core.c 195.2 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>
L
Linus Torvalds 已提交
74

75
#include <asm/tlb.h>
76
#include <asm/irq_regs.h>
77
#include <asm/mutex.h>
G
Glauber Costa 已提交
78 79 80
#ifdef CONFIG_PARAVIRT
#include <asm/paravirt.h>
#endif
L
Linus Torvalds 已提交
81

82
#include "sched.h"
83
#include "../workqueue_sched.h"
84

85
#define CREATE_TRACE_POINTS
86
#include <trace/events/sched.h>
87

88
void start_bandwidth_timer(struct hrtimer *period_timer, ktime_t period)
89
{
90 91
	unsigned long delta;
	ktime_t soft, hard, now;
92

93 94 95 96 97 98
	for (;;) {
		if (hrtimer_active(period_timer))
			break;

		now = hrtimer_cb_get_time(period_timer);
		hrtimer_forward(period_timer, now, period);
99

100 101 102 103 104 105 106 107
		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);
	}
}

108 109
DEFINE_MUTEX(sched_domains_mutex);
DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
110

111
static void update_rq_clock_task(struct rq *rq, s64 delta);
112

113
void update_rq_clock(struct rq *rq)
114
{
115
	s64 delta;
116

117
	if (rq->skip_clock_update > 0)
118
		return;
119

120 121 122
	delta = sched_clock_cpu(cpu_of(rq)) - rq->clock;
	rq->clock += delta;
	update_rq_clock_task(rq, delta);
123 124
}

I
Ingo Molnar 已提交
125 126 127
/*
 * Debugging: various feature bits
 */
P
Peter Zijlstra 已提交
128 129 130 131

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

I
Ingo Molnar 已提交
132
const_debug unsigned int sysctl_sched_features =
133
#include "features.h"
P
Peter Zijlstra 已提交
134 135 136 137 138 139 140 141
	0;

#undef SCHED_FEAT

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

142
static __read_mostly char *sched_feat_names[] = {
143
#include "features.h"
P
Peter Zijlstra 已提交
144 145 146 147 148
	NULL
};

#undef SCHED_FEAT

L
Li Zefan 已提交
149
static int sched_feat_show(struct seq_file *m, void *v)
P
Peter Zijlstra 已提交
150 151 152
{
	int i;

153
	for (i = 0; i < __SCHED_FEAT_NR; i++) {
L
Li Zefan 已提交
154 155 156
		if (!(sysctl_sched_features & (1UL << i)))
			seq_puts(m, "NO_");
		seq_printf(m, "%s ", sched_feat_names[i]);
P
Peter Zijlstra 已提交
157
	}
L
Li Zefan 已提交
158
	seq_puts(m, "\n");
P
Peter Zijlstra 已提交
159

L
Li Zefan 已提交
160
	return 0;
P
Peter Zijlstra 已提交
161 162
}

163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192
#ifdef HAVE_JUMP_LABEL

#define jump_label_key__true  jump_label_key_enabled
#define jump_label_key__false jump_label_key_disabled

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

struct jump_label_key sched_feat_keys[__SCHED_FEAT_NR] = {
#include "features.h"
};

#undef SCHED_FEAT

static void sched_feat_disable(int i)
{
	if (jump_label_enabled(&sched_feat_keys[i]))
		jump_label_dec(&sched_feat_keys[i]);
}

static void sched_feat_enable(int i)
{
	if (!jump_label_enabled(&sched_feat_keys[i]))
		jump_label_inc(&sched_feat_keys[i]);
}
#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
193 194 195 196 197
static ssize_t
sched_feat_write(struct file *filp, const char __user *ubuf,
		size_t cnt, loff_t *ppos)
{
	char buf[64];
198
	char *cmp;
P
Peter Zijlstra 已提交
199 200 201 202 203 204 205 206 207 208
	int neg = 0;
	int i;

	if (cnt > 63)
		cnt = 63;

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

	buf[cnt] = 0;
209
	cmp = strstrip(buf);
P
Peter Zijlstra 已提交
210

H
Hillf Danton 已提交
211
	if (strncmp(cmp, "NO_", 3) == 0) {
P
Peter Zijlstra 已提交
212 213 214 215
		neg = 1;
		cmp += 3;
	}

216
	for (i = 0; i < __SCHED_FEAT_NR; i++) {
217
		if (strcmp(cmp, sched_feat_names[i]) == 0) {
218
			if (neg) {
P
Peter Zijlstra 已提交
219
				sysctl_sched_features &= ~(1UL << i);
220 221
				sched_feat_disable(i);
			} else {
P
Peter Zijlstra 已提交
222
				sysctl_sched_features |= (1UL << i);
223 224
				sched_feat_enable(i);
			}
P
Peter Zijlstra 已提交
225 226 227 228
			break;
		}
	}

229
	if (i == __SCHED_FEAT_NR)
P
Peter Zijlstra 已提交
230 231
		return -EINVAL;

232
	*ppos += cnt;
P
Peter Zijlstra 已提交
233 234 235 236

	return cnt;
}

L
Li Zefan 已提交
237 238 239 240 241
static int sched_feat_open(struct inode *inode, struct file *filp)
{
	return single_open(filp, sched_feat_show, NULL);
}

242
static const struct file_operations sched_feat_fops = {
L
Li Zefan 已提交
243 244 245 246 247
	.open		= sched_feat_open,
	.write		= sched_feat_write,
	.read		= seq_read,
	.llseek		= seq_lseek,
	.release	= single_release,
P
Peter Zijlstra 已提交
248 249 250 251 252 253 254 255 256 257
};

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);
258
#endif /* CONFIG_SCHED_DEBUG */
I
Ingo Molnar 已提交
259

260 261 262 263 264 265
/*
 * 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;

266 267 268 269 270 271 272 273
/*
 * 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
274
/*
P
Peter Zijlstra 已提交
275
 * period over which we measure -rt task cpu usage in us.
P
Peter Zijlstra 已提交
276 277
 * default: 1s
 */
P
Peter Zijlstra 已提交
278
unsigned int sysctl_sched_rt_period = 1000000;
P
Peter Zijlstra 已提交
279

280
__read_mostly int scheduler_running;
281

P
Peter Zijlstra 已提交
282 283 284 285 286
/*
 * 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
287 288


L
Linus Torvalds 已提交
289

290
/*
291
 * __task_rq_lock - lock the rq @p resides on.
292
 */
293
static inline struct rq *__task_rq_lock(struct task_struct *p)
294 295
	__acquires(rq->lock)
{
296 297
	struct rq *rq;

298 299
	lockdep_assert_held(&p->pi_lock);

300
	for (;;) {
301
		rq = task_rq(p);
302
		raw_spin_lock(&rq->lock);
P
Peter Zijlstra 已提交
303
		if (likely(rq == task_rq(p)))
304
			return rq;
305
		raw_spin_unlock(&rq->lock);
306 307 308
	}
}

L
Linus Torvalds 已提交
309
/*
310
 * task_rq_lock - lock p->pi_lock and lock the rq @p resides on.
L
Linus Torvalds 已提交
311
 */
312
static struct rq *task_rq_lock(struct task_struct *p, unsigned long *flags)
313
	__acquires(p->pi_lock)
L
Linus Torvalds 已提交
314 315
	__acquires(rq->lock)
{
316
	struct rq *rq;
L
Linus Torvalds 已提交
317

318
	for (;;) {
319
		raw_spin_lock_irqsave(&p->pi_lock, *flags);
320
		rq = task_rq(p);
321
		raw_spin_lock(&rq->lock);
P
Peter Zijlstra 已提交
322
		if (likely(rq == task_rq(p)))
323
			return rq;
324 325
		raw_spin_unlock(&rq->lock);
		raw_spin_unlock_irqrestore(&p->pi_lock, *flags);
L
Linus Torvalds 已提交
326 327 328
	}
}

A
Alexey Dobriyan 已提交
329
static void __task_rq_unlock(struct rq *rq)
330 331
	__releases(rq->lock)
{
332
	raw_spin_unlock(&rq->lock);
333 334
}

335 336
static inline void
task_rq_unlock(struct rq *rq, struct task_struct *p, unsigned long *flags)
L
Linus Torvalds 已提交
337
	__releases(rq->lock)
338
	__releases(p->pi_lock)
L
Linus Torvalds 已提交
339
{
340 341
	raw_spin_unlock(&rq->lock);
	raw_spin_unlock_irqrestore(&p->pi_lock, *flags);
L
Linus Torvalds 已提交
342 343 344
}

/*
345
 * this_rq_lock - lock this runqueue and disable interrupts.
L
Linus Torvalds 已提交
346
 */
A
Alexey Dobriyan 已提交
347
static struct rq *this_rq_lock(void)
L
Linus Torvalds 已提交
348 349
	__acquires(rq->lock)
{
350
	struct rq *rq;
L
Linus Torvalds 已提交
351 352 353

	local_irq_disable();
	rq = this_rq();
354
	raw_spin_lock(&rq->lock);
L
Linus Torvalds 已提交
355 356 357 358

	return rq;
}

P
Peter Zijlstra 已提交
359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386
#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());

387
	raw_spin_lock(&rq->lock);
388
	update_rq_clock(rq);
P
Peter Zijlstra 已提交
389
	rq->curr->sched_class->task_tick(rq, rq->curr, 1);
390
	raw_spin_unlock(&rq->lock);
P
Peter Zijlstra 已提交
391 392 393 394

	return HRTIMER_NORESTART;
}

395
#ifdef CONFIG_SMP
396 397 398 399
/*
 * called from hardirq (IPI) context
 */
static void __hrtick_start(void *arg)
400
{
401
	struct rq *rq = arg;
402

403
	raw_spin_lock(&rq->lock);
404 405
	hrtimer_restart(&rq->hrtick_timer);
	rq->hrtick_csd_pending = 0;
406
	raw_spin_unlock(&rq->lock);
407 408
}

409 410 411 412 413
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
414
void hrtick_start(struct rq *rq, u64 delay)
415
{
416 417
	struct hrtimer *timer = &rq->hrtick_timer;
	ktime_t time = ktime_add_ns(timer->base->get_time(), delay);
418

419
	hrtimer_set_expires(timer, time);
420 421 422 423

	if (rq == this_rq()) {
		hrtimer_restart(timer);
	} else if (!rq->hrtick_csd_pending) {
424
		__smp_call_function_single(cpu_of(rq), &rq->hrtick_csd, 0);
425 426
		rq->hrtick_csd_pending = 1;
	}
427 428 429 430 431 432 433 434 435 436 437 438 439 440
}

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:
441
		hrtick_clear(cpu_rq(cpu));
442 443 444 445 446 447
		return NOTIFY_OK;
	}

	return NOTIFY_DONE;
}

448
static __init void init_hrtick(void)
449 450 451
{
	hotcpu_notifier(hotplug_hrtick, 0);
}
452 453 454 455 456 457
#else
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
458
void hrtick_start(struct rq *rq, u64 delay)
459
{
460
	__hrtimer_start_range_ns(&rq->hrtick_timer, ns_to_ktime(delay), 0,
461
			HRTIMER_MODE_REL_PINNED, 0);
462
}
463

A
Andrew Morton 已提交
464
static inline void init_hrtick(void)
P
Peter Zijlstra 已提交
465 466
{
}
467
#endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
468

469
static void init_rq_hrtick(struct rq *rq)
P
Peter Zijlstra 已提交
470
{
471 472
#ifdef CONFIG_SMP
	rq->hrtick_csd_pending = 0;
P
Peter Zijlstra 已提交
473

474 475 476 477
	rq->hrtick_csd.flags = 0;
	rq->hrtick_csd.func = __hrtick_start;
	rq->hrtick_csd.info = rq;
#endif
P
Peter Zijlstra 已提交
478

479 480
	hrtimer_init(&rq->hrtick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	rq->hrtick_timer.function = hrtick;
P
Peter Zijlstra 已提交
481
}
A
Andrew Morton 已提交
482
#else	/* CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
483 484 485 486 487 488 489 490
static inline void hrtick_clear(struct rq *rq)
{
}

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

491 492 493
static inline void init_hrtick(void)
{
}
A
Andrew Morton 已提交
494
#endif	/* CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
495

I
Ingo Molnar 已提交
496 497 498 499 500 501 502 503 504 505 506 507 508
/*
 * 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

509
void resched_task(struct task_struct *p)
I
Ingo Molnar 已提交
510 511 512
{
	int cpu;

513
	assert_raw_spin_locked(&task_rq(p)->lock);
I
Ingo Molnar 已提交
514

515
	if (test_tsk_need_resched(p))
I
Ingo Molnar 已提交
516 517
		return;

518
	set_tsk_need_resched(p);
I
Ingo Molnar 已提交
519 520 521 522 523 524 525 526 527 528 529

	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);
}

530
void resched_cpu(int cpu)
I
Ingo Molnar 已提交
531 532 533 534
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

535
	if (!raw_spin_trylock_irqsave(&rq->lock, flags))
I
Ingo Molnar 已提交
536 537
		return;
	resched_task(cpu_curr(cpu));
538
	raw_spin_unlock_irqrestore(&rq->lock, flags);
I
Ingo Molnar 已提交
539
}
540 541

#ifdef CONFIG_NO_HZ
542 543 544 545 546 547 548 549 550 551 552 553 554 555
/*
 * 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;

556
	rcu_read_lock();
557
	for_each_domain(cpu, sd) {
558 559 560 561 562 563
		for_each_cpu(i, sched_domain_span(sd)) {
			if (!idle_cpu(i)) {
				cpu = i;
				goto unlock;
			}
		}
564
	}
565 566
unlock:
	rcu_read_unlock();
567 568
	return cpu;
}
569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594
/*
 * 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;
595 596

	/*
597 598 599
	 * 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()
600
	 */
601
	set_tsk_need_resched(rq->idle);
602

603 604 605 606
	/* NEED_RESCHED must be visible before we test polling */
	smp_mb();
	if (!tsk_is_polling(rq->idle))
		smp_send_reschedule(cpu);
607 608
}

609
static inline bool got_nohz_idle_kick(void)
610
{
611 612
	int cpu = smp_processor_id();
	return idle_cpu(cpu) && test_bit(NOHZ_BALANCE_KICK, nohz_flags(cpu));
613 614
}

615
#else /* CONFIG_NO_HZ */
616

617
static inline bool got_nohz_idle_kick(void)
P
Peter Zijlstra 已提交
618
{
619
	return false;
P
Peter Zijlstra 已提交
620 621
}

622
#endif /* CONFIG_NO_HZ */
623

624
void sched_avg_update(struct rq *rq)
625
{
626 627 628
	s64 period = sched_avg_period();

	while ((s64)(rq->clock - rq->age_stamp) > period) {
629 630 631 632 633 634
		/*
		 * 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));
635 636 637
		rq->age_stamp += period;
		rq->rt_avg /= 2;
	}
638 639
}

640
#else /* !CONFIG_SMP */
641
void resched_task(struct task_struct *p)
642
{
643
	assert_raw_spin_locked(&task_rq(p)->lock);
644
	set_tsk_need_resched(p);
645
}
646
#endif /* CONFIG_SMP */
647

648 649
#if defined(CONFIG_RT_GROUP_SCHED) || (defined(CONFIG_FAIR_GROUP_SCHED) && \
			(defined(CONFIG_SMP) || defined(CONFIG_CFS_BANDWIDTH)))
650
/*
651 652 653 654
 * 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.
655
 */
656
int walk_tg_tree_from(struct task_group *from,
657
			     tg_visitor down, tg_visitor up, void *data)
658 659
{
	struct task_group *parent, *child;
P
Peter Zijlstra 已提交
660
	int ret;
661

662 663
	parent = from;

664
down:
P
Peter Zijlstra 已提交
665 666
	ret = (*down)(parent, data);
	if (ret)
667
		goto out;
668 669 670 671 672 673 674
	list_for_each_entry_rcu(child, &parent->children, siblings) {
		parent = child;
		goto down;

up:
		continue;
	}
P
Peter Zijlstra 已提交
675
	ret = (*up)(parent, data);
676 677
	if (ret || parent == from)
		goto out;
678 679 680 681 682

	child = parent;
	parent = parent->parent;
	if (parent)
		goto up;
683
out:
P
Peter Zijlstra 已提交
684
	return ret;
685 686
}

687
int tg_nop(struct task_group *tg, void *data)
P
Peter Zijlstra 已提交
688
{
689
	return 0;
P
Peter Zijlstra 已提交
690
}
691 692
#endif

693
void update_cpu_load(struct rq *this_rq);
694

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;
895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914
#endif
#ifdef CONFIG_PARAVIRT_TIME_ACCOUNTING
	if (static_branch((&paravirt_steal_rq_enabled))) {
		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 1267 1268 1269 1270
static int select_fallback_rq(int cpu, struct task_struct *p)
{
	int dest_cpu;
	const struct cpumask *nodemask = cpumask_of_node(cpu_to_node(cpu));

	/* Look for allowed, online CPU in same node. */
	for_each_cpu_and(dest_cpu, nodemask, cpu_active_mask)
1271
		if (cpumask_test_cpu(dest_cpu, tsk_cpus_allowed(p)))
1272 1273 1274
			return dest_cpu;

	/* Any allowed, online CPU? */
1275
	dest_cpu = cpumask_any_and(tsk_cpus_allowed(p), cpu_active_mask);
1276 1277 1278 1279
	if (dest_cpu < nr_cpu_ids)
		return dest_cpu;

	/* No more Mr. Nice Guy. */
1280 1281 1282 1283 1284 1285 1286
	dest_cpu = cpuset_cpus_allowed_fallback(p);
	/*
	 * Don't tell them about moving exiting tasks or
	 * kernel threads (both mm NULL), since they never
	 * leave kernel.
	 */
	if (p->mm && printk_ratelimit()) {
1287
		printk_sched("process %d (%s) no longer affine to cpu%d\n",
1288
				task_pid_nr(p), p->comm, cpu);
1289 1290 1291 1292 1293
	}

	return dest_cpu;
}

1294
/*
1295
 * The caller (fork, wakeup) owns p->pi_lock, ->cpus_allowed is stable.
1296
 */
1297
static inline
1298
int select_task_rq(struct task_struct *p, int sd_flags, int wake_flags)
1299
{
1300
	int cpu = p->sched_class->select_task_rq(p, sd_flags, wake_flags);
1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311

	/*
	 * 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 ]
	 */
1312
	if (unlikely(!cpumask_test_cpu(cpu, tsk_cpus_allowed(p)) ||
P
Peter Zijlstra 已提交
1313
		     !cpu_online(cpu)))
1314
		cpu = select_fallback_rq(task_cpu(p), p);
1315 1316

	return cpu;
1317
}
1318 1319 1320 1321 1322 1323

static void update_avg(u64 *avg, u64 sample)
{
	s64 diff = sample - *avg;
	*avg += diff >> 3;
}
1324 1325
#endif

P
Peter Zijlstra 已提交
1326
static void
1327
ttwu_stat(struct task_struct *p, int cpu, int wake_flags)
T
Tejun Heo 已提交
1328
{
P
Peter Zijlstra 已提交
1329
#ifdef CONFIG_SCHEDSTATS
1330 1331
	struct rq *rq = this_rq();

P
Peter Zijlstra 已提交
1332 1333 1334 1335 1336 1337 1338 1339 1340 1341
#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);
1342
		rcu_read_lock();
P
Peter Zijlstra 已提交
1343 1344 1345 1346 1347 1348
		for_each_domain(this_cpu, sd) {
			if (cpumask_test_cpu(cpu, sched_domain_span(sd))) {
				schedstat_inc(sd, ttwu_wake_remote);
				break;
			}
		}
1349
		rcu_read_unlock();
P
Peter Zijlstra 已提交
1350
	}
1351 1352 1353 1354

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

P
Peter Zijlstra 已提交
1355 1356 1357
#endif /* CONFIG_SMP */

	schedstat_inc(rq, ttwu_count);
T
Tejun Heo 已提交
1358
	schedstat_inc(p, se.statistics.nr_wakeups);
P
Peter Zijlstra 已提交
1359 1360

	if (wake_flags & WF_SYNC)
T
Tejun Heo 已提交
1361
		schedstat_inc(p, se.statistics.nr_wakeups_sync);
P
Peter Zijlstra 已提交
1362 1363 1364 1365 1366 1367

#endif /* CONFIG_SCHEDSTATS */
}

static void ttwu_activate(struct rq *rq, struct task_struct *p, int en_flags)
{
T
Tejun Heo 已提交
1368
	activate_task(rq, p, en_flags);
P
Peter Zijlstra 已提交
1369
	p->on_rq = 1;
1370 1371 1372 1373

	/* 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
1374 1375
}

1376 1377 1378
/*
 * Mark the task runnable and perform wakeup-preemption.
 */
1379
static void
1380
ttwu_do_wakeup(struct rq *rq, struct task_struct *p, int wake_flags)
T
Tejun Heo 已提交
1381
{
1382
	trace_sched_wakeup(p, true);
T
Tejun Heo 已提交
1383 1384 1385 1386 1387 1388 1389
	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);

1390
	if (rq->idle_stamp) {
T
Tejun Heo 已提交
1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402
		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
}

1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435
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;
}

1436
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
1437
static void sched_ttwu_pending(void)
1438 1439
{
	struct rq *rq = this_rq();
P
Peter Zijlstra 已提交
1440 1441
	struct llist_node *llist = llist_del_all(&rq->wake_list);
	struct task_struct *p;
1442 1443 1444

	raw_spin_lock(&rq->lock);

P
Peter Zijlstra 已提交
1445 1446 1447
	while (llist) {
		p = llist_entry(llist, struct task_struct, wake_entry);
		llist = llist_next(llist);
1448 1449 1450 1451 1452 1453 1454 1455
		ttwu_do_activate(rq, p, 0);
	}

	raw_spin_unlock(&rq->lock);
}

void scheduler_ipi(void)
{
1456
	if (llist_empty(&this_rq()->wake_list) && !got_nohz_idle_kick())
1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472
		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
1473
	sched_ttwu_pending();
1474 1475 1476 1477

	/*
	 * Check if someone kicked us for doing the nohz idle load balance.
	 */
1478 1479
	if (unlikely(got_nohz_idle_kick() && !need_resched())) {
		this_rq()->idle_balance = 1;
1480
		raise_softirq_irqoff(SCHED_SOFTIRQ);
1481
	}
1482
	irq_exit();
1483 1484 1485 1486
}

static void ttwu_queue_remote(struct task_struct *p, int cpu)
{
P
Peter Zijlstra 已提交
1487
	if (llist_add(&p->wake_entry, &cpu_rq(cpu)->wake_list))
1488 1489
		smp_send_reschedule(cpu);
}
1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508

#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 */
1509

1510
bool cpus_share_cache(int this_cpu, int that_cpu)
1511 1512 1513
{
	return per_cpu(sd_llc_id, this_cpu) == per_cpu(sd_llc_id, that_cpu);
}
1514
#endif /* CONFIG_SMP */
1515

1516 1517 1518 1519
static void ttwu_queue(struct task_struct *p, int cpu)
{
	struct rq *rq = cpu_rq(cpu);

1520
#if defined(CONFIG_SMP)
1521
	if (sched_feat(TTWU_QUEUE) && !cpus_share_cache(smp_processor_id(), cpu)) {
1522
		sched_clock_cpu(cpu); /* sync clocks x-cpu */
1523 1524 1525 1526 1527
		ttwu_queue_remote(p, cpu);
		return;
	}
#endif

1528 1529 1530
	raw_spin_lock(&rq->lock);
	ttwu_do_activate(rq, p, 0);
	raw_spin_unlock(&rq->lock);
T
Tejun Heo 已提交
1531 1532 1533
}

/**
L
Linus Torvalds 已提交
1534
 * try_to_wake_up - wake up a thread
T
Tejun Heo 已提交
1535
 * @p: the thread to be awakened
L
Linus Torvalds 已提交
1536
 * @state: the mask of task states that can be woken
T
Tejun Heo 已提交
1537
 * @wake_flags: wake modifier flags (WF_*)
L
Linus Torvalds 已提交
1538 1539 1540 1541 1542 1543 1544
 *
 * 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
1545 1546
 * 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
1547
 */
1548 1549
static int
try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
L
Linus Torvalds 已提交
1550 1551
{
	unsigned long flags;
1552
	int cpu, success = 0;
P
Peter Zijlstra 已提交
1553

1554
	smp_wmb();
1555
	raw_spin_lock_irqsave(&p->pi_lock, flags);
P
Peter Zijlstra 已提交
1556
	if (!(p->state & state))
L
Linus Torvalds 已提交
1557 1558
		goto out;

1559
	success = 1; /* we're going to change ->state */
L
Linus Torvalds 已提交
1560 1561
	cpu = task_cpu(p);

1562 1563
	if (p->on_rq && ttwu_remote(p, wake_flags))
		goto stat;
L
Linus Torvalds 已提交
1564 1565

#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
1566
	/*
1567 1568
	 * 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
1569
	 */
1570 1571 1572
	while (p->on_cpu) {
#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
		/*
1573 1574 1575 1576 1577
		 * 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.
1578
		 */
1579
		if (ttwu_activate_remote(p, wake_flags))
1580
			goto stat;
1581
#else
1582
		cpu_relax();
1583
#endif
1584
	}
1585
	/*
1586
	 * Pairs with the smp_wmb() in finish_lock_switch().
1587
	 */
1588
	smp_rmb();
L
Linus Torvalds 已提交
1589

1590
	p->sched_contributes_to_load = !!task_contributes_to_load(p);
P
Peter Zijlstra 已提交
1591
	p->state = TASK_WAKING;
1592

1593
	if (p->sched_class->task_waking)
1594
		p->sched_class->task_waking(p);
1595

1596
	cpu = select_task_rq(p, SD_BALANCE_WAKE, wake_flags);
1597 1598
	if (task_cpu(p) != cpu) {
		wake_flags |= WF_MIGRATED;
1599
		set_task_cpu(p, cpu);
1600
	}
L
Linus Torvalds 已提交
1601 1602
#endif /* CONFIG_SMP */

1603 1604
	ttwu_queue(p, cpu);
stat:
1605
	ttwu_stat(p, cpu, wake_flags);
L
Linus Torvalds 已提交
1606
out:
1607
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
1608 1609 1610 1611

	return success;
}

T
Tejun Heo 已提交
1612 1613 1614 1615
/**
 * try_to_wake_up_local - try to wake up a local task with rq lock held
 * @p: the thread to be awakened
 *
1616
 * Put @p on the run-queue if it's not already there. The caller must
T
Tejun Heo 已提交
1617
 * ensure that this_rq() is locked, @p is bound to this_rq() and not
1618
 * the current task.
T
Tejun Heo 已提交
1619 1620 1621 1622 1623 1624 1625 1626 1627
 */
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);

1628 1629 1630 1631 1632 1633
	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
1634
	if (!(p->state & TASK_NORMAL))
1635
		goto out;
T
Tejun Heo 已提交
1636

P
Peter Zijlstra 已提交
1637
	if (!p->on_rq)
P
Peter Zijlstra 已提交
1638 1639
		ttwu_activate(rq, p, ENQUEUE_WAKEUP);

1640
	ttwu_do_wakeup(rq, p, 0);
1641
	ttwu_stat(p, smp_processor_id(), 0);
1642 1643
out:
	raw_spin_unlock(&p->pi_lock);
T
Tejun Heo 已提交
1644 1645
}

1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656
/**
 * 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.
 */
1657
int wake_up_process(struct task_struct *p)
L
Linus Torvalds 已提交
1658
{
1659
	return try_to_wake_up(p, TASK_ALL, 0);
L
Linus Torvalds 已提交
1660 1661 1662
}
EXPORT_SYMBOL(wake_up_process);

1663
int wake_up_state(struct task_struct *p, unsigned int state)
L
Linus Torvalds 已提交
1664 1665 1666 1667 1668 1669 1670
{
	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
1671 1672 1673 1674 1675
 *
 * __sched_fork() is basic setup used by init_idle() too:
 */
static void __sched_fork(struct task_struct *p)
{
P
Peter Zijlstra 已提交
1676 1677 1678
	p->on_rq			= 0;

	p->se.on_rq			= 0;
I
Ingo Molnar 已提交
1679 1680
	p->se.exec_start		= 0;
	p->se.sum_exec_runtime		= 0;
1681
	p->se.prev_sum_exec_runtime	= 0;
1682
	p->se.nr_migrations		= 0;
P
Peter Zijlstra 已提交
1683
	p->se.vruntime			= 0;
P
Peter Zijlstra 已提交
1684
	INIT_LIST_HEAD(&p->se.group_node);
I
Ingo Molnar 已提交
1685 1686

#ifdef CONFIG_SCHEDSTATS
1687
	memset(&p->se.statistics, 0, sizeof(p->se.statistics));
I
Ingo Molnar 已提交
1688
#endif
N
Nick Piggin 已提交
1689

P
Peter Zijlstra 已提交
1690
	INIT_LIST_HEAD(&p->rt.run_list);
N
Nick Piggin 已提交
1691

1692 1693 1694
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&p->preempt_notifiers);
#endif
I
Ingo Molnar 已提交
1695 1696 1697 1698 1699
}

/*
 * fork()/clone()-time setup:
 */
1700
void sched_fork(struct task_struct *p)
I
Ingo Molnar 已提交
1701
{
1702
	unsigned long flags;
I
Ingo Molnar 已提交
1703 1704 1705
	int cpu = get_cpu();

