core.c 193.1 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>
G
Glauber Costa 已提交
77 78 79
#ifdef CONFIG_PARAVIRT
#include <asm/paravirt.h>
#endif
L
Linus Torvalds 已提交
80

81
#include "sched.h"
82
#include "../workqueue_sched.h"
83

84
#define CREATE_TRACE_POINTS
85
#include <trace/events/sched.h>
86

87
void start_bandwidth_timer(struct hrtimer *period_timer, ktime_t period)
88
{
89 90
	unsigned long delta;
	ktime_t soft, hard, now;
91

92 93 94 95 96 97
	for (;;) {
		if (hrtimer_active(period_timer))
			break;

		now = hrtimer_cb_get_time(period_timer);
		hrtimer_forward(period_timer, now, period);
98

99 100 101 102 103 104 105 106
		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);
	}
}

107 108
DEFINE_MUTEX(sched_domains_mutex);
DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
109

110
static void update_rq_clock_task(struct rq *rq, s64 delta);
111

112
void update_rq_clock(struct rq *rq)
113
{
114
	s64 delta;
115

116
	if (rq->skip_clock_update > 0)
117
		return;
118

119 120 121
	delta = sched_clock_cpu(cpu_of(rq)) - rq->clock;
	rq->clock += delta;
	update_rq_clock_task(rq, delta);
122 123
}

I
Ingo Molnar 已提交
124 125 126
/*
 * Debugging: various feature bits
 */
P
Peter Zijlstra 已提交
127 128 129 130

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

I
Ingo Molnar 已提交
131
const_debug unsigned int sysctl_sched_features =
132
#include "features.h"
P
Peter Zijlstra 已提交
133 134 135 136 137 138 139 140
	0;

#undef SCHED_FEAT

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

141
static __read_mostly char *sched_feat_names[] = {
142
#include "features.h"
P
Peter Zijlstra 已提交
143 144 145 146 147
	NULL
};

#undef SCHED_FEAT

L
Li Zefan 已提交
148
static int sched_feat_show(struct seq_file *m, void *v)
P
Peter Zijlstra 已提交
149 150 151 152
{
	int i;

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

L
Li Zefan 已提交
159
	return 0;
P
Peter Zijlstra 已提交
160 161 162 163 164 165 166
}

static ssize_t
sched_feat_write(struct file *filp, const char __user *ubuf,
		size_t cnt, loff_t *ppos)
{
	char buf[64];
167
	char *cmp;
P
Peter Zijlstra 已提交
168 169 170 171 172 173 174 175 176 177
	int neg = 0;
	int i;

	if (cnt > 63)
		cnt = 63;

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

	buf[cnt] = 0;
178
	cmp = strstrip(buf);
P
Peter Zijlstra 已提交
179

H
Hillf Danton 已提交
180
	if (strncmp(cmp, "NO_", 3) == 0) {
P
Peter Zijlstra 已提交
181 182 183 184 185
		neg = 1;
		cmp += 3;
	}

	for (i = 0; sched_feat_names[i]; i++) {
186
		if (strcmp(cmp, sched_feat_names[i]) == 0) {
P
Peter Zijlstra 已提交
187 188 189 190 191 192 193 194 195 196 197
			if (neg)
				sysctl_sched_features &= ~(1UL << i);
			else
				sysctl_sched_features |= (1UL << i);
			break;
		}
	}

	if (!sched_feat_names[i])
		return -EINVAL;

198
	*ppos += cnt;
P
Peter Zijlstra 已提交
199 200 201 202

	return cnt;
}

L
Li Zefan 已提交
203 204 205 206 207
static int sched_feat_open(struct inode *inode, struct file *filp)
{
	return single_open(filp, sched_feat_show, NULL);
}

208
static const struct file_operations sched_feat_fops = {
L
Li Zefan 已提交
209 210 211 212 213
	.open		= sched_feat_open,
	.write		= sched_feat_write,
	.read		= seq_read,
	.llseek		= seq_lseek,
	.release	= single_release,
P
Peter Zijlstra 已提交
214 215 216 217 218 219 220 221 222 223 224 225 226
};

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);

#endif

227 228 229 230 231 232
/*
 * 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;

233 234 235 236 237 238 239 240
/*
 * 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;

241 242 243 244 245
/*
 * period over which we measure -rt task cpu usage in us.
 * default: 1s
 */
unsigned int sysctl_sched_rt_period = 1000000;
246

247 248 249 250 251 252 253
__read_mostly int scheduler_running;

/*
 * part of the period that we allow rt tasks to run in us.
 * default: 0.95s
 */
int sysctl_sched_rt_runtime = 950000;
254 255


L
Linus Torvalds 已提交
256

257
/*
258
 * __task_rq_lock - lock the rq @p resides on.
259
 */
260
static inline struct rq *__task_rq_lock(struct task_struct *p)
261 262
	__acquires(rq->lock)
{
263 264
	struct rq *rq;

265 266
	lockdep_assert_held(&p->pi_lock);

267
	for (;;) {
268
		rq = task_rq(p);
269
		raw_spin_lock(&rq->lock);
P
Peter Zijlstra 已提交
270
		if (likely(rq == task_rq(p)))
271
			return rq;
272
		raw_spin_unlock(&rq->lock);
273 274 275
	}
}

L
Linus Torvalds 已提交
276
/*
277
 * task_rq_lock - lock p->pi_lock and lock the rq @p resides on.
L
Linus Torvalds 已提交
278
 */
279
static struct rq *task_rq_lock(struct task_struct *p, unsigned long *flags)
280
	__acquires(p->pi_lock)
L
Linus Torvalds 已提交
281 282
	__acquires(rq->lock)
{
283
	struct rq *rq;
L
Linus Torvalds 已提交
284

285
	for (;;) {
286
		raw_spin_lock_irqsave(&p->pi_lock, *flags);
287
		rq = task_rq(p);
288
		raw_spin_lock(&rq->lock);
P
Peter Zijlstra 已提交
289
		if (likely(rq == task_rq(p)))
290
			return rq;
291 292
		raw_spin_unlock(&rq->lock);
		raw_spin_unlock_irqrestore(&p->pi_lock, *flags);
L
Linus Torvalds 已提交
293 294 295
	}
}

A
Alexey Dobriyan 已提交
296
static void __task_rq_unlock(struct rq *rq)
297 298
	__releases(rq->lock)
{
299
	raw_spin_unlock(&rq->lock);
300 301
}

302 303
static inline void
task_rq_unlock(struct rq *rq, struct task_struct *p, unsigned long *flags)
L
Linus Torvalds 已提交
304
	__releases(rq->lock)
305
	__releases(p->pi_lock)
L
Linus Torvalds 已提交
306
{
307 308
	raw_spin_unlock(&rq->lock);
	raw_spin_unlock_irqrestore(&p->pi_lock, *flags);
L
Linus Torvalds 已提交
309 310 311
}

/*
312
 * this_rq_lock - lock this runqueue and disable interrupts.
L
Linus Torvalds 已提交
313
 */
A
Alexey Dobriyan 已提交
314
static struct rq *this_rq_lock(void)
L
Linus Torvalds 已提交
315 316
	__acquires(rq->lock)
{
317
	struct rq *rq;
L
Linus Torvalds 已提交
318 319 320

	local_irq_disable();
	rq = this_rq();
321
	raw_spin_lock(&rq->lock);
L
Linus Torvalds 已提交
322 323 324 325

	return rq;
}

P
Peter Zijlstra 已提交
326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353
#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());

354
	raw_spin_lock(&rq->lock);
355
	update_rq_clock(rq);
P
Peter Zijlstra 已提交
356
	rq->curr->sched_class->task_tick(rq, rq->curr, 1);
357
	raw_spin_unlock(&rq->lock);
P
Peter Zijlstra 已提交
358 359 360 361

	return HRTIMER_NORESTART;
}

362
#ifdef CONFIG_SMP
363 364 365 366
/*
 * called from hardirq (IPI) context
 */
static void __hrtick_start(void *arg)
367
{
368
	struct rq *rq = arg;
369

370
	raw_spin_lock(&rq->lock);
371 372
	hrtimer_restart(&rq->hrtick_timer);
	rq->hrtick_csd_pending = 0;
373
	raw_spin_unlock(&rq->lock);
374 375
}

376 377 378 379 380
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
381
void hrtick_start(struct rq *rq, u64 delay)
382
{
383 384
	struct hrtimer *timer = &rq->hrtick_timer;
	ktime_t time = ktime_add_ns(timer->base->get_time(), delay);
385

386
	hrtimer_set_expires(timer, time);
387 388 389 390

	if (rq == this_rq()) {
		hrtimer_restart(timer);
	} else if (!rq->hrtick_csd_pending) {
391
		__smp_call_function_single(cpu_of(rq), &rq->hrtick_csd, 0);
392 393
		rq->hrtick_csd_pending = 1;
	}
394 395 396 397 398 399 400 401 402 403 404 405 406 407
}

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:
408
		hrtick_clear(cpu_rq(cpu));
409 410 411 412 413 414
		return NOTIFY_OK;
	}

	return NOTIFY_DONE;
}

415
static __init void init_hrtick(void)
416 417 418
{
	hotcpu_notifier(hotplug_hrtick, 0);
}
419 420 421 422 423 424
#else
/*
 * Called to set the hrtick timer state.
 *
 * called with rq->lock held and irqs disabled
 */
425
void hrtick_start(struct rq *rq, u64 delay)
426
{
427
	__hrtimer_start_range_ns(&rq->hrtick_timer, ns_to_ktime(delay), 0,
428
			HRTIMER_MODE_REL_PINNED, 0);
429
}
430

A
Andrew Morton 已提交
431
static inline void init_hrtick(void)
P
Peter Zijlstra 已提交
432 433
{
}
434
#endif /* CONFIG_SMP */
P
Peter Zijlstra 已提交
435

436
static void init_rq_hrtick(struct rq *rq)
P
Peter Zijlstra 已提交
437
{
438 439
#ifdef CONFIG_SMP
	rq->hrtick_csd_pending = 0;
P
Peter Zijlstra 已提交
440

441 442 443 444
	rq->hrtick_csd.flags = 0;
	rq->hrtick_csd.func = __hrtick_start;
	rq->hrtick_csd.info = rq;
#endif
P
Peter Zijlstra 已提交
445

446 447
	hrtimer_init(&rq->hrtick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	rq->hrtick_timer.function = hrtick;
P
Peter Zijlstra 已提交
448
}
A
Andrew Morton 已提交
449
#else	/* CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
450 451 452 453 454 455 456 457
static inline void hrtick_clear(struct rq *rq)
{
}

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

458 459 460
static inline void init_hrtick(void)
{
}
A
Andrew Morton 已提交
461
#endif	/* CONFIG_SCHED_HRTICK */
P
Peter Zijlstra 已提交
462

I
Ingo Molnar 已提交
463 464 465 466 467 468 469 470 471 472 473 474 475
/*
 * 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

476
void resched_task(struct task_struct *p)
I
Ingo Molnar 已提交
477 478 479
{
	int cpu;

480
	assert_raw_spin_locked(&task_rq(p)->lock);
I
Ingo Molnar 已提交
481

482
	if (test_tsk_need_resched(p))
I
Ingo Molnar 已提交
483 484
		return;

485
	set_tsk_need_resched(p);
I
Ingo Molnar 已提交
486 487 488 489 490 491 492 493 494 495 496

	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);
}

497
void resched_cpu(int cpu)
I
Ingo Molnar 已提交
498 499 500 501
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

502
	if (!raw_spin_trylock_irqsave(&rq->lock, flags))
I
Ingo Molnar 已提交
503 504
		return;
	resched_task(cpu_curr(cpu));
505
	raw_spin_unlock_irqrestore(&rq->lock, flags);
I
Ingo Molnar 已提交
506
}
507 508

#ifdef CONFIG_NO_HZ
509 510 511 512 513 514 515 516 517 518 519 520 521 522
/*
 * 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;

523
	rcu_read_lock();
524
	for_each_domain(cpu, sd) {
525 526 527 528 529 530
		for_each_cpu(i, sched_domain_span(sd)) {
			if (!idle_cpu(i)) {
				cpu = i;
				goto unlock;
			}
		}
531
	}
532 533
unlock:
	rcu_read_unlock();
534 535
	return cpu;
}
536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567
/*
 * 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;

	/*
	 * 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()
	 */
568
	set_tsk_need_resched(rq->idle);
569 570 571 572 573 574

	/* NEED_RESCHED must be visible before we test polling */
	smp_mb();
	if (!tsk_is_polling(rq->idle))
		smp_send_reschedule(cpu);
}
M
Mike Galbraith 已提交
575

576 577
static inline bool got_nohz_idle_kick(void)
{
578 579
	int cpu = smp_processor_id();
	return idle_cpu(cpu) && test_bit(NOHZ_BALANCE_KICK, nohz_flags(cpu));
580 581 582 583 584 585 586 587 588
}

#else /* CONFIG_NO_HZ */

static inline bool got_nohz_idle_kick(void)
{
	return false;
}

589
#endif /* CONFIG_NO_HZ */
590

591
void sched_avg_update(struct rq *rq)
592 593 594 595
{
	s64 period = sched_avg_period();

	while ((s64)(rq->clock - rq->age_stamp) > period) {
596 597 598 599 600 601
		/*
		 * 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));
602 603 604 605 606
		rq->age_stamp += period;
		rq->rt_avg /= 2;
	}
}

607
#else /* !CONFIG_SMP */
608
void resched_task(struct task_struct *p)
I
Ingo Molnar 已提交
609
{
610
	assert_raw_spin_locked(&task_rq(p)->lock);
611
	set_tsk_need_resched(p);
I
Ingo Molnar 已提交
612
}
613
#endif /* CONFIG_SMP */
I
Ingo Molnar 已提交
614

615 616
#if defined(CONFIG_RT_GROUP_SCHED) || (defined(CONFIG_FAIR_GROUP_SCHED) && \
			(defined(CONFIG_SMP) || defined(CONFIG_CFS_BANDWIDTH)))
617
/*
618 619 620 621
 * 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.
622
 */
623
int walk_tg_tree_from(struct task_group *from,
624
			     tg_visitor down, tg_visitor up, void *data)
625 626
{
	struct task_group *parent, *child;
P
Peter Zijlstra 已提交
627
	int ret;
628

629 630
	parent = from;

631
down:
P
Peter Zijlstra 已提交
632 633
	ret = (*down)(parent, data);
	if (ret)
634
		goto out;
635 636 637 638 639
	list_for_each_entry_rcu(child, &parent->children, siblings) {
		parent = child;
		goto down;

up:
640 641 642 643 644
		continue;
	}
	ret = (*up)(parent, data);
	if (ret || parent == from)
		goto out;
645

646 647 648 649 650 651
	child = parent;
	parent = parent->parent;
	if (parent)
		goto up;
out:
	return ret;
652 653
}

654
int tg_nop(struct task_group *tg, void *data)
655
{
656
	return 0;
657
}
658 659 660
#endif

void update_cpu_load(struct rq *this_rq);
661

662 663
static void set_load_weight(struct task_struct *p)
{
N
Nikhil Rao 已提交
664 665 666
	int prio = p->static_prio - MAX_RT_PRIO;
	struct load_weight *load = &p->se.load;

I
Ingo Molnar 已提交
667 668 669 670
	/*
	 * SCHED_IDLE tasks get minimal weight:
	 */
	if (p->policy == SCHED_IDLE) {
671
		load->weight = scale_load(WEIGHT_IDLEPRIO);
N
Nikhil Rao 已提交
672
		load->inv_weight = WMULT_IDLEPRIO;
I
Ingo Molnar 已提交
673 674
		return;
	}
675

676
	load->weight = scale_load(prio_to_weight[prio]);
N
Nikhil Rao 已提交
677
	load->inv_weight = prio_to_wmult[prio];
678 679
}

680
static void enqueue_task(struct rq *rq, struct task_struct *p, int flags)
681
{
682
	update_rq_clock(rq);
I
Ingo Molnar 已提交
683
	sched_info_queued(p);
684
	p->sched_class->enqueue_task(rq, p, flags);
685 686
}

687
static void dequeue_task(struct rq *rq, struct task_struct *p, int flags)
688
{
689
	update_rq_clock(rq);
690
	sched_info_dequeued(p);
691
	p->sched_class->dequeue_task(rq, p, flags);
692 693
}

694 695 696
/*
 * activate_task - move a task to the runqueue.
 */
697
void activate_task(struct rq *rq, struct task_struct *p, int flags)
698 699 700 701
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible--;

702
	enqueue_task(rq, p, flags);
703 704 705 706 707
}

/*
 * deactivate_task - remove a task from the runqueue.
 */
708
void deactivate_task(struct rq *rq, struct task_struct *p, int flags)
709 710 711 712
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible++;

713
	dequeue_task(rq, p, flags);
714 715
}

716 717
#ifdef CONFIG_IRQ_TIME_ACCOUNTING

718 719 720 721 722 723 724
/*
 * 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
725 726 727
 * 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.
728
 */
729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744
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;
}

745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782
#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)
783 784 785
{
	return per_cpu(cpu_softirq_time, cpu) + per_cpu(cpu_hardirq_time, cpu);
}
786
#endif /* CONFIG_64BIT */
787

788 789 790 791
/*
 * Called before incrementing preempt_count on {soft,}irq_enter
 * and before decrementing preempt_count on {soft,}irq_exit.
 */
792 793 794
void account_system_vtime(struct task_struct *curr)
{
	unsigned long flags;
795
	s64 delta;
796 797 798 799 800 801 802 803
	int cpu;

	if (!sched_clock_irqtime)
		return;

	local_irq_save(flags);

	cpu = smp_processor_id();
804 805 806
	delta = sched_clock_cpu(cpu) - __this_cpu_read(irq_start_time);
	__this_cpu_add(irq_start_time, delta);

807
	irq_time_write_begin();
808 809 810 811 812 813 814
	/*
	 * 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())
815
		__this_cpu_add(cpu_hardirq_time, delta);
816
	else if (in_serving_softirq() && curr != this_cpu_ksoftirqd())
817
		__this_cpu_add(cpu_softirq_time, delta);
818

819
	irq_time_write_end();
820 821
	local_irq_restore(flags);
}
I
Ingo Molnar 已提交
822
EXPORT_SYMBOL_GPL(account_system_vtime);
823

G
Glauber Costa 已提交
824 825 826 827
#endif /* CONFIG_IRQ_TIME_ACCOUNTING */

#ifdef CONFIG_PARAVIRT
static inline u64 steal_ticks(u64 steal)
828
{
G
Glauber Costa 已提交
829 830
	if (unlikely(steal > NSEC_PER_SEC))
		return div_u64(steal, TICK_NSEC);
831

G
Glauber Costa 已提交
832 833 834 835
	return __iter_div_u64_rem(steal, TICK_NSEC, &steal);
}
#endif

836
static void update_rq_clock_task(struct rq *rq, s64 delta)
837
{
838 839 840 841 842 843 844 845
/*
 * 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
846
	irq_delta = irq_time_read(cpu_of(rq)) - rq->prev_irq_time;
847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867

	/*
	 * 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;
868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887
#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

888 889
	rq->clock_task += delta;

890 891 892 893
#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
894 895
}

896
#ifdef CONFIG_IRQ_TIME_ACCOUNTING
897 898
static int irqtime_account_hi_update(void)
{
899
	u64 *cpustat = kcpustat_this_cpu->cpustat;
900 901 902 903 904 905
	unsigned long flags;
	u64 latest_ns;
	int ret = 0;

	local_irq_save(flags);
	latest_ns = this_cpu_read(cpu_hardirq_time);
906
	if (cputime64_gt(nsecs_to_cputime64(latest_ns), cpustat[CPUTIME_IRQ]))
907 908 909 910 911 912 913
		ret = 1;
	local_irq_restore(flags);
	return ret;
}

static int irqtime_account_si_update(void)
{
914
	u64 *cpustat = kcpustat_this_cpu->cpustat;
915 916 917 918 919 920
	unsigned long flags;
	u64 latest_ns;
	int ret = 0;

	local_irq_save(flags);
	latest_ns = this_cpu_read(cpu_softirq_time);
921
	if (cputime64_gt(nsecs_to_cputime64(latest_ns), cpustat[CPUTIME_SOFTIRQ]))
922 923 924 925 926
		ret = 1;
	local_irq_restore(flags);
	return ret;
}

927
#else /* CONFIG_IRQ_TIME_ACCOUNTING */
928

929 930
#define sched_clock_irqtime	(0)

931
#endif
932

933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962
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;
	}
}

963
/*
I
Ingo Molnar 已提交
964
 * __normal_prio - return the priority that is based on the static prio
965 966 967
 */
static inline int __normal_prio(struct task_struct *p)
{
I
Ingo Molnar 已提交
968
	return p->static_prio;
969 970
}

971 972 973 974 975 976 977
/*
 * 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.
 */
978
static inline int normal_prio(struct task_struct *p)
979 980 981
{
	int prio;

982
	if (task_has_rt_policy(p))
983 984 985 986 987 988 989 990 991 992 993 994 995
		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.
 */
996
static int effective_prio(struct task_struct *p)
997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008
{
	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
1009 1010 1011 1012
/**
 * task_curr - is this task currently executing on a CPU?
 * @p: the task in question.
 */
1013
inline int task_curr(const struct task_struct *p)
L
Linus Torvalds 已提交
1014 1015 1016 1017
{
	return cpu_curr(task_cpu(p)) == p;
}

1018 1019
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
1020
				       int oldprio)
1021 1022 1023
{
	if (prev_class != p->sched_class) {
		if (prev_class->switched_from)
P
Peter Zijlstra 已提交
1024 1025 1026 1027
			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);
1028 1029
}

1030
void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags)
1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050
{
	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
1051
	if (rq->curr->on_rq && test_tsk_need_resched(rq->curr))
1052 1053 1054
		rq->skip_clock_update = 1;
}

L
Linus Torvalds 已提交
1055
#ifdef CONFIG_SMP
I
Ingo Molnar 已提交
1056
void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
I
Ingo Molnar 已提交
1057
{
1058 1059 1060 1061 1062
#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
1063 1064
	WARN_ON_ONCE(p->state != TASK_RUNNING && p->state != TASK_WAKING &&
			!(task_thread_info(p)->preempt_count & PREEMPT_ACTIVE));
1065 1066

#ifdef CONFIG_LOCKDEP
1067 1068 1069 1070 1071 1072 1073 1074 1075 1076
	/*
	 * 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().
	 */
1077 1078 1079
	WARN_ON_ONCE(debug_locks && !(lockdep_is_held(&p->pi_lock) ||
				      lockdep_is_held(&task_rq(p)->lock)));
#endif
1080 1081
#endif

1082
	trace_sched_migrate_task(p, new_cpu);
1083

1084 1085
	if (task_cpu(p) != new_cpu) {
		p->se.nr_migrations++;
1086
		perf_sw_event(PERF_COUNT_SW_CPU_MIGRATIONS, 1, NULL, 0);
1087
	}
I
Ingo Molnar 已提交
1088 1089

	__set_task_cpu(p, new_cpu);
I
Ingo Molnar 已提交
1090 1091
}

1092
struct migration_arg {
1093
	struct task_struct *task;
L
Linus Torvalds 已提交
1094
	int dest_cpu;
1095
};
L
Linus Torvalds 已提交
1096

1097 1098
static int migration_cpu_stop(void *data);

L
Linus Torvalds 已提交
1099 1100 1101
/*
 * wait_task_inactive - wait for a thread to unschedule.
 *
R
Roland McGrath 已提交
1102 1103 1104 1105 1106 1107 1108
 * 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
1109 1110 1111 1112 1113 1114
 * 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
1115
unsigned long wait_task_inactive(struct task_struct *p, long match_state)
L
Linus Torvalds 已提交
1116 1117
{
	unsigned long flags;
I
Ingo Molnar 已提交
1118
	int running, on_rq;
R
Roland McGrath 已提交
1119
	unsigned long ncsw;
1120
	struct rq *rq;
L
Linus Torvalds 已提交
1121

1122 1123 1124 1125 1126 1127 1128 1129
	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);
1130

1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141
		/*
		 * 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
1142 1143 1144
		while (task_running(rq, p)) {
			if (match_state && unlikely(p->state != match_state))
				return 0;
1145
			cpu_relax();
R
Roland McGrath 已提交
1146
		}
1147

1148 1149 1150 1151 1152 1153
		/*
		 * 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);
1154
		trace_sched_wait_task(p);
1155
		running = task_running(rq, p);
P
Peter Zijlstra 已提交
1156
		on_rq = p->on_rq;
R
Roland McGrath 已提交
1157
		ncsw = 0;
1158
		if (!match_state || p->state == match_state)
1159
			ncsw = p->nvcsw | LONG_MIN; /* sets MSB */
1160
		task_rq_unlock(rq, p, &flags);
1161

R
Roland McGrath 已提交
1162 1163 1164 1165 1166 1167
		/*
		 * If it changed from the expected state, bail out now.
		 */
		if (unlikely(!ncsw))
			break;

1168 1169 1170 1171 1172 1173 1174 1175 1176 1177
		/*
		 * 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;
		}
1178

1179 1180 1181 1182 1183
		/*
		 * It's not enough that it's not actively running,
		 * it must be off the runqueue _entirely_, and not
		 * preempted!
		 *
1184
		 * So if it was still runnable (but just not actively
1185 1186 1187 1188
		 * running right now), it's preempted, and we should
		 * yield - it could be a while.
		 */
		if (unlikely(on_rq)) {
1189 1190 1191 1192
			ktime_t to = ktime_set(0, NSEC_PER_SEC/HZ);

			set_current_state(TASK_UNINTERRUPTIBLE);
			schedule_hrtimeout(&to, HRTIMER_MODE_REL);
1193 1194
			continue;
		}
1195

1196 1197 1198 1199 1200 1201 1202
		/*
		 * 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
1203 1204

	return ncsw;
L
Linus Torvalds 已提交
1205 1206 1207 1208 1209 1210 1211 1212 1213
}

/***
 * 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
1214
 * NOTE: this function doesn't have to take the runqueue lock,
L
Linus Torvalds 已提交
1215 1216 1217 1218 1219
 * 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.
 */
1220
void kick_process(struct task_struct *p)
L
Linus Torvalds 已提交
1221 1222 1223 1224 1225 1226 1227 1228 1229
{
	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
1230
EXPORT_SYMBOL_GPL(kick_process);
N
Nick Piggin 已提交
1231
#endif /* CONFIG_SMP */
L
Linus Torvalds 已提交
1232

1233
#ifdef CONFIG_SMP
1234
/*
1235
 * ->cpus_allowed is protected by both rq->lock and p->pi_lock
1236
 */
1237 1238 1239 1240 1241 1242 1243
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)
1244
		if (cpumask_test_cpu(dest_cpu, tsk_cpus_allowed(p)))
1245 1246 1247
			return dest_cpu;

	/* Any allowed, online CPU? */
1248
	dest_cpu = cpumask_any_and(tsk_cpus_allowed(p), cpu_active_mask);
1249 1250 1251 1252
	if (dest_cpu < nr_cpu_ids)
		return dest_cpu;

	/* No more Mr. Nice Guy. */
1253 1254 1255 1256 1257 1258 1259 1260 1261
	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()) {
		printk(KERN_INFO "process %d (%s) no longer affine to cpu%d\n",
				task_pid_nr(p), p->comm, cpu);
1262 1263 1264 1265 1266
	}

	return dest_cpu;
}

1267
/*
1268
 * The caller (fork, wakeup) owns p->pi_lock, ->cpus_allowed is stable.
1269
 */
1270
static inline
1271
int select_task_rq(struct task_struct *p, int sd_flags, int wake_flags)
1272
{
1273
	int cpu = p->sched_class->select_task_rq(p, sd_flags, wake_flags);
1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284

	/*
	 * 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 ]
	 */
1285
	if (unlikely(!cpumask_test_cpu(cpu, tsk_cpus_allowed(p)) ||
P
Peter Zijlstra 已提交
1286
		     !cpu_online(cpu)))
1287
		cpu = select_fallback_rq(task_cpu(p), p);
1288 1289

	return cpu;
1290
}
1291 1292 1293 1294 1295 1296

static void update_avg(u64 *avg, u64 sample)
{
	s64 diff = sample - *avg;
	*avg += diff >> 3;
}
1297 1298
#endif

P
Peter Zijlstra 已提交
1299
static void
1300
ttwu_stat(struct task_struct *p, int cpu, int wake_flags)
T
Tejun Heo 已提交
1301
{
P
Peter Zijlstra 已提交
1302
#ifdef CONFIG_SCHEDSTATS
1303 1304
	struct rq *rq = this_rq();

P
Peter Zijlstra 已提交
1305 1306 1307 1308 1309 1310 1311 1312 1313 1314
#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);
1315
		rcu_read_lock();
P
Peter Zijlstra 已提交
1316 1317 1318 1319 1320 1321
		for_each_domain(this_cpu, sd) {
			if (cpumask_test_cpu(cpu, sched_domain_span(sd))) {
				schedstat_inc(sd, ttwu_wake_remote);
				break;
			}
		}
1322
		rcu_read_unlock();
P
Peter Zijlstra 已提交
1323
	}
1324 1325 1326 1327

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

P
Peter Zijlstra 已提交
1328 1329 1330
#endif /* CONFIG_SMP */

	schedstat_inc(rq, ttwu_count);
T
Tejun Heo 已提交
1331
	schedstat_inc(p, se.statistics.nr_wakeups);
P
Peter Zijlstra 已提交
1332 1333

	if (wake_flags & WF_SYNC)
T
Tejun Heo 已提交
1334
		schedstat_inc(p, se.statistics.nr_wakeups_sync);
P
Peter Zijlstra 已提交
1335 1336 1337 1338 1339 1340

#endif /* CONFIG_SCHEDSTATS */
}

static void ttwu_activate(struct rq *rq, struct task_struct *p, int en_flags)
{
T
Tejun Heo 已提交
1341
	activate_task(rq, p, en_flags);
P
Peter Zijlstra 已提交
1342
	p->on_rq = 1;
1343 1344 1345 1346

