core.c 205.3 KB
Newer Older
L
Linus Torvalds 已提交
1
/*
2
 *  kernel/sched/core.c
L
Linus Torvalds 已提交
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
 *
 *  Kernel scheduler and related syscalls
 *
 *  Copyright (C) 1991-2002  Linus Torvalds
 *
 *  1996-12-23  Modified by Dave Grothe to fix bugs in semaphores and
 *		make semaphores SMP safe
 *  1998-11-19	Implemented schedule_timeout() and related stuff
 *		by Andrea Arcangeli
 *  2002-01-04	New ultra-scalable O(1) scheduler by Ingo Molnar:
 *		hybrid priority-list and round-robin design with
 *		an array-switch method of distributing timeslices
 *		and per-CPU runqueues.  Cleanups and useful suggestions
 *		by Davide Libenzi, preemptible kernel bits by Robert Love.
 *  2003-09-03	Interactivity tuning by Con Kolivas.
 *  2004-04-02	Scheduler domains code by Nick Piggin
I
Ingo Molnar 已提交
19 20 21 22 23 24
 *  2007-04-15  Work begun on replacing all interactivity tuning with a
 *              fair scheduling design by Con Kolivas.
 *  2007-05-05  Load balancing (smp-nice) and other improvements
 *              by Peter Williams
 *  2007-05-06  Interactivity improvements to CFS by Mike Galbraith
 *  2007-07-01  Group scheduling enhancements by Srivatsa Vaddagiri
25 26
 *  2007-11-29  RT balancing improvements by Steven Rostedt, Gregory Haskins,
 *              Thomas Gleixner, Mike Kravetz
L
Linus Torvalds 已提交
27 28 29 30 31 32
 */

#include <linux/mm.h>
#include <linux/module.h>
#include <linux/nmi.h>
#include <linux/init.h>
33
#include <linux/uaccess.h>
L
Linus Torvalds 已提交
34 35 36
#include <linux/highmem.h>
#include <asm/mmu_context.h>
#include <linux/interrupt.h>
37
#include <linux/capability.h>
L
Linus Torvalds 已提交
38 39
#include <linux/completion.h>
#include <linux/kernel_stat.h>
40
#include <linux/debug_locks.h>
41
#include <linux/perf_event.h>
L
Linus Torvalds 已提交
42 43 44
#include <linux/security.h>
#include <linux/notifier.h>
#include <linux/profile.h>
45
#include <linux/freezer.h>
46
#include <linux/vmalloc.h>
L
Linus Torvalds 已提交
47 48
#include <linux/blkdev.h>
#include <linux/delay.h>
49
#include <linux/pid_namespace.h>
L
Linus Torvalds 已提交
50 51 52 53 54 55 56
#include <linux/smp.h>
#include <linux/threads.h>
#include <linux/timer.h>
#include <linux/rcupdate.h>
#include <linux/cpu.h>
#include <linux/cpuset.h>
#include <linux/percpu.h>
57
#include <linux/proc_fs.h>
L
Linus Torvalds 已提交
58
#include <linux/seq_file.h>
59
#include <linux/sysctl.h>
L
Linus Torvalds 已提交
60 61
#include <linux/syscalls.h>
#include <linux/times.h>
62
#include <linux/tsacct_kern.h>
63
#include <linux/kprobes.h>
64
#include <linux/delayacct.h>
65
#include <linux/unistd.h>
J
Jens Axboe 已提交
66
#include <linux/pagemap.h>
P
Peter Zijlstra 已提交
67
#include <linux/hrtimer.h>
R
Reynes Philippe 已提交
68
#include <linux/tick.h>
P
Peter Zijlstra 已提交
69 70
#include <linux/debugfs.h>
#include <linux/ctype.h>
71
#include <linux/ftrace.h>
72
#include <linux/slab.h>
73
#include <linux/init_task.h>
A
Al Viro 已提交
74
#include <linux/binfmts.h>
75
#include <linux/context_tracking.h>
76
#include <linux/compiler.h>
L
Linus Torvalds 已提交
77

78
#include <asm/switch_to.h>
79
#include <asm/tlb.h>
80
#include <asm/irq_regs.h>
81
#include <asm/mutex.h>
G
Glauber Costa 已提交
82 83 84
#ifdef CONFIG_PARAVIRT
#include <asm/paravirt.h>
#endif
L
Linus Torvalds 已提交
85

86
#include "sched.h"
87
#include "../workqueue_internal.h"
88
#include "../smpboot.h"
89

90
#define CREATE_TRACE_POINTS
91
#include <trace/events/sched.h>
92

93 94
DEFINE_MUTEX(sched_domains_mutex);
DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
95

96
static void update_rq_clock_task(struct rq *rq, s64 delta);
97

98
void update_rq_clock(struct rq *rq)
99
{
100
	s64 delta;
101

102 103 104
	lockdep_assert_held(&rq->lock);

	if (rq->clock_skip_update & RQCF_ACT_SKIP)
105
		return;
106

107
	delta = sched_clock_cpu(cpu_of(rq)) - rq->clock;
108 109
	if (delta < 0)
		return;
110 111
	rq->clock += delta;
	update_rq_clock_task(rq, delta);
112 113
}

I
Ingo Molnar 已提交
114 115 116
/*
 * Debugging: various feature bits
 */
P
Peter Zijlstra 已提交
117 118 119 120

#define SCHED_FEAT(name, enabled)	\
	(1UL << __SCHED_FEAT_##name) * enabled |

I
Ingo Molnar 已提交
121
const_debug unsigned int sysctl_sched_features =
122
#include "features.h"
P
Peter Zijlstra 已提交
123 124 125 126 127 128 129 130
	0;

#undef SCHED_FEAT

#ifdef CONFIG_SCHED_DEBUG
#define SCHED_FEAT(name, enabled)	\
	#name ,

131
static const char * const sched_feat_names[] = {
132
#include "features.h"
P
Peter Zijlstra 已提交
133 134 135 136
};

#undef SCHED_FEAT

L
Li Zefan 已提交
137
static int sched_feat_show(struct seq_file *m, void *v)
P
Peter Zijlstra 已提交
138 139 140
{
	int i;

141
	for (i = 0; i < __SCHED_FEAT_NR; i++) {
L
Li Zefan 已提交
142 143 144
		if (!(sysctl_sched_features & (1UL << i)))
			seq_puts(m, "NO_");
		seq_printf(m, "%s ", sched_feat_names[i]);
P
Peter Zijlstra 已提交
145
	}
L
Li Zefan 已提交
146
	seq_puts(m, "\n");
P
Peter Zijlstra 已提交
147

L
Li Zefan 已提交
148
	return 0;
P
Peter Zijlstra 已提交
149 150
}

151 152
#ifdef HAVE_JUMP_LABEL

153 154
#define jump_label_key__true  STATIC_KEY_INIT_TRUE
#define jump_label_key__false STATIC_KEY_INIT_FALSE
155 156 157 158

#define SCHED_FEAT(name, enabled)	\
	jump_label_key__##enabled ,

159
struct static_key sched_feat_keys[__SCHED_FEAT_NR] = {
160 161 162 163 164 165 166
#include "features.h"
};

#undef SCHED_FEAT

static void sched_feat_disable(int i)
{
167
	static_key_disable(&sched_feat_keys[i]);
168 169 170 171
}

static void sched_feat_enable(int i)
{
172
	static_key_enable(&sched_feat_keys[i]);
173 174 175 176 177 178
}
#else
static void sched_feat_disable(int i) { };
static void sched_feat_enable(int i) { };
#endif /* HAVE_JUMP_LABEL */

179
static int sched_feat_set(char *cmp)
P
Peter Zijlstra 已提交
180 181
{
	int i;
182
	int neg = 0;
P
Peter Zijlstra 已提交
183

H
Hillf Danton 已提交
184
	if (strncmp(cmp, "NO_", 3) == 0) {
P
Peter Zijlstra 已提交
185 186 187 188
		neg = 1;
		cmp += 3;
	}

189
	for (i = 0; i < __SCHED_FEAT_NR; i++) {
190
		if (strcmp(cmp, sched_feat_names[i]) == 0) {
191
			if (neg) {
P
Peter Zijlstra 已提交
192
				sysctl_sched_features &= ~(1UL << i);
193 194
				sched_feat_disable(i);
			} else {
P
Peter Zijlstra 已提交
195
				sysctl_sched_features |= (1UL << i);
196 197
				sched_feat_enable(i);
			}
P
Peter Zijlstra 已提交
198 199 200 201
			break;
		}
	}

202 203 204 205 206 207 208 209 210 211
	return i;
}

static ssize_t
sched_feat_write(struct file *filp, const char __user *ubuf,
		size_t cnt, loff_t *ppos)
{
	char buf[64];
	char *cmp;
	int i;
212
	struct inode *inode;
213 214 215 216 217 218 219 220 221 222

	if (cnt > 63)
		cnt = 63;

	if (copy_from_user(&buf, ubuf, cnt))
		return -EFAULT;

	buf[cnt] = 0;
	cmp = strstrip(buf);

223 224 225
	/* Ensure the static_key remains in a consistent state */
	inode = file_inode(filp);
	mutex_lock(&inode->i_mutex);
226
	i = sched_feat_set(cmp);
227
	mutex_unlock(&inode->i_mutex);
228
	if (i == __SCHED_FEAT_NR)
P
Peter Zijlstra 已提交
229 230
		return -EINVAL;

231
	*ppos += cnt;
P
Peter Zijlstra 已提交
232 233 234 235

	return cnt;
}

L
Li Zefan 已提交
236 237 238 239 240
static int sched_feat_open(struct inode *inode, struct file *filp)
{
	return single_open(filp, sched_feat_show, NULL);
}

241
static const struct file_operations sched_feat_fops = {
L
Li Zefan 已提交
242 243 244 245 246
	.open		= sched_feat_open,
	.write		= sched_feat_write,
	.read		= seq_read,
	.llseek		= seq_lseek,
	.release	= single_release,
P
Peter Zijlstra 已提交
247 248 249 250 251 252 253 254 255 256
};

static __init int sched_init_debug(void)
{
	debugfs_create_file("sched_features", 0644, NULL, NULL,
			&sched_feat_fops);

	return 0;
}
late_initcall(sched_init_debug);
257
#endif /* CONFIG_SCHED_DEBUG */
I
Ingo Molnar 已提交
258

259 260 261 262 263 264
/*
 * Number of tasks to iterate in a single balance run.
 * Limited because this is done with IRQs disabled.
 */
const_debug unsigned int sysctl_sched_nr_migrate = 32;

265 266 267 268 269 270 271 272
/*
 * period over which we average the RT time consumption, measured
 * in ms.
 *
 * default: 1s
 */
const_debug unsigned int sysctl_sched_time_avg = MSEC_PER_SEC;

P
Peter Zijlstra 已提交
273
/*
P
Peter Zijlstra 已提交
274
 * period over which we measure -rt task cpu usage in us.
P
Peter Zijlstra 已提交
275 276
 * default: 1s
 */
P
Peter Zijlstra 已提交
277
unsigned int sysctl_sched_rt_period = 1000000;
P
Peter Zijlstra 已提交
278

279
__read_mostly int scheduler_running;
280

P
Peter Zijlstra 已提交
281 282 283 284 285
/*
 * part of the period that we allow rt tasks to run in us.
 * default: 0.95s
 */
int sysctl_sched_rt_runtime = 950000;
P
Peter Zijlstra 已提交
286

287 288 289
/* cpus with isolated domains */
cpumask_var_t cpu_isolated_map;

L
Linus Torvalds 已提交
290
/*
291
 * this_rq_lock - lock this runqueue and disable interrupts.
L
Linus Torvalds 已提交
292
 */
A
Alexey Dobriyan 已提交
293
static struct rq *this_rq_lock(void)
L
Linus Torvalds 已提交
294 295
	__acquires(rq->lock)
{
296
	struct rq *rq;
L
Linus Torvalds 已提交
297 298 299

	local_irq_disable();
	rq = this_rq();
300
	raw_spin_lock(&rq->lock);
L
Linus Torvalds 已提交
301 302 303 304

	return rq;
}

P
Peter Zijlstra 已提交
305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325
#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);

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

326
	raw_spin_lock(&rq->lock);
327
	update_rq_clock(rq);
P
Peter Zijlstra 已提交
328
	rq->curr->sched_class->task_tick(rq, rq->curr, 1);
329
	raw_spin_unlock(&rq->lock);
P
Peter Zijlstra 已提交
330 331 332 333

	return HRTIMER_NORESTART;
}

334
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
335

336
static void __hrtick_restart(struct rq *rq)
P
Peter Zijlstra 已提交
337 338 339
{
	struct hrtimer *timer = &rq->hrtick_timer;

340
	hrtimer_start_expires(timer, HRTIMER_MODE_ABS_PINNED);
P
Peter Zijlstra 已提交
341 342
}

343 344 345 346
/*
 * called from hardirq (IPI) context
 */
static void __hrtick_start(void *arg)
347
{
348
	struct rq *rq = arg;
349

350
	raw_spin_lock(&rq->lock);
P
Peter Zijlstra 已提交
351
	__hrtick_restart(rq);
352
	rq->hrtick_csd_pending = 0;
353
	raw_spin_unlock(&rq->lock);
354 355
}

356 357 358 359 360
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
361
void hrtick_start(struct rq *rq, u64 delay)
362
{
363
	struct hrtimer *timer = &rq->hrtick_timer;
364 365 366 367 368 369 370 371 372
	ktime_t time;
	s64 delta;

	/*
	 * Don't schedule slices shorter than 10000ns, that just
	 * doesn't make sense and can cause timer DoS.
	 */
	delta = max_t(s64, delay, 10000LL);
	time = ktime_add_ns(timer->base->get_time(), delta);
373

374
	hrtimer_set_expires(timer, time);
375 376

	if (rq == this_rq()) {
P
Peter Zijlstra 已提交
377
		__hrtick_restart(rq);
378
	} else if (!rq->hrtick_csd_pending) {
379
		smp_call_function_single_async(cpu_of(rq), &rq->hrtick_csd);
380 381
		rq->hrtick_csd_pending = 1;
	}
382 383 384 385 386 387 388 389 390 391 392 393 394 395
}

static int
hotplug_hrtick(struct notifier_block *nfb, unsigned long action, void *hcpu)
{
	int cpu = (int)(long)hcpu;

	switch (action) {
	case CPU_UP_CANCELED:
	case CPU_UP_CANCELED_FROZEN:
	case CPU_DOWN_PREPARE:
	case CPU_DOWN_PREPARE_FROZEN:
	case CPU_DEAD:
	case CPU_DEAD_FROZEN:
396
		hrtick_clear(cpu_rq(cpu));
397 398 399 400 401 402
		return NOTIFY_OK;
	}

	return NOTIFY_DONE;
}

403
static __init void init_hrtick(void)
404 405 406
{
	hotcpu_notifier(hotplug_hrtick, 0);
}
407 408 409 410 411 412
#else
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
413
void hrtick_start(struct rq *rq, u64 delay)
414
{
W
Wanpeng Li 已提交
415 416 417 418 419
	/*
	 * Don't schedule slices shorter than 10000ns, that just
	 * doesn't make sense. Rely on vruntime for fairness.
	 */
	delay = max_t(u64, delay, 10000LL);
420 421
	hrtimer_start(&rq->hrtick_timer, ns_to_ktime(delay),
		      HRTIMER_MODE_REL_PINNED);
422
}
423

A
Andrew Morton 已提交
424
static inline void init_hrtick(void)
P
Peter Zijlstra 已提交
425 426
{
}
427
#endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
428

429
static void init_rq_hrtick(struct rq *rq)
P
Peter Zijlstra 已提交
430
{
431 432
#ifdef CONFIG_SMP
	rq->hrtick_csd_pending = 0;
P
Peter Zijlstra 已提交
433

434 435 436 437
	rq->hrtick_csd.flags = 0;
	rq->hrtick_csd.func = __hrtick_start;
	rq->hrtick_csd.info = rq;
#endif
P
Peter Zijlstra 已提交
438

439 440
	hrtimer_init(&rq->hrtick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	rq->hrtick_timer.function = hrtick;
P
Peter Zijlstra 已提交
441
}
A
Andrew Morton 已提交
442
#else	/* CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
443 444 445 446 447 448 449 450
static inline void hrtick_clear(struct rq *rq)
{
}

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

451 452 453
static inline void init_hrtick(void)
{
}
A
Andrew Morton 已提交
454
#endif	/* CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
455

456 457 458 459 460 461 462 463 464 465 466 467 468 469
/*
 * cmpxchg based fetch_or, macro so it works for different integer types
 */
#define fetch_or(ptr, val)						\
({	typeof(*(ptr)) __old, __val = *(ptr);				\
 	for (;;) {							\
 		__old = cmpxchg((ptr), __val, __val | (val));		\
 		if (__old == __val)					\
 			break;						\
 		__val = __old;						\
 	}								\
 	__old;								\
})

470
#if defined(CONFIG_SMP) && defined(TIF_POLLING_NRFLAG)
471 472 473 474 475 476 477 478 479 480
/*
 * Atomically set TIF_NEED_RESCHED and test for TIF_POLLING_NRFLAG,
 * this avoids any races wrt polling state changes and thereby avoids
 * spurious IPIs.
 */
static bool set_nr_and_not_polling(struct task_struct *p)
{
	struct thread_info *ti = task_thread_info(p);
	return !(fetch_or(&ti->flags, _TIF_NEED_RESCHED) & _TIF_POLLING_NRFLAG);
}
481 482 483 484 485 486 487 488 489 490

/*
 * 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);
491
	typeof(ti->flags) old, val = READ_ONCE(ti->flags);
492 493 494 495 496 497 498 499 500 501 502 503 504 505

	for (;;) {
		if (!(val & _TIF_POLLING_NRFLAG))
			return false;
		if (val & _TIF_NEED_RESCHED)
			return true;
		old = cmpxchg(&ti->flags, val, val | _TIF_NEED_RESCHED);
		if (old == val)
			break;
		val = old;
	}
	return true;
}

506 507 508 509 510 511
#else
static bool set_nr_and_not_polling(struct task_struct *p)
{
	set_tsk_need_resched(p);
	return true;
}
512 513 514 515 516 517 518

#ifdef CONFIG_SMP
static bool set_nr_if_polling(struct task_struct *p)
{
	return false;
}
#endif
519 520
#endif

521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566
void wake_q_add(struct wake_q_head *head, struct task_struct *task)
{
	struct wake_q_node *node = &task->wake_q;

	/*
	 * Atomically grab the task, if ->wake_q is !nil already it means
	 * its already queued (either by us or someone else) and will get the
	 * wakeup due to that.
	 *
	 * This cmpxchg() implies a full barrier, which pairs with the write
	 * barrier implied by the wakeup in wake_up_list().
	 */
	if (cmpxchg(&node->next, NULL, WAKE_Q_TAIL))
		return;

	get_task_struct(task);

	/*
	 * The head is context local, there can be no concurrency.
	 */
	*head->lastp = node;
	head->lastp = &node->next;
}

void wake_up_q(struct wake_q_head *head)
{
	struct wake_q_node *node = head->first;

	while (node != WAKE_Q_TAIL) {
		struct task_struct *task;

		task = container_of(node, struct task_struct, wake_q);
		BUG_ON(!task);
		/* task can safely be re-inserted now */
		node = node->next;
		task->wake_q.next = NULL;

		/*
		 * wake_up_process() implies a wmb() to pair with the queueing
		 * in wake_q_add() so as not to miss wakeups.
		 */
		wake_up_process(task);
		put_task_struct(task);
	}
}

I
Ingo Molnar 已提交
567
/*
568
 * resched_curr - mark rq's current task 'to be rescheduled now'.
I
Ingo Molnar 已提交
569 570 571 572 573
 *
 * 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.
 */
574
void resched_curr(struct rq *rq)
I
Ingo Molnar 已提交
575
{
576
	struct task_struct *curr = rq->curr;
I
Ingo Molnar 已提交
577 578
	int cpu;

579
	lockdep_assert_held(&rq->lock);
I
Ingo Molnar 已提交
580

581
	if (test_tsk_need_resched(curr))
I
Ingo Molnar 已提交
582 583
		return;

584
	cpu = cpu_of(rq);
585

586
	if (cpu == smp_processor_id()) {
587
		set_tsk_need_resched(curr);
588
		set_preempt_need_resched();
I
Ingo Molnar 已提交
589
		return;
590
	}
I
Ingo Molnar 已提交
591

592
	if (set_nr_and_not_polling(curr))
I
Ingo Molnar 已提交
593
		smp_send_reschedule(cpu);
594 595
	else
		trace_sched_wake_idle_without_ipi(cpu);
I
Ingo Molnar 已提交
596 597
}

598
void resched_cpu(int cpu)
I
Ingo Molnar 已提交
599 600 601 602
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

603
	if (!raw_spin_trylock_irqsave(&rq->lock, flags))
I
Ingo Molnar 已提交
604
		return;
605
	resched_curr(rq);
606
	raw_spin_unlock_irqrestore(&rq->lock, flags);
I
Ingo Molnar 已提交
607
}
608

609
#ifdef CONFIG_SMP
610
#ifdef CONFIG_NO_HZ_COMMON
611 612 613 614 615 616 617 618
/*
 * In the semi idle case, use the nearest busy cpu for migrating timers
 * from an idle cpu.  This is good for power-savings.
 *
 * We don't do similar optimization for completely idle system, as
 * 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).
 */
619
int get_nohz_timer_target(void)
620
{
621
	int i, cpu = smp_processor_id();
622 623
	struct sched_domain *sd;

624
	if (!idle_cpu(cpu) && is_housekeeping_cpu(cpu))
625 626
		return cpu;

627
	rcu_read_lock();
628
	for_each_domain(cpu, sd) {
629
		for_each_cpu(i, sched_domain_span(sd)) {
630
			if (!idle_cpu(i) && is_housekeeping_cpu(cpu)) {
631 632 633 634
				cpu = i;
				goto unlock;
			}
		}
635
	}
636 637 638

	if (!is_housekeeping_cpu(cpu))
		cpu = housekeeping_any_cpu();
639 640
unlock:
	rcu_read_unlock();
641 642
	return cpu;
}
643 644 645 646 647 648 649 650 651 652
/*
 * 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.
 */
653
static void wake_up_idle_cpu(int cpu)
654 655 656 657 658 659
{
	struct rq *rq = cpu_rq(cpu);

	if (cpu == smp_processor_id())
		return;

660
	if (set_nr_and_not_polling(rq->idle))
661
		smp_send_reschedule(cpu);
662 663
	else
		trace_sched_wake_idle_without_ipi(cpu);
664 665
}

666
static bool wake_up_full_nohz_cpu(int cpu)
667
{
668 669 670 671 672 673
	/*
	 * 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.
	 */
674
	if (tick_nohz_full_cpu(cpu)) {
675 676
		if (cpu != smp_processor_id() ||
		    tick_nohz_tick_stopped())
677
			tick_nohz_full_kick_cpu(cpu);
678 679 680 681 682 683 684 685
		return true;
	}

	return false;
}

void wake_up_nohz_cpu(int cpu)
{
686
	if (!wake_up_full_nohz_cpu(cpu))
687 688 689
		wake_up_idle_cpu(cpu);
}

690
static inline bool got_nohz_idle_kick(void)
691
{
692
	int cpu = smp_processor_id();
693 694 695 696 697 698 699 700 701 702 703 704 705

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

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

	/*
	 * We can't run Idle Load Balance on this CPU for this time so we
	 * cancel it and clear NOHZ_BALANCE_KICK
	 */
	clear_bit(NOHZ_BALANCE_KICK, nohz_flags(cpu));
	return false;
706 707
}

708
#else /* CONFIG_NO_HZ_COMMON */
709

710
static inline bool got_nohz_idle_kick(void)
P
Peter Zijlstra 已提交
711
{
712
	return false;
P
Peter Zijlstra 已提交
713 714
}

715
#endif /* CONFIG_NO_HZ_COMMON */
716

717 718 719
#ifdef CONFIG_NO_HZ_FULL
bool sched_can_stop_tick(void)
{
720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736
	/*
	 * FIFO realtime policy runs the highest priority task. Other runnable
	 * tasks are of a lower priority. The scheduler tick does nothing.
	 */
	if (current->policy == SCHED_FIFO)
		return true;

	/*
	 * Round-robin realtime tasks time slice with other tasks at the same
	 * realtime priority. Is this task the only one at this priority?
	 */
	if (current->policy == SCHED_RR) {
		struct sched_rt_entity *rt_se = &current->rt;

		return rt_se->run_list.prev == rt_se->run_list.next;
	}

737 738 739 740 741
	/*
	 * More than one running task need preemption.
	 * nr_running update is assumed to be visible
	 * after IPI is sent from wakers.
	 */
742 743
	if (this_rq()->nr_running > 1)
		return false;
744

745
	return true;
746 747
}
#endif /* CONFIG_NO_HZ_FULL */
748

749
void sched_avg_update(struct rq *rq)
750
{
751 752
	s64 period = sched_avg_period();

753
	while ((s64)(rq_clock(rq) - rq->age_stamp) > period) {
754 755 756 757 758 759
		/*
		 * Inline assembly required to prevent the compiler
		 * optimising this loop into a divmod call.
		 * See __iter_div_u64_rem() for another example of this.
		 */
		asm("" : "+rm" (rq->age_stamp));
760 761 762
		rq->age_stamp += period;
		rq->rt_avg /= 2;
	}
763 764
}

765
#endif /* CONFIG_SMP */
766

767 768
#if defined(CONFIG_RT_GROUP_SCHED) || (defined(CONFIG_FAIR_GROUP_SCHED) && \
			(defined(CONFIG_SMP) || defined(CONFIG_CFS_BANDWIDTH)))
769
/*
770 771 772 773
 * 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.
774
 */
775
int walk_tg_tree_from(struct task_group *from,
776
			     tg_visitor down, tg_visitor up, void *data)
777 778
{
	struct task_group *parent, *child;
P
Peter Zijlstra 已提交
779
	int ret;
780

781 782
	parent = from;

783
down:
P
Peter Zijlstra 已提交
784 785
	ret = (*down)(parent, data);
	if (ret)
786
		goto out;
787 788 789 790 791 792 793
	list_for_each_entry_rcu(child, &parent->children, siblings) {
		parent = child;
		goto down;

up:
		continue;
	}
P
Peter Zijlstra 已提交
794
	ret = (*up)(parent, data);
795 796
	if (ret || parent == from)
		goto out;
797 798 799 800 801

	child = parent;
	parent = parent->parent;
	if (parent)
		goto up;
802
out:
P
Peter Zijlstra 已提交
803
	return ret;
804 805
}

806
int tg_nop(struct task_group *tg, void *data)
P
Peter Zijlstra 已提交
807
{
808
	return 0;
P
Peter Zijlstra 已提交
809
}
810 811
#endif

812 813
static void set_load_weight(struct task_struct *p)
{
N
Nikhil Rao 已提交
814 815 816
	int prio = p->static_prio - MAX_RT_PRIO;
	struct load_weight *load = &p->se.load;

I
Ingo Molnar 已提交
817 818 819
	/*
	 * SCHED_IDLE tasks get minimal weight:
	 */
820
	if (idle_policy(p->policy)) {
821
		load->weight = scale_load(WEIGHT_IDLEPRIO);
N
Nikhil Rao 已提交
822
		load->inv_weight = WMULT_IDLEPRIO;
I
Ingo Molnar 已提交
823 824
		return;
	}
825

826
	load->weight = scale_load(prio_to_weight[prio]);
N
Nikhil Rao 已提交
827
	load->inv_weight = prio_to_wmult[prio];
828 829
}

830
static void enqueue_task(struct rq *rq, struct task_struct *p, int flags)
831
{
832
	update_rq_clock(rq);
833
	sched_info_queued(rq, p);
834
	p->sched_class->enqueue_task(rq, p, flags);
835 836
}

837
static void dequeue_task(struct rq *rq, struct task_struct *p, int flags)
838
{
839
	update_rq_clock(rq);
840
	sched_info_dequeued(rq, p);
841
	p->sched_class->dequeue_task(rq, p, flags);
842 843
}

844
void activate_task(struct rq *rq, struct task_struct *p, int flags)
845 846 847 848
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible--;

849
	enqueue_task(rq, p, flags);
850 851
}

852
void deactivate_task(struct rq *rq, struct task_struct *p, int flags)
853 854 855 856
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible++;

857
	dequeue_task(rq, p, flags);
858 859
}

860
static void update_rq_clock_task(struct rq *rq, s64 delta)
861
{
862 863 864 865 866 867 868 869
/*
 * In theory, the compile should just see 0 here, and optimize out the call
 * to sched_rt_avg_update. But I don't trust it...
 */
#if defined(CONFIG_IRQ_TIME_ACCOUNTING) || defined(CONFIG_PARAVIRT_TIME_ACCOUNTING)
	s64 steal = 0, irq_delta = 0;
#endif
#ifdef CONFIG_IRQ_TIME_ACCOUNTING
870
	irq_delta = irq_time_read(cpu_of(rq)) - rq->prev_irq_time;
871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891

	/*
	 * 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;
892 893
#endif
#ifdef CONFIG_PARAVIRT_TIME_ACCOUNTING
894
	if (static_key_false((&paravirt_steal_rq_enabled))) {
895 896 897 898 899 900 901 902 903 904 905
		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

906 907
	rq->clock_task += delta;

908
#if defined(CONFIG_IRQ_TIME_ACCOUNTING) || defined(CONFIG_PARAVIRT_TIME_ACCOUNTING)
909
	if ((irq_delta + steal) && sched_feat(NONTASK_CAPACITY))
910 911
		sched_rt_avg_update(rq, irq_delta + steal);
#endif
912 913
}

914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943
void sched_set_stop_task(int cpu, struct task_struct *stop)
{
	struct sched_param param = { .sched_priority = MAX_RT_PRIO - 1 };
	struct task_struct *old_stop = cpu_rq(cpu)->stop;

	if (stop) {
		/*
		 * Make it appear like a SCHED_FIFO task, its something
		 * userspace knows about and won't get confused about.
		 *
		 * Also, it will make PI more or less work without too
		 * much confusion -- but then, stop work should not
		 * rely on PI working anyway.
		 */
		sched_setscheduler_nocheck(stop, SCHED_FIFO, &param);

		stop->sched_class = &stop_sched_class;
	}

	cpu_rq(cpu)->stop = stop;

	if (old_stop) {
		/*
		 * Reset it back to a normal scheduling class so that
		 * it can die in pieces.
		 */
		old_stop->sched_class = &rt_sched_class;
	}
}

944
/*
I
Ingo Molnar 已提交
945
 * __normal_prio - return the priority that is based on the static prio
946 947 948
 */
static inline int __normal_prio(struct task_struct *p)
{
I
Ingo Molnar 已提交
949
	return p->static_prio;
950 951
}

952 953 954 955 956 957 958
/*
 * 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.
 */
959
static inline int normal_prio(struct task_struct *p)
960 961 962
{
	int prio;

963 964 965
	if (task_has_dl_policy(p))
		prio = MAX_DL_PRIO-1;
	else if (task_has_rt_policy(p))
966 967 968 969 970 971 972 973 974 975 976 977 978
		prio = MAX_RT_PRIO-1 - p->rt_priority;
	else
		prio = __normal_prio(p);
	return prio;
}

/*
 * Calculate the current priority, i.e. the priority
 * taken into account by the scheduler. This value might
 * be boosted by RT tasks, or might be boosted by
 * interactivity modifiers. Will be RT if the task got
 * RT-boosted. If not then it returns p->normal_prio.
 */
979
static int effective_prio(struct task_struct *p)
980 981 982 983 984 985 986 987 988 989 990 991
{
	p->normal_prio = normal_prio(p);
	/*
	 * If we are RT tasks or we were boosted to RT priority,
	 * keep the priority unchanged. Otherwise, update priority
	 * to the normal priority:
	 */
	if (!rt_prio(p->prio))
		return p->normal_prio;
	return p->prio;
}

L
Linus Torvalds 已提交
992 993 994
/**
 * task_curr - is this task currently executing on a CPU?
 * @p: the task in question.
995 996
 *
 * Return: 1 if the task is currently executing. 0 otherwise.
L
Linus Torvalds 已提交
997
 */
998
inline int task_curr(const struct task_struct *p)
L
Linus Torvalds 已提交
999 1000 1001 1002
{
	return cpu_curr(task_cpu(p)) == p;
}

1003
/*
1004 1005 1006 1007 1008
 * 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().
1009
 */
1010 1011
static inline void check_class_changed(struct rq *rq, struct task_struct *p,
				       const struct sched_class *prev_class,
P
Peter Zijlstra 已提交
1012
				       int oldprio)
1013 1014 1015
{
	if (prev_class != p->sched_class) {
		if (prev_class->switched_from)
P
Peter Zijlstra 已提交
1016
			prev_class->switched_from(rq, p);
1017

P
Peter Zijlstra 已提交
1018
		p->sched_class->switched_to(rq, p);
1019
	} else if (oldprio != p->prio || dl_task(p))
P
Peter Zijlstra 已提交
1020
		p->sched_class->prio_changed(rq, p, oldprio);
1021 1022
}

1023
void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags)
1024 1025 1026 1027 1028 1029 1030 1031 1032 1033
{
	const struct sched_class *class;

	if (p->sched_class == rq->curr->sched_class) {
		rq->curr->sched_class->check_preempt_curr(rq, p, flags);
	} else {
		for_each_class(class) {
			if (class == rq->curr->sched_class)
				break;
			if (class == p->sched_class) {
1034
				resched_curr(rq);
1035 1036 1037 1038 1039 1040 1041 1042 1043
				break;
			}
		}
	}

	/*
	 * A queue event has occurred, and we're going to schedule.  In
	 * this case, we can save a useless back to back clock update.
	 */
1044
	if (task_on_rq_queued(rq->curr) && test_tsk_need_resched(rq->curr))
1045
		rq_clock_skip_update(rq, true);
1046 1047
}

L
Linus Torvalds 已提交
1048
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067
/*
 * 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.
 */
1068
static struct rq *move_queued_task(struct rq *rq, struct task_struct *p, int new_cpu)
P
Peter Zijlstra 已提交
1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101
{
	lockdep_assert_held(&rq->lock);

	dequeue_task(rq, p, 0);
	p->on_rq = TASK_ON_RQ_MIGRATING;
	set_task_cpu(p, new_cpu);
	raw_spin_unlock(&rq->lock);

	rq = cpu_rq(new_cpu);

	raw_spin_lock(&rq->lock);
	BUG_ON(task_cpu(p) != new_cpu);
	p->on_rq = TASK_ON_RQ_QUEUED;
	enqueue_task(rq, p, 0);
	check_preempt_curr(rq, p, 0);

	return rq;
}

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

/*
 * Move (not current) task off this cpu, onto dest cpu. We're doing
 * 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.
 */
1102
static struct rq *__migrate_task(struct rq *rq, struct task_struct *p, int dest_cpu)
P
Peter Zijlstra 已提交
1103 1104
{
	if (unlikely(!cpu_active(dest_cpu)))
1105
		return rq;
P
Peter Zijlstra 已提交
1106 1107 1108

	/* Affinity changed (again). */
	if (!cpumask_test_cpu(dest_cpu, tsk_cpus_allowed(p)))
1109
		return rq;
P
Peter Zijlstra 已提交
1110

1111 1112 1113
	rq = move_queued_task(rq, p, dest_cpu);

	return rq;
P
Peter Zijlstra 已提交
1114 1115 1116 1117 1118 1119 1120 1121 1122 1123
}

/*
 * 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;
1124 1125
	struct task_struct *p = arg->task;
	struct rq *rq = this_rq();
P
Peter Zijlstra 已提交
1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137

	/*
	 * The original target cpu might have gone down and we might
	 * be on another cpu but it doesn't matter.
	 */
	local_irq_disable();
	/*
	 * We need to explicitly wake pending tasks before running
	 * __migrate_task() such that we will not miss enforcing cpus_allowed
	 * during wakeups, see set_cpus_allowed_ptr()'s TASK_WAKING test.
	 */
	sched_ttwu_pending();
1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150

	raw_spin_lock(&p->pi_lock);
	raw_spin_lock(&rq->lock);
	/*
	 * 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.
	 */
	if (task_rq(p) == rq && task_on_rq_queued(p))
		rq = __migrate_task(rq, p, arg->dest_cpu);
	raw_spin_unlock(&rq->lock);
	raw_spin_unlock(&p->pi_lock);

P
Peter Zijlstra 已提交
1151 1152 1153 1154
	local_irq_enable();
	return 0;
}

1155 1156 1157 1158 1159
/*
 * sched_class::set_cpus_allowed must do the below, but is not required to
 * actually call this function.
 */
void set_cpus_allowed_common(struct task_struct *p, const struct cpumask *new_mask)
P
Peter Zijlstra 已提交
1160 1161 1162 1163 1164
{
	cpumask_copy(&p->cpus_allowed, new_mask);
	p->nr_cpus_allowed = cpumask_weight(new_mask);
}

1165 1166
void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
{
1167 1168 1169
	struct rq *rq = task_rq(p);
	bool queued, running;

1170
	lockdep_assert_held(&p->pi_lock);
1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185

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

	if (queued) {
		/*
		 * Because __kthread_bind() calls this on blocked tasks without
		 * holding rq->lock.
		 */
		lockdep_assert_held(&rq->lock);
		dequeue_task(rq, p, 0);
	}
	if (running)
		put_prev_task(rq, p);

1186
	p->sched_class->set_cpus_allowed(p, new_mask);
1187 1188 1189 1190 1191

	if (running)
		p->sched_class->set_curr_task(rq);
	if (queued)
		enqueue_task(rq, p, 0);
1192 1193
}

P
Peter Zijlstra 已提交
1194 1195 1196 1197 1198 1199 1200 1201 1202
/*
 * 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.
 */
1203 1204
static int __set_cpus_allowed_ptr(struct task_struct *p,
				  const struct cpumask *new_mask, bool check)
P
Peter Zijlstra 已提交
1205 1206 1207 1208 1209 1210 1211 1212
{
	unsigned long flags;
	struct rq *rq;
	unsigned int dest_cpu;
	int ret = 0;

	rq = task_rq_lock(p, &flags);

1213 1214 1215 1216 1217 1218 1219 1220 1221
	/*
	 * Must re-check here, to close a race against __kthread_bind(),
	 * sched_setaffinity() is not guaranteed to observe the flag.
	 */
	if (check && (p->flags & PF_NO_SETAFFINITY)) {
		ret = -EINVAL;
		goto out;
	}

P
Peter Zijlstra 已提交
1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243
	if (cpumask_equal(&p->cpus_allowed, new_mask))
		goto out;

	if (!cpumask_intersects(new_mask, cpu_active_mask)) {
		ret = -EINVAL;
		goto out;
	}

	do_set_cpus_allowed(p, new_mask);

	/* Can the task run on the task's current CPU? If so, we're done */
	if (cpumask_test_cpu(task_cpu(p), new_mask))
		goto out;

	dest_cpu = cpumask_any_and(cpu_active_mask, new_mask);
	if (task_running(rq, p) || p->state == TASK_WAKING) {
		struct migration_arg arg = { p, dest_cpu };
		/* Need help from migration thread: drop lock and wait. */
		task_rq_unlock(rq, p, &flags);
		stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg);
		tlb_migrate_finish(p->mm);
		return 0;
1244 1245 1246 1247 1248 1249
	} else if (task_on_rq_queued(p)) {
		/*
		 * OK, since we're going to drop the lock immediately
		 * afterwards anyway.
		 */
		lockdep_unpin_lock(&rq->lock);
1250
		rq = move_queued_task(rq, p, dest_cpu);
1251 1252
		lockdep_pin_lock(&rq->lock);
	}
P
Peter Zijlstra 已提交
1253 1254 1255 1256 1257
out:
	task_rq_unlock(rq, p, &flags);

	return ret;
}
1258 1259 1260 1261 1262

int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
{
	return __set_cpus_allowed_ptr(p, new_mask, false);
}
P
Peter Zijlstra 已提交
1263 1264
EXPORT_SYMBOL_GPL(set_cpus_allowed_ptr);