	__sched_fork(p);
1706
	/*
1707
	 * We mark the process as running here. This guarantees that
1708 1709 1710
	 * nobody will actually run it, and a signal or other external
	 * event cannot wake it up and insert it on the runqueue either.
	 */
1711
	p->state = TASK_RUNNING;
I
Ingo Molnar 已提交
1712

1713 1714 1715 1716 1717
	/*
	 * Make sure we do not leak PI boosting priority to the child.
	 */
	p->prio = current->normal_prio;

1718 1719 1720 1721
	/*
	 * Revert to default priority/policy on fork if requested.
	 */
	if (unlikely(p->sched_reset_on_fork)) {
1722
		if (task_has_rt_policy(p)) {
1723
			p->policy = SCHED_NORMAL;
1724
			p->static_prio = NICE_TO_PRIO(0);
1725 1726 1727 1728 1729 1730
			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);
1731

1732 1733 1734 1735 1736 1737
		/*
		 * We don't need the reset flag anymore after the fork. It has
		 * fulfilled its duty:
		 */
		p->sched_reset_on_fork = 0;
	}
1738

H
Hiroshi Shimamoto 已提交
1739 1740
	if (!rt_prio(p->prio))
		p->sched_class = &fair_sched_class;
1741

P
Peter Zijlstra 已提交
1742 1743 1744
	if (p->sched_class->task_fork)
		p->sched_class->task_fork(p);

1745 1746 1747 1748 1749 1750 1751
	/*
	 * 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.
	 */
1752
	raw_spin_lock_irqsave(&p->pi_lock, flags);
1753
	set_task_cpu(p, cpu);
1754
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
1755

1756
#if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT)
I
Ingo Molnar 已提交
1757
	if (likely(sched_info_on()))
1758
		memset(&p->sched_info, 0, sizeof(p->sched_info));
L
Linus Torvalds 已提交
1759
#endif
P
Peter Zijlstra 已提交
1760 1761
#if defined(CONFIG_SMP)
	p->on_cpu = 0;
1762
#endif
1763
#ifdef CONFIG_PREEMPT_COUNT
1764
	/* Want to start with kernel preemption disabled. */
A
Al Viro 已提交
1765
	task_thread_info(p)->preempt_count = 1;
L
Linus Torvalds 已提交
1766
#endif
1767
#ifdef CONFIG_SMP
1768
	plist_node_init(&p->pushable_tasks, MAX_PRIO);
1769
#endif
1770

N
Nick Piggin 已提交
1771
	put_cpu();
L
Linus Torvalds 已提交
1772 1773 1774 1775 1776 1777 1778 1779 1780
}

/*
 * 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.
 */
1781
void wake_up_new_task(struct task_struct *p)
L
Linus Torvalds 已提交
1782 1783
{
	unsigned long flags;
I
Ingo Molnar 已提交
1784
	struct rq *rq;
1785

1786
	raw_spin_lock_irqsave(&p->pi_lock, flags);
1787 1788 1789 1790 1791 1792
#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
	 */
1793
	set_task_cpu(p, select_task_rq(p, SD_BALANCE_FORK, 0));
1794 1795
#endif

1796
	rq = __task_rq_lock(p);
P
Peter Zijlstra 已提交
1797
	activate_task(rq, p, 0);
P
Peter Zijlstra 已提交
1798
	p->on_rq = 1;
1799
	trace_sched_wakeup_new(p, true);
P
Peter Zijlstra 已提交
1800
	check_preempt_curr(rq, p, WF_FORK);
1801
#ifdef CONFIG_SMP
1802 1803
	if (p->sched_class->task_woken)
		p->sched_class->task_woken(rq, p);
1804
#endif
1805
	task_rq_unlock(rq, p, &flags);
L
Linus Torvalds 已提交
1806 1807
}

1808 1809 1810
#ifdef CONFIG_PREEMPT_NOTIFIERS

/**
1811
 * preempt_notifier_register - tell me when current is being preempted & rescheduled
R
Randy Dunlap 已提交
1812
 * @notifier: notifier struct to register
1813 1814 1815 1816 1817 1818 1819 1820 1821
 */
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
1822
 * @notifier: notifier struct to unregister
1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851
 *
 * 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);
}

1852
#else /* !CONFIG_PREEMPT_NOTIFIERS */
1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863

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)
{
}

1864
#endif /* CONFIG_PREEMPT_NOTIFIERS */
1865

1866 1867 1868
/**
 * prepare_task_switch - prepare to switch tasks
 * @rq: the runqueue preparing to switch
R
Randy Dunlap 已提交
1869
 * @prev: the current task that is being switched out
1870 1871 1872 1873 1874 1875 1876 1877 1878
 * @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.
 */
1879 1880 1881
static inline void
prepare_task_switch(struct rq *rq, struct task_struct *prev,
		    struct task_struct *next)
1882
{
1883 1884
	sched_info_switch(prev, next);
	perf_event_task_sched_out(prev, next);
1885
	fire_sched_out_preempt_notifiers(prev, next);
1886 1887
	prepare_lock_switch(rq, next);
	prepare_arch_switch(next);
1888
	trace_sched_switch(prev, next);
1889 1890
}

L
Linus Torvalds 已提交
1891 1892
/**
 * finish_task_switch - clean up after a task-switch
1893
 * @rq: runqueue associated with task-switch
L
Linus Torvalds 已提交
1894 1895
 * @prev: the thread we just switched away from.
 *
1896 1897 1898 1899
 * 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
1900 1901
 *
 * Note that we may have delayed dropping an mm in context_switch(). If
I
Ingo Molnar 已提交
1902
 * so, we finish that here outside of the runqueue lock. (Doing it
L
Linus Torvalds 已提交
1903 1904 1905
 * with the lock held can cause deadlocks; see schedule() for
 * details.)
 */
A
Alexey Dobriyan 已提交
1906
static void finish_task_switch(struct rq *rq, struct task_struct *prev)
L
Linus Torvalds 已提交
1907 1908 1909
	__releases(rq->lock)
{
	struct mm_struct *mm = rq->prev_mm;
O
Oleg Nesterov 已提交
1910
	long prev_state;
L
Linus Torvalds 已提交
1911 1912 1913 1914 1915

	rq->prev_mm = NULL;

	/*
	 * A task struct has one reference for the use as "current".
1916
	 * If a task dies, then it sets TASK_DEAD in tsk->state and calls
O
Oleg Nesterov 已提交
1917 1918
	 * schedule one last time. The schedule call will never return, and
	 * the scheduled task must drop that reference.
1919
	 * The test for TASK_DEAD must occur while the runqueue locks are
L
Linus Torvalds 已提交
1920 1921 1922 1923 1924
	 * 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
1925
	prev_state = prev->state;
1926
	finish_arch_switch(prev);
1927 1928 1929
#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
	local_irq_disable();
#endif /* __ARCH_WANT_INTERRUPTS_ON_CTXSW */
1930
	perf_event_task_sched_in(prev, current);
1931 1932 1933
#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
	local_irq_enable();
#endif /* __ARCH_WANT_INTERRUPTS_ON_CTXSW */
1934
	finish_lock_switch(rq, prev);
S
Steven Rostedt 已提交
1935

1936
	fire_sched_in_preempt_notifiers(current);
L
Linus Torvalds 已提交
1937 1938
	if (mm)
		mmdrop(mm);
1939
	if (unlikely(prev_state == TASK_DEAD)) {
1940 1941 1942
		/*
		 * Remove function-return probe instances associated with this
		 * task and put them back on the free list.
I
Ingo Molnar 已提交
1943
		 */
1944
		kprobe_flush_task(prev);
L
Linus Torvalds 已提交
1945
		put_task_struct(prev);
1946
	}
L
Linus Torvalds 已提交
1947 1948
}

1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963
#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;

1964
		raw_spin_lock_irqsave(&rq->lock, flags);
1965 1966
		if (rq->curr->sched_class->post_schedule)
			rq->curr->sched_class->post_schedule(rq);
1967
		raw_spin_unlock_irqrestore(&rq->lock, flags);
1968 1969 1970 1971 1972 1973

		rq->post_schedule = 0;
	}
}

#else
1974

1975 1976 1977 1978 1979 1980
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
1981 1982
}

1983 1984
#endif

L
Linus Torvalds 已提交
1985 1986 1987 1988
/**
 * schedule_tail - first thing a freshly forked thread must call.
 * @prev: the thread we just switched away from.
 */
1989
asmlinkage void schedule_tail(struct task_struct *prev)
L
Linus Torvalds 已提交
1990 1991
	__releases(rq->lock)
{
1992 1993
	struct rq *rq = this_rq();

1994
	finish_task_switch(rq, prev);
1995

1996 1997 1998 1999 2000
	/*
	 * FIXME: do we need to worry about rq being invalidated by the
	 * task_switch?
	 */
	post_schedule(rq);
2001

2002 2003 2004 2005
#ifdef __ARCH_WANT_UNLOCKED_CTXSW
	/* In this case, finish_task_switch does not reenable preemption */
	preempt_enable();
#endif
L
Linus Torvalds 已提交
2006
	if (current->set_child_tid)
2007
		put_user(task_pid_vnr(current), current->set_child_tid);
L
Linus Torvalds 已提交
2008 2009 2010 2011 2012 2013
}

/*
 * context_switch - switch to the new MM and the new
 * thread's register state.
 */
I
Ingo Molnar 已提交
2014
static inline void
2015
context_switch(struct rq *rq, struct task_struct *prev,
2016
	       struct task_struct *next)
L
Linus Torvalds 已提交
2017
{
I
Ingo Molnar 已提交
2018
	struct mm_struct *mm, *oldmm;
L
Linus Torvalds 已提交
2019

2020
	prepare_task_switch(rq, prev, next);
2021

I
Ingo Molnar 已提交
2022 2023
	mm = next->mm;
	oldmm = prev->active_mm;
2024 2025 2026 2027 2028
	/*
	 * For paravirt, this is coupled with an exit in switch_to to
	 * combine the page table reload and the switch backend into
	 * one hypercall.
	 */
2029
	arch_start_context_switch(prev);
2030

2031
	if (!mm) {
L
Linus Torvalds 已提交
2032 2033 2034 2035 2036 2037
		next->active_mm = oldmm;
		atomic_inc(&oldmm->mm_count);
		enter_lazy_tlb(oldmm, next);
	} else
		switch_mm(oldmm, mm, next);

2038
	if (!prev->mm) {
L
Linus Torvalds 已提交
2039 2040 2041
		prev->active_mm = NULL;
		rq->prev_mm = oldmm;
	}
2042 2043 2044 2045 2046 2047 2048
	/*
	 * 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
2049
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
2050
#endif
L
Linus Torvalds 已提交
2051 2052 2053 2054

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

I
Ingo Molnar 已提交
2055 2056 2057 2058 2059 2060 2061
	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
2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078
}

/*
 * 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;
2079
}
L
Linus Torvalds 已提交
2080 2081

unsigned long nr_uninterruptible(void)
2082
{
L
Linus Torvalds 已提交
2083
	unsigned long i, sum = 0;
2084

2085
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2086
		sum += cpu_rq(i)->nr_uninterruptible;
2087 2088

	/*
L
Linus Torvalds 已提交
2089 2090
	 * Since we read the counters lockless, it might be slightly
	 * inaccurate. Do not allow it to go below zero though:
2091
	 */
L
Linus Torvalds 已提交
2092 2093
	if (unlikely((long)sum < 0))
		sum = 0;
2094

L
Linus Torvalds 已提交
2095
	return sum;
2096 2097
}

L
Linus Torvalds 已提交
2098
unsigned long long nr_context_switches(void)
2099
{
2100 2101
	int i;
	unsigned long long sum = 0;
2102

2103
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2104
		sum += cpu_rq(i)->nr_switches;
2105

L
Linus Torvalds 已提交
2106 2107
	return sum;
}
2108

L
Linus Torvalds 已提交
2109 2110 2111
unsigned long nr_iowait(void)
{
	unsigned long i, sum = 0;
2112

2113
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2114
		sum += atomic_read(&cpu_rq(i)->nr_iowait);
2115

L
Linus Torvalds 已提交
2116 2117
	return sum;
}
2118

2119
unsigned long nr_iowait_cpu(int cpu)
2120
{
2121
	struct rq *this = cpu_rq(cpu);
2122 2123
	return atomic_read(&this->nr_iowait);
}
2124

2125 2126 2127 2128 2129
unsigned long this_cpu_load(void)
{
	struct rq *this = this_rq();
	return this->cpu_load[0];
}
2130

2131

2132 2133 2134 2135 2136
/* Variables and functions for calc_load */
static atomic_long_t calc_load_tasks;
static unsigned long calc_load_update;
unsigned long avenrun[3];
EXPORT_SYMBOL(avenrun);
2137

2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152
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;
}

2153 2154 2155 2156 2157 2158 2159 2160 2161
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;
}

2162 2163 2164 2165 2166 2167 2168 2169
#ifdef CONFIG_NO_HZ
/*
 * For NO_HZ we delay the active fold to the next LOAD_FREQ update.
 *
 * When making the ILB scale, we should try to pull this in as well.
 */
static atomic_long_t calc_load_tasks_idle;

2170
void calc_load_account_idle(struct rq *this_rq)
2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190
{
	long delta;

	delta = calc_load_fold_active(this_rq);
	if (delta)
		atomic_long_add(delta, &calc_load_tasks_idle);
}

static long calc_load_fold_idle(void)
{
	long delta = 0;

	/*
	 * Its got a race, we don't care...
	 */
	if (atomic_long_read(&calc_load_tasks_idle))
		delta = atomic_long_xchg(&calc_load_tasks_idle, 0);

	return delta;
}
2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 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 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312

/**
 * 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.
 */
static void calc_global_nohz(unsigned long ticks)
{
	long delta, active, n;

	if (time_before(jiffies, calc_load_update))
		return;

	/*
	 * If we crossed a calc_load_update boundary, make sure to fold
	 * any pending idle changes, the respective CPUs might have
	 * missed the tick driven calc_load_account_active() update
	 * due to NO_HZ.
	 */
	delta = calc_load_fold_idle();
	if (delta)
		atomic_long_add(delta, &calc_load_tasks);

	/*
	 * If we were idle for multiple load cycles, apply them.
	 */
	if (ticks >= LOAD_FREQ) {
		n = ticks / LOAD_FREQ;

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

		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);

		calc_load_update += n * LOAD_FREQ;
	}

	/*
	 * Its possible the remainder of the above division also crosses
	 * a LOAD_FREQ period, the regular check in calc_global_load()
	 * which comes after this will take care of that.
	 *
	 * Consider us being 11 ticks before a cycle completion, and us
	 * sleeping for 4*LOAD_FREQ + 22 ticks, then the above code will
	 * age us 4 cycles, and the test in calc_global_load() will
	 * pick up the final one.
	 */
}
2313
#else
2314
void calc_load_account_idle(struct rq *this_rq)
2315 2316 2317 2318 2319 2320 2321
{
}

static inline long calc_load_fold_idle(void)
{
	return 0;
}
2322 2323 2324 2325

static void calc_global_nohz(unsigned long ticks)
{
}
2326 2327
#endif

2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340
/**
 * 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;
2341 2342 2343
}

/*
2344 2345
 * calc_load - update the avenrun load estimates 10 ticks after the
 * CPUs have updated calc_load_tasks.
2346
 */
2347
void calc_global_load(unsigned long ticks)
2348
{
2349
	long active;
L
Linus Torvalds 已提交
2350

2351 2352 2353
	calc_global_nohz(ticks);

	if (time_before(jiffies, calc_load_update + 10))
2354
		return;
L
Linus Torvalds 已提交
2355

2356 2357
	active = atomic_long_read(&calc_load_tasks);
	active = active > 0 ? active * FIXED_1 : 0;
L
Linus Torvalds 已提交
2358

2359 2360 2361
	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
2362

2363 2364
	calc_load_update += LOAD_FREQ;
}
L
Linus Torvalds 已提交
2365

2366
/*
2367 2368
 * Called from update_cpu_load() to periodically update this CPU's
 * active count.
2369 2370 2371
 */
static void calc_load_account_active(struct rq *this_rq)
{
2372
	long delta;
2373

2374 2375
	if (time_before(jiffies, this_rq->calc_load_update))
		return;
2376

2377 2378 2379
	delta  = calc_load_fold_active(this_rq);
	delta += calc_load_fold_idle();
	if (delta)
2380
		atomic_long_add(delta, &calc_load_tasks);
2381 2382

	this_rq->calc_load_update += LOAD_FREQ;
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
/*
 * 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;
}

2452
/*
I
Ingo Molnar 已提交
2453
 * Update rq->cpu_load[] statistics. This function is usually called every
2454 2455
 * scheduler tick (TICK_NSEC). With tickless idle this will not be called
 * every tick. We fix it up based on jiffies.
2456
 */
2457
void update_cpu_load(struct rq *this_rq)
2458
{
2459
	unsigned long this_load = this_rq->load.weight;
2460 2461
	unsigned long curr_jiffies = jiffies;
	unsigned long pending_updates;
I
Ingo Molnar 已提交
2462
	int i, scale;
2463

I
Ingo Molnar 已提交
2464
	this_rq->nr_load_updates++;
2465

2466 2467 2468 2469 2470 2471 2472
	/* Avoid repeated calls on same jiffy, when moving in and out of idle */
	if (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;

I
Ingo Molnar 已提交
2473
	/* Update our load: */
2474 2475
	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
2476
		unsigned long old_load, new_load;
2477

I
Ingo Molnar 已提交
2478
		/* scale is effectively 1 << i now, and >> i divides by scale */
2479

I
Ingo Molnar 已提交
2480
		old_load = this_rq->cpu_load[i];
2481
		old_load = decay_load_missed(old_load, pending_updates - 1, i);
I
Ingo Molnar 已提交
2482
		new_load = this_load;
I
Ingo Molnar 已提交
2483 2484 2485 2486 2487 2488
		/*
		 * 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)
2489 2490 2491
			new_load += scale - 1;

		this_rq->cpu_load[i] = (old_load * (scale - 1) + new_load) >> i;
I
Ingo Molnar 已提交
2492
	}
2493 2494

	sched_avg_update(this_rq);
2495 2496 2497 2498 2499
}

static void update_cpu_load_active(struct rq *this_rq)
{
	update_cpu_load(this_rq);
2500

2501
	calc_load_account_active(this_rq);
2502 2503
}

I
Ingo Molnar 已提交
2504
#ifdef CONFIG_SMP
2505

2506
/*
P
Peter Zijlstra 已提交
2507 2508
 * sched_exec - execve() is a valuable balancing opportunity, because at
 * this point the task has the smallest effective memory and cache footprint.
2509
 */
P
Peter Zijlstra 已提交
2510
void sched_exec(void)
2511
{
P
Peter Zijlstra 已提交
2512
	struct task_struct *p = current;
L
Linus Torvalds 已提交
2513
	unsigned long flags;
2514
	int dest_cpu;
2515

2516
	raw_spin_lock_irqsave(&p->pi_lock, flags);
2517
	dest_cpu = p->sched_class->select_task_rq(p, SD_BALANCE_EXEC, 0);
2518 2519
	if (dest_cpu == smp_processor_id())
		goto unlock;
P
Peter Zijlstra 已提交
2520

2521
	if (likely(cpu_active(dest_cpu))) {
2522
		struct migration_arg arg = { p, dest_cpu };
2523

2524 2525
		raw_spin_unlock_irqrestore(&p->pi_lock, flags);
		stop_one_cpu(task_cpu(p), migration_cpu_stop, &arg);
L
Linus Torvalds 已提交
2526 2527
		return;
	}
2528
unlock:
2529
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
2530
}
I
Ingo Molnar 已提交
2531

L
Linus Torvalds 已提交
2532 2533 2534
#endif

DEFINE_PER_CPU(struct kernel_stat, kstat);
2535
DEFINE_PER_CPU(struct kernel_cpustat, kernel_cpustat);
L
Linus Torvalds 已提交
2536 2537

EXPORT_PER_CPU_SYMBOL(kstat);
2538
EXPORT_PER_CPU_SYMBOL(kernel_cpustat);
L
Linus Torvalds 已提交
2539 2540

/*
2541
 * Return any ns on the sched_clock that have not yet been accounted in
2542
 * @p in case that task is currently running.
2543 2544
 *
 * Called with task_rq_lock() held on @rq.
L
Linus Torvalds 已提交
2545
 */
2546 2547 2548 2549 2550 2551
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);
2552
		ns = rq->clock_task - p->se.exec_start;
2553 2554 2555 2556 2557 2558 2559
		if ((s64)ns < 0)
			ns = 0;
	}

	return ns;
}

2560
unsigned long long task_delta_exec(struct task_struct *p)
L
Linus Torvalds 已提交
2561 2562
{
	unsigned long flags;
2563
	struct rq *rq;
2564
	u64 ns = 0;
2565

2566
	rq = task_rq_lock(p, &flags);
2567
	ns = do_task_delta_exec(p, rq);
2568
	task_rq_unlock(rq, p, &flags);
2569

2570 2571
	return ns;
}
2572

2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585
/*
 * 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);
2586
	task_rq_unlock(rq, p, &flags);
2587 2588 2589

	return ns;
}
2590

2591 2592 2593 2594 2595
#ifdef CONFIG_CGROUP_CPUACCT
struct cgroup_subsys cpuacct_subsys;
struct cpuacct root_cpuacct;
#endif

2596 2597
static inline void task_group_account_field(struct task_struct *p, int index,
					    u64 tmp)
2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626
{
#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
2627 2628 2629 2630
/*
 * 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
2631
 * @cputime_scaled: cputime scaled by cpu frequency
L
Linus Torvalds 已提交
2632
 */
2633 2634
void account_user_time(struct task_struct *p, cputime_t cputime,
		       cputime_t cputime_scaled)
L
Linus Torvalds 已提交
2635
{
2636
	int index;
L
Linus Torvalds 已提交
2637

2638
	/* Add user time to process. */
2639 2640
	p->utime += cputime;
	p->utimescaled += cputime_scaled;
2641
	account_group_user_time(p, cputime);
L
Linus Torvalds 已提交
2642

2643
	index = (TASK_NICE(p) > 0) ? CPUTIME_NICE : CPUTIME_USER;
2644

L
Linus Torvalds 已提交
2645
	/* Add user time to cpustat. */
2646
	task_group_account_field(p, index, (__force u64) cputime);
2647

2648 2649
	/* Account for user time used */
	acct_update_integrals(p);
L
Linus Torvalds 已提交
2650 2651
}

2652 2653 2654 2655
/*
 * 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
2656
 * @cputime_scaled: cputime scaled by cpu frequency
2657
 */
2658 2659
static void account_guest_time(struct task_struct *p, cputime_t cputime,
			       cputime_t cputime_scaled)
2660
{
2661
	u64 *cpustat = kcpustat_this_cpu->cpustat;
2662

2663
	/* Add guest time to process. */
2664 2665
	p->utime += cputime;
	p->utimescaled += cputime_scaled;
2666
	account_group_user_time(p, cputime);
2667
	p->gtime += cputime;
2668

2669
	/* Add guest time to cpustat. */
2670
	if (TASK_NICE(p) > 0) {
2671 2672
		cpustat[CPUTIME_NICE] += (__force u64) cputime;
		cpustat[CPUTIME_GUEST_NICE] += (__force u64) cputime;
2673
	} else {
2674 2675
		cpustat[CPUTIME_USER] += (__force u64) cputime;
		cpustat[CPUTIME_GUEST] += (__force u64) cputime;
2676
	}
2677 2678
}

2679 2680 2681 2682 2683 2684 2685 2686 2687
/*
 * 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,
2688
			cputime_t cputime_scaled, int index)
2689 2690
{
	/* Add system time to process. */
2691 2692
	p->stime += cputime;
	p->stimescaled += cputime_scaled;
2693 2694 2695
	account_group_system_time(p, cputime);

	/* Add system time to cpustat. */
2696
	task_group_account_field(p, index, (__force u64) cputime);
2697 2698 2699 2700 2701

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

L
Linus Torvalds 已提交
2702 2703 2704 2705 2706
/*
 * 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
2707
 * @cputime_scaled: cputime scaled by cpu frequency
L
Linus Torvalds 已提交
2708 2709
 */
void account_system_time(struct task_struct *p, int hardirq_offset,
2710
			 cputime_t cputime, cputime_t cputime_scaled)
L
Linus Torvalds 已提交
2711
{
2712
	int index;
L
Linus Torvalds 已提交
2713

2714
	if ((p->flags & PF_VCPU) && (irq_count() - hardirq_offset == 0)) {
2715
		account_guest_time(p, cputime, cputime_scaled);
2716 2717
		return;
	}
2718

L
Linus Torvalds 已提交
2719
	if (hardirq_count() - hardirq_offset)
2720
		index = CPUTIME_IRQ;
2721
	else if (in_serving_softirq())
2722
		index = CPUTIME_SOFTIRQ;
L
Linus Torvalds 已提交
2723
	else
2724
		index = CPUTIME_SYSTEM;
2725

2726
	__account_system_time(p, cputime, cputime_scaled, index);
L
Linus Torvalds 已提交
2727 2728
}

2729
/*
L
Linus Torvalds 已提交
2730
 * Account for involuntary wait time.
2731
 * @cputime: the cpu time spent in involuntary wait
2732
 */
2733
void account_steal_time(cputime_t cputime)
2734
{
2735
	u64 *cpustat = kcpustat_this_cpu->cpustat;
2736

2737
	cpustat[CPUTIME_STEAL] += (__force u64) cputime;
2738 2739
}

L
Linus Torvalds 已提交
2740
/*
2741 2742
 * Account for idle time.
 * @cputime: the cpu time spent in idle wait
L
Linus Torvalds 已提交
2743
 */
2744
void account_idle_time(cputime_t cputime)
L
Linus Torvalds 已提交
2745
{
2746
	u64 *cpustat = kcpustat_this_cpu->cpustat;
2747
	struct rq *rq = this_rq();
L
Linus Torvalds 已提交
2748

2749
	if (atomic_read(&rq->nr_iowait) > 0)
2750
		cpustat[CPUTIME_IOWAIT] += (__force u64) cputime;
2751
	else
2752
		cpustat[CPUTIME_IDLE] += (__force u64) cputime;
L
Linus Torvalds 已提交
2753 2754
}

G
Glauber Costa 已提交
2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773
static __always_inline bool steal_account_process_tick(void)
{
#ifdef CONFIG_PARAVIRT
	if (static_branch(&paravirt_steal_enabled)) {
		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;
}

2774 2775
#ifndef CONFIG_VIRT_CPU_ACCOUNTING

2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801
#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);
2802
	u64 *cpustat = kcpustat_this_cpu->cpustat;
2803

G
Glauber Costa 已提交
2804 2805 2806
	if (steal_account_process_tick())
		return;

2807
	if (irqtime_account_hi_update()) {
2808
		cpustat[CPUTIME_IRQ] += (__force u64) cputime_one_jiffy;
2809
	} else if (irqtime_account_si_update()) {
2810
		cpustat[CPUTIME_SOFTIRQ] += (__force u64) cputime_one_jiffy;
2811 2812 2813 2814 2815 2816 2817
	} 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,
2818
					CPUTIME_SOFTIRQ);
2819 2820 2821 2822 2823 2824 2825 2826
	} 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,
2827
					CPUTIME_SYSTEM);
2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838
	}
}

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);
}
2839
#else /* CONFIG_IRQ_TIME_ACCOUNTING */
2840 2841 2842
static void irqtime_account_idle_ticks(int ticks) {}
static void irqtime_account_process_tick(struct task_struct *p, int user_tick,
						struct rq *rq) {}
2843
#endif /* CONFIG_IRQ_TIME_ACCOUNTING */
2844 2845 2846 2847 2848 2849 2850 2851

/*
 * 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)
{
2852
	cputime_t one_jiffy_scaled = cputime_to_scaled(cputime_one_jiffy);
2853 2854
	struct rq *rq = this_rq();

2855 2856 2857 2858 2859
	if (sched_clock_irqtime) {
		irqtime_account_process_tick(p, user_tick, rq);
		return;
	}

G
Glauber Costa 已提交
2860 2861 2862
	if (steal_account_process_tick())
		return;

2863
	if (user_tick)
2864
		account_user_time(p, cputime_one_jiffy, one_jiffy_scaled);
2865
	else if ((p != rq->idle) || (irq_count() != HARDIRQ_OFFSET))
2866
		account_system_time(p, HARDIRQ_OFFSET, cputime_one_jiffy,
2867 2868
				    one_jiffy_scaled);
	else
2869
		account_idle_time(cputime_one_jiffy);
2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887
}

/*
 * 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)
{
2888 2889 2890 2891 2892 2893

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

2894
	account_idle_time(jiffies_to_cputime(ticks));
L
Linus Torvalds 已提交
2895 2896
}

2897 2898
#endif

2899 2900 2901 2902
/*
 * Use precise platform statistics if available:
 */
#ifdef CONFIG_VIRT_CPU_ACCOUNTING
2903
void task_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
2904
{
2905 2906
	*ut = p->utime;
	*st = p->stime;
2907 2908
}

2909
void thread_group_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
2910
{
2911 2912 2913 2914 2915 2916
	struct task_cputime cputime;

	thread_group_cputime(p, &cputime);

