core.c 195 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1
/*
2
 *  kernel/sched/core.c
Linus Torvalds's avatar
Linus Torvalds committed
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
Ingo Molnar's avatar
Ingo Molnar committed
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
Linus Torvalds's avatar
Linus Torvalds committed
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>
Linus Torvalds's avatar
Linus Torvalds committed
34 35 36
#include <linux/highmem.h>
#include <asm/mmu_context.h>
#include <linux/interrupt.h>
37
#include <linux/capability.h>
Linus Torvalds's avatar
Linus Torvalds committed
38 39
#include <linux/completion.h>
#include <linux/kernel_stat.h>
40
#include <linux/debug_locks.h>
41
#include <linux/perf_event.h>
Linus Torvalds's avatar
Linus Torvalds committed
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>
Linus Torvalds's avatar
Linus Torvalds committed
47 48
#include <linux/blkdev.h>
#include <linux/delay.h>
49
#include <linux/pid_namespace.h>
Linus Torvalds's avatar
Linus Torvalds committed
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>
Linus Torvalds's avatar
Linus Torvalds committed
58
#include <linux/seq_file.h>
59
#include <linux/sysctl.h>
Linus Torvalds's avatar
Linus Torvalds committed
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>
Jens Axboe's avatar
Jens Axboe committed
66
#include <linux/pagemap.h>
67
#include <linux/hrtimer.h>
68
#include <linux/tick.h>
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>
Al Viro's avatar
Al Viro committed
74
#include <linux/binfmts.h>
Linus Torvalds's avatar
Linus Torvalds committed
75

76
#include <asm/switch_to.h>
77
#include <asm/tlb.h>
78
#include <asm/irq_regs.h>
79
#include <asm/mutex.h>
80 81 82
#ifdef CONFIG_PARAVIRT
#include <asm/paravirt.h>
#endif
Linus Torvalds's avatar
Linus Torvalds committed
83

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

87
#define CREATE_TRACE_POINTS
88
#include <trace/events/sched.h>
89

90
void start_bandwidth_timer(struct hrtimer *period_timer, ktime_t period)
91
{
92 93
	unsigned long delta;
	ktime_t soft, hard, now;
94

95 96 97 98 99 100
	for (;;) {
		if (hrtimer_active(period_timer))
			break;

		now = hrtimer_cb_get_time(period_timer);
		hrtimer_forward(period_timer, now, period);
101

102 103 104 105 106 107 108 109
		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);
	}
}

110 111
DEFINE_MUTEX(sched_domains_mutex);
DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
112

113
static void update_rq_clock_task(struct rq *rq, s64 delta);
114

115
void update_rq_clock(struct rq *rq)
116
{
117
	s64 delta;
118

119
	if (rq->skip_clock_update > 0)
120
		return;
121

122 123 124
	delta = sched_clock_cpu(cpu_of(rq)) - rq->clock;
	rq->clock += delta;
	update_rq_clock_task(rq, delta);
125 126
}

127 128 129
/*
 * Debugging: various feature bits
 */
130 131 132 133

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

134
const_debug unsigned int sysctl_sched_features =
135
#include "features.h"
136 137 138 139 140 141 142 143
	0;

#undef SCHED_FEAT

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

144
static __read_mostly char *sched_feat_names[] = {
145
#include "features.h"
146 147 148 149 150
	NULL
};

#undef SCHED_FEAT

151
static int sched_feat_show(struct seq_file *m, void *v)
152 153 154
{
	int i;

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

162
	return 0;
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 */

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;
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);
212

Hillf Danton's avatar
Hillf Danton committed
213
	if (strncmp(cmp, "NO_", 3) == 0) {
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) {
221
				sysctl_sched_features &= ~(1UL << i);
222 223
				sched_feat_disable(i);
			} else {
224
				sysctl_sched_features |= (1UL << i);
225 226
				sched_feat_enable(i);
			}
227 228 229 230
			break;
		}
	}

231
	if (i == __SCHED_FEAT_NR)
232 233
		return -EINVAL;

234
	*ppos += cnt;
235 236 237 238

	return cnt;
}

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 = {
245 246 247 248 249
	.open		= sched_feat_open,
	.write		= sched_feat_write,
	.read		= seq_read,
	.llseek		= seq_lseek,
	.release	= single_release,
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 */
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;

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

282
__read_mostly int scheduler_running;
283

Peter Zijlstra's avatar
Peter Zijlstra committed
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;
Peter Zijlstra's avatar
Peter Zijlstra committed
289 290


Linus Torvalds's avatar
Linus Torvalds committed
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);
305
		if (likely(rq == task_rq(p)))
306
			return rq;
307
		raw_spin_unlock(&rq->lock);
308 309 310
	}
}

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

320
	for (;;) {
321
		raw_spin_lock_irqsave(&p->pi_lock, *flags);
322
		rq = task_rq(p);
323
		raw_spin_lock(&rq->lock);
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);
Linus Torvalds's avatar
Linus Torvalds committed
328 329 330
	}
}

Alexey Dobriyan's avatar
Alexey Dobriyan committed
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)
Linus Torvalds's avatar
Linus Torvalds committed
339
	__releases(rq->lock)
340
	__releases(p->pi_lock)
Linus Torvalds's avatar
Linus Torvalds committed
341
{
342 343
	raw_spin_unlock(&rq->lock);
	raw_spin_unlock_irqrestore(&p->pi_lock, *flags);
Linus Torvalds's avatar
Linus Torvalds committed
344 345 346
}

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

	local_irq_disable();
	rq = this_rq();
356
	raw_spin_lock(&rq->lock);
Linus Torvalds's avatar
Linus Torvalds committed
357 358 359 360

	return rq;
}

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);
391
	rq->curr->sched_class->task_tick(rq, rq->curr, 1);
392
	raw_spin_unlock(&rq->lock);
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

Andrew Morton's avatar
Andrew Morton committed
466
static inline void init_hrtick(void)
467 468
{
}
469
#endif /* CONFIG_SMP */
470

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

476 477 478 479
	rq->hrtick_csd.flags = 0;
	rq->hrtick_csd.func = __hrtick_start;
	rq->hrtick_csd.info = rq;
#endif
480

481 482
	hrtimer_init(&rq->hrtick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	rq->hrtick_timer.function = hrtick;
483
}
Andrew Morton's avatar
Andrew Morton committed
484
#else	/* CONFIG_SCHED_HRTICK */
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)
{
}
Andrew Morton's avatar
Andrew Morton committed
496
#endif	/* CONFIG_SCHED_HRTICK */
497

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)
512 513 514
{
	int cpu;

515
	assert_raw_spin_locked(&task_rq(p)->lock);
516

517
	if (test_tsk_need_resched(p))
518 519
		return;

520
	set_tsk_need_resched(p);
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)
533 534 535 536
{
	struct rq *rq = cpu_rq(cpu);
	unsigned long flags;

537
	if (!raw_spin_trylock_irqsave(&rq->lock, flags))
538 539
		return;
	resched_task(cpu_curr(cpu));
540
	raw_spin_unlock_irqrestore(&rq->lock, flags);
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)
620
{
621
	return false;
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;
662
	int ret;
663

664 665
	parent = from;

666
down:
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;
	}
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:
686
	return ret;
687 688
}

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

695
void update_cpu_load(struct rq *this_rq);
696

697 698
static void set_load_weight(struct task_struct *p)
{
699 700 701
	int prio = p->static_prio - MAX_RT_PRIO;
	struct load_weight *load = &p->se.load;

Ingo Molnar's avatar
Ingo Molnar committed
702 703 704 705
	/*
	 * SCHED_IDLE tasks get minimal weight:
	 */
	if (p->policy == SCHED_IDLE) {
706
		load->weight = scale_load(WEIGHT_IDLEPRIO);
707
		load->inv_weight = WMULT_IDLEPRIO;
Ingo Molnar's avatar
Ingo Molnar committed
708 709
		return;
	}
710

711
	load->weight = scale_load(prio_to_weight[prio]);
712
	load->inv_weight = prio_to_wmult[prio];
713 714
}

715
static void enqueue_task(struct rq *rq, struct task_struct *p, int flags)
716
{
717
	update_rq_clock(rq);
Ingo Molnar's avatar
Ingo Molnar committed
718
	sched_info_queued(p);
719
	p->sched_class->enqueue_task(rq, p, flags);
720 721
}

722
static void dequeue_task(struct rq *rq, struct task_struct *p, int flags)
723
{
724
	update_rq_clock(rq);
725
	sched_info_dequeued(p);
726
	p->sched_class->dequeue_task(rq, p, flags);
727 728
}

729
void activate_task(struct rq *rq, struct task_struct *p, int flags)
730 731 732 733
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible--;

734
	enqueue_task(rq, p, flags);
735 736
}

737
void deactivate_task(struct rq *rq, struct task_struct *p, int flags)
738 739 740 741
{
	if (task_contributes_to_load(p))
		rq->nr_uninterruptible++;

742
	dequeue_task(rq, p, flags);
743 744
}

745 746
#ifdef CONFIG_IRQ_TIME_ACCOUNTING

747 748 749 750 751 752 753
/*
 * 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
754 755 756
 * 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.
757
 */
758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773
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;
}

774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811
#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)
812 813 814
{
	return per_cpu(cpu_softirq_time, cpu) + per_cpu(cpu_hardirq_time, cpu);
}
815
#endif /* CONFIG_64BIT */
816

817 818 819 820
/*
 * Called before incrementing preempt_count on {soft,}irq_enter
 * and before decrementing preempt_count on {soft,}irq_exit.
 */
821 822 823
void account_system_vtime(struct task_struct *curr)
{
	unsigned long flags;
824
	s64 delta;
825 826 827 828 829 830 831 832
	int cpu;

	if (!sched_clock_irqtime)
		return;

	local_irq_save(flags);

	cpu = smp_processor_id();
833 834 835
	delta = sched_clock_cpu(cpu) - __this_cpu_read(irq_start_time);
	__this_cpu_add(irq_start_time, delta);

836
	irq_time_write_begin();
837 838 839 840 841 842 843
	/*
	 * 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())
844
		__this_cpu_add(cpu_hardirq_time, delta);
845
	else if (in_serving_softirq() && curr != this_cpu_ksoftirqd())
846
		__this_cpu_add(cpu_softirq_time, delta);
847

848
	irq_time_write_end();
849 850
	local_irq_restore(flags);
}
851
EXPORT_SYMBOL_GPL(account_system_vtime);
852

853 854 855 856
#endif /* CONFIG_IRQ_TIME_ACCOUNTING */

#ifdef CONFIG_PARAVIRT
static inline u64 steal_ticks(u64 steal)
857
{
858 859
	if (unlikely(steal > NSEC_PER_SEC))
		return div_u64(steal, TICK_NSEC);
860

861 862 863 864
	return __iter_div_u64_rem(steal, TICK_NSEC, &steal);
}
#endif

865
static void update_rq_clock_task(struct rq *rq, s64 delta)
866
{
867 868 869 870 871 872 873 874
/*
 * 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
875
	irq_delta = irq_time_read(cpu_of(rq)) - rq->prev_irq_time;
876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896

	/*
	 * 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;
897 898
#endif
#ifdef CONFIG_PARAVIRT_TIME_ACCOUNTING
899
	if (static_key_false((&paravirt_steal_rq_enabled))) {
900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916
		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

917 918
	rq->clock_task += delta;

919 920 921 922
#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
923 924
}

925
#ifdef CONFIG_IRQ_TIME_ACCOUNTING
926 927
static int irqtime_account_hi_update(void)
{
928
	u64 *cpustat = kcpustat_this_cpu->cpustat;
929 930 931 932 933 934
	unsigned long flags;
	u64 latest_ns;
	int ret = 0;

	local_irq_save(flags);
	latest_ns = this_cpu_read(cpu_hardirq_time);
935
	if (nsecs_to_cputime64(latest_ns) > cpustat[CPUTIME_IRQ])
936 937 938 939 940 941 942
		ret = 1;
	local_irq_restore(flags);
	return ret;
}

static int irqtime_account_si_update(void)
{
943
	u64 *cpustat = kcpustat_this_cpu->cpustat;
944 945 946 947 948 949
	unsigned long flags;
	u64 latest_ns;
	int ret = 0;

	local_irq_save(flags);
	latest_ns = this_cpu_read(cpu_softirq_time);
950
	if (nsecs_to_cputime64(latest_ns) > cpustat[CPUTIME_SOFTIRQ])
951 952 953 954 955
		ret = 1;
	local_irq_restore(flags);
	return ret;
}

956
#else /* CONFIG_IRQ_TIME_ACCOUNTING */
957

958 959
#define sched_clock_irqtime	(0)

960
#endif
961

962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991
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;
	}
}

992
/*
Ingo Molnar's avatar
Ingo Molnar committed
993
 * __normal_prio - return the priority that is based on the static prio
994 995 996
 */
static inline int __normal_prio(struct task_struct *p)
{
Ingo Molnar's avatar
Ingo Molnar committed
997
	return p->static_prio;
998 999
}

1000 1001 1002 1003 1004 1005 1006
/*
 * 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.
 */
1007
static inline int normal_prio(struct task_struct *p)
1008 1009 1010
{
	int prio;

1011
	if (task_has_rt_policy(p))
1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024
		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.
 */
1025
static int effective_prio(struct task_struct *p)
1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037
{
	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;
}

Linus Torvalds's avatar
Linus Torvalds committed
1038 1039 1040 1041
/**
 * task_curr - is this task currently executing on a CPU?
 * @p: the task in question.
 */
1042
inline int task_curr(const struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
1043 1044 1045 1046
{
	return cpu_curr(task_cpu(p)) == p;
}

1047 1048
static inline void check_class_changed(struct rq *rq, struct task_struct *p,
				       const struct sched_class *prev_class,
1049
				       int oldprio)
1050 1051 1052
{
	if (prev_class != p->sched_class) {
		if (prev_class->switched_from)
1053 1054 1055 1056
			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);
1057 1058
}

1059
void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags)
1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079
{
	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.
	 */
Peter Zijlstra's avatar
Peter Zijlstra committed
1080
	if (rq->curr->on_rq && test_tsk_need_resched(rq->curr))
1081 1082 1083
		rq->skip_clock_update = 1;
}

Linus Torvalds's avatar
Linus Torvalds committed
1084
#ifdef CONFIG_SMP
Ingo Molnar's avatar
Ingo Molnar committed
1085
void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
Ingo Molnar's avatar
Ingo Molnar committed
1086
{
1087 1088 1089 1090 1091
#ifdef CONFIG_SCHED_DEBUG
	/*
	 * We should never call set_task_cpu() on a blocked task,
	 * ttwu() will sort out the placement.
	 */
Peter Zijlstra's avatar
Peter Zijlstra committed
1092 1093
	WARN_ON_ONCE(p->state != TASK_RUNNING && p->state != TASK_WAKING &&
			!(task_thread_info(p)->preempt_count & PREEMPT_ACTIVE));
1094 1095

#ifdef CONFIG_LOCKDEP
1096 1097 1098 1099 1100 1101 1102 1103 1104 1105
	/*
	 * The caller should hold either p->pi_lock or rq->lock, when changing
	 * a task's CPU. ->pi_lock for waking tasks, rq->lock for runnable tasks.
	 *
	 * sched_move_task() holds both and thus holding either pins the cgroup,
	 * see set_task_rq().
	 *
	 * Furthermore, all task_rq users should acquire both locks, see
	 * task_rq_lock().
	 */
1106 1107 1108
	WARN_ON_ONCE(debug_locks && !(lockdep_is_held(&p->pi_lock) ||
				      lockdep_is_held(&task_rq(p)->lock)));
#endif
1109 1110
#endif

1111
	trace_sched_migrate_task(p, new_cpu);
1112

1113 1114
	if (task_cpu(p) != new_cpu) {
		p->se.nr_migrations++;
1115
		perf_sw_event(PERF_COUNT_SW_CPU_MIGRATIONS, 1, NULL, 0);
1116
	}
Ingo Molnar's avatar
Ingo Molnar committed
1117 1118

	__set_task_cpu(p, new_cpu);
Ingo Molnar's avatar
Ingo Molnar committed
1119 1120
}

1121
struct migration_arg {
1122
	struct task_struct *task;
Linus Torvalds's avatar
Linus Torvalds committed
1123
	int dest_cpu;
1124
};
Linus Torvalds's avatar
Linus Torvalds committed
1125

1126 1127
static int migration_cpu_stop(void *data);

Linus Torvalds's avatar
Linus Torvalds committed
1128 1129 1130
/*
 * wait_task_inactive - wait for a thread to unschedule.
 *
1131 1132 1133 1134 1135 1136 1137
 * 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.
 *
Linus Torvalds's avatar
Linus Torvalds committed
1138 1139 1140 1141 1142 1143
 * 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.
 */
1144
unsigned long wait_task_inactive(struct task_struct *p, long match_state)
Linus Torvalds's avatar
Linus Torvalds committed
1145 1146
{
	unsigned long flags;
Ingo Molnar's avatar
Ingo Molnar committed
1147
	int running, on_rq;
1148
	unsigned long ncsw;
1149
	struct rq *rq;
Linus Torvalds's avatar
Linus Torvalds committed
1150

1151 1152 1153 1154 1155 1156 1157 1158
	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);
1159

1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170
		/*
		 * 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!
		 */
1171 1172 1173
		while (task_running(rq, p)) {
			if (match_state && unlikely(p->state != match_state))
				return 0;
1174
			cpu_relax();
1175
		}
1176

1177 1178 1179 1180 1181 1182
		/*
		 * 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);
1183
		trace_sched_wait_task(p);
1184
		running = task_running(rq, p);
Peter Zijlstra's avatar
Peter Zijlstra committed
1185
		on_rq = p->on_rq;
1186
		ncsw = 0;
1187
		if (!match_state || p->state == match_state)
1188
			ncsw = p->nvcsw | LONG_MIN; /* sets MSB */
1189
		task_rq_unlock(rq, p, &flags);
1190

1191 1192 1193 1194 1195 1196
		/*
		 * If it changed from the expected state, bail out now.
		 */
		if (unlikely(!ncsw))
			break;

1197 1198 1199 1200 1201 1202 1203 1204 1205 1206
		/*
		 * 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;
		}
1207

1208 1209 1210 1211 1212
		/*
		 * It's not enough that it's not actively running,
		 * it must be off the runqueue _entirely_, and not
		 * preempted!
		 *
1213
		 * So if it was still runnable (but just not actively
1214 1215 1216 1217
		 * running right now), it's preempted, and we should
		 * yield - it could be a while.
		 */
		if (unlikely(on_rq)) {
1218 1219 1220 1221
			ktime_t to = ktime_set(0, NSEC_PER_SEC/HZ);

			set_current_state(TASK_UNINTERRUPTIBLE);
			schedule_hrtimeout(&to, HRTIMER_MODE_REL);
1222 1223
			continue;
		}
1224

1225 1226 1227 1228 1229 1230 1231
		/*
		 * 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;
	}
1232 1233

	return ncsw;
Linus Torvalds's avatar
Linus Torvalds committed
1234 1235 1236 1237 1238 1239 1240 1241 1242
}

/***
 * 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.)
 *
Lucas De Marchi's avatar
Lucas De Marchi committed
1243
 * NOTE: this function doesn't have to take the runqueue lock,
Linus Torvalds's avatar
Linus Torvalds committed
1244 1245 1246 1247 1248
 * 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.
 */
1249
void kick_process(struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
1250 1251 1252 1253 1254 1255 1256 1257 1258
{
	int cpu;

	preempt_disable();
	cpu = task_cpu(p);
	if ((cpu != smp_processor_id()) && task_curr(p))
		smp_send_reschedule(cpu);
	preempt_enable();
}
Rusty Russell's avatar
Rusty Russell committed
1259
EXPORT_SYMBOL_GPL(kick_process);
1260
#endif /* CONFIG_SMP */
Linus Torvalds's avatar
Linus Torvalds committed
1261

1262
#ifdef CONFIG_SMP
1263
/*
1264
 * ->cpus_allowed is protected by both rq->lock and p->pi_lock
1265
 */
1266 1267 1268
static int select_fallback_rq(int cpu, struct task_struct *p)
{
	const struct cpumask *nodemask = cpumask_of_node(cpu_to_node(cpu));
1269 1270
	enum { cpuset, possible, fail } state = cpuset;
	int dest_cpu;
1271 1272

	/* Look for allowed, online CPU in same node. */
1273
	for_each_cpu(dest_cpu, nodemask) {
1274 1275 1276 1277
		if (!cpu_online(dest_cpu))
			continue;
		if (!cpu_active(dest_cpu))
			continue;
1278
		if (cpumask_test_cpu(dest_cpu, tsk_cpus_allowed(p)))
1279
			return dest_cpu;
1280
	}
1281

1282 1283
	for (;;) {
		/* Any allowed, online CPU? */
1284
		for_each_cpu(dest_cpu, tsk_cpus_allowed(p)) {
1285 1286 1287 1288 1289 1290
			if (!cpu_online(dest_cpu))
				continue;
			if (!cpu_active(dest_cpu))
				continue;
			goto out;
		}
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 1319 1320
		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);
		}
1321 1322 1323 1324 1325
	}

	return dest_cpu;
}

1326
/*
1327
 * The caller (fork, wakeup) owns p->pi_lock, ->cpus_allowed is stable.
1328
 */
1329
static inline
1330
int select_task_rq(struct task_struct *p, int sd_flags, int wake_flags)
1331
{
1332
	int cpu = p->sched_class->select_task_rq(p, sd_flags, wake_flags);
1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343

	/*
	 * 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 ]
	 */
1344
	if (unlikely(!cpumask_test_cpu(cpu, tsk_cpus_allowed(p)) ||
Peter Zijlstra's avatar
Peter Zijlstra committed
1345
		     !cpu_online(cpu)))
1346
		cpu = select_fallback_rq(task_cpu(p), p);
1347 1348

	return cpu;
1349
}
1350 1351 1352 1353 1354 1355

static void update_avg(u64 *avg, u64 sample)
{
	s64 diff = sample - *avg;
	*avg += diff >> 3;
}
1356 1357
#endif

1358
static void
1359
ttwu_stat(struct task_struct *p, int cpu, int wake_flags)
Tejun Heo's avatar
Tejun Heo committed
1360
{
1361
#ifdef CONFIG_SCHEDSTATS
1362 1363
	struct rq *rq = this_rq();

1364 1365 1366 1367 1368 1369 1370 1371 1372 1373
#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);
1374
		rcu_read_lock();
1375 1376 1377 1378 1379 1380
		for_each_domain(this_cpu, sd) {
			if (cpumask_test_cpu(cpu, sched_domain_span(sd))) {
				schedstat_inc(sd, ttwu_wake_remote);
				break;
			}
		}
1381
		rcu_read_unlock();
1382
	}
1383 1384 1385 1386

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

1387 1388 1389
#endif /* CONFIG_SMP */

	schedstat_inc(rq, ttwu_count);
Tejun Heo's avatar
Tejun Heo committed
1390
	schedstat_inc(p, se.statistics.nr_wakeups);
1391 1392

	if (wake_flags & WF_SYNC)
Tejun Heo's avatar
Tejun Heo committed
1393
		schedstat_inc(p, se.statistics.nr_wakeups_sync);
1394 1395 1396 1397 1398 1399

#endif /* CONFIG_SCHEDSTATS */
}

static void ttwu_activate(struct rq *rq, struct task_struct *p, int en_flags)
{
Tejun Heo's avatar
Tejun Heo committed
1400
	activate_task(rq, p, en_flags);
Peter Zijlstra's avatar
Peter Zijlstra committed
1401
	p->on_rq = 1;
1402 1403 1404 1405

	/* if a worker is waking up, notify workqueue */
	if (p->flags & PF_WQ_WORKER)
		wq_worker_waking_up(p, cpu_of(rq));
Tejun Heo's avatar
Tejun Heo committed
1406 1407
}

1408 1409 1410
/*
 * Mark the task runnable and perform wakeup-preemption.
 */
1411
static void
1412
ttwu_do_wakeup(struct rq *rq, struct task_struct *p, int wake_flags)
Tejun Heo's avatar
Tejun Heo committed
1413
{
1414
	trace_sched_wakeup(p, true);
Tejun Heo's avatar
Tejun Heo committed
1415 1416 1417 1418 1419 1420 1421
	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);

1422
	if (rq->idle_stamp) {
Tejun Heo's avatar
Tejun Heo committed
1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434
		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
}

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 1466 1467
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;
}

1468
#ifdef CONFIG_SMP
1469
static void sched_ttwu_pending(void)
1470 1471
{
	struct rq *rq = this_rq();
1472 1473
	struct llist_node *llist = llist_del_all(&rq->wake_list);
	struct task_struct *p;
1474 1475 1476

	raw_spin_lock(&rq->lock);

1477 1478 1479
	while (llist) {
		p = llist_entry(llist, struct task_struct, wake_entry);
		llist = llist_next(llist);
1480 1481 1482 1483 1484 1485 1486 1487
		ttwu_do_activate(rq, p, 0);
	}

	raw_spin_unlock(&rq->lock);
}

void scheduler_ipi(void)
{
1488
	if (llist_empty(&this_rq()->wake_list) && !got_nohz_idle_kick())
1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504
		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();
1505
	sched_ttwu_pending();
1506 1507 1508 1509

	/*
	 * Check if someone kicked us for doing the nohz idle load balance.
	 */
1510 1511
	if (unlikely(got_nohz_idle_kick() && !need_resched())) {
		this_rq()->idle_balance = 1;
1512
		raise_softirq_irqoff(SCHED_SOFTIRQ);
1513
	}
1514
	irq_exit();
1515 1516 1517 1518
}

static void ttwu_queue_remote(struct task_struct *p, int cpu)
{
1519
	if (llist_add(&p->wake_entry, &cpu_rq(cpu)->wake_list))
1520 1521
		smp_send_reschedule(cpu);
}
1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540

#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 */
1541

1542
bool cpus_share_cache(int this_cpu, int that_cpu)
1543 1544 1545
{
	return per_cpu(sd_llc_id, this_cpu) == per_cpu(sd_llc_id, that_cpu);
}
1546
#endif /* CONFIG_SMP */
1547

1548 1549 1550 1551
static void ttwu_queue(struct task_struct *p, int cpu)
{
	struct rq *rq = cpu_rq(cpu);

1552
#if defined(CONFIG_SMP)
1553
	if (sched_feat(TTWU_QUEUE) && !cpus_share_cache(smp_processor_id(), cpu)) {
1554
		sched_clock_cpu(cpu); /* sync clocks x-cpu */
1555 1556 1557 1558 1559
		ttwu_queue_remote(p, cpu);
		return;
	}
#endif

1560 1561 1562
	raw_spin_lock(&rq->lock);
	ttwu_do_activate(rq, p, 0);
	raw_spin_unlock(&rq->lock);
Tejun Heo's avatar
Tejun Heo committed
1563 1564 1565
}

/**
Linus Torvalds's avatar
Linus Torvalds committed
1566
 * try_to_wake_up - wake up a thread
Tejun Heo's avatar
Tejun Heo committed
1567
 * @p: the thread to be awakened
Linus Torvalds's avatar
Linus Torvalds committed
1568
 * @state: the mask of task states that can be woken
Tejun Heo's avatar
Tejun Heo committed
1569
 * @wake_flags: wake modifier flags (WF_*)
Linus Torvalds's avatar
Linus Torvalds committed
1570 1571 1572 1573 1574 1575 1576
 *
 * 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.
 *
Tejun Heo's avatar
Tejun Heo committed
1577 1578
 * Returns %true if @p was woken up, %false if it was already running
 * or @state didn't match @p's state.
Linus Torvalds's avatar
Linus Torvalds committed
1579
 */
1580 1581
static int
try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
Linus Torvalds's avatar
Linus Torvalds committed
1582 1583
{
	unsigned long flags;
1584
	int cpu, success = 0;
1585

1586
	smp_wmb();
1587
	raw_spin_lock_irqsave(&p->pi_lock, flags);
Peter Zijlstra's avatar
Peter Zijlstra committed
1588
	if (!(p->state & state))
Linus Torvalds's avatar
Linus Torvalds committed
1589 1590
		goto out;

1591
	success = 1; /* we're going to change ->state */
Linus Torvalds's avatar
Linus Torvalds committed
1592 1593
	cpu = task_cpu(p);

1594 1595
	if (p->on_rq && ttwu_remote(p, wake_flags))
		goto stat;
Linus Torvalds's avatar
Linus Torvalds committed
1596 1597

#ifdef CONFIG_SMP
Peter Zijlstra's avatar
Peter Zijlstra committed
1598
	/*
1599 1600
	 * If the owning (remote) cpu is still in the middle of schedule() with
	 * this task as prev, wait until its done referencing the task.
Peter Zijlstra's avatar
Peter Zijlstra committed
1601
	 */
1602 1603 1604
	while (p->on_cpu) {
#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
		/*
1605 1606 1607 1608 1609
		 * 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.
1610
		 */
1611
		if (ttwu_activate_remote(p, wake_flags))
1612
			goto stat;
1613
#else
1614
		cpu_relax();
1615
#endif
1616
	}
1617
	/*
1618
	 * Pairs with the smp_wmb() in finish_lock_switch().
1619
	 */
1620
	smp_rmb();
Linus Torvalds's avatar
Linus Torvalds committed
1621

1622
	p->sched_contributes_to_load = !!task_contributes_to_load(p);
Peter Zijlstra's avatar
Peter Zijlstra committed
1623
	p->state = TASK_WAKING;
1624

1625
	if (p->sched_class->task_waking)
1626
		p->sched_class->task_waking(p);
1627

1628
	cpu = select_task_rq(p, SD_BALANCE_WAKE, wake_flags);
1629 1630
	if (task_cpu(p) != cpu) {
		wake_flags |= WF_MIGRATED;
1631
		set_task_cpu(p, cpu);
1632
	}
Linus Torvalds's avatar
Linus Torvalds committed
1633 1634
#endif /* CONFIG_SMP */

1635 1636
	ttwu_queue(p, cpu);
stat:
1637
	ttwu_stat(p, cpu, wake_flags);
Linus Torvalds's avatar
Linus Torvalds committed
1638
out:
1639
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
1640 1641 1642 1643

	return success;
}

Tejun Heo's avatar
Tejun Heo committed
1644 1645 1646 1647
/**
 * try_to_wake_up_local - try to wake up a local task with rq lock held
 * @p: the thread to be awakened
 *
1648
 * Put @p on the run-queue if it's not already there. The caller must
Tejun Heo's avatar
Tejun Heo committed
1649
 * ensure that this_rq() is locked, @p is bound to this_rq() and not
1650
 * the current task.
Tejun Heo's avatar
Tejun Heo committed
1651 1652 1653 1654 1655 1656 1657 1658 1659
 */
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);

1660 1661 1662 1663 1664 1665
	if (!raw_spin_trylock(&p->pi_lock)) {
		raw_spin_unlock(&rq->lock);
		raw_spin_lock(&p->pi_lock);
		raw_spin_lock(&rq->lock);
	}

Tejun Heo's avatar
Tejun Heo committed
1666
	if (!(p->state & TASK_NORMAL))
1667
		goto out;
Tejun Heo's avatar
Tejun Heo committed
1668