I
Ingo Molnar 已提交
1265
void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
I
Ingo Molnar 已提交
1266
{
1267 1268 1269 1270 1271
#ifdef CONFIG_SCHED_DEBUG
	/*
	 * We should never call set_task_cpu() on a blocked task,
	 * ttwu() will sort out the placement.
	 */
P
Peter Zijlstra 已提交
1272
	WARN_ON_ONCE(p->state != TASK_RUNNING && p->state != TASK_WAKING &&
O
Oleg Nesterov 已提交
1273
			!p->on_rq);
1274 1275

#ifdef CONFIG_LOCKDEP
1276 1277 1278 1279 1280
	/*
	 * The caller should hold either p->pi_lock or rq->lock, when changing
	 * a task's CPU. ->pi_lock for waking tasks, rq->lock for runnable tasks.
	 *
	 * sched_move_task() holds both and thus holding either pins the cgroup,
P
Peter Zijlstra 已提交
1281
	 * see task_group().
1282 1283 1284 1285
	 *
	 * Furthermore, all task_rq users should acquire both locks, see
	 * task_rq_lock().
	 */
1286 1287 1288
	WARN_ON_ONCE(debug_locks && !(lockdep_is_held(&p->pi_lock) ||
				      lockdep_is_held(&task_rq(p)->lock)));
#endif
1289 1290
#endif

1291
	trace_sched_migrate_task(p, new_cpu);
1292

1293
	if (task_cpu(p) != new_cpu) {
1294 1295
		if (p->sched_class->migrate_task_rq)
			p->sched_class->migrate_task_rq(p, new_cpu);
1296
		p->se.nr_migrations++;
1297
		perf_event_task_migrate(p);
1298
	}
I
Ingo Molnar 已提交
1299 1300

	__set_task_cpu(p, new_cpu);
I
Ingo Molnar 已提交
1301 1302
}

1303 1304
static void __migrate_swap_task(struct task_struct *p, int cpu)
{
1305
	if (task_on_rq_queued(p)) {
1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338
		struct rq *src_rq, *dst_rq;

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

		deactivate_task(src_rq, p, 0);
		set_task_cpu(p, cpu);
		activate_task(dst_rq, p, 0);
		check_preempt_curr(dst_rq, p, 0);
	} 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
		 * previous cpu our targer instead of where it really is.
		 */
		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;

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

1339 1340
	double_raw_lock(&arg->src_task->pi_lock,
			&arg->dst_task->pi_lock);
1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360
	double_rq_lock(src_rq, dst_rq);
	if (task_cpu(arg->dst_task) != arg->dst_cpu)
		goto unlock;

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

	if (!cpumask_test_cpu(arg->dst_cpu, tsk_cpus_allowed(arg->src_task)))
		goto unlock;

	if (!cpumask_test_cpu(arg->src_cpu, tsk_cpus_allowed(arg->dst_task)))
		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);
1361 1362
	raw_spin_unlock(&arg->dst_task->pi_lock);
	raw_spin_unlock(&arg->src_task->pi_lock);
1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384

	return ret;
}

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

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

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

1385 1386 1387 1388
	/*
	 * 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.
	 */
1389 1390 1391 1392 1393 1394 1395 1396 1397
	if (!cpu_active(arg.src_cpu) || !cpu_active(arg.dst_cpu))
		goto out;

	if (!cpumask_test_cpu(arg.dst_cpu, tsk_cpus_allowed(arg.src_task)))
		goto out;

	if (!cpumask_test_cpu(arg.src_cpu, tsk_cpus_allowed(arg.dst_task)))
		goto out;

1398
	trace_sched_swap_numa(cur, arg.src_cpu, p, arg.dst_cpu);
1399 1400 1401 1402 1403 1404
	ret = stop_two_cpus(arg.dst_cpu, arg.src_cpu, migrate_swap_stop, &arg);

out:
	return ret;
}

L
Linus Torvalds 已提交
1405 1406 1407
/*
 * wait_task_inactive - wait for a thread to unschedule.
 *
R
Roland McGrath 已提交
1408 1409 1410 1411 1412 1413 1414
 * If @match_state is nonzero, it's the @p->state value just checked and
 * not expected to change.  If it changes, i.e. @p might have woken up,
 * then return zero.  When we succeed in waiting for @p to be off its CPU,
 * we return a positive number (its total switch count).  If a second call
 * a short while later returns the same number, the caller can be sure that
 * @p has remained unscheduled the whole time.
 *
L
Linus Torvalds 已提交
1415 1416 1417 1418 1419 1420
 * The caller must ensure that the task *will* unschedule sometime soon,
 * else this function might spin for a *long* time. This function can't
 * be called with interrupts off, or it may introduce deadlock with
 * smp_call_function() if an IPI is sent by the same process we are
 * waiting to become inactive.
 */
R
Roland McGrath 已提交
1421
unsigned long wait_task_inactive(struct task_struct *p, long match_state)
L
Linus Torvalds 已提交
1422 1423
{
	unsigned long flags;
1424
	int running, queued;
R
Roland McGrath 已提交
1425
	unsigned long ncsw;
1426
	struct rq *rq;
L
Linus Torvalds 已提交
1427

1428 1429 1430 1431 1432 1433 1434 1435
	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);
1436

1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447
		/*
		 * If the task is actively running on another CPU
		 * still, just relax and busy-wait without holding
		 * any locks.
		 *
		 * NOTE! Since we don't hold any locks, it's not
		 * even sure that "rq" stays as the right runqueue!
		 * But we don't care, since "task_running()" will
		 * return false if the runqueue has changed and p
		 * is actually now running somewhere else!
		 */
R
Roland McGrath 已提交
1448 1449 1450
		while (task_running(rq, p)) {
			if (match_state && unlikely(p->state != match_state))
				return 0;
1451
			cpu_relax();
R
Roland McGrath 已提交
1452
		}
1453

1454 1455 1456 1457 1458 1459
		/*
		 * 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.
		 */
		rq = task_rq_lock(p, &flags);
1460
		trace_sched_wait_task(p);
1461
		running = task_running(rq, p);
1462
		queued = task_on_rq_queued(p);
R
Roland McGrath 已提交
1463
		ncsw = 0;
1464
		if (!match_state || p->state == match_state)
1465
			ncsw = p->nvcsw | LONG_MIN; /* sets MSB */
1466
		task_rq_unlock(rq, p, &flags);
1467

R
Roland McGrath 已提交
1468 1469 1470 1471 1472 1473
		/*
		 * If it changed from the expected state, bail out now.
		 */
		if (unlikely(!ncsw))
			break;

1474 1475 1476 1477 1478 1479 1480 1481 1482 1483
		/*
		 * 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;
		}
1484

1485 1486 1487 1488 1489
		/*
		 * It's not enough that it's not actively running,
		 * it must be off the runqueue _entirely_, and not
		 * preempted!
		 *
1490
		 * So if it was still runnable (but just not actively
1491 1492 1493
		 * running right now), it's preempted, and we should
		 * yield - it could be a while.
		 */
1494
		if (unlikely(queued)) {
1495 1496 1497 1498
			ktime_t to = ktime_set(0, NSEC_PER_SEC/HZ);

			set_current_state(TASK_UNINTERRUPTIBLE);
			schedule_hrtimeout(&to, HRTIMER_MODE_REL);
1499 1500
			continue;
		}
1501

1502 1503 1504 1505 1506 1507 1508
		/*
		 * Ahh, all good. It wasn't running, and it wasn't
		 * runnable, which means that it will never become
		 * running in the future either. We're all done!
		 */
		break;
	}
R
Roland McGrath 已提交
1509 1510

	return ncsw;
L
Linus Torvalds 已提交
1511 1512 1513 1514 1515 1516 1517 1518 1519
}

/***
 * kick_process - kick a running thread to enter/exit the kernel
 * @p: the to-be-kicked thread
 *
 * Cause a process which is running on another CPU to enter
 * kernel-mode, without any delay. (to get signals handled.)
 *
L
Lucas De Marchi 已提交
1520
 * NOTE: this function doesn't have to take the runqueue lock,
L
Linus Torvalds 已提交
1521 1522 1523 1524 1525
 * 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.
 */
1526
void kick_process(struct task_struct *p)
L
Linus Torvalds 已提交
1527 1528 1529 1530 1531 1532 1533 1534 1535
{
	int cpu;

	preempt_disable();
	cpu = task_cpu(p);
	if ((cpu != smp_processor_id()) && task_curr(p))
		smp_send_reschedule(cpu);
	preempt_enable();
}
R
Rusty Russell 已提交
1536
EXPORT_SYMBOL_GPL(kick_process);
L
Linus Torvalds 已提交
1537

1538
/*
1539
 * ->cpus_allowed is protected by both rq->lock and p->pi_lock
1540
 */
1541 1542
static int select_fallback_rq(int cpu, struct task_struct *p)
{
1543 1544
	int nid = cpu_to_node(cpu);
	const struct cpumask *nodemask = NULL;
1545 1546
	enum { cpuset, possible, fail } state = cpuset;
	int dest_cpu;
1547

1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564
	/*
	 * 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.
	 */
	if (nid != -1) {
		nodemask = cpumask_of_node(nid);

		/* Look for allowed, online CPU in same node. */
		for_each_cpu(dest_cpu, nodemask) {
			if (!cpu_online(dest_cpu))
				continue;
			if (!cpu_active(dest_cpu))
				continue;
			if (cpumask_test_cpu(dest_cpu, tsk_cpus_allowed(p)))
				return dest_cpu;
		}
1565
	}
1566

1567 1568
	for (;;) {
		/* Any allowed, online CPU? */
1569
		for_each_cpu(dest_cpu, tsk_cpus_allowed(p)) {
1570 1571 1572 1573 1574 1575
			if (!cpu_online(dest_cpu))
				continue;
			if (!cpu_active(dest_cpu))
				continue;
			goto out;
		}
1576

1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602
		switch (state) {
		case cpuset:
			/* No more Mr. Nice Guy. */
			cpuset_cpus_allowed_fallback(p);
			state = possible;
			break;

		case possible:
			do_set_cpus_allowed(p, cpu_possible_mask);
			state = fail;
			break;

		case fail:
			BUG();
			break;
		}
	}

out:
	if (state != cpuset) {
		/*
		 * Don't tell them about moving exiting tasks or
		 * kernel threads (both mm NULL), since they never
		 * leave kernel.
		 */
		if (p->mm && printk_ratelimit()) {
1603
			printk_deferred("process %d (%s) no longer affine to cpu%d\n",
1604 1605
					task_pid_nr(p), p->comm, cpu);
		}
1606 1607 1608 1609 1610
	}

	return dest_cpu;
}

1611
/*
1612
 * The caller (fork, wakeup) owns p->pi_lock, ->cpus_allowed is stable.
1613
 */
1614
static inline
1615
int select_task_rq(struct task_struct *p, int cpu, int sd_flags, int wake_flags)
1616
{
1617 1618
	lockdep_assert_held(&p->pi_lock);

1619 1620
	if (p->nr_cpus_allowed > 1)
		cpu = p->sched_class->select_task_rq(p, cpu, sd_flags, wake_flags);
1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631

	/*
	 * In order not to call set_task_cpu() on a blocking task we need
	 * to rely on ttwu() to place the task on a valid ->cpus_allowed
	 * cpu.
	 *
	 * 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 ]
	 */
1632
	if (unlikely(!cpumask_test_cpu(cpu, tsk_cpus_allowed(p)) ||
P
Peter Zijlstra 已提交
1633
		     !cpu_online(cpu)))
1634
		cpu = select_fallback_rq(task_cpu(p), p);
1635 1636

	return cpu;
1637
}
1638 1639 1640 1641 1642 1643

static void update_avg(u64 *avg, u64 sample)
{
	s64 diff = sample - *avg;
	*avg += diff >> 3;
}
1644 1645 1646 1647 1648 1649 1650 1651 1652

#else

static inline int __set_cpus_allowed_ptr(struct task_struct *p,
					 const struct cpumask *new_mask, bool check)
{
	return set_cpus_allowed_ptr(p, new_mask);
}

P
Peter Zijlstra 已提交
1653
#endif /* CONFIG_SMP */
1654

P
Peter Zijlstra 已提交
1655
static void
1656
ttwu_stat(struct task_struct *p, int cpu, int wake_flags)
T
Tejun Heo 已提交
1657
{
P
Peter Zijlstra 已提交
1658
#ifdef CONFIG_SCHEDSTATS
1659 1660
	struct rq *rq = this_rq();

P
Peter Zijlstra 已提交
1661 1662 1663 1664 1665 1666 1667 1668 1669 1670
#ifdef CONFIG_SMP
	int this_cpu = smp_processor_id();

	if (cpu == this_cpu) {
		schedstat_inc(rq, ttwu_local);
		schedstat_inc(p, se.statistics.nr_wakeups_local);
	} else {
		struct sched_domain *sd;

		schedstat_inc(p, se.statistics.nr_wakeups_remote);
1671
		rcu_read_lock();
P
Peter Zijlstra 已提交
1672 1673 1674 1675 1676 1677
		for_each_domain(this_cpu, sd) {
			if (cpumask_test_cpu(cpu, sched_domain_span(sd))) {
				schedstat_inc(sd, ttwu_wake_remote);
				break;
			}
		}
1678
		rcu_read_unlock();
P
Peter Zijlstra 已提交
1679
	}
1680 1681 1682 1683

	if (wake_flags & WF_MIGRATED)
		schedstat_inc(p, se.statistics.nr_wakeups_migrate);

P
Peter Zijlstra 已提交
1684 1685 1686
#endif /* CONFIG_SMP */

	schedstat_inc(rq, ttwu_count);
T
Tejun Heo 已提交
1687
	schedstat_inc(p, se.statistics.nr_wakeups);
P
Peter Zijlstra 已提交
1688 1689

	if (wake_flags & WF_SYNC)
T
Tejun Heo 已提交
1690
		schedstat_inc(p, se.statistics.nr_wakeups_sync);
P
Peter Zijlstra 已提交
1691 1692 1693 1694 1695 1696

#endif /* CONFIG_SCHEDSTATS */
}

static void ttwu_activate(struct rq *rq, struct task_struct *p, int en_flags)
{
T
Tejun Heo 已提交
1697
	activate_task(rq, p, en_flags);
1698
	p->on_rq = TASK_ON_RQ_QUEUED;
1699 1700 1701 1702

	/* if a worker is waking up, notify workqueue */
	if (p->flags & PF_WQ_WORKER)
		wq_worker_waking_up(p, cpu_of(rq));
T
Tejun Heo 已提交
1703 1704
}

1705 1706 1707
/*
 * Mark the task runnable and perform wakeup-preemption.
 */
1708
static void
1709
ttwu_do_wakeup(struct rq *rq, struct task_struct *p, int wake_flags)
T
Tejun Heo 已提交
1710 1711 1712
{
	check_preempt_curr(rq, p, wake_flags);
	p->state = TASK_RUNNING;
1713 1714
	trace_sched_wakeup(p);

T
Tejun Heo 已提交
1715
#ifdef CONFIG_SMP
1716 1717
	if (p->sched_class->task_woken) {
		/*
1718 1719
		 * Our task @p is fully woken up and running; so its safe to
		 * drop the rq->lock, hereafter rq is only used for statistics.
1720
		 */
1721
		lockdep_unpin_lock(&rq->lock);
T
Tejun Heo 已提交
1722
		p->sched_class->task_woken(rq, p);
1723
		lockdep_pin_lock(&rq->lock);
1724
	}
T
Tejun Heo 已提交
1725

1726
	if (rq->idle_stamp) {
1727
		u64 delta = rq_clock(rq) - rq->idle_stamp;
1728
		u64 max = 2*rq->max_idle_balance_cost;
T
Tejun Heo 已提交
1729

1730 1731 1732
		update_avg(&rq->avg_idle, delta);

		if (rq->avg_idle > max)
T
Tejun Heo 已提交
1733
			rq->avg_idle = max;
1734

T
Tejun Heo 已提交
1735 1736 1737 1738 1739
		rq->idle_stamp = 0;
	}
#endif
}

1740 1741 1742
static void
ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags)
{
1743 1744
	lockdep_assert_held(&rq->lock);

1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765
#ifdef CONFIG_SMP
	if (p->sched_contributes_to_load)
		rq->nr_uninterruptible--;
#endif

	ttwu_activate(rq, p, ENQUEUE_WAKEUP | ENQUEUE_WAKING);
	ttwu_do_wakeup(rq, p, wake_flags);
}

/*
 * Called in case the task @p isn't fully descheduled from its runqueue,
 * in this case we must do a remote wakeup. Its a 'light' wakeup though,
 * since all we need to do is flip p->state to TASK_RUNNING, since
 * the task is still ->on_rq.
 */
static int ttwu_remote(struct task_struct *p, int wake_flags)
{
	struct rq *rq;
	int ret = 0;

	rq = __task_rq_lock(p);
1766
	if (task_on_rq_queued(p)) {
1767 1768
		/* check_preempt_curr() may use rq clock */
		update_rq_clock(rq);
1769 1770 1771 1772 1773 1774 1775 1776
		ttwu_do_wakeup(rq, p, wake_flags);
		ret = 1;
	}
	__task_rq_unlock(rq);

	return ret;
}

1777
#ifdef CONFIG_SMP
1778
void sched_ttwu_pending(void)
1779 1780
{
	struct rq *rq = this_rq();
P
Peter Zijlstra 已提交
1781 1782
	struct llist_node *llist = llist_del_all(&rq->wake_list);
	struct task_struct *p;
1783
	unsigned long flags;
1784

1785 1786 1787 1788
	if (!llist)
		return;

	raw_spin_lock_irqsave(&rq->lock, flags);
1789
	lockdep_pin_lock(&rq->lock);
1790

P
Peter Zijlstra 已提交
1791 1792 1793
	while (llist) {
		p = llist_entry(llist, struct task_struct, wake_entry);
		llist = llist_next(llist);
1794 1795 1796
		ttwu_do_activate(rq, p, 0);
	}

1797
	lockdep_unpin_lock(&rq->lock);
1798
	raw_spin_unlock_irqrestore(&rq->lock, flags);
1799 1800 1801 1802
}

void scheduler_ipi(void)
{
1803 1804 1805 1806 1807
	/*
	 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
	 * TIF_NEED_RESCHED remotely (for the first time) will also send
	 * this IPI.
	 */
1808
	preempt_fold_need_resched();
1809

1810
	if (llist_empty(&this_rq()->wake_list) && !got_nohz_idle_kick())
1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826
		return;

	/*
	 * Not all reschedule IPI handlers call irq_enter/irq_exit, since
	 * traditionally all their work was done from the interrupt return
	 * path. Now that we actually do some work, we need to make sure
	 * we do call them.
	 *
	 * Some archs already do call them, luckily irq_enter/exit nest
	 * properly.
	 *
	 * Arguably we should visit all archs and update all handlers,
	 * however a fair share of IPIs are still resched only so this would
	 * somewhat pessimize the simple resched case.
	 */
	irq_enter();
P
Peter Zijlstra 已提交
1827
	sched_ttwu_pending();
1828 1829 1830 1831

	/*
	 * Check if someone kicked us for doing the nohz idle load balance.
	 */
1832
	if (unlikely(got_nohz_idle_kick())) {
1833
		this_rq()->idle_balance = 1;
1834
		raise_softirq_irqoff(SCHED_SOFTIRQ);
1835
	}
1836
	irq_exit();
1837 1838 1839 1840
}

static void ttwu_queue_remote(struct task_struct *p, int cpu)
{
1841 1842 1843 1844 1845 1846 1847 1848
	struct rq *rq = cpu_rq(cpu);

	if (llist_add(&p->wake_entry, &cpu_rq(cpu)->wake_list)) {
		if (!set_nr_if_polling(rq->idle))
			smp_send_reschedule(cpu);
		else
			trace_sched_wake_idle_without_ipi(cpu);
	}
1849
}
1850

1851 1852 1853 1854 1855
void wake_up_if_idle(int cpu)
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

1856 1857 1858 1859
	rcu_read_lock();

	if (!is_idle_task(rcu_dereference(rq->curr)))
		goto out;
1860 1861 1862 1863 1864 1865 1866 1867 1868 1869

	if (set_nr_if_polling(rq->idle)) {
		trace_sched_wake_idle_without_ipi(cpu);
	} else {
		raw_spin_lock_irqsave(&rq->lock, flags);
		if (is_idle_task(rq->curr))
			smp_send_reschedule(cpu);
		/* Else cpu is not in idle, do nothing here */
		raw_spin_unlock_irqrestore(&rq->lock, flags);
	}
1870 1871 1872

out:
	rcu_read_unlock();
1873 1874
}

1875
bool cpus_share_cache(int this_cpu, int that_cpu)
1876 1877 1878
{
	return per_cpu(sd_llc_id, this_cpu) == per_cpu(sd_llc_id, that_cpu);
}
1879
#endif /* CONFIG_SMP */
1880

1881 1882 1883 1884
static void ttwu_queue(struct task_struct *p, int cpu)
{
	struct rq *rq = cpu_rq(cpu);

1885
#if defined(CONFIG_SMP)
1886
	if (sched_feat(TTWU_QUEUE) && !cpus_share_cache(smp_processor_id(), cpu)) {
1887
		sched_clock_cpu(cpu); /* sync clocks x-cpu */
1888 1889 1890 1891 1892
		ttwu_queue_remote(p, cpu);
		return;
	}
#endif

1893
	raw_spin_lock(&rq->lock);
1894
	lockdep_pin_lock(&rq->lock);
1895
	ttwu_do_activate(rq, p, 0);
1896
	lockdep_unpin_lock(&rq->lock);
1897
	raw_spin_unlock(&rq->lock);
T
Tejun Heo 已提交
1898 1899 1900
}

/**
L
Linus Torvalds 已提交
1901
 * try_to_wake_up - wake up a thread
T
Tejun Heo 已提交
1902
 * @p: the thread to be awakened
L
Linus Torvalds 已提交
1903
 * @state: the mask of task states that can be woken
T
Tejun Heo 已提交
1904
 * @wake_flags: wake modifier flags (WF_*)
L
Linus Torvalds 已提交
1905 1906 1907 1908 1909 1910 1911
 *
 * Put it on the run-queue if it's not already there. The "current"
 * thread is always on the run-queue (except when the actual
 * re-schedule is in progress), and as such you're allowed to do
 * the simpler "current->state = TASK_RUNNING" to mark yourself
 * runnable without the overhead of this.
 *
1912
 * Return: %true if @p was woken up, %false if it was already running.
T
Tejun Heo 已提交
1913
 * or @state didn't match @p's state.
L
Linus Torvalds 已提交
1914
 */
1915 1916
static int
try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
L
Linus Torvalds 已提交
1917 1918
{
	unsigned long flags;
1919
	int cpu, success = 0;
P
Peter Zijlstra 已提交
1920

1921 1922 1923 1924 1925 1926 1927
	/*
	 * If we are going to wake up a thread waiting for CONDITION we
	 * need to ensure that CONDITION=1 done by the caller can not be
	 * reordered with p->state check below. This pairs with mb() in
	 * set_current_state() the waiting thread does.
	 */
	smp_mb__before_spinlock();
1928
	raw_spin_lock_irqsave(&p->pi_lock, flags);
P
Peter Zijlstra 已提交
1929
	if (!(p->state & state))
L
Linus Torvalds 已提交
1930 1931
		goto out;

1932 1933
	trace_sched_waking(p);

1934
	success = 1; /* we're going to change ->state */
L
Linus Torvalds 已提交
1935 1936
	cpu = task_cpu(p);

1937 1938
	if (p->on_rq && ttwu_remote(p, wake_flags))
		goto stat;
L
Linus Torvalds 已提交
1939 1940

#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
1941
	/*
1942 1943
	 * If the owning (remote) cpu is still in the middle of schedule() with
	 * this task as prev, wait until its done referencing the task.
P
Peter Zijlstra 已提交
1944
	 */
1945
	while (p->on_cpu)
1946
		cpu_relax();
1947
	/*
1948
	 * Pairs with the smp_wmb() in finish_lock_switch().
1949
	 */
1950
	smp_rmb();
L
Linus Torvalds 已提交
1951

1952
	p->sched_contributes_to_load = !!task_contributes_to_load(p);
P
Peter Zijlstra 已提交
1953
	p->state = TASK_WAKING;
1954

1955
	if (p->sched_class->task_waking)
1956
		p->sched_class->task_waking(p);
1957

1958
	cpu = select_task_rq(p, p->wake_cpu, SD_BALANCE_WAKE, wake_flags);
1959 1960
	if (task_cpu(p) != cpu) {
		wake_flags |= WF_MIGRATED;
1961
		set_task_cpu(p, cpu);
1962
	}
L
Linus Torvalds 已提交
1963 1964
#endif /* CONFIG_SMP */

1965 1966
	ttwu_queue(p, cpu);
stat:
1967
	ttwu_stat(p, cpu, wake_flags);
L
Linus Torvalds 已提交
1968
out:
1969
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
1970 1971 1972 1973

	return success;
}

T
Tejun Heo 已提交
1974 1975 1976 1977
/**
 * try_to_wake_up_local - try to wake up a local task with rq lock held
 * @p: the thread to be awakened
 *
1978
 * Put @p on the run-queue if it's not already there. The caller must
T
Tejun Heo 已提交
1979
 * ensure that this_rq() is locked, @p is bound to this_rq() and not
1980
 * the current task.
T
Tejun Heo 已提交
1981 1982 1983 1984 1985
 */
static void try_to_wake_up_local(struct task_struct *p)
{
	struct rq *rq = task_rq(p);

1986 1987 1988 1989
	if (WARN_ON_ONCE(rq != this_rq()) ||
	    WARN_ON_ONCE(p == current))
		return;

T
Tejun Heo 已提交
1990 1991
	lockdep_assert_held(&rq->lock);

1992
	if (!raw_spin_trylock(&p->pi_lock)) {
1993 1994 1995 1996 1997 1998 1999
		/*
		 * This is OK, because current is on_cpu, which avoids it being
		 * picked for load-balance and preemption/IRQs are still
		 * disabled avoiding further scheduler activity on it and we've
		 * not yet picked a replacement task.
		 */
		lockdep_unpin_lock(&rq->lock);
2000 2001 2002
		raw_spin_unlock(&rq->lock);
		raw_spin_lock(&p->pi_lock);
		raw_spin_lock(&rq->lock);
2003
		lockdep_pin_lock(&rq->lock);
2004 2005
	}

T
Tejun Heo 已提交
2006
	if (!(p->state & TASK_NORMAL))
2007
		goto out;
T
Tejun Heo 已提交
2008

2009 2010
	trace_sched_waking(p);

2011
	if (!task_on_rq_queued(p))
P
Peter Zijlstra 已提交
2012 2013
		ttwu_activate(rq, p, ENQUEUE_WAKEUP);

2014
	ttwu_do_wakeup(rq, p, 0);
2015
	ttwu_stat(p, smp_processor_id(), 0);
2016 2017
out:
	raw_spin_unlock(&p->pi_lock);
T
Tejun Heo 已提交
2018 2019
}

2020 2021 2022 2023 2024
/**
 * 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
2025 2026 2027
 * processes.
 *
 * Return: 1 if the process was woken up, 0 if it was already running.
2028 2029 2030 2031
 *
 * It may be assumed that this function implies a write memory barrier before
 * changing the task state if and only if any tasks are woken up.
 */
2032
int wake_up_process(struct task_struct *p)
L
Linus Torvalds 已提交
2033
{
2034 2035
	WARN_ON(task_is_stopped_or_traced(p));
	return try_to_wake_up(p, TASK_NORMAL, 0);
L
Linus Torvalds 已提交
2036 2037 2038
}
EXPORT_SYMBOL(wake_up_process);

2039
int wake_up_state(struct task_struct *p, unsigned int state)
L
Linus Torvalds 已提交
2040 2041 2042 2043
{
	return try_to_wake_up(p, state, 0);
}

2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055
/*
 * This function clears the sched_dl_entity static params.
 */
void __dl_clear_params(struct task_struct *p)
{
	struct sched_dl_entity *dl_se = &p->dl;

	dl_se->dl_runtime = 0;
	dl_se->dl_deadline = 0;
	dl_se->dl_period = 0;
	dl_se->flags = 0;
	dl_se->dl_bw = 0;
2056 2057 2058 2059

	dl_se->dl_throttled = 0;
	dl_se->dl_new = 1;
	dl_se->dl_yielded = 0;
2060 2061
}

L
Linus Torvalds 已提交
2062 2063 2064
/*
 * Perform scheduler related setup for a newly forked process p.
 * p is forked by current.
I
Ingo Molnar 已提交
2065 2066 2067
 *
 * __sched_fork() is basic setup used by init_idle() too:
 */
2068
static void __sched_fork(unsigned long clone_flags, struct task_struct *p)
I
Ingo Molnar 已提交
2069
{
P
Peter Zijlstra 已提交
2070 2071 2072
	p->on_rq			= 0;

	p->se.on_rq			= 0;
I
Ingo Molnar 已提交
2073 2074
	p->se.exec_start		= 0;
	p->se.sum_exec_runtime		= 0;
2075
	p->se.prev_sum_exec_runtime	= 0;
2076
	p->se.nr_migrations		= 0;
P
Peter Zijlstra 已提交
2077
	p->se.vruntime			= 0;
P
Peter Zijlstra 已提交
2078
	INIT_LIST_HEAD(&p->se.group_node);
I
Ingo Molnar 已提交
2079 2080

#ifdef CONFIG_SCHEDSTATS
2081
	memset(&p->se.statistics, 0, sizeof(p->se.statistics));
I
Ingo Molnar 已提交
2082
#endif
N
Nick Piggin 已提交
2083

2084
	RB_CLEAR_NODE(&p->dl.rb_node);
2085
	init_dl_task_timer(&p->dl);
2086
	__dl_clear_params(p);
2087

P
Peter Zijlstra 已提交
2088
	INIT_LIST_HEAD(&p->rt.run_list);
N
Nick Piggin 已提交
2089

2090 2091 2092
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&p->preempt_notifiers);
#endif
2093 2094 2095

#ifdef CONFIG_NUMA_BALANCING
	if (p->mm && atomic_read(&p->mm->mm_users) == 1) {
2096
		p->mm->numa_next_scan = jiffies + msecs_to_jiffies(sysctl_numa_balancing_scan_delay);
2097 2098 2099
		p->mm->numa_scan_seq = 0;
	}

2100 2101 2102 2103 2104
	if (clone_flags & CLONE_VM)
		p->numa_preferred_nid = current->numa_preferred_nid;
	else
		p->numa_preferred_nid = -1;

2105 2106
	p->node_stamp = 0ULL;
	p->numa_scan_seq = p->mm ? p->mm->numa_scan_seq : 0;
2107
	p->numa_scan_period = sysctl_numa_balancing_scan_delay;
2108
	p->numa_work.next = &p->numa_work;
2109
	p->numa_faults = NULL;
2110 2111
	p->last_task_numa_placement = 0;
	p->last_sum_exec_runtime = 0;
2112 2113

	p->numa_group = NULL;
2114
#endif /* CONFIG_NUMA_BALANCING */
I
Ingo Molnar 已提交
2115 2116
}

2117 2118
DEFINE_STATIC_KEY_FALSE(sched_numa_balancing);

2119
#ifdef CONFIG_NUMA_BALANCING
2120

2121 2122
void set_numabalancing_state(bool enabled)
{
2123 2124 2125 2126
	if (enabled)
		static_branch_enable(&sched_numa_balancing);
	else
		static_branch_disable(&sched_numa_balancing);
2127
}
2128 2129 2130 2131 2132 2133 2134

#ifdef CONFIG_PROC_SYSCTL
int sysctl_numa_balancing(struct ctl_table *table, int write,
			 void __user *buffer, size_t *lenp, loff_t *ppos)
{
	struct ctl_table t;
	int err;
2135
	int state = static_branch_likely(&sched_numa_balancing);
2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150

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

	t = *table;
	t.data = &state;
	err = proc_dointvec_minmax(&t, write, buffer, lenp, ppos);
	if (err < 0)
		return err;
	if (write)
		set_numabalancing_state(state);
	return err;
}
#endif
#endif
I
Ingo Molnar 已提交
2151 2152 2153 2154

/*
 * fork()/clone()-time setup:
 */
2155
int sched_fork(unsigned long clone_flags, struct task_struct *p)
I
Ingo Molnar 已提交
2156
{
2157
	unsigned long flags;
I
Ingo Molnar 已提交
2158 2159
	int cpu = get_cpu();

2160
	__sched_fork(clone_flags, p);
2161
	/*
2162
	 * We mark the process as running here. This guarantees that
2163 2164 2165
	 * nobody will actually run it, and a signal or other external
	 * event cannot wake it up and insert it on the runqueue either.
	 */
2166
	p->state = TASK_RUNNING;
I
Ingo Molnar 已提交
2167

2168 2169 2170 2171 2172
	/*
	 * Make sure we do not leak PI boosting priority to the child.
	 */
	p->prio = current->normal_prio;

2173 2174 2175 2176
	/*
	 * Revert to default priority/policy on fork if requested.
	 */
	if (unlikely(p->sched_reset_on_fork)) {
2177
		if (task_has_dl_policy(p) || task_has_rt_policy(p)) {
2178
			p->policy = SCHED_NORMAL;
2179
			p->static_prio = NICE_TO_PRIO(0);
2180 2181 2182 2183 2184 2185
			p->rt_priority = 0;
		} else if (PRIO_TO_NICE(p->static_prio) < 0)
			p->static_prio = NICE_TO_PRIO(0);

		p->prio = p->normal_prio = __normal_prio(p);
		set_load_weight(p);
2186

2187 2188 2189 2190 2191 2192
		/*
		 * We don't need the reset flag anymore after the fork. It has
		 * fulfilled its duty:
		 */
		p->sched_reset_on_fork = 0;
	}
2193

2194 2195 2196 2197 2198 2199
	if (dl_prio(p->prio)) {
		put_cpu();
		return -EAGAIN;
	} else if (rt_prio(p->prio)) {
		p->sched_class = &rt_sched_class;
	} else {
H
Hiroshi Shimamoto 已提交
2200
		p->sched_class = &fair_sched_class;
2201
	}
2202

P
Peter Zijlstra 已提交
2203 2204 2205
	if (p->sched_class->task_fork)
		p->sched_class->task_fork(p);

2206 2207 2208 2209 2210 2211 2212
	/*
	 * The child is not yet in the pid-hash so no cgroup attach races,
	 * and the cgroup is pinned to this child due to cgroup_fork()
	 * is ran before sched_fork().
	 *
	 * Silence PROVE_RCU.
	 */
2213
	raw_spin_lock_irqsave(&p->pi_lock, flags);
2214
	set_task_cpu(p, cpu);
2215
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
2216

2217
#ifdef CONFIG_SCHED_INFO
I
Ingo Molnar 已提交
2218
	if (likely(sched_info_on()))
2219
		memset(&p->sched_info, 0, sizeof(p->sched_info));
L
Linus Torvalds 已提交
2220
#endif
P
Peter Zijlstra 已提交
2221 2222
#if defined(CONFIG_SMP)
	p->on_cpu = 0;
2223
#endif
2224
	init_task_preempt_count(p);
2225
#ifdef CONFIG_SMP
2226
	plist_node_init(&p->pushable_tasks, MAX_PRIO);
2227
	RB_CLEAR_NODE(&p->pushable_dl_tasks);
2228
#endif
2229

N
Nick Piggin 已提交
2230
	put_cpu();
2231
	return 0;
L
Linus Torvalds 已提交
2232 2233
}

2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252
unsigned long to_ratio(u64 period, u64 runtime)
{
	if (runtime == RUNTIME_INF)
		return 1ULL << 20;

	/*
	 * 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;

	return div64_u64(runtime << 20, period);
}

#ifdef CONFIG_SMP
inline struct dl_bw *dl_bw_of(int i)
{
2253 2254
	RCU_LOCKDEP_WARN(!rcu_read_lock_sched_held(),
			 "sched RCU must be held");
2255 2256 2257
	return &cpu_rq(i)->rd->dl_bw;
}

2258
static inline int dl_bw_cpus(int i)
2259
{
2260 2261 2262
	struct root_domain *rd = cpu_rq(i)->rd;
	int cpus = 0;

2263 2264
	RCU_LOCKDEP_WARN(!rcu_read_lock_sched_held(),
			 "sched RCU must be held");
2265 2266 2267 2268
	for_each_cpu_and(i, rd->span, cpu_active_mask)
		cpus++;

	return cpus;
2269 2270 2271 2272 2273 2274 2275
}
#else
inline struct dl_bw *dl_bw_of(int i)
{
	return &cpu_rq(i)->dl.dl_bw;
}

2276
static inline int dl_bw_cpus(int i)
2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288
{
	return 1;
}
#endif

/*
 * We must be sure that accepting a new task (or allowing changing the
 * parameters of an existing one) is consistent with the bandwidth
 * constraints. If yes, this function also accordingly updates the currently
 * allocated bandwidth to reflect the new situation.
 *
 * This function is called while holding p's rq->lock.
2289 2290 2291
 *
 * XXX we should delay bw change until the task's 0-lag point, see
 * __setparam_dl().
2292 2293 2294 2295 2296 2297
 */
static int dl_overflow(struct task_struct *p, int policy,
		       const struct sched_attr *attr)
{

	struct dl_bw *dl_b = dl_bw_of(task_cpu(p));
2298
	u64 period = attr->sched_period ?: attr->sched_deadline;
2299 2300
	u64 runtime = attr->sched_runtime;
	u64 new_bw = dl_policy(policy) ? to_ratio(period, runtime) : 0;
2301
	int cpus, err = -1;
2302 2303 2304 2305 2306 2307 2308 2309 2310 2311

	if (new_bw == p->dl.dl_bw)
		return 0;

	/*
	 * Either if a task, enters, leave, or stays -deadline but changes
	 * its parameters, we may need to update accordingly the total
	 * allocated bandwidth of the container.
	 */
	raw_spin_lock(&dl_b->lock);
2312
	cpus = dl_bw_cpus(task_cpu(p));
2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332
	if (dl_policy(policy) && !task_has_dl_policy(p) &&
	    !__dl_overflow(dl_b, cpus, 0, new_bw)) {
		__dl_add(dl_b, new_bw);
		err = 0;
	} else if (dl_policy(policy) && task_has_dl_policy(p) &&
		   !__dl_overflow(dl_b, cpus, p->dl.dl_bw, new_bw)) {
		__dl_clear(dl_b, p->dl.dl_bw);
		__dl_add(dl_b, new_bw);
		err = 0;
	} else if (!dl_policy(policy) && task_has_dl_policy(p)) {
		__dl_clear(dl_b, p->dl.dl_bw);
		err = 0;
	}
	raw_spin_unlock(&dl_b->lock);

	return err;
}

extern void init_dl_bw(struct dl_bw *dl_b);

L
Linus Torvalds 已提交
2333 2334 2335 2336 2337 2338 2339
/*
 * 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.
 */
2340
void wake_up_new_task(struct task_struct *p)
L
Linus Torvalds 已提交
2341 2342
{
	unsigned long flags;
I
Ingo Molnar 已提交
2343
	struct rq *rq;
2344

