core.c 284 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 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
#include <linux/highmem.h>
#include <linux/hrtimer_api.h>
#include <linux/ktime_api.h>
#include <linux/sched/signal.h>
#include <linux/syscalls_api.h>
#include <linux/debug_locks.h>
#include <linux/prefetch.h>
#include <linux/capability.h>
#include <linux/pgtable_api.h>
#include <linux/wait_bit.h>
#include <linux/jiffies.h>
#include <linux/spinlock_api.h>
#include <linux/cpumask_api.h>
#include <linux/lockdep_api.h>
#include <linux/hardirq.h>
#include <linux/softirq.h>
#include <linux/refcount_api.h>
#include <linux/topology.h>
#include <linux/sched/clock.h>
#include <linux/sched/cond_resched.h>
29
#include <linux/sched/cputime.h>
30
#include <linux/sched/debug.h>
31 32
#include <linux/sched/hotplug.h>
#include <linux/sched/init.h>
33 34 35 36 37 38
#include <linux/sched/isolation.h>
#include <linux/sched/loadavg.h>
#include <linux/sched/mm.h>
#include <linux/sched/nohz.h>
#include <linux/sched/rseq_api.h>
#include <linux/sched/rt.h>
Linus Torvalds's avatar
Linus Torvalds committed
39

40
#include <linux/blkdev.h>
41 42 43 44 45 46 47
#include <linux/context_tracking.h>
#include <linux/cpuset.h>
#include <linux/delayacct.h>
#include <linux/init_task.h>
#include <linux/interrupt.h>
#include <linux/ioprio.h>
#include <linux/kallsyms.h>
48
#include <linux/kcov.h>
49 50 51 52 53 54 55 56 57 58 59 60
#include <linux/kprobes.h>
#include <linux/llist_api.h>
#include <linux/mmu_context.h>
#include <linux/mmzone.h>
#include <linux/mutex_api.h>
#include <linux/nmi.h>
#include <linux/nospec.h>
#include <linux/perf_event_api.h>
#include <linux/profile.h>
#include <linux/psi.h>
#include <linux/rcuwait_api.h>
#include <linux/sched/wake_q.h>
61
#include <linux/scs.h>
62 63 64 65 66 67 68
#include <linux/slab.h>
#include <linux/syscalls.h>
#include <linux/vtime.h>
#include <linux/wait_api.h>
#include <linux/workqueue_api.h>

#ifdef CONFIG_PREEMPT_DYNAMIC
69 70 71
# ifdef CONFIG_GENERIC_ENTRY
#  include <linux/entry-common.h>
# endif
72 73 74
#endif

#include <uapi/linux/sched/types.h>
75

76
#include <asm/switch_to.h>
77
#include <asm/tlb.h>
Linus Torvalds's avatar
Linus Torvalds committed
78

79
#define CREATE_TRACE_POINTS
80
#include <linux/sched/rseq_api.h>
81 82 83
#include <trace/events/sched.h>
#undef CREATE_TRACE_POINTS

84
#include "sched.h"
85 86
#include "stats.h"
#include "autogroup.h"
87

88
#include "autogroup.h"
89
#include "pelt.h"
90
#include "smp.h"
91
#include "stats.h"
Linus Torvalds's avatar
Linus Torvalds committed
92

93
#include "../workqueue_internal.h"
94
#include "../../io_uring/io-wq.h"
95
#include "../smpboot.h"
96

97 98 99 100 101 102 103 104 105
/*
 * 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);
106
EXPORT_TRACEPOINT_SYMBOL_GPL(pelt_thermal_tp);
107
EXPORT_TRACEPOINT_SYMBOL_GPL(sched_cpu_capacity_tp);
108
EXPORT_TRACEPOINT_SYMBOL_GPL(sched_overutilized_tp);
109 110
EXPORT_TRACEPOINT_SYMBOL_GPL(sched_util_est_cfs_tp);
EXPORT_TRACEPOINT_SYMBOL_GPL(sched_util_est_se_tp);
111
EXPORT_TRACEPOINT_SYMBOL_GPL(sched_update_nr_running_tp);
112

113
DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
114

115
#ifdef CONFIG_SCHED_DEBUG
116 117
/*
 * Debugging: various feature bits
118 119 120 121
 *
 * 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.
122
 */
123 124
#define SCHED_FEAT(name, enabled)	\
	(1UL << __SCHED_FEAT_##name) * enabled |
125
const_debug unsigned int sysctl_sched_features =
126
#include "features.h"
127 128
	0;
#undef SCHED_FEAT
129 130 131 132 133 134 135 136 137 138 139

/*
 * 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 */
140

141 142 143 144
/*
 * Number of tasks to iterate in a single balance run.
 * Limited because this is done with IRQs disabled.
 */
145
const_debug unsigned int sysctl_sched_nr_migrate = SCHED_NR_MIGRATE_BREAK;
146

147
__read_mostly int scheduler_running;
148

Peter Zijlstra's avatar
Peter Zijlstra committed
149 150 151 152
#ifdef CONFIG_SCHED_CORE

DEFINE_STATIC_KEY_FALSE(__sched_core_enabled);

153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175
/* 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 */
176
static inline bool prio_less(struct task_struct *a, struct task_struct *b, bool in_fi)
177 178 179 180 181 182 183 184 185 186 187 188 189
{

	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);

190 191
	if (pa == MAX_RT_PRIO + MAX_NICE)	/* fair */
		return cfs_prio_less(a, b, in_fi);
192 193 194 195 196 197 198 199 200 201 202 203 204

	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 */
205
	if (prio_less(b, a, !!task_rq(a)->core->core_forceidle_count))
206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231
		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;
}

232
void sched_core_enqueue(struct rq *rq, struct task_struct *p)
233 234 235 236 237 238 239 240 241
{
	rq->core->core_task_seq++;

	if (!p->core_cookie)
		return;

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

242
void sched_core_dequeue(struct rq *rq, struct task_struct *p, int flags)
243 244 245
{
	rq->core->core_task_seq++;

246 247 248 249
	if (sched_core_enqueued(p)) {
		rb_erase(&p->core_node, &rq->core_tree);
		RB_CLEAR_NODE(&p->core_node);
	}
250

251 252 253 254 255 256 257 258
	/*
	 * 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);
259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277
}

/*
 * 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);
}

278 279 280 281 282 283 284 285 286 287 288 289 290 291 292
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
293 294 295 296 297 298 299 300 301 302 303 304 305 306
/*
 * 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);
307
static atomic_t sched_core_count;
Peter Zijlstra's avatar
Peter Zijlstra committed
308 309
static struct cpumask sched_core_mask;

310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329
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
330 331
static void __sched_core_flip(bool enabled)
{
332 333
	unsigned long flags;
	int cpu, t;
Peter Zijlstra's avatar
Peter Zijlstra committed
334 335 336 337 338 339 340 341 342 343

	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);

344
		sched_core_lock(cpu, &flags);
Peter Zijlstra's avatar
Peter Zijlstra committed
345 346 347 348

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

349 350
		cpu_rq(cpu)->core->core_forceidle_start = 0;

351
		sched_core_unlock(cpu, &flags);
Peter Zijlstra's avatar
Peter Zijlstra committed
352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367

		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();
}

368
static void sched_core_assert_empty(void)
Peter Zijlstra's avatar
Peter Zijlstra committed
369
{
370
	int cpu;
Peter Zijlstra's avatar
Peter Zijlstra committed
371

372 373 374 375 376 377
	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
378 379 380 381 382 383 384
	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);
385
	sched_core_assert_empty();
Peter Zijlstra's avatar
Peter Zijlstra committed
386 387 388 389
}

static void __sched_core_disable(void)
{
390
	sched_core_assert_empty();
Peter Zijlstra's avatar
Peter Zijlstra committed
391 392 393 394 395 396
	__sched_core_flip(false);
	static_branch_disable(&__sched_core_enabled);
}

void sched_core_get(void)
{
397 398 399
	if (atomic_inc_not_zero(&sched_core_count))
		return;

Peter Zijlstra's avatar
Peter Zijlstra committed
400
	mutex_lock(&sched_core_mutex);
401
	if (!atomic_read(&sched_core_count))
Peter Zijlstra's avatar
Peter Zijlstra committed
402
		__sched_core_enable();
403 404 405

	smp_mb__before_atomic();
	atomic_inc(&sched_core_count);
Peter Zijlstra's avatar
Peter Zijlstra committed
406 407 408
	mutex_unlock(&sched_core_mutex);
}

409
static void __sched_core_put(struct work_struct *work)
Peter Zijlstra's avatar
Peter Zijlstra committed
410
{
411
	if (atomic_dec_and_mutex_lock(&sched_core_count, &sched_core_mutex)) {
Peter Zijlstra's avatar
Peter Zijlstra committed
412
		__sched_core_disable();
413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429
		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
430 431
}

432 433 434
#else /* !CONFIG_SCHED_CORE */

static inline void sched_core_enqueue(struct rq *rq, struct task_struct *p) { }
435 436
static inline void
sched_core_dequeue(struct rq *rq, struct task_struct *p, int flags) { }
437

Peter Zijlstra's avatar
Peter Zijlstra committed
438 439
#endif /* CONFIG_SCHED_CORE */

440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455
/*
 * 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
456
 * always looks at the local rq data structures to find the most eligible task
457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479
 * 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
480
 *  - sched_move_task():	p->sched_task_group
481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531
 *  - 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()
 *
 */

532 533
void raw_spin_rq_lock_nested(struct rq *rq, int subclass)
{
534 535
	raw_spinlock_t *lock;

Peter Zijlstra's avatar
Peter Zijlstra committed
536 537
	/* Matches synchronize_rcu() in __sched_core_enable() */
	preempt_disable();
538 539
	if (sched_core_disabled()) {
		raw_spin_lock_nested(&rq->__lock, subclass);
Peter Zijlstra's avatar
Peter Zijlstra committed
540 541
		/* preempt_count *MUST* be > 1 */
		preempt_enable_no_resched();
542 543 544 545
		return;
	}

	for (;;) {
546
		lock = __rq_lockp(rq);
547
		raw_spin_lock_nested(lock, subclass);
548
		if (likely(lock == __rq_lockp(rq))) {
Peter Zijlstra's avatar
Peter Zijlstra committed
549 550
			/* preempt_count *MUST* be > 1 */
			preempt_enable_no_resched();
551
			return;
Peter Zijlstra's avatar
Peter Zijlstra committed
552
		}
553 554
		raw_spin_unlock(lock);
	}
555 556 557 558
}

bool raw_spin_rq_trylock(struct rq *rq)
{
559 560 561
	raw_spinlock_t *lock;
	bool ret;

Peter Zijlstra's avatar
Peter Zijlstra committed
562 563 564 565 566 567 568
	/* Matches synchronize_rcu() in __sched_core_enable() */
	preempt_disable();
	if (sched_core_disabled()) {
		ret = raw_spin_trylock(&rq->__lock);
		preempt_enable();
		return ret;
	}
569 570

	for (;;) {
571
		lock = __rq_lockp(rq);
572
		ret = raw_spin_trylock(lock);
573
		if (!ret || (likely(lock == __rq_lockp(rq)))) {
Peter Zijlstra's avatar
Peter Zijlstra committed
574
			preempt_enable();
575
			return ret;
Peter Zijlstra's avatar
Peter Zijlstra committed
576
		}
577 578
		raw_spin_unlock(lock);
	}
579 580 581 582 583 584 585
}

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

586 587 588 589 590 591 592 593 594 595 596 597
#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);
598 599
	if (__rq_lockp(rq1) != __rq_lockp(rq2))
		raw_spin_rq_lock_nested(rq2, SINGLE_DEPTH_NESTING);
600

601
	double_rq_clock_clear_update(rq1, rq2);
602 603 604
}
#endif

605 606 607
/*
 * __task_rq_lock - lock the rq @p resides on.
 */
608
struct rq *__task_rq_lock(struct task_struct *p, struct rq_flags *rf)
609 610 611 612 613 614 615 616
	__acquires(rq->lock)
{
	struct rq *rq;

	lockdep_assert_held(&p->pi_lock);

	for (;;) {
		rq = task_rq(p);
617
		raw_spin_rq_lock(rq);
618
		if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) {
619
			rq_pin_lock(rq, rf);
620 621
			return rq;
		}
622
		raw_spin_rq_unlock(rq);
623 624 625 626 627 628 629 630 631

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

/*
 * task_rq_lock - lock p->pi_lock and lock the rq @p resides on.
 */
632
struct rq *task_rq_lock(struct task_struct *p, struct rq_flags *rf)
633 634 635 636 637 638
	__acquires(p->pi_lock)
	__acquires(rq->lock)
{
	struct rq *rq;

	for (;;) {
639
		raw_spin_lock_irqsave(&p->pi_lock, rf->flags);
640
		rq = task_rq(p);
641
		raw_spin_rq_lock(rq);
642 643 644 645 646 647 648 649 650 651
		/*
		 *	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)
		 *
652
		 * If we observe the old CPU in task_rq_lock(), the acquire of
653 654
		 * the old rq->lock will fully serialize against the stores.
		 *
655 656 657
		 * 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.
658 659
		 */
		if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) {
660
			rq_pin_lock(rq, rf);
661 662
			return rq;
		}
663
		raw_spin_rq_unlock(rq);
664
		raw_spin_unlock_irqrestore(&p->pi_lock, rf->flags);
665 666 667 668 669 670

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

671 672 673 674 675 676 677 678 679 680
/*
 * 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...
 */
681 682
	s64 __maybe_unused steal = 0, irq_delta = 0;

683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721
#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;

722
#ifdef CONFIG_HAVE_SCHED_AVG_IRQ
723
	if ((irq_delta + steal) && sched_feat(NONTASK_CAPACITY))
724
		update_irq_load_avg(rq, irq_delta + steal);
725
#endif
726
	update_rq_clock_pelt(rq, delta);
727 728 729 730 731 732
}

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

733
	lockdep_assert_rq_held(rq);
734 735 736 737 738

	if (rq->clock_update_flags & RQCF_ACT_SKIP)
		return;

#ifdef CONFIG_SCHED_DEBUG
739 740
	if (sched_feat(WARN_DOUBLE_CLOCK))
		SCHED_WARN_ON(rq->clock_update_flags & RQCF_UPDATED);
741 742
	rq->clock_update_flags |= RQCF_UPDATED;
#endif
743

744 745 746 747 748 749 750
	delta = sched_clock_cpu(cpu_of(rq)) - rq->clock;
	if (delta < 0)
		return;
	rq->clock += delta;
	update_rq_clock_task(rq, delta);
}

751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768
#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);
769
	struct rq_flags rf;
770 771 772

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

773
	rq_lock(rq, &rf);
774
	update_rq_clock(rq);
775
	rq->curr->sched_class->task_tick(rq, rq->curr, 1);
776
	rq_unlock(rq, &rf);
777 778 779 780

	return HRTIMER_NORESTART;
}

781
#ifdef CONFIG_SMP
Peter Zijlstra's avatar
Peter Zijlstra committed
782

783
static void __hrtick_restart(struct rq *rq)
Peter Zijlstra's avatar
Peter Zijlstra committed
784 785
{
	struct hrtimer *timer = &rq->hrtick_timer;
786
	ktime_t time = rq->hrtick_time;
Peter Zijlstra's avatar
Peter Zijlstra committed
787

788
	hrtimer_start(timer, time, HRTIMER_MODE_ABS_PINNED_HARD);
Peter Zijlstra's avatar
Peter Zijlstra committed
789 790
}

791 792 793 794
/*
 * called from hardirq (IPI) context
 */
static void __hrtick_start(void *arg)
795
{
796
	struct rq *rq = arg;
797
	struct rq_flags rf;
798

799
	rq_lock(rq, &rf);
Peter Zijlstra's avatar
Peter Zijlstra committed
800
	__hrtick_restart(rq);
801
	rq_unlock(rq, &rf);
802 803
}

804 805 806 807 808
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
809
void hrtick_start(struct rq *rq, u64 delay)
810
{
811
	struct hrtimer *timer = &rq->hrtick_timer;
812 813 814 815 816 817 818
	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);
819
	rq->hrtick_time = ktime_add_ns(timer->base->get_time(), delta);
820

821
	if (rq == this_rq())
Peter Zijlstra's avatar
Peter Zijlstra committed
822
		__hrtick_restart(rq);
823
	else
824
		smp_call_function_single_async(cpu_of(rq), &rq->hrtick_csd);
825 826
}

827 828 829 830 831 832
#else
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
833
void hrtick_start(struct rq *rq, u64 delay)
834
{
835 836 837 838 839
	/*
	 * Don't schedule slices shorter than 10000ns, that just
	 * doesn't make sense. Rely on vruntime for fairness.
	 */
	delay = max_t(u64, delay, 10000LL);
840
	hrtimer_start(&rq->hrtick_timer, ns_to_ktime(delay),
841
		      HRTIMER_MODE_REL_PINNED_HARD);
842
}
843

844
#endif /* CONFIG_SMP */
845

846
static void hrtick_rq_init(struct rq *rq)
847
{
848
#ifdef CONFIG_SMP
849
	INIT_CSD(&rq->hrtick_csd, __hrtick_start, rq);
850
#endif
851
	hrtimer_init(&rq->hrtick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL_HARD);
852
	rq->hrtick_timer.function = hrtick;
853
}
Andrew Morton's avatar
Andrew Morton committed
854
#else	/* CONFIG_SCHED_HRTICK */
855 856 857 858
static inline void hrtick_clear(struct rq *rq)
{
}

859
static inline void hrtick_rq_init(struct rq *rq)
860 861
{
}
Andrew Morton's avatar
Andrew Morton committed
862
#endif	/* CONFIG_SCHED_HRTICK */
863

864 865 866 867 868 869 870
/*
 * 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);				\
871
		typeof(*_ptr) _val = *_ptr;				\
872
									\
873 874 875
		do {							\
		} while (!try_cmpxchg(_ptr, &_val, _val | _mask));	\
	_val;								\
876 877
})

878
#if defined(CONFIG_SMP) && defined(TIF_POLLING_NRFLAG)
879 880 881 882 883
/*
 * Atomically set TIF_NEED_RESCHED and test for TIF_POLLING_NRFLAG,
 * this avoids any races wrt polling state changes and thereby avoids
 * spurious IPIs.
 */
884
static inline bool set_nr_and_not_polling(struct task_struct *p)
885 886 887 888
{
	struct thread_info *ti = task_thread_info(p);
	return !(fetch_or(&ti->flags, _TIF_NEED_RESCHED) & _TIF_POLLING_NRFLAG);
}
889 890 891 892 893 894 895 896 897 898

/*
 * 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);
899
	typeof(ti->flags) val = READ_ONCE(ti->flags);
900 901 902 903 904 905

	for (;;) {
		if (!(val & _TIF_POLLING_NRFLAG))
			return false;
		if (val & _TIF_NEED_RESCHED)
			return true;
906
		if (try_cmpxchg(&ti->flags, &val, val | _TIF_NEED_RESCHED))
907 908 909 910 911
			break;
	}
	return true;
}

912
#else
913
static inline bool set_nr_and_not_polling(struct task_struct *p)
914 915 916 917
{
	set_tsk_need_resched(p);
	return true;
}
918 919

#ifdef CONFIG_SMP
920
static inline bool set_nr_if_polling(struct task_struct *p)
921 922 923 924
{
	return false;
}
#endif
925 926
#endif

927
static bool __wake_q_add(struct wake_q_head *head, struct task_struct *task)
928 929 930 931 932
{
	struct wake_q_node *node = &task->wake_q;

	/*
	 * Atomically grab the task, if ->wake_q is !nil already it means
933
	 * it's already queued (either by us or someone else) and will get the
934 935
	 * wakeup due to that.
	 *
936 937
	 * 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.
938
	 */
939
	smp_mb__before_atomic();
940
	if (unlikely(cmpxchg_relaxed(&node->next, NULL, WAKE_Q_TAIL)))
941
		return false;
942 943 944 945 946 947

	/*
	 * The head is context local, there can be no concurrency.
	 */
	*head->lastp = node;
	head->lastp = &node->next;
948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989
	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);
990 991 992 993 994 995 996 997 998 999
}

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);
1000
		/* Task can safely be re-inserted now: */
1001 1002 1003 1004
		node = node->next;
		task->wake_q.next = NULL;

		/*
1005 1006
		 * wake_up_process() executes a full barrier, which pairs with
		 * the queueing in wake_q_add() so as not to miss wakeups.
1007 1008 1009 1010 1011 1012
		 */
		wake_up_process(task);
		put_task_struct(task);
	}
}

1013
/*
1014
 * resched_curr - mark rq's current task 'to be rescheduled now'.
1015 1016 1017 1018 1019
 *
 * 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.
 */
1020
void resched_curr(struct rq *rq)
1021
{
1022
	struct task_struct *curr = rq->curr;
1023 1024
	int cpu;

1025
	lockdep_assert_rq_held(rq);
1026

1027
	if (test_tsk_need_resched(curr))
1028 1029
		return;

1030
	cpu = cpu_of(rq);
1031

1032
	if (cpu == smp_processor_id()) {
1033
		set_tsk_need_resched(curr);
1034
		set_preempt_need_resched();
1035
		return;
1036
	}
1037

1038
	if (set_nr_and_not_polling(curr))
1039
		smp_send_reschedule(cpu);
1040 1041
	else
		trace_sched_wake_idle_without_ipi(cpu);
1042 1043
}

1044
void resched_cpu(int cpu)
1045 1046 1047 1048
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

1049
	raw_spin_rq_lock_irqsave(rq, flags);
1050 1051
	if (cpu_online(cpu) || cpu == smp_processor_id())
		resched_curr(rq);
1052
	raw_spin_rq_unlock_irqrestore(rq, flags);
1053
}
1054

1055
#ifdef CONFIG_SMP
1056
#ifdef CONFIG_NO_HZ_COMMON
1057
/*
1058 1059
 * In the semi idle case, use the nearest busy CPU for migrating timers
 * from an idle CPU.  This is good for power-savings.
1060 1061
 *
 * We don't do similar optimization for completely idle system, as
1062 1063
 * 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).
1064
 */
1065
int get_nohz_timer_target(void)
1066
{
1067
	int i, cpu = smp_processor_id(), default_cpu = -1;
1068
	struct sched_domain *sd;
1069
	const struct cpumask *hk_mask;
1070

1071
	if (housekeeping_cpu(cpu, HK_TYPE_TIMER)) {
1072 1073 1074 1075
		if (!idle_cpu(cpu))
			return cpu;
		default_cpu = cpu;
	}
1076

1077
	hk_mask = housekeeping_cpumask(HK_TYPE_TIMER);
1078

1079
	rcu_read_lock();
1080
	for_each_domain(cpu, sd) {
1081
		for_each_cpu_and(i, sched_domain_span(sd), hk_mask) {
1082 1083 1084
			if (cpu == i)
				continue;

1085
			if (!idle_cpu(i)) {
1086 1087 1088 1089
				cpu = i;
				goto unlock;
			}
		}
1090
	}
1091

1092
	if (default_cpu == -1)
1093
		default_cpu = housekeeping_any_cpu(HK_TYPE_TIMER);
1094
	cpu = default_cpu;
1095 1096
unlock:
	rcu_read_unlock();
1097 1098
	return cpu;
}
1099

1100 1101 1102 1103 1104 1105 1106 1107 1108 1109
/*
 * 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.
 */
1110
static void wake_up_idle_cpu(int cpu)
1111 1112 1113 1114 1115 1116
{
	struct rq *rq = cpu_rq(cpu);

	if (cpu == smp_processor_id())
		return;

1117
	if (set_nr_and_not_polling(rq->idle))
1118
		smp_send_reschedule(cpu);
1119 1120
	else
		trace_sched_wake_idle_without_ipi(cpu);
1121 1122
}

1123
static bool wake_up_full_nohz_cpu(int cpu)
1124
{
1125 1126 1127 1128 1129 1130
	/*
	 * 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.
	 */
1131 1132
	if (cpu_is_offline(cpu))
		return true;  /* Don't try to wake offline CPUs. */
1133
	if (tick_nohz_full_cpu(cpu)) {
1134 1135
		if (cpu != smp_processor_id() ||
		    tick_nohz_tick_stopped())
1136
			tick_nohz_full_kick_cpu(cpu);
1137 1138 1139 1140 1141 1142
		return true;
	}

	return false;
}

1143 1144 1145 1146 1147
/*
 * 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.
 */
1148 1149
void wake_up_nohz_cpu(int cpu)
{
1150
	if (!wake_up_full_nohz_cpu(cpu))
1151 1152 1153
		wake_up_idle_cpu(cpu);
}

1154
static void nohz_csd_func(void *info)
1155
{
1156 1157 1158
	struct rq *rq = info;
	int cpu = cpu_of(rq);
	unsigned int flags;
1159 1160

	/*
1161
	 * Release the rq::nohz_csd.
1162
	 */
1163
	flags = atomic_fetch_andnot(NOHZ_KICK_MASK | NOHZ_NEWILB_KICK, nohz_flags(cpu));
1164
	WARN_ON(!(flags & NOHZ_KICK_MASK));
1165

1166 1167 1168
	rq->idle_balance = idle_cpu(cpu);
	if (rq->idle_balance && !need_resched()) {
		rq->nohz_idle_balance = flags;
1169 1170
		raise_softirq_irqoff(SCHED_SOFTIRQ);
	}
1171 1172
}

1173
#endif /* CONFIG_NO_HZ_COMMON */
1174

1175
#ifdef CONFIG_NO_HZ_FULL
1176
bool sched_can_stop_tick(struct rq *rq)
1177
{
1178 1179 1180 1181 1182 1183
	int fifo_nr_running;

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

1184
	/*
1185
	 * If there are more than one RR tasks, we need the tick to affect the
1186
	 * actual RR behaviour.
1187
	 */
1188 1189 1190 1191 1192
	if (rq->rt.rr_nr_running) {
		if (rq->rt.rr_nr_running == 1)
			return true;
		else
			return false;
1193 1194
	}

1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208
	/*
	 * 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)
1209
		return false;
1210

1211
	return true;
1212 1213
}
#endif /* CONFIG_NO_HZ_FULL */
1214
#endif /* CONFIG_SMP */
1215

1216 1217
#if defined(CONFIG_RT_GROUP_SCHED) || (defined(CONFIG_FAIR_GROUP_SCHED) && \
			(defined(CONFIG_SMP) || defined(CONFIG_CFS_BANDWIDTH)))
1218
/*
1219 1220 1221 1222
 * 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.
1223
 */
1224
int walk_tg_tree_from(struct task_group *from,
1225
			     tg_visitor down, tg_visitor up, void *data)
1226 1227
{
	struct task_group *parent, *child;
1228
	int ret;
1229

1230 1231
	parent = from;

1232
down:
1233 1234
	ret = (*down)(parent, data);
	if (ret)
1235
		goto out;
1236 1237 1238 1239 1240 1241 1242
	list_for_each_entry_rcu(child, &parent->children, siblings) {
		parent = child;
		goto down;

up:
		continue;
	}
1243
	ret = (*up)(parent, data);
1244 1245
	if (ret || parent == from)
		goto out;
1246 1247 1248 1249 1250

	child = parent;
	parent = parent->parent;
	if (parent)
		goto up;
1251
out:
1252
	return ret;
1253 1254
}

1255
int tg_nop(struct task_group *tg, void *data)
1256
{
1257
	return 0;
1258
}
1259 1260
#endif

1261
static void set_load_weight(struct task_struct *p, bool update_load)
1262
{
1263 1264 1265
	int prio = p->static_prio - MAX_RT_PRIO;
	struct load_weight *load = &p->se.load;

Ingo Molnar's avatar
Ingo Molnar committed
1266 1267 1268
	/*
	 * SCHED_IDLE tasks get minimal weight:
	 */
1269
	if (task_has_idle_policy(p)) {
1270
		load->weight = scale_load(WEIGHT_IDLEPRIO);
1271
		load->inv_weight = WMULT_IDLEPRIO;
Ingo Molnar's avatar
Ingo Molnar committed
1272 1273
		return;
	}
1274

1275 1276 1277 1278 1279 1280 1281 1282 1283 1284
	/*
	 * 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];
	}
1285 1286
}

1287
#ifdef CONFIG_UCLAMP_TASK
1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299
/*
 * 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);

1300
/* Max allowed minimum utilization */
1301
static unsigned int __maybe_unused sysctl_sched_uclamp_util_min = SCHED_CAPACITY_SCALE;
1302 1303

/* Max allowed maximum utilization */
1304
static unsigned int __maybe_unused sysctl_sched_uclamp_util_max = SCHED_CAPACITY_SCALE;
1305

1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320
/*
 * 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.
 */
1321
static unsigned int sysctl_sched_uclamp_util_min_rt_default = SCHED_CAPACITY_SCALE;
1322

1323 1324
/* All clamps are required to be less or equal than these values */
static struct uclamp_se uclamp_default[UCLAMP_CNT];
1325

1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345
/*
 * 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);

1346 1347 1348 1349 1350 1351 1352 1353
/* 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)
{
1354
	return min_t(unsigned int, clamp_value / UCLAMP_BUCKET_DELTA, UCLAMP_BUCKETS - 1);
1355 1356
}

1357
static inline unsigned int uclamp_none(enum uclamp_id clamp_id)
1358 1359 1360 1361 1362 1363
{
	if (clamp_id == UCLAMP_MIN)
		return 0;
	return SCHED_CAPACITY_SCALE;
}

1364 1365
static inline void uclamp_se_set(struct uclamp_se *uc_se,
				 unsigned int value, bool user_defined)
1366 1367 1368
{
	uc_se->value = value;
	uc_se->bucket_id = uclamp_bucket_id(value);
1369
	uc_se->user_defined = user_defined;
1370 1371
}

1372
static inline unsigned int
1373
uclamp_idle_value(struct rq *rq, enum uclamp_id clamp_id,
1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388
		  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);
}

1389
static inline void uclamp_idle_reset(struct rq *rq, enum uclamp_id clamp_id,
1390 1391 1392 1393 1394 1395 1396 1397 1398
				     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);
}

1399
static inline
1400
unsigned int uclamp_rq_max_value(struct rq *rq, enum uclamp_id clamp_id,
1401
				   unsigned int clamp_value)
1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416
{
	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 */
1417
	return uclamp_idle_value(rq, clamp_id, clamp_value);
1418 1419
}

1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450
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);
}

1451
static inline struct uclamp_se
1452
uclamp_tg_restrict(struct task_struct *p, enum uclamp_id clamp_id)
1453
{
1454
	/* Copy by value as we could modify it */
1455 1456
	struct uclamp_se uc_req = p->uclamp_req[clamp_id];
#ifdef CONFIG_UCLAMP_TASK_GROUP
1457
	unsigned int tg_min, tg_max, value;
1458 1459 1460 1461 1462 1463 1464 1465 1466 1467

	/*
	 * 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;

1468 1469 1470 1471 1472
	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);
1473 1474 1475 1476 1477
#endif

	return uc_req;
}

1478 1479 1480 1481
/*
 * The effective clamp bucket index of a task depends on, by increasing
 * priority:
 * - the task specific clamp value, when explicitly requested from userspace
1482 1483
 * - the task group effective clamp value, for tasks not either in the root
 *   group or in an autogroup
1484 1485 1486
 * - the system default clamp value, defined by the sysadmin
 */
static inline struct uclamp_se
1487
uclamp_eff_get(struct task_struct *p, enum uclamp_id clamp_id)
1488
{
1489
	struct uclamp_se uc_req = uclamp_tg_restrict(p, clamp_id);
1490 1491 1492 1493 1494 1495 1496 1497 1498
	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;
}

1499
unsigned long uclamp_eff_value(struct task_struct *p, enum uclamp_id clamp_id)
1500 1501 1502 1503 1504
{
	struct uclamp_se uc_eff;

	/* Task currently refcounted: use back-annotated (effective) value */
	if (p->uclamp[clamp_id].active)
1505
		return (unsigned long)p->uclamp[clamp_id].value;
1506 1507 1508

	uc_eff = uclamp_eff_get(p, clamp_id);

1509
	return (unsigned long)uc_eff.value;
1510 1511
}

1512 1513 1514 1515
/*
 * 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.
1516 1517 1518 1519 1520
 *
 * 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.
1521 1522
 */
static inline void uclamp_rq_inc_id(struct rq *rq, struct task_struct *p,
1523
				    enum uclamp_id clamp_id)
1524 1525 1526 1527 1528
{
	struct uclamp_rq *uc_rq = &rq->uclamp[clamp_id];
	struct uclamp_se *uc_se = &p->uclamp[clamp_id];
	struct uclamp_bucket *bucket;

1529
	lockdep_assert_rq_held(rq);
1530

1531 1532 1533
	/* Update task effective clamp */
	p->uclamp[clamp_id] = uclamp_eff_get(p, clamp_id);

1534 1535
	bucket = &uc_rq->bucket[uc_se->bucket_id];
	bucket->tasks++;
1536
	uc_se->active = true;
1537

1538 1539
	uclamp_idle_reset(rq, clamp_id, uc_se->value);

1540 1541 1542 1543 1544 1545 1546
	/*
	 * 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;

1547
	if (uc_se->value > READ_ONCE(uc_rq->value))
1548
		WRITE_ONCE(uc_rq->value, uc_se->value);
1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560
}

/*
 * 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,
1561
				    enum uclamp_id clamp_id)
1562 1563 1564 1565
{
	struct uclamp_rq *uc_rq = &rq->uclamp[clamp_id];
	struct uclamp_se *uc_se = &p->uclamp[clamp_id];
	struct uclamp_bucket *bucket;
1566
	unsigned int bkt_clamp;
1567 1568
	unsigned int rq_clamp;

1569
	lockdep_assert_rq_held(rq);
1570

1571 1572 1573 1574 1575 1576 1577 1578
	/*
	 * 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.
	 *
1579
	 * Need to be careful of the following enqueue/dequeue ordering
1580 1581 1582 1583 1584 1585
	 * problem too
	 *
	 *	enqueue(taskA)
	 *	// sched_uclamp_used gets enabled
	 *	enqueue(taskB)
	 *	dequeue(taskA)
1586
	 *	// Must not decrement bucket->tasks here
1587 1588 1589 1590 1591 1592 1593 1594 1595 1596
	 *	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;

1597
	bucket = &uc_rq->bucket[uc_se->bucket_id];
1598

1599 1600 1601
	SCHED_WARN_ON(!bucket->tasks);
	if (likely(bucket->tasks))
		bucket->tasks--;
1602

1603
	uc_se->active = false;
1604

1605 1606 1607 1608 1609 1610
	/*
	 * 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.
	 */
1611 1612 1613 1614 1615 1616 1617 1618 1619
	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);
1620 1621 1622 1623
	if (bucket->value >= rq_clamp) {
		bkt_clamp = uclamp_rq_max_value(rq, clamp_id, uc_se->value);
		WRITE_ONCE(uc_rq->value, bkt_clamp);
	}
1624 1625 1626 1627
}

static inline void uclamp_rq_inc(struct rq *rq, struct task_struct *p)
{
1628
	enum uclamp_id clamp_id;
1629

1630 1631 1632 1633 1634 1635 1636 1637 1638
	/*
	 * 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;

1639 1640 1641 1642 1643
	if (unlikely(!p->sched_class->uclamp_enabled))
		return;

	for_each_clamp_id(clamp_id)
		uclamp_rq_inc_id(rq, p, clamp_id);
1644 1645 1646 1647

	/* Reset clamp idle holding when there is one RUNNABLE task */
	if (rq->uclamp_flags & UCLAMP_FLAG_IDLE)
		rq->uclamp_flags &= ~UCLAMP_FLAG_IDLE;
1648 1649 1650 1651
}

static inline void uclamp_rq_dec(struct rq *rq, struct task_struct *p)
{
1652
	enum uclamp_id clamp_id;
1653

1654 1655 1656 1657 1658 1659 1660 1661 1662
	/*
	 * 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;

1663 1664 1665 1666 1667 1668 1669
	if (unlikely(!p->sched_class->uclamp_enabled))
		return;

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

1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686
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;
}

1687
static inline void
1688
uclamp_update_active(struct task_struct *p)
1689
{
1690
	enum uclamp_id clamp_id;
1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709
	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.
	 */
1710 1711
	for_each_clamp_id(clamp_id)
		uclamp_rq_reinc_id(rq, p, clamp_id);
1712 1713 1714 1715

	task_rq_unlock(rq, p, &rf);
}

1716
#ifdef CONFIG_UCLAMP_TASK_GROUP
1717
static inline void
1718
uclamp_update_active_tasks(struct cgroup_subsys_state *css)
1719 1720 1721 1722 1723
{
	struct css_task_iter it;
	struct task_struct *p;

	css_task_iter_start(css, 0, &it);
1724 1725
	while ((p = css_task_iter_next(&it)))
		uclamp_update_active(p);
1726 1727 1728
	css_task_iter_end(&it);
}

1729
static void cpu_util_update_eff(struct cgroup_subsys_state *css);
1730 1731 1732 1733 1734
#endif

#ifdef CONFIG_SYSCTL
#ifdef CONFIG_UCLAMP_TASK
#ifdef CONFIG_UCLAMP_TASK_GROUP
1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751
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

1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778
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();
}

1779
static int sysctl_sched_uclamp_handler(struct ctl_table *table, int write,
1780
				void *buffer, size_t *lenp, loff_t *ppos)
1781
{
1782
	bool update_root_tg = false;
1783
	int old_min, old_max, old_min_rt;
1784 1785
	int result;