Peter Zijlstra's avatar
Peter Zijlstra committed
1669
	if (!p->on_rq)
1670 1671
		ttwu_activate(rq, p, ENQUEUE_WAKEUP);

1672
	ttwu_do_wakeup(rq, p, 0);
1673
	ttwu_stat(p, smp_processor_id(), 0);
1674 1675
out:
	raw_spin_unlock(&p->pi_lock);
Tejun Heo's avatar
Tejun Heo committed
1676 1677
}

1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688
/**
 * 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.
 */
1689
int wake_up_process(struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
1690
{
1691
	return try_to_wake_up(p, TASK_ALL, 0);
Linus Torvalds's avatar
Linus Torvalds committed
1692 1693 1694
}
EXPORT_SYMBOL(wake_up_process);

1695
int wake_up_state(struct task_struct *p, unsigned int state)
Linus Torvalds's avatar
Linus Torvalds committed
1696 1697 1698 1699 1700 1701 1702
{
	return try_to_wake_up(p, state, 0);
}

/*
 * Perform scheduler related setup for a newly forked process p.
 * p is forked by current.
Ingo Molnar's avatar
Ingo Molnar committed
1703 1704 1705 1706 1707
 *
 * __sched_fork() is basic setup used by init_idle() too:
 */
static void __sched_fork(struct task_struct *p)
{
Peter Zijlstra's avatar
Peter Zijlstra committed
1708 1709 1710
	p->on_rq			= 0;

	p->se.on_rq			= 0;
Ingo Molnar's avatar
Ingo Molnar committed
1711 1712
	p->se.exec_start		= 0;
	p->se.sum_exec_runtime		= 0;
1713
	p->se.prev_sum_exec_runtime	= 0;
1714
	p->se.nr_migrations		= 0;
1715
	p->se.vruntime			= 0;
Peter Zijlstra's avatar
Peter Zijlstra committed
1716
	INIT_LIST_HEAD(&p->se.group_node);
1717 1718

#ifdef CONFIG_SCHEDSTATS
1719
	memset(&p->se.statistics, 0, sizeof(p->se.statistics));
1720
#endif
1721

Peter Zijlstra's avatar
Peter Zijlstra committed
1722
	INIT_LIST_HEAD(&p->rt.run_list);
1723

1724 1725 1726
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&p->preempt_notifiers);
#endif
Ingo Molnar's avatar
Ingo Molnar committed
1727 1728 1729 1730 1731
}

/*
 * fork()/clone()-time setup:
 */
1732
void sched_fork(struct task_struct *p)
Ingo Molnar's avatar
Ingo Molnar committed
1733
{
1734
	unsigned long flags;
Ingo Molnar's avatar
Ingo Molnar committed
1735 1736 1737
	int cpu = get_cpu();

	__sched_fork(p);
1738
	/*
1739
	 * We mark the process as running here. This guarantees that
1740 1741 1742
	 * nobody will actually run it, and a signal or other external
	 * event cannot wake it up and insert it on the runqueue either.
	 */
1743
	p->state = TASK_RUNNING;
Ingo Molnar's avatar
Ingo Molnar committed
1744

1745 1746 1747 1748 1749
	/*
	 * Make sure we do not leak PI boosting priority to the child.
	 */
	p->prio = current->normal_prio;

1750 1751 1752 1753
	/*
	 * Revert to default priority/policy on fork if requested.
	 */
	if (unlikely(p->sched_reset_on_fork)) {
1754
		if (task_has_rt_policy(p)) {
1755
			p->policy = SCHED_NORMAL;
1756
			p->static_prio = NICE_TO_PRIO(0);
1757 1758 1759 1760 1761 1762
			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);
1763

1764 1765 1766 1767 1768 1769
		/*
		 * We don't need the reset flag anymore after the fork. It has
		 * fulfilled its duty:
		 */
		p->sched_reset_on_fork = 0;
	}
1770

1771 1772
	if (!rt_prio(p->prio))
		p->sched_class = &fair_sched_class;
1773

1774 1775 1776
	if (p->sched_class->task_fork)
		p->sched_class->task_fork(p);

1777 1778 1779 1780 1781 1782 1783
	/*
	 * 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.
	 */
1784
	raw_spin_lock_irqsave(&p->pi_lock, flags);
1785
	set_task_cpu(p, cpu);
1786
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
1787

1788
#if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT)
Ingo Molnar's avatar
Ingo Molnar committed
1789
	if (likely(sched_info_on()))
1790
		memset(&p->sched_info, 0, sizeof(p->sched_info));
Linus Torvalds's avatar
Linus Torvalds committed
1791
#endif
1792 1793
#if defined(CONFIG_SMP)
	p->on_cpu = 0;
1794
#endif
1795
#ifdef CONFIG_PREEMPT_COUNT
1796
	/* Want to start with kernel preemption disabled. */
1797
	task_thread_info(p)->preempt_count = 1;
Linus Torvalds's avatar
Linus Torvalds committed
1798
#endif
1799
#ifdef CONFIG_SMP
1800
	plist_node_init(&p->pushable_tasks, MAX_PRIO);
1801
#endif
1802

1803
	put_cpu();
Linus Torvalds's avatar
Linus Torvalds committed
1804 1805 1806 1807 1808 1809 1810 1811 1812
}

/*
 * 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.
 */
1813
void wake_up_new_task(struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
1814 1815
{
	unsigned long flags;
Ingo Molnar's avatar
Ingo Molnar committed
1816
	struct rq *rq;
1817

1818
	raw_spin_lock_irqsave(&p->pi_lock, flags);
1819 1820 1821 1822 1823 1824
#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
	 */
1825
	set_task_cpu(p, select_task_rq(p, SD_BALANCE_FORK, 0));
1826 1827
#endif

1828
	rq = __task_rq_lock(p);
1829
	activate_task(rq, p, 0);
Peter Zijlstra's avatar
Peter Zijlstra committed
1830
	p->on_rq = 1;
1831
	trace_sched_wakeup_new(p, true);
Peter Zijlstra's avatar
Peter Zijlstra committed
1832
	check_preempt_curr(rq, p, WF_FORK);
1833
#ifdef CONFIG_SMP
1834 1835
	if (p->sched_class->task_woken)
		p->sched_class->task_woken(rq, p);
1836
#endif
1837
	task_rq_unlock(rq, p, &flags);
Linus Torvalds's avatar
Linus Torvalds committed
1838 1839
}

1840 1841 1842
#ifdef CONFIG_PREEMPT_NOTIFIERS

/**
1843
 * preempt_notifier_register - tell me when current is being preempted & rescheduled
1844
 * @notifier: notifier struct to register
1845 1846 1847 1848 1849 1850 1851 1852 1853
 */
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
1854
 * @notifier: notifier struct to unregister
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 1882 1883
 *
 * 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);
}

1884
#else /* !CONFIG_PREEMPT_NOTIFIERS */
1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895

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)
{
}

1896
#endif /* CONFIG_PREEMPT_NOTIFIERS */
1897

1898 1899 1900
/**
 * prepare_task_switch - prepare to switch tasks
 * @rq: the runqueue preparing to switch
1901
 * @prev: the current task that is being switched out
1902 1903 1904 1905 1906 1907 1908 1909 1910
 * @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.
 */
1911 1912 1913
static inline void
prepare_task_switch(struct rq *rq, struct task_struct *prev,
		    struct task_struct *next)
1914
{
1915 1916
	sched_info_switch(prev, next);
	perf_event_task_sched_out(prev, next);
1917
	fire_sched_out_preempt_notifiers(prev, next);
1918 1919
	prepare_lock_switch(rq, next);
	prepare_arch_switch(next);
1920
	trace_sched_switch(prev, next);
1921 1922
}

Linus Torvalds's avatar
Linus Torvalds committed
1923 1924
/**
 * finish_task_switch - clean up after a task-switch
1925
 * @rq: runqueue associated with task-switch
Linus Torvalds's avatar
Linus Torvalds committed
1926 1927
 * @prev: the thread we just switched away from.
 *
1928 1929 1930 1931
 * 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.
Linus Torvalds's avatar
Linus Torvalds committed
1932 1933
 *
 * Note that we may have delayed dropping an mm in context_switch(). If
Ingo Molnar's avatar
Ingo Molnar committed
1934
 * so, we finish that here outside of the runqueue lock. (Doing it
Linus Torvalds's avatar
Linus Torvalds committed
1935 1936 1937
 * with the lock held can cause deadlocks; see schedule() for
 * details.)
 */
Alexey Dobriyan's avatar
Alexey Dobriyan committed
1938
static void finish_task_switch(struct rq *rq, struct task_struct *prev)
Linus Torvalds's avatar
Linus Torvalds committed
1939 1940 1941
	__releases(rq->lock)
{
	struct mm_struct *mm = rq->prev_mm;
Oleg Nesterov's avatar
Oleg Nesterov committed
1942
	long prev_state;
Linus Torvalds's avatar
Linus Torvalds committed
1943 1944 1945 1946 1947

	rq->prev_mm = NULL;

	/*
	 * A task struct has one reference for the use as "current".
1948
	 * If a task dies, then it sets TASK_DEAD in tsk->state and calls
Oleg Nesterov's avatar
Oleg Nesterov committed
1949 1950
	 * schedule one last time. The schedule call will never return, and
	 * the scheduled task must drop that reference.
1951
	 * The test for TASK_DEAD must occur while the runqueue locks are
Linus Torvalds's avatar
Linus Torvalds committed
1952 1953 1954 1955 1956
	 * 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>
	 */
Oleg Nesterov's avatar
Oleg Nesterov committed
1957
	prev_state = prev->state;
1958
	finish_arch_switch(prev);
1959 1960 1961
#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
	local_irq_disable();
#endif /* __ARCH_WANT_INTERRUPTS_ON_CTXSW */
1962
	perf_event_task_sched_in(prev, current);
1963 1964 1965
#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
	local_irq_enable();
#endif /* __ARCH_WANT_INTERRUPTS_ON_CTXSW */
1966
	finish_lock_switch(rq, prev);
1967
	finish_arch_post_lock_switch();
Steven Rostedt's avatar
Steven Rostedt committed
1968

1969
	fire_sched_in_preempt_notifiers(current);
Linus Torvalds's avatar
Linus Torvalds committed
1970 1971
	if (mm)
		mmdrop(mm);
1972
	if (unlikely(prev_state == TASK_DEAD)) {
1973 1974 1975
		/*
		 * Remove function-return probe instances associated with this
		 * task and put them back on the free list.
Ingo Molnar's avatar
Ingo Molnar committed
1976
		 */
1977
		kprobe_flush_task(prev);
Linus Torvalds's avatar
Linus Torvalds committed
1978
		put_task_struct(prev);
1979
	}
Linus Torvalds's avatar
Linus Torvalds committed
1980 1981
}

1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996
#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;

1997
		raw_spin_lock_irqsave(&rq->lock, flags);
1998 1999
		if (rq->curr->sched_class->post_schedule)
			rq->curr->sched_class->post_schedule(rq);
2000
		raw_spin_unlock_irqrestore(&rq->lock, flags);
2001 2002 2003 2004 2005 2006

		rq->post_schedule = 0;
	}
}

#else
2007

2008 2009 2010 2011 2012 2013
static inline void pre_schedule(struct rq *rq, struct task_struct *p)
{
}

static inline void post_schedule(struct rq *rq)
{
Linus Torvalds's avatar
Linus Torvalds committed
2014 2015
}

2016 2017
#endif

Linus Torvalds's avatar
Linus Torvalds committed
2018 2019 2020 2021
/**
 * schedule_tail - first thing a freshly forked thread must call.
 * @prev: the thread we just switched away from.
 */
2022
asmlinkage void schedule_tail(struct task_struct *prev)
Linus Torvalds's avatar
Linus Torvalds committed
2023 2024
	__releases(rq->lock)
{
2025 2026
	struct rq *rq = this_rq();

2027
	finish_task_switch(rq, prev);
2028

2029 2030 2031 2032 2033
	/*
	 * FIXME: do we need to worry about rq being invalidated by the
	 * task_switch?
	 */
	post_schedule(rq);
2034

2035 2036 2037 2038
#ifdef __ARCH_WANT_UNLOCKED_CTXSW
	/* In this case, finish_task_switch does not reenable preemption */
	preempt_enable();
#endif
Linus Torvalds's avatar
Linus Torvalds committed
2039
	if (current->set_child_tid)
2040
		put_user(task_pid_vnr(current), current->set_child_tid);
Linus Torvalds's avatar
Linus Torvalds committed
2041 2042 2043 2044 2045 2046
}

/*
 * context_switch - switch to the new MM and the new
 * thread's register state.
 */
Ingo Molnar's avatar
Ingo Molnar committed
2047
static inline void
2048
context_switch(struct rq *rq, struct task_struct *prev,
2049
	       struct task_struct *next)
Linus Torvalds's avatar
Linus Torvalds committed
2050
{
Ingo Molnar's avatar
Ingo Molnar committed
2051
	struct mm_struct *mm, *oldmm;
Linus Torvalds's avatar
Linus Torvalds committed
2052

2053
	prepare_task_switch(rq, prev, next);
2054

Ingo Molnar's avatar
Ingo Molnar committed
2055 2056
	mm = next->mm;
	oldmm = prev->active_mm;
2057 2058 2059 2060 2061
	/*
	 * For paravirt, this is coupled with an exit in switch_to to
	 * combine the page table reload and the switch backend into
	 * one hypercall.
	 */
2062
	arch_start_context_switch(prev);
2063

2064
	if (!mm) {
Linus Torvalds's avatar
Linus Torvalds committed
2065 2066 2067 2068 2069 2070
		next->active_mm = oldmm;
		atomic_inc(&oldmm->mm_count);
		enter_lazy_tlb(oldmm, next);
	} else
		switch_mm(oldmm, mm, next);

2071
	if (!prev->mm) {
Linus Torvalds's avatar
Linus Torvalds committed
2072 2073 2074
		prev->active_mm = NULL;
		rq->prev_mm = oldmm;
	}
2075 2076 2077 2078 2079 2080 2081
	/*
	 * 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
2082
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
2083
#endif
Linus Torvalds's avatar
Linus Torvalds committed
2084 2085 2086 2087

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

Ingo Molnar's avatar
Ingo Molnar committed
2088 2089 2090 2091 2092 2093 2094
	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);
Linus Torvalds's avatar
Linus Torvalds committed
2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111
}

/*
 * 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;
2112
}
Linus Torvalds's avatar
Linus Torvalds committed
2113 2114

unsigned long nr_uninterruptible(void)
2115
{
Linus Torvalds's avatar
Linus Torvalds committed
2116
	unsigned long i, sum = 0;
2117

2118
	for_each_possible_cpu(i)
Linus Torvalds's avatar
Linus Torvalds committed
2119
		sum += cpu_rq(i)->nr_uninterruptible;
2120 2121

	/*
Linus Torvalds's avatar
Linus Torvalds committed
2122 2123
	 * Since we read the counters lockless, it might be slightly
	 * inaccurate. Do not allow it to go below zero though:
2124
	 */
Linus Torvalds's avatar
Linus Torvalds committed
2125 2126
	if (unlikely((long)sum < 0))
		sum = 0;
2127

Linus Torvalds's avatar
Linus Torvalds committed
2128
	return sum;
2129 2130
}

Linus Torvalds's avatar
Linus Torvalds committed
2131
unsigned long long nr_context_switches(void)
2132
{
2133 2134
	int i;
	unsigned long long sum = 0;
2135

2136
	for_each_possible_cpu(i)
Linus Torvalds's avatar
Linus Torvalds committed
2137
		sum += cpu_rq(i)->nr_switches;
2138

Linus Torvalds's avatar
Linus Torvalds committed
2139 2140
	return sum;
}
2141

Linus Torvalds's avatar
Linus Torvalds committed
2142 2143 2144
unsigned long nr_iowait(void)
{
	unsigned long i, sum = 0;
2145

2146
	for_each_possible_cpu(i)
Linus Torvalds's avatar
Linus Torvalds committed
2147
		sum += atomic_read(&cpu_rq(i)->nr_iowait);
2148

Linus Torvalds's avatar
Linus Torvalds committed
2149 2150
	return sum;
}
2151

2152
unsigned long nr_iowait_cpu(int cpu)
2153
{
2154
	struct rq *this = cpu_rq(cpu);
2155 2156
	return atomic_read(&this->nr_iowait);
}
2157

2158 2159 2160 2161 2162
unsigned long this_cpu_load(void)
{
	struct rq *this = this_rq();
	return this->cpu_load[0];
}
2163

2164

2165 2166 2167 2168 2169
/* Variables and functions for calc_load */
static atomic_long_t calc_load_tasks;
static unsigned long calc_load_update;
unsigned long avenrun[3];
EXPORT_SYMBOL(avenrun);
2170

2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185
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;
}

2186 2187 2188 2189 2190 2191 2192 2193 2194
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;
}

2195 2196 2197 2198 2199 2200 2201 2202
#ifdef CONFIG_NO_HZ
/*
 * For NO_HZ we delay the active fold to the next LOAD_FREQ update.
 *
 * When making the ILB scale, we should try to pull this in as well.
 */
static atomic_long_t calc_load_tasks_idle;

2203
void calc_load_account_idle(struct rq *this_rq)
2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223
{
	long delta;

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

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

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

	return delta;
}
2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301

/**
 * 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.
 */
2302
static void calc_global_nohz(void)
2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316
{
	long delta, active, n;

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

	/*
2317
	 * It could be the one fold was all it took, we done!
2318
	 */
2319 2320
	if (time_before(jiffies, calc_load_update + 10))
		return;
2321

2322 2323 2324 2325 2326
	/*
	 * Catch-up, fold however many we are behind still
	 */
	delta = jiffies - calc_load_update - 10;
	n = 1 + (delta / LOAD_FREQ);
2327

2328 2329
	active = atomic_long_read(&calc_load_tasks);
	active = active > 0 ? active * FIXED_1 : 0;
2330

2331 2332 2333
	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);
2334

2335
	calc_load_update += n * LOAD_FREQ;
2336
}
2337
#else
2338
void calc_load_account_idle(struct rq *this_rq)
2339 2340 2341 2342 2343 2344 2345
{
}

static inline long calc_load_fold_idle(void)
{
	return 0;
}
2346

2347
static void calc_global_nohz(void)
2348 2349
{
}
2350 2351
#endif

2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364
/**
 * 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;
2365 2366 2367
}

/*
2368 2369
 * calc_load - update the avenrun load estimates 10 ticks after the
 * CPUs have updated calc_load_tasks.
2370
 */
2371
void calc_global_load(unsigned long ticks)
2372
{
2373
	long active;
Linus Torvalds's avatar
Linus Torvalds committed
2374

2375
	if (time_before(jiffies, calc_load_update + 10))
2376
		return;
Linus Torvalds's avatar
Linus Torvalds committed
2377

2378 2379
	active = atomic_long_read(&calc_load_tasks);
	active = active > 0 ? active * FIXED_1 : 0;
Linus Torvalds's avatar
Linus Torvalds committed
2380

2381 2382 2383
	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);
Ingo Molnar's avatar
Ingo Molnar committed
2384

2385
	calc_load_update += LOAD_FREQ;
2386 2387 2388 2389 2390 2391 2392 2393 2394 2395

	/*
	 * Account one period with whatever state we found before
	 * folding in the nohz state and ageing the entire idle period.
	 *
	 * This avoids loosing a sample when we go idle between 
	 * calc_load_account_active() (10 ticks ago) and now and thus
	 * under-accounting.
	 */
	calc_global_nohz();
2396
}
Linus Torvalds's avatar
Linus Torvalds committed
2397

2398
/*
2399 2400
 * Called from update_cpu_load() to periodically update this CPU's
 * active count.
2401 2402 2403
 */
static void calc_load_account_active(struct rq *this_rq)
{
2404
	long delta;
2405

2406 2407
	if (time_before(jiffies, this_rq->calc_load_update))
		return;
2408

2409 2410 2411
	delta  = calc_load_fold_active(this_rq);
	delta += calc_load_fold_idle();
	if (delta)
2412
		atomic_long_add(delta, &calc_load_tasks);
2413 2414

	this_rq->calc_load_update += LOAD_FREQ;
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 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483
/*
 * 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;
}

2484
/*
Ingo Molnar's avatar
Ingo Molnar committed
2485
 * Update rq->cpu_load[] statistics. This function is usually called every
2486 2487
 * scheduler tick (TICK_NSEC). With tickless idle this will not be called
 * every tick. We fix it up based on jiffies.
2488
 */
2489
void update_cpu_load(struct rq *this_rq)
2490
{
2491
	unsigned long this_load = this_rq->load.weight;
2492 2493
	unsigned long curr_jiffies = jiffies;
	unsigned long pending_updates;
Ingo Molnar's avatar
Ingo Molnar committed
2494
	int i, scale;
2495

Ingo Molnar's avatar
Ingo Molnar committed
2496
	this_rq->nr_load_updates++;
2497

2498 2499 2500 2501 2502 2503 2504
	/* Avoid repeated calls on same jiffy, when moving in and out of idle */
	if (curr_jiffies == this_rq->last_load_update_tick)
		return;

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

Ingo Molnar's avatar
Ingo Molnar committed
2505
	/* Update our load: */
2506 2507
	this_rq->cpu_load[0] = this_load; /* Fasttrack for idx 0 */
	for (i = 1, scale = 2; i < CPU_LOAD_IDX_MAX; i++, scale += scale) {
Ingo Molnar's avatar
Ingo Molnar committed
2508
		unsigned long old_load, new_load;
2509

Ingo Molnar's avatar
Ingo Molnar committed
2510
		/* scale is effectively 1 << i now, and >> i divides by scale */
2511

Ingo Molnar's avatar
Ingo Molnar committed
2512
		old_load = this_rq->cpu_load[i];
2513
		old_load = decay_load_missed(old_load, pending_updates - 1, i);
Ingo Molnar's avatar
Ingo Molnar committed
2514
		new_load = this_load;
Ingo Molnar's avatar
Ingo Molnar committed
2515 2516 2517 2518 2519 2520
		/*
		 * 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)
2521 2522 2523
			new_load += scale - 1;

		this_rq->cpu_load[i] = (old_load * (scale - 1) + new_load) >> i;
Ingo Molnar's avatar
Ingo Molnar committed
2524
	}
2525 2526

	sched_avg_update(this_rq);
2527 2528 2529 2530 2531
}

static void update_cpu_load_active(struct rq *this_rq)
{
	update_cpu_load(this_rq);
2532

2533
	calc_load_account_active(this_rq);
2534 2535
}

Ingo Molnar's avatar
Ingo Molnar committed
2536
#ifdef CONFIG_SMP
2537

2538
/*
2539 2540
 * sched_exec - execve() is a valuable balancing opportunity, because at
 * this point the task has the smallest effective memory and cache footprint.
2541
 */
2542
void sched_exec(void)
2543
{
2544
	struct task_struct *p = current;
Linus Torvalds's avatar
Linus Torvalds committed
2545
	unsigned long flags;
2546
	int dest_cpu;
2547

2548
	raw_spin_lock_irqsave(&p->pi_lock, flags);
2549
	dest_cpu = p->sched_class->select_task_rq(p, SD_BALANCE_EXEC, 0);
2550 2551
	if (dest_cpu == smp_processor_id())
		goto unlock;
2552

2553
	if (likely(cpu_active(dest_cpu))) {
2554
		struct migration_arg arg = { p, dest_cpu };
2555

2556 2557
		raw_spin_unlock_irqrestore(&p->pi_lock, flags);
		stop_one_cpu(task_cpu(p), migration_cpu_stop, &arg);
Linus Torvalds's avatar
Linus Torvalds committed
2558 2559
		return;
	}
2560
unlock:
2561
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
2562
}
Ingo Molnar's avatar
Ingo Molnar committed
2563

Linus Torvalds's avatar
Linus Torvalds committed
2564 2565 2566
#endif

DEFINE_PER_CPU(struct kernel_stat, kstat);
2567
DEFINE_PER_CPU(struct kernel_cpustat, kernel_cpustat);
Linus Torvalds's avatar
Linus Torvalds committed
2568 2569

EXPORT_PER_CPU_SYMBOL(kstat);
2570
EXPORT_PER_CPU_SYMBOL(kernel_cpustat);
Linus Torvalds's avatar
Linus Torvalds committed
2571 2572

/*
2573
 * Return any ns on the sched_clock that have not yet been accounted in
2574
 * @p in case that task is currently running.
2575 2576
 *
 * Called with task_rq_lock() held on @rq.
Linus Torvalds's avatar
Linus Torvalds committed
2577
 */
2578 2579 2580 2581 2582 2583
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);
2584
		ns = rq->clock_task - p->se.exec_start;
2585 2586 2587 2588 2589 2590 2591
		if ((s64)ns < 0)
			ns = 0;
	}

	return ns;
}

2592
unsigned long long task_delta_exec(struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
2593 2594
{
	unsigned long flags;
2595
	struct rq *rq;
2596
	u64 ns = 0;
2597

2598
	rq = task_rq_lock(p, &flags);
2599
	ns = do_task_delta_exec(p, rq);
2600
	task_rq_unlock(rq, p, &flags);
2601

2602 2603
	return ns;
}
2604

2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617
/*
 * 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);
2618
	task_rq_unlock(rq, p, &flags);
2619 2620 2621

	return ns;
}
2622

2623 2624 2625 2626 2627
#ifdef CONFIG_CGROUP_CPUACCT
struct cgroup_subsys cpuacct_subsys;
struct cpuacct root_cpuacct;
#endif

2628 2629
static inline void task_group_account_field(struct task_struct *p, int index,
					    u64 tmp)
2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658
{
#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
}


Linus Torvalds's avatar
Linus Torvalds committed
2659 2660 2661 2662
/*
 * 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
2663
 * @cputime_scaled: cputime scaled by cpu frequency
Linus Torvalds's avatar
Linus Torvalds committed
2664
 */
2665 2666
void account_user_time(struct task_struct *p, cputime_t cputime,
		       cputime_t cputime_scaled)
Linus Torvalds's avatar
Linus Torvalds committed
2667
{
2668
	int index;
Linus Torvalds's avatar
Linus Torvalds committed
2669

2670
	/* Add user time to process. */
2671 2672
	p->utime += cputime;
	p->utimescaled += cputime_scaled;
2673
	account_group_user_time(p, cputime);
Linus Torvalds's avatar
Linus Torvalds committed
2674

2675
	index = (TASK_NICE(p) > 0) ? CPUTIME_NICE : CPUTIME_USER;
2676

Linus Torvalds's avatar
Linus Torvalds committed
2677
	/* Add user time to cpustat. */
2678
	task_group_account_field(p, index, (__force u64) cputime);
2679

2680 2681
	/* Account for user time used */
	acct_update_integrals(p);
Linus Torvalds's avatar
Linus Torvalds committed
2682 2683
}

2684 2685 2686 2687
/*
 * 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
2688
 * @cputime_scaled: cputime scaled by cpu frequency
2689
 */
2690 2691
static void account_guest_time(struct task_struct *p, cputime_t cputime,
			       cputime_t cputime_scaled)
2692
{
2693
	u64 *cpustat = kcpustat_this_cpu->cpustat;
2694

2695
	/* Add guest time to process. */
2696 2697
	p->utime += cputime;
	p->utimescaled += cputime_scaled;
2698
	account_group_user_time(p, cputime);
2699
	p->gtime += cputime;
2700

2701
	/* Add guest time to cpustat. */
2702
	if (TASK_NICE(p) > 0) {
2703 2704
		cpustat[CPUTIME_NICE] += (__force u64) cputime;
		cpustat[CPUTIME_GUEST_NICE] += (__force u64) cputime;
2705
	} else {
2706 2707
		cpustat[CPUTIME_USER] += (__force u64) cputime;
		cpustat[CPUTIME_GUEST] += (__force u64) cputime;
2708
	}
2709 2710
}

2711 2712 2713 2714 2715 2716 2717 2718 2719
/*
 * 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,
2720
			cputime_t cputime_scaled, int index)
2721 2722
{
	/* Add system time to process. */
2723 2724
	p->stime += cputime;
	p->stimescaled += cputime_scaled;
2725 2726 2727
	account_group_system_time(p, cputime);

	/* Add system time to cpustat. */
2728
	task_group_account_field(p, index, (__force u64) cputime);
2729 2730 2731 2732 2733

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

Linus Torvalds's avatar
Linus Torvalds committed
2734 2735 2736 2737 2738
/*
 * 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
2739
 * @cputime_scaled: cputime scaled by cpu frequency
Linus Torvalds's avatar
Linus Torvalds committed
2740 2741
 */
void account_system_time(struct task_struct *p, int hardirq_offset,
2742
			 cputime_t cputime, cputime_t cputime_scaled)
Linus Torvalds's avatar
Linus Torvalds committed
2743
{
2744
	int index;
Linus Torvalds's avatar
Linus Torvalds committed
2745

2746
	if ((p->flags & PF_VCPU) && (irq_count() - hardirq_offset == 0)) {
2747
		account_guest_time(p, cputime, cputime_scaled);
2748 2749
		return;
	}
2750

Linus Torvalds's avatar
Linus Torvalds committed
2751
	if (hardirq_count() - hardirq_offset)
2752
		index = CPUTIME_IRQ;
2753
	else if (in_serving_softirq())
2754
		index = CPUTIME_SOFTIRQ;
Linus Torvalds's avatar
Linus Torvalds committed
2755
	else
2756
		index = CPUTIME_SYSTEM;
2757

2758
	__account_system_time(p, cputime, cputime_scaled, index);
Linus Torvalds's avatar
Linus Torvalds committed
2759 2760
}

2761
/*
Linus Torvalds's avatar
Linus Torvalds committed
2762
 * Account for involuntary wait time.
2763
 * @cputime: the cpu time spent in involuntary wait
2764
 */
2765
void account_steal_time(cputime_t cputime)
2766
{
2767
	u64 *cpustat = kcpustat_this_cpu->cpustat;
2768

2769
	cpustat[CPUTIME_STEAL] += (__force u64) cputime;
2770 2771
}

Linus Torvalds's avatar
Linus Torvalds committed
2772
/*
2773 2774
 * Account for idle time.
 * @cputime: the cpu time spent in idle wait
Linus Torvalds's avatar
Linus Torvalds committed
2775
 */
2776
void account_idle_time(cputime_t cputime)
Linus Torvalds's avatar
Linus Torvalds committed
2777
{
2778
	u64 *cpustat = kcpustat_this_cpu->cpustat;
2779
	struct rq *rq = this_rq();
Linus Torvalds's avatar
Linus Torvalds committed
2780

2781
	if (atomic_read(&rq->nr_iowait) > 0)
2782
		cpustat[CPUTIME_IOWAIT] += (__force u64) cputime;
2783
	else
2784
		cpustat[CPUTIME_IDLE] += (__force u64) cputime;
Linus Torvalds's avatar
Linus Torvalds committed
2785 2786
}

2787 2788 2789
static __always_inline bool steal_account_process_tick(void)
{
#ifdef CONFIG_PARAVIRT
2790
	if (static_key_false(&paravirt_steal_enabled)) {
2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805
		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;
}