2345
	raw_spin_lock_irqsave(&p->pi_lock, flags);
2346 2347
	/* Initialize new task's runnable average */
	init_entity_runnable_average(&p->se);
2348 2349 2350 2351 2352 2353
#ifdef CONFIG_SMP
	/*
	 * Fork balancing, do it here and not earlier because:
	 *  - cpus_allowed can change in the fork path
	 *  - any previously selected cpu might disappear through hotplug
	 */
2354
	set_task_cpu(p, select_task_rq(p, task_cpu(p), SD_BALANCE_FORK, 0));
2355 2356
#endif

2357
	rq = __task_rq_lock(p);
P
Peter Zijlstra 已提交
2358
	activate_task(rq, p, 0);
2359
	p->on_rq = TASK_ON_RQ_QUEUED;
2360
	trace_sched_wakeup_new(p);
P
Peter Zijlstra 已提交
2361
	check_preempt_curr(rq, p, WF_FORK);
2362
#ifdef CONFIG_SMP
2363 2364
	if (p->sched_class->task_woken)
		p->sched_class->task_woken(rq, p);
2365
#endif
2366
	task_rq_unlock(rq, p, &flags);
L
Linus Torvalds 已提交
2367 2368
}

2369 2370
#ifdef CONFIG_PREEMPT_NOTIFIERS

2371 2372
static struct static_key preempt_notifier_key = STATIC_KEY_INIT_FALSE;

2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384
void preempt_notifier_inc(void)
{
	static_key_slow_inc(&preempt_notifier_key);
}
EXPORT_SYMBOL_GPL(preempt_notifier_inc);

void preempt_notifier_dec(void)
{
	static_key_slow_dec(&preempt_notifier_key);
}
EXPORT_SYMBOL_GPL(preempt_notifier_dec);

2385
/**
2386
 * preempt_notifier_register - tell me when current is being preempted & rescheduled
R
Randy Dunlap 已提交
2387
 * @notifier: notifier struct to register
2388 2389 2390
 */
void preempt_notifier_register(struct preempt_notifier *notifier)
{
2391 2392 2393
	if (!static_key_false(&preempt_notifier_key))
		WARN(1, "registering preempt_notifier while notifiers disabled\n");

2394 2395 2396 2397 2398 2399
	hlist_add_head(&notifier->link, &current->preempt_notifiers);
}
EXPORT_SYMBOL_GPL(preempt_notifier_register);

/**
 * preempt_notifier_unregister - no longer interested in preemption notifications
R
Randy Dunlap 已提交
2400
 * @notifier: notifier struct to unregister
2401
 *
2402
 * This is *not* safe to call from within a preemption notifier.
2403 2404 2405 2406 2407 2408 2409
 */
void preempt_notifier_unregister(struct preempt_notifier *notifier)
{
	hlist_del(&notifier->link);
}
EXPORT_SYMBOL_GPL(preempt_notifier_unregister);

2410
static void __fire_sched_in_preempt_notifiers(struct task_struct *curr)
2411 2412 2413
{
	struct preempt_notifier *notifier;

2414
	hlist_for_each_entry(notifier, &curr->preempt_notifiers, link)
2415 2416 2417
		notifier->ops->sched_in(notifier, raw_smp_processor_id());
}

2418 2419 2420 2421 2422 2423
static __always_inline void fire_sched_in_preempt_notifiers(struct task_struct *curr)
{
	if (static_key_false(&preempt_notifier_key))
		__fire_sched_in_preempt_notifiers(curr);
}

2424
static void
2425 2426
__fire_sched_out_preempt_notifiers(struct task_struct *curr,
				   struct task_struct *next)
2427 2428 2429
{
	struct preempt_notifier *notifier;

2430
	hlist_for_each_entry(notifier, &curr->preempt_notifiers, link)
2431 2432 2433
		notifier->ops->sched_out(notifier, next);
}

2434 2435 2436 2437 2438 2439 2440 2441
static __always_inline void
fire_sched_out_preempt_notifiers(struct task_struct *curr,
				 struct task_struct *next)
{
	if (static_key_false(&preempt_notifier_key))
		__fire_sched_out_preempt_notifiers(curr, next);
}

2442
#else /* !CONFIG_PREEMPT_NOTIFIERS */
2443

2444
static inline void fire_sched_in_preempt_notifiers(struct task_struct *curr)
2445 2446 2447
{
}

2448
static inline void
2449 2450 2451 2452 2453
fire_sched_out_preempt_notifiers(struct task_struct *curr,
				 struct task_struct *next)
{
}

2454
#endif /* CONFIG_PREEMPT_NOTIFIERS */
2455

2456 2457 2458
/**
 * prepare_task_switch - prepare to switch tasks
 * @rq: the runqueue preparing to switch
R
Randy Dunlap 已提交
2459
 * @prev: the current task that is being switched out
2460 2461 2462 2463 2464 2465 2466 2467 2468
 * @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.
 */
2469 2470 2471
static inline void
prepare_task_switch(struct rq *rq, struct task_struct *prev,
		    struct task_struct *next)
2472
{
2473
	sched_info_switch(rq, prev, next);
2474
	perf_event_task_sched_out(prev, next);
2475
	fire_sched_out_preempt_notifiers(prev, next);
2476 2477 2478 2479
	prepare_lock_switch(rq, next);
	prepare_arch_switch(next);
}

L
Linus Torvalds 已提交
2480 2481 2482 2483
/**
 * finish_task_switch - clean up after a task-switch
 * @prev: the thread we just switched away from.
 *
2484 2485 2486 2487
 * finish_task_switch must be called after the context switch, paired
 * with a prepare_task_switch call before the context switch.
 * finish_task_switch will reconcile locking set up by prepare_task_switch,
 * and do any other architecture-specific cleanup actions.
L
Linus Torvalds 已提交
2488 2489
 *
 * Note that we may have delayed dropping an mm in context_switch(). If
I
Ingo Molnar 已提交
2490
 * so, we finish that here outside of the runqueue lock. (Doing it
L
Linus Torvalds 已提交
2491 2492
 * with the lock held can cause deadlocks; see schedule() for
 * details.)
2493 2494 2495 2496 2497
 *
 * The context switch have flipped the stack from under us and restored the
 * local variables which were saved when this task called schedule() in the
 * past. prev == current is still correct but we need to recalculate this_rq
 * because prev may have moved to another CPU.
L
Linus Torvalds 已提交
2498
 */
2499
static struct rq *finish_task_switch(struct task_struct *prev)
L
Linus Torvalds 已提交
2500 2501
	__releases(rq->lock)
{
2502
	struct rq *rq = this_rq();
L
Linus Torvalds 已提交
2503
	struct mm_struct *mm = rq->prev_mm;
O
Oleg Nesterov 已提交
2504
	long prev_state;
L
Linus Torvalds 已提交
2505

2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517
	/*
	 * 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.
	 */

L
Linus Torvalds 已提交
2518 2519 2520 2521
	rq->prev_mm = NULL;

	/*
	 * A task struct has one reference for the use as "current".
2522
	 * If a task dies, then it sets TASK_DEAD in tsk->state and calls
O
Oleg Nesterov 已提交
2523 2524
	 * schedule one last time. The schedule call will never return, and
	 * the scheduled task must drop that reference.
2525 2526 2527 2528 2529
	 *
	 * We must observe prev->state before clearing prev->on_cpu (in
	 * finish_lock_switch), otherwise a concurrent wakeup can get prev
	 * running on another CPU and we could rave with its RUNNING -> DEAD
	 * transition, resulting in a double drop.
L
Linus Torvalds 已提交
2530
	 */
O
Oleg Nesterov 已提交
2531
	prev_state = prev->state;
2532
	vtime_task_switch(prev);
2533
	perf_event_task_sched_in(prev, current);
2534
	finish_lock_switch(rq, prev);
2535
	finish_arch_post_lock_switch();
S
Steven Rostedt 已提交
2536

2537
	fire_sched_in_preempt_notifiers(current);
L
Linus Torvalds 已提交
2538 2539
	if (mm)
		mmdrop(mm);
2540
	if (unlikely(prev_state == TASK_DEAD)) {
2541 2542 2543
		if (prev->sched_class->task_dead)
			prev->sched_class->task_dead(prev);

2544 2545 2546
		/*
		 * Remove function-return probe instances associated with this
		 * task and put them back on the free list.
I
Ingo Molnar 已提交
2547
		 */
2548
		kprobe_flush_task(prev);
L
Linus Torvalds 已提交
2549
		put_task_struct(prev);
2550
	}
2551

2552
	tick_nohz_task_switch();
2553
	return rq;
L
Linus Torvalds 已提交
2554 2555
}

2556 2557 2558
#ifdef CONFIG_SMP

/* rq->lock is NOT held, but preemption is disabled */
2559
static void __balance_callback(struct rq *rq)
2560
{
2561 2562 2563
	struct callback_head *head, *next;
	void (*func)(struct rq *rq);
	unsigned long flags;
2564

2565 2566 2567 2568 2569 2570 2571 2572
	raw_spin_lock_irqsave(&rq->lock, flags);
	head = rq->balance_callback;
	rq->balance_callback = NULL;
	while (head) {
		func = (void (*)(struct rq *))head->func;
		next = head->next;
		head->next = NULL;
		head = next;
2573

2574
		func(rq);
2575
	}
2576 2577 2578 2579 2580 2581 2582
	raw_spin_unlock_irqrestore(&rq->lock, flags);
}

static inline void balance_callback(struct rq *rq)
{
	if (unlikely(rq->balance_callback))
		__balance_callback(rq);
2583 2584 2585
}

#else
2586

2587
static inline void balance_callback(struct rq *rq)
2588
{
L
Linus Torvalds 已提交
2589 2590
}

2591 2592
#endif

L
Linus Torvalds 已提交
2593 2594 2595 2596
/**
 * schedule_tail - first thing a freshly forked thread must call.
 * @prev: the thread we just switched away from.
 */
2597
asmlinkage __visible void schedule_tail(struct task_struct *prev)
L
Linus Torvalds 已提交
2598 2599
	__releases(rq->lock)
{
2600
	struct rq *rq;
2601

2602 2603 2604 2605 2606 2607 2608 2609 2610
	/*
	 * 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).
	 */

2611
	rq = finish_task_switch(prev);
2612
	balance_callback(rq);
2613
	preempt_enable();
2614

L
Linus Torvalds 已提交
2615
	if (current->set_child_tid)
2616
		put_user(task_pid_vnr(current), current->set_child_tid);
L
Linus Torvalds 已提交
2617 2618 2619
}

/*
2620
 * context_switch - switch to the new MM and the new thread's register state.
L
Linus Torvalds 已提交
2621
 */
2622
static inline struct rq *
2623
context_switch(struct rq *rq, struct task_struct *prev,
2624
	       struct task_struct *next)
L
Linus Torvalds 已提交
2625
{
I
Ingo Molnar 已提交
2626
	struct mm_struct *mm, *oldmm;
L
Linus Torvalds 已提交
2627

2628
	prepare_task_switch(rq, prev, next);
2629

I
Ingo Molnar 已提交
2630 2631
	mm = next->mm;
	oldmm = prev->active_mm;
2632 2633 2634 2635 2636
	/*
	 * For paravirt, this is coupled with an exit in switch_to to
	 * combine the page table reload and the switch backend into
	 * one hypercall.
	 */
2637
	arch_start_context_switch(prev);
2638

2639
	if (!mm) {
L
Linus Torvalds 已提交
2640 2641 2642 2643 2644 2645
		next->active_mm = oldmm;
		atomic_inc(&oldmm->mm_count);
		enter_lazy_tlb(oldmm, next);
	} else
		switch_mm(oldmm, mm, next);

2646
	if (!prev->mm) {
L
Linus Torvalds 已提交
2647 2648 2649
		prev->active_mm = NULL;
		rq->prev_mm = oldmm;
	}
2650 2651 2652 2653 2654 2655
	/*
	 * 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:
	 */
2656
	lockdep_unpin_lock(&rq->lock);
2657
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
L
Linus Torvalds 已提交
2658 2659 2660

	/* Here we just switch the register state and the stack. */
	switch_to(prev, next, prev);
I
Ingo Molnar 已提交
2661
	barrier();
2662 2663

	return finish_task_switch(prev);
L
Linus Torvalds 已提交
2664 2665 2666
}

/*
2667
 * nr_running and nr_context_switches:
L
Linus Torvalds 已提交
2668 2669
 *
 * externally visible scheduler statistics: current number of runnable
2670
 * threads, total number of context switches performed since bootup.
L
Linus Torvalds 已提交
2671 2672 2673 2674 2675 2676 2677 2678 2679
 */
unsigned long nr_running(void)
{
	unsigned long i, sum = 0;

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

	return sum;
2680
}
L
Linus Torvalds 已提交
2681

2682 2683
/*
 * Check if only the current task is running on the cpu.
2684 2685 2686 2687 2688 2689 2690 2691 2692 2693
 *
 * Caution: this function does not check that the caller has disabled
 * preemption, thus the result might have a time-of-check-to-time-of-use
 * race.  The caller is responsible to use it correctly, for example:
 *
 * - from a non-preemptable section (of course)
 *
 * - from a thread that is bound to a single CPU
 *
 * - in a loop with very short iterations (e.g. a polling loop)
2694 2695 2696
 */
bool single_task_running(void)
{
2697
	return raw_rq()->nr_running == 1;
2698 2699 2700
}
EXPORT_SYMBOL(single_task_running);

L
Linus Torvalds 已提交
2701
unsigned long long nr_context_switches(void)
2702
{
2703 2704
	int i;
	unsigned long long sum = 0;
2705

2706
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2707
		sum += cpu_rq(i)->nr_switches;
2708

L
Linus Torvalds 已提交
2709 2710
	return sum;
}
2711

L
Linus Torvalds 已提交
2712 2713 2714
unsigned long nr_iowait(void)
{
	unsigned long i, sum = 0;
2715

2716
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2717
		sum += atomic_read(&cpu_rq(i)->nr_iowait);
2718

L
Linus Torvalds 已提交
2719 2720
	return sum;
}
2721

2722
unsigned long nr_iowait_cpu(int cpu)
2723
{
2724
	struct rq *this = cpu_rq(cpu);
2725 2726
	return atomic_read(&this->nr_iowait);
}
2727

2728 2729
void get_iowait_load(unsigned long *nr_waiters, unsigned long *load)
{
2730 2731 2732
	struct rq *rq = this_rq();
	*nr_waiters = atomic_read(&rq->nr_iowait);
	*load = rq->load.weight;
2733 2734
}

I
Ingo Molnar 已提交
2735
#ifdef CONFIG_SMP
2736

2737
/*
P
Peter Zijlstra 已提交
2738 2739
 * sched_exec - execve() is a valuable balancing opportunity, because at
 * this point the task has the smallest effective memory and cache footprint.
2740
 */
P
Peter Zijlstra 已提交
2741
void sched_exec(void)
2742
{
P
Peter Zijlstra 已提交
2743
	struct task_struct *p = current;
L
Linus Torvalds 已提交
2744
	unsigned long flags;
2745
	int dest_cpu;
2746

2747
	raw_spin_lock_irqsave(&p->pi_lock, flags);
2748
	dest_cpu = p->sched_class->select_task_rq(p, task_cpu(p), SD_BALANCE_EXEC, 0);
2749 2750
	if (dest_cpu == smp_processor_id())
		goto unlock;
P
Peter Zijlstra 已提交
2751

2752
	if (likely(cpu_active(dest_cpu))) {
2753
		struct migration_arg arg = { p, dest_cpu };
2754

2755 2756
		raw_spin_unlock_irqrestore(&p->pi_lock, flags);
		stop_one_cpu(task_cpu(p), migration_cpu_stop, &arg);
L
Linus Torvalds 已提交
2757 2758
		return;
	}
2759
unlock:
2760
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
2761
}
I
Ingo Molnar 已提交
2762

L
Linus Torvalds 已提交
2763 2764 2765
#endif

DEFINE_PER_CPU(struct kernel_stat, kstat);
2766
DEFINE_PER_CPU(struct kernel_cpustat, kernel_cpustat);
L
Linus Torvalds 已提交
2767 2768

EXPORT_PER_CPU_SYMBOL(kstat);
2769
EXPORT_PER_CPU_SYMBOL(kernel_cpustat);
L
Linus Torvalds 已提交
2770

2771 2772 2773 2774 2775 2776 2777 2778 2779
/*
 * 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)
{
	unsigned long flags;
	struct rq *rq;
2780
	u64 ns;
2781

2782 2783 2784 2785 2786 2787 2788 2789 2790
#if defined(CONFIG_64BIT) && defined(CONFIG_SMP)
	/*
	 * 64-bit doesn't need locks to atomically read a 64bit value.
	 * So we have a optimization chance when the task's delta_exec is 0.
	 * Reading ->on_cpu is racy, but this is ok.
	 *
	 * 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
	 * indistinguishable from the read occurring a few cycles earlier.
2791 2792
	 * If we see ->on_cpu without ->on_rq, the task is leaving, and has
	 * been accounted, so we're correct here as well.
2793
	 */
2794
	if (!p->on_cpu || !task_on_rq_queued(p))
2795 2796 2797
		return p->se.sum_exec_runtime;
#endif

2798
	rq = task_rq_lock(p, &flags);
2799 2800 2801 2802 2803 2804 2805 2806 2807 2808
	/*
	 * 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)) {
		update_rq_clock(rq);
		p->sched_class->update_curr(rq);
	}
	ns = p->se.sum_exec_runtime;
2809
	task_rq_unlock(rq, p, &flags);
2810 2811 2812

	return ns;
}
2813

2814 2815 2816 2817 2818 2819 2820 2821
/*
 * This function gets called by the timer code, with HZ frequency.
 * We call it with interrupts disabled.
 */
void scheduler_tick(void)
{
	int cpu = smp_processor_id();
	struct rq *rq = cpu_rq(cpu);
I
Ingo Molnar 已提交
2822
	struct task_struct *curr = rq->curr;
2823 2824

	sched_clock_tick();
I
Ingo Molnar 已提交
2825

2826
	raw_spin_lock(&rq->lock);
2827
	update_rq_clock(rq);
P
Peter Zijlstra 已提交
2828
	curr->sched_class->task_tick(rq, curr, 0);
2829
	update_cpu_load_active(rq);
2830
	calc_global_load_tick(rq);
2831
	raw_spin_unlock(&rq->lock);
2832

2833
	perf_event_task_tick();
2834

2835
#ifdef CONFIG_SMP
2836
	rq->idle_balance = idle_cpu(cpu);
2837
	trigger_load_balance(rq);
2838
#endif
2839
	rq_last_tick_reset(rq);
L
Linus Torvalds 已提交
2840 2841
}

2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852
#ifdef CONFIG_NO_HZ_FULL
/**
 * scheduler_tick_max_deferment
 *
 * Keep at least one tick per second when a single
 * active task is running because the scheduler doesn't
 * yet completely support full dynticks environment.
 *
 * This makes sure that uptime, CFS vruntime, load
 * balancing, etc... continue to move forward, even
 * with a very low granularity.
2853 2854
 *
 * Return: Maximum deferment in nanoseconds.
2855 2856 2857 2858
 */
u64 scheduler_tick_max_deferment(void)
{
	struct rq *rq = this_rq();
2859
	unsigned long next, now = READ_ONCE(jiffies);
2860 2861 2862 2863 2864 2865

	next = rq->last_sched_tick + HZ;

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

2866
	return jiffies_to_nsecs(next - now);
L
Linus Torvalds 已提交
2867
}
2868
#endif
L
Linus Torvalds 已提交
2869

2870
notrace unsigned long get_parent_ip(unsigned long addr)
2871 2872 2873 2874 2875 2876 2877 2878
{
	if (in_lock_functions(addr)) {
		addr = CALLER_ADDR2;
		if (in_lock_functions(addr))
			addr = CALLER_ADDR3;
	}
	return addr;
}
L
Linus Torvalds 已提交
2879

2880 2881 2882
#if defined(CONFIG_PREEMPT) && (defined(CONFIG_DEBUG_PREEMPT) || \
				defined(CONFIG_PREEMPT_TRACER))

2883
void preempt_count_add(int val)
L
Linus Torvalds 已提交
2884
{
2885
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
2886 2887 2888
	/*
	 * Underflow?
	 */
2889 2890
	if (DEBUG_LOCKS_WARN_ON((preempt_count() < 0)))
		return;
2891
#endif
2892
	__preempt_count_add(val);
2893
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
2894 2895 2896
	/*
	 * Spinlock count overflowing soon?
	 */
2897 2898
	DEBUG_LOCKS_WARN_ON((preempt_count() & PREEMPT_MASK) >=
				PREEMPT_MASK - 10);
2899
#endif
2900 2901 2902 2903 2904 2905 2906
	if (preempt_count() == val) {
		unsigned long ip = get_parent_ip(CALLER_ADDR1);
#ifdef CONFIG_DEBUG_PREEMPT
		current->preempt_disable_ip = ip;
#endif
		trace_preempt_off(CALLER_ADDR0, ip);
	}
L
Linus Torvalds 已提交
2907
}
2908
EXPORT_SYMBOL(preempt_count_add);
2909
NOKPROBE_SYMBOL(preempt_count_add);
L
Linus Torvalds 已提交
2910

2911
void preempt_count_sub(int val)
L
Linus Torvalds 已提交
2912
{
2913
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
2914 2915 2916
	/*
	 * Underflow?
	 */
2917
	if (DEBUG_LOCKS_WARN_ON(val > preempt_count()))
2918
		return;
L
Linus Torvalds 已提交
2919 2920 2921
	/*
	 * Is the spinlock portion underflowing?
	 */
2922 2923 2924
	if (DEBUG_LOCKS_WARN_ON((val < PREEMPT_MASK) &&
			!(preempt_count() & PREEMPT_MASK)))
		return;
2925
#endif
2926

2927 2928
	if (preempt_count() == val)
		trace_preempt_on(CALLER_ADDR0, get_parent_ip(CALLER_ADDR1));
2929
	__preempt_count_sub(val);
L
Linus Torvalds 已提交
2930
}
2931
EXPORT_SYMBOL(preempt_count_sub);
2932
NOKPROBE_SYMBOL(preempt_count_sub);
L
Linus Torvalds 已提交
2933 2934 2935 2936

#endif

/*
I
Ingo Molnar 已提交
2937
 * Print scheduling while atomic bug:
L
Linus Torvalds 已提交
2938
 */
I
Ingo Molnar 已提交
2939
static noinline void __schedule_bug(struct task_struct *prev)
L
Linus Torvalds 已提交
2940
{
2941 2942 2943
	if (oops_in_progress)
		return;

P
Peter Zijlstra 已提交
2944 2945
	printk(KERN_ERR "BUG: scheduling while atomic: %s/%d/0x%08x\n",
		prev->comm, prev->pid, preempt_count());
2946

I
Ingo Molnar 已提交
2947
	debug_show_held_locks(prev);
2948
	print_modules();
I
Ingo Molnar 已提交
2949 2950
	if (irqs_disabled())
		print_irqtrace_events(prev);
2951 2952 2953 2954 2955 2956 2957
#ifdef CONFIG_DEBUG_PREEMPT
	if (in_atomic_preempt_off()) {
		pr_err("Preemption disabled at:");
		print_ip_sym(current->preempt_disable_ip);
		pr_cont("\n");
	}
#endif
2958
	dump_stack();
2959
	add_taint(TAINT_WARN, LOCKDEP_STILL_OK);
I
Ingo Molnar 已提交
2960
}
L
Linus Torvalds 已提交
2961

I
Ingo Molnar 已提交
2962 2963 2964 2965 2966
/*
 * Various schedule()-time debugging checks and statistics:
 */
static inline void schedule_debug(struct task_struct *prev)
{
2967 2968 2969
#ifdef CONFIG_SCHED_STACK_END_CHECK
	BUG_ON(unlikely(task_stack_end_corrupted(prev)));
#endif
2970 2971

	if (unlikely(in_atomic_preempt_off()))
I
Ingo Molnar 已提交
2972
		__schedule_bug(prev);
2973
	rcu_sleep_check();
I
Ingo Molnar 已提交
2974

L
Linus Torvalds 已提交
2975 2976
	profile_hit(SCHED_PROFILING, __builtin_return_address(0));

2977
	schedstat_inc(this_rq(), sched_count);
I
Ingo Molnar 已提交
2978 2979 2980 2981 2982 2983
}

/*
 * Pick up the highest-prio task:
 */
static inline struct task_struct *
2984
pick_next_task(struct rq *rq, struct task_struct *prev)
I
Ingo Molnar 已提交
2985
{
2986
	const struct sched_class *class = &fair_sched_class;
I
Ingo Molnar 已提交
2987
	struct task_struct *p;
L
Linus Torvalds 已提交
2988 2989

	/*
I
Ingo Molnar 已提交
2990 2991
	 * Optimization: we know that if all tasks are in
	 * the fair class we can call that function directly:
L
Linus Torvalds 已提交
2992
	 */
2993
	if (likely(prev->sched_class == class &&
2994
		   rq->nr_running == rq->cfs.h_nr_running)) {
2995
		p = fair_sched_class.pick_next_task(rq, prev);
2996 2997 2998 2999 3000 3001 3002 3003
		if (unlikely(p == RETRY_TASK))
			goto again;

		/* assumes fair_sched_class->next == idle_sched_class */
		if (unlikely(!p))
			p = idle_sched_class.pick_next_task(rq, prev);

		return p;
L
Linus Torvalds 已提交
3004 3005
	}

3006
again:
3007
	for_each_class(class) {
3008
		p = class->pick_next_task(rq, prev);
3009 3010 3011
		if (p) {
			if (unlikely(p == RETRY_TASK))
				goto again;
I
Ingo Molnar 已提交
3012
			return p;
3013
		}
I
Ingo Molnar 已提交
3014
	}
3015 3016

	BUG(); /* the idle class will always have a runnable task */
I
Ingo Molnar 已提交
3017
}
L
Linus Torvalds 已提交
3018

I
Ingo Molnar 已提交
3019
/*
3020
 * __schedule() is the main scheduler function.
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
 *
 * The main means of driving the scheduler and thus entering this function are:
 *
 *   1. Explicit blocking: mutex, semaphore, waitqueue, etc.
 *
 *   2. TIF_NEED_RESCHED flag is checked on interrupt and userspace return
 *      paths. For example, see arch/x86/entry_64.S.
 *
 *      To drive preemption between tasks, the scheduler sets the flag in timer
 *      interrupt handler scheduler_tick().
 *
 *   3. Wakeups don't really cause entry into schedule(). They add a
 *      task to the run-queue and that's it.
 *
 *      Now, if the new task added to the run-queue preempts the current
 *      task, then the wakeup sets TIF_NEED_RESCHED and schedule() gets
 *      called on the nearest possible occasion:
 *
 *       - If the kernel is preemptible (CONFIG_PREEMPT=y):
 *
 *         - in syscall or exception context, at the next outmost
 *           preempt_enable(). (this might be as soon as the wake_up()'s
 *           spin_unlock()!)
 *
 *         - in IRQ context, return from interrupt-handler to
 *           preemptible context
 *
 *       - If the kernel is not preemptible (CONFIG_PREEMPT is not set)
 *         then at the next:
 *
 *          - cond_resched() call
 *          - explicit schedule() call
 *          - return from syscall or exception to user-space
 *          - return from interrupt-handler to user-space
3055
 *
3056
 * WARNING: must be called with preemption disabled!
I
Ingo Molnar 已提交
3057
 */
3058
static void __sched __schedule(bool preempt)
I
Ingo Molnar 已提交
3059 3060
{
	struct task_struct *prev, *next;
3061
	unsigned long *switch_count;
I
Ingo Molnar 已提交
3062
	struct rq *rq;
3063
	int cpu;
I
Ingo Molnar 已提交
3064 3065 3066

	cpu = smp_processor_id();
	rq = cpu_rq(cpu);
3067
	rcu_note_context_switch();
I
Ingo Molnar 已提交
3068 3069
	prev = rq->curr;

3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080
	/*
	 * do_exit() calls schedule() with preemption disabled as an exception;
	 * however we must fix that up, otherwise the next task will see an
	 * inconsistent (higher) preempt count.
	 *
	 * It also avoids the below schedule_debug() test from complaining
	 * about this.
	 */
	if (unlikely(prev->state == TASK_DEAD))
		preempt_enable_no_resched_notrace();

I
Ingo Molnar 已提交
3081
	schedule_debug(prev);
L
Linus Torvalds 已提交
3082

3083
	if (sched_feat(HRTICK))
M
Mike Galbraith 已提交
3084
		hrtick_clear(rq);
P
Peter Zijlstra 已提交
3085

3086 3087 3088 3089 3090 3091
	/*
	 * Make sure that signal_pending_state()->signal_pending() below
	 * can't be reordered with __set_current_state(TASK_INTERRUPTIBLE)
	 * done by the caller to avoid the race with signal_wake_up().
	 */
	smp_mb__before_spinlock();
3092
	raw_spin_lock_irq(&rq->lock);
3093
	lockdep_pin_lock(&rq->lock);
L
Linus Torvalds 已提交
3094

3095 3096
	rq->clock_skip_update <<= 1; /* promote REQ to ACT */

3097
	switch_count = &prev->nivcsw;
3098
	if (!preempt && prev->state) {
T
Tejun Heo 已提交
3099
		if (unlikely(signal_pending_state(prev->state, prev))) {
L
Linus Torvalds 已提交
3100
			prev->state = TASK_RUNNING;
T
Tejun Heo 已提交
3101
		} else {
3102 3103 3104
			deactivate_task(rq, prev, DEQUEUE_SLEEP);
			prev->on_rq = 0;

T
Tejun Heo 已提交
3105
			/*
3106 3107 3108
			 * If a worker went to sleep, notify and ask workqueue
			 * whether it wants to wake up a task to maintain
			 * concurrency.
T
Tejun Heo 已提交
3109 3110 3111 3112 3113 3114 3115 3116 3117
			 */
			if (prev->flags & PF_WQ_WORKER) {
				struct task_struct *to_wakeup;

				to_wakeup = wq_worker_sleeping(prev, cpu);
				if (to_wakeup)
					try_to_wake_up_local(to_wakeup);
			}
		}
I
Ingo Molnar 已提交
3118
		switch_count = &prev->nvcsw;
L
Linus Torvalds 已提交
3119 3120
	}

3121
	if (task_on_rq_queued(prev))
3122 3123 3124
		update_rq_clock(rq);

	next = pick_next_task(rq, prev);
3125
	clear_tsk_need_resched(prev);
3126
	clear_preempt_need_resched();
3127
	rq->clock_skip_update = 0;
L
Linus Torvalds 已提交
3128 3129 3130 3131 3132 3133

	if (likely(prev != next)) {
		rq->nr_switches++;
		rq->curr = next;
		++*switch_count;

3134
		trace_sched_switch(preempt, prev, next);
3135 3136
		rq = context_switch(rq, prev, next); /* unlocks the rq */
		cpu = cpu_of(rq);
3137 3138
	} else {
		lockdep_unpin_lock(&rq->lock);
3139
		raw_spin_unlock_irq(&rq->lock);
3140
	}
L
Linus Torvalds 已提交
3141

3142
	balance_callback(rq);
L
Linus Torvalds 已提交
3143
}
3144

3145 3146
static inline void sched_submit_work(struct task_struct *tsk)
{
3147
	if (!tsk->state || tsk_is_pi_blocked(tsk))
3148 3149 3150 3151 3152 3153 3154 3155 3156
		return;
	/*
	 * If we are going to sleep and we have plugged IO queued,
	 * make sure to submit it to avoid deadlocks.
	 */
	if (blk_needs_flush_plug(tsk))
		blk_schedule_flush_plug(tsk);
}

3157
asmlinkage __visible void __sched schedule(void)
3158
{
3159 3160 3161
	struct task_struct *tsk = current;

	sched_submit_work(tsk);
3162
	do {
3163
		preempt_disable();
3164
		__schedule(false);
3165
		sched_preempt_enable_no_resched();
3166
	} while (need_resched());
3167
}
L
Linus Torvalds 已提交
3168 3169
EXPORT_SYMBOL(schedule);

3170
#ifdef CONFIG_CONTEXT_TRACKING
3171
asmlinkage __visible void __sched schedule_user(void)
3172 3173 3174 3175 3176 3177
{
	/*
	 * 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.
3178 3179
	 *
	 * NB: There are buggy callers of this function.  Ideally we
3180
	 * should warn if prev_state != CONTEXT_USER, but that will trigger
3181
	 * too frequently to make sense yet.
3182
	 */
3183
	enum ctx_state prev_state = exception_enter();
3184
	schedule();
3185
	exception_exit(prev_state);
3186 3187 3188
}
#endif

3189 3190 3191 3192 3193 3194 3195
/**
 * schedule_preempt_disabled - called with preemption disabled
 *
 * Returns with preemption disabled. Note: preempt_count must be 1
 */
void __sched schedule_preempt_disabled(void)
{
3196
	sched_preempt_enable_no_resched();
3197 3198 3199 3200
	schedule();
	preempt_disable();
}

3201
static void __sched notrace preempt_schedule_common(void)
3202 3203
{
	do {
3204
		preempt_active_enter();
3205
		__schedule(true);
3206
		preempt_active_exit();
3207 3208 3209 3210 3211 3212 3213 3214

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

L
Linus Torvalds 已提交
3215 3216
#ifdef CONFIG_PREEMPT
/*
3217
 * this is the entry point to schedule() from in-kernel preemption
I
Ingo Molnar 已提交
3218
 * off of preempt_enable. Kernel preemptions off return from interrupt
L
Linus Torvalds 已提交
3219 3220
 * occur there and call schedule directly.
 */
3221
asmlinkage __visible void __sched notrace preempt_schedule(void)
L
Linus Torvalds 已提交
3222 3223 3224
{
	/*
	 * If there is a non-zero preempt_count or interrupts are disabled,
I
Ingo Molnar 已提交
3225
	 * we do not want to preempt the current task. Just return..
L
Linus Torvalds 已提交
3226
	 */
3227
	if (likely(!preemptible()))
L
Linus Torvalds 已提交
3228 3229
		return;

3230
	preempt_schedule_common();
L
Linus Torvalds 已提交
3231
}
3232
NOKPROBE_SYMBOL(preempt_schedule);
L
Linus Torvalds 已提交
3233
EXPORT_SYMBOL(preempt_schedule);
3234 3235

/**
3236
 * preempt_schedule_notrace - preempt_schedule called by tracing
3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248
 *
 * 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.
 */
3249
asmlinkage __visible void __sched notrace preempt_schedule_notrace(void)
3250 3251 3252 3253 3254 3255 3256
{
	enum ctx_state prev_ctx;

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

	do {
3257 3258 3259 3260 3261 3262 3263
		/*
		 * Use raw __prempt_count() ops that don't call function.
		 * We can't call functions before disabling preemption which
		 * disarm preemption tracing recursions.
		 */
		__preempt_count_add(PREEMPT_ACTIVE + PREEMPT_DISABLE_OFFSET);
		barrier();
3264 3265 3266 3267 3268 3269
		/*
		 * Needs preempt disabled in case user_exit() is traced
		 * and the tracer calls preempt_enable_notrace() causing
		 * an infinite recursion.
		 */
		prev_ctx = exception_enter();
3270
		__schedule(true);
3271 3272 3273
		exception_exit(prev_ctx);

		barrier();
3274
		__preempt_count_sub(PREEMPT_ACTIVE + PREEMPT_DISABLE_OFFSET);
3275 3276
	} while (need_resched());
}
3277
EXPORT_SYMBOL_GPL(preempt_schedule_notrace);
3278

3279
#endif /* CONFIG_PREEMPT */
L
Linus Torvalds 已提交
3280 3281

/*
3282
 * this is the entry point to schedule() from kernel preemption
L
Linus Torvalds 已提交
3283 3284 3285 3286
 * off of irq context.
 * Note, that this is called and return with irqs disabled. This will
 * protect us against recursive calling from irq.
 */
3287
asmlinkage __visible void __sched preempt_schedule_irq(void)
L
Linus Torvalds 已提交
3288
{
3289
	enum ctx_state prev_state;
3290

3291
	/* Catch callers which need to be fixed */
3292
	BUG_ON(preempt_count() || !irqs_disabled());
L
Linus Torvalds 已提交
3293

3294 3295
	prev_state = exception_enter();

3296
	do {
3297
		preempt_active_enter();
3298
		local_irq_enable();
3299
		__schedule(true);
3300
		local_irq_disable();
3301
		preempt_active_exit();
3302
	} while (need_resched());
3303 3304

	exception_exit(prev_state);
L
Linus Torvalds 已提交
3305 3306
}

P
Peter Zijlstra 已提交
3307
int default_wake_function(wait_queue_t *curr, unsigned mode, int wake_flags,
I
Ingo Molnar 已提交
3308
			  void *key)
L
Linus Torvalds 已提交
3309
{
P
Peter Zijlstra 已提交
3310
	return try_to_wake_up(curr->private, mode, wake_flags);
L
Linus Torvalds 已提交
3311 3312 3313
}
EXPORT_SYMBOL(default_wake_function);

3314 3315 3316 3317 3318 3319 3320 3321 3322 3323
#ifdef CONFIG_RT_MUTEXES

/*
 * rt_mutex_setprio - set the current priority of a task
 * @p: task
 * @prio: prio value (kernel-internal form)
 *
 * This function changes the 'effective' priority of a task. It does
 * not touch ->normal_prio like __setscheduler().
 *
3324 3325
 * Used by the rt_mutex code to implement priority inheritance
 * logic. Call site only calls if the priority of the task changed.
3326
 */
3327
void rt_mutex_setprio(struct task_struct *p, int prio)
3328
{
3329
	int oldprio, queued, running, enqueue_flag = 0;
3330
	struct rq *rq;
3331
	const struct sched_class *prev_class;
3332

3333
	BUG_ON(prio > MAX_PRIO);
3334

3335
	rq = __task_rq_lock(p);
3336

3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354
	/*
	 * 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;
	}

3355
	trace_sched_pi_setprio(p, prio);
3356
	oldprio = p->prio;
3357
	prev_class = p->sched_class;
3358
	queued = task_on_rq_queued(p);
3359
	running = task_current(rq, p);
3360
	if (queued)
3361
		dequeue_task(rq, p, 0);
3362
	if (running)
3363
		put_prev_task(rq, p);
I
Ingo Molnar 已提交
3364

3365 3366 3367 3368 3369 3370 3371 3372 3373 3374
	/*
	 * 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)) {
3375 3376 3377
		struct task_struct *pi_task = rt_mutex_get_top_task(p);
		if (!dl_prio(p->normal_prio) ||
		    (pi_task && dl_entity_preempt(&pi_task->dl, &p->dl))) {
3378 3379 3380 3381
			p->dl.dl_boosted = 1;
			enqueue_flag = ENQUEUE_REPLENISH;
		} else
			p->dl.dl_boosted = 0;
3382
		p->sched_class = &dl_sched_class;
3383 3384 3385 3386 3387
	} else if (rt_prio(prio)) {
		if (dl_prio(oldprio))
			p->dl.dl_boosted = 0;
		if (oldprio < prio)
			enqueue_flag = ENQUEUE_HEAD;
I
Ingo Molnar 已提交
3388
		p->sched_class = &rt_sched_class;
3389 3390 3391
	} else {
		if (dl_prio(oldprio))
			p->dl.dl_boosted = 0;
3392 3393
		if (rt_prio(oldprio))
			p->rt.timeout = 0;
I
Ingo Molnar 已提交
3394
		p->sched_class = &fair_sched_class;
3395
	}
I
Ingo Molnar 已提交
3396

3397 3398
	p->prio = prio;

3399 3400
	if (running)
		p->sched_class->set_curr_task(rq);
3401
	if (queued)
3402
		enqueue_task(rq, p, enqueue_flag);
3403

