core.c 206.3 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>
L
Linus Torvalds 已提交
75

76
#include <asm/switch_to.h>
77
#include <asm/tlb.h>
78
#include <asm/irq_regs.h>
79
#include <asm/mutex.h>
G
Glauber Costa 已提交
80 81 82
#ifdef CONFIG_PARAVIRT
#include <asm/paravirt.h>
#endif
L
Linus Torvalds 已提交
83

84
#include "sched.h"
85
#include "../workqueue_sched.h"
86
#include "../smpboot.h"
87

88
#define CREATE_TRACE_POINTS
89
#include <trace/events/sched.h>
90

91
void start_bandwidth_timer(struct hrtimer *period_timer, ktime_t period)
92
{
93 94
	unsigned long delta;
	ktime_t soft, hard, now;
95

96 97 98 99 100 101
	for (;;) {
		if (hrtimer_active(period_timer))
			break;

		now = hrtimer_cb_get_time(period_timer);
		hrtimer_forward(period_timer, now, period);
102

103 104 105 106 107 108 109 110
		soft = hrtimer_get_softexpires(period_timer);
		hard = hrtimer_get_expires(period_timer);
		delta = ktime_to_ns(ktime_sub(hard, soft));
		__hrtimer_start_range_ns(period_timer, soft, delta,
					 HRTIMER_MODE_ABS_PINNED, 0);
	}
}

111 112
DEFINE_MUTEX(sched_domains_mutex);
DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
113

114
static void update_rq_clock_task(struct rq *rq, s64 delta);
115

116
void update_rq_clock(struct rq *rq)
117
{
118
	s64 delta;
119

120
	if (rq->skip_clock_update > 0)
121
		return;
122

123 124 125
	delta = sched_clock_cpu(cpu_of(rq)) - rq->clock;
	rq->clock += delta;
	update_rq_clock_task(rq, delta);
126 127
}

I
Ingo Molnar 已提交
128 129 130
/*
 * Debugging: various feature bits
 */
P
Peter Zijlstra 已提交
131 132 133 134

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

I
Ingo Molnar 已提交
135
const_debug unsigned int sysctl_sched_features =
136
#include "features.h"
P
Peter Zijlstra 已提交
137 138 139 140 141 142 143 144
	0;

#undef SCHED_FEAT

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

145
static const char * const sched_feat_names[] = {
146
#include "features.h"
P
Peter Zijlstra 已提交
147 148 149 150
};

#undef SCHED_FEAT

L
Li Zefan 已提交
151
static int sched_feat_show(struct seq_file *m, void *v)
P
Peter Zijlstra 已提交
152 153 154
{
	int i;

155
	for (i = 0; i < __SCHED_FEAT_NR; i++) {
L
Li Zefan 已提交
156 157 158
		if (!(sysctl_sched_features & (1UL << i)))
			seq_puts(m, "NO_");
		seq_printf(m, "%s ", sched_feat_names[i]);
P
Peter Zijlstra 已提交
159
	}
L
Li Zefan 已提交
160
	seq_puts(m, "\n");
P
Peter Zijlstra 已提交
161

L
Li Zefan 已提交
162
	return 0;
P
Peter Zijlstra 已提交
163 164
}

165 166
#ifdef HAVE_JUMP_LABEL

167 168
#define jump_label_key__true  STATIC_KEY_INIT_TRUE
#define jump_label_key__false STATIC_KEY_INIT_FALSE
169 170 171 172

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

173
struct static_key sched_feat_keys[__SCHED_FEAT_NR] = {
174 175 176 177 178 179 180
#include "features.h"
};

#undef SCHED_FEAT

static void sched_feat_disable(int i)
{
181 182
	if (static_key_enabled(&sched_feat_keys[i]))
		static_key_slow_dec(&sched_feat_keys[i]);
183 184 185 186
}

static void sched_feat_enable(int i)
{
187 188
	if (!static_key_enabled(&sched_feat_keys[i]))
		static_key_slow_inc(&sched_feat_keys[i]);
189 190 191 192 193 194
}
#else
static void sched_feat_disable(int i) { };
static void sched_feat_enable(int i) { };
#endif /* HAVE_JUMP_LABEL */

P
Peter Zijlstra 已提交
195 196 197 198 199
static ssize_t
sched_feat_write(struct file *filp, const char __user *ubuf,
		size_t cnt, loff_t *ppos)
{
	char buf[64];
200
	char *cmp;
P
Peter Zijlstra 已提交
201 202 203 204 205 206 207 208 209 210
	int neg = 0;
	int i;

	if (cnt > 63)
		cnt = 63;

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

	buf[cnt] = 0;
211
	cmp = strstrip(buf);
P
Peter Zijlstra 已提交
212

H
Hillf Danton 已提交
213
	if (strncmp(cmp, "NO_", 3) == 0) {
P
Peter Zijlstra 已提交
214 215 216 217
		neg = 1;
		cmp += 3;
	}

218
	for (i = 0; i < __SCHED_FEAT_NR; i++) {
219
		if (strcmp(cmp, sched_feat_names[i]) == 0) {
220
			if (neg) {
P
Peter Zijlstra 已提交
221
				sysctl_sched_features &= ~(1UL << i);
222 223
				sched_feat_disable(i);
			} else {
P
Peter Zijlstra 已提交
224
				sysctl_sched_features |= (1UL << i);
225 226
				sched_feat_enable(i);
			}
P
Peter Zijlstra 已提交
227 228 229 230
			break;
		}
	}

231
	if (i == __SCHED_FEAT_NR)
P
Peter Zijlstra 已提交
232 233
		return -EINVAL;

234
	*ppos += cnt;
P
Peter Zijlstra 已提交
235 236 237 238

	return cnt;
}

L
Li Zefan 已提交
239 240 241 242 243
static int sched_feat_open(struct inode *inode, struct file *filp)
{
	return single_open(filp, sched_feat_show, NULL);
}

244
static const struct file_operations sched_feat_fops = {
L
Li Zefan 已提交
245 246 247 248 249
	.open		= sched_feat_open,
	.write		= sched_feat_write,
	.read		= seq_read,
	.llseek		= seq_lseek,
	.release	= single_release,
P
Peter Zijlstra 已提交
250 251 252 253 254 255 256 257 258 259
};

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);
260
#endif /* CONFIG_SCHED_DEBUG */
I
Ingo Molnar 已提交
261

262 263 264 265 266 267
/*
 * 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;

268 269 270 271 272 273 274 275
/*
 * 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
276
/*
P
Peter Zijlstra 已提交
277
 * period over which we measure -rt task cpu usage in us.
P
Peter Zijlstra 已提交
278 279
 * default: 1s
 */
P
Peter Zijlstra 已提交
280
unsigned int sysctl_sched_rt_period = 1000000;
P
Peter Zijlstra 已提交
281

282
__read_mostly int scheduler_running;
283

P
Peter Zijlstra 已提交
284 285 286 287 288
/*
 * 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
289 290


L
Linus Torvalds 已提交
291

292
/*
293
 * __task_rq_lock - lock the rq @p resides on.
294
 */
295
static inline struct rq *__task_rq_lock(struct task_struct *p)
296 297
	__acquires(rq->lock)
{
298 299
	struct rq *rq;

300 301
	lockdep_assert_held(&p->pi_lock);

302
	for (;;) {
303
		rq = task_rq(p);
304
		raw_spin_lock(&rq->lock);
P
Peter Zijlstra 已提交
305
		if (likely(rq == task_rq(p)))
306
			return rq;
307
		raw_spin_unlock(&rq->lock);
308 309 310
	}
}

L
Linus Torvalds 已提交
311
/*
312
 * task_rq_lock - lock p->pi_lock and lock the rq @p resides on.
L
Linus Torvalds 已提交
313
 */
314
static struct rq *task_rq_lock(struct task_struct *p, unsigned long *flags)
315
	__acquires(p->pi_lock)
L
Linus Torvalds 已提交
316 317
	__acquires(rq->lock)
{
318
	struct rq *rq;
L
Linus Torvalds 已提交
319

320
	for (;;) {
321
		raw_spin_lock_irqsave(&p->pi_lock, *flags);
322
		rq = task_rq(p);
323
		raw_spin_lock(&rq->lock);
P
Peter Zijlstra 已提交
324
		if (likely(rq == task_rq(p)))
325
			return rq;
326 327
		raw_spin_unlock(&rq->lock);
		raw_spin_unlock_irqrestore(&p->pi_lock, *flags);
L
Linus Torvalds 已提交
328 329 330
	}
}

A
Alexey Dobriyan 已提交
331
static void __task_rq_unlock(struct rq *rq)
332 333
	__releases(rq->lock)
{
334
	raw_spin_unlock(&rq->lock);
335 336
}

337 338
static inline void
task_rq_unlock(struct rq *rq, struct task_struct *p, unsigned long *flags)
L
Linus Torvalds 已提交
339
	__releases(rq->lock)
340
	__releases(p->pi_lock)
L
Linus Torvalds 已提交
341
{
342 343
	raw_spin_unlock(&rq->lock);
	raw_spin_unlock_irqrestore(&p->pi_lock, *flags);
L
Linus Torvalds 已提交
344 345 346
}

/*
347
 * this_rq_lock - lock this runqueue and disable interrupts.
L
Linus Torvalds 已提交
348
 */
A
Alexey Dobriyan 已提交
349
static struct rq *this_rq_lock(void)
L
Linus Torvalds 已提交
350 351
	__acquires(rq->lock)
{
352
	struct rq *rq;
L
Linus Torvalds 已提交
353 354 355

	local_irq_disable();
	rq = this_rq();
356
	raw_spin_lock(&rq->lock);
L
Linus Torvalds 已提交
357 358 359 360

	return rq;
}

P
Peter Zijlstra 已提交
361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388
#ifdef CONFIG_SCHED_HRTICK
/*
 * Use HR-timers to deliver accurate preemption points.
 *
 * Its all a bit involved since we cannot program an hrt while holding the
 * rq->lock. So what we do is store a state in in rq->hrtick_* and ask for a
 * reschedule event.
 *
 * When we get rescheduled we reprogram the hrtick_timer outside of the
 * rq->lock.
 */

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());

389
	raw_spin_lock(&rq->lock);
390
	update_rq_clock(rq);
P
Peter Zijlstra 已提交
391
	rq->curr->sched_class->task_tick(rq, rq->curr, 1);
392
	raw_spin_unlock(&rq->lock);
P
Peter Zijlstra 已提交
393 394 395 396

	return HRTIMER_NORESTART;
}

397
#ifdef CONFIG_SMP
398 399 400 401
/*
 * called from hardirq (IPI) context
 */
static void __hrtick_start(void *arg)
402
{
403
	struct rq *rq = arg;
404

405
	raw_spin_lock(&rq->lock);
406 407
	hrtimer_restart(&rq->hrtick_timer);
	rq->hrtick_csd_pending = 0;
408
	raw_spin_unlock(&rq->lock);
409 410
}

411 412 413 414 415
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
416
void hrtick_start(struct rq *rq, u64 delay)
417
{
418 419
	struct hrtimer *timer = &rq->hrtick_timer;
	ktime_t time = ktime_add_ns(timer->base->get_time(), delay);
420

421
	hrtimer_set_expires(timer, time);
422 423 424 425

	if (rq == this_rq()) {
		hrtimer_restart(timer);
	} else if (!rq->hrtick_csd_pending) {
426
		__smp_call_function_single(cpu_of(rq), &rq->hrtick_csd, 0);
427 428
		rq->hrtick_csd_pending = 1;
	}
429 430 431 432 433 434 435 436 437 438 439 440 441 442
}

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:
443
		hrtick_clear(cpu_rq(cpu));
444 445 446 447 448 449
		return NOTIFY_OK;
	}

	return NOTIFY_DONE;
}

450
static __init void init_hrtick(void)
451 452 453
{
	hotcpu_notifier(hotplug_hrtick, 0);
}
454 455 456 457 458 459
#else
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
460
void hrtick_start(struct rq *rq, u64 delay)
461
{
462
	__hrtimer_start_range_ns(&rq->hrtick_timer, ns_to_ktime(delay), 0,
463
			HRTIMER_MODE_REL_PINNED, 0);
464
}
465

A
Andrew Morton 已提交
466
static inline void init_hrtick(void)
P
Peter Zijlstra 已提交
467 468
{
}
469
#endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
470

471
static void init_rq_hrtick(struct rq *rq)
P
Peter Zijlstra 已提交
472
{
473 474
#ifdef CONFIG_SMP
	rq->hrtick_csd_pending = 0;
P
Peter Zijlstra 已提交
475

476 477 478 479
	rq->hrtick_csd.flags = 0;
	rq->hrtick_csd.func = __hrtick_start;
	rq->hrtick_csd.info = rq;
#endif
P
Peter Zijlstra 已提交
480

481 482
	hrtimer_init(&rq->hrtick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	rq->hrtick_timer.function = hrtick;
P
Peter Zijlstra 已提交
483
}
A
Andrew Morton 已提交
484
#else	/* CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
485 486 487 488 489 490 491 492
static inline void hrtick_clear(struct rq *rq)
{
}

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

493 494 495
static inline void init_hrtick(void)
{
}
A
Andrew Morton 已提交
496
#endif	/* CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
497

I
Ingo Molnar 已提交
498 499 500 501 502 503 504 505 506 507 508 509 510
/*
 * resched_task - mark a task 'to be rescheduled now'.
 *
 * 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.
 */
#ifdef CONFIG_SMP

#ifndef tsk_is_polling
#define tsk_is_polling(t) test_tsk_thread_flag(t, TIF_POLLING_NRFLAG)
#endif

511
void resched_task(struct task_struct *p)
I
Ingo Molnar 已提交
512 513 514
{
	int cpu;

515
	assert_raw_spin_locked(&task_rq(p)->lock);
I
Ingo Molnar 已提交
516

517
	if (test_tsk_need_resched(p))
I
Ingo Molnar 已提交
518 519
		return;

520
	set_tsk_need_resched(p);
I
Ingo Molnar 已提交
521 522 523 524 525 526 527 528 529 530 531

	cpu = task_cpu(p);
	if (cpu == smp_processor_id())
		return;

	/* NEED_RESCHED must be visible before we test polling */
	smp_mb();
	if (!tsk_is_polling(p))
		smp_send_reschedule(cpu);
}

532
void resched_cpu(int cpu)
I
Ingo Molnar 已提交
533 534 535 536
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

537
	if (!raw_spin_trylock_irqsave(&rq->lock, flags))
I
Ingo Molnar 已提交
538 539
		return;
	resched_task(cpu_curr(cpu));
540
	raw_spin_unlock_irqrestore(&rq->lock, flags);
I
Ingo Molnar 已提交
541
}
542 543

#ifdef CONFIG_NO_HZ
544 545 546 547 548 549 550 551 552 553 554 555 556 557
/*
 * 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).
 */
int get_nohz_timer_target(void)
{
	int cpu = smp_processor_id();
	int i;
	struct sched_domain *sd;

558
	rcu_read_lock();
559
	for_each_domain(cpu, sd) {
560 561 562 563 564 565
		for_each_cpu(i, sched_domain_span(sd)) {
			if (!idle_cpu(i)) {
				cpu = i;
				goto unlock;
			}
		}
566
	}
567 568
unlock:
	rcu_read_unlock();
569 570
	return cpu;
}
571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596
/*
 * 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.
 */
void wake_up_idle_cpu(int cpu)
{
	struct rq *rq = cpu_rq(cpu);

	if (cpu == smp_processor_id())
		return;

	/*
	 * This is safe, as this function is called with the timer
	 * wheel base lock of (cpu) held. When the CPU is on the way
	 * to idle and has not yet set rq->curr to idle then it will
	 * be serialized on the timer wheel base lock and take the new
	 * timer into account automatically.
	 */
	if (rq->curr != rq->idle)
		return;
597 598

	/*
599 600 601
	 * We can set TIF_RESCHED on the idle task of the other CPU
	 * lockless. The worst case is that the other CPU runs the
	 * idle task through an additional NOOP schedule()
602
	 */
603
	set_tsk_need_resched(rq->idle);
604

605 606 607 608
	/* NEED_RESCHED must be visible before we test polling */
	smp_mb();
	if (!tsk_is_polling(rq->idle))
		smp_send_reschedule(cpu);
609 610
}

611
static inline bool got_nohz_idle_kick(void)
612
{
613 614
	int cpu = smp_processor_id();
	return idle_cpu(cpu) && test_bit(NOHZ_BALANCE_KICK, nohz_flags(cpu));
615 616
}

617
#else /* CONFIG_NO_HZ */
618

619
static inline bool got_nohz_idle_kick(void)
P
Peter Zijlstra 已提交
620
{
621
	return false;
P
Peter Zijlstra 已提交
622 623
}

624
#endif /* CONFIG_NO_HZ */
625

626
void sched_avg_update(struct rq *rq)
627
{
628 629 630
	s64 period = sched_avg_period();

	while ((s64)(rq->clock - rq->age_stamp) > period) {
631 632 633 634 635 636
		/*
		 * 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));
637 638 639
		rq->age_stamp += period;
		rq->rt_avg /= 2;
	}
640 641
}

642
#else /* !CONFIG_SMP */
643
void resched_task(struct task_struct *p)
644
{
645
	assert_raw_spin_locked(&task_rq(p)->lock);
646
	set_tsk_need_resched(p);
647
}
648
#endif /* CONFIG_SMP */
649

650 651
#if defined(CONFIG_RT_GROUP_SCHED) || (defined(CONFIG_FAIR_GROUP_SCHED) && \
			(defined(CONFIG_SMP) || defined(CONFIG_CFS_BANDWIDTH)))
652
/*
653 654 655 656
 * 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.
657
 */
658
int walk_tg_tree_from(struct task_group *from,
659
			     tg_visitor down, tg_visitor up, void *data)
660 661
{
	struct task_group *parent, *child;
P
Peter Zijlstra 已提交
662
	int ret;
663

664 665
	parent = from;

666
down:
P
Peter Zijlstra 已提交
667 668
	ret = (*down)(parent, data);
	if (ret)
669
		goto out;
670 671 672 673 674 675 676
	list_for_each_entry_rcu(child, &parent->children, siblings) {
		parent = child;
		goto down;

up:
		continue;
	}
P
Peter Zijlstra 已提交
677
	ret = (*up)(parent, data);
678 679
	if (ret || parent == from)
		goto out;
680 681 682 683 684

	child = parent;
	parent = parent->parent;
	if (parent)
		goto up;
685
out:
P
Peter Zijlstra 已提交
686
	return ret;
687 688
}

689
int tg_nop(struct task_group *tg, void *data)
P
Peter Zijlstra 已提交
690
{
691
	return 0;
P
Peter Zijlstra 已提交
692
}
693 694
#endif

695 696
static void set_load_weight(struct task_struct *p)
{
N
Nikhil Rao 已提交
697 698 699
	int prio = p->static_prio - MAX_RT_PRIO;
	struct load_weight *load = &p->se.load;

I
Ingo Molnar 已提交
700 701 702 703
	/*
	 * SCHED_IDLE tasks get minimal weight:
	 */
	if (p->policy == SCHED_IDLE) {
704
		load->weight = scale_load(WEIGHT_IDLEPRIO);
N
Nikhil Rao 已提交
705
		load->inv_weight = WMULT_IDLEPRIO;
I
Ingo Molnar 已提交
706 707
		return;
	}
708

709
	load->weight = scale_load(prio_to_weight[prio]);
N
Nikhil Rao 已提交
710
	load->inv_weight = prio_to_wmult[prio];
711 712
}

713
static void enqueue_task(struct rq *rq, struct task_struct *p, int flags)
714
{
715
	update_rq_clock(rq);
I
Ingo Molnar 已提交
716
	sched_info_queued(p);
717
	p->sched_class->enqueue_task(rq, p, flags);
718 719
}

720
static void dequeue_task(struct rq *rq, struct task_struct *p, int flags)
721
{
722
	update_rq_clock(rq);
723
	sched_info_dequeued(p);
724
	p->sched_class->dequeue_task(rq, p, flags);
725 726
}

727
void activate_task(struct rq *rq, struct task_struct *p, int flags)
728 729 730 731
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible--;

732
	enqueue_task(rq, p, flags);
733 734
}

735
void deactivate_task(struct rq *rq, struct task_struct *p, int flags)
736 737 738 739
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible++;

740
	dequeue_task(rq, p, flags);
741 742
}

743 744
#ifdef CONFIG_IRQ_TIME_ACCOUNTING

745 746 747 748 749 750 751
/*
 * There are no locks covering percpu hardirq/softirq time.
 * They are only modified in account_system_vtime, on corresponding CPU
 * with interrupts disabled. So, writes are safe.
 * They are read and saved off onto struct rq in update_rq_clock().
 * This may result in other CPU reading this CPU's irq time and can
 * race with irq/account_system_vtime on this CPU. We would either get old
752 753 754
 * or new value with a side effect of accounting a slice of irq time to wrong
 * task when irq is in progress while we read rq->clock. That is a worthy
 * compromise in place of having locks on each irq in account_system_time.
755
 */
756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771
static DEFINE_PER_CPU(u64, cpu_hardirq_time);
static DEFINE_PER_CPU(u64, cpu_softirq_time);

static DEFINE_PER_CPU(u64, irq_start_time);
static int sched_clock_irqtime;

void enable_sched_clock_irqtime(void)
{
	sched_clock_irqtime = 1;
}

void disable_sched_clock_irqtime(void)
{
	sched_clock_irqtime = 0;
}

772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809
#ifndef CONFIG_64BIT
static DEFINE_PER_CPU(seqcount_t, irq_time_seq);

static inline void irq_time_write_begin(void)
{
	__this_cpu_inc(irq_time_seq.sequence);
	smp_wmb();
}

static inline void irq_time_write_end(void)
{
	smp_wmb();
	__this_cpu_inc(irq_time_seq.sequence);
}

static inline u64 irq_time_read(int cpu)
{
	u64 irq_time;
	unsigned seq;

	do {
		seq = read_seqcount_begin(&per_cpu(irq_time_seq, cpu));
		irq_time = per_cpu(cpu_softirq_time, cpu) +
			   per_cpu(cpu_hardirq_time, cpu);
	} while (read_seqcount_retry(&per_cpu(irq_time_seq, cpu), seq));

	return irq_time;
}
#else /* CONFIG_64BIT */
static inline void irq_time_write_begin(void)
{
}

static inline void irq_time_write_end(void)
{
}

static inline u64 irq_time_read(int cpu)
810 811 812
{
	return per_cpu(cpu_softirq_time, cpu) + per_cpu(cpu_hardirq_time, cpu);
}
813
#endif /* CONFIG_64BIT */
814

815 816 817 818
/*
 * Called before incrementing preempt_count on {soft,}irq_enter
 * and before decrementing preempt_count on {soft,}irq_exit.
 */
819 820 821
void account_system_vtime(struct task_struct *curr)
{
	unsigned long flags;
822
	s64 delta;
823 824 825 826 827 828 829 830
	int cpu;

	if (!sched_clock_irqtime)
		return;

	local_irq_save(flags);

	cpu = smp_processor_id();
831 832 833
	delta = sched_clock_cpu(cpu) - __this_cpu_read(irq_start_time);
	__this_cpu_add(irq_start_time, delta);

834
	irq_time_write_begin();
835 836 837 838 839 840 841
	/*
	 * We do not account for softirq time from ksoftirqd here.
	 * We want to continue accounting softirq time to ksoftirqd thread
	 * in that case, so as not to confuse scheduler with a special task
	 * that do not consume any time, but still wants to run.
	 */
	if (hardirq_count())
842
		__this_cpu_add(cpu_hardirq_time, delta);
843
	else if (in_serving_softirq() && curr != this_cpu_ksoftirqd())
844
		__this_cpu_add(cpu_softirq_time, delta);
845

846
	irq_time_write_end();
847 848
	local_irq_restore(flags);
}
I
Ingo Molnar 已提交
849
EXPORT_SYMBOL_GPL(account_system_vtime);
850

G
Glauber Costa 已提交
851 852 853 854
#endif /* CONFIG_IRQ_TIME_ACCOUNTING */

#ifdef CONFIG_PARAVIRT
static inline u64 steal_ticks(u64 steal)
855
{
G
Glauber Costa 已提交
856 857
	if (unlikely(steal > NSEC_PER_SEC))
		return div_u64(steal, TICK_NSEC);
858

G
Glauber Costa 已提交
859 860 861 862
	return __iter_div_u64_rem(steal, TICK_NSEC, &steal);
}
#endif

863
static void update_rq_clock_task(struct rq *rq, s64 delta)
864
{
865 866 867 868 869 870 871 872
/*
 * 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
873
	irq_delta = irq_time_read(cpu_of(rq)) - rq->prev_irq_time;
874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894

	/*
	 * 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;
895 896
#endif
#ifdef CONFIG_PARAVIRT_TIME_ACCOUNTING
897
	if (static_key_false((&paravirt_steal_rq_enabled))) {
898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914
		u64 st;

		steal = paravirt_steal_clock(cpu_of(rq));
		steal -= rq->prev_steal_time_rq;

		if (unlikely(steal > delta))
			steal = delta;

		st = steal_ticks(steal);
		steal = st * TICK_NSEC;

		rq->prev_steal_time_rq += steal;

		delta -= steal;
	}
#endif

915 916
	rq->clock_task += delta;

917 918 919 920
#if defined(CONFIG_IRQ_TIME_ACCOUNTING) || defined(CONFIG_PARAVIRT_TIME_ACCOUNTING)
	if ((irq_delta + steal) && sched_feat(NONTASK_POWER))
		sched_rt_avg_update(rq, irq_delta + steal);
#endif
921 922
}

923
#ifdef CONFIG_IRQ_TIME_ACCOUNTING
924 925
static int irqtime_account_hi_update(void)
{
926
	u64 *cpustat = kcpustat_this_cpu->cpustat;
927 928 929 930 931 932
	unsigned long flags;
	u64 latest_ns;
	int ret = 0;

	local_irq_save(flags);
	latest_ns = this_cpu_read(cpu_hardirq_time);
933
	if (nsecs_to_cputime64(latest_ns) > cpustat[CPUTIME_IRQ])
934 935 936 937 938 939 940
		ret = 1;
	local_irq_restore(flags);
	return ret;
}

static int irqtime_account_si_update(void)
{
941
	u64 *cpustat = kcpustat_this_cpu->cpustat;
942 943 944 945 946 947
	unsigned long flags;
	u64 latest_ns;
	int ret = 0;

	local_irq_save(flags);
	latest_ns = this_cpu_read(cpu_softirq_time);
948
	if (nsecs_to_cputime64(latest_ns) > cpustat[CPUTIME_SOFTIRQ])
949 950 951 952 953
		ret = 1;
	local_irq_restore(flags);
	return ret;
}

954
#else /* CONFIG_IRQ_TIME_ACCOUNTING */
955

956 957
#define sched_clock_irqtime	(0)

958
#endif
959

960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989
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;
	}
}

990
/*
I
Ingo Molnar 已提交
991
 * __normal_prio - return the priority that is based on the static prio
992 993 994
 */
static inline int __normal_prio(struct task_struct *p)
{
I
Ingo Molnar 已提交
995
	return p->static_prio;
996 997
}

998 999 1000 1001 1002 1003 1004
/*
 * 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.
 */
1005
static inline int normal_prio(struct task_struct *p)
1006 1007 1008
{
	int prio;

1009
	if (task_has_rt_policy(p))
1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022
		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.
 */
1023
static int effective_prio(struct task_struct *p)
1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035
{
	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
1036 1037 1038 1039
/**
 * task_curr - is this task currently executing on a CPU?
 * @p: the task in question.
 */
1040
inline int task_curr(const struct task_struct *p)
L
Linus Torvalds 已提交
1041 1042 1043 1044
{
	return cpu_curr(task_cpu(p)) == p;
}

1045 1046
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
1047
				       int oldprio)
1048 1049 1050
{
	if (prev_class != p->sched_class) {
		if (prev_class->switched_from)
P
Peter Zijlstra 已提交
1051 1052 1053 1054
			prev_class->switched_from(rq, p);
		p->sched_class->switched_to(rq, p);
	} else if (oldprio != p->prio)
		p->sched_class->prio_changed(rq, p, oldprio);
1055 1056
}

1057
void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags)
1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077
{
	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) {
				resched_task(rq->curr);
				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.
	 */
P
Peter Zijlstra 已提交
1078
	if (rq->curr->on_rq && test_tsk_need_resched(rq->curr))
1079 1080 1081
		rq->skip_clock_update = 1;
}

L
Linus Torvalds 已提交
1082
#ifdef CONFIG_SMP
I
Ingo Molnar 已提交
1083
void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
I
Ingo Molnar 已提交
1084
{
1085 1086 1087 1088 1089
#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
1090 1091
	WARN_ON_ONCE(p->state != TASK_RUNNING && p->state != TASK_WAKING &&
			!(task_thread_info(p)->preempt_count & PREEMPT_ACTIVE));
1092 1093

#ifdef CONFIG_LOCKDEP
1094 1095 1096 1097 1098 1099 1100 1101 1102 1103
	/*
	 * 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,
	 * see set_task_rq().
	 *
	 * Furthermore, all task_rq users should acquire both locks, see
	 * task_rq_lock().
	 */
1104 1105 1106
	WARN_ON_ONCE(debug_locks && !(lockdep_is_held(&p->pi_lock) ||
				      lockdep_is_held(&task_rq(p)->lock)));
#endif
1107 1108
#endif

1109
	trace_sched_migrate_task(p, new_cpu);
1110

1111 1112
	if (task_cpu(p) != new_cpu) {
		p->se.nr_migrations++;
1113
		perf_sw_event(PERF_COUNT_SW_CPU_MIGRATIONS, 1, NULL, 0);
1114
	}
I
Ingo Molnar 已提交
1115 1116

	__set_task_cpu(p, new_cpu);
I
Ingo Molnar 已提交
1117 1118
}

1119
struct migration_arg {
1120
	struct task_struct *task;
L
Linus Torvalds 已提交
1121
	int dest_cpu;
1122
};
L
Linus Torvalds 已提交
1123

1124 1125
static int migration_cpu_stop(void *data);

L
Linus Torvalds 已提交
1126 1127 1128
/*
 * wait_task_inactive - wait for a thread to unschedule.
 *
R
Roland McGrath 已提交
1129 1130 1131 1132 1133 1134 1135
 * 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
1136 1137 1138 1139 1140 1141
 * 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
1142
unsigned long wait_task_inactive(struct task_struct *p, long match_state)
L
Linus Torvalds 已提交
1143 1144
{
	unsigned long flags;
I
Ingo Molnar 已提交
1145
	int running, on_rq;
R
Roland McGrath 已提交
1146
	unsigned long ncsw;
1147
	struct rq *rq;
L
Linus Torvalds 已提交
1148

1149 1150 1151 1152 1153 1154 1155 1156
	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);
1157

1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168
		/*
		 * 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
1169 1170 1171
		while (task_running(rq, p)) {
			if (match_state && unlikely(p->state != match_state))
				return 0;
1172
			cpu_relax();
R
Roland McGrath 已提交
1173
		}
1174

1175 1176 1177 1178 1179 1180
		/*
		 * 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);
1181
		trace_sched_wait_task(p);
1182
		running = task_running(rq, p);
P
Peter Zijlstra 已提交
1183
		on_rq = p->on_rq;
R
Roland McGrath 已提交
1184
		ncsw = 0;
1185
		if (!match_state || p->state == match_state)
1186
			ncsw = p->nvcsw | LONG_MIN; /* sets MSB */
1187
		task_rq_unlock(rq, p, &flags);
1188

R
Roland McGrath 已提交
1189 1190 1191 1192 1193 1194
		/*
		 * If it changed from the expected state, bail out now.
		 */
		if (unlikely(!ncsw))
			break;

1195 1196 1197 1198 1199 1200 1201 1202 1203 1204
		/*
		 * 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;
		}
1205

1206 1207 1208 1209 1210
		/*
		 * It's not enough that it's not actively running,
		 * it must be off the runqueue _entirely_, and not
		 * preempted!
		 *
1211
		 * So if it was still runnable (but just not actively
1212 1213 1214 1215
		 * running right now), it's preempted, and we should
		 * yield - it could be a while.
		 */
		if (unlikely(on_rq)) {
1216 1217 1218 1219
			ktime_t to = ktime_set(0, NSEC_PER_SEC/HZ);

			set_current_state(TASK_UNINTERRUPTIBLE);
			schedule_hrtimeout(&to, HRTIMER_MODE_REL);
1220 1221
			continue;
		}
1222

1223 1224 1225 1226 1227 1228 1229
		/*
		 * 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
1230 1231

	return ncsw;
L
Linus Torvalds 已提交
1232 1233 1234 1235 1236 1237 1238 1239 1240
}

/***
 * 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
1241
 * NOTE: this function doesn't have to take the runqueue lock,
L
Linus Torvalds 已提交
1242 1243 1244 1245 1246
 * 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.
 */
1247
void kick_process(struct task_struct *p)
L
Linus Torvalds 已提交
1248 1249 1250 1251 1252 1253 1254 1255 1256
{
	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
1257
EXPORT_SYMBOL_GPL(kick_process);
N
Nick Piggin 已提交
1258
#endif /* CONFIG_SMP */
L
Linus Torvalds 已提交
1259

1260
#ifdef CONFIG_SMP
1261
/*
1262
 * ->cpus_allowed is protected by both rq->lock and p->pi_lock
1263
 */
1264 1265 1266
static int select_fallback_rq(int cpu, struct task_struct *p)
{
	const struct cpumask *nodemask = cpumask_of_node(cpu_to_node(cpu));
1267 1268
	enum { cpuset, possible, fail } state = cpuset;
	int dest_cpu;
1269 1270

	/* Look for allowed, online CPU in same node. */
1271
	for_each_cpu(dest_cpu, nodemask) {
1272 1273 1274 1275
		if (!cpu_online(dest_cpu))
			continue;
		if (!cpu_active(dest_cpu))
			continue;
1276
		if (cpumask_test_cpu(dest_cpu, tsk_cpus_allowed(p)))
1277
			return dest_cpu;
1278
	}
1279

1280 1281
	for (;;) {
		/* Any allowed, online CPU? */
1282
		for_each_cpu(dest_cpu, tsk_cpus_allowed(p)) {
1283 1284 1285 1286 1287 1288
			if (!cpu_online(dest_cpu))
				continue;
			if (!cpu_active(dest_cpu))
				continue;
			goto out;
		}
1289

1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318
		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()) {
			printk_sched("process %d (%s) no longer affine to cpu%d\n",
					task_pid_nr(p), p->comm, cpu);
		}
1319 1320 1321 1322 1323
	}

	return dest_cpu;
}

1324
/*
1325
 * The caller (fork, wakeup) owns p->pi_lock, ->cpus_allowed is stable.
1326
 */
1327
static inline
1328
int select_task_rq(struct task_struct *p, int sd_flags, int wake_flags)
1329
{
1330
	int cpu = p->sched_class->select_task_rq(p, sd_flags, wake_flags);
1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341

	/*
	 * 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 ]
	 */
1342
	if (unlikely(!cpumask_test_cpu(cpu, tsk_cpus_allowed(p)) ||
P
Peter Zijlstra 已提交
1343
		     !cpu_online(cpu)))
1344
		cpu = select_fallback_rq(task_cpu(p), p);
1345 1346

	return cpu;
1347
}
1348 1349 1350 1351 1352 1353

static void update_avg(u64 *avg, u64 sample)
{
	s64 diff = sample - *avg;
	*avg += diff >> 3;
}
1354 1355
#endif

P
Peter Zijlstra 已提交
1356
static void
1357
ttwu_stat(struct task_struct *p, int cpu, int wake_flags)
T
Tejun Heo 已提交
1358
{
P
Peter Zijlstra 已提交
1359
#ifdef CONFIG_SCHEDSTATS
1360 1361
	struct rq *rq = this_rq();

P
Peter Zijlstra 已提交
1362 1363 1364 1365 1366 1367 1368 1369 1370 1371
#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);
1372
		rcu_read_lock();
P
Peter Zijlstra 已提交
1373 1374 1375 1376 1377 1378
		for_each_domain(this_cpu, sd) {
			if (cpumask_test_cpu(cpu, sched_domain_span(sd))) {
				schedstat_inc(sd, ttwu_wake_remote);
				break;
			}
		}
1379
		rcu_read_unlock();
P
Peter Zijlstra 已提交
1380
	}
1381 1382 1383 1384

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

P
Peter Zijlstra 已提交
1385 1386 1387
#endif /* CONFIG_SMP */

	schedstat_inc(rq, ttwu_count);
T
Tejun Heo 已提交
1388
	schedstat_inc(p, se.statistics.nr_wakeups);
P
Peter Zijlstra 已提交
1389 1390

	if (wake_flags & WF_SYNC)
T
Tejun Heo 已提交
1391
		schedstat_inc(p, se.statistics.nr_wakeups_sync);
P
Peter Zijlstra 已提交
1392 1393 1394 1395 1396 1397