2806 2807
#ifndef CONFIG_VIRT_CPU_ACCOUNTING

2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833
#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);
2834
	u64 *cpustat = kcpustat_this_cpu->cpustat;
2835

2836 2837 2838
	if (steal_account_process_tick())
		return;

2839
	if (irqtime_account_hi_update()) {
2840
		cpustat[CPUTIME_IRQ] += (__force u64) cputime_one_jiffy;
2841
	} else if (irqtime_account_si_update()) {
2842
		cpustat[CPUTIME_SOFTIRQ] += (__force u64) cputime_one_jiffy;
2843 2844 2845 2846 2847 2848 2849
	} 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,
2850
					CPUTIME_SOFTIRQ);
2851 2852 2853 2854 2855 2856 2857 2858
	} 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,
2859
					CPUTIME_SYSTEM);
2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870
	}
}

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);
}
2871
#else /* CONFIG_IRQ_TIME_ACCOUNTING */
2872 2873 2874
static void irqtime_account_idle_ticks(int ticks) {}
static void irqtime_account_process_tick(struct task_struct *p, int user_tick,
						struct rq *rq) {}
2875
#endif /* CONFIG_IRQ_TIME_ACCOUNTING */
2876 2877 2878 2879 2880 2881 2882 2883

/*
 * 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)
{
2884
	cputime_t one_jiffy_scaled = cputime_to_scaled(cputime_one_jiffy);
2885 2886
	struct rq *rq = this_rq();

2887 2888 2889 2890 2891
	if (sched_clock_irqtime) {
		irqtime_account_process_tick(p, user_tick, rq);
		return;
	}

2892 2893 2894
	if (steal_account_process_tick())
		return;

2895
	if (user_tick)
2896
		account_user_time(p, cputime_one_jiffy, one_jiffy_scaled);
2897
	else if ((p != rq->idle) || (irq_count() != HARDIRQ_OFFSET))
2898
		account_system_time(p, HARDIRQ_OFFSET, cputime_one_jiffy,
2899 2900
				    one_jiffy_scaled);
	else
2901
		account_idle_time(cputime_one_jiffy);
2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919
}

/*
 * 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)
{
2920 2921 2922 2923 2924 2925

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

2926
	account_idle_time(jiffies_to_cputime(ticks));
Linus Torvalds's avatar
Linus Torvalds committed
2927 2928
}

2929 2930
#endif

2931 2932 2933 2934
/*
 * Use precise platform statistics if available:
 */
#ifdef CONFIG_VIRT_CPU_ACCOUNTING
2935
void task_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
2936
{
2937 2938
	*ut = p->utime;
	*st = p->stime;
2939 2940
}

2941
void thread_group_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
2942
{
2943 2944 2945 2946 2947 2948
	struct task_cputime cputime;

	thread_group_cputime(p, &cputime);

	*ut = cputime.utime;
	*st = cputime.stime;
2949 2950
}
#else
2951 2952

#ifndef nsecs_to_cputime
2953
# define nsecs_to_cputime(__nsecs)	nsecs_to_jiffies(__nsecs)
2954 2955
#endif

2956
void task_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
2957
{
2958
	cputime_t rtime, utime = p->utime, total = utime + p->stime;
2959 2960 2961 2962

	/*
	 * Use CFS's precise accounting:
	 */
2963
	rtime = nsecs_to_cputime(p->se.sum_exec_runtime);
2964 2965

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

2968 2969 2970
		temp *= (__force u64) utime;
		do_div(temp, (__force u32) total);
		utime = (__force cputime_t) temp;
2971 2972
	} else
		utime = rtime;
2973

2974 2975 2976
	/*
	 * Compare with previous values, to keep monotonicity:
	 */
2977
	p->prev_utime = max(p->prev_utime, utime);
2978
	p->prev_stime = max(p->prev_stime, rtime - p->prev_utime);
2979

2980 2981
	*ut = p->prev_utime;
	*st = p->prev_stime;
2982 2983
}

2984 2985 2986 2987
/*
 * Must be called with siglock held.
 */
void thread_group_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
2988
{
2989 2990 2991
	struct signal_struct *sig = p->signal;
	struct task_cputime cputime;
	cputime_t rtime, utime, total;
2992

2993
	thread_group_cputime(p, &cputime);
2994

2995
	total = cputime.utime + cputime.stime;
2996
	rtime = nsecs_to_cputime(cputime.sum_exec_runtime);
2997

2998
	if (total) {
2999
		u64 temp = (__force u64) rtime;
3000

3001 3002 3003
		temp *= (__force u64) cputime.utime;
		do_div(temp, (__force u32) total);
		utime = (__force cputime_t) temp;
3004 3005 3006 3007
	} else
		utime = rtime;

	sig->prev_utime = max(sig->prev_utime, utime);
3008
	sig->prev_stime = max(sig->prev_stime, rtime - sig->prev_utime);
3009 3010 3011

	*ut = sig->prev_utime;
	*st = sig->prev_stime;
3012 3013 3014
}
#endif

3015 3016 3017 3018 3019 3020 3021 3022
/*
 * 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);
Ingo Molnar's avatar
Ingo Molnar committed
3023
	struct task_struct *curr = rq->curr;
3024 3025

	sched_clock_tick();
Ingo Molnar's avatar
Ingo Molnar committed
3026

3027
	raw_spin_lock(&rq->lock);
3028
	update_rq_clock(rq);
3029
	update_cpu_load_active(rq);
Peter Zijlstra's avatar
Peter Zijlstra committed
3030
	curr->sched_class->task_tick(rq, curr, 0);
3031
	raw_spin_unlock(&rq->lock);
3032

3033
	perf_event_task_tick();
3034

3035
#ifdef CONFIG_SMP
3036
	rq->idle_balance = idle_cpu(cpu);
Ingo Molnar's avatar
Ingo Molnar committed
3037
	trigger_load_balance(rq, cpu);
3038
#endif
Linus Torvalds's avatar
Linus Torvalds committed
3039 3040
}

3041
notrace unsigned long get_parent_ip(unsigned long addr)
3042 3043 3044 3045 3046 3047 3048 3049
{
	if (in_lock_functions(addr)) {
		addr = CALLER_ADDR2;
		if (in_lock_functions(addr))
			addr = CALLER_ADDR3;
	}
	return addr;
}
Linus Torvalds's avatar
Linus Torvalds committed
3050

3051 3052 3053
#if defined(CONFIG_PREEMPT) && (defined(CONFIG_DEBUG_PREEMPT) || \
				defined(CONFIG_PREEMPT_TRACER))

3054
void __kprobes add_preempt_count(int val)
Linus Torvalds's avatar
Linus Torvalds committed
3055
{
3056
#ifdef CONFIG_DEBUG_PREEMPT
Linus Torvalds's avatar
Linus Torvalds committed
3057 3058 3059
	/*
	 * Underflow?
	 */
3060 3061
	if (DEBUG_LOCKS_WARN_ON((preempt_count() < 0)))
		return;
3062
#endif
Linus Torvalds's avatar
Linus Torvalds committed
3063
	preempt_count() += val;
3064
#ifdef CONFIG_DEBUG_PREEMPT
Linus Torvalds's avatar
Linus Torvalds committed
3065 3066 3067
	/*
	 * Spinlock count overflowing soon?
	 */
3068 3069
	DEBUG_LOCKS_WARN_ON((preempt_count() & PREEMPT_MASK) >=
				PREEMPT_MASK - 10);
3070 3071 3072
#endif
	if (preempt_count() == val)
		trace_preempt_off(CALLER_ADDR0, get_parent_ip(CALLER_ADDR1));
Linus Torvalds's avatar
Linus Torvalds committed
3073 3074 3075
}
EXPORT_SYMBOL(add_preempt_count);

3076
void __kprobes sub_preempt_count(int val)
Linus Torvalds's avatar
Linus Torvalds committed
3077
{
3078
#ifdef CONFIG_DEBUG_PREEMPT
Linus Torvalds's avatar
Linus Torvalds committed
3079 3080 3081
	/*
	 * Underflow?
	 */
3082
	if (DEBUG_LOCKS_WARN_ON(val > preempt_count()))
3083
		return;
Linus Torvalds's avatar
Linus Torvalds committed
3084 3085 3086
	/*
	 * Is the spinlock portion underflowing?
	 */
3087 3088 3089
	if (DEBUG_LOCKS_WARN_ON((val < PREEMPT_MASK) &&
			!(preempt_count() & PREEMPT_MASK)))
		return;
3090
#endif
3091

3092 3093
	if (preempt_count() == val)
		trace_preempt_on(CALLER_ADDR0, get_parent_ip(CALLER_ADDR1));
Linus Torvalds's avatar
Linus Torvalds committed
3094 3095 3096 3097 3098 3099 3100
	preempt_count() -= val;
}
EXPORT_SYMBOL(sub_preempt_count);

#endif

/*
Ingo Molnar's avatar
Ingo Molnar committed
3101
 * Print scheduling while atomic bug:
Linus Torvalds's avatar
Linus Torvalds committed
3102
 */
Ingo Molnar's avatar
Ingo Molnar committed
3103
static noinline void __schedule_bug(struct task_struct *prev)
Linus Torvalds's avatar
Linus Torvalds committed
3104
{
3105 3106 3107
	if (oops_in_progress)
		return;

3108 3109
	printk(KERN_ERR "BUG: scheduling while atomic: %s/%d/0x%08x\n",
		prev->comm, prev->pid, preempt_count());
3110

Ingo Molnar's avatar
Ingo Molnar committed
3111
	debug_show_held_locks(prev);
3112
	print_modules();
Ingo Molnar's avatar
Ingo Molnar committed
3113 3114
	if (irqs_disabled())
		print_irqtrace_events(prev);
3115
	dump_stack();
Ingo Molnar's avatar
Ingo Molnar committed
3116
}
Linus Torvalds's avatar
Linus Torvalds committed
3117

Ingo Molnar's avatar
Ingo Molnar committed
3118 3119 3120 3121 3122
/*
 * Various schedule()-time debugging checks and statistics:
 */
static inline void schedule_debug(struct task_struct *prev)
{
Linus Torvalds's avatar
Linus Torvalds committed
3123
	/*
Ingo Molnar's avatar
Ingo Molnar committed
3124
	 * Test if we are atomic. Since do_exit() needs to call into
Linus Torvalds's avatar
Linus Torvalds committed
3125 3126 3127
	 * schedule() atomically, we ignore that path for now.
	 * Otherwise, whine if we are scheduling when we should not be.
	 */
3128
	if (unlikely(in_atomic_preempt_off() && !prev->exit_state))
Ingo Molnar's avatar
Ingo Molnar committed
3129
		__schedule_bug(prev);
3130
	rcu_sleep_check();
Ingo Molnar's avatar
Ingo Molnar committed
3131

Linus Torvalds's avatar
Linus Torvalds committed
3132 3133
	profile_hit(SCHED_PROFILING, __builtin_return_address(0));

3134
	schedstat_inc(this_rq(), sched_count);
Ingo Molnar's avatar
Ingo Molnar committed
3135 3136
}

3137
static void put_prev_task(struct rq *rq, struct task_struct *prev)
3138
{
3139
	if (prev->on_rq || rq->skip_clock_update < 0)
3140
		update_rq_clock(rq);
3141
	prev->sched_class->put_prev_task(rq, prev);
3142 3143
}

Ingo Molnar's avatar
Ingo Molnar committed
3144 3145 3146 3147
/*
 * Pick up the highest-prio task:
 */
static inline struct task_struct *
3148
pick_next_task(struct rq *rq)
Ingo Molnar's avatar
Ingo Molnar committed
3149
{
3150
	const struct sched_class *class;
Ingo Molnar's avatar
Ingo Molnar committed
3151
	struct task_struct *p;
Linus Torvalds's avatar
Linus Torvalds committed
3152 3153

	/*
Ingo Molnar's avatar
Ingo Molnar committed
3154 3155
	 * Optimization: we know that if all tasks are in
	 * the fair class we can call that function directly:
Linus Torvalds's avatar
Linus Torvalds committed
3156
	 */
3157
	if (likely(rq->nr_running == rq->cfs.h_nr_running)) {
3158
		p = fair_sched_class.pick_next_task(rq);
Ingo Molnar's avatar
Ingo Molnar committed
3159 3160
		if (likely(p))
			return p;
Linus Torvalds's avatar
Linus Torvalds committed
3161 3162
	}

3163
	for_each_class(class) {
3164
		p = class->pick_next_task(rq);
Ingo Molnar's avatar
Ingo Molnar committed
3165 3166 3167
		if (p)
			return p;
	}
3168 3169

	BUG(); /* the idle class will always have a runnable task */
Ingo Molnar's avatar
Ingo Molnar committed
3170
}
Linus Torvalds's avatar
Linus Torvalds committed
3171

Ingo Molnar's avatar
Ingo Molnar committed
3172
/*
3173
 * __schedule() is the main scheduler function.
Ingo Molnar's avatar
Ingo Molnar committed
3174
 */
3175
static void __sched __schedule(void)
Ingo Molnar's avatar
Ingo Molnar committed
3176 3177
{
	struct task_struct *prev, *next;
3178
	unsigned long *switch_count;
Ingo Molnar's avatar
Ingo Molnar committed
3179
	struct rq *rq;
3180
	int cpu;
Ingo Molnar's avatar
Ingo Molnar committed
3181

3182 3183
need_resched:
	preempt_disable();
Ingo Molnar's avatar
Ingo Molnar committed
3184 3185
	cpu = smp_processor_id();
	rq = cpu_rq(cpu);
3186
	rcu_note_context_switch(cpu);
Ingo Molnar's avatar
Ingo Molnar committed
3187 3188 3189
	prev = rq->curr;

	schedule_debug(prev);
Linus Torvalds's avatar
Linus Torvalds committed
3190

3191
	if (sched_feat(HRTICK))
Mike Galbraith's avatar
Mike Galbraith committed
3192
		hrtick_clear(rq);
3193

3194
	raw_spin_lock_irq(&rq->lock);
Linus Torvalds's avatar
Linus Torvalds committed
3195

3196
	switch_count = &prev->nivcsw;
Linus Torvalds's avatar
Linus Torvalds committed
3197
	if (prev->state && !(preempt_count() & PREEMPT_ACTIVE)) {
Tejun Heo's avatar
Tejun Heo committed
3198
		if (unlikely(signal_pending_state(prev->state, prev))) {
Linus Torvalds's avatar
Linus Torvalds committed
3199
			prev->state = TASK_RUNNING;
Tejun Heo's avatar
Tejun Heo committed
3200
		} else {
3201 3202 3203
			deactivate_task(rq, prev, DEQUEUE_SLEEP);
			prev->on_rq = 0;

Tejun Heo's avatar
Tejun Heo committed
3204
			/*
3205 3206 3207
			 * If a worker went to sleep, notify and ask workqueue
			 * whether it wants to wake up a task to maintain
			 * concurrency.
Tejun Heo's avatar
Tejun Heo committed
3208 3209 3210 3211 3212 3213 3214 3215 3216
			 */
			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);
			}
		}
Ingo Molnar's avatar
Ingo Molnar committed
3217
		switch_count = &prev->nvcsw;
Linus Torvalds's avatar
Linus Torvalds committed
3218 3219
	}

3220
	pre_schedule(rq, prev);
3221

Ingo Molnar's avatar
Ingo Molnar committed
3222
	if (unlikely(!rq->nr_running))
Linus Torvalds's avatar
Linus Torvalds committed
3223 3224
		idle_balance(cpu, rq);

3225
	put_prev_task(rq, prev);
3226
	next = pick_next_task(rq);
3227 3228
	clear_tsk_need_resched(prev);
	rq->skip_clock_update = 0;
Linus Torvalds's avatar
Linus Torvalds committed
3229 3230 3231 3232 3233 3234

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

Ingo Molnar's avatar
Ingo Molnar committed
3235
		context_switch(rq, prev, next); /* unlocks the rq */
3236
		/*
3237 3238 3239 3240
		 * 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.
3241 3242 3243
		 */
		cpu = smp_processor_id();
		rq = cpu_rq(cpu);
Linus Torvalds's avatar
Linus Torvalds committed
3244
	} else
3245
		raw_spin_unlock_irq(&rq->lock);
Linus Torvalds's avatar
Linus Torvalds committed
3246

3247
	post_schedule(rq);
Linus Torvalds's avatar
Linus Torvalds committed
3248

3249
	sched_preempt_enable_no_resched();
3250
	if (need_resched())
Linus Torvalds's avatar
Linus Torvalds committed
3251 3252
		goto need_resched;
}
3253

3254 3255
static inline void sched_submit_work(struct task_struct *tsk)
{
3256
	if (!tsk->state || tsk_is_pi_blocked(tsk))
3257 3258 3259 3260 3261 3262 3263 3264 3265
		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);
}

Simon Kirby's avatar
Simon Kirby committed
3266
asmlinkage void __sched schedule(void)
3267
{
3268 3269 3270
	struct task_struct *tsk = current;

	sched_submit_work(tsk);
3271 3272
	__schedule();
}
Linus Torvalds's avatar
Linus Torvalds committed
3273 3274
EXPORT_SYMBOL(schedule);

3275 3276 3277 3278 3279 3280 3281
/**
 * schedule_preempt_disabled - called with preemption disabled
 *
 * Returns with preemption disabled. Note: preempt_count must be 1
 */
void __sched schedule_preempt_disabled(void)
{
3282
	sched_preempt_enable_no_resched();
3283 3284 3285 3286
	schedule();
	preempt_disable();
}

3287
#ifdef CONFIG_MUTEX_SPIN_ON_OWNER
3288

3289 3290 3291
static inline bool owner_running(struct mutex *lock, struct task_struct *owner)
{
	if (lock->owner != owner)
3292
		return false;
3293 3294

	/*
3295 3296 3297 3298
	 * 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.
3299
	 */
3300
	barrier();
3301

3302
	return owner->on_cpu;
3303
}
3304

3305 3306 3307 3308 3309 3310 3311 3312
/*
 * 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;
3313

3314
	rcu_read_lock();
3315 3316
	while (owner_running(lock, owner)) {
		if (need_resched())
3317
			break;
3318

3319
		arch_mutex_cpu_relax();
3320
	}
3321
	rcu_read_unlock();
3322

3323
	/*
3324 3325 3326
	 * 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.
3327
	 */
3328
	return lock->owner == NULL;
3329 3330 3331
}
#endif

Linus Torvalds's avatar
Linus Torvalds committed
3332 3333
#ifdef CONFIG_PREEMPT
/*
3334
 * this is the entry point to schedule() from in-kernel preemption
Ingo Molnar's avatar
Ingo Molnar committed
3335
 * off of preempt_enable. Kernel preemptions off return from interrupt
Linus Torvalds's avatar
Linus Torvalds committed
3336 3337
 * occur there and call schedule directly.
 */
3338
asmlinkage void __sched notrace preempt_schedule(void)
Linus Torvalds's avatar
Linus Torvalds committed
3339 3340
{
	struct thread_info *ti = current_thread_info();
3341

Linus Torvalds's avatar
Linus Torvalds committed
3342 3343
	/*
	 * If there is a non-zero preempt_count or interrupts are disabled,
Ingo Molnar's avatar
Ingo Molnar committed
3344
	 * we do not want to preempt the current task. Just return..
Linus Torvalds's avatar
Linus Torvalds committed
3345
	 */
3346
	if (likely(ti->preempt_count || irqs_disabled()))
Linus Torvalds's avatar
Linus Torvalds committed
3347 3348
		return;

3349
	do {
3350
		add_preempt_count_notrace(PREEMPT_ACTIVE);
3351
		__schedule();
3352
		sub_preempt_count_notrace(PREEMPT_ACTIVE);
Linus Torvalds's avatar
Linus Torvalds committed
3353

3354 3355 3356 3357 3358
		/*
		 * Check again in case we missed a preemption opportunity
		 * between schedule and now.
		 */
		barrier();
3359
	} while (need_resched());
Linus Torvalds's avatar
Linus Torvalds committed
3360 3361 3362 3363
}
EXPORT_SYMBOL(preempt_schedule);

/*
3364
 * this is the entry point to schedule() from kernel preemption
Linus Torvalds's avatar
Linus Torvalds committed
3365 3366 3367 3368 3369 3370 3371
 * 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();
3372

3373
	/* Catch callers which need to be fixed */
Linus Torvalds's avatar
Linus Torvalds committed
3374 3375
	BUG_ON(ti->preempt_count || !irqs_disabled());

3376 3377 3378
	do {
		add_preempt_count(PREEMPT_ACTIVE);
		local_irq_enable();
3379
		__schedule();
3380 3381
		local_irq_disable();
		sub_preempt_count(PREEMPT_ACTIVE);
Linus Torvalds's avatar
Linus Torvalds committed
3382

3383 3384 3385 3386 3387
		/*
		 * Check again in case we missed a preemption opportunity
		 * between schedule and now.
		 */
		barrier();
3388
	} while (need_resched());
Linus Torvalds's avatar
Linus Torvalds committed
3389 3390 3391 3392
}

#endif /* CONFIG_PREEMPT */

3393
int default_wake_function(wait_queue_t *curr, unsigned mode, int wake_flags,
Ingo Molnar's avatar
Ingo Molnar committed
3394
			  void *key)
Linus Torvalds's avatar
Linus Torvalds committed
3395
{
3396
	return try_to_wake_up(curr->private, mode, wake_flags);
Linus Torvalds's avatar
Linus Torvalds committed
3397 3398 3399 3400
}
EXPORT_SYMBOL(default_wake_function);

/*
Ingo Molnar's avatar
Ingo Molnar committed
3401 3402
 * The core wakeup function. Non-exclusive wakeups (nr_exclusive == 0) just
 * wake everything up. If it's an exclusive wakeup (nr_exclusive == small +ve
Linus Torvalds's avatar
Linus Torvalds committed
3403 3404 3405
 * 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
Ingo Molnar's avatar
Ingo Molnar committed
3406
 * started to run but is not in state TASK_RUNNING. try_to_wake_up() returns
Linus Torvalds's avatar
Linus Torvalds committed
3407 3408
 * zero in this (rare) case, and we handle it by continuing to scan the queue.
 */
3409
static void __wake_up_common(wait_queue_head_t *q, unsigned int mode,
3410
			int nr_exclusive, int wake_flags, void *key)
Linus Torvalds's avatar
Linus Torvalds committed
3411
{
3412
	wait_queue_t *curr, *next;
Linus Torvalds's avatar
Linus Torvalds committed
3413

3414
	list_for_each_entry_safe(curr, next, &q->task_list, task_list) {
3415 3416
		unsigned flags = curr->flags;

3417
		if (curr->func(curr, mode, wake_flags, key) &&
3418
				(flags & WQ_FLAG_EXCLUSIVE) && !--nr_exclusive)
Linus Torvalds's avatar
Linus Torvalds committed
3419 3420 3421 3422 3423 3424 3425 3426 3427
			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
3428
 * @key: is directly passed to the wakeup function
3429 3430 3431
 *
 * 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.
Linus Torvalds's avatar
Linus Torvalds committed
3432
 */
3433
void __wake_up(wait_queue_head_t *q, unsigned int mode,
Ingo Molnar's avatar
Ingo Molnar committed
3434
			int nr_exclusive, void *key)
Linus Torvalds's avatar
Linus Torvalds committed
3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446
{
	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.
 */
3447
void __wake_up_locked(wait_queue_head_t *q, unsigned int mode, int nr)
Linus Torvalds's avatar
Linus Torvalds committed
3448
{
3449
	__wake_up_common(q, mode, nr, 0, NULL);
Linus Torvalds's avatar
Linus Torvalds committed
3450
}
3451
EXPORT_SYMBOL_GPL(__wake_up_locked);
Linus Torvalds's avatar
Linus Torvalds committed
3452

3453 3454 3455 3456
void __wake_up_locked_key(wait_queue_head_t *q, unsigned int mode, void *key)
{
	__wake_up_common(q, mode, 1, 0, key);
}
3457
EXPORT_SYMBOL_GPL(__wake_up_locked_key);
3458

Linus Torvalds's avatar
Linus Torvalds committed
3459
/**
3460
 * __wake_up_sync_key - wake up threads blocked on a waitqueue.
Linus Torvalds's avatar
Linus Torvalds committed
3461 3462 3463
 * @q: the waitqueue
 * @mode: which threads
 * @nr_exclusive: how many wake-one or wake-many threads to wake up
3464
 * @key: opaque value to be passed to wakeup targets
Linus Torvalds's avatar
Linus Torvalds committed
3465 3466 3467 3468 3469 3470 3471
 *
 * 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.
3472 3473 3474
 *
 * 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.
Linus Torvalds's avatar
Linus Torvalds committed
3475
 */
3476 3477
void __wake_up_sync_key(wait_queue_head_t *q, unsigned int mode,
			int nr_exclusive, void *key)
Linus Torvalds's avatar
Linus Torvalds committed
3478 3479
{
	unsigned long flags;
3480
	int wake_flags = WF_SYNC;
Linus Torvalds's avatar
Linus Torvalds committed
3481 3482 3483 3484 3485

	if (unlikely(!q))
		return;

	if (unlikely(!nr_exclusive))
3486
		wake_flags = 0;
Linus Torvalds's avatar
Linus Torvalds committed
3487 3488

	spin_lock_irqsave(&q->lock, flags);
3489
	__wake_up_common(q, mode, nr_exclusive, wake_flags, key);
Linus Torvalds's avatar
Linus Torvalds committed
3490 3491
	spin_unlock_irqrestore(&q->lock, flags);
}
3492 3493 3494 3495 3496 3497 3498 3499 3500
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);
}
Linus Torvalds's avatar
Linus Torvalds committed
3501 3502
EXPORT_SYMBOL_GPL(__wake_up_sync);	/* For internal use only */

3503 3504 3505 3506 3507 3508 3509 3510
/**
 * 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.
3511 3512 3513
 *
 * 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.
3514
 */
3515
void complete(struct completion *x)
Linus Torvalds's avatar
Linus Torvalds committed
3516 3517 3518 3519 3520
{
	unsigned long flags;

	spin_lock_irqsave(&x->wait.lock, flags);
	x->done++;
3521
	__wake_up_common(&x->wait, TASK_NORMAL, 1, 0, NULL);
Linus Torvalds's avatar
Linus Torvalds committed
3522 3523 3524 3525
	spin_unlock_irqrestore(&x->wait.lock, flags);
}
EXPORT_SYMBOL(complete);

3526 3527 3528 3529 3530
/**
 * 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.
3531 3532 3533
 *
 * 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.
3534
 */
3535
void complete_all(struct completion *x)
Linus Torvalds's avatar
Linus Torvalds committed
3536 3537 3538 3539 3540
{
	unsigned long flags;

	spin_lock_irqsave(&x->wait.lock, flags);
	x->done += UINT_MAX/2;
3541
	__wake_up_common(&x->wait, TASK_NORMAL, 0, 0, NULL);
Linus Torvalds's avatar
Linus Torvalds committed
3542 3543 3544 3545
	spin_unlock_irqrestore(&x->wait.lock, flags);
}
EXPORT_SYMBOL(complete_all);

3546 3547
static inline long __sched
do_wait_for_common(struct completion *x, long timeout, int state)
Linus Torvalds's avatar
Linus Torvalds committed
3548 3549 3550 3551
{
	if (!x->done) {
		DECLARE_WAITQUEUE(wait, current);

3552
		__add_wait_queue_tail_exclusive(&x->wait, &wait);
Linus Torvalds's avatar
Linus Torvalds committed
3553
		do {
3554
			if (signal_pending_state(state, current)) {
3555 3556
				timeout = -ERESTARTSYS;
				break;
3557 3558
			}
			__set_current_state(state);
Linus Torvalds's avatar
Linus Torvalds committed
3559 3560 3561
			spin_unlock_irq(&x->wait.lock);
			timeout = schedule_timeout(timeout);
			spin_lock_irq(&x->wait.lock);
3562
		} while (!x->done && timeout);
Linus Torvalds's avatar
Linus Torvalds committed
3563
		__remove_wait_queue(&x->wait, &wait);
3564 3565
		if (!x->done)
			return timeout;
Linus Torvalds's avatar
Linus Torvalds committed
3566 3567
	}
	x->done--;
3568
	return timeout ?: 1;
Linus Torvalds's avatar
Linus Torvalds committed
3569 3570
}

3571 3572
static long __sched
wait_for_common(struct completion *x, long timeout, int state)
Linus Torvalds's avatar
Linus Torvalds committed
3573 3574 3575 3576
{
	might_sleep();

	spin_lock_irq(&x->wait.lock);
3577
	timeout = do_wait_for_common(x, timeout, state);
Linus Torvalds's avatar
Linus Torvalds committed
3578
	spin_unlock_irq(&x->wait.lock);
3579 3580
	return timeout;
}
Linus Torvalds's avatar
Linus Torvalds committed
3581

3582 3583 3584 3585 3586 3587 3588 3589 3590 3591
/**
 * 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().
 */
3592
void __sched wait_for_completion(struct completion *x)
3593 3594
{
	wait_for_common(x, MAX_SCHEDULE_TIMEOUT, TASK_UNINTERRUPTIBLE);
Linus Torvalds's avatar
Linus Torvalds committed
3595
}
3596
EXPORT_SYMBOL(wait_for_completion);
Linus Torvalds's avatar
Linus Torvalds committed
3597

3598 3599 3600 3601 3602 3603 3604 3605
/**
 * 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.
3606 3607 3608
 *
 * The return value is 0 if timed out, and positive (at least 1, or number of
 * jiffies left till timeout) if completed.
3609
 */
3610
unsigned long __sched
3611
wait_for_completion_timeout(struct completion *x, unsigned long timeout)
Linus Torvalds's avatar
Linus Torvalds committed
3612
{
3613
	return wait_for_common(x, timeout, TASK_UNINTERRUPTIBLE);
Linus Torvalds's avatar
Linus Torvalds committed
3614
}
3615
EXPORT_SYMBOL(wait_for_completion_timeout);
Linus Torvalds's avatar
Linus Torvalds committed
3616

3617 3618 3619 3620 3621 3622
/**
 * 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.
3623 3624
 *
 * The return value is -ERESTARTSYS if interrupted, 0 if completed.
3625
 */
3626
int __sched wait_for_completion_interruptible(struct completion *x)
3627
{
3628 3629 3630 3631
	long t = wait_for_common(x, MAX_SCHEDULE_TIMEOUT, TASK_INTERRUPTIBLE);
	if (t == -ERESTARTSYS)
		return t;
	return 0;
3632
}
3633
EXPORT_SYMBOL(wait_for_completion_interruptible);
Linus Torvalds's avatar
Linus Torvalds committed
3634

3635 3636 3637 3638 3639 3640 3641
/**
 * 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.
3642 3643 3644
 *
 * The return value is -ERESTARTSYS if interrupted, 0 if timed out,
 * positive (at least 1, or number of jiffies left till timeout) if completed.
3645
 */
3646
long __sched
3647 3648
wait_for_completion_interruptible_timeout(struct completion *x,
					  unsigned long timeout)
3649
{
3650
	return wait_for_common(x, timeout, TASK_INTERRUPTIBLE);
3651
}
3652
EXPORT_SYMBOL(wait_for_completion_interruptible_timeout);
Linus Torvalds's avatar
Linus Torvalds committed
3653