1786
	mutex_lock(&uclamp_mutex);
1787 1788
	old_min = sysctl_sched_uclamp_util_min;
	old_max = sysctl_sched_uclamp_util_max;
1789
	old_min_rt = sysctl_sched_uclamp_util_min_rt_default;
1790 1791 1792 1793 1794 1795 1796 1797

	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 ||
1798 1799 1800
	    sysctl_sched_uclamp_util_max > SCHED_CAPACITY_SCALE	||
	    sysctl_sched_uclamp_util_min_rt_default > SCHED_CAPACITY_SCALE) {

1801 1802 1803 1804 1805 1806
		result = -EINVAL;
		goto undo;
	}

	if (old_min != sysctl_sched_uclamp_util_min) {
		uclamp_se_set(&uclamp_default[UCLAMP_MIN],
1807
			      sysctl_sched_uclamp_util_min, false);
1808
		update_root_tg = true;
1809 1810 1811
	}
	if (old_max != sysctl_sched_uclamp_util_max) {
		uclamp_se_set(&uclamp_default[UCLAMP_MAX],
1812
			      sysctl_sched_uclamp_util_max, false);
1813
		update_root_tg = true;
1814 1815
	}

1816 1817
	if (update_root_tg) {
		static_branch_enable(&sched_uclamp_used);
1818
		uclamp_update_root_tg();
1819
	}
1820

1821 1822 1823 1824
	if (old_min_rt != sysctl_sched_uclamp_util_min_rt_default) {
		static_branch_enable(&sched_uclamp_used);
		uclamp_sync_util_min_rt_default();
	}
1825

1826
	/*
1827 1828 1829
	 * 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.
1830
	 */
1831

1832 1833 1834 1835 1836
	goto done;

undo:
	sysctl_sched_uclamp_util_min = old_min;
	sysctl_sched_uclamp_util_max = old_max;
1837
	sysctl_sched_uclamp_util_min_rt_default = old_min_rt;
1838
done:
1839
	mutex_unlock(&uclamp_mutex);
1840 1841 1842

	return result;
}
1843 1844
#endif
#endif
1845

1846 1847 1848
static int uclamp_validate(struct task_struct *p,
			   const struct sched_attr *attr)
{
1849 1850
	int util_min = p->uclamp_req[UCLAMP_MIN].value;
	int util_max = p->uclamp_req[UCLAMP_MAX].value;
1851

1852 1853
	if (attr->sched_flags & SCHED_FLAG_UTIL_CLAMP_MIN) {
		util_min = attr->sched_util_min;
1854

1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866
		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)
1867 1868
		return -EINVAL;

1869 1870 1871 1872 1873 1874 1875 1876 1877
	/*
	 * 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);

1878 1879 1880
	return 0;
}

1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905
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;
}

1906 1907 1908
static void __setscheduler_uclamp(struct task_struct *p,
				  const struct sched_attr *attr)
{
1909
	enum uclamp_id clamp_id;
1910 1911 1912

	for_each_clamp_id(clamp_id) {
		struct uclamp_se *uc_se = &p->uclamp_req[clamp_id];
1913
		unsigned int value;
1914

1915
		if (!uclamp_reset(attr, clamp_id, uc_se))
1916 1917
			continue;

1918 1919 1920 1921
		/*
		 * RT by default have a 100% boost value that could be modified
		 * at runtime.
		 */
1922
		if (unlikely(rt_task(p) && clamp_id == UCLAMP_MIN))
1923
			value = sysctl_sched_uclamp_util_min_rt_default;
1924
		else
1925 1926 1927
			value = uclamp_none(clamp_id);

		uclamp_se_set(uc_se, value, false);
1928 1929 1930

	}

1931 1932 1933
	if (likely(!(attr->sched_flags & SCHED_FLAG_UTIL_CLAMP)))
		return;

1934 1935
	if (attr->sched_flags & SCHED_FLAG_UTIL_CLAMP_MIN &&
	    attr->sched_util_min != -1) {
1936 1937 1938 1939
		uclamp_se_set(&p->uclamp_req[UCLAMP_MIN],
			      attr->sched_util_min, true);
	}

1940 1941
	if (attr->sched_flags & SCHED_FLAG_UTIL_CLAMP_MAX &&
	    attr->sched_util_max != -1) {
1942 1943 1944 1945 1946
		uclamp_se_set(&p->uclamp_req[UCLAMP_MAX],
			      attr->sched_util_max, true);
	}
}

1947 1948
static void uclamp_fork(struct task_struct *p)
{
1949
	enum uclamp_id clamp_id;
1950

1951 1952 1953 1954
	/*
	 * We don't need to hold task_rq_lock() when updating p->uclamp_* here
	 * as the task is still at its early fork stages.
	 */
1955 1956
	for_each_clamp_id(clamp_id)
		p->uclamp[clamp_id].active = false;
1957 1958 1959 1960 1961

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

	for_each_clamp_id(clamp_id) {
1962 1963
		uclamp_se_set(&p->uclamp_req[clamp_id],
			      uclamp_none(clamp_id), false);
1964
	}
1965 1966
}

1967 1968 1969 1970 1971
static void uclamp_post_fork(struct task_struct *p)
{
	uclamp_update_util_min_rt_default(p);
}

1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982
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)
		};
	}

1983
	rq->uclamp_flags = UCLAMP_FLAG_IDLE;
1984 1985
}

1986 1987
static void __init init_uclamp(void)
{
1988
	struct uclamp_se uc_max = {};
1989
	enum uclamp_id clamp_id;
1990 1991
	int cpu;

1992 1993
	for_each_possible_cpu(cpu)
		init_uclamp_rq(cpu_rq(cpu));
1994 1995

	for_each_clamp_id(clamp_id) {
1996
		uclamp_se_set(&init_task.uclamp_req[clamp_id],
1997
			      uclamp_none(clamp_id), false);
1998
	}
1999 2000

	/* System defaults allow max clamp values for both indexes */
2001
	uclamp_se_set(&uc_max, uclamp_none(UCLAMP_MAX), false);
2002
	for_each_clamp_id(clamp_id) {
2003
		uclamp_default[clamp_id] = uc_max;
2004 2005
#ifdef CONFIG_UCLAMP_TASK_GROUP
		root_task_group.uclamp_req[clamp_id] = uc_max;
2006
		root_task_group.uclamp[clamp_id] = uc_max;
2007 2008
#endif
	}
2009 2010 2011 2012 2013
}

#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) { }
2014 2015 2016 2017 2018 2019 2020
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) { }
2021
static inline void uclamp_fork(struct task_struct *p) { }
2022
static inline void uclamp_post_fork(struct task_struct *p) { }
2023 2024 2025
static inline void init_uclamp(void) { }
#endif /* CONFIG_UCLAMP_TASK */

2026 2027 2028 2029 2030
bool sched_task_on_rq(struct task_struct *p)
{
	return task_on_rq_queued(p);
}

2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049
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;
}

2050
static inline void enqueue_task(struct rq *rq, struct task_struct *p, int flags)
2051
{
2052 2053 2054
	if (!(flags & ENQUEUE_NOCLOCK))
		update_rq_clock(rq);

2055
	if (!(flags & ENQUEUE_RESTORE)) {
2056
		sched_info_enqueue(rq, p);
2057 2058
		psi_enqueue(p, flags & ENQUEUE_WAKEUP);
	}
2059

2060
	uclamp_rq_inc(rq, p);
2061
	p->sched_class->enqueue_task(rq, p, flags);
2062 2063 2064

	if (sched_core_enabled(rq))
		sched_core_enqueue(rq, p);
2065 2066
}

2067
static inline void dequeue_task(struct rq *rq, struct task_struct *p, int flags)
2068
{
2069
	if (sched_core_enabled(rq))
2070
		sched_core_dequeue(rq, p, flags);
2071

2072 2073 2074
	if (!(flags & DEQUEUE_NOCLOCK))
		update_rq_clock(rq);

2075
	if (!(flags & DEQUEUE_SAVE)) {
2076
		sched_info_dequeue(rq, p);
2077 2078
		psi_dequeue(p, flags & DEQUEUE_SLEEP);
	}
2079

2080
	uclamp_rq_dec(rq, p);
2081
	p->sched_class->dequeue_task(rq, p, flags);
2082 2083
}

2084
void activate_task(struct rq *rq, struct task_struct *p, int flags)
2085
{
2086
	enqueue_task(rq, p, flags);
2087 2088

	p->on_rq = TASK_ON_RQ_QUEUED;
2089 2090
}

2091
void deactivate_task(struct rq *rq, struct task_struct *p, int flags)
2092
{
2093 2094
	p->on_rq = (flags & DEQUEUE_SLEEP) ? 0 : TASK_ON_RQ_MIGRATING;

2095
	dequeue_task(rq, p, flags);
2096 2097
}

2098
static inline int __normal_prio(int policy, int rt_prio, int nice)
2099
{
2100 2101 2102 2103 2104 2105 2106 2107 2108 2109
	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;
2110 2111
}

2112 2113 2114 2115 2116 2117 2118
/*
 * 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.
 */
2119
static inline int normal_prio(struct task_struct *p)
2120
{
2121
	return __normal_prio(p->policy, p->rt_priority, PRIO_TO_NICE(p->static_prio));
2122 2123 2124 2125 2126 2127 2128 2129 2130
}

/*
 * 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.
 */
2131
static int effective_prio(struct task_struct *p)
2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143
{
	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
2144 2145 2146
/**
 * task_curr - is this task currently executing on a CPU?
 * @p: the task in question.
2147 2148
 *
 * Return: 1 if the task is currently executing. 0 otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
2149
 */
2150
inline int task_curr(const struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
2151 2152 2153 2154
{
	return cpu_curr(task_cpu(p)) == p;
}

2155
/*
2156 2157 2158 2159 2160
 * 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().
2161
 */
2162 2163
static inline void check_class_changed(struct rq *rq, struct task_struct *p,
				       const struct sched_class *prev_class,
2164
				       int oldprio)
2165 2166 2167
{
	if (prev_class != p->sched_class) {
		if (prev_class->switched_from)
2168
			prev_class->switched_from(rq, p);
2169

2170
		p->sched_class->switched_to(rq, p);
2171
	} else if (oldprio != p->prio || dl_task(p))
2172
		p->sched_class->prio_changed(rq, p, oldprio);
2173 2174
}

2175
void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags)
2176
{
2177
	if (p->sched_class == rq->curr->sched_class)
2178
		rq->curr->sched_class->check_preempt_curr(rq, p, flags);
2179
	else if (sched_class_above(p->sched_class, rq->curr->sched_class))
2180
		resched_curr(rq);
2181 2182 2183 2184 2185

	/*
	 * A queue event has occurred, and we're going to schedule.  In
	 * this case, we can save a useless back to back clock update.
	 */
2186
	if (task_on_rq_queued(rq->curr) && test_tsk_need_resched(rq->curr))
2187
		rq_clock_skip_update(rq);
2188 2189
}

Linus Torvalds's avatar
Linus Torvalds committed
2190
#ifdef CONFIG_SMP
2191

2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214
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)
{
2215 2216 2217 2218
	struct task_struct *p = current;

	if (p->migration_disabled) {
		p->migration_disabled++;
2219
		return;
2220
	}
2221

2222 2223 2224 2225
	preempt_disable();
	this_rq()->nr_pinned++;
	p->migration_disabled = 1;
	preempt_enable();
2226 2227 2228 2229 2230 2231 2232
}
EXPORT_SYMBOL_GPL(migrate_disable);

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

2233 2234
	if (p->migration_disabled > 1) {
		p->migration_disabled--;
2235
		return;
2236
	}
2237

2238 2239 2240
	if (WARN_ON_ONCE(!p->migration_disabled))
		return;

2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252
	/*
	 * 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.
	 */
2253
	barrier();
2254
	p->migration_disabled = 0;
2255
	this_rq()->nr_pinned--;
2256
	preempt_enable();
2257 2258 2259
}
EXPORT_SYMBOL_GPL(migrate_enable);

2260 2261 2262 2263 2264
static inline bool rq_has_pinned_tasks(struct rq *rq)
{
	return rq->nr_pinned;
}

2265
/*
2266
 * Per-CPU kthreads are allowed to run on !active && online CPUs, see
2267 2268 2269 2270
 * __set_cpus_allowed_ptr() and select_fallback_rq().
 */
static inline bool is_cpu_allowed(struct task_struct *p, int cpu)
{
2271
	/* When not in the task's cpumask, no point in looking further. */
2272
	if (!cpumask_test_cpu(cpu, p->cpus_ptr))
2273 2274
		return false;

2275 2276
	/* migrate_disabled() must be allowed to finish. */
	if (is_migration_disabled(p))
2277 2278
		return cpu_online(cpu);

2279 2280
	/* Non kernel threads are not allowed during either online or offline. */
	if (!(p->flags & PF_KTHREAD))
2281
		return cpu_active(cpu) && task_cpu_possible(cpu, p);
2282 2283 2284 2285 2286 2287

	/* 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. */
2288
	if (cpu_dying(cpu))
2289 2290 2291 2292
		return false;

	/* But are allowed during online. */
	return cpu_online(cpu);
2293 2294
}

Peter Zijlstra's avatar
Peter Zijlstra committed
2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313
/*
 * 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.
 */
2314 2315
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
2316
{
2317
	lockdep_assert_rq_held(rq);
Peter Zijlstra's avatar
Peter Zijlstra committed
2318

2319
	deactivate_task(rq, p, DEQUEUE_NOCLOCK);
Peter Zijlstra's avatar
Peter Zijlstra committed
2320
	set_task_cpu(p, new_cpu);
2321
	rq_unlock(rq, rf);
Peter Zijlstra's avatar
Peter Zijlstra committed
2322 2323 2324

	rq = cpu_rq(new_cpu);

2325
	rq_lock(rq, rf);
2326
	WARN_ON_ONCE(task_cpu(p) != new_cpu);
2327
	activate_task(rq, p, 0);
Peter Zijlstra's avatar
Peter Zijlstra committed
2328 2329 2330 2331 2332 2333
	check_preempt_curr(rq, p, 0);

	return rq;
}

struct migration_arg {
2334 2335 2336 2337 2338
	struct task_struct		*task;
	int				dest_cpu;
	struct set_affinity_pending	*pending;
};

2339 2340 2341 2342
/*
 * @refs: number of wait_for_completion()
 * @stop_pending: is @stop_work in use
 */
2343 2344
struct set_affinity_pending {
	refcount_t		refs;
2345
	unsigned int		stop_pending;
2346 2347 2348
	struct completion	done;
	struct cpu_stop_work	stop_work;
	struct migration_arg	arg;
Peter Zijlstra's avatar
Peter Zijlstra committed
2349 2350 2351
};

/*
2352
 * Move (not current) task off this CPU, onto the destination CPU. We're doing
Peter Zijlstra's avatar
Peter Zijlstra committed
2353 2354 2355 2356 2357 2358 2359
 * 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.
 */
2360 2361
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
2362 2363
{
	/* Affinity changed (again). */
2364
	if (!is_cpu_allowed(p, dest_cpu))
2365
		return rq;
Peter Zijlstra's avatar
Peter Zijlstra committed
2366

2367
	update_rq_clock(rq);
2368
	rq = move_queued_task(rq, rf, p, dest_cpu);
2369 2370

	return rq;
Peter Zijlstra's avatar
Peter Zijlstra committed
2371 2372 2373 2374 2375 2376 2377 2378 2379 2380
}

/*
 * 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;
2381
	struct set_affinity_pending *pending = arg->pending;
2382 2383
	struct task_struct *p = arg->task;
	struct rq *rq = this_rq();
2384
	bool complete = false;
2385
	struct rq_flags rf;
Peter Zijlstra's avatar
Peter Zijlstra committed
2386 2387

	/*
2388 2389
	 * 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
2390
	 */
2391
	local_irq_save(rf.flags);
Peter Zijlstra's avatar
Peter Zijlstra committed
2392 2393
	/*
	 * We need to explicitly wake pending tasks before running
2394
	 * __migrate_task() such that we will not miss enforcing cpus_ptr
Peter Zijlstra's avatar
Peter Zijlstra committed
2395 2396
	 * during wakeups, see set_cpus_allowed_ptr()'s TASK_WAKING test.
	 */
2397
	flush_smp_call_function_queue();
2398 2399

	raw_spin_lock(&p->pi_lock);
2400
	rq_lock(rq, &rf);
2401

2402 2403 2404 2405 2406 2407
	/*
	 * 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);

2408 2409 2410 2411 2412
	/*
	 * 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.
	 */
2413
	if (task_rq(p) == rq) {
2414 2415 2416 2417
		if (is_migration_disabled(p))
			goto out;

		if (pending) {
2418
			p->migration_pending = NULL;
2419 2420
			complete = true;

2421 2422 2423
			if (cpumask_test_cpu(task_cpu(p), &p->cpus_mask))
				goto out;
		}
2424

2425
		if (task_on_rq_queued(p))
2426
			rq = __migrate_task(rq, &rf, p, arg->dest_cpu);
2427
		else
2428
			p->wake_cpu = arg->dest_cpu;
2429

2430 2431 2432 2433 2434 2435 2436
		/*
		 * 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.
		 */

2437
	} else if (pending) {
2438 2439 2440 2441 2442 2443 2444 2445 2446
		/*
		 * 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.
		 */

2447 2448 2449 2450 2451
		/*
		 * 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.
		 */
2452
		if (cpumask_test_cpu(task_cpu(p), p->cpus_ptr)) {
2453
			p->migration_pending = NULL;
2454 2455 2456 2457
			complete = true;
			goto out;
		}

2458 2459 2460 2461 2462
		/*
		 * When migrate_enable() hits a rq mis-match we can't reliably
		 * determine is_migration_disabled() and so have to chase after
		 * it.
		 */
2463
		WARN_ON_ONCE(!pending->stop_pending);
2464 2465 2466 2467
		task_rq_unlock(rq, p, &rf);
		stop_one_cpu_nowait(task_cpu(p), migration_cpu_stop,
				    &pending->arg, &pending->stop_work);
		return 0;
2468
	}
2469
out:
2470 2471
	if (pending)
		pending->stop_pending = false;
2472 2473 2474 2475 2476
	task_rq_unlock(rq, p, &rf);

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

Peter Zijlstra's avatar
Peter Zijlstra committed
2477 2478 2479
	return 0;
}

2480 2481 2482 2483 2484 2485
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);
2486
	raw_spin_rq_lock(rq);
2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499

	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);
2500

2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515
	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;
2516
	raw_spin_rq_unlock(rq);
2517 2518 2519
	raw_spin_unlock_irq(&p->pi_lock);

	put_task_struct(p);
Peter Zijlstra's avatar
Peter Zijlstra committed
2520 2521 2522
	return 0;
}

2523 2524 2525 2526
/*
 * sched_class::set_cpus_allowed must do the below, but is not required to
 * actually call this function.
 */
2527
void set_cpus_allowed_common(struct task_struct *p, const struct cpumask *new_mask, u32 flags)
Peter Zijlstra's avatar
Peter Zijlstra committed
2528
{
2529 2530 2531 2532 2533
	if (flags & (SCA_MIGRATE_ENABLE | SCA_MIGRATE_DISABLE)) {
		p->cpus_ptr = new_mask;
		return;
	}

2534
	cpumask_copy(&p->cpus_mask, new_mask);
Peter Zijlstra's avatar
Peter Zijlstra committed
2535 2536 2537
	p->nr_cpus_allowed = cpumask_weight(new_mask);
}

2538 2539
static void
__do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask, u32 flags)
2540
{
2541 2542 2543
	struct rq *rq = task_rq(p);
	bool queued, running;

2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559
	/*
	 * 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);
2560 2561 2562 2563 2564 2565 2566 2567 2568

	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.
		 */
2569
		lockdep_assert_rq_held(rq);
2570
		dequeue_task(rq, p, DEQUEUE_SAVE | DEQUEUE_NOCLOCK);
2571 2572 2573 2574
	}
	if (running)
		put_prev_task(rq, p);

2575
	p->sched_class->set_cpus_allowed(p, new_mask, flags);
2576 2577

	if (queued)
2578
		enqueue_task(rq, p, ENQUEUE_RESTORE | ENQUEUE_NOCLOCK);
2579
	if (running)
2580
		set_next_task(rq, p);
2581 2582
}

2583 2584 2585 2586 2587
void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
{
	__do_set_cpus_allowed(p, new_mask, 0);
}

2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601
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;
}

2602 2603 2604 2605 2606 2607 2608 2609 2610
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;
}

2611 2612
void release_user_cpus_ptr(struct task_struct *p)
{
2613
	kfree(clear_user_cpus_ptr(p));
2614 2615
}

2616
/*
2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668
 * 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
2669 2670
 * occurs after the stopper bailed out due to the targeted task still being
 * Migrate-Disable. Consider:
2671 2672 2673
 *
 *     Initial conditions: P0->cpus_mask = [0, 1]
 *
2674 2675 2676 2677
 *     CPU0		  P1				P2
 *     <P0>
 *       migrate_disable();
 *       <preempted>
2678 2679
 *                        set_cpus_allowed_ptr(P0, [1]);
 *                          <blocks>
2680 2681 2682 2683
 *     <migration/0>
 *       migration_cpu_stop()
 *         is_migration_disabled()
 *           <bails>
2684 2685 2686 2687 2688 2689 2690
 *                                                       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.
2691 2692 2693 2694 2695
 */
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;
2696
	bool stop_pending, complete = false;
2697 2698 2699

	/* 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)) {
2700 2701 2702 2703 2704 2705 2706 2707
		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);
		}

2708 2709 2710 2711
		/*
		 * If there are pending waiters, but no pending stop_work,
		 * then complete now.
		 */
2712
		pending = p->migration_pending;
2713
		if (pending && !pending->stop_pending) {
2714 2715 2716
			p->migration_pending = NULL;
			complete = true;
		}
2717

2718 2719
		task_rq_unlock(rq, p, rf);

2720 2721 2722 2723 2724
		if (push_task) {
			stop_one_cpu_nowait(rq->cpu, push_cpu_stop,
					    p, &rq->push_work);
		}

2725
		if (complete)
2726
			complete_all(&pending->done);
2727 2728 2729 2730 2731 2732 2733

		return 0;
	}

	if (!(flags & SCA_MIGRATE_ENABLE)) {
		/* serialized by p->pi_lock */
		if (!p->migration_pending) {
2734
			/* Install the request */
2735 2736
			refcount_set(&my_pending.refs, 1);
			init_completion(&my_pending.done);
2737 2738
			my_pending.arg = (struct migration_arg) {
				.task = p,
2739
				.dest_cpu = dest_cpu,
2740 2741 2742
				.pending = &my_pending,
			};

2743 2744 2745 2746
			p->migration_pending = &my_pending;
		} else {
			pending = p->migration_pending;
			refcount_inc(&pending->refs);
2747 2748 2749 2750 2751 2752 2753 2754 2755
			/*
			 * 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;
2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775
		}
	}
	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;
	}

2776
	if (task_on_cpu(rq, p) || READ_ONCE(p->__state) == TASK_WAKING) {
2777
		/*
2778 2779 2780
		 * 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.
2781
		 */
2782 2783 2784
		stop_pending = pending->stop_pending;
		if (!stop_pending)
			pending->stop_pending = true;
2785 2786 2787

		if (flags & SCA_MIGRATE_ENABLE)
			p->migration_flags &= ~MDF_PUSH;
2788

2789
		task_rq_unlock(rq, p, rf);
2790

2791 2792 2793 2794
		if (!stop_pending) {
			stop_one_cpu_nowait(cpu_of(rq), migration_cpu_stop,
					    &pending->arg, &pending->stop_work);
		}
2795

2796 2797
		if (flags & SCA_MIGRATE_ENABLE)
			return 0;
2798 2799 2800 2801 2802 2803
	} else {

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

2804 2805 2806 2807
			if (!pending->stop_pending) {
				p->migration_pending = NULL;
				complete = true;
			}
2808 2809 2810 2811 2812 2813 2814 2815 2816 2817
		}
		task_rq_unlock(rq, p, rf);

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

	wait_for_completion(&pending->done);

	if (refcount_dec_and_test(&pending->refs))
2818
		wake_up_var(&pending->refs); /* No UaF, just an address */
2819

2820 2821 2822 2823
	/*
	 * Block the original owner of &pending until all subsequent callers
	 * have seen the completion and decremented the refcount
	 */
2824 2825
	wait_var_event(&my_pending.refs, !refcount_read(&my_pending.refs));

2826 2827 2828
	/* ARGH */
	WARN_ON_ONCE(my_pending.stop_pending);

2829 2830 2831
	return 0;
}

Peter Zijlstra's avatar
Peter Zijlstra committed
2832
/*
2833
 * Called with both p->pi_lock and rq->lock held; drops both before returning.
Peter Zijlstra's avatar
Peter Zijlstra committed
2834
 */
2835 2836 2837 2838 2839 2840 2841
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
2842
{
2843
	const struct cpumask *cpu_allowed_mask = task_cpu_possible_mask(p);
2844
	const struct cpumask *cpu_valid_mask = cpu_active_mask;
2845
	bool kthread = p->flags & PF_KTHREAD;
2846
	struct cpumask *user_mask = NULL;
Peter Zijlstra's avatar
Peter Zijlstra committed
2847 2848 2849
	unsigned int dest_cpu;
	int ret = 0;

2850
	update_rq_clock(rq);
Peter Zijlstra's avatar
Peter Zijlstra committed
2851

2852
	if (kthread || is_migration_disabled(p)) {
2853
		/*
2854 2855 2856
		 * 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.
2857 2858 2859 2860 2861
		 *
		 * 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.
2862 2863 2864 2865
		 */
		cpu_valid_mask = cpu_online_mask;
	}

2866 2867 2868 2869 2870
	if (!kthread && !cpumask_subset(new_mask, cpu_allowed_mask)) {
		ret = -EINVAL;
		goto out;
	}

2871 2872 2873 2874
	/*
	 * Must re-check here, to close a race against __kthread_bind(),
	 * sched_setaffinity() is not guaranteed to observe the flag.
	 */
2875
	if ((flags & SCA_CHECK) && (p->flags & PF_NO_SETAFFINITY)) {
2876 2877 2878 2879
		ret = -EINVAL;
		goto out;
	}

2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890
	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
2891

2892 2893 2894 2895 2896 2897
	/*
	 * 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);
2898
	if (dest_cpu >= nr_cpu_ids) {
Peter Zijlstra's avatar
Peter Zijlstra committed
2899 2900 2901 2902
		ret = -EINVAL;
		goto out;
	}

2903
	__do_set_cpus_allowed(p, new_mask, flags);
Peter Zijlstra's avatar
Peter Zijlstra committed
2904

2905 2906 2907 2908 2909 2910 2911 2912
	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
2913 2914

out:
2915
	task_rq_unlock(rq, p, rf);
Peter Zijlstra's avatar
Peter Zijlstra committed
2916 2917 2918

	return ret;
}
2919

2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938
/*
 * 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);
}

2939 2940
int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
{
2941
	return __set_cpus_allowed_ptr(p, new_mask, 0);
2942
}
Peter Zijlstra's avatar
Peter Zijlstra committed
2943 2944
EXPORT_SYMBOL_GPL(set_cpus_allowed_ptr);

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 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076
/*
 * 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
3077
void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
Ingo Molnar's avatar
Ingo Molnar committed
3078
{
3079
#ifdef CONFIG_SCHED_DEBUG
3080 3081
	unsigned int state = READ_ONCE(p->__state);

3082 3083 3084 3085
	/*
	 * We should never call set_task_cpu() on a blocked task,
	 * ttwu() will sort out the placement.
	 */
3086
	WARN_ON_ONCE(state != TASK_RUNNING && state != TASK_WAKING && !p->on_rq);
3087

3088 3089 3090 3091 3092
	/*
	 * 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.
	 */
3093
	WARN_ON_ONCE(state == TASK_RUNNING &&
3094 3095 3096
		     p->sched_class == &fair_sched_class &&
		     (p->on_rq && !task_on_rq_migrating(p)));

3097
#ifdef CONFIG_LOCKDEP
3098 3099 3100 3101 3102
	/*
	 * 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,
3103
	 * see task_group().
3104 3105 3106 3107
	 *
	 * Furthermore, all task_rq users should acquire both locks, see
	 * task_rq_lock().
	 */
3108
	WARN_ON_ONCE(debug_locks && !(lockdep_is_held(&p->pi_lock) ||
3109
				      lockdep_is_held(__rq_lockp(task_rq(p)))));
3110
#endif
3111 3112 3113 3114
	/*
	 * Clearly, migrating tasks to offline CPUs is a fairly daft thing.
	 */
	WARN_ON_ONCE(!cpu_online(new_cpu));
3115 3116

	WARN_ON_ONCE(is_migration_disabled(p));
3117 3118
#endif

3119
	trace_sched_migrate_task(p, new_cpu);
3120

3121
	if (task_cpu(p) != new_cpu) {
3122
		if (p->sched_class->migrate_task_rq)
3123
			p->sched_class->migrate_task_rq(p, new_cpu);
3124
		p->se.nr_migrations++;
3125
		rseq_migrate(p);
3126
		perf_event_task_migrate(p);
3127
	}
Ingo Molnar's avatar
Ingo Molnar committed
3128 3129

	__set_task_cpu(p, new_cpu);
Ingo Molnar's avatar
Ingo Molnar committed
3130 3131
}

3132
#ifdef CONFIG_NUMA_BALANCING
3133 3134
static void __migrate_swap_task(struct task_struct *p, int cpu)
{
3135
	if (task_on_rq_queued(p)) {
3136
		struct rq *src_rq, *dst_rq;
3137
		struct rq_flags srf, drf;
3138 3139 3140 3141

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

3142 3143 3144
		rq_pin_lock(src_rq, &srf);
		rq_pin_lock(dst_rq, &drf);

3145 3146 3147 3148
		deactivate_task(src_rq, p, 0);
		set_task_cpu(p, cpu);
		activate_task(dst_rq, p, 0);
		check_preempt_curr(dst_rq, p, 0);
3149 3150 3151 3152

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

3153 3154 3155 3156
	} 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
3157
		 * previous CPU our target instead of where it really is.
3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173
		 */
		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;

3174 3175 3176
	if (!cpu_active(arg->src_cpu) || !cpu_active(arg->dst_cpu))
		return -EAGAIN;

3177 3178 3179
	src_rq = cpu_rq(arg->src_cpu);
	dst_rq = cpu_rq(arg->dst_cpu);

3180 3181
	double_raw_lock(&arg->src_task->pi_lock,
			&arg->dst_task->pi_lock);
3182
	double_rq_lock(src_rq, dst_rq);
3183

3184 3185 3186 3187 3188 3189
	if (task_cpu(arg->dst_task) != arg->dst_cpu)
		goto unlock;

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

3190
	if (!cpumask_test_cpu(arg->dst_cpu, arg->src_task->cpus_ptr))
3191 3192
		goto unlock;

3193
	if (!cpumask_test_cpu(arg->src_cpu, arg->dst_task->cpus_ptr))
3194 3195 3196 3197 3198 3199 3200 3201 3202
		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);
3203 3204
	raw_spin_unlock(&arg->dst_task->pi_lock);
	raw_spin_unlock(&arg->src_task->pi_lock);
3205 3206 3207 3208 3209 3210 3211

	return ret;
}

/*
 * Cross migrate two tasks
 */
3212 3213
int migrate_swap(struct task_struct *cur, struct task_struct *p,
		int target_cpu, int curr_cpu)
3214 3215 3216 3217 3218 3219
{
	struct migration_swap_arg arg;
	int ret = -EINVAL;

	arg = (struct migration_swap_arg){
		.src_task = cur,
3220
		.src_cpu = curr_cpu,
3221
		.dst_task = p,
3222
		.dst_cpu = target_cpu,
3223 3224 3225 3226 3227
	};

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

3228 3229 3230 3231
	/*
	 * 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.
	 */
3232 3233 3234
	if (!cpu_active(arg.src_cpu) || !cpu_active(arg.dst_cpu))
		goto out;

3235
	if (!cpumask_test_cpu(arg.dst_cpu, arg.src_task->cpus_ptr))
3236 3237
		goto out;

3238
	if (!cpumask_test_cpu(arg.src_cpu, arg.dst_task->cpus_ptr))
3239 3240
		goto out;

3241
	trace_sched_swap_numa(cur, arg.src_cpu, p, arg.dst_cpu);
3242 3243 3244 3245 3246
	ret = stop_two_cpus(arg.dst_cpu, arg.src_cpu, migrate_swap_stop, &arg);

out:
	return ret;
}
3247
#endif /* CONFIG_NUMA_BALANCING */
3248

Linus Torvalds's avatar
Linus Torvalds committed
3249 3250 3251
/*
 * wait_task_inactive - wait for a thread to unschedule.
 *
3252 3253 3254 3255 3256 3257
 * Wait for the thread to block in any of the states set in @match_state.
 * 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.
3258
 *
Linus Torvalds's avatar
Linus Torvalds committed
3259 3260 3261 3262 3263 3264
 * 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.
 */
3265
unsigned long wait_task_inactive(struct task_struct *p, unsigned int match_state)
Linus Torvalds's avatar
Linus Torvalds committed
3266
{
3267
	int running, queued;
3268
	struct rq_flags rf;
3269
	unsigned long ncsw;
3270
	struct rq *rq;
Linus Torvalds's avatar
Linus Torvalds committed
3271

3272 3273 3274 3275 3276 3277 3278 3279
	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);
3280

3281 3282 3283 3284 3285 3286 3287
		/*
		 * 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!
3288
		 * But we don't care, since "task_on_cpu()" will
3289 3290 3291
		 * return false if the runqueue has changed and p
		 * is actually now running somewhere else!
		 */
3292
		while (task_on_cpu(rq, p)) {
3293
			if (!(READ_ONCE(p->__state) & match_state))
3294
				return 0;
3295
			cpu_relax();
3296
		}
3297

3298 3299 3300 3301 3302
		/*
		 * 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.
		 */
3303
		rq = task_rq_lock(p, &rf);
3304
		trace_sched_wait_task(p);
3305
		running = task_on_cpu(rq, p);
3306
		queued = task_on_rq_queued(p);
3307
		ncsw = 0;
3308
		if (READ_ONCE(p->__state) & match_state)
3309
			ncsw = p->nvcsw | LONG_MIN; /* sets MSB */
3310
		task_rq_unlock(rq, p, &rf);
3311

3312 3313 3314 3315 3316 3317
		/*
		 * If it changed from the expected state, bail out now.
		 */
		if (unlikely(!ncsw))
			break;

3318 3319 3320 3321 3322 3323 3324 3325 3326 3327
		/*
		 * 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;
		}
3328

3329 3330 3331 3332 3333
		/*
		 * It's not enough that it's not actively running,
		 * it must be off the runqueue _entirely_, and not
		 * preempted!
		 *
3334
		 * So if it was still runnable (but just not actively
3335 3336 3337
		 * running right now), it's preempted, and we should
		 * yield - it could be a while.
		 */
3338
		if (unlikely(queued)) {
3339
			ktime_t to = NSEC_PER_SEC / HZ;
3340 3341

			set_current_state(TASK_UNINTERRUPTIBLE);
3342
			schedule_hrtimeout(&to, HRTIMER_MODE_REL_HARD);
3343 3344
			continue;
		}
3345

3346 3347 3348 3349 3350 3351 3352
		/*
		 * 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;
	}
3353 3354

	return ncsw;
Linus Torvalds's avatar
Linus Torvalds committed
3355 3356 3357 3358 3359 3360 3361 3362 3363
}

/***
 * 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
3364
 * NOTE: this function doesn't have to take the runqueue lock,
Linus Torvalds's avatar
Linus Torvalds committed
3365 3366 3367 3368 3369
 * 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.
 */
3370
void kick_process(struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
3371 3372 3373 3374 3375 3376 3377 3378 3379
{
	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
3380
EXPORT_SYMBOL_GPL(kick_process);
Linus Torvalds's avatar
Linus Torvalds committed
3381

3382
/*
3383
 * ->cpus_ptr is protected by both rq->lock and p->pi_lock
3384 3385 3386 3387 3388
 *
 * A few notes on cpu_active vs cpu_online:
 *
 *  - cpu_active must be a subset of cpu_online
 *
3389
 *  - on CPU-up we allow per-CPU kthreads on the online && !active CPU,
3390
 *    see __set_cpus_allowed_ptr(). At this point the newly online
3391
 *    CPU isn't yet part of the sched domains, and balancing will not
3392 3393
 *    see it.
 *
3394
 *  - on CPU-down we clear cpu_active() to mask the sched domains and
3395
 *    avoid the load balancer to place new tasks on the to be removed
3396
 *    CPU. Existing tasks will remain running there and will be taken
3397 3398 3399 3400 3401 3402
 *    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.
3403
 */
3404 3405
static int select_fallback_rq(int cpu, struct task_struct *p)
{
3406 3407
	int nid = cpu_to_node(cpu);
	const struct cpumask *nodemask = NULL;
3408 3409
	enum { cpuset, possible, fail } state = cpuset;
	int dest_cpu;
3410

3411
	/*
3412 3413 3414
	 * 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.
3415 3416 3417 3418 3419 3420
	 */
	if (nid != -1) {
		nodemask = cpumask_of_node(nid);

