core.c 205.5 KB
Newer Older
L
Linus Torvalds 已提交
1
/*
2
 *  kernel/sched/core.c
L
Linus Torvalds 已提交
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
 *
 *  Kernel scheduler and related syscalls
 *
 *  Copyright (C) 1991-2002  Linus Torvalds
 *
 *  1996-12-23  Modified by Dave Grothe to fix bugs in semaphores and
 *		make semaphores SMP safe
 *  1998-11-19	Implemented schedule_timeout() and related stuff
 *		by Andrea Arcangeli
 *  2002-01-04	New ultra-scalable O(1) scheduler by Ingo Molnar:
 *		hybrid priority-list and round-robin design with
 *		an array-switch method of distributing timeslices
 *		and per-CPU runqueues.  Cleanups and useful suggestions
 *		by Davide Libenzi, preemptible kernel bits by Robert Love.
 *  2003-09-03	Interactivity tuning by Con Kolivas.
 *  2004-04-02	Scheduler domains code by Nick Piggin
I
Ingo Molnar 已提交
19 20 21 22 23 24
 *  2007-04-15  Work begun on replacing all interactivity tuning with a
 *              fair scheduling design by Con Kolivas.
 *  2007-05-05  Load balancing (smp-nice) and other improvements
 *              by Peter Williams
 *  2007-05-06  Interactivity improvements to CFS by Mike Galbraith
 *  2007-07-01  Group scheduling enhancements by Srivatsa Vaddagiri
25 26
 *  2007-11-29  RT balancing improvements by Steven Rostedt, Gregory Haskins,
 *              Thomas Gleixner, Mike Kravetz
L
Linus Torvalds 已提交
27 28 29 30 31 32
 */

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

#undef SCHED_FEAT

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

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

#undef SCHED_FEAT

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

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

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

165 166
#ifdef HAVE_JUMP_LABEL

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

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

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

#undef SCHED_FEAT

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

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

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

	if (cnt > 63)
		cnt = 63;

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

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

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

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

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

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

	return cnt;
}

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

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

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

	return 0;
}
late_initcall(sched_init_debug);
260
#endif /* CONFIG_SCHED_DEBUG */
I
Ingo Molnar 已提交
261

262 263 264 265 266 267
/*
 * Number of tasks to iterate in a single balance run.
 * Limited because this is done with IRQs disabled.
 */
const_debug unsigned int sysctl_sched_nr_migrate = 32;

268 269 270 271 272 273 274 275
/*
 * period over which we average the RT time consumption, measured
 * in ms.
 *
 * default: 1s
 */
const_debug unsigned int sysctl_sched_time_avg = MSEC_PER_SEC;

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

282
__read_mostly int scheduler_running;
283

P
Peter Zijlstra 已提交
284 285 286 287 288
/*
 * part of the period that we allow rt tasks to run in us.
 * default: 0.95s
 */
int sysctl_sched_rt_runtime = 950000;
P
Peter Zijlstra 已提交
289 290


L
Linus Torvalds 已提交
291

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

300 301
	lockdep_assert_held(&p->pi_lock);

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

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

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

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

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

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

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

	return rq;
}

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

static void hrtick_clear(struct rq *rq)
{
	if (hrtimer_active(&rq->hrtick_timer))
		hrtimer_cancel(&rq->hrtick_timer);
}

/*
 * High-resolution timer tick.
 * Runs from hardirq context with interrupts disabled.
 */
static enum hrtimer_restart hrtick(struct hrtimer *timer)
{
	struct rq *rq = container_of(timer, struct rq, hrtick_timer);

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

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

	return HRTIMER_NORESTART;
}

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

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

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

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

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

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

	switch (action) {
	case CPU_UP_CANCELED:
	case CPU_UP_CANCELED_FROZEN:
	case CPU_DOWN_PREPARE:
	case CPU_DOWN_PREPARE_FROZEN:
	case CPU_DEAD:
	case CPU_DEAD_FROZEN:
443
		hrtick_clear(cpu_rq(cpu));
444 445 446 447 448 449
		return NOTIFY_OK;
	}

	return NOTIFY_DONE;
}

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

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

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

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

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

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

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

I
Ingo Molnar 已提交
498 499 500 501 502 503 504 505 506 507 508 509 510
/*
 * resched_task - mark a task 'to be rescheduled now'.
 *
 * On UP this means the setting of the need_resched flag, on SMP it
 * might also involve a cross-CPU call to trigger the scheduler on
 * the target CPU.
 */
#ifdef CONFIG_SMP

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

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

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

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

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

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

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

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

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

#ifdef CONFIG_NO_HZ
544 545 546 547 548 549 550 551 552 553 554 555 556 557
/*
 * In the semi idle case, use the nearest busy cpu for migrating timers
 * from an idle cpu.  This is good for power-savings.
 *
 * We don't do similar optimization for completely idle system, as
 * selecting an idle cpu will add more delays to the timers than intended
 * (as that cpu's timer base may not be uptodate wrt jiffies etc).
 */
int get_nohz_timer_target(void)
{
	int cpu = smp_processor_id();
	int i;
	struct sched_domain *sd;

558
	rcu_read_lock();
559
	for_each_domain(cpu, sd) {
560 561 562 563 564 565
		for_each_cpu(i, sched_domain_span(sd)) {
			if (!idle_cpu(i)) {
				cpu = i;
				goto unlock;
			}
		}
566
	}
567 568
unlock:
	rcu_read_unlock();
569 570
	return cpu;
}
571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596
/*
 * When add_timer_on() enqueues a timer into the timer wheel of an
 * idle CPU then this timer might expire before the next timer event
 * which is scheduled to wake up that CPU. In case of a completely
 * idle system the next event might even be infinite time into the
 * future. wake_up_idle_cpu() ensures that the CPU is woken up and
 * leaves the inner idle loop so the newly added timer is taken into
 * account when the CPU goes back to idle and evaluates the timer
 * wheel for the next timer event.
 */
void wake_up_idle_cpu(int cpu)
{
	struct rq *rq = cpu_rq(cpu);

	if (cpu == smp_processor_id())
		return;

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

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

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

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

617
#else /* CONFIG_NO_HZ */
618

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

624
#endif /* CONFIG_NO_HZ */
625

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

	while ((s64)(rq->clock - rq->age_stamp) > period) {
631 632 633 634 635 636
		/*
		 * Inline assembly required to prevent the compiler
		 * optimising this loop into a divmod call.
		 * See __iter_div_u64_rem() for another example of this.
		 */
		asm("" : "+rm" (rq->age_stamp));
637 638 639
		rq->age_stamp += period;
		rq->rt_avg /= 2;
	}
640 641
}

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

650 651
#if defined(CONFIG_RT_GROUP_SCHED) || (defined(CONFIG_FAIR_GROUP_SCHED) && \
			(defined(CONFIG_SMP) || defined(CONFIG_CFS_BANDWIDTH)))
652
/*
653 654 655 656
 * Iterate task_group tree rooted at *from, calling @down when first entering a
 * node and @up when leaving it for the final time.
 *
 * Caller must hold rcu_lock or sufficient equivalent.
657
 */
658
int walk_tg_tree_from(struct task_group *from,
659
			     tg_visitor down, tg_visitor up, void *data)
660 661
{
	struct task_group *parent, *child;
P
Peter Zijlstra 已提交
662
	int ret;
663

664 665
	parent = from;

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

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

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

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

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

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

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

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

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

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

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

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

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

743 744
#ifdef CONFIG_IRQ_TIME_ACCOUNTING

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

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

void enable_sched_clock_irqtime(void)
{
	sched_clock_irqtime = 1;
}

void disable_sched_clock_irqtime(void)
{
	sched_clock_irqtime = 0;
}

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

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

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

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

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

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

static inline void irq_time_write_end(void)
{
}

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

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

	if (!sched_clock_irqtime)
		return;

	local_irq_save(flags);

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

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

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

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

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

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

863
static void update_rq_clock_task(struct rq *rq, s64 delta)
864
{
865 866 867 868 869 870 871 872
/*
 * In theory, the compile should just see 0 here, and optimize out the call
 * to sched_rt_avg_update. But I don't trust it...
 */
#if defined(CONFIG_IRQ_TIME_ACCOUNTING) || defined(CONFIG_PARAVIRT_TIME_ACCOUNTING)
	s64 steal = 0, irq_delta = 0;
#endif
#ifdef CONFIG_IRQ_TIME_ACCOUNTING
873
	irq_delta = irq_time_read(cpu_of(rq)) - rq->prev_irq_time;
874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894

	/*
	 * Since irq_time is only updated on {soft,}irq_exit, we might run into
	 * this case when a previous update_rq_clock() happened inside a
	 * {soft,}irq region.
	 *
	 * When this happens, we stop ->clock_task and only update the
	 * prev_irq_time stamp to account for the part that fit, so that a next
	 * update will consume the rest. This ensures ->clock_task is
	 * monotonic.
	 *
	 * It does however cause some slight miss-attribution of {soft,}irq
	 * time, a more accurate solution would be to update the irq_time using
	 * the current rq->clock timestamp, except that would require using
	 * atomic ops.
	 */
	if (irq_delta > delta)
		irq_delta = delta;

	rq->prev_irq_time += irq_delta;
	delta -= irq_delta;
895 896
#endif
#ifdef CONFIG_PARAVIRT_TIME_ACCOUNTING
897
	if (static_key_false((&paravirt_steal_rq_enabled))) {
898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914
		u64 st;

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

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

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

		rq->prev_steal_time_rq += steal;

		delta -= steal;
	}
#endif

915 916
	rq->clock_task += delta;

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

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

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

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

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

954
#else /* CONFIG_IRQ_TIME_ACCOUNTING */
955

956 957
#define sched_clock_irqtime	(0)

958
#endif
959

960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989
void sched_set_stop_task(int cpu, struct task_struct *stop)
{
	struct sched_param param = { .sched_priority = MAX_RT_PRIO - 1 };
	struct task_struct *old_stop = cpu_rq(cpu)->stop;

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

		stop->sched_class = &stop_sched_class;
	}

	cpu_rq(cpu)->stop = stop;

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

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

998 999 1000 1001 1002 1003 1004
/*
 * Calculate the expected normal priority: i.e. priority
 * without taking RT-inheritance into account. Might be
 * boosted by interactivity modifiers. Changes upon fork,
 * setprio syscalls, and whenever the interactivity
 * estimator recalculates.
 */
1005
static inline int normal_prio(struct task_struct *p)
1006 1007 1008
{
	int prio;

1009
	if (task_has_rt_policy(p))
1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022
		prio = MAX_RT_PRIO-1 - p->rt_priority;
	else
		prio = __normal_prio(p);
	return prio;
}

/*
 * Calculate the current priority, i.e. the priority
 * taken into account by the scheduler. This value might
 * be boosted by RT tasks, or might be boosted by
 * interactivity modifiers. Will be RT if the task got
 * RT-boosted. If not then it returns p->normal_prio.
 */
1023
static int effective_prio(struct task_struct *p)
1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035
{
	p->normal_prio = normal_prio(p);
	/*
	 * If we are RT tasks or we were boosted to RT priority,
	 * keep the priority unchanged. Otherwise, update priority
	 * to the normal priority:
	 */
	if (!rt_prio(p->prio))
		return p->normal_prio;
	return p->prio;
}

L
Linus Torvalds 已提交
1036 1037 1038 1039
/**
 * task_curr - is this task currently executing on a CPU?
 * @p: the task in question.
 */
1040
inline int task_curr(const struct task_struct *p)
L
Linus Torvalds 已提交
1041 1042 1043 1044
{
	return cpu_curr(task_cpu(p)) == p;
}

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

1057
void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags)
1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077
{
	const struct sched_class *class;

	if (p->sched_class == rq->curr->sched_class) {
		rq->curr->sched_class->check_preempt_curr(rq, p, flags);
	} else {
		for_each_class(class) {
			if (class == rq->curr->sched_class)
				break;
			if (class == p->sched_class) {
				resched_task(rq->curr);
				break;
			}
		}
	}

	/*
	 * A queue event has occurred, and we're going to schedule.  In
	 * this case, we can save a useless back to back clock update.
	 */
P
Peter Zijlstra 已提交
1078
	if (rq->curr->on_rq && test_tsk_need_resched(rq->curr))
1079 1080 1081
		rq->skip_clock_update = 1;
}

L
Linus Torvalds 已提交
1082
#ifdef CONFIG_SMP
I
Ingo Molnar 已提交
1083
void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
I
Ingo Molnar 已提交
1084
{
1085 1086 1087 1088 1089
#ifdef CONFIG_SCHED_DEBUG
	/*
	 * We should never call set_task_cpu() on a blocked task,
	 * ttwu() will sort out the placement.
	 */
P
Peter Zijlstra 已提交
1090 1091
	WARN_ON_ONCE(p->state != TASK_RUNNING && p->state != TASK_WAKING &&
			!(task_thread_info(p)->preempt_count & PREEMPT_ACTIVE));
1092 1093

#ifdef CONFIG_LOCKDEP
1094 1095 1096 1097 1098
	/*
	 * The caller should hold either p->pi_lock or rq->lock, when changing
	 * a task's CPU. ->pi_lock for waking tasks, rq->lock for runnable tasks.
	 *
	 * sched_move_task() holds both and thus holding either pins the cgroup,
P
Peter Zijlstra 已提交
1099
	 * see task_group().
1100 1101 1102 1103
	 *
	 * Furthermore, all task_rq users should acquire both locks, see
	 * task_rq_lock().
	 */
1104 1105 1106
	WARN_ON_ONCE(debug_locks && !(lockdep_is_held(&p->pi_lock) ||
				      lockdep_is_held(&task_rq(p)->lock)));
#endif
1107 1108
#endif

1109
	trace_sched_migrate_task(p, new_cpu);
1110

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

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

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

1124 1125
static int migration_cpu_stop(void *data);

L
Linus Torvalds 已提交
1126 1127 1128
/*
 * wait_task_inactive - wait for a thread to unschedule.
 *
R
Roland McGrath 已提交
1129 1130 1131 1132 1133 1134 1135
 * If @match_state is nonzero, it's the @p->state value just checked and
 * not expected to change.  If it changes, i.e. @p might have woken up,
 * then return zero.  When we succeed in waiting for @p to be off its CPU,
 * we return a positive number (its total switch count).  If a second call
 * a short while later returns the same number, the caller can be sure that
 * @p has remained unscheduled the whole time.
 *
L
Linus Torvalds 已提交
1136 1137 1138 1139 1140 1141
 * The caller must ensure that the task *will* unschedule sometime soon,
 * else this function might spin for a *long* time. This function can't
 * be called with interrupts off, or it may introduce deadlock with
 * smp_call_function() if an IPI is sent by the same process we are
 * waiting to become inactive.
 */
R
Roland McGrath 已提交
1142
unsigned long wait_task_inactive(struct task_struct *p, long match_state)
L
Linus Torvalds 已提交
1143 1144
{
	unsigned long flags;
I
Ingo Molnar 已提交
1145
	int running, on_rq;
R
Roland McGrath 已提交
1146
	unsigned long ncsw;
1147
	struct rq *rq;
L
Linus Torvalds 已提交
1148

1149 1150 1151 1152 1153 1154 1155 1156
	for (;;) {
		/*
		 * We do the initial early heuristics without holding
		 * any task-queue locks at all. We'll only try to get
		 * the runqueue lock when things look like they will
		 * work out!
		 */
		rq = task_rq(p);
1157

1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168
		/*
		 * If the task is actively running on another CPU
		 * still, just relax and busy-wait without holding
		 * any locks.
		 *
		 * NOTE! Since we don't hold any locks, it's not
		 * even sure that "rq" stays as the right runqueue!
		 * But we don't care, since "task_running()" will
		 * return false if the runqueue has changed and p
		 * is actually now running somewhere else!
		 */
R
Roland McGrath 已提交
1169 1170 1171
		while (task_running(rq, p)) {
			if (match_state && unlikely(p->state != match_state))
				return 0;
1172
			cpu_relax();
R
Roland McGrath 已提交
1173
		}
1174

1175 1176 1177 1178 1179 1180
		/*
		 * Ok, time to look more closely! We need the rq
		 * lock now, to be *sure*. If we're wrong, we'll
		 * just go back and repeat.
		 */
		rq = task_rq_lock(p, &flags);
1181
		trace_sched_wait_task(p);
1182
		running = task_running(rq, p);
P
Peter Zijlstra 已提交
1183
		on_rq = p->on_rq;
R
Roland McGrath 已提交
1184
		ncsw = 0;
1185
		if (!match_state || p->state == match_state)
1186
			ncsw = p->nvcsw | LONG_MIN; /* sets MSB */
1187
		task_rq_unlock(rq, p, &flags);
1188

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

1195 1196 1197 1198 1199 1200 1201 1202 1203 1204
		/*
		 * Was it really running after all now that we
		 * checked with the proper locks actually held?
		 *
		 * Oops. Go back and try again..
		 */
		if (unlikely(running)) {
			cpu_relax();
			continue;
		}
1205

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

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

1223 1224 1225 1226 1227 1228 1229
		/*
		 * Ahh, all good. It wasn't running, and it wasn't
		 * runnable, which means that it will never become
		 * running in the future either. We're all done!
		 */
		break;
	}
R
Roland McGrath 已提交
1230 1231

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

/***
 * kick_process - kick a running thread to enter/exit the kernel
 * @p: the to-be-kicked thread
 *
 * Cause a process which is running on another CPU to enter
 * kernel-mode, without any delay. (to get signals handled.)
 *
L
Lucas De Marchi 已提交
1241
 * NOTE: this function doesn't have to take the runqueue lock,
L
Linus Torvalds 已提交
1242 1243 1244 1245 1246
 * because all it wants to ensure is that the remote task enters
 * the kernel. If the IPI races and the task has been migrated
 * to another CPU then no harm is done and the purpose has been
 * achieved as well.
 */
1247
void kick_process(struct task_struct *p)
L
Linus Torvalds 已提交
1248 1249 1250 1251 1252 1253 1254 1255 1256
{
	int cpu;

	preempt_disable();
	cpu = task_cpu(p);
	if ((cpu != smp_processor_id()) && task_curr(p))
		smp_send_reschedule(cpu);
	preempt_enable();
}
R
Rusty Russell 已提交
1257
EXPORT_SYMBOL_GPL(kick_process);
N
Nick Piggin 已提交
1258
#endif /* CONFIG_SMP */
L
Linus Torvalds 已提交
1259

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

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

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

1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318
		switch (state) {
		case cpuset:
			/* No more Mr. Nice Guy. */
			cpuset_cpus_allowed_fallback(p);
			state = possible;
			break;

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

		case fail:
			BUG();
			break;
		}
	}

out:
	if (state != cpuset) {
		/*
		 * Don't tell them about moving exiting tasks or
		 * kernel threads (both mm NULL), since they never
		 * leave kernel.
		 */
		if (p->mm && printk_ratelimit()) {
			printk_sched("process %d (%s) no longer affine to cpu%d\n",
					task_pid_nr(p), p->comm, cpu);
		}
1319 1320 1321 1322 1323
	}

	return dest_cpu;
}

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

	/*
	 * In order not to call set_task_cpu() on a blocking task we need
	 * to rely on ttwu() to place the task on a valid ->cpus_allowed
	 * cpu.
	 *
	 * Since this is common to all placement strategies, this lives here.
	 *
	 * [ this allows ->select_task() to simply return task_cpu(p) and
	 *   not worry about this generic constraint ]
	 */
1342
	if (unlikely(!cpumask_test_cpu(cpu, tsk_cpus_allowed(p)) ||
P
Peter Zijlstra 已提交
1343
		     !cpu_online(cpu)))
1344
		cpu = select_fallback_rq(task_cpu(p), p);
1345 1346

	return cpu;
1347
}
1348 1349 1350 1351 1352 1353

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

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

P
Peter Zijlstra 已提交
1362 1363 1364 1365 1366 1367 1368 1369 1370 1371
#ifdef CONFIG_SMP
	int this_cpu = smp_processor_id();

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

		schedstat_inc(p, se.statistics.nr_wakeups_remote);
1372
		rcu_read_lock();
P
Peter Zijlstra 已提交
1373 1374 1375 1376 1377 1378
		for_each_domain(this_cpu, sd) {
			if (cpumask_test_cpu(cpu, sched_domain_span(sd))) {
				schedstat_inc(sd, ttwu_wake_remote);
				break;
			}
		}
1379
		rcu_read_unlock();
P
Peter Zijlstra 已提交
1380
	}
1381 1382 1383 1384

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

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

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

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

#endif /* CONFIG_SCHEDSTATS */
}

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

	/* if a worker is waking up, notify workqueue */
	if (p->flags & PF_WQ_WORKER)
		wq_worker_waking_up(p, cpu_of(rq));
T
Tejun Heo 已提交
1404 1405
}

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

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

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

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

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

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

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

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

	return ret;
}

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

	raw_spin_lock(&rq->lock);

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

	raw_spin_unlock(&rq->lock);
}

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

	/*
	 * Not all reschedule IPI handlers call irq_enter/irq_exit, since
	 * traditionally all their work was done from the interrupt return
	 * path. Now that we actually do some work, we need to make sure
	 * we do call them.
	 *
	 * Some archs already do call them, luckily irq_enter/exit nest
	 * properly.
	 *
	 * Arguably we should visit all archs and update all handlers,
	 * however a fair share of IPIs are still resched only so this would
	 * somewhat pessimize the simple resched case.
	 */
	irq_enter();
P
Peter Zijlstra 已提交
1503
	sched_ttwu_pending();
1504 1505 1506 1507

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

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

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

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

	return ret;

}
#endif /* __ARCH_WANT_INTERRUPTS_ON_CTXSW */
1539

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

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

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

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

/**
L
Linus Torvalds 已提交
1564
 * try_to_wake_up - wake up a thread
T
Tejun Heo 已提交
1565
 * @p: the thread to be awakened
L
Linus Torvalds 已提交
1566
 * @state: the mask of task states that can be woken
T
Tejun Heo 已提交
1567
 * @wake_flags: wake modifier flags (WF_*)
L
Linus Torvalds 已提交
1568 1569 1570 1571 1572 1573 1574
 *
 * Put it on the run-queue if it's not already there. The "current"
 * thread is always on the run-queue (except when the actual
 * re-schedule is in progress), and as such you're allowed to do
 * the simpler "current->state = TASK_RUNNING" to mark yourself
 * runnable without the overhead of this.
 *
T
Tejun Heo 已提交
1575 1576
 * Returns %true if @p was woken up, %false if it was already running
 * or @state didn't match @p's state.
L
Linus Torvalds 已提交
1577
 */
1578 1579
static int
try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
L
Linus Torvalds 已提交
1580 1581
{
	unsigned long flags;
1582
	int cpu, success = 0;
P
Peter Zijlstra 已提交
1583

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

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

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

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

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

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

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

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

	return success;
}

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

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

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

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

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

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

1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686
/**
 * wake_up_process - Wake up a specific process
 * @p: The process to be woken up.
 *
 * Attempt to wake up the nominated process and move it to the set of runnable
 * processes.  Returns 1 if the process was woken up, 0 if it was already
 * running.
 *
 * It may be assumed that this function implies a write memory barrier before
 * changing the task state if and only if any tasks are woken up.
 */
1687
int wake_up_process(struct task_struct *p)
L
Linus Torvalds 已提交
1688
{
1689
	return try_to_wake_up(p, TASK_ALL, 0);
L
Linus Torvalds 已提交
1690 1691 1692
}
EXPORT_SYMBOL(wake_up_process);

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

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

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

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

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

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

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

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

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

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

		p->prio = p->normal_prio = __normal_prio(p);
		set_load_weight(p);
1761

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

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

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

1775 1776 1777 1778 1779 1780 1781
	/*
	 * The child is not yet in the pid-hash so no cgroup attach races,
	 * and the cgroup is pinned to this child due to cgroup_fork()
	 * is ran before sched_fork().
	 *
	 * Silence PROVE_RCU.
	 */
1782
	raw_spin_lock_irqsave(&p->pi_lock, flags);
1783
	set_task_cpu(p, cpu);
1784
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
1785

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

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

/*
 * wake_up_new_task - wake up a newly created task for the first time.
 *
 * This function will do some initial scheduler statistics housekeeping
 * that must be done for every newly created context, then puts the task
 * on the runqueue and wakes it.
 */
1811
void wake_up_new_task(struct task_struct *p)
L
Linus Torvalds 已提交
1812 1813
{
	unsigned long flags;
I
Ingo Molnar 已提交
1814
	struct rq *rq;
1815

1816
	raw_spin_lock_irqsave(&p->pi_lock, flags);
1817 1818 1819 1820 1821 1822
#ifdef CONFIG_SMP
	/*
	 * Fork balancing, do it here and not earlier because:
	 *  - cpus_allowed can change in the fork path
	 *  - any previously selected cpu might disappear through hotplug
	 */
1823
	set_task_cpu(p, select_task_rq(p, SD_BALANCE_FORK, 0));
1824 1825
#endif

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

1838 1839 1840
#ifdef CONFIG_PREEMPT_NOTIFIERS

/**
1841
 * preempt_notifier_register - tell me when current is being preempted & rescheduled
R
Randy Dunlap 已提交
1842
 * @notifier: notifier struct to register
1843 1844 1845 1846 1847 1848 1849 1850 1851
 */
void preempt_notifier_register(struct preempt_notifier *notifier)
{
	hlist_add_head(&notifier->link, &current->preempt_notifiers);
}
EXPORT_SYMBOL_GPL(preempt_notifier_register);

/**
 * preempt_notifier_unregister - no longer interested in preemption notifications
R
Randy Dunlap 已提交
1852
 * @notifier: notifier struct to unregister
1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881
 *
 * This is safe to call from within a preemption notifier.
 */
void preempt_notifier_unregister(struct preempt_notifier *notifier)
{
	hlist_del(&notifier->link);
}
EXPORT_SYMBOL_GPL(preempt_notifier_unregister);

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

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

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

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

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

static void fire_sched_in_preempt_notifiers(struct task_struct *curr)
{
}

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

1894
#endif /* CONFIG_PREEMPT_NOTIFIERS */
1895

1896 1897 1898
/**
 * prepare_task_switch - prepare to switch tasks
 * @rq: the runqueue preparing to switch
R
Randy Dunlap 已提交
1899
 * @prev: the current task that is being switched out
1900 1901 1902 1903 1904 1905 1906 1907 1908
 * @next: the task we are going to switch to.
 *
 * This is called with the rq lock held and interrupts off. It must
 * be paired with a subsequent finish_task_switch after the context
 * switch.
 *
 * prepare_task_switch sets up locking and calls architecture specific
 * hooks.
 */
1909 1910 1911
static inline void
prepare_task_switch(struct rq *rq, struct task_struct *prev,
		    struct task_struct *next)
1912
{
1913
	trace_sched_switch(prev, next);
1914 1915
	sched_info_switch(prev, next);
	perf_event_task_sched_out(prev, next);
1916
	fire_sched_out_preempt_notifiers(prev, next);
1917 1918 1919 1920
	prepare_lock_switch(rq, next);
	prepare_arch_switch(next);
}

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

	rq->prev_mm = NULL;

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

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

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

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

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

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

		rq->post_schedule = 0;
	}
}

#else
2005

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

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

2014 2015
#endif

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

2025
	finish_task_switch(rq, prev);
2026

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

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

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

2051
	prepare_task_switch(rq, prev, next);
2052

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

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

2069
	if (!prev->mm) {
L
Linus Torvalds 已提交
2070 2071 2072
		prev->active_mm = NULL;
		rq->prev_mm = oldmm;
	}
2073 2074 2075 2076 2077 2078 2079
	/*
	 * Since the runqueue lock will be released by the next
	 * task (which is an invalid locking op but in the case
	 * of the scheduler it's an obvious special-case), so we
	 * do an early lockdep release here:
	 */
#ifndef __ARCH_WANT_UNLOCKED_CTXSW
2080
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
2081
#endif
L
Linus Torvalds 已提交
2082 2083

	/* Here we just switch the register state and the stack. */
2084
	rcu_switch(prev, next);
L
Linus Torvalds 已提交
2085 2086
	switch_to(prev, next, prev);

I
Ingo Molnar 已提交
2087 2088 2089 2090 2091 2092 2093
	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
2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110
}

/*
 * 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;
2111
}
L
Linus Torvalds 已提交
2112 2113

unsigned long nr_uninterruptible(void)
2114
{
L
Linus Torvalds 已提交
2115
	unsigned long i, sum = 0;
2116

2117
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2118
		sum += cpu_rq(i)->nr_uninterruptible;
2119 2120

	/*
L
Linus Torvalds 已提交
2121 2122
	 * Since we read the counters lockless, it might be slightly
	 * inaccurate. Do not allow it to go below zero though:
2123
	 */
L
Linus Torvalds 已提交
2124 2125
	if (unlikely((long)sum < 0))
		sum = 0;
2126

L
Linus Torvalds 已提交
2127
	return sum;
2128 2129
}

L
Linus Torvalds 已提交
2130
unsigned long long nr_context_switches(void)
2131
{
2132 2133
	int i;
	unsigned long long sum = 0;
2134

2135
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2136
		sum += cpu_rq(i)->nr_switches;
2137

L
Linus Torvalds 已提交
2138 2139
	return sum;
}
2140

L
Linus Torvalds 已提交
2141 2142 2143
unsigned long nr_iowait(void)
{
	unsigned long i, sum = 0;
2144

2145
	for_each_possible_cpu(i)
L
Linus Torvalds 已提交
2146
		sum += atomic_read(&cpu_rq(i)->nr_iowait);
2147