3654 3655 3656 3657 3658 3659
/**
 * 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.
3660 3661
 *
 * The return value is -ERESTARTSYS if interrupted, 0 if completed.
3662
 */
3663 3664 3665 3666 3667 3668 3669 3670 3671
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);

3672 3673 3674 3675 3676 3677 3678 3679
/**
 * 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.
3680 3681 3682
 *
 * The return value is -ERESTARTSYS if interrupted, 0 if timed out,
 * positive (at least 1, or number of jiffies left till timeout) if completed.
3683
 */
3684
long __sched
3685 3686 3687 3688 3689 3690 3691
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);

3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705
/**
 *	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)
{
3706
	unsigned long flags;
3707 3708
	int ret = 1;

3709
	spin_lock_irqsave(&x->wait.lock, flags);
3710 3711 3712 3713
	if (!x->done)
		ret = 0;
	else
		x->done--;
3714
	spin_unlock_irqrestore(&x->wait.lock, flags);
3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728
	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)
{
3729
	unsigned long flags;
3730 3731
	int ret = 1;

3732
	spin_lock_irqsave(&x->wait.lock, flags);
3733 3734
	if (!x->done)
		ret = 0;
3735
	spin_unlock_irqrestore(&x->wait.lock, flags);
3736 3737 3738 3739
	return ret;
}
EXPORT_SYMBOL(completion_done);

3740 3741
static long __sched
sleep_on_common(wait_queue_head_t *q, int state, long timeout)
Linus Torvalds's avatar
Linus Torvalds committed
3742
{
3743 3744 3745 3746
	unsigned long flags;
	wait_queue_t wait;

	init_waitqueue_entry(&wait, current);
Linus Torvalds's avatar
Linus Torvalds committed
3747

3748
	__set_current_state(state);
Linus Torvalds's avatar
Linus Torvalds committed
3749

3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763
	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);
Linus Torvalds's avatar
Linus Torvalds committed
3764 3765 3766
}
EXPORT_SYMBOL(interruptible_sleep_on);

3767
long __sched
Ingo Molnar's avatar
Ingo Molnar committed
3768
interruptible_sleep_on_timeout(wait_queue_head_t *q, long timeout)
Linus Torvalds's avatar
Linus Torvalds committed
3769
{
3770
	return sleep_on_common(q, TASK_INTERRUPTIBLE, timeout);
Linus Torvalds's avatar
Linus Torvalds committed
3771 3772 3773
}
EXPORT_SYMBOL(interruptible_sleep_on_timeout);

3774
void __sched sleep_on(wait_queue_head_t *q)
Linus Torvalds's avatar
Linus Torvalds committed
3775
{
3776
	sleep_on_common(q, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
Linus Torvalds's avatar
Linus Torvalds committed
3777 3778 3779
}
EXPORT_SYMBOL(sleep_on);

3780
long __sched sleep_on_timeout(wait_queue_head_t *q, long timeout)
Linus Torvalds's avatar
Linus Torvalds committed
3781
{
3782
	return sleep_on_common(q, TASK_UNINTERRUPTIBLE, timeout);
Linus Torvalds's avatar
Linus Torvalds committed
3783 3784 3785
}
EXPORT_SYMBOL(sleep_on_timeout);

3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797
#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.
 */
3798
void rt_mutex_setprio(struct task_struct *p, int prio)
3799
{
3800
	int oldprio, on_rq, running;
3801
	struct rq *rq;
3802
	const struct sched_class *prev_class;
3803 3804 3805

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

3806
	rq = __task_rq_lock(p);
3807

3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825
	/*
	 * 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;
	}

3826
	trace_sched_pi_setprio(p, prio);
3827
	oldprio = p->prio;
3828
	prev_class = p->sched_class;
Peter Zijlstra's avatar
Peter Zijlstra committed
3829
	on_rq = p->on_rq;
3830
	running = task_current(rq, p);
3831
	if (on_rq)
3832
		dequeue_task(rq, p, 0);
3833 3834
	if (running)
		p->sched_class->put_prev_task(rq, p);
Ingo Molnar's avatar
Ingo Molnar committed
3835 3836 3837 3838 3839 3840

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

3841 3842
	p->prio = prio;

3843 3844
	if (running)
		p->sched_class->set_curr_task(rq);
3845
	if (on_rq)
3846
		enqueue_task(rq, p, oldprio < prio ? ENQUEUE_HEAD : 0);
3847

3848
	check_class_changed(rq, p, prev_class, oldprio);
3849
out_unlock:
3850
	__task_rq_unlock(rq);
3851 3852
}
#endif
3853
void set_user_nice(struct task_struct *p, long nice)
Linus Torvalds's avatar
Linus Torvalds committed
3854
{
Ingo Molnar's avatar
Ingo Molnar committed
3855
	int old_prio, delta, on_rq;
Linus Torvalds's avatar
Linus Torvalds committed
3856
	unsigned long flags;
3857
	struct rq *rq;
Linus Torvalds's avatar
Linus Torvalds committed
3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869

	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
Ingo Molnar's avatar
Ingo Molnar committed
3870
	 * SCHED_FIFO/SCHED_RR:
Linus Torvalds's avatar
Linus Torvalds committed
3871
	 */
3872
	if (task_has_rt_policy(p)) {
Linus Torvalds's avatar
Linus Torvalds committed
3873 3874 3875
		p->static_prio = NICE_TO_PRIO(nice);
		goto out_unlock;
	}
Peter Zijlstra's avatar
Peter Zijlstra committed
3876
	on_rq = p->on_rq;
3877
	if (on_rq)
3878
		dequeue_task(rq, p, 0);
Linus Torvalds's avatar
Linus Torvalds committed
3879 3880

	p->static_prio = NICE_TO_PRIO(nice);
3881
	set_load_weight(p);
3882 3883 3884
	old_prio = p->prio;
	p->prio = effective_prio(p);
	delta = p->prio - old_prio;
Linus Torvalds's avatar
Linus Torvalds committed
3885

Ingo Molnar's avatar
Ingo Molnar committed
3886
	if (on_rq) {
3887
		enqueue_task(rq, p, 0);
Linus Torvalds's avatar
Linus Torvalds committed
3888
		/*
3889 3890
		 * If the task increased its priority or is running and
		 * lowered its priority, then reschedule its CPU:
Linus Torvalds's avatar
Linus Torvalds committed
3891
		 */
3892
		if (delta < 0 || (delta > 0 && task_running(rq, p)))
Linus Torvalds's avatar
Linus Torvalds committed
3893 3894 3895
			resched_task(rq->curr);
	}
out_unlock:
3896
	task_rq_unlock(rq, p, &flags);
Linus Torvalds's avatar
Linus Torvalds committed
3897 3898 3899
}
EXPORT_SYMBOL(set_user_nice);

3900 3901 3902 3903 3904
/*
 * can_nice - check if a task can reduce its nice value
 * @p: task
 * @nice: nice value
 */
3905
int can_nice(const struct task_struct *p, const int nice)
3906
{
3907 3908
	/* convert nice value [19,-20] to rlimit style value [1,40] */
	int nice_rlim = 20 - nice;
3909

3910
	return (nice_rlim <= task_rlimit(p, RLIMIT_NICE) ||
3911 3912 3913
		capable(CAP_SYS_NICE));
}

Linus Torvalds's avatar
Linus Torvalds committed
3914 3915 3916 3917 3918 3919 3920 3921 3922
#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.
 */
3923
SYSCALL_DEFINE1(nice, int, increment)
Linus Torvalds's avatar
Linus Torvalds committed
3924
{
3925
	long nice, retval;
Linus Torvalds's avatar
Linus Torvalds committed
3926 3927 3928 3929 3930 3931

	/*
	 * 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.
	 */
3932 3933
	if (increment < -40)
		increment = -40;
Linus Torvalds's avatar
Linus Torvalds committed
3934 3935 3936
	if (increment > 40)
		increment = 40;

3937
	nice = TASK_NICE(current) + increment;
Linus Torvalds's avatar
Linus Torvalds committed
3938 3939 3940 3941 3942
	if (nice < -20)
		nice = -20;
	if (nice > 19)
		nice = 19;

3943 3944 3945
	if (increment < 0 && !can_nice(current, nice))
		return -EPERM;

Linus Torvalds's avatar
Linus Torvalds committed
3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963
	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.
 */
3964
int task_prio(const struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
3965 3966 3967 3968 3969 3970 3971 3972
{
	return p->prio - MAX_RT_PRIO;
}

/**
 * task_nice - return the nice value of a given task.
 * @p: the task in question.
 */
3973
int task_nice(const struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
3974 3975 3976
{
	return TASK_NICE(p);
}
Pavel Roskin's avatar
Pavel Roskin committed
3977
EXPORT_SYMBOL(task_nice);
Linus Torvalds's avatar
Linus Torvalds committed
3978 3979 3980 3981 3982 3983 3984

/**
 * idle_cpu - is a given cpu idle currently?
 * @cpu: the processor in question.
 */
int idle_cpu(int cpu)
{
Thomas Gleixner's avatar
Thomas Gleixner committed
3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998
	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;
Linus Torvalds's avatar
Linus Torvalds committed
3999 4000 4001 4002 4003 4004
}

/**
 * idle_task - return the idle task for a given cpu.
 * @cpu: the processor in question.
 */
4005
struct task_struct *idle_task(int cpu)
Linus Torvalds's avatar
Linus Torvalds committed
4006 4007 4008 4009 4010 4011 4012 4013
{
	return cpu_rq(cpu)->idle;
}

/**
 * find_process_by_pid - find a process with a matching PID value.
 * @pid: the pid in question.
 */
Alexey Dobriyan's avatar
Alexey Dobriyan committed
4014
static struct task_struct *find_process_by_pid(pid_t pid)
Linus Torvalds's avatar
Linus Torvalds committed
4015
{
4016
	return pid ? find_task_by_vpid(pid) : current;
Linus Torvalds's avatar
Linus Torvalds committed
4017 4018 4019
}

/* Actually do priority change: must hold rq lock. */
Ingo Molnar's avatar
Ingo Molnar committed
4020 4021
static void
__setscheduler(struct rq *rq, struct task_struct *p, int policy, int prio)
Linus Torvalds's avatar
Linus Torvalds committed
4022 4023 4024
{
	p->policy = policy;
	p->rt_priority = prio;
4025 4026 4027
	p->normal_prio = normal_prio(p);
	/* we are holding p->pi_lock already */
	p->prio = rt_mutex_getprio(p);
4028 4029 4030 4031
	if (rt_prio(p->prio))
		p->sched_class = &rt_sched_class;
	else
		p->sched_class = &fair_sched_class;
4032
	set_load_weight(p);
Linus Torvalds's avatar
Linus Torvalds committed
4033 4034
}

4035 4036 4037 4038 4039 4040 4041 4042 4043 4044
/*
 * 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);
4045
	if (cred->user_ns == pcred->user_ns)
4046 4047 4048 4049
		match = (cred->euid == pcred->euid ||
			 cred->euid == pcred->uid);
	else
		match = false;
4050 4051 4052 4053
	rcu_read_unlock();
	return match;
}

4054
static int __sched_setscheduler(struct task_struct *p, int policy,
4055
				const struct sched_param *param, bool user)
Linus Torvalds's avatar
Linus Torvalds committed
4056
{
4057
	int retval, oldprio, oldpolicy = -1, on_rq, running;
Linus Torvalds's avatar
Linus Torvalds committed
4058
	unsigned long flags;
4059
	const struct sched_class *prev_class;
4060
	struct rq *rq;
4061
	int reset_on_fork;
Linus Torvalds's avatar
Linus Torvalds committed
4062

4063 4064
	/* may grab non-irq protected spin_locks */
	BUG_ON(in_interrupt());
Linus Torvalds's avatar
Linus Torvalds committed
4065 4066
recheck:
	/* double check policy once rq lock held */
4067 4068
	if (policy < 0) {
		reset_on_fork = p->sched_reset_on_fork;
Linus Torvalds's avatar
Linus Torvalds committed
4069
		policy = oldpolicy = p->policy;
4070 4071 4072 4073 4074 4075 4076 4077 4078 4079
	} 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;
	}

Linus Torvalds's avatar
Linus Torvalds committed
4080 4081
	/*
	 * Valid priorities for SCHED_FIFO and SCHED_RR are
Ingo Molnar's avatar
Ingo Molnar committed
4082 4083
	 * 1..MAX_USER_RT_PRIO-1, valid priority for SCHED_NORMAL,
	 * SCHED_BATCH and SCHED_IDLE is 0.
Linus Torvalds's avatar
Linus Torvalds committed
4084 4085
	 */
	if (param->sched_priority < 0 ||
Ingo Molnar's avatar
Ingo Molnar committed
4086
	    (p->mm && param->sched_priority > MAX_USER_RT_PRIO-1) ||
4087
	    (!p->mm && param->sched_priority > MAX_RT_PRIO-1))
Linus Torvalds's avatar
Linus Torvalds committed
4088
		return -EINVAL;
4089
	if (rt_policy(policy) != (param->sched_priority != 0))
Linus Torvalds's avatar
Linus Torvalds committed
4090 4091
		return -EINVAL;

4092 4093 4094
	/*
	 * Allow unprivileged RT tasks to decrease priority:
	 */
4095
	if (user && !capable(CAP_SYS_NICE)) {
4096
		if (rt_policy(policy)) {
4097 4098
			unsigned long rlim_rtprio =
					task_rlimit(p, RLIMIT_RTPRIO);
4099 4100 4101 4102 4103 4104 4105 4106 4107 4108

			/* 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;
		}
4109

Ingo Molnar's avatar
Ingo Molnar committed
4110
		/*
4111 4112
		 * Treat SCHED_IDLE as nice 20. Only allow a switch to
		 * SCHED_NORMAL if the RLIMIT_NICE would normally permit it.
Ingo Molnar's avatar
Ingo Molnar committed
4113
		 */
4114 4115 4116 4117
		if (p->policy == SCHED_IDLE && policy != SCHED_IDLE) {
			if (!can_nice(p, TASK_NICE(p)))
				return -EPERM;
		}
4118

4119
		/* can't change other user's priorities */
4120
		if (!check_same_owner(p))
4121
			return -EPERM;
4122 4123 4124 4125

		/* Normal users shall not reset the sched_reset_on_fork flag */
		if (p->sched_reset_on_fork && !reset_on_fork)
			return -EPERM;
4126
	}
Linus Torvalds's avatar
Linus Torvalds committed
4127

4128
	if (user) {
4129
		retval = security_task_setscheduler(p);
4130 4131 4132 4133
		if (retval)
			return retval;
	}

4134 4135 4136
	/*
	 * make sure no PI-waiters arrive (or leave) while we are
	 * changing the priority of the task:
4137
	 *
Lucas De Marchi's avatar
Lucas De Marchi committed
4138
	 * To be able to change p->policy safely, the appropriate
Linus Torvalds's avatar
Linus Torvalds committed
4139 4140
	 * runqueue lock must be held.
	 */
4141
	rq = task_rq_lock(p, &flags);
4142

4143 4144 4145 4146
	/*
	 * Changing the policy of the stop threads its a very bad idea
	 */
	if (p == rq->stop) {
4147
		task_rq_unlock(rq, p, &flags);
4148 4149 4150
		return -EINVAL;
	}

4151 4152 4153 4154 4155 4156 4157 4158 4159 4160 4161
	/*
	 * If not changing anything there's no need to proceed further:
	 */
	if (unlikely(policy == p->policy && (!rt_policy(policy) ||
			param->sched_priority == p->rt_priority))) {

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

4162 4163 4164 4165 4166 4167 4168
#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) &&
4169 4170
				task_group(p)->rt_bandwidth.rt_runtime == 0 &&
				!task_group_is_autogroup(task_group(p))) {
4171
			task_rq_unlock(rq, p, &flags);
4172 4173 4174 4175 4176
			return -EPERM;
		}
	}
#endif

Linus Torvalds's avatar
Linus Torvalds committed
4177 4178 4179
	/* recheck policy now with rq lock held */
	if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) {
		policy = oldpolicy = -1;
4180
		task_rq_unlock(rq, p, &flags);
Linus Torvalds's avatar
Linus Torvalds committed
4181 4182
		goto recheck;
	}
Peter Zijlstra's avatar
Peter Zijlstra committed
4183
	on_rq = p->on_rq;
4184
	running = task_current(rq, p);
4185
	if (on_rq)
4186
		dequeue_task(rq, p, 0);
4187 4188
	if (running)
		p->sched_class->put_prev_task(rq, p);
4189

4190 4191
	p->sched_reset_on_fork = reset_on_fork;

Linus Torvalds's avatar
Linus Torvalds committed
4192
	oldprio = p->prio;
4193
	prev_class = p->sched_class;
Ingo Molnar's avatar
Ingo Molnar committed
4194
	__setscheduler(rq, p, policy, param->sched_priority);
4195

4196 4197
	if (running)
		p->sched_class->set_curr_task(rq);
4198
	if (on_rq)
4199
		enqueue_task(rq, p, 0);
4200

4201
	check_class_changed(rq, p, prev_class, oldprio);
4202
	task_rq_unlock(rq, p, &flags);
4203

4204 4205
	rt_mutex_adjust_pi(p);

Linus Torvalds's avatar
Linus Torvalds committed
4206 4207
	return 0;
}
4208 4209 4210 4211 4212 4213 4214 4215 4216 4217

/**
 * 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,
4218
		       const struct sched_param *param)
4219 4220 4221
{
	return __sched_setscheduler(p, policy, param, true);
}
Linus Torvalds's avatar
Linus Torvalds committed
4222 4223
EXPORT_SYMBOL_GPL(sched_setscheduler);

4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235
/**
 * 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,
4236
			       const struct sched_param *param)
4237 4238 4239 4240
{
	return __sched_setscheduler(p, policy, param, false);
}

Ingo Molnar's avatar
Ingo Molnar committed
4241 4242
static int
do_sched_setscheduler(pid_t pid, int policy, struct sched_param __user *param)
Linus Torvalds's avatar
Linus Torvalds committed
4243 4244 4245
{
	struct sched_param lparam;
	struct task_struct *p;
4246
	int retval;
Linus Torvalds's avatar
Linus Torvalds committed
4247 4248 4249 4250 4251

	if (!param || pid < 0)
		return -EINVAL;
	if (copy_from_user(&lparam, param, sizeof(struct sched_param)))
		return -EFAULT;
4252 4253 4254

	rcu_read_lock();
	retval = -ESRCH;
Linus Torvalds's avatar
Linus Torvalds committed
4255
	p = find_process_by_pid(pid);
4256 4257 4258
	if (p != NULL)
		retval = sched_setscheduler(p, policy, &lparam);
	rcu_read_unlock();
4259

Linus Torvalds's avatar
Linus Torvalds committed
4260 4261 4262 4263 4264 4265 4266 4267 4268
	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.
 */
4269 4270
SYSCALL_DEFINE3(sched_setscheduler, pid_t, pid, int, policy,
		struct sched_param __user *, param)
Linus Torvalds's avatar
Linus Torvalds committed
4271
{
4272 4273 4274 4275
	/* negative values for policy are not valid */
	if (policy < 0)
		return -EINVAL;

Linus Torvalds's avatar
Linus Torvalds committed
4276 4277 4278 4279 4280 4281 4282 4283
	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.
 */
4284
SYSCALL_DEFINE2(sched_setparam, pid_t, pid, struct sched_param __user *, param)
Linus Torvalds's avatar
Linus Torvalds committed
4285 4286 4287 4288 4289 4290 4291 4292
{
	return do_sched_setscheduler(pid, -1, param);
}

/**
 * sys_sched_getscheduler - get the policy (scheduling class) of a thread
 * @pid: the pid in question.
 */
4293
SYSCALL_DEFINE1(sched_getscheduler, pid_t, pid)
Linus Torvalds's avatar
Linus Torvalds committed
4294
{
4295
	struct task_struct *p;
4296
	int retval;
Linus Torvalds's avatar
Linus Torvalds committed
4297 4298

	if (pid < 0)
4299
		return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
4300 4301

	retval = -ESRCH;
4302
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
4303 4304 4305 4306
	p = find_process_by_pid(pid);
	if (p) {
		retval = security_task_getscheduler(p);
		if (!retval)
4307 4308
			retval = p->policy
				| (p->sched_reset_on_fork ? SCHED_RESET_ON_FORK : 0);
Linus Torvalds's avatar
Linus Torvalds committed
4309
	}
4310
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
4311 4312 4313 4314
	return retval;
}

/**
4315
 * sys_sched_getparam - get the RT priority of a thread
Linus Torvalds's avatar
Linus Torvalds committed
4316 4317 4318
 * @pid: the pid in question.
 * @param: structure containing the RT priority.
 */
4319
SYSCALL_DEFINE2(sched_getparam, pid_t, pid, struct sched_param __user *, param)
Linus Torvalds's avatar
Linus Torvalds committed
4320 4321
{
	struct sched_param lp;
4322
	struct task_struct *p;
4323
	int retval;
Linus Torvalds's avatar
Linus Torvalds committed
4324 4325

	if (!param || pid < 0)
4326
		return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
4327

4328
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
4329 4330 4331 4332 4333 4334 4335 4336 4337 4338
	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;
4339
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
4340 4341 4342 4343 4344 4345 4346 4347 4348

	/*
	 * 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:
4349
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
4350 4351 4352
	return retval;
}

4353
long sched_setaffinity(pid_t pid, const struct cpumask *in_mask)
Linus Torvalds's avatar
Linus Torvalds committed
4354
{
4355
	cpumask_var_t cpus_allowed, new_mask;
4356 4357
	struct task_struct *p;
	int retval;
Linus Torvalds's avatar
Linus Torvalds committed
4358

4359
	get_online_cpus();
4360
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
4361 4362 4363

	p = find_process_by_pid(pid);
	if (!p) {
4364
		rcu_read_unlock();
4365
		put_online_cpus();
Linus Torvalds's avatar
Linus Torvalds committed
4366 4367 4368
		return -ESRCH;
	}

4369
	/* Prevent p going away */
Linus Torvalds's avatar
Linus Torvalds committed
4370
	get_task_struct(p);
4371
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
4372

4373 4374 4375 4376 4377 4378 4379 4380
	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;
	}
Linus Torvalds's avatar
Linus Torvalds committed
4381
	retval = -EPERM;
4382
	if (!check_same_owner(p) && !ns_capable(task_user_ns(p), CAP_SYS_NICE))
Linus Torvalds's avatar
Linus Torvalds committed
4383 4384
		goto out_unlock;

4385
	retval = security_task_setscheduler(p);
4386 4387 4388
	if (retval)
		goto out_unlock;

4389 4390
	cpuset_cpus_allowed(p, cpus_allowed);
	cpumask_and(new_mask, in_mask, cpus_allowed);
Peter Zijlstra's avatar
Peter Zijlstra committed
4391
again:
4392
	retval = set_cpus_allowed_ptr(p, new_mask);
Linus Torvalds's avatar
Linus Torvalds committed
4393

Paul Menage's avatar
Paul Menage committed
4394
	if (!retval) {
4395 4396
		cpuset_cpus_allowed(p, cpus_allowed);
		if (!cpumask_subset(new_mask, cpus_allowed)) {
Paul Menage's avatar
Paul Menage committed
4397 4398 4399 4400 4401
			/*
			 * We must have raced with a concurrent cpuset
			 * update. Just reset the cpus_allowed to the
			 * cpuset's cpus_allowed
			 */
4402
			cpumask_copy(new_mask, cpus_allowed);
Paul Menage's avatar
Paul Menage committed
4403 4404 4405
			goto again;
		}
	}
Linus Torvalds's avatar
Linus Torvalds committed
4406
out_unlock:
4407 4408 4409 4410
	free_cpumask_var(new_mask);
out_free_cpus_allowed:
	free_cpumask_var(cpus_allowed);
out_put_task:
Linus Torvalds's avatar
Linus Torvalds committed
4411
	put_task_struct(p);
4412
	put_online_cpus();
Linus Torvalds's avatar
Linus Torvalds committed
4413 4414 4415 4416
	return retval;
}

static int get_user_cpu_mask(unsigned long __user *user_mask_ptr, unsigned len,
4417
			     struct cpumask *new_mask)
Linus Torvalds's avatar
Linus Torvalds committed
4418
{
4419 4420 4421 4422 4423
	if (len < cpumask_size())
		cpumask_clear(new_mask);
	else if (len > cpumask_size())
		len = cpumask_size();

Linus Torvalds's avatar
Linus Torvalds committed
4424 4425 4426 4427 4428 4429 4430 4431 4432
	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
 */
4433 4434
SYSCALL_DEFINE3(sched_setaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
Linus Torvalds's avatar
Linus Torvalds committed
4435
{
4436
	cpumask_var_t new_mask;
Linus Torvalds's avatar
Linus Torvalds committed
4437 4438
	int retval;

4439 4440
	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL))
		return -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
4441

4442 4443 4444 4445 4446
	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;
Linus Torvalds's avatar
Linus Torvalds committed
4447 4448
}

4449
long sched_getaffinity(pid_t pid, struct cpumask *mask)
Linus Torvalds's avatar
Linus Torvalds committed
4450
{
4451
	struct task_struct *p;
4452
	unsigned long flags;
Linus Torvalds's avatar
Linus Torvalds committed
4453 4454
	int retval;

4455
	get_online_cpus();
4456
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
4457 4458 4459 4460 4461 4462

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

4463 4464 4465 4466
	retval = security_task_getscheduler(p);
	if (retval)
		goto out_unlock;

4467
	raw_spin_lock_irqsave(&p->pi_lock, flags);
4468
	cpumask_and(mask, &p->cpus_allowed, cpu_online_mask);
4469
	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
4470 4471

out_unlock:
4472
	rcu_read_unlock();
4473
	put_online_cpus();
Linus Torvalds's avatar
Linus Torvalds committed
4474

4475
	return retval;
Linus Torvalds's avatar
Linus Torvalds committed
4476 4477 4478 4479 4480 4481 4482 4483
}

/**
 * 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
 */
4484 4485
SYSCALL_DEFINE3(sched_getaffinity, pid_t, pid, unsigned int, len,
		unsigned long __user *, user_mask_ptr)
Linus Torvalds's avatar
Linus Torvalds committed
4486 4487
{
	int ret;
4488
	cpumask_var_t mask;
Linus Torvalds's avatar
Linus Torvalds committed
4489

4490
	if ((len * BITS_PER_BYTE) < nr_cpu_ids)
4491 4492
		return -EINVAL;
	if (len & (sizeof(unsigned long)-1))
Linus Torvalds's avatar
Linus Torvalds committed
4493 4494
		return -EINVAL;

4495 4496
	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
		return -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
4497

4498 4499
	ret = sched_getaffinity(pid, mask);
	if (ret == 0) {
4500
		size_t retlen = min_t(size_t, len, cpumask_size());
4501 4502

		if (copy_to_user(user_mask_ptr, mask, retlen))
4503 4504
			ret = -EFAULT;
		else
4505
			ret = retlen;
4506 4507
	}
	free_cpumask_var(mask);
Linus Torvalds's avatar
Linus Torvalds committed
4508

4509
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
4510 4511 4512 4513 4514
}

/**
 * sys_sched_yield - yield the current processor to other threads.
 *
Ingo Molnar's avatar
Ingo Molnar committed
4515 4516
 * This function yields the current CPU to other tasks. If there are no
 * other threads running on this CPU then this function will return.
Linus Torvalds's avatar
Linus Torvalds committed
4517
 */
4518
SYSCALL_DEFINE0(sched_yield)
Linus Torvalds's avatar
Linus Torvalds committed
4519
{
4520
	struct rq *rq = this_rq_lock();
Linus Torvalds's avatar
Linus Torvalds committed
4521

4522
	schedstat_inc(rq, yld_count);
4523
	current->sched_class->yield_task(rq);
Linus Torvalds's avatar
Linus Torvalds committed
4524 4525 4526 4527 4528 4529

	/*
	 * Since we are going to call schedule() anyway, there's
	 * no need to preempt or enable interrupts:
	 */
	__release(rq->lock);
4530
	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
4531
	do_raw_spin_unlock(&rq->lock);
4532
	sched_preempt_enable_no_resched();
Linus Torvalds's avatar
Linus Torvalds committed
4533 4534 4535 4536 4537 4538

	schedule();

	return 0;
}

4539 4540 4541 4542 4543
static inline int should_resched(void)
{
	return need_resched() && !(preempt_count() & PREEMPT_ACTIVE);
}

Andrew Morton's avatar
Andrew Morton committed
4544
static void __cond_resched(void)
Linus Torvalds's avatar
Linus Torvalds committed
4545
{
4546
	add_preempt_count(PREEMPT_ACTIVE);
4547
	__schedule();
4548
	sub_preempt_count(PREEMPT_ACTIVE);
Linus Torvalds's avatar
Linus Torvalds committed
4549 4550
}

4551
int __sched _cond_resched(void)
Linus Torvalds's avatar
Linus Torvalds committed
4552
{
4553
	if (should_resched()) {
Linus Torvalds's avatar
Linus Torvalds committed
4554 4555 4556 4557 4558
		__cond_resched();
		return 1;
	}
	return 0;
}
4559
EXPORT_SYMBOL(_cond_resched);
Linus Torvalds's avatar
Linus Torvalds committed
4560 4561

/*
4562
 * __cond_resched_lock() - if a reschedule is pending, drop the given lock,
Linus Torvalds's avatar
Linus Torvalds committed
4563 4564
 * call schedule, and on return reacquire the lock.
 *
Ingo Molnar's avatar
Ingo Molnar committed
4565
 * This works OK both with and without CONFIG_PREEMPT. We do strange low-level
Linus Torvalds's avatar
Linus Torvalds committed
4566 4567 4568
 * operations here to prevent schedule() from being called twice (once via
 * spin_unlock(), once by hand).
 */
4569
int __cond_resched_lock(spinlock_t *lock)
Linus Torvalds's avatar
Linus Torvalds committed
4570
{
4571
	int resched = should_resched();
Jan Kara's avatar
Jan Kara committed
4572 4573
	int ret = 0;

4574 4575
	lockdep_assert_held(lock);

Nick Piggin's avatar
Nick Piggin committed
4576
	if (spin_needbreak(lock) || resched) {
Linus Torvalds's avatar
Linus Torvalds committed
4577
		spin_unlock(lock);
4578
		if (resched)
Nick Piggin's avatar
Nick Piggin committed
4579 4580 4581
			__cond_resched();
		else
			cpu_relax();
Jan Kara's avatar
Jan Kara committed
4582
		ret = 1;
Linus Torvalds's avatar
Linus Torvalds committed
4583 4584
		spin_lock(lock);
	}
Jan Kara's avatar
Jan Kara committed
4585
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
4586
}
4587
EXPORT_SYMBOL(__cond_resched_lock);
Linus Torvalds's avatar
Linus Torvalds committed
4588

4589
int __sched __cond_resched_softirq(void)
Linus Torvalds's avatar
Linus Torvalds committed
4590 4591 4592
{
	BUG_ON(!in_softirq());