P
Peter Zijlstra 已提交
3404
	check_class_changed(rq, p, prev_class, oldprio);
3405
out_unlock:
3406
	preempt_disable(); /* avoid rq from going away on us */
3407
	__task_rq_unlock(rq);
3408 3409 3410

	balance_callback(rq);
	preempt_enable();
3411 3412
}
#endif
3413

3414
void set_user_nice(struct task_struct *p, long nice)
L
Linus Torvalds 已提交
3415
{
3416
	int old_prio, delta, queued;
L
Linus Torvalds 已提交
3417
	unsigned long flags;
3418
	struct rq *rq;
L
Linus Torvalds 已提交
3419

3420
	if (task_nice(p) == nice || nice < MIN_NICE || nice > MAX_NICE)
L
Linus Torvalds 已提交
3421 3422 3423 3424 3425 3426 3427 3428 3429 3430
		return;
	/*
	 * We have to be careful, if called from sys_setpriority(),
	 * the task might be in the middle of scheduling on another CPU.
	 */
	rq = task_rq_lock(p, &flags);
	/*
	 * The RT priorities are set via sched_setscheduler(), but we still
	 * allow the 'normal' nice value to be set - but as expected
	 * it wont have any effect on scheduling until the task is
3431
	 * SCHED_DEADLINE, SCHED_FIFO or SCHED_RR:
L
Linus Torvalds 已提交
3432
	 */
3433
	if (task_has_dl_policy(p) || task_has_rt_policy(p)) {
L
Linus Torvalds 已提交
3434 3435 3436
		p->static_prio = NICE_TO_PRIO(nice);
		goto out_unlock;
	}
3437 3438
	queued = task_on_rq_queued(p);
	if (queued)
3439
		dequeue_task(rq, p, 0);
L
Linus Torvalds 已提交
3440 3441

	p->static_prio = NICE_TO_PRIO(nice);
3442
	set_load_weight(p);
3443 3444 3445
	old_prio = p->prio;
	p->prio = effective_prio(p);
	delta = p->prio - old_prio;
L
Linus Torvalds 已提交
3446

3447
	if (queued) {
3448
		enqueue_task(rq, p, 0);
L
Linus Torvalds 已提交
3449
		/*
3450 3451
		 * If the task increased its priority or is running and
		 * lowered its priority, then reschedule its CPU:
L
Linus Torvalds 已提交
3452
		 */
3453
		if (delta < 0 || (delta > 0 && task_running(rq, p)))
3454
			resched_curr(rq);
L
Linus Torvalds 已提交
3455 3456
	}
out_unlock:
3457
	task_rq_unlock(rq, p, &flags);
L
Linus Torvalds 已提交
3458 3459 3460
}
EXPORT_SYMBOL(set_user_nice);

M
Matt Mackall 已提交
3461 3462 3463 3464 3465
/*
 * can_nice - check if a task can reduce its nice value
 * @p: task
 * @nice: nice value
 */
3466
int can_nice(const struct task_struct *p, const int nice)
M
Matt Mackall 已提交
3467
{
3468
	/* convert nice value [19,-20] to rlimit style value [1,40] */
3469
	int nice_rlim = nice_to_rlimit(nice);
3470

3471
	return (nice_rlim <= task_rlimit(p, RLIMIT_NICE) ||
M
Matt Mackall 已提交
3472 3473 3474
		capable(CAP_SYS_NICE));
}

L
Linus Torvalds 已提交
3475 3476 3477 3478 3479 3480 3481 3482 3483
#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.
 */
3484
SYSCALL_DEFINE1(nice, int, increment)
L
Linus Torvalds 已提交
3485
{
3486
	long nice, retval;
L
Linus Torvalds 已提交
3487 3488 3489 3490 3491 3492

	/*
	 * 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.
	 */
3493
	increment = clamp(increment, -NICE_WIDTH, NICE_WIDTH);
3494
	nice = task_nice(current) + increment;
L
Linus Torvalds 已提交
3495

3496
	nice = clamp_val(nice, MIN_NICE, MAX_NICE);
M
Matt Mackall 已提交
3497 3498 3499
	if (increment < 0 && !can_nice(current, nice))
		return -EPERM;

L
Linus Torvalds 已提交
3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513
	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.
 *
3514
 * Return: The priority value as seen by users in /proc.
L
Linus Torvalds 已提交
3515 3516 3517
 * RT tasks are offset by -200. Normal tasks are centered
 * around 0, value goes from -16 to +15.
 */
3518
int task_prio(const struct task_struct *p)
L
Linus Torvalds 已提交
3519 3520 3521 3522 3523 3524 3525
{
	return p->prio - MAX_RT_PRIO;
}

/**
 * idle_cpu - is a given cpu idle currently?
 * @cpu: the processor in question.
3526 3527
 *
 * Return: 1 if the CPU is currently idle. 0 otherwise.
L
Linus Torvalds 已提交
3528 3529 3530
 */
int idle_cpu(int cpu)
{
T
Thomas Gleixner 已提交
3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544
	struct rq *rq = cpu_rq(cpu);

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

	if (rq->nr_running)
		return 0;

#ifdef CONFIG_SMP
	if (!llist_empty(&rq->wake_list))
		return 0;
#endif

	return 1;
L
Linus Torvalds 已提交
3545 3546 3547 3548 3549
}

/**
 * idle_task - return the idle task for a given cpu.
 * @cpu: the processor in question.
3550 3551
 *
 * Return: The idle task for the cpu @cpu.
L
Linus Torvalds 已提交
3552
 */
3553
struct task_struct *idle_task(int cpu)
L
Linus Torvalds 已提交
3554 3555 3556 3557 3558 3559 3560
{
	return cpu_rq(cpu)->idle;
}

/**
 * find_process_by_pid - find a process with a matching PID value.
 * @pid: the pid in question.
3561 3562
 *
 * The task of @pid, if found. %NULL otherwise.
L
Linus Torvalds 已提交
3563
 */
A
Alexey Dobriyan 已提交
3564
static struct task_struct *find_process_by_pid(pid_t pid)
L
Linus Torvalds 已提交
3565
{
3566
	return pid ? find_task_by_vpid(pid) : current;
L
Linus Torvalds 已提交
3567 3568
}

3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583
/*
 * This function initializes the sched_dl_entity of a newly becoming
 * SCHED_DEADLINE task.
 *
 * Only the static values are considered here, the actual runtime and the
 * absolute deadline will be properly calculated when the task is enqueued
 * for the first time with its new policy.
 */
static void
__setparam_dl(struct task_struct *p, const struct sched_attr *attr)
{
	struct sched_dl_entity *dl_se = &p->dl;

	dl_se->dl_runtime = attr->sched_runtime;
	dl_se->dl_deadline = attr->sched_deadline;
3584
	dl_se->dl_period = attr->sched_period ?: dl_se->dl_deadline;
3585
	dl_se->flags = attr->sched_flags;
3586
	dl_se->dl_bw = to_ratio(dl_se->dl_period, dl_se->dl_runtime);
3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606

	/*
	 * Changing the parameters of a task is 'tricky' and we're not doing
	 * the correct thing -- also see task_dead_dl() and switched_from_dl().
	 *
	 * What we SHOULD do is delay the bandwidth release until the 0-lag
	 * point. This would include retaining the task_struct until that time
	 * and change dl_overflow() to not immediately decrement the current
	 * amount.
	 *
	 * Instead we retain the current runtime/deadline and let the new
	 * parameters take effect after the current reservation period lapses.
	 * This is safe (albeit pessimistic) because the 0-lag point is always
	 * before the current scheduling deadline.
	 *
	 * We can still have temporary overloads because we do not delay the
	 * change in bandwidth until that time; so admission control is
	 * not on the safe side. It does however guarantee tasks will never
	 * consume more than promised.
	 */
3607 3608
}

3609 3610 3611 3612 3613 3614
/*
 * sched_setparam() passes in -1 for its policy, to let the functions
 * it calls know not to change it.
 */
#define SETPARAM_POLICY	-1

3615 3616
static void __setscheduler_params(struct task_struct *p,
		const struct sched_attr *attr)
L
Linus Torvalds 已提交
3617
{
3618 3619
	int policy = attr->sched_policy;

3620
	if (policy == SETPARAM_POLICY)
3621 3622
		policy = p->policy;

L
Linus Torvalds 已提交
3623
	p->policy = policy;
3624

3625 3626
	if (dl_policy(policy))
		__setparam_dl(p, attr);
3627
	else if (fair_policy(policy))
3628 3629
		p->static_prio = NICE_TO_PRIO(attr->sched_nice);

3630 3631 3632 3633 3634 3635
	/*
	 * __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;
3636
	p->normal_prio = normal_prio(p);
3637 3638
	set_load_weight(p);
}
3639

3640 3641
/* Actually do priority change: must hold pi & rq lock. */
static void __setscheduler(struct rq *rq, struct task_struct *p,
3642
			   const struct sched_attr *attr, bool keep_boost)
3643 3644
{
	__setscheduler_params(p, attr);
3645

3646
	/*
3647 3648
	 * Keep a potential priority boosting if called from
	 * sched_setscheduler().
3649
	 */
3650 3651 3652 3653
	if (keep_boost)
		p->prio = rt_mutex_get_effective_prio(p, normal_prio(p));
	else
		p->prio = normal_prio(p);
3654

3655 3656 3657
	if (dl_prio(p->prio))
		p->sched_class = &dl_sched_class;
	else if (rt_prio(p->prio))
3658 3659 3660
		p->sched_class = &rt_sched_class;
	else
		p->sched_class = &fair_sched_class;
L
Linus Torvalds 已提交
3661
}
3662 3663 3664 3665 3666 3667 3668 3669 3670

static void
__getparam_dl(struct task_struct *p, struct sched_attr *attr)
{
	struct sched_dl_entity *dl_se = &p->dl;

	attr->sched_priority = p->rt_priority;
	attr->sched_runtime = dl_se->dl_runtime;
	attr->sched_deadline = dl_se->dl_deadline;
3671
	attr->sched_period = dl_se->dl_period;
3672 3673 3674 3675 3676 3677
	attr->sched_flags = dl_se->flags;
}

/*
 * This function validates the new parameters of a -deadline task.
 * We ask for the deadline not being zero, and greater or equal
3678
 * than the runtime, as well as the period of being zero or
3679
 * greater than deadline. Furthermore, we have to be sure that
3680 3681 3682 3683
 * user parameters are above the internal resolution of 1us (we
 * check sched_runtime only since it is always the smaller one) and
 * below 2^63 ns (we have to check both sched_deadline and
 * sched_period, as the latter can be zero).
3684 3685 3686 3687
 */
static bool
__checkparam_dl(const struct sched_attr *attr)
{
3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713
	/* deadline != 0 */
	if (attr->sched_deadline == 0)
		return false;

	/*
	 * Since we truncate DL_SCALE bits, make sure we're at least
	 * that big.
	 */
	if (attr->sched_runtime < (1ULL << DL_SCALE))
		return false;

	/*
	 * Since we use the MSB for wrap-around and sign issues, make
	 * sure it's not set (mind that period can be equal to zero).
	 */
	if (attr->sched_deadline & (1ULL << 63) ||
	    attr->sched_period & (1ULL << 63))
		return false;

	/* runtime <= deadline <= period (if period != 0) */
	if ((attr->sched_period != 0 &&
	     attr->sched_period < attr->sched_deadline) ||
	    attr->sched_deadline < attr->sched_runtime)
		return false;

	return true;
3714 3715
}

3716 3717 3718 3719 3720 3721 3722 3723 3724 3725
/*
 * check the target process has a UID that matches the current process's
 */
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);
3726 3727
	match = (uid_eq(cred->euid, pcred->euid) ||
		 uid_eq(cred->euid, pcred->uid));
3728 3729 3730 3731
	rcu_read_unlock();
	return match;
}

3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745
static bool dl_param_changed(struct task_struct *p,
		const struct sched_attr *attr)
{
	struct sched_dl_entity *dl_se = &p->dl;

	if (dl_se->dl_runtime != attr->sched_runtime ||
		dl_se->dl_deadline != attr->sched_deadline ||
		dl_se->dl_period != attr->sched_period ||
		dl_se->flags != attr->sched_flags)
		return true;

	return false;
}

3746 3747
static int __sched_setscheduler(struct task_struct *p,
				const struct sched_attr *attr,
3748
				bool user, bool pi)
L
Linus Torvalds 已提交
3749
{
3750 3751
	int newprio = dl_policy(attr->sched_policy) ? MAX_DL_PRIO - 1 :
		      MAX_RT_PRIO - 1 - attr->sched_priority;
3752
	int retval, oldprio, oldpolicy = -1, queued, running;
3753
	int new_effective_prio, policy = attr->sched_policy;
L
Linus Torvalds 已提交
3754
	unsigned long flags;
3755
	const struct sched_class *prev_class;
3756
	struct rq *rq;
3757
	int reset_on_fork;
L
Linus Torvalds 已提交
3758

3759 3760
	/* may grab non-irq protected spin_locks */
	BUG_ON(in_interrupt());
L
Linus Torvalds 已提交
3761 3762
recheck:
	/* double check policy once rq lock held */
3763 3764
	if (policy < 0) {
		reset_on_fork = p->sched_reset_on_fork;
L
Linus Torvalds 已提交
3765
		policy = oldpolicy = p->policy;
3766
	} else {
3767
		reset_on_fork = !!(attr->sched_flags & SCHED_FLAG_RESET_ON_FORK);
3768

3769
		if (!valid_policy(policy))
3770 3771 3772
			return -EINVAL;
	}

3773 3774 3775
	if (attr->sched_flags & ~(SCHED_FLAG_RESET_ON_FORK))
		return -EINVAL;

L
Linus Torvalds 已提交
3776 3777
	/*
	 * Valid priorities for SCHED_FIFO and SCHED_RR are
I
Ingo Molnar 已提交
3778 3779
	 * 1..MAX_USER_RT_PRIO-1, valid priority for SCHED_NORMAL,
	 * SCHED_BATCH and SCHED_IDLE is 0.
L
Linus Torvalds 已提交
3780
	 */
3781
	if ((p->mm && attr->sched_priority > MAX_USER_RT_PRIO-1) ||
3782
	    (!p->mm && attr->sched_priority > MAX_RT_PRIO-1))
L
Linus Torvalds 已提交
3783
		return -EINVAL;
3784 3785
	if ((dl_policy(policy) && !__checkparam_dl(attr)) ||
	    (rt_policy(policy) != (attr->sched_priority != 0)))
L
Linus Torvalds 已提交
3786 3787
		return -EINVAL;

3788 3789 3790
	/*
	 * Allow unprivileged RT tasks to decrease priority:
	 */
3791
	if (user && !capable(CAP_SYS_NICE)) {
3792
		if (fair_policy(policy)) {
3793
			if (attr->sched_nice < task_nice(p) &&
3794
			    !can_nice(p, attr->sched_nice))
3795 3796 3797
				return -EPERM;
		}

3798
		if (rt_policy(policy)) {
3799 3800
			unsigned long rlim_rtprio =
					task_rlimit(p, RLIMIT_RTPRIO);
3801 3802 3803 3804 3805 3806

			/* can't set/change the rt policy */
			if (policy != p->policy && !rlim_rtprio)
				return -EPERM;

			/* can't increase priority */
3807 3808
			if (attr->sched_priority > p->rt_priority &&
			    attr->sched_priority > rlim_rtprio)
3809 3810
				return -EPERM;
		}
3811

3812 3813 3814 3815 3816 3817 3818 3819 3820
		 /*
		  * Can't set/change SCHED_DEADLINE policy at all for now
		  * (safest behavior); in the future we would like to allow
		  * unprivileged DL tasks to increase their relative deadline
		  * or reduce their runtime (both ways reducing utilization)
		  */
		if (dl_policy(policy))
			return -EPERM;

I
Ingo Molnar 已提交
3821
		/*
3822 3823
		 * Treat SCHED_IDLE as nice 20. Only allow a switch to
		 * SCHED_NORMAL if the RLIMIT_NICE would normally permit it.
I
Ingo Molnar 已提交
3824
		 */
3825
		if (idle_policy(p->policy) && !idle_policy(policy)) {
3826
			if (!can_nice(p, task_nice(p)))
3827 3828
				return -EPERM;
		}
3829

3830
		/* can't change other user's priorities */
3831
		if (!check_same_owner(p))
3832
			return -EPERM;
3833 3834 3835 3836

		/* Normal users shall not reset the sched_reset_on_fork flag */
		if (p->sched_reset_on_fork && !reset_on_fork)
			return -EPERM;
3837
	}
L
Linus Torvalds 已提交
3838

3839
	if (user) {
3840
		retval = security_task_setscheduler(p);
3841 3842 3843 3844
		if (retval)
			return retval;
	}

3845 3846 3847
	/*
	 * make sure no PI-waiters arrive (or leave) while we are
	 * changing the priority of the task:
3848
	 *
L
Lucas De Marchi 已提交
3849
	 * To be able to change p->policy safely, the appropriate
L
Linus Torvalds 已提交
3850 3851
	 * runqueue lock must be held.
	 */
3852
	rq = task_rq_lock(p, &flags);
3853

3854 3855 3856 3857
	/*
	 * Changing the policy of the stop threads its a very bad idea
	 */
	if (p == rq->stop) {
3858
		task_rq_unlock(rq, p, &flags);
3859 3860 3861
		return -EINVAL;
	}

3862
	/*
3863 3864
	 * If not changing anything there's no need to proceed further,
	 * but store a possible modification of reset_on_fork.
3865
	 */
3866
	if (unlikely(policy == p->policy)) {
3867
		if (fair_policy(policy) && attr->sched_nice != task_nice(p))
3868 3869 3870
			goto change;
		if (rt_policy(policy) && attr->sched_priority != p->rt_priority)
			goto change;
3871
		if (dl_policy(policy) && dl_param_changed(p, attr))
3872
			goto change;
3873

3874
		p->sched_reset_on_fork = reset_on_fork;
3875
		task_rq_unlock(rq, p, &flags);
3876 3877
		return 0;
	}
3878
change:
3879

3880
	if (user) {
3881
#ifdef CONFIG_RT_GROUP_SCHED
3882 3883 3884 3885 3886
		/*
		 * Do not allow realtime tasks into groups that have no runtime
		 * assigned.
		 */
		if (rt_bandwidth_enabled() && rt_policy(policy) &&
3887 3888
				task_group(p)->rt_bandwidth.rt_runtime == 0 &&
				!task_group_is_autogroup(task_group(p))) {
3889
			task_rq_unlock(rq, p, &flags);
3890 3891 3892
			return -EPERM;
		}
#endif
3893 3894 3895 3896 3897 3898 3899 3900 3901
#ifdef CONFIG_SMP
		if (dl_bandwidth_enabled() && dl_policy(policy)) {
			cpumask_t *span = rq->rd->span;

			/*
			 * Don't allow tasks with an affinity mask smaller than
			 * the entire root_domain to become SCHED_DEADLINE. We
			 * will also fail if there's no bandwidth available.
			 */
3902 3903
			if (!cpumask_subset(span, &p->cpus_allowed) ||
			    rq->rd->dl_bw.bw == 0) {
3904 3905 3906 3907 3908 3909
				task_rq_unlock(rq, p, &flags);
				return -EPERM;
			}
		}
#endif
	}
3910

L
Linus Torvalds 已提交
3911 3912 3913
	/* recheck policy now with rq lock held */
	if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) {
		policy = oldpolicy = -1;
3914
		task_rq_unlock(rq, p, &flags);
L
Linus Torvalds 已提交
3915 3916
		goto recheck;
	}
3917 3918 3919 3920 3921 3922

	/*
	 * If setscheduling to SCHED_DEADLINE (or changing the parameters
	 * of a SCHED_DEADLINE task) we need to check if enough bandwidth
	 * is available.
	 */
3923
	if ((dl_policy(policy) || dl_task(p)) && dl_overflow(p, policy, attr)) {
3924 3925 3926 3927
		task_rq_unlock(rq, p, &flags);
		return -EBUSY;
	}

3928 3929 3930
	p->sched_reset_on_fork = reset_on_fork;
	oldprio = p->prio;

3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944
	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.
		 */
		new_effective_prio = rt_mutex_get_effective_prio(p, newprio);
		if (new_effective_prio == oldprio) {
			__setscheduler_params(p, attr);
			task_rq_unlock(rq, p, &flags);
			return 0;
		}
3945 3946
	}

3947
	queued = task_on_rq_queued(p);
3948
	running = task_current(rq, p);
3949
	if (queued)
3950
		dequeue_task(rq, p, 0);
3951
	if (running)
3952
		put_prev_task(rq, p);
3953

3954
	prev_class = p->sched_class;
3955
	__setscheduler(rq, p, attr, pi);
3956

3957 3958
	if (running)
		p->sched_class->set_curr_task(rq);
3959
	if (queued) {
3960 3961 3962 3963 3964 3965
		/*
		 * We enqueue to tail when the priority of a task is
		 * increased (user space view).
		 */
		enqueue_task(rq, p, oldprio <= p->prio ? ENQUEUE_HEAD : 0);
	}
3966

P
Peter Zijlstra 已提交
3967
	check_class_changed(rq, p, prev_class, oldprio);
3968
	preempt_disable(); /* avoid rq from going away on us */
3969
	task_rq_unlock(rq, p, &flags);
3970

3971 3972
	if (pi)
		rt_mutex_adjust_pi(p);
3973

3974 3975 3976 3977 3978
	/*
	 * Run balance callbacks after we've adjusted the PI chain.
	 */
	balance_callback(rq);
	preempt_enable();
3979

L
Linus Torvalds 已提交
3980 3981
	return 0;
}
3982

3983 3984 3985 3986 3987 3988 3989 3990 3991
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),
	};

3992 3993
	/* Fixup the legacy SCHED_RESET_ON_FORK hack. */
	if ((policy != SETPARAM_POLICY) && (policy & SCHED_RESET_ON_FORK)) {
3994 3995 3996 3997 3998
		attr.sched_flags |= SCHED_FLAG_RESET_ON_FORK;
		policy &= ~SCHED_RESET_ON_FORK;
		attr.sched_policy = policy;
	}

3999
	return __sched_setscheduler(p, &attr, check, true);
4000
}
4001 4002 4003 4004 4005 4006
/**
 * 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.
 *
4007 4008
 * Return: 0 on success. An error code otherwise.
 *
4009 4010 4011
 * NOTE that the task may be already dead.
 */
int sched_setscheduler(struct task_struct *p, int policy,
4012
		       const struct sched_param *param)
4013
{
4014
	return _sched_setscheduler(p, policy, param, true);
4015
}
L
Linus Torvalds 已提交
4016 4017
EXPORT_SYMBOL_GPL(sched_setscheduler);

4018 4019
int sched_setattr(struct task_struct *p, const struct sched_attr *attr)
{
4020
	return __sched_setscheduler(p, attr, true, true);
4021 4022 4023
}
EXPORT_SYMBOL_GPL(sched_setattr);

4024 4025 4026 4027 4028 4029 4030 4031 4032 4033
/**
 * 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.
4034 4035
 *
 * Return: 0 on success. An error code otherwise.
4036 4037
 */
int sched_setscheduler_nocheck(struct task_struct *p, int policy,
4038
			       const struct sched_param *param)
4039
{
4040
	return _sched_setscheduler(p, policy, param, false);
4041 4042
}

I
Ingo Molnar 已提交
4043 4044
static int
do_sched_setscheduler(pid_t pid, int policy, struct sched_param __user *param)
L
Linus Torvalds 已提交
4045 4046 4047
{
	struct sched_param lparam;
	struct task_struct *p;
4048
	int retval;
L
Linus Torvalds 已提交
4049 4050 4051 4052 4053

	if (!param || pid < 0)
		return -EINVAL;
	if (copy_from_user(&lparam, param, sizeof(struct sched_param)))
		return -EFAULT;
4054 4055 4056

	rcu_read_lock();
	retval = -ESRCH;
L
Linus Torvalds 已提交
4057
	p = find_process_by_pid(pid);
4058 4059 4060
	if (p != NULL)
		retval = sched_setscheduler(p, policy, &lparam);
	rcu_read_unlock();
4061

L
Linus Torvalds 已提交
4062 4063 4064
	return retval;
}

4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126
/*
 * Mimics kernel/events/core.c perf_copy_attr().
 */
static int sched_copy_attr(struct sched_attr __user *uattr,
			   struct sched_attr *attr)
{
	u32 size;
	int ret;

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

	/*
	 * zero the full structure, so that a short copy will be nice.
	 */
	memset(attr, 0, sizeof(*attr));

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

	if (size > PAGE_SIZE)	/* silly large */
		goto err_size;

	if (!size)		/* abi compat */
		size = SCHED_ATTR_SIZE_VER0;

	if (size < SCHED_ATTR_SIZE_VER0)
		goto err_size;

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

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

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

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

	/*
	 * XXX: do we want to be lenient like existing syscalls; or do we want
	 * to be strict and return an error on out-of-bounds values?
	 */
4127
	attr->sched_nice = clamp(attr->sched_nice, MIN_NICE, MAX_NICE);
4128

4129
	return 0;
4130 4131 4132

err_size:
	put_user(sizeof(*attr), &uattr->size);
4133
	return -E2BIG;
4134 4135
}

L
Linus Torvalds 已提交
4136 4137 4138 4139 4140
/**
 * 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.
4141 4142
 *
 * Return: 0 on success. An error code otherwise.
L
Linus Torvalds 已提交
4143
 */
4144 4145
SYSCALL_DEFINE3(sched_setscheduler, pid_t, pid, int, policy,
		struct sched_param __user *, param)
L
Linus Torvalds 已提交
4146
{
4147 4148 4149 4150
	/* negative values for policy are not valid */
	if (policy < 0)
		return -EINVAL;

L
Linus Torvalds 已提交
4151 4152 4153 4154 4155 4156 4157
	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.
4158 4159
 *
 * Return: 0 on success. An error code otherwise.
L
Linus Torvalds 已提交
4160
 */
4161
SYSCALL_DEFINE2(sched_setparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4162
{
4163
	return do_sched_setscheduler(pid, SETPARAM_POLICY, param);
L
Linus Torvalds 已提交
4164 4165
}

4166 4167 4168
/**
 * sys_sched_setattr - same as above, but with extended sched_attr
 * @pid: the pid in question.
J
Juri Lelli 已提交
4169
 * @uattr: structure containing the extended parameters.
4170
 * @flags: for future extension.
4171
 */
4172 4173
SYSCALL_DEFINE3(sched_setattr, pid_t, pid, struct sched_attr __user *, uattr,
			       unsigned int, flags)
4174 4175 4176 4177 4178
{
	struct sched_attr attr;
	struct task_struct *p;
	int retval;

4179
	if (!uattr || pid < 0 || flags)
4180 4181
		return -EINVAL;

4182 4183 4184
	retval = sched_copy_attr(uattr, &attr);
	if (retval)
		return retval;
4185

4186
	if ((int)attr.sched_policy < 0)
4187
		return -EINVAL;
4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198

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

	return retval;
}

L
Linus Torvalds 已提交
4199 4200 4201
/**
 * sys_sched_getscheduler - get the policy (scheduling class) of a thread
 * @pid: the pid in question.
4202 4203 4204
 *
 * Return: On success, the policy of the thread. Otherwise, a negative error
 * code.
L
Linus Torvalds 已提交
4205
 */
4206
SYSCALL_DEFINE1(sched_getscheduler, pid_t, pid)
L
Linus Torvalds 已提交
4207
{
4208
	struct task_struct *p;
4209
	int retval;
L
Linus Torvalds 已提交
4210 4211

	if (pid < 0)
4212
		return -EINVAL;
L
Linus Torvalds 已提交
4213 4214

	retval = -ESRCH;
4215
	rcu_read_lock();
L
Linus Torvalds 已提交
4216 4217 4218 4219
	p = find_process_by_pid(pid);
	if (p) {
		retval = security_task_getscheduler(p);
		if (!retval)
4220 4221
			retval = p->policy
				| (p->sched_reset_on_fork ? SCHED_RESET_ON_FORK : 0);
L
Linus Torvalds 已提交
4222
	}
4223
	rcu_read_unlock();
L
Linus Torvalds 已提交
4224 4225 4226 4227
	return retval;
}

/**
4228
 * sys_sched_getparam - get the RT priority of a thread
L
Linus Torvalds 已提交
4229 4230
 * @pid: the pid in question.
 * @param: structure containing the RT priority.
4231 4232 4233
 *
 * Return: On success, 0 and the RT priority is in @param. Otherwise, an error
 * code.
L
Linus Torvalds 已提交
4234
 */
4235
SYSCALL_DEFINE2(sched_getparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4236
{
4237
	struct sched_param lp = { .sched_priority = 0 };
4238
	struct task_struct *p;
4239
	int retval;
L
Linus Torvalds 已提交
4240 4241

	if (!param || pid < 0)
4242
		return -EINVAL;
L
Linus Torvalds 已提交
4243

4244
	rcu_read_lock();
L
Linus Torvalds 已提交
4245 4246 4247 4248 4249 4250 4251 4252 4253
	p = find_process_by_pid(pid);
	retval = -ESRCH;
	if (!p)
		goto out_unlock;

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

4254 4255
	if (task_has_rt_policy(p))
		lp.sched_priority = p->rt_priority;
4256
	rcu_read_unlock();
L
Linus Torvalds 已提交
4257 4258 4259 4260 4261 4262 4263 4264 4265

	/*
	 * 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:
4266
	rcu_read_unlock();
L
Linus Torvalds 已提交
4267 4268 4269
	return retval;
}

4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292
static int sched_read_attr(struct sched_attr __user *uattr,
			   struct sched_attr *attr,
			   unsigned int usize)
{
	int ret;

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

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

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

		for (; addr < end; addr++) {
			if (*addr)
4293
				return -EFBIG;
4294 4295 4296 4297 4298
		}

		attr->size = usize;
	}

4299
	ret = copy_to_user(uattr, attr, attr->size);
4300 4301 4302
	if (ret)
		return -EFAULT;

4303
	return 0;
4304 4305 4306
}

/**
4307
 * sys_sched_getattr - similar to sched_getparam, but with sched_attr
4308
 * @pid: the pid in question.
J
Juri Lelli 已提交
4309
 * @uattr: structure containing the extended parameters.
4310
 * @size: sizeof(attr) for fwd/bwd comp.
4311
 * @flags: for future extension.
4312
 */
4313 4314
SYSCALL_DEFINE4(sched_getattr, pid_t, pid, struct sched_attr __user *, uattr,
		unsigned int, size, unsigned int, flags)
4315 4316 4317 4318 4319 4320 4321 4322
{
	struct sched_attr attr = {
		.size = sizeof(struct sched_attr),
	};
	struct task_struct *p;
	int retval;

	if (!uattr || pid < 0 || size > PAGE_SIZE ||
4323
	    size < SCHED_ATTR_SIZE_VER0 || flags)
4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334 4335 4336
		return -EINVAL;

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

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

	attr.sched_policy = p->policy;
4337 4338
	if (p->sched_reset_on_fork)
		attr.sched_flags |= SCHED_FLAG_RESET_ON_FORK;
4339 4340 4341
	if (task_has_dl_policy(p))
		__getparam_dl(p, &attr);
	else if (task_has_rt_policy(p))
4342 4343
		attr.sched_priority = p->rt_priority;
	else
4344
		attr.sched_nice = task_nice(p);
4345 4346 4347 4348 4349 4350 4351 4352 4353 4354 4355

	rcu_read_unlock();

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

out_unlock:
	rcu_read_unlock();
	return retval;
}

4356
long sched_setaffinity(pid_t pid, const struct cpumask *in_mask)
L
Linus Torvalds 已提交
4357
{
4358
	cpumask_var_t cpus_allowed, new_mask;
4359 4360
	struct task_struct *p;
	int retval;
L
Linus Torvalds 已提交
4361

4362
	rcu_read_lock();
L
Linus Torvalds 已提交
4363 4364 4365

	p = find_process_by_pid(pid);
	if (!p) {
4366
		rcu_read_unlock();
L
Linus Torvalds 已提交
4367 4368 4369
		return -ESRCH;
	}

4370
	/* Prevent p going away */
L
Linus Torvalds 已提交
4371
	get_task_struct(p);
4372
	rcu_read_unlock();
L
Linus Torvalds 已提交
4373

4374 4375 4376 4377
	if (p->flags & PF_NO_SETAFFINITY) {
		retval = -EINVAL;
		goto out_put_task;
	}
4378 4379 4380 4381 4382 4383 4384 4385
	if (!alloc_cpumask_var(&cpus_allowed, GFP_KERNEL)) {
		retval = -ENOMEM;
		goto out_put_task;
	}
	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL)) {
		retval = -ENOMEM;
		goto out_free_cpus_allowed;
	}
L
Linus Torvalds 已提交
4386
	retval = -EPERM;
E
Eric W. Biederman 已提交
4387 4388 4389 4390
	if (!check_same_owner(p)) {
		rcu_read_lock();
		if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
			rcu_read_unlock();
4391
			goto out_free_new_mask;
E
Eric W. Biederman 已提交
4392 4393 4394
		}
		rcu_read_unlock();
	}
L
Linus Torvalds 已提交
4395

4396
	retval = security_task_setscheduler(p);
4397
	if (retval)
4398
		goto out_free_new_mask;
4399

4400 4401 4402 4403

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

4404 4405 4406 4407 4408 4409 4410
	/*
	 * Since bandwidth control happens on root_domain basis,
	 * if admission test is enabled, we only admit -deadline
	 * tasks allowed to run on all the CPUs in the task's
	 * root_domain.
	 */
#ifdef CONFIG_SMP
4411 4412 4413
	if (task_has_dl_policy(p) && dl_bandwidth_enabled()) {
		rcu_read_lock();
		if (!cpumask_subset(task_rq(p)->rd->span, new_mask)) {
4414
			retval = -EBUSY;
4415
			rcu_read_unlock();
4416
			goto out_free_new_mask;
4417
		}
4418
		rcu_read_unlock();
4419 4420
	}
#endif
P
Peter Zijlstra 已提交
4421
again:
4422
	retval = __set_cpus_allowed_ptr(p, new_mask, true);
L
Linus Torvalds 已提交
4423

P
Paul Menage 已提交
4424
	if (!retval) {
4425 4426
		cpuset_cpus_allowed(p, cpus_allowed);
		if (!cpumask_subset(new_mask, cpus_allowed)) {
P
Paul Menage 已提交
4427 4428 4429 4430 4431
			/*
			 * We must have raced with a concurrent cpuset
			 * update. Just reset the cpus_allowed to the
			 * cpuset's cpus_allowed
			 */
4432
			cpumask_copy(new_mask, cpus_allowed);
P
Paul Menage 已提交
4433 4434 4435
			goto again;
		}
	}
4436
out_free_new_mask:
4437 4438 4439 4440
	free_cpumask_var(new_mask);
out_free_cpus_allowed:
	free_cpumask_var(cpus_allowed);
out_put_task:
L
Linus Torvalds 已提交
4441 4442 4443 4444 4445
	put_task_struct(p);
	return retval;
}

static int get_user_cpu_mask(unsigned long __user *user_mask_ptr, unsigned len,
4446
			     struct cpumask *new_mask)
L
Linus Torvalds 已提交
4447
{
4448 4449 4450 4451 4452
	if (len < cpumask_size())
		cpumask_clear(new_mask);
	else if (len > cpumask_size())
		len = cpumask_size();

L
Linus Torvalds 已提交
4453 4454 4455 4456 4457 4458 4459 4460
	return copy_from_user(new_mask, user_mask_ptr, len) ? -EFAULT : 0;
}

/**
 * sys_sched_setaffinity - set the cpu affinity of a process
 * @pid: pid of the process
 * @len: length in bytes of the bitmask pointed to by user_mask_ptr
 * @user_mask_ptr: user-space pointer to the new cpu mask
4461 4462
 *
 * Return: 0 on success. An error code otherwise.
L
Linus Torvalds 已提交
4463
 */
4464 4465
SYSCALL_DEFINE3(sched_setaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
4466
{
4467
	cpumask_var_t new_mask;
L
Linus Torvalds 已提交
4468 4469
	int retval;

4470 4471
	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
4472

4473 4474 4475 4476 4477
	retval = get_user_cpu_mask(user_mask_ptr, len, new_mask);
	if (retval == 0)
		retval = sched_setaffinity(pid, new_mask);
	free_cpumask_var(new_mask);
	return retval;
L
Linus Torvalds 已提交
4478 4479
}

4480
long sched_getaffinity(pid_t pid, struct cpumask *mask)
L
Linus Torvalds 已提交
4481
{
4482
	struct task_struct *p;
4483
	unsigned long flags;
L
Linus Torvalds 已提交
4484 4485
	int retval;

4486
	rcu_read_lock();
L
Linus Torvalds 已提交
4487 4488 4489 4490 4491 4492

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

4493 4494 4495 4496
	retval = security_task_getscheduler(p);
	if (retval)
		goto out_unlock;

4497
	raw_spin_lock_irqsave(&p->pi_lock, flags);
4498
	cpumask_and(mask, &p->cpus_allowed, cpu_active_mask);
4499
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
4500 4501

out_unlock:
4502
	rcu_read_unlock();
L
Linus Torvalds 已提交
4503

4504
	return retval;
L
Linus Torvalds 已提交
4505 4506 4507 4508 4509 4510 4511
}

/**
 * sys_sched_getaffinity - get the cpu affinity of a process
 * @pid: pid of the process
 * @len: length in bytes of the bitmask pointed to by user_mask_ptr
 * @user_mask_ptr: user-space pointer to hold the current cpu mask
4512 4513
 *
 * Return: 0 on success. An error code otherwise.
L
Linus Torvalds 已提交
4514
 */
4515 4516
SYSCALL_DEFINE3(sched_getaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
4517 4518
{
	int ret;
4519
	cpumask_var_t mask;
L
Linus Torvalds 已提交
4520

A
Anton Blanchard 已提交
4521
	if ((len * BITS_PER_BYTE) < nr_cpu_ids)
4522 4523
		return -EINVAL;
	if (len & (sizeof(unsigned long)-1))
L
Linus Torvalds 已提交
4524 4525
		return -EINVAL;

4526 4527
	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
4528

4529 4530
	ret = sched_getaffinity(pid, mask);
	if (ret == 0) {
4531
		size_t retlen = min_t(size_t, len, cpumask_size());
4532 4533

		if (copy_to_user(user_mask_ptr, mask, retlen))
4534 4535
			ret = -EFAULT;
		else
4536
			ret = retlen;
4537 4538
	}
	free_cpumask_var(mask);
L
Linus Torvalds 已提交
4539

4540
	return ret;
L
Linus Torvalds 已提交
4541 4542 4543 4544 4545
}

/**
 * sys_sched_yield - yield the current processor to other threads.
 *
I
Ingo Molnar 已提交
4546 4547
 * This function yields the current CPU to other tasks. If there are no
 * other threads running on this CPU then this function will return.
4548 4549
 *
 * Return: 0.
L
Linus Torvalds 已提交
4550
 */
4551
SYSCALL_DEFINE0(sched_yield)
L
Linus Torvalds 已提交
4552
{
4553
	struct rq *rq = this_rq_lock();
L
Linus Torvalds 已提交
4554

4555
	schedstat_inc(rq, yld_count);
4556
	current->sched_class->yield_task(rq);
L
Linus Torvalds 已提交
4557 4558 4559 4560 4561 4562

	/*
	 * Since we are going to call schedule() anyway, there's
	 * no need to preempt or enable interrupts:
	 */
	__release(rq->lock);
4563
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
4564
	do_raw_spin_unlock(&rq->lock);
4565
	sched_preempt_enable_no_resched();
L
Linus Torvalds 已提交
4566 4567 4568 4569 4570 4571

	schedule();

	return 0;
}

4572
int __sched _cond_resched(void)
L
Linus Torvalds 已提交
4573
{
4574
	if (should_resched(0)) {
4575
		preempt_schedule_common();
L
Linus Torvalds 已提交
4576 4577 4578 4579
		return 1;
	}
	return 0;
}
4580
EXPORT_SYMBOL(_cond_resched);
L
Linus Torvalds 已提交
4581 4582

