core.c 276 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 10 11 12
#define CREATE_TRACE_POINTS
#include <trace/events/sched.h>
#undef CREATE_TRACE_POINTS

13
#include "sched.h"
Linus Torvalds's avatar
Linus Torvalds committed
14

15
#include <linux/nospec.h>
16
#include <linux/blkdev.h>
17
#include <linux/kcov.h>
18
#include <linux/scs.h>
19

20
#include <asm/switch_to.h>
21
#include <asm/tlb.h>
Linus Torvalds's avatar
Linus Torvalds committed
22

23
#include "../workqueue_internal.h"
24
#include "../../fs/io-wq.h"
25
#include "../smpboot.h"
26

27
#include "pelt.h"
28
#include "smp.h"
29

30 31 32 33 34 35 36 37 38
/*
 * 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);
39
EXPORT_TRACEPOINT_SYMBOL_GPL(pelt_thermal_tp);
40
EXPORT_TRACEPOINT_SYMBOL_GPL(sched_cpu_capacity_tp);
41
EXPORT_TRACEPOINT_SYMBOL_GPL(sched_overutilized_tp);
42 43
EXPORT_TRACEPOINT_SYMBOL_GPL(sched_util_est_cfs_tp);
EXPORT_TRACEPOINT_SYMBOL_GPL(sched_util_est_se_tp);
44
EXPORT_TRACEPOINT_SYMBOL_GPL(sched_update_nr_running_tp);
45

46
DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
47

48
#ifdef CONFIG_SCHED_DEBUG
49 50
/*
 * Debugging: various feature bits
51 52 53 54
 *
 * 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.
55
 */
56 57
#define SCHED_FEAT(name, enabled)	\
	(1UL << __SCHED_FEAT_##name) * enabled |
58
const_debug unsigned int sysctl_sched_features =
59
#include "features.h"
60 61
	0;
#undef SCHED_FEAT
62 63 64 65 66 67 68 69 70 71 72

/*
 * Print a warning if need_resched is set for the given duration (if
 * LATENCY_WARN is enabled).
 *
 * If sysctl_resched_latency_warn_once is set, only one warning will be shown
 * per boot.
 */
__read_mostly int sysctl_resched_latency_warn_ms = 100;
__read_mostly int sysctl_resched_latency_warn_once = 1;
#endif /* CONFIG_SCHED_DEBUG */
73

74 75 76 77
/*
 * Number of tasks to iterate in a single balance run.
 * Limited because this is done with IRQs disabled.
 */
78 79 80
#ifdef CONFIG_PREEMPT_RT
const_debug unsigned int sysctl_sched_nr_migrate = 8;
#else
81
const_debug unsigned int sysctl_sched_nr_migrate = 32;
82
#endif
83

Peter Zijlstra's avatar
Peter Zijlstra committed
84
/*
85
 * period over which we measure -rt task CPU usage in us.
Peter Zijlstra's avatar
Peter Zijlstra committed
86 87
 * default: 1s
 */
Peter Zijlstra's avatar
Peter Zijlstra committed
88
unsigned int sysctl_sched_rt_period = 1000000;
Peter Zijlstra's avatar
Peter Zijlstra committed
89

90
__read_mostly int scheduler_running;
91

Peter Zijlstra's avatar
Peter Zijlstra committed
92 93 94 95
#ifdef CONFIG_SCHED_CORE

DEFINE_STATIC_KEY_FALSE(__sched_core_enabled);

96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118
/* kernel prio, less is more */
static inline int __task_prio(struct task_struct *p)
{
	if (p->sched_class == &stop_sched_class) /* trumps deadline */
		return -2;

	if (rt_prio(p->prio)) /* includes deadline */
		return p->prio; /* [-1, 99] */

	if (p->sched_class == &idle_sched_class)
		return MAX_RT_PRIO + NICE_WIDTH; /* 140 */

	return MAX_RT_PRIO + MAX_NICE; /* 120, squash fair */
}

/*
 * l(a,b)
 * le(a,b) := !l(b,a)
 * g(a,b)  := l(b,a)
 * ge(a,b) := !l(a,b)
 */

/* real prio, less is less */
119
static inline bool prio_less(struct task_struct *a, struct task_struct *b, bool in_fi)
120 121 122 123 124 125 126 127 128 129 130 131 132
{

	int pa = __task_prio(a), pb = __task_prio(b);

	if (-pa < -pb)
		return true;

	if (-pb < -pa)
		return false;

	if (pa == -1) /* dl_prio() doesn't work because of stop_class above */
		return !dl_time_before(a->dl.deadline, b->dl.deadline);

133 134
	if (pa == MAX_RT_PRIO + MAX_NICE)	/* fair */
		return cfs_prio_less(a, b, in_fi);
135 136 137 138 139 140 141 142 143 144 145 146 147

	return false;
}

static inline bool __sched_core_less(struct task_struct *a, struct task_struct *b)
{
	if (a->core_cookie < b->core_cookie)
		return true;

	if (a->core_cookie > b->core_cookie)
		return false;

	/* flip prio, so high prio is leftmost */
148
	if (prio_less(b, a, !!task_rq(a)->core->core_forceidle_count))
149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174
		return true;

	return false;
}

#define __node_2_sc(node) rb_entry((node), struct task_struct, core_node)

static inline bool rb_sched_core_less(struct rb_node *a, const struct rb_node *b)
{
	return __sched_core_less(__node_2_sc(a), __node_2_sc(b));
}

static inline int rb_sched_core_cmp(const void *key, const struct rb_node *node)
{
	const struct task_struct *p = __node_2_sc(node);
	unsigned long cookie = (unsigned long)key;

	if (cookie < p->core_cookie)
		return -1;

	if (cookie > p->core_cookie)
		return 1;

	return 0;
}

175
void sched_core_enqueue(struct rq *rq, struct task_struct *p)
176 177 178 179 180 181 182 183 184
{
	rq->core->core_task_seq++;

	if (!p->core_cookie)
		return;

	rb_add(&p->core_node, &rq->core_tree, rb_sched_core_less);
}

185
void sched_core_dequeue(struct rq *rq, struct task_struct *p, int flags)
186 187 188
{
	rq->core->core_task_seq++;

189 190 191 192
	if (sched_core_enqueued(p)) {
		rb_erase(&p->core_node, &rq->core_tree);
		RB_CLEAR_NODE(&p->core_node);
	}
193

194 195 196 197 198 199 200 201
	/*
	 * Migrating the last task off the cpu, with the cpu in forced idle
	 * state. Reschedule to create an accounting edge for forced idle,
	 * and re-examine whether the core is still in forced idle state.
	 */
	if (!(flags & DEQUEUE_SAVE) && rq->nr_running == 1 &&
	    rq->core->core_forceidle_count && rq->curr == rq->idle)
		resched_curr(rq);
202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220
}

/*
 * Find left-most (aka, highest priority) task matching @cookie.
 */
static struct task_struct *sched_core_find(struct rq *rq, unsigned long cookie)
{
	struct rb_node *node;

	node = rb_find_first((void *)cookie, &rq->core_tree, rb_sched_core_cmp);
	/*
	 * The idle task always matches any cookie!
	 */
	if (!node)
		return idle_sched_class.pick_task(rq);

	return __node_2_sc(node);
}

221 222 223 224 225 226 227 228 229 230 231 232 233 234 235
static struct task_struct *sched_core_next(struct task_struct *p, unsigned long cookie)
{
	struct rb_node *node = &p->core_node;

	node = rb_next(node);
	if (!node)
		return NULL;

	p = container_of(node, struct task_struct, core_node);
	if (p->core_cookie != cookie)
		return NULL;

	return p;
}

Peter Zijlstra's avatar
Peter Zijlstra committed
236 237 238 239 240 241 242 243 244 245 246 247 248 249
/*
 * Magic required such that:
 *
 *	raw_spin_rq_lock(rq);
 *	...
 *	raw_spin_rq_unlock(rq);
 *
 * ends up locking and unlocking the _same_ lock, and all CPUs
 * always agree on what rq has what lock.
 *
 * XXX entirely possible to selectively enable cores, don't bother for now.
 */

static DEFINE_MUTEX(sched_core_mutex);
250
static atomic_t sched_core_count;
Peter Zijlstra's avatar
Peter Zijlstra committed
251 252
static struct cpumask sched_core_mask;

253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272
static void sched_core_lock(int cpu, unsigned long *flags)
{
	const struct cpumask *smt_mask = cpu_smt_mask(cpu);
	int t, i = 0;

	local_irq_save(*flags);
	for_each_cpu(t, smt_mask)
		raw_spin_lock_nested(&cpu_rq(t)->__lock, i++);
}

static void sched_core_unlock(int cpu, unsigned long *flags)
{
	const struct cpumask *smt_mask = cpu_smt_mask(cpu);
	int t;

	for_each_cpu(t, smt_mask)
		raw_spin_unlock(&cpu_rq(t)->__lock);
	local_irq_restore(*flags);
}

Peter Zijlstra's avatar
Peter Zijlstra committed
273 274
static void __sched_core_flip(bool enabled)
{
275 276
	unsigned long flags;
	int cpu, t;
Peter Zijlstra's avatar
Peter Zijlstra committed
277 278 279 280 281 282 283 284 285 286

	cpus_read_lock();

	/*
	 * Toggle the online cores, one by one.
	 */
	cpumask_copy(&sched_core_mask, cpu_online_mask);
	for_each_cpu(cpu, &sched_core_mask) {
		const struct cpumask *smt_mask = cpu_smt_mask(cpu);

287
		sched_core_lock(cpu, &flags);
Peter Zijlstra's avatar
Peter Zijlstra committed
288 289 290 291

		for_each_cpu(t, smt_mask)
			cpu_rq(t)->core_enabled = enabled;

292 293
		cpu_rq(cpu)->core->core_forceidle_start = 0;

294
		sched_core_unlock(cpu, &flags);
Peter Zijlstra's avatar
Peter Zijlstra committed
295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310

		cpumask_andnot(&sched_core_mask, &sched_core_mask, smt_mask);
	}

	/*
	 * Toggle the offline CPUs.
	 */
	cpumask_copy(&sched_core_mask, cpu_possible_mask);
	cpumask_andnot(&sched_core_mask, &sched_core_mask, cpu_online_mask);

	for_each_cpu(cpu, &sched_core_mask)
		cpu_rq(cpu)->core_enabled = enabled;

	cpus_read_unlock();
}

311
static void sched_core_assert_empty(void)
Peter Zijlstra's avatar
Peter Zijlstra committed
312
{
313
	int cpu;
Peter Zijlstra's avatar
Peter Zijlstra committed
314

315 316 317 318 319 320
	for_each_possible_cpu(cpu)
		WARN_ON_ONCE(!RB_EMPTY_ROOT(&cpu_rq(cpu)->core_tree));
}

static void __sched_core_enable(void)
{
Peter Zijlstra's avatar
Peter Zijlstra committed
321 322 323 324 325 326 327
	static_branch_enable(&__sched_core_enabled);
	/*
	 * Ensure all previous instances of raw_spin_rq_*lock() have finished
	 * and future ones will observe !sched_core_disabled().
	 */
	synchronize_rcu();
	__sched_core_flip(true);
328
	sched_core_assert_empty();
Peter Zijlstra's avatar
Peter Zijlstra committed
329 330 331 332
}

static void __sched_core_disable(void)
{
333
	sched_core_assert_empty();
Peter Zijlstra's avatar
Peter Zijlstra committed
334 335 336 337 338 339
	__sched_core_flip(false);
	static_branch_disable(&__sched_core_enabled);
}

void sched_core_get(void)
{
340 341 342
	if (atomic_inc_not_zero(&sched_core_count))
		return;

Peter Zijlstra's avatar
Peter Zijlstra committed
343
	mutex_lock(&sched_core_mutex);
344
	if (!atomic_read(&sched_core_count))
Peter Zijlstra's avatar
Peter Zijlstra committed
345
		__sched_core_enable();
346 347 348

	smp_mb__before_atomic();
	atomic_inc(&sched_core_count);
Peter Zijlstra's avatar
Peter Zijlstra committed
349 350 351
	mutex_unlock(&sched_core_mutex);
}

352
static void __sched_core_put(struct work_struct *work)
Peter Zijlstra's avatar
Peter Zijlstra committed
353
{
354
	if (atomic_dec_and_mutex_lock(&sched_core_count, &sched_core_mutex)) {
Peter Zijlstra's avatar
Peter Zijlstra committed
355
		__sched_core_disable();
356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372
		mutex_unlock(&sched_core_mutex);
	}
}

void sched_core_put(void)
{
	static DECLARE_WORK(_work, __sched_core_put);

	/*
	 * "There can be only one"
	 *
	 * Either this is the last one, or we don't actually need to do any
	 * 'work'. If it is the last *again*, we rely on
	 * WORK_STRUCT_PENDING_BIT.
	 */
	if (!atomic_add_unless(&sched_core_count, -1, 1))
		schedule_work(&_work);
Peter Zijlstra's avatar
Peter Zijlstra committed
373 374
}

375 376 377
#else /* !CONFIG_SCHED_CORE */

static inline void sched_core_enqueue(struct rq *rq, struct task_struct *p) { }
378 379
static inline void
sched_core_dequeue(struct rq *rq, struct task_struct *p, int flags) { }
380

Peter Zijlstra's avatar
Peter Zijlstra committed
381 382
#endif /* CONFIG_SCHED_CORE */

Peter Zijlstra's avatar
Peter Zijlstra committed
383 384 385 386 387
/*
 * 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
388

389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405

/*
 * Serialization rules:
 *
 * Lock order:
 *
 *   p->pi_lock
 *     rq->lock
 *       hrtimer_cpu_base->lock (hrtimer_start() for bandwidth controls)
 *
 *  rq1->lock
 *    rq2->lock  where: rq1 < rq2
 *
 * Regular state:
 *
 * Normal scheduling state is serialized by rq->lock. __schedule() takes the
 * local CPU's rq->lock, it optionally removes the task from the runqueue and
406
 * always looks at the local rq data structures to find the most eligible task
407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482
 * to run next.
 *
 * Task enqueue is also under rq->lock, possibly taken from another CPU.
 * Wakeups from another LLC domain might use an IPI to transfer the enqueue to
 * the local CPU to avoid bouncing the runqueue state around [ see
 * ttwu_queue_wakelist() ]
 *
 * Task wakeup, specifically wakeups that involve migration, are horribly
 * complicated to avoid having to take two rq->locks.
 *
 * Special state:
 *
 * System-calls and anything external will use task_rq_lock() which acquires
 * both p->pi_lock and rq->lock. As a consequence the state they change is
 * stable while holding either lock:
 *
 *  - sched_setaffinity()/
 *    set_cpus_allowed_ptr():	p->cpus_ptr, p->nr_cpus_allowed
 *  - set_user_nice():		p->se.load, p->*prio
 *  - __sched_setscheduler():	p->sched_class, p->policy, p->*prio,
 *				p->se.load, p->rt_priority,
 *				p->dl.dl_{runtime, deadline, period, flags, bw, density}
 *  - sched_setnuma():		p->numa_preferred_nid
 *  - sched_move_task()/
 *    cpu_cgroup_fork():	p->sched_task_group
 *  - uclamp_update_active()	p->uclamp*
 *
 * p->state <- TASK_*:
 *
 *   is changed locklessly using set_current_state(), __set_current_state() or
 *   set_special_state(), see their respective comments, or by
 *   try_to_wake_up(). This latter uses p->pi_lock to serialize against
 *   concurrent self.
 *
 * p->on_rq <- { 0, 1 = TASK_ON_RQ_QUEUED, 2 = TASK_ON_RQ_MIGRATING }:
 *
 *   is set by activate_task() and cleared by deactivate_task(), under
 *   rq->lock. Non-zero indicates the task is runnable, the special
 *   ON_RQ_MIGRATING state is used for migration without holding both
 *   rq->locks. It indicates task_cpu() is not stable, see task_rq_lock().
 *
 * p->on_cpu <- { 0, 1 }:
 *
 *   is set by prepare_task() and cleared by finish_task() such that it will be
 *   set before p is scheduled-in and cleared after p is scheduled-out, both
 *   under rq->lock. Non-zero indicates the task is running on its CPU.
 *
 *   [ The astute reader will observe that it is possible for two tasks on one
 *     CPU to have ->on_cpu = 1 at the same time. ]
 *
 * task_cpu(p): is changed by set_task_cpu(), the rules are:
 *
 *  - Don't call set_task_cpu() on a blocked task:
 *
 *    We don't care what CPU we're not running on, this simplifies hotplug,
 *    the CPU assignment of blocked tasks isn't required to be valid.
 *
 *  - for try_to_wake_up(), called under p->pi_lock:
 *
 *    This allows try_to_wake_up() to only take one rq->lock, see its comment.
 *
 *  - for migration called under rq->lock:
 *    [ see task_on_rq_migrating() in task_rq_lock() ]
 *
 *    o move_queued_task()
 *    o detach_task()
 *
 *  - for migration called under double_rq_lock():
 *
 *    o __migrate_swap_task()
 *    o push_rt_task() / pull_rt_task()
 *    o push_dl_task() / pull_dl_task()
 *    o dl_task_offline_migration()
 *
 */

483 484
void raw_spin_rq_lock_nested(struct rq *rq, int subclass)
{
485 486
	raw_spinlock_t *lock;

Peter Zijlstra's avatar
Peter Zijlstra committed
487 488
	/* Matches synchronize_rcu() in __sched_core_enable() */
	preempt_disable();
489 490
	if (sched_core_disabled()) {
		raw_spin_lock_nested(&rq->__lock, subclass);
Peter Zijlstra's avatar
Peter Zijlstra committed
491 492
		/* preempt_count *MUST* be > 1 */
		preempt_enable_no_resched();
493 494 495 496
		return;
	}

	for (;;) {
497
		lock = __rq_lockp(rq);
498
		raw_spin_lock_nested(lock, subclass);
499
		if (likely(lock == __rq_lockp(rq))) {
Peter Zijlstra's avatar
Peter Zijlstra committed
500 501
			/* preempt_count *MUST* be > 1 */
			preempt_enable_no_resched();
502
			return;
Peter Zijlstra's avatar
Peter Zijlstra committed
503
		}
504 505
		raw_spin_unlock(lock);
	}
506 507 508 509
}

bool raw_spin_rq_trylock(struct rq *rq)
{
510 511 512
	raw_spinlock_t *lock;
	bool ret;

Peter Zijlstra's avatar
Peter Zijlstra committed
513 514 515 516 517 518 519
	/* Matches synchronize_rcu() in __sched_core_enable() */
	preempt_disable();
	if (sched_core_disabled()) {
		ret = raw_spin_trylock(&rq->__lock);
		preempt_enable();
		return ret;
	}
520 521

	for (;;) {
522
		lock = __rq_lockp(rq);
523
		ret = raw_spin_trylock(lock);
524
		if (!ret || (likely(lock == __rq_lockp(rq)))) {
Peter Zijlstra's avatar
Peter Zijlstra committed
525
			preempt_enable();
526
			return ret;
Peter Zijlstra's avatar
Peter Zijlstra committed
527
		}
528 529
		raw_spin_unlock(lock);
	}
530 531 532 533 534 535 536
}

void raw_spin_rq_unlock(struct rq *rq)
{
	raw_spin_unlock(rq_lockp(rq));
}

537 538 539 540 541 542 543 544 545 546 547 548
#ifdef CONFIG_SMP
/*
 * double_rq_lock - safely lock two runqueues
 */
void double_rq_lock(struct rq *rq1, struct rq *rq2)
{
	lockdep_assert_irqs_disabled();

	if (rq_order_less(rq2, rq1))
		swap(rq1, rq2);

	raw_spin_rq_lock(rq1);
549
	if (__rq_lockp(rq1) == __rq_lockp(rq2))
550 551 552 553 554 555
		return;

	raw_spin_rq_lock_nested(rq2, SINGLE_DEPTH_NESTING);
}
#endif

556 557 558
/*
 * __task_rq_lock - lock the rq @p resides on.
 */
559
struct rq *__task_rq_lock(struct task_struct *p, struct rq_flags *rf)
560 561 562 563 564 565 566 567
	__acquires(rq->lock)
{
	struct rq *rq;

	lockdep_assert_held(&p->pi_lock);

	for (;;) {
		rq = task_rq(p);
568
		raw_spin_rq_lock(rq);
569
		if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) {
570
			rq_pin_lock(rq, rf);
571 572
			return rq;
		}
573
		raw_spin_rq_unlock(rq);
574 575 576 577 578 579 580 581 582

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

/*
 * task_rq_lock - lock p->pi_lock and lock the rq @p resides on.
 */
583
struct rq *task_rq_lock(struct task_struct *p, struct rq_flags *rf)
584 585 586 587 588 589
	__acquires(p->pi_lock)
	__acquires(rq->lock)
{
	struct rq *rq;

	for (;;) {
590
		raw_spin_lock_irqsave(&p->pi_lock, rf->flags);
591
		rq = task_rq(p);
592
		raw_spin_rq_lock(rq);
593 594 595 596 597 598 599 600 601 602
		/*
		 *	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)
		 *
603
		 * If we observe the old CPU in task_rq_lock(), the acquire of
604 605
		 * the old rq->lock will fully serialize against the stores.
		 *
606 607 608
		 * 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.
609 610
		 */
		if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) {
611
			rq_pin_lock(rq, rf);
612 613
			return rq;
		}
614
		raw_spin_rq_unlock(rq);
615
		raw_spin_unlock_irqrestore(&p->pi_lock, rf->flags);
616 617 618 619 620 621

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

622 623 624 625 626 627 628 629 630 631
/*
 * 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...
 */
632 633
	s64 __maybe_unused steal = 0, irq_delta = 0;

634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672
#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;

673
#ifdef CONFIG_HAVE_SCHED_AVG_IRQ
674
	if ((irq_delta + steal) && sched_feat(NONTASK_CAPACITY))
675
		update_irq_load_avg(rq, irq_delta + steal);
676
#endif
677
	update_rq_clock_pelt(rq, delta);
678 679 680 681 682 683
}

void update_rq_clock(struct rq *rq)
{
	s64 delta;

684
	lockdep_assert_rq_held(rq);
685 686 687 688 689

	if (rq->clock_update_flags & RQCF_ACT_SKIP)
		return;

#ifdef CONFIG_SCHED_DEBUG
690 691
	if (sched_feat(WARN_DOUBLE_CLOCK))
		SCHED_WARN_ON(rq->clock_update_flags & RQCF_UPDATED);
692 693
	rq->clock_update_flags |= RQCF_UPDATED;
#endif
694

695 696 697 698 699 700 701
	delta = sched_clock_cpu(cpu_of(rq)) - rq->clock;
	if (delta < 0)
		return;
	rq->clock += delta;
	update_rq_clock_task(rq, delta);
}

702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719
#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);
720
	struct rq_flags rf;
721 722 723

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

724
	rq_lock(rq, &rf);
725
	update_rq_clock(rq);
726
	rq->curr->sched_class->task_tick(rq, rq->curr, 1);
727
	rq_unlock(rq, &rf);
728 729 730 731

	return HRTIMER_NORESTART;
}

732
#ifdef CONFIG_SMP
Peter Zijlstra's avatar
Peter Zijlstra committed
733

734
static void __hrtick_restart(struct rq *rq)
Peter Zijlstra's avatar
Peter Zijlstra committed
735 736
{
	struct hrtimer *timer = &rq->hrtick_timer;
737
	ktime_t time = rq->hrtick_time;
Peter Zijlstra's avatar
Peter Zijlstra committed
738

739
	hrtimer_start(timer, time, HRTIMER_MODE_ABS_PINNED_HARD);
Peter Zijlstra's avatar
Peter Zijlstra committed
740 741
}

742 743 744 745
/*
 * called from hardirq (IPI) context
 */
static void __hrtick_start(void *arg)
746
{
747
	struct rq *rq = arg;
748
	struct rq_flags rf;
749

750
	rq_lock(rq, &rf);
Peter Zijlstra's avatar
Peter Zijlstra committed
751
	__hrtick_restart(rq);
752
	rq_unlock(rq, &rf);
753 754
}

755 756 757 758 759
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
760
void hrtick_start(struct rq *rq, u64 delay)
761
{
762
	struct hrtimer *timer = &rq->hrtick_timer;
763 764 765 766 767 768 769
	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);
770
	rq->hrtick_time = ktime_add_ns(timer->base->get_time(), delta);
771

772
	if (rq == this_rq())
Peter Zijlstra's avatar
Peter Zijlstra committed
773
		__hrtick_restart(rq);
774
	else
775
		smp_call_function_single_async(cpu_of(rq), &rq->hrtick_csd);
776 777
}

778 779 780 781 782 783
#else
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
784
void hrtick_start(struct rq *rq, u64 delay)
785
{
786 787 788 789 790
	/*
	 * Don't schedule slices shorter than 10000ns, that just
	 * doesn't make sense. Rely on vruntime for fairness.
	 */
	delay = max_t(u64, delay, 10000LL);
791
	hrtimer_start(&rq->hrtick_timer, ns_to_ktime(delay),
792
		      HRTIMER_MODE_REL_PINNED_HARD);
793
}
794

795
#endif /* CONFIG_SMP */
796

797
static void hrtick_rq_init(struct rq *rq)
798
{
799
#ifdef CONFIG_SMP
800
	INIT_CSD(&rq->hrtick_csd, __hrtick_start, rq);
801
#endif
802
	hrtimer_init(&rq->hrtick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL_HARD);
803
	rq->hrtick_timer.function = hrtick;
804
}
Andrew Morton's avatar
Andrew Morton committed
805
#else	/* CONFIG_SCHED_HRTICK */
806 807 808 809
static inline void hrtick_clear(struct rq *rq)
{
}

810
static inline void hrtick_rq_init(struct rq *rq)
811 812
{
}
Andrew Morton's avatar
Andrew Morton committed
813
#endif	/* CONFIG_SCHED_HRTICK */
814

815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832
/*
 * 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;								\
})

833
#if defined(CONFIG_SMP) && defined(TIF_POLLING_NRFLAG)
834 835 836 837 838 839 840 841 842 843
/*
 * 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);
}
844 845 846 847 848 849 850 851 852 853

/*
 * 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);
854
	typeof(ti->flags) old, val = READ_ONCE(ti->flags);
855 856 857 858 859 860 861 862 863 864 865 866 867 868

	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;
}

869 870 871 872 873 874
#else
static bool set_nr_and_not_polling(struct task_struct *p)
{
	set_tsk_need_resched(p);
	return true;
}
875 876 877 878 879 880 881

#ifdef CONFIG_SMP
static bool set_nr_if_polling(struct task_struct *p)
{
	return false;
}
#endif
882 883
#endif

884
static bool __wake_q_add(struct wake_q_head *head, struct task_struct *task)
885 886 887 888 889
{
	struct wake_q_node *node = &task->wake_q;

	/*
	 * Atomically grab the task, if ->wake_q is !nil already it means
890
	 * it's already queued (either by us or someone else) and will get the
891 892
	 * wakeup due to that.
	 *
893 894
	 * 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.
895
	 */
896
	smp_mb__before_atomic();
897
	if (unlikely(cmpxchg_relaxed(&node->next, NULL, WAKE_Q_TAIL)))
898
		return false;
899 900 901 902 903 904

	/*
	 * The head is context local, there can be no concurrency.
	 */
	*head->lastp = node;
	head->lastp = &node->next;
905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946
	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);
947 948 949 950 951 952 953 954 955 956
}

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);
957
		/* Task can safely be re-inserted now: */
958 959 960 961
		node = node->next;
		task->wake_q.next = NULL;

		/*
962 963
		 * wake_up_process() executes a full barrier, which pairs with
		 * the queueing in wake_q_add() so as not to miss wakeups.
964 965 966 967 968 969
		 */
		wake_up_process(task);
		put_task_struct(task);
	}
}

970
/*
971
 * resched_curr - mark rq's current task 'to be rescheduled now'.
972 973 974 975 976
 *
 * 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.
 */
977
void resched_curr(struct rq *rq)
978
{
979
	struct task_struct *curr = rq->curr;
980 981
	int cpu;

982
	lockdep_assert_rq_held(rq);
983

984
	if (test_tsk_need_resched(curr))
985 986
		return;

987
	cpu = cpu_of(rq);
988

989
	if (cpu == smp_processor_id()) {
990
		set_tsk_need_resched(curr);
991
		set_preempt_need_resched();
992
		return;
993
	}
994

995
	if (set_nr_and_not_polling(curr))
996
		smp_send_reschedule(cpu);
997 998
	else
		trace_sched_wake_idle_without_ipi(cpu);
999 1000
}

1001
void resched_cpu(int cpu)
1002 1003 1004 1005
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

1006
	raw_spin_rq_lock_irqsave(rq, flags);
1007 1008
	if (cpu_online(cpu) || cpu == smp_processor_id())
		resched_curr(rq);
1009
	raw_spin_rq_unlock_irqrestore(rq, flags);
1010
}
1011

1012
#ifdef CONFIG_SMP
1013
#ifdef CONFIG_NO_HZ_COMMON
1014
/*
1015 1016
 * In the semi idle case, use the nearest busy CPU for migrating timers
 * from an idle CPU.  This is good for power-savings.
1017 1018
 *
 * We don't do similar optimization for completely idle system, as
1019 1020
 * 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).
1021
 */
1022
int get_nohz_timer_target(void)
1023
{
1024
	int i, cpu = smp_processor_id(), default_cpu = -1;
1025
	struct sched_domain *sd;
1026
	const struct cpumask *hk_mask;
1027

1028
	if (housekeeping_cpu(cpu, HK_TYPE_TIMER)) {
1029 1030 1031 1032
		if (!idle_cpu(cpu))
			return cpu;
		default_cpu = cpu;
	}
1033

1034
	hk_mask = housekeeping_cpumask(HK_TYPE_TIMER);
1035

1036
	rcu_read_lock();
1037
	for_each_domain(cpu, sd) {
1038
		for_each_cpu_and(i, sched_domain_span(sd), hk_mask) {
1039 1040 1041
			if (cpu == i)
				continue;

1042
			if (!idle_cpu(i)) {
1043 1044 1045 1046
				cpu = i;
				goto unlock;
			}
		}
1047
	}
1048

1049
	if (default_cpu == -1)
1050
		default_cpu = housekeeping_any_cpu(HK_TYPE_TIMER);
1051
	cpu = default_cpu;
1052 1053
unlock:
	rcu_read_unlock();
1054 1055
	return cpu;
}
1056

1057 1058 1059 1060 1061 1062 1063 1064 1065 1066
/*
 * 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.
 */
1067
static void wake_up_idle_cpu(int cpu)
1068 1069 1070 1071 1072 1073
{
	struct rq *rq = cpu_rq(cpu);

	if (cpu == smp_processor_id())
		return;

1074
	if (set_nr_and_not_polling(rq->idle))
1075
		smp_send_reschedule(cpu);
1076 1077
	else
		trace_sched_wake_idle_without_ipi(cpu);
1078 1079
}

1080
static bool wake_up_full_nohz_cpu(int cpu)
1081
{
1082 1083 1084 1085 1086 1087
	/*
	 * 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.
	 */
1088 1089
	if (cpu_is_offline(cpu))
		return true;  /* Don't try to wake offline CPUs. */
1090
	if (tick_nohz_full_cpu(cpu)) {
1091 1092
		if (cpu != smp_processor_id() ||
		    tick_nohz_tick_stopped())
1093
			tick_nohz_full_kick_cpu(cpu);
1094 1095 1096 1097 1098 1099
		return true;
	}

	return false;
}

1100 1101 1102 1103 1104
/*
 * 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.
 */
1105 1106
void wake_up_nohz_cpu(int cpu)
{
1107
	if (!wake_up_full_nohz_cpu(cpu))
1108 1109 1110
		wake_up_idle_cpu(cpu);
}

1111
static void nohz_csd_func(void *info)
1112
{
1113 1114 1115
	struct rq *rq = info;
	int cpu = cpu_of(rq);
	unsigned int flags;
1116 1117

	/*
1118
	 * Release the rq::nohz_csd.
1119
	 */
1120
	flags = atomic_fetch_andnot(NOHZ_KICK_MASK | NOHZ_NEWILB_KICK, nohz_flags(cpu));
1121
	WARN_ON(!(flags & NOHZ_KICK_MASK));
1122

1123 1124 1125
	rq->idle_balance = idle_cpu(cpu);
	if (rq->idle_balance && !need_resched()) {
		rq->nohz_idle_balance = flags;
1126 1127
		raise_softirq_irqoff(SCHED_SOFTIRQ);
	}
1128 1129
}

1130
#endif /* CONFIG_NO_HZ_COMMON */
1131

1132
#ifdef CONFIG_NO_HZ_FULL
1133
bool sched_can_stop_tick(struct rq *rq)
1134
{
1135 1136 1137 1138 1139 1140
	int fifo_nr_running;

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

1141
	/*
1142
	 * If there are more than one RR tasks, we need the tick to affect the
1143
	 * actual RR behaviour.
1144
	 */
1145 1146 1147 1148 1149
	if (rq->rt.rr_nr_running) {
		if (rq->rt.rr_nr_running == 1)
			return true;
		else
			return false;
1150 1151
	}

1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165
	/*
	 * 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)
1166
		return false;
1167

1168
	return true;
1169 1170
}
#endif /* CONFIG_NO_HZ_FULL */
1171
#endif /* CONFIG_SMP */
1172

1173 1174
#if defined(CONFIG_RT_GROUP_SCHED) || (defined(CONFIG_FAIR_GROUP_SCHED) && \
			(defined(CONFIG_SMP) || defined(CONFIG_CFS_BANDWIDTH)))
1175
/*
1176 1177 1178 1179
 * 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.
1180
 */
1181
int walk_tg_tree_from(struct task_group *from,
1182
			     tg_visitor down, tg_visitor up, void *data)
1183 1184
{
	struct task_group *parent, *child;
1185
	int ret;
1186

1187 1188
	parent = from;

1189
down:
1190 1191
	ret = (*down)(parent, data);
	if (ret)
1192
		goto out;
1193 1194 1195 1196 1197 1198 1199
	list_for_each_entry_rcu(child, &parent->children, siblings) {
		parent = child;
		goto down;

up:
		continue;
	}
1200
	ret = (*up)(parent, data);
1201 1202
	if (ret || parent == from)
		goto out;
1203 1204 1205 1206 1207

	child = parent;
	parent = parent->parent;
	if (parent)
		goto up;
1208
out:
1209
	return ret;
1210 1211
}

1212
int tg_nop(struct task_group *tg, void *data)
1213
{
1214
	return 0;
1215
}
1216 1217
#endif

1218
static void set_load_weight(struct task_struct *p, bool update_load)
1219
{
1220 1221 1222
	int prio = p->static_prio - MAX_RT_PRIO;
	struct load_weight *load = &p->se.load;

Ingo Molnar's avatar
Ingo Molnar committed
1223 1224 1225
	/*
	 * SCHED_IDLE tasks get minimal weight:
	 */
1226
	if (task_has_idle_policy(p)) {
1227
		load->weight = scale_load(WEIGHT_IDLEPRIO);
1228
		load->inv_weight = WMULT_IDLEPRIO;
Ingo Molnar's avatar
Ingo Molnar committed
1229 1230
		return;
	}
1231

1232 1233 1234 1235 1236 1237 1238 1239 1240 1241
	/*
	 * 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];
	}
1242 1243
}

1244
#ifdef CONFIG_UCLAMP_TASK
1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256
/*
 * 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);

1257 1258 1259 1260 1261 1262
/* 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;

1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279
/*
 * By default RT tasks run at the maximum performance point/capacity of the
 * system. Uclamp enforces this by always setting UCLAMP_MIN of RT tasks to
 * SCHED_CAPACITY_SCALE.
 *
 * This knob allows admins to change the default behavior when uclamp is being
 * used. In battery powered devices, particularly, running at the maximum
 * capacity and frequency will increase energy consumption and shorten the
 * battery life.
 *
 * This knob only affects RT tasks that their uclamp_se->user_defined == false.
 *
 * This knob will not override the system default sched_util_clamp_min defined
 * above.
 */
unsigned int sysctl_sched_uclamp_util_min_rt_default = SCHED_CAPACITY_SCALE;

1280 1281
/* All clamps are required to be less or equal than these values */
static struct uclamp_se uclamp_default[UCLAMP_CNT];
1282

1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302
/*
 * This static key is used to reduce the uclamp overhead in the fast path. It
 * primarily disables the call to uclamp_rq_{inc, dec}() in
 * enqueue/dequeue_task().
 *
 * This allows users to continue to enable uclamp in their kernel config with
 * minimum uclamp overhead in the fast path.
 *
 * As soon as userspace modifies any of the uclamp knobs, the static key is
 * enabled, since we have an actual users that make use of uclamp
 * functionality.
 *
 * The knobs that would enable this static key are:
 *
 *   * A task modifying its uclamp value with sched_setattr().
 *   * An admin modifying the sysctl_sched_uclamp_{min, max} via procfs.
 *   * An admin modifying the cgroup cpu.uclamp.{min, max}
 */
DEFINE_STATIC_KEY_FALSE(sched_uclamp_used);

1303 1304 1305 1306 1307 1308 1309 1310
/* 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)
{
1311
	return min_t(unsigned int, clamp_value / UCLAMP_BUCKET_DELTA, UCLAMP_BUCKETS - 1);
1312 1313
}

1314
static inline unsigned int uclamp_none(enum uclamp_id clamp_id)
1315 1316 1317 1318 1319 1320
{
	if (clamp_id == UCLAMP_MIN)
		return 0;
	return SCHED_CAPACITY_SCALE;
}

1321 1322
static inline void uclamp_se_set(struct uclamp_se *uc_se,
				 unsigned int value, bool user_defined)
1323 1324 1325
{
	uc_se->value = value;
	uc_se->bucket_id = uclamp_bucket_id(value);
1326
	uc_se->user_defined = user_defined;
1327 1328
}

1329
static inline unsigned int
1330
uclamp_idle_value(struct rq *rq, enum uclamp_id clamp_id,
1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345
		  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);
}

1346
static inline void uclamp_idle_reset(struct rq *rq, enum uclamp_id clamp_id,
1347 1348 1349 1350 1351 1352 1353 1354 1355
				     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);
}

1356
static inline
1357
unsigned int uclamp_rq_max_value(struct rq *rq, enum uclamp_id clamp_id,
1358
				   unsigned int clamp_value)
1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373
{
	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 */
1374
	return uclamp_idle_value(rq, clamp_id, clamp_value);
1375 1376
}

1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434
static void __uclamp_update_util_min_rt_default(struct task_struct *p)
{
	unsigned int default_util_min;
	struct uclamp_se *uc_se;

	lockdep_assert_held(&p->pi_lock);

	uc_se = &p->uclamp_req[UCLAMP_MIN];

	/* Only sync if user didn't override the default */
	if (uc_se->user_defined)
		return;

	default_util_min = sysctl_sched_uclamp_util_min_rt_default;
	uclamp_se_set(uc_se, default_util_min, false);
}

static void uclamp_update_util_min_rt_default(struct task_struct *p)
{
	struct rq_flags rf;
	struct rq *rq;

	if (!rt_task(p))
		return;

	/* Protect updates to p->uclamp_* */
	rq = task_rq_lock(p, &rf);
	__uclamp_update_util_min_rt_default(p);
	task_rq_unlock(rq, p, &rf);
}

static void uclamp_sync_util_min_rt_default(void)
{
	struct task_struct *g, *p;

	/*
	 * copy_process()			sysctl_uclamp
	 *					  uclamp_min_rt = X;
	 *   write_lock(&tasklist_lock)		  read_lock(&tasklist_lock)
	 *   // link thread			  smp_mb__after_spinlock()
	 *   write_unlock(&tasklist_lock)	  read_unlock(&tasklist_lock);
	 *   sched_post_fork()			  for_each_process_thread()
	 *     __uclamp_sync_rt()		    __uclamp_sync_rt()
	 *
	 * Ensures that either sched_post_fork() will observe the new
	 * uclamp_min_rt or for_each_process_thread() will observe the new
	 * task.
	 */
	read_lock(&tasklist_lock);
	smp_mb__after_spinlock();
	read_unlock(&tasklist_lock);

	rcu_read_lock();
	for_each_process_thread(g, p)
		uclamp_update_util_min_rt_default(p);
	rcu_read_unlock();
}

1435
static inline struct uclamp_se
1436
uclamp_tg_restrict(struct task_struct *p, enum uclamp_id clamp_id)
1437
{
1438
	/* Copy by value as we could modify it */
1439 1440
	struct uclamp_se uc_req = p->uclamp_req[clamp_id];
#ifdef CONFIG_UCLAMP_TASK_GROUP
1441
	unsigned int tg_min, tg_max, value;
1442 1443 1444 1445 1446 1447 1448 1449 1450 1451

	/*
	 * 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;

1452 1453 1454 1455 1456
	tg_min = task_group(p)->uclamp[UCLAMP_MIN].value;
	tg_max = task_group(p)->uclamp[UCLAMP_MAX].value;
	value = uc_req.value;
	value = clamp(value, tg_min, tg_max);
	uclamp_se_set(&uc_req, value, false);
1457 1458 1459 1460 1461
#endif

	return uc_req;
}

1462 1463 1464 1465
/*
 * The effective clamp bucket index of a task depends on, by increasing
 * priority:
 * - the task specific clamp value, when explicitly requested from userspace
1466 1467
 * - the task group effective clamp value, for tasks not either in the root
 *   group or in an autogroup
1468 1469 1470
 * - the system default clamp value, defined by the sysadmin
 */
static inline struct uclamp_se
1471
uclamp_eff_get(struct task_struct *p, enum uclamp_id clamp_id)
1472
{
1473
	struct uclamp_se uc_req = uclamp_tg_restrict(p, clamp_id);
1474 1475 1476 1477 1478 1479 1480 1481 1482
	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;
}

1483
unsigned long uclamp_eff_value(struct task_struct *p, enum uclamp_id clamp_id)
1484 1485 1486 1487 1488
{
	struct uclamp_se uc_eff;

	/* Task currently refcounted: use back-annotated (effective) value */
	if (p->uclamp[clamp_id].active)
1489
		return (unsigned long)p->uclamp[clamp_id].value;
1490 1491 1492

	uc_eff = uclamp_eff_get(p, clamp_id);

1493
	return (unsigned long)uc_eff.value;
1494 1495
}

1496 1497 1498 1499
/*
 * 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.
1500 1501 1502 1503 1504
 *
 * 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.
1505 1506
 */
static inline void uclamp_rq_inc_id(struct rq *rq, struct task_struct *p,
1507
				    enum uclamp_id clamp_id)
1508 1509 1510 1511 1512
{
	struct uclamp_rq *uc_rq = &rq->uclamp[clamp_id];
	struct uclamp_se *uc_se = &p->uclamp[clamp_id];
	struct uclamp_bucket *bucket;

1513
	lockdep_assert_rq_held(rq);
1514

1515 1516 1517
	/* Update task effective clamp */
	p->uclamp[clamp_id] = uclamp_eff_get(p, clamp_id);

1518 1519
	bucket = &uc_rq->bucket[uc_se->bucket_id];
	bucket->tasks++;
1520
	uc_se->active = true;
1521

1522 1523
	uclamp_idle_reset(rq, clamp_id, uc_se->value);

1524 1525 1526 1527 1528 1529 1530
	/*
	 * 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;

1531
	if (uc_se->value > READ_ONCE(uc_rq->value))
1532
		WRITE_ONCE(uc_rq->value, uc_se->value);
1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544
}

/*
 * 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,
1545
				    enum uclamp_id clamp_id)
1546 1547 1548 1549
{
	struct uclamp_rq *uc_rq = &rq->uclamp[clamp_id];
	struct uclamp_se *uc_se = &p->uclamp[clamp_id];
	struct uclamp_bucket *bucket;
1550
	unsigned int bkt_clamp;
1551 1552
	unsigned int rq_clamp;

1553
	lockdep_assert_rq_held(rq);
1554

1555 1556 1557 1558 1559 1560 1561 1562
	/*
	 * If sched_uclamp_used was enabled after task @p was enqueued,
	 * we could end up with unbalanced call to uclamp_rq_dec_id().
	 *
	 * In this case the uc_se->active flag should be false since no uclamp
	 * accounting was performed at enqueue time and we can just return
	 * here.
	 *
1563
	 * Need to be careful of the following enqueue/dequeue ordering
1564 1565 1566 1567 1568 1569
	 * problem too
	 *
	 *	enqueue(taskA)
	 *	// sched_uclamp_used gets enabled
	 *	enqueue(taskB)
	 *	dequeue(taskA)
1570
	 *	// Must not decrement bucket->tasks here
1571 1572 1573 1574 1575 1576 1577 1578 1579 1580
	 *	dequeue(taskB)
	 *
	 * where we could end up with stale data in uc_se and
	 * bucket[uc_se->bucket_id].
	 *
	 * The following check here eliminates the possibility of such race.
	 */
	if (unlikely(!uc_se->active))
		return;