	/* 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
1347 1348
}

1349 1350 1351
/*
 * Mark the task runnable and perform wakeup-preemption.
 */
1352
static void
1353
ttwu_do_wakeup(struct rq *rq, struct task_struct *p, int wake_flags)
T
Tejun Heo 已提交
1354
{
1355
	trace_sched_wakeup(p, true);
T
Tejun Heo 已提交
1356 1357 1358 1359 1360 1361 1362
	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);

1363
	if (rq->idle_stamp) {
T
Tejun Heo 已提交
1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375
		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
}

1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408
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;
}

1409
#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
1410
static void sched_ttwu_pending(void)
1411 1412
{
	struct rq *rq = this_rq();
P
Peter Zijlstra 已提交
1413 1414
	struct llist_node *llist = llist_del_all(&rq->wake_list);
	struct task_struct *p;
1415 1416 1417

	raw_spin_lock(&rq->lock);

P
Peter Zijlstra 已提交
1418 1419 1420
	while (llist) {
		p = llist_entry(llist, struct task_struct, wake_entry);
		llist = llist_next(llist);
1421 1422 1423 1424 1425 1426 1427 1428
		ttwu_do_activate(rq, p, 0);
	}

	raw_spin_unlock(&rq->lock);
}

void scheduler_ipi(void)
{
1429
	if (llist_empty(&this_rq()->wake_list) && !got_nohz_idle_kick())
1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445
		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
1446
	sched_ttwu_pending();
1447 1448 1449 1450

	/*
	 * Check if someone kicked us for doing the nohz idle load balance.
	 */
1451 1452
	if (unlikely(got_nohz_idle_kick() && !need_resched())) {
		this_rq()->idle_balance = 1;
1453
		raise_softirq_irqoff(SCHED_SOFTIRQ);
1454
	}
1455
	irq_exit();
1456 1457 1458 1459
}

static void ttwu_queue_remote(struct task_struct *p, int cpu)
{
P
Peter Zijlstra 已提交
1460
	if (llist_add(&p->wake_entry, &cpu_rq(cpu)->wake_list))
1461 1462
		smp_send_reschedule(cpu);
}
1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482

#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 */
#endif /* CONFIG_SMP */
1483

1484 1485 1486 1487
static void ttwu_queue(struct task_struct *p, int cpu)
{
	struct rq *rq = cpu_rq(cpu);

1488
#if defined(CONFIG_SMP)
1489
	if (sched_feat(TTWU_QUEUE) && cpu != smp_processor_id()) {
1490
		sched_clock_cpu(cpu); /* sync clocks x-cpu */
1491 1492 1493 1494 1495
		ttwu_queue_remote(p, cpu);
		return;
	}
#endif

1496 1497 1498
	raw_spin_lock(&rq->lock);
	ttwu_do_activate(rq, p, 0);
	raw_spin_unlock(&rq->lock);
T
Tejun Heo 已提交
1499 1500 1501
}

/**
L
Linus Torvalds 已提交
1502
 * try_to_wake_up - wake up a thread
T
Tejun Heo 已提交
1503
 * @p: the thread to be awakened
L
Linus Torvalds 已提交
1504
 * @state: the mask of task states that can be woken
T
Tejun Heo 已提交
1505
 * @wake_flags: wake modifier flags (WF_*)
L
Linus Torvalds 已提交
1506 1507 1508 1509 1510 1511 1512
 *
 * 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
1513 1514
 * 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
1515
 */
1516 1517
static int
try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
L
Linus Torvalds 已提交
1518 1519
{
	unsigned long flags;
1520
	int cpu, success = 0;
P
Peter Zijlstra 已提交
1521

1522
	smp_wmb();
1523
	raw_spin_lock_irqsave(&p->pi_lock, flags);
P
Peter Zijlstra 已提交
1524
	if (!(p->state & state))
L
Linus Torvalds 已提交
1525 1526
		goto out;

1527
	success = 1; /* we're going to change ->state */
L
Linus Torvalds 已提交
1528 1529
	cpu = task_cpu(p);

1530 1531
	if (p->on_rq && ttwu_remote(p, wake_flags))
		goto stat;
L
Linus Torvalds 已提交
1532 1533

#ifdef CONFIG_SMP
P
Peter Zijlstra 已提交
1534
	/*
1535 1536
	 * 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
1537
	 */
1538 1539 1540
	while (p->on_cpu) {
#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
		/*
1541 1542 1543 1544 1545
		 * 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.
1546
		 */
1547
		if (ttwu_activate_remote(p, wake_flags))
1548
			goto stat;
1549
#else
1550
		cpu_relax();
1551
#endif
1552
	}
1553
	/*
1554
	 * Pairs with the smp_wmb() in finish_lock_switch().
1555
	 */
1556
	smp_rmb();
L
Linus Torvalds 已提交
1557

1558
	p->sched_contributes_to_load = !!task_contributes_to_load(p);
P
Peter Zijlstra 已提交
1559
	p->state = TASK_WAKING;
1560

1561
	if (p->sched_class->task_waking)
1562
		p->sched_class->task_waking(p);
1563

1564
	cpu = select_task_rq(p, SD_BALANCE_WAKE, wake_flags);
1565 1566
	if (task_cpu(p) != cpu) {
		wake_flags |= WF_MIGRATED;
1567
		set_task_cpu(p, cpu);
1568
	}
L
Linus Torvalds 已提交
1569 1570
#endif /* CONFIG_SMP */

1571 1572
	ttwu_queue(p, cpu);
stat:
1573
	ttwu_stat(p, cpu, wake_flags);
L
Linus Torvalds 已提交
1574
out:
1575
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
1576 1577 1578 1579

	return success;
}

T
Tejun Heo 已提交
1580 1581 1582 1583
/**
 * try_to_wake_up_local - try to wake up a local task with rq lock held
 * @p: the thread to be awakened
 *
1584
 * Put @p on the run-queue if it's not already there. The caller must
T
Tejun Heo 已提交
1585
 * ensure that this_rq() is locked, @p is bound to this_rq() and not
1586
 * the current task.
T
Tejun Heo 已提交
1587 1588 1589 1590 1591 1592 1593 1594 1595
 */
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);

1596 1597 1598 1599 1600 1601
	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
1602
	if (!(p->state & TASK_NORMAL))
1603
		goto out;
T
Tejun Heo 已提交
1604

P
Peter Zijlstra 已提交
1605
	if (!p->on_rq)
P
Peter Zijlstra 已提交
1606 1607
		ttwu_activate(rq, p, ENQUEUE_WAKEUP);

1608
	ttwu_do_wakeup(rq, p, 0);
1609
	ttwu_stat(p, smp_processor_id(), 0);
1610 1611
out:
	raw_spin_unlock(&p->pi_lock);
T
Tejun Heo 已提交
1612 1613
}

1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624
/**
 * 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.
 */
1625
int wake_up_process(struct task_struct *p)
L
Linus Torvalds 已提交
1626
{
1627
	return try_to_wake_up(p, TASK_ALL, 0);
L
Linus Torvalds 已提交
1628 1629 1630
}
EXPORT_SYMBOL(wake_up_process);

1631
int wake_up_state(struct task_struct *p, unsigned int state)
L
Linus Torvalds 已提交
1632 1633 1634 1635 1636 1637 1638
{
	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
1639 1640 1641 1642 1643
 *
 * __sched_fork() is basic setup used by init_idle() too:
 */
static void __sched_fork(struct task_struct *p)
{
P
Peter Zijlstra 已提交
1644 1645 1646
	p->on_rq			= 0;

	p->se.on_rq			= 0;
I
Ingo Molnar 已提交
1647 1648
	p->se.exec_start		= 0;
	p->se.sum_exec_runtime		= 0;
1649
	p->se.prev_sum_exec_runtime	= 0;
1650
	p->se.nr_migrations		= 0;
P
Peter Zijlstra 已提交
1651
	p->se.vruntime			= 0;
P
Peter Zijlstra 已提交
1652
	INIT_LIST_HEAD(&p->se.group_node);
I
Ingo Molnar 已提交
1653 1654

#ifdef CONFIG_SCHEDSTATS
1655
	memset(&p->se.statistics, 0, sizeof(p->se.statistics));
I
Ingo Molnar 已提交
1656
#endif
N
Nick Piggin 已提交
1657

P
Peter Zijlstra 已提交
1658
	INIT_LIST_HEAD(&p->rt.run_list);
N
Nick Piggin 已提交
1659

1660 1661 1662
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&p->preempt_notifiers);
#endif
I
Ingo Molnar 已提交
1663 1664 1665 1666 1667
}

/*
 * fork()/clone()-time setup:
 */
1668
void sched_fork(struct task_struct *p)
I
Ingo Molnar 已提交
1669
{
1670
	unsigned long flags;
I
Ingo Molnar 已提交
1671 1672 1673
	int cpu = get_cpu();

	__sched_fork(p);
1674
	/*
1675
	 * We mark the process as running here. This guarantees that
1676 1677 1678
	 * nobody will actually run it, and a signal or other external
	 * event cannot wake it up and insert it on the runqueue either.
	 */
1679
	p->state = TASK_RUNNING;
I
Ingo Molnar 已提交
1680

1681 1682 1683 1684 1685
	/*
	 * Make sure we do not leak PI boosting priority to the child.
	 */
	p->prio = current->normal_prio;

1686 1687 1688 1689
	/*
	 * Revert to default priority/policy on fork if requested.
	 */
	if (unlikely(p->sched_reset_on_fork)) {
1690
		if (task_has_rt_policy(p)) {
1691
			p->policy = SCHED_NORMAL;
1692
			p->static_prio = NICE_TO_PRIO(0);
1693 1694 1695 1696 1697 1698
			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);
1699

1700 1701 1702 1703 1704 1705
		/*
		 * We don't need the reset flag anymore after the fork. It has
		 * fulfilled its duty:
		 */
		p->sched_reset_on_fork = 0;
	}
1706

H
Hiroshi Shimamoto 已提交
1707 1708
	if (!rt_prio(p->prio))
		p->sched_class = &fair_sched_class;
1709

P
Peter Zijlstra 已提交
1710 1711 1712
	if (p->sched_class->task_fork)
		p->sched_class->task_fork(p);

1713 1714 1715 1716 1717 1718 1719
	/*
	 * 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.
	 */
1720
	raw_spin_lock_irqsave(&p->pi_lock, flags);
1721
	set_task_cpu(p, cpu);
1722
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
1723

1724
#if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT)
I
Ingo Molnar 已提交
1725
	if (likely(sched_info_on()))
1726
		memset(&p->sched_info, 0, sizeof(p->sched_info));
L
Linus Torvalds 已提交
1727
#endif
P
Peter Zijlstra 已提交
1728 1729
#if defined(CONFIG_SMP)
	p->on_cpu = 0;
1730
#endif
1731
#ifdef CONFIG_PREEMPT_COUNT
1732
	/* Want to start with kernel preemption disabled. */
A
Al Viro 已提交
1733
	task_thread_info(p)->preempt_count = 1;
L
Linus Torvalds 已提交
1734
#endif
1735
#ifdef CONFIG_SMP
1736
	plist_node_init(&p->pushable_tasks, MAX_PRIO);
1737
#endif
1738

N
Nick Piggin 已提交
1739
	put_cpu();
L
Linus Torvalds 已提交
1740 1741 1742 1743 1744 1745 1746 1747 1748
}

/*
 * 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.
 */
1749
void wake_up_new_task(struct task_struct *p)
L
Linus Torvalds 已提交
1750 1751
{
	unsigned long flags;
I
Ingo Molnar 已提交
1752
	struct rq *rq;
1753

1754
	raw_spin_lock_irqsave(&p->pi_lock, flags);
1755 1756 1757 1758 1759 1760
#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
	 */
1761
	set_task_cpu(p, select_task_rq(p, SD_BALANCE_FORK, 0));
1762 1763
#endif

1764
	rq = __task_rq_lock(p);
P
Peter Zijlstra 已提交
1765
	activate_task(rq, p, 0);
P
Peter Zijlstra 已提交
1766
	p->on_rq = 1;
1767
	trace_sched_wakeup_new(p, true);
P
Peter Zijlstra 已提交
1768
	check_preempt_curr(rq, p, WF_FORK);
1769
#ifdef CONFIG_SMP
1770 1771
	if (p->sched_class->task_woken)
		p->sched_class->task_woken(rq, p);
1772
#endif
1773
	task_rq_unlock(rq, p, &flags);
L
Linus Torvalds 已提交
1774 1775
}

1776 1777 1778
#ifdef CONFIG_PREEMPT_NOTIFIERS

/**
1779
 * preempt_notifier_register - tell me when current is being preempted & rescheduled
R
Randy Dunlap 已提交
1780
 * @notifier: notifier struct to register
1781 1782 1783 1784 1785 1786 1787 1788 1789
 */
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
1790
 * @notifier: notifier struct to unregister
1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819
 *
 * 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);
}

1820
#else /* !CONFIG_PREEMPT_NOTIFIERS */
1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831

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)
{
}

1832
#endif /* CONFIG_PREEMPT_NOTIFIERS */
1833

1834 1835 1836
/**
 * prepare_task_switch - prepare to switch tasks
 * @rq: the runqueue preparing to switch
R
Randy Dunlap 已提交
1837
 * @prev: the current task that is being switched out
1838 1839 1840 1841 1842 1843 1844 1845 1846
 * @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.
 */
1847 1848 1849
static inline void
prepare_task_switch(struct rq *rq, struct task_struct *prev,
		    struct task_struct *next)
1850
{
1851 1852
	sched_info_switch(prev, next);
	perf_event_task_sched_out(prev, next);
1853
	fire_sched_out_preempt_notifiers(prev, next);
1854 1855
	prepare_lock_switch(rq, next);
	prepare_arch_switch(next);
1856
	trace_sched_switch(prev, next);
1857 1858
}

L
Linus Torvalds 已提交
1859 1860
/**
 * finish_task_switch - clean up after a task-switch
1861
 * @rq: runqueue associated with task-switch
L
Linus Torvalds 已提交
1862 1863
 * @prev: the thread we just switched away from.
 *
1864 1865 1866 1867
 * 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
1868 1869
 *
 * Note that we may have delayed dropping an mm in context_switch(). If
I
Ingo Molnar 已提交
1870
 * so, we finish that here outside of the runqueue lock. (Doing it
L
Linus Torvalds 已提交
1871 1872 1873
 * with the lock held can cause deadlocks; see schedule() for
 * details.)
 */
A
Alexey Dobriyan 已提交
1874
static void finish_task_switch(struct rq *rq, struct task_struct *prev)
L
Linus Torvalds 已提交
1875 1876 1877
	__releases(rq->lock)
{
	struct mm_struct *mm = rq->prev_mm;
O
Oleg Nesterov 已提交
1878
	long prev_state;
L
Linus Torvalds 已提交
1879 1880 1881 1882 1883

	rq->prev_mm = NULL;

	/*
	 * A task struct has one reference for the use as "current".
1884
	 * If a task dies, then it sets TASK_DEAD in tsk->state and calls
O
Oleg Nesterov 已提交
1885 1886
	 * schedule one last time. The schedule call will never return, and
	 * the scheduled task must drop that reference.
1887
	 * The test for TASK_DEAD must occur while the runqueue locks are
L
Linus Torvalds 已提交
1888 1889 1890 1891 1892
	 * 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
1893
	prev_state = prev->state;
1894
	finish_arch_switch(prev);
1895 1896 1897
#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
	local_irq_disable();
#endif /* __ARCH_WANT_INTERRUPTS_ON_CTXSW */
1898
	perf_event_task_sched_in(prev, current);
1899 1900 1901
#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
	local_irq_enable();
#endif /* __ARCH_WANT_INTERRUPTS_ON_CTXSW */
1902
	finish_lock_switch(rq, prev);
S
Steven Rostedt 已提交
1903

1904
	fire_sched_in_preempt_notifiers(current);
L
Linus Torvalds 已提交
1905 1906
	if (mm)
		mmdrop(mm);
1907
	if (unlikely(prev_state == TASK_DEAD)) {
1908 1909 1910
		/*
		 * Remove function-return probe instances associated with this
		 * task and put them back on the free list.
I
Ingo Molnar 已提交
1911
		 */
1912
		kprobe_flush_task(prev);
L
Linus Torvalds 已提交
1913
		put_task_struct(prev);
1914
	}
L
Linus Torvalds 已提交
1915 1916
}

1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931
#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;

1932
		raw_spin_lock_irqsave(&rq->lock, flags);
1933 1934
		if (rq->curr->sched_class->post_schedule)
			rq->curr->sched_class->post_schedule(rq);
1935
		raw_spin_unlock_irqrestore(&rq->lock, flags);
1936 1937 1938 1939 1940 1941

		rq->post_schedule = 0;
	}
}

#else
1942

1943 1944 1945 1946 1947 1948
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
1949 1950
}

1951 1952
#endif

L
Linus Torvalds 已提交
1953 1954 1955 1956
/**
 * schedule_tail - first thing a freshly forked thread must call.
 * @prev: the thread we just switched away from.
 */
1957
asmlinkage void schedule_tail(struct task_struct *prev)
L
Linus Torvalds 已提交
1958 1959
	__releases(rq->lock)
{
1960 1961
	struct rq *rq = this_rq();

1962
	finish_task_switch(rq, prev);
1963

1964 1965 1966 1967 1968
	/*
	 * FIXME: do we need to worry about rq being invalidated by the
	 * task_switch?
	 */
	post_schedule(rq);
1969

1970 1971 1972 1973
#ifdef __ARCH_WANT_UNLOCKED_CTXSW
	/* In this case, finish_task_switch does not reenable preemption */
	preempt_enable();
#endif
L
Linus Torvalds 已提交
1974
	if (current->set_child_tid)
1975
		put_user(task_pid_vnr(current), current->set_child_tid);
L
Linus Torvalds 已提交
1976 1977 1978 1979 1980 1981
}

/*
 * context_switch - switch to the new MM and the new
 * thread's register state.
 */
I
Ingo Molnar 已提交
1982
static inline void
1983
context_switch(struct rq *rq, struct task_struct *prev,
1984
	       struct task_struct *next)
L
Linus Torvalds 已提交
1985
{
I
Ingo Molnar 已提交
1986
	struct mm_struct *mm, *oldmm;
L
Linus Torvalds 已提交
1987

1988
	prepare_task_switch(rq, prev, next);
1989

I
Ingo Molnar 已提交
1990 1991
	mm = next->mm;
	oldmm = prev->active_mm;
1992 1993 1994 1995 1996
	/*
	 * For paravirt, this is coupled with an exit in switch_to to
	 * combine the page table reload and the switch backend into
	 * one hypercall.
	 */
1997
	arch_start_context_switch(prev);
1998

1999
	if (!mm) {
L
Linus Torvalds 已提交
2000 2001 2002 2003 2004 2005
		next->active_mm = oldmm;
		atomic_inc(&oldmm->mm_count);
		enter_lazy_tlb(oldmm, next);
	} else
		switch_mm(oldmm, mm, next);

2006
	if (!prev->mm) {
L
Linus Torvalds 已提交
2007 2008 2009
		prev->active_mm = NULL;
		rq->prev_mm = oldmm;
	}
2010 2011 2012 2013 2014 2015 2016
	/*
	 * 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
2017
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
2018
#endif
L
Linus Torvalds 已提交
2019 2020 2021 2022

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

I
Ingo Molnar 已提交
2023 2024 2025 2026 2027 2028 2029
	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
2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046
}

/*
 * 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;
2047
}
L
Linus Torvalds 已提交
2048 2049

unsigned long nr_uninterruptible(void)
2050
{
L
Linus Torvalds 已提交
2051
	unsigned long i, sum = 0;
2052

2053
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2054
		sum += cpu_rq(i)->nr_uninterruptible;
2055 2056

	/*
L
Linus Torvalds 已提交
2057 2058
	 * Since we read the counters lockless, it might be slightly
	 * inaccurate. Do not allow it to go below zero though:
2059
	 */
L
Linus Torvalds 已提交
2060 2061
	if (unlikely((long)sum < 0))
		sum = 0;
2062

L
Linus Torvalds 已提交
2063
	return sum;
2064 2065
}

L
Linus Torvalds 已提交
2066
unsigned long long nr_context_switches(void)
2067
{
2068 2069
	int i;
	unsigned long long sum = 0;
2070

2071
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2072
		sum += cpu_rq(i)->nr_switches;
2073

L
Linus Torvalds 已提交
2074 2075
	return sum;
}
2076

L
Linus Torvalds 已提交
2077 2078 2079
unsigned long nr_iowait(void)
{
	unsigned long i, sum = 0;
2080

2081
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2082
		sum += atomic_read(&cpu_rq(i)->nr_iowait);
2083

L
Linus Torvalds 已提交
2084 2085
	return sum;
}
2086

2087
unsigned long nr_iowait_cpu(int cpu)
2088
{
2089
	struct rq *this = cpu_rq(cpu);
2090 2091
	return atomic_read(&this->nr_iowait);
}
2092

2093 2094 2095 2096 2097
unsigned long this_cpu_load(void)
{
	struct rq *this = this_rq();
	return this->cpu_load[0];
}
2098

2099

2100 2101 2102 2103 2104
/* 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);
2105

2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120
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;
}

2121 2122 2123 2124 2125 2126 2127 2128 2129
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;
}

2130 2131 2132 2133 2134 2135 2136 2137
#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;

2138
void calc_load_account_idle(struct rq *this_rq)
2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158
{
	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;
}
2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 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

/**
 * 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.
	 */
}
2281
#else
2282
void calc_load_account_idle(struct rq *this_rq)
2283 2284 2285 2286 2287 2288 2289
{
}

static inline long calc_load_fold_idle(void)
{
	return 0;
}
2290 2291 2292 2293

static void calc_global_nohz(unsigned long ticks)
{
}
2294 2295
#endif

2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308
/**
 * 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;
2309 2310 2311
}

/*
2312 2313
 * calc_load - update the avenrun load estimates 10 ticks after the
 * CPUs have updated calc_load_tasks.
2314
 */
2315
void calc_global_load(unsigned long ticks)
2316
{
2317
	long active;
L
Linus Torvalds 已提交
2318

2319 2320 2321
	calc_global_nohz(ticks);

	if (time_before(jiffies, calc_load_update + 10))
2322
		return;
L
Linus Torvalds 已提交
2323

2324 2325
	active = atomic_long_read(&calc_load_tasks);
	active = active > 0 ? active * FIXED_1 : 0;
L
Linus Torvalds 已提交
2326

2327 2328 2329
	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
2330

2331 2332
	calc_load_update += LOAD_FREQ;
}
L
Linus Torvalds 已提交
2333

2334
/*
2335 2336
 * Called from update_cpu_load() to periodically update this CPU's
 * active count.
2337 2338 2339
 */
static void calc_load_account_active(struct rq *this_rq)
{
2340
	long delta;
2341

2342 2343
	if (time_before(jiffies, this_rq->calc_load_update))
		return;
2344

2345 2346 2347
	delta  = calc_load_fold_active(this_rq);
	delta += calc_load_fold_idle();
	if (delta)
2348
		atomic_long_add(delta, &calc_load_tasks);
2349 2350

	this_rq->calc_load_update += LOAD_FREQ;
2351 2352
}

2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419
/*
 * 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;
}

2420
/*
I
Ingo Molnar 已提交
2421
 * Update rq->cpu_load[] statistics. This function is usually called every
2422 2423
 * scheduler tick (TICK_NSEC). With tickless idle this will not be called
 * every tick. We fix it up based on jiffies.
2424
 */
2425
void update_cpu_load(struct rq *this_rq)
2426
{
2427
	unsigned long this_load = this_rq->load.weight;
2428 2429
	unsigned long curr_jiffies = jiffies;
	unsigned long pending_updates;
I
Ingo Molnar 已提交
2430
	int i, scale;
2431

I
Ingo Molnar 已提交
2432
	this_rq->nr_load_updates++;
2433

2434 2435 2436 2437 2438 2439 2440
	/* 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
2441
	/* Update our load: */
2442 2443
	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
2444
		unsigned long old_load, new_load;
2445

I
Ingo Molnar 已提交
2446
		/* scale is effectively 1 << i now, and >> i divides by scale */
2447

I
Ingo Molnar 已提交
2448
		old_load = this_rq->cpu_load[i];
2449
		old_load = decay_load_missed(old_load, pending_updates - 1, i);
I
Ingo Molnar 已提交
2450
		new_load = this_load;
I
Ingo Molnar 已提交
2451 2452 2453 2454 2455 2456
		/*
		 * 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)
2457 2458 2459
			new_load += scale - 1;

		this_rq->cpu_load[i] = (old_load * (scale - 1) + new_load) >> i;
I
Ingo Molnar 已提交
2460
	}
2461 2462

	sched_avg_update(this_rq);
2463 2464 2465 2466 2467
}

static void update_cpu_load_active(struct rq *this_rq)
{
	update_cpu_load(this_rq);
2468

2469
	calc_load_account_active(this_rq);
2470 2471
}

I
Ingo Molnar 已提交
2472
#ifdef CONFIG_SMP
2473

2474
/*
P
Peter Zijlstra 已提交
2475 2476
 * sched_exec - execve() is a valuable balancing opportunity, because at
 * this point the task has the smallest effective memory and cache footprint.
2477
 */
P
Peter Zijlstra 已提交
2478
void sched_exec(void)
2479
{
P
Peter Zijlstra 已提交
2480
	struct task_struct *p = current;
L
Linus Torvalds 已提交
2481
	unsigned long flags;
2482
	int dest_cpu;
2483

2484
	raw_spin_lock_irqsave(&p->pi_lock, flags);
2485
	dest_cpu = p->sched_class->select_task_rq(p, SD_BALANCE_EXEC, 0);
2486 2487
	if (dest_cpu == smp_processor_id())
		goto unlock;
P
Peter Zijlstra 已提交
2488

2489
	if (likely(cpu_active(dest_cpu))) {
2490
		struct migration_arg arg = { p, dest_cpu };
2491

2492 2493
		raw_spin_unlock_irqrestore(&p->pi_lock, flags);
		stop_one_cpu(task_cpu(p), migration_cpu_stop, &arg);
L
Linus Torvalds 已提交
2494 2495
		return;
	}
2496
unlock:
2497
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
2498
}
I
Ingo Molnar 已提交
2499

L
Linus Torvalds 已提交
2500 2501 2502
#endif

DEFINE_PER_CPU(struct kernel_stat, kstat);
2503
DEFINE_PER_CPU(struct kernel_cpustat, kernel_cpustat);
L
Linus Torvalds 已提交
2504 2505

EXPORT_PER_CPU_SYMBOL(kstat);
2506
EXPORT_PER_CPU_SYMBOL(kernel_cpustat);
L
Linus Torvalds 已提交
2507 2508

/*
2509
 * Return any ns on the sched_clock that have not yet been accounted in
2510
 * @p in case that task is currently running.
2511 2512
 *
 * Called with task_rq_lock() held on @rq.
L
Linus Torvalds 已提交
2513
 */
2514 2515 2516 2517 2518 2519
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);
2520
		ns = rq->clock_task - p->se.exec_start;
2521 2522 2523 2524 2525 2526 2527
		if ((s64)ns < 0)
			ns = 0;
	}

	return ns;
}

2528
unsigned long long task_delta_exec(struct task_struct *p)
L
Linus Torvalds 已提交
2529 2530
{
	unsigned long flags;
2531
	struct rq *rq;
2532
	u64 ns = 0;
2533

2534
	rq = task_rq_lock(p, &flags);
2535
	ns = do_task_delta_exec(p, rq);
2536
	task_rq_unlock(rq, p, &flags);
2537

2538 2539
	return ns;
}
2540

2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553
/*
 * 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);
2554
	task_rq_unlock(rq, p, &flags);
2555 2556 2557

	return ns;
}
2558

L
Linus Torvalds 已提交
2559 2560 2561 2562
/*
 * 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
2563
 * @cputime_scaled: cputime scaled by cpu frequency
L
Linus Torvalds 已提交
2564
 */
2565 2566
void account_user_time(struct task_struct *p, cputime_t cputime,
		       cputime_t cputime_scaled)
L
Linus Torvalds 已提交
2567
{
2568 2569 2570
	u64 *cpustat = kcpustat_this_cpu->cpustat;
	u64 tmp;
	int index;
L
Linus Torvalds 已提交
2571

2572
	/* Add user time to process. */
L
Linus Torvalds 已提交
2573
	p->utime = cputime_add(p->utime, cputime);
2574
	p->utimescaled = cputime_add(p->utimescaled, cputime_scaled);
2575
	account_group_user_time(p, cputime);
L
Linus Torvalds 已提交
2576 2577 2578

	/* Add user time to cpustat. */
	tmp = cputime_to_cputime64(cputime);
2579 2580 2581

	index = (TASK_NICE(p) > 0) ? CPUTIME_NICE : CPUTIME_USER;
	cpustat[index] += tmp;
2582 2583

	cpuacct_update_stats(p, CPUACCT_STAT_USER, cputime);
2584 2585
	/* Account for user time used */
	acct_update_integrals(p);
L
Linus Torvalds 已提交
2586 2587
}

2588 2589 2590 2591
/*
 * 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
2592
 * @cputime_scaled: cputime scaled by cpu frequency
2593
 */
2594 2595
static void account_guest_time(struct task_struct *p, cputime_t cputime,
			       cputime_t cputime_scaled)
2596
{
2597 2598
	u64 tmp;
	u64 *cpustat = kcpustat_this_cpu->cpustat;
2599 2600 2601

	tmp = cputime_to_cputime64(cputime);

2602
	/* Add guest time to process. */
2603
	p->utime = cputime_add(p->utime, cputime);
2604
	p->utimescaled = cputime_add(p->utimescaled, cputime_scaled);
2605
	account_group_user_time(p, cputime);
2606 2607
	p->gtime = cputime_add(p->gtime, cputime);

