task_struct *p)
{
	return false;
}
#endif
407 408
#endif

409
static bool __wake_q_add(struct wake_q_head *head, struct task_struct *task)
410 411 412 413 414 415 416 417
{
	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.
	 *
418 419
	 * 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.
420
	 */
421
	smp_mb__before_atomic();
422
	if (unlikely(cmpxchg_relaxed(&node->next, NULL, WAKE_Q_TAIL)))
423
		return false;
424 425 426 427 428 429

	/*
	 * The head is context local, there can be no concurrency.
	 */
	*head->lastp = node;
	head->lastp = &node->next;
430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471
	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);
472 473 474 475 476 477 478 479 480 481 482
}

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);
483
		/* Task can safely be re-inserted now: */
484 485 486 487
		node = node->next;
		task->wake_q.next = NULL;

		/*
488 489
		 * wake_up_process() executes a full barrier, which pairs with
		 * the queueing in wake_q_add() so as not to miss wakeups.
490 491 492 493 494 495
		 */
		wake_up_process(task);
		put_task_struct(task);
	}
}

496
/*
497
 * resched_curr - mark rq's current task 'to be rescheduled now'.
498 499 500 501 502
 *
 * 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.
 */
503
void resched_curr(struct rq *rq)
504
{
505
	struct task_struct *curr = rq->curr;
506 507
	int cpu;

508
	lockdep_assert_held(&rq->lock);
509

510
	if (test_tsk_need_resched(curr))
511 512
		return;

513
	cpu = cpu_of(rq);
514

515
	if (cpu == smp_processor_id()) {
516
		set_tsk_need_resched(curr);
517
		set_preempt_need_resched();
518
		return;
519
	}
520

521
	if (set_nr_and_not_polling(curr))
522
		smp_send_reschedule(cpu);
523 524
	else
		trace_sched_wake_idle_without_ipi(cpu);
525 526
}

527
void resched_cpu(int cpu)
528 529 530 531
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

532
	raw_spin_lock_irqsave(&rq->lock, flags);
533 534
	if (cpu_online(cpu) || cpu == smp_processor_id())
		resched_curr(rq);
535
	raw_spin_unlock_irqrestore(&rq->lock, flags);
536
}
537

538
#ifdef CONFIG_SMP
539
#ifdef CONFIG_NO_HZ_COMMON
540
/*
541 542
 * In the semi idle case, use the nearest busy CPU for migrating timers
 * from an idle CPU.  This is good for power-savings.
543 544
 *
 * We don't do similar optimization for completely idle system, as
545 546
 * 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).
547
 */
548
int get_nohz_timer_target(void)
549
{
550
	int i, cpu = smp_processor_id(), default_cpu = -1;
551 552
	struct sched_domain *sd;

553 554 555 556 557
	if (housekeeping_cpu(cpu, HK_FLAG_TIMER)) {
		if (!idle_cpu(cpu))
			return cpu;
		default_cpu = cpu;
	}
558

559
	rcu_read_lock();
560
	for_each_domain(cpu, sd) {
561 562
		for_each_cpu_and(i, sched_domain_span(sd),
			housekeeping_cpumask(HK_FLAG_TIMER)) {
563 564 565
			if (cpu == i)
				continue;

566
			if (!idle_cpu(i)) {
567 568 569 570
				cpu = i;
				goto unlock;
			}
		}
571
	}
572

573 574 575
	if (default_cpu == -1)
		default_cpu = housekeeping_any_cpu(HK_FLAG_TIMER);
	cpu = default_cpu;
576 577
unlock:
	rcu_read_unlock();
578 579
	return cpu;
}
580

581 582 583 584 585 586 587 588 589 590
/*
 * 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.
 */
591
static void wake_up_idle_cpu(int cpu)
592 593 594 595 596 597
{
	struct rq *rq = cpu_rq(cpu);

	if (cpu == smp_processor_id())
		return;

598
	if (set_nr_and_not_polling(rq->idle))
599
		smp_send_reschedule(cpu);
600 601
	else
		trace_sched_wake_idle_without_ipi(cpu);
602 603
}

604
static bool wake_up_full_nohz_cpu(int cpu)
605
{
606 607 608 609 610 611
	/*
	 * 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.
	 */
612 613
	if (cpu_is_offline(cpu))
		return true;  /* Don't try to wake offline CPUs. */
614
	if (tick_nohz_full_cpu(cpu)) {
615 616
		if (cpu != smp_processor_id() ||
		    tick_nohz_tick_stopped())
617
			tick_nohz_full_kick_cpu(cpu);
618 619 620 621 622 623
		return true;
	}

	return false;
}

624 625 626 627 628
/*
 * 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.
 */
629 630
void wake_up_nohz_cpu(int cpu)
{
631
	if (!wake_up_full_nohz_cpu(cpu))
632 633 634
		wake_up_idle_cpu(cpu);
}

635
static inline bool got_nohz_idle_kick(void)
636
{
637
	int cpu = smp_processor_id();
638

639
	if (!(atomic_read(nohz_flags(cpu)) & NOHZ_KICK_MASK))
640 641 642 643 644 645 646 647 648
		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
	 */
649
	atomic_andnot(NOHZ_KICK_MASK, nohz_flags(cpu));
650
	return false;
651 652
}

653
#else /* CONFIG_NO_HZ_COMMON */
654

655
static inline bool got_nohz_idle_kick(void)
656
{
657
	return false;
658 659
}

660
#endif /* CONFIG_NO_HZ_COMMON */
661

662
#ifdef CONFIG_NO_HZ_FULL
663
bool sched_can_stop_tick(struct rq *rq)
664
{
665 666 667 668 669 670
	int fifo_nr_running;

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

671
	/*
672 673
	 * If there are more than one RR tasks, we need the tick to effect the
	 * actual RR behaviour.
674
	 */
675 676 677 678 679
	if (rq->rt.rr_nr_running) {
		if (rq->rt.rr_nr_running == 1)
			return true;
		else
			return false;
680 681
	}

682 683 684 685 686 687 688 689 690 691 692 693 694 695
	/*
	 * 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)
696
		return false;
697

698
	return true;
699 700
}
#endif /* CONFIG_NO_HZ_FULL */
701
#endif /* CONFIG_SMP */
702

703 704
#if defined(CONFIG_RT_GROUP_SCHED) || (defined(CONFIG_FAIR_GROUP_SCHED) && \
			(defined(CONFIG_SMP) || defined(CONFIG_CFS_BANDWIDTH)))
705
/*
706 707 708 709
 * 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.
710
 */
711
int walk_tg_tree_from(struct task_group *from,
712
			     tg_visitor down, tg_visitor up, void *data)
713 714
{
	struct task_group *parent, *child;
715
	int ret;
716

717 718
	parent = from;

719
down:
720 721
	ret = (*down)(parent, data);
	if (ret)
722
		goto out;
723 724 725 726 727 728 729
	list_for_each_entry_rcu(child, &parent->children, siblings) {
		parent = child;
		goto down;

up:
		continue;
	}
730
	ret = (*up)(parent, data);
731 732
	if (ret || parent == from)
		goto out;
733 734 735 736 737

	child = parent;
	parent = parent->parent;
	if (parent)
		goto up;
738
out:
739
	return ret;
740 741
}

742
int tg_nop(struct task_group *tg, void *data)
743
{
744
	return 0;
745
}
746 747
#endif

748
static void set_load_weight(struct task_struct *p, bool update_load)
749
{
750 751 752
	int prio = p->static_prio - MAX_RT_PRIO;
	struct load_weight *load = &p->se.load;

Ingo Molnar's avatar
Ingo Molnar committed
753 754 755
	/*
	 * SCHED_IDLE tasks get minimal weight:
	 */
756
	if (task_has_idle_policy(p)) {
757
		load->weight = scale_load(WEIGHT_IDLEPRIO);
758
		load->inv_weight = WMULT_IDLEPRIO;
Ingo Molnar's avatar
Ingo Molnar committed
759 760
		return;
	}
761

762 763 764 765 766 767 768 769 770 771
	/*
	 * 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];
	}
772 773
}

774
#ifdef CONFIG_UCLAMP_TASK
775 776 777 778 779 780 781 782 783 784 785 786
/*
 * Serializes updates of utilization clamp values
 *
 * The (slow-path) user-space triggers utilization clamp value updates which
 * can require updates on (fast-path) scheduler's data structures used to
 * support enqueue/dequeue operations.
 * While the per-CPU rq lock protects fast-path update operations, user-space
 * requests are serialized using a mutex to reduce the risk of conflicting
 * updates or API abuses.
 */
static DEFINE_MUTEX(uclamp_mutex);

787 788 789 790 791 792 793 794
/* Max allowed minimum utilization */
unsigned int sysctl_sched_uclamp_util_min = SCHED_CAPACITY_SCALE;

/* Max allowed maximum utilization */
unsigned int sysctl_sched_uclamp_util_max = SCHED_CAPACITY_SCALE;

/* All clamps are required to be less or equal than these values */
static struct uclamp_se uclamp_default[UCLAMP_CNT];
795 796 797 798 799 800 801 802 803 804 805 806

/* Integer rounded range for each bucket */
#define UCLAMP_BUCKET_DELTA DIV_ROUND_CLOSEST(SCHED_CAPACITY_SCALE, UCLAMP_BUCKETS)

#define for_each_clamp_id(clamp_id) \
	for ((clamp_id) = 0; (clamp_id) < UCLAMP_CNT; (clamp_id)++)

static inline unsigned int uclamp_bucket_id(unsigned int clamp_value)
{
	return clamp_value / UCLAMP_BUCKET_DELTA;
}

807 808 809 810 811
static inline unsigned int uclamp_bucket_base_value(unsigned int clamp_value)
{
	return UCLAMP_BUCKET_DELTA * uclamp_bucket_id(clamp_value);
}

812
static inline unsigned int uclamp_none(enum uclamp_id clamp_id)
813 814 815 816 817 818
{
	if (clamp_id == UCLAMP_MIN)
		return 0;
	return SCHED_CAPACITY_SCALE;
}

819 820
static inline void uclamp_se_set(struct uclamp_se *uc_se,
				 unsigned int value, bool user_defined)
821 822 823
{
	uc_se->value = value;
	uc_se->bucket_id = uclamp_bucket_id(value);
824
	uc_se->user_defined = user_defined;
825 826
}

827
static inline unsigned int
828
uclamp_idle_value(struct rq *rq, enum uclamp_id clamp_id,
829 830 831 832 833 834 835 836 837 838 839 840 841 842 843
		  unsigned int clamp_value)
{
	/*
	 * Avoid blocked utilization pushing up the frequency when we go
	 * idle (which drops the max-clamp) by retaining the last known
	 * max-clamp.
	 */
	if (clamp_id == UCLAMP_MAX) {
		rq->uclamp_flags |= UCLAMP_FLAG_IDLE;
		return clamp_value;
	}

	return uclamp_none(UCLAMP_MIN);
}

844
static inline void uclamp_idle_reset(struct rq *rq, enum uclamp_id clamp_id,
845 846 847 848 849 850 851 852 853
				     unsigned int clamp_value)
{
	/* Reset max-clamp retention only on idle exit */
	if (!(rq->uclamp_flags & UCLAMP_FLAG_IDLE))
		return;

	WRITE_ONCE(rq->uclamp[clamp_id].value, clamp_value);
}

854
static inline
855
unsigned int uclamp_rq_max_value(struct rq *rq, enum uclamp_id clamp_id,
856
				   unsigned int clamp_value)
857 858 859 860 861 862 863 864 865 866 867 868 869 870 871
{
	struct uclamp_bucket *bucket = rq->uclamp[clamp_id].bucket;
	int bucket_id = UCLAMP_BUCKETS - 1;

	/*
	 * Since both min and max clamps are max aggregated, find the
	 * top most bucket with tasks in.
	 */
	for ( ; bucket_id >= 0; bucket_id--) {
		if (!bucket[bucket_id].tasks)
			continue;
		return bucket[bucket_id].value;
	}

	/* No tasks -- default clamp values */
872
	return uclamp_idle_value(rq, clamp_id, clamp_value);
873 874
}

875
static inline struct uclamp_se
876
uclamp_tg_restrict(struct task_struct *p, enum uclamp_id clamp_id)
877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898
{
	struct uclamp_se uc_req = p->uclamp_req[clamp_id];
#ifdef CONFIG_UCLAMP_TASK_GROUP
	struct uclamp_se uc_max;

	/*
	 * Tasks in autogroups or root task group will be
	 * restricted by system defaults.
	 */
	if (task_group_is_autogroup(task_group(p)))
		return uc_req;
	if (task_group(p) == &root_task_group)
		return uc_req;

	uc_max = task_group(p)->uclamp[clamp_id];
	if (uc_req.value > uc_max.value || !uc_req.user_defined)
		return uc_max;
#endif

	return uc_req;
}

899 900 901 902
/*
 * The effective clamp bucket index of a task depends on, by increasing
 * priority:
 * - the task specific clamp value, when explicitly requested from userspace
903 904
 * - the task group effective clamp value, for tasks not either in the root
 *   group or in an autogroup
905 906 907
 * - the system default clamp value, defined by the sysadmin
 */
static inline struct uclamp_se
908
uclamp_eff_get(struct task_struct *p, enum uclamp_id clamp_id)
909
{
910
	struct uclamp_se uc_req = uclamp_tg_restrict(p, clamp_id);
911 912 913 914 915 916 917 918 919
	struct uclamp_se uc_max = uclamp_default[clamp_id];

	/* System default restrictions always apply */
	if (unlikely(uc_req.value > uc_max.value))
		return uc_max;

	return uc_req;
}

920
unsigned long uclamp_eff_value(struct task_struct *p, enum uclamp_id clamp_id)
921 922 923 924 925
{
	struct uclamp_se uc_eff;

	/* Task currently refcounted: use back-annotated (effective) value */
	if (p->uclamp[clamp_id].active)
926
		return (unsigned long)p->uclamp[clamp_id].value;
927 928 929

	uc_eff = uclamp_eff_get(p, clamp_id);

930
	return (unsigned long)uc_eff.value;
931 932
}

933 934 935 936
/*
 * When a task is enqueued on a rq, the clamp bucket currently defined by the
 * task's uclamp::bucket_id is refcounted on that rq. This also immediately
 * updates the rq's clamp value if required.
937 938 939 940 941
 *
 * Tasks can have a task-specific value requested from user-space, track
 * within each bucket the maximum value for tasks refcounted in it.
 * This "local max aggregation" allows to track the exact "requested" value
 * for each bucket when all its RUNNABLE tasks require the same clamp.
942 943
 */
static inline void uclamp_rq_inc_id(struct rq *rq, struct task_struct *p,
944
				    enum uclamp_id clamp_id)
945 946 947 948 949 950 951
{
	struct uclamp_rq *uc_rq = &rq->uclamp[clamp_id];
	struct uclamp_se *uc_se = &p->uclamp[clamp_id];
	struct uclamp_bucket *bucket;

	lockdep_assert_held(&rq->lock);

952 953 954
	/* Update task effective clamp */
	p->uclamp[clamp_id] = uclamp_eff_get(p, clamp_id);

955 956
	bucket = &uc_rq->bucket[uc_se->bucket_id];
	bucket->tasks++;
957
	uc_se->active = true;
958

959 960
	uclamp_idle_reset(rq, clamp_id, uc_se->value);

961 962 963 964 965 966 967
	/*
	 * Local max aggregation: rq buckets always track the max
	 * "requested" clamp value of its RUNNABLE tasks.
	 */
	if (bucket->tasks == 1 || uc_se->value > bucket->value)
		bucket->value = uc_se->value;

968
	if (uc_se->value > READ_ONCE(uc_rq->value))
969
		WRITE_ONCE(uc_rq->value, uc_se->value);
970 971 972 973 974 975 976 977 978 979 980 981
}

/*
 * When a task is dequeued from a rq, the clamp bucket refcounted by the task
 * is released. If this is the last task reference counting the rq's max
 * active clamp value, then the rq's clamp value is updated.
 *
 * Both refcounted tasks and rq's cached clamp values are expected to be
 * always valid. If it's detected they are not, as defensive programming,
 * enforce the expected state and warn.
 */
static inline void uclamp_rq_dec_id(struct rq *rq, struct task_struct *p,
982
				    enum uclamp_id clamp_id)
983 984 985 986
{
	struct uclamp_rq *uc_rq = &rq->uclamp[clamp_id];
	struct uclamp_se *uc_se = &p->uclamp[clamp_id];
	struct uclamp_bucket *bucket;
987
	unsigned int bkt_clamp;
988 989 990 991 992 993 994 995
	unsigned int rq_clamp;

	lockdep_assert_held(&rq->lock);

	bucket = &uc_rq->bucket[uc_se->bucket_id];
	SCHED_WARN_ON(!bucket->tasks);
	if (likely(bucket->tasks))
		bucket->tasks--;
996
	uc_se->active = false;
997

998 999 1000 1001 1002 1003
	/*
	 * Keep "local max aggregation" simple and accept to (possibly)
	 * overboost some RUNNABLE tasks in the same bucket.
	 * The rq clamp bucket value is reset to its base value whenever
	 * there are no more RUNNABLE tasks refcounting it.
	 */
1004 1005 1006 1007 1008 1009 1010 1011 1012
	if (likely(bucket->tasks))
		return;

	rq_clamp = READ_ONCE(uc_rq->value);
	/*
	 * Defensive programming: this should never happen. If it happens,
	 * e.g. due to future modification, warn and fixup the expected value.
	 */
	SCHED_WARN_ON(bucket->value > rq_clamp);
1013 1014 1015 1016
	if (bucket->value >= rq_clamp) {
		bkt_clamp = uclamp_rq_max_value(rq, clamp_id, uc_se->value);
		WRITE_ONCE(uc_rq->value, bkt_clamp);
	}
1017 1018 1019 1020
}

static inline void uclamp_rq_inc(struct rq *rq, struct task_struct *p)
{
1021
	enum uclamp_id clamp_id;
1022 1023 1024 1025 1026 1027

	if (unlikely(!p->sched_class->uclamp_enabled))
		return;

	for_each_clamp_id(clamp_id)
		uclamp_rq_inc_id(rq, p, clamp_id);
1028 1029 1030 1031

	/* Reset clamp idle holding when there is one RUNNABLE task */
	if (rq->uclamp_flags & UCLAMP_FLAG_IDLE)
		rq->uclamp_flags &= ~UCLAMP_FLAG_IDLE;
1032 1033 1034 1035
}

static inline void uclamp_rq_dec(struct rq *rq, struct task_struct *p)
{
1036
	enum uclamp_id clamp_id;
1037 1038 1039 1040 1041 1042 1043 1044

	if (unlikely(!p->sched_class->uclamp_enabled))
		return;

	for_each_clamp_id(clamp_id)
		uclamp_rq_dec_id(rq, p, clamp_id);
}

1045
static inline void
1046
uclamp_update_active(struct task_struct *p, enum uclamp_id clamp_id)
1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066
{
	struct rq_flags rf;
	struct rq *rq;

	/*
	 * Lock the task and the rq where the task is (or was) queued.
	 *
	 * We might lock the (previous) rq of a !RUNNABLE task, but that's the
	 * price to pay to safely serialize util_{min,max} updates with
	 * enqueues, dequeues and migration operations.
	 * This is the same locking schema used by __set_cpus_allowed_ptr().
	 */
	rq = task_rq_lock(p, &rf);

	/*
	 * Setting the clamp bucket is serialized by task_rq_lock().
	 * If the task is not yet RUNNABLE and its task_struct is not
	 * affecting a valid clamp bucket, the next time it's enqueued,
	 * it will already see the updated clamp bucket value.
	 */
1067
	if (p->uclamp[clamp_id].active) {
1068 1069 1070 1071 1072 1073 1074
		uclamp_rq_dec_id(rq, p, clamp_id);
		uclamp_rq_inc_id(rq, p, clamp_id);
	}

	task_rq_unlock(rq, p, &rf);
}

1075
#ifdef CONFIG_UCLAMP_TASK_GROUP
1076 1077 1078 1079
static inline void
uclamp_update_active_tasks(struct cgroup_subsys_state *css,
			   unsigned int clamps)
{
1080
	enum uclamp_id clamp_id;
1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093
	struct css_task_iter it;
	struct task_struct *p;

	css_task_iter_start(css, 0, &it);
	while ((p = css_task_iter_next(&it))) {
		for_each_clamp_id(clamp_id) {
			if ((0x1 << clamp_id) & clamps)
				uclamp_update_active(p, clamp_id);
		}
	}
	css_task_iter_end(&it);
}

1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111
static void cpu_util_update_eff(struct cgroup_subsys_state *css);
static void uclamp_update_root_tg(void)
{
	struct task_group *tg = &root_task_group;

	uclamp_se_set(&tg->uclamp_req[UCLAMP_MIN],
		      sysctl_sched_uclamp_util_min, false);
	uclamp_se_set(&tg->uclamp_req[UCLAMP_MAX],
		      sysctl_sched_uclamp_util_max, false);

	rcu_read_lock();
	cpu_util_update_eff(&root_task_group.css);
	rcu_read_unlock();
}
#else
static void uclamp_update_root_tg(void) { }
#endif

1112 1113 1114 1115
int sysctl_sched_uclamp_handler(struct ctl_table *table, int write,
				void __user *buffer, size_t *lenp,
				loff_t *ppos)
{
1116
	bool update_root_tg = false;
1117 1118 1119
	int old_min, old_max;
	int result;

1120
	mutex_lock(&uclamp_mutex);
1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137
	old_min = sysctl_sched_uclamp_util_min;
	old_max = sysctl_sched_uclamp_util_max;

	result = proc_dointvec(table, write, buffer, lenp, ppos);
	if (result)
		goto undo;
	if (!write)
		goto done;

	if (sysctl_sched_uclamp_util_min > sysctl_sched_uclamp_util_max ||
	    sysctl_sched_uclamp_util_max > SCHED_CAPACITY_SCALE) {
		result = -EINVAL;
		goto undo;
	}

	if (old_min != sysctl_sched_uclamp_util_min) {
		uclamp_se_set(&uclamp_default[UCLAMP_MIN],
1138
			      sysctl_sched_uclamp_util_min, false);
1139
		update_root_tg = true;
1140 1141 1142
	}
	if (old_max != sysctl_sched_uclamp_util_max) {
		uclamp_se_set(&uclamp_default[UCLAMP_MAX],
1143
			      sysctl_sched_uclamp_util_max, false);
1144
		update_root_tg = true;
1145 1146
	}

1147 1148 1149
	if (update_root_tg)
		uclamp_update_root_tg();

1150
	/*
1151 1152 1153
	 * We update all RUNNABLE tasks only when task groups are in use.
	 * Otherwise, keep it simple and do just a lazy update at each next
	 * task enqueue time.
1154
	 */
1155

1156 1157 1158 1159 1160 1161
	goto done;

undo:
	sysctl_sched_uclamp_util_min = old_min;
	sysctl_sched_uclamp_util_max = old_max;
done:
1162
	mutex_unlock(&uclamp_mutex);
1163 1164 1165 1166

	return result;
}

1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188
static int uclamp_validate(struct task_struct *p,
			   const struct sched_attr *attr)
{
	unsigned int lower_bound = p->uclamp_req[UCLAMP_MIN].value;
	unsigned int upper_bound = p->uclamp_req[UCLAMP_MAX].value;

	if (attr->sched_flags & SCHED_FLAG_UTIL_CLAMP_MIN)
		lower_bound = attr->sched_util_min;
	if (attr->sched_flags & SCHED_FLAG_UTIL_CLAMP_MAX)
		upper_bound = attr->sched_util_max;

	if (lower_bound > upper_bound)
		return -EINVAL;
	if (upper_bound > SCHED_CAPACITY_SCALE)
		return -EINVAL;

	return 0;
}

static void __setscheduler_uclamp(struct task_struct *p,
				  const struct sched_attr *attr)
{
1189
	enum uclamp_id clamp_id;
1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209

	/*
	 * On scheduling class change, reset to default clamps for tasks
	 * without a task-specific value.
	 */
	for_each_clamp_id(clamp_id) {
		struct uclamp_se *uc_se = &p->uclamp_req[clamp_id];
		unsigned int clamp_value = uclamp_none(clamp_id);

		/* Keep using defined clamps across class changes */
		if (uc_se->user_defined)
			continue;

		/* By default, RT tasks always get 100% boost */
		if (unlikely(rt_task(p) && clamp_id == UCLAMP_MIN))
			clamp_value = uclamp_none(UCLAMP_MAX);

		uclamp_se_set(uc_se, clamp_value, false);
	}

1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223
	if (likely(!(attr->sched_flags & SCHED_FLAG_UTIL_CLAMP)))
		return;

	if (attr->sched_flags & SCHED_FLAG_UTIL_CLAMP_MIN) {
		uclamp_se_set(&p->uclamp_req[UCLAMP_MIN],
			      attr->sched_util_min, true);
	}

	if (attr->sched_flags & SCHED_FLAG_UTIL_CLAMP_MAX) {
		uclamp_se_set(&p->uclamp_req[UCLAMP_MAX],
			      attr->sched_util_max, true);
	}
}

1224 1225
static void uclamp_fork(struct task_struct *p)
{
1226
	enum uclamp_id clamp_id;
1227 1228 1229

	for_each_clamp_id(clamp_id)
		p->uclamp[clamp_id].active = false;
1230 1231 1232 1233 1234

	if (likely(!p->sched_reset_on_fork))
		return;

	for_each_clamp_id(clamp_id) {
1235 1236
		uclamp_se_set(&p->uclamp_req[clamp_id],
			      uclamp_none(clamp_id), false);
1237
	}
1238 1239
}

1240 1241
static void __init init_uclamp(void)
{
1242
	struct uclamp_se uc_max = {};
1243
	enum uclamp_id clamp_id;
1244 1245
	int cpu;

1246 1247
	mutex_init(&uclamp_mutex);

1248
	for_each_possible_cpu(cpu) {
1249 1250
		memset(&cpu_rq(cpu)->uclamp, 0,
				sizeof(struct uclamp_rq)*UCLAMP_CNT);
1251 1252
		cpu_rq(cpu)->uclamp_flags = 0;
	}
1253 1254

	for_each_clamp_id(clamp_id) {
1255
		uclamp_se_set(&init_task.uclamp_req[clamp_id],
1256
			      uclamp_none(clamp_id), false);
1257
	}
1258 1259

	/* System defaults allow max clamp values for both indexes */
1260
	uclamp_se_set(&uc_max, uclamp_none(UCLAMP_MAX), false);
1261
	for_each_clamp_id(clamp_id) {
1262
		uclamp_default[clamp_id] = uc_max;
1263 1264
#ifdef CONFIG_UCLAMP_TASK_GROUP
		root_task_group.uclamp_req[clamp_id] = uc_max;
1265
		root_task_group.uclamp[clamp_id] = uc_max;
1266 1267
#endif
	}
1268 1269 1270 1271 1272
}

#else /* CONFIG_UCLAMP_TASK */
static inline void uclamp_rq_inc(struct rq *rq, struct task_struct *p) { }
static inline void uclamp_rq_dec(struct rq *rq, struct task_struct *p) { }
1273 1274 1275 1276 1277 1278 1279
static inline int uclamp_validate(struct task_struct *p,
				  const struct sched_attr *attr)
{
	return -EOPNOTSUPP;
}
static void __setscheduler_uclamp(struct task_struct *p,
				  const struct sched_attr *attr) { }
1280
static inline void uclamp_fork(struct task_struct *p) { }
1281 1282 1283
static inline void init_uclamp(void) { }
#endif /* CONFIG_UCLAMP_TASK */

1284
static inline void enqueue_task(struct rq *rq, struct task_struct *p, int flags)
1285
{
1286 1287 1288
	if (!(flags & ENQUEUE_NOCLOCK))
		update_rq_clock(rq);

1289
	if (!(flags & ENQUEUE_RESTORE)) {
1290
		sched_info_queued(rq, p);
1291 1292
		psi_enqueue(p, flags & ENQUEUE_WAKEUP);
	}
1293

1294
	uclamp_rq_inc(rq, p);
1295
	p->sched_class->enqueue_task(rq, p, flags);
1296 1297
}

1298
static inline void dequeue_task(struct rq *rq, struct task_struct *p, int flags)
1299
{
1300 1301 1302
	if (!(flags & DEQUEUE_NOCLOCK))
		update_rq_clock(rq);

1303
	if (!(flags & DEQUEUE_SAVE)) {
1304
		sched_info_dequeued(rq, p);
1305 1306
		psi_dequeue(p, flags & DEQUEUE_SLEEP);
	}
1307

1308
	uclamp_rq_dec(rq, p);
1309
	p->sched_class->dequeue_task(rq, p, flags);
1310 1311
}

1312
void activate_task(struct rq *rq, struct task_struct *p, int flags)
1313 1314 1315 1316
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible--;

1317
	enqueue_task(rq, p, flags);
1318 1319

	p->on_rq = TASK_ON_RQ_QUEUED;
1320 1321
}

1322
void deactivate_task(struct rq *rq, struct task_struct *p, int flags)
1323
{
1324 1325
	p->on_rq = (flags & DEQUEUE_SLEEP) ? 0 : TASK_ON_RQ_MIGRATING;

1326 1327 1328
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible++;

1329
	dequeue_task(rq, p, flags);
1330 1331
}

1332
/*
Ingo Molnar's avatar
Ingo Molnar committed
1333
 * __normal_prio - return the priority that is based on the static prio
1334 1335 1336
 */
static inline int __normal_prio(struct task_struct *p)
{
Ingo Molnar's avatar
Ingo Molnar committed
1337
	return p->static_prio;
1338 1339
}