		/* Look for allowed, online CPU in same node. */
		for_each_cpu(dest_cpu, nodemask) {
3421
			if (is_cpu_allowed(p, dest_cpu))
3422 3423
				return dest_cpu;
		}
3424
	}
3425

3426 3427
	for (;;) {
		/* Any allowed, online CPU? */
3428
		for_each_cpu(dest_cpu, p->cpus_ptr) {
3429
			if (!is_cpu_allowed(p, dest_cpu))
3430
				continue;
3431

3432 3433
			goto out;
		}
3434

3435
		/* No more Mr. Nice Guy. */
3436 3437
		switch (state) {
		case cpuset:
3438
			if (cpuset_cpus_allowed_fallback(p)) {
3439 3440 3441
				state = possible;
				break;
			}
3442
			fallthrough;
3443
		case possible:
3444 3445 3446 3447 3448 3449
			/*
			 * XXX When called from select_task_rq() we only
			 * hold p->pi_lock and again violate locking order.
			 *
			 * More yuck to audit.
			 */
3450
			do_set_cpus_allowed(p, task_cpu_possible_mask(p));
3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466
			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()) {
3467
			printk_deferred("process %d (%s) no longer affine to cpu%d\n",
3468 3469
					task_pid_nr(p), p->comm, cpu);
		}
3470 3471 3472 3473 3474
	}

	return dest_cpu;
}

3475
/*
3476
 * The caller (fork, wakeup) owns p->pi_lock, ->cpus_ptr is stable.
3477
 */
3478
static inline
3479
int select_task_rq(struct task_struct *p, int cpu, int wake_flags)
3480
{
3481 3482
	lockdep_assert_held(&p->pi_lock);

3483
	if (p->nr_cpus_allowed > 1 && !is_migration_disabled(p))
3484
		cpu = p->sched_class->select_task_rq(p, cpu, wake_flags);
3485
	else
3486
		cpu = cpumask_any(p->cpus_ptr);
3487 3488 3489

	/*
	 * In order not to call set_task_cpu() on a blocking task we need
3490
	 * to rely on ttwu() to place the task on a valid ->cpus_ptr
3491
	 * CPU.
3492 3493 3494 3495 3496 3497
	 *
	 * 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 ]
	 */
3498
	if (unlikely(!is_cpu_allowed(p, cpu)))
3499
		cpu = select_fallback_rq(task_cpu(p), p);
3500 3501

	return cpu;
3502
}
3503

3504 3505
void sched_set_stop_task(int cpu, struct task_struct *stop)
{
3506
	static struct lock_class_key stop_pi_lock;
3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521
	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;
3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535

		/*
		 * 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);
3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548
	}

	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;
	}
}

3549
#else /* CONFIG_SMP */
3550 3551

static inline int __set_cpus_allowed_ptr(struct task_struct *p,
3552 3553
					 const struct cpumask *new_mask,
					 u32 flags)
3554 3555 3556 3557
{
	return set_cpus_allowed_ptr(p, new_mask);
}

3558 3559
static inline void migrate_disable_switch(struct rq *rq, struct task_struct *p) { }

3560 3561 3562 3563 3564
static inline bool rq_has_pinned_tasks(struct rq *rq)
{
	return false;
}

3565
#endif /* !CONFIG_SMP */
3566

3567
static void
3568
ttwu_stat(struct task_struct *p, int cpu, int wake_flags)
Tejun Heo's avatar
Tejun Heo committed
3569
{
3570
	struct rq *rq;
3571

3572 3573 3574 3575
	if (!schedstat_enabled())
		return;

	rq = this_rq();
3576

3577 3578
#ifdef CONFIG_SMP
	if (cpu == rq->cpu) {
3579
		__schedstat_inc(rq->ttwu_local);
3580
		__schedstat_inc(p->stats.nr_wakeups_local);
3581 3582 3583
	} else {
		struct sched_domain *sd;

3584
		__schedstat_inc(p->stats.nr_wakeups_remote);
3585
		rcu_read_lock();
3586
		for_each_domain(rq->cpu, sd) {
3587
			if (cpumask_test_cpu(cpu, sched_domain_span(sd))) {
3588
				__schedstat_inc(sd->ttwu_wake_remote);
3589 3590 3591
				break;
			}
		}
3592
		rcu_read_unlock();
3593
	}
3594 3595

	if (wake_flags & WF_MIGRATED)
3596
		__schedstat_inc(p->stats.nr_wakeups_migrate);
3597 3598
#endif /* CONFIG_SMP */

3599
	__schedstat_inc(rq->ttwu_count);
3600
	__schedstat_inc(p->stats.nr_wakeups);
3601 3602

	if (wake_flags & WF_SYNC)
3603
		__schedstat_inc(p->stats.nr_wakeups_sync);
3604 3605
}

3606 3607 3608
/*
 * Mark the task runnable and perform wakeup-preemption.
 */
3609
static void ttwu_do_wakeup(struct rq *rq, struct task_struct *p, int wake_flags,
3610
			   struct rq_flags *rf)
Tejun Heo's avatar
Tejun Heo committed
3611 3612
{
	check_preempt_curr(rq, p, wake_flags);
3613
	WRITE_ONCE(p->__state, TASK_RUNNING);
3614 3615
	trace_sched_wakeup(p);

Tejun Heo's avatar
Tejun Heo committed
3616
#ifdef CONFIG_SMP
3617 3618
	if (p->sched_class->task_woken) {
		/*
3619
		 * Our task @p is fully woken up and running; so it's safe to
3620
		 * drop the rq->lock, hereafter rq is only used for statistics.
3621
		 */
3622
		rq_unpin_lock(rq, rf);
Tejun Heo's avatar
Tejun Heo committed
3623
		p->sched_class->task_woken(rq, p);
3624
		rq_repin_lock(rq, rf);
3625
	}
Tejun Heo's avatar
Tejun Heo committed
3626

3627
	if (rq->idle_stamp) {
3628
		u64 delta = rq_clock(rq) - rq->idle_stamp;
3629
		u64 max = 2*rq->max_idle_balance_cost;
Tejun Heo's avatar
Tejun Heo committed
3630

3631 3632 3633
		update_avg(&rq->avg_idle, delta);

		if (rq->avg_idle > max)
Tejun Heo's avatar
Tejun Heo committed
3634
			rq->avg_idle = max;
3635

3636 3637 3638
		rq->wake_stamp = jiffies;
		rq->wake_avg_idle = rq->avg_idle / 2;

Tejun Heo's avatar
Tejun Heo committed
3639 3640 3641 3642 3643
		rq->idle_stamp = 0;
	}
#endif
}

3644
static void
3645
ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags,
3646
		 struct rq_flags *rf)
3647
{
3648
	int en_flags = ENQUEUE_WAKEUP | ENQUEUE_NOCLOCK;
3649

3650
	lockdep_assert_rq_held(rq);
3651

3652 3653
	if (p->sched_contributes_to_load)
		rq->nr_uninterruptible--;
3654

3655
#ifdef CONFIG_SMP
3656
	if (wake_flags & WF_MIGRATED)
3657
		en_flags |= ENQUEUE_MIGRATED;
3658
	else
3659
#endif
3660 3661 3662 3663
	if (p->in_iowait) {
		delayacct_blkio_end(p);
		atomic_dec(&task_rq(p)->nr_iowait);
	}
3664

3665
	activate_task(rq, p, en_flags);
3666
	ttwu_do_wakeup(rq, p, wake_flags, rf);
3667 3668 3669
}

/*
3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692
 * 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.
3693
 */
3694
static int ttwu_runnable(struct task_struct *p, int wake_flags)
3695
{
3696
	struct rq_flags rf;
3697 3698 3699
	struct rq *rq;
	int ret = 0;

3700
	rq = __task_rq_lock(p, &rf);
3701
	if (task_on_rq_queued(p)) {
3702 3703
		/* check_preempt_curr() may use rq clock */
		update_rq_clock(rq);
3704
		ttwu_do_wakeup(rq, p, wake_flags, &rf);
3705 3706
		ret = 1;
	}
3707
	__task_rq_unlock(rq, &rf);
3708 3709 3710 3711

	return ret;
}

3712
#ifdef CONFIG_SMP
3713
void sched_ttwu_pending(void *arg)
3714
{
3715
	struct llist_node *llist = arg;
3716
	struct rq *rq = this_rq();
3717
	struct task_struct *p, *t;
3718
	struct rq_flags rf;
3719

3720 3721 3722
	if (!llist)
		return;

3723 3724 3725 3726 3727 3728 3729
	/*
	 * 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);

3730
	rq_lock_irqsave(rq, &rf);
3731
	update_rq_clock(rq);
3732

3733
	llist_for_each_entry_safe(p, t, llist, wake_entry.llist) {
3734 3735 3736 3737 3738 3739
		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));

3740
		ttwu_do_activate(rq, p, p->sched_remote_wakeup ? WF_MIGRATED : 0, &rf);
3741
	}
3742

3743
	rq_unlock_irqrestore(rq, &rf);
3744 3745
}

3746
void send_call_function_single_ipi(int cpu)
3747
{
3748
	struct rq *rq = cpu_rq(cpu);
3749

3750 3751 3752 3753
	if (!set_nr_if_polling(rq->idle))
		arch_send_call_function_single_ipi(cpu);
	else
		trace_sched_wake_idle_without_ipi(cpu);
3754 3755
}

3756 3757 3758 3759 3760 3761 3762
/*
 * 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)
3763
{
3764 3765
	struct rq *rq = cpu_rq(cpu);

3766 3767
	p->sched_remote_wakeup = !!(wake_flags & WF_MIGRATED);

3768
	WRITE_ONCE(rq->ttwu_pending, 1);
3769
	__smp_call_single_queue(cpu, &p->wake_entry.llist);
3770
}
3771

3772 3773 3774
void wake_up_if_idle(int cpu)
{
	struct rq *rq = cpu_rq(cpu);
3775
	struct rq_flags rf;
3776

3777 3778 3779 3780
	rcu_read_lock();

	if (!is_idle_task(rcu_dereference(rq->curr)))
		goto out;
3781

3782 3783 3784 3785 3786
	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);
3787 3788 3789

out:
	rcu_read_unlock();
3790 3791
}

3792
bool cpus_share_cache(int this_cpu, int that_cpu)
3793
{
3794 3795 3796
	if (this_cpu == that_cpu)
		return true;

3797 3798
	return per_cpu(sd_llc_id, this_cpu) == per_cpu(sd_llc_id, that_cpu);
}
3799

3800
static inline bool ttwu_queue_cond(struct task_struct *p, int cpu)
3801
{
3802 3803 3804 3805 3806 3807 3808
	/*
	 * Do not complicate things with the async wake_list while the CPU is
	 * in hotplug state.
	 */
	if (!cpu_active(cpu))
		return false;

3809 3810 3811 3812
	/* Ensure the task will still be allowed to run on the CPU. */
	if (!cpumask_test_cpu(cpu, p->cpus_ptr))
		return false;

3813 3814 3815 3816 3817 3818 3819
	/*
	 * 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;

3820 3821 3822
	if (cpu == smp_processor_id())
		return false;

3823
	/*
3824 3825 3826 3827 3828
	 * If the wakee cpu is idle, or the task is descheduling and the
	 * only running task on the CPU, then use the wakelist to offload
	 * the task activation to the idle (or soon-to-be-idle) CPU as
	 * the current CPU is likely busy. nr_running is checked to
	 * avoid unnecessary task stacking.
3829 3830 3831 3832
	 *
	 * Note that we can only get here with (wakee) p->on_rq=0,
	 * p->on_cpu can be whatever, we've done the dequeue, so
	 * the wakee has been accounted out of ->nr_running.
3833
	 */
3834
	if (!cpu_rq(cpu)->nr_running)
3835 3836 3837 3838 3839 3840
		return true;

	return false;
}

static bool ttwu_queue_wakelist(struct task_struct *p, int cpu, int wake_flags)
3841
{
3842
	if (sched_feat(TTWU_QUEUE) && ttwu_queue_cond(p, cpu)) {
3843
		sched_clock_cpu(cpu); /* Sync clocks across CPUs */
3844
		__ttwu_queue_wakelist(p, cpu, wake_flags);
3845 3846 3847 3848 3849
		return true;
	}

	return false;
}
3850 3851 3852 3853 3854 3855 3856 3857

#else /* !CONFIG_SMP */

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

3858
#endif /* CONFIG_SMP */
3859

3860
static void ttwu_queue(struct task_struct *p, int cpu, int wake_flags)
3861 3862
{
	struct rq *rq = cpu_rq(cpu);
3863
	struct rq_flags rf;
3864

3865
	if (ttwu_queue_wakelist(p, cpu, wake_flags))
3866 3867
		return;

3868
	rq_lock(rq, &rf);
3869
	update_rq_clock(rq);
3870
	ttwu_do_activate(rq, p, wake_flags, &rf);
3871
	rq_unlock(rq, &rf);
Tejun Heo's avatar
Tejun Heo committed
3872 3873
}

3874 3875 3876 3877 3878
/*
 * 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.
3879 3880 3881 3882 3883 3884 3885 3886
 *
 * 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.
3887 3888 3889 3890
 */
static __always_inline
bool ttwu_state_match(struct task_struct *p, unsigned int state, int *success)
{
3891 3892 3893 3894 3895
	if (IS_ENABLED(CONFIG_DEBUG_PREEMPT)) {
		WARN_ON_ONCE((state & TASK_RTLOCK_WAIT) &&
			     state != TASK_RTLOCK_WAIT);
	}

3896 3897 3898 3899
	if (READ_ONCE(p->__state) & state) {
		*success = 1;
		return true;
	}
3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918 3919

#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
3920 3921 3922
	return false;
}

3923 3924 3925 3926 3927 3928
/*
 * Notes on Program-Order guarantees on SMP systems.
 *
 *  MIGRATION
 *
 * The basic program-order guarantee on SMP systems is that when a task [t]
3929 3930
 * migrates, all its activity on its old CPU [c0] happens-before any subsequent
 * execution on its new CPU [c1].
3931 3932 3933 3934 3935 3936 3937 3938
 *
 * 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
 *
3939
 * Release/acquire chaining guarantees that B happens after A and C after B.
3940
 * Note: the CPU doing B need not be c0 or c1
3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970
 *
 * 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:
 *
3971 3972
 *   1) smp_store_release(X->on_cpu, 0)   -- finish_task()
 *   2) smp_cond_load_acquire(!X->on_cpu) -- try_to_wake_up()
3973 3974 3975 3976 3977 3978 3979 3980 3981 3982
 *
 * 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);
 *
3983
 *                    smp_cond_load_acquire(&X->on_cpu, !VAL);
3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000
 *                    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
 *
 *
4001 4002 4003
 * 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().
4004 4005
 */

Tejun Heo's avatar
Tejun Heo committed
4006
/**
Linus Torvalds's avatar
Linus Torvalds committed
4007
 * try_to_wake_up - wake up a thread
Tejun Heo's avatar
Tejun Heo committed
4008
 * @p: the thread to be awakened
Linus Torvalds's avatar
Linus Torvalds committed
4009
 * @state: the mask of task states that can be woken
Tejun Heo's avatar
Tejun Heo committed
4010
 * @wake_flags: wake modifier flags (WF_*)
Linus Torvalds's avatar
Linus Torvalds committed
4011
 *
4012 4013 4014
 * Conceptually does:
 *
 *   If (@state & @p->state) @p->state = TASK_RUNNING.
Linus Torvalds's avatar
Linus Torvalds committed
4015
 *
4016 4017
 * If the task was not queued/runnable, also place it back on a runqueue.
 *
4018 4019 4020 4021
 * 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().
4022
 *
4023
 * Uses p->pi_lock to serialize against concurrent wake-ups.
4024
 *
4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038
 * 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.
4039
 *
4040 4041
 * Return: %true if @p->state changes (an actual wakeup was done),
 *	   %false otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
4042
 */
4043 4044
static int
try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
Linus Torvalds's avatar
Linus Torvalds committed
4045 4046
{
	unsigned long flags;
4047
	int cpu, success = 0;
4048

4049
	preempt_disable();
4050 4051 4052 4053
	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
4054
		 * case the whole 'p->on_rq && ttwu_runnable()' case below
4055 4056 4057 4058 4059 4060 4061
		 * 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).
		 */
4062
		if (!ttwu_state_match(p, state, &success))
4063
			goto out;
4064 4065

		trace_sched_waking(p);
4066
		WRITE_ONCE(p->__state, TASK_RUNNING);
4067 4068 4069 4070
		trace_sched_wakeup(p);
		goto out;
	}

4071 4072 4073
	/*
	 * 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
4074 4075
	 * reordered with p->state check below. This pairs with smp_store_mb()
	 * in set_current_state() that the waiting thread does.
4076
	 */
4077
	raw_spin_lock_irqsave(&p->pi_lock, flags);
4078
	smp_mb__after_spinlock();
4079
	if (!ttwu_state_match(p, state, &success))
4080
		goto unlock;
Linus Torvalds's avatar
Linus Torvalds committed
4081

4082 4083
	trace_sched_waking(p);

4084 4085 4086 4087 4088
	/*
	 * 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.
	 *
4089 4090 4091 4092 4093 4094 4095 4096
	 * 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
4097 4098
	 *
	 * [task p]
4099
	 *   STORE p->state = UNINTERRUPTIBLE	  LOAD p->on_rq
4100
	 *
4101 4102
	 * Pairs with the LOCK+smp_mb__after_spinlock() on rq->lock in
	 * __schedule().  See the comment for smp_mb__after_spinlock().
4103 4104
	 *
	 * A similar smb_rmb() lives in try_invoke_on_locked_down_task().
4105 4106
	 */
	smp_rmb();
4107
	if (READ_ONCE(p->on_rq) && ttwu_runnable(p, wake_flags))
4108
		goto unlock;
Linus Torvalds's avatar
Linus Torvalds committed
4109 4110

#ifdef CONFIG_SMP
4111 4112 4113 4114 4115 4116 4117
	/*
	 * 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.
	 *
4118 4119 4120 4121 4122 4123 4124 4125
	 * __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
4126
	 *
4127 4128
	 * Pairs with the LOCK+smp_mb__after_spinlock() on rq->lock in
	 * __schedule().  See the comment for smp_mb__after_spinlock().
4129 4130 4131 4132
	 *
	 * 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().
4133
	 */
4134 4135 4136 4137 4138 4139 4140 4141
	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().
	 */
4142
	WRITE_ONCE(p->__state, TASK_WAKING);
4143

4144 4145 4146 4147 4148 4149
	/*
	 * 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.
4150 4151 4152 4153 4154 4155 4156 4157 4158 4159 4160 4161
	 *
	 * 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.
4162
	 */
4163
	if (smp_load_acquire(&p->on_cpu) &&
4164
	    ttwu_queue_wakelist(p, task_cpu(p), wake_flags))
4165 4166
		goto unlock;

Peter Zijlstra's avatar
Peter Zijlstra committed
4167
	/*
4168
	 * If the owning (remote) CPU is still in the middle of schedule() with
4169
	 * this task as prev, wait until it's done referencing the task.
4170
	 *
4171
	 * Pairs with the smp_store_release() in finish_task().
4172 4173 4174
	 *
	 * This ensures that tasks getting woken will be fully ordered against
	 * their previous state and preserve Program Order.
4175
	 */
4176
	smp_cond_load_acquire(&p->on_cpu, !VAL);
Linus Torvalds's avatar
Linus Torvalds committed
4177

4178
	cpu = select_task_rq(p, p->wake_cpu, wake_flags | WF_TTWU);
4179
	if (task_cpu(p) != cpu) {
4180 4181 4182 4183 4184
		if (p->in_iowait) {
			delayacct_blkio_end(p);
			atomic_dec(&task_rq(p)->nr_iowait);
		}

4185
		wake_flags |= WF_MIGRATED;
4186
		psi_ttwu_dequeue(p);
4187
		set_task_cpu(p, cpu);
4188
	}
4189 4190
#else
	cpu = task_cpu(p);
Linus Torvalds's avatar
Linus Torvalds committed
4191 4192
#endif /* CONFIG_SMP */

4193
	ttwu_queue(p, cpu, wake_flags);
4194
unlock:
4195
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
4196 4197
out:
	if (success)
4198
		ttwu_stat(p, task_cpu(p), wake_flags);
4199
	preempt_enable();
Linus Torvalds's avatar
Linus Torvalds committed
4200 4201 4202 4203

	return success;
}

4204
/**
4205
 * task_call_func - Invoke a function on task in fixed state
4206
 * @p: Process for which the function is to be invoked, can be @current.
4207 4208 4209
 * @func: Function to invoke.
 * @arg: Argument to function.
 *
4210 4211 4212 4213
 * 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.
4214 4215
 *
 * Returns:
4216
 *   Whatever @func returns
4217
 */
4218
int task_call_func(struct task_struct *p, task_call_f func, void *arg)
4219
{
4220 4221
	struct rq *rq = NULL;
	unsigned int state;
4222
	struct rq_flags rf;
4223
	int ret;
4224

4225
	raw_spin_lock_irqsave(&p->pi_lock, rf.flags);
4226 4227 4228 4229 4230 4231 4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242

	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)
4243
		rq = __task_rq_lock(p, &rf);
4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257

	/*
	 * 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)
4258
		rq_unlock(rq, &rf);
4259

4260
	raw_spin_unlock_irqrestore(&p->pi_lock, rf.flags);
4261 4262 4263
	return ret;
}

4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295
/**
 * cpu_curr_snapshot - Return a snapshot of the currently running task
 * @cpu: The CPU on which to snapshot the task.
 *
 * Returns the task_struct pointer of the task "currently" running on
 * the specified CPU.  If the same task is running on that CPU throughout,
 * the return value will be a pointer to that task's task_struct structure.
 * If the CPU did any context switches even vaguely concurrently with the
 * execution of this function, the return value will be a pointer to the
 * task_struct structure of a randomly chosen task that was running on
 * that CPU somewhere around the time that this function was executing.
 *
 * If the specified CPU was offline, the return value is whatever it
 * is, perhaps a pointer to the task_struct structure of that CPU's idle
 * task, but there is no guarantee.  Callers wishing a useful return
 * value must take some action to ensure that the specified CPU remains
 * online throughout.
 *
 * This function executes full memory barriers before and after fetching
 * the pointer, which permits the caller to confine this function's fetch
 * with respect to the caller's accesses to other shared variables.
 */
struct task_struct *cpu_curr_snapshot(int cpu)
{
	struct task_struct *t;

	smp_mb(); /* Pairing determined by caller's synchronization design. */
	t = rcu_dereference(cpu_curr(cpu));
	smp_mb(); /* Pairing determined by caller's synchronization design. */
	return t;
}

4296 4297 4298 4299 4300
/**
 * 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
4301 4302 4303
 * processes.
 *
 * Return: 1 if the process was woken up, 0 if it was already running.
4304
 *
4305
 * This function executes a full memory barrier before accessing the task state.
4306
 */
4307
int wake_up_process(struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
4308
{
4309
	return try_to_wake_up(p, TASK_NORMAL, 0);
Linus Torvalds's avatar
Linus Torvalds committed
4310 4311 4312
}
EXPORT_SYMBOL(wake_up_process);

4313
int wake_up_state(struct task_struct *p, unsigned int state)
Linus Torvalds's avatar
Linus Torvalds committed
4314 4315 4316 4317 4318 4319 4320
{
	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
4321 4322 4323
 *
 * __sched_fork() is basic setup used by init_idle() too:
 */
4324
static void __sched_fork(unsigned long clone_flags, struct task_struct *p)
Ingo Molnar's avatar
Ingo Molnar committed
4325
{
Peter Zijlstra's avatar
Peter Zijlstra committed
4326 4327 4328
	p->on_rq			= 0;

	p->se.on_rq			= 0;
Ingo Molnar's avatar
Ingo Molnar committed
4329 4330
	p->se.exec_start		= 0;
	p->se.sum_exec_runtime		= 0;
4331
	p->se.prev_sum_exec_runtime	= 0;
4332
	p->se.nr_migrations		= 0;
4333
	p->se.vruntime			= 0;
Peter Zijlstra's avatar
Peter Zijlstra committed
4334
	INIT_LIST_HEAD(&p->se.group_node);
4335

4336 4337 4338 4339
#ifdef CONFIG_FAIR_GROUP_SCHED
	p->se.cfs_rq			= NULL;
#endif

4340
#ifdef CONFIG_SCHEDSTATS
4341
	/* Even if schedstat is disabled, there should not be garbage */
4342
	memset(&p->stats, 0, sizeof(p->stats));
4343
#endif
4344

4345
	RB_CLEAR_NODE(&p->dl.rb_node);
4346
	init_dl_task_timer(&p->dl);
4347
	init_dl_inactive_task_timer(&p->dl);
4348
	__dl_clear_params(p);
4349

Peter Zijlstra's avatar
Peter Zijlstra committed
4350
	INIT_LIST_HEAD(&p->rt.run_list);
4351 4352 4353 4354
	p->rt.timeout		= 0;
	p->rt.time_slice	= sched_rr_timeslice;
	p->rt.on_rq		= 0;
	p->rt.on_list		= 0;
4355

4356 4357 4358
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&p->preempt_notifiers);
#endif
4359

4360 4361 4362
#ifdef CONFIG_COMPACTION
	p->capture_control = NULL;
#endif
4363
	init_numa_balancing(clone_flags, p);
4364
#ifdef CONFIG_SMP
4365
	p->wake_entry.u_flags = CSD_TYPE_TTWU;
4366
	p->migration_pending = NULL;
4367
#endif
Ingo Molnar's avatar
Ingo Molnar committed
4368 4369
}

4370 4371
DEFINE_STATIC_KEY_FALSE(sched_numa_balancing);

4372
#ifdef CONFIG_NUMA_BALANCING
4373

4374 4375 4376
int sysctl_numa_balancing_mode;

static void __set_numabalancing_state(bool enabled)
4377 4378
{
	if (enabled)
4379
		static_branch_enable(&sched_numa_balancing);
4380
	else
4381
		static_branch_disable(&sched_numa_balancing);
4382
}
4383

4384 4385 4386 4387 4388 4389 4390 4391 4392
void set_numabalancing_state(bool enabled)
{
	if (enabled)
		sysctl_numa_balancing_mode = NUMA_BALANCING_NORMAL;
	else
		sysctl_numa_balancing_mode = NUMA_BALANCING_DISABLED;
	__set_numabalancing_state(enabled);
}

4393 4394
#ifdef CONFIG_PROC_SYSCTL
int sysctl_numa_balancing(struct ctl_table *table, int write,
4395
			  void *buffer, size_t *lenp, loff_t *ppos)
4396 4397 4398
{
	struct ctl_table t;
	int err;
4399
	int state = sysctl_numa_balancing_mode;
4400 4401 4402 4403 4404 4405 4406 4407 4408

	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;
4409 4410 4411 4412
	if (write) {
		sysctl_numa_balancing_mode = state;
		__set_numabalancing_state(state);
	}
4413 4414 4415 4416
	return err;
}
#endif
#endif
Ingo Molnar's avatar
Ingo Molnar committed
4417

4418 4419
#ifdef CONFIG_SCHEDSTATS

4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444
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")) {
4445
		set_schedstats(true);
4446 4447
		ret = 1;
	} else if (!strcmp(str, "disable")) {
4448
		set_schedstats(false);
4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459
		ret = 1;
	}
out:
	if (!ret)
		pr_warn("Unable to parse schedstats=\n");

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

#ifdef CONFIG_PROC_SYSCTL
4460
static int sysctl_schedstats(struct ctl_table *table, int write, void *buffer,
4461
		size_t *lenp, loff_t *ppos)
4462 4463 4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478
{
	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;
}
4479 4480
#endif /* CONFIG_PROC_SYSCTL */
#endif /* CONFIG_SCHEDSTATS */
Ingo Molnar's avatar
Ingo Molnar committed
4481

4482 4483 4484
#ifdef CONFIG_SYSCTL
static struct ctl_table sched_core_sysctls[] = {
#ifdef CONFIG_SCHEDSTATS
4485 4486 4487 4488 4489 4490 4491 4492 4493
	{
		.procname       = "sched_schedstats",
		.data           = NULL,
		.maxlen         = sizeof(unsigned int),
		.mode           = 0644,
		.proc_handler   = sysctl_schedstats,
		.extra1         = SYSCTL_ZERO,
		.extra2         = SYSCTL_ONE,
	},
4494 4495 4496 4497 4498 4499 4500 4501 4502 4503 4504 4505 4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517
#endif /* CONFIG_SCHEDSTATS */
#ifdef CONFIG_UCLAMP_TASK
	{
		.procname       = "sched_util_clamp_min",
		.data           = &sysctl_sched_uclamp_util_min,
		.maxlen         = sizeof(unsigned int),
		.mode           = 0644,
		.proc_handler   = sysctl_sched_uclamp_handler,
	},
	{
		.procname       = "sched_util_clamp_max",
		.data           = &sysctl_sched_uclamp_util_max,
		.maxlen         = sizeof(unsigned int),
		.mode           = 0644,
		.proc_handler   = sysctl_sched_uclamp_handler,
	},
	{
		.procname       = "sched_util_clamp_min_rt_default",
		.data           = &sysctl_sched_uclamp_util_min_rt_default,
		.maxlen         = sizeof(unsigned int),
		.mode           = 0644,
		.proc_handler   = sysctl_sched_uclamp_handler,
	},
#endif /* CONFIG_UCLAMP_TASK */
4518 4519
	{}
};
4520
static int __init sched_core_sysctl_init(void)
4521
{
4522
	register_sysctl_init("kernel", sched_core_sysctls);
4523 4524
	return 0;
}
4525 4526
late_initcall(sched_core_sysctl_init);
#endif /* CONFIG_SYSCTL */
Ingo Molnar's avatar
Ingo Molnar committed
4527 4528 4529 4530

/*
 * fork()/clone()-time setup:
 */
4531
int sched_fork(unsigned long clone_flags, struct task_struct *p)
Ingo Molnar's avatar
Ingo Molnar committed
4532
{
4533
	__sched_fork(clone_flags, p);
4534
	/*
4535
	 * We mark the process as NEW here. This guarantees that
4536 4537 4538
	 * nobody will actually run it, and a signal or other external
	 * event cannot wake it up and insert it on the runqueue either.
	 */
4539
	p->__state = TASK_NEW;
Ingo Molnar's avatar
Ingo Molnar committed
4540

4541 4542 4543 4544 4545
	/*
	 * Make sure we do not leak PI boosting priority to the child.
	 */
	p->prio = current->normal_prio;

4546 4547
	uclamp_fork(p);

4548 4549 4550 4551
	/*
	 * Revert to default priority/policy on fork if requested.
	 */
	if (unlikely(p->sched_reset_on_fork)) {
4552
		if (task_has_dl_policy(p) || task_has_rt_policy(p)) {
4553
			p->policy = SCHED_NORMAL;
4554
			p->static_prio = NICE_TO_PRIO(0);
4555 4556 4557 4558
			p->rt_priority = 0;
		} else if (PRIO_TO_NICE(p->static_prio) < 0)
			p->static_prio = NICE_TO_PRIO(0);

4559
		p->prio = p->normal_prio = p->static_prio;
4560
		set_load_weight(p, false);
4561

4562 4563 4564 4565 4566 4567
		/*
		 * We don't need the reset flag anymore after the fork. It has
		 * fulfilled its duty:
		 */
		p->sched_reset_on_fork = 0;
	}
4568

4569
	if (dl_prio(p->prio))
4570
		return -EAGAIN;
4571
	else if (rt_prio(p->prio))
4572
		p->sched_class = &rt_sched_class;
4573
	else
4574
		p->sched_class = &fair_sched_class;
4575

4576
	init_entity_runnable_average(&p->se);
4577

4578

4579
#ifdef CONFIG_SCHED_INFO
Ingo Molnar's avatar
Ingo Molnar committed
4580
	if (likely(sched_info_on()))
4581
		memset(&p->sched_info, 0, sizeof(p->sched_info));
Linus Torvalds's avatar
Linus Torvalds committed
4582
#endif
4583 4584
#if defined(CONFIG_SMP)
	p->on_cpu = 0;
4585
#endif
4586
	init_task_preempt_count(p);
4587
#ifdef CONFIG_SMP
4588
	plist_node_init(&p->pushable_tasks, MAX_PRIO);
4589
	RB_CLEAR_NODE(&p->pushable_dl_tasks);
4590
#endif
4591
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
4592 4593
}

4594
void sched_cgroup_fork(struct task_struct *p, struct kernel_clone_args *kargs)
4595
{
4596 4597
	unsigned long flags;

4598 4599 4600 4601
	/*
	 * Because we're not yet on the pid-hash, p->pi_lock isn't strictly
	 * required yet, but lockdep gets upset if rules are violated.
	 */
4602 4603
	raw_spin_lock_irqsave(&p->pi_lock, flags);
#ifdef CONFIG_CGROUP_SCHED
4604 4605 4606 4607 4608 4609 4610
	if (1) {
		struct task_group *tg;
		tg = container_of(kargs->cset->subsys[cpu_cgrp_id],
				  struct task_group, css);
		tg = autogroup_task_group(p, tg);
		p->sched_task_group = tg;
	}
4611 4612 4613 4614 4615 4616 4617 4618 4619 4620
#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);
4621
}
4622

4623 4624
void sched_post_fork(struct task_struct *p)
{
4625 4626 4627
	uclamp_post_fork(p);
}

4628 4629 4630
unsigned long to_ratio(u64 period, u64 runtime)
{
	if (runtime == RUNTIME_INF)
4631
		return BW_UNIT;
4632 4633 4634 4635 4636 4637 4638 4639 4640

	/*
	 * 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;

4641
	return div64_u64(runtime << BW_SHIFT, period);
4642 4643
}

Linus Torvalds's avatar
Linus Torvalds committed
4644 4645 4646 4647 4648 4649 4650
/*
 * 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.
 */
4651
void wake_up_new_task(struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
4652
{
4653
	struct rq_flags rf;
Ingo Molnar's avatar
Ingo Molnar committed
4654
	struct rq *rq;
4655

4656
	raw_spin_lock_irqsave(&p->pi_lock, rf.flags);
4657
	WRITE_ONCE(p->__state, TASK_RUNNING);
4658 4659 4660
#ifdef CONFIG_SMP
	/*
	 * Fork balancing, do it here and not earlier because:
4661
	 *  - cpus_ptr can change in the fork path
4662
	 *  - any previously selected CPU might disappear through hotplug
4663 4664 4665
	 *
	 * Use __set_task_cpu() to avoid calling sched_class::migrate_task_rq,
	 * as we're not fully set-up yet.
4666
	 */
4667
	p->recent_used_cpu = task_cpu(p);
4668
	rseq_migrate(p);
4669
	__set_task_cpu(p, select_task_rq(p, task_cpu(p), WF_FORK));
4670
#endif
4671
	rq = __task_rq_lock(p, &rf);
4672
	update_rq_clock(rq);
4673
	post_init_entity_util_avg(p);
4674

4675
	activate_task(rq, p, ENQUEUE_NOCLOCK);
4676
	trace_sched_wakeup_new(p);
Peter Zijlstra's avatar
Peter Zijlstra committed
4677
	check_preempt_curr(rq, p, WF_FORK);
4678
#ifdef CONFIG_SMP
4679 4680
	if (p->sched_class->task_woken) {
		/*
4681
		 * Nothing relies on rq->lock after this, so it's fine to
4682 4683
		 * drop it.
		 */
4684
		rq_unpin_lock(rq, &rf);
4685
		p->sched_class->task_woken(rq, p);
4686
		rq_repin_lock(rq, &rf);
4687
	}
4688
#endif
4689
	task_rq_unlock(rq, p, &rf);
Linus Torvalds's avatar
Linus Torvalds committed
4690 4691
}

4692 4693
#ifdef CONFIG_PREEMPT_NOTIFIERS

4694
static DEFINE_STATIC_KEY_FALSE(preempt_notifier_key);
4695

4696 4697
void preempt_notifier_inc(void)
{
4698
	static_branch_inc(&preempt_notifier_key);
4699 4700 4701 4702 4703
}
EXPORT_SYMBOL_GPL(preempt_notifier_inc);

void preempt_notifier_dec(void)
{
4704
	static_branch_dec(&preempt_notifier_key);
4705 4706 4707
}
EXPORT_SYMBOL_GPL(preempt_notifier_dec);

4708
/**
4709
 * preempt_notifier_register - tell me when current is being preempted & rescheduled
4710
 * @notifier: notifier struct to register
4711 4712 4713
 */
void preempt_notifier_register(struct preempt_notifier *notifier)
{
4714
	if (!static_branch_unlikely(&preempt_notifier_key))
4715 4716
		WARN(1, "registering preempt_notifier while notifiers disabled\n");

4717 4718 4719 4720 4721 4722
	hlist_add_head(&notifier->link, &current->preempt_notifiers);
}
EXPORT_SYMBOL_GPL(preempt_notifier_register);

/**
 * preempt_notifier_unregister - no longer interested in preemption notifications
4723
 * @notifier: notifier struct to unregister
4724
 *
4725
 * This is *not* safe to call from within a preemption notifier.
4726 4727 4728 4729 4730 4731 4732
 */
void preempt_notifier_unregister(struct preempt_notifier *notifier)
{
	hlist_del(&notifier->link);
}
EXPORT_SYMBOL_GPL(preempt_notifier_unregister);