#endif /* CONFIG_SCHEDSTATS */
}

static void ttwu_activate(struct rq *rq, struct task_struct *p, int en_flags)
{
T
Tejun Heo 已提交
1398
	activate_task(rq, p, en_flags);
P
Peter Zijlstra 已提交
1399
	p->on_rq = 1;
1400 1401 1402 1403

	/* 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
1404 1405
}

1406 1407 1408
/*
 * Mark the task runnable and perform wakeup-preemption.
 */
1409
static void
1410
ttwu_do_wakeup(struct rq *rq, struct task_struct *p, int wake_flags)
T
Tejun Heo 已提交
1411
{
1412
	trace_sched_wakeup(p, true);
T
Tejun Heo 已提交
1413 1414 1415 1416 1417 1418 1419
	check_preempt_curr(rq, p, wake_flags);

	p->state = TASK_RUNNING;
#ifdef CONFIG_SMP
	if (p->sched_class->task_woken)
		p->sched_class->task_woken(rq, p);

1420
	if (rq->idle_stamp) {
T
Tejun Heo 已提交
1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432
		u64 delta = rq->clock - rq->idle_stamp;
		u64 max = 2*sysctl_sched_migration_cost;

		if (delta > max)
			rq->avg_idle = max;
		else
			update_avg(&rq->avg_idle, delta);
		rq->idle_stamp = 0;
	}
#endif
}

1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465
static void
ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags)
{
#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);
	if (p->on_rq) {
		ttwu_do_wakeup(rq, p, wake_flags);
		ret = 1;
	}
	__task_rq_unlock(rq);

	return ret;
}

1466
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
1467
static void sched_ttwu_pending(void)
1468 1469
{
	struct rq *rq = this_rq();
P
Peter Zijlstra 已提交
1470 1471
	struct llist_node *llist = llist_del_all(&rq->wake_list);
	struct task_struct *p;
1472 1473 1474

	raw_spin_lock(&rq->lock);

P
Peter Zijlstra 已提交
1475 1476 1477
	while (llist) {
		p = llist_entry(llist, struct task_struct, wake_entry);
		llist = llist_next(llist);
1478 1479 1480 1481 1482 1483 1484 1485
		ttwu_do_activate(rq, p, 0);
	}

	raw_spin_unlock(&rq->lock);
}

void scheduler_ipi(void)
{
1486
	if (llist_empty(&this_rq()->wake_list) && !got_nohz_idle_kick())
1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502
		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
1503
	sched_ttwu_pending();
1504 1505 1506 1507

	/*
	 * Check if someone kicked us for doing the nohz idle load balance.
	 */
1508 1509
	if (unlikely(got_nohz_idle_kick() && !need_resched())) {
		this_rq()->idle_balance = 1;
1510
		raise_softirq_irqoff(SCHED_SOFTIRQ);
1511
	}
1512
	irq_exit();
1513 1514 1515 1516
}

static void ttwu_queue_remote(struct task_struct *p, int cpu)
{
P
Peter Zijlstra 已提交
1517
	if (llist_add(&p->wake_entry, &cpu_rq(cpu)->wake_list))
1518 1519
		smp_send_reschedule(cpu);
}
1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538

#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
static int ttwu_activate_remote(struct task_struct *p, int wake_flags)
{
	struct rq *rq;
	int ret = 0;

	rq = __task_rq_lock(p);
	if (p->on_cpu) {
		ttwu_activate(rq, p, ENQUEUE_WAKEUP);
		ttwu_do_wakeup(rq, p, wake_flags);
		ret = 1;
	}
	__task_rq_unlock(rq);

	return ret;

}
#endif /* __ARCH_WANT_INTERRUPTS_ON_CTXSW */
1539

1540
bool cpus_share_cache(int this_cpu, int that_cpu)
1541 1542 1543
{
	return per_cpu(sd_llc_id, this_cpu) == per_cpu(sd_llc_id, that_cpu);
}
1544
#endif /* CONFIG_SMP */
1545

1546 1547 1548 1549
static void ttwu_queue(struct task_struct *p, int cpu)
{
	struct rq *rq = cpu_rq(cpu);

1550
#if defined(CONFIG_SMP)
1551
	if (sched_feat(TTWU_QUEUE) && !cpus_share_cache(smp_processor_id(), cpu)) {
1552
		sched_clock_cpu(cpu); /* sync clocks x-cpu */
1553 1554 1555 1556 1557
		ttwu_queue_remote(p, cpu);
		return;
	}
#endif

1558 1559 1560
	raw_spin_lock(&rq->lock);
	ttwu_do_activate(rq, p, 0);
	raw_spin_unlock(&rq->lock);
T
Tejun Heo 已提交
1561 1562 1563
}

/**
L
Linus Torvalds 已提交
1564
 * try_to_wake_up - wake up a thread
T
Tejun Heo 已提交
1565
 * @p: the thread to be awakened
L
Linus Torvalds 已提交
1566
 * @state: the mask of task states that can be woken
T
Tejun Heo 已提交
1567
 * @wake_flags: wake modifier flags (WF_*)
L
Linus Torvalds 已提交
1568 1569 1570 1571 1572 1573 1574
 *
 * 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.
 *
T
Tejun Heo 已提交
1575 1576
 * Returns %true if @p was woken up, %false if it was already running
 * or @state didn't match @p's state.
L
Linus Torvalds 已提交
1577
 */
1578 1579
static int
try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
L
Linus Torvalds 已提交
1580 1581
{
	unsigned long flags;
1582
	int cpu, success = 0;
P
Peter Zijlstra 已提交
1583

1584
	smp_wmb();
1585
	raw_spin_lock_irqsave(&p->pi_lock, flags);
P
Peter Zijlstra 已提交
1586
	if (!(p->state & state))
L
Linus Torvalds 已提交
1587 1588
		goto out;

1589
	success = 1; /* we're going to change ->state */
L
Linus Torvalds 已提交
1590 1591
	cpu = task_cpu(p);

1592 1593
	if (p->on_rq && ttwu_remote(p, wake_flags))
		goto stat;
L
Linus Torvalds 已提交
1594 1595

#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
1596
	/*
1597 1598
	 * 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
1599
	 */
1600 1601 1602
	while (p->on_cpu) {
#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
		/*
1603 1604 1605 1606 1607
		 * In case the architecture enables interrupts in
		 * context_switch(), we cannot busy wait, since that
		 * would lead to deadlocks when an interrupt hits and
		 * tries to wake up @prev. So bail and do a complete
		 * remote wakeup.
1608
		 */
1609
		if (ttwu_activate_remote(p, wake_flags))
1610
			goto stat;
1611
#else
1612
		cpu_relax();
1613
#endif
1614
	}
1615
	/*
1616
	 * Pairs with the smp_wmb() in finish_lock_switch().
1617
	 */
1618
	smp_rmb();
L
Linus Torvalds 已提交
1619

1620
	p->sched_contributes_to_load = !!task_contributes_to_load(p);
P
Peter Zijlstra 已提交
1621
	p->state = TASK_WAKING;
1622

1623
	if (p->sched_class->task_waking)
1624
		p->sched_class->task_waking(p);
1625

1626
	cpu = select_task_rq(p, SD_BALANCE_WAKE, wake_flags);
1627 1628
	if (task_cpu(p) != cpu) {
		wake_flags |= WF_MIGRATED;
1629
		set_task_cpu(p, cpu);
1630
	}
L
Linus Torvalds 已提交
1631 1632
#endif /* CONFIG_SMP */

1633 1634
	ttwu_queue(p, cpu);
stat:
1635
	ttwu_stat(p, cpu, wake_flags);
L
Linus Torvalds 已提交
1636
out:
1637
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
1638 1639 1640 1641

	return success;
}

T
Tejun Heo 已提交
1642 1643 1644 1645
/**
 * try_to_wake_up_local - try to wake up a local task with rq lock held
 * @p: the thread to be awakened
 *
1646
 * Put @p on the run-queue if it's not already there. The caller must
T
Tejun Heo 已提交
1647
 * ensure that this_rq() is locked, @p is bound to this_rq() and not
1648
 * the current task.
T
Tejun Heo 已提交
1649 1650 1651 1652 1653 1654 1655 1656 1657
 */
static void try_to_wake_up_local(struct task_struct *p)
{
	struct rq *rq = task_rq(p);

	BUG_ON(rq != this_rq());
	BUG_ON(p == current);
	lockdep_assert_held(&rq->lock);

1658 1659 1660 1661 1662 1663
	if (!raw_spin_trylock(&p->pi_lock)) {
		raw_spin_unlock(&rq->lock);
		raw_spin_lock(&p->pi_lock);
		raw_spin_lock(&rq->lock);
	}

T
Tejun Heo 已提交
1664
	if (!(p->state & TASK_NORMAL))
1665
		goto out;
T
Tejun Heo 已提交
1666

P
Peter Zijlstra 已提交
1667
	if (!p->on_rq)
P
Peter Zijlstra 已提交
1668 1669
		ttwu_activate(rq, p, ENQUEUE_WAKEUP);

1670
	ttwu_do_wakeup(rq, p, 0);
1671
	ttwu_stat(p, smp_processor_id(), 0);
1672 1673
out:
	raw_spin_unlock(&p->pi_lock);
T
Tejun Heo 已提交
1674 1675
}

1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686
/**
 * 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
 * processes.  Returns 1 if the process was woken up, 0 if it was already
 * running.
 *
 * 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.
 */
1687
int wake_up_process(struct task_struct *p)
L
Linus Torvalds 已提交
1688
{
1689
	return try_to_wake_up(p, TASK_ALL, 0);
L
Linus Torvalds 已提交
1690 1691 1692
}
EXPORT_SYMBOL(wake_up_process);

1693
int wake_up_state(struct task_struct *p, unsigned int state)
L
Linus Torvalds 已提交
1694 1695 1696 1697 1698 1699 1700
{
	return try_to_wake_up(p, state, 0);
}

/*
 * Perform scheduler related setup for a newly forked process p.
 * p is forked by current.
I
Ingo Molnar 已提交
1701 1702 1703 1704 1705
 *
 * __sched_fork() is basic setup used by init_idle() too:
 */
static void __sched_fork(struct task_struct *p)
{
P
Peter Zijlstra 已提交
1706 1707 1708
	p->on_rq			= 0;

	p->se.on_rq			= 0;
I
Ingo Molnar 已提交
1709 1710
	p->se.exec_start		= 0;
	p->se.sum_exec_runtime		= 0;
1711
	p->se.prev_sum_exec_runtime	= 0;
1712
	p->se.nr_migrations		= 0;
P
Peter Zijlstra 已提交
1713
	p->se.vruntime			= 0;
P
Peter Zijlstra 已提交
1714
	INIT_LIST_HEAD(&p->se.group_node);
I
Ingo Molnar 已提交
1715 1716

#ifdef CONFIG_SCHEDSTATS
1717
	memset(&p->se.statistics, 0, sizeof(p->se.statistics));
I
Ingo Molnar 已提交
1718
#endif
N
Nick Piggin 已提交
1719

P
Peter Zijlstra 已提交
1720
	INIT_LIST_HEAD(&p->rt.run_list);
N
Nick Piggin 已提交
1721

1722 1723 1724
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&p->preempt_notifiers);
#endif
I
Ingo Molnar 已提交
1725 1726 1727 1728 1729
}

/*
 * fork()/clone()-time setup:
 */
1730
void sched_fork(struct task_struct *p)
I
Ingo Molnar 已提交
1731
{
1732
	unsigned long flags;
I
Ingo Molnar 已提交
1733 1734 1735
	int cpu = get_cpu();

	__sched_fork(p);
1736
	/*
1737
	 * We mark the process as running here. This guarantees that
1738 1739 1740
	 * nobody will actually run it, and a signal or other external
	 * event cannot wake it up and insert it on the runqueue either.
	 */
1741
	p->state = TASK_RUNNING;
I
Ingo Molnar 已提交
1742

1743 1744 1745 1746 1747
	/*
	 * Make sure we do not leak PI boosting priority to the child.
	 */
	p->prio = current->normal_prio;

1748 1749 1750 1751
	/*
	 * Revert to default priority/policy on fork if requested.
	 */
	if (unlikely(p->sched_reset_on_fork)) {
1752
		if (task_has_rt_policy(p)) {
1753
			p->policy = SCHED_NORMAL;
1754
			p->static_prio = NICE_TO_PRIO(0);
1755 1756 1757 1758 1759 1760
			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);
1761

1762 1763 1764 1765 1766 1767
		/*
		 * We don't need the reset flag anymore after the fork. It has
		 * fulfilled its duty:
		 */
		p->sched_reset_on_fork = 0;
	}
1768

H
Hiroshi Shimamoto 已提交
1769 1770
	if (!rt_prio(p->prio))
		p->sched_class = &fair_sched_class;
1771

P
Peter Zijlstra 已提交
1772 1773 1774
	if (p->sched_class->task_fork)
		p->sched_class->task_fork(p);

1775 1776 1777 1778 1779 1780 1781
	/*
	 * 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.
	 */
1782
	raw_spin_lock_irqsave(&p->pi_lock, flags);
1783
	set_task_cpu(p, cpu);
1784
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
1785

1786
#if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT)
I
Ingo Molnar 已提交
1787
	if (likely(sched_info_on()))
1788
		memset(&p->sched_info, 0, sizeof(p->sched_info));
L
Linus Torvalds 已提交
1789
#endif
P
Peter Zijlstra 已提交
1790 1791
#if defined(CONFIG_SMP)
	p->on_cpu = 0;
1792
#endif
1793
#ifdef CONFIG_PREEMPT_COUNT
1794
	/* Want to start with kernel preemption disabled. */
A
Al Viro 已提交
1795
	task_thread_info(p)->preempt_count = 1;
L
Linus Torvalds 已提交
1796
#endif
1797
#ifdef CONFIG_SMP
1798
	plist_node_init(&p->pushable_tasks, MAX_PRIO);
1799
#endif
1800

N
Nick Piggin 已提交
1801
	put_cpu();
L
Linus Torvalds 已提交
1802 1803 1804 1805 1806 1807 1808 1809 1810
}

/*
 * 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.
 */
1811
void wake_up_new_task(struct task_struct *p)
L
Linus Torvalds 已提交
1812 1813
{
	unsigned long flags;
I
Ingo Molnar 已提交
1814
	struct rq *rq;
1815

1816
	raw_spin_lock_irqsave(&p->pi_lock, flags);
1817 1818 1819 1820 1821 1822
#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
	 */
1823
	set_task_cpu(p, select_task_rq(p, SD_BALANCE_FORK, 0));
1824 1825
#endif

1826
	rq = __task_rq_lock(p);
P
Peter Zijlstra 已提交
1827
	activate_task(rq, p, 0);
P
Peter Zijlstra 已提交
1828
	p->on_rq = 1;
1829
	trace_sched_wakeup_new(p, true);
P
Peter Zijlstra 已提交
1830
	check_preempt_curr(rq, p, WF_FORK);
1831
#ifdef CONFIG_SMP
1832 1833
	if (p->sched_class->task_woken)
		p->sched_class->task_woken(rq, p);
1834
#endif
1835
	task_rq_unlock(rq, p, &flags);
L
Linus Torvalds 已提交
1836 1837
}

1838 1839 1840
#ifdef CONFIG_PREEMPT_NOTIFIERS

/**
1841
 * preempt_notifier_register - tell me when current is being preempted & rescheduled
R
Randy Dunlap 已提交
1842
 * @notifier: notifier struct to register
1843 1844 1845 1846 1847 1848 1849 1850 1851
 */
void preempt_notifier_register(struct preempt_notifier *notifier)
{
	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
1852
 * @notifier: notifier struct to unregister
1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881
 *
 * This is safe to call from within a preemption notifier.
 */
void preempt_notifier_unregister(struct preempt_notifier *notifier)
{
	hlist_del(&notifier->link);
}
EXPORT_SYMBOL_GPL(preempt_notifier_unregister);

static void fire_sched_in_preempt_notifiers(struct task_struct *curr)
{
	struct preempt_notifier *notifier;
	struct hlist_node *node;

	hlist_for_each_entry(notifier, node, &curr->preempt_notifiers, link)
		notifier->ops->sched_in(notifier, raw_smp_processor_id());
}

static void
fire_sched_out_preempt_notifiers(struct task_struct *curr,
				 struct task_struct *next)
{
	struct preempt_notifier *notifier;
	struct hlist_node *node;

	hlist_for_each_entry(notifier, node, &curr->preempt_notifiers, link)
		notifier->ops->sched_out(notifier, next);
}

1882
#else /* !CONFIG_PREEMPT_NOTIFIERS */
1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893

static void fire_sched_in_preempt_notifiers(struct task_struct *curr)
{
}

static void
fire_sched_out_preempt_notifiers(struct task_struct *curr,
				 struct task_struct *next)
{
}

1894
#endif /* CONFIG_PREEMPT_NOTIFIERS */
1895

1896 1897 1898
/**
 * prepare_task_switch - prepare to switch tasks
 * @rq: the runqueue preparing to switch
R
Randy Dunlap 已提交
1899
 * @prev: the current task that is being switched out
1900 1901 1902 1903 1904 1905 1906 1907 1908
 * @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.
 */
1909 1910 1911
static inline void
prepare_task_switch(struct rq *rq, struct task_struct *prev,
		    struct task_struct *next)
1912
{
1913 1914
	sched_info_switch(prev, next);
	perf_event_task_sched_out(prev, next);
1915
	fire_sched_out_preempt_notifiers(prev, next);
1916 1917
	prepare_lock_switch(rq, next);
	prepare_arch_switch(next);
1918
	trace_sched_switch(prev, next);
1919 1920
}

L
Linus Torvalds 已提交
1921 1922
/**
 * finish_task_switch - clean up after a task-switch
1923
 * @rq: runqueue associated with task-switch
L
Linus Torvalds 已提交
1924 1925
 * @prev: the thread we just switched away from.
 *
1926 1927 1928 1929
 * 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
1930 1931
 *
 * Note that we may have delayed dropping an mm in context_switch(). If
I
Ingo Molnar 已提交
1932
 * so, we finish that here outside of the runqueue lock. (Doing it
L
Linus Torvalds 已提交
1933 1934 1935
 * with the lock held can cause deadlocks; see schedule() for
 * details.)
 */
A
Alexey Dobriyan 已提交
1936
static void finish_task_switch(struct rq *rq, struct task_struct *prev)
L
Linus Torvalds 已提交
1937 1938 1939
	__releases(rq->lock)
{
	struct mm_struct *mm = rq->prev_mm;
O
Oleg Nesterov 已提交
1940
	long prev_state;
L
Linus Torvalds 已提交
1941 1942 1943 1944 1945

	rq->prev_mm = NULL;

	/*
	 * A task struct has one reference for the use as "current".
1946
	 * If a task dies, then it sets TASK_DEAD in tsk->state and calls
O
Oleg Nesterov 已提交
1947 1948
	 * schedule one last time. The schedule call will never return, and
	 * the scheduled task must drop that reference.
1949
	 * The test for TASK_DEAD must occur while the runqueue locks are
L
Linus Torvalds 已提交
1950 1951 1952 1953 1954
	 * still held, otherwise prev could be scheduled on another cpu, die
	 * there before we look at prev->state, and then the reference would
	 * be dropped twice.
	 *		Manfred Spraul <manfred@colorfullife.com>
	 */
O
Oleg Nesterov 已提交
1955
	prev_state = prev->state;
1956
	finish_arch_switch(prev);
1957 1958 1959
#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
	local_irq_disable();
#endif /* __ARCH_WANT_INTERRUPTS_ON_CTXSW */
1960
	perf_event_task_sched_in(prev, current);
1961 1962 1963
#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
	local_irq_enable();
#endif /* __ARCH_WANT_INTERRUPTS_ON_CTXSW */
1964
	finish_lock_switch(rq, prev);
1965
	finish_arch_post_lock_switch();
S
Steven Rostedt 已提交
1966

1967
	fire_sched_in_preempt_notifiers(current);
L
Linus Torvalds 已提交
1968 1969
	if (mm)
		mmdrop(mm);
1970
	if (unlikely(prev_state == TASK_DEAD)) {
1971 1972 1973
		/*
		 * Remove function-return probe instances associated with this
		 * task and put them back on the free list.
I
Ingo Molnar 已提交
1974
		 */
1975
		kprobe_flush_task(prev);
L
Linus Torvalds 已提交
1976
		put_task_struct(prev);
1977
	}
L
Linus Torvalds 已提交
1978 1979
}

1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994
#ifdef CONFIG_SMP

/* assumes rq->lock is held */
static inline void pre_schedule(struct rq *rq, struct task_struct *prev)
{
	if (prev->sched_class->pre_schedule)
		prev->sched_class->pre_schedule(rq, prev);
}

/* rq->lock is NOT held, but preemption is disabled */
static inline void post_schedule(struct rq *rq)
{
	if (rq->post_schedule) {
		unsigned long flags;

1995
		raw_spin_lock_irqsave(&rq->lock, flags);
1996 1997
		if (rq->curr->sched_class->post_schedule)
			rq->curr->sched_class->post_schedule(rq);
1998
		raw_spin_unlock_irqrestore(&rq->lock, flags);
1999 2000 2001 2002 2003 2004

		rq->post_schedule = 0;
	}
}

#else
2005

2006 2007 2008 2009 2010 2011
static inline void pre_schedule(struct rq *rq, struct task_struct *p)
{
}

static inline void post_schedule(struct rq *rq)
{
L
Linus Torvalds 已提交
2012 2013
}

2014 2015
#endif

L
Linus Torvalds 已提交
2016 2017 2018 2019
/**
 * schedule_tail - first thing a freshly forked thread must call.
 * @prev: the thread we just switched away from.
 */
2020
asmlinkage void schedule_tail(struct task_struct *prev)
L
Linus Torvalds 已提交
2021 2022
	__releases(rq->lock)
{
2023 2024
	struct rq *rq = this_rq();

2025
	finish_task_switch(rq, prev);
2026

2027 2028 2029 2030 2031
	/*
	 * FIXME: do we need to worry about rq being invalidated by the
	 * task_switch?
	 */
	post_schedule(rq);
2032

2033 2034 2035 2036
#ifdef __ARCH_WANT_UNLOCKED_CTXSW
	/* In this case, finish_task_switch does not reenable preemption */
	preempt_enable();
#endif
L
Linus Torvalds 已提交
2037
	if (current->set_child_tid)
2038
		put_user(task_pid_vnr(current), current->set_child_tid);
L
Linus Torvalds 已提交
2039 2040 2041 2042 2043 2044
}

/*
 * context_switch - switch to the new MM and the new
 * thread's register state.
 */
I
Ingo Molnar 已提交
2045
static inline void
2046
context_switch(struct rq *rq, struct task_struct *prev,
2047
	       struct task_struct *next)
L
Linus Torvalds 已提交
2048
{
I
Ingo Molnar 已提交
2049
	struct mm_struct *mm, *oldmm;
L
Linus Torvalds 已提交
2050

2051
	prepare_task_switch(rq, prev, next);
2052

I
Ingo Molnar 已提交
2053 2054
	mm = next->mm;
	oldmm = prev->active_mm;
2055 2056 2057 2058 2059
	/*
	 * For paravirt, this is coupled with an exit in switch_to to
	 * combine the page table reload and the switch backend into
	 * one hypercall.
	 */
2060
	arch_start_context_switch(prev);
2061

2062
	if (!mm) {
L
Linus Torvalds 已提交
2063 2064 2065 2066 2067 2068
		next->active_mm = oldmm;
		atomic_inc(&oldmm->mm_count);
		enter_lazy_tlb(oldmm, next);
	} else
		switch_mm(oldmm, mm, next);

2069
	if (!prev->mm) {
L
Linus Torvalds 已提交
2070 2071 2072
		prev->active_mm = NULL;
		rq->prev_mm = oldmm;
	}
2073 2074 2075 2076 2077 2078 2079
	/*
	 * 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:
	 */
#ifndef __ARCH_WANT_UNLOCKED_CTXSW
2080
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
2081
#endif
L
Linus Torvalds 已提交
2082 2083 2084 2085

	/* Here we just switch the register state and the stack. */
	switch_to(prev, next, prev);

I
Ingo Molnar 已提交
2086 2087 2088 2089 2090 2091 2092
	barrier();
	/*
	 * this_rq must be evaluated again because prev may have moved
	 * CPUs since it called schedule(), thus the 'rq' on its stack
	 * frame will be invalid.
	 */
	finish_task_switch(this_rq(), prev);
L
Linus Torvalds 已提交
2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109
}

/*
 * nr_running, nr_uninterruptible and nr_context_switches:
 *
 * externally visible scheduler statistics: current number of runnable
 * threads, current number of uninterruptible-sleeping threads, total
 * number of context switches performed since bootup.
 */
unsigned long nr_running(void)
{
	unsigned long i, sum = 0;

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

	return sum;
2110
}
L
Linus Torvalds 已提交
2111 2112

unsigned long nr_uninterruptible(void)
2113
{
L
Linus Torvalds 已提交
2114
	unsigned long i, sum = 0;
2115

2116
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2117
		sum += cpu_rq(i)->nr_uninterruptible;
2118 2119

	/*
L
Linus Torvalds 已提交
2120 2121
	 * Since we read the counters lockless, it might be slightly
	 * inaccurate. Do not allow it to go below zero though:
2122
	 */
L
Linus Torvalds 已提交
2123 2124
	if (unlikely((long)sum < 0))
		sum = 0;
2125

L
Linus Torvalds 已提交
2126
	return sum;
2127 2128
}

L
Linus Torvalds 已提交
2129
unsigned long long nr_context_switches(void)
2130
{
2131 2132
	int i;
	unsigned long long sum = 0;
2133

2134
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2135
		sum += cpu_rq(i)->nr_switches;
2136

L
Linus Torvalds 已提交
2137 2138
	return sum;
}
2139

L
Linus Torvalds 已提交
2140 2141 2142
unsigned long nr_iowait(void)
{
	unsigned long i, sum = 0;
2143

2144
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2145
		sum += atomic_read(&cpu_rq(i)->nr_iowait);
2146

L
Linus Torvalds 已提交
2147 2148
	return sum;
}
2149

2150
unsigned long nr_iowait_cpu(int cpu)
2151
{
2152
	struct rq *this = cpu_rq(cpu);
2153 2154
	return atomic_read(&this->nr_iowait);
}
2155

2156 2157 2158 2159 2160
unsigned long this_cpu_load(void)
{
	struct rq *this = this_rq();
	return this->cpu_load[0];
}
2161

2162

2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209
/*
 * Global load-average calculations
 *
 * We take a distributed and async approach to calculating the global load-avg
 * in order to minimize overhead.
 *
 * The global load average is an exponentially decaying average of nr_running +
 * nr_uninterruptible.
 *
 * Once every LOAD_FREQ:
 *
 *   nr_active = 0;
 *   for_each_possible_cpu(cpu)
 *   	nr_active += cpu_of(cpu)->nr_running + cpu_of(cpu)->nr_uninterruptible;
 *
 *   avenrun[n] = avenrun[0] * exp_n + nr_active * (1 - exp_n)
 *
 * Due to a number of reasons the above turns in the mess below:
 *
 *  - for_each_possible_cpu() is prohibitively expensive on machines with
 *    serious number of cpus, therefore we need to take a distributed approach
 *    to calculating nr_active.
 *
 *        \Sum_i x_i(t) = \Sum_i x_i(t) - x_i(t_0) | x_i(t_0) := 0
 *                      = \Sum_i { \Sum_j=1 x_i(t_j) - x_i(t_j-1) }
 *
 *    So assuming nr_active := 0 when we start out -- true per definition, we
 *    can simply take per-cpu deltas and fold those into a global accumulate
 *    to obtain the same result. See calc_load_fold_active().
 *
 *    Furthermore, in order to avoid synchronizing all per-cpu delta folding
 *    across the machine, we assume 10 ticks is sufficient time for every
 *    cpu to have completed this task.
 *
 *    This places an upper-bound on the IRQ-off latency of the machine. Then
 *    again, being late doesn't loose the delta, just wrecks the sample.
 *
 *  - cpu_rq()->nr_uninterruptible isn't accurately tracked per-cpu because
 *    this would add another cross-cpu cacheline miss and atomic operation
 *    to the wakeup path. Instead we increment on whatever cpu the task ran
 *    when it went into uninterruptible state and decrement on whatever cpu
 *    did the wakeup. This means that only the sum of nr_uninterruptible over
 *    all cpus yields the correct result.
 *
 *  This covers the NO_HZ=n code, for extra head-aches, see the comment below.
 */

2210 2211 2212 2213
/* Variables and functions for calc_load */
static atomic_long_t calc_load_tasks;
static unsigned long calc_load_update;
unsigned long avenrun[3];
2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229
EXPORT_SYMBOL(avenrun); /* should be removed */

/**
 * get_avenrun - get the load average array
 * @loads:	pointer to dest load array
 * @offset:	offset to add
 * @shift:	shift count to shift the result left
 *
 * These values are estimates at best, so no need for locking.
 */
void get_avenrun(unsigned long *loads, unsigned long offset, int shift)
{
	loads[0] = (avenrun[0] + offset) << shift;
	loads[1] = (avenrun[1] + offset) << shift;
	loads[2] = (avenrun[2] + offset) << shift;
}
2230

2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245
static long calc_load_fold_active(struct rq *this_rq)
{
	long nr_active, delta = 0;

	nr_active = this_rq->nr_running;
	nr_active += (long) this_rq->nr_uninterruptible;

	if (nr_active != this_rq->calc_load_active) {
		delta = nr_active - this_rq->calc_load_active;
		this_rq->calc_load_active = nr_active;
	}

	return delta;
}

2246 2247 2248
/*
 * a1 = a0 * e + a * (1 - e)
 */
2249 2250 2251 2252 2253 2254 2255 2256 2257
static unsigned long
calc_load(unsigned long load, unsigned long exp, unsigned long active)
{
	load *= exp;
	load += active * (FIXED_1 - exp);
	load += 1UL << (FSHIFT - 1);
	return load >> FSHIFT;
}

2258 2259
#ifdef CONFIG_NO_HZ
/*
2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297
 * Handle NO_HZ for the global load-average.
 *
 * Since the above described distributed algorithm to compute the global
 * load-average relies on per-cpu sampling from the tick, it is affected by
 * NO_HZ.
 *
 * The basic idea is to fold the nr_active delta into a global idle-delta upon
 * entering NO_HZ state such that we can include this as an 'extra' cpu delta
 * when we read the global state.
 *
 * Obviously reality has to ruin such a delightfully simple scheme:
 *
 *  - When we go NO_HZ idle during the window, we can negate our sample
 *    contribution, causing under-accounting.
 *
 *    We avoid this by keeping two idle-delta counters and flipping them
 *    when the window starts, thus separating old and new NO_HZ load.
 *
 *    The only trick is the slight shift in index flip for read vs write.
 *
 *        0s            5s            10s           15s
 *          +10           +10           +10           +10
 *        |-|-----------|-|-----------|-|-----------|-|
 *    r:0 0 1           1 0           0 1           1 0
 *    w:0 1 1           0 0           1 1           0 0
 *
 *    This ensures we'll fold the old idle contribution in this window while
 *    accumlating the new one.
 *
 *  - When we wake up from NO_HZ idle during the window, we push up our
 *    contribution, since we effectively move our sample point to a known
 *    busy state.
 *
 *    This is solved by pushing the window forward, and thus skipping the
 *    sample, for this cpu (effectively using the idle-delta for this cpu which
 *    was in effect at the time the window opened). This also solves the issue
 *    of having to deal with a cpu having been in NOHZ idle for multiple
 *    LOAD_FREQ intervals.
2298 2299 2300
 *
 * When making the ILB scale, we should try to pull this in as well.
 */
2301 2302
static atomic_long_t calc_load_idle[2];
static int calc_load_idx;
2303

2304
static inline int calc_load_write_idx(void)
2305
{
2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331
	int idx = calc_load_idx;

	/*
	 * See calc_global_nohz(), if we observe the new index, we also
	 * need to observe the new update time.
	 */
	smp_rmb();

	/*
	 * If the folding window started, make sure we start writing in the
	 * next idle-delta.
	 */
	if (!time_before(jiffies, calc_load_update))
		idx++;

	return idx & 1;
}

static inline int calc_load_read_idx(void)
{
	return calc_load_idx & 1;
}

void calc_load_enter_idle(void)
{
	struct rq *this_rq = this_rq();
2332 2333
	long delta;

2334 2335 2336 2337
	/*
	 * We're going into NOHZ mode, if there's any pending delta, fold it
	 * into the pending idle delta.
	 */
2338
	delta = calc_load_fold_active(this_rq);
2339 2340 2341 2342
	if (delta) {
		int idx = calc_load_write_idx();
		atomic_long_add(delta, &calc_load_idle[idx]);
	}
2343 2344
}

2345
void calc_load_exit_idle(void)
2346
{
2347 2348 2349 2350 2351 2352 2353
	struct rq *this_rq = this_rq();

	/*
	 * If we're still before the sample window, we're done.
	 */
	if (time_before(jiffies, this_rq->calc_load_update))
		return;
2354 2355

	/*
2356 2357 2358
	 * We woke inside or after the sample window, this means we're already
	 * accounted through the nohz accounting, so skip the entire deal and
	 * sync up for the next window.
2359
	 */
2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371
	this_rq->calc_load_update = calc_load_update;
	if (time_before(jiffies, this_rq->calc_load_update + 10))
		this_rq->calc_load_update += LOAD_FREQ;
}

static long calc_load_fold_idle(void)
{
	int idx = calc_load_read_idx();
	long delta = 0;

	if (atomic_long_read(&calc_load_idle[idx]))
		delta = atomic_long_xchg(&calc_load_idle[idx], 0);
2372 2373 2374

	return delta;
}
2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452

/**
 * fixed_power_int - compute: x^n, in O(log n) time
 *
 * @x:         base of the power
 * @frac_bits: fractional bits of @x
 * @n:         power to raise @x to.
 *
 * By exploiting the relation between the definition of the natural power
 * function: x^n := x*x*...*x (x multiplied by itself for n times), and
 * the binary encoding of numbers used by computers: n := \Sum n_i * 2^i,
 * (where: n_i \elem {0, 1}, the binary vector representing n),
 * we find: x^n := x^(\Sum n_i * 2^i) := \Prod x^(n_i * 2^i), which is
 * of course trivially computable in O(log_2 n), the length of our binary
 * vector.
 */
static unsigned long
fixed_power_int(unsigned long x, unsigned int frac_bits, unsigned int n)
{
	unsigned long result = 1UL << frac_bits;

	if (n) for (;;) {
		if (n & 1) {
			result *= x;
			result += 1UL << (frac_bits - 1);
			result >>= frac_bits;
		}
		n >>= 1;
		if (!n)
			break;
		x *= x;
		x += 1UL << (frac_bits - 1);
		x >>= frac_bits;
	}

	return result;
}

/*
 * a1 = a0 * e + a * (1 - e)
 *
 * a2 = a1 * e + a * (1 - e)
 *    = (a0 * e + a * (1 - e)) * e + a * (1 - e)
 *    = a0 * e^2 + a * (1 - e) * (1 + e)
 *
 * a3 = a2 * e + a * (1 - e)
 *    = (a0 * e^2 + a * (1 - e) * (1 + e)) * e + a * (1 - e)
 *    = a0 * e^3 + a * (1 - e) * (1 + e + e^2)
 *
 *  ...
 *
 * an = a0 * e^n + a * (1 - e) * (1 + e + ... + e^n-1) [1]
 *    = a0 * e^n + a * (1 - e) * (1 - e^n)/(1 - e)
 *    = a0 * e^n + a * (1 - e^n)
 *
 * [1] application of the geometric series:
 *
 *              n         1 - x^(n+1)
 *     S_n := \Sum x^i = -------------
 *             i=0          1 - x
 */
static unsigned long
calc_load_n(unsigned long load, unsigned long exp,
	    unsigned long active, unsigned int n)
{

	return calc_load(load, fixed_power_int(exp, FSHIFT, n), active);
}

/*
 * NO_HZ can leave us missing all per-cpu ticks calling
 * calc_load_account_active(), but since an idle CPU folds its delta into
 * calc_load_tasks_idle per calc_load_account_idle(), all we need to do is fold
 * in the pending idle delta if our idle period crossed a load cycle boundary.
 *
 * Once we've updated the global active value, we need to apply the exponential
 * weights adjusted to the number of cycles missed.
 */
2453
static void calc_global_nohz(void)
2454 2455 2456
{
	long delta, active, n;

2457 2458 2459 2460 2461 2462
	if (!time_before(jiffies, calc_load_update + 10)) {
		/*
		 * Catch-up, fold however many we are behind still
		 */
		delta = jiffies - calc_load_update - 10;
		n = 1 + (delta / LOAD_FREQ);
2463

2464 2465
		active = atomic_long_read(&calc_load_tasks);
		active = active > 0 ? active * FIXED_1 : 0;
2466

2467 2468 2469
		avenrun[0] = calc_load_n(avenrun[0], EXP_1, active, n);
		avenrun[1] = calc_load_n(avenrun[1], EXP_5, active, n);
		avenrun[2] = calc_load_n(avenrun[2], EXP_15, active, n);
2470