4593
	if (should_resched()) {
4594
		local_bh_enable();
Linus Torvalds's avatar
Linus Torvalds committed
4595 4596 4597 4598 4599 4600
		__cond_resched();
		local_bh_disable();
		return 1;
	}
	return 0;
}
4601
EXPORT_SYMBOL(__cond_resched_softirq);
Linus Torvalds's avatar
Linus Torvalds committed
4602 4603 4604 4605

/**
 * yield - yield the current processor to other threads.
 *
Peter Zijlstra's avatar
Peter Zijlstra committed
4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623
 * 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!
Linus Torvalds's avatar
Linus Torvalds committed
4624 4625 4626 4627 4628 4629 4630 4631
 */
void __sched yield(void)
{
	set_current_state(TASK_RUNNING);
	sys_sched_yield();
}
EXPORT_SYMBOL(yield);

4632 4633 4634 4635
/**
 * yield_to - yield the current processor to another thread in
 * your thread group, or accelerate that thread toward the
 * processor it's on.
4636 4637
 * @p: target task
 * @preempt: whether task preemption is allowed or not
4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671
 *
 * 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);
4672
	if (yielded) {
4673
		schedstat_inc(rq, yld_count);
4674 4675 4676 4677 4678 4679
		/*
		 * Make p's CPU reschedule; pick_next_entity takes care of
		 * fairness.
		 */
		if (preempt && rq != p_rq)
			resched_task(p_rq->curr);
4680 4681 4682 4683 4684 4685 4686
	} 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;
4687
	}
4688 4689 4690 4691 4692 4693 4694 4695 4696 4697 4698 4699

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

	if (yielded)
		schedule();

	return yielded;
}
EXPORT_SYMBOL_GPL(yield_to);

Linus Torvalds's avatar
Linus Torvalds committed
4700
/*
Ingo Molnar's avatar
Ingo Molnar committed
4701
 * This task is about to go to sleep on IO. Increment rq->nr_iowait so
Linus Torvalds's avatar
Linus Torvalds committed
4702 4703 4704 4705
 * that process accounting knows that this is a task in IO wait state.
 */
void __sched io_schedule(void)
{
4706
	struct rq *rq = raw_rq();
Linus Torvalds's avatar
Linus Torvalds committed
4707

4708
	delayacct_blkio_start();
Linus Torvalds's avatar
Linus Torvalds committed
4709
	atomic_inc(&rq->nr_iowait);
4710
	blk_flush_plug(current);
4711
	current->in_iowait = 1;
Linus Torvalds's avatar
Linus Torvalds committed
4712
	schedule();
4713
	current->in_iowait = 0;
Linus Torvalds's avatar
Linus Torvalds committed
4714
	atomic_dec(&rq->nr_iowait);
4715
	delayacct_blkio_end();
Linus Torvalds's avatar
Linus Torvalds committed
4716 4717 4718 4719 4720
}
EXPORT_SYMBOL(io_schedule);

long __sched io_schedule_timeout(long timeout)
{
4721
	struct rq *rq = raw_rq();
Linus Torvalds's avatar
Linus Torvalds committed
4722 4723
	long ret;

4724
	delayacct_blkio_start();
Linus Torvalds's avatar
Linus Torvalds committed
4725
	atomic_inc(&rq->nr_iowait);
4726
	blk_flush_plug(current);
4727
	current->in_iowait = 1;
Linus Torvalds's avatar
Linus Torvalds committed
4728
	ret = schedule_timeout(timeout);
4729
	current->in_iowait = 0;
Linus Torvalds's avatar
Linus Torvalds committed
4730
	atomic_dec(&rq->nr_iowait);
4731
	delayacct_blkio_end();
Linus Torvalds's avatar
Linus Torvalds committed
4732 4733 4734 4735 4736 4737 4738 4739 4740 4741
	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.
 */
4742
SYSCALL_DEFINE1(sched_get_priority_max, int, policy)
Linus Torvalds's avatar
Linus Torvalds committed
4743 4744 4745 4746 4747 4748 4749 4750 4751
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = MAX_USER_RT_PRIO-1;
		break;
	case SCHED_NORMAL:
4752
	case SCHED_BATCH:
Ingo Molnar's avatar
Ingo Molnar committed
4753
	case SCHED_IDLE:
Linus Torvalds's avatar
Linus Torvalds committed
4754 4755 4756 4757 4758 4759 4760 4761 4762 4763 4764 4765 4766
		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.
 */
4767
SYSCALL_DEFINE1(sched_get_priority_min, int, policy)
Linus Torvalds's avatar
Linus Torvalds committed
4768 4769 4770 4771 4772 4773 4774 4775 4776
{
	int ret = -EINVAL;

	switch (policy) {
	case SCHED_FIFO:
	case SCHED_RR:
		ret = 1;
		break;
	case SCHED_NORMAL:
4777
	case SCHED_BATCH:
Ingo Molnar's avatar
Ingo Molnar committed
4778
	case SCHED_IDLE:
Linus Torvalds's avatar
Linus Torvalds committed
4779 4780 4781 4782 4783 4784 4785 4786 4787 4788 4789 4790 4791
		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.
 */
4792
SYSCALL_DEFINE2(sched_rr_get_interval, pid_t, pid,
4793
		struct timespec __user *, interval)
Linus Torvalds's avatar
Linus Torvalds committed
4794
{
4795
	struct task_struct *p;
Dmitry Adamushko's avatar
Dmitry Adamushko committed
4796
	unsigned int time_slice;
4797 4798
	unsigned long flags;
	struct rq *rq;
4799
	int retval;
Linus Torvalds's avatar
Linus Torvalds committed
4800 4801 4802
	struct timespec t;

	if (pid < 0)
4803
		return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
4804 4805

	retval = -ESRCH;
4806
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
4807 4808 4809 4810 4811 4812 4813 4814
	p = find_process_by_pid(pid);
	if (!p)
		goto out_unlock;

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

4815 4816
	rq = task_rq_lock(p, &flags);
	time_slice = p->sched_class->get_rr_interval(rq, p);
4817
	task_rq_unlock(rq, p, &flags);
Dmitry Adamushko's avatar
Dmitry Adamushko committed
4818

4819
	rcu_read_unlock();
Dmitry Adamushko's avatar
Dmitry Adamushko committed
4820
	jiffies_to_timespec(time_slice, &t);
Linus Torvalds's avatar
Linus Torvalds committed
4821 4822
	retval = copy_to_user(interval, &t, sizeof(t)) ? -EFAULT : 0;
	return retval;
4823

Linus Torvalds's avatar
Linus Torvalds committed
4824
out_unlock:
4825
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
4826 4827 4828
	return retval;
}

4829
static const char stat_nam[] = TASK_STATE_TO_CHAR_STR;
4830

4831
void sched_show_task(struct task_struct *p)
Linus Torvalds's avatar
Linus Torvalds committed
4832 4833
{
	unsigned long free = 0;
4834
	unsigned state;
Linus Torvalds's avatar
Linus Torvalds committed
4835 4836

	state = p->state ? __ffs(p->state) + 1 : 0;
4837
	printk(KERN_INFO "%-15.15s %c", p->comm,
4838
		state < sizeof(stat_nam) - 1 ? stat_nam[state] : '?');
4839
#if BITS_PER_LONG == 32
Linus Torvalds's avatar
Linus Torvalds committed
4840
	if (state == TASK_RUNNING)
4841
		printk(KERN_CONT " running  ");
Linus Torvalds's avatar
Linus Torvalds committed
4842
	else
4843
		printk(KERN_CONT " %08lx ", thread_saved_pc(p));
Linus Torvalds's avatar
Linus Torvalds committed
4844 4845
#else
	if (state == TASK_RUNNING)
4846
		printk(KERN_CONT "  running task    ");
Linus Torvalds's avatar
Linus Torvalds committed
4847
	else
4848
		printk(KERN_CONT " %016lx ", thread_saved_pc(p));
Linus Torvalds's avatar
Linus Torvalds committed
4849 4850
#endif
#ifdef CONFIG_DEBUG_STACK_USAGE
4851
	free = stack_not_used(p);
Linus Torvalds's avatar
Linus Torvalds committed
4852
#endif
4853
	printk(KERN_CONT "%5lu %5d %6d 0x%08lx\n", free,
4854
		task_pid_nr(p), task_pid_nr(rcu_dereference(p->real_parent)),
4855
		(unsigned long)task_thread_info(p)->flags);
Linus Torvalds's avatar
Linus Torvalds committed
4856

4857
	show_stack(p, NULL);
Linus Torvalds's avatar
Linus Torvalds committed
4858 4859
}

4860
void show_state_filter(unsigned long state_filter)
Linus Torvalds's avatar
Linus Torvalds committed
4861
{
4862
	struct task_struct *g, *p;
Linus Torvalds's avatar
Linus Torvalds committed
4863

4864
#if BITS_PER_LONG == 32
4865 4866
	printk(KERN_INFO
		"  task                PC stack   pid father\n");
Linus Torvalds's avatar
Linus Torvalds committed
4867
#else
4868 4869
	printk(KERN_INFO
		"  task                        PC stack   pid father\n");
Linus Torvalds's avatar
Linus Torvalds committed
4870
#endif
4871
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
4872 4873 4874
	do_each_thread(g, p) {
		/*
		 * reset the NMI-timeout, listing all files on a slow
Lucas De Marchi's avatar
Lucas De Marchi committed
4875
		 * console might take a lot of time:
Linus Torvalds's avatar
Linus Torvalds committed
4876 4877
		 */
		touch_nmi_watchdog();
4878
		if (!state_filter || (p->state & state_filter))
4879
			sched_show_task(p);
Linus Torvalds's avatar
Linus Torvalds committed
4880 4881
	} while_each_thread(g, p);

4882 4883
	touch_all_softlockup_watchdogs();

Ingo Molnar's avatar
Ingo Molnar committed
4884 4885 4886
#ifdef CONFIG_SCHED_DEBUG
	sysrq_sched_debug_show();
#endif
4887
	rcu_read_unlock();
4888 4889 4890
	/*
	 * Only show locks if all tasks are dumped:
	 */
4891
	if (!state_filter)
4892
		debug_show_all_locks();
Linus Torvalds's avatar
Linus Torvalds committed
4893 4894
}

4895 4896
void __cpuinit init_idle_bootup_task(struct task_struct *idle)
{
Ingo Molnar's avatar
Ingo Molnar committed
4897
	idle->sched_class = &idle_sched_class;
4898 4899
}

4900 4901 4902 4903 4904 4905 4906 4907
/**
 * 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.
 */
4908
void __cpuinit init_idle(struct task_struct *idle, int cpu)
Linus Torvalds's avatar
Linus Torvalds committed
4909
{
4910
	struct rq *rq = cpu_rq(cpu);
Linus Torvalds's avatar
Linus Torvalds committed
4911 4912
	unsigned long flags;

4913
	raw_spin_lock_irqsave(&rq->lock, flags);
4914

Ingo Molnar's avatar
Ingo Molnar committed
4915
	__sched_fork(idle);
4916
	idle->state = TASK_RUNNING;
Ingo Molnar's avatar
Ingo Molnar committed
4917 4918
	idle->se.exec_start = sched_clock();

4919
	do_set_cpus_allowed(idle, cpumask_of(cpu));
4920 4921 4922 4923 4924 4925 4926 4927 4928 4929 4930
	/*
	 * 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();
Ingo Molnar's avatar
Ingo Molnar committed
4931
	__set_task_cpu(idle, cpu);
4932
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
4933 4934

	rq->curr = rq->idle = idle;
4935 4936
#if defined(CONFIG_SMP)
	idle->on_cpu = 1;
4937
#endif
4938
	raw_spin_unlock_irqrestore(&rq->lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
4939 4940

	/* Set the preempt count _outside_ the spinlocks! */
4941
	task_thread_info(idle)->preempt_count = 0;
4942

Ingo Molnar's avatar
Ingo Molnar committed
4943 4944 4945 4946
	/*
	 * The idle tasks have their own, simple scheduling class:
	 */
	idle->sched_class = &idle_sched_class;
4947
	ftrace_graph_init_idle_task(idle, cpu);
4948 4949 4950
#if defined(CONFIG_SMP)
	sprintf(idle->comm, "%s/%d", INIT_TASK_COMM, cpu);
#endif
4951 4952
}

Linus Torvalds's avatar
Linus Torvalds committed
4953
#ifdef CONFIG_SMP
4954 4955 4956 4957
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);
4958 4959 4960

	cpumask_copy(&p->cpus_allowed, new_mask);
	p->rt.nr_cpus_allowed = cpumask_weight(new_mask);
4961 4962
}

Linus Torvalds's avatar
Linus Torvalds committed
4963 4964 4965
/*
 * This is how migration works:
 *
4966 4967 4968 4969 4970 4971
 * 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
Linus Torvalds's avatar
Linus Torvalds committed
4972
 *    it and puts it into the right queue.
4973 4974
 * 5) stopper completes and stop_one_cpu() returns and the migration
 *    is done.
Linus Torvalds's avatar
Linus Torvalds committed
4975 4976 4977 4978 4979 4980 4981 4982
 */

/*
 * 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
Ingo Molnar's avatar
Ingo Molnar committed
4983
 * task must not exit() & deallocate itself prematurely. The
Linus Torvalds's avatar
Linus Torvalds committed
4984 4985
 * call is not atomic; no spinlocks may be held.
 */
4986
int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
Linus Torvalds's avatar
Linus Torvalds committed
4987 4988
{
	unsigned long flags;
4989
	struct rq *rq;
4990
	unsigned int dest_cpu;
4991
	int ret = 0;
Linus Torvalds's avatar
Linus Torvalds committed
4992 4993

	rq = task_rq_lock(p, &flags);
4994

4995 4996 4997
	if (cpumask_equal(&p->cpus_allowed, new_mask))
		goto out;

4998
	if (!cpumask_intersects(new_mask, cpu_active_mask)) {
Linus Torvalds's avatar
Linus Torvalds committed
4999 5000 5001 5002
		ret = -EINVAL;
		goto out;
	}

5003
	if (unlikely((p->flags & PF_THREAD_BOUND) && p != current)) {
5004 5005 5006 5007
		ret = -EINVAL;
		goto out;
	}

5008
	do_set_cpus_allowed(p, new_mask);
5009

Linus Torvalds's avatar
Linus Torvalds committed
5010
	/* Can the task run on the task's current CPU? If so, we're done */
5011
	if (cpumask_test_cpu(task_cpu(p), new_mask))
Linus Torvalds's avatar
Linus Torvalds committed
5012 5013
		goto out;

5014
	dest_cpu = cpumask_any_and(cpu_active_mask, new_mask);
5015
	if (p->on_rq) {
5016
		struct migration_arg arg = { p, dest_cpu };
Linus Torvalds's avatar
Linus Torvalds committed
5017
		/* Need help from migration thread: drop lock and wait. */
5018
		task_rq_unlock(rq, p, &flags);
5019
		stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg);
Linus Torvalds's avatar
Linus Torvalds committed
5020 5021 5022 5023
		tlb_migrate_finish(p->mm);
		return 0;
	}
out:
5024
	task_rq_unlock(rq, p, &flags);
5025

Linus Torvalds's avatar
Linus Torvalds committed
5026 5027
	return ret;
}
5028
EXPORT_SYMBOL_GPL(set_cpus_allowed_ptr);
Linus Torvalds's avatar
Linus Torvalds committed
5029 5030

/*
Ingo Molnar's avatar
Ingo Molnar committed
5031
 * Move (not current) task off this cpu, onto dest cpu. We're doing
Linus Torvalds's avatar
Linus Torvalds committed
5032 5033 5034 5035 5036 5037
 * 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.
5038 5039
 *
 * Returns non-zero if task was successfully migrated.
Linus Torvalds's avatar
Linus Torvalds committed
5040
 */
5041
static int __migrate_task(struct task_struct *p, int src_cpu, int dest_cpu)
Linus Torvalds's avatar
Linus Torvalds committed
5042
{
5043
	struct rq *rq_dest, *rq_src;
5044
	int ret = 0;
Linus Torvalds's avatar
Linus Torvalds committed
5045

5046
	if (unlikely(!cpu_active(dest_cpu)))
5047
		return ret;
Linus Torvalds's avatar
Linus Torvalds committed
5048 5049 5050 5051

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

5052
	raw_spin_lock(&p->pi_lock);
Linus Torvalds's avatar
Linus Torvalds committed
5053 5054 5055
	double_rq_lock(rq_src, rq_dest);
	/* Already moved. */
	if (task_cpu(p) != src_cpu)
5056
		goto done;
Linus Torvalds's avatar
Linus Torvalds committed
5057
	/* Affinity changed (again). */
5058
	if (!cpumask_test_cpu(dest_cpu, tsk_cpus_allowed(p)))
5059
		goto fail;
Linus Torvalds's avatar
Linus Torvalds committed
5060

5061 5062 5063 5064
	/*
	 * If we're not on a rq, the next wake-up will ensure we're
	 * placed properly.
	 */
Peter Zijlstra's avatar
Peter Zijlstra committed
5065
	if (p->on_rq) {
5066
		dequeue_task(rq_src, p, 0);
5067
		set_task_cpu(p, dest_cpu);
5068
		enqueue_task(rq_dest, p, 0);
5069
		check_preempt_curr(rq_dest, p, 0);
Linus Torvalds's avatar
Linus Torvalds committed
5070
	}
5071
done:
5072
	ret = 1;
5073
fail:
Linus Torvalds's avatar
Linus Torvalds committed
5074
	double_rq_unlock(rq_src, rq_dest);
5075
	raw_spin_unlock(&p->pi_lock);
5076
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
5077 5078 5079
}

/*
5080 5081 5082
 * 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.
Linus Torvalds's avatar
Linus Torvalds committed
5083
 */
5084
static int migration_cpu_stop(void *data)
Linus Torvalds's avatar
Linus Torvalds committed
5085
{
5086
	struct migration_arg *arg = data;
5087

5088 5089 5090 5091
	/*
	 * The original target cpu might have gone down and we might
	 * be on another cpu but it doesn't matter.
	 */
5092
	local_irq_disable();
5093
	__migrate_task(arg->task, raw_smp_processor_id(), arg->dest_cpu);
5094
	local_irq_enable();
Linus Torvalds's avatar
Linus Torvalds committed
5095
	return 0;
5096 5097
}

Linus Torvalds's avatar
Linus Torvalds committed
5098
#ifdef CONFIG_HOTPLUG_CPU
5099

5100
/*
5101 5102
 * Ensures that the idle task is using init_mm right before its cpu goes
 * offline.
5103
 */
5104
void idle_task_exit(void)
Linus Torvalds's avatar
Linus Torvalds committed
5105
{
5106
	struct mm_struct *mm = current->active_mm;
5107

5108
	BUG_ON(cpu_online(smp_processor_id()));
5109

5110 5111 5112
	if (mm != &init_mm)
		switch_mm(mm, &init_mm, current);
	mmdrop(mm);
Linus Torvalds's avatar
Linus Torvalds committed
5113 5114 5115 5116 5117 5118 5119 5120 5121
}

/*
 * While a dead CPU has no uninterruptible tasks queued at this point,
 * it might still have a nonzero ->nr_uninterruptible counter, because
 * for performance reasons the counter is not stricly tracking tasks to
 * their home CPUs. So we just add the counter to another CPU's counter,
 * to keep the global sum constant after CPU-down:
 */
5122
static void migrate_nr_uninterruptible(struct rq *rq_src)
Linus Torvalds's avatar
Linus Torvalds committed
5123
{
5124
	struct rq *rq_dest = cpu_rq(cpumask_any(cpu_active_mask));
Linus Torvalds's avatar
Linus Torvalds committed
5125 5126 5127 5128 5129

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

Ingo Molnar's avatar
Ingo Molnar committed
5130
/*
5131
 * remove the tasks which were accounted by rq from calc_load_tasks.
Linus Torvalds's avatar
Linus Torvalds committed
5132
 */
5133
static void calc_global_load_remove(struct rq *rq)
Linus Torvalds's avatar
Linus Torvalds committed
5134
{
5135 5136
	atomic_long_sub(rq->calc_load_active, &calc_load_tasks);
	rq->calc_load_active = 0;
Linus Torvalds's avatar
Linus Torvalds committed
5137 5138
}

5139
/*
5140 5141 5142 5143 5144 5145
 * 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.
Linus Torvalds's avatar
Linus Torvalds committed
5146
 */
5147
static void migrate_tasks(unsigned int dead_cpu)
Linus Torvalds's avatar
Linus Torvalds committed
5148
{
5149
	struct rq *rq = cpu_rq(dead_cpu);
5150 5151
	struct task_struct *next, *stop = rq->stop;
	int dest_cpu;
Linus Torvalds's avatar
Linus Torvalds committed
5152 5153

	/*
5154 5155 5156 5157 5158 5159 5160
	 * 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.
Linus Torvalds's avatar
Linus Torvalds committed
5161
	 */
5162
	rq->stop = NULL;
5163

5164 5165 5166
	/* Ensure any throttled groups are reachable by pick_next_task */
	unthrottle_offline_cfs_rqs(rq);

Ingo Molnar's avatar
Ingo Molnar committed
5167
	for ( ; ; ) {
5168 5169 5170 5171 5172
		/*
		 * There's this thread running, bail when that's the only
		 * remaining thread.
		 */
		if (rq->nr_running == 1)
Ingo Molnar's avatar
Ingo Molnar committed
5173
			break;
5174

5175
		next = pick_next_task(rq);
5176
		BUG_ON(!next);
Dmitry Adamushko's avatar
Dmitry Adamushko committed
5177
		next->sched_class->put_prev_task(rq, next);
5178

5179 5180 5181 5182 5183 5184 5185
		/* 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);
Linus Torvalds's avatar
Linus Torvalds committed
5186
	}
5187

5188
	rq->stop = stop;
5189
}
5190

Linus Torvalds's avatar
Linus Torvalds committed
5191 5192
#endif /* CONFIG_HOTPLUG_CPU */

5193 5194 5195
#if defined(CONFIG_SCHED_DEBUG) && defined(CONFIG_SYSCTL)

static struct ctl_table sd_ctl_dir[] = {
5196 5197
	{
		.procname	= "sched_domain",
5198
		.mode		= 0555,
5199
	},
5200
	{}
5201 5202 5203
};

static struct ctl_table sd_ctl_root[] = {
5204 5205
	{
		.procname	= "kernel",
5206
		.mode		= 0555,
5207 5208
		.child		= sd_ctl_dir,
	},
5209
	{}
5210 5211 5212 5213 5214
};

static struct ctl_table *sd_alloc_ctl_entry(int n)
{
	struct ctl_table *entry =
5215
		kcalloc(n, sizeof(struct ctl_table), GFP_KERNEL);
5216 5217 5218 5219

	return entry;
}

5220 5221
static void sd_free_ctl_entry(struct ctl_table **tablep)
{
5222
	struct ctl_table *entry;
5223

5224 5225 5226
	/*
	 * In the intermediate directories, both the child directory and
	 * procname are dynamically allocated and could fail but the mode
Ingo Molnar's avatar
Ingo Molnar committed
5227
	 * will always be set. In the lowest directory the names are
5228 5229 5230
	 * static strings and all have proc handlers.
	 */
	for (entry = *tablep; entry->mode; entry++) {
5231 5232
		if (entry->child)
			sd_free_ctl_entry(&entry->child);
5233 5234 5235
		if (entry->proc_handler == NULL)
			kfree(entry->procname);
	}
5236 5237 5238 5239 5240

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

5241
static void
5242
set_table_entry(struct ctl_table *entry,
5243
		const char *procname, void *data, int maxlen,
5244
		umode_t mode, proc_handler *proc_handler)
5245 5246 5247 5248 5249 5250 5251 5252 5253 5254 5255
{
	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)
{
5256
	struct ctl_table *table = sd_alloc_ctl_entry(13);
5257

5258 5259 5260
	if (table == NULL)
		return NULL;

5261
	set_table_entry(&table[0], "min_interval", &sd->min_interval,
5262
		sizeof(long), 0644, proc_doulongvec_minmax);
5263
	set_table_entry(&table[1], "max_interval", &sd->max_interval,
5264
		sizeof(long), 0644, proc_doulongvec_minmax);
5265
	set_table_entry(&table[2], "busy_idx", &sd->busy_idx,
5266
		sizeof(int), 0644, proc_dointvec_minmax);
5267
	set_table_entry(&table[3], "idle_idx", &sd->idle_idx,
5268
		sizeof(int), 0644, proc_dointvec_minmax);
5269
	set_table_entry(&table[4], "newidle_idx", &sd->newidle_idx,
5270
		sizeof(int), 0644, proc_dointvec_minmax);
5271
	set_table_entry(&table[5], "wake_idx", &sd->wake_idx,
5272
		sizeof(int), 0644, proc_dointvec_minmax);
5273
	set_table_entry(&table[6], "forkexec_idx", &sd->forkexec_idx,
5274
		sizeof(int), 0644, proc_dointvec_minmax);
5275
	set_table_entry(&table[7], "busy_factor", &sd->busy_factor,
5276
		sizeof(int), 0644, proc_dointvec_minmax);
5277
	set_table_entry(&table[8], "imbalance_pct", &sd->imbalance_pct,
5278
		sizeof(int), 0644, proc_dointvec_minmax);
5279
	set_table_entry(&table[9], "cache_nice_tries",
5280 5281
		&sd->cache_nice_tries,
		sizeof(int), 0644, proc_dointvec_minmax);
5282
	set_table_entry(&table[10], "flags", &sd->flags,
5283
		sizeof(int), 0644, proc_dointvec_minmax);
5284 5285 5286
	set_table_entry(&table[11], "name", sd->name,
		CORENAME_MAX_SIZE, 0444, proc_dostring);
	/* &table[12] is terminator */
5287 5288 5289 5290

	return table;
}

5291
static ctl_table *sd_alloc_ctl_cpu_table(int cpu)
5292 5293 5294 5295 5296 5297 5298 5299 5300
{
	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);
5301 5302
	if (table == NULL)
		return NULL;
5303 5304 5305 5306 5307

	i = 0;
	for_each_domain(cpu, sd) {
		snprintf(buf, 32, "domain%d", i);
		entry->procname = kstrdup(buf, GFP_KERNEL);
5308
		entry->mode = 0555;
5309 5310 5311 5312 5313 5314 5315 5316
		entry->child = sd_alloc_ctl_domain_table(sd);
		entry++;
		i++;
	}
	return table;
}

static struct ctl_table_header *sd_sysctl_header;
5317
static void register_sched_domain_sysctl(void)
5318
{
5319
	int i, cpu_num = num_possible_cpus();
5320 5321 5322
	struct ctl_table *entry = sd_alloc_ctl_entry(cpu_num + 1);
	char buf[32];

5323 5324 5325
	WARN_ON(sd_ctl_dir[0].child);
	sd_ctl_dir[0].child = entry;

5326 5327 5328
	if (entry == NULL)
		return;

5329
	for_each_possible_cpu(i) {
5330 5331
		snprintf(buf, 32, "cpu%d", i);
		entry->procname = kstrdup(buf, GFP_KERNEL);
5332
		entry->mode = 0555;
5333
		entry->child = sd_alloc_ctl_cpu_table(i);
5334
		entry++;
5335
	}
5336 5337

	WARN_ON(sd_sysctl_header);
5338 5339
	sd_sysctl_header = register_sysctl_table(sd_ctl_root);
}
5340

5341
/* may be called multiple times per register */
5342 5343
static void unregister_sched_domain_sysctl(void)
{
5344 5345
	if (sd_sysctl_header)
		unregister_sysctl_table(sd_sysctl_header);
5346
	sd_sysctl_header = NULL;
5347 5348
	if (sd_ctl_dir[0].child)
		sd_free_ctl_entry(&sd_ctl_dir[0].child);
5349
}
5350
#else
5351 5352 5353 5354
static void register_sched_domain_sysctl(void)
{
}
static void unregister_sched_domain_sysctl(void)
5355 5356 5357 5358
{
}
#endif

5359 5360 5361 5362 5363
static void set_rq_online(struct rq *rq)
{
	if (!rq->online) {
		const struct sched_class *class;

5364
		cpumask_set_cpu(rq->cpu, rq->rd->online);
5365 5366 5367 5368 5369 5370 5371 5372 5373 5374 5375 5376 5377 5378 5379 5380 5381 5382 5383
		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);
		}

5384
		cpumask_clear_cpu(rq->cpu, rq->rd->online);
5385 5386 5387 5388
		rq->online = 0;
	}
}

Linus Torvalds's avatar
Linus Torvalds committed
5389 5390 5391 5392
/*
 * migration_call - callback that gets triggered when a CPU is added.
 * Here we can start up the necessary migration thread for the new CPU.
 */
5393 5394
static int __cpuinit
migration_call(struct notifier_block *nfb, unsigned long action, void *hcpu)
Linus Torvalds's avatar
Linus Torvalds committed
5395
{
5396
	int cpu = (long)hcpu;
Linus Torvalds's avatar
Linus Torvalds committed
5397
	unsigned long flags;
5398
	struct rq *rq = cpu_rq(cpu);
Linus Torvalds's avatar
Linus Torvalds committed
5399

5400
	switch (action & ~CPU_TASKS_FROZEN) {
5401

Linus Torvalds's avatar
Linus Torvalds committed
5402
	case CPU_UP_PREPARE:
5403
		rq->calc_load_update = calc_load_update;
Linus Torvalds's avatar
Linus Torvalds committed
5404
		break;
5405

Linus Torvalds's avatar
Linus Torvalds committed
5406
	case CPU_ONLINE:
5407
		/* Update our root-domain */
5408
		raw_spin_lock_irqsave(&rq->lock, flags);
5409
		if (rq->rd) {
5410
			BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
5411 5412

			set_rq_online(rq);
5413
		}
5414
		raw_spin_unlock_irqrestore(&rq->lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
5415
		break;
5416

Linus Torvalds's avatar
Linus Torvalds committed
5417
#ifdef CONFIG_HOTPLUG_CPU
5418
	case CPU_DYING:
5419
		sched_ttwu_pending();
5420
		/* Update our root-domain */
5421
		raw_spin_lock_irqsave(&rq->lock, flags);
5422
		if (rq->rd) {
5423
			BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
5424
			set_rq_offline(rq);
5425
		}
5426 5427
		migrate_tasks(cpu);
		BUG_ON(rq->nr_running != 1); /* the migration thread */
5428
		raw_spin_unlock_irqrestore(&rq->lock, flags);
5429 5430 5431

		migrate_nr_uninterruptible(rq);
		calc_global_load_remove(rq);
5432
		break;
Linus Torvalds's avatar
Linus Torvalds committed
5433 5434
#endif
	}
5435 5436 5437

	update_max_interval();

Linus Torvalds's avatar
Linus Torvalds committed
5438 5439 5440
	return NOTIFY_OK;
}

5441 5442 5443
/*
 * Register at high priority so that task migration (migrate_all_tasks)
 * happens before everything else.  This has to be lower priority than
5444
 * the notifier in the perf_event subsystem, though.
Linus Torvalds's avatar
Linus Torvalds committed
5445
 */
5446
static struct notifier_block __cpuinitdata migration_notifier = {
Linus Torvalds's avatar
Linus Torvalds committed
5447
	.notifier_call = migration_call,
5448
	.priority = CPU_PRI_MIGRATION,
Linus Torvalds's avatar
Linus Torvalds committed
5449 5450
};