1340 1341 1342 1343 1344 1345 1346
/*
 * 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.
 */
1347
static inline int normal_prio(struct task_struct *p)
1348 1349 1350
{
	int prio;

1351 1352 1353
	if (task_has_dl_policy(p))
		prio = MAX_DL_PRIO-1;
	else if (task_has_rt_policy(p))
1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366
		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.
 */
1367
static int effective_prio(struct task_struct *p)
1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379
{
	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
1380 1381 1382
/**
 * task_curr - is this task currently executing on a CPU?
 * @p: the task in question.
1383 1384
 *
 * Return: 1 if the task is currently executing. 0 otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
1385
 */
1386
inline int task_curr(const struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
1387 1388 1389 1390
{
	return cpu_curr(task_cpu(p)) == p;
}

1391
/*
1392 1393 1394 1395 1396
 * 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().
1397
 */
1398 1399
static inline void check_class_changed(struct rq *rq, struct task_struct *p,
				       const struct sched_class *prev_class,
1400
				       int oldprio)
1401 1402 1403
{
	if (prev_class != p->sched_class) {
		if (prev_class->switched_from)
1404
			prev_class->switched_from(rq, p);
1405

1406
		p->sched_class->switched_to(rq, p);
1407
	} else if (oldprio != p->prio || dl_task(p))
1408
		p->sched_class->prio_changed(rq, p, oldprio);
1409 1410
}

1411
void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags)
1412 1413 1414 1415 1416 1417 1418 1419 1420 1421
{
	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) {
1422
				resched_curr(rq);
1423 1424 1425 1426 1427 1428 1429 1430 1431
				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.
	 */
1432
	if (task_on_rq_queued(rq->curr) && test_tsk_need_resched(rq->curr))
1433
		rq_clock_skip_update(rq);
1434 1435
}

Linus Torvalds's avatar
Linus Torvalds committed
1436
#ifdef CONFIG_SMP
1437 1438

/*
1439
 * Per-CPU kthreads are allowed to run on !active && online CPUs, see
1440 1441 1442 1443
 * __set_cpus_allowed_ptr() and select_fallback_rq().
 */
static inline bool is_cpu_allowed(struct task_struct *p, int cpu)
{
1444
	if (!cpumask_test_cpu(cpu, p->cpus_ptr))
1445 1446 1447 1448 1449 1450 1451 1452
		return false;

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

	return cpu_active(cpu);
}

Peter Zijlstra's avatar
Peter Zijlstra committed
1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471
/*
 * 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.
 */
1472 1473
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
1474 1475 1476
{
	lockdep_assert_held(&rq->lock);

1477
	WRITE_ONCE(p->on_rq, TASK_ON_RQ_MIGRATING);
1478
	dequeue_task(rq, p, DEQUEUE_NOCLOCK);
Peter Zijlstra's avatar
Peter Zijlstra committed
1479
	set_task_cpu(p, new_cpu);
1480
	rq_unlock(rq, rf);
Peter Zijlstra's avatar
Peter Zijlstra committed
1481 1482 1483

	rq = cpu_rq(new_cpu);

1484
	rq_lock(rq, rf);
Peter Zijlstra's avatar
Peter Zijlstra committed
1485 1486
	BUG_ON(task_cpu(p) != new_cpu);
	enqueue_task(rq, p, 0);
1487
	p->on_rq = TASK_ON_RQ_QUEUED;
Peter Zijlstra's avatar
Peter Zijlstra committed
1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498
	check_preempt_curr(rq, p, 0);

	return rq;
}

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

/*
1499
 * Move (not current) task off this CPU, onto the destination CPU. We're doing
Peter Zijlstra's avatar
Peter Zijlstra committed
1500 1501 1502 1503 1504 1505 1506
 * 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.
 */
1507 1508
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
1509 1510
{
	/* Affinity changed (again). */
1511
	if (!is_cpu_allowed(p, dest_cpu))
1512
		return rq;
Peter Zijlstra's avatar
Peter Zijlstra committed
1513

1514
	update_rq_clock(rq);
1515
	rq = move_queued_task(rq, rf, p, dest_cpu);
1516 1517

	return rq;
Peter Zijlstra's avatar
Peter Zijlstra committed
1518 1519 1520 1521 1522 1523 1524 1525 1526 1527
}

/*
 * 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;
1528 1529
	struct task_struct *p = arg->task;
	struct rq *rq = this_rq();
1530
	struct rq_flags rf;
Peter Zijlstra's avatar
Peter Zijlstra committed
1531 1532

	/*
1533 1534
	 * 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
1535 1536 1537 1538
	 */
	local_irq_disable();
	/*
	 * We need to explicitly wake pending tasks before running
1539
	 * __migrate_task() such that we will not miss enforcing cpus_ptr
Peter Zijlstra's avatar
Peter Zijlstra committed
1540 1541 1542
	 * during wakeups, see set_cpus_allowed_ptr()'s TASK_WAKING test.
	 */
	sched_ttwu_pending();
1543 1544

	raw_spin_lock(&p->pi_lock);
1545
	rq_lock(rq, &rf);
1546 1547 1548 1549 1550
	/*
	 * 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.
	 */
1551 1552
	if (task_rq(p) == rq) {
		if (task_on_rq_queued(p))
1553
			rq = __migrate_task(rq, &rf, p, arg->dest_cpu);
1554 1555 1556
		else
			p->wake_cpu = arg->dest_cpu;
	}
1557
	rq_unlock(rq, &rf);
1558 1559
	raw_spin_unlock(&p->pi_lock);

Peter Zijlstra's avatar
Peter Zijlstra committed
1560 1561 1562 1563
	local_irq_enable();
	return 0;
}

1564 1565 1566 1567 1568
/*
 * 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
1569
{
1570
	cpumask_copy(&p->cpus_mask, new_mask);
Peter Zijlstra's avatar
Peter Zijlstra committed
1571 1572 1573
	p->nr_cpus_allowed = cpumask_weight(new_mask);
}

1574 1575
void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
{
1576 1577 1578
	struct rq *rq = task_rq(p);
	bool queued, running;

1579
	lockdep_assert_held(&p->pi_lock);
1580 1581 1582 1583 1584 1585 1586 1587 1588 1589

	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);
1590
		dequeue_task(rq, p, DEQUEUE_SAVE | DEQUEUE_NOCLOCK);
1591 1592 1593 1594
	}
	if (running)
		put_prev_task(rq, p);

1595
	p->sched_class->set_cpus_allowed(p, new_mask);
1596 1597

	if (queued)
1598
		enqueue_task(rq, p, ENQUEUE_RESTORE | ENQUEUE_NOCLOCK);
1599
	if (running)
1600
		set_next_task(rq, p);
1601 1602
}

Peter Zijlstra's avatar
Peter Zijlstra committed
1603 1604 1605 1606 1607 1608 1609 1610 1611
/*
 * 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.
 */
1612 1613
static int __set_cpus_allowed_ptr(struct task_struct *p,
				  const struct cpumask *new_mask, bool check)
Peter Zijlstra's avatar
Peter Zijlstra committed
1614
{
1615
	const struct cpumask *cpu_valid_mask = cpu_active_mask;
Peter Zijlstra's avatar
Peter Zijlstra committed
1616
	unsigned int dest_cpu;
1617 1618
	struct rq_flags rf;
	struct rq *rq;
Peter Zijlstra's avatar
Peter Zijlstra committed
1619 1620
	int ret = 0;

1621
	rq = task_rq_lock(p, &rf);
1622
	update_rq_clock(rq);
Peter Zijlstra's avatar
Peter Zijlstra committed
1623

1624 1625 1626 1627 1628 1629 1630
	if (p->flags & PF_KTHREAD) {
		/*
		 * Kernel threads are allowed on online && !active CPUs
		 */
		cpu_valid_mask = cpu_online_mask;
	}

1631 1632 1633 1634 1635 1636 1637 1638 1639
	/*
	 * 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;
	}

1640
	if (cpumask_equal(p->cpus_ptr, new_mask))
Peter Zijlstra's avatar
Peter Zijlstra committed
1641 1642
		goto out;

1643 1644 1645 1646 1647 1648
	/*
	 * Picking a ~random cpu helps in cases where we are changing affinity
	 * for groups of tasks (ie. cpuset), so that load balancing is not
	 * immediately required to distribute the tasks within their new mask.
	 */
	dest_cpu = cpumask_any_and_distribute(cpu_valid_mask, new_mask);
1649
	if (dest_cpu >= nr_cpu_ids) {
Peter Zijlstra's avatar
Peter Zijlstra committed
1650 1651 1652 1653 1654 1655
		ret = -EINVAL;
		goto out;
	}

	do_set_cpus_allowed(p, new_mask);

1656 1657 1658
	if (p->flags & PF_KTHREAD) {
		/*
		 * For kernel threads that do indeed end up on online &&
1659
		 * !active we want to ensure they are strict per-CPU threads.
1660 1661 1662 1663 1664 1665
		 */
		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
1666 1667 1668 1669 1670 1671 1672
	/* 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;

	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. */
1673
		task_rq_unlock(rq, p, &rf);
Peter Zijlstra's avatar
Peter Zijlstra committed
1674 1675
		stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg);
		return 0;
1676 1677 1678 1679 1680
	} else if (task_on_rq_queued(p)) {
		/*
		 * OK, since we're going to drop the lock immediately
		 * afterwards anyway.
		 */
1681
		rq = move_queued_task(rq, &rf, p, dest_cpu);
1682
	}
Peter Zijlstra's avatar
Peter Zijlstra committed
1683
out:
1684
	task_rq_unlock(rq, p, &rf);
Peter Zijlstra's avatar
Peter Zijlstra committed
1685 1686 1687

	return ret;
}
1688 1689 1690 1691 1692

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
1693 1694
EXPORT_SYMBOL_GPL(set_cpus_allowed_ptr);

Ingo Molnar's avatar
Ingo Molnar committed
1695
void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
Ingo Molnar's avatar
Ingo Molnar committed
1696
{
1697 1698 1699 1700 1701
#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
1702
	WARN_ON_ONCE(p->state != TASK_RUNNING && p->state != TASK_WAKING &&
1703
			!p->on_rq);
1704

1705 1706 1707 1708 1709 1710 1711 1712 1713
	/*
	 * 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)));

1714
#ifdef CONFIG_LOCKDEP
1715 1716 1717 1718 1719
	/*
	 * 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,
1720
	 * see task_group().
1721 1722 1723 1724
	 *
	 * Furthermore, all task_rq users should acquire both locks, see
	 * task_rq_lock().
	 */
1725 1726 1727
	WARN_ON_ONCE(debug_locks && !(lockdep_is_held(&p->pi_lock) ||
				      lockdep_is_held(&task_rq(p)->lock)));
#endif
1728 1729 1730 1731
	/*
	 * Clearly, migrating tasks to offline CPUs is a fairly daft thing.
	 */
	WARN_ON_ONCE(!cpu_online(new_cpu));
1732 1733
#endif

1734
	trace_sched_migrate_task(p, new_cpu);
1735

1736
	if (task_cpu(p) != new_cpu) {
1737
		if (p->sched_class->migrate_task_rq)
1738
			p->sched_class->migrate_task_rq(p, new_cpu);
1739
		p->se.nr_migrations++;
1740
		rseq_migrate(p);
1741
		perf_event_task_migrate(p);
1742
	}
Ingo Molnar's avatar
Ingo Molnar committed
1743 1744

	__set_task_cpu(p, new_cpu);
Ingo Molnar's avatar
Ingo Molnar committed
1745 1746
}

1747
#ifdef CONFIG_NUMA_BALANCING
1748 1749
static void __migrate_swap_task(struct task_struct *p, int cpu)
{
1750
	if (task_on_rq_queued(p)) {
1751
		struct rq *src_rq, *dst_rq;
1752
		struct rq_flags srf, drf;
1753 1754 1755 1756

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

1757 1758 1759
		rq_pin_lock(src_rq, &srf);
		rq_pin_lock(dst_rq, &drf);

1760 1761 1762 1763
		deactivate_task(src_rq, p, 0);
		set_task_cpu(p, cpu);
		activate_task(dst_rq, p, 0);
		check_preempt_curr(dst_rq, p, 0);
1764 1765 1766 1767

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

1768 1769 1770 1771
	} 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
1772
		 * previous CPU our target instead of where it really is.
1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788
		 */
		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;

1789 1790 1791
	if (!cpu_active(arg->src_cpu) || !cpu_active(arg->dst_cpu))
		return -EAGAIN;

1792 1793 1794
	src_rq = cpu_rq(arg->src_cpu);
	dst_rq = cpu_rq(arg->dst_cpu);

1795 1796
	double_raw_lock(&arg->src_task->pi_lock,
			&arg->dst_task->pi_lock);
1797
	double_rq_lock(src_rq, dst_rq);
1798

1799 1800 1801 1802 1803 1804
	if (task_cpu(arg->dst_task) != arg->dst_cpu)
		goto unlock;

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

1805
	if (!cpumask_test_cpu(arg->dst_cpu, arg->src_task->cpus_ptr))
1806 1807
		goto unlock;

1808
	if (!cpumask_test_cpu(arg->src_cpu, arg->dst_task->cpus_ptr))
1809 1810 1811 1812 1813 1814 1815 1816 1817
		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);
1818 1819
	raw_spin_unlock(&arg->dst_task->pi_lock);
	raw_spin_unlock(&arg->src_task->pi_lock);
1820 1821 1822 1823 1824 1825 1826

	return ret;
}

/*
 * Cross migrate two tasks
 */
1827 1828
int migrate_swap(struct task_struct *cur, struct task_struct *p,
		int target_cpu, int curr_cpu)
1829 1830 1831 1832 1833 1834
{
	struct migration_swap_arg arg;
	int ret = -EINVAL;

	arg = (struct migration_swap_arg){
		.src_task = cur,
1835
		.src_cpu = curr_cpu,
1836
		.dst_task = p,
1837
		.dst_cpu = target_cpu,
1838 1839 1840 1841 1842
	};

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

1843 1844 1845 1846
	/*
	 * 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.
	 */
1847 1848 1849
	if (!cpu_active(arg.src_cpu) || !cpu_active(arg.dst_cpu))
		goto out;

1850
	if (!cpumask_test_cpu(arg.dst_cpu, arg.src_task->cpus_ptr))
1851 1852
		goto out;

1853
	if (!cpumask_test_cpu(arg.src_cpu, arg.dst_task->cpus_ptr))
1854 1855
		goto out;

1856
	trace_sched_swap_numa(cur, arg.src_cpu, p, arg.dst_cpu);
1857 1858 1859 1860 1861
	ret = stop_two_cpus(arg.dst_cpu, arg.src_cpu, migrate_swap_stop, &arg);

out:
	return ret;
}
1862
#endif /* CONFIG_NUMA_BALANCING */
1863

Linus Torvalds's avatar
Linus Torvalds committed
1864 1865 1866
/*
 * wait_task_inactive - wait for a thread to unschedule.
 *
1867 1868 1869 1870 1871 1872 1873
 * 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
1874 1875 1876 1877 1878 1879
 * 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.
 */
1880
unsigned long wait_task_inactive(struct task_struct *p, long match_state)
Linus Torvalds's avatar
Linus Torvalds committed
1881
{
1882
	int running, queued;
1883
	struct rq_flags rf;
1884
	unsigned long ncsw;
1885
	struct rq *rq;
Linus Torvalds's avatar
Linus Torvalds committed
1886

1887 1888 1889 1890 1891 1892 1893 1894
	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);
1895

1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906
		/*
		 * 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!
		 */
1907 1908 1909
		while (task_running(rq, p)) {
			if (match_state && unlikely(p->state != match_state))
				return 0;
1910
			cpu_relax();
1911
		}
1912

1913 1914 1915 1916 1917
		/*
		 * 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.
		 */
1918
		rq = task_rq_lock(p, &rf);
1919
		trace_sched_wait_task(p);
1920
		running = task_running(rq, p);
1921
		queued = task_on_rq_queued(p);
1922
		ncsw = 0;
1923
		if (!match_state || p->state == match_state)
1924
			ncsw = p->nvcsw | LONG_MIN; /* sets MSB */
1925
		task_rq_unlock(rq, p, &rf);
1926

1927 1928 1929 1930 1931 1932
		/*
		 * If it changed from the expected state, bail out now.
		 */
		if (unlikely(!ncsw))
			break;

1933 1934 1935 1936 1937 1938 1939 1940 1941 1942
		/*
		 * 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;
		}
1943

1944 1945 1946 1947 1948
		/*
		 * It's not enough that it's not actively running,
		 * it must be off the runqueue _entirely_, and not
		 * preempted!
		 *
1949
		 * So if it was still runnable (but just not actively
1950 1951 1952
		 * running right now), it's preempted, and we should
		 * yield - it could be a while.
		 */
1953
		if (unlikely(queued)) {
1954
			ktime_t to = NSEC_PER_SEC / HZ;
1955 1956 1957

			set_current_state(TASK_UNINTERRUPTIBLE);
			schedule_hrtimeout(&to, HRTIMER_MODE_REL);
1958 1959
			continue;
		}
1960

1961 1962 1963 1964 1965 1966 1967
		/*
		 * 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;
	}
1968 1969

	return ncsw;
Linus Torvalds's avatar
Linus Torvalds committed
1970 1971 1972 1973 1974 1975 1976 1977 1978
}

/***
 * 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
1979
 * NOTE: this function doesn't have to take the runqueue lock,
Linus Torvalds's avatar
Linus Torvalds committed
1980 1981 1982 1983 1984
 * 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.
 */
1985
void kick_process(struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
1986 1987 1988 1989 1990 1991 1992 1993 1994
{
	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
1995
EXPORT_SYMBOL_GPL(kick_process);
Linus Torvalds's avatar
Linus Torvalds committed
1996

1997
/*
1998
 * ->cpus_ptr is protected by both rq->lock and p->pi_lock
1999 2000 2001 2002 2003
 *
 * A few notes on cpu_active vs cpu_online:
 *
 *  - cpu_active must be a subset of cpu_online
 *
2004
 *  - on CPU-up we allow per-CPU kthreads on the online && !active CPU,
2005
 *    see __set_cpus_allowed_ptr(). At this point the newly online
2006
 *    CPU isn't yet part of the sched domains, and balancing will not
2007 2008
 *    see it.
 *
2009
 *  - on CPU-down we clear cpu_active() to mask the sched domains and
2010
 *    avoid the load balancer to place new tasks on the to be removed
2011
 *    CPU. Existing tasks will remain running there and will be taken
2012 2013 2014 2015 2016 2017
 *    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.
2018
 */
2019 2020
static int select_fallback_rq(int cpu, struct task_struct *p)
{
2021 2022
	int nid = cpu_to_node(cpu);
	const struct cpumask *nodemask = NULL;
2023 2024
	enum { cpuset, possible, fail } state = cpuset;
	int dest_cpu;
2025

2026
	/*
2027 2028 2029
	 * 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.
2030 2031 2032 2033 2034 2035 2036 2037
	 */
	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;
2038
			if (cpumask_test_cpu(dest_cpu, p->cpus_ptr))
2039 2040
				return dest_cpu;
		}
2041
	}
2042

2043 2044
	for (;;) {
		/* Any allowed, online CPU? */
2045
		for_each_cpu(dest_cpu, p->cpus_ptr) {
2046
			if (!is_cpu_allowed(p, dest_cpu))
2047
				continue;
2048

2049 2050
			goto out;
		}
2051

2052
		/* No more Mr. Nice Guy. */
2053 2054
		switch (state) {
		case cpuset:
2055 2056 2057 2058 2059
			if (IS_ENABLED(CONFIG_CPUSETS)) {
				cpuset_cpus_allowed_fallback(p);
				state = possible;
				break;
			}
2060
			/* Fall-through */
2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079
		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()) {
2080
			printk_deferred("process %d (%s) no longer affine to cpu%d\n",
2081 2082
					task_pid_nr(p), p->comm, cpu);
		}
2083 2084 2085 2086 2087
	}

	return dest_cpu;
}

2088
/*
2089
 * The caller (fork, wakeup) owns p->pi_lock, ->cpus_ptr is stable.
2090
 */
2091
static inline
2092
int select_task_rq(struct task_struct *p, int cpu, int sd_flags, int wake_flags)
2093
{
2094 2095
	lockdep_assert_held(&p->pi_lock);

2096
	if (p->nr_cpus_allowed > 1)
2097
		cpu = p->sched_class->select_task_rq(p, cpu, sd_flags, wake_flags);
2098
	else
2099
		cpu = cpumask_any(p->cpus_ptr);
2100 2101 2102

	/*
	 * In order not to call set_task_cpu() on a blocking task we need
2103
	 * to rely on ttwu() to place the task on a valid ->cpus_ptr
2104
	 * CPU.
2105 2106 2107 2108 2109 2110
	 *
	 * 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 ]
	 */
2111
	if (unlikely(!is_cpu_allowed(p, cpu)))
2112
		cpu = select_fallback_rq(task_cpu(p), p);
2113 2114

	return cpu;
2115
}
2116

2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146
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;
	}
}

2147 2148 2149 2150 2151 2152 2153 2154
#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
2155
#endif /* CONFIG_SMP */
2156

2157
static void
2158
ttwu_stat(struct task_struct *p, int cpu, int wake_flags)
Tejun Heo's avatar
Tejun Heo committed
2159
{
2160
	struct rq *rq;
2161

2162 2163 2164 2165
	if (!schedstat_enabled())
		return;

	rq = this_rq();
2166

2167 2168
#ifdef CONFIG_SMP
	if (cpu == rq->cpu) {
2169 2170
		__schedstat_inc(rq->ttwu_local);
		__schedstat_inc(p->se.statistics.nr_wakeups_local);
2171 2172 2173
	} else {
		struct sched_domain *sd;

2174
		__schedstat_inc(p->se.statistics.nr_wakeups_remote);
2175
		rcu_read_lock();
2176
		for_each_domain(rq->cpu, sd) {
2177
			if (cpumask_test_cpu(cpu, sched_domain_span(sd))) {
2178
				__schedstat_inc(sd->ttwu_wake_remote);
2179 2180 2181
				break;
			}
		}
2182
		rcu_read_unlock();
2183
	}
2184 2185

	if (wake_flags & WF_MIGRATED)
2186
		__schedstat_inc(p->se.statistics.nr_wakeups_migrate);
2187 2188
#endif /* CONFIG_SMP */

2189 2190
	__schedstat_inc(rq->ttwu_count);
	__schedstat_inc(p->se.statistics.nr_wakeups);
2191 2192

	if (wake_flags & WF_SYNC)
2193
		__schedstat_inc(p->se.statistics.nr_wakeups_sync);
2194 2195
}

2196 2197 2198
/*
 * Mark the task runnable and perform wakeup-preemption.
 */
2199
static void ttwu_do_wakeup(struct rq *rq, struct task_struct *p, int wake_flags,
2200
			   struct rq_flags *rf)
Tejun Heo's avatar
Tejun Heo committed
2201 2202 2203
{
	check_preempt_curr(rq, p, wake_flags);
	p->state = TASK_RUNNING;
2204 2205
	trace_sched_wakeup(p);

Tejun Heo's avatar
Tejun Heo committed
2206
#ifdef CONFIG_SMP
2207 2208
	if (p->sched_class->task_woken) {
		/*
2209 2210
		 * Our task @p is fully woken up and running; so its safe to
		 * drop the rq->lock, hereafter rq is only used for statistics.
2211
		 */
2212
		rq_unpin_lock(rq, rf);
Tejun Heo's avatar
Tejun Heo committed
2213
		p->sched_class->task_woken(rq, p);
2214
		rq_repin_lock(rq, rf);
2215
	}
Tejun Heo's avatar
Tejun Heo committed
2216

2217
	if (rq->idle_stamp) {
2218
		u64 delta = rq_clock(rq) - rq->idle_stamp;
2219
		u64 max = 2*rq->max_idle_balance_cost;
Tejun Heo's avatar
Tejun Heo committed
2220

2221 2222 2223
		update_avg(&rq->avg_idle, delta);

		if (rq->avg_idle > max)
Tejun Heo's avatar
Tejun Heo committed
2224
			rq->avg_idle = max;
2225

Tejun Heo's avatar
Tejun Heo committed
2226 2227 2228 2229 2230
		rq->idle_stamp = 0;
	}
#endif
}

2231
static void
2232
ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags,
2233
		 struct rq_flags *rf)
2234
{
2235
	int en_flags = ENQUEUE_WAKEUP | ENQUEUE_NOCLOCK;
2236

2237 2238
	lockdep_assert_held(&rq->lock);

2239 2240 2241
#ifdef CONFIG_SMP
	if (p->sched_contributes_to_load)
		rq->nr_uninterruptible--;
2242 2243

	if (wake_flags & WF_MIGRATED)
2244
		en_flags |= ENQUEUE_MIGRATED;
2245 2246
#endif

2247
	activate_task(rq, p, en_flags);
2248
	ttwu_do_wakeup(rq, p, wake_flags, rf);
2249 2250 2251 2252 2253 2254 2255 2256 2257 2258
}

/*
 * 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)
{
2259
	struct rq_flags rf;
2260 2261 2262
	struct rq *rq;
	int ret = 0;

2263
	rq = __task_rq_lock(p, &rf);
2264
	if (task_on_rq_queued(p)) {
2265 2266
		/* check_preempt_curr() may use rq clock */
		update_rq_clock(rq);
2267
		ttwu_do_wakeup(rq, p, wake_flags, &rf);
2268 2269
		ret = 1;
	}
2270
	__task_rq_unlock(rq, &rf);
2271 2272 2273 2274

	return ret;
}

2275
#ifdef CONFIG_SMP
2276
void sched_ttwu_pending(void)
2277 2278
{
	struct rq *rq = this_rq();
2279
	struct llist_node *llist = llist_del_all(&rq->wake_list);
2280
	struct task_struct *p, *t;
2281
	struct rq_flags rf;
2282

2283 2284 2285
	if (!llist)
		return;

2286
	rq_lock_irqsave(rq, &rf);
2287
	update_rq_clock(rq);
2288

2289 2290
	llist_for_each_entry_safe(p, t, llist, wake_entry)
		ttwu_do_activate(rq, p, p->sched_remote_wakeup ? WF_MIGRATED : 0, &rf);
2291

2292
	rq_unlock_irqrestore(rq, &rf);
2293 2294 2295 2296
}

void scheduler_ipi(void)
{
2297 2298 2299 2300 2301
	/*
	 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
	 * TIF_NEED_RESCHED remotely (for the first time) will also send
	 * this IPI.
	 */
2302
	preempt_fold_need_resched();
2303

2304
	if (llist_empty(&this_rq()->wake_list) && !got_nohz_idle_kick())
2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320
		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();
2321
	sched_ttwu_pending();
2322 2323 2324 2325

	/*
	 * Check if someone kicked us for doing the nohz idle load balance.
	 */
2326
	if (unlikely(got_nohz_idle_kick())) {
2327
		this_rq()->idle_balance = 1;
2328
		raise_softirq_irqoff(SCHED_SOFTIRQ);
2329
	}
2330
	irq_exit();
2331 2332
}

2333
static void ttwu_queue_remote(struct task_struct *p, int cpu, int wake_flags)
2334
{
2335 2336
	struct rq *rq = cpu_rq(cpu);

2337 2338
	p->sched_remote_wakeup = !!(wake_flags & WF_MIGRATED);

2339 2340 2341 2342 2343 2344
	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);
	}
2345
}
2346

2347 2348 2349
void wake_up_if_idle(int cpu)
{
	struct rq *rq = cpu_rq(cpu);
2350
	struct rq_flags rf;
2351

2352 2353 2354 2355
	rcu_read_lock();

	if (!is_idle_task(rcu_dereference(rq->curr)))
		goto out;
2356 2357 2358 2359

	if (set_nr_if_polling(rq->idle)) {
		trace_sched_wake_idle_without_ipi(cpu);
	} else {
2360
		rq_lock_irqsave(rq, &rf);
2361 2362
		if (is_idle_task(rq->curr))
			smp_send_reschedule(cpu);
2363
		/* Else CPU is not idle, do nothing here: */
2364
		rq_unlock_irqrestore(rq, &rf);
2365
	}
2366 2367 2368

out:
	rcu_read_unlock();
2369 2370
}

2371
bool cpus_share_cache(int this_cpu, int that_cpu)
2372 2373 2374
{
	return per_cpu(sd_llc_id, this_cpu) == per_cpu(sd_llc_id, that_cpu);
}
2375
#endif /* CONFIG_SMP */
2376

2377
static void ttwu_queue(struct task_struct *p, int cpu, int wake_flags)
2378 2379
{
	struct rq *rq = cpu_rq(cpu);
2380
	struct rq_flags rf;
2381

2382
#if defined(CONFIG_SMP)
2383
	if (sched_feat(TTWU_QUEUE) && !cpus_share_cache(smp_processor_id(), cpu)) {
2384
		sched_clock_cpu(cpu); /* Sync clocks across CPUs */
2385
		ttwu_queue_remote(p, cpu, wake_flags);
2386 2387 2388 2389
		return;
	}
#endif

2390
	rq_lock(rq, &rf);
2391
	update_rq_clock(rq);
2392
	ttwu_do_activate(rq, p, wake_flags, &rf);
2393
	rq_unlock(rq, &rf);
Tejun Heo's avatar
Tejun Heo committed
2394 2395
}