1581
	bucket = &uc_rq->bucket[uc_se->bucket_id];
1582

1583 1584 1585
	SCHED_WARN_ON(!bucket->tasks);
	if (likely(bucket->tasks))
		bucket->tasks--;
1586

1587
	uc_se->active = false;
1588

1589 1590 1591 1592 1593 1594
	/*
	 * 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.
	 */
1595 1596 1597 1598 1599 1600 1601 1602 1603
	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);
1604 1605 1606 1607
	if (bucket->value >= rq_clamp) {
		bkt_clamp = uclamp_rq_max_value(rq, clamp_id, uc_se->value);
		WRITE_ONCE(uc_rq->value, bkt_clamp);
	}
1608 1609 1610 1611
}

static inline void uclamp_rq_inc(struct rq *rq, struct task_struct *p)
{
1612
	enum uclamp_id clamp_id;
1613

1614 1615 1616 1617 1618 1619 1620 1621 1622
	/*
	 * Avoid any overhead until uclamp is actually used by the userspace.
	 *
	 * The condition is constructed such that a NOP is generated when
	 * sched_uclamp_used is disabled.
	 */
	if (!static_branch_unlikely(&sched_uclamp_used))
		return;

1623 1624 1625 1626 1627
	if (unlikely(!p->sched_class->uclamp_enabled))
		return;

	for_each_clamp_id(clamp_id)
		uclamp_rq_inc_id(rq, p, clamp_id);
1628 1629 1630 1631

	/* Reset clamp idle holding when there is one RUNNABLE task */
	if (rq->uclamp_flags & UCLAMP_FLAG_IDLE)
		rq->uclamp_flags &= ~UCLAMP_FLAG_IDLE;
1632 1633 1634 1635
}

static inline void uclamp_rq_dec(struct rq *rq, struct task_struct *p)
{
1636
	enum uclamp_id clamp_id;
1637

1638 1639 1640 1641 1642 1643 1644 1645 1646
	/*
	 * Avoid any overhead until uclamp is actually used by the userspace.
	 *
	 * The condition is constructed such that a NOP is generated when
	 * sched_uclamp_used is disabled.
	 */
	if (!static_branch_unlikely(&sched_uclamp_used))
		return;

1647 1648 1649 1650 1651 1652 1653
	if (unlikely(!p->sched_class->uclamp_enabled))
		return;

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

1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670
static inline void uclamp_rq_reinc_id(struct rq *rq, struct task_struct *p,
				      enum uclamp_id clamp_id)
{
	if (!p->uclamp[clamp_id].active)
		return;

	uclamp_rq_dec_id(rq, p, clamp_id);
	uclamp_rq_inc_id(rq, p, clamp_id);

	/*
	 * Make sure to clear the idle flag if we've transiently reached 0
	 * active tasks on rq.
	 */
	if (clamp_id == UCLAMP_MAX && (rq->uclamp_flags & UCLAMP_FLAG_IDLE))
		rq->uclamp_flags &= ~UCLAMP_FLAG_IDLE;
}

1671
static inline void
1672
uclamp_update_active(struct task_struct *p)
1673
{
1674
	enum uclamp_id clamp_id;
1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693
	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.
	 */
1694 1695
	for_each_clamp_id(clamp_id)
		uclamp_rq_reinc_id(rq, p, clamp_id);
1696 1697 1698 1699

	task_rq_unlock(rq, p, &rf);
}

1700
#ifdef CONFIG_UCLAMP_TASK_GROUP
1701
static inline void
1702
uclamp_update_active_tasks(struct cgroup_subsys_state *css)
1703 1704 1705 1706 1707
{
	struct css_task_iter it;
	struct task_struct *p;

	css_task_iter_start(css, 0, &it);
1708 1709
	while ((p = css_task_iter_next(&it)))
		uclamp_update_active(p);
1710 1711 1712
	css_task_iter_end(&it);
}

1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730
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

1731
int sysctl_sched_uclamp_handler(struct ctl_table *table, int write,
1732
				void *buffer, size_t *lenp, loff_t *ppos)
1733
{
1734
	bool update_root_tg = false;
1735
	int old_min, old_max, old_min_rt;
1736 1737
	int result;

1738
	mutex_lock(&uclamp_mutex);
1739 1740
	old_min = sysctl_sched_uclamp_util_min;
	old_max = sysctl_sched_uclamp_util_max;
1741
	old_min_rt = sysctl_sched_uclamp_util_min_rt_default;
1742 1743 1744 1745 1746 1747 1748 1749

	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 ||
1750 1751 1752
	    sysctl_sched_uclamp_util_max > SCHED_CAPACITY_SCALE	||
	    sysctl_sched_uclamp_util_min_rt_default > SCHED_CAPACITY_SCALE) {

1753 1754 1755 1756 1757 1758
		result = -EINVAL;
		goto undo;
	}

	if (old_min != sysctl_sched_uclamp_util_min) {
		uclamp_se_set(&uclamp_default[UCLAMP_MIN],
1759
			      sysctl_sched_uclamp_util_min, false);
1760
		update_root_tg = true;
1761 1762 1763
	}
	if (old_max != sysctl_sched_uclamp_util_max) {
		uclamp_se_set(&uclamp_default[UCLAMP_MAX],
1764
			      sysctl_sched_uclamp_util_max, false);
1765
		update_root_tg = true;
1766 1767
	}

1768 1769
	if (update_root_tg) {
		static_branch_enable(&sched_uclamp_used);
1770
		uclamp_update_root_tg();
1771
	}
1772

1773 1774 1775 1776
	if (old_min_rt != sysctl_sched_uclamp_util_min_rt_default) {
		static_branch_enable(&sched_uclamp_used);
		uclamp_sync_util_min_rt_default();
	}
1777

1778
	/*
1779 1780 1781
	 * 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.
1782
	 */
1783

1784 1785 1786 1787 1788
	goto done;

undo:
	sysctl_sched_uclamp_util_min = old_min;
	sysctl_sched_uclamp_util_max = old_max;
1789
	sysctl_sched_uclamp_util_min_rt_default = old_min_rt;
1790
done:
1791
	mutex_unlock(&uclamp_mutex);
1792 1793 1794 1795

	return result;
}

1796 1797 1798
static int uclamp_validate(struct task_struct *p,
			   const struct sched_attr *attr)
{
1799 1800
	int util_min = p->uclamp_req[UCLAMP_MIN].value;
	int util_max = p->uclamp_req[UCLAMP_MAX].value;
1801

1802 1803
	if (attr->sched_flags & SCHED_FLAG_UTIL_CLAMP_MIN) {
		util_min = attr->sched_util_min;
1804

1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816
		if (util_min + 1 > SCHED_CAPACITY_SCALE + 1)
			return -EINVAL;
	}

	if (attr->sched_flags & SCHED_FLAG_UTIL_CLAMP_MAX) {
		util_max = attr->sched_util_max;

		if (util_max + 1 > SCHED_CAPACITY_SCALE + 1)
			return -EINVAL;
	}

	if (util_min != -1 && util_max != -1 && util_min > util_max)
1817 1818
		return -EINVAL;

1819 1820 1821 1822 1823 1824 1825 1826 1827
	/*
	 * We have valid uclamp attributes; make sure uclamp is enabled.
	 *
	 * We need to do that here, because enabling static branches is a
	 * blocking operation which obviously cannot be done while holding
	 * scheduler locks.
	 */
	static_branch_enable(&sched_uclamp_used);

1828 1829 1830
	return 0;
}

1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855
static bool uclamp_reset(const struct sched_attr *attr,
			 enum uclamp_id clamp_id,
			 struct uclamp_se *uc_se)
{
	/* Reset on sched class change for a non user-defined clamp value. */
	if (likely(!(attr->sched_flags & SCHED_FLAG_UTIL_CLAMP)) &&
	    !uc_se->user_defined)
		return true;

	/* Reset on sched_util_{min,max} == -1. */
	if (clamp_id == UCLAMP_MIN &&
	    attr->sched_flags & SCHED_FLAG_UTIL_CLAMP_MIN &&
	    attr->sched_util_min == -1) {
		return true;
	}

	if (clamp_id == UCLAMP_MAX &&
	    attr->sched_flags & SCHED_FLAG_UTIL_CLAMP_MAX &&
	    attr->sched_util_max == -1) {
		return true;
	}

	return false;
}

1856 1857 1858
static void __setscheduler_uclamp(struct task_struct *p,
				  const struct sched_attr *attr)
{
1859
	enum uclamp_id clamp_id;
1860 1861 1862

	for_each_clamp_id(clamp_id) {
		struct uclamp_se *uc_se = &p->uclamp_req[clamp_id];
1863
		unsigned int value;
1864

1865
		if (!uclamp_reset(attr, clamp_id, uc_se))
1866 1867
			continue;

1868 1869 1870 1871
		/*
		 * RT by default have a 100% boost value that could be modified
		 * at runtime.
		 */
1872
		if (unlikely(rt_task(p) && clamp_id == UCLAMP_MIN))
1873
			value = sysctl_sched_uclamp_util_min_rt_default;
1874
		else
1875 1876 1877
			value = uclamp_none(clamp_id);

		uclamp_se_set(uc_se, value, false);
1878 1879 1880

	}

1881 1882 1883
	if (likely(!(attr->sched_flags & SCHED_FLAG_UTIL_CLAMP)))
		return;

1884 1885
	if (attr->sched_flags & SCHED_FLAG_UTIL_CLAMP_MIN &&
	    attr->sched_util_min != -1) {
1886 1887 1888 1889
		uclamp_se_set(&p->uclamp_req[UCLAMP_MIN],
			      attr->sched_util_min, true);
	}

1890 1891
	if (attr->sched_flags & SCHED_FLAG_UTIL_CLAMP_MAX &&
	    attr->sched_util_max != -1) {
1892 1893 1894 1895 1896
		uclamp_se_set(&p->uclamp_req[UCLAMP_MAX],
			      attr->sched_util_max, true);
	}
}

1897 1898
static void uclamp_fork(struct task_struct *p)
{
1899
	enum uclamp_id clamp_id;
1900

1901 1902 1903 1904
	/*
	 * We don't need to hold task_rq_lock() when updating p->uclamp_* here
	 * as the task is still at its early fork stages.
	 */
1905 1906
	for_each_clamp_id(clamp_id)
		p->uclamp[clamp_id].active = false;
1907 1908 1909 1910 1911

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

	for_each_clamp_id(clamp_id) {
1912 1913
		uclamp_se_set(&p->uclamp_req[clamp_id],
			      uclamp_none(clamp_id), false);
1914
	}
1915 1916
}

1917 1918 1919 1920 1921
static void uclamp_post_fork(struct task_struct *p)
{
	uclamp_update_util_min_rt_default(p);
}

1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932
static void __init init_uclamp_rq(struct rq *rq)
{
	enum uclamp_id clamp_id;
	struct uclamp_rq *uc_rq = rq->uclamp;

	for_each_clamp_id(clamp_id) {
		uc_rq[clamp_id] = (struct uclamp_rq) {
			.value = uclamp_none(clamp_id)
		};
	}

1933
	rq->uclamp_flags = UCLAMP_FLAG_IDLE;
1934 1935
}

1936 1937
static void __init init_uclamp(void)
{
1938
	struct uclamp_se uc_max = {};
1939
	enum uclamp_id clamp_id;
1940 1941
	int cpu;

1942 1943
	for_each_possible_cpu(cpu)
		init_uclamp_rq(cpu_rq(cpu));
1944 1945

	for_each_clamp_id(clamp_id) {
1946
		uclamp_se_set(&init_task.uclamp_req[clamp_id],
1947
			      uclamp_none(clamp_id), false);
1948
	}
1949 1950

	/* System defaults allow max clamp values for both indexes */
1951
	uclamp_se_set(&uc_max, uclamp_none(UCLAMP_MAX), false);
1952
	for_each_clamp_id(clamp_id) {
1953
		uclamp_default[clamp_id] = uc_max;
1954 1955
#ifdef CONFIG_UCLAMP_TASK_GROUP
		root_task_group.uclamp_req[clamp_id] = uc_max;
1956
		root_task_group.uclamp[clamp_id] = uc_max;
1957 1958
#endif
	}
1959 1960 1961 1962 1963
}

#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) { }
1964 1965 1966 1967 1968 1969 1970
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) { }
1971
static inline void uclamp_fork(struct task_struct *p) { }
1972
static inline void uclamp_post_fork(struct task_struct *p) { }
1973 1974 1975
static inline void init_uclamp(void) { }
#endif /* CONFIG_UCLAMP_TASK */

1976 1977 1978 1979 1980
bool sched_task_on_rq(struct task_struct *p)
{
	return task_on_rq_queued(p);
}

1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999
unsigned long get_wchan(struct task_struct *p)
{
	unsigned long ip = 0;
	unsigned int state;

	if (!p || p == current)
		return 0;

	/* Only get wchan if task is blocked and we can keep it that way. */
	raw_spin_lock_irq(&p->pi_lock);
	state = READ_ONCE(p->__state);
	smp_rmb(); /* see try_to_wake_up() */
	if (state != TASK_RUNNING && state != TASK_WAKING && !p->on_rq)
		ip = __get_wchan(p);
	raw_spin_unlock_irq(&p->pi_lock);

	return ip;
}

2000
static inline void enqueue_task(struct rq *rq, struct task_struct *p, int flags)
2001
{
2002 2003 2004
	if (!(flags & ENQUEUE_NOCLOCK))
		update_rq_clock(rq);

2005
	if (!(flags & ENQUEUE_RESTORE)) {
2006
		sched_info_enqueue(rq, p);
2007 2008
		psi_enqueue(p, flags & ENQUEUE_WAKEUP);
	}
2009

2010
	uclamp_rq_inc(rq, p);
2011
	p->sched_class->enqueue_task(rq, p, flags);
2012 2013 2014

	if (sched_core_enabled(rq))
		sched_core_enqueue(rq, p);
2015 2016
}

2017
static inline void dequeue_task(struct rq *rq, struct task_struct *p, int flags)
2018
{
2019
	if (sched_core_enabled(rq))
2020
		sched_core_dequeue(rq, p, flags);
2021

2022 2023 2024
	if (!(flags & DEQUEUE_NOCLOCK))
		update_rq_clock(rq);

2025
	if (!(flags & DEQUEUE_SAVE)) {
2026
		sched_info_dequeue(rq, p);
2027 2028
		psi_dequeue(p, flags & DEQUEUE_SLEEP);
	}
2029

2030
	uclamp_rq_dec(rq, p);
2031
	p->sched_class->dequeue_task(rq, p, flags);
2032 2033
}

2034
void activate_task(struct rq *rq, struct task_struct *p, int flags)
2035
{
2036
	enqueue_task(rq, p, flags);
2037 2038

	p->on_rq = TASK_ON_RQ_QUEUED;
2039 2040
}

2041
void deactivate_task(struct rq *rq, struct task_struct *p, int flags)
2042
{
2043 2044
	p->on_rq = (flags & DEQUEUE_SLEEP) ? 0 : TASK_ON_RQ_MIGRATING;

2045
	dequeue_task(rq, p, flags);
2046 2047
}

2048
static inline int __normal_prio(int policy, int rt_prio, int nice)
2049
{
2050 2051 2052 2053 2054 2055 2056 2057 2058 2059
	int prio;

	if (dl_policy(policy))
		prio = MAX_DL_PRIO - 1;
	else if (rt_policy(policy))
		prio = MAX_RT_PRIO - 1 - rt_prio;
	else
		prio = NICE_TO_PRIO(nice);

	return prio;
2060 2061
}

2062 2063 2064 2065 2066 2067 2068
/*
 * 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.
 */
2069
static inline int normal_prio(struct task_struct *p)
2070
{
2071
	return __normal_prio(p->policy, p->rt_priority, PRIO_TO_NICE(p->static_prio));
2072 2073 2074 2075 2076 2077 2078 2079 2080
}

/*
 * 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.
 */
2081
static int effective_prio(struct task_struct *p)
2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093
{
	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
2094 2095 2096
/**
 * task_curr - is this task currently executing on a CPU?
 * @p: the task in question.
2097 2098
 *
 * Return: 1 if the task is currently executing. 0 otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
2099
 */
2100
inline int task_curr(const struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
2101 2102 2103 2104
{
	return cpu_curr(task_cpu(p)) == p;
}

2105
/*
2106 2107 2108 2109 2110
 * 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().
2111
 */
2112 2113
static inline void check_class_changed(struct rq *rq, struct task_struct *p,
				       const struct sched_class *prev_class,
2114
				       int oldprio)
2115 2116 2117
{
	if (prev_class != p->sched_class) {
		if (prev_class->switched_from)
2118
			prev_class->switched_from(rq, p);
2119

2120
		p->sched_class->switched_to(rq, p);
2121
	} else if (oldprio != p->prio || dl_task(p))
2122
		p->sched_class->prio_changed(rq, p, oldprio);
2123 2124
}

2125
void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags)
2126
{
2127
	if (p->sched_class == rq->curr->sched_class)
2128
		rq->curr->sched_class->check_preempt_curr(rq, p, flags);
2129 2130
	else if (p->sched_class > rq->curr->sched_class)
		resched_curr(rq);
2131 2132 2133 2134 2135

	/*
	 * A queue event has occurred, and we're going to schedule.  In
	 * this case, we can save a useless back to back clock update.
	 */
2136
	if (task_on_rq_queued(rq->curr) && test_tsk_need_resched(rq->curr))
2137
		rq_clock_skip_update(rq);
2138 2139
}

Linus Torvalds's avatar
Linus Torvalds committed
2140
#ifdef CONFIG_SMP
2141

2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164
static void
__do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask, u32 flags);

static int __set_cpus_allowed_ptr(struct task_struct *p,
				  const struct cpumask *new_mask,
				  u32 flags);

static void migrate_disable_switch(struct rq *rq, struct task_struct *p)
{
	if (likely(!p->migration_disabled))
		return;

	if (p->cpus_ptr != &p->cpus_mask)
		return;

	/*
	 * Violates locking rules! see comment in __do_set_cpus_allowed().
	 */
	__do_set_cpus_allowed(p, cpumask_of(rq->cpu), SCA_MIGRATE_DISABLE);
}

void migrate_disable(void)
{
2165 2166 2167 2168
	struct task_struct *p = current;

	if (p->migration_disabled) {
		p->migration_disabled++;
2169
		return;
2170
	}
2171

2172 2173 2174 2175
	preempt_disable();
	this_rq()->nr_pinned++;
	p->migration_disabled = 1;
	preempt_enable();
2176 2177 2178 2179 2180 2181 2182
}
EXPORT_SYMBOL_GPL(migrate_disable);

void migrate_enable(void)
{
	struct task_struct *p = current;

2183 2184
	if (p->migration_disabled > 1) {
		p->migration_disabled--;
2185
		return;
2186
	}
2187

2188 2189 2190
	if (WARN_ON_ONCE(!p->migration_disabled))
		return;

2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202
	/*
	 * Ensure stop_task runs either before or after this, and that
	 * __set_cpus_allowed_ptr(SCA_MIGRATE_ENABLE) doesn't schedule().
	 */
	preempt_disable();
	if (p->cpus_ptr != &p->cpus_mask)
		__set_cpus_allowed_ptr(p, &p->cpus_mask, SCA_MIGRATE_ENABLE);
	/*
	 * Mustn't clear migration_disabled() until cpus_ptr points back at the
	 * regular cpus_mask, otherwise things that race (eg.
	 * select_fallback_rq) get confused.
	 */
2203
	barrier();
2204
	p->migration_disabled = 0;
2205
	this_rq()->nr_pinned--;
2206
	preempt_enable();
2207 2208 2209
}
EXPORT_SYMBOL_GPL(migrate_enable);

2210 2211 2212 2213 2214
static inline bool rq_has_pinned_tasks(struct rq *rq)
{
	return rq->nr_pinned;
}

2215
/*
2216
 * Per-CPU kthreads are allowed to run on !active && online CPUs, see
2217 2218 2219 2220
 * __set_cpus_allowed_ptr() and select_fallback_rq().
 */
static inline bool is_cpu_allowed(struct task_struct *p, int cpu)
{
2221
	/* When not in the task's cpumask, no point in looking further. */
2222
	if (!cpumask_test_cpu(cpu, p->cpus_ptr))
2223 2224
		return false;

2225 2226
	/* migrate_disabled() must be allowed to finish. */
	if (is_migration_disabled(p))
2227 2228
		return cpu_online(cpu);

2229 2230
	/* Non kernel threads are not allowed during either online or offline. */
	if (!(p->flags & PF_KTHREAD))
2231
		return cpu_active(cpu) && task_cpu_possible(cpu, p);
2232 2233 2234 2235 2236 2237

	/* KTHREAD_IS_PER_CPU is always allowed. */
	if (kthread_is_per_cpu(p))
		return cpu_online(cpu);

	/* Regular kernel threads don't get to stay during offline. */
2238
	if (cpu_dying(cpu))
2239 2240 2241 2242
		return false;

	/* But are allowed during online. */
	return cpu_online(cpu);
2243 2244
}

Peter Zijlstra's avatar
Peter Zijlstra committed
2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263
/*
 * 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.
 */
2264 2265
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
2266
{
2267
	lockdep_assert_rq_held(rq);
Peter Zijlstra's avatar
Peter Zijlstra committed
2268

2269
	deactivate_task(rq, p, DEQUEUE_NOCLOCK);
Peter Zijlstra's avatar
Peter Zijlstra committed
2270
	set_task_cpu(p, new_cpu);
2271
	rq_unlock(rq, rf);
Peter Zijlstra's avatar
Peter Zijlstra committed
2272 2273 2274

	rq = cpu_rq(new_cpu);

2275
	rq_lock(rq, rf);
Peter Zijlstra's avatar
Peter Zijlstra committed
2276
	BUG_ON(task_cpu(p) != new_cpu);
2277
	activate_task(rq, p, 0);
Peter Zijlstra's avatar
Peter Zijlstra committed
2278 2279 2280 2281 2282 2283
	check_preempt_curr(rq, p, 0);

	return rq;
}

struct migration_arg {
2284 2285 2286 2287 2288
	struct task_struct		*task;
	int				dest_cpu;
	struct set_affinity_pending	*pending;
};

2289 2290 2291 2292
/*
 * @refs: number of wait_for_completion()
 * @stop_pending: is @stop_work in use
 */
2293 2294
struct set_affinity_pending {
	refcount_t		refs;
2295
	unsigned int		stop_pending;
2296 2297 2298
	struct completion	done;
	struct cpu_stop_work	stop_work;
	struct migration_arg	arg;
Peter Zijlstra's avatar
Peter Zijlstra committed
2299 2300 2301
};

/*
2302
 * Move (not current) task off this CPU, onto the destination CPU. We're doing
Peter Zijlstra's avatar
Peter Zijlstra committed
2303 2304 2305 2306 2307 2308 2309
 * 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.
 */
2310 2311
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
2312 2313
{
	/* Affinity changed (again). */
2314
	if (!is_cpu_allowed(p, dest_cpu))
2315
		return rq;
Peter Zijlstra's avatar
Peter Zijlstra committed
2316

2317
	update_rq_clock(rq);
2318
	rq = move_queued_task(rq, rf, p, dest_cpu);
2319 2320

	return rq;
Peter Zijlstra's avatar
Peter Zijlstra committed
2321 2322 2323 2324 2325 2326 2327 2328 2329 2330
}

/*
 * 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;
2331
	struct set_affinity_pending *pending = arg->pending;
2332 2333
	struct task_struct *p = arg->task;
	struct rq *rq = this_rq();
2334
	bool complete = false;
2335
	struct rq_flags rf;
Peter Zijlstra's avatar
Peter Zijlstra committed
2336 2337

	/*
2338 2339
	 * 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
2340
	 */
2341
	local_irq_save(rf.flags);
Peter Zijlstra's avatar
Peter Zijlstra committed
2342 2343
	/*
	 * We need to explicitly wake pending tasks before running
2344
	 * __migrate_task() such that we will not miss enforcing cpus_ptr
Peter Zijlstra's avatar
Peter Zijlstra committed
2345 2346
	 * during wakeups, see set_cpus_allowed_ptr()'s TASK_WAKING test.
	 */
2347
	flush_smp_call_function_from_idle();
2348 2349

	raw_spin_lock(&p->pi_lock);
2350
	rq_lock(rq, &rf);
2351

2352 2353 2354 2355 2356 2357
	/*
	 * If we were passed a pending, then ->stop_pending was set, thus
	 * p->migration_pending must have remained stable.
	 */
	WARN_ON_ONCE(pending && pending != p->migration_pending);

2358 2359 2360 2361 2362
	/*
	 * 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.
	 */
2363
	if (task_rq(p) == rq) {
2364 2365 2366 2367
		if (is_migration_disabled(p))
			goto out;

		if (pending) {
2368
			p->migration_pending = NULL;
2369 2370
			complete = true;

2371 2372 2373
			if (cpumask_test_cpu(task_cpu(p), &p->cpus_mask))
				goto out;
		}
2374

2375
		if (task_on_rq_queued(p))
2376
			rq = __migrate_task(rq, &rf, p, arg->dest_cpu);
2377
		else
2378
			p->wake_cpu = arg->dest_cpu;
2379

2380 2381 2382 2383 2384 2385 2386
		/*
		 * XXX __migrate_task() can fail, at which point we might end
		 * up running on a dodgy CPU, AFAICT this can only happen
		 * during CPU hotplug, at which point we'll get pushed out
		 * anyway, so it's probably not a big deal.
		 */

2387
	} else if (pending) {
2388 2389 2390 2391 2392 2393 2394 2395 2396
		/*
		 * This happens when we get migrated between migrate_enable()'s
		 * preempt_enable() and scheduling the stopper task. At that
		 * point we're a regular task again and not current anymore.
		 *
		 * A !PREEMPT kernel has a giant hole here, which makes it far
		 * more likely.
		 */

2397 2398 2399 2400 2401
		/*
		 * The task moved before the stopper got to run. We're holding
		 * ->pi_lock, so the allowed mask is stable - if it got
		 * somewhere allowed, we're done.
		 */
2402
		if (cpumask_test_cpu(task_cpu(p), p->cpus_ptr)) {
2403
			p->migration_pending = NULL;
2404 2405 2406 2407
			complete = true;
			goto out;
		}

2408 2409 2410 2411 2412
		/*
		 * When migrate_enable() hits a rq mis-match we can't reliably
		 * determine is_migration_disabled() and so have to chase after
		 * it.
		 */
2413
		WARN_ON_ONCE(!pending->stop_pending);
2414 2415 2416 2417
		task_rq_unlock(rq, p, &rf);
		stop_one_cpu_nowait(task_cpu(p), migration_cpu_stop,
				    &pending->arg, &pending->stop_work);
		return 0;
2418
	}
2419
out:
2420 2421
	if (pending)
		pending->stop_pending = false;
2422 2423 2424 2425 2426
	task_rq_unlock(rq, p, &rf);

	if (complete)
		complete_all(&pending->done);

Peter Zijlstra's avatar
Peter Zijlstra committed
2427 2428 2429
	return 0;
}

2430 2431 2432 2433 2434 2435
int push_cpu_stop(void *arg)
{
	struct rq *lowest_rq = NULL, *rq = this_rq();
	struct task_struct *p = arg;

	raw_spin_lock_irq(&p->pi_lock);
2436
	raw_spin_rq_lock(rq);
2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449

	if (task_rq(p) != rq)
		goto out_unlock;

	if (is_migration_disabled(p)) {
		p->migration_flags |= MDF_PUSH;
		goto out_unlock;
	}

	p->migration_flags &= ~MDF_PUSH;

	if (p->sched_class->find_lock_rq)
		lowest_rq = p->sched_class->find_lock_rq(p, rq);
2450

2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465
	if (!lowest_rq)
		goto out_unlock;

	// XXX validate p is still the highest prio task
	if (task_rq(p) == rq) {
		deactivate_task(rq, p, 0);
		set_task_cpu(p, lowest_rq->cpu);
		activate_task(lowest_rq, p, 0);
		resched_curr(lowest_rq);
	}

	double_unlock_balance(rq, lowest_rq);

out_unlock:
	rq->push_busy = false;
2466
	raw_spin_rq_unlock(rq);
2467 2468 2469
	raw_spin_unlock_irq(&p->pi_lock);

	put_task_struct(p);
Peter Zijlstra's avatar
Peter Zijlstra committed
2470 2471 2472
	return 0;
}

2473 2474 2475 2476
/*
 * sched_class::set_cpus_allowed must do the below, but is not required to
 * actually call this function.
 */
2477
void set_cpus_allowed_common(struct task_struct *p, const struct cpumask *new_mask, u32 flags)
Peter Zijlstra's avatar
Peter Zijlstra committed
2478
{
2479 2480 2481 2482 2483
	if (flags & (SCA_MIGRATE_ENABLE | SCA_MIGRATE_DISABLE)) {
		p->cpus_ptr = new_mask;
		return;
	}

2484
	cpumask_copy(&p->cpus_mask, new_mask);
Peter Zijlstra's avatar
Peter Zijlstra committed
2485 2486 2487
	p->nr_cpus_allowed = cpumask_weight(new_mask);
}

2488 2489
static void
__do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask, u32 flags)
2490
{
2491 2492 2493
	struct rq *rq = task_rq(p);
	bool queued, running;

2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509
	/*
	 * This here violates the locking rules for affinity, since we're only
	 * supposed to change these variables while holding both rq->lock and
	 * p->pi_lock.
	 *
	 * HOWEVER, it magically works, because ttwu() is the only code that
	 * accesses these variables under p->pi_lock and only does so after
	 * smp_cond_load_acquire(&p->on_cpu, !VAL), and we're in __schedule()
	 * before finish_task().
	 *
	 * XXX do further audits, this smells like something putrid.
	 */
	if (flags & SCA_MIGRATE_DISABLE)
		SCHED_WARN_ON(!p->on_cpu);
	else
		lockdep_assert_held(&p->pi_lock);
2510 2511 2512 2513 2514 2515 2516 2517 2518

	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.
		 */
2519
		lockdep_assert_rq_held(rq);
2520
		dequeue_task(rq, p, DEQUEUE_SAVE | DEQUEUE_NOCLOCK);
2521 2522 2523 2524
	}
	if (running)
		put_prev_task(rq, p);

2525
	p->sched_class->set_cpus_allowed(p, new_mask, flags);
2526 2527

	if (queued)
2528
		enqueue_task(rq, p, ENQUEUE_RESTORE | ENQUEUE_NOCLOCK);
2529
	if (running)
2530
		set_next_task(rq, p);
2531 2532
}

2533 2534 2535 2536 2537
void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
{
	__do_set_cpus_allowed(p, new_mask, 0);
}

2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551
int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src,
		      int node)
{
	if (!src->user_cpus_ptr)
		return 0;

	dst->user_cpus_ptr = kmalloc_node(cpumask_size(), GFP_KERNEL, node);
	if (!dst->user_cpus_ptr)
		return -ENOMEM;

	cpumask_copy(dst->user_cpus_ptr, src->user_cpus_ptr);
	return 0;
}

2552 2553 2554 2555 2556 2557 2558 2559 2560
static inline struct cpumask *clear_user_cpus_ptr(struct task_struct *p)
{
	struct cpumask *user_mask = NULL;

	swap(p->user_cpus_ptr, user_mask);

	return user_mask;
}

2561 2562
void release_user_cpus_ptr(struct task_struct *p)
{
2563
	kfree(clear_user_cpus_ptr(p));
2564 2565
}

2566
/*
2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618
 * This function is wildly self concurrent; here be dragons.
 *
 *
 * When given a valid mask, __set_cpus_allowed_ptr() must block until the
 * designated task is enqueued on an allowed CPU. If that task is currently
 * running, we have to kick it out using the CPU stopper.
 *
 * Migrate-Disable comes along and tramples all over our nice sandcastle.
 * Consider:
 *
 *     Initial conditions: P0->cpus_mask = [0, 1]
 *
 *     P0@CPU0                  P1
 *
 *     migrate_disable();
 *     <preempted>
 *                              set_cpus_allowed_ptr(P0, [1]);
 *
 * P1 *cannot* return from this set_cpus_allowed_ptr() call until P0 executes
 * its outermost migrate_enable() (i.e. it exits its Migrate-Disable region).
 * This means we need the following scheme:
 *
 *     P0@CPU0                  P1
 *
 *     migrate_disable();
 *     <preempted>
 *                              set_cpus_allowed_ptr(P0, [1]);
 *                                <blocks>
 *     <resumes>
 *     migrate_enable();
 *       __set_cpus_allowed_ptr();
 *       <wakes local stopper>
 *                         `--> <woken on migration completion>
 *
 * Now the fun stuff: there may be several P1-like tasks, i.e. multiple
 * concurrent set_cpus_allowed_ptr(P0, [*]) calls. CPU affinity changes of any
 * task p are serialized by p->pi_lock, which we can leverage: the one that
 * should come into effect at the end of the Migrate-Disable region is the last
 * one. This means we only need to track a single cpumask (i.e. p->cpus_mask),
 * but we still need to properly signal those waiting tasks at the appropriate
 * moment.
 *
 * This is implemented using struct set_affinity_pending. The first
 * __set_cpus_allowed_ptr() caller within a given Migrate-Disable region will
 * setup an instance of that struct and install it on the targeted task_struct.
 * Any and all further callers will reuse that instance. Those then wait for
 * a completion signaled at the tail of the CPU stopper callback (1), triggered
 * on the end of the Migrate-Disable region (i.e. outermost migrate_enable()).
 *
 *
 * (1) In the cases covered above. There is one more where the completion is
 * signaled within affine_move_task() itself: when a subsequent affinity request
2619 2620
 * occurs after the stopper bailed out due to the targeted task still being
 * Migrate-Disable. Consider:
2621 2622 2623
 *
 *     Initial conditions: P0->cpus_mask = [0, 1]
 *
2624 2625 2626 2627
 *     CPU0		  P1				P2
 *     <P0>
 *       migrate_disable();
 *       <preempted>
2628 2629
 *                        set_cpus_allowed_ptr(P0, [1]);
 *                          <blocks>
2630 2631 2632 2633
 *     <migration/0>
 *       migration_cpu_stop()
 *         is_migration_disabled()
 *           <bails>
2634 2635 2636 2637 2638 2639 2640
 *                                                       set_cpus_allowed_ptr(P0, [0, 1]);
 *                                                         <signal completion>
 *                          <awakes>
 *
 * Note that the above is safe vs a concurrent migrate_enable(), as any
 * pending affinity completion is preceded by an uninstallation of
 * p->migration_pending done with p->pi_lock held.
2641 2642 2643 2644 2645
 */
static int affine_move_task(struct rq *rq, struct task_struct *p, struct rq_flags *rf,
			    int dest_cpu, unsigned int flags)
{
	struct set_affinity_pending my_pending = { }, *pending = NULL;
2646
	bool stop_pending, complete = false;
2647 2648 2649

	/* Can the task run on the task's current CPU? If so, we're done */
	if (cpumask_test_cpu(task_cpu(p), &p->cpus_mask)) {
2650 2651 2652 2653 2654 2655 2656 2657
		struct task_struct *push_task = NULL;

		if ((flags & SCA_MIGRATE_ENABLE) &&
		    (p->migration_flags & MDF_PUSH) && !rq->push_busy) {
			rq->push_busy = true;
			push_task = get_task_struct(p);
		}

2658 2659 2660 2661
		/*
		 * If there are pending waiters, but no pending stop_work,
		 * then complete now.
		 */
2662
		pending = p->migration_pending;
2663
		if (pending && !pending->stop_pending) {
2664 2665 2666
			p->migration_pending = NULL;
			complete = true;
		}
2667

2668 2669
		task_rq_unlock(rq, p, rf);

2670 2671 2672 2673 2674
		if (push_task) {
			stop_one_cpu_nowait(rq->cpu, push_cpu_stop,
					    p, &rq->push_work);
		}

2675
		if (complete)
2676
			complete_all(&pending->done);
2677 2678 2679 2680 2681 2682 2683

		return 0;
	}

	if (!(flags & SCA_MIGRATE_ENABLE)) {
		/* serialized by p->pi_lock */
		if (!p->migration_pending) {
2684
			/* Install the request */
2685 2686
			refcount_set(&my_pending.refs, 1);
			init_completion(&my_pending.done);
2687 2688
			my_pending.arg = (struct migration_arg) {
				.task = p,
2689
				.dest_cpu = dest_cpu,
2690 2691 2692
				.pending = &my_pending,
			};

2693 2694 2695 2696
			p->migration_pending = &my_pending;
		} else {
			pending = p->migration_pending;
			refcount_inc(&pending->refs);
2697 2698 2699 2700 2701 2702 2703 2704 2705
			/*
			 * Affinity has changed, but we've already installed a
			 * pending. migration_cpu_stop() *must* see this, else
			 * we risk a completion of the pending despite having a
			 * task on a disallowed CPU.
			 *
			 * Serialized by p->pi_lock, so this is safe.
			 */
			pending->arg.dest_cpu = dest_cpu;
2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725
		}
	}
	pending = p->migration_pending;
	/*
	 * - !MIGRATE_ENABLE:
	 *   we'll have installed a pending if there wasn't one already.
	 *
	 * - MIGRATE_ENABLE:
	 *   we're here because the current CPU isn't matching anymore,
	 *   the only way that can happen is because of a concurrent
	 *   set_cpus_allowed_ptr() call, which should then still be
	 *   pending completion.
	 *
	 * Either way, we really should have a @pending here.
	 */
	if (WARN_ON_ONCE(!pending)) {
		task_rq_unlock(rq, p, rf);
		return -EINVAL;
	}

2726
	if (task_running(rq, p) || READ_ONCE(p->__state) == TASK_WAKING) {
2727
		/*
2728 2729 2730
		 * MIGRATE_ENABLE gets here because 'p == current', but for
		 * anything else we cannot do is_migration_disabled(), punt
		 * and have the stopper function handle it all race-free.
2731
		 */
2732 2733 2734
		stop_pending = pending->stop_pending;
		if (!stop_pending)
			pending->stop_pending = true;
2735 2736 2737

		if (flags & SCA_MIGRATE_ENABLE)
			p->migration_flags &= ~MDF_PUSH;
2738

2739
		task_rq_unlock(rq, p, rf);
2740

2741 2742 2743 2744
		if (!stop_pending) {
			stop_one_cpu_nowait(cpu_of(rq), migration_cpu_stop,
					    &pending->arg, &pending->stop_work);
		}
2745

2746 2747
		if (flags & SCA_MIGRATE_ENABLE)
			return 0;
2748 2749 2750 2751 2752 2753
	} else {

		if (!is_migration_disabled(p)) {
			if (task_on_rq_queued(p))
				rq = move_queued_task(rq, rf, p, dest_cpu);

2754 2755 2756 2757
			if (!pending->stop_pending) {
				p->migration_pending = NULL;
				complete = true;
			}
2758 2759 2760 2761 2762 2763 2764 2765 2766 2767
		}
		task_rq_unlock(rq, p, rf);

		if (complete)
			complete_all(&pending->done);
	}

	wait_for_completion(&pending->done);

	if (refcount_dec_and_test(&pending->refs))
2768
		wake_up_var(&pending->refs); /* No UaF, just an address */
2769

2770 2771 2772 2773
	/*
	 * Block the original owner of &pending until all subsequent callers
	 * have seen the completion and decremented the refcount
	 */
2774 2775
	wait_var_event(&my_pending.refs, !refcount_read(&my_pending.refs));

2776 2777 2778
	/* ARGH */
	WARN_ON_ONCE(my_pending.stop_pending);

2779 2780 2781
	return 0;
}

Peter Zijlstra's avatar
Peter Zijlstra committed
2782
/*
2783
 * Called with both p->pi_lock and rq->lock held; drops both before returning.
Peter Zijlstra's avatar
Peter Zijlstra committed
2784
 */
2785 2786 2787 2788 2789 2790 2791
static int __set_cpus_allowed_ptr_locked(struct task_struct *p,
					 const struct cpumask *new_mask,
					 u32 flags,
					 struct rq *rq,
					 struct rq_flags *rf)
	__releases(rq->lock)
	__releases(p->pi_lock)
Peter Zijlstra's avatar
Peter Zijlstra committed
2792
{
2793
	const struct cpumask *cpu_allowed_mask = task_cpu_possible_mask(p);
2794
	const struct cpumask *cpu_valid_mask = cpu_active_mask;
2795
	bool kthread = p->flags & PF_KTHREAD;
2796
	struct cpumask *user_mask = NULL;
Peter Zijlstra's avatar
Peter Zijlstra committed
2797 2798 2799
	unsigned int dest_cpu;
	int ret = 0;

2800
	update_rq_clock(rq);
Peter Zijlstra's avatar
Peter Zijlstra committed
2801

2802
	if (kthread || is_migration_disabled(p)) {
2803
		/*
2804 2805 2806
		 * Kernel threads are allowed on online && !active CPUs,
		 * however, during cpu-hot-unplug, even these might get pushed
		 * away if not KTHREAD_IS_PER_CPU.
2807 2808 2809 2810 2811
		 *
		 * Specifically, migration_disabled() tasks must not fail the
		 * cpumask_any_and_distribute() pick below, esp. so on
		 * SCA_MIGRATE_ENABLE, otherwise we'll not call
		 * set_cpus_allowed_common() and actually reset p->cpus_ptr.
2812 2813 2814 2815
		 */
		cpu_valid_mask = cpu_online_mask;
	}

2816 2817 2818 2819 2820
	if (!kthread && !cpumask_subset(new_mask, cpu_allowed_mask)) {
		ret = -EINVAL;
		goto out;
	}

2821 2822 2823 2824
	/*
	 * Must re-check here, to close a race against __kthread_bind(),
	 * sched_setaffinity() is not guaranteed to observe the flag.
	 */
2825
	if ((flags & SCA_CHECK) && (p->flags & PF_NO_SETAFFINITY)) {
2826 2827 2828 2829
		ret = -EINVAL;
		goto out;
	}

2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840
	if (!(flags & SCA_MIGRATE_ENABLE)) {
		if (cpumask_equal(&p->cpus_mask, new_mask))
			goto out;

		if (WARN_ON_ONCE(p == current &&
				 is_migration_disabled(p) &&
				 !cpumask_test_cpu(task_cpu(p), new_mask))) {
			ret = -EBUSY;
			goto out;
		}
	}
Peter Zijlstra's avatar
Peter Zijlstra committed
2841

2842 2843 2844 2845 2846 2847
	/*
	 * 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);
2848
	if (dest_cpu >= nr_cpu_ids) {
Peter Zijlstra's avatar
Peter Zijlstra committed
2849 2850 2851 2852
		ret = -EINVAL;
		goto out;
	}

2853
	__do_set_cpus_allowed(p, new_mask, flags);
Peter Zijlstra's avatar
Peter Zijlstra committed
2854

2855 2856 2857 2858 2859 2860 2861 2862
	if (flags & SCA_USER)
		user_mask = clear_user_cpus_ptr(p);

	ret = affine_move_task(rq, p, rf, dest_cpu, flags);

	kfree(user_mask);

	return ret;
Peter Zijlstra's avatar
Peter Zijlstra committed
2863 2864

out:
2865
	task_rq_unlock(rq, p, rf);
Peter Zijlstra's avatar
Peter Zijlstra committed
2866 2867 2868

	return ret;
}
2869

2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888
/*
 * 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.
 */
static int __set_cpus_allowed_ptr(struct task_struct *p,
				  const struct cpumask *new_mask, u32 flags)
{
	struct rq_flags rf;
	struct rq *rq;

	rq = task_rq_lock(p, &rf);
	return __set_cpus_allowed_ptr_locked(p, new_mask, flags, rq, &rf);
}

2889 2890
int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
{
2891
	return __set_cpus_allowed_ptr(p, new_mask, 0);
2892
}
Peter Zijlstra's avatar
Peter Zijlstra committed
2893 2894
EXPORT_SYMBOL_GPL(set_cpus_allowed_ptr);