5451 5452 5453 5454
static int __cpuinit sched_cpu_active(struct notifier_block *nfb,
				      unsigned long action, void *hcpu)
{
	switch (action & ~CPU_TASKS_FROZEN) {
5455
	case CPU_STARTING:
5456 5457 5458 5459 5460 5461 5462 5463 5464 5465 5466 5467 5468 5469 5470 5471 5472 5473 5474 5475
	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;
	}
}

5476
static int __init migration_init(void)
Linus Torvalds's avatar
Linus Torvalds committed
5477 5478
{
	void *cpu = (void *)(long)smp_processor_id();
5479
	int err;
5480

5481
	/* Initialize migration for the boot CPU */
5482 5483
	err = migration_call(&migration_notifier, CPU_UP_PREPARE, cpu);
	BUG_ON(err == NOTIFY_BAD);
Linus Torvalds's avatar
Linus Torvalds committed
5484 5485
	migration_call(&migration_notifier, CPU_ONLINE, cpu);
	register_cpu_notifier(&migration_notifier);
5486

5487 5488 5489 5490
	/* Register cpu active notifiers */
	cpu_notifier(sched_cpu_active, CPU_PRI_SCHED_ACTIVE);
	cpu_notifier(sched_cpu_inactive, CPU_PRI_SCHED_INACTIVE);

5491
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
5492
}
5493
early_initcall(migration_init);
Linus Torvalds's avatar
Linus Torvalds committed
5494 5495 5496
#endif

#ifdef CONFIG_SMP
5497

5498 5499
static cpumask_var_t sched_domains_tmpmask; /* sched_domains_mutex */

5500
#ifdef CONFIG_SCHED_DEBUG
5501

5502 5503 5504 5505 5506 5507 5508 5509 5510 5511
static __read_mostly int sched_domain_debug_enabled;

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

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

5512
static int sched_domain_debug_one(struct sched_domain *sd, int cpu, int level,
5513
				  struct cpumask *groupmask)
Linus Torvalds's avatar
Linus Torvalds committed
5514
{
5515
	struct sched_group *group = sd->groups;
5516
	char str[256];
Linus Torvalds's avatar
Linus Torvalds committed
5517

Rusty Russell's avatar
Rusty Russell committed
5518
	cpulist_scnprintf(str, sizeof(str), sched_domain_span(sd));
5519
	cpumask_clear(groupmask);
5520 5521 5522 5523

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

	if (!(sd->flags & SD_LOAD_BALANCE)) {
5524
		printk("does not load-balance\n");
5525
		if (sd->parent)
5526 5527
			printk(KERN_ERR "ERROR: !SD_LOAD_BALANCE domain"
					" has parent");
5528
		return -1;
Nick Piggin's avatar
Nick Piggin committed
5529 5530
	}

5531
	printk(KERN_CONT "span %s level %s\n", str, sd->name);
5532

5533
	if (!cpumask_test_cpu(cpu, sched_domain_span(sd))) {
5534 5535
		printk(KERN_ERR "ERROR: domain->span does not contain "
				"CPU%d\n", cpu);
5536
	}
5537
	if (!cpumask_test_cpu(cpu, sched_group_cpus(group))) {
5538 5539
		printk(KERN_ERR "ERROR: domain->groups does not contain"
				" CPU%d\n", cpu);
5540
	}
Linus Torvalds's avatar
Linus Torvalds committed
5541

5542
	printk(KERN_DEBUG "%*s groups:", level + 1, "");
Linus Torvalds's avatar
Linus Torvalds committed
5543
	do {
5544
		if (!group) {
5545 5546
			printk("\n");
			printk(KERN_ERR "ERROR: group is NULL\n");
Linus Torvalds's avatar
Linus Torvalds committed
5547 5548 5549
			break;
		}

5550
		if (!group->sgp->power) {
5551 5552 5553
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: domain->cpu_power not "
					"set\n");
5554 5555
			break;
		}
Linus Torvalds's avatar
Linus Torvalds committed
5556

5557
		if (!cpumask_weight(sched_group_cpus(group))) {
5558 5559
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: empty group\n");
5560 5561
			break;
		}
Linus Torvalds's avatar
Linus Torvalds committed
5562

5563
		if (cpumask_intersects(groupmask, sched_group_cpus(group))) {
5564 5565
			printk(KERN_CONT "\n");
			printk(KERN_ERR "ERROR: repeated CPUs\n");
5566 5567
			break;
		}
Linus Torvalds's avatar
Linus Torvalds committed
5568

5569
		cpumask_or(groupmask, groupmask, sched_group_cpus(group));
Linus Torvalds's avatar
Linus Torvalds committed
5570

Rusty Russell's avatar
Rusty Russell committed
5571
		cpulist_scnprintf(str, sizeof(str), sched_group_cpus(group));
5572

5573
		printk(KERN_CONT " %s", str);
5574
		if (group->sgp->power != SCHED_POWER_SCALE) {
5575
			printk(KERN_CONT " (cpu_power = %d)",
5576
				group->sgp->power);
5577
		}
Linus Torvalds's avatar
Linus Torvalds committed
5578

5579 5580
		group = group->next;
	} while (group != sd->groups);
5581
	printk(KERN_CONT "\n");
Linus Torvalds's avatar
Linus Torvalds committed
5582

5583
	if (!cpumask_equal(sched_domain_span(sd), groupmask))
5584
		printk(KERN_ERR "ERROR: groups don't span domain->span\n");
Linus Torvalds's avatar
Linus Torvalds committed
5585

5586 5587
	if (sd->parent &&
	    !cpumask_subset(groupmask, sched_domain_span(sd->parent)))
5588 5589
		printk(KERN_ERR "ERROR: parent span is not a superset "
			"of domain->span\n");
5590 5591
	return 0;
}
Linus Torvalds's avatar
Linus Torvalds committed
5592

5593 5594 5595
static void sched_domain_debug(struct sched_domain *sd, int cpu)
{
	int level = 0;
Linus Torvalds's avatar
Linus Torvalds committed
5596

5597 5598 5599
	if (!sched_domain_debug_enabled)
		return;

5600 5601 5602 5603
	if (!sd) {
		printk(KERN_DEBUG "CPU%d attaching NULL sched-domain.\n", cpu);
		return;
	}
Linus Torvalds's avatar
Linus Torvalds committed
5604

5605 5606 5607
	printk(KERN_DEBUG "CPU%d attaching sched-domain:\n", cpu);

	for (;;) {
5608
		if (sched_domain_debug_one(sd, cpu, level, sched_domains_tmpmask))
5609
			break;
Linus Torvalds's avatar
Linus Torvalds committed
5610 5611
		level++;
		sd = sd->parent;
5612
		if (!sd)
5613 5614
			break;
	}
Linus Torvalds's avatar
Linus Torvalds committed
5615
}
5616
#else /* !CONFIG_SCHED_DEBUG */
5617
# define sched_domain_debug(sd, cpu) do { } while (0)
5618
#endif /* CONFIG_SCHED_DEBUG */
Linus Torvalds's avatar
Linus Torvalds committed
5619

5620
static int sd_degenerate(struct sched_domain *sd)
5621
{
5622
	if (cpumask_weight(sched_domain_span(sd)) == 1)
5623 5624 5625 5626 5627 5628
		return 1;

	/* Following flags need at least 2 groups */
	if (sd->flags & (SD_LOAD_BALANCE |
			 SD_BALANCE_NEWIDLE |
			 SD_BALANCE_FORK |
5629 5630 5631
			 SD_BALANCE_EXEC |
			 SD_SHARE_CPUPOWER |
			 SD_SHARE_PKG_RESOURCES)) {
5632 5633 5634 5635 5636
		if (sd->groups != sd->groups->next)
			return 0;
	}

	/* Following flags don't use groups */
5637
	if (sd->flags & (SD_WAKE_AFFINE))
5638 5639 5640 5641 5642
		return 0;

	return 1;
}

5643 5644
static int
sd_parent_degenerate(struct sched_domain *sd, struct sched_domain *parent)
5645 5646 5647 5648 5649 5650
{
	unsigned long cflags = sd->flags, pflags = parent->flags;

	if (sd_degenerate(parent))
		return 1;

5651
	if (!cpumask_equal(sched_domain_span(sd), sched_domain_span(parent)))
5652 5653 5654 5655 5656 5657 5658
		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 |
5659 5660 5661
				SD_BALANCE_EXEC |
				SD_SHARE_CPUPOWER |
				SD_SHARE_PKG_RESOURCES);
5662 5663
		if (nr_node_ids == 1)
			pflags &= ~SD_SERIALIZE;
5664 5665 5666 5667 5668 5669 5670
	}
	if (~cflags & pflags)
		return 0;

	return 1;
}

5671
static void free_rootdomain(struct rcu_head *rcu)
5672
{
5673
	struct root_domain *rd = container_of(rcu, struct root_domain, rcu);
5674

5675
	cpupri_cleanup(&rd->cpupri);
5676 5677 5678 5679 5680 5681
	free_cpumask_var(rd->rto_mask);
	free_cpumask_var(rd->online);
	free_cpumask_var(rd->span);
	kfree(rd);
}

5682 5683
static void rq_attach_root(struct rq *rq, struct root_domain *rd)
{
Ingo Molnar's avatar
Ingo Molnar committed
5684
	struct root_domain *old_rd = NULL;
5685 5686
	unsigned long flags;

5687
	raw_spin_lock_irqsave(&rq->lock, flags);
5688 5689

	if (rq->rd) {
Ingo Molnar's avatar
Ingo Molnar committed
5690
		old_rd = rq->rd;
5691

5692
		if (cpumask_test_cpu(rq->cpu, old_rd->online))
5693
			set_rq_offline(rq);
5694

5695
		cpumask_clear_cpu(rq->cpu, old_rd->span);
5696

Ingo Molnar's avatar
Ingo Molnar committed
5697 5698 5699 5700 5701 5702 5703
		/*
		 * 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;
5704 5705 5706 5707 5708
	}

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

5709
	cpumask_set_cpu(rq->cpu, rd->span);
5710
	if (cpumask_test_cpu(rq->cpu, cpu_active_mask))
5711
		set_rq_online(rq);
5712

5713
	raw_spin_unlock_irqrestore(&rq->lock, flags);
Ingo Molnar's avatar
Ingo Molnar committed
5714 5715

	if (old_rd)
5716
		call_rcu_sched(&old_rd->rcu, free_rootdomain);
5717 5718
}

5719
static int init_rootdomain(struct root_domain *rd)
5720 5721 5722
{
	memset(rd, 0, sizeof(*rd));

5723
	if (!alloc_cpumask_var(&rd->span, GFP_KERNEL))
5724
		goto out;
5725
	if (!alloc_cpumask_var(&rd->online, GFP_KERNEL))
5726
		goto free_span;
5727
	if (!alloc_cpumask_var(&rd->rto_mask, GFP_KERNEL))
5728
		goto free_online;
5729

5730
	if (cpupri_init(&rd->cpupri) != 0)
5731
		goto free_rto_mask;
5732
	return 0;
5733

5734 5735
free_rto_mask:
	free_cpumask_var(rd->rto_mask);
5736 5737 5738 5739
free_online:
	free_cpumask_var(rd->online);
free_span:
	free_cpumask_var(rd->span);
5740
out:
5741
	return -ENOMEM;
5742 5743
}

5744 5745 5746 5747 5748 5749
/*
 * 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;

5750 5751
static void init_defrootdomain(void)
{
5752
	init_rootdomain(&def_root_domain);
5753

5754 5755 5756
	atomic_set(&def_root_domain.refcount, 1);
}

5757
static struct root_domain *alloc_rootdomain(void)
5758 5759 5760 5761 5762 5763 5764
{
	struct root_domain *rd;

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

5765
	if (init_rootdomain(rd) != 0) {
5766 5767 5768
		kfree(rd);
		return NULL;
	}
5769 5770 5771 5772

	return rd;
}

5773 5774 5775 5776 5777 5778 5779 5780 5781 5782 5783 5784 5785 5786 5787 5788 5789 5790 5791
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);
}

5792 5793 5794
static void free_sched_domain(struct rcu_head *rcu)
{
	struct sched_domain *sd = container_of(rcu, struct sched_domain, rcu);
5795 5796 5797 5798 5799 5800 5801 5802

	/*
	 * 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)) {
5803
		kfree(sd->groups->sgp);
5804
		kfree(sd->groups);
5805
	}
5806 5807 5808 5809 5810 5811 5812 5813 5814 5815 5816 5817 5818 5819
	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);
}

5820 5821 5822 5823 5824 5825 5826
/*
 * 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
5827
 * two cpus are in the same cache domain, see cpus_share_cache().
5828 5829 5830 5831 5832 5833 5834 5835 5836 5837 5838 5839 5840 5841 5842 5843 5844
 */
DEFINE_PER_CPU(struct sched_domain *, sd_llc);
DEFINE_PER_CPU(int, sd_llc_id);

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

	sd = highest_flag_domain(cpu, SD_SHARE_PKG_RESOURCES);
	if (sd)
		id = cpumask_first(sched_domain_span(sd));

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

Linus Torvalds's avatar
Linus Torvalds committed
5845
/*
Ingo Molnar's avatar
Ingo Molnar committed
5846
 * Attach the domain 'sd' to 'cpu' as its base domain. Callers must
Linus Torvalds's avatar
Linus Torvalds committed
5847 5848
 * hold the hotplug lock.
 */
Ingo Molnar's avatar
Ingo Molnar committed
5849 5850
static void
cpu_attach_domain(struct sched_domain *sd, struct root_domain *rd, int cpu)
Linus Torvalds's avatar
Linus Torvalds committed
5851
{
5852
	struct rq *rq = cpu_rq(cpu);
5853 5854 5855
	struct sched_domain *tmp;

	/* Remove the sched domains which do not contribute to scheduling. */
5856
	for (tmp = sd; tmp; ) {
5857 5858 5859
		struct sched_domain *parent = tmp->parent;
		if (!parent)
			break;
5860

5861
		if (sd_parent_degenerate(tmp, parent)) {
5862
			tmp->parent = parent->parent;
5863 5864
			if (parent->parent)
				parent->parent->child = tmp;
5865
			destroy_sched_domain(parent, cpu);
5866 5867
		} else
			tmp = tmp->parent;
5868 5869
	}

5870
	if (sd && sd_degenerate(sd)) {
5871
		tmp = sd;
5872
		sd = sd->parent;
5873
		destroy_sched_domain(tmp, cpu);
5874 5875 5876
		if (sd)
			sd->child = NULL;
	}
Linus Torvalds's avatar
Linus Torvalds committed
5877

5878
	sched_domain_debug(sd, cpu);
Linus Torvalds's avatar
Linus Torvalds committed
5879

5880
	rq_attach_root(rq, rd);
5881
	tmp = rq->sd;
Nick Piggin's avatar
Nick Piggin committed
5882
	rcu_assign_pointer(rq->sd, sd);
5883
	destroy_sched_domains(tmp, cpu);
5884 5885

	update_top_cache_domain(cpu);
Linus Torvalds's avatar
Linus Torvalds committed
5886 5887 5888
}

/* cpus with isolated domains */
5889
static cpumask_var_t cpu_isolated_map;
Linus Torvalds's avatar
Linus Torvalds committed
5890 5891 5892 5893

/* Setup the mask of cpus configured for isolated domains */
static int __init isolated_cpu_setup(char *str)
{
5894
	alloc_bootmem_cpumask_var(&cpu_isolated_map);
Rusty Russell's avatar
Rusty Russell committed
5895
	cpulist_parse(str, cpu_isolated_map);
Linus Torvalds's avatar
Linus Torvalds committed
5896 5897 5898
	return 1;
}

Ingo Molnar's avatar
Ingo Molnar committed
5899
__setup("isolcpus=", isolated_cpu_setup);
Linus Torvalds's avatar
Linus Torvalds committed
5900

5901
#ifdef CONFIG_NUMA
5902

5903 5904 5905 5906 5907
/**
 * find_next_best_node - find the next node to include in a sched_domain
 * @node: node whose sched_domain we're building
 * @used_nodes: nodes already in the sched_domain
 *
Ingo Molnar's avatar
Ingo Molnar committed
5908
 * Find the next node to include in a given scheduling domain. Simply
5909 5910 5911 5912
 * finds the closest node not already in the @used_nodes map.
 *
 * Should use nodemask_t.
 */
5913
static int find_next_best_node(int node, nodemask_t *used_nodes)
5914
{
5915
	int i, n, val, min_val, best_node = -1;
5916 5917 5918

	min_val = INT_MAX;

5919
	for (i = 0; i < nr_node_ids; i++) {
5920
		/* Start at @node */
5921
		n = (node + i) % nr_node_ids;
5922 5923 5924 5925 5926

		if (!nr_cpus_node(n))
			continue;

		/* Skip already used nodes */
5927
		if (node_isset(n, *used_nodes))
5928 5929 5930 5931 5932 5933 5934 5935 5936 5937 5938
			continue;

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

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

5939 5940
	if (best_node != -1)
		node_set(best_node, *used_nodes);
5941 5942 5943 5944 5945 5946
	return best_node;
}

/**
 * sched_domain_node_span - get a cpumask for a node's sched_domain
 * @node: node whose cpumask we're constructing
5947
 * @span: resulting cpumask
5948
 *
Ingo Molnar's avatar
Ingo Molnar committed
5949
 * Given a node, construct a good cpumask for its sched_domain to span. It
5950 5951 5952
 * should be one that prevents unnecessary balancing, but also spreads tasks
 * out optimally.
 */
5953
static void sched_domain_node_span(int node, struct cpumask *span)
5954
{
5955
	nodemask_t used_nodes;
5956
	int i;
5957

5958
	cpumask_clear(span);
5959
	nodes_clear(used_nodes);
5960

5961
	cpumask_or(span, span, cpumask_of_node(node));
5962
	node_set(node, used_nodes);
5963 5964

	for (i = 1; i < SD_NODES_PER_DOMAIN; i++) {
5965
		int next_node = find_next_best_node(node, &used_nodes);
5966 5967
		if (next_node < 0)
			break;
5968
		cpumask_or(span, span, cpumask_of_node(next_node));
5969 5970
	}
}
5971 5972 5973 5974 5975 5976 5977 5978 5979

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

	sched_domain_node_span(cpu_to_node(cpu), sched_domains_tmpmask);

	return sched_domains_tmpmask;
}
5980 5981 5982 5983 5984

static const struct cpumask *cpu_allnodes_mask(int cpu)
{
	return cpu_possible_mask;
}
5985
#endif /* CONFIG_NUMA */
5986

5987 5988 5989 5990 5991
static const struct cpumask *cpu_cpu_mask(int cpu)
{
	return cpumask_of_node(cpu_to_node(cpu));
}

5992
int sched_smt_power_savings = 0, sched_mc_power_savings = 0;
5993

5994 5995 5996
struct sd_data {
	struct sched_domain **__percpu sd;
	struct sched_group **__percpu sg;
5997
	struct sched_group_power **__percpu sgp;
5998 5999
};

6000
struct s_data {
6001
	struct sched_domain ** __percpu sd;
6002 6003 6004
	struct root_domain	*rd;
};

6005 6006
enum s_alloc {
	sa_rootdomain,
6007
	sa_sd,
6008
	sa_sd_storage,
6009 6010 6011
	sa_none,
};

6012 6013 6014
struct sched_domain_topology_level;

typedef struct sched_domain *(*sched_domain_init_f)(struct sched_domain_topology_level *tl, int cpu);
6015 6016
typedef const struct cpumask *(*sched_domain_mask_f)(int cpu);

6017 6018
#define SDTL_OVERLAP	0x01

6019
struct sched_domain_topology_level {
6020 6021
	sched_domain_init_f init;
	sched_domain_mask_f mask;
6022
	int		    flags;
6023
	struct sd_data      data;
6024 6025
};

6026 6027 6028 6029 6030 6031 6032 6033 6034 6035 6036 6037 6038 6039 6040 6041 6042 6043 6044
static int
build_overlap_sched_groups(struct sched_domain *sd, int cpu)
{
	struct sched_group *first = NULL, *last = NULL, *groups = NULL, *sg;
	const struct cpumask *span = sched_domain_span(sd);
	struct cpumask *covered = sched_domains_tmpmask;
	struct sd_data *sdd = sd->private;
	struct sched_domain *child;
	int i;

	cpumask_clear(covered);

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

		if (cpumask_test_cpu(i, covered))
			continue;

		sg = kzalloc_node(sizeof(struct sched_group) + cpumask_size(),
6045
				GFP_KERNEL, cpu_to_node(cpu));
6046 6047 6048 6049 6050 6051 6052 6053 6054 6055 6056 6057 6058 6059 6060 6061 6062 6063 6064 6065 6066 6067 6068 6069 6070 6071 6072 6073 6074 6075 6076 6077 6078 6079 6080 6081 6082 6083

		if (!sg)
			goto fail;

		sg_span = sched_group_cpus(sg);

		child = *per_cpu_ptr(sdd->sd, i);
		if (child->child) {
			child = child->child;
			cpumask_copy(sg_span, sched_domain_span(child));
		} else
			cpumask_set_cpu(i, sg_span);

		cpumask_or(covered, covered, sg_span);

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

		if (cpumask_test_cpu(cpu, sg_span))
			groups = sg;

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

	return 0;

fail:
	free_sched_groups(first, 0);

	return -ENOMEM;
}

6084
static int get_group(int cpu, struct sd_data *sdd, struct sched_group **sg)
Linus Torvalds's avatar
Linus Torvalds committed
6085
{
6086 6087
	struct sched_domain *sd = *per_cpu_ptr(sdd->sd, cpu);
	struct sched_domain *child = sd->child;
Linus Torvalds's avatar
Linus Torvalds committed
6088

6089 6090
	if (child)
		cpu = cpumask_first(sched_domain_span(child));
6091

6092
	if (sg) {
6093
		*sg = *per_cpu_ptr(sdd->sg, cpu);
6094
		(*sg)->sgp = *per_cpu_ptr(sdd->sgp, cpu);
6095
		atomic_set(&(*sg)->sgp->ref, 1); /* for claim_allocations */
6096
	}
6097 6098

	return cpu;
6099 6100
}

6101
/*
6102 6103 6104
 * 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.
6105 6106
 *
 * Assumes the sched_domain tree is fully constructed
6107
 */
6108 6109
static int
build_sched_groups(struct sched_domain *sd, int cpu)
Linus Torvalds's avatar
Linus Torvalds committed
6110
{
6111 6112 6113
	struct sched_group *first = NULL, *last = NULL;
	struct sd_data *sdd = sd->private;
	const struct cpumask *span = sched_domain_span(sd);
6114
	struct cpumask *covered;
6115
	int i;
6116

6117 6118 6119 6120 6121 6122
	get_group(cpu, sdd, &sd->groups);
	atomic_inc(&sd->groups->ref);

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

6123 6124 6125
	lockdep_assert_held(&sched_domains_mutex);
	covered = sched_domains_tmpmask;

6126
	cpumask_clear(covered);
6127

6128 6129 6130 6131
	for_each_cpu(i, span) {
		struct sched_group *sg;
		int group = get_group(i, sdd, &sg);
		int j;
6132

6133 6134
		if (cpumask_test_cpu(i, covered))
			continue;
6135

6136
		cpumask_clear(sched_group_cpus(sg));
6137
		sg->sgp->power = 0;
6138

6139 6140 6141
		for_each_cpu(j, span) {
			if (get_group(j, sdd, NULL) != group)
				continue;
6142

6143 6144 6145
			cpumask_set_cpu(j, covered);
			cpumask_set_cpu(j, sched_group_cpus(sg));
		}
6146

6147 6148 6149 6150 6151 6152 6153
		if (!first)
			first = sg;
		if (last)
			last->next = sg;
		last = sg;
	}
	last->next = first;
6154 6155

	return 0;
6156
}
6157

6158 6159 6160 6161 6162 6163 6164 6165 6166 6167 6168 6169
/*
 * 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)
{
6170
	struct sched_group *sg = sd->groups;
6171

6172 6173 6174 6175 6176 6177
	WARN_ON(!sd || !sg);

	do {
		sg->group_weight = cpumask_weight(sched_group_cpus(sg));
		sg = sg->next;
	} while (sg != sd->groups);
6178

6179 6180
	if (cpu != group_first_cpu(sg))
		return;
6181

6182
	update_group_power(sd, cpu);
6183
	atomic_set(&sg->sgp->nr_busy_cpus, sg->group_weight);
6184 6185
}

6186 6187 6188
int __weak arch_sd_sibling_asym_packing(void)
{
       return 0*SD_ASYM_PACKING;
6189 6190
}

6191 6192 6193 6194 6195
/*
 * Initializers for schedule domains
 * Non-inlined to reduce accumulated stack pressure in build_sched_domains()
 */

6196 6197 6198 6199 6200 6201
#ifdef CONFIG_SCHED_DEBUG
# define SD_INIT_NAME(sd, type)		sd->name = #type
#else
# define SD_INIT_NAME(sd, type)		do { } while (0)
#endif

6202 6203 6204 6205 6206 6207 6208 6209 6210
#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;							\
6211 6212 6213 6214 6215 6216 6217 6218 6219 6220 6221 6222 6223
}

SD_INIT_FUNC(CPU)
#ifdef CONFIG_NUMA
 SD_INIT_FUNC(ALLNODES)
 SD_INIT_FUNC(NODE)
#endif
#ifdef CONFIG_SCHED_SMT
 SD_INIT_FUNC(SIBLING)
#endif
#ifdef CONFIG_SCHED_MC
 SD_INIT_FUNC(MC)
#endif
6224 6225 6226
#ifdef CONFIG_SCHED_BOOK
 SD_INIT_FUNC(BOOK)
#endif
6227

6228
static int default_relax_domain_level = -1;
6229
int sched_domain_level_max;
6230 6231 6232

static int __init setup_relax_domain_level(char *str)
{
6233 6234 6235
	unsigned long val;

	val = simple_strtoul(str, NULL, 0);
6236
	if (val < sched_domain_level_max)
6237 6238
		default_relax_domain_level = val;

6239 6240 6241 6242 6243 6244 6245 6246 6247 6248 6249 6250 6251 6252 6253 6254 6255 6256
	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 */
6257
		sd->flags &= ~(SD_BALANCE_WAKE|SD_BALANCE_NEWIDLE);
6258 6259
	} else {
		/* turn on idle balance on this domain */
6260
		sd->flags |= (SD_BALANCE_WAKE|SD_BALANCE_NEWIDLE);
6261 6262 6263
	}
}

6264 6265 6266
static void __sdt_free(const struct cpumask *cpu_map);
static int __sdt_alloc(const struct cpumask *cpu_map);

6267 6268 6269 6270 6271
static void __free_domain_allocs(struct s_data *d, enum s_alloc what,
				 const struct cpumask *cpu_map)
{
	switch (what) {
	case sa_rootdomain:
6272 6273
		if (!atomic_read(&d->rd->refcount))
			free_rootdomain(&d->rd->rcu); /* fall through */
6274 6275
	case sa_sd:
		free_percpu(d->sd); /* fall through */
6276
	case sa_sd_storage:
6277
		__sdt_free(cpu_map); /* fall through */
6278 6279 6280 6281
	case sa_none:
		break;
	}
}
6282

6283 6284 6285
static enum s_alloc __visit_domain_allocation_hell(struct s_data *d,
						   const struct cpumask *cpu_map)
{
6286 6287
	memset(d, 0, sizeof(*d));

6288 6289
	if (__sdt_alloc(cpu_map))
		return sa_sd_storage;
6290 6291 6292
	d->sd = alloc_percpu(struct sched_domain *);
	if (!d->sd)
		return sa_sd_storage;
6293
	d->rd = alloc_rootdomain();
6294
	if (!d->rd)
6295
		return sa_sd;
6296 6297
	return sa_rootdomain;
}
6298

6299 6300 6301 6302 6303 6304 6305 6306 6307 6308 6309 6310
/*
 * 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;

6311
	if (atomic_read(&(*per_cpu_ptr(sdd->sg, cpu))->ref))
6312
		*per_cpu_ptr(sdd->sg, cpu) = NULL;
6313 6314

	if (atomic_read(&(*per_cpu_ptr(sdd->sgp, cpu))->ref))
6315
		*per_cpu_ptr(sdd->sgp, cpu) = NULL;
6316 6317
}

6318 6319
#ifdef CONFIG_SCHED_SMT
static const struct cpumask *cpu_smt_mask(int cpu)
6320
{
6321
	return topology_thread_cpumask(cpu);
6322
}
6323
#endif
6324

6325 6326 6327
/*
 * Topology list, bottom-up.
 */
6328
static struct sched_domain_topology_level default_topology[] = {
6329 6330
#ifdef CONFIG_SCHED_SMT
	{ sd_init_SIBLING, cpu_smt_mask, },
6331
#endif
6332
#ifdef CONFIG_SCHED_MC
6333
	{ sd_init_MC, cpu_coregroup_mask, },
6334
#endif
6335 6336 6337 6338 6339
#ifdef CONFIG_SCHED_BOOK
	{ sd_init_BOOK, cpu_book_mask, },
#endif
	{ sd_init_CPU, cpu_cpu_mask, },
#ifdef CONFIG_NUMA
6340
	{ sd_init_NODE, cpu_node_mask, SDTL_OVERLAP, },
6341
	{ sd_init_ALLNODES, cpu_allnodes_mask, },
Linus Torvalds's avatar
Linus Torvalds committed
6342
#endif
6343 6344 6345 6346 6347
	{ NULL, },
};

static struct sched_domain_topology_level *sched_domain_topology = default_topology;

6348 6349 6350 6351 6352 6353 6354 6355 6356 6357 6358 6359 6360 6361 6362 6363
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;

6364 6365 6366 6367
		sdd->sgp = alloc_percpu(struct sched_group_power *);
		if (!sdd->sgp)
			return -ENOMEM;

6368 6369 6370
		for_each_cpu(j, cpu_map) {
			struct sched_domain *sd;
			struct sched_group *sg;
6371
			struct sched_group_power *sgp;
6372 6373 6374 6375 6376 6377 6378 6379 6380 6381 6382 6383 6384 6385

		       	sd = kzalloc_node(sizeof(struct sched_domain) + cpumask_size(),
					GFP_KERNEL, cpu_to_node(j));
			if (!sd)
				return -ENOMEM;

			*per_cpu_ptr(sdd->sd, j) = sd;

			sg = kzalloc_node(sizeof(struct sched_group) + cpumask_size(),
					GFP_KERNEL, cpu_to_node(j));
			if (!sg)
				return -ENOMEM;

			*per_cpu_ptr(sdd->sg, j) = sg;
6386 6387 6388 6389 6390 6391 6392

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

			*per_cpu_ptr(sdd->sgp, j) = sgp;
6393 6394 6395 6396 6397 6398 6399 6400 6401 6402 6403 6404 6405 6406 6407
		}
	}