/*
4583
 * __cond_resched_lock() - if a reschedule is pending, drop the given lock,
L
Linus Torvalds 已提交
4584 4585
 * call schedule, and on return reacquire the lock.
 *
I
Ingo Molnar 已提交
4586
 * This works OK both with and without CONFIG_PREEMPT. We do strange low-level
L
Linus Torvalds 已提交
4587 4588 4589
 * operations here to prevent schedule() from being called twice (once via
 * spin_unlock(), once by hand).
 */
4590
int __cond_resched_lock(spinlock_t *lock)
L
Linus Torvalds 已提交
4591
{
4592
	int resched = should_resched(PREEMPT_LOCK_OFFSET);
J
Jan Kara 已提交
4593 4594
	int ret = 0;

4595 4596
	lockdep_assert_held(lock);

4597
	if (spin_needbreak(lock) || resched) {
L
Linus Torvalds 已提交
4598
		spin_unlock(lock);
P
Peter Zijlstra 已提交
4599
		if (resched)
4600
			preempt_schedule_common();
N
Nick Piggin 已提交
4601 4602
		else
			cpu_relax();
J
Jan Kara 已提交
4603
		ret = 1;
L
Linus Torvalds 已提交
4604 4605
		spin_lock(lock);
	}
J
Jan Kara 已提交
4606
	return ret;
L
Linus Torvalds 已提交
4607
}
4608
EXPORT_SYMBOL(__cond_resched_lock);
L
Linus Torvalds 已提交
4609

4610
int __sched __cond_resched_softirq(void)
L
Linus Torvalds 已提交
4611 4612 4613
{
	BUG_ON(!in_softirq());

4614
	if (should_resched(SOFTIRQ_DISABLE_OFFSET)) {
4615
		local_bh_enable();
4616
		preempt_schedule_common();
L
Linus Torvalds 已提交
4617 4618 4619 4620 4621
		local_bh_disable();
		return 1;
	}
	return 0;
}
4622
EXPORT_SYMBOL(__cond_resched_softirq);
L
Linus Torvalds 已提交
4623 4624 4625 4626

/**
 * yield - yield the current processor to other threads.
 *
P
Peter Zijlstra 已提交
4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643 4644
 * Do not ever use this function, there's a 99% chance you're doing it wrong.
 *
 * The scheduler is at all times free to pick the calling task as the most
 * eligible task to run, if removing the yield() call from your code breaks
 * it, its already broken.
 *
 * Typical broken usage is:
 *
 * while (!event)
 * 	yield();
 *
 * where one assumes that yield() will let 'the other' process run that will
 * make event true. If the current task is a SCHED_FIFO task that will never
 * happen. Never use yield() as a progress guarantee!!
 *
 * If you want to use yield() to wait for something, use wait_event().
 * If you want to use yield() to be 'nice' for others, use cond_resched().
 * If you still want to use yield(), do not!
L
Linus Torvalds 已提交
4645 4646 4647 4648 4649 4650 4651 4652
 */
void __sched yield(void)
{
	set_current_state(TASK_RUNNING);
	sys_sched_yield();
}
EXPORT_SYMBOL(yield);

4653 4654 4655 4656
/**
 * yield_to - yield the current processor to another thread in
 * your thread group, or accelerate that thread toward the
 * processor it's on.
R
Randy Dunlap 已提交
4657 4658
 * @p: target task
 * @preempt: whether task preemption is allowed or not
4659 4660 4661 4662
 *
 * 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.
 *
4663
 * Return:
4664 4665 4666
 *	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.
4667
 */
4668
int __sched yield_to(struct task_struct *p, bool preempt)
4669 4670 4671 4672
{
	struct task_struct *curr = current;
	struct rq *rq, *p_rq;
	unsigned long flags;
4673
	int yielded = 0;
4674 4675 4676 4677 4678 4679

	local_irq_save(flags);
	rq = this_rq();

again:
	p_rq = task_rq(p);
4680 4681 4682 4683 4684 4685 4686 4687 4688
	/*
	 * 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;
	}

4689
	double_rq_lock(rq, p_rq);
4690
	if (task_rq(p) != p_rq) {
4691 4692 4693 4694 4695
		double_rq_unlock(rq, p_rq);
		goto again;
	}

	if (!curr->sched_class->yield_to_task)
4696
		goto out_unlock;
4697 4698

	if (curr->sched_class != p->sched_class)
4699
		goto out_unlock;
4700 4701

	if (task_running(p_rq, p) || p->state)
4702
		goto out_unlock;
4703 4704

	yielded = curr->sched_class->yield_to_task(rq, p, preempt);
4705
	if (yielded) {
4706
		schedstat_inc(rq, yld_count);
4707 4708 4709 4710 4711
		/*
		 * Make p's CPU reschedule; pick_next_entity takes care of
		 * fairness.
		 */
		if (preempt && rq != p_rq)
4712
			resched_curr(p_rq);
4713
	}
4714

4715
out_unlock:
4716
	double_rq_unlock(rq, p_rq);
4717
out_irq:
4718 4719
	local_irq_restore(flags);

4720
	if (yielded > 0)
4721 4722 4723 4724 4725 4726
		schedule();

	return yielded;
}
EXPORT_SYMBOL_GPL(yield_to);

L
Linus Torvalds 已提交
4727
/*
I
Ingo Molnar 已提交
4728
 * This task is about to go to sleep on IO. Increment rq->nr_iowait so
L
Linus Torvalds 已提交
4729 4730 4731 4732
 * that process accounting knows that this is a task in IO wait state.
 */
long __sched io_schedule_timeout(long timeout)
{
4733 4734
	int old_iowait = current->in_iowait;
	struct rq *rq;
L
Linus Torvalds 已提交
4735 4736
	long ret;

4737
	current->in_iowait = 1;
4738
	blk_schedule_flush_plug(current);
4739

4740
	delayacct_blkio_start();
4741
	rq = raw_rq();
L
Linus Torvalds 已提交
4742 4743
	atomic_inc(&rq->nr_iowait);
	ret = schedule_timeout(timeout);
4744
	current->in_iowait = old_iowait;
L
Linus Torvalds 已提交
4745
	atomic_dec(&rq->nr_iowait);
4746
	delayacct_blkio_end();
4747

L
Linus Torvalds 已提交
4748 4749
	return ret;
}
4750
EXPORT_SYMBOL(io_schedule_timeout);
L
Linus Torvalds 已提交
4751 4752 4753 4754 4755

/**
 * sys_sched_get_priority_max - return maximum RT priority.
 * @policy: scheduling class.
 *
4756 4757 4758
 * Return: On success, this syscall returns the maximum
 * rt_priority that can be used by a given scheduling class.
 * On failure, a negative error code is returned.
L
Linus Torvalds 已提交
4759
 */
4760
SYSCALL_DEFINE1(sched_get_priority_max, int, policy)
L
Linus Torvalds 已提交
4761 4762 4763 4764 4765 4766 4767 4768
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = MAX_USER_RT_PRIO-1;
		break;
4769
	case SCHED_DEADLINE:
L
Linus Torvalds 已提交
4770
	case SCHED_NORMAL:
4771
	case SCHED_BATCH:
I
Ingo Molnar 已提交
4772
	case SCHED_IDLE:
L
Linus Torvalds 已提交
4773 4774 4775 4776 4777 4778 4779 4780 4781 4782
		ret = 0;
		break;
	}
	return ret;
}

/**
 * sys_sched_get_priority_min - return minimum RT priority.
 * @policy: scheduling class.
 *
4783 4784 4785
 * Return: On success, this syscall returns the minimum
 * rt_priority that can be used by a given scheduling class.
 * On failure, a negative error code is returned.
L
Linus Torvalds 已提交
4786
 */
4787
SYSCALL_DEFINE1(sched_get_priority_min, int, policy)
L
Linus Torvalds 已提交
4788 4789 4790 4791 4792 4793 4794 4795
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = 1;
		break;
4796
	case SCHED_DEADLINE:
L
Linus Torvalds 已提交
4797
	case SCHED_NORMAL:
4798
	case SCHED_BATCH:
I
Ingo Molnar 已提交
4799
	case SCHED_IDLE:
L
Linus Torvalds 已提交
4800 4801 4802 4803 4804 4805 4806 4807 4808 4809 4810 4811
		ret = 0;
	}
	return ret;
}

/**
 * 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.
4812 4813 4814
 *
 * Return: On success, 0 and the timeslice is in @interval. Otherwise,
 * an error code.
L
Linus Torvalds 已提交
4815
 */
4816
SYSCALL_DEFINE2(sched_rr_get_interval, pid_t, pid,
4817
		struct timespec __user *, interval)
L
Linus Torvalds 已提交
4818
{
4819
	struct task_struct *p;
D
Dmitry Adamushko 已提交
4820
	unsigned int time_slice;
4821 4822
	unsigned long flags;
	struct rq *rq;
4823
	int retval;
L
Linus Torvalds 已提交
4824 4825 4826
	struct timespec t;

	if (pid < 0)
4827
		return -EINVAL;
L
Linus Torvalds 已提交
4828 4829

	retval = -ESRCH;
4830
	rcu_read_lock();
L
Linus Torvalds 已提交
4831 4832 4833 4834 4835 4836 4837 4838
	p = find_process_by_pid(pid);
	if (!p)
		goto out_unlock;

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

4839
	rq = task_rq_lock(p, &flags);
4840 4841 4842
	time_slice = 0;
	if (p->sched_class->get_rr_interval)
		time_slice = p->sched_class->get_rr_interval(rq, p);
4843
	task_rq_unlock(rq, p, &flags);
D
Dmitry Adamushko 已提交
4844

4845
	rcu_read_unlock();
D
Dmitry Adamushko 已提交
4846
	jiffies_to_timespec(time_slice, &t);
L
Linus Torvalds 已提交
4847 4848
	retval = copy_to_user(interval, &t, sizeof(t)) ? -EFAULT : 0;
	return retval;
4849

L
Linus Torvalds 已提交
4850
out_unlock:
4851
	rcu_read_unlock();
L
Linus Torvalds 已提交
4852 4853 4854
	return retval;
}

4855
static const char stat_nam[] = TASK_STATE_TO_CHAR_STR;
4856

4857
void sched_show_task(struct task_struct *p)
L
Linus Torvalds 已提交
4858 4859
{
	unsigned long free = 0;
4860
	int ppid;
4861
	unsigned long state = p->state;
L
Linus Torvalds 已提交
4862

4863 4864
	if (state)
		state = __ffs(state) + 1;
4865
	printk(KERN_INFO "%-15.15s %c", p->comm,
4866
		state < sizeof(stat_nam) - 1 ? stat_nam[state] : '?');
4867
#if BITS_PER_LONG == 32
L
Linus Torvalds 已提交
4868
	if (state == TASK_RUNNING)
P
Peter Zijlstra 已提交
4869
		printk(KERN_CONT " running  ");
L
Linus Torvalds 已提交
4870
	else
P
Peter Zijlstra 已提交
4871
		printk(KERN_CONT " %08lx ", thread_saved_pc(p));
L
Linus Torvalds 已提交
4872 4873
#else
	if (state == TASK_RUNNING)
P
Peter Zijlstra 已提交
4874
		printk(KERN_CONT "  running task    ");
L
Linus Torvalds 已提交
4875
	else
P
Peter Zijlstra 已提交
4876
		printk(KERN_CONT " %016lx ", thread_saved_pc(p));
L
Linus Torvalds 已提交
4877 4878
#endif
#ifdef CONFIG_DEBUG_STACK_USAGE
4879
	free = stack_not_used(p);
L
Linus Torvalds 已提交
4880
#endif
4881
	ppid = 0;
4882
	rcu_read_lock();
4883 4884
	if (pid_alive(p))
		ppid = task_pid_nr(rcu_dereference(p->real_parent));
4885
	rcu_read_unlock();
P
Peter Zijlstra 已提交
4886
	printk(KERN_CONT "%5lu %5d %6d 0x%08lx\n", free,
4887
		task_pid_nr(p), ppid,
4888
		(unsigned long)task_thread_info(p)->flags);
L
Linus Torvalds 已提交
4889

4890
	print_worker_info(KERN_INFO, p);
4891
	show_stack(p, NULL);
L
Linus Torvalds 已提交
4892 4893
}

I
Ingo Molnar 已提交
4894
void show_state_filter(unsigned long state_filter)
L
Linus Torvalds 已提交
4895
{
4896
	struct task_struct *g, *p;
L
Linus Torvalds 已提交
4897

4898
#if BITS_PER_LONG == 32
P
Peter Zijlstra 已提交
4899 4900
	printk(KERN_INFO
		"  task                PC stack   pid father\n");
L
Linus Torvalds 已提交
4901
#else
P
Peter Zijlstra 已提交
4902 4903
	printk(KERN_INFO
		"  task                        PC stack   pid father\n");
L
Linus Torvalds 已提交
4904
#endif
4905
	rcu_read_lock();
4906
	for_each_process_thread(g, p) {
L
Linus Torvalds 已提交
4907 4908
		/*
		 * reset the NMI-timeout, listing all files on a slow
L
Lucas De Marchi 已提交
4909
		 * console might take a lot of time:
L
Linus Torvalds 已提交
4910 4911
		 */
		touch_nmi_watchdog();
I
Ingo Molnar 已提交
4912
		if (!state_filter || (p->state & state_filter))
4913
			sched_show_task(p);
4914
	}
L
Linus Torvalds 已提交
4915

4916 4917
	touch_all_softlockup_watchdogs();

I
Ingo Molnar 已提交
4918 4919 4920
#ifdef CONFIG_SCHED_DEBUG
	sysrq_sched_debug_show();
#endif
4921
	rcu_read_unlock();
I
Ingo Molnar 已提交
4922 4923 4924
	/*
	 * Only show locks if all tasks are dumped:
	 */
4925
	if (!state_filter)
I
Ingo Molnar 已提交
4926
		debug_show_all_locks();
L
Linus Torvalds 已提交
4927 4928
}

4929
void init_idle_bootup_task(struct task_struct *idle)
I
Ingo Molnar 已提交
4930
{
I
Ingo Molnar 已提交
4931
	idle->sched_class = &idle_sched_class;
I
Ingo Molnar 已提交
4932 4933
}

4934 4935 4936 4937 4938 4939 4940 4941
/**
 * init_idle - set up an idle thread for a given CPU
 * @idle: task in question
 * @cpu: cpu the idle task belongs to
 *
 * NOTE: this function does not set the idle thread's NEED_RESCHED
 * flag, to make booting more robust.
 */
4942
void init_idle(struct task_struct *idle, int cpu)
L
Linus Torvalds 已提交
4943
{
4944
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
4945 4946
	unsigned long flags;

4947 4948
	raw_spin_lock_irqsave(&idle->pi_lock, flags);
	raw_spin_lock(&rq->lock);
4949

4950
	__sched_fork(0, idle);
4951
	idle->state = TASK_RUNNING;
I
Ingo Molnar 已提交
4952 4953
	idle->se.exec_start = sched_clock();

4954 4955 4956 4957 4958 4959 4960 4961 4962
#ifdef CONFIG_SMP
	/*
	 * Its possible that init_idle() gets called multiple times on a task,
	 * in that case do_set_cpus_allowed() will not do the right thing.
	 *
	 * And since this is boot we can forgo the serialization.
	 */
	set_cpus_allowed_common(idle, cpumask_of(cpu));
#endif
4963 4964 4965 4966 4967 4968 4969 4970 4971 4972 4973
	/*
	 * We're having a chicken and egg problem, even though we are
	 * holding rq->lock, the cpu isn't yet set to this cpu so the
	 * lockdep check in task_group() will fail.
	 *
	 * Similar case to sched_fork(). / Alternatively we could
	 * use task_rq_lock() here and obtain the other rq->lock.
	 *
	 * Silence PROVE_RCU
	 */
	rcu_read_lock();
I
Ingo Molnar 已提交
4974
	__set_task_cpu(idle, cpu);
4975
	rcu_read_unlock();
L
Linus Torvalds 已提交
4976 4977

	rq->curr = rq->idle = idle;
4978
	idle->on_rq = TASK_ON_RQ_QUEUED;
4979
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
4980
	idle->on_cpu = 1;
4981
#endif
4982 4983
	raw_spin_unlock(&rq->lock);
	raw_spin_unlock_irqrestore(&idle->pi_lock, flags);
L
Linus Torvalds 已提交
4984 4985

	/* Set the preempt count _outside_ the spinlocks! */
4986
	init_idle_preempt_count(idle, cpu);
4987

I
Ingo Molnar 已提交
4988 4989 4990 4991
	/*
	 * The idle tasks have their own, simple scheduling class:
	 */
	idle->sched_class = &idle_sched_class;
4992
	ftrace_graph_init_idle_task(idle, cpu);
4993
	vtime_init_idle(idle, cpu);
4994
#ifdef CONFIG_SMP
4995 4996
	sprintf(idle->comm, "%s/%d", INIT_TASK_COMM, cpu);
#endif
I
Ingo Molnar 已提交
4997 4998
}

4999 5000 5001 5002 5003 5004 5005
int cpuset_cpumask_can_shrink(const struct cpumask *cur,
			      const struct cpumask *trial)
{
	int ret = 1, trial_cpus;
	struct dl_bw *cur_dl_b;
	unsigned long flags;

5006 5007 5008
	if (!cpumask_weight(cur))
		return ret;

5009
	rcu_read_lock_sched();
5010 5011 5012 5013 5014 5015 5016 5017
	cur_dl_b = dl_bw_of(cpumask_any(cur));
	trial_cpus = cpumask_weight(trial);

	raw_spin_lock_irqsave(&cur_dl_b->lock, flags);
	if (cur_dl_b->bw != -1 &&
	    cur_dl_b->bw * trial_cpus < cur_dl_b->total_bw)
		ret = 0;
	raw_spin_unlock_irqrestore(&cur_dl_b->lock, flags);
5018
	rcu_read_unlock_sched();
5019 5020 5021 5022

	return ret;
}

5023 5024 5025 5026 5027 5028 5029 5030 5031 5032 5033 5034 5035 5036 5037 5038 5039 5040 5041 5042 5043 5044 5045 5046
int task_can_attach(struct task_struct *p,
		    const struct cpumask *cs_cpus_allowed)
{
	int ret = 0;

	/*
	 * Kthreads which disallow setaffinity shouldn't be moved
	 * to a new cpuset; we don't want to change their cpu
	 * affinity and isolating such threads by their set of
	 * allowed nodes is unnecessary.  Thus, cpusets are not
	 * applicable for such threads.  This prevents checking for
	 * success of set_cpus_allowed_ptr() on all attached tasks
	 * before cpus_allowed may be changed.
	 */
	if (p->flags & PF_NO_SETAFFINITY) {
		ret = -EINVAL;
		goto out;
	}

#ifdef CONFIG_SMP
	if (dl_task(p) && !cpumask_intersects(task_rq(p)->rd->span,
					      cs_cpus_allowed)) {
		unsigned int dest_cpu = cpumask_any_and(cpu_active_mask,
							cs_cpus_allowed);
5047
		struct dl_bw *dl_b;
5048 5049 5050 5051
		bool overflow;
		int cpus;
		unsigned long flags;

5052 5053
		rcu_read_lock_sched();
		dl_b = dl_bw_of(dest_cpu);
5054 5055 5056 5057 5058 5059 5060 5061 5062 5063 5064 5065 5066 5067 5068
		raw_spin_lock_irqsave(&dl_b->lock, flags);
		cpus = dl_bw_cpus(dest_cpu);
		overflow = __dl_overflow(dl_b, cpus, 0, p->dl.dl_bw);
		if (overflow)
			ret = -EBUSY;
		else {
			/*
			 * We reserve space for this task in the destination
			 * root_domain, as we can't fail after this point.
			 * We will free resources in the source root_domain
			 * later on (see set_cpus_allowed_dl()).
			 */
			__dl_add(dl_b, p->dl.dl_bw);
		}
		raw_spin_unlock_irqrestore(&dl_b->lock, flags);
5069
		rcu_read_unlock_sched();
5070 5071 5072 5073 5074 5075 5076

	}
#endif
out:
	return ret;
}

L
Linus Torvalds 已提交
5077 5078
#ifdef CONFIG_SMP

5079 5080 5081 5082 5083 5084 5085 5086 5087 5088 5089 5090 5091 5092 5093
#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;

	if (!cpumask_test_cpu(target_cpu, tsk_cpus_allowed(p)))
		return -EINVAL;

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

5094
	trace_sched_move_numa(p, curr_cpu, target_cpu);
5095 5096
	return stop_one_cpu(curr_cpu, migration_cpu_stop, &arg);
}
5097 5098 5099 5100 5101 5102 5103 5104 5105

/*
 * 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)
{
	struct rq *rq;
	unsigned long flags;
5106
	bool queued, running;
5107 5108

	rq = task_rq_lock(p, &flags);
5109
	queued = task_on_rq_queued(p);
5110 5111
	running = task_current(rq, p);

5112
	if (queued)
5113 5114
		dequeue_task(rq, p, 0);
	if (running)
5115
		put_prev_task(rq, p);
5116 5117 5118 5119 5120

	p->numa_preferred_nid = nid;

	if (running)
		p->sched_class->set_curr_task(rq);
5121
	if (queued)
5122 5123 5124
		enqueue_task(rq, p, 0);
	task_rq_unlock(rq, p, &flags);
}
P
Peter Zijlstra 已提交
5125
#endif /* CONFIG_NUMA_BALANCING */
5126

L
Linus Torvalds 已提交
5127
#ifdef CONFIG_HOTPLUG_CPU
5128
/*
5129 5130
 * Ensures that the idle task is using init_mm right before its cpu goes
 * offline.
5131
 */
5132
void idle_task_exit(void)
L
Linus Torvalds 已提交
5133
{
5134
	struct mm_struct *mm = current->active_mm;
5135

5136
	BUG_ON(cpu_online(smp_processor_id()));
5137

5138
	if (mm != &init_mm) {
5139
		switch_mm(mm, &init_mm, current);
5140 5141
		finish_arch_post_lock_switch();
	}
5142
	mmdrop(mm);
L
Linus Torvalds 已提交
5143 5144 5145
}

/*
5146 5147 5148 5149 5150
 * Since this CPU is going 'away' for a while, fold any nr_active delta
 * we might have. Assumes we're called after migrate_tasks() so that the
 * nr_active count is stable.
 *
 * Also see the comment "Global load-average calculations".
L
Linus Torvalds 已提交
5151
 */
5152
static void calc_load_migrate(struct rq *rq)
L
Linus Torvalds 已提交
5153
{
5154 5155 5156
	long delta = calc_load_fold_active(rq);
	if (delta)
		atomic_long_add(delta, &calc_load_tasks);
L
Linus Torvalds 已提交
5157 5158
}

5159 5160 5161 5162 5163 5164 5165 5166 5167 5168 5169 5170 5171 5172 5173 5174
static void put_prev_task_fake(struct rq *rq, struct task_struct *prev)
{
}

static const struct sched_class fake_sched_class = {
	.put_prev_task = put_prev_task_fake,
};

static struct task_struct fake_task = {
	/*
	 * Avoid pull_{rt,dl}_task()
	 */
	.prio = MAX_PRIO + 1,
	.sched_class = &fake_sched_class,
};

5175
/*
5176 5177 5178 5179 5180 5181
 * Migrate all tasks from the rq, sleeping tasks will be migrated by
 * try_to_wake_up()->select_task_rq().
 *
 * Called with rq->lock held even though we'er in stop_machine() and
 * there's no concurrency possible, we hold the required locks anyway
 * because of lock validation efforts.
L
Linus Torvalds 已提交
5182
 */
5183
static void migrate_tasks(struct rq *dead_rq)
L
Linus Torvalds 已提交
5184
{
5185
	struct rq *rq = dead_rq;
5186 5187
	struct task_struct *next, *stop = rq->stop;
	int dest_cpu;
L
Linus Torvalds 已提交
5188 5189

	/*
5190 5191 5192 5193 5194 5195 5196
	 * Fudge the rq selection such that the below task selection loop
	 * doesn't get stuck on the currently eligible stop task.
	 *
	 * We're currently inside stop_machine() and the rq is either stuck
	 * in the stop_machine_cpu_stop() loop, or we're executing this code,
	 * either way we should never end up calling schedule() until we're
	 * done here.
L
Linus Torvalds 已提交
5197
	 */
5198
	rq->stop = NULL;
5199

5200 5201 5202 5203 5204 5205 5206
	/*
	 * put_prev_task() and pick_next_task() sched
	 * class method both need to have an up-to-date
	 * value of rq->clock[_task]
	 */
	update_rq_clock(rq);

5207
	for (;;) {
5208 5209 5210 5211 5212
		/*
		 * There's this thread running, bail when that's the only
		 * remaining thread.
		 */
		if (rq->nr_running == 1)
I
Ingo Molnar 已提交
5213
			break;
5214

5215
		/*
W
Wanpeng Li 已提交
5216
		 * pick_next_task assumes pinned rq->lock.
5217 5218
		 */
		lockdep_pin_lock(&rq->lock);
5219
		next = pick_next_task(rq, &fake_task);
5220
		BUG_ON(!next);
D
Dmitry Adamushko 已提交
5221
		next->sched_class->put_prev_task(rq, next);
5222

W
Wanpeng Li 已提交
5223 5224 5225 5226 5227 5228 5229 5230 5231 5232 5233 5234 5235 5236 5237 5238 5239 5240 5241 5242 5243 5244 5245 5246
		/*
		 * Rules for changing task_struct::cpus_allowed are holding
		 * both pi_lock and rq->lock, such that holding either
		 * stabilizes the mask.
		 *
		 * Drop rq->lock is not quite as disastrous as it usually is
		 * because !cpu_active at this point, which means load-balance
		 * will not interfere. Also, stop-machine.
		 */
		lockdep_unpin_lock(&rq->lock);
		raw_spin_unlock(&rq->lock);
		raw_spin_lock(&next->pi_lock);
		raw_spin_lock(&rq->lock);

		/*
		 * Since we're inside stop-machine, _nothing_ should have
		 * changed the task, WARN if weird stuff happened, because in
		 * that case the above rq->lock drop is a fail too.
		 */
		if (WARN_ON(task_rq(next) != rq || !task_on_rq_queued(next))) {
			raw_spin_unlock(&next->pi_lock);
			continue;
		}

5247
		/* Find suitable destination for @next, with force if needed. */
5248
		dest_cpu = select_fallback_rq(dead_rq->cpu, next);
5249

5250 5251 5252 5253 5254 5255
		rq = __migrate_task(rq, next, dest_cpu);
		if (rq != dead_rq) {
			raw_spin_unlock(&rq->lock);
			rq = dead_rq;
			raw_spin_lock(&rq->lock);
		}
W
Wanpeng Li 已提交
5256
		raw_spin_unlock(&next->pi_lock);
L
Linus Torvalds 已提交
5257
	}
5258

5259
	rq->stop = stop;
5260
}
L
Linus Torvalds 已提交
5261 5262
#endif /* CONFIG_HOTPLUG_CPU */

5263 5264 5265
#if defined(CONFIG_SCHED_DEBUG) && defined(CONFIG_SYSCTL)

static struct ctl_table sd_ctl_dir[] = {
5266 5267
	{
		.procname	= "sched_domain",
5268
		.mode		= 0555,
5269
	},
5270
	{}
5271 5272 5273
};

static struct ctl_table sd_ctl_root[] = {
5274 5275
	{
		.procname	= "kernel",
5276
		.mode		= 0555,
5277 5278
		.child		= sd_ctl_dir,
	},
5279
	{}
5280 5281 5282 5283 5284
};

static struct ctl_table *sd_alloc_ctl_entry(int n)
{
	struct ctl_table *entry =
5285
		kcalloc(n, sizeof(struct ctl_table), GFP_KERNEL);
5286 5287 5288 5289

	return entry;
}

5290 5291
static void sd_free_ctl_entry(struct ctl_table **tablep)
{
5292
	struct ctl_table *entry;
5293

5294 5295 5296
	/*
	 * In the intermediate directories, both the child directory and
	 * procname are dynamically allocated and could fail but the mode
I
Ingo Molnar 已提交
5297
	 * will always be set. In the lowest directory the names are
5298 5299 5300
	 * static strings and all have proc handlers.
	 */
	for (entry = *tablep; entry->mode; entry++) {
5301 5302
		if (entry->child)
			sd_free_ctl_entry(&entry->child);
5303 5304 5305
		if (entry->proc_handler == NULL)
			kfree(entry->procname);
	}
5306 5307 5308 5309 5310

	kfree(*tablep);
	*tablep = NULL;
}

5311
static int min_load_idx = 0;
5312
static int max_load_idx = CPU_LOAD_IDX_MAX-1;
5313

5314
static void
5315
set_table_entry(struct ctl_table *entry,
5316
		const char *procname, void *data, int maxlen,
5317 5318
		umode_t mode, proc_handler *proc_handler,
		bool load_idx)
5319 5320 5321 5322 5323 5324
{
	entry->procname = procname;
	entry->data = data;
	entry->maxlen = maxlen;
	entry->mode = mode;
	entry->proc_handler = proc_handler;
5325 5326 5327 5328 5329

	if (load_idx) {
		entry->extra1 = &min_load_idx;
		entry->extra2 = &max_load_idx;
	}
5330 5331 5332 5333 5334
}

static struct ctl_table *
sd_alloc_ctl_domain_table(struct sched_domain *sd)
{
5335
	struct ctl_table *table = sd_alloc_ctl_entry(14);
5336

5337 5338 5339
	if (table == NULL)
		return NULL;

5340
	set_table_entry(&table[0], "min_interval", &sd->min_interval,
5341
		sizeof(long), 0644, proc_doulongvec_minmax, false);
5342
	set_table_entry(&table[1], "max_interval", &sd->max_interval,
5343
		sizeof(long), 0644, proc_doulongvec_minmax, false);
5344
	set_table_entry(&table[2], "busy_idx", &sd->busy_idx,
5345
		sizeof(int), 0644, proc_dointvec_minmax, true);
5346
	set_table_entry(&table[3], "idle_idx", &sd->idle_idx,
5347
		sizeof(int), 0644, proc_dointvec_minmax, true);
5348
	set_table_entry(&table[4], "newidle_idx", &sd->newidle_idx,
5349
		sizeof(int), 0644, proc_dointvec_minmax, true);
5350
	set_table_entry(&table[5], "wake_idx", &sd->wake_idx,
5351
		sizeof(int), 0644, proc_dointvec_minmax, true);
5352
	set_table_entry(&table[6], "forkexec_idx", &sd->forkexec_idx,
5353
		sizeof(int), 0644, proc_dointvec_minmax, true);
5354
	set_table_entry(&table[7], "busy_factor", &sd->busy_factor,
5355
		sizeof(int), 0644, proc_dointvec_minmax, false);
5356
	set_table_entry(&table[8], "imbalance_pct", &sd->imbalance_pct,
5357
		sizeof(int), 0644, proc_dointvec_minmax, false);
5358
	set_table_entry(&table[9], "cache_nice_tries",
5359
		&sd->cache_nice_tries,
5360
		sizeof(int), 0644, proc_dointvec_minmax, false);
5361
	set_table_entry(&table[10], "flags", &sd->flags,
5362
		sizeof(int), 0644, proc_dointvec_minmax, false);
5363 5364 5365 5366
	set_table_entry(&table[11], "max_newidle_lb_cost",
		&sd->max_newidle_lb_cost,
		sizeof(long), 0644, proc_doulongvec_minmax, false);
	set_table_entry(&table[12], "name", sd->name,
5367
		CORENAME_MAX_SIZE, 0444, proc_dostring, false);
5368
	/* &table[13] is terminator */
5369 5370 5371 5372

	return table;
}

5373
static struct ctl_table *sd_alloc_ctl_cpu_table(int cpu)
5374 5375 5376 5377 5378 5379 5380 5381 5382
{
	struct ctl_table *entry, *table;
	struct sched_domain *sd;
	int domain_num = 0, i;
	char buf[32];

	for_each_domain(cpu, sd)
		domain_num++;
	entry = table = sd_alloc_ctl_entry(domain_num + 1);
5383 5384
	if (table == NULL)
		return NULL;
5385 5386 5387 5388 5389

	i = 0;
	for_each_domain(cpu, sd) {
		snprintf(buf, 32, "domain%d", i);
		entry->procname = kstrdup(buf, GFP_KERNEL);
5390
		entry->mode = 0555;
5391 5392 5393 5394 5395 5396 5397 5398
		entry->child = sd_alloc_ctl_domain_table(sd);
		entry++;
		i++;
	}
	return table;
}

static struct ctl_table_header *sd_sysctl_header;
5399
static void register_sched_domain_sysctl(void)
5400
{
5401
	int i, cpu_num = num_possible_cpus();
5402 5403 5404
	struct ctl_table *entry = sd_alloc_ctl_entry(cpu_num + 1);
	char buf[32];

5405 5406 5407
	WARN_ON(sd_ctl_dir[0].child);
	sd_ctl_dir[0].child = entry;

5408 5409 5410
	if (entry == NULL)
		return;

5411
	for_each_possible_cpu(i) {
5412 5413
		snprintf(buf, 32, "cpu%d", i);
		entry->procname = kstrdup(buf, GFP_KERNEL);
5414
		entry->mode = 0555;
5415
		entry->child = sd_alloc_ctl_cpu_table(i);
5416
		entry++;
5417
	}
5418 5419

	WARN_ON(sd_sysctl_header);
5420 5421
	sd_sysctl_header = register_sysctl_table(sd_ctl_root);
}
5422

5423
/* may be called multiple times per register */
5424 5425
static void unregister_sched_domain_sysctl(void)
{
5426
	unregister_sysctl_table(sd_sysctl_header);
5427
	sd_sysctl_header = NULL;
5428 5429
	if (sd_ctl_dir[0].child)
		sd_free_ctl_entry(&sd_ctl_dir[0].child);
5430
}
5431
#else
5432 5433 5434 5435
static void register_sched_domain_sysctl(void)
{
}
static void unregister_sched_domain_sysctl(void)
5436 5437
{
}
P
Peter Zijlstra 已提交
5438
#endif /* CONFIG_SCHED_DEBUG && CONFIG_SYSCTL */
5439

5440 5441 5442 5443 5444
static void set_rq_online(struct rq *rq)
{
	if (!rq->online) {
		const struct sched_class *class;

5445
		cpumask_set_cpu(rq->cpu, rq->rd->online);
5446 5447 5448 5449 5450 5451 5452 5453 5454 5455 5456 5457 5458 5459 5460 5461 5462 5463 5464
		rq->online = 1;

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

static void set_rq_offline(struct rq *rq)
{
	if (rq->online) {
		const struct sched_class *class;

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

5465
		cpumask_clear_cpu(rq->cpu, rq->rd->online);
5466 5467 5468 5469
		rq->online = 0;
	}
}

L
Linus Torvalds 已提交
5470 5471 5472 5473
/*
 * migration_call - callback that gets triggered when a CPU is added.
 * Here we can start up the necessary migration thread for the new CPU.
 */
5474
static int
5475
migration_call(struct notifier_block *nfb, unsigned long action, void *hcpu)
L
Linus Torvalds 已提交
5476
{
5477
	int cpu = (long)hcpu;
L
Linus Torvalds 已提交
5478
	unsigned long flags;
5479
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
5480

5481
	switch (action & ~CPU_TASKS_FROZEN) {
5482

L
Linus Torvalds 已提交
5483
	case CPU_UP_PREPARE:
5484
		rq->calc_load_update = calc_load_update;
L
Linus Torvalds 已提交
5485
		break;
5486

L
Linus Torvalds 已提交
5487
	case CPU_ONLINE:
5488
		/* Update our root-domain */
5489
		raw_spin_lock_irqsave(&rq->lock, flags);
5490
		if (rq->rd) {
5491
			BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
5492 5493

			set_rq_online(rq);
5494
		}
5495
		raw_spin_unlock_irqrestore(&rq->lock, flags);
L
Linus Torvalds 已提交
5496
		break;
5497

L
Linus Torvalds 已提交
5498
#ifdef CONFIG_HOTPLUG_CPU
5499
	case CPU_DYING:
5500
		sched_ttwu_pending();
G
Gregory Haskins 已提交
5501
		/* Update our root-domain */
5502
		raw_spin_lock_irqsave(&rq->lock, flags);
G
Gregory Haskins 已提交
5503
		if (rq->rd) {
5504
			BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
5505
			set_rq_offline(rq);
G
Gregory Haskins 已提交
5506
		}
5507
		migrate_tasks(rq);
5508
		BUG_ON(rq->nr_running != 1); /* the migration thread */
5509
		raw_spin_unlock_irqrestore(&rq->lock, flags);
5510
		break;
5511

5512
	case CPU_DEAD:
5513
		calc_load_migrate(rq);
G
Gregory Haskins 已提交
5514
		break;
L
Linus Torvalds 已提交
5515 5516
#endif
	}
5517 5518 5519

	update_max_interval();

L
Linus Torvalds 已提交
5520 5521 5522
	return NOTIFY_OK;
}

5523 5524 5525
/*
 * Register at high priority so that task migration (migrate_all_tasks)
 * happens before everything else.  This has to be lower priority than
5526
 * the notifier in the perf_event subsystem, though.
L
Linus Torvalds 已提交
5527
 */
5528
static struct notifier_block migration_notifier = {
L
Linus Torvalds 已提交
5529
	.notifier_call = migration_call,
5530
	.priority = CPU_PRI_MIGRATION,
L
Linus Torvalds 已提交
5531 5532
};

5533
static void set_cpu_rq_start_time(void)
5534 5535 5536 5537 5538 5539
{
	int cpu = smp_processor_id();
	struct rq *rq = cpu_rq(cpu);
	rq->age_stamp = sched_clock_cpu(cpu);
}

5540
static int sched_cpu_active(struct notifier_block *nfb,
5541 5542 5543
				      unsigned long action, void *hcpu)
{
	switch (action & ~CPU_TASKS_FROZEN) {
5544 5545 5546
	case CPU_STARTING:
		set_cpu_rq_start_time();
		return NOTIFY_OK;
5547 5548 5549 5550 5551 5552 5553 5554
	case CPU_ONLINE:
		/*
		 * At this point a starting CPU has marked itself as online via
		 * set_cpu_online(). But it might not yet have marked itself
		 * as active, which is essential from here on.
		 *
		 * Thus, fall-through and help the starting CPU along.
		 */
5555 5556 5557 5558 5559 5560 5561 5562
	case CPU_DOWN_FAILED:
		set_cpu_active((long)hcpu, true);
		return NOTIFY_OK;
	default:
		return NOTIFY_DONE;
	}
}

5563
static int sched_cpu_inactive(struct notifier_block *nfb,
5564 5565 5566 5567
					unsigned long action, void *hcpu)
{
	switch (action & ~CPU_TASKS_FROZEN) {
	case CPU_DOWN_PREPARE:
5568
		set_cpu_active((long)hcpu, false);
5569
		return NOTIFY_OK;
5570 5571
	default:
		return NOTIFY_DONE;
5572 5573 5574
	}
}

5575
static int __init migration_init(void)
L
Linus Torvalds 已提交
5576 5577
{
	void *cpu = (void *)(long)smp_processor_id();
5578
	int err;
5579

5580
	/* Initialize migration for the boot CPU */
5581 5582
	err = migration_call(&migration_notifier, CPU_UP_PREPARE, cpu);
	BUG_ON(err == NOTIFY_BAD);
L
Linus Torvalds 已提交
5583 5584
	migration_call(&migration_notifier, CPU_ONLINE, cpu);
	register_cpu_notifier(&migration_notifier);
5585