2608
	/* Add guest time to cpustat. */
2609
	if (TASK_NICE(p) > 0) {
2610 2611
		cpustat[CPUTIME_NICE] += tmp;
		cpustat[CPUTIME_GUEST_NICE] += tmp;
2612
	} else {
2613 2614
		cpustat[CPUTIME_USER] += tmp;
		cpustat[CPUTIME_GUEST] += tmp;
2615
	}
2616 2617
}

2618 2619 2620 2621 2622 2623 2624 2625 2626
/*
 * 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,
2627
			cputime_t cputime_scaled, int index)
2628
{
2629 2630
	u64 tmp = cputime_to_cputime64(cputime);
	u64 *cpustat = kcpustat_this_cpu->cpustat;
2631 2632 2633 2634 2635 2636 2637

	/* Add system time to process. */
	p->stime = cputime_add(p->stime, cputime);
	p->stimescaled = cputime_add(p->stimescaled, cputime_scaled);
	account_group_system_time(p, cputime);

	/* Add system time to cpustat. */
2638
	cpustat[index] += tmp;
2639 2640 2641 2642 2643 2644
	cpuacct_update_stats(p, CPUACCT_STAT_SYSTEM, cputime);

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

L
Linus Torvalds 已提交
2645 2646 2647 2648 2649
/*
 * 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
2650
 * @cputime_scaled: cputime scaled by cpu frequency
L
Linus Torvalds 已提交
2651 2652
 */
void account_system_time(struct task_struct *p, int hardirq_offset,
2653
			 cputime_t cputime, cputime_t cputime_scaled)
L
Linus Torvalds 已提交
2654
{
2655
	int index;
L
Linus Torvalds 已提交
2656

2657
	if ((p->flags & PF_VCPU) && (irq_count() - hardirq_offset == 0)) {
2658
		account_guest_time(p, cputime, cputime_scaled);
2659 2660
		return;
	}
2661

L
Linus Torvalds 已提交
2662
	if (hardirq_count() - hardirq_offset)
2663
		index = CPUTIME_IRQ;
2664
	else if (in_serving_softirq())
2665
		index = CPUTIME_SOFTIRQ;
L
Linus Torvalds 已提交
2666
	else
2667
		index = CPUTIME_SYSTEM;
2668

2669
	__account_system_time(p, cputime, cputime_scaled, index);
L
Linus Torvalds 已提交
2670 2671
}

2672
/*
L
Linus Torvalds 已提交
2673
 * Account for involuntary wait time.
2674
 * @cputime: the cpu time spent in involuntary wait
2675
 */
2676
void account_steal_time(cputime_t cputime)
2677
{
2678 2679
	u64 *cpustat = kcpustat_this_cpu->cpustat;
	u64 cputime64 = cputime_to_cputime64(cputime);
2680

2681
	cpustat[CPUTIME_STEAL] += cputime64;
2682 2683
}

L
Linus Torvalds 已提交
2684
/*
2685 2686
 * Account for idle time.
 * @cputime: the cpu time spent in idle wait
L
Linus Torvalds 已提交
2687
 */
2688
void account_idle_time(cputime_t cputime)
L
Linus Torvalds 已提交
2689
{
2690 2691
	u64 *cpustat = kcpustat_this_cpu->cpustat;
	u64 cputime64 = cputime_to_cputime64(cputime);
2692
	struct rq *rq = this_rq();
L
Linus Torvalds 已提交
2693

2694
	if (atomic_read(&rq->nr_iowait) > 0)
2695
		cpustat[CPUTIME_IOWAIT] += cputime64;
2696
	else
2697
		cpustat[CPUTIME_IDLE] += cputime64;
L
Linus Torvalds 已提交
2698 2699
}

G
Glauber Costa 已提交
2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718
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;
}

2719 2720
#ifndef CONFIG_VIRT_CPU_ACCOUNTING

2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746
#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);
2747 2748
	u64 tmp = cputime_to_cputime64(cputime_one_jiffy);
	u64 *cpustat = kcpustat_this_cpu->cpustat;
2749

G
Glauber Costa 已提交
2750 2751 2752
	if (steal_account_process_tick())
		return;

2753
	if (irqtime_account_hi_update()) {
2754
		cpustat[CPUTIME_IRQ] += tmp;
2755
	} else if (irqtime_account_si_update()) {
2756
		cpustat[CPUTIME_SOFTIRQ] += tmp;
2757 2758 2759 2760 2761 2762 2763
	} 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,
2764
					CPUTIME_SOFTIRQ);
2765 2766 2767 2768 2769 2770 2771 2772
	} 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,
2773
					CPUTIME_SYSTEM);
2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784
	}
}

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);
}
2785
#else /* CONFIG_IRQ_TIME_ACCOUNTING */
2786 2787 2788
static void irqtime_account_idle_ticks(int ticks) {}
static void irqtime_account_process_tick(struct task_struct *p, int user_tick,
						struct rq *rq) {}
2789
#endif /* CONFIG_IRQ_TIME_ACCOUNTING */
2790 2791 2792 2793 2794 2795 2796 2797

/*
 * 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)
{
2798
	cputime_t one_jiffy_scaled = cputime_to_scaled(cputime_one_jiffy);
2799 2800
	struct rq *rq = this_rq();

2801 2802 2803 2804 2805
	if (sched_clock_irqtime) {
		irqtime_account_process_tick(p, user_tick, rq);
		return;
	}

G
Glauber Costa 已提交
2806 2807 2808
	if (steal_account_process_tick())
		return;

2809
	if (user_tick)
2810
		account_user_time(p, cputime_one_jiffy, one_jiffy_scaled);
2811
	else if ((p != rq->idle) || (irq_count() != HARDIRQ_OFFSET))
2812
		account_system_time(p, HARDIRQ_OFFSET, cputime_one_jiffy,
2813 2814
				    one_jiffy_scaled);
	else
2815
		account_idle_time(cputime_one_jiffy);
2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833
}

/*
 * 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)
{
2834 2835 2836 2837 2838 2839

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

2840
	account_idle_time(jiffies_to_cputime(ticks));
L
Linus Torvalds 已提交
2841 2842
}

2843 2844
#endif

2845 2846 2847 2848
/*
 * Use precise platform statistics if available:
 */
#ifdef CONFIG_VIRT_CPU_ACCOUNTING
2849
void task_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
2850
{
2851 2852
	*ut = p->utime;
	*st = p->stime;
2853 2854
}

2855
void thread_group_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
2856
{
2857 2858 2859 2860 2861 2862
	struct task_cputime cputime;

	thread_group_cputime(p, &cputime);

	*ut = cputime.utime;
	*st = cputime.stime;
2863 2864
}
#else
2865 2866

#ifndef nsecs_to_cputime
2867
# define nsecs_to_cputime(__nsecs)	nsecs_to_jiffies(__nsecs)
2868 2869
#endif

2870
void task_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
2871
{
2872
	cputime_t rtime, utime = p->utime, total = cputime_add(utime, p->stime);
2873 2874 2875 2876

	/*
	 * Use CFS's precise accounting:
	 */
2877
	rtime = nsecs_to_cputime(p->se.sum_exec_runtime);
2878 2879

	if (total) {
2880
		u64 temp = rtime;
2881

2882
		temp *= utime;
2883
		do_div(temp, total);
2884 2885 2886
		utime = (cputime_t)temp;
	} else
		utime = rtime;
2887

2888 2889 2890
	/*
	 * Compare with previous values, to keep monotonicity:
	 */
2891
	p->prev_utime = max(p->prev_utime, utime);
2892
	p->prev_stime = max(p->prev_stime, cputime_sub(rtime, p->prev_utime));
2893

2894 2895
	*ut = p->prev_utime;
	*st = p->prev_stime;
2896 2897
}

2898 2899 2900 2901
/*
 * Must be called with siglock held.
 */
void thread_group_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
2902
{
2903 2904 2905
	struct signal_struct *sig = p->signal;
	struct task_cputime cputime;
	cputime_t rtime, utime, total;
2906

2907
	thread_group_cputime(p, &cputime);
2908

2909 2910
	total = cputime_add(cputime.utime, cputime.stime);
	rtime = nsecs_to_cputime(cputime.sum_exec_runtime);
2911

2912
	if (total) {
2913
		u64 temp = rtime;
2914

2915
		temp *= cputime.utime;
2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926
		do_div(temp, total);
		utime = (cputime_t)temp;
	} else
		utime = rtime;

	sig->prev_utime = max(sig->prev_utime, utime);
	sig->prev_stime = max(sig->prev_stime,
			      cputime_sub(rtime, sig->prev_utime));

	*ut = sig->prev_utime;
	*st = sig->prev_stime;
2927 2928 2929
}
#endif

2930 2931 2932 2933 2934 2935 2936 2937
/*
 * 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
2938
	struct task_struct *curr = rq->curr;
2939 2940

	sched_clock_tick();
I
Ingo Molnar 已提交
2941

2942
	raw_spin_lock(&rq->lock);
2943
	update_rq_clock(rq);
2944
	update_cpu_load_active(rq);
P
Peter Zijlstra 已提交
2945
	curr->sched_class->task_tick(rq, curr, 0);
2946
	raw_spin_unlock(&rq->lock);
2947

2948
	perf_event_task_tick();
2949

2950
#ifdef CONFIG_SMP
2951
	rq->idle_balance = idle_cpu(cpu);
I
Ingo Molnar 已提交
2952
	trigger_load_balance(rq, cpu);
2953
#endif
L
Linus Torvalds 已提交
2954 2955
}

2956
notrace unsigned long get_parent_ip(unsigned long addr)
2957 2958 2959 2960 2961 2962 2963 2964
{
	if (in_lock_functions(addr)) {
		addr = CALLER_ADDR2;
		if (in_lock_functions(addr))
			addr = CALLER_ADDR3;
	}
	return addr;
}
L
Linus Torvalds 已提交
2965

2966 2967 2968
#if defined(CONFIG_PREEMPT) && (defined(CONFIG_DEBUG_PREEMPT) || \
				defined(CONFIG_PREEMPT_TRACER))

2969
void __kprobes add_preempt_count(int val)
L
Linus Torvalds 已提交
2970
{
2971
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
2972 2973 2974
	/*
	 * Underflow?
	 */
2975 2976
	if (DEBUG_LOCKS_WARN_ON((preempt_count() < 0)))
		return;
2977
#endif
L
Linus Torvalds 已提交
2978
	preempt_count() += val;
2979
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
2980 2981 2982
	/*
	 * Spinlock count overflowing soon?
	 */
2983 2984
	DEBUG_LOCKS_WARN_ON((preempt_count() & PREEMPT_MASK) >=
				PREEMPT_MASK - 10);
2985 2986 2987
#endif
	if (preempt_count() == val)
		trace_preempt_off(CALLER_ADDR0, get_parent_ip(CALLER_ADDR1));
L
Linus Torvalds 已提交
2988 2989 2990
}
EXPORT_SYMBOL(add_preempt_count);

2991
void __kprobes sub_preempt_count(int val)
L
Linus Torvalds 已提交
2992
{
2993
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
2994 2995 2996
	/*
	 * Underflow?
	 */
2997
	if (DEBUG_LOCKS_WARN_ON(val > preempt_count()))
2998
		return;
L
Linus Torvalds 已提交
2999 3000 3001
	/*
	 * Is the spinlock portion underflowing?
	 */
3002 3003 3004
	if (DEBUG_LOCKS_WARN_ON((val < PREEMPT_MASK) &&
			!(preempt_count() & PREEMPT_MASK)))
		return;
3005
#endif
3006

3007 3008
	if (preempt_count() == val)
		trace_preempt_on(CALLER_ADDR0, get_parent_ip(CALLER_ADDR1));
L
Linus Torvalds 已提交
3009 3010 3011 3012 3013 3014 3015
	preempt_count() -= val;
}
EXPORT_SYMBOL(sub_preempt_count);

#endif

/*
I
Ingo Molnar 已提交
3016
 * Print scheduling while atomic bug:
L
Linus Torvalds 已提交
3017
 */
I
Ingo Molnar 已提交
3018
static noinline void __schedule_bug(struct task_struct *prev)
L
Linus Torvalds 已提交
3019
{
3020 3021
	struct pt_regs *regs = get_irq_regs();

P
Peter Zijlstra 已提交
3022 3023
	printk(KERN_ERR "BUG: scheduling while atomic: %s/%d/0x%08x\n",
		prev->comm, prev->pid, preempt_count());
3024

I
Ingo Molnar 已提交
3025
	debug_show_held_locks(prev);
3026
	print_modules();
I
Ingo Molnar 已提交
3027 3028
	if (irqs_disabled())
		print_irqtrace_events(prev);
3029 3030 3031 3032 3033

	if (regs)
		show_regs(regs);
	else
		dump_stack();
I
Ingo Molnar 已提交
3034
}
L
Linus Torvalds 已提交
3035

I
Ingo Molnar 已提交
3036 3037 3038 3039 3040
/*
 * Various schedule()-time debugging checks and statistics:
 */
static inline void schedule_debug(struct task_struct *prev)
{
L
Linus Torvalds 已提交
3041
	/*
I
Ingo Molnar 已提交
3042
	 * Test if we are atomic. Since do_exit() needs to call into
L
Linus Torvalds 已提交
3043 3044 3045
	 * schedule() atomically, we ignore that path for now.
	 * Otherwise, whine if we are scheduling when we should not be.
	 */
3046
	if (unlikely(in_atomic_preempt_off() && !prev->exit_state))
I
Ingo Molnar 已提交
3047
		__schedule_bug(prev);
3048
	rcu_sleep_check();
I
Ingo Molnar 已提交
3049

L
Linus Torvalds 已提交
3050 3051
	profile_hit(SCHED_PROFILING, __builtin_return_address(0));

3052
	schedstat_inc(this_rq(), sched_count);
I
Ingo Molnar 已提交
3053 3054
}

P
Peter Zijlstra 已提交
3055
static void put_prev_task(struct rq *rq, struct task_struct *prev)
M
Mike Galbraith 已提交
3056
{
3057
	if (prev->on_rq || rq->skip_clock_update < 0)
3058
		update_rq_clock(rq);
P
Peter Zijlstra 已提交
3059
	prev->sched_class->put_prev_task(rq, prev);
M
Mike Galbraith 已提交
3060 3061
}

I
Ingo Molnar 已提交
3062 3063 3064 3065
/*
 * Pick up the highest-prio task:
 */
static inline struct task_struct *
3066
pick_next_task(struct rq *rq)
I
Ingo Molnar 已提交
3067
{
3068
	const struct sched_class *class;
I
Ingo Molnar 已提交
3069
	struct task_struct *p;
L
Linus Torvalds 已提交
3070 3071

	/*
I
Ingo Molnar 已提交
3072 3073
	 * Optimization: we know that if all tasks are in
	 * the fair class we can call that function directly:
L
Linus Torvalds 已提交
3074
	 */
3075
	if (likely(rq->nr_running == rq->cfs.h_nr_running)) {
3076
		p = fair_sched_class.pick_next_task(rq);
I
Ingo Molnar 已提交
3077 3078
		if (likely(p))
			return p;
L
Linus Torvalds 已提交
3079 3080
	}

3081
	for_each_class(class) {
3082
		p = class->pick_next_task(rq);
I
Ingo Molnar 已提交
3083 3084 3085
		if (p)
			return p;
	}
3086 3087

	BUG(); /* the idle class will always have a runnable task */
I
Ingo Molnar 已提交
3088
}
L
Linus Torvalds 已提交
3089

I
Ingo Molnar 已提交
3090
/*
3091
 * __schedule() is the main scheduler function.
I
Ingo Molnar 已提交
3092
 */
3093
static void __sched __schedule(void)
I
Ingo Molnar 已提交
3094 3095
{
	struct task_struct *prev, *next;
3096
	unsigned long *switch_count;
I
Ingo Molnar 已提交
3097
	struct rq *rq;
3098
	int cpu;
I
Ingo Molnar 已提交
3099

3100 3101
need_resched:
	preempt_disable();
I
Ingo Molnar 已提交
3102 3103
	cpu = smp_processor_id();
	rq = cpu_rq(cpu);
3104
	rcu_note_context_switch(cpu);
I
Ingo Molnar 已提交
3105 3106 3107
	prev = rq->curr;

	schedule_debug(prev);
L
Linus Torvalds 已提交
3108

3109
	if (sched_feat(HRTICK))
M
Mike Galbraith 已提交
3110
		hrtick_clear(rq);
P
Peter Zijlstra 已提交
3111

3112
	raw_spin_lock_irq(&rq->lock);
L
Linus Torvalds 已提交
3113

3114
	switch_count = &prev->nivcsw;
L
Linus Torvalds 已提交
3115
	if (prev->state && !(preempt_count() & PREEMPT_ACTIVE)) {
T
Tejun Heo 已提交
3116
		if (unlikely(signal_pending_state(prev->state, prev))) {
L
Linus Torvalds 已提交
3117
			prev->state = TASK_RUNNING;
T
Tejun Heo 已提交
3118
		} else {
3119 3120 3121
			deactivate_task(rq, prev, DEQUEUE_SLEEP);
			prev->on_rq = 0;

T
Tejun Heo 已提交
3122
			/*
3123 3124 3125
			 * 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
3126 3127 3128 3129 3130 3131 3132 3133 3134
			 */
			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
3135
		switch_count = &prev->nvcsw;
L
Linus Torvalds 已提交
3136 3137
	}

3138
	pre_schedule(rq, prev);
3139

I
Ingo Molnar 已提交
3140
	if (unlikely(!rq->nr_running))
L
Linus Torvalds 已提交
3141 3142
		idle_balance(cpu, rq);

M
Mike Galbraith 已提交
3143
	put_prev_task(rq, prev);
3144
	next = pick_next_task(rq);
3145 3146
	clear_tsk_need_resched(prev);
	rq->skip_clock_update = 0;
L
Linus Torvalds 已提交
3147 3148 3149 3150 3151 3152

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

I
Ingo Molnar 已提交
3153
		context_switch(rq, prev, next); /* unlocks the rq */
P
Peter Zijlstra 已提交
3154
		/*
3155 3156 3157 3158
		 * 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
3159 3160 3161
		 */
		cpu = smp_processor_id();
		rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
3162
	} else
3163
		raw_spin_unlock_irq(&rq->lock);
L
Linus Torvalds 已提交
3164

3165
	post_schedule(rq);
L
Linus Torvalds 已提交
3166 3167

	preempt_enable_no_resched();
3168
	if (need_resched())
L
Linus Torvalds 已提交
3169 3170
		goto need_resched;
}
3171

3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183
static inline void sched_submit_work(struct task_struct *tsk)
{
	if (!tsk->state)
		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
3184
asmlinkage void __sched schedule(void)
3185
{
3186 3187 3188
	struct task_struct *tsk = current;

	sched_submit_work(tsk);
3189 3190
	__schedule();
}
L
Linus Torvalds 已提交
3191 3192
EXPORT_SYMBOL(schedule);

3193
#ifdef CONFIG_MUTEX_SPIN_ON_OWNER
3194

3195 3196 3197
static inline bool owner_running(struct mutex *lock, struct task_struct *owner)
{
	if (lock->owner != owner)
3198
		return false;
3199 3200

	/*
3201 3202 3203 3204
	 * 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.
3205
	 */
3206
	barrier();
3207

3208
	return owner->on_cpu;
3209
}
3210

3211 3212 3213 3214 3215 3216 3217 3218
/*
 * 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;
3219

3220
	rcu_read_lock();
3221 3222
	while (owner_running(lock, owner)) {
		if (need_resched())
3223
			break;
3224

3225
		arch_mutex_cpu_relax();
3226
	}
3227
	rcu_read_unlock();
3228

3229
	/*
3230 3231 3232
	 * 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.
3233
	 */
3234
	return lock->owner == NULL;
3235 3236 3237
}
#endif

L
Linus Torvalds 已提交
3238 3239
#ifdef CONFIG_PREEMPT
/*
3240
 * this is the entry point to schedule() from in-kernel preemption
I
Ingo Molnar 已提交
3241
 * off of preempt_enable. Kernel preemptions off return from interrupt
L
Linus Torvalds 已提交
3242 3243
 * occur there and call schedule directly.
 */
3244
asmlinkage void __sched notrace preempt_schedule(void)
L
Linus Torvalds 已提交
3245 3246
{
	struct thread_info *ti = current_thread_info();
3247

L
Linus Torvalds 已提交
3248 3249
	/*
	 * If there is a non-zero preempt_count or interrupts are disabled,
I
Ingo Molnar 已提交
3250
	 * we do not want to preempt the current task. Just return..
L
Linus Torvalds 已提交
3251
	 */
N
Nick Piggin 已提交
3252
	if (likely(ti->preempt_count || irqs_disabled()))
L
Linus Torvalds 已提交
3253 3254
		return;

3255
	do {
3256
		add_preempt_count_notrace(PREEMPT_ACTIVE);
3257
		__schedule();
3258
		sub_preempt_count_notrace(PREEMPT_ACTIVE);
L
Linus Torvalds 已提交
3259

3260 3261 3262 3263 3264
		/*
		 * Check again in case we missed a preemption opportunity
		 * between schedule and now.
		 */
		barrier();
3265
	} while (need_resched());
L
Linus Torvalds 已提交
3266 3267 3268 3269
}
EXPORT_SYMBOL(preempt_schedule);

/*
3270
 * this is the entry point to schedule() from kernel preemption
L
Linus Torvalds 已提交
3271 3272 3273 3274 3275 3276 3277
 * 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();
3278

3279
	/* Catch callers which need to be fixed */
L
Linus Torvalds 已提交
3280 3281
	BUG_ON(ti->preempt_count || !irqs_disabled());

3282 3283 3284
	do {
		add_preempt_count(PREEMPT_ACTIVE);
		local_irq_enable();
3285
		__schedule();
3286 3287
		local_irq_disable();
		sub_preempt_count(PREEMPT_ACTIVE);
L
Linus Torvalds 已提交
3288

3289 3290 3291 3292 3293
		/*
		 * Check again in case we missed a preemption opportunity
		 * between schedule and now.
		 */
		barrier();
3294
	} while (need_resched());
L
Linus Torvalds 已提交
3295 3296 3297 3298
}

#endif /* CONFIG_PREEMPT */

P
Peter Zijlstra 已提交
3299
int default_wake_function(wait_queue_t *curr, unsigned mode, int wake_flags,
I
Ingo Molnar 已提交
3300
			  void *key)
L
Linus Torvalds 已提交
3301
{
P
Peter Zijlstra 已提交
3302
	return try_to_wake_up(curr->private, mode, wake_flags);
L
Linus Torvalds 已提交
3303 3304 3305 3306
}
EXPORT_SYMBOL(default_wake_function);

/*
I
Ingo Molnar 已提交
3307 3308
 * 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
3309 3310 3311
 * 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
3312
 * started to run but is not in state TASK_RUNNING. try_to_wake_up() returns
L
Linus Torvalds 已提交
3313 3314
 * zero in this (rare) case, and we handle it by continuing to scan the queue.
 */
3315
static void __wake_up_common(wait_queue_head_t *q, unsigned int mode,
P
Peter Zijlstra 已提交
3316
			int nr_exclusive, int wake_flags, void *key)
L
Linus Torvalds 已提交
3317
{
3318
	wait_queue_t *curr, *next;
L
Linus Torvalds 已提交
3319

3320
	list_for_each_entry_safe(curr, next, &q->task_list, task_list) {
3321 3322
		unsigned flags = curr->flags;

P
Peter Zijlstra 已提交
3323
		if (curr->func(curr, mode, wake_flags, key) &&
3324
				(flags & WQ_FLAG_EXCLUSIVE) && !--nr_exclusive)
L
Linus Torvalds 已提交
3325 3326 3327 3328 3329 3330 3331 3332 3333
			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
3334
 * @key: is directly passed to the wakeup function
3335 3336 3337
 *
 * 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
3338
 */
3339
void __wake_up(wait_queue_head_t *q, unsigned int mode,
I
Ingo Molnar 已提交
3340
			int nr_exclusive, void *key)
L
Linus Torvalds 已提交
3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352
{
	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.
 */
3353
void __wake_up_locked(wait_queue_head_t *q, unsigned int mode)
L
Linus Torvalds 已提交
3354 3355 3356
{
	__wake_up_common(q, mode, 1, 0, NULL);
}
3357
EXPORT_SYMBOL_GPL(__wake_up_locked);
L
Linus Torvalds 已提交
3358

3359 3360 3361 3362
void __wake_up_locked_key(wait_queue_head_t *q, unsigned int mode, void *key)
{
	__wake_up_common(q, mode, 1, 0, key);
}
3363
EXPORT_SYMBOL_GPL(__wake_up_locked_key);
3364

L
Linus Torvalds 已提交
3365
/**
3366
 * __wake_up_sync_key - wake up threads blocked on a waitqueue.
L
Linus Torvalds 已提交
3367 3368 3369
 * @q: the waitqueue
 * @mode: which threads
 * @nr_exclusive: how many wake-one or wake-many threads to wake up
3370
 * @key: opaque value to be passed to wakeup targets
L
Linus Torvalds 已提交
3371 3372 3373 3374 3375 3376 3377
 *
 * 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.
3378 3379 3380
 *
 * 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
3381
 */
3382 3383
void __wake_up_sync_key(wait_queue_head_t *q, unsigned int mode,
			int nr_exclusive, void *key)
L
Linus Torvalds 已提交
3384 3385
{
	unsigned long flags;
P
Peter Zijlstra 已提交
3386
	int wake_flags = WF_SYNC;
L
Linus Torvalds 已提交
3387 3388 3389 3390 3391

	if (unlikely(!q))
		return;

	if (unlikely(!nr_exclusive))
P
Peter Zijlstra 已提交
3392
		wake_flags = 0;
L
Linus Torvalds 已提交
3393 3394

	spin_lock_irqsave(&q->lock, flags);
P
Peter Zijlstra 已提交
3395
	__wake_up_common(q, mode, nr_exclusive, wake_flags, key);
L
Linus Torvalds 已提交
3396 3397
	spin_unlock_irqrestore(&q->lock, flags);
}
3398 3399 3400 3401 3402 3403 3404 3405 3406
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
3407 3408
EXPORT_SYMBOL_GPL(__wake_up_sync);	/* For internal use only */

3409 3410 3411 3412 3413 3414 3415 3416
/**
 * 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.
3417 3418 3419
 *
 * 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.
3420
 */
3421
void complete(struct completion *x)
L
Linus Torvalds 已提交
3422 3423 3424 3425 3426
{
	unsigned long flags;

	spin_lock_irqsave(&x->wait.lock, flags);
	x->done++;
3427
	__wake_up_common(&x->wait, TASK_NORMAL, 1, 0, NULL);
L
Linus Torvalds 已提交
3428 3429 3430 3431
	spin_unlock_irqrestore(&x->wait.lock, flags);
}
EXPORT_SYMBOL(complete);

3432 3433 3434 3435 3436
/**
 * 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.
3437 3438 3439
 *
 * 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.
3440
 */
3441
void complete_all(struct completion *x)
L
Linus Torvalds 已提交
3442 3443 3444 3445 3446
{
	unsigned long flags;

	spin_lock_irqsave(&x->wait.lock, flags);
	x->done += UINT_MAX/2;
3447
	__wake_up_common(&x->wait, TASK_NORMAL, 0, 0, NULL);
L
Linus Torvalds 已提交
3448 3449 3450 3451
	spin_unlock_irqrestore(&x->wait.lock, flags);
}
EXPORT_SYMBOL(complete_all);

3452 3453
static inline long __sched
do_wait_for_common(struct completion *x, long timeout, int state)
L
Linus Torvalds 已提交
3454 3455 3456 3457
{
	if (!x->done) {
		DECLARE_WAITQUEUE(wait, current);

C
Changli Gao 已提交
3458
		__add_wait_queue_tail_exclusive(&x->wait, &wait);
L
Linus Torvalds 已提交
3459
		do {
3460
			if (signal_pending_state(state, current)) {
3461 3462
				timeout = -ERESTARTSYS;
				break;
3463 3464
			}
			__set_current_state(state);
L
Linus Torvalds 已提交
3465 3466 3467
			spin_unlock_irq(&x->wait.lock);
			timeout = schedule_timeout(timeout);
			spin_lock_irq(&x->wait.lock);
3468
		} while (!x->done && timeout);
L
Linus Torvalds 已提交
3469
		__remove_wait_queue(&x->wait, &wait);
3470 3471
		if (!x->done)
			return timeout;
L
Linus Torvalds 已提交
3472 3473
	}
	x->done--;
3474
	return timeout ?: 1;
L
Linus Torvalds 已提交
3475 3476
}

3477 3478
static long __sched
wait_for_common(struct completion *x, long timeout, int state)
L
Linus Torvalds 已提交
3479 3480 3481 3482
{
	might_sleep();

	spin_lock_irq(&x->wait.lock);
3483
	timeout = do_wait_for_common(x, timeout, state);
L
Linus Torvalds 已提交
3484
	spin_unlock_irq(&x->wait.lock);
3485 3486
	return timeout;
}
L
Linus Torvalds 已提交
3487

3488 3489 3490 3491 3492 3493 3494 3495 3496 3497
/**
 * 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().
 */
3498
void __sched wait_for_completion(struct completion *x)
3499 3500
{
	wait_for_common(x, MAX_SCHEDULE_TIMEOUT, TASK_UNINTERRUPTIBLE);
L
Linus Torvalds 已提交
3501
}
3502
EXPORT_SYMBOL(wait_for_completion);
L
Linus Torvalds 已提交
3503

3504 3505 3506 3507 3508 3509 3510 3511
/**
 * 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.
3512 3513 3514
 *
 * The return value is 0 if timed out, and positive (at least 1, or number of
 * jiffies left till timeout) if completed.
3515
 */
3516
unsigned long __sched
3517
wait_for_completion_timeout(struct completion *x, unsigned long timeout)
L
Linus Torvalds 已提交
3518
{
3519
	return wait_for_common(x, timeout, TASK_UNINTERRUPTIBLE);
L
Linus Torvalds 已提交
3520
}
3521
EXPORT_SYMBOL(wait_for_completion_timeout);
L
Linus Torvalds 已提交
3522