4733
static void __fire_sched_in_preempt_notifiers(struct task_struct *curr)
4734 4735 4736
{
	struct preempt_notifier *notifier;

4737
	hlist_for_each_entry(notifier, &curr->preempt_notifiers, link)
4738 4739 4740
		notifier->ops->sched_in(notifier, raw_smp_processor_id());
}

4741 4742
static __always_inline void fire_sched_in_preempt_notifiers(struct task_struct *curr)
{
4743
	if (static_branch_unlikely(&preempt_notifier_key))
4744 4745 4746
		__fire_sched_in_preempt_notifiers(curr);
}

4747
static void
4748 4749
__fire_sched_out_preempt_notifiers(struct task_struct *curr,
				   struct task_struct *next)
4750 4751 4752
{
	struct preempt_notifier *notifier;

4753
	hlist_for_each_entry(notifier, &curr->preempt_notifiers, link)
4754 4755 4756
		notifier->ops->sched_out(notifier, next);
}

4757 4758 4759 4760
static __always_inline void
fire_sched_out_preempt_notifiers(struct task_struct *curr,
				 struct task_struct *next)
{
4761
	if (static_branch_unlikely(&preempt_notifier_key))
4762 4763 4764
		__fire_sched_out_preempt_notifiers(curr, next);
}

4765
#else /* !CONFIG_PREEMPT_NOTIFIERS */
4766

4767
static inline void fire_sched_in_preempt_notifiers(struct task_struct *curr)
4768 4769 4770
{
}

4771
static inline void
4772 4773 4774 4775 4776
fire_sched_out_preempt_notifiers(struct task_struct *curr,
				 struct task_struct *next)
{
}

4777
#endif /* CONFIG_PREEMPT_NOTIFIERS */
4778

4779 4780 4781 4782 4783 4784
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.
4785
	 *
4786 4787
	 * See the smp_load_acquire(&p->on_cpu) case in ttwu() and
	 * its ordering comment.
4788
	 */
4789
	WRITE_ONCE(next->on_cpu, 1);
4790 4791 4792 4793 4794 4795 4796
#endif
}

static inline void finish_task(struct task_struct *prev)
{
#ifdef CONFIG_SMP
	/*
4797 4798 4799
	 * 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
4800 4801 4802 4803 4804 4805 4806 4807 4808 4809 4810
	 * 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
}

4811 4812 4813 4814 4815 4816 4817
#ifdef CONFIG_SMP

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

4818
	lockdep_assert_rq_held(rq);
4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829

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

		func(rq);
	}
}

4830 4831
static void balance_push(struct rq *rq);

4832 4833 4834 4835 4836 4837 4838 4839 4840 4841 4842
/*
 * balance_push_callback is a right abuse of the callback interface and plays
 * by significantly different rules.
 *
 * Where the normal balance_callback's purpose is to be ran in the same context
 * that queued it (only later, when it's safe to drop rq->lock again),
 * balance_push_callback is specifically targeted at __schedule().
 *
 * This abuse is tolerated because it places all the unlikely/odd cases behind
 * a single test, namely: rq->balance_callback == NULL.
 */
4843 4844 4845 4846 4847
struct callback_head balance_push_callback = {
	.next = NULL,
	.func = (void (*)(struct callback_head *))balance_push,
};

4848 4849
static inline struct callback_head *
__splice_balance_callbacks(struct rq *rq, bool split)
4850 4851 4852
{
	struct callback_head *head = rq->balance_callback;

4853 4854 4855
	if (likely(!head))
		return NULL;

4856
	lockdep_assert_rq_held(rq);
4857 4858 4859 4860 4861 4862 4863 4864 4865 4866 4867
	/*
	 * Must not take balance_push_callback off the list when
	 * splice_balance_callbacks() and balance_callbacks() are not
	 * in the same rq->lock section.
	 *
	 * In that case it would be possible for __schedule() to interleave
	 * and observe the list empty.
	 */
	if (split && head == &balance_push_callback)
		head = NULL;
	else
4868 4869 4870 4871 4872
		rq->balance_callback = NULL;

	return head;
}

4873 4874 4875 4876 4877
static inline struct callback_head *splice_balance_callbacks(struct rq *rq)
{
	return __splice_balance_callbacks(rq, true);
}

4878 4879
static void __balance_callbacks(struct rq *rq)
{
4880
	do_balance_callbacks(rq, __splice_balance_callbacks(rq, false));
4881 4882 4883 4884 4885 4886 4887
}

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

	if (unlikely(head)) {
4888
		raw_spin_rq_lock_irqsave(rq, flags);
4889
		do_balance_callbacks(rq, head);
4890
		raw_spin_rq_unlock_irqrestore(rq, flags);
4891 4892 4893 4894 4895 4896 4897 4898 4899 4900 4901 4902 4903 4904 4905 4906 4907 4908 4909 4910
	}
}

#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

4911 4912
static inline void
prepare_lock_switch(struct rq *rq, struct task_struct *next, struct rq_flags *rf)
4913
{
4914 4915 4916 4917 4918 4919 4920
	/*
	 * 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);
4921
	spin_release(&__rq_lockp(rq)->dep_map, _THIS_IP_);
4922 4923
#ifdef CONFIG_DEBUG_SPINLOCK
	/* this is a valid case when another task releases the spinlock */
4924
	rq_lockp(rq)->owner = next;
4925
#endif
4926 4927 4928 4929
}

static inline void finish_lock_switch(struct rq *rq)
{
4930 4931 4932 4933 4934
	/*
	 * If we are tracking spinlock dependencies then we have to
	 * fix up the runqueue lock - which gets 'carried over' from
	 * prev into current:
	 */
4935
	spin_acquire(&__rq_lockp(rq)->dep_map, 0, 0, _THIS_IP_);
4936
	__balance_callbacks(rq);
4937
	raw_spin_rq_unlock_irq(rq);
4938 4939
}

4940 4941 4942 4943 4944 4945 4946 4947 4948 4949 4950 4951
/*
 * 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

4952 4953 4954 4955 4956 4957 4958 4959 4960 4961 4962 4963 4964 4965 4966 4967
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
}

4968 4969 4970
/**
 * prepare_task_switch - prepare to switch tasks
 * @rq: the runqueue preparing to switch
4971
 * @prev: the current task that is being switched out
4972 4973 4974 4975 4976 4977 4978 4979 4980
 * @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.
 */
4981 4982 4983
static inline void
prepare_task_switch(struct rq *rq, struct task_struct *prev,
		    struct task_struct *next)
4984
{
4985
	kcov_prepare_switch(prev);
4986
	sched_info_switch(rq, prev, next);
4987
	perf_event_task_sched_out(prev, next);
4988
	rseq_preempt(prev);
4989
	fire_sched_out_preempt_notifiers(prev, next);
4990
	kmap_local_sched_out();
4991
	prepare_task(next);
4992 4993 4994
	prepare_arch_switch(next);
}

Linus Torvalds's avatar
Linus Torvalds committed
4995 4996 4997 4998
/**
 * finish_task_switch - clean up after a task-switch
 * @prev: the thread we just switched away from.
 *
4999 5000 5001 5002
 * 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
5003 5004
 *
 * Note that we may have delayed dropping an mm in context_switch(). If
Ingo Molnar's avatar
Ingo Molnar committed
5005
 * so, we finish that here outside of the runqueue lock. (Doing it
Linus Torvalds's avatar
Linus Torvalds committed
5006 5007
 * with the lock held can cause deadlocks; see schedule() for
 * details.)
5008 5009 5010 5011 5012
 *
 * 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
5013
 */
5014
static struct rq *finish_task_switch(struct task_struct *prev)
Linus Torvalds's avatar
Linus Torvalds committed
5015 5016
	__releases(rq->lock)
{
5017
	struct rq *rq = this_rq();
Linus Torvalds's avatar
Linus Torvalds committed
5018
	struct mm_struct *mm = rq->prev_mm;
5019
	unsigned int prev_state;
Linus Torvalds's avatar
Linus Torvalds committed
5020

5021 5022 5023 5024 5025 5026 5027 5028 5029 5030 5031
	/*
	 * 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.
	 */
5032 5033 5034 5035
	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);
5036

Linus Torvalds's avatar
Linus Torvalds committed
5037 5038 5039 5040
	rq->prev_mm = NULL;

	/*
	 * A task struct has one reference for the use as "current".
5041
	 * If a task dies, then it sets TASK_DEAD in tsk->state and calls
Oleg Nesterov's avatar
Oleg Nesterov committed
5042 5043
	 * schedule one last time. The schedule call will never return, and
	 * the scheduled task must drop that reference.
5044 5045
	 *
	 * We must observe prev->state before clearing prev->on_cpu (in
5046
	 * finish_task), otherwise a concurrent wakeup can get prev
5047 5048
	 * 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
5049
	 */
5050
	prev_state = READ_ONCE(prev->__state);
5051
	vtime_task_switch(prev);
5052
	perf_event_task_sched_in(prev, current);
5053
	finish_task(prev);
5054
	tick_nohz_task_switch();
5055
	finish_lock_switch(rq);
5056
	finish_arch_post_lock_switch();
5057
	kcov_finish_switch(current);
5058 5059 5060 5061 5062 5063 5064 5065
	/*
	 * 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
5066

5067
	fire_sched_in_preempt_notifiers(current);
5068
	/*
5069 5070 5071 5072 5073 5074 5075 5076 5077 5078
	 * 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.
5079
	 */
5080 5081
	if (mm) {
		membarrier_mm_sync_core_before_usermode(mm);
5082
		mmdrop_sched(mm);
5083
	}
5084 5085 5086
	if (unlikely(prev_state == TASK_DEAD)) {
		if (prev->sched_class->task_dead)
			prev->sched_class->task_dead(prev);
5087

5088 5089 5090
		/* Task is done with its stack. */
		put_task_stack(prev);

5091
		put_task_struct_rcu_user(prev);
5092
	}
5093

5094
	return rq;
Linus Torvalds's avatar
Linus Torvalds committed
5095 5096 5097 5098 5099 5100
}

/**
 * schedule_tail - first thing a freshly forked thread must call.
 * @prev: the thread we just switched away from.
 */
5101
asmlinkage __visible void schedule_tail(struct task_struct *prev)
Linus Torvalds's avatar
Linus Torvalds committed
5102 5103
	__releases(rq->lock)
{
5104 5105 5106 5107 5108 5109 5110 5111 5112
	/*
	 * 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).
	 */

5113
	finish_task_switch(prev);
5114
	preempt_enable();
5115

Linus Torvalds's avatar
Linus Torvalds committed
5116
	if (current->set_child_tid)
5117
		put_user(task_pid_vnr(current), current->set_child_tid);
5118 5119

	calculate_sigpending();
Linus Torvalds's avatar
Linus Torvalds committed
5120 5121 5122
}

/*
5123
 * context_switch - switch to the new MM and the new thread's register state.
Linus Torvalds's avatar
Linus Torvalds committed
5124
 */
5125
static __always_inline struct rq *
5126
context_switch(struct rq *rq, struct task_struct *prev,
5127
	       struct task_struct *next, struct rq_flags *rf)
Linus Torvalds's avatar
Linus Torvalds committed
5128
{
5129
	prepare_task_switch(rq, prev, next);
5130

5131 5132 5133 5134 5135
	/*
	 * For paravirt, this is coupled with an exit in switch_to to
	 * combine the page table reload and the switch backend into
	 * one hypercall.
	 */
5136
	arch_start_context_switch(prev);
5137

5138
	/*
5139 5140 5141 5142 5143
	 * kernel -> kernel   lazy + transfer active
	 *   user -> kernel   lazy + mmgrab() active
	 *
	 * kernel ->   user   switch + mmdrop() active
	 *   user ->   user   switch
5144
	 */
5145 5146 5147 5148 5149 5150 5151 5152 5153
	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
5154
		membarrier_switch_mm(rq, prev->active_mm, next->mm);
5155 5156
		/*
		 * sys_membarrier() requires an smp_mb() between setting
5157
		 * rq->curr / membarrier_switch_mm() and returning to userspace.
5158 5159 5160 5161 5162 5163
		 *
		 * 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
5164

5165 5166 5167 5168 5169
		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
5170
	}
5171

5172
	rq->clock_update_flags &= ~(RQCF_ACT_SKIP|RQCF_REQ_SKIP);
5173

5174
	prepare_lock_switch(rq, next, rf);
Linus Torvalds's avatar
Linus Torvalds committed
5175 5176 5177

	/* Here we just switch the register state and the stack. */
	switch_to(prev, next, prev);
Ingo Molnar's avatar
Ingo Molnar committed
5178
	barrier();
5179 5180

	return finish_task_switch(prev);
Linus Torvalds's avatar
Linus Torvalds committed
5181 5182 5183
}

/*
5184
 * nr_running and nr_context_switches:
Linus Torvalds's avatar
Linus Torvalds committed
5185 5186
 *
 * externally visible scheduler statistics: current number of runnable
5187
 * threads, total number of context switches performed since bootup.
Linus Torvalds's avatar
Linus Torvalds committed
5188
 */
5189
unsigned int nr_running(void)
Linus Torvalds's avatar
Linus Torvalds committed
5190
{
5191
	unsigned int i, sum = 0;
Linus Torvalds's avatar
Linus Torvalds committed
5192 5193 5194 5195 5196

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

	return sum;
5197
}
Linus Torvalds's avatar
Linus Torvalds committed
5198

5199
/*
5200
 * Check if only the current task is running on the CPU.
5201 5202 5203 5204 5205
 *
 * 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:
 *
5206
 * - from a non-preemptible section (of course)
5207 5208 5209 5210
 *
 * - from a thread that is bound to a single CPU
 *
 * - in a loop with very short iterations (e.g. a polling loop)
5211 5212 5213
 */
bool single_task_running(void)
{
5214
	return raw_rq()->nr_running == 1;
5215 5216 5217
}
EXPORT_SYMBOL(single_task_running);

Linus Torvalds's avatar
Linus Torvalds committed
5218
unsigned long long nr_context_switches(void)
5219
{
5220 5221
	int i;
	unsigned long long sum = 0;
5222

5223
	for_each_possible_cpu(i)
Linus Torvalds's avatar
Linus Torvalds committed
5224
		sum += cpu_rq(i)->nr_switches;
5225

Linus Torvalds's avatar
Linus Torvalds committed
5226 5227
	return sum;
}
5228

5229 5230 5231 5232 5233 5234 5235
/*
 * 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.
 */

5236
unsigned int nr_iowait_cpu(int cpu)
5237 5238 5239 5240
{
	return atomic_read(&cpu_rq(cpu)->nr_iowait);
}

5241
/*
5242
 * IO-wait accounting, and how it's mostly bollocks (on SMP).
5243 5244 5245 5246 5247 5248 5249 5250 5251 5252 5253 5254 5255 5256 5257 5258 5259 5260 5261 5262 5263 5264 5265 5266 5267 5268 5269 5270
 *
 * 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'.
 */

5271
unsigned int nr_iowait(void)
Linus Torvalds's avatar
Linus Torvalds committed
5272
{
5273
	unsigned int i, sum = 0;
5274

5275
	for_each_possible_cpu(i)
5276
		sum += nr_iowait_cpu(i);
5277

Linus Torvalds's avatar
Linus Torvalds committed
5278 5279
	return sum;
}
5280

Ingo Molnar's avatar
Ingo Molnar committed
5281
#ifdef CONFIG_SMP
5282

5283
/*
5284 5285
 * sched_exec - execve() is a valuable balancing opportunity, because at
 * this point the task has the smallest effective memory and cache footprint.
5286
 */
5287
void sched_exec(void)
5288
{
5289
	struct task_struct *p = current;
Linus Torvalds's avatar
Linus Torvalds committed
5290
	unsigned long flags;
5291
	int dest_cpu;
5292

5293
	raw_spin_lock_irqsave(&p->pi_lock, flags);
5294
	dest_cpu = p->sched_class->select_task_rq(p, task_cpu(p), WF_EXEC);
5295 5296
	if (dest_cpu == smp_processor_id())
		goto unlock;
5297

5298
	if (likely(cpu_active(dest_cpu))) {
5299
		struct migration_arg arg = { p, dest_cpu };
5300

5301 5302
		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
5303 5304
		return;
	}
5305
unlock:
5306
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
5307
}
Ingo Molnar's avatar
Ingo Molnar committed
5308

Linus Torvalds's avatar
Linus Torvalds committed
5309 5310 5311
#endif

DEFINE_PER_CPU(struct kernel_stat, kstat);
5312
DEFINE_PER_CPU(struct kernel_cpustat, kernel_cpustat);
Linus Torvalds's avatar
Linus Torvalds committed
5313 5314

EXPORT_PER_CPU_SYMBOL(kstat);
5315
EXPORT_PER_CPU_SYMBOL(kernel_cpustat);
Linus Torvalds's avatar
Linus Torvalds committed
5316

5317 5318 5319 5320 5321 5322 5323 5324 5325 5326 5327 5328 5329 5330 5331 5332 5333
/*
 * 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);
}

5334 5335 5336 5337 5338 5339 5340
/*
 * 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)
{
5341
	struct rq_flags rf;
5342
	struct rq *rq;
5343
	u64 ns;
5344

5345 5346
#if defined(CONFIG_64BIT) && defined(CONFIG_SMP)
	/*
5347
	 * 64-bit doesn't need locks to atomically read a 64-bit value.
5348 5349 5350
	 * So we have a optimization chance when the task's delta_exec is 0.
	 * Reading ->on_cpu is racy, but this is ok.
	 *
5351 5352
	 * 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
5353
	 * indistinguishable from the read occurring a few cycles earlier.
5354 5355
	 * If we see ->on_cpu without ->on_rq, the task is leaving, and has
	 * been accounted, so we're correct here as well.
5356
	 */
5357
	if (!p->on_cpu || !task_on_rq_queued(p))
5358 5359 5360
		return p->se.sum_exec_runtime;
#endif

5361
	rq = task_rq_lock(p, &rf);
5362 5363 5364 5365 5366 5367
	/*
	 * 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)) {
5368
		prefetch_curr_exec_start(p);
5369 5370 5371 5372
		update_rq_clock(rq);
		p->sched_class->update_curr(rq);
	}
	ns = p->se.sum_exec_runtime;
5373
	task_rq_unlock(rq, p, &rf);
5374 5375 5376

	return ns;
}
5377

5378 5379 5380 5381 5382 5383 5384 5385 5386 5387 5388 5389 5390 5391 5392 5393 5394 5395 5396 5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410 5411 5412 5413 5414 5415 5416 5417 5418 5419 5420 5421 5422 5423 5424 5425 5426
#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 */

5427 5428 5429 5430 5431 5432 5433 5434
/*
 * 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
5435
	struct task_struct *curr = rq->curr;
5436
	struct rq_flags rf;
5437
	unsigned long thermal_pressure;
5438
	u64 resched_latency;
5439

5440
	arch_scale_freq_tick();
5441
	sched_clock_tick();
Ingo Molnar's avatar
Ingo Molnar committed
5442

5443 5444
	rq_lock(rq, &rf);

5445
	update_rq_clock(rq);
5446
	thermal_pressure = arch_scale_thermal_pressure(cpu_of(rq));
5447
	update_thermal_load_avg(rq_clock_thermal(rq), rq, thermal_pressure);
Peter Zijlstra's avatar
Peter Zijlstra committed
5448
	curr->sched_class->task_tick(rq, curr, 0);
5449 5450
	if (sched_feat(LATENCY_WARN))
		resched_latency = cpu_resched_latency(rq);
5451
	calc_global_load_tick(rq);
5452
	sched_core_tick(rq);
5453 5454

	rq_unlock(rq, &rf);
5455

5456 5457 5458
	if (sched_feat(LATENCY_WARN) && resched_latency)
		resched_latency_warn(cpu, resched_latency);

5459
	perf_event_task_tick();
5460

5461
#ifdef CONFIG_SMP
5462
	rq->idle_balance = idle_cpu(cpu);
5463
	trigger_load_balance(rq);
5464
#endif
Linus Torvalds's avatar
Linus Torvalds committed
5465 5466
}

5467
#ifdef CONFIG_NO_HZ_FULL
5468 5469 5470

struct tick_work {
	int			cpu;
5471
	atomic_t		state;
5472 5473
	struct delayed_work	work;
};
5474 5475 5476 5477 5478 5479 5480 5481 5482 5483 5484 5485 5486 5487 5488 5489 5490 5491 5492 5493 5494 5495 5496 5497 5498 5499 5500
/* 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.
 */
5501 5502 5503 5504 5505 5506 5507 5508 5509

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);
5510
	struct task_struct *curr;
5511
	struct rq_flags rf;
5512
	u64 delta;
5513
	int os;
5514 5515 5516 5517 5518 5519 5520 5521

	/*
	 * 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.
	 */
5522
	if (!tick_nohz_tick_stopped_cpu(cpu))
5523
		goto out_requeue;
5524

5525 5526
	rq_lock_irq(rq, &rf);
	curr = rq->curr;
5527
	if (cpu_is_offline(cpu))
5528
		goto out_unlock;
5529

5530 5531
	update_rq_clock(rq);

5532 5533 5534 5535 5536 5537 5538 5539
	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);
	}
5540 5541
	curr->sched_class->task_tick(rq, curr, 0);

5542
	calc_load_nohz_remote(rq);
5543 5544 5545
out_unlock:
	rq_unlock_irq(rq, &rf);
out_requeue:
5546

5547 5548 5549
	/*
	 * Run the remote tick once per second (1Hz). This arbitrary
	 * frequency is large enough to avoid overload but short enough
5550 5551
	 * to keep scheduler internal stats reasonably up to date.  But
	 * first update state to reflect hotplug activity if required.
5552
	 */
5553 5554 5555 5556
	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);
5557 5558 5559 5560
}

static void sched_tick_start(int cpu)
{
5561
	int os;
5562 5563
	struct tick_work *twork;

5564
	if (housekeeping_cpu(cpu, HK_TYPE_TICK))
5565 5566 5567 5568 5569
		return;

	WARN_ON_ONCE(!tick_work_cpu);

	twork = per_cpu_ptr(tick_work_cpu, cpu);
5570 5571 5572 5573 5574 5575 5576
	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);
	}
5577 5578 5579 5580 5581 5582
}

#ifdef CONFIG_HOTPLUG_CPU
static void sched_tick_stop(int cpu)
{
	struct tick_work *twork;
5583
	int os;
5584

5585
	if (housekeeping_cpu(cpu, HK_TYPE_TICK))
5586 5587 5588 5589 5590
		return;

	WARN_ON_ONCE(!tick_work_cpu);

	twork = per_cpu_ptr(tick_work_cpu, cpu);
5591 5592 5593 5594
	/* 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. */
5595 5596 5597 5598 5599 5600 5601 5602 5603 5604 5605 5606 5607
}
#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) { }
5608
#endif
Linus Torvalds's avatar
Linus Torvalds committed
5609

5610
#if defined(CONFIG_PREEMPTION) && (defined(CONFIG_DEBUG_PREEMPT) || \
5611
				defined(CONFIG_TRACE_PREEMPT_TOGGLE))
5612 5613 5614 5615 5616 5617 5618 5619 5620 5621 5622 5623 5624 5625
/*
 * 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);
	}
}
5626

5627
void preempt_count_add(int val)
Linus Torvalds's avatar
Linus Torvalds committed
5628
{
5629
#ifdef CONFIG_DEBUG_PREEMPT
Linus Torvalds's avatar
Linus Torvalds committed
5630 5631 5632
	/*
	 * Underflow?
	 */
5633 5634
	if (DEBUG_LOCKS_WARN_ON((preempt_count() < 0)))
		return;
5635
#endif
5636
	__preempt_count_add(val);
5637
#ifdef CONFIG_DEBUG_PREEMPT
Linus Torvalds's avatar
Linus Torvalds committed
5638 5639 5640
	/*
	 * Spinlock count overflowing soon?
	 */
5641 5642
	DEBUG_LOCKS_WARN_ON((preempt_count() & PREEMPT_MASK) >=
				PREEMPT_MASK - 10);
5643
#endif
5644
	preempt_latency_start(val);
Linus Torvalds's avatar
Linus Torvalds committed
5645
}
5646
EXPORT_SYMBOL(preempt_count_add);
5647
NOKPROBE_SYMBOL(preempt_count_add);
Linus Torvalds's avatar
Linus Torvalds committed
5648

5649 5650 5651 5652 5653 5654 5655 5656 5657 5658
/*
 * 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());
}

5659
void preempt_count_sub(int val)
Linus Torvalds's avatar
Linus Torvalds committed
5660
{
5661
#ifdef CONFIG_DEBUG_PREEMPT
Linus Torvalds's avatar
Linus Torvalds committed
5662 5663 5664
	/*
	 * Underflow?
	 */
5665
	if (DEBUG_LOCKS_WARN_ON(val > preempt_count()))
5666
		return;
Linus Torvalds's avatar
Linus Torvalds committed
5667 5668 5669
	/*
	 * Is the spinlock portion underflowing?
	 */
5670 5671 5672
	if (DEBUG_LOCKS_WARN_ON((val < PREEMPT_MASK) &&
			!(preempt_count() & PREEMPT_MASK)))
		return;
5673
#endif
5674

5675
	preempt_latency_stop(val);
5676
	__preempt_count_sub(val);
Linus Torvalds's avatar
Linus Torvalds committed
5677
}
5678
EXPORT_SYMBOL(preempt_count_sub);
5679
NOKPROBE_SYMBOL(preempt_count_sub);
Linus Torvalds's avatar
Linus Torvalds committed
5680

5681 5682 5683
#else
static inline void preempt_latency_start(int val) { }
static inline void preempt_latency_stop(int val) { }
Linus Torvalds's avatar
Linus Torvalds committed
5684 5685
#endif

5686 5687 5688 5689 5690 5691 5692 5693 5694
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
5695
/*
Ingo Molnar's avatar
Ingo Molnar committed
5696
 * Print scheduling while atomic bug:
Linus Torvalds's avatar
Linus Torvalds committed
5697
 */
Ingo Molnar's avatar
Ingo Molnar committed
5698
static noinline void __schedule_bug(struct task_struct *prev)
Linus Torvalds's avatar
Linus Torvalds committed
5699
{
5700 5701 5702
	/* Save this before calling printk(), since that will clobber it */
	unsigned long preempt_disable_ip = get_preempt_disable_ip(current);

5703 5704 5705
	if (oops_in_progress)
		return;

5706 5707
	printk(KERN_ERR "BUG: scheduling while atomic: %s/%d/0x%08x\n",
		prev->comm, prev->pid, preempt_count());
5708

Ingo Molnar's avatar
Ingo Molnar committed
5709
	debug_show_held_locks(prev);
5710
	print_modules();
Ingo Molnar's avatar
Ingo Molnar committed
5711 5712
	if (irqs_disabled())
		print_irqtrace_events(prev);
5713 5714
	if (IS_ENABLED(CONFIG_DEBUG_PREEMPT)
	    && in_atomic_preempt_off()) {
5715
		pr_err("Preemption disabled at:");
5716
		print_ip_sym(KERN_ERR, preempt_disable_ip);
5717
	}
5718 5719 5720
	if (panic_on_warn)
		panic("scheduling while atomic\n");

5721
	dump_stack();
5722
	add_taint(TAINT_WARN, LOCKDEP_STILL_OK);
Ingo Molnar's avatar
Ingo Molnar committed
5723
}
Linus Torvalds's avatar
Linus Torvalds committed
5724

Ingo Molnar's avatar
Ingo Molnar committed
5725 5726 5727
/*
 * Various schedule()-time debugging checks and statistics:
 */
5728
static inline void schedule_debug(struct task_struct *prev, bool preempt)
Ingo Molnar's avatar
Ingo Molnar committed
5729
{
5730
#ifdef CONFIG_SCHED_STACK_END_CHECK
5731 5732
	if (task_stack_end_corrupted(prev))
		panic("corrupted stack end detected inside scheduler\n");
5733 5734 5735

	if (task_scs_end_corrupted(prev))
		panic("corrupted shadow stack detected inside scheduler\n");
5736
#endif
5737

5738
#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
5739
	if (!preempt && READ_ONCE(prev->__state) && prev->non_block_count) {
5740 5741 5742 5743 5744 5745 5746
		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

5747
	if (unlikely(in_atomic_preempt_off())) {
Ingo Molnar's avatar
Ingo Molnar committed
5748
		__schedule_bug(prev);
5749 5750
		preempt_count_set(PREEMPT_DISABLED);
	}
5751
	rcu_sleep_check();
5752
	SCHED_WARN_ON(ct_state() == CONTEXT_USER);
Ingo Molnar's avatar
Ingo Molnar committed
5753

Linus Torvalds's avatar
Linus Torvalds committed
5754 5755
	profile_hit(SCHED_PROFILING, __builtin_return_address(0));

5756
	schedstat_inc(this_rq()->sched_count);
Ingo Molnar's avatar
Ingo Molnar committed
5757 5758
}

5759 5760 5761 5762 5763 5764 5765 5766 5767 5768 5769 5770 5771 5772 5773 5774 5775 5776 5777 5778 5779 5780
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
5781 5782 5783 5784
/*
 * Pick up the highest-prio task:
 */
static inline struct task_struct *
5785
__pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
Ingo Molnar's avatar
Ingo Molnar committed
5786
{
5787
	const struct sched_class *class;
Ingo Molnar's avatar
Ingo Molnar committed
5788
	struct task_struct *p;
Linus Torvalds's avatar
Linus Torvalds committed
5789 5790

	/*
5791 5792
	 * 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
5793
	 * higher scheduling class, because otherwise those lose the
5794
	 * opportunity to pull in more work from other CPUs.
Linus Torvalds's avatar
Linus Torvalds committed
5795
	 */
5796
	if (likely(!sched_class_above(prev->sched_class, &fair_sched_class) &&
5797 5798
		   rq->nr_running == rq->cfs.h_nr_running)) {

5799
		p = pick_next_task_fair(rq, prev, rf);
5800
		if (unlikely(p == RETRY_TASK))
5801
			goto restart;
5802

5803
		/* Assume the next prioritized class is idle_sched_class */
5804
		if (!p) {
5805
			put_prev_task(rq, prev);
5806
			p = pick_next_task_idle(rq);
5807
		}
5808 5809

		return p;
Linus Torvalds's avatar
Linus Torvalds committed
5810 5811
	}

5812
restart:
5813
	put_prev_task_balance(rq, prev, rf);
5814

5815
	for_each_class(class) {
5816
		p = class->pick_next_task(rq);
5817
		if (p)
Ingo Molnar's avatar
Ingo Molnar committed
5818 5819
			return p;
	}
5820

5821
	BUG(); /* The idle class should always have a runnable task. */
Ingo Molnar's avatar
Ingo Molnar committed
5822
}
Linus Torvalds's avatar
Linus Torvalds committed
5823

Peter Zijlstra's avatar
Peter Zijlstra committed
5824
#ifdef CONFIG_SCHED_CORE
5825 5826 5827 5828 5829 5830 5831 5832 5833 5834 5835 5836 5837 5838 5839 5840 5841 5842
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;
}

5843
static inline struct task_struct *pick_task(struct rq *rq)
5844
{
5845 5846
	const struct sched_class *class;
	struct task_struct *p;
5847

5848 5849 5850 5851
	for_each_class(class) {
		p = class->pick_task(rq);
		if (p)
			return p;
5852 5853
	}

5854
	BUG(); /* The idle class should always have a runnable task. */
5855 5856
}

5857 5858
extern void task_vruntime_update(struct rq *rq, struct task_struct *p, bool in_fi);

5859 5860
static void queue_core_balance(struct rq *rq);

5861 5862 5863
static struct task_struct *
pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
{
5864
	struct task_struct *next, *p, *max = NULL;
5865
	const struct cpumask *smt_mask;
5866
	bool fi_before = false;
5867
	bool core_clock_updated = (rq == rq->core);
5868 5869 5870
	unsigned long cookie;
	int i, cpu, occ = 0;
	struct rq *rq_i;
5871 5872 5873 5874 5875 5876 5877 5878 5879 5880 5881 5882 5883 5884 5885 5886 5887 5888 5889 5890 5891 5892 5893 5894 5895 5896 5897 5898 5899 5900 5901 5902 5903 5904 5905 5906 5907 5908 5909
	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;
5910
		goto out;
5911 5912 5913 5914 5915
	}

	put_prev_task_balance(rq, prev, rf);

	smt_mask = cpu_smt_mask(cpu);
5916 5917 5918 5919
	need_sync = !!rq->core->core_cookie;

	/* reset state */
	rq->core->core_cookie = 0UL;
5920 5921 5922 5923 5924 5925 5926 5927 5928 5929
	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;
5930 5931 5932
		need_sync = true;
		fi_before = true;
	}
5933 5934 5935 5936 5937 5938 5939 5940 5941 5942 5943 5944 5945

	/*
	 * 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++;

5946 5947 5948 5949 5950
	/*
	 * Optimize for common case where this CPU has no cookies
	 * and there are no cookied tasks running on siblings.
	 */
	if (!need_sync) {
5951
		next = pick_task(rq);
5952 5953
		if (!next->core_cookie) {
			rq->core_pick = NULL;
5954 5955 5956 5957 5958 5959
			/*
			 * For robustness, update the min_vruntime_fi for
			 * unconstrained picks as well.
			 */
			WARN_ON_ONCE(fi_before);
			task_vruntime_update(rq, next, false);
5960
			goto out_set_next;
5961
		}
5962
	}
5963

5964 5965 5966 5967 5968 5969 5970 5971
	/*
	 * 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);
5972

5973 5974 5975 5976 5977 5978
		/*
		 * 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))
5979
			update_rq_clock(rq_i);
5980 5981 5982 5983

		p = rq_i->core_pick = pick_task(rq_i);
		if (!max || prio_less(max, p, fi_before))
			max = p;
5984 5985
	}

5986 5987
	cookie = rq->core->core_cookie = max->core_cookie;

5988
	/*
5989 5990
	 * For each thread: try and find a runnable task that matches @max or
	 * force idle.
5991
	 */
5992 5993 5994
	for_each_cpu(i, smt_mask) {
		rq_i = cpu_rq(i);
		p = rq_i->core_pick;
5995

5996 5997 5998 5999
		if (!cookie_equals(p, cookie)) {
			p = NULL;
			if (cookie)
				p = sched_core_find(rq_i, cookie);
6000
			if (!p)
6001 6002
				p = idle_sched_class.pick_task(rq_i);
		}
6003

6004
		rq_i->core_pick = p;
6005

6006 6007
		if (p == rq_i->idle) {
			if (rq_i->nr_running) {
6008
				rq->core->core_forceidle_count++;
6009 6010 6011
				if (!fi_before)
					rq->core->core_forceidle_seq++;
			}
6012 6013
		} else {
			occ++;
6014 6015 6016
		}
	}

6017
	if (schedstat_enabled() && rq->core->core_forceidle_count) {
6018
		rq->core->core_forceidle_start = rq_clock(rq->core);
6019 6020 6021
		rq->core->core_forceidle_occupation = occ;
	}

6022 6023 6024 6025 6026 6027 6028 6029 6030 6031 6032 6033 6034 6035 6036 6037
	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) {
6038
		rq_i = cpu_rq(i);
6039 6040 6041 6042 6043 6044 6045 6046 6047 6048 6049

		/*
		 * 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;

6050 6051 6052 6053 6054 6055 6056 6057
		/*
		 * 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
		 */
6058 6059
		if (!(fi_before && rq->core->core_forceidle_count))
			task_vruntime_update(rq_i, rq_i->core_pick, !!rq->core->core_forceidle_count);
6060

6061 6062
		rq_i->core_pick->core_occupation = occ;

6063 6064 6065 6066 6067 6068 6069 6070 6071 6072 6073 6074 6075 6076 6077 6078
		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);
	}

6079
out_set_next:
6080
	set_next_task(rq, next);
6081 6082 6083 6084
out:
	if (rq->core->core_forceidle_count && next == rq->idle)
		queue_core_balance(rq);

6085 6086
	return next;
}
Peter Zijlstra's avatar
Peter Zijlstra committed
6087

6088 6089 6090 6091 6092 6093 6094 6095 6096 6097 6098 6099 6100 6101 6102 6103 6104 6105 6106 6107 6108 6109 6110 6111 6112
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;

6113
		if (!is_cpu_allowed(p, this))
6114 6115 6116 6117 6118 6119 6120 6121 6122 6123 6124 6125 6126 6127 6128 6129 6130 6131 6132 6133 6134 6135 6136 6137 6138 6139 6140 6141 6142 6143 6144 6145 6146 6147 6148 6149 6150 6151 6152 6153 6154 6155 6156 6157 6158 6159 6160 6161 6162 6163 6164 6165 6166 6167 6168 6169 6170 6171 6172 6173 6174 6175 6176 6177 6178
			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);

6179
static void queue_core_balance(struct rq *rq)
6180 6181 6182 6183 6184 6185 6186 6187 6188 6189 6190 6191 6192
{
	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);
}

6193
static void sched_core_cpu_starting(unsigned int cpu)
Peter Zijlstra's avatar
Peter Zijlstra committed
6194 6195
{
	const struct cpumask *smt_mask = cpu_smt_mask(cpu);
6196 6197 6198
	struct rq *rq = cpu_rq(cpu), *core_rq = NULL;
	unsigned long flags;
	int t;
Peter Zijlstra's avatar
Peter Zijlstra committed
6199