2396 2397 2398 2399 2400 2401
/*
 * Notes on Program-Order guarantees on SMP systems.
 *
 *  MIGRATION
 *
 * The basic program-order guarantee on SMP systems is that when a task [t]
2402 2403
 * migrates, all its activity on its old CPU [c0] happens-before any subsequent
 * execution on its new CPU [c1].
2404 2405 2406 2407 2408 2409 2410 2411
 *
 * 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
 *
2412
 * Release/acquire chaining guarantees that B happens after A and C after B.
2413
 * Note: the CPU doing B need not be c0 or c1
2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444
 *
 * 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)
2445
 *   2) smp_cond_load_acquire(!X->on_cpu)
2446 2447 2448 2449 2450 2451 2452 2453 2454 2455
 *
 * 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);
 *
2456
 *                    smp_cond_load_acquire(&X->on_cpu, !VAL);
2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473
 *                    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
 *
 *
2474 2475 2476
 * 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().
2477 2478
 */

Tejun Heo's avatar
Tejun Heo committed
2479
/**
Linus Torvalds's avatar
Linus Torvalds committed
2480
 * try_to_wake_up - wake up a thread
Tejun Heo's avatar
Tejun Heo committed
2481
 * @p: the thread to be awakened
Linus Torvalds's avatar
Linus Torvalds committed
2482
 * @state: the mask of task states that can be woken
Tejun Heo's avatar
Tejun Heo committed
2483
 * @wake_flags: wake modifier flags (WF_*)
Linus Torvalds's avatar
Linus Torvalds committed
2484
 *
2485
 * If (@state & @p->state) @p->state = TASK_RUNNING.
Linus Torvalds's avatar
Linus Torvalds committed
2486
 *
2487 2488 2489 2490 2491
 * 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().
 *
2492 2493 2494
 * This function executes a full memory barrier before accessing the task
 * state; see set_current_state().
 *
2495 2496
 * Return: %true if @p->state changes (an actual wakeup was done),
 *	   %false otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
2497
 */
2498 2499
static int
try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
Linus Torvalds's avatar
Linus Torvalds committed
2500 2501
{
	unsigned long flags;
2502
	int cpu, success = 0;
2503

2504
	preempt_disable();
2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517
	if (p == current) {
		/*
		 * We're waking current, this means 'p->on_rq' and 'task_cpu(p)
		 * == smp_processor_id()'. Together this means we can special
		 * case the whole 'p->on_rq && ttwu_remote()' case below
		 * without taking any locks.
		 *
		 * In particular:
		 *  - we rely on Program-Order guarantees for all the ordering,
		 *  - we're serialized against set_special_state() by virtue of
		 *    it disabling IRQs (this allows not taking ->pi_lock).
		 */
		if (!(p->state & state))
2518
			goto out;
2519 2520 2521 2522 2523 2524 2525 2526 2527

		success = 1;
		cpu = task_cpu(p);
		trace_sched_waking(p);
		p->state = TASK_RUNNING;
		trace_sched_wakeup(p);
		goto out;
	}

2528 2529 2530 2531 2532 2533
	/*
	 * 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.
	 */
2534
	raw_spin_lock_irqsave(&p->pi_lock, flags);
2535
	smp_mb__after_spinlock();
Peter Zijlstra's avatar
Peter Zijlstra committed
2536
	if (!(p->state & state))
2537
		goto unlock;
Linus Torvalds's avatar
Linus Torvalds committed
2538

2539 2540
	trace_sched_waking(p);

2541 2542
	/* We're going to change ->state: */
	success = 1;
Linus Torvalds's avatar
Linus Torvalds committed
2543 2544
	cpu = task_cpu(p);

2545 2546 2547 2548 2549
	/*
	 * 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.
	 *
2550 2551 2552 2553 2554 2555 2556 2557
	 * 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
2558 2559
	 *
	 * [task p]
2560
	 *   STORE p->state = UNINTERRUPTIBLE	  LOAD p->on_rq
2561
	 *
2562 2563
	 * Pairs with the LOCK+smp_mb__after_spinlock() on rq->lock in
	 * __schedule().  See the comment for smp_mb__after_spinlock().
2564 2565
	 */
	smp_rmb();
2566
	if (p->on_rq && ttwu_remote(p, wake_flags))
2567
		goto unlock;
Linus Torvalds's avatar
Linus Torvalds committed
2568 2569

#ifdef CONFIG_SMP
2570 2571 2572 2573 2574 2575 2576
	/*
	 * 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.
	 *
2577 2578 2579 2580 2581 2582 2583 2584
	 * __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
2585
	 *
2586 2587
	 * Pairs with the LOCK+smp_mb__after_spinlock() on rq->lock in
	 * __schedule().  See the comment for smp_mb__after_spinlock().
2588 2589 2590
	 */
	smp_rmb();

Peter Zijlstra's avatar
Peter Zijlstra committed
2591
	/*
2592
	 * If the owning (remote) CPU is still in the middle of schedule() with
2593
	 * this task as prev, wait until its done referencing the task.
2594
	 *
2595
	 * Pairs with the smp_store_release() in finish_task().
2596 2597 2598
	 *
	 * This ensures that tasks getting woken will be fully ordered against
	 * their previous state and preserve Program Order.
2599
	 */
2600
	smp_cond_load_acquire(&p->on_cpu, !VAL);
Linus Torvalds's avatar
Linus Torvalds committed
2601

2602
	p->sched_contributes_to_load = !!task_contributes_to_load(p);
Peter Zijlstra's avatar
Peter Zijlstra committed
2603
	p->state = TASK_WAKING;
2604

2605
	if (p->in_iowait) {
2606
		delayacct_blkio_end(p);
2607 2608 2609
		atomic_dec(&task_rq(p)->nr_iowait);
	}

2610
	cpu = select_task_rq(p, p->wake_cpu, SD_BALANCE_WAKE, wake_flags);
2611 2612
	if (task_cpu(p) != cpu) {
		wake_flags |= WF_MIGRATED;
2613
		psi_ttwu_dequeue(p);
2614
		set_task_cpu(p, cpu);
2615
	}
2616 2617 2618 2619

#else /* CONFIG_SMP */

	if (p->in_iowait) {
2620
		delayacct_blkio_end(p);
2621 2622 2623
		atomic_dec(&task_rq(p)->nr_iowait);
	}

Linus Torvalds's avatar
Linus Torvalds committed
2624 2625
#endif /* CONFIG_SMP */

2626
	ttwu_queue(p, cpu, wake_flags);
2627
unlock:
2628
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
2629 2630 2631
out:
	if (success)
		ttwu_stat(p, cpu, wake_flags);
2632
	preempt_enable();
Linus Torvalds's avatar
Linus Torvalds committed
2633 2634 2635 2636

	return success;
}

2637 2638 2639 2640 2641
/**
 * 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
2642 2643 2644
 * processes.
 *
 * Return: 1 if the process was woken up, 0 if it was already running.
2645
 *
2646
 * This function executes a full memory barrier before accessing the task state.
2647
 */
2648
int wake_up_process(struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
2649
{
2650
	return try_to_wake_up(p, TASK_NORMAL, 0);
Linus Torvalds's avatar
Linus Torvalds committed
2651 2652 2653
}
EXPORT_SYMBOL(wake_up_process);

2654
int wake_up_state(struct task_struct *p, unsigned int state)
Linus Torvalds's avatar
Linus Torvalds committed
2655 2656 2657 2658 2659 2660 2661
{
	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
2662 2663 2664
 *
 * __sched_fork() is basic setup used by init_idle() too:
 */
2665
static void __sched_fork(unsigned long clone_flags, struct task_struct *p)
Ingo Molnar's avatar
Ingo Molnar committed
2666
{
Peter Zijlstra's avatar
Peter Zijlstra committed
2667 2668 2669
	p->on_rq			= 0;

	p->se.on_rq			= 0;
Ingo Molnar's avatar
Ingo Molnar committed
2670 2671
	p->se.exec_start		= 0;
	p->se.sum_exec_runtime		= 0;
2672
	p->se.prev_sum_exec_runtime	= 0;
2673
	p->se.nr_migrations		= 0;
2674
	p->se.vruntime			= 0;
Peter Zijlstra's avatar
Peter Zijlstra committed
2675
	INIT_LIST_HEAD(&p->se.group_node);
2676

2677 2678 2679 2680
#ifdef CONFIG_FAIR_GROUP_SCHED
	p->se.cfs_rq			= NULL;
#endif

2681
#ifdef CONFIG_SCHEDSTATS
2682
	/* Even if schedstat is disabled, there should not be garbage */
2683
	memset(&p->se.statistics, 0, sizeof(p->se.statistics));
2684
#endif
2685

2686
	RB_CLEAR_NODE(&p->dl.rb_node);
2687
	init_dl_task_timer(&p->dl);
2688
	init_dl_inactive_task_timer(&p->dl);
2689
	__dl_clear_params(p);
2690

Peter Zijlstra's avatar
Peter Zijlstra committed
2691
	INIT_LIST_HEAD(&p->rt.run_list);
2692 2693 2694 2695
	p->rt.timeout		= 0;
	p->rt.time_slice	= sched_rr_timeslice;
	p->rt.on_rq		= 0;
	p->rt.on_list		= 0;
2696

2697 2698 2699
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&p->preempt_notifiers);
#endif
2700

2701 2702 2703
#ifdef CONFIG_COMPACTION
	p->capture_control = NULL;
#endif
2704
	init_numa_balancing(clone_flags, p);
Ingo Molnar's avatar
Ingo Molnar committed
2705 2706
}

2707 2708
DEFINE_STATIC_KEY_FALSE(sched_numa_balancing);

2709
#ifdef CONFIG_NUMA_BALANCING
2710

2711 2712 2713
void set_numabalancing_state(bool enabled)
{
	if (enabled)
2714
		static_branch_enable(&sched_numa_balancing);
2715
	else
2716
		static_branch_disable(&sched_numa_balancing);
2717
}
2718 2719 2720 2721 2722 2723 2724

#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;
2725
	int state = static_branch_likely(&sched_numa_balancing);
2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740

	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
2741

2742 2743
#ifdef CONFIG_SCHEDSTATS

2744
DEFINE_STATIC_KEY_FALSE(sched_schedstats);
2745
static bool __initdata __sched_schedstats = false;
2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768

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;

2769 2770 2771 2772 2773
	/*
	 * 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.
	 */
2774
	if (!strcmp(str, "enable")) {
2775
		__sched_schedstats = true;
2776 2777
		ret = 1;
	} else if (!strcmp(str, "disable")) {
2778
		__sched_schedstats = false;
2779 2780 2781 2782 2783 2784 2785 2786 2787 2788
		ret = 1;
	}
out:
	if (!ret)
		pr_warn("Unable to parse schedstats=\n");

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

2789 2790 2791 2792 2793
static void __init init_schedstats(void)
{
	set_schedstats(__sched_schedstats);
}

2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813
#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;
}
2814 2815 2816 2817
#endif /* CONFIG_PROC_SYSCTL */
#else  /* !CONFIG_SCHEDSTATS */
static inline void init_schedstats(void) {}
#endif /* CONFIG_SCHEDSTATS */
Ingo Molnar's avatar
Ingo Molnar committed
2818 2819 2820 2821

/*
 * fork()/clone()-time setup:
 */
2822
int sched_fork(unsigned long clone_flags, struct task_struct *p)
Ingo Molnar's avatar
Ingo Molnar committed
2823
{
2824
	unsigned long flags;
Ingo Molnar's avatar
Ingo Molnar committed
2825

2826
	__sched_fork(clone_flags, p);
2827
	/*
2828
	 * We mark the process as NEW here. This guarantees that
2829 2830 2831
	 * nobody will actually run it, and a signal or other external
	 * event cannot wake it up and insert it on the runqueue either.
	 */
2832
	p->state = TASK_NEW;
Ingo Molnar's avatar
Ingo Molnar committed
2833

2834 2835 2836 2837 2838
	/*
	 * Make sure we do not leak PI boosting priority to the child.
	 */
	p->prio = current->normal_prio;

2839 2840
	uclamp_fork(p);

2841 2842 2843 2844
	/*
	 * Revert to default priority/policy on fork if requested.
	 */
	if (unlikely(p->sched_reset_on_fork)) {
2845
		if (task_has_dl_policy(p) || task_has_rt_policy(p)) {
2846
			p->policy = SCHED_NORMAL;
2847
			p->static_prio = NICE_TO_PRIO(0);
2848 2849 2850 2851 2852
			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);
2853
		set_load_weight(p, false);
2854

2855 2856 2857 2858 2859 2860
		/*
		 * We don't need the reset flag anymore after the fork. It has
		 * fulfilled its duty:
		 */
		p->sched_reset_on_fork = 0;
	}
2861

2862
	if (dl_prio(p->prio))
2863
		return -EAGAIN;
2864
	else if (rt_prio(p->prio))
2865
		p->sched_class = &rt_sched_class;
2866
	else
2867
		p->sched_class = &fair_sched_class;
2868

2869
	init_entity_runnable_average(&p->se);
2870

2871 2872 2873 2874 2875 2876 2877
	/*
	 * 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.
	 */
2878
	raw_spin_lock_irqsave(&p->pi_lock, flags);
2879
	/*
2880
	 * We're setting the CPU for the first time, we don't migrate,
2881 2882
	 * so use __set_task_cpu().
	 */
2883
	__set_task_cpu(p, smp_processor_id());
2884 2885
	if (p->sched_class->task_fork)
		p->sched_class->task_fork(p);
2886
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
2887

2888
#ifdef CONFIG_SCHED_INFO
Ingo Molnar's avatar
Ingo Molnar committed
2889
	if (likely(sched_info_on()))
2890
		memset(&p->sched_info, 0, sizeof(p->sched_info));
Linus Torvalds's avatar
Linus Torvalds committed
2891
#endif
2892 2893
#if defined(CONFIG_SMP)
	p->on_cpu = 0;
2894
#endif
2895
	init_task_preempt_count(p);
2896
#ifdef CONFIG_SMP
2897
	plist_node_init(&p->pushable_tasks, MAX_PRIO);
2898
	RB_CLEAR_NODE(&p->pushable_dl_tasks);
2899
#endif
2900
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
2901 2902
}

2903 2904 2905
unsigned long to_ratio(u64 period, u64 runtime)
{
	if (runtime == RUNTIME_INF)
2906
		return BW_UNIT;
2907 2908 2909 2910 2911 2912 2913 2914 2915

	/*
	 * 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;

2916
	return div64_u64(runtime << BW_SHIFT, period);
2917 2918
}

Linus Torvalds's avatar
Linus Torvalds committed
2919 2920 2921 2922 2923 2924 2925
/*
 * 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.
 */
2926
void wake_up_new_task(struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
2927
{
2928
	struct rq_flags rf;
Ingo Molnar's avatar
Ingo Molnar committed
2929
	struct rq *rq;
2930

2931
	raw_spin_lock_irqsave(&p->pi_lock, rf.flags);
2932
	p->state = TASK_RUNNING;
2933 2934 2935
#ifdef CONFIG_SMP
	/*
	 * Fork balancing, do it here and not earlier because:
2936
	 *  - cpus_ptr can change in the fork path
2937
	 *  - any previously selected CPU might disappear through hotplug
2938 2939 2940
	 *
	 * Use __set_task_cpu() to avoid calling sched_class::migrate_task_rq,
	 * as we're not fully set-up yet.
2941
	 */
2942
	p->recent_used_cpu = task_cpu(p);
2943
	__set_task_cpu(p, select_task_rq(p, task_cpu(p), SD_BALANCE_FORK, 0));
2944
#endif
2945
	rq = __task_rq_lock(p, &rf);
2946
	update_rq_clock(rq);
2947
	post_init_entity_util_avg(p);
2948

2949
	activate_task(rq, p, ENQUEUE_NOCLOCK);
2950
	trace_sched_wakeup_new(p);
Peter Zijlstra's avatar
Peter Zijlstra committed
2951
	check_preempt_curr(rq, p, WF_FORK);
2952
#ifdef CONFIG_SMP
2953 2954 2955 2956 2957
	if (p->sched_class->task_woken) {
		/*
		 * Nothing relies on rq->lock after this, so its fine to
		 * drop it.
		 */
2958
		rq_unpin_lock(rq, &rf);
2959
		p->sched_class->task_woken(rq, p);
2960
		rq_repin_lock(rq, &rf);
2961
	}
2962
#endif
2963
	task_rq_unlock(rq, p, &rf);
Linus Torvalds's avatar
Linus Torvalds committed
2964 2965
}

2966 2967
#ifdef CONFIG_PREEMPT_NOTIFIERS

2968
static DEFINE_STATIC_KEY_FALSE(preempt_notifier_key);
2969

2970 2971
void preempt_notifier_inc(void)
{
2972
	static_branch_inc(&preempt_notifier_key);
2973 2974 2975 2976 2977
}
EXPORT_SYMBOL_GPL(preempt_notifier_inc);

void preempt_notifier_dec(void)
{
2978
	static_branch_dec(&preempt_notifier_key);
2979 2980 2981
}
EXPORT_SYMBOL_GPL(preempt_notifier_dec);

2982
/**
2983
 * preempt_notifier_register - tell me when current is being preempted & rescheduled
2984
 * @notifier: notifier struct to register
2985 2986 2987
 */
void preempt_notifier_register(struct preempt_notifier *notifier)
{
2988
	if (!static_branch_unlikely(&preempt_notifier_key))
2989 2990
		WARN(1, "registering preempt_notifier while notifiers disabled\n");

2991 2992 2993 2994 2995 2996
	hlist_add_head(&notifier->link, &current->preempt_notifiers);
}
EXPORT_SYMBOL_GPL(preempt_notifier_register);

/**
 * preempt_notifier_unregister - no longer interested in preemption notifications
2997
 * @notifier: notifier struct to unregister
2998
 *
2999
 * This is *not* safe to call from within a preemption notifier.
3000 3001 3002 3003 3004 3005 3006
 */
void preempt_notifier_unregister(struct preempt_notifier *notifier)
{
	hlist_del(&notifier->link);
}
EXPORT_SYMBOL_GPL(preempt_notifier_unregister);

3007
static void __fire_sched_in_preempt_notifiers(struct task_struct *curr)
3008 3009 3010
{
	struct preempt_notifier *notifier;

3011
	hlist_for_each_entry(notifier, &curr->preempt_notifiers, link)
3012 3013 3014
		notifier->ops->sched_in(notifier, raw_smp_processor_id());
}

3015 3016
static __always_inline void fire_sched_in_preempt_notifiers(struct task_struct *curr)
{
3017
	if (static_branch_unlikely(&preempt_notifier_key))
3018 3019 3020
		__fire_sched_in_preempt_notifiers(curr);
}

3021
static void
3022 3023
__fire_sched_out_preempt_notifiers(struct task_struct *curr,
				   struct task_struct *next)
3024 3025 3026
{
	struct preempt_notifier *notifier;

3027
	hlist_for_each_entry(notifier, &curr->preempt_notifiers, link)
3028 3029 3030
		notifier->ops->sched_out(notifier, next);
}

3031 3032 3033 3034
static __always_inline void
fire_sched_out_preempt_notifiers(struct task_struct *curr,
				 struct task_struct *next)
{
3035
	if (static_branch_unlikely(&preempt_notifier_key))
3036 3037 3038
		__fire_sched_out_preempt_notifiers(curr, next);
}

3039
#else /* !CONFIG_PREEMPT_NOTIFIERS */
3040

3041
static inline void fire_sched_in_preempt_notifiers(struct task_struct *curr)
3042 3043 3044
{
}

3045
static inline void
3046 3047 3048 3049 3050
fire_sched_out_preempt_notifiers(struct task_struct *curr,
				 struct task_struct *next)
{
}

3051
#endif /* CONFIG_PREEMPT_NOTIFIERS */
3052

3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080
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
}

3081 3082
static inline void
prepare_lock_switch(struct rq *rq, struct task_struct *next, struct rq_flags *rf)
3083
{
3084 3085 3086 3087 3088 3089 3090
	/*
	 * 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);
3091
	spin_release(&rq->lock.dep_map, _THIS_IP_);
3092 3093
#ifdef CONFIG_DEBUG_SPINLOCK
	/* this is a valid case when another task releases the spinlock */
3094
	rq->lock.owner = next;
3095
#endif
3096 3097 3098 3099
}

static inline void finish_lock_switch(struct rq *rq)
{
3100 3101 3102 3103 3104 3105 3106 3107 3108
	/*
	 * 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);
}

3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120
/*
 * 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

3121 3122 3123
/**
 * prepare_task_switch - prepare to switch tasks
 * @rq: the runqueue preparing to switch
3124
 * @prev: the current task that is being switched out
3125 3126 3127 3128 3129 3130 3131 3132 3133
 * @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.
 */
3134 3135 3136
static inline void
prepare_task_switch(struct rq *rq, struct task_struct *prev,
		    struct task_struct *next)
3137
{
3138
	kcov_prepare_switch(prev);
3139
	sched_info_switch(rq, prev, next);
3140
	perf_event_task_sched_out(prev, next);
3141
	rseq_preempt(prev);
3142
	fire_sched_out_preempt_notifiers(prev, next);
3143
	prepare_task(next);
3144 3145 3146
	prepare_arch_switch(next);
}

Linus Torvalds's avatar
Linus Torvalds committed
3147 3148 3149 3150
/**
 * finish_task_switch - clean up after a task-switch
 * @prev: the thread we just switched away from.
 *
3151 3152 3153 3154
 * 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
3155 3156
 *
 * Note that we may have delayed dropping an mm in context_switch(). If
Ingo Molnar's avatar
Ingo Molnar committed
3157
 * so, we finish that here outside of the runqueue lock. (Doing it
Linus Torvalds's avatar
Linus Torvalds committed
3158 3159
 * with the lock held can cause deadlocks; see schedule() for
 * details.)
3160 3161 3162 3163 3164
 *
 * 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
3165
 */
3166
static struct rq *finish_task_switch(struct task_struct *prev)
Linus Torvalds's avatar
Linus Torvalds committed
3167 3168
	__releases(rq->lock)
{
3169
	struct rq *rq = this_rq();
Linus Torvalds's avatar
Linus Torvalds committed
3170
	struct mm_struct *mm = rq->prev_mm;
Oleg Nesterov's avatar
Oleg Nesterov committed
3171
	long prev_state;
Linus Torvalds's avatar
Linus Torvalds committed
3172

3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183
	/*
	 * 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.
	 */
3184 3185 3186 3187
	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);
3188

Linus Torvalds's avatar
Linus Torvalds committed
3189 3190 3191 3192
	rq->prev_mm = NULL;

	/*
	 * A task struct has one reference for the use as "current".
3193
	 * If a task dies, then it sets TASK_DEAD in tsk->state and calls
Oleg Nesterov's avatar
Oleg Nesterov committed
3194 3195
	 * schedule one last time. The schedule call will never return, and
	 * the scheduled task must drop that reference.
3196 3197
	 *
	 * We must observe prev->state before clearing prev->on_cpu (in
3198
	 * finish_task), otherwise a concurrent wakeup can get prev
3199 3200
	 * 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
3201
	 */
Oleg Nesterov's avatar
Oleg Nesterov committed
3202
	prev_state = prev->state;
3203
	vtime_task_switch(prev);
3204
	perf_event_task_sched_in(prev, current);
3205 3206
	finish_task(prev);
	finish_lock_switch(rq);
3207
	finish_arch_post_lock_switch();
3208
	kcov_finish_switch(current);
Steven Rostedt's avatar
Steven Rostedt committed
3209

3210
	fire_sched_in_preempt_notifiers(current);
3211
	/*
3212 3213 3214 3215 3216 3217 3218 3219 3220 3221
	 * 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.
3222
	 */
3223 3224
	if (mm) {
		membarrier_mm_sync_core_before_usermode(mm);
Linus Torvalds's avatar
Linus Torvalds committed
3225
		mmdrop(mm);
3226
	}
3227 3228 3229
	if (unlikely(prev_state == TASK_DEAD)) {
		if (prev->sched_class->task_dead)
			prev->sched_class->task_dead(prev);
3230

3231 3232 3233 3234 3235 3236 3237 3238 3239
		/*
		 * 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);

3240
		put_task_struct_rcu_user(prev);
3241
	}
3242

3243
	tick_nohz_task_switch();
3244
	return rq;
Linus Torvalds's avatar
Linus Torvalds committed
3245 3246
}

3247 3248 3249
#ifdef CONFIG_SMP

/* rq->lock is NOT held, but preemption is disabled */
3250
static void __balance_callback(struct rq *rq)
3251
{
3252 3253 3254
	struct callback_head *head, *next;
	void (*func)(struct rq *rq);
	unsigned long flags;
3255

3256 3257 3258 3259 3260 3261 3262 3263
	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;
3264

3265
		func(rq);
3266
	}
3267 3268 3269 3270 3271 3272 3273
	raw_spin_unlock_irqrestore(&rq->lock, flags);
}

static inline void balance_callback(struct rq *rq)
{
	if (unlikely(rq->balance_callback))
		__balance_callback(rq);
3274 3275 3276
}

#else
3277

3278
static inline void balance_callback(struct rq *rq)
3279
{
Linus Torvalds's avatar
Linus Torvalds committed
3280 3281
}

3282 3283
#endif

Linus Torvalds's avatar
Linus Torvalds committed
3284 3285 3286 3287
/**
 * schedule_tail - first thing a freshly forked thread must call.
 * @prev: the thread we just switched away from.
 */
3288
asmlinkage __visible void schedule_tail(struct task_struct *prev)
Linus Torvalds's avatar
Linus Torvalds committed
3289 3290
	__releases(rq->lock)
{
3291
	struct rq *rq;
3292

3293 3294 3295 3296 3297 3298 3299 3300 3301
	/*
	 * 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).
	 */

3302
	rq = finish_task_switch(prev);
3303
	balance_callback(rq);
3304
	preempt_enable();
3305

Linus Torvalds's avatar
Linus Torvalds committed
3306
	if (current->set_child_tid)
3307
		put_user(task_pid_vnr(current), current->set_child_tid);
3308 3309

	calculate_sigpending();
Linus Torvalds's avatar
Linus Torvalds committed
3310 3311 3312
}

/*
3313
 * context_switch - switch to the new MM and the new thread's register state.
Linus Torvalds's avatar
Linus Torvalds committed
3314
 */
3315
static __always_inline struct rq *
3316
context_switch(struct rq *rq, struct task_struct *prev,
3317
	       struct task_struct *next, struct rq_flags *rf)
Linus Torvalds's avatar
Linus Torvalds committed
3318
{
3319
	prepare_task_switch(rq, prev, next);
3320

3321 3322 3323 3324 3325
	/*
	 * For paravirt, this is coupled with an exit in switch_to to
	 * combine the page table reload and the switch backend into
	 * one hypercall.
	 */
3326
	arch_start_context_switch(prev);
3327

3328
	/*
3329 3330 3331 3332 3333
	 * kernel -> kernel   lazy + transfer active
	 *   user -> kernel   lazy + mmgrab() active
	 *
	 * kernel ->   user   switch + mmdrop() active
	 *   user ->   user   switch
3334
	 */
3335 3336 3337 3338 3339 3340 3341 3342 3343
	if (!next->mm) {                                // to kernel
		enter_lazy_tlb(prev->active_mm, next);

		next->active_mm = prev->active_mm;
		if (prev->mm)                           // from user
			mmgrab(prev->active_mm);
		else
			prev->active_mm = NULL;
	} else {                                        // to user
3344
		membarrier_switch_mm(rq, prev->active_mm, next->mm);
3345 3346
		/*
		 * sys_membarrier() requires an smp_mb() between setting
3347
		 * rq->curr / membarrier_switch_mm() and returning to userspace.
3348 3349 3350 3351 3352 3353
		 *
		 * The below provides this either through switch_mm(), or in
		 * case 'prev->active_mm == next->mm' through
		 * finish_task_switch()'s mmdrop().
		 */
		switch_mm_irqs_off(prev->active_mm, next->mm, next);
Linus Torvalds's avatar
Linus Torvalds committed
3354

3355 3356 3357 3358 3359
		if (!prev->mm) {                        // from kernel
			/* will mmdrop() in finish_task_switch(). */
			rq->prev_mm = prev->active_mm;
			prev->active_mm = NULL;
		}
Linus Torvalds's avatar
Linus Torvalds committed
3360
	}
3361

3362
	rq->clock_update_flags &= ~(RQCF_ACT_SKIP|RQCF_REQ_SKIP);
3363

3364
	prepare_lock_switch(rq, next, rf);
Linus Torvalds's avatar
Linus Torvalds committed
3365 3366 3367

	/* Here we just switch the register state and the stack. */
	switch_to(prev, next, prev);
Ingo Molnar's avatar
Ingo Molnar committed
3368
	barrier();
3369 3370

	return finish_task_switch(prev);
Linus Torvalds's avatar
Linus Torvalds committed
3371 3372 3373
}

/*
3374
 * nr_running and nr_context_switches:
Linus Torvalds's avatar
Linus Torvalds committed
3375 3376
 *
 * externally visible scheduler statistics: current number of runnable
3377
 * threads, total number of context switches performed since bootup.
Linus Torvalds's avatar
Linus Torvalds committed
3378 3379 3380 3381 3382 3383 3384 3385 3386
 */
unsigned long nr_running(void)
{
	unsigned long i, sum = 0;

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

	return sum;
3387
}
Linus Torvalds's avatar
Linus Torvalds committed
3388

3389
/*
3390
 * Check if only the current task is running on the CPU.
3391 3392 3393 3394 3395
 *
 * 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:
 *
3396
 * - from a non-preemptible section (of course)
3397 3398 3399 3400
 *
 * - from a thread that is bound to a single CPU
 *
 * - in a loop with very short iterations (e.g. a polling loop)
3401 3402 3403
 */
bool single_task_running(void)
{
3404
	return raw_rq()->nr_running == 1;
3405 3406 3407
}
EXPORT_SYMBOL(single_task_running);