	*ut = cputime.utime;
	*st = cputime.stime;
2917 2918
}
#else
2919 2920

#ifndef nsecs_to_cputime
2921
# define nsecs_to_cputime(__nsecs)	nsecs_to_jiffies(__nsecs)
2922 2923
#endif

2924
void task_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
2925
{
2926
	cputime_t rtime, utime = p->utime, total = utime + p->stime;
2927 2928 2929 2930

	/*
	 * Use CFS's precise accounting:
	 */
2931
	rtime = nsecs_to_cputime(p->se.sum_exec_runtime);
2932 2933

	if (total) {
2934
		u64 temp = (__force u64) rtime;
2935

2936 2937 2938
		temp *= (__force u64) utime;
		do_div(temp, (__force u32) total);
		utime = (__force cputime_t) temp;
2939 2940
	} else
		utime = rtime;
2941

2942 2943 2944
	/*
	 * Compare with previous values, to keep monotonicity:
	 */
2945
	p->prev_utime = max(p->prev_utime, utime);
2946
	p->prev_stime = max(p->prev_stime, rtime - p->prev_utime);
2947

2948 2949
	*ut = p->prev_utime;
	*st = p->prev_stime;
2950 2951
}

2952 2953 2954 2955
/*
 * Must be called with siglock held.
 */
void thread_group_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
2956
{
2957 2958 2959
	struct signal_struct *sig = p->signal;
	struct task_cputime cputime;
	cputime_t rtime, utime, total;
2960

2961
	thread_group_cputime(p, &cputime);
2962

2963
	total = cputime.utime + cputime.stime;
2964
	rtime = nsecs_to_cputime(cputime.sum_exec_runtime);
2965

2966
	if (total) {
2967
		u64 temp = (__force u64) rtime;
2968

2969 2970 2971
		temp *= (__force u64) cputime.utime;
		do_div(temp, (__force u32) total);
		utime = (__force cputime_t) temp;
2972 2973 2974 2975
	} else
		utime = rtime;

	sig->prev_utime = max(sig->prev_utime, utime);
2976
	sig->prev_stime = max(sig->prev_stime, rtime - sig->prev_utime);
2977 2978 2979

	*ut = sig->prev_utime;
	*st = sig->prev_stime;
2980 2981 2982
}
#endif

2983 2984 2985 2986 2987 2988 2989 2990
/*
 * 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
2991
	struct task_struct *curr = rq->curr;
2992 2993

	sched_clock_tick();
I
Ingo Molnar 已提交
2994

2995
	raw_spin_lock(&rq->lock);
2996
	update_rq_clock(rq);
2997
	update_cpu_load_active(rq);
P
Peter Zijlstra 已提交
2998
	curr->sched_class->task_tick(rq, curr, 0);
2999
	raw_spin_unlock(&rq->lock);
3000

3001
	perf_event_task_tick();
3002

3003
#ifdef CONFIG_SMP
3004
	rq->idle_balance = idle_cpu(cpu);
I
Ingo Molnar 已提交
3005
	trigger_load_balance(rq, cpu);
3006
#endif
L
Linus Torvalds 已提交
3007 3008
}

3009
notrace unsigned long get_parent_ip(unsigned long addr)
3010 3011 3012 3013 3014 3015 3016 3017
{
	if (in_lock_functions(addr)) {
		addr = CALLER_ADDR2;
		if (in_lock_functions(addr))
			addr = CALLER_ADDR3;
	}
	return addr;
}
L
Linus Torvalds 已提交
3018

3019 3020 3021
#if defined(CONFIG_PREEMPT) && (defined(CONFIG_DEBUG_PREEMPT) || \
				defined(CONFIG_PREEMPT_TRACER))

3022
void __kprobes add_preempt_count(int val)
L
Linus Torvalds 已提交
3023
{
3024
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3025 3026 3027
	/*
	 * Underflow?
	 */
3028 3029
	if (DEBUG_LOCKS_WARN_ON((preempt_count() < 0)))
		return;
3030
#endif
L
Linus Torvalds 已提交
3031
	preempt_count() += val;
3032
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3033 3034 3035
	/*
	 * Spinlock count overflowing soon?
	 */
3036 3037
	DEBUG_LOCKS_WARN_ON((preempt_count() & PREEMPT_MASK) >=
				PREEMPT_MASK - 10);
3038 3039 3040
#endif
	if (preempt_count() == val)
		trace_preempt_off(CALLER_ADDR0, get_parent_ip(CALLER_ADDR1));
L
Linus Torvalds 已提交
3041 3042 3043
}
EXPORT_SYMBOL(add_preempt_count);

3044
void __kprobes sub_preempt_count(int val)
L
Linus Torvalds 已提交
3045
{
3046
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3047 3048 3049
	/*
	 * Underflow?
	 */
3050
	if (DEBUG_LOCKS_WARN_ON(val > preempt_count()))
3051
		return;
L
Linus Torvalds 已提交
3052 3053 3054
	/*
	 * Is the spinlock portion underflowing?
	 */
3055 3056 3057
	if (DEBUG_LOCKS_WARN_ON((val < PREEMPT_MASK) &&
			!(preempt_count() & PREEMPT_MASK)))
		return;
3058
#endif
3059

3060 3061
	if (preempt_count() == val)
		trace_preempt_on(CALLER_ADDR0, get_parent_ip(CALLER_ADDR1));
L
Linus Torvalds 已提交
3062 3063 3064 3065 3066 3067 3068
	preempt_count() -= val;
}
EXPORT_SYMBOL(sub_preempt_count);

#endif

/*
I
Ingo Molnar 已提交
3069
 * Print scheduling while atomic bug:
L
Linus Torvalds 已提交
3070
 */
I
Ingo Molnar 已提交
3071
static noinline void __schedule_bug(struct task_struct *prev)
L
Linus Torvalds 已提交
3072
{
3073 3074
	struct pt_regs *regs = get_irq_regs();

3075 3076 3077
	if (oops_in_progress)
		return;

P
Peter Zijlstra 已提交
3078 3079
	printk(KERN_ERR "BUG: scheduling while atomic: %s/%d/0x%08x\n",
		prev->comm, prev->pid, preempt_count());
3080

I
Ingo Molnar 已提交
3081
	debug_show_held_locks(prev);
3082
	print_modules();
I
Ingo Molnar 已提交
3083 3084
	if (irqs_disabled())
		print_irqtrace_events(prev);
3085 3086 3087 3088 3089

	if (regs)
		show_regs(regs);
	else
		dump_stack();
I
Ingo Molnar 已提交
3090
}
L
Linus Torvalds 已提交
3091

I
Ingo Molnar 已提交
3092 3093 3094 3095 3096
/*
 * Various schedule()-time debugging checks and statistics:
 */
static inline void schedule_debug(struct task_struct *prev)
{
L
Linus Torvalds 已提交
3097
	/*
I
Ingo Molnar 已提交
3098
	 * Test if we are atomic. Since do_exit() needs to call into
L
Linus Torvalds 已提交
3099 3100 3101
	 * schedule() atomically, we ignore that path for now.
	 * Otherwise, whine if we are scheduling when we should not be.
	 */
3102
	if (unlikely(in_atomic_preempt_off() && !prev->exit_state))
I
Ingo Molnar 已提交
3103
		__schedule_bug(prev);
3104
	rcu_sleep_check();
I
Ingo Molnar 已提交
3105

L
Linus Torvalds 已提交
3106 3107
	profile_hit(SCHED_PROFILING, __builtin_return_address(0));

3108
	schedstat_inc(this_rq(), sched_count);
I
Ingo Molnar 已提交
3109 3110
}

P
Peter Zijlstra 已提交
3111
static void put_prev_task(struct rq *rq, struct task_struct *prev)
M
Mike Galbraith 已提交
3112
{
3113
	if (prev->on_rq || rq->skip_clock_update < 0)
3114
		update_rq_clock(rq);
P
Peter Zijlstra 已提交
3115
	prev->sched_class->put_prev_task(rq, prev);
M
Mike Galbraith 已提交
3116 3117
}

I
Ingo Molnar 已提交
3118 3119 3120 3121
/*
 * Pick up the highest-prio task:
 */
static inline struct task_struct *
3122
pick_next_task(struct rq *rq)
I
Ingo Molnar 已提交
3123
{
3124
	const struct sched_class *class;
I
Ingo Molnar 已提交
3125
	struct task_struct *p;
L
Linus Torvalds 已提交
3126 3127

	/*
I
Ingo Molnar 已提交
3128 3129
	 * Optimization: we know that if all tasks are in
	 * the fair class we can call that function directly:
L
Linus Torvalds 已提交
3130
	 */
3131
	if (likely(rq->nr_running == rq->cfs.h_nr_running)) {
3132
		p = fair_sched_class.pick_next_task(rq);
I
Ingo Molnar 已提交
3133 3134
		if (likely(p))
			return p;
L
Linus Torvalds 已提交
3135 3136
	}

3137
	for_each_class(class) {
3138
		p = class->pick_next_task(rq);
I
Ingo Molnar 已提交
3139 3140 3141
		if (p)
			return p;
	}
3142 3143

	BUG(); /* the idle class will always have a runnable task */
I
Ingo Molnar 已提交
3144
}
L
Linus Torvalds 已提交
3145

I
Ingo Molnar 已提交
3146
/*
3147
 * __schedule() is the main scheduler function.
I
Ingo Molnar 已提交
3148
 */
3149
static void __sched __schedule(void)
I
Ingo Molnar 已提交
3150 3151
{
	struct task_struct *prev, *next;
3152
	unsigned long *switch_count;
I
Ingo Molnar 已提交
3153
	struct rq *rq;
3154
	int cpu;
I
Ingo Molnar 已提交
3155

3156 3157
need_resched:
	preempt_disable();
I
Ingo Molnar 已提交
3158 3159
	cpu = smp_processor_id();
	rq = cpu_rq(cpu);
3160
	rcu_note_context_switch(cpu);
I
Ingo Molnar 已提交
3161 3162 3163
	prev = rq->curr;

	schedule_debug(prev);
L
Linus Torvalds 已提交
3164

3165
	if (sched_feat(HRTICK))
M
Mike Galbraith 已提交
3166
		hrtick_clear(rq);
P
Peter Zijlstra 已提交
3167

3168
	raw_spin_lock_irq(&rq->lock);
L
Linus Torvalds 已提交
3169

3170
	switch_count = &prev->nivcsw;
L
Linus Torvalds 已提交
3171
	if (prev->state && !(preempt_count() & PREEMPT_ACTIVE)) {
T
Tejun Heo 已提交
3172
		if (unlikely(signal_pending_state(prev->state, prev))) {
L
Linus Torvalds 已提交
3173
			prev->state = TASK_RUNNING;
T
Tejun Heo 已提交
3174
		} else {
3175 3176 3177
			deactivate_task(rq, prev, DEQUEUE_SLEEP);
			prev->on_rq = 0;

T
Tejun Heo 已提交
3178
			/*
3179 3180 3181
			 * 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
3182 3183 3184 3185 3186 3187 3188 3189 3190
			 */
			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
3191
		switch_count = &prev->nvcsw;
L
Linus Torvalds 已提交
3192 3193
	}

3194
	pre_schedule(rq, prev);
3195

I
Ingo Molnar 已提交
3196
	if (unlikely(!rq->nr_running))
L
Linus Torvalds 已提交
3197 3198
		idle_balance(cpu, rq);

M
Mike Galbraith 已提交
3199
	put_prev_task(rq, prev);
3200
	next = pick_next_task(rq);
3201 3202
	clear_tsk_need_resched(prev);
	rq->skip_clock_update = 0;
L
Linus Torvalds 已提交
3203 3204 3205 3206 3207 3208

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

I
Ingo Molnar 已提交
3209
		context_switch(rq, prev, next); /* unlocks the rq */
P
Peter Zijlstra 已提交
3210
		/*
3211 3212 3213 3214
		 * 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
3215 3216 3217
		 */
		cpu = smp_processor_id();
		rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
3218
	} else
3219
		raw_spin_unlock_irq(&rq->lock);
L
Linus Torvalds 已提交
3220

3221
	post_schedule(rq);
L
Linus Torvalds 已提交
3222

3223
	sched_preempt_enable_no_resched();
3224
	if (need_resched())
L
Linus Torvalds 已提交
3225 3226
		goto need_resched;
}
3227

3228 3229
static inline void sched_submit_work(struct task_struct *tsk)
{
3230
	if (!tsk->state || tsk_is_pi_blocked(tsk))
3231 3232 3233 3234 3235 3236 3237 3238 3239
		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
3240
asmlinkage void __sched schedule(void)
3241
{
3242 3243 3244
	struct task_struct *tsk = current;

	sched_submit_work(tsk);
3245 3246
	__schedule();
}
L
Linus Torvalds 已提交
3247 3248
EXPORT_SYMBOL(schedule);

3249 3250 3251 3252 3253 3254 3255
/**
 * schedule_preempt_disabled - called with preemption disabled
 *
 * Returns with preemption disabled. Note: preempt_count must be 1
 */
void __sched schedule_preempt_disabled(void)
{
3256
	sched_preempt_enable_no_resched();
3257 3258 3259 3260
	schedule();
	preempt_disable();
}

3261
#ifdef CONFIG_MUTEX_SPIN_ON_OWNER
3262

3263 3264 3265
static inline bool owner_running(struct mutex *lock, struct task_struct *owner)
{
	if (lock->owner != owner)
3266
		return false;
3267 3268

	/*
3269 3270 3271 3272
	 * 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.
3273
	 */
3274
	barrier();
3275

3276
	return owner->on_cpu;
3277
}
3278

3279 3280 3281 3282 3283 3284 3285 3286
/*
 * 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;
3287

3288
	rcu_read_lock();
3289 3290
	while (owner_running(lock, owner)) {
		if (need_resched())
3291
			break;
3292

3293
		arch_mutex_cpu_relax();
3294
	}
3295
	rcu_read_unlock();
3296

3297
	/*
3298 3299 3300
	 * 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.
3301
	 */
3302
	return lock->owner == NULL;
3303 3304 3305
}
#endif

L
Linus Torvalds 已提交
3306 3307
#ifdef CONFIG_PREEMPT
/*
3308
 * this is the entry point to schedule() from in-kernel preemption
I
Ingo Molnar 已提交
3309
 * off of preempt_enable. Kernel preemptions off return from interrupt
L
Linus Torvalds 已提交
3310 3311
 * occur there and call schedule directly.
 */
3312
asmlinkage void __sched notrace preempt_schedule(void)
L
Linus Torvalds 已提交
3313 3314
{
	struct thread_info *ti = current_thread_info();
3315

L
Linus Torvalds 已提交
3316 3317
	/*
	 * If there is a non-zero preempt_count or interrupts are disabled,
I
Ingo Molnar 已提交
3318
	 * we do not want to preempt the current task. Just return..
L
Linus Torvalds 已提交
3319
	 */
N
Nick Piggin 已提交
3320
	if (likely(ti->preempt_count || irqs_disabled()))
L
Linus Torvalds 已提交
3321 3322
		return;

3323
	do {
3324
		add_preempt_count_notrace(PREEMPT_ACTIVE);
3325
		__schedule();
3326
		sub_preempt_count_notrace(PREEMPT_ACTIVE);
L
Linus Torvalds 已提交
3327

3328 3329 3330 3331 3332
		/*
		 * Check again in case we missed a preemption opportunity
		 * between schedule and now.
		 */
		barrier();
3333
	} while (need_resched());
L
Linus Torvalds 已提交
3334 3335 3336 3337
}
EXPORT_SYMBOL(preempt_schedule);

/*
3338
 * this is the entry point to schedule() from kernel preemption
L
Linus Torvalds 已提交
3339 3340 3341 3342 3343 3344 3345
 * 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();
3346

3347
	/* Catch callers which need to be fixed */
L
Linus Torvalds 已提交
3348 3349
	BUG_ON(ti->preempt_count || !irqs_disabled());

3350 3351 3352
	do {
		add_preempt_count(PREEMPT_ACTIVE);
		local_irq_enable();
3353
		__schedule();
3354 3355
		local_irq_disable();
		sub_preempt_count(PREEMPT_ACTIVE);
L
Linus Torvalds 已提交
3356

3357 3358 3359 3360 3361
		/*
		 * Check again in case we missed a preemption opportunity
		 * between schedule and now.
		 */
		barrier();
3362
	} while (need_resched());
L
Linus Torvalds 已提交
3363 3364 3365 3366
}

#endif /* CONFIG_PREEMPT */

P
Peter Zijlstra 已提交
3367
int default_wake_function(wait_queue_t *curr, unsigned mode, int wake_flags,
I
Ingo Molnar 已提交
3368
			  void *key)
L
Linus Torvalds 已提交
3369
{
P
Peter Zijlstra 已提交
3370
	return try_to_wake_up(curr->private, mode, wake_flags);
L
Linus Torvalds 已提交
3371 3372 3373 3374
}
EXPORT_SYMBOL(default_wake_function);

/*
I
Ingo Molnar 已提交
3375 3376
 * 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
3377 3378 3379
 * 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
3380
 * started to run but is not in state TASK_RUNNING. try_to_wake_up() returns
L
Linus Torvalds 已提交
3381 3382
 * zero in this (rare) case, and we handle it by continuing to scan the queue.
 */
3383
static void __wake_up_common(wait_queue_head_t *q, unsigned int mode,
P
Peter Zijlstra 已提交
3384
			int nr_exclusive, int wake_flags, void *key)
L
Linus Torvalds 已提交
3385
{
3386
	wait_queue_t *curr, *next;
L
Linus Torvalds 已提交
3387

3388
	list_for_each_entry_safe(curr, next, &q->task_list, task_list) {
3389 3390
		unsigned flags = curr->flags;

P
Peter Zijlstra 已提交
3391
		if (curr->func(curr, mode, wake_flags, key) &&
3392
				(flags & WQ_FLAG_EXCLUSIVE) && !--nr_exclusive)
L
Linus Torvalds 已提交
3393 3394 3395 3396 3397 3398 3399 3400 3401
			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
3402
 * @key: is directly passed to the wakeup function
3403 3404 3405
 *
 * 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
3406
 */
3407
void __wake_up(wait_queue_head_t *q, unsigned int mode,
I
Ingo Molnar 已提交
3408
			int nr_exclusive, void *key)
L
Linus Torvalds 已提交
3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420
{
	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.
 */
3421
void __wake_up_locked(wait_queue_head_t *q, unsigned int mode, int nr)
L
Linus Torvalds 已提交
3422
{
3423
	__wake_up_common(q, mode, nr, 0, NULL);
L
Linus Torvalds 已提交
3424
}
3425
EXPORT_SYMBOL_GPL(__wake_up_locked);
L
Linus Torvalds 已提交
3426

3427 3428 3429 3430
void __wake_up_locked_key(wait_queue_head_t *q, unsigned int mode, void *key)
{
	__wake_up_common(q, mode, 1, 0, key);
}
3431
EXPORT_SYMBOL_GPL(__wake_up_locked_key);
3432

L
Linus Torvalds 已提交
3433
/**
3434
 * __wake_up_sync_key - wake up threads blocked on a waitqueue.
L
Linus Torvalds 已提交
3435 3436 3437
 * @q: the waitqueue
 * @mode: which threads
 * @nr_exclusive: how many wake-one or wake-many threads to wake up
3438
 * @key: opaque value to be passed to wakeup targets
L
Linus Torvalds 已提交
3439 3440 3441 3442 3443 3444 3445
 *
 * 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.
3446 3447 3448
 *
 * 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
3449
 */
3450 3451
void __wake_up_sync_key(wait_queue_head_t *q, unsigned int mode,
			int nr_exclusive, void *key)
L
Linus Torvalds 已提交
3452 3453
{
	unsigned long flags;
P
Peter Zijlstra 已提交
3454
	int wake_flags = WF_SYNC;
L
Linus Torvalds 已提交
3455 3456 3457 3458 3459

	if (unlikely(!q))
		return;

	if (unlikely(!nr_exclusive))
P
Peter Zijlstra 已提交
3460
		wake_flags = 0;
L
Linus Torvalds 已提交
3461 3462

	spin_lock_irqsave(&q->lock, flags);
P
Peter Zijlstra 已提交
3463
	__wake_up_common(q, mode, nr_exclusive, wake_flags, key);
L
Linus Torvalds 已提交
3464 3465
	spin_unlock_irqrestore(&q->lock, flags);
}
3466 3467 3468 3469 3470 3471 3472 3473 3474
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
3475 3476
EXPORT_SYMBOL_GPL(__wake_up_sync);	/* For internal use only */

3477 3478 3479 3480 3481 3482 3483 3484
/**
 * 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.
3485 3486 3487
 *
 * 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.
3488
 */
3489
void complete(struct completion *x)
L
Linus Torvalds 已提交
3490 3491 3492 3493 3494
{
	unsigned long flags;

	spin_lock_irqsave(&x->wait.lock, flags);
	x->done++;
3495
	__wake_up_common(&x->wait, TASK_NORMAL, 1, 0, NULL);
L
Linus Torvalds 已提交
3496 3497 3498 3499
	spin_unlock_irqrestore(&x->wait.lock, flags);
}
EXPORT_SYMBOL(complete);

3500 3501 3502 3503 3504
/**
 * 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.
3505 3506 3507
 *
 * 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.
3508
 */
3509
void complete_all(struct completion *x)
L
Linus Torvalds 已提交
3510 3511 3512 3513 3514
{
	unsigned long flags;

	spin_lock_irqsave(&x->wait.lock, flags);
	x->done += UINT_MAX/2;
3515
	__wake_up_common(&x->wait, TASK_NORMAL, 0, 0, NULL);
L
Linus Torvalds 已提交
3516 3517 3518 3519
	spin_unlock_irqrestore(&x->wait.lock, flags);
}
EXPORT_SYMBOL(complete_all);

3520 3521
static inline long __sched
do_wait_for_common(struct completion *x, long timeout, int state)
L
Linus Torvalds 已提交
3522 3523 3524 3525
{
	if (!x->done) {
		DECLARE_WAITQUEUE(wait, current);

C
Changli Gao 已提交
3526
		__add_wait_queue_tail_exclusive(&x->wait, &wait);
L
Linus Torvalds 已提交
3527
		do {
3528
			if (signal_pending_state(state, current)) {
3529 3530
				timeout = -ERESTARTSYS;
				break;
3531 3532
			}
			__set_current_state(state);
L
Linus Torvalds 已提交
3533 3534 3535
			spin_unlock_irq(&x->wait.lock);
			timeout = schedule_timeout(timeout);
			spin_lock_irq(&x->wait.lock);
3536
		} while (!x->done && timeout);
L
Linus Torvalds 已提交
3537
		__remove_wait_queue(&x->wait, &wait);
3538 3539
		if (!x->done)
			return timeout;
L
Linus Torvalds 已提交
3540 3541
	}
	x->done--;
3542
	return timeout ?: 1;
L
Linus Torvalds 已提交
3543 3544
}

3545 3546
static long __sched
wait_for_common(struct completion *x, long timeout, int state)
L
Linus Torvalds 已提交
3547 3548 3549 3550
{
	might_sleep();

	spin_lock_irq(&x->wait.lock);
3551
	timeout = do_wait_for_common(x, timeout, state);
L
Linus Torvalds 已提交
3552
	spin_unlock_irq(&x->wait.lock);
3553 3554
	return timeout;
}
L
Linus Torvalds 已提交
3555

3556 3557 3558 3559 3560 3561 3562 3563 3564 3565
/**
 * 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().
 */
3566
void __sched wait_for_completion(struct completion *x)
3567 3568
{
	wait_for_common(x, MAX_SCHEDULE_TIMEOUT, TASK_UNINTERRUPTIBLE);
L
Linus Torvalds 已提交
3569
}
3570
EXPORT_SYMBOL(wait_for_completion);
L
Linus Torvalds 已提交
3571

3572 3573 3574 3575 3576 3577 3578 3579
/**
 * 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.
3580 3581 3582
 *
 * The return value is 0 if timed out, and positive (at least 1, or number of
 * jiffies left till timeout) if completed.
3583
 */
3584
unsigned long __sched
3585
wait_for_completion_timeout(struct completion *x, unsigned long timeout)
L
Linus Torvalds 已提交
3586
{
3587
	return wait_for_common(x, timeout, TASK_UNINTERRUPTIBLE);
L
Linus Torvalds 已提交
3588
}
3589
EXPORT_SYMBOL(wait_for_completion_timeout);
L
Linus Torvalds 已提交
3590

3591 3592 3593 3594 3595 3596
/**
 * 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.
3597 3598
 *
 * The return value is -ERESTARTSYS if interrupted, 0 if completed.
3599
 */
3600
int __sched wait_for_completion_interruptible(struct completion *x)
I
Ingo Molnar 已提交
3601
{
3602 3603 3604 3605
	long t = wait_for_common(x, MAX_SCHEDULE_TIMEOUT, TASK_INTERRUPTIBLE);
	if (t == -ERESTARTSYS)
		return t;
	return 0;
I
Ingo Molnar 已提交
3606
}
3607
EXPORT_SYMBOL(wait_for_completion_interruptible);
L
Linus Torvalds 已提交
3608

3609 3610 3611 3612 3613 3614 3615
/**
 * 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.
3616 3617 3618
 *
 * The return value is -ERESTARTSYS if interrupted, 0 if timed out,
 * positive (at least 1, or number of jiffies left till timeout) if completed.
3619
 */
3620
long __sched
3621 3622
wait_for_completion_interruptible_timeout(struct completion *x,
					  unsigned long timeout)
I
Ingo Molnar 已提交
3623
{
3624
	return wait_for_common(x, timeout, TASK_INTERRUPTIBLE);
I
Ingo Molnar 已提交
3625
}
3626
EXPORT_SYMBOL(wait_for_completion_interruptible_timeout);
L
Linus Torvalds 已提交
3627

3628 3629 3630 3631 3632 3633
/**
 * 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.
3634 3635
 *
 * The return value is -ERESTARTSYS if interrupted, 0 if completed.
3636
 */
M
Matthew Wilcox 已提交
3637 3638 3639 3640 3641 3642 3643 3644 3645
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);

3646 3647 3648 3649 3650 3651 3652 3653
/**
 * 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.
3654 3655 3656
 *
 * The return value is -ERESTARTSYS if interrupted, 0 if timed out,
 * positive (at least 1, or number of jiffies left till timeout) if completed.
3657
 */
3658
long __sched
3659 3660 3661 3662 3663 3664 3665
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);

3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679
/**
 *	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)
{
3680
	unsigned long flags;
3681 3682
	int ret = 1;

3683
	spin_lock_irqsave(&x->wait.lock, flags);
3684 3685 3686 3687
	if (!x->done)
		ret = 0;
	else
		x->done--;
3688
	spin_unlock_irqrestore(&x->wait.lock, flags);
3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702
	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)
{
3703
	unsigned long flags;
3704 3705
	int ret = 1;

3706
	spin_lock_irqsave(&x->wait.lock, flags);
3707 3708
	if (!x->done)
		ret = 0;
3709
	spin_unlock_irqrestore(&x->wait.lock, flags);
3710 3711 3712 3713
	return ret;
}
EXPORT_SYMBOL(completion_done);

3714 3715
static long __sched
sleep_on_common(wait_queue_head_t *q, int state, long timeout)
L
Linus Torvalds 已提交
3716
{
I
Ingo Molnar 已提交
3717 3718 3719 3720
	unsigned long flags;
	wait_queue_t wait;

	init_waitqueue_entry(&wait, current);
L
Linus Torvalds 已提交
3721

3722
	__set_current_state(state);
L
Linus Torvalds 已提交
3723

3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737
	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
3738 3739 3740
}
EXPORT_SYMBOL(interruptible_sleep_on);

I
Ingo Molnar 已提交
3741
long __sched
I
Ingo Molnar 已提交
3742
interruptible_sleep_on_timeout(wait_queue_head_t *q, long timeout)
L
Linus Torvalds 已提交
3743
{
3744
	return sleep_on_common(q, TASK_INTERRUPTIBLE, timeout);
L
Linus Torvalds 已提交
3745 3746 3747
}
EXPORT_SYMBOL(interruptible_sleep_on_timeout);

I
Ingo Molnar 已提交
3748
void __sched sleep_on(wait_queue_head_t *q)
L
Linus Torvalds 已提交
3749
{
3750
	sleep_on_common(q, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
L
Linus Torvalds 已提交
3751 3752 3753
}
EXPORT_SYMBOL(sleep_on);

I
Ingo Molnar 已提交
3754
long __sched sleep_on_timeout(wait_queue_head_t *q, long timeout)
L
Linus Torvalds 已提交
3755
{
3756
	return sleep_on_common(q, TASK_UNINTERRUPTIBLE, timeout);
L
Linus Torvalds 已提交
3757 3758 3759
}
EXPORT_SYMBOL(sleep_on_timeout);

3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771
#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.
 */
3772
void rt_mutex_setprio(struct task_struct *p, int prio)
3773
{
3774
	int oldprio, on_rq, running;
3775
	struct rq *rq;
3776
	const struct sched_class *prev_class;
3777 3778 3779

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

3780
	rq = __task_rq_lock(p);
3781

3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799
	/*
	 * 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;
	}