6200
	sched_core_lock(cpu, &flags);
Peter Zijlstra's avatar
Peter Zijlstra committed
6201

6202 6203 6204 6205 6206 6207 6208 6209 6210 6211 6212 6213 6214 6215
	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
6216
		}
6217
	}
Peter Zijlstra's avatar
Peter Zijlstra committed
6218

6219 6220
	if (WARN_ON_ONCE(!core_rq)) /* whoopsie */
		goto unlock;
Peter Zijlstra's avatar
Peter Zijlstra committed
6221

6222 6223 6224
	/* install and validate core_rq */
	for_each_cpu(t, smt_mask) {
		rq = cpu_rq(t);
Peter Zijlstra's avatar
Peter Zijlstra committed
6225

6226
		if (t == cpu)
Peter Zijlstra's avatar
Peter Zijlstra committed
6227
			rq->core = core_rq;
6228 6229

		WARN_ON_ONCE(rq->core != core_rq);
Peter Zijlstra's avatar
Peter Zijlstra committed
6230
	}
6231 6232 6233

unlock:
	sched_core_unlock(cpu, &flags);
Peter Zijlstra's avatar
Peter Zijlstra committed
6234
}
6235 6236 6237 6238 6239 6240 6241 6242 6243 6244 6245 6246 6247 6248 6249 6250 6251 6252 6253 6254 6255 6256 6257 6258 6259 6260 6261 6262 6263 6264 6265 6266

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 */
6267 6268 6269 6270 6271 6272 6273 6274 6275 6276 6277 6278 6279
	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;
6280 6281 6282 6283 6284 6285 6286 6287 6288 6289 6290 6291 6292 6293 6294 6295 6296 6297 6298

	/* 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
6299 6300 6301
#else /* !CONFIG_SCHED_CORE */

static inline void sched_core_cpu_starting(unsigned int cpu) {}
6302 6303
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
6304

6305 6306 6307 6308 6309 6310
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
6311 6312
#endif /* CONFIG_SCHED_CORE */

6313 6314 6315 6316 6317 6318 6319 6320 6321 6322
/*
 * 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
6323 6324 6325 6326 6327 6328 6329
#define SM_RTLOCK_WAIT		0x2

#ifndef CONFIG_PREEMPT_RT
# define SM_MASK_PREEMPT	(~0U)
#else
# define SM_MASK_PREEMPT	SM_PREEMPT
#endif
6330

Ingo Molnar's avatar
Ingo Molnar committed
6331
/*
6332
 * __schedule() is the main scheduler function.
6333 6334 6335 6336 6337 6338 6339 6340 6341 6342 6343 6344 6345 6346 6347 6348 6349 6350
 *
 * 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:
 *
6351
 *       - If the kernel is preemptible (CONFIG_PREEMPTION=y):
6352 6353 6354 6355 6356 6357 6358 6359
 *
 *         - 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
 *
6360
 *       - If the kernel is not preemptible (CONFIG_PREEMPTION is not set)
6361 6362 6363 6364 6365 6366
 *         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
6367
 *
6368
 * WARNING: must be called with preemption disabled!
Ingo Molnar's avatar
Ingo Molnar committed
6369
 */
6370
static void __sched notrace __schedule(unsigned int sched_mode)
Ingo Molnar's avatar
Ingo Molnar committed
6371 6372
{
	struct task_struct *prev, *next;
6373
	unsigned long *switch_count;
6374
	unsigned long prev_state;
6375
	struct rq_flags rf;
Ingo Molnar's avatar
Ingo Molnar committed
6376
	struct rq *rq;
6377
	int cpu;
Ingo Molnar's avatar
Ingo Molnar committed
6378 6379 6380 6381 6382

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

6383
	schedule_debug(prev, !!sched_mode);
Linus Torvalds's avatar
Linus Torvalds committed
6384

6385
	if (sched_feat(HRTICK) || sched_feat(HRTICK_DL))
Mike Galbraith's avatar
Mike Galbraith committed
6386
		hrtick_clear(rq);
6387

6388
	local_irq_disable();
6389
	rcu_note_context_switch(!!sched_mode);
6390

6391 6392 6393
	/*
	 * Make sure that signal_pending_state()->signal_pending() below
	 * can't be reordered with __set_current_state(TASK_INTERRUPTIBLE)
6394 6395 6396 6397 6398 6399 6400 6401
	 * 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)
6402
	 *
6403
	 * Also, the membarrier system call requires a full memory barrier
6404
	 * after coming from user-space, before storing to rq->curr.
6405
	 */
6406
	rq_lock(rq, &rf);
6407
	smp_mb__after_spinlock();
Linus Torvalds's avatar
Linus Torvalds committed
6408

6409 6410
	/* Promote REQ to ACT */
	rq->clock_update_flags <<= 1;
6411
	update_rq_clock(rq);
6412

6413
	switch_count = &prev->nivcsw;
6414

6415
	/*
6416
	 * We must load prev->state once (task_struct::state is volatile), such
6417
	 * that we form a control dependency vs deactivate_task() below.
6418
	 */
6419
	prev_state = READ_ONCE(prev->__state);
6420
	if (!(sched_mode & SM_MASK_PREEMPT) && prev_state) {
6421
		if (signal_pending_state(prev_state, prev)) {
6422
			WRITE_ONCE(prev->__state, TASK_RUNNING);
Tejun Heo's avatar
Tejun Heo committed
6423
		} else {
6424 6425 6426
			prev->sched_contributes_to_load =
				(prev_state & TASK_UNINTERRUPTIBLE) &&
				!(prev_state & TASK_NOLOAD) &&
6427
				!(prev_state & TASK_FROZEN);
6428 6429 6430 6431 6432 6433

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

			/*
			 * __schedule()			ttwu()
6434 6435 6436 6437 6438 6439
			 *   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.
6440 6441 6442
			 *
			 * After this, schedule() must not care about p->state any more.
			 */
6443
			deactivate_task(rq, prev, DEQUEUE_SLEEP | DEQUEUE_NOCLOCK);
6444

6445 6446 6447 6448
			if (prev->in_iowait) {
				atomic_inc(&rq->nr_iowait);
				delayacct_blkio_start();
			}
Tejun Heo's avatar
Tejun Heo committed
6449
		}
Ingo Molnar's avatar
Ingo Molnar committed
6450
		switch_count = &prev->nvcsw;
Linus Torvalds's avatar
Linus Torvalds committed
6451 6452
	}

6453
	next = pick_next_task(rq, prev, &rf);
6454
	clear_tsk_need_resched(prev);
6455
	clear_preempt_need_resched();
6456 6457 6458
#ifdef CONFIG_SCHED_DEBUG
	rq->last_seen_need_resched_ns = 0;
#endif
Linus Torvalds's avatar
Linus Torvalds committed
6459 6460 6461

	if (likely(prev != next)) {
		rq->nr_switches++;
6462 6463 6464 6465 6466
		/*
		 * 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);
6467 6468 6469
		/*
		 * The membarrier system call requires each architecture
		 * to have a full memory barrier after updating
6470 6471 6472 6473 6474 6475 6476 6477 6478 6479
		 * 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),
6480
		 */
Linus Torvalds's avatar
Linus Torvalds committed
6481 6482
		++*switch_count;

6483
		migrate_disable_switch(rq, prev);
6484 6485
		psi_sched_switch(prev, next, !task_on_rq_queued(prev));

6486
		trace_sched_switch(sched_mode & SM_MASK_PREEMPT, prev, next, prev_state);
6487 6488 6489

		/* Also unlocks the rq: */
		rq = context_switch(rq, prev, next, &rf);
6490
	} else {
6491
		rq->clock_update_flags &= ~(RQCF_ACT_SKIP|RQCF_REQ_SKIP);
Linus Torvalds's avatar
Linus Torvalds committed
6492

6493 6494
		rq_unpin_lock(rq, &rf);
		__balance_callbacks(rq);
6495
		raw_spin_rq_unlock_irq(rq);
6496
	}
Linus Torvalds's avatar
Linus Torvalds committed
6497
}
6498

6499 6500
void __noreturn do_task_dead(void)
{
6501
	/* Causes final put_task_struct in finish_task_switch(): */
6502
	set_special_state(TASK_DEAD);
6503 6504 6505 6506

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

6507
	__schedule(SM_NONE);
6508
	BUG();
6509 6510

	/* Avoid "noreturn function does return" - but don't continue if BUG() is a NOP: */
6511
	for (;;)
6512
		cpu_relax();
6513 6514
}

6515 6516
static inline void sched_submit_work(struct task_struct *tsk)
{
6517 6518
	unsigned int task_flags;

6519
	if (task_is_running(tsk))
6520
		return;
6521

6522
	task_flags = tsk->flags;
6523
	/*
6524 6525
	 * If a worker goes to sleep, notify and ask workqueue whether it
	 * wants to wake up a task to maintain concurrency.
6526
	 */
6527 6528
	if (task_flags & (PF_WQ_WORKER | PF_IO_WORKER)) {
		if (task_flags & PF_WQ_WORKER)
6529 6530 6531
			wq_worker_sleeping(tsk);
		else
			io_wq_worker_sleeping(tsk);
6532 6533
	}

6534 6535 6536 6537 6538 6539
	/*
	 * spinlock and rwlock must not flush block requests.  This will
	 * deadlock if the callback attempts to acquire a lock which is
	 * already acquired.
	 */
	SCHED_WARN_ON(current->__state & TASK_RTLOCK_WAIT);
6540

6541 6542 6543 6544
	/*
	 * If we are going to sleep and we have plugged IO queued,
	 * make sure to submit it to avoid deadlocks.
	 */
6545
	blk_flush_plug(tsk->plug, true);
6546 6547
}

6548 6549
static void sched_update_worker(struct task_struct *tsk)
{
6550 6551 6552 6553 6554 6555
	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);
	}
6556 6557
}

6558
asmlinkage __visible void __sched schedule(void)
6559
{
6560 6561 6562
	struct task_struct *tsk = current;

	sched_submit_work(tsk);
6563
	do {
6564
		preempt_disable();
6565
		__schedule(SM_NONE);
6566
		sched_preempt_enable_no_resched();
6567
	} while (need_resched());
6568
	sched_update_worker(tsk);
6569
}
Linus Torvalds's avatar
Linus Torvalds committed
6570 6571
EXPORT_SYMBOL(schedule);

6572 6573 6574 6575 6576 6577 6578 6579 6580 6581 6582 6583 6584 6585 6586 6587 6588 6589 6590
/*
 * 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.
	 */
6591
	WARN_ON_ONCE(current->__state);
6592
	do {
6593
		__schedule(SM_NONE);
6594 6595 6596
	} while (need_resched());
}

6597
#if defined(CONFIG_CONTEXT_TRACKING_USER) && !defined(CONFIG_HAVE_CONTEXT_TRACKING_USER_OFFSTACK)
6598
asmlinkage __visible void __sched schedule_user(void)
6599 6600 6601 6602 6603 6604
{
	/*
	 * 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.
6605 6606
	 *
	 * NB: There are buggy callers of this function.  Ideally we
6607
	 * should warn if prev_state != CONTEXT_USER, but that will trigger
6608
	 * too frequently to make sense yet.
6609
	 */
6610
	enum ctx_state prev_state = exception_enter();
6611
	schedule();
6612
	exception_exit(prev_state);
6613 6614 6615
}
#endif

6616 6617 6618 6619 6620 6621 6622
/**
 * schedule_preempt_disabled - called with preemption disabled
 *
 * Returns with preemption disabled. Note: preempt_count must be 1
 */
void __sched schedule_preempt_disabled(void)
{
6623
	sched_preempt_enable_no_resched();
6624 6625 6626 6627
	schedule();
	preempt_disable();
}

6628 6629 6630 6631 6632 6633 6634 6635 6636 6637 6638 6639
#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

6640
static void __sched notrace preempt_schedule_common(void)
6641 6642
{
	do {
6643 6644 6645 6646 6647 6648 6649 6650 6651 6652 6653 6654 6655
		/*
		 * 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.
		 */
6656
		preempt_disable_notrace();
6657
		preempt_latency_start(1);
6658
		__schedule(SM_PREEMPT);
6659
		preempt_latency_stop(1);
6660
		preempt_enable_no_resched_notrace();
6661 6662 6663 6664 6665 6666 6667 6668

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

6669
#ifdef CONFIG_PREEMPTION
Linus Torvalds's avatar
Linus Torvalds committed
6670
/*
6671 6672
 * This is the entry point to schedule() from in-kernel preemption
 * off of preempt_enable.
Linus Torvalds's avatar
Linus Torvalds committed
6673
 */
6674
asmlinkage __visible void __sched notrace preempt_schedule(void)
Linus Torvalds's avatar
Linus Torvalds committed
6675 6676 6677
{
	/*
	 * If there is a non-zero preempt_count or interrupts are disabled,
Ingo Molnar's avatar
Ingo Molnar committed
6678
	 * we do not want to preempt the current task. Just return..
Linus Torvalds's avatar
Linus Torvalds committed
6679
	 */
6680
	if (likely(!preemptible()))
Linus Torvalds's avatar
Linus Torvalds committed
6681
		return;
6682
	preempt_schedule_common();
Linus Torvalds's avatar
Linus Torvalds committed
6683
}
6684
NOKPROBE_SYMBOL(preempt_schedule);
Linus Torvalds's avatar
Linus Torvalds committed
6685
EXPORT_SYMBOL(preempt_schedule);
6686

6687
#ifdef CONFIG_PREEMPT_DYNAMIC
6688
#if defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
6689 6690 6691 6692 6693
#ifndef preempt_schedule_dynamic_enabled
#define preempt_schedule_dynamic_enabled	preempt_schedule
#define preempt_schedule_dynamic_disabled	NULL
#endif
DEFINE_STATIC_CALL(preempt_schedule, preempt_schedule_dynamic_enabled);
6694
EXPORT_STATIC_CALL_TRAMP(preempt_schedule);
6695 6696 6697 6698 6699 6700 6701 6702 6703 6704 6705
#elif defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
static DEFINE_STATIC_KEY_TRUE(sk_dynamic_preempt_schedule);
void __sched notrace dynamic_preempt_schedule(void)
{
	if (!static_branch_unlikely(&sk_dynamic_preempt_schedule))
		return;
	preempt_schedule();
}
NOKPROBE_SYMBOL(dynamic_preempt_schedule);
EXPORT_SYMBOL(dynamic_preempt_schedule);
#endif
6706 6707
#endif

6708
/**
6709
 * preempt_schedule_notrace - preempt_schedule called by tracing
6710 6711 6712 6713 6714 6715 6716 6717 6718 6719 6720 6721
 *
 * 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.
 */
6722
asmlinkage __visible void __sched notrace preempt_schedule_notrace(void)
6723 6724 6725 6726 6727 6728 6729
{
	enum ctx_state prev_ctx;

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

	do {
6730 6731 6732 6733 6734 6735 6736 6737 6738 6739 6740 6741 6742
		/*
		 * 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.
		 */
6743
		preempt_disable_notrace();
6744
		preempt_latency_start(1);
6745 6746 6747 6748 6749 6750
		/*
		 * Needs preempt disabled in case user_exit() is traced
		 * and the tracer calls preempt_enable_notrace() causing
		 * an infinite recursion.
		 */
		prev_ctx = exception_enter();
6751
		__schedule(SM_PREEMPT);
6752 6753
		exception_exit(prev_ctx);

6754
		preempt_latency_stop(1);
6755
		preempt_enable_no_resched_notrace();
6756 6757
	} while (need_resched());
}
6758
EXPORT_SYMBOL_GPL(preempt_schedule_notrace);
6759

6760
#ifdef CONFIG_PREEMPT_DYNAMIC
6761
#if defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
6762 6763 6764
#ifndef preempt_schedule_notrace_dynamic_enabled
#define preempt_schedule_notrace_dynamic_enabled	preempt_schedule_notrace
#define preempt_schedule_notrace_dynamic_disabled	NULL
6765
#endif
6766
DEFINE_STATIC_CALL(preempt_schedule_notrace, preempt_schedule_notrace_dynamic_enabled);
6767
EXPORT_STATIC_CALL_TRAMP(preempt_schedule_notrace);
6768 6769 6770
#elif defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
static DEFINE_STATIC_KEY_TRUE(sk_dynamic_preempt_schedule_notrace);
void __sched notrace dynamic_preempt_schedule_notrace(void)
6771
{
6772 6773 6774
	if (!static_branch_unlikely(&sk_dynamic_preempt_schedule_notrace))
		return;
	preempt_schedule_notrace();
6775
}
6776 6777 6778
NOKPROBE_SYMBOL(dynamic_preempt_schedule_notrace);
EXPORT_SYMBOL(dynamic_preempt_schedule_notrace);
#endif
6779
#endif
6780

6781
#endif /* CONFIG_PREEMPTION */
6782

Linus Torvalds's avatar
Linus Torvalds committed
6783
/*
6784
 * This is the entry point to schedule() from kernel preemption
Linus Torvalds's avatar
Linus Torvalds committed
6785 6786 6787 6788
 * off of irq context.
 * Note, that this is called and return with irqs disabled. This will
 * protect us against recursive calling from irq.
 */
6789
asmlinkage __visible void __sched preempt_schedule_irq(void)
Linus Torvalds's avatar
Linus Torvalds committed
6790
{
6791
	enum ctx_state prev_state;
6792

6793
	/* Catch callers which need to be fixed */
6794
	BUG_ON(preempt_count() || !irqs_disabled());
Linus Torvalds's avatar
Linus Torvalds committed
6795

6796 6797
	prev_state = exception_enter();

6798
	do {
6799
		preempt_disable();
6800
		local_irq_enable();
6801
		__schedule(SM_PREEMPT);
6802
		local_irq_disable();
6803
		sched_preempt_enable_no_resched();
6804
	} while (need_resched());
6805 6806

	exception_exit(prev_state);
Linus Torvalds's avatar
Linus Torvalds committed
6807 6808
}

6809
int default_wake_function(wait_queue_entry_t *curr, unsigned mode, int wake_flags,
Ingo Molnar's avatar
Ingo Molnar committed
6810
			  void *key)
Linus Torvalds's avatar
Linus Torvalds committed
6811
{
6812
	WARN_ON_ONCE(IS_ENABLED(CONFIG_SCHED_DEBUG) && wake_flags & ~WF_SYNC);
6813
	return try_to_wake_up(curr->private, mode, wake_flags);
Linus Torvalds's avatar
Linus Torvalds committed
6814 6815 6816
}
EXPORT_SYMBOL(default_wake_function);

6817 6818 6819 6820 6821 6822 6823 6824 6825 6826 6827 6828
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;
}

6829 6830
#ifdef CONFIG_RT_MUTEXES

6831 6832 6833 6834 6835 6836 6837 6838 6839 6840 6841 6842 6843 6844 6845
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);
}

6846 6847
/*
 * rt_mutex_setprio - set the current priority of a task
6848 6849
 * @p: task to boost
 * @pi_task: donor task
6850 6851 6852 6853
 *
 * This function changes the 'effective' priority of a task. It does
 * not touch ->normal_prio like __setscheduler().
 *
6854 6855
 * Used by the rt_mutex code to implement priority inheritance
 * logic. Call site only calls if the priority of the task changed.
6856
 */
6857
void rt_mutex_setprio(struct task_struct *p, struct task_struct *pi_task)
6858
{
6859
	int prio, oldprio, queued, running, queue_flag =
6860
		DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
6861
	const struct sched_class *prev_class;
6862 6863
	struct rq_flags rf;
	struct rq *rq;
6864

6865 6866 6867 6868 6869 6870 6871 6872
	/* 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;
6873

6874
	rq = __task_rq_lock(p, &rf);
6875
	update_rq_clock(rq);
6876 6877 6878 6879 6880 6881 6882 6883
	/*
	 * 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
6884
	 * points to a blocked task -- which guarantees the task is present.
6885 6886 6887 6888 6889 6890 6891 6892
	 */
	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;
6893

6894 6895 6896 6897 6898 6899 6900 6901 6902 6903 6904 6905 6906 6907 6908 6909 6910 6911
	/*
	 * 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;
	}

6912
	trace_sched_pi_setprio(p, pi_task);
6913
	oldprio = p->prio;
6914 6915 6916 6917

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

6918
	prev_class = p->sched_class;
6919
	queued = task_on_rq_queued(p);
6920
	running = task_current(rq, p);
6921
	if (queued)
6922
		dequeue_task(rq, p, queue_flag);
6923
	if (running)
6924
		put_prev_task(rq, p);
Ingo Molnar's avatar
Ingo Molnar committed
6925

6926 6927 6928 6929 6930 6931 6932 6933 6934 6935
	/*
	 * 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)) {
6936
		if (!dl_prio(p->normal_prio) ||
6937 6938
		    (pi_task && dl_prio(pi_task->prio) &&
		     dl_entity_preempt(&pi_task->dl, &p->dl))) {
6939
			p->dl.pi_se = pi_task->dl.pi_se;
6940
			queue_flag |= ENQUEUE_REPLENISH;
6941 6942 6943
		} else {
			p->dl.pi_se = &p->dl;
		}
6944 6945
	} else if (rt_prio(prio)) {
		if (dl_prio(oldprio))
6946
			p->dl.pi_se = &p->dl;
6947
		if (oldprio < prio)
6948
			queue_flag |= ENQUEUE_HEAD;
6949 6950
	} else {
		if (dl_prio(oldprio))
6951
			p->dl.pi_se = &p->dl;
6952 6953
		if (rt_prio(oldprio))
			p->rt.timeout = 0;
6954
	}
Ingo Molnar's avatar
Ingo Molnar committed
6955

6956
	__setscheduler_prio(p, prio);
6957

6958
	if (queued)
6959
		enqueue_task(rq, p, queue_flag);
6960
	if (running)
6961
		set_next_task(rq, p);
6962

6963
	check_class_changed(rq, p, prev_class, oldprio);
6964
out_unlock:
6965 6966
	/* Avoid rq from going away on us: */
	preempt_disable();
6967

6968 6969
	rq_unpin_lock(rq, &rf);
	__balance_callbacks(rq);
6970
	raw_spin_rq_unlock(rq);
6971

6972
	preempt_enable();
6973
}
6974 6975 6976 6977 6978
#else
static inline int rt_effective_prio(struct task_struct *p, int prio)
{
	return prio;
}
6979
#endif
6980

6981
void set_user_nice(struct task_struct *p, long nice)
Linus Torvalds's avatar
Linus Torvalds committed
6982
{
6983
	bool queued, running;
6984
	int old_prio;
6985
	struct rq_flags rf;
6986
	struct rq *rq;
Linus Torvalds's avatar
Linus Torvalds committed
6987

6988
	if (task_nice(p) == nice || nice < MIN_NICE || nice > MAX_NICE)
Linus Torvalds's avatar
Linus Torvalds committed
6989 6990 6991 6992 6993
		return;
	/*
	 * We have to be careful, if called from sys_setpriority(),
	 * the task might be in the middle of scheduling on another CPU.
	 */
6994
	rq = task_rq_lock(p, &rf);
6995 6996
	update_rq_clock(rq);

Linus Torvalds's avatar
Linus Torvalds committed
6997 6998 6999
	/*
	 * The RT priorities are set via sched_setscheduler(), but we still
	 * allow the 'normal' nice value to be set - but as expected
7000
	 * it won't have any effect on scheduling until the task is
7001
	 * SCHED_DEADLINE, SCHED_FIFO or SCHED_RR:
Linus Torvalds's avatar
Linus Torvalds committed
7002
	 */
7003
	if (task_has_dl_policy(p) || task_has_rt_policy(p)) {
Linus Torvalds's avatar
Linus Torvalds committed
7004 7005 7006
		p->static_prio = NICE_TO_PRIO(nice);
		goto out_unlock;
	}
7007
	queued = task_on_rq_queued(p);
7008
	running = task_current(rq, p);
7009
	if (queued)
7010
		dequeue_task(rq, p, DEQUEUE_SAVE | DEQUEUE_NOCLOCK);
7011 7012
	if (running)
		put_prev_task(rq, p);
Linus Torvalds's avatar
Linus Torvalds committed
7013 7014

	p->static_prio = NICE_TO_PRIO(nice);
7015
	set_load_weight(p, true);
7016 7017
	old_prio = p->prio;
	p->prio = effective_prio(p);
Linus Torvalds's avatar
Linus Torvalds committed
7018

7019
	if (queued)
7020
		enqueue_task(rq, p, ENQUEUE_RESTORE | ENQUEUE_NOCLOCK);
7021
	if (running)
7022
		set_next_task(rq, p);
7023 7024 7025 7026 7027 7028 7029

	/*
	 * 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
7030
out_unlock:
7031
	task_rq_unlock(rq, p, &rf);
Linus Torvalds's avatar
Linus Torvalds committed
7032 7033 7034
}
EXPORT_SYMBOL(set_user_nice);

7035
/*
7036 7037 7038 7039
 * is_nice_reduction - check if nice value is an actual reduction
 *
 * Similar to can_nice() but does not perform a capability check.
 *
7040 7041 7042
 * @p: task
 * @nice: nice value
 */
7043
static bool is_nice_reduction(const struct task_struct *p, const int nice)
7044
{
7045
	/* Convert nice value [19,-20] to rlimit style value [1,40]: */
7046
	int nice_rlim = nice_to_rlimit(nice);
7047

7048 7049 7050 7051 7052 7053 7054 7055 7056 7057 7058
	return (nice_rlim <= task_rlimit(p, RLIMIT_NICE));
}

/*
 * can_nice - check if a task can reduce its nice value
 * @p: task
 * @nice: nice value
 */
int can_nice(const struct task_struct *p, const int nice)
{
	return is_nice_reduction(p, nice) || capable(CAP_SYS_NICE);
7059 7060
}

Linus Torvalds's avatar
Linus Torvalds committed
7061 7062 7063 7064 7065 7066 7067 7068 7069
#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.
 */
7070
SYSCALL_DEFINE1(nice, int, increment)
Linus Torvalds's avatar
Linus Torvalds committed
7071
{
7072
	long nice, retval;
Linus Torvalds's avatar
Linus Torvalds committed
7073 7074 7075 7076 7077 7078

	/*
	 * 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.
	 */
7079
	increment = clamp(increment, -NICE_WIDTH, NICE_WIDTH);
7080
	nice = task_nice(current) + increment;
Linus Torvalds's avatar
Linus Torvalds committed
7081

7082
	nice = clamp_val(nice, MIN_NICE, MAX_NICE);
7083 7084 7085
	if (increment < 0 && !can_nice(current, nice))
		return -EPERM;

Linus Torvalds's avatar
Linus Torvalds committed
7086 7087 7088 7089 7090 7091 7092 7093 7094 7095 7096 7097 7098 7099
	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.
 *
7100
 * Return: The priority value as seen by users in /proc.
7101 7102 7103 7104 7105 7106
 *
 * 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
7107
 */
7108
int task_prio(const struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
7109 7110 7111 7112 7113
{
	return p->prio - MAX_RT_PRIO;
}

/**
7114
 * idle_cpu - is a given CPU idle currently?
Linus Torvalds's avatar
Linus Torvalds committed
7115
 * @cpu: the processor in question.
7116 7117
 *
 * Return: 1 if the CPU is currently idle. 0 otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
7118 7119 7120
 */
int idle_cpu(int cpu)
{
Thomas Gleixner's avatar
Thomas Gleixner committed
7121 7122 7123 7124 7125 7126 7127 7128 7129
	struct rq *rq = cpu_rq(cpu);

	if (rq->curr != rq->idle)
		return 0;

	if (rq->nr_running)
		return 0;

#ifdef CONFIG_SMP
7130
	if (rq->ttwu_pending)
Thomas Gleixner's avatar
Thomas Gleixner committed
7131 7132 7133 7134
		return 0;
#endif

	return 1;
Linus Torvalds's avatar
Linus Torvalds committed
7135 7136
}

7137 7138 7139 7140 7141 7142 7143 7144 7145 7146 7147
/**
 * 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;

7148 7149 7150
	if (vcpu_is_preempted(cpu))
		return 0;

Thomas Gleixner's avatar
Thomas Gleixner committed
7151
	return 1;
Linus Torvalds's avatar
Linus Torvalds committed
7152 7153 7154
}

/**
7155
 * idle_task - return the idle task for a given CPU.
Linus Torvalds's avatar
Linus Torvalds committed
7156
 * @cpu: the processor in question.
7157
 *
7158
 * Return: The idle task for the CPU @cpu.
Linus Torvalds's avatar
Linus Torvalds committed
7159
 */
7160
struct task_struct *idle_task(int cpu)
Linus Torvalds's avatar
Linus Torvalds committed
7161 7162 7163 7164
{
	return cpu_rq(cpu)->idle;
}

7165 7166 7167 7168 7169 7170 7171 7172 7173 7174 7175 7176 7177 7178 7179 7180 7181 7182 7183 7184 7185
#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.
 */
7186
unsigned long effective_cpu_util(int cpu, unsigned long util_cfs,
7187
				 enum cpu_util_type type,
7188 7189
				 struct task_struct *p)
{
7190
	unsigned long dl_util, util, irq, max;
7191 7192
	struct rq *rq = cpu_rq(cpu);

7193 7194
	max = arch_scale_cpu_capacity(cpu);

7195 7196 7197 7198 7199 7200 7201 7202 7203 7204 7205 7206 7207 7208 7209 7210 7211 7212 7213 7214 7215 7216 7217 7218 7219 7220 7221 7222 7223 7224 7225 7226 7227 7228 7229 7230 7231 7232 7233 7234 7235 7236 7237 7238 7239 7240 7241 7242 7243 7244 7245 7246 7247 7248 7249 7250 7251 7252 7253 7254 7255 7256 7257 7258 7259 7260 7261 7262 7263 7264 7265 7266 7267 7268 7269 7270 7271 7272
	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);
}
7273

7274
unsigned long sched_cpu_util(int cpu)
7275
{
7276
	return effective_cpu_util(cpu, cpu_util_cfs(cpu), ENERGY_UTIL, NULL);
7277
}
7278 7279
#endif /* CONFIG_SMP */

Linus Torvalds's avatar
Linus Torvalds committed
7280 7281 7282
/**
 * find_process_by_pid - find a process with a matching PID value.
 * @pid: the pid in question.
7283 7284
 *
 * The task of @pid, if found. %NULL otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
7285
 */
Alexey Dobriyan's avatar
Alexey Dobriyan committed
7286
static struct task_struct *find_process_by_pid(pid_t pid)
Linus Torvalds's avatar
Linus Torvalds committed
7287
{
7288
	return pid ? find_task_by_vpid(pid) : current;
Linus Torvalds's avatar
Linus Torvalds committed
7289 7290
}

7291 7292 7293 7294 7295 7296
/*
 * sched_setparam() passes in -1 for its policy, to let the functions
 * it calls know not to change it.
 */
#define SETPARAM_POLICY	-1

7297 7298
static void __setscheduler_params(struct task_struct *p,
		const struct sched_attr *attr)
Linus Torvalds's avatar
Linus Torvalds committed
7299
{
7300 7301
	int policy = attr->sched_policy;

7302
	if (policy == SETPARAM_POLICY)
7303 7304
		policy = p->policy;

Linus Torvalds's avatar
Linus Torvalds committed
7305
	p->policy = policy;
7306

7307 7308
	if (dl_policy(policy))
		__setparam_dl(p, attr);
7309
	else if (fair_policy(policy))
7310 7311
		p->static_prio = NICE_TO_PRIO(attr->sched_nice);

7312 7313 7314 7315 7316 7317
	/*
	 * __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;
7318
	p->normal_prio = normal_prio(p);
7319
	set_load_weight(p, true);
7320
}
7321

7322
/*
7323
 * Check the target process has a UID that matches the current process's:
7324 7325 7326 7327 7328 7329 7330 7331
 */
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);
7332 7333
	match = (uid_eq(cred->euid, pcred->euid) ||
		 uid_eq(cred->euid, pcred->uid));
7334 7335 7336 7337
	rcu_read_unlock();
	return match;
}

7338 7339 7340 7341 7342 7343 7344 7345 7346 7347 7348 7349 7350 7351 7352 7353 7354 7355 7356 7357 7358 7359 7360 7361 7362 7363 7364 7365 7366 7367 7368 7369 7370 7371 7372 7373 7374 7375 7376 7377 7378 7379 7380 7381 7382 7383 7384 7385 7386 7387 7388 7389 7390 7391 7392 7393 7394 7395 7396 7397 7398 7399 7400
/*
 * Allow unprivileged RT tasks to decrease priority.
 * Only issue a capable test if needed and only once to avoid an audit
 * event on permitted non-privileged operations:
 */
static int user_check_sched_setscheduler(struct task_struct *p,
					 const struct sched_attr *attr,
					 int policy, int reset_on_fork)
{
	if (fair_policy(policy)) {
		if (attr->sched_nice < task_nice(p) &&
		    !is_nice_reduction(p, attr->sched_nice))
			goto req_priv;
	}

	if (rt_policy(policy)) {
		unsigned long rlim_rtprio = task_rlimit(p, RLIMIT_RTPRIO);

		/* Can't set/change the rt policy: */
		if (policy != p->policy && !rlim_rtprio)
			goto req_priv;

		/* Can't increase priority: */
		if (attr->sched_priority > p->rt_priority &&
		    attr->sched_priority > rlim_rtprio)
			goto req_priv;
	}

	/*
	 * 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))
		goto req_priv;

	/*
	 * Treat SCHED_IDLE as nice 20. Only allow a switch to
	 * SCHED_NORMAL if the RLIMIT_NICE would normally permit it.
	 */
	if (task_has_idle_policy(p) && !idle_policy(policy)) {
		if (!is_nice_reduction(p, task_nice(p)))
			goto req_priv;
	}

	/* Can't change other user's priorities: */
	if (!check_same_owner(p))
		goto req_priv;

	/* Normal users shall not reset the sched_reset_on_fork flag: */
	if (p->sched_reset_on_fork && !reset_on_fork)
		goto req_priv;

	return 0;

req_priv:
	if (!capable(CAP_SYS_NICE))
		return -EPERM;

	return 0;
}

7401 7402
static int __sched_setscheduler(struct task_struct *p,
				const struct sched_attr *attr,
7403
				bool user, bool pi)
Linus Torvalds's avatar
Linus Torvalds committed
7404
{
7405 7406
	int oldpolicy = -1, policy = attr->sched_policy;
	int retval, oldprio, newprio, queued, running;
7407
	const struct sched_class *prev_class;
7408
	struct callback_head *head;
7409
	struct rq_flags rf;
7410
	int reset_on_fork;
7411
	int queue_flags = DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
7412
	struct rq *rq;
Linus Torvalds's avatar
Linus Torvalds committed
7413

7414 7415
	/* The pi code expects interrupts enabled */
	BUG_ON(pi && in_interrupt());
Linus Torvalds's avatar
Linus Torvalds committed
7416
recheck:
7417
	/* Double check policy once rq lock held: */
7418 7419
	if (policy < 0) {
		reset_on_fork = p->sched_reset_on_fork;
Linus Torvalds's avatar
Linus Torvalds committed
7420
		policy = oldpolicy = p->policy;
7421
	} else {
7422
		reset_on_fork = !!(attr->sched_flags & SCHED_FLAG_RESET_ON_FORK);
7423

7424
		if (!valid_policy(policy))
7425 7426 7427
			return -EINVAL;
	}

7428
	if (attr->sched_flags & ~(SCHED_FLAG_ALL | SCHED_FLAG_SUGOV))
7429 7430
		return -EINVAL;

Linus Torvalds's avatar
Linus Torvalds committed
7431 7432
	/*
	 * Valid priorities for SCHED_FIFO and SCHED_RR are
7433
	 * 1..MAX_RT_PRIO-1, valid priority for SCHED_NORMAL,
Ingo Molnar's avatar
Ingo Molnar committed
7434
	 * SCHED_BATCH and SCHED_IDLE is 0.
Linus Torvalds's avatar
Linus Torvalds committed
7435
	 */
7436
	if (attr->sched_priority > MAX_RT_PRIO-1)
Linus Torvalds's avatar
Linus Torvalds committed
7437
		return -EINVAL;
7438 7439
	if ((dl_policy(policy) && !__checkparam_dl(attr)) ||
	    (rt_policy(policy) != (attr->sched_priority != 0)))
Linus Torvalds's avatar
Linus Torvalds committed
7440 7441
		return -EINVAL;

7442
	if (user) {
7443 7444 7445 7446
		retval = user_check_sched_setscheduler(p, attr, policy, reset_on_fork);
		if (retval)
			return retval;

7447 7448 7449
		if (attr->sched_flags & SCHED_FLAG_SUGOV)
			return -EINVAL;

7450
		retval = security_task_setscheduler(p);
7451 7452 7453 7454
		if (retval)
			return retval;
	}

7455 7456 7457 7458 7459 7460 7461
	/* Update task specific "requested" clamps */
	if (attr->sched_flags & SCHED_FLAG_UTIL_CLAMP) {
		retval = uclamp_validate(p, attr);
		if (retval)
			return retval;
	}

7462 7463 7464
	if (pi)
		cpuset_read_lock();

7465
	/*
7466
	 * Make sure no PI-waiters arrive (or leave) while we are
7467
	 * changing the priority of the task:
7468
	 *
Lucas De Marchi's avatar
Lucas De Marchi committed
7469
	 * To be able to change p->policy safely, the appropriate
Linus Torvalds's avatar
Linus Torvalds committed
7470 7471
	 * runqueue lock must be held.
	 */
7472
	rq = task_rq_lock(p, &rf);
7473
	update_rq_clock(rq);
7474