2471 2472
		calc_load_update += n * LOAD_FREQ;
	}
2473

2474 2475 2476 2477 2478 2479 2480 2481 2482
	/*
	 * Flip the idle index...
	 *
	 * Make sure we first write the new time then flip the index, so that
	 * calc_load_write_idx() will see the new time when it reads the new
	 * index, this avoids a double flip messing things up.
	 */
	smp_wmb();
	calc_load_idx++;
2483
}
2484
#else /* !CONFIG_NO_HZ */
2485

2486 2487
static inline long calc_load_fold_idle(void) { return 0; }
static inline void calc_global_nohz(void) { }
2488

2489
#endif /* CONFIG_NO_HZ */
2490 2491

/*
2492 2493
 * calc_load - update the avenrun load estimates 10 ticks after the
 * CPUs have updated calc_load_tasks.
2494
 */
2495
void calc_global_load(unsigned long ticks)
2496
{
2497
	long active, delta;
L
Linus Torvalds 已提交
2498

2499
	if (time_before(jiffies, calc_load_update + 10))
2500
		return;
L
Linus Torvalds 已提交
2501

2502 2503 2504 2505 2506 2507 2508
	/*
	 * Fold the 'old' idle-delta to include all NO_HZ cpus.
	 */
	delta = calc_load_fold_idle();
	if (delta)
		atomic_long_add(delta, &calc_load_tasks);

2509 2510
	active = atomic_long_read(&calc_load_tasks);
	active = active > 0 ? active * FIXED_1 : 0;
L
Linus Torvalds 已提交
2511

2512 2513 2514
	avenrun[0] = calc_load(avenrun[0], EXP_1, active);
	avenrun[1] = calc_load(avenrun[1], EXP_5, active);
	avenrun[2] = calc_load(avenrun[2], EXP_15, active);
I
Ingo Molnar 已提交
2515

2516
	calc_load_update += LOAD_FREQ;
2517 2518

	/*
2519
	 * In case we idled for multiple LOAD_FREQ intervals, catch up in bulk.
2520 2521
	 */
	calc_global_nohz();
2522
}
L
Linus Torvalds 已提交
2523

2524
/*
2525 2526
 * Called from update_cpu_load() to periodically update this CPU's
 * active count.
2527 2528 2529
 */
static void calc_load_account_active(struct rq *this_rq)
{
2530
	long delta;
2531

2532 2533
	if (time_before(jiffies, this_rq->calc_load_update))
		return;
2534

2535 2536
	delta  = calc_load_fold_active(this_rq);
	if (delta)
2537
		atomic_long_add(delta, &calc_load_tasks);
2538 2539

	this_rq->calc_load_update += LOAD_FREQ;
2540 2541
}

2542 2543 2544 2545
/*
 * End of global load-average stuff
 */

2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612
/*
 * The exact cpuload at various idx values, calculated at every tick would be
 * load = (2^idx - 1) / 2^idx * load + 1 / 2^idx * cur_load
 *
 * If a cpu misses updates for n-1 ticks (as it was idle) and update gets called
 * on nth tick when cpu may be busy, then we have:
 * load = ((2^idx - 1) / 2^idx)^(n-1) * load
 * load = (2^idx - 1) / 2^idx) * load + 1 / 2^idx * cur_load
 *
 * decay_load_missed() below does efficient calculation of
 * load = ((2^idx - 1) / 2^idx)^(n-1) * load
 * avoiding 0..n-1 loop doing load = ((2^idx - 1) / 2^idx) * load
 *
 * The calculation is approximated on a 128 point scale.
 * degrade_zero_ticks is the number of ticks after which load at any
 * particular idx is approximated to be zero.
 * degrade_factor is a precomputed table, a row for each load idx.
 * Each column corresponds to degradation factor for a power of two ticks,
 * based on 128 point scale.
 * Example:
 * row 2, col 3 (=12) says that the degradation at load idx 2 after
 * 8 ticks is 12/128 (which is an approximation of exact factor 3^8/4^8).
 *
 * With this power of 2 load factors, we can degrade the load n times
 * by looking at 1 bits in n and doing as many mult/shift instead of
 * n mult/shifts needed by the exact degradation.
 */
#define DEGRADE_SHIFT		7
static const unsigned char
		degrade_zero_ticks[CPU_LOAD_IDX_MAX] = {0, 8, 32, 64, 128};
static const unsigned char
		degrade_factor[CPU_LOAD_IDX_MAX][DEGRADE_SHIFT + 1] = {
					{0, 0, 0, 0, 0, 0, 0, 0},
					{64, 32, 8, 0, 0, 0, 0, 0},
					{96, 72, 40, 12, 1, 0, 0},
					{112, 98, 75, 43, 15, 1, 0},
					{120, 112, 98, 76, 45, 16, 2} };

/*
 * Update cpu_load for any missed ticks, due to tickless idle. The backlog
 * would be when CPU is idle and so we just decay the old load without
 * adding any new load.
 */
static unsigned long
decay_load_missed(unsigned long load, unsigned long missed_updates, int idx)
{
	int j = 0;

	if (!missed_updates)
		return load;

	if (missed_updates >= degrade_zero_ticks[idx])
		return 0;

	if (idx == 1)
		return load >> missed_updates;

	while (missed_updates) {
		if (missed_updates % 2)
			load = (load * degrade_factor[idx][j]) >> DEGRADE_SHIFT;

		missed_updates >>= 1;
		j++;
	}
	return load;
}

2613
/*
I
Ingo Molnar 已提交
2614
 * Update rq->cpu_load[] statistics. This function is usually called every
2615 2616
 * scheduler tick (TICK_NSEC). With tickless idle this will not be called
 * every tick. We fix it up based on jiffies.
2617
 */
2618 2619
static void __update_cpu_load(struct rq *this_rq, unsigned long this_load,
			      unsigned long pending_updates)
2620
{
I
Ingo Molnar 已提交
2621
	int i, scale;
2622

I
Ingo Molnar 已提交
2623
	this_rq->nr_load_updates++;
2624

I
Ingo Molnar 已提交
2625
	/* Update our load: */
2626 2627
	this_rq->cpu_load[0] = this_load; /* Fasttrack for idx 0 */
	for (i = 1, scale = 2; i < CPU_LOAD_IDX_MAX; i++, scale += scale) {
I
Ingo Molnar 已提交
2628
		unsigned long old_load, new_load;
2629

I
Ingo Molnar 已提交
2630
		/* scale is effectively 1 << i now, and >> i divides by scale */
2631

I
Ingo Molnar 已提交
2632
		old_load = this_rq->cpu_load[i];
2633
		old_load = decay_load_missed(old_load, pending_updates - 1, i);
I
Ingo Molnar 已提交
2634
		new_load = this_load;
I
Ingo Molnar 已提交
2635 2636 2637 2638 2639 2640
		/*
		 * Round up the averaging division if load is increasing. This
		 * prevents us from getting stuck on 9 if the load is 10, for
		 * example.
		 */
		if (new_load > old_load)
2641 2642 2643
			new_load += scale - 1;

		this_rq->cpu_load[i] = (old_load * (scale - 1) + new_load) >> i;
I
Ingo Molnar 已提交
2644
	}
2645 2646

	sched_avg_update(this_rq);
2647 2648
}

2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662
#ifdef CONFIG_NO_HZ
/*
 * There is no sane way to deal with nohz on smp when using jiffies because the
 * cpu doing the jiffies update might drift wrt the cpu doing the jiffy reading
 * causing off-by-one errors in observed deltas; {0,2} instead of {1,1}.
 *
 * Therefore we cannot use the delta approach from the regular tick since that
 * would seriously skew the load calculation. However we'll make do for those
 * updates happening while idle (nohz_idle_balance) or coming out of idle
 * (tick_nohz_idle_exit).
 *
 * This means we might still be one tick off for nohz periods.
 */

2663 2664 2665 2666 2667 2668
/*
 * Called from nohz_idle_balance() to update the load ratings before doing the
 * idle balance.
 */
void update_idle_cpu_load(struct rq *this_rq)
{
2669
	unsigned long curr_jiffies = ACCESS_ONCE(jiffies);
2670 2671 2672 2673
	unsigned long load = this_rq->load.weight;
	unsigned long pending_updates;

	/*
2674
	 * bail if there's load or we're actually up-to-date.
2675 2676 2677 2678 2679 2680 2681 2682 2683 2684
	 */
	if (load || curr_jiffies == this_rq->last_load_update_tick)
		return;

	pending_updates = curr_jiffies - this_rq->last_load_update_tick;
	this_rq->last_load_update_tick = curr_jiffies;

	__update_cpu_load(this_rq, load, pending_updates);
}

2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710
/*
 * Called from tick_nohz_idle_exit() -- try and fix up the ticks we missed.
 */
void update_cpu_load_nohz(void)
{
	struct rq *this_rq = this_rq();
	unsigned long curr_jiffies = ACCESS_ONCE(jiffies);
	unsigned long pending_updates;

	if (curr_jiffies == this_rq->last_load_update_tick)
		return;

	raw_spin_lock(&this_rq->lock);
	pending_updates = curr_jiffies - this_rq->last_load_update_tick;
	if (pending_updates) {
		this_rq->last_load_update_tick = curr_jiffies;
		/*
		 * We were idle, this means load 0, the current load might be
		 * !0 due to remote wakeups and the sort.
		 */
		__update_cpu_load(this_rq, 0, pending_updates);
	}
	raw_spin_unlock(&this_rq->lock);
}
#endif /* CONFIG_NO_HZ */

2711 2712 2713
/*
 * Called from scheduler_tick()
 */
2714 2715
static void update_cpu_load_active(struct rq *this_rq)
{
2716
	/*
2717
	 * See the mess around update_idle_cpu_load() / update_cpu_load_nohz().
2718 2719 2720
	 */
	this_rq->last_load_update_tick = jiffies;
	__update_cpu_load(this_rq, this_rq->load.weight, 1);
2721

2722
	calc_load_account_active(this_rq);
2723 2724
}

I
Ingo Molnar 已提交
2725
#ifdef CONFIG_SMP
2726

2727
/*
P
Peter Zijlstra 已提交
2728 2729
 * sched_exec - execve() is a valuable balancing opportunity, because at
 * this point the task has the smallest effective memory and cache footprint.
2730
 */
P
Peter Zijlstra 已提交
2731
void sched_exec(void)
2732
{
P
Peter Zijlstra 已提交
2733
	struct task_struct *p = current;
L
Linus Torvalds 已提交
2734
	unsigned long flags;
2735
	int dest_cpu;
2736

2737
	raw_spin_lock_irqsave(&p->pi_lock, flags);
2738
	dest_cpu = p->sched_class->select_task_rq(p, SD_BALANCE_EXEC, 0);
2739 2740
	if (dest_cpu == smp_processor_id())
		goto unlock;
P
Peter Zijlstra 已提交
2741

2742
	if (likely(cpu_active(dest_cpu))) {
2743
		struct migration_arg arg = { p, dest_cpu };
2744

2745 2746
		raw_spin_unlock_irqrestore(&p->pi_lock, flags);
		stop_one_cpu(task_cpu(p), migration_cpu_stop, &arg);
L
Linus Torvalds 已提交
2747 2748
		return;
	}
2749
unlock:
2750
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
2751
}
I
Ingo Molnar 已提交
2752

L
Linus Torvalds 已提交
2753 2754 2755
#endif

DEFINE_PER_CPU(struct kernel_stat, kstat);
2756
DEFINE_PER_CPU(struct kernel_cpustat, kernel_cpustat);
L
Linus Torvalds 已提交
2757 2758

EXPORT_PER_CPU_SYMBOL(kstat);
2759
EXPORT_PER_CPU_SYMBOL(kernel_cpustat);
L
Linus Torvalds 已提交
2760 2761

/*
2762
 * Return any ns on the sched_clock that have not yet been accounted in
2763
 * @p in case that task is currently running.
2764 2765
 *
 * Called with task_rq_lock() held on @rq.
L
Linus Torvalds 已提交
2766
 */
2767 2768 2769 2770 2771 2772
static u64 do_task_delta_exec(struct task_struct *p, struct rq *rq)
{
	u64 ns = 0;

	if (task_current(rq, p)) {
		update_rq_clock(rq);
2773
		ns = rq->clock_task - p->se.exec_start;
2774 2775 2776 2777 2778 2779 2780
		if ((s64)ns < 0)
			ns = 0;
	}

	return ns;
}

2781
unsigned long long task_delta_exec(struct task_struct *p)
L
Linus Torvalds 已提交
2782 2783
{
	unsigned long flags;
2784
	struct rq *rq;
2785
	u64 ns = 0;
2786

2787
	rq = task_rq_lock(p, &flags);
2788
	ns = do_task_delta_exec(p, rq);
2789
	task_rq_unlock(rq, p, &flags);
2790

2791 2792
	return ns;
}
2793

2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806
/*
 * 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;
	u64 ns = 0;

	rq = task_rq_lock(p, &flags);
	ns = p->se.sum_exec_runtime + do_task_delta_exec(p, rq);
2807
	task_rq_unlock(rq, p, &flags);
2808 2809 2810

	return ns;
}
2811

2812 2813 2814 2815 2816
#ifdef CONFIG_CGROUP_CPUACCT
struct cgroup_subsys cpuacct_subsys;
struct cpuacct root_cpuacct;
#endif

2817 2818
static inline void task_group_account_field(struct task_struct *p, int index,
					    u64 tmp)
2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847
{
#ifdef CONFIG_CGROUP_CPUACCT
	struct kernel_cpustat *kcpustat;
	struct cpuacct *ca;
#endif
	/*
	 * Since all updates are sure to touch the root cgroup, we
	 * get ourselves ahead and touch it first. If the root cgroup
	 * is the only cgroup, then nothing else should be necessary.
	 *
	 */
	__get_cpu_var(kernel_cpustat).cpustat[index] += tmp;

#ifdef CONFIG_CGROUP_CPUACCT
	if (unlikely(!cpuacct_subsys.active))
		return;

	rcu_read_lock();
	ca = task_ca(p);
	while (ca && (ca != &root_cpuacct)) {
		kcpustat = this_cpu_ptr(ca->cpustat);
		kcpustat->cpustat[index] += tmp;
		ca = parent_ca(ca);
	}
	rcu_read_unlock();
#endif
}


L
Linus Torvalds 已提交
2848 2849 2850 2851
/*
 * Account user cpu time to a process.
 * @p: the process that the cpu time gets accounted to
 * @cputime: the cpu time spent in user space since the last update
2852
 * @cputime_scaled: cputime scaled by cpu frequency
L
Linus Torvalds 已提交
2853
 */
2854 2855
void account_user_time(struct task_struct *p, cputime_t cputime,
		       cputime_t cputime_scaled)
L
Linus Torvalds 已提交
2856
{
2857
	int index;
L
Linus Torvalds 已提交
2858

2859
	/* Add user time to process. */
2860 2861
	p->utime += cputime;
	p->utimescaled += cputime_scaled;
2862
	account_group_user_time(p, cputime);
L
Linus Torvalds 已提交
2863

2864
	index = (TASK_NICE(p) > 0) ? CPUTIME_NICE : CPUTIME_USER;
2865

L
Linus Torvalds 已提交
2866
	/* Add user time to cpustat. */
2867
	task_group_account_field(p, index, (__force u64) cputime);
2868

2869 2870
	/* Account for user time used */
	acct_update_integrals(p);
L
Linus Torvalds 已提交
2871 2872
}

2873 2874 2875 2876
/*
 * Account guest cpu time to a process.
 * @p: the process that the cpu time gets accounted to
 * @cputime: the cpu time spent in virtual machine since the last update
2877
 * @cputime_scaled: cputime scaled by cpu frequency
2878
 */
2879 2880
static void account_guest_time(struct task_struct *p, cputime_t cputime,
			       cputime_t cputime_scaled)
2881
{
2882
	u64 *cpustat = kcpustat_this_cpu->cpustat;
2883

2884
	/* Add guest time to process. */
2885 2886
	p->utime += cputime;
	p->utimescaled += cputime_scaled;
2887
	account_group_user_time(p, cputime);
2888
	p->gtime += cputime;
2889

2890
	/* Add guest time to cpustat. */
2891
	if (TASK_NICE(p) > 0) {
2892 2893
		cpustat[CPUTIME_NICE] += (__force u64) cputime;
		cpustat[CPUTIME_GUEST_NICE] += (__force u64) cputime;
2894
	} else {
2895 2896
		cpustat[CPUTIME_USER] += (__force u64) cputime;
		cpustat[CPUTIME_GUEST] += (__force u64) cputime;
2897
	}
2898 2899
}

2900 2901 2902 2903 2904 2905 2906 2907 2908
/*
 * Account system cpu time to a process and desired cpustat field
 * @p: the process that the cpu time gets accounted to
 * @cputime: the cpu time spent in kernel space since the last update
 * @cputime_scaled: cputime scaled by cpu frequency
 * @target_cputime64: pointer to cpustat field that has to be updated
 */
static inline
void __account_system_time(struct task_struct *p, cputime_t cputime,
2909
			cputime_t cputime_scaled, int index)
2910 2911
{
	/* Add system time to process. */
2912 2913
	p->stime += cputime;
	p->stimescaled += cputime_scaled;
2914 2915 2916
	account_group_system_time(p, cputime);

	/* Add system time to cpustat. */
2917
	task_group_account_field(p, index, (__force u64) cputime);
2918 2919 2920 2921 2922

	/* Account for system time used */
	acct_update_integrals(p);
}

L
Linus Torvalds 已提交
2923 2924 2925 2926 2927
/*
 * Account system cpu time to a process.
 * @p: the process that the cpu time gets accounted to
 * @hardirq_offset: the offset to subtract from hardirq_count()
 * @cputime: the cpu time spent in kernel space since the last update
2928
 * @cputime_scaled: cputime scaled by cpu frequency
L
Linus Torvalds 已提交
2929 2930
 */
void account_system_time(struct task_struct *p, int hardirq_offset,
2931
			 cputime_t cputime, cputime_t cputime_scaled)
L
Linus Torvalds 已提交
2932
{
2933
	int index;
L
Linus Torvalds 已提交
2934

2935
	if ((p->flags & PF_VCPU) && (irq_count() - hardirq_offset == 0)) {
2936
		account_guest_time(p, cputime, cputime_scaled);
2937 2938
		return;
	}
2939

L
Linus Torvalds 已提交
2940
	if (hardirq_count() - hardirq_offset)
2941
		index = CPUTIME_IRQ;
2942
	else if (in_serving_softirq())
2943
		index = CPUTIME_SOFTIRQ;
L
Linus Torvalds 已提交
2944
	else
2945
		index = CPUTIME_SYSTEM;
2946

2947
	__account_system_time(p, cputime, cputime_scaled, index);
L
Linus Torvalds 已提交
2948 2949
}

2950
/*
L
Linus Torvalds 已提交
2951
 * Account for involuntary wait time.
2952
 * @cputime: the cpu time spent in involuntary wait
2953
 */
2954
void account_steal_time(cputime_t cputime)
2955
{
2956
	u64 *cpustat = kcpustat_this_cpu->cpustat;
2957

2958
	cpustat[CPUTIME_STEAL] += (__force u64) cputime;
2959 2960
}

L
Linus Torvalds 已提交
2961
/*
2962 2963
 * Account for idle time.
 * @cputime: the cpu time spent in idle wait
L
Linus Torvalds 已提交
2964
 */
2965
void account_idle_time(cputime_t cputime)
L
Linus Torvalds 已提交
2966
{
2967
	u64 *cpustat = kcpustat_this_cpu->cpustat;
2968
	struct rq *rq = this_rq();
L
Linus Torvalds 已提交
2969

2970
	if (atomic_read(&rq->nr_iowait) > 0)
2971
		cpustat[CPUTIME_IOWAIT] += (__force u64) cputime;
2972
	else
2973
		cpustat[CPUTIME_IDLE] += (__force u64) cputime;
L
Linus Torvalds 已提交
2974 2975
}

G
Glauber Costa 已提交
2976 2977 2978
static __always_inline bool steal_account_process_tick(void)
{
#ifdef CONFIG_PARAVIRT
2979
	if (static_key_false(&paravirt_steal_enabled)) {
G
Glauber Costa 已提交
2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994
		u64 steal, st = 0;

		steal = paravirt_steal_clock(smp_processor_id());
		steal -= this_rq()->prev_steal_time;

		st = steal_ticks(steal);
		this_rq()->prev_steal_time += st * TICK_NSEC;

		account_steal_time(st);
		return st;
	}
#endif
	return false;
}

2995 2996
#ifndef CONFIG_VIRT_CPU_ACCOUNTING

2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022
#ifdef CONFIG_IRQ_TIME_ACCOUNTING
/*
 * Account a tick to a process and cpustat
 * @p: the process that the cpu time gets accounted to
 * @user_tick: is the tick from userspace
 * @rq: the pointer to rq
 *
 * Tick demultiplexing follows the order
 * - pending hardirq update
 * - pending softirq update
 * - user_time
 * - idle_time
 * - system time
 *   - check for guest_time
 *   - else account as system_time
 *
 * Check for hardirq is done both for system and user time as there is
 * no timer going off while we are on hardirq and hence we may never get an
 * opportunity to update it solely in system time.
 * p->stime and friends are only updated on system time and not on irq
 * softirq as those do not count in task exec_runtime any more.
 */
static void irqtime_account_process_tick(struct task_struct *p, int user_tick,
						struct rq *rq)
{
	cputime_t one_jiffy_scaled = cputime_to_scaled(cputime_one_jiffy);
3023
	u64 *cpustat = kcpustat_this_cpu->cpustat;
3024

G
Glauber Costa 已提交
3025 3026 3027
	if (steal_account_process_tick())
		return;

3028
	if (irqtime_account_hi_update()) {
3029
		cpustat[CPUTIME_IRQ] += (__force u64) cputime_one_jiffy;
3030
	} else if (irqtime_account_si_update()) {
3031
		cpustat[CPUTIME_SOFTIRQ] += (__force u64) cputime_one_jiffy;
3032 3033 3034 3035 3036 3037 3038
	} else if (this_cpu_ksoftirqd() == p) {
		/*
		 * ksoftirqd time do not get accounted in cpu_softirq_time.
		 * So, we have to handle it separately here.
		 * Also, p->stime needs to be updated for ksoftirqd.
		 */
		__account_system_time(p, cputime_one_jiffy, one_jiffy_scaled,
3039
					CPUTIME_SOFTIRQ);
3040 3041 3042 3043 3044 3045 3046 3047
	} else if (user_tick) {
		account_user_time(p, cputime_one_jiffy, one_jiffy_scaled);
	} else if (p == rq->idle) {
		account_idle_time(cputime_one_jiffy);
	} else if (p->flags & PF_VCPU) { /* System time or guest time */
		account_guest_time(p, cputime_one_jiffy, one_jiffy_scaled);
	} else {
		__account_system_time(p, cputime_one_jiffy, one_jiffy_scaled,
3048
					CPUTIME_SYSTEM);
3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059
	}
}

static void irqtime_account_idle_ticks(int ticks)
{
	int i;
	struct rq *rq = this_rq();

	for (i = 0; i < ticks; i++)
		irqtime_account_process_tick(current, 0, rq);
}
3060
#else /* CONFIG_IRQ_TIME_ACCOUNTING */
3061 3062 3063
static void irqtime_account_idle_ticks(int ticks) {}
static void irqtime_account_process_tick(struct task_struct *p, int user_tick,
						struct rq *rq) {}
3064
#endif /* CONFIG_IRQ_TIME_ACCOUNTING */
3065 3066 3067 3068 3069 3070 3071 3072

/*
 * Account a single tick of cpu time.
 * @p: the process that the cpu time gets accounted to
 * @user_tick: indicates if the tick is a user or a system tick
 */
void account_process_tick(struct task_struct *p, int user_tick)
{
3073
	cputime_t one_jiffy_scaled = cputime_to_scaled(cputime_one_jiffy);
3074 3075
	struct rq *rq = this_rq();

3076 3077 3078 3079 3080
	if (sched_clock_irqtime) {
		irqtime_account_process_tick(p, user_tick, rq);
		return;
	}

G
Glauber Costa 已提交
3081 3082 3083
	if (steal_account_process_tick())
		return;

3084
	if (user_tick)
3085
		account_user_time(p, cputime_one_jiffy, one_jiffy_scaled);
3086
	else if ((p != rq->idle) || (irq_count() != HARDIRQ_OFFSET))
3087
		account_system_time(p, HARDIRQ_OFFSET, cputime_one_jiffy,
3088 3089
				    one_jiffy_scaled);
	else
3090
		account_idle_time(cputime_one_jiffy);
3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108
}

/*
 * Account multiple ticks of steal time.
 * @p: the process from which the cpu time has been stolen
 * @ticks: number of stolen ticks
 */
void account_steal_ticks(unsigned long ticks)
{
	account_steal_time(jiffies_to_cputime(ticks));
}

/*
 * Account multiple ticks of idle time.
 * @ticks: number of stolen ticks
 */
void account_idle_ticks(unsigned long ticks)
{
3109 3110 3111 3112 3113 3114

	if (sched_clock_irqtime) {
		irqtime_account_idle_ticks(ticks);
		return;
	}

3115
	account_idle_time(jiffies_to_cputime(ticks));
L
Linus Torvalds 已提交
3116 3117
}

3118 3119
#endif

3120 3121 3122 3123
/*
 * Use precise platform statistics if available:
 */
#ifdef CONFIG_VIRT_CPU_ACCOUNTING
3124
void task_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
3125
{
3126 3127
	*ut = p->utime;
	*st = p->stime;
3128 3129
}

3130
void thread_group_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
3131
{
3132 3133 3134 3135 3136 3137
	struct task_cputime cputime;

	thread_group_cputime(p, &cputime);

	*ut = cputime.utime;
	*st = cputime.stime;
3138 3139
}
#else
3140 3141

#ifndef nsecs_to_cputime
3142
# define nsecs_to_cputime(__nsecs)	nsecs_to_jiffies(__nsecs)
3143 3144
#endif

3145
void task_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
3146
{
3147
	cputime_t rtime, utime = p->utime, total = utime + p->stime;
3148 3149 3150 3151

	/*
	 * Use CFS's precise accounting:
	 */
3152
	rtime = nsecs_to_cputime(p->se.sum_exec_runtime);
3153 3154

	if (total) {
3155
		u64 temp = (__force u64) rtime;
3156

3157 3158 3159
		temp *= (__force u64) utime;
		do_div(temp, (__force u32) total);
		utime = (__force cputime_t) temp;
3160 3161
	} else
		utime = rtime;
3162

3163 3164 3165
	/*
	 * Compare with previous values, to keep monotonicity:
	 */
3166
	p->prev_utime = max(p->prev_utime, utime);
3167
	p->prev_stime = max(p->prev_stime, rtime - p->prev_utime);
3168

3169 3170
	*ut = p->prev_utime;
	*st = p->prev_stime;
3171 3172
}

3173 3174 3175 3176
/*
 * Must be called with siglock held.
 */
void thread_group_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
3177
{
3178 3179 3180
	struct signal_struct *sig = p->signal;
	struct task_cputime cputime;
	cputime_t rtime, utime, total;
3181

3182
	thread_group_cputime(p, &cputime);
3183

3184
	total = cputime.utime + cputime.stime;
3185
	rtime = nsecs_to_cputime(cputime.sum_exec_runtime);
3186

3187
	if (total) {
3188
		u64 temp = (__force u64) rtime;
3189

3190 3191 3192
		temp *= (__force u64) cputime.utime;
		do_div(temp, (__force u32) total);
		utime = (__force cputime_t) temp;
3193 3194 3195 3196
	} else
		utime = rtime;

	sig->prev_utime = max(sig->prev_utime, utime);
3197
	sig->prev_stime = max(sig->prev_stime, rtime - sig->prev_utime);
3198 3199 3200

	*ut = sig->prev_utime;
	*st = sig->prev_stime;
3201 3202 3203
}
#endif

3204 3205 3206 3207 3208 3209 3210 3211
/*
 * 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
3212
	struct task_struct *curr = rq->curr;
3213 3214

	sched_clock_tick();
I
Ingo Molnar 已提交
3215

3216
	raw_spin_lock(&rq->lock);
3217
	update_rq_clock(rq);
3218
	update_cpu_load_active(rq);
P
Peter Zijlstra 已提交
3219
	curr->sched_class->task_tick(rq, curr, 0);
3220
	raw_spin_unlock(&rq->lock);
3221

3222
	perf_event_task_tick();
3223

3224
#ifdef CONFIG_SMP
3225
	rq->idle_balance = idle_cpu(cpu);
I
Ingo Molnar 已提交
3226
	trigger_load_balance(rq, cpu);
3227
#endif
L
Linus Torvalds 已提交
3228 3229
}

3230
notrace unsigned long get_parent_ip(unsigned long addr)
3231 3232 3233 3234 3235 3236 3237 3238
{
	if (in_lock_functions(addr)) {
		addr = CALLER_ADDR2;
		if (in_lock_functions(addr))
			addr = CALLER_ADDR3;
	}
	return addr;
}
L
Linus Torvalds 已提交
3239

3240 3241 3242
#if defined(CONFIG_PREEMPT) && (defined(CONFIG_DEBUG_PREEMPT) || \
				defined(CONFIG_PREEMPT_TRACER))

3243
void __kprobes add_preempt_count(int val)
L
Linus Torvalds 已提交
3244
{
3245
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3246 3247 3248
	/*
	 * Underflow?
	 */
3249 3250
	if (DEBUG_LOCKS_WARN_ON((preempt_count() < 0)))
		return;
3251
#endif
L
Linus Torvalds 已提交
3252
	preempt_count() += val;
3253
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3254 3255 3256
	/*
	 * Spinlock count overflowing soon?
	 */
3257 3258
	DEBUG_LOCKS_WARN_ON((preempt_count() & PREEMPT_MASK) >=
				PREEMPT_MASK - 10);
3259 3260 3261
#endif
	if (preempt_count() == val)
		trace_preempt_off(CALLER_ADDR0, get_parent_ip(CALLER_ADDR1));
L
Linus Torvalds 已提交
3262 3263 3264
}
EXPORT_SYMBOL(add_preempt_count);

3265
void __kprobes sub_preempt_count(int val)
L
Linus Torvalds 已提交
3266
{
3267
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3268 3269 3270
	/*
	 * Underflow?
	 */
3271
	if (DEBUG_LOCKS_WARN_ON(val > preempt_count()))
3272
		return;
L
Linus Torvalds 已提交
3273 3274 3275
	/*
	 * Is the spinlock portion underflowing?
	 */
3276 3277 3278
	if (DEBUG_LOCKS_WARN_ON((val < PREEMPT_MASK) &&
			!(preempt_count() & PREEMPT_MASK)))
		return;
3279
#endif
3280

3281 3282
	if (preempt_count() == val)
		trace_preempt_on(CALLER_ADDR0, get_parent_ip(CALLER_ADDR1));
L
Linus Torvalds 已提交
3283 3284 3285 3286 3287 3288 3289
	preempt_count() -= val;
}
EXPORT_SYMBOL(sub_preempt_count);

#endif

/*
I
Ingo Molnar 已提交
3290
 * Print scheduling while atomic bug:
L
Linus Torvalds 已提交
3291
 */
I
Ingo Molnar 已提交
3292
static noinline void __schedule_bug(struct task_struct *prev)
L
Linus Torvalds 已提交
3293
{
3294 3295 3296
	if (oops_in_progress)
		return;

P
Peter Zijlstra 已提交
3297 3298
	printk(KERN_ERR "BUG: scheduling while atomic: %s/%d/0x%08x\n",
		prev->comm, prev->pid, preempt_count());
3299

I
Ingo Molnar 已提交
3300
	debug_show_held_locks(prev);
3301
	print_modules();
I
Ingo Molnar 已提交
3302 3303
	if (irqs_disabled())
		print_irqtrace_events(prev);
3304
	dump_stack();
3305
	add_taint(TAINT_WARN);
I
Ingo Molnar 已提交
3306
}
L
Linus Torvalds 已提交
3307

I
Ingo Molnar 已提交
3308 3309 3310 3311 3312
/*
 * Various schedule()-time debugging checks and statistics:
 */
static inline void schedule_debug(struct task_struct *prev)
{
L
Linus Torvalds 已提交
3313
	/*
I
Ingo Molnar 已提交
3314
	 * Test if we are atomic. Since do_exit() needs to call into
L
Linus Torvalds 已提交
3315 3316 3317
	 * schedule() atomically, we ignore that path for now.
	 * Otherwise, whine if we are scheduling when we should not be.
	 */
3318
	if (unlikely(in_atomic_preempt_off() && !prev->exit_state))
I
Ingo Molnar 已提交
3319
		__schedule_bug(prev);
3320
	rcu_sleep_check();
I
Ingo Molnar 已提交
3321

L
Linus Torvalds 已提交
3322 3323
	profile_hit(SCHED_PROFILING, __builtin_return_address(0));

3324
	schedstat_inc(this_rq(), sched_count);
I
Ingo Molnar 已提交
3325 3326
}

P
Peter Zijlstra 已提交
3327
static void put_prev_task(struct rq *rq, struct task_struct *prev)
M
Mike Galbraith 已提交
3328
{
3329
	if (prev->on_rq || rq->skip_clock_update < 0)
3330
		update_rq_clock(rq);
P
Peter Zijlstra 已提交
3331
	prev->sched_class->put_prev_task(rq, prev);
M
Mike Galbraith 已提交
3332 3333
}

I
Ingo Molnar 已提交
3334 3335 3336 3337
/*
 * Pick up the highest-prio task:
 */
static inline struct task_struct *
3338
pick_next_task(struct rq *rq)
I
Ingo Molnar 已提交
3339
{
3340
	const struct sched_class *class;
I
Ingo Molnar 已提交
3341
	struct task_struct *p;
L
Linus Torvalds 已提交
3342 3343

	/*
I
Ingo Molnar 已提交
3344 3345
	 * Optimization: we know that if all tasks are in
	 * the fair class we can call that function directly:
L
Linus Torvalds 已提交
3346
	 */
3347
	if (likely(rq->nr_running == rq->cfs.h_nr_running)) {
3348
		p = fair_sched_class.pick_next_task(rq);
I
Ingo Molnar 已提交
3349 3350
		if (likely(p))
			return p;
L
Linus Torvalds 已提交
3351 3352
	}

3353
	for_each_class(class) {
3354
		p = class->pick_next_task(rq);
I
Ingo Molnar 已提交
3355 3356 3357
		if (p)
			return p;
	}
3358 3359

	BUG(); /* the idle class will always have a runnable task */
I
Ingo Molnar 已提交
3360
}
L
Linus Torvalds 已提交
3361

I
Ingo Molnar 已提交
3362
/*
3363
 * __schedule() is the main scheduler function.
I
Ingo Molnar 已提交
3364
 */
3365
static void __sched __schedule(void)
I
Ingo Molnar 已提交
3366 3367
{
	struct task_struct *prev, *next;
3368
	unsigned long *switch_count;
I
Ingo Molnar 已提交
3369
	struct rq *rq;
3370
	int cpu;
I
Ingo Molnar 已提交
3371

3372 3373
need_resched:
	preempt_disable();
I
Ingo Molnar 已提交
3374 3375
	cpu = smp_processor_id();
	rq = cpu_rq(cpu);
3376
	rcu_note_context_switch(cpu);
I
Ingo Molnar 已提交
3377 3378 3379
	prev = rq->curr;

	schedule_debug(prev);
L
Linus Torvalds 已提交
3380

3381
	if (sched_feat(HRTICK))
M
Mike Galbraith 已提交
3382
		hrtick_clear(rq);
P
Peter Zijlstra 已提交
3383

3384
	raw_spin_lock_irq(&rq->lock);
L
Linus Torvalds 已提交
3385

3386
	switch_count = &prev->nivcsw;
L
Linus Torvalds 已提交
3387
	if (prev->state && !(preempt_count() & PREEMPT_ACTIVE)) {
T
Tejun Heo 已提交
3388
		if (unlikely(signal_pending_state(prev->state, prev))) {
L
Linus Torvalds 已提交
3389
			prev->state = TASK_RUNNING;
T
Tejun Heo 已提交
3390
		} else {
3391 3392 3393
			deactivate_task(rq, prev, DEQUEUE_SLEEP);
			prev->on_rq = 0;

T
Tejun Heo 已提交
3394
			/*
3395 3396 3397
			 * 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
3398 3399 3400 3401 3402 3403 3404 3405 3406
			 */
			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
3407
		switch_count = &prev->nvcsw;
L
Linus Torvalds 已提交
3408 3409
	}

3410
	pre_schedule(rq, prev);
3411

I
Ingo Molnar 已提交
3412
	if (unlikely(!rq->nr_running))
L
Linus Torvalds 已提交
3413 3414
		idle_balance(cpu, rq);

M
Mike Galbraith 已提交
3415
	put_prev_task(rq, prev);
3416
	next = pick_next_task(rq);
3417 3418
	clear_tsk_need_resched(prev);
	rq->skip_clock_update = 0;
L
Linus Torvalds 已提交
3419 3420 3421 3422 3423 3424