3800
	trace_sched_pi_setprio(p, prio);
3801
	oldprio = p->prio;
3802
	prev_class = p->sched_class;
P
Peter Zijlstra 已提交
3803
	on_rq = p->on_rq;
3804
	running = task_current(rq, p);
3805
	if (on_rq)
3806
		dequeue_task(rq, p, 0);
3807 3808
	if (running)
		p->sched_class->put_prev_task(rq, p);
I
Ingo Molnar 已提交
3809 3810 3811 3812 3813 3814

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

3815 3816
	p->prio = prio;

3817 3818
	if (running)
		p->sched_class->set_curr_task(rq);
P
Peter Zijlstra 已提交
3819
	if (on_rq)
3820
		enqueue_task(rq, p, oldprio < prio ? ENQUEUE_HEAD : 0);
3821

P
Peter Zijlstra 已提交
3822
	check_class_changed(rq, p, prev_class, oldprio);
3823
out_unlock:
3824
	__task_rq_unlock(rq);
3825 3826
}
#endif
3827
void set_user_nice(struct task_struct *p, long nice)
L
Linus Torvalds 已提交
3828
{
I
Ingo Molnar 已提交
3829
	int old_prio, delta, on_rq;
L
Linus Torvalds 已提交
3830
	unsigned long flags;
3831
	struct rq *rq;
L
Linus Torvalds 已提交
3832 3833 3834 3835 3836 3837 3838 3839 3840 3841 3842 3843

	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
3844
	 * SCHED_FIFO/SCHED_RR:
L
Linus Torvalds 已提交
3845
	 */
3846
	if (task_has_rt_policy(p)) {
L
Linus Torvalds 已提交
3847 3848 3849
		p->static_prio = NICE_TO_PRIO(nice);
		goto out_unlock;
	}
P
Peter Zijlstra 已提交
3850
	on_rq = p->on_rq;
3851
	if (on_rq)
3852
		dequeue_task(rq, p, 0);
L
Linus Torvalds 已提交
3853 3854

	p->static_prio = NICE_TO_PRIO(nice);
3855
	set_load_weight(p);
3856 3857 3858
	old_prio = p->prio;
	p->prio = effective_prio(p);
	delta = p->prio - old_prio;
L
Linus Torvalds 已提交
3859

I
Ingo Molnar 已提交
3860
	if (on_rq) {
3861
		enqueue_task(rq, p, 0);
L
Linus Torvalds 已提交
3862
		/*
3863 3864
		 * If the task increased its priority or is running and
		 * lowered its priority, then reschedule its CPU:
L
Linus Torvalds 已提交
3865
		 */
3866
		if (delta < 0 || (delta > 0 && task_running(rq, p)))
L
Linus Torvalds 已提交
3867 3868 3869
			resched_task(rq->curr);
	}
out_unlock:
3870
	task_rq_unlock(rq, p, &flags);
L
Linus Torvalds 已提交
3871 3872 3873
}
EXPORT_SYMBOL(set_user_nice);

M
Matt Mackall 已提交
3874 3875 3876 3877 3878
/*
 * can_nice - check if a task can reduce its nice value
 * @p: task
 * @nice: nice value
 */
3879
int can_nice(const struct task_struct *p, const int nice)
M
Matt Mackall 已提交
3880
{
3881 3882
	/* convert nice value [19,-20] to rlimit style value [1,40] */
	int nice_rlim = 20 - nice;
3883

3884
	return (nice_rlim <= task_rlimit(p, RLIMIT_NICE) ||
M
Matt Mackall 已提交
3885 3886 3887
		capable(CAP_SYS_NICE));
}

L
Linus Torvalds 已提交
3888 3889 3890 3891 3892 3893 3894 3895 3896
#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.
 */
3897
SYSCALL_DEFINE1(nice, int, increment)
L
Linus Torvalds 已提交
3898
{
3899
	long nice, retval;
L
Linus Torvalds 已提交
3900 3901 3902 3903 3904 3905

	/*
	 * 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
3906 3907
	if (increment < -40)
		increment = -40;
L
Linus Torvalds 已提交
3908 3909 3910
	if (increment > 40)
		increment = 40;

3911
	nice = TASK_NICE(current) + increment;
L
Linus Torvalds 已提交
3912 3913 3914 3915 3916
	if (nice < -20)
		nice = -20;
	if (nice > 19)
		nice = 19;

M
Matt Mackall 已提交
3917 3918 3919
	if (increment < 0 && !can_nice(current, nice))
		return -EPERM;

L
Linus Torvalds 已提交
3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937
	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.
 */
3938
int task_prio(const struct task_struct *p)
L
Linus Torvalds 已提交
3939 3940 3941 3942 3943 3944 3945 3946
{
	return p->prio - MAX_RT_PRIO;
}

/**
 * task_nice - return the nice value of a given task.
 * @p: the task in question.
 */
3947
int task_nice(const struct task_struct *p)
L
Linus Torvalds 已提交
3948 3949 3950
{
	return TASK_NICE(p);
}
P
Pavel Roskin 已提交
3951
EXPORT_SYMBOL(task_nice);
L
Linus Torvalds 已提交
3952 3953 3954 3955 3956 3957 3958

/**
 * idle_cpu - is a given cpu idle currently?
 * @cpu: the processor in question.
 */
int idle_cpu(int cpu)
{
T
Thomas Gleixner 已提交
3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972
	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
3973 3974 3975 3976 3977 3978
}

/**
 * idle_task - return the idle task for a given cpu.
 * @cpu: the processor in question.
 */
3979
struct task_struct *idle_task(int cpu)
L
Linus Torvalds 已提交
3980 3981 3982 3983 3984 3985 3986 3987
{
	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
3988
static struct task_struct *find_process_by_pid(pid_t pid)
L
Linus Torvalds 已提交
3989
{
3990
	return pid ? find_task_by_vpid(pid) : current;
L
Linus Torvalds 已提交
3991 3992 3993
}

/* Actually do priority change: must hold rq lock. */
I
Ingo Molnar 已提交
3994 3995
static void
__setscheduler(struct rq *rq, struct task_struct *p, int policy, int prio)
L
Linus Torvalds 已提交
3996 3997 3998
{
	p->policy = policy;
	p->rt_priority = prio;
3999 4000 4001
	p->normal_prio = normal_prio(p);
	/* we are holding p->pi_lock already */
	p->prio = rt_mutex_getprio(p);
4002 4003 4004 4005
	if (rt_prio(p->prio))
		p->sched_class = &rt_sched_class;
	else
		p->sched_class = &fair_sched_class;
4006
	set_load_weight(p);
L
Linus Torvalds 已提交
4007 4008
}

4009 4010 4011 4012 4013 4014 4015 4016 4017 4018
/*
 * 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);
4019 4020 4021 4022 4023
	if (cred->user->user_ns == pcred->user->user_ns)
		match = (cred->euid == pcred->euid ||
			 cred->euid == pcred->uid);
	else
		match = false;
4024 4025 4026 4027
	rcu_read_unlock();
	return match;
}

4028
static int __sched_setscheduler(struct task_struct *p, int policy,
4029
				const struct sched_param *param, bool user)
L
Linus Torvalds 已提交
4030
{
4031
	int retval, oldprio, oldpolicy = -1, on_rq, running;
L
Linus Torvalds 已提交
4032
	unsigned long flags;
4033
	const struct sched_class *prev_class;
4034
	struct rq *rq;
4035
	int reset_on_fork;
L
Linus Torvalds 已提交
4036

4037 4038
	/* may grab non-irq protected spin_locks */
	BUG_ON(in_interrupt());
L
Linus Torvalds 已提交
4039 4040
recheck:
	/* double check policy once rq lock held */
4041 4042
	if (policy < 0) {
		reset_on_fork = p->sched_reset_on_fork;
L
Linus Torvalds 已提交
4043
		policy = oldpolicy = p->policy;
4044 4045 4046 4047 4048 4049 4050 4051 4052 4053
	} 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
4054 4055
	/*
	 * Valid priorities for SCHED_FIFO and SCHED_RR are
I
Ingo Molnar 已提交
4056 4057
	 * 1..MAX_USER_RT_PRIO-1, valid priority for SCHED_NORMAL,
	 * SCHED_BATCH and SCHED_IDLE is 0.
L
Linus Torvalds 已提交
4058 4059
	 */
	if (param->sched_priority < 0 ||
I
Ingo Molnar 已提交
4060
	    (p->mm && param->sched_priority > MAX_USER_RT_PRIO-1) ||
4061
	    (!p->mm && param->sched_priority > MAX_RT_PRIO-1))
L
Linus Torvalds 已提交
4062
		return -EINVAL;
4063
	if (rt_policy(policy) != (param->sched_priority != 0))
L
Linus Torvalds 已提交
4064 4065
		return -EINVAL;

4066 4067 4068
	/*
	 * Allow unprivileged RT tasks to decrease priority:
	 */
4069
	if (user && !capable(CAP_SYS_NICE)) {
4070
		if (rt_policy(policy)) {
4071 4072
			unsigned long rlim_rtprio =
					task_rlimit(p, RLIMIT_RTPRIO);
4073 4074 4075 4076 4077 4078 4079 4080 4081 4082

			/* 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;
		}
4083

I
Ingo Molnar 已提交
4084
		/*
4085 4086
		 * 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
4087
		 */
4088 4089 4090 4091
		if (p->policy == SCHED_IDLE && policy != SCHED_IDLE) {
			if (!can_nice(p, TASK_NICE(p)))
				return -EPERM;
		}
4092

4093
		/* can't change other user's priorities */
4094
		if (!check_same_owner(p))
4095
			return -EPERM;
4096 4097 4098 4099

		/* Normal users shall not reset the sched_reset_on_fork flag */
		if (p->sched_reset_on_fork && !reset_on_fork)
			return -EPERM;
4100
	}
L
Linus Torvalds 已提交
4101

4102
	if (user) {
4103
		retval = security_task_setscheduler(p);
4104 4105 4106 4107
		if (retval)
			return retval;
	}

4108 4109 4110
	/*
	 * make sure no PI-waiters arrive (or leave) while we are
	 * changing the priority of the task:
4111
	 *
L
Lucas De Marchi 已提交
4112
	 * To be able to change p->policy safely, the appropriate
L
Linus Torvalds 已提交
4113 4114
	 * runqueue lock must be held.
	 */
4115
	rq = task_rq_lock(p, &flags);
4116

4117 4118 4119 4120
	/*
	 * Changing the policy of the stop threads its a very bad idea
	 */
	if (p == rq->stop) {
4121
		task_rq_unlock(rq, p, &flags);
4122 4123 4124
		return -EINVAL;
	}

4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135
	/*
	 * 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;
	}

4136 4137 4138 4139 4140 4141 4142
#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) &&
4143 4144
				task_group(p)->rt_bandwidth.rt_runtime == 0 &&
				!task_group_is_autogroup(task_group(p))) {
4145
			task_rq_unlock(rq, p, &flags);
4146 4147 4148 4149 4150
			return -EPERM;
		}
	}
#endif

L
Linus Torvalds 已提交
4151 4152 4153
	/* recheck policy now with rq lock held */
	if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) {
		policy = oldpolicy = -1;
4154
		task_rq_unlock(rq, p, &flags);
L
Linus Torvalds 已提交
4155 4156
		goto recheck;
	}
P
Peter Zijlstra 已提交
4157
	on_rq = p->on_rq;
4158
	running = task_current(rq, p);
4159
	if (on_rq)
4160
		dequeue_task(rq, p, 0);
4161 4162
	if (running)
		p->sched_class->put_prev_task(rq, p);
4163

4164 4165
	p->sched_reset_on_fork = reset_on_fork;

L
Linus Torvalds 已提交
4166
	oldprio = p->prio;
4167
	prev_class = p->sched_class;
I
Ingo Molnar 已提交
4168
	__setscheduler(rq, p, policy, param->sched_priority);
4169

4170 4171
	if (running)
		p->sched_class->set_curr_task(rq);
P
Peter Zijlstra 已提交
4172
	if (on_rq)
4173
		enqueue_task(rq, p, 0);
4174

P
Peter Zijlstra 已提交
4175
	check_class_changed(rq, p, prev_class, oldprio);
4176
	task_rq_unlock(rq, p, &flags);
4177

4178 4179
	rt_mutex_adjust_pi(p);

L
Linus Torvalds 已提交
4180 4181
	return 0;
}
4182 4183 4184 4185 4186 4187 4188 4189 4190 4191

/**
 * 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,
4192
		       const struct sched_param *param)
4193 4194 4195
{
	return __sched_setscheduler(p, policy, param, true);
}
L
Linus Torvalds 已提交
4196 4197
EXPORT_SYMBOL_GPL(sched_setscheduler);

4198 4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209
/**
 * 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,
4210
			       const struct sched_param *param)
4211 4212 4213 4214
{
	return __sched_setscheduler(p, policy, param, false);
}

I
Ingo Molnar 已提交
4215 4216
static int
do_sched_setscheduler(pid_t pid, int policy, struct sched_param __user *param)
L
Linus Torvalds 已提交
4217 4218 4219
{
	struct sched_param lparam;
	struct task_struct *p;
4220
	int retval;
L
Linus Torvalds 已提交
4221 4222 4223 4224 4225

	if (!param || pid < 0)
		return -EINVAL;
	if (copy_from_user(&lparam, param, sizeof(struct sched_param)))
		return -EFAULT;
4226 4227 4228

	rcu_read_lock();
	retval = -ESRCH;
L
Linus Torvalds 已提交
4229
	p = find_process_by_pid(pid);
4230 4231 4232
	if (p != NULL)
		retval = sched_setscheduler(p, policy, &lparam);
	rcu_read_unlock();
4233

L
Linus Torvalds 已提交
4234 4235 4236 4237 4238 4239 4240 4241 4242
	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.
 */
4243 4244
SYSCALL_DEFINE3(sched_setscheduler, pid_t, pid, int, policy,
		struct sched_param __user *, param)
L
Linus Torvalds 已提交
4245
{
4246 4247 4248 4249
	/* negative values for policy are not valid */
	if (policy < 0)
		return -EINVAL;

L
Linus Torvalds 已提交
4250 4251 4252 4253 4254 4255 4256 4257
	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.
 */
4258
SYSCALL_DEFINE2(sched_setparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4259 4260 4261 4262 4263 4264 4265 4266
{
	return do_sched_setscheduler(pid, -1, param);
}

/**
 * sys_sched_getscheduler - get the policy (scheduling class) of a thread
 * @pid: the pid in question.
 */
4267
SYSCALL_DEFINE1(sched_getscheduler, pid_t, pid)
L
Linus Torvalds 已提交
4268
{
4269
	struct task_struct *p;
4270
	int retval;
L
Linus Torvalds 已提交
4271 4272

	if (pid < 0)
4273
		return -EINVAL;
L
Linus Torvalds 已提交
4274 4275

	retval = -ESRCH;
4276
	rcu_read_lock();
L
Linus Torvalds 已提交
4277 4278 4279 4280
	p = find_process_by_pid(pid);
	if (p) {
		retval = security_task_getscheduler(p);
		if (!retval)
4281 4282
			retval = p->policy
				| (p->sched_reset_on_fork ? SCHED_RESET_ON_FORK : 0);
L
Linus Torvalds 已提交
4283
	}
4284
	rcu_read_unlock();
L
Linus Torvalds 已提交
4285 4286 4287 4288
	return retval;
}

/**
4289
 * sys_sched_getparam - get the RT priority of a thread
L
Linus Torvalds 已提交
4290 4291 4292
 * @pid: the pid in question.
 * @param: structure containing the RT priority.
 */
4293
SYSCALL_DEFINE2(sched_getparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4294 4295
{
	struct sched_param lp;
4296
	struct task_struct *p;
4297
	int retval;
L
Linus Torvalds 已提交
4298 4299

	if (!param || pid < 0)
4300
		return -EINVAL;
L
Linus Torvalds 已提交
4301

4302
	rcu_read_lock();
L
Linus Torvalds 已提交
4303 4304 4305 4306 4307 4308 4309 4310 4311 4312
	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;
4313
	rcu_read_unlock();
L
Linus Torvalds 已提交
4314 4315 4316 4317 4318 4319 4320 4321 4322

	/*
	 * 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:
4323
	rcu_read_unlock();
L
Linus Torvalds 已提交
4324 4325 4326
	return retval;
}

4327
long sched_setaffinity(pid_t pid, const struct cpumask *in_mask)
L
Linus Torvalds 已提交
4328
{
4329
	cpumask_var_t cpus_allowed, new_mask;
4330 4331
	struct task_struct *p;
	int retval;
L
Linus Torvalds 已提交
4332

4333
	get_online_cpus();
4334
	rcu_read_lock();
L
Linus Torvalds 已提交
4335 4336 4337

	p = find_process_by_pid(pid);
	if (!p) {
4338
		rcu_read_unlock();
4339
		put_online_cpus();
L
Linus Torvalds 已提交
4340 4341 4342
		return -ESRCH;
	}

4343
	/* Prevent p going away */
L
Linus Torvalds 已提交
4344
	get_task_struct(p);
4345
	rcu_read_unlock();
L
Linus Torvalds 已提交
4346

4347 4348 4349 4350 4351 4352 4353 4354
	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
4355
	retval = -EPERM;
4356
	if (!check_same_owner(p) && !ns_capable(task_user_ns(p), CAP_SYS_NICE))
L
Linus Torvalds 已提交
4357 4358
		goto out_unlock;

4359
	retval = security_task_setscheduler(p);
4360 4361 4362
	if (retval)
		goto out_unlock;

4363 4364
	cpuset_cpus_allowed(p, cpus_allowed);
	cpumask_and(new_mask, in_mask, cpus_allowed);
P
Peter Zijlstra 已提交
4365
again:
4366
	retval = set_cpus_allowed_ptr(p, new_mask);
L
Linus Torvalds 已提交
4367

P
Paul Menage 已提交
4368
	if (!retval) {
4369 4370
		cpuset_cpus_allowed(p, cpus_allowed);
		if (!cpumask_subset(new_mask, cpus_allowed)) {
P
Paul Menage 已提交
4371 4372 4373 4374 4375
			/*
			 * We must have raced with a concurrent cpuset
			 * update. Just reset the cpus_allowed to the
			 * cpuset's cpus_allowed
			 */
4376
			cpumask_copy(new_mask, cpus_allowed);
P
Paul Menage 已提交
4377 4378 4379
			goto again;
		}
	}
L
Linus Torvalds 已提交
4380
out_unlock:
4381 4382 4383 4384
	free_cpumask_var(new_mask);
out_free_cpus_allowed:
	free_cpumask_var(cpus_allowed);
out_put_task:
L
Linus Torvalds 已提交
4385
	put_task_struct(p);
4386
	put_online_cpus();
L
Linus Torvalds 已提交
4387 4388 4389 4390
	return retval;
}

static int get_user_cpu_mask(unsigned long __user *user_mask_ptr, unsigned len,
4391
			     struct cpumask *new_mask)
L
Linus Torvalds 已提交
4392
{
4393 4394 4395 4396 4397
	if (len < cpumask_size())
		cpumask_clear(new_mask);
	else if (len > cpumask_size())
		len = cpumask_size();

L
Linus Torvalds 已提交
4398 4399 4400 4401 4402 4403 4404 4405 4406
	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
 */
4407 4408
SYSCALL_DEFINE3(sched_setaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
4409
{
4410
	cpumask_var_t new_mask;
L
Linus Torvalds 已提交
4411 4412
	int retval;

4413 4414
	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
4415

4416 4417 4418 4419 4420
	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
4421 4422
}

4423
long sched_getaffinity(pid_t pid, struct cpumask *mask)
L
Linus Torvalds 已提交
4424
{
4425
	struct task_struct *p;
4426
	unsigned long flags;
L
Linus Torvalds 已提交
4427 4428
	int retval;

4429
	get_online_cpus();
4430
	rcu_read_lock();
L
Linus Torvalds 已提交
4431 4432 4433 4434 4435 4436

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

4437 4438 4439 4440
	retval = security_task_getscheduler(p);
	if (retval)
		goto out_unlock;

4441
	raw_spin_lock_irqsave(&p->pi_lock, flags);
4442
	cpumask_and(mask, &p->cpus_allowed, cpu_online_mask);
4443
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
4444 4445

out_unlock:
4446
	rcu_read_unlock();
4447
	put_online_cpus();
L
Linus Torvalds 已提交
4448

4449
	return retval;
L
Linus Torvalds 已提交
4450 4451 4452 4453 4454 4455 4456 4457
}

/**
 * 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
 */
4458 4459
SYSCALL_DEFINE3(sched_getaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
4460 4461
{
	int ret;
4462
	cpumask_var_t mask;
L
Linus Torvalds 已提交
4463

A
Anton Blanchard 已提交
4464
	if ((len * BITS_PER_BYTE) < nr_cpu_ids)
4465 4466
		return -EINVAL;
	if (len & (sizeof(unsigned long)-1))
L
Linus Torvalds 已提交
4467 4468
		return -EINVAL;

4469 4470
	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
4471

4472 4473
	ret = sched_getaffinity(pid, mask);
	if (ret == 0) {
4474
		size_t retlen = min_t(size_t, len, cpumask_size());
4475 4476

		if (copy_to_user(user_mask_ptr, mask, retlen))
4477 4478
			ret = -EFAULT;
		else
4479
			ret = retlen;
4480 4481
	}
	free_cpumask_var(mask);
L
Linus Torvalds 已提交
4482

4483
	return ret;
L
Linus Torvalds 已提交
4484 4485 4486 4487 4488
}

/**
 * sys_sched_yield - yield the current processor to other threads.
 *
I
Ingo Molnar 已提交
4489 4490
 * 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
4491
 */
4492
SYSCALL_DEFINE0(sched_yield)
L
Linus Torvalds 已提交
4493
{
4494
	struct rq *rq = this_rq_lock();
L
Linus Torvalds 已提交
4495

4496
	schedstat_inc(rq, yld_count);
4497
	current->sched_class->yield_task(rq);
L
Linus Torvalds 已提交
4498 4499 4500 4501 4502 4503

	/*
	 * Since we are going to call schedule() anyway, there's
	 * no need to preempt or enable interrupts:
	 */
	__release(rq->lock);
4504
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
4505
	do_raw_spin_unlock(&rq->lock);
4506
	sched_preempt_enable_no_resched();
L
Linus Torvalds 已提交
4507 4508 4509 4510 4511 4512

	schedule();

	return 0;
}

P
Peter Zijlstra 已提交
4513 4514 4515 4516 4517
static inline int should_resched(void)
{
	return need_resched() && !(preempt_count() & PREEMPT_ACTIVE);
}

A
Andrew Morton 已提交
4518
static void __cond_resched(void)
L
Linus Torvalds 已提交
4519
{
4520
	add_preempt_count(PREEMPT_ACTIVE);
4521
	__schedule();
4522
	sub_preempt_count(PREEMPT_ACTIVE);
L
Linus Torvalds 已提交
4523 4524
}

4525
int __sched _cond_resched(void)
L
Linus Torvalds 已提交
4526
{
P
Peter Zijlstra 已提交
4527
	if (should_resched()) {
L
Linus Torvalds 已提交
4528 4529 4530 4531 4532
		__cond_resched();
		return 1;
	}
	return 0;
}
4533
EXPORT_SYMBOL(_cond_resched);
L
Linus Torvalds 已提交
4534 4535

/*
4536
 * __cond_resched_lock() - if a reschedule is pending, drop the given lock,
L
Linus Torvalds 已提交
4537 4538
 * call schedule, and on return reacquire the lock.
 *
I
Ingo Molnar 已提交
4539
 * This works OK both with and without CONFIG_PREEMPT. We do strange low-level
L
Linus Torvalds 已提交
4540 4541 4542
 * operations here to prevent schedule() from being called twice (once via
 * spin_unlock(), once by hand).
 */
4543
int __cond_resched_lock(spinlock_t *lock)
L
Linus Torvalds 已提交
4544
{
P
Peter Zijlstra 已提交
4545
	int resched = should_resched();
J
Jan Kara 已提交
4546 4547
	int ret = 0;

4548 4549
	lockdep_assert_held(lock);

N
Nick Piggin 已提交
4550
	if (spin_needbreak(lock) || resched) {
L
Linus Torvalds 已提交
4551
		spin_unlock(lock);
P
Peter Zijlstra 已提交
4552
		if (resched)
N
Nick Piggin 已提交
4553 4554 4555
			__cond_resched();
		else
			cpu_relax();
J
Jan Kara 已提交
4556
		ret = 1;
L
Linus Torvalds 已提交
4557 4558
		spin_lock(lock);
	}
J
Jan Kara 已提交
4559
	return ret;
L
Linus Torvalds 已提交
4560
}
4561
EXPORT_SYMBOL(__cond_resched_lock);
L
Linus Torvalds 已提交
4562

4563
int __sched __cond_resched_softirq(void)
L
Linus Torvalds 已提交
4564 4565 4566
{
	BUG_ON(!in_softirq());

P
Peter Zijlstra 已提交
4567
	if (should_resched()) {
4568
		local_bh_enable();
L
Linus Torvalds 已提交
4569 4570 4571 4572 4573 4574
		__cond_resched();
		local_bh_disable();
		return 1;
	}
	return 0;
}
4575
EXPORT_SYMBOL(__cond_resched_softirq);
L
Linus Torvalds 已提交
4576 4577 4578 4579

/**
 * yield - yield the current processor to other threads.
 *
P
Peter Zijlstra 已提交
4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597
 * 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
4598 4599 4600 4601 4602 4603 4604 4605
 */
void __sched yield(void)
{
	set_current_state(TASK_RUNNING);
	sys_sched_yield();
}
EXPORT_SYMBOL(yield);

4606 4607 4608 4609
/**
 * 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
4610 4611
 * @p: target task
 * @preempt: whether task preemption is allowed or not
4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643 4644 4645
 *
 * 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);
4646
	if (yielded) {
4647
		schedstat_inc(rq, yld_count);
4648 4649 4650 4651 4652 4653
		/*
		 * Make p's CPU reschedule; pick_next_entity takes care of
		 * fairness.
		 */
		if (preempt && rq != p_rq)
			resched_task(p_rq->curr);
4654 4655 4656 4657 4658 4659 4660
	} 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;
4661
	}
4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673

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
4674
/*
I
Ingo Molnar 已提交
4675
 * This task is about to go to sleep on IO. Increment rq->nr_iowait so
L
Linus Torvalds 已提交
4676 4677 4678 4679
 * that process accounting knows that this is a task in IO wait state.
 */
void __sched io_schedule(void)
{
4680
	struct rq *rq = raw_rq();
L
Linus Torvalds 已提交
4681

4682
	delayacct_blkio_start();
L
Linus Torvalds 已提交
4683
	atomic_inc(&rq->nr_iowait);
4684
	blk_flush_plug(current);
4685
	current->in_iowait = 1;
L
Linus Torvalds 已提交
4686
	schedule();
4687
	current->in_iowait = 0;
L
Linus Torvalds 已提交
4688
	atomic_dec(&rq->nr_iowait);
4689
	delayacct_blkio_end();
L
Linus Torvalds 已提交
4690 4691 4692 4693 4694
}
EXPORT_SYMBOL(io_schedule);

long __sched io_schedule_timeout(long timeout)
{
4695
	struct rq *rq = raw_rq();
L
Linus Torvalds 已提交
4696 4697
	long ret;

4698
	delayacct_blkio_start();
L
Linus Torvalds 已提交
4699
	atomic_inc(&rq->nr_iowait);
4700
	blk_flush_plug(current);
4701
	current->in_iowait = 1;
L
Linus Torvalds 已提交
4702
	ret = schedule_timeout(timeout);
4703
	current->in_iowait = 0;
L
Linus Torvalds 已提交
4704
	atomic_dec(&rq->nr_iowait);
4705
	delayacct_blkio_end();
L
Linus Torvalds 已提交
4706 4707 4708 4709 4710 4711 4712 4713 4714 4715
	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.
 */
4716
SYSCALL_DEFINE1(sched_get_priority_max, int, policy)
L
Linus Torvalds 已提交
4717 4718 4719 4720 4721 4722 4723 4724 4725
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = MAX_USER_RT_PRIO-1;
		break;
	case SCHED_NORMAL:
4726
	case SCHED_BATCH:
I
Ingo Molnar 已提交
4727
	case SCHED_IDLE:
L
Linus Torvalds 已提交
4728 4729 4730 4731 4732 4733 4734 4735 4736 4737 4738 4739 4740
		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.
 */
4741
SYSCALL_DEFINE1(sched_get_priority_min, int, policy)
L
Linus Torvalds 已提交
4742 4743 4744 4745 4746 4747 4748 4749 4750
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = 1;
		break;
	case SCHED_NORMAL:
4751
	case SCHED_BATCH:
I
Ingo Molnar 已提交
4752
	case SCHED_IDLE:
L
Linus Torvalds 已提交
4753 4754 4755 4756 4757 4758 4759 4760 4761 4762 4763 4764 4765
		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.
 */
4766
SYSCALL_DEFINE2(sched_rr_get_interval, pid_t, pid,
4767
		struct timespec __user *, interval)
L
Linus Torvalds 已提交
4768
{
4769
	struct task_struct *p;
D
Dmitry Adamushko 已提交
4770
	unsigned int time_slice;
4771 4772
	unsigned long flags;
	struct rq *rq;
4773
	int retval;
L
Linus Torvalds 已提交
4774 4775 4776
	struct timespec t;

	if (pid < 0)
4777
		return -EINVAL;
L
Linus Torvalds 已提交
4778 4779

	retval = -ESRCH;
4780
	rcu_read_lock();
L
Linus Torvalds 已提交
4781 4782 4783 4784 4785 4786 4787 4788
	p = find_process_by_pid(pid);
	if (!p)
		goto out_unlock;