Linus Torvalds's avatar
Linus Torvalds committed
3408
unsigned long long nr_context_switches(void)
3409
{
3410 3411
	int i;
	unsigned long long sum = 0;
3412

3413
	for_each_possible_cpu(i)
Linus Torvalds's avatar
Linus Torvalds committed
3414
		sum += cpu_rq(i)->nr_switches;
3415

Linus Torvalds's avatar
Linus Torvalds committed
3416 3417
	return sum;
}
3418

3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430
/*
 * 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);
}

3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460
/*
 * 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
3461 3462 3463
unsigned long nr_iowait(void)
{
	unsigned long i, sum = 0;
3464

3465
	for_each_possible_cpu(i)
3466
		sum += nr_iowait_cpu(i);
3467

Linus Torvalds's avatar
Linus Torvalds committed
3468 3469
	return sum;
}
3470

Ingo Molnar's avatar
Ingo Molnar committed
3471
#ifdef CONFIG_SMP
3472

3473
/*
3474 3475
 * sched_exec - execve() is a valuable balancing opportunity, because at
 * this point the task has the smallest effective memory and cache footprint.
3476
 */
3477
void sched_exec(void)
3478
{
3479
	struct task_struct *p = current;
Linus Torvalds's avatar
Linus Torvalds committed
3480
	unsigned long flags;
3481
	int dest_cpu;
3482

3483
	raw_spin_lock_irqsave(&p->pi_lock, flags);
3484
	dest_cpu = p->sched_class->select_task_rq(p, task_cpu(p), SD_BALANCE_EXEC, 0);
3485 3486
	if (dest_cpu == smp_processor_id())
		goto unlock;
3487

3488
	if (likely(cpu_active(dest_cpu))) {
3489
		struct migration_arg arg = { p, dest_cpu };
3490

3491 3492
		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
3493 3494
		return;
	}
3495
unlock:
3496
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
3497
}
Ingo Molnar's avatar
Ingo Molnar committed
3498

Linus Torvalds's avatar
Linus Torvalds committed
3499 3500 3501
#endif

DEFINE_PER_CPU(struct kernel_stat, kstat);
3502
DEFINE_PER_CPU(struct kernel_cpustat, kernel_cpustat);
Linus Torvalds's avatar
Linus Torvalds committed
3503 3504

EXPORT_PER_CPU_SYMBOL(kstat);
3505
EXPORT_PER_CPU_SYMBOL(kernel_cpustat);
Linus Torvalds's avatar
Linus Torvalds committed
3506

3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523
/*
 * 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);
}

3524 3525 3526 3527 3528 3529 3530
/*
 * 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)
{
3531
	struct rq_flags rf;
3532
	struct rq *rq;
3533
	u64 ns;
3534

3535 3536
#if defined(CONFIG_64BIT) && defined(CONFIG_SMP)
	/*
3537
	 * 64-bit doesn't need locks to atomically read a 64-bit value.
3538 3539 3540
	 * So we have a optimization chance when the task's delta_exec is 0.
	 * Reading ->on_cpu is racy, but this is ok.
	 *
3541 3542
	 * 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
3543
	 * indistinguishable from the read occurring a few cycles earlier.
3544 3545
	 * If we see ->on_cpu without ->on_rq, the task is leaving, and has
	 * been accounted, so we're correct here as well.
3546
	 */
3547
	if (!p->on_cpu || !task_on_rq_queued(p))
3548 3549 3550
		return p->se.sum_exec_runtime;
#endif

3551
	rq = task_rq_lock(p, &rf);
3552 3553 3554 3555 3556 3557
	/*
	 * 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)) {
3558
		prefetch_curr_exec_start(p);
3559 3560 3561 3562
		update_rq_clock(rq);
		p->sched_class->update_curr(rq);
	}
	ns = p->se.sum_exec_runtime;
3563
	task_rq_unlock(rq, p, &rf);
3564 3565 3566

	return ns;
}
3567

3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578
DEFINE_PER_CPU(unsigned long, thermal_pressure);

void arch_set_thermal_pressure(struct cpumask *cpus,
			       unsigned long th_pressure)
{
	int cpu;

	for_each_cpu(cpu, cpus)
		WRITE_ONCE(per_cpu(thermal_pressure, cpu), th_pressure);
}

3579 3580 3581 3582 3583 3584 3585 3586
/*
 * 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
3587
	struct task_struct *curr = rq->curr;
3588
	struct rq_flags rf;
3589
	unsigned long thermal_pressure;
3590

3591
	arch_scale_freq_tick();
3592
	sched_clock_tick();
Ingo Molnar's avatar
Ingo Molnar committed
3593

3594 3595
	rq_lock(rq, &rf);

3596
	update_rq_clock(rq);
3597
	thermal_pressure = arch_scale_thermal_pressure(cpu_of(rq));
3598
	update_thermal_load_avg(rq_clock_thermal(rq), rq, thermal_pressure);
Peter Zijlstra's avatar
Peter Zijlstra committed
3599
	curr->sched_class->task_tick(rq, curr, 0);
3600
	calc_global_load_tick(rq);
3601
	psi_task_tick(rq);
3602 3603

	rq_unlock(rq, &rf);
3604

3605
	perf_event_task_tick();
3606

3607
#ifdef CONFIG_SMP
3608
	rq->idle_balance = idle_cpu(cpu);
3609
	trigger_load_balance(rq);
3610
#endif
Linus Torvalds's avatar
Linus Torvalds committed
3611 3612
}

3613
#ifdef CONFIG_NO_HZ_FULL
3614 3615 3616

struct tick_work {
	int			cpu;
3617
	atomic_t		state;
3618 3619
	struct delayed_work	work;
};
3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646
/* Values for ->state, see diagram below. */
#define TICK_SCHED_REMOTE_OFFLINE	0
#define TICK_SCHED_REMOTE_OFFLINING	1
#define TICK_SCHED_REMOTE_RUNNING	2

/*
 * State diagram for ->state:
 *
 *
 *          TICK_SCHED_REMOTE_OFFLINE
 *                    |   ^
 *                    |   |
 *                    |   | sched_tick_remote()
 *                    |   |
 *                    |   |
 *                    +--TICK_SCHED_REMOTE_OFFLINING
 *                    |   ^
 *                    |   |
 * sched_tick_start() |   | sched_tick_stop()
 *                    |   |
 *                    V   |
 *          TICK_SCHED_REMOTE_RUNNING
 *
 *
 * Other transitions get WARN_ON_ONCE(), except that sched_tick_remote()
 * and sched_tick_start() are happy to leave the state in RUNNING.
 */
3647 3648 3649 3650 3651 3652 3653 3654 3655

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);
3656
	struct task_struct *curr;
3657
	struct rq_flags rf;
3658
	u64 delta;
3659
	int os;
3660 3661 3662 3663 3664 3665 3666 3667

	/*
	 * 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.
	 */
3668
	if (!tick_nohz_tick_stopped_cpu(cpu))
3669
		goto out_requeue;
3670

3671 3672
	rq_lock_irq(rq, &rf);
	curr = rq->curr;
3673
	if (cpu_is_offline(cpu))
3674
		goto out_unlock;
3675

3676 3677
	update_rq_clock(rq);

3678 3679 3680 3681 3682 3683 3684 3685
	if (!is_idle_task(curr)) {
		/*
		 * Make sure the next tick runs within a reasonable
		 * amount of time.
		 */
		delta = rq_clock_task(rq) - curr->se.exec_start;
		WARN_ON_ONCE(delta > (u64)NSEC_PER_SEC * 3);
	}
3686 3687
	curr->sched_class->task_tick(rq, curr, 0);

3688
	calc_load_nohz_remote(rq);
3689 3690 3691
out_unlock:
	rq_unlock_irq(rq, &rf);
out_requeue:
3692

3693 3694 3695
	/*
	 * Run the remote tick once per second (1Hz). This arbitrary
	 * frequency is large enough to avoid overload but short enough
3696 3697
	 * to keep scheduler internal stats reasonably up to date.  But
	 * first update state to reflect hotplug activity if required.
3698
	 */
3699 3700 3701 3702
	os = atomic_fetch_add_unless(&twork->state, -1, TICK_SCHED_REMOTE_RUNNING);
	WARN_ON_ONCE(os == TICK_SCHED_REMOTE_OFFLINE);
	if (os == TICK_SCHED_REMOTE_RUNNING)
		queue_delayed_work(system_unbound_wq, dwork, HZ);
3703 3704 3705 3706
}

static void sched_tick_start(int cpu)
{
3707
	int os;
3708 3709 3710 3711 3712 3713 3714 3715
	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);
3716 3717 3718 3719 3720 3721 3722
	os = atomic_xchg(&twork->state, TICK_SCHED_REMOTE_RUNNING);
	WARN_ON_ONCE(os == TICK_SCHED_REMOTE_RUNNING);
	if (os == TICK_SCHED_REMOTE_OFFLINE) {
		twork->cpu = cpu;
		INIT_DELAYED_WORK(&twork->work, sched_tick_remote);
		queue_delayed_work(system_unbound_wq, &twork->work, HZ);
	}
3723 3724 3725 3726 3727 3728
}

#ifdef CONFIG_HOTPLUG_CPU
static void sched_tick_stop(int cpu)
{
	struct tick_work *twork;
3729
	int os;
3730 3731 3732 3733 3734 3735 3736

	if (housekeeping_cpu(cpu, HK_FLAG_TICK))
		return;

	WARN_ON_ONCE(!tick_work_cpu);

	twork = per_cpu_ptr(tick_work_cpu, cpu);
3737 3738 3739 3740
	/* There cannot be competing actions, but don't rely on stop-machine. */
	os = atomic_xchg(&twork->state, TICK_SCHED_REMOTE_OFFLINING);
	WARN_ON_ONCE(os != TICK_SCHED_REMOTE_RUNNING);
	/* Don't cancel, as this would mess up the state machine. */
3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753
}
#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) { }
3754
#endif
Linus Torvalds's avatar
Linus Torvalds committed
3755

3756
#if defined(CONFIG_PREEMPTION) && (defined(CONFIG_DEBUG_PREEMPT) || \
3757
				defined(CONFIG_TRACE_PREEMPT_TOGGLE))
3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771
/*
 * 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);
	}
}
3772

3773
void preempt_count_add(int val)
Linus Torvalds's avatar
Linus Torvalds committed
3774
{
3775
#ifdef CONFIG_DEBUG_PREEMPT
Linus Torvalds's avatar
Linus Torvalds committed
3776 3777 3778
	/*
	 * Underflow?
	 */
3779 3780
	if (DEBUG_LOCKS_WARN_ON((preempt_count() < 0)))
		return;
3781
#endif
3782
	__preempt_count_add(val);
3783
#ifdef CONFIG_DEBUG_PREEMPT
Linus Torvalds's avatar
Linus Torvalds committed
3784 3785 3786
	/*
	 * Spinlock count overflowing soon?
	 */
3787 3788
	DEBUG_LOCKS_WARN_ON((preempt_count() & PREEMPT_MASK) >=
				PREEMPT_MASK - 10);
3789
#endif
3790
	preempt_latency_start(val);
Linus Torvalds's avatar
Linus Torvalds committed
3791
}
3792
EXPORT_SYMBOL(preempt_count_add);
3793
NOKPROBE_SYMBOL(preempt_count_add);
Linus Torvalds's avatar
Linus Torvalds committed
3794

3795 3796 3797 3798 3799 3800 3801 3802 3803 3804
/*
 * 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());
}

3805
void preempt_count_sub(int val)
Linus Torvalds's avatar
Linus Torvalds committed
3806
{
3807
#ifdef CONFIG_DEBUG_PREEMPT
Linus Torvalds's avatar
Linus Torvalds committed
3808 3809 3810
	/*
	 * Underflow?
	 */
3811
	if (DEBUG_LOCKS_WARN_ON(val > preempt_count()))
3812
		return;
Linus Torvalds's avatar
Linus Torvalds committed
3813 3814 3815
	/*
	 * Is the spinlock portion underflowing?
	 */
3816 3817 3818
	if (DEBUG_LOCKS_WARN_ON((val < PREEMPT_MASK) &&
			!(preempt_count() & PREEMPT_MASK)))
		return;
3819
#endif
3820

3821
	preempt_latency_stop(val);
3822
	__preempt_count_sub(val);
Linus Torvalds's avatar
Linus Torvalds committed
3823
}
3824
EXPORT_SYMBOL(preempt_count_sub);
3825
NOKPROBE_SYMBOL(preempt_count_sub);
Linus Torvalds's avatar
Linus Torvalds committed
3826

3827 3828 3829
#else
static inline void preempt_latency_start(int val) { }
static inline void preempt_latency_stop(int val) { }
Linus Torvalds's avatar
Linus Torvalds committed
3830 3831
#endif

3832 3833 3834 3835 3836 3837 3838 3839 3840
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
3841
/*
Ingo Molnar's avatar
Ingo Molnar committed
3842
 * Print scheduling while atomic bug:
Linus Torvalds's avatar
Linus Torvalds committed
3843
 */
Ingo Molnar's avatar
Ingo Molnar committed
3844
static noinline void __schedule_bug(struct task_struct *prev)
Linus Torvalds's avatar
Linus Torvalds committed
3845
{
3846 3847 3848
	/* Save this before calling printk(), since that will clobber it */
	unsigned long preempt_disable_ip = get_preempt_disable_ip(current);

3849 3850 3851
	if (oops_in_progress)
		return;

3852 3853
	printk(KERN_ERR "BUG: scheduling while atomic: %s/%d/0x%08x\n",
		prev->comm, prev->pid, preempt_count());
3854

Ingo Molnar's avatar
Ingo Molnar committed
3855
	debug_show_held_locks(prev);
3856
	print_modules();
Ingo Molnar's avatar
Ingo Molnar committed
3857 3858
	if (irqs_disabled())
		print_irqtrace_events(prev);
3859 3860
	if (IS_ENABLED(CONFIG_DEBUG_PREEMPT)
	    && in_atomic_preempt_off()) {
3861
		pr_err("Preemption disabled at:");
3862
		print_ip_sym(preempt_disable_ip);
3863 3864
		pr_cont("\n");
	}
3865 3866 3867
	if (panic_on_warn)
		panic("scheduling while atomic\n");

3868
	dump_stack();
3869
	add_taint(TAINT_WARN, LOCKDEP_STILL_OK);
Ingo Molnar's avatar
Ingo Molnar committed
3870
}
Linus Torvalds's avatar
Linus Torvalds committed
3871

Ingo Molnar's avatar
Ingo Molnar committed
3872 3873 3874
/*
 * Various schedule()-time debugging checks and statistics:
 */
3875
static inline void schedule_debug(struct task_struct *prev, bool preempt)
Ingo Molnar's avatar
Ingo Molnar committed
3876
{
3877
#ifdef CONFIG_SCHED_STACK_END_CHECK
3878 3879
	if (task_stack_end_corrupted(prev))
		panic("corrupted stack end detected inside scheduler\n");
3880
#endif
3881

3882 3883 3884 3885 3886 3887 3888 3889 3890
#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
	if (!preempt && prev->state && prev->non_block_count) {
		printk(KERN_ERR "BUG: scheduling in a non-blocking section: %s/%d/%i\n",
			prev->comm, prev->pid, prev->non_block_count);
		dump_stack();
		add_taint(TAINT_WARN, LOCKDEP_STILL_OK);
	}
#endif

3891
	if (unlikely(in_atomic_preempt_off())) {
Ingo Molnar's avatar
Ingo Molnar committed
3892
		__schedule_bug(prev);
3893 3894
		preempt_count_set(PREEMPT_DISABLED);
	}
3895
	rcu_sleep_check();
Ingo Molnar's avatar
Ingo Molnar committed
3896

Linus Torvalds's avatar
Linus Torvalds committed
3897 3898
	profile_hit(SCHED_PROFILING, __builtin_return_address(0));

3899
	schedstat_inc(this_rq()->sched_count);
Ingo Molnar's avatar
Ingo Molnar committed
3900 3901
}

3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923
static void put_prev_task_balance(struct rq *rq, struct task_struct *prev,
				  struct rq_flags *rf)
{
#ifdef CONFIG_SMP
	const struct sched_class *class;
	/*
	 * We must do the balancing pass before put_prev_task(), such
	 * that when we release the rq->lock the task is in the same
	 * state as before we took rq->lock.
	 *
	 * We can terminate the balance pass as soon as we know there is
	 * a runnable task of @class priority or higher.
	 */
	for_class_range(class, prev->sched_class, &idle_sched_class) {
		if (class->balance(rq, prev, rf))
			break;
	}
#endif

	put_prev_task(rq, prev);
}

Ingo Molnar's avatar
Ingo Molnar committed
3924 3925 3926 3927
/*
 * Pick up the highest-prio task:
 */
static inline struct task_struct *
3928
pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
Ingo Molnar's avatar
Ingo Molnar committed
3929
{
3930
	const struct sched_class *class;
Ingo Molnar's avatar
Ingo Molnar committed
3931
	struct task_struct *p;
Linus Torvalds's avatar
Linus Torvalds committed
3932 3933

	/*
3934 3935 3936 3937
	 * 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
3938
	 */
3939 3940 3941 3942
	if (likely((prev->sched_class == &idle_sched_class ||
		    prev->sched_class == &fair_sched_class) &&
		   rq->nr_running == rq->cfs.h_nr_running)) {

3943
		p = pick_next_task_fair(rq, prev, rf);
3944
		if (unlikely(p == RETRY_TASK))
3945
			goto restart;
3946

3947
		/* Assumes fair_sched_class->next == idle_sched_class */
3948
		if (!p) {
3949
			put_prev_task(rq, prev);
3950
			p = pick_next_task_idle(rq);
3951
		}
3952 3953

		return p;
Linus Torvalds's avatar
Linus Torvalds committed
3954 3955
	}

3956
restart:
3957
	put_prev_task_balance(rq, prev, rf);
3958

3959
	for_each_class(class) {
3960
		p = class->pick_next_task(rq);
3961
		if (p)
Ingo Molnar's avatar
Ingo Molnar committed
3962 3963
			return p;
	}
3964

3965 3966
	/* The idle class should always have a runnable task: */
	BUG();
Ingo Molnar's avatar
Ingo Molnar committed
3967
}
Linus Torvalds's avatar
Linus Torvalds committed
3968

Ingo Molnar's avatar
Ingo Molnar committed
3969
/*
3970
 * __schedule() is the main scheduler function.
3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988
 *
 * 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:
 *
3989
 *       - If the kernel is preemptible (CONFIG_PREEMPTION=y):
3990 3991 3992 3993 3994 3995 3996 3997
 *
 *         - 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
 *
3998
 *       - If the kernel is not preemptible (CONFIG_PREEMPTION is not set)
3999 4000 4001 4002 4003 4004
 *         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
4005
 *
4006
 * WARNING: must be called with preemption disabled!
Ingo Molnar's avatar
Ingo Molnar committed
4007
 */
4008
static void __sched notrace __schedule(bool preempt)
Ingo Molnar's avatar
Ingo Molnar committed
4009 4010
{
	struct task_struct *prev, *next;
4011
	unsigned long *switch_count;
4012
	struct rq_flags rf;
Ingo Molnar's avatar
Ingo Molnar committed
4013
	struct rq *rq;
4014
	int cpu;
Ingo Molnar's avatar
Ingo Molnar committed
4015 4016 4017 4018 4019

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

4020
	schedule_debug(prev, preempt);
Linus Torvalds's avatar
Linus Torvalds committed
4021

4022
	if (sched_feat(HRTICK))
Mike Galbraith's avatar
Mike Galbraith committed
4023
		hrtick_clear(rq);
4024

4025
	local_irq_disable();
4026
	rcu_note_context_switch(preempt);
4027

4028 4029 4030 4031
	/*
	 * 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().
4032 4033 4034
	 *
	 * The membarrier system call requires a full memory barrier
	 * after coming from user-space, before storing to rq->curr.
4035
	 */
4036
	rq_lock(rq, &rf);
4037
	smp_mb__after_spinlock();
Linus Torvalds's avatar
Linus Torvalds committed
4038

4039 4040
	/* Promote REQ to ACT */
	rq->clock_update_flags <<= 1;
4041
	update_rq_clock(rq);
4042

4043
	switch_count = &prev->nivcsw;
4044
	if (!preempt && prev->state) {
4045
		if (signal_pending_state(prev->state, prev)) {
Linus Torvalds's avatar
Linus Torvalds committed
4046
			prev->state = TASK_RUNNING;
Tejun Heo's avatar
Tejun Heo committed
4047
		} else {
4048
			deactivate_task(rq, prev, DEQUEUE_SLEEP | DEQUEUE_NOCLOCK);
4049

4050 4051 4052 4053
			if (prev->in_iowait) {
				atomic_inc(&rq->nr_iowait);
				delayacct_blkio_start();
			}
Tejun Heo's avatar
Tejun Heo committed
4054
		}
Ingo Molnar's avatar
Ingo Molnar committed
4055
		switch_count = &prev->nvcsw;
Linus Torvalds's avatar
Linus Torvalds committed
4056 4057
	}

4058
	next = pick_next_task(rq, prev, &rf);
4059
	clear_tsk_need_resched(prev);
4060
	clear_preempt_need_resched();
Linus Torvalds's avatar
Linus Torvalds committed
4061 4062 4063

	if (likely(prev != next)) {
		rq->nr_switches++;
4064 4065 4066 4067 4068
		/*
		 * RCU users of rcu_dereference(rq->curr) may not see
		 * changes to task_struct made by pick_next_task().
		 */
		RCU_INIT_POINTER(rq->curr, next);
4069 4070 4071
		/*
		 * The membarrier system call requires each architecture
		 * to have a full memory barrier after updating
4072 4073 4074 4075 4076 4077 4078 4079 4080 4081
		 * 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),
4082
		 */
Linus Torvalds's avatar
Linus Torvalds committed
4083 4084
		++*switch_count;

4085 4086
		psi_sched_switch(prev, next, !task_on_rq_queued(prev));

4087
		trace_sched_switch(preempt, prev, next);
4088 4089 4090

		/* Also unlocks the rq: */
		rq = context_switch(rq, prev, next, &rf);
4091
	} else {
4092
		rq->clock_update_flags &= ~(RQCF_ACT_SKIP|RQCF_REQ_SKIP);
4093
		rq_unlock_irq(rq, &rf);
4094
	}
Linus Torvalds's avatar
Linus Torvalds committed
4095

4096
	balance_callback(rq);
Linus Torvalds's avatar
Linus Torvalds committed
4097
}
4098

4099 4100
void __noreturn do_task_dead(void)
{
4101
	/* Causes final put_task_struct in finish_task_switch(): */
4102
	set_special_state(TASK_DEAD);
4103 4104 4105 4106

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

4107 4108
	__schedule(false);
	BUG();
4109 4110

	/* Avoid "noreturn function does return" - but don't continue if BUG() is a NOP: */
4111
	for (;;)
4112
		cpu_relax();
4113 4114
}

4115 4116
static inline void sched_submit_work(struct task_struct *tsk)
{
4117
	if (!tsk->state)
4118
		return;
4119 4120 4121 4122 4123 4124

	/*
	 * 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
4125 4126
	 * in the possible wakeup of a kworker and because wq_worker_sleeping()
	 * requires it.
4127
	 */
4128
	if (tsk->flags & (PF_WQ_WORKER | PF_IO_WORKER)) {
4129
		preempt_disable();
4130 4131 4132 4133
		if (tsk->flags & PF_WQ_WORKER)
			wq_worker_sleeping(tsk);
		else
			io_wq_worker_sleeping(tsk);
4134 4135 4136
		preempt_enable_no_resched();
	}

4137 4138 4139
	if (tsk_is_pi_blocked(tsk))
		return;

4140 4141 4142 4143 4144 4145 4146 4147
	/*
	 * 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);
}

4148 4149
static void sched_update_worker(struct task_struct *tsk)
{
4150 4151 4152 4153 4154 4155
	if (tsk->flags & (PF_WQ_WORKER | PF_IO_WORKER)) {
		if (tsk->flags & PF_WQ_WORKER)
			wq_worker_running(tsk);
		else
			io_wq_worker_running(tsk);
	}
4156 4157
}

4158
asmlinkage __visible void __sched schedule(void)
4159
{
4160 4161 4162
	struct task_struct *tsk = current;

	sched_submit_work(tsk);
4163
	do {
4164
		preempt_disable();
4165
		__schedule(false);
4166
		sched_preempt_enable_no_resched();
4167
	} while (need_resched());
4168
	sched_update_worker(tsk);
4169
}
Linus Torvalds's avatar
Linus Torvalds committed
4170 4171
EXPORT_SYMBOL(schedule);

4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196
/*
 * 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());
}

4197
#ifdef CONFIG_CONTEXT_TRACKING
4198
asmlinkage __visible void __sched schedule_user(void)
4199 4200 4201 4202 4203 4204
{
	/*
	 * 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.
4205 4206
	 *
	 * NB: There are buggy callers of this function.  Ideally we
4207
	 * should warn if prev_state != CONTEXT_USER, but that will trigger
4208
	 * too frequently to make sense yet.
4209
	 */
4210
	enum ctx_state prev_state = exception_enter();
4211
	schedule();
4212
	exception_exit(prev_state);
4213 4214 4215
}
#endif

4216 4217 4218 4219 4220 4221 4222
/**
 * schedule_preempt_disabled - called with preemption disabled
 *
 * Returns with preemption disabled. Note: preempt_count must be 1
 */
void __sched schedule_preempt_disabled(void)
{
4223
	sched_preempt_enable_no_resched();
4224 4225 4226 4227
	schedule();
	preempt_disable();
}

4228
static void __sched notrace preempt_schedule_common(void)
4229 4230
{
	do {
4231 4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242 4243
		/*
		 * 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.
		 */
4244
		preempt_disable_notrace();
4245
		preempt_latency_start(1);
4246
		__schedule(true);
4247
		preempt_latency_stop(1);
4248
		preempt_enable_no_resched_notrace();
4249 4250 4251 4252 4253 4254 4255 4256

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

4257
#ifdef CONFIG_PREEMPTION
Linus Torvalds's avatar
Linus Torvalds committed
4258
/*
4259 4260
 * This is the entry point to schedule() from in-kernel preemption
 * off of preempt_enable.
Linus Torvalds's avatar
Linus Torvalds committed
4261
 */
4262
asmlinkage __visible void __sched notrace preempt_schedule(void)
Linus Torvalds's avatar
Linus Torvalds committed
4263 4264 4265
{
	/*
	 * If there is a non-zero preempt_count or interrupts are disabled,
Ingo Molnar's avatar
Ingo Molnar committed
4266
	 * we do not want to preempt the current task. Just return..
Linus Torvalds's avatar
Linus Torvalds committed
4267
	 */
4268
	if (likely(!preemptible()))
Linus Torvalds's avatar
Linus Torvalds committed
4269 4270
		return;

4271
	preempt_schedule_common();
Linus Torvalds's avatar
Linus Torvalds committed
4272
}
4273
NOKPROBE_SYMBOL(preempt_schedule);
Linus Torvalds's avatar
Linus Torvalds committed
4274
EXPORT_SYMBOL(preempt_schedule);
4275 4276

/**
4277
 * preempt_schedule_notrace - preempt_schedule called by tracing
4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289
 *
 * 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.
 */
4290
asmlinkage __visible void __sched notrace preempt_schedule_notrace(void)
4291 4292 4293 4294 4295 4296 4297
{
	enum ctx_state prev_ctx;

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

	do {
4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310
		/*
		 * 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.
		 */
4311
		preempt_disable_notrace();
4312
		preempt_latency_start(1);
4313 4314 4315 4316 4317 4318
		/*
		 * Needs preempt disabled in case user_exit() is traced
		 * and the tracer calls preempt_enable_notrace() causing
		 * an infinite recursion.
		 */
		prev_ctx = exception_enter();
4319
		__schedule(true);
4320 4321
		exception_exit(prev_ctx);

4322
		preempt_latency_stop(1);
4323
		preempt_enable_no_resched_notrace();
4324 4325
	} while (need_resched());
}
4326
EXPORT_SYMBOL_GPL(preempt_schedule_notrace);
4327

4328
#endif /* CONFIG_PREEMPTION */
Linus Torvalds's avatar
Linus Torvalds committed
4329 4330

/*
4331
 * This is the entry point to schedule() from kernel preemption
Linus Torvalds's avatar
Linus Torvalds committed
4332 4333 4334 4335
 * off of irq context.
 * Note, that this is called and return with irqs disabled. This will
 * protect us against recursive calling from irq.
 */
4336
asmlinkage __visible void __sched preempt_schedule_irq(void)
Linus Torvalds's avatar
Linus Torvalds committed
4337
{
4338
	enum ctx_state prev_state;
4339

4340
	/* Catch callers which need to be fixed */
4341
	BUG_ON(preempt_count() || !irqs_disabled());
Linus Torvalds's avatar
Linus Torvalds committed
4342

4343 4344
	prev_state = exception_enter();

4345
	do {
4346
		preempt_disable();
4347
		local_irq_enable();
4348
		__schedule(true);
4349
		local_irq_disable();
4350
		sched_preempt_enable_no_resched();
4351
	} while (need_resched());
4352 4353

	exception_exit(prev_state);
Linus Torvalds's avatar
Linus Torvalds committed
4354 4355
}