2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026
/*
 * Change a given task's CPU affinity to the intersection of its current
 * affinity mask and @subset_mask, writing the resulting mask to @new_mask
 * and pointing @p->user_cpus_ptr to a copy of the old mask.
 * If the resulting mask is empty, leave the affinity unchanged and return
 * -EINVAL.
 */
static int restrict_cpus_allowed_ptr(struct task_struct *p,
				     struct cpumask *new_mask,
				     const struct cpumask *subset_mask)
{
	struct cpumask *user_mask = NULL;
	struct rq_flags rf;
	struct rq *rq;
	int err;

	if (!p->user_cpus_ptr) {
		user_mask = kmalloc(cpumask_size(), GFP_KERNEL);
		if (!user_mask)
			return -ENOMEM;
	}

	rq = task_rq_lock(p, &rf);

	/*
	 * Forcefully restricting the affinity of a deadline task is
	 * likely to cause problems, so fail and noisily override the
	 * mask entirely.
	 */
	if (task_has_dl_policy(p) && dl_bandwidth_enabled()) {
		err = -EPERM;
		goto err_unlock;
	}

	if (!cpumask_and(new_mask, &p->cpus_mask, subset_mask)) {
		err = -EINVAL;
		goto err_unlock;
	}

	/*
	 * We're about to butcher the task affinity, so keep track of what
	 * the user asked for in case we're able to restore it later on.
	 */
	if (user_mask) {
		cpumask_copy(user_mask, p->cpus_ptr);
		p->user_cpus_ptr = user_mask;
	}

	return __set_cpus_allowed_ptr_locked(p, new_mask, 0, rq, &rf);

err_unlock:
	task_rq_unlock(rq, p, &rf);
	kfree(user_mask);
	return err;
}

/*
 * Restrict the CPU affinity of task @p so that it is a subset of
 * task_cpu_possible_mask() and point @p->user_cpu_ptr to a copy of the
 * old affinity mask. If the resulting mask is empty, we warn and walk
 * up the cpuset hierarchy until we find a suitable mask.
 */
void force_compatible_cpus_allowed_ptr(struct task_struct *p)
{
	cpumask_var_t new_mask;
	const struct cpumask *override_mask = task_cpu_possible_mask(p);

	alloc_cpumask_var(&new_mask, GFP_KERNEL);

	/*
	 * __migrate_task() can fail silently in the face of concurrent
	 * offlining of the chosen destination CPU, so take the hotplug
	 * lock to ensure that the migration succeeds.
	 */
	cpus_read_lock();
	if (!cpumask_available(new_mask))
		goto out_set_mask;

	if (!restrict_cpus_allowed_ptr(p, new_mask, override_mask))
		goto out_free_mask;

	/*
	 * We failed to find a valid subset of the affinity mask for the
	 * task, so override it based on its cpuset hierarchy.
	 */
	cpuset_cpus_allowed(p, new_mask);
	override_mask = new_mask;

out_set_mask:
	if (printk_ratelimit()) {
		printk_deferred("Overriding affinity for process %d (%s) to CPUs %*pbl\n",
				task_pid_nr(p), p->comm,
				cpumask_pr_args(override_mask));
	}

	WARN_ON(set_cpus_allowed_ptr(p, override_mask));
out_free_mask:
	cpus_read_unlock();
	free_cpumask_var(new_mask);
}

static int
__sched_setaffinity(struct task_struct *p, const struct cpumask *mask);

/*
 * Restore the affinity of a task @p which was previously restricted by a
 * call to force_compatible_cpus_allowed_ptr(). This will clear (and free)
 * @p->user_cpus_ptr.
 *
 * It is the caller's responsibility to serialise this with any calls to
 * force_compatible_cpus_allowed_ptr(@p).
 */
void relax_compatible_cpus_allowed_ptr(struct task_struct *p)
{
	struct cpumask *user_mask = p->user_cpus_ptr;
	unsigned long flags;

	/*
	 * Try to restore the old affinity mask. If this fails, then
	 * we free the mask explicitly to avoid it being inherited across
	 * a subsequent fork().
	 */
	if (!user_mask || !__sched_setaffinity(p, user_mask))
		return;

	raw_spin_lock_irqsave(&p->pi_lock, flags);
	user_mask = clear_user_cpus_ptr(p);
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);

	kfree(user_mask);
}

Ingo Molnar's avatar
Ingo Molnar committed
3027
void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
Ingo Molnar's avatar
Ingo Molnar committed
3028
{
3029
#ifdef CONFIG_SCHED_DEBUG
3030 3031
	unsigned int state = READ_ONCE(p->__state);

3032 3033 3034 3035
	/*
	 * We should never call set_task_cpu() on a blocked task,
	 * ttwu() will sort out the placement.
	 */
3036
	WARN_ON_ONCE(state != TASK_RUNNING && state != TASK_WAKING && !p->on_rq);
3037

3038 3039 3040 3041 3042
	/*
	 * 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.
	 */
3043
	WARN_ON_ONCE(state == TASK_RUNNING &&
3044 3045 3046
		     p->sched_class == &fair_sched_class &&
		     (p->on_rq && !task_on_rq_migrating(p)));

3047
#ifdef CONFIG_LOCKDEP
3048 3049 3050 3051 3052
	/*
	 * 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,
3053
	 * see task_group().
3054 3055 3056 3057
	 *
	 * Furthermore, all task_rq users should acquire both locks, see
	 * task_rq_lock().
	 */
3058
	WARN_ON_ONCE(debug_locks && !(lockdep_is_held(&p->pi_lock) ||
3059
				      lockdep_is_held(__rq_lockp(task_rq(p)))));
3060
#endif
3061 3062 3063 3064
	/*
	 * Clearly, migrating tasks to offline CPUs is a fairly daft thing.
	 */
	WARN_ON_ONCE(!cpu_online(new_cpu));
3065 3066

	WARN_ON_ONCE(is_migration_disabled(p));
3067 3068
#endif

3069
	trace_sched_migrate_task(p, new_cpu);
3070

3071
	if (task_cpu(p) != new_cpu) {
3072
		if (p->sched_class->migrate_task_rq)
3073
			p->sched_class->migrate_task_rq(p, new_cpu);
3074
		p->se.nr_migrations++;
3075
		rseq_migrate(p);
3076
		perf_event_task_migrate(p);
3077
	}
Ingo Molnar's avatar
Ingo Molnar committed
3078 3079

	__set_task_cpu(p, new_cpu);
Ingo Molnar's avatar
Ingo Molnar committed
3080 3081
}

3082
#ifdef CONFIG_NUMA_BALANCING
3083 3084
static void __migrate_swap_task(struct task_struct *p, int cpu)
{
3085
	if (task_on_rq_queued(p)) {
3086
		struct rq *src_rq, *dst_rq;
3087
		struct rq_flags srf, drf;
3088 3089 3090 3091

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

3092 3093 3094
		rq_pin_lock(src_rq, &srf);
		rq_pin_lock(dst_rq, &drf);

3095 3096 3097 3098
		deactivate_task(src_rq, p, 0);
		set_task_cpu(p, cpu);
		activate_task(dst_rq, p, 0);
		check_preempt_curr(dst_rq, p, 0);
3099 3100 3101 3102

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

3103 3104 3105 3106
	} 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
3107
		 * previous CPU our target instead of where it really is.
3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123
		 */
		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;

3124 3125 3126
	if (!cpu_active(arg->src_cpu) || !cpu_active(arg->dst_cpu))
		return -EAGAIN;

3127 3128 3129
	src_rq = cpu_rq(arg->src_cpu);
	dst_rq = cpu_rq(arg->dst_cpu);

3130 3131
	double_raw_lock(&arg->src_task->pi_lock,
			&arg->dst_task->pi_lock);
3132
	double_rq_lock(src_rq, dst_rq);
3133

3134 3135 3136 3137 3138 3139
	if (task_cpu(arg->dst_task) != arg->dst_cpu)
		goto unlock;

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

3140
	if (!cpumask_test_cpu(arg->dst_cpu, arg->src_task->cpus_ptr))
3141 3142
		goto unlock;

3143
	if (!cpumask_test_cpu(arg->src_cpu, arg->dst_task->cpus_ptr))
3144 3145 3146 3147 3148 3149 3150 3151 3152
		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);
3153 3154
	raw_spin_unlock(&arg->dst_task->pi_lock);
	raw_spin_unlock(&arg->src_task->pi_lock);
3155 3156 3157 3158 3159 3160 3161

	return ret;
}

/*
 * Cross migrate two tasks
 */
3162 3163
int migrate_swap(struct task_struct *cur, struct task_struct *p,
		int target_cpu, int curr_cpu)
3164 3165 3166 3167 3168 3169
{
	struct migration_swap_arg arg;
	int ret = -EINVAL;

	arg = (struct migration_swap_arg){
		.src_task = cur,
3170
		.src_cpu = curr_cpu,
3171
		.dst_task = p,
3172
		.dst_cpu = target_cpu,
3173 3174 3175 3176 3177
	};

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

3178 3179 3180 3181
	/*
	 * 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.
	 */
3182 3183 3184
	if (!cpu_active(arg.src_cpu) || !cpu_active(arg.dst_cpu))
		goto out;

3185
	if (!cpumask_test_cpu(arg.dst_cpu, arg.src_task->cpus_ptr))
3186 3187
		goto out;

3188
	if (!cpumask_test_cpu(arg.src_cpu, arg.dst_task->cpus_ptr))
3189 3190
		goto out;

3191
	trace_sched_swap_numa(cur, arg.src_cpu, p, arg.dst_cpu);
3192 3193 3194 3195 3196
	ret = stop_two_cpus(arg.dst_cpu, arg.src_cpu, migrate_swap_stop, &arg);

out:
	return ret;
}
3197
#endif /* CONFIG_NUMA_BALANCING */
3198

Linus Torvalds's avatar
Linus Torvalds committed
3199 3200 3201
/*
 * wait_task_inactive - wait for a thread to unschedule.
 *
3202 3203 3204 3205 3206 3207 3208
 * 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
3209 3210 3211 3212 3213 3214
 * 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.
 */
3215
unsigned long wait_task_inactive(struct task_struct *p, unsigned int match_state)
Linus Torvalds's avatar
Linus Torvalds committed
3216
{
3217
	int running, queued;
3218
	struct rq_flags rf;
3219
	unsigned long ncsw;
3220
	struct rq *rq;
Linus Torvalds's avatar
Linus Torvalds committed
3221

3222 3223 3224 3225 3226 3227 3228 3229
	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);
3230

3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241
		/*
		 * 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!
		 */
3242
		while (task_running(rq, p)) {
3243
			if (match_state && unlikely(READ_ONCE(p->__state) != match_state))
3244
				return 0;
3245
			cpu_relax();
3246
		}
3247

3248 3249 3250 3251 3252
		/*
		 * 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.
		 */
3253
		rq = task_rq_lock(p, &rf);
3254
		trace_sched_wait_task(p);
3255
		running = task_running(rq, p);
3256
		queued = task_on_rq_queued(p);
3257
		ncsw = 0;
3258
		if (!match_state || READ_ONCE(p->__state) == match_state)
3259
			ncsw = p->nvcsw | LONG_MIN; /* sets MSB */
3260
		task_rq_unlock(rq, p, &rf);
3261

3262 3263 3264 3265 3266 3267
		/*
		 * If it changed from the expected state, bail out now.
		 */
		if (unlikely(!ncsw))
			break;

3268 3269 3270 3271 3272 3273 3274 3275 3276 3277
		/*
		 * 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;
		}
3278

3279 3280 3281 3282 3283
		/*
		 * It's not enough that it's not actively running,
		 * it must be off the runqueue _entirely_, and not
		 * preempted!
		 *
3284
		 * So if it was still runnable (but just not actively
3285 3286 3287
		 * running right now), it's preempted, and we should
		 * yield - it could be a while.
		 */
3288
		if (unlikely(queued)) {
3289
			ktime_t to = NSEC_PER_SEC / HZ;
3290 3291

			set_current_state(TASK_UNINTERRUPTIBLE);
3292
			schedule_hrtimeout(&to, HRTIMER_MODE_REL_HARD);
3293 3294
			continue;
		}
3295

3296 3297 3298 3299 3300 3301 3302
		/*
		 * 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;
	}
3303 3304

	return ncsw;
Linus Torvalds's avatar
Linus Torvalds committed
3305 3306 3307 3308 3309 3310 3311 3312 3313
}

/***
 * 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
3314
 * NOTE: this function doesn't have to take the runqueue lock,
Linus Torvalds's avatar
Linus Torvalds committed
3315 3316 3317 3318 3319
 * 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.
 */
3320
void kick_process(struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
3321 3322 3323 3324 3325 3326 3327 3328 3329
{
	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
3330
EXPORT_SYMBOL_GPL(kick_process);
Linus Torvalds's avatar
Linus Torvalds committed
3331

3332
/*
3333
 * ->cpus_ptr is protected by both rq->lock and p->pi_lock
3334 3335 3336 3337 3338
 *
 * A few notes on cpu_active vs cpu_online:
 *
 *  - cpu_active must be a subset of cpu_online
 *
3339
 *  - on CPU-up we allow per-CPU kthreads on the online && !active CPU,
3340
 *    see __set_cpus_allowed_ptr(). At this point the newly online
3341
 *    CPU isn't yet part of the sched domains, and balancing will not
3342 3343
 *    see it.
 *
3344
 *  - on CPU-down we clear cpu_active() to mask the sched domains and
3345
 *    avoid the load balancer to place new tasks on the to be removed
3346
 *    CPU. Existing tasks will remain running there and will be taken
3347 3348 3349 3350 3351 3352
 *    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.
3353
 */
3354 3355
static int select_fallback_rq(int cpu, struct task_struct *p)
{
3356 3357
	int nid = cpu_to_node(cpu);
	const struct cpumask *nodemask = NULL;
3358 3359
	enum { cpuset, possible, fail } state = cpuset;
	int dest_cpu;
3360

3361
	/*
3362 3363 3364
	 * 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.
3365 3366 3367 3368 3369 3370
	 */
	if (nid != -1) {
		nodemask = cpumask_of_node(nid);

		/* Look for allowed, online CPU in same node. */
		for_each_cpu(dest_cpu, nodemask) {
3371
			if (is_cpu_allowed(p, dest_cpu))
3372 3373
				return dest_cpu;
		}
3374
	}
3375

3376 3377
	for (;;) {
		/* Any allowed, online CPU? */
3378
		for_each_cpu(dest_cpu, p->cpus_ptr) {
3379
			if (!is_cpu_allowed(p, dest_cpu))
3380
				continue;
3381

3382 3383
			goto out;
		}
3384

3385
		/* No more Mr. Nice Guy. */
3386 3387
		switch (state) {
		case cpuset:
3388
			if (cpuset_cpus_allowed_fallback(p)) {
3389 3390 3391
				state = possible;
				break;
			}
3392
			fallthrough;
3393
		case possible:
3394 3395 3396 3397 3398 3399
			/*
			 * XXX When called from select_task_rq() we only
			 * hold p->pi_lock and again violate locking order.
			 *
			 * More yuck to audit.
			 */
3400
			do_set_cpus_allowed(p, task_cpu_possible_mask(p));
3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416
			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()) {
3417
			printk_deferred("process %d (%s) no longer affine to cpu%d\n",
3418 3419
					task_pid_nr(p), p->comm, cpu);
		}
3420 3421 3422 3423 3424
	}

	return dest_cpu;
}

3425
/*
3426
 * The caller (fork, wakeup) owns p->pi_lock, ->cpus_ptr is stable.
3427
 */
3428
static inline
3429
int select_task_rq(struct task_struct *p, int cpu, int wake_flags)
3430
{
3431 3432
	lockdep_assert_held(&p->pi_lock);

3433
	if (p->nr_cpus_allowed > 1 && !is_migration_disabled(p))
3434
		cpu = p->sched_class->select_task_rq(p, cpu, wake_flags);
3435
	else
3436
		cpu = cpumask_any(p->cpus_ptr);
3437 3438 3439

	/*
	 * In order not to call set_task_cpu() on a blocking task we need
3440
	 * to rely on ttwu() to place the task on a valid ->cpus_ptr
3441
	 * CPU.
3442 3443 3444 3445 3446 3447
	 *
	 * 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 ]
	 */
3448
	if (unlikely(!is_cpu_allowed(p, cpu)))
3449
		cpu = select_fallback_rq(task_cpu(p), p);
3450 3451

	return cpu;
3452
}
3453

3454 3455
void sched_set_stop_task(int cpu, struct task_struct *stop)
{
3456
	static struct lock_class_key stop_pi_lock;
3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471
	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;
3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485

		/*
		 * The PI code calls rt_mutex_setprio() with ->pi_lock held to
		 * adjust the effective priority of a task. As a result,
		 * rt_mutex_setprio() can trigger (RT) balancing operations,
		 * which can then trigger wakeups of the stop thread to push
		 * around the current task.
		 *
		 * The stop task itself will never be part of the PI-chain, it
		 * never blocks, therefore that ->pi_lock recursion is safe.
		 * Tell lockdep about this by placing the stop->pi_lock in its
		 * own class.
		 */
		lockdep_set_class(&stop->pi_lock, &stop_pi_lock);
3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498
	}

	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;
	}
}

3499
#else /* CONFIG_SMP */
3500 3501

static inline int __set_cpus_allowed_ptr(struct task_struct *p,
3502 3503
					 const struct cpumask *new_mask,
					 u32 flags)
3504 3505 3506 3507
{
	return set_cpus_allowed_ptr(p, new_mask);
}

3508 3509
static inline void migrate_disable_switch(struct rq *rq, struct task_struct *p) { }

3510 3511 3512 3513 3514
static inline bool rq_has_pinned_tasks(struct rq *rq)
{
	return false;
}

3515
#endif /* !CONFIG_SMP */
3516

3517
static void
3518
ttwu_stat(struct task_struct *p, int cpu, int wake_flags)
Tejun Heo's avatar
Tejun Heo committed
3519
{
3520
	struct rq *rq;
3521

3522 3523 3524 3525
	if (!schedstat_enabled())
		return;

	rq = this_rq();
3526

3527 3528
#ifdef CONFIG_SMP
	if (cpu == rq->cpu) {
3529
		__schedstat_inc(rq->ttwu_local);
3530
		__schedstat_inc(p->stats.nr_wakeups_local);
3531 3532 3533
	} else {
		struct sched_domain *sd;

3534
		__schedstat_inc(p->stats.nr_wakeups_remote);
3535
		rcu_read_lock();
3536
		for_each_domain(rq->cpu, sd) {
3537
			if (cpumask_test_cpu(cpu, sched_domain_span(sd))) {
3538
				__schedstat_inc(sd->ttwu_wake_remote);
3539 3540 3541
				break;
			}
		}
3542
		rcu_read_unlock();
3543
	}
3544 3545

	if (wake_flags & WF_MIGRATED)
3546
		__schedstat_inc(p->stats.nr_wakeups_migrate);
3547 3548
#endif /* CONFIG_SMP */

3549
	__schedstat_inc(rq->ttwu_count);
3550
	__schedstat_inc(p->stats.nr_wakeups);
3551 3552

	if (wake_flags & WF_SYNC)
3553
		__schedstat_inc(p->stats.nr_wakeups_sync);
3554 3555
}

3556 3557 3558
/*
 * Mark the task runnable and perform wakeup-preemption.
 */
3559
static void ttwu_do_wakeup(struct rq *rq, struct task_struct *p, int wake_flags,
3560
			   struct rq_flags *rf)
Tejun Heo's avatar
Tejun Heo committed
3561 3562
{
	check_preempt_curr(rq, p, wake_flags);
3563
	WRITE_ONCE(p->__state, TASK_RUNNING);
3564 3565
	trace_sched_wakeup(p);

Tejun Heo's avatar
Tejun Heo committed
3566
#ifdef CONFIG_SMP
3567 3568
	if (p->sched_class->task_woken) {
		/*
3569
		 * Our task @p is fully woken up and running; so it's safe to
3570
		 * drop the rq->lock, hereafter rq is only used for statistics.
3571
		 */
3572
		rq_unpin_lock(rq, rf);
Tejun Heo's avatar
Tejun Heo committed
3573
		p->sched_class->task_woken(rq, p);
3574
		rq_repin_lock(rq, rf);
3575
	}
Tejun Heo's avatar
Tejun Heo committed
3576

3577
	if (rq->idle_stamp) {
3578
		u64 delta = rq_clock(rq) - rq->idle_stamp;
3579
		u64 max = 2*rq->max_idle_balance_cost;
Tejun Heo's avatar
Tejun Heo committed
3580

3581 3582 3583
		update_avg(&rq->avg_idle, delta);

		if (rq->avg_idle > max)
Tejun Heo's avatar
Tejun Heo committed
3584
			rq->avg_idle = max;
3585

3586 3587 3588
		rq->wake_stamp = jiffies;
		rq->wake_avg_idle = rq->avg_idle / 2;

Tejun Heo's avatar
Tejun Heo committed
3589 3590 3591 3592 3593
		rq->idle_stamp = 0;
	}
#endif
}

3594
static void
3595
ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags,
3596
		 struct rq_flags *rf)
3597
{
3598
	int en_flags = ENQUEUE_WAKEUP | ENQUEUE_NOCLOCK;
3599

3600
	lockdep_assert_rq_held(rq);
3601

3602 3603
	if (p->sched_contributes_to_load)
		rq->nr_uninterruptible--;
3604

3605
#ifdef CONFIG_SMP
3606
	if (wake_flags & WF_MIGRATED)
3607
		en_flags |= ENQUEUE_MIGRATED;
3608
	else
3609
#endif
3610 3611 3612 3613
	if (p->in_iowait) {
		delayacct_blkio_end(p);
		atomic_dec(&task_rq(p)->nr_iowait);
	}
3614

3615
	activate_task(rq, p, en_flags);
3616
	ttwu_do_wakeup(rq, p, wake_flags, rf);
3617 3618 3619
}

/*
3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642
 * Consider @p being inside a wait loop:
 *
 *   for (;;) {
 *      set_current_state(TASK_UNINTERRUPTIBLE);
 *
 *      if (CONDITION)
 *         break;
 *
 *      schedule();
 *   }
 *   __set_current_state(TASK_RUNNING);
 *
 * between set_current_state() and schedule(). In this case @p is still
 * runnable, so all that needs doing is change p->state back to TASK_RUNNING in
 * an atomic manner.
 *
 * By taking task_rq(p)->lock we serialize against schedule(), if @p->on_rq
 * then schedule() must still happen and p->state can be changed to
 * TASK_RUNNING. Otherwise we lost the race, schedule() has happened, and we
 * need to do a full wakeup with enqueue.
 *
 * Returns: %true when the wakeup is done,
 *          %false otherwise.
3643
 */
3644
static int ttwu_runnable(struct task_struct *p, int wake_flags)
3645
{
3646
	struct rq_flags rf;
3647 3648 3649
	struct rq *rq;
	int ret = 0;

3650
	rq = __task_rq_lock(p, &rf);
3651
	if (task_on_rq_queued(p)) {
3652 3653
		/* check_preempt_curr() may use rq clock */
		update_rq_clock(rq);
3654
		ttwu_do_wakeup(rq, p, wake_flags, &rf);
3655 3656
		ret = 1;
	}
3657
	__task_rq_unlock(rq, &rf);
3658 3659 3660 3661

	return ret;
}

3662
#ifdef CONFIG_SMP
3663
void sched_ttwu_pending(void *arg)
3664
{
3665
	struct llist_node *llist = arg;
3666
	struct rq *rq = this_rq();
3667
	struct task_struct *p, *t;
3668
	struct rq_flags rf;
3669

3670 3671 3672
	if (!llist)
		return;

3673 3674 3675 3676 3677 3678 3679
	/*
	 * rq::ttwu_pending racy indication of out-standing wakeups.
	 * Races such that false-negatives are possible, since they
	 * are shorter lived that false-positives would be.
	 */
	WRITE_ONCE(rq->ttwu_pending, 0);

3680
	rq_lock_irqsave(rq, &rf);
3681
	update_rq_clock(rq);
3682

3683
	llist_for_each_entry_safe(p, t, llist, wake_entry.llist) {
3684 3685 3686 3687 3688 3689
		if (WARN_ON_ONCE(p->on_cpu))
			smp_cond_load_acquire(&p->on_cpu, !VAL);

		if (WARN_ON_ONCE(task_cpu(p) != cpu_of(rq)))
			set_task_cpu(p, cpu_of(rq));

3690
		ttwu_do_activate(rq, p, p->sched_remote_wakeup ? WF_MIGRATED : 0, &rf);
3691
	}
3692

3693
	rq_unlock_irqrestore(rq, &rf);
3694 3695
}

3696
void send_call_function_single_ipi(int cpu)
3697
{
3698
	struct rq *rq = cpu_rq(cpu);
3699

3700 3701 3702 3703
	if (!set_nr_if_polling(rq->idle))
		arch_send_call_function_single_ipi(cpu);
	else
		trace_sched_wake_idle_without_ipi(cpu);
3704 3705
}

3706 3707 3708 3709 3710 3711 3712
/*
 * Queue a task on the target CPUs wake_list and wake the CPU via IPI if
 * necessary. The wakee CPU on receipt of the IPI will queue the task
 * via sched_ttwu_wakeup() for activation so the wakee incurs the cost
 * of the wakeup instead of the waker.
 */
static void __ttwu_queue_wakelist(struct task_struct *p, int cpu, int wake_flags)
3713
{
3714 3715
	struct rq *rq = cpu_rq(cpu);

3716 3717
	p->sched_remote_wakeup = !!(wake_flags & WF_MIGRATED);

3718
	WRITE_ONCE(rq->ttwu_pending, 1);
3719
	__smp_call_single_queue(cpu, &p->wake_entry.llist);
3720
}
3721

3722 3723 3724
void wake_up_if_idle(int cpu)
{
	struct rq *rq = cpu_rq(cpu);
3725
	struct rq_flags rf;
3726

3727 3728 3729 3730
	rcu_read_lock();

	if (!is_idle_task(rcu_dereference(rq->curr)))
		goto out;
3731

3732 3733 3734 3735 3736
	rq_lock_irqsave(rq, &rf);
	if (is_idle_task(rq->curr))
		resched_curr(rq);
	/* Else CPU is not idle, do nothing here: */
	rq_unlock_irqrestore(rq, &rf);
3737 3738 3739

out:
	rcu_read_unlock();
3740 3741
}

3742
bool cpus_share_cache(int this_cpu, int that_cpu)
3743
{
3744 3745 3746
	if (this_cpu == that_cpu)
		return true;

3747 3748
	return per_cpu(sd_llc_id, this_cpu) == per_cpu(sd_llc_id, that_cpu);
}
3749

3750 3751
static inline bool ttwu_queue_cond(int cpu, int wake_flags)
{
3752 3753 3754 3755 3756 3757 3758
	/*
	 * Do not complicate things with the async wake_list while the CPU is
	 * in hotplug state.
	 */
	if (!cpu_active(cpu))
		return false;

3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771
	/*
	 * If the CPU does not share cache, then queue the task on the
	 * remote rqs wakelist to avoid accessing remote data.
	 */
	if (!cpus_share_cache(smp_processor_id(), cpu))
		return true;

	/*
	 * If the task is descheduling and the only running task on the
	 * CPU then use the wakelist to offload the task activation to
	 * the soon-to-be-idle CPU as the current CPU is likely busy.
	 * nr_running is checked to avoid unnecessary task stacking.
	 */
3772
	if ((wake_flags & WF_ON_CPU) && cpu_rq(cpu)->nr_running <= 1)
3773 3774 3775 3776 3777 3778
		return true;

	return false;
}

static bool ttwu_queue_wakelist(struct task_struct *p, int cpu, int wake_flags)
3779
{
3780
	if (sched_feat(TTWU_QUEUE) && ttwu_queue_cond(cpu, wake_flags)) {
3781 3782 3783
		if (WARN_ON_ONCE(cpu == smp_processor_id()))
			return false;

3784
		sched_clock_cpu(cpu); /* Sync clocks across CPUs */
3785
		__ttwu_queue_wakelist(p, cpu, wake_flags);
3786 3787 3788 3789 3790
		return true;
	}

	return false;
}
3791 3792 3793 3794 3795 3796 3797 3798

#else /* !CONFIG_SMP */

static inline bool ttwu_queue_wakelist(struct task_struct *p, int cpu, int wake_flags)
{
	return false;
}

3799
#endif /* CONFIG_SMP */
3800

3801
static void ttwu_queue(struct task_struct *p, int cpu, int wake_flags)
3802 3803
{
	struct rq *rq = cpu_rq(cpu);
3804
	struct rq_flags rf;
3805

3806
	if (ttwu_queue_wakelist(p, cpu, wake_flags))
3807 3808
		return;

3809
	rq_lock(rq, &rf);
3810
	update_rq_clock(rq);
3811
	ttwu_do_activate(rq, p, wake_flags, &rf);
3812
	rq_unlock(rq, &rf);
Tejun Heo's avatar
Tejun Heo committed
3813 3814
}

3815 3816 3817 3818 3819
/*
 * Invoked from try_to_wake_up() to check whether the task can be woken up.
 *
 * The caller holds p::pi_lock if p != current or has preemption
 * disabled when p == current.
3820 3821 3822 3823 3824 3825 3826 3827
 *
 * The rules of PREEMPT_RT saved_state:
 *
 *   The related locking code always holds p::pi_lock when updating
 *   p::saved_state, which means the code is fully serialized in both cases.
 *
 *   The lock wait and lock wakeups happen via TASK_RTLOCK_WAIT. No other
 *   bits set. This allows to distinguish all wakeup scenarios.
3828 3829 3830 3831
 */
static __always_inline
bool ttwu_state_match(struct task_struct *p, unsigned int state, int *success)
{
3832 3833 3834 3835 3836
	if (IS_ENABLED(CONFIG_DEBUG_PREEMPT)) {
		WARN_ON_ONCE((state & TASK_RTLOCK_WAIT) &&
			     state != TASK_RTLOCK_WAIT);
	}

3837 3838 3839 3840
	if (READ_ONCE(p->__state) & state) {
		*success = 1;
		return true;
	}
3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860

#ifdef CONFIG_PREEMPT_RT
	/*
	 * Saved state preserves the task state across blocking on
	 * an RT lock.  If the state matches, set p::saved_state to
	 * TASK_RUNNING, but do not wake the task because it waits
	 * for a lock wakeup. Also indicate success because from
	 * the regular waker's point of view this has succeeded.
	 *
	 * After acquiring the lock the task will restore p::__state
	 * from p::saved_state which ensures that the regular
	 * wakeup is not lost. The restore will also set
	 * p::saved_state to TASK_RUNNING so any further tests will
	 * not result in false positives vs. @success
	 */
	if (p->saved_state & state) {
		p->saved_state = TASK_RUNNING;
		*success = 1;
	}
#endif
3861 3862 3863
	return false;
}

3864 3865 3866 3867 3868 3869
/*
 * Notes on Program-Order guarantees on SMP systems.
 *
 *  MIGRATION
 *
 * The basic program-order guarantee on SMP systems is that when a task [t]
3870 3871
 * migrates, all its activity on its old CPU [c0] happens-before any subsequent
 * execution on its new CPU [c1].
3872 3873 3874 3875 3876 3877 3878 3879
 *
 * 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
 *
3880
 * Release/acquire chaining guarantees that B happens after A and C after B.
3881
 * Note: the CPU doing B need not be c0 or c1
3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911
 *
 * 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:
 *
3912 3913
 *   1) smp_store_release(X->on_cpu, 0)   -- finish_task()
 *   2) smp_cond_load_acquire(!X->on_cpu) -- try_to_wake_up()
3914 3915 3916 3917 3918 3919 3920 3921 3922 3923
 *
 * 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);
 *
3924
 *                    smp_cond_load_acquire(&X->on_cpu, !VAL);
3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941
 *                    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
 *
 *
3942 3943 3944
 * 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().
3945 3946
 */

Tejun Heo's avatar
Tejun Heo committed
3947
/**
Linus Torvalds's avatar
Linus Torvalds committed
3948
 * try_to_wake_up - wake up a thread
Tejun Heo's avatar
Tejun Heo committed
3949
 * @p: the thread to be awakened
Linus Torvalds's avatar
Linus Torvalds committed
3950
 * @state: the mask of task states that can be woken
Tejun Heo's avatar
Tejun Heo committed
3951
 * @wake_flags: wake modifier flags (WF_*)
Linus Torvalds's avatar
Linus Torvalds committed
3952
 *
3953 3954 3955
 * Conceptually does:
 *
 *   If (@state & @p->state) @p->state = TASK_RUNNING.
Linus Torvalds's avatar
Linus Torvalds committed
3956
 *
3957 3958
 * If the task was not queued/runnable, also place it back on a runqueue.
 *
3959 3960 3961 3962
 * This function is atomic against schedule() which would dequeue the task.
 *
 * It issues a full memory barrier before accessing @p->state, see the comment
 * with set_current_state().
3963
 *
3964
 * Uses p->pi_lock to serialize against concurrent wake-ups.
3965
 *
3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979
 * Relies on p->pi_lock stabilizing:
 *  - p->sched_class
 *  - p->cpus_ptr
 *  - p->sched_task_group
 * in order to do migration, see its use of select_task_rq()/set_task_cpu().
 *
 * Tries really hard to only take one task_rq(p)->lock for performance.
 * Takes rq->lock in:
 *  - ttwu_runnable()    -- old rq, unavoidable, see comment there;
 *  - ttwu_queue()       -- new rq, for enqueue of the task;
 *  - psi_ttwu_dequeue() -- much sadness :-( accounting will kill us.
 *
 * As a consequence we race really badly with just about everything. See the
 * many memory barriers and their comments for details.
3980
 *
3981 3982
 * Return: %true if @p->state changes (an actual wakeup was done),
 *	   %false otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
3983
 */
3984 3985
static int
try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
Linus Torvalds's avatar
Linus Torvalds committed
3986 3987
{
	unsigned long flags;
3988
	int cpu, success = 0;
3989

3990
	preempt_disable();
3991 3992 3993 3994
	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
3995
		 * case the whole 'p->on_rq && ttwu_runnable()' case below
3996 3997 3998 3999 4000 4001 4002
		 * 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).
		 */
4003
		if (!ttwu_state_match(p, state, &success))
4004
			goto out;
4005 4006

		trace_sched_waking(p);
4007
		WRITE_ONCE(p->__state, TASK_RUNNING);
4008 4009 4010 4011
		trace_sched_wakeup(p);
		goto out;
	}

4012 4013 4014
	/*
	 * 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
4015 4016
	 * reordered with p->state check below. This pairs with smp_store_mb()
	 * in set_current_state() that the waiting thread does.
4017
	 */
4018
	raw_spin_lock_irqsave(&p->pi_lock, flags);
4019
	smp_mb__after_spinlock();
4020
	if (!ttwu_state_match(p, state, &success))
4021
		goto unlock;
Linus Torvalds's avatar
Linus Torvalds committed
4022

4023 4024
	trace_sched_waking(p);

4025 4026 4027 4028 4029
	/*
	 * 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.
	 *
4030 4031 4032 4033 4034 4035 4036 4037
	 * 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
4038 4039
	 *
	 * [task p]
4040
	 *   STORE p->state = UNINTERRUPTIBLE	  LOAD p->on_rq
4041
	 *
4042 4043
	 * Pairs with the LOCK+smp_mb__after_spinlock() on rq->lock in
	 * __schedule().  See the comment for smp_mb__after_spinlock().
4044 4045
	 *
	 * A similar smb_rmb() lives in try_invoke_on_locked_down_task().
4046 4047
	 */
	smp_rmb();
4048
	if (READ_ONCE(p->on_rq) && ttwu_runnable(p, wake_flags))
4049
		goto unlock;
Linus Torvalds's avatar
Linus Torvalds committed
4050 4051

#ifdef CONFIG_SMP
4052 4053 4054 4055 4056 4057 4058
	/*
	 * 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.
	 *
4059 4060 4061 4062 4063 4064 4065 4066
	 * __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
4067
	 *
4068 4069
	 * Pairs with the LOCK+smp_mb__after_spinlock() on rq->lock in
	 * __schedule().  See the comment for smp_mb__after_spinlock().
4070 4071 4072 4073
	 *
	 * Form a control-dep-acquire with p->on_rq == 0 above, to ensure
	 * schedule()'s deactivate_task() has 'happened' and p will no longer
	 * care about it's own p->state. See the comment in __schedule().
4074
	 */
4075 4076 4077 4078 4079 4080 4081 4082
	smp_acquire__after_ctrl_dep();

	/*
	 * We're doing the wakeup (@success == 1), they did a dequeue (p->on_rq
	 * == 0), which means we need to do an enqueue, change p->state to
	 * TASK_WAKING such that we can unlock p->pi_lock before doing the
	 * enqueue, such as ttwu_queue_wakelist().
	 */
4083
	WRITE_ONCE(p->__state, TASK_WAKING);
4084

4085 4086 4087 4088 4089 4090
	/*
	 * If the owning (remote) CPU is still in the middle of schedule() with
	 * this task as prev, considering queueing p on the remote CPUs wake_list
	 * which potentially sends an IPI instead of spinning on p->on_cpu to
	 * let the waker make forward progress. This is safe because IRQs are
	 * disabled and the IPI will deliver after on_cpu is cleared.
4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102
	 *
	 * Ensure we load task_cpu(p) after p->on_cpu:
	 *
	 * set_task_cpu(p, cpu);
	 *   STORE p->cpu = @cpu
	 * __schedule() (switch to task 'p')
	 *   LOCK rq->lock
	 *   smp_mb__after_spin_lock()		smp_cond_load_acquire(&p->on_cpu)
	 *   STORE p->on_cpu = 1		LOAD p->cpu
	 *
	 * to ensure we observe the correct CPU on which the task is currently
	 * scheduling.
4103
	 */
4104
	if (smp_load_acquire(&p->on_cpu) &&
4105
	    ttwu_queue_wakelist(p, task_cpu(p), wake_flags | WF_ON_CPU))
4106 4107
		goto unlock;

Peter Zijlstra's avatar
Peter Zijlstra committed
4108
	/*
4109
	 * If the owning (remote) CPU is still in the middle of schedule() with
4110
	 * this task as prev, wait until it's done referencing the task.
4111
	 *
4112
	 * Pairs with the smp_store_release() in finish_task().
4113 4114 4115
	 *
	 * This ensures that tasks getting woken will be fully ordered against
	 * their previous state and preserve Program Order.
4116
	 */
4117
	smp_cond_load_acquire(&p->on_cpu, !VAL);
Linus Torvalds's avatar
Linus Torvalds committed
4118

4119
	cpu = select_task_rq(p, p->wake_cpu, wake_flags | WF_TTWU);
4120
	if (task_cpu(p) != cpu) {
4121 4122 4123 4124 4125
		if (p->in_iowait) {
			delayacct_blkio_end(p);
			atomic_dec(&task_rq(p)->nr_iowait);
		}

4126
		wake_flags |= WF_MIGRATED;
4127
		psi_ttwu_dequeue(p);
4128
		set_task_cpu(p, cpu);
4129
	}
4130 4131
#else
	cpu = task_cpu(p);
Linus Torvalds's avatar
Linus Torvalds committed
4132 4133
#endif /* CONFIG_SMP */

4134
	ttwu_queue(p, cpu, wake_flags);
4135
unlock:
4136
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
4137 4138
out:
	if (success)
4139
		ttwu_stat(p, task_cpu(p), wake_flags);
4140
	preempt_enable();
Linus Torvalds's avatar
Linus Torvalds committed
4141 4142 4143 4144

	return success;
}

4145
/**
4146
 * task_call_func - Invoke a function on task in fixed state
4147
 * @p: Process for which the function is to be invoked, can be @current.
4148 4149 4150
 * @func: Function to invoke.
 * @arg: Argument to function.
 *
4151 4152 4153 4154
 * Fix the task in it's current state by avoiding wakeups and or rq operations
 * and call @func(@arg) on it.  This function can use ->on_rq and task_curr()
 * to work out what the state is, if required.  Given that @func can be invoked
 * with a runqueue lock held, it had better be quite lightweight.
4155 4156
 *
 * Returns:
4157
 *   Whatever @func returns
4158
 */
4159
int task_call_func(struct task_struct *p, task_call_f func, void *arg)
4160
{
4161 4162
	struct rq *rq = NULL;
	unsigned int state;
4163
	struct rq_flags rf;
4164
	int ret;
4165

4166
	raw_spin_lock_irqsave(&p->pi_lock, rf.flags);
4167 4168 4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182 4183

	state = READ_ONCE(p->__state);

	/*
	 * Ensure we load p->on_rq after p->__state, otherwise it would be
	 * possible to, falsely, observe p->on_rq == 0.
	 *
	 * See try_to_wake_up() for a longer comment.
	 */
	smp_rmb();

	/*
	 * Since pi->lock blocks try_to_wake_up(), we don't need rq->lock when
	 * the task is blocked. Make sure to check @state since ttwu() can drop
	 * locks at the end, see ttwu_queue_wakelist().
	 */
	if (state == TASK_RUNNING || state == TASK_WAKING || p->on_rq)
4184
		rq = __task_rq_lock(p, &rf);
4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198

	/*
	 * At this point the task is pinned; either:
	 *  - blocked and we're holding off wakeups	 (pi->lock)
	 *  - woken, and we're holding off enqueue	 (rq->lock)
	 *  - queued, and we're holding off schedule	 (rq->lock)
	 *  - running, and we're holding off de-schedule (rq->lock)
	 *
	 * The called function (@func) can use: task_curr(), p->on_rq and
	 * p->__state to differentiate between these states.
	 */
	ret = func(p, arg);

	if (rq)
4199
		rq_unlock(rq, &rf);
4200

4201
	raw_spin_unlock_irqrestore(&p->pi_lock, rf.flags);
4202 4203 4204
	return ret;
}

4205 4206 4207 4208 4209
/**
 * 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
4210 4211 4212
 * processes.
 *
 * Return: 1 if the process was woken up, 0 if it was already running.
4213
 *
4214
 * This function executes a full memory barrier before accessing the task state.
4215
 */
4216
int wake_up_process(struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
4217
{
4218
	return try_to_wake_up(p, TASK_NORMAL, 0);
Linus Torvalds's avatar
Linus Torvalds committed
4219 4220 4221
}
EXPORT_SYMBOL(wake_up_process);

4222
int wake_up_state(struct task_struct *p, unsigned int state)
Linus Torvalds's avatar
Linus Torvalds committed
4223 4224 4225 4226 4227 4228 4229
{
	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
4230 4231 4232
 *
 * __sched_fork() is basic setup used by init_idle() too:
 */
4233
static void __sched_fork(unsigned long clone_flags, struct task_struct *p)
Ingo Molnar's avatar
Ingo Molnar committed
4234
{
Peter Zijlstra's avatar
Peter Zijlstra committed
4235 4236 4237
	p->on_rq			= 0;

	p->se.on_rq			= 0;
Ingo Molnar's avatar
Ingo Molnar committed
4238 4239
	p->se.exec_start		= 0;
	p->se.sum_exec_runtime		= 0;
4240
	p->se.prev_sum_exec_runtime	= 0;
4241
	p->se.nr_migrations		= 0;
4242
	p->se.vruntime			= 0;
Peter Zijlstra's avatar
Peter Zijlstra committed
4243
	INIT_LIST_HEAD(&p->se.group_node);
4244

4245 4246 4247 4248
#ifdef CONFIG_FAIR_GROUP_SCHED
	p->se.cfs_rq			= NULL;
#endif

4249
#ifdef CONFIG_SCHEDSTATS
4250
	/* Even if schedstat is disabled, there should not be garbage */
4251
	memset(&p->stats, 0, sizeof(p->stats));
4252
#endif
4253

4254
	RB_CLEAR_NODE(&p->dl.rb_node);
4255
	init_dl_task_timer(&p->dl);
4256
	init_dl_inactive_task_timer(&p->dl);
4257
	__dl_clear_params(p);
4258

Peter Zijlstra's avatar
Peter Zijlstra committed
4259
	INIT_LIST_HEAD(&p->rt.run_list);
4260 4261 4262 4263
	p->rt.timeout		= 0;
	p->rt.time_slice	= sched_rr_timeslice;
	p->rt.on_rq		= 0;
	p->rt.on_list		= 0;
4264

4265 4266 4267
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&p->preempt_notifiers);
#endif
4268