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

4789 4790
	rq = task_rq_lock(p, &flags);
	time_slice = p->sched_class->get_rr_interval(rq, p);
4791
	task_rq_unlock(rq, p, &flags);
D
Dmitry Adamushko 已提交
4792

4793
	rcu_read_unlock();
D
Dmitry Adamushko 已提交
4794
	jiffies_to_timespec(time_slice, &t);
L
Linus Torvalds 已提交
4795 4796
	retval = copy_to_user(interval, &t, sizeof(t)) ? -EFAULT : 0;
	return retval;
4797

L
Linus Torvalds 已提交
4798
out_unlock:
4799
	rcu_read_unlock();
L
Linus Torvalds 已提交
4800 4801 4802
	return retval;
}

4803
static const char stat_nam[] = TASK_STATE_TO_CHAR_STR;
4804

4805
void sched_show_task(struct task_struct *p)
L
Linus Torvalds 已提交
4806 4807
{
	unsigned long free = 0;
4808
	unsigned state;
L
Linus Torvalds 已提交
4809 4810

	state = p->state ? __ffs(p->state) + 1 : 0;
4811
	printk(KERN_INFO "%-15.15s %c", p->comm,
4812
		state < sizeof(stat_nam) - 1 ? stat_nam[state] : '?');
4813
#if BITS_PER_LONG == 32
L
Linus Torvalds 已提交
4814
	if (state == TASK_RUNNING)
P
Peter Zijlstra 已提交
4815
		printk(KERN_CONT " running  ");
L
Linus Torvalds 已提交
4816
	else
P
Peter Zijlstra 已提交
4817
		printk(KERN_CONT " %08lx ", thread_saved_pc(p));
L
Linus Torvalds 已提交
4818 4819
#else
	if (state == TASK_RUNNING)
P
Peter Zijlstra 已提交
4820
		printk(KERN_CONT "  running task    ");
L
Linus Torvalds 已提交
4821
	else
P
Peter Zijlstra 已提交
4822
		printk(KERN_CONT " %016lx ", thread_saved_pc(p));
L
Linus Torvalds 已提交
4823 4824
#endif
#ifdef CONFIG_DEBUG_STACK_USAGE
4825
	free = stack_not_used(p);
L
Linus Torvalds 已提交
4826
#endif
P
Peter Zijlstra 已提交
4827
	printk(KERN_CONT "%5lu %5d %6d 0x%08lx\n", free,
4828
		task_pid_nr(p), task_pid_nr(rcu_dereference(p->real_parent)),
4829
		(unsigned long)task_thread_info(p)->flags);
L
Linus Torvalds 已提交
4830

4831
	show_stack(p, NULL);
L
Linus Torvalds 已提交
4832 4833
}

I
Ingo Molnar 已提交
4834
void show_state_filter(unsigned long state_filter)
L
Linus Torvalds 已提交
4835
{
4836
	struct task_struct *g, *p;
L
Linus Torvalds 已提交
4837

4838
#if BITS_PER_LONG == 32
P
Peter Zijlstra 已提交
4839 4840
	printk(KERN_INFO
		"  task                PC stack   pid father\n");
L
Linus Torvalds 已提交
4841
#else
P
Peter Zijlstra 已提交
4842 4843
	printk(KERN_INFO
		"  task                        PC stack   pid father\n");
L
Linus Torvalds 已提交
4844
#endif
4845
	rcu_read_lock();
L
Linus Torvalds 已提交
4846 4847 4848
	do_each_thread(g, p) {
		/*
		 * reset the NMI-timeout, listing all files on a slow
L
Lucas De Marchi 已提交
4849
		 * console might take a lot of time:
L
Linus Torvalds 已提交
4850 4851
		 */
		touch_nmi_watchdog();
I
Ingo Molnar 已提交
4852
		if (!state_filter || (p->state & state_filter))
4853
			sched_show_task(p);
L
Linus Torvalds 已提交
4854 4855
	} while_each_thread(g, p);

4856 4857
	touch_all_softlockup_watchdogs();

I
Ingo Molnar 已提交
4858 4859 4860
#ifdef CONFIG_SCHED_DEBUG
	sysrq_sched_debug_show();
#endif
4861
	rcu_read_unlock();
I
Ingo Molnar 已提交
4862 4863 4864
	/*
	 * Only show locks if all tasks are dumped:
	 */
4865
	if (!state_filter)
I
Ingo Molnar 已提交
4866
		debug_show_all_locks();
L
Linus Torvalds 已提交
4867 4868
}

I
Ingo Molnar 已提交
4869 4870
void __cpuinit init_idle_bootup_task(struct task_struct *idle)
{
I
Ingo Molnar 已提交
4871
	idle->sched_class = &idle_sched_class;
I
Ingo Molnar 已提交
4872 4873
}

4874 4875 4876 4877 4878 4879 4880 4881
/**
 * 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.
 */
4882
void __cpuinit init_idle(struct task_struct *idle, int cpu)
L
Linus Torvalds 已提交
4883
{
4884
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
4885 4886
	unsigned long flags;

4887
	raw_spin_lock_irqsave(&rq->lock, flags);
4888

I
Ingo Molnar 已提交
4889
	__sched_fork(idle);
4890
	idle->state = TASK_RUNNING;
I
Ingo Molnar 已提交
4891 4892
	idle->se.exec_start = sched_clock();

4893
	do_set_cpus_allowed(idle, cpumask_of(cpu));
4894 4895 4896 4897 4898 4899 4900 4901 4902 4903 4904
	/*
	 * 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
4905
	__set_task_cpu(idle, cpu);
4906
	rcu_read_unlock();
L
Linus Torvalds 已提交
4907 4908

	rq->curr = rq->idle = idle;
P
Peter Zijlstra 已提交
4909 4910
#if defined(CONFIG_SMP)
	idle->on_cpu = 1;
4911
#endif
4912
	raw_spin_unlock_irqrestore(&rq->lock, flags);
L
Linus Torvalds 已提交
4913 4914

	/* Set the preempt count _outside_ the spinlocks! */
A
Al Viro 已提交
4915
	task_thread_info(idle)->preempt_count = 0;
4916

I
Ingo Molnar 已提交
4917 4918 4919 4920
	/*
	 * The idle tasks have their own, simple scheduling class:
	 */
	idle->sched_class = &idle_sched_class;
4921
	ftrace_graph_init_idle_task(idle, cpu);
4922 4923 4924
#if defined(CONFIG_SMP)
	sprintf(idle->comm, "%s/%d", INIT_TASK_COMM, cpu);
#endif
I
Ingo Molnar 已提交
4925 4926
}

L
Linus Torvalds 已提交
4927
#ifdef CONFIG_SMP
4928 4929 4930 4931
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);
4932 4933 4934

	cpumask_copy(&p->cpus_allowed, new_mask);
	p->rt.nr_cpus_allowed = cpumask_weight(new_mask);
4935 4936
}

L
Linus Torvalds 已提交
4937 4938 4939
/*
 * This is how migration works:
 *
4940 4941 4942 4943 4944 4945
 * 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
4946
 *    it and puts it into the right queue.
4947 4948
 * 5) stopper completes and stop_one_cpu() returns and the migration
 *    is done.
L
Linus Torvalds 已提交
4949 4950 4951 4952 4953 4954 4955 4956
 */

/*
 * 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
4957
 * task must not exit() & deallocate itself prematurely. The
L
Linus Torvalds 已提交
4958 4959
 * call is not atomic; no spinlocks may be held.
 */
4960
int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
L
Linus Torvalds 已提交
4961 4962
{
	unsigned long flags;
4963
	struct rq *rq;
4964
	unsigned int dest_cpu;
4965
	int ret = 0;
L
Linus Torvalds 已提交
4966 4967

	rq = task_rq_lock(p, &flags);
4968

4969 4970 4971
	if (cpumask_equal(&p->cpus_allowed, new_mask))
		goto out;

4972
	if (!cpumask_intersects(new_mask, cpu_active_mask)) {
L
Linus Torvalds 已提交
4973 4974 4975 4976
		ret = -EINVAL;
		goto out;
	}

4977
	if (unlikely((p->flags & PF_THREAD_BOUND) && p != current)) {
4978 4979 4980 4981
		ret = -EINVAL;
		goto out;
	}

4982
	do_set_cpus_allowed(p, new_mask);
4983

L
Linus Torvalds 已提交
4984
	/* Can the task run on the task's current CPU? If so, we're done */
4985
	if (cpumask_test_cpu(task_cpu(p), new_mask))
L
Linus Torvalds 已提交
4986 4987
		goto out;

4988
	dest_cpu = cpumask_any_and(cpu_active_mask, new_mask);
4989
	if (p->on_rq) {
4990
		struct migration_arg arg = { p, dest_cpu };
L
Linus Torvalds 已提交
4991
		/* Need help from migration thread: drop lock and wait. */
4992
		task_rq_unlock(rq, p, &flags);
4993
		stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg);
L
Linus Torvalds 已提交
4994 4995 4996 4997
		tlb_migrate_finish(p->mm);
		return 0;
	}
out:
4998
	task_rq_unlock(rq, p, &flags);
4999

L
Linus Torvalds 已提交
5000 5001
	return ret;
}
5002
EXPORT_SYMBOL_GPL(set_cpus_allowed_ptr);
L
Linus Torvalds 已提交
5003 5004

/*
I
Ingo Molnar 已提交
5005
 * Move (not current) task off this cpu, onto dest cpu. We're doing
L
Linus Torvalds 已提交
5006 5007 5008 5009 5010 5011
 * 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.
5012 5013
 *
 * Returns non-zero if task was successfully migrated.
L
Linus Torvalds 已提交
5014
 */
5015
static int __migrate_task(struct task_struct *p, int src_cpu, int dest_cpu)
L
Linus Torvalds 已提交
5016
{
5017
	struct rq *rq_dest, *rq_src;
5018
	int ret = 0;
L
Linus Torvalds 已提交
5019

5020
	if (unlikely(!cpu_active(dest_cpu)))
5021
		return ret;
L
Linus Torvalds 已提交
5022 5023 5024 5025

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

5026
	raw_spin_lock(&p->pi_lock);
L
Linus Torvalds 已提交
5027 5028 5029
	double_rq_lock(rq_src, rq_dest);
	/* Already moved. */
	if (task_cpu(p) != src_cpu)
L
Linus Torvalds 已提交
5030
		goto done;
L
Linus Torvalds 已提交
5031
	/* Affinity changed (again). */
5032
	if (!cpumask_test_cpu(dest_cpu, tsk_cpus_allowed(p)))
L
Linus Torvalds 已提交
5033
		goto fail;
L
Linus Torvalds 已提交
5034

5035 5036 5037 5038
	/*
	 * If we're not on a rq, the next wake-up will ensure we're
	 * placed properly.
	 */
P
Peter Zijlstra 已提交
5039
	if (p->on_rq) {
5040
		dequeue_task(rq_src, p, 0);
5041
		set_task_cpu(p, dest_cpu);
5042
		enqueue_task(rq_dest, p, 0);
5043
		check_preempt_curr(rq_dest, p, 0);
L
Linus Torvalds 已提交
5044
	}
L
Linus Torvalds 已提交
5045
done:
5046
	ret = 1;
L
Linus Torvalds 已提交
5047
fail:
L
Linus Torvalds 已提交
5048
	double_rq_unlock(rq_src, rq_dest);
5049
	raw_spin_unlock(&p->pi_lock);
5050
	return ret;
L
Linus Torvalds 已提交
5051 5052 5053
}

/*
5054 5055 5056
 * 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
5057
 */
5058
static int migration_cpu_stop(void *data)
L
Linus Torvalds 已提交
5059
{
5060
	struct migration_arg *arg = data;
5061

5062 5063 5064 5065
	/*
	 * The original target cpu might have gone down and we might
	 * be on another cpu but it doesn't matter.
	 */
5066
	local_irq_disable();
5067
	__migrate_task(arg->task, raw_smp_processor_id(), arg->dest_cpu);
5068
	local_irq_enable();
L
Linus Torvalds 已提交
5069
	return 0;
5070 5071
}

L
Linus Torvalds 已提交
5072
#ifdef CONFIG_HOTPLUG_CPU
5073

5074
/*
5075 5076
 * Ensures that the idle task is using init_mm right before its cpu goes
 * offline.
5077
 */
5078
void idle_task_exit(void)
L
Linus Torvalds 已提交
5079
{
5080
	struct mm_struct *mm = current->active_mm;
5081

5082
	BUG_ON(cpu_online(smp_processor_id()));
5083

5084 5085 5086
	if (mm != &init_mm)
		switch_mm(mm, &init_mm, current);
	mmdrop(mm);
L
Linus Torvalds 已提交
5087 5088 5089 5090 5091 5092 5093 5094 5095
}

/*
 * 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:
 */
5096
static void migrate_nr_uninterruptible(struct rq *rq_src)
L
Linus Torvalds 已提交
5097
{
5098
	struct rq *rq_dest = cpu_rq(cpumask_any(cpu_active_mask));
L
Linus Torvalds 已提交
5099 5100 5101 5102 5103

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

I
Ingo Molnar 已提交
5104
/*
5105
 * remove the tasks which were accounted by rq from calc_load_tasks.
L
Linus Torvalds 已提交
5106
 */
5107
static void calc_global_load_remove(struct rq *rq)
L
Linus Torvalds 已提交
5108
{
5109 5110
	atomic_long_sub(rq->calc_load_active, &calc_load_tasks);
	rq->calc_load_active = 0;
L
Linus Torvalds 已提交
5111 5112
}

5113
/*
5114 5115 5116 5117 5118 5119
 * 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
5120
 */
5121
static void migrate_tasks(unsigned int dead_cpu)
L
Linus Torvalds 已提交
5122
{
5123
	struct rq *rq = cpu_rq(dead_cpu);
5124 5125
	struct task_struct *next, *stop = rq->stop;
	int dest_cpu;
L
Linus Torvalds 已提交
5126 5127

	/*
5128 5129 5130 5131 5132 5133 5134
	 * 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
5135
	 */
5136
	rq->stop = NULL;
5137

5138 5139 5140
	/* Ensure any throttled groups are reachable by pick_next_task */
	unthrottle_offline_cfs_rqs(rq);

I
Ingo Molnar 已提交
5141
	for ( ; ; ) {
5142 5143 5144 5145 5146
		/*
		 * There's this thread running, bail when that's the only
		 * remaining thread.
		 */
		if (rq->nr_running == 1)
I
Ingo Molnar 已提交
5147
			break;
5148

5149
		next = pick_next_task(rq);
5150
		BUG_ON(!next);
D
Dmitry Adamushko 已提交
5151
		next->sched_class->put_prev_task(rq, next);
5152

5153 5154 5155 5156 5157 5158 5159
		/* 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
5160
	}
5161

5162
	rq->stop = stop;
5163
}
5164

L
Linus Torvalds 已提交
5165 5166
#endif /* CONFIG_HOTPLUG_CPU */

5167 5168 5169
#if defined(CONFIG_SCHED_DEBUG) && defined(CONFIG_SYSCTL)

static struct ctl_table sd_ctl_dir[] = {
5170 5171
	{
		.procname	= "sched_domain",
5172
		.mode		= 0555,
5173
	},
5174
	{}
5175 5176 5177
};

static struct ctl_table sd_ctl_root[] = {
5178 5179
	{
		.procname	= "kernel",
5180
		.mode		= 0555,
5181 5182
		.child		= sd_ctl_dir,
	},
5183
	{}
5184 5185 5186 5187 5188
};

static struct ctl_table *sd_alloc_ctl_entry(int n)
{
	struct ctl_table *entry =
5189
		kcalloc(n, sizeof(struct ctl_table), GFP_KERNEL);
5190 5191 5192 5193

	return entry;
}

5194 5195
static void sd_free_ctl_entry(struct ctl_table **tablep)
{
5196
	struct ctl_table *entry;
5197

5198 5199 5200
	/*
	 * 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
5201
	 * will always be set. In the lowest directory the names are
5202 5203 5204
	 * static strings and all have proc handlers.
	 */
	for (entry = *tablep; entry->mode; entry++) {
5205 5206
		if (entry->child)
			sd_free_ctl_entry(&entry->child);
5207 5208 5209
		if (entry->proc_handler == NULL)
			kfree(entry->procname);
	}
5210 5211 5212 5213 5214

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

5215
static void
5216
set_table_entry(struct ctl_table *entry,
5217
		const char *procname, void *data, int maxlen,
5218
		umode_t mode, proc_handler *proc_handler)
5219 5220 5221 5222 5223 5224 5225 5226 5227 5228 5229
{
	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)
{
5230
	struct ctl_table *table = sd_alloc_ctl_entry(13);
5231

5232 5233 5234
	if (table == NULL)
		return NULL;

5235
	set_table_entry(&table[0], "min_interval", &sd->min_interval,
5236
		sizeof(long), 0644, proc_doulongvec_minmax);
5237
	set_table_entry(&table[1], "max_interval", &sd->max_interval,
5238
		sizeof(long), 0644, proc_doulongvec_minmax);
5239
	set_table_entry(&table[2], "busy_idx", &sd->busy_idx,
5240
		sizeof(int), 0644, proc_dointvec_minmax);
5241
	set_table_entry(&table[3], "idle_idx", &sd->idle_idx,
5242
		sizeof(int), 0644, proc_dointvec_minmax);
5243
	set_table_entry(&table[4], "newidle_idx", &sd->newidle_idx,
5244
		sizeof(int), 0644, proc_dointvec_minmax);
5245
	set_table_entry(&table[5], "wake_idx", &sd->wake_idx,
5246
		sizeof(int), 0644, proc_dointvec_minmax);
5247
	set_table_entry(&table[6], "forkexec_idx", &sd->forkexec_idx,
5248
		sizeof(int), 0644, proc_dointvec_minmax);
5249
	set_table_entry(&table[7], "busy_factor", &sd->busy_factor,
5250
		sizeof(int), 0644, proc_dointvec_minmax);
5251
	set_table_entry(&table[8], "imbalance_pct", &sd->imbalance_pct,
5252
		sizeof(int), 0644, proc_dointvec_minmax);
5253
	set_table_entry(&table[9], "cache_nice_tries",
5254 5255
		&sd->cache_nice_tries,
		sizeof(int), 0644, proc_dointvec_minmax);
5256
	set_table_entry(&table[10], "flags", &sd->flags,
5257
		sizeof(int), 0644, proc_dointvec_minmax);
5258 5259 5260
	set_table_entry(&table[11], "name", sd->name,
		CORENAME_MAX_SIZE, 0444, proc_dostring);
	/* &table[12] is terminator */
5261 5262 5263 5264

	return table;
}

5265
static ctl_table *sd_alloc_ctl_cpu_table(int cpu)
5266 5267 5268 5269 5270 5271 5272 5273 5274
{
	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);
5275 5276
	if (table == NULL)
		return NULL;
5277 5278 5279 5280 5281

	i = 0;
	for_each_domain(cpu, sd) {
		snprintf(buf, 32, "domain%d", i);
		entry->procname = kstrdup(buf, GFP_KERNEL);
5282
		entry->mode = 0555;
5283 5284 5285 5286 5287 5288 5289 5290
		entry->child = sd_alloc_ctl_domain_table(sd);
		entry++;
		i++;
	}
	return table;
}

static struct ctl_table_header *sd_sysctl_header;
5291
static void register_sched_domain_sysctl(void)
5292
{
5293
	int i, cpu_num = num_possible_cpus();
5294 5295 5296
	struct ctl_table *entry = sd_alloc_ctl_entry(cpu_num + 1);
	char buf[32];

5297 5298 5299
	WARN_ON(sd_ctl_dir[0].child);
	sd_ctl_dir[0].child = entry;

5300 5301 5302
	if (entry == NULL)
		return;

5303
	for_each_possible_cpu(i) {
5304 5305
		snprintf(buf, 32, "cpu%d", i);
		entry->procname = kstrdup(buf, GFP_KERNEL);
5306
		entry->mode = 0555;
5307
		entry->child = sd_alloc_ctl_cpu_table(i);
5308
		entry++;
5309
	}
5310 5311

	WARN_ON(sd_sysctl_header);
5312 5313
	sd_sysctl_header = register_sysctl_table(sd_ctl_root);
}
5314

5315
/* may be called multiple times per register */
5316 5317
static void unregister_sched_domain_sysctl(void)
{
5318 5319
	if (sd_sysctl_header)
		unregister_sysctl_table(sd_sysctl_header);
5320
	sd_sysctl_header = NULL;
5321 5322
	if (sd_ctl_dir[0].child)
		sd_free_ctl_entry(&sd_ctl_dir[0].child);
5323
}
5324
#else
5325 5326 5327 5328
static void register_sched_domain_sysctl(void)
{
}
static void unregister_sched_domain_sysctl(void)
5329 5330 5331 5332
{
}
#endif

5333 5334 5335 5336 5337
static void set_rq_online(struct rq *rq)
{
	if (!rq->online) {
		const struct sched_class *class;

5338
		cpumask_set_cpu(rq->cpu, rq->rd->online);
5339 5340 5341 5342 5343 5344 5345 5346 5347 5348 5349 5350 5351 5352 5353 5354 5355 5356 5357
		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);
		}

5358
		cpumask_clear_cpu(rq->cpu, rq->rd->online);
5359 5360 5361 5362
		rq->online = 0;
	}
}

L
Linus Torvalds 已提交
5363 5364 5365 5366
/*
 * migration_call - callback that gets triggered when a CPU is added.
 * Here we can start up the necessary migration thread for the new CPU.
 */
5367 5368
static int __cpuinit
migration_call(struct notifier_block *nfb, unsigned long action, void *hcpu)
L
Linus Torvalds 已提交
5369
{
5370
	int cpu = (long)hcpu;
L
Linus Torvalds 已提交
5371
	unsigned long flags;
5372
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
5373

5374
	switch (action & ~CPU_TASKS_FROZEN) {
5375

L
Linus Torvalds 已提交
5376
	case CPU_UP_PREPARE:
5377
		rq->calc_load_update = calc_load_update;
L
Linus Torvalds 已提交
5378
		break;
5379

L
Linus Torvalds 已提交
5380
	case CPU_ONLINE:
5381
		/* Update our root-domain */
5382
		raw_spin_lock_irqsave(&rq->lock, flags);
5383
		if (rq->rd) {
5384
			BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
5385 5386

			set_rq_online(rq);
5387
		}
5388
		raw_spin_unlock_irqrestore(&rq->lock, flags);
L
Linus Torvalds 已提交
5389
		break;
5390

L
Linus Torvalds 已提交
5391
#ifdef CONFIG_HOTPLUG_CPU
5392
	case CPU_DYING:
5393
		sched_ttwu_pending();
G
Gregory Haskins 已提交
5394
		/* Update our root-domain */
5395
		raw_spin_lock_irqsave(&rq->lock, flags);
G
Gregory Haskins 已提交
5396
		if (rq->rd) {
5397
			BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
5398
			set_rq_offline(rq);
G
Gregory Haskins 已提交
5399
		}
5400 5401
		migrate_tasks(cpu);
		BUG_ON(rq->nr_running != 1); /* the migration thread */
5402
		raw_spin_unlock_irqrestore(&rq->lock, flags);
5403 5404 5405

		migrate_nr_uninterruptible(rq);
		calc_global_load_remove(rq);
G
Gregory Haskins 已提交
5406
		break;
L
Linus Torvalds 已提交
5407 5408
#endif
	}
5409 5410 5411

	update_max_interval();

L
Linus Torvalds 已提交
5412 5413 5414
	return NOTIFY_OK;
}

5415 5416 5417
/*
 * Register at high priority so that task migration (migrate_all_tasks)
 * happens before everything else.  This has to be lower priority than
5418
 * the notifier in the perf_event subsystem, though.
L
Linus Torvalds 已提交
5419
 */
5420
static struct notifier_block __cpuinitdata migration_notifier = {
L
Linus Torvalds 已提交
5421
	.notifier_call = migration_call,
5422
	.priority = CPU_PRI_MIGRATION,
L
Linus Torvalds 已提交
5423 5424
};

5425 5426 5427 5428
static int __cpuinit sched_cpu_active(struct notifier_block *nfb,
				      unsigned long action, void *hcpu)
{
	switch (action & ~CPU_TASKS_FROZEN) {
5429
	case CPU_STARTING:
5430 5431 5432 5433 5434 5435 5436 5437 5438 5439 5440 5441 5442 5443 5444 5445 5446 5447 5448 5449
	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;
	}
}

5450
static int __init migration_init(void)
L
Linus Torvalds 已提交
5451 5452
{
	void *cpu = (void *)(long)smp_processor_id();
5453
	int err;
5454

5455
	/* Initialize migration for the boot CPU */
5456 5457
	err = migration_call(&migration_notifier, CPU_UP_PREPARE, cpu);
	BUG_ON(err == NOTIFY_BAD);
L
Linus Torvalds 已提交
5458 5459
	migration_call(&migration_notifier, CPU_ONLINE, cpu);
	register_cpu_notifier(&migration_notifier);
5460

5461 5462 5463 5464
	/* Register cpu active notifiers */
	cpu_notifier(sched_cpu_active, CPU_PRI_SCHED_ACTIVE);
	cpu_notifier(sched_cpu_inactive, CPU_PRI_SCHED_INACTIVE);

5465
	return 0;
L
Linus Torvalds 已提交
5466
}
5467
early_initcall(migration_init);
L
Linus Torvalds 已提交
5468 5469 5470
#endif

#ifdef CONFIG_SMP
5471

5472 5473
static cpumask_var_t sched_domains_tmpmask; /* sched_domains_mutex */

5474
#ifdef CONFIG_SCHED_DEBUG
I
Ingo Molnar 已提交
5475

5476 5477 5478 5479 5480 5481 5482 5483 5484 5485
static __read_mostly int sched_domain_debug_enabled;

static int __init sched_domain_debug_setup(char *str)
{
	sched_domain_debug_enabled = 1;

	return 0;
}
early_param("sched_debug", sched_domain_debug_setup);

5486
static int sched_domain_debug_one(struct sched_domain *sd, int cpu, int level,
5487
				  struct cpumask *groupmask)
L
Linus Torvalds 已提交
5488
{
I
Ingo Molnar 已提交
5489
	struct sched_group *group = sd->groups;
5490
	char str[256];
L
Linus Torvalds 已提交
5491

R
Rusty Russell 已提交
5492
	cpulist_scnprintf(str, sizeof(str), sched_domain_span(sd));
5493
	cpumask_clear(groupmask);
I
Ingo Molnar 已提交
5494 5495 5496 5497

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

	if (!(sd->flags & SD_LOAD_BALANCE)) {
P
Peter Zijlstra 已提交
5498
		printk("does not load-balance\n");
I
Ingo Molnar 已提交
5499
		if (sd->parent)
P
Peter Zijlstra 已提交
5500 5501
			printk(KERN_ERR "ERROR: !SD_LOAD_BALANCE domain"
					" has parent");
I
Ingo Molnar 已提交
5502
		return -1;
N
Nick Piggin 已提交
5503 5504
	}

P
Peter Zijlstra 已提交
5505
	printk(KERN_CONT "span %s level %s\n", str, sd->name);
I
Ingo Molnar 已提交
5506

5507
	if (!cpumask_test_cpu(cpu, sched_domain_span(sd))) {
P
Peter Zijlstra 已提交
5508 5509
		printk(KERN_ERR "ERROR: domain->span does not contain "
				"CPU%d\n", cpu);
I
Ingo Molnar 已提交
5510
	}
5511
	if (!cpumask_test_cpu(cpu, sched_group_cpus(group))) {
P
Peter Zijlstra 已提交
5512 5513
		printk(KERN_ERR "ERROR: domain->groups does not contain"
				" CPU%d\n", cpu);
I
Ingo Molnar 已提交
5514
	}
L
Linus Torvalds 已提交
5515

I
Ingo Molnar 已提交
5516
	printk(KERN_DEBUG "%*s groups:", level + 1, "");
L
Linus Torvalds 已提交
5517
	do {
I
Ingo Molnar 已提交
5518
		if (!group) {
P
Peter Zijlstra 已提交
5519 5520
			printk("\n");
			printk(KERN_ERR "ERROR: group is NULL\n");
L
Linus Torvalds 已提交
5521 5522 5523
			break;
		}

5524
		if (!group->sgp->power) {
P
Peter Zijlstra 已提交
5525 5526 5527
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: domain->cpu_power not "
					"set\n");
I
Ingo Molnar 已提交
5528 5529
			break;
		}
L
Linus Torvalds 已提交
5530

5531
		if (!cpumask_weight(sched_group_cpus(group))) {
P
Peter Zijlstra 已提交
5532 5533
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: empty group\n");
I
Ingo Molnar 已提交
5534 5535
			break;
		}
L
Linus Torvalds 已提交
5536