3523 3524 3525 3526 3527 3528
/**
 * 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.
3529 3530
 *
 * The return value is -ERESTARTSYS if interrupted, 0 if completed.
3531
 */
3532
int __sched wait_for_completion_interruptible(struct completion *x)
I
Ingo Molnar 已提交
3533
{
3534 3535 3536 3537
	long t = wait_for_common(x, MAX_SCHEDULE_TIMEOUT, TASK_INTERRUPTIBLE);
	if (t == -ERESTARTSYS)
		return t;
	return 0;
I
Ingo Molnar 已提交
3538
}
3539
EXPORT_SYMBOL(wait_for_completion_interruptible);
L
Linus Torvalds 已提交
3540

3541 3542 3543 3544 3545 3546 3547
/**
 * 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.
3548 3549 3550
 *
 * The return value is -ERESTARTSYS if interrupted, 0 if timed out,
 * positive (at least 1, or number of jiffies left till timeout) if completed.
3551
 */
3552
long __sched
3553 3554
wait_for_completion_interruptible_timeout(struct completion *x,
					  unsigned long timeout)
I
Ingo Molnar 已提交
3555
{
3556
	return wait_for_common(x, timeout, TASK_INTERRUPTIBLE);
I
Ingo Molnar 已提交
3557
}
3558
EXPORT_SYMBOL(wait_for_completion_interruptible_timeout);
L
Linus Torvalds 已提交
3559

3560 3561 3562 3563 3564 3565
/**
 * 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.
3566 3567
 *
 * The return value is -ERESTARTSYS if interrupted, 0 if completed.
3568
 */
M
Matthew Wilcox 已提交
3569 3570 3571 3572 3573 3574 3575 3576 3577
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);

3578 3579 3580 3581 3582 3583 3584 3585
/**
 * 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.
3586 3587 3588
 *
 * The return value is -ERESTARTSYS if interrupted, 0 if timed out,
 * positive (at least 1, or number of jiffies left till timeout) if completed.
3589
 */
3590
long __sched
3591 3592 3593 3594 3595 3596 3597
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);

3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611
/**
 *	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)
{
3612
	unsigned long flags;
3613 3614
	int ret = 1;

3615
	spin_lock_irqsave(&x->wait.lock, flags);
3616 3617 3618 3619
	if (!x->done)
		ret = 0;
	else
		x->done--;
3620
	spin_unlock_irqrestore(&x->wait.lock, flags);
3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634
	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)
{
3635
	unsigned long flags;
3636 3637
	int ret = 1;

3638
	spin_lock_irqsave(&x->wait.lock, flags);
3639 3640
	if (!x->done)
		ret = 0;
3641
	spin_unlock_irqrestore(&x->wait.lock, flags);
3642 3643 3644 3645
	return ret;
}
EXPORT_SYMBOL(completion_done);

3646 3647
static long __sched
sleep_on_common(wait_queue_head_t *q, int state, long timeout)
L
Linus Torvalds 已提交
3648
{
I
Ingo Molnar 已提交
3649 3650 3651 3652
	unsigned long flags;
	wait_queue_t wait;

	init_waitqueue_entry(&wait, current);
L
Linus Torvalds 已提交
3653

3654
	__set_current_state(state);
L
Linus Torvalds 已提交
3655

3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669
	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
3670 3671 3672
}
EXPORT_SYMBOL(interruptible_sleep_on);

I
Ingo Molnar 已提交
3673
long __sched
I
Ingo Molnar 已提交
3674
interruptible_sleep_on_timeout(wait_queue_head_t *q, long timeout)
L
Linus Torvalds 已提交
3675
{
3676
	return sleep_on_common(q, TASK_INTERRUPTIBLE, timeout);
L
Linus Torvalds 已提交
3677 3678 3679
}
EXPORT_SYMBOL(interruptible_sleep_on_timeout);

I
Ingo Molnar 已提交
3680
void __sched sleep_on(wait_queue_head_t *q)
L
Linus Torvalds 已提交
3681
{
3682
	sleep_on_common(q, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
L
Linus Torvalds 已提交
3683 3684 3685
}
EXPORT_SYMBOL(sleep_on);

I
Ingo Molnar 已提交
3686
long __sched sleep_on_timeout(wait_queue_head_t *q, long timeout)
L
Linus Torvalds 已提交
3687
{
3688
	return sleep_on_common(q, TASK_UNINTERRUPTIBLE, timeout);
L
Linus Torvalds 已提交
3689 3690 3691
}
EXPORT_SYMBOL(sleep_on_timeout);

3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703
#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.
 */
3704
void rt_mutex_setprio(struct task_struct *p, int prio)
3705
{
3706
	int oldprio, on_rq, running;
3707
	struct rq *rq;
3708
	const struct sched_class *prev_class;
3709 3710 3711

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

3712
	rq = __task_rq_lock(p);
3713

3714
	trace_sched_pi_setprio(p, prio);
3715
	oldprio = p->prio;
3716
	prev_class = p->sched_class;
P
Peter Zijlstra 已提交
3717
	on_rq = p->on_rq;
3718
	running = task_current(rq, p);
3719
	if (on_rq)
3720
		dequeue_task(rq, p, 0);
3721 3722
	if (running)
		p->sched_class->put_prev_task(rq, p);
I
Ingo Molnar 已提交
3723 3724 3725 3726 3727 3728

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

3729 3730
	p->prio = prio;

3731 3732
	if (running)
		p->sched_class->set_curr_task(rq);
P
Peter Zijlstra 已提交
3733
	if (on_rq)
3734
		enqueue_task(rq, p, oldprio < prio ? ENQUEUE_HEAD : 0);
3735

P
Peter Zijlstra 已提交
3736
	check_class_changed(rq, p, prev_class, oldprio);
3737
	__task_rq_unlock(rq);
3738 3739 3740 3741
}

#endif

3742
void set_user_nice(struct task_struct *p, long nice)
L
Linus Torvalds 已提交
3743
{
I
Ingo Molnar 已提交
3744
	int old_prio, delta, on_rq;
L
Linus Torvalds 已提交
3745
	unsigned long flags;
3746
	struct rq *rq;
L
Linus Torvalds 已提交
3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758

	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
3759
	 * SCHED_FIFO/SCHED_RR:
L
Linus Torvalds 已提交
3760
	 */
3761
	if (task_has_rt_policy(p)) {
L
Linus Torvalds 已提交
3762 3763 3764
		p->static_prio = NICE_TO_PRIO(nice);
		goto out_unlock;
	}
P
Peter Zijlstra 已提交
3765
	on_rq = p->on_rq;
3766
	if (on_rq)
3767
		dequeue_task(rq, p, 0);
L
Linus Torvalds 已提交
3768 3769

	p->static_prio = NICE_TO_PRIO(nice);
3770
	set_load_weight(p);
3771 3772 3773
	old_prio = p->prio;
	p->prio = effective_prio(p);
	delta = p->prio - old_prio;
L
Linus Torvalds 已提交
3774

I
Ingo Molnar 已提交
3775
	if (on_rq) {
3776
		enqueue_task(rq, p, 0);
L
Linus Torvalds 已提交
3777
		/*
3778 3779
		 * If the task increased its priority or is running and
		 * lowered its priority, then reschedule its CPU:
L
Linus Torvalds 已提交
3780
		 */
3781
		if (delta < 0 || (delta > 0 && task_running(rq, p)))
L
Linus Torvalds 已提交
3782 3783 3784
			resched_task(rq->curr);
	}
out_unlock:
3785
	task_rq_unlock(rq, p, &flags);
L
Linus Torvalds 已提交
3786 3787 3788
}
EXPORT_SYMBOL(set_user_nice);

M
Matt Mackall 已提交
3789 3790 3791 3792 3793
/*
 * can_nice - check if a task can reduce its nice value
 * @p: task
 * @nice: nice value
 */
3794
int can_nice(const struct task_struct *p, const int nice)
M
Matt Mackall 已提交
3795
{
3796 3797
	/* convert nice value [19,-20] to rlimit style value [1,40] */
	int nice_rlim = 20 - nice;
3798

3799
	return (nice_rlim <= task_rlimit(p, RLIMIT_NICE) ||
M
Matt Mackall 已提交
3800 3801 3802
		capable(CAP_SYS_NICE));
}

L
Linus Torvalds 已提交
3803 3804 3805 3806 3807 3808 3809 3810 3811
#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.
 */
3812
SYSCALL_DEFINE1(nice, int, increment)
L
Linus Torvalds 已提交
3813
{
3814
	long nice, retval;
L
Linus Torvalds 已提交
3815 3816 3817 3818 3819 3820

	/*
	 * 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
3821 3822
	if (increment < -40)
		increment = -40;
L
Linus Torvalds 已提交
3823 3824 3825
	if (increment > 40)
		increment = 40;

3826
	nice = TASK_NICE(current) + increment;
L
Linus Torvalds 已提交
3827 3828 3829 3830 3831
	if (nice < -20)
		nice = -20;
	if (nice > 19)
		nice = 19;

M
Matt Mackall 已提交
3832 3833 3834
	if (increment < 0 && !can_nice(current, nice))
		return -EPERM;

L
Linus Torvalds 已提交
3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852
	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.
 */
3853
int task_prio(const struct task_struct *p)
L
Linus Torvalds 已提交
3854 3855 3856 3857 3858 3859 3860 3861
{
	return p->prio - MAX_RT_PRIO;
}

/**
 * task_nice - return the nice value of a given task.
 * @p: the task in question.
 */
3862
int task_nice(const struct task_struct *p)
L
Linus Torvalds 已提交
3863 3864 3865
{
	return TASK_NICE(p);
}
P
Pavel Roskin 已提交
3866
EXPORT_SYMBOL(task_nice);
L
Linus Torvalds 已提交
3867 3868 3869 3870 3871 3872 3873

/**
 * idle_cpu - is a given cpu idle currently?
 * @cpu: the processor in question.
 */
int idle_cpu(int cpu)
{
T
Thomas Gleixner 已提交
3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887
	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
3888 3889 3890 3891 3892 3893
}

/**
 * idle_task - return the idle task for a given cpu.
 * @cpu: the processor in question.
 */
3894
struct task_struct *idle_task(int cpu)
L
Linus Torvalds 已提交
3895 3896 3897 3898 3899 3900 3901 3902
{
	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
3903
static struct task_struct *find_process_by_pid(pid_t pid)
L
Linus Torvalds 已提交
3904
{
3905
	return pid ? find_task_by_vpid(pid) : current;
L
Linus Torvalds 已提交
3906 3907 3908
}

/* Actually do priority change: must hold rq lock. */
I
Ingo Molnar 已提交
3909 3910
static void
__setscheduler(struct rq *rq, struct task_struct *p, int policy, int prio)
L
Linus Torvalds 已提交
3911 3912 3913
{
	p->policy = policy;
	p->rt_priority = prio;
3914 3915 3916
	p->normal_prio = normal_prio(p);
	/* we are holding p->pi_lock already */
	p->prio = rt_mutex_getprio(p);
3917 3918 3919 3920
	if (rt_prio(p->prio))
		p->sched_class = &rt_sched_class;
	else
		p->sched_class = &fair_sched_class;
3921
	set_load_weight(p);
L
Linus Torvalds 已提交
3922 3923
}

3924 3925 3926 3927 3928 3929 3930 3931 3932 3933
/*
 * 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);
3934 3935 3936 3937 3938
	if (cred->user->user_ns == pcred->user->user_ns)
		match = (cred->euid == pcred->euid ||
			 cred->euid == pcred->uid);
	else
		match = false;
3939 3940 3941 3942
	rcu_read_unlock();
	return match;
}

3943
static int __sched_setscheduler(struct task_struct *p, int policy,
3944
				const struct sched_param *param, bool user)
L
Linus Torvalds 已提交
3945
{
3946
	int retval, oldprio, oldpolicy = -1, on_rq, running;
L
Linus Torvalds 已提交
3947
	unsigned long flags;
3948
	const struct sched_class *prev_class;
3949
	struct rq *rq;
3950
	int reset_on_fork;
L
Linus Torvalds 已提交
3951

3952 3953
	/* may grab non-irq protected spin_locks */
	BUG_ON(in_interrupt());
L
Linus Torvalds 已提交
3954 3955
recheck:
	/* double check policy once rq lock held */
3956 3957
	if (policy < 0) {
		reset_on_fork = p->sched_reset_on_fork;
L
Linus Torvalds 已提交
3958
		policy = oldpolicy = p->policy;
3959 3960 3961 3962 3963 3964 3965 3966 3967 3968
	} 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
3969 3970
	/*
	 * Valid priorities for SCHED_FIFO and SCHED_RR are
I
Ingo Molnar 已提交
3971 3972
	 * 1..MAX_USER_RT_PRIO-1, valid priority for SCHED_NORMAL,
	 * SCHED_BATCH and SCHED_IDLE is 0.
L
Linus Torvalds 已提交
3973 3974
	 */
	if (param->sched_priority < 0 ||
I
Ingo Molnar 已提交
3975
	    (p->mm && param->sched_priority > MAX_USER_RT_PRIO-1) ||
3976
	    (!p->mm && param->sched_priority > MAX_RT_PRIO-1))
L
Linus Torvalds 已提交
3977
		return -EINVAL;
3978
	if (rt_policy(policy) != (param->sched_priority != 0))
L
Linus Torvalds 已提交
3979 3980
		return -EINVAL;

3981 3982 3983
	/*
	 * Allow unprivileged RT tasks to decrease priority:
	 */
3984
	if (user && !capable(CAP_SYS_NICE)) {
3985
		if (rt_policy(policy)) {
3986 3987
			unsigned long rlim_rtprio =
					task_rlimit(p, RLIMIT_RTPRIO);
3988 3989 3990 3991 3992 3993 3994 3995 3996 3997

			/* 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;
		}
3998

I
Ingo Molnar 已提交
3999
		/*
4000 4001
		 * 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
4002
		 */
4003 4004 4005 4006
		if (p->policy == SCHED_IDLE && policy != SCHED_IDLE) {
			if (!can_nice(p, TASK_NICE(p)))
				return -EPERM;
		}
4007

4008
		/* can't change other user's priorities */
4009
		if (!check_same_owner(p))
4010
			return -EPERM;
4011 4012 4013 4014

		/* Normal users shall not reset the sched_reset_on_fork flag */
		if (p->sched_reset_on_fork && !reset_on_fork)
			return -EPERM;
4015
	}
L
Linus Torvalds 已提交
4016

4017
	if (user) {
4018
		retval = security_task_setscheduler(p);
4019 4020 4021 4022
		if (retval)
			return retval;
	}

4023 4024 4025
	/*
	 * make sure no PI-waiters arrive (or leave) while we are
	 * changing the priority of the task:
4026
	 *
L
Lucas De Marchi 已提交
4027
	 * To be able to change p->policy safely, the appropriate
L
Linus Torvalds 已提交
4028 4029
	 * runqueue lock must be held.
	 */
4030
	rq = task_rq_lock(p, &flags);
4031

4032 4033 4034 4035
	/*
	 * Changing the policy of the stop threads its a very bad idea
	 */
	if (p == rq->stop) {
4036
		task_rq_unlock(rq, p, &flags);
4037 4038 4039
		return -EINVAL;
	}

4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050
	/*
	 * 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;
	}

4051 4052 4053 4054 4055 4056 4057
#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) &&
4058 4059
				task_group(p)->rt_bandwidth.rt_runtime == 0 &&
				!task_group_is_autogroup(task_group(p))) {
4060
			task_rq_unlock(rq, p, &flags);
4061 4062 4063 4064 4065
			return -EPERM;
		}
	}
#endif

L
Linus Torvalds 已提交
4066 4067 4068
	/* recheck policy now with rq lock held */
	if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) {
		policy = oldpolicy = -1;
4069
		task_rq_unlock(rq, p, &flags);
L
Linus Torvalds 已提交
4070 4071
		goto recheck;
	}
P
Peter Zijlstra 已提交
4072
	on_rq = p->on_rq;
4073
	running = task_current(rq, p);
4074
	if (on_rq)
4075
		deactivate_task(rq, p, 0);
4076 4077
	if (running)
		p->sched_class->put_prev_task(rq, p);
4078

4079 4080
	p->sched_reset_on_fork = reset_on_fork;

L
Linus Torvalds 已提交
4081
	oldprio = p->prio;
4082
	prev_class = p->sched_class;
I
Ingo Molnar 已提交
4083
	__setscheduler(rq, p, policy, param->sched_priority);
4084

4085 4086
	if (running)
		p->sched_class->set_curr_task(rq);
P
Peter Zijlstra 已提交
4087
	if (on_rq)
I
Ingo Molnar 已提交
4088
		activate_task(rq, p, 0);
4089

P
Peter Zijlstra 已提交
4090
	check_class_changed(rq, p, prev_class, oldprio);
4091
	task_rq_unlock(rq, p, &flags);
4092

4093 4094
	rt_mutex_adjust_pi(p);

L
Linus Torvalds 已提交
4095 4096
	return 0;
}
4097 4098 4099 4100 4101 4102 4103 4104 4105 4106

/**
 * 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,
4107
		       const struct sched_param *param)
4108 4109 4110
{
	return __sched_setscheduler(p, policy, param, true);
}
L
Linus Torvalds 已提交
4111 4112
EXPORT_SYMBOL_GPL(sched_setscheduler);

4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124
/**
 * 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,
4125
			       const struct sched_param *param)
4126 4127 4128 4129
{
	return __sched_setscheduler(p, policy, param, false);
}

I
Ingo Molnar 已提交
4130 4131
static int
do_sched_setscheduler(pid_t pid, int policy, struct sched_param __user *param)
L
Linus Torvalds 已提交
4132 4133 4134
{
	struct sched_param lparam;
	struct task_struct *p;
4135
	int retval;
L
Linus Torvalds 已提交
4136 4137 4138 4139 4140

	if (!param || pid < 0)
		return -EINVAL;
	if (copy_from_user(&lparam, param, sizeof(struct sched_param)))
		return -EFAULT;
4141 4142 4143

	rcu_read_lock();
	retval = -ESRCH;
L
Linus Torvalds 已提交
4144
	p = find_process_by_pid(pid);
4145 4146 4147
	if (p != NULL)
		retval = sched_setscheduler(p, policy, &lparam);
	rcu_read_unlock();
4148

L
Linus Torvalds 已提交
4149 4150 4151 4152 4153 4154 4155 4156 4157
	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.
 */
4158 4159
SYSCALL_DEFINE3(sched_setscheduler, pid_t, pid, int, policy,
		struct sched_param __user *, param)
L
Linus Torvalds 已提交
4160
{
4161 4162 4163 4164
	/* negative values for policy are not valid */
	if (policy < 0)
		return -EINVAL;

L
Linus Torvalds 已提交
4165 4166 4167 4168 4169 4170 4171 4172
	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.
 */
4173
SYSCALL_DEFINE2(sched_setparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4174 4175 4176 4177 4178 4179 4180 4181
{
	return do_sched_setscheduler(pid, -1, param);
}

/**
 * sys_sched_getscheduler - get the policy (scheduling class) of a thread
 * @pid: the pid in question.
 */
4182
SYSCALL_DEFINE1(sched_getscheduler, pid_t, pid)
L
Linus Torvalds 已提交
4183
{
4184
	struct task_struct *p;
4185
	int retval;
L
Linus Torvalds 已提交
4186 4187

	if (pid < 0)
4188
		return -EINVAL;
L
Linus Torvalds 已提交
4189 4190

	retval = -ESRCH;
4191
	rcu_read_lock();
L
Linus Torvalds 已提交
4192 4193 4194 4195
	p = find_process_by_pid(pid);
	if (p) {
		retval = security_task_getscheduler(p);
		if (!retval)
4196 4197
			retval = p->policy
				| (p->sched_reset_on_fork ? SCHED_RESET_ON_FORK : 0);
L
Linus Torvalds 已提交
4198
	}
4199
	rcu_read_unlock();
L
Linus Torvalds 已提交
4200 4201 4202 4203
	return retval;
}

/**
4204
 * sys_sched_getparam - get the RT priority of a thread
L
Linus Torvalds 已提交
4205 4206 4207
 * @pid: the pid in question.
 * @param: structure containing the RT priority.
 */
4208
SYSCALL_DEFINE2(sched_getparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4209 4210
{
	struct sched_param lp;
4211
	struct task_struct *p;
4212
	int retval;
L
Linus Torvalds 已提交
4213 4214

	if (!param || pid < 0)
4215
		return -EINVAL;
L
Linus Torvalds 已提交
4216

4217
	rcu_read_lock();
L
Linus Torvalds 已提交
4218 4219 4220 4221 4222 4223 4224 4225 4226 4227
	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;
4228
	rcu_read_unlock();
L
Linus Torvalds 已提交
4229 4230 4231 4232 4233 4234 4235 4236 4237

	/*
	 * 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:
4238
	rcu_read_unlock();
L
Linus Torvalds 已提交
4239 4240 4241
	return retval;
}

4242
long sched_setaffinity(pid_t pid, const struct cpumask *in_mask)
L
Linus Torvalds 已提交
4243
{
4244
	cpumask_var_t cpus_allowed, new_mask;
4245 4246
	struct task_struct *p;
	int retval;
L
Linus Torvalds 已提交
4247

4248
	get_online_cpus();
4249
	rcu_read_lock();
L
Linus Torvalds 已提交
4250 4251 4252

	p = find_process_by_pid(pid);
	if (!p) {
4253
		rcu_read_unlock();
4254
		put_online_cpus();
L
Linus Torvalds 已提交
4255 4256 4257
		return -ESRCH;
	}

4258
	/* Prevent p going away */
L
Linus Torvalds 已提交
4259
	get_task_struct(p);
4260
	rcu_read_unlock();
L
Linus Torvalds 已提交
4261

4262 4263 4264 4265 4266 4267 4268 4269
	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
4270
	retval = -EPERM;
4271
	if (!check_same_owner(p) && !task_ns_capable(p, CAP_SYS_NICE))
L
Linus Torvalds 已提交
4272 4273
		goto out_unlock;

4274
	retval = security_task_setscheduler(p);
4275 4276 4277
	if (retval)
		goto out_unlock;

4278 4279
	cpuset_cpus_allowed(p, cpus_allowed);
	cpumask_and(new_mask, in_mask, cpus_allowed);
P
Peter Zijlstra 已提交
4280
again:
4281
	retval = set_cpus_allowed_ptr(p, new_mask);
L
Linus Torvalds 已提交
4282

P
Paul Menage 已提交
4283
	if (!retval) {
4284 4285
		cpuset_cpus_allowed(p, cpus_allowed);
		if (!cpumask_subset(new_mask, cpus_allowed)) {
P
Paul Menage 已提交
4286 4287 4288 4289 4290
			/*
			 * We must have raced with a concurrent cpuset
			 * update. Just reset the cpus_allowed to the
			 * cpuset's cpus_allowed
			 */
4291
			cpumask_copy(new_mask, cpus_allowed);
P
Paul Menage 已提交
4292 4293 4294
			goto again;
		}
	}
L
Linus Torvalds 已提交
4295
out_unlock:
4296 4297 4298 4299
	free_cpumask_var(new_mask);
out_free_cpus_allowed:
	free_cpumask_var(cpus_allowed);
out_put_task:
L
Linus Torvalds 已提交
4300
	put_task_struct(p);
4301
	put_online_cpus();
L
Linus Torvalds 已提交
4302 4303 4304 4305
	return retval;
}

static int get_user_cpu_mask(unsigned long __user *user_mask_ptr, unsigned len,
4306
			     struct cpumask *new_mask)
L
Linus Torvalds 已提交
4307
{
4308 4309 4310 4311 4312
	if (len < cpumask_size())
		cpumask_clear(new_mask);
	else if (len > cpumask_size())
		len = cpumask_size();

L
Linus Torvalds 已提交
4313 4314 4315 4316 4317 4318 4319 4320 4321
	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
 */
4322 4323
SYSCALL_DEFINE3(sched_setaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
4324
{
4325
	cpumask_var_t new_mask;
L
Linus Torvalds 已提交
4326 4327
	int retval;

4328 4329
	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
4330

4331 4332 4333 4334 4335
	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
4336 4337
}

4338
long sched_getaffinity(pid_t pid, struct cpumask *mask)
L
Linus Torvalds 已提交
4339
{
4340
	struct task_struct *p;
4341
	unsigned long flags;
L
Linus Torvalds 已提交
4342 4343
	int retval;

4344
	get_online_cpus();
4345
	rcu_read_lock();
L
Linus Torvalds 已提交
4346 4347 4348 4349 4350 4351

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

4352 4353 4354 4355
	retval = security_task_getscheduler(p);
	if (retval)
		goto out_unlock;

4356
	raw_spin_lock_irqsave(&p->pi_lock, flags);
4357
	cpumask_and(mask, &p->cpus_allowed, cpu_online_mask);
4358
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
4359 4360

out_unlock:
4361
	rcu_read_unlock();
4362
	put_online_cpus();
L
Linus Torvalds 已提交
4363

4364
	return retval;
L
Linus Torvalds 已提交
4365 4366 4367 4368 4369 4370 4371 4372
}

/**
 * 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
 */
4373 4374
SYSCALL_DEFINE3(sched_getaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
4375 4376
{
	int ret;
4377
	cpumask_var_t mask;
L
Linus Torvalds 已提交
4378

A
Anton Blanchard 已提交
4379
	if ((len * BITS_PER_BYTE) < nr_cpu_ids)
4380 4381
		return -EINVAL;
	if (len & (sizeof(unsigned long)-1))
L
Linus Torvalds 已提交
4382 4383
		return -EINVAL;

4384 4385
	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
4386

4387 4388
	ret = sched_getaffinity(pid, mask);
	if (ret == 0) {
4389
		size_t retlen = min_t(size_t, len, cpumask_size());
4390 4391

		if (copy_to_user(user_mask_ptr, mask, retlen))
4392 4393
			ret = -EFAULT;
		else
4394
			ret = retlen;
4395 4396
	}
	free_cpumask_var(mask);
L
Linus Torvalds 已提交
4397

4398
	return ret;
L
Linus Torvalds 已提交
4399 4400 4401 4402 4403
}

/**
 * sys_sched_yield - yield the current processor to other threads.
 *
I
Ingo Molnar 已提交
4404 4405
 * 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
4406
 */
4407
SYSCALL_DEFINE0(sched_yield)
L
Linus Torvalds 已提交
4408
{
4409
	struct rq *rq = this_rq_lock();
L
Linus Torvalds 已提交
4410

4411
	schedstat_inc(rq, yld_count);
4412
	current->sched_class->yield_task(rq);
L
Linus Torvalds 已提交
4413 4414 4415 4416 4417 4418

	/*
	 * Since we are going to call schedule() anyway, there's
	 * no need to preempt or enable interrupts:
	 */
	__release(rq->lock);
4419
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
4420
	do_raw_spin_unlock(&rq->lock);
L
Linus Torvalds 已提交
4421 4422 4423 4424 4425 4426 4427
	preempt_enable_no_resched();

	schedule();

	return 0;
}

P
Peter Zijlstra 已提交
4428 4429 4430 4431 4432
static inline int should_resched(void)
{
	return need_resched() && !(preempt_count() & PREEMPT_ACTIVE);
}

A
Andrew Morton 已提交
4433
static void __cond_resched(void)
L
Linus Torvalds 已提交
4434
{
4435
	add_preempt_count(PREEMPT_ACTIVE);
4436
	__schedule();
4437
	sub_preempt_count(PREEMPT_ACTIVE);
L
Linus Torvalds 已提交
4438 4439
}

4440
int __sched _cond_resched(void)
L
Linus Torvalds 已提交
4441
{
P
Peter Zijlstra 已提交
4442
	if (should_resched()) {
L
Linus Torvalds 已提交
4443 4444 4445 4446 4447
		__cond_resched();
		return 1;
	}
	return 0;
}
4448
EXPORT_SYMBOL(_cond_resched);
L
Linus Torvalds 已提交
4449 4450

/*
4451
 * __cond_resched_lock() - if a reschedule is pending, drop the given lock,
L
Linus Torvalds 已提交
4452 4453
 * call schedule, and on return reacquire the lock.
 *
I
Ingo Molnar 已提交
4454
 * This works OK both with and without CONFIG_PREEMPT. We do strange low-level
L
Linus Torvalds 已提交
4455 4456 4457
 * operations here to prevent schedule() from being called twice (once via
 * spin_unlock(), once by hand).
 */
4458
int __cond_resched_lock(spinlock_t *lock)
L
Linus Torvalds 已提交
4459
{
P
Peter Zijlstra 已提交
4460
	int resched = should_resched();
J
Jan Kara 已提交
4461 4462
	int ret = 0;

4463 4464
	lockdep_assert_held(lock);

N
Nick Piggin 已提交
4465
	if (spin_needbreak(lock) || resched) {
L
Linus Torvalds 已提交
4466
		spin_unlock(lock);
P
Peter Zijlstra 已提交
4467
		if (resched)
N
Nick Piggin 已提交
4468 4469 4470
			__cond_resched();
		else
			cpu_relax();
J
Jan Kara 已提交
4471
		ret = 1;
L
Linus Torvalds 已提交
4472 4473
		spin_lock(lock);
	}
J
Jan Kara 已提交
4474
	return ret;
L
Linus Torvalds 已提交
4475
}
4476
EXPORT_SYMBOL(__cond_resched_lock);
L
Linus Torvalds 已提交
4477

4478
int __sched __cond_resched_softirq(void)
L
Linus Torvalds 已提交
4479 4480 4481
{
	BUG_ON(!in_softirq());

P
Peter Zijlstra 已提交
4482
	if (should_resched()) {
4483
		local_bh_enable();
L
Linus Torvalds 已提交
4484 4485 4486 4487 4488 4489
		__cond_resched();
		local_bh_disable();
		return 1;
	}
	return 0;
}
4490
EXPORT_SYMBOL(__cond_resched_softirq);
L
Linus Torvalds 已提交
4491 4492 4493 4494