L
Linus Torvalds 已提交
2148 2149
	return sum;
}
2150

2151
unsigned long nr_iowait_cpu(int cpu)
2152
{
2153
	struct rq *this = cpu_rq(cpu);
2154 2155
	return atomic_read(&this->nr_iowait);
}
2156

2157 2158 2159 2160 2161
unsigned long this_cpu_load(void)
{
	struct rq *this = this_rq();
	return this->cpu_load[0];
}
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
/*
 * Global load-average calculations
 *
 * We take a distributed and async approach to calculating the global load-avg
 * in order to minimize overhead.
 *
 * The global load average is an exponentially decaying average of nr_running +
 * nr_uninterruptible.
 *
 * Once every LOAD_FREQ:
 *
 *   nr_active = 0;
 *   for_each_possible_cpu(cpu)
 *   	nr_active += cpu_of(cpu)->nr_running + cpu_of(cpu)->nr_uninterruptible;
 *
 *   avenrun[n] = avenrun[0] * exp_n + nr_active * (1 - exp_n)
 *
 * Due to a number of reasons the above turns in the mess below:
 *
 *  - for_each_possible_cpu() is prohibitively expensive on machines with
 *    serious number of cpus, therefore we need to take a distributed approach
 *    to calculating nr_active.
 *
 *        \Sum_i x_i(t) = \Sum_i x_i(t) - x_i(t_0) | x_i(t_0) := 0
 *                      = \Sum_i { \Sum_j=1 x_i(t_j) - x_i(t_j-1) }
 *
 *    So assuming nr_active := 0 when we start out -- true per definition, we
 *    can simply take per-cpu deltas and fold those into a global accumulate
 *    to obtain the same result. See calc_load_fold_active().
 *
 *    Furthermore, in order to avoid synchronizing all per-cpu delta folding
 *    across the machine, we assume 10 ticks is sufficient time for every
 *    cpu to have completed this task.
 *
 *    This places an upper-bound on the IRQ-off latency of the machine. Then
 *    again, being late doesn't loose the delta, just wrecks the sample.
 *
 *  - cpu_rq()->nr_uninterruptible isn't accurately tracked per-cpu because
 *    this would add another cross-cpu cacheline miss and atomic operation
 *    to the wakeup path. Instead we increment on whatever cpu the task ran
 *    when it went into uninterruptible state and decrement on whatever cpu
 *    did the wakeup. This means that only the sum of nr_uninterruptible over
 *    all cpus yields the correct result.
 *
 *  This covers the NO_HZ=n code, for extra head-aches, see the comment below.
 */

2211 2212 2213 2214
/* Variables and functions for calc_load */
static atomic_long_t calc_load_tasks;
static unsigned long calc_load_update;
unsigned long avenrun[3];
2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230
EXPORT_SYMBOL(avenrun); /* should be removed */

/**
 * get_avenrun - get the load average array
 * @loads:	pointer to dest load array
 * @offset:	offset to add
 * @shift:	shift count to shift the result left
 *
 * These values are estimates at best, so no need for locking.
 */
void get_avenrun(unsigned long *loads, unsigned long offset, int shift)
{
	loads[0] = (avenrun[0] + offset) << shift;
	loads[1] = (avenrun[1] + offset) << shift;
	loads[2] = (avenrun[2] + offset) << shift;
}
2231

2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246
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;
}

2247 2248 2249
/*
 * a1 = a0 * e + a * (1 - e)
 */
2250 2251 2252 2253 2254 2255 2256 2257 2258
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;
}

2259 2260
#ifdef CONFIG_NO_HZ
/*
2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298
 * Handle NO_HZ for the global load-average.
 *
 * Since the above described distributed algorithm to compute the global
 * load-average relies on per-cpu sampling from the tick, it is affected by
 * NO_HZ.
 *
 * The basic idea is to fold the nr_active delta into a global idle-delta upon
 * entering NO_HZ state such that we can include this as an 'extra' cpu delta
 * when we read the global state.
 *
 * Obviously reality has to ruin such a delightfully simple scheme:
 *
 *  - When we go NO_HZ idle during the window, we can negate our sample
 *    contribution, causing under-accounting.
 *
 *    We avoid this by keeping two idle-delta counters and flipping them
 *    when the window starts, thus separating old and new NO_HZ load.
 *
 *    The only trick is the slight shift in index flip for read vs write.
 *
 *        0s            5s            10s           15s
 *          +10           +10           +10           +10
 *        |-|-----------|-|-----------|-|-----------|-|
 *    r:0 0 1           1 0           0 1           1 0
 *    w:0 1 1           0 0           1 1           0 0
 *
 *    This ensures we'll fold the old idle contribution in this window while
 *    accumlating the new one.
 *
 *  - When we wake up from NO_HZ idle during the window, we push up our
 *    contribution, since we effectively move our sample point to a known
 *    busy state.
 *
 *    This is solved by pushing the window forward, and thus skipping the
 *    sample, for this cpu (effectively using the idle-delta for this cpu which
 *    was in effect at the time the window opened). This also solves the issue
 *    of having to deal with a cpu having been in NOHZ idle for multiple
 *    LOAD_FREQ intervals.
2299 2300 2301
 *
 * When making the ILB scale, we should try to pull this in as well.
 */
2302 2303
static atomic_long_t calc_load_idle[2];
static int calc_load_idx;
2304

2305
static inline int calc_load_write_idx(void)
2306
{
2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332
	int idx = calc_load_idx;

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

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

	return idx & 1;
}

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

void calc_load_enter_idle(void)
{
	struct rq *this_rq = this_rq();
2333 2334
	long delta;

2335 2336 2337 2338
	/*
	 * We're going into NOHZ mode, if there's any pending delta, fold it
	 * into the pending idle delta.
	 */
2339
	delta = calc_load_fold_active(this_rq);
2340 2341 2342 2343
	if (delta) {
		int idx = calc_load_write_idx();
		atomic_long_add(delta, &calc_load_idle[idx]);
	}
2344 2345
}

2346
void calc_load_exit_idle(void)
2347
{
2348 2349 2350 2351 2352 2353 2354
	struct rq *this_rq = this_rq();

	/*
	 * If we're still before the sample window, we're done.
	 */
	if (time_before(jiffies, this_rq->calc_load_update))
		return;
2355 2356

	/*
2357 2358 2359
	 * We woke inside or after the sample window, this means we're already
	 * accounted through the nohz accounting, so skip the entire deal and
	 * sync up for the next window.
2360
	 */
2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372
	this_rq->calc_load_update = calc_load_update;
	if (time_before(jiffies, this_rq->calc_load_update + 10))
		this_rq->calc_load_update += LOAD_FREQ;
}

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

	if (atomic_long_read(&calc_load_idle[idx]))
		delta = atomic_long_xchg(&calc_load_idle[idx], 0);
2373 2374 2375

	return delta;
}
2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453

/**
 * 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.
 */
2454
static void calc_global_nohz(void)
2455 2456 2457
{
	long delta, active, n;

2458 2459 2460 2461 2462 2463
	if (!time_before(jiffies, calc_load_update + 10)) {
		/*
		 * Catch-up, fold however many we are behind still
		 */
		delta = jiffies - calc_load_update - 10;
		n = 1 + (delta / LOAD_FREQ);
2464

2465 2466
		active = atomic_long_read(&calc_load_tasks);
		active = active > 0 ? active * FIXED_1 : 0;
2467

2468 2469 2470
		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);
2471

2472 2473
		calc_load_update += n * LOAD_FREQ;
	}
2474

2475 2476 2477 2478 2479 2480 2481 2482 2483
	/*
	 * Flip the idle index...
	 *
	 * Make sure we first write the new time then flip the index, so that
	 * calc_load_write_idx() will see the new time when it reads the new
	 * index, this avoids a double flip messing things up.
	 */
	smp_wmb();
	calc_load_idx++;
2484
}
2485
#else /* !CONFIG_NO_HZ */
2486

2487 2488
static inline long calc_load_fold_idle(void) { return 0; }
static inline void calc_global_nohz(void) { }
2489

2490
#endif /* CONFIG_NO_HZ */
2491 2492

/*
2493 2494
 * calc_load - update the avenrun load estimates 10 ticks after the
 * CPUs have updated calc_load_tasks.
2495
 */
2496
void calc_global_load(unsigned long ticks)
2497
{
2498
	long active, delta;
L
Linus Torvalds 已提交
2499

2500
	if (time_before(jiffies, calc_load_update + 10))
2501
		return;
L
Linus Torvalds 已提交
2502

2503 2504 2505 2506 2507 2508 2509
	/*
	 * Fold the 'old' idle-delta to include all NO_HZ cpus.
	 */
	delta = calc_load_fold_idle();
	if (delta)
		atomic_long_add(delta, &calc_load_tasks);

2510 2511
	active = atomic_long_read(&calc_load_tasks);
	active = active > 0 ? active * FIXED_1 : 0;
L
Linus Torvalds 已提交
2512

2513 2514 2515
	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
2516

2517
	calc_load_update += LOAD_FREQ;
2518 2519

	/*
2520
	 * In case we idled for multiple LOAD_FREQ intervals, catch up in bulk.
2521 2522
	 */
	calc_global_nohz();
2523
}
L
Linus Torvalds 已提交
2524

2525
/*
2526 2527
 * Called from update_cpu_load() to periodically update this CPU's
 * active count.
2528 2529 2530
 */
static void calc_load_account_active(struct rq *this_rq)
{
2531
	long delta;
2532

2533 2534
	if (time_before(jiffies, this_rq->calc_load_update))
		return;
2535

2536 2537
	delta  = calc_load_fold_active(this_rq);
	if (delta)
2538
		atomic_long_add(delta, &calc_load_tasks);
2539 2540

	this_rq->calc_load_update += LOAD_FREQ;
2541 2542
}

2543 2544 2545 2546
/*
 * End of global load-average stuff
 */

2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613
/*
 * 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;
}

2614
/*
I
Ingo Molnar 已提交
2615
 * Update rq->cpu_load[] statistics. This function is usually called every
2616 2617
 * scheduler tick (TICK_NSEC). With tickless idle this will not be called
 * every tick. We fix it up based on jiffies.
2618
 */
2619 2620
static void __update_cpu_load(struct rq *this_rq, unsigned long this_load,
			      unsigned long pending_updates)
2621
{
I
Ingo Molnar 已提交
2622
	int i, scale;
2623

I
Ingo Molnar 已提交
2624
	this_rq->nr_load_updates++;
2625

I
Ingo Molnar 已提交
2626
	/* Update our load: */
2627 2628
	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
2629
		unsigned long old_load, new_load;
2630

I
Ingo Molnar 已提交
2631
		/* scale is effectively 1 << i now, and >> i divides by scale */
2632

I
Ingo Molnar 已提交
2633
		old_load = this_rq->cpu_load[i];
2634
		old_load = decay_load_missed(old_load, pending_updates - 1, i);
I
Ingo Molnar 已提交
2635
		new_load = this_load;
I
Ingo Molnar 已提交
2636 2637 2638 2639 2640 2641
		/*
		 * 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)
2642 2643 2644
			new_load += scale - 1;

		this_rq->cpu_load[i] = (old_load * (scale - 1) + new_load) >> i;
I
Ingo Molnar 已提交
2645
	}
2646 2647

	sched_avg_update(this_rq);
2648 2649
}

2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663
#ifdef CONFIG_NO_HZ
/*
 * There is no sane way to deal with nohz on smp when using jiffies because the
 * cpu doing the jiffies update might drift wrt the cpu doing the jiffy reading
 * causing off-by-one errors in observed deltas; {0,2} instead of {1,1}.
 *
 * Therefore we cannot use the delta approach from the regular tick since that
 * would seriously skew the load calculation. However we'll make do for those
 * updates happening while idle (nohz_idle_balance) or coming out of idle
 * (tick_nohz_idle_exit).
 *
 * This means we might still be one tick off for nohz periods.
 */

2664 2665 2666 2667 2668 2669
/*
 * Called from nohz_idle_balance() to update the load ratings before doing the
 * idle balance.
 */
void update_idle_cpu_load(struct rq *this_rq)
{
2670
	unsigned long curr_jiffies = ACCESS_ONCE(jiffies);
2671 2672 2673 2674
	unsigned long load = this_rq->load.weight;
	unsigned long pending_updates;

	/*
2675
	 * bail if there's load or we're actually up-to-date.
2676 2677 2678 2679 2680 2681 2682 2683 2684 2685
	 */
	if (load || curr_jiffies == this_rq->last_load_update_tick)
		return;

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

	__update_cpu_load(this_rq, load, pending_updates);
}

2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711
/*
 * Called from tick_nohz_idle_exit() -- try and fix up the ticks we missed.
 */
void update_cpu_load_nohz(void)
{
	struct rq *this_rq = this_rq();
	unsigned long curr_jiffies = ACCESS_ONCE(jiffies);
	unsigned long pending_updates;

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

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

2712 2713 2714
/*
 * Called from scheduler_tick()
 */
2715 2716
static void update_cpu_load_active(struct rq *this_rq)
{
2717
	/*
2718
	 * See the mess around update_idle_cpu_load() / update_cpu_load_nohz().
2719 2720 2721
	 */
	this_rq->last_load_update_tick = jiffies;
	__update_cpu_load(this_rq, this_rq->load.weight, 1);
2722

2723
	calc_load_account_active(this_rq);
2724 2725
}

I
Ingo Molnar 已提交
2726
#ifdef CONFIG_SMP
2727

2728
/*
P
Peter Zijlstra 已提交
2729 2730
 * sched_exec - execve() is a valuable balancing opportunity, because at
 * this point the task has the smallest effective memory and cache footprint.
2731
 */
P
Peter Zijlstra 已提交
2732
void sched_exec(void)
2733
{
P
Peter Zijlstra 已提交
2734
	struct task_struct *p = current;
L
Linus Torvalds 已提交
2735
	unsigned long flags;
2736
	int dest_cpu;
2737

2738
	raw_spin_lock_irqsave(&p->pi_lock, flags);
2739
	dest_cpu = p->sched_class->select_task_rq(p, SD_BALANCE_EXEC, 0);
2740 2741
	if (dest_cpu == smp_processor_id())
		goto unlock;
P
Peter Zijlstra 已提交
2742

2743
	if (likely(cpu_active(dest_cpu))) {
2744
		struct migration_arg arg = { p, dest_cpu };
2745

2746 2747
		raw_spin_unlock_irqrestore(&p->pi_lock, flags);
		stop_one_cpu(task_cpu(p), migration_cpu_stop, &arg);
L
Linus Torvalds 已提交
2748 2749
		return;
	}
2750
unlock:
2751
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
2752
}
I
Ingo Molnar 已提交
2753

L
Linus Torvalds 已提交
2754 2755 2756
#endif

DEFINE_PER_CPU(struct kernel_stat, kstat);
2757
DEFINE_PER_CPU(struct kernel_cpustat, kernel_cpustat);
L
Linus Torvalds 已提交
2758 2759

EXPORT_PER_CPU_SYMBOL(kstat);
2760
EXPORT_PER_CPU_SYMBOL(kernel_cpustat);
L
Linus Torvalds 已提交
2761 2762

/*
2763
 * Return any ns on the sched_clock that have not yet been accounted in
2764
 * @p in case that task is currently running.
2765 2766
 *
 * Called with task_rq_lock() held on @rq.
L
Linus Torvalds 已提交
2767
 */
2768 2769 2770 2771 2772 2773
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);
2774
		ns = rq->clock_task - p->se.exec_start;
2775 2776 2777 2778 2779 2780 2781
		if ((s64)ns < 0)
			ns = 0;
	}

	return ns;
}

2782
unsigned long long task_delta_exec(struct task_struct *p)
L
Linus Torvalds 已提交
2783 2784
{
	unsigned long flags;
2785
	struct rq *rq;
2786
	u64 ns = 0;
2787

2788
	rq = task_rq_lock(p, &flags);
2789
	ns = do_task_delta_exec(p, rq);
2790
	task_rq_unlock(rq, p, &flags);
2791

2792 2793
	return ns;
}
2794

2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807
/*
 * 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);
2808
	task_rq_unlock(rq, p, &flags);
2809 2810 2811

	return ns;
}
2812

2813 2814 2815 2816 2817
#ifdef CONFIG_CGROUP_CPUACCT
struct cgroup_subsys cpuacct_subsys;
struct cpuacct root_cpuacct;
#endif

2818 2819
static inline void task_group_account_field(struct task_struct *p, int index,
					    u64 tmp)
2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848
{
#ifdef CONFIG_CGROUP_CPUACCT
	struct kernel_cpustat *kcpustat;
	struct cpuacct *ca;
#endif
	/*
	 * Since all updates are sure to touch the root cgroup, we
	 * get ourselves ahead and touch it first. If the root cgroup
	 * is the only cgroup, then nothing else should be necessary.
	 *
	 */
	__get_cpu_var(kernel_cpustat).cpustat[index] += tmp;

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

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


L
Linus Torvalds 已提交
2849 2850 2851 2852
/*
 * 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
2853
 * @cputime_scaled: cputime scaled by cpu frequency
L
Linus Torvalds 已提交
2854
 */
2855 2856
void account_user_time(struct task_struct *p, cputime_t cputime,
		       cputime_t cputime_scaled)
L
Linus Torvalds 已提交
2857
{
2858
	int index;
L
Linus Torvalds 已提交
2859

2860
	/* Add user time to process. */
2861 2862
	p->utime += cputime;
	p->utimescaled += cputime_scaled;
2863
	account_group_user_time(p, cputime);
L
Linus Torvalds 已提交
2864

2865
	index = (TASK_NICE(p) > 0) ? CPUTIME_NICE : CPUTIME_USER;
2866

L
Linus Torvalds 已提交
2867
	/* Add user time to cpustat. */
2868
	task_group_account_field(p, index, (__force u64) cputime);
2869

2870 2871
	/* Account for user time used */
	acct_update_integrals(p);
L
Linus Torvalds 已提交
2872 2873
}

2874 2875 2876 2877
/*
 * 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
2878
 * @cputime_scaled: cputime scaled by cpu frequency
2879
 */
2880 2881
static void account_guest_time(struct task_struct *p, cputime_t cputime,
			       cputime_t cputime_scaled)
2882
{
2883
	u64 *cpustat = kcpustat_this_cpu->cpustat;
2884

2885
	/* Add guest time to process. */
2886 2887
	p->utime += cputime;
	p->utimescaled += cputime_scaled;
2888
	account_group_user_time(p, cputime);
2889
	p->gtime += cputime;
2890

2891
	/* Add guest time to cpustat. */
2892
	if (TASK_NICE(p) > 0) {
2893 2894
		cpustat[CPUTIME_NICE] += (__force u64) cputime;
		cpustat[CPUTIME_GUEST_NICE] += (__force u64) cputime;
2895
	} else {
2896 2897
		cpustat[CPUTIME_USER] += (__force u64) cputime;
		cpustat[CPUTIME_GUEST] += (__force u64) cputime;
2898
	}
2899 2900
}

2901 2902 2903 2904 2905 2906 2907 2908 2909
/*
 * 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,
2910
			cputime_t cputime_scaled, int index)
2911 2912
{
	/* Add system time to process. */
2913 2914
	p->stime += cputime;
	p->stimescaled += cputime_scaled;
2915 2916 2917
	account_group_system_time(p, cputime);

	/* Add system time to cpustat. */
2918
	task_group_account_field(p, index, (__force u64) cputime);
2919 2920 2921 2922 2923

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

L
Linus Torvalds 已提交
2924 2925 2926 2927 2928
/*
 * 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
2929
 * @cputime_scaled: cputime scaled by cpu frequency
L
Linus Torvalds 已提交
2930 2931
 */
void account_system_time(struct task_struct *p, int hardirq_offset,
2932
			 cputime_t cputime, cputime_t cputime_scaled)
L
Linus Torvalds 已提交
2933
{
2934
	int index;
L
Linus Torvalds 已提交
2935

2936
	if ((p->flags & PF_VCPU) && (irq_count() - hardirq_offset == 0)) {
2937
		account_guest_time(p, cputime, cputime_scaled);
2938 2939
		return;
	}
2940

L
Linus Torvalds 已提交
2941
	if (hardirq_count() - hardirq_offset)
2942
		index = CPUTIME_IRQ;
2943
	else if (in_serving_softirq())
2944
		index = CPUTIME_SOFTIRQ;
L
Linus Torvalds 已提交
2945
	else
2946
		index = CPUTIME_SYSTEM;
2947

2948
	__account_system_time(p, cputime, cputime_scaled, index);
L
Linus Torvalds 已提交
2949 2950
}

2951
/*
L
Linus Torvalds 已提交
2952
 * Account for involuntary wait time.
2953
 * @cputime: the cpu time spent in involuntary wait
2954
 */
2955
void account_steal_time(cputime_t cputime)
2956
{
2957
	u64 *cpustat = kcpustat_this_cpu->cpustat;
2958

2959
	cpustat[CPUTIME_STEAL] += (__force u64) cputime;
2960 2961
}

L
Linus Torvalds 已提交
2962
/*
2963 2964
 * Account for idle time.
 * @cputime: the cpu time spent in idle wait
L
Linus Torvalds 已提交
2965
 */
2966
void account_idle_time(cputime_t cputime)
L
Linus Torvalds 已提交
2967
{
2968
	u64 *cpustat = kcpustat_this_cpu->cpustat;
2969
	struct rq *rq = this_rq();
L
Linus Torvalds 已提交
2970

2971
	if (atomic_read(&rq->nr_iowait) > 0)
2972
		cpustat[CPUTIME_IOWAIT] += (__force u64) cputime;
2973
	else
2974
		cpustat[CPUTIME_IDLE] += (__force u64) cputime;
L
Linus Torvalds 已提交
2975 2976
}

G
Glauber Costa 已提交
2977 2978 2979
static __always_inline bool steal_account_process_tick(void)
{
#ifdef CONFIG_PARAVIRT
2980
	if (static_key_false(&paravirt_steal_enabled)) {
G
Glauber Costa 已提交
2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995
		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;
}

2996 2997
#ifndef CONFIG_VIRT_CPU_ACCOUNTING

2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023
#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);
3024
	u64 *cpustat = kcpustat_this_cpu->cpustat;
3025

G
Glauber Costa 已提交
3026 3027 3028
	if (steal_account_process_tick())
		return;

3029
	if (irqtime_account_hi_update()) {
3030
		cpustat[CPUTIME_IRQ] += (__force u64) cputime_one_jiffy;
3031
	} else if (irqtime_account_si_update()) {
3032
		cpustat[CPUTIME_SOFTIRQ] += (__force u64) cputime_one_jiffy;
3033 3034 3035 3036 3037 3038 3039
	} 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,
3040
					CPUTIME_SOFTIRQ);
3041 3042 3043 3044 3045 3046 3047 3048
	} 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,
3049
					CPUTIME_SYSTEM);
3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060
	}
}

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);
}
3061
#else /* CONFIG_IRQ_TIME_ACCOUNTING */
3062 3063 3064
static void irqtime_account_idle_ticks(int ticks) {}
static void irqtime_account_process_tick(struct task_struct *p, int user_tick,
						struct rq *rq) {}
3065
#endif /* CONFIG_IRQ_TIME_ACCOUNTING */
3066 3067 3068 3069 3070 3071 3072 3073

/*
 * 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)
{
3074
	cputime_t one_jiffy_scaled = cputime_to_scaled(cputime_one_jiffy);
3075 3076
	struct rq *rq = this_rq();

3077 3078 3079 3080 3081
	if (sched_clock_irqtime) {
		irqtime_account_process_tick(p, user_tick, rq);
		return;
	}

G
Glauber Costa 已提交
3082 3083 3084
	if (steal_account_process_tick())
		return;

3085
	if (user_tick)
3086
		account_user_time(p, cputime_one_jiffy, one_jiffy_scaled);
3087
	else if ((p != rq->idle) || (irq_count() != HARDIRQ_OFFSET))
3088
		account_system_time(p, HARDIRQ_OFFSET, cputime_one_jiffy,
3089 3090
				    one_jiffy_scaled);
	else
3091
		account_idle_time(cputime_one_jiffy);
3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109
}

/*
 * 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)
{
3110 3111 3112 3113 3114 3115

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

3116
	account_idle_time(jiffies_to_cputime(ticks));
L
Linus Torvalds 已提交
3117 3118
}

3119 3120
#endif

3121 3122 3123 3124
/*
 * Use precise platform statistics if available:
 */
#ifdef CONFIG_VIRT_CPU_ACCOUNTING
3125
void task_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
3126
{
3127 3128
	*ut = p->utime;
	*st = p->stime;
3129 3130
}

3131
void thread_group_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
3132
{
3133 3134 3135 3136 3137 3138
	struct task_cputime cputime;

	thread_group_cputime(p, &cputime);

	*ut = cputime.utime;
	*st = cputime.stime;
3139 3140
}
#else
3141 3142

#ifndef nsecs_to_cputime
3143
# define nsecs_to_cputime(__nsecs)	nsecs_to_jiffies(__nsecs)
3144 3145
#endif

3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159
static cputime_t scale_utime(cputime_t utime, cputime_t rtime, cputime_t total)
{
	u64 temp = (__force u64) rtime;

	temp *= (__force u64) utime;

	if (sizeof(cputime_t) == 4)
		temp = div_u64(temp, (__force u32) total);
	else
		temp = div64_u64(temp, (__force u64) total);

	return (__force cputime_t) temp;
}

3160
void task_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
3161
{
3162
	cputime_t rtime, utime = p->utime, total = utime + p->stime;
3163 3164 3165 3166

	/*
	 * Use CFS's precise accounting:
	 */
3167
	rtime = nsecs_to_cputime(p->se.sum_exec_runtime);
3168

3169 3170 3171
	if (total)
		utime = scale_utime(utime, rtime, total);
	else
3172
		utime = rtime;
3173

3174 3175 3176
	/*
	 * Compare with previous values, to keep monotonicity:
	 */
3177
	p->prev_utime = max(p->prev_utime, utime);
3178
	p->prev_stime = max(p->prev_stime, rtime - p->prev_utime);
3179

3180 3181
	*ut = p->prev_utime;
	*st = p->prev_stime;
3182 3183
}

3184 3185 3186 3187
/*
 * Must be called with siglock held.
 */
void thread_group_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
3188
{
3189 3190 3191
	struct signal_struct *sig = p->signal;
	struct task_cputime cputime;
	cputime_t rtime, utime, total;
3192

3193
	thread_group_cputime(p, &cputime);
3194

3195
	total = cputime.utime + cputime.stime;
3196
	rtime = nsecs_to_cputime(cputime.sum_exec_runtime);
3197

3198 3199 3200
	if (total)
		utime = scale_utime(cputime.utime, rtime, total);
	else
3201 3202 3203
		utime = rtime;

	sig->prev_utime = max(sig->prev_utime, utime);
3204
	sig->prev_stime = max(sig->prev_stime, rtime - sig->prev_utime);
3205 3206 3207

	*ut = sig->prev_utime;
	*st = sig->prev_stime;
3208 3209 3210
}
#endif

3211 3212 3213 3214 3215 3216 3217 3218
/*
 * 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
3219
	struct task_struct *curr = rq->curr;
3220 3221

	sched_clock_tick();
I
Ingo Molnar 已提交
3222

3223
	raw_spin_lock(&rq->lock);
3224
	update_rq_clock(rq);
3225
	update_cpu_load_active(rq);
P
Peter Zijlstra 已提交
3226
	curr->sched_class->task_tick(rq, curr, 0);
3227
	raw_spin_unlock(&rq->lock);
3228

3229
	perf_event_task_tick();
3230

3231
#ifdef CONFIG_SMP
3232
	rq->idle_balance = idle_cpu(cpu);
I
Ingo Molnar 已提交
3233
	trigger_load_balance(rq, cpu);
3234
#endif
L
Linus Torvalds 已提交
3235 3236
}

3237
notrace unsigned long get_parent_ip(unsigned long addr)
3238 3239 3240 3241 3242 3243 3244 3245
{
	if (in_lock_functions(addr)) {
		addr = CALLER_ADDR2;
		if (in_lock_functions(addr))
			addr = CALLER_ADDR3;
	}
	return addr;
}
L
Linus Torvalds 已提交
3246

3247 3248 3249
#if defined(CONFIG_PREEMPT) && (defined(CONFIG_DEBUG_PREEMPT) || \
				defined(CONFIG_PREEMPT_TRACER))

3250
void __kprobes add_preempt_count(int val)
L
Linus Torvalds 已提交
3251
{
3252
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3253 3254 3255
	/*
	 * Underflow?
	 */
3256 3257
	if (DEBUG_LOCKS_WARN_ON((preempt_count() < 0)))
		return;
3258
#endif
L
Linus Torvalds 已提交
3259
	preempt_count() += val;
3260
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3261 3262 3263
	/*
	 * Spinlock count overflowing soon?
	 */
3264 3265
	DEBUG_LOCKS_WARN_ON((preempt_count() & PREEMPT_MASK) >=
				PREEMPT_MASK - 10);
3266 3267 3268
#endif
	if (preempt_count() == val)
		trace_preempt_off(CALLER_ADDR0, get_parent_ip(CALLER_ADDR1));
L
Linus Torvalds 已提交
3269 3270 3271
}
EXPORT_SYMBOL(add_preempt_count);

3272
void __kprobes sub_preempt_count(int val)
L
Linus Torvalds 已提交
3273
{
3274
#ifdef CONFIG_DEBUG_PREEMPT
L
Linus Torvalds 已提交
3275 3276 3277
	/*
	 * Underflow?
	 */
3278
	if (DEBUG_LOCKS_WARN_ON(val > preempt_count()))
3279
		return;
L
Linus Torvalds 已提交
3280 3281 3282
	/*
	 * Is the spinlock portion underflowing?
	 */
3283 3284 3285
	if (DEBUG_LOCKS_WARN_ON((val < PREEMPT_MASK) &&
			!(preempt_count() & PREEMPT_MASK)))
		return;
3286
#endif
3287