7475
	/*
7476
	 * Changing the policy of the stop threads its a very bad idea:
7477 7478
	 */
	if (p == rq->stop) {
7479 7480
		retval = -EINVAL;
		goto unlock;
7481 7482
	}

7483
	/*
7484 7485
	 * If not changing anything there's no need to proceed further,
	 * but store a possible modification of reset_on_fork.
7486
	 */
7487
	if (unlikely(policy == p->policy)) {
7488
		if (fair_policy(policy) && attr->sched_nice != task_nice(p))
7489 7490 7491
			goto change;
		if (rt_policy(policy) && attr->sched_priority != p->rt_priority)
			goto change;
7492
		if (dl_policy(policy) && dl_param_changed(p, attr))
7493
			goto change;
7494 7495
		if (attr->sched_flags & SCHED_FLAG_UTIL_CLAMP)
			goto change;
7496

7497
		p->sched_reset_on_fork = reset_on_fork;
7498 7499
		retval = 0;
		goto unlock;
7500
	}
7501
change:
7502

7503
	if (user) {
7504
#ifdef CONFIG_RT_GROUP_SCHED
7505 7506 7507 7508 7509
		/*
		 * Do not allow realtime tasks into groups that have no runtime
		 * assigned.
		 */
		if (rt_bandwidth_enabled() && rt_policy(policy) &&
7510 7511
				task_group(p)->rt_bandwidth.rt_runtime == 0 &&
				!task_group_is_autogroup(task_group(p))) {
7512 7513
			retval = -EPERM;
			goto unlock;
7514 7515
		}
#endif
7516
#ifdef CONFIG_SMP
7517 7518
		if (dl_bandwidth_enabled() && dl_policy(policy) &&
				!(attr->sched_flags & SCHED_FLAG_SUGOV)) {
7519 7520 7521 7522 7523 7524 7525
			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.
			 */
7526
			if (!cpumask_subset(span, p->cpus_ptr) ||
7527
			    rq->rd->dl_bw.bw == 0) {
7528 7529
				retval = -EPERM;
				goto unlock;
7530 7531 7532 7533
			}
		}
#endif
	}
7534

7535
	/* Re-check policy now with rq lock held: */
Linus Torvalds's avatar
Linus Torvalds committed
7536 7537
	if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) {
		policy = oldpolicy = -1;
7538
		task_rq_unlock(rq, p, &rf);
7539 7540
		if (pi)
			cpuset_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
7541 7542
		goto recheck;
	}
7543 7544 7545 7546 7547 7548

	/*
	 * If setscheduling to SCHED_DEADLINE (or changing the parameters
	 * of a SCHED_DEADLINE task) we need to check if enough bandwidth
	 * is available.
	 */
7549
	if ((dl_policy(policy) || dl_task(p)) && sched_dl_overflow(p, policy, attr)) {
7550 7551
		retval = -EBUSY;
		goto unlock;
7552 7553
	}

7554 7555 7556
	p->sched_reset_on_fork = reset_on_fork;
	oldprio = p->prio;

7557
	newprio = __normal_prio(policy, attr->sched_priority, attr->sched_nice);
7558 7559 7560 7561 7562 7563 7564 7565
	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.
		 */
7566 7567
		newprio = rt_effective_prio(p, newprio);
		if (newprio == oldprio)
7568
			queue_flags &= ~DEQUEUE_MOVE;
7569 7570
	}

7571
	queued = task_on_rq_queued(p);
7572
	running = task_current(rq, p);
7573
	if (queued)
7574
		dequeue_task(rq, p, queue_flags);
7575
	if (running)
7576
		put_prev_task(rq, p);
7577

7578
	prev_class = p->sched_class;
7579

7580 7581 7582 7583
	if (!(attr->sched_flags & SCHED_FLAG_KEEP_PARAMS)) {
		__setscheduler_params(p, attr);
		__setscheduler_prio(p, newprio);
	}
7584
	__setscheduler_uclamp(p, attr);
7585

7586
	if (queued) {
7587 7588 7589 7590
		/*
		 * We enqueue to tail when the priority of a task is
		 * increased (user space view).
		 */
7591 7592
		if (oldprio < p->prio)
			queue_flags |= ENQUEUE_HEAD;
7593

7594
		enqueue_task(rq, p, queue_flags);
7595
	}
7596
	if (running)
7597
		set_next_task(rq, p);
7598

7599
	check_class_changed(rq, p, prev_class, oldprio);
7600 7601 7602

	/* Avoid rq from going away on us: */
	preempt_disable();
7603
	head = splice_balance_callbacks(rq);
7604
	task_rq_unlock(rq, p, &rf);
7605

7606 7607
	if (pi) {
		cpuset_read_unlock();
7608
		rt_mutex_adjust_pi(p);
7609
	}
7610

7611
	/* Run balance callbacks after we've adjusted the PI chain: */
7612
	balance_callbacks(rq, head);
7613
	preempt_enable();
7614

Linus Torvalds's avatar
Linus Torvalds committed
7615
	return 0;
7616 7617 7618

unlock:
	task_rq_unlock(rq, p, &rf);
7619 7620
	if (pi)
		cpuset_read_unlock();
7621
	return retval;
Linus Torvalds's avatar
Linus Torvalds committed
7622
}
7623

7624 7625 7626 7627 7628 7629 7630 7631 7632
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),
	};

7633 7634
	/* Fixup the legacy SCHED_RESET_ON_FORK hack. */
	if ((policy != SETPARAM_POLICY) && (policy & SCHED_RESET_ON_FORK)) {
7635 7636 7637 7638 7639
		attr.sched_flags |= SCHED_FLAG_RESET_ON_FORK;
		policy &= ~SCHED_RESET_ON_FORK;
		attr.sched_policy = policy;
	}

7640
	return __sched_setscheduler(p, &attr, check, true);
7641
}
7642 7643 7644 7645 7646 7647
/**
 * 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.
 *
7648 7649
 * Use sched_set_fifo(), read its comment.
 *
7650 7651
 * Return: 0 on success. An error code otherwise.
 *
7652 7653 7654
 * NOTE that the task may be already dead.
 */
int sched_setscheduler(struct task_struct *p, int policy,
7655
		       const struct sched_param *param)
7656
{
7657
	return _sched_setscheduler(p, policy, param, true);
7658
}
Linus Torvalds's avatar
Linus Torvalds committed
7659

7660 7661
int sched_setattr(struct task_struct *p, const struct sched_attr *attr)
{
7662
	return __sched_setscheduler(p, attr, true, true);
7663 7664
}

7665 7666 7667 7668
int sched_setattr_nocheck(struct task_struct *p, const struct sched_attr *attr)
{
	return __sched_setscheduler(p, attr, false, true);
}
7669
EXPORT_SYMBOL_GPL(sched_setattr_nocheck);
7670

7671 7672 7673 7674 7675 7676 7677 7678 7679 7680
/**
 * 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.
7681 7682
 *
 * Return: 0 on success. An error code otherwise.
7683 7684
 */
int sched_setscheduler_nocheck(struct task_struct *p, int policy,
7685
			       const struct sched_param *param)
7686
{
7687
	return _sched_setscheduler(p, policy, param, false);
7688 7689
}

7690 7691 7692 7693 7694 7695 7696 7697 7698 7699 7700 7701 7702 7703 7704 7705 7706 7707
/*
 * 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.
 */
7708
void sched_set_fifo(struct task_struct *p)
7709 7710
{
	struct sched_param sp = { .sched_priority = MAX_RT_PRIO / 2 };
7711
	WARN_ON_ONCE(sched_setscheduler_nocheck(p, SCHED_FIFO, &sp) != 0);
7712 7713 7714 7715 7716 7717
}
EXPORT_SYMBOL_GPL(sched_set_fifo);

/*
 * For when you don't much care about FIFO, but want to be above SCHED_NORMAL.
 */
7718
void sched_set_fifo_low(struct task_struct *p)
7719 7720
{
	struct sched_param sp = { .sched_priority = 1 };
7721
	WARN_ON_ONCE(sched_setscheduler_nocheck(p, SCHED_FIFO, &sp) != 0);
7722 7723 7724
}
EXPORT_SYMBOL_GPL(sched_set_fifo_low);

7725
void sched_set_normal(struct task_struct *p, int nice)
7726 7727 7728 7729 7730
{
	struct sched_attr attr = {
		.sched_policy = SCHED_NORMAL,
		.sched_nice = nice,
	};
7731
	WARN_ON_ONCE(sched_setattr_nocheck(p, &attr) != 0);
7732 7733
}
EXPORT_SYMBOL_GPL(sched_set_normal);
7734

Ingo Molnar's avatar
Ingo Molnar committed
7735 7736
static int
do_sched_setscheduler(pid_t pid, int policy, struct sched_param __user *param)
Linus Torvalds's avatar
Linus Torvalds committed
7737 7738 7739
{
	struct sched_param lparam;
	struct task_struct *p;
7740
	int retval;
Linus Torvalds's avatar
Linus Torvalds committed
7741 7742 7743 7744 7745

	if (!param || pid < 0)
		return -EINVAL;
	if (copy_from_user(&lparam, param, sizeof(struct sched_param)))
		return -EFAULT;
7746 7747 7748

	rcu_read_lock();
	retval = -ESRCH;
Linus Torvalds's avatar
Linus Torvalds committed
7749
	p = find_process_by_pid(pid);
7750 7751
	if (likely(p))
		get_task_struct(p);
7752
	rcu_read_unlock();
7753

7754 7755 7756 7757 7758
	if (likely(p)) {
		retval = sched_setscheduler(p, policy, &lparam);
		put_task_struct(p);
	}

Linus Torvalds's avatar
Linus Torvalds committed
7759 7760 7761
	return retval;
}

7762 7763 7764
/*
 * Mimics kernel/events/core.c perf_copy_attr().
 */
7765
static int sched_copy_attr(struct sched_attr __user *uattr, struct sched_attr *attr)
7766 7767 7768 7769
{
	u32 size;
	int ret;

7770
	/* Zero the full structure, so that a short copy will be nice: */
7771 7772 7773 7774 7775 7776
	memset(attr, 0, sizeof(*attr));

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

7777 7778
	/* ABI compatibility quirk: */
	if (!size)
7779
		size = SCHED_ATTR_SIZE_VER0;
7780
	if (size < SCHED_ATTR_SIZE_VER0 || size > PAGE_SIZE)
7781 7782
		goto err_size;

7783 7784 7785 7786 7787
	ret = copy_struct_from_user(attr, sizeof(*attr), uattr, size);
	if (ret) {
		if (ret == -E2BIG)
			goto err_size;
		return ret;
7788 7789
	}

7790 7791 7792 7793
	if ((attr->sched_flags & SCHED_FLAG_UTIL_CLAMP) &&
	    size < SCHED_ATTR_SIZE_VER1)
		return -EINVAL;

7794
	/*
7795
	 * XXX: Do we want to be lenient like existing syscalls; or do we want
7796 7797
	 * to be strict and return an error on out-of-bounds values?
	 */
7798
	attr->sched_nice = clamp(attr->sched_nice, MIN_NICE, MAX_NICE);
7799

7800
	return 0;
7801 7802 7803

err_size:
	put_user(sizeof(*attr), &uattr->size);
7804
	return -E2BIG;
7805 7806
}

7807 7808 7809 7810 7811 7812 7813 7814 7815 7816
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
7817 7818 7819 7820 7821
/**
 * 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.
7822 7823
 *
 * Return: 0 on success. An error code otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
7824
 */
7825
SYSCALL_DEFINE3(sched_setscheduler, pid_t, pid, int, policy, struct sched_param __user *, param)
Linus Torvalds's avatar
Linus Torvalds committed
7826
{
7827 7828 7829
	if (policy < 0)
		return -EINVAL;

Linus Torvalds's avatar
Linus Torvalds committed
7830 7831 7832 7833 7834 7835 7836
	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.
7837 7838
 *
 * Return: 0 on success. An error code otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
7839
 */
7840
SYSCALL_DEFINE2(sched_setparam, pid_t, pid, struct sched_param __user *, param)
Linus Torvalds's avatar
Linus Torvalds committed
7841
{
7842
	return do_sched_setscheduler(pid, SETPARAM_POLICY, param);
Linus Torvalds's avatar
Linus Torvalds committed
7843 7844
}

7845 7846 7847
/**
 * sys_sched_setattr - same as above, but with extended sched_attr
 * @pid: the pid in question.
7848
 * @uattr: structure containing the extended parameters.
7849
 * @flags: for future extension.
7850
 */
7851 7852
SYSCALL_DEFINE3(sched_setattr, pid_t, pid, struct sched_attr __user *, uattr,
			       unsigned int, flags)
7853 7854 7855 7856 7857
{
	struct sched_attr attr;
	struct task_struct *p;
	int retval;

7858
	if (!uattr || pid < 0 || flags)
7859 7860
		return -EINVAL;

7861 7862 7863
	retval = sched_copy_attr(uattr, &attr);
	if (retval)
		return retval;
7864

7865
	if ((int)attr.sched_policy < 0)
7866
		return -EINVAL;
7867 7868
	if (attr.sched_flags & SCHED_FLAG_KEEP_POLICY)
		attr.sched_policy = SETPARAM_POLICY;
7869 7870 7871 7872

	rcu_read_lock();
	retval = -ESRCH;
	p = find_process_by_pid(pid);
7873 7874
	if (likely(p))
		get_task_struct(p);
7875 7876
	rcu_read_unlock();

7877
	if (likely(p)) {
7878 7879
		if (attr.sched_flags & SCHED_FLAG_KEEP_PARAMS)
			get_params(p, &attr);
7880 7881 7882 7883
		retval = sched_setattr(p, &attr);
		put_task_struct(p);
	}

7884 7885 7886
	return retval;
}

Linus Torvalds's avatar
Linus Torvalds committed
7887 7888 7889
/**
 * sys_sched_getscheduler - get the policy (scheduling class) of a thread
 * @pid: the pid in question.
7890 7891 7892
 *
 * Return: On success, the policy of the thread. Otherwise, a negative error
 * code.
Linus Torvalds's avatar
Linus Torvalds committed
7893
 */
7894
SYSCALL_DEFINE1(sched_getscheduler, pid_t, pid)
Linus Torvalds's avatar
Linus Torvalds committed
7895
{
7896
	struct task_struct *p;
7897
	int retval;
Linus Torvalds's avatar
Linus Torvalds committed
7898 7899

	if (pid < 0)
7900
		return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
7901 7902

	retval = -ESRCH;
7903
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
7904 7905 7906 7907
	p = find_process_by_pid(pid);
	if (p) {
		retval = security_task_getscheduler(p);
		if (!retval)
7908 7909
			retval = p->policy
				| (p->sched_reset_on_fork ? SCHED_RESET_ON_FORK : 0);
Linus Torvalds's avatar
Linus Torvalds committed
7910
	}
7911
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
7912 7913 7914 7915
	return retval;
}

/**
7916
 * sys_sched_getparam - get the RT priority of a thread
Linus Torvalds's avatar
Linus Torvalds committed
7917 7918
 * @pid: the pid in question.
 * @param: structure containing the RT priority.
7919 7920 7921
 *
 * Return: On success, 0 and the RT priority is in @param. Otherwise, an error
 * code.
Linus Torvalds's avatar
Linus Torvalds committed
7922
 */
7923
SYSCALL_DEFINE2(sched_getparam, pid_t, pid, struct sched_param __user *, param)
Linus Torvalds's avatar
Linus Torvalds committed
7924
{
7925
	struct sched_param lp = { .sched_priority = 0 };
7926
	struct task_struct *p;
7927
	int retval;
Linus Torvalds's avatar
Linus Torvalds committed
7928 7929

	if (!param || pid < 0)
7930
		return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
7931

7932
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
7933 7934 7935 7936 7937 7938 7939 7940 7941
	p = find_process_by_pid(pid);
	retval = -ESRCH;
	if (!p)
		goto out_unlock;

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

7942 7943
	if (task_has_rt_policy(p))
		lp.sched_priority = p->rt_priority;
7944
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
7945 7946 7947 7948 7949 7950 7951 7952 7953

	/*
	 * 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:
7954
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
7955 7956 7957
	return retval;
}

7958 7959 7960 7961 7962 7963 7964 7965 7966 7967 7968 7969
/*
 * 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)
7970
{
7971
	unsigned int ksize = sizeof(*kattr);
7972

7973
	if (!access_ok(uattr, usize))
7974 7975 7976
		return -EFAULT;

	/*
7977 7978 7979 7980 7981 7982 7983 7984 7985 7986 7987
	 * 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.
7988
	 */
7989
	kattr->size = min(usize, ksize);
7990

7991
	if (copy_to_user(uattr, kattr, kattr->size))
7992 7993
		return -EFAULT;

7994
	return 0;
7995 7996 7997
}

/**
7998
 * sys_sched_getattr - similar to sched_getparam, but with sched_attr
7999
 * @pid: the pid in question.
8000
 * @uattr: structure containing the extended parameters.
8001
 * @usize: sizeof(attr) for fwd/bwd comp.
8002
 * @flags: for future extension.
8003
 */
8004
SYSCALL_DEFINE4(sched_getattr, pid_t, pid, struct sched_attr __user *, uattr,
8005
		unsigned int, usize, unsigned int, flags)
8006
{
8007
	struct sched_attr kattr = { };
8008 8009 8010
	struct task_struct *p;
	int retval;

8011 8012
	if (!uattr || pid < 0 || usize > PAGE_SIZE ||
	    usize < SCHED_ATTR_SIZE_VER0 || flags)
8013 8014 8015 8016 8017 8018 8019 8020 8021 8022 8023 8024
		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;

8025
	kattr.sched_policy = p->policy;
8026
	if (p->sched_reset_on_fork)
8027
		kattr.sched_flags |= SCHED_FLAG_RESET_ON_FORK;
8028
	get_params(p, &kattr);
8029
	kattr.sched_flags &= SCHED_FLAG_ALL;
8030

8031
#ifdef CONFIG_UCLAMP_TASK
8032 8033 8034 8035 8036
	/*
	 * 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.
	 */
8037 8038
	kattr.sched_util_min = p->uclamp_req[UCLAMP_MIN].value;
	kattr.sched_util_max = p->uclamp_req[UCLAMP_MAX].value;
8039 8040
#endif

8041 8042
	rcu_read_unlock();

8043
	return sched_attr_copy_to_user(uattr, &kattr, usize);
8044 8045 8046 8047 8048 8049

out_unlock:
	rcu_read_unlock();
	return retval;
}

8050 8051
#ifdef CONFIG_SMP
int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
Linus Torvalds's avatar
Linus Torvalds committed
8052
{
8053 8054 8055 8056 8057 8058 8059 8060 8061 8062 8063 8064 8065 8066 8067 8068 8069 8070 8071 8072 8073 8074 8075
	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

8076 8077
static int
__sched_setaffinity(struct task_struct *p, const struct cpumask *mask)
Linus Torvalds's avatar
Linus Torvalds committed
8078
{
8079
	int retval;
8080
	cpumask_var_t cpus_allowed, new_mask;
Linus Torvalds's avatar
Linus Torvalds committed
8081

8082 8083
	if (!alloc_cpumask_var(&cpus_allowed, GFP_KERNEL))
		return -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
8084

8085 8086 8087 8088
	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL)) {
		retval = -ENOMEM;
		goto out_free_cpus_allowed;
	}
8089 8090

	cpuset_cpus_allowed(p, cpus_allowed);
8091
	cpumask_and(new_mask, mask, cpus_allowed);
8092

8093 8094 8095
	retval = dl_task_check_affinity(p, new_mask);
	if (retval)
		goto out_free_new_mask;
Peter Zijlstra's avatar
Peter Zijlstra committed
8096
again:
8097
	retval = __set_cpus_allowed_ptr(p, new_mask, SCA_CHECK | SCA_USER);
8098 8099
	if (retval)
		goto out_free_new_mask;
Linus Torvalds's avatar
Linus Torvalds committed
8100

8101 8102 8103 8104 8105 8106 8107 8108
	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
8109
	}
8110

8111
out_free_new_mask:
8112 8113 8114
	free_cpumask_var(new_mask);
out_free_cpus_allowed:
	free_cpumask_var(cpus_allowed);
8115 8116 8117 8118 8119
	return retval;
}

long sched_setaffinity(pid_t pid, const struct cpumask *in_mask)
{
8120 8121
	struct task_struct *p;
	int retval;
Linus Torvalds's avatar
Linus Torvalds committed
8122

8123
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
8124 8125 8126

	p = find_process_by_pid(pid);
	if (!p) {
8127
		rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
8128 8129 8130
		return -ESRCH;
	}

8131
	/* Prevent p going away */
Linus Torvalds's avatar
Linus Torvalds committed
8132
	get_task_struct(p);
8133
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
8134

8135 8136 8137 8138
	if (p->flags & PF_NO_SETAFFINITY) {
		retval = -EINVAL;
		goto out_put_task;
	}
8139

8140 8141 8142 8143
	if (!check_same_owner(p)) {
		rcu_read_lock();
		if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
			rcu_read_unlock();
8144 8145
			retval = -EPERM;
			goto out_put_task;
8146 8147 8148
		}
		rcu_read_unlock();
	}
Linus Torvalds's avatar
Linus Torvalds committed
8149

8150
	retval = security_task_setscheduler(p);
8151
	if (retval)
8152
		goto out_put_task;
Linus Torvalds's avatar
Linus Torvalds committed
8153

8154
	retval = __sched_setaffinity(p, in_mask);
8155
out_put_task:
Linus Torvalds's avatar
Linus Torvalds committed
8156 8157 8158 8159 8160
	put_task_struct(p);
	return retval;
}

static int get_user_cpu_mask(unsigned long __user *user_mask_ptr, unsigned len,
8161
			     struct cpumask *new_mask)
Linus Torvalds's avatar
Linus Torvalds committed
8162
{
8163 8164 8165 8166 8167
	if (len < cpumask_size())
		cpumask_clear(new_mask);
	else if (len > cpumask_size())
		len = cpumask_size();

Linus Torvalds's avatar
Linus Torvalds committed
8168 8169 8170 8171
	return copy_from_user(new_mask, user_mask_ptr, len) ? -EFAULT : 0;
}

/**
8172
 * sys_sched_setaffinity - set the CPU affinity of a process
Linus Torvalds's avatar
Linus Torvalds committed
8173 8174
 * @pid: pid of the process
 * @len: length in bytes of the bitmask pointed to by user_mask_ptr
8175
 * @user_mask_ptr: user-space pointer to the new CPU mask
8176 8177
 *
 * Return: 0 on success. An error code otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
8178
 */
8179 8180
SYSCALL_DEFINE3(sched_setaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
Linus Torvalds's avatar
Linus Torvalds committed
8181
{
8182
	cpumask_var_t new_mask;
Linus Torvalds's avatar
Linus Torvalds committed
8183 8184
	int retval;

8185 8186
	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL))
		return -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
8187

8188 8189 8190 8191 8192
	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
8193 8194
}

8195
long sched_getaffinity(pid_t pid, struct cpumask *mask)
Linus Torvalds's avatar
Linus Torvalds committed
8196
{
8197
	struct task_struct *p;
8198
	unsigned long flags;
Linus Torvalds's avatar
Linus Torvalds committed
8199 8200
	int retval;

8201
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
8202 8203 8204 8205 8206 8207

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

8208 8209 8210 8211
	retval = security_task_getscheduler(p);
	if (retval)
		goto out_unlock;

8212
	raw_spin_lock_irqsave(&p->pi_lock, flags);
8213
	cpumask_and(mask, &p->cpus_mask, cpu_active_mask);
8214
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
8215 8216

out_unlock:
8217
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
8218

8219
	return retval;
Linus Torvalds's avatar
Linus Torvalds committed
8220 8221 8222
}

/**
8223
 * sys_sched_getaffinity - get the CPU affinity of a process
Linus Torvalds's avatar
Linus Torvalds committed
8224 8225
 * @pid: pid of the process
 * @len: length in bytes of the bitmask pointed to by user_mask_ptr
8226
 * @user_mask_ptr: user-space pointer to hold the current CPU mask
8227
 *
8228 8229
 * Return: size of CPU mask copied to user_mask_ptr on success. An
 * error code otherwise.
Linus Torvalds's avatar
Linus Torvalds committed
8230
 */
8231 8232
SYSCALL_DEFINE3(sched_getaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
Linus Torvalds's avatar
Linus Torvalds committed
8233 8234
{
	int ret;
8235
	cpumask_var_t mask;
Linus Torvalds's avatar
Linus Torvalds committed
8236

8237
	if ((len * BITS_PER_BYTE) < nr_cpu_ids)
8238 8239
		return -EINVAL;
	if (len & (sizeof(unsigned long)-1))
Linus Torvalds's avatar
Linus Torvalds committed
8240 8241
		return -EINVAL;

8242 8243
	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
		return -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
8244

8245 8246
	ret = sched_getaffinity(pid, mask);
	if (ret == 0) {
8247
		unsigned int retlen = min(len, cpumask_size());
8248 8249

		if (copy_to_user(user_mask_ptr, mask, retlen))
8250 8251
			ret = -EFAULT;
		else
8252
			ret = retlen;
8253 8254
	}
	free_cpumask_var(mask);
Linus Torvalds's avatar
Linus Torvalds committed
8255

8256
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
8257 8258
}

8259
static void do_sched_yield(void)
Linus Torvalds's avatar
Linus Torvalds committed
8260
{
8261 8262 8263
	struct rq_flags rf;
	struct rq *rq;

8264
	rq = this_rq_lock_irq(&rf);
Linus Torvalds's avatar
Linus Torvalds committed
8265

8266
	schedstat_inc(rq->yld_count);
8267
	current->sched_class->yield_task(rq);
Linus Torvalds's avatar
Linus Torvalds committed
8268

8269
	preempt_disable();
8270
	rq_unlock_irq(rq, &rf);
8271
	sched_preempt_enable_no_resched();
Linus Torvalds's avatar
Linus Torvalds committed
8272 8273

	schedule();
8274
}
Linus Torvalds's avatar
Linus Torvalds committed
8275

8276 8277 8278 8279 8280 8281 8282 8283
/**
 * 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.
 */
8284 8285 8286
SYSCALL_DEFINE0(sched_yield)
{
	do_sched_yield();
Linus Torvalds's avatar
Linus Torvalds committed
8287 8288 8289
	return 0;
}

8290 8291
#if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
int __sched __cond_resched(void)
Linus Torvalds's avatar
Linus Torvalds committed
8292
{
8293
	if (should_resched(0)) {
8294
		preempt_schedule_common();
Linus Torvalds's avatar
Linus Torvalds committed
8295 8296
		return 1;
	}
8297 8298 8299 8300 8301 8302 8303 8304 8305 8306 8307
	/*
	 * 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.
	 */
8308
#ifndef CONFIG_PREEMPT_RCU
8309
	rcu_all_qs();
8310
#endif
Linus Torvalds's avatar
Linus Torvalds committed
8311 8312
	return 0;
}
8313 8314 8315 8316
EXPORT_SYMBOL(__cond_resched);
#endif

#ifdef CONFIG_PREEMPT_DYNAMIC
8317
#if defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
8318 8319
#define cond_resched_dynamic_enabled	__cond_resched
#define cond_resched_dynamic_disabled	((void *)&__static_call_return0)
8320
DEFINE_STATIC_CALL_RET0(cond_resched, __cond_resched);
8321
EXPORT_STATIC_CALL_TRAMP(cond_resched);
8322

8323 8324
#define might_resched_dynamic_enabled	__cond_resched
#define might_resched_dynamic_disabled	((void *)&__static_call_return0)
8325
DEFINE_STATIC_CALL_RET0(might_resched, __cond_resched);
8326
EXPORT_STATIC_CALL_TRAMP(might_resched);
8327 8328 8329 8330 8331 8332 8333 8334 8335 8336 8337 8338 8339 8340 8341 8342 8343 8344 8345
#elif defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
static DEFINE_STATIC_KEY_FALSE(sk_dynamic_cond_resched);
int __sched dynamic_cond_resched(void)
{
	if (!static_branch_unlikely(&sk_dynamic_cond_resched))
		return 0;
	return __cond_resched();
}
EXPORT_SYMBOL(dynamic_cond_resched);

static DEFINE_STATIC_KEY_FALSE(sk_dynamic_might_resched);
int __sched dynamic_might_resched(void)
{
	if (!static_branch_unlikely(&sk_dynamic_might_resched))
		return 0;
	return __cond_resched();
}
EXPORT_SYMBOL(dynamic_might_resched);
#endif
8346
#endif
Linus Torvalds's avatar
Linus Torvalds committed
8347 8348

/*
8349
 * __cond_resched_lock() - if a reschedule is pending, drop the given lock,
Linus Torvalds's avatar
Linus Torvalds committed
8350 8351
 * call schedule, and on return reacquire the lock.
 *
8352
 * This works OK both with and without CONFIG_PREEMPTION. We do strange low-level
Linus Torvalds's avatar
Linus Torvalds committed
8353 8354 8355
 * operations here to prevent schedule() from being called twice (once via
 * spin_unlock(), once by hand).
 */
8356
int __cond_resched_lock(spinlock_t *lock)
Linus Torvalds's avatar
Linus Torvalds committed
8357
{
8358
	int resched = should_resched(PREEMPT_LOCK_OFFSET);
Jan Kara's avatar
Jan Kara committed
8359 8360
	int ret = 0;

8361 8362
	lockdep_assert_held(lock);

8363
	if (spin_needbreak(lock) || resched) {
Linus Torvalds's avatar
Linus Torvalds committed
8364
		spin_unlock(lock);
8365
		if (!_cond_resched())
Nick Piggin's avatar
Nick Piggin committed
8366
			cpu_relax();
Jan Kara's avatar
Jan Kara committed
8367
		ret = 1;
Linus Torvalds's avatar
Linus Torvalds committed
8368 8369
		spin_lock(lock);
	}
Jan Kara's avatar
Jan Kara committed
8370
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
8371
}
8372
EXPORT_SYMBOL(__cond_resched_lock);
Linus Torvalds's avatar
Linus Torvalds committed
8373

8374 8375 8376 8377 8378 8379 8380 8381 8382
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);
8383
		if (!_cond_resched())
8384 8385 8386 8387 8388 8389 8390 8391 8392 8393 8394 8395 8396 8397 8398 8399 8400
			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);
8401
		if (!_cond_resched())
8402 8403 8404 8405 8406 8407 8408 8409
			cpu_relax();
		ret = 1;
		write_lock(lock);
	}
	return ret;
}
EXPORT_SYMBOL(__cond_resched_rwlock_write);

8410 8411
#ifdef CONFIG_PREEMPT_DYNAMIC

8412
#ifdef CONFIG_GENERIC_ENTRY
8413
#include <linux/entry-common.h>
8414
#endif
8415 8416 8417 8418 8419 8420 8421 8422 8423 8424 8425 8426 8427 8428 8429 8430 8431 8432 8433 8434 8435 8436 8437 8438 8439 8440 8441 8442 8443 8444 8445 8446 8447 8448 8449 8450 8451 8452 8453 8454 8455 8456 8457 8458 8459 8460 8461 8462 8463 8464 8465 8466 8467 8468

/*
 * 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
 */

enum {
	preempt_dynamic_undefined = -1,
	preempt_dynamic_none,
	preempt_dynamic_voluntary,
	preempt_dynamic_full,
};

int preempt_dynamic_mode = preempt_dynamic_undefined;

int sched_dynamic_mode(const char *str)
{
	if (!strcmp(str, "none"))
		return preempt_dynamic_none;

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

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

	return -EINVAL;
}

8469
#if defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
8470 8471
#define preempt_dynamic_enable(f)	static_call_update(f, f##_dynamic_enabled)
#define preempt_dynamic_disable(f)	static_call_update(f, f##_dynamic_disabled)
8472 8473 8474 8475 8476 8477
#elif defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
#define preempt_dynamic_enable(f)	static_key_enable(&sk_dynamic_##f.key)
#define preempt_dynamic_disable(f)	static_key_disable(&sk_dynamic_##f.key)
#else
#error "Unsupported PREEMPT_DYNAMIC mechanism"
#endif
8478

8479 8480 8481 8482 8483 8484
void sched_dynamic_update(int mode)
{
	/*
	 * Avoid {NONE,VOLUNTARY} -> FULL transitions from ever ending up in
	 * the ZERO state, which is invalid.
	 */
8485 8486 8487 8488 8489
	preempt_dynamic_enable(cond_resched);
	preempt_dynamic_enable(might_resched);
	preempt_dynamic_enable(preempt_schedule);
	preempt_dynamic_enable(preempt_schedule_notrace);
	preempt_dynamic_enable(irqentry_exit_cond_resched);
8490 8491 8492

	switch (mode) {
	case preempt_dynamic_none:
8493 8494 8495 8496 8497
		preempt_dynamic_enable(cond_resched);
		preempt_dynamic_disable(might_resched);
		preempt_dynamic_disable(preempt_schedule);
		preempt_dynamic_disable(preempt_schedule_notrace);
		preempt_dynamic_disable(irqentry_exit_cond_resched);
8498 8499 8500 8501
		pr_info("Dynamic Preempt: none\n");
		break;

	case preempt_dynamic_voluntary:
8502 8503 8504 8505 8506
		preempt_dynamic_enable(cond_resched);
		preempt_dynamic_enable(might_resched);
		preempt_dynamic_disable(preempt_schedule);
		preempt_dynamic_disable(preempt_schedule_notrace);
		preempt_dynamic_disable(irqentry_exit_cond_resched);
8507 8508 8509 8510
		pr_info("Dynamic Preempt: voluntary\n");
		break;

	case preempt_dynamic_full:
8511 8512 8513 8514 8515
		preempt_dynamic_disable(cond_resched);
		preempt_dynamic_disable(might_resched);
		preempt_dynamic_enable(preempt_schedule);
		preempt_dynamic_enable(preempt_schedule_notrace);
		preempt_dynamic_enable(irqentry_exit_cond_resched);
8516 8517 8518 8519 8520 8521 8522 8523 8524 8525 8526 8527 8528 8529 8530 8531 8532 8533 8534 8535 8536 8537 8538 8539 8540 8541 8542 8543 8544 8545 8546 8547 8548 8549 8550 8551
		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);
		return 0;
	}

	sched_dynamic_update(mode);
	return 1;
}
__setup("preempt=", setup_preempt_mode);

static void __init preempt_dynamic_init(void)
{
	if (preempt_dynamic_mode == preempt_dynamic_undefined) {
		if (IS_ENABLED(CONFIG_PREEMPT_NONE)) {
			sched_dynamic_update(preempt_dynamic_none);
		} else if (IS_ENABLED(CONFIG_PREEMPT_VOLUNTARY)) {
			sched_dynamic_update(preempt_dynamic_voluntary);
		} else {
			/* Default static call setting, nothing to do */
			WARN_ON_ONCE(!IS_ENABLED(CONFIG_PREEMPT));
			preempt_dynamic_mode = preempt_dynamic_full;
			pr_info("Dynamic Preempt: full\n");
		}
	}
}

8552 8553 8554 8555 8556 8557 8558 8559 8560 8561 8562 8563
#define PREEMPT_MODEL_ACCESSOR(mode) \
	bool preempt_model_##mode(void)						 \
	{									 \
		WARN_ON_ONCE(preempt_dynamic_mode == preempt_dynamic_undefined); \
		return preempt_dynamic_mode == preempt_dynamic_##mode;		 \
	}									 \
	EXPORT_SYMBOL_GPL(preempt_model_##mode)

PREEMPT_MODEL_ACCESSOR(none);
PREEMPT_MODEL_ACCESSOR(voluntary);
PREEMPT_MODEL_ACCESSOR(full);

8564 8565 8566 8567 8568 8569
#else /* !CONFIG_PREEMPT_DYNAMIC */

static inline void preempt_dynamic_init(void) { }

#endif /* #ifdef CONFIG_PREEMPT_DYNAMIC */

Linus Torvalds's avatar
Linus Torvalds committed
8570 8571 8572
/**
 * yield - yield the current processor to other threads.
 *
Peter Zijlstra's avatar
Peter Zijlstra committed
8573 8574 8575 8576
 * 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
8577
 * it, it's already broken.
Peter Zijlstra's avatar
Peter Zijlstra committed
8578 8579 8580 8581
 *
 * Typical broken usage is:
 *
 * while (!event)
8582
 *	yield();
Peter Zijlstra's avatar
Peter Zijlstra committed
8583 8584 8585 8586 8587 8588 8589 8590
 *
 * 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
8591 8592 8593 8594
 */
void __sched yield(void)
{
	set_current_state(TASK_RUNNING);
8595
	do_sched_yield();
Linus Torvalds's avatar
Linus Torvalds committed
8596 8597 8598
}
EXPORT_SYMBOL(yield);

8599 8600 8601 8602
/**
 * yield_to - yield the current processor to another thread in
 * your thread group, or accelerate that thread toward the
 * processor it's on.
8603 8604
 * @p: target task
 * @preempt: whether task preemption is allowed or not
8605 8606 8607 8608
 *
 * 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.
 *
8609
 * Return:
8610 8611 8612
 *	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.
8613
 */
8614
int __sched yield_to(struct task_struct *p, bool preempt)
8615 8616 8617 8618
{
	struct task_struct *curr = current;
	struct rq *rq, *p_rq;
	unsigned long flags;
8619
	int yielded = 0;
8620 8621 8622 8623 8624 8625

	local_irq_save(flags);
	rq = this_rq();

again:
	p_rq = task_rq(p);
8626 8627 8628 8629 8630 8631 8632 8633 8634
	/*
	 * 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;
	}