/**
 * yield - yield the current processor to other threads.
 *
4495
 * This is a shortcut for kernel-space yielding - it marks the
L
Linus Torvalds 已提交
4496 4497 4498 4499 4500 4501 4502 4503 4504
 * thread runnable and calls sys_sched_yield().
 */
void __sched yield(void)
{
	set_current_state(TASK_RUNNING);
	sys_sched_yield();
}
EXPORT_SYMBOL(yield);

4505 4506 4507 4508
/**
 * 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
4509 4510
 * @p: target task
 * @preempt: whether task preemption is allowed or not
4511 4512 4513 4514 4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536 4537 4538 4539 4540 4541 4542 4543 4544
 *
 * 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);
4545
	if (yielded) {
4546
		schedstat_inc(rq, yld_count);
4547 4548 4549 4550 4551 4552
		/*
		 * Make p's CPU reschedule; pick_next_entity takes care of
		 * fairness.
		 */
		if (preempt && rq != p_rq)
			resched_task(p_rq->curr);
4553 4554 4555 4556 4557 4558 4559
	} 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;
4560
	}
4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572

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
4573
/*
I
Ingo Molnar 已提交
4574
 * This task is about to go to sleep on IO. Increment rq->nr_iowait so
L
Linus Torvalds 已提交
4575 4576 4577 4578
 * that process accounting knows that this is a task in IO wait state.
 */
void __sched io_schedule(void)
{
4579
	struct rq *rq = raw_rq();
L
Linus Torvalds 已提交
4580

4581
	delayacct_blkio_start();
L
Linus Torvalds 已提交
4582
	atomic_inc(&rq->nr_iowait);
4583
	blk_flush_plug(current);
4584
	current->in_iowait = 1;
L
Linus Torvalds 已提交
4585
	schedule();
4586
	current->in_iowait = 0;
L
Linus Torvalds 已提交
4587
	atomic_dec(&rq->nr_iowait);
4588
	delayacct_blkio_end();
L
Linus Torvalds 已提交
4589 4590 4591 4592 4593
}
EXPORT_SYMBOL(io_schedule);

long __sched io_schedule_timeout(long timeout)
{
4594
	struct rq *rq = raw_rq();
L
Linus Torvalds 已提交
4595 4596
	long ret;

4597
	delayacct_blkio_start();
L
Linus Torvalds 已提交
4598
	atomic_inc(&rq->nr_iowait);
4599
	blk_flush_plug(current);
4600
	current->in_iowait = 1;
L
Linus Torvalds 已提交
4601
	ret = schedule_timeout(timeout);
4602
	current->in_iowait = 0;
L
Linus Torvalds 已提交
4603
	atomic_dec(&rq->nr_iowait);
4604
	delayacct_blkio_end();
L
Linus Torvalds 已提交
4605 4606 4607 4608 4609 4610 4611 4612 4613 4614
	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.
 */
4615
SYSCALL_DEFINE1(sched_get_priority_max, int, policy)
L
Linus Torvalds 已提交
4616 4617 4618 4619 4620 4621 4622 4623 4624
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = MAX_USER_RT_PRIO-1;
		break;
	case SCHED_NORMAL:
4625
	case SCHED_BATCH:
I
Ingo Molnar 已提交
4626
	case SCHED_IDLE:
L
Linus Torvalds 已提交
4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639
		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.
 */
4640
SYSCALL_DEFINE1(sched_get_priority_min, int, policy)
L
Linus Torvalds 已提交
4641 4642 4643 4644 4645 4646 4647 4648 4649
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = 1;
		break;
	case SCHED_NORMAL:
4650
	case SCHED_BATCH:
I
Ingo Molnar 已提交
4651
	case SCHED_IDLE:
L
Linus Torvalds 已提交
4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664
		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.
 */
4665
SYSCALL_DEFINE2(sched_rr_get_interval, pid_t, pid,
4666
		struct timespec __user *, interval)
L
Linus Torvalds 已提交
4667
{
4668
	struct task_struct *p;
D
Dmitry Adamushko 已提交
4669
	unsigned int time_slice;
4670 4671
	unsigned long flags;
	struct rq *rq;
4672
	int retval;
L
Linus Torvalds 已提交
4673 4674 4675
	struct timespec t;

	if (pid < 0)
4676
		return -EINVAL;
L
Linus Torvalds 已提交
4677 4678

	retval = -ESRCH;
4679
	rcu_read_lock();
L
Linus Torvalds 已提交
4680 4681 4682 4683 4684 4685 4686 4687
	p = find_process_by_pid(pid);
	if (!p)
		goto out_unlock;

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

4688 4689
	rq = task_rq_lock(p, &flags);
	time_slice = p->sched_class->get_rr_interval(rq, p);
4690
	task_rq_unlock(rq, p, &flags);
D
Dmitry Adamushko 已提交
4691

4692
	rcu_read_unlock();
D
Dmitry Adamushko 已提交
4693
	jiffies_to_timespec(time_slice, &t);
L
Linus Torvalds 已提交
4694 4695
	retval = copy_to_user(interval, &t, sizeof(t)) ? -EFAULT : 0;
	return retval;
4696

L
Linus Torvalds 已提交
4697
out_unlock:
4698
	rcu_read_unlock();
L
Linus Torvalds 已提交
4699 4700 4701
	return retval;
}

4702
static const char stat_nam[] = TASK_STATE_TO_CHAR_STR;
4703

4704
void sched_show_task(struct task_struct *p)
L
Linus Torvalds 已提交
4705 4706
{
	unsigned long free = 0;
4707
	unsigned state;
L
Linus Torvalds 已提交
4708 4709

	state = p->state ? __ffs(p->state) + 1 : 0;
4710
	printk(KERN_INFO "%-15.15s %c", p->comm,
4711
		state < sizeof(stat_nam) - 1 ? stat_nam[state] : '?');
4712
#if BITS_PER_LONG == 32
L
Linus Torvalds 已提交
4713
	if (state == TASK_RUNNING)
P
Peter Zijlstra 已提交
4714
		printk(KERN_CONT " running  ");
L
Linus Torvalds 已提交
4715
	else
P
Peter Zijlstra 已提交
4716
		printk(KERN_CONT " %08lx ", thread_saved_pc(p));
L
Linus Torvalds 已提交
4717 4718
#else
	if (state == TASK_RUNNING)
P
Peter Zijlstra 已提交
4719
		printk(KERN_CONT "  running task    ");
L
Linus Torvalds 已提交
4720
	else
P
Peter Zijlstra 已提交
4721
		printk(KERN_CONT " %016lx ", thread_saved_pc(p));
L
Linus Torvalds 已提交
4722 4723
#endif
#ifdef CONFIG_DEBUG_STACK_USAGE
4724
	free = stack_not_used(p);
L
Linus Torvalds 已提交
4725
#endif
P
Peter Zijlstra 已提交
4726
	printk(KERN_CONT "%5lu %5d %6d 0x%08lx\n", free,
4727 4728
		task_pid_nr(p), task_pid_nr(p->real_parent),
		(unsigned long)task_thread_info(p)->flags);
L
Linus Torvalds 已提交
4729

4730
	show_stack(p, NULL);
L
Linus Torvalds 已提交
4731 4732
}

I
Ingo Molnar 已提交
4733
void show_state_filter(unsigned long state_filter)
L
Linus Torvalds 已提交
4734
{
4735
	struct task_struct *g, *p;
L
Linus Torvalds 已提交
4736

4737
#if BITS_PER_LONG == 32
P
Peter Zijlstra 已提交
4738 4739
	printk(KERN_INFO
		"  task                PC stack   pid father\n");
L
Linus Torvalds 已提交
4740
#else
P
Peter Zijlstra 已提交
4741 4742
	printk(KERN_INFO
		"  task                        PC stack   pid father\n");
L
Linus Torvalds 已提交
4743
#endif
4744
	rcu_read_lock();
L
Linus Torvalds 已提交
4745 4746 4747
	do_each_thread(g, p) {
		/*
		 * reset the NMI-timeout, listing all files on a slow
L
Lucas De Marchi 已提交
4748
		 * console might take a lot of time:
L
Linus Torvalds 已提交
4749 4750
		 */
		touch_nmi_watchdog();
I
Ingo Molnar 已提交
4751
		if (!state_filter || (p->state & state_filter))
4752
			sched_show_task(p);
L
Linus Torvalds 已提交
4753 4754
	} while_each_thread(g, p);

4755 4756
	touch_all_softlockup_watchdogs();

I
Ingo Molnar 已提交
4757 4758 4759
#ifdef CONFIG_SCHED_DEBUG
	sysrq_sched_debug_show();
#endif
4760
	rcu_read_unlock();
I
Ingo Molnar 已提交
4761 4762 4763
	/*
	 * Only show locks if all tasks are dumped:
	 */
4764
	if (!state_filter)
I
Ingo Molnar 已提交
4765
		debug_show_all_locks();
L
Linus Torvalds 已提交
4766 4767
}

I
Ingo Molnar 已提交
4768 4769
void __cpuinit init_idle_bootup_task(struct task_struct *idle)
{
I
Ingo Molnar 已提交
4770
	idle->sched_class = &idle_sched_class;
I
Ingo Molnar 已提交
4771 4772
}

4773 4774 4775 4776 4777 4778 4779 4780
/**
 * 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.
 */
4781
void __cpuinit init_idle(struct task_struct *idle, int cpu)
L
Linus Torvalds 已提交
4782
{
4783
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
4784 4785
	unsigned long flags;

4786
	raw_spin_lock_irqsave(&rq->lock, flags);
4787

I
Ingo Molnar 已提交
4788
	__sched_fork(idle);
4789
	idle->state = TASK_RUNNING;
I
Ingo Molnar 已提交
4790 4791
	idle->se.exec_start = sched_clock();

4792
	do_set_cpus_allowed(idle, cpumask_of(cpu));
4793 4794 4795 4796 4797 4798 4799 4800 4801 4802 4803
	/*
	 * 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
4804
	__set_task_cpu(idle, cpu);
4805
	rcu_read_unlock();
L
Linus Torvalds 已提交
4806 4807

	rq->curr = rq->idle = idle;
P
Peter Zijlstra 已提交
4808 4809
#if defined(CONFIG_SMP)
	idle->on_cpu = 1;
4810
#endif
4811
	raw_spin_unlock_irqrestore(&rq->lock, flags);
L
Linus Torvalds 已提交
4812 4813

	/* Set the preempt count _outside_ the spinlocks! */
A
Al Viro 已提交
4814
	task_thread_info(idle)->preempt_count = 0;
J
Jonathan Corbet 已提交
4815

I
Ingo Molnar 已提交
4816 4817 4818 4819
	/*
	 * The idle tasks have their own, simple scheduling class:
	 */
	idle->sched_class = &idle_sched_class;
4820
	ftrace_graph_init_idle_task(idle, cpu);
4821 4822 4823
#if defined(CONFIG_SMP)
	sprintf(idle->comm, "%s/%d", INIT_TASK_COMM, cpu);
#endif
L
Linus Torvalds 已提交
4824 4825 4826
}

#ifdef CONFIG_SMP
4827 4828 4829 4830
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);
4831 4832 4833

	cpumask_copy(&p->cpus_allowed, new_mask);
	p->rt.nr_cpus_allowed = cpumask_weight(new_mask);
4834 4835
}

L
Linus Torvalds 已提交
4836 4837 4838
/*
 * This is how migration works:
 *
4839 4840 4841 4842 4843 4844
 * 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
4845
 *    it and puts it into the right queue.
4846 4847
 * 5) stopper completes and stop_one_cpu() returns and the migration
 *    is done.
L
Linus Torvalds 已提交
4848 4849 4850 4851 4852 4853 4854 4855
 */

/*
 * 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
4856
 * task must not exit() & deallocate itself prematurely. The
L
Linus Torvalds 已提交
4857 4858
 * call is not atomic; no spinlocks may be held.
 */
4859
int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
L
Linus Torvalds 已提交
4860 4861
{
	unsigned long flags;
4862
	struct rq *rq;
4863
	unsigned int dest_cpu;
4864
	int ret = 0;
L
Linus Torvalds 已提交
4865 4866

	rq = task_rq_lock(p, &flags);
4867

4868 4869 4870
	if (cpumask_equal(&p->cpus_allowed, new_mask))
		goto out;

4871
	if (!cpumask_intersects(new_mask, cpu_active_mask)) {
L
Linus Torvalds 已提交
4872 4873 4874 4875
		ret = -EINVAL;
		goto out;
	}

4876
	if (unlikely((p->flags & PF_THREAD_BOUND) && p != current)) {
4877 4878 4879 4880
		ret = -EINVAL;
		goto out;
	}

4881
	do_set_cpus_allowed(p, new_mask);
4882

L
Linus Torvalds 已提交
4883
	/* Can the task run on the task's current CPU? If so, we're done */
4884
	if (cpumask_test_cpu(task_cpu(p), new_mask))
L
Linus Torvalds 已提交
4885 4886
		goto out;

4887
	dest_cpu = cpumask_any_and(cpu_active_mask, new_mask);
4888
	if (p->on_rq) {
4889
		struct migration_arg arg = { p, dest_cpu };
L
Linus Torvalds 已提交
4890
		/* Need help from migration thread: drop lock and wait. */
4891
		task_rq_unlock(rq, p, &flags);
4892
		stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg);
L
Linus Torvalds 已提交
4893 4894 4895 4896
		tlb_migrate_finish(p->mm);
		return 0;
	}
out:
4897
	task_rq_unlock(rq, p, &flags);
4898

L
Linus Torvalds 已提交
4899 4900
	return ret;
}
4901
EXPORT_SYMBOL_GPL(set_cpus_allowed_ptr);
L
Linus Torvalds 已提交
4902 4903

/*
I
Ingo Molnar 已提交
4904
 * Move (not current) task off this cpu, onto dest cpu. We're doing
L
Linus Torvalds 已提交
4905 4906 4907 4908 4909 4910
 * 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.
4911 4912
 *
 * Returns non-zero if task was successfully migrated.
L
Linus Torvalds 已提交
4913
 */
4914
static int __migrate_task(struct task_struct *p, int src_cpu, int dest_cpu)
L
Linus Torvalds 已提交
4915
{
4916
	struct rq *rq_dest, *rq_src;
4917
	int ret = 0;
L
Linus Torvalds 已提交
4918

4919
	if (unlikely(!cpu_active(dest_cpu)))
4920
		return ret;
L
Linus Torvalds 已提交
4921 4922 4923 4924

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

4925
	raw_spin_lock(&p->pi_lock);
L
Linus Torvalds 已提交
4926 4927 4928
	double_rq_lock(rq_src, rq_dest);
	/* Already moved. */
	if (task_cpu(p) != src_cpu)
L
Linus Torvalds 已提交
4929
		goto done;
L
Linus Torvalds 已提交
4930
	/* Affinity changed (again). */
4931
	if (!cpumask_test_cpu(dest_cpu, tsk_cpus_allowed(p)))
L
Linus Torvalds 已提交
4932
		goto fail;
L
Linus Torvalds 已提交
4933

4934 4935 4936 4937
	/*
	 * If we're not on a rq, the next wake-up will ensure we're
	 * placed properly.
	 */
P
Peter Zijlstra 已提交
4938
	if (p->on_rq) {
4939
		deactivate_task(rq_src, p, 0);
4940
		set_task_cpu(p, dest_cpu);
I
Ingo Molnar 已提交
4941
		activate_task(rq_dest, p, 0);
4942
		check_preempt_curr(rq_dest, p, 0);
L
Linus Torvalds 已提交
4943
	}
L
Linus Torvalds 已提交
4944
done:
4945
	ret = 1;
L
Linus Torvalds 已提交
4946
fail:
L
Linus Torvalds 已提交
4947
	double_rq_unlock(rq_src, rq_dest);
4948
	raw_spin_unlock(&p->pi_lock);
4949
	return ret;
L
Linus Torvalds 已提交
4950 4951 4952
}

/*
4953 4954 4955
 * 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
4956
 */
4957
static int migration_cpu_stop(void *data)
L
Linus Torvalds 已提交
4958
{
4959
	struct migration_arg *arg = data;
4960

4961 4962 4963 4964
	/*
	 * The original target cpu might have gone down and we might
	 * be on another cpu but it doesn't matter.
	 */
4965
	local_irq_disable();
4966
	__migrate_task(arg->task, raw_smp_processor_id(), arg->dest_cpu);
4967
	local_irq_enable();
L
Linus Torvalds 已提交
4968
	return 0;
4969 4970
}

L
Linus Torvalds 已提交
4971
#ifdef CONFIG_HOTPLUG_CPU
4972

4973
/*
4974 4975
 * Ensures that the idle task is using init_mm right before its cpu goes
 * offline.
4976
 */
4977
void idle_task_exit(void)
L
Linus Torvalds 已提交
4978
{
4979
	struct mm_struct *mm = current->active_mm;
4980

4981
	BUG_ON(cpu_online(smp_processor_id()));
4982

4983 4984 4985
	if (mm != &init_mm)
		switch_mm(mm, &init_mm, current);
	mmdrop(mm);
L
Linus Torvalds 已提交
4986 4987 4988 4989 4990 4991 4992 4993 4994
}

/*
 * 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:
 */
4995
static void migrate_nr_uninterruptible(struct rq *rq_src)
L
Linus Torvalds 已提交
4996
{
4997
	struct rq *rq_dest = cpu_rq(cpumask_any(cpu_active_mask));
L
Linus Torvalds 已提交
4998 4999 5000 5001 5002

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

I
Ingo Molnar 已提交
5003
/*
5004
 * remove the tasks which were accounted by rq from calc_load_tasks.
L
Linus Torvalds 已提交
5005
 */
5006
static void calc_global_load_remove(struct rq *rq)
L
Linus Torvalds 已提交
5007
{
5008 5009
	atomic_long_sub(rq->calc_load_active, &calc_load_tasks);
	rq->calc_load_active = 0;
L
Linus Torvalds 已提交
5010 5011
}

5012
/*
5013 5014 5015 5016 5017 5018
 * 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
5019
 */
5020
static void migrate_tasks(unsigned int dead_cpu)
L
Linus Torvalds 已提交
5021
{
5022
	struct rq *rq = cpu_rq(dead_cpu);
5023 5024
	struct task_struct *next, *stop = rq->stop;
	int dest_cpu;
L
Linus Torvalds 已提交
5025 5026

	/*
5027 5028 5029 5030 5031 5032 5033
	 * 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
5034
	 */
5035
	rq->stop = NULL;
5036

5037 5038 5039
	/* Ensure any throttled groups are reachable by pick_next_task */
	unthrottle_offline_cfs_rqs(rq);

I
Ingo Molnar 已提交
5040
	for ( ; ; ) {
5041 5042 5043 5044 5045
		/*
		 * There's this thread running, bail when that's the only
		 * remaining thread.
		 */
		if (rq->nr_running == 1)
I
Ingo Molnar 已提交
5046
			break;
5047

5048
		next = pick_next_task(rq);
5049
		BUG_ON(!next);
D
Dmitry Adamushko 已提交
5050
		next->sched_class->put_prev_task(rq, next);
5051

5052 5053 5054 5055 5056 5057 5058
		/* 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
5059
	}
5060

5061
	rq->stop = stop;
5062
}
5063

L
Linus Torvalds 已提交
5064 5065
#endif /* CONFIG_HOTPLUG_CPU */

5066 5067 5068
#if defined(CONFIG_SCHED_DEBUG) && defined(CONFIG_SYSCTL)

static struct ctl_table sd_ctl_dir[] = {
5069 5070
	{
		.procname	= "sched_domain",
5071
		.mode		= 0555,
5072
	},
5073
	{}
5074 5075 5076
};

static struct ctl_table sd_ctl_root[] = {
5077 5078
	{
		.procname	= "kernel",
5079
		.mode		= 0555,
5080 5081
		.child		= sd_ctl_dir,
	},
5082
	{}
5083 5084 5085 5086 5087
};

static struct ctl_table *sd_alloc_ctl_entry(int n)
{
	struct ctl_table *entry =
5088
		kcalloc(n, sizeof(struct ctl_table), GFP_KERNEL);
5089 5090 5091 5092

	return entry;
}

5093 5094
static void sd_free_ctl_entry(struct ctl_table **tablep)
{
5095
	struct ctl_table *entry;
5096

5097 5098 5099
	/*
	 * 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
5100
	 * will always be set. In the lowest directory the names are
5101 5102 5103
	 * static strings and all have proc handlers.
	 */
	for (entry = *tablep; entry->mode; entry++) {
5104 5105
		if (entry->child)
			sd_free_ctl_entry(&entry->child);
5106 5107 5108
		if (entry->proc_handler == NULL)
			kfree(entry->procname);
	}
5109 5110 5111 5112 5113

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

5114
static void
5115
set_table_entry(struct ctl_table *entry,
5116 5117 5118 5119 5120 5121 5122 5123 5124 5125 5126 5127 5128
		const char *procname, void *data, int maxlen,
		mode_t mode, proc_handler *proc_handler)
{
	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)
{
5129
	struct ctl_table *table = sd_alloc_ctl_entry(13);
5130

5131 5132 5133
	if (table == NULL)
		return NULL;

5134
	set_table_entry(&table[0], "min_interval", &sd->min_interval,
5135
		sizeof(long), 0644, proc_doulongvec_minmax);
5136
	set_table_entry(&table[1], "max_interval", &sd->max_interval,
5137
		sizeof(long), 0644, proc_doulongvec_minmax);
5138
	set_table_entry(&table[2], "busy_idx", &sd->busy_idx,
5139
		sizeof(int), 0644, proc_dointvec_minmax);
5140
	set_table_entry(&table[3], "idle_idx", &sd->idle_idx,
5141
		sizeof(int), 0644, proc_dointvec_minmax);
5142
	set_table_entry(&table[4], "newidle_idx", &sd->newidle_idx,
5143
		sizeof(int), 0644, proc_dointvec_minmax);
5144
	set_table_entry(&table[5], "wake_idx", &sd->wake_idx,
5145
		sizeof(int), 0644, proc_dointvec_minmax);
5146
	set_table_entry(&table[6], "forkexec_idx", &sd->forkexec_idx,
5147
		sizeof(int), 0644, proc_dointvec_minmax);
5148
	set_table_entry(&table[7], "busy_factor", &sd->busy_factor,
5149
		sizeof(int), 0644, proc_dointvec_minmax);
5150
	set_table_entry(&table[8], "imbalance_pct", &sd->imbalance_pct,
5151
		sizeof(int), 0644, proc_dointvec_minmax);
5152
	set_table_entry(&table[9], "cache_nice_tries",
5153 5154
		&sd->cache_nice_tries,
		sizeof(int), 0644, proc_dointvec_minmax);
5155
	set_table_entry(&table[10], "flags", &sd->flags,
5156
		sizeof(int), 0644, proc_dointvec_minmax);
5157 5158 5159
	set_table_entry(&table[11], "name", sd->name,
		CORENAME_MAX_SIZE, 0444, proc_dostring);
	/* &table[12] is terminator */
5160 5161 5162 5163

	return table;
}

5164
static ctl_table *sd_alloc_ctl_cpu_table(int cpu)
5165 5166 5167 5168 5169 5170 5171 5172 5173
{
	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);
5174 5175
	if (table == NULL)
		return NULL;
5176 5177 5178 5179 5180

	i = 0;
	for_each_domain(cpu, sd) {
		snprintf(buf, 32, "domain%d", i);
		entry->procname = kstrdup(buf, GFP_KERNEL);
5181
		entry->mode = 0555;
5182 5183 5184 5185 5186 5187 5188 5189
		entry->child = sd_alloc_ctl_domain_table(sd);
		entry++;
		i++;
	}
	return table;
}

static struct ctl_table_header *sd_sysctl_header;
5190
static void register_sched_domain_sysctl(void)
5191
{
5192
	int i, cpu_num = num_possible_cpus();
5193 5194 5195
	struct ctl_table *entry = sd_alloc_ctl_entry(cpu_num + 1);
	char buf[32];

5196 5197 5198
	WARN_ON(sd_ctl_dir[0].child);
	sd_ctl_dir[0].child = entry;

5199 5200 5201
	if (entry == NULL)
		return;

5202
	for_each_possible_cpu(i) {
5203 5204
		snprintf(buf, 32, "cpu%d", i);
		entry->procname = kstrdup(buf, GFP_KERNEL);
5205
		entry->mode = 0555;
5206
		entry->child = sd_alloc_ctl_cpu_table(i);
5207
		entry++;
5208
	}
5209 5210

	WARN_ON(sd_sysctl_header);
5211 5212
	sd_sysctl_header = register_sysctl_table(sd_ctl_root);
}
5213

5214
/* may be called multiple times per register */
5215 5216
static void unregister_sched_domain_sysctl(void)
{
5217 5218
	if (sd_sysctl_header)
		unregister_sysctl_table(sd_sysctl_header);
5219
	sd_sysctl_header = NULL;
5220 5221
	if (sd_ctl_dir[0].child)
		sd_free_ctl_entry(&sd_ctl_dir[0].child);
5222
}
5223
#else
5224 5225 5226 5227
static void register_sched_domain_sysctl(void)
{
}
static void unregister_sched_domain_sysctl(void)
5228 5229 5230 5231
{
}
#endif

5232 5233 5234 5235 5236
static void set_rq_online(struct rq *rq)
{
	if (!rq->online) {
		const struct sched_class *class;

5237
		cpumask_set_cpu(rq->cpu, rq->rd->online);
5238 5239 5240 5241 5242 5243 5244 5245 5246 5247 5248 5249 5250 5251 5252 5253 5254 5255 5256
		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);
		}

5257
		cpumask_clear_cpu(rq->cpu, rq->rd->online);
5258 5259 5260 5261
		rq->online = 0;
	}
}

L
Linus Torvalds 已提交
5262 5263 5264 5265
/*
 * migration_call - callback that gets triggered when a CPU is added.
 * Here we can start up the necessary migration thread for the new CPU.
 */
5266 5267
static int __cpuinit
migration_call(struct notifier_block *nfb, unsigned long action, void *hcpu)
L
Linus Torvalds 已提交
5268
{
5269
	int cpu = (long)hcpu;
L
Linus Torvalds 已提交
5270
	unsigned long flags;
5271
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
5272

5273
	switch (action & ~CPU_TASKS_FROZEN) {
5274

L
Linus Torvalds 已提交
5275
	case CPU_UP_PREPARE:
5276
		rq->calc_load_update = calc_load_update;
L
Linus Torvalds 已提交
5277
		break;
5278

L
Linus Torvalds 已提交
5279
	case CPU_ONLINE:
5280
		/* Update our root-domain */
5281
		raw_spin_lock_irqsave(&rq->lock, flags);
5282
		if (rq->rd) {
5283
			BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
5284 5285

			set_rq_online(rq);
5286
		}
5287
		raw_spin_unlock_irqrestore(&rq->lock, flags);
L
Linus Torvalds 已提交
5288
		break;
5289

L
Linus Torvalds 已提交
5290
#ifdef CONFIG_HOTPLUG_CPU
5291
	case CPU_DYING:
5292
		sched_ttwu_pending();
G
Gregory Haskins 已提交
5293
		/* Update our root-domain */
5294
		raw_spin_lock_irqsave(&rq->lock, flags);
G
Gregory Haskins 已提交
5295
		if (rq->rd) {
5296
			BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
5297
			set_rq_offline(rq);
G
Gregory Haskins 已提交
5298
		}
5299 5300
		migrate_tasks(cpu);
		BUG_ON(rq->nr_running != 1); /* the migration thread */
5301
		raw_spin_unlock_irqrestore(&rq->lock, flags);
5302 5303 5304

		migrate_nr_uninterruptible(rq);
		calc_global_load_remove(rq);
G
Gregory Haskins 已提交
5305
		break;
L
Linus Torvalds 已提交
5306 5307
#endif
	}
5308 5309 5310

	update_max_interval();

L
Linus Torvalds 已提交
5311 5312 5313
	return NOTIFY_OK;
}

5314 5315 5316
/*
 * Register at high priority so that task migration (migrate_all_tasks)
 * happens before everything else.  This has to be lower priority than
5317
 * the notifier in the perf_event subsystem, though.
L
Linus Torvalds 已提交
5318
 */
5319
static struct notifier_block __cpuinitdata migration_notifier = {
L
Linus Torvalds 已提交
5320
	.notifier_call = migration_call,
5321
	.priority = CPU_PRI_MIGRATION,
L
Linus Torvalds 已提交
5322 5323
};

5324 5325 5326 5327 5328 5329 5330 5331 5332 5333 5334 5335 5336 5337 5338 5339 5340 5341 5342 5343 5344 5345 5346 5347 5348
static int __cpuinit sched_cpu_active(struct notifier_block *nfb,
				      unsigned long action, void *hcpu)
{
	switch (action & ~CPU_TASKS_FROZEN) {
	case CPU_ONLINE:
	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;
	}
}

5349
static int __init migration_init(void)
L
Linus Torvalds 已提交
5350 5351
{
	void *cpu = (void *)(long)smp_processor_id();
5352
	int err;
5353

5354
	/* Initialize migration for the boot CPU */
5355 5356
	err = migration_call(&migration_notifier, CPU_UP_PREPARE, cpu);
	BUG_ON(err == NOTIFY_BAD);
L
Linus Torvalds 已提交
5357 5358
	migration_call(&migration_notifier, CPU_ONLINE, cpu);
	register_cpu_notifier(&migration_notifier);
5359

5360 5361 5362 5363
	/* Register cpu active notifiers */
	cpu_notifier(sched_cpu_active, CPU_PRI_SCHED_ACTIVE);
	cpu_notifier(sched_cpu_inactive, CPU_PRI_SCHED_INACTIVE);

5364
	return 0;
L
Linus Torvalds 已提交
5365
}
5366
early_initcall(migration_init);
L
Linus Torvalds 已提交
5367 5368 5369
#endif

#ifdef CONFIG_SMP
5370

5371 5372
static cpumask_var_t sched_domains_tmpmask; /* sched_domains_mutex */

5373
#ifdef CONFIG_SCHED_DEBUG
I
Ingo Molnar 已提交
5374

5375 5376 5377 5378 5379 5380 5381 5382 5383 5384
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);