3288 3289
	if (preempt_count() == val)
		trace_preempt_on(CALLER_ADDR0, get_parent_ip(CALLER_ADDR1));
L
Linus Torvalds 已提交
3290 3291 3292 3293 3294 3295 3296
	preempt_count() -= val;
}
EXPORT_SYMBOL(sub_preempt_count);

#endif

/*
I
Ingo Molnar 已提交
3297
 * Print scheduling while atomic bug:
L
Linus Torvalds 已提交
3298
 */
I
Ingo Molnar 已提交
3299
static noinline void __schedule_bug(struct task_struct *prev)
L
Linus Torvalds 已提交
3300
{
3301 3302 3303
	if (oops_in_progress)
		return;

P
Peter Zijlstra 已提交
3304 3305
	printk(KERN_ERR "BUG: scheduling while atomic: %s/%d/0x%08x\n",
		prev->comm, prev->pid, preempt_count());
3306

I
Ingo Molnar 已提交
3307
	debug_show_held_locks(prev);
3308
	print_modules();
I
Ingo Molnar 已提交
3309 3310
	if (irqs_disabled())
		print_irqtrace_events(prev);
3311
	dump_stack();
3312
	add_taint(TAINT_WARN);
I
Ingo Molnar 已提交
3313
}
L
Linus Torvalds 已提交
3314

I
Ingo Molnar 已提交
3315 3316 3317 3318 3319
/*
 * Various schedule()-time debugging checks and statistics:
 */
static inline void schedule_debug(struct task_struct *prev)
{
L
Linus Torvalds 已提交
3320
	/*
I
Ingo Molnar 已提交
3321
	 * Test if we are atomic. Since do_exit() needs to call into
L
Linus Torvalds 已提交
3322 3323 3324
	 * schedule() atomically, we ignore that path for now.
	 * Otherwise, whine if we are scheduling when we should not be.
	 */
3325
	if (unlikely(in_atomic_preempt_off() && !prev->exit_state))
I
Ingo Molnar 已提交
3326
		__schedule_bug(prev);
3327
	rcu_sleep_check();
I
Ingo Molnar 已提交
3328

L
Linus Torvalds 已提交
3329 3330
	profile_hit(SCHED_PROFILING, __builtin_return_address(0));

3331
	schedstat_inc(this_rq(), sched_count);
I
Ingo Molnar 已提交
3332 3333
}

P
Peter Zijlstra 已提交
3334
static void put_prev_task(struct rq *rq, struct task_struct *prev)
M
Mike Galbraith 已提交
3335
{
3336
	if (prev->on_rq || rq->skip_clock_update < 0)
3337
		update_rq_clock(rq);
P
Peter Zijlstra 已提交
3338
	prev->sched_class->put_prev_task(rq, prev);
M
Mike Galbraith 已提交
3339 3340
}

I
Ingo Molnar 已提交
3341 3342 3343 3344
/*
 * Pick up the highest-prio task:
 */
static inline struct task_struct *
3345
pick_next_task(struct rq *rq)
I
Ingo Molnar 已提交
3346
{
3347
	const struct sched_class *class;
I
Ingo Molnar 已提交
3348
	struct task_struct *p;
L
Linus Torvalds 已提交
3349 3350

	/*
I
Ingo Molnar 已提交
3351 3352
	 * Optimization: we know that if all tasks are in
	 * the fair class we can call that function directly:
L
Linus Torvalds 已提交
3353
	 */
3354
	if (likely(rq->nr_running == rq->cfs.h_nr_running)) {
3355
		p = fair_sched_class.pick_next_task(rq);
I
Ingo Molnar 已提交
3356 3357
		if (likely(p))
			return p;
L
Linus Torvalds 已提交
3358 3359
	}

3360
	for_each_class(class) {
3361
		p = class->pick_next_task(rq);
I
Ingo Molnar 已提交
3362 3363 3364
		if (p)
			return p;
	}
3365 3366

	BUG(); /* the idle class will always have a runnable task */
I
Ingo Molnar 已提交
3367
}
L
Linus Torvalds 已提交
3368

I
Ingo Molnar 已提交
3369
/*
3370
 * __schedule() is the main scheduler function.
I
Ingo Molnar 已提交
3371
 */
3372
static void __sched __schedule(void)
I
Ingo Molnar 已提交
3373 3374
{
	struct task_struct *prev, *next;
3375
	unsigned long *switch_count;
I
Ingo Molnar 已提交
3376
	struct rq *rq;
3377
	int cpu;
I
Ingo Molnar 已提交
3378

3379 3380
need_resched:
	preempt_disable();
I
Ingo Molnar 已提交
3381 3382
	cpu = smp_processor_id();
	rq = cpu_rq(cpu);
3383
	rcu_note_context_switch(cpu);
I
Ingo Molnar 已提交
3384 3385 3386
	prev = rq->curr;

	schedule_debug(prev);
L
Linus Torvalds 已提交
3387

3388
	if (sched_feat(HRTICK))
M
Mike Galbraith 已提交
3389
		hrtick_clear(rq);
P
Peter Zijlstra 已提交
3390

3391
	raw_spin_lock_irq(&rq->lock);
L
Linus Torvalds 已提交
3392

3393
	switch_count = &prev->nivcsw;
L
Linus Torvalds 已提交
3394
	if (prev->state && !(preempt_count() & PREEMPT_ACTIVE)) {
T
Tejun Heo 已提交
3395
		if (unlikely(signal_pending_state(prev->state, prev))) {
L
Linus Torvalds 已提交
3396
			prev->state = TASK_RUNNING;
T
Tejun Heo 已提交
3397
		} else {
3398 3399 3400
			deactivate_task(rq, prev, DEQUEUE_SLEEP);
			prev->on_rq = 0;

T
Tejun Heo 已提交
3401
			/*
3402 3403 3404
			 * 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
3405 3406 3407 3408 3409 3410 3411 3412 3413
			 */
			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
3414
		switch_count = &prev->nvcsw;
L
Linus Torvalds 已提交
3415 3416
	}

3417
	pre_schedule(rq, prev);
3418

I
Ingo Molnar 已提交
3419
	if (unlikely(!rq->nr_running))
L
Linus Torvalds 已提交
3420 3421
		idle_balance(cpu, rq);

M
Mike Galbraith 已提交
3422
	put_prev_task(rq, prev);
3423
	next = pick_next_task(rq);
3424 3425
	clear_tsk_need_resched(prev);
	rq->skip_clock_update = 0;
L
Linus Torvalds 已提交
3426 3427 3428 3429 3430 3431

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

I
Ingo Molnar 已提交
3432
		context_switch(rq, prev, next); /* unlocks the rq */
P
Peter Zijlstra 已提交
3433
		/*
3434 3435 3436 3437
		 * 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
3438 3439 3440
		 */
		cpu = smp_processor_id();
		rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
3441
	} else
3442
		raw_spin_unlock_irq(&rq->lock);
L
Linus Torvalds 已提交
3443

3444
	post_schedule(rq);
L
Linus Torvalds 已提交
3445

3446
	sched_preempt_enable_no_resched();
3447
	if (need_resched())
L
Linus Torvalds 已提交
3448 3449
		goto need_resched;
}
3450

3451 3452
static inline void sched_submit_work(struct task_struct *tsk)
{
3453
	if (!tsk->state || tsk_is_pi_blocked(tsk))
3454 3455 3456 3457 3458 3459 3460 3461 3462
		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
3463
asmlinkage void __sched schedule(void)
3464
{
3465 3466 3467
	struct task_struct *tsk = current;

	sched_submit_work(tsk);
3468 3469
	__schedule();
}
L
Linus Torvalds 已提交
3470 3471
EXPORT_SYMBOL(schedule);

3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486
#ifdef CONFIG_RCU_USER_QS
asmlinkage void __sched schedule_user(void)
{
	/*
	 * If we come here after a random call to set_need_resched(),
	 * or we have been woken up remotely but the IPI has not yet arrived,
	 * we haven't yet exited the RCU idle mode. Do it here manually until
	 * we find a better solution.
	 */
	rcu_user_exit();
	schedule();
	rcu_user_enter();
}
#endif

3487 3488 3489 3490 3491 3492 3493
/**
 * schedule_preempt_disabled - called with preemption disabled
 *
 * Returns with preemption disabled. Note: preempt_count must be 1
 */
void __sched schedule_preempt_disabled(void)
{
3494
	sched_preempt_enable_no_resched();
3495 3496 3497 3498
	schedule();
	preempt_disable();
}

3499
#ifdef CONFIG_MUTEX_SPIN_ON_OWNER
3500

3501 3502 3503
static inline bool owner_running(struct mutex *lock, struct task_struct *owner)
{
	if (lock->owner != owner)
3504
		return false;
3505 3506

	/*
3507 3508 3509 3510
	 * 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.
3511
	 */
3512
	barrier();
3513

3514
	return owner->on_cpu;
3515
}
3516

3517 3518 3519 3520 3521 3522 3523 3524
/*
 * 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;
3525

3526
	rcu_read_lock();
3527 3528
	while (owner_running(lock, owner)) {
		if (need_resched())
3529
			break;
3530

3531
		arch_mutex_cpu_relax();
3532
	}
3533
	rcu_read_unlock();
3534

3535
	/*
3536 3537 3538
	 * 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.
3539
	 */
3540
	return lock->owner == NULL;
3541 3542 3543
}
#endif

L
Linus Torvalds 已提交
3544 3545
#ifdef CONFIG_PREEMPT
/*
3546
 * this is the entry point to schedule() from in-kernel preemption
I
Ingo Molnar 已提交
3547
 * off of preempt_enable. Kernel preemptions off return from interrupt
L
Linus Torvalds 已提交
3548 3549
 * occur there and call schedule directly.
 */
3550
asmlinkage void __sched notrace preempt_schedule(void)
L
Linus Torvalds 已提交
3551 3552
{
	struct thread_info *ti = current_thread_info();
3553

L
Linus Torvalds 已提交
3554 3555
	/*
	 * If there is a non-zero preempt_count or interrupts are disabled,
I
Ingo Molnar 已提交
3556
	 * we do not want to preempt the current task. Just return..
L
Linus Torvalds 已提交
3557
	 */
N
Nick Piggin 已提交
3558
	if (likely(ti->preempt_count || irqs_disabled()))
L
Linus Torvalds 已提交
3559 3560
		return;

3561
	do {
3562
		add_preempt_count_notrace(PREEMPT_ACTIVE);
3563
		__schedule();
3564
		sub_preempt_count_notrace(PREEMPT_ACTIVE);
L
Linus Torvalds 已提交
3565

3566 3567 3568 3569 3570
		/*
		 * Check again in case we missed a preemption opportunity
		 * between schedule and now.
		 */
		barrier();
3571
	} while (need_resched());
L
Linus Torvalds 已提交
3572 3573 3574 3575
}
EXPORT_SYMBOL(preempt_schedule);

/*
3576
 * this is the entry point to schedule() from kernel preemption
L
Linus Torvalds 已提交
3577 3578 3579 3580 3581 3582 3583
 * 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();
3584

3585
	/* Catch callers which need to be fixed */
L
Linus Torvalds 已提交
3586 3587
	BUG_ON(ti->preempt_count || !irqs_disabled());

3588
	rcu_user_exit();
3589 3590 3591
	do {
		add_preempt_count(PREEMPT_ACTIVE);
		local_irq_enable();
3592
		__schedule();
3593 3594
		local_irq_disable();
		sub_preempt_count(PREEMPT_ACTIVE);
L
Linus Torvalds 已提交
3595

3596 3597 3598 3599 3600
		/*
		 * Check again in case we missed a preemption opportunity
		 * between schedule and now.
		 */
		barrier();
3601
	} while (need_resched());
L
Linus Torvalds 已提交
3602 3603 3604 3605
}

#endif /* CONFIG_PREEMPT */

P
Peter Zijlstra 已提交
3606
int default_wake_function(wait_queue_t *curr, unsigned mode, int wake_flags,
I
Ingo Molnar 已提交
3607
			  void *key)
L
Linus Torvalds 已提交
3608
{
P
Peter Zijlstra 已提交
3609
	return try_to_wake_up(curr->private, mode, wake_flags);
L
Linus Torvalds 已提交
3610 3611 3612 3613
}
EXPORT_SYMBOL(default_wake_function);

/*
I
Ingo Molnar 已提交
3614 3615
 * 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
3616 3617 3618
 * 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
3619
 * started to run but is not in state TASK_RUNNING. try_to_wake_up() returns
L
Linus Torvalds 已提交
3620 3621
 * zero in this (rare) case, and we handle it by continuing to scan the queue.
 */
3622
static void __wake_up_common(wait_queue_head_t *q, unsigned int mode,
P
Peter Zijlstra 已提交
3623
			int nr_exclusive, int wake_flags, void *key)
L
Linus Torvalds 已提交
3624
{
3625
	wait_queue_t *curr, *next;
L
Linus Torvalds 已提交
3626

3627
	list_for_each_entry_safe(curr, next, &q->task_list, task_list) {
3628 3629
		unsigned flags = curr->flags;

P
Peter Zijlstra 已提交
3630
		if (curr->func(curr, mode, wake_flags, key) &&
3631
				(flags & WQ_FLAG_EXCLUSIVE) && !--nr_exclusive)
L
Linus Torvalds 已提交
3632 3633 3634 3635 3636 3637 3638 3639 3640
			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
3641
 * @key: is directly passed to the wakeup function
3642 3643 3644
 *
 * 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
3645
 */
3646
void __wake_up(wait_queue_head_t *q, unsigned int mode,
I
Ingo Molnar 已提交
3647
			int nr_exclusive, void *key)
L
Linus Torvalds 已提交
3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659
{
	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.
 */
3660
void __wake_up_locked(wait_queue_head_t *q, unsigned int mode, int nr)
L
Linus Torvalds 已提交
3661
{
3662
	__wake_up_common(q, mode, nr, 0, NULL);
L
Linus Torvalds 已提交
3663
}
3664
EXPORT_SYMBOL_GPL(__wake_up_locked);
L
Linus Torvalds 已提交
3665

3666 3667 3668 3669
void __wake_up_locked_key(wait_queue_head_t *q, unsigned int mode, void *key)
{
	__wake_up_common(q, mode, 1, 0, key);
}
3670
EXPORT_SYMBOL_GPL(__wake_up_locked_key);
3671

L
Linus Torvalds 已提交
3672
/**
3673
 * __wake_up_sync_key - wake up threads blocked on a waitqueue.
L
Linus Torvalds 已提交
3674 3675 3676
 * @q: the waitqueue
 * @mode: which threads
 * @nr_exclusive: how many wake-one or wake-many threads to wake up
3677
 * @key: opaque value to be passed to wakeup targets
L
Linus Torvalds 已提交
3678 3679 3680 3681 3682 3683 3684
 *
 * 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.
3685 3686 3687
 *
 * 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
3688
 */
3689 3690
void __wake_up_sync_key(wait_queue_head_t *q, unsigned int mode,
			int nr_exclusive, void *key)
L
Linus Torvalds 已提交
3691 3692
{
	unsigned long flags;
P
Peter Zijlstra 已提交
3693
	int wake_flags = WF_SYNC;
L
Linus Torvalds 已提交
3694 3695 3696 3697 3698

	if (unlikely(!q))
		return;

	if (unlikely(!nr_exclusive))
P
Peter Zijlstra 已提交
3699
		wake_flags = 0;
L
Linus Torvalds 已提交
3700 3701

	spin_lock_irqsave(&q->lock, flags);
P
Peter Zijlstra 已提交
3702
	__wake_up_common(q, mode, nr_exclusive, wake_flags, key);
L
Linus Torvalds 已提交
3703 3704
	spin_unlock_irqrestore(&q->lock, flags);
}
3705 3706 3707 3708 3709 3710 3711 3712 3713
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
3714 3715
EXPORT_SYMBOL_GPL(__wake_up_sync);	/* For internal use only */

3716 3717 3718 3719 3720 3721 3722 3723
/**
 * 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.
3724 3725 3726
 *
 * 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.
3727
 */
3728
void complete(struct completion *x)
L
Linus Torvalds 已提交
3729 3730 3731 3732 3733
{
	unsigned long flags;

	spin_lock_irqsave(&x->wait.lock, flags);
	x->done++;
3734
	__wake_up_common(&x->wait, TASK_NORMAL, 1, 0, NULL);
L
Linus Torvalds 已提交
3735 3736 3737 3738
	spin_unlock_irqrestore(&x->wait.lock, flags);
}
EXPORT_SYMBOL(complete);

3739 3740 3741 3742 3743
/**
 * 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.
3744 3745 3746
 *
 * 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.
3747
 */
3748
void complete_all(struct completion *x)
L
Linus Torvalds 已提交
3749 3750 3751 3752 3753
{
	unsigned long flags;

	spin_lock_irqsave(&x->wait.lock, flags);
	x->done += UINT_MAX/2;
3754
	__wake_up_common(&x->wait, TASK_NORMAL, 0, 0, NULL);
L
Linus Torvalds 已提交
3755 3756 3757 3758
	spin_unlock_irqrestore(&x->wait.lock, flags);
}
EXPORT_SYMBOL(complete_all);

3759 3760
static inline long __sched
do_wait_for_common(struct completion *x, long timeout, int state)
L
Linus Torvalds 已提交
3761 3762 3763 3764
{
	if (!x->done) {
		DECLARE_WAITQUEUE(wait, current);

C
Changli Gao 已提交
3765
		__add_wait_queue_tail_exclusive(&x->wait, &wait);
L
Linus Torvalds 已提交
3766
		do {
3767
			if (signal_pending_state(state, current)) {
3768 3769
				timeout = -ERESTARTSYS;
				break;
3770 3771
			}
			__set_current_state(state);
L
Linus Torvalds 已提交
3772 3773 3774
			spin_unlock_irq(&x->wait.lock);
			timeout = schedule_timeout(timeout);
			spin_lock_irq(&x->wait.lock);
3775
		} while (!x->done && timeout);
L
Linus Torvalds 已提交
3776
		__remove_wait_queue(&x->wait, &wait);
3777 3778
		if (!x->done)
			return timeout;
L
Linus Torvalds 已提交
3779 3780
	}
	x->done--;
3781
	return timeout ?: 1;
L
Linus Torvalds 已提交
3782 3783
}

3784 3785
static long __sched
wait_for_common(struct completion *x, long timeout, int state)
L
Linus Torvalds 已提交
3786 3787 3788 3789
{
	might_sleep();

	spin_lock_irq(&x->wait.lock);
3790
	timeout = do_wait_for_common(x, timeout, state);
L
Linus Torvalds 已提交
3791
	spin_unlock_irq(&x->wait.lock);
3792 3793
	return timeout;
}
L
Linus Torvalds 已提交
3794

3795 3796 3797 3798 3799 3800 3801 3802 3803 3804
/**
 * 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().
 */
3805
void __sched wait_for_completion(struct completion *x)
3806 3807
{
	wait_for_common(x, MAX_SCHEDULE_TIMEOUT, TASK_UNINTERRUPTIBLE);
L
Linus Torvalds 已提交
3808
}
3809
EXPORT_SYMBOL(wait_for_completion);
L
Linus Torvalds 已提交
3810

3811 3812 3813 3814 3815 3816 3817 3818
/**
 * 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.
3819 3820 3821
 *
 * The return value is 0 if timed out, and positive (at least 1, or number of
 * jiffies left till timeout) if completed.
3822
 */
3823
unsigned long __sched
3824
wait_for_completion_timeout(struct completion *x, unsigned long timeout)
L
Linus Torvalds 已提交
3825
{
3826
	return wait_for_common(x, timeout, TASK_UNINTERRUPTIBLE);
L
Linus Torvalds 已提交
3827
}
3828
EXPORT_SYMBOL(wait_for_completion_timeout);
L
Linus Torvalds 已提交
3829

3830 3831 3832 3833 3834 3835
/**
 * 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.
3836 3837
 *
 * The return value is -ERESTARTSYS if interrupted, 0 if completed.
3838
 */
3839
int __sched wait_for_completion_interruptible(struct completion *x)
I
Ingo Molnar 已提交
3840
{
3841 3842 3843 3844
	long t = wait_for_common(x, MAX_SCHEDULE_TIMEOUT, TASK_INTERRUPTIBLE);
	if (t == -ERESTARTSYS)
		return t;
	return 0;
I
Ingo Molnar 已提交
3845
}
3846
EXPORT_SYMBOL(wait_for_completion_interruptible);
L
Linus Torvalds 已提交
3847

3848 3849 3850 3851 3852 3853 3854
/**
 * 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.
3855 3856 3857
 *
 * The return value is -ERESTARTSYS if interrupted, 0 if timed out,
 * positive (at least 1, or number of jiffies left till timeout) if completed.
3858
 */
3859
long __sched
3860 3861
wait_for_completion_interruptible_timeout(struct completion *x,
					  unsigned long timeout)
I
Ingo Molnar 已提交
3862
{
3863
	return wait_for_common(x, timeout, TASK_INTERRUPTIBLE);
I
Ingo Molnar 已提交
3864
}
3865
EXPORT_SYMBOL(wait_for_completion_interruptible_timeout);
L
Linus Torvalds 已提交
3866

3867 3868 3869 3870 3871 3872
/**
 * 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.
3873 3874
 *
 * The return value is -ERESTARTSYS if interrupted, 0 if completed.
3875
 */
M
Matthew Wilcox 已提交
3876 3877 3878 3879 3880 3881 3882 3883 3884
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);

3885 3886 3887 3888 3889 3890 3891 3892
/**
 * 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.
3893 3894 3895
 *
 * The return value is -ERESTARTSYS if interrupted, 0 if timed out,
 * positive (at least 1, or number of jiffies left till timeout) if completed.
3896
 */
3897
long __sched
3898 3899 3900 3901 3902 3903 3904
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);

3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918
/**
 *	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)
{
3919
	unsigned long flags;
3920 3921
	int ret = 1;

3922
	spin_lock_irqsave(&x->wait.lock, flags);
3923 3924 3925 3926
	if (!x->done)
		ret = 0;
	else
		x->done--;
3927
	spin_unlock_irqrestore(&x->wait.lock, flags);
3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941
	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)
{
3942
	unsigned long flags;
3943 3944
	int ret = 1;

3945
	spin_lock_irqsave(&x->wait.lock, flags);
3946 3947
	if (!x->done)
		ret = 0;
3948
	spin_unlock_irqrestore(&x->wait.lock, flags);
3949 3950 3951 3952
	return ret;
}
EXPORT_SYMBOL(completion_done);

3953 3954
static long __sched
sleep_on_common(wait_queue_head_t *q, int state, long timeout)
L
Linus Torvalds 已提交
3955
{
I
Ingo Molnar 已提交
3956 3957 3958 3959
	unsigned long flags;
	wait_queue_t wait;

	init_waitqueue_entry(&wait, current);
L
Linus Torvalds 已提交
3960

3961
	__set_current_state(state);
L
Linus Torvalds 已提交
3962

3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976
	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
3977 3978 3979
}
EXPORT_SYMBOL(interruptible_sleep_on);

I
Ingo Molnar 已提交
3980
long __sched
I
Ingo Molnar 已提交
3981
interruptible_sleep_on_timeout(wait_queue_head_t *q, long timeout)
L
Linus Torvalds 已提交
3982
{
3983
	return sleep_on_common(q, TASK_INTERRUPTIBLE, timeout);
L
Linus Torvalds 已提交
3984 3985 3986
}
EXPORT_SYMBOL(interruptible_sleep_on_timeout);

I
Ingo Molnar 已提交
3987
void __sched sleep_on(wait_queue_head_t *q)
L
Linus Torvalds 已提交
3988
{
3989
	sleep_on_common(q, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
L
Linus Torvalds 已提交
3990 3991 3992
}
EXPORT_SYMBOL(sleep_on);

I
Ingo Molnar 已提交
3993
long __sched sleep_on_timeout(wait_queue_head_t *q, long timeout)
L
Linus Torvalds 已提交
3994
{
3995
	return sleep_on_common(q, TASK_UNINTERRUPTIBLE, timeout);
L
Linus Torvalds 已提交
3996 3997 3998
}
EXPORT_SYMBOL(sleep_on_timeout);

3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010
#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.
 */
4011
void rt_mutex_setprio(struct task_struct *p, int prio)
4012
{
4013
	int oldprio, on_rq, running;
4014
	struct rq *rq;
4015
	const struct sched_class *prev_class;
4016 4017 4018

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

4019
	rq = __task_rq_lock(p);
4020

4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038
	/*
	 * Idle task boosting is a nono in general. There is one
	 * exception, when PREEMPT_RT and NOHZ is active:
	 *
	 * The idle task calls get_next_timer_interrupt() and holds
	 * the timer wheel base->lock on the CPU and another CPU wants
	 * to access the timer (probably to cancel it). We can safely
	 * ignore the boosting request, as the idle CPU runs this code
	 * with interrupts disabled and will complete the lock
	 * protected section without being interrupted. So there is no
	 * real need to boost.
	 */
	if (unlikely(p == rq->idle)) {
		WARN_ON(p != rq->curr);
		WARN_ON(p->pi_blocked_on);
		goto out_unlock;
	}

4039
	trace_sched_pi_setprio(p, prio);
4040
	oldprio = p->prio;
4041
	prev_class = p->sched_class;
P
Peter Zijlstra 已提交
4042
	on_rq = p->on_rq;
4043
	running = task_current(rq, p);
4044
	if (on_rq)
4045
		dequeue_task(rq, p, 0);
4046 4047
	if (running)
		p->sched_class->put_prev_task(rq, p);
I
Ingo Molnar 已提交
4048 4049 4050 4051 4052 4053

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

4054 4055
	p->prio = prio;

4056 4057
	if (running)
		p->sched_class->set_curr_task(rq);
P
Peter Zijlstra 已提交
4058
	if (on_rq)
4059
		enqueue_task(rq, p, oldprio < prio ? ENQUEUE_HEAD : 0);
4060

P
Peter Zijlstra 已提交
4061
	check_class_changed(rq, p, prev_class, oldprio);
4062
out_unlock:
4063
	__task_rq_unlock(rq);
4064 4065
}
#endif
4066
void set_user_nice(struct task_struct *p, long nice)
L
Linus Torvalds 已提交
4067
{
I
Ingo Molnar 已提交
4068
	int old_prio, delta, on_rq;
L
Linus Torvalds 已提交
4069
	unsigned long flags;
4070
	struct rq *rq;
L
Linus Torvalds 已提交
4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081 4082

	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
4083
	 * SCHED_FIFO/SCHED_RR:
L
Linus Torvalds 已提交
4084
	 */
4085
	if (task_has_rt_policy(p)) {
L
Linus Torvalds 已提交
4086 4087 4088
		p->static_prio = NICE_TO_PRIO(nice);
		goto out_unlock;
	}
P
Peter Zijlstra 已提交
4089
	on_rq = p->on_rq;
4090
	if (on_rq)
4091
		dequeue_task(rq, p, 0);
L
Linus Torvalds 已提交
4092 4093

	p->static_prio = NICE_TO_PRIO(nice);
4094
	set_load_weight(p);
4095 4096 4097
	old_prio = p->prio;
	p->prio = effective_prio(p);
	delta = p->prio - old_prio;
L
Linus Torvalds 已提交
4098

I
Ingo Molnar 已提交
4099
	if (on_rq) {
4100
		enqueue_task(rq, p, 0);
L
Linus Torvalds 已提交
4101
		/*
4102 4103
		 * If the task increased its priority or is running and
		 * lowered its priority, then reschedule its CPU:
L
Linus Torvalds 已提交
4104
		 */
4105
		if (delta < 0 || (delta > 0 && task_running(rq, p)))
L
Linus Torvalds 已提交
4106 4107 4108
			resched_task(rq->curr);
	}
out_unlock:
4109
	task_rq_unlock(rq, p, &flags);
L
Linus Torvalds 已提交
4110 4111 4112
}
EXPORT_SYMBOL(set_user_nice);

M
Matt Mackall 已提交
4113 4114 4115 4116 4117
/*
 * can_nice - check if a task can reduce its nice value
 * @p: task
 * @nice: nice value
 */
4118
int can_nice(const struct task_struct *p, const int nice)
M
Matt Mackall 已提交
4119
{
4120 4121
	/* convert nice value [19,-20] to rlimit style value [1,40] */
	int nice_rlim = 20 - nice;
4122

4123
	return (nice_rlim <= task_rlimit(p, RLIMIT_NICE) ||
M
Matt Mackall 已提交
4124 4125 4126
		capable(CAP_SYS_NICE));
}

L
Linus Torvalds 已提交
4127 4128 4129 4130 4131 4132 4133 4134 4135
#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.
 */
4136
SYSCALL_DEFINE1(nice, int, increment)
L
Linus Torvalds 已提交
4137
{
4138
	long nice, retval;
L
Linus Torvalds 已提交
4139 4140 4141 4142 4143 4144

	/*
	 * 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
4145 4146
	if (increment < -40)
		increment = -40;
L
Linus Torvalds 已提交
4147 4148 4149
	if (increment > 40)
		increment = 40;

4150
	nice = TASK_NICE(current) + increment;
L
Linus Torvalds 已提交
4151 4152 4153 4154 4155
	if (nice < -20)
		nice = -20;
	if (nice > 19)
		nice = 19;

M
Matt Mackall 已提交
4156 4157 4158
	if (increment < 0 && !can_nice(current, nice))
		return -EPERM;