4356
int default_wake_function(wait_queue_entry_t *curr, unsigned mode, int wake_flags,
Ingo Molnar's avatar
Ingo Molnar committed
4357
			  void *key)
Linus Torvalds's avatar
Linus Torvalds committed
4358
{
4359
	return try_to_wake_up(curr->private, mode, wake_flags);
Linus Torvalds's avatar
Linus Torvalds committed
4360 4361 4362
}
EXPORT_SYMBOL(default_wake_function);

4363 4364
#ifdef CONFIG_RT_MUTEXES

4365 4366 4367 4368 4369 4370 4371 4372 4373 4374 4375 4376 4377 4378 4379
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);
}

4380 4381
/*
 * rt_mutex_setprio - set the current priority of a task
4382 4383
 * @p: task to boost
 * @pi_task: donor task
4384 4385 4386 4387
 *
 * This function changes the 'effective' priority of a task. It does
 * not touch ->normal_prio like __setscheduler().
 *
4388 4389
 * Used by the rt_mutex code to implement priority inheritance
 * logic. Call site only calls if the priority of the task changed.
4390
 */
4391
void rt_mutex_setprio(struct task_struct *p, struct task_struct *pi_task)
4392
{
4393
	int prio, oldprio, queued, running, queue_flag =
4394
		DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
4395
	const struct sched_class *prev_class;
4396 4397
	struct rq_flags rf;
	struct rq *rq;
4398

4399 4400 4401 4402 4403 4404 4405 4406
	/* 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;
4407

4408
	rq = __task_rq_lock(p, &rf);
4409
	update_rq_clock(rq);
4410 4411 4412 4413 4414 4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426
	/*
	 * 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;
4427

4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445
	/*
	 * 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;
	}

4446
	trace_sched_pi_setprio(p, pi_task);
4447
	oldprio = p->prio;
4448 4449 4450 4451

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

4452
	prev_class = p->sched_class;
4453
	queued = task_on_rq_queued(p);
4454
	running = task_current(rq, p);
4455
	if (queued)
4456
		dequeue_task(rq, p, queue_flag);
4457
	if (running)
4458
		put_prev_task(rq, p);
Ingo Molnar's avatar
Ingo Molnar committed
4459

4460 4461 4462 4463 4464 4465 4466 4467 4468 4469
	/*
	 * 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)) {
4470 4471
		if (!dl_prio(p->normal_prio) ||
		    (pi_task && dl_entity_preempt(&pi_task->dl, &p->dl))) {
4472
			p->dl.dl_boosted = 1;
4473
			queue_flag |= ENQUEUE_REPLENISH;
4474 4475
		} else
			p->dl.dl_boosted = 0;
4476
		p->sched_class = &dl_sched_class;
4477 4478 4479 4480
	} else if (rt_prio(prio)) {
		if (dl_prio(oldprio))
			p->dl.dl_boosted = 0;
		if (oldprio < prio)
4481
			queue_flag |= ENQUEUE_HEAD;
Ingo Molnar's avatar
Ingo Molnar committed
4482
		p->sched_class = &rt_sched_class;
4483 4484 4485
	} else {
		if (dl_prio(oldprio))
			p->dl.dl_boosted = 0;
4486 4487
		if (rt_prio(oldprio))
			p->rt.timeout = 0;
Ingo Molnar's avatar
Ingo Molnar committed
4488
		p->sched_class = &fair_sched_class;
4489
	}
Ingo Molnar's avatar
Ingo Molnar committed
4490

4491 4492
	p->prio = prio;

4493
	if (queued)
4494
		enqueue_task(rq, p, queue_flag);
4495
	if (running)
4496
		set_next_task(rq, p);
4497

4498
	check_class_changed(rq, p, prev_class, oldprio);
4499
out_unlock:
4500 4501
	/* Avoid rq from going away on us: */
	preempt_disable();
4502
	__task_rq_unlock(rq, &rf);
4503 4504 4505

	balance_callback(rq);
	preempt_enable();
4506
}
4507 4508 4509 4510 4511
#else
static inline int rt_effective_prio(struct task_struct *p, int prio)
{
	return prio;
}
4512
#endif
4513

4514
void set_user_nice(struct task_struct *p, long nice)
Linus Torvalds's avatar
Linus Torvalds committed
4515
{
4516
	bool queued, running;
4517
	int old_prio;
4518
	struct rq_flags rf;
4519
	struct rq *rq;
Linus Torvalds's avatar
Linus Torvalds committed
4520

4521
	if (task_nice(p) == nice || nice < MIN_NICE || nice > MAX_NICE)
Linus Torvalds's avatar
Linus Torvalds committed
4522 4523 4524 4525 4526
		return;
	/*
	 * We have to be careful, if called from sys_setpriority(),
	 * the task might be in the middle of scheduling on another CPU.
	 */
4527
	rq = task_rq_lock(p, &rf);
4528 4529
	update_rq_clock(rq);

Linus Torvalds's avatar
Linus Torvalds committed
4530 4531 4532 4533
	/*
	 * 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
4534
	 * SCHED_DEADLINE, SCHED_FIFO or SCHED_RR:
Linus Torvalds's avatar
Linus Torvalds committed
4535
	 */
4536
	if (task_has_dl_policy(p) || task_has_rt_policy(p)) {
Linus Torvalds's avatar
Linus Torvalds committed
4537 4538 4539
		p->static_prio = NICE_TO_PRIO(nice);
		goto out_unlock;
	}
4540
	queued = task_on_rq_queued(p);
4541
	running = task_current(rq, p);
4542
	if (queued)
4543
		dequeue_task(rq, p, DEQUEUE_SAVE | DEQUEUE_NOCLOCK);
4544 4545
	if (running)
		put_prev_task(rq, p);
Linus Torvalds's avatar
Linus Torvalds committed
4546 4547

	p->static_prio = NICE_TO_PRIO(nice);
4548
	set_load_weight(p, true);
4549 4550
	old_prio = p->prio;
	p->prio = effective_prio(p);
Linus Torvalds's avatar
Linus Torvalds committed
4551

4552
	if (queued)
4553
		enqueue_task(rq, p, ENQUEUE_RESTORE | ENQUEUE_NOCLOCK);
4554
	if (running)
4555
		set_next_task(rq, p);
4556 4557 4558 4559 4560 4561 4562

	/*
	 * If the task increased its priority or is running and
	 * lowered its priority, then reschedule its CPU:
	 */
	p->sched_class->prio_changed(rq, p, old_prio);

Linus Torvalds's avatar
Linus Torvalds committed
4563
out_unlock:
4564
	task_rq_unlock(rq, p, &rf);
Linus Torvalds's avatar
Linus Torvalds committed
4565 4566 4567
}
EXPORT_SYMBOL(set_user_nice);

4568 4569 4570 4571 4572
/*
 * can_nice - check if a task can reduce its nice value
 * @p: task
 * @nice: nice value
 */
4573
int can_nice(const struct task_struct *p, const int nice)
4574
{
4575
	/* Convert nice value [19,-20] to rlimit style value [1,40]: */
4576
	int nice_rlim = nice_to_rlimit(nice);
4577

4578
	return (nice_rlim <= task_rlimit(p, RLIMIT_NICE) ||
4579 4580 4581
		capable(CAP_SYS_NICE));
}

Linus Torvalds's avatar
Linus Torvalds committed
4582 4583 4584 4585 4586 4587 4588 4589 4590
#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.
 */
4591
SYSCALL_DEFINE1(nice, int, increment)
Linus Torvalds's avatar
Linus Torvalds committed
4592
{
4593
	long nice, retval;
Linus Torvalds's avatar
Linus Torvalds committed
4594 4595 4596 4597 4598 4599

	/*
	 * 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.
	 */
4600
	increment = clamp(increment, -NICE_WIDTH, NICE_WIDTH);
4601
	nice = task_nice(current) + increment;
Linus Torvalds's avatar
Linus Torvalds committed
4602

4603
	nice = clamp_val(nice, MIN_NICE, MAX_NICE);
4604 4605 4606
	if (increment < 0 && !can_nice(current, nice))
		return -EPERM;

Linus Torvalds's avatar
Linus Torvalds committed
4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620
	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.
 *
4621
 * Return: The priority value as seen by users in /proc.
Linus Torvalds's avatar
Linus Torvalds committed
4622 4623 4624
 * RT tasks are offset by -200. Normal tasks are centered
 * around 0, value goes from -16 to +15.
 */
4625
int task_prio(const struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
4626 4627 4628 4629 4630
{
	return p->prio - MAX_RT_PRIO;
}

/**
4631
 * idle_cpu - is a given CPU idle currently?
Linus Torvalds's avatar
Linus Torvalds committed
4632
 * @cpu: the processor in question.
4633 4634
 *
 * Return: 1 if the CPU is currently idle. 0 otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
4635 4636 4637
 */
int idle_cpu(int cpu)
{
Thomas Gleixner's avatar
Thomas Gleixner committed
4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650 4651
	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
4652 4653
}

4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664
/**
 * 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;

4665 4666 4667
	if (vcpu_is_preempted(cpu))
		return 0;

Thomas Gleixner's avatar
Thomas Gleixner committed
4668
	return 1;
Linus Torvalds's avatar
Linus Torvalds committed
4669 4670 4671
}

/**
4672
 * idle_task - return the idle task for a given CPU.
Linus Torvalds's avatar
Linus Torvalds committed
4673
 * @cpu: the processor in question.
4674
 *
4675
 * Return: The idle task for the CPU @cpu.
Linus Torvalds's avatar
Linus Torvalds committed
4676
 */
4677
struct task_struct *idle_task(int cpu)
Linus Torvalds's avatar
Linus Torvalds committed
4678 4679 4680 4681 4682 4683 4684
{
	return cpu_rq(cpu)->idle;
}

/**
 * find_process_by_pid - find a process with a matching PID value.
 * @pid: the pid in question.
4685 4686
 *
 * The task of @pid, if found. %NULL otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
4687
 */
Alexey Dobriyan's avatar
Alexey Dobriyan committed
4688
static struct task_struct *find_process_by_pid(pid_t pid)
Linus Torvalds's avatar
Linus Torvalds committed
4689
{
4690
	return pid ? find_task_by_vpid(pid) : current;
Linus Torvalds's avatar
Linus Torvalds committed
4691 4692
}

4693 4694 4695 4696 4697 4698
/*
 * sched_setparam() passes in -1 for its policy, to let the functions
 * it calls know not to change it.
 */
#define SETPARAM_POLICY	-1

4699 4700
static void __setscheduler_params(struct task_struct *p,
		const struct sched_attr *attr)
Linus Torvalds's avatar
Linus Torvalds committed
4701
{
4702 4703
	int policy = attr->sched_policy;

4704
	if (policy == SETPARAM_POLICY)
4705 4706
		policy = p->policy;

Linus Torvalds's avatar
Linus Torvalds committed
4707
	p->policy = policy;
4708

4709 4710
	if (dl_policy(policy))
		__setparam_dl(p, attr);
4711
	else if (fair_policy(policy))
4712 4713
		p->static_prio = NICE_TO_PRIO(attr->sched_nice);

4714 4715 4716 4717 4718 4719
	/*
	 * __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;
4720
	p->normal_prio = normal_prio(p);
4721
	set_load_weight(p, true);
4722
}
4723

4724 4725
/* Actually do priority change: must hold pi & rq lock. */
static void __setscheduler(struct rq *rq, struct task_struct *p,
4726
			   const struct sched_attr *attr, bool keep_boost)
4727
{
4728 4729 4730 4731 4732 4733 4734
	/*
	 * If params can't change scheduling class changes aren't allowed
	 * either.
	 */
	if (attr->sched_flags & SCHED_FLAG_KEEP_PARAMS)
		return;

4735
	__setscheduler_params(p, attr);
4736

4737
	/*
4738 4739
	 * Keep a potential priority boosting if called from
	 * sched_setscheduler().
4740
	 */
4741
	p->prio = normal_prio(p);
4742
	if (keep_boost)
4743
		p->prio = rt_effective_prio(p, p->prio);
4744

4745 4746 4747
	if (dl_prio(p->prio))
		p->sched_class = &dl_sched_class;
	else if (rt_prio(p->prio))
4748 4749 4750
		p->sched_class = &rt_sched_class;
	else
		p->sched_class = &fair_sched_class;
Linus Torvalds's avatar
Linus Torvalds committed
4751
}
4752

4753
/*
4754
 * Check the target process has a UID that matches the current process's:
4755 4756 4757 4758 4759 4760 4761 4762
 */
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);
4763 4764
	match = (uid_eq(cred->euid, pcred->euid) ||
		 uid_eq(cred->euid, pcred->uid));
4765 4766 4767 4768
	rcu_read_unlock();
	return match;
}

4769 4770
static int __sched_setscheduler(struct task_struct *p,
				const struct sched_attr *attr,
4771
				bool user, bool pi)
Linus Torvalds's avatar
Linus Torvalds committed
4772
{
4773 4774
	int newprio = dl_policy(attr->sched_policy) ? MAX_DL_PRIO - 1 :
		      MAX_RT_PRIO - 1 - attr->sched_priority;
4775
	int retval, oldprio, oldpolicy = -1, queued, running;
4776
	int new_effective_prio, policy = attr->sched_policy;
4777
	const struct sched_class *prev_class;
4778
	struct rq_flags rf;
4779
	int reset_on_fork;
4780
	int queue_flags = DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
4781
	struct rq *rq;
Linus Torvalds's avatar
Linus Torvalds committed
4782

4783 4784
	/* The pi code expects interrupts enabled */
	BUG_ON(pi && in_interrupt());
Linus Torvalds's avatar
Linus Torvalds committed
4785
recheck:
4786
	/* Double check policy once rq lock held: */
4787 4788
	if (policy < 0) {
		reset_on_fork = p->sched_reset_on_fork;
Linus Torvalds's avatar
Linus Torvalds committed
4789
		policy = oldpolicy = p->policy;
4790
	} else {
4791
		reset_on_fork = !!(attr->sched_flags & SCHED_FLAG_RESET_ON_FORK);
4792

4793
		if (!valid_policy(policy))
4794 4795 4796
			return -EINVAL;
	}

4797
	if (attr->sched_flags & ~(SCHED_FLAG_ALL | SCHED_FLAG_SUGOV))
4798 4799
		return -EINVAL;

Linus Torvalds's avatar
Linus Torvalds committed
4800 4801
	/*
	 * Valid priorities for SCHED_FIFO and SCHED_RR are
Ingo Molnar's avatar
Ingo Molnar committed
4802 4803
	 * 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
4804
	 */
4805
	if ((p->mm && attr->sched_priority > MAX_USER_RT_PRIO-1) ||
4806
	    (!p->mm && attr->sched_priority > MAX_RT_PRIO-1))
Linus Torvalds's avatar
Linus Torvalds committed
4807
		return -EINVAL;
4808 4809
	if ((dl_policy(policy) && !__checkparam_dl(attr)) ||
	    (rt_policy(policy) != (attr->sched_priority != 0)))
Linus Torvalds's avatar
Linus Torvalds committed
4810 4811
		return -EINVAL;

4812 4813 4814
	/*
	 * Allow unprivileged RT tasks to decrease priority:
	 */
4815
	if (user && !capable(CAP_SYS_NICE)) {
4816
		if (fair_policy(policy)) {
4817
			if (attr->sched_nice < task_nice(p) &&
4818
			    !can_nice(p, attr->sched_nice))
4819 4820 4821
				return -EPERM;
		}

4822
		if (rt_policy(policy)) {
4823 4824
			unsigned long rlim_rtprio =
					task_rlimit(p, RLIMIT_RTPRIO);
4825

4826
			/* Can't set/change the rt policy: */
4827 4828 4829
			if (policy != p->policy && !rlim_rtprio)
				return -EPERM;

4830
			/* Can't increase priority: */
4831 4832
			if (attr->sched_priority > p->rt_priority &&
			    attr->sched_priority > rlim_rtprio)
4833 4834
				return -EPERM;
		}
4835

4836 4837 4838 4839 4840 4841 4842 4843 4844
		 /*
		  * 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
4845
		/*
4846 4847
		 * 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
4848
		 */
4849
		if (task_has_idle_policy(p) && !idle_policy(policy)) {
4850
			if (!can_nice(p, task_nice(p)))
4851 4852
				return -EPERM;
		}
4853

4854
		/* Can't change other user's priorities: */
4855
		if (!check_same_owner(p))
4856
			return -EPERM;
4857

4858
		/* Normal users shall not reset the sched_reset_on_fork flag: */
4859 4860
		if (p->sched_reset_on_fork && !reset_on_fork)
			return -EPERM;
4861
	}
Linus Torvalds's avatar
Linus Torvalds committed
4862

4863
	if (user) {
4864 4865 4866
		if (attr->sched_flags & SCHED_FLAG_SUGOV)
			return -EINVAL;

4867
		retval = security_task_setscheduler(p);
4868 4869 4870 4871
		if (retval)
			return retval;
	}

4872 4873 4874 4875 4876 4877 4878
	/* Update task specific "requested" clamps */
	if (attr->sched_flags & SCHED_FLAG_UTIL_CLAMP) {
		retval = uclamp_validate(p, attr);
		if (retval)
			return retval;
	}

4879 4880 4881
	if (pi)
		cpuset_read_lock();

4882
	/*
4883
	 * Make sure no PI-waiters arrive (or leave) while we are
4884
	 * changing the priority of the task:
4885
	 *
Lucas De Marchi's avatar
Lucas De Marchi committed
4886
	 * To be able to change p->policy safely, the appropriate
Linus Torvalds's avatar
Linus Torvalds committed
4887 4888
	 * runqueue lock must be held.
	 */
4889
	rq = task_rq_lock(p, &rf);
4890
	update_rq_clock(rq);
4891

4892
	/*
4893
	 * Changing the policy of the stop threads its a very bad idea:
4894 4895
	 */
	if (p == rq->stop) {
4896 4897
		retval = -EINVAL;
		goto unlock;
4898 4899
	}

4900
	/*
4901 4902
	 * If not changing anything there's no need to proceed further,
	 * but store a possible modification of reset_on_fork.
4903
	 */
4904
	if (unlikely(policy == p->policy)) {
4905
		if (fair_policy(policy) && attr->sched_nice != task_nice(p))
4906 4907 4908
			goto change;
		if (rt_policy(policy) && attr->sched_priority != p->rt_priority)
			goto change;
4909
		if (dl_policy(policy) && dl_param_changed(p, attr))
4910
			goto change;
4911 4912
		if (attr->sched_flags & SCHED_FLAG_UTIL_CLAMP)
			goto change;
4913

4914
		p->sched_reset_on_fork = reset_on_fork;
4915 4916
		retval = 0;
		goto unlock;
4917
	}
4918
change:
4919

4920
	if (user) {
4921
#ifdef CONFIG_RT_GROUP_SCHED
4922 4923 4924 4925 4926
		/*
		 * Do not allow realtime tasks into groups that have no runtime
		 * assigned.
		 */
		if (rt_bandwidth_enabled() && rt_policy(policy) &&
4927 4928
				task_group(p)->rt_bandwidth.rt_runtime == 0 &&
				!task_group_is_autogroup(task_group(p))) {
4929 4930
			retval = -EPERM;
			goto unlock;
4931 4932
		}
#endif
4933
#ifdef CONFIG_SMP
4934 4935
		if (dl_bandwidth_enabled() && dl_policy(policy) &&
				!(attr->sched_flags & SCHED_FLAG_SUGOV)) {
4936 4937 4938 4939 4940 4941 4942
			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.
			 */
4943
			if (!cpumask_subset(span, p->cpus_ptr) ||
4944
			    rq->rd->dl_bw.bw == 0) {
4945 4946
				retval = -EPERM;
				goto unlock;
4947 4948 4949 4950
			}
		}
#endif
	}
4951

4952
	/* Re-check policy now with rq lock held: */
Linus Torvalds's avatar
Linus Torvalds committed
4953 4954
	if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) {
		policy = oldpolicy = -1;
4955
		task_rq_unlock(rq, p, &rf);
4956 4957
		if (pi)
			cpuset_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
4958 4959
		goto recheck;
	}
4960 4961 4962 4963 4964 4965

	/*
	 * If setscheduling to SCHED_DEADLINE (or changing the parameters
	 * of a SCHED_DEADLINE task) we need to check if enough bandwidth
	 * is available.
	 */
4966
	if ((dl_policy(policy) || dl_task(p)) && sched_dl_overflow(p, policy, attr)) {
4967 4968
		retval = -EBUSY;
		goto unlock;
4969 4970
	}

4971 4972 4973
	p->sched_reset_on_fork = reset_on_fork;
	oldprio = p->prio;

4974 4975 4976 4977 4978 4979 4980 4981
	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.
		 */
4982
		new_effective_prio = rt_effective_prio(p, newprio);
4983 4984
		if (new_effective_prio == oldprio)
			queue_flags &= ~DEQUEUE_MOVE;
4985 4986
	}

4987
	queued = task_on_rq_queued(p);
4988
	running = task_current(rq, p);
4989
	if (queued)
4990
		dequeue_task(rq, p, queue_flags);
4991
	if (running)
4992
		put_prev_task(rq, p);
4993

4994
	prev_class = p->sched_class;
4995

4996
	__setscheduler(rq, p, attr, pi);
4997
	__setscheduler_uclamp(p, attr);
4998

4999
	if (queued) {
5000 5001 5002 5003
		/*
		 * We enqueue to tail when the priority of a task is
		 * increased (user space view).
		 */
5004 5005
		if (oldprio < p->prio)
			queue_flags |= ENQUEUE_HEAD;
5006

5007
		enqueue_task(rq, p, queue_flags);
5008
	}
5009
	if (running)
5010
		set_next_task(rq, p);
5011

5012
	check_class_changed(rq, p, prev_class, oldprio);
5013 5014 5015

	/* Avoid rq from going away on us: */
	preempt_disable();
5016
	task_rq_unlock(rq, p, &rf);
5017

5018 5019
	if (pi) {
		cpuset_read_unlock();
5020
		rt_mutex_adjust_pi(p);
5021
	}
5022

5023
	/* Run balance callbacks after we've adjusted the PI chain: */
5024 5025
	balance_callback(rq);
	preempt_enable();
5026

Linus Torvalds's avatar
Linus Torvalds committed
5027
	return 0;
5028 5029 5030

unlock:
	task_rq_unlock(rq, p, &rf);
5031 5032
	if (pi)
		cpuset_read_unlock();
5033
	return retval;
Linus Torvalds's avatar
Linus Torvalds committed
5034
}
5035

5036 5037 5038 5039 5040 5041 5042 5043 5044
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),
	};

5045 5046
	/* Fixup the legacy SCHED_RESET_ON_FORK hack. */
	if ((policy != SETPARAM_POLICY) && (policy & SCHED_RESET_ON_FORK)) {
5047 5048 5049 5050 5051
		attr.sched_flags |= SCHED_FLAG_RESET_ON_FORK;
		policy &= ~SCHED_RESET_ON_FORK;
		attr.sched_policy = policy;
	}

5052
	return __sched_setscheduler(p, &attr, check, true);
5053
}
5054 5055 5056 5057 5058 5059
/**
 * 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.
 *
5060 5061
 * Return: 0 on success. An error code otherwise.
 *
5062 5063 5064
 * NOTE that the task may be already dead.
 */
int sched_setscheduler(struct task_struct *p, int policy,
5065
		       const struct sched_param *param)
5066
{
5067
	return _sched_setscheduler(p, policy, param, true);
5068
}
Linus Torvalds's avatar
Linus Torvalds committed
5069 5070
EXPORT_SYMBOL_GPL(sched_setscheduler);

5071 5072
int sched_setattr(struct task_struct *p, const struct sched_attr *attr)
{
5073
	return __sched_setscheduler(p, attr, true, true);
5074 5075 5076
}
EXPORT_SYMBOL_GPL(sched_setattr);

5077 5078 5079 5080 5081
int sched_setattr_nocheck(struct task_struct *p, const struct sched_attr *attr)
{
	return __sched_setscheduler(p, attr, false, true);
}

5082 5083 5084 5085 5086 5087 5088 5089 5090 5091
/**
 * 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.
5092 5093
 *
 * Return: 0 on success. An error code otherwise.
5094 5095
 */
int sched_setscheduler_nocheck(struct task_struct *p, int policy,
5096
			       const struct sched_param *param)
5097
{
5098
	return _sched_setscheduler(p, policy, param, false);
5099
}
5100
EXPORT_SYMBOL_GPL(sched_setscheduler_nocheck);
5101

Ingo Molnar's avatar
Ingo Molnar committed
5102 5103
static int
do_sched_setscheduler(pid_t pid, int policy, struct sched_param __user *param)
Linus Torvalds's avatar
Linus Torvalds committed
5104 5105 5106
{
	struct sched_param lparam;
	struct task_struct *p;
5107
	int retval;
Linus Torvalds's avatar
Linus Torvalds committed
5108 5109 5110 5111 5112

	if (!param || pid < 0)
		return -EINVAL;
	if (copy_from_user(&lparam, param, sizeof(struct sched_param)))
		return -EFAULT;
5113 5114 5115

	rcu_read_lock();
	retval = -ESRCH;
Linus Torvalds's avatar
Linus Torvalds committed
5116
	p = find_process_by_pid(pid);
5117 5118
	if (likely(p))
		get_task_struct(p);
5119
	rcu_read_unlock();
5120

5121 5122 5123 5124 5125
	if (likely(p)) {
		retval = sched_setscheduler(p, policy, &lparam);
		put_task_struct(p);
	}

Linus Torvalds's avatar
Linus Torvalds committed
5126 5127 5128
	return retval;
}

5129 5130 5131
/*
 * Mimics kernel/events/core.c perf_copy_attr().
 */
5132
static int sched_copy_attr(struct sched_attr __user *uattr, struct sched_attr *attr)
5133 5134 5135 5136
{
	u32 size;
	int ret;

5137
	/* Zero the full structure, so that a short copy will be nice: */
5138 5139 5140 5141 5142 5143
	memset(attr, 0, sizeof(*attr));

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

5144 5145
	/* ABI compatibility quirk: */
	if (!size)
5146
		size = SCHED_ATTR_SIZE_VER0;
5147
	if (size < SCHED_ATTR_SIZE_VER0 || size > PAGE_SIZE)
5148 5149
		goto err_size;

5150 5151 5152 5153 5154
	ret = copy_struct_from_user(attr, sizeof(*attr), uattr, size);
	if (ret) {
		if (ret == -E2BIG)
			goto err_size;
		return ret;
5155 5156
	}

5157 5158 5159 5160
	if ((attr->sched_flags & SCHED_FLAG_UTIL_CLAMP) &&
	    size < SCHED_ATTR_SIZE_VER1)
		return -EINVAL;

5161
	/*
5162
	 * XXX: Do we want to be lenient like existing syscalls; or do we want
5163 5164
	 * to be strict and return an error on out-of-bounds values?
	 */
5165
	attr->sched_nice = clamp(attr->sched_nice, MIN_NICE, MAX_NICE);
5166

5167
	return 0;
5168 5169 5170

err_size:
	put_user(sizeof(*attr), &uattr->size);
5171
	return -E2BIG;
5172 5173
}

Linus Torvalds's avatar
Linus Torvalds committed
5174 5175 5176 5177 5178
/**
 * 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.
5179 5180
 *
 * Return: 0 on success. An error code otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
5181
 */
5182
SYSCALL_DEFINE3(sched_setscheduler, pid_t, pid, int, policy, struct sched_param __user *, param)
Linus Torvalds's avatar
Linus Torvalds committed
5183
{
5184 5185 5186
	if (policy < 0)
		return -EINVAL;

Linus Torvalds's avatar
Linus Torvalds committed
5187 5188 5189 5190 5191 5192 5193
	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.
5194 5195
 *
 * Return: 0 on success. An error code otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
5196
 */
5197
SYSCALL_DEFINE2(sched_setparam, pid_t, pid, struct sched_param __user *, param)
Linus Torvalds's avatar
Linus Torvalds committed
5198
{
5199
	return do_sched_setscheduler(pid, SETPARAM_POLICY, param);
Linus Torvalds's avatar
Linus Torvalds committed
5200 5201
}

5202 5203 5204
/**
 * sys_sched_setattr - same as above, but with extended sched_attr
 * @pid: the pid in question.
5205
 * @uattr: structure containing the extended parameters.
5206
 * @flags: for future extension.
5207
 */
5208 5209
SYSCALL_DEFINE3(sched_setattr, pid_t, pid, struct sched_attr __user *, uattr,
			       unsigned int, flags)
5210 5211 5212 5213 5214
{
	struct sched_attr attr;
	struct task_struct *p;
	int retval;

5215
	if (!uattr || pid < 0 || flags)
5216 5217
		return -EINVAL;

5218 5219 5220
	retval = sched_copy_attr(uattr, &attr);
	if (retval)
		return retval;
5221

5222
	if ((int)attr.sched_policy < 0)
5223
		return -EINVAL;
5224 5225
	if (attr.sched_flags & SCHED_FLAG_KEEP_POLICY)
		attr.sched_policy = SETPARAM_POLICY;
5226 5227 5228 5229

	rcu_read_lock();
	retval = -ESRCH;
	p = find_process_by_pid(pid);
5230 5231
	if (likely(p))
		get_task_struct(p);
5232 5233
	rcu_read_unlock();

5234 5235 5236 5237 5238
	if (likely(p)) {
		retval = sched_setattr(p, &attr);
		put_task_struct(p);
	}