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

I
Ingo Molnar 已提交
3425
		context_switch(rq, prev, next); /* unlocks the rq */
P
Peter Zijlstra 已提交
3426
		/*
3427 3428 3429 3430
		 * 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 it can be moved to another cpu/rq.
P
Peter Zijlstra 已提交
3431 3432 3433
		 */
		cpu = smp_processor_id();
		rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
3434
	} else
3435
		raw_spin_unlock_irq(&rq->lock);
L
Linus Torvalds 已提交
3436

3437
	post_schedule(rq);
L
Linus Torvalds 已提交
3438

3439
	sched_preempt_enable_no_resched();
3440
	if (need_resched())
L
Linus Torvalds 已提交
3441 3442
		goto need_resched;
}
3443

3444 3445
static inline void sched_submit_work(struct task_struct *tsk)
{
3446
	if (!tsk->state || tsk_is_pi_blocked(tsk))
3447 3448 3449 3450 3451 3452 3453 3454 3455
		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);
}

S
Simon Kirby 已提交
3456
asmlinkage void __sched schedule(void)
3457
{
3458 3459 3460
	struct task_struct *tsk = current;

	sched_submit_work(tsk);
3461 3462
	__schedule();
}
L
Linus Torvalds 已提交
3463 3464
EXPORT_SYMBOL(schedule);

3465 3466 3467 3468 3469 3470 3471
/**
 * schedule_preempt_disabled - called with preemption disabled
 *
 * Returns with preemption disabled. Note: preempt_count must be 1
 */
void __sched schedule_preempt_disabled(void)
{
3472
	sched_preempt_enable_no_resched();
3473 3474 3475 3476
	schedule();
	preempt_disable();
}

3477
#ifdef CONFIG_MUTEX_SPIN_ON_OWNER
3478

3479 3480 3481
static inline bool owner_running(struct mutex *lock, struct task_struct *owner)
{
	if (lock->owner != owner)
3482
		return false;
3483 3484

	/*
3485 3486 3487 3488
	 * Ensure we emit the owner->on_cpu, dereference _after_ checking
	 * lock->owner still matches owner, if that fails, owner might
	 * point to free()d memory, if it still matches, the rcu_read_lock()
	 * ensures the memory stays valid.
3489
	 */
3490
	barrier();
3491

3492
	return owner->on_cpu;
3493
}
3494

3495 3496 3497 3498 3499 3500 3501 3502
/*
 * Look out! "owner" is an entirely speculative pointer
 * access and not reliable.
 */
int mutex_spin_on_owner(struct mutex *lock, struct task_struct *owner)
{
	if (!sched_feat(OWNER_SPIN))
		return 0;
3503

3504
	rcu_read_lock();
3505 3506
	while (owner_running(lock, owner)) {
		if (need_resched())
3507
			break;
3508

3509
		arch_mutex_cpu_relax();
3510
	}
3511
	rcu_read_unlock();
3512

3513
	/*
3514 3515 3516
	 * We break out the loop above on need_resched() and when the
	 * owner changed, which is a sign for heavy contention. Return
	 * success only when lock->owner is NULL.
3517
	 */
3518
	return lock->owner == NULL;
3519 3520 3521
}
#endif

L
Linus Torvalds 已提交
3522 3523
#ifdef CONFIG_PREEMPT
/*
3524
 * this is the entry point to schedule() from in-kernel preemption
I
Ingo Molnar 已提交
3525
 * off of preempt_enable. Kernel preemptions off return from interrupt
L
Linus Torvalds 已提交
3526 3527
 * occur there and call schedule directly.
 */
3528
asmlinkage void __sched notrace preempt_schedule(void)
L
Linus Torvalds 已提交
3529 3530
{
	struct thread_info *ti = current_thread_info();
3531

L
Linus Torvalds 已提交
3532 3533
	/*
	 * If there is a non-zero preempt_count or interrupts are disabled,
I
Ingo Molnar 已提交
3534
	 * we do not want to preempt the current task. Just return..
L
Linus Torvalds 已提交
3535
	 */
N
Nick Piggin 已提交
3536
	if (likely(ti->preempt_count || irqs_disabled()))
L
Linus Torvalds 已提交
3537 3538
		return;

3539
	do {
3540
		add_preempt_count_notrace(PREEMPT_ACTIVE);
3541
		__schedule();
3542
		sub_preempt_count_notrace(PREEMPT_ACTIVE);
L
Linus Torvalds 已提交
3543

3544 3545 3546 3547 3548
		/*
		 * Check again in case we missed a preemption opportunity
		 * between schedule and now.
		 */
		barrier();
3549
	} while (need_resched());
L
Linus Torvalds 已提交
3550 3551 3552 3553
}
EXPORT_SYMBOL(preempt_schedule);

/*
3554
 * this is the entry point to schedule() from kernel preemption
L
Linus Torvalds 已提交
3555 3556 3557 3558 3559 3560 3561
 * off of irq context.
 * Note, that this is called and return with irqs disabled. This will
 * protect us against recursive calling from irq.
 */
asmlinkage void __sched preempt_schedule_irq(void)
{
	struct thread_info *ti = current_thread_info();
3562

3563
	/* Catch callers which need to be fixed */
L
Linus Torvalds 已提交
3564 3565
	BUG_ON(ti->preempt_count || !irqs_disabled());

3566 3567 3568
	do {
		add_preempt_count(PREEMPT_ACTIVE);
		local_irq_enable();
3569
		__schedule();
3570 3571
		local_irq_disable();
		sub_preempt_count(PREEMPT_ACTIVE);
L
Linus Torvalds 已提交
3572

3573 3574 3575 3576 3577
		/*
		 * Check again in case we missed a preemption opportunity
		 * between schedule and now.
		 */
		barrier();
3578
	} while (need_resched());
L
Linus Torvalds 已提交
3579 3580 3581 3582
}

#endif /* CONFIG_PREEMPT */

P
Peter Zijlstra 已提交
3583
int default_wake_function(wait_queue_t *curr, unsigned mode, int wake_flags,
I
Ingo Molnar 已提交
3584
			  void *key)
L
Linus Torvalds 已提交
3585
{
P
Peter Zijlstra 已提交
3586
	return try_to_wake_up(curr->private, mode, wake_flags);
L
Linus Torvalds 已提交
3587 3588 3589 3590
}
EXPORT_SYMBOL(default_wake_function);

/*
I
Ingo Molnar 已提交
3591 3592
 * The core wakeup function. Non-exclusive wakeups (nr_exclusive == 0) just
 * wake everything up. If it's an exclusive wakeup (nr_exclusive == small +ve
L
Linus Torvalds 已提交
3593 3594 3595
 * number) then we wake all the non-exclusive tasks and one exclusive task.
 *
 * There are circumstances in which we can try to wake a task which has already
I
Ingo Molnar 已提交
3596
 * started to run but is not in state TASK_RUNNING. try_to_wake_up() returns
L
Linus Torvalds 已提交
3597 3598
 * zero in this (rare) case, and we handle it by continuing to scan the queue.
 */
3599
static void __wake_up_common(wait_queue_head_t *q, unsigned int mode,
P
Peter Zijlstra 已提交
3600
			int nr_exclusive, int wake_flags, void *key)
L
Linus Torvalds 已提交
3601
{
3602
	wait_queue_t *curr, *next;
L
Linus Torvalds 已提交
3603

3604
	list_for_each_entry_safe(curr, next, &q->task_list, task_list) {
3605 3606
		unsigned flags = curr->flags;

P
Peter Zijlstra 已提交
3607
		if (curr->func(curr, mode, wake_flags, key) &&
3608
				(flags & WQ_FLAG_EXCLUSIVE) && !--nr_exclusive)
L
Linus Torvalds 已提交
3609 3610 3611 3612 3613 3614 3615 3616 3617
			break;
	}
}

/**
 * __wake_up - wake up threads blocked on a waitqueue.
 * @q: the waitqueue
 * @mode: which threads
 * @nr_exclusive: how many wake-one or wake-many threads to wake up
3618
 * @key: is directly passed to the wakeup function
3619 3620 3621
 *
 * 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.
L
Linus Torvalds 已提交
3622
 */
3623
void __wake_up(wait_queue_head_t *q, unsigned int mode,
I
Ingo Molnar 已提交
3624
			int nr_exclusive, void *key)
L
Linus Torvalds 已提交
3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636
{
	unsigned long flags;

	spin_lock_irqsave(&q->lock, flags);
	__wake_up_common(q, mode, nr_exclusive, 0, key);
	spin_unlock_irqrestore(&q->lock, flags);
}
EXPORT_SYMBOL(__wake_up);

/*
 * Same as __wake_up but called with the spinlock in wait_queue_head_t held.
 */
3637
void __wake_up_locked(wait_queue_head_t *q, unsigned int mode, int nr)
L
Linus Torvalds 已提交
3638
{
3639
	__wake_up_common(q, mode, nr, 0, NULL);
L
Linus Torvalds 已提交
3640
}
3641
EXPORT_SYMBOL_GPL(__wake_up_locked);
L
Linus Torvalds 已提交
3642

3643 3644 3645 3646
void __wake_up_locked_key(wait_queue_head_t *q, unsigned int mode, void *key)
{
	__wake_up_common(q, mode, 1, 0, key);
}
3647
EXPORT_SYMBOL_GPL(__wake_up_locked_key);
3648

L
Linus Torvalds 已提交
3649
/**
3650
 * __wake_up_sync_key - wake up threads blocked on a waitqueue.
L
Linus Torvalds 已提交
3651 3652 3653
 * @q: the waitqueue
 * @mode: which threads
 * @nr_exclusive: how many wake-one or wake-many threads to wake up
3654
 * @key: opaque value to be passed to wakeup targets
L
Linus Torvalds 已提交
3655 3656 3657 3658 3659 3660 3661
 *
 * The sync wakeup differs that the waker knows that it will schedule
 * away soon, so while the target thread will be woken up, it will not
 * be migrated to another CPU - ie. the two threads are 'synchronized'
 * with each other. This can prevent needless bouncing between CPUs.
 *
 * On UP it can prevent extra preemption.
3662 3663 3664
 *
 * 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.
L
Linus Torvalds 已提交
3665
 */
3666 3667
void __wake_up_sync_key(wait_queue_head_t *q, unsigned int mode,
			int nr_exclusive, void *key)
L
Linus Torvalds 已提交
3668 3669
{
	unsigned long flags;
P
Peter Zijlstra 已提交
3670
	int wake_flags = WF_SYNC;
L
Linus Torvalds 已提交
3671 3672 3673 3674 3675

	if (unlikely(!q))
		return;

	if (unlikely(!nr_exclusive))
P
Peter Zijlstra 已提交
3676
		wake_flags = 0;
L
Linus Torvalds 已提交
3677 3678

	spin_lock_irqsave(&q->lock, flags);
P
Peter Zijlstra 已提交
3679
	__wake_up_common(q, mode, nr_exclusive, wake_flags, key);
L
Linus Torvalds 已提交
3680 3681
	spin_unlock_irqrestore(&q->lock, flags);
}
3682 3683 3684 3685 3686 3687 3688 3689 3690
EXPORT_SYMBOL_GPL(__wake_up_sync_key);

/*
 * __wake_up_sync - see __wake_up_sync_key()
 */
void __wake_up_sync(wait_queue_head_t *q, unsigned int mode, int nr_exclusive)
{
	__wake_up_sync_key(q, mode, nr_exclusive, NULL);
}
L
Linus Torvalds 已提交
3691 3692
EXPORT_SYMBOL_GPL(__wake_up_sync);	/* For internal use only */

3693 3694 3695 3696 3697 3698 3699 3700
/**
 * complete: - signals a single thread waiting on this completion
 * @x:  holds the state of this particular completion
 *
 * This will wake up a single thread waiting on this completion. Threads will be
 * awakened in the same order in which they were queued.
 *
 * See also complete_all(), wait_for_completion() and related routines.
3701 3702 3703
 *
 * 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.
3704
 */
3705
void complete(struct completion *x)
L
Linus Torvalds 已提交
3706 3707 3708 3709 3710
{
	unsigned long flags;

	spin_lock_irqsave(&x->wait.lock, flags);
	x->done++;
3711
	__wake_up_common(&x->wait, TASK_NORMAL, 1, 0, NULL);
L
Linus Torvalds 已提交
3712 3713 3714 3715
	spin_unlock_irqrestore(&x->wait.lock, flags);
}
EXPORT_SYMBOL(complete);

3716 3717 3718 3719 3720
/**
 * complete_all: - signals all threads waiting on this completion
 * @x:  holds the state of this particular completion
 *
 * This will wake up all threads waiting on this particular completion event.
3721 3722 3723
 *
 * 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.
3724
 */
3725
void complete_all(struct completion *x)
L
Linus Torvalds 已提交
3726 3727 3728 3729 3730
{
	unsigned long flags;

	spin_lock_irqsave(&x->wait.lock, flags);
	x->done += UINT_MAX/2;
3731
	__wake_up_common(&x->wait, TASK_NORMAL, 0, 0, NULL);
L
Linus Torvalds 已提交
3732 3733 3734 3735
	spin_unlock_irqrestore(&x->wait.lock, flags);
}
EXPORT_SYMBOL(complete_all);

3736 3737
static inline long __sched
do_wait_for_common(struct completion *x, long timeout, int state)
L
Linus Torvalds 已提交
3738 3739 3740 3741
{
	if (!x->done) {
		DECLARE_WAITQUEUE(wait, current);

C
Changli Gao 已提交
3742
		__add_wait_queue_tail_exclusive(&x->wait, &wait);
L
Linus Torvalds 已提交
3743
		do {
3744
			if (signal_pending_state(state, current)) {
3745 3746
				timeout = -ERESTARTSYS;
				break;
3747 3748
			}
			__set_current_state(state);
L
Linus Torvalds 已提交
3749 3750 3751
			spin_unlock_irq(&x->wait.lock);
			timeout = schedule_timeout(timeout);
			spin_lock_irq(&x->wait.lock);
3752
		} while (!x->done && timeout);
L
Linus Torvalds 已提交
3753
		__remove_wait_queue(&x->wait, &wait);
3754 3755
		if (!x->done)
			return timeout;
L
Linus Torvalds 已提交
3756 3757
	}
	x->done--;
3758
	return timeout ?: 1;
L
Linus Torvalds 已提交
3759 3760
}

3761 3762
static long __sched
wait_for_common(struct completion *x, long timeout, int state)
L
Linus Torvalds 已提交
3763 3764 3765 3766
{
	might_sleep();

	spin_lock_irq(&x->wait.lock);
3767
	timeout = do_wait_for_common(x, timeout, state);
L
Linus Torvalds 已提交
3768
	spin_unlock_irq(&x->wait.lock);
3769 3770
	return timeout;
}
L
Linus Torvalds 已提交
3771

3772 3773 3774 3775 3776 3777 3778 3779 3780 3781
/**
 * wait_for_completion: - waits for completion of a task
 * @x:  holds the state of this particular completion
 *
 * This waits to be signaled for completion of a specific task. It is NOT
 * interruptible and there is no timeout.
 *
 * See also similar routines (i.e. wait_for_completion_timeout()) with timeout
 * and interrupt capability. Also see complete().
 */
3782
void __sched wait_for_completion(struct completion *x)
3783 3784
{
	wait_for_common(x, MAX_SCHEDULE_TIMEOUT, TASK_UNINTERRUPTIBLE);
L
Linus Torvalds 已提交
3785
}
3786
EXPORT_SYMBOL(wait_for_completion);
L
Linus Torvalds 已提交
3787

3788 3789 3790 3791 3792 3793 3794 3795
/**
 * wait_for_completion_timeout: - waits for completion of a task (w/timeout)
 * @x:  holds the state of this particular completion
 * @timeout:  timeout value in jiffies
 *
 * This waits for either a completion of a specific task to be signaled or for a
 * specified timeout to expire. The timeout is in jiffies. It is not
 * interruptible.
3796 3797 3798
 *
 * The return value is 0 if timed out, and positive (at least 1, or number of
 * jiffies left till timeout) if completed.
3799
 */
3800
unsigned long __sched
3801
wait_for_completion_timeout(struct completion *x, unsigned long timeout)
L
Linus Torvalds 已提交
3802
{
3803
	return wait_for_common(x, timeout, TASK_UNINTERRUPTIBLE);
L
Linus Torvalds 已提交
3804
}
3805
EXPORT_SYMBOL(wait_for_completion_timeout);
L
Linus Torvalds 已提交
3806

3807 3808 3809 3810 3811 3812
/**
 * wait_for_completion_interruptible: - waits for completion of a task (w/intr)
 * @x:  holds the state of this particular completion
 *
 * This waits for completion of a specific task to be signaled. It is
 * interruptible.
3813 3814
 *
 * The return value is -ERESTARTSYS if interrupted, 0 if completed.
3815
 */
3816
int __sched wait_for_completion_interruptible(struct completion *x)
I
Ingo Molnar 已提交
3817
{
3818 3819 3820 3821
	long t = wait_for_common(x, MAX_SCHEDULE_TIMEOUT, TASK_INTERRUPTIBLE);
	if (t == -ERESTARTSYS)
		return t;
	return 0;
I
Ingo Molnar 已提交
3822
}
3823
EXPORT_SYMBOL(wait_for_completion_interruptible);
L
Linus Torvalds 已提交
3824

3825 3826 3827 3828 3829 3830 3831
/**
 * wait_for_completion_interruptible_timeout: - waits for completion (w/(to,intr))
 * @x:  holds the state of this particular completion
 * @timeout:  timeout value in jiffies
 *
 * This waits for either a completion of a specific task to be signaled or for a
 * specified timeout to expire. It is interruptible. The timeout is in jiffies.
3832 3833 3834
 *
 * The return value is -ERESTARTSYS if interrupted, 0 if timed out,
 * positive (at least 1, or number of jiffies left till timeout) if completed.
3835
 */
3836
long __sched
3837 3838
wait_for_completion_interruptible_timeout(struct completion *x,
					  unsigned long timeout)
I
Ingo Molnar 已提交
3839
{
3840
	return wait_for_common(x, timeout, TASK_INTERRUPTIBLE);
I
Ingo Molnar 已提交
3841
}
3842
EXPORT_SYMBOL(wait_for_completion_interruptible_timeout);
L
Linus Torvalds 已提交
3843

3844 3845 3846 3847 3848 3849
/**
 * wait_for_completion_killable: - waits for completion of a task (killable)
 * @x:  holds the state of this particular completion
 *
 * This waits to be signaled for completion of a specific task. It can be
 * interrupted by a kill signal.
3850 3851
 *
 * The return value is -ERESTARTSYS if interrupted, 0 if completed.
3852
 */
M
Matthew Wilcox 已提交
3853 3854 3855 3856 3857 3858 3859 3860 3861
int __sched wait_for_completion_killable(struct completion *x)
{
	long t = wait_for_common(x, MAX_SCHEDULE_TIMEOUT, TASK_KILLABLE);
	if (t == -ERESTARTSYS)
		return t;
	return 0;
}
EXPORT_SYMBOL(wait_for_completion_killable);

3862 3863 3864 3865 3866 3867 3868 3869
/**
 * wait_for_completion_killable_timeout: - waits for completion of a task (w/(to,killable))
 * @x:  holds the state of this particular completion
 * @timeout:  timeout value in jiffies
 *
 * This waits for either a completion of a specific task to be
 * signaled or for a specified timeout to expire. It can be
 * interrupted by a kill signal. The timeout is in jiffies.
3870 3871 3872
 *
 * The return value is -ERESTARTSYS if interrupted, 0 if timed out,
 * positive (at least 1, or number of jiffies left till timeout) if completed.
3873
 */
3874
long __sched
3875 3876 3877 3878 3879 3880 3881
wait_for_completion_killable_timeout(struct completion *x,
				     unsigned long timeout)
{
	return wait_for_common(x, timeout, TASK_KILLABLE);
}
EXPORT_SYMBOL(wait_for_completion_killable_timeout);

3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895
/**
 *	try_wait_for_completion - try to decrement a completion without blocking
 *	@x:	completion structure
 *
 *	Returns: 0 if a decrement cannot be done without blocking
 *		 1 if a decrement succeeded.
 *
 *	If a completion is being used as a counting completion,
 *	attempt to decrement the counter without blocking. This
 *	enables us to avoid waiting if the resource the completion
 *	is protecting is not available.
 */
bool try_wait_for_completion(struct completion *x)
{
3896
	unsigned long flags;
3897 3898
	int ret = 1;

3899
	spin_lock_irqsave(&x->wait.lock, flags);
3900 3901 3902 3903
	if (!x->done)
		ret = 0;
	else
		x->done--;
3904
	spin_unlock_irqrestore(&x->wait.lock, flags);
3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918
	return ret;
}
EXPORT_SYMBOL(try_wait_for_completion);

/**
 *	completion_done - Test to see if a completion has any waiters
 *	@x:	completion structure
 *
 *	Returns: 0 if there are waiters (wait_for_completion() in progress)
 *		 1 if there are no waiters.
 *
 */
bool completion_done(struct completion *x)
{
3919
	unsigned long flags;
3920 3921
	int ret = 1;

3922
	spin_lock_irqsave(&x->wait.lock, flags);
3923 3924
	if (!x->done)
		ret = 0;
3925
	spin_unlock_irqrestore(&x->wait.lock, flags);
3926 3927 3928 3929
	return ret;
}
EXPORT_SYMBOL(completion_done);

3930 3931
static long __sched
sleep_on_common(wait_queue_head_t *q, int state, long timeout)
L
Linus Torvalds 已提交
3932
{
I
Ingo Molnar 已提交
3933 3934 3935 3936
	unsigned long flags;
	wait_queue_t wait;

	init_waitqueue_entry(&wait, current);
L
Linus Torvalds 已提交
3937

3938
	__set_current_state(state);
L
Linus Torvalds 已提交
3939

3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953
	spin_lock_irqsave(&q->lock, flags);
	__add_wait_queue(q, &wait);
	spin_unlock(&q->lock);
	timeout = schedule_timeout(timeout);
	spin_lock_irq(&q->lock);
	__remove_wait_queue(q, &wait);
	spin_unlock_irqrestore(&q->lock, flags);

	return timeout;
}

void __sched interruptible_sleep_on(wait_queue_head_t *q)
{
	sleep_on_common(q, TASK_INTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
L
Linus Torvalds 已提交
3954 3955 3956
}
EXPORT_SYMBOL(interruptible_sleep_on);

I
Ingo Molnar 已提交
3957
long __sched
I
Ingo Molnar 已提交
3958
interruptible_sleep_on_timeout(wait_queue_head_t *q, long timeout)
L
Linus Torvalds 已提交
3959
{
3960
	return sleep_on_common(q, TASK_INTERRUPTIBLE, timeout);
L
Linus Torvalds 已提交
3961 3962 3963
}
EXPORT_SYMBOL(interruptible_sleep_on_timeout);

I
Ingo Molnar 已提交
3964
void __sched sleep_on(wait_queue_head_t *q)
L
Linus Torvalds 已提交
3965
{
3966
	sleep_on_common(q, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
L
Linus Torvalds 已提交
3967 3968 3969
}
EXPORT_SYMBOL(sleep_on);

I
Ingo Molnar 已提交
3970
long __sched sleep_on_timeout(wait_queue_head_t *q, long timeout)
L
Linus Torvalds 已提交
3971
{
3972
	return sleep_on_common(q, TASK_UNINTERRUPTIBLE, timeout);
L
Linus Torvalds 已提交
3973 3974 3975
}
EXPORT_SYMBOL(sleep_on_timeout);

3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987
#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().
 *
 * Used by the rt_mutex code to implement priority inheritance logic.
 */
3988
void rt_mutex_setprio(struct task_struct *p, int prio)
3989
{
3990
	int oldprio, on_rq, running;
3991
	struct rq *rq;
3992
	const struct sched_class *prev_class;
3993 3994 3995

	BUG_ON(prio < 0 || prio > MAX_PRIO);

3996
	rq = __task_rq_lock(p);
3997

3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015
	/*
	 * 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;
	}

4016
	trace_sched_pi_setprio(p, prio);
4017
	oldprio = p->prio;
4018
	prev_class = p->sched_class;
P
Peter Zijlstra 已提交
4019
	on_rq = p->on_rq;
4020
	running = task_current(rq, p);
4021
	if (on_rq)
4022
		dequeue_task(rq, p, 0);
4023 4024
	if (running)
		p->sched_class->put_prev_task(rq, p);
I
Ingo Molnar 已提交
4025 4026 4027 4028 4029 4030

	if (rt_prio(prio))
		p->sched_class = &rt_sched_class;
	else
		p->sched_class = &fair_sched_class;

4031 4032
	p->prio = prio;

4033 4034
	if (running)
		p->sched_class->set_curr_task(rq);
P
Peter Zijlstra 已提交
4035
	if (on_rq)
4036
		enqueue_task(rq, p, oldprio < prio ? ENQUEUE_HEAD : 0);
4037

P
Peter Zijlstra 已提交
4038
	check_class_changed(rq, p, prev_class, oldprio);
4039
out_unlock:
4040
	__task_rq_unlock(rq);
4041 4042
}
#endif
4043
void set_user_nice(struct task_struct *p, long nice)
L
Linus Torvalds 已提交
4044
{
I
Ingo Molnar 已提交
4045
	int old_prio, delta, on_rq;
L
Linus Torvalds 已提交
4046
	unsigned long flags;
4047
	struct rq *rq;
L
Linus Torvalds 已提交
4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059

	if (TASK_NICE(p) == nice || nice < -20 || nice > 19)
		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
I
Ingo Molnar 已提交
4060
	 * SCHED_FIFO/SCHED_RR:
L
Linus Torvalds 已提交
4061
	 */
4062
	if (task_has_rt_policy(p)) {
L
Linus Torvalds 已提交
4063 4064 4065
		p->static_prio = NICE_TO_PRIO(nice);
		goto out_unlock;
	}
P
Peter Zijlstra 已提交
4066
	on_rq = p->on_rq;
4067
	if (on_rq)
4068
		dequeue_task(rq, p, 0);
L
Linus Torvalds 已提交
4069 4070

	p->static_prio = NICE_TO_PRIO(nice);
4071
	set_load_weight(p);
4072 4073 4074
	old_prio = p->prio;
	p->prio = effective_prio(p);
	delta = p->prio - old_prio;
L
Linus Torvalds 已提交
4075

I
Ingo Molnar 已提交
4076
	if (on_rq) {
4077
		enqueue_task(rq, p, 0);
L
Linus Torvalds 已提交
4078
		/*
4079 4080
		 * If the task increased its priority or is running and
		 * lowered its priority, then reschedule its CPU:
L
Linus Torvalds 已提交
4081
		 */
4082
		if (delta < 0 || (delta > 0 && task_running(rq, p)))
L
Linus Torvalds 已提交
4083 4084 4085
			resched_task(rq->curr);
	}
out_unlock:
4086
	task_rq_unlock(rq, p, &flags);
L
Linus Torvalds 已提交
4087 4088 4089
}
EXPORT_SYMBOL(set_user_nice);

M
Matt Mackall 已提交
4090 4091 4092 4093 4094
/*
 * can_nice - check if a task can reduce its nice value
 * @p: task
 * @nice: nice value
 */
4095
int can_nice(const struct task_struct *p, const int nice)
M
Matt Mackall 已提交
4096
{
4097 4098
	/* convert nice value [19,-20] to rlimit style value [1,40] */
	int nice_rlim = 20 - nice;
4099

4100
	return (nice_rlim <= task_rlimit(p, RLIMIT_NICE) ||
M
Matt Mackall 已提交
4101 4102 4103
		capable(CAP_SYS_NICE));
}

L
Linus Torvalds 已提交
4104 4105 4106 4107 4108 4109 4110 4111 4112
#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.
 */
4113
SYSCALL_DEFINE1(nice, int, increment)
L
Linus Torvalds 已提交
4114
{
4115
	long nice, retval;
L
Linus Torvalds 已提交
4116 4117 4118 4119 4120 4121

	/*
	 * 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.
	 */
M
Matt Mackall 已提交
4122 4123
	if (increment < -40)
		increment = -40;
L
Linus Torvalds 已提交
4124 4125 4126
	if (increment > 40)
		increment = 40;

4127
	nice = TASK_NICE(current) + increment;
L
Linus Torvalds 已提交
4128 4129 4130 4131 4132
	if (nice < -20)
		nice = -20;
	if (nice > 19)
		nice = 19;

M
Matt Mackall 已提交
4133 4134 4135
	if (increment < 0 && !can_nice(current, nice))
		return -EPERM;

L
Linus Torvalds 已提交
4136 4137 4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153
	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.
 *
 * This is the priority value as seen by users in /proc.
 * RT tasks are offset by -200. Normal tasks are centered
 * around 0, value goes from -16 to +15.
 */
4154
int task_prio(const struct task_struct *p)
L
Linus Torvalds 已提交
4155 4156 4157 4158 4159 4160 4161 4162
{
	return p->prio - MAX_RT_PRIO;
}

/**
 * task_nice - return the nice value of a given task.
 * @p: the task in question.
 */
4163
int task_nice(const struct task_struct *p)
L
Linus Torvalds 已提交
4164 4165 4166
{
	return TASK_NICE(p);
}
P
Pavel Roskin 已提交
4167
EXPORT_SYMBOL(task_nice);
L
Linus Torvalds 已提交
4168 4169 4170 4171 4172 4173 4174

/**
 * idle_cpu - is a given cpu idle currently?
 * @cpu: the processor in question.
 */
int idle_cpu(int cpu)
{
T
Thomas Gleixner 已提交
4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188
	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
4189 4190 4191 4192 4193 4194
}

/**
 * idle_task - return the idle task for a given cpu.
 * @cpu: the processor in question.
 */
4195
struct task_struct *idle_task(int cpu)
L
Linus Torvalds 已提交
4196 4197 4198 4199 4200 4201 4202 4203
{
	return cpu_rq(cpu)->idle;
}

/**
 * find_process_by_pid - find a process with a matching PID value.
 * @pid: the pid in question.
 */
A
Alexey Dobriyan 已提交
4204
static struct task_struct *find_process_by_pid(pid_t pid)
L
Linus Torvalds 已提交
4205
{
4206
	return pid ? find_task_by_vpid(pid) : current;
L
Linus Torvalds 已提交
4207 4208 4209
}

/* Actually do priority change: must hold rq lock. */
I
Ingo Molnar 已提交
4210 4211
static void
__setscheduler(struct rq *rq, struct task_struct *p, int policy, int prio)
L
Linus Torvalds 已提交
4212 4213 4214
{
	p->policy = policy;
	p->rt_priority = prio;
4215 4216 4217
	p->normal_prio = normal_prio(p);
	/* we are holding p->pi_lock already */
	p->prio = rt_mutex_getprio(p);
4218 4219 4220 4221
	if (rt_prio(p->prio))
		p->sched_class = &rt_sched_class;
	else
		p->sched_class = &fair_sched_class;
4222
	set_load_weight(p);
L
Linus Torvalds 已提交
4223 4224
}

4225 4226 4227 4228 4229 4230 4231 4232 4233 4234
/*
 * 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);
4235 4236
	match = (uid_eq(cred->euid, pcred->euid) ||
		 uid_eq(cred->euid, pcred->uid));
4237 4238 4239 4240
	rcu_read_unlock();
	return match;
}

4241
static int __sched_setscheduler(struct task_struct *p, int policy,
4242
				const struct sched_param *param, bool user)
L
Linus Torvalds 已提交
4243
{
4244
	int retval, oldprio, oldpolicy = -1, on_rq, running;
L
Linus Torvalds 已提交
4245
	unsigned long flags;
4246
	const struct sched_class *prev_class;
4247
	struct rq *rq;
4248
	int reset_on_fork;
L
Linus Torvalds 已提交
4249

4250 4251
	/* may grab non-irq protected spin_locks */
	BUG_ON(in_interrupt());
L
Linus Torvalds 已提交
4252 4253
recheck:
	/* double check policy once rq lock held */
4254 4255
	if (policy < 0) {
		reset_on_fork = p->sched_reset_on_fork;
L
Linus Torvalds 已提交
4256
		policy = oldpolicy = p->policy;
4257 4258 4259 4260 4261 4262 4263 4264 4265 4266
	} else {
		reset_on_fork = !!(policy & SCHED_RESET_ON_FORK);
		policy &= ~SCHED_RESET_ON_FORK;

		if (policy != SCHED_FIFO && policy != SCHED_RR &&
				policy != SCHED_NORMAL && policy != SCHED_BATCH &&
				policy != SCHED_IDLE)
			return -EINVAL;
	}

L
Linus Torvalds 已提交
4267 4268
	/*
	 * Valid priorities for SCHED_FIFO and SCHED_RR are
I
Ingo Molnar 已提交
4269 4270
	 * 1..MAX_USER_RT_PRIO-1, valid priority for SCHED_NORMAL,
	 * SCHED_BATCH and SCHED_IDLE is 0.
L
Linus Torvalds 已提交
4271 4272
	 */
	if (param->sched_priority < 0 ||
I
Ingo Molnar 已提交
4273
	    (p->mm && param->sched_priority > MAX_USER_RT_PRIO-1) ||
4274
	    (!p->mm && param->sched_priority > MAX_RT_PRIO-1))
L
Linus Torvalds 已提交
4275
		return -EINVAL;
4276
	if (rt_policy(policy) != (param->sched_priority != 0))
L
Linus Torvalds 已提交
4277 4278
		return -EINVAL;

4279 4280 4281
	/*
	 * Allow unprivileged RT tasks to decrease priority:
	 */
4282
	if (user && !capable(CAP_SYS_NICE)) {
4283
		if (rt_policy(policy)) {
4284 4285
			unsigned long rlim_rtprio =
					task_rlimit(p, RLIMIT_RTPRIO);
4286 4287 4288 4289 4290 4291 4292 4293 4294 4295

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

			/* can't increase priority */
			if (param->sched_priority > p->rt_priority &&
			    param->sched_priority > rlim_rtprio)
				return -EPERM;
		}
4296

I
Ingo Molnar 已提交
4297
		/*
4298 4299
		 * 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
4300
		 */
4301 4302 4303 4304
		if (p->policy == SCHED_IDLE && policy != SCHED_IDLE) {
			if (!can_nice(p, TASK_NICE(p)))
				return -EPERM;
		}
4305

4306
		/* can't change other user's priorities */
4307
		if (!check_same_owner(p))
4308
			return -EPERM;
4309 4310 4311 4312

		/* Normal users shall not reset the sched_reset_on_fork flag */
		if (p->sched_reset_on_fork && !reset_on_fork)
			return -EPERM;
4313
	}
L
Linus Torvalds 已提交
4314

4315
	if (user) {
4316
		retval = security_task_setscheduler(p);
4317 4318 4319 4320
		if (retval)
			return retval;
	}

4321 4322 4323
	/*
	 * make sure no PI-waiters arrive (or leave) while we are
	 * changing the priority of the task:
4324
	 *
L
Lucas De Marchi 已提交
4325
	 * To be able to change p->policy safely, the appropriate
L
Linus Torvalds 已提交
4326 4327
	 * runqueue lock must be held.
	 */
4328
	rq = task_rq_lock(p, &flags);
4329

4330 4331 4332 4333
	/*
	 * Changing the policy of the stop threads its a very bad idea
	 */
	if (p == rq->stop) {
4334
		task_rq_unlock(rq, p, &flags);
4335 4336 4337
		return -EINVAL;
	}

4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348
	/*
	 * If not changing anything there's no need to proceed further:
	 */
	if (unlikely(policy == p->policy && (!rt_policy(policy) ||
			param->sched_priority == p->rt_priority))) {

		__task_rq_unlock(rq);
		raw_spin_unlock_irqrestore(&p->pi_lock, flags);
		return 0;
	}

4349 4350 4351 4352 4353 4354 4355
#ifdef CONFIG_RT_GROUP_SCHED
	if (user) {
		/*
		 * Do not allow realtime tasks into groups that have no runtime
		 * assigned.
		 */
		if (rt_bandwidth_enabled() && rt_policy(policy) &&
4356 4357
				task_group(p)->rt_bandwidth.rt_runtime == 0 &&
				!task_group_is_autogroup(task_group(p))) {
4358
			task_rq_unlock(rq, p, &flags);
4359 4360 4361 4362 4363
			return -EPERM;
		}
	}
#endif

L
Linus Torvalds 已提交
4364 4365 4366
	/* recheck policy now with rq lock held */
	if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) {
		policy = oldpolicy = -1;
4367
		task_rq_unlock(rq, p, &flags);
L
Linus Torvalds 已提交
4368 4369
		goto recheck;
	}
P
Peter Zijlstra 已提交
4370
	on_rq = p->on_rq;
4371
	running = task_current(rq, p);
4372
	if (on_rq)
4373
		dequeue_task(rq, p, 0);
4374 4375
	if (running)
		p->sched_class->put_prev_task(rq, p);
4376

4377 4378
	p->sched_reset_on_fork = reset_on_fork;