	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) {
6408 6409 6410
			struct sched_domain *sd = *per_cpu_ptr(sdd->sd, j);
			if (sd && (sd->flags & SD_OVERLAP))
				free_sched_groups(sd->groups, 0);
6411
			kfree(*per_cpu_ptr(sdd->sd, j));
6412
			kfree(*per_cpu_ptr(sdd->sg, j));
6413
			kfree(*per_cpu_ptr(sdd->sgp, j));
6414 6415 6416
		}
		free_percpu(sdd->sd);
		free_percpu(sdd->sg);
6417
		free_percpu(sdd->sgp);
6418 6419 6420
	}
}

6421 6422
struct sched_domain *build_sched_domain(struct sched_domain_topology_level *tl,
		struct s_data *d, const struct cpumask *cpu_map,
6423
		struct sched_domain_attr *attr, struct sched_domain *child,
6424 6425
		int cpu)
{
6426
	struct sched_domain *sd = tl->init(tl, cpu);
6427
	if (!sd)
6428
		return child;
6429 6430 6431

	set_domain_attribute(sd, attr);
	cpumask_and(sched_domain_span(sd), cpu_map, tl->mask(cpu));
6432 6433 6434
	if (child) {
		sd->level = child->level + 1;
		sched_domain_level_max = max(sched_domain_level_max, sd->level);
6435
		child->parent = sd;
6436
	}
6437
	sd->child = child;
6438 6439 6440 6441

	return sd;
}

6442 6443 6444 6445
/*
 * Build sched domains for a given set of cpus and attach the sched domains
 * to the individual cpus
 */
6446 6447
static int build_sched_domains(const struct cpumask *cpu_map,
			       struct sched_domain_attr *attr)
6448 6449
{
	enum s_alloc alloc_state = sa_none;
6450
	struct sched_domain *sd;
6451
	struct s_data d;
6452
	int i, ret = -ENOMEM;
6453

6454 6455 6456
	alloc_state = __visit_domain_allocation_hell(&d, cpu_map);
	if (alloc_state != sa_rootdomain)
		goto error;
6457

6458
	/* Set up domains for cpus specified by the cpu_map. */
6459
	for_each_cpu(i, cpu_map) {
6460 6461
		struct sched_domain_topology_level *tl;

6462
		sd = NULL;
6463
		for (tl = sched_domain_topology; tl->init; tl++) {
6464
			sd = build_sched_domain(tl, &d, cpu_map, attr, sd, i);
6465 6466
			if (tl->flags & SDTL_OVERLAP || sched_feat(FORCE_SD_OVERLAP))
				sd->flags |= SD_OVERLAP;
6467 6468
			if (cpumask_equal(cpu_map, sched_domain_span(sd)))
				break;
6469
		}
6470

6471 6472 6473
		while (sd->child)
			sd = sd->child;

6474
		*per_cpu_ptr(d.sd, i) = sd;
6475 6476 6477 6478 6479 6480
	}

	/* 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));
6481 6482 6483 6484 6485 6486 6487
			if (sd->flags & SD_OVERLAP) {
				if (build_overlap_sched_groups(sd, i))
					goto error;
			} else {
				if (build_sched_groups(sd, i))
					goto error;
			}
6488
		}
6489
	}
6490

Linus Torvalds's avatar
Linus Torvalds committed
6491
	/* Calculate CPU power for physical packages and nodes */
6492 6493 6494
	for (i = nr_cpumask_bits-1; i >= 0; i--) {
		if (!cpumask_test_cpu(i, cpu_map))
			continue;
6495

6496 6497
		for (sd = *per_cpu_ptr(d.sd, i); sd; sd = sd->parent) {
			claim_allocations(i, sd);
6498
			init_sched_groups_power(i, sd);
6499
		}
6500
	}
6501

Linus Torvalds's avatar
Linus Torvalds committed
6502
	/* Attach the domains */
6503
	rcu_read_lock();
6504
	for_each_cpu(i, cpu_map) {
6505
		sd = *per_cpu_ptr(d.sd, i);
6506
		cpu_attach_domain(sd, d.rd, i);
Linus Torvalds's avatar
Linus Torvalds committed
6507
	}
6508
	rcu_read_unlock();
6509

6510
	ret = 0;
6511
error:
6512
	__free_domain_allocs(&d, alloc_state, cpu_map);
6513
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
6514
}
6515

6516
static cpumask_var_t *doms_cur;	/* current sched domains */
6517
static int ndoms_cur;		/* number of sched domains in 'doms_cur' */
Ingo Molnar's avatar
Ingo Molnar committed
6518 6519
static struct sched_domain_attr *dattr_cur;
				/* attribues of custom domains in 'doms_cur' */
6520 6521 6522

/*
 * Special case: If a kmalloc of a doms_cur partition (array of
6523 6524
 * cpumask) fails, then fallback to a single sched domain,
 * as determined by the single cpumask fallback_doms.
6525
 */
6526
static cpumask_var_t fallback_doms;
6527

6528 6529 6530 6531 6532 6533
/*
 * 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)
6534
{
6535
	return 0;
6536 6537
}

6538 6539 6540 6541 6542 6543 6544 6545 6546 6547 6548 6549 6550 6551 6552 6553 6554 6555 6556 6557 6558 6559 6560 6561 6562
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);
}

6563
/*
Ingo Molnar's avatar
Ingo Molnar committed
6564
 * Set up scheduler domains and groups. Callers must hold the hotplug lock.
6565 6566
 * For now this just excludes isolated cpus, but could be used to
 * exclude other special cases in the future.
6567
 */
6568
static int init_sched_domains(const struct cpumask *cpu_map)
6569
{
6570 6571
	int err;

6572
	arch_update_cpu_topology();
6573
	ndoms_cur = 1;
6574
	doms_cur = alloc_sched_domains(ndoms_cur);
6575
	if (!doms_cur)
6576 6577
		doms_cur = &fallback_doms;
	cpumask_andnot(doms_cur[0], cpu_map, cpu_isolated_map);
6578
	dattr_cur = NULL;
6579
	err = build_sched_domains(doms_cur[0], NULL);
6580
	register_sched_domain_sysctl();
6581 6582

	return err;
6583 6584 6585 6586 6587 6588
}

/*
 * Detach sched domains from a group of cpus specified in cpu_map
 * These cpus will now be attached to the NULL domain
 */
6589
static void detach_destroy_domains(const struct cpumask *cpu_map)
6590 6591 6592
{
	int i;

6593
	rcu_read_lock();
6594
	for_each_cpu(i, cpu_map)
6595
		cpu_attach_domain(NULL, &def_root_domain, i);
6596
	rcu_read_unlock();
6597 6598
}

6599 6600 6601 6602 6603 6604 6605 6606 6607 6608 6609 6610 6611 6612 6613 6614
/* 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));
}

6615 6616
/*
 * Partition sched domains as specified by the 'ndoms_new'
Ingo Molnar's avatar
Ingo Molnar committed
6617
 * cpumasks in the array doms_new[] of cpumasks. This compares
6618 6619 6620
 * doms_new[] to the current sched domain partitioning, doms_cur[].
 * It destroys each deleted domain and builds each new domain.
 *
6621
 * 'doms_new' is an array of cpumask_var_t's of length 'ndoms_new'.
Ingo Molnar's avatar
Ingo Molnar committed
6622 6623 6624
 * 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
6625 6626 6627
 * current 'doms_cur' domains and in the new 'doms_new', we can leave
 * it as it is.
 *
6628 6629 6630 6631 6632 6633
 * 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.
6634
 *
6635
 * If doms_new == NULL it will be replaced with cpu_online_mask.
6636 6637
 * ndoms_new == 0 is a special case for destroying existing domains,
 * and it will not create the default domain.
6638
 *
6639 6640
 * Call with hotplug lock held
 */
6641
void partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
6642
			     struct sched_domain_attr *dattr_new)
6643
{
6644
	int i, j, n;
6645
	int new_topology;
6646

6647
	mutex_lock(&sched_domains_mutex);
6648

6649 6650 6651
	/* always unregister in case we don't destroy any domains */
	unregister_sched_domain_sysctl();

6652 6653 6654
	/* Let architecture update cpu core mappings. */
	new_topology = arch_update_cpu_topology();

6655
	n = doms_new ? ndoms_new : 0;
6656 6657 6658

	/* Destroy deleted domains */
	for (i = 0; i < ndoms_cur; i++) {
6659
		for (j = 0; j < n && !new_topology; j++) {
6660
			if (cpumask_equal(doms_cur[i], doms_new[j])
6661
			    && dattrs_equal(dattr_cur, i, dattr_new, j))
6662 6663 6664
				goto match1;
		}
		/* no match - a current sched domain not in new doms_new[] */
6665
		detach_destroy_domains(doms_cur[i]);
6666 6667 6668 6669
match1:
		;
	}

6670 6671
	if (doms_new == NULL) {
		ndoms_cur = 0;
6672
		doms_new = &fallback_doms;
6673
		cpumask_andnot(doms_new[0], cpu_active_mask, cpu_isolated_map);
6674
		WARN_ON_ONCE(dattr_new);
6675 6676
	}

6677 6678
	/* Build new domains */
	for (i = 0; i < ndoms_new; i++) {
6679
		for (j = 0; j < ndoms_cur && !new_topology; j++) {
6680
			if (cpumask_equal(doms_new[i], doms_cur[j])
6681
			    && dattrs_equal(dattr_new, i, dattr_cur, j))
6682 6683 6684
				goto match2;
		}
		/* no match - add a new doms_new */
6685
		build_sched_domains(doms_new[i], dattr_new ? dattr_new + i : NULL);
6686 6687 6688 6689 6690
match2:
		;
	}

	/* Remember the new sched domains */
6691 6692
	if (doms_cur != &fallback_doms)
		free_sched_domains(doms_cur, ndoms_cur);
6693
	kfree(dattr_cur);	/* kfree(NULL) is safe */
6694
	doms_cur = doms_new;
6695
	dattr_cur = dattr_new;
6696
	ndoms_cur = ndoms_new;
6697 6698

	register_sched_domain_sysctl();
6699

6700
	mutex_unlock(&sched_domains_mutex);
6701 6702
}

6703
#if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT)
6704
static void reinit_sched_domains(void)
6705
{
6706
	get_online_cpus();
6707 6708 6709 6710

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

6711
	rebuild_sched_domains();
6712
	put_online_cpus();
6713 6714 6715 6716
}

static ssize_t sched_power_savings_store(const char *buf, size_t count, int smt)
{
6717
	unsigned int level = 0;
6718

6719 6720 6721 6722 6723 6724 6725 6726 6727 6728 6729
	if (sscanf(buf, "%u", &level) != 1)
		return -EINVAL;

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

	if (level >= MAX_POWERSAVINGS_BALANCE_LEVELS)
6730 6731 6732
		return -EINVAL;

	if (smt)
6733
		sched_smt_power_savings = level;
6734
	else
6735
		sched_mc_power_savings = level;
6736

6737
	reinit_sched_domains();
6738

6739
	return count;
6740 6741 6742
}

#ifdef CONFIG_SCHED_MC
6743 6744 6745
static ssize_t sched_mc_power_savings_show(struct device *dev,
					   struct device_attribute *attr,
					   char *buf)
6746
{
6747
	return sprintf(buf, "%u\n", sched_mc_power_savings);
6748
}
6749 6750
static ssize_t sched_mc_power_savings_store(struct device *dev,
					    struct device_attribute *attr,
6751
					    const char *buf, size_t count)
6752 6753 6754
{
	return sched_power_savings_store(buf, count, 0);
}
6755 6756 6757
static DEVICE_ATTR(sched_mc_power_savings, 0644,
		   sched_mc_power_savings_show,
		   sched_mc_power_savings_store);
6758 6759 6760
#endif

#ifdef CONFIG_SCHED_SMT
6761 6762 6763
static ssize_t sched_smt_power_savings_show(struct device *dev,
					    struct device_attribute *attr,
					    char *buf)
6764
{
6765
	return sprintf(buf, "%u\n", sched_smt_power_savings);
6766
}
6767 6768
static ssize_t sched_smt_power_savings_store(struct device *dev,
					    struct device_attribute *attr,
6769
					     const char *buf, size_t count)
6770 6771 6772
{
	return sched_power_savings_store(buf, count, 1);
}
6773
static DEVICE_ATTR(sched_smt_power_savings, 0644,
6774
		   sched_smt_power_savings_show,
Adrian Bunk's avatar
Adrian Bunk committed
6775 6776 6777
		   sched_smt_power_savings_store);
#endif

6778
int __init sched_create_sysfs_power_savings_entries(struct device *dev)
Adrian Bunk's avatar
Adrian Bunk committed
6779 6780 6781 6782 6783
{
	int err = 0;

#ifdef CONFIG_SCHED_SMT
	if (smt_capable())
6784
		err = device_create_file(dev, &dev_attr_sched_smt_power_savings);
Adrian Bunk's avatar
Adrian Bunk committed
6785 6786 6787
#endif
#ifdef CONFIG_SCHED_MC
	if (!err && mc_capable())
6788
		err = device_create_file(dev, &dev_attr_sched_mc_power_savings);
Adrian Bunk's avatar
Adrian Bunk committed
6789 6790 6791
#endif
	return err;
}
6792
#endif /* CONFIG_SCHED_MC || CONFIG_SCHED_SMT */
6793

Linus Torvalds's avatar
Linus Torvalds committed
6794
/*
6795 6796 6797
 * Update cpusets according to cpu_active mask.  If cpusets are
 * disabled, cpuset_update_active_cpus() becomes a simple wrapper
 * around partition_sched_domains().
Linus Torvalds's avatar
Linus Torvalds committed
6798
 */
6799 6800
static int cpuset_cpu_active(struct notifier_block *nfb, unsigned long action,
			     void *hcpu)
6801
{
6802
	switch (action & ~CPU_TASKS_FROZEN) {
6803
	case CPU_ONLINE:
6804
	case CPU_DOWN_FAILED:
6805
		cpuset_update_active_cpus();
6806
		return NOTIFY_OK;
6807 6808 6809 6810
	default:
		return NOTIFY_DONE;
	}
}
6811

6812 6813
static int cpuset_cpu_inactive(struct notifier_block *nfb, unsigned long action,
			       void *hcpu)
6814 6815 6816 6817 6818
{
	switch (action & ~CPU_TASKS_FROZEN) {
	case CPU_DOWN_PREPARE:
		cpuset_update_active_cpus();
		return NOTIFY_OK;
6819 6820 6821 6822 6823
	default:
		return NOTIFY_DONE;
	}
}

Linus Torvalds's avatar
Linus Torvalds committed
6824 6825
void __init sched_init_smp(void)
{
6826 6827 6828
	cpumask_var_t non_isolated_cpus;

	alloc_cpumask_var(&non_isolated_cpus, GFP_KERNEL);
6829
	alloc_cpumask_var(&fallback_doms, GFP_KERNEL);
6830

6831
	get_online_cpus();
6832
	mutex_lock(&sched_domains_mutex);
6833
	init_sched_domains(cpu_active_mask);
6834 6835 6836
	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);
6837
	mutex_unlock(&sched_domains_mutex);
6838
	put_online_cpus();
6839

6840 6841
	hotcpu_notifier(cpuset_cpu_active, CPU_PRI_CPUSET_ACTIVE);
	hotcpu_notifier(cpuset_cpu_inactive, CPU_PRI_CPUSET_INACTIVE);
6842 6843 6844 6845

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

6846
	init_hrtick();
6847 6848

	/* Move init over to a non-isolated CPU */
6849
	if (set_cpus_allowed_ptr(current, non_isolated_cpus) < 0)
6850
		BUG();
6851
	sched_init_granularity();
6852
	free_cpumask_var(non_isolated_cpus);
6853

6854
	init_sched_rt_class();
Linus Torvalds's avatar
Linus Torvalds committed
6855 6856 6857 6858
}
#else
void __init sched_init_smp(void)
{
6859
	sched_init_granularity();
Linus Torvalds's avatar
Linus Torvalds committed
6860 6861 6862
}
#endif /* CONFIG_SMP */

6863 6864
const_debug unsigned int sysctl_timer_migration = 1;

Linus Torvalds's avatar
Linus Torvalds committed
6865 6866 6867 6868 6869 6870 6871
int in_sched_functions(unsigned long addr)
{
	return in_lock_functions(addr) ||
		(addr >= (unsigned long)__sched_text_start
		&& addr < (unsigned long)__sched_text_end);
}

6872 6873
#ifdef CONFIG_CGROUP_SCHED
struct task_group root_task_group;
6874
#endif
Peter Zijlstra's avatar
Peter Zijlstra committed
6875

6876
DECLARE_PER_CPU(cpumask_var_t, load_balance_tmpmask);
Peter Zijlstra's avatar
Peter Zijlstra committed
6877

Linus Torvalds's avatar
Linus Torvalds committed
6878 6879
void __init sched_init(void)
{
Ingo Molnar's avatar
Ingo Molnar committed
6880
	int i, j;
6881 6882 6883 6884 6885 6886 6887
	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 **);
6888
#endif
6889
#ifdef CONFIG_CPUMASK_OFFSTACK
6890
	alloc_size += num_possible_cpus() * cpumask_size();
6891 6892
#endif
	if (alloc_size) {
6893
		ptr = (unsigned long)kzalloc(alloc_size, GFP_NOWAIT);
6894 6895

#ifdef CONFIG_FAIR_GROUP_SCHED
6896
		root_task_group.se = (struct sched_entity **)ptr;
6897 6898
		ptr += nr_cpu_ids * sizeof(void **);

6899
		root_task_group.cfs_rq = (struct cfs_rq **)ptr;
6900
		ptr += nr_cpu_ids * sizeof(void **);
6901

6902
#endif /* CONFIG_FAIR_GROUP_SCHED */
6903
#ifdef CONFIG_RT_GROUP_SCHED
6904
		root_task_group.rt_se = (struct sched_rt_entity **)ptr;
6905 6906
		ptr += nr_cpu_ids * sizeof(void **);

6907
		root_task_group.rt_rq = (struct rt_rq **)ptr;
6908 6909
		ptr += nr_cpu_ids * sizeof(void **);

6910
#endif /* CONFIG_RT_GROUP_SCHED */
6911 6912 6913 6914 6915 6916
#ifdef CONFIG_CPUMASK_OFFSTACK
		for_each_possible_cpu(i) {
			per_cpu(load_balance_tmpmask, i) = (void *)ptr;
			ptr += cpumask_size();
		}
#endif /* CONFIG_CPUMASK_OFFSTACK */
6917
	}
Ingo Molnar's avatar
Ingo Molnar committed
6918

6919 6920 6921 6922
#ifdef CONFIG_SMP
	init_defrootdomain();
#endif

6923 6924 6925 6926
	init_rt_bandwidth(&def_rt_bandwidth,
			global_rt_period(), global_rt_runtime());

#ifdef CONFIG_RT_GROUP_SCHED
6927
	init_rt_bandwidth(&root_task_group.rt_bandwidth,
6928
			global_rt_period(), global_rt_runtime());
6929
#endif /* CONFIG_RT_GROUP_SCHED */
6930

Dhaval Giani's avatar
Dhaval Giani committed
6931
#ifdef CONFIG_CGROUP_SCHED
6932 6933
	list_add(&root_task_group.list, &task_groups);
	INIT_LIST_HEAD(&root_task_group.children);
6934
	INIT_LIST_HEAD(&root_task_group.siblings);
6935
	autogroup_init(&init_task);
6936

Dhaval Giani's avatar
Dhaval Giani committed
6937
#endif /* CONFIG_CGROUP_SCHED */
Peter Zijlstra's avatar
Peter Zijlstra committed
6938

6939 6940 6941 6942 6943 6944
#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
6945
	for_each_possible_cpu(i) {
6946
		struct rq *rq;
Linus Torvalds's avatar
Linus Torvalds committed
6947 6948

		rq = cpu_rq(i);
6949
		raw_spin_lock_init(&rq->lock);
Nick Piggin's avatar
Nick Piggin committed
6950
		rq->nr_running = 0;
6951 6952
		rq->calc_load_active = 0;
		rq->calc_load_update = jiffies + LOAD_FREQ;
6953
		init_cfs_rq(&rq->cfs);
Peter Zijlstra's avatar
Peter Zijlstra committed
6954
		init_rt_rq(&rq->rt, rq);
Ingo Molnar's avatar
Ingo Molnar committed
6955
#ifdef CONFIG_FAIR_GROUP_SCHED
6956
		root_task_group.shares = ROOT_TASK_GROUP_LOAD;
Peter Zijlstra's avatar
Peter Zijlstra committed
6957
		INIT_LIST_HEAD(&rq->leaf_cfs_rq_list);
Dhaval Giani's avatar
Dhaval Giani committed
6958
		/*
6959
		 * How much cpu bandwidth does root_task_group get?
Dhaval Giani's avatar
Dhaval Giani committed
6960 6961 6962 6963
		 *
		 * 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
6964
		 * root_task_group and its child task-groups in a fair manner,
Dhaval Giani's avatar
Dhaval Giani committed
6965 6966 6967
		 * based on each entity's (task or task-group's) weight
		 * (se->load.weight).
		 *
6968
		 * In other words, if root_task_group has 10 tasks of weight
Dhaval Giani's avatar
Dhaval Giani committed
6969 6970 6971
		 * 1024) and two child groups A0 and A1 (of weight 1024 each),
		 * then A0's share of the cpu resource is:
		 *
6972
		 *	A0's bandwidth = 1024 / (10*1024 + 1024 + 1024) = 8.33%
Dhaval Giani's avatar
Dhaval Giani committed
6973
		 *
6974 6975
		 * We achieve this by letting root_task_group's tasks sit
		 * directly in rq->cfs (i.e root_task_group->se[] = NULL).
Dhaval Giani's avatar
Dhaval Giani committed
6976
		 */
6977
		init_cfs_bandwidth(&root_task_group.cfs_bandwidth);
6978
		init_tg_cfs_entry(&root_task_group, &rq->cfs, NULL, i, NULL);
Dhaval Giani's avatar
Dhaval Giani committed
6979 6980 6981
#endif /* CONFIG_FAIR_GROUP_SCHED */

		rq->rt.rt_runtime = def_rt_bandwidth.rt_runtime;
6982
#ifdef CONFIG_RT_GROUP_SCHED
Peter Zijlstra's avatar
Peter Zijlstra committed
6983
		INIT_LIST_HEAD(&rq->leaf_rt_rq_list);
6984
		init_tg_rt_entry(&root_task_group, &rq->rt, NULL, i, NULL);
Ingo Molnar's avatar
Ingo Molnar committed
6985
#endif
Linus Torvalds's avatar
Linus Torvalds committed
6986

Ingo Molnar's avatar
Ingo Molnar committed
6987 6988
		for (j = 0; j < CPU_LOAD_IDX_MAX; j++)
			rq->cpu_load[j] = 0;
6989 6990 6991

		rq->last_load_update_tick = jiffies;

Linus Torvalds's avatar
Linus Torvalds committed
6992
#ifdef CONFIG_SMP
Nick Piggin's avatar
Nick Piggin committed
6993
		rq->sd = NULL;
6994
		rq->rd = NULL;
6995
		rq->cpu_power = SCHED_POWER_SCALE;
6996
		rq->post_schedule = 0;
Linus Torvalds's avatar
Linus Torvalds committed
6997
		rq->active_balance = 0;
Ingo Molnar's avatar
Ingo Molnar committed
6998
		rq->next_balance = jiffies;
Linus Torvalds's avatar
Linus Torvalds committed
6999
		rq->push_cpu = 0;
7000
		rq->cpu = i;
7001
		rq->online = 0;
7002 7003
		rq->idle_stamp = 0;
		rq->avg_idle = 2*sysctl_sched_migration_cost;
7004 7005 7006

		INIT_LIST_HEAD(&rq->cfs_tasks);

7007
		rq_attach_root(rq, &def_root_domain);
7008
#ifdef CONFIG_NO_HZ
7009
		rq->nohz_flags = 0;
7010
#endif
Linus Torvalds's avatar
Linus Torvalds committed
7011
#endif
7012
		init_rq_hrtick(rq);
Linus Torvalds's avatar
Linus Torvalds committed
7013 7014 7015
		atomic_set(&rq->nr_iowait, 0);
	}

7016
	set_load_weight(&init_task);
7017

7018 7019 7020 7021
#ifdef CONFIG_PREEMPT_NOTIFIERS
	INIT_HLIST_HEAD(&init_task.preempt_notifiers);
#endif

7022
#ifdef CONFIG_RT_MUTEXES
7023
	plist_head_init(&init_task.pi_waiters);
7024 7025
#endif

Linus Torvalds's avatar
Linus Torvalds committed
7026 7027 7028 7029 7030 7031 7032 7033 7034 7035 7036 7037 7038
	/*
	 * 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());
7039 7040 7041

	calc_load_update = jiffies + LOAD_FREQ;

Ingo Molnar's avatar
Ingo Molnar committed
7042 7043 7044 7045
	/*
	 * During early bootup we pretend to be a normal task:
	 */
	current->sched_class = &fair_sched_class;
7046

7047
#ifdef CONFIG_SMP
7048
	zalloc_cpumask_var(&sched_domains_tmpmask, GFP_NOWAIT);
7049 7050 7051
	/* May be allocated at isolcpus cmdline parse time */
	if (cpu_isolated_map == NULL)
		zalloc_cpumask_var(&cpu_isolated_map, GFP_NOWAIT);
7052 7053
#endif
	init_sched_fair_class();
7054

7055
	scheduler_running = 1;
Linus Torvalds's avatar
Linus Torvalds committed
7056 7057
}

7058
#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
7059 7060
static inline int preempt_count_equals(int preempt_offset)
{
7061
	int nested = (preempt_count() & ~PREEMPT_ACTIVE) + rcu_preempt_depth();
7062

Arnd Bergmann's avatar
Arnd Bergmann committed
7063
	return (nested == preempt_offset);
7064 7065
}

7066
void __might_sleep(const char *file, int line, int preempt_offset)
Linus Torvalds's avatar
Linus Torvalds committed
7067 7068 7069
{
	static unsigned long prev_jiffy;	/* ratelimiting */

7070
	rcu_sleep_check(); /* WARN_ON_ONCE() by default, no rate limit reqd. */
7071 7072
	if ((preempt_count_equals(preempt_offset) && !irqs_disabled()) ||
	    system_state != SYSTEM_RUNNING || oops_in_progress)
7073 7074 7075 7076 7077
		return;
	if (time_before(jiffies, prev_jiffy + HZ) && prev_jiffy)
		return;
	prev_jiffy = jiffies;

7078 7079 7080 7081 7082 7083 7084
	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);
7085 7086 7087 7088 7089

	debug_show_held_locks(current);
	if (irqs_disabled())
		print_irqtrace_events(current);
	dump_stack();
Linus Torvalds's avatar
Linus Torvalds committed
7090 7091 7092 7093 7094
}
EXPORT_SYMBOL(__might_sleep);
#endif

#ifdef CONFIG_MAGIC_SYSRQ
7095 7096
static void normalize_task(struct rq *rq, struct task_struct *p)
{
7097 7098
	const struct sched_class *prev_class = p->sched_class;
	int old_prio = p->prio;
7099
	int on_rq;
7100

Peter Zijlstra's avatar
Peter Zijlstra committed
7101
	on_rq = p->on_rq;
7102
	if (on_rq)
7103
		dequeue_task(rq, p, 0);
7104 7105
	__setscheduler(rq, p, SCHED_NORMAL, 0);
	if (on_rq) {
7106
		enqueue_task(rq, p, 0);
7107 7108
		resched_task(rq->curr);
	}
7109 7110

	check_class_changed(rq, p, prev_class, old_prio);
7111 7112
}

Linus Torvalds's avatar
Linus Torvalds committed
7113 7114
void normalize_rt_tasks(void)
{
7115
	struct task_struct *g, *p;
Linus Torvalds's avatar
Linus Torvalds committed
7116
	unsigned long flags;
7117
	struct rq *rq;
Linus Torvalds's avatar
Linus Torvalds committed
7118

7119
	read_lock_irqsave(&tasklist_lock, flags);
7120
	do_each_thread(g, p) {
7121 7122 7123 7124 7125 7126
		/*
		 * Only normalize user tasks:
		 */
		if (!p->mm)
			continue;

7127 7128
		p->se.exec_start		= 0;
#ifdef CONFIG_SCHEDSTATS
7129 7130 7131
		p->se.statistics.wait_start	= 0;
		p->se.statistics.sleep_start	= 0;
		p->se.statistics.block_start	= 0;
7132
#endif
Ingo Molnar's avatar
Ingo Molnar committed
7133 7134 7135 7136 7137 7138 7139 7140

		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);
Linus Torvalds's avatar
Linus Torvalds committed
7141
			continue;
Ingo Molnar's avatar
Ingo Molnar committed
7142
		}
Linus Torvalds's avatar
Linus Torvalds committed
7143

7144
		raw_spin_lock(&p->pi_lock);
7145
		rq = __task_rq_lock(p);
Linus Torvalds's avatar
Linus Torvalds committed
7146

7147
		normalize_task(rq, p);
7148

7149
		__task_rq_unlock(rq);
7150
		raw_spin_unlock(&p->pi_lock);
7151 7152
	} while_each_thread(g, p);

7153
	read_unlock_irqrestore(&tasklist_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
7154 7155 7156
}

#endif /* CONFIG_MAGIC_SYSRQ */
7157

7158
#if defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB)
7159
/*
7160
 * These functions are only useful for the IA64 MCA handling, or kdb.
7161 7162 7163 7164 7165 7166 7167 7168 7169 7170 7171 7172 7173 7174
 *
 * 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!
 */
7175
struct task_struct *curr_task(int cpu)
7176 7177 7178 7179
{
	return cpu_curr(cpu);
}

7180 7181 7182
#endif /* defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB) */

#ifdef CONFIG_IA64
7183 7184 7185 7186 7187 7188
/**
 * 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
Ingo Molnar's avatar
Ingo Molnar committed
7189 7190
 * 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
7191 7192 7193 7194 7195 7196 7197
 * 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!
 */
7198
void set_curr_task(int cpu, struct task_struct *p)
7199 7200 7201 7202 7203
{
	cpu_curr(cpu) = p;
}

#endif
7204

Dhaval Giani's avatar
Dhaval Giani committed
7205
#ifdef CONFIG_CGROUP_SCHED
7206 7207 7208
/* task_group_lock serializes the addition/removal of task groups */
static DEFINE_SPINLOCK(task_group_lock);

7209 7210 7211 7212
static void free_sched_group(struct task_group *tg)
{
	free_fair_sched_group(tg);
	free_rt_sched_group(tg);
7213
	autogroup_free(tg);
7214 7215 7216 7217
	kfree(tg);
}

/* allocate runqueue etc for a new task group */
7218
struct task_group *sched_create_group(struct task_group *parent)
7219 7220 7221 7222 7223 7224 7225 7226
{
	struct task_group *tg;
	unsigned long flags;

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

7227
	if (!alloc_fair_sched_group(tg, parent))
7228 7229
		goto err;

7230
	if (!alloc_rt_sched_group(tg, parent))
7231 7232
		goto err;

7233
	spin_lock_irqsave(&task_group_lock, flags);
Peter Zijlstra's avatar
Peter Zijlstra committed
7234
	list_add_rcu(&tg->list, &task_groups);
Peter Zijlstra's avatar
Peter Zijlstra committed
7235 7236 7237 7238 7239