5586 5587 5588 5589
	/* Register cpu active notifiers */
	cpu_notifier(sched_cpu_active, CPU_PRI_SCHED_ACTIVE);
	cpu_notifier(sched_cpu_inactive, CPU_PRI_SCHED_INACTIVE);

5590
	return 0;
L
Linus Torvalds 已提交
5591
}
5592
early_initcall(migration_init);
5593

5594 5595
static cpumask_var_t sched_domains_tmpmask; /* sched_domains_mutex */

5596
#ifdef CONFIG_SCHED_DEBUG
I
Ingo Molnar 已提交
5597

5598
static __read_mostly int sched_debug_enabled;
5599

5600
static int __init sched_debug_setup(char *str)
5601
{
5602
	sched_debug_enabled = 1;
5603 5604 5605

	return 0;
}
5606 5607 5608 5609 5610 5611
early_param("sched_debug", sched_debug_setup);

static inline bool sched_debug(void)
{
	return sched_debug_enabled;
}
5612

5613
static int sched_domain_debug_one(struct sched_domain *sd, int cpu, int level,
5614
				  struct cpumask *groupmask)
L
Linus Torvalds 已提交
5615
{
I
Ingo Molnar 已提交
5616
	struct sched_group *group = sd->groups;
L
Linus Torvalds 已提交
5617

5618
	cpumask_clear(groupmask);
I
Ingo Molnar 已提交
5619 5620 5621 5622

	printk(KERN_DEBUG "%*s domain %d: ", level, "", level);

	if (!(sd->flags & SD_LOAD_BALANCE)) {
P
Peter Zijlstra 已提交
5623
		printk("does not load-balance\n");
I
Ingo Molnar 已提交
5624
		if (sd->parent)
P
Peter Zijlstra 已提交
5625 5626
			printk(KERN_ERR "ERROR: !SD_LOAD_BALANCE domain"
					" has parent");
I
Ingo Molnar 已提交
5627
		return -1;
N
Nick Piggin 已提交
5628 5629
	}

5630 5631
	printk(KERN_CONT "span %*pbl level %s\n",
	       cpumask_pr_args(sched_domain_span(sd)), sd->name);
I
Ingo Molnar 已提交
5632

5633
	if (!cpumask_test_cpu(cpu, sched_domain_span(sd))) {
P
Peter Zijlstra 已提交
5634 5635
		printk(KERN_ERR "ERROR: domain->span does not contain "
				"CPU%d\n", cpu);
I
Ingo Molnar 已提交
5636
	}
5637
	if (!cpumask_test_cpu(cpu, sched_group_cpus(group))) {
P
Peter Zijlstra 已提交
5638 5639
		printk(KERN_ERR "ERROR: domain->groups does not contain"
				" CPU%d\n", cpu);
I
Ingo Molnar 已提交
5640
	}
L
Linus Torvalds 已提交
5641

I
Ingo Molnar 已提交
5642
	printk(KERN_DEBUG "%*s groups:", level + 1, "");
L
Linus Torvalds 已提交
5643
	do {
I
Ingo Molnar 已提交
5644
		if (!group) {
P
Peter Zijlstra 已提交
5645 5646
			printk("\n");
			printk(KERN_ERR "ERROR: group is NULL\n");
L
Linus Torvalds 已提交
5647 5648 5649
			break;
		}

5650
		if (!cpumask_weight(sched_group_cpus(group))) {
P
Peter Zijlstra 已提交
5651 5652
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: empty group\n");
I
Ingo Molnar 已提交
5653 5654
			break;
		}
L
Linus Torvalds 已提交
5655

5656 5657
		if (!(sd->flags & SD_OVERLAP) &&
		    cpumask_intersects(groupmask, sched_group_cpus(group))) {
P
Peter Zijlstra 已提交
5658 5659
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: repeated CPUs\n");
I
Ingo Molnar 已提交
5660 5661
			break;
		}
L
Linus Torvalds 已提交
5662

5663
		cpumask_or(groupmask, groupmask, sched_group_cpus(group));
L
Linus Torvalds 已提交
5664

5665 5666
		printk(KERN_CONT " %*pbl",
		       cpumask_pr_args(sched_group_cpus(group)));
5667
		if (group->sgc->capacity != SCHED_CAPACITY_SCALE) {
5668 5669
			printk(KERN_CONT " (cpu_capacity = %d)",
				group->sgc->capacity);
5670
		}
L
Linus Torvalds 已提交
5671

I
Ingo Molnar 已提交
5672 5673
		group = group->next;
	} while (group != sd->groups);
P
Peter Zijlstra 已提交
5674
	printk(KERN_CONT "\n");
L
Linus Torvalds 已提交
5675

5676
	if (!cpumask_equal(sched_domain_span(sd), groupmask))
P
Peter Zijlstra 已提交
5677
		printk(KERN_ERR "ERROR: groups don't span domain->span\n");
L
Linus Torvalds 已提交
5678

5679 5680
	if (sd->parent &&
	    !cpumask_subset(groupmask, sched_domain_span(sd->parent)))
P
Peter Zijlstra 已提交
5681 5682
		printk(KERN_ERR "ERROR: parent span is not a superset "
			"of domain->span\n");
I
Ingo Molnar 已提交
5683 5684
	return 0;
}
L
Linus Torvalds 已提交
5685

I
Ingo Molnar 已提交
5686 5687 5688
static void sched_domain_debug(struct sched_domain *sd, int cpu)
{
	int level = 0;
L
Linus Torvalds 已提交
5689

5690
	if (!sched_debug_enabled)
5691 5692
		return;

I
Ingo Molnar 已提交
5693 5694 5695 5696
	if (!sd) {
		printk(KERN_DEBUG "CPU%d attaching NULL sched-domain.\n", cpu);
		return;
	}
L
Linus Torvalds 已提交
5697

I
Ingo Molnar 已提交
5698 5699 5700
	printk(KERN_DEBUG "CPU%d attaching sched-domain:\n", cpu);

	for (;;) {
5701
		if (sched_domain_debug_one(sd, cpu, level, sched_domains_tmpmask))
I
Ingo Molnar 已提交
5702
			break;
L
Linus Torvalds 已提交
5703 5704
		level++;
		sd = sd->parent;
5705
		if (!sd)
I
Ingo Molnar 已提交
5706 5707
			break;
	}
L
Linus Torvalds 已提交
5708
}
5709
#else /* !CONFIG_SCHED_DEBUG */
5710
# define sched_domain_debug(sd, cpu) do { } while (0)
5711 5712 5713 5714
static inline bool sched_debug(void)
{
	return false;
}
5715
#endif /* CONFIG_SCHED_DEBUG */
L
Linus Torvalds 已提交
5716

5717
static int sd_degenerate(struct sched_domain *sd)
5718
{
5719
	if (cpumask_weight(sched_domain_span(sd)) == 1)
5720 5721 5722 5723 5724 5725
		return 1;

	/* Following flags need at least 2 groups */
	if (sd->flags & (SD_LOAD_BALANCE |
			 SD_BALANCE_NEWIDLE |
			 SD_BALANCE_FORK |
5726
			 SD_BALANCE_EXEC |
5727
			 SD_SHARE_CPUCAPACITY |
5728 5729
			 SD_SHARE_PKG_RESOURCES |
			 SD_SHARE_POWERDOMAIN)) {
5730 5731 5732 5733 5734
		if (sd->groups != sd->groups->next)
			return 0;
	}

	/* Following flags don't use groups */
5735
	if (sd->flags & (SD_WAKE_AFFINE))
5736 5737 5738 5739 5740
		return 0;

	return 1;
}

5741 5742
static int
sd_parent_degenerate(struct sched_domain *sd, struct sched_domain *parent)
5743 5744 5745 5746 5747 5748
{
	unsigned long cflags = sd->flags, pflags = parent->flags;

	if (sd_degenerate(parent))
		return 1;

5749
	if (!cpumask_equal(sched_domain_span(sd), sched_domain_span(parent)))
5750 5751 5752 5753 5754 5755 5756
		return 0;

	/* Flags needing groups don't count if only 1 group in parent */
	if (parent->groups == parent->groups->next) {
		pflags &= ~(SD_LOAD_BALANCE |
				SD_BALANCE_NEWIDLE |
				SD_BALANCE_FORK |
5757
				SD_BALANCE_EXEC |
5758
				SD_SHARE_CPUCAPACITY |
5759
				SD_SHARE_PKG_RESOURCES |
5760 5761
				SD_PREFER_SIBLING |
				SD_SHARE_POWERDOMAIN);
5762 5763
		if (nr_node_ids == 1)
			pflags &= ~SD_SERIALIZE;
5764 5765 5766 5767 5768 5769 5770
	}
	if (~cflags & pflags)
		return 0;

	return 1;
}

5771
static void free_rootdomain(struct rcu_head *rcu)
5772
{
5773
	struct root_domain *rd = container_of(rcu, struct root_domain, rcu);
5774

5775
	cpupri_cleanup(&rd->cpupri);
5776
	cpudl_cleanup(&rd->cpudl);
5777
	free_cpumask_var(rd->dlo_mask);
5778 5779 5780 5781 5782 5783
	free_cpumask_var(rd->rto_mask);
	free_cpumask_var(rd->online);
	free_cpumask_var(rd->span);
	kfree(rd);
}

G
Gregory Haskins 已提交
5784 5785
static void rq_attach_root(struct rq *rq, struct root_domain *rd)
{
I
Ingo Molnar 已提交
5786
	struct root_domain *old_rd = NULL;
G
Gregory Haskins 已提交
5787 5788
	unsigned long flags;

5789
	raw_spin_lock_irqsave(&rq->lock, flags);
G
Gregory Haskins 已提交
5790 5791

	if (rq->rd) {
I
Ingo Molnar 已提交
5792
		old_rd = rq->rd;
G
Gregory Haskins 已提交
5793

5794
		if (cpumask_test_cpu(rq->cpu, old_rd->online))
5795
			set_rq_offline(rq);
G
Gregory Haskins 已提交
5796

5797
		cpumask_clear_cpu(rq->cpu, old_rd->span);
5798

I
Ingo Molnar 已提交
5799
		/*
5800
		 * If we dont want to free the old_rd yet then
I
Ingo Molnar 已提交
5801 5802 5803 5804 5805
		 * set old_rd to NULL to skip the freeing later
		 * in this function:
		 */
		if (!atomic_dec_and_test(&old_rd->refcount))
			old_rd = NULL;
G
Gregory Haskins 已提交
5806 5807 5808 5809 5810
	}

	atomic_inc(&rd->refcount);
	rq->rd = rd;

5811
	cpumask_set_cpu(rq->cpu, rd->span);
5812
	if (cpumask_test_cpu(rq->cpu, cpu_active_mask))
5813
		set_rq_online(rq);
G
Gregory Haskins 已提交
5814

5815
	raw_spin_unlock_irqrestore(&rq->lock, flags);
I
Ingo Molnar 已提交
5816 5817

	if (old_rd)
5818
		call_rcu_sched(&old_rd->rcu, free_rootdomain);
G
Gregory Haskins 已提交
5819 5820
}

5821
static int init_rootdomain(struct root_domain *rd)
G
Gregory Haskins 已提交
5822 5823 5824
{
	memset(rd, 0, sizeof(*rd));

5825
	if (!alloc_cpumask_var(&rd->span, GFP_KERNEL))
5826
		goto out;
5827
	if (!alloc_cpumask_var(&rd->online, GFP_KERNEL))
5828
		goto free_span;
5829
	if (!alloc_cpumask_var(&rd->dlo_mask, GFP_KERNEL))
5830
		goto free_online;
5831 5832
	if (!alloc_cpumask_var(&rd->rto_mask, GFP_KERNEL))
		goto free_dlo_mask;
5833

5834
	init_dl_bw(&rd->dl_bw);
5835 5836
	if (cpudl_init(&rd->cpudl) != 0)
		goto free_dlo_mask;
5837

5838
	if (cpupri_init(&rd->cpupri) != 0)
5839
		goto free_rto_mask;
5840
	return 0;
5841

5842 5843
free_rto_mask:
	free_cpumask_var(rd->rto_mask);
5844 5845
free_dlo_mask:
	free_cpumask_var(rd->dlo_mask);
5846 5847 5848 5849
free_online:
	free_cpumask_var(rd->online);
free_span:
	free_cpumask_var(rd->span);
5850
out:
5851
	return -ENOMEM;
G
Gregory Haskins 已提交
5852 5853
}

5854 5855 5856 5857 5858 5859
/*
 * By default the system creates a single root-domain with all cpus as
 * members (mimicking the global state we have today).
 */
struct root_domain def_root_domain;

G
Gregory Haskins 已提交
5860 5861
static void init_defrootdomain(void)
{
5862
	init_rootdomain(&def_root_domain);
5863

G
Gregory Haskins 已提交
5864 5865 5866
	atomic_set(&def_root_domain.refcount, 1);
}

5867
static struct root_domain *alloc_rootdomain(void)
G
Gregory Haskins 已提交
5868 5869 5870 5871 5872 5873 5874
{
	struct root_domain *rd;

	rd = kmalloc(sizeof(*rd), GFP_KERNEL);
	if (!rd)
		return NULL;

5875
	if (init_rootdomain(rd) != 0) {
5876 5877 5878
		kfree(rd);
		return NULL;
	}
G
Gregory Haskins 已提交
5879 5880 5881 5882

	return rd;
}

5883
static void free_sched_groups(struct sched_group *sg, int free_sgc)
5884 5885 5886 5887 5888 5889 5890 5891 5892 5893
{
	struct sched_group *tmp, *first;

	if (!sg)
		return;

	first = sg;
	do {
		tmp = sg->next;

5894 5895
		if (free_sgc && atomic_dec_and_test(&sg->sgc->ref))
			kfree(sg->sgc);
5896 5897 5898 5899 5900 5901

		kfree(sg);
		sg = tmp;
	} while (sg != first);
}

5902 5903 5904
static void free_sched_domain(struct rcu_head *rcu)
{
	struct sched_domain *sd = container_of(rcu, struct sched_domain, rcu);
5905 5906 5907 5908 5909 5910 5911 5912

	/*
	 * If its an overlapping domain it has private groups, iterate and
	 * nuke them all.
	 */
	if (sd->flags & SD_OVERLAP) {
		free_sched_groups(sd->groups, 1);
	} else if (atomic_dec_and_test(&sd->groups->ref)) {
5913
		kfree(sd->groups->sgc);
5914
		kfree(sd->groups);
5915
	}
5916 5917 5918 5919 5920 5921 5922 5923 5924 5925 5926 5927 5928 5929
	kfree(sd);
}

static void destroy_sched_domain(struct sched_domain *sd, int cpu)
{
	call_rcu(&sd->rcu, free_sched_domain);
}

static void destroy_sched_domains(struct sched_domain *sd, int cpu)
{
	for (; sd; sd = sd->parent)
		destroy_sched_domain(sd, cpu);
}

5930 5931 5932 5933 5934 5935 5936
/*
 * Keep a special pointer to the highest sched_domain that has
 * SD_SHARE_PKG_RESOURCE set (Last Level Cache Domain) for this
 * allows us to avoid some pointer chasing select_idle_sibling().
 *
 * Also keep a unique ID per domain (we use the first cpu number in
 * the cpumask of the domain), this allows us to quickly tell if
5937
 * two cpus are in the same cache domain, see cpus_share_cache().
5938 5939
 */
DEFINE_PER_CPU(struct sched_domain *, sd_llc);
5940
DEFINE_PER_CPU(int, sd_llc_size);
5941
DEFINE_PER_CPU(int, sd_llc_id);
5942
DEFINE_PER_CPU(struct sched_domain *, sd_numa);
5943 5944
DEFINE_PER_CPU(struct sched_domain *, sd_busy);
DEFINE_PER_CPU(struct sched_domain *, sd_asym);
5945 5946 5947 5948

static void update_top_cache_domain(int cpu)
{
	struct sched_domain *sd;
5949
	struct sched_domain *busy_sd = NULL;
5950
	int id = cpu;
5951
	int size = 1;
5952 5953

	sd = highest_flag_domain(cpu, SD_SHARE_PKG_RESOURCES);
5954
	if (sd) {
5955
		id = cpumask_first(sched_domain_span(sd));
5956
		size = cpumask_weight(sched_domain_span(sd));
5957
		busy_sd = sd->parent; /* sd_busy */
5958
	}
5959
	rcu_assign_pointer(per_cpu(sd_busy, cpu), busy_sd);
5960 5961

	rcu_assign_pointer(per_cpu(sd_llc, cpu), sd);
5962
	per_cpu(sd_llc_size, cpu) = size;
5963
	per_cpu(sd_llc_id, cpu) = id;
5964 5965 5966

	sd = lowest_flag_domain(cpu, SD_NUMA);
	rcu_assign_pointer(per_cpu(sd_numa, cpu), sd);
5967 5968 5969

	sd = highest_flag_domain(cpu, SD_ASYM_PACKING);
	rcu_assign_pointer(per_cpu(sd_asym, cpu), sd);
5970 5971
}

L
Linus Torvalds 已提交
5972
/*
I
Ingo Molnar 已提交
5973
 * Attach the domain 'sd' to 'cpu' as its base domain. Callers must
L
Linus Torvalds 已提交
5974 5975
 * hold the hotplug lock.
 */
I
Ingo Molnar 已提交
5976 5977
static void
cpu_attach_domain(struct sched_domain *sd, struct root_domain *rd, int cpu)
L
Linus Torvalds 已提交
5978
{
5979
	struct rq *rq = cpu_rq(cpu);
5980 5981 5982
	struct sched_domain *tmp;

	/* Remove the sched domains which do not contribute to scheduling. */
5983
	for (tmp = sd; tmp; ) {
5984 5985 5986
		struct sched_domain *parent = tmp->parent;
		if (!parent)
			break;
5987

5988
		if (sd_parent_degenerate(tmp, parent)) {
5989
			tmp->parent = parent->parent;
5990 5991
			if (parent->parent)
				parent->parent->child = tmp;
5992 5993 5994 5995 5996 5997 5998
			/*
			 * Transfer SD_PREFER_SIBLING down in case of a
			 * degenerate parent; the spans match for this
			 * so the property transfers.
			 */
			if (parent->flags & SD_PREFER_SIBLING)
				tmp->flags |= SD_PREFER_SIBLING;
5999
			destroy_sched_domain(parent, cpu);
6000 6001
		} else
			tmp = tmp->parent;
6002 6003
	}

6004
	if (sd && sd_degenerate(sd)) {
6005
		tmp = sd;
6006
		sd = sd->parent;
6007
		destroy_sched_domain(tmp, cpu);
6008 6009 6010
		if (sd)
			sd->child = NULL;
	}
L
Linus Torvalds 已提交
6011

6012
	sched_domain_debug(sd, cpu);
L
Linus Torvalds 已提交
6013

G
Gregory Haskins 已提交
6014
	rq_attach_root(rq, rd);
6015
	tmp = rq->sd;
N
Nick Piggin 已提交
6016
	rcu_assign_pointer(rq->sd, sd);
6017
	destroy_sched_domains(tmp, cpu);
6018 6019

	update_top_cache_domain(cpu);
L
Linus Torvalds 已提交
6020 6021 6022 6023 6024
}

/* Setup the mask of cpus configured for isolated domains */
static int __init isolated_cpu_setup(char *str)
{
R
Rusty Russell 已提交
6025
	alloc_bootmem_cpumask_var(&cpu_isolated_map);
R
Rusty Russell 已提交
6026
	cpulist_parse(str, cpu_isolated_map);
L
Linus Torvalds 已提交
6027 6028 6029
	return 1;
}

I
Ingo Molnar 已提交
6030
__setup("isolcpus=", isolated_cpu_setup);
L
Linus Torvalds 已提交
6031

6032
struct s_data {
6033
	struct sched_domain ** __percpu sd;
6034 6035 6036
	struct root_domain	*rd;
};

6037 6038
enum s_alloc {
	sa_rootdomain,
6039
	sa_sd,
6040
	sa_sd_storage,
6041 6042 6043
	sa_none,
};

P
Peter Zijlstra 已提交
6044 6045 6046 6047 6048 6049 6050 6051 6052 6053 6054 6055 6056 6057 6058 6059 6060 6061 6062 6063 6064 6065 6066 6067 6068 6069 6070 6071 6072 6073 6074 6075 6076 6077 6078 6079 6080 6081
/*
 * Build an iteration mask that can exclude certain CPUs from the upwards
 * domain traversal.
 *
 * Asymmetric node setups can result in situations where the domain tree is of
 * unequal depth, make sure to skip domains that already cover the entire
 * range.
 *
 * In that case build_sched_domains() will have terminated the iteration early
 * and our sibling sd spans will be empty. Domains should always include the
 * cpu they're built on, so check that.
 *
 */
static void build_group_mask(struct sched_domain *sd, struct sched_group *sg)
{
	const struct cpumask *span = sched_domain_span(sd);
	struct sd_data *sdd = sd->private;
	struct sched_domain *sibling;
	int i;

	for_each_cpu(i, span) {
		sibling = *per_cpu_ptr(sdd->sd, i);
		if (!cpumask_test_cpu(i, sched_domain_span(sibling)))
			continue;

		cpumask_set_cpu(i, sched_group_mask(sg));
	}
}

/*
 * Return the canonical balance cpu for this group, this is the first cpu
 * of this group that's also in the iteration mask.
 */
int group_balance_cpu(struct sched_group *sg)
{
	return cpumask_first_and(sched_group_cpus(sg), sched_group_mask(sg));
}

6082 6083 6084 6085 6086 6087 6088
static int
build_overlap_sched_groups(struct sched_domain *sd, int cpu)
{
	struct sched_group *first = NULL, *last = NULL, *groups = NULL, *sg;
	const struct cpumask *span = sched_domain_span(sd);
	struct cpumask *covered = sched_domains_tmpmask;
	struct sd_data *sdd = sd->private;
6089
	struct sched_domain *sibling;
6090 6091 6092 6093 6094 6095 6096 6097 6098 6099
	int i;

	cpumask_clear(covered);

	for_each_cpu(i, span) {
		struct cpumask *sg_span;

		if (cpumask_test_cpu(i, covered))
			continue;

6100
		sibling = *per_cpu_ptr(sdd->sd, i);
P
Peter Zijlstra 已提交
6101 6102

		/* See the comment near build_group_mask(). */
6103
		if (!cpumask_test_cpu(i, sched_domain_span(sibling)))
P
Peter Zijlstra 已提交
6104 6105
			continue;

6106
		sg = kzalloc_node(sizeof(struct sched_group) + cpumask_size(),
6107
				GFP_KERNEL, cpu_to_node(cpu));
6108 6109 6110 6111 6112

		if (!sg)
			goto fail;

		sg_span = sched_group_cpus(sg);
6113 6114 6115
		if (sibling->child)
			cpumask_copy(sg_span, sched_domain_span(sibling->child));
		else
6116 6117 6118 6119
			cpumask_set_cpu(i, sg_span);

		cpumask_or(covered, covered, sg_span);

6120 6121
		sg->sgc = *per_cpu_ptr(sdd->sgc, i);
		if (atomic_inc_return(&sg->sgc->ref) == 1)
P
Peter Zijlstra 已提交
6122 6123
			build_group_mask(sd, sg);

6124
		/*
6125
		 * Initialize sgc->capacity such that even if we mess up the
6126 6127 6128
		 * domains and no possible iteration will get us here, we won't
		 * die on a /0 trap.
		 */
6129
		sg->sgc->capacity = SCHED_CAPACITY_SCALE * cpumask_weight(sg_span);
6130

P
Peter Zijlstra 已提交
6131 6132 6133 6134 6135
		/*
		 * Make sure the first group of this domain contains the
		 * canonical balance cpu. Otherwise the sched_domain iteration
		 * breaks. See update_sg_lb_stats().
		 */
P
Peter Zijlstra 已提交
6136
		if ((!groups && cpumask_test_cpu(cpu, sg_span)) ||
P
Peter Zijlstra 已提交
6137
		    group_balance_cpu(sg) == cpu)
6138 6139 6140 6141 6142 6143 6144 6145 6146 6147 6148 6149 6150 6151 6152 6153 6154 6155 6156
			groups = sg;

		if (!first)
			first = sg;
		if (last)
			last->next = sg;
		last = sg;
		last->next = first;
	}
	sd->groups = groups;

	return 0;

fail:
	free_sched_groups(first, 0);

	return -ENOMEM;
}

6157
static int get_group(int cpu, struct sd_data *sdd, struct sched_group **sg)
L
Linus Torvalds 已提交
6158
{
6159 6160
	struct sched_domain *sd = *per_cpu_ptr(sdd->sd, cpu);
	struct sched_domain *child = sd->child;
L
Linus Torvalds 已提交
6161

6162 6163
	if (child)
		cpu = cpumask_first(sched_domain_span(child));
6164

6165
	if (sg) {
6166
		*sg = *per_cpu_ptr(sdd->sg, cpu);
6167 6168
		(*sg)->sgc = *per_cpu_ptr(sdd->sgc, cpu);
		atomic_set(&(*sg)->sgc->ref, 1); /* for claim_allocations */
6169
	}
6170 6171

	return cpu;
6172 6173
}

6174
/*
6175 6176
 * build_sched_groups will build a circular linked list of the groups
 * covered by the given span, and will set each group's ->cpumask correctly,
6177
 * and ->cpu_capacity to 0.
6178 6179
 *
 * Assumes the sched_domain tree is fully constructed
6180
 */
6181 6182
static int
build_sched_groups(struct sched_domain *sd, int cpu)
L
Linus Torvalds 已提交
6183
{
6184 6185 6186
	struct sched_group *first = NULL, *last = NULL;
	struct sd_data *sdd = sd->private;
	const struct cpumask *span = sched_domain_span(sd);
6187
	struct cpumask *covered;
6188
	int i;
6189

6190 6191 6192
	get_group(cpu, sdd, &sd->groups);
	atomic_inc(&sd->groups->ref);

6193
	if (cpu != cpumask_first(span))
6194 6195
		return 0;

6196 6197 6198
	lockdep_assert_held(&sched_domains_mutex);
	covered = sched_domains_tmpmask;

6199
	cpumask_clear(covered);
6200

6201 6202
	for_each_cpu(i, span) {
		struct sched_group *sg;
6203
		int group, j;
6204

6205 6206
		if (cpumask_test_cpu(i, covered))
			continue;
6207

6208
		group = get_group(i, sdd, &sg);
P
Peter Zijlstra 已提交
6209
		cpumask_setall(sched_group_mask(sg));
6210

6211 6212 6213
		for_each_cpu(j, span) {
			if (get_group(j, sdd, NULL) != group)
				continue;
6214

6215 6216 6217
			cpumask_set_cpu(j, covered);
			cpumask_set_cpu(j, sched_group_cpus(sg));
		}
6218

6219 6220 6221 6222 6223 6224 6225
		if (!first)
			first = sg;
		if (last)
			last->next = sg;
		last = sg;
	}
	last->next = first;
6226 6227

	return 0;
6228
}
6229

6230
/*
6231
 * Initialize sched groups cpu_capacity.
6232
 *
6233
 * cpu_capacity indicates the capacity of sched group, which is used while
6234
 * distributing the load between different sched groups in a sched domain.
6235 6236 6237 6238
 * Typically cpu_capacity for all the groups in a sched domain will be same
 * unless there are asymmetries in the topology. If there are asymmetries,
 * group having more cpu_capacity will pickup more load compared to the
 * group having less cpu_capacity.
6239
 */
6240
static void init_sched_groups_capacity(int cpu, struct sched_domain *sd)
6241
{
6242
	struct sched_group *sg = sd->groups;
6243

6244
	WARN_ON(!sg);
6245 6246 6247 6248 6249

	do {
		sg->group_weight = cpumask_weight(sched_group_cpus(sg));
		sg = sg->next;
	} while (sg != sd->groups);
6250

P
Peter Zijlstra 已提交
6251
	if (cpu != group_balance_cpu(sg))
6252
		return;
6253

6254 6255
	update_group_capacity(sd, cpu);
	atomic_set(&sg->sgc->nr_busy_cpus, sg->group_weight);
6256 6257
}

6258 6259 6260 6261 6262
/*
 * Initializers for schedule domains
 * Non-inlined to reduce accumulated stack pressure in build_sched_domains()
 */

6263
static int default_relax_domain_level = -1;
6264
int sched_domain_level_max;
6265 6266 6267

static int __init setup_relax_domain_level(char *str)
{
6268 6269
	if (kstrtoint(str, 0, &default_relax_domain_level))
		pr_warn("Unable to set relax_domain_level\n");
6270

6271 6272 6273 6274 6275 6276 6277 6278 6279 6280 6281 6282 6283 6284 6285 6286 6287 6288
	return 1;
}
__setup("relax_domain_level=", setup_relax_domain_level);

static void set_domain_attribute(struct sched_domain *sd,
				 struct sched_domain_attr *attr)
{
	int request;

	if (!attr || attr->relax_domain_level < 0) {
		if (default_relax_domain_level < 0)
			return;
		else
			request = default_relax_domain_level;
	} else
		request = attr->relax_domain_level;
	if (request < sd->level) {
		/* turn off idle balance on this domain */
6289
		sd->flags &= ~(SD_BALANCE_WAKE|SD_BALANCE_NEWIDLE);
6290 6291
	} else {
		/* turn on idle balance on this domain */
6292
		sd->flags |= (SD_BALANCE_WAKE|SD_BALANCE_NEWIDLE);
6293 6294 6295
	}
}

6296 6297 6298
static void __sdt_free(const struct cpumask *cpu_map);
static int __sdt_alloc(const struct cpumask *cpu_map);

6299 6300 6301 6302 6303
static void __free_domain_allocs(struct s_data *d, enum s_alloc what,
				 const struct cpumask *cpu_map)
{
	switch (what) {
	case sa_rootdomain:
6304 6305
		if (!atomic_read(&d->rd->refcount))
			free_rootdomain(&d->rd->rcu); /* fall through */
6306 6307
	case sa_sd:
		free_percpu(d->sd); /* fall through */
6308
	case sa_sd_storage:
6309
		__sdt_free(cpu_map); /* fall through */
6310 6311 6312 6313
	case sa_none:
		break;
	}
}
6314

6315 6316 6317
static enum s_alloc __visit_domain_allocation_hell(struct s_data *d,
						   const struct cpumask *cpu_map)
{
6318 6319
	memset(d, 0, sizeof(*d));

6320 6321
	if (__sdt_alloc(cpu_map))
		return sa_sd_storage;
6322 6323 6324
	d->sd = alloc_percpu(struct sched_domain *);
	if (!d->sd)
		return sa_sd_storage;
6325
	d->rd = alloc_rootdomain();
6326
	if (!d->rd)
6327
		return sa_sd;
6328 6329
	return sa_rootdomain;
}
G
Gregory Haskins 已提交
6330

6331 6332 6333 6334 6335 6336 6337 6338 6339 6340 6341 6342
/*
 * NULL the sd_data elements we've used to build the sched_domain and
 * sched_group structure so that the subsequent __free_domain_allocs()
 * will not free the data we're using.
 */
static void claim_allocations(int cpu, struct sched_domain *sd)
{
	struct sd_data *sdd = sd->private;

	WARN_ON_ONCE(*per_cpu_ptr(sdd->sd, cpu) != sd);
	*per_cpu_ptr(sdd->sd, cpu) = NULL;

6343
	if (atomic_read(&(*per_cpu_ptr(sdd->sg, cpu))->ref))
6344
		*per_cpu_ptr(sdd->sg, cpu) = NULL;
6345

6346 6347
	if (atomic_read(&(*per_cpu_ptr(sdd->sgc, cpu))->ref))
		*per_cpu_ptr(sdd->sgc, cpu) = NULL;
6348 6349
}

6350 6351
#ifdef CONFIG_NUMA
static int sched_domains_numa_levels;
6352
enum numa_topology_type sched_numa_topology_type;
6353
static int *sched_domains_numa_distance;
6354
int sched_max_numa_distance;
6355 6356
static struct cpumask ***sched_domains_numa_masks;
static int sched_domains_curr_level;
6357
#endif
6358

6359 6360 6361
/*
 * SD_flags allowed in topology descriptions.
 *
6362
 * SD_SHARE_CPUCAPACITY      - describes SMT topologies
6363 6364
 * SD_SHARE_PKG_RESOURCES - describes shared caches
 * SD_NUMA                - describes NUMA topologies
6365
 * SD_SHARE_POWERDOMAIN   - describes shared power domain
6366 6367 6368 6369 6370
 *
 * Odd one out:
 * SD_ASYM_PACKING        - describes SMT quirks
 */
#define TOPOLOGY_SD_FLAGS		\
6371
	(SD_SHARE_CPUCAPACITY |		\
6372 6373
	 SD_SHARE_PKG_RESOURCES |	\
	 SD_NUMA |			\
6374 6375
	 SD_ASYM_PACKING |		\
	 SD_SHARE_POWERDOMAIN)
6376 6377

static struct sched_domain *
6378
sd_init(struct sched_domain_topology_level *tl, int cpu)
6379 6380
{
	struct sched_domain *sd = *per_cpu_ptr(tl->data.sd, cpu);
6381 6382 6383 6384 6385 6386 6387 6388 6389 6390 6391 6392 6393 6394 6395 6396
	int sd_weight, sd_flags = 0;

#ifdef CONFIG_NUMA
	/*
	 * Ugly hack to pass state to sd_numa_mask()...
	 */
	sched_domains_curr_level = tl->numa_level;
#endif

	sd_weight = cpumask_weight(tl->mask(cpu));

	if (tl->sd_flags)
		sd_flags = (*tl->sd_flags)();
	if (WARN_ONCE(sd_flags & ~TOPOLOGY_SD_FLAGS,
			"wrong sd_flags in topology description\n"))
		sd_flags &= ~TOPOLOGY_SD_FLAGS;
6397 6398 6399 6400 6401

	*sd = (struct sched_domain){
		.min_interval		= sd_weight,
		.max_interval		= 2*sd_weight,
		.busy_factor		= 32,
6402
		.imbalance_pct		= 125,
6403 6404 6405 6406

		.cache_nice_tries	= 0,
		.busy_idx		= 0,
		.idle_idx		= 0,
6407 6408 6409 6410 6411 6412
		.newidle_idx		= 0,
		.wake_idx		= 0,
		.forkexec_idx		= 0,

		.flags			= 1*SD_LOAD_BALANCE
					| 1*SD_BALANCE_NEWIDLE
6413 6414
					| 1*SD_BALANCE_EXEC
					| 1*SD_BALANCE_FORK
6415
					| 0*SD_BALANCE_WAKE
6416
					| 1*SD_WAKE_AFFINE
6417
					| 0*SD_SHARE_CPUCAPACITY
6418
					| 0*SD_SHARE_PKG_RESOURCES
6419
					| 0*SD_SERIALIZE
6420
					| 0*SD_PREFER_SIBLING
6421 6422
					| 0*SD_NUMA
					| sd_flags
6423
					,
6424

6425 6426
		.last_balance		= jiffies,
		.balance_interval	= sd_weight,
6427
		.smt_gain		= 0,
6428 6429
		.max_newidle_lb_cost	= 0,
		.next_decay_max_lb_cost	= jiffies,
6430 6431 6432
#ifdef CONFIG_SCHED_DEBUG
		.name			= tl->name,
#endif
6433 6434 6435
	};

	/*
6436
	 * Convert topological properties into behaviour.
6437
	 */
6438

6439
	if (sd->flags & SD_SHARE_CPUCAPACITY) {
6440
		sd->flags |= SD_PREFER_SIBLING;
6441 6442 6443 6444 6445 6446 6447 6448 6449 6450 6451 6452 6453 6454 6455 6456 6457 6458 6459 6460 6461 6462 6463 6464 6465 6466 6467 6468 6469 6470
		sd->imbalance_pct = 110;
		sd->smt_gain = 1178; /* ~15% */

	} else if (sd->flags & SD_SHARE_PKG_RESOURCES) {
		sd->imbalance_pct = 117;
		sd->cache_nice_tries = 1;
		sd->busy_idx = 2;

#ifdef CONFIG_NUMA
	} else if (sd->flags & SD_NUMA) {
		sd->cache_nice_tries = 2;
		sd->busy_idx = 3;
		sd->idle_idx = 2;

		sd->flags |= SD_SERIALIZE;
		if (sched_domains_numa_distance[tl->numa_level] > RECLAIM_DISTANCE) {
			sd->flags &= ~(SD_BALANCE_EXEC |
				       SD_BALANCE_FORK |
				       SD_WAKE_AFFINE);
		}

#endif
	} else {
		sd->flags |= SD_PREFER_SIBLING;
		sd->cache_nice_tries = 1;
		sd->busy_idx = 2;
		sd->idle_idx = 1;
	}

	sd->private = &tl->data;
6471 6472 6473 6474

	return sd;
}

6475 6476 6477 6478 6479 6480 6481 6482 6483 6484 6485 6486 6487 6488
/*
 * Topology list, bottom-up.
 */
static struct sched_domain_topology_level default_topology[] = {
#ifdef CONFIG_SCHED_SMT
	{ cpu_smt_mask, cpu_smt_flags, SD_INIT_NAME(SMT) },
#endif
#ifdef CONFIG_SCHED_MC
	{ cpu_coregroup_mask, cpu_core_flags, SD_INIT_NAME(MC) },
#endif
	{ cpu_cpu_mask, SD_INIT_NAME(DIE) },
	{ NULL, },
};

6489 6490
static struct sched_domain_topology_level *sched_domain_topology =
	default_topology;
6491 6492 6493 6494 6495 6496 6497 6498 6499 6500 6501

#define for_each_sd_topology(tl)			\
	for (tl = sched_domain_topology; tl->mask; tl++)

void set_sched_topology(struct sched_domain_topology_level *tl)
{
	sched_domain_topology = tl;
}

#ifdef CONFIG_NUMA

6502 6503 6504 6505 6506
static const struct cpumask *sd_numa_mask(int cpu)
{
	return sched_domains_numa_masks[sched_domains_curr_level][cpu_to_node(cpu)];
}

6507 6508 6509 6510 6511 6512 6513 6514 6515 6516 6517 6518 6519 6520 6521 6522 6523 6524 6525 6526 6527
static void sched_numa_warn(const char *str)
{
	static int done = false;
	int i,j;

	if (done)
		return;

	done = true;

	printk(KERN_WARNING "ERROR: %s\n\n", str);

	for (i = 0; i < nr_node_ids; i++) {
		printk(KERN_WARNING "  ");
		for (j = 0; j < nr_node_ids; j++)
			printk(KERN_CONT "%02d ", node_distance(i,j));
		printk(KERN_CONT "\n");
	}
	printk(KERN_WARNING "\n");
}

6528
bool find_numa_distance(int distance)
6529 6530 6531 6532 6533 6534 6535 6536 6537 6538 6539 6540 6541 6542
{
	int i;

	if (distance == node_distance(0, 0))
		return true;

	for (i = 0; i < sched_domains_numa_levels; i++) {
		if (sched_domains_numa_distance[i] == distance)
			return true;
	}

	return false;
}

6543 6544 6545 6546 6547 6548 6549 6550 6551 6552 6553 6554 6555 6556 6557 6558 6559 6560 6561 6562 6563 6564 6565 6566 6567
/*
 * A system can have three types of NUMA topology:
 * NUMA_DIRECT: all nodes are directly connected, or not a NUMA system
 * NUMA_GLUELESS_MESH: some nodes reachable through intermediary nodes
 * NUMA_BACKPLANE: nodes can reach other nodes through a backplane
 *
 * The difference between a glueless mesh topology and a backplane
 * topology lies in whether communication between not directly
 * connected nodes goes through intermediary nodes (where programs
 * could run), or through backplane controllers. This affects
 * placement of programs.
 *
 * The type of topology can be discerned with the following tests:
 * - If the maximum distance between any nodes is 1 hop, the system
 *   is directly connected.
 * - If for two nodes A and B, located N > 1 hops away from each other,
 *   there is an intermediary node C, which is < N hops away from both
 *   nodes A and B, the system is a glueless mesh.
 */
static void init_numa_topology_type(void)
{
	int a, b, c, n;

	n = sched_max_numa_distance;