L
Linus Torvalds 已提交
4379
	oldprio = p->prio;
4380
	prev_class = p->sched_class;
I
Ingo Molnar 已提交
4381
	__setscheduler(rq, p, policy, param->sched_priority);
4382

4383 4384
	if (running)
		p->sched_class->set_curr_task(rq);
P
Peter Zijlstra 已提交
4385
	if (on_rq)
4386
		enqueue_task(rq, p, 0);
4387

P
Peter Zijlstra 已提交
4388
	check_class_changed(rq, p, prev_class, oldprio);
4389
	task_rq_unlock(rq, p, &flags);
4390

4391 4392
	rt_mutex_adjust_pi(p);

L
Linus Torvalds 已提交
4393 4394
	return 0;
}
4395 4396 4397 4398 4399 4400 4401 4402 4403 4404

/**
 * 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.
 *
 * NOTE that the task may be already dead.
 */
int sched_setscheduler(struct task_struct *p, int policy,
4405
		       const struct sched_param *param)
4406 4407 4408
{
	return __sched_setscheduler(p, policy, param, true);
}
L
Linus Torvalds 已提交
4409 4410
EXPORT_SYMBOL_GPL(sched_setscheduler);

4411 4412 4413 4414 4415 4416 4417 4418 4419 4420 4421 4422
/**
 * 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.
 */
int sched_setscheduler_nocheck(struct task_struct *p, int policy,
4423
			       const struct sched_param *param)
4424 4425 4426 4427
{
	return __sched_setscheduler(p, policy, param, false);
}

I
Ingo Molnar 已提交
4428 4429
static int
do_sched_setscheduler(pid_t pid, int policy, struct sched_param __user *param)
L
Linus Torvalds 已提交
4430 4431 4432
{
	struct sched_param lparam;
	struct task_struct *p;
4433
	int retval;
L
Linus Torvalds 已提交
4434 4435 4436 4437 4438

	if (!param || pid < 0)
		return -EINVAL;
	if (copy_from_user(&lparam, param, sizeof(struct sched_param)))
		return -EFAULT;
4439 4440 4441

	rcu_read_lock();
	retval = -ESRCH;
L
Linus Torvalds 已提交
4442
	p = find_process_by_pid(pid);
4443 4444 4445
	if (p != NULL)
		retval = sched_setscheduler(p, policy, &lparam);
	rcu_read_unlock();
4446

L
Linus Torvalds 已提交
4447 4448 4449 4450 4451 4452 4453 4454 4455
	return retval;
}

/**
 * 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.
 */
4456 4457
SYSCALL_DEFINE3(sched_setscheduler, pid_t, pid, int, policy,
		struct sched_param __user *, param)
L
Linus Torvalds 已提交
4458
{
4459 4460 4461 4462
	/* negative values for policy are not valid */
	if (policy < 0)
		return -EINVAL;

L
Linus Torvalds 已提交
4463 4464 4465 4466 4467 4468 4469 4470
	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.
 */
4471
SYSCALL_DEFINE2(sched_setparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4472 4473 4474 4475 4476 4477 4478 4479
{
	return do_sched_setscheduler(pid, -1, param);
}

/**
 * sys_sched_getscheduler - get the policy (scheduling class) of a thread
 * @pid: the pid in question.
 */
4480
SYSCALL_DEFINE1(sched_getscheduler, pid_t, pid)
L
Linus Torvalds 已提交
4481
{
4482
	struct task_struct *p;
4483
	int retval;
L
Linus Torvalds 已提交
4484 4485

	if (pid < 0)
4486
		return -EINVAL;
L
Linus Torvalds 已提交
4487 4488

	retval = -ESRCH;
4489
	rcu_read_lock();
L
Linus Torvalds 已提交
4490 4491 4492 4493
	p = find_process_by_pid(pid);
	if (p) {
		retval = security_task_getscheduler(p);
		if (!retval)
4494 4495
			retval = p->policy
				| (p->sched_reset_on_fork ? SCHED_RESET_ON_FORK : 0);
L
Linus Torvalds 已提交
4496
	}
4497
	rcu_read_unlock();
L
Linus Torvalds 已提交
4498 4499 4500 4501
	return retval;
}

/**
4502
 * sys_sched_getparam - get the RT priority of a thread
L
Linus Torvalds 已提交
4503 4504 4505
 * @pid: the pid in question.
 * @param: structure containing the RT priority.
 */
4506
SYSCALL_DEFINE2(sched_getparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4507 4508
{
	struct sched_param lp;
4509
	struct task_struct *p;
4510
	int retval;
L
Linus Torvalds 已提交
4511 4512

	if (!param || pid < 0)
4513
		return -EINVAL;
L
Linus Torvalds 已提交
4514

4515
	rcu_read_lock();
L
Linus Torvalds 已提交
4516 4517 4518 4519 4520 4521 4522 4523 4524 4525
	p = find_process_by_pid(pid);
	retval = -ESRCH;
	if (!p)
		goto out_unlock;

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

	lp.sched_priority = p->rt_priority;
4526
	rcu_read_unlock();
L
Linus Torvalds 已提交
4527 4528 4529 4530 4531 4532 4533 4534 4535

	/*
	 * 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:
4536
	rcu_read_unlock();
L
Linus Torvalds 已提交
4537 4538 4539
	return retval;
}

4540
long sched_setaffinity(pid_t pid, const struct cpumask *in_mask)
L
Linus Torvalds 已提交
4541
{
4542
	cpumask_var_t cpus_allowed, new_mask;
4543 4544
	struct task_struct *p;
	int retval;
L
Linus Torvalds 已提交
4545

4546
	get_online_cpus();
4547
	rcu_read_lock();
L
Linus Torvalds 已提交
4548 4549 4550

	p = find_process_by_pid(pid);
	if (!p) {
4551
		rcu_read_unlock();
4552
		put_online_cpus();
L
Linus Torvalds 已提交
4553 4554 4555
		return -ESRCH;
	}

4556
	/* Prevent p going away */
L
Linus Torvalds 已提交
4557
	get_task_struct(p);
4558
	rcu_read_unlock();
L
Linus Torvalds 已提交
4559

4560 4561 4562 4563 4564 4565 4566 4567
	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
4568
	retval = -EPERM;
4569
	if (!check_same_owner(p) && !ns_capable(task_user_ns(p), CAP_SYS_NICE))
L
Linus Torvalds 已提交
4570 4571
		goto out_unlock;

4572
	retval = security_task_setscheduler(p);
4573 4574 4575
	if (retval)
		goto out_unlock;

4576 4577
	cpuset_cpus_allowed(p, cpus_allowed);
	cpumask_and(new_mask, in_mask, cpus_allowed);
P
Peter Zijlstra 已提交
4578
again:
4579
	retval = set_cpus_allowed_ptr(p, new_mask);
L
Linus Torvalds 已提交
4580

P
Paul Menage 已提交
4581
	if (!retval) {
4582 4583
		cpuset_cpus_allowed(p, cpus_allowed);
		if (!cpumask_subset(new_mask, cpus_allowed)) {
P
Paul Menage 已提交
4584 4585 4586 4587 4588
			/*
			 * We must have raced with a concurrent cpuset
			 * update. Just reset the cpus_allowed to the
			 * cpuset's cpus_allowed
			 */
4589
			cpumask_copy(new_mask, cpus_allowed);
P
Paul Menage 已提交
4590 4591 4592
			goto again;
		}
	}
L
Linus Torvalds 已提交
4593
out_unlock:
4594 4595 4596 4597
	free_cpumask_var(new_mask);
out_free_cpus_allowed:
	free_cpumask_var(cpus_allowed);
out_put_task:
L
Linus Torvalds 已提交
4598
	put_task_struct(p);
4599
	put_online_cpus();
L
Linus Torvalds 已提交
4600 4601 4602 4603
	return retval;
}

static int get_user_cpu_mask(unsigned long __user *user_mask_ptr, unsigned len,
4604
			     struct cpumask *new_mask)
L
Linus Torvalds 已提交
4605
{
4606 4607 4608 4609 4610
	if (len < cpumask_size())
		cpumask_clear(new_mask);
	else if (len > cpumask_size())
		len = cpumask_size();

L
Linus Torvalds 已提交
4611 4612 4613 4614 4615 4616 4617 4618 4619
	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
 */
4620 4621
SYSCALL_DEFINE3(sched_setaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
4622
{
4623
	cpumask_var_t new_mask;
L
Linus Torvalds 已提交
4624 4625
	int retval;

4626 4627
	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
4628

4629 4630 4631 4632 4633
	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
4634 4635
}

4636
long sched_getaffinity(pid_t pid, struct cpumask *mask)
L
Linus Torvalds 已提交
4637
{
4638
	struct task_struct *p;
4639
	unsigned long flags;
L
Linus Torvalds 已提交
4640 4641
	int retval;

4642
	get_online_cpus();
4643
	rcu_read_lock();
L
Linus Torvalds 已提交
4644 4645 4646 4647 4648 4649

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

4650 4651 4652 4653
	retval = security_task_getscheduler(p);
	if (retval)
		goto out_unlock;

4654
	raw_spin_lock_irqsave(&p->pi_lock, flags);
4655
	cpumask_and(mask, &p->cpus_allowed, cpu_online_mask);
4656
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
4657 4658

out_unlock:
4659
	rcu_read_unlock();
4660
	put_online_cpus();
L
Linus Torvalds 已提交
4661

4662
	return retval;
L
Linus Torvalds 已提交
4663 4664 4665 4666 4667 4668 4669 4670
}

/**
 * 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
 */
4671 4672
SYSCALL_DEFINE3(sched_getaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
4673 4674
{
	int ret;
4675
	cpumask_var_t mask;
L
Linus Torvalds 已提交
4676

A
Anton Blanchard 已提交
4677
	if ((len * BITS_PER_BYTE) < nr_cpu_ids)
4678 4679
		return -EINVAL;
	if (len & (sizeof(unsigned long)-1))
L
Linus Torvalds 已提交
4680 4681
		return -EINVAL;

4682 4683
	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
4684

4685 4686
	ret = sched_getaffinity(pid, mask);
	if (ret == 0) {
4687
		size_t retlen = min_t(size_t, len, cpumask_size());
4688 4689

		if (copy_to_user(user_mask_ptr, mask, retlen))
4690 4691
			ret = -EFAULT;
		else
4692
			ret = retlen;
4693 4694
	}
	free_cpumask_var(mask);
L
Linus Torvalds 已提交
4695

4696
	return ret;
L
Linus Torvalds 已提交
4697 4698 4699 4700 4701
}

/**
 * sys_sched_yield - yield the current processor to other threads.
 *
I
Ingo Molnar 已提交
4702 4703
 * This function yields the current CPU to other tasks. If there are no
 * other threads running on this CPU then this function will return.
L
Linus Torvalds 已提交
4704
 */
4705
SYSCALL_DEFINE0(sched_yield)
L
Linus Torvalds 已提交
4706
{
4707
	struct rq *rq = this_rq_lock();
L
Linus Torvalds 已提交
4708

4709
	schedstat_inc(rq, yld_count);
4710
	current->sched_class->yield_task(rq);
L
Linus Torvalds 已提交
4711 4712 4713 4714 4715 4716

	/*
	 * Since we are going to call schedule() anyway, there's
	 * no need to preempt or enable interrupts:
	 */
	__release(rq->lock);
4717
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
4718
	do_raw_spin_unlock(&rq->lock);
4719
	sched_preempt_enable_no_resched();
L
Linus Torvalds 已提交
4720 4721 4722 4723 4724 4725

	schedule();

	return 0;
}

P
Peter Zijlstra 已提交
4726 4727 4728 4729 4730
static inline int should_resched(void)
{
	return need_resched() && !(preempt_count() & PREEMPT_ACTIVE);
}

A
Andrew Morton 已提交
4731
static void __cond_resched(void)
L
Linus Torvalds 已提交
4732
{
4733
	add_preempt_count(PREEMPT_ACTIVE);
4734
	__schedule();
4735
	sub_preempt_count(PREEMPT_ACTIVE);
L
Linus Torvalds 已提交
4736 4737
}

4738
int __sched _cond_resched(void)
L
Linus Torvalds 已提交
4739
{
P
Peter Zijlstra 已提交
4740
	if (should_resched()) {
L
Linus Torvalds 已提交
4741 4742 4743 4744 4745
		__cond_resched();
		return 1;
	}
	return 0;
}
4746
EXPORT_SYMBOL(_cond_resched);
L
Linus Torvalds 已提交
4747 4748

/*
4749
 * __cond_resched_lock() - if a reschedule is pending, drop the given lock,
L
Linus Torvalds 已提交
4750 4751
 * call schedule, and on return reacquire the lock.
 *
I
Ingo Molnar 已提交
4752
 * This works OK both with and without CONFIG_PREEMPT. We do strange low-level
L
Linus Torvalds 已提交
4753 4754 4755
 * operations here to prevent schedule() from being called twice (once via
 * spin_unlock(), once by hand).
 */
4756
int __cond_resched_lock(spinlock_t *lock)
L
Linus Torvalds 已提交
4757
{
P
Peter Zijlstra 已提交
4758
	int resched = should_resched();
J
Jan Kara 已提交
4759 4760
	int ret = 0;

4761 4762
	lockdep_assert_held(lock);

N
Nick Piggin 已提交
4763
	if (spin_needbreak(lock) || resched) {
L
Linus Torvalds 已提交
4764
		spin_unlock(lock);
P
Peter Zijlstra 已提交
4765
		if (resched)
N
Nick Piggin 已提交
4766 4767 4768
			__cond_resched();
		else
			cpu_relax();
J
Jan Kara 已提交
4769
		ret = 1;
L
Linus Torvalds 已提交
4770 4771
		spin_lock(lock);
	}
J
Jan Kara 已提交
4772
	return ret;
L
Linus Torvalds 已提交
4773
}
4774
EXPORT_SYMBOL(__cond_resched_lock);
L
Linus Torvalds 已提交
4775

4776
int __sched __cond_resched_softirq(void)
L
Linus Torvalds 已提交
4777 4778 4779
{
	BUG_ON(!in_softirq());

P
Peter Zijlstra 已提交
4780
	if (should_resched()) {
4781
		local_bh_enable();
L
Linus Torvalds 已提交
4782 4783 4784 4785 4786 4787
		__cond_resched();
		local_bh_disable();
		return 1;
	}
	return 0;
}
4788
EXPORT_SYMBOL(__cond_resched_softirq);
L
Linus Torvalds 已提交
4789 4790 4791 4792

/**
 * yield - yield the current processor to other threads.
 *
P
Peter Zijlstra 已提交
4793 4794 4795 4796 4797 4798 4799 4800 4801 4802 4803 4804 4805 4806 4807 4808 4809 4810
 * 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
4811 4812 4813 4814 4815 4816 4817 4818
 */
void __sched yield(void)
{
	set_current_state(TASK_RUNNING);
	sys_sched_yield();
}
EXPORT_SYMBOL(yield);

4819 4820 4821 4822
/**
 * 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
4823 4824
 * @p: target task
 * @preempt: whether task preemption is allowed or not
4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836 4837 4838 4839 4840 4841 4842 4843 4844 4845 4846 4847 4848 4849 4850 4851 4852 4853 4854 4855 4856 4857 4858
 *
 * 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.
 *
 * Returns true if we indeed boosted the target task.
 */
bool __sched yield_to(struct task_struct *p, bool preempt)
{
	struct task_struct *curr = current;
	struct rq *rq, *p_rq;
	unsigned long flags;
	bool yielded = 0;

	local_irq_save(flags);
	rq = this_rq();

again:
	p_rq = task_rq(p);
	double_rq_lock(rq, p_rq);
	while (task_rq(p) != p_rq) {
		double_rq_unlock(rq, p_rq);
		goto again;
	}

	if (!curr->sched_class->yield_to_task)
		goto out;

	if (curr->sched_class != p->sched_class)
		goto out;

	if (task_running(p_rq, p) || p->state)
		goto out;

	yielded = curr->sched_class->yield_to_task(rq, p, preempt);
4859
	if (yielded) {
4860
		schedstat_inc(rq, yld_count);
4861 4862 4863 4864 4865 4866
		/*
		 * Make p's CPU reschedule; pick_next_entity takes care of
		 * fairness.
		 */
		if (preempt && rq != p_rq)
			resched_task(p_rq->curr);
4867 4868 4869 4870 4871 4872 4873
	} else {
		/*
		 * We might have set it in task_yield_fair(), but are
		 * not going to schedule(), so don't want to skip
		 * the next update.
		 */
		rq->skip_clock_update = 0;
4874
	}
4875 4876 4877 4878 4879 4880 4881 4882 4883 4884 4885 4886

out:
	double_rq_unlock(rq, p_rq);
	local_irq_restore(flags);

	if (yielded)
		schedule();

	return yielded;
}
EXPORT_SYMBOL_GPL(yield_to);

L
Linus Torvalds 已提交
4887
/*
I
Ingo Molnar 已提交
4888
 * This task is about to go to sleep on IO. Increment rq->nr_iowait so
L
Linus Torvalds 已提交
4889 4890 4891 4892
 * that process accounting knows that this is a task in IO wait state.
 */
void __sched io_schedule(void)
{
4893
	struct rq *rq = raw_rq();
L
Linus Torvalds 已提交
4894

4895
	delayacct_blkio_start();
L
Linus Torvalds 已提交
4896
	atomic_inc(&rq->nr_iowait);
4897
	blk_flush_plug(current);
4898
	current->in_iowait = 1;
L
Linus Torvalds 已提交
4899
	schedule();
4900
	current->in_iowait = 0;
L
Linus Torvalds 已提交
4901
	atomic_dec(&rq->nr_iowait);
4902
	delayacct_blkio_end();
L
Linus Torvalds 已提交
4903 4904 4905 4906 4907
}
EXPORT_SYMBOL(io_schedule);

long __sched io_schedule_timeout(long timeout)
{
4908
	struct rq *rq = raw_rq();
L
Linus Torvalds 已提交
4909 4910
	long ret;

4911
	delayacct_blkio_start();
L
Linus Torvalds 已提交
4912
	atomic_inc(&rq->nr_iowait);
4913
	blk_flush_plug(current);
4914
	current->in_iowait = 1;
L
Linus Torvalds 已提交
4915
	ret = schedule_timeout(timeout);
4916
	current->in_iowait = 0;
L
Linus Torvalds 已提交
4917
	atomic_dec(&rq->nr_iowait);
4918
	delayacct_blkio_end();
L
Linus Torvalds 已提交
4919 4920 4921 4922 4923 4924 4925 4926 4927 4928
	return ret;
}

/**
 * sys_sched_get_priority_max - return maximum RT priority.
 * @policy: scheduling class.
 *
 * this syscall returns the maximum rt_priority that can be used
 * by a given scheduling class.
 */
4929
SYSCALL_DEFINE1(sched_get_priority_max, int, policy)
L
Linus Torvalds 已提交
4930 4931 4932 4933 4934 4935 4936 4937 4938
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = MAX_USER_RT_PRIO-1;
		break;
	case SCHED_NORMAL:
4939
	case SCHED_BATCH:
I
Ingo Molnar 已提交
4940
	case SCHED_IDLE:
L
Linus Torvalds 已提交
4941 4942 4943 4944 4945 4946 4947 4948 4949 4950 4951 4952 4953
		ret = 0;
		break;
	}
	return ret;
}

/**
 * sys_sched_get_priority_min - return minimum RT priority.
 * @policy: scheduling class.
 *
 * this syscall returns the minimum rt_priority that can be used
 * by a given scheduling class.
 */
4954
SYSCALL_DEFINE1(sched_get_priority_min, int, policy)
L
Linus Torvalds 已提交
4955 4956 4957 4958 4959 4960 4961 4962 4963
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = 1;
		break;
	case SCHED_NORMAL:
4964
	case SCHED_BATCH:
I
Ingo Molnar 已提交
4965
	case SCHED_IDLE:
L
Linus Torvalds 已提交
4966 4967 4968 4969 4970 4971 4972 4973 4974 4975 4976 4977 4978
		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.
 */
4979
SYSCALL_DEFINE2(sched_rr_get_interval, pid_t, pid,
4980
		struct timespec __user *, interval)
L
Linus Torvalds 已提交
4981
{
4982
	struct task_struct *p;
D
Dmitry Adamushko 已提交
4983
	unsigned int time_slice;
4984 4985
	unsigned long flags;
	struct rq *rq;
4986
	int retval;
L
Linus Torvalds 已提交
4987 4988 4989
	struct timespec t;

	if (pid < 0)
4990
		return -EINVAL;
L
Linus Torvalds 已提交
4991 4992

	retval = -ESRCH;
4993
	rcu_read_lock();
L
Linus Torvalds 已提交
4994 4995 4996 4997 4998 4999 5000 5001
	p = find_process_by_pid(pid);
	if (!p)
		goto out_unlock;

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

5002 5003
	rq = task_rq_lock(p, &flags);
	time_slice = p->sched_class->get_rr_interval(rq, p);
5004
	task_rq_unlock(rq, p, &flags);
D
Dmitry Adamushko 已提交
5005

5006
	rcu_read_unlock();
D
Dmitry Adamushko 已提交
5007
	jiffies_to_timespec(time_slice, &t);
L
Linus Torvalds 已提交
5008 5009
	retval = copy_to_user(interval, &t, sizeof(t)) ? -EFAULT : 0;
	return retval;
5010

L
Linus Torvalds 已提交
5011
out_unlock:
5012
	rcu_read_unlock();
L
Linus Torvalds 已提交
5013 5014 5015
	return retval;
}

5016
static const char stat_nam[] = TASK_STATE_TO_CHAR_STR;
5017

5018
void sched_show_task(struct task_struct *p)
L
Linus Torvalds 已提交
5019 5020
{
	unsigned long free = 0;
5021
	unsigned state;
L
Linus Torvalds 已提交
5022 5023

	state = p->state ? __ffs(p->state) + 1 : 0;
5024
	printk(KERN_INFO "%-15.15s %c", p->comm,
5025
		state < sizeof(stat_nam) - 1 ? stat_nam[state] : '?');
5026
#if BITS_PER_LONG == 32
L
Linus Torvalds 已提交
5027
	if (state == TASK_RUNNING)
P
Peter Zijlstra 已提交
5028
		printk(KERN_CONT " running  ");
L
Linus Torvalds 已提交
5029
	else
P
Peter Zijlstra 已提交
5030
		printk(KERN_CONT " %08lx ", thread_saved_pc(p));
L
Linus Torvalds 已提交
5031 5032
#else
	if (state == TASK_RUNNING)
P
Peter Zijlstra 已提交
5033
		printk(KERN_CONT "  running task    ");
L
Linus Torvalds 已提交
5034
	else
P
Peter Zijlstra 已提交
5035
		printk(KERN_CONT " %016lx ", thread_saved_pc(p));
L
Linus Torvalds 已提交
5036 5037
#endif
#ifdef CONFIG_DEBUG_STACK_USAGE
5038
	free = stack_not_used(p);
L
Linus Torvalds 已提交
5039
#endif
P
Peter Zijlstra 已提交
5040
	printk(KERN_CONT "%5lu %5d %6d 0x%08lx\n", free,
5041
		task_pid_nr(p), task_pid_nr(rcu_dereference(p->real_parent)),
5042
		(unsigned long)task_thread_info(p)->flags);
L
Linus Torvalds 已提交
5043

5044
	show_stack(p, NULL);
L
Linus Torvalds 已提交
5045 5046
}

I
Ingo Molnar 已提交
5047
void show_state_filter(unsigned long state_filter)
L
Linus Torvalds 已提交
5048
{
5049
	struct task_struct *g, *p;
L
Linus Torvalds 已提交
5050

5051
#if BITS_PER_LONG == 32
P
Peter Zijlstra 已提交
5052 5053
	printk(KERN_INFO
		"  task                PC stack   pid father\n");
L
Linus Torvalds 已提交
5054
#else
P
Peter Zijlstra 已提交
5055 5056
	printk(KERN_INFO
		"  task                        PC stack   pid father\n");
L
Linus Torvalds 已提交
5057
#endif
5058
	rcu_read_lock();
L
Linus Torvalds 已提交
5059 5060 5061
	do_each_thread(g, p) {
		/*
		 * reset the NMI-timeout, listing all files on a slow
L
Lucas De Marchi 已提交
5062
		 * console might take a lot of time:
L
Linus Torvalds 已提交
5063 5064
		 */
		touch_nmi_watchdog();
I
Ingo Molnar 已提交
5065
		if (!state_filter || (p->state & state_filter))
5066
			sched_show_task(p);
L
Linus Torvalds 已提交
5067 5068
	} while_each_thread(g, p);

5069 5070
	touch_all_softlockup_watchdogs();

I
Ingo Molnar 已提交
5071 5072 5073
#ifdef CONFIG_SCHED_DEBUG
	sysrq_sched_debug_show();
#endif
5074
	rcu_read_unlock();
I
Ingo Molnar 已提交
5075 5076 5077
	/*
	 * Only show locks if all tasks are dumped:
	 */
5078
	if (!state_filter)
I
Ingo Molnar 已提交
5079
		debug_show_all_locks();
L
Linus Torvalds 已提交
5080 5081
}

I
Ingo Molnar 已提交
5082 5083
void __cpuinit init_idle_bootup_task(struct task_struct *idle)
{
I
Ingo Molnar 已提交
5084
	idle->sched_class = &idle_sched_class;
I
Ingo Molnar 已提交
5085 5086
}

5087 5088 5089 5090 5091 5092 5093 5094
/**
 * 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.
 */
5095
void __cpuinit init_idle(struct task_struct *idle, int cpu)
L
Linus Torvalds 已提交
5096
{
5097
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
5098 5099
	unsigned long flags;

5100
	raw_spin_lock_irqsave(&rq->lock, flags);
5101

I
Ingo Molnar 已提交
5102
	__sched_fork(idle);
5103
	idle->state = TASK_RUNNING;
I
Ingo Molnar 已提交
5104 5105
	idle->se.exec_start = sched_clock();

5106
	do_set_cpus_allowed(idle, cpumask_of(cpu));
5107 5108 5109 5110 5111 5112 5113 5114 5115 5116 5117
	/*
	 * 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
5118
	__set_task_cpu(idle, cpu);
5119
	rcu_read_unlock();
L
Linus Torvalds 已提交
5120 5121

	rq->curr = rq->idle = idle;
P
Peter Zijlstra 已提交
5122 5123
#if defined(CONFIG_SMP)
	idle->on_cpu = 1;
5124
#endif
5125
	raw_spin_unlock_irqrestore(&rq->lock, flags);
L
Linus Torvalds 已提交
5126 5127

	/* Set the preempt count _outside_ the spinlocks! */
A
Al Viro 已提交
5128
	task_thread_info(idle)->preempt_count = 0;
5129

I
Ingo Molnar 已提交
5130 5131 5132 5133
	/*
	 * The idle tasks have their own, simple scheduling class:
	 */
	idle->sched_class = &idle_sched_class;
5134
	ftrace_graph_init_idle_task(idle, cpu);
5135 5136 5137
#if defined(CONFIG_SMP)
	sprintf(idle->comm, "%s/%d", INIT_TASK_COMM, cpu);
#endif
I
Ingo Molnar 已提交
5138 5139
}

L
Linus Torvalds 已提交
5140
#ifdef CONFIG_SMP
5141 5142 5143 5144
void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
{
	if (p->sched_class && p->sched_class->set_cpus_allowed)
		p->sched_class->set_cpus_allowed(p, new_mask);
5145 5146

	cpumask_copy(&p->cpus_allowed, new_mask);
5147
	p->nr_cpus_allowed = cpumask_weight(new_mask);
5148 5149
}

L
Linus Torvalds 已提交
5150 5151 5152
/*
 * This is how migration works:
 *
5153 5154 5155 5156 5157 5158
 * 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
L
Linus Torvalds 已提交
5159
 *    it and puts it into the right queue.
5160 5161
 * 5) stopper completes and stop_one_cpu() returns and the migration
 *    is done.
L
Linus Torvalds 已提交
5162 5163 5164 5165 5166 5167 5168 5169
 */

/*
 * 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
I
Ingo Molnar 已提交
5170
 * task must not exit() & deallocate itself prematurely. The
L
Linus Torvalds 已提交
5171 5172
 * call is not atomic; no spinlocks may be held.
 */
5173
int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
L
Linus Torvalds 已提交
5174 5175
{
	unsigned long flags;
5176
	struct rq *rq;
5177
	unsigned int dest_cpu;
5178
	int ret = 0;
L
Linus Torvalds 已提交
5179 5180

	rq = task_rq_lock(p, &flags);
5181

5182 5183 5184
	if (cpumask_equal(&p->cpus_allowed, new_mask))
		goto out;

5185
	if (!cpumask_intersects(new_mask, cpu_active_mask)) {
L
Linus Torvalds 已提交
5186 5187 5188 5189
		ret = -EINVAL;
		goto out;
	}

5190
	if (unlikely((p->flags & PF_THREAD_BOUND) && p != current)) {
5191 5192 5193 5194
		ret = -EINVAL;
		goto out;
	}

5195
	do_set_cpus_allowed(p, new_mask);
5196

L
Linus Torvalds 已提交
5197
	/* Can the task run on the task's current CPU? If so, we're done */
5198
	if (cpumask_test_cpu(task_cpu(p), new_mask))
L
Linus Torvalds 已提交
5199 5200
		goto out;

5201
	dest_cpu = cpumask_any_and(cpu_active_mask, new_mask);
5202
	if (p->on_rq) {
5203
		struct migration_arg arg = { p, dest_cpu };
L
Linus Torvalds 已提交
5204
		/* Need help from migration thread: drop lock and wait. */
5205
		task_rq_unlock(rq, p, &flags);
5206
		stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg);
L
Linus Torvalds 已提交
5207 5208 5209 5210
		tlb_migrate_finish(p->mm);
		return 0;
	}
out:
5211
	task_rq_unlock(rq, p, &flags);
5212

L
Linus Torvalds 已提交
5213 5214
	return ret;
}
5215
EXPORT_SYMBOL_GPL(set_cpus_allowed_ptr);
L
Linus Torvalds 已提交
5216 5217

/*
I
Ingo Molnar 已提交
5218
 * Move (not current) task off this cpu, onto dest cpu. We're doing
L
Linus Torvalds 已提交
5219 5220 5221 5222 5223 5224
 * 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.
5225 5226
 *
 * Returns non-zero if task was successfully migrated.
L
Linus Torvalds 已提交
5227
 */
5228
static int __migrate_task(struct task_struct *p, int src_cpu, int dest_cpu)
L
Linus Torvalds 已提交
5229
{
5230
	struct rq *rq_dest, *rq_src;
5231
	int ret = 0;
L
Linus Torvalds 已提交
5232

5233
	if (unlikely(!cpu_active(dest_cpu)))
5234
		return ret;
L
Linus Torvalds 已提交
5235 5236 5237 5238

	rq_src = cpu_rq(src_cpu);
	rq_dest = cpu_rq(dest_cpu);

5239
	raw_spin_lock(&p->pi_lock);
L
Linus Torvalds 已提交
5240 5241 5242
	double_rq_lock(rq_src, rq_dest);
	/* Already moved. */
	if (task_cpu(p) != src_cpu)
L
Linus Torvalds 已提交
5243
		goto done;
L
Linus Torvalds 已提交
5244
	/* Affinity changed (again). */
5245
	if (!cpumask_test_cpu(dest_cpu, tsk_cpus_allowed(p)))
L
Linus Torvalds 已提交
5246
		goto fail;
L
Linus Torvalds 已提交
5247

5248 5249 5250 5251
	/*
	 * If we're not on a rq, the next wake-up will ensure we're
	 * placed properly.
	 */
P
Peter Zijlstra 已提交
5252
	if (p->on_rq) {
5253
		dequeue_task(rq_src, p, 0);
5254
		set_task_cpu(p, dest_cpu);
5255
		enqueue_task(rq_dest, p, 0);
5256
		check_preempt_curr(rq_dest, p, 0);
L
Linus Torvalds 已提交
5257
	}
L
Linus Torvalds 已提交
5258
done:
5259
	ret = 1;
L
Linus Torvalds 已提交
5260
fail:
L
Linus Torvalds 已提交
5261
	double_rq_unlock(rq_src, rq_dest);
5262
	raw_spin_unlock(&p->pi_lock);
5263
	return ret;
L
Linus Torvalds 已提交
5264 5265 5266
}

/*
5267 5268 5269
 * 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.
L
Linus Torvalds 已提交
5270
 */
5271
static int migration_cpu_stop(void *data)
L
Linus Torvalds 已提交
5272
{
5273
	struct migration_arg *arg = data;
5274

5275 5276 5277 5278
	/*
	 * The original target cpu might have gone down and we might
	 * be on another cpu but it doesn't matter.
	 */
5279
	local_irq_disable();
5280
	__migrate_task(arg->task, raw_smp_processor_id(), arg->dest_cpu);
5281
	local_irq_enable();
L
Linus Torvalds 已提交
5282
	return 0;
5283 5284
}

L
Linus Torvalds 已提交
5285
#ifdef CONFIG_HOTPLUG_CPU
5286

5287
/*
5288 5289
 * Ensures that the idle task is using init_mm right before its cpu goes
 * offline.
5290
 */
5291
void idle_task_exit(void)
L
Linus Torvalds 已提交
5292
{
5293
	struct mm_struct *mm = current->active_mm;
5294

5295
	BUG_ON(cpu_online(smp_processor_id()));
5296

5297 5298 5299
	if (mm != &init_mm)
		switch_mm(mm, &init_mm, current);
	mmdrop(mm);
L
Linus Torvalds 已提交
5300 5301 5302 5303 5304 5305 5306 5307 5308
}

/*
 * While a dead CPU has no uninterruptible tasks queued at this point,
 * it might still have a nonzero ->nr_uninterruptible counter, because
 * for performance reasons the counter is not stricly tracking tasks to
 * their home CPUs. So we just add the counter to another CPU's counter,
 * to keep the global sum constant after CPU-down:
 */
5309
static void migrate_nr_uninterruptible(struct rq *rq_src)
L
Linus Torvalds 已提交
5310
{
5311
	struct rq *rq_dest = cpu_rq(cpumask_any(cpu_active_mask));
L
Linus Torvalds 已提交
5312 5313 5314 5315 5316

	rq_dest->nr_uninterruptible += rq_src->nr_uninterruptible;
	rq_src->nr_uninterruptible = 0;
}

I
Ingo Molnar 已提交
5317
/*
5318
 * remove the tasks which were accounted by rq from calc_load_tasks.
L
Linus Torvalds 已提交
5319
 */
5320
static void calc_global_load_remove(struct rq *rq)
L
Linus Torvalds 已提交
5321
{
5322 5323
	atomic_long_sub(rq->calc_load_active, &calc_load_tasks);
	rq->calc_load_active = 0;
L
Linus Torvalds 已提交
5324 5325
}

5326
/*
5327 5328 5329 5330 5331 5332
 * 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
5333
 */
5334
static void migrate_tasks(unsigned int dead_cpu)
L
Linus Torvalds 已提交
5335
{
5336
	struct rq *rq = cpu_rq(dead_cpu);
5337 5338
	struct task_struct *next, *stop = rq->stop;
	int dest_cpu;
L
Linus Torvalds 已提交
5339 5340

	/*
5341 5342 5343 5344 5345 5346 5347
	 * 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
5348
	 */
5349
	rq->stop = NULL;
5350

5351 5352 5353
	/* Ensure any throttled groups are reachable by pick_next_task */
	unthrottle_offline_cfs_rqs(rq);

I
Ingo Molnar 已提交
5354
	for ( ; ; ) {
5355 5356 5357 5358 5359
		/*
		 * There's this thread running, bail when that's the only
		 * remaining thread.
		 */
		if (rq->nr_running == 1)
I
Ingo Molnar 已提交
5360
			break;
5361

5362
		next = pick_next_task(rq);
5363
		BUG_ON(!next);
D
Dmitry Adamushko 已提交
5364
		next->sched_class->put_prev_task(rq, next);
5365

5366 5367 5368 5369 5370 5371 5372
		/* Find suitable destination for @next, with force if needed. */
		dest_cpu = select_fallback_rq(dead_cpu, next);
		raw_spin_unlock(&rq->lock);

		__migrate_task(next, dead_cpu, dest_cpu);

		raw_spin_lock(&rq->lock);
L
Linus Torvalds 已提交
5373
	}
5374

5375
	rq->stop = stop;
5376
}
5377

L
Linus Torvalds 已提交
5378 5379
#endif /* CONFIG_HOTPLUG_CPU */

5380 5381 5382
#if defined(CONFIG_SCHED_DEBUG) && defined(CONFIG_SYSCTL)

static struct ctl_table sd_ctl_dir[] = {
5383 5384
	{
		.procname	= "sched_domain",
5385
		.mode		= 0555,
5386
	},
5387
	{}
5388 5389 5390
};

static struct ctl_table sd_ctl_root[] = {
5391 5392
	{
		.procname	= "kernel",
5393
		.mode		= 0555,
5394 5395
		.child		= sd_ctl_dir,
	},
5396
	{}
5397 5398 5399 5400 5401
};

static struct ctl_table *sd_alloc_ctl_entry(int n)
{
	struct ctl_table *entry =
5402
		kcalloc(n, sizeof(struct ctl_table), GFP_KERNEL);
5403 5404 5405 5406

	return entry;
}

5407 5408
static void sd_free_ctl_entry(struct ctl_table **tablep)
{
5409
	struct ctl_table *entry;
5410