5239 5240 5241
	return retval;
}

Linus Torvalds's avatar
Linus Torvalds committed
5242 5243 5244
/**
 * sys_sched_getscheduler - get the policy (scheduling class) of a thread
 * @pid: the pid in question.
5245 5246 5247
 *
 * Return: On success, the policy of the thread. Otherwise, a negative error
 * code.
Linus Torvalds's avatar
Linus Torvalds committed
5248
 */
5249
SYSCALL_DEFINE1(sched_getscheduler, pid_t, pid)
Linus Torvalds's avatar
Linus Torvalds committed
5250
{
5251
	struct task_struct *p;
5252
	int retval;
Linus Torvalds's avatar
Linus Torvalds committed
5253 5254

	if (pid < 0)
5255
		return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
5256 5257

	retval = -ESRCH;
5258
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
5259 5260 5261 5262
	p = find_process_by_pid(pid);
	if (p) {
		retval = security_task_getscheduler(p);
		if (!retval)
5263 5264
			retval = p->policy
				| (p->sched_reset_on_fork ? SCHED_RESET_ON_FORK : 0);
Linus Torvalds's avatar
Linus Torvalds committed
5265
	}
5266
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
5267 5268 5269 5270
	return retval;
}

/**
5271
 * sys_sched_getparam - get the RT priority of a thread
Linus Torvalds's avatar
Linus Torvalds committed
5272 5273
 * @pid: the pid in question.
 * @param: structure containing the RT priority.
5274 5275 5276
 *
 * Return: On success, 0 and the RT priority is in @param. Otherwise, an error
 * code.
Linus Torvalds's avatar
Linus Torvalds committed
5277
 */
5278
SYSCALL_DEFINE2(sched_getparam, pid_t, pid, struct sched_param __user *, param)
Linus Torvalds's avatar
Linus Torvalds committed
5279
{
5280
	struct sched_param lp = { .sched_priority = 0 };
5281
	struct task_struct *p;
5282
	int retval;
Linus Torvalds's avatar
Linus Torvalds committed
5283 5284

	if (!param || pid < 0)
5285
		return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
5286

5287
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
5288 5289 5290 5291 5292 5293 5294 5295 5296
	p = find_process_by_pid(pid);
	retval = -ESRCH;
	if (!p)
		goto out_unlock;

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

5297 5298
	if (task_has_rt_policy(p))
		lp.sched_priority = p->rt_priority;
5299
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
5300 5301 5302 5303 5304 5305 5306 5307 5308

	/*
	 * 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:
5309
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
5310 5311 5312
	return retval;
}

5313 5314 5315 5316 5317 5318 5319 5320 5321 5322 5323 5324
/*
 * Copy the kernel size attribute structure (which might be larger
 * than what user-space knows about) to user-space.
 *
 * Note that all cases are valid: user-space buffer can be larger or
 * smaller than the kernel-space buffer. The usual case is that both
 * have the same size.
 */
static int
sched_attr_copy_to_user(struct sched_attr __user *uattr,
			struct sched_attr *kattr,
			unsigned int usize)
5325
{
5326
	unsigned int ksize = sizeof(*kattr);
5327

5328
	if (!access_ok(uattr, usize))
5329 5330 5331
		return -EFAULT;

	/*
5332 5333 5334 5335 5336 5337 5338 5339 5340 5341 5342
	 * sched_getattr() ABI forwards and backwards compatibility:
	 *
	 * If usize == ksize then we just copy everything to user-space and all is good.
	 *
	 * If usize < ksize then we only copy as much as user-space has space for,
	 * this keeps ABI compatibility as well. We skip the rest.
	 *
	 * If usize > ksize then user-space is using a newer version of the ABI,
	 * which part the kernel doesn't know about. Just ignore it - tooling can
	 * detect the kernel's knowledge of attributes from the attr->size value
	 * which is set to ksize in this case.
5343
	 */
5344
	kattr->size = min(usize, ksize);
5345

5346
	if (copy_to_user(uattr, kattr, kattr->size))
5347 5348
		return -EFAULT;

5349
	return 0;
5350 5351 5352
}

/**
5353
 * sys_sched_getattr - similar to sched_getparam, but with sched_attr
5354
 * @pid: the pid in question.
5355
 * @uattr: structure containing the extended parameters.
5356
 * @usize: sizeof(attr) for fwd/bwd comp.
5357
 * @flags: for future extension.
5358
 */
5359
SYSCALL_DEFINE4(sched_getattr, pid_t, pid, struct sched_attr __user *, uattr,
5360
		unsigned int, usize, unsigned int, flags)
5361
{
5362
	struct sched_attr kattr = { };
5363 5364 5365
	struct task_struct *p;
	int retval;

5366 5367
	if (!uattr || pid < 0 || usize > PAGE_SIZE ||
	    usize < SCHED_ATTR_SIZE_VER0 || flags)
5368 5369 5370 5371 5372 5373 5374 5375 5376 5377 5378 5379
		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;

5380
	kattr.sched_policy = p->policy;
5381
	if (p->sched_reset_on_fork)
5382
		kattr.sched_flags |= SCHED_FLAG_RESET_ON_FORK;
5383
	if (task_has_dl_policy(p))
5384
		__getparam_dl(p, &kattr);
5385
	else if (task_has_rt_policy(p))
5386
		kattr.sched_priority = p->rt_priority;
5387
	else
5388
		kattr.sched_nice = task_nice(p);
5389

5390
#ifdef CONFIG_UCLAMP_TASK
5391 5392
	kattr.sched_util_min = p->uclamp_req[UCLAMP_MIN].value;
	kattr.sched_util_max = p->uclamp_req[UCLAMP_MAX].value;
5393 5394
#endif

5395 5396
	rcu_read_unlock();

5397
	return sched_attr_copy_to_user(uattr, &kattr, usize);
5398 5399 5400 5401 5402 5403

out_unlock:
	rcu_read_unlock();
	return retval;
}

5404
long sched_setaffinity(pid_t pid, const struct cpumask *in_mask)
Linus Torvalds's avatar
Linus Torvalds committed
5405
{
5406
	cpumask_var_t cpus_allowed, new_mask;
5407 5408
	struct task_struct *p;
	int retval;
Linus Torvalds's avatar
Linus Torvalds committed
5409

5410
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
5411 5412 5413

	p = find_process_by_pid(pid);
	if (!p) {
5414
		rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
5415 5416 5417
		return -ESRCH;
	}

5418
	/* Prevent p going away */
Linus Torvalds's avatar
Linus Torvalds committed
5419
	get_task_struct(p);
5420
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
5421

5422 5423 5424 5425
	if (p->flags & PF_NO_SETAFFINITY) {
		retval = -EINVAL;
		goto out_put_task;
	}
5426 5427 5428 5429 5430 5431 5432 5433
	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
5434
	retval = -EPERM;
5435 5436 5437 5438
	if (!check_same_owner(p)) {
		rcu_read_lock();
		if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
			rcu_read_unlock();
5439
			goto out_free_new_mask;
5440 5441 5442
		}
		rcu_read_unlock();
	}
Linus Torvalds's avatar
Linus Torvalds committed
5443

5444
	retval = security_task_setscheduler(p);
5445
	if (retval)
5446
		goto out_free_new_mask;
5447

5448 5449 5450 5451

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

5452 5453 5454 5455 5456 5457 5458
	/*
	 * 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
5459 5460 5461
	if (task_has_dl_policy(p) && dl_bandwidth_enabled()) {
		rcu_read_lock();
		if (!cpumask_subset(task_rq(p)->rd->span, new_mask)) {
5462
			retval = -EBUSY;
5463
			rcu_read_unlock();
5464
			goto out_free_new_mask;
5465
		}
5466
		rcu_read_unlock();
5467 5468
	}
#endif
Peter Zijlstra's avatar
Peter Zijlstra committed
5469
again:
5470
	retval = __set_cpus_allowed_ptr(p, new_mask, true);
Linus Torvalds's avatar
Linus Torvalds committed
5471

Paul Menage's avatar
Paul Menage committed
5472
	if (!retval) {
5473 5474
		cpuset_cpus_allowed(p, cpus_allowed);
		if (!cpumask_subset(new_mask, cpus_allowed)) {
Paul Menage's avatar
Paul Menage committed
5475 5476 5477 5478 5479
			/*
			 * We must have raced with a concurrent cpuset
			 * update. Just reset the cpus_allowed to the
			 * cpuset's cpus_allowed
			 */
5480
			cpumask_copy(new_mask, cpus_allowed);
Paul Menage's avatar
Paul Menage committed
5481 5482 5483
			goto again;
		}
	}
5484
out_free_new_mask:
5485 5486 5487 5488
	free_cpumask_var(new_mask);
out_free_cpus_allowed:
	free_cpumask_var(cpus_allowed);
out_put_task:
Linus Torvalds's avatar
Linus Torvalds committed
5489 5490 5491 5492 5493
	put_task_struct(p);
	return retval;
}

static int get_user_cpu_mask(unsigned long __user *user_mask_ptr, unsigned len,
5494
			     struct cpumask *new_mask)
Linus Torvalds's avatar
Linus Torvalds committed
5495
{
5496 5497 5498 5499 5500
	if (len < cpumask_size())
		cpumask_clear(new_mask);
	else if (len > cpumask_size())
		len = cpumask_size();

Linus Torvalds's avatar
Linus Torvalds committed
5501 5502 5503 5504
	return copy_from_user(new_mask, user_mask_ptr, len) ? -EFAULT : 0;
}

/**
5505
 * sys_sched_setaffinity - set the CPU affinity of a process
Linus Torvalds's avatar
Linus Torvalds committed
5506 5507
 * @pid: pid of the process
 * @len: length in bytes of the bitmask pointed to by user_mask_ptr
5508
 * @user_mask_ptr: user-space pointer to the new CPU mask
5509 5510
 *
 * Return: 0 on success. An error code otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
5511
 */
5512 5513
SYSCALL_DEFINE3(sched_setaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
Linus Torvalds's avatar
Linus Torvalds committed
5514
{
5515
	cpumask_var_t new_mask;
Linus Torvalds's avatar
Linus Torvalds committed
5516 5517
	int retval;

5518 5519
	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL))
		return -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
5520

5521 5522 5523 5524 5525
	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
5526 5527
}

5528
long sched_getaffinity(pid_t pid, struct cpumask *mask)
Linus Torvalds's avatar
Linus Torvalds committed
5529
{
5530
	struct task_struct *p;
5531
	unsigned long flags;
Linus Torvalds's avatar
Linus Torvalds committed
5532 5533
	int retval;

5534
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
5535 5536 5537 5538 5539 5540

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

5541 5542 5543 5544
	retval = security_task_getscheduler(p);
	if (retval)
		goto out_unlock;

5545
	raw_spin_lock_irqsave(&p->pi_lock, flags);
5546
	cpumask_and(mask, &p->cpus_mask, cpu_active_mask);
5547
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
5548 5549

out_unlock:
5550
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
5551

5552
	return retval;
Linus Torvalds's avatar
Linus Torvalds committed
5553 5554 5555
}

/**
5556
 * sys_sched_getaffinity - get the CPU affinity of a process
Linus Torvalds's avatar
Linus Torvalds committed
5557 5558
 * @pid: pid of the process
 * @len: length in bytes of the bitmask pointed to by user_mask_ptr
5559
 * @user_mask_ptr: user-space pointer to hold the current CPU mask
5560
 *
5561 5562
 * Return: size of CPU mask copied to user_mask_ptr on success. An
 * error code otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
5563
 */
5564 5565
SYSCALL_DEFINE3(sched_getaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
Linus Torvalds's avatar
Linus Torvalds committed
5566 5567
{
	int ret;
5568
	cpumask_var_t mask;
Linus Torvalds's avatar
Linus Torvalds committed
5569

5570
	if ((len * BITS_PER_BYTE) < nr_cpu_ids)
5571 5572
		return -EINVAL;
	if (len & (sizeof(unsigned long)-1))
Linus Torvalds's avatar
Linus Torvalds committed
5573 5574
		return -EINVAL;

5575 5576
	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
		return -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
5577

5578 5579
	ret = sched_getaffinity(pid, mask);
	if (ret == 0) {
5580
		unsigned int retlen = min(len, cpumask_size());
5581 5582

		if (copy_to_user(user_mask_ptr, mask, retlen))
5583 5584
			ret = -EFAULT;
		else
5585
			ret = retlen;
5586 5587
	}
	free_cpumask_var(mask);
Linus Torvalds's avatar
Linus Torvalds committed
5588

5589
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
5590 5591 5592 5593 5594
}

/**
 * sys_sched_yield - yield the current processor to other threads.
 *
Ingo Molnar's avatar
Ingo Molnar committed
5595 5596
 * This function yields the current CPU to other tasks. If there are no
 * other threads running on this CPU then this function will return.
5597 5598
 *
 * Return: 0.
Linus Torvalds's avatar
Linus Torvalds committed
5599
 */
5600
static void do_sched_yield(void)
Linus Torvalds's avatar
Linus Torvalds committed
5601
{
5602 5603 5604
	struct rq_flags rf;
	struct rq *rq;

5605
	rq = this_rq_lock_irq(&rf);
Linus Torvalds's avatar
Linus Torvalds committed
5606

5607
	schedstat_inc(rq->yld_count);
5608
	current->sched_class->yield_task(rq);
Linus Torvalds's avatar
Linus Torvalds committed
5609 5610 5611 5612 5613

	/*
	 * Since we are going to call schedule() anyway, there's
	 * no need to preempt or enable interrupts:
	 */
5614 5615
	preempt_disable();
	rq_unlock(rq, &rf);
5616
	sched_preempt_enable_no_resched();
Linus Torvalds's avatar
Linus Torvalds committed
5617 5618

	schedule();
5619
}
Linus Torvalds's avatar
Linus Torvalds committed
5620

5621 5622 5623
SYSCALL_DEFINE0(sched_yield)
{
	do_sched_yield();
Linus Torvalds's avatar
Linus Torvalds committed
5624 5625 5626
	return 0;
}

5627
#ifndef CONFIG_PREEMPTION
5628
int __sched _cond_resched(void)
Linus Torvalds's avatar
Linus Torvalds committed
5629
{
5630
	if (should_resched(0)) {
5631
		preempt_schedule_common();
Linus Torvalds's avatar
Linus Torvalds committed
5632 5633
		return 1;
	}
5634
	rcu_all_qs();
Linus Torvalds's avatar
Linus Torvalds committed
5635 5636
	return 0;
}
5637
EXPORT_SYMBOL(_cond_resched);
5638
#endif
Linus Torvalds's avatar
Linus Torvalds committed
5639 5640

/*
5641
 * __cond_resched_lock() - if a reschedule is pending, drop the given lock,
Linus Torvalds's avatar
Linus Torvalds committed
5642 5643
 * call schedule, and on return reacquire the lock.
 *
5644
 * This works OK both with and without CONFIG_PREEMPTION. We do strange low-level
Linus Torvalds's avatar
Linus Torvalds committed
5645 5646 5647
 * operations here to prevent schedule() from being called twice (once via
 * spin_unlock(), once by hand).
 */
5648
int __cond_resched_lock(spinlock_t *lock)
Linus Torvalds's avatar
Linus Torvalds committed
5649
{
5650
	int resched = should_resched(PREEMPT_LOCK_OFFSET);
Jan Kara's avatar
Jan Kara committed
5651 5652
	int ret = 0;

5653 5654
	lockdep_assert_held(lock);

5655
	if (spin_needbreak(lock) || resched) {
Linus Torvalds's avatar
Linus Torvalds committed
5656
		spin_unlock(lock);
5657
		if (resched)
5658
			preempt_schedule_common();
Nick Piggin's avatar
Nick Piggin committed
5659 5660
		else
			cpu_relax();
Jan Kara's avatar
Jan Kara committed
5661
		ret = 1;
Linus Torvalds's avatar
Linus Torvalds committed
5662 5663
		spin_lock(lock);
	}
Jan Kara's avatar
Jan Kara committed
5664
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
5665
}
5666
EXPORT_SYMBOL(__cond_resched_lock);
Linus Torvalds's avatar
Linus Torvalds committed
5667 5668 5669 5670

/**
 * yield - yield the current processor to other threads.
 *
Peter Zijlstra's avatar
Peter Zijlstra committed
5671 5672 5673 5674 5675 5676 5677 5678 5679
 * 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)
5680
 *	yield();
Peter Zijlstra's avatar
Peter Zijlstra committed
5681 5682 5683 5684 5685 5686 5687 5688
 *
 * 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
5689 5690 5691 5692
 */
void __sched yield(void)
{
	set_current_state(TASK_RUNNING);
5693
	do_sched_yield();
Linus Torvalds's avatar
Linus Torvalds committed
5694 5695 5696
}
EXPORT_SYMBOL(yield);

5697 5698 5699 5700
/**
 * yield_to - yield the current processor to another thread in
 * your thread group, or accelerate that thread toward the
 * processor it's on.
5701 5702
 * @p: target task
 * @preempt: whether task preemption is allowed or not
5703 5704 5705 5706
 *
 * 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.
 *
5707
 * Return:
5708 5709 5710
 *	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.
5711
 */
5712
int __sched yield_to(struct task_struct *p, bool preempt)
5713 5714 5715 5716
{
	struct task_struct *curr = current;
	struct rq *rq, *p_rq;
	unsigned long flags;
5717
	int yielded = 0;
5718 5719 5720 5721 5722 5723

	local_irq_save(flags);
	rq = this_rq();

again:
	p_rq = task_rq(p);
5724 5725 5726 5727 5728 5729 5730 5731 5732
	/*
	 * 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;
	}

5733
	double_rq_lock(rq, p_rq);
5734
	if (task_rq(p) != p_rq) {
5735 5736 5737 5738 5739
		double_rq_unlock(rq, p_rq);
		goto again;
	}

	if (!curr->sched_class->yield_to_task)
5740
		goto out_unlock;
5741 5742

	if (curr->sched_class != p->sched_class)
5743
		goto out_unlock;
5744 5745

	if (task_running(p_rq, p) || p->state)
5746
		goto out_unlock;
5747 5748

	yielded = curr->sched_class->yield_to_task(rq, p, preempt);
5749
	if (yielded) {
5750
		schedstat_inc(rq->yld_count);
5751 5752 5753 5754 5755
		/*
		 * Make p's CPU reschedule; pick_next_entity takes care of
		 * fairness.
		 */
		if (preempt && rq != p_rq)
5756
			resched_curr(p_rq);
5757
	}
5758

5759
out_unlock:
5760
	double_rq_unlock(rq, p_rq);
5761
out_irq:
5762 5763
	local_irq_restore(flags);

5764
	if (yielded > 0)
5765 5766 5767 5768 5769 5770
		schedule();

	return yielded;
}
EXPORT_SYMBOL_GPL(yield_to);

5771 5772 5773 5774 5775 5776 5777 5778 5779 5780 5781 5782 5783 5784 5785
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
5786
/*
Ingo Molnar's avatar
Ingo Molnar committed
5787
 * This task is about to go to sleep on IO. Increment rq->nr_iowait so
Linus Torvalds's avatar
Linus Torvalds committed
5788 5789 5790 5791
 * that process accounting knows that this is a task in IO wait state.
 */
long __sched io_schedule_timeout(long timeout)
{
5792
	int token;
Linus Torvalds's avatar
Linus Torvalds committed
5793 5794
	long ret;

5795
	token = io_schedule_prepare();
Linus Torvalds's avatar
Linus Torvalds committed
5796
	ret = schedule_timeout(timeout);
5797
	io_schedule_finish(token);
5798

Linus Torvalds's avatar
Linus Torvalds committed
5799 5800
	return ret;
}
5801
EXPORT_SYMBOL(io_schedule_timeout);
Linus Torvalds's avatar
Linus Torvalds committed
5802

5803
void __sched io_schedule(void)
5804 5805 5806 5807 5808 5809 5810 5811 5812
{
	int token;

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

Linus Torvalds's avatar
Linus Torvalds committed
5813 5814 5815 5816
/**
 * sys_sched_get_priority_max - return maximum RT priority.
 * @policy: scheduling class.
 *
5817 5818 5819
 * 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
5820
 */
5821
SYSCALL_DEFINE1(sched_get_priority_max, int, policy)
Linus Torvalds's avatar
Linus Torvalds committed
5822 5823 5824 5825 5826 5827 5828 5829
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = MAX_USER_RT_PRIO-1;
		break;
5830
	case SCHED_DEADLINE:
Linus Torvalds's avatar
Linus Torvalds committed
5831
	case SCHED_NORMAL:
5832
	case SCHED_BATCH:
Ingo Molnar's avatar
Ingo Molnar committed
5833
	case SCHED_IDLE:
Linus Torvalds's avatar
Linus Torvalds committed
5834 5835 5836 5837 5838 5839 5840 5841 5842 5843
		ret = 0;
		break;
	}
	return ret;
}

/**
 * sys_sched_get_priority_min - return minimum RT priority.
 * @policy: scheduling class.
 *
5844 5845 5846
 * 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
5847
 */
5848
SYSCALL_DEFINE1(sched_get_priority_min, int, policy)
Linus Torvalds's avatar
Linus Torvalds committed
5849 5850 5851 5852 5853 5854 5855 5856
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = 1;
		break;
5857
	case SCHED_DEADLINE:
Linus Torvalds's avatar
Linus Torvalds committed
5858
	case SCHED_NORMAL:
5859
	case SCHED_BATCH:
Ingo Molnar's avatar
Ingo Molnar committed
5860
	case SCHED_IDLE:
Linus Torvalds's avatar
Linus Torvalds committed
5861 5862 5863 5864 5865
		ret = 0;
	}
	return ret;
}

5866
static int sched_rr_get_interval(pid_t pid, struct timespec64 *t)
Linus Torvalds's avatar
Linus Torvalds committed
5867
{
5868
	struct task_struct *p;
Dmitry Adamushko's avatar
Dmitry Adamushko committed
5869
	unsigned int time_slice;
5870
	struct rq_flags rf;
5871
	struct rq *rq;
5872
	int retval;
Linus Torvalds's avatar
Linus Torvalds committed
5873 5874

	if (pid < 0)
5875
		return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
5876 5877

	retval = -ESRCH;
5878
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
5879 5880 5881 5882 5883 5884 5885 5886
	p = find_process_by_pid(pid);
	if (!p)
		goto out_unlock;

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

5887
	rq = task_rq_lock(p, &rf);
5888 5889 5890
	time_slice = 0;
	if (p->sched_class->get_rr_interval)
		time_slice = p->sched_class->get_rr_interval(rq, p);
5891
	task_rq_unlock(rq, p, &rf);
Dmitry Adamushko's avatar
Dmitry Adamushko committed
5892

5893
	rcu_read_unlock();
5894 5895
	jiffies_to_timespec64(time_slice, t);
	return 0;
5896

Linus Torvalds's avatar
Linus Torvalds committed
5897
out_unlock:
5898
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
5899 5900 5901
	return retval;
}

5902 5903 5904 5905 5906 5907 5908 5909 5910 5911 5912
/**
 * 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.
 */
5913
SYSCALL_DEFINE2(sched_rr_get_interval, pid_t, pid,
5914
		struct __kernel_timespec __user *, interval)
5915 5916 5917 5918 5919 5920 5921 5922 5923 5924
{
	struct timespec64 t;
	int retval = sched_rr_get_interval(pid, &t);

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

	return retval;
}

5925
#ifdef CONFIG_COMPAT_32BIT_TIME
5926 5927
SYSCALL_DEFINE2(sched_rr_get_interval_time32, pid_t, pid,
		struct old_timespec32 __user *, interval)
5928 5929 5930 5931 5932
{
	struct timespec64 t;
	int retval = sched_rr_get_interval(pid, &t);

	if (retval == 0)
5933
		retval = put_old_timespec32(&t, interval);
5934 5935 5936 5937
	return retval;
}
#endif

5938
void sched_show_task(struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
5939 5940
{
	unsigned long free = 0;
5941
	int ppid;
5942

5943 5944
	if (!try_get_task_stack(p))
		return;
5945 5946 5947 5948

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

	if (p->state == TASK_RUNNING)
5949
		printk(KERN_CONT "  running task    ");
Linus Torvalds's avatar
Linus Torvalds committed
5950
#ifdef CONFIG_DEBUG_STACK_USAGE
5951
	free = stack_not_used(p);
Linus Torvalds's avatar
Linus Torvalds committed
5952
#endif
5953
	ppid = 0;
5954
	rcu_read_lock();
5955 5956
	if (pid_alive(p))
		ppid = task_pid_nr(rcu_dereference(p->real_parent));
5957
	rcu_read_unlock();
5958
	printk(KERN_CONT "%5lu %5d %6d 0x%08lx\n", free,
5959
		task_pid_nr(p), ppid,
5960
		(unsigned long)task_thread_info(p)->flags);
Linus Torvalds's avatar
Linus Torvalds committed
5961

5962
	print_worker_info(KERN_INFO, p);
5963
	show_stack(p, NULL);
5964
	put_task_stack(p);
Linus Torvalds's avatar
Linus Torvalds committed
5965
}
5966
EXPORT_SYMBOL_GPL(sched_show_task);
Linus Torvalds's avatar
Linus Torvalds committed
5967

5968 5969 5970 5971 5972 5973 5974 5975 5976 5977 5978 5979 5980 5981 5982 5983 5984 5985 5986 5987 5988 5989
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;
}


5990
void show_state_filter(unsigned long state_filter)
Linus Torvalds's avatar
Linus Torvalds committed
5991
{
5992
	struct task_struct *g, *p;
Linus Torvalds's avatar
Linus Torvalds committed
5993

5994
#if BITS_PER_LONG == 32
5995 5996
	printk(KERN_INFO
		"  task                PC stack   pid father\n");
Linus Torvalds's avatar
Linus Torvalds committed
5997
#else
5998 5999
	printk(KERN_INFO
		"  task                        PC stack   pid father\n");
Linus Torvalds's avatar
Linus Torvalds committed
6000
#endif
6001
	rcu_read_lock();
6002
	for_each_process_thread(g, p) {
Linus Torvalds's avatar
Linus Torvalds committed
6003 6004
		/*
		 * reset the NMI-timeout, listing all files on a slow
Lucas De Marchi's avatar
Lucas De Marchi committed
6005
		 * console might take a lot of time:
6006 6007 6008
		 * 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
6009 6010
		 */
		touch_nmi_watchdog();
6011
		touch_all_softlockup_watchdogs();
6012
		if (state_filter_match(state_filter, p))
6013
			sched_show_task(p);
6014
	}
Linus Torvalds's avatar
Linus Torvalds committed
6015

Ingo Molnar's avatar
Ingo Molnar committed
6016
#ifdef CONFIG_SCHED_DEBUG
6017 6018
	if (!state_filter)
		sysrq_sched_debug_show();
Ingo Molnar's avatar
Ingo Molnar committed
6019
#endif
6020
	rcu_read_unlock();
6021 6022 6023
	/*
	 * Only show locks if all tasks are dumped:
	 */
6024
	if (!state_filter)
6025
		debug_show_all_locks();
Linus Torvalds's avatar
Linus Torvalds committed
6026 6027
}

6028 6029 6030
/**
 * init_idle - set up an idle thread for a given CPU
 * @idle: task in question
6031
 * @cpu: CPU the idle task belongs to
6032 6033 6034 6035
 *
 * NOTE: this function does not set the idle thread's NEED_RESCHED
 * flag, to make booting more robust.
 */
6036
void init_idle(struct task_struct *idle, int cpu)
Linus Torvalds's avatar
Linus Torvalds committed
6037
{
6038
	struct rq *rq = cpu_rq(cpu);
Linus Torvalds's avatar
Linus Torvalds committed
6039 6040
	unsigned long flags;

6041 6042
	__sched_fork(0, idle);

6043 6044
	raw_spin_lock_irqsave(&idle->pi_lock, flags);
	raw_spin_lock(&rq->lock);
6045

6046
	idle->state = TASK_RUNNING;
Ingo Molnar's avatar
Ingo Molnar committed
6047
	idle->se.exec_start = sched_clock();
6048
	idle->flags |= PF_IDLE;
Ingo Molnar's avatar
Ingo Molnar committed
6049

6050 6051
	kasan_unpoison_task_stack(idle);

6052 6053 6054 6055 6056 6057 6058 6059 6060
#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
6061 6062
	/*
	 * We're having a chicken and egg problem, even though we are
6063
	 * holding rq->lock, the CPU isn't yet set to this CPU so the
6064 6065 6066 6067 6068 6069 6070 6071
	 * 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
6072
	__set_task_cpu(idle, cpu);
6073
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
6074

6075 6076
	rq->idle = idle;
	rcu_assign_pointer(rq->curr, idle);
6077
	idle->on_rq = TASK_ON_RQ_QUEUED;
6078
#ifdef CONFIG_SMP
6079
	idle->on_cpu = 1;
6080
#endif
6081 6082
	raw_spin_unlock(&rq->lock);
	raw_spin_unlock_irqrestore(&idle->pi_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
6083 6084

	/* Set the preempt count _outside_ the spinlocks! */
6085
	init_idle_preempt_count(idle, cpu);
6086

Ingo Molnar's avatar
Ingo Molnar committed
6087 6088 6089 6090
	/*
	 * The idle tasks have their own, simple scheduling class:
	 */
	idle->sched_class = &idle_sched_class;
6091
	ftrace_graph_init_idle_task(idle, cpu);
6092
	vtime_init_idle(idle, cpu);
6093
#ifdef CONFIG_SMP
6094 6095
	sprintf(idle->comm, "%s/%d", INIT_TASK_COMM, cpu);
#endif
6096 6097
}

6098 6099
#ifdef CONFIG_SMP

6100 6101 6102
int cpuset_cpumask_can_shrink(const struct cpumask *cur,
			      const struct cpumask *trial)
{
6103
	int ret = 1;
6104

6105 6106 6107
	if (!cpumask_weight(cur))
		return ret;