4269 4270 4271
#ifdef CONFIG_COMPACTION
	p->capture_control = NULL;
#endif
4272
	init_numa_balancing(clone_flags, p);
4273
#ifdef CONFIG_SMP
4274
	p->wake_entry.u_flags = CSD_TYPE_TTWU;
4275
	p->migration_pending = NULL;
4276
#endif
Ingo Molnar's avatar
Ingo Molnar committed
4277 4278
}

4279 4280
DEFINE_STATIC_KEY_FALSE(sched_numa_balancing);

4281
#ifdef CONFIG_NUMA_BALANCING
4282

4283 4284 4285
void set_numabalancing_state(bool enabled)
{
	if (enabled)
4286
		static_branch_enable(&sched_numa_balancing);
4287
	else
4288
		static_branch_disable(&sched_numa_balancing);
4289
}
4290 4291 4292

#ifdef CONFIG_PROC_SYSCTL
int sysctl_numa_balancing(struct ctl_table *table, int write,
4293
			  void *buffer, size_t *lenp, loff_t *ppos)
4294 4295 4296
{
	struct ctl_table t;
	int err;
4297
	int state = static_branch_likely(&sched_numa_balancing);
4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312

	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
4313

4314 4315
#ifdef CONFIG_SCHEDSTATS

4316 4317 4318 4319 4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334 4335 4336 4337 4338 4339 4340
DEFINE_STATIC_KEY_FALSE(sched_schedstats);

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;

	if (!strcmp(str, "enable")) {
4341
		set_schedstats(true);
4342 4343
		ret = 1;
	} else if (!strcmp(str, "disable")) {
4344
		set_schedstats(false);
4345 4346 4347 4348 4349 4350 4351 4352 4353 4354 4355
		ret = 1;
	}
out:
	if (!ret)
		pr_warn("Unable to parse schedstats=\n");

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

#ifdef CONFIG_PROC_SYSCTL
4356 4357
int sysctl_schedstats(struct ctl_table *table, int write, void *buffer,
		size_t *lenp, loff_t *ppos)
4358 4359 4360 4361 4362 4363 4364 4365 4366 4367 4368 4369 4370 4371 4372 4373 4374
{
	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;
}
4375 4376
#endif /* CONFIG_PROC_SYSCTL */
#endif /* CONFIG_SCHEDSTATS */
Ingo Molnar's avatar
Ingo Molnar committed
4377 4378 4379 4380

/*
 * fork()/clone()-time setup:
 */
4381
int sched_fork(unsigned long clone_flags, struct task_struct *p)
Ingo Molnar's avatar
Ingo Molnar committed
4382
{
4383
	__sched_fork(clone_flags, p);
4384
	/*
4385
	 * We mark the process as NEW here. This guarantees that
4386 4387 4388
	 * nobody will actually run it, and a signal or other external
	 * event cannot wake it up and insert it on the runqueue either.
	 */
4389
	p->__state = TASK_NEW;
Ingo Molnar's avatar
Ingo Molnar committed
4390

4391 4392 4393 4394 4395
	/*
	 * Make sure we do not leak PI boosting priority to the child.
	 */
	p->prio = current->normal_prio;

4396 4397
	uclamp_fork(p);

4398 4399 4400 4401
	/*
	 * Revert to default priority/policy on fork if requested.
	 */
	if (unlikely(p->sched_reset_on_fork)) {
4402
		if (task_has_dl_policy(p) || task_has_rt_policy(p)) {
4403
			p->policy = SCHED_NORMAL;
4404
			p->static_prio = NICE_TO_PRIO(0);
4405 4406 4407 4408
			p->rt_priority = 0;
		} else if (PRIO_TO_NICE(p->static_prio) < 0)
			p->static_prio = NICE_TO_PRIO(0);

4409
		p->prio = p->normal_prio = p->static_prio;
4410
		set_load_weight(p, false);
4411

4412 4413 4414 4415 4416 4417
		/*
		 * We don't need the reset flag anymore after the fork. It has
		 * fulfilled its duty:
		 */
		p->sched_reset_on_fork = 0;
	}
4418

4419
	if (dl_prio(p->prio))
4420
		return -EAGAIN;
4421
	else if (rt_prio(p->prio))
4422
		p->sched_class = &rt_sched_class;
4423
	else
4424
		p->sched_class = &fair_sched_class;
4425

4426
	init_entity_runnable_average(&p->se);
4427

4428
#ifdef CONFIG_SCHED_INFO
Ingo Molnar's avatar
Ingo Molnar committed
4429
	if (likely(sched_info_on()))
4430
		memset(&p->sched_info, 0, sizeof(p->sched_info));
Linus Torvalds's avatar
Linus Torvalds committed
4431
#endif
4432 4433
#if defined(CONFIG_SMP)
	p->on_cpu = 0;
4434
#endif
4435
	init_task_preempt_count(p);
4436
#ifdef CONFIG_SMP
4437
	plist_node_init(&p->pushable_tasks, MAX_PRIO);
4438
	RB_CLEAR_NODE(&p->pushable_dl_tasks);
4439
#endif
4440
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
4441 4442
}

4443
void sched_post_fork(struct task_struct *p, struct kernel_clone_args *kargs)
4444
{
4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465
	unsigned long flags;
#ifdef CONFIG_CGROUP_SCHED
	struct task_group *tg;
#endif

	raw_spin_lock_irqsave(&p->pi_lock, flags);
#ifdef CONFIG_CGROUP_SCHED
	tg = container_of(kargs->cset->subsys[cpu_cgrp_id],
			  struct task_group, css);
	p->sched_task_group = autogroup_task_group(p, tg);
#endif
	rseq_migrate(p);
	/*
	 * We're setting the CPU for the first time, we don't migrate,
	 * so use __set_task_cpu().
	 */
	__set_task_cpu(p, smp_processor_id());
	if (p->sched_class->task_fork)
		p->sched_class->task_fork(p);
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);

4466 4467 4468
	uclamp_post_fork(p);
}

4469 4470 4471
unsigned long to_ratio(u64 period, u64 runtime)
{
	if (runtime == RUNTIME_INF)
4472
		return BW_UNIT;
4473 4474 4475 4476 4477 4478 4479 4480 4481

	/*
	 * 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;

4482
	return div64_u64(runtime << BW_SHIFT, period);
4483 4484
}

Linus Torvalds's avatar
Linus Torvalds committed
4485 4486 4487 4488 4489 4490 4491
/*
 * 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.
 */
4492
void wake_up_new_task(struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
4493
{
4494
	struct rq_flags rf;
Ingo Molnar's avatar
Ingo Molnar committed
4495
	struct rq *rq;
4496

4497
	raw_spin_lock_irqsave(&p->pi_lock, rf.flags);
4498
	WRITE_ONCE(p->__state, TASK_RUNNING);
4499 4500 4501
#ifdef CONFIG_SMP
	/*
	 * Fork balancing, do it here and not earlier because:
4502
	 *  - cpus_ptr can change in the fork path
4503
	 *  - any previously selected CPU might disappear through hotplug
4504 4505 4506
	 *
	 * Use __set_task_cpu() to avoid calling sched_class::migrate_task_rq,
	 * as we're not fully set-up yet.
4507
	 */
4508
	p->recent_used_cpu = task_cpu(p);
4509
	rseq_migrate(p);
4510
	__set_task_cpu(p, select_task_rq(p, task_cpu(p), WF_FORK));
4511
#endif
4512
	rq = __task_rq_lock(p, &rf);
4513
	update_rq_clock(rq);
4514
	post_init_entity_util_avg(p);
4515

4516
	activate_task(rq, p, ENQUEUE_NOCLOCK);
4517
	trace_sched_wakeup_new(p);
Peter Zijlstra's avatar
Peter Zijlstra committed
4518
	check_preempt_curr(rq, p, WF_FORK);
4519
#ifdef CONFIG_SMP
4520 4521
	if (p->sched_class->task_woken) {
		/*
4522
		 * Nothing relies on rq->lock after this, so it's fine to
4523 4524
		 * drop it.
		 */
4525
		rq_unpin_lock(rq, &rf);
4526
		p->sched_class->task_woken(rq, p);
4527
		rq_repin_lock(rq, &rf);
4528
	}
4529
#endif
4530
	task_rq_unlock(rq, p, &rf);
Linus Torvalds's avatar
Linus Torvalds committed
4531 4532
}

4533 4534
#ifdef CONFIG_PREEMPT_NOTIFIERS

4535
static DEFINE_STATIC_KEY_FALSE(preempt_notifier_key);
4536

4537 4538
void preempt_notifier_inc(void)
{
4539
	static_branch_inc(&preempt_notifier_key);
4540 4541 4542 4543 4544
}
EXPORT_SYMBOL_GPL(preempt_notifier_inc);

void preempt_notifier_dec(void)
{
4545
	static_branch_dec(&preempt_notifier_key);
4546 4547 4548
}
EXPORT_SYMBOL_GPL(preempt_notifier_dec);

4549
/**
4550
 * preempt_notifier_register - tell me when current is being preempted & rescheduled
4551
 * @notifier: notifier struct to register
4552 4553 4554
 */
void preempt_notifier_register(struct preempt_notifier *notifier)
{
4555
	if (!static_branch_unlikely(&preempt_notifier_key))
4556 4557
		WARN(1, "registering preempt_notifier while notifiers disabled\n");

4558 4559 4560 4561 4562 4563
	hlist_add_head(&notifier->link, &current->preempt_notifiers);
}
EXPORT_SYMBOL_GPL(preempt_notifier_register);

/**
 * preempt_notifier_unregister - no longer interested in preemption notifications
4564
 * @notifier: notifier struct to unregister
4565
 *
4566
 * This is *not* safe to call from within a preemption notifier.
4567 4568 4569 4570 4571 4572 4573
 */
void preempt_notifier_unregister(struct preempt_notifier *notifier)
{
	hlist_del(&notifier->link);
}
EXPORT_SYMBOL_GPL(preempt_notifier_unregister);

4574
static void __fire_sched_in_preempt_notifiers(struct task_struct *curr)
4575 4576 4577
{
	struct preempt_notifier *notifier;

4578
	hlist_for_each_entry(notifier, &curr->preempt_notifiers, link)
4579 4580 4581
		notifier->ops->sched_in(notifier, raw_smp_processor_id());
}

4582 4583
static __always_inline void fire_sched_in_preempt_notifiers(struct task_struct *curr)
{
4584
	if (static_branch_unlikely(&preempt_notifier_key))
4585 4586 4587
		__fire_sched_in_preempt_notifiers(curr);
}

4588
static void
4589 4590
__fire_sched_out_preempt_notifiers(struct task_struct *curr,
				   struct task_struct *next)
4591 4592 4593
{
	struct preempt_notifier *notifier;

4594
	hlist_for_each_entry(notifier, &curr->preempt_notifiers, link)
4595 4596 4597
		notifier->ops->sched_out(notifier, next);
}

4598 4599 4600 4601
static __always_inline void
fire_sched_out_preempt_notifiers(struct task_struct *curr,
				 struct task_struct *next)
{
4602
	if (static_branch_unlikely(&preempt_notifier_key))
4603 4604 4605
		__fire_sched_out_preempt_notifiers(curr, next);
}

4606
#else /* !CONFIG_PREEMPT_NOTIFIERS */
4607

4608
static inline void fire_sched_in_preempt_notifiers(struct task_struct *curr)
4609 4610 4611
{
}

4612
static inline void
4613 4614 4615 4616 4617
fire_sched_out_preempt_notifiers(struct task_struct *curr,
				 struct task_struct *next)
{
}

4618
#endif /* CONFIG_PREEMPT_NOTIFIERS */
4619

4620 4621 4622 4623 4624 4625
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.
4626 4627
	 *
	 * See the ttwu() WF_ON_CPU case and its ordering comment.
4628
	 */
4629
	WRITE_ONCE(next->on_cpu, 1);
4630 4631 4632 4633 4634 4635 4636
#endif
}

static inline void finish_task(struct task_struct *prev)
{
#ifdef CONFIG_SMP
	/*
4637 4638 4639
	 * This must be the very last reference to @prev from this CPU. After
	 * p->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
4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650
	 * 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
}

4651 4652 4653 4654 4655 4656 4657
#ifdef CONFIG_SMP

static void do_balance_callbacks(struct rq *rq, struct callback_head *head)
{
	void (*func)(struct rq *rq);
	struct callback_head *next;

4658
	lockdep_assert_rq_held(rq);
4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669

	while (head) {
		func = (void (*)(struct rq *))head->func;
		next = head->next;
		head->next = NULL;
		head = next;

		func(rq);
	}
}

4670 4671 4672 4673 4674 4675 4676
static void balance_push(struct rq *rq);

struct callback_head balance_push_callback = {
	.next = NULL,
	.func = (void (*)(struct callback_head *))balance_push,
};

4677 4678 4679 4680
static inline struct callback_head *splice_balance_callbacks(struct rq *rq)
{
	struct callback_head *head = rq->balance_callback;

4681
	lockdep_assert_rq_held(rq);
4682
	if (head)
4683 4684 4685 4686 4687 4688 4689 4690 4691 4692 4693 4694 4695 4696 4697
		rq->balance_callback = NULL;

	return head;
}

static void __balance_callbacks(struct rq *rq)
{
	do_balance_callbacks(rq, splice_balance_callbacks(rq));
}

static inline void balance_callbacks(struct rq *rq, struct callback_head *head)
{
	unsigned long flags;

	if (unlikely(head)) {
4698
		raw_spin_rq_lock_irqsave(rq, flags);
4699
		do_balance_callbacks(rq, head);
4700
		raw_spin_rq_unlock_irqrestore(rq, flags);
4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715 4716 4717 4718 4719 4720
	}
}

#else

static inline void __balance_callbacks(struct rq *rq)
{
}

static inline struct callback_head *splice_balance_callbacks(struct rq *rq)
{
	return NULL;
}

static inline void balance_callbacks(struct rq *rq, struct callback_head *head)
{
}

#endif

4721 4722
static inline void
prepare_lock_switch(struct rq *rq, struct task_struct *next, struct rq_flags *rf)
4723
{
4724 4725 4726 4727 4728 4729 4730
	/*
	 * 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);
4731
	spin_release(&__rq_lockp(rq)->dep_map, _THIS_IP_);
4732 4733
#ifdef CONFIG_DEBUG_SPINLOCK
	/* this is a valid case when another task releases the spinlock */
4734
	rq_lockp(rq)->owner = next;
4735
#endif
4736 4737 4738 4739
}

static inline void finish_lock_switch(struct rq *rq)
{
4740 4741 4742 4743 4744
	/*
	 * If we are tracking spinlock dependencies then we have to
	 * fix up the runqueue lock - which gets 'carried over' from
	 * prev into current:
	 */
4745
	spin_acquire(&__rq_lockp(rq)->dep_map, 0, 0, _THIS_IP_);
4746
	__balance_callbacks(rq);
4747
	raw_spin_rq_unlock_irq(rq);
4748 4749
}

4750 4751 4752 4753 4754 4755 4756 4757 4758 4759 4760 4761
/*
 * 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

4762 4763 4764 4765 4766 4767 4768 4769 4770 4771 4772 4773 4774 4775 4776 4777
static inline void kmap_local_sched_out(void)
{
#ifdef CONFIG_KMAP_LOCAL
	if (unlikely(current->kmap_ctrl.idx))
		__kmap_local_sched_out();
#endif
}

static inline void kmap_local_sched_in(void)
{
#ifdef CONFIG_KMAP_LOCAL
	if (unlikely(current->kmap_ctrl.idx))
		__kmap_local_sched_in();
#endif
}

4778 4779 4780
/**
 * prepare_task_switch - prepare to switch tasks
 * @rq: the runqueue preparing to switch
4781
 * @prev: the current task that is being switched out
4782 4783 4784 4785 4786 4787 4788 4789 4790
 * @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.
 */
4791 4792 4793
static inline void
prepare_task_switch(struct rq *rq, struct task_struct *prev,
		    struct task_struct *next)
4794
{
4795
	kcov_prepare_switch(prev);
4796
	sched_info_switch(rq, prev, next);
4797
	perf_event_task_sched_out(prev, next);
4798
	rseq_preempt(prev);
4799
	fire_sched_out_preempt_notifiers(prev, next);
4800
	kmap_local_sched_out();
4801
	prepare_task(next);
4802 4803 4804
	prepare_arch_switch(next);
}

Linus Torvalds's avatar
Linus Torvalds committed
4805 4806 4807 4808
/**
 * finish_task_switch - clean up after a task-switch
 * @prev: the thread we just switched away from.
 *
4809 4810 4811 4812
 * 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
4813 4814
 *
 * Note that we may have delayed dropping an mm in context_switch(). If
Ingo Molnar's avatar
Ingo Molnar committed
4815
 * so, we finish that here outside of the runqueue lock. (Doing it
Linus Torvalds's avatar
Linus Torvalds committed
4816 4817
 * with the lock held can cause deadlocks; see schedule() for
 * details.)
4818 4819 4820 4821 4822
 *
 * 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
4823
 */
4824
static struct rq *finish_task_switch(struct task_struct *prev)
Linus Torvalds's avatar
Linus Torvalds committed
4825 4826
	__releases(rq->lock)
{
4827
	struct rq *rq = this_rq();
Linus Torvalds's avatar
Linus Torvalds committed
4828
	struct mm_struct *mm = rq->prev_mm;
Oleg Nesterov's avatar
Oleg Nesterov committed
4829
	long prev_state;
Linus Torvalds's avatar
Linus Torvalds committed
4830

4831 4832 4833 4834 4835 4836 4837 4838 4839 4840 4841
	/*
	 * 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.
	 */
4842 4843 4844 4845
	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);
4846

Linus Torvalds's avatar
Linus Torvalds committed
4847 4848 4849 4850
	rq->prev_mm = NULL;

	/*
	 * A task struct has one reference for the use as "current".
4851
	 * If a task dies, then it sets TASK_DEAD in tsk->state and calls
Oleg Nesterov's avatar
Oleg Nesterov committed
4852 4853
	 * schedule one last time. The schedule call will never return, and
	 * the scheduled task must drop that reference.
4854 4855
	 *
	 * We must observe prev->state before clearing prev->on_cpu (in
4856
	 * finish_task), otherwise a concurrent wakeup can get prev
4857 4858
	 * 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
4859
	 */
4860
	prev_state = READ_ONCE(prev->__state);
4861
	vtime_task_switch(prev);
4862
	perf_event_task_sched_in(prev, current);
4863
	finish_task(prev);
4864
	tick_nohz_task_switch();
4865
	finish_lock_switch(rq);
4866
	finish_arch_post_lock_switch();
4867
	kcov_finish_switch(current);
4868 4869 4870 4871 4872 4873 4874 4875
	/*
	 * kmap_local_sched_out() is invoked with rq::lock held and
	 * interrupts disabled. There is no requirement for that, but the
	 * sched out code does not have an interrupt enabled section.
	 * Restoring the maps on sched in does not require interrupts being
	 * disabled either.
	 */
	kmap_local_sched_in();
Steven Rostedt's avatar
Steven Rostedt committed
4876

4877
	fire_sched_in_preempt_notifiers(current);
4878
	/*
4879 4880 4881 4882 4883 4884 4885 4886 4887 4888
	 * 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.
4889
	 */
4890 4891
	if (mm) {
		membarrier_mm_sync_core_before_usermode(mm);
4892
		mmdrop_sched(mm);
4893
	}
4894 4895 4896
	if (unlikely(prev_state == TASK_DEAD)) {
		if (prev->sched_class->task_dead)
			prev->sched_class->task_dead(prev);
4897

4898 4899 4900
		/* Task is done with its stack. */
		put_task_stack(prev);

4901
		put_task_struct_rcu_user(prev);
4902
	}
4903

4904
	return rq;
Linus Torvalds's avatar
Linus Torvalds committed
4905 4906 4907 4908 4909 4910
}

/**
 * schedule_tail - first thing a freshly forked thread must call.
 * @prev: the thread we just switched away from.
 */
4911
asmlinkage __visible void schedule_tail(struct task_struct *prev)
Linus Torvalds's avatar
Linus Torvalds committed
4912 4913
	__releases(rq->lock)
{
4914 4915 4916 4917 4918 4919 4920 4921 4922
	/*
	 * 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).
	 */

4923
	finish_task_switch(prev);
4924
	preempt_enable();
4925

Linus Torvalds's avatar
Linus Torvalds committed
4926
	if (current->set_child_tid)
4927
		put_user(task_pid_vnr(current), current->set_child_tid);
4928 4929

	calculate_sigpending();
Linus Torvalds's avatar
Linus Torvalds committed
4930 4931 4932
}

/*
4933
 * context_switch - switch to the new MM and the new thread's register state.
Linus Torvalds's avatar
Linus Torvalds committed
4934
 */
4935
static __always_inline struct rq *
4936
context_switch(struct rq *rq, struct task_struct *prev,
4937
	       struct task_struct *next, struct rq_flags *rf)
Linus Torvalds's avatar
Linus Torvalds committed
4938
{
4939
	prepare_task_switch(rq, prev, next);
4940

4941 4942 4943 4944 4945
	/*
	 * For paravirt, this is coupled with an exit in switch_to to
	 * combine the page table reload and the switch backend into
	 * one hypercall.
	 */
4946
	arch_start_context_switch(prev);
4947

4948
	/*
4949 4950 4951 4952 4953
	 * kernel -> kernel   lazy + transfer active
	 *   user -> kernel   lazy + mmgrab() active
	 *
	 * kernel ->   user   switch + mmdrop() active
	 *   user ->   user   switch
4954
	 */
4955 4956 4957 4958 4959 4960 4961 4962 4963
	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
4964
		membarrier_switch_mm(rq, prev->active_mm, next->mm);
4965 4966
		/*
		 * sys_membarrier() requires an smp_mb() between setting
4967
		 * rq->curr / membarrier_switch_mm() and returning to userspace.
4968 4969 4970 4971 4972 4973
		 *
		 * 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
4974

4975 4976 4977 4978 4979
		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
4980
	}
4981

4982
	rq->clock_update_flags &= ~(RQCF_ACT_SKIP|RQCF_REQ_SKIP);
4983

4984
	prepare_lock_switch(rq, next, rf);
Linus Torvalds's avatar
Linus Torvalds committed
4985 4986 4987

	/* Here we just switch the register state and the stack. */
	switch_to(prev, next, prev);
Ingo Molnar's avatar
Ingo Molnar committed
4988
	barrier();
4989 4990

	return finish_task_switch(prev);
Linus Torvalds's avatar
Linus Torvalds committed
4991 4992 4993
}

/*
4994
 * nr_running and nr_context_switches:
Linus Torvalds's avatar
Linus Torvalds committed
4995 4996
 *
 * externally visible scheduler statistics: current number of runnable
4997
 * threads, total number of context switches performed since bootup.
Linus Torvalds's avatar
Linus Torvalds committed
4998
 */
4999
unsigned int nr_running(void)
Linus Torvalds's avatar
Linus Torvalds committed
5000
{
5001
	unsigned int i, sum = 0;
Linus Torvalds's avatar
Linus Torvalds committed
5002 5003 5004 5005 5006

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

	return sum;
5007
}
Linus Torvalds's avatar
Linus Torvalds committed
5008

5009
/*
5010
 * Check if only the current task is running on the CPU.
5011 5012 5013 5014 5015
 *
 * 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:
 *
5016
 * - from a non-preemptible section (of course)
5017 5018 5019 5020
 *
 * - from a thread that is bound to a single CPU
 *
 * - in a loop with very short iterations (e.g. a polling loop)
5021 5022 5023
 */
bool single_task_running(void)
{
5024
	return raw_rq()->nr_running == 1;
5025 5026 5027
}
EXPORT_SYMBOL(single_task_running);

Linus Torvalds's avatar
Linus Torvalds committed
5028
unsigned long long nr_context_switches(void)
5029
{
5030 5031
	int i;
	unsigned long long sum = 0;
5032

5033
	for_each_possible_cpu(i)
Linus Torvalds's avatar
Linus Torvalds committed
5034
		sum += cpu_rq(i)->nr_switches;
5035

Linus Torvalds's avatar
Linus Torvalds committed
5036 5037
	return sum;
}
5038

5039 5040 5041 5042 5043 5044 5045
/*
 * 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.
 */

5046
unsigned int nr_iowait_cpu(int cpu)
5047 5048 5049 5050
{
	return atomic_read(&cpu_rq(cpu)->nr_iowait);
}

5051
/*
5052
 * IO-wait accounting, and how it's mostly bollocks (on SMP).
5053 5054 5055 5056 5057 5058 5059 5060 5061 5062 5063 5064 5065 5066 5067 5068 5069 5070 5071 5072 5073 5074 5075 5076 5077 5078 5079 5080
 *
 * 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'.
 */

5081
unsigned int nr_iowait(void)
Linus Torvalds's avatar
Linus Torvalds committed
5082
{
5083
	unsigned int i, sum = 0;
5084

5085
	for_each_possible_cpu(i)
5086
		sum += nr_iowait_cpu(i);
5087

Linus Torvalds's avatar
Linus Torvalds committed
5088 5089
	return sum;
}
5090

Ingo Molnar's avatar
Ingo Molnar committed
5091
#ifdef CONFIG_SMP
5092

5093
/*
5094 5095
 * sched_exec - execve() is a valuable balancing opportunity, because at
 * this point the task has the smallest effective memory and cache footprint.
5096
 */
5097
void sched_exec(void)
5098
{
5099
	struct task_struct *p = current;
Linus Torvalds's avatar
Linus Torvalds committed
5100
	unsigned long flags;
5101
	int dest_cpu;
5102

5103
	raw_spin_lock_irqsave(&p->pi_lock, flags);
5104
	dest_cpu = p->sched_class->select_task_rq(p, task_cpu(p), WF_EXEC);
5105 5106
	if (dest_cpu == smp_processor_id())
		goto unlock;
5107

5108
	if (likely(cpu_active(dest_cpu))) {
5109
		struct migration_arg arg = { p, dest_cpu };
5110

5111 5112
		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
5113 5114
		return;
	}
5115
unlock:
5116
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
5117
}
Ingo Molnar's avatar
Ingo Molnar committed
5118

Linus Torvalds's avatar
Linus Torvalds committed
5119 5120 5121
#endif

DEFINE_PER_CPU(struct kernel_stat, kstat);
5122
DEFINE_PER_CPU(struct kernel_cpustat, kernel_cpustat);
Linus Torvalds's avatar
Linus Torvalds committed
5123 5124

EXPORT_PER_CPU_SYMBOL(kstat);
5125
EXPORT_PER_CPU_SYMBOL(kernel_cpustat);
Linus Torvalds's avatar
Linus Torvalds committed
5126

5127 5128 5129 5130 5131 5132 5133 5134 5135 5136 5137 5138 5139 5140 5141 5142 5143
/*
 * 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);
}

5144 5145 5146 5147 5148 5149 5150
/*
 * 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)
{
5151
	struct rq_flags rf;
5152
	struct rq *rq;
5153
	u64 ns;
5154

5155 5156
#if defined(CONFIG_64BIT) && defined(CONFIG_SMP)
	/*
5157
	 * 64-bit doesn't need locks to atomically read a 64-bit value.
5158 5159 5160
	 * So we have a optimization chance when the task's delta_exec is 0.
	 * Reading ->on_cpu is racy, but this is ok.
	 *
5161 5162
	 * 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
5163
	 * indistinguishable from the read occurring a few cycles earlier.
5164 5165
	 * If we see ->on_cpu without ->on_rq, the task is leaving, and has
	 * been accounted, so we're correct here as well.
5166
	 */
5167
	if (!p->on_cpu || !task_on_rq_queued(p))
5168 5169 5170
		return p->se.sum_exec_runtime;
#endif

5171
	rq = task_rq_lock(p, &rf);
5172 5173 5174 5175 5176 5177
	/*
	 * 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)) {
5178
		prefetch_curr_exec_start(p);
5179 5180 5181 5182
		update_rq_clock(rq);
		p->sched_class->update_curr(rq);
	}
	ns = p->se.sum_exec_runtime;
5183
	task_rq_unlock(rq, p, &rf);
5184 5185 5186

	return ns;
}
5187

5188 5189 5190 5191 5192 5193 5194 5195 5196 5197 5198 5199 5200 5201 5202 5203 5204 5205 5206 5207 5208 5209 5210 5211 5212 5213 5214 5215 5216 5217 5218 5219 5220 5221 5222 5223 5224 5225 5226 5227 5228 5229 5230 5231 5232 5233 5234 5235 5236
#ifdef CONFIG_SCHED_DEBUG
static u64 cpu_resched_latency(struct rq *rq)
{
	int latency_warn_ms = READ_ONCE(sysctl_resched_latency_warn_ms);
	u64 resched_latency, now = rq_clock(rq);
	static bool warned_once;

	if (sysctl_resched_latency_warn_once && warned_once)
		return 0;

	if (!need_resched() || !latency_warn_ms)
		return 0;

	if (system_state == SYSTEM_BOOTING)
		return 0;

	if (!rq->last_seen_need_resched_ns) {
		rq->last_seen_need_resched_ns = now;
		rq->ticks_without_resched = 0;
		return 0;
	}

	rq->ticks_without_resched++;
	resched_latency = now - rq->last_seen_need_resched_ns;
	if (resched_latency <= latency_warn_ms * NSEC_PER_MSEC)
		return 0;

	warned_once = true;

	return resched_latency;
}

static int __init setup_resched_latency_warn_ms(char *str)
{
	long val;

	if ((kstrtol(str, 0, &val))) {
		pr_warn("Unable to set resched_latency_warn_ms\n");
		return 1;
	}

	sysctl_resched_latency_warn_ms = val;
	return 1;
}
__setup("resched_latency_warn_ms=", setup_resched_latency_warn_ms);
#else
static inline u64 cpu_resched_latency(struct rq *rq) { return 0; }
#endif /* CONFIG_SCHED_DEBUG */

5237 5238 5239 5240 5241 5242 5243 5244
/*
 * 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
5245
	struct task_struct *curr = rq->curr;
5246
	struct rq_flags rf;
5247
	unsigned long thermal_pressure;
5248
	u64 resched_latency;
5249

5250
	arch_scale_freq_tick();
5251
	sched_clock_tick();
Ingo Molnar's avatar
Ingo Molnar committed
5252

5253 5254
	rq_lock(rq, &rf);

5255
	update_rq_clock(rq);
5256
	thermal_pressure = arch_scale_thermal_pressure(cpu_of(rq));
5257
	update_thermal_load_avg(rq_clock_thermal(rq), rq, thermal_pressure);
Peter Zijlstra's avatar
Peter Zijlstra committed
5258
	curr->sched_class->task_tick(rq, curr, 0);
5259 5260
	if (sched_feat(LATENCY_WARN))
		resched_latency = cpu_resched_latency(rq);
5261
	calc_global_load_tick(rq);
5262
	sched_core_tick(rq);
5263 5264

	rq_unlock(rq, &rf);
5265

5266 5267 5268
	if (sched_feat(LATENCY_WARN) && resched_latency)
		resched_latency_warn(cpu, resched_latency);

5269
	perf_event_task_tick();
5270

5271
#ifdef CONFIG_SMP
5272
	rq->idle_balance = idle_cpu(cpu);
5273
	trigger_load_balance(rq);
5274
#endif
Linus Torvalds's avatar
Linus Torvalds committed
5275 5276
}

5277
#ifdef CONFIG_NO_HZ_FULL
5278 5279 5280

struct tick_work {
	int			cpu;
5281
	atomic_t		state;
5282 5283
	struct delayed_work	work;
};
5284 5285 5286 5287 5288 5289 5290 5291 5292 5293 5294 5295 5296 5297 5298 5299 5300 5301 5302 5303 5304 5305 5306 5307 5308 5309 5310
/* 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.
 */
5311 5312 5313 5314 5315 5316 5317 5318 5319

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);
5320
	struct task_struct *curr;
5321
	struct rq_flags rf;
5322
	u64 delta;
5323
	int os;
5324 5325 5326 5327 5328 5329 5330 5331

	/*
	 * 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.
	 */
5332
	if (!tick_nohz_tick_stopped_cpu(cpu))
5333
		goto out_requeue;
5334

5335 5336
	rq_lock_irq(rq, &rf);
	curr = rq->curr;
5337
	if (cpu_is_offline(cpu))
5338
		goto out_unlock;
5339

5340 5341
	update_rq_clock(rq);

5342 5343 5344 5345 5346 5347 5348 5349
	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);
	}
5350 5351
	curr->sched_class->task_tick(rq, curr, 0);

5352
	calc_load_nohz_remote(rq);
5353 5354 5355
out_unlock:
	rq_unlock_irq(rq, &rf);
out_requeue:
5356

5357 5358 5359
	/*
	 * Run the remote tick once per second (1Hz). This arbitrary
	 * frequency is large enough to avoid overload but short enough
5360 5361
	 * to keep scheduler internal stats reasonably up to date.  But
	 * first update state to reflect hotplug activity if required.
5362
	 */
5363 5364 5365 5366
	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);
5367 5368 5369 5370
}

static void sched_tick_start(int cpu)
{
5371
	int os;
5372 5373
	struct tick_work *twork;

5374
	if (housekeeping_cpu(cpu, HK_TYPE_TICK))
5375 5376 5377 5378 5379
		return;

	WARN_ON_ONCE(!tick_work_cpu);

	twork = per_cpu_ptr(tick_work_cpu, cpu);
5380 5381 5382 5383 5384 5385 5386
	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);
	}
5387 5388 5389 5390 5391 5392
}

#ifdef CONFIG_HOTPLUG_CPU
static void sched_tick_stop(int cpu)
{
	struct tick_work *twork;
5393
	int os;
5394

5395
	if (housekeeping_cpu(cpu, HK_TYPE_TICK))
5396 5397 5398 5399 5400
		return;

	WARN_ON_ONCE(!tick_work_cpu);

	twork = per_cpu_ptr(tick_work_cpu, cpu);
5401 5402 5403 5404
	/* 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. */
5405 5406 5407 5408 5409 5410 5411 5412 5413 5414 5415 5416 5417
}
#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) { }
5418
#endif
Linus Torvalds's avatar
Linus Torvalds committed
5419

5420
#if defined(CONFIG_PREEMPTION) && (defined(CONFIG_DEBUG_PREEMPT) || \
5421
				defined(CONFIG_TRACE_PREEMPT_TOGGLE))
5422 5423 5424 5425 5426 5427 5428 5429 5430 5431 5432 5433 5434 5435
/*
 * 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);
	}
}
5436

5437
void preempt_count_add(int val)
Linus Torvalds's avatar
Linus Torvalds committed
5438
{
5439
#ifdef CONFIG_DEBUG_PREEMPT
Linus Torvalds's avatar
Linus Torvalds committed
5440 5441 5442
	/*
	 * Underflow?
	 */
5443 5444
	if (DEBUG_LOCKS_WARN_ON((preempt_count() < 0)))
		return;
5445
#endif
5446
	__preempt_count_add(val);
5447
#ifdef CONFIG_DEBUG_PREEMPT
Linus Torvalds's avatar
Linus Torvalds committed
5448 5449 5450
	/*
	 * Spinlock count overflowing soon?
	 */
5451 5452
	DEBUG_LOCKS_WARN_ON((preempt_count() & PREEMPT_MASK) >=
				PREEMPT_MASK - 10);
5453
#endif
5454
	preempt_latency_start(val);
Linus Torvalds's avatar
Linus Torvalds committed
5455
}
5456
EXPORT_SYMBOL(preempt_count_add);
5457
NOKPROBE_SYMBOL(preempt_count_add);
Linus Torvalds's avatar
Linus Torvalds committed
5458

5459 5460 5461 5462 5463 5464 5465 5466 5467 5468
/*
 * 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());
}

5469
void preempt_count_sub(int val)
Linus Torvalds's avatar
Linus Torvalds committed
5470
{
5471
#ifdef CONFIG_DEBUG_PREEMPT
Linus Torvalds's avatar
Linus Torvalds committed
5472 5473 5474
	/*
	 * Underflow?
	 */
5475
	if (DEBUG_LOCKS_WARN_ON(val > preempt_count()))
5476
		return;
Linus Torvalds's avatar
Linus Torvalds committed
5477 5478 5479
	/*
	 * Is the spinlock portion underflowing?
	 */
5480 5481 5482
	if (DEBUG_LOCKS_WARN_ON((val < PREEMPT_MASK) &&
			!(preempt_count() & PREEMPT_MASK)))
		return;
5483
#endif
5484

5485
	preempt_latency_stop(val);
5486
	__preempt_count_sub(val);
Linus Torvalds's avatar
Linus Torvalds committed
5487
}
5488
EXPORT_SYMBOL(preempt_count_sub);
5489
NOKPROBE_SYMBOL(preempt_count_sub);
Linus Torvalds's avatar
Linus Torvalds committed
5490

5491 5492 5493
#else
static inline void preempt_latency_start(int val) { }
static inline void preempt_latency_stop(int val) { }
Linus Torvalds's avatar
Linus Torvalds committed
5494 5495
#endif

5496 5497 5498 5499 5500 5501 5502 5503 5504
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
5505
/*
Ingo Molnar's avatar
Ingo Molnar committed
5506
 * Print scheduling while atomic bug:
Linus Torvalds's avatar
Linus Torvalds committed
5507
 */
Ingo Molnar's avatar
Ingo Molnar committed
5508
static noinline void __schedule_bug(struct task_struct *prev)
Linus Torvalds's avatar
Linus Torvalds committed
5509
{
5510 5511 5512
	/* Save this before calling printk(), since that will clobber it */
	unsigned long preempt_disable_ip = get_preempt_disable_ip(current);

5513 5514 5515
	if (oops_in_progress)
		return;

5516 5517
	printk(KERN_ERR "BUG: scheduling while atomic: %s/%d/0x%08x\n",
		prev->comm, prev->pid, preempt_count());
5518

Ingo Molnar's avatar
Ingo Molnar committed
5519
	debug_show_held_locks(prev);
5520
	print_modules();
Ingo Molnar's avatar
Ingo Molnar committed
5521 5522
	if (irqs_disabled())
		print_irqtrace_events(prev);
5523 5524
	if (IS_ENABLED(CONFIG_DEBUG_PREEMPT)
	    && in_atomic_preempt_off()) {
5525
		pr_err("Preemption disabled at:");
5526
		print_ip_sym(KERN_ERR, preempt_disable_ip);
5527
	}
5528 5529 5530
	if (panic_on_warn)
		panic("scheduling while atomic\n");

5531
	dump_stack();
5532
	add_taint(TAINT_WARN, LOCKDEP_STILL_OK);
Ingo Molnar's avatar
Ingo Molnar committed
5533
}
Linus Torvalds's avatar
Linus Torvalds committed
5534

Ingo Molnar's avatar
Ingo Molnar committed
5535 5536 5537
/*
 * Various schedule()-time debugging checks and statistics:
 */
5538
static inline void schedule_debug(struct task_struct *prev, bool preempt)
Ingo Molnar's avatar
Ingo Molnar committed
5539
{
5540
#ifdef CONFIG_SCHED_STACK_END_CHECK
5541 5542
	if (task_stack_end_corrupted(prev))
		panic("corrupted stack end detected inside scheduler\n");
5543 5544 5545

	if (task_scs_end_corrupted(prev))
		panic("corrupted shadow stack detected inside scheduler\n");
5546
#endif
5547

5548
#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
5549
	if (!preempt && READ_ONCE(prev->__state) && prev->non_block_count) {
5550 5551 5552 5553 5554 5555 5556
		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

5557
	if (unlikely(in_atomic_preempt_off())) {
Ingo Molnar's avatar
Ingo Molnar committed
5558
		__schedule_bug(prev);
5559 5560
		preempt_count_set(PREEMPT_DISABLED);
	}
5561
	rcu_sleep_check();
5562
	SCHED_WARN_ON(ct_state() == CONTEXT_USER);
Ingo Molnar's avatar
Ingo Molnar committed
5563

Linus Torvalds's avatar
Linus Torvalds committed
5564 5565
	profile_hit(SCHED_PROFILING, __builtin_return_address(0));

5566
	schedstat_inc(this_rq()->sched_count);
Ingo Molnar's avatar
Ingo Molnar committed
5567 5568
}

5569 5570 5571 5572 5573 5574 5575 5576 5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587 5588 5589 5590
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
5591 5592 5593 5594
/*
 * Pick up the highest-prio task:
 */
static inline struct task_struct *
5595
__pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
Ingo Molnar's avatar
Ingo Molnar committed
5596
{
5597
	const struct sched_class *class;
Ingo Molnar's avatar
Ingo Molnar committed
5598
	struct task_struct *p;
Linus Torvalds's avatar
Linus Torvalds committed
5599 5600

	/*
5601 5602
	 * 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
5603
	 * higher scheduling class, because otherwise those lose the
5604
	 * opportunity to pull in more work from other CPUs.
Linus Torvalds's avatar
Linus Torvalds committed
5605
	 */
5606
	if (likely(prev->sched_class <= &fair_sched_class &&
5607 5608
		   rq->nr_running == rq->cfs.h_nr_running)) {

5609
		p = pick_next_task_fair(rq, prev, rf);
5610
		if (unlikely(p == RETRY_TASK))
5611
			goto restart;
5612

5613
		/* Assume the next prioritized class is idle_sched_class */
5614
		if (!p) {
5615
			put_prev_task(rq, prev);
5616
			p = pick_next_task_idle(rq);
5617
		}
5618 5619

		return p;
Linus Torvalds's avatar
Linus Torvalds committed
5620 5621
	}

5622
restart:
5623
	put_prev_task_balance(rq, prev, rf);
5624

5625
	for_each_class(class) {
5626
		p = class->pick_next_task(rq);
5627
		if (p)
Ingo Molnar's avatar
Ingo Molnar committed
5628 5629
			return p;
	}
5630

5631
	BUG(); /* The idle class should always have a runnable task. */
Ingo Molnar's avatar
Ingo Molnar committed
5632
}
Linus Torvalds's avatar
Linus Torvalds committed
5633

Peter Zijlstra's avatar
Peter Zijlstra committed
5634
#ifdef CONFIG_SCHED_CORE
5635 5636 5637 5638 5639 5640 5641 5642 5643 5644 5645 5646 5647 5648 5649 5650 5651 5652
static inline bool is_task_rq_idle(struct task_struct *t)
{
	return (task_rq(t)->idle == t);
}

static inline bool cookie_equals(struct task_struct *a, unsigned long cookie)
{
	return is_task_rq_idle(a) || (a->core_cookie == cookie);
}

static inline bool cookie_match(struct task_struct *a, struct task_struct *b)
{
	if (is_task_rq_idle(a) || is_task_rq_idle(b))
		return true;

	return a->core_cookie == b->core_cookie;
}

5653
static inline struct task_struct *pick_task(struct rq *rq)
5654
{
5655 5656
	const struct sched_class *class;
	struct task_struct *p;
5657

5658 5659 5660 5661
	for_each_class(class) {
		p = class->pick_task(rq);
		if (p)
			return p;
5662 5663
	}

5664
	BUG(); /* The idle class should always have a runnable task. */
5665 5666
}

5667 5668
extern void task_vruntime_update(struct rq *rq, struct task_struct *p, bool in_fi);

5669 5670 5671
static struct task_struct *
pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
{
5672
	struct task_struct *next, *p, *max = NULL;
5673
	const struct cpumask *smt_mask;
5674
	bool fi_before = false;
5675
	bool core_clock_updated = (rq == rq->core);
5676 5677 5678
	unsigned long cookie;
	int i, cpu, occ = 0;
	struct rq *rq_i;
5679 5680 5681 5682 5683 5684 5685 5686 5687 5688 5689 5690 5691 5692 5693 5694 5695 5696 5697 5698 5699 5700 5701 5702 5703 5704 5705 5706 5707 5708 5709 5710 5711 5712 5713 5714 5715 5716 5717 5718 5719 5720 5721 5722 5723
	bool need_sync;

	if (!sched_core_enabled(rq))
		return __pick_next_task(rq, prev, rf);

	cpu = cpu_of(rq);

	/* Stopper task is switching into idle, no need core-wide selection. */
	if (cpu_is_offline(cpu)) {
		/*
		 * Reset core_pick so that we don't enter the fastpath when
		 * coming online. core_pick would already be migrated to
		 * another cpu during offline.
		 */
		rq->core_pick = NULL;
		return __pick_next_task(rq, prev, rf);
	}