5411 5412 5413
	/*
	 * 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
5414
	 * will always be set. In the lowest directory the names are
5415 5416 5417
	 * static strings and all have proc handlers.
	 */
	for (entry = *tablep; entry->mode; entry++) {
5418 5419
		if (entry->child)
			sd_free_ctl_entry(&entry->child);
5420 5421 5422
		if (entry->proc_handler == NULL)
			kfree(entry->procname);
	}
5423 5424 5425 5426 5427

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

5428
static void
5429
set_table_entry(struct ctl_table *entry,
5430
		const char *procname, void *data, int maxlen,
5431
		umode_t mode, proc_handler *proc_handler)
5432 5433 5434 5435 5436 5437 5438 5439 5440 5441 5442
{
	entry->procname = procname;
	entry->data = data;
	entry->maxlen = maxlen;
	entry->mode = mode;
	entry->proc_handler = proc_handler;
}

static struct ctl_table *
sd_alloc_ctl_domain_table(struct sched_domain *sd)
{
5443
	struct ctl_table *table = sd_alloc_ctl_entry(13);
5444

5445 5446 5447
	if (table == NULL)
		return NULL;

5448
	set_table_entry(&table[0], "min_interval", &sd->min_interval,
5449
		sizeof(long), 0644, proc_doulongvec_minmax);
5450
	set_table_entry(&table[1], "max_interval", &sd->max_interval,
5451
		sizeof(long), 0644, proc_doulongvec_minmax);
5452
	set_table_entry(&table[2], "busy_idx", &sd->busy_idx,
5453
		sizeof(int), 0644, proc_dointvec_minmax);
5454
	set_table_entry(&table[3], "idle_idx", &sd->idle_idx,
5455
		sizeof(int), 0644, proc_dointvec_minmax);
5456
	set_table_entry(&table[4], "newidle_idx", &sd->newidle_idx,
5457
		sizeof(int), 0644, proc_dointvec_minmax);
5458
	set_table_entry(&table[5], "wake_idx", &sd->wake_idx,
5459
		sizeof(int), 0644, proc_dointvec_minmax);
5460
	set_table_entry(&table[6], "forkexec_idx", &sd->forkexec_idx,
5461
		sizeof(int), 0644, proc_dointvec_minmax);
5462
	set_table_entry(&table[7], "busy_factor", &sd->busy_factor,
5463
		sizeof(int), 0644, proc_dointvec_minmax);
5464
	set_table_entry(&table[8], "imbalance_pct", &sd->imbalance_pct,
5465
		sizeof(int), 0644, proc_dointvec_minmax);
5466
	set_table_entry(&table[9], "cache_nice_tries",
5467 5468
		&sd->cache_nice_tries,
		sizeof(int), 0644, proc_dointvec_minmax);
5469
	set_table_entry(&table[10], "flags", &sd->flags,
5470
		sizeof(int), 0644, proc_dointvec_minmax);
5471 5472 5473
	set_table_entry(&table[11], "name", sd->name,
		CORENAME_MAX_SIZE, 0444, proc_dostring);
	/* &table[12] is terminator */
5474 5475 5476 5477

	return table;
}

5478
static ctl_table *sd_alloc_ctl_cpu_table(int cpu)
5479 5480 5481 5482 5483 5484 5485 5486 5487
{
	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);
5488 5489
	if (table == NULL)
		return NULL;
5490 5491 5492 5493 5494

	i = 0;
	for_each_domain(cpu, sd) {
		snprintf(buf, 32, "domain%d", i);
		entry->procname = kstrdup(buf, GFP_KERNEL);
5495
		entry->mode = 0555;
5496 5497 5498 5499 5500 5501 5502 5503
		entry->child = sd_alloc_ctl_domain_table(sd);
		entry++;
		i++;
	}
	return table;
}

static struct ctl_table_header *sd_sysctl_header;
5504
static void register_sched_domain_sysctl(void)
5505
{
5506
	int i, cpu_num = num_possible_cpus();
5507 5508 5509
	struct ctl_table *entry = sd_alloc_ctl_entry(cpu_num + 1);
	char buf[32];

5510 5511 5512
	WARN_ON(sd_ctl_dir[0].child);
	sd_ctl_dir[0].child = entry;

5513 5514 5515
	if (entry == NULL)
		return;

5516
	for_each_possible_cpu(i) {
5517 5518
		snprintf(buf, 32, "cpu%d", i);
		entry->procname = kstrdup(buf, GFP_KERNEL);
5519
		entry->mode = 0555;
5520
		entry->child = sd_alloc_ctl_cpu_table(i);
5521
		entry++;
5522
	}
5523 5524

	WARN_ON(sd_sysctl_header);
5525 5526
	sd_sysctl_header = register_sysctl_table(sd_ctl_root);
}
5527

5528
/* may be called multiple times per register */
5529 5530
static void unregister_sched_domain_sysctl(void)
{
5531 5532
	if (sd_sysctl_header)
		unregister_sysctl_table(sd_sysctl_header);
5533
	sd_sysctl_header = NULL;
5534 5535
	if (sd_ctl_dir[0].child)
		sd_free_ctl_entry(&sd_ctl_dir[0].child);
5536
}
5537
#else
5538 5539 5540 5541
static void register_sched_domain_sysctl(void)
{
}
static void unregister_sched_domain_sysctl(void)
5542 5543 5544 5545
{
}
#endif

5546 5547 5548 5549 5550
static void set_rq_online(struct rq *rq)
{
	if (!rq->online) {
		const struct sched_class *class;

5551
		cpumask_set_cpu(rq->cpu, rq->rd->online);
5552 5553 5554 5555 5556 5557 5558 5559 5560 5561 5562 5563 5564 5565 5566 5567 5568 5569 5570
		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);
		}

5571
		cpumask_clear_cpu(rq->cpu, rq->rd->online);
5572 5573 5574 5575
		rq->online = 0;
	}
}

L
Linus Torvalds 已提交
5576 5577 5578 5579
/*
 * migration_call - callback that gets triggered when a CPU is added.
 * Here we can start up the necessary migration thread for the new CPU.
 */
5580 5581
static int __cpuinit
migration_call(struct notifier_block *nfb, unsigned long action, void *hcpu)
L
Linus Torvalds 已提交
5582
{
5583
	int cpu = (long)hcpu;
L
Linus Torvalds 已提交
5584
	unsigned long flags;
5585
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
5586

5587
	switch (action & ~CPU_TASKS_FROZEN) {
5588

L
Linus Torvalds 已提交
5589
	case CPU_UP_PREPARE:
5590
		rq->calc_load_update = calc_load_update;
L
Linus Torvalds 已提交
5591
		break;
5592

L
Linus Torvalds 已提交
5593
	case CPU_ONLINE:
5594
		/* Update our root-domain */
5595
		raw_spin_lock_irqsave(&rq->lock, flags);
5596
		if (rq->rd) {
5597
			BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
5598 5599

			set_rq_online(rq);
5600
		}
5601
		raw_spin_unlock_irqrestore(&rq->lock, flags);
L
Linus Torvalds 已提交
5602
		break;
5603

L
Linus Torvalds 已提交
5604
#ifdef CONFIG_HOTPLUG_CPU
5605
	case CPU_DYING:
5606
		sched_ttwu_pending();
G
Gregory Haskins 已提交
5607
		/* Update our root-domain */
5608
		raw_spin_lock_irqsave(&rq->lock, flags);
G
Gregory Haskins 已提交
5609
		if (rq->rd) {
5610
			BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
5611
			set_rq_offline(rq);
G
Gregory Haskins 已提交
5612
		}
5613 5614
		migrate_tasks(cpu);
		BUG_ON(rq->nr_running != 1); /* the migration thread */
5615
		raw_spin_unlock_irqrestore(&rq->lock, flags);
5616 5617 5618

		migrate_nr_uninterruptible(rq);
		calc_global_load_remove(rq);
G
Gregory Haskins 已提交
5619
		break;
L
Linus Torvalds 已提交
5620 5621
#endif
	}
5622 5623 5624

	update_max_interval();

L
Linus Torvalds 已提交
5625 5626 5627
	return NOTIFY_OK;
}

5628 5629 5630
/*
 * Register at high priority so that task migration (migrate_all_tasks)
 * happens before everything else.  This has to be lower priority than
5631
 * the notifier in the perf_event subsystem, though.
L
Linus Torvalds 已提交
5632
 */
5633
static struct notifier_block __cpuinitdata migration_notifier = {
L
Linus Torvalds 已提交
5634
	.notifier_call = migration_call,
5635
	.priority = CPU_PRI_MIGRATION,
L
Linus Torvalds 已提交
5636 5637
};

5638 5639 5640 5641
static int __cpuinit sched_cpu_active(struct notifier_block *nfb,
				      unsigned long action, void *hcpu)
{
	switch (action & ~CPU_TASKS_FROZEN) {
5642
	case CPU_STARTING:
5643 5644 5645 5646 5647 5648 5649 5650 5651 5652 5653 5654 5655 5656 5657 5658 5659 5660 5661 5662
	case CPU_DOWN_FAILED:
		set_cpu_active((long)hcpu, true);
		return NOTIFY_OK;
	default:
		return NOTIFY_DONE;
	}
}

static int __cpuinit sched_cpu_inactive(struct notifier_block *nfb,
					unsigned long action, void *hcpu)
{
	switch (action & ~CPU_TASKS_FROZEN) {
	case CPU_DOWN_PREPARE:
		set_cpu_active((long)hcpu, false);
		return NOTIFY_OK;
	default:
		return NOTIFY_DONE;
	}
}

5663
static int __init migration_init(void)
L
Linus Torvalds 已提交
5664 5665
{
	void *cpu = (void *)(long)smp_processor_id();
5666
	int err;
5667

5668
	/* Initialize migration for the boot CPU */
5669 5670
	err = migration_call(&migration_notifier, CPU_UP_PREPARE, cpu);
	BUG_ON(err == NOTIFY_BAD);
L
Linus Torvalds 已提交
5671 5672
	migration_call(&migration_notifier, CPU_ONLINE, cpu);
	register_cpu_notifier(&migration_notifier);
5673

5674 5675 5676 5677
	/* Register cpu active notifiers */
	cpu_notifier(sched_cpu_active, CPU_PRI_SCHED_ACTIVE);
	cpu_notifier(sched_cpu_inactive, CPU_PRI_SCHED_INACTIVE);

5678
	return 0;
L
Linus Torvalds 已提交
5679
}
5680
early_initcall(migration_init);
L
Linus Torvalds 已提交
5681 5682 5683
#endif

#ifdef CONFIG_SMP
5684

5685 5686
static cpumask_var_t sched_domains_tmpmask; /* sched_domains_mutex */

5687
#ifdef CONFIG_SCHED_DEBUG
I
Ingo Molnar 已提交
5688

5689
static __read_mostly int sched_debug_enabled;
5690

5691
static int __init sched_debug_setup(char *str)
5692
{
5693
	sched_debug_enabled = 1;
5694 5695 5696

	return 0;
}
5697 5698 5699 5700 5701 5702
early_param("sched_debug", sched_debug_setup);

static inline bool sched_debug(void)
{
	return sched_debug_enabled;
}
5703

5704
static int sched_domain_debug_one(struct sched_domain *sd, int cpu, int level,
5705
				  struct cpumask *groupmask)
L
Linus Torvalds 已提交
5706
{
I
Ingo Molnar 已提交
5707
	struct sched_group *group = sd->groups;
5708
	char str[256];
L
Linus Torvalds 已提交
5709

R
Rusty Russell 已提交
5710
	cpulist_scnprintf(str, sizeof(str), sched_domain_span(sd));
5711
	cpumask_clear(groupmask);
I
Ingo Molnar 已提交
5712 5713 5714 5715

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

	if (!(sd->flags & SD_LOAD_BALANCE)) {
P
Peter Zijlstra 已提交
5716
		printk("does not load-balance\n");
I
Ingo Molnar 已提交
5717
		if (sd->parent)
P
Peter Zijlstra 已提交
5718 5719
			printk(KERN_ERR "ERROR: !SD_LOAD_BALANCE domain"
					" has parent");
I
Ingo Molnar 已提交
5720
		return -1;
N
Nick Piggin 已提交
5721 5722
	}

P
Peter Zijlstra 已提交
5723
	printk(KERN_CONT "span %s level %s\n", str, sd->name);
I
Ingo Molnar 已提交
5724

5725
	if (!cpumask_test_cpu(cpu, sched_domain_span(sd))) {
P
Peter Zijlstra 已提交
5726 5727
		printk(KERN_ERR "ERROR: domain->span does not contain "
				"CPU%d\n", cpu);
I
Ingo Molnar 已提交
5728
	}
5729
	if (!cpumask_test_cpu(cpu, sched_group_cpus(group))) {
P
Peter Zijlstra 已提交
5730 5731
		printk(KERN_ERR "ERROR: domain->groups does not contain"
				" CPU%d\n", cpu);
I
Ingo Molnar 已提交
5732
	}
L
Linus Torvalds 已提交
5733

I
Ingo Molnar 已提交
5734
	printk(KERN_DEBUG "%*s groups:", level + 1, "");
L
Linus Torvalds 已提交
5735
	do {
I
Ingo Molnar 已提交
5736
		if (!group) {
P
Peter Zijlstra 已提交
5737 5738
			printk("\n");
			printk(KERN_ERR "ERROR: group is NULL\n");
L
Linus Torvalds 已提交
5739 5740 5741
			break;
		}

5742 5743 5744 5745 5746 5747
		/*
		 * Even though we initialize ->power to something semi-sane,
		 * we leave power_orig unset. This allows us to detect if
		 * domain iteration is still funny without causing /0 traps.
		 */
		if (!group->sgp->power_orig) {
P
Peter Zijlstra 已提交
5748 5749 5750
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: domain->cpu_power not "
					"set\n");
I
Ingo Molnar 已提交
5751 5752
			break;
		}
L
Linus Torvalds 已提交
5753

5754
		if (!cpumask_weight(sched_group_cpus(group))) {
P
Peter Zijlstra 已提交
5755 5756
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: empty group\n");
I
Ingo Molnar 已提交
5757 5758
			break;
		}
L
Linus Torvalds 已提交
5759

5760 5761
		if (!(sd->flags & SD_OVERLAP) &&
		    cpumask_intersects(groupmask, sched_group_cpus(group))) {
P
Peter Zijlstra 已提交
5762 5763
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: repeated CPUs\n");
I
Ingo Molnar 已提交
5764 5765
			break;
		}
L
Linus Torvalds 已提交
5766

5767
		cpumask_or(groupmask, groupmask, sched_group_cpus(group));
L
Linus Torvalds 已提交
5768

R
Rusty Russell 已提交
5769
		cpulist_scnprintf(str, sizeof(str), sched_group_cpus(group));
5770

P
Peter Zijlstra 已提交
5771
		printk(KERN_CONT " %s", str);
5772
		if (group->sgp->power != SCHED_POWER_SCALE) {
P
Peter Zijlstra 已提交
5773
			printk(KERN_CONT " (cpu_power = %d)",
5774
				group->sgp->power);
5775
		}
L
Linus Torvalds 已提交
5776

I
Ingo Molnar 已提交
5777 5778
		group = group->next;
	} while (group != sd->groups);
P
Peter Zijlstra 已提交
5779
	printk(KERN_CONT "\n");
L
Linus Torvalds 已提交
5780

5781
	if (!cpumask_equal(sched_domain_span(sd), groupmask))
P
Peter Zijlstra 已提交
5782
		printk(KERN_ERR "ERROR: groups don't span domain->span\n");
L
Linus Torvalds 已提交
5783

5784 5785
	if (sd->parent &&
	    !cpumask_subset(groupmask, sched_domain_span(sd->parent)))
P
Peter Zijlstra 已提交
5786 5787
		printk(KERN_ERR "ERROR: parent span is not a superset "
			"of domain->span\n");
I
Ingo Molnar 已提交
5788 5789
	return 0;
}
L
Linus Torvalds 已提交
5790

I
Ingo Molnar 已提交
5791 5792 5793
static void sched_domain_debug(struct sched_domain *sd, int cpu)
{
	int level = 0;
L
Linus Torvalds 已提交
5794

5795
	if (!sched_debug_enabled)
5796 5797
		return;

I
Ingo Molnar 已提交
5798 5799 5800 5801
	if (!sd) {
		printk(KERN_DEBUG "CPU%d attaching NULL sched-domain.\n", cpu);
		return;
	}
L
Linus Torvalds 已提交
5802

I
Ingo Molnar 已提交
5803 5804 5805
	printk(KERN_DEBUG "CPU%d attaching sched-domain:\n", cpu);

	for (;;) {
5806
		if (sched_domain_debug_one(sd, cpu, level, sched_domains_tmpmask))
I
Ingo Molnar 已提交
5807
			break;
L
Linus Torvalds 已提交
5808 5809
		level++;
		sd = sd->parent;
5810
		if (!sd)
I
Ingo Molnar 已提交
5811 5812
			break;
	}
L
Linus Torvalds 已提交
5813
}
5814
#else /* !CONFIG_SCHED_DEBUG */
5815
# define sched_domain_debug(sd, cpu) do { } while (0)
5816 5817 5818 5819
static inline bool sched_debug(void)
{
	return false;
}
5820
#endif /* CONFIG_SCHED_DEBUG */
L
Linus Torvalds 已提交
5821

5822
static int sd_degenerate(struct sched_domain *sd)
5823
{
5824
	if (cpumask_weight(sched_domain_span(sd)) == 1)
5825 5826 5827 5828 5829 5830
		return 1;

	/* Following flags need at least 2 groups */
	if (sd->flags & (SD_LOAD_BALANCE |
			 SD_BALANCE_NEWIDLE |
			 SD_BALANCE_FORK |
5831 5832 5833
			 SD_BALANCE_EXEC |
			 SD_SHARE_CPUPOWER |
			 SD_SHARE_PKG_RESOURCES)) {
5834 5835 5836 5837 5838
		if (sd->groups != sd->groups->next)
			return 0;
	}

	/* Following flags don't use groups */
5839
	if (sd->flags & (SD_WAKE_AFFINE))
5840 5841 5842 5843 5844
		return 0;

	return 1;
}

5845 5846
static int
sd_parent_degenerate(struct sched_domain *sd, struct sched_domain *parent)
5847 5848 5849 5850 5851 5852
{
	unsigned long cflags = sd->flags, pflags = parent->flags;

	if (sd_degenerate(parent))
		return 1;

5853
	if (!cpumask_equal(sched_domain_span(sd), sched_domain_span(parent)))
5854 5855 5856 5857 5858 5859 5860
		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 |
5861 5862 5863
				SD_BALANCE_EXEC |
				SD_SHARE_CPUPOWER |
				SD_SHARE_PKG_RESOURCES);
5864 5865
		if (nr_node_ids == 1)
			pflags &= ~SD_SERIALIZE;
5866 5867 5868 5869 5870 5871 5872
	}
	if (~cflags & pflags)
		return 0;

	return 1;
}

5873
static void free_rootdomain(struct rcu_head *rcu)
5874
{
5875
	struct root_domain *rd = container_of(rcu, struct root_domain, rcu);
5876

5877
	cpupri_cleanup(&rd->cpupri);
5878 5879 5880 5881 5882 5883
	free_cpumask_var(rd->rto_mask);
	free_cpumask_var(rd->online);
	free_cpumask_var(rd->span);
	kfree(rd);
}

G
Gregory Haskins 已提交
5884 5885
static void rq_attach_root(struct rq *rq, struct root_domain *rd)
{
I
Ingo Molnar 已提交
5886
	struct root_domain *old_rd = NULL;
G
Gregory Haskins 已提交
5887 5888
	unsigned long flags;

5889
	raw_spin_lock_irqsave(&rq->lock, flags);
G
Gregory Haskins 已提交
5890 5891

	if (rq->rd) {
I
Ingo Molnar 已提交
5892
		old_rd = rq->rd;
G
Gregory Haskins 已提交
5893

5894
		if (cpumask_test_cpu(rq->cpu, old_rd->online))
5895
			set_rq_offline(rq);
G
Gregory Haskins 已提交
5896

5897
		cpumask_clear_cpu(rq->cpu, old_rd->span);
5898

I
Ingo Molnar 已提交
5899 5900 5901 5902 5903 5904 5905
		/*
		 * If we dont want to free the old_rt yet then
		 * 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
5906 5907 5908 5909 5910
	}

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

5911
	cpumask_set_cpu(rq->cpu, rd->span);
5912
	if (cpumask_test_cpu(rq->cpu, cpu_active_mask))
5913
		set_rq_online(rq);
G
Gregory Haskins 已提交
5914

5915
	raw_spin_unlock_irqrestore(&rq->lock, flags);
I
Ingo Molnar 已提交
5916 5917

	if (old_rd)
5918
		call_rcu_sched(&old_rd->rcu, free_rootdomain);
G
Gregory Haskins 已提交
5919 5920
}

5921
static int init_rootdomain(struct root_domain *rd)
G
Gregory Haskins 已提交
5922 5923 5924
{
	memset(rd, 0, sizeof(*rd));

5925
	if (!alloc_cpumask_var(&rd->span, GFP_KERNEL))
5926
		goto out;
5927
	if (!alloc_cpumask_var(&rd->online, GFP_KERNEL))
5928
		goto free_span;
5929
	if (!alloc_cpumask_var(&rd->rto_mask, GFP_KERNEL))
5930
		goto free_online;
5931

5932
	if (cpupri_init(&rd->cpupri) != 0)
5933
		goto free_rto_mask;
5934
	return 0;
5935

5936 5937
free_rto_mask:
	free_cpumask_var(rd->rto_mask);
5938 5939 5940 5941
free_online:
	free_cpumask_var(rd->online);
free_span:
	free_cpumask_var(rd->span);
5942
out:
5943
	return -ENOMEM;
G
Gregory Haskins 已提交
5944 5945
}

5946 5947 5948 5949 5950 5951
/*
 * 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
5952 5953
static void init_defrootdomain(void)
{
5954
	init_rootdomain(&def_root_domain);
5955

G
Gregory Haskins 已提交
5956 5957 5958
	atomic_set(&def_root_domain.refcount, 1);
}

5959
static struct root_domain *alloc_rootdomain(void)
G
Gregory Haskins 已提交
5960 5961 5962 5963 5964 5965 5966
{
	struct root_domain *rd;

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

5967
	if (init_rootdomain(rd) != 0) {
5968 5969 5970
		kfree(rd);
		return NULL;
	}
G
Gregory Haskins 已提交
5971 5972 5973 5974

	return rd;
}

5975 5976 5977 5978 5979 5980 5981 5982 5983 5984 5985 5986 5987 5988 5989 5990 5991 5992 5993
static void free_sched_groups(struct sched_group *sg, int free_sgp)
{
	struct sched_group *tmp, *first;

	if (!sg)
		return;

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

		if (free_sgp && atomic_dec_and_test(&sg->sgp->ref))
			kfree(sg->sgp);

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

5994 5995 5996
static void free_sched_domain(struct rcu_head *rcu)
{
	struct sched_domain *sd = container_of(rcu, struct sched_domain, rcu);
5997 5998 5999 6000 6001 6002 6003 6004

	/*
	 * 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)) {
6005
		kfree(sd->groups->sgp);
6006
		kfree(sd->groups);
6007
	}
6008 6009 6010 6011 6012 6013 6014 6015 6016 6017 6018 6019 6020 6021
	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);
}

6022 6023 6024 6025 6026
/*
 * 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().
 *
6027 6028 6029 6030 6031
 * Iterate domains and sched_groups downward, assigning CPUs to be
 * select_idle_sibling() hw buddy.  Cross-wiring hw makes bouncing
 * due to random perturbation self canceling, ie sw buddies pull
 * their counterpart to their CPU's hw counterpart.
 *
6032 6033
 * 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
6034
 * two cpus are in the same cache domain, see cpus_share_cache().
6035 6036 6037 6038 6039 6040 6041 6042 6043 6044
 */
DEFINE_PER_CPU(struct sched_domain *, sd_llc);
DEFINE_PER_CPU(int, sd_llc_id);

static void update_top_cache_domain(int cpu)
{
	struct sched_domain *sd;
	int id = cpu;

	sd = highest_flag_domain(cpu, SD_SHARE_PKG_RESOURCES);
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
	if (sd) {
		struct sched_domain *tmp = sd;
		struct sched_group *sg, *prev;
		bool right;

		/*
		 * Traverse to first CPU in group, and count hops
		 * to cpu from there, switching direction on each
		 * hop, never ever pointing the last CPU rightward.
		 */
		do {
			id = cpumask_first(sched_domain_span(tmp));
			prev = sg = tmp->groups;
			right = 1;

			while (cpumask_first(sched_group_cpus(sg)) != id)
				sg = sg->next;

			while (!cpumask_test_cpu(cpu, sched_group_cpus(sg))) {
				prev = sg;
				sg = sg->next;
				right = !right;
			}

			/* A CPU went down, never point back to domain start. */
			if (right && cpumask_first(sched_group_cpus(sg->next)) == id)
				right = false;

			sg = right ? sg->next : prev;
			tmp->idle_buddy = cpumask_first(sched_group_cpus(sg));
		} while ((tmp = tmp->child));

6077
		id = cpumask_first(sched_domain_span(sd));
6078
	}
6079 6080 6081 6082 6083

	rcu_assign_pointer(per_cpu(sd_llc, cpu), sd);
	per_cpu(sd_llc_id, cpu) = id;
}

L
Linus Torvalds 已提交
6084
/*
I
Ingo Molnar 已提交
6085
 * Attach the domain 'sd' to 'cpu' as its base domain. Callers must
L
Linus Torvalds 已提交
6086 6087
 * hold the hotplug lock.
 */
I
Ingo Molnar 已提交
6088 6089
static void
cpu_attach_domain(struct sched_domain *sd, struct root_domain *rd, int cpu)
L
Linus Torvalds 已提交
6090
{
6091
	struct rq *rq = cpu_rq(cpu);
6092 6093 6094
	struct sched_domain *tmp;

	/* Remove the sched domains which do not contribute to scheduling. */
6095
	for (tmp = sd; tmp; ) {
6096 6097 6098
		struct sched_domain *parent = tmp->parent;
		if (!parent)
			break;
6099

6100
		if (sd_parent_degenerate(tmp, parent)) {
6101
			tmp->parent = parent->parent;
6102 6103
			if (parent->parent)
				parent->parent->child = tmp;
6104
			destroy_sched_domain(parent, cpu);
6105 6106
		} else
			tmp = tmp->parent;
6107 6108
	}

6109
	if (sd && sd_degenerate(sd)) {
6110
		tmp = sd;
6111
		sd = sd->parent;
6112
		destroy_sched_domain(tmp, cpu);
6113 6114 6115
		if (sd)
			sd->child = NULL;
	}
L
Linus Torvalds 已提交
6116

6117
	sched_domain_debug(sd, cpu);
L
Linus Torvalds 已提交
6118

G
Gregory Haskins 已提交
6119
	rq_attach_root(rq, rd);
6120
	tmp = rq->sd;
N
Nick Piggin 已提交
6121
	rcu_assign_pointer(rq->sd, sd);
6122
	destroy_sched_domains(tmp, cpu);
6123 6124

	update_top_cache_domain(cpu);
L
Linus Torvalds 已提交
6125 6126 6127
}

/* cpus with isolated domains */
6128
static cpumask_var_t cpu_isolated_map;
L
Linus Torvalds 已提交
6129 6130 6131 6132

/* Setup the mask of cpus configured for isolated domains */
static int __init isolated_cpu_setup(char *str)
{
R
Rusty Russell 已提交
6133
	alloc_bootmem_cpumask_var(&cpu_isolated_map);
R
Rusty Russell 已提交
6134
	cpulist_parse(str, cpu_isolated_map);
L
Linus Torvalds 已提交
6135 6136 6137
	return 1;
}

I
Ingo Molnar 已提交
6138
__setup("isolcpus=", isolated_cpu_setup);
L
Linus Torvalds 已提交
6139

6140 6141 6142 6143 6144
static const struct cpumask *cpu_cpu_mask(int cpu)
{
	return cpumask_of_node(cpu_to_node(cpu));
}

6145 6146 6147
struct sd_data {
	struct sched_domain **__percpu sd;
	struct sched_group **__percpu sg;
6148
	struct sched_group_power **__percpu sgp;
6149 6150
};

6151
struct s_data {
6152
	struct sched_domain ** __percpu sd;
6153 6154 6155
	struct root_domain	*rd;
};

6156 6157
enum s_alloc {
	sa_rootdomain,
6158
	sa_sd,
6159
	sa_sd_storage,
6160 6161 6162
	sa_none,
};

6163 6164 6165
struct sched_domain_topology_level;

typedef struct sched_domain *(*sched_domain_init_f)(struct sched_domain_topology_level *tl, int cpu);
6166 6167
typedef const struct cpumask *(*sched_domain_mask_f)(int cpu);

6168 6169
#define SDTL_OVERLAP	0x01

6170
struct sched_domain_topology_level {
6171 6172
	sched_domain_init_f init;
	sched_domain_mask_f mask;
6173
	int		    flags;
6174
	int		    numa_level;
6175
	struct sd_data      data;
6176 6177
};

P
Peter Zijlstra 已提交
6178 6179 6180 6181 6182 6183 6184 6185 6186 6187 6188 6189 6190 6191 6192 6193 6194 6195 6196 6197 6198 6199 6200 6201 6202 6203 6204 6205 6206 6207 6208 6209 6210 6211 6212 6213 6214 6215
/*
 * 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));
}

6216 6217 6218 6219 6220 6221 6222 6223 6224 6225 6226 6227 6228 6229 6230 6231 6232 6233
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;
	struct sched_domain *child;
	int i;

	cpumask_clear(covered);

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

		if (cpumask_test_cpu(i, covered))
			continue;

P
Peter Zijlstra 已提交
6234 6235 6236 6237 6238 6239
		child = *per_cpu_ptr(sdd->sd, i);

		/* See the comment near build_group_mask(). */
		if (!cpumask_test_cpu(i, sched_domain_span(child)))
			continue;

6240
		sg = kzalloc_node(sizeof(struct sched_group) + cpumask_size(),
6241
				GFP_KERNEL, cpu_to_node(cpu));
6242 6243 6244 6245 6246 6247 6248 6249 6250 6251 6252 6253 6254

		if (!sg)
			goto fail;

		sg_span = sched_group_cpus(sg);
		if (child->child) {
			child = child->child;
			cpumask_copy(sg_span, sched_domain_span(child));
		} else
			cpumask_set_cpu(i, sg_span);

		cpumask_or(covered, covered, sg_span);

P
Peter Zijlstra 已提交
6255
		sg->sgp = *per_cpu_ptr(sdd->sgp, i);
P
Peter Zijlstra 已提交
6256 6257 6258
		if (atomic_inc_return(&sg->sgp->ref) == 1)
			build_group_mask(sd, sg);

6259 6260 6261 6262 6263 6264
		/*
		 * Initialize sgp->power such that even if we mess up the
		 * domains and no possible iteration will get us here, we won't
		 * die on a /0 trap.
		 */
		sg->sgp->power = SCHED_POWER_SCALE * cpumask_weight(sg_span);
6265

P
Peter Zijlstra 已提交
6266 6267 6268 6269 6270
		/*
		 * 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
6271
		if ((!groups && cpumask_test_cpu(cpu, sg_span)) ||
P
Peter Zijlstra 已提交
6272
		    group_balance_cpu(sg) == cpu)
6273 6274 6275 6276 6277 6278 6279 6280 6281 6282 6283 6284 6285 6286 6287 6288 6289 6290 6291
			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;
}

6292
static int get_group(int cpu, struct sd_data *sdd, struct sched_group **sg)
L
Linus Torvalds 已提交
6293
{
6294 6295
	struct sched_domain *sd = *per_cpu_ptr(sdd->sd, cpu);
	struct sched_domain *child = sd->child;
L
Linus Torvalds 已提交
6296

6297 6298
	if (child)
		cpu = cpumask_first(sched_domain_span(child));
6299

6300
	if (sg) {
6301
		*sg = *per_cpu_ptr(sdd->sg, cpu);
6302
		(*sg)->sgp = *per_cpu_ptr(sdd->sgp, cpu);
6303
		atomic_set(&(*sg)->sgp->ref, 1); /* for claim_allocations */
6304
	}
6305 6306

	return cpu;
6307 6308
}

6309
/*
6310 6311 6312
 * 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,
 * and ->cpu_power to 0.
6313 6314
 *
 * Assumes the sched_domain tree is fully constructed
6315
 */
6316 6317
static int
build_sched_groups(struct sched_domain *sd, int cpu)
L
Linus Torvalds 已提交
6318
{
6319 6320 6321
	struct sched_group *first = NULL, *last = NULL;
	struct sd_data *sdd = sd->private;
	const struct cpumask *span = sched_domain_span(sd);
6322
	struct cpumask *covered;
6323
	int i;
6324

6325 6326 6327 6328 6329 6330
	get_group(cpu, sdd, &sd->groups);
	atomic_inc(&sd->groups->ref);

	if (cpu != cpumask_first(sched_domain_span(sd)))
		return 0;

6331 6332 6333
	lockdep_assert_held(&sched_domains_mutex);
	covered = sched_domains_tmpmask;

6334
	cpumask_clear(covered);
6335

6336 6337 6338 6339
	for_each_cpu(i, span) {
		struct sched_group *sg;
		int group = get_group(i, sdd, &sg);
		int j;
6340

6341 6342
		if (cpumask_test_cpu(i, covered))
			continue;
6343

6344
		cpumask_clear(sched_group_cpus(sg));
6345
		sg->sgp->power = 0;
P
Peter Zijlstra 已提交
6346
		cpumask_setall(sched_group_mask(sg));
6347

6348 6349 6350
		for_each_cpu(j, span) {
			if (get_group(j, sdd, NULL) != group)
				continue;
6351

6352 6353 6354
			cpumask_set_cpu(j, covered);
			cpumask_set_cpu(j, sched_group_cpus(sg));
		}
6355

6356 6357 6358 6359 6360 6361 6362
		if (!first)
			first = sg;
		if (last)
			last->next = sg;
		last = sg;
	}
	last->next = first;
6363 6364

	return 0;
6365
}
6366

6367 6368 6369 6370 6371 6372 6373 6374 6375 6376 6377 6378
/*
 * Initialize sched groups cpu_power.
 *
 * cpu_power indicates the capacity of sched group, which is used while
 * distributing the load between different sched groups in a sched domain.
 * Typically cpu_power 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_power will pickup more load compared to the group having
 * less cpu_power.
 */
static void init_sched_groups_power(int cpu, struct sched_domain *sd)
{
6379
	struct sched_group *sg = sd->groups;
6380

6381 6382 6383 6384 6385 6386
	WARN_ON(!sd || !sg);

	do {
		sg->group_weight = cpumask_weight(sched_group_cpus(sg));
		sg = sg->next;
	} while (sg != sd->groups);
6387

P
Peter Zijlstra 已提交
6388
	if (cpu != group_balance_cpu(sg))
6389
		return;
6390

6391
	update_group_power(sd, cpu);
6392
	atomic_set(&sg->sgp->nr_busy_cpus, sg->group_weight);
6393 6394
}

6395 6396 6397
int __weak arch_sd_sibling_asym_packing(void)
{
       return 0*SD_ASYM_PACKING;
6398 6399
}

6400 6401 6402 6403 6404
/*
 * Initializers for schedule domains
 * Non-inlined to reduce accumulated stack pressure in build_sched_domains()
 */

6405 6406 6407 6408 6409 6410
#ifdef CONFIG_SCHED_DEBUG
# define SD_INIT_NAME(sd, type)		sd->name = #type
#else
# define SD_INIT_NAME(sd, type)		do { } while (0)
#endif

6411 6412 6413 6414 6415 6416 6417 6418 6419
#define SD_INIT_FUNC(type)						\
static noinline struct sched_domain *					\
sd_init_##type(struct sched_domain_topology_level *tl, int cpu) 	\
{									\
	struct sched_domain *sd = *per_cpu_ptr(tl->data.sd, cpu);	\
	*sd = SD_##type##_INIT;						\
	SD_INIT_NAME(sd, type);						\
	sd->private = &tl->data;					\
	return sd;							\
6420 6421 6422 6423 6424 6425 6426 6427 6428
}

SD_INIT_FUNC(CPU)
#ifdef CONFIG_SCHED_SMT
 SD_INIT_FUNC(SIBLING)
#endif
#ifdef CONFIG_SCHED_MC
 SD_INIT_FUNC(MC)
#endif
6429 6430 6431
#ifdef CONFIG_SCHED_BOOK
 SD_INIT_FUNC(BOOK)
#endif
6432