5385
static int sched_domain_debug_one(struct sched_domain *sd, int cpu, int level,
5386
				  struct cpumask *groupmask)
L
Linus Torvalds 已提交
5387
{
I
Ingo Molnar 已提交
5388
	struct sched_group *group = sd->groups;
5389
	char str[256];
L
Linus Torvalds 已提交
5390

R
Rusty Russell 已提交
5391
	cpulist_scnprintf(str, sizeof(str), sched_domain_span(sd));
5392
	cpumask_clear(groupmask);
I
Ingo Molnar 已提交
5393 5394 5395 5396

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

	if (!(sd->flags & SD_LOAD_BALANCE)) {
P
Peter Zijlstra 已提交
5397
		printk("does not load-balance\n");
I
Ingo Molnar 已提交
5398
		if (sd->parent)
P
Peter Zijlstra 已提交
5399 5400
			printk(KERN_ERR "ERROR: !SD_LOAD_BALANCE domain"
					" has parent");
I
Ingo Molnar 已提交
5401
		return -1;
N
Nick Piggin 已提交
5402 5403
	}

P
Peter Zijlstra 已提交
5404
	printk(KERN_CONT "span %s level %s\n", str, sd->name);
I
Ingo Molnar 已提交
5405

5406
	if (!cpumask_test_cpu(cpu, sched_domain_span(sd))) {
P
Peter Zijlstra 已提交
5407 5408
		printk(KERN_ERR "ERROR: domain->span does not contain "
				"CPU%d\n", cpu);
I
Ingo Molnar 已提交
5409
	}
5410
	if (!cpumask_test_cpu(cpu, sched_group_cpus(group))) {
P
Peter Zijlstra 已提交
5411 5412
		printk(KERN_ERR "ERROR: domain->groups does not contain"
				" CPU%d\n", cpu);
I
Ingo Molnar 已提交
5413
	}
L
Linus Torvalds 已提交
5414

I
Ingo Molnar 已提交
5415
	printk(KERN_DEBUG "%*s groups:", level + 1, "");
L
Linus Torvalds 已提交
5416
	do {
I
Ingo Molnar 已提交
5417
		if (!group) {
P
Peter Zijlstra 已提交
5418 5419
			printk("\n");
			printk(KERN_ERR "ERROR: group is NULL\n");
L
Linus Torvalds 已提交
5420 5421 5422
			break;
		}

5423
		if (!group->sgp->power) {
P
Peter Zijlstra 已提交
5424 5425 5426
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: domain->cpu_power not "
					"set\n");
I
Ingo Molnar 已提交
5427 5428
			break;
		}
L
Linus Torvalds 已提交
5429

5430
		if (!cpumask_weight(sched_group_cpus(group))) {
P
Peter Zijlstra 已提交
5431 5432
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: empty group\n");
I
Ingo Molnar 已提交
5433 5434
			break;
		}
L
Linus Torvalds 已提交
5435

5436
		if (cpumask_intersects(groupmask, sched_group_cpus(group))) {
P
Peter Zijlstra 已提交
5437 5438
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: repeated CPUs\n");
I
Ingo Molnar 已提交
5439 5440
			break;
		}
L
Linus Torvalds 已提交
5441

5442
		cpumask_or(groupmask, groupmask, sched_group_cpus(group));
L
Linus Torvalds 已提交
5443

R
Rusty Russell 已提交
5444
		cpulist_scnprintf(str, sizeof(str), sched_group_cpus(group));
5445

P
Peter Zijlstra 已提交
5446
		printk(KERN_CONT " %s", str);
5447
		if (group->sgp->power != SCHED_POWER_SCALE) {
P
Peter Zijlstra 已提交
5448
			printk(KERN_CONT " (cpu_power = %d)",
5449
				group->sgp->power);
5450
		}
L
Linus Torvalds 已提交
5451

I
Ingo Molnar 已提交
5452 5453
		group = group->next;
	} while (group != sd->groups);
P
Peter Zijlstra 已提交
5454
	printk(KERN_CONT "\n");
L
Linus Torvalds 已提交
5455

5456
	if (!cpumask_equal(sched_domain_span(sd), groupmask))
P
Peter Zijlstra 已提交
5457
		printk(KERN_ERR "ERROR: groups don't span domain->span\n");
L
Linus Torvalds 已提交
5458

5459 5460
	if (sd->parent &&
	    !cpumask_subset(groupmask, sched_domain_span(sd->parent)))
P
Peter Zijlstra 已提交
5461 5462
		printk(KERN_ERR "ERROR: parent span is not a superset "
			"of domain->span\n");
I
Ingo Molnar 已提交
5463 5464
	return 0;
}
L
Linus Torvalds 已提交
5465

I
Ingo Molnar 已提交
5466 5467 5468
static void sched_domain_debug(struct sched_domain *sd, int cpu)
{
	int level = 0;
L
Linus Torvalds 已提交
5469

5470 5471 5472
	if (!sched_domain_debug_enabled)
		return;

I
Ingo Molnar 已提交
5473 5474 5475 5476
	if (!sd) {
		printk(KERN_DEBUG "CPU%d attaching NULL sched-domain.\n", cpu);
		return;
	}
L
Linus Torvalds 已提交
5477

I
Ingo Molnar 已提交
5478 5479 5480
	printk(KERN_DEBUG "CPU%d attaching sched-domain:\n", cpu);

	for (;;) {
5481
		if (sched_domain_debug_one(sd, cpu, level, sched_domains_tmpmask))
I
Ingo Molnar 已提交
5482
			break;
L
Linus Torvalds 已提交
5483 5484
		level++;
		sd = sd->parent;
5485
		if (!sd)
I
Ingo Molnar 已提交
5486 5487
			break;
	}
L
Linus Torvalds 已提交
5488
}
5489
#else /* !CONFIG_SCHED_DEBUG */
5490
# define sched_domain_debug(sd, cpu) do { } while (0)
5491
#endif /* CONFIG_SCHED_DEBUG */
L
Linus Torvalds 已提交
5492

5493
static int sd_degenerate(struct sched_domain *sd)
5494
{
5495
	if (cpumask_weight(sched_domain_span(sd)) == 1)
5496 5497 5498 5499 5500 5501
		return 1;

	/* Following flags need at least 2 groups */
	if (sd->flags & (SD_LOAD_BALANCE |
			 SD_BALANCE_NEWIDLE |
			 SD_BALANCE_FORK |
5502 5503 5504
			 SD_BALANCE_EXEC |
			 SD_SHARE_CPUPOWER |
			 SD_SHARE_PKG_RESOURCES)) {
5505 5506 5507 5508 5509
		if (sd->groups != sd->groups->next)
			return 0;
	}

	/* Following flags don't use groups */
5510
	if (sd->flags & (SD_WAKE_AFFINE))
5511 5512 5513 5514 5515
		return 0;

	return 1;
}

5516 5517
static int
sd_parent_degenerate(struct sched_domain *sd, struct sched_domain *parent)
5518 5519 5520 5521 5522 5523
{
	unsigned long cflags = sd->flags, pflags = parent->flags;

	if (sd_degenerate(parent))
		return 1;

5524
	if (!cpumask_equal(sched_domain_span(sd), sched_domain_span(parent)))
5525 5526 5527 5528 5529 5530 5531
		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 |
5532 5533 5534
				SD_BALANCE_EXEC |
				SD_SHARE_CPUPOWER |
				SD_SHARE_PKG_RESOURCES);
5535 5536
		if (nr_node_ids == 1)
			pflags &= ~SD_SERIALIZE;
5537 5538 5539 5540 5541 5542 5543
	}
	if (~cflags & pflags)
		return 0;

	return 1;
}

5544
static void free_rootdomain(struct rcu_head *rcu)
5545
{
5546
	struct root_domain *rd = container_of(rcu, struct root_domain, rcu);
5547

5548
	cpupri_cleanup(&rd->cpupri);
5549 5550 5551 5552 5553 5554
	free_cpumask_var(rd->rto_mask);
	free_cpumask_var(rd->online);
	free_cpumask_var(rd->span);
	kfree(rd);
}

G
Gregory Haskins 已提交
5555 5556
static void rq_attach_root(struct rq *rq, struct root_domain *rd)
{
I
Ingo Molnar 已提交
5557
	struct root_domain *old_rd = NULL;
G
Gregory Haskins 已提交
5558 5559
	unsigned long flags;

5560
	raw_spin_lock_irqsave(&rq->lock, flags);
G
Gregory Haskins 已提交
5561 5562

	if (rq->rd) {
I
Ingo Molnar 已提交
5563
		old_rd = rq->rd;
G
Gregory Haskins 已提交
5564

5565
		if (cpumask_test_cpu(rq->cpu, old_rd->online))
5566
			set_rq_offline(rq);
G
Gregory Haskins 已提交
5567

5568
		cpumask_clear_cpu(rq->cpu, old_rd->span);
5569

I
Ingo Molnar 已提交
5570 5571 5572 5573 5574 5575 5576
		/*
		 * 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
5577 5578 5579 5580 5581
	}

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

5582
	cpumask_set_cpu(rq->cpu, rd->span);
5583
	if (cpumask_test_cpu(rq->cpu, cpu_active_mask))
5584
		set_rq_online(rq);
G
Gregory Haskins 已提交
5585

5586
	raw_spin_unlock_irqrestore(&rq->lock, flags);
I
Ingo Molnar 已提交
5587 5588

	if (old_rd)
5589
		call_rcu_sched(&old_rd->rcu, free_rootdomain);
G
Gregory Haskins 已提交
5590 5591
}

5592
static int init_rootdomain(struct root_domain *rd)
G
Gregory Haskins 已提交
5593 5594 5595
{
	memset(rd, 0, sizeof(*rd));

5596
	if (!alloc_cpumask_var(&rd->span, GFP_KERNEL))
5597
		goto out;
5598
	if (!alloc_cpumask_var(&rd->online, GFP_KERNEL))
5599
		goto free_span;
5600
	if (!alloc_cpumask_var(&rd->rto_mask, GFP_KERNEL))
5601
		goto free_online;
5602

5603
	if (cpupri_init(&rd->cpupri) != 0)
5604
		goto free_rto_mask;
5605
	return 0;
5606

5607 5608
free_rto_mask:
	free_cpumask_var(rd->rto_mask);
5609 5610 5611 5612
free_online:
	free_cpumask_var(rd->online);
free_span:
	free_cpumask_var(rd->span);
5613
out:
5614
	return -ENOMEM;
G
Gregory Haskins 已提交
5615 5616
}

5617 5618 5619 5620 5621 5622
/*
 * 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
5623 5624
static void init_defrootdomain(void)
{
5625
	init_rootdomain(&def_root_domain);
5626

G
Gregory Haskins 已提交
5627 5628 5629
	atomic_set(&def_root_domain.refcount, 1);
}

5630
static struct root_domain *alloc_rootdomain(void)
G
Gregory Haskins 已提交
5631 5632 5633 5634 5635 5636 5637
{
	struct root_domain *rd;

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

5638
	if (init_rootdomain(rd) != 0) {
5639 5640 5641
		kfree(rd);
		return NULL;
	}
G
Gregory Haskins 已提交
5642 5643 5644 5645

	return rd;
}

5646 5647 5648 5649 5650 5651 5652 5653 5654 5655 5656 5657 5658 5659 5660 5661 5662 5663 5664
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);
}

5665 5666 5667
static void free_sched_domain(struct rcu_head *rcu)
{
	struct sched_domain *sd = container_of(rcu, struct sched_domain, rcu);
5668 5669 5670 5671 5672 5673 5674 5675

	/*
	 * 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)) {
5676
		kfree(sd->groups->sgp);
5677
		kfree(sd->groups);
5678
	}
5679 5680 5681 5682 5683 5684 5685 5686 5687 5688 5689 5690 5691 5692
	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);
}

L
Linus Torvalds 已提交
5693
/*
I
Ingo Molnar 已提交
5694
 * Attach the domain 'sd' to 'cpu' as its base domain. Callers must
L
Linus Torvalds 已提交
5695 5696
 * hold the hotplug lock.
 */
I
Ingo Molnar 已提交
5697 5698
static void
cpu_attach_domain(struct sched_domain *sd, struct root_domain *rd, int cpu)
L
Linus Torvalds 已提交
5699
{
5700
	struct rq *rq = cpu_rq(cpu);
5701 5702 5703
	struct sched_domain *tmp;

	/* Remove the sched domains which do not contribute to scheduling. */
5704
	for (tmp = sd; tmp; ) {
5705 5706 5707
		struct sched_domain *parent = tmp->parent;
		if (!parent)
			break;
5708

5709
		if (sd_parent_degenerate(tmp, parent)) {
5710
			tmp->parent = parent->parent;
5711 5712
			if (parent->parent)
				parent->parent->child = tmp;
5713
			destroy_sched_domain(parent, cpu);
5714 5715
		} else
			tmp = tmp->parent;
5716 5717
	}

5718
	if (sd && sd_degenerate(sd)) {
5719
		tmp = sd;
5720
		sd = sd->parent;
5721
		destroy_sched_domain(tmp, cpu);
5722 5723 5724
		if (sd)
			sd->child = NULL;
	}
L
Linus Torvalds 已提交
5725

5726
	sched_domain_debug(sd, cpu);
L
Linus Torvalds 已提交
5727

G
Gregory Haskins 已提交
5728
	rq_attach_root(rq, rd);
5729
	tmp = rq->sd;
N
Nick Piggin 已提交
5730
	rcu_assign_pointer(rq->sd, sd);
5731
	destroy_sched_domains(tmp, cpu);
L
Linus Torvalds 已提交
5732 5733 5734
}

/* cpus with isolated domains */
5735
static cpumask_var_t cpu_isolated_map;
L
Linus Torvalds 已提交
5736 5737 5738 5739

/* Setup the mask of cpus configured for isolated domains */
static int __init isolated_cpu_setup(char *str)
{
R
Rusty Russell 已提交
5740
	alloc_bootmem_cpumask_var(&cpu_isolated_map);
R
Rusty Russell 已提交
5741
	cpulist_parse(str, cpu_isolated_map);
L
Linus Torvalds 已提交
5742 5743 5744
	return 1;
}

I
Ingo Molnar 已提交
5745
__setup("isolcpus=", isolated_cpu_setup);
L
Linus Torvalds 已提交
5746

5747
#ifdef CONFIG_NUMA
5748

5749 5750 5751 5752 5753
/**
 * 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
5754
 * Find the next node to include in a given scheduling domain. Simply
5755 5756 5757 5758
 * finds the closest node not already in the @used_nodes map.
 *
 * Should use nodemask_t.
 */
5759
static int find_next_best_node(int node, nodemask_t *used_nodes)
5760
{
5761
	int i, n, val, min_val, best_node = -1;
5762 5763 5764

	min_val = INT_MAX;

5765
	for (i = 0; i < nr_node_ids; i++) {
5766
		/* Start at @node */
5767
		n = (node + i) % nr_node_ids;
5768 5769 5770 5771 5772

		if (!nr_cpus_node(n))
			continue;

		/* Skip already used nodes */
5773
		if (node_isset(n, *used_nodes))
5774 5775 5776 5777 5778 5779 5780 5781 5782 5783 5784
			continue;

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

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

5785 5786
	if (best_node != -1)
		node_set(best_node, *used_nodes);
5787 5788 5789 5790 5791 5792
	return best_node;
}

/**
 * sched_domain_node_span - get a cpumask for a node's sched_domain
 * @node: node whose cpumask we're constructing
5793
 * @span: resulting cpumask
5794
 *
I
Ingo Molnar 已提交
5795
 * Given a node, construct a good cpumask for its sched_domain to span. It
5796 5797 5798
 * should be one that prevents unnecessary balancing, but also spreads tasks
 * out optimally.
 */
5799
static void sched_domain_node_span(int node, struct cpumask *span)
5800
{
5801
	nodemask_t used_nodes;
5802
	int i;
5803

5804
	cpumask_clear(span);
5805
	nodes_clear(used_nodes);
5806

5807
	cpumask_or(span, span, cpumask_of_node(node));
5808
	node_set(node, used_nodes);
5809 5810

	for (i = 1; i < SD_NODES_PER_DOMAIN; i++) {
5811
		int next_node = find_next_best_node(node, &used_nodes);
5812 5813
		if (next_node < 0)
			break;
5814
		cpumask_or(span, span, cpumask_of_node(next_node));
5815 5816
	}
}
5817 5818 5819 5820 5821 5822 5823 5824 5825

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;
}
5826 5827 5828 5829 5830

static const struct cpumask *cpu_allnodes_mask(int cpu)
{
	return cpu_possible_mask;
}
5831
#endif /* CONFIG_NUMA */
5832

5833 5834 5835 5836 5837
static const struct cpumask *cpu_cpu_mask(int cpu)
{
	return cpumask_of_node(cpu_to_node(cpu));
}

5838
int sched_smt_power_savings = 0, sched_mc_power_savings = 0;
5839

5840 5841 5842
struct sd_data {
	struct sched_domain **__percpu sd;
	struct sched_group **__percpu sg;
5843
	struct sched_group_power **__percpu sgp;
5844 5845
};

5846
struct s_data {
5847
	struct sched_domain ** __percpu sd;
5848 5849 5850
	struct root_domain	*rd;
};

5851 5852
enum s_alloc {
	sa_rootdomain,
5853
	sa_sd,
5854
	sa_sd_storage,
5855 5856 5857
	sa_none,
};

5858 5859 5860
struct sched_domain_topology_level;

typedef struct sched_domain *(*sched_domain_init_f)(struct sched_domain_topology_level *tl, int cpu);
5861 5862
typedef const struct cpumask *(*sched_domain_mask_f)(int cpu);

5863 5864
#define SDTL_OVERLAP	0x01

5865
struct sched_domain_topology_level {
5866 5867
	sched_domain_init_f init;
	sched_domain_mask_f mask;
5868
	int		    flags;
5869
	struct sd_data      data;
5870 5871
};

5872 5873 5874 5875 5876 5877 5878 5879 5880 5881 5882 5883 5884 5885 5886 5887 5888 5889 5890
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(),
5891
				GFP_KERNEL, cpu_to_node(cpu));
5892 5893 5894 5895 5896 5897 5898 5899 5900 5901 5902 5903 5904 5905 5906 5907 5908 5909 5910 5911 5912 5913 5914 5915 5916 5917 5918 5919 5920 5921 5922 5923 5924 5925 5926 5927 5928 5929

		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;
}

5930
static int get_group(int cpu, struct sd_data *sdd, struct sched_group **sg)
L
Linus Torvalds 已提交
5931
{
5932 5933
	struct sched_domain *sd = *per_cpu_ptr(sdd->sd, cpu);
	struct sched_domain *child = sd->child;
L
Linus Torvalds 已提交
5934

5935 5936
	if (child)
		cpu = cpumask_first(sched_domain_span(child));
5937

5938
	if (sg) {
5939
		*sg = *per_cpu_ptr(sdd->sg, cpu);
5940
		(*sg)->sgp = *per_cpu_ptr(sdd->sgp, cpu);
5941
		atomic_set(&(*sg)->sgp->ref, 1); /* for claim_allocations */
5942
	}
5943 5944

	return cpu;
5945 5946
}

5947
/*
5948 5949 5950
 * 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.
5951 5952
 *
 * Assumes the sched_domain tree is fully constructed
5953
 */
5954 5955
static int
build_sched_groups(struct sched_domain *sd, int cpu)
L
Linus Torvalds 已提交
5956
{
5957 5958 5959
	struct sched_group *first = NULL, *last = NULL;
	struct sd_data *sdd = sd->private;
	const struct cpumask *span = sched_domain_span(sd);
5960
	struct cpumask *covered;
5961
	int i;
5962

5963 5964 5965 5966 5967 5968
	get_group(cpu, sdd, &sd->groups);
	atomic_inc(&sd->groups->ref);

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

5969 5970 5971
	lockdep_assert_held(&sched_domains_mutex);
	covered = sched_domains_tmpmask;

5972
	cpumask_clear(covered);
5973

5974 5975 5976 5977
	for_each_cpu(i, span) {
		struct sched_group *sg;
		int group = get_group(i, sdd, &sg);
		int j;
5978

5979 5980
		if (cpumask_test_cpu(i, covered))
			continue;
5981

5982
		cpumask_clear(sched_group_cpus(sg));
5983
		sg->sgp->power = 0;
5984

5985 5986 5987
		for_each_cpu(j, span) {
			if (get_group(j, sdd, NULL) != group)
				continue;
5988

5989 5990 5991
			cpumask_set_cpu(j, covered);
			cpumask_set_cpu(j, sched_group_cpus(sg));
		}
5992

5993 5994 5995 5996 5997 5998 5999
		if (!first)
			first = sg;
		if (last)
			last->next = sg;
		last = sg;
	}
	last->next = first;
6000 6001

	return 0;
6002
}
6003

6004 6005 6006 6007 6008 6009 6010 6011 6012 6013 6014 6015
/*
 * 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)
{
6016
	struct sched_group *sg = sd->groups;
6017

6018 6019 6020 6021 6022 6023
	WARN_ON(!sd || !sg);

	do {
		sg->group_weight = cpumask_weight(sched_group_cpus(sg));
		sg = sg->next;
	} while (sg != sd->groups);
6024

6025 6026
	if (cpu != group_first_cpu(sg))
		return;
6027

6028
	update_group_power(sd, cpu);
6029
	atomic_set(&sg->sgp->nr_busy_cpus, sg->group_weight);
6030 6031
}

6032 6033 6034 6035 6036
int __weak arch_sd_sibling_asym_packing(void)
{
       return 0*SD_ASYM_PACKING;
}

6037 6038 6039 6040 6041
/*
 * Initializers for schedule domains
 * Non-inlined to reduce accumulated stack pressure in build_sched_domains()
 */

6042 6043 6044 6045 6046 6047
#ifdef CONFIG_SCHED_DEBUG
# define SD_INIT_NAME(sd, type)		sd->name = #type
#else
# define SD_INIT_NAME(sd, type)		do { } while (0)
#endif

6048 6049 6050 6051 6052 6053 6054 6055 6056
#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;							\
6057 6058 6059 6060 6061 6062 6063 6064 6065 6066 6067 6068 6069
}

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
6070 6071 6072
#ifdef CONFIG_SCHED_BOOK
 SD_INIT_FUNC(BOOK)
#endif
6073

6074
static int default_relax_domain_level = -1;
6075
int sched_domain_level_max;
6076 6077 6078

static int __init setup_relax_domain_level(char *str)
{
6079 6080 6081
	unsigned long val;

	val = simple_strtoul(str, NULL, 0);
6082
	if (val < sched_domain_level_max)
6083 6084
		default_relax_domain_level = val;

6085 6086 6087 6088 6089 6090 6091 6092 6093 6094 6095 6096 6097 6098 6099 6100 6101 6102
	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 */
6103
		sd->flags &= ~(SD_BALANCE_WAKE|SD_BALANCE_NEWIDLE);
6104 6105
	} else {
		/* turn on idle balance on this domain */
6106
		sd->flags |= (SD_BALANCE_WAKE|SD_BALANCE_NEWIDLE);
6107 6108 6109
	}
}

6110 6111 6112
static void __sdt_free(const struct cpumask *cpu_map);
static int __sdt_alloc(const struct cpumask *cpu_map);

6113 6114 6115 6116 6117
static void __free_domain_allocs(struct s_data *d, enum s_alloc what,
				 const struct cpumask *cpu_map)
{
	switch (what) {
	case sa_rootdomain:
6118 6119
		if (!atomic_read(&d->rd->refcount))
			free_rootdomain(&d->rd->rcu); /* fall through */
6120 6121
	case sa_sd:
		free_percpu(d->sd); /* fall through */
6122
	case sa_sd_storage:
6123
		__sdt_free(cpu_map); /* fall through */
6124 6125 6126 6127
	case sa_none:
		break;
	}
}
6128

6129 6130 6131
static enum s_alloc __visit_domain_allocation_hell(struct s_data *d,
						   const struct cpumask *cpu_map)
{
6132 6133
	memset(d, 0, sizeof(*d));

6134 6135
	if (__sdt_alloc(cpu_map))
		return sa_sd_storage;
6136 6137 6138
	d->sd = alloc_percpu(struct sched_domain *);
	if (!d->sd)
		return sa_sd_storage;
6139
	d->rd = alloc_rootdomain();
6140
	if (!d->rd)
6141
		return sa_sd;
6142 6143
	return sa_rootdomain;
}
G
Gregory Haskins 已提交
6144

6145 6146 6147 6148 6149 6150 6151 6152 6153 6154 6155 6156
/*
 * 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;

6157
	if (atomic_read(&(*per_cpu_ptr(sdd->sg, cpu))->ref))
6158
		*per_cpu_ptr(sdd->sg, cpu) = NULL;
6159 6160

	if (atomic_read(&(*per_cpu_ptr(sdd->sgp, cpu))->ref))
6161
		*per_cpu_ptr(sdd->sgp, cpu) = NULL;
6162 6163
}

6164 6165
#ifdef CONFIG_SCHED_SMT
static const struct cpumask *cpu_smt_mask(int cpu)
6166
{
6167
	return topology_thread_cpumask(cpu);
6168
}
6169
#endif
6170

6171 6172 6173
/*
 * Topology list, bottom-up.
 */
6174
static struct sched_domain_topology_level default_topology[] = {
6175 6176
#ifdef CONFIG_SCHED_SMT
	{ sd_init_SIBLING, cpu_smt_mask, },
6177
#endif
6178
#ifdef CONFIG_SCHED_MC
6179
	{ sd_init_MC, cpu_coregroup_mask, },
6180
#endif
6181 6182 6183 6184 6185
#ifdef CONFIG_SCHED_BOOK
	{ sd_init_BOOK, cpu_book_mask, },
#endif
	{ sd_init_CPU, cpu_cpu_mask, },
#ifdef CONFIG_NUMA
6186
	{ sd_init_NODE, cpu_node_mask, SDTL_OVERLAP, },
6187
	{ sd_init_ALLNODES, cpu_allnodes_mask, },
L
Linus Torvalds 已提交
6188
#endif
6189 6190 6191 6192 6193
	{ NULL, },
};

static struct sched_domain_topology_level *sched_domain_topology = default_topology;

6194 6195 6196 6197 6198 6199 6200 6201 6202 6203 6204 6205 6206 6207 6208 6209
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;

6210 6211 6212 6213
		sdd->sgp = alloc_percpu(struct sched_group_power *);
		if (!sdd->sgp)
			return -ENOMEM;

6214 6215 6216
		for_each_cpu(j, cpu_map) {
			struct sched_domain *sd;
			struct sched_group *sg;
6217
			struct sched_group_power *sgp;
6218 6219 6220 6221 6222 6223 6224 6225 6226 6227 6228 6229 6230 6231

		       	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;
6232 6233 6234 6235 6236 6237 6238

			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;
6239 6240 6241 6242 6243 6244 6245 6246 6247 6248 6249 6250 6251 6252 6253
		}
	}

	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) {
6254 6255 6256
			struct sched_domain *sd = *per_cpu_ptr(sdd->sd, j);
			if (sd && (sd->flags & SD_OVERLAP))
				free_sched_groups(sd->groups, 0);
6257
			kfree(*per_cpu_ptr(sdd->sd, j));
6258
			kfree(*per_cpu_ptr(sdd->sg, j));
6259
			kfree(*per_cpu_ptr(sdd->sgp, j));
6260 6261 6262
		}
		free_percpu(sdd->sd);
		free_percpu(sdd->sg);
6263
		free_percpu(sdd->sgp);
6264 6265 6266
	}
}

6267 6268
struct sched_domain *build_sched_domain(struct sched_domain_topology_level *tl,
		struct s_data *d, const struct cpumask *cpu_map,
6269
		struct sched_domain_attr *attr, struct sched_domain *child,
6270 6271
		int cpu)
{
6272
	struct sched_domain *sd = tl->init(tl, cpu);
6273
	if (!sd)
6274
		return child;
6275 6276 6277

	set_domain_attribute(sd, attr);
	cpumask_and(sched_domain_span(sd), cpu_map, tl->mask(cpu));
6278 6279 6280
	if (child) {
		sd->level = child->level + 1;
		sched_domain_level_max = max(sched_domain_level_max, sd->level);
6281
		child->parent = sd;
6282
	}
6283
	sd->child = child;
6284 6285 6286 6287

	return sd;
}

6288 6289 6290 6291
/*
 * Build sched domains for a given set of cpus and attach the sched domains
 * to the individual cpus
 */
6292 6293
static int build_sched_domains(const struct cpumask *cpu_map,
			       struct sched_domain_attr *attr)
6294 6295
{
	enum s_alloc alloc_state = sa_none;
6296
	struct sched_domain *sd;
6297
	struct s_data d;
6298
	int i, ret = -ENOMEM;
6299

6300 6301 6302
	alloc_state = __visit_domain_allocation_hell(&d, cpu_map);
	if (alloc_state != sa_rootdomain)
		goto error;
6303

6304
	/* Set up domains for cpus specified by the cpu_map. */
6305
	for_each_cpu(i, cpu_map) {
6306 6307
		struct sched_domain_topology_level *tl;

6308
		sd = NULL;
6309
		for (tl = sched_domain_topology; tl->init; tl++) {
6310
			sd = build_sched_domain(tl, &d, cpu_map, attr, sd, i);
6311 6312
			if (tl->flags & SDTL_OVERLAP || sched_feat(FORCE_SD_OVERLAP))
				sd->flags |= SD_OVERLAP;
6313 6314
			if (cpumask_equal(cpu_map, sched_domain_span(sd)))
				break;
6315
		}
6316

6317 6318 6319
		while (sd->child)
			sd = sd->child;

6320
		*per_cpu_ptr(d.sd, i) = sd;
6321 6322 6323 6324 6325 6326
	}

	/* 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));
6327 6328 6329 6330 6331 6332 6333
			if (sd->flags & SD_OVERLAP) {
				if (build_overlap_sched_groups(sd, i))
					goto error;
			} else {
				if (build_sched_groups(sd, i))
					goto error;
			}
6334
		}
6335
	}
6336