L
Linus Torvalds 已提交
4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174 4175 4176
	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.
 */
4177
int task_prio(const struct task_struct *p)
L
Linus Torvalds 已提交
4178 4179 4180 4181 4182 4183 4184 4185
{
	return p->prio - MAX_RT_PRIO;
}

/**
 * task_nice - return the nice value of a given task.
 * @p: the task in question.
 */
4186
int task_nice(const struct task_struct *p)
L
Linus Torvalds 已提交
4187 4188 4189
{
	return TASK_NICE(p);
}
P
Pavel Roskin 已提交
4190
EXPORT_SYMBOL(task_nice);
L
Linus Torvalds 已提交
4191 4192 4193 4194 4195 4196 4197

/**
 * idle_cpu - is a given cpu idle currently?
 * @cpu: the processor in question.
 */
int idle_cpu(int cpu)
{
T
Thomas Gleixner 已提交
4198 4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210 4211
	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
4212 4213 4214 4215 4216 4217
}

/**
 * idle_task - return the idle task for a given cpu.
 * @cpu: the processor in question.
 */
4218
struct task_struct *idle_task(int cpu)
L
Linus Torvalds 已提交
4219 4220 4221 4222 4223 4224 4225 4226
{
	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
4227
static struct task_struct *find_process_by_pid(pid_t pid)
L
Linus Torvalds 已提交
4228
{
4229
	return pid ? find_task_by_vpid(pid) : current;
L
Linus Torvalds 已提交
4230 4231 4232
}

/* Actually do priority change: must hold rq lock. */
I
Ingo Molnar 已提交
4233 4234
static void
__setscheduler(struct rq *rq, struct task_struct *p, int policy, int prio)
L
Linus Torvalds 已提交
4235 4236 4237
{
	p->policy = policy;
	p->rt_priority = prio;
4238 4239 4240
	p->normal_prio = normal_prio(p);
	/* we are holding p->pi_lock already */
	p->prio = rt_mutex_getprio(p);
4241 4242 4243 4244
	if (rt_prio(p->prio))
		p->sched_class = &rt_sched_class;
	else
		p->sched_class = &fair_sched_class;
4245
	set_load_weight(p);
L
Linus Torvalds 已提交
4246 4247
}

4248 4249 4250 4251 4252 4253 4254 4255 4256 4257
/*
 * 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);
4258 4259
	match = (uid_eq(cred->euid, pcred->euid) ||
		 uid_eq(cred->euid, pcred->uid));
4260 4261 4262 4263
	rcu_read_unlock();
	return match;
}

4264
static int __sched_setscheduler(struct task_struct *p, int policy,
4265
				const struct sched_param *param, bool user)
L
Linus Torvalds 已提交
4266
{
4267
	int retval, oldprio, oldpolicy = -1, on_rq, running;
L
Linus Torvalds 已提交
4268
	unsigned long flags;
4269
	const struct sched_class *prev_class;
4270
	struct rq *rq;
4271
	int reset_on_fork;
L
Linus Torvalds 已提交
4272

4273 4274
	/* may grab non-irq protected spin_locks */
	BUG_ON(in_interrupt());
L
Linus Torvalds 已提交
4275 4276
recheck:
	/* double check policy once rq lock held */
4277 4278
	if (policy < 0) {
		reset_on_fork = p->sched_reset_on_fork;
L
Linus Torvalds 已提交
4279
		policy = oldpolicy = p->policy;
4280 4281 4282 4283 4284 4285 4286 4287 4288 4289
	} 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
4290 4291
	/*
	 * Valid priorities for SCHED_FIFO and SCHED_RR are
I
Ingo Molnar 已提交
4292 4293
	 * 1..MAX_USER_RT_PRIO-1, valid priority for SCHED_NORMAL,
	 * SCHED_BATCH and SCHED_IDLE is 0.
L
Linus Torvalds 已提交
4294 4295
	 */
	if (param->sched_priority < 0 ||
I
Ingo Molnar 已提交
4296
	    (p->mm && param->sched_priority > MAX_USER_RT_PRIO-1) ||
4297
	    (!p->mm && param->sched_priority > MAX_RT_PRIO-1))
L
Linus Torvalds 已提交
4298
		return -EINVAL;
4299
	if (rt_policy(policy) != (param->sched_priority != 0))
L
Linus Torvalds 已提交
4300 4301
		return -EINVAL;

4302 4303 4304
	/*
	 * Allow unprivileged RT tasks to decrease priority:
	 */
4305
	if (user && !capable(CAP_SYS_NICE)) {
4306
		if (rt_policy(policy)) {
4307 4308
			unsigned long rlim_rtprio =
					task_rlimit(p, RLIMIT_RTPRIO);
4309 4310 4311 4312 4313 4314 4315 4316 4317 4318

			/* 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;
		}
4319

I
Ingo Molnar 已提交
4320
		/*
4321 4322
		 * 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
4323
		 */
4324 4325 4326 4327
		if (p->policy == SCHED_IDLE && policy != SCHED_IDLE) {
			if (!can_nice(p, TASK_NICE(p)))
				return -EPERM;
		}
4328

4329
		/* can't change other user's priorities */
4330
		if (!check_same_owner(p))
4331
			return -EPERM;
4332 4333 4334 4335

		/* Normal users shall not reset the sched_reset_on_fork flag */
		if (p->sched_reset_on_fork && !reset_on_fork)
			return -EPERM;
4336
	}
L
Linus Torvalds 已提交
4337

4338
	if (user) {
4339
		retval = security_task_setscheduler(p);
4340 4341 4342 4343
		if (retval)
			return retval;
	}

4344 4345 4346
	/*
	 * make sure no PI-waiters arrive (or leave) while we are
	 * changing the priority of the task:
4347
	 *
L
Lucas De Marchi 已提交
4348
	 * To be able to change p->policy safely, the appropriate
L
Linus Torvalds 已提交
4349 4350
	 * runqueue lock must be held.
	 */
4351
	rq = task_rq_lock(p, &flags);
4352

4353 4354 4355 4356
	/*
	 * Changing the policy of the stop threads its a very bad idea
	 */
	if (p == rq->stop) {
4357
		task_rq_unlock(rq, p, &flags);
4358 4359 4360
		return -EINVAL;
	}

4361 4362 4363 4364 4365
	/*
	 * 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))) {
4366
		task_rq_unlock(rq, p, &flags);
4367 4368 4369
		return 0;
	}

4370 4371 4372 4373 4374 4375 4376
#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) &&
4377 4378
				task_group(p)->rt_bandwidth.rt_runtime == 0 &&
				!task_group_is_autogroup(task_group(p))) {
4379
			task_rq_unlock(rq, p, &flags);
4380 4381 4382 4383 4384
			return -EPERM;
		}
	}
#endif

L
Linus Torvalds 已提交
4385 4386 4387
	/* recheck policy now with rq lock held */
	if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) {
		policy = oldpolicy = -1;
4388
		task_rq_unlock(rq, p, &flags);
L
Linus Torvalds 已提交
4389 4390
		goto recheck;
	}
P
Peter Zijlstra 已提交
4391
	on_rq = p->on_rq;
4392
	running = task_current(rq, p);
4393
	if (on_rq)
4394
		dequeue_task(rq, p, 0);
4395 4396
	if (running)
		p->sched_class->put_prev_task(rq, p);
4397

4398 4399
	p->sched_reset_on_fork = reset_on_fork;

L
Linus Torvalds 已提交
4400
	oldprio = p->prio;
4401
	prev_class = p->sched_class;
I
Ingo Molnar 已提交
4402
	__setscheduler(rq, p, policy, param->sched_priority);
4403

4404 4405
	if (running)
		p->sched_class->set_curr_task(rq);
P
Peter Zijlstra 已提交
4406
	if (on_rq)
4407
		enqueue_task(rq, p, 0);
4408

P
Peter Zijlstra 已提交
4409
	check_class_changed(rq, p, prev_class, oldprio);
4410
	task_rq_unlock(rq, p, &flags);
4411

4412 4413
	rt_mutex_adjust_pi(p);

L
Linus Torvalds 已提交
4414 4415
	return 0;
}
4416 4417 4418 4419 4420 4421 4422 4423 4424 4425

/**
 * 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,
4426
		       const struct sched_param *param)
4427 4428 4429
{
	return __sched_setscheduler(p, policy, param, true);
}
L
Linus Torvalds 已提交
4430 4431
EXPORT_SYMBOL_GPL(sched_setscheduler);

4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443
/**
 * 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,
4444
			       const struct sched_param *param)
4445 4446 4447 4448
{
	return __sched_setscheduler(p, policy, param, false);
}

I
Ingo Molnar 已提交
4449 4450
static int
do_sched_setscheduler(pid_t pid, int policy, struct sched_param __user *param)
L
Linus Torvalds 已提交
4451 4452 4453
{
	struct sched_param lparam;
	struct task_struct *p;
4454
	int retval;
L
Linus Torvalds 已提交
4455 4456 4457 4458 4459

	if (!param || pid < 0)
		return -EINVAL;
	if (copy_from_user(&lparam, param, sizeof(struct sched_param)))
		return -EFAULT;
4460 4461 4462

	rcu_read_lock();
	retval = -ESRCH;
L
Linus Torvalds 已提交
4463
	p = find_process_by_pid(pid);
4464 4465 4466
	if (p != NULL)
		retval = sched_setscheduler(p, policy, &lparam);
	rcu_read_unlock();
4467

L
Linus Torvalds 已提交
4468 4469 4470 4471 4472 4473 4474 4475 4476
	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.
 */
4477 4478
SYSCALL_DEFINE3(sched_setscheduler, pid_t, pid, int, policy,
		struct sched_param __user *, param)
L
Linus Torvalds 已提交
4479
{
4480 4481 4482 4483
	/* negative values for policy are not valid */
	if (policy < 0)
		return -EINVAL;

L
Linus Torvalds 已提交
4484 4485 4486 4487 4488 4489 4490 4491
	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.
 */
4492
SYSCALL_DEFINE2(sched_setparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4493 4494 4495 4496 4497 4498 4499 4500
{
	return do_sched_setscheduler(pid, -1, param);
}

/**
 * sys_sched_getscheduler - get the policy (scheduling class) of a thread
 * @pid: the pid in question.
 */
4501
SYSCALL_DEFINE1(sched_getscheduler, pid_t, pid)
L
Linus Torvalds 已提交
4502
{
4503
	struct task_struct *p;
4504
	int retval;
L
Linus Torvalds 已提交
4505 4506

	if (pid < 0)
4507
		return -EINVAL;
L
Linus Torvalds 已提交
4508 4509

	retval = -ESRCH;
4510
	rcu_read_lock();
L
Linus Torvalds 已提交
4511 4512 4513 4514
	p = find_process_by_pid(pid);
	if (p) {
		retval = security_task_getscheduler(p);
		if (!retval)
4515 4516
			retval = p->policy
				| (p->sched_reset_on_fork ? SCHED_RESET_ON_FORK : 0);
L
Linus Torvalds 已提交
4517
	}
4518
	rcu_read_unlock();
L
Linus Torvalds 已提交
4519 4520 4521 4522
	return retval;
}

/**
4523
 * sys_sched_getparam - get the RT priority of a thread
L
Linus Torvalds 已提交
4524 4525 4526
 * @pid: the pid in question.
 * @param: structure containing the RT priority.
 */
4527
SYSCALL_DEFINE2(sched_getparam, pid_t, pid, struct sched_param __user *, param)
L
Linus Torvalds 已提交
4528 4529
{
	struct sched_param lp;
4530
	struct task_struct *p;
4531
	int retval;
L
Linus Torvalds 已提交
4532 4533

	if (!param || pid < 0)
4534
		return -EINVAL;
L
Linus Torvalds 已提交
4535

4536
	rcu_read_lock();
L
Linus Torvalds 已提交
4537 4538 4539 4540 4541 4542 4543 4544 4545 4546
	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;
4547
	rcu_read_unlock();
L
Linus Torvalds 已提交
4548 4549 4550 4551 4552 4553 4554 4555 4556

	/*
	 * 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:
4557
	rcu_read_unlock();
L
Linus Torvalds 已提交
4558 4559 4560
	return retval;
}

4561
long sched_setaffinity(pid_t pid, const struct cpumask *in_mask)
L
Linus Torvalds 已提交
4562
{
4563
	cpumask_var_t cpus_allowed, new_mask;
4564 4565
	struct task_struct *p;
	int retval;
L
Linus Torvalds 已提交
4566

4567
	get_online_cpus();
4568
	rcu_read_lock();
L
Linus Torvalds 已提交
4569 4570 4571

	p = find_process_by_pid(pid);
	if (!p) {
4572
		rcu_read_unlock();
4573
		put_online_cpus();
L
Linus Torvalds 已提交
4574 4575 4576
		return -ESRCH;
	}

4577
	/* Prevent p going away */
L
Linus Torvalds 已提交
4578
	get_task_struct(p);
4579
	rcu_read_unlock();
L
Linus Torvalds 已提交
4580

4581 4582 4583 4584 4585 4586 4587 4588
	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
4589
	retval = -EPERM;
4590
	if (!check_same_owner(p) && !ns_capable(task_user_ns(p), CAP_SYS_NICE))
L
Linus Torvalds 已提交
4591 4592
		goto out_unlock;

4593
	retval = security_task_setscheduler(p);
4594 4595 4596
	if (retval)
		goto out_unlock;

4597 4598
	cpuset_cpus_allowed(p, cpus_allowed);
	cpumask_and(new_mask, in_mask, cpus_allowed);
P
Peter Zijlstra 已提交
4599
again:
4600
	retval = set_cpus_allowed_ptr(p, new_mask);
L
Linus Torvalds 已提交
4601

P
Paul Menage 已提交
4602
	if (!retval) {
4603 4604
		cpuset_cpus_allowed(p, cpus_allowed);
		if (!cpumask_subset(new_mask, cpus_allowed)) {
P
Paul Menage 已提交
4605 4606 4607 4608 4609
			/*
			 * We must have raced with a concurrent cpuset
			 * update. Just reset the cpus_allowed to the
			 * cpuset's cpus_allowed
			 */
4610
			cpumask_copy(new_mask, cpus_allowed);
P
Paul Menage 已提交
4611 4612 4613
			goto again;
		}
	}
L
Linus Torvalds 已提交
4614
out_unlock:
4615 4616 4617 4618
	free_cpumask_var(new_mask);
out_free_cpus_allowed:
	free_cpumask_var(cpus_allowed);
out_put_task:
L
Linus Torvalds 已提交
4619
	put_task_struct(p);
4620
	put_online_cpus();
L
Linus Torvalds 已提交
4621 4622 4623 4624
	return retval;
}

static int get_user_cpu_mask(unsigned long __user *user_mask_ptr, unsigned len,
4625
			     struct cpumask *new_mask)
L
Linus Torvalds 已提交
4626
{
4627 4628 4629 4630 4631
	if (len < cpumask_size())
		cpumask_clear(new_mask);
	else if (len > cpumask_size())
		len = cpumask_size();

L
Linus Torvalds 已提交
4632 4633 4634 4635 4636 4637 4638 4639 4640
	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
 */
4641 4642
SYSCALL_DEFINE3(sched_setaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
4643
{
4644
	cpumask_var_t new_mask;
L
Linus Torvalds 已提交
4645 4646
	int retval;

4647 4648
	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
4649

4650 4651 4652 4653 4654
	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
4655 4656
}

4657
long sched_getaffinity(pid_t pid, struct cpumask *mask)
L
Linus Torvalds 已提交
4658
{
4659
	struct task_struct *p;
4660
	unsigned long flags;
L
Linus Torvalds 已提交
4661 4662
	int retval;

4663
	get_online_cpus();
4664
	rcu_read_lock();
L
Linus Torvalds 已提交
4665 4666 4667 4668 4669 4670

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

4671 4672 4673 4674
	retval = security_task_getscheduler(p);
	if (retval)
		goto out_unlock;

4675
	raw_spin_lock_irqsave(&p->pi_lock, flags);
4676
	cpumask_and(mask, &p->cpus_allowed, cpu_online_mask);
4677
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
L
Linus Torvalds 已提交
4678 4679

out_unlock:
4680
	rcu_read_unlock();
4681
	put_online_cpus();
L
Linus Torvalds 已提交
4682

4683
	return retval;
L
Linus Torvalds 已提交
4684 4685 4686 4687 4688 4689 4690 4691
}

/**
 * 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
 */
4692 4693
SYSCALL_DEFINE3(sched_getaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
L
Linus Torvalds 已提交
4694 4695
{
	int ret;
4696
	cpumask_var_t mask;
L
Linus Torvalds 已提交
4697

A
Anton Blanchard 已提交
4698
	if ((len * BITS_PER_BYTE) < nr_cpu_ids)
4699 4700
		return -EINVAL;
	if (len & (sizeof(unsigned long)-1))
L
Linus Torvalds 已提交
4701 4702
		return -EINVAL;

4703 4704
	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
		return -ENOMEM;
L
Linus Torvalds 已提交
4705

4706 4707
	ret = sched_getaffinity(pid, mask);
	if (ret == 0) {
4708
		size_t retlen = min_t(size_t, len, cpumask_size());
4709 4710

		if (copy_to_user(user_mask_ptr, mask, retlen))
4711 4712
			ret = -EFAULT;
		else
4713
			ret = retlen;
4714 4715
	}
	free_cpumask_var(mask);
L
Linus Torvalds 已提交
4716

4717
	return ret;
L
Linus Torvalds 已提交
4718 4719 4720 4721 4722
}

/**
 * sys_sched_yield - yield the current processor to other threads.
 *
I
Ingo Molnar 已提交
4723 4724
 * 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
4725
 */
4726
SYSCALL_DEFINE0(sched_yield)
L
Linus Torvalds 已提交
4727
{
4728
	struct rq *rq = this_rq_lock();
L
Linus Torvalds 已提交
4729

4730
	schedstat_inc(rq, yld_count);
4731
	current->sched_class->yield_task(rq);
L
Linus Torvalds 已提交
4732 4733 4734 4735 4736 4737

	/*
	 * Since we are going to call schedule() anyway, there's
	 * no need to preempt or enable interrupts:
	 */
	__release(rq->lock);
4738
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
4739
	do_raw_spin_unlock(&rq->lock);
4740
	sched_preempt_enable_no_resched();
L
Linus Torvalds 已提交
4741 4742 4743 4744 4745 4746

	schedule();

	return 0;
}

P
Peter Zijlstra 已提交
4747 4748 4749 4750 4751
static inline int should_resched(void)
{
	return need_resched() && !(preempt_count() & PREEMPT_ACTIVE);
}

A
Andrew Morton 已提交
4752
static void __cond_resched(void)
L
Linus Torvalds 已提交
4753
{
4754
	add_preempt_count(PREEMPT_ACTIVE);
4755
	__schedule();
4756
	sub_preempt_count(PREEMPT_ACTIVE);
L
Linus Torvalds 已提交
4757 4758
}

4759
int __sched _cond_resched(void)
L
Linus Torvalds 已提交
4760
{
P
Peter Zijlstra 已提交
4761
	if (should_resched()) {
L
Linus Torvalds 已提交
4762 4763 4764 4765 4766
		__cond_resched();
		return 1;
	}
	return 0;
}
4767
EXPORT_SYMBOL(_cond_resched);
L
Linus Torvalds 已提交
4768 4769

/*
4770
 * __cond_resched_lock() - if a reschedule is pending, drop the given lock,
L
Linus Torvalds 已提交
4771 4772
 * call schedule, and on return reacquire the lock.
 *
I
Ingo Molnar 已提交
4773
 * This works OK both with and without CONFIG_PREEMPT. We do strange low-level
L
Linus Torvalds 已提交
4774 4775 4776
 * operations here to prevent schedule() from being called twice (once via
 * spin_unlock(), once by hand).
 */
4777
int __cond_resched_lock(spinlock_t *lock)
L
Linus Torvalds 已提交
4778
{
P
Peter Zijlstra 已提交
4779
	int resched = should_resched();
J
Jan Kara 已提交
4780 4781
	int ret = 0;

4782 4783
	lockdep_assert_held(lock);

N
Nick Piggin 已提交
4784
	if (spin_needbreak(lock) || resched) {
L
Linus Torvalds 已提交
4785
		spin_unlock(lock);
P
Peter Zijlstra 已提交
4786
		if (resched)
N
Nick Piggin 已提交
4787 4788 4789
			__cond_resched();
		else
			cpu_relax();
J
Jan Kara 已提交
4790
		ret = 1;
L
Linus Torvalds 已提交
4791 4792
		spin_lock(lock);
	}
J
Jan Kara 已提交
4793
	return ret;
L
Linus Torvalds 已提交
4794
}
4795
EXPORT_SYMBOL(__cond_resched_lock);
L
Linus Torvalds 已提交
4796

4797
int __sched __cond_resched_softirq(void)
L
Linus Torvalds 已提交
4798 4799 4800
{
	BUG_ON(!in_softirq());

P
Peter Zijlstra 已提交
4801
	if (should_resched()) {
4802
		local_bh_enable();
L
Linus Torvalds 已提交
4803 4804 4805 4806 4807 4808
		__cond_resched();
		local_bh_disable();
		return 1;
	}
	return 0;
}
4809
EXPORT_SYMBOL(__cond_resched_softirq);
L
Linus Torvalds 已提交
4810 4811 4812 4813

/**
 * yield - yield the current processor to other threads.
 *
P
Peter Zijlstra 已提交
4814 4815 4816 4817 4818 4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830 4831
 * Do not ever use this function, there's a 99% chance you're doing it wrong.
 *
 * The scheduler is at all times free to pick the calling task as the most
 * eligible task to run, if removing the yield() call from your code breaks
 * it, its already broken.
 *
 * Typical broken usage is:
 *
 * while (!event)
 * 	yield();
 *
 * where one assumes that yield() will let 'the other' process run that will
 * make event true. If the current task is a SCHED_FIFO task that will never
 * happen. Never use yield() as a progress guarantee!!
 *
 * If you want to use yield() to wait for something, use wait_event().
 * If you want to use yield() to be 'nice' for others, use cond_resched().
 * If you still want to use yield(), do not!
L
Linus Torvalds 已提交
4832 4833 4834 4835 4836 4837 4838 4839
 */
void __sched yield(void)
{
	set_current_state(TASK_RUNNING);
	sys_sched_yield();
}
EXPORT_SYMBOL(yield);

4840 4841 4842 4843
/**
 * 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
4844 4845
 * @p: target task
 * @preempt: whether task preemption is allowed or not
4846 4847 4848 4849 4850 4851 4852 4853 4854 4855 4856 4857 4858 4859 4860 4861 4862 4863 4864 4865 4866 4867 4868 4869 4870 4871 4872 4873 4874 4875 4876 4877 4878 4879
 *
 * 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);
4880
	if (yielded) {
4881
		schedstat_inc(rq, yld_count);
4882 4883 4884 4885 4886 4887
		/*
		 * Make p's CPU reschedule; pick_next_entity takes care of
		 * fairness.
		 */
		if (preempt && rq != p_rq)
			resched_task(p_rq->curr);
4888 4889 4890 4891 4892 4893 4894
	} 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;
4895
	}
4896 4897 4898 4899 4900 4901 4902 4903 4904 4905 4906 4907

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
4908
/*
I
Ingo Molnar 已提交
4909
 * This task is about to go to sleep on IO. Increment rq->nr_iowait so
L
Linus Torvalds 已提交
4910 4911 4912 4913
 * that process accounting knows that this is a task in IO wait state.
 */
void __sched io_schedule(void)
{
4914
	struct rq *rq = raw_rq();
L
Linus Torvalds 已提交
4915

4916
	delayacct_blkio_start();
L
Linus Torvalds 已提交
4917
	atomic_inc(&rq->nr_iowait);
4918
	blk_flush_plug(current);
4919
	current->in_iowait = 1;
L
Linus Torvalds 已提交
4920
	schedule();
4921
	current->in_iowait = 0;
L
Linus Torvalds 已提交
4922
	atomic_dec(&rq->nr_iowait);
4923
	delayacct_blkio_end();
L
Linus Torvalds 已提交
4924 4925 4926 4927 4928
}
EXPORT_SYMBOL(io_schedule);

long __sched io_schedule_timeout(long timeout)
{
4929
	struct rq *rq = raw_rq();
L
Linus Torvalds 已提交
4930 4931
	long ret;

4932
	delayacct_blkio_start();
L
Linus Torvalds 已提交
4933
	atomic_inc(&rq->nr_iowait);
4934
	blk_flush_plug(current);
4935
	current->in_iowait = 1;
L
Linus Torvalds 已提交
4936
	ret = schedule_timeout(timeout);
4937
	current->in_iowait = 0;
L
Linus Torvalds 已提交
4938
	atomic_dec(&rq->nr_iowait);
4939
	delayacct_blkio_end();
L
Linus Torvalds 已提交
4940 4941 4942 4943 4944 4945 4946 4947 4948 4949
	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.
 */
4950
SYSCALL_DEFINE1(sched_get_priority_max, int, policy)
L
Linus Torvalds 已提交
4951 4952 4953 4954 4955 4956 4957 4958 4959
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = MAX_USER_RT_PRIO-1;
		break;
	case SCHED_NORMAL:
4960
	case SCHED_BATCH:
I
Ingo Molnar 已提交
4961
	case SCHED_IDLE:
L
Linus Torvalds 已提交
4962 4963 4964 4965 4966 4967 4968 4969 4970 4971 4972 4973 4974
		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.
 */
4975
SYSCALL_DEFINE1(sched_get_priority_min, int, policy)
L
Linus Torvalds 已提交
4976 4977 4978 4979 4980 4981 4982 4983 4984
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = 1;
		break;
	case SCHED_NORMAL:
4985
	case SCHED_BATCH:
I
Ingo Molnar 已提交
4986
	case SCHED_IDLE:
L
Linus Torvalds 已提交
4987 4988 4989 4990 4991 4992 4993 4994 4995 4996 4997 4998 4999
		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.
 */
5000
SYSCALL_DEFINE2(sched_rr_get_interval, pid_t, pid,
5001
		struct timespec __user *, interval)
L
Linus Torvalds 已提交
5002
{
5003
	struct task_struct *p;
D
Dmitry Adamushko 已提交
5004
	unsigned int time_slice;
5005 5006
	unsigned long flags;
	struct rq *rq;
5007
	int retval;
L
Linus Torvalds 已提交
5008 5009 5010
	struct timespec t;

	if (pid < 0)
5011
		return -EINVAL;
L
Linus Torvalds 已提交
5012 5013

	retval = -ESRCH;
5014
	rcu_read_lock();
L
Linus Torvalds 已提交
5015 5016 5017 5018 5019 5020 5021 5022
	p = find_process_by_pid(pid);
	if (!p)
		goto out_unlock;

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

5023 5024
	rq = task_rq_lock(p, &flags);
	time_slice = p->sched_class->get_rr_interval(rq, p);
5025
	task_rq_unlock(rq, p, &flags);
D
Dmitry Adamushko 已提交
5026

5027
	rcu_read_unlock();
D
Dmitry Adamushko 已提交
5028
	jiffies_to_timespec(time_slice, &t);
L
Linus Torvalds 已提交
5029 5030
	retval = copy_to_user(interval, &t, sizeof(t)) ? -EFAULT : 0;
	return retval;
5031

L
Linus Torvalds 已提交
5032
out_unlock:
5033
	rcu_read_unlock();
L
Linus Torvalds 已提交
5034 5035 5036
	return retval;
}

5037
static const char stat_nam[] = TASK_STATE_TO_CHAR_STR;
5038

5039
void sched_show_task(struct task_struct *p)
L
Linus Torvalds 已提交
5040 5041
{
	unsigned long free = 0;
5042
	unsigned state;
L
Linus Torvalds 已提交
5043 5044

	state = p->state ? __ffs(p->state) + 1 : 0;
5045
	printk(KERN_INFO "%-15.15s %c", p->comm,
5046
		state < sizeof(stat_nam) - 1 ? stat_nam[state] : '?');
5047
#if BITS_PER_LONG == 32
L
Linus Torvalds 已提交
5048
	if (state == TASK_RUNNING)
P
Peter Zijlstra 已提交
5049
		printk(KERN_CONT " running  ");
L
Linus Torvalds 已提交
5050
	else
P
Peter Zijlstra 已提交
5051
		printk(KERN_CONT " %08lx ", thread_saved_pc(p));
L
Linus Torvalds 已提交
5052 5053
#else
	if (state == TASK_RUNNING)
P
Peter Zijlstra 已提交
5054
		printk(KERN_CONT "  running task    ");
L
Linus Torvalds 已提交
5055
	else
P
Peter Zijlstra 已提交
5056
		printk(KERN_CONT " %016lx ", thread_saved_pc(p));
L
Linus Torvalds 已提交
5057 5058
#endif
#ifdef CONFIG_DEBUG_STACK_USAGE
5059
	free = stack_not_used(p);
L
Linus Torvalds 已提交
5060
#endif
P
Peter Zijlstra 已提交
5061
	printk(KERN_CONT "%5lu %5d %6d 0x%08lx\n", free,
5062
		task_pid_nr(p), task_pid_nr(rcu_dereference(p->real_parent)),
5063
		(unsigned long)task_thread_info(p)->flags);
L
Linus Torvalds 已提交
5064

5065
	show_stack(p, NULL);
L
Linus Torvalds 已提交
5066 5067
}

I
Ingo Molnar 已提交
5068
void show_state_filter(unsigned long state_filter)
L
Linus Torvalds 已提交
5069
{
5070
	struct task_struct *g, *p;
L
Linus Torvalds 已提交
5071