6433
static int default_relax_domain_level = -1;
6434
int sched_domain_level_max;
6435 6436 6437

static int __init setup_relax_domain_level(char *str)
{
6438 6439
	if (kstrtoint(str, 0, &default_relax_domain_level))
		pr_warn("Unable to set relax_domain_level\n");
6440

6441 6442 6443 6444 6445 6446 6447 6448 6449 6450 6451 6452 6453 6454 6455 6456 6457 6458
	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 */
6459
		sd->flags &= ~(SD_BALANCE_WAKE|SD_BALANCE_NEWIDLE);
6460 6461
	} else {
		/* turn on idle balance on this domain */
6462
		sd->flags |= (SD_BALANCE_WAKE|SD_BALANCE_NEWIDLE);
6463 6464 6465
	}
}

6466 6467 6468
static void __sdt_free(const struct cpumask *cpu_map);
static int __sdt_alloc(const struct cpumask *cpu_map);

6469 6470 6471 6472 6473
static void __free_domain_allocs(struct s_data *d, enum s_alloc what,
				 const struct cpumask *cpu_map)
{
	switch (what) {
	case sa_rootdomain:
6474 6475
		if (!atomic_read(&d->rd->refcount))
			free_rootdomain(&d->rd->rcu); /* fall through */
6476 6477
	case sa_sd:
		free_percpu(d->sd); /* fall through */
6478
	case sa_sd_storage:
6479
		__sdt_free(cpu_map); /* fall through */
6480 6481 6482 6483
	case sa_none:
		break;
	}
}
6484

6485 6486 6487
static enum s_alloc __visit_domain_allocation_hell(struct s_data *d,
						   const struct cpumask *cpu_map)
{
6488 6489
	memset(d, 0, sizeof(*d));

6490 6491
	if (__sdt_alloc(cpu_map))
		return sa_sd_storage;
6492 6493 6494
	d->sd = alloc_percpu(struct sched_domain *);
	if (!d->sd)
		return sa_sd_storage;
6495
	d->rd = alloc_rootdomain();
6496
	if (!d->rd)
6497
		return sa_sd;
6498 6499
	return sa_rootdomain;
}
G
Gregory Haskins 已提交
6500

6501 6502 6503 6504 6505 6506 6507 6508 6509 6510 6511 6512
/*
 * 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;

6513
	if (atomic_read(&(*per_cpu_ptr(sdd->sg, cpu))->ref))
6514
		*per_cpu_ptr(sdd->sg, cpu) = NULL;
6515 6516

	if (atomic_read(&(*per_cpu_ptr(sdd->sgp, cpu))->ref))
6517
		*per_cpu_ptr(sdd->sgp, cpu) = NULL;
6518 6519
}

6520 6521
#ifdef CONFIG_SCHED_SMT
static const struct cpumask *cpu_smt_mask(int cpu)
6522
{
6523
	return topology_thread_cpumask(cpu);
6524
}
6525
#endif
6526

6527 6528 6529
/*
 * Topology list, bottom-up.
 */
6530
static struct sched_domain_topology_level default_topology[] = {
6531 6532
#ifdef CONFIG_SCHED_SMT
	{ sd_init_SIBLING, cpu_smt_mask, },
6533
#endif
6534
#ifdef CONFIG_SCHED_MC
6535
	{ sd_init_MC, cpu_coregroup_mask, },
6536
#endif
6537 6538 6539 6540
#ifdef CONFIG_SCHED_BOOK
	{ sd_init_BOOK, cpu_book_mask, },
#endif
	{ sd_init_CPU, cpu_cpu_mask, },
6541 6542 6543 6544 6545
	{ NULL, },
};

static struct sched_domain_topology_level *sched_domain_topology = default_topology;

6546 6547 6548 6549 6550 6551 6552 6553 6554
#ifdef CONFIG_NUMA

static int sched_domains_numa_levels;
static int *sched_domains_numa_distance;
static struct cpumask ***sched_domains_numa_masks;
static int sched_domains_curr_level;

static inline int sd_local_flags(int level)
{
6555
	if (sched_domains_numa_distance[level] > RECLAIM_DISTANCE)
6556 6557 6558 6559 6560 6561 6562 6563 6564 6565 6566 6567 6568 6569 6570 6571 6572
		return 0;

	return SD_BALANCE_EXEC | SD_BALANCE_FORK | SD_WAKE_AFFINE;
}

static struct sched_domain *
sd_numa_init(struct sched_domain_topology_level *tl, int cpu)
{
	struct sched_domain *sd = *per_cpu_ptr(tl->data.sd, cpu);
	int level = tl->numa_level;
	int sd_weight = cpumask_weight(
			sched_domains_numa_masks[level][cpu_to_node(cpu)]);

	*sd = (struct sched_domain){
		.min_interval		= sd_weight,
		.max_interval		= 2*sd_weight,
		.busy_factor		= 32,
6573
		.imbalance_pct		= 125,
6574 6575 6576 6577 6578 6579 6580 6581 6582 6583 6584 6585 6586 6587 6588 6589 6590 6591 6592 6593 6594 6595 6596 6597 6598 6599 6600 6601 6602 6603 6604 6605 6606 6607 6608 6609 6610 6611 6612
		.cache_nice_tries	= 2,
		.busy_idx		= 3,
		.idle_idx		= 2,
		.newidle_idx		= 0,
		.wake_idx		= 0,
		.forkexec_idx		= 0,

		.flags			= 1*SD_LOAD_BALANCE
					| 1*SD_BALANCE_NEWIDLE
					| 0*SD_BALANCE_EXEC
					| 0*SD_BALANCE_FORK
					| 0*SD_BALANCE_WAKE
					| 0*SD_WAKE_AFFINE
					| 0*SD_PREFER_LOCAL
					| 0*SD_SHARE_CPUPOWER
					| 0*SD_SHARE_PKG_RESOURCES
					| 1*SD_SERIALIZE
					| 0*SD_PREFER_SIBLING
					| sd_local_flags(level)
					,
		.last_balance		= jiffies,
		.balance_interval	= sd_weight,
	};
	SD_INIT_NAME(sd, NUMA);
	sd->private = &tl->data;

	/*
	 * Ugly hack to pass state to sd_numa_mask()...
	 */
	sched_domains_curr_level = tl->numa_level;

	return sd;
}

static const struct cpumask *sd_numa_mask(int cpu)
{
	return sched_domains_numa_masks[sched_domains_curr_level][cpu_to_node(cpu)];
}

6613 6614 6615 6616 6617 6618 6619 6620 6621 6622 6623 6624 6625 6626 6627 6628 6629 6630 6631 6632 6633 6634 6635 6636 6637 6638 6639 6640 6641 6642 6643 6644 6645 6646 6647 6648
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");
}

static bool find_numa_distance(int distance)
{
	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;
}

6649 6650 6651 6652 6653 6654 6655 6656 6657 6658 6659 6660 6661 6662 6663 6664 6665 6666 6667 6668 6669
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++) {
6670 6671 6672 6673 6674 6675 6676 6677 6678 6679 6680 6681 6682 6683 6684 6685 6686 6687 6688 6689 6690 6691 6692 6693
			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;
6694
		}
6695 6696 6697 6698 6699 6700

		/*
		 * In case of sched_debug() we verify the above assumption.
		 */
		if (!sched_debug())
			break;
6701 6702 6703 6704 6705 6706 6707 6708 6709 6710 6711 6712 6713 6714 6715 6716 6717 6718 6719 6720 6721 6722 6723 6724
	}
	/*
	 * 'level' contains the number of unique distances, excluding the
	 * identity distance node_distance(i,i).
	 *
	 * The sched_domains_nume_distance[] array includes the actual distance
	 * numbers.
	 */

	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++) {
6725
			struct cpumask *mask = kzalloc(cpumask_size(), GFP_KERNEL);
6726 6727 6728 6729 6730 6731
			if (!mask)
				return;

			sched_domains_numa_masks[i][j] = mask;

			for (k = 0; k < nr_node_ids; k++) {
6732
				if (node_distance(j, k) > sched_domains_numa_distance[i])
6733 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
					continue;

				cpumask_or(mask, mask, cpumask_of_node(k));
			}
		}
	}

	tl = kzalloc((ARRAY_SIZE(default_topology) + level) *
			sizeof(struct sched_domain_topology_level), GFP_KERNEL);
	if (!tl)
		return;

	/*
	 * Copy the default topology bits..
	 */
	for (i = 0; default_topology[i].init; i++)
		tl[i] = default_topology[i];

	/*
	 * .. and append 'j' levels of NUMA goodness.
	 */
	for (j = 0; j < level; i++, j++) {
		tl[i] = (struct sched_domain_topology_level){
			.init = sd_numa_init,
			.mask = sd_numa_mask,
			.flags = SDTL_OVERLAP,
			.numa_level = j,
		};
	}

	sched_domain_topology = tl;
}
#else
static inline void sched_init_numa(void)
{
}
#endif /* CONFIG_NUMA */

6771 6772 6773 6774 6775 6776 6777 6778 6779 6780 6781 6782 6783 6784 6785 6786
static int __sdt_alloc(const struct cpumask *cpu_map)
{
	struct sched_domain_topology_level *tl;
	int j;

	for (tl = sched_domain_topology; tl->init; tl++) {
		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;

6787 6788 6789 6790
		sdd->sgp = alloc_percpu(struct sched_group_power *);
		if (!sdd->sgp)
			return -ENOMEM;

6791 6792 6793
		for_each_cpu(j, cpu_map) {
			struct sched_domain *sd;
			struct sched_group *sg;
6794
			struct sched_group_power *sgp;
6795 6796 6797 6798 6799 6800 6801 6802 6803 6804 6805 6806 6807

		       	sd = kzalloc_node(sizeof(struct sched_domain) + cpumask_size(),
					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;

6808 6809
			sg->next = sg;

6810
			*per_cpu_ptr(sdd->sg, j) = sg;
6811

P
Peter Zijlstra 已提交
6812
			sgp = kzalloc_node(sizeof(struct sched_group_power) + cpumask_size(),
6813 6814 6815 6816 6817
					GFP_KERNEL, cpu_to_node(j));
			if (!sgp)
				return -ENOMEM;

			*per_cpu_ptr(sdd->sgp, j) = sgp;
6818 6819 6820 6821 6822 6823 6824 6825 6826 6827 6828 6829 6830 6831 6832
		}
	}

	return 0;
}

static void __sdt_free(const struct cpumask *cpu_map)
{
	struct sched_domain_topology_level *tl;
	int j;

	for (tl = sched_domain_topology; tl->init; tl++) {
		struct sd_data *sdd = &tl->data;

		for_each_cpu(j, cpu_map) {
6833 6834 6835 6836 6837 6838 6839 6840 6841 6842 6843 6844 6845
			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));
			if (sdd->sgp)
				kfree(*per_cpu_ptr(sdd->sgp, j));
6846 6847
		}
		free_percpu(sdd->sd);
6848
		sdd->sd = NULL;
6849
		free_percpu(sdd->sg);
6850
		sdd->sg = NULL;
6851
		free_percpu(sdd->sgp);
6852
		sdd->sgp = NULL;
6853 6854 6855
	}
}

6856 6857
struct sched_domain *build_sched_domain(struct sched_domain_topology_level *tl,
		struct s_data *d, const struct cpumask *cpu_map,
6858
		struct sched_domain_attr *attr, struct sched_domain *child,
6859 6860
		int cpu)
{
6861
	struct sched_domain *sd = tl->init(tl, cpu);
6862
	if (!sd)
6863
		return child;
6864 6865

	cpumask_and(sched_domain_span(sd), cpu_map, tl->mask(cpu));
6866 6867 6868
	if (child) {
		sd->level = child->level + 1;
		sched_domain_level_max = max(sched_domain_level_max, sd->level);
6869
		child->parent = sd;
6870
	}
6871
	sd->child = child;
6872
	set_domain_attribute(sd, attr);
6873 6874 6875 6876

	return sd;
}

6877 6878 6879 6880
/*
 * Build sched domains for a given set of cpus and attach the sched domains
 * to the individual cpus
 */
6881 6882
static int build_sched_domains(const struct cpumask *cpu_map,
			       struct sched_domain_attr *attr)
6883 6884
{
	enum s_alloc alloc_state = sa_none;
6885
	struct sched_domain *sd;
6886
	struct s_data d;
6887
	int i, ret = -ENOMEM;
6888

6889 6890 6891
	alloc_state = __visit_domain_allocation_hell(&d, cpu_map);
	if (alloc_state != sa_rootdomain)
		goto error;
6892

6893
	/* Set up domains for cpus specified by the cpu_map. */
6894
	for_each_cpu(i, cpu_map) {
6895 6896
		struct sched_domain_topology_level *tl;

6897
		sd = NULL;
6898
		for (tl = sched_domain_topology; tl->init; tl++) {
6899
			sd = build_sched_domain(tl, &d, cpu_map, attr, sd, i);
6900 6901
			if (tl->flags & SDTL_OVERLAP || sched_feat(FORCE_SD_OVERLAP))
				sd->flags |= SD_OVERLAP;
6902 6903
			if (cpumask_equal(cpu_map, sched_domain_span(sd)))
				break;
6904
		}
6905

6906 6907 6908
		while (sd->child)
			sd = sd->child;

6909
		*per_cpu_ptr(d.sd, i) = sd;
6910 6911 6912 6913 6914 6915
	}

	/* 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));
6916 6917 6918 6919 6920 6921 6922
			if (sd->flags & SD_OVERLAP) {
				if (build_overlap_sched_groups(sd, i))
					goto error;
			} else {
				if (build_sched_groups(sd, i))
					goto error;
			}
6923
		}
6924
	}
6925

L
Linus Torvalds 已提交
6926
	/* Calculate CPU power for physical packages and nodes */
6927 6928 6929
	for (i = nr_cpumask_bits-1; i >= 0; i--) {
		if (!cpumask_test_cpu(i, cpu_map))
			continue;
6930

6931 6932
		for (sd = *per_cpu_ptr(d.sd, i); sd; sd = sd->parent) {
			claim_allocations(i, sd);
6933
			init_sched_groups_power(i, sd);
6934
		}
6935
	}
6936

L
Linus Torvalds 已提交
6937
	/* Attach the domains */
6938
	rcu_read_lock();
6939
	for_each_cpu(i, cpu_map) {
6940
		sd = *per_cpu_ptr(d.sd, i);
6941
		cpu_attach_domain(sd, d.rd, i);
L
Linus Torvalds 已提交
6942
	}
6943
	rcu_read_unlock();
6944

6945
	ret = 0;
6946
error:
6947
	__free_domain_allocs(&d, alloc_state, cpu_map);
6948
	return ret;
L
Linus Torvalds 已提交
6949
}
P
Paul Jackson 已提交
6950

6951
static cpumask_var_t *doms_cur;	/* current sched domains */
P
Paul Jackson 已提交
6952
static int ndoms_cur;		/* number of sched domains in 'doms_cur' */
I
Ingo Molnar 已提交
6953 6954
static struct sched_domain_attr *dattr_cur;
				/* attribues of custom domains in 'doms_cur' */
P
Paul Jackson 已提交
6955 6956 6957

/*
 * Special case: If a kmalloc of a doms_cur partition (array of
6958 6959
 * cpumask) fails, then fallback to a single sched domain,
 * as determined by the single cpumask fallback_doms.
P
Paul Jackson 已提交
6960
 */
6961
static cpumask_var_t fallback_doms;
P
Paul Jackson 已提交
6962

6963 6964 6965 6966 6967 6968
/*
 * 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.
 */
int __attribute__((weak)) arch_update_cpu_topology(void)
6969
{
6970
	return 0;
6971 6972
}

6973 6974 6975 6976 6977 6978 6979 6980 6981 6982 6983 6984 6985 6986 6987 6988 6989 6990 6991 6992 6993 6994 6995 6996 6997
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);
}

6998
/*
I
Ingo Molnar 已提交
6999
 * Set up scheduler domains and groups. Callers must hold the hotplug lock.
P
Paul Jackson 已提交
7000 7001
 * For now this just excludes isolated cpus, but could be used to
 * exclude other special cases in the future.
7002
 */
7003
static int init_sched_domains(const struct cpumask *cpu_map)
7004
{
7005 7006
	int err;

7007
	arch_update_cpu_topology();
P
Paul Jackson 已提交
7008
	ndoms_cur = 1;
7009
	doms_cur = alloc_sched_domains(ndoms_cur);
P
Paul Jackson 已提交
7010
	if (!doms_cur)
7011 7012
		doms_cur = &fallback_doms;
	cpumask_andnot(doms_cur[0], cpu_map, cpu_isolated_map);
7013
	err = build_sched_domains(doms_cur[0], NULL);
7014
	register_sched_domain_sysctl();
7015 7016

	return err;
7017 7018 7019 7020 7021 7022
}

/*
 * Detach sched domains from a group of cpus specified in cpu_map
 * These cpus will now be attached to the NULL domain
 */
7023
static void detach_destroy_domains(const struct cpumask *cpu_map)
7024 7025 7026
{
	int i;

7027
	rcu_read_lock();
7028
	for_each_cpu(i, cpu_map)
G
Gregory Haskins 已提交
7029
		cpu_attach_domain(NULL, &def_root_domain, i);
7030
	rcu_read_unlock();
7031 7032
}

7033 7034 7035 7036 7037 7038 7039 7040 7041 7042 7043 7044 7045 7046 7047 7048
/* 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
7049 7050
/*
 * Partition sched domains as specified by the 'ndoms_new'
I
Ingo Molnar 已提交
7051
 * cpumasks in the array doms_new[] of cpumasks. This compares
P
Paul Jackson 已提交
7052 7053 7054
 * doms_new[] to the current sched domain partitioning, doms_cur[].
 * It destroys each deleted domain and builds each new domain.
 *
7055
 * 'doms_new' is an array of cpumask_var_t's of length 'ndoms_new'.
I
Ingo Molnar 已提交
7056 7057 7058
 * 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
7059 7060 7061
 * current 'doms_cur' domains and in the new 'doms_new', we can leave
 * it as it is.
 *
7062 7063 7064 7065 7066 7067
 * 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
7068
 *
7069
 * If doms_new == NULL it will be replaced with cpu_online_mask.
7070 7071
 * ndoms_new == 0 is a special case for destroying existing domains,
 * and it will not create the default domain.
7072
 *
P
Paul Jackson 已提交
7073 7074
 * Call with hotplug lock held
 */
7075
void partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
7076
			     struct sched_domain_attr *dattr_new)
P
Paul Jackson 已提交
7077
{
7078
	int i, j, n;
7079
	int new_topology;
P
Paul Jackson 已提交
7080

7081
	mutex_lock(&sched_domains_mutex);
7082

7083 7084 7085
	/* always unregister in case we don't destroy any domains */
	unregister_sched_domain_sysctl();

7086 7087 7088
	/* Let architecture update cpu core mappings. */
	new_topology = arch_update_cpu_topology();

7089
	n = doms_new ? ndoms_new : 0;
P
Paul Jackson 已提交
7090 7091 7092

	/* Destroy deleted domains */
	for (i = 0; i < ndoms_cur; i++) {
7093
		for (j = 0; j < n && !new_topology; j++) {
7094
			if (cpumask_equal(doms_cur[i], doms_new[j])
7095
			    && dattrs_equal(dattr_cur, i, dattr_new, j))
P
Paul Jackson 已提交
7096 7097 7098
				goto match1;
		}
		/* no match - a current sched domain not in new doms_new[] */
7099
		detach_destroy_domains(doms_cur[i]);
P
Paul Jackson 已提交
7100 7101 7102 7103
match1:
		;
	}

7104 7105
	if (doms_new == NULL) {
		ndoms_cur = 0;
7106
		doms_new = &fallback_doms;
7107
		cpumask_andnot(doms_new[0], cpu_active_mask, cpu_isolated_map);
7108
		WARN_ON_ONCE(dattr_new);
7109 7110
	}

P
Paul Jackson 已提交
7111 7112
	/* Build new domains */
	for (i = 0; i < ndoms_new; i++) {
7113
		for (j = 0; j < ndoms_cur && !new_topology; j++) {
7114
			if (cpumask_equal(doms_new[i], doms_cur[j])
7115
			    && dattrs_equal(dattr_new, i, dattr_cur, j))
P
Paul Jackson 已提交
7116 7117 7118
				goto match2;
		}
		/* no match - add a new doms_new */
7119
		build_sched_domains(doms_new[i], dattr_new ? dattr_new + i : NULL);
P
Paul Jackson 已提交
7120 7121 7122 7123 7124
match2:
		;
	}

	/* Remember the new sched domains */
7125 7126
	if (doms_cur != &fallback_doms)
		free_sched_domains(doms_cur, ndoms_cur);
7127
	kfree(dattr_cur);	/* kfree(NULL) is safe */
P
Paul Jackson 已提交
7128
	doms_cur = doms_new;
7129
	dattr_cur = dattr_new;
P
Paul Jackson 已提交
7130
	ndoms_cur = ndoms_new;
7131 7132

	register_sched_domain_sysctl();
7133

7134
	mutex_unlock(&sched_domains_mutex);
P
Paul Jackson 已提交
7135 7136
}

7137 7138
static int num_cpus_frozen;	/* used to mark begin/end of suspend/resume */

L
Linus Torvalds 已提交
7139
/*
7140 7141 7142
 * Update cpusets according to cpu_active mask.  If cpusets are
 * disabled, cpuset_update_active_cpus() becomes a simple wrapper
 * around partition_sched_domains().
7143 7144 7145
 *
 * 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
7146
 */
7147 7148
static int cpuset_cpu_active(struct notifier_block *nfb, unsigned long action,
			     void *hcpu)
7149
{
7150 7151 7152 7153 7154 7155 7156 7157 7158 7159 7160 7161 7162 7163 7164 7165 7166 7167 7168 7169 7170 7171
	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.
		 */

7172
	case CPU_ONLINE:
7173
	case CPU_DOWN_FAILED:
7174
		cpuset_update_active_cpus(true);
7175
		break;
7176 7177 7178
	default:
		return NOTIFY_DONE;
	}
7179
	return NOTIFY_OK;
7180
}
7181

7182 7183
static int cpuset_cpu_inactive(struct notifier_block *nfb, unsigned long action,
			       void *hcpu)
7184
{
7185
	switch (action) {
7186
	case CPU_DOWN_PREPARE:
7187
		cpuset_update_active_cpus(false);
7188 7189 7190 7191 7192
		break;
	case CPU_DOWN_PREPARE_FROZEN:
		num_cpus_frozen++;
		partition_sched_domains(1, NULL, NULL);
		break;
7193 7194 7195
	default:
		return NOTIFY_DONE;
	}
7196
	return NOTIFY_OK;
7197 7198
}

L
Linus Torvalds 已提交
7199 7200
void __init sched_init_smp(void)
{
7201 7202 7203
	cpumask_var_t non_isolated_cpus;

	alloc_cpumask_var(&non_isolated_cpus, GFP_KERNEL);
7204
	alloc_cpumask_var(&fallback_doms, GFP_KERNEL);
7205

7206 7207
	sched_init_numa();

7208
	get_online_cpus();
7209
	mutex_lock(&sched_domains_mutex);
7210
	init_sched_domains(cpu_active_mask);
7211 7212 7213
	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);
7214
	mutex_unlock(&sched_domains_mutex);
7215
	put_online_cpus();
7216

7217 7218
	hotcpu_notifier(cpuset_cpu_active, CPU_PRI_CPUSET_ACTIVE);
	hotcpu_notifier(cpuset_cpu_inactive, CPU_PRI_CPUSET_INACTIVE);
7219 7220 7221 7222

	/* RT runtime code needs to handle some hotplug events */
	hotcpu_notifier(update_runtime, 0);

7223
	init_hrtick();
7224 7225

	/* Move init over to a non-isolated CPU */
7226
	if (set_cpus_allowed_ptr(current, non_isolated_cpus) < 0)
7227
		BUG();
I
Ingo Molnar 已提交
7228
	sched_init_granularity();
7229
	free_cpumask_var(non_isolated_cpus);
7230

7231
	init_sched_rt_class();
L
Linus Torvalds 已提交
7232 7233 7234 7235
}
#else
void __init sched_init_smp(void)
{
I
Ingo Molnar 已提交
7236
	sched_init_granularity();
L
Linus Torvalds 已提交
7237 7238 7239
}
#endif /* CONFIG_SMP */

7240 7241
const_debug unsigned int sysctl_timer_migration = 1;

L
Linus Torvalds 已提交
7242 7243 7244 7245 7246 7247 7248
int in_sched_functions(unsigned long addr)
{
	return in_lock_functions(addr) ||
		(addr >= (unsigned long)__sched_text_start
		&& addr < (unsigned long)__sched_text_end);
}

7249 7250
#ifdef CONFIG_CGROUP_SCHED
struct task_group root_task_group;
7251
#endif
P
Peter Zijlstra 已提交
7252

7253
DECLARE_PER_CPU(cpumask_var_t, load_balance_tmpmask);
P
Peter Zijlstra 已提交
7254

L
Linus Torvalds 已提交
7255 7256
void __init sched_init(void)
{
I
Ingo Molnar 已提交
7257
	int i, j;
7258 7259 7260 7261 7262 7263 7264
	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 **);
7265
#endif
7266
#ifdef CONFIG_CPUMASK_OFFSTACK
7267
	alloc_size += num_possible_cpus() * cpumask_size();
7268 7269
#endif
	if (alloc_size) {
7270
		ptr = (unsigned long)kzalloc(alloc_size, GFP_NOWAIT);
7271 7272

#ifdef CONFIG_FAIR_GROUP_SCHED
7273
		root_task_group.se = (struct sched_entity **)ptr;
7274 7275
		ptr += nr_cpu_ids * sizeof(void **);

7276
		root_task_group.cfs_rq = (struct cfs_rq **)ptr;
7277
		ptr += nr_cpu_ids * sizeof(void **);
7278

7279
#endif /* CONFIG_FAIR_GROUP_SCHED */
7280
#ifdef CONFIG_RT_GROUP_SCHED
7281
		root_task_group.rt_se = (struct sched_rt_entity **)ptr;
7282 7283
		ptr += nr_cpu_ids * sizeof(void **);

7284
		root_task_group.rt_rq = (struct rt_rq **)ptr;
7285 7286
		ptr += nr_cpu_ids * sizeof(void **);

7287
#endif /* CONFIG_RT_GROUP_SCHED */
7288 7289 7290 7291 7292 7293
#ifdef CONFIG_CPUMASK_OFFSTACK
		for_each_possible_cpu(i) {
			per_cpu(load_balance_tmpmask, i) = (void *)ptr;
			ptr += cpumask_size();
		}
#endif /* CONFIG_CPUMASK_OFFSTACK */
7294
	}
I
Ingo Molnar 已提交
7295

G
Gregory Haskins 已提交
7296 7297 7298 7299
#ifdef CONFIG_SMP
	init_defrootdomain();
#endif

7300 7301 7302 7303
	init_rt_bandwidth(&def_rt_bandwidth,
			global_rt_period(), global_rt_runtime());

#ifdef CONFIG_RT_GROUP_SCHED
7304
	init_rt_bandwidth(&root_task_group.rt_bandwidth,
7305
			global_rt_period(), global_rt_runtime());
7306
#endif /* CONFIG_RT_GROUP_SCHED */
7307

D
Dhaval Giani 已提交
7308
#ifdef CONFIG_CGROUP_SCHED
7309 7310
	list_add(&root_task_group.list, &task_groups);
	INIT_LIST_HEAD(&root_task_group.children);
7311
	INIT_LIST_HEAD(&root_task_group.siblings);
7312
	autogroup_init(&init_task);
7313

D
Dhaval Giani 已提交
7314
#endif /* CONFIG_CGROUP_SCHED */
P
Peter Zijlstra 已提交
7315

7316 7317 7318 7319 7320 7321
#ifdef CONFIG_CGROUP_CPUACCT
	root_cpuacct.cpustat = &kernel_cpustat;
	root_cpuacct.cpuusage = alloc_percpu(u64);
	/* Too early, not expected to fail */
	BUG_ON(!root_cpuacct.cpuusage);
#endif
7322
	for_each_possible_cpu(i) {
7323
		struct rq *rq;
L
Linus Torvalds 已提交
7324 7325

		rq = cpu_rq(i);
7326
		raw_spin_lock_init(&rq->lock);
N
Nick Piggin 已提交
7327
		rq->nr_running = 0;
7328 7329
		rq->calc_load_active = 0;
		rq->calc_load_update = jiffies + LOAD_FREQ;
7330
		init_cfs_rq(&rq->cfs);
P
Peter Zijlstra 已提交
7331
		init_rt_rq(&rq->rt, rq);
I
Ingo Molnar 已提交
7332
#ifdef CONFIG_FAIR_GROUP_SCHED
7333
		root_task_group.shares = ROOT_TASK_GROUP_LOAD;
P
Peter Zijlstra 已提交
7334
		INIT_LIST_HEAD(&rq->leaf_cfs_rq_list);
D
Dhaval Giani 已提交
7335
		/*
7336
		 * How much cpu bandwidth does root_task_group get?
D
Dhaval Giani 已提交
7337 7338 7339 7340
		 *
		 * 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
7341
		 * root_task_group and its child task-groups in a fair manner,
D
Dhaval Giani 已提交
7342 7343 7344
		 * based on each entity's (task or task-group's) weight
		 * (se->load.weight).
		 *
7345
		 * In other words, if root_task_group has 10 tasks of weight
D
Dhaval Giani 已提交
7346 7347 7348
		 * 1024) and two child groups A0 and A1 (of weight 1024 each),
		 * then A0's share of the cpu resource is:
		 *
7349
		 *	A0's bandwidth = 1024 / (10*1024 + 1024 + 1024) = 8.33%
D
Dhaval Giani 已提交
7350
		 *
7351 7352
		 * 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
7353
		 */
7354
		init_cfs_bandwidth(&root_task_group.cfs_bandwidth);
7355
		init_tg_cfs_entry(&root_task_group, &rq->cfs, NULL, i, NULL);
D
Dhaval Giani 已提交
7356 7357 7358
#endif /* CONFIG_FAIR_GROUP_SCHED */

		rq->rt.rt_runtime = def_rt_bandwidth.rt_runtime;
7359
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
7360
		INIT_LIST_HEAD(&rq->leaf_rt_rq_list);
7361
		init_tg_rt_entry(&root_task_group, &rq->rt, NULL, i, NULL);
I
Ingo Molnar 已提交
7362
#endif
L
Linus Torvalds 已提交
7363

I
Ingo Molnar 已提交
7364 7365
		for (j = 0; j < CPU_LOAD_IDX_MAX; j++)
			rq->cpu_load[j] = 0;
7366 7367 7368

		rq->last_load_update_tick = jiffies;

L
Linus Torvalds 已提交
7369
#ifdef CONFIG_SMP
N
Nick Piggin 已提交
7370
		rq->sd = NULL;
G
Gregory Haskins 已提交
7371
		rq->rd = NULL;
7372
		rq->cpu_power = SCHED_POWER_SCALE;
7373
		rq->post_schedule = 0;
L
Linus Torvalds 已提交
7374
		rq->active_balance = 0;
I
Ingo Molnar 已提交
7375
		rq->next_balance = jiffies;
L
Linus Torvalds 已提交
7376
		rq->push_cpu = 0;
7377
		rq->cpu = i;
7378
		rq->online = 0;
7379 7380
		rq->idle_stamp = 0;
		rq->avg_idle = 2*sysctl_sched_migration_cost;
7381 7382 7383

		INIT_LIST_HEAD(&rq->cfs_tasks);

7384
		rq_attach_root(rq, &def_root_domain);
7385
#ifdef CONFIG_NO_HZ
7386
		rq->nohz_flags = 0;
7387
#endif
L
Linus Torvalds 已提交
7388
#endif
P
Peter Zijlstra 已提交
7389
		init_rq_hrtick(rq);
L
Linus Torvalds 已提交
7390 7391 7392
		atomic_set(&rq->nr_iowait, 0);
	}

7393
	set_load_weight(&init_task);
7394

7395 7396 7397 7398
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&init_task.preempt_notifiers);
#endif

7399
#ifdef CONFIG_RT_MUTEXES
7400
	plist_head_init(&init_task.pi_waiters);
7401 7402
#endif

L
Linus Torvalds 已提交
7403 7404 7405 7406 7407 7408 7409 7410 7411 7412 7413 7414 7415
	/*
	 * The boot idle thread does lazy MMU switching as well:
	 */
	atomic_inc(&init_mm.mm_count);
	enter_lazy_tlb(&init_mm, current);

	/*
	 * Make us the idle thread. Technically, schedule() should not be
	 * called from this thread, however somewhere below it might be,
	 * but because we are the idle thread, we just pick up running again
	 * when this runqueue becomes "idle".
	 */
	init_idle(current, smp_processor_id());
7416 7417 7418

	calc_load_update = jiffies + LOAD_FREQ;

I
Ingo Molnar 已提交
7419 7420 7421 7422
	/*
	 * During early bootup we pretend to be a normal task:
	 */
	current->sched_class = &fair_sched_class;
7423

7424
#ifdef CONFIG_SMP
7425
	zalloc_cpumask_var(&sched_domains_tmpmask, GFP_NOWAIT);
R
Rusty Russell 已提交
7426 7427 7428
	/* May be allocated at isolcpus cmdline parse time */
	if (cpu_isolated_map == NULL)
		zalloc_cpumask_var(&cpu_isolated_map, GFP_NOWAIT);
7429
	idle_thread_set_boot_cpu();
7430 7431
#endif
	init_sched_fair_class();
7432

7433
	scheduler_running = 1;
L
Linus Torvalds 已提交
7434 7435
}

7436
#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
7437 7438
static inline int preempt_count_equals(int preempt_offset)
{
7439
	int nested = (preempt_count() & ~PREEMPT_ACTIVE) + rcu_preempt_depth();
7440

A
Arnd Bergmann 已提交
7441
	return (nested == preempt_offset);
7442 7443
}

7444
void __might_sleep(const char *file, int line, int preempt_offset)
L
Linus Torvalds 已提交
7445 7446 7447
{
	static unsigned long prev_jiffy;	/* ratelimiting */

7448
	rcu_sleep_check(); /* WARN_ON_ONCE() by default, no rate limit reqd. */
7449 7450
	if ((preempt_count_equals(preempt_offset) && !irqs_disabled()) ||
	    system_state != SYSTEM_RUNNING || oops_in_progress)
I
Ingo Molnar 已提交
7451 7452 7453 7454 7455
		return;
	if (time_before(jiffies, prev_jiffy + HZ) && prev_jiffy)
		return;
	prev_jiffy = jiffies;

P
Peter Zijlstra 已提交
7456 7457 7458 7459 7460 7461 7462
	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
7463 7464 7465 7466 7467

	debug_show_held_locks(current);
	if (irqs_disabled())
		print_irqtrace_events(current);
	dump_stack();
L
Linus Torvalds 已提交
7468 7469 7470 7471 7472
}
EXPORT_SYMBOL(__might_sleep);
#endif

#ifdef CONFIG_MAGIC_SYSRQ
7473 7474
static void normalize_task(struct rq *rq, struct task_struct *p)
{
P
Peter Zijlstra 已提交
7475 7476
	const struct sched_class *prev_class = p->sched_class;
	int old_prio = p->prio;
7477
	int on_rq;
7478

P
Peter Zijlstra 已提交
7479
	on_rq = p->on_rq;
7480
	if (on_rq)
7481
		dequeue_task(rq, p, 0);
7482 7483
	__setscheduler(rq, p, SCHED_NORMAL, 0);
	if (on_rq) {
7484
		enqueue_task(rq, p, 0);
7485 7486
		resched_task(rq->curr);
	}
P
Peter Zijlstra 已提交
7487 7488

	check_class_changed(rq, p, prev_class, old_prio);
7489 7490
}

L
Linus Torvalds 已提交
7491 7492
void normalize_rt_tasks(void)
{
7493
	struct task_struct *g, *p;
L
Linus Torvalds 已提交
7494
	unsigned long flags;
7495
	struct rq *rq;
L
Linus Torvalds 已提交
7496

7497
	read_lock_irqsave(&tasklist_lock, flags);
7498
	do_each_thread(g, p) {
7499 7500 7501 7502 7503 7504
		/*
		 * Only normalize user tasks:
		 */
		if (!p->mm)
			continue;

I
Ingo Molnar 已提交
7505 7506
		p->se.exec_start		= 0;
#ifdef CONFIG_SCHEDSTATS
7507 7508 7509
		p->se.statistics.wait_start	= 0;
		p->se.statistics.sleep_start	= 0;
		p->se.statistics.block_start	= 0;
I
Ingo Molnar 已提交
7510
#endif
I
Ingo Molnar 已提交
7511 7512 7513 7514 7515 7516 7517 7518

		if (!rt_task(p)) {
			/*
			 * Renice negative nice level userspace
			 * tasks back to 0:
			 */
			if (TASK_NICE(p) < 0 && p->mm)
				set_user_nice(p, 0);
L
Linus Torvalds 已提交
7519
			continue;
I
Ingo Molnar 已提交
7520
		}
L
Linus Torvalds 已提交
7521