8635
	double_rq_lock(rq, p_rq);
8636
	if (task_rq(p) != p_rq) {
8637 8638 8639 8640 8641
		double_rq_unlock(rq, p_rq);
		goto again;
	}

	if (!curr->sched_class->yield_to_task)
8642
		goto out_unlock;
8643 8644

	if (curr->sched_class != p->sched_class)
8645
		goto out_unlock;
8646

8647
	if (task_on_cpu(p_rq, p) || !task_is_running(p))
8648
		goto out_unlock;
8649

8650
	yielded = curr->sched_class->yield_to_task(rq, p);
8651
	if (yielded) {
8652
		schedstat_inc(rq->yld_count);
8653 8654 8655 8656 8657
		/*
		 * Make p's CPU reschedule; pick_next_entity takes care of
		 * fairness.
		 */
		if (preempt && rq != p_rq)
8658
			resched_curr(p_rq);
8659
	}
8660

8661
out_unlock:
8662
	double_rq_unlock(rq, p_rq);
8663
out_irq:
8664 8665
	local_irq_restore(flags);

8666
	if (yielded > 0)
8667 8668 8669 8670 8671 8672
		schedule();

	return yielded;
}
EXPORT_SYMBOL_GPL(yield_to);

8673 8674 8675 8676 8677
int io_schedule_prepare(void)
{
	int old_iowait = current->in_iowait;

	current->in_iowait = 1;
8678
	blk_flush_plug(current->plug, true);
8679 8680 8681 8682 8683 8684 8685 8686
	return old_iowait;
}

void io_schedule_finish(int token)
{
	current->in_iowait = token;
}

Linus Torvalds's avatar
Linus Torvalds committed
8687
/*
Ingo Molnar's avatar
Ingo Molnar committed
8688
 * This task is about to go to sleep on IO. Increment rq->nr_iowait so
Linus Torvalds's avatar
Linus Torvalds committed
8689 8690 8691 8692
 * that process accounting knows that this is a task in IO wait state.
 */
long __sched io_schedule_timeout(long timeout)
{
8693
	int token;
Linus Torvalds's avatar
Linus Torvalds committed
8694 8695
	long ret;

8696
	token = io_schedule_prepare();
Linus Torvalds's avatar
Linus Torvalds committed
8697
	ret = schedule_timeout(timeout);
8698
	io_schedule_finish(token);
8699

Linus Torvalds's avatar
Linus Torvalds committed
8700 8701
	return ret;
}
8702
EXPORT_SYMBOL(io_schedule_timeout);
Linus Torvalds's avatar
Linus Torvalds committed
8703

8704
void __sched io_schedule(void)
8705 8706 8707 8708 8709 8710 8711 8712 8713
{
	int token;

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

Linus Torvalds's avatar
Linus Torvalds committed
8714 8715 8716 8717
/**
 * sys_sched_get_priority_max - return maximum RT priority.
 * @policy: scheduling class.
 *
8718 8719 8720
 * 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
8721
 */
8722
SYSCALL_DEFINE1(sched_get_priority_max, int, policy)
Linus Torvalds's avatar
Linus Torvalds committed
8723 8724 8725 8726 8727 8728
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
8729
		ret = MAX_RT_PRIO-1;
Linus Torvalds's avatar
Linus Torvalds committed
8730
		break;
8731
	case SCHED_DEADLINE:
Linus Torvalds's avatar
Linus Torvalds committed
8732
	case SCHED_NORMAL:
8733
	case SCHED_BATCH:
Ingo Molnar's avatar
Ingo Molnar committed
8734
	case SCHED_IDLE:
Linus Torvalds's avatar
Linus Torvalds committed
8735 8736 8737 8738 8739 8740 8741 8742 8743 8744
		ret = 0;
		break;
	}
	return ret;
}

/**
 * sys_sched_get_priority_min - return minimum RT priority.
 * @policy: scheduling class.
 *
8745 8746 8747
 * 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
8748
 */
8749
SYSCALL_DEFINE1(sched_get_priority_min, int, policy)
Linus Torvalds's avatar
Linus Torvalds committed
8750 8751 8752 8753 8754 8755 8756 8757
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = 1;
		break;
8758
	case SCHED_DEADLINE:
Linus Torvalds's avatar
Linus Torvalds committed
8759
	case SCHED_NORMAL:
8760
	case SCHED_BATCH:
Ingo Molnar's avatar
Ingo Molnar committed
8761
	case SCHED_IDLE:
Linus Torvalds's avatar
Linus Torvalds committed
8762 8763 8764 8765 8766
		ret = 0;
	}
	return ret;
}

8767
static int sched_rr_get_interval(pid_t pid, struct timespec64 *t)
Linus Torvalds's avatar
Linus Torvalds committed
8768
{
8769
	struct task_struct *p;
Dmitry Adamushko's avatar
Dmitry Adamushko committed
8770
	unsigned int time_slice;
8771
	struct rq_flags rf;
8772
	struct rq *rq;
8773
	int retval;
Linus Torvalds's avatar
Linus Torvalds committed
8774 8775

	if (pid < 0)
8776
		return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
8777 8778

	retval = -ESRCH;
8779
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
8780 8781 8782 8783 8784 8785 8786 8787
	p = find_process_by_pid(pid);
	if (!p)
		goto out_unlock;

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

8788
	rq = task_rq_lock(p, &rf);
8789 8790 8791
	time_slice = 0;
	if (p->sched_class->get_rr_interval)
		time_slice = p->sched_class->get_rr_interval(rq, p);
8792
	task_rq_unlock(rq, p, &rf);
Dmitry Adamushko's avatar
Dmitry Adamushko committed
8793

8794
	rcu_read_unlock();
8795 8796
	jiffies_to_timespec64(time_slice, t);
	return 0;
8797

Linus Torvalds's avatar
Linus Torvalds committed
8798
out_unlock:
8799
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
8800 8801 8802
	return retval;
}

8803 8804 8805 8806 8807 8808 8809 8810 8811 8812 8813
/**
 * 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.
 */
8814
SYSCALL_DEFINE2(sched_rr_get_interval, pid_t, pid,
8815
		struct __kernel_timespec __user *, interval)
8816 8817 8818 8819 8820 8821 8822 8823 8824 8825
{
	struct timespec64 t;
	int retval = sched_rr_get_interval(pid, &t);

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

	return retval;
}

8826
#ifdef CONFIG_COMPAT_32BIT_TIME
8827 8828
SYSCALL_DEFINE2(sched_rr_get_interval_time32, pid_t, pid,
		struct old_timespec32 __user *, interval)
8829 8830 8831 8832 8833
{
	struct timespec64 t;
	int retval = sched_rr_get_interval(pid, &t);

	if (retval == 0)
8834
		retval = put_old_timespec32(&t, interval);
8835 8836 8837 8838
	return retval;
}
#endif

8839
void sched_show_task(struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
8840 8841
{
	unsigned long free = 0;
8842
	int ppid;
8843

8844 8845
	if (!try_get_task_stack(p))
		return;
8846

8847
	pr_info("task:%-15.15s state:%c", p->comm, task_state_to_char(p));
8848

8849
	if (task_is_running(p))
8850
		pr_cont("  running task    ");
Linus Torvalds's avatar
Linus Torvalds committed
8851
#ifdef CONFIG_DEBUG_STACK_USAGE
8852
	free = stack_not_used(p);
Linus Torvalds's avatar
Linus Torvalds committed
8853
#endif
8854
	ppid = 0;
8855
	rcu_read_lock();
8856 8857
	if (pid_alive(p))
		ppid = task_pid_nr(rcu_dereference(p->real_parent));
8858
	rcu_read_unlock();
8859
	pr_cont(" stack:%-5lu pid:%-5d ppid:%-6d flags:0x%08lx\n",
8860
		free, task_pid_nr(p), ppid,
8861
		read_task_thread_flags(p));
Linus Torvalds's avatar
Linus Torvalds committed
8862

8863
	print_worker_info(KERN_INFO, p);
8864
	print_stop_info(KERN_INFO, p);
8865
	show_stack(p, NULL, KERN_INFO);
8866
	put_task_stack(p);
Linus Torvalds's avatar
Linus Torvalds committed
8867
}
8868
EXPORT_SYMBOL_GPL(sched_show_task);
Linus Torvalds's avatar
Linus Torvalds committed
8869

8870 8871 8872
static inline bool
state_filter_match(unsigned long state_filter, struct task_struct *p)
{
8873 8874
	unsigned int state = READ_ONCE(p->__state);

8875 8876 8877 8878 8879
	/* no filter, everything matches */
	if (!state_filter)
		return true;

	/* filter, but doesn't match */
8880
	if (!(state & state_filter))
8881 8882 8883 8884 8885 8886
		return false;

	/*
	 * When looking for TASK_UNINTERRUPTIBLE skip TASK_IDLE (allows
	 * TASK_KILLABLE).
	 */
8887
	if (state_filter == TASK_UNINTERRUPTIBLE && (state & TASK_NOLOAD))
8888 8889 8890 8891 8892 8893
		return false;

	return true;
}


8894
void show_state_filter(unsigned int state_filter)
Linus Torvalds's avatar
Linus Torvalds committed
8895
{
8896
	struct task_struct *g, *p;
Linus Torvalds's avatar
Linus Torvalds committed
8897

8898
	rcu_read_lock();
8899
	for_each_process_thread(g, p) {
Linus Torvalds's avatar
Linus Torvalds committed
8900 8901
		/*
		 * reset the NMI-timeout, listing all files on a slow
Lucas De Marchi's avatar
Lucas De Marchi committed
8902
		 * console might take a lot of time:
8903 8904 8905
		 * 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
8906 8907
		 */
		touch_nmi_watchdog();
8908
		touch_all_softlockup_watchdogs();
8909
		if (state_filter_match(state_filter, p))
8910
			sched_show_task(p);
8911
	}
Linus Torvalds's avatar
Linus Torvalds committed
8912

Ingo Molnar's avatar
Ingo Molnar committed
8913
#ifdef CONFIG_SCHED_DEBUG
8914 8915
	if (!state_filter)
		sysrq_sched_debug_show();
Ingo Molnar's avatar
Ingo Molnar committed
8916
#endif
8917
	rcu_read_unlock();
8918 8919 8920
	/*
	 * Only show locks if all tasks are dumped:
	 */
8921
	if (!state_filter)
8922
		debug_show_all_locks();
Linus Torvalds's avatar
Linus Torvalds committed
8923 8924
}

8925 8926 8927
/**
 * init_idle - set up an idle thread for a given CPU
 * @idle: task in question
8928
 * @cpu: CPU the idle task belongs to
8929 8930 8931 8932
 *
 * NOTE: this function does not set the idle thread's NEED_RESCHED
 * flag, to make booting more robust.
 */
8933
void __init init_idle(struct task_struct *idle, int cpu)
Linus Torvalds's avatar
Linus Torvalds committed
8934
{
8935
	struct rq *rq = cpu_rq(cpu);
Linus Torvalds's avatar
Linus Torvalds committed
8936 8937
	unsigned long flags;

8938 8939
	__sched_fork(0, idle);

8940
	raw_spin_lock_irqsave(&idle->pi_lock, flags);
8941
	raw_spin_rq_lock(rq);
8942

8943
	idle->__state = TASK_RUNNING;
Ingo Molnar's avatar
Ingo Molnar committed
8944
	idle->se.exec_start = sched_clock();
8945 8946 8947 8948 8949 8950
	/*
	 * 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
8951

8952 8953
#ifdef CONFIG_SMP
	/*
8954
	 * It's possible that init_idle() gets called multiple times on a task,
8955 8956 8957 8958
	 * in that case do_set_cpus_allowed() will not do the right thing.
	 *
	 * And since this is boot we can forgo the serialization.
	 */
8959
	set_cpus_allowed_common(idle, cpumask_of(cpu), 0);
8960
#endif
8961 8962
	/*
	 * We're having a chicken and egg problem, even though we are
8963
	 * holding rq->lock, the CPU isn't yet set to this CPU so the
8964 8965 8966 8967 8968 8969 8970 8971
	 * 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
8972
	__set_task_cpu(idle, cpu);
8973
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
8974

8975 8976
	rq->idle = idle;
	rcu_assign_pointer(rq->curr, idle);
8977
	idle->on_rq = TASK_ON_RQ_QUEUED;
8978
#ifdef CONFIG_SMP
8979
	idle->on_cpu = 1;
8980
#endif
8981
	raw_spin_rq_unlock(rq);
8982
	raw_spin_unlock_irqrestore(&idle->pi_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
8983 8984

	/* Set the preempt count _outside_ the spinlocks! */
8985
	init_idle_preempt_count(idle, cpu);
8986

Ingo Molnar's avatar
Ingo Molnar committed
8987 8988 8989 8990
	/*
	 * The idle tasks have their own, simple scheduling class:
	 */
	idle->sched_class = &idle_sched_class;
8991
	ftrace_graph_init_idle_task(idle, cpu);
8992
	vtime_init_idle(idle, cpu);
8993
#ifdef CONFIG_SMP
8994 8995
	sprintf(idle->comm, "%s/%d", INIT_TASK_COMM, cpu);
#endif
8996 8997
}

8998 8999
#ifdef CONFIG_SMP

9000 9001 9002
int cpuset_cpumask_can_shrink(const struct cpumask *cur,
			      const struct cpumask *trial)
{
9003
	int ret = 1;
9004

9005
	if (cpumask_empty(cur))
9006 9007
		return ret;

9008
	ret = dl_cpuset_cpumask_can_shrink(cur, trial);
9009 9010 9011 9012

	return ret;
}

9013
int task_can_attach(struct task_struct *p,
9014
		    const struct cpumask *cs_effective_cpus)
9015 9016 9017 9018 9019
{
	int ret = 0;

	/*
	 * Kthreads which disallow setaffinity shouldn't be moved
9020
	 * to a new cpuset; we don't want to change their CPU
9021 9022 9023 9024
	 * 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
9025
	 * before cpus_mask may be changed.
9026 9027 9028 9029 9030 9031 9032
	 */
	if (p->flags & PF_NO_SETAFFINITY) {
		ret = -EINVAL;
		goto out;
	}

	if (dl_task(p) && !cpumask_intersects(task_rq(p)->rd->span,
9033 9034
					      cs_effective_cpus)) {
		int cpu = cpumask_any_and(cpu_active_mask, cs_effective_cpus);
9035

9036 9037
		if (unlikely(cpu >= nr_cpu_ids))
			return -EINVAL;
9038 9039
		ret = dl_cpu_busy(cpu, p);
	}
9040 9041 9042 9043 9044

out:
	return ret;
}

9045
bool sched_smp_initialized __read_mostly;
9046

9047 9048 9049 9050 9051 9052 9053 9054 9055 9056
#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;

9057
	if (!cpumask_test_cpu(target_cpu, p->cpus_ptr))
9058 9059 9060 9061
		return -EINVAL;

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

9062
	trace_sched_move_numa(p, curr_cpu, target_cpu);
9063 9064
	return stop_one_cpu(curr_cpu, migration_cpu_stop, &arg);
}
9065 9066 9067 9068 9069 9070 9071

/*
 * 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)
{
9072
	bool queued, running;
9073 9074
	struct rq_flags rf;
	struct rq *rq;
9075

9076
	rq = task_rq_lock(p, &rf);
9077
	queued = task_on_rq_queued(p);
9078 9079
	running = task_current(rq, p);

9080
	if (queued)
9081
		dequeue_task(rq, p, DEQUEUE_SAVE);
9082
	if (running)
9083
		put_prev_task(rq, p);
9084 9085 9086

	p->numa_preferred_nid = nid;

9087
	if (queued)
9088
		enqueue_task(rq, p, ENQUEUE_RESTORE | ENQUEUE_NOCLOCK);
9089
	if (running)
9090
		set_next_task(rq, p);
9091
	task_rq_unlock(rq, p, &rf);
9092
}
Peter Zijlstra's avatar
Peter Zijlstra committed
9093
#endif /* CONFIG_NUMA_BALANCING */
9094

Linus Torvalds's avatar
Linus Torvalds committed
9095
#ifdef CONFIG_HOTPLUG_CPU
9096
/*
9097
 * Ensure that the idle task is using init_mm right before its CPU goes
9098
 * offline.
9099
 */
9100
void idle_task_exit(void)
Linus Torvalds's avatar
Linus Torvalds committed
9101
{
9102
	struct mm_struct *mm = current->active_mm;
9103

9104
	BUG_ON(cpu_online(smp_processor_id()));
9105
	BUG_ON(current != this_rq()->idle);
9106

9107
	if (mm != &init_mm) {
9108
		switch_mm(mm, &init_mm, current);
9109 9110
		finish_arch_post_lock_switch();
	}
9111 9112

	/* finish_cpu(), as ran on the BP, will clean up the active_mm state */
Linus Torvalds's avatar
Linus Torvalds committed
9113 9114
}

9115
static int __balance_push_cpu_stop(void *arg)
Linus Torvalds's avatar
Linus Torvalds committed
9116
{
9117 9118 9119 9120
	struct task_struct *p = arg;
	struct rq *rq = this_rq();
	struct rq_flags rf;
	int cpu;
Linus Torvalds's avatar
Linus Torvalds committed
9121

9122 9123
	raw_spin_lock_irq(&p->pi_lock);
	rq_lock(rq, &rf);
9124

9125 9126 9127 9128 9129
	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);
9130
	}
9131

9132 9133 9134 9135 9136 9137
	rq_unlock(rq, &rf);
	raw_spin_unlock_irq(&p->pi_lock);

	put_task_struct(p);

	return 0;
9138
}
9139

9140 9141
static DEFINE_PER_CPU(struct cpu_stop_work, push_work);

9142
/*
9143
 * Ensure we only run per-cpu kthreads once the CPU goes !active.
9144 9145 9146
 *
 * 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
9147
 */
9148
static void balance_push(struct rq *rq)
Linus Torvalds's avatar
Linus Torvalds committed
9149
{
9150 9151
	struct task_struct *push_task = rq->curr;

9152
	lockdep_assert_rq_held(rq);
9153

9154 9155 9156 9157
	/*
	 * Ensure the thing is persistent until balance_push_set(.on = false);
	 */
	rq->balance_callback = &balance_push_callback;
Linus Torvalds's avatar
Linus Torvalds committed
9158

9159
	/*
9160 9161
	 * Only active while going offline and when invoked on the outgoing
	 * CPU.
9162
	 */
9163
	if (!cpu_dying(rq->cpu) || rq != this_rq())
9164 9165
		return;

Linus Torvalds's avatar
Linus Torvalds committed
9166
	/*
9167 9168
	 * 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
9169
	 */
9170
	if (kthread_is_per_cpu(push_task) ||
9171 9172
	    is_migration_disabled(push_task)) {

9173 9174 9175 9176 9177 9178
		/*
		 * 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.
9179 9180 9181 9182
		 *
		 * 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.
9183
		 */
9184 9185
		if (!rq->nr_running && !rq_has_pinned_tasks(rq) &&
		    rcuwait_active(&rq->hotplug_wait)) {
9186
			raw_spin_rq_unlock(rq);
9187
			rcuwait_wake_up(&rq->hotplug_wait);
9188
			raw_spin_rq_lock(rq);
9189
		}
9190
		return;
9191
	}
9192

9193
	get_task_struct(push_task);
9194
	/*
9195 9196
	 * Temporarily drop rq->lock such that we can wake-up the stop task.
	 * Both preemption and IRQs are still disabled.
9197
	 */
9198
	raw_spin_rq_unlock(rq);
9199 9200 9201 9202 9203
	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
9204
	 * which kthread_is_per_cpu() and will push this task away.
9205
	 */
9206
	raw_spin_rq_lock(rq);
9207
}
9208

9209 9210 9211 9212
static void balance_push_set(int cpu, bool on)
{
	struct rq *rq = cpu_rq(cpu);
	struct rq_flags rf;
9213

9214
	rq_lock_irqsave(rq, &rf);
9215 9216
	if (on) {
		WARN_ON_ONCE(rq->balance_callback);
9217
		rq->balance_callback = &balance_push_callback;
9218
	} else if (rq->balance_callback == &balance_push_callback) {
9219
		rq->balance_callback = NULL;
9220
	}
9221 9222
	rq_unlock_irqrestore(rq, &rf);
}
9223

9224 9225 9226 9227 9228 9229 9230 9231 9232
/*
 * 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();
9233

9234 9235
	rcuwait_wait_event(&rq->hotplug_wait,
			   rq->nr_running == 1 && !rq_has_pinned_tasks(rq),
9236 9237
			   TASK_UNINTERRUPTIBLE);
}
9238

9239
#else
9240

9241 9242
static inline void balance_push(struct rq *rq)
{
9243 9244
}

9245 9246 9247 9248
static inline void balance_push_set(int cpu, bool on)
{
}

9249 9250
static inline void balance_hotplug_wait(void)
{
9251
}
9252

Linus Torvalds's avatar
Linus Torvalds committed
9253 9254
#endif /* CONFIG_HOTPLUG_CPU */

9255
void set_rq_online(struct rq *rq)
9256 9257 9258 9259
{
	if (!rq->online) {
		const struct sched_class *class;

9260
		cpumask_set_cpu(rq->cpu, rq->rd->online);
9261 9262 9263 9264 9265 9266 9267 9268 9269
		rq->online = 1;

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

9270
void set_rq_offline(struct rq *rq)
9271 9272 9273 9274 9275 9276 9277 9278 9279
{
	if (rq->online) {
		const struct sched_class *class;

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

9280
		cpumask_clear_cpu(rq->cpu, rq->rd->online);
9281 9282 9283 9284
		rq->online = 0;
	}
}

9285 9286 9287 9288
/*
 * used to mark begin/end of suspend/resume:
 */
static int num_cpus_frozen;
9289

Linus Torvalds's avatar
Linus Torvalds committed
9290
/*
9291 9292 9293
 * Update cpusets according to cpu_active mask.  If cpusets are
 * disabled, cpuset_update_active_cpus() becomes a simple wrapper
 * around partition_sched_domains().
9294 9295 9296
 *
 * 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
9297
 */
9298
static void cpuset_cpu_active(void)
9299
{
9300
	if (cpuhp_tasks_frozen) {
9301 9302 9303 9304 9305 9306
		/*
		 * 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.
		 */
9307 9308
		partition_sched_domains(1, NULL, NULL);
		if (--num_cpus_frozen)
9309
			return;
9310 9311 9312 9313 9314
		/*
		 * This is the last CPU online operation. So fall through and
		 * restore the original sched domains by considering the
		 * cpuset configurations.
		 */
9315
		cpuset_force_rebuild();
9316
	}
9317
	cpuset_update_active_cpus();
9318
}
9319

9320
static int cpuset_cpu_inactive(unsigned int cpu)
9321
{
9322
	if (!cpuhp_tasks_frozen) {
9323 9324 9325 9326
		int ret = dl_cpu_busy(cpu, NULL);

		if (ret)
			return ret;
9327
		cpuset_update_active_cpus();
9328
	} else {
9329 9330
		num_cpus_frozen++;
		partition_sched_domains(1, NULL, NULL);
9331
	}
9332
	return 0;
9333 9334
}

9335
int sched_cpu_activate(unsigned int cpu)
9336
{
9337
	struct rq *rq = cpu_rq(cpu);
9338
	struct rq_flags rf;
9339

9340
	/*
9341 9342
	 * Clear the balance_push callback and prepare to schedule
	 * regular tasks.
9343
	 */
9344 9345
	balance_push_set(cpu, false);

9346 9347
#ifdef CONFIG_SCHED_SMT
	/*
9348
	 * When going up, increment the number of cores with SMT present.
9349
	 */
9350 9351
	if (cpumask_weight(cpu_smt_mask(cpu)) == 2)
		static_branch_inc_cpuslocked(&sched_smt_present);
9352
#endif
9353
	set_cpu_active(cpu, true);
9354

9355
	if (sched_smp_initialized) {
9356
		sched_update_numa(cpu, true);
9357
		sched_domains_numa_masks_set(cpu);
9358
		cpuset_cpu_active();
9359
	}
9360 9361 9362 9363 9364

	/*
	 * Put the rq online, if not already. This happens:
	 *
	 * 1) In the early boot process, because we build the real domains
9365
	 *    after all CPUs have been brought up.
9366 9367 9368 9369
	 *
	 * 2) At runtime, if cpuset_cpu_active() fails to rebuild the
	 *    domains.
	 */
9370
	rq_lock_irqsave(rq, &rf);
9371 9372 9373 9374
	if (rq->rd) {
		BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
		set_rq_online(rq);
	}
9375
	rq_unlock_irqrestore(rq, &rf);
9376

9377
	return 0;
9378 9379
}

9380
int sched_cpu_deactivate(unsigned int cpu)
9381
{
9382 9383
	struct rq *rq = cpu_rq(cpu);
	struct rq_flags rf;
9384 9385
	int ret;

9386 9387 9388 9389 9390 9391
	/*
	 * Remove CPU from nohz.idle_cpus_mask to prevent participating in
	 * load balancing when not active
	 */
	nohz_balance_exit_idle(rq);

9392
	set_cpu_active(cpu, false);
9393 9394 9395 9396 9397 9398 9399

	/*
	 * 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().
	 */
9400 9401
	balance_push_set(cpu, true);

9402
	/*
9403 9404 9405
	 * 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.
9406
	 *
9407 9408 9409
	 * Specifically, we rely on ttwu to no longer target this CPU, see
	 * ttwu_queue_cond() and is_cpu_allowed().
	 *
9410 9411
	 * Do sync before park smpboot threads to take care the rcu boost case.
	 */
9412
	synchronize_rcu();
9413

9414 9415 9416 9417 9418 9419 9420 9421
	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);

9422 9423 9424 9425 9426 9427
#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);
9428 9429

	sched_core_cpu_deactivate(cpu);
9430 9431
#endif

9432 9433 9434
	if (!sched_smp_initialized)
		return 0;

9435
	sched_update_numa(cpu, false);
9436 9437
	ret = cpuset_cpu_inactive(cpu);
	if (ret) {
9438
		balance_push_set(cpu, false);
9439
		set_cpu_active(cpu, true);
9440
		sched_update_numa(cpu, true);
9441
		return ret;
9442
	}
9443 9444
	sched_domains_numa_masks_clear(cpu);
	return 0;
9445 9446
}

9447 9448 9449 9450 9451 9452 9453 9454
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();
}

9455 9456
int sched_cpu_starting(unsigned int cpu)
{
Peter Zijlstra's avatar
Peter Zijlstra committed
9457
	sched_core_cpu_starting(cpu);
9458
	sched_rq_cpu_starting(cpu);
9459
	sched_tick_start(cpu);
9460
	return 0;
9461 9462
}

9463
#ifdef CONFIG_HOTPLUG_CPU
9464 9465 9466 9467 9468 9469 9470 9471 9472 9473 9474 9475 9476 9477 9478 9479 9480 9481 9482 9483 9484 9485 9486 9487 9488 9489 9490 9491 9492 9493 9494 9495 9496 9497 9498

/*
 * 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);
}

9499 9500 9501 9502 9503
static void dump_rq_tasks(struct rq *rq, const char *loglvl)
{
	struct task_struct *g, *p;
	int cpu = cpu_of(rq);

9504
	lockdep_assert_rq_held(rq);
9505 9506 9507 9508 9509 9510 9511 9512 9513 9514 9515 9516 9517

	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);
	}
}

9518 9519 9520
int sched_cpu_dying(unsigned int cpu)
{
	struct rq *rq = cpu_rq(cpu);
9521
	struct rq_flags rf;
9522 9523

	/* Handle pending wakeups and then migrate everything off */
9524
	sched_tick_stop(cpu);
9525 9526

	rq_lock_irqsave(rq, &rf);
9527 9528 9529 9530
	if (rq->nr_running != 1 || rq_has_pinned_tasks(rq)) {
		WARN(true, "Dying CPU not properly vacated!");
		dump_rq_tasks(rq, KERN_WARNING);
	}
9531 9532
	rq_unlock_irqrestore(rq, &rf);

9533 9534
	calc_load_migrate(rq);
	update_max_interval();
9535
	hrtick_clear(rq);
9536
	sched_core_cpu_dying(cpu);
9537 9538 9539 9540
	return 0;
}
#endif

Linus Torvalds's avatar
Linus Torvalds committed
9541 9542
void __init sched_init_smp(void)
{
9543
	sched_init_numa(NUMA_NO_NODE);
9544

9545 9546
	/*
	 * There's no userspace yet to cause hotplug operations; hence all the
9547
	 * CPU masks are stable and all blatant races in the below code cannot
9548
	 * happen.
9549
	 */
9550
	mutex_lock(&sched_domains_mutex);
9551
	sched_init_domains(cpu_active_mask);
9552
	mutex_unlock(&sched_domains_mutex);
9553

9554
	/* Move init over to a non-isolated CPU */
9555
	if (set_cpus_allowed_ptr(current, housekeeping_cpumask(HK_TYPE_DOMAIN)) < 0)
9556
		BUG();
9557
	current->flags &= ~PF_NO_SETAFFINITY;
9558
	sched_init_granularity();
9559

9560
	init_sched_rt_class();
9561
	init_sched_dl_class();
9562

9563
	sched_smp_initialized = true;
Linus Torvalds's avatar
Linus Torvalds committed
9564
}
9565 9566 9567

static int __init migration_init(void)
{
9568
	sched_cpu_starting(smp_processor_id());
9569
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
9570
}
9571 9572
early_initcall(migration_init);

Linus Torvalds's avatar
Linus Torvalds committed
9573 9574 9575
#else
void __init sched_init_smp(void)
{
9576
	sched_init_granularity();
Linus Torvalds's avatar
Linus Torvalds committed
9577 9578 9579 9580 9581 9582 9583 9584 9585 9586
}
#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);
}

9587
#ifdef CONFIG_CGROUP_SCHED
9588 9589 9590 9591
/*
 * Default task group.
 * Every task in system belongs to this group at bootup.
 */
9592
struct task_group root_task_group;
9593
LIST_HEAD(task_groups);
9594 9595 9596

/* Cacheline aligned slab cache for task_group */
static struct kmem_cache *task_group_cache __read_mostly;
9597
#endif
Peter Zijlstra's avatar
Peter Zijlstra committed
9598

Linus Torvalds's avatar
Linus Torvalds committed
9599 9600
void __init sched_init(void)
{
9601
	unsigned long ptr = 0;
9602
	int i;
9603

9604
	/* Make sure the linker didn't screw up */
9605 9606 9607
	BUG_ON(&idle_sched_class != &fair_sched_class + 1 ||
	       &fair_sched_class != &rt_sched_class + 1 ||
	       &rt_sched_class   != &dl_sched_class + 1);
9608
#ifdef CONFIG_SMP
9609
	BUG_ON(&dl_sched_class != &stop_sched_class + 1);
9610 9611
#endif

9612
	wait_bit_init();
9613

9614
#ifdef CONFIG_FAIR_GROUP_SCHED
9615
	ptr += 2 * nr_cpu_ids * sizeof(void **);
9616 9617
#endif
#ifdef CONFIG_RT_GROUP_SCHED
9618
	ptr += 2 * nr_cpu_ids * sizeof(void **);
9619
#endif
9620 9621
	if (ptr) {
		ptr = (unsigned long)kzalloc(ptr, GFP_NOWAIT);
9622 9623

#ifdef CONFIG_FAIR_GROUP_SCHED
9624
		root_task_group.se = (struct sched_entity **)ptr;
9625 9626
		ptr += nr_cpu_ids * sizeof(void **);

9627
		root_task_group.cfs_rq = (struct cfs_rq **)ptr;
9628
		ptr += nr_cpu_ids * sizeof(void **);
9629

9630 9631
		root_task_group.shares = ROOT_TASK_GROUP_LOAD;
		init_cfs_bandwidth(&root_task_group.cfs_bandwidth);
9632
#endif /* CONFIG_FAIR_GROUP_SCHED */
9633
#ifdef CONFIG_RT_GROUP_SCHED
9634
		root_task_group.rt_se = (struct sched_rt_entity **)ptr;
9635 9636
		ptr += nr_cpu_ids * sizeof(void **);

9637
		root_task_group.rt_rq = (struct rt_rq **)ptr;
9638 9639
		ptr += nr_cpu_ids * sizeof(void **);

9640
#endif /* CONFIG_RT_GROUP_SCHED */
9641
	}
Ingo Molnar's avatar
Ingo Molnar committed
9642

9643
	init_rt_bandwidth(&def_rt_bandwidth, global_rt_period(), global_rt_runtime());
9644

9645 9646 9647 9648
#ifdef CONFIG_SMP
	init_defrootdomain();
#endif

9649
#ifdef CONFIG_RT_GROUP_SCHED
9650
	init_rt_bandwidth(&root_task_group.rt_bandwidth,
9651
			global_rt_period(), global_rt_runtime());
9652
#endif /* CONFIG_RT_GROUP_SCHED */
9653

Dhaval Giani's avatar
Dhaval Giani committed
9654
#ifdef CONFIG_CGROUP_SCHED
9655 9656
	task_group_cache = KMEM_CACHE(task_group, 0);

9657 9658
	list_add(&root_task_group.list, &task_groups);
	INIT_LIST_HEAD(&root_task_group.children);
9659
	INIT_LIST_HEAD(&root_task_group.siblings);
9660
	autogroup_init(&init_task);
Dhaval Giani's avatar
Dhaval Giani committed
9661
#endif /* CONFIG_CGROUP_SCHED */
Peter Zijlstra's avatar
Peter Zijlstra committed
9662

9663
	for_each_possible_cpu(i) {
9664
		struct rq *rq;
Linus Torvalds's avatar
Linus Torvalds committed
9665 9666

		rq = cpu_rq(i);
9667
		raw_spin_lock_init(&rq->__lock);
Nick Piggin's avatar
Nick Piggin committed
9668
		rq->nr_running = 0;
9669 9670
		rq->calc_load_active = 0;
		rq->calc_load_update = jiffies + LOAD_FREQ;
9671
		init_cfs_rq(&rq->cfs);
9672 9673
		init_rt_rq(&rq->rt);
		init_dl_rq(&rq->dl);
Ingo Molnar's avatar
Ingo Molnar committed
9674
#ifdef CONFIG_FAIR_GROUP_SCHED
Peter Zijlstra's avatar
Peter Zijlstra committed
9675
		INIT_LIST_HEAD(&rq->leaf_cfs_rq_list);
9676
		rq->tmp_alone_branch = &rq->leaf_cfs_rq_list;
Dhaval Giani's avatar
Dhaval Giani committed
9677
		/*
9678
		 * How much CPU bandwidth does root_task_group get?
Dhaval Giani's avatar
Dhaval Giani committed
9679 9680
		 *
		 * In case of task-groups formed thr' the cgroup filesystem, it
9681 9682
		 * gets 100% of the CPU resources in the system. This overall
		 * system CPU resource is divided among the tasks of
9683
		 * root_task_group and its child task-groups in a fair manner,
Dhaval Giani's avatar
Dhaval Giani committed
9684 9685 9686
		 * based on each entity's (task or task-group's) weight
		 * (se->load.weight).
		 *
9687
		 * In other words, if root_task_group has 10 tasks of weight
Dhaval Giani's avatar
Dhaval Giani committed
9688
		 * 1024) and two child groups A0 and A1 (of weight 1024 each),
9689
		 * then A0's share of the CPU resource is:
Dhaval Giani's avatar
Dhaval Giani committed
9690
		 *
9691
		 *	A0's bandwidth = 1024 / (10*1024 + 1024 + 1024) = 8.33%
Dhaval Giani's avatar
Dhaval Giani committed
9692
		 *
9693 9694
		 * 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
9695
		 */
9696
		init_tg_cfs_entry(&root_task_group, &rq->cfs, NULL, i, NULL);
Dhaval Giani's avatar
Dhaval Giani committed
9697 9698 9699
#endif /* CONFIG_FAIR_GROUP_SCHED */

		rq->rt.rt_runtime = def_rt_bandwidth.rt_runtime;
9700
#ifdef CONFIG_RT_GROUP_SCHED
9701
		init_tg_rt_entry(&root_task_group, &rq->rt, NULL, i, NULL);
Ingo Molnar's avatar
Ingo Molnar committed
9702
#endif
Linus Torvalds's avatar
Linus Torvalds committed
9703
#ifdef CONFIG_SMP
Nick Piggin's avatar
Nick Piggin committed
9704
		rq->sd = NULL;
9705
		rq->rd = NULL;
9706
		rq->cpu_capacity = rq->cpu_capacity_orig = SCHED_CAPACITY_SCALE;
9707
		rq->balance_callback = &balance_push_callback;
Linus Torvalds's avatar
Linus Torvalds committed
9708
		rq->active_balance = 0;
Ingo Molnar's avatar
Ingo Molnar committed
9709
		rq->next_balance = jiffies;
Linus Torvalds's avatar
Linus Torvalds committed
9710
		rq->push_cpu = 0;
9711
		rq->cpu = i;
9712
		rq->online = 0;
9713 9714
		rq->idle_stamp = 0;
		rq->avg_idle = 2*sysctl_sched_migration_cost;
9715 9716
		rq->wake_stamp = jiffies;
		rq->wake_avg_idle = rq->avg_idle;
9717
		rq->max_idle_balance_cost = sysctl_sched_migration_cost;
9718 9719 9720