5072
#if BITS_PER_LONG == 32
P
Peter Zijlstra 已提交
5073 5074
	printk(KERN_INFO
		"  task                PC stack   pid father\n");
L
Linus Torvalds 已提交
5075
#else
P
Peter Zijlstra 已提交
5076 5077
	printk(KERN_INFO
		"  task                        PC stack   pid father\n");
L
Linus Torvalds 已提交
5078
#endif
5079
	rcu_read_lock();
L
Linus Torvalds 已提交
5080 5081 5082
	do_each_thread(g, p) {
		/*
		 * reset the NMI-timeout, listing all files on a slow
L
Lucas De Marchi 已提交
5083
		 * console might take a lot of time:
L
Linus Torvalds 已提交
5084 5085
		 */
		touch_nmi_watchdog();
I
Ingo Molnar 已提交
5086
		if (!state_filter || (p->state & state_filter))
5087
			sched_show_task(p);
L
Linus Torvalds 已提交
5088 5089
	} while_each_thread(g, p);

5090 5091
	touch_all_softlockup_watchdogs();

I
Ingo Molnar 已提交
5092 5093 5094
#ifdef CONFIG_SCHED_DEBUG
	sysrq_sched_debug_show();
#endif
5095
	rcu_read_unlock();
I
Ingo Molnar 已提交
5096 5097 5098
	/*
	 * Only show locks if all tasks are dumped:
	 */
5099
	if (!state_filter)
I
Ingo Molnar 已提交
5100
		debug_show_all_locks();
L
Linus Torvalds 已提交
5101 5102
}

I
Ingo Molnar 已提交
5103 5104
void __cpuinit init_idle_bootup_task(struct task_struct *idle)
{
I
Ingo Molnar 已提交
5105
	idle->sched_class = &idle_sched_class;
I
Ingo Molnar 已提交
5106 5107
}

5108 5109 5110 5111 5112 5113 5114 5115
/**
 * 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.
 */
5116
void __cpuinit init_idle(struct task_struct *idle, int cpu)
L
Linus Torvalds 已提交
5117
{
5118
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
5119 5120
	unsigned long flags;

5121
	raw_spin_lock_irqsave(&rq->lock, flags);
5122

I
Ingo Molnar 已提交
5123
	__sched_fork(idle);
5124
	idle->state = TASK_RUNNING;
I
Ingo Molnar 已提交
5125 5126
	idle->se.exec_start = sched_clock();

5127
	do_set_cpus_allowed(idle, cpumask_of(cpu));
5128 5129 5130 5131 5132 5133 5134 5135 5136 5137 5138
	/*
	 * 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
5139
	__set_task_cpu(idle, cpu);
5140
	rcu_read_unlock();
L
Linus Torvalds 已提交
5141 5142

	rq->curr = rq->idle = idle;
P
Peter Zijlstra 已提交
5143 5144
#if defined(CONFIG_SMP)
	idle->on_cpu = 1;
5145
#endif
5146
	raw_spin_unlock_irqrestore(&rq->lock, flags);
L
Linus Torvalds 已提交
5147 5148

	/* Set the preempt count _outside_ the spinlocks! */
A
Al Viro 已提交
5149
	task_thread_info(idle)->preempt_count = 0;
5150

I
Ingo Molnar 已提交
5151 5152 5153 5154
	/*
	 * The idle tasks have their own, simple scheduling class:
	 */
	idle->sched_class = &idle_sched_class;
5155
	ftrace_graph_init_idle_task(idle, cpu);
5156 5157 5158
#if defined(CONFIG_SMP)
	sprintf(idle->comm, "%s/%d", INIT_TASK_COMM, cpu);
#endif
I
Ingo Molnar 已提交
5159 5160
}

L
Linus Torvalds 已提交
5161
#ifdef CONFIG_SMP
5162 5163 5164 5165
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);
5166 5167

	cpumask_copy(&p->cpus_allowed, new_mask);
5168
	p->nr_cpus_allowed = cpumask_weight(new_mask);
5169 5170
}

L
Linus Torvalds 已提交
5171 5172 5173
/*
 * This is how migration works:
 *
5174 5175 5176 5177 5178 5179
 * 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
5180
 *    it and puts it into the right queue.
5181 5182
 * 5) stopper completes and stop_one_cpu() returns and the migration
 *    is done.
L
Linus Torvalds 已提交
5183 5184 5185 5186 5187 5188 5189 5190
 */

/*
 * 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
5191
 * task must not exit() & deallocate itself prematurely. The
L
Linus Torvalds 已提交
5192 5193
 * call is not atomic; no spinlocks may be held.
 */
5194
int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
L
Linus Torvalds 已提交
5195 5196
{
	unsigned long flags;
5197
	struct rq *rq;
5198
	unsigned int dest_cpu;
5199
	int ret = 0;
L
Linus Torvalds 已提交
5200 5201

	rq = task_rq_lock(p, &flags);
5202

5203 5204 5205
	if (cpumask_equal(&p->cpus_allowed, new_mask))
		goto out;

5206
	if (!cpumask_intersects(new_mask, cpu_active_mask)) {
L
Linus Torvalds 已提交
5207 5208 5209 5210
		ret = -EINVAL;
		goto out;
	}

5211
	if (unlikely((p->flags & PF_THREAD_BOUND) && p != current)) {
5212 5213 5214 5215
		ret = -EINVAL;
		goto out;
	}

5216
	do_set_cpus_allowed(p, new_mask);
5217

L
Linus Torvalds 已提交
5218
	/* Can the task run on the task's current CPU? If so, we're done */
5219
	if (cpumask_test_cpu(task_cpu(p), new_mask))
L
Linus Torvalds 已提交
5220 5221
		goto out;

5222
	dest_cpu = cpumask_any_and(cpu_active_mask, new_mask);
5223
	if (p->on_rq) {
5224
		struct migration_arg arg = { p, dest_cpu };
L
Linus Torvalds 已提交
5225
		/* Need help from migration thread: drop lock and wait. */
5226
		task_rq_unlock(rq, p, &flags);
5227
		stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg);
L
Linus Torvalds 已提交
5228 5229 5230 5231
		tlb_migrate_finish(p->mm);
		return 0;
	}
out:
5232
	task_rq_unlock(rq, p, &flags);
5233

L
Linus Torvalds 已提交
5234 5235
	return ret;
}
5236
EXPORT_SYMBOL_GPL(set_cpus_allowed_ptr);
L
Linus Torvalds 已提交
5237 5238

/*
I
Ingo Molnar 已提交
5239
 * Move (not current) task off this cpu, onto dest cpu. We're doing
L
Linus Torvalds 已提交
5240 5241 5242 5243 5244 5245
 * 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.
5246 5247
 *
 * Returns non-zero if task was successfully migrated.
L
Linus Torvalds 已提交
5248
 */
5249
static int __migrate_task(struct task_struct *p, int src_cpu, int dest_cpu)
L
Linus Torvalds 已提交
5250
{
5251
	struct rq *rq_dest, *rq_src;
5252
	int ret = 0;
L
Linus Torvalds 已提交
5253

5254
	if (unlikely(!cpu_active(dest_cpu)))
5255
		return ret;
L
Linus Torvalds 已提交
5256 5257 5258 5259

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

5260
	raw_spin_lock(&p->pi_lock);
L
Linus Torvalds 已提交
5261 5262 5263
	double_rq_lock(rq_src, rq_dest);
	/* Already moved. */
	if (task_cpu(p) != src_cpu)
L
Linus Torvalds 已提交
5264
		goto done;
L
Linus Torvalds 已提交
5265
	/* Affinity changed (again). */
5266
	if (!cpumask_test_cpu(dest_cpu, tsk_cpus_allowed(p)))
L
Linus Torvalds 已提交
5267
		goto fail;
L
Linus Torvalds 已提交
5268

5269 5270 5271 5272
	/*
	 * If we're not on a rq, the next wake-up will ensure we're
	 * placed properly.
	 */
P
Peter Zijlstra 已提交
5273
	if (p->on_rq) {
5274
		dequeue_task(rq_src, p, 0);
5275
		set_task_cpu(p, dest_cpu);
5276
		enqueue_task(rq_dest, p, 0);
5277
		check_preempt_curr(rq_dest, p, 0);
L
Linus Torvalds 已提交
5278
	}
L
Linus Torvalds 已提交
5279
done:
5280
	ret = 1;
L
Linus Torvalds 已提交
5281
fail:
L
Linus Torvalds 已提交
5282
	double_rq_unlock(rq_src, rq_dest);
5283
	raw_spin_unlock(&p->pi_lock);
5284
	return ret;
L
Linus Torvalds 已提交
5285 5286 5287
}

/*
5288 5289 5290
 * 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
5291
 */
5292
static int migration_cpu_stop(void *data)
L
Linus Torvalds 已提交
5293
{
5294
	struct migration_arg *arg = data;
5295

5296 5297 5298 5299
	/*
	 * The original target cpu might have gone down and we might
	 * be on another cpu but it doesn't matter.
	 */
5300
	local_irq_disable();
5301
	__migrate_task(arg->task, raw_smp_processor_id(), arg->dest_cpu);
5302
	local_irq_enable();
L
Linus Torvalds 已提交
5303
	return 0;
5304 5305
}

L
Linus Torvalds 已提交
5306
#ifdef CONFIG_HOTPLUG_CPU
5307

5308
/*
5309 5310
 * Ensures that the idle task is using init_mm right before its cpu goes
 * offline.
5311
 */
5312
void idle_task_exit(void)
L
Linus Torvalds 已提交
5313
{
5314
	struct mm_struct *mm = current->active_mm;
5315

5316
	BUG_ON(cpu_online(smp_processor_id()));
5317

5318 5319 5320
	if (mm != &init_mm)
		switch_mm(mm, &init_mm, current);
	mmdrop(mm);
L
Linus Torvalds 已提交
5321 5322 5323
}

/*
5324 5325 5326 5327 5328
 * Since this CPU is going 'away' for a while, fold any nr_active delta
 * we might have. Assumes we're called after migrate_tasks() so that the
 * nr_active count is stable.
 *
 * Also see the comment "Global load-average calculations".
L
Linus Torvalds 已提交
5329
 */
5330
static void calc_load_migrate(struct rq *rq)
L
Linus Torvalds 已提交
5331
{
5332 5333 5334
	long delta = calc_load_fold_active(rq);
	if (delta)
		atomic_long_add(delta, &calc_load_tasks);
L
Linus Torvalds 已提交
5335 5336
}

5337
/*
5338 5339 5340 5341 5342 5343
 * 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
5344
 */
5345
static void migrate_tasks(unsigned int dead_cpu)
L
Linus Torvalds 已提交
5346
{
5347
	struct rq *rq = cpu_rq(dead_cpu);
5348 5349
	struct task_struct *next, *stop = rq->stop;
	int dest_cpu;
L
Linus Torvalds 已提交
5350 5351

	/*
5352 5353 5354 5355 5356 5357 5358
	 * 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
5359
	 */
5360
	rq->stop = NULL;
5361

I
Ingo Molnar 已提交
5362
	for ( ; ; ) {
5363 5364 5365 5366 5367
		/*
		 * There's this thread running, bail when that's the only
		 * remaining thread.
		 */
		if (rq->nr_running == 1)
I
Ingo Molnar 已提交
5368
			break;
5369

5370
		next = pick_next_task(rq);
5371
		BUG_ON(!next);
D
Dmitry Adamushko 已提交
5372
		next->sched_class->put_prev_task(rq, next);
5373

5374 5375 5376 5377 5378 5379 5380
		/* 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
5381
	}
5382

5383
	rq->stop = stop;
5384
}
5385

L
Linus Torvalds 已提交
5386 5387
#endif /* CONFIG_HOTPLUG_CPU */

5388 5389 5390
#if defined(CONFIG_SCHED_DEBUG) && defined(CONFIG_SYSCTL)

static struct ctl_table sd_ctl_dir[] = {
5391 5392
	{
		.procname	= "sched_domain",
5393
		.mode		= 0555,
5394
	},
5395
	{}
5396 5397 5398
};

static struct ctl_table sd_ctl_root[] = {
5399 5400
	{
		.procname	= "kernel",
5401
		.mode		= 0555,
5402 5403
		.child		= sd_ctl_dir,
	},
5404
	{}
5405 5406 5407 5408 5409
};

static struct ctl_table *sd_alloc_ctl_entry(int n)
{
	struct ctl_table *entry =
5410
		kcalloc(n, sizeof(struct ctl_table), GFP_KERNEL);
5411 5412 5413 5414

	return entry;
}

5415 5416
static void sd_free_ctl_entry(struct ctl_table **tablep)
{
5417
	struct ctl_table *entry;
5418

5419 5420 5421
	/*
	 * 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
5422
	 * will always be set. In the lowest directory the names are
5423 5424 5425
	 * static strings and all have proc handlers.
	 */
	for (entry = *tablep; entry->mode; entry++) {
5426 5427
		if (entry->child)
			sd_free_ctl_entry(&entry->child);
5428 5429 5430
		if (entry->proc_handler == NULL)
			kfree(entry->procname);
	}
5431 5432 5433 5434 5435

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

5436
static void
5437
set_table_entry(struct ctl_table *entry,
5438
		const char *procname, void *data, int maxlen,
5439
		umode_t mode, proc_handler *proc_handler)
5440 5441 5442 5443 5444 5445 5446 5447 5448 5449 5450
{
	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)
{
5451
	struct ctl_table *table = sd_alloc_ctl_entry(13);
5452

5453 5454 5455
	if (table == NULL)
		return NULL;

5456
	set_table_entry(&table[0], "min_interval", &sd->min_interval,
5457
		sizeof(long), 0644, proc_doulongvec_minmax);
5458
	set_table_entry(&table[1], "max_interval", &sd->max_interval,
5459
		sizeof(long), 0644, proc_doulongvec_minmax);
5460
	set_table_entry(&table[2], "busy_idx", &sd->busy_idx,
5461
		sizeof(int), 0644, proc_dointvec_minmax);
5462
	set_table_entry(&table[3], "idle_idx", &sd->idle_idx,
5463
		sizeof(int), 0644, proc_dointvec_minmax);
5464
	set_table_entry(&table[4], "newidle_idx", &sd->newidle_idx,
5465
		sizeof(int), 0644, proc_dointvec_minmax);
5466
	set_table_entry(&table[5], "wake_idx", &sd->wake_idx,
5467
		sizeof(int), 0644, proc_dointvec_minmax);
5468
	set_table_entry(&table[6], "forkexec_idx", &sd->forkexec_idx,
5469
		sizeof(int), 0644, proc_dointvec_minmax);
5470
	set_table_entry(&table[7], "busy_factor", &sd->busy_factor,
5471
		sizeof(int), 0644, proc_dointvec_minmax);
5472
	set_table_entry(&table[8], "imbalance_pct", &sd->imbalance_pct,
5473
		sizeof(int), 0644, proc_dointvec_minmax);
5474
	set_table_entry(&table[9], "cache_nice_tries",
5475 5476
		&sd->cache_nice_tries,
		sizeof(int), 0644, proc_dointvec_minmax);
5477
	set_table_entry(&table[10], "flags", &sd->flags,
5478
		sizeof(int), 0644, proc_dointvec_minmax);
5479 5480 5481
	set_table_entry(&table[11], "name", sd->name,
		CORENAME_MAX_SIZE, 0444, proc_dostring);
	/* &table[12] is terminator */
5482 5483 5484 5485

	return table;
}

5486
static ctl_table *sd_alloc_ctl_cpu_table(int cpu)
5487 5488 5489 5490 5491 5492 5493 5494 5495
{
	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);
5496 5497
	if (table == NULL)
		return NULL;
5498 5499 5500 5501 5502

	i = 0;
	for_each_domain(cpu, sd) {
		snprintf(buf, 32, "domain%d", i);
		entry->procname = kstrdup(buf, GFP_KERNEL);
5503
		entry->mode = 0555;
5504 5505 5506 5507 5508 5509 5510 5511
		entry->child = sd_alloc_ctl_domain_table(sd);
		entry++;
		i++;
	}
	return table;
}

static struct ctl_table_header *sd_sysctl_header;
5512
static void register_sched_domain_sysctl(void)
5513
{
5514
	int i, cpu_num = num_possible_cpus();
5515 5516 5517
	struct ctl_table *entry = sd_alloc_ctl_entry(cpu_num + 1);
	char buf[32];

5518 5519 5520
	WARN_ON(sd_ctl_dir[0].child);
	sd_ctl_dir[0].child = entry;

5521 5522 5523
	if (entry == NULL)
		return;

5524
	for_each_possible_cpu(i) {
5525 5526
		snprintf(buf, 32, "cpu%d", i);
		entry->procname = kstrdup(buf, GFP_KERNEL);
5527
		entry->mode = 0555;
5528
		entry->child = sd_alloc_ctl_cpu_table(i);
5529
		entry++;
5530
	}
5531 5532

	WARN_ON(sd_sysctl_header);
5533 5534
	sd_sysctl_header = register_sysctl_table(sd_ctl_root);
}
5535

5536
/* may be called multiple times per register */
5537 5538
static void unregister_sched_domain_sysctl(void)
{
5539 5540
	if (sd_sysctl_header)
		unregister_sysctl_table(sd_sysctl_header);
5541
	sd_sysctl_header = NULL;
5542 5543
	if (sd_ctl_dir[0].child)
		sd_free_ctl_entry(&sd_ctl_dir[0].child);
5544
}
5545
#else
5546 5547 5548 5549
static void register_sched_domain_sysctl(void)
{
}
static void unregister_sched_domain_sysctl(void)
5550 5551 5552 5553
{
}
#endif

5554 5555 5556 5557 5558
static void set_rq_online(struct rq *rq)
{
	if (!rq->online) {
		const struct sched_class *class;

5559
		cpumask_set_cpu(rq->cpu, rq->rd->online);
5560 5561 5562 5563 5564 5565 5566 5567 5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578
		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);
		}

5579
		cpumask_clear_cpu(rq->cpu, rq->rd->online);
5580 5581 5582 5583
		rq->online = 0;
	}
}

L
Linus Torvalds 已提交
5584 5585 5586 5587
/*
 * migration_call - callback that gets triggered when a CPU is added.
 * Here we can start up the necessary migration thread for the new CPU.
 */
5588 5589
static int __cpuinit
migration_call(struct notifier_block *nfb, unsigned long action, void *hcpu)
L
Linus Torvalds 已提交
5590
{
5591
	int cpu = (long)hcpu;
L
Linus Torvalds 已提交
5592
	unsigned long flags;
5593
	struct rq *rq = cpu_rq(cpu);
L
Linus Torvalds 已提交
5594

5595
	switch (action & ~CPU_TASKS_FROZEN) {
5596

L
Linus Torvalds 已提交
5597
	case CPU_UP_PREPARE:
5598
		rq->calc_load_update = calc_load_update;
L
Linus Torvalds 已提交
5599
		break;
5600

L
Linus Torvalds 已提交
5601
	case CPU_ONLINE:
5602
		/* Update our root-domain */
5603
		raw_spin_lock_irqsave(&rq->lock, flags);
5604
		if (rq->rd) {
5605
			BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
5606 5607

			set_rq_online(rq);
5608
		}
5609
		raw_spin_unlock_irqrestore(&rq->lock, flags);
L
Linus Torvalds 已提交
5610
		break;
5611

L
Linus Torvalds 已提交
5612
#ifdef CONFIG_HOTPLUG_CPU
5613
	case CPU_DYING:
5614
		sched_ttwu_pending();
G
Gregory Haskins 已提交
5615
		/* Update our root-domain */
5616
		raw_spin_lock_irqsave(&rq->lock, flags);
G
Gregory Haskins 已提交
5617
		if (rq->rd) {
5618
			BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
5619
			set_rq_offline(rq);
G
Gregory Haskins 已提交
5620
		}
5621 5622
		migrate_tasks(cpu);
		BUG_ON(rq->nr_running != 1); /* the migration thread */
5623
		raw_spin_unlock_irqrestore(&rq->lock, flags);
5624
		break;
5625

5626
	case CPU_DEAD:
5627
		calc_load_migrate(rq);
G
Gregory Haskins 已提交
5628
		break;
L
Linus Torvalds 已提交
5629 5630
#endif
	}
5631 5632 5633

	update_max_interval();

L
Linus Torvalds 已提交
5634 5635 5636
	return NOTIFY_OK;
}

5637 5638 5639
/*
 * Register at high priority so that task migration (migrate_all_tasks)
 * happens before everything else.  This has to be lower priority than
5640
 * the notifier in the perf_event subsystem, though.
L
Linus Torvalds 已提交
5641
 */
5642
static struct notifier_block __cpuinitdata migration_notifier = {
L
Linus Torvalds 已提交
5643
	.notifier_call = migration_call,
5644
	.priority = CPU_PRI_MIGRATION,
L
Linus Torvalds 已提交
5645 5646
};

5647 5648 5649 5650
static int __cpuinit sched_cpu_active(struct notifier_block *nfb,
				      unsigned long action, void *hcpu)
{
	switch (action & ~CPU_TASKS_FROZEN) {
5651
	case CPU_STARTING:
5652 5653 5654 5655 5656 5657 5658 5659 5660 5661 5662 5663 5664 5665 5666 5667 5668 5669 5670 5671
	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;
	}
}

5672
static int __init migration_init(void)
L
Linus Torvalds 已提交
5673 5674
{
	void *cpu = (void *)(long)smp_processor_id();
5675
	int err;
5676

5677
	/* Initialize migration for the boot CPU */
5678 5679
	err = migration_call(&migration_notifier, CPU_UP_PREPARE, cpu);
	BUG_ON(err == NOTIFY_BAD);
L
Linus Torvalds 已提交
5680 5681
	migration_call(&migration_notifier, CPU_ONLINE, cpu);
	register_cpu_notifier(&migration_notifier);
5682

5683 5684 5685 5686
	/* Register cpu active notifiers */
	cpu_notifier(sched_cpu_active, CPU_PRI_SCHED_ACTIVE);
	cpu_notifier(sched_cpu_inactive, CPU_PRI_SCHED_INACTIVE);

5687
	return 0;
L
Linus Torvalds 已提交
5688
}
5689
early_initcall(migration_init);
L
Linus Torvalds 已提交
5690 5691 5692
#endif

#ifdef CONFIG_SMP
5693

5694 5695
static cpumask_var_t sched_domains_tmpmask; /* sched_domains_mutex */

5696
#ifdef CONFIG_SCHED_DEBUG
I
Ingo Molnar 已提交
5697

5698
static __read_mostly int sched_debug_enabled;
5699

5700
static int __init sched_debug_setup(char *str)
5701
{
5702
	sched_debug_enabled = 1;
5703 5704 5705

	return 0;
}
5706 5707 5708 5709 5710 5711
early_param("sched_debug", sched_debug_setup);

static inline bool sched_debug(void)
{
	return sched_debug_enabled;
}
5712

5713
static int sched_domain_debug_one(struct sched_domain *sd, int cpu, int level,
5714
				  struct cpumask *groupmask)
L
Linus Torvalds 已提交
5715
{
I
Ingo Molnar 已提交
5716
	struct sched_group *group = sd->groups;
5717
	char str[256];
L
Linus Torvalds 已提交
5718

R
Rusty Russell 已提交
5719
	cpulist_scnprintf(str, sizeof(str), sched_domain_span(sd));
5720
	cpumask_clear(groupmask);
I
Ingo Molnar 已提交
5721 5722 5723 5724

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

	if (!(sd->flags & SD_LOAD_BALANCE)) {
P
Peter Zijlstra 已提交
5725
		printk("does not load-balance\n");
I
Ingo Molnar 已提交
5726
		if (sd->parent)
P
Peter Zijlstra 已提交
5727 5728
			printk(KERN_ERR "ERROR: !SD_LOAD_BALANCE domain"
					" has parent");
I
Ingo Molnar 已提交
5729
		return -1;
N
Nick Piggin 已提交
5730 5731
	}

P
Peter Zijlstra 已提交
5732
	printk(KERN_CONT "span %s level %s\n", str, sd->name);
I
Ingo Molnar 已提交
5733

5734
	if (!cpumask_test_cpu(cpu, sched_domain_span(sd))) {
P
Peter Zijlstra 已提交
5735 5736
		printk(KERN_ERR "ERROR: domain->span does not contain "
				"CPU%d\n", cpu);
I
Ingo Molnar 已提交
5737
	}
5738
	if (!cpumask_test_cpu(cpu, sched_group_cpus(group))) {
P
Peter Zijlstra 已提交
5739 5740
		printk(KERN_ERR "ERROR: domain->groups does not contain"
				" CPU%d\n", cpu);
I
Ingo Molnar 已提交
5741
	}
L
Linus Torvalds 已提交
5742

I
Ingo Molnar 已提交
5743
	printk(KERN_DEBUG "%*s groups:", level + 1, "");
L
Linus Torvalds 已提交
5744
	do {
I
Ingo Molnar 已提交
5745
		if (!group) {
P
Peter Zijlstra 已提交
5746 5747
			printk("\n");
			printk(KERN_ERR "ERROR: group is NULL\n");
L
Linus Torvalds 已提交
5748 5749 5750
			break;
		}

5751 5752 5753 5754 5755 5756
		/*
		 * Even though we initialize ->power to something semi-sane,
		 * we leave power_orig unset. This allows us to detect if
		 * domain iteration is still funny without causing /0 traps.
		 */
		if (!group->sgp->power_orig) {
P
Peter Zijlstra 已提交
5757 5758 5759
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: domain->cpu_power not "
					"set\n");
I
Ingo Molnar 已提交
5760 5761
			break;
		}
L
Linus Torvalds 已提交
5762

5763
		if (!cpumask_weight(sched_group_cpus(group))) {
P
Peter Zijlstra 已提交
5764 5765
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: empty group\n");
I
Ingo Molnar 已提交
5766 5767
			break;
		}
L
Linus Torvalds 已提交
5768

5769 5770
		if (!(sd->flags & SD_OVERLAP) &&
		    cpumask_intersects(groupmask, sched_group_cpus(group))) {
P
Peter Zijlstra 已提交
5771 5772
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: repeated CPUs\n");
I
Ingo Molnar 已提交
5773 5774
			break;
		}
L
Linus Torvalds 已提交
5775

5776
		cpumask_or(groupmask, groupmask, sched_group_cpus(group));
L
Linus Torvalds 已提交
5777

R
Rusty Russell 已提交
5778
		cpulist_scnprintf(str, sizeof(str), sched_group_cpus(group));
5779

P
Peter Zijlstra 已提交
5780
		printk(KERN_CONT " %s", str);
5781
		if (group->sgp->power != SCHED_POWER_SCALE) {
P
Peter Zijlstra 已提交
5782
			printk(KERN_CONT " (cpu_power = %d)",
5783
				group->sgp->power);
5784
		}
L
Linus Torvalds 已提交
5785

I
Ingo Molnar 已提交
5786 5787
		group = group->next;
	} while (group != sd->groups);
P
Peter Zijlstra 已提交
5788
	printk(KERN_CONT "\n");
L
Linus Torvalds 已提交
5789

5790
	if (!cpumask_equal(sched_domain_span(sd), groupmask))
P
Peter Zijlstra 已提交
5791
		printk(KERN_ERR "ERROR: groups don't span domain->span\n");
L
Linus Torvalds 已提交
5792

5793 5794
	if (sd->parent &&
	    !cpumask_subset(groupmask, sched_domain_span(sd->parent)))
P
Peter Zijlstra 已提交
5795 5796
		printk(KERN_ERR "ERROR: parent span is not a superset "
			"of domain->span\n");
I
Ingo Molnar 已提交
5797 5798
	return 0;
}
L
Linus Torvalds 已提交
5799

I
Ingo Molnar 已提交
5800 5801 5802
static void sched_domain_debug(struct sched_domain *sd, int cpu)
{
	int level = 0;
L
Linus Torvalds 已提交
5803

5804
	if (!sched_debug_enabled)
5805 5806
		return;

I
Ingo Molnar 已提交
5807 5808 5809 5810
	if (!sd) {
		printk(KERN_DEBUG "CPU%d attaching NULL sched-domain.\n", cpu);
		return;
	}
L
Linus Torvalds 已提交
5811

I
Ingo Molnar 已提交
5812 5813 5814
	printk(KERN_DEBUG "CPU%d attaching sched-domain:\n", cpu);

	for (;;) {
5815
		if (sched_domain_debug_one(sd, cpu, level, sched_domains_tmpmask))
I
Ingo Molnar 已提交
5816
			break;
L
Linus Torvalds 已提交
5817 5818
		level++;
		sd = sd->parent;
5819
		if (!sd)
I
Ingo Molnar 已提交
5820 5821
			break;
	}
L
Linus Torvalds 已提交
5822
}
5823
#else /* !CONFIG_SCHED_DEBUG */
5824
# define sched_domain_debug(sd, cpu) do { } while (0)
5825 5826 5827 5828
static inline bool sched_debug(void)
{
	return false;
}
5829
#endif /* CONFIG_SCHED_DEBUG */
L
Linus Torvalds 已提交
5830

5831
static int sd_degenerate(struct sched_domain *sd)
5832
{
5833
	if (cpumask_weight(sched_domain_span(sd)) == 1)
5834 5835 5836 5837 5838 5839
		return 1;

	/* Following flags need at least 2 groups */
	if (sd->flags & (SD_LOAD_BALANCE |
			 SD_BALANCE_NEWIDLE |
			 SD_BALANCE_FORK |
5840 5841 5842
			 SD_BALANCE_EXEC |
			 SD_SHARE_CPUPOWER |
			 SD_SHARE_PKG_RESOURCES)) {
5843 5844 5845 5846 5847
		if (sd->groups != sd->groups->next)
			return 0;
	}