	/*
	 * If there were no {en,de}queues since we picked (IOW, the task
	 * pointers are all still valid), and we haven't scheduled the last
	 * pick yet, do so now.
	 *
	 * rq->core_pick can be NULL if no selection was made for a CPU because
	 * it was either offline or went offline during a sibling's core-wide
	 * selection. In this case, do a core-wide selection.
	 */
	if (rq->core->core_pick_seq == rq->core->core_task_seq &&
	    rq->core->core_pick_seq != rq->core_sched_seq &&
	    rq->core_pick) {
		WRITE_ONCE(rq->core_sched_seq, rq->core->core_pick_seq);

		next = rq->core_pick;
		if (next != prev) {
			put_prev_task(rq, prev);
			set_next_task(rq, next);
		}

		rq->core_pick = NULL;
		return next;
	}

	put_prev_task_balance(rq, prev, rf);

	smt_mask = cpu_smt_mask(cpu);
5724 5725 5726 5727
	need_sync = !!rq->core->core_cookie;

	/* reset state */
	rq->core->core_cookie = 0UL;
5728 5729 5730 5731 5732 5733 5734 5735 5736 5737
	if (rq->core->core_forceidle_count) {
		if (!core_clock_updated) {
			update_rq_clock(rq->core);
			core_clock_updated = true;
		}
		sched_core_account_forceidle(rq);
		/* reset after accounting force idle */
		rq->core->core_forceidle_start = 0;
		rq->core->core_forceidle_count = 0;
		rq->core->core_forceidle_occupation = 0;
5738 5739 5740
		need_sync = true;
		fi_before = true;
	}
5741 5742 5743 5744 5745 5746 5747 5748 5749 5750 5751 5752 5753

	/*
	 * core->core_task_seq, core->core_pick_seq, rq->core_sched_seq
	 *
	 * @task_seq guards the task state ({en,de}queues)
	 * @pick_seq is the @task_seq we did a selection on
	 * @sched_seq is the @pick_seq we scheduled
	 *
	 * However, preemptions can cause multiple picks on the same task set.
	 * 'Fix' this by also increasing @task_seq for every pick.
	 */
	rq->core->core_task_seq++;

5754 5755 5756 5757 5758
	/*
	 * Optimize for common case where this CPU has no cookies
	 * and there are no cookied tasks running on siblings.
	 */
	if (!need_sync) {
5759
		next = pick_task(rq);
5760 5761
		if (!next->core_cookie) {
			rq->core_pick = NULL;
5762 5763 5764 5765 5766 5767
			/*
			 * For robustness, update the min_vruntime_fi for
			 * unconstrained picks as well.
			 */
			WARN_ON_ONCE(fi_before);
			task_vruntime_update(rq, next, false);
5768 5769
			goto done;
		}
5770
	}
5771

5772 5773 5774 5775 5776 5777 5778 5779
	/*
	 * For each thread: do the regular task pick and find the max prio task
	 * amongst them.
	 *
	 * Tie-break prio towards the current CPU
	 */
	for_each_cpu_wrap(i, smt_mask, cpu) {
		rq_i = cpu_rq(i);
5780

5781 5782 5783 5784 5785 5786
		/*
		 * Current cpu always has its clock updated on entrance to
		 * pick_next_task(). If the current cpu is not the core,
		 * the core may also have been updated above.
		 */
		if (i != cpu && (rq_i != rq->core || !core_clock_updated))
5787
			update_rq_clock(rq_i);
5788 5789 5790 5791

		p = rq_i->core_pick = pick_task(rq_i);
		if (!max || prio_less(max, p, fi_before))
			max = p;
5792 5793
	}

5794 5795
	cookie = rq->core->core_cookie = max->core_cookie;

5796
	/*
5797 5798
	 * For each thread: try and find a runnable task that matches @max or
	 * force idle.
5799
	 */
5800 5801 5802
	for_each_cpu(i, smt_mask) {
		rq_i = cpu_rq(i);
		p = rq_i->core_pick;
5803

5804 5805 5806 5807
		if (!cookie_equals(p, cookie)) {
			p = NULL;
			if (cookie)
				p = sched_core_find(rq_i, cookie);
5808
			if (!p)
5809 5810
				p = idle_sched_class.pick_task(rq_i);
		}
5811

5812
		rq_i->core_pick = p;
5813

5814 5815
		if (p == rq_i->idle) {
			if (rq_i->nr_running) {
5816
				rq->core->core_forceidle_count++;
5817 5818 5819
				if (!fi_before)
					rq->core->core_forceidle_seq++;
			}
5820 5821
		} else {
			occ++;
5822 5823 5824
		}
	}

5825 5826 5827 5828 5829 5830
	if (schedstat_enabled() && rq->core->core_forceidle_count) {
		if (cookie)
			rq->core->core_forceidle_start = rq_clock(rq->core);
		rq->core->core_forceidle_occupation = occ;
	}

5831 5832 5833 5834 5835 5836 5837 5838 5839 5840 5841 5842 5843 5844 5845 5846
	rq->core->core_pick_seq = rq->core->core_task_seq;
	next = rq->core_pick;
	rq->core_sched_seq = rq->core->core_pick_seq;

	/* Something should have been selected for current CPU */
	WARN_ON_ONCE(!next);

	/*
	 * Reschedule siblings
	 *
	 * NOTE: L1TF -- at this point we're no longer running the old task and
	 * sending an IPI (below) ensures the sibling will no longer be running
	 * their task. This ensures there is no inter-sibling overlap between
	 * non-matching user state.
	 */
	for_each_cpu(i, smt_mask) {
5847
		rq_i = cpu_rq(i);
5848 5849 5850 5851 5852 5853 5854 5855 5856 5857 5858

		/*
		 * An online sibling might have gone offline before a task
		 * could be picked for it, or it might be offline but later
		 * happen to come online, but its too late and nothing was
		 * picked for it.  That's Ok - it will pick tasks for itself,
		 * so ignore it.
		 */
		if (!rq_i->core_pick)
			continue;

5859 5860 5861 5862 5863 5864 5865 5866
		/*
		 * Update for new !FI->FI transitions, or if continuing to be in !FI:
		 * fi_before     fi      update?
		 *  0            0       1
		 *  0            1       1
		 *  1            0       1
		 *  1            1       0
		 */
5867 5868
		if (!(fi_before && rq->core->core_forceidle_count))
			task_vruntime_update(rq_i, rq_i->core_pick, !!rq->core->core_forceidle_count);
5869

5870 5871
		rq_i->core_pick->core_occupation = occ;

5872 5873 5874 5875 5876 5877 5878 5879 5880 5881 5882 5883 5884 5885 5886 5887 5888 5889 5890 5891
		if (i == cpu) {
			rq_i->core_pick = NULL;
			continue;
		}

		/* Did we break L1TF mitigation requirements? */
		WARN_ON_ONCE(!cookie_match(next, rq_i->core_pick));

		if (rq_i->curr == rq_i->core_pick) {
			rq_i->core_pick = NULL;
			continue;
		}

		resched_curr(rq_i);
	}

done:
	set_next_task(rq, next);
	return next;
}
Peter Zijlstra's avatar
Peter Zijlstra committed
5892

5893 5894 5895 5896 5897 5898 5899 5900 5901 5902 5903 5904 5905 5906 5907 5908 5909 5910 5911 5912 5913 5914 5915 5916 5917 5918 5919 5920 5921 5922 5923 5924 5925 5926 5927 5928 5929 5930 5931 5932 5933 5934 5935 5936 5937 5938 5939 5940 5941 5942 5943 5944 5945 5946 5947 5948 5949 5950 5951 5952 5953 5954 5955 5956 5957 5958 5959 5960 5961 5962 5963 5964 5965 5966 5967 5968 5969 5970 5971 5972 5973 5974 5975 5976 5977 5978 5979 5980 5981 5982 5983 5984 5985 5986 5987 5988 5989 5990 5991 5992 5993 5994 5995 5996 5997
static bool try_steal_cookie(int this, int that)
{
	struct rq *dst = cpu_rq(this), *src = cpu_rq(that);
	struct task_struct *p;
	unsigned long cookie;
	bool success = false;

	local_irq_disable();
	double_rq_lock(dst, src);

	cookie = dst->core->core_cookie;
	if (!cookie)
		goto unlock;

	if (dst->curr != dst->idle)
		goto unlock;

	p = sched_core_find(src, cookie);
	if (p == src->idle)
		goto unlock;

	do {
		if (p == src->core_pick || p == src->curr)
			goto next;

		if (!cpumask_test_cpu(this, &p->cpus_mask))
			goto next;

		if (p->core_occupation > dst->idle->core_occupation)
			goto next;

		deactivate_task(src, p, 0);
		set_task_cpu(p, this);
		activate_task(dst, p, 0);

		resched_curr(dst);

		success = true;
		break;

next:
		p = sched_core_next(p, cookie);
	} while (p);

unlock:
	double_rq_unlock(dst, src);
	local_irq_enable();

	return success;
}

static bool steal_cookie_task(int cpu, struct sched_domain *sd)
{
	int i;

	for_each_cpu_wrap(i, sched_domain_span(sd), cpu) {
		if (i == cpu)
			continue;

		if (need_resched())
			break;

		if (try_steal_cookie(cpu, i))
			return true;
	}

	return false;
}

static void sched_core_balance(struct rq *rq)
{
	struct sched_domain *sd;
	int cpu = cpu_of(rq);

	preempt_disable();
	rcu_read_lock();
	raw_spin_rq_unlock_irq(rq);
	for_each_domain(cpu, sd) {
		if (need_resched())
			break;

		if (steal_cookie_task(cpu, sd))
			break;
	}
	raw_spin_rq_lock_irq(rq);
	rcu_read_unlock();
	preempt_enable();
}

static DEFINE_PER_CPU(struct callback_head, core_balance_head);

void queue_core_balance(struct rq *rq)
{
	if (!sched_core_enabled(rq))
		return;

	if (!rq->core->core_cookie)
		return;

	if (!rq->nr_running) /* not forced idle */
		return;

	queue_balance_callback(rq, &per_cpu(core_balance_head, rq->cpu), sched_core_balance);
}

5998
static void sched_core_cpu_starting(unsigned int cpu)
Peter Zijlstra's avatar
Peter Zijlstra committed
5999 6000
{
	const struct cpumask *smt_mask = cpu_smt_mask(cpu);
6001 6002 6003
	struct rq *rq = cpu_rq(cpu), *core_rq = NULL;
	unsigned long flags;
	int t;
Peter Zijlstra's avatar
Peter Zijlstra committed
6004

6005
	sched_core_lock(cpu, &flags);
Peter Zijlstra's avatar
Peter Zijlstra committed
6006

6007 6008 6009 6010 6011 6012 6013 6014 6015 6016 6017 6018 6019 6020
	WARN_ON_ONCE(rq->core != rq);

	/* if we're the first, we'll be our own leader */
	if (cpumask_weight(smt_mask) == 1)
		goto unlock;

	/* find the leader */
	for_each_cpu(t, smt_mask) {
		if (t == cpu)
			continue;
		rq = cpu_rq(t);
		if (rq->core == rq) {
			core_rq = rq;
			break;
Peter Zijlstra's avatar
Peter Zijlstra committed
6021
		}
6022
	}
Peter Zijlstra's avatar
Peter Zijlstra committed
6023

6024 6025
	if (WARN_ON_ONCE(!core_rq)) /* whoopsie */
		goto unlock;
Peter Zijlstra's avatar
Peter Zijlstra committed
6026

6027 6028 6029
	/* install and validate core_rq */
	for_each_cpu(t, smt_mask) {
		rq = cpu_rq(t);
Peter Zijlstra's avatar
Peter Zijlstra committed
6030

6031
		if (t == cpu)
Peter Zijlstra's avatar
Peter Zijlstra committed
6032
			rq->core = core_rq;
6033 6034

		WARN_ON_ONCE(rq->core != core_rq);
Peter Zijlstra's avatar
Peter Zijlstra committed
6035
	}
6036 6037 6038

unlock:
	sched_core_unlock(cpu, &flags);
Peter Zijlstra's avatar
Peter Zijlstra committed
6039
}
6040 6041 6042 6043 6044 6045 6046 6047 6048 6049 6050 6051 6052 6053 6054 6055 6056 6057 6058 6059 6060 6061 6062 6063 6064 6065 6066 6067 6068 6069 6070 6071

static void sched_core_cpu_deactivate(unsigned int cpu)
{
	const struct cpumask *smt_mask = cpu_smt_mask(cpu);
	struct rq *rq = cpu_rq(cpu), *core_rq = NULL;
	unsigned long flags;
	int t;

	sched_core_lock(cpu, &flags);

	/* if we're the last man standing, nothing to do */
	if (cpumask_weight(smt_mask) == 1) {
		WARN_ON_ONCE(rq->core != rq);
		goto unlock;
	}

	/* if we're not the leader, nothing to do */
	if (rq->core != rq)
		goto unlock;

	/* find a new leader */
	for_each_cpu(t, smt_mask) {
		if (t == cpu)
			continue;
		core_rq = cpu_rq(t);
		break;
	}

	if (WARN_ON_ONCE(!core_rq)) /* impossible */
		goto unlock;

	/* copy the shared state to the new leader */
6072 6073 6074 6075 6076 6077 6078 6079 6080 6081 6082 6083 6084
	core_rq->core_task_seq             = rq->core_task_seq;
	core_rq->core_pick_seq             = rq->core_pick_seq;
	core_rq->core_cookie               = rq->core_cookie;
	core_rq->core_forceidle_count      = rq->core_forceidle_count;
	core_rq->core_forceidle_seq        = rq->core_forceidle_seq;
	core_rq->core_forceidle_occupation = rq->core_forceidle_occupation;

	/*
	 * Accounting edge for forced idle is handled in pick_next_task().
	 * Don't need another one here, since the hotplug thread shouldn't
	 * have a cookie.
	 */
	core_rq->core_forceidle_start = 0;
6085 6086 6087 6088 6089 6090 6091 6092 6093 6094 6095 6096 6097 6098 6099 6100 6101 6102 6103

	/* install new leader */
	for_each_cpu(t, smt_mask) {
		rq = cpu_rq(t);
		rq->core = core_rq;
	}

unlock:
	sched_core_unlock(cpu, &flags);
}

static inline void sched_core_cpu_dying(unsigned int cpu)
{
	struct rq *rq = cpu_rq(cpu);

	if (rq->core != rq)
		rq->core = rq;
}

Peter Zijlstra's avatar
Peter Zijlstra committed
6104 6105 6106
#else /* !CONFIG_SCHED_CORE */

static inline void sched_core_cpu_starting(unsigned int cpu) {}
6107 6108
static inline void sched_core_cpu_deactivate(unsigned int cpu) {}
static inline void sched_core_cpu_dying(unsigned int cpu) {}
Peter Zijlstra's avatar
Peter Zijlstra committed
6109

6110 6111 6112 6113 6114 6115
static struct task_struct *
pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
{
	return __pick_next_task(rq, prev, rf);
}

Peter Zijlstra's avatar
Peter Zijlstra committed
6116 6117
#endif /* CONFIG_SCHED_CORE */

6118 6119 6120 6121 6122 6123 6124 6125 6126 6127
/*
 * Constants for the sched_mode argument of __schedule().
 *
 * The mode argument allows RT enabled kernels to differentiate a
 * preemption from blocking on an 'sleeping' spin/rwlock. Note that
 * SM_MASK_PREEMPT for !RT has all bits set, which allows the compiler to
 * optimize the AND operation out and just check for zero.
 */
#define SM_NONE			0x0
#define SM_PREEMPT		0x1
6128 6129 6130 6131 6132 6133 6134
#define SM_RTLOCK_WAIT		0x2

#ifndef CONFIG_PREEMPT_RT
# define SM_MASK_PREEMPT	(~0U)
#else
# define SM_MASK_PREEMPT	SM_PREEMPT
#endif
6135

Ingo Molnar's avatar
Ingo Molnar committed
6136
/*
6137
 * __schedule() is the main scheduler function.
6138 6139 6140 6141 6142 6143 6144 6145 6146 6147 6148 6149 6150 6151 6152 6153 6154 6155
 *
 * 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:
 *
6156
 *       - If the kernel is preemptible (CONFIG_PREEMPTION=y):
6157 6158 6159 6160 6161 6162 6163 6164
 *
 *         - 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
 *
6165
 *       - If the kernel is not preemptible (CONFIG_PREEMPTION is not set)
6166 6167 6168 6169 6170 6171
 *         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
6172
 *
6173
 * WARNING: must be called with preemption disabled!
Ingo Molnar's avatar
Ingo Molnar committed
6174
 */
6175
static void __sched notrace __schedule(unsigned int sched_mode)
Ingo Molnar's avatar
Ingo Molnar committed
6176 6177
{
	struct task_struct *prev, *next;
6178
	unsigned long *switch_count;
6179
	unsigned long prev_state;
6180
	struct rq_flags rf;
Ingo Molnar's avatar
Ingo Molnar committed
6181
	struct rq *rq;
6182
	int cpu;
Ingo Molnar's avatar
Ingo Molnar committed
6183 6184 6185 6186 6187

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

6188
	schedule_debug(prev, !!sched_mode);
Linus Torvalds's avatar
Linus Torvalds committed
6189

6190
	if (sched_feat(HRTICK) || sched_feat(HRTICK_DL))
Mike Galbraith's avatar
Mike Galbraith committed
6191
		hrtick_clear(rq);
6192

6193
	local_irq_disable();
6194
	rcu_note_context_switch(!!sched_mode);
6195

6196 6197 6198
	/*
	 * Make sure that signal_pending_state()->signal_pending() below
	 * can't be reordered with __set_current_state(TASK_INTERRUPTIBLE)
6199 6200 6201 6202 6203 6204 6205 6206
	 * done by the caller to avoid the race with signal_wake_up():
	 *
	 * __set_current_state(@state)		signal_wake_up()
	 * schedule()				  set_tsk_thread_flag(p, TIF_SIGPENDING)
	 *					  wake_up_state(p, state)
	 *   LOCK rq->lock			    LOCK p->pi_state
	 *   smp_mb__after_spinlock()		    smp_mb__after_spinlock()
	 *     if (signal_pending_state())	    if (p->state & @state)
6207
	 *
6208
	 * Also, the membarrier system call requires a full memory barrier
6209
	 * after coming from user-space, before storing to rq->curr.
6210
	 */
6211
	rq_lock(rq, &rf);
6212
	smp_mb__after_spinlock();
Linus Torvalds's avatar
Linus Torvalds committed
6213

6214 6215
	/* Promote REQ to ACT */
	rq->clock_update_flags <<= 1;
6216
	update_rq_clock(rq);
6217

6218
	switch_count = &prev->nivcsw;
6219

6220
	/*
6221 6222 6223 6224 6225
	 * We must load prev->state once (task_struct::state is volatile), such
	 * that:
	 *
	 *  - we form a control dependency vs deactivate_task() below.
	 *  - ptrace_{,un}freeze_traced() can change ->state underneath us.
6226
	 */
6227
	prev_state = READ_ONCE(prev->__state);
6228
	if (!(sched_mode & SM_MASK_PREEMPT) && prev_state) {
6229
		if (signal_pending_state(prev_state, prev)) {
6230
			WRITE_ONCE(prev->__state, TASK_RUNNING);
Tejun Heo's avatar
Tejun Heo committed
6231
		} else {
6232 6233 6234 6235 6236 6237 6238 6239 6240 6241
			prev->sched_contributes_to_load =
				(prev_state & TASK_UNINTERRUPTIBLE) &&
				!(prev_state & TASK_NOLOAD) &&
				!(prev->flags & PF_FROZEN);

			if (prev->sched_contributes_to_load)
				rq->nr_uninterruptible++;

			/*
			 * __schedule()			ttwu()
6242 6243 6244 6245 6246 6247
			 *   prev_state = prev->state;    if (p->on_rq && ...)
			 *   if (prev_state)		    goto out;
			 *     p->on_rq = 0;		  smp_acquire__after_ctrl_dep();
			 *				  p->state = TASK_WAKING
			 *
			 * Where __schedule() and ttwu() have matching control dependencies.
6248 6249 6250
			 *
			 * After this, schedule() must not care about p->state any more.
			 */
6251
			deactivate_task(rq, prev, DEQUEUE_SLEEP | DEQUEUE_NOCLOCK);
6252

6253 6254 6255 6256
			if (prev->in_iowait) {
				atomic_inc(&rq->nr_iowait);
				delayacct_blkio_start();
			}
Tejun Heo's avatar
Tejun Heo committed
6257
		}
Ingo Molnar's avatar
Ingo Molnar committed
6258
		switch_count = &prev->nvcsw;
Linus Torvalds's avatar
Linus Torvalds committed
6259 6260
	}

6261
	next = pick_next_task(rq, prev, &rf);
6262
	clear_tsk_need_resched(prev);
6263
	clear_preempt_need_resched();
6264 6265 6266
#ifdef CONFIG_SCHED_DEBUG
	rq->last_seen_need_resched_ns = 0;
#endif
Linus Torvalds's avatar
Linus Torvalds committed
6267 6268 6269

	if (likely(prev != next)) {
		rq->nr_switches++;
6270 6271 6272 6273 6274
		/*
		 * 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);
6275 6276 6277
		/*
		 * The membarrier system call requires each architecture
		 * to have a full memory barrier after updating
6278 6279 6280 6281 6282 6283 6284 6285 6286 6287
		 * 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),
6288
		 */
Linus Torvalds's avatar
Linus Torvalds committed
6289 6290
		++*switch_count;

6291
		migrate_disable_switch(rq, prev);
6292 6293
		psi_sched_switch(prev, next, !task_on_rq_queued(prev));

6294
		trace_sched_switch(sched_mode & SM_MASK_PREEMPT, prev, next);
6295 6296 6297

		/* Also unlocks the rq: */
		rq = context_switch(rq, prev, next, &rf);
6298
	} else {
6299
		rq->clock_update_flags &= ~(RQCF_ACT_SKIP|RQCF_REQ_SKIP);
Linus Torvalds's avatar
Linus Torvalds committed
6300

6301 6302
		rq_unpin_lock(rq, &rf);
		__balance_callbacks(rq);
6303
		raw_spin_rq_unlock_irq(rq);
6304
	}
Linus Torvalds's avatar
Linus Torvalds committed
6305
}
6306

6307 6308
void __noreturn do_task_dead(void)
{
6309
	/* Causes final put_task_struct in finish_task_switch(): */
6310
	set_special_state(TASK_DEAD);
6311 6312 6313 6314

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

6315
	__schedule(SM_NONE);
6316
	BUG();
6317 6318

	/* Avoid "noreturn function does return" - but don't continue if BUG() is a NOP: */
6319
	for (;;)
6320
		cpu_relax();
6321 6322
}

6323 6324
static inline void sched_submit_work(struct task_struct *tsk)
{
6325 6326
	unsigned int task_flags;

6327
	if (task_is_running(tsk))
6328
		return;
6329

6330
	task_flags = tsk->flags;
6331
	/*
6332 6333
	 * If a worker goes to sleep, notify and ask workqueue whether it
	 * wants to wake up a task to maintain concurrency.
6334
	 */
6335 6336
	if (task_flags & (PF_WQ_WORKER | PF_IO_WORKER)) {
		if (task_flags & PF_WQ_WORKER)
6337 6338 6339
			wq_worker_sleeping(tsk);
		else
			io_wq_worker_sleeping(tsk);
6340 6341
	}

6342 6343 6344
	if (tsk_is_pi_blocked(tsk))
		return;

6345 6346 6347 6348 6349
	/*
	 * 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))
6350
		blk_flush_plug(tsk->plug, true);
6351 6352
}

6353 6354
static void sched_update_worker(struct task_struct *tsk)
{
6355 6356 6357 6358 6359 6360
	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);
	}
6361 6362
}

6363
asmlinkage __visible void __sched schedule(void)
6364
{
6365 6366 6367
	struct task_struct *tsk = current;

	sched_submit_work(tsk);
6368
	do {
6369
		preempt_disable();
6370
		__schedule(SM_NONE);
6371
		sched_preempt_enable_no_resched();
6372
	} while (need_resched());
6373
	sched_update_worker(tsk);
6374
}
Linus Torvalds's avatar
Linus Torvalds committed
6375 6376
EXPORT_SYMBOL(schedule);

6377 6378 6379 6380 6381 6382 6383 6384 6385 6386 6387 6388 6389 6390 6391 6392 6393 6394 6395
/*
 * 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.
	 */
6396
	WARN_ON_ONCE(current->__state);
6397
	do {
6398
		__schedule(SM_NONE);
6399 6400 6401
	} while (need_resched());
}

6402
#if defined(CONFIG_CONTEXT_TRACKING) && !defined(CONFIG_HAVE_CONTEXT_TRACKING_OFFSTACK)
6403
asmlinkage __visible void __sched schedule_user(void)
6404 6405 6406 6407 6408 6409
{
	/*
	 * 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.
6410 6411
	 *
	 * NB: There are buggy callers of this function.  Ideally we
6412
	 * should warn if prev_state != CONTEXT_USER, but that will trigger
6413
	 * too frequently to make sense yet.
6414
	 */
6415
	enum ctx_state prev_state = exception_enter();
6416
	schedule();
6417
	exception_exit(prev_state);
6418 6419 6420
}
#endif

6421 6422 6423 6424 6425 6426 6427
/**
 * schedule_preempt_disabled - called with preemption disabled
 *
 * Returns with preemption disabled. Note: preempt_count must be 1
 */
void __sched schedule_preempt_disabled(void)
{
6428
	sched_preempt_enable_no_resched();
6429 6430 6431 6432
	schedule();
	preempt_disable();
}

6433 6434 6435 6436 6437 6438 6439 6440 6441 6442 6443 6444
#ifdef CONFIG_PREEMPT_RT
void __sched notrace schedule_rtlock(void)
{
	do {
		preempt_disable();
		__schedule(SM_RTLOCK_WAIT);
		sched_preempt_enable_no_resched();
	} while (need_resched());
}
NOKPROBE_SYMBOL(schedule_rtlock);
#endif

6445
static void __sched notrace preempt_schedule_common(void)
6446 6447
{
	do {
6448 6449 6450 6451 6452 6453 6454 6455 6456 6457 6458 6459 6460
		/*
		 * 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.
		 */
6461
		preempt_disable_notrace();
6462
		preempt_latency_start(1);
6463
		__schedule(SM_PREEMPT);
6464
		preempt_latency_stop(1);
6465
		preempt_enable_no_resched_notrace();
6466 6467 6468 6469 6470 6471 6472 6473

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

6474
#ifdef CONFIG_PREEMPTION
Linus Torvalds's avatar
Linus Torvalds committed
6475
/*
6476 6477
 * This is the entry point to schedule() from in-kernel preemption
 * off of preempt_enable.
Linus Torvalds's avatar
Linus Torvalds committed
6478
 */
6479
asmlinkage __visible void __sched notrace preempt_schedule(void)
Linus Torvalds's avatar
Linus Torvalds committed
6480 6481 6482
{
	/*
	 * If there is a non-zero preempt_count or interrupts are disabled,
Ingo Molnar's avatar
Ingo Molnar committed
6483
	 * we do not want to preempt the current task. Just return..
Linus Torvalds's avatar
Linus Torvalds committed
6484
	 */
6485
	if (likely(!preemptible()))
Linus Torvalds's avatar
Linus Torvalds committed
6486 6487
		return;

6488
	preempt_schedule_common();
Linus Torvalds's avatar
Linus Torvalds committed
6489
}
6490
NOKPROBE_SYMBOL(preempt_schedule);
Linus Torvalds's avatar
Linus Torvalds committed
6491
EXPORT_SYMBOL(preempt_schedule);
6492

6493 6494
#ifdef CONFIG_PREEMPT_DYNAMIC
DEFINE_STATIC_CALL(preempt_schedule, __preempt_schedule_func);
6495
EXPORT_STATIC_CALL_TRAMP(preempt_schedule);
6496 6497 6498
#endif


6499
/**
6500
 * preempt_schedule_notrace - preempt_schedule called by tracing
6501 6502 6503 6504 6505 6506 6507 6508 6509 6510 6511 6512
 *
 * 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.
 */
6513
asmlinkage __visible void __sched notrace preempt_schedule_notrace(void)
6514 6515 6516 6517 6518 6519 6520
{
	enum ctx_state prev_ctx;

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

	do {
6521 6522 6523 6524 6525 6526 6527 6528 6529 6530 6531 6532 6533
		/*
		 * 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.
		 */
6534
		preempt_disable_notrace();
6535
		preempt_latency_start(1);
6536 6537 6538 6539 6540 6541
		/*
		 * Needs preempt disabled in case user_exit() is traced
		 * and the tracer calls preempt_enable_notrace() causing
		 * an infinite recursion.
		 */
		prev_ctx = exception_enter();
6542
		__schedule(SM_PREEMPT);
6543 6544
		exception_exit(prev_ctx);

6545
		preempt_latency_stop(1);
6546
		preempt_enable_no_resched_notrace();
6547 6548
	} while (need_resched());
}
6549
EXPORT_SYMBOL_GPL(preempt_schedule_notrace);
6550

6551 6552
#ifdef CONFIG_PREEMPT_DYNAMIC
DEFINE_STATIC_CALL(preempt_schedule_notrace, __preempt_schedule_notrace_func);
6553
EXPORT_STATIC_CALL_TRAMP(preempt_schedule_notrace);
6554 6555
#endif

6556
#endif /* CONFIG_PREEMPTION */
Linus Torvalds's avatar
Linus Torvalds committed
6557

6558 6559 6560 6561 6562 6563 6564 6565 6566 6567 6568 6569 6570 6571 6572 6573 6574 6575 6576 6577 6578 6579 6580 6581 6582 6583 6584 6585 6586 6587 6588 6589 6590
#ifdef CONFIG_PREEMPT_DYNAMIC

#include <linux/entry-common.h>

/*
 * SC:cond_resched
 * SC:might_resched
 * SC:preempt_schedule
 * SC:preempt_schedule_notrace
 * SC:irqentry_exit_cond_resched
 *
 *
 * NONE:
 *   cond_resched               <- __cond_resched
 *   might_resched              <- RET0
 *   preempt_schedule           <- NOP
 *   preempt_schedule_notrace   <- NOP
 *   irqentry_exit_cond_resched <- NOP
 *
 * VOLUNTARY:
 *   cond_resched               <- __cond_resched
 *   might_resched              <- __cond_resched
 *   preempt_schedule           <- NOP
 *   preempt_schedule_notrace   <- NOP
 *   irqentry_exit_cond_resched <- NOP
 *
 * FULL:
 *   cond_resched               <- RET0
 *   might_resched              <- RET0
 *   preempt_schedule           <- preempt_schedule
 *   preempt_schedule_notrace   <- preempt_schedule_notrace
 *   irqentry_exit_cond_resched <- irqentry_exit_cond_resched
 */
6591 6592

enum {
6593 6594
	preempt_dynamic_undefined = -1,
	preempt_dynamic_none,
6595 6596 6597 6598
	preempt_dynamic_voluntary,
	preempt_dynamic_full,
};

6599
int preempt_dynamic_mode = preempt_dynamic_undefined;
6600

6601
int sched_dynamic_mode(const char *str)
6602
{
6603
	if (!strcmp(str, "none"))
6604
		return preempt_dynamic_none;
6605 6606

	if (!strcmp(str, "voluntary"))
6607
		return preempt_dynamic_voluntary;
6608 6609

	if (!strcmp(str, "full"))
6610
		return preempt_dynamic_full;
6611

6612
	return -EINVAL;
6613 6614
}

6615
void sched_dynamic_update(int mode)
6616 6617 6618 6619 6620 6621 6622 6623 6624 6625 6626 6627 6628
{
	/*
	 * Avoid {NONE,VOLUNTARY} -> FULL transitions from ever ending up in
	 * the ZERO state, which is invalid.
	 */
	static_call_update(cond_resched, __cond_resched);
	static_call_update(might_resched, __cond_resched);
	static_call_update(preempt_schedule, __preempt_schedule_func);
	static_call_update(preempt_schedule_notrace, __preempt_schedule_notrace_func);
	static_call_update(irqentry_exit_cond_resched, irqentry_exit_cond_resched);

	switch (mode) {
	case preempt_dynamic_none:
6629
		static_call_update(cond_resched, __cond_resched);
6630 6631 6632 6633
		static_call_update(might_resched, (void *)&__static_call_return0);
		static_call_update(preempt_schedule, NULL);
		static_call_update(preempt_schedule_notrace, NULL);
		static_call_update(irqentry_exit_cond_resched, NULL);
6634 6635 6636 6637
		pr_info("Dynamic Preempt: none\n");
		break;

	case preempt_dynamic_voluntary:
6638 6639
		static_call_update(cond_resched, __cond_resched);
		static_call_update(might_resched, __cond_resched);
6640 6641 6642
		static_call_update(preempt_schedule, NULL);
		static_call_update(preempt_schedule_notrace, NULL);
		static_call_update(irqentry_exit_cond_resched, NULL);
6643 6644 6645 6646
		pr_info("Dynamic Preempt: voluntary\n");
		break;

	case preempt_dynamic_full:
6647 6648
		static_call_update(cond_resched, (void *)&__static_call_return0);
		static_call_update(might_resched, (void *)&__static_call_return0);
6649 6650 6651
		static_call_update(preempt_schedule, __preempt_schedule_func);
		static_call_update(preempt_schedule_notrace, __preempt_schedule_notrace_func);
		static_call_update(irqentry_exit_cond_resched, irqentry_exit_cond_resched);
6652 6653 6654 6655 6656 6657 6658 6659 6660 6661 6662 6663
		pr_info("Dynamic Preempt: full\n");
		break;
	}

	preempt_dynamic_mode = mode;
}

static int __init setup_preempt_mode(char *str)
{
	int mode = sched_dynamic_mode(str);
	if (mode < 0) {
		pr_warn("Dynamic Preempt: unsupported mode: %s\n", str);
6664
		return 0;
6665
	}
6666 6667

	sched_dynamic_update(mode);
6668
	return 1;
6669 6670 6671
}
__setup("preempt=", setup_preempt_mode);

6672 6673 6674
static void __init preempt_dynamic_init(void)
{
	if (preempt_dynamic_mode == preempt_dynamic_undefined) {
6675
		if (IS_ENABLED(CONFIG_PREEMPT_NONE)) {
6676
			sched_dynamic_update(preempt_dynamic_none);
6677
		} else if (IS_ENABLED(CONFIG_PREEMPT_VOLUNTARY)) {
6678 6679 6680
			sched_dynamic_update(preempt_dynamic_voluntary);
		} else {
			/* Default static call setting, nothing to do */
6681
			WARN_ON_ONCE(!IS_ENABLED(CONFIG_PREEMPT));
6682 6683 6684 6685 6686 6687 6688 6689 6690 6691 6692
			preempt_dynamic_mode = preempt_dynamic_full;
			pr_info("Dynamic Preempt: full\n");
		}
	}
}

#else /* !CONFIG_PREEMPT_DYNAMIC */

static inline void preempt_dynamic_init(void) { }

#endif /* #ifdef CONFIG_PREEMPT_DYNAMIC */
6693

Linus Torvalds's avatar
Linus Torvalds committed
6694
/*
6695
 * This is the entry point to schedule() from kernel preemption
Linus Torvalds's avatar
Linus Torvalds committed
6696 6697 6698 6699
 * off of irq context.
 * Note, that this is called and return with irqs disabled. This will
 * protect us against recursive calling from irq.
 */
6700
asmlinkage __visible void __sched preempt_schedule_irq(void)
Linus Torvalds's avatar
Linus Torvalds committed
6701
{
6702
	enum ctx_state prev_state;
6703

6704
	/* Catch callers which need to be fixed */
6705
	BUG_ON(preempt_count() || !irqs_disabled());
Linus Torvalds's avatar
Linus Torvalds committed
6706

6707 6708
	prev_state = exception_enter();

6709
	do {
6710
		preempt_disable();
6711
		local_irq_enable();
6712
		__schedule(SM_PREEMPT);
6713
		local_irq_disable();
6714
		sched_preempt_enable_no_resched();
6715
	} while (need_resched());
6716 6717

	exception_exit(prev_state);
Linus Torvalds's avatar
Linus Torvalds committed
6718 6719
}

6720
int default_wake_function(wait_queue_entry_t *curr, unsigned mode, int wake_flags,
Ingo Molnar's avatar
Ingo Molnar committed
6721
			  void *key)
Linus Torvalds's avatar
Linus Torvalds committed
6722
{
6723
	WARN_ON_ONCE(IS_ENABLED(CONFIG_SCHED_DEBUG) && wake_flags & ~WF_SYNC);
6724
	return try_to_wake_up(curr->private, mode, wake_flags);
Linus Torvalds's avatar
Linus Torvalds committed
6725 6726 6727
}
EXPORT_SYMBOL(default_wake_function);

6728 6729 6730 6731 6732 6733 6734 6735 6736 6737 6738 6739
static void __setscheduler_prio(struct task_struct *p, int prio)
{
	if (dl_prio(prio))
		p->sched_class = &dl_sched_class;
	else if (rt_prio(prio))
		p->sched_class = &rt_sched_class;
	else
		p->sched_class = &fair_sched_class;

	p->prio = prio;
}

6740 6741
#ifdef CONFIG_RT_MUTEXES

6742 6743 6744 6745 6746 6747 6748 6749 6750 6751 6752 6753 6754 6755 6756
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);
}

6757 6758
/*
 * rt_mutex_setprio - set the current priority of a task
6759 6760
 * @p: task to boost
 * @pi_task: donor task
6761 6762 6763 6764
 *
 * This function changes the 'effective' priority of a task. It does
 * not touch ->normal_prio like __setscheduler().
 *
6765 6766
 * Used by the rt_mutex code to implement priority inheritance
 * logic. Call site only calls if the priority of the task changed.
6767
 */
6768
void rt_mutex_setprio(struct task_struct *p, struct task_struct *pi_task)
6769
{
6770
	int prio, oldprio, queued, running, queue_flag =
6771
		DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
6772
	const struct sched_class *prev_class;
6773 6774
	struct rq_flags rf;
	struct rq *rq;
6775

6776 6777 6778 6779 6780 6781 6782 6783
	/* 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;
6784

6785
	rq = __task_rq_lock(p, &rf);
6786
	update_rq_clock(rq);
6787 6788 6789 6790 6791 6792 6793 6794
	/*
	 * 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
6795
	 * points to a blocked task -- which guarantees the task is present.
6796 6797 6798 6799 6800 6801 6802 6803
	 */
	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;
6804

6805 6806 6807 6808 6809 6810 6811 6812 6813 6814 6815 6816 6817 6818 6819 6820 6821 6822
	/*
	 * 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;
	}

6823
	trace_sched_pi_setprio(p, pi_task);
6824
	oldprio = p->prio;
6825 6826 6827 6828

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

6829
	prev_class = p->sched_class;
6830
	queued = task_on_rq_queued(p);
6831
	running = task_current(rq, p);
6832
	if (queued)
6833
		dequeue_task(rq, p, queue_flag);
6834
	if (running)
6835
		put_prev_task(rq, p);
Ingo Molnar's avatar
Ingo Molnar committed
6836

6837 6838 6839 6840 6841 6842 6843 6844 6845 6846
	/*
	 * 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)) {
6847
		if (!dl_prio(p->normal_prio) ||
6848 6849
		    (pi_task && dl_prio(pi_task->prio) &&
		     dl_entity_preempt(&pi_task->dl, &p->dl))) {
6850
			p->dl.pi_se = pi_task->dl.pi_se;
6851
			queue_flag |= ENQUEUE_REPLENISH;
6852 6853 6854
		} else {
			p->dl.pi_se = &p->dl;
		}
6855 6856
	} else if (rt_prio(prio)) {
		if (dl_prio(oldprio))
6857
			p->dl.pi_se = &p->dl;
6858
		if (oldprio < prio)
6859
			queue_flag |= ENQUEUE_HEAD;
6860 6861
	} else {
		if (dl_prio(oldprio))
6862
			p->dl.pi_se = &p->dl;
6863 6864
		if (rt_prio(oldprio))
			p->rt.timeout = 0;
6865
	}
Ingo Molnar's avatar
Ingo Molnar committed
6866

6867
	__setscheduler_prio(p, prio);
6868

6869
	if (queued)
6870
		enqueue_task(rq, p, queue_flag);
6871
	if (running)
6872
		set_next_task(rq, p);
6873

6874
	check_class_changed(rq, p, prev_class, oldprio);
6875
out_unlock:
6876 6877
	/* Avoid rq from going away on us: */
	preempt_disable();
6878

6879 6880
	rq_unpin_lock(rq, &rf);
	__balance_callbacks(rq);
6881
	raw_spin_rq_unlock(rq);
6882

6883
	preempt_enable();
6884
}
6885 6886 6887 6888 6889
#else
static inline int rt_effective_prio(struct task_struct *p, int prio)
{
	return prio;
}
6890
#endif
6891

6892
void set_user_nice(struct task_struct *p, long nice)
Linus Torvalds's avatar
Linus Torvalds committed
6893
{
6894
	bool queued, running;
6895
	int old_prio;
6896
	struct rq_flags rf;
6897
	struct rq *rq;
Linus Torvalds's avatar
Linus Torvalds committed
6898

6899
	if (task_nice(p) == nice || nice < MIN_NICE || nice > MAX_NICE)
Linus Torvalds's avatar
Linus Torvalds committed
6900 6901 6902 6903 6904
		return;
	/*
	 * We have to be careful, if called from sys_setpriority(),
	 * the task might be in the middle of scheduling on another CPU.
	 */
6905
	rq = task_rq_lock(p, &rf);
6906 6907
	update_rq_clock(rq);

Linus Torvalds's avatar
Linus Torvalds committed
6908 6909 6910
	/*
	 * The RT priorities are set via sched_setscheduler(), but we still
	 * allow the 'normal' nice value to be set - but as expected
6911
	 * it won't have any effect on scheduling until the task is
6912
	 * SCHED_DEADLINE, SCHED_FIFO or SCHED_RR:
Linus Torvalds's avatar
Linus Torvalds committed
6913
	 */
6914
	if (task_has_dl_policy(p) || task_has_rt_policy(p)) {
Linus Torvalds's avatar
Linus Torvalds committed
6915 6916 6917
		p->static_prio = NICE_TO_PRIO(nice);
		goto out_unlock;
	}
6918
	queued = task_on_rq_queued(p);
6919
	running = task_current(rq, p);
6920
	if (queued)
6921
		dequeue_task(rq, p, DEQUEUE_SAVE | DEQUEUE_NOCLOCK);
6922 6923
	if (running)
		put_prev_task(rq, p);
Linus Torvalds's avatar
Linus Torvalds committed
6924 6925

	p->static_prio = NICE_TO_PRIO(nice);
6926
	set_load_weight(p, true);
6927 6928
	old_prio = p->prio;
	p->prio = effective_prio(p);
Linus Torvalds's avatar
Linus Torvalds committed
6929

6930
	if (queued)
6931
		enqueue_task(rq, p, ENQUEUE_RESTORE | ENQUEUE_NOCLOCK);
6932
	if (running)
6933
		set_next_task(rq, p);
6934 6935 6936 6937 6938 6939 6940

	/*
	 * 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
6941
out_unlock:
6942
	task_rq_unlock(rq, p, &rf);
Linus Torvalds's avatar
Linus Torvalds committed
6943 6944 6945
}
EXPORT_SYMBOL(set_user_nice);

6946 6947 6948 6949 6950
/*
 * can_nice - check if a task can reduce its nice value
 * @p: task
 * @nice: nice value
 */
6951
int can_nice(const struct task_struct *p, const int nice)
6952
{
6953
	/* Convert nice value [19,-20] to rlimit style value [1,40]: */
6954
	int nice_rlim = nice_to_rlimit(nice);
6955

6956
	return (nice_rlim <= task_rlimit(p, RLIMIT_NICE) ||
6957 6958 6959
		capable(CAP_SYS_NICE));
}

Linus Torvalds's avatar
Linus Torvalds committed
6960 6961 6962 6963 6964 6965 6966 6967 6968
#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.
 */
6969
SYSCALL_DEFINE1(nice, int, increment)
Linus Torvalds's avatar
Linus Torvalds committed
6970
{
6971
	long nice, retval;
Linus Torvalds's avatar
Linus Torvalds committed
6972 6973 6974 6975 6976 6977

	/*
	 * 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.
	 */
6978
	increment = clamp(increment, -NICE_WIDTH, NICE_WIDTH);
6979
	nice = task_nice(current) + increment;
Linus Torvalds's avatar
Linus Torvalds committed
6980