7522
		raw_spin_lock(&p->pi_lock);
7523
		rq = __task_rq_lock(p);
L
Linus Torvalds 已提交
7524

7525
		normalize_task(rq, p);
7526

7527
		__task_rq_unlock(rq);
7528
		raw_spin_unlock(&p->pi_lock);
7529 7530
	} while_each_thread(g, p);

7531
	read_unlock_irqrestore(&tasklist_lock, flags);
L
Linus Torvalds 已提交
7532 7533 7534
}

#endif /* CONFIG_MAGIC_SYSRQ */
7535

7536
#if defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB)
7537
/*
7538
 * These functions are only useful for the IA64 MCA handling, or kdb.
7539 7540 7541 7542 7543 7544 7545 7546 7547 7548 7549 7550 7551 7552
 *
 * 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!
 */
7553
struct task_struct *curr_task(int cpu)
7554 7555 7556 7557
{
	return cpu_curr(cpu);
}

7558 7559 7560
#endif /* defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB) */

#ifdef CONFIG_IA64
7561 7562 7563 7564 7565 7566
/**
 * 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
7567 7568
 * 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
7569 7570 7571 7572 7573 7574 7575
 * 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!
 */
7576
void set_curr_task(int cpu, struct task_struct *p)
7577 7578 7579 7580 7581
{
	cpu_curr(cpu) = p;
}

#endif
S
Srivatsa Vaddagiri 已提交
7582

D
Dhaval Giani 已提交
7583
#ifdef CONFIG_CGROUP_SCHED
7584 7585 7586
/* task_group_lock serializes the addition/removal of task groups */
static DEFINE_SPINLOCK(task_group_lock);

7587 7588 7589 7590
static void free_sched_group(struct task_group *tg)
{
	free_fair_sched_group(tg);
	free_rt_sched_group(tg);
7591
	autogroup_free(tg);
7592 7593 7594 7595
	kfree(tg);
}

/* allocate runqueue etc for a new task group */
7596
struct task_group *sched_create_group(struct task_group *parent)
7597 7598 7599 7600 7601 7602 7603 7604
{
	struct task_group *tg;
	unsigned long flags;

	tg = kzalloc(sizeof(*tg), GFP_KERNEL);
	if (!tg)
		return ERR_PTR(-ENOMEM);

7605
	if (!alloc_fair_sched_group(tg, parent))
7606 7607
		goto err;

7608
	if (!alloc_rt_sched_group(tg, parent))
7609 7610
		goto err;

7611
	spin_lock_irqsave(&task_group_lock, flags);
P
Peter Zijlstra 已提交
7612
	list_add_rcu(&tg->list, &task_groups);
P
Peter Zijlstra 已提交
7613 7614 7615 7616 7617

	WARN_ON(!parent); /* root should already exist */

	tg->parent = parent;
	INIT_LIST_HEAD(&tg->children);
7618
	list_add_rcu(&tg->siblings, &parent->children);
7619
	spin_unlock_irqrestore(&task_group_lock, flags);
S
Srivatsa Vaddagiri 已提交
7620

7621
	return tg;
S
Srivatsa Vaddagiri 已提交
7622 7623

err:
P
Peter Zijlstra 已提交
7624
	free_sched_group(tg);
S
Srivatsa Vaddagiri 已提交
7625 7626 7627
	return ERR_PTR(-ENOMEM);
}

7628
/* rcu callback to free various structures associated with a task group */
P
Peter Zijlstra 已提交
7629
static void free_sched_group_rcu(struct rcu_head *rhp)
S
Srivatsa Vaddagiri 已提交
7630 7631
{
	/* now it should be safe to free those cfs_rqs */
P
Peter Zijlstra 已提交
7632
	free_sched_group(container_of(rhp, struct task_group, rcu));
S
Srivatsa Vaddagiri 已提交
7633 7634
}

7635
/* Destroy runqueue etc associated with a task group */
7636
void sched_destroy_group(struct task_group *tg)
S
Srivatsa Vaddagiri 已提交
7637
{
7638
	unsigned long flags;
7639
	int i;
S
Srivatsa Vaddagiri 已提交
7640

7641 7642
	/* end participation in shares distribution */
	for_each_possible_cpu(i)
7643
		unregister_fair_sched_group(tg, i);
7644 7645

	spin_lock_irqsave(&task_group_lock, flags);
P
Peter Zijlstra 已提交
7646
	list_del_rcu(&tg->list);
P
Peter Zijlstra 已提交
7647
	list_del_rcu(&tg->siblings);
7648
	spin_unlock_irqrestore(&task_group_lock, flags);
7649 7650

	/* wait for possible concurrent references to cfs_rqs complete */
P
Peter Zijlstra 已提交
7651
	call_rcu(&tg->rcu, free_sched_group_rcu);
S
Srivatsa Vaddagiri 已提交
7652 7653
}

7654
/* change task's runqueue when it moves between groups.
I
Ingo Molnar 已提交
7655 7656 7657
 *	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.
7658 7659
 */
void sched_move_task(struct task_struct *tsk)
S
Srivatsa Vaddagiri 已提交
7660 7661 7662 7663 7664 7665 7666
{
	int on_rq, running;
	unsigned long flags;
	struct rq *rq;

	rq = task_rq_lock(tsk, &flags);

7667
	running = task_current(rq, tsk);
P
Peter Zijlstra 已提交
7668
	on_rq = tsk->on_rq;
S
Srivatsa Vaddagiri 已提交
7669

7670
	if (on_rq)
S
Srivatsa Vaddagiri 已提交
7671
		dequeue_task(rq, tsk, 0);
7672 7673
	if (unlikely(running))
		tsk->sched_class->put_prev_task(rq, tsk);
S
Srivatsa Vaddagiri 已提交
7674

P
Peter Zijlstra 已提交
7675
#ifdef CONFIG_FAIR_GROUP_SCHED
7676 7677 7678
	if (tsk->sched_class->task_move_group)
		tsk->sched_class->task_move_group(tsk, on_rq);
	else
P
Peter Zijlstra 已提交
7679
#endif
7680
		set_task_rq(tsk, task_cpu(tsk));
P
Peter Zijlstra 已提交
7681

7682 7683 7684
	if (unlikely(running))
		tsk->sched_class->set_curr_task(rq);
	if (on_rq)
7685
		enqueue_task(rq, tsk, 0);
S
Srivatsa Vaddagiri 已提交
7686

7687
	task_rq_unlock(rq, tsk, &flags);
S
Srivatsa Vaddagiri 已提交
7688
}
D
Dhaval Giani 已提交
7689
#endif /* CONFIG_CGROUP_SCHED */
S
Srivatsa Vaddagiri 已提交
7690

7691
#if defined(CONFIG_RT_GROUP_SCHED) || defined(CONFIG_CFS_BANDWIDTH)
P
Peter Zijlstra 已提交
7692 7693 7694
static unsigned long to_ratio(u64 period, u64 runtime)
{
	if (runtime == RUNTIME_INF)
P
Peter Zijlstra 已提交
7695
		return 1ULL << 20;
P
Peter Zijlstra 已提交
7696

P
Peter Zijlstra 已提交
7697
	return div64_u64(runtime << 20, period);
P
Peter Zijlstra 已提交
7698
}
7699 7700 7701 7702 7703 7704 7705
#endif

#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
7706

P
Peter Zijlstra 已提交
7707 7708
/* Must be called with tasklist_lock held */
static inline int tg_has_rt_tasks(struct task_group *tg)
7709
{
P
Peter Zijlstra 已提交
7710
	struct task_struct *g, *p;
7711

P
Peter Zijlstra 已提交
7712
	do_each_thread(g, p) {
7713
		if (rt_task(p) && task_rq(p)->rt.tg == tg)
P
Peter Zijlstra 已提交
7714 7715
			return 1;
	} while_each_thread(g, p);
7716

P
Peter Zijlstra 已提交
7717 7718
	return 0;
}
7719

P
Peter Zijlstra 已提交
7720 7721 7722 7723 7724
struct rt_schedulable_data {
	struct task_group *tg;
	u64 rt_period;
	u64 rt_runtime;
};
7725

7726
static int tg_rt_schedulable(struct task_group *tg, void *data)
P
Peter Zijlstra 已提交
7727 7728 7729 7730 7731
{
	struct rt_schedulable_data *d = data;
	struct task_group *child;
	unsigned long total, sum = 0;
	u64 period, runtime;
7732

P
Peter Zijlstra 已提交
7733 7734
	period = ktime_to_ns(tg->rt_bandwidth.rt_period);
	runtime = tg->rt_bandwidth.rt_runtime;
7735

P
Peter Zijlstra 已提交
7736 7737 7738
	if (tg == d->tg) {
		period = d->rt_period;
		runtime = d->rt_runtime;
7739 7740
	}

7741 7742 7743 7744 7745
	/*
	 * Cannot have more runtime than the period.
	 */
	if (runtime > period && runtime != RUNTIME_INF)
		return -EINVAL;
P
Peter Zijlstra 已提交
7746

7747 7748 7749
	/*
	 * Ensure we don't starve existing RT tasks.
	 */
P
Peter Zijlstra 已提交
7750 7751
	if (rt_bandwidth_enabled() && !runtime && tg_has_rt_tasks(tg))
		return -EBUSY;
P
Peter Zijlstra 已提交
7752

P
Peter Zijlstra 已提交
7753
	total = to_ratio(period, runtime);
P
Peter Zijlstra 已提交
7754

7755 7756 7757 7758 7759
	/*
	 * 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
7760

7761 7762 7763
	/*
	 * The sum of our children's runtime should not exceed our own.
	 */
P
Peter Zijlstra 已提交
7764 7765 7766
	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
7767

P
Peter Zijlstra 已提交
7768 7769 7770 7771
		if (child == d->tg) {
			period = d->rt_period;
			runtime = d->rt_runtime;
		}
P
Peter Zijlstra 已提交
7772

P
Peter Zijlstra 已提交
7773
		sum += to_ratio(period, runtime);
P
Peter Zijlstra 已提交
7774
	}
P
Peter Zijlstra 已提交
7775

P
Peter Zijlstra 已提交
7776 7777 7778 7779
	if (sum > total)
		return -EINVAL;

	return 0;
P
Peter Zijlstra 已提交
7780 7781
}

P
Peter Zijlstra 已提交
7782
static int __rt_schedulable(struct task_group *tg, u64 period, u64 runtime)
7783
{
7784 7785
	int ret;

P
Peter Zijlstra 已提交
7786 7787 7788 7789 7790 7791
	struct rt_schedulable_data data = {
		.tg = tg,
		.rt_period = period,
		.rt_runtime = runtime,
	};

7792 7793 7794 7795 7796
	rcu_read_lock();
	ret = walk_tg_tree(tg_rt_schedulable, tg_nop, &data);
	rcu_read_unlock();

	return ret;
7797 7798
}

7799
static int tg_set_rt_bandwidth(struct task_group *tg,
7800
		u64 rt_period, u64 rt_runtime)
P
Peter Zijlstra 已提交
7801
{
P
Peter Zijlstra 已提交
7802
	int i, err = 0;
P
Peter Zijlstra 已提交
7803 7804

	mutex_lock(&rt_constraints_mutex);
7805
	read_lock(&tasklist_lock);
P
Peter Zijlstra 已提交
7806 7807
	err = __rt_schedulable(tg, rt_period, rt_runtime);
	if (err)
P
Peter Zijlstra 已提交
7808
		goto unlock;
P
Peter Zijlstra 已提交
7809

7810
	raw_spin_lock_irq(&tg->rt_bandwidth.rt_runtime_lock);
7811 7812
	tg->rt_bandwidth.rt_period = ns_to_ktime(rt_period);
	tg->rt_bandwidth.rt_runtime = rt_runtime;
P
Peter Zijlstra 已提交
7813 7814 7815 7816

	for_each_possible_cpu(i) {
		struct rt_rq *rt_rq = tg->rt_rq[i];

7817
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
7818
		rt_rq->rt_runtime = rt_runtime;
7819
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
7820
	}
7821
	raw_spin_unlock_irq(&tg->rt_bandwidth.rt_runtime_lock);
P
Peter Zijlstra 已提交
7822
unlock:
7823
	read_unlock(&tasklist_lock);
P
Peter Zijlstra 已提交
7824 7825 7826
	mutex_unlock(&rt_constraints_mutex);

	return err;
P
Peter Zijlstra 已提交
7827 7828
}

7829 7830 7831 7832 7833 7834 7835 7836 7837
int sched_group_set_rt_runtime(struct task_group *tg, long rt_runtime_us)
{
	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;

7838
	return tg_set_rt_bandwidth(tg, rt_period, rt_runtime);
7839 7840
}

P
Peter Zijlstra 已提交
7841 7842 7843 7844
long sched_group_rt_runtime(struct task_group *tg)
{
	u64 rt_runtime_us;

7845
	if (tg->rt_bandwidth.rt_runtime == RUNTIME_INF)
P
Peter Zijlstra 已提交
7846 7847
		return -1;

7848
	rt_runtime_us = tg->rt_bandwidth.rt_runtime;
P
Peter Zijlstra 已提交
7849 7850 7851
	do_div(rt_runtime_us, NSEC_PER_USEC);
	return rt_runtime_us;
}
7852 7853 7854 7855 7856 7857 7858 7859

int sched_group_set_rt_period(struct task_group *tg, long rt_period_us)
{
	u64 rt_runtime, rt_period;

	rt_period = (u64)rt_period_us * NSEC_PER_USEC;
	rt_runtime = tg->rt_bandwidth.rt_runtime;

7860 7861 7862
	if (rt_period == 0)
		return -EINVAL;

7863
	return tg_set_rt_bandwidth(tg, rt_period, rt_runtime);
7864 7865 7866 7867 7868 7869 7870 7871 7872 7873 7874 7875 7876
}

long sched_group_rt_period(struct task_group *tg)
{
	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;
}

static int sched_rt_global_constraints(void)
{
7877
	u64 runtime, period;
7878 7879
	int ret = 0;

7880 7881 7882
	if (sysctl_sched_rt_period <= 0)
		return -EINVAL;

7883 7884 7885 7886 7887 7888 7889 7890
	runtime = global_rt_runtime();
	period = global_rt_period();

	/*
	 * Sanity check on the sysctl variables.
	 */
	if (runtime > period && runtime != RUNTIME_INF)
		return -EINVAL;
7891

7892
	mutex_lock(&rt_constraints_mutex);
P
Peter Zijlstra 已提交
7893
	read_lock(&tasklist_lock);
7894
	ret = __rt_schedulable(NULL, 0, 0);
P
Peter Zijlstra 已提交
7895
	read_unlock(&tasklist_lock);
7896 7897 7898 7899
	mutex_unlock(&rt_constraints_mutex);

	return ret;
}
7900 7901 7902 7903 7904 7905 7906 7907 7908 7909

int sched_rt_can_attach(struct task_group *tg, struct task_struct *tsk)
{
	/* 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;
}

7910
#else /* !CONFIG_RT_GROUP_SCHED */
7911 7912
static int sched_rt_global_constraints(void)
{
P
Peter Zijlstra 已提交
7913 7914 7915
	unsigned long flags;
	int i;

7916 7917 7918
	if (sysctl_sched_rt_period <= 0)
		return -EINVAL;

7919 7920 7921 7922 7923 7924 7925
	/*
	 * There's always some RT tasks in the root group
	 * -- migration, kstopmachine etc..
	 */
	if (sysctl_sched_rt_runtime == 0)
		return -EBUSY;

7926
	raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags);
P
Peter Zijlstra 已提交
7927 7928 7929
	for_each_possible_cpu(i) {
		struct rt_rq *rt_rq = &cpu_rq(i)->rt;

7930
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
7931
		rt_rq->rt_runtime = global_rt_runtime();
7932
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
7933
	}
7934
	raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags);
P
Peter Zijlstra 已提交
7935

7936 7937
	return 0;
}
7938
#endif /* CONFIG_RT_GROUP_SCHED */
7939 7940

int sched_rt_handler(struct ctl_table *table, int write,
7941
		void __user *buffer, size_t *lenp,
7942 7943 7944 7945 7946 7947 7948 7949 7950 7951
		loff_t *ppos)
{
	int ret;
	int old_period, old_runtime;
	static DEFINE_MUTEX(mutex);

	mutex_lock(&mutex);
	old_period = sysctl_sched_rt_period;
	old_runtime = sysctl_sched_rt_runtime;

7952
	ret = proc_dointvec(table, write, buffer, lenp, ppos);
7953 7954 7955 7956 7957 7958 7959 7960 7961 7962 7963 7964 7965 7966 7967 7968

	if (!ret && write) {
		ret = sched_rt_global_constraints();
		if (ret) {
			sysctl_sched_rt_period = old_period;
			sysctl_sched_rt_runtime = old_runtime;
		} else {
			def_rt_bandwidth.rt_runtime = global_rt_runtime();
			def_rt_bandwidth.rt_period =
				ns_to_ktime(global_rt_period());
		}
	}
	mutex_unlock(&mutex);

	return ret;
}
7969

7970
#ifdef CONFIG_CGROUP_SCHED
7971 7972

/* return corresponding task_group object of a cgroup */
7973
static inline struct task_group *cgroup_tg(struct cgroup *cgrp)
7974
{
7975 7976
	return container_of(cgroup_subsys_state(cgrp, cpu_cgroup_subsys_id),
			    struct task_group, css);
7977 7978
}

7979
static struct cgroup_subsys_state *cpu_cgroup_create(struct cgroup *cgrp)
7980
{
7981
	struct task_group *tg, *parent;
7982

7983
	if (!cgrp->parent) {
7984
		/* This is early initialization for the top cgroup */
7985
		return &root_task_group.css;
7986 7987
	}

7988 7989
	parent = cgroup_tg(cgrp->parent);
	tg = sched_create_group(parent);
7990 7991 7992 7993 7994 7995
	if (IS_ERR(tg))
		return ERR_PTR(-ENOMEM);

	return &tg->css;
}

7996
static void cpu_cgroup_destroy(struct cgroup *cgrp)
7997
{
7998
	struct task_group *tg = cgroup_tg(cgrp);
7999 8000 8001 8002

	sched_destroy_group(tg);
}

8003
static int cpu_cgroup_can_attach(struct cgroup *cgrp,
8004
				 struct cgroup_taskset *tset)
8005
{
8006 8007 8008
	struct task_struct *task;

	cgroup_taskset_for_each(task, cgrp, tset) {
8009
#ifdef CONFIG_RT_GROUP_SCHED
8010 8011
		if (!sched_rt_can_attach(cgroup_tg(cgrp), task))
			return -EINVAL;
8012
#else
8013 8014 8015
		/* We don't support RT-tasks being in separate groups */
		if (task->sched_class != &fair_sched_class)
			return -EINVAL;
8016
#endif
8017
	}
8018 8019
	return 0;
}
8020

8021
static void cpu_cgroup_attach(struct cgroup *cgrp,
8022
			      struct cgroup_taskset *tset)
8023
{
8024 8025 8026 8027
	struct task_struct *task;

	cgroup_taskset_for_each(task, cgrp, tset)
		sched_move_task(task);
8028 8029
}

8030
static void
8031 8032
cpu_cgroup_exit(struct cgroup *cgrp, struct cgroup *old_cgrp,
		struct task_struct *task)
8033 8034 8035 8036 8037 8038 8039 8040 8041 8042 8043 8044
{
	/*
	 * cgroup_exit() is called in the copy_process() failure path.
	 * Ignore this case since the task hasn't ran yet, this avoids
	 * trying to poke a half freed task state from generic code.
	 */
	if (!(task->flags & PF_EXITING))
		return;

	sched_move_task(task);
}

8045
#ifdef CONFIG_FAIR_GROUP_SCHED
8046
static int cpu_shares_write_u64(struct cgroup *cgrp, struct cftype *cftype,
8047
				u64 shareval)
8048
{
8049
	return sched_group_set_shares(cgroup_tg(cgrp), scale_load(shareval));
8050 8051
}

8052
static u64 cpu_shares_read_u64(struct cgroup *cgrp, struct cftype *cft)
8053
{
8054
	struct task_group *tg = cgroup_tg(cgrp);
8055

8056
	return (u64) scale_load_down(tg->shares);
8057
}
8058 8059

#ifdef CONFIG_CFS_BANDWIDTH
8060 8061
static DEFINE_MUTEX(cfs_constraints_mutex);

8062 8063 8064
const u64 max_cfs_quota_period = 1 * NSEC_PER_SEC; /* 1s */
const u64 min_cfs_quota_period = 1 * NSEC_PER_MSEC; /* 1ms */

8065 8066
static int __cfs_schedulable(struct task_group *tg, u64 period, u64 runtime);

8067 8068
static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota)
{
8069
	int i, ret = 0, runtime_enabled, runtime_was_enabled;
8070
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
8071 8072 8073 8074 8075 8076 8077 8078 8079 8080 8081 8082 8083 8084 8085 8086 8087 8088 8089 8090

	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;

8091 8092 8093 8094 8095
	mutex_lock(&cfs_constraints_mutex);
	ret = __cfs_schedulable(tg, period, quota);
	if (ret)
		goto out_unlock;

8096
	runtime_enabled = quota != RUNTIME_INF;
8097 8098
	runtime_was_enabled = cfs_b->quota != RUNTIME_INF;
	account_cfs_bandwidth_used(runtime_enabled, runtime_was_enabled);
8099 8100 8101
	raw_spin_lock_irq(&cfs_b->lock);
	cfs_b->period = ns_to_ktime(period);
	cfs_b->quota = quota;
8102

P
Paul Turner 已提交
8103
	__refill_cfs_bandwidth_runtime(cfs_b);
8104 8105 8106 8107 8108 8109
	/* restart the period timer (if active) to handle new period expiry */
	if (runtime_enabled && cfs_b->timer_active) {
		/* force a reprogram */
		cfs_b->timer_active = 0;
		__start_cfs_bandwidth(cfs_b);
	}
8110 8111 8112 8113
	raw_spin_unlock_irq(&cfs_b->lock);

	for_each_possible_cpu(i) {
		struct cfs_rq *cfs_rq = tg->cfs_rq[i];
8114
		struct rq *rq = cfs_rq->rq;
8115 8116

		raw_spin_lock_irq(&rq->lock);
8117
		cfs_rq->runtime_enabled = runtime_enabled;
8118
		cfs_rq->runtime_remaining = 0;
8119

8120
		if (cfs_rq->throttled)
8121
			unthrottle_cfs_rq(cfs_rq);
8122 8123
		raw_spin_unlock_irq(&rq->lock);
	}
8124 8125
out_unlock:
	mutex_unlock(&cfs_constraints_mutex);
8126

8127
	return ret;
8128 8129 8130 8131 8132 8133
}

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

8134
	period = ktime_to_ns(tg->cfs_bandwidth.period);
8135 8136 8137 8138 8139 8140 8141 8142 8143 8144 8145 8146
	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;

8147
	if (tg->cfs_bandwidth.quota == RUNTIME_INF)
8148 8149
		return -1;

8150
	quota_us = tg->cfs_bandwidth.quota;
8151 8152 8153 8154 8155 8156 8157 8158 8159 8160
	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;
8161
	quota = tg->cfs_bandwidth.quota;
8162 8163 8164 8165 8166 8167 8168 8169

	return tg_set_cfs_bandwidth(tg, period, quota);
}

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

8170
	cfs_period_us = ktime_to_ns(tg->cfs_bandwidth.period);
8171 8172 8173 8174 8175 8176 8177 8178 8179 8180 8181 8182 8183 8184 8185 8186 8187 8188 8189 8190 8191 8192 8193 8194 8195 8196 8197
	do_div(cfs_period_us, NSEC_PER_USEC);

	return cfs_period_us;
}

static s64 cpu_cfs_quota_read_s64(struct cgroup *cgrp, struct cftype *cft)
{
	return tg_get_cfs_quota(cgroup_tg(cgrp));
}

static int cpu_cfs_quota_write_s64(struct cgroup *cgrp, struct cftype *cftype,
				s64 cfs_quota_us)
{
	return tg_set_cfs_quota(cgroup_tg(cgrp), cfs_quota_us);
}

static u64 cpu_cfs_period_read_u64(struct cgroup *cgrp, struct cftype *cft)
{
	return tg_get_cfs_period(cgroup_tg(cgrp));
}

static int cpu_cfs_period_write_u64(struct cgroup *cgrp, struct cftype *cftype,
				u64 cfs_period_us)
{
	return tg_set_cfs_period(cgroup_tg(cgrp), cfs_period_us);
}

8198 8199 8200 8201 8202 8203 8204 8205 8206 8207 8208 8209 8210 8211 8212 8213 8214 8215 8216 8217 8218 8219 8220 8221 8222 8223 8224 8225 8226 8227 8228 8229
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;
8230
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
8231 8232 8233 8234 8235
	s64 quota = 0, parent_quota = -1;

	if (!tg->parent) {
		quota = RUNTIME_INF;
	} else {
8236
		struct cfs_bandwidth *parent_b = &tg->parent->cfs_bandwidth;
8237 8238 8239 8240 8241 8242 8243 8244 8245 8246 8247 8248 8249 8250 8251 8252 8253 8254 8255 8256

		quota = normalize_cfs_quota(tg, d);
		parent_quota = parent_b->hierarchal_quota;

		/*
		 * 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;
	}
	cfs_b->hierarchal_quota = quota;

	return 0;
}

static int __cfs_schedulable(struct task_group *tg, u64 period, u64 quota)
{
8257
	int ret;
8258 8259 8260 8261 8262 8263 8264 8265 8266 8267 8268
	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);
	}

8269 8270 8271 8272 8273
	rcu_read_lock();
	ret = walk_tg_tree(tg_cfs_schedulable_down, tg_nop, &data);
	rcu_read_unlock();

	return ret;
8274
}
8275 8276 8277 8278 8279

static int cpu_stats_show(struct cgroup *cgrp, struct cftype *cft,
		struct cgroup_map_cb *cb)
{
	struct task_group *tg = cgroup_tg(cgrp);
8280
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
8281 8282 8283 8284 8285 8286 8287

	cb->fill(cb, "nr_periods", cfs_b->nr_periods);
	cb->fill(cb, "nr_throttled", cfs_b->nr_throttled);
	cb->fill(cb, "throttled_time", cfs_b->throttled_time);

	return 0;
}
8288
#endif /* CONFIG_CFS_BANDWIDTH */
8289
#endif /* CONFIG_FAIR_GROUP_SCHED */
8290

8291
#ifdef CONFIG_RT_GROUP_SCHED
M
Mirco Tischler 已提交
8292
static int cpu_rt_runtime_write(struct cgroup *cgrp, struct cftype *cft,
8293
				s64 val)
P
Peter Zijlstra 已提交
8294
{
8295
	return sched_group_set_rt_runtime(cgroup_tg(cgrp), val);
P
Peter Zijlstra 已提交
8296 8297
}

8298
static s64 cpu_rt_runtime_read(struct cgroup *cgrp, struct cftype *cft)
P
Peter Zijlstra 已提交
8299
{
8300
	return sched_group_rt_runtime(cgroup_tg(cgrp));
P
Peter Zijlstra 已提交
8301
}
8302 8303 8304 8305 8306 8307 8308 8309 8310 8311 8312

static int cpu_rt_period_write_uint(struct cgroup *cgrp, struct cftype *cftype,
		u64 rt_period_us)
{
	return sched_group_set_rt_period(cgroup_tg(cgrp), rt_period_us);
}

static u64 cpu_rt_period_read_uint(struct cgroup *cgrp, struct cftype *cft)
{
	return sched_group_rt_period(cgroup_tg(cgrp));
}
8313
#endif /* CONFIG_RT_GROUP_SCHED */
P
Peter Zijlstra 已提交
8314

8315
static struct cftype cpu_files[] = {
8316
#ifdef CONFIG_FAIR_GROUP_SCHED
8317 8318
	{
		.name = "shares",
8319 8320
		.read_u64 = cpu_shares_read_u64,
		.write_u64 = cpu_shares_write_u64,
8321
	},
8322
#endif
8323 8324 8325 8326 8327 8328 8329 8330 8331 8332 8333
#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,
	},
8334 8335 8336 8337
	{
		.name = "stat",
		.read_map = cpu_stats_show,
	},
8338
#endif
8339
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
8340
	{
P
Peter Zijlstra 已提交
8341
		.name = "rt_runtime_us",
8342 8343
		.read_s64 = cpu_rt_runtime_read,
		.write_s64 = cpu_rt_runtime_write,
P
Peter Zijlstra 已提交
8344
	},
8345 8346
	{
		.name = "rt_period_us",
8347 8348
		.read_u64 = cpu_rt_period_read_uint,
		.write_u64 = cpu_rt_period_write_uint,
8349
	},
8350
#endif
8351
	{ }	/* terminate */
8352 8353 8354
};

struct cgroup_subsys cpu_cgroup_subsys = {
I
Ingo Molnar 已提交
8355 8356 8357
	.name		= "cpu",
	.create		= cpu_cgroup_create,
	.destroy	= cpu_cgroup_destroy,
8358 8359
	.can_attach	= cpu_cgroup_can_attach,
	.attach		= cpu_cgroup_attach,
8360
	.exit		= cpu_cgroup_exit,
I
Ingo Molnar 已提交
8361
	.subsys_id	= cpu_cgroup_subsys_id,
8362
	.base_cftypes	= cpu_files,
8363 8364 8365
	.early_init	= 1,
};

8366
#endif	/* CONFIG_CGROUP_SCHED */
8367 8368 8369 8370 8371 8372 8373 8374 8375 8376 8377

#ifdef CONFIG_CGROUP_CPUACCT

/*
 * CPU accounting code for task groups.
 *
 * Based on the work by Paul Menage (menage@google.com) and Balbir Singh
 * (balbir@in.ibm.com).
 */

/* create a new cpu accounting group */
8378
static struct cgroup_subsys_state *cpuacct_create(struct cgroup *cgrp)
8379
{
8380
	struct cpuacct *ca;
8381

8382 8383 8384 8385
	if (!cgrp->parent)
		return &root_cpuacct.css;

	ca = kzalloc(sizeof(*ca), GFP_KERNEL);
8386
	if (!ca)
8387
		goto out;
8388 8389

	ca->cpuusage = alloc_percpu(u64);
8390 8391 8392
	if (!ca->cpuusage)
		goto out_free_ca;

8393 8394 8395
	ca->cpustat = alloc_percpu(struct kernel_cpustat);
	if (!ca->cpustat)
		goto out_free_cpuusage;
8396

8397
	return &ca->css;
8398

8399
out_free_cpuusage:
8400 8401 8402 8403 8404
	free_percpu(ca->cpuusage);
out_free_ca:
	kfree(ca);
out:
	return ERR_PTR(-ENOMEM);
8405 8406 8407
}

/* destroy an existing cpu accounting group */
8408
static void cpuacct_destroy(struct cgroup *cgrp)
8409
{
8410
	struct cpuacct *ca = cgroup_ca(cgrp);
8411

8412
	free_percpu(ca->cpustat);
8413 8414 8415 8416
	free_percpu(ca->cpuusage);
	kfree(ca);
}

8417 8418
static u64 cpuacct_cpuusage_read(struct cpuacct *ca, int cpu)
{
8419
	u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
8420 8421 8422 8423 8424 8425
	u64 data;

#ifndef CONFIG_64BIT
	/*
	 * Take rq->lock to make 64-bit read safe on 32-bit platforms.
	 */
8426
	raw_spin_lock_irq(&cpu_rq(cpu)->lock);
8427
	data = *cpuusage;
8428
	raw_spin_unlock_irq(&cpu_rq(cpu)->lock);
8429 8430 8431 8432 8433 8434 8435 8436 8437
#else
	data = *cpuusage;
#endif

	return data;
}

static void cpuacct_cpuusage_write(struct cpuacct *ca, int cpu, u64 val)
{
8438
	u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
8439 8440 8441 8442 8443

#ifndef CONFIG_64BIT
	/*
	 * Take rq->lock to make 64-bit write safe on 32-bit platforms.
	 */
8444
	raw_spin_lock_irq(&cpu_rq(cpu)->lock);
8445
	*cpuusage = val;
8446
	raw_spin_unlock_irq(&cpu_rq(cpu)->lock);
8447 8448 8449 8450 8451
#else
	*cpuusage = val;
#endif
}

8452
/* return total cpu usage (in nanoseconds) of a group */
8453
static u64 cpuusage_read(struct cgroup *cgrp, struct cftype *cft)
8454
{
8455
	struct cpuacct *ca = cgroup_ca(cgrp);
8456 8457 8458
	u64 totalcpuusage = 0;
	int i;

8459 8460
	for_each_present_cpu(i)
		totalcpuusage += cpuacct_cpuusage_read(ca, i);
8461 8462 8463 8464

	return totalcpuusage;
}

8465 8466 8467 8468 8469 8470 8471 8472 8473 8474 8475 8476
static int cpuusage_write(struct cgroup *cgrp, struct cftype *cftype,
								u64 reset)
{
	struct cpuacct *ca = cgroup_ca(cgrp);
	int err = 0;
	int i;

	if (reset) {
		err = -EINVAL;
		goto out;
	}

8477 8478
	for_each_present_cpu(i)
		cpuacct_cpuusage_write(ca, i, 0);
8479 8480 8481 8482 8483

out:
	return err;
}

8484 8485 8486 8487 8488 8489 8490 8491 8492 8493 8494 8495 8496 8497 8498
static int cpuacct_percpu_seq_read(struct cgroup *cgroup, struct cftype *cft,
				   struct seq_file *m)
{
	struct cpuacct *ca = cgroup_ca(cgroup);
	u64 percpu;
	int i;

	for_each_present_cpu(i) {
		percpu = cpuacct_cpuusage_read(ca, i);
		seq_printf(m, "%llu ", (unsigned long long) percpu);
	}
	seq_printf(m, "\n");
	return 0;
}

8499 8500 8501 8502 8503 8504
static const char *cpuacct_stat_desc[] = {
	[CPUACCT_STAT_USER] = "user",
	[CPUACCT_STAT_SYSTEM] = "system",
};

static int cpuacct_stats_show(struct cgroup *cgrp, struct cftype *cft,
8505
			      struct cgroup_map_cb *cb)
8506 8507
{
	struct cpuacct *ca = cgroup_ca(cgrp);
8508 8509
	int cpu;
	s64 val = 0;
8510

8511 8512 8513 8514
	for_each_online_cpu(cpu) {
		struct kernel_cpustat *kcpustat = per_cpu_ptr(ca->cpustat, cpu);
		val += kcpustat->cpustat[CPUTIME_USER];
		val += kcpustat->cpustat[CPUTIME_NICE];
8515
	}
8516 8517
	val = cputime64_to_clock_t(val);
	cb->fill(cb, cpuacct_stat_desc[CPUACCT_STAT_USER], val);
8518

8519 8520 8521 8522 8523 8524
	val = 0;
	for_each_online_cpu(cpu) {
		struct kernel_cpustat *kcpustat = per_cpu_ptr(ca->cpustat, cpu);
		val += kcpustat->cpustat[CPUTIME_SYSTEM];
		val += kcpustat->cpustat[CPUTIME_IRQ];
		val += kcpustat->cpustat[CPUTIME_SOFTIRQ];
8525
	}
8526 8527 8528 8529

	val = cputime64_to_clock_t(val);
	cb->fill(cb, cpuacct_stat_desc[CPUACCT_STAT_SYSTEM], val);

8530 8531 8532
	return 0;
}

8533 8534 8535
static struct cftype files[] = {
	{
		.name = "usage",
8536 8537
		.read_u64 = cpuusage_read,
		.write_u64 = cpuusage_write,
8538
	},
8539 8540 8541 8542
	{
		.name = "usage_percpu",
		.read_seq_string = cpuacct_percpu_seq_read,
	},
8543 8544 8545 8546
	{
		.name = "stat",
		.read_map = cpuacct_stats_show,
	},
8547
	{ }	/* terminate */
8548 8549 8550 8551 8552 8553 8554
};

/*
 * charge this task's execution time to its accounting group.
 *
 * called with rq->lock held.
 */
8555
void cpuacct_charge(struct task_struct *tsk, u64 cputime)
8556 8557
{
	struct cpuacct *ca;
8558
	int cpu;
8559

L
Li Zefan 已提交
8560
	if (unlikely(!cpuacct_subsys.active))
8561 8562
		return;

8563
	cpu = task_cpu(tsk);
8564 8565 8566

	rcu_read_lock();

8567 8568
	ca = task_ca(tsk);

8569
	for (; ca; ca = parent_ca(ca)) {
8570
		u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
8571 8572
		*cpuusage += cputime;
	}
8573 8574

	rcu_read_unlock();
8575 8576 8577 8578 8579 8580 8581
}

struct cgroup_subsys cpuacct_subsys = {
	.name = "cpuacct",
	.create = cpuacct_create,
	.destroy = cpuacct_destroy,
	.subsys_id = cpuacct_subsys_id,
8582
	.base_cftypes = files,
8583 8584
};
#endif	/* CONFIG_CGROUP_CPUACCT */