5537
		if (cpumask_intersects(groupmask, sched_group_cpus(group))) {
P
Peter Zijlstra 已提交
5538 5539
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: repeated CPUs\n");
I
Ingo Molnar 已提交
5540 5541
			break;
		}
L
Linus Torvalds 已提交
5542

5543
		cpumask_or(groupmask, groupmask, sched_group_cpus(group));
L
Linus Torvalds 已提交
5544

R
Rusty Russell 已提交
5545
		cpulist_scnprintf(str, sizeof(str), sched_group_cpus(group));
5546

P
Peter Zijlstra 已提交
5547
		printk(KERN_CONT " %s", str);
5548
		if (group->sgp->power != SCHED_POWER_SCALE) {
P
Peter Zijlstra 已提交
5549
			printk(KERN_CONT " (cpu_power = %d)",
5550
				group->sgp->power);
5551
		}
L
Linus Torvalds 已提交
5552

I
Ingo Molnar 已提交
5553 5554
		group = group->next;
	} while (group != sd->groups);
P
Peter Zijlstra 已提交
5555
	printk(KERN_CONT "\n");
L
Linus Torvalds 已提交
5556

5557
	if (!cpumask_equal(sched_domain_span(sd), groupmask))
P
Peter Zijlstra 已提交
5558
		printk(KERN_ERR "ERROR: groups don't span domain->span\n");
L
Linus Torvalds 已提交
5559

5560 5561
	if (sd->parent &&
	    !cpumask_subset(groupmask, sched_domain_span(sd->parent)))
P
Peter Zijlstra 已提交
5562 5563
		printk(KERN_ERR "ERROR: parent span is not a superset "
			"of domain->span\n");
I
Ingo Molnar 已提交
5564 5565
	return 0;
}
L
Linus Torvalds 已提交
5566

I
Ingo Molnar 已提交
5567 5568 5569
static void sched_domain_debug(struct sched_domain *sd, int cpu)
{
	int level = 0;
L
Linus Torvalds 已提交
5570

5571 5572 5573
	if (!sched_domain_debug_enabled)
		return;

I
Ingo Molnar 已提交
5574 5575 5576 5577
	if (!sd) {
		printk(KERN_DEBUG "CPU%d attaching NULL sched-domain.\n", cpu);
		return;
	}
L
Linus Torvalds 已提交
5578

I
Ingo Molnar 已提交
5579 5580 5581
	printk(KERN_DEBUG "CPU%d attaching sched-domain:\n", cpu);

	for (;;) {
5582
		if (sched_domain_debug_one(sd, cpu, level, sched_domains_tmpmask))
I
Ingo Molnar 已提交
5583
			break;
L
Linus Torvalds 已提交
5584 5585
		level++;
		sd = sd->parent;
5586
		if (!sd)
I
Ingo Molnar 已提交
5587 5588
			break;
	}
L
Linus Torvalds 已提交
5589
}
5590
#else /* !CONFIG_SCHED_DEBUG */
5591
# define sched_domain_debug(sd, cpu) do { } while (0)
5592
#endif /* CONFIG_SCHED_DEBUG */
L
Linus Torvalds 已提交
5593

5594
static int sd_degenerate(struct sched_domain *sd)
5595
{
5596
	if (cpumask_weight(sched_domain_span(sd)) == 1)
5597 5598 5599 5600 5601 5602
		return 1;

	/* Following flags need at least 2 groups */
	if (sd->flags & (SD_LOAD_BALANCE |
			 SD_BALANCE_NEWIDLE |
			 SD_BALANCE_FORK |
5603 5604 5605
			 SD_BALANCE_EXEC |
			 SD_SHARE_CPUPOWER |
			 SD_SHARE_PKG_RESOURCES)) {
5606 5607 5608 5609 5610
		if (sd->groups != sd->groups->next)
			return 0;
	}

	/* Following flags don't use groups */
5611
	if (sd->flags & (SD_WAKE_AFFINE))
5612 5613 5614 5615 5616
		return 0;

	return 1;
}

5617 5618
static int
sd_parent_degenerate(struct sched_domain *sd, struct sched_domain *parent)
5619 5620 5621 5622 5623 5624
{
	unsigned long cflags = sd->flags, pflags = parent->flags;

	if (sd_degenerate(parent))
		return 1;

5625
	if (!cpumask_equal(sched_domain_span(sd), sched_domain_span(parent)))
5626 5627 5628 5629 5630 5631 5632
		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 |
5633 5634 5635
				SD_BALANCE_EXEC |
				SD_SHARE_CPUPOWER |
				SD_SHARE_PKG_RESOURCES);
5636 5637
		if (nr_node_ids == 1)
			pflags &= ~SD_SERIALIZE;
5638 5639 5640 5641 5642 5643 5644
	}
	if (~cflags & pflags)
		return 0;

	return 1;
}

5645
static void free_rootdomain(struct rcu_head *rcu)
5646
{
5647
	struct root_domain *rd = container_of(rcu, struct root_domain, rcu);
5648

5649
	cpupri_cleanup(&rd->cpupri);
5650 5651 5652 5653 5654 5655
	free_cpumask_var(rd->rto_mask);
	free_cpumask_var(rd->online);
	free_cpumask_var(rd->span);
	kfree(rd);
}

G
Gregory Haskins 已提交
5656 5657
static void rq_attach_root(struct rq *rq, struct root_domain *rd)
{
I
Ingo Molnar 已提交
5658
	struct root_domain *old_rd = NULL;
G
Gregory Haskins 已提交
5659 5660
	unsigned long flags;

5661
	raw_spin_lock_irqsave(&rq->lock, flags);
G
Gregory Haskins 已提交
5662 5663

	if (rq->rd) {
I
Ingo Molnar 已提交
5664
		old_rd = rq->rd;
G
Gregory Haskins 已提交
5665

5666
		if (cpumask_test_cpu(rq->cpu, old_rd->online))
5667
			set_rq_offline(rq);
G
Gregory Haskins 已提交
5668

5669
		cpumask_clear_cpu(rq->cpu, old_rd->span);
5670

I
Ingo Molnar 已提交
5671 5672 5673 5674 5675 5676 5677
		/*
		 * 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
5678 5679 5680 5681 5682
	}

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

5683
	cpumask_set_cpu(rq->cpu, rd->span);
5684
	if (cpumask_test_cpu(rq->cpu, cpu_active_mask))
5685
		set_rq_online(rq);
G
Gregory Haskins 已提交
5686

5687
	raw_spin_unlock_irqrestore(&rq->lock, flags);
I
Ingo Molnar 已提交
5688 5689

	if (old_rd)
5690
		call_rcu_sched(&old_rd->rcu, free_rootdomain);
G
Gregory Haskins 已提交
5691 5692
}

5693
static int init_rootdomain(struct root_domain *rd)
G
Gregory Haskins 已提交
5694 5695 5696
{
	memset(rd, 0, sizeof(*rd));

5697
	if (!alloc_cpumask_var(&rd->span, GFP_KERNEL))
5698
		goto out;
5699
	if (!alloc_cpumask_var(&rd->online, GFP_KERNEL))
5700
		goto free_span;
5701
	if (!alloc_cpumask_var(&rd->rto_mask, GFP_KERNEL))
5702
		goto free_online;
5703

5704
	if (cpupri_init(&rd->cpupri) != 0)
5705
		goto free_rto_mask;
5706
	return 0;
5707

5708 5709
free_rto_mask:
	free_cpumask_var(rd->rto_mask);
5710 5711 5712 5713
free_online:
	free_cpumask_var(rd->online);
free_span:
	free_cpumask_var(rd->span);
5714
out:
5715
	return -ENOMEM;
G
Gregory Haskins 已提交
5716 5717
}

5718 5719 5720 5721 5722 5723
/*
 * 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
5724 5725
static void init_defrootdomain(void)
{
5726
	init_rootdomain(&def_root_domain);
5727

G
Gregory Haskins 已提交
5728 5729 5730
	atomic_set(&def_root_domain.refcount, 1);
}

5731
static struct root_domain *alloc_rootdomain(void)
G
Gregory Haskins 已提交
5732 5733 5734 5735 5736 5737 5738
{
	struct root_domain *rd;

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

5739
	if (init_rootdomain(rd) != 0) {
5740 5741 5742
		kfree(rd);
		return NULL;
	}
G
Gregory Haskins 已提交
5743 5744 5745 5746

	return rd;
}

5747 5748 5749 5750 5751 5752 5753 5754 5755 5756 5757 5758 5759 5760 5761 5762 5763 5764 5765
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);
}

5766 5767 5768
static void free_sched_domain(struct rcu_head *rcu)
{
	struct sched_domain *sd = container_of(rcu, struct sched_domain, rcu);
5769 5770 5771 5772 5773 5774 5775 5776

	/*
	 * 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)) {
5777
		kfree(sd->groups->sgp);
5778
		kfree(sd->groups);
5779
	}
5780 5781 5782 5783 5784 5785 5786 5787 5788 5789 5790 5791 5792 5793
	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);
}

5794 5795 5796 5797 5798 5799 5800
/*
 * Keep a special pointer to the highest sched_domain that has
 * SD_SHARE_PKG_RESOURCE set (Last Level Cache Domain) for this
 * allows us to avoid some pointer chasing select_idle_sibling().
 *
 * Also keep a unique ID per domain (we use the first cpu number in
 * the cpumask of the domain), this allows us to quickly tell if
5801
 * two cpus are in the same cache domain, see cpus_share_cache().
5802 5803 5804 5805 5806 5807 5808 5809 5810 5811 5812 5813 5814 5815 5816 5817 5818
 */
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);
	if (sd)
		id = cpumask_first(sched_domain_span(sd));

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

L
Linus Torvalds 已提交
5819
/*
I
Ingo Molnar 已提交
5820
 * Attach the domain 'sd' to 'cpu' as its base domain. Callers must
L
Linus Torvalds 已提交
5821 5822
 * hold the hotplug lock.
 */
I
Ingo Molnar 已提交
5823 5824
static void
cpu_attach_domain(struct sched_domain *sd, struct root_domain *rd, int cpu)
L
Linus Torvalds 已提交
5825
{
5826
	struct rq *rq = cpu_rq(cpu);
5827 5828 5829
	struct sched_domain *tmp;

	/* Remove the sched domains which do not contribute to scheduling. */
5830
	for (tmp = sd; tmp; ) {
5831 5832 5833
		struct sched_domain *parent = tmp->parent;
		if (!parent)
			break;
5834

5835
		if (sd_parent_degenerate(tmp, parent)) {
5836
			tmp->parent = parent->parent;
5837 5838
			if (parent->parent)
				parent->parent->child = tmp;
5839
			destroy_sched_domain(parent, cpu);
5840 5841
		} else
			tmp = tmp->parent;
5842 5843
	}

5844
	if (sd && sd_degenerate(sd)) {
5845
		tmp = sd;
5846
		sd = sd->parent;
5847
		destroy_sched_domain(tmp, cpu);
5848 5849 5850
		if (sd)
			sd->child = NULL;
	}
L
Linus Torvalds 已提交
5851

5852
	sched_domain_debug(sd, cpu);
L
Linus Torvalds 已提交
5853

G
Gregory Haskins 已提交
5854
	rq_attach_root(rq, rd);
5855
	tmp = rq->sd;
N
Nick Piggin 已提交
5856
	rcu_assign_pointer(rq->sd, sd);
5857
	destroy_sched_domains(tmp, cpu);
5858 5859

	update_top_cache_domain(cpu);
L
Linus Torvalds 已提交
5860 5861 5862
}

/* cpus with isolated domains */
5863
static cpumask_var_t cpu_isolated_map;
L
Linus Torvalds 已提交
5864 5865 5866 5867

/* Setup the mask of cpus configured for isolated domains */
static int __init isolated_cpu_setup(char *str)
{
R
Rusty Russell 已提交
5868
	alloc_bootmem_cpumask_var(&cpu_isolated_map);
R
Rusty Russell 已提交
5869
	cpulist_parse(str, cpu_isolated_map);
L
Linus Torvalds 已提交
5870 5871 5872
	return 1;
}

I
Ingo Molnar 已提交
5873
__setup("isolcpus=", isolated_cpu_setup);
L
Linus Torvalds 已提交
5874

5875
#ifdef CONFIG_NUMA
5876

5877 5878 5879 5880 5881
/**
 * find_next_best_node - find the next node to include in a sched_domain
 * @node: node whose sched_domain we're building
 * @used_nodes: nodes already in the sched_domain
 *
I
Ingo Molnar 已提交
5882
 * Find the next node to include in a given scheduling domain. Simply
5883 5884 5885 5886
 * finds the closest node not already in the @used_nodes map.
 *
 * Should use nodemask_t.
 */
5887
static int find_next_best_node(int node, nodemask_t *used_nodes)
5888
{
5889
	int i, n, val, min_val, best_node = -1;
5890 5891 5892

	min_val = INT_MAX;

5893
	for (i = 0; i < nr_node_ids; i++) {
5894
		/* Start at @node */
5895
		n = (node + i) % nr_node_ids;
5896 5897 5898 5899 5900

		if (!nr_cpus_node(n))
			continue;

		/* Skip already used nodes */
5901
		if (node_isset(n, *used_nodes))
5902 5903 5904 5905 5906 5907 5908 5909 5910 5911 5912
			continue;

		/* Simple min distance search */
		val = node_distance(node, n);

		if (val < min_val) {
			min_val = val;
			best_node = n;
		}
	}

5913 5914
	if (best_node != -1)
		node_set(best_node, *used_nodes);
5915 5916 5917 5918 5919 5920
	return best_node;
}

/**
 * sched_domain_node_span - get a cpumask for a node's sched_domain
 * @node: node whose cpumask we're constructing
5921
 * @span: resulting cpumask
5922
 *
I
Ingo Molnar 已提交
5923
 * Given a node, construct a good cpumask for its sched_domain to span. It
5924 5925 5926
 * should be one that prevents unnecessary balancing, but also spreads tasks
 * out optimally.
 */
5927
static void sched_domain_node_span(int node, struct cpumask *span)
5928
{
5929
	nodemask_t used_nodes;
5930
	int i;
5931

5932
	cpumask_clear(span);
5933
	nodes_clear(used_nodes);
5934

5935
	cpumask_or(span, span, cpumask_of_node(node));
5936
	node_set(node, used_nodes);
5937 5938

	for (i = 1; i < SD_NODES_PER_DOMAIN; i++) {
5939
		int next_node = find_next_best_node(node, &used_nodes);
5940 5941
		if (next_node < 0)
			break;
5942
		cpumask_or(span, span, cpumask_of_node(next_node));
5943 5944
	}
}
5945 5946 5947 5948 5949 5950 5951 5952 5953

static const struct cpumask *cpu_node_mask(int cpu)
{
	lockdep_assert_held(&sched_domains_mutex);

	sched_domain_node_span(cpu_to_node(cpu), sched_domains_tmpmask);

	return sched_domains_tmpmask;
}
5954 5955 5956 5957 5958

static const struct cpumask *cpu_allnodes_mask(int cpu)
{
	return cpu_possible_mask;
}
5959
#endif /* CONFIG_NUMA */
5960

5961 5962 5963 5964 5965
static const struct cpumask *cpu_cpu_mask(int cpu)
{
	return cpumask_of_node(cpu_to_node(cpu));
}

5966
int sched_smt_power_savings = 0, sched_mc_power_savings = 0;
5967

5968 5969 5970
struct sd_data {
	struct sched_domain **__percpu sd;
	struct sched_group **__percpu sg;
5971
	struct sched_group_power **__percpu sgp;
5972 5973
};

5974
struct s_data {
5975
	struct sched_domain ** __percpu sd;
5976 5977 5978
	struct root_domain	*rd;
};

5979 5980
enum s_alloc {
	sa_rootdomain,
5981
	sa_sd,
5982
	sa_sd_storage,
5983 5984 5985
	sa_none,
};

5986 5987 5988
struct sched_domain_topology_level;

typedef struct sched_domain *(*sched_domain_init_f)(struct sched_domain_topology_level *tl, int cpu);
5989 5990
typedef const struct cpumask *(*sched_domain_mask_f)(int cpu);

5991 5992
#define SDTL_OVERLAP	0x01

5993
struct sched_domain_topology_level {
5994 5995
	sched_domain_init_f init;
	sched_domain_mask_f mask;
5996
	int		    flags;
5997
	struct sd_data      data;
5998 5999
};

6000 6001 6002 6003 6004 6005 6006 6007 6008 6009 6010 6011 6012 6013 6014 6015 6016 6017 6018
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;

		sg = kzalloc_node(sizeof(struct sched_group) + cpumask_size(),
6019
				GFP_KERNEL, cpu_to_node(cpu));
6020 6021 6022 6023 6024 6025 6026 6027 6028 6029 6030 6031 6032 6033 6034 6035 6036 6037 6038 6039 6040 6041 6042 6043 6044 6045 6046 6047 6048 6049 6050 6051 6052 6053 6054 6055 6056 6057

		if (!sg)
			goto fail;

		sg_span = sched_group_cpus(sg);

		child = *per_cpu_ptr(sdd->sd, i);
		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);

		sg->sgp = *per_cpu_ptr(sdd->sgp, cpumask_first(sg_span));
		atomic_inc(&sg->sgp->ref);

		if (cpumask_test_cpu(cpu, sg_span))
			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;
}

6058
static int get_group(int cpu, struct sd_data *sdd, struct sched_group **sg)
L
Linus Torvalds 已提交
6059
{
6060 6061
	struct sched_domain *sd = *per_cpu_ptr(sdd->sd, cpu);
	struct sched_domain *child = sd->child;
L
Linus Torvalds 已提交
6062

6063 6064
	if (child)
		cpu = cpumask_first(sched_domain_span(child));
6065

6066
	if (sg) {
6067
		*sg = *per_cpu_ptr(sdd->sg, cpu);
6068
		(*sg)->sgp = *per_cpu_ptr(sdd->sgp, cpu);
6069
		atomic_set(&(*sg)->sgp->ref, 1); /* for claim_allocations */
6070
	}
6071 6072

	return cpu;
6073 6074
}

6075
/*
6076 6077 6078
 * 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.
6079 6080
 *
 * Assumes the sched_domain tree is fully constructed
6081
 */
6082 6083
static int
build_sched_groups(struct sched_domain *sd, int cpu)
L
Linus Torvalds 已提交
6084
{
6085 6086 6087
	struct sched_group *first = NULL, *last = NULL;
	struct sd_data *sdd = sd->private;
	const struct cpumask *span = sched_domain_span(sd);
6088
	struct cpumask *covered;
6089
	int i;
6090

6091 6092 6093 6094 6095 6096
	get_group(cpu, sdd, &sd->groups);
	atomic_inc(&sd->groups->ref);

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

6097 6098 6099
	lockdep_assert_held(&sched_domains_mutex);
	covered = sched_domains_tmpmask;

6100
	cpumask_clear(covered);
6101

6102 6103 6104 6105
	for_each_cpu(i, span) {
		struct sched_group *sg;
		int group = get_group(i, sdd, &sg);
		int j;
6106

6107 6108
		if (cpumask_test_cpu(i, covered))
			continue;
6109

6110
		cpumask_clear(sched_group_cpus(sg));
6111
		sg->sgp->power = 0;
6112

6113 6114 6115
		for_each_cpu(j, span) {
			if (get_group(j, sdd, NULL) != group)
				continue;
6116

6117 6118 6119
			cpumask_set_cpu(j, covered);
			cpumask_set_cpu(j, sched_group_cpus(sg));
		}
6120

6121 6122 6123 6124 6125 6126 6127
		if (!first)
			first = sg;
		if (last)
			last->next = sg;
		last = sg;
	}
	last->next = first;
6128 6129

	return 0;
6130
}
6131

6132 6133 6134 6135 6136 6137 6138 6139 6140 6141 6142 6143
/*
 * 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)
{
6144
	struct sched_group *sg = sd->groups;
6145

6146 6147 6148 6149 6150 6151
	WARN_ON(!sd || !sg);

	do {
		sg->group_weight = cpumask_weight(sched_group_cpus(sg));
		sg = sg->next;
	} while (sg != sd->groups);
6152

6153 6154
	if (cpu != group_first_cpu(sg))
		return;
6155

6156
	update_group_power(sd, cpu);
6157
	atomic_set(&sg->sgp->nr_busy_cpus, sg->group_weight);
6158 6159
}

6160 6161 6162
int __weak arch_sd_sibling_asym_packing(void)
{
       return 0*SD_ASYM_PACKING;
6163 6164
}

6165 6166 6167 6168 6169
/*
 * Initializers for schedule domains
 * Non-inlined to reduce accumulated stack pressure in build_sched_domains()
 */

6170 6171 6172 6173 6174 6175
#ifdef CONFIG_SCHED_DEBUG
# define SD_INIT_NAME(sd, type)		sd->name = #type
#else
# define SD_INIT_NAME(sd, type)		do { } while (0)
#endif

6176 6177 6178 6179 6180 6181 6182 6183 6184
#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;							\
6185 6186 6187 6188 6189 6190 6191 6192 6193 6194 6195 6196 6197
}

SD_INIT_FUNC(CPU)
#ifdef CONFIG_NUMA
 SD_INIT_FUNC(ALLNODES)
 SD_INIT_FUNC(NODE)
#endif
#ifdef CONFIG_SCHED_SMT
 SD_INIT_FUNC(SIBLING)
#endif
#ifdef CONFIG_SCHED_MC
 SD_INIT_FUNC(MC)
#endif
6198 6199 6200
#ifdef CONFIG_SCHED_BOOK
 SD_INIT_FUNC(BOOK)
#endif
6201

6202
static int default_relax_domain_level = -1;
6203
int sched_domain_level_max;
6204 6205 6206

static int __init setup_relax_domain_level(char *str)
{
6207 6208 6209
	unsigned long val;

	val = simple_strtoul(str, NULL, 0);
6210
	if (val < sched_domain_level_max)
6211 6212
		default_relax_domain_level = val;

6213 6214 6215 6216 6217 6218 6219 6220 6221 6222 6223 6224 6225 6226 6227 6228 6229 6230
	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 */
6231
		sd->flags &= ~(SD_BALANCE_WAKE|SD_BALANCE_NEWIDLE);
6232 6233
	} else {
		/* turn on idle balance on this domain */
6234
		sd->flags |= (SD_BALANCE_WAKE|SD_BALANCE_NEWIDLE);
6235 6236 6237
	}
}

6238 6239 6240
static void __sdt_free(const struct cpumask *cpu_map);
static int __sdt_alloc(const struct cpumask *cpu_map);

6241 6242 6243 6244 6245
static void __free_domain_allocs(struct s_data *d, enum s_alloc what,
				 const struct cpumask *cpu_map)
{
	switch (what) {
	case sa_rootdomain:
6246 6247
		if (!atomic_read(&d->rd->refcount))
			free_rootdomain(&d->rd->rcu); /* fall through */
6248 6249
	case sa_sd:
		free_percpu(d->sd); /* fall through */
6250
	case sa_sd_storage:
6251
		__sdt_free(cpu_map); /* fall through */
6252 6253 6254 6255
	case sa_none:
		break;
	}
}
6256

6257 6258 6259
static enum s_alloc __visit_domain_allocation_hell(struct s_data *d,
						   const struct cpumask *cpu_map)
{
6260 6261
	memset(d, 0, sizeof(*d));

6262 6263
	if (__sdt_alloc(cpu_map))
		return sa_sd_storage;
6264 6265 6266
	d->sd = alloc_percpu(struct sched_domain *);
	if (!d->sd)
		return sa_sd_storage;
6267
	d->rd = alloc_rootdomain();
6268
	if (!d->rd)
6269
		return sa_sd;
6270 6271
	return sa_rootdomain;
}
G
Gregory Haskins 已提交
6272

6273 6274 6275 6276 6277 6278 6279 6280 6281 6282 6283 6284
/*
 * 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;

6285
	if (atomic_read(&(*per_cpu_ptr(sdd->sg, cpu))->ref))
6286
		*per_cpu_ptr(sdd->sg, cpu) = NULL;
6287 6288

	if (atomic_read(&(*per_cpu_ptr(sdd->sgp, cpu))->ref))
6289
		*per_cpu_ptr(sdd->sgp, cpu) = NULL;
6290 6291
}

6292 6293
#ifdef CONFIG_SCHED_SMT
static const struct cpumask *cpu_smt_mask(int cpu)
6294
{
6295
	return topology_thread_cpumask(cpu);
6296
}
6297
#endif
6298

6299 6300 6301
/*
 * Topology list, bottom-up.
 */
6302
static struct sched_domain_topology_level default_topology[] = {
6303 6304
#ifdef CONFIG_SCHED_SMT
	{ sd_init_SIBLING, cpu_smt_mask, },
6305
#endif
6306
#ifdef CONFIG_SCHED_MC
6307
	{ sd_init_MC, cpu_coregroup_mask, },
6308
#endif
6309 6310 6311 6312 6313
#ifdef CONFIG_SCHED_BOOK
	{ sd_init_BOOK, cpu_book_mask, },
#endif
	{ sd_init_CPU, cpu_cpu_mask, },
#ifdef CONFIG_NUMA
6314
	{ sd_init_NODE, cpu_node_mask, SDTL_OVERLAP, },
6315
	{ sd_init_ALLNODES, cpu_allnodes_mask, },
L
Linus Torvalds 已提交
6316
#endif
6317 6318 6319 6320 6321
	{ NULL, },
};

static struct sched_domain_topology_level *sched_domain_topology = default_topology;

6322 6323 6324 6325 6326 6327 6328 6329 6330 6331 6332 6333 6334 6335 6336 6337
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;

6338 6339 6340 6341
		sdd->sgp = alloc_percpu(struct sched_group_power *);
		if (!sdd->sgp)
			return -ENOMEM;

6342 6343 6344
		for_each_cpu(j, cpu_map) {
			struct sched_domain *sd;
			struct sched_group *sg;
6345
			struct sched_group_power *sgp;
6346 6347 6348 6349 6350 6351 6352 6353 6354 6355 6356 6357 6358 6359

		       	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;

			*per_cpu_ptr(sdd->sg, j) = sg;
6360 6361 6362 6363 6364 6365 6366

			sgp = kzalloc_node(sizeof(struct sched_group_power),
					GFP_KERNEL, cpu_to_node(j));
			if (!sgp)
				return -ENOMEM;

			*per_cpu_ptr(sdd->sgp, j) = sgp;
6367 6368 6369 6370 6371 6372 6373 6374 6375 6376 6377 6378 6379 6380 6381
		}
	}

	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) {
6382 6383 6384
			struct sched_domain *sd = *per_cpu_ptr(sdd->sd, j);
			if (sd && (sd->flags & SD_OVERLAP))
				free_sched_groups(sd->groups, 0);
6385
			kfree(*per_cpu_ptr(sdd->sd, j));
6386
			kfree(*per_cpu_ptr(sdd->sg, j));
6387
			kfree(*per_cpu_ptr(sdd->sgp, j));
6388 6389 6390
		}
		free_percpu(sdd->sd);
		free_percpu(sdd->sg);
6391
		free_percpu(sdd->sgp);
6392 6393 6394
	}
}

6395 6396
struct sched_domain *build_sched_domain(struct sched_domain_topology_level *tl,
		struct s_data *d, const struct cpumask *cpu_map,
6397
		struct sched_domain_attr *attr, struct sched_domain *child,
6398 6399
		int cpu)
{
6400
	struct sched_domain *sd = tl->init(tl, cpu);
6401
	if (!sd)
6402
		return child;
6403 6404 6405

	set_domain_attribute(sd, attr);
	cpumask_and(sched_domain_span(sd), cpu_map, tl->mask(cpu));
6406 6407 6408
	if (child) {
		sd->level = child->level + 1;
		sched_domain_level_max = max(sched_domain_level_max, sd->level);
6409
		child->parent = sd;
6410
	}
6411
	sd->child = child;
6412 6413 6414 6415

	return sd;
}

6416 6417 6418 6419
/*
 * Build sched domains for a given set of cpus and attach the sched domains
 * to the individual cpus
 */
6420 6421
static int build_sched_domains(const struct cpumask *cpu_map,
			       struct sched_domain_attr *attr)
6422 6423
{
	enum s_alloc alloc_state = sa_none;
6424
	struct sched_domain *sd;
6425
	struct s_data d;
6426
	int i, ret = -ENOMEM;
6427

6428 6429 6430
	alloc_state = __visit_domain_allocation_hell(&d, cpu_map);
	if (alloc_state != sa_rootdomain)
		goto error;
6431

6432
	/* Set up domains for cpus specified by the cpu_map. */
6433
	for_each_cpu(i, cpu_map) {
6434 6435
		struct sched_domain_topology_level *tl;

6436
		sd = NULL;
6437
		for (tl = sched_domain_topology; tl->init; tl++) {
6438
			sd = build_sched_domain(tl, &d, cpu_map, attr, sd, i);
6439 6440
			if (tl->flags & SDTL_OVERLAP || sched_feat(FORCE_SD_OVERLAP))
				sd->flags |= SD_OVERLAP;
6441 6442
			if (cpumask_equal(cpu_map, sched_domain_span(sd)))
				break;
6443
		}
6444

6445 6446 6447
		while (sd->child)
			sd = sd->child;

6448
		*per_cpu_ptr(d.sd, i) = sd;
6449 6450 6451 6452 6453 6454
	}

	/* 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));
6455 6456 6457 6458 6459 6460 6461
			if (sd->flags & SD_OVERLAP) {
				if (build_overlap_sched_groups(sd, i))
					goto error;
			} else {
				if (build_sched_groups(sd, i))
					goto error;
			}
6462
		}
6463
	}
6464