6981
	nice = clamp_val(nice, MIN_NICE, MAX_NICE);
6982 6983 6984
	if (increment < 0 && !can_nice(current, nice))
		return -EPERM;

Linus Torvalds's avatar
Linus Torvalds committed
6985 6986 6987 6988 6989 6990 6991 6992 6993 6994 6995 6996 6997 6998
	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.
 *
6999
 * Return: The priority value as seen by users in /proc.
7000 7001 7002 7003 7004 7005
 *
 * sched policy         return value   kernel prio    user prio/nice
 *
 * normal, batch, idle     [0 ... 39]  [100 ... 139]          0/[-20 ... 19]
 * fifo, rr             [-2 ... -100]     [98 ... 0]  [1 ... 99]
 * deadline                     -101             -1           0
Linus Torvalds's avatar
Linus Torvalds committed
7006
 */
7007
int task_prio(const struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
7008 7009 7010 7011 7012
{
	return p->prio - MAX_RT_PRIO;
}

/**
7013
 * idle_cpu - is a given CPU idle currently?
Linus Torvalds's avatar
Linus Torvalds committed
7014
 * @cpu: the processor in question.
7015 7016
 *
 * Return: 1 if the CPU is currently idle. 0 otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
7017 7018 7019
 */
int idle_cpu(int cpu)
{
Thomas Gleixner's avatar
Thomas Gleixner committed
7020 7021 7022 7023 7024 7025 7026 7027 7028
	struct rq *rq = cpu_rq(cpu);

	if (rq->curr != rq->idle)
		return 0;

	if (rq->nr_running)
		return 0;

#ifdef CONFIG_SMP
7029
	if (rq->ttwu_pending)
Thomas Gleixner's avatar
Thomas Gleixner committed
7030 7031 7032 7033
		return 0;
#endif

	return 1;
Linus Torvalds's avatar
Linus Torvalds committed
7034 7035
}

7036 7037 7038 7039 7040 7041 7042 7043 7044 7045 7046
/**
 * 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;

7047 7048 7049
	if (vcpu_is_preempted(cpu))
		return 0;

Thomas Gleixner's avatar
Thomas Gleixner committed
7050
	return 1;
Linus Torvalds's avatar
Linus Torvalds committed
7051 7052 7053
}

/**
7054
 * idle_task - return the idle task for a given CPU.
Linus Torvalds's avatar
Linus Torvalds committed
7055
 * @cpu: the processor in question.
7056
 *
7057
 * Return: The idle task for the CPU @cpu.
Linus Torvalds's avatar
Linus Torvalds committed
7058
 */
7059
struct task_struct *idle_task(int cpu)
Linus Torvalds's avatar
Linus Torvalds committed
7060 7061 7062 7063
{
	return cpu_rq(cpu)->idle;
}

7064 7065 7066 7067 7068 7069 7070 7071 7072 7073 7074 7075 7076 7077 7078 7079 7080 7081 7082 7083 7084
#ifdef CONFIG_SMP
/*
 * This function computes an effective utilization for the given CPU, to be
 * used for frequency selection given the linear relation: f = u * f_max.
 *
 * The scheduler tracks the following metrics:
 *
 *   cpu_util_{cfs,rt,dl,irq}()
 *   cpu_bw_dl()
 *
 * Where the cfs,rt and dl util numbers are tracked with the same metric and
 * synchronized windows and are thus directly comparable.
 *
 * The cfs,rt,dl utilization are the running times measured with rq->clock_task
 * which excludes things like IRQ and steal-time. These latter are then accrued
 * in the irq utilization.
 *
 * The DL bandwidth number otoh is not a measured metric but a value computed
 * based on the task model parameters and gives the minimal utilization
 * required to meet deadlines.
 */
7085 7086
unsigned long effective_cpu_util(int cpu, unsigned long util_cfs,
				 unsigned long max, enum cpu_util_type type,
7087 7088 7089 7090 7091 7092 7093 7094 7095 7096 7097 7098 7099 7100 7101 7102 7103 7104 7105 7106 7107 7108 7109 7110 7111 7112 7113 7114 7115 7116 7117 7118 7119 7120 7121 7122 7123 7124 7125 7126 7127 7128 7129 7130 7131 7132 7133 7134 7135 7136 7137 7138 7139 7140 7141 7142 7143 7144 7145 7146 7147 7148 7149 7150 7151 7152 7153 7154 7155 7156 7157 7158 7159 7160 7161 7162 7163 7164 7165 7166 7167 7168 7169
				 struct task_struct *p)
{
	unsigned long dl_util, util, irq;
	struct rq *rq = cpu_rq(cpu);

	if (!uclamp_is_used() &&
	    type == FREQUENCY_UTIL && rt_rq_is_runnable(&rq->rt)) {
		return max;
	}

	/*
	 * Early check to see if IRQ/steal time saturates the CPU, can be
	 * because of inaccuracies in how we track these -- see
	 * update_irq_load_avg().
	 */
	irq = cpu_util_irq(rq);
	if (unlikely(irq >= max))
		return max;

	/*
	 * Because the time spend on RT/DL tasks is visible as 'lost' time to
	 * CFS tasks and we use the same metric to track the effective
	 * utilization (PELT windows are synchronized) we can directly add them
	 * to obtain the CPU's actual utilization.
	 *
	 * CFS and RT utilization can be boosted or capped, depending on
	 * utilization clamp constraints requested by currently RUNNABLE
	 * tasks.
	 * When there are no CFS RUNNABLE tasks, clamps are released and
	 * frequency will be gracefully reduced with the utilization decay.
	 */
	util = util_cfs + cpu_util_rt(rq);
	if (type == FREQUENCY_UTIL)
		util = uclamp_rq_util_with(rq, util, p);

	dl_util = cpu_util_dl(rq);

	/*
	 * For frequency selection we do not make cpu_util_dl() a permanent part
	 * of this sum because we want to use cpu_bw_dl() later on, but we need
	 * to check if the CFS+RT+DL sum is saturated (ie. no idle time) such
	 * that we select f_max when there is no idle time.
	 *
	 * NOTE: numerical errors or stop class might cause us to not quite hit
	 * saturation when we should -- something for later.
	 */
	if (util + dl_util >= max)
		return max;

	/*
	 * OTOH, for energy computation we need the estimated running time, so
	 * include util_dl and ignore dl_bw.
	 */
	if (type == ENERGY_UTIL)
		util += dl_util;

	/*
	 * There is still idle time; further improve the number by using the
	 * irq metric. Because IRQ/steal time is hidden from the task clock we
	 * need to scale the task numbers:
	 *
	 *              max - irq
	 *   U' = irq + --------- * U
	 *                 max
	 */
	util = scale_irq_capacity(util, irq, max);
	util += irq;

	/*
	 * Bandwidth required by DEADLINE must always be granted while, for
	 * FAIR and RT, we use blocked utilization of IDLE CPUs as a mechanism
	 * to gracefully reduce the frequency when no tasks show up for longer
	 * periods of time.
	 *
	 * Ideally we would like to set bw_dl as min/guaranteed freq and util +
	 * bw_dl as requested freq. However, cpufreq is not yet ready for such
	 * an interface. So, we only do the latter for now.
	 */
	if (type == FREQUENCY_UTIL)
		util += cpu_bw_dl(rq);

	return min(max, util);
}
7170 7171 7172

unsigned long sched_cpu_util(int cpu, unsigned long max)
{
7173
	return effective_cpu_util(cpu, cpu_util_cfs(cpu), max,
7174 7175
				  ENERGY_UTIL, NULL);
}
7176 7177
#endif /* CONFIG_SMP */

Linus Torvalds's avatar
Linus Torvalds committed
7178 7179 7180
/**
 * find_process_by_pid - find a process with a matching PID value.
 * @pid: the pid in question.
7181 7182
 *
 * The task of @pid, if found. %NULL otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
7183
 */
Alexey Dobriyan's avatar
Alexey Dobriyan committed
7184
static struct task_struct *find_process_by_pid(pid_t pid)
Linus Torvalds's avatar
Linus Torvalds committed
7185
{
7186
	return pid ? find_task_by_vpid(pid) : current;
Linus Torvalds's avatar
Linus Torvalds committed
7187 7188
}

7189 7190 7191 7192 7193 7194
/*
 * sched_setparam() passes in -1 for its policy, to let the functions
 * it calls know not to change it.
 */
#define SETPARAM_POLICY	-1

7195 7196
static void __setscheduler_params(struct task_struct *p,
		const struct sched_attr *attr)
Linus Torvalds's avatar
Linus Torvalds committed
7197
{
7198 7199
	int policy = attr->sched_policy;

7200
	if (policy == SETPARAM_POLICY)
7201 7202
		policy = p->policy;

Linus Torvalds's avatar
Linus Torvalds committed
7203
	p->policy = policy;
7204

7205 7206
	if (dl_policy(policy))
		__setparam_dl(p, attr);
7207
	else if (fair_policy(policy))
7208 7209
		p->static_prio = NICE_TO_PRIO(attr->sched_nice);

7210 7211 7212 7213 7214 7215
	/*
	 * __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;
7216
	p->normal_prio = normal_prio(p);
7217
	set_load_weight(p, true);
7218
}
7219

7220
/*
7221
 * Check the target process has a UID that matches the current process's:
7222 7223 7224 7225 7226 7227 7228 7229
 */
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);
7230 7231
	match = (uid_eq(cred->euid, pcred->euid) ||
		 uid_eq(cred->euid, pcred->uid));
7232 7233 7234 7235
	rcu_read_unlock();
	return match;
}

7236 7237
static int __sched_setscheduler(struct task_struct *p,
				const struct sched_attr *attr,
7238
				bool user, bool pi)
Linus Torvalds's avatar
Linus Torvalds committed
7239
{
7240 7241
	int oldpolicy = -1, policy = attr->sched_policy;
	int retval, oldprio, newprio, queued, running;
7242
	const struct sched_class *prev_class;
7243
	struct callback_head *head;
7244
	struct rq_flags rf;
7245
	int reset_on_fork;
7246
	int queue_flags = DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
7247
	struct rq *rq;
Linus Torvalds's avatar
Linus Torvalds committed
7248

7249 7250
	/* The pi code expects interrupts enabled */
	BUG_ON(pi && in_interrupt());
Linus Torvalds's avatar
Linus Torvalds committed
7251
recheck:
7252
	/* Double check policy once rq lock held: */
7253 7254
	if (policy < 0) {
		reset_on_fork = p->sched_reset_on_fork;
Linus Torvalds's avatar
Linus Torvalds committed
7255
		policy = oldpolicy = p->policy;
7256
	} else {
7257
		reset_on_fork = !!(attr->sched_flags & SCHED_FLAG_RESET_ON_FORK);
7258

7259
		if (!valid_policy(policy))
7260 7261 7262
			return -EINVAL;
	}

7263
	if (attr->sched_flags & ~(SCHED_FLAG_ALL | SCHED_FLAG_SUGOV))
7264 7265
		return -EINVAL;

Linus Torvalds's avatar
Linus Torvalds committed
7266 7267
	/*
	 * Valid priorities for SCHED_FIFO and SCHED_RR are
7268
	 * 1..MAX_RT_PRIO-1, valid priority for SCHED_NORMAL,
Ingo Molnar's avatar
Ingo Molnar committed
7269
	 * SCHED_BATCH and SCHED_IDLE is 0.
Linus Torvalds's avatar
Linus Torvalds committed
7270
	 */
7271
	if (attr->sched_priority > MAX_RT_PRIO-1)
Linus Torvalds's avatar
Linus Torvalds committed
7272
		return -EINVAL;
7273 7274
	if ((dl_policy(policy) && !__checkparam_dl(attr)) ||
	    (rt_policy(policy) != (attr->sched_priority != 0)))
Linus Torvalds's avatar
Linus Torvalds committed
7275 7276
		return -EINVAL;

7277 7278 7279
	/*
	 * Allow unprivileged RT tasks to decrease priority:
	 */
7280
	if (user && !capable(CAP_SYS_NICE)) {
7281
		if (fair_policy(policy)) {
7282
			if (attr->sched_nice < task_nice(p) &&
7283
			    !can_nice(p, attr->sched_nice))
7284 7285 7286
				return -EPERM;
		}

7287
		if (rt_policy(policy)) {
7288 7289
			unsigned long rlim_rtprio =
					task_rlimit(p, RLIMIT_RTPRIO);
7290

7291
			/* Can't set/change the rt policy: */
7292 7293 7294
			if (policy != p->policy && !rlim_rtprio)
				return -EPERM;

7295
			/* Can't increase priority: */
7296 7297
			if (attr->sched_priority > p->rt_priority &&
			    attr->sched_priority > rlim_rtprio)
7298 7299
				return -EPERM;
		}
7300

7301 7302 7303 7304 7305 7306 7307 7308 7309
		 /*
		  * 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
7310
		/*
7311 7312
		 * 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
7313
		 */
7314
		if (task_has_idle_policy(p) && !idle_policy(policy)) {
7315
			if (!can_nice(p, task_nice(p)))
7316 7317
				return -EPERM;
		}
7318

7319
		/* Can't change other user's priorities: */
7320
		if (!check_same_owner(p))
7321
			return -EPERM;
7322

7323
		/* Normal users shall not reset the sched_reset_on_fork flag: */
7324 7325
		if (p->sched_reset_on_fork && !reset_on_fork)
			return -EPERM;
7326
	}
Linus Torvalds's avatar
Linus Torvalds committed
7327

7328
	if (user) {
7329 7330 7331
		if (attr->sched_flags & SCHED_FLAG_SUGOV)
			return -EINVAL;

7332
		retval = security_task_setscheduler(p);
7333 7334 7335 7336
		if (retval)
			return retval;
	}

7337 7338 7339 7340 7341 7342 7343
	/* Update task specific "requested" clamps */
	if (attr->sched_flags & SCHED_FLAG_UTIL_CLAMP) {
		retval = uclamp_validate(p, attr);
		if (retval)
			return retval;
	}

7344 7345 7346
	if (pi)
		cpuset_read_lock();

7347
	/*
7348
	 * Make sure no PI-waiters arrive (or leave) while we are
7349
	 * changing the priority of the task:
7350
	 *
Lucas De Marchi's avatar
Lucas De Marchi committed
7351
	 * To be able to change p->policy safely, the appropriate
Linus Torvalds's avatar
Linus Torvalds committed
7352 7353
	 * runqueue lock must be held.
	 */
7354
	rq = task_rq_lock(p, &rf);
7355
	update_rq_clock(rq);
7356

7357
	/*
7358
	 * Changing the policy of the stop threads its a very bad idea:
7359 7360
	 */
	if (p == rq->stop) {
7361 7362
		retval = -EINVAL;
		goto unlock;
7363 7364
	}

7365
	/*
7366 7367
	 * If not changing anything there's no need to proceed further,
	 * but store a possible modification of reset_on_fork.
7368
	 */
7369
	if (unlikely(policy == p->policy)) {
7370
		if (fair_policy(policy) && attr->sched_nice != task_nice(p))
7371 7372 7373
			goto change;
		if (rt_policy(policy) && attr->sched_priority != p->rt_priority)
			goto change;
7374
		if (dl_policy(policy) && dl_param_changed(p, attr))
7375
			goto change;
7376 7377
		if (attr->sched_flags & SCHED_FLAG_UTIL_CLAMP)
			goto change;
7378

7379
		p->sched_reset_on_fork = reset_on_fork;
7380 7381
		retval = 0;
		goto unlock;
7382
	}
7383
change:
7384

7385
	if (user) {
7386
#ifdef CONFIG_RT_GROUP_SCHED
7387 7388 7389 7390 7391
		/*
		 * Do not allow realtime tasks into groups that have no runtime
		 * assigned.
		 */
		if (rt_bandwidth_enabled() && rt_policy(policy) &&
7392 7393
				task_group(p)->rt_bandwidth.rt_runtime == 0 &&
				!task_group_is_autogroup(task_group(p))) {
7394 7395
			retval = -EPERM;
			goto unlock;
7396 7397
		}
#endif
7398
#ifdef CONFIG_SMP
7399 7400
		if (dl_bandwidth_enabled() && dl_policy(policy) &&
				!(attr->sched_flags & SCHED_FLAG_SUGOV)) {
7401 7402 7403 7404 7405 7406 7407
			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.
			 */
7408
			if (!cpumask_subset(span, p->cpus_ptr) ||
7409
			    rq->rd->dl_bw.bw == 0) {
7410 7411
				retval = -EPERM;
				goto unlock;
7412 7413 7414 7415
			}
		}
#endif
	}
7416

7417
	/* Re-check policy now with rq lock held: */
Linus Torvalds's avatar
Linus Torvalds committed
7418 7419
	if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) {
		policy = oldpolicy = -1;
7420
		task_rq_unlock(rq, p, &rf);
7421 7422
		if (pi)
			cpuset_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
7423 7424
		goto recheck;
	}
7425 7426 7427 7428 7429 7430

	/*
	 * If setscheduling to SCHED_DEADLINE (or changing the parameters
	 * of a SCHED_DEADLINE task) we need to check if enough bandwidth
	 * is available.
	 */
7431
	if ((dl_policy(policy) || dl_task(p)) && sched_dl_overflow(p, policy, attr)) {
7432 7433
		retval = -EBUSY;
		goto unlock;
7434 7435
	}

7436 7437 7438
	p->sched_reset_on_fork = reset_on_fork;
	oldprio = p->prio;

7439
	newprio = __normal_prio(policy, attr->sched_priority, attr->sched_nice);
7440 7441 7442 7443 7444 7445 7446 7447
	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.
		 */
7448 7449
		newprio = rt_effective_prio(p, newprio);
		if (newprio == oldprio)
7450
			queue_flags &= ~DEQUEUE_MOVE;
7451 7452
	}

7453
	queued = task_on_rq_queued(p);
7454
	running = task_current(rq, p);
7455
	if (queued)
7456
		dequeue_task(rq, p, queue_flags);
7457
	if (running)
7458
		put_prev_task(rq, p);
7459

7460
	prev_class = p->sched_class;
7461

7462 7463 7464 7465
	if (!(attr->sched_flags & SCHED_FLAG_KEEP_PARAMS)) {
		__setscheduler_params(p, attr);
		__setscheduler_prio(p, newprio);
	}
7466
	__setscheduler_uclamp(p, attr);
7467

7468
	if (queued) {
7469 7470 7471 7472
		/*
		 * We enqueue to tail when the priority of a task is
		 * increased (user space view).
		 */
7473 7474
		if (oldprio < p->prio)
			queue_flags |= ENQUEUE_HEAD;
7475

7476
		enqueue_task(rq, p, queue_flags);
7477
	}
7478
	if (running)
7479
		set_next_task(rq, p);
7480

7481
	check_class_changed(rq, p, prev_class, oldprio);
7482 7483 7484

	/* Avoid rq from going away on us: */
	preempt_disable();
7485
	head = splice_balance_callbacks(rq);
7486
	task_rq_unlock(rq, p, &rf);
7487

7488 7489
	if (pi) {
		cpuset_read_unlock();
7490
		rt_mutex_adjust_pi(p);
7491
	}
7492

7493
	/* Run balance callbacks after we've adjusted the PI chain: */
7494
	balance_callbacks(rq, head);
7495
	preempt_enable();
7496

Linus Torvalds's avatar
Linus Torvalds committed
7497
	return 0;
7498 7499 7500

unlock:
	task_rq_unlock(rq, p, &rf);
7501 7502
	if (pi)
		cpuset_read_unlock();
7503
	return retval;
Linus Torvalds's avatar
Linus Torvalds committed
7504
}
7505

7506 7507 7508 7509 7510 7511 7512 7513 7514
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),
	};

7515 7516
	/* Fixup the legacy SCHED_RESET_ON_FORK hack. */
	if ((policy != SETPARAM_POLICY) && (policy & SCHED_RESET_ON_FORK)) {
7517 7518 7519 7520 7521
		attr.sched_flags |= SCHED_FLAG_RESET_ON_FORK;
		policy &= ~SCHED_RESET_ON_FORK;
		attr.sched_policy = policy;
	}

7522
	return __sched_setscheduler(p, &attr, check, true);
7523
}
7524 7525 7526 7527 7528 7529
/**
 * 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.
 *
7530 7531
 * Use sched_set_fifo(), read its comment.
 *
7532 7533
 * Return: 0 on success. An error code otherwise.
 *
7534 7535 7536
 * NOTE that the task may be already dead.
 */
int sched_setscheduler(struct task_struct *p, int policy,
7537
		       const struct sched_param *param)
7538
{
7539
	return _sched_setscheduler(p, policy, param, true);
7540
}
Linus Torvalds's avatar
Linus Torvalds committed
7541

7542 7543
int sched_setattr(struct task_struct *p, const struct sched_attr *attr)
{
7544
	return __sched_setscheduler(p, attr, true, true);
7545 7546
}

7547 7548 7549 7550
int sched_setattr_nocheck(struct task_struct *p, const struct sched_attr *attr)
{
	return __sched_setscheduler(p, attr, false, true);
}
7551
EXPORT_SYMBOL_GPL(sched_setattr_nocheck);
7552

7553 7554 7555 7556 7557 7558 7559 7560 7561 7562
/**
 * 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.
7563 7564
 *
 * Return: 0 on success. An error code otherwise.
7565 7566
 */
int sched_setscheduler_nocheck(struct task_struct *p, int policy,
7567
			       const struct sched_param *param)
7568
{
7569
	return _sched_setscheduler(p, policy, param, false);
7570 7571
}

7572 7573 7574 7575 7576 7577 7578 7579 7580 7581 7582 7583 7584 7585 7586 7587 7588 7589
/*
 * SCHED_FIFO is a broken scheduler model; that is, it is fundamentally
 * incapable of resource management, which is the one thing an OS really should
 * be doing.
 *
 * This is of course the reason it is limited to privileged users only.
 *
 * Worse still; it is fundamentally impossible to compose static priority
 * workloads. You cannot take two correctly working static prio workloads
 * and smash them together and still expect them to work.
 *
 * For this reason 'all' FIFO tasks the kernel creates are basically at:
 *
 *   MAX_RT_PRIO / 2
 *
 * The administrator _MUST_ configure the system, the kernel simply doesn't
 * know enough information to make a sensible choice.
 */
7590
void sched_set_fifo(struct task_struct *p)
7591 7592
{
	struct sched_param sp = { .sched_priority = MAX_RT_PRIO / 2 };
7593
	WARN_ON_ONCE(sched_setscheduler_nocheck(p, SCHED_FIFO, &sp) != 0);
7594 7595 7596 7597 7598 7599
}
EXPORT_SYMBOL_GPL(sched_set_fifo);

/*
 * For when you don't much care about FIFO, but want to be above SCHED_NORMAL.
 */
7600
void sched_set_fifo_low(struct task_struct *p)
7601 7602
{
	struct sched_param sp = { .sched_priority = 1 };
7603
	WARN_ON_ONCE(sched_setscheduler_nocheck(p, SCHED_FIFO, &sp) != 0);
7604 7605 7606
}
EXPORT_SYMBOL_GPL(sched_set_fifo_low);

7607
void sched_set_normal(struct task_struct *p, int nice)
7608 7609 7610 7611 7612
{
	struct sched_attr attr = {
		.sched_policy = SCHED_NORMAL,
		.sched_nice = nice,
	};
7613
	WARN_ON_ONCE(sched_setattr_nocheck(p, &attr) != 0);
7614 7615
}
EXPORT_SYMBOL_GPL(sched_set_normal);
7616

Ingo Molnar's avatar
Ingo Molnar committed
7617 7618
static int
do_sched_setscheduler(pid_t pid, int policy, struct sched_param __user *param)
Linus Torvalds's avatar
Linus Torvalds committed
7619 7620 7621
{
	struct sched_param lparam;
	struct task_struct *p;
7622
	int retval;
Linus Torvalds's avatar
Linus Torvalds committed
7623 7624 7625 7626 7627

	if (!param || pid < 0)
		return -EINVAL;
	if (copy_from_user(&lparam, param, sizeof(struct sched_param)))
		return -EFAULT;
7628 7629 7630

	rcu_read_lock();
	retval = -ESRCH;
Linus Torvalds's avatar
Linus Torvalds committed
7631
	p = find_process_by_pid(pid);
7632 7633
	if (likely(p))
		get_task_struct(p);
7634
	rcu_read_unlock();
7635

7636 7637 7638 7639 7640
	if (likely(p)) {
		retval = sched_setscheduler(p, policy, &lparam);
		put_task_struct(p);
	}

Linus Torvalds's avatar
Linus Torvalds committed
7641 7642 7643
	return retval;
}

7644 7645 7646
/*
 * Mimics kernel/events/core.c perf_copy_attr().
 */
7647
static int sched_copy_attr(struct sched_attr __user *uattr, struct sched_attr *attr)
7648 7649 7650 7651
{
	u32 size;
	int ret;

7652
	/* Zero the full structure, so that a short copy will be nice: */
7653 7654 7655 7656 7657 7658
	memset(attr, 0, sizeof(*attr));

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

7659 7660
	/* ABI compatibility quirk: */
	if (!size)
7661
		size = SCHED_ATTR_SIZE_VER0;
7662
	if (size < SCHED_ATTR_SIZE_VER0 || size > PAGE_SIZE)
7663 7664
		goto err_size;

7665 7666 7667 7668 7669
	ret = copy_struct_from_user(attr, sizeof(*attr), uattr, size);
	if (ret) {
		if (ret == -E2BIG)
			goto err_size;
		return ret;
7670 7671
	}

7672 7673 7674 7675
	if ((attr->sched_flags & SCHED_FLAG_UTIL_CLAMP) &&
	    size < SCHED_ATTR_SIZE_VER1)
		return -EINVAL;

7676
	/*
7677
	 * XXX: Do we want to be lenient like existing syscalls; or do we want
7678 7679
	 * to be strict and return an error on out-of-bounds values?
	 */
7680
	attr->sched_nice = clamp(attr->sched_nice, MIN_NICE, MAX_NICE);
7681

7682
	return 0;
7683 7684 7685

err_size:
	put_user(sizeof(*attr), &uattr->size);
7686
	return -E2BIG;
7687 7688
}

7689 7690 7691 7692 7693 7694 7695 7696 7697 7698
static void get_params(struct task_struct *p, struct sched_attr *attr)
{
	if (task_has_dl_policy(p))
		__getparam_dl(p, attr);
	else if (task_has_rt_policy(p))
		attr->sched_priority = p->rt_priority;
	else
		attr->sched_nice = task_nice(p);
}

Linus Torvalds's avatar
Linus Torvalds committed
7699 7700 7701 7702 7703
/**
 * 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.
7704 7705
 *
 * Return: 0 on success. An error code otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
7706
 */
7707
SYSCALL_DEFINE3(sched_setscheduler, pid_t, pid, int, policy, struct sched_param __user *, param)
Linus Torvalds's avatar
Linus Torvalds committed
7708
{
7709 7710 7711
	if (policy < 0)
		return -EINVAL;

Linus Torvalds's avatar
Linus Torvalds committed
7712 7713 7714 7715 7716 7717 7718
	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.
7719 7720
 *
 * Return: 0 on success. An error code otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
7721
 */
7722
SYSCALL_DEFINE2(sched_setparam, pid_t, pid, struct sched_param __user *, param)
Linus Torvalds's avatar
Linus Torvalds committed
7723
{
7724
	return do_sched_setscheduler(pid, SETPARAM_POLICY, param);
Linus Torvalds's avatar
Linus Torvalds committed
7725 7726
}

7727 7728 7729
/**
 * sys_sched_setattr - same as above, but with extended sched_attr
 * @pid: the pid in question.
7730
 * @uattr: structure containing the extended parameters.
7731
 * @flags: for future extension.
7732
 */
7733 7734
SYSCALL_DEFINE3(sched_setattr, pid_t, pid, struct sched_attr __user *, uattr,
			       unsigned int, flags)
7735 7736 7737 7738 7739
{
	struct sched_attr attr;
	struct task_struct *p;
	int retval;

7740
	if (!uattr || pid < 0 || flags)
7741 7742
		return -EINVAL;

7743 7744 7745
	retval = sched_copy_attr(uattr, &attr);
	if (retval)
		return retval;
7746

7747
	if ((int)attr.sched_policy < 0)
7748
		return -EINVAL;
7749 7750
	if (attr.sched_flags & SCHED_FLAG_KEEP_POLICY)
		attr.sched_policy = SETPARAM_POLICY;
7751 7752 7753 7754

	rcu_read_lock();
	retval = -ESRCH;
	p = find_process_by_pid(pid);
7755 7756
	if (likely(p))
		get_task_struct(p);
7757 7758
	rcu_read_unlock();

7759
	if (likely(p)) {
7760 7761
		if (attr.sched_flags & SCHED_FLAG_KEEP_PARAMS)
			get_params(p, &attr);
7762 7763 7764 7765
		retval = sched_setattr(p, &attr);
		put_task_struct(p);
	}

7766 7767 7768
	return retval;
}

Linus Torvalds's avatar
Linus Torvalds committed
7769 7770 7771
/**
 * sys_sched_getscheduler - get the policy (scheduling class) of a thread
 * @pid: the pid in question.
7772 7773 7774
 *
 * Return: On success, the policy of the thread. Otherwise, a negative error
 * code.
Linus Torvalds's avatar
Linus Torvalds committed
7775
 */
7776
SYSCALL_DEFINE1(sched_getscheduler, pid_t, pid)
Linus Torvalds's avatar
Linus Torvalds committed
7777
{
7778
	struct task_struct *p;
7779
	int retval;
Linus Torvalds's avatar
Linus Torvalds committed
7780 7781

	if (pid < 0)
7782
		return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
7783 7784

	retval = -ESRCH;
7785
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
7786 7787 7788 7789
	p = find_process_by_pid(pid);
	if (p) {
		retval = security_task_getscheduler(p);
		if (!retval)
7790 7791
			retval = p->policy
				| (p->sched_reset_on_fork ? SCHED_RESET_ON_FORK : 0);
Linus Torvalds's avatar
Linus Torvalds committed
7792
	}
7793
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
7794 7795 7796 7797
	return retval;
}

/**
7798
 * sys_sched_getparam - get the RT priority of a thread
Linus Torvalds's avatar
Linus Torvalds committed
7799 7800
 * @pid: the pid in question.
 * @param: structure containing the RT priority.
7801 7802 7803
 *
 * Return: On success, 0 and the RT priority is in @param. Otherwise, an error
 * code.
Linus Torvalds's avatar
Linus Torvalds committed
7804
 */
7805
SYSCALL_DEFINE2(sched_getparam, pid_t, pid, struct sched_param __user *, param)
Linus Torvalds's avatar
Linus Torvalds committed
7806
{
7807
	struct sched_param lp = { .sched_priority = 0 };
7808
	struct task_struct *p;
7809
	int retval;
Linus Torvalds's avatar
Linus Torvalds committed
7810 7811

	if (!param || pid < 0)
7812
		return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
7813

7814
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
7815 7816 7817 7818 7819 7820 7821 7822 7823
	p = find_process_by_pid(pid);
	retval = -ESRCH;
	if (!p)
		goto out_unlock;

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

7824 7825
	if (task_has_rt_policy(p))
		lp.sched_priority = p->rt_priority;
7826
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
7827 7828 7829 7830 7831 7832 7833 7834 7835

	/*
	 * 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:
7836
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
7837 7838 7839
	return retval;
}

7840 7841 7842 7843 7844 7845 7846 7847 7848 7849 7850 7851
/*
 * 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)
7852
{
7853
	unsigned int ksize = sizeof(*kattr);
7854

7855
	if (!access_ok(uattr, usize))
7856 7857 7858
		return -EFAULT;

	/*
7859 7860 7861 7862 7863 7864 7865 7866 7867 7868 7869
	 * 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.
7870
	 */
7871
	kattr->size = min(usize, ksize);
7872

7873
	if (copy_to_user(uattr, kattr, kattr->size))
7874 7875
		return -EFAULT;

7876
	return 0;
7877 7878 7879
}

/**
7880
 * sys_sched_getattr - similar to sched_getparam, but with sched_attr
7881
 * @pid: the pid in question.
7882
 * @uattr: structure containing the extended parameters.
7883
 * @usize: sizeof(attr) for fwd/bwd comp.
7884
 * @flags: for future extension.
7885
 */
7886
SYSCALL_DEFINE4(sched_getattr, pid_t, pid, struct sched_attr __user *, uattr,
7887
		unsigned int, usize, unsigned int, flags)
7888
{
7889
	struct sched_attr kattr = { };
7890 7891 7892
	struct task_struct *p;
	int retval;

7893 7894
	if (!uattr || pid < 0 || usize > PAGE_SIZE ||
	    usize < SCHED_ATTR_SIZE_VER0 || flags)
7895 7896 7897 7898 7899 7900 7901 7902 7903 7904 7905 7906
		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;

7907
	kattr.sched_policy = p->policy;
7908
	if (p->sched_reset_on_fork)
7909
		kattr.sched_flags |= SCHED_FLAG_RESET_ON_FORK;
7910
	get_params(p, &kattr);
7911
	kattr.sched_flags &= SCHED_FLAG_ALL;
7912

7913
#ifdef CONFIG_UCLAMP_TASK
7914 7915 7916 7917 7918
	/*
	 * This could race with another potential updater, but this is fine
	 * because it'll correctly read the old or the new value. We don't need
	 * to guarantee who wins the race as long as it doesn't return garbage.
	 */
7919 7920
	kattr.sched_util_min = p->uclamp_req[UCLAMP_MIN].value;
	kattr.sched_util_max = p->uclamp_req[UCLAMP_MAX].value;
7921 7922
#endif

7923 7924
	rcu_read_unlock();

7925
	return sched_attr_copy_to_user(uattr, &kattr, usize);
7926 7927 7928 7929 7930 7931

out_unlock:
	rcu_read_unlock();
	return retval;
}

7932 7933
#ifdef CONFIG_SMP
int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
Linus Torvalds's avatar
Linus Torvalds committed
7934
{
7935 7936 7937 7938 7939 7940 7941 7942 7943 7944 7945 7946 7947 7948 7949 7950 7951 7952 7953 7954 7955 7956 7957
	int ret = 0;

	/*
	 * If the task isn't a deadline task or admission control is
	 * disabled then we don't care about affinity changes.
	 */
	if (!task_has_dl_policy(p) || !dl_bandwidth_enabled())
		return 0;

	/*
	 * 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.
	 */
	rcu_read_lock();
	if (!cpumask_subset(task_rq(p)->rd->span, mask))
		ret = -EBUSY;
	rcu_read_unlock();
	return ret;
}
#endif

7958 7959
static int
__sched_setaffinity(struct task_struct *p, const struct cpumask *mask)
Linus Torvalds's avatar
Linus Torvalds committed
7960
{
7961
	int retval;
7962
	cpumask_var_t cpus_allowed, new_mask;
Linus Torvalds's avatar
Linus Torvalds committed
7963

7964 7965
	if (!alloc_cpumask_var(&cpus_allowed, GFP_KERNEL))
		return -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
7966

7967 7968 7969 7970
	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL)) {
		retval = -ENOMEM;
		goto out_free_cpus_allowed;
	}
7971 7972

	cpuset_cpus_allowed(p, cpus_allowed);
7973
	cpumask_and(new_mask, mask, cpus_allowed);
7974

7975 7976 7977
	retval = dl_task_check_affinity(p, new_mask);
	if (retval)
		goto out_free_new_mask;
Peter Zijlstra's avatar
Peter Zijlstra committed
7978
again:
7979
	retval = __set_cpus_allowed_ptr(p, new_mask, SCA_CHECK | SCA_USER);
7980 7981
	if (retval)
		goto out_free_new_mask;
Linus Torvalds's avatar
Linus Torvalds committed
7982

7983 7984 7985 7986 7987 7988 7989 7990
	cpuset_cpus_allowed(p, cpus_allowed);
	if (!cpumask_subset(new_mask, cpus_allowed)) {
		/*
		 * We must have raced with a concurrent cpuset update.
		 * Just reset the cpumask to the cpuset's cpus_allowed.
		 */
		cpumask_copy(new_mask, cpus_allowed);
		goto again;
Paul Menage's avatar
Paul Menage committed
7991
	}
7992

7993
out_free_new_mask:
7994 7995 7996
	free_cpumask_var(new_mask);
out_free_cpus_allowed:
	free_cpumask_var(cpus_allowed);
7997 7998 7999 8000 8001
	return retval;
}

long sched_setaffinity(pid_t pid, const struct cpumask *in_mask)
{
8002 8003
	struct task_struct *p;
	int retval;
Linus Torvalds's avatar
Linus Torvalds committed
8004

8005
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
8006 8007 8008

	p = find_process_by_pid(pid);
	if (!p) {
8009
		rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
8010 8011 8012
		return -ESRCH;
	}

8013
	/* Prevent p going away */
Linus Torvalds's avatar
Linus Torvalds committed
8014
	get_task_struct(p);
8015
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
8016

8017 8018 8019 8020
	if (p->flags & PF_NO_SETAFFINITY) {
		retval = -EINVAL;
		goto out_put_task;
	}
8021

8022 8023 8024 8025
	if (!check_same_owner(p)) {
		rcu_read_lock();
		if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
			rcu_read_unlock();
8026 8027
			retval = -EPERM;
			goto out_put_task;
8028 8029 8030
		}
		rcu_read_unlock();
	}
Linus Torvalds's avatar
Linus Torvalds committed
8031

8032
	retval = security_task_setscheduler(p);
8033
	if (retval)
8034
		goto out_put_task;
Linus Torvalds's avatar
Linus Torvalds committed
8035

8036
	retval = __sched_setaffinity(p, in_mask);
8037
out_put_task:
Linus Torvalds's avatar
Linus Torvalds committed
8038 8039 8040 8041 8042
	put_task_struct(p);
	return retval;
}

static int get_user_cpu_mask(unsigned long __user *user_mask_ptr, unsigned len,
8043
			     struct cpumask *new_mask)
Linus Torvalds's avatar
Linus Torvalds committed
8044
{
8045 8046 8047 8048 8049
	if (len < cpumask_size())
		cpumask_clear(new_mask);
	else if (len > cpumask_size())
		len = cpumask_size();

Linus Torvalds's avatar
Linus Torvalds committed
8050 8051 8052 8053
	return copy_from_user(new_mask, user_mask_ptr, len) ? -EFAULT : 0;
}

/**
8054
 * sys_sched_setaffinity - set the CPU affinity of a process
Linus Torvalds's avatar
Linus Torvalds committed
8055 8056
 * @pid: pid of the process
 * @len: length in bytes of the bitmask pointed to by user_mask_ptr
8057
 * @user_mask_ptr: user-space pointer to the new CPU mask
8058 8059
 *
 * Return: 0 on success. An error code otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
8060
 */
8061 8062
SYSCALL_DEFINE3(sched_setaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
Linus Torvalds's avatar
Linus Torvalds committed
8063
{
8064
	cpumask_var_t new_mask;
Linus Torvalds's avatar
Linus Torvalds committed
8065 8066
	int retval;

8067 8068
	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL))
		return -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
8069

8070 8071 8072 8073 8074
	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
8075 8076
}

8077
long sched_getaffinity(pid_t pid, struct cpumask *mask)
Linus Torvalds's avatar
Linus Torvalds committed
8078
{
8079
	struct task_struct *p;
8080
	unsigned long flags;
Linus Torvalds's avatar
Linus Torvalds committed
8081 8082
	int retval;

8083
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
8084 8085 8086 8087 8088 8089

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

8090 8091 8092 8093
	retval = security_task_getscheduler(p);
	if (retval)
		goto out_unlock;

8094
	raw_spin_lock_irqsave(&p->pi_lock, flags);
8095
	cpumask_and(mask, &p->cpus_mask, cpu_active_mask);
8096
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
8097 8098

out_unlock:
8099
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
8100

8101
	return retval;
Linus Torvalds's avatar
Linus Torvalds committed
8102 8103 8104
}

/**
8105
 * sys_sched_getaffinity - get the CPU affinity of a process
Linus Torvalds's avatar
Linus Torvalds committed
8106 8107
 * @pid: pid of the process
 * @len: length in bytes of the bitmask pointed to by user_mask_ptr
8108
 * @user_mask_ptr: user-space pointer to hold the current CPU mask
8109
 *
8110 8111
 * Return: size of CPU mask copied to user_mask_ptr on success. An
 * error code otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
8112
 */
8113 8114
SYSCALL_DEFINE3(sched_getaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
Linus Torvalds's avatar
Linus Torvalds committed
8115 8116
{
	int ret;
8117
	cpumask_var_t mask;
Linus Torvalds's avatar
Linus Torvalds committed
8118

8119
	if ((len * BITS_PER_BYTE) < nr_cpu_ids)
8120 8121
		return -EINVAL;
	if (len & (sizeof(unsigned long)-1))
Linus Torvalds's avatar
Linus Torvalds committed
8122 8123
		return -EINVAL;

8124 8125
	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
		return -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
8126

8127 8128
	ret = sched_getaffinity(pid, mask);
	if (ret == 0) {
8129
		unsigned int retlen = min(len, cpumask_size());
8130 8131

		if (copy_to_user(user_mask_ptr, mask, retlen))
8132 8133
			ret = -EFAULT;
		else
8134
			ret = retlen;
8135 8136
	}
	free_cpumask_var(mask);
Linus Torvalds's avatar
Linus Torvalds committed
8137

8138
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
8139 8140
}

8141
static void do_sched_yield(void)
Linus Torvalds's avatar
Linus Torvalds committed
8142
{
8143 8144 8145
	struct rq_flags rf;
	struct rq *rq;

8146
	rq = this_rq_lock_irq(&rf);
Linus Torvalds's avatar
Linus Torvalds committed
8147

8148
	schedstat_inc(rq->yld_count);
8149
	current->sched_class->yield_task(rq);
Linus Torvalds's avatar
Linus Torvalds committed
8150

8151
	preempt_disable();
8152
	rq_unlock_irq(rq, &rf);
8153
	sched_preempt_enable_no_resched();
Linus Torvalds's avatar
Linus Torvalds committed
8154 8155

	schedule();
8156
}
Linus Torvalds's avatar
Linus Torvalds committed
8157

8158 8159 8160 8161 8162 8163 8164 8165
/**
 * sys_sched_yield - yield the current processor to other threads.
 *
 * This function yields the current CPU to other tasks. If there are no
 * other threads running on this CPU then this function will return.
 *
 * Return: 0.
 */
8166 8167 8168
SYSCALL_DEFINE0(sched_yield)
{
	do_sched_yield();
Linus Torvalds's avatar
Linus Torvalds committed
8169 8170 8171
	return 0;
}

8172 8173
#if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
int __sched __cond_resched(void)
Linus Torvalds's avatar
Linus Torvalds committed
8174
{
8175
	if (should_resched(0)) {
8176
		preempt_schedule_common();
Linus Torvalds's avatar
Linus Torvalds committed
8177 8178
		return 1;
	}
8179 8180 8181 8182 8183 8184 8185 8186 8187 8188 8189
	/*
	 * In preemptible kernels, ->rcu_read_lock_nesting tells the tick
	 * whether the current CPU is in an RCU read-side critical section,
	 * so the tick can report quiescent states even for CPUs looping
	 * in kernel context.  In contrast, in non-preemptible kernels,
	 * RCU readers leave no in-memory hints, which means that CPU-bound
	 * processes executing in kernel context might never report an
	 * RCU quiescent state.  Therefore, the following code causes
	 * cond_resched() to report a quiescent state, but only when RCU
	 * is in urgent need of one.
	 */
8190
#ifndef CONFIG_PREEMPT_RCU
8191
	rcu_all_qs();
8192
#endif
Linus Torvalds's avatar
Linus Torvalds committed
8193 8194
	return 0;
}
8195 8196 8197 8198 8199
EXPORT_SYMBOL(__cond_resched);
#endif

#ifdef CONFIG_PREEMPT_DYNAMIC
DEFINE_STATIC_CALL_RET0(cond_resched, __cond_resched);
8200
EXPORT_STATIC_CALL_TRAMP(cond_resched);
8201 8202

DEFINE_STATIC_CALL_RET0(might_resched, __cond_resched);
8203
EXPORT_STATIC_CALL_TRAMP(might_resched);
8204
#endif
Linus Torvalds's avatar
Linus Torvalds committed
8205 8206