L
Linus Torvalds 已提交
6337
	/* Calculate CPU power for physical packages and nodes */
6338 6339 6340
	for (i = nr_cpumask_bits-1; i >= 0; i--) {
		if (!cpumask_test_cpu(i, cpu_map))
			continue;
6341

6342 6343
		for (sd = *per_cpu_ptr(d.sd, i); sd; sd = sd->parent) {
			claim_allocations(i, sd);
6344
			init_sched_groups_power(i, sd);
6345
		}
6346
	}
6347

L
Linus Torvalds 已提交
6348
	/* Attach the domains */
6349
	rcu_read_lock();
6350
	for_each_cpu(i, cpu_map) {
6351
		sd = *per_cpu_ptr(d.sd, i);
6352
		cpu_attach_domain(sd, d.rd, i);
L
Linus Torvalds 已提交
6353
	}
6354
	rcu_read_unlock();
6355

6356
	ret = 0;
6357
error:
6358
	__free_domain_allocs(&d, alloc_state, cpu_map);
6359
	return ret;
L
Linus Torvalds 已提交
6360
}
P
Paul Jackson 已提交
6361

6362
static cpumask_var_t *doms_cur;	/* current sched domains */
P
Paul Jackson 已提交
6363
static int ndoms_cur;		/* number of sched domains in 'doms_cur' */
I
Ingo Molnar 已提交
6364 6365
static struct sched_domain_attr *dattr_cur;
				/* attribues of custom domains in 'doms_cur' */
P
Paul Jackson 已提交
6366 6367 6368

/*
 * Special case: If a kmalloc of a doms_cur partition (array of
6369 6370
 * cpumask) fails, then fallback to a single sched domain,
 * as determined by the single cpumask fallback_doms.
P
Paul Jackson 已提交
6371
 */
6372
static cpumask_var_t fallback_doms;
P
Paul Jackson 已提交
6373

6374 6375 6376 6377 6378 6379
/*
 * 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)
6380
{
6381
	return 0;
6382 6383
}

6384 6385 6386 6387 6388 6389 6390 6391 6392 6393 6394 6395 6396 6397 6398 6399 6400 6401 6402 6403 6404 6405 6406 6407 6408
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);
}

6409
/*
I
Ingo Molnar 已提交
6410
 * Set up scheduler domains and groups. Callers must hold the hotplug lock.
P
Paul Jackson 已提交
6411 6412
 * For now this just excludes isolated cpus, but could be used to
 * exclude other special cases in the future.
6413
 */
6414
static int init_sched_domains(const struct cpumask *cpu_map)
6415
{
6416 6417
	int err;

6418
	arch_update_cpu_topology();
P
Paul Jackson 已提交
6419
	ndoms_cur = 1;
6420
	doms_cur = alloc_sched_domains(ndoms_cur);
P
Paul Jackson 已提交
6421
	if (!doms_cur)
6422 6423
		doms_cur = &fallback_doms;
	cpumask_andnot(doms_cur[0], cpu_map, cpu_isolated_map);
6424
	dattr_cur = NULL;
6425
	err = build_sched_domains(doms_cur[0], NULL);
6426
	register_sched_domain_sysctl();
6427 6428

	return err;
6429 6430 6431 6432 6433 6434
}

/*
 * Detach sched domains from a group of cpus specified in cpu_map
 * These cpus will now be attached to the NULL domain
 */
6435
static void detach_destroy_domains(const struct cpumask *cpu_map)
6436 6437 6438
{
	int i;

6439
	rcu_read_lock();
6440
	for_each_cpu(i, cpu_map)
G
Gregory Haskins 已提交
6441
		cpu_attach_domain(NULL, &def_root_domain, i);
6442
	rcu_read_unlock();
6443 6444
}

6445 6446 6447 6448 6449 6450 6451 6452 6453 6454 6455 6456 6457 6458 6459 6460
/* 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
6461 6462
/*
 * Partition sched domains as specified by the 'ndoms_new'
I
Ingo Molnar 已提交
6463
 * cpumasks in the array doms_new[] of cpumasks. This compares
P
Paul Jackson 已提交
6464 6465 6466
 * doms_new[] to the current sched domain partitioning, doms_cur[].
 * It destroys each deleted domain and builds each new domain.
 *
6467
 * 'doms_new' is an array of cpumask_var_t's of length 'ndoms_new'.
I
Ingo Molnar 已提交
6468 6469 6470
 * 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
6471 6472 6473
 * current 'doms_cur' domains and in the new 'doms_new', we can leave
 * it as it is.
 *
6474 6475 6476 6477 6478 6479
 * 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
6480
 *
6481
 * If doms_new == NULL it will be replaced with cpu_online_mask.
6482 6483
 * ndoms_new == 0 is a special case for destroying existing domains,
 * and it will not create the default domain.
6484
 *
P
Paul Jackson 已提交
6485 6486
 * Call with hotplug lock held
 */
6487
void partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
6488
			     struct sched_domain_attr *dattr_new)
P
Paul Jackson 已提交
6489
{
6490
	int i, j, n;
6491
	int new_topology;
P
Paul Jackson 已提交
6492

6493
	mutex_lock(&sched_domains_mutex);
6494

6495 6496 6497
	/* always unregister in case we don't destroy any domains */
	unregister_sched_domain_sysctl();

6498 6499 6500
	/* Let architecture update cpu core mappings. */
	new_topology = arch_update_cpu_topology();

6501
	n = doms_new ? ndoms_new : 0;
P
Paul Jackson 已提交
6502 6503 6504

	/* Destroy deleted domains */
	for (i = 0; i < ndoms_cur; i++) {
6505
		for (j = 0; j < n && !new_topology; j++) {
6506
			if (cpumask_equal(doms_cur[i], doms_new[j])
6507
			    && dattrs_equal(dattr_cur, i, dattr_new, j))
P
Paul Jackson 已提交
6508 6509 6510
				goto match1;
		}
		/* no match - a current sched domain not in new doms_new[] */
6511
		detach_destroy_domains(doms_cur[i]);
P
Paul Jackson 已提交
6512 6513 6514 6515
match1:
		;
	}

6516 6517
	if (doms_new == NULL) {
		ndoms_cur = 0;
6518
		doms_new = &fallback_doms;
6519
		cpumask_andnot(doms_new[0], cpu_active_mask, cpu_isolated_map);
6520
		WARN_ON_ONCE(dattr_new);
6521 6522
	}

P
Paul Jackson 已提交
6523 6524
	/* Build new domains */
	for (i = 0; i < ndoms_new; i++) {
6525
		for (j = 0; j < ndoms_cur && !new_topology; j++) {
6526
			if (cpumask_equal(doms_new[i], doms_cur[j])
6527
			    && dattrs_equal(dattr_new, i, dattr_cur, j))
P
Paul Jackson 已提交
6528 6529 6530
				goto match2;
		}
		/* no match - add a new doms_new */
6531
		build_sched_domains(doms_new[i], dattr_new ? dattr_new + i : NULL);
P
Paul Jackson 已提交
6532 6533 6534 6535 6536
match2:
		;
	}

	/* Remember the new sched domains */
6537 6538
	if (doms_cur != &fallback_doms)
		free_sched_domains(doms_cur, ndoms_cur);
6539
	kfree(dattr_cur);	/* kfree(NULL) is safe */
P
Paul Jackson 已提交
6540
	doms_cur = doms_new;
6541
	dattr_cur = dattr_new;
P
Paul Jackson 已提交
6542
	ndoms_cur = ndoms_new;
6543 6544

	register_sched_domain_sysctl();
6545

6546
	mutex_unlock(&sched_domains_mutex);
P
Paul Jackson 已提交
6547 6548
}

6549
#if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT)
6550
static void reinit_sched_domains(void)
6551
{
6552
	get_online_cpus();
6553 6554 6555 6556

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

6557
	rebuild_sched_domains();
6558
	put_online_cpus();
6559 6560 6561 6562
}

static ssize_t sched_power_savings_store(const char *buf, size_t count, int smt)
{
6563
	unsigned int level = 0;
6564

6565 6566 6567 6568 6569 6570 6571 6572 6573 6574 6575
	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)
6576 6577 6578
		return -EINVAL;

	if (smt)
6579
		sched_smt_power_savings = level;
6580
	else
6581
		sched_mc_power_savings = level;
6582

6583
	reinit_sched_domains();
6584

6585
	return count;
6586 6587 6588
}

#ifdef CONFIG_SCHED_MC
6589
static ssize_t sched_mc_power_savings_show(struct sysdev_class *class,
6590
					   struct sysdev_class_attribute *attr,
6591
					   char *page)
6592 6593 6594
{
	return sprintf(page, "%u\n", sched_mc_power_savings);
}
6595
static ssize_t sched_mc_power_savings_store(struct sysdev_class *class,
6596
					    struct sysdev_class_attribute *attr,
6597
					    const char *buf, size_t count)
6598 6599 6600
{
	return sched_power_savings_store(buf, count, 0);
}
6601 6602 6603
static SYSDEV_CLASS_ATTR(sched_mc_power_savings, 0644,
			 sched_mc_power_savings_show,
			 sched_mc_power_savings_store);
6604 6605 6606
#endif

#ifdef CONFIG_SCHED_SMT
6607
static ssize_t sched_smt_power_savings_show(struct sysdev_class *dev,
6608
					    struct sysdev_class_attribute *attr,
6609
					    char *page)
6610 6611 6612
{
	return sprintf(page, "%u\n", sched_smt_power_savings);
}
6613
static ssize_t sched_smt_power_savings_store(struct sysdev_class *dev,
6614
					     struct sysdev_class_attribute *attr,
6615
					     const char *buf, size_t count)
6616 6617 6618
{
	return sched_power_savings_store(buf, count, 1);
}
6619 6620
static SYSDEV_CLASS_ATTR(sched_smt_power_savings, 0644,
		   sched_smt_power_savings_show,
A
Adrian Bunk 已提交
6621 6622 6623
		   sched_smt_power_savings_store);
#endif

6624
int __init sched_create_sysfs_power_savings_entries(struct sysdev_class *cls)
A
Adrian Bunk 已提交
6625 6626 6627 6628 6629 6630 6631 6632 6633 6634 6635 6636 6637 6638 6639
{
	int err = 0;

#ifdef CONFIG_SCHED_SMT
	if (smt_capable())
		err = sysfs_create_file(&cls->kset.kobj,
					&attr_sched_smt_power_savings.attr);
#endif
#ifdef CONFIG_SCHED_MC
	if (!err && mc_capable())
		err = sysfs_create_file(&cls->kset.kobj,
					&attr_sched_mc_power_savings.attr);
#endif
	return err;
}
6640
#endif /* CONFIG_SCHED_MC || CONFIG_SCHED_SMT */
6641

L
Linus Torvalds 已提交
6642
/*
6643 6644 6645
 * 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
6646
 */
6647 6648
static int cpuset_cpu_active(struct notifier_block *nfb, unsigned long action,
			     void *hcpu)
6649
{
6650
	switch (action & ~CPU_TASKS_FROZEN) {
6651
	case CPU_ONLINE:
6652
	case CPU_DOWN_FAILED:
6653
		cpuset_update_active_cpus();
6654
		return NOTIFY_OK;
6655 6656 6657 6658
	default:
		return NOTIFY_DONE;
	}
}
6659

6660 6661
static int cpuset_cpu_inactive(struct notifier_block *nfb, unsigned long action,
			       void *hcpu)
6662 6663 6664 6665 6666
{
	switch (action & ~CPU_TASKS_FROZEN) {
	case CPU_DOWN_PREPARE:
		cpuset_update_active_cpus();
		return NOTIFY_OK;
6667 6668 6669 6670 6671
	default:
		return NOTIFY_DONE;
	}
}

L
Linus Torvalds 已提交
6672 6673
void __init sched_init_smp(void)
{
6674 6675 6676
	cpumask_var_t non_isolated_cpus;

	alloc_cpumask_var(&non_isolated_cpus, GFP_KERNEL);
6677
	alloc_cpumask_var(&fallback_doms, GFP_KERNEL);
6678

6679
	get_online_cpus();
6680
	mutex_lock(&sched_domains_mutex);
6681
	init_sched_domains(cpu_active_mask);
6682 6683 6684
	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);
6685
	mutex_unlock(&sched_domains_mutex);
6686
	put_online_cpus();
6687

6688 6689
	hotcpu_notifier(cpuset_cpu_active, CPU_PRI_CPUSET_ACTIVE);
	hotcpu_notifier(cpuset_cpu_inactive, CPU_PRI_CPUSET_INACTIVE);
6690 6691 6692 6693

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

6694
	init_hrtick();
6695 6696

	/* Move init over to a non-isolated CPU */
6697
	if (set_cpus_allowed_ptr(current, non_isolated_cpus) < 0)
6698
		BUG();
I
Ingo Molnar 已提交
6699
	sched_init_granularity();
6700
	free_cpumask_var(non_isolated_cpus);
6701

6702
	init_sched_rt_class();
L
Linus Torvalds 已提交
6703 6704 6705 6706
}
#else
void __init sched_init_smp(void)
{
I
Ingo Molnar 已提交
6707
	sched_init_granularity();
L
Linus Torvalds 已提交
6708 6709 6710
}
#endif /* CONFIG_SMP */

6711 6712
const_debug unsigned int sysctl_timer_migration = 1;

L
Linus Torvalds 已提交
6713 6714 6715 6716 6717 6718 6719
int in_sched_functions(unsigned long addr)
{
	return in_lock_functions(addr) ||
		(addr >= (unsigned long)__sched_text_start
		&& addr < (unsigned long)__sched_text_end);
}

6720 6721
#ifdef CONFIG_CGROUP_SCHED
struct task_group root_task_group;
6722
#endif
P
Peter Zijlstra 已提交
6723

6724
DECLARE_PER_CPU(cpumask_var_t, load_balance_tmpmask);
P
Peter Zijlstra 已提交
6725

L
Linus Torvalds 已提交
6726 6727
void __init sched_init(void)
{
I
Ingo Molnar 已提交
6728
	int i, j;
6729 6730 6731 6732 6733 6734 6735
	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 **);
6736
#endif
6737
#ifdef CONFIG_CPUMASK_OFFSTACK
6738
	alloc_size += num_possible_cpus() * cpumask_size();
6739 6740
#endif
	if (alloc_size) {
6741
		ptr = (unsigned long)kzalloc(alloc_size, GFP_NOWAIT);
6742 6743

#ifdef CONFIG_FAIR_GROUP_SCHED
6744
		root_task_group.se = (struct sched_entity **)ptr;
6745 6746
		ptr += nr_cpu_ids * sizeof(void **);

6747
		root_task_group.cfs_rq = (struct cfs_rq **)ptr;
6748
		ptr += nr_cpu_ids * sizeof(void **);
6749

6750
#endif /* CONFIG_FAIR_GROUP_SCHED */
6751
#ifdef CONFIG_RT_GROUP_SCHED
6752
		root_task_group.rt_se = (struct sched_rt_entity **)ptr;
6753 6754
		ptr += nr_cpu_ids * sizeof(void **);

6755
		root_task_group.rt_rq = (struct rt_rq **)ptr;
6756 6757
		ptr += nr_cpu_ids * sizeof(void **);

6758
#endif /* CONFIG_RT_GROUP_SCHED */
6759 6760 6761 6762 6763 6764
#ifdef CONFIG_CPUMASK_OFFSTACK
		for_each_possible_cpu(i) {
			per_cpu(load_balance_tmpmask, i) = (void *)ptr;
			ptr += cpumask_size();
		}
#endif /* CONFIG_CPUMASK_OFFSTACK */
6765
	}
I
Ingo Molnar 已提交
6766

G
Gregory Haskins 已提交
6767 6768 6769 6770
#ifdef CONFIG_SMP
	init_defrootdomain();
#endif

6771 6772 6773 6774
	init_rt_bandwidth(&def_rt_bandwidth,
			global_rt_period(), global_rt_runtime());

#ifdef CONFIG_RT_GROUP_SCHED
6775
	init_rt_bandwidth(&root_task_group.rt_bandwidth,
6776
			global_rt_period(), global_rt_runtime());
6777
#endif /* CONFIG_RT_GROUP_SCHED */
6778

D
Dhaval Giani 已提交
6779
#ifdef CONFIG_CGROUP_SCHED
6780 6781
	list_add(&root_task_group.list, &task_groups);
	INIT_LIST_HEAD(&root_task_group.children);
6782
	INIT_LIST_HEAD(&root_task_group.siblings);
6783
	autogroup_init(&init_task);
D
Dhaval Giani 已提交
6784
#endif /* CONFIG_CGROUP_SCHED */
P
Peter Zijlstra 已提交
6785

6786
	for_each_possible_cpu(i) {
6787
		struct rq *rq;
L
Linus Torvalds 已提交
6788 6789

		rq = cpu_rq(i);
6790
		raw_spin_lock_init(&rq->lock);
N
Nick Piggin 已提交
6791
		rq->nr_running = 0;
6792 6793
		rq->calc_load_active = 0;
		rq->calc_load_update = jiffies + LOAD_FREQ;
6794
		init_cfs_rq(&rq->cfs);
P
Peter Zijlstra 已提交
6795
		init_rt_rq(&rq->rt, rq);
I
Ingo Molnar 已提交
6796
#ifdef CONFIG_FAIR_GROUP_SCHED
6797
		root_task_group.shares = ROOT_TASK_GROUP_LOAD;
P
Peter Zijlstra 已提交
6798
		INIT_LIST_HEAD(&rq->leaf_cfs_rq_list);
D
Dhaval Giani 已提交
6799
		/*
6800
		 * How much cpu bandwidth does root_task_group get?
D
Dhaval Giani 已提交
6801 6802 6803 6804
		 *
		 * 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
6805
		 * root_task_group and its child task-groups in a fair manner,
D
Dhaval Giani 已提交
6806 6807 6808
		 * based on each entity's (task or task-group's) weight
		 * (se->load.weight).
		 *
6809
		 * In other words, if root_task_group has 10 tasks of weight
D
Dhaval Giani 已提交
6810 6811 6812
		 * 1024) and two child groups A0 and A1 (of weight 1024 each),
		 * then A0's share of the cpu resource is:
		 *
6813
		 *	A0's bandwidth = 1024 / (10*1024 + 1024 + 1024) = 8.33%
D
Dhaval Giani 已提交
6814
		 *
6815 6816
		 * 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
6817
		 */
6818
		init_cfs_bandwidth(&root_task_group.cfs_bandwidth);
6819
		init_tg_cfs_entry(&root_task_group, &rq->cfs, NULL, i, NULL);
D
Dhaval Giani 已提交
6820 6821 6822
#endif /* CONFIG_FAIR_GROUP_SCHED */

		rq->rt.rt_runtime = def_rt_bandwidth.rt_runtime;
6823
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
6824
		INIT_LIST_HEAD(&rq->leaf_rt_rq_list);
6825
		init_tg_rt_entry(&root_task_group, &rq->rt, NULL, i, NULL);
I
Ingo Molnar 已提交
6826
#endif
L
Linus Torvalds 已提交
6827

I
Ingo Molnar 已提交
6828 6829
		for (j = 0; j < CPU_LOAD_IDX_MAX; j++)
			rq->cpu_load[j] = 0;
6830 6831 6832

		rq->last_load_update_tick = jiffies;

L
Linus Torvalds 已提交
6833
#ifdef CONFIG_SMP
N
Nick Piggin 已提交
6834
		rq->sd = NULL;
G
Gregory Haskins 已提交
6835
		rq->rd = NULL;
6836
		rq->cpu_power = SCHED_POWER_SCALE;
6837
		rq->post_schedule = 0;
L
Linus Torvalds 已提交
6838
		rq->active_balance = 0;
I
Ingo Molnar 已提交
6839
		rq->next_balance = jiffies;
L
Linus Torvalds 已提交
6840
		rq->push_cpu = 0;
6841
		rq->cpu = i;
6842
		rq->online = 0;
6843 6844
		rq->idle_stamp = 0;
		rq->avg_idle = 2*sysctl_sched_migration_cost;
6845
		rq_attach_root(rq, &def_root_domain);
6846
#ifdef CONFIG_NO_HZ
6847
		rq->nohz_flags = 0;
6848
#endif
L
Linus Torvalds 已提交
6849
#endif
P
Peter Zijlstra 已提交
6850
		init_rq_hrtick(rq);
L
Linus Torvalds 已提交
6851 6852 6853
		atomic_set(&rq->nr_iowait, 0);
	}

6854
	set_load_weight(&init_task);
6855

6856 6857 6858 6859
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&init_task.preempt_notifiers);
#endif

6860
#ifdef CONFIG_RT_MUTEXES
6861
	plist_head_init(&init_task.pi_waiters);
6862 6863
#endif

L
Linus Torvalds 已提交
6864 6865 6866 6867 6868 6869 6870 6871 6872 6873 6874 6875 6876
	/*
	 * 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());
6877 6878 6879

	calc_load_update = jiffies + LOAD_FREQ;

I
Ingo Molnar 已提交
6880 6881 6882 6883
	/*
	 * During early bootup we pretend to be a normal task:
	 */
	current->sched_class = &fair_sched_class;
6884

6885
#ifdef CONFIG_SMP
6886
	zalloc_cpumask_var(&sched_domains_tmpmask, GFP_NOWAIT);
R
Rusty Russell 已提交
6887 6888 6889
	/* May be allocated at isolcpus cmdline parse time */
	if (cpu_isolated_map == NULL)
		zalloc_cpumask_var(&cpu_isolated_map, GFP_NOWAIT);
6890 6891
#endif
	init_sched_fair_class();
6892

6893
	scheduler_running = 1;
L
Linus Torvalds 已提交
6894 6895
}

6896
#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
6897 6898
static inline int preempt_count_equals(int preempt_offset)
{
6899
	int nested = (preempt_count() & ~PREEMPT_ACTIVE) + rcu_preempt_depth();
6900

A
Arnd Bergmann 已提交
6901
	return (nested == preempt_offset);
6902 6903
}

6904
void __might_sleep(const char *file, int line, int preempt_offset)
L
Linus Torvalds 已提交
6905 6906 6907
{
	static unsigned long prev_jiffy;	/* ratelimiting */

6908
	rcu_sleep_check(); /* WARN_ON_ONCE() by default, no rate limit reqd. */
6909 6910
	if ((preempt_count_equals(preempt_offset) && !irqs_disabled()) ||
	    system_state != SYSTEM_RUNNING || oops_in_progress)
I
Ingo Molnar 已提交
6911 6912 6913 6914 6915
		return;
	if (time_before(jiffies, prev_jiffy + HZ) && prev_jiffy)
		return;
	prev_jiffy = jiffies;

P
Peter Zijlstra 已提交
6916 6917 6918 6919 6920 6921 6922
	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
6923 6924 6925 6926 6927

	debug_show_held_locks(current);
	if (irqs_disabled())
		print_irqtrace_events(current);
	dump_stack();
L
Linus Torvalds 已提交
6928 6929 6930 6931 6932
}
EXPORT_SYMBOL(__might_sleep);
#endif

#ifdef CONFIG_MAGIC_SYSRQ
6933 6934
static void normalize_task(struct rq *rq, struct task_struct *p)
{
P
Peter Zijlstra 已提交
6935 6936
	const struct sched_class *prev_class = p->sched_class;
	int old_prio = p->prio;
6937
	int on_rq;
6938

P
Peter Zijlstra 已提交
6939
	on_rq = p->on_rq;
6940 6941 6942 6943 6944 6945 6946
	if (on_rq)
		deactivate_task(rq, p, 0);
	__setscheduler(rq, p, SCHED_NORMAL, 0);
	if (on_rq) {
		activate_task(rq, p, 0);
		resched_task(rq->curr);
	}
P
Peter Zijlstra 已提交
6947 6948

	check_class_changed(rq, p, prev_class, old_prio);
6949 6950
}

L
Linus Torvalds 已提交
6951 6952
void normalize_rt_tasks(void)
{
6953
	struct task_struct *g, *p;
L
Linus Torvalds 已提交
6954
	unsigned long flags;
6955
	struct rq *rq;
L
Linus Torvalds 已提交
6956

6957
	read_lock_irqsave(&tasklist_lock, flags);
6958
	do_each_thread(g, p) {
6959 6960 6961 6962 6963 6964
		/*
		 * Only normalize user tasks:
		 */
		if (!p->mm)
			continue;

I
Ingo Molnar 已提交
6965 6966
		p->se.exec_start		= 0;
#ifdef CONFIG_SCHEDSTATS
6967 6968 6969
		p->se.statistics.wait_start	= 0;
		p->se.statistics.sleep_start	= 0;
		p->se.statistics.block_start	= 0;
I
Ingo Molnar 已提交
6970
#endif
I
Ingo Molnar 已提交
6971 6972 6973 6974 6975 6976 6977 6978

		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
6979
			continue;
I
Ingo Molnar 已提交
6980
		}
L
Linus Torvalds 已提交
6981

6982
		raw_spin_lock(&p->pi_lock);
6983
		rq = __task_rq_lock(p);
L
Linus Torvalds 已提交
6984

6985
		normalize_task(rq, p);
6986

6987
		__task_rq_unlock(rq);
6988
		raw_spin_unlock(&p->pi_lock);
6989 6990
	} while_each_thread(g, p);

6991
	read_unlock_irqrestore(&tasklist_lock, flags);
L
Linus Torvalds 已提交
6992 6993 6994
}

#endif /* CONFIG_MAGIC_SYSRQ */
6995

6996
#if defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB)
6997
/*
6998
 * These functions are only useful for the IA64 MCA handling, or kdb.
6999 7000 7001 7002 7003 7004 7005 7006 7007 7008 7009 7010 7011 7012
 *
 * 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!
 */
7013
struct task_struct *curr_task(int cpu)
7014 7015 7016 7017
{
	return cpu_curr(cpu);
}

7018 7019 7020
#endif /* defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB) */

#ifdef CONFIG_IA64
7021 7022 7023 7024 7025 7026
/**
 * 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
7027 7028
 * 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
7029 7030 7031 7032 7033 7034 7035
 * 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!
 */
7036
void set_curr_task(int cpu, struct task_struct *p)
7037 7038 7039 7040 7041
{
	cpu_curr(cpu) = p;
}

#endif
S
Srivatsa Vaddagiri 已提交
7042

7043
#ifdef CONFIG_RT_GROUP_SCHED
7044 7045
#else /* !CONFIG_RT_GROUP_SCHED */
#endif /* CONFIG_RT_GROUP_SCHED */
7046

D
Dhaval Giani 已提交
7047
#ifdef CONFIG_CGROUP_SCHED
7048 7049 7050
/* task_group_lock serializes the addition/removal of task groups */
static DEFINE_SPINLOCK(task_group_lock);

7051 7052 7053 7054
static void free_sched_group(struct task_group *tg)
{
	free_fair_sched_group(tg);
	free_rt_sched_group(tg);
7055
	autogroup_free(tg);
7056 7057 7058 7059
	kfree(tg);
}

/* allocate runqueue etc for a new task group */
7060
struct task_group *sched_create_group(struct task_group *parent)
7061 7062 7063 7064 7065 7066 7067 7068
{
	struct task_group *tg;
	unsigned long flags;

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

7069
	if (!alloc_fair_sched_group(tg, parent))
7070 7071
		goto err;

7072
	if (!alloc_rt_sched_group(tg, parent))
7073 7074
		goto err;

7075
	spin_lock_irqsave(&task_group_lock, flags);
P
Peter Zijlstra 已提交
7076
	list_add_rcu(&tg->list, &task_groups);
P
Peter Zijlstra 已提交
7077 7078 7079 7080 7081

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

	tg->parent = parent;
	INIT_LIST_HEAD(&tg->children);
7082
	list_add_rcu(&tg->siblings, &parent->children);
7083
	spin_unlock_irqrestore(&task_group_lock, flags);
S
Srivatsa Vaddagiri 已提交
7084

7085
	return tg;
S
Srivatsa Vaddagiri 已提交
7086 7087

err:
P
Peter Zijlstra 已提交
7088
	free_sched_group(tg);
S
Srivatsa Vaddagiri 已提交
7089 7090 7091
	return ERR_PTR(-ENOMEM);
}

7092
/* rcu callback to free various structures associated with a task group */
P
Peter Zijlstra 已提交
7093
static void free_sched_group_rcu(struct rcu_head *rhp)
S
Srivatsa Vaddagiri 已提交
7094 7095
{
	/* now it should be safe to free those cfs_rqs */
P
Peter Zijlstra 已提交
7096
	free_sched_group(container_of(rhp, struct task_group, rcu));
S
Srivatsa Vaddagiri 已提交
7097 7098
}

7099
/* Destroy runqueue etc associated with a task group */
7100
void sched_destroy_group(struct task_group *tg)
S
Srivatsa Vaddagiri 已提交
7101
{
7102
	unsigned long flags;
7103
	int i;
S
Srivatsa Vaddagiri 已提交
7104

7105 7106
	/* end participation in shares distribution */
	for_each_possible_cpu(i)
7107
		unregister_fair_sched_group(tg, i);
7108 7109

	spin_lock_irqsave(&task_group_lock, flags);
P
Peter Zijlstra 已提交
7110
	list_del_rcu(&tg->list);
P
Peter Zijlstra 已提交
7111
	list_del_rcu(&tg->siblings);
7112
	spin_unlock_irqrestore(&task_group_lock, flags);
7113 7114

	/* wait for possible concurrent references to cfs_rqs complete */
P
Peter Zijlstra 已提交
7115
	call_rcu(&tg->rcu, free_sched_group_rcu);
S
Srivatsa Vaddagiri 已提交
7116 7117
}

7118
/* change task's runqueue when it moves between groups.
I
Ingo Molnar 已提交
7119 7120 7121
 *	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.
7122 7123
 */
void sched_move_task(struct task_struct *tsk)
S
Srivatsa Vaddagiri 已提交
7124 7125 7126 7127 7128 7129 7130
{
	int on_rq, running;
	unsigned long flags;
	struct rq *rq;

	rq = task_rq_lock(tsk, &flags);