6568
	if (sched_domains_numa_levels <= 1) {
6569
		sched_numa_topology_type = NUMA_DIRECT;
6570 6571
		return;
	}
6572 6573 6574 6575 6576 6577 6578 6579 6580 6581 6582 6583 6584 6585 6586 6587 6588 6589 6590 6591 6592 6593 6594

	for_each_online_node(a) {
		for_each_online_node(b) {
			/* Find two nodes furthest removed from each other. */
			if (node_distance(a, b) < n)
				continue;

			/* Is there an intermediary node between a and b? */
			for_each_online_node(c) {
				if (node_distance(a, c) < n &&
				    node_distance(b, c) < n) {
					sched_numa_topology_type =
							NUMA_GLUELESS_MESH;
					return;
				}
			}

			sched_numa_topology_type = NUMA_BACKPLANE;
			return;
		}
	}
}

6595 6596 6597 6598 6599 6600 6601 6602 6603 6604 6605 6606 6607 6608 6609 6610 6611 6612 6613 6614 6615
static void sched_init_numa(void)
{
	int next_distance, curr_distance = node_distance(0, 0);
	struct sched_domain_topology_level *tl;
	int level = 0;
	int i, j, k;

	sched_domains_numa_distance = kzalloc(sizeof(int) * nr_node_ids, GFP_KERNEL);
	if (!sched_domains_numa_distance)
		return;

	/*
	 * O(nr_nodes^2) deduplicating selection sort -- in order to find the
	 * unique distances in the node_distance() table.
	 *
	 * Assumes node_distance(0,j) includes all distances in
	 * node_distance(i,j) in order to avoid cubic time.
	 */
	next_distance = curr_distance;
	for (i = 0; i < nr_node_ids; i++) {
		for (j = 0; j < nr_node_ids; j++) {
6616 6617 6618 6619 6620 6621 6622 6623 6624 6625 6626 6627 6628 6629 6630 6631 6632 6633 6634 6635 6636 6637 6638 6639
			for (k = 0; k < nr_node_ids; k++) {
				int distance = node_distance(i, k);

				if (distance > curr_distance &&
				    (distance < next_distance ||
				     next_distance == curr_distance))
					next_distance = distance;

				/*
				 * While not a strong assumption it would be nice to know
				 * about cases where if node A is connected to B, B is not
				 * equally connected to A.
				 */
				if (sched_debug() && node_distance(k, i) != distance)
					sched_numa_warn("Node-distance not symmetric");

				if (sched_debug() && i && !find_numa_distance(distance))
					sched_numa_warn("Node-0 not representative");
			}
			if (next_distance != curr_distance) {
				sched_domains_numa_distance[level++] = next_distance;
				sched_domains_numa_levels = level;
				curr_distance = next_distance;
			} else break;
6640
		}
6641 6642 6643 6644 6645 6646

		/*
		 * In case of sched_debug() we verify the above assumption.
		 */
		if (!sched_debug())
			break;
6647
	}
6648 6649 6650 6651

	if (!level)
		return;

6652 6653 6654 6655
	/*
	 * 'level' contains the number of unique distances, excluding the
	 * identity distance node_distance(i,i).
	 *
V
Viresh Kumar 已提交
6656
	 * The sched_domains_numa_distance[] array includes the actual distance
6657 6658 6659
	 * numbers.
	 */

6660 6661 6662 6663 6664 6665 6666 6667 6668 6669 6670
	/*
	 * Here, we should temporarily reset sched_domains_numa_levels to 0.
	 * If it fails to allocate memory for array sched_domains_numa_masks[][],
	 * the array will contain less then 'level' members. This could be
	 * dangerous when we use it to iterate array sched_domains_numa_masks[][]
	 * in other functions.
	 *
	 * We reset it to 'level' at the end of this function.
	 */
	sched_domains_numa_levels = 0;

6671 6672 6673 6674 6675 6676 6677 6678 6679 6680 6681 6682 6683 6684 6685
	sched_domains_numa_masks = kzalloc(sizeof(void *) * level, GFP_KERNEL);
	if (!sched_domains_numa_masks)
		return;

	/*
	 * Now for each level, construct a mask per node which contains all
	 * cpus of nodes that are that many hops away from us.
	 */
	for (i = 0; i < level; i++) {
		sched_domains_numa_masks[i] =
			kzalloc(nr_node_ids * sizeof(void *), GFP_KERNEL);
		if (!sched_domains_numa_masks[i])
			return;

		for (j = 0; j < nr_node_ids; j++) {
6686
			struct cpumask *mask = kzalloc(cpumask_size(), GFP_KERNEL);
6687 6688 6689 6690 6691 6692
			if (!mask)
				return;

			sched_domains_numa_masks[i][j] = mask;

			for (k = 0; k < nr_node_ids; k++) {
6693
				if (node_distance(j, k) > sched_domains_numa_distance[i])
6694 6695 6696 6697 6698 6699 6700
					continue;

				cpumask_or(mask, mask, cpumask_of_node(k));
			}
		}
	}

6701 6702 6703
	/* Compute default topology size */
	for (i = 0; sched_domain_topology[i].mask; i++);

6704
	tl = kzalloc((i + level + 1) *
6705 6706 6707 6708 6709 6710 6711
			sizeof(struct sched_domain_topology_level), GFP_KERNEL);
	if (!tl)
		return;

	/*
	 * Copy the default topology bits..
	 */
6712 6713
	for (i = 0; sched_domain_topology[i].mask; i++)
		tl[i] = sched_domain_topology[i];
6714 6715 6716 6717 6718 6719 6720

	/*
	 * .. and append 'j' levels of NUMA goodness.
	 */
	for (j = 0; j < level; i++, j++) {
		tl[i] = (struct sched_domain_topology_level){
			.mask = sd_numa_mask,
6721
			.sd_flags = cpu_numa_flags,
6722 6723
			.flags = SDTL_OVERLAP,
			.numa_level = j,
6724
			SD_INIT_NAME(NUMA)
6725 6726 6727 6728
		};
	}

	sched_domain_topology = tl;
6729 6730

	sched_domains_numa_levels = level;
6731
	sched_max_numa_distance = sched_domains_numa_distance[level - 1];
6732 6733

	init_numa_topology_type();
6734
}
6735 6736 6737 6738 6739 6740 6741 6742 6743 6744 6745 6746 6747 6748 6749 6750 6751 6752 6753 6754 6755 6756 6757 6758 6759 6760 6761 6762 6763 6764 6765 6766 6767 6768 6769 6770 6771 6772 6773 6774 6775 6776 6777 6778 6779 6780 6781

static void sched_domains_numa_masks_set(int cpu)
{
	int i, j;
	int node = cpu_to_node(cpu);

	for (i = 0; i < sched_domains_numa_levels; i++) {
		for (j = 0; j < nr_node_ids; j++) {
			if (node_distance(j, node) <= sched_domains_numa_distance[i])
				cpumask_set_cpu(cpu, sched_domains_numa_masks[i][j]);
		}
	}
}

static void sched_domains_numa_masks_clear(int cpu)
{
	int i, j;
	for (i = 0; i < sched_domains_numa_levels; i++) {
		for (j = 0; j < nr_node_ids; j++)
			cpumask_clear_cpu(cpu, sched_domains_numa_masks[i][j]);
	}
}

/*
 * Update sched_domains_numa_masks[level][node] array when new cpus
 * are onlined.
 */
static int sched_domains_numa_masks_update(struct notifier_block *nfb,
					   unsigned long action,
					   void *hcpu)
{
	int cpu = (long)hcpu;

	switch (action & ~CPU_TASKS_FROZEN) {
	case CPU_ONLINE:
		sched_domains_numa_masks_set(cpu);
		break;

	case CPU_DEAD:
		sched_domains_numa_masks_clear(cpu);
		break;

	default:
		return NOTIFY_DONE;
	}

	return NOTIFY_OK;
6782 6783 6784 6785 6786
}
#else
static inline void sched_init_numa(void)
{
}
6787 6788 6789 6790 6791 6792 6793

static int sched_domains_numa_masks_update(struct notifier_block *nfb,
					   unsigned long action,
					   void *hcpu)
{
	return 0;
}
6794 6795
#endif /* CONFIG_NUMA */

6796 6797 6798 6799 6800
static int __sdt_alloc(const struct cpumask *cpu_map)
{
	struct sched_domain_topology_level *tl;
	int j;

6801
	for_each_sd_topology(tl) {
6802 6803 6804 6805 6806 6807 6808 6809 6810 6811
		struct sd_data *sdd = &tl->data;

		sdd->sd = alloc_percpu(struct sched_domain *);
		if (!sdd->sd)
			return -ENOMEM;

		sdd->sg = alloc_percpu(struct sched_group *);
		if (!sdd->sg)
			return -ENOMEM;

6812 6813
		sdd->sgc = alloc_percpu(struct sched_group_capacity *);
		if (!sdd->sgc)
6814 6815
			return -ENOMEM;

6816 6817 6818
		for_each_cpu(j, cpu_map) {
			struct sched_domain *sd;
			struct sched_group *sg;
6819
			struct sched_group_capacity *sgc;
6820

P
Peter Zijlstra 已提交
6821
			sd = kzalloc_node(sizeof(struct sched_domain) + cpumask_size(),
6822 6823 6824 6825 6826 6827 6828 6829 6830 6831 6832
					GFP_KERNEL, cpu_to_node(j));
			if (!sd)
				return -ENOMEM;

			*per_cpu_ptr(sdd->sd, j) = sd;

			sg = kzalloc_node(sizeof(struct sched_group) + cpumask_size(),
					GFP_KERNEL, cpu_to_node(j));
			if (!sg)
				return -ENOMEM;

6833 6834
			sg->next = sg;

6835
			*per_cpu_ptr(sdd->sg, j) = sg;
6836

6837
			sgc = kzalloc_node(sizeof(struct sched_group_capacity) + cpumask_size(),
6838
					GFP_KERNEL, cpu_to_node(j));
6839
			if (!sgc)
6840 6841
				return -ENOMEM;

6842
			*per_cpu_ptr(sdd->sgc, j) = sgc;
6843 6844 6845 6846 6847 6848 6849 6850 6851 6852 6853
		}
	}

	return 0;
}

static void __sdt_free(const struct cpumask *cpu_map)
{
	struct sched_domain_topology_level *tl;
	int j;

6854
	for_each_sd_topology(tl) {
6855 6856 6857
		struct sd_data *sdd = &tl->data;

		for_each_cpu(j, cpu_map) {
6858 6859 6860 6861 6862 6863 6864 6865 6866 6867 6868
			struct sched_domain *sd;

			if (sdd->sd) {
				sd = *per_cpu_ptr(sdd->sd, j);
				if (sd && (sd->flags & SD_OVERLAP))
					free_sched_groups(sd->groups, 0);
				kfree(*per_cpu_ptr(sdd->sd, j));
			}

			if (sdd->sg)
				kfree(*per_cpu_ptr(sdd->sg, j));
6869 6870
			if (sdd->sgc)
				kfree(*per_cpu_ptr(sdd->sgc, j));
6871 6872
		}
		free_percpu(sdd->sd);
6873
		sdd->sd = NULL;
6874
		free_percpu(sdd->sg);
6875
		sdd->sg = NULL;
6876 6877
		free_percpu(sdd->sgc);
		sdd->sgc = NULL;
6878 6879 6880
	}
}

6881
struct sched_domain *build_sched_domain(struct sched_domain_topology_level *tl,
6882 6883
		const struct cpumask *cpu_map, struct sched_domain_attr *attr,
		struct sched_domain *child, int cpu)
6884
{
6885
	struct sched_domain *sd = sd_init(tl, cpu);
6886
	if (!sd)
6887
		return child;
6888 6889

	cpumask_and(sched_domain_span(sd), cpu_map, tl->mask(cpu));
6890 6891 6892
	if (child) {
		sd->level = child->level + 1;
		sched_domain_level_max = max(sched_domain_level_max, sd->level);
6893
		child->parent = sd;
6894
		sd->child = child;
P
Peter Zijlstra 已提交
6895 6896 6897 6898 6899 6900 6901 6902 6903 6904 6905 6906 6907 6908

		if (!cpumask_subset(sched_domain_span(child),
				    sched_domain_span(sd))) {
			pr_err("BUG: arch topology borken\n");
#ifdef CONFIG_SCHED_DEBUG
			pr_err("     the %s domain not a subset of the %s domain\n",
					child->name, sd->name);
#endif
			/* Fixup, ensure @sd has at least @child cpus. */
			cpumask_or(sched_domain_span(sd),
				   sched_domain_span(sd),
				   sched_domain_span(child));
		}

6909
	}
6910
	set_domain_attribute(sd, attr);
6911 6912 6913 6914

	return sd;
}

6915 6916 6917 6918
/*
 * Build sched domains for a given set of cpus and attach the sched domains
 * to the individual cpus
 */
6919 6920
static int build_sched_domains(const struct cpumask *cpu_map,
			       struct sched_domain_attr *attr)
6921
{
6922
	enum s_alloc alloc_state;
6923
	struct sched_domain *sd;
6924
	struct s_data d;
6925
	int i, ret = -ENOMEM;
6926

6927 6928 6929
	alloc_state = __visit_domain_allocation_hell(&d, cpu_map);
	if (alloc_state != sa_rootdomain)
		goto error;
6930

6931
	/* Set up domains for cpus specified by the cpu_map. */
6932
	for_each_cpu(i, cpu_map) {
6933 6934
		struct sched_domain_topology_level *tl;

6935
		sd = NULL;
6936
		for_each_sd_topology(tl) {
6937
			sd = build_sched_domain(tl, cpu_map, attr, sd, i);
6938 6939
			if (tl == sched_domain_topology)
				*per_cpu_ptr(d.sd, i) = sd;
6940 6941
			if (tl->flags & SDTL_OVERLAP || sched_feat(FORCE_SD_OVERLAP))
				sd->flags |= SD_OVERLAP;
6942 6943
			if (cpumask_equal(cpu_map, sched_domain_span(sd)))
				break;
6944
		}
6945 6946 6947 6948 6949 6950
	}

	/* Build the groups for the domains */
	for_each_cpu(i, cpu_map) {
		for (sd = *per_cpu_ptr(d.sd, i); sd; sd = sd->parent) {
			sd->span_weight = cpumask_weight(sched_domain_span(sd));
6951 6952 6953 6954 6955 6956 6957
			if (sd->flags & SD_OVERLAP) {
				if (build_overlap_sched_groups(sd, i))
					goto error;
			} else {
				if (build_sched_groups(sd, i))
					goto error;
			}
6958
		}
6959
	}
6960

6961
	/* Calculate CPU capacity for physical packages and nodes */
6962 6963 6964
	for (i = nr_cpumask_bits-1; i >= 0; i--) {
		if (!cpumask_test_cpu(i, cpu_map))
			continue;
6965

6966 6967
		for (sd = *per_cpu_ptr(d.sd, i); sd; sd = sd->parent) {
			claim_allocations(i, sd);
6968
			init_sched_groups_capacity(i, sd);
6969
		}
6970
	}
6971

L
Linus Torvalds 已提交
6972
	/* Attach the domains */
6973
	rcu_read_lock();
6974
	for_each_cpu(i, cpu_map) {
6975
		sd = *per_cpu_ptr(d.sd, i);
6976
		cpu_attach_domain(sd, d.rd, i);
L
Linus Torvalds 已提交
6977
	}
6978
	rcu_read_unlock();
6979

6980
	ret = 0;
6981
error:
6982
	__free_domain_allocs(&d, alloc_state, cpu_map);
6983
	return ret;
L
Linus Torvalds 已提交
6984
}
P
Paul Jackson 已提交
6985

6986
static cpumask_var_t *doms_cur;	/* current sched domains */
P
Paul Jackson 已提交
6987
static int ndoms_cur;		/* number of sched domains in 'doms_cur' */
I
Ingo Molnar 已提交
6988 6989
static struct sched_domain_attr *dattr_cur;
				/* attribues of custom domains in 'doms_cur' */
P
Paul Jackson 已提交
6990 6991 6992

/*
 * Special case: If a kmalloc of a doms_cur partition (array of
6993 6994
 * cpumask) fails, then fallback to a single sched domain,
 * as determined by the single cpumask fallback_doms.
P
Paul Jackson 已提交
6995
 */
6996
static cpumask_var_t fallback_doms;
P
Paul Jackson 已提交
6997

6998 6999 7000 7001 7002
/*
 * arch_update_cpu_topology lets virtualized architectures update the
 * cpu core maps. It is supposed to return 1 if the topology changed
 * or 0 if it stayed the same.
 */
7003
int __weak arch_update_cpu_topology(void)
7004
{
7005
	return 0;
7006 7007
}

7008 7009 7010 7011 7012 7013 7014 7015 7016 7017 7018 7019 7020 7021 7022 7023 7024 7025 7026 7027 7028 7029 7030 7031 7032
cpumask_var_t *alloc_sched_domains(unsigned int ndoms)
{
	int i;
	cpumask_var_t *doms;

	doms = kmalloc(sizeof(*doms) * ndoms, GFP_KERNEL);
	if (!doms)
		return NULL;
	for (i = 0; i < ndoms; i++) {
		if (!alloc_cpumask_var(&doms[i], GFP_KERNEL)) {
			free_sched_domains(doms, i);
			return NULL;
		}
	}
	return doms;
}

void free_sched_domains(cpumask_var_t doms[], unsigned int ndoms)
{
	unsigned int i;
	for (i = 0; i < ndoms; i++)
		free_cpumask_var(doms[i]);
	kfree(doms);
}

7033
/*
I
Ingo Molnar 已提交
7034
 * Set up scheduler domains and groups. Callers must hold the hotplug lock.
P
Paul Jackson 已提交
7035 7036
 * For now this just excludes isolated cpus, but could be used to
 * exclude other special cases in the future.
7037
 */
7038
static int init_sched_domains(const struct cpumask *cpu_map)
7039
{
7040 7041
	int err;

7042
	arch_update_cpu_topology();
P
Paul Jackson 已提交
7043
	ndoms_cur = 1;
7044
	doms_cur = alloc_sched_domains(ndoms_cur);
P
Paul Jackson 已提交
7045
	if (!doms_cur)
7046 7047
		doms_cur = &fallback_doms;
	cpumask_andnot(doms_cur[0], cpu_map, cpu_isolated_map);
7048
	err = build_sched_domains(doms_cur[0], NULL);
7049
	register_sched_domain_sysctl();
7050 7051

	return err;
7052 7053 7054 7055 7056 7057
}

/*
 * Detach sched domains from a group of cpus specified in cpu_map
 * These cpus will now be attached to the NULL domain
 */
7058
static void detach_destroy_domains(const struct cpumask *cpu_map)
7059 7060 7061
{
	int i;

7062
	rcu_read_lock();
7063
	for_each_cpu(i, cpu_map)
G
Gregory Haskins 已提交
7064
		cpu_attach_domain(NULL, &def_root_domain, i);
7065
	rcu_read_unlock();
7066 7067
}

7068 7069 7070 7071 7072 7073 7074 7075 7076 7077 7078 7079 7080 7081 7082 7083
/* handle null as "default" */
static int dattrs_equal(struct sched_domain_attr *cur, int idx_cur,
			struct sched_domain_attr *new, int idx_new)
{
	struct sched_domain_attr tmp;

	/* fast path */
	if (!new && !cur)
		return 1;

	tmp = SD_ATTR_INIT;
	return !memcmp(cur ? (cur + idx_cur) : &tmp,
			new ? (new + idx_new) : &tmp,
			sizeof(struct sched_domain_attr));
}

P
Paul Jackson 已提交
7084 7085
/*
 * Partition sched domains as specified by the 'ndoms_new'
I
Ingo Molnar 已提交
7086
 * cpumasks in the array doms_new[] of cpumasks. This compares
P
Paul Jackson 已提交
7087 7088 7089
 * doms_new[] to the current sched domain partitioning, doms_cur[].
 * It destroys each deleted domain and builds each new domain.
 *
7090
 * 'doms_new' is an array of cpumask_var_t's of length 'ndoms_new'.
I
Ingo Molnar 已提交
7091 7092 7093
 * The masks don't intersect (don't overlap.) We should setup one
 * sched domain for each mask. CPUs not in any of the cpumasks will
 * not be load balanced. If the same cpumask appears both in the
P
Paul Jackson 已提交
7094 7095 7096
 * current 'doms_cur' domains and in the new 'doms_new', we can leave
 * it as it is.
 *
7097 7098 7099 7100 7101 7102
 * The passed in 'doms_new' should be allocated using
 * alloc_sched_domains.  This routine takes ownership of it and will
 * free_sched_domains it when done with it. If the caller failed the
 * alloc call, then it can pass in doms_new == NULL && ndoms_new == 1,
 * and partition_sched_domains() will fallback to the single partition
 * 'fallback_doms', it also forces the domains to be rebuilt.
P
Paul Jackson 已提交
7103
 *
7104
 * If doms_new == NULL it will be replaced with cpu_online_mask.
7105 7106
 * ndoms_new == 0 is a special case for destroying existing domains,
 * and it will not create the default domain.
7107
 *
P
Paul Jackson 已提交
7108 7109
 * Call with hotplug lock held
 */
7110
void partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
7111
			     struct sched_domain_attr *dattr_new)
P
Paul Jackson 已提交
7112
{
7113
	int i, j, n;
7114
	int new_topology;
P
Paul Jackson 已提交
7115

7116
	mutex_lock(&sched_domains_mutex);
7117

7118 7119 7120
	/* always unregister in case we don't destroy any domains */
	unregister_sched_domain_sysctl();

7121 7122 7123
	/* Let architecture update cpu core mappings. */
	new_topology = arch_update_cpu_topology();

7124
	n = doms_new ? ndoms_new : 0;
P
Paul Jackson 已提交
7125 7126 7127

	/* Destroy deleted domains */
	for (i = 0; i < ndoms_cur; i++) {
7128
		for (j = 0; j < n && !new_topology; j++) {
7129
			if (cpumask_equal(doms_cur[i], doms_new[j])
7130
			    && dattrs_equal(dattr_cur, i, dattr_new, j))
P
Paul Jackson 已提交
7131 7132 7133
				goto match1;
		}
		/* no match - a current sched domain not in new doms_new[] */
7134
		detach_destroy_domains(doms_cur[i]);
P
Paul Jackson 已提交
7135 7136 7137 7138
match1:
		;
	}

7139
	n = ndoms_cur;
7140
	if (doms_new == NULL) {
7141
		n = 0;
7142
		doms_new = &fallback_doms;
7143
		cpumask_andnot(doms_new[0], cpu_active_mask, cpu_isolated_map);
7144
		WARN_ON_ONCE(dattr_new);
7145 7146
	}

P
Paul Jackson 已提交
7147 7148
	/* Build new domains */
	for (i = 0; i < ndoms_new; i++) {
7149
		for (j = 0; j < n && !new_topology; j++) {
7150
			if (cpumask_equal(doms_new[i], doms_cur[j])
7151
			    && dattrs_equal(dattr_new, i, dattr_cur, j))
P
Paul Jackson 已提交
7152 7153 7154
				goto match2;
		}
		/* no match - add a new doms_new */
7155
		build_sched_domains(doms_new[i], dattr_new ? dattr_new + i : NULL);
P
Paul Jackson 已提交
7156 7157 7158 7159 7160
match2:
		;
	}

	/* Remember the new sched domains */
7161 7162
	if (doms_cur != &fallback_doms)
		free_sched_domains(doms_cur, ndoms_cur);
7163
	kfree(dattr_cur);	/* kfree(NULL) is safe */
P
Paul Jackson 已提交
7164
	doms_cur = doms_new;
7165
	dattr_cur = dattr_new;
P
Paul Jackson 已提交
7166
	ndoms_cur = ndoms_new;
7167 7168

	register_sched_domain_sysctl();
7169

7170
	mutex_unlock(&sched_domains_mutex);
P
Paul Jackson 已提交
7171 7172
}

7173 7174
static int num_cpus_frozen;	/* used to mark begin/end of suspend/resume */

L
Linus Torvalds 已提交
7175
/*
7176 7177 7178
 * Update cpusets according to cpu_active mask.  If cpusets are
 * disabled, cpuset_update_active_cpus() becomes a simple wrapper
 * around partition_sched_domains().
7179 7180 7181
 *
 * If we come here as part of a suspend/resume, don't touch cpusets because we
 * want to restore it back to its original state upon resume anyway.
L
Linus Torvalds 已提交
7182
 */
7183 7184
static int cpuset_cpu_active(struct notifier_block *nfb, unsigned long action,
			     void *hcpu)
7185
{
7186 7187 7188 7189 7190 7191 7192 7193 7194 7195 7196 7197 7198 7199 7200 7201 7202 7203 7204 7205 7206 7207
	switch (action) {
	case CPU_ONLINE_FROZEN:
	case CPU_DOWN_FAILED_FROZEN:

		/*
		 * 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.
		 */
		num_cpus_frozen--;
		if (likely(num_cpus_frozen)) {
			partition_sched_domains(1, NULL, NULL);
			break;
		}

		/*
		 * This is the last CPU online operation. So fall through and
		 * restore the original sched domains by considering the
		 * cpuset configurations.
		 */

7208
	case CPU_ONLINE:
7209
		cpuset_update_active_cpus(true);
7210
		break;
7211 7212 7213
	default:
		return NOTIFY_DONE;
	}
7214
	return NOTIFY_OK;
7215
}
7216

7217 7218
static int cpuset_cpu_inactive(struct notifier_block *nfb, unsigned long action,
			       void *hcpu)
7219
{
7220 7221 7222
	unsigned long flags;
	long cpu = (long)hcpu;
	struct dl_bw *dl_b;
7223 7224
	bool overflow;
	int cpus;
7225

7226
	switch (action) {
7227
	case CPU_DOWN_PREPARE:
7228 7229
		rcu_read_lock_sched();
		dl_b = dl_bw_of(cpu);
7230

7231 7232 7233 7234
		raw_spin_lock_irqsave(&dl_b->lock, flags);
		cpus = dl_bw_cpus(cpu);
		overflow = __dl_overflow(dl_b, cpus, 0, 0);
		raw_spin_unlock_irqrestore(&dl_b->lock, flags);
7235

7236
		rcu_read_unlock_sched();
7237

7238 7239
		if (overflow)
			return notifier_from_errno(-EBUSY);
7240
		cpuset_update_active_cpus(false);
7241 7242 7243 7244 7245
		break;
	case CPU_DOWN_PREPARE_FROZEN:
		num_cpus_frozen++;
		partition_sched_domains(1, NULL, NULL);
		break;
7246 7247 7248
	default:
		return NOTIFY_DONE;
	}
7249
	return NOTIFY_OK;
7250 7251
}

L
Linus Torvalds 已提交
7252 7253
void __init sched_init_smp(void)
{
7254 7255 7256
	cpumask_var_t non_isolated_cpus;

	alloc_cpumask_var(&non_isolated_cpus, GFP_KERNEL);
7257
	alloc_cpumask_var(&fallback_doms, GFP_KERNEL);
7258

7259 7260 7261
	/* nohz_full won't take effect without isolating the cpus. */
	tick_nohz_full_add_cpus_to(cpu_isolated_map);

7262 7263
	sched_init_numa();

7264 7265 7266 7267 7268
	/*
	 * There's no userspace yet to cause hotplug operations; hence all the
	 * cpu masks are stable and all blatant races in the below code cannot
	 * happen.
	 */
7269
	mutex_lock(&sched_domains_mutex);
7270
	init_sched_domains(cpu_active_mask);
7271 7272 7273
	cpumask_andnot(non_isolated_cpus, cpu_possible_mask, cpu_isolated_map);
	if (cpumask_empty(non_isolated_cpus))
		cpumask_set_cpu(smp_processor_id(), non_isolated_cpus);
7274
	mutex_unlock(&sched_domains_mutex);
7275

7276
	hotcpu_notifier(sched_domains_numa_masks_update, CPU_PRI_SCHED_ACTIVE);
7277 7278
	hotcpu_notifier(cpuset_cpu_active, CPU_PRI_CPUSET_ACTIVE);
	hotcpu_notifier(cpuset_cpu_inactive, CPU_PRI_CPUSET_INACTIVE);
7279

7280
	init_hrtick();
7281 7282

	/* Move init over to a non-isolated CPU */
7283
	if (set_cpus_allowed_ptr(current, non_isolated_cpus) < 0)
7284
		BUG();
I
Ingo Molnar 已提交
7285
	sched_init_granularity();
7286
	free_cpumask_var(non_isolated_cpus);
7287

7288
	init_sched_rt_class();
7289
	init_sched_dl_class();
L
Linus Torvalds 已提交
7290 7291 7292 7293
}
#else
void __init sched_init_smp(void)
{
I
Ingo Molnar 已提交
7294
	sched_init_granularity();
L
Linus Torvalds 已提交
7295 7296 7297 7298 7299 7300 7301 7302 7303 7304
}
#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);
}

7305
#ifdef CONFIG_CGROUP_SCHED
7306 7307 7308 7309
/*
 * Default task group.
 * Every task in system belongs to this group at bootup.
 */
7310
struct task_group root_task_group;
7311
LIST_HEAD(task_groups);
7312
#endif
P
Peter Zijlstra 已提交
7313

7314
DECLARE_PER_CPU(cpumask_var_t, load_balance_mask);
P
Peter Zijlstra 已提交
7315

L
Linus Torvalds 已提交
7316 7317
void __init sched_init(void)
{
I
Ingo Molnar 已提交
7318
	int i, j;
7319 7320 7321 7322 7323 7324 7325 7326 7327
	unsigned long alloc_size = 0, ptr;

#ifdef CONFIG_FAIR_GROUP_SCHED
	alloc_size += 2 * nr_cpu_ids * sizeof(void **);
#endif
#ifdef CONFIG_RT_GROUP_SCHED
	alloc_size += 2 * nr_cpu_ids * sizeof(void **);
#endif
	if (alloc_size) {
7328
		ptr = (unsigned long)kzalloc(alloc_size, GFP_NOWAIT);
7329 7330

#ifdef CONFIG_FAIR_GROUP_SCHED
7331
		root_task_group.se = (struct sched_entity **)ptr;
7332 7333
		ptr += nr_cpu_ids * sizeof(void **);

7334
		root_task_group.cfs_rq = (struct cfs_rq **)ptr;
7335
		ptr += nr_cpu_ids * sizeof(void **);
7336

7337
#endif /* CONFIG_FAIR_GROUP_SCHED */
7338
#ifdef CONFIG_RT_GROUP_SCHED
7339
		root_task_group.rt_se = (struct sched_rt_entity **)ptr;
7340 7341
		ptr += nr_cpu_ids * sizeof(void **);

7342
		root_task_group.rt_rq = (struct rt_rq **)ptr;
7343 7344
		ptr += nr_cpu_ids * sizeof(void **);

7345
#endif /* CONFIG_RT_GROUP_SCHED */
7346
	}
7347
#ifdef CONFIG_CPUMASK_OFFSTACK
7348 7349 7350
	for_each_possible_cpu(i) {
		per_cpu(load_balance_mask, i) = (cpumask_var_t)kzalloc_node(
			cpumask_size(), GFP_KERNEL, cpu_to_node(i));
7351
	}
7352
#endif /* CONFIG_CPUMASK_OFFSTACK */
I
Ingo Molnar 已提交
7353

7354 7355 7356
	init_rt_bandwidth(&def_rt_bandwidth,
			global_rt_period(), global_rt_runtime());
	init_dl_bandwidth(&def_dl_bandwidth,
7357
			global_rt_period(), global_rt_runtime());
7358

G
Gregory Haskins 已提交
7359 7360 7361 7362
#ifdef CONFIG_SMP
	init_defrootdomain();
#endif

7363
#ifdef CONFIG_RT_GROUP_SCHED
7364
	init_rt_bandwidth(&root_task_group.rt_bandwidth,
7365
			global_rt_period(), global_rt_runtime());
7366
#endif /* CONFIG_RT_GROUP_SCHED */
7367

D
Dhaval Giani 已提交
7368
#ifdef CONFIG_CGROUP_SCHED
7369 7370
	list_add(&root_task_group.list, &task_groups);
	INIT_LIST_HEAD(&root_task_group.children);
7371
	INIT_LIST_HEAD(&root_task_group.siblings);
7372
	autogroup_init(&init_task);
7373

D
Dhaval Giani 已提交
7374
#endif /* CONFIG_CGROUP_SCHED */
P
Peter Zijlstra 已提交
7375

7376
	for_each_possible_cpu(i) {
7377
		struct rq *rq;
L
Linus Torvalds 已提交
7378 7379

		rq = cpu_rq(i);
7380
		raw_spin_lock_init(&rq->lock);
N
Nick Piggin 已提交
7381
		rq->nr_running = 0;
7382 7383
		rq->calc_load_active = 0;
		rq->calc_load_update = jiffies + LOAD_FREQ;
7384
		init_cfs_rq(&rq->cfs);
7385 7386
		init_rt_rq(&rq->rt);
		init_dl_rq(&rq->dl);
I
Ingo Molnar 已提交
7387
#ifdef CONFIG_FAIR_GROUP_SCHED
7388
		root_task_group.shares = ROOT_TASK_GROUP_LOAD;
P
Peter Zijlstra 已提交
7389
		INIT_LIST_HEAD(&rq->leaf_cfs_rq_list);
D
Dhaval Giani 已提交
7390
		/*
7391
		 * How much cpu bandwidth does root_task_group get?
D
Dhaval Giani 已提交
7392 7393 7394 7395
		 *
		 * In case of task-groups formed thr' the cgroup filesystem, it
		 * gets 100% of the cpu resources in the system. This overall
		 * system cpu resource is divided among the tasks of
7396
		 * root_task_group and its child task-groups in a fair manner,
D
Dhaval Giani 已提交
7397 7398 7399
		 * based on each entity's (task or task-group's) weight
		 * (se->load.weight).
		 *
7400
		 * In other words, if root_task_group has 10 tasks of weight
D
Dhaval Giani 已提交
7401 7402 7403
		 * 1024) and two child groups A0 and A1 (of weight 1024 each),
		 * then A0's share of the cpu resource is:
		 *
7404
		 *	A0's bandwidth = 1024 / (10*1024 + 1024 + 1024) = 8.33%
D
Dhaval Giani 已提交
7405
		 *
7406 7407
		 * We achieve this by letting root_task_group's tasks sit
		 * directly in rq->cfs (i.e root_task_group->se[] = NULL).
D
Dhaval Giani 已提交
7408
		 */
7409
		init_cfs_bandwidth(&root_task_group.cfs_bandwidth);
7410
		init_tg_cfs_entry(&root_task_group, &rq->cfs, NULL, i, NULL);
D
Dhaval Giani 已提交
7411 7412 7413
#endif /* CONFIG_FAIR_GROUP_SCHED */

		rq->rt.rt_runtime = def_rt_bandwidth.rt_runtime;
7414
#ifdef CONFIG_RT_GROUP_SCHED
7415
		init_tg_rt_entry(&root_task_group, &rq->rt, NULL, i, NULL);
I
Ingo Molnar 已提交
7416
#endif
L
Linus Torvalds 已提交
7417

I
Ingo Molnar 已提交
7418 7419
		for (j = 0; j < CPU_LOAD_IDX_MAX; j++)
			rq->cpu_load[j] = 0;
7420 7421 7422

		rq->last_load_update_tick = jiffies;

L
Linus Torvalds 已提交
7423
#ifdef CONFIG_SMP
N
Nick Piggin 已提交
7424
		rq->sd = NULL;
G
Gregory Haskins 已提交
7425
		rq->rd = NULL;
7426
		rq->cpu_capacity = rq->cpu_capacity_orig = SCHED_CAPACITY_SCALE;
7427
		rq->balance_callback = NULL;
L
Linus Torvalds 已提交
7428
		rq->active_balance = 0;
I
Ingo Molnar 已提交
7429
		rq->next_balance = jiffies;
L
Linus Torvalds 已提交
7430
		rq->push_cpu = 0;
7431
		rq->cpu = i;
7432
		rq->online = 0;
7433 7434
		rq->idle_stamp = 0;
		rq->avg_idle = 2*sysctl_sched_migration_cost;
7435
		rq->max_idle_balance_cost = sysctl_sched_migration_cost;
7436 7437 7438

		INIT_LIST_HEAD(&rq->cfs_tasks);

7439
		rq_attach_root(rq, &def_root_domain);
7440
#ifdef CONFIG_NO_HZ_COMMON
7441
		rq->nohz_flags = 0;
7442
#endif
7443 7444 7445
#ifdef CONFIG_NO_HZ_FULL
		rq->last_sched_tick = 0;
#endif
L
Linus Torvalds 已提交
7446
#endif
P
Peter Zijlstra 已提交
7447
		init_rq_hrtick(rq);
L
Linus Torvalds 已提交
7448 7449 7450
		atomic_set(&rq->nr_iowait, 0);
	}

7451
	set_load_weight(&init_task);
7452

7453 7454 7455 7456
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&init_task.preempt_notifiers);
#endif

L
Linus Torvalds 已提交
7457 7458 7459 7460 7461 7462
	/*
	 * The boot idle thread does lazy MMU switching as well:
	 */
	atomic_inc(&init_mm.mm_count);
	enter_lazy_tlb(&init_mm, current);

7463 7464 7465 7466 7467
	/*
	 * During early bootup we pretend to be a normal task:
	 */
	current->sched_class = &fair_sched_class;

L
Linus Torvalds 已提交
7468 7469 7470 7471 7472 7473 7474
	/*
	 * 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());
7475 7476 7477

	calc_load_update = jiffies + LOAD_FREQ;

7478
#ifdef CONFIG_SMP
7479
	zalloc_cpumask_var(&sched_domains_tmpmask, GFP_NOWAIT);
R
Rusty Russell 已提交
7480 7481 7482
	/* May be allocated at isolcpus cmdline parse time */
	if (cpu_isolated_map == NULL)
		zalloc_cpumask_var(&cpu_isolated_map, GFP_NOWAIT);
7483
	idle_thread_set_boot_cpu();
7484
	set_cpu_rq_start_time();
7485 7486
#endif
	init_sched_fair_class();
7487

7488
	scheduler_running = 1;
L
Linus Torvalds 已提交
7489 7490
}

7491
#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
7492 7493
static inline int preempt_count_equals(int preempt_offset)
{
7494
	int nested = (preempt_count() & ~PREEMPT_ACTIVE) + rcu_preempt_depth();
7495

A
Arnd Bergmann 已提交
7496
	return (nested == preempt_offset);
7497 7498
}

7499
void __might_sleep(const char *file, int line, int preempt_offset)
L
Linus Torvalds 已提交
7500
{
P
Peter Zijlstra 已提交
7501 7502 7503 7504 7505
	/*
	 * 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.
	 */
7506
	WARN_ONCE(current->state != TASK_RUNNING && current->task_state_change,
P
Peter Zijlstra 已提交
7507 7508 7509 7510
			"do not call blocking ops when !TASK_RUNNING; "
			"state=%lx set at [<%p>] %pS\n",
			current->state,
			(void *)current->task_state_change,
7511
			(void *)current->task_state_change);
P
Peter Zijlstra 已提交
7512