	/* Following flags don't use groups */
5848
	if (sd->flags & (SD_WAKE_AFFINE))
5849 5850 5851 5852 5853
		return 0;

	return 1;
}

5854 5855
static int
sd_parent_degenerate(struct sched_domain *sd, struct sched_domain *parent)
5856 5857 5858 5859 5860 5861
{
	unsigned long cflags = sd->flags, pflags = parent->flags;

	if (sd_degenerate(parent))
		return 1;

5862
	if (!cpumask_equal(sched_domain_span(sd), sched_domain_span(parent)))
5863 5864 5865 5866 5867 5868 5869
		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 |
5870 5871 5872
				SD_BALANCE_EXEC |
				SD_SHARE_CPUPOWER |
				SD_SHARE_PKG_RESOURCES);
5873 5874
		if (nr_node_ids == 1)
			pflags &= ~SD_SERIALIZE;
5875 5876 5877 5878 5879 5880 5881
	}
	if (~cflags & pflags)
		return 0;

	return 1;
}

5882
static void free_rootdomain(struct rcu_head *rcu)
5883
{
5884
	struct root_domain *rd = container_of(rcu, struct root_domain, rcu);
5885

5886
	cpupri_cleanup(&rd->cpupri);
5887 5888 5889 5890 5891 5892
	free_cpumask_var(rd->rto_mask);
	free_cpumask_var(rd->online);
	free_cpumask_var(rd->span);
	kfree(rd);
}

G
Gregory Haskins 已提交
5893 5894
static void rq_attach_root(struct rq *rq, struct root_domain *rd)
{
I
Ingo Molnar 已提交
5895
	struct root_domain *old_rd = NULL;
G
Gregory Haskins 已提交
5896 5897
	unsigned long flags;

5898
	raw_spin_lock_irqsave(&rq->lock, flags);
G
Gregory Haskins 已提交
5899 5900

	if (rq->rd) {
I
Ingo Molnar 已提交
5901
		old_rd = rq->rd;
G
Gregory Haskins 已提交
5902

5903
		if (cpumask_test_cpu(rq->cpu, old_rd->online))
5904
			set_rq_offline(rq);
G
Gregory Haskins 已提交
5905

5906
		cpumask_clear_cpu(rq->cpu, old_rd->span);
5907

I
Ingo Molnar 已提交
5908 5909 5910 5911 5912 5913 5914
		/*
		 * 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
5915 5916 5917 5918 5919
	}

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

5920
	cpumask_set_cpu(rq->cpu, rd->span);
5921
	if (cpumask_test_cpu(rq->cpu, cpu_active_mask))
5922
		set_rq_online(rq);
G
Gregory Haskins 已提交
5923

5924
	raw_spin_unlock_irqrestore(&rq->lock, flags);
I
Ingo Molnar 已提交
5925 5926

	if (old_rd)
5927
		call_rcu_sched(&old_rd->rcu, free_rootdomain);
G
Gregory Haskins 已提交
5928 5929
}

5930
static int init_rootdomain(struct root_domain *rd)
G
Gregory Haskins 已提交
5931 5932 5933
{
	memset(rd, 0, sizeof(*rd));

5934
	if (!alloc_cpumask_var(&rd->span, GFP_KERNEL))
5935
		goto out;
5936
	if (!alloc_cpumask_var(&rd->online, GFP_KERNEL))
5937
		goto free_span;
5938
	if (!alloc_cpumask_var(&rd->rto_mask, GFP_KERNEL))
5939
		goto free_online;
5940

5941
	if (cpupri_init(&rd->cpupri) != 0)
5942
		goto free_rto_mask;
5943
	return 0;
5944

5945 5946
free_rto_mask:
	free_cpumask_var(rd->rto_mask);
5947 5948 5949 5950
free_online:
	free_cpumask_var(rd->online);
free_span:
	free_cpumask_var(rd->span);
5951
out:
5952
	return -ENOMEM;
G
Gregory Haskins 已提交
5953 5954
}

5955 5956 5957 5958 5959 5960
/*
 * 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
5961 5962
static void init_defrootdomain(void)
{
5963
	init_rootdomain(&def_root_domain);
5964

G
Gregory Haskins 已提交
5965 5966 5967
	atomic_set(&def_root_domain.refcount, 1);
}

5968
static struct root_domain *alloc_rootdomain(void)
G
Gregory Haskins 已提交
5969 5970 5971 5972 5973 5974 5975
{
	struct root_domain *rd;

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

5976
	if (init_rootdomain(rd) != 0) {
5977 5978 5979
		kfree(rd);
		return NULL;
	}
G
Gregory Haskins 已提交
5980 5981 5982 5983

	return rd;
}

5984 5985 5986 5987 5988 5989 5990 5991 5992 5993 5994 5995 5996 5997 5998 5999 6000 6001 6002
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);
}

6003 6004 6005
static void free_sched_domain(struct rcu_head *rcu)
{
	struct sched_domain *sd = container_of(rcu, struct sched_domain, rcu);
6006 6007 6008 6009 6010 6011 6012 6013

	/*
	 * 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)) {
6014
		kfree(sd->groups->sgp);
6015
		kfree(sd->groups);
6016
	}
6017 6018 6019 6020 6021 6022 6023 6024 6025 6026 6027 6028 6029 6030
	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);
}

6031 6032 6033 6034 6035 6036 6037
/*
 * Keep a special pointer to the highest sched_domain that has
 * SD_SHARE_PKG_RESOURCE set (Last Level Cache Domain) for this
 * allows us to avoid some pointer chasing select_idle_sibling().
 *
 * Also keep a unique ID per domain (we use the first cpu number in
 * the cpumask of the domain), this allows us to quickly tell if
6038
 * two cpus are in the same cache domain, see cpus_share_cache().
6039 6040 6041 6042 6043 6044 6045 6046 6047 6048
 */
DEFINE_PER_CPU(struct sched_domain *, sd_llc);
DEFINE_PER_CPU(int, sd_llc_id);

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

	sd = highest_flag_domain(cpu, SD_SHARE_PKG_RESOURCES);
6049
	if (sd)
6050 6051 6052 6053 6054 6055
		id = cpumask_first(sched_domain_span(sd));

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

L
Linus Torvalds 已提交
6056
/*
I
Ingo Molnar 已提交
6057
 * Attach the domain 'sd' to 'cpu' as its base domain. Callers must
L
Linus Torvalds 已提交
6058 6059
 * hold the hotplug lock.
 */
I
Ingo Molnar 已提交
6060 6061
static void
cpu_attach_domain(struct sched_domain *sd, struct root_domain *rd, int cpu)
L
Linus Torvalds 已提交
6062
{
6063
	struct rq *rq = cpu_rq(cpu);
6064 6065 6066
	struct sched_domain *tmp;

	/* Remove the sched domains which do not contribute to scheduling. */
6067
	for (tmp = sd; tmp; ) {
6068 6069 6070
		struct sched_domain *parent = tmp->parent;
		if (!parent)
			break;
6071

6072
		if (sd_parent_degenerate(tmp, parent)) {
6073
			tmp->parent = parent->parent;
6074 6075
			if (parent->parent)
				parent->parent->child = tmp;
6076
			destroy_sched_domain(parent, cpu);
6077 6078
		} else
			tmp = tmp->parent;
6079 6080
	}

6081
	if (sd && sd_degenerate(sd)) {
6082
		tmp = sd;
6083
		sd = sd->parent;
6084
		destroy_sched_domain(tmp, cpu);
6085 6086 6087
		if (sd)
			sd->child = NULL;
	}
L
Linus Torvalds 已提交
6088

6089
	sched_domain_debug(sd, cpu);
L
Linus Torvalds 已提交
6090

G
Gregory Haskins 已提交
6091
	rq_attach_root(rq, rd);
6092
	tmp = rq->sd;
N
Nick Piggin 已提交
6093
	rcu_assign_pointer(rq->sd, sd);
6094
	destroy_sched_domains(tmp, cpu);
6095 6096

	update_top_cache_domain(cpu);
L
Linus Torvalds 已提交
6097 6098 6099
}

/* cpus with isolated domains */
6100
static cpumask_var_t cpu_isolated_map;
L
Linus Torvalds 已提交
6101 6102 6103 6104

/* Setup the mask of cpus configured for isolated domains */
static int __init isolated_cpu_setup(char *str)
{
R
Rusty Russell 已提交
6105
	alloc_bootmem_cpumask_var(&cpu_isolated_map);
R
Rusty Russell 已提交
6106
	cpulist_parse(str, cpu_isolated_map);
L
Linus Torvalds 已提交
6107 6108 6109
	return 1;
}

I
Ingo Molnar 已提交
6110
__setup("isolcpus=", isolated_cpu_setup);
L
Linus Torvalds 已提交
6111

6112 6113 6114 6115 6116
static const struct cpumask *cpu_cpu_mask(int cpu)
{
	return cpumask_of_node(cpu_to_node(cpu));
}

6117 6118 6119
struct sd_data {
	struct sched_domain **__percpu sd;
	struct sched_group **__percpu sg;
6120
	struct sched_group_power **__percpu sgp;
6121 6122
};

6123
struct s_data {
6124
	struct sched_domain ** __percpu sd;
6125 6126 6127
	struct root_domain	*rd;
};

6128 6129
enum s_alloc {
	sa_rootdomain,
6130
	sa_sd,
6131
	sa_sd_storage,
6132 6133 6134
	sa_none,
};

6135 6136 6137
struct sched_domain_topology_level;

typedef struct sched_domain *(*sched_domain_init_f)(struct sched_domain_topology_level *tl, int cpu);
6138 6139
typedef const struct cpumask *(*sched_domain_mask_f)(int cpu);

6140 6141
#define SDTL_OVERLAP	0x01

6142
struct sched_domain_topology_level {
6143 6144
	sched_domain_init_f init;
	sched_domain_mask_f mask;
6145
	int		    flags;
6146
	int		    numa_level;
6147
	struct sd_data      data;
6148 6149
};

P
Peter Zijlstra 已提交
6150 6151 6152 6153 6154 6155 6156 6157 6158 6159 6160 6161 6162 6163 6164 6165 6166 6167 6168 6169 6170 6171 6172 6173 6174 6175 6176 6177 6178 6179 6180 6181 6182 6183 6184 6185 6186 6187
/*
 * Build an iteration mask that can exclude certain CPUs from the upwards
 * domain traversal.
 *
 * Asymmetric node setups can result in situations where the domain tree is of
 * unequal depth, make sure to skip domains that already cover the entire
 * range.
 *
 * In that case build_sched_domains() will have terminated the iteration early
 * and our sibling sd spans will be empty. Domains should always include the
 * cpu they're built on, so check that.
 *
 */
static void build_group_mask(struct sched_domain *sd, struct sched_group *sg)
{
	const struct cpumask *span = sched_domain_span(sd);
	struct sd_data *sdd = sd->private;
	struct sched_domain *sibling;
	int i;

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

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

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

6188 6189 6190 6191 6192 6193 6194 6195 6196 6197 6198 6199 6200 6201 6202 6203 6204 6205
static int
build_overlap_sched_groups(struct sched_domain *sd, int cpu)
{
	struct sched_group *first = NULL, *last = NULL, *groups = NULL, *sg;
	const struct cpumask *span = sched_domain_span(sd);
	struct cpumask *covered = sched_domains_tmpmask;
	struct sd_data *sdd = sd->private;
	struct sched_domain *child;
	int i;

	cpumask_clear(covered);

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

		if (cpumask_test_cpu(i, covered))
			continue;

P
Peter Zijlstra 已提交
6206 6207 6208 6209 6210 6211
		child = *per_cpu_ptr(sdd->sd, i);

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

6212
		sg = kzalloc_node(sizeof(struct sched_group) + cpumask_size(),
6213
				GFP_KERNEL, cpu_to_node(cpu));
6214 6215 6216 6217 6218 6219 6220 6221 6222 6223 6224 6225 6226

		if (!sg)
			goto fail;

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

		cpumask_or(covered, covered, sg_span);

P
Peter Zijlstra 已提交
6227
		sg->sgp = *per_cpu_ptr(sdd->sgp, i);
P
Peter Zijlstra 已提交
6228 6229 6230
		if (atomic_inc_return(&sg->sgp->ref) == 1)
			build_group_mask(sd, sg);

6231 6232 6233 6234 6235 6236
		/*
		 * Initialize sgp->power such that even if we mess up the
		 * domains and no possible iteration will get us here, we won't
		 * die on a /0 trap.
		 */
		sg->sgp->power = SCHED_POWER_SCALE * cpumask_weight(sg_span);
6237

P
Peter Zijlstra 已提交
6238 6239 6240 6241 6242
		/*
		 * Make sure the first group of this domain contains the
		 * canonical balance cpu. Otherwise the sched_domain iteration
		 * breaks. See update_sg_lb_stats().
		 */
P
Peter Zijlstra 已提交
6243
		if ((!groups && cpumask_test_cpu(cpu, sg_span)) ||
P
Peter Zijlstra 已提交
6244
		    group_balance_cpu(sg) == cpu)
6245 6246 6247 6248 6249 6250 6251 6252 6253 6254 6255 6256 6257 6258 6259 6260 6261 6262 6263
			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;
}

6264
static int get_group(int cpu, struct sd_data *sdd, struct sched_group **sg)
L
Linus Torvalds 已提交
6265
{
6266 6267
	struct sched_domain *sd = *per_cpu_ptr(sdd->sd, cpu);
	struct sched_domain *child = sd->child;
L
Linus Torvalds 已提交
6268

6269 6270
	if (child)
		cpu = cpumask_first(sched_domain_span(child));
6271

6272
	if (sg) {
6273
		*sg = *per_cpu_ptr(sdd->sg, cpu);
6274
		(*sg)->sgp = *per_cpu_ptr(sdd->sgp, cpu);
6275
		atomic_set(&(*sg)->sgp->ref, 1); /* for claim_allocations */
6276
	}
6277 6278

	return cpu;
6279 6280
}

6281
/*
6282 6283 6284
 * 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.
6285 6286
 *
 * Assumes the sched_domain tree is fully constructed
6287
 */
6288 6289
static int
build_sched_groups(struct sched_domain *sd, int cpu)
L
Linus Torvalds 已提交
6290
{
6291 6292 6293
	struct sched_group *first = NULL, *last = NULL;
	struct sd_data *sdd = sd->private;
	const struct cpumask *span = sched_domain_span(sd);
6294
	struct cpumask *covered;
6295
	int i;
6296

6297 6298 6299 6300 6301 6302
	get_group(cpu, sdd, &sd->groups);
	atomic_inc(&sd->groups->ref);

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

6303 6304 6305
	lockdep_assert_held(&sched_domains_mutex);
	covered = sched_domains_tmpmask;

6306
	cpumask_clear(covered);
6307

6308 6309 6310 6311
	for_each_cpu(i, span) {
		struct sched_group *sg;
		int group = get_group(i, sdd, &sg);
		int j;
6312

6313 6314
		if (cpumask_test_cpu(i, covered))
			continue;
6315

6316
		cpumask_clear(sched_group_cpus(sg));
6317
		sg->sgp->power = 0;
P
Peter Zijlstra 已提交
6318
		cpumask_setall(sched_group_mask(sg));
6319

6320 6321 6322
		for_each_cpu(j, span) {
			if (get_group(j, sdd, NULL) != group)
				continue;
6323

6324 6325 6326
			cpumask_set_cpu(j, covered);
			cpumask_set_cpu(j, sched_group_cpus(sg));
		}
6327

6328 6329 6330 6331 6332 6333 6334
		if (!first)
			first = sg;
		if (last)
			last->next = sg;
		last = sg;
	}
	last->next = first;
6335 6336

	return 0;
6337
}
6338

6339 6340 6341 6342 6343 6344 6345 6346 6347 6348 6349 6350
/*
 * 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)
{
6351
	struct sched_group *sg = sd->groups;
6352

6353 6354 6355 6356 6357 6358
	WARN_ON(!sd || !sg);

	do {
		sg->group_weight = cpumask_weight(sched_group_cpus(sg));
		sg = sg->next;
	} while (sg != sd->groups);
6359

P
Peter Zijlstra 已提交
6360
	if (cpu != group_balance_cpu(sg))
6361
		return;
6362

6363
	update_group_power(sd, cpu);
6364
	atomic_set(&sg->sgp->nr_busy_cpus, sg->group_weight);
6365 6366
}

6367 6368 6369
int __weak arch_sd_sibling_asym_packing(void)
{
       return 0*SD_ASYM_PACKING;
6370 6371
}

6372 6373 6374 6375 6376
/*
 * Initializers for schedule domains
 * Non-inlined to reduce accumulated stack pressure in build_sched_domains()
 */

6377 6378 6379 6380 6381 6382
#ifdef CONFIG_SCHED_DEBUG
# define SD_INIT_NAME(sd, type)		sd->name = #type
#else
# define SD_INIT_NAME(sd, type)		do { } while (0)
#endif

6383 6384 6385 6386 6387 6388 6389 6390 6391
#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;							\
6392 6393 6394 6395 6396 6397 6398 6399 6400
}

SD_INIT_FUNC(CPU)
#ifdef CONFIG_SCHED_SMT
 SD_INIT_FUNC(SIBLING)
#endif
#ifdef CONFIG_SCHED_MC
 SD_INIT_FUNC(MC)
#endif
6401 6402 6403
#ifdef CONFIG_SCHED_BOOK
 SD_INIT_FUNC(BOOK)
#endif
6404

6405
static int default_relax_domain_level = -1;
6406
int sched_domain_level_max;
6407 6408 6409

static int __init setup_relax_domain_level(char *str)
{
6410 6411
	if (kstrtoint(str, 0, &default_relax_domain_level))
		pr_warn("Unable to set relax_domain_level\n");
6412

6413 6414 6415 6416 6417 6418 6419 6420 6421 6422 6423 6424 6425 6426 6427 6428 6429 6430
	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 */
6431
		sd->flags &= ~(SD_BALANCE_WAKE|SD_BALANCE_NEWIDLE);
6432 6433
	} else {
		/* turn on idle balance on this domain */
6434
		sd->flags |= (SD_BALANCE_WAKE|SD_BALANCE_NEWIDLE);
6435 6436 6437
	}
}

6438 6439 6440
static void __sdt_free(const struct cpumask *cpu_map);
static int __sdt_alloc(const struct cpumask *cpu_map);

6441 6442 6443 6444 6445
static void __free_domain_allocs(struct s_data *d, enum s_alloc what,
				 const struct cpumask *cpu_map)
{
	switch (what) {
	case sa_rootdomain:
6446 6447
		if (!atomic_read(&d->rd->refcount))
			free_rootdomain(&d->rd->rcu); /* fall through */
6448 6449
	case sa_sd:
		free_percpu(d->sd); /* fall through */
6450
	case sa_sd_storage:
6451
		__sdt_free(cpu_map); /* fall through */
6452 6453 6454 6455
	case sa_none:
		break;
	}
}
6456

6457 6458 6459
static enum s_alloc __visit_domain_allocation_hell(struct s_data *d,
						   const struct cpumask *cpu_map)
{
6460 6461
	memset(d, 0, sizeof(*d));

6462 6463
	if (__sdt_alloc(cpu_map))
		return sa_sd_storage;
6464 6465 6466
	d->sd = alloc_percpu(struct sched_domain *);
	if (!d->sd)
		return sa_sd_storage;
6467
	d->rd = alloc_rootdomain();
6468
	if (!d->rd)
6469
		return sa_sd;
6470 6471
	return sa_rootdomain;
}
G
Gregory Haskins 已提交
6472

6473 6474 6475 6476 6477 6478 6479 6480 6481 6482 6483 6484
/*
 * 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;

6485
	if (atomic_read(&(*per_cpu_ptr(sdd->sg, cpu))->ref))
6486
		*per_cpu_ptr(sdd->sg, cpu) = NULL;
6487 6488

	if (atomic_read(&(*per_cpu_ptr(sdd->sgp, cpu))->ref))
6489
		*per_cpu_ptr(sdd->sgp, cpu) = NULL;
6490 6491
}

6492 6493
#ifdef CONFIG_SCHED_SMT
static const struct cpumask *cpu_smt_mask(int cpu)
6494
{
6495
	return topology_thread_cpumask(cpu);
6496
}
6497
#endif
6498

6499 6500 6501
/*
 * Topology list, bottom-up.
 */
6502
static struct sched_domain_topology_level default_topology[] = {
6503 6504
#ifdef CONFIG_SCHED_SMT
	{ sd_init_SIBLING, cpu_smt_mask, },
6505
#endif
6506
#ifdef CONFIG_SCHED_MC
6507
	{ sd_init_MC, cpu_coregroup_mask, },
6508
#endif
6509 6510 6511 6512
#ifdef CONFIG_SCHED_BOOK
	{ sd_init_BOOK, cpu_book_mask, },
#endif
	{ sd_init_CPU, cpu_cpu_mask, },
6513 6514 6515 6516 6517
	{ NULL, },
};

static struct sched_domain_topology_level *sched_domain_topology = default_topology;

6518 6519 6520 6521 6522 6523 6524 6525 6526
#ifdef CONFIG_NUMA

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

static inline int sd_local_flags(int level)
{
6527
	if (sched_domains_numa_distance[level] > RECLAIM_DISTANCE)
6528 6529 6530 6531 6532 6533 6534 6535 6536 6537 6538 6539 6540 6541 6542 6543 6544
		return 0;

	return SD_BALANCE_EXEC | SD_BALANCE_FORK | SD_WAKE_AFFINE;
}

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

	*sd = (struct sched_domain){
		.min_interval		= sd_weight,
		.max_interval		= 2*sd_weight,
		.busy_factor		= 32,
6545
		.imbalance_pct		= 125,
6546 6547 6548 6549 6550 6551 6552 6553 6554 6555 6556 6557 6558 6559 6560 6561 6562 6563 6564 6565 6566 6567 6568 6569 6570 6571 6572 6573 6574 6575 6576 6577 6578 6579 6580 6581 6582 6583 6584
		.cache_nice_tries	= 2,
		.busy_idx		= 3,
		.idle_idx		= 2,
		.newidle_idx		= 0,
		.wake_idx		= 0,
		.forkexec_idx		= 0,

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

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

	return sd;
}

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

6585 6586 6587 6588 6589 6590 6591 6592 6593 6594 6595 6596 6597 6598 6599 6600 6601 6602 6603 6604 6605 6606 6607 6608 6609 6610 6611 6612 6613 6614 6615 6616 6617 6618 6619 6620
static void sched_numa_warn(const char *str)
{
	static int done = false;
	int i,j;

	if (done)
		return;

	done = true;

	printk(KERN_WARNING "ERROR: %s\n\n", str);

	for (i = 0; i < nr_node_ids; i++) {
		printk(KERN_WARNING "  ");
		for (j = 0; j < nr_node_ids; j++)
			printk(KERN_CONT "%02d ", node_distance(i,j));
		printk(KERN_CONT "\n");
	}
	printk(KERN_WARNING "\n");
}

static bool find_numa_distance(int distance)
{
	int i;

	if (distance == node_distance(0, 0))
		return true;

	for (i = 0; i < sched_domains_numa_levels; i++) {
		if (sched_domains_numa_distance[i] == distance)
			return true;
	}

	return false;
}

6621 6622 6623 6624 6625 6626 6627 6628 6629 6630 6631 6632 6633 6634 6635 6636 6637 6638 6639 6640 6641
static void sched_init_numa(void)
{
	int next_distance, curr_distance = node_distance(0, 0);
	struct sched_domain_topology_level *tl;
	int level = 0;
	int i, j, k;

	sched_domains_numa_distance = kzalloc(sizeof(int) * nr_node_ids, GFP_KERNEL);
	if (!sched_domains_numa_distance)
		return;

	/*
	 * O(nr_nodes^2) deduplicating selection sort -- in order to find the
	 * unique distances in the node_distance() table.
	 *
	 * Assumes node_distance(0,j) includes all distances in
	 * node_distance(i,j) in order to avoid cubic time.
	 */
	next_distance = curr_distance;
	for (i = 0; i < nr_node_ids; i++) {
		for (j = 0; j < nr_node_ids; j++) {
6642 6643 6644 6645 6646 6647 6648 6649 6650 6651 6652 6653 6654 6655 6656 6657 6658 6659 6660 6661 6662 6663 6664 6665
			for (k = 0; k < nr_node_ids; k++) {
				int distance = node_distance(i, k);

				if (distance > curr_distance &&
				    (distance < next_distance ||
				     next_distance == curr_distance))
					next_distance = distance;

				/*
				 * While not a strong assumption it would be nice to know
				 * about cases where if node A is connected to B, B is not
				 * equally connected to A.
				 */
				if (sched_debug() && node_distance(k, i) != distance)
					sched_numa_warn("Node-distance not symmetric");

				if (sched_debug() && i && !find_numa_distance(distance))
					sched_numa_warn("Node-0 not representative");
			}
			if (next_distance != curr_distance) {
				sched_domains_numa_distance[level++] = next_distance;
				sched_domains_numa_levels = level;
				curr_distance = next_distance;
			} else break;
6666
		}
6667 6668 6669 6670 6671 6672

		/*
		 * In case of sched_debug() we verify the above assumption.
		 */
		if (!sched_debug())
			break;
6673 6674 6675 6676 6677 6678 6679 6680 6681 6682 6683 6684 6685 6686 6687 6688 6689 6690 6691 6692 6693 6694 6695 6696
	}
	/*
	 * 'level' contains the number of unique distances, excluding the
	 * identity distance node_distance(i,i).
	 *
	 * The sched_domains_nume_distance[] array includes the actual distance
	 * numbers.
	 */

	sched_domains_numa_masks = kzalloc(sizeof(void *) * level, GFP_KERNEL);
	if (!sched_domains_numa_masks)
		return;

	/*
	 * Now for each level, construct a mask per node which contains all
	 * cpus of nodes that are that many hops away from us.
	 */
	for (i = 0; i < level; i++) {
		sched_domains_numa_masks[i] =
			kzalloc(nr_node_ids * sizeof(void *), GFP_KERNEL);
		if (!sched_domains_numa_masks[i])
			return;

		for (j = 0; j < nr_node_ids; j++) {
6697
			struct cpumask *mask = kzalloc(cpumask_size(), GFP_KERNEL);
6698 6699 6700 6701 6702 6703
			if (!mask)
				return;

			sched_domains_numa_masks[i][j] = mask;

			for (k = 0; k < nr_node_ids; k++) {
6704
				if (node_distance(j, k) > sched_domains_numa_distance[i])
6705 6706 6707 6708 6709 6710 6711 6712 6713 6714 6715 6716 6717 6718 6719 6720 6721 6722 6723 6724 6725 6726 6727 6728 6729 6730 6731 6732 6733 6734 6735 6736 6737 6738 6739 6740 6741 6742
					continue;

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

	tl = kzalloc((ARRAY_SIZE(default_topology) + level) *
			sizeof(struct sched_domain_topology_level), GFP_KERNEL);
	if (!tl)
		return;

	/*
	 * Copy the default topology bits..
	 */
	for (i = 0; default_topology[i].init; i++)
		tl[i] = default_topology[i];

	/*
	 * .. and append 'j' levels of NUMA goodness.
	 */
	for (j = 0; j < level; i++, j++) {
		tl[i] = (struct sched_domain_topology_level){
			.init = sd_numa_init,
			.mask = sd_numa_mask,
			.flags = SDTL_OVERLAP,
			.numa_level = j,
		};
	}

	sched_domain_topology = tl;
}
#else
static inline void sched_init_numa(void)
{
}
#endif /* CONFIG_NUMA */

6743 6744 6745 6746 6747 6748 6749 6750 6751 6752 6753 6754 6755 6756 6757 6758
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;

6759 6760 6761 6762
		sdd->sgp = alloc_percpu(struct sched_group_power *);
		if (!sdd->sgp)
			return -ENOMEM;

6763 6764 6765
		for_each_cpu(j, cpu_map) {
			struct sched_domain *sd;
			struct sched_group *sg;
6766
			struct sched_group_power *sgp;
6767 6768 6769 6770 6771 6772 6773 6774 6775 6776 6777 6778 6779

		       	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;

6780 6781
			sg->next = sg;

6782
			*per_cpu_ptr(sdd->sg, j) = sg;
6783

P
Peter Zijlstra 已提交
6784
			sgp = kzalloc_node(sizeof(struct sched_group_power) + cpumask_size(),
6785 6786 6787 6788 6789
					GFP_KERNEL, cpu_to_node(j));
			if (!sgp)
				return -ENOMEM;

			*per_cpu_ptr(sdd->sgp, j) = sgp;
6790 6791 6792 6793 6794 6795 6796 6797 6798 6799 6800 6801 6802 6803 6804
		}
	}

	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) {
6805 6806 6807 6808 6809 6810 6811 6812 6813 6814 6815 6816 6817
			struct sched_domain *sd;

			if (sdd->sd) {
				sd = *per_cpu_ptr(sdd->sd, j);
				if (sd && (sd->flags & SD_OVERLAP))
					free_sched_groups(sd->groups, 0);
				kfree(*per_cpu_ptr(sdd->sd, j));
			}

			if (sdd->sg)
				kfree(*per_cpu_ptr(sdd->sg, j));
			if (sdd->sgp)
				kfree(*per_cpu_ptr(sdd->sgp, j));
6818 6819
		}
		free_percpu(sdd->sd);
6820
		sdd->sd = NULL;
6821
		free_percpu(sdd->sg);
6822
		sdd->sg = NULL;
6823
		free_percpu(sdd->sgp);
6824
		sdd->sgp = NULL;
6825 6826 6827
	}
}