L
Linus Torvalds 已提交
6465
	/* Calculate CPU power for physical packages and nodes */
6466 6467 6468
	for (i = nr_cpumask_bits-1; i >= 0; i--) {
		if (!cpumask_test_cpu(i, cpu_map))
			continue;
6469

6470 6471
		for (sd = *per_cpu_ptr(d.sd, i); sd; sd = sd->parent) {
			claim_allocations(i, sd);
6472
			init_sched_groups_power(i, sd);
6473
		}
6474
	}
6475

L
Linus Torvalds 已提交
6476
	/* Attach the domains */
6477
	rcu_read_lock();
6478
	for_each_cpu(i, cpu_map) {
6479
		sd = *per_cpu_ptr(d.sd, i);
6480
		cpu_attach_domain(sd, d.rd, i);
L
Linus Torvalds 已提交
6481
	}
6482
	rcu_read_unlock();
6483

6484
	ret = 0;
6485
error:
6486
	__free_domain_allocs(&d, alloc_state, cpu_map);
6487
	return ret;
L
Linus Torvalds 已提交
6488
}
P
Paul Jackson 已提交
6489

6490
static cpumask_var_t *doms_cur;	/* current sched domains */
P
Paul Jackson 已提交
6491
static int ndoms_cur;		/* number of sched domains in 'doms_cur' */
I
Ingo Molnar 已提交
6492 6493
static struct sched_domain_attr *dattr_cur;
				/* attribues of custom domains in 'doms_cur' */
P
Paul Jackson 已提交
6494 6495 6496

/*
 * Special case: If a kmalloc of a doms_cur partition (array of
6497 6498
 * cpumask) fails, then fallback to a single sched domain,
 * as determined by the single cpumask fallback_doms.
P
Paul Jackson 已提交
6499
 */
6500
static cpumask_var_t fallback_doms;
P
Paul Jackson 已提交
6501

6502 6503 6504 6505 6506 6507
/*
 * 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)
6508
{
6509
	return 0;
6510 6511
}

6512 6513 6514 6515 6516 6517 6518 6519 6520 6521 6522 6523 6524 6525 6526 6527 6528 6529 6530 6531 6532 6533 6534 6535 6536
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);
}

6537
/*
I
Ingo Molnar 已提交
6538
 * Set up scheduler domains and groups. Callers must hold the hotplug lock.
P
Paul Jackson 已提交
6539 6540
 * For now this just excludes isolated cpus, but could be used to
 * exclude other special cases in the future.
6541
 */
6542
static int init_sched_domains(const struct cpumask *cpu_map)
6543
{
6544 6545
	int err;

6546
	arch_update_cpu_topology();
P
Paul Jackson 已提交
6547
	ndoms_cur = 1;
6548
	doms_cur = alloc_sched_domains(ndoms_cur);
P
Paul Jackson 已提交
6549
	if (!doms_cur)
6550 6551
		doms_cur = &fallback_doms;
	cpumask_andnot(doms_cur[0], cpu_map, cpu_isolated_map);
6552
	dattr_cur = NULL;
6553
	err = build_sched_domains(doms_cur[0], NULL);
6554
	register_sched_domain_sysctl();
6555 6556

	return err;
6557 6558 6559 6560 6561 6562
}

/*
 * Detach sched domains from a group of cpus specified in cpu_map
 * These cpus will now be attached to the NULL domain
 */
6563
static void detach_destroy_domains(const struct cpumask *cpu_map)
6564 6565 6566
{
	int i;

6567
	rcu_read_lock();
6568
	for_each_cpu(i, cpu_map)
G
Gregory Haskins 已提交
6569
		cpu_attach_domain(NULL, &def_root_domain, i);
6570
	rcu_read_unlock();
6571 6572
}

6573 6574 6575 6576 6577 6578 6579 6580 6581 6582 6583 6584 6585 6586 6587 6588
/* 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
6589 6590
/*
 * Partition sched domains as specified by the 'ndoms_new'
I
Ingo Molnar 已提交
6591
 * cpumasks in the array doms_new[] of cpumasks. This compares
P
Paul Jackson 已提交
6592 6593 6594
 * doms_new[] to the current sched domain partitioning, doms_cur[].
 * It destroys each deleted domain and builds each new domain.
 *
6595
 * 'doms_new' is an array of cpumask_var_t's of length 'ndoms_new'.
I
Ingo Molnar 已提交
6596 6597 6598
 * 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
6599 6600 6601
 * current 'doms_cur' domains and in the new 'doms_new', we can leave
 * it as it is.
 *
6602 6603 6604 6605 6606 6607
 * 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
6608
 *
6609
 * If doms_new == NULL it will be replaced with cpu_online_mask.
6610 6611
 * ndoms_new == 0 is a special case for destroying existing domains,
 * and it will not create the default domain.
6612
 *
P
Paul Jackson 已提交
6613 6614
 * Call with hotplug lock held
 */
6615
void partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
6616
			     struct sched_domain_attr *dattr_new)
P
Paul Jackson 已提交
6617
{
6618
	int i, j, n;
6619
	int new_topology;
P
Paul Jackson 已提交
6620

6621
	mutex_lock(&sched_domains_mutex);
6622

6623 6624 6625
	/* always unregister in case we don't destroy any domains */
	unregister_sched_domain_sysctl();

6626 6627 6628
	/* Let architecture update cpu core mappings. */
	new_topology = arch_update_cpu_topology();

6629
	n = doms_new ? ndoms_new : 0;
P
Paul Jackson 已提交
6630 6631 6632

	/* Destroy deleted domains */
	for (i = 0; i < ndoms_cur; i++) {
6633
		for (j = 0; j < n && !new_topology; j++) {
6634
			if (cpumask_equal(doms_cur[i], doms_new[j])
6635
			    && dattrs_equal(dattr_cur, i, dattr_new, j))
P
Paul Jackson 已提交
6636 6637 6638
				goto match1;
		}
		/* no match - a current sched domain not in new doms_new[] */
6639
		detach_destroy_domains(doms_cur[i]);
P
Paul Jackson 已提交
6640 6641 6642 6643
match1:
		;
	}

6644 6645
	if (doms_new == NULL) {
		ndoms_cur = 0;
6646
		doms_new = &fallback_doms;
6647
		cpumask_andnot(doms_new[0], cpu_active_mask, cpu_isolated_map);
6648
		WARN_ON_ONCE(dattr_new);
6649 6650
	}

P
Paul Jackson 已提交
6651 6652
	/* Build new domains */
	for (i = 0; i < ndoms_new; i++) {
6653
		for (j = 0; j < ndoms_cur && !new_topology; j++) {
6654
			if (cpumask_equal(doms_new[i], doms_cur[j])
6655
			    && dattrs_equal(dattr_new, i, dattr_cur, j))
P
Paul Jackson 已提交
6656 6657 6658
				goto match2;
		}
		/* no match - add a new doms_new */
6659
		build_sched_domains(doms_new[i], dattr_new ? dattr_new + i : NULL);
P
Paul Jackson 已提交
6660 6661 6662 6663 6664
match2:
		;
	}

	/* Remember the new sched domains */
6665 6666
	if (doms_cur != &fallback_doms)
		free_sched_domains(doms_cur, ndoms_cur);
6667
	kfree(dattr_cur);	/* kfree(NULL) is safe */
P
Paul Jackson 已提交
6668
	doms_cur = doms_new;
6669
	dattr_cur = dattr_new;
P
Paul Jackson 已提交
6670
	ndoms_cur = ndoms_new;
6671 6672

	register_sched_domain_sysctl();
6673

6674
	mutex_unlock(&sched_domains_mutex);
P
Paul Jackson 已提交
6675 6676
}

6677
#if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT)
6678
static void reinit_sched_domains(void)
6679
{
6680
	get_online_cpus();
6681 6682 6683 6684

	/* Destroy domains first to force the rebuild */
	partition_sched_domains(0, NULL, NULL);

6685
	rebuild_sched_domains();
6686
	put_online_cpus();
6687 6688 6689 6690
}

static ssize_t sched_power_savings_store(const char *buf, size_t count, int smt)
{
6691
	unsigned int level = 0;
6692

6693 6694 6695 6696 6697 6698 6699 6700 6701 6702 6703
	if (sscanf(buf, "%u", &level) != 1)
		return -EINVAL;

	/*
	 * level is always be positive so don't check for
	 * level < POWERSAVINGS_BALANCE_NONE which is 0
	 * What happens on 0 or 1 byte write,
	 * need to check for count as well?
	 */

	if (level >= MAX_POWERSAVINGS_BALANCE_LEVELS)
6704 6705 6706
		return -EINVAL;

	if (smt)
6707
		sched_smt_power_savings = level;
6708
	else
6709
		sched_mc_power_savings = level;
6710

6711
	reinit_sched_domains();
6712

6713
	return count;
6714 6715 6716
}

#ifdef CONFIG_SCHED_MC
6717 6718 6719
static ssize_t sched_mc_power_savings_show(struct device *dev,
					   struct device_attribute *attr,
					   char *buf)
6720
{
6721
	return sprintf(buf, "%u\n", sched_mc_power_savings);
6722
}
6723 6724
static ssize_t sched_mc_power_savings_store(struct device *dev,
					    struct device_attribute *attr,
6725
					    const char *buf, size_t count)
6726 6727 6728
{
	return sched_power_savings_store(buf, count, 0);
}
6729 6730 6731
static DEVICE_ATTR(sched_mc_power_savings, 0644,
		   sched_mc_power_savings_show,
		   sched_mc_power_savings_store);
6732 6733 6734
#endif

#ifdef CONFIG_SCHED_SMT
6735 6736 6737
static ssize_t sched_smt_power_savings_show(struct device *dev,
					    struct device_attribute *attr,
					    char *buf)
6738
{
6739
	return sprintf(buf, "%u\n", sched_smt_power_savings);
6740
}
6741 6742
static ssize_t sched_smt_power_savings_store(struct device *dev,
					    struct device_attribute *attr,
6743
					     const char *buf, size_t count)
6744 6745 6746
{
	return sched_power_savings_store(buf, count, 1);
}
6747
static DEVICE_ATTR(sched_smt_power_savings, 0644,
6748
		   sched_smt_power_savings_show,
A
Adrian Bunk 已提交
6749 6750 6751
		   sched_smt_power_savings_store);
#endif

6752
int __init sched_create_sysfs_power_savings_entries(struct device *dev)
A
Adrian Bunk 已提交
6753 6754 6755 6756 6757
{
	int err = 0;

#ifdef CONFIG_SCHED_SMT
	if (smt_capable())
6758
		err = device_create_file(dev, &dev_attr_sched_smt_power_savings);
A
Adrian Bunk 已提交
6759 6760 6761
#endif
#ifdef CONFIG_SCHED_MC
	if (!err && mc_capable())
6762
		err = device_create_file(dev, &dev_attr_sched_mc_power_savings);
A
Adrian Bunk 已提交
6763 6764 6765
#endif
	return err;
}
6766
#endif /* CONFIG_SCHED_MC || CONFIG_SCHED_SMT */
6767

L
Linus Torvalds 已提交
6768
/*
6769 6770 6771
 * Update cpusets according to cpu_active mask.  If cpusets are
 * disabled, cpuset_update_active_cpus() becomes a simple wrapper
 * around partition_sched_domains().
L
Linus Torvalds 已提交
6772
 */
6773 6774
static int cpuset_cpu_active(struct notifier_block *nfb, unsigned long action,
			     void *hcpu)
6775
{
6776
	switch (action & ~CPU_TASKS_FROZEN) {
6777
	case CPU_ONLINE:
6778
	case CPU_DOWN_FAILED:
6779
		cpuset_update_active_cpus();
6780
		return NOTIFY_OK;
6781 6782 6783 6784
	default:
		return NOTIFY_DONE;
	}
}
6785

6786 6787
static int cpuset_cpu_inactive(struct notifier_block *nfb, unsigned long action,
			       void *hcpu)
6788 6789 6790 6791 6792
{
	switch (action & ~CPU_TASKS_FROZEN) {
	case CPU_DOWN_PREPARE:
		cpuset_update_active_cpus();
		return NOTIFY_OK;
6793 6794 6795 6796 6797
	default:
		return NOTIFY_DONE;
	}
}

L
Linus Torvalds 已提交
6798 6799
void __init sched_init_smp(void)
{
6800 6801 6802
	cpumask_var_t non_isolated_cpus;

	alloc_cpumask_var(&non_isolated_cpus, GFP_KERNEL);
6803
	alloc_cpumask_var(&fallback_doms, GFP_KERNEL);
6804

6805
	get_online_cpus();
6806
	mutex_lock(&sched_domains_mutex);
6807
	init_sched_domains(cpu_active_mask);
6808 6809 6810
	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);
6811
	mutex_unlock(&sched_domains_mutex);
6812
	put_online_cpus();
6813

6814 6815
	hotcpu_notifier(cpuset_cpu_active, CPU_PRI_CPUSET_ACTIVE);
	hotcpu_notifier(cpuset_cpu_inactive, CPU_PRI_CPUSET_INACTIVE);
6816 6817 6818 6819

	/* RT runtime code needs to handle some hotplug events */
	hotcpu_notifier(update_runtime, 0);

6820
	init_hrtick();
6821 6822

	/* Move init over to a non-isolated CPU */
6823
	if (set_cpus_allowed_ptr(current, non_isolated_cpus) < 0)
6824
		BUG();
I
Ingo Molnar 已提交
6825
	sched_init_granularity();
6826
	free_cpumask_var(non_isolated_cpus);
6827

6828
	init_sched_rt_class();
L
Linus Torvalds 已提交
6829 6830 6831 6832
}
#else
void __init sched_init_smp(void)
{
I
Ingo Molnar 已提交
6833
	sched_init_granularity();
L
Linus Torvalds 已提交
6834 6835 6836
}
#endif /* CONFIG_SMP */

6837 6838
const_debug unsigned int sysctl_timer_migration = 1;

L
Linus Torvalds 已提交
6839 6840 6841 6842 6843 6844 6845
int in_sched_functions(unsigned long addr)
{
	return in_lock_functions(addr) ||
		(addr >= (unsigned long)__sched_text_start
		&& addr < (unsigned long)__sched_text_end);
}

6846 6847
#ifdef CONFIG_CGROUP_SCHED
struct task_group root_task_group;
6848
#endif
P
Peter Zijlstra 已提交
6849

6850
DECLARE_PER_CPU(cpumask_var_t, load_balance_tmpmask);
P
Peter Zijlstra 已提交
6851

L
Linus Torvalds 已提交
6852 6853
void __init sched_init(void)
{
I
Ingo Molnar 已提交
6854
	int i, j;
6855 6856 6857 6858 6859 6860 6861
	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 **);
6862
#endif
6863
#ifdef CONFIG_CPUMASK_OFFSTACK
6864
	alloc_size += num_possible_cpus() * cpumask_size();
6865 6866
#endif
	if (alloc_size) {
6867
		ptr = (unsigned long)kzalloc(alloc_size, GFP_NOWAIT);
6868 6869

#ifdef CONFIG_FAIR_GROUP_SCHED
6870
		root_task_group.se = (struct sched_entity **)ptr;
6871 6872
		ptr += nr_cpu_ids * sizeof(void **);

6873
		root_task_group.cfs_rq = (struct cfs_rq **)ptr;
6874
		ptr += nr_cpu_ids * sizeof(void **);
6875

6876
#endif /* CONFIG_FAIR_GROUP_SCHED */
6877
#ifdef CONFIG_RT_GROUP_SCHED
6878
		root_task_group.rt_se = (struct sched_rt_entity **)ptr;
6879 6880
		ptr += nr_cpu_ids * sizeof(void **);

6881
		root_task_group.rt_rq = (struct rt_rq **)ptr;
6882 6883
		ptr += nr_cpu_ids * sizeof(void **);

6884
#endif /* CONFIG_RT_GROUP_SCHED */
6885 6886 6887 6888 6889 6890
#ifdef CONFIG_CPUMASK_OFFSTACK
		for_each_possible_cpu(i) {
			per_cpu(load_balance_tmpmask, i) = (void *)ptr;
			ptr += cpumask_size();
		}
#endif /* CONFIG_CPUMASK_OFFSTACK */
6891
	}
I
Ingo Molnar 已提交
6892

G
Gregory Haskins 已提交
6893 6894 6895 6896
#ifdef CONFIG_SMP
	init_defrootdomain();
#endif

6897 6898 6899 6900
	init_rt_bandwidth(&def_rt_bandwidth,
			global_rt_period(), global_rt_runtime());

#ifdef CONFIG_RT_GROUP_SCHED
6901
	init_rt_bandwidth(&root_task_group.rt_bandwidth,
6902
			global_rt_period(), global_rt_runtime());
6903
#endif /* CONFIG_RT_GROUP_SCHED */
6904

D
Dhaval Giani 已提交
6905
#ifdef CONFIG_CGROUP_SCHED
6906 6907
	list_add(&root_task_group.list, &task_groups);
	INIT_LIST_HEAD(&root_task_group.children);
6908
	INIT_LIST_HEAD(&root_task_group.siblings);
6909
	autogroup_init(&init_task);
6910

D
Dhaval Giani 已提交
6911
#endif /* CONFIG_CGROUP_SCHED */
P
Peter Zijlstra 已提交
6912

6913 6914 6915 6916 6917 6918
#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
6919
	for_each_possible_cpu(i) {
6920
		struct rq *rq;
L
Linus Torvalds 已提交
6921 6922

		rq = cpu_rq(i);
6923
		raw_spin_lock_init(&rq->lock);
N
Nick Piggin 已提交
6924
		rq->nr_running = 0;
6925 6926
		rq->calc_load_active = 0;
		rq->calc_load_update = jiffies + LOAD_FREQ;
6927
		init_cfs_rq(&rq->cfs);
P
Peter Zijlstra 已提交
6928
		init_rt_rq(&rq->rt, rq);
I
Ingo Molnar 已提交
6929
#ifdef CONFIG_FAIR_GROUP_SCHED
6930
		root_task_group.shares = ROOT_TASK_GROUP_LOAD;
P
Peter Zijlstra 已提交
6931
		INIT_LIST_HEAD(&rq->leaf_cfs_rq_list);
D
Dhaval Giani 已提交
6932
		/*
6933
		 * How much cpu bandwidth does root_task_group get?
D
Dhaval Giani 已提交
6934 6935 6936 6937
		 *
		 * 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
6938
		 * root_task_group and its child task-groups in a fair manner,
D
Dhaval Giani 已提交
6939 6940 6941
		 * based on each entity's (task or task-group's) weight
		 * (se->load.weight).
		 *
6942
		 * In other words, if root_task_group has 10 tasks of weight
D
Dhaval Giani 已提交
6943 6944 6945
		 * 1024) and two child groups A0 and A1 (of weight 1024 each),
		 * then A0's share of the cpu resource is:
		 *
6946
		 *	A0's bandwidth = 1024 / (10*1024 + 1024 + 1024) = 8.33%
D
Dhaval Giani 已提交
6947
		 *
6948 6949
		 * 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
6950
		 */
6951
		init_cfs_bandwidth(&root_task_group.cfs_bandwidth);
6952
		init_tg_cfs_entry(&root_task_group, &rq->cfs, NULL, i, NULL);
D
Dhaval Giani 已提交
6953 6954 6955
#endif /* CONFIG_FAIR_GROUP_SCHED */

		rq->rt.rt_runtime = def_rt_bandwidth.rt_runtime;
6956
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
6957
		INIT_LIST_HEAD(&rq->leaf_rt_rq_list);
6958
		init_tg_rt_entry(&root_task_group, &rq->rt, NULL, i, NULL);
I
Ingo Molnar 已提交
6959
#endif
L
Linus Torvalds 已提交
6960

I
Ingo Molnar 已提交
6961 6962
		for (j = 0; j < CPU_LOAD_IDX_MAX; j++)
			rq->cpu_load[j] = 0;
6963 6964 6965

		rq->last_load_update_tick = jiffies;

L
Linus Torvalds 已提交
6966
#ifdef CONFIG_SMP
N
Nick Piggin 已提交
6967
		rq->sd = NULL;
G
Gregory Haskins 已提交
6968
		rq->rd = NULL;
6969
		rq->cpu_power = SCHED_POWER_SCALE;
6970
		rq->post_schedule = 0;
L
Linus Torvalds 已提交
6971
		rq->active_balance = 0;
I
Ingo Molnar 已提交
6972
		rq->next_balance = jiffies;
L
Linus Torvalds 已提交
6973
		rq->push_cpu = 0;
6974
		rq->cpu = i;
6975
		rq->online = 0;
6976 6977
		rq->idle_stamp = 0;
		rq->avg_idle = 2*sysctl_sched_migration_cost;
6978 6979 6980

		INIT_LIST_HEAD(&rq->cfs_tasks);

6981
		rq_attach_root(rq, &def_root_domain);
6982
#ifdef CONFIG_NO_HZ
6983
		rq->nohz_flags = 0;
6984
#endif
L
Linus Torvalds 已提交
6985
#endif
P
Peter Zijlstra 已提交
6986
		init_rq_hrtick(rq);
L
Linus Torvalds 已提交
6987 6988 6989
		atomic_set(&rq->nr_iowait, 0);
	}

6990
	set_load_weight(&init_task);
6991

6992 6993 6994 6995
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&init_task.preempt_notifiers);
#endif

6996
#ifdef CONFIG_RT_MUTEXES
6997
	plist_head_init(&init_task.pi_waiters);
6998 6999
#endif

L
Linus Torvalds 已提交
7000 7001 7002 7003 7004 7005 7006 7007 7008 7009 7010 7011 7012
	/*
	 * 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());
7013 7014 7015

	calc_load_update = jiffies + LOAD_FREQ;

I
Ingo Molnar 已提交
7016 7017 7018 7019
	/*
	 * During early bootup we pretend to be a normal task:
	 */
	current->sched_class = &fair_sched_class;
7020

7021
#ifdef CONFIG_SMP
7022
	zalloc_cpumask_var(&sched_domains_tmpmask, GFP_NOWAIT);
R
Rusty Russell 已提交
7023 7024 7025
	/* May be allocated at isolcpus cmdline parse time */
	if (cpu_isolated_map == NULL)
		zalloc_cpumask_var(&cpu_isolated_map, GFP_NOWAIT);
7026 7027
#endif
	init_sched_fair_class();
7028

7029
	scheduler_running = 1;
L
Linus Torvalds 已提交
7030 7031
}

7032
#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
7033 7034
static inline int preempt_count_equals(int preempt_offset)
{
7035
	int nested = (preempt_count() & ~PREEMPT_ACTIVE) + rcu_preempt_depth();
7036

A
Arnd Bergmann 已提交
7037
	return (nested == preempt_offset);
7038 7039
}

7040
void __might_sleep(const char *file, int line, int preempt_offset)
L
Linus Torvalds 已提交
7041 7042 7043
{
	static unsigned long prev_jiffy;	/* ratelimiting */

7044
	rcu_sleep_check(); /* WARN_ON_ONCE() by default, no rate limit reqd. */
7045 7046
	if ((preempt_count_equals(preempt_offset) && !irqs_disabled()) ||
	    system_state != SYSTEM_RUNNING || oops_in_progress)
I
Ingo Molnar 已提交
7047 7048 7049 7050 7051
		return;
	if (time_before(jiffies, prev_jiffy + HZ) && prev_jiffy)
		return;
	prev_jiffy = jiffies;

P
Peter Zijlstra 已提交
7052 7053 7054 7055 7056 7057 7058
	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
7059 7060 7061 7062 7063

	debug_show_held_locks(current);
	if (irqs_disabled())
		print_irqtrace_events(current);
	dump_stack();
L
Linus Torvalds 已提交
7064 7065 7066 7067 7068
}
EXPORT_SYMBOL(__might_sleep);
#endif

#ifdef CONFIG_MAGIC_SYSRQ
7069 7070
static void normalize_task(struct rq *rq, struct task_struct *p)
{
P
Peter Zijlstra 已提交
7071 7072
	const struct sched_class *prev_class = p->sched_class;
	int old_prio = p->prio;
7073
	int on_rq;
7074

P
Peter Zijlstra 已提交
7075
	on_rq = p->on_rq;
7076
	if (on_rq)
7077
		dequeue_task(rq, p, 0);
7078 7079
	__setscheduler(rq, p, SCHED_NORMAL, 0);
	if (on_rq) {
7080
		enqueue_task(rq, p, 0);
7081 7082
		resched_task(rq->curr);
	}
P
Peter Zijlstra 已提交
7083 7084

	check_class_changed(rq, p, prev_class, old_prio);
7085 7086
}

L
Linus Torvalds 已提交
7087 7088
void normalize_rt_tasks(void)
{
7089
	struct task_struct *g, *p;
L
Linus Torvalds 已提交
7090
	unsigned long flags;
7091
	struct rq *rq;
L
Linus Torvalds 已提交
7092

7093
	read_lock_irqsave(&tasklist_lock, flags);
7094
	do_each_thread(g, p) {
7095 7096 7097 7098 7099 7100
		/*
		 * Only normalize user tasks:
		 */
		if (!p->mm)
			continue;

I
Ingo Molnar 已提交
7101 7102
		p->se.exec_start		= 0;
#ifdef CONFIG_SCHEDSTATS
7103 7104 7105
		p->se.statistics.wait_start	= 0;
		p->se.statistics.sleep_start	= 0;
		p->se.statistics.block_start	= 0;
I
Ingo Molnar 已提交
7106
#endif
I
Ingo Molnar 已提交
7107 7108 7109 7110 7111 7112 7113 7114

		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
7115
			continue;
I
Ingo Molnar 已提交
7116
		}
L
Linus Torvalds 已提交
7117

7118
		raw_spin_lock(&p->pi_lock);
7119
		rq = __task_rq_lock(p);
L
Linus Torvalds 已提交
7120

7121
		normalize_task(rq, p);
7122

7123
		__task_rq_unlock(rq);
7124
		raw_spin_unlock(&p->pi_lock);
7125 7126
	} while_each_thread(g, p);

7127
	read_unlock_irqrestore(&tasklist_lock, flags);
L
Linus Torvalds 已提交
7128 7129 7130
}

#endif /* CONFIG_MAGIC_SYSRQ */
7131

7132
#if defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB)
7133
/*
7134
 * These functions are only useful for the IA64 MCA handling, or kdb.
7135 7136 7137 7138 7139 7140 7141 7142 7143 7144 7145 7146 7147 7148
 *
 * 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!
 */
7149
struct task_struct *curr_task(int cpu)
7150 7151 7152 7153
{
	return cpu_curr(cpu);
}

7154 7155 7156
#endif /* defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB) */

#ifdef CONFIG_IA64
7157 7158 7159 7160 7161 7162
/**
 * 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
7163 7164
 * 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
7165 7166 7167 7168 7169 7170 7171
 * 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!
 */
7172
void set_curr_task(int cpu, struct task_struct *p)
7173 7174 7175 7176 7177
{
	cpu_curr(cpu) = p;
}

#endif
S
Srivatsa Vaddagiri 已提交
7178

D
Dhaval Giani 已提交
7179
#ifdef CONFIG_CGROUP_SCHED
7180 7181 7182
/* task_group_lock serializes the addition/removal of task groups */
static DEFINE_SPINLOCK(task_group_lock);

7183 7184 7185 7186
static void free_sched_group(struct task_group *tg)
{
	free_fair_sched_group(tg);
	free_rt_sched_group(tg);
7187
	autogroup_free(tg);
7188 7189 7190 7191
	kfree(tg);
}

/* allocate runqueue etc for a new task group */
7192
struct task_group *sched_create_group(struct task_group *parent)
7193 7194 7195 7196 7197 7198 7199 7200
{
	struct task_group *tg;
	unsigned long flags;

	tg = kzalloc(sizeof(*tg), GFP_KERNEL);
	if (!tg)
		return ERR_PTR(-ENOMEM);

7201
	if (!alloc_fair_sched_group(tg, parent))
7202 7203
		goto err;

7204
	if (!alloc_rt_sched_group(tg, parent))
7205 7206
		goto err;

7207
	spin_lock_irqsave(&task_group_lock, flags);
P
Peter Zijlstra 已提交
7208
	list_add_rcu(&tg->list, &task_groups);
P
Peter Zijlstra 已提交
7209 7210 7211 7212 7213

	WARN_ON(!parent); /* root should already exist */

	tg->parent = parent;
	INIT_LIST_HEAD(&tg->children);
7214
	list_add_rcu(&tg->siblings, &parent->children);
7215
	spin_unlock_irqrestore(&task_group_lock, flags);
S
Srivatsa Vaddagiri 已提交
7216

7217
	return tg;
S
Srivatsa Vaddagiri 已提交
7218 7219

err:
P
Peter Zijlstra 已提交
7220
	free_sched_group(tg);
S
Srivatsa Vaddagiri 已提交
7221 7222 7223
	return ERR_PTR(-ENOMEM);
}