7513 7514 7515 7516 7517
	___might_sleep(file, line, preempt_offset);
}
EXPORT_SYMBOL(__might_sleep);

void ___might_sleep(const char *file, int line, int preempt_offset)
L
Linus Torvalds 已提交
7518 7519 7520
{
	static unsigned long prev_jiffy;	/* ratelimiting */

7521
	rcu_sleep_check(); /* WARN_ON_ONCE() by default, no rate limit reqd. */
7522 7523
	if ((preempt_count_equals(preempt_offset) && !irqs_disabled() &&
	     !is_idle_task(current)) ||
7524
	    system_state != SYSTEM_RUNNING || oops_in_progress)
I
Ingo Molnar 已提交
7525 7526 7527 7528 7529
		return;
	if (time_before(jiffies, prev_jiffy + HZ) && prev_jiffy)
		return;
	prev_jiffy = jiffies;

P
Peter Zijlstra 已提交
7530 7531 7532 7533 7534 7535 7536
	printk(KERN_ERR
		"BUG: sleeping function called from invalid context at %s:%d\n",
			file, line);
	printk(KERN_ERR
		"in_atomic(): %d, irqs_disabled(): %d, pid: %d, name: %s\n",
			in_atomic(), irqs_disabled(),
			current->pid, current->comm);
I
Ingo Molnar 已提交
7537

7538 7539 7540
	if (task_stack_end_corrupted(current))
		printk(KERN_EMERG "Thread overran stack, or stack corrupted\n");

I
Ingo Molnar 已提交
7541 7542 7543
	debug_show_held_locks(current);
	if (irqs_disabled())
		print_irqtrace_events(current);
7544 7545 7546 7547 7548 7549 7550
#ifdef CONFIG_DEBUG_PREEMPT
	if (!preempt_count_equals(preempt_offset)) {
		pr_err("Preemption disabled at:");
		print_ip_sym(current->preempt_disable_ip);
		pr_cont("\n");
	}
#endif
I
Ingo Molnar 已提交
7551
	dump_stack();
L
Linus Torvalds 已提交
7552
}
7553
EXPORT_SYMBOL(___might_sleep);
L
Linus Torvalds 已提交
7554 7555 7556
#endif

#ifdef CONFIG_MAGIC_SYSRQ
7557
void normalize_rt_tasks(void)
7558
{
7559
	struct task_struct *g, *p;
7560 7561 7562
	struct sched_attr attr = {
		.sched_policy = SCHED_NORMAL,
	};
L
Linus Torvalds 已提交
7563

7564
	read_lock(&tasklist_lock);
7565
	for_each_process_thread(g, p) {
7566 7567 7568
		/*
		 * Only normalize user tasks:
		 */
7569
		if (p->flags & PF_KTHREAD)
7570 7571
			continue;

I
Ingo Molnar 已提交
7572 7573
		p->se.exec_start		= 0;
#ifdef CONFIG_SCHEDSTATS
7574 7575 7576
		p->se.statistics.wait_start	= 0;
		p->se.statistics.sleep_start	= 0;
		p->se.statistics.block_start	= 0;
I
Ingo Molnar 已提交
7577
#endif
I
Ingo Molnar 已提交
7578

7579
		if (!dl_task(p) && !rt_task(p)) {
I
Ingo Molnar 已提交
7580 7581 7582 7583
			/*
			 * Renice negative nice level userspace
			 * tasks back to 0:
			 */
7584
			if (task_nice(p) < 0)
I
Ingo Molnar 已提交
7585
				set_user_nice(p, 0);
L
Linus Torvalds 已提交
7586
			continue;
I
Ingo Molnar 已提交
7587
		}
L
Linus Torvalds 已提交
7588

7589
		__sched_setscheduler(p, &attr, false, false);
7590
	}
7591
	read_unlock(&tasklist_lock);
L
Linus Torvalds 已提交
7592 7593 7594
}

#endif /* CONFIG_MAGIC_SYSRQ */
7595

7596
#if defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB)
7597
/*
7598
 * These functions are only useful for the IA64 MCA handling, or kdb.
7599 7600 7601 7602 7603 7604 7605 7606 7607 7608 7609 7610 7611
 *
 * 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.
 */

/**
 * curr_task - return the current task for a given cpu.
 * @cpu: the processor in question.
 *
 * ONLY VALID WHEN THE WHOLE SYSTEM IS STOPPED!
7612 7613
 *
 * Return: The current task for @cpu.
7614
 */
7615
struct task_struct *curr_task(int cpu)
7616 7617 7618 7619
{
	return cpu_curr(cpu);
}

7620 7621 7622
#endif /* defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB) */

#ifdef CONFIG_IA64
7623 7624 7625 7626 7627 7628
/**
 * set_curr_task - set the current task for a given cpu.
 * @cpu: the processor in question.
 * @p: the task pointer to set.
 *
 * Description: This function must only be used when non-maskable interrupts
I
Ingo Molnar 已提交
7629 7630
 * are serviced on a separate stack. It allows the architecture to switch the
 * notion of the current task on a cpu in a non-blocking manner. This function
7631 7632 7633 7634 7635 7636 7637
 * 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!
 */
7638
void set_curr_task(int cpu, struct task_struct *p)
7639 7640 7641 7642 7643
{
	cpu_curr(cpu) = p;
}

#endif
S
Srivatsa Vaddagiri 已提交
7644

D
Dhaval Giani 已提交
7645
#ifdef CONFIG_CGROUP_SCHED
7646 7647 7648
/* task_group_lock serializes the addition/removal of task groups */
static DEFINE_SPINLOCK(task_group_lock);

7649 7650 7651 7652
static void free_sched_group(struct task_group *tg)
{
	free_fair_sched_group(tg);
	free_rt_sched_group(tg);
7653
	autogroup_free(tg);
7654 7655 7656 7657
	kfree(tg);
}

/* allocate runqueue etc for a new task group */
7658
struct task_group *sched_create_group(struct task_group *parent)
7659 7660 7661 7662 7663 7664 7665
{
	struct task_group *tg;

	tg = kzalloc(sizeof(*tg), GFP_KERNEL);
	if (!tg)
		return ERR_PTR(-ENOMEM);

7666
	if (!alloc_fair_sched_group(tg, parent))
7667 7668
		goto err;

7669
	if (!alloc_rt_sched_group(tg, parent))
7670 7671
		goto err;

7672 7673 7674 7675 7676 7677 7678 7679 7680 7681 7682
	return tg;

err:
	free_sched_group(tg);
	return ERR_PTR(-ENOMEM);
}

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

7683
	spin_lock_irqsave(&task_group_lock, flags);
P
Peter Zijlstra 已提交
7684
	list_add_rcu(&tg->list, &task_groups);
P
Peter Zijlstra 已提交
7685 7686 7687 7688 7689

	WARN_ON(!parent); /* root should already exist */

	tg->parent = parent;
	INIT_LIST_HEAD(&tg->children);
7690
	list_add_rcu(&tg->siblings, &parent->children);
7691
	spin_unlock_irqrestore(&task_group_lock, flags);
S
Srivatsa Vaddagiri 已提交
7692 7693
}

7694
/* rcu callback to free various structures associated with a task group */
P
Peter Zijlstra 已提交
7695
static void free_sched_group_rcu(struct rcu_head *rhp)
S
Srivatsa Vaddagiri 已提交
7696 7697
{
	/* now it should be safe to free those cfs_rqs */
P
Peter Zijlstra 已提交
7698
	free_sched_group(container_of(rhp, struct task_group, rcu));
S
Srivatsa Vaddagiri 已提交
7699 7700
}

7701
/* Destroy runqueue etc associated with a task group */
7702
void sched_destroy_group(struct task_group *tg)
7703 7704 7705 7706 7707 7708
{
	/* wait for possible concurrent references to cfs_rqs complete */
	call_rcu(&tg->rcu, free_sched_group_rcu);
}

void sched_offline_group(struct task_group *tg)
S
Srivatsa Vaddagiri 已提交
7709
{
7710
	unsigned long flags;
7711
	int i;
S
Srivatsa Vaddagiri 已提交
7712

7713 7714
	/* end participation in shares distribution */
	for_each_possible_cpu(i)
7715
		unregister_fair_sched_group(tg, i);
7716 7717

	spin_lock_irqsave(&task_group_lock, flags);
P
Peter Zijlstra 已提交
7718
	list_del_rcu(&tg->list);
P
Peter Zijlstra 已提交
7719
	list_del_rcu(&tg->siblings);
7720
	spin_unlock_irqrestore(&task_group_lock, flags);
S
Srivatsa Vaddagiri 已提交
7721 7722
}

7723
/* change task's runqueue when it moves between groups.
I
Ingo Molnar 已提交
7724 7725 7726
 *	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.
7727 7728
 */
void sched_move_task(struct task_struct *tsk)
S
Srivatsa Vaddagiri 已提交
7729
{
P
Peter Zijlstra 已提交
7730
	struct task_group *tg;
7731
	int queued, running;
S
Srivatsa Vaddagiri 已提交
7732 7733 7734 7735 7736
	unsigned long flags;
	struct rq *rq;

	rq = task_rq_lock(tsk, &flags);

7737
	running = task_current(rq, tsk);
7738
	queued = task_on_rq_queued(tsk);
S
Srivatsa Vaddagiri 已提交
7739

7740
	if (queued)
S
Srivatsa Vaddagiri 已提交
7741
		dequeue_task(rq, tsk, 0);
7742
	if (unlikely(running))
7743
		put_prev_task(rq, tsk);
S
Srivatsa Vaddagiri 已提交
7744

7745 7746 7747 7748 7749 7750
	/*
	 * All callers are synchronized by task_rq_lock(); we do not use RCU
	 * which is pointless here. Thus, we pass "true" to task_css_check()
	 * to prevent lockdep warnings.
	 */
	tg = container_of(task_css_check(tsk, cpu_cgrp_id, true),
P
Peter Zijlstra 已提交
7751 7752 7753 7754
			  struct task_group, css);
	tg = autogroup_task_group(tsk, tg);
	tsk->sched_task_group = tg;

P
Peter Zijlstra 已提交
7755
#ifdef CONFIG_FAIR_GROUP_SCHED
7756
	if (tsk->sched_class->task_move_group)
7757
		tsk->sched_class->task_move_group(tsk);
7758
	else
P
Peter Zijlstra 已提交
7759
#endif
7760
		set_task_rq(tsk, task_cpu(tsk));
P
Peter Zijlstra 已提交
7761

7762 7763
	if (unlikely(running))
		tsk->sched_class->set_curr_task(rq);
7764
	if (queued)
7765
		enqueue_task(rq, tsk, 0);
S
Srivatsa Vaddagiri 已提交
7766

7767
	task_rq_unlock(rq, tsk, &flags);
S
Srivatsa Vaddagiri 已提交
7768
}
D
Dhaval Giani 已提交
7769
#endif /* CONFIG_CGROUP_SCHED */
S
Srivatsa Vaddagiri 已提交
7770

7771 7772 7773 7774 7775
#ifdef CONFIG_RT_GROUP_SCHED
/*
 * Ensure that the real time constraints are schedulable.
 */
static DEFINE_MUTEX(rt_constraints_mutex);
P
Peter Zijlstra 已提交
7776

P
Peter Zijlstra 已提交
7777 7778
/* Must be called with tasklist_lock held */
static inline int tg_has_rt_tasks(struct task_group *tg)
7779
{
P
Peter Zijlstra 已提交
7780
	struct task_struct *g, *p;
7781

7782 7783 7784 7785 7786 7787
	/*
	 * Autogroups do not have RT tasks; see autogroup_create().
	 */
	if (task_group_is_autogroup(tg))
		return 0;

7788
	for_each_process_thread(g, p) {
7789
		if (rt_task(p) && task_group(p) == tg)
P
Peter Zijlstra 已提交
7790
			return 1;
7791
	}
7792

P
Peter Zijlstra 已提交
7793 7794
	return 0;
}
7795

P
Peter Zijlstra 已提交
7796 7797 7798 7799 7800
struct rt_schedulable_data {
	struct task_group *tg;
	u64 rt_period;
	u64 rt_runtime;
};
7801

7802
static int tg_rt_schedulable(struct task_group *tg, void *data)
P
Peter Zijlstra 已提交
7803 7804 7805 7806 7807
{
	struct rt_schedulable_data *d = data;
	struct task_group *child;
	unsigned long total, sum = 0;
	u64 period, runtime;
7808

P
Peter Zijlstra 已提交
7809 7810
	period = ktime_to_ns(tg->rt_bandwidth.rt_period);
	runtime = tg->rt_bandwidth.rt_runtime;
7811

P
Peter Zijlstra 已提交
7812 7813 7814
	if (tg == d->tg) {
		period = d->rt_period;
		runtime = d->rt_runtime;
7815 7816
	}

7817 7818 7819 7820 7821
	/*
	 * Cannot have more runtime than the period.
	 */
	if (runtime > period && runtime != RUNTIME_INF)
		return -EINVAL;
P
Peter Zijlstra 已提交
7822

7823 7824 7825
	/*
	 * Ensure we don't starve existing RT tasks.
	 */
P
Peter Zijlstra 已提交
7826 7827
	if (rt_bandwidth_enabled() && !runtime && tg_has_rt_tasks(tg))
		return -EBUSY;
P
Peter Zijlstra 已提交
7828

P
Peter Zijlstra 已提交
7829
	total = to_ratio(period, runtime);
P
Peter Zijlstra 已提交
7830

7831 7832 7833 7834 7835
	/*
	 * Nobody can have more than the global setting allows.
	 */
	if (total > to_ratio(global_rt_period(), global_rt_runtime()))
		return -EINVAL;
P
Peter Zijlstra 已提交
7836

7837 7838 7839
	/*
	 * The sum of our children's runtime should not exceed our own.
	 */
P
Peter Zijlstra 已提交
7840 7841 7842
	list_for_each_entry_rcu(child, &tg->children, siblings) {
		period = ktime_to_ns(child->rt_bandwidth.rt_period);
		runtime = child->rt_bandwidth.rt_runtime;
P
Peter Zijlstra 已提交
7843

P
Peter Zijlstra 已提交
7844 7845 7846 7847
		if (child == d->tg) {
			period = d->rt_period;
			runtime = d->rt_runtime;
		}
P
Peter Zijlstra 已提交
7848

P
Peter Zijlstra 已提交
7849
		sum += to_ratio(period, runtime);
P
Peter Zijlstra 已提交
7850
	}
P
Peter Zijlstra 已提交
7851

P
Peter Zijlstra 已提交
7852 7853 7854 7855
	if (sum > total)
		return -EINVAL;

	return 0;
P
Peter Zijlstra 已提交
7856 7857
}

P
Peter Zijlstra 已提交
7858
static int __rt_schedulable(struct task_group *tg, u64 period, u64 runtime)
7859
{
7860 7861
	int ret;

P
Peter Zijlstra 已提交
7862 7863 7864 7865 7866 7867
	struct rt_schedulable_data data = {
		.tg = tg,
		.rt_period = period,
		.rt_runtime = runtime,
	};

7868 7869 7870 7871 7872
	rcu_read_lock();
	ret = walk_tg_tree(tg_rt_schedulable, tg_nop, &data);
	rcu_read_unlock();

	return ret;
7873 7874
}

7875
static int tg_set_rt_bandwidth(struct task_group *tg,
7876
		u64 rt_period, u64 rt_runtime)
P
Peter Zijlstra 已提交
7877
{
P
Peter Zijlstra 已提交
7878
	int i, err = 0;
P
Peter Zijlstra 已提交
7879

7880 7881 7882 7883 7884 7885 7886 7887 7888 7889 7890
	/*
	 * Disallowing the root group RT runtime is BAD, it would disallow the
	 * kernel creating (and or operating) RT threads.
	 */
	if (tg == &root_task_group && rt_runtime == 0)
		return -EINVAL;

	/* No period doesn't make any sense. */
	if (rt_period == 0)
		return -EINVAL;

P
Peter Zijlstra 已提交
7891
	mutex_lock(&rt_constraints_mutex);
7892
	read_lock(&tasklist_lock);
P
Peter Zijlstra 已提交
7893 7894
	err = __rt_schedulable(tg, rt_period, rt_runtime);
	if (err)
P
Peter Zijlstra 已提交
7895
		goto unlock;
P
Peter Zijlstra 已提交
7896

7897
	raw_spin_lock_irq(&tg->rt_bandwidth.rt_runtime_lock);
7898 7899
	tg->rt_bandwidth.rt_period = ns_to_ktime(rt_period);
	tg->rt_bandwidth.rt_runtime = rt_runtime;
P
Peter Zijlstra 已提交
7900 7901 7902 7903

	for_each_possible_cpu(i) {
		struct rt_rq *rt_rq = tg->rt_rq[i];

7904
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
7905
		rt_rq->rt_runtime = rt_runtime;
7906
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
7907
	}
7908
	raw_spin_unlock_irq(&tg->rt_bandwidth.rt_runtime_lock);
P
Peter Zijlstra 已提交
7909
unlock:
7910
	read_unlock(&tasklist_lock);
P
Peter Zijlstra 已提交
7911 7912 7913
	mutex_unlock(&rt_constraints_mutex);

	return err;
P
Peter Zijlstra 已提交
7914 7915
}

7916
static int sched_group_set_rt_runtime(struct task_group *tg, long rt_runtime_us)
7917 7918 7919 7920 7921 7922 7923 7924
{
	u64 rt_runtime, rt_period;

	rt_period = ktime_to_ns(tg->rt_bandwidth.rt_period);
	rt_runtime = (u64)rt_runtime_us * NSEC_PER_USEC;
	if (rt_runtime_us < 0)
		rt_runtime = RUNTIME_INF;

7925
	return tg_set_rt_bandwidth(tg, rt_period, rt_runtime);
7926 7927
}

7928
static long sched_group_rt_runtime(struct task_group *tg)
P
Peter Zijlstra 已提交
7929 7930 7931
{
	u64 rt_runtime_us;

7932
	if (tg->rt_bandwidth.rt_runtime == RUNTIME_INF)
P
Peter Zijlstra 已提交
7933 7934
		return -1;

7935
	rt_runtime_us = tg->rt_bandwidth.rt_runtime;
P
Peter Zijlstra 已提交
7936 7937 7938
	do_div(rt_runtime_us, NSEC_PER_USEC);
	return rt_runtime_us;
}
7939

7940
static int sched_group_set_rt_period(struct task_group *tg, u64 rt_period_us)
7941 7942 7943
{
	u64 rt_runtime, rt_period;

7944
	rt_period = rt_period_us * NSEC_PER_USEC;
7945 7946
	rt_runtime = tg->rt_bandwidth.rt_runtime;

7947
	return tg_set_rt_bandwidth(tg, rt_period, rt_runtime);
7948 7949
}

7950
static long sched_group_rt_period(struct task_group *tg)
7951 7952 7953 7954 7955 7956 7957
{
	u64 rt_period_us;

	rt_period_us = ktime_to_ns(tg->rt_bandwidth.rt_period);
	do_div(rt_period_us, NSEC_PER_USEC);
	return rt_period_us;
}
7958
#endif /* CONFIG_RT_GROUP_SCHED */
7959

7960
#ifdef CONFIG_RT_GROUP_SCHED
7961 7962 7963 7964 7965
static int sched_rt_global_constraints(void)
{
	int ret = 0;

	mutex_lock(&rt_constraints_mutex);
P
Peter Zijlstra 已提交
7966
	read_lock(&tasklist_lock);
7967
	ret = __rt_schedulable(NULL, 0, 0);
P
Peter Zijlstra 已提交
7968
	read_unlock(&tasklist_lock);
7969 7970 7971 7972
	mutex_unlock(&rt_constraints_mutex);

	return ret;
}
7973

7974
static int sched_rt_can_attach(struct task_group *tg, struct task_struct *tsk)
7975 7976 7977 7978 7979 7980 7981 7982
{
	/* Don't accept realtime tasks when there is no way for them to run */
	if (rt_task(tsk) && tg->rt_bandwidth.rt_runtime == 0)
		return 0;

	return 1;
}

7983
#else /* !CONFIG_RT_GROUP_SCHED */
7984 7985
static int sched_rt_global_constraints(void)
{
P
Peter Zijlstra 已提交
7986
	unsigned long flags;
7987
	int i, ret = 0;
7988

7989
	raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags);
P
Peter Zijlstra 已提交
7990 7991 7992
	for_each_possible_cpu(i) {
		struct rt_rq *rt_rq = &cpu_rq(i)->rt;

7993
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
7994
		rt_rq->rt_runtime = global_rt_runtime();
7995
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
7996
	}
7997
	raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags);
P
Peter Zijlstra 已提交
7998

7999
	return ret;
8000
}
8001
#endif /* CONFIG_RT_GROUP_SCHED */
8002

8003
static int sched_dl_global_validate(void)
8004
{
8005 8006
	u64 runtime = global_rt_runtime();
	u64 period = global_rt_period();
8007
	u64 new_bw = to_ratio(period, runtime);
8008
	struct dl_bw *dl_b;
8009
	int cpu, ret = 0;
8010
	unsigned long flags;
8011 8012 8013 8014 8015 8016 8017 8018 8019 8020

	/*
	 * Here we want to check the bandwidth not being set to some
	 * value smaller than the currently allocated bandwidth in
	 * any of the root_domains.
	 *
	 * FIXME: Cycling on all the CPUs is overdoing, but simpler than
	 * cycling on root_domains... Discussion on different/better
	 * solutions is welcome!
	 */
8021
	for_each_possible_cpu(cpu) {
8022 8023
		rcu_read_lock_sched();
		dl_b = dl_bw_of(cpu);
8024

8025
		raw_spin_lock_irqsave(&dl_b->lock, flags);
8026 8027
		if (new_bw < dl_b->total_bw)
			ret = -EBUSY;
8028
		raw_spin_unlock_irqrestore(&dl_b->lock, flags);
8029

8030 8031
		rcu_read_unlock_sched();

8032 8033
		if (ret)
			break;
8034 8035
	}

8036
	return ret;
8037 8038
}

8039
static void sched_dl_do_global(void)
8040
{
8041
	u64 new_bw = -1;
8042
	struct dl_bw *dl_b;
8043
	int cpu;
8044
	unsigned long flags;
8045

8046 8047 8048 8049 8050 8051 8052 8053 8054 8055
	def_dl_bandwidth.dl_period = global_rt_period();
	def_dl_bandwidth.dl_runtime = global_rt_runtime();

	if (global_rt_runtime() != RUNTIME_INF)
		new_bw = to_ratio(global_rt_period(), global_rt_runtime());

	/*
	 * FIXME: As above...
	 */
	for_each_possible_cpu(cpu) {
8056 8057
		rcu_read_lock_sched();
		dl_b = dl_bw_of(cpu);
8058

8059
		raw_spin_lock_irqsave(&dl_b->lock, flags);
8060
		dl_b->bw = new_bw;
8061
		raw_spin_unlock_irqrestore(&dl_b->lock, flags);
8062 8063

		rcu_read_unlock_sched();
8064
	}
8065 8066 8067 8068 8069 8070 8071
}

static int sched_rt_global_validate(void)
{
	if (sysctl_sched_rt_period <= 0)
		return -EINVAL;

8072 8073
	if ((sysctl_sched_rt_runtime != RUNTIME_INF) &&
		(sysctl_sched_rt_runtime > sysctl_sched_rt_period))
8074 8075 8076 8077 8078 8079 8080 8081 8082
		return -EINVAL;

	return 0;
}

static void sched_rt_do_global(void)
{
	def_rt_bandwidth.rt_runtime = global_rt_runtime();
	def_rt_bandwidth.rt_period = ns_to_ktime(global_rt_period());
8083 8084
}

8085
int sched_rt_handler(struct ctl_table *table, int write,
8086
		void __user *buffer, size_t *lenp,
8087 8088 8089 8090
		loff_t *ppos)
{
	int old_period, old_runtime;
	static DEFINE_MUTEX(mutex);
8091
	int ret;
8092 8093 8094 8095 8096

	mutex_lock(&mutex);
	old_period = sysctl_sched_rt_period;
	old_runtime = sysctl_sched_rt_runtime;

8097
	ret = proc_dointvec(table, write, buffer, lenp, ppos);
8098 8099

	if (!ret && write) {
8100 8101 8102 8103
		ret = sched_rt_global_validate();
		if (ret)
			goto undo;

8104
		ret = sched_dl_global_validate();
8105 8106 8107
		if (ret)
			goto undo;

8108
		ret = sched_rt_global_constraints();
8109 8110 8111 8112 8113 8114 8115 8116 8117 8118
		if (ret)
			goto undo;

		sched_rt_do_global();
		sched_dl_do_global();
	}
	if (0) {
undo:
		sysctl_sched_rt_period = old_period;
		sysctl_sched_rt_runtime = old_runtime;
8119 8120 8121 8122 8123
	}
	mutex_unlock(&mutex);

	return ret;
}
8124

8125
int sched_rr_handler(struct ctl_table *table, int write,
8126 8127 8128 8129 8130 8131 8132 8133
		void __user *buffer, size_t *lenp,
		loff_t *ppos)
{
	int ret;
	static DEFINE_MUTEX(mutex);

	mutex_lock(&mutex);
	ret = proc_dointvec(table, write, buffer, lenp, ppos);
8134 8135
	/* make sure that internally we keep jiffies */
	/* also, writing zero resets timeslice to default */
8136
	if (!ret && write) {
8137 8138
		sched_rr_timeslice = sched_rr_timeslice <= 0 ?
			RR_TIMESLICE : msecs_to_jiffies(sched_rr_timeslice);
8139 8140 8141 8142 8143
	}
	mutex_unlock(&mutex);
	return ret;
}

8144
#ifdef CONFIG_CGROUP_SCHED
8145

8146
static inline struct task_group *css_tg(struct cgroup_subsys_state *css)
8147
{
8148
	return css ? container_of(css, struct task_group, css) : NULL;
8149 8150
}

8151 8152
static struct cgroup_subsys_state *
cpu_cgroup_css_alloc(struct cgroup_subsys_state *parent_css)
8153
{
8154 8155
	struct task_group *parent = css_tg(parent_css);
	struct task_group *tg;
8156

8157
	if (!parent) {
8158
		/* This is early initialization for the top cgroup */
8159
		return &root_task_group.css;
8160 8161
	}

8162
	tg = sched_create_group(parent);
8163 8164 8165 8166 8167 8168
	if (IS_ERR(tg))
		return ERR_PTR(-ENOMEM);

	return &tg->css;
}

8169
static int cpu_cgroup_css_online(struct cgroup_subsys_state *css)
8170
{
8171
	struct task_group *tg = css_tg(css);
T
Tejun Heo 已提交
8172
	struct task_group *parent = css_tg(css->parent);
8173

T
Tejun Heo 已提交
8174 8175
	if (parent)
		sched_online_group(tg, parent);
8176 8177 8178
	return 0;
}

8179
static void cpu_cgroup_css_free(struct cgroup_subsys_state *css)
8180
{
8181
	struct task_group *tg = css_tg(css);
8182 8183 8184 8185

	sched_destroy_group(tg);
}

8186
static void cpu_cgroup_css_offline(struct cgroup_subsys_state *css)
8187
{
8188
	struct task_group *tg = css_tg(css);
8189 8190 8191 8192

	sched_offline_group(tg);
}

8193
static void cpu_cgroup_fork(struct task_struct *task, void *private)
8194 8195 8196 8197
{
	sched_move_task(task);
}

8198
static int cpu_cgroup_can_attach(struct cgroup_subsys_state *css,
8199
				 struct cgroup_taskset *tset)
8200
{
8201 8202
	struct task_struct *task;

8203
	cgroup_taskset_for_each(task, tset) {
8204
#ifdef CONFIG_RT_GROUP_SCHED
8205
		if (!sched_rt_can_attach(css_tg(css), task))
8206
			return -EINVAL;
8207
#else
8208 8209 8210
		/* We don't support RT-tasks being in separate groups */
		if (task->sched_class != &fair_sched_class)
			return -EINVAL;
8211
#endif
8212
	}
8213 8214
	return 0;
}
8215

8216
static void cpu_cgroup_attach(struct cgroup_subsys_state *css,
8217
			      struct cgroup_taskset *tset)
8218
{
8219 8220
	struct task_struct *task;

8221
	cgroup_taskset_for_each(task, tset)
8222
		sched_move_task(task);
8223 8224
}

8225 8226 8227
static void cpu_cgroup_exit(struct cgroup_subsys_state *css,
			    struct cgroup_subsys_state *old_css,
			    struct task_struct *task)
8228 8229 8230 8231
{
	sched_move_task(task);
}

8232
#ifdef CONFIG_FAIR_GROUP_SCHED
8233 8234
static int cpu_shares_write_u64(struct cgroup_subsys_state *css,
				struct cftype *cftype, u64 shareval)
8235
{
8236
	return sched_group_set_shares(css_tg(css), scale_load(shareval));
8237 8238
}

8239 8240
static u64 cpu_shares_read_u64(struct cgroup_subsys_state *css,
			       struct cftype *cft)
8241
{
8242
	struct task_group *tg = css_tg(css);
8243

8244
	return (u64) scale_load_down(tg->shares);
8245
}
8246 8247

#ifdef CONFIG_CFS_BANDWIDTH
8248 8249
static DEFINE_MUTEX(cfs_constraints_mutex);

8250 8251 8252
const u64 max_cfs_quota_period = 1 * NSEC_PER_SEC; /* 1s */
const u64 min_cfs_quota_period = 1 * NSEC_PER_MSEC; /* 1ms */

8253 8254
static int __cfs_schedulable(struct task_group *tg, u64 period, u64 runtime);

8255 8256
static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota)
{
8257
	int i, ret = 0, runtime_enabled, runtime_was_enabled;
8258
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
8259 8260 8261 8262 8263 8264 8265 8266 8267 8268 8269 8270 8271 8272 8273 8274 8275 8276 8277 8278

	if (tg == &root_task_group)
		return -EINVAL;

	/*
	 * Ensure we have at some amount of bandwidth every period.  This is
	 * to prevent reaching a state of large arrears when throttled via
	 * entity_tick() resulting in prolonged exit starvation.
	 */
	if (quota < min_cfs_quota_period || period < min_cfs_quota_period)
		return -EINVAL;

	/*
	 * Likewise, bound things on the otherside by preventing insane quota
	 * periods.  This also allows us to normalize in computing quota
	 * feasibility.
	 */
	if (period > max_cfs_quota_period)
		return -EINVAL;

8279 8280 8281 8282 8283
	/*
	 * Prevent race between setting of cfs_rq->runtime_enabled and
	 * unthrottle_offline_cfs_rqs().
	 */
	get_online_cpus();
8284 8285 8286 8287 8288
	mutex_lock(&cfs_constraints_mutex);
	ret = __cfs_schedulable(tg, period, quota);
	if (ret)
		goto out_unlock;

8289
	runtime_enabled = quota != RUNTIME_INF;
8290
	runtime_was_enabled = cfs_b->quota != RUNTIME_INF;
8291 8292 8293 8294 8295 8296
	/*
	 * 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();
8297 8298 8299
	raw_spin_lock_irq(&cfs_b->lock);
	cfs_b->period = ns_to_ktime(period);
	cfs_b->quota = quota;
8300

P
Paul Turner 已提交
8301
	__refill_cfs_bandwidth_runtime(cfs_b);
8302
	/* restart the period timer (if active) to handle new period expiry */
P
Peter Zijlstra 已提交
8303 8304
	if (runtime_enabled)
		start_cfs_bandwidth(cfs_b);
8305 8306
	raw_spin_unlock_irq(&cfs_b->lock);

8307
	for_each_online_cpu(i) {
8308
		struct cfs_rq *cfs_rq = tg->cfs_rq[i];
8309
		struct rq *rq = cfs_rq->rq;
8310 8311

		raw_spin_lock_irq(&rq->lock);
8312
		cfs_rq->runtime_enabled = runtime_enabled;
8313
		cfs_rq->runtime_remaining = 0;
8314

8315
		if (cfs_rq->throttled)
8316
			unthrottle_cfs_rq(cfs_rq);
8317 8318
		raw_spin_unlock_irq(&rq->lock);
	}
8319 8320
	if (runtime_was_enabled && !runtime_enabled)
		cfs_bandwidth_usage_dec();
8321 8322
out_unlock:
	mutex_unlock(&cfs_constraints_mutex);
8323
	put_online_cpus();
8324

8325
	return ret;
8326 8327 8328 8329 8330 8331
}

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

8332
	period = ktime_to_ns(tg->cfs_bandwidth.period);
8333 8334 8335 8336 8337 8338 8339 8340 8341 8342 8343 8344
	if (cfs_quota_us < 0)
		quota = RUNTIME_INF;
	else
		quota = (u64)cfs_quota_us * NSEC_PER_USEC;

	return tg_set_cfs_bandwidth(tg, period, quota);
}

long tg_get_cfs_quota(struct task_group *tg)
{
	u64 quota_us;

8345
	if (tg->cfs_bandwidth.quota == RUNTIME_INF)
8346 8347
		return -1;

8348
	quota_us = tg->cfs_bandwidth.quota;
8349 8350 8351 8352 8353 8354 8355 8356 8357 8358
	do_div(quota_us, NSEC_PER_USEC);

	return quota_us;
}

int tg_set_cfs_period(struct task_group *tg, long cfs_period_us)
{
	u64 quota, period;

	period = (u64)cfs_period_us * NSEC_PER_USEC;
8359
	quota = tg->cfs_bandwidth.quota;
8360 8361 8362 8363 8364 8365 8366 8367

	return tg_set_cfs_bandwidth(tg, period, quota);
}

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

8368
	cfs_period_us = ktime_to_ns(tg->cfs_bandwidth.period);
8369 8370 8371 8372 8373
	do_div(cfs_period_us, NSEC_PER_USEC);

	return cfs_period_us;
}

8374 8375
static s64 cpu_cfs_quota_read_s64(struct cgroup_subsys_state *css,
				  struct cftype *cft)
8376
{
8377
	return tg_get_cfs_quota(css_tg(css));
8378 8379
}

8380 8381
static int cpu_cfs_quota_write_s64(struct cgroup_subsys_state *css,
				   struct cftype *cftype, s64 cfs_quota_us)
8382
{
8383
	return tg_set_cfs_quota(css_tg(css), cfs_quota_us);
8384 8385
}

8386 8387
static u64 cpu_cfs_period_read_u64(struct cgroup_subsys_state *css,
				   struct cftype *cft)
8388
{
8389
	return tg_get_cfs_period(css_tg(css));
8390 8391
}

8392 8393
static int cpu_cfs_period_write_u64(struct cgroup_subsys_state *css,
				    struct cftype *cftype, u64 cfs_period_us)
8394
{
8395
	return tg_set_cfs_period(css_tg(css), cfs_period_us);
8396 8397
}

8398 8399 8400 8401 8402 8403 8404 8405 8406 8407 8408 8409 8410 8411 8412 8413 8414 8415 8416 8417 8418 8419 8420 8421 8422 8423 8424 8425 8426 8427 8428 8429
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;
8430
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
8431 8432 8433 8434 8435
	s64 quota = 0, parent_quota = -1;

	if (!tg->parent) {
		quota = RUNTIME_INF;
	} else {
8436
		struct cfs_bandwidth *parent_b = &tg->parent->cfs_bandwidth;
8437 8438

		quota = normalize_cfs_quota(tg, d);
8439
		parent_quota = parent_b->hierarchical_quota;
8440 8441 8442 8443 8444 8445 8446 8447 8448 8449

		/*
		 * ensure max(child_quota) <= parent_quota, inherit when no
		 * limit is set
		 */
		if (quota == RUNTIME_INF)
			quota = parent_quota;
		else if (parent_quota != RUNTIME_INF && quota > parent_quota)
			return -EINVAL;
	}
8450
	cfs_b->hierarchical_quota = quota;
8451 8452 8453 8454 8455 8456

	return 0;
}

static int __cfs_schedulable(struct task_group *tg, u64 period, u64 quota)
{
8457
	int ret;
8458 8459 8460 8461 8462 8463 8464 8465 8466 8467 8468
	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);
	}

8469 8470 8471 8472 8473
	rcu_read_lock();
	ret = walk_tg_tree(tg_cfs_schedulable_down, tg_nop, &data);
	rcu_read_unlock();

	return ret;
8474
}
8475

8476
static int cpu_stats_show(struct seq_file *sf, void *v)
8477
{
8478
	struct task_group *tg = css_tg(seq_css(sf));
8479
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
8480

8481 8482 8483
	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);
8484 8485 8486

	return 0;
}
8487
#endif /* CONFIG_CFS_BANDWIDTH */
8488
#endif /* CONFIG_FAIR_GROUP_SCHED */
8489

8490
#ifdef CONFIG_RT_GROUP_SCHED
8491 8492
static int cpu_rt_runtime_write(struct cgroup_subsys_state *css,
				struct cftype *cft, s64 val)
P
Peter Zijlstra 已提交
8493
{
8494
	return sched_group_set_rt_runtime(css_tg(css), val);
P
Peter Zijlstra 已提交
8495 8496
}

8497 8498
static s64 cpu_rt_runtime_read(struct cgroup_subsys_state *css,
			       struct cftype *cft)
P
Peter Zijlstra 已提交
8499
{
8500
	return sched_group_rt_runtime(css_tg(css));
P
Peter Zijlstra 已提交
8501
}
8502

8503 8504
static int cpu_rt_period_write_uint(struct cgroup_subsys_state *css,
				    struct cftype *cftype, u64 rt_period_us)
8505
{
8506
	return sched_group_set_rt_period(css_tg(css), rt_period_us);
8507 8508
}

8509 8510
static u64 cpu_rt_period_read_uint(struct cgroup_subsys_state *css,
				   struct cftype *cft)
8511
{
8512
	return sched_group_rt_period(css_tg(css));
8513
}
8514
#endif /* CONFIG_RT_GROUP_SCHED */
P
Peter Zijlstra 已提交
8515

8516
static struct cftype cpu_files[] = {
8517
#ifdef CONFIG_FAIR_GROUP_SCHED
8518 8519
	{
		.name = "shares",
8520 8521
		.read_u64 = cpu_shares_read_u64,
		.write_u64 = cpu_shares_write_u64,
8522
	},
8523
#endif
8524 8525 8526 8527 8528 8529 8530 8531 8532 8533 8534
#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,
	},
8535 8536
	{
		.name = "stat",
8537
		.seq_show = cpu_stats_show,
8538
	},
8539
#endif
8540
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
8541
	{
P
Peter Zijlstra 已提交
8542
		.name = "rt_runtime_us",
8543 8544
		.read_s64 = cpu_rt_runtime_read,
		.write_s64 = cpu_rt_runtime_write,
P
Peter Zijlstra 已提交
8545
	},
8546 8547
	{
		.name = "rt_period_us",
8548 8549
		.read_u64 = cpu_rt_period_read_uint,
		.write_u64 = cpu_rt_period_write_uint,
8550
	},
8551
#endif
8552
	{ }	/* terminate */
8553 8554
};

8555
struct cgroup_subsys cpu_cgrp_subsys = {
8556 8557
	.css_alloc	= cpu_cgroup_css_alloc,
	.css_free	= cpu_cgroup_css_free,
8558 8559
	.css_online	= cpu_cgroup_css_online,
	.css_offline	= cpu_cgroup_css_offline,
8560
	.fork		= cpu_cgroup_fork,
8561 8562
	.can_attach	= cpu_cgroup_can_attach,
	.attach		= cpu_cgroup_attach,
8563
	.exit		= cpu_cgroup_exit,
8564
	.legacy_cftypes	= cpu_files,
8565 8566 8567
	.early_init	= 1,
};

8568
#endif	/* CONFIG_CGROUP_SCHED */
8569

8570 8571 8572 8573 8574
void dump_cpu_task(int cpu)
{
	pr_info("Task dump for CPU %d:\n", cpu);
	sched_show_task(cpu_curr(cpu));
}