7131
	running = task_current(rq, tsk);
P
Peter Zijlstra 已提交
7132
	on_rq = tsk->on_rq;
S
Srivatsa Vaddagiri 已提交
7133

7134
	if (on_rq)
S
Srivatsa Vaddagiri 已提交
7135
		dequeue_task(rq, tsk, 0);
7136 7137
	if (unlikely(running))
		tsk->sched_class->put_prev_task(rq, tsk);
S
Srivatsa Vaddagiri 已提交
7138

P
Peter Zijlstra 已提交
7139
#ifdef CONFIG_FAIR_GROUP_SCHED
7140 7141 7142
	if (tsk->sched_class->task_move_group)
		tsk->sched_class->task_move_group(tsk, on_rq);
	else
P
Peter Zijlstra 已提交
7143
#endif
7144
		set_task_rq(tsk, task_cpu(tsk));
P
Peter Zijlstra 已提交
7145

7146 7147 7148
	if (unlikely(running))
		tsk->sched_class->set_curr_task(rq);
	if (on_rq)
7149
		enqueue_task(rq, tsk, 0);
S
Srivatsa Vaddagiri 已提交
7150

7151
	task_rq_unlock(rq, tsk, &flags);
S
Srivatsa Vaddagiri 已提交
7152
}
D
Dhaval Giani 已提交
7153
#endif /* CONFIG_CGROUP_SCHED */
S
Srivatsa Vaddagiri 已提交
7154

7155 7156
#ifdef CONFIG_FAIR_GROUP_SCHED
#endif
7157

7158
#if defined(CONFIG_RT_GROUP_SCHED) || defined(CONFIG_CFS_BANDWIDTH)
P
Peter Zijlstra 已提交
7159 7160 7161
static unsigned long to_ratio(u64 period, u64 runtime)
{
	if (runtime == RUNTIME_INF)
P
Peter Zijlstra 已提交
7162
		return 1ULL << 20;
P
Peter Zijlstra 已提交
7163

P
Peter Zijlstra 已提交
7164
	return div64_u64(runtime << 20, period);
P
Peter Zijlstra 已提交
7165
}
7166 7167 7168 7169 7170 7171 7172
#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
7173

P
Peter Zijlstra 已提交
7174 7175
/* Must be called with tasklist_lock held */
static inline int tg_has_rt_tasks(struct task_group *tg)
7176
{
P
Peter Zijlstra 已提交
7177
	struct task_struct *g, *p;
7178

P
Peter Zijlstra 已提交
7179
	do_each_thread(g, p) {
7180
		if (rt_task(p) && task_rq(p)->rt.tg == tg)
P
Peter Zijlstra 已提交
7181 7182
			return 1;
	} while_each_thread(g, p);
7183

P
Peter Zijlstra 已提交
7184 7185
	return 0;
}
7186

P
Peter Zijlstra 已提交
7187 7188 7189 7190 7191
struct rt_schedulable_data {
	struct task_group *tg;
	u64 rt_period;
	u64 rt_runtime;
};
7192

7193
static int tg_rt_schedulable(struct task_group *tg, void *data)
P
Peter Zijlstra 已提交
7194 7195 7196 7197 7198
{
	struct rt_schedulable_data *d = data;
	struct task_group *child;
	unsigned long total, sum = 0;
	u64 period, runtime;
7199

P
Peter Zijlstra 已提交
7200 7201
	period = ktime_to_ns(tg->rt_bandwidth.rt_period);
	runtime = tg->rt_bandwidth.rt_runtime;
7202

P
Peter Zijlstra 已提交
7203 7204 7205
	if (tg == d->tg) {
		period = d->rt_period;
		runtime = d->rt_runtime;
7206 7207
	}

7208 7209 7210 7211 7212
	/*
	 * Cannot have more runtime than the period.
	 */
	if (runtime > period && runtime != RUNTIME_INF)
		return -EINVAL;
P
Peter Zijlstra 已提交
7213

7214 7215 7216
	/*
	 * Ensure we don't starve existing RT tasks.
	 */
P
Peter Zijlstra 已提交
7217 7218
	if (rt_bandwidth_enabled() && !runtime && tg_has_rt_tasks(tg))
		return -EBUSY;
P
Peter Zijlstra 已提交
7219

P
Peter Zijlstra 已提交
7220
	total = to_ratio(period, runtime);
P
Peter Zijlstra 已提交
7221

7222 7223 7224 7225 7226
	/*
	 * 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
7227

7228 7229 7230
	/*
	 * The sum of our children's runtime should not exceed our own.
	 */
P
Peter Zijlstra 已提交
7231 7232 7233
	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
7234

P
Peter Zijlstra 已提交
7235 7236 7237 7238
		if (child == d->tg) {
			period = d->rt_period;
			runtime = d->rt_runtime;
		}
P
Peter Zijlstra 已提交
7239

P
Peter Zijlstra 已提交
7240
		sum += to_ratio(period, runtime);
P
Peter Zijlstra 已提交
7241
	}
P
Peter Zijlstra 已提交
7242

P
Peter Zijlstra 已提交
7243 7244 7245 7246
	if (sum > total)
		return -EINVAL;

	return 0;
P
Peter Zijlstra 已提交
7247 7248
}

P
Peter Zijlstra 已提交
7249
static int __rt_schedulable(struct task_group *tg, u64 period, u64 runtime)
7250
{
7251 7252
	int ret;

P
Peter Zijlstra 已提交
7253 7254 7255 7256 7257 7258
	struct rt_schedulable_data data = {
		.tg = tg,
		.rt_period = period,
		.rt_runtime = runtime,
	};

7259 7260 7261 7262 7263
	rcu_read_lock();
	ret = walk_tg_tree(tg_rt_schedulable, tg_nop, &data);
	rcu_read_unlock();

	return ret;
7264 7265
}

7266
static int tg_set_rt_bandwidth(struct task_group *tg,
7267
		u64 rt_period, u64 rt_runtime)
P
Peter Zijlstra 已提交
7268
{
P
Peter Zijlstra 已提交
7269
	int i, err = 0;
P
Peter Zijlstra 已提交
7270 7271

	mutex_lock(&rt_constraints_mutex);
7272
	read_lock(&tasklist_lock);
P
Peter Zijlstra 已提交
7273 7274
	err = __rt_schedulable(tg, rt_period, rt_runtime);
	if (err)
P
Peter Zijlstra 已提交
7275
		goto unlock;
P
Peter Zijlstra 已提交
7276

7277
	raw_spin_lock_irq(&tg->rt_bandwidth.rt_runtime_lock);
7278 7279
	tg->rt_bandwidth.rt_period = ns_to_ktime(rt_period);
	tg->rt_bandwidth.rt_runtime = rt_runtime;
P
Peter Zijlstra 已提交
7280 7281 7282 7283

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

7284
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
7285
		rt_rq->rt_runtime = rt_runtime;
7286
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
7287
	}
7288
	raw_spin_unlock_irq(&tg->rt_bandwidth.rt_runtime_lock);
P
Peter Zijlstra 已提交
7289
unlock:
7290
	read_unlock(&tasklist_lock);
P
Peter Zijlstra 已提交
7291 7292 7293
	mutex_unlock(&rt_constraints_mutex);

	return err;
P
Peter Zijlstra 已提交
7294 7295
}

7296 7297 7298 7299 7300 7301 7302 7303 7304
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;

7305
	return tg_set_rt_bandwidth(tg, rt_period, rt_runtime);
7306 7307
}

P
Peter Zijlstra 已提交
7308 7309 7310 7311
long sched_group_rt_runtime(struct task_group *tg)
{
	u64 rt_runtime_us;

7312
	if (tg->rt_bandwidth.rt_runtime == RUNTIME_INF)
P
Peter Zijlstra 已提交
7313 7314
		return -1;

7315
	rt_runtime_us = tg->rt_bandwidth.rt_runtime;
P
Peter Zijlstra 已提交
7316 7317 7318
	do_div(rt_runtime_us, NSEC_PER_USEC);
	return rt_runtime_us;
}
7319 7320 7321 7322 7323 7324 7325 7326

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;

7327 7328 7329
	if (rt_period == 0)
		return -EINVAL;

7330
	return tg_set_rt_bandwidth(tg, rt_period, rt_runtime);
7331 7332 7333 7334 7335 7336 7337 7338 7339 7340 7341 7342 7343
}

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)
{
7344
	u64 runtime, period;
7345 7346
	int ret = 0;

7347 7348 7349
	if (sysctl_sched_rt_period <= 0)
		return -EINVAL;

7350 7351 7352 7353 7354 7355 7356 7357
	runtime = global_rt_runtime();
	period = global_rt_period();

	/*
	 * Sanity check on the sysctl variables.
	 */
	if (runtime > period && runtime != RUNTIME_INF)
		return -EINVAL;
7358

7359
	mutex_lock(&rt_constraints_mutex);
P
Peter Zijlstra 已提交
7360
	read_lock(&tasklist_lock);
7361
	ret = __rt_schedulable(NULL, 0, 0);
P
Peter Zijlstra 已提交
7362
	read_unlock(&tasklist_lock);
7363 7364 7365 7366
	mutex_unlock(&rt_constraints_mutex);

	return ret;
}
7367 7368 7369 7370 7371 7372 7373 7374 7375 7376

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;
}

7377
#else /* !CONFIG_RT_GROUP_SCHED */
7378 7379
static int sched_rt_global_constraints(void)
{
P
Peter Zijlstra 已提交
7380 7381 7382
	unsigned long flags;
	int i;

7383 7384 7385
	if (sysctl_sched_rt_period <= 0)
		return -EINVAL;

7386 7387 7388 7389 7390 7391 7392
	/*
	 * There's always some RT tasks in the root group
	 * -- migration, kstopmachine etc..
	 */
	if (sysctl_sched_rt_runtime == 0)
		return -EBUSY;

7393
	raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags);
P
Peter Zijlstra 已提交
7394 7395 7396
	for_each_possible_cpu(i) {
		struct rt_rq *rt_rq = &cpu_rq(i)->rt;

7397
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
7398
		rt_rq->rt_runtime = global_rt_runtime();
7399
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
7400
	}
7401
	raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags);
P
Peter Zijlstra 已提交
7402

7403 7404
	return 0;
}
7405
#endif /* CONFIG_RT_GROUP_SCHED */
7406 7407

int sched_rt_handler(struct ctl_table *table, int write,
7408
		void __user *buffer, size_t *lenp,
7409 7410 7411 7412 7413 7414 7415 7416 7417 7418
		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;

7419
	ret = proc_dointvec(table, write, buffer, lenp, ppos);
7420 7421 7422 7423 7424 7425 7426 7427 7428 7429 7430 7431 7432 7433 7434 7435

	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;
}
7436

7437
#ifdef CONFIG_CGROUP_SCHED
7438 7439

/* return corresponding task_group object of a cgroup */
7440
static inline struct task_group *cgroup_tg(struct cgroup *cgrp)
7441
{
7442 7443
	return container_of(cgroup_subsys_state(cgrp, cpu_cgroup_subsys_id),
			    struct task_group, css);
7444 7445 7446
}

static struct cgroup_subsys_state *
7447
cpu_cgroup_create(struct cgroup_subsys *ss, struct cgroup *cgrp)
7448
{
7449
	struct task_group *tg, *parent;
7450

7451
	if (!cgrp->parent) {
7452
		/* This is early initialization for the top cgroup */
7453
		return &root_task_group.css;
7454 7455
	}

7456 7457
	parent = cgroup_tg(cgrp->parent);
	tg = sched_create_group(parent);
7458 7459 7460 7461 7462 7463
	if (IS_ERR(tg))
		return ERR_PTR(-ENOMEM);

	return &tg->css;
}

I
Ingo Molnar 已提交
7464 7465
static void
cpu_cgroup_destroy(struct cgroup_subsys *ss, struct cgroup *cgrp)
7466
{
7467
	struct task_group *tg = cgroup_tg(cgrp);
7468 7469 7470 7471

	sched_destroy_group(tg);
}

I
Ingo Molnar 已提交
7472
static int
7473
cpu_cgroup_can_attach_task(struct cgroup *cgrp, struct task_struct *tsk)
7474
{
7475
#ifdef CONFIG_RT_GROUP_SCHED
7476
	if (!sched_rt_can_attach(cgroup_tg(cgrp), tsk))
7477 7478
		return -EINVAL;
#else
7479 7480 7481
	/* We don't support RT-tasks being in separate groups */
	if (tsk->sched_class != &fair_sched_class)
		return -EINVAL;
7482
#endif
7483 7484
	return 0;
}
7485 7486

static void
7487
cpu_cgroup_attach_task(struct cgroup *cgrp, struct task_struct *tsk)
7488 7489 7490 7491
{
	sched_move_task(tsk);
}

7492
static void
7493 7494
cpu_cgroup_exit(struct cgroup_subsys *ss, struct cgroup *cgrp,
		struct cgroup *old_cgrp, struct task_struct *task)
7495 7496 7497 7498 7499 7500 7501 7502 7503 7504 7505 7506
{
	/*
	 * 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);
}

7507
#ifdef CONFIG_FAIR_GROUP_SCHED
7508
static int cpu_shares_write_u64(struct cgroup *cgrp, struct cftype *cftype,
7509
				u64 shareval)
7510
{
7511
	return sched_group_set_shares(cgroup_tg(cgrp), scale_load(shareval));
7512 7513
}

7514
static u64 cpu_shares_read_u64(struct cgroup *cgrp, struct cftype *cft)
7515
{
7516
	struct task_group *tg = cgroup_tg(cgrp);
7517

7518
	return (u64) scale_load_down(tg->shares);
7519
}
7520 7521

#ifdef CONFIG_CFS_BANDWIDTH
7522 7523
static DEFINE_MUTEX(cfs_constraints_mutex);

7524 7525 7526
const u64 max_cfs_quota_period = 1 * NSEC_PER_SEC; /* 1s */
const u64 min_cfs_quota_period = 1 * NSEC_PER_MSEC; /* 1ms */

7527 7528
static int __cfs_schedulable(struct task_group *tg, u64 period, u64 runtime);

7529 7530
static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota)
{
7531
	int i, ret = 0, runtime_enabled, runtime_was_enabled;
7532
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
7533 7534 7535 7536 7537 7538 7539 7540 7541 7542 7543 7544 7545 7546 7547 7548 7549 7550 7551 7552

	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;

7553 7554 7555 7556 7557
	mutex_lock(&cfs_constraints_mutex);
	ret = __cfs_schedulable(tg, period, quota);
	if (ret)
		goto out_unlock;

7558
	runtime_enabled = quota != RUNTIME_INF;
7559 7560
	runtime_was_enabled = cfs_b->quota != RUNTIME_INF;
	account_cfs_bandwidth_used(runtime_enabled, runtime_was_enabled);
7561 7562 7563
	raw_spin_lock_irq(&cfs_b->lock);
	cfs_b->period = ns_to_ktime(period);
	cfs_b->quota = quota;
7564

P
Paul Turner 已提交
7565
	__refill_cfs_bandwidth_runtime(cfs_b);
7566 7567 7568 7569 7570 7571
	/* 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);
	}
7572 7573 7574 7575
	raw_spin_unlock_irq(&cfs_b->lock);

	for_each_possible_cpu(i) {
		struct cfs_rq *cfs_rq = tg->cfs_rq[i];
7576
		struct rq *rq = cfs_rq->rq;
7577 7578

		raw_spin_lock_irq(&rq->lock);
7579
		cfs_rq->runtime_enabled = runtime_enabled;
7580
		cfs_rq->runtime_remaining = 0;
7581

7582
		if (cfs_rq->throttled)
7583
			unthrottle_cfs_rq(cfs_rq);
7584 7585
		raw_spin_unlock_irq(&rq->lock);
	}
7586 7587
out_unlock:
	mutex_unlock(&cfs_constraints_mutex);
7588

7589
	return ret;
7590 7591 7592 7593 7594 7595
}

int tg_set_cfs_quota(struct task_group *tg, long cfs_quota_us)
{
	u64 quota, period;

7596
	period = ktime_to_ns(tg->cfs_bandwidth.period);
7597 7598 7599 7600 7601 7602 7603 7604 7605 7606 7607 7608
	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;

7609
	if (tg->cfs_bandwidth.quota == RUNTIME_INF)
7610 7611
		return -1;

7612
	quota_us = tg->cfs_bandwidth.quota;
7613 7614 7615 7616 7617 7618 7619 7620 7621 7622
	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;
7623
	quota = tg->cfs_bandwidth.quota;
7624 7625 7626 7627 7628 7629 7630 7631 7632 7633 7634

	if (period <= 0)
		return -EINVAL;

	return tg_set_cfs_bandwidth(tg, period, quota);
}

long tg_get_cfs_period(struct task_group *tg)
{
	u64 cfs_period_us;

7635
	cfs_period_us = ktime_to_ns(tg->cfs_bandwidth.period);
7636 7637 7638 7639 7640 7641 7642 7643 7644 7645 7646 7647 7648 7649 7650 7651 7652 7653 7654 7655 7656 7657 7658 7659 7660 7661 7662
	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);
}

7663 7664 7665 7666 7667 7668 7669 7670 7671 7672 7673 7674 7675 7676 7677 7678 7679 7680 7681 7682 7683 7684 7685 7686 7687 7688 7689 7690 7691 7692 7693 7694
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;
7695
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
7696 7697 7698 7699 7700
	s64 quota = 0, parent_quota = -1;

	if (!tg->parent) {
		quota = RUNTIME_INF;
	} else {
7701
		struct cfs_bandwidth *parent_b = &tg->parent->cfs_bandwidth;
7702 7703 7704 7705 7706 7707 7708 7709 7710 7711 7712 7713 7714 7715 7716 7717 7718 7719 7720 7721

		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)
{
7722
	int ret;
7723 7724 7725 7726 7727 7728 7729 7730 7731 7732 7733
	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);
	}

7734 7735 7736 7737 7738
	rcu_read_lock();
	ret = walk_tg_tree(tg_cfs_schedulable_down, tg_nop, &data);
	rcu_read_unlock();

	return ret;
7739
}
7740 7741 7742 7743 7744

static int cpu_stats_show(struct cgroup *cgrp, struct cftype *cft,
		struct cgroup_map_cb *cb)
{
	struct task_group *tg = cgroup_tg(cgrp);
7745
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
7746 7747 7748 7749 7750 7751 7752

	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;
}
7753
#endif /* CONFIG_CFS_BANDWIDTH */
7754
#endif /* CONFIG_FAIR_GROUP_SCHED */
7755

7756
#ifdef CONFIG_RT_GROUP_SCHED
M
Mirco Tischler 已提交
7757
static int cpu_rt_runtime_write(struct cgroup *cgrp, struct cftype *cft,
7758
				s64 val)
P
Peter Zijlstra 已提交
7759
{
7760
	return sched_group_set_rt_runtime(cgroup_tg(cgrp), val);
P
Peter Zijlstra 已提交
7761 7762
}

7763
static s64 cpu_rt_runtime_read(struct cgroup *cgrp, struct cftype *cft)
P
Peter Zijlstra 已提交
7764
{
7765
	return sched_group_rt_runtime(cgroup_tg(cgrp));
P
Peter Zijlstra 已提交
7766
}
7767 7768 7769 7770 7771 7772 7773 7774 7775 7776 7777

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));
}
7778
#endif /* CONFIG_RT_GROUP_SCHED */
P
Peter Zijlstra 已提交
7779

7780
static struct cftype cpu_files[] = {
7781
#ifdef CONFIG_FAIR_GROUP_SCHED
7782 7783
	{
		.name = "shares",
7784 7785
		.read_u64 = cpu_shares_read_u64,
		.write_u64 = cpu_shares_write_u64,
7786
	},
7787
#endif
7788 7789 7790 7791 7792 7793 7794 7795 7796 7797 7798
#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,
	},
7799 7800 7801 7802
	{
		.name = "stat",
		.read_map = cpu_stats_show,
	},
7803
#endif
7804
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
7805
	{
P
Peter Zijlstra 已提交
7806
		.name = "rt_runtime_us",
7807 7808
		.read_s64 = cpu_rt_runtime_read,
		.write_s64 = cpu_rt_runtime_write,
P
Peter Zijlstra 已提交
7809
	},
7810 7811
	{
		.name = "rt_period_us",
7812 7813
		.read_u64 = cpu_rt_period_read_uint,
		.write_u64 = cpu_rt_period_write_uint,
7814
	},
7815
#endif
7816 7817 7818 7819
};

static int cpu_cgroup_populate(struct cgroup_subsys *ss, struct cgroup *cont)
{
7820
	return cgroup_add_files(cont, ss, cpu_files, ARRAY_SIZE(cpu_files));
7821 7822 7823
}

struct cgroup_subsys cpu_cgroup_subsys = {
I
Ingo Molnar 已提交
7824 7825 7826
	.name		= "cpu",
	.create		= cpu_cgroup_create,
	.destroy	= cpu_cgroup_destroy,
7827 7828
	.can_attach_task = cpu_cgroup_can_attach_task,
	.attach_task	= cpu_cgroup_attach_task,
7829
	.exit		= cpu_cgroup_exit,
I
Ingo Molnar 已提交
7830 7831
	.populate	= cpu_cgroup_populate,
	.subsys_id	= cpu_cgroup_subsys_id,
7832 7833 7834
	.early_init	= 1,
};

7835
#endif	/* CONFIG_CGROUP_SCHED */
7836 7837 7838 7839 7840 7841 7842 7843 7844 7845

#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).
 */

7846
/* track cpu usage of a group of tasks and its child groups */
7847 7848 7849
struct cpuacct {
	struct cgroup_subsys_state css;
	/* cpuusage holds pointer to a u64-type object on every cpu */
7850
	u64 __percpu *cpuusage;
7851
	struct percpu_counter cpustat[CPUACCT_STAT_NSTATS];
7852 7853 7854 7855 7856
};

struct cgroup_subsys cpuacct_subsys;

/* return cpu accounting group corresponding to this container */
7857
static inline struct cpuacct *cgroup_ca(struct cgroup *cgrp)
7858
{
7859
	return container_of(cgroup_subsys_state(cgrp, cpuacct_subsys_id),
7860 7861 7862 7863 7864 7865 7866 7867 7868 7869
			    struct cpuacct, css);
}

/* return cpu accounting group to which this task belongs */
static inline struct cpuacct *task_ca(struct task_struct *tsk)
{
	return container_of(task_subsys_state(tsk, cpuacct_subsys_id),
			    struct cpuacct, css);
}

7870 7871 7872 7873 7874 7875 7876
static inline struct cpuacct *parent_ca(struct cpuacct *ca)
{
	if (!ca || !ca->css.cgroup->parent)
		return NULL;
	return cgroup_ca(ca->css.cgroup->parent);
}

7877 7878
/* create a new cpu accounting group */
static struct cgroup_subsys_state *cpuacct_create(
7879
	struct cgroup_subsys *ss, struct cgroup *cgrp)
7880 7881
{
	struct cpuacct *ca = kzalloc(sizeof(*ca), GFP_KERNEL);
7882
	int i;
7883 7884

	if (!ca)
7885
		goto out;
7886 7887

	ca->cpuusage = alloc_percpu(u64);
7888 7889 7890 7891 7892 7893
	if (!ca->cpuusage)
		goto out_free_ca;

	for (i = 0; i < CPUACCT_STAT_NSTATS; i++)
		if (percpu_counter_init(&ca->cpustat[i], 0))
			goto out_free_counters;
7894 7895

	return &ca->css;
7896 7897 7898 7899 7900 7901 7902 7903 7904

out_free_counters:
	while (--i >= 0)
		percpu_counter_destroy(&ca->cpustat[i]);
	free_percpu(ca->cpuusage);
out_free_ca:
	kfree(ca);
out:
	return ERR_PTR(-ENOMEM);
7905 7906 7907
}

/* destroy an existing cpu accounting group */
I
Ingo Molnar 已提交
7908
static void
7909
cpuacct_destroy(struct cgroup_subsys *ss, struct cgroup *cgrp)
7910
{
7911
	struct cpuacct *ca = cgroup_ca(cgrp);
7912
	int i;
7913

7914 7915
	for (i = 0; i < CPUACCT_STAT_NSTATS; i++)
		percpu_counter_destroy(&ca->cpustat[i]);
7916 7917 7918 7919
	free_percpu(ca->cpuusage);
	kfree(ca);
}

7920 7921
static u64 cpuacct_cpuusage_read(struct cpuacct *ca, int cpu)
{
7922
	u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
7923 7924 7925 7926 7927 7928
	u64 data;

#ifndef CONFIG_64BIT
	/*
	 * Take rq->lock to make 64-bit read safe on 32-bit platforms.
	 */
7929
	raw_spin_lock_irq(&cpu_rq(cpu)->lock);
7930
	data = *cpuusage;
7931
	raw_spin_unlock_irq(&cpu_rq(cpu)->lock);
7932 7933 7934 7935 7936 7937 7938 7939 7940
#else
	data = *cpuusage;
#endif

	return data;
}

static void cpuacct_cpuusage_write(struct cpuacct *ca, int cpu, u64 val)
{
7941
	u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
7942 7943 7944 7945 7946

#ifndef CONFIG_64BIT
	/*
	 * Take rq->lock to make 64-bit write safe on 32-bit platforms.
	 */
7947
	raw_spin_lock_irq(&cpu_rq(cpu)->lock);
7948
	*cpuusage = val;
7949
	raw_spin_unlock_irq(&cpu_rq(cpu)->lock);
7950 7951 7952 7953 7954
#else
	*cpuusage = val;
#endif
}

7955
/* return total cpu usage (in nanoseconds) of a group */
7956
static u64 cpuusage_read(struct cgroup *cgrp, struct cftype *cft)
7957
{
7958
	struct cpuacct *ca = cgroup_ca(cgrp);
7959 7960 7961
	u64 totalcpuusage = 0;
	int i;

7962 7963
	for_each_present_cpu(i)
		totalcpuusage += cpuacct_cpuusage_read(ca, i);
7964 7965 7966 7967

	return totalcpuusage;
}

7968 7969 7970 7971 7972 7973 7974 7975 7976 7977 7978 7979
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;
	}

7980 7981
	for_each_present_cpu(i)
		cpuacct_cpuusage_write(ca, i, 0);
7982 7983 7984 7985 7986

out:
	return err;
}

7987 7988 7989 7990 7991 7992 7993 7994 7995 7996 7997 7998 7999 8000 8001
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;
}

8002 8003 8004 8005 8006 8007 8008 8009 8010 8011 8012 8013 8014 8015 8016 8017 8018 8019 8020
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,
		struct cgroup_map_cb *cb)
{
	struct cpuacct *ca = cgroup_ca(cgrp);
	int i;

	for (i = 0; i < CPUACCT_STAT_NSTATS; i++) {
		s64 val = percpu_counter_read(&ca->cpustat[i]);
		val = cputime64_to_clock_t(val);
		cb->fill(cb, cpuacct_stat_desc[i], val);
	}
	return 0;
}

8021 8022 8023
static struct cftype files[] = {
	{
		.name = "usage",
8024 8025
		.read_u64 = cpuusage_read,
		.write_u64 = cpuusage_write,
8026
	},
8027 8028 8029 8030
	{
		.name = "usage_percpu",
		.read_seq_string = cpuacct_percpu_seq_read,
	},
8031 8032 8033 8034
	{
		.name = "stat",
		.read_map = cpuacct_stats_show,
	},
8035 8036
};

8037
static int cpuacct_populate(struct cgroup_subsys *ss, struct cgroup *cgrp)
8038
{
8039
	return cgroup_add_files(cgrp, ss, files, ARRAY_SIZE(files));
8040 8041 8042 8043 8044 8045 8046
}

/*
 * charge this task's execution time to its accounting group.
 *
 * called with rq->lock held.
 */
8047
void cpuacct_charge(struct task_struct *tsk, u64 cputime)
8048 8049
{
	struct cpuacct *ca;
8050
	int cpu;
8051

L
Li Zefan 已提交
8052
	if (unlikely(!cpuacct_subsys.active))
8053 8054
		return;

8055
	cpu = task_cpu(tsk);
8056 8057 8058

	rcu_read_lock();

8059 8060
	ca = task_ca(tsk);

8061
	for (; ca; ca = parent_ca(ca)) {
8062
		u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
8063 8064
		*cpuusage += cputime;
	}
8065 8066

	rcu_read_unlock();
8067 8068
}

8069 8070 8071 8072 8073 8074 8075 8076 8077 8078 8079 8080 8081 8082 8083 8084 8085
/*
 * When CONFIG_VIRT_CPU_ACCOUNTING is enabled one jiffy can be very large
 * in cputime_t units. As a result, cpuacct_update_stats calls
 * percpu_counter_add with values large enough to always overflow the
 * per cpu batch limit causing bad SMP scalability.
 *
 * To fix this we scale percpu_counter_batch by cputime_one_jiffy so we
 * batch the same amount of time with CONFIG_VIRT_CPU_ACCOUNTING disabled
 * and enabled. We cap it at INT_MAX which is the largest allowed batch value.
 */
#ifdef CONFIG_SMP
#define CPUACCT_BATCH	\
	min_t(long, percpu_counter_batch * cputime_one_jiffy, INT_MAX)
#else
#define CPUACCT_BATCH	0
#endif

8086 8087 8088
/*
 * Charge the system/user time to the task's accounting group.
 */
8089
void cpuacct_update_stats(struct task_struct *tsk,
8090 8091 8092
		enum cpuacct_stat_index idx, cputime_t val)
{
	struct cpuacct *ca;
8093
	int batch = CPUACCT_BATCH;
8094 8095 8096 8097 8098 8099 8100 8101

	if (unlikely(!cpuacct_subsys.active))
		return;

	rcu_read_lock();
	ca = task_ca(tsk);

	do {
8102
		__percpu_counter_add(&ca->cpustat[idx], val, batch);
8103
		ca = parent_ca(ca);
8104 8105 8106 8107
	} while (ca);
	rcu_read_unlock();
}

8108 8109 8110 8111 8112 8113 8114 8115
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 */