6108
	ret = dl_cpuset_cpumask_can_shrink(cur, trial);
6109 6110 6111 6112

	return ret;
}

6113 6114 6115 6116 6117 6118 6119
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
6120
	 * to a new cpuset; we don't want to change their CPU
6121 6122 6123 6124
	 * 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
6125
	 * before cpus_mask may be changed.
6126 6127 6128 6129 6130 6131 6132
	 */
	if (p->flags & PF_NO_SETAFFINITY) {
		ret = -EINVAL;
		goto out;
	}

	if (dl_task(p) && !cpumask_intersects(task_rq(p)->rd->span,
6133 6134
					      cs_cpus_allowed))
		ret = dl_task_can_attach(p, cs_cpus_allowed);
6135 6136 6137 6138 6139

out:
	return ret;
}

6140
bool sched_smp_initialized __read_mostly;
6141

6142 6143 6144 6145 6146 6147 6148 6149 6150 6151
#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;

6152
	if (!cpumask_test_cpu(target_cpu, p->cpus_ptr))
6153 6154 6155 6156
		return -EINVAL;

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

6157
	trace_sched_move_numa(p, curr_cpu, target_cpu);
6158 6159
	return stop_one_cpu(curr_cpu, migration_cpu_stop, &arg);
}
6160 6161 6162 6163 6164 6165 6166

/*
 * 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)
{
6167
	bool queued, running;
6168 6169
	struct rq_flags rf;
	struct rq *rq;
6170

6171
	rq = task_rq_lock(p, &rf);
6172
	queued = task_on_rq_queued(p);
6173 6174
	running = task_current(rq, p);

6175
	if (queued)
6176
		dequeue_task(rq, p, DEQUEUE_SAVE);
6177
	if (running)
6178
		put_prev_task(rq, p);
6179 6180 6181

	p->numa_preferred_nid = nid;

6182
	if (queued)
6183
		enqueue_task(rq, p, ENQUEUE_RESTORE | ENQUEUE_NOCLOCK);
6184
	if (running)
6185
		set_next_task(rq, p);
6186
	task_rq_unlock(rq, p, &rf);
6187
}
Peter Zijlstra's avatar
Peter Zijlstra committed
6188
#endif /* CONFIG_NUMA_BALANCING */
6189

Linus Torvalds's avatar
Linus Torvalds committed
6190
#ifdef CONFIG_HOTPLUG_CPU
6191
/*
6192
 * Ensure that the idle task is using init_mm right before its CPU goes
6193
 * offline.
6194
 */
6195
void idle_task_exit(void)
Linus Torvalds's avatar
Linus Torvalds committed
6196
{
6197
	struct mm_struct *mm = current->active_mm;
6198

6199
	BUG_ON(cpu_online(smp_processor_id()));
6200
	BUG_ON(current != this_rq()->idle);
6201

6202
	if (mm != &init_mm) {
6203
		switch_mm(mm, &init_mm, current);
6204 6205
		finish_arch_post_lock_switch();
	}
6206 6207

	/* finish_cpu(), as ran on the BP, will clean up the active_mm state */
Linus Torvalds's avatar
Linus Torvalds committed
6208 6209 6210
}

/*
6211 6212
 * 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
6213 6214 6215
 * 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.
6216 6217
 *
 * Also see the comment "Global load-average calculations".
Linus Torvalds's avatar
Linus Torvalds committed
6218
 */
6219
static void calc_load_migrate(struct rq *rq)
Linus Torvalds's avatar
Linus Torvalds committed
6220
{
6221
	long delta = calc_load_fold_active(rq, 1);
6222 6223
	if (delta)
		atomic_long_add(delta, &calc_load_tasks);
Linus Torvalds's avatar
Linus Torvalds committed
6224 6225
}

6226
static struct task_struct *__pick_migrate_task(struct rq *rq)
6227
{
6228 6229
	const struct sched_class *class;
	struct task_struct *next;
6230

6231
	for_each_class(class) {
6232
		next = class->pick_next_task(rq);
6233
		if (next) {
6234
			next->sched_class->put_prev_task(rq, next);
6235 6236 6237
			return next;
		}
	}
6238

6239 6240 6241
	/* The idle class should always have a runnable task */
	BUG();
}
6242

6243
/*
6244 6245 6246 6247 6248 6249
 * 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
6250
 */
6251
static void migrate_tasks(struct rq *dead_rq, struct rq_flags *rf)
Linus Torvalds's avatar
Linus Torvalds committed
6252
{
6253
	struct rq *rq = dead_rq;
6254
	struct task_struct *next, *stop = rq->stop;
6255
	struct rq_flags orf = *rf;
6256
	int dest_cpu;
Linus Torvalds's avatar
Linus Torvalds committed
6257 6258

	/*
6259 6260 6261 6262 6263 6264 6265
	 * 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
6266
	 */
6267
	rq->stop = NULL;
6268

6269 6270 6271 6272 6273 6274 6275
	/*
	 * 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);

6276
	for (;;) {
6277 6278
		/*
		 * There's this thread running, bail when that's the only
6279
		 * remaining thread:
6280 6281
		 */
		if (rq->nr_running == 1)
Ingo Molnar's avatar
Ingo Molnar committed
6282
			break;
6283

6284
		next = __pick_migrate_task(rq);
6285

6286
		/*
6287
		 * Rules for changing task_struct::cpus_mask are holding
6288 6289 6290 6291 6292 6293 6294
		 * 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.
		 */
6295
		rq_unlock(rq, rf);
6296
		raw_spin_lock(&next->pi_lock);
6297
		rq_relock(rq, rf);
6298 6299 6300 6301 6302 6303 6304 6305 6306 6307 6308

		/*
		 * 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;
		}

6309
		/* Find suitable destination for @next, with force if needed. */
6310
		dest_cpu = select_fallback_rq(dead_rq->cpu, next);
6311
		rq = __migrate_task(rq, rf, next, dest_cpu);
6312
		if (rq != dead_rq) {
6313
			rq_unlock(rq, rf);
6314
			rq = dead_rq;
6315 6316
			*rf = orf;
			rq_relock(rq, rf);
6317
		}
6318
		raw_spin_unlock(&next->pi_lock);
Linus Torvalds's avatar
Linus Torvalds committed
6319
	}
6320

6321
	rq->stop = stop;
6322
}
Linus Torvalds's avatar
Linus Torvalds committed
6323 6324
#endif /* CONFIG_HOTPLUG_CPU */

6325
void set_rq_online(struct rq *rq)
6326 6327 6328 6329
{
	if (!rq->online) {
		const struct sched_class *class;

6330
		cpumask_set_cpu(rq->cpu, rq->rd->online);
6331 6332 6333 6334 6335 6336 6337 6338 6339
		rq->online = 1;

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

6340
void set_rq_offline(struct rq *rq)
6341 6342 6343 6344 6345 6346 6347 6348 6349
{
	if (rq->online) {
		const struct sched_class *class;

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

6350
		cpumask_clear_cpu(rq->cpu, rq->rd->online);
6351 6352 6353 6354
		rq->online = 0;
	}
}

6355 6356 6357 6358
/*
 * used to mark begin/end of suspend/resume:
 */
static int num_cpus_frozen;
6359

Linus Torvalds's avatar
Linus Torvalds committed
6360
/*
6361 6362 6363
 * Update cpusets according to cpu_active mask.  If cpusets are
 * disabled, cpuset_update_active_cpus() becomes a simple wrapper
 * around partition_sched_domains().
6364 6365 6366
 *
 * 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
6367
 */
6368
static void cpuset_cpu_active(void)
6369
{
6370
	if (cpuhp_tasks_frozen) {
6371 6372 6373 6374 6375 6376
		/*
		 * 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.
		 */
6377 6378
		partition_sched_domains(1, NULL, NULL);
		if (--num_cpus_frozen)
6379
			return;
6380 6381 6382 6383 6384
		/*
		 * This is the last CPU online operation. So fall through and
		 * restore the original sched domains by considering the
		 * cpuset configurations.
		 */
6385
		cpuset_force_rebuild();
6386
	}
6387
	cpuset_update_active_cpus();
6388
}
6389

6390
static int cpuset_cpu_inactive(unsigned int cpu)
6391
{
6392
	if (!cpuhp_tasks_frozen) {
6393
		if (dl_cpu_busy(cpu))
6394
			return -EBUSY;
6395
		cpuset_update_active_cpus();
6396
	} else {
6397 6398
		num_cpus_frozen++;
		partition_sched_domains(1, NULL, NULL);
6399
	}
6400
	return 0;
6401 6402
}

6403
int sched_cpu_activate(unsigned int cpu)
6404
{
6405
	struct rq *rq = cpu_rq(cpu);
6406
	struct rq_flags rf;
6407

6408 6409
#ifdef CONFIG_SCHED_SMT
	/*
6410
	 * When going up, increment the number of cores with SMT present.
6411
	 */
6412 6413
	if (cpumask_weight(cpu_smt_mask(cpu)) == 2)
		static_branch_inc_cpuslocked(&sched_smt_present);
6414
#endif
6415
	set_cpu_active(cpu, true);
6416

6417
	if (sched_smp_initialized) {
6418
		sched_domains_numa_masks_set(cpu);
6419
		cpuset_cpu_active();
6420
	}
6421 6422 6423 6424 6425

	/*
	 * Put the rq online, if not already. This happens:
	 *
	 * 1) In the early boot process, because we build the real domains
6426
	 *    after all CPUs have been brought up.
6427 6428 6429 6430
	 *
	 * 2) At runtime, if cpuset_cpu_active() fails to rebuild the
	 *    domains.
	 */
6431
	rq_lock_irqsave(rq, &rf);
6432 6433 6434 6435
	if (rq->rd) {
		BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
		set_rq_online(rq);
	}
6436
	rq_unlock_irqrestore(rq, &rf);
6437

6438
	return 0;
6439 6440
}

6441
int sched_cpu_deactivate(unsigned int cpu)
6442 6443 6444
{
	int ret;

6445
	set_cpu_active(cpu, false);
6446 6447 6448 6449 6450 6451 6452
	/*
	 * 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.
	 */
6453
	synchronize_rcu();
6454

6455 6456 6457 6458 6459 6460 6461 6462
#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

6463 6464 6465 6466 6467 6468 6469
	if (!sched_smp_initialized)
		return 0;

	ret = cpuset_cpu_inactive(cpu);
	if (ret) {
		set_cpu_active(cpu, true);
		return ret;
6470
	}
6471 6472
	sched_domains_numa_masks_clear(cpu);
	return 0;
6473 6474
}

6475 6476 6477 6478 6479 6480 6481 6482
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();
}

6483 6484
int sched_cpu_starting(unsigned int cpu)
{
6485
	sched_rq_cpu_starting(cpu);
6486
	sched_tick_start(cpu);
6487
	return 0;
6488 6489
}

6490 6491 6492 6493
#ifdef CONFIG_HOTPLUG_CPU
int sched_cpu_dying(unsigned int cpu)
{
	struct rq *rq = cpu_rq(cpu);
6494
	struct rq_flags rf;
6495 6496 6497

	/* Handle pending wakeups and then migrate everything off */
	sched_ttwu_pending();
6498
	sched_tick_stop(cpu);
6499 6500

	rq_lock_irqsave(rq, &rf);
6501 6502 6503 6504
	if (rq->rd) {
		BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
		set_rq_offline(rq);
	}
6505
	migrate_tasks(rq, &rf);
6506
	BUG_ON(rq->nr_running != 1);
6507 6508
	rq_unlock_irqrestore(rq, &rf);

6509 6510
	calc_load_migrate(rq);
	update_max_interval();
6511
	nohz_balance_exit_idle(rq);
6512
	hrtick_clear(rq);
6513 6514 6515 6516
	return 0;
}
#endif

Linus Torvalds's avatar
Linus Torvalds committed
6517 6518
void __init sched_init_smp(void)
{
6519 6520
	sched_init_numa();

6521 6522
	/*
	 * There's no userspace yet to cause hotplug operations; hence all the
6523
	 * CPU masks are stable and all blatant races in the below code cannot
6524
	 * happen.
6525
	 */
6526
	mutex_lock(&sched_domains_mutex);
6527
	sched_init_domains(cpu_active_mask);
6528
	mutex_unlock(&sched_domains_mutex);
6529

6530
	/* Move init over to a non-isolated CPU */
6531
	if (set_cpus_allowed_ptr(current, housekeeping_cpumask(HK_FLAG_DOMAIN)) < 0)
6532
		BUG();
6533
	sched_init_granularity();
6534

6535
	init_sched_rt_class();
6536
	init_sched_dl_class();
6537

6538
	sched_smp_initialized = true;
Linus Torvalds's avatar
Linus Torvalds committed
6539
}
6540 6541 6542

static int __init migration_init(void)
{
6543
	sched_cpu_starting(smp_processor_id());
6544
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
6545
}
6546 6547
early_initcall(migration_init);

Linus Torvalds's avatar
Linus Torvalds committed
6548 6549 6550
#else
void __init sched_init_smp(void)
{
6551
	sched_init_granularity();
Linus Torvalds's avatar
Linus Torvalds committed
6552 6553 6554 6555 6556 6557 6558 6559 6560 6561
}
#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);
}

6562
#ifdef CONFIG_CGROUP_SCHED
6563 6564 6565 6566
/*
 * Default task group.
 * Every task in system belongs to this group at bootup.
 */
6567
struct task_group root_task_group;
6568
LIST_HEAD(task_groups);
6569 6570 6571

/* Cacheline aligned slab cache for task_group */
static struct kmem_cache *task_group_cache __read_mostly;
6572
#endif
Peter Zijlstra's avatar
Peter Zijlstra committed
6573

6574
DECLARE_PER_CPU(cpumask_var_t, load_balance_mask);
6575
DECLARE_PER_CPU(cpumask_var_t, select_idle_mask);
Peter Zijlstra's avatar
Peter Zijlstra committed
6576

Linus Torvalds's avatar
Linus Torvalds committed
6577 6578
void __init sched_init(void)
{
6579
	unsigned long ptr = 0;
6580
	int i;
6581

6582
	wait_bit_init();
6583

6584
#ifdef CONFIG_FAIR_GROUP_SCHED
6585
	ptr += 2 * nr_cpu_ids * sizeof(void **);
6586 6587
#endif
#ifdef CONFIG_RT_GROUP_SCHED
6588
	ptr += 2 * nr_cpu_ids * sizeof(void **);
6589
#endif
6590 6591
	if (ptr) {
		ptr = (unsigned long)kzalloc(ptr, GFP_NOWAIT);
6592 6593

#ifdef CONFIG_FAIR_GROUP_SCHED
6594
		root_task_group.se = (struct sched_entity **)ptr;
6595 6596
		ptr += nr_cpu_ids * sizeof(void **);

6597
		root_task_group.cfs_rq = (struct cfs_rq **)ptr;
6598
		ptr += nr_cpu_ids * sizeof(void **);
6599

6600 6601
		root_task_group.shares = ROOT_TASK_GROUP_LOAD;
		init_cfs_bandwidth(&root_task_group.cfs_bandwidth);
6602
#endif /* CONFIG_FAIR_GROUP_SCHED */
6603
#ifdef CONFIG_RT_GROUP_SCHED
6604
		root_task_group.rt_se = (struct sched_rt_entity **)ptr;
6605 6606
		ptr += nr_cpu_ids * sizeof(void **);

6607
		root_task_group.rt_rq = (struct rt_rq **)ptr;
6608 6609
		ptr += nr_cpu_ids * sizeof(void **);

6610
#endif /* CONFIG_RT_GROUP_SCHED */
6611
	}
6612
#ifdef CONFIG_CPUMASK_OFFSTACK
6613 6614 6615
	for_each_possible_cpu(i) {
		per_cpu(load_balance_mask, i) = (cpumask_var_t)kzalloc_node(
			cpumask_size(), GFP_KERNEL, cpu_to_node(i));
6616 6617
		per_cpu(select_idle_mask, i) = (cpumask_var_t)kzalloc_node(
			cpumask_size(), GFP_KERNEL, cpu_to_node(i));
6618
	}
6619
#endif /* CONFIG_CPUMASK_OFFSTACK */
Ingo Molnar's avatar
Ingo Molnar committed
6620

6621 6622
	init_rt_bandwidth(&def_rt_bandwidth, global_rt_period(), global_rt_runtime());
	init_dl_bandwidth(&def_dl_bandwidth, global_rt_period(), global_rt_runtime());
6623

6624 6625 6626 6627
#ifdef CONFIG_SMP
	init_defrootdomain();
#endif

6628
#ifdef CONFIG_RT_GROUP_SCHED
6629
	init_rt_bandwidth(&root_task_group.rt_bandwidth,
6630
			global_rt_period(), global_rt_runtime());
6631
#endif /* CONFIG_RT_GROUP_SCHED */
6632

Dhaval Giani's avatar
Dhaval Giani committed
6633
#ifdef CONFIG_CGROUP_SCHED
6634 6635
	task_group_cache = KMEM_CACHE(task_group, 0);

6636 6637
	list_add(&root_task_group.list, &task_groups);
	INIT_LIST_HEAD(&root_task_group.children);
6638
	INIT_LIST_HEAD(&root_task_group.siblings);
6639
	autogroup_init(&init_task);
Dhaval Giani's avatar
Dhaval Giani committed
6640
#endif /* CONFIG_CGROUP_SCHED */
Peter Zijlstra's avatar
Peter Zijlstra committed
6641

6642
	for_each_possible_cpu(i) {
6643
		struct rq *rq;
Linus Torvalds's avatar
Linus Torvalds committed
6644 6645

		rq = cpu_rq(i);
6646
		raw_spin_lock_init(&rq->lock);
Nick Piggin's avatar
Nick Piggin committed
6647
		rq->nr_running = 0;
6648 6649
		rq->calc_load_active = 0;
		rq->calc_load_update = jiffies + LOAD_FREQ;
6650
		init_cfs_rq(&rq->cfs);
6651 6652
		init_rt_rq(&rq->rt);
		init_dl_rq(&rq->dl);
Ingo Molnar's avatar
Ingo Molnar committed
6653
#ifdef CONFIG_FAIR_GROUP_SCHED
Peter Zijlstra's avatar
Peter Zijlstra committed
6654
		INIT_LIST_HEAD(&rq->leaf_cfs_rq_list);
6655
		rq->tmp_alone_branch = &rq->leaf_cfs_rq_list;
Dhaval Giani's avatar
Dhaval Giani committed
6656
		/*
6657
		 * How much CPU bandwidth does root_task_group get?
Dhaval Giani's avatar
Dhaval Giani committed
6658 6659
		 *
		 * In case of task-groups formed thr' the cgroup filesystem, it
6660 6661
		 * gets 100% of the CPU resources in the system. This overall
		 * system CPU resource is divided among the tasks of
6662
		 * root_task_group and its child task-groups in a fair manner,
Dhaval Giani's avatar
Dhaval Giani committed
6663 6664 6665
		 * based on each entity's (task or task-group's) weight
		 * (se->load.weight).
		 *
6666
		 * In other words, if root_task_group has 10 tasks of weight
Dhaval Giani's avatar
Dhaval Giani committed
6667
		 * 1024) and two child groups A0 and A1 (of weight 1024 each),
6668
		 * then A0's share of the CPU resource is:
Dhaval Giani's avatar
Dhaval Giani committed
6669
		 *
6670
		 *	A0's bandwidth = 1024 / (10*1024 + 1024 + 1024) = 8.33%
Dhaval Giani's avatar
Dhaval Giani committed
6671
		 *
6672 6673
		 * 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
6674
		 */
6675
		init_tg_cfs_entry(&root_task_group, &rq->cfs, NULL, i, NULL);
Dhaval Giani's avatar
Dhaval Giani committed
6676 6677 6678
#endif /* CONFIG_FAIR_GROUP_SCHED */

		rq->rt.rt_runtime = def_rt_bandwidth.rt_runtime;
6679
#ifdef CONFIG_RT_GROUP_SCHED
6680
		init_tg_rt_entry(&root_task_group, &rq->rt, NULL, i, NULL);
Ingo Molnar's avatar
Ingo Molnar committed
6681
#endif
Linus Torvalds's avatar
Linus Torvalds committed
6682
#ifdef CONFIG_SMP
Nick Piggin's avatar
Nick Piggin committed
6683
		rq->sd = NULL;
6684
		rq->rd = NULL;
6685
		rq->cpu_capacity = rq->cpu_capacity_orig = SCHED_CAPACITY_SCALE;
6686
		rq->balance_callback = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
6687
		rq->active_balance = 0;
Ingo Molnar's avatar
Ingo Molnar committed
6688
		rq->next_balance = jiffies;
Linus Torvalds's avatar
Linus Torvalds committed
6689
		rq->push_cpu = 0;
6690
		rq->cpu = i;
6691
		rq->online = 0;
6692 6693
		rq->idle_stamp = 0;
		rq->avg_idle = 2*sysctl_sched_migration_cost;
6694
		rq->max_idle_balance_cost = sysctl_sched_migration_cost;
6695 6696 6697

		INIT_LIST_HEAD(&rq->cfs_tasks);

6698
		rq_attach_root(rq, &def_root_domain);
6699
#ifdef CONFIG_NO_HZ_COMMON
6700
		rq->last_blocked_load_update_tick = jiffies;
6701
		atomic_set(&rq->nohz_flags, 0);
6702
#endif
6703
#endif /* CONFIG_SMP */
6704
		hrtick_rq_init(rq);
Linus Torvalds's avatar
Linus Torvalds committed
6705 6706 6707
		atomic_set(&rq->nr_iowait, 0);
	}

6708
	set_load_weight(&init_task, false);
6709

Linus Torvalds's avatar
Linus Torvalds committed
6710 6711 6712
	/*
	 * The boot idle thread does lazy MMU switching as well:
	 */
6713
	mmgrab(&init_mm);
Linus Torvalds's avatar
Linus Torvalds committed
6714 6715 6716 6717 6718 6719 6720 6721 6722
	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());
6723 6724 6725

	calc_load_update = jiffies + LOAD_FREQ;

6726
#ifdef CONFIG_SMP
6727
	idle_thread_set_boot_cpu();
6728 6729
#endif
	init_sched_fair_class();
6730

6731 6732
	init_schedstats();

6733 6734
	psi_init();

6735 6736
	init_uclamp();

6737
	scheduler_running = 1;
Linus Torvalds's avatar
Linus Torvalds committed
6738 6739
}

6740
#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
6741 6742
static inline int preempt_count_equals(int preempt_offset)
{
6743
	int nested = preempt_count() + rcu_preempt_depth();
6744

Arnd Bergmann's avatar
Arnd Bergmann committed
6745
	return (nested == preempt_offset);
6746 6747
}

6748
void __might_sleep(const char *file, int line, int preempt_offset)
Linus Torvalds's avatar
Linus Torvalds committed
6749
{
Peter Zijlstra's avatar
Peter Zijlstra committed
6750 6751 6752 6753 6754
	/*
	 * 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.
	 */
6755
	WARN_ONCE(current->state != TASK_RUNNING && current->task_state_change,
Peter Zijlstra's avatar
Peter Zijlstra committed
6756 6757 6758 6759
			"do not call blocking ops when !TASK_RUNNING; "
			"state=%lx set at [<%p>] %pS\n",
			current->state,
			(void *)current->task_state_change,
6760
			(void *)current->task_state_change);
Peter Zijlstra's avatar
Peter Zijlstra committed
6761

6762 6763 6764 6765 6766
	___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
6767
{
6768 6769 6770
	/* Ratelimiting timestamp: */
	static unsigned long prev_jiffy;

6771
	unsigned long preempt_disable_ip;
Linus Torvalds's avatar
Linus Torvalds committed
6772

6773 6774 6775
	/* WARN_ON_ONCE() by default, no rate limit required: */
	rcu_sleep_check();

6776
	if ((preempt_count_equals(preempt_offset) && !irqs_disabled() &&
6777
	     !is_idle_task(current) && !current->non_block_count) ||
6778 6779
	    system_state == SYSTEM_BOOTING || system_state > SYSTEM_RUNNING ||
	    oops_in_progress)
6780
		return;
6781

6782 6783 6784 6785
	if (time_before(jiffies, prev_jiffy + HZ) && prev_jiffy)
		return;
	prev_jiffy = jiffies;

6786
	/* Save this before calling printk(), since that will clobber it: */
6787 6788
	preempt_disable_ip = get_preempt_disable_ip(current);

6789 6790 6791 6792
	printk(KERN_ERR
		"BUG: sleeping function called from invalid context at %s:%d\n",
			file, line);
	printk(KERN_ERR
6793 6794
		"in_atomic(): %d, irqs_disabled(): %d, non_block: %d, pid: %d, name: %s\n",
			in_atomic(), irqs_disabled(), current->non_block_count,
6795
			current->pid, current->comm);
6796

6797 6798 6799
	if (task_stack_end_corrupted(current))
		printk(KERN_EMERG "Thread overran stack, or stack corrupted\n");

6800 6801 6802
	debug_show_held_locks(current);
	if (irqs_disabled())
		print_irqtrace_events(current);
6803 6804
	if (IS_ENABLED(CONFIG_DEBUG_PREEMPT)
	    && !preempt_count_equals(preempt_offset)) {
6805
		pr_err("Preemption disabled at:");
6806
		print_ip_sym(preempt_disable_ip);
6807 6808
		pr_cont("\n");
	}
6809
	dump_stack();
6810
	add_taint(TAINT_WARN, LOCKDEP_STILL_OK);
Linus Torvalds's avatar
Linus Torvalds committed
6811
}
6812
EXPORT_SYMBOL(___might_sleep);
6813 6814 6815 6816 6817 6818 6819 6820 6821 6822 6823 6824 6825 6826 6827 6828 6829 6830 6831 6832 6833 6834 6835 6836 6837 6838 6839 6840

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
6841 6842 6843
#endif

#ifdef CONFIG_MAGIC_SYSRQ
6844
void normalize_rt_tasks(void)
6845
{
6846
	struct task_struct *g, *p;
6847 6848 6849
	struct sched_attr attr = {
		.sched_policy = SCHED_NORMAL,
	};
Linus Torvalds's avatar
Linus Torvalds committed
6850

6851
	read_lock(&tasklist_lock);
6852
	for_each_process_thread(g, p) {
6853 6854 6855
		/*
		 * Only normalize user tasks:
		 */
6856
		if (p->flags & PF_KTHREAD)
6857 6858
			continue;

6859 6860 6861 6862
		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
6863

6864
		if (!dl_task(p) && !rt_task(p)) {
Ingo Molnar's avatar
Ingo Molnar committed
6865 6866 6867 6868
			/*
			 * Renice negative nice level userspace
			 * tasks back to 0:
			 */
6869
			if (task_nice(p) < 0)
Ingo Molnar's avatar
Ingo Molnar committed
6870
				set_user_nice(p, 0);
Linus Torvalds's avatar
Linus Torvalds committed
6871
			continue;
Ingo Molnar's avatar
Ingo Molnar committed
6872
		}
Linus Torvalds's avatar
Linus Torvalds committed
6873

6874
		__sched_setscheduler(p, &attr, false, false);
6875
	}
6876
	read_unlock(&tasklist_lock);
Linus Torvalds's avatar
Linus Torvalds committed
6877 6878 6879
}

#endif /* CONFIG_MAGIC_SYSRQ */
6880

6881
#if defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB)
6882
/*
6883
 * These functions are only useful for the IA64 MCA handling, or kdb.
6884 6885 6886 6887 6888 6889 6890 6891 6892
 *
 * 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.
 */

/**
6893
 * curr_task - return the current task for a given CPU.
6894 6895 6896
 * @cpu: the processor in question.
 *
 * ONLY VALID WHEN THE WHOLE SYSTEM IS STOPPED!
6897 6898
 *
 * Return: The current task for @cpu.
6899
 */
6900
struct task_struct *curr_task(int cpu)
6901 6902 6903 6904
{
	return cpu_curr(cpu);
}

6905 6906 6907
#endif /* defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB) */

#ifdef CONFIG_IA64
6908
/**
6909
 * ia64_set_curr_task - set the current task for a given CPU.
6910 6911 6912 6913
 * @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
6914
 * are serviced on a separate stack. It allows the architecture to switch the
6915
 * notion of the current task on a CPU in a non-blocking manner. This function
6916 6917 6918 6919 6920 6921 6922
 * 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!
 */
6923
void ia64_set_curr_task(int cpu, struct task_struct *p)
6924 6925 6926 6927 6928
{
	cpu_curr(cpu) = p;
}

#endif
6929

Dhaval Giani's avatar
Dhaval Giani committed
6930
#ifdef CONFIG_CGROUP_SCHED
6931 6932 6933
/* task_group_lock serializes the addition/removal of task groups */
static DEFINE_SPINLOCK(task_group_lock);

6934 6935 6936 6937
static inline void alloc_uclamp_sched_group(struct task_group *tg,
					    struct task_group *parent)
{
#ifdef CONFIG_UCLAMP_TASK_GROUP
6938
	enum uclamp_id clamp_id;
6939 6940 6941 6942

	for_each_clamp_id(clamp_id) {
		uclamp_se_set(&tg->uclamp_req[clamp_id],
			      uclamp_none(clamp_id), false);
6943
		tg->uclamp[clamp_id] = parent->uclamp[clamp_id];
6944 6945 6946 6947
	}
#endif
}

6948
static void sched_free_group(struct task_group *tg)
6949 6950 6951
{
	free_fair_sched_group(tg);
	free_rt_sched_group(tg);
6952
	autogroup_free(tg);
6953
	kmem_cache_free(task_group_cache, tg);
6954 6955 6956
}

/* allocate runqueue etc for a new task group */
6957
struct task_group *sched_create_group(struct task_group *parent)
6958 6959 6960
{
	struct task_group *tg;