/*
8207
 * __cond_resched_lock() - if a reschedule is pending, drop the given lock,
Linus Torvalds's avatar
Linus Torvalds committed
8208 8209
 * call schedule, and on return reacquire the lock.
 *
8210
 * This works OK both with and without CONFIG_PREEMPTION. We do strange low-level
Linus Torvalds's avatar
Linus Torvalds committed
8211 8212 8213
 * operations here to prevent schedule() from being called twice (once via
 * spin_unlock(), once by hand).
 */
8214
int __cond_resched_lock(spinlock_t *lock)
Linus Torvalds's avatar
Linus Torvalds committed
8215
{
8216
	int resched = should_resched(PREEMPT_LOCK_OFFSET);
Jan Kara's avatar
Jan Kara committed
8217 8218
	int ret = 0;

8219 8220
	lockdep_assert_held(lock);

8221
	if (spin_needbreak(lock) || resched) {
Linus Torvalds's avatar
Linus Torvalds committed
8222
		spin_unlock(lock);
8223
		if (resched)
8224
			preempt_schedule_common();
Nick Piggin's avatar
Nick Piggin committed
8225 8226
		else
			cpu_relax();
Jan Kara's avatar
Jan Kara committed
8227
		ret = 1;
Linus Torvalds's avatar
Linus Torvalds committed
8228 8229
		spin_lock(lock);
	}
Jan Kara's avatar
Jan Kara committed
8230
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
8231
}
8232
EXPORT_SYMBOL(__cond_resched_lock);
Linus Torvalds's avatar
Linus Torvalds committed
8233

8234 8235 8236 8237 8238 8239 8240 8241 8242 8243 8244 8245 8246 8247 8248 8249 8250 8251 8252 8253 8254 8255 8256 8257 8258 8259 8260 8261 8262 8263 8264 8265 8266 8267 8268 8269 8270 8271 8272 8273
int __cond_resched_rwlock_read(rwlock_t *lock)
{
	int resched = should_resched(PREEMPT_LOCK_OFFSET);
	int ret = 0;

	lockdep_assert_held_read(lock);

	if (rwlock_needbreak(lock) || resched) {
		read_unlock(lock);
		if (resched)
			preempt_schedule_common();
		else
			cpu_relax();
		ret = 1;
		read_lock(lock);
	}
	return ret;
}
EXPORT_SYMBOL(__cond_resched_rwlock_read);

int __cond_resched_rwlock_write(rwlock_t *lock)
{
	int resched = should_resched(PREEMPT_LOCK_OFFSET);
	int ret = 0;

	lockdep_assert_held_write(lock);

	if (rwlock_needbreak(lock) || resched) {
		write_unlock(lock);
		if (resched)
			preempt_schedule_common();
		else
			cpu_relax();
		ret = 1;
		write_lock(lock);
	}
	return ret;
}
EXPORT_SYMBOL(__cond_resched_rwlock_write);

Linus Torvalds's avatar
Linus Torvalds committed
8274 8275 8276
/**
 * yield - yield the current processor to other threads.
 *
Peter Zijlstra's avatar
Peter Zijlstra committed
8277 8278 8279 8280
 * 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
8281
 * it, it's already broken.
Peter Zijlstra's avatar
Peter Zijlstra committed
8282 8283 8284 8285
 *
 * Typical broken usage is:
 *
 * while (!event)
8286
 *	yield();
Peter Zijlstra's avatar
Peter Zijlstra committed
8287 8288 8289 8290 8291 8292 8293 8294
 *
 * 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
8295 8296 8297 8298
 */
void __sched yield(void)
{
	set_current_state(TASK_RUNNING);
8299
	do_sched_yield();
Linus Torvalds's avatar
Linus Torvalds committed
8300 8301 8302
}
EXPORT_SYMBOL(yield);

8303 8304 8305 8306
/**
 * yield_to - yield the current processor to another thread in
 * your thread group, or accelerate that thread toward the
 * processor it's on.
8307 8308
 * @p: target task
 * @preempt: whether task preemption is allowed or not
8309 8310 8311 8312
 *
 * 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.
 *
8313
 * Return:
8314 8315 8316
 *	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.
8317
 */
8318
int __sched yield_to(struct task_struct *p, bool preempt)
8319 8320 8321 8322
{
	struct task_struct *curr = current;
	struct rq *rq, *p_rq;
	unsigned long flags;
8323
	int yielded = 0;
8324 8325 8326 8327 8328 8329

	local_irq_save(flags);
	rq = this_rq();

again:
	p_rq = task_rq(p);
8330 8331 8332 8333 8334 8335 8336 8337 8338
	/*
	 * 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;
	}

8339
	double_rq_lock(rq, p_rq);
8340
	if (task_rq(p) != p_rq) {
8341 8342 8343 8344 8345
		double_rq_unlock(rq, p_rq);
		goto again;
	}

	if (!curr->sched_class->yield_to_task)
8346
		goto out_unlock;
8347 8348

	if (curr->sched_class != p->sched_class)
8349
		goto out_unlock;
8350

8351
	if (task_running(p_rq, p) || !task_is_running(p))
8352
		goto out_unlock;
8353

8354
	yielded = curr->sched_class->yield_to_task(rq, p);
8355
	if (yielded) {
8356
		schedstat_inc(rq->yld_count);
8357 8358 8359 8360 8361
		/*
		 * Make p's CPU reschedule; pick_next_entity takes care of
		 * fairness.
		 */
		if (preempt && rq != p_rq)
8362
			resched_curr(p_rq);
8363
	}
8364

8365
out_unlock:
8366
	double_rq_unlock(rq, p_rq);
8367
out_irq:
8368 8369
	local_irq_restore(flags);

8370
	if (yielded > 0)
8371 8372 8373 8374 8375 8376
		schedule();

	return yielded;
}
EXPORT_SYMBOL_GPL(yield_to);

8377 8378 8379 8380 8381
int io_schedule_prepare(void)
{
	int old_iowait = current->in_iowait;

	current->in_iowait = 1;
8382 8383
	if (current->plug)
		blk_flush_plug(current->plug, true);
8384 8385 8386 8387 8388 8389 8390 8391 8392

	return old_iowait;
}

void io_schedule_finish(int token)
{
	current->in_iowait = token;
}

Linus Torvalds's avatar
Linus Torvalds committed
8393
/*
Ingo Molnar's avatar
Ingo Molnar committed
8394
 * This task is about to go to sleep on IO. Increment rq->nr_iowait so
Linus Torvalds's avatar
Linus Torvalds committed
8395 8396 8397 8398
 * that process accounting knows that this is a task in IO wait state.
 */
long __sched io_schedule_timeout(long timeout)
{
8399
	int token;
Linus Torvalds's avatar
Linus Torvalds committed
8400 8401
	long ret;

8402
	token = io_schedule_prepare();
Linus Torvalds's avatar
Linus Torvalds committed
8403
	ret = schedule_timeout(timeout);
8404
	io_schedule_finish(token);
8405

Linus Torvalds's avatar
Linus Torvalds committed
8406 8407
	return ret;
}
8408
EXPORT_SYMBOL(io_schedule_timeout);
Linus Torvalds's avatar
Linus Torvalds committed
8409

8410
void __sched io_schedule(void)
8411 8412 8413 8414 8415 8416 8417 8418 8419
{
	int token;

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

Linus Torvalds's avatar
Linus Torvalds committed
8420 8421 8422 8423
/**
 * sys_sched_get_priority_max - return maximum RT priority.
 * @policy: scheduling class.
 *
8424 8425 8426
 * 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
8427
 */
8428
SYSCALL_DEFINE1(sched_get_priority_max, int, policy)
Linus Torvalds's avatar
Linus Torvalds committed
8429 8430 8431 8432 8433 8434
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
8435
		ret = MAX_RT_PRIO-1;
Linus Torvalds's avatar
Linus Torvalds committed
8436
		break;
8437
	case SCHED_DEADLINE:
Linus Torvalds's avatar
Linus Torvalds committed
8438
	case SCHED_NORMAL:
8439
	case SCHED_BATCH:
Ingo Molnar's avatar
Ingo Molnar committed
8440
	case SCHED_IDLE:
Linus Torvalds's avatar
Linus Torvalds committed
8441 8442 8443 8444 8445 8446 8447 8448 8449 8450
		ret = 0;
		break;
	}
	return ret;
}

/**
 * sys_sched_get_priority_min - return minimum RT priority.
 * @policy: scheduling class.
 *
8451 8452 8453
 * 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
8454
 */
8455
SYSCALL_DEFINE1(sched_get_priority_min, int, policy)
Linus Torvalds's avatar
Linus Torvalds committed
8456 8457 8458 8459 8460 8461 8462 8463
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = 1;
		break;
8464
	case SCHED_DEADLINE:
Linus Torvalds's avatar
Linus Torvalds committed
8465
	case SCHED_NORMAL:
8466
	case SCHED_BATCH:
Ingo Molnar's avatar
Ingo Molnar committed
8467
	case SCHED_IDLE:
Linus Torvalds's avatar
Linus Torvalds committed
8468 8469 8470 8471 8472
		ret = 0;
	}
	return ret;
}

8473
static int sched_rr_get_interval(pid_t pid, struct timespec64 *t)
Linus Torvalds's avatar
Linus Torvalds committed
8474
{
8475
	struct task_struct *p;
Dmitry Adamushko's avatar
Dmitry Adamushko committed
8476
	unsigned int time_slice;
8477
	struct rq_flags rf;
8478
	struct rq *rq;
8479
	int retval;
Linus Torvalds's avatar
Linus Torvalds committed
8480 8481

	if (pid < 0)
8482
		return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
8483 8484

	retval = -ESRCH;
8485
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
8486 8487 8488 8489 8490 8491 8492 8493
	p = find_process_by_pid(pid);
	if (!p)
		goto out_unlock;

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

8494
	rq = task_rq_lock(p, &rf);
8495 8496 8497
	time_slice = 0;
	if (p->sched_class->get_rr_interval)
		time_slice = p->sched_class->get_rr_interval(rq, p);
8498
	task_rq_unlock(rq, p, &rf);
Dmitry Adamushko's avatar
Dmitry Adamushko committed
8499

8500
	rcu_read_unlock();
8501 8502
	jiffies_to_timespec64(time_slice, t);
	return 0;
8503

Linus Torvalds's avatar
Linus Torvalds committed
8504
out_unlock:
8505
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
8506 8507 8508
	return retval;
}

8509 8510 8511 8512 8513 8514 8515 8516 8517 8518 8519
/**
 * 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.
 */
8520
SYSCALL_DEFINE2(sched_rr_get_interval, pid_t, pid,
8521
		struct __kernel_timespec __user *, interval)
8522 8523 8524 8525 8526 8527 8528 8529 8530 8531
{
	struct timespec64 t;
	int retval = sched_rr_get_interval(pid, &t);

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

	return retval;
}

8532
#ifdef CONFIG_COMPAT_32BIT_TIME
8533 8534
SYSCALL_DEFINE2(sched_rr_get_interval_time32, pid_t, pid,
		struct old_timespec32 __user *, interval)
8535 8536 8537 8538 8539
{
	struct timespec64 t;
	int retval = sched_rr_get_interval(pid, &t);

	if (retval == 0)
8540
		retval = put_old_timespec32(&t, interval);
8541 8542 8543 8544
	return retval;
}
#endif

8545
void sched_show_task(struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
8546 8547
{
	unsigned long free = 0;
8548
	int ppid;
8549

8550 8551
	if (!try_get_task_stack(p))
		return;
8552

8553
	pr_info("task:%-15.15s state:%c", p->comm, task_state_to_char(p));
8554

8555
	if (task_is_running(p))
8556
		pr_cont("  running task    ");
Linus Torvalds's avatar
Linus Torvalds committed
8557
#ifdef CONFIG_DEBUG_STACK_USAGE
8558
	free = stack_not_used(p);
Linus Torvalds's avatar
Linus Torvalds committed
8559
#endif
8560
	ppid = 0;
8561
	rcu_read_lock();
8562 8563
	if (pid_alive(p))
		ppid = task_pid_nr(rcu_dereference(p->real_parent));
8564
	rcu_read_unlock();
8565 8566
	pr_cont(" stack:%5lu pid:%5d ppid:%6d flags:0x%08lx\n",
		free, task_pid_nr(p), ppid,
8567
		read_task_thread_flags(p));
Linus Torvalds's avatar
Linus Torvalds committed
8568

8569
	print_worker_info(KERN_INFO, p);
8570
	print_stop_info(KERN_INFO, p);
8571
	show_stack(p, NULL, KERN_INFO);
8572
	put_task_stack(p);
Linus Torvalds's avatar
Linus Torvalds committed
8573
}
8574
EXPORT_SYMBOL_GPL(sched_show_task);
Linus Torvalds's avatar
Linus Torvalds committed
8575

8576 8577 8578
static inline bool
state_filter_match(unsigned long state_filter, struct task_struct *p)
{
8579 8580
	unsigned int state = READ_ONCE(p->__state);

8581 8582 8583 8584 8585
	/* no filter, everything matches */
	if (!state_filter)
		return true;

	/* filter, but doesn't match */
8586
	if (!(state & state_filter))
8587 8588 8589 8590 8591 8592
		return false;

	/*
	 * When looking for TASK_UNINTERRUPTIBLE skip TASK_IDLE (allows
	 * TASK_KILLABLE).
	 */
8593
	if (state_filter == TASK_UNINTERRUPTIBLE && state == TASK_IDLE)
8594 8595 8596 8597 8598 8599
		return false;

	return true;
}


8600
void show_state_filter(unsigned int state_filter)
Linus Torvalds's avatar
Linus Torvalds committed
8601
{
8602
	struct task_struct *g, *p;
Linus Torvalds's avatar
Linus Torvalds committed
8603

8604
	rcu_read_lock();
8605
	for_each_process_thread(g, p) {
Linus Torvalds's avatar
Linus Torvalds committed
8606 8607
		/*
		 * reset the NMI-timeout, listing all files on a slow
Lucas De Marchi's avatar
Lucas De Marchi committed
8608
		 * console might take a lot of time:
8609 8610 8611
		 * 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
8612 8613
		 */
		touch_nmi_watchdog();
8614
		touch_all_softlockup_watchdogs();
8615
		if (state_filter_match(state_filter, p))
8616
			sched_show_task(p);
8617
	}
Linus Torvalds's avatar
Linus Torvalds committed
8618

Ingo Molnar's avatar
Ingo Molnar committed
8619
#ifdef CONFIG_SCHED_DEBUG
8620 8621
	if (!state_filter)
		sysrq_sched_debug_show();
Ingo Molnar's avatar
Ingo Molnar committed
8622
#endif
8623
	rcu_read_unlock();
8624 8625 8626
	/*
	 * Only show locks if all tasks are dumped:
	 */
8627
	if (!state_filter)
8628
		debug_show_all_locks();
Linus Torvalds's avatar
Linus Torvalds committed
8629 8630
}

8631 8632 8633
/**
 * init_idle - set up an idle thread for a given CPU
 * @idle: task in question
8634
 * @cpu: CPU the idle task belongs to
8635 8636 8637 8638
 *
 * NOTE: this function does not set the idle thread's NEED_RESCHED
 * flag, to make booting more robust.
 */
8639
void __init init_idle(struct task_struct *idle, int cpu)
Linus Torvalds's avatar
Linus Torvalds committed
8640
{
8641
	struct rq *rq = cpu_rq(cpu);
Linus Torvalds's avatar
Linus Torvalds committed
8642 8643
	unsigned long flags;

8644 8645
	__sched_fork(0, idle);

8646
	raw_spin_lock_irqsave(&idle->pi_lock, flags);
8647
	raw_spin_rq_lock(rq);
8648

8649
	idle->__state = TASK_RUNNING;
Ingo Molnar's avatar
Ingo Molnar committed
8650
	idle->se.exec_start = sched_clock();
8651 8652 8653 8654 8655 8656
	/*
	 * PF_KTHREAD should already be set at this point; regardless, make it
	 * look like a proper per-CPU kthread.
	 */
	idle->flags |= PF_IDLE | PF_KTHREAD | PF_NO_SETAFFINITY;
	kthread_set_per_cpu(idle, cpu);
Ingo Molnar's avatar
Ingo Molnar committed
8657

8658 8659
#ifdef CONFIG_SMP
	/*
8660
	 * It's possible that init_idle() gets called multiple times on a task,
8661 8662 8663 8664
	 * in that case do_set_cpus_allowed() will not do the right thing.
	 *
	 * And since this is boot we can forgo the serialization.
	 */
8665
	set_cpus_allowed_common(idle, cpumask_of(cpu), 0);
8666
#endif
8667 8668
	/*
	 * We're having a chicken and egg problem, even though we are
8669
	 * holding rq->lock, the CPU isn't yet set to this CPU so the
8670 8671 8672 8673 8674 8675 8676 8677
	 * 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
8678
	__set_task_cpu(idle, cpu);
8679
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
8680

8681 8682
	rq->idle = idle;
	rcu_assign_pointer(rq->curr, idle);
8683
	idle->on_rq = TASK_ON_RQ_QUEUED;
8684
#ifdef CONFIG_SMP
8685
	idle->on_cpu = 1;
8686
#endif
8687
	raw_spin_rq_unlock(rq);
8688
	raw_spin_unlock_irqrestore(&idle->pi_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
8689 8690

	/* Set the preempt count _outside_ the spinlocks! */
8691
	init_idle_preempt_count(idle, cpu);
8692

Ingo Molnar's avatar
Ingo Molnar committed
8693 8694 8695 8696
	/*
	 * The idle tasks have their own, simple scheduling class:
	 */
	idle->sched_class = &idle_sched_class;
8697
	ftrace_graph_init_idle_task(idle, cpu);
8698
	vtime_init_idle(idle, cpu);
8699
#ifdef CONFIG_SMP
8700 8701
	sprintf(idle->comm, "%s/%d", INIT_TASK_COMM, cpu);
#endif
8702 8703
}

8704 8705
#ifdef CONFIG_SMP

8706 8707 8708
int cpuset_cpumask_can_shrink(const struct cpumask *cur,
			      const struct cpumask *trial)
{
8709
	int ret = 1;
8710

8711
	if (cpumask_empty(cur))
8712 8713
		return ret;

8714
	ret = dl_cpuset_cpumask_can_shrink(cur, trial);
8715 8716 8717 8718

	return ret;
}

8719 8720 8721 8722 8723 8724 8725
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
8726
	 * to a new cpuset; we don't want to change their CPU
8727 8728 8729 8730
	 * 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
8731
	 * before cpus_mask may be changed.
8732 8733 8734 8735 8736 8737 8738
	 */
	if (p->flags & PF_NO_SETAFFINITY) {
		ret = -EINVAL;
		goto out;
	}

	if (dl_task(p) && !cpumask_intersects(task_rq(p)->rd->span,
8739 8740
					      cs_cpus_allowed))
		ret = dl_task_can_attach(p, cs_cpus_allowed);
8741 8742 8743 8744 8745

out:
	return ret;
}

8746
bool sched_smp_initialized __read_mostly;
8747

8748 8749 8750 8751 8752 8753 8754 8755 8756 8757
#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;

8758
	if (!cpumask_test_cpu(target_cpu, p->cpus_ptr))
8759 8760 8761 8762
		return -EINVAL;

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

8763
	trace_sched_move_numa(p, curr_cpu, target_cpu);
8764 8765
	return stop_one_cpu(curr_cpu, migration_cpu_stop, &arg);
}
8766 8767 8768 8769 8770 8771 8772

/*
 * 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)
{
8773
	bool queued, running;
8774 8775
	struct rq_flags rf;
	struct rq *rq;
8776

8777
	rq = task_rq_lock(p, &rf);
8778
	queued = task_on_rq_queued(p);
8779 8780
	running = task_current(rq, p);

8781
	if (queued)
8782
		dequeue_task(rq, p, DEQUEUE_SAVE);
8783
	if (running)
8784
		put_prev_task(rq, p);
8785 8786 8787

	p->numa_preferred_nid = nid;

8788
	if (queued)
8789
		enqueue_task(rq, p, ENQUEUE_RESTORE | ENQUEUE_NOCLOCK);
8790
	if (running)
8791
		set_next_task(rq, p);
8792
	task_rq_unlock(rq, p, &rf);
8793
}
Peter Zijlstra's avatar
Peter Zijlstra committed
8794
#endif /* CONFIG_NUMA_BALANCING */
8795

Linus Torvalds's avatar
Linus Torvalds committed
8796
#ifdef CONFIG_HOTPLUG_CPU
8797
/*
8798
 * Ensure that the idle task is using init_mm right before its CPU goes
8799
 * offline.
8800
 */
8801
void idle_task_exit(void)
Linus Torvalds's avatar
Linus Torvalds committed
8802
{
8803
	struct mm_struct *mm = current->active_mm;
8804

8805
	BUG_ON(cpu_online(smp_processor_id()));
8806
	BUG_ON(current != this_rq()->idle);
8807

8808
	if (mm != &init_mm) {
8809
		switch_mm(mm, &init_mm, current);
8810 8811
		finish_arch_post_lock_switch();
	}
8812 8813

	/* finish_cpu(), as ran on the BP, will clean up the active_mm state */
Linus Torvalds's avatar
Linus Torvalds committed
8814 8815
}

8816
static int __balance_push_cpu_stop(void *arg)
Linus Torvalds's avatar
Linus Torvalds committed
8817
{
8818 8819 8820 8821
	struct task_struct *p = arg;
	struct rq *rq = this_rq();
	struct rq_flags rf;
	int cpu;
Linus Torvalds's avatar
Linus Torvalds committed
8822

8823 8824
	raw_spin_lock_irq(&p->pi_lock);
	rq_lock(rq, &rf);
8825

8826 8827 8828 8829 8830
	update_rq_clock(rq);

	if (task_rq(p) == rq && task_on_rq_queued(p)) {
		cpu = select_fallback_rq(rq->cpu, p);
		rq = __migrate_task(rq, &rf, p, cpu);
8831
	}
8832

8833 8834 8835 8836 8837 8838
	rq_unlock(rq, &rf);
	raw_spin_unlock_irq(&p->pi_lock);

	put_task_struct(p);

	return 0;
8839
}
8840

8841 8842
static DEFINE_PER_CPU(struct cpu_stop_work, push_work);

8843
/*
8844
 * Ensure we only run per-cpu kthreads once the CPU goes !active.
8845 8846 8847
 *
 * This is enabled below SCHED_AP_ACTIVE; when !cpu_active(), but only
 * effective when the hotplug motion is down.
Linus Torvalds's avatar
Linus Torvalds committed
8848
 */
8849
static void balance_push(struct rq *rq)
Linus Torvalds's avatar
Linus Torvalds committed
8850
{
8851 8852
	struct task_struct *push_task = rq->curr;

8853
	lockdep_assert_rq_held(rq);
8854

8855 8856 8857 8858
	/*
	 * Ensure the thing is persistent until balance_push_set(.on = false);
	 */
	rq->balance_callback = &balance_push_callback;
Linus Torvalds's avatar
Linus Torvalds committed
8859

8860
	/*
8861 8862
	 * Only active while going offline and when invoked on the outgoing
	 * CPU.
8863
	 */
8864
	if (!cpu_dying(rq->cpu) || rq != this_rq())
8865 8866
		return;

Linus Torvalds's avatar
Linus Torvalds committed
8867
	/*
8868 8869
	 * Both the cpu-hotplug and stop task are in this case and are
	 * required to complete the hotplug process.
Linus Torvalds's avatar
Linus Torvalds committed
8870
	 */
8871
	if (kthread_is_per_cpu(push_task) ||
8872 8873
	    is_migration_disabled(push_task)) {

8874 8875 8876 8877 8878 8879
		/*
		 * If this is the idle task on the outgoing CPU try to wake
		 * up the hotplug control thread which might wait for the
		 * last task to vanish. The rcuwait_active() check is
		 * accurate here because the waiter is pinned on this CPU
		 * and can't obviously be running in parallel.
8880 8881 8882 8883
		 *
		 * On RT kernels this also has to check whether there are
		 * pinned and scheduled out tasks on the runqueue. They
		 * need to leave the migrate disabled section first.
8884
		 */
8885 8886
		if (!rq->nr_running && !rq_has_pinned_tasks(rq) &&
		    rcuwait_active(&rq->hotplug_wait)) {
8887
			raw_spin_rq_unlock(rq);
8888
			rcuwait_wake_up(&rq->hotplug_wait);
8889
			raw_spin_rq_lock(rq);
8890
		}
8891
		return;
8892
	}
8893

8894
	get_task_struct(push_task);
8895
	/*
8896 8897
	 * Temporarily drop rq->lock such that we can wake-up the stop task.
	 * Both preemption and IRQs are still disabled.
8898
	 */
8899
	raw_spin_rq_unlock(rq);
8900 8901 8902 8903 8904
	stop_one_cpu_nowait(rq->cpu, __balance_push_cpu_stop, push_task,
			    this_cpu_ptr(&push_work));
	/*
	 * At this point need_resched() is true and we'll take the loop in
	 * schedule(). The next pick is obviously going to be the stop task
8905
	 * which kthread_is_per_cpu() and will push this task away.
8906
	 */
8907
	raw_spin_rq_lock(rq);
8908
}
8909

8910 8911 8912 8913
static void balance_push_set(int cpu, bool on)
{
	struct rq *rq = cpu_rq(cpu);
	struct rq_flags rf;
8914

8915
	rq_lock_irqsave(rq, &rf);
8916 8917
	if (on) {
		WARN_ON_ONCE(rq->balance_callback);
8918
		rq->balance_callback = &balance_push_callback;
8919
	} else if (rq->balance_callback == &balance_push_callback) {
8920
		rq->balance_callback = NULL;
8921
	}
8922 8923
	rq_unlock_irqrestore(rq, &rf);
}
8924

8925 8926 8927 8928 8929 8930 8931 8932 8933
/*
 * Invoked from a CPUs hotplug control thread after the CPU has been marked
 * inactive. All tasks which are not per CPU kernel threads are either
 * pushed off this CPU now via balance_push() or placed on a different CPU
 * during wakeup. Wait until the CPU is quiescent.
 */
static void balance_hotplug_wait(void)
{
	struct rq *rq = this_rq();
8934

8935 8936
	rcuwait_wait_event(&rq->hotplug_wait,
			   rq->nr_running == 1 && !rq_has_pinned_tasks(rq),
8937 8938
			   TASK_UNINTERRUPTIBLE);
}
8939

8940
#else
8941

8942 8943
static inline void balance_push(struct rq *rq)
{
8944 8945
}

8946 8947 8948 8949
static inline void balance_push_set(int cpu, bool on)
{
}

8950 8951
static inline void balance_hotplug_wait(void)
{
8952
}
8953

Linus Torvalds's avatar
Linus Torvalds committed
8954 8955
#endif /* CONFIG_HOTPLUG_CPU */

8956
void set_rq_online(struct rq *rq)
8957 8958 8959 8960
{
	if (!rq->online) {
		const struct sched_class *class;

8961
		cpumask_set_cpu(rq->cpu, rq->rd->online);
8962 8963 8964 8965 8966 8967 8968 8969 8970
		rq->online = 1;

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

8971
void set_rq_offline(struct rq *rq)
8972 8973 8974 8975 8976 8977 8978 8979 8980
{
	if (rq->online) {
		const struct sched_class *class;

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

8981
		cpumask_clear_cpu(rq->cpu, rq->rd->online);
8982 8983 8984 8985
		rq->online = 0;
	}
}

8986 8987 8988 8989
/*
 * used to mark begin/end of suspend/resume:
 */
static int num_cpus_frozen;
8990

Linus Torvalds's avatar
Linus Torvalds committed
8991
/*
8992 8993 8994
 * Update cpusets according to cpu_active mask.  If cpusets are
 * disabled, cpuset_update_active_cpus() becomes a simple wrapper
 * around partition_sched_domains().
8995 8996 8997
 *
 * 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
8998
 */
8999
static void cpuset_cpu_active(void)
9000
{
9001
	if (cpuhp_tasks_frozen) {
9002 9003 9004 9005 9006 9007
		/*
		 * 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.
		 */
9008 9009
		partition_sched_domains(1, NULL, NULL);
		if (--num_cpus_frozen)
9010
			return;
9011 9012 9013 9014 9015
		/*
		 * This is the last CPU online operation. So fall through and
		 * restore the original sched domains by considering the
		 * cpuset configurations.
		 */
9016
		cpuset_force_rebuild();
9017
	}
9018
	cpuset_update_active_cpus();
9019
}
9020

9021
static int cpuset_cpu_inactive(unsigned int cpu)
9022
{
9023
	if (!cpuhp_tasks_frozen) {
9024
		if (dl_cpu_busy(cpu))
9025
			return -EBUSY;
9026
		cpuset_update_active_cpus();
9027
	} else {
9028 9029
		num_cpus_frozen++;
		partition_sched_domains(1, NULL, NULL);
9030
	}
9031
	return 0;
9032 9033
}

9034
int sched_cpu_activate(unsigned int cpu)
9035
{
9036
	struct rq *rq = cpu_rq(cpu);
9037
	struct rq_flags rf;
9038

9039
	/*
9040 9041
	 * Clear the balance_push callback and prepare to schedule
	 * regular tasks.
9042
	 */
9043 9044
	balance_push_set(cpu, false);

9045 9046
#ifdef CONFIG_SCHED_SMT
	/*
9047
	 * When going up, increment the number of cores with SMT present.
9048
	 */
9049 9050
	if (cpumask_weight(cpu_smt_mask(cpu)) == 2)
		static_branch_inc_cpuslocked(&sched_smt_present);
9051
#endif
9052
	set_cpu_active(cpu, true);
9053

9054
	if (sched_smp_initialized) {
9055
		sched_update_numa(cpu, true);
9056
		sched_domains_numa_masks_set(cpu);
9057
		cpuset_cpu_active();
9058
	}
9059 9060 9061 9062 9063

	/*
	 * Put the rq online, if not already. This happens:
	 *
	 * 1) In the early boot process, because we build the real domains
9064
	 *    after all CPUs have been brought up.
9065 9066 9067 9068
	 *
	 * 2) At runtime, if cpuset_cpu_active() fails to rebuild the
	 *    domains.
	 */
9069
	rq_lock_irqsave(rq, &rf);
9070 9071 9072 9073
	if (rq->rd) {
		BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
		set_rq_online(rq);
	}
9074
	rq_unlock_irqrestore(rq, &rf);
9075

9076
	return 0;
9077 9078
}

9079
int sched_cpu_deactivate(unsigned int cpu)
9080
{
9081 9082
	struct rq *rq = cpu_rq(cpu);
	struct rq_flags rf;
9083 9084
	int ret;

9085 9086 9087 9088 9089 9090
	/*
	 * Remove CPU from nohz.idle_cpus_mask to prevent participating in
	 * load balancing when not active
	 */
	nohz_balance_exit_idle(rq);

9091
	set_cpu_active(cpu, false);
9092 9093 9094 9095 9096 9097 9098

	/*
	 * From this point forward, this CPU will refuse to run any task that
	 * is not: migrate_disable() or KTHREAD_IS_PER_CPU, and will actively
	 * push those tasks away until this gets cleared, see
	 * sched_cpu_dying().
	 */
9099 9100
	balance_push_set(cpu, true);

9101
	/*
9102 9103 9104
	 * We've cleared cpu_active_mask / set balance_push, wait for all
	 * preempt-disabled and RCU users of this state to go away such that
	 * all new such users will observe it.
9105
	 *
9106 9107 9108
	 * Specifically, we rely on ttwu to no longer target this CPU, see
	 * ttwu_queue_cond() and is_cpu_allowed().
	 *
9109 9110
	 * Do sync before park smpboot threads to take care the rcu boost case.
	 */
9111
	synchronize_rcu();
9112

9113 9114 9115 9116 9117 9118 9119 9120
	rq_lock_irqsave(rq, &rf);
	if (rq->rd) {
		update_rq_clock(rq);
		BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
		set_rq_offline(rq);
	}
	rq_unlock_irqrestore(rq, &rf);

9121 9122 9123 9124 9125 9126
#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);
9127 9128

	sched_core_cpu_deactivate(cpu);
9129 9130
#endif

9131 9132 9133
	if (!sched_smp_initialized)
		return 0;

9134
	sched_update_numa(cpu, false);
9135 9136
	ret = cpuset_cpu_inactive(cpu);
	if (ret) {
9137
		balance_push_set(cpu, false);
9138
		set_cpu_active(cpu, true);
9139
		sched_update_numa(cpu, true);
9140
		return ret;
9141
	}
9142 9143
	sched_domains_numa_masks_clear(cpu);
	return 0;
9144 9145
}

9146 9147 9148 9149 9150 9151 9152 9153
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();
}

9154 9155
int sched_cpu_starting(unsigned int cpu)
{
Peter Zijlstra's avatar
Peter Zijlstra committed
9156
	sched_core_cpu_starting(cpu);
9157
	sched_rq_cpu_starting(cpu);
9158
	sched_tick_start(cpu);
9159
	return 0;
9160 9161
}

9162
#ifdef CONFIG_HOTPLUG_CPU
9163 9164 9165 9166 9167 9168 9169 9170 9171 9172 9173 9174 9175 9176 9177 9178 9179 9180 9181 9182 9183 9184 9185 9186 9187 9188 9189 9190 9191 9192 9193 9194 9195 9196 9197

/*
 * Invoked immediately before the stopper thread is invoked to bring the
 * CPU down completely. At this point all per CPU kthreads except the
 * hotplug thread (current) and the stopper thread (inactive) have been
 * either parked or have been unbound from the outgoing CPU. Ensure that
 * any of those which might be on the way out are gone.
 *
 * If after this point a bound task is being woken on this CPU then the
 * responsible hotplug callback has failed to do it's job.
 * sched_cpu_dying() will catch it with the appropriate fireworks.
 */
int sched_cpu_wait_empty(unsigned int cpu)
{
	balance_hotplug_wait();
	return 0;
}

/*
 * Since this CPU is going 'away' for a while, fold any nr_active delta we
 * might have. Called from the CPU stopper task after ensuring that the
 * stopper is the last running task on the CPU, so 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.
 *
 * Also see the comment "Global load-average calculations".
 */
static void calc_load_migrate(struct rq *rq)
{
	long delta = calc_load_fold_active(rq, 1);

	if (delta)
		atomic_long_add(delta, &calc_load_tasks);
}

9198 9199 9200 9201 9202
static void dump_rq_tasks(struct rq *rq, const char *loglvl)
{
	struct task_struct *g, *p;
	int cpu = cpu_of(rq);

9203
	lockdep_assert_rq_held(rq);
9204 9205 9206 9207 9208 9209 9210 9211 9212 9213 9214 9215 9216

	printk("%sCPU%d enqueued tasks (%u total):\n", loglvl, cpu, rq->nr_running);
	for_each_process_thread(g, p) {
		if (task_cpu(p) != cpu)
			continue;

		if (!task_on_rq_queued(p))
			continue;

		printk("%s\tpid: %d, name: %s\n", loglvl, p->pid, p->comm);
	}
}

9217 9218 9219
int sched_cpu_dying(unsigned int cpu)
{
	struct rq *rq = cpu_rq(cpu);
9220
	struct rq_flags rf;
9221 9222

	/* Handle pending wakeups and then migrate everything off */
9223
	sched_tick_stop(cpu);
9224 9225

	rq_lock_irqsave(rq, &rf);
9226 9227 9228 9229
	if (rq->nr_running != 1 || rq_has_pinned_tasks(rq)) {
		WARN(true, "Dying CPU not properly vacated!");
		dump_rq_tasks(rq, KERN_WARNING);
	}
9230 9231
	rq_unlock_irqrestore(rq, &rf);

9232 9233
	calc_load_migrate(rq);
	update_max_interval();
9234
	hrtick_clear(rq);
9235
	sched_core_cpu_dying(cpu);
9236 9237 9238 9239
	return 0;
}
#endif

Linus Torvalds's avatar
Linus Torvalds committed
9240 9241
void __init sched_init_smp(void)
{
9242
	sched_init_numa(NUMA_NO_NODE);
9243

9244 9245
	/*
	 * There's no userspace yet to cause hotplug operations; hence all the
9246
	 * CPU masks are stable and all blatant races in the below code cannot
9247
	 * happen.
9248
	 */
9249
	mutex_lock(&sched_domains_mutex);
9250
	sched_init_domains(cpu_active_mask);
9251
	mutex_unlock(&sched_domains_mutex);
9252

9253
	/* Move init over to a non-isolated CPU */
9254
	if (set_cpus_allowed_ptr(current, housekeeping_cpumask(HK_TYPE_DOMAIN)) < 0)
9255
		BUG();
9256
	current->flags &= ~PF_NO_SETAFFINITY;
9257
	sched_init_granularity();
9258

9259
	init_sched_rt_class();
9260
	init_sched_dl_class();
9261

9262
	sched_smp_initialized = true;
Linus Torvalds's avatar
Linus Torvalds committed
9263
}
9264 9265 9266

static int __init migration_init(void)
{
9267
	sched_cpu_starting(smp_processor_id());
9268
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
9269
}
9270 9271
early_initcall(migration_init);

Linus Torvalds's avatar
Linus Torvalds committed
9272 9273 9274
#else
void __init sched_init_smp(void)
{
9275
	sched_init_granularity();
Linus Torvalds's avatar
Linus Torvalds committed
9276 9277 9278 9279 9280 9281 9282 9283 9284 9285
}
#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);
}

9286
#ifdef CONFIG_CGROUP_SCHED
9287 9288 9289 9290
/*
 * Default task group.
 * Every task in system belongs to this group at bootup.
 */
9291
struct task_group root_task_group;
9292
LIST_HEAD(task_groups);
9293 9294 9295

/* Cacheline aligned slab cache for task_group */
static struct kmem_cache *task_group_cache __read_mostly;
9296
#endif
Peter Zijlstra's avatar
Peter Zijlstra committed
9297

9298
DECLARE_PER_CPU(cpumask_var_t, load_balance_mask);
9299
DECLARE_PER_CPU(cpumask_var_t, select_idle_mask);
Peter Zijlstra's avatar
Peter Zijlstra committed
9300

Linus Torvalds's avatar
Linus Torvalds committed
9301 9302
void __init sched_init(void)
{
9303
	unsigned long ptr = 0;
9304
	int i;
9305

9306 9307 9308 9309 9310 9311 9312 9313
	/* Make sure the linker didn't screw up */
	BUG_ON(&idle_sched_class + 1 != &fair_sched_class ||
	       &fair_sched_class + 1 != &rt_sched_class ||
	       &rt_sched_class + 1   != &dl_sched_class);
#ifdef CONFIG_SMP
	BUG_ON(&dl_sched_class + 1 != &stop_sched_class);
#endif

9314
	wait_bit_init();
9315

9316
#ifdef CONFIG_FAIR_GROUP_SCHED
9317
	ptr += 2 * nr_cpu_ids * sizeof(void **);
9318 9319
#endif
#ifdef CONFIG_RT_GROUP_SCHED
9320
	ptr += 2 * nr_cpu_ids * sizeof(void **);
9321
#endif
9322 9323
	if (ptr) {
		ptr = (unsigned long)kzalloc(ptr, GFP_NOWAIT);
9324 9325

#ifdef CONFIG_FAIR_GROUP_SCHED
9326
		root_task_group.se = (struct sched_entity **)ptr;
9327 9328
		ptr += nr_cpu_ids * sizeof(void **);

9329
		root_task_group.cfs_rq = (struct cfs_rq **)ptr;
9330
		ptr += nr_cpu_ids * sizeof(void **);
9331

9332 9333
		root_task_group.shares = ROOT_TASK_GROUP_LOAD;
		init_cfs_bandwidth(&root_task_group.cfs_bandwidth);
9334
#endif /* CONFIG_FAIR_GROUP_SCHED */
9335
#ifdef CONFIG_RT_GROUP_SCHED
9336
		root_task_group.rt_se = (struct sched_rt_entity **)ptr;
9337 9338
		ptr += nr_cpu_ids * sizeof(void **);

9339
		root_task_group.rt_rq = (struct rt_rq **)ptr;
9340 9341
		ptr += nr_cpu_ids * sizeof(void **);

9342
#endif /* CONFIG_RT_GROUP_SCHED */
9343
	}
9344
#ifdef CONFIG_CPUMASK_OFFSTACK
9345 9346 9347
	for_each_possible_cpu(i) {
		per_cpu(load_balance_mask, i) = (cpumask_var_t)kzalloc_node(
			cpumask_size(), GFP_KERNEL, cpu_to_node(i));
9348 9349
		per_cpu(select_idle_mask, i) = (cpumask_var_t)kzalloc_node(
			cpumask_size(), GFP_KERNEL, cpu_to_node(i));
9350
	}
9351
#endif /* CONFIG_CPUMASK_OFFSTACK */
Ingo Molnar's avatar
Ingo Molnar committed
9352

9353 9354
	init_rt_bandwidth(&def_rt_bandwidth, global_rt_period(), global_rt_runtime());
	init_dl_bandwidth(&def_dl_bandwidth, global_rt_period(), global_rt_runtime());
9355

9356 9357 9358 9359
#ifdef CONFIG_SMP
	init_defrootdomain();
#endif

9360
#ifdef CONFIG_RT_GROUP_SCHED
9361
	init_rt_bandwidth(&root_task_group.rt_bandwidth,
9362
			global_rt_period(), global_rt_runtime());
9363
#endif /* CONFIG_RT_GROUP_SCHED */
9364

Dhaval Giani's avatar
Dhaval Giani committed
9365
#ifdef CONFIG_CGROUP_SCHED
9366 9367
	task_group_cache = KMEM_CACHE(task_group, 0);

9368 9369
	list_add(&root_task_group.list, &task_groups);
	INIT_LIST_HEAD(&root_task_group.children);
9370
	INIT_LIST_HEAD(&root_task_group.siblings);
9371
	autogroup_init(&init_task);
Dhaval Giani's avatar
Dhaval Giani committed
9372
#endif /* CONFIG_CGROUP_SCHED */
Peter Zijlstra's avatar
Peter Zijlstra committed
9373

9374
	for_each_possible_cpu(i) {
9375
		struct rq *rq;
Linus Torvalds's avatar
Linus Torvalds committed
9376 9377

		rq = cpu_rq(i);
9378
		raw_spin_lock_init(&rq->__lock);
Nick Piggin's avatar
Nick Piggin committed
9379
		rq->nr_running = 0;
9380 9381
		rq->calc_load_active = 0;
		rq->calc_load_update = jiffies + LOAD_FREQ;
9382
		init_cfs_rq(&rq->cfs);
9383 9384
		init_rt_rq(&rq->rt);
		init_dl_rq(&rq->dl);
Ingo Molnar's avatar
Ingo Molnar committed
9385
#ifdef CONFIG_FAIR_GROUP_SCHED
Peter Zijlstra's avatar
Peter Zijlstra committed
9386
		INIT_LIST_HEAD(&rq->leaf_cfs_rq_list);
9387
		rq->tmp_alone_branch = &rq->leaf_cfs_rq_list;
Dhaval Giani's avatar
Dhaval Giani committed
9388
		/*
9389
		 * How much CPU bandwidth does root_task_group get?
Dhaval Giani's avatar
Dhaval Giani committed
9390 9391
		 *
		 * In case of task-groups formed thr' the cgroup filesystem, it
9392 9393
		 * gets 100% of the CPU resources in the system. This overall
		 * system CPU resource is divided among the tasks of
9394
		 * root_task_group and its child task-groups in a fair manner,
Dhaval Giani's avatar
Dhaval Giani committed
9395 9396 9397
		 * based on each entity's (task or task-group's) weight
		 * (se->load.weight).
		 *
9398
		 * In other words, if root_task_group has 10 tasks of weight
Dhaval Giani's avatar
Dhaval Giani committed
9399
		 * 1024) and two child groups A0 and A1 (of weight 1024 each),
9400
		 * then A0's share of the CPU resource is:
Dhaval Giani's avatar
Dhaval Giani committed
9401
		 *
9402
		 *	A0's bandwidth = 1024 / (10*1024 + 1024 + 1024) = 8.33%
Dhaval Giani's avatar
Dhaval Giani committed
9403
		 *
9404 9405
		 * 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
9406
		 */
9407
		init_tg_cfs_entry(&root_task_group, &rq->cfs, NULL, i, NULL);
Dhaval Giani's avatar
Dhaval Giani committed
9408 9409 9410
#endif /* CONFIG_FAIR_GROUP_SCHED */

		rq->rt.rt_runtime = def_rt_bandwidth.rt_runtime;
9411
#ifdef CONFIG_RT_GROUP_SCHED
9412
		init_tg_rt_entry(&root_task_group, &rq->rt, NULL, i, NULL);
Ingo Molnar's avatar
Ingo Molnar committed
9413
#endif
Linus Torvalds's avatar
Linus Torvalds committed
9414
#ifdef CONFIG_SMP
Nick Piggin's avatar
Nick Piggin committed
9415
		rq->sd = NULL;
9416
		rq->rd = NULL;
9417
		rq->cpu_capacity = rq->cpu_capacity_orig = SCHED_CAPACITY_SCALE;
9418
		rq->balance_callback = &balance_push_callback;
Linus Torvalds's avatar
Linus Torvalds committed
9419
		rq->active_balance = 0;
Ingo Molnar's avatar
Ingo Molnar committed
9420
		rq->next_balance = jiffies;
Linus Torvalds's avatar
Linus Torvalds committed
9421
		rq->push_cpu = 0;
9422
		rq->cpu = i;
9423
		rq->online = 0;
9424 9425
		rq->idle_stamp = 0;
		rq->avg_idle = 2*sysctl_sched_migration_cost;
9426 9427
		rq->wake_stamp = jiffies;
		rq->wake_avg_idle = rq->avg_idle;
9428
		rq->max_idle_balance_cost = sysctl_sched_migration_cost;
9429 9430 9431