6828 6829
struct sched_domain *build_sched_domain(struct sched_domain_topology_level *tl,
		struct s_data *d, const struct cpumask *cpu_map,
6830
		struct sched_domain_attr *attr, struct sched_domain *child,
6831 6832
		int cpu)
{
6833
	struct sched_domain *sd = tl->init(tl, cpu);
6834
	if (!sd)
6835
		return child;
6836 6837

	cpumask_and(sched_domain_span(sd), cpu_map, tl->mask(cpu));
6838 6839 6840
	if (child) {
		sd->level = child->level + 1;
		sched_domain_level_max = max(sched_domain_level_max, sd->level);
6841
		child->parent = sd;
6842
	}
6843
	sd->child = child;
6844
	set_domain_attribute(sd, attr);
6845 6846 6847 6848

	return sd;
}

6849 6850 6851 6852
/*
 * Build sched domains for a given set of cpus and attach the sched domains
 * to the individual cpus
 */
6853 6854
static int build_sched_domains(const struct cpumask *cpu_map,
			       struct sched_domain_attr *attr)
6855 6856
{
	enum s_alloc alloc_state = sa_none;
6857
	struct sched_domain *sd;
6858
	struct s_data d;
6859
	int i, ret = -ENOMEM;
6860

6861 6862 6863
	alloc_state = __visit_domain_allocation_hell(&d, cpu_map);
	if (alloc_state != sa_rootdomain)
		goto error;
6864

6865
	/* Set up domains for cpus specified by the cpu_map. */
6866
	for_each_cpu(i, cpu_map) {
6867 6868
		struct sched_domain_topology_level *tl;

6869
		sd = NULL;
6870
		for (tl = sched_domain_topology; tl->init; tl++) {
6871
			sd = build_sched_domain(tl, &d, cpu_map, attr, sd, i);
6872 6873
			if (tl->flags & SDTL_OVERLAP || sched_feat(FORCE_SD_OVERLAP))
				sd->flags |= SD_OVERLAP;
6874 6875
			if (cpumask_equal(cpu_map, sched_domain_span(sd)))
				break;
6876
		}
6877

6878 6879 6880
		while (sd->child)
			sd = sd->child;

6881
		*per_cpu_ptr(d.sd, i) = sd;
6882 6883 6884 6885 6886 6887
	}

	/* 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));
6888 6889 6890 6891 6892 6893 6894
			if (sd->flags & SD_OVERLAP) {
				if (build_overlap_sched_groups(sd, i))
					goto error;
			} else {
				if (build_sched_groups(sd, i))
					goto error;
			}
6895
		}
6896
	}
6897

L
Linus Torvalds 已提交
6898
	/* Calculate CPU power for physical packages and nodes */
6899 6900 6901
	for (i = nr_cpumask_bits-1; i >= 0; i--) {
		if (!cpumask_test_cpu(i, cpu_map))
			continue;
6902

6903 6904
		for (sd = *per_cpu_ptr(d.sd, i); sd; sd = sd->parent) {
			claim_allocations(i, sd);
6905
			init_sched_groups_power(i, sd);
6906
		}
6907
	}
6908

L
Linus Torvalds 已提交
6909
	/* Attach the domains */
6910
	rcu_read_lock();
6911
	for_each_cpu(i, cpu_map) {
6912
		sd = *per_cpu_ptr(d.sd, i);
6913
		cpu_attach_domain(sd, d.rd, i);
L
Linus Torvalds 已提交
6914
	}
6915
	rcu_read_unlock();
6916

6917
	ret = 0;
6918
error:
6919
	__free_domain_allocs(&d, alloc_state, cpu_map);
6920
	return ret;
L
Linus Torvalds 已提交
6921
}
P
Paul Jackson 已提交
6922

6923
static cpumask_var_t *doms_cur;	/* current sched domains */
P
Paul Jackson 已提交
6924
static int ndoms_cur;		/* number of sched domains in 'doms_cur' */
I
Ingo Molnar 已提交
6925 6926
static struct sched_domain_attr *dattr_cur;
				/* attribues of custom domains in 'doms_cur' */
P
Paul Jackson 已提交
6927 6928 6929

/*
 * Special case: If a kmalloc of a doms_cur partition (array of
6930 6931
 * cpumask) fails, then fallback to a single sched domain,
 * as determined by the single cpumask fallback_doms.
P
Paul Jackson 已提交
6932
 */
6933
static cpumask_var_t fallback_doms;
P
Paul Jackson 已提交
6934

6935 6936 6937 6938 6939 6940
/*
 * 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)
6941
{
6942
	return 0;
6943 6944
}

6945 6946 6947 6948 6949 6950 6951 6952 6953 6954 6955 6956 6957 6958 6959 6960 6961 6962 6963 6964 6965 6966 6967 6968 6969
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);
}

6970
/*
I
Ingo Molnar 已提交
6971
 * Set up scheduler domains and groups. Callers must hold the hotplug lock.
P
Paul Jackson 已提交
6972 6973
 * For now this just excludes isolated cpus, but could be used to
 * exclude other special cases in the future.
6974
 */
6975
static int init_sched_domains(const struct cpumask *cpu_map)
6976
{
6977 6978
	int err;

6979
	arch_update_cpu_topology();
P
Paul Jackson 已提交
6980
	ndoms_cur = 1;
6981
	doms_cur = alloc_sched_domains(ndoms_cur);
P
Paul Jackson 已提交
6982
	if (!doms_cur)
6983 6984
		doms_cur = &fallback_doms;
	cpumask_andnot(doms_cur[0], cpu_map, cpu_isolated_map);
6985
	err = build_sched_domains(doms_cur[0], NULL);
6986
	register_sched_domain_sysctl();
6987 6988

	return err;
6989 6990 6991 6992 6993 6994
}

/*
 * Detach sched domains from a group of cpus specified in cpu_map
 * These cpus will now be attached to the NULL domain
 */
6995
static void detach_destroy_domains(const struct cpumask *cpu_map)
6996 6997 6998
{
	int i;

6999
	rcu_read_lock();
7000
	for_each_cpu(i, cpu_map)
G
Gregory Haskins 已提交
7001
		cpu_attach_domain(NULL, &def_root_domain, i);
7002
	rcu_read_unlock();
7003 7004
}

7005 7006 7007 7008 7009 7010 7011 7012 7013 7014 7015 7016 7017 7018 7019 7020
/* 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
7021 7022
/*
 * Partition sched domains as specified by the 'ndoms_new'
I
Ingo Molnar 已提交
7023
 * cpumasks in the array doms_new[] of cpumasks. This compares
P
Paul Jackson 已提交
7024 7025 7026
 * doms_new[] to the current sched domain partitioning, doms_cur[].
 * It destroys each deleted domain and builds each new domain.
 *
7027
 * 'doms_new' is an array of cpumask_var_t's of length 'ndoms_new'.
I
Ingo Molnar 已提交
7028 7029 7030
 * 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
7031 7032 7033
 * current 'doms_cur' domains and in the new 'doms_new', we can leave
 * it as it is.
 *
7034 7035 7036 7037 7038 7039
 * 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
7040
 *
7041
 * If doms_new == NULL it will be replaced with cpu_online_mask.
7042 7043
 * ndoms_new == 0 is a special case for destroying existing domains,
 * and it will not create the default domain.
7044
 *
P
Paul Jackson 已提交
7045 7046
 * Call with hotplug lock held
 */
7047
void partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
7048
			     struct sched_domain_attr *dattr_new)
P
Paul Jackson 已提交
7049
{
7050
	int i, j, n;
7051
	int new_topology;
P
Paul Jackson 已提交
7052

7053
	mutex_lock(&sched_domains_mutex);
7054

7055 7056 7057
	/* always unregister in case we don't destroy any domains */
	unregister_sched_domain_sysctl();

7058 7059 7060
	/* Let architecture update cpu core mappings. */
	new_topology = arch_update_cpu_topology();

7061
	n = doms_new ? ndoms_new : 0;
P
Paul Jackson 已提交
7062 7063 7064

	/* Destroy deleted domains */
	for (i = 0; i < ndoms_cur; i++) {
7065
		for (j = 0; j < n && !new_topology; j++) {
7066
			if (cpumask_equal(doms_cur[i], doms_new[j])
7067
			    && dattrs_equal(dattr_cur, i, dattr_new, j))
P
Paul Jackson 已提交
7068 7069 7070
				goto match1;
		}
		/* no match - a current sched domain not in new doms_new[] */
7071
		detach_destroy_domains(doms_cur[i]);
P
Paul Jackson 已提交
7072 7073 7074 7075
match1:
		;
	}

7076 7077
	if (doms_new == NULL) {
		ndoms_cur = 0;
7078
		doms_new = &fallback_doms;
7079
		cpumask_andnot(doms_new[0], cpu_active_mask, cpu_isolated_map);
7080
		WARN_ON_ONCE(dattr_new);
7081 7082
	}

P
Paul Jackson 已提交
7083 7084
	/* Build new domains */
	for (i = 0; i < ndoms_new; i++) {
7085
		for (j = 0; j < ndoms_cur && !new_topology; j++) {
7086
			if (cpumask_equal(doms_new[i], doms_cur[j])
7087
			    && dattrs_equal(dattr_new, i, dattr_cur, j))
P
Paul Jackson 已提交
7088 7089 7090
				goto match2;
		}
		/* no match - add a new doms_new */
7091
		build_sched_domains(doms_new[i], dattr_new ? dattr_new + i : NULL);
P
Paul Jackson 已提交
7092 7093 7094 7095 7096
match2:
		;
	}

	/* Remember the new sched domains */
7097 7098
	if (doms_cur != &fallback_doms)
		free_sched_domains(doms_cur, ndoms_cur);
7099
	kfree(dattr_cur);	/* kfree(NULL) is safe */
P
Paul Jackson 已提交
7100
	doms_cur = doms_new;
7101
	dattr_cur = dattr_new;
P
Paul Jackson 已提交
7102
	ndoms_cur = ndoms_new;
7103 7104

	register_sched_domain_sysctl();
7105

7106
	mutex_unlock(&sched_domains_mutex);
P
Paul Jackson 已提交
7107 7108
}

7109 7110
static int num_cpus_frozen;	/* used to mark begin/end of suspend/resume */

L
Linus Torvalds 已提交
7111
/*
7112 7113 7114
 * Update cpusets according to cpu_active mask.  If cpusets are
 * disabled, cpuset_update_active_cpus() becomes a simple wrapper
 * around partition_sched_domains().
7115 7116 7117
 *
 * If we come here as part of a suspend/resume, don't touch cpusets because we
 * want to restore it back to its original state upon resume anyway.
L
Linus Torvalds 已提交
7118
 */
7119 7120
static int cpuset_cpu_active(struct notifier_block *nfb, unsigned long action,
			     void *hcpu)
7121
{
7122 7123 7124 7125 7126 7127 7128 7129 7130 7131 7132 7133 7134 7135 7136 7137 7138 7139 7140 7141 7142 7143
	switch (action) {
	case CPU_ONLINE_FROZEN:
	case CPU_DOWN_FAILED_FROZEN:

		/*
		 * num_cpus_frozen tracks how many CPUs are involved in suspend
		 * resume sequence. As long as this is not the last online
		 * operation in the resume sequence, just build a single sched
		 * domain, ignoring cpusets.
		 */
		num_cpus_frozen--;
		if (likely(num_cpus_frozen)) {
			partition_sched_domains(1, NULL, NULL);
			break;
		}

		/*
		 * This is the last CPU online operation. So fall through and
		 * restore the original sched domains by considering the
		 * cpuset configurations.
		 */

7144
	case CPU_ONLINE:
7145
	case CPU_DOWN_FAILED:
7146
		cpuset_update_active_cpus(true);
7147
		break;
7148 7149 7150
	default:
		return NOTIFY_DONE;
	}
7151
	return NOTIFY_OK;
7152
}
7153

7154 7155
static int cpuset_cpu_inactive(struct notifier_block *nfb, unsigned long action,
			       void *hcpu)
7156
{
7157
	switch (action) {
7158
	case CPU_DOWN_PREPARE:
7159
		cpuset_update_active_cpus(false);
7160 7161 7162 7163 7164
		break;
	case CPU_DOWN_PREPARE_FROZEN:
		num_cpus_frozen++;
		partition_sched_domains(1, NULL, NULL);
		break;
7165 7166 7167
	default:
		return NOTIFY_DONE;
	}
7168
	return NOTIFY_OK;
7169 7170
}

L
Linus Torvalds 已提交
7171 7172
void __init sched_init_smp(void)
{
7173 7174 7175
	cpumask_var_t non_isolated_cpus;

	alloc_cpumask_var(&non_isolated_cpus, GFP_KERNEL);
7176
	alloc_cpumask_var(&fallback_doms, GFP_KERNEL);
7177

7178 7179
	sched_init_numa();

7180
	get_online_cpus();
7181
	mutex_lock(&sched_domains_mutex);
7182
	init_sched_domains(cpu_active_mask);
7183 7184 7185
	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);
7186
	mutex_unlock(&sched_domains_mutex);
7187
	put_online_cpus();
7188

7189 7190
	hotcpu_notifier(cpuset_cpu_active, CPU_PRI_CPUSET_ACTIVE);
	hotcpu_notifier(cpuset_cpu_inactive, CPU_PRI_CPUSET_INACTIVE);
7191 7192 7193 7194

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

7195
	init_hrtick();
7196 7197

	/* Move init over to a non-isolated CPU */
7198
	if (set_cpus_allowed_ptr(current, non_isolated_cpus) < 0)
7199
		BUG();
I
Ingo Molnar 已提交
7200
	sched_init_granularity();
7201
	free_cpumask_var(non_isolated_cpus);
7202

7203
	init_sched_rt_class();
L
Linus Torvalds 已提交
7204 7205 7206 7207
}
#else
void __init sched_init_smp(void)
{
I
Ingo Molnar 已提交
7208
	sched_init_granularity();
L
Linus Torvalds 已提交
7209 7210 7211
}
#endif /* CONFIG_SMP */

7212 7213
const_debug unsigned int sysctl_timer_migration = 1;

L
Linus Torvalds 已提交
7214 7215 7216 7217 7218 7219 7220
int in_sched_functions(unsigned long addr)
{
	return in_lock_functions(addr) ||
		(addr >= (unsigned long)__sched_text_start
		&& addr < (unsigned long)__sched_text_end);
}

7221 7222
#ifdef CONFIG_CGROUP_SCHED
struct task_group root_task_group;
7223
LIST_HEAD(task_groups);
7224
#endif
P
Peter Zijlstra 已提交
7225

7226
DECLARE_PER_CPU(cpumask_var_t, load_balance_tmpmask);
P
Peter Zijlstra 已提交
7227

L
Linus Torvalds 已提交
7228 7229
void __init sched_init(void)
{
I
Ingo Molnar 已提交
7230
	int i, j;
7231 7232 7233 7234 7235 7236 7237
	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 **);
7238
#endif
7239
#ifdef CONFIG_CPUMASK_OFFSTACK
7240
	alloc_size += num_possible_cpus() * cpumask_size();
7241 7242
#endif
	if (alloc_size) {
7243
		ptr = (unsigned long)kzalloc(alloc_size, GFP_NOWAIT);
7244 7245

#ifdef CONFIG_FAIR_GROUP_SCHED
7246
		root_task_group.se = (struct sched_entity **)ptr;
7247 7248
		ptr += nr_cpu_ids * sizeof(void **);

7249
		root_task_group.cfs_rq = (struct cfs_rq **)ptr;
7250
		ptr += nr_cpu_ids * sizeof(void **);
7251

7252
#endif /* CONFIG_FAIR_GROUP_SCHED */
7253
#ifdef CONFIG_RT_GROUP_SCHED
7254
		root_task_group.rt_se = (struct sched_rt_entity **)ptr;
7255 7256
		ptr += nr_cpu_ids * sizeof(void **);

7257
		root_task_group.rt_rq = (struct rt_rq **)ptr;
7258 7259
		ptr += nr_cpu_ids * sizeof(void **);

7260
#endif /* CONFIG_RT_GROUP_SCHED */
7261 7262 7263 7264 7265 7266
#ifdef CONFIG_CPUMASK_OFFSTACK
		for_each_possible_cpu(i) {
			per_cpu(load_balance_tmpmask, i) = (void *)ptr;
			ptr += cpumask_size();
		}
#endif /* CONFIG_CPUMASK_OFFSTACK */
7267
	}
I
Ingo Molnar 已提交
7268

G
Gregory Haskins 已提交
7269 7270 7271 7272
#ifdef CONFIG_SMP
	init_defrootdomain();
#endif

7273 7274 7275 7276
	init_rt_bandwidth(&def_rt_bandwidth,
			global_rt_period(), global_rt_runtime());

#ifdef CONFIG_RT_GROUP_SCHED
7277
	init_rt_bandwidth(&root_task_group.rt_bandwidth,
7278
			global_rt_period(), global_rt_runtime());
7279
#endif /* CONFIG_RT_GROUP_SCHED */
7280

D
Dhaval Giani 已提交
7281
#ifdef CONFIG_CGROUP_SCHED
7282 7283
	list_add(&root_task_group.list, &task_groups);
	INIT_LIST_HEAD(&root_task_group.children);
7284
	INIT_LIST_HEAD(&root_task_group.siblings);
7285
	autogroup_init(&init_task);
7286

D
Dhaval Giani 已提交
7287
#endif /* CONFIG_CGROUP_SCHED */
P
Peter Zijlstra 已提交
7288

7289 7290 7291 7292 7293 7294
#ifdef CONFIG_CGROUP_CPUACCT
	root_cpuacct.cpustat = &kernel_cpustat;
	root_cpuacct.cpuusage = alloc_percpu(u64);
	/* Too early, not expected to fail */
	BUG_ON(!root_cpuacct.cpuusage);
#endif
7295
	for_each_possible_cpu(i) {
7296
		struct rq *rq;
L
Linus Torvalds 已提交
7297 7298

		rq = cpu_rq(i);
7299
		raw_spin_lock_init(&rq->lock);
N
Nick Piggin 已提交
7300
		rq->nr_running = 0;
7301 7302
		rq->calc_load_active = 0;
		rq->calc_load_update = jiffies + LOAD_FREQ;
7303
		init_cfs_rq(&rq->cfs);
P
Peter Zijlstra 已提交
7304
		init_rt_rq(&rq->rt, rq);
I
Ingo Molnar 已提交
7305
#ifdef CONFIG_FAIR_GROUP_SCHED
7306
		root_task_group.shares = ROOT_TASK_GROUP_LOAD;
P
Peter Zijlstra 已提交
7307
		INIT_LIST_HEAD(&rq->leaf_cfs_rq_list);
D
Dhaval Giani 已提交
7308
		/*
7309
		 * How much cpu bandwidth does root_task_group get?
D
Dhaval Giani 已提交
7310 7311 7312 7313
		 *
		 * 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
7314
		 * root_task_group and its child task-groups in a fair manner,
D
Dhaval Giani 已提交
7315 7316 7317
		 * based on each entity's (task or task-group's) weight
		 * (se->load.weight).
		 *
7318
		 * In other words, if root_task_group has 10 tasks of weight
D
Dhaval Giani 已提交
7319 7320 7321
		 * 1024) and two child groups A0 and A1 (of weight 1024 each),
		 * then A0's share of the cpu resource is:
		 *
7322
		 *	A0's bandwidth = 1024 / (10*1024 + 1024 + 1024) = 8.33%
D
Dhaval Giani 已提交
7323
		 *
7324 7325
		 * 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
7326
		 */
7327
		init_cfs_bandwidth(&root_task_group.cfs_bandwidth);
7328
		init_tg_cfs_entry(&root_task_group, &rq->cfs, NULL, i, NULL);
D
Dhaval Giani 已提交
7329 7330 7331
#endif /* CONFIG_FAIR_GROUP_SCHED */

		rq->rt.rt_runtime = def_rt_bandwidth.rt_runtime;
7332
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
7333
		INIT_LIST_HEAD(&rq->leaf_rt_rq_list);
7334
		init_tg_rt_entry(&root_task_group, &rq->rt, NULL, i, NULL);
I
Ingo Molnar 已提交
7335
#endif
L
Linus Torvalds 已提交
7336

I
Ingo Molnar 已提交
7337 7338
		for (j = 0; j < CPU_LOAD_IDX_MAX; j++)
			rq->cpu_load[j] = 0;
7339 7340 7341

		rq->last_load_update_tick = jiffies;

L
Linus Torvalds 已提交
7342
#ifdef CONFIG_SMP
N
Nick Piggin 已提交
7343
		rq->sd = NULL;
G
Gregory Haskins 已提交
7344
		rq->rd = NULL;
7345
		rq->cpu_power = SCHED_POWER_SCALE;
7346
		rq->post_schedule = 0;
L
Linus Torvalds 已提交
7347
		rq->active_balance = 0;
I
Ingo Molnar 已提交
7348
		rq->next_balance = jiffies;
L
Linus Torvalds 已提交
7349
		rq->push_cpu = 0;
7350
		rq->cpu = i;
7351
		rq->online = 0;
7352 7353
		rq->idle_stamp = 0;
		rq->avg_idle = 2*sysctl_sched_migration_cost;
7354 7355 7356

		INIT_LIST_HEAD(&rq->cfs_tasks);

7357
		rq_attach_root(rq, &def_root_domain);
7358
#ifdef CONFIG_NO_HZ
7359
		rq->nohz_flags = 0;
7360
#endif
L
Linus Torvalds 已提交
7361
#endif
P
Peter Zijlstra 已提交
7362
		init_rq_hrtick(rq);
L
Linus Torvalds 已提交
7363 7364 7365
		atomic_set(&rq->nr_iowait, 0);
	}

7366
	set_load_weight(&init_task);
7367

7368 7369 7370 7371
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&init_task.preempt_notifiers);
#endif

7372
#ifdef CONFIG_RT_MUTEXES
7373
	plist_head_init(&init_task.pi_waiters);
7374 7375
#endif

L
Linus Torvalds 已提交
7376 7377 7378 7379 7380 7381 7382 7383 7384 7385 7386 7387 7388
	/*
	 * 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());
7389 7390 7391

	calc_load_update = jiffies + LOAD_FREQ;

I
Ingo Molnar 已提交
7392 7393 7394 7395
	/*
	 * During early bootup we pretend to be a normal task:
	 */
	current->sched_class = &fair_sched_class;
7396

7397
#ifdef CONFIG_SMP
7398
	zalloc_cpumask_var(&sched_domains_tmpmask, GFP_NOWAIT);
R
Rusty Russell 已提交
7399 7400 7401
	/* May be allocated at isolcpus cmdline parse time */
	if (cpu_isolated_map == NULL)
		zalloc_cpumask_var(&cpu_isolated_map, GFP_NOWAIT);
7402
	idle_thread_set_boot_cpu();
7403 7404
#endif
	init_sched_fair_class();
7405

7406
	scheduler_running = 1;
L
Linus Torvalds 已提交
7407 7408
}

7409
#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
7410 7411
static inline int preempt_count_equals(int preempt_offset)
{
7412
	int nested = (preempt_count() & ~PREEMPT_ACTIVE) + rcu_preempt_depth();
7413

A
Arnd Bergmann 已提交
7414
	return (nested == preempt_offset);
7415 7416
}

7417
void __might_sleep(const char *file, int line, int preempt_offset)
L
Linus Torvalds 已提交
7418 7419 7420
{
	static unsigned long prev_jiffy;	/* ratelimiting */

7421
	rcu_sleep_check(); /* WARN_ON_ONCE() by default, no rate limit reqd. */
7422 7423
	if ((preempt_count_equals(preempt_offset) && !irqs_disabled()) ||
	    system_state != SYSTEM_RUNNING || oops_in_progress)
I
Ingo Molnar 已提交
7424 7425 7426 7427 7428
		return;
	if (time_before(jiffies, prev_jiffy + HZ) && prev_jiffy)
		return;
	prev_jiffy = jiffies;

P
Peter Zijlstra 已提交
7429 7430 7431 7432 7433 7434 7435
	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
7436 7437 7438 7439 7440

	debug_show_held_locks(current);
	if (irqs_disabled())
		print_irqtrace_events(current);
	dump_stack();
L
Linus Torvalds 已提交
7441 7442 7443 7444 7445
}
EXPORT_SYMBOL(__might_sleep);
#endif

#ifdef CONFIG_MAGIC_SYSRQ
7446 7447
static void normalize_task(struct rq *rq, struct task_struct *p)
{
P
Peter Zijlstra 已提交
7448 7449
	const struct sched_class *prev_class = p->sched_class;
	int old_prio = p->prio;
7450
	int on_rq;
7451

P
Peter Zijlstra 已提交
7452
	on_rq = p->on_rq;
7453
	if (on_rq)
7454
		dequeue_task(rq, p, 0);
7455 7456
	__setscheduler(rq, p, SCHED_NORMAL, 0);
	if (on_rq) {
7457
		enqueue_task(rq, p, 0);
7458 7459
		resched_task(rq->curr);
	}
P
Peter Zijlstra 已提交
7460 7461

	check_class_changed(rq, p, prev_class, old_prio);
7462 7463
}

L
Linus Torvalds 已提交
7464 7465
void normalize_rt_tasks(void)
{
7466
	struct task_struct *g, *p;
L
Linus Torvalds 已提交
7467
	unsigned long flags;
7468
	struct rq *rq;
L
Linus Torvalds 已提交
7469

7470
	read_lock_irqsave(&tasklist_lock, flags);
7471
	do_each_thread(g, p) {
7472 7473 7474 7475 7476 7477
		/*
		 * Only normalize user tasks:
		 */
		if (!p->mm)
			continue;

I
Ingo Molnar 已提交
7478 7479
		p->se.exec_start		= 0;
#ifdef CONFIG_SCHEDSTATS
7480 7481 7482
		p->se.statistics.wait_start	= 0;
		p->se.statistics.sleep_start	= 0;
		p->se.statistics.block_start	= 0;
I
Ingo Molnar 已提交
7483
#endif
I
Ingo Molnar 已提交
7484 7485 7486 7487 7488 7489 7490 7491

		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
7492
			continue;
I
Ingo Molnar 已提交
7493
		}
L
Linus Torvalds 已提交
7494

7495
		raw_spin_lock(&p->pi_lock);
7496
		rq = __task_rq_lock(p);
L
Linus Torvalds 已提交
7497

7498
		normalize_task(rq, p);
7499

7500
		__task_rq_unlock(rq);
7501
		raw_spin_unlock(&p->pi_lock);
7502 7503
	} while_each_thread(g, p);

7504
	read_unlock_irqrestore(&tasklist_lock, flags);
L
Linus Torvalds 已提交
7505 7506 7507
}

#endif /* CONFIG_MAGIC_SYSRQ */
7508

7509
#if defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB)
7510
/*
7511
 * These functions are only useful for the IA64 MCA handling, or kdb.
7512 7513 7514 7515 7516 7517 7518 7519 7520 7521 7522 7523 7524 7525
 *
 * 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!
 */
7526
struct task_struct *curr_task(int cpu)
7527 7528 7529 7530
{
	return cpu_curr(cpu);
}

7531 7532 7533
#endif /* defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB) */

#ifdef CONFIG_IA64
7534 7535 7536 7537 7538 7539
/**
 * 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
7540 7541
 * 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
7542 7543 7544 7545 7546 7547 7548
 * 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!
 */
7549
void set_curr_task(int cpu, struct task_struct *p)
7550 7551 7552 7553 7554
{
	cpu_curr(cpu) = p;
}

#endif
S
Srivatsa Vaddagiri 已提交
7555

D
Dhaval Giani 已提交
7556
#ifdef CONFIG_CGROUP_SCHED
7557 7558 7559
/* task_group_lock serializes the addition/removal of task groups */
static DEFINE_SPINLOCK(task_group_lock);

7560 7561 7562 7563
static void free_sched_group(struct task_group *tg)
{
	free_fair_sched_group(tg);
	free_rt_sched_group(tg);
7564
	autogroup_free(tg);
7565 7566 7567 7568
	kfree(tg);
}

/* allocate runqueue etc for a new task group */
7569
struct task_group *sched_create_group(struct task_group *parent)
7570 7571 7572 7573 7574 7575 7576 7577
{
	struct task_group *tg;
	unsigned long flags;

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

7578
	if (!alloc_fair_sched_group(tg, parent))
7579 7580
		goto err;

7581
	if (!alloc_rt_sched_group(tg, parent))
7582 7583
		goto err;

7584
	spin_lock_irqsave(&task_group_lock, flags);
P
Peter Zijlstra 已提交
7585
	list_add_rcu(&tg->list, &task_groups);
P
Peter Zijlstra 已提交
7586 7587 7588 7589 7590

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

	tg->parent = parent;
	INIT_LIST_HEAD(&tg->children);
7591
	list_add_rcu(&tg->siblings, &parent->children);
7592
	spin_unlock_irqrestore(&task_group_lock, flags);
S
Srivatsa Vaddagiri 已提交
7593

7594
	return tg;
S
Srivatsa Vaddagiri 已提交
7595 7596

err:
P
Peter Zijlstra 已提交
7597
	free_sched_group(tg);
S
Srivatsa Vaddagiri 已提交
7598 7599 7600
	return ERR_PTR(-ENOMEM);
}

7601
/* rcu callback to free various structures associated with a task group */
P
Peter Zijlstra 已提交
7602
static void free_sched_group_rcu(struct rcu_head *rhp)
S
Srivatsa Vaddagiri 已提交
7603 7604
{
	/* now it should be safe to free those cfs_rqs */
P
Peter Zijlstra 已提交
7605
	free_sched_group(container_of(rhp, struct task_group, rcu));
S
Srivatsa Vaddagiri 已提交
7606 7607
}