6961
	tg = kmem_cache_alloc(task_group_cache, GFP_KERNEL | __GFP_ZERO);
6962 6963 6964
	if (!tg)
		return ERR_PTR(-ENOMEM);

6965
	if (!alloc_fair_sched_group(tg, parent))
6966 6967
		goto err;

6968
	if (!alloc_rt_sched_group(tg, parent))
6969 6970
		goto err;

6971 6972
	alloc_uclamp_sched_group(tg, parent);

6973 6974 6975
	return tg;

err:
6976
	sched_free_group(tg);
6977 6978 6979 6980 6981 6982 6983
	return ERR_PTR(-ENOMEM);
}

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

6984
	spin_lock_irqsave(&task_group_lock, flags);
Peter Zijlstra's avatar
Peter Zijlstra committed
6985
	list_add_rcu(&tg->list, &task_groups);
Peter Zijlstra's avatar
Peter Zijlstra committed
6986

6987 6988
	/* Root should already exist: */
	WARN_ON(!parent);
Peter Zijlstra's avatar
Peter Zijlstra committed
6989 6990 6991

	tg->parent = parent;
	INIT_LIST_HEAD(&tg->children);
6992
	list_add_rcu(&tg->siblings, &parent->children);
6993
	spin_unlock_irqrestore(&task_group_lock, flags);
6994 6995

	online_fair_sched_group(tg);
6996 6997
}

6998
/* rcu callback to free various structures associated with a task group */
6999
static void sched_free_group_rcu(struct rcu_head *rhp)
7000
{
7001
	/* Now it should be safe to free those cfs_rqs: */
7002
	sched_free_group(container_of(rhp, struct task_group, rcu));
7003 7004
}

7005
void sched_destroy_group(struct task_group *tg)
7006
{
7007
	/* Wait for possible concurrent references to cfs_rqs complete: */
7008
	call_rcu(&tg->rcu, sched_free_group_rcu);
7009 7010 7011
}

void sched_offline_group(struct task_group *tg)
7012
{
7013
	unsigned long flags;
7014

7015
	/* End participation in shares distribution: */
7016
	unregister_fair_sched_group(tg);
7017 7018

	spin_lock_irqsave(&task_group_lock, flags);
Peter Zijlstra's avatar
Peter Zijlstra committed
7019
	list_del_rcu(&tg->list);
Peter Zijlstra's avatar
Peter Zijlstra committed
7020
	list_del_rcu(&tg->siblings);
7021
	spin_unlock_irqrestore(&task_group_lock, flags);
7022 7023
}

7024
static void sched_change_group(struct task_struct *tsk, int type)
7025
{
7026
	struct task_group *tg;
7027

7028 7029 7030 7031 7032 7033
	/*
	 * 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),
7034 7035 7036 7037
			  struct task_group, css);
	tg = autogroup_task_group(tsk, tg);
	tsk->sched_task_group = tg;

Peter Zijlstra's avatar
Peter Zijlstra committed
7038
#ifdef CONFIG_FAIR_GROUP_SCHED
7039 7040
	if (tsk->sched_class->task_change_group)
		tsk->sched_class->task_change_group(tsk, type);
7041
	else
Peter Zijlstra's avatar
Peter Zijlstra committed
7042
#endif
7043
		set_task_rq(tsk, task_cpu(tsk));
7044 7045 7046 7047 7048 7049 7050 7051 7052 7053 7054
}

/*
 * 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)
{
7055 7056
	int queued, running, queue_flags =
		DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
7057 7058 7059 7060
	struct rq_flags rf;
	struct rq *rq;

	rq = task_rq_lock(tsk, &rf);
7061
	update_rq_clock(rq);
7062 7063 7064 7065 7066

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

	if (queued)
7067
		dequeue_task(rq, tsk, queue_flags);
7068
	if (running)
7069 7070 7071
		put_prev_task(rq, tsk);

	sched_change_group(tsk, TASK_MOVE_GROUP);
Peter Zijlstra's avatar
Peter Zijlstra committed
7072

7073
	if (queued)
7074
		enqueue_task(rq, tsk, queue_flags);
7075
	if (running) {
7076
		set_next_task(rq, tsk);
7077 7078 7079 7080 7081 7082 7083
		/*
		 * After changing group, the running task may have joined a
		 * throttled one but it's still the running task. Trigger a
		 * resched to make sure that task can still run.
		 */
		resched_curr(rq);
	}
7084

7085
	task_rq_unlock(rq, tsk, &rf);
7086
}
7087

7088
static inline struct task_group *css_tg(struct cgroup_subsys_state *css)
7089
{
7090
	return css ? container_of(css, struct task_group, css) : NULL;
7091 7092
}

7093 7094
static struct cgroup_subsys_state *
cpu_cgroup_css_alloc(struct cgroup_subsys_state *parent_css)
7095
{
7096 7097
	struct task_group *parent = css_tg(parent_css);
	struct task_group *tg;
7098

7099
	if (!parent) {
7100
		/* This is early initialization for the top cgroup */
7101
		return &root_task_group.css;
7102 7103
	}

7104
	tg = sched_create_group(parent);
7105 7106 7107 7108 7109 7110
	if (IS_ERR(tg))
		return ERR_PTR(-ENOMEM);

	return &tg->css;
}

7111 7112 7113 7114 7115 7116 7117 7118
/* 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);
7119 7120 7121 7122 7123 7124

#ifdef CONFIG_UCLAMP_TASK_GROUP
	/* Propagate the effective uclamp value for the new group */
	cpu_util_update_eff(css);
#endif

7125 7126 7127
	return 0;
}

7128
static void cpu_cgroup_css_released(struct cgroup_subsys_state *css)
7129
{
7130
	struct task_group *tg = css_tg(css);
7131

7132
	sched_offline_group(tg);
7133 7134
}

7135
static void cpu_cgroup_css_free(struct cgroup_subsys_state *css)
7136
{
7137
	struct task_group *tg = css_tg(css);
7138

7139 7140 7141 7142
	/*
	 * Relies on the RCU grace period between css_released() and this.
	 */
	sched_free_group(tg);
7143 7144
}

7145 7146 7147 7148
/*
 * 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.
 */
7149
static void cpu_cgroup_fork(struct task_struct *task)
7150
{
7151 7152 7153 7154 7155
	struct rq_flags rf;
	struct rq *rq;

	rq = task_rq_lock(task, &rf);

7156
	update_rq_clock(rq);
7157 7158 7159
	sched_change_group(task, TASK_SET_GROUP);

	task_rq_unlock(rq, task, &rf);
7160 7161
}

7162
static int cpu_cgroup_can_attach(struct cgroup_taskset *tset)
7163
{
7164
	struct task_struct *task;
7165
	struct cgroup_subsys_state *css;
7166
	int ret = 0;
7167

7168
	cgroup_taskset_for_each(task, css, tset) {
7169
#ifdef CONFIG_RT_GROUP_SCHED
7170
		if (!sched_rt_can_attach(css_tg(css), task))
7171
			return -EINVAL;
7172
#endif
7173 7174 7175 7176 7177 7178 7179 7180 7181 7182 7183 7184 7185 7186 7187 7188
		/*
		 * 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;
7189
	}
7190
	return ret;
7191
}
7192

7193
static void cpu_cgroup_attach(struct cgroup_taskset *tset)
7194
{
7195
	struct task_struct *task;
7196
	struct cgroup_subsys_state *css;
7197

7198
	cgroup_taskset_for_each(task, css, tset)
7199
		sched_move_task(task);
7200 7201
}

7202
#ifdef CONFIG_UCLAMP_TASK_GROUP
7203 7204 7205 7206 7207 7208
static void cpu_util_update_eff(struct cgroup_subsys_state *css)
{
	struct cgroup_subsys_state *top_css = css;
	struct uclamp_se *uc_parent = NULL;
	struct uclamp_se *uc_se = NULL;
	unsigned int eff[UCLAMP_CNT];
7209
	enum uclamp_id clamp_id;
7210 7211 7212 7213 7214 7215 7216 7217 7218 7219 7220 7221 7222 7223 7224 7225 7226 7227 7228 7229 7230 7231 7232 7233 7234 7235 7236 7237
	unsigned int clamps;

	css_for_each_descendant_pre(css, top_css) {
		uc_parent = css_tg(css)->parent
			? css_tg(css)->parent->uclamp : NULL;

		for_each_clamp_id(clamp_id) {
			/* Assume effective clamps matches requested clamps */
			eff[clamp_id] = css_tg(css)->uclamp_req[clamp_id].value;
			/* Cap effective clamps with parent's effective clamps */
			if (uc_parent &&
			    eff[clamp_id] > uc_parent[clamp_id].value) {
				eff[clamp_id] = uc_parent[clamp_id].value;
			}
		}
		/* Ensure protection is always capped by limit */
		eff[UCLAMP_MIN] = min(eff[UCLAMP_MIN], eff[UCLAMP_MAX]);

		/* Propagate most restrictive effective clamps */
		clamps = 0x0;
		uc_se = css_tg(css)->uclamp;
		for_each_clamp_id(clamp_id) {
			if (eff[clamp_id] == uc_se[clamp_id].value)
				continue;
			uc_se[clamp_id].value = eff[clamp_id];
			uc_se[clamp_id].bucket_id = uclamp_bucket_id(eff[clamp_id]);
			clamps |= (0x1 << clamp_id);
		}
7238
		if (!clamps) {
7239
			css = css_rightmost_descendant(css);
7240 7241 7242 7243 7244
			continue;
		}

		/* Immediately update descendants RUNNABLE tasks */
		uclamp_update_active_tasks(css, clamps);
7245 7246
	}
}
7247 7248 7249 7250 7251 7252 7253 7254 7255 7256 7257 7258 7259 7260 7261 7262 7263 7264 7265 7266 7267 7268 7269 7270 7271 7272 7273 7274 7275 7276 7277 7278

/*
 * Integer 10^N with a given N exponent by casting to integer the literal "1eN"
 * C expression. Since there is no way to convert a macro argument (N) into a
 * character constant, use two levels of macros.
 */
#define _POW10(exp) ((unsigned int)1e##exp)
#define POW10(exp) _POW10(exp)

struct uclamp_request {
#define UCLAMP_PERCENT_SHIFT	2
#define UCLAMP_PERCENT_SCALE	(100 * POW10(UCLAMP_PERCENT_SHIFT))
	s64 percent;
	u64 util;
	int ret;
};

static inline struct uclamp_request
capacity_from_percent(char *buf)
{
	struct uclamp_request req = {
		.percent = UCLAMP_PERCENT_SCALE,
		.util = SCHED_CAPACITY_SCALE,
		.ret = 0,
	};

	buf = strim(buf);
	if (strcmp(buf, "max")) {
		req.ret = cgroup_parse_float(buf, UCLAMP_PERCENT_SHIFT,
					     &req.percent);
		if (req.ret)
			return req;
7279
		if ((u64)req.percent > UCLAMP_PERCENT_SCALE) {
7280 7281 7282 7283 7284 7285 7286 7287 7288 7289 7290 7291 7292 7293 7294 7295 7296 7297 7298 7299 7300 7301 7302 7303 7304 7305 7306 7307 7308 7309 7310 7311 7312 7313 7314
			req.ret = -ERANGE;
			return req;
		}

		req.util = req.percent << SCHED_CAPACITY_SHIFT;
		req.util = DIV_ROUND_CLOSEST_ULL(req.util, UCLAMP_PERCENT_SCALE);
	}

	return req;
}

static ssize_t cpu_uclamp_write(struct kernfs_open_file *of, char *buf,
				size_t nbytes, loff_t off,
				enum uclamp_id clamp_id)
{
	struct uclamp_request req;
	struct task_group *tg;

	req = capacity_from_percent(buf);
	if (req.ret)
		return req.ret;

	mutex_lock(&uclamp_mutex);
	rcu_read_lock();

	tg = css_tg(of_css(of));
	if (tg->uclamp_req[clamp_id].value != req.util)
		uclamp_se_set(&tg->uclamp_req[clamp_id], req.util, false);

	/*
	 * Because of not recoverable conversion rounding we keep track of the
	 * exact requested value
	 */
	tg->uclamp_pct[clamp_id] = req.percent;

7315 7316 7317
	/* Update effective clamps to track the most restrictive value */
	cpu_util_update_eff(of_css(of));

7318 7319 7320 7321 7322 7323 7324 7325 7326 7327 7328 7329 7330 7331 7332 7333 7334 7335 7336 7337 7338 7339 7340 7341 7342 7343 7344 7345 7346 7347 7348 7349 7350 7351 7352 7353 7354 7355 7356 7357 7358 7359 7360 7361 7362 7363 7364 7365 7366 7367 7368 7369 7370 7371 7372 7373
	rcu_read_unlock();
	mutex_unlock(&uclamp_mutex);

	return nbytes;
}

static ssize_t cpu_uclamp_min_write(struct kernfs_open_file *of,
				    char *buf, size_t nbytes,
				    loff_t off)
{
	return cpu_uclamp_write(of, buf, nbytes, off, UCLAMP_MIN);
}

static ssize_t cpu_uclamp_max_write(struct kernfs_open_file *of,
				    char *buf, size_t nbytes,
				    loff_t off)
{
	return cpu_uclamp_write(of, buf, nbytes, off, UCLAMP_MAX);
}

static inline void cpu_uclamp_print(struct seq_file *sf,
				    enum uclamp_id clamp_id)
{
	struct task_group *tg;
	u64 util_clamp;
	u64 percent;
	u32 rem;

	rcu_read_lock();
	tg = css_tg(seq_css(sf));
	util_clamp = tg->uclamp_req[clamp_id].value;
	rcu_read_unlock();

	if (util_clamp == SCHED_CAPACITY_SCALE) {
		seq_puts(sf, "max\n");
		return;
	}

	percent = tg->uclamp_pct[clamp_id];
	percent = div_u64_rem(percent, POW10(UCLAMP_PERCENT_SHIFT), &rem);
	seq_printf(sf, "%llu.%0*u\n", percent, UCLAMP_PERCENT_SHIFT, rem);
}

static int cpu_uclamp_min_show(struct seq_file *sf, void *v)
{
	cpu_uclamp_print(sf, UCLAMP_MIN);
	return 0;
}

static int cpu_uclamp_max_show(struct seq_file *sf, void *v)
{
	cpu_uclamp_print(sf, UCLAMP_MAX);
	return 0;
}
#endif /* CONFIG_UCLAMP_TASK_GROUP */

7374
#ifdef CONFIG_FAIR_GROUP_SCHED
7375 7376
static int cpu_shares_write_u64(struct cgroup_subsys_state *css,
				struct cftype *cftype, u64 shareval)
7377
{
7378 7379
	if (shareval > scale_load_down(ULONG_MAX))
		shareval = MAX_SHARES;
7380
	return sched_group_set_shares(css_tg(css), scale_load(shareval));
7381 7382
}

7383 7384
static u64 cpu_shares_read_u64(struct cgroup_subsys_state *css,
			       struct cftype *cft)
7385
{
7386
	struct task_group *tg = css_tg(css);
7387

7388
	return (u64) scale_load_down(tg->shares);
7389
}
7390 7391

#ifdef CONFIG_CFS_BANDWIDTH
7392 7393
static DEFINE_MUTEX(cfs_constraints_mutex);

7394
const u64 max_cfs_quota_period = 1 * NSEC_PER_SEC; /* 1s */
7395
static const u64 min_cfs_quota_period = 1 * NSEC_PER_MSEC; /* 1ms */
7396

7397 7398
static int __cfs_schedulable(struct task_group *tg, u64 period, u64 runtime);

7399 7400
static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota)
{
7401
	int i, ret = 0, runtime_enabled, runtime_was_enabled;
7402
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
7403 7404 7405 7406 7407 7408 7409 7410 7411 7412 7413 7414 7415 7416 7417 7418 7419 7420 7421 7422

	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;

7423 7424 7425 7426 7427
	/*
	 * Prevent race between setting of cfs_rq->runtime_enabled and
	 * unthrottle_offline_cfs_rqs().
	 */
	get_online_cpus();
7428 7429 7430 7431 7432
	mutex_lock(&cfs_constraints_mutex);
	ret = __cfs_schedulable(tg, period, quota);
	if (ret)
		goto out_unlock;

7433
	runtime_enabled = quota != RUNTIME_INF;
7434
	runtime_was_enabled = cfs_b->quota != RUNTIME_INF;
7435 7436 7437 7438 7439 7440
	/*
	 * 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();
7441 7442 7443
	raw_spin_lock_irq(&cfs_b->lock);
	cfs_b->period = ns_to_ktime(period);
	cfs_b->quota = quota;
7444

Paul Turner's avatar
Paul Turner committed
7445
	__refill_cfs_bandwidth_runtime(cfs_b);
7446 7447

	/* Restart the period timer (if active) to handle new period expiry: */
7448 7449
	if (runtime_enabled)
		start_cfs_bandwidth(cfs_b);
7450

7451 7452
	raw_spin_unlock_irq(&cfs_b->lock);

7453
	for_each_online_cpu(i) {
7454
		struct cfs_rq *cfs_rq = tg->cfs_rq[i];
7455
		struct rq *rq = cfs_rq->rq;
7456
		struct rq_flags rf;
7457

7458
		rq_lock_irq(rq, &rf);
7459
		cfs_rq->runtime_enabled = runtime_enabled;
7460
		cfs_rq->runtime_remaining = 0;
7461

7462
		if (cfs_rq->throttled)
7463
			unthrottle_cfs_rq(cfs_rq);
7464
		rq_unlock_irq(rq, &rf);
7465
	}
7466 7467
	if (runtime_was_enabled && !runtime_enabled)
		cfs_bandwidth_usage_dec();
7468 7469
out_unlock:
	mutex_unlock(&cfs_constraints_mutex);
7470
	put_online_cpus();
7471

7472
	return ret;
7473 7474
}

7475
static int tg_set_cfs_quota(struct task_group *tg, long cfs_quota_us)
7476 7477 7478
{
	u64 quota, period;

7479
	period = ktime_to_ns(tg->cfs_bandwidth.period);
7480 7481
	if (cfs_quota_us < 0)
		quota = RUNTIME_INF;
7482
	else if ((u64)cfs_quota_us <= U64_MAX / NSEC_PER_USEC)
7483
		quota = (u64)cfs_quota_us * NSEC_PER_USEC;
7484 7485
	else
		return -EINVAL;
7486 7487 7488 7489

	return tg_set_cfs_bandwidth(tg, period, quota);
}

7490
static long tg_get_cfs_quota(struct task_group *tg)
7491 7492 7493
{
	u64 quota_us;

7494
	if (tg->cfs_bandwidth.quota == RUNTIME_INF)
7495 7496
		return -1;

7497
	quota_us = tg->cfs_bandwidth.quota;
7498 7499 7500 7501 7502
	do_div(quota_us, NSEC_PER_USEC);

	return quota_us;
}

7503
static int tg_set_cfs_period(struct task_group *tg, long cfs_period_us)
7504 7505 7506
{
	u64 quota, period;

7507 7508 7509
	if ((u64)cfs_period_us > U64_MAX / NSEC_PER_USEC)
		return -EINVAL;

7510
	period = (u64)cfs_period_us * NSEC_PER_USEC;
7511
	quota = tg->cfs_bandwidth.quota;
7512 7513 7514 7515

	return tg_set_cfs_bandwidth(tg, period, quota);
}

7516
static long tg_get_cfs_period(struct task_group *tg)
7517 7518 7519
{
	u64 cfs_period_us;

7520
	cfs_period_us = ktime_to_ns(tg->cfs_bandwidth.period);
7521 7522 7523 7524 7525
	do_div(cfs_period_us, NSEC_PER_USEC);

	return cfs_period_us;
}

7526 7527
static s64 cpu_cfs_quota_read_s64(struct cgroup_subsys_state *css,
				  struct cftype *cft)
7528
{
7529
	return tg_get_cfs_quota(css_tg(css));
7530 7531
}

7532 7533
static int cpu_cfs_quota_write_s64(struct cgroup_subsys_state *css,
				   struct cftype *cftype, s64 cfs_quota_us)
7534
{
7535
	return tg_set_cfs_quota(css_tg(css), cfs_quota_us);
7536 7537
}

7538 7539
static u64 cpu_cfs_period_read_u64(struct cgroup_subsys_state *css,
				   struct cftype *cft)
7540
{
7541
	return tg_get_cfs_period(css_tg(css));
7542 7543
}

7544 7545
static int cpu_cfs_period_write_u64(struct cgroup_subsys_state *css,
				    struct cftype *cftype, u64 cfs_period_us)
7546
{
7547
	return tg_set_cfs_period(css_tg(css), cfs_period_us);
7548 7549
}

7550 7551 7552 7553 7554 7555 7556 7557 7558 7559 7560 7561 7562 7563 7564 7565 7566 7567 7568 7569 7570 7571 7572 7573 7574 7575 7576 7577 7578 7579 7580 7581
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;
7582
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
7583 7584 7585 7586 7587
	s64 quota = 0, parent_quota = -1;

	if (!tg->parent) {
		quota = RUNTIME_INF;
	} else {
7588
		struct cfs_bandwidth *parent_b = &tg->parent->cfs_bandwidth;
7589 7590

		quota = normalize_cfs_quota(tg, d);
7591
		parent_quota = parent_b->hierarchical_quota;
7592 7593

		/*
7594 7595
		 * Ensure max(child_quota) <= parent_quota.  On cgroup2,
		 * always take the min.  On cgroup1, only inherit when no
7596
		 * limit is set:
7597
		 */
7598 7599 7600 7601 7602 7603 7604 7605
		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;
		}
7606
	}
7607
	cfs_b->hierarchical_quota = quota;
7608 7609 7610 7611 7612 7613

	return 0;
}

static int __cfs_schedulable(struct task_group *tg, u64 period, u64 quota)
{
7614
	int ret;
7615 7616 7617 7618 7619 7620 7621 7622 7623 7624 7625
	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);
	}

7626 7627 7628 7629 7630
	rcu_read_lock();
	ret = walk_tg_tree(tg_cfs_schedulable_down, tg_nop, &data);
	rcu_read_unlock();

	return ret;
7631
}
7632

7633
static int cpu_cfs_stat_show(struct seq_file *sf, void *v)
7634
{
7635
	struct task_group *tg = css_tg(seq_css(sf));
7636
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
7637

7638 7639 7640
	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);
7641

7642 7643 7644 7645 7646 7647 7648 7649 7650 7651
	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);
	}

7652 7653
	return 0;
}
7654
#endif /* CONFIG_CFS_BANDWIDTH */
7655
#endif /* CONFIG_FAIR_GROUP_SCHED */
7656

7657
#ifdef CONFIG_RT_GROUP_SCHED
7658 7659
static int cpu_rt_runtime_write(struct cgroup_subsys_state *css,
				struct cftype *cft, s64 val)
Peter Zijlstra's avatar
Peter Zijlstra committed
7660
{
7661
	return sched_group_set_rt_runtime(css_tg(css), val);
Peter Zijlstra's avatar
Peter Zijlstra committed
7662 7663
}

7664 7665
static s64 cpu_rt_runtime_read(struct cgroup_subsys_state *css,
			       struct cftype *cft)
Peter Zijlstra's avatar
Peter Zijlstra committed
7666
{
7667
	return sched_group_rt_runtime(css_tg(css));
Peter Zijlstra's avatar
Peter Zijlstra committed
7668
}
7669

7670 7671
static int cpu_rt_period_write_uint(struct cgroup_subsys_state *css,
				    struct cftype *cftype, u64 rt_period_us)
7672
{
7673
	return sched_group_set_rt_period(css_tg(css), rt_period_us);
7674 7675
}

7676 7677
static u64 cpu_rt_period_read_uint(struct cgroup_subsys_state *css,
				   struct cftype *cft)
7678
{
7679
	return sched_group_rt_period(css_tg(css));
7680
}
7681
#endif /* CONFIG_RT_GROUP_SCHED */
Peter Zijlstra's avatar
Peter Zijlstra committed
7682

7683
static struct cftype cpu_legacy_files[] = {
7684
#ifdef CONFIG_FAIR_GROUP_SCHED
7685 7686
	{
		.name = "shares",
7687 7688
		.read_u64 = cpu_shares_read_u64,
		.write_u64 = cpu_shares_write_u64,
7689
	},
7690
#endif
7691 7692 7693 7694 7695 7696 7697 7698 7699 7700 7701
#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,
	},
7702 7703
	{
		.name = "stat",
7704
		.seq_show = cpu_cfs_stat_show,
7705
	},
7706
#endif
7707
#ifdef CONFIG_RT_GROUP_SCHED
Peter Zijlstra's avatar
Peter Zijlstra committed
7708
	{
Peter Zijlstra's avatar
Peter Zijlstra committed
7709
		.name = "rt_runtime_us",
7710 7711
		.read_s64 = cpu_rt_runtime_read,
		.write_s64 = cpu_rt_runtime_write,
Peter Zijlstra's avatar
Peter Zijlstra committed
7712
	},
7713 7714
	{
		.name = "rt_period_us",
7715 7716
		.read_u64 = cpu_rt_period_read_uint,
		.write_u64 = cpu_rt_period_write_uint,
7717
	},
7718 7719 7720 7721 7722 7723 7724 7725 7726 7727 7728 7729 7730 7731
#endif
#ifdef CONFIG_UCLAMP_TASK_GROUP
	{
		.name = "uclamp.min",
		.flags = CFTYPE_NOT_ON_ROOT,
		.seq_show = cpu_uclamp_min_show,
		.write = cpu_uclamp_min_write,
	},
	{
		.name = "uclamp.max",
		.flags = CFTYPE_NOT_ON_ROOT,
		.seq_show = cpu_uclamp_max_show,
		.write = cpu_uclamp_max_write,
	},
7732
#endif
7733
	{ }	/* Terminate */
7734 7735
};

7736 7737
static int cpu_extra_stat_show(struct seq_file *sf,
			       struct cgroup_subsys_state *css)
7738 7739 7740
{
#ifdef CONFIG_CFS_BANDWIDTH
	{
7741
		struct task_group *tg = css_tg(css);
7742 7743 7744 7745 7746 7747 7748 7749 7750 7751 7752 7753 7754 7755 7756 7757 7758 7759 7760 7761 7762 7763 7764 7765 7766 7767 7768 7769 7770 7771 7772 7773 7774 7775 7776 7777 7778 7779 7780 7781 7782 7783 7784 7785 7786 7787 7788 7789 7790 7791 7792 7793 7794 7795 7796 7797 7798 7799 7800 7801 7802 7803 7804 7805 7806 7807
		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;
7808
	int idx;
7809 7810 7811 7812

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

7813 7814 7815 7816
	idx = NICE_TO_PRIO(nice) - MAX_RT_PRIO;
	idx = array_index_nospec(idx, 40);
	weight = sched_prio_to_weight[idx];

7817 7818 7819 7820 7821 7822 7823 7824 7825 7826 7827 7828 7829 7830 7831 7832 7833 7834 7835 7836 7837
	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 */

7838
	if (sscanf(buf, "%20s %llu", tok, periodp) < 1)
7839 7840 7841 7842 7843 7844 7845 7846 7847 7848 7849 7850 7851 7852 7853 7854 7855 7856 7857 7858 7859 7860 7861 7862 7863 7864 7865 7866 7867 7868 7869 7870 7871 7872 7873 7874 7875 7876 7877 7878 7879 7880 7881 7882 7883 7884 7885 7886 7887 7888 7889 7890 7891 7892 7893 7894 7895 7896 7897 7898
		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,
	},
7899 7900 7901 7902 7903 7904 7905 7906 7907 7908 7909 7910 7911 7912
#endif
#ifdef CONFIG_UCLAMP_TASK_GROUP
	{
		.name = "uclamp.min",
		.flags = CFTYPE_NOT_ON_ROOT,
		.seq_show = cpu_uclamp_min_show,
		.write = cpu_uclamp_min_write,
	},
	{
		.name = "uclamp.max",
		.flags = CFTYPE_NOT_ON_ROOT,
		.seq_show = cpu_uclamp_max_show,
		.write = cpu_uclamp_max_write,
	},
7913 7914 7915 7916
#endif
	{ }	/* terminate */
};

7917
struct cgroup_subsys cpu_cgrp_subsys = {
7918
	.css_alloc	= cpu_cgroup_css_alloc,
7919
	.css_online	= cpu_cgroup_css_online,
7920
	.css_released	= cpu_cgroup_css_released,
7921
	.css_free	= cpu_cgroup_css_free,
7922
	.css_extra_stat_show = cpu_extra_stat_show,
7923
	.fork		= cpu_cgroup_fork,
7924 7925
	.can_attach	= cpu_cgroup_can_attach,
	.attach		= cpu_cgroup_attach,
7926
	.legacy_cftypes	= cpu_legacy_files,
7927
	.dfl_cftypes	= cpu_files,
7928
	.early_init	= true,
7929
	.threaded	= true,
7930 7931
};

7932
#endif	/* CONFIG_CGROUP_SCHED */
7933

7934 7935 7936 7937 7938
void dump_cpu_task(int cpu)
{
	pr_info("Task dump for CPU %d:\n", cpu);
	sched_show_task(cpu_curr(cpu));
}
7939 7940 7941 7942 7943 7944 7945 7946 7947 7948 7949 7950 7951 7952 7953 7954 7955 7956 7957 7958 7959 7960 7961 7962 7963 7964 7965 7966 7967 7968 7969 7970 7971 7972 7973 7974 7975 7976 7977 7978 7979

/*
 * 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,
};
7980 7981

#undef CREATE_TRACE_POINTS