		INIT_LIST_HEAD(&rq->cfs_tasks);

9432
		rq_attach_root(rq, &def_root_domain);
9433
#ifdef CONFIG_NO_HZ_COMMON
9434
		rq->last_blocked_load_update_tick = jiffies;
9435
		atomic_set(&rq->nohz_flags, 0);
9436

9437
		INIT_CSD(&rq->nohz_csd, nohz_csd_func, rq);
9438
#endif
9439 9440
#ifdef CONFIG_HOTPLUG_CPU
		rcuwait_init(&rq->hotplug_wait);
9441
#endif
9442
#endif /* CONFIG_SMP */
9443
		hrtick_rq_init(rq);
Linus Torvalds's avatar
Linus Torvalds committed
9444
		atomic_set(&rq->nr_iowait, 0);
Peter Zijlstra's avatar
Peter Zijlstra committed
9445 9446

#ifdef CONFIG_SCHED_CORE
9447
		rq->core = rq;
9448
		rq->core_pick = NULL;
Peter Zijlstra's avatar
Peter Zijlstra committed
9449
		rq->core_enabled = 0;
9450
		rq->core_tree = RB_ROOT;
9451 9452 9453
		rq->core_forceidle_count = 0;
		rq->core_forceidle_occupation = 0;
		rq->core_forceidle_start = 0;
9454 9455

		rq->core_cookie = 0UL;
Peter Zijlstra's avatar
Peter Zijlstra committed
9456
#endif
Linus Torvalds's avatar
Linus Torvalds committed
9457 9458
	}

9459
	set_load_weight(&init_task, false);
9460

Linus Torvalds's avatar
Linus Torvalds committed
9461 9462 9463
	/*
	 * The boot idle thread does lazy MMU switching as well:
	 */
9464
	mmgrab(&init_mm);
Linus Torvalds's avatar
Linus Torvalds committed
9465 9466
	enter_lazy_tlb(&init_mm, current);

9467 9468 9469 9470 9471 9472
	/*
	 * The idle task doesn't need the kthread struct to function, but it
	 * is dressed up as a per-CPU kthread and thus needs to play the part
	 * if we want to avoid special-casing it in code that deals with per-CPU
	 * kthreads.
	 */
9473
	WARN_ON(!set_kthread_struct(current));
9474

Linus Torvalds's avatar
Linus Torvalds committed
9475 9476 9477 9478 9479 9480 9481
	/*
	 * 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());
9482 9483 9484

	calc_load_update = jiffies + LOAD_FREQ;

9485
#ifdef CONFIG_SMP
9486
	idle_thread_set_boot_cpu();
9487
	balance_push_set(smp_processor_id(), false);
9488 9489
#endif
	init_sched_fair_class();
9490

9491 9492
	psi_init();

9493 9494
	init_uclamp();

9495 9496
	preempt_dynamic_init();

9497
	scheduler_running = 1;
Linus Torvalds's avatar
Linus Torvalds committed
9498 9499
}

9500
#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
9501

9502
void __might_sleep(const char *file, int line)
Linus Torvalds's avatar
Linus Torvalds committed
9503
{
9504
	unsigned int state = get_current_state();
Peter Zijlstra's avatar
Peter Zijlstra committed
9505 9506 9507 9508 9509
	/*
	 * 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.
	 */
9510
	WARN_ONCE(state != TASK_RUNNING && current->task_state_change,
Peter Zijlstra's avatar
Peter Zijlstra committed
9511
			"do not call blocking ops when !TASK_RUNNING; "
9512
			"state=%x set at [<%p>] %pS\n", state,
Peter Zijlstra's avatar
Peter Zijlstra committed
9513
			(void *)current->task_state_change,
9514
			(void *)current->task_state_change);
Peter Zijlstra's avatar
Peter Zijlstra committed
9515

9516
	__might_resched(file, line, 0);
9517 9518 9519
}
EXPORT_SYMBOL(__might_sleep);

9520 9521 9522 9523 9524 9525 9526 9527 9528 9529 9530 9531
static void print_preempt_disable_ip(int preempt_offset, unsigned long ip)
{
	if (!IS_ENABLED(CONFIG_DEBUG_PREEMPT))
		return;

	if (preempt_count() == preempt_offset)
		return;

	pr_err("Preemption disabled at:");
	print_ip_sym(KERN_ERR, ip);
}

9532 9533 9534 9535 9536 9537 9538 9539 9540 9541
static inline bool resched_offsets_ok(unsigned int offsets)
{
	unsigned int nested = preempt_count();

	nested += rcu_preempt_depth() << MIGHT_RESCHED_RCU_SHIFT;

	return nested == offsets;
}

void __might_resched(const char *file, int line, unsigned int offsets)
Linus Torvalds's avatar
Linus Torvalds committed
9542
{
9543 9544 9545
	/* Ratelimiting timestamp: */
	static unsigned long prev_jiffy;

9546
	unsigned long preempt_disable_ip;
Linus Torvalds's avatar
Linus Torvalds committed
9547

9548 9549 9550
	/* WARN_ON_ONCE() by default, no rate limit required: */
	rcu_sleep_check();

9551
	if ((resched_offsets_ok(offsets) && !irqs_disabled() &&
9552
	     !is_idle_task(current) && !current->non_block_count) ||
9553 9554
	    system_state == SYSTEM_BOOTING || system_state > SYSTEM_RUNNING ||
	    oops_in_progress)
9555
		return;
9556

9557 9558 9559 9560
	if (time_before(jiffies, prev_jiffy + HZ) && prev_jiffy)
		return;
	prev_jiffy = jiffies;

9561
	/* Save this before calling printk(), since that will clobber it: */
9562 9563
	preempt_disable_ip = get_preempt_disable_ip(current);

9564 9565 9566 9567 9568
	pr_err("BUG: sleeping function called from invalid context at %s:%d\n",
	       file, line);
	pr_err("in_atomic(): %d, irqs_disabled(): %d, non_block: %d, pid: %d, name: %s\n",
	       in_atomic(), irqs_disabled(), current->non_block_count,
	       current->pid, current->comm);
9569
	pr_err("preempt_count: %x, expected: %x\n", preempt_count(),
9570
	       offsets & MIGHT_RESCHED_PREEMPT_MASK);
9571 9572

	if (IS_ENABLED(CONFIG_PREEMPT_RCU)) {
9573 9574
		pr_err("RCU nest depth: %d, expected: %u\n",
		       rcu_preempt_depth(), offsets >> MIGHT_RESCHED_RCU_SHIFT);
9575
	}
9576

9577
	if (task_stack_end_corrupted(current))
9578
		pr_emerg("Thread overran stack, or stack corrupted\n");
9579

9580 9581 9582
	debug_show_held_locks(current);
	if (irqs_disabled())
		print_irqtrace_events(current);
9583

9584 9585
	print_preempt_disable_ip(offsets & MIGHT_RESCHED_PREEMPT_MASK,
				 preempt_disable_ip);
9586

9587
	dump_stack();
9588
	add_taint(TAINT_WARN, LOCKDEP_STILL_OK);
Linus Torvalds's avatar
Linus Torvalds committed
9589
}
9590
EXPORT_SYMBOL(__might_resched);
9591 9592 9593 9594 9595 9596 9597 9598 9599 9600 9601 9602 9603 9604 9605 9606 9607 9608 9609 9610 9611 9612 9613 9614 9615 9616 9617 9618

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);
9619 9620 9621 9622 9623 9624 9625 9626 9627 9628 9629 9630 9631 9632 9633 9634 9635 9636 9637 9638 9639 9640 9641 9642 9643 9644 9645 9646 9647 9648 9649 9650 9651

#ifdef CONFIG_SMP
void __cant_migrate(const char *file, int line)
{
	static unsigned long prev_jiffy;

	if (irqs_disabled())
		return;

	if (is_migration_disabled(current))
		return;

	if (!IS_ENABLED(CONFIG_PREEMPT_COUNT))
		return;

	if (preempt_count() > 0)
		return;

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

	pr_err("BUG: assuming non migratable context at %s:%d\n", file, line);
	pr_err("in_atomic(): %d, irqs_disabled(): %d, migration_disabled() %u pid: %d, name: %s\n",
	       in_atomic(), irqs_disabled(), is_migration_disabled(current),
	       current->pid, current->comm);

	debug_show_held_locks(current);
	dump_stack();
	add_taint(TAINT_WARN, LOCKDEP_STILL_OK);
}
EXPORT_SYMBOL_GPL(__cant_migrate);
#endif
Linus Torvalds's avatar
Linus Torvalds committed
9652 9653 9654
#endif

#ifdef CONFIG_MAGIC_SYSRQ
9655
void normalize_rt_tasks(void)
9656
{
9657
	struct task_struct *g, *p;
9658 9659 9660
	struct sched_attr attr = {
		.sched_policy = SCHED_NORMAL,
	};
Linus Torvalds's avatar
Linus Torvalds committed
9661

9662
	read_lock(&tasklist_lock);
9663
	for_each_process_thread(g, p) {
9664 9665 9666
		/*
		 * Only normalize user tasks:
		 */
9667
		if (p->flags & PF_KTHREAD)
9668 9669
			continue;

9670
		p->se.exec_start = 0;
9671 9672 9673
		schedstat_set(p->stats.wait_start,  0);
		schedstat_set(p->stats.sleep_start, 0);
		schedstat_set(p->stats.block_start, 0);
Ingo Molnar's avatar
Ingo Molnar committed
9674

9675
		if (!dl_task(p) && !rt_task(p)) {
Ingo Molnar's avatar
Ingo Molnar committed
9676 9677 9678 9679
			/*
			 * Renice negative nice level userspace
			 * tasks back to 0:
			 */
9680
			if (task_nice(p) < 0)
Ingo Molnar's avatar
Ingo Molnar committed
9681
				set_user_nice(p, 0);
Linus Torvalds's avatar
Linus Torvalds committed
9682
			continue;
Ingo Molnar's avatar
Ingo Molnar committed
9683
		}
Linus Torvalds's avatar
Linus Torvalds committed
9684

9685
		__sched_setscheduler(p, &attr, false, false);
9686
	}
9687
	read_unlock(&tasklist_lock);
Linus Torvalds's avatar
Linus Torvalds committed
9688 9689 9690
}

#endif /* CONFIG_MAGIC_SYSRQ */
9691

9692
#if defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB)
9693
/*
9694
 * These functions are only useful for the IA64 MCA handling, or kdb.
9695 9696 9697 9698 9699 9700 9701 9702 9703
 *
 * 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.
 */

/**
9704
 * curr_task - return the current task for a given CPU.
9705 9706 9707
 * @cpu: the processor in question.
 *
 * ONLY VALID WHEN THE WHOLE SYSTEM IS STOPPED!
9708 9709
 *
 * Return: The current task for @cpu.
9710
 */
9711
struct task_struct *curr_task(int cpu)
9712 9713 9714 9715
{
	return cpu_curr(cpu);
}

9716 9717 9718
#endif /* defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB) */

#ifdef CONFIG_IA64
9719
/**
9720
 * ia64_set_curr_task - set the current task for a given CPU.
9721 9722 9723 9724
 * @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
9725
 * are serviced on a separate stack. It allows the architecture to switch the
9726
 * notion of the current task on a CPU in a non-blocking manner. This function
9727 9728 9729 9730 9731 9732 9733
 * 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!
 */
9734
void ia64_set_curr_task(int cpu, struct task_struct *p)
9735 9736 9737 9738 9739
{
	cpu_curr(cpu) = p;
}

#endif
9740

Dhaval Giani's avatar
Dhaval Giani committed
9741
#ifdef CONFIG_CGROUP_SCHED
9742 9743 9744
/* task_group_lock serializes the addition/removal of task groups */
static DEFINE_SPINLOCK(task_group_lock);

9745 9746 9747 9748
static inline void alloc_uclamp_sched_group(struct task_group *tg,
					    struct task_group *parent)
{
#ifdef CONFIG_UCLAMP_TASK_GROUP
9749
	enum uclamp_id clamp_id;
9750 9751 9752 9753

	for_each_clamp_id(clamp_id) {
		uclamp_se_set(&tg->uclamp_req[clamp_id],
			      uclamp_none(clamp_id), false);
9754
		tg->uclamp[clamp_id] = parent->uclamp[clamp_id];
9755 9756 9757 9758
	}
#endif
}

9759
static void sched_free_group(struct task_group *tg)
9760 9761 9762
{
	free_fair_sched_group(tg);
	free_rt_sched_group(tg);
9763
	autogroup_free(tg);
9764
	kmem_cache_free(task_group_cache, tg);
9765 9766
}

9767 9768 9769 9770 9771 9772 9773 9774 9775 9776 9777 9778 9779 9780 9781 9782
static void sched_free_group_rcu(struct rcu_head *rcu)
{
	sched_free_group(container_of(rcu, struct task_group, rcu));
}

static void sched_unregister_group(struct task_group *tg)
{
	unregister_fair_sched_group(tg);
	unregister_rt_sched_group(tg);
	/*
	 * We have to wait for yet another RCU grace period to expire, as
	 * print_cfs_stats() might run concurrently.
	 */
	call_rcu(&tg->rcu, sched_free_group_rcu);
}

9783
/* allocate runqueue etc for a new task group */
9784
struct task_group *sched_create_group(struct task_group *parent)
9785 9786 9787
{
	struct task_group *tg;

9788
	tg = kmem_cache_alloc(task_group_cache, GFP_KERNEL | __GFP_ZERO);
9789 9790 9791
	if (!tg)
		return ERR_PTR(-ENOMEM);

9792
	if (!alloc_fair_sched_group(tg, parent))
9793 9794
		goto err;

9795
	if (!alloc_rt_sched_group(tg, parent))
9796 9797
		goto err;

9798 9799
	alloc_uclamp_sched_group(tg, parent);

9800 9801 9802
	return tg;

err:
9803
	sched_free_group(tg);
9804 9805 9806 9807 9808 9809 9810
	return ERR_PTR(-ENOMEM);
}

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

9811
	spin_lock_irqsave(&task_group_lock, flags);
Peter Zijlstra's avatar
Peter Zijlstra committed
9812
	list_add_rcu(&tg->list, &task_groups);
Peter Zijlstra's avatar
Peter Zijlstra committed
9813

9814 9815
	/* Root should already exist: */
	WARN_ON(!parent);
Peter Zijlstra's avatar
Peter Zijlstra committed
9816 9817 9818

	tg->parent = parent;
	INIT_LIST_HEAD(&tg->children);
9819
	list_add_rcu(&tg->siblings, &parent->children);
9820
	spin_unlock_irqrestore(&task_group_lock, flags);
9821 9822

	online_fair_sched_group(tg);
9823 9824
}

9825
/* rcu callback to free various structures associated with a task group */
9826
static void sched_unregister_group_rcu(struct rcu_head *rhp)
9827
{
9828
	/* Now it should be safe to free those cfs_rqs: */
9829
	sched_unregister_group(container_of(rhp, struct task_group, rcu));
9830 9831
}

9832
void sched_destroy_group(struct task_group *tg)
9833
{
9834
	/* Wait for possible concurrent references to cfs_rqs complete: */
9835
	call_rcu(&tg->rcu, sched_unregister_group_rcu);
9836 9837
}

9838
void sched_release_group(struct task_group *tg)
9839
{
9840
	unsigned long flags;
9841

9842 9843 9844 9845 9846 9847 9848 9849 9850 9851 9852 9853 9854
	/*
	 * Unlink first, to avoid walk_tg_tree_from() from finding us (via
	 * sched_cfs_period_timer()).
	 *
	 * For this to be effective, we have to wait for all pending users of
	 * this task group to leave their RCU critical section to ensure no new
	 * user will see our dying task group any more. Specifically ensure
	 * that tg_unthrottle_up() won't add decayed cfs_rq's to it.
	 *
	 * We therefore defer calling unregister_fair_sched_group() to
	 * sched_unregister_group() which is guarantied to get called only after the
	 * current RCU grace period has expired.
	 */
9855
	spin_lock_irqsave(&task_group_lock, flags);
Peter Zijlstra's avatar
Peter Zijlstra committed
9856
	list_del_rcu(&tg->list);
Peter Zijlstra's avatar
Peter Zijlstra committed
9857
	list_del_rcu(&tg->siblings);
9858
	spin_unlock_irqrestore(&task_group_lock, flags);
9859 9860
}

9861
static void sched_change_group(struct task_struct *tsk, int type)
9862
{
9863
	struct task_group *tg;
9864

9865 9866 9867 9868 9869 9870
	/*
	 * 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),
9871 9872 9873 9874
			  struct task_group, css);
	tg = autogroup_task_group(tsk, tg);
	tsk->sched_task_group = tg;

Peter Zijlstra's avatar
Peter Zijlstra committed
9875
#ifdef CONFIG_FAIR_GROUP_SCHED
9876 9877
	if (tsk->sched_class->task_change_group)
		tsk->sched_class->task_change_group(tsk, type);
9878
	else
Peter Zijlstra's avatar
Peter Zijlstra committed
9879
#endif
9880
		set_task_rq(tsk, task_cpu(tsk));
9881 9882 9883 9884 9885 9886 9887 9888 9889 9890 9891
}

/*
 * 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)
{
9892 9893
	int queued, running, queue_flags =
		DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
9894 9895 9896 9897
	struct rq_flags rf;
	struct rq *rq;

	rq = task_rq_lock(tsk, &rf);
9898
	update_rq_clock(rq);
9899 9900 9901 9902 9903

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

	if (queued)
9904
		dequeue_task(rq, tsk, queue_flags);
9905
	if (running)
9906 9907 9908
		put_prev_task(rq, tsk);

	sched_change_group(tsk, TASK_MOVE_GROUP);
Peter Zijlstra's avatar
Peter Zijlstra committed
9909

9910
	if (queued)
9911
		enqueue_task(rq, tsk, queue_flags);
9912
	if (running) {
9913
		set_next_task(rq, tsk);
9914 9915 9916 9917 9918 9919 9920
		/*
		 * 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);
	}
9921

9922
	task_rq_unlock(rq, tsk, &rf);
9923
}
9924

9925
static inline struct task_group *css_tg(struct cgroup_subsys_state *css)
9926
{
9927
	return css ? container_of(css, struct task_group, css) : NULL;
9928 9929
}

9930 9931
static struct cgroup_subsys_state *
cpu_cgroup_css_alloc(struct cgroup_subsys_state *parent_css)
9932
{
9933 9934
	struct task_group *parent = css_tg(parent_css);
	struct task_group *tg;
9935

9936
	if (!parent) {
9937
		/* This is early initialization for the top cgroup */
9938
		return &root_task_group.css;
9939 9940
	}

9941
	tg = sched_create_group(parent);
9942 9943 9944 9945 9946 9947
	if (IS_ERR(tg))
		return ERR_PTR(-ENOMEM);

	return &tg->css;
}

9948 9949 9950 9951 9952 9953 9954 9955
/* 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);
9956 9957 9958

#ifdef CONFIG_UCLAMP_TASK_GROUP
	/* Propagate the effective uclamp value for the new group */
9959 9960
	mutex_lock(&uclamp_mutex);
	rcu_read_lock();
9961
	cpu_util_update_eff(css);
9962 9963
	rcu_read_unlock();
	mutex_unlock(&uclamp_mutex);
9964 9965
#endif

9966 9967 9968
	return 0;
}

9969
static void cpu_cgroup_css_released(struct cgroup_subsys_state *css)
9970
{
9971
	struct task_group *tg = css_tg(css);
9972

9973
	sched_release_group(tg);
9974 9975
}

9976
static void cpu_cgroup_css_free(struct cgroup_subsys_state *css)
9977
{
9978
	struct task_group *tg = css_tg(css);
9979

9980 9981 9982
	/*
	 * Relies on the RCU grace period between css_released() and this.
	 */
9983
	sched_unregister_group(tg);
9984 9985
}

9986 9987 9988 9989
/*
 * 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.
 */
9990
static void cpu_cgroup_fork(struct task_struct *task)
9991
{
9992 9993 9994 9995 9996
	struct rq_flags rf;
	struct rq *rq;

	rq = task_rq_lock(task, &rf);

9997
	update_rq_clock(rq);
9998 9999 10000
	sched_change_group(task, TASK_SET_GROUP);

	task_rq_unlock(rq, task, &rf);
10001 10002
}

10003
static int cpu_cgroup_can_attach(struct cgroup_taskset *tset)
10004
{
10005
	struct task_struct *task;
10006
	struct cgroup_subsys_state *css;
10007
	int ret = 0;
10008

10009
	cgroup_taskset_for_each(task, css, tset) {
10010
#ifdef CONFIG_RT_GROUP_SCHED
10011
		if (!sched_rt_can_attach(css_tg(css), task))
10012
			return -EINVAL;
10013
#endif
10014
		/*
10015
		 * Serialize against wake_up_new_task() such that if it's
10016 10017 10018 10019 10020 10021 10022 10023
		 * 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.
		 */
10024
		if (READ_ONCE(task->__state) == TASK_NEW)
10025 10026 10027 10028 10029
			ret = -EINVAL;
		raw_spin_unlock_irq(&task->pi_lock);

		if (ret)
			break;
10030
	}
10031
	return ret;
10032
}
10033

10034
static void cpu_cgroup_attach(struct cgroup_taskset *tset)
10035
{
10036
	struct task_struct *task;
10037
	struct cgroup_subsys_state *css;
10038

10039
	cgroup_taskset_for_each(task, css, tset)
10040
		sched_move_task(task);
10041 10042
}

10043
#ifdef CONFIG_UCLAMP_TASK_GROUP
10044 10045 10046 10047 10048 10049
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];
10050
	enum uclamp_id clamp_id;
10051 10052
	unsigned int clamps;

10053 10054 10055
	lockdep_assert_held(&uclamp_mutex);
	SCHED_WARN_ON(!rcu_read_lock_held());

10056 10057 10058 10059 10060 10061 10062 10063 10064 10065 10066 10067 10068 10069 10070 10071 10072 10073 10074 10075 10076 10077 10078 10079 10080 10081
	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);
		}
10082
		if (!clamps) {
10083
			css = css_rightmost_descendant(css);
10084 10085 10086 10087
			continue;
		}

		/* Immediately update descendants RUNNABLE tasks */
10088
		uclamp_update_active_tasks(css);
10089 10090
	}
}
10091 10092 10093 10094 10095 10096 10097 10098 10099 10100 10101 10102 10103 10104 10105 10106 10107 10108 10109 10110 10111 10112 10113 10114 10115 10116 10117 10118 10119 10120 10121 10122

/*
 * 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;
10123
		if ((u64)req.percent > UCLAMP_PERCENT_SCALE) {
10124 10125 10126 10127 10128 10129 10130 10131 10132 10133 10134 10135 10136 10137 10138 10139 10140 10141 10142 10143 10144 10145
			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;

10146 10147
	static_branch_enable(&sched_uclamp_used);

10148 10149 10150 10151 10152 10153 10154 10155 10156 10157 10158 10159 10160
	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;

10161 10162 10163
	/* Update effective clamps to track the most restrictive value */
	cpu_util_update_eff(of_css(of));

10164 10165 10166 10167 10168 10169 10170 10171 10172 10173 10174 10175 10176 10177 10178 10179 10180 10181 10182 10183 10184 10185 10186 10187 10188 10189 10190 10191 10192 10193 10194 10195 10196 10197 10198 10199 10200 10201 10202 10203 10204 10205 10206 10207 10208 10209 10210 10211 10212 10213 10214 10215 10216 10217 10218 10219
	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 */

10220
#ifdef CONFIG_FAIR_GROUP_SCHED
10221 10222
static int cpu_shares_write_u64(struct cgroup_subsys_state *css,
				struct cftype *cftype, u64 shareval)
10223
{
10224 10225
	if (shareval > scale_load_down(ULONG_MAX))
		shareval = MAX_SHARES;
10226
	return sched_group_set_shares(css_tg(css), scale_load(shareval));
10227 10228
}

10229 10230
static u64 cpu_shares_read_u64(struct cgroup_subsys_state *css,
			       struct cftype *cft)
10231
{
10232
	struct task_group *tg = css_tg(css);
10233

10234
	return (u64) scale_load_down(tg->shares);
10235
}
10236 10237

#ifdef CONFIG_CFS_BANDWIDTH
10238 10239
static DEFINE_MUTEX(cfs_constraints_mutex);

10240
const u64 max_cfs_quota_period = 1 * NSEC_PER_SEC; /* 1s */
10241
static const u64 min_cfs_quota_period = 1 * NSEC_PER_MSEC; /* 1ms */
10242 10243
/* More than 203 days if BW_SHIFT equals 20. */
static const u64 max_cfs_runtime = MAX_BW * NSEC_PER_USEC;
10244

10245 10246
static int __cfs_schedulable(struct task_group *tg, u64 period, u64 runtime);

10247 10248
static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota,
				u64 burst)
10249
{
10250
	int i, ret = 0, runtime_enabled, runtime_was_enabled;
10251
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
10252 10253 10254 10255 10256 10257 10258 10259 10260 10261 10262 10263 10264

	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;

	/*
Ingo Molnar's avatar
Ingo Molnar committed
10265
	 * Likewise, bound things on the other side by preventing insane quota
10266 10267 10268 10269 10270 10271
	 * periods.  This also allows us to normalize in computing quota
	 * feasibility.
	 */
	if (period > max_cfs_quota_period)
		return -EINVAL;

10272 10273 10274 10275 10276 10277
	/*
	 * Bound quota to defend quota against overflow during bandwidth shift.
	 */
	if (quota != RUNTIME_INF && quota > max_cfs_runtime)
		return -EINVAL;

10278 10279 10280 10281
	if (quota != RUNTIME_INF && (burst > quota ||
				     burst + quota > max_cfs_runtime))
		return -EINVAL;

10282 10283 10284 10285
	/*
	 * Prevent race between setting of cfs_rq->runtime_enabled and
	 * unthrottle_offline_cfs_rqs().
	 */
10286
	cpus_read_lock();
10287 10288 10289 10290 10291
	mutex_lock(&cfs_constraints_mutex);
	ret = __cfs_schedulable(tg, period, quota);
	if (ret)
		goto out_unlock;

10292
	runtime_enabled = quota != RUNTIME_INF;
10293
	runtime_was_enabled = cfs_b->quota != RUNTIME_INF;
10294 10295 10296 10297 10298 10299
	/*
	 * 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();
10300 10301 10302
	raw_spin_lock_irq(&cfs_b->lock);
	cfs_b->period = ns_to_ktime(period);
	cfs_b->quota = quota;
10303
	cfs_b->burst = burst;
10304

Paul Turner's avatar
Paul Turner committed
10305
	__refill_cfs_bandwidth_runtime(cfs_b);
10306 10307

	/* Restart the period timer (if active) to handle new period expiry: */
10308 10309
	if (runtime_enabled)
		start_cfs_bandwidth(cfs_b);
10310

10311 10312
	raw_spin_unlock_irq(&cfs_b->lock);

10313
	for_each_online_cpu(i) {
10314
		struct cfs_rq *cfs_rq = tg->cfs_rq[i];
10315
		struct rq *rq = cfs_rq->rq;
10316
		struct rq_flags rf;
10317

10318
		rq_lock_irq(rq, &rf);
10319
		cfs_rq->runtime_enabled = runtime_enabled;
10320
		cfs_rq->runtime_remaining = 0;
10321

10322
		if (cfs_rq->throttled)
10323
			unthrottle_cfs_rq(cfs_rq);
10324
		rq_unlock_irq(rq, &rf);
10325
	}
10326 10327
	if (runtime_was_enabled && !runtime_enabled)
		cfs_bandwidth_usage_dec();
10328 10329
out_unlock:
	mutex_unlock(&cfs_constraints_mutex);
10330
	cpus_read_unlock();
10331

10332
	return ret;
10333 10334
}

10335
static int tg_set_cfs_quota(struct task_group *tg, long cfs_quota_us)
10336
{
10337
	u64 quota, period, burst;
10338

10339
	period = ktime_to_ns(tg->cfs_bandwidth.period);
10340
	burst = tg->cfs_bandwidth.burst;
10341 10342
	if (cfs_quota_us < 0)
		quota = RUNTIME_INF;
10343
	else if ((u64)cfs_quota_us <= U64_MAX / NSEC_PER_USEC)
10344
		quota = (u64)cfs_quota_us * NSEC_PER_USEC;
10345 10346
	else
		return -EINVAL;
10347

10348
	return tg_set_cfs_bandwidth(tg, period, quota, burst);
10349 10350
}

10351
static long tg_get_cfs_quota(struct task_group *tg)
10352 10353 10354
{
	u64 quota_us;

10355
	if (tg->cfs_bandwidth.quota == RUNTIME_INF)
10356 10357
		return -1;

10358
	quota_us = tg->cfs_bandwidth.quota;
10359 10360 10361 10362 10363
	do_div(quota_us, NSEC_PER_USEC);

	return quota_us;
}

10364
static int tg_set_cfs_period(struct task_group *tg, long cfs_period_us)
10365
{
10366
	u64 quota, period, burst;
10367

10368 10369 10370
	if ((u64)cfs_period_us > U64_MAX / NSEC_PER_USEC)
		return -EINVAL;

10371
	period = (u64)cfs_period_us * NSEC_PER_USEC;
10372
	quota = tg->cfs_bandwidth.quota;
10373
	burst = tg->cfs_bandwidth.burst;
10374

10375
	return tg_set_cfs_bandwidth(tg, period, quota, burst);
10376 10377
}

10378
static long tg_get_cfs_period(struct task_group *tg)
10379 10380 10381
{
	u64 cfs_period_us;

10382
	cfs_period_us = ktime_to_ns(tg->cfs_bandwidth.period);
10383 10384 10385 10386 10387
	do_div(cfs_period_us, NSEC_PER_USEC);

	return cfs_period_us;
}

10388 10389 10390 10391 10392 10393 10394 10395 10396 10397 10398 10399 10400 10401 10402 10403 10404 10405 10406 10407 10408 10409 10410 10411
static int tg_set_cfs_burst(struct task_group *tg, long cfs_burst_us)
{
	u64 quota, period, burst;

	if ((u64)cfs_burst_us > U64_MAX / NSEC_PER_USEC)
		return -EINVAL;

	burst = (u64)cfs_burst_us * NSEC_PER_USEC;
	period = ktime_to_ns(tg->cfs_bandwidth.period);
	quota = tg->cfs_bandwidth.quota;

	return tg_set_cfs_bandwidth(tg, period, quota, burst);
}

static long tg_get_cfs_burst(struct task_group *tg)
{
	u64 burst_us;

	burst_us = tg->cfs_bandwidth.burst;
	do_div(burst_us, NSEC_PER_USEC);

	return burst_us;
}

10412 10413
static s64 cpu_cfs_quota_read_s64(struct cgroup_subsys_state *css,
				  struct cftype *cft)
10414
{
10415
	return tg_get_cfs_quota(css_tg(css));
10416 10417
}

10418 10419
static int cpu_cfs_quota_write_s64(struct cgroup_subsys_state *css,
				   struct cftype *cftype, s64 cfs_quota_us)
10420
{
10421
	return tg_set_cfs_quota(css_tg(css), cfs_quota_us);
10422 10423
}

10424 10425
static u64 cpu_cfs_period_read_u64(struct cgroup_subsys_state *css,
				   struct cftype *cft)
10426
{
10427
	return tg_get_cfs_period(css_tg(css));
10428 10429
}

10430 10431
static int cpu_cfs_period_write_u64(struct cgroup_subsys_state *css,
				    struct cftype *cftype, u64 cfs_period_us)
10432
{
10433
	return tg_set_cfs_period(css_tg(css), cfs_period_us);
10434 10435
}

10436 10437 10438 10439 10440 10441 10442 10443 10444 10445 10446 10447
static u64 cpu_cfs_burst_read_u64(struct cgroup_subsys_state *css,
				  struct cftype *cft)
{
	return tg_get_cfs_burst(css_tg(css));
}

static int cpu_cfs_burst_write_u64(struct cgroup_subsys_state *css,
				   struct cftype *cftype, u64 cfs_burst_us)
{
	return tg_set_cfs_burst(css_tg(css), cfs_burst_us);
}

10448 10449 10450 10451 10452 10453 10454 10455 10456 10457 10458 10459 10460 10461 10462 10463 10464 10465 10466 10467 10468 10469 10470 10471 10472 10473 10474 10475 10476 10477 10478 10479
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;
10480
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
10481 10482 10483 10484 10485
	s64 quota = 0, parent_quota = -1;

	if (!tg->parent) {
		quota = RUNTIME_INF;
	} else {
10486
		struct cfs_bandwidth *parent_b = &tg->parent->cfs_bandwidth;
10487 10488

		quota = normalize_cfs_quota(tg, d);
10489
		parent_quota = parent_b->hierarchical_quota;
10490 10491

		/*
10492 10493
		 * Ensure max(child_quota) <= parent_quota.  On cgroup2,
		 * always take the min.  On cgroup1, only inherit when no
10494
		 * limit is set:
10495
		 */
10496 10497 10498 10499 10500 10501 10502 10503
		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;
		}
10504
	}
10505
	cfs_b->hierarchical_quota = quota;
10506 10507 10508 10509 10510 10511

	return 0;
}

static int __cfs_schedulable(struct task_group *tg, u64 period, u64 quota)
{
10512
	int ret;
10513 10514 10515 10516 10517 10518 10519 10520 10521 10522 10523
	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);
	}

10524 10525 10526 10527 10528
	rcu_read_lock();
	ret = walk_tg_tree(tg_cfs_schedulable_down, tg_nop, &data);
	rcu_read_unlock();

	return ret;
10529
}
10530

10531
static int cpu_cfs_stat_show(struct seq_file *sf, void *v)
10532
{
10533
	struct task_group *tg = css_tg(seq_css(sf));
10534
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
10535

10536 10537 10538
	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);
10539

10540
	if (schedstat_enabled() && tg != &root_task_group) {
10541
		struct sched_statistics *stats;
10542 10543 10544
		u64 ws = 0;
		int i;

10545 10546 10547 10548
		for_each_possible_cpu(i) {
			stats = __schedstats_from_se(tg->se[i]);
			ws += schedstat_val(stats->wait_sum);
		}
10549 10550 10551 10552

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

10553 10554 10555
	seq_printf(sf, "nr_bursts %d\n", cfs_b->nr_burst);
	seq_printf(sf, "burst_time %llu\n", cfs_b->burst_time);

10556 10557
	return 0;
}
10558
#endif /* CONFIG_CFS_BANDWIDTH */
10559
#endif /* CONFIG_FAIR_GROUP_SCHED */
10560

10561
#ifdef CONFIG_RT_GROUP_SCHED
10562 10563
static int cpu_rt_runtime_write(struct cgroup_subsys_state *css,
				struct cftype *cft, s64 val)
Peter Zijlstra's avatar
Peter Zijlstra committed
10564
{
10565
	return sched_group_set_rt_runtime(css_tg(css), val);
Peter Zijlstra's avatar
Peter Zijlstra committed
10566 10567
}

10568 10569
static s64 cpu_rt_runtime_read(struct cgroup_subsys_state *css,
			       struct cftype *cft)
Peter Zijlstra's avatar
Peter Zijlstra committed
10570
{
10571
	return sched_group_rt_runtime(css_tg(css));
Peter Zijlstra's avatar
Peter Zijlstra committed
10572
}
10573

10574 10575
static int cpu_rt_period_write_uint(struct cgroup_subsys_state *css,
				    struct cftype *cftype, u64 rt_period_us)
10576
{
10577
	return sched_group_set_rt_period(css_tg(css), rt_period_us);
10578 10579
}

10580 10581
static u64 cpu_rt_period_read_uint(struct cgroup_subsys_state *css,
				   struct cftype *cft)
10582
{
10583
	return sched_group_rt_period(css_tg(css));
10584
}
10585
#endif /* CONFIG_RT_GROUP_SCHED */
Peter Zijlstra's avatar
Peter Zijlstra committed
10586

10587 10588 10589 10590 10591 10592 10593 10594 10595 10596 10597 10598 10599 10600
#ifdef CONFIG_FAIR_GROUP_SCHED
static s64 cpu_idle_read_s64(struct cgroup_subsys_state *css,
			       struct cftype *cft)
{
	return css_tg(css)->idle;
}

static int cpu_idle_write_s64(struct cgroup_subsys_state *css,
				struct cftype *cft, s64 idle)
{
	return sched_group_set_idle(css_tg(css), idle);
}
#endif

10601
static struct cftype cpu_legacy_files[] = {
10602
#ifdef CONFIG_FAIR_GROUP_SCHED
10603 10604
	{
		.name = "shares",
10605 10606
		.read_u64 = cpu_shares_read_u64,
		.write_u64 = cpu_shares_write_u64,
10607
	},
10608 10609 10610 10611 10612
	{
		.name = "idle",
		.read_s64 = cpu_idle_read_s64,
		.write_s64 = cpu_idle_write_s64,
	},
10613
#endif
10614 10615 10616 10617 10618 10619 10620 10621 10622 10623 10624
#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,
	},
10625 10626 10627 10628 10629
	{
		.name = "cfs_burst_us",
		.read_u64 = cpu_cfs_burst_read_u64,
		.write_u64 = cpu_cfs_burst_write_u64,
	},
10630 10631
	{
		.name = "stat",
10632
		.seq_show = cpu_cfs_stat_show,
10633
	},
10634
#endif
10635
#ifdef CONFIG_RT_GROUP_SCHED
Peter Zijlstra's avatar
Peter Zijlstra committed
10636
	{
Peter Zijlstra's avatar
Peter Zijlstra committed
10637
		.name = "rt_runtime_us",
10638 10639
		.read_s64 = cpu_rt_runtime_read,
		.write_s64 = cpu_rt_runtime_write,
Peter Zijlstra's avatar
Peter Zijlstra committed
10640
	},
10641 10642
	{
		.name = "rt_period_us",
10643 10644
		.read_u64 = cpu_rt_period_read_uint,
		.write_u64 = cpu_rt_period_write_uint,
10645
	},
10646 10647 10648 10649 10650 10651 10652 10653 10654 10655 10656 10657 10658 10659
#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,
	},
10660
#endif
10661
	{ }	/* Terminate */
10662 10663
};

10664 10665
static int cpu_extra_stat_show(struct seq_file *sf,
			       struct cgroup_subsys_state *css)
10666 10667 10668
{
#ifdef CONFIG_CFS_BANDWIDTH
	{
10669
		struct task_group *tg = css_tg(css);
10670
		struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
10671
		u64 throttled_usec, burst_usec;
10672 10673 10674

		throttled_usec = cfs_b->throttled_time;
		do_div(throttled_usec, NSEC_PER_USEC);
10675 10676
		burst_usec = cfs_b->burst_time;
		do_div(burst_usec, NSEC_PER_USEC);
10677 10678 10679

		seq_printf(sf, "nr_periods %d\n"
			   "nr_throttled %d\n"
10680 10681 10682
			   "throttled_usec %llu\n"
			   "nr_bursts %d\n"
			   "burst_usec %llu\n",
10683
			   cfs_b->nr_periods, cfs_b->nr_throttled,
10684
			   throttled_usec, cfs_b->nr_burst, burst_usec);
10685 10686 10687 10688 10689 10690 10691 10692 10693 10694 10695 10696 10697 10698 10699 10700 10701 10702 10703 10704 10705 10706 10707 10708 10709 10710 10711 10712 10713 10714 10715 10716 10717 10718 10719 10720 10721 10722 10723 10724 10725 10726 10727 10728 10729 10730 10731 10732 10733 10734 10735 10736 10737 10738 10739
	}
#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;
10740
	int idx;
10741 10742 10743 10744

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

10745 10746 10747 10748
	idx = NICE_TO_PRIO(nice) - MAX_RT_PRIO;
	idx = array_index_nospec(idx, 40);
	weight = sched_prio_to_weight[idx];

10749 10750 10751 10752 10753 10754 10755 10756 10757 10758 10759 10760 10761 10762 10763 10764 10765 10766 10767 10768 10769
	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 */

10770
	if (sscanf(buf, "%20s %llu", tok, periodp) < 1)
10771 10772 10773 10774 10775 10776 10777 10778 10779 10780 10781 10782 10783 10784 10785 10786 10787 10788 10789 10790 10791 10792 10793 10794 10795 10796 10797 10798
		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);
10799
	u64 burst = tg_get_cfs_burst(tg);
10800 10801 10802 10803 10804
	u64 quota;
	int ret;

	ret = cpu_period_quota_parse(buf, &period, &quota);
	if (!ret)
10805
		ret = tg_set_cfs_bandwidth(tg, period, quota, burst);
10806 10807 10808 10809 10810 10811 10812 10813 10814 10815 10816 10817 10818 10819 10820 10821 10822 10823
	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,
	},
10824 10825 10826 10827 10828 10829
	{
		.name = "idle",
		.flags = CFTYPE_NOT_ON_ROOT,
		.read_s64 = cpu_idle_read_s64,
		.write_s64 = cpu_idle_write_s64,
	},
10830 10831 10832 10833 10834 10835 10836 10837
#endif
#ifdef CONFIG_CFS_BANDWIDTH
	{
		.name = "max",
		.flags = CFTYPE_NOT_ON_ROOT,
		.seq_show = cpu_max_show,
		.write = cpu_max_write,
	},
10838 10839 10840 10841 10842 10843
	{
		.name = "max.burst",
		.flags = CFTYPE_NOT_ON_ROOT,
		.read_u64 = cpu_cfs_burst_read_u64,
		.write_u64 = cpu_cfs_burst_write_u64,
	},
10844 10845 10846 10847 10848 10849 10850 10851 10852 10853 10854 10855 10856 10857
#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,
	},
10858 10859 10860 10861
#endif
	{ }	/* terminate */
};

10862
struct cgroup_subsys cpu_cgrp_subsys = {
10863
	.css_alloc	= cpu_cgroup_css_alloc,
10864
	.css_online	= cpu_cgroup_css_online,
10865
	.css_released	= cpu_cgroup_css_released,
10866
	.css_free	= cpu_cgroup_css_free,
10867
	.css_extra_stat_show = cpu_extra_stat_show,
10868
	.fork		= cpu_cgroup_fork,
10869 10870
	.can_attach	= cpu_cgroup_can_attach,
	.attach		= cpu_cgroup_attach,
10871
	.legacy_cftypes	= cpu_legacy_files,
10872
	.dfl_cftypes	= cpu_files,
10873
	.early_init	= true,
10874
	.threaded	= true,
10875 10876
};

10877
#endif	/* CONFIG_CGROUP_SCHED */
10878

10879 10880 10881 10882 10883
void dump_cpu_task(int cpu)
{
	pr_info("Task dump for CPU %d:\n", cpu);
	sched_show_task(cpu_curr(cpu));
}
10884 10885 10886 10887 10888 10889 10890 10891 10892 10893 10894 10895 10896 10897 10898 10899 10900 10901 10902 10903 10904 10905 10906 10907 10908 10909 10910 10911 10912 10913 10914 10915 10916 10917 10918 10919 10920 10921 10922 10923 10924

/*
 * 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,
};
10925

10926 10927 10928 10929
void call_trace_sched_update_nr_running(struct rq *rq, int count)
{
        trace_sched_update_nr_running_tp(rq, count);
}