		INIT_LIST_HEAD(&rq->cfs_tasks);

9721
		rq_attach_root(rq, &def_root_domain);
9722
#ifdef CONFIG_NO_HZ_COMMON
9723
		rq->last_blocked_load_update_tick = jiffies;
9724
		atomic_set(&rq->nohz_flags, 0);
9725

9726
		INIT_CSD(&rq->nohz_csd, nohz_csd_func, rq);
9727
#endif
9728 9729
#ifdef CONFIG_HOTPLUG_CPU
		rcuwait_init(&rq->hotplug_wait);
9730
#endif
9731
#endif /* CONFIG_SMP */
9732
		hrtick_rq_init(rq);
Linus Torvalds's avatar
Linus Torvalds committed
9733
		atomic_set(&rq->nr_iowait, 0);
Peter Zijlstra's avatar
Peter Zijlstra committed
9734 9735

#ifdef CONFIG_SCHED_CORE
9736
		rq->core = rq;
9737
		rq->core_pick = NULL;
Peter Zijlstra's avatar
Peter Zijlstra committed
9738
		rq->core_enabled = 0;
9739
		rq->core_tree = RB_ROOT;
9740 9741 9742
		rq->core_forceidle_count = 0;
		rq->core_forceidle_occupation = 0;
		rq->core_forceidle_start = 0;
9743 9744

		rq->core_cookie = 0UL;
Peter Zijlstra's avatar
Peter Zijlstra committed
9745
#endif
Linus Torvalds's avatar
Linus Torvalds committed
9746 9747
	}

9748
	set_load_weight(&init_task, false);
9749

Linus Torvalds's avatar
Linus Torvalds committed
9750 9751 9752
	/*
	 * The boot idle thread does lazy MMU switching as well:
	 */
9753
	mmgrab(&init_mm);
Linus Torvalds's avatar
Linus Torvalds committed
9754 9755
	enter_lazy_tlb(&init_mm, current);

9756 9757 9758 9759 9760 9761
	/*
	 * 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.
	 */
9762
	WARN_ON(!set_kthread_struct(current));
9763

Linus Torvalds's avatar
Linus Torvalds committed
9764 9765 9766 9767 9768 9769 9770
	/*
	 * 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());
9771 9772 9773

	calc_load_update = jiffies + LOAD_FREQ;

9774
#ifdef CONFIG_SMP
9775
	idle_thread_set_boot_cpu();
9776
	balance_push_set(smp_processor_id(), false);
9777 9778
#endif
	init_sched_fair_class();
9779

9780 9781
	psi_init();

9782 9783
	init_uclamp();

9784 9785
	preempt_dynamic_init();

9786
	scheduler_running = 1;
Linus Torvalds's avatar
Linus Torvalds committed
9787 9788
}

9789
#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
9790

9791
void __might_sleep(const char *file, int line)
Linus Torvalds's avatar
Linus Torvalds committed
9792
{
9793
	unsigned int state = get_current_state();
Peter Zijlstra's avatar
Peter Zijlstra committed
9794 9795 9796 9797 9798
	/*
	 * 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.
	 */
9799
	WARN_ONCE(state != TASK_RUNNING && current->task_state_change,
Peter Zijlstra's avatar
Peter Zijlstra committed
9800
			"do not call blocking ops when !TASK_RUNNING; "
9801
			"state=%x set at [<%p>] %pS\n", state,
Peter Zijlstra's avatar
Peter Zijlstra committed
9802
			(void *)current->task_state_change,
9803
			(void *)current->task_state_change);
Peter Zijlstra's avatar
Peter Zijlstra committed
9804

9805
	__might_resched(file, line, 0);
9806 9807 9808
}
EXPORT_SYMBOL(__might_sleep);

9809 9810 9811 9812 9813 9814 9815 9816 9817 9818 9819 9820
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);
}

9821 9822 9823 9824 9825 9826 9827 9828 9829 9830
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
9831
{
9832 9833 9834
	/* Ratelimiting timestamp: */
	static unsigned long prev_jiffy;

9835
	unsigned long preempt_disable_ip;
Linus Torvalds's avatar
Linus Torvalds committed
9836

9837 9838 9839
	/* WARN_ON_ONCE() by default, no rate limit required: */
	rcu_sleep_check();

9840
	if ((resched_offsets_ok(offsets) && !irqs_disabled() &&
9841
	     !is_idle_task(current) && !current->non_block_count) ||
9842 9843
	    system_state == SYSTEM_BOOTING || system_state > SYSTEM_RUNNING ||
	    oops_in_progress)
9844
		return;
9845

9846 9847 9848 9849
	if (time_before(jiffies, prev_jiffy + HZ) && prev_jiffy)
		return;
	prev_jiffy = jiffies;

9850
	/* Save this before calling printk(), since that will clobber it: */
9851 9852
	preempt_disable_ip = get_preempt_disable_ip(current);

9853 9854 9855 9856 9857
	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);
9858
	pr_err("preempt_count: %x, expected: %x\n", preempt_count(),
9859
	       offsets & MIGHT_RESCHED_PREEMPT_MASK);
9860 9861

	if (IS_ENABLED(CONFIG_PREEMPT_RCU)) {
9862 9863
		pr_err("RCU nest depth: %d, expected: %u\n",
		       rcu_preempt_depth(), offsets >> MIGHT_RESCHED_RCU_SHIFT);
9864
	}
9865

9866
	if (task_stack_end_corrupted(current))
9867
		pr_emerg("Thread overran stack, or stack corrupted\n");
9868

9869 9870 9871
	debug_show_held_locks(current);
	if (irqs_disabled())
		print_irqtrace_events(current);
9872

9873 9874
	print_preempt_disable_ip(offsets & MIGHT_RESCHED_PREEMPT_MASK,
				 preempt_disable_ip);
9875

9876
	dump_stack();
9877
	add_taint(TAINT_WARN, LOCKDEP_STILL_OK);
Linus Torvalds's avatar
Linus Torvalds committed
9878
}
9879
EXPORT_SYMBOL(__might_resched);
9880 9881 9882 9883 9884 9885 9886 9887 9888 9889 9890 9891 9892 9893 9894 9895 9896 9897 9898 9899 9900 9901 9902 9903 9904 9905 9906 9907

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);
9908 9909 9910 9911 9912 9913 9914 9915 9916 9917 9918 9919 9920 9921 9922 9923 9924 9925 9926 9927 9928 9929 9930 9931 9932 9933 9934 9935 9936 9937 9938 9939 9940

#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
9941 9942 9943
#endif

#ifdef CONFIG_MAGIC_SYSRQ
9944
void normalize_rt_tasks(void)
9945
{
9946
	struct task_struct *g, *p;
9947 9948 9949
	struct sched_attr attr = {
		.sched_policy = SCHED_NORMAL,
	};
Linus Torvalds's avatar
Linus Torvalds committed
9950

9951
	read_lock(&tasklist_lock);
9952
	for_each_process_thread(g, p) {
9953 9954 9955
		/*
		 * Only normalize user tasks:
		 */
9956
		if (p->flags & PF_KTHREAD)
9957 9958
			continue;

9959
		p->se.exec_start = 0;
9960 9961 9962
		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
9963

9964
		if (!dl_task(p) && !rt_task(p)) {
Ingo Molnar's avatar
Ingo Molnar committed
9965 9966 9967 9968
			/*
			 * Renice negative nice level userspace
			 * tasks back to 0:
			 */
9969
			if (task_nice(p) < 0)
Ingo Molnar's avatar
Ingo Molnar committed
9970
				set_user_nice(p, 0);
Linus Torvalds's avatar
Linus Torvalds committed
9971
			continue;
Ingo Molnar's avatar
Ingo Molnar committed
9972
		}
Linus Torvalds's avatar
Linus Torvalds committed
9973

9974
		__sched_setscheduler(p, &attr, false, false);
9975
	}
9976
	read_unlock(&tasklist_lock);
Linus Torvalds's avatar
Linus Torvalds committed
9977 9978 9979
}

#endif /* CONFIG_MAGIC_SYSRQ */
9980

9981
#if defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB)
9982
/*
9983
 * These functions are only useful for the IA64 MCA handling, or kdb.
9984 9985 9986 9987 9988 9989 9990 9991 9992
 *
 * 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.
 */

/**
9993
 * curr_task - return the current task for a given CPU.
9994 9995 9996
 * @cpu: the processor in question.
 *
 * ONLY VALID WHEN THE WHOLE SYSTEM IS STOPPED!
9997 9998
 *
 * Return: The current task for @cpu.
9999
 */
10000
struct task_struct *curr_task(int cpu)
10001 10002 10003 10004
{
	return cpu_curr(cpu);
}

10005 10006 10007
#endif /* defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB) */

#ifdef CONFIG_IA64
10008
/**
10009
 * ia64_set_curr_task - set the current task for a given CPU.
10010 10011 10012 10013
 * @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
10014
 * are serviced on a separate stack. It allows the architecture to switch the
10015
 * notion of the current task on a CPU in a non-blocking manner. This function
10016 10017 10018 10019 10020 10021 10022
 * 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!
 */
10023
void ia64_set_curr_task(int cpu, struct task_struct *p)
10024 10025 10026 10027 10028
{
	cpu_curr(cpu) = p;
}

#endif
10029

Dhaval Giani's avatar
Dhaval Giani committed
10030
#ifdef CONFIG_CGROUP_SCHED
10031 10032 10033
/* task_group_lock serializes the addition/removal of task groups */
static DEFINE_SPINLOCK(task_group_lock);

10034 10035 10036 10037
static inline void alloc_uclamp_sched_group(struct task_group *tg,
					    struct task_group *parent)
{
#ifdef CONFIG_UCLAMP_TASK_GROUP
10038
	enum uclamp_id clamp_id;
10039 10040 10041 10042

	for_each_clamp_id(clamp_id) {
		uclamp_se_set(&tg->uclamp_req[clamp_id],
			      uclamp_none(clamp_id), false);
10043
		tg->uclamp[clamp_id] = parent->uclamp[clamp_id];
10044 10045 10046 10047
	}
#endif
}

10048
static void sched_free_group(struct task_group *tg)
10049 10050 10051
{
	free_fair_sched_group(tg);
	free_rt_sched_group(tg);
10052
	autogroup_free(tg);
10053
	kmem_cache_free(task_group_cache, tg);
10054 10055
}

10056 10057 10058 10059 10060 10061 10062 10063 10064 10065 10066 10067 10068 10069 10070 10071
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);
}

10072
/* allocate runqueue etc for a new task group */
10073
struct task_group *sched_create_group(struct task_group *parent)
10074 10075 10076
{
	struct task_group *tg;

10077
	tg = kmem_cache_alloc(task_group_cache, GFP_KERNEL | __GFP_ZERO);
10078 10079 10080
	if (!tg)
		return ERR_PTR(-ENOMEM);

10081
	if (!alloc_fair_sched_group(tg, parent))
10082 10083
		goto err;

10084
	if (!alloc_rt_sched_group(tg, parent))
10085 10086
		goto err;

10087 10088
	alloc_uclamp_sched_group(tg, parent);

10089 10090 10091
	return tg;

err:
10092
	sched_free_group(tg);
10093 10094 10095 10096 10097 10098 10099
	return ERR_PTR(-ENOMEM);
}

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

10100
	spin_lock_irqsave(&task_group_lock, flags);
Peter Zijlstra's avatar
Peter Zijlstra committed
10101
	list_add_rcu(&tg->list, &task_groups);
Peter Zijlstra's avatar
Peter Zijlstra committed
10102

10103 10104
	/* Root should already exist: */
	WARN_ON(!parent);
Peter Zijlstra's avatar
Peter Zijlstra committed
10105 10106 10107

	tg->parent = parent;
	INIT_LIST_HEAD(&tg->children);
10108
	list_add_rcu(&tg->siblings, &parent->children);
10109
	spin_unlock_irqrestore(&task_group_lock, flags);
10110 10111

	online_fair_sched_group(tg);
10112 10113
}

10114
/* rcu callback to free various structures associated with a task group */
10115
static void sched_unregister_group_rcu(struct rcu_head *rhp)
10116
{
10117
	/* Now it should be safe to free those cfs_rqs: */
10118
	sched_unregister_group(container_of(rhp, struct task_group, rcu));
10119 10120
}

10121
void sched_destroy_group(struct task_group *tg)
10122
{
10123
	/* Wait for possible concurrent references to cfs_rqs complete: */
10124
	call_rcu(&tg->rcu, sched_unregister_group_rcu);
10125 10126
}

10127
void sched_release_group(struct task_group *tg)
10128
{
10129
	unsigned long flags;
10130

10131 10132 10133 10134 10135 10136 10137 10138 10139 10140 10141 10142 10143
	/*
	 * 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.
	 */
10144
	spin_lock_irqsave(&task_group_lock, flags);
Peter Zijlstra's avatar
Peter Zijlstra committed
10145
	list_del_rcu(&tg->list);
Peter Zijlstra's avatar
Peter Zijlstra committed
10146
	list_del_rcu(&tg->siblings);
10147
	spin_unlock_irqrestore(&task_group_lock, flags);
10148 10149
}

10150
static void sched_change_group(struct task_struct *tsk)
10151
{
10152
	struct task_group *tg;
10153

10154 10155 10156 10157 10158 10159
	/*
	 * 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),
10160 10161 10162 10163
			  struct task_group, css);
	tg = autogroup_task_group(tsk, tg);
	tsk->sched_task_group = tg;

Peter Zijlstra's avatar
Peter Zijlstra committed
10164
#ifdef CONFIG_FAIR_GROUP_SCHED
10165
	if (tsk->sched_class->task_change_group)
10166
		tsk->sched_class->task_change_group(tsk);
10167
	else
Peter Zijlstra's avatar
Peter Zijlstra committed
10168
#endif
10169
		set_task_rq(tsk, task_cpu(tsk));
10170 10171 10172 10173 10174 10175 10176 10177 10178 10179 10180
}

/*
 * 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)
{
10181 10182
	int queued, running, queue_flags =
		DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
10183 10184 10185 10186
	struct rq_flags rf;
	struct rq *rq;

	rq = task_rq_lock(tsk, &rf);
10187
	update_rq_clock(rq);
10188 10189 10190 10191 10192

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

	if (queued)
10193
		dequeue_task(rq, tsk, queue_flags);
10194
	if (running)
10195 10196
		put_prev_task(rq, tsk);

10197
	sched_change_group(tsk);
Peter Zijlstra's avatar
Peter Zijlstra committed
10198

10199
	if (queued)
10200
		enqueue_task(rq, tsk, queue_flags);
10201
	if (running) {
10202
		set_next_task(rq, tsk);
10203 10204 10205 10206 10207 10208 10209
		/*
		 * 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);
	}
10210

10211
	task_rq_unlock(rq, tsk, &rf);
10212
}
10213

10214
static inline struct task_group *css_tg(struct cgroup_subsys_state *css)
10215
{
10216
	return css ? container_of(css, struct task_group, css) : NULL;
10217 10218
}

10219 10220
static struct cgroup_subsys_state *
cpu_cgroup_css_alloc(struct cgroup_subsys_state *parent_css)
10221
{
10222 10223
	struct task_group *parent = css_tg(parent_css);
	struct task_group *tg;
10224

10225
	if (!parent) {
10226
		/* This is early initialization for the top cgroup */
10227
		return &root_task_group.css;
10228 10229
	}

10230
	tg = sched_create_group(parent);
10231 10232 10233 10234 10235 10236
	if (IS_ERR(tg))
		return ERR_PTR(-ENOMEM);

	return &tg->css;
}

10237 10238 10239 10240 10241 10242 10243 10244
/* 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);
10245 10246 10247

#ifdef CONFIG_UCLAMP_TASK_GROUP
	/* Propagate the effective uclamp value for the new group */
10248 10249
	mutex_lock(&uclamp_mutex);
	rcu_read_lock();
10250
	cpu_util_update_eff(css);
10251 10252
	rcu_read_unlock();
	mutex_unlock(&uclamp_mutex);
10253 10254
#endif

10255 10256 10257
	return 0;
}

10258
static void cpu_cgroup_css_released(struct cgroup_subsys_state *css)
10259
{
10260
	struct task_group *tg = css_tg(css);
10261

10262
	sched_release_group(tg);
10263 10264
}

10265
static void cpu_cgroup_css_free(struct cgroup_subsys_state *css)
10266
{
10267
	struct task_group *tg = css_tg(css);
10268

10269 10270 10271
	/*
	 * Relies on the RCU grace period between css_released() and this.
	 */
10272
	sched_unregister_group(tg);
10273 10274
}

10275
#ifdef CONFIG_RT_GROUP_SCHED
10276
static int cpu_cgroup_can_attach(struct cgroup_taskset *tset)
10277
{
10278
	struct task_struct *task;
10279
	struct cgroup_subsys_state *css;
10280

10281
	cgroup_taskset_for_each(task, css, tset) {
10282
		if (!sched_rt_can_attach(css_tg(css), task))
10283 10284
			return -EINVAL;
	}
10285
	return 0;
10286
}
10287
#endif
10288

10289
static void cpu_cgroup_attach(struct cgroup_taskset *tset)
10290
{
10291
	struct task_struct *task;
10292
	struct cgroup_subsys_state *css;
10293

10294
	cgroup_taskset_for_each(task, css, tset)
10295
		sched_move_task(task);
10296 10297
}

10298
#ifdef CONFIG_UCLAMP_TASK_GROUP
10299 10300 10301 10302 10303 10304
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];
10305
	enum uclamp_id clamp_id;
10306 10307
	unsigned int clamps;

10308 10309 10310
	lockdep_assert_held(&uclamp_mutex);
	SCHED_WARN_ON(!rcu_read_lock_held());

10311 10312 10313 10314 10315 10316 10317 10318 10319 10320 10321 10322 10323 10324 10325 10326 10327 10328 10329 10330 10331 10332 10333 10334 10335 10336
	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);
		}
10337
		if (!clamps) {
10338
			css = css_rightmost_descendant(css);
10339 10340 10341 10342
			continue;
		}

		/* Immediately update descendants RUNNABLE tasks */
10343
		uclamp_update_active_tasks(css);
10344 10345
	}
}
10346 10347 10348 10349 10350 10351 10352 10353 10354 10355 10356 10357 10358 10359 10360 10361 10362 10363 10364 10365 10366 10367 10368 10369 10370 10371 10372 10373 10374 10375 10376 10377

/*
 * 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;
10378
		if ((u64)req.percent > UCLAMP_PERCENT_SCALE) {
10379 10380 10381 10382 10383 10384 10385 10386 10387 10388 10389 10390 10391 10392 10393 10394 10395 10396 10397 10398 10399 10400
			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;

10401 10402
	static_branch_enable(&sched_uclamp_used);

10403 10404 10405 10406 10407 10408 10409 10410 10411 10412 10413 10414 10415
	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;

10416 10417 10418
	/* Update effective clamps to track the most restrictive value */
	cpu_util_update_eff(of_css(of));

10419 10420 10421 10422 10423 10424 10425 10426 10427 10428 10429 10430 10431 10432 10433 10434 10435 10436 10437 10438 10439 10440 10441 10442 10443 10444 10445 10446 10447 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
	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 */

10475
#ifdef CONFIG_FAIR_GROUP_SCHED
10476 10477
static int cpu_shares_write_u64(struct cgroup_subsys_state *css,
				struct cftype *cftype, u64 shareval)
10478
{
10479 10480
	if (shareval > scale_load_down(ULONG_MAX))
		shareval = MAX_SHARES;
10481
	return sched_group_set_shares(css_tg(css), scale_load(shareval));
10482 10483
}

10484 10485
static u64 cpu_shares_read_u64(struct cgroup_subsys_state *css,
			       struct cftype *cft)
10486
{
10487
	struct task_group *tg = css_tg(css);
10488

10489
	return (u64) scale_load_down(tg->shares);
10490
}
10491 10492

#ifdef CONFIG_CFS_BANDWIDTH
10493 10494
static DEFINE_MUTEX(cfs_constraints_mutex);

10495
const u64 max_cfs_quota_period = 1 * NSEC_PER_SEC; /* 1s */
10496
static const u64 min_cfs_quota_period = 1 * NSEC_PER_MSEC; /* 1ms */
10497 10498
/* More than 203 days if BW_SHIFT equals 20. */
static const u64 max_cfs_runtime = MAX_BW * NSEC_PER_USEC;
10499

10500 10501
static int __cfs_schedulable(struct task_group *tg, u64 period, u64 runtime);

10502 10503
static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota,
				u64 burst)
10504
{
10505
	int i, ret = 0, runtime_enabled, runtime_was_enabled;
10506
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
10507 10508 10509 10510 10511 10512 10513 10514 10515 10516 10517 10518 10519

	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
10520
	 * Likewise, bound things on the other side by preventing insane quota
10521 10522 10523 10524 10525 10526
	 * periods.  This also allows us to normalize in computing quota
	 * feasibility.
	 */
	if (period > max_cfs_quota_period)
		return -EINVAL;

10527 10528 10529 10530 10531 10532
	/*
	 * Bound quota to defend quota against overflow during bandwidth shift.
	 */
	if (quota != RUNTIME_INF && quota > max_cfs_runtime)
		return -EINVAL;

10533 10534 10535 10536
	if (quota != RUNTIME_INF && (burst > quota ||
				     burst + quota > max_cfs_runtime))
		return -EINVAL;

10537 10538 10539 10540
	/*
	 * Prevent race between setting of cfs_rq->runtime_enabled and
	 * unthrottle_offline_cfs_rqs().
	 */
10541
	cpus_read_lock();
10542 10543 10544 10545 10546
	mutex_lock(&cfs_constraints_mutex);
	ret = __cfs_schedulable(tg, period, quota);
	if (ret)
		goto out_unlock;

10547
	runtime_enabled = quota != RUNTIME_INF;
10548
	runtime_was_enabled = cfs_b->quota != RUNTIME_INF;
10549 10550 10551 10552 10553 10554
	/*
	 * 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();
10555 10556 10557
	raw_spin_lock_irq(&cfs_b->lock);
	cfs_b->period = ns_to_ktime(period);
	cfs_b->quota = quota;
10558
	cfs_b->burst = burst;
10559

Paul Turner's avatar
Paul Turner committed
10560
	__refill_cfs_bandwidth_runtime(cfs_b);
10561 10562

	/* Restart the period timer (if active) to handle new period expiry: */
10563 10564
	if (runtime_enabled)
		start_cfs_bandwidth(cfs_b);
10565

10566 10567
	raw_spin_unlock_irq(&cfs_b->lock);

10568
	for_each_online_cpu(i) {
10569
		struct cfs_rq *cfs_rq = tg->cfs_rq[i];
10570
		struct rq *rq = cfs_rq->rq;
10571
		struct rq_flags rf;
10572

10573
		rq_lock_irq(rq, &rf);
10574
		cfs_rq->runtime_enabled = runtime_enabled;
10575
		cfs_rq->runtime_remaining = 0;
10576

10577
		if (cfs_rq->throttled)
10578
			unthrottle_cfs_rq(cfs_rq);
10579
		rq_unlock_irq(rq, &rf);
10580
	}
10581 10582
	if (runtime_was_enabled && !runtime_enabled)
		cfs_bandwidth_usage_dec();
10583 10584
out_unlock:
	mutex_unlock(&cfs_constraints_mutex);
10585
	cpus_read_unlock();
10586

10587
	return ret;
10588 10589
}

10590
static int tg_set_cfs_quota(struct task_group *tg, long cfs_quota_us)
10591
{
10592
	u64 quota, period, burst;
10593

10594
	period = ktime_to_ns(tg->cfs_bandwidth.period);
10595
	burst = tg->cfs_bandwidth.burst;
10596 10597
	if (cfs_quota_us < 0)
		quota = RUNTIME_INF;
10598
	else if ((u64)cfs_quota_us <= U64_MAX / NSEC_PER_USEC)
10599
		quota = (u64)cfs_quota_us * NSEC_PER_USEC;
10600 10601
	else
		return -EINVAL;
10602

10603
	return tg_set_cfs_bandwidth(tg, period, quota, burst);
10604 10605
}

10606
static long tg_get_cfs_quota(struct task_group *tg)
10607 10608 10609
{
	u64 quota_us;

10610
	if (tg->cfs_bandwidth.quota == RUNTIME_INF)
10611 10612
		return -1;

10613
	quota_us = tg->cfs_bandwidth.quota;
10614 10615 10616 10617 10618
	do_div(quota_us, NSEC_PER_USEC);

	return quota_us;
}

10619
static int tg_set_cfs_period(struct task_group *tg, long cfs_period_us)
10620
{
10621
	u64 quota, period, burst;
10622

10623 10624 10625
	if ((u64)cfs_period_us > U64_MAX / NSEC_PER_USEC)
		return -EINVAL;

10626
	period = (u64)cfs_period_us * NSEC_PER_USEC;
10627
	quota = tg->cfs_bandwidth.quota;
10628
	burst = tg->cfs_bandwidth.burst;
10629

10630
	return tg_set_cfs_bandwidth(tg, period, quota, burst);
10631 10632
}

10633
static long tg_get_cfs_period(struct task_group *tg)
10634 10635 10636
{
	u64 cfs_period_us;

10637
	cfs_period_us = ktime_to_ns(tg->cfs_bandwidth.period);
10638 10639 10640 10641 10642
	do_div(cfs_period_us, NSEC_PER_USEC);

	return cfs_period_us;
}

10643 10644 10645 10646 10647 10648 10649 10650 10651 10652 10653 10654 10655 10656 10657 10658 10659 10660 10661 10662 10663 10664 10665 10666
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;
}

10667 10668
static s64 cpu_cfs_quota_read_s64(struct cgroup_subsys_state *css,
				  struct cftype *cft)
10669
{
10670
	return tg_get_cfs_quota(css_tg(css));
10671 10672
}

10673 10674
static int cpu_cfs_quota_write_s64(struct cgroup_subsys_state *css,
				   struct cftype *cftype, s64 cfs_quota_us)
10675
{
10676
	return tg_set_cfs_quota(css_tg(css), cfs_quota_us);
10677 10678
}

10679 10680
static u64 cpu_cfs_period_read_u64(struct cgroup_subsys_state *css,
				   struct cftype *cft)
10681
{
10682
	return tg_get_cfs_period(css_tg(css));
10683 10684
}

10685 10686
static int cpu_cfs_period_write_u64(struct cgroup_subsys_state *css,
				    struct cftype *cftype, u64 cfs_period_us)
10687
{
10688
	return tg_set_cfs_period(css_tg(css), cfs_period_us);
10689 10690
}

10691 10692 10693 10694 10695 10696 10697 10698 10699 10700 10701 10702
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);
}

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
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;
10735
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
10736 10737 10738 10739 10740
	s64 quota = 0, parent_quota = -1;

	if (!tg->parent) {
		quota = RUNTIME_INF;
	} else {
10741
		struct cfs_bandwidth *parent_b = &tg->parent->cfs_bandwidth;
10742 10743

		quota = normalize_cfs_quota(tg, d);
10744
		parent_quota = parent_b->hierarchical_quota;
10745 10746

		/*
10747 10748
		 * Ensure max(child_quota) <= parent_quota.  On cgroup2,
		 * always take the min.  On cgroup1, only inherit when no
10749
		 * limit is set:
10750
		 */
10751 10752 10753 10754 10755 10756 10757 10758
		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;
		}
10759
	}
10760
	cfs_b->hierarchical_quota = quota;
10761 10762 10763 10764 10765 10766

	return 0;
}

static int __cfs_schedulable(struct task_group *tg, u64 period, u64 quota)
{
10767
	int ret;
10768 10769 10770 10771 10772 10773 10774 10775 10776 10777 10778
	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);
	}

10779 10780 10781 10782 10783
	rcu_read_lock();
	ret = walk_tg_tree(tg_cfs_schedulable_down, tg_nop, &data);
	rcu_read_unlock();

	return ret;
10784
}
10785

10786
static int cpu_cfs_stat_show(struct seq_file *sf, void *v)
10787
{
10788
	struct task_group *tg = css_tg(seq_css(sf));
10789
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
10790

10791 10792 10793
	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);
10794

10795
	if (schedstat_enabled() && tg != &root_task_group) {
10796
		struct sched_statistics *stats;
10797 10798 10799
		u64 ws = 0;
		int i;

10800 10801 10802 10803
		for_each_possible_cpu(i) {
			stats = __schedstats_from_se(tg->se[i]);
			ws += schedstat_val(stats->wait_sum);
		}
10804 10805 10806 10807

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

10808 10809 10810
	seq_printf(sf, "nr_bursts %d\n", cfs_b->nr_burst);
	seq_printf(sf, "burst_time %llu\n", cfs_b->burst_time);

10811 10812
	return 0;
}
10813
#endif /* CONFIG_CFS_BANDWIDTH */
10814
#endif /* CONFIG_FAIR_GROUP_SCHED */
10815

10816
#ifdef CONFIG_RT_GROUP_SCHED
10817 10818
static int cpu_rt_runtime_write(struct cgroup_subsys_state *css,
				struct cftype *cft, s64 val)
Peter Zijlstra's avatar
Peter Zijlstra committed
10819
{
10820
	return sched_group_set_rt_runtime(css_tg(css), val);
Peter Zijlstra's avatar
Peter Zijlstra committed
10821 10822
}

10823 10824
static s64 cpu_rt_runtime_read(struct cgroup_subsys_state *css,
			       struct cftype *cft)
Peter Zijlstra's avatar
Peter Zijlstra committed
10825
{
10826
	return sched_group_rt_runtime(css_tg(css));
Peter Zijlstra's avatar
Peter Zijlstra committed
10827
}
10828

10829 10830
static int cpu_rt_period_write_uint(struct cgroup_subsys_state *css,
				    struct cftype *cftype, u64 rt_period_us)
10831
{
10832
	return sched_group_set_rt_period(css_tg(css), rt_period_us);
10833 10834
}

10835 10836
static u64 cpu_rt_period_read_uint(struct cgroup_subsys_state *css,
				   struct cftype *cft)
10837
{
10838
	return sched_group_rt_period(css_tg(css));
10839
}
10840
#endif /* CONFIG_RT_GROUP_SCHED */
Peter Zijlstra's avatar
Peter Zijlstra committed
10841

10842 10843 10844 10845 10846 10847 10848 10849 10850 10851 10852 10853 10854 10855
#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

10856
static struct cftype cpu_legacy_files[] = {
10857
#ifdef CONFIG_FAIR_GROUP_SCHED
10858 10859
	{
		.name = "shares",
10860 10861
		.read_u64 = cpu_shares_read_u64,
		.write_u64 = cpu_shares_write_u64,
10862
	},
10863 10864 10865 10866 10867
	{
		.name = "idle",
		.read_s64 = cpu_idle_read_s64,
		.write_s64 = cpu_idle_write_s64,
	},
10868
#endif
10869 10870 10871 10872 10873 10874 10875 10876 10877 10878 10879
#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,
	},
10880 10881 10882 10883 10884
	{
		.name = "cfs_burst_us",
		.read_u64 = cpu_cfs_burst_read_u64,
		.write_u64 = cpu_cfs_burst_write_u64,
	},
10885 10886
	{
		.name = "stat",
10887
		.seq_show = cpu_cfs_stat_show,
10888
	},
10889
#endif
10890
#ifdef CONFIG_RT_GROUP_SCHED
Peter Zijlstra's avatar
Peter Zijlstra committed
10891
	{
Peter Zijlstra's avatar
Peter Zijlstra committed
10892
		.name = "rt_runtime_us",
10893 10894
		.read_s64 = cpu_rt_runtime_read,
		.write_s64 = cpu_rt_runtime_write,
Peter Zijlstra's avatar
Peter Zijlstra committed
10895
	},
10896 10897
	{
		.name = "rt_period_us",
10898 10899
		.read_u64 = cpu_rt_period_read_uint,
		.write_u64 = cpu_rt_period_write_uint,
10900
	},
10901 10902 10903 10904 10905 10906 10907 10908 10909 10910 10911 10912 10913 10914
#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,
	},
10915
#endif
10916
	{ }	/* Terminate */
10917 10918
};

10919 10920
static int cpu_extra_stat_show(struct seq_file *sf,
			       struct cgroup_subsys_state *css)
10921 10922 10923
{
#ifdef CONFIG_CFS_BANDWIDTH
	{
10924
		struct task_group *tg = css_tg(css);
10925
		struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
10926
		u64 throttled_usec, burst_usec;
10927 10928 10929

		throttled_usec = cfs_b->throttled_time;
		do_div(throttled_usec, NSEC_PER_USEC);
10930 10931
		burst_usec = cfs_b->burst_time;
		do_div(burst_usec, NSEC_PER_USEC);
10932 10933 10934

		seq_printf(sf, "nr_periods %d\n"
			   "nr_throttled %d\n"
10935 10936 10937
			   "throttled_usec %llu\n"
			   "nr_bursts %d\n"
			   "burst_usec %llu\n",
10938
			   cfs_b->nr_periods, cfs_b->nr_throttled,
10939
			   throttled_usec, cfs_b->nr_burst, burst_usec);
10940 10941 10942 10943 10944 10945 10946 10947 10948 10949 10950 10951 10952 10953 10954 10955 10956 10957 10958 10959 10960 10961 10962 10963 10964 10965 10966 10967 10968 10969 10970 10971 10972 10973 10974 10975 10976 10977 10978 10979 10980 10981 10982 10983 10984 10985 10986 10987 10988 10989 10990 10991 10992 10993 10994
	}
#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;
10995
	int idx;
10996 10997 10998 10999

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

11000 11001 11002 11003
	idx = NICE_TO_PRIO(nice) - MAX_RT_PRIO;
	idx = array_index_nospec(idx, 40);
	weight = sched_prio_to_weight[idx];

11004 11005 11006 11007 11008 11009 11010 11011 11012 11013 11014 11015 11016 11017 11018 11019 11020 11021 11022 11023 11024
	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 */

11025
	if (sscanf(buf, "%20s %llu", tok, periodp) < 1)
11026 11027 11028 11029 11030 11031 11032 11033 11034 11035 11036 11037 11038 11039 11040 11041 11042 11043 11044 11045 11046 11047 11048 11049 11050 11051 11052 11053
		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);
11054
	u64 burst = tg_get_cfs_burst(tg);
11055 11056 11057 11058 11059
	u64 quota;
	int ret;

	ret = cpu_period_quota_parse(buf, &period, &quota);
	if (!ret)
11060
		ret = tg_set_cfs_bandwidth(tg, period, quota, burst);
11061 11062 11063 11064 11065 11066 11067 11068 11069 11070 11071 11072 11073 11074 11075 11076 11077 11078
	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,
	},
11079 11080 11081 11082 11083 11084
	{
		.name = "idle",
		.flags = CFTYPE_NOT_ON_ROOT,
		.read_s64 = cpu_idle_read_s64,
		.write_s64 = cpu_idle_write_s64,
	},
11085 11086 11087 11088 11089 11090 11091 11092
#endif
#ifdef CONFIG_CFS_BANDWIDTH
	{
		.name = "max",
		.flags = CFTYPE_NOT_ON_ROOT,
		.seq_show = cpu_max_show,
		.write = cpu_max_write,
	},
11093 11094 11095 11096 11097 11098
	{
		.name = "max.burst",
		.flags = CFTYPE_NOT_ON_ROOT,
		.read_u64 = cpu_cfs_burst_read_u64,
		.write_u64 = cpu_cfs_burst_write_u64,
	},
11099 11100 11101 11102 11103 11104 11105 11106 11107 11108 11109 11110 11111 11112
#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,
	},
11113 11114 11115 11116
#endif
	{ }	/* terminate */
};

11117
struct cgroup_subsys cpu_cgrp_subsys = {
11118
	.css_alloc	= cpu_cgroup_css_alloc,
11119
	.css_online	= cpu_cgroup_css_online,
11120
	.css_released	= cpu_cgroup_css_released,
11121
	.css_free	= cpu_cgroup_css_free,
11122
	.css_extra_stat_show = cpu_extra_stat_show,
11123
#ifdef CONFIG_RT_GROUP_SCHED
11124
	.can_attach	= cpu_cgroup_can_attach,
11125
#endif
11126
	.attach		= cpu_cgroup_attach,
11127
	.legacy_cftypes	= cpu_legacy_files,
11128
	.dfl_cftypes	= cpu_files,
11129
	.early_init	= true,
11130
	.threaded	= true,
11131 11132
};

11133
#endif	/* CONFIG_CGROUP_SCHED */
11134

11135 11136 11137 11138 11139
void dump_cpu_task(int cpu)
{
	pr_info("Task dump for CPU %d:\n", cpu);
	sched_show_task(cpu_curr(cpu));
}
11140 11141 11142 11143 11144 11145 11146 11147 11148 11149 11150 11151 11152 11153 11154 11155 11156 11157 11158 11159 11160 11161 11162 11163 11164 11165 11166 11167 11168 11169 11170 11171 11172 11173 11174 11175 11176 11177 11178 11179 11180

/*
 * 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,
};
11181

11182 11183 11184 11185
void call_trace_sched_update_nr_running(struct rq *rq, int count)
{
        trace_sched_update_nr_running_tp(rq, count);
}