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

	tg->parent = parent;
	INIT_LIST_HEAD(&tg->children);
7240
	list_add_rcu(&tg->siblings, &parent->children);
7241
	spin_unlock_irqrestore(&task_group_lock, flags);
7242

7243
	return tg;
7244 7245

err:
Peter Zijlstra's avatar
Peter Zijlstra committed
7246
	free_sched_group(tg);
7247 7248 7249
	return ERR_PTR(-ENOMEM);
}

7250
/* rcu callback to free various structures associated with a task group */
Peter Zijlstra's avatar
Peter Zijlstra committed
7251
static void free_sched_group_rcu(struct rcu_head *rhp)
7252 7253
{
	/* now it should be safe to free those cfs_rqs */
Peter Zijlstra's avatar
Peter Zijlstra committed
7254
	free_sched_group(container_of(rhp, struct task_group, rcu));
7255 7256
}

7257
/* Destroy runqueue etc associated with a task group */
7258
void sched_destroy_group(struct task_group *tg)
7259
{
7260
	unsigned long flags;
7261
	int i;
7262

7263 7264
	/* end participation in shares distribution */
	for_each_possible_cpu(i)
7265
		unregister_fair_sched_group(tg, i);
7266 7267

	spin_lock_irqsave(&task_group_lock, flags);
Peter Zijlstra's avatar
Peter Zijlstra committed
7268
	list_del_rcu(&tg->list);
Peter Zijlstra's avatar
Peter Zijlstra committed
7269
	list_del_rcu(&tg->siblings);
7270
	spin_unlock_irqrestore(&task_group_lock, flags);
7271 7272

	/* wait for possible concurrent references to cfs_rqs complete */
Peter Zijlstra's avatar
Peter Zijlstra committed
7273
	call_rcu(&tg->rcu, free_sched_group_rcu);
7274 7275
}

7276
/* change task's runqueue when it moves between groups.
Ingo Molnar's avatar
Ingo Molnar committed
7277 7278 7279
 *	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.
7280 7281
 */
void sched_move_task(struct task_struct *tsk)
7282 7283 7284 7285 7286 7287 7288
{
	int on_rq, running;
	unsigned long flags;
	struct rq *rq;

	rq = task_rq_lock(tsk, &flags);

7289
	running = task_current(rq, tsk);
Peter Zijlstra's avatar
Peter Zijlstra committed
7290
	on_rq = tsk->on_rq;
7291

7292
	if (on_rq)
7293
		dequeue_task(rq, tsk, 0);
7294 7295
	if (unlikely(running))
		tsk->sched_class->put_prev_task(rq, tsk);
7296

Peter Zijlstra's avatar
Peter Zijlstra committed
7297
#ifdef CONFIG_FAIR_GROUP_SCHED
7298 7299 7300
	if (tsk->sched_class->task_move_group)
		tsk->sched_class->task_move_group(tsk, on_rq);
	else
Peter Zijlstra's avatar
Peter Zijlstra committed
7301
#endif
7302
		set_task_rq(tsk, task_cpu(tsk));
Peter Zijlstra's avatar
Peter Zijlstra committed
7303

7304 7305 7306
	if (unlikely(running))
		tsk->sched_class->set_curr_task(rq);
	if (on_rq)
7307
		enqueue_task(rq, tsk, 0);
7308

7309
	task_rq_unlock(rq, tsk, &flags);
7310
}
Dhaval Giani's avatar
Dhaval Giani committed
7311
#endif /* CONFIG_CGROUP_SCHED */
7312

7313
#if defined(CONFIG_RT_GROUP_SCHED) || defined(CONFIG_CFS_BANDWIDTH)
Peter Zijlstra's avatar
Peter Zijlstra committed
7314 7315 7316
static unsigned long to_ratio(u64 period, u64 runtime)
{
	if (runtime == RUNTIME_INF)
Peter Zijlstra's avatar
Peter Zijlstra committed
7317
		return 1ULL << 20;
Peter Zijlstra's avatar
Peter Zijlstra committed
7318

Peter Zijlstra's avatar
Peter Zijlstra committed
7319
	return div64_u64(runtime << 20, period);
Peter Zijlstra's avatar
Peter Zijlstra committed
7320
}
7321 7322 7323 7324 7325 7326 7327
#endif

#ifdef CONFIG_RT_GROUP_SCHED
/*
 * Ensure that the real time constraints are schedulable.
 */
static DEFINE_MUTEX(rt_constraints_mutex);
Peter Zijlstra's avatar
Peter Zijlstra committed
7328

Peter Zijlstra's avatar
Peter Zijlstra committed
7329 7330
/* Must be called with tasklist_lock held */
static inline int tg_has_rt_tasks(struct task_group *tg)
7331
{
Peter Zijlstra's avatar
Peter Zijlstra committed
7332
	struct task_struct *g, *p;
7333

Peter Zijlstra's avatar
Peter Zijlstra committed
7334
	do_each_thread(g, p) {
7335
		if (rt_task(p) && task_rq(p)->rt.tg == tg)
Peter Zijlstra's avatar
Peter Zijlstra committed
7336 7337
			return 1;
	} while_each_thread(g, p);
7338

Peter Zijlstra's avatar
Peter Zijlstra committed
7339 7340
	return 0;
}
7341

Peter Zijlstra's avatar
Peter Zijlstra committed
7342 7343 7344 7345 7346
struct rt_schedulable_data {
	struct task_group *tg;
	u64 rt_period;
	u64 rt_runtime;
};
7347

7348
static int tg_rt_schedulable(struct task_group *tg, void *data)
Peter Zijlstra's avatar
Peter Zijlstra committed
7349 7350 7351 7352 7353
{
	struct rt_schedulable_data *d = data;
	struct task_group *child;
	unsigned long total, sum = 0;
	u64 period, runtime;
7354

Peter Zijlstra's avatar
Peter Zijlstra committed
7355 7356
	period = ktime_to_ns(tg->rt_bandwidth.rt_period);
	runtime = tg->rt_bandwidth.rt_runtime;
7357

Peter Zijlstra's avatar
Peter Zijlstra committed
7358 7359 7360
	if (tg == d->tg) {
		period = d->rt_period;
		runtime = d->rt_runtime;
7361 7362
	}

7363 7364 7365 7366 7367
	/*
	 * Cannot have more runtime than the period.
	 */
	if (runtime > period && runtime != RUNTIME_INF)
		return -EINVAL;
Peter Zijlstra's avatar
Peter Zijlstra committed
7368

7369 7370 7371
	/*
	 * Ensure we don't starve existing RT tasks.
	 */
Peter Zijlstra's avatar
Peter Zijlstra committed
7372 7373
	if (rt_bandwidth_enabled() && !runtime && tg_has_rt_tasks(tg))
		return -EBUSY;
Peter Zijlstra's avatar
Peter Zijlstra committed
7374

Peter Zijlstra's avatar
Peter Zijlstra committed
7375
	total = to_ratio(period, runtime);
Peter Zijlstra's avatar
Peter Zijlstra committed
7376

7377 7378 7379 7380 7381
	/*
	 * Nobody can have more than the global setting allows.
	 */
	if (total > to_ratio(global_rt_period(), global_rt_runtime()))
		return -EINVAL;
Peter Zijlstra's avatar
Peter Zijlstra committed
7382

7383 7384 7385
	/*
	 * The sum of our children's runtime should not exceed our own.
	 */
Peter Zijlstra's avatar
Peter Zijlstra committed
7386 7387 7388
	list_for_each_entry_rcu(child, &tg->children, siblings) {
		period = ktime_to_ns(child->rt_bandwidth.rt_period);
		runtime = child->rt_bandwidth.rt_runtime;
Peter Zijlstra's avatar
Peter Zijlstra committed
7389

Peter Zijlstra's avatar
Peter Zijlstra committed
7390 7391 7392 7393
		if (child == d->tg) {
			period = d->rt_period;
			runtime = d->rt_runtime;
		}
Peter Zijlstra's avatar
Peter Zijlstra committed
7394

Peter Zijlstra's avatar
Peter Zijlstra committed
7395
		sum += to_ratio(period, runtime);
Peter Zijlstra's avatar
Peter Zijlstra committed
7396
	}
Peter Zijlstra's avatar
Peter Zijlstra committed
7397

Peter Zijlstra's avatar
Peter Zijlstra committed
7398 7399 7400 7401
	if (sum > total)
		return -EINVAL;

	return 0;
Peter Zijlstra's avatar
Peter Zijlstra committed
7402 7403
}

Peter Zijlstra's avatar
Peter Zijlstra committed
7404
static int __rt_schedulable(struct task_group *tg, u64 period, u64 runtime)
7405
{
7406 7407
	int ret;

Peter Zijlstra's avatar
Peter Zijlstra committed
7408 7409 7410 7411 7412 7413
	struct rt_schedulable_data data = {
		.tg = tg,
		.rt_period = period,
		.rt_runtime = runtime,
	};

7414 7415 7416 7417 7418
	rcu_read_lock();
	ret = walk_tg_tree(tg_rt_schedulable, tg_nop, &data);
	rcu_read_unlock();

	return ret;
7419 7420
}

7421
static int tg_set_rt_bandwidth(struct task_group *tg,
7422
		u64 rt_period, u64 rt_runtime)
Peter Zijlstra's avatar
Peter Zijlstra committed
7423
{
7424
	int i, err = 0;
Peter Zijlstra's avatar
Peter Zijlstra committed
7425 7426

	mutex_lock(&rt_constraints_mutex);
7427
	read_lock(&tasklist_lock);
Peter Zijlstra's avatar
Peter Zijlstra committed
7428 7429
	err = __rt_schedulable(tg, rt_period, rt_runtime);
	if (err)
Peter Zijlstra's avatar
Peter Zijlstra committed
7430
		goto unlock;
7431

7432
	raw_spin_lock_irq(&tg->rt_bandwidth.rt_runtime_lock);
7433 7434
	tg->rt_bandwidth.rt_period = ns_to_ktime(rt_period);
	tg->rt_bandwidth.rt_runtime = rt_runtime;
7435 7436 7437 7438

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

7439
		raw_spin_lock(&rt_rq->rt_runtime_lock);
7440
		rt_rq->rt_runtime = rt_runtime;
7441
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
7442
	}
7443
	raw_spin_unlock_irq(&tg->rt_bandwidth.rt_runtime_lock);
Peter Zijlstra's avatar
Peter Zijlstra committed
7444
unlock:
7445
	read_unlock(&tasklist_lock);
Peter Zijlstra's avatar
Peter Zijlstra committed
7446 7447 7448
	mutex_unlock(&rt_constraints_mutex);

	return err;
Peter Zijlstra's avatar
Peter Zijlstra committed
7449 7450
}

7451 7452 7453 7454 7455 7456 7457 7458 7459
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;

7460
	return tg_set_rt_bandwidth(tg, rt_period, rt_runtime);
7461 7462
}

Peter Zijlstra's avatar
Peter Zijlstra committed
7463 7464 7465 7466
long sched_group_rt_runtime(struct task_group *tg)
{
	u64 rt_runtime_us;

7467
	if (tg->rt_bandwidth.rt_runtime == RUNTIME_INF)
Peter Zijlstra's avatar
Peter Zijlstra committed
7468 7469
		return -1;

7470
	rt_runtime_us = tg->rt_bandwidth.rt_runtime;
Peter Zijlstra's avatar
Peter Zijlstra committed
7471 7472 7473
	do_div(rt_runtime_us, NSEC_PER_USEC);
	return rt_runtime_us;
}
7474 7475 7476 7477 7478 7479 7480 7481

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;

7482 7483 7484
	if (rt_period == 0)
		return -EINVAL;

7485
	return tg_set_rt_bandwidth(tg, rt_period, rt_runtime);
7486 7487 7488 7489 7490 7491 7492 7493 7494 7495 7496 7497 7498
}

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)
{
7499
	u64 runtime, period;
7500 7501
	int ret = 0;

7502 7503 7504
	if (sysctl_sched_rt_period <= 0)
		return -EINVAL;

7505 7506 7507 7508 7509 7510 7511 7512
	runtime = global_rt_runtime();
	period = global_rt_period();

	/*
	 * Sanity check on the sysctl variables.
	 */
	if (runtime > period && runtime != RUNTIME_INF)
		return -EINVAL;
7513

7514
	mutex_lock(&rt_constraints_mutex);
Peter Zijlstra's avatar
Peter Zijlstra committed
7515
	read_lock(&tasklist_lock);
7516
	ret = __rt_schedulable(NULL, 0, 0);
Peter Zijlstra's avatar
Peter Zijlstra committed
7517
	read_unlock(&tasklist_lock);
7518 7519 7520 7521
	mutex_unlock(&rt_constraints_mutex);

	return ret;
}
7522 7523 7524 7525 7526 7527 7528 7529 7530 7531

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;
}

7532
#else /* !CONFIG_RT_GROUP_SCHED */
7533 7534
static int sched_rt_global_constraints(void)
{
7535 7536 7537
	unsigned long flags;
	int i;

7538 7539 7540
	if (sysctl_sched_rt_period <= 0)
		return -EINVAL;

7541 7542 7543 7544 7545 7546 7547
	/*
	 * There's always some RT tasks in the root group
	 * -- migration, kstopmachine etc..
	 */
	if (sysctl_sched_rt_runtime == 0)
		return -EBUSY;

7548
	raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags);
7549 7550 7551
	for_each_possible_cpu(i) {
		struct rt_rq *rt_rq = &cpu_rq(i)->rt;

7552
		raw_spin_lock(&rt_rq->rt_runtime_lock);
7553
		rt_rq->rt_runtime = global_rt_runtime();
7554
		raw_spin_unlock(&rt_rq->rt_runtime_lock);
7555
	}
7556
	raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags);
7557

7558 7559
	return 0;
}
7560
#endif /* CONFIG_RT_GROUP_SCHED */
7561 7562

int sched_rt_handler(struct ctl_table *table, int write,
7563
		void __user *buffer, size_t *lenp,
7564 7565 7566 7567 7568 7569 7570 7571 7572 7573
		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;

7574
	ret = proc_dointvec(table, write, buffer, lenp, ppos);
7575 7576 7577 7578 7579 7580 7581 7582 7583 7584 7585 7586 7587 7588 7589 7590

	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;
}
7591

7592
#ifdef CONFIG_CGROUP_SCHED
7593 7594

/* return corresponding task_group object of a cgroup */
7595
static inline struct task_group *cgroup_tg(struct cgroup *cgrp)
7596
{
7597 7598
	return container_of(cgroup_subsys_state(cgrp, cpu_cgroup_subsys_id),
			    struct task_group, css);
7599 7600
}

7601
static struct cgroup_subsys_state *cpu_cgroup_create(struct cgroup *cgrp)
7602
{
7603
	struct task_group *tg, *parent;
7604

7605
	if (!cgrp->parent) {
7606
		/* This is early initialization for the top cgroup */
7607
		return &root_task_group.css;
7608 7609
	}

7610 7611
	parent = cgroup_tg(cgrp->parent);
	tg = sched_create_group(parent);
7612 7613 7614 7615 7616 7617
	if (IS_ERR(tg))
		return ERR_PTR(-ENOMEM);

	return &tg->css;
}

7618
static void cpu_cgroup_destroy(struct cgroup *cgrp)
7619
{
7620
	struct task_group *tg = cgroup_tg(cgrp);
7621 7622 7623 7624

	sched_destroy_group(tg);
}

7625
static int cpu_cgroup_can_attach(struct cgroup *cgrp,
7626
				 struct cgroup_taskset *tset)
7627
{
7628 7629 7630
	struct task_struct *task;

	cgroup_taskset_for_each(task, cgrp, tset) {
7631
#ifdef CONFIG_RT_GROUP_SCHED
7632 7633
		if (!sched_rt_can_attach(cgroup_tg(cgrp), task))
			return -EINVAL;
7634
#else
7635 7636 7637
		/* We don't support RT-tasks being in separate groups */
		if (task->sched_class != &fair_sched_class)
			return -EINVAL;
7638
#endif
7639
	}
7640 7641
	return 0;
}
7642

7643
static void cpu_cgroup_attach(struct cgroup *cgrp,
7644
			      struct cgroup_taskset *tset)
7645
{
7646 7647 7648 7649
	struct task_struct *task;

	cgroup_taskset_for_each(task, cgrp, tset)
		sched_move_task(task);
7650 7651
}

7652
static void
7653 7654
cpu_cgroup_exit(struct cgroup *cgrp, struct cgroup *old_cgrp,
		struct task_struct *task)
7655 7656 7657 7658 7659 7660 7661 7662 7663 7664 7665 7666
{
	/*
	 * 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);
}

7667
#ifdef CONFIG_FAIR_GROUP_SCHED
7668
static int cpu_shares_write_u64(struct cgroup *cgrp, struct cftype *cftype,
7669
				u64 shareval)
7670
{
7671
	return sched_group_set_shares(cgroup_tg(cgrp), scale_load(shareval));
7672 7673
}

7674
static u64 cpu_shares_read_u64(struct cgroup *cgrp, struct cftype *cft)
7675
{
7676
	struct task_group *tg = cgroup_tg(cgrp);
7677

7678
	return (u64) scale_load_down(tg->shares);
7679
}
7680 7681

#ifdef CONFIG_CFS_BANDWIDTH
7682 7683
static DEFINE_MUTEX(cfs_constraints_mutex);

7684 7685 7686
const u64 max_cfs_quota_period = 1 * NSEC_PER_SEC; /* 1s */
const u64 min_cfs_quota_period = 1 * NSEC_PER_MSEC; /* 1ms */

7687 7688
static int __cfs_schedulable(struct task_group *tg, u64 period, u64 runtime);

7689 7690
static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota)
{
7691
	int i, ret = 0, runtime_enabled, runtime_was_enabled;
7692
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
7693 7694 7695 7696 7697 7698 7699 7700 7701 7702 7703 7704 7705 7706 7707 7708 7709 7710 7711 7712

	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;

7713 7714 7715 7716 7717
	mutex_lock(&cfs_constraints_mutex);
	ret = __cfs_schedulable(tg, period, quota);
	if (ret)
		goto out_unlock;

7718
	runtime_enabled = quota != RUNTIME_INF;
7719 7720
	runtime_was_enabled = cfs_b->quota != RUNTIME_INF;
	account_cfs_bandwidth_used(runtime_enabled, runtime_was_enabled);
7721 7722 7723
	raw_spin_lock_irq(&cfs_b->lock);
	cfs_b->period = ns_to_ktime(period);
	cfs_b->quota = quota;
7724

Paul Turner's avatar
Paul Turner committed
7725
	__refill_cfs_bandwidth_runtime(cfs_b);
7726 7727 7728 7729 7730 7731
	/* 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);
	}
7732 7733 7734 7735
	raw_spin_unlock_irq(&cfs_b->lock);

	for_each_possible_cpu(i) {
		struct cfs_rq *cfs_rq = tg->cfs_rq[i];
7736
		struct rq *rq = cfs_rq->rq;
7737 7738

		raw_spin_lock_irq(&rq->lock);
7739
		cfs_rq->runtime_enabled = runtime_enabled;
7740
		cfs_rq->runtime_remaining = 0;
7741

7742
		if (cfs_rq->throttled)
7743
			unthrottle_cfs_rq(cfs_rq);
7744 7745
		raw_spin_unlock_irq(&rq->lock);
	}
7746 7747
out_unlock:
	mutex_unlock(&cfs_constraints_mutex);
7748

7749
	return ret;
7750 7751 7752 7753 7754 7755
}

int tg_set_cfs_quota(struct task_group *tg, long cfs_quota_us)
{
	u64 quota, period;

7756
	period = ktime_to_ns(tg->cfs_bandwidth.period);
7757 7758 7759 7760 7761 7762 7763 7764 7765 7766 7767 7768
	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;

7769
	if (tg->cfs_bandwidth.quota == RUNTIME_INF)
7770 7771
		return -1;

7772
	quota_us = tg->cfs_bandwidth.quota;
7773 7774 7775 7776 7777 7778 7779 7780 7781 7782
	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;
7783
	quota = tg->cfs_bandwidth.quota;
7784 7785 7786 7787 7788 7789 7790 7791

	return tg_set_cfs_bandwidth(tg, period, quota);
}

long tg_get_cfs_period(struct task_group *tg)
{
	u64 cfs_period_us;

7792
	cfs_period_us = ktime_to_ns(tg->cfs_bandwidth.period);
7793 7794 7795 7796 7797 7798 7799 7800 7801 7802 7803 7804 7805 7806 7807 7808 7809 7810 7811 7812 7813 7814 7815 7816 7817 7818 7819
	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);
}

7820 7821 7822 7823 7824 7825 7826 7827 7828 7829 7830 7831 7832 7833 7834 7835 7836 7837 7838 7839 7840 7841 7842 7843 7844 7845 7846 7847 7848 7849 7850 7851
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;
7852
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
7853 7854 7855 7856 7857
	s64 quota = 0, parent_quota = -1;

	if (!tg->parent) {
		quota = RUNTIME_INF;
	} else {
7858
		struct cfs_bandwidth *parent_b = &tg->parent->cfs_bandwidth;
7859 7860 7861 7862 7863 7864 7865 7866 7867 7868 7869 7870 7871 7872 7873 7874 7875 7876 7877 7878

		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)
{
7879
	int ret;
7880 7881 7882 7883 7884 7885 7886 7887 7888 7889 7890
	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);
	}

7891 7892 7893 7894 7895
	rcu_read_lock();
	ret = walk_tg_tree(tg_cfs_schedulable_down, tg_nop, &data);
	rcu_read_unlock();

	return ret;
7896
}
7897 7898 7899 7900 7901

static int cpu_stats_show(struct cgroup *cgrp, struct cftype *cft,
		struct cgroup_map_cb *cb)
{
	struct task_group *tg = cgroup_tg(cgrp);
7902
	struct cfs_bandwidth *cfs_b = &tg->cfs_bandwidth;
7903 7904 7905 7906 7907 7908 7909

	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;
}
7910
#endif /* CONFIG_CFS_BANDWIDTH */
7911
#endif /* CONFIG_FAIR_GROUP_SCHED */
7912

7913
#ifdef CONFIG_RT_GROUP_SCHED
7914
static int cpu_rt_runtime_write(struct cgroup *cgrp, struct cftype *cft,
7915
				s64 val)
Peter Zijlstra's avatar
Peter Zijlstra committed
7916
{
7917
	return sched_group_set_rt_runtime(cgroup_tg(cgrp), val);
Peter Zijlstra's avatar
Peter Zijlstra committed
7918 7919
}

7920
static s64 cpu_rt_runtime_read(struct cgroup *cgrp, struct cftype *cft)
Peter Zijlstra's avatar
Peter Zijlstra committed
7921
{
7922
	return sched_group_rt_runtime(cgroup_tg(cgrp));
Peter Zijlstra's avatar
Peter Zijlstra committed
7923
}
7924 7925 7926 7927 7928 7929 7930 7931 7932 7933 7934

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));
}
7935
#endif /* CONFIG_RT_GROUP_SCHED */
Peter Zijlstra's avatar
Peter Zijlstra committed
7936

7937
static struct cftype cpu_files[] = {
7938
#ifdef CONFIG_FAIR_GROUP_SCHED
7939 7940
	{
		.name = "shares",
7941 7942
		.read_u64 = cpu_shares_read_u64,
		.write_u64 = cpu_shares_write_u64,
7943
	},
7944
#endif
7945 7946 7947 7948 7949 7950 7951 7952 7953 7954 7955
#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,
	},
7956 7957 7958 7959
	{
		.name = "stat",
		.read_map = cpu_stats_show,
	},
7960
#endif
7961
#ifdef CONFIG_RT_GROUP_SCHED
Peter Zijlstra's avatar
Peter Zijlstra committed
7962
	{
Peter Zijlstra's avatar
Peter Zijlstra committed
7963
		.name = "rt_runtime_us",
7964 7965
		.read_s64 = cpu_rt_runtime_read,
		.write_s64 = cpu_rt_runtime_write,
Peter Zijlstra's avatar
Peter Zijlstra committed
7966
	},
7967 7968
	{
		.name = "rt_period_us",
7969 7970
		.read_u64 = cpu_rt_period_read_uint,
		.write_u64 = cpu_rt_period_write_uint,
7971
	},
7972
#endif
7973 7974 7975 7976
};

static int cpu_cgroup_populate(struct cgroup_subsys *ss, struct cgroup *cont)
{
7977
	return cgroup_add_files(cont, ss, cpu_files, ARRAY_SIZE(cpu_files));
7978 7979 7980
}

struct cgroup_subsys cpu_cgroup_subsys = {
7981 7982 7983
	.name		= "cpu",
	.create		= cpu_cgroup_create,
	.destroy	= cpu_cgroup_destroy,
7984 7985
	.can_attach	= cpu_cgroup_can_attach,
	.attach		= cpu_cgroup_attach,
7986
	.exit		= cpu_cgroup_exit,
7987 7988
	.populate	= cpu_cgroup_populate,
	.subsys_id	= cpu_cgroup_subsys_id,
7989 7990 7991
	.early_init	= 1,
};

7992
#endif	/* CONFIG_CGROUP_SCHED */
7993 7994 7995 7996 7997 7998 7999 8000 8001 8002 8003

#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 */
8004
static struct cgroup_subsys_state *cpuacct_create(struct cgroup *cgrp)
8005
{
8006
	struct cpuacct *ca;
8007

8008 8009 8010 8011
	if (!cgrp->parent)
		return &root_cpuacct.css;

	ca = kzalloc(sizeof(*ca), GFP_KERNEL);
8012
	if (!ca)
8013
		goto out;
8014 8015

	ca->cpuusage = alloc_percpu(u64);
8016 8017 8018
	if (!ca->cpuusage)
		goto out_free_ca;

8019 8020 8021
	ca->cpustat = alloc_percpu(struct kernel_cpustat);
	if (!ca->cpustat)
		goto out_free_cpuusage;
8022

8023
	return &ca->css;
8024

8025
out_free_cpuusage:
8026 8027 8028 8029 8030
	free_percpu(ca->cpuusage);
out_free_ca:
	kfree(ca);
out:
	return ERR_PTR(-ENOMEM);
8031 8032 8033
}

/* destroy an existing cpu accounting group */
8034
static void cpuacct_destroy(struct cgroup *cgrp)
8035
{
8036
	struct cpuacct *ca = cgroup_ca(cgrp);
8037

8038
	free_percpu(ca->cpustat);
8039 8040 8041 8042
	free_percpu(ca->cpuusage);
	kfree(ca);
}

8043 8044
static u64 cpuacct_cpuusage_read(struct cpuacct *ca, int cpu)
{
8045
	u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
8046 8047 8048 8049 8050 8051
	u64 data;

#ifndef CONFIG_64BIT
	/*
	 * Take rq->lock to make 64-bit read safe on 32-bit platforms.
	 */
8052
	raw_spin_lock_irq(&cpu_rq(cpu)->lock);
8053
	data = *cpuusage;
8054
	raw_spin_unlock_irq(&cpu_rq(cpu)->lock);
8055 8056 8057 8058 8059 8060 8061 8062 8063
#else
	data = *cpuusage;
#endif

	return data;
}

static void cpuacct_cpuusage_write(struct cpuacct *ca, int cpu, u64 val)
{
8064
	u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
8065 8066 8067 8068 8069

#ifndef CONFIG_64BIT
	/*
	 * Take rq->lock to make 64-bit write safe on 32-bit platforms.
	 */
8070
	raw_spin_lock_irq(&cpu_rq(cpu)->lock);
8071
	*cpuusage = val;
8072
	raw_spin_unlock_irq(&cpu_rq(cpu)->lock);
8073 8074 8075 8076 8077
#else
	*cpuusage = val;
#endif
}

8078
/* return total cpu usage (in nanoseconds) of a group */
8079
static u64 cpuusage_read(struct cgroup *cgrp, struct cftype *cft)
8080
{
8081
	struct cpuacct *ca = cgroup_ca(cgrp);
8082 8083 8084
	u64 totalcpuusage = 0;
	int i;

8085 8086
	for_each_present_cpu(i)
		totalcpuusage += cpuacct_cpuusage_read(ca, i);
8087 8088 8089 8090

	return totalcpuusage;
}

8091 8092 8093 8094 8095 8096 8097 8098 8099 8100 8101 8102
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;
	}

8103 8104
	for_each_present_cpu(i)
		cpuacct_cpuusage_write(ca, i, 0);
8105 8106 8107 8108 8109

out:
	return err;
}

8110 8111 8112 8113 8114 8115 8116 8117 8118 8119 8120 8121 8122 8123 8124
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;
}

8125 8126 8127 8128 8129 8130
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,
8131
			      struct cgroup_map_cb *cb)
8132 8133
{
	struct cpuacct *ca = cgroup_ca(cgrp);
8134 8135
	int cpu;
	s64 val = 0;
8136

8137 8138 8139 8140
	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];
8141
	}
8142 8143
	val = cputime64_to_clock_t(val);
	cb->fill(cb, cpuacct_stat_desc[CPUACCT_STAT_USER], val);
8144

8145 8146 8147 8148 8149 8150
	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];
8151
	}
8152 8153 8154 8155

	val = cputime64_to_clock_t(val);
	cb->fill(cb, cpuacct_stat_desc[CPUACCT_STAT_SYSTEM], val);

8156 8157 8158
	return 0;
}

8159 8160 8161
static struct cftype files[] = {
	{
		.name = "usage",
8162 8163
		.read_u64 = cpuusage_read,
		.write_u64 = cpuusage_write,
8164
	},
8165 8166 8167 8168
	{
		.name = "usage_percpu",
		.read_seq_string = cpuacct_percpu_seq_read,
	},
8169 8170 8171 8172
	{
		.name = "stat",
		.read_map = cpuacct_stats_show,
	},
8173 8174
};

8175
static int cpuacct_populate(struct cgroup_subsys *ss, struct cgroup *cgrp)
8176
{
8177
	return cgroup_add_files(cgrp, ss, files, ARRAY_SIZE(files));
8178 8179 8180 8181 8182 8183 8184
}

/*
 * charge this task's execution time to its accounting group.
 *
 * called with rq->lock held.
 */
8185
void cpuacct_charge(struct task_struct *tsk, u64 cputime)
8186 8187
{
	struct cpuacct *ca;
8188
	int cpu;
8189

Li Zefan's avatar
Li Zefan committed
8190
	if (unlikely(!cpuacct_subsys.active))
8191 8192
		return;

8193
	cpu = task_cpu(tsk);
8194 8195 8196

	rcu_read_lock();

8197 8198
	ca = task_ca(tsk);

8199
	for (; ca; ca = parent_ca(ca)) {
8200
		u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
8201 8202
		*cpuusage += cputime;
	}
8203 8204

	rcu_read_unlock();
8205 8206 8207 8208 8209 8210 8211 8212 8213 8214
}

struct cgroup_subsys cpuacct_subsys = {
	.name = "cpuacct",
	.create = cpuacct_create,
	.destroy = cpuacct_destroy,
	.populate = cpuacct_populate,
	.subsys_id = cpuacct_subsys_id,
};
#endif	/* CONFIG_CGROUP_CPUACCT */