7608
/* Destroy runqueue etc associated with a task group */
7609
void sched_destroy_group(struct task_group *tg)
S
Srivatsa Vaddagiri 已提交
7610
{
7611
	unsigned long flags;
7612
	int i;
S
Srivatsa Vaddagiri 已提交
7613

7614 7615
	/* end participation in shares distribution */
	for_each_possible_cpu(i)
7616
		unregister_fair_sched_group(tg, i);
7617 7618

	spin_lock_irqsave(&task_group_lock, flags);
P
Peter Zijlstra 已提交
7619
	list_del_rcu(&tg->list);
P
Peter Zijlstra 已提交
7620
	list_del_rcu(&tg->siblings);
7621
	spin_unlock_irqrestore(&task_group_lock, flags);
7622 7623

	/* wait for possible concurrent references to cfs_rqs complete */
P
Peter Zijlstra 已提交
7624
	call_rcu(&tg->rcu, free_sched_group_rcu);
S
Srivatsa Vaddagiri 已提交
7625 7626
}

7627
/* change task's runqueue when it moves between groups.
I
Ingo Molnar 已提交
7628 7629 7630
 *	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.
7631 7632
 */
void sched_move_task(struct task_struct *tsk)
S
Srivatsa Vaddagiri 已提交
7633
{
P
Peter Zijlstra 已提交
7634
	struct task_group *tg;
S
Srivatsa Vaddagiri 已提交
7635 7636 7637 7638 7639 7640
	int on_rq, running;
	unsigned long flags;
	struct rq *rq;

	rq = task_rq_lock(tsk, &flags);

7641
	running = task_current(rq, tsk);
P
Peter Zijlstra 已提交
7642
	on_rq = tsk->on_rq;
S
Srivatsa Vaddagiri 已提交
7643

7644
	if (on_rq)
S
Srivatsa Vaddagiri 已提交
7645
		dequeue_task(rq, tsk, 0);
7646 7647
	if (unlikely(running))
		tsk->sched_class->put_prev_task(rq, tsk);
S
Srivatsa Vaddagiri 已提交
7648

P
Peter Zijlstra 已提交
7649 7650 7651 7652 7653 7654
	tg = container_of(task_subsys_state_check(tsk, cpu_cgroup_subsys_id,
				lockdep_is_held(&tsk->sighand->siglock)),
			  struct task_group, css);
	tg = autogroup_task_group(tsk, tg);
	tsk->sched_task_group = tg;

P
Peter Zijlstra 已提交
7655
#ifdef CONFIG_FAIR_GROUP_SCHED
7656 7657 7658
	if (tsk->sched_class->task_move_group)
		tsk->sched_class->task_move_group(tsk, on_rq);
	else
P
Peter Zijlstra 已提交
7659
#endif
7660
		set_task_rq(tsk, task_cpu(tsk));
P
Peter Zijlstra 已提交
7661

7662 7663 7664
	if (unlikely(running))
		tsk->sched_class->set_curr_task(rq);
	if (on_rq)
7665
		enqueue_task(rq, tsk, 0);
S
Srivatsa Vaddagiri 已提交
7666

7667
	task_rq_unlock(rq, tsk, &flags);
S
Srivatsa Vaddagiri 已提交
7668
}
D
Dhaval Giani 已提交
7669
#endif /* CONFIG_CGROUP_SCHED */
S
Srivatsa Vaddagiri 已提交
7670

7671
#if defined(CONFIG_RT_GROUP_SCHED) || defined(CONFIG_CFS_BANDWIDTH)
P
Peter Zijlstra 已提交
7672 7673 7674
static unsigned long to_ratio(u64 period, u64 runtime)
{
	if (runtime == RUNTIME_INF)
P
Peter Zijlstra 已提交
7675
		return 1ULL << 20;
P
Peter Zijlstra 已提交
7676

P
Peter Zijlstra 已提交
7677
	return div64_u64(runtime << 20, period);
P
Peter Zijlstra 已提交
7678
}
7679 7680 7681 7682 7683 7684 7685
#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
7686

P
Peter Zijlstra 已提交
7687 7688
/* Must be called with tasklist_lock held */
static inline int tg_has_rt_tasks(struct task_group *tg)
7689
{
P
Peter Zijlstra 已提交
7690
	struct task_struct *g, *p;
7691

P
Peter Zijlstra 已提交
7692
	do_each_thread(g, p) {
7693
		if (rt_task(p) && task_rq(p)->rt.tg == tg)
P
Peter Zijlstra 已提交
7694 7695
			return 1;
	} while_each_thread(g, p);
7696

P
Peter Zijlstra 已提交
7697 7698
	return 0;
}
7699

P
Peter Zijlstra 已提交
7700 7701 7702 7703 7704
struct rt_schedulable_data {
	struct task_group *tg;
	u64 rt_period;
	u64 rt_runtime;
};
7705

7706
static int tg_rt_schedulable(struct task_group *tg, void *data)
P
Peter Zijlstra 已提交
7707 7708 7709 7710 7711
{
	struct rt_schedulable_data *d = data;
	struct task_group *child;
	unsigned long total, sum = 0;
	u64 period, runtime;
7712

P
Peter Zijlstra 已提交
7713 7714
	period = ktime_to_ns(tg->rt_bandwidth.rt_period);
	runtime = tg->rt_bandwidth.rt_runtime;
7715

P
Peter Zijlstra 已提交
7716 7717 7718
	if (tg == d->tg) {
		period = d->rt_period;
		runtime = d->rt_runtime;
7719 7720
	}

7721 7722 7723 7724 7725
	/*
	 * Cannot have more runtime than the period.
	 */
	if (runtime > period && runtime != RUNTIME_INF)
		return -EINVAL;
P
Peter Zijlstra 已提交
7726

7727 7728 7729
	/*
	 * Ensure we don't starve existing RT tasks.
	 */
P
Peter Zijlstra 已提交
7730 7731
	if (rt_bandwidth_enabled() && !runtime && tg_has_rt_tasks(tg))
		return -EBUSY;
P
Peter Zijlstra 已提交
7732

P
Peter Zijlstra 已提交
7733
	total = to_ratio(period, runtime);
P
Peter Zijlstra 已提交
7734

7735 7736 7737 7738 7739
	/*
	 * 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
7740

7741 7742 7743
	/*
	 * The sum of our children's runtime should not exceed our own.
	 */
P
Peter Zijlstra 已提交
7744 7745 7746
	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
7747

P
Peter Zijlstra 已提交
7748 7749 7750 7751
		if (child == d->tg) {
			period = d->rt_period;
			runtime = d->rt_runtime;
		}
P
Peter Zijlstra 已提交
7752

P
Peter Zijlstra 已提交
7753
		sum += to_ratio(period, runtime);
P
Peter Zijlstra 已提交
7754
	}
P
Peter Zijlstra 已提交
7755

P
Peter Zijlstra 已提交
7756 7757 7758 7759
	if (sum > total)
		return -EINVAL;

	return 0;
P
Peter Zijlstra 已提交
7760 7761
}

P
Peter Zijlstra 已提交
7762
static int __rt_schedulable(struct task_group *tg, u64 period, u64 runtime)
7763
{
7764 7765
	int ret;

P
Peter Zijlstra 已提交
7766 7767 7768 7769 7770 7771
	struct rt_schedulable_data data = {
		.tg = tg,
		.rt_period = period,
		.rt_runtime = runtime,
	};

7772 7773 7774 7775 7776
	rcu_read_lock();
	ret = walk_tg_tree(tg_rt_schedulable, tg_nop, &data);
	rcu_read_unlock();

	return ret;
7777 7778
}

7779
static int tg_set_rt_bandwidth(struct task_group *tg,
7780
		u64 rt_period, u64 rt_runtime)
P
Peter Zijlstra 已提交
7781
{
P
Peter Zijlstra 已提交
7782
	int i, err = 0;
P
Peter Zijlstra 已提交
7783 7784

	mutex_lock(&rt_constraints_mutex);
7785
	read_lock(&tasklist_lock);
P
Peter Zijlstra 已提交
7786 7787
	err = __rt_schedulable(tg, rt_period, rt_runtime);
	if (err)
P
Peter Zijlstra 已提交
7788
		goto unlock;
P
Peter Zijlstra 已提交
7789

7790
	raw_spin_lock_irq(&tg->rt_bandwidth.rt_runtime_lock);
7791 7792
	tg->rt_bandwidth.rt_period = ns_to_ktime(rt_period);
	tg->rt_bandwidth.rt_runtime = rt_runtime;
P
Peter Zijlstra 已提交
7793 7794 7795 7796

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

7797
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
7798
		rt_rq->rt_runtime = rt_runtime;
7799
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
7800
	}
7801
	raw_spin_unlock_irq(&tg->rt_bandwidth.rt_runtime_lock);
P
Peter Zijlstra 已提交
7802
unlock:
7803
	read_unlock(&tasklist_lock);
P
Peter Zijlstra 已提交
7804 7805 7806
	mutex_unlock(&rt_constraints_mutex);

	return err;
P
Peter Zijlstra 已提交
7807 7808
}

7809 7810 7811 7812 7813 7814 7815 7816 7817
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;

7818
	return tg_set_rt_bandwidth(tg, rt_period, rt_runtime);
7819 7820
}

P
Peter Zijlstra 已提交
7821 7822 7823 7824
long sched_group_rt_runtime(struct task_group *tg)
{
	u64 rt_runtime_us;

7825
	if (tg->rt_bandwidth.rt_runtime == RUNTIME_INF)
P
Peter Zijlstra 已提交
7826 7827
		return -1;

7828
	rt_runtime_us = tg->rt_bandwidth.rt_runtime;
P
Peter Zijlstra 已提交
7829 7830 7831
	do_div(rt_runtime_us, NSEC_PER_USEC);
	return rt_runtime_us;
}
7832 7833 7834 7835 7836 7837 7838 7839

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;

7840 7841 7842
	if (rt_period == 0)
		return -EINVAL;

7843
	return tg_set_rt_bandwidth(tg, rt_period, rt_runtime);
7844 7845 7846 7847 7848 7849 7850 7851 7852 7853 7854 7855 7856
}

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)
{
7857
	u64 runtime, period;
7858 7859
	int ret = 0;

7860 7861 7862
	if (sysctl_sched_rt_period <= 0)
		return -EINVAL;

7863 7864 7865 7866 7867 7868 7869 7870
	runtime = global_rt_runtime();
	period = global_rt_period();

	/*
	 * Sanity check on the sysctl variables.
	 */
	if (runtime > period && runtime != RUNTIME_INF)
		return -EINVAL;
7871

7872
	mutex_lock(&rt_constraints_mutex);
P
Peter Zijlstra 已提交
7873
	read_lock(&tasklist_lock);
7874
	ret = __rt_schedulable(NULL, 0, 0);
P
Peter Zijlstra 已提交
7875
	read_unlock(&tasklist_lock);
7876 7877 7878 7879
	mutex_unlock(&rt_constraints_mutex);

	return ret;
}
7880 7881 7882 7883 7884 7885 7886 7887 7888 7889

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;
}

7890
#else /* !CONFIG_RT_GROUP_SCHED */
7891 7892
static int sched_rt_global_constraints(void)
{
P
Peter Zijlstra 已提交
7893 7894 7895
	unsigned long flags;
	int i;

7896 7897 7898
	if (sysctl_sched_rt_period <= 0)
		return -EINVAL;

7899 7900 7901 7902 7903 7904 7905
	/*
	 * There's always some RT tasks in the root group
	 * -- migration, kstopmachine etc..
	 */
	if (sysctl_sched_rt_runtime == 0)
		return -EBUSY;

7906
	raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags);
P
Peter Zijlstra 已提交
7907 7908 7909
	for_each_possible_cpu(i) {
		struct rt_rq *rt_rq = &cpu_rq(i)->rt;

7910
		raw_spin_lock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
7911
		rt_rq->rt_runtime = global_rt_runtime();
7912
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
P
Peter Zijlstra 已提交
7913
	}
7914
	raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags);
P
Peter Zijlstra 已提交
7915

7916 7917
	return 0;
}
7918
#endif /* CONFIG_RT_GROUP_SCHED */
7919 7920

int sched_rt_handler(struct ctl_table *table, int write,
7921
		void __user *buffer, size_t *lenp,
7922 7923 7924 7925 7926 7927 7928 7929 7930 7931
		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;

7932
	ret = proc_dointvec(table, write, buffer, lenp, ppos);
7933 7934 7935 7936 7937 7938 7939 7940 7941 7942 7943 7944 7945 7946 7947 7948

	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;
}
7949

7950
#ifdef CONFIG_CGROUP_SCHED
7951 7952

/* return corresponding task_group object of a cgroup */
7953
static inline struct task_group *cgroup_tg(struct cgroup *cgrp)
7954
{
7955 7956
	return container_of(cgroup_subsys_state(cgrp, cpu_cgroup_subsys_id),
			    struct task_group, css);
7957 7958
}

7959
static struct cgroup_subsys_state *cpu_cgroup_create(struct cgroup *cgrp)
7960
{
7961
	struct task_group *tg, *parent;
7962

7963
	if (!cgrp->parent) {
7964
		/* This is early initialization for the top cgroup */
7965
		return &root_task_group.css;
7966 7967
	}

7968 7969
	parent = cgroup_tg(cgrp->parent);
	tg = sched_create_group(parent);
7970 7971 7972 7973 7974 7975
	if (IS_ERR(tg))
		return ERR_PTR(-ENOMEM);

	return &tg->css;
}

7976
static void cpu_cgroup_destroy(struct cgroup *cgrp)
7977
{
7978
	struct task_group *tg = cgroup_tg(cgrp);
7979 7980 7981 7982

	sched_destroy_group(tg);
}

7983
static int cpu_cgroup_can_attach(struct cgroup *cgrp,
7984
				 struct cgroup_taskset *tset)
7985
{
7986 7987 7988
	struct task_struct *task;

	cgroup_taskset_for_each(task, cgrp, tset) {
7989
#ifdef CONFIG_RT_GROUP_SCHED
7990 7991
		if (!sched_rt_can_attach(cgroup_tg(cgrp), task))
			return -EINVAL;
7992
#else
7993 7994 7995
		/* We don't support RT-tasks being in separate groups */
		if (task->sched_class != &fair_sched_class)
			return -EINVAL;
7996
#endif
7997
	}
7998 7999
	return 0;
}
8000

8001
static void cpu_cgroup_attach(struct cgroup *cgrp,
8002
			      struct cgroup_taskset *tset)
8003
{
8004 8005 8006 8007
	struct task_struct *task;

	cgroup_taskset_for_each(task, cgrp, tset)
		sched_move_task(task);
8008 8009
}

8010
static void
8011 8012
cpu_cgroup_exit(struct cgroup *cgrp, struct cgroup *old_cgrp,
		struct task_struct *task)
8013 8014 8015 8016 8017 8018 8019 8020 8021 8022 8023 8024
{
	/*
	 * 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);
}

8025
#ifdef CONFIG_FAIR_GROUP_SCHED
8026
static int cpu_shares_write_u64(struct cgroup *cgrp, struct cftype *cftype,
8027
				u64 shareval)
8028
{
8029
	return sched_group_set_shares(cgroup_tg(cgrp), scale_load(shareval));
8030 8031
}

8032
static u64 cpu_shares_read_u64(struct cgroup *cgrp, struct cftype *cft)
8033
{
8034
	struct task_group *tg = cgroup_tg(cgrp);
8035

8036
	return (u64) scale_load_down(tg->shares);
8037
}
8038 8039

#ifdef CONFIG_CFS_BANDWIDTH
8040 8041
static DEFINE_MUTEX(cfs_constraints_mutex);

8042 8043 8044
const u64 max_cfs_quota_period = 1 * NSEC_PER_SEC; /* 1s */
const u64 min_cfs_quota_period = 1 * NSEC_PER_MSEC; /* 1ms */

8045 8046
static int __cfs_schedulable(struct task_group *tg, u64 period, u64 runtime);

8047 8048
static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota)
{
8049
	int i, ret = 0, runtime_enabled, runtime_was_enabled;
8050
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
8051 8052 8053 8054 8055 8056 8057 8058 8059 8060 8061 8062 8063 8064 8065 8066 8067 8068 8069 8070

	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;

8071 8072 8073 8074 8075
	mutex_lock(&cfs_constraints_mutex);
	ret = __cfs_schedulable(tg, period, quota);
	if (ret)
		goto out_unlock;

8076
	runtime_enabled = quota != RUNTIME_INF;
8077 8078
	runtime_was_enabled = cfs_b->quota != RUNTIME_INF;
	account_cfs_bandwidth_used(runtime_enabled, runtime_was_enabled);
8079 8080 8081
	raw_spin_lock_irq(&cfs_b->lock);
	cfs_b->period = ns_to_ktime(period);
	cfs_b->quota = quota;
8082

P
Paul Turner 已提交
8083
	__refill_cfs_bandwidth_runtime(cfs_b);
8084 8085 8086 8087 8088 8089
	/* 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);
	}
8090 8091 8092 8093
	raw_spin_unlock_irq(&cfs_b->lock);

	for_each_possible_cpu(i) {
		struct cfs_rq *cfs_rq = tg->cfs_rq[i];
8094
		struct rq *rq = cfs_rq->rq;
8095 8096

		raw_spin_lock_irq(&rq->lock);
8097
		cfs_rq->runtime_enabled = runtime_enabled;
8098
		cfs_rq->runtime_remaining = 0;
8099

8100
		if (cfs_rq->throttled)
8101
			unthrottle_cfs_rq(cfs_rq);
8102 8103
		raw_spin_unlock_irq(&rq->lock);
	}
8104 8105
out_unlock:
	mutex_unlock(&cfs_constraints_mutex);
8106

8107
	return ret;
8108 8109 8110 8111 8112 8113
}

int tg_set_cfs_quota(struct task_group *tg, long cfs_quota_us)
{
	u64 quota, period;

8114
	period = ktime_to_ns(tg->cfs_bandwidth.period);
8115 8116 8117 8118 8119 8120 8121 8122 8123 8124 8125 8126
	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;

8127
	if (tg->cfs_bandwidth.quota == RUNTIME_INF)
8128 8129
		return -1;

8130
	quota_us = tg->cfs_bandwidth.quota;
8131 8132 8133 8134 8135 8136 8137 8138 8139 8140
	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;
8141
	quota = tg->cfs_bandwidth.quota;
8142 8143 8144 8145 8146 8147 8148 8149

	return tg_set_cfs_bandwidth(tg, period, quota);
}

long tg_get_cfs_period(struct task_group *tg)
{
	u64 cfs_period_us;

8150
	cfs_period_us = ktime_to_ns(tg->cfs_bandwidth.period);
8151 8152 8153 8154 8155 8156 8157 8158 8159 8160 8161 8162 8163 8164 8165 8166 8167 8168 8169 8170 8171 8172 8173 8174 8175 8176 8177
	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);
}

8178 8179 8180 8181 8182 8183 8184 8185 8186 8187 8188 8189 8190 8191 8192 8193 8194 8195 8196 8197 8198 8199 8200 8201 8202 8203 8204 8205 8206 8207 8208 8209
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;
8210
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
8211 8212 8213 8214 8215
	s64 quota = 0, parent_quota = -1;

	if (!tg->parent) {
		quota = RUNTIME_INF;
	} else {
8216
		struct cfs_bandwidth *parent_b = &tg->parent->cfs_bandwidth;
8217 8218 8219 8220 8221 8222 8223 8224 8225 8226 8227 8228 8229 8230 8231 8232 8233 8234 8235 8236

		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)
{
8237
	int ret;
8238 8239 8240 8241 8242 8243 8244 8245 8246 8247 8248
	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);
	}

8249 8250 8251 8252 8253
	rcu_read_lock();
	ret = walk_tg_tree(tg_cfs_schedulable_down, tg_nop, &data);
	rcu_read_unlock();

	return ret;
8254
}
8255 8256 8257 8258 8259

static int cpu_stats_show(struct cgroup *cgrp, struct cftype *cft,
		struct cgroup_map_cb *cb)
{
	struct task_group *tg = cgroup_tg(cgrp);
8260
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
8261 8262 8263 8264 8265 8266 8267

	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;
}
8268
#endif /* CONFIG_CFS_BANDWIDTH */
8269
#endif /* CONFIG_FAIR_GROUP_SCHED */
8270

8271
#ifdef CONFIG_RT_GROUP_SCHED
M
Mirco Tischler 已提交
8272
static int cpu_rt_runtime_write(struct cgroup *cgrp, struct cftype *cft,
8273
				s64 val)
P
Peter Zijlstra 已提交
8274
{
8275
	return sched_group_set_rt_runtime(cgroup_tg(cgrp), val);
P
Peter Zijlstra 已提交
8276 8277
}

8278
static s64 cpu_rt_runtime_read(struct cgroup *cgrp, struct cftype *cft)
P
Peter Zijlstra 已提交
8279
{
8280
	return sched_group_rt_runtime(cgroup_tg(cgrp));
P
Peter Zijlstra 已提交
8281
}
8282 8283 8284 8285 8286 8287 8288 8289 8290 8291 8292

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));
}
8293
#endif /* CONFIG_RT_GROUP_SCHED */
P
Peter Zijlstra 已提交
8294

8295
static struct cftype cpu_files[] = {
8296
#ifdef CONFIG_FAIR_GROUP_SCHED
8297 8298
	{
		.name = "shares",
8299 8300
		.read_u64 = cpu_shares_read_u64,
		.write_u64 = cpu_shares_write_u64,
8301
	},
8302
#endif
8303 8304 8305 8306 8307 8308 8309 8310 8311 8312 8313
#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,
	},
8314 8315 8316 8317
	{
		.name = "stat",
		.read_map = cpu_stats_show,
	},
8318
#endif
8319
#ifdef CONFIG_RT_GROUP_SCHED
P
Peter Zijlstra 已提交
8320
	{
P
Peter Zijlstra 已提交
8321
		.name = "rt_runtime_us",
8322 8323
		.read_s64 = cpu_rt_runtime_read,
		.write_s64 = cpu_rt_runtime_write,
P
Peter Zijlstra 已提交
8324
	},
8325 8326
	{
		.name = "rt_period_us",
8327 8328
		.read_u64 = cpu_rt_period_read_uint,
		.write_u64 = cpu_rt_period_write_uint,
8329
	},
8330
#endif
8331
	{ }	/* terminate */
8332 8333 8334
};

struct cgroup_subsys cpu_cgroup_subsys = {
I
Ingo Molnar 已提交
8335 8336 8337
	.name		= "cpu",
	.create		= cpu_cgroup_create,
	.destroy	= cpu_cgroup_destroy,
8338 8339
	.can_attach	= cpu_cgroup_can_attach,
	.attach		= cpu_cgroup_attach,
8340
	.exit		= cpu_cgroup_exit,
I
Ingo Molnar 已提交
8341
	.subsys_id	= cpu_cgroup_subsys_id,
8342
	.base_cftypes	= cpu_files,
8343 8344 8345
	.early_init	= 1,
};

8346
#endif	/* CONFIG_CGROUP_SCHED */
8347 8348 8349 8350 8351 8352 8353 8354 8355 8356 8357

#ifdef CONFIG_CGROUP_CPUACCT

/*
 * CPU accounting code for task groups.
 *
 * Based on the work by Paul Menage (menage@google.com) and Balbir Singh
 * (balbir@in.ibm.com).
 */

/* create a new cpu accounting group */
8358
static struct cgroup_subsys_state *cpuacct_create(struct cgroup *cgrp)
8359
{
8360
	struct cpuacct *ca;
8361

8362 8363 8364 8365
	if (!cgrp->parent)
		return &root_cpuacct.css;

	ca = kzalloc(sizeof(*ca), GFP_KERNEL);
8366
	if (!ca)
8367
		goto out;
8368 8369

	ca->cpuusage = alloc_percpu(u64);
8370 8371 8372
	if (!ca->cpuusage)
		goto out_free_ca;

8373 8374 8375
	ca->cpustat = alloc_percpu(struct kernel_cpustat);
	if (!ca->cpustat)
		goto out_free_cpuusage;
8376

8377
	return &ca->css;
8378

8379
out_free_cpuusage:
8380 8381 8382 8383 8384
	free_percpu(ca->cpuusage);
out_free_ca:
	kfree(ca);
out:
	return ERR_PTR(-ENOMEM);
8385 8386 8387
}

/* destroy an existing cpu accounting group */
8388
static void cpuacct_destroy(struct cgroup *cgrp)
8389
{
8390
	struct cpuacct *ca = cgroup_ca(cgrp);
8391

8392
	free_percpu(ca->cpustat);
8393 8394 8395 8396
	free_percpu(ca->cpuusage);
	kfree(ca);
}

8397 8398
static u64 cpuacct_cpuusage_read(struct cpuacct *ca, int cpu)
{
8399
	u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
8400 8401 8402 8403 8404 8405
	u64 data;

#ifndef CONFIG_64BIT
	/*
	 * Take rq->lock to make 64-bit read safe on 32-bit platforms.
	 */
8406
	raw_spin_lock_irq(&cpu_rq(cpu)->lock);
8407
	data = *cpuusage;
8408
	raw_spin_unlock_irq(&cpu_rq(cpu)->lock);
8409 8410 8411 8412 8413 8414 8415 8416 8417
#else
	data = *cpuusage;
#endif

	return data;
}

static void cpuacct_cpuusage_write(struct cpuacct *ca, int cpu, u64 val)
{
8418
	u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
8419 8420 8421 8422 8423

#ifndef CONFIG_64BIT
	/*
	 * Take rq->lock to make 64-bit write safe on 32-bit platforms.
	 */
8424
	raw_spin_lock_irq(&cpu_rq(cpu)->lock);
8425
	*cpuusage = val;
8426
	raw_spin_unlock_irq(&cpu_rq(cpu)->lock);
8427 8428 8429 8430 8431
#else
	*cpuusage = val;
#endif
}

8432
/* return total cpu usage (in nanoseconds) of a group */
8433
static u64 cpuusage_read(struct cgroup *cgrp, struct cftype *cft)
8434
{
8435
	struct cpuacct *ca = cgroup_ca(cgrp);
8436 8437 8438
	u64 totalcpuusage = 0;
	int i;

8439 8440
	for_each_present_cpu(i)
		totalcpuusage += cpuacct_cpuusage_read(ca, i);
8441 8442 8443 8444

	return totalcpuusage;
}

8445 8446 8447 8448 8449 8450 8451 8452 8453 8454 8455 8456
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;
	}

8457 8458
	for_each_present_cpu(i)
		cpuacct_cpuusage_write(ca, i, 0);
8459 8460 8461 8462 8463

out:
	return err;
}

8464 8465 8466 8467 8468 8469 8470 8471 8472 8473 8474 8475 8476 8477 8478
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;
}

8479 8480 8481 8482 8483 8484
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,
8485
			      struct cgroup_map_cb *cb)
8486 8487
{
	struct cpuacct *ca = cgroup_ca(cgrp);
8488 8489
	int cpu;
	s64 val = 0;
8490

8491 8492 8493 8494
	for_each_online_cpu(cpu) {
		struct kernel_cpustat *kcpustat = per_cpu_ptr(ca->cpustat, cpu);
		val += kcpustat->cpustat[CPUTIME_USER];
		val += kcpustat->cpustat[CPUTIME_NICE];
8495
	}
8496 8497
	val = cputime64_to_clock_t(val);
	cb->fill(cb, cpuacct_stat_desc[CPUACCT_STAT_USER], val);
8498

8499 8500 8501 8502 8503 8504
	val = 0;
	for_each_online_cpu(cpu) {
		struct kernel_cpustat *kcpustat = per_cpu_ptr(ca->cpustat, cpu);
		val += kcpustat->cpustat[CPUTIME_SYSTEM];
		val += kcpustat->cpustat[CPUTIME_IRQ];
		val += kcpustat->cpustat[CPUTIME_SOFTIRQ];
8505
	}
8506 8507 8508 8509

	val = cputime64_to_clock_t(val);
	cb->fill(cb, cpuacct_stat_desc[CPUACCT_STAT_SYSTEM], val);

8510 8511 8512
	return 0;
}

8513 8514 8515
static struct cftype files[] = {
	{
		.name = "usage",
8516 8517
		.read_u64 = cpuusage_read,
		.write_u64 = cpuusage_write,
8518
	},
8519 8520 8521 8522
	{
		.name = "usage_percpu",
		.read_seq_string = cpuacct_percpu_seq_read,
	},
8523 8524 8525 8526
	{
		.name = "stat",
		.read_map = cpuacct_stats_show,
	},
8527
	{ }	/* terminate */
8528 8529 8530 8531 8532 8533 8534
};

/*
 * charge this task's execution time to its accounting group.
 *
 * called with rq->lock held.
 */
8535
void cpuacct_charge(struct task_struct *tsk, u64 cputime)
8536 8537
{
	struct cpuacct *ca;
8538
	int cpu;
8539

L
Li Zefan 已提交
8540
	if (unlikely(!cpuacct_subsys.active))
8541 8542
		return;

8543
	cpu = task_cpu(tsk);
8544 8545 8546

	rcu_read_lock();

8547 8548
	ca = task_ca(tsk);

8549
	for (; ca; ca = parent_ca(ca)) {
8550
		u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
8551 8552
		*cpuusage += cputime;
	}
8553 8554

	rcu_read_unlock();
8555 8556 8557 8558 8559 8560 8561
}

struct cgroup_subsys cpuacct_subsys = {
	.name = "cpuacct",
	.create = cpuacct_create,
	.destroy = cpuacct_destroy,
	.subsys_id = cpuacct_subsys_id,
8562
	.base_cftypes = files,
8563 8564
};
#endif	/* CONFIG_CGROUP_CPUACCT */