7224
/* rcu callback to free various structures associated with a task group */
P
Peter Zijlstra 已提交
7225
static void free_sched_group_rcu(struct rcu_head *rhp)
S
Srivatsa Vaddagiri 已提交
7226 7227
{
	/* now it should be safe to free those cfs_rqs */
P
Peter Zijlstra 已提交
7228
	free_sched_group(container_of(rhp, struct task_group, rcu));
S
Srivatsa Vaddagiri 已提交
7229 7230
}

7231
/* Destroy runqueue etc associated with a task group */
7232
void sched_destroy_group(struct task_group *tg)
S
Srivatsa Vaddagiri 已提交
7233
{
7234
	unsigned long flags;
7235
	int i;
S
Srivatsa Vaddagiri 已提交
7236

7237 7238
	/* end participation in shares distribution */
	for_each_possible_cpu(i)
7239
		unregister_fair_sched_group(tg, i);
7240 7241

	spin_lock_irqsave(&task_group_lock, flags);
P
Peter Zijlstra 已提交
7242
	list_del_rcu(&tg->list);
P
Peter Zijlstra 已提交
7243
	list_del_rcu(&tg->siblings);
7244
	spin_unlock_irqrestore(&task_group_lock, flags);
7245 7246

	/* wait for possible concurrent references to cfs_rqs complete */
P
Peter Zijlstra 已提交
7247
	call_rcu(&tg->rcu, free_sched_group_rcu);
S
Srivatsa Vaddagiri 已提交
7248 7249
}

7250
/* change task's runqueue when it moves between groups.
I
Ingo Molnar 已提交
7251 7252 7253
 *	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.
7254 7255
 */
void sched_move_task(struct task_struct *tsk)
S
Srivatsa Vaddagiri 已提交
7256 7257 7258 7259 7260 7261 7262
{
	int on_rq, running;
	unsigned long flags;
	struct rq *rq;

	rq = task_rq_lock(tsk, &flags);

7263
	running = task_current(rq, tsk);
P
Peter Zijlstra 已提交
7264
	on_rq = tsk->on_rq;
S
Srivatsa Vaddagiri 已提交
7265

7266
	if (on_rq)
S
Srivatsa Vaddagiri 已提交
7267
		dequeue_task(rq, tsk, 0);
7268 7269
	if (unlikely(running))
		tsk->sched_class->put_prev_task(rq, tsk);
S
Srivatsa Vaddagiri 已提交
7270

P
Peter Zijlstra 已提交
7271
#ifdef CONFIG_FAIR_GROUP_SCHED
7272 7273 7274
	if (tsk->sched_class->task_move_group)
		tsk->sched_class->task_move_group(tsk, on_rq);
	else
P
Peter Zijlstra 已提交
7275
#endif
7276
		set_task_rq(tsk, task_cpu(tsk));
P
Peter Zijlstra 已提交
7277

7278 7279 7280
	if (unlikely(running))
		tsk->sched_class->set_curr_task(rq);
	if (on_rq)
7281
		enqueue_task(rq, tsk, 0);
S
Srivatsa Vaddagiri 已提交
7282

7283
	task_rq_unlock(rq, tsk, &flags);
S
Srivatsa Vaddagiri 已提交
7284
}
D
Dhaval Giani 已提交
7285
#endif /* CONFIG_CGROUP_SCHED */
S
Srivatsa Vaddagiri 已提交
7286

7287
#if defined(CONFIG_RT_GROUP_SCHED) || defined(CONFIG_CFS_BANDWIDTH)
P
Peter Zijlstra 已提交
7288 7289 7290
static unsigned long to_ratio(u64 period, u64 runtime)
{
	if (runtime == RUNTIME_INF)
P
Peter Zijlstra 已提交
7291
		return 1ULL << 20;
P
Peter Zijlstra 已提交
7292

P
Peter Zijlstra 已提交
7293
	return div64_u64(runtime << 20, period);
P
Peter Zijlstra 已提交
7294
}
7295 7296 7297 7298 7299 7300 7301
#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
7302

P
Peter Zijlstra 已提交
7303 7304
/* Must be called with tasklist_lock held */
static inline int tg_has_rt_tasks(struct task_group *tg)
7305
{
P
Peter Zijlstra 已提交
7306
	struct task_struct *g, *p;
7307

P
Peter Zijlstra 已提交
7308
	do_each_thread(g, p) {
7309
		if (rt_task(p) && task_rq(p)->rt.tg == tg)
P
Peter Zijlstra 已提交
7310 7311
			return 1;
	} while_each_thread(g, p);
7312

P
Peter Zijlstra 已提交
7313 7314
	return 0;
}
7315

P
Peter Zijlstra 已提交
7316 7317 7318 7319 7320
struct rt_schedulable_data {
	struct task_group *tg;
	u64 rt_period;
	u64 rt_runtime;
};
7321

7322
static int tg_rt_schedulable(struct task_group *tg, void *data)
P
Peter Zijlstra 已提交
7323 7324 7325 7326 7327
{
	struct rt_schedulable_data *d = data;
	struct task_group *child;
	unsigned long total, sum = 0;
	u64 period, runtime;
7328

P
Peter Zijlstra 已提交
7329 7330
	period = ktime_to_ns(tg->rt_bandwidth.rt_period);
	runtime = tg->rt_bandwidth.rt_runtime;
7331

P
Peter Zijlstra 已提交
7332 7333 7334
	if (tg == d->tg) {
		period = d->rt_period;
		runtime = d->rt_runtime;
7335 7336
	}

7337 7338 7339 7340 7341
	/*
	 * Cannot have more runtime than the period.
	 */
	if (runtime > period && runtime != RUNTIME_INF)
		return -EINVAL;
P
Peter Zijlstra 已提交
7342

7343 7344 7345
	/*
	 * Ensure we don't starve existing RT tasks.
	 */
P
Peter Zijlstra 已提交
7346 7347
	if (rt_bandwidth_enabled() && !runtime && tg_has_rt_tasks(tg))
		return -EBUSY;
P
Peter Zijlstra 已提交
7348

P
Peter Zijlstra 已提交
7349
	total = to_ratio(period, runtime);
P
Peter Zijlstra 已提交
7350

7351 7352 7353 7354 7355
	/*
	 * 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
7356

7357 7358 7359
	/*
	 * The sum of our children's runtime should not exceed our own.
	 */
P
Peter Zijlstra 已提交
7360 7361 7362
	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
7363

P
Peter Zijlstra 已提交
7364 7365 7366 7367
		if (child == d->tg) {
			period = d->rt_period;
			runtime = d->rt_runtime;
		}
P
Peter Zijlstra 已提交
7368

P
Peter Zijlstra 已提交
7369
		sum += to_ratio(period, runtime);
P
Peter Zijlstra 已提交
7370
	}
P
Peter Zijlstra 已提交
7371

P
Peter Zijlstra 已提交
7372 7373 7374 7375
	if (sum > total)
		return -EINVAL;

	return 0;
P
Peter Zijlstra 已提交
7376 7377
}

P
Peter Zijlstra 已提交
7378
static int __rt_schedulable(struct task_group *tg, u64 period, u64 runtime)
7379
{
7380 7381
	int ret;

P
Peter Zijlstra 已提交
7382 7383 7384 7385 7386 7387
	struct rt_schedulable_data data = {
		.tg = tg,
		.rt_period = period,
		.rt_runtime = runtime,
	};

7388 7389 7390 7391 7392
	rcu_read_lock();
	ret = walk_tg_tree(tg_rt_schedulable, tg_nop, &data);
	rcu_read_unlock();

	return ret;
7393 7394
}

7395
static int tg_set_rt_bandwidth(struct task_group *tg,
7396
		u64 rt_period, u64 rt_runtime)
P
Peter Zijlstra 已提交
7397
{
P
Peter Zijlstra 已提交
7398
	int i, err = 0;
P
Peter Zijlstra 已提交
7399 7400

	mutex_lock(&rt_constraints_mutex);
7401
	read_lock(&tasklist_lock);
P
Peter Zijlstra 已提交
7402 7403
	err = __rt_schedulable(tg, rt_period, rt_runtime);
	if (err)
P
Peter Zijlstra 已提交
7404
		goto unlock;
P
Peter Zijlstra 已提交
7405

7406
	raw_spin_lock_irq(&tg->rt_bandwidth.rt_runtime_lock);
7407 7408
	tg->rt_bandwidth.rt_period = ns_to_ktime(rt_period);
	tg->rt_bandwidth.rt_runtime = rt_runtime;
P
Peter Zijlstra 已提交
7409 7410 7411 7412

	for_each_possible_cpu(i) {
		struct rt_rq *rt_rq = tg->rt_rq[i];

7413
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
7414
		rt_rq->rt_runtime = rt_runtime;
7415
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
7416
	}
7417
	raw_spin_unlock_irq(&tg->rt_bandwidth.rt_runtime_lock);
P
Peter Zijlstra 已提交
7418
unlock:
7419
	read_unlock(&tasklist_lock);
P
Peter Zijlstra 已提交
7420 7421 7422
	mutex_unlock(&rt_constraints_mutex);

	return err;
P
Peter Zijlstra 已提交
7423 7424
}

7425 7426 7427 7428 7429 7430 7431 7432 7433
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;

7434
	return tg_set_rt_bandwidth(tg, rt_period, rt_runtime);
7435 7436
}

P
Peter Zijlstra 已提交
7437 7438 7439 7440
long sched_group_rt_runtime(struct task_group *tg)
{
	u64 rt_runtime_us;

7441
	if (tg->rt_bandwidth.rt_runtime == RUNTIME_INF)
P
Peter Zijlstra 已提交
7442 7443
		return -1;

7444
	rt_runtime_us = tg->rt_bandwidth.rt_runtime;
P
Peter Zijlstra 已提交
7445 7446 7447
	do_div(rt_runtime_us, NSEC_PER_USEC);
	return rt_runtime_us;
}
7448 7449 7450 7451 7452 7453 7454 7455

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;

7456 7457 7458
	if (rt_period == 0)
		return -EINVAL;

7459
	return tg_set_rt_bandwidth(tg, rt_period, rt_runtime);
7460 7461 7462 7463 7464 7465 7466 7467 7468 7469 7470 7471 7472
}

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)
{
7473
	u64 runtime, period;
7474 7475
	int ret = 0;

7476 7477 7478
	if (sysctl_sched_rt_period <= 0)
		return -EINVAL;

7479 7480 7481 7482 7483 7484 7485 7486
	runtime = global_rt_runtime();
	period = global_rt_period();

	/*
	 * Sanity check on the sysctl variables.
	 */
	if (runtime > period && runtime != RUNTIME_INF)
		return -EINVAL;
7487

7488
	mutex_lock(&rt_constraints_mutex);
P
Peter Zijlstra 已提交
7489
	read_lock(&tasklist_lock);
7490
	ret = __rt_schedulable(NULL, 0, 0);
P
Peter Zijlstra 已提交
7491
	read_unlock(&tasklist_lock);
7492 7493 7494 7495
	mutex_unlock(&rt_constraints_mutex);

	return ret;
}
7496 7497 7498 7499 7500 7501 7502 7503 7504 7505

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;
}

7506
#else /* !CONFIG_RT_GROUP_SCHED */
7507 7508
static int sched_rt_global_constraints(void)
{
P
Peter Zijlstra 已提交
7509 7510 7511
	unsigned long flags;
	int i;

7512 7513 7514
	if (sysctl_sched_rt_period <= 0)
		return -EINVAL;

7515 7516 7517 7518 7519 7520 7521
	/*
	 * There's always some RT tasks in the root group
	 * -- migration, kstopmachine etc..
	 */
	if (sysctl_sched_rt_runtime == 0)
		return -EBUSY;

7522
	raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags);
P
Peter Zijlstra 已提交
7523 7524 7525
	for_each_possible_cpu(i) {
		struct rt_rq *rt_rq = &cpu_rq(i)->rt;

7526
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
7527
		rt_rq->rt_runtime = global_rt_runtime();
7528
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
7529
	}
7530
	raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags);
P
Peter Zijlstra 已提交
7531

7532 7533
	return 0;
}
7534
#endif /* CONFIG_RT_GROUP_SCHED */
7535 7536

int sched_rt_handler(struct ctl_table *table, int write,
7537
		void __user *buffer, size_t *lenp,
7538 7539 7540 7541 7542 7543 7544 7545 7546 7547
		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;

7548
	ret = proc_dointvec(table, write, buffer, lenp, ppos);
7549 7550 7551 7552 7553 7554 7555 7556 7557 7558 7559 7560 7561 7562 7563 7564

	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;
}
7565

7566
#ifdef CONFIG_CGROUP_SCHED
7567 7568

/* return corresponding task_group object of a cgroup */
7569
static inline struct task_group *cgroup_tg(struct cgroup *cgrp)
7570
{
7571 7572
	return container_of(cgroup_subsys_state(cgrp, cpu_cgroup_subsys_id),
			    struct task_group, css);
7573 7574 7575
}

static struct cgroup_subsys_state *
7576
cpu_cgroup_create(struct cgroup_subsys *ss, struct cgroup *cgrp)
7577
{
7578
	struct task_group *tg, *parent;
7579

7580
	if (!cgrp->parent) {
7581
		/* This is early initialization for the top cgroup */
7582
		return &root_task_group.css;
7583 7584
	}

7585 7586
	parent = cgroup_tg(cgrp->parent);
	tg = sched_create_group(parent);
7587 7588 7589 7590 7591 7592
	if (IS_ERR(tg))
		return ERR_PTR(-ENOMEM);

	return &tg->css;
}

I
Ingo Molnar 已提交
7593 7594
static void
cpu_cgroup_destroy(struct cgroup_subsys *ss, struct cgroup *cgrp)
7595
{
7596
	struct task_group *tg = cgroup_tg(cgrp);
7597 7598 7599 7600

	sched_destroy_group(tg);
}

7601 7602
static int cpu_cgroup_can_attach(struct cgroup_subsys *ss, struct cgroup *cgrp,
				 struct cgroup_taskset *tset)
7603
{
7604 7605 7606
	struct task_struct *task;

	cgroup_taskset_for_each(task, cgrp, tset) {
7607
#ifdef CONFIG_RT_GROUP_SCHED
7608 7609
		if (!sched_rt_can_attach(cgroup_tg(cgrp), task))
			return -EINVAL;
7610
#else
7611 7612 7613
		/* We don't support RT-tasks being in separate groups */
		if (task->sched_class != &fair_sched_class)
			return -EINVAL;
7614
#endif
7615
	}
7616 7617
	return 0;
}
7618

7619 7620
static void cpu_cgroup_attach(struct cgroup_subsys *ss, struct cgroup *cgrp,
			      struct cgroup_taskset *tset)
7621
{
7622 7623 7624 7625
	struct task_struct *task;

	cgroup_taskset_for_each(task, cgrp, tset)
		sched_move_task(task);
7626 7627
}

7628
static void
7629 7630
cpu_cgroup_exit(struct cgroup_subsys *ss, struct cgroup *cgrp,
		struct cgroup *old_cgrp, struct task_struct *task)
7631 7632 7633 7634 7635 7636 7637 7638 7639 7640 7641 7642
{
	/*
	 * 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);
}

7643
#ifdef CONFIG_FAIR_GROUP_SCHED
7644
static int cpu_shares_write_u64(struct cgroup *cgrp, struct cftype *cftype,
7645
				u64 shareval)
7646
{
7647
	return sched_group_set_shares(cgroup_tg(cgrp), scale_load(shareval));
7648 7649
}

7650
static u64 cpu_shares_read_u64(struct cgroup *cgrp, struct cftype *cft)
7651
{
7652
	struct task_group *tg = cgroup_tg(cgrp);
7653

7654
	return (u64) scale_load_down(tg->shares);
7655
}
7656 7657

#ifdef CONFIG_CFS_BANDWIDTH
7658 7659
static DEFINE_MUTEX(cfs_constraints_mutex);

7660 7661 7662
const u64 max_cfs_quota_period = 1 * NSEC_PER_SEC; /* 1s */
const u64 min_cfs_quota_period = 1 * NSEC_PER_MSEC; /* 1ms */

7663 7664
static int __cfs_schedulable(struct task_group *tg, u64 period, u64 runtime);

7665 7666
static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota)
{
7667
	int i, ret = 0, runtime_enabled, runtime_was_enabled;
7668
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
7669 7670 7671 7672 7673 7674 7675 7676 7677 7678 7679 7680 7681 7682 7683 7684 7685 7686 7687 7688

	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;

7689 7690 7691 7692 7693
	mutex_lock(&cfs_constraints_mutex);
	ret = __cfs_schedulable(tg, period, quota);
	if (ret)
		goto out_unlock;

7694
	runtime_enabled = quota != RUNTIME_INF;
7695 7696
	runtime_was_enabled = cfs_b->quota != RUNTIME_INF;
	account_cfs_bandwidth_used(runtime_enabled, runtime_was_enabled);
7697 7698 7699
	raw_spin_lock_irq(&cfs_b->lock);
	cfs_b->period = ns_to_ktime(period);
	cfs_b->quota = quota;
7700

P
Paul Turner 已提交
7701
	__refill_cfs_bandwidth_runtime(cfs_b);
7702 7703 7704 7705 7706 7707
	/* 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);
	}
7708 7709 7710 7711
	raw_spin_unlock_irq(&cfs_b->lock);

	for_each_possible_cpu(i) {
		struct cfs_rq *cfs_rq = tg->cfs_rq[i];
7712
		struct rq *rq = cfs_rq->rq;
7713 7714

		raw_spin_lock_irq(&rq->lock);
7715
		cfs_rq->runtime_enabled = runtime_enabled;
7716
		cfs_rq->runtime_remaining = 0;
7717

7718
		if (cfs_rq->throttled)
7719
			unthrottle_cfs_rq(cfs_rq);
7720 7721
		raw_spin_unlock_irq(&rq->lock);
	}
7722 7723
out_unlock:
	mutex_unlock(&cfs_constraints_mutex);
7724

7725
	return ret;
7726 7727 7728 7729 7730 7731
}

int tg_set_cfs_quota(struct task_group *tg, long cfs_quota_us)
{
	u64 quota, period;

7732
	period = ktime_to_ns(tg->cfs_bandwidth.period);
7733 7734 7735 7736 7737 7738 7739 7740 7741 7742 7743 7744
	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;

7745
	if (tg->cfs_bandwidth.quota == RUNTIME_INF)
7746 7747
		return -1;

7748
	quota_us = tg->cfs_bandwidth.quota;
7749 7750 7751 7752 7753 7754 7755 7756 7757 7758
	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;
7759
	quota = tg->cfs_bandwidth.quota;
7760 7761 7762 7763 7764 7765 7766 7767

	return tg_set_cfs_bandwidth(tg, period, quota);
}

long tg_get_cfs_period(struct task_group *tg)
{
	u64 cfs_period_us;

7768
	cfs_period_us = ktime_to_ns(tg->cfs_bandwidth.period);
7769 7770 7771 7772 7773 7774 7775 7776 7777 7778 7779 7780 7781 7782 7783 7784 7785 7786 7787 7788 7789 7790 7791 7792 7793 7794 7795
	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);
}

7796 7797 7798 7799 7800 7801 7802 7803 7804 7805 7806 7807 7808 7809 7810 7811 7812 7813 7814 7815 7816 7817 7818 7819 7820 7821 7822 7823 7824 7825 7826 7827
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;
7828
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
7829 7830 7831 7832 7833
	s64 quota = 0, parent_quota = -1;

	if (!tg->parent) {
		quota = RUNTIME_INF;
	} else {
7834
		struct cfs_bandwidth *parent_b = &tg->parent->cfs_bandwidth;
7835 7836 7837 7838 7839 7840 7841 7842 7843 7844 7845 7846 7847 7848 7849 7850 7851 7852 7853 7854

		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)
{
7855
	int ret;
7856 7857 7858 7859 7860 7861 7862 7863 7864 7865 7866
	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);
	}

7867 7868 7869 7870 7871
	rcu_read_lock();
	ret = walk_tg_tree(tg_cfs_schedulable_down, tg_nop, &data);
	rcu_read_unlock();

	return ret;
7872
}
7873 7874 7875 7876 7877

static int cpu_stats_show(struct cgroup *cgrp, struct cftype *cft,
		struct cgroup_map_cb *cb)
{
	struct task_group *tg = cgroup_tg(cgrp);
7878
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
7879 7880 7881 7882 7883 7884 7885

	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;
}
7886
#endif /* CONFIG_CFS_BANDWIDTH */
7887
#endif /* CONFIG_FAIR_GROUP_SCHED */
7888

7889
#ifdef CONFIG_RT_GROUP_SCHED
M
Mirco Tischler 已提交
7890
static int cpu_rt_runtime_write(struct cgroup *cgrp, struct cftype *cft,
7891
				s64 val)
P
Peter Zijlstra 已提交
7892
{
7893
	return sched_group_set_rt_runtime(cgroup_tg(cgrp), val);
P
Peter Zijlstra 已提交
7894 7895
}

7896
static s64 cpu_rt_runtime_read(struct cgroup *cgrp, struct cftype *cft)
P
Peter Zijlstra 已提交
7897
{
7898
	return sched_group_rt_runtime(cgroup_tg(cgrp));
P
Peter Zijlstra 已提交
7899
}
7900 7901 7902 7903 7904 7905 7906 7907 7908 7909 7910

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));
}
7911
#endif /* CONFIG_RT_GROUP_SCHED */
P
Peter Zijlstra 已提交
7912

7913
static struct cftype cpu_files[] = {
7914
#ifdef CONFIG_FAIR_GROUP_SCHED
7915 7916
	{
		.name = "shares",
7917 7918
		.read_u64 = cpu_shares_read_u64,
		.write_u64 = cpu_shares_write_u64,
7919
	},
7920
#endif
7921 7922 7923 7924 7925 7926 7927 7928 7929 7930 7931
#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,
	},
7932 7933 7934 7935
	{
		.name = "stat",
		.read_map = cpu_stats_show,
	},
7936
#endif
7937
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
7938
	{
P
Peter Zijlstra 已提交
7939
		.name = "rt_runtime_us",
7940 7941
		.read_s64 = cpu_rt_runtime_read,
		.write_s64 = cpu_rt_runtime_write,
P
Peter Zijlstra 已提交
7942
	},
7943 7944
	{
		.name = "rt_period_us",
7945 7946
		.read_u64 = cpu_rt_period_read_uint,
		.write_u64 = cpu_rt_period_write_uint,
7947
	},
7948
#endif
7949 7950 7951 7952
};

static int cpu_cgroup_populate(struct cgroup_subsys *ss, struct cgroup *cont)
{
7953
	return cgroup_add_files(cont, ss, cpu_files, ARRAY_SIZE(cpu_files));
7954 7955 7956
}

struct cgroup_subsys cpu_cgroup_subsys = {
I
Ingo Molnar 已提交
7957 7958 7959
	.name		= "cpu",
	.create		= cpu_cgroup_create,
	.destroy	= cpu_cgroup_destroy,
7960 7961
	.can_attach	= cpu_cgroup_can_attach,
	.attach		= cpu_cgroup_attach,
7962
	.exit		= cpu_cgroup_exit,
I
Ingo Molnar 已提交
7963 7964
	.populate	= cpu_cgroup_populate,
	.subsys_id	= cpu_cgroup_subsys_id,
7965 7966 7967
	.early_init	= 1,
};

7968
#endif	/* CONFIG_CGROUP_SCHED */
7969 7970 7971 7972 7973 7974 7975 7976 7977 7978 7979 7980

#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 */
static struct cgroup_subsys_state *cpuacct_create(
7981
	struct cgroup_subsys *ss, struct cgroup *cgrp)
7982
{
7983
	struct cpuacct *ca;
7984

7985 7986 7987 7988
	if (!cgrp->parent)
		return &root_cpuacct.css;

	ca = kzalloc(sizeof(*ca), GFP_KERNEL);
7989
	if (!ca)
7990
		goto out;
7991 7992

	ca->cpuusage = alloc_percpu(u64);
7993 7994 7995
	if (!ca->cpuusage)
		goto out_free_ca;

7996 7997 7998
	ca->cpustat = alloc_percpu(struct kernel_cpustat);
	if (!ca->cpustat)
		goto out_free_cpuusage;
7999

8000
	return &ca->css;
8001

8002
out_free_cpuusage:
8003 8004 8005 8006 8007
	free_percpu(ca->cpuusage);
out_free_ca:
	kfree(ca);
out:
	return ERR_PTR(-ENOMEM);
8008 8009 8010
}

/* destroy an existing cpu accounting group */
I
Ingo Molnar 已提交
8011
static void
8012
cpuacct_destroy(struct cgroup_subsys *ss, struct cgroup *cgrp)
8013
{
8014
	struct cpuacct *ca = cgroup_ca(cgrp);
8015

8016
	free_percpu(ca->cpustat);
8017 8018 8019 8020
	free_percpu(ca->cpuusage);
	kfree(ca);
}

8021 8022
static u64 cpuacct_cpuusage_read(struct cpuacct *ca, int cpu)
{
8023
	u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
8024 8025 8026 8027 8028 8029
	u64 data;

#ifndef CONFIG_64BIT
	/*
	 * Take rq->lock to make 64-bit read safe on 32-bit platforms.
	 */
8030
	raw_spin_lock_irq(&cpu_rq(cpu)->lock);
8031
	data = *cpuusage;
8032
	raw_spin_unlock_irq(&cpu_rq(cpu)->lock);
8033 8034 8035 8036 8037 8038 8039 8040 8041
#else
	data = *cpuusage;
#endif

	return data;
}

static void cpuacct_cpuusage_write(struct cpuacct *ca, int cpu, u64 val)
{
8042
	u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
8043 8044 8045 8046 8047

#ifndef CONFIG_64BIT
	/*
	 * Take rq->lock to make 64-bit write safe on 32-bit platforms.
	 */
8048
	raw_spin_lock_irq(&cpu_rq(cpu)->lock);
8049
	*cpuusage = val;
8050
	raw_spin_unlock_irq(&cpu_rq(cpu)->lock);
8051 8052 8053 8054 8055
#else
	*cpuusage = val;
#endif
}

8056
/* return total cpu usage (in nanoseconds) of a group */
8057
static u64 cpuusage_read(struct cgroup *cgrp, struct cftype *cft)
8058
{
8059
	struct cpuacct *ca = cgroup_ca(cgrp);
8060 8061 8062
	u64 totalcpuusage = 0;
	int i;

8063 8064
	for_each_present_cpu(i)
		totalcpuusage += cpuacct_cpuusage_read(ca, i);
8065 8066 8067 8068

	return totalcpuusage;
}

8069 8070 8071 8072 8073 8074 8075 8076 8077 8078 8079 8080
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;
	}

8081 8082
	for_each_present_cpu(i)
		cpuacct_cpuusage_write(ca, i, 0);
8083 8084 8085 8086 8087

out:
	return err;
}

8088 8089 8090 8091 8092 8093 8094 8095 8096 8097 8098 8099 8100 8101 8102
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;
}

8103 8104 8105 8106 8107 8108
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,
8109
			      struct cgroup_map_cb *cb)
8110 8111
{
	struct cpuacct *ca = cgroup_ca(cgrp);
8112 8113
	int cpu;
	s64 val = 0;
8114

8115 8116 8117 8118
	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];
8119
	}
8120 8121
	val = cputime64_to_clock_t(val);
	cb->fill(cb, cpuacct_stat_desc[CPUACCT_STAT_USER], val);
8122

8123 8124 8125 8126 8127 8128
	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];
8129
	}
8130 8131 8132 8133

	val = cputime64_to_clock_t(val);
	cb->fill(cb, cpuacct_stat_desc[CPUACCT_STAT_SYSTEM], val);

8134 8135 8136
	return 0;
}

8137 8138 8139
static struct cftype files[] = {
	{
		.name = "usage",
8140 8141
		.read_u64 = cpuusage_read,
		.write_u64 = cpuusage_write,
8142
	},
8143 8144 8145 8146
	{
		.name = "usage_percpu",
		.read_seq_string = cpuacct_percpu_seq_read,
	},
8147 8148 8149 8150
	{
		.name = "stat",
		.read_map = cpuacct_stats_show,
	},
8151 8152
};

8153
static int cpuacct_populate(struct cgroup_subsys *ss, struct cgroup *cgrp)
8154
{
8155
	return cgroup_add_files(cgrp, ss, files, ARRAY_SIZE(files));
8156 8157 8158 8159 8160 8161 8162
}

/*
 * charge this task's execution time to its accounting group.
 *
 * called with rq->lock held.
 */
8163
void cpuacct_charge(struct task_struct *tsk, u64 cputime)
8164 8165
{
	struct cpuacct *ca;
8166
	int cpu;
8167

L
Li Zefan 已提交
8168
	if (unlikely(!cpuacct_subsys.active))
8169 8170
		return;

8171
	cpu = task_cpu(tsk);
8172 8173 8174

	rcu_read_lock();

8175 8176
	ca = task_ca(tsk);

8177
	for (; ca; ca = parent_ca(ca)) {
8178
		u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
8179 8180
		*cpuusage += cputime;
	}
8181 8182

	rcu_read_unlock();
8183 8184 8185 8186 8187 8188 8189 8190 8191 8192
}

struct cgroup_subsys cpuacct_subsys = {
	.name = "cpuacct",
	.create = cpuacct_create,
	.destroy = cpuacct_destroy,
	.populate = cpuacct_populate,
	.subsys_id = cpuacct_subsys_id,
};
#endif	/* CONFIG_CGROUP